Netdev Archive on lore.kernel.org
help / color / mirror / Atom feed
* NET: r8168/r8169 identifying fix
@ 2020-04-13 10:30 Lauri Jakku
  2020-04-13 10:58 ` Leon Romanovsky
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 10:30 UTC (permalink / raw)
  To: netdev, Heiner Kallweit, nic_swsd

[-- Attachment #1: Type: text/plain, Size: 6110 bytes --]

From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
 drivers/net/phy/mdio_bus.c                | 11 +++-
 2 files changed, 72 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..1ea6f121b561 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	new_bus->write = r8169_mdio_write_reg;
 
 	ret = mdiobus_register(new_bus);
+	dev_info(&pdev->dev,
+		 "mdiobus_register: %s, %d\n",
+		 new_bus->id, ret);
 	if (ret)
 		return ret;
 
 	tp->phydev = mdiobus_get_phy(new_bus, 0);
+
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id      = tp->phydev->phy_id;
+			phydrv_id      = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked    = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+
+			dev_debug(&pdev->dev,
+				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
+				new_bus->id, phydev_id, phydev_masked,
+				phydrv_id, drv_masked);
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match    = phydev_masked & drv_masked;
+			drv_match    = drv_match == drv_masked;
+
+			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
+				  new_bus->id, phydev_match, drv_match);
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
@@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	u64_stats_init(&tp->rx_stats.syncp);
 	u64_stats_init(&tp->tx_stats.syncp);
 
+	dev_dbg(&pdev->dev, "init: MAC\n");
 	rtl_init_mac_address(tp);
 
 	dev->ethtool_ops = &rtl8169_ethtool_ops;
@@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	dev->hw_features |= NETIF_F_RXFCS;
 
 	jumbo_max = rtl_jumbo_max(tp);
+	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
 	if (jumbo_max)
 		dev->max_mtu = jumbo_max;
 
+	dev_dbg(&pdev->dev, "init: irq mask\n");
 	rtl_set_irq_mask(tp);
 
 	tp->fw_name = rtl_chip_infos[chipset].fw_name;
+	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
 
 	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
 					    &tp->counters_phys_addr,
@@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (!tp->counters)
 		return -ENOMEM;
 
+	dev_dbg(&pdev->dev, "init: set driver data\n");
 	pci_set_drvdata(pdev, dev);
 
+	dev_dbg(&pdev->dev, "init: register mdio\n");
 	rc = r8169_mdio_register(tp);
+	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
 	if (rc)
 		return rc;
 
 	/* chip gets powered up in rtl_open() */
+	dev_dbg(&pdev->dev, "init: pll pwr down\n");
 	rtl_pll_power_down(tp);
 
 	rc = register_netdev(dev);
+	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
 	if (rc)
 		goto err_mdio_unregister;
 
@@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (pci_dev_run_wake(pdev))
 		pm_runtime_put_sync(&pdev->dev);
 
+	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
+
 	return 0;
 
 err_mdio_unregister:
diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
index 522760c8bca6..719ea48164f6 100644
--- a/drivers/net/phy/mdio_bus.c
+++ b/drivers/net/phy/mdio_bus.c
@@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
 struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 {
 	struct mdio_device *mdiodev = bus->mdio_map[addr];
+	struct phy_device *rv = NULL;
+
+	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
 
 	if (!mdiodev)
 		return NULL;
@@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
 		return NULL;
 
-	return container_of(mdiodev, struct phy_device, mdio);
+	rv = container_of(mdiodev, struct phy_device, mdio);
+	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
+		 bus->id, addr, mdiodev, rv);
+	return rv;
 }
 EXPORT_SYMBOL(mdiobus_get_phy);
 
@@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
 	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
 
 	bus->state = MDIOBUS_REGISTERED;
-	pr_info("%s: probed\n", bus->name);
+	pr_info("%s: probed (mdiobus_register)\n", bus->name);
 	return 0;
 
 error:
+	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
 	while (--i >= 0) {
 		mdiodev = bus->mdio_map[i];
 		if (!mdiodev)
-- 
2.26.0



[-- Attachment #2: NET-r8169-module-enchansments.patch --]
[-- Type: text/x-patch, Size: 6109 bytes --]

From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
 drivers/net/phy/mdio_bus.c                | 11 +++-
 2 files changed, 72 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..1ea6f121b561 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	new_bus->write = r8169_mdio_write_reg;
 
 	ret = mdiobus_register(new_bus);
+	dev_info(&pdev->dev,
+		 "mdiobus_register: %s, %d\n",
+		 new_bus->id, ret);
 	if (ret)
 		return ret;
 
 	tp->phydev = mdiobus_get_phy(new_bus, 0);
+
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id      = tp->phydev->phy_id;
+			phydrv_id      = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked    = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+
+			dev_debug(&pdev->dev,
+				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
+				new_bus->id, phydev_id, phydev_masked,
+				phydrv_id, drv_masked);
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match    = phydev_masked & drv_masked;
+			drv_match    = drv_match == drv_masked;
+
+			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
+				  new_bus->id, phydev_match, drv_match);
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
@@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	u64_stats_init(&tp->rx_stats.syncp);
 	u64_stats_init(&tp->tx_stats.syncp);
 
+	dev_dbg(&pdev->dev, "init: MAC\n");
 	rtl_init_mac_address(tp);
 
 	dev->ethtool_ops = &rtl8169_ethtool_ops;
@@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	dev->hw_features |= NETIF_F_RXFCS;
 
 	jumbo_max = rtl_jumbo_max(tp);
+	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
 	if (jumbo_max)
 		dev->max_mtu = jumbo_max;
 
+	dev_dbg(&pdev->dev, "init: irq mask\n");
 	rtl_set_irq_mask(tp);
 
 	tp->fw_name = rtl_chip_infos[chipset].fw_name;
+	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
 
 	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
 					    &tp->counters_phys_addr,
@@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (!tp->counters)
 		return -ENOMEM;
 
+	dev_dbg(&pdev->dev, "init: set driver data\n");
 	pci_set_drvdata(pdev, dev);
 
+	dev_dbg(&pdev->dev, "init: register mdio\n");
 	rc = r8169_mdio_register(tp);
+	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
 	if (rc)
 		return rc;
 
 	/* chip gets powered up in rtl_open() */
+	dev_dbg(&pdev->dev, "init: pll pwr down\n");
 	rtl_pll_power_down(tp);
 
 	rc = register_netdev(dev);
+	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
 	if (rc)
 		goto err_mdio_unregister;
 
@@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (pci_dev_run_wake(pdev))
 		pm_runtime_put_sync(&pdev->dev);
 
+	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
+
 	return 0;
 
 err_mdio_unregister:
diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
index 522760c8bca6..719ea48164f6 100644
--- a/drivers/net/phy/mdio_bus.c
+++ b/drivers/net/phy/mdio_bus.c
@@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
 struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 {
 	struct mdio_device *mdiodev = bus->mdio_map[addr];
+	struct phy_device *rv = NULL;
+
+	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
 
 	if (!mdiodev)
 		return NULL;
@@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
 		return NULL;
 
-	return container_of(mdiodev, struct phy_device, mdio);
+	rv = container_of(mdiodev, struct phy_device, mdio);
+	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
+		 bus->id, addr, mdiodev, rv);
+	return rv;
 }
 EXPORT_SYMBOL(mdiobus_get_phy);
 
@@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
 	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
 
 	bus->state = MDIOBUS_REGISTERED;
-	pr_info("%s: probed\n", bus->name);
+	pr_info("%s: probed (mdiobus_register)\n", bus->name);
 	return 0;
 
 error:
+	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
 	while (--i >= 0) {
 		mdiodev = bus->mdio_map[i];
 		if (!mdiodev)
-- 
2.26.0


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 10:30 NET: r8168/r8169 identifying fix Lauri Jakku
@ 2020-04-13 10:58 ` Leon Romanovsky
  2020-04-13 11:02   ` Lauri Jakku
  2020-04-13 11:06   ` Lauri Jakku
  0 siblings, 2 replies; 57+ messages in thread
From: Leon Romanovsky @ 2020-04-13 10:58 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: netdev, Heiner Kallweit, nic_swsd

On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
> From: Lauri Jakku <lja@iki.fi>
> Date: Mon, 13 Apr 2020 13:18:35 +0300
> Subject: [PATCH] NET: r8168/r8169 identifying fix
>
> The driver installation determination made properly by
> checking PHY vs DRIVER id's.
> ---
>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>  drivers/net/phy/mdio_bus.c                | 11 +++-
>  2 files changed, 72 insertions(+), 9 deletions(-)

I would say that most of the code is debug prints.

>
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..1ea6f121b561 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  {
>  	struct pci_dev *pdev = tp->pci_dev;
>  	struct mii_bus *new_bus;
> +	u32 phydev_id = 0;
> +	u32 phydrv_id = 0;
> +	u32 phydrv_id_mask = 0;
>  	int ret;
>
>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  	new_bus->write = r8169_mdio_write_reg;
>
>  	ret = mdiobus_register(new_bus);
> +	dev_info(&pdev->dev,
> +		 "mdiobus_register: %s, %d\n",
> +		 new_bus->id, ret);
>  	if (ret)
>  		return ret;
>
>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
> +
>  	if (!tp->phydev) {
>  		mdiobus_unregister(new_bus);
>  		return -ENODEV;
> -	} else if (!tp->phydev->drv) {
> -		/* Most chip versions fail with the genphy driver.
> -		 * Therefore ensure that the dedicated PHY driver is loaded.
> -		 */
> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> -		mdiobus_unregister(new_bus);
> -		return -EUNATCH;
> +	} else {
> +		/* tp -> phydev ok */
> +		int everything_OK = 0;
> +
> +		/* Check driver id versus phy */
> +
> +		if (tp->phydev->drv) {
> +			u32 phydev_masked = 0xBEEFDEAD;
> +			u32 drv_masked = ~0;
> +			u32 phydev_match = ~0;
> +			u32 drv_match = 0xDEADBEEF;
> +
> +			phydev_id      = tp->phydev->phy_id;
> +			phydrv_id      = tp->phydev->drv->phy_id;
> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
> +
> +			drv_masked    = phydrv_id & phydrv_id_mask;
> +			phydev_masked = phydev_id & phydrv_id_mask;

Please don't do vertical space alignment, every change later near such
lines will be whitespace disaster.

Thanks

> +
> +			dev_debug(&pdev->dev,
> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
> +				new_bus->id, phydev_id, phydev_masked,
> +				phydrv_id, drv_masked);
> +
> +			phydev_match = phydev_masked & drv_masked;
> +			phydev_match = phydev_match == phydev_masked;
> +
> +			drv_match    = phydev_masked & drv_masked;
> +			drv_match    = drv_match == drv_masked;
> +
> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
> +				  new_bus->id, phydev_match, drv_match);
> +
> +			everything_OK = (phydev_match == drv_match);
> +		}
> +
> +		if (!everything_OK) {
> +			/* Most chip versions fail with the genphy driver.
> +			 * Therefore ensure that the dedicated PHY driver
> +			 * is loaded.
> +			 */
> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> +			mdiobus_unregister(new_bus);
> +			return -EUNATCH;
> +		}
>  	}
>
>  	/* PHY will be woken up in rtl_open() */
> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	u64_stats_init(&tp->rx_stats.syncp);
>  	u64_stats_init(&tp->tx_stats.syncp);
>
> +	dev_dbg(&pdev->dev, "init: MAC\n");
>  	rtl_init_mac_address(tp);
>
>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	dev->hw_features |= NETIF_F_RXFCS;
>
>  	jumbo_max = rtl_jumbo_max(tp);
> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>  	if (jumbo_max)
>  		dev->max_mtu = jumbo_max;
>
> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>  	rtl_set_irq_mask(tp);
>
>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>
>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>  					    &tp->counters_phys_addr,
> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (!tp->counters)
>  		return -ENOMEM;
>
> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>  	pci_set_drvdata(pdev, dev);
>
> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>  	rc = r8169_mdio_register(tp);
> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>  	if (rc)
>  		return rc;
>
>  	/* chip gets powered up in rtl_open() */
> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>  	rtl_pll_power_down(tp);
>
>  	rc = register_netdev(dev);
> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>  	if (rc)
>  		goto err_mdio_unregister;
>
> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (pci_dev_run_wake(pdev))
>  		pm_runtime_put_sync(&pdev->dev);
>
> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
> +
>  	return 0;
>
>  err_mdio_unregister:
> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
> index 522760c8bca6..719ea48164f6 100644
> --- a/drivers/net/phy/mdio_bus.c
> +++ b/drivers/net/phy/mdio_bus.c
> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  {
>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
> +	struct phy_device *rv = NULL;
> +
> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>
>  	if (!mdiodev)
>  		return NULL;
> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>  		return NULL;
>
> -	return container_of(mdiodev, struct phy_device, mdio);
> +	rv = container_of(mdiodev, struct phy_device, mdio);
> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
> +		 bus->id, addr, mdiodev, rv);
> +	return rv;
>  }
>  EXPORT_SYMBOL(mdiobus_get_phy);
>
> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>
>  	bus->state = MDIOBUS_REGISTERED;
> -	pr_info("%s: probed\n", bus->name);
> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>  	return 0;
>
>  error:
> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>  	while (--i >= 0) {
>  		mdiodev = bus->mdio_map[i];
>  		if (!mdiodev)
> --
> 2.26.0
>
>

> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
> From: Lauri Jakku <lja@iki.fi>
> Date: Mon, 13 Apr 2020 13:18:35 +0300
> Subject: [PATCH] NET: r8168/r8169 identifying fix
>
> The driver installation determination made properly by
> checking PHY vs DRIVER id's.
> ---
>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>  drivers/net/phy/mdio_bus.c                | 11 +++-
>  2 files changed, 72 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..1ea6f121b561 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  {
>  	struct pci_dev *pdev = tp->pci_dev;
>  	struct mii_bus *new_bus;
> +	u32 phydev_id = 0;
> +	u32 phydrv_id = 0;
> +	u32 phydrv_id_mask = 0;
>  	int ret;
>
>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  	new_bus->write = r8169_mdio_write_reg;
>
>  	ret = mdiobus_register(new_bus);
> +	dev_info(&pdev->dev,
> +		 "mdiobus_register: %s, %d\n",
> +		 new_bus->id, ret);
>  	if (ret)
>  		return ret;
>
>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
> +
>  	if (!tp->phydev) {
>  		mdiobus_unregister(new_bus);
>  		return -ENODEV;
> -	} else if (!tp->phydev->drv) {
> -		/* Most chip versions fail with the genphy driver.
> -		 * Therefore ensure that the dedicated PHY driver is loaded.
> -		 */
> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> -		mdiobus_unregister(new_bus);
> -		return -EUNATCH;
> +	} else {
> +		/* tp -> phydev ok */
> +		int everything_OK = 0;
> +
> +		/* Check driver id versus phy */
> +
> +		if (tp->phydev->drv) {
> +			u32 phydev_masked = 0xBEEFDEAD;
> +			u32 drv_masked = ~0;
> +			u32 phydev_match = ~0;
> +			u32 drv_match = 0xDEADBEEF;
> +
> +			phydev_id      = tp->phydev->phy_id;
> +			phydrv_id      = tp->phydev->drv->phy_id;
> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
> +
> +			drv_masked    = phydrv_id & phydrv_id_mask;
> +			phydev_masked = phydev_id & phydrv_id_mask;
> +
> +			dev_debug(&pdev->dev,
> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
> +				new_bus->id, phydev_id, phydev_masked,
> +				phydrv_id, drv_masked);
> +
> +			phydev_match = phydev_masked & drv_masked;
> +			phydev_match = phydev_match == phydev_masked;
> +
> +			drv_match    = phydev_masked & drv_masked;
> +			drv_match    = drv_match == drv_masked;
> +
> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
> +				  new_bus->id, phydev_match, drv_match);
> +
> +			everything_OK = (phydev_match == drv_match);
> +		}
> +
> +		if (!everything_OK) {
> +			/* Most chip versions fail with the genphy driver.
> +			 * Therefore ensure that the dedicated PHY driver
> +			 * is loaded.
> +			 */
> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> +			mdiobus_unregister(new_bus);
> +			return -EUNATCH;
> +		}
>  	}
>
>  	/* PHY will be woken up in rtl_open() */
> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	u64_stats_init(&tp->rx_stats.syncp);
>  	u64_stats_init(&tp->tx_stats.syncp);
>
> +	dev_dbg(&pdev->dev, "init: MAC\n");
>  	rtl_init_mac_address(tp);
>
>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	dev->hw_features |= NETIF_F_RXFCS;
>
>  	jumbo_max = rtl_jumbo_max(tp);
> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>  	if (jumbo_max)
>  		dev->max_mtu = jumbo_max;
>
> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>  	rtl_set_irq_mask(tp);
>
>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>
>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>  					    &tp->counters_phys_addr,
> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (!tp->counters)
>  		return -ENOMEM;
>
> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>  	pci_set_drvdata(pdev, dev);
>
> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>  	rc = r8169_mdio_register(tp);
> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>  	if (rc)
>  		return rc;
>
>  	/* chip gets powered up in rtl_open() */
> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>  	rtl_pll_power_down(tp);
>
>  	rc = register_netdev(dev);
> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>  	if (rc)
>  		goto err_mdio_unregister;
>
> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (pci_dev_run_wake(pdev))
>  		pm_runtime_put_sync(&pdev->dev);
>
> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
> +
>  	return 0;
>
>  err_mdio_unregister:
> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
> index 522760c8bca6..719ea48164f6 100644
> --- a/drivers/net/phy/mdio_bus.c
> +++ b/drivers/net/phy/mdio_bus.c
> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  {
>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
> +	struct phy_device *rv = NULL;
> +
> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>
>  	if (!mdiodev)
>  		return NULL;
> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>  		return NULL;
>
> -	return container_of(mdiodev, struct phy_device, mdio);
> +	rv = container_of(mdiodev, struct phy_device, mdio);
> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
> +		 bus->id, addr, mdiodev, rv);
> +	return rv;
>  }
>  EXPORT_SYMBOL(mdiobus_get_phy);
>
> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>
>  	bus->state = MDIOBUS_REGISTERED;
> -	pr_info("%s: probed\n", bus->name);
> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>  	return 0;
>
>  error:
> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>  	while (--i >= 0) {
>  		mdiodev = bus->mdio_map[i];
>  		if (!mdiodev)
> --
> 2.26.0
>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 10:58 ` Leon Romanovsky
@ 2020-04-13 11:02   ` Lauri Jakku
  2020-04-13 11:34     ` Leon Romanovsky
  2020-04-13 11:06   ` Lauri Jakku
  1 sibling, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 11:02 UTC (permalink / raw)
  To: Leon Romanovsky; +Cc: netdev, Heiner Kallweit, nic_swsd

Hi,

Comments inline.

On 2020-04-13 13:58, Leon Romanovsky wrote:
> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>  2 files changed, 72 insertions(+), 9 deletions(-)
> 
> I would say that most of the code is debug prints.
> 

I tought that they are helpful to keep, they are using the debug calls, so
they are not visible if user does not like those.

>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..1ea6f121b561 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	new_bus->write = r8169_mdio_write_reg;
>>
>>  	ret = mdiobus_register(new_bus);
>> +	dev_info(&pdev->dev,
>> +		 "mdiobus_register: %s, %d\n",
>> +		 new_bus->id, ret);
>>  	if (ret)
>>  		return ret;
>>
>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>> +
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id      = tp->phydev->phy_id;
>> +			phydrv_id      = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked    = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
> 
> Please don't do vertical space alignment, every change later near such
> lines will be whitespace disaster.
>

Ok, i'll reformat this.

 
> Thanks
> 
>> +
>> +			dev_debug(&pdev->dev,
>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>> +				new_bus->id, phydev_id, phydev_masked,
>> +				phydrv_id, drv_masked);
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match    = phydev_masked & drv_masked;
>> +			drv_match    = drv_match == drv_masked;
>> +
>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>> +				  new_bus->id, phydev_match, drv_match);
>> +
>> +			everything_OK = (phydev_match == drv_match);
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>
>>  	/* PHY will be woken up in rtl_open() */
>> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	u64_stats_init(&tp->rx_stats.syncp);
>>  	u64_stats_init(&tp->tx_stats.syncp);
>>
>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>  	rtl_init_mac_address(tp);
>>
>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	dev->hw_features |= NETIF_F_RXFCS;
>>
>>  	jumbo_max = rtl_jumbo_max(tp);
>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>  	if (jumbo_max)
>>  		dev->max_mtu = jumbo_max;
>>
>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>  	rtl_set_irq_mask(tp);
>>
>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>
>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>  					    &tp->counters_phys_addr,
>> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (!tp->counters)
>>  		return -ENOMEM;
>>
>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>  	pci_set_drvdata(pdev, dev);
>>
>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>  	rc = r8169_mdio_register(tp);
>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>  	if (rc)
>>  		return rc;
>>
>>  	/* chip gets powered up in rtl_open() */
>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>  	rtl_pll_power_down(tp);
>>
>>  	rc = register_netdev(dev);
>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>  	if (rc)
>>  		goto err_mdio_unregister;
>>
>> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (pci_dev_run_wake(pdev))
>>  		pm_runtime_put_sync(&pdev->dev);
>>
>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>> +
>>  	return 0;
>>
>>  err_mdio_unregister:
>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>> index 522760c8bca6..719ea48164f6 100644
>> --- a/drivers/net/phy/mdio_bus.c
>> +++ b/drivers/net/phy/mdio_bus.c
>> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  {
>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>> +	struct phy_device *rv = NULL;
>> +
>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>
>>  	if (!mdiodev)
>>  		return NULL;
>> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>  		return NULL;
>>
>> -	return container_of(mdiodev, struct phy_device, mdio);
>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>> +		 bus->id, addr, mdiodev, rv);
>> +	return rv;
>>  }
>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>
>> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>
>>  	bus->state = MDIOBUS_REGISTERED;
>> -	pr_info("%s: probed\n", bus->name);
>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>  	return 0;
>>
>>  error:
>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>  	while (--i >= 0) {
>>  		mdiodev = bus->mdio_map[i];
>>  		if (!mdiodev)
>> --
>> 2.26.0
>>
>>
> 
>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..1ea6f121b561 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	new_bus->write = r8169_mdio_write_reg;
>>
>>  	ret = mdiobus_register(new_bus);
>> +	dev_info(&pdev->dev,
>> +		 "mdiobus_register: %s, %d\n",
>> +		 new_bus->id, ret);
>>  	if (ret)
>>  		return ret;
>>
>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>> +
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id      = tp->phydev->phy_id;
>> +			phydrv_id      = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked    = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
>> +
>> +			dev_debug(&pdev->dev,
>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>> +				new_bus->id, phydev_id, phydev_masked,
>> +				phydrv_id, drv_masked);
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match    = phydev_masked & drv_masked;
>> +			drv_match    = drv_match == drv_masked;
>> +
>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>> +				  new_bus->id, phydev_match, drv_match);
>> +
>> +			everything_OK = (phydev_match == drv_match);
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>
>>  	/* PHY will be woken up in rtl_open() */
>> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	u64_stats_init(&tp->rx_stats.syncp);
>>  	u64_stats_init(&tp->tx_stats.syncp);
>>
>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>  	rtl_init_mac_address(tp);
>>
>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	dev->hw_features |= NETIF_F_RXFCS;
>>
>>  	jumbo_max = rtl_jumbo_max(tp);
>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>  	if (jumbo_max)
>>  		dev->max_mtu = jumbo_max;
>>
>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>  	rtl_set_irq_mask(tp);
>>
>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>
>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>  					    &tp->counters_phys_addr,
>> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (!tp->counters)
>>  		return -ENOMEM;
>>
>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>  	pci_set_drvdata(pdev, dev);
>>
>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>  	rc = r8169_mdio_register(tp);
>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>  	if (rc)
>>  		return rc;
>>
>>  	/* chip gets powered up in rtl_open() */
>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>  	rtl_pll_power_down(tp);
>>
>>  	rc = register_netdev(dev);
>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>  	if (rc)
>>  		goto err_mdio_unregister;
>>
>> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (pci_dev_run_wake(pdev))
>>  		pm_runtime_put_sync(&pdev->dev);
>>
>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>> +
>>  	return 0;
>>
>>  err_mdio_unregister:
>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>> index 522760c8bca6..719ea48164f6 100644
>> --- a/drivers/net/phy/mdio_bus.c
>> +++ b/drivers/net/phy/mdio_bus.c
>> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  {
>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>> +	struct phy_device *rv = NULL;
>> +
>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>
>>  	if (!mdiodev)
>>  		return NULL;
>> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>  		return NULL;
>>
>> -	return container_of(mdiodev, struct phy_device, mdio);
>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>> +		 bus->id, addr, mdiodev, rv);
>> +	return rv;
>>  }
>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>
>> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>
>>  	bus->state = MDIOBUS_REGISTERED;
>> -	pr_info("%s: probed\n", bus->name);
>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>  	return 0;
>>
>>  error:
>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>  	while (--i >= 0) {
>>  		mdiodev = bus->mdio_map[i];
>>  		if (!mdiodev)
>> --
>> 2.26.0
>>
> 

Br,
Lauri Jakku

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 10:58 ` Leon Romanovsky
  2020-04-13 11:02   ` Lauri Jakku
@ 2020-04-13 11:06   ` Lauri Jakku
  2020-04-13 11:28     ` Heiner Kallweit
  1 sibling, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 11:06 UTC (permalink / raw)
  To: Leon Romanovsky; +Cc: netdev, Heiner Kallweit, nic_swsd

[-- Attachment #1: Type: text/plain, Size: 19722 bytes --]

Hi,

Modified as suggested:

From f4fba16025260bf08f0df867c3de51803ddb78ef Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
 drivers/net/phy/mdio_bus.c                | 11 +++-
 2 files changed, 72 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..2384da7d2988 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	new_bus->write = r8169_mdio_write_reg;
 
 	ret = mdiobus_register(new_bus);
+	dev_info(&pdev->dev,
+		 "mdiobus_register: %s, %d\n",
+		 new_bus->id, ret);
 	if (ret)
 		return ret;
 
 	tp->phydev = mdiobus_get_phy(new_bus, 0);
+
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id = tp->phydev->phy_id;
+			phydrv_id = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+
+			dev_debug(&pdev->dev,
+				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
+				new_bus->id, phydev_id, phydev_masked,
+				phydrv_id, drv_masked);
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match = phydev_masked & drv_masked;
+			drv_match = drv_match == drv_masked;
+
+			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
+				  new_bus->id, phydev_match, drv_match);
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
@@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	u64_stats_init(&tp->rx_stats.syncp);
 	u64_stats_init(&tp->tx_stats.syncp);
 
+	dev_dbg(&pdev->dev, "init: MAC\n");
 	rtl_init_mac_address(tp);
 
 	dev->ethtool_ops = &rtl8169_ethtool_ops;
@@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	dev->hw_features |= NETIF_F_RXFCS;
 
 	jumbo_max = rtl_jumbo_max(tp);
+	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
 	if (jumbo_max)
 		dev->max_mtu = jumbo_max;
 
+	dev_dbg(&pdev->dev, "init: irq mask\n");
 	rtl_set_irq_mask(tp);
 
 	tp->fw_name = rtl_chip_infos[chipset].fw_name;
+	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
 
 	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
 					    &tp->counters_phys_addr,
@@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (!tp->counters)
 		return -ENOMEM;
 
+	dev_dbg(&pdev->dev, "init: set driver data\n");
 	pci_set_drvdata(pdev, dev);
 
+	dev_dbg(&pdev->dev, "init: register mdio\n");
 	rc = r8169_mdio_register(tp);
+	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
 	if (rc)
 		return rc;
 
 	/* chip gets powered up in rtl_open() */
+	dev_dbg(&pdev->dev, "init: pll pwr down\n");
 	rtl_pll_power_down(tp);
 
 	rc = register_netdev(dev);
+	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
 	if (rc)
 		goto err_mdio_unregister;
 
@@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (pci_dev_run_wake(pdev))
 		pm_runtime_put_sync(&pdev->dev);
 
+	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
+
 	return 0;
 
 err_mdio_unregister:
diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
index 522760c8bca6..719ea48164f6 100644
--- a/drivers/net/phy/mdio_bus.c
+++ b/drivers/net/phy/mdio_bus.c
@@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
 struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 {
 	struct mdio_device *mdiodev = bus->mdio_map[addr];
+	struct phy_device *rv = NULL;
+
+	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
 
 	if (!mdiodev)
 		return NULL;
@@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
 		return NULL;
 
-	return container_of(mdiodev, struct phy_device, mdio);
+	rv = container_of(mdiodev, struct phy_device, mdio);
+	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
+		 bus->id, addr, mdiodev, rv);
+	return rv;
 }
 EXPORT_SYMBOL(mdiobus_get_phy);
 
@@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
 	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
 
 	bus->state = MDIOBUS_REGISTERED;
-	pr_info("%s: probed\n", bus->name);
+	pr_info("%s: probed (mdiobus_register)\n", bus->name);
 	return 0;
 
 error:
+	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
 	while (--i >= 0) {
 		mdiodev = bus->mdio_map[i];
 		if (!mdiodev)
-- 
2.26.0



On 2020-04-13 13:58, Leon Romanovsky wrote:
> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>  2 files changed, 72 insertions(+), 9 deletions(-)
> 
> I would say that most of the code is debug prints.
> 
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..1ea6f121b561 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	new_bus->write = r8169_mdio_write_reg;
>>
>>  	ret = mdiobus_register(new_bus);
>> +	dev_info(&pdev->dev,
>> +		 "mdiobus_register: %s, %d\n",
>> +		 new_bus->id, ret);
>>  	if (ret)
>>  		return ret;
>>
>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>> +
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id      = tp->phydev->phy_id;
>> +			phydrv_id      = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked    = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
> 
> Please don't do vertical space alignment, every change later near such
> lines will be whitespace disaster.
> 
> Thanks
> 
>> +
>> +			dev_debug(&pdev->dev,
>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>> +				new_bus->id, phydev_id, phydev_masked,
>> +				phydrv_id, drv_masked);
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match    = phydev_masked & drv_masked;
>> +			drv_match    = drv_match == drv_masked;
>> +
>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>> +				  new_bus->id, phydev_match, drv_match);
>> +
>> +			everything_OK = (phydev_match == drv_match);
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>
>>  	/* PHY will be woken up in rtl_open() */
>> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	u64_stats_init(&tp->rx_stats.syncp);
>>  	u64_stats_init(&tp->tx_stats.syncp);
>>
>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>  	rtl_init_mac_address(tp);
>>
>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	dev->hw_features |= NETIF_F_RXFCS;
>>
>>  	jumbo_max = rtl_jumbo_max(tp);
>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>  	if (jumbo_max)
>>  		dev->max_mtu = jumbo_max;
>>
>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>  	rtl_set_irq_mask(tp);
>>
>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>
>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>  					    &tp->counters_phys_addr,
>> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (!tp->counters)
>>  		return -ENOMEM;
>>
>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>  	pci_set_drvdata(pdev, dev);
>>
>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>  	rc = r8169_mdio_register(tp);
>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>  	if (rc)
>>  		return rc;
>>
>>  	/* chip gets powered up in rtl_open() */
>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>  	rtl_pll_power_down(tp);
>>
>>  	rc = register_netdev(dev);
>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>  	if (rc)
>>  		goto err_mdio_unregister;
>>
>> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (pci_dev_run_wake(pdev))
>>  		pm_runtime_put_sync(&pdev->dev);
>>
>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>> +
>>  	return 0;
>>
>>  err_mdio_unregister:
>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>> index 522760c8bca6..719ea48164f6 100644
>> --- a/drivers/net/phy/mdio_bus.c
>> +++ b/drivers/net/phy/mdio_bus.c
>> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  {
>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>> +	struct phy_device *rv = NULL;
>> +
>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>
>>  	if (!mdiodev)
>>  		return NULL;
>> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>  		return NULL;
>>
>> -	return container_of(mdiodev, struct phy_device, mdio);
>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>> +		 bus->id, addr, mdiodev, rv);
>> +	return rv;
>>  }
>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>
>> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>
>>  	bus->state = MDIOBUS_REGISTERED;
>> -	pr_info("%s: probed\n", bus->name);
>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>  	return 0;
>>
>>  error:
>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>  	while (--i >= 0) {
>>  		mdiodev = bus->mdio_map[i];
>>  		if (!mdiodev)
>> --
>> 2.26.0
>>
>>
> 
>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..1ea6f121b561 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	new_bus->write = r8169_mdio_write_reg;
>>
>>  	ret = mdiobus_register(new_bus);
>> +	dev_info(&pdev->dev,
>> +		 "mdiobus_register: %s, %d\n",
>> +		 new_bus->id, ret);
>>  	if (ret)
>>  		return ret;
>>
>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>> +
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id      = tp->phydev->phy_id;
>> +			phydrv_id      = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked    = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
>> +
>> +			dev_debug(&pdev->dev,
>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>> +				new_bus->id, phydev_id, phydev_masked,
>> +				phydrv_id, drv_masked);
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match    = phydev_masked & drv_masked;
>> +			drv_match    = drv_match == drv_masked;
>> +
>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>> +				  new_bus->id, phydev_match, drv_match);
>> +
>> +			everything_OK = (phydev_match == drv_match);
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>
>>  	/* PHY will be woken up in rtl_open() */
>> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	u64_stats_init(&tp->rx_stats.syncp);
>>  	u64_stats_init(&tp->tx_stats.syncp);
>>
>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>  	rtl_init_mac_address(tp);
>>
>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	dev->hw_features |= NETIF_F_RXFCS;
>>
>>  	jumbo_max = rtl_jumbo_max(tp);
>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>  	if (jumbo_max)
>>  		dev->max_mtu = jumbo_max;
>>
>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>  	rtl_set_irq_mask(tp);
>>
>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>
>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>  					    &tp->counters_phys_addr,
>> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (!tp->counters)
>>  		return -ENOMEM;
>>
>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>  	pci_set_drvdata(pdev, dev);
>>
>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>  	rc = r8169_mdio_register(tp);
>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>  	if (rc)
>>  		return rc;
>>
>>  	/* chip gets powered up in rtl_open() */
>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>  	rtl_pll_power_down(tp);
>>
>>  	rc = register_netdev(dev);
>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>  	if (rc)
>>  		goto err_mdio_unregister;
>>
>> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (pci_dev_run_wake(pdev))
>>  		pm_runtime_put_sync(&pdev->dev);
>>
>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>> +
>>  	return 0;
>>
>>  err_mdio_unregister:
>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>> index 522760c8bca6..719ea48164f6 100644
>> --- a/drivers/net/phy/mdio_bus.c
>> +++ b/drivers/net/phy/mdio_bus.c
>> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  {
>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>> +	struct phy_device *rv = NULL;
>> +
>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>
>>  	if (!mdiodev)
>>  		return NULL;
>> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>  		return NULL;
>>
>> -	return container_of(mdiodev, struct phy_device, mdio);
>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>> +		 bus->id, addr, mdiodev, rv);
>> +	return rv;
>>  }
>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>
>> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>
>>  	bus->state = MDIOBUS_REGISTERED;
>> -	pr_info("%s: probed\n", bus->name);
>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>  	return 0;
>>
>>  error:
>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>  	while (--i >= 0) {
>>  		mdiodev = bus->mdio_map[i];
>>  		if (!mdiodev)
>> --
>> 2.26.0
>>
> 

-- 
Br,
Lauri J.

[-- Attachment #2: NET-r8169-module-enchansments.patch --]
[-- Type: text/x-patch, Size: 6090 bytes --]

From f4fba16025260bf08f0df867c3de51803ddb78ef Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
 drivers/net/phy/mdio_bus.c                | 11 +++-
 2 files changed, 72 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..2384da7d2988 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	new_bus->write = r8169_mdio_write_reg;
 
 	ret = mdiobus_register(new_bus);
+	dev_info(&pdev->dev,
+		 "mdiobus_register: %s, %d\n",
+		 new_bus->id, ret);
 	if (ret)
 		return ret;
 
 	tp->phydev = mdiobus_get_phy(new_bus, 0);
+
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id = tp->phydev->phy_id;
+			phydrv_id = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+
+			dev_debug(&pdev->dev,
+				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
+				new_bus->id, phydev_id, phydev_masked,
+				phydrv_id, drv_masked);
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match = phydev_masked & drv_masked;
+			drv_match = drv_match == drv_masked;
+
+			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
+				  new_bus->id, phydev_match, drv_match);
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
@@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	u64_stats_init(&tp->rx_stats.syncp);
 	u64_stats_init(&tp->tx_stats.syncp);
 
+	dev_dbg(&pdev->dev, "init: MAC\n");
 	rtl_init_mac_address(tp);
 
 	dev->ethtool_ops = &rtl8169_ethtool_ops;
@@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	dev->hw_features |= NETIF_F_RXFCS;
 
 	jumbo_max = rtl_jumbo_max(tp);
+	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
 	if (jumbo_max)
 		dev->max_mtu = jumbo_max;
 
+	dev_dbg(&pdev->dev, "init: irq mask\n");
 	rtl_set_irq_mask(tp);
 
 	tp->fw_name = rtl_chip_infos[chipset].fw_name;
+	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
 
 	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
 					    &tp->counters_phys_addr,
@@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (!tp->counters)
 		return -ENOMEM;
 
+	dev_dbg(&pdev->dev, "init: set driver data\n");
 	pci_set_drvdata(pdev, dev);
 
+	dev_dbg(&pdev->dev, "init: register mdio\n");
 	rc = r8169_mdio_register(tp);
+	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
 	if (rc)
 		return rc;
 
 	/* chip gets powered up in rtl_open() */
+	dev_dbg(&pdev->dev, "init: pll pwr down\n");
 	rtl_pll_power_down(tp);
 
 	rc = register_netdev(dev);
+	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
 	if (rc)
 		goto err_mdio_unregister;
 
@@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (pci_dev_run_wake(pdev))
 		pm_runtime_put_sync(&pdev->dev);
 
+	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
+
 	return 0;
 
 err_mdio_unregister:
diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
index 522760c8bca6..719ea48164f6 100644
--- a/drivers/net/phy/mdio_bus.c
+++ b/drivers/net/phy/mdio_bus.c
@@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
 struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 {
 	struct mdio_device *mdiodev = bus->mdio_map[addr];
+	struct phy_device *rv = NULL;
+
+	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
 
 	if (!mdiodev)
 		return NULL;
@@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
 		return NULL;
 
-	return container_of(mdiodev, struct phy_device, mdio);
+	rv = container_of(mdiodev, struct phy_device, mdio);
+	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
+		 bus->id, addr, mdiodev, rv);
+	return rv;
 }
 EXPORT_SYMBOL(mdiobus_get_phy);
 
@@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
 	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
 
 	bus->state = MDIOBUS_REGISTERED;
-	pr_info("%s: probed\n", bus->name);
+	pr_info("%s: probed (mdiobus_register)\n", bus->name);
 	return 0;
 
 error:
+	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
 	while (--i >= 0) {
 		mdiodev = bus->mdio_map[i];
 		if (!mdiodev)
-- 
2.26.0


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 11:06   ` Lauri Jakku
@ 2020-04-13 11:28     ` Heiner Kallweit
  2020-04-13 11:40       ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-13 11:28 UTC (permalink / raw)
  To: Lauri Jakku, Leon Romanovsky; +Cc: netdev, nic_swsd

On 13.04.2020 13:06, Lauri Jakku wrote:
> Hi,
> 
> Modified as suggested:
> 
>>From f4fba16025260bf08f0df867c3de51803ddb78ef Mon Sep 17 00:00:00 2001
> From: Lauri Jakku <lja@iki.fi>
> Date: Mon, 13 Apr 2020 13:18:35 +0300
> Subject: [PATCH] NET: r8168/r8169 identifying fix
> 
> The driver installation determination made properly by
> checking PHY vs DRIVER id's.
> ---
>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>  drivers/net/phy/mdio_bus.c                | 11 +++-
>  2 files changed, 72 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..2384da7d2988 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  {
>  	struct pci_dev *pdev = tp->pci_dev;
>  	struct mii_bus *new_bus;
> +	u32 phydev_id = 0;
> +	u32 phydrv_id = 0;
> +	u32 phydrv_id_mask = 0;
>  	int ret;
>  
>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  	new_bus->write = r8169_mdio_write_reg;
>  
>  	ret = mdiobus_register(new_bus);
> +	dev_info(&pdev->dev,
> +		 "mdiobus_register: %s, %d\n",
> +		 new_bus->id, ret);
>  	if (ret)
>  		return ret;
>  
>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
> +
>  	if (!tp->phydev) {
>  		mdiobus_unregister(new_bus);
>  		return -ENODEV;
> -	} else if (!tp->phydev->drv) {
> -		/* Most chip versions fail with the genphy driver.
> -		 * Therefore ensure that the dedicated PHY driver is loaded.
> -		 */
> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> -		mdiobus_unregister(new_bus);
> -		return -EUNATCH;
> +	} else {
> +		/* tp -> phydev ok */
> +		int everything_OK = 0;
> +
> +		/* Check driver id versus phy */
> +
> +		if (tp->phydev->drv) {
> +			u32 phydev_masked = 0xBEEFDEAD;
> +			u32 drv_masked = ~0;
> +			u32 phydev_match = ~0;
> +			u32 drv_match = 0xDEADBEEF;
> +
> +			phydev_id = tp->phydev->phy_id;
> +			phydrv_id = tp->phydev->drv->phy_id;
> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
> +
> +			drv_masked = phydrv_id & phydrv_id_mask;
> +			phydev_masked = phydev_id & phydrv_id_mask;
> +
> +			dev_debug(&pdev->dev,
> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
> +				new_bus->id, phydev_id, phydev_masked,
> +				phydrv_id, drv_masked);
> +
> +			phydev_match = phydev_masked & drv_masked;
> +			phydev_match = phydev_match == phydev_masked;
> +
> +			drv_match = phydev_masked & drv_masked;
> +			drv_match = drv_match == drv_masked;
> +
> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
> +				  new_bus->id, phydev_match, drv_match);
> +
> +			everything_OK = (phydev_match == drv_match);

You're doing again what phy_bus_match() did. It's unclear what this should
be good for. Also it would be helpful if you could explain what the actual
issue is you're trying to fix, incl. the commit that caused the regression.
There is one known issue in 5.4.31 with chip versions with RTL8208 PHY.
This is fixed in 5.4.32.
Apart from that having r8169 in initramfs can cause issues if realtek.ko
isn't included too. Most distro's take care by checking softdeps in their
initramfs mgmt tools. One known exception is Gentoo, as their genkernel
tool doesn't check softdeps.


> +		}
> +
> +		if (!everything_OK) {
> +			/* Most chip versions fail with the genphy driver.
> +			 * Therefore ensure that the dedicated PHY driver
> +			 * is loaded.
> +			 */
> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> +			mdiobus_unregister(new_bus);
> +			return -EUNATCH;
> +		}
>  	}
>  
>  	/* PHY will be woken up in rtl_open() */
> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	u64_stats_init(&tp->rx_stats.syncp);
>  	u64_stats_init(&tp->tx_stats.syncp);
>  
> +	dev_dbg(&pdev->dev, "init: MAC\n");
>  	rtl_init_mac_address(tp);
>  
>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	dev->hw_features |= NETIF_F_RXFCS;
>  
>  	jumbo_max = rtl_jumbo_max(tp);
> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>  	if (jumbo_max)
>  		dev->max_mtu = jumbo_max;
>  
> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>  	rtl_set_irq_mask(tp);
>  
>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>  
>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>  					    &tp->counters_phys_addr,
> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (!tp->counters)
>  		return -ENOMEM;
>  
> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>  	pci_set_drvdata(pdev, dev);
>  
> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>  	rc = r8169_mdio_register(tp);
> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>  	if (rc)
>  		return rc;
>  
>  	/* chip gets powered up in rtl_open() */
> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>  	rtl_pll_power_down(tp);
>  
>  	rc = register_netdev(dev);
> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>  	if (rc)
>  		goto err_mdio_unregister;
>  
> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (pci_dev_run_wake(pdev))
>  		pm_runtime_put_sync(&pdev->dev);
>  
> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
> +
>  	return 0;
>  
>  err_mdio_unregister:
> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
> index 522760c8bca6..719ea48164f6 100644
> --- a/drivers/net/phy/mdio_bus.c
> +++ b/drivers/net/phy/mdio_bus.c
> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  {
>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
> +	struct phy_device *rv = NULL;
> +
> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>  
>  	if (!mdiodev)
>  		return NULL;
> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>  		return NULL;
>  
> -	return container_of(mdiodev, struct phy_device, mdio);
> +	rv = container_of(mdiodev, struct phy_device, mdio);
> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
> +		 bus->id, addr, mdiodev, rv);
> +	return rv;
>  }
>  EXPORT_SYMBOL(mdiobus_get_phy);
>  
> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>  
>  	bus->state = MDIOBUS_REGISTERED;
> -	pr_info("%s: probed\n", bus->name);
> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>  	return 0;
>  
>  error:
> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>  	while (--i >= 0) {
>  		mdiodev = bus->mdio_map[i];
>  		if (!mdiodev)
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 11:02   ` Lauri Jakku
@ 2020-04-13 11:34     ` Leon Romanovsky
  2020-04-13 11:46       ` Lauri Jakku
  2020-04-13 12:01       ` Lauri Jakku
  0 siblings, 2 replies; 57+ messages in thread
From: Leon Romanovsky @ 2020-04-13 11:34 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: netdev, Heiner Kallweit, nic_swsd

On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
> Hi,
>
> Comments inline.
>
> On 2020-04-13 13:58, Leon Romanovsky wrote:
> > On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
> >> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
> >> From: Lauri Jakku <lja@iki.fi>
> >> Date: Mon, 13 Apr 2020 13:18:35 +0300
> >> Subject: [PATCH] NET: r8168/r8169 identifying fix
> >>
> >> The driver installation determination made properly by
> >> checking PHY vs DRIVER id's.
> >> ---
> >>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
> >>  drivers/net/phy/mdio_bus.c                | 11 +++-
> >>  2 files changed, 72 insertions(+), 9 deletions(-)
> >
> > I would say that most of the code is debug prints.
> >
>
> I tought that they are helpful to keep, they are using the debug calls, so
> they are not visible if user does not like those.

You are missing the point of who are your users.

Users want to have working device and the code. They don't need or like
to debug their kernel.

Thanks

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 11:28     ` Heiner Kallweit
@ 2020-04-13 11:40       ` Lauri Jakku
  2020-04-13 11:57         ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 11:40 UTC (permalink / raw)
  To: Heiner Kallweit, Leon Romanovsky; +Cc: netdev, nic_swsd

Hi,

What i did was fix my problem: The old way of just checking the pointer, did not work correctly. I do have

02:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 03)
03:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 03)

NIC's integrated to motherboard. Without this patch: They will not get recognised: at boot, or with rmmod/modprobe cycle, neither
works.

--> The oldway is not what has been documented sequence.




On 2020-04-13 14:28, Heiner Kallweit wrote:
> On 13.04.2020 13:06, Lauri Jakku wrote:
>> Hi,
>>
>> Modified as suggested:
>>
>> >From f4fba16025260bf08f0df867c3de51803ddb78ef Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..2384da7d2988 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>  
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	new_bus->write = r8169_mdio_write_reg;
>>  
>>  	ret = mdiobus_register(new_bus);
>> +	dev_info(&pdev->dev,
>> +		 "mdiobus_register: %s, %d\n",
>> +		 new_bus->id, ret);
>>  	if (ret)
>>  		return ret;
>>  
>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>> +
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id = tp->phydev->phy_id;
>> +			phydrv_id = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
>> +
>> +			dev_debug(&pdev->dev,
>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>> +				new_bus->id, phydev_id, phydev_masked,
>> +				phydrv_id, drv_masked);
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match = phydev_masked & drv_masked;
>> +			drv_match = drv_match == drv_masked;
>> +
>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>> +				  new_bus->id, phydev_match, drv_match);
>> +
>> +			everything_OK = (phydev_match == drv_match);
> 
> You're doing again what phy_bus_match() did. It's unclear what this should
> be good for. Also it would be helpful if you could explain what the actual
> issue is you're trying to fix, incl. the commit that caused the regression.
> There is one known issue in 5.4.31 with chip versions with RTL8208 PHY.
> This is fixed in 5.4.32.
> Apart from that having r8169 in initramfs can cause issues if realtek.ko
> isn't included too. Most distro's take care by checking softdeps in their
> initramfs mgmt tools. One known exception is Gentoo, as their genkernel
> tool doesn't check softdeps.
> 
> 
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>  
>>  	/* PHY will be woken up in rtl_open() */
>> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	u64_stats_init(&tp->rx_stats.syncp);
>>  	u64_stats_init(&tp->tx_stats.syncp);
>>  
>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>  	rtl_init_mac_address(tp);
>>  
>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	dev->hw_features |= NETIF_F_RXFCS;
>>  
>>  	jumbo_max = rtl_jumbo_max(tp);
>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>  	if (jumbo_max)
>>  		dev->max_mtu = jumbo_max;
>>  
>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>  	rtl_set_irq_mask(tp);
>>  
>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>  
>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>  					    &tp->counters_phys_addr,
>> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (!tp->counters)
>>  		return -ENOMEM;
>>  
>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>  	pci_set_drvdata(pdev, dev);
>>  
>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>  	rc = r8169_mdio_register(tp);
>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>  	if (rc)
>>  		return rc;
>>  
>>  	/* chip gets powered up in rtl_open() */
>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>  	rtl_pll_power_down(tp);
>>  
>>  	rc = register_netdev(dev);
>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>  	if (rc)
>>  		goto err_mdio_unregister;
>>  
>> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (pci_dev_run_wake(pdev))
>>  		pm_runtime_put_sync(&pdev->dev);
>>  
>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>> +
>>  	return 0;
>>  
>>  err_mdio_unregister:
>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>> index 522760c8bca6..719ea48164f6 100644
>> --- a/drivers/net/phy/mdio_bus.c
>> +++ b/drivers/net/phy/mdio_bus.c
>> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  {
>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>> +	struct phy_device *rv = NULL;
>> +
>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>  
>>  	if (!mdiodev)
>>  		return NULL;
>> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>  		return NULL;
>>  
>> -	return container_of(mdiodev, struct phy_device, mdio);
>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>> +		 bus->id, addr, mdiodev, rv);
>> +	return rv;
>>  }
>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>  
>> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>  
>>  	bus->state = MDIOBUS_REGISTERED;
>> -	pr_info("%s: probed\n", bus->name);
>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>  	return 0;
>>  
>>  error:
>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>  	while (--i >= 0) {
>>  		mdiodev = bus->mdio_map[i];
>>  		if (!mdiodev)
>>
> 

-- 
Br,
Lauri J.

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 11:34     ` Leon Romanovsky
@ 2020-04-13 11:46       ` Lauri Jakku
       [not found]         ` <d3adc7f2-06bb-45bc-ab02-3d443999cefd@gmail.com>
  2020-04-13 12:01       ` Lauri Jakku
  1 sibling, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 11:46 UTC (permalink / raw)
  To: Leon Romanovsky; +Cc: netdev, Heiner Kallweit, nic_swsd

Hi,

Fair enough, i'll strip them.

-lja

On 2020-04-13 14:34, Leon Romanovsky wrote:
> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>> Hi,
>>
>> Comments inline.
>>
>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>> From: Lauri Jakku <lja@iki.fi>
>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>
>>>> The driver installation determination made properly by
>>>> checking PHY vs DRIVER id's.
>>>> ---
>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>
>>> I would say that most of the code is debug prints.
>>>
>>
>> I tought that they are helpful to keep, they are using the debug calls, so
>> they are not visible if user does not like those.
> 
> You are missing the point of who are your users.
> 
> Users want to have working device and the code. They don't need or like
> to debug their kernel.
> 
> Thanks
> 

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 11:40       ` Lauri Jakku
@ 2020-04-13 11:57         ` Heiner Kallweit
  2020-04-13 12:04           ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-13 11:57 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 13.04.2020 13:40, Lauri Jakku wrote:
> Hi,
> 
> What i did was fix my problem: The old way of just checking the pointer, did not work correctly. I do have
> 
> 02:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 03)
> 03:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 03)
> 
> NIC's integrated to motherboard. Without this patch: They will not get recognised: at boot, or with rmmod/modprobe cycle, neither
> works.
> 
Helpful would be:
- full dmesg of last known good kernel version
- full dmesg of failing kernel version
- info whether r8169.ko and/or realtek.ko are in your initramfs

> --> The oldway is not what has been documented sequence.
> 
> 
> 
> 
> On 2020-04-13 14:28, Heiner Kallweit wrote:
>> On 13.04.2020 13:06, Lauri Jakku wrote:
>>> Hi,
>>>
>>> Modified as suggested:
>>>
>>> >From f4fba16025260bf08f0df867c3de51803ddb78ef Mon Sep 17 00:00:00 2001
>>> From: Lauri Jakku <lja@iki.fi>
>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>
>>> The driver installation determination made properly by
>>> checking PHY vs DRIVER id's.
>>> ---
>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>> index bf5bf05970a2..2384da7d2988 100644
>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  {
>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>  	struct mii_bus *new_bus;
>>> +	u32 phydev_id = 0;
>>> +	u32 phydrv_id = 0;
>>> +	u32 phydrv_id_mask = 0;
>>>  	int ret;
>>>  
>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  	new_bus->write = r8169_mdio_write_reg;
>>>  
>>>  	ret = mdiobus_register(new_bus);
>>> +	dev_info(&pdev->dev,
>>> +		 "mdiobus_register: %s, %d\n",
>>> +		 new_bus->id, ret);
>>>  	if (ret)
>>>  		return ret;
>>>  
>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>> +
>>>  	if (!tp->phydev) {
>>>  		mdiobus_unregister(new_bus);
>>>  		return -ENODEV;
>>> -	} else if (!tp->phydev->drv) {
>>> -		/* Most chip versions fail with the genphy driver.
>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>> -		 */
>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> -		mdiobus_unregister(new_bus);
>>> -		return -EUNATCH;
>>> +	} else {
>>> +		/* tp -> phydev ok */
>>> +		int everything_OK = 0;
>>> +
>>> +		/* Check driver id versus phy */
>>> +
>>> +		if (tp->phydev->drv) {
>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>> +			u32 drv_masked = ~0;
>>> +			u32 phydev_match = ~0;
>>> +			u32 drv_match = 0xDEADBEEF;
>>> +
>>> +			phydev_id = tp->phydev->phy_id;
>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>> +
>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>> +
>>> +			dev_debug(&pdev->dev,
>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>> +				new_bus->id, phydev_id, phydev_masked,
>>> +				phydrv_id, drv_masked);
>>> +
>>> +			phydev_match = phydev_masked & drv_masked;
>>> +			phydev_match = phydev_match == phydev_masked;
>>> +
>>> +			drv_match = phydev_masked & drv_masked;
>>> +			drv_match = drv_match == drv_masked;
>>> +
>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>> +				  new_bus->id, phydev_match, drv_match);
>>> +
>>> +			everything_OK = (phydev_match == drv_match);
>>
>> You're doing again what phy_bus_match() did. It's unclear what this should
>> be good for. Also it would be helpful if you could explain what the actual
>> issue is you're trying to fix, incl. the commit that caused the regression.
>> There is one known issue in 5.4.31 with chip versions with RTL8208 PHY.
>> This is fixed in 5.4.32.
>> Apart from that having r8169 in initramfs can cause issues if realtek.ko
>> isn't included too. Most distro's take care by checking softdeps in their
>> initramfs mgmt tools. One known exception is Gentoo, as their genkernel
>> tool doesn't check softdeps.
>>
>>
>>> +		}
>>> +
>>> +		if (!everything_OK) {
>>> +			/* Most chip versions fail with the genphy driver.
>>> +			 * Therefore ensure that the dedicated PHY driver
>>> +			 * is loaded.
>>> +			 */
>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> +			mdiobus_unregister(new_bus);
>>> +			return -EUNATCH;
>>> +		}
>>>  	}
>>>  
>>>  	/* PHY will be woken up in rtl_open() */
>>> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>  
>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>  	rtl_init_mac_address(tp);
>>>  
>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>  
>>>  	jumbo_max = rtl_jumbo_max(tp);
>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>  	if (jumbo_max)
>>>  		dev->max_mtu = jumbo_max;
>>>  
>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>  	rtl_set_irq_mask(tp);
>>>  
>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>  
>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>  					    &tp->counters_phys_addr,
>>> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	if (!tp->counters)
>>>  		return -ENOMEM;
>>>  
>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>  	pci_set_drvdata(pdev, dev);
>>>  
>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>  	rc = r8169_mdio_register(tp);
>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>  	if (rc)
>>>  		return rc;
>>>  
>>>  	/* chip gets powered up in rtl_open() */
>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>  	rtl_pll_power_down(tp);
>>>  
>>>  	rc = register_netdev(dev);
>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>  	if (rc)
>>>  		goto err_mdio_unregister;
>>>  
>>> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	if (pci_dev_run_wake(pdev))
>>>  		pm_runtime_put_sync(&pdev->dev);
>>>  
>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>> +
>>>  	return 0;
>>>  
>>>  err_mdio_unregister:
>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>> index 522760c8bca6..719ea48164f6 100644
>>> --- a/drivers/net/phy/mdio_bus.c
>>> +++ b/drivers/net/phy/mdio_bus.c
>>> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>  {
>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>> +	struct phy_device *rv = NULL;
>>> +
>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>  
>>>  	if (!mdiodev)
>>>  		return NULL;
>>> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>  		return NULL;
>>>  
>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>> +		 bus->id, addr, mdiodev, rv);
>>> +	return rv;
>>>  }
>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>  
>>> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>  
>>>  	bus->state = MDIOBUS_REGISTERED;
>>> -	pr_info("%s: probed\n", bus->name);
>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>  	return 0;
>>>  
>>>  error:
>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>  	while (--i >= 0) {
>>>  		mdiodev = bus->mdio_map[i];
>>>  		if (!mdiodev)
>>>
>>
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 11:34     ` Leon Romanovsky
  2020-04-13 11:46       ` Lauri Jakku
@ 2020-04-13 12:01       ` Lauri Jakku
  2020-04-13 12:18         ` Leon Romanovsky
  1 sibling, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 12:01 UTC (permalink / raw)
  To: Leon Romanovsky; +Cc: netdev, Heiner Kallweit, nic_swsd

[-- Attachment #1: Type: text/plain, Size: 8050 bytes --]

Hi,

I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.


Latest version:

From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
 drivers/net/phy/mdio_bus.c                |  15 ++-
 2 files changed, 119 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..5e992f285527 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -61,6 +61,11 @@
 #define R8169_MSG_DEFAULT \
 	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
 
+
+/*
+#define R8169_PROBE_DEBUG
+*/
+	
 /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
    The RTL chips use a 64 element hash table based on the Ethernet CRC. */
 #define	MC_FILTER_LIMIT	32
@@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	new_bus->write = r8169_mdio_write_reg;
 
 	ret = mdiobus_register(new_bus);
+
+#ifdef R8169_PROBE_DEBUG
+	dev_info(&pdev->dev,
+		 "mdiobus_register: %s, %d\n",
+		 new_bus->id, ret);
+#endif
 	if (ret)
 		return ret;
 
 	tp->phydev = mdiobus_get_phy(new_bus, 0);
+
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id = tp->phydev->phy_id;
+			phydrv_id = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+		
+#ifdef R8169_PROBE_DEBUG
+			dev_debug(&pdev->dev,
+				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
+				new_bus->id, phydev_id, phydev_masked,
+				phydrv_id, drv_masked);
+#endif
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match = phydev_masked & drv_masked;
+			drv_match = drv_match == drv_masked;
+
+#ifdef R8169_PROBE_DEBUG
+			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
+				  new_bus->id, phydev_match, drv_match);
+#endif
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
@@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	u64_stats_init(&tp->rx_stats.syncp);
 	u64_stats_init(&tp->tx_stats.syncp);
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: MAC\n");
+#endif
+
 	rtl_init_mac_address(tp);
 
 	dev->ethtool_ops = &rtl8169_ethtool_ops;
@@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	dev->hw_features |= NETIF_F_RXFCS;
 
 	jumbo_max = rtl_jumbo_max(tp);
+
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
+#endif
+
 	if (jumbo_max)
 		dev->max_mtu = jumbo_max;
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: irq mask\n");
+#endif
+	
 	rtl_set_irq_mask(tp);
 
 	tp->fw_name = rtl_chip_infos[chipset].fw_name;
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
+#endif
+
 	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
 					    &tp->counters_phys_addr,
 					    GFP_KERNEL);
 	if (!tp->counters)
 		return -ENOMEM;
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: set driver data\n");
+#endif
+
 	pci_set_drvdata(pdev, dev);
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: register mdio\n");
+#endif
+
 	rc = r8169_mdio_register(tp);
+
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
+#endif
+
 	if (rc)
 		return rc;
 
 	/* chip gets powered up in rtl_open() */
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: pll pwr down\n");
+#endif
+
 	rtl_pll_power_down(tp);
 
 	rc = register_netdev(dev);
+
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
+#endif
+
 	if (rc)
 		goto err_mdio_unregister;
 
@@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (pci_dev_run_wake(pdev))
 		pm_runtime_put_sync(&pdev->dev);
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
+#endif
+
 	return 0;
 
 err_mdio_unregister:
diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
index 522760c8bca6..41777f379a57 100644
--- a/drivers/net/phy/mdio_bus.c
+++ b/drivers/net/phy/mdio_bus.c
@@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
 struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 {
 	struct mdio_device *mdiodev = bus->mdio_map[addr];
-
+	struct phy_device *rv = NULL;
+/*
+	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
+*/
 	if (!mdiodev)
 		return NULL;
 
 	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
 		return NULL;
 
-	return container_of(mdiodev, struct phy_device, mdio);
+	rv = container_of(mdiodev, struct phy_device, mdio);
+/*
+ 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
+		 bus->id, addr, mdiodev, rv);
+*/
+	return rv;
 }
 EXPORT_SYMBOL(mdiobus_get_phy);
 
@@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
 	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
 
 	bus->state = MDIOBUS_REGISTERED;
-	pr_info("%s: probed\n", bus->name);
+	pr_info("%s: probed (mdiobus_register)\n", bus->name);
 	return 0;
 
 error:
+	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
 	while (--i >= 0) {
 		mdiodev = bus->mdio_map[i];
 		if (!mdiodev)
-- 
2.26.0




On 2020-04-13 14:34, Leon Romanovsky wrote:
> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>> Hi,
>>
>> Comments inline.
>>
>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>> From: Lauri Jakku <lja@iki.fi>
>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>
>>>> The driver installation determination made properly by
>>>> checking PHY vs DRIVER id's.
>>>> ---
>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>
>>> I would say that most of the code is debug prints.
>>>
>>
>> I tought that they are helpful to keep, they are using the debug calls, so
>> they are not visible if user does not like those.
> 
> You are missing the point of who are your users.
> 
> Users want to have working device and the code. They don't need or like
> to debug their kernel.
> 
> Thanks
> 

-- 
Br,
Lauri J.

[-- Attachment #2: NET-r8169-module-enchansments.patch --]
[-- Type: text/x-patch, Size: 6759 bytes --]

From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
 drivers/net/phy/mdio_bus.c                |  15 ++-
 2 files changed, 119 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..5e992f285527 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -61,6 +61,11 @@
 #define R8169_MSG_DEFAULT \
 	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
 
+
+/*
+#define R8169_PROBE_DEBUG
+*/
+	
 /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
    The RTL chips use a 64 element hash table based on the Ethernet CRC. */
 #define	MC_FILTER_LIMIT	32
@@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	new_bus->write = r8169_mdio_write_reg;
 
 	ret = mdiobus_register(new_bus);
+
+#ifdef R8169_PROBE_DEBUG
+	dev_info(&pdev->dev,
+		 "mdiobus_register: %s, %d\n",
+		 new_bus->id, ret);
+#endif
 	if (ret)
 		return ret;
 
 	tp->phydev = mdiobus_get_phy(new_bus, 0);
+
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id = tp->phydev->phy_id;
+			phydrv_id = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+		
+#ifdef R8169_PROBE_DEBUG
+			dev_debug(&pdev->dev,
+				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
+				new_bus->id, phydev_id, phydev_masked,
+				phydrv_id, drv_masked);
+#endif
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match = phydev_masked & drv_masked;
+			drv_match = drv_match == drv_masked;
+
+#ifdef R8169_PROBE_DEBUG
+			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
+				  new_bus->id, phydev_match, drv_match);
+#endif
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
@@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	u64_stats_init(&tp->rx_stats.syncp);
 	u64_stats_init(&tp->tx_stats.syncp);
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: MAC\n");
+#endif
+
 	rtl_init_mac_address(tp);
 
 	dev->ethtool_ops = &rtl8169_ethtool_ops;
@@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	dev->hw_features |= NETIF_F_RXFCS;
 
 	jumbo_max = rtl_jumbo_max(tp);
+
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
+#endif
+
 	if (jumbo_max)
 		dev->max_mtu = jumbo_max;
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: irq mask\n");
+#endif
+	
 	rtl_set_irq_mask(tp);
 
 	tp->fw_name = rtl_chip_infos[chipset].fw_name;
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
+#endif
+
 	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
 					    &tp->counters_phys_addr,
 					    GFP_KERNEL);
 	if (!tp->counters)
 		return -ENOMEM;
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: set driver data\n");
+#endif
+
 	pci_set_drvdata(pdev, dev);
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: register mdio\n");
+#endif
+
 	rc = r8169_mdio_register(tp);
+
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
+#endif
+
 	if (rc)
 		return rc;
 
 	/* chip gets powered up in rtl_open() */
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: pll pwr down\n");
+#endif
+
 	rtl_pll_power_down(tp);
 
 	rc = register_netdev(dev);
+
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
+#endif
+
 	if (rc)
 		goto err_mdio_unregister;
 
@@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (pci_dev_run_wake(pdev))
 		pm_runtime_put_sync(&pdev->dev);
 
+#ifdef R816X_PROBE_DEBUG
+	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
+#endif
+
 	return 0;
 
 err_mdio_unregister:
diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
index 522760c8bca6..41777f379a57 100644
--- a/drivers/net/phy/mdio_bus.c
+++ b/drivers/net/phy/mdio_bus.c
@@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
 struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
 {
 	struct mdio_device *mdiodev = bus->mdio_map[addr];
-
+	struct phy_device *rv = NULL;
+/*
+	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
+*/
 	if (!mdiodev)
 		return NULL;
 
 	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
 		return NULL;
 
-	return container_of(mdiodev, struct phy_device, mdio);
+	rv = container_of(mdiodev, struct phy_device, mdio);
+/*
+ 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
+		 bus->id, addr, mdiodev, rv);
+*/
+	return rv;
 }
 EXPORT_SYMBOL(mdiobus_get_phy);
 
@@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
 	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
 
 	bus->state = MDIOBUS_REGISTERED;
-	pr_info("%s: probed\n", bus->name);
+	pr_info("%s: probed (mdiobus_register)\n", bus->name);
 	return 0;
 
 error:
+	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
 	while (--i >= 0) {
 		mdiodev = bus->mdio_map[i];
 		if (!mdiodev)
-- 
2.26.0


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 11:57         ` Heiner Kallweit
@ 2020-04-13 12:04           ` Lauri Jakku
  0 siblings, 0 replies; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 12:04 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi, ok i'll provide today.. i recall that 5.3.x was ok, and from 5.4.x i found that there was not, but i'll recheck and get logs.


On 2020-04-13 14:57, Heiner Kallweit wrote:
> On 13.04.2020 13:40, Lauri Jakku wrote:
>> Hi,
>>
>> What i did was fix my problem: The old way of just checking the pointer, did not work correctly. I do have
>>
>> 02:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 03)
>> 03:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 03)
>>
>> NIC's integrated to motherboard. Without this patch: They will not get recognised: at boot, or with rmmod/modprobe cycle, neither
>> works.
>>
> Helpful would be:
> - full dmesg of last known good kernel version
> - full dmesg of failing kernel version
> - info whether r8169.ko and/or realtek.ko are in your initramfs
> 
>> --> The oldway is not what has been documented sequence.
>>
>>
>>
>>
>> On 2020-04-13 14:28, Heiner Kallweit wrote:
>>> On 13.04.2020 13:06, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> Modified as suggested:
>>>>
>>>> >From f4fba16025260bf08f0df867c3de51803ddb78ef Mon Sep 17 00:00:00 2001
>>>> From: Lauri Jakku <lja@iki.fi>
>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>
>>>> The driver installation determination made properly by
>>>> checking PHY vs DRIVER id's.
>>>> ---
>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>
>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>> index bf5bf05970a2..2384da7d2988 100644
>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  {
>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>  	struct mii_bus *new_bus;
>>>> +	u32 phydev_id = 0;
>>>> +	u32 phydrv_id = 0;
>>>> +	u32 phydrv_id_mask = 0;
>>>>  	int ret;
>>>>  
>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>> @@ -5165,20 +5168,62 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>  
>>>>  	ret = mdiobus_register(new_bus);
>>>> +	dev_info(&pdev->dev,
>>>> +		 "mdiobus_register: %s, %d\n",
>>>> +		 new_bus->id, ret);
>>>>  	if (ret)
>>>>  		return ret;
>>>>  
>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>> +
>>>>  	if (!tp->phydev) {
>>>>  		mdiobus_unregister(new_bus);
>>>>  		return -ENODEV;
>>>> -	} else if (!tp->phydev->drv) {
>>>> -		/* Most chip versions fail with the genphy driver.
>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>> -		 */
>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> -		mdiobus_unregister(new_bus);
>>>> -		return -EUNATCH;
>>>> +	} else {
>>>> +		/* tp -> phydev ok */
>>>> +		int everything_OK = 0;
>>>> +
>>>> +		/* Check driver id versus phy */
>>>> +
>>>> +		if (tp->phydev->drv) {
>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>> +			u32 drv_masked = ~0;
>>>> +			u32 phydev_match = ~0;
>>>> +			u32 drv_match = 0xDEADBEEF;
>>>> +
>>>> +			phydev_id = tp->phydev->phy_id;
>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>> +
>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>> +
>>>> +			dev_debug(&pdev->dev,
>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>> +				phydrv_id, drv_masked);
>>>> +
>>>> +			phydev_match = phydev_masked & drv_masked;
>>>> +			phydev_match = phydev_match == phydev_masked;
>>>> +
>>>> +			drv_match = phydev_masked & drv_masked;
>>>> +			drv_match = drv_match == drv_masked;
>>>> +
>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>> +				  new_bus->id, phydev_match, drv_match);
>>>> +
>>>> +			everything_OK = (phydev_match == drv_match);
>>>
>>> You're doing again what phy_bus_match() did. It's unclear what this should
>>> be good for. Also it would be helpful if you could explain what the actual
>>> issue is you're trying to fix, incl. the commit that caused the regression.
>>> There is one known issue in 5.4.31 with chip versions with RTL8208 PHY.
>>> This is fixed in 5.4.32.
>>> Apart from that having r8169 in initramfs can cause issues if realtek.ko
>>> isn't included too. Most distro's take care by checking softdeps in their
>>> initramfs mgmt tools. One known exception is Gentoo, as their genkernel
>>> tool doesn't check softdeps.
>>>
>>>
>>>> +		}
>>>> +
>>>> +		if (!everything_OK) {
>>>> +			/* Most chip versions fail with the genphy driver.
>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>> +			 * is loaded.
>>>> +			 */
>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> +			mdiobus_unregister(new_bus);
>>>> +			return -EUNATCH;
>>>> +		}
>>>>  	}
>>>>  
>>>>  	/* PHY will be woken up in rtl_open() */
>>>> @@ -5435,6 +5480,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>  
>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>  	rtl_init_mac_address(tp);
>>>>  
>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>> @@ -5483,12 +5529,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>  
>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>  	if (jumbo_max)
>>>>  		dev->max_mtu = jumbo_max;
>>>>  
>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>  	rtl_set_irq_mask(tp);
>>>>  
>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>  
>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>  					    &tp->counters_phys_addr,
>>>> @@ -5496,16 +5545,21 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	if (!tp->counters)
>>>>  		return -ENOMEM;
>>>>  
>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>  	pci_set_drvdata(pdev, dev);
>>>>  
>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>  	rc = r8169_mdio_register(tp);
>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>  	if (rc)
>>>>  		return rc;
>>>>  
>>>>  	/* chip gets powered up in rtl_open() */
>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>  	rtl_pll_power_down(tp);
>>>>  
>>>>  	rc = register_netdev(dev);
>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>  	if (rc)
>>>>  		goto err_mdio_unregister;
>>>>  
>>>> @@ -5525,6 +5579,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	if (pci_dev_run_wake(pdev))
>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>  
>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>> +
>>>>  	return 0;
>>>>  
>>>>  err_mdio_unregister:
>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>> index 522760c8bca6..719ea48164f6 100644
>>>> --- a/drivers/net/phy/mdio_bus.c
>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>> @@ -112,6 +112,9 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>  {
>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>> +	struct phy_device *rv = NULL;
>>>> +
>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>  
>>>>  	if (!mdiodev)
>>>>  		return NULL;
>>>> @@ -119,7 +122,10 @@ struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>  		return NULL;
>>>>  
>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>> +	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>> +		 bus->id, addr, mdiodev, rv);
>>>> +	return rv;
>>>>  }
>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>  
>>>> @@ -645,10 +651,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>  
>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>> -	pr_info("%s: probed\n", bus->name);
>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>  	return 0;
>>>>  
>>>>  error:
>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>  	while (--i >= 0) {
>>>>  		mdiodev = bus->mdio_map[i];
>>>>  		if (!mdiodev)
>>>>
>>>
>>
> 

-- 
Br,
Lauri J.

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 12:01       ` Lauri Jakku
@ 2020-04-13 12:18         ` Leon Romanovsky
  2020-04-13 14:44           ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Leon Romanovsky @ 2020-04-13 12:18 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: netdev, Heiner Kallweit, nic_swsd

On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
> Hi,
>
> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>

Please stop to do top-posting.

>
> Latest version:
>
> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
> From: Lauri Jakku <lja@iki.fi>
> Date: Mon, 13 Apr 2020 13:18:35 +0300
> Subject: [PATCH] NET: r8168/r8169 identifying fix
>
> The driver installation determination made properly by
> checking PHY vs DRIVER id's.
> ---
>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>  drivers/net/phy/mdio_bus.c                |  15 ++-
>  2 files changed, 119 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..5e992f285527 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -61,6 +61,11 @@
>  #define R8169_MSG_DEFAULT \
>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>
> +
> +/*
> +#define R8169_PROBE_DEBUG
> +*/

Of course not, it is even worse than before.
If user recompiles module, he will add prints.

Thanks

> +
>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>  #define	MC_FILTER_LIMIT	32
> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  {
>  	struct pci_dev *pdev = tp->pci_dev;
>  	struct mii_bus *new_bus;
> +	u32 phydev_id = 0;
> +	u32 phydrv_id = 0;
> +	u32 phydrv_id_mask = 0;
>  	int ret;
>
>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  	new_bus->write = r8169_mdio_write_reg;
>
>  	ret = mdiobus_register(new_bus);
> +
> +#ifdef R8169_PROBE_DEBUG
> +	dev_info(&pdev->dev,
> +		 "mdiobus_register: %s, %d\n",
> +		 new_bus->id, ret);
> +#endif
>  	if (ret)
>  		return ret;
>
>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
> +
>  	if (!tp->phydev) {
>  		mdiobus_unregister(new_bus);
>  		return -ENODEV;
> -	} else if (!tp->phydev->drv) {
> -		/* Most chip versions fail with the genphy driver.
> -		 * Therefore ensure that the dedicated PHY driver is loaded.
> -		 */
> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> -		mdiobus_unregister(new_bus);
> -		return -EUNATCH;
> +	} else {
> +		/* tp -> phydev ok */
> +		int everything_OK = 0;
> +
> +		/* Check driver id versus phy */
> +
> +		if (tp->phydev->drv) {
> +			u32 phydev_masked = 0xBEEFDEAD;
> +			u32 drv_masked = ~0;
> +			u32 phydev_match = ~0;
> +			u32 drv_match = 0xDEADBEEF;
> +
> +			phydev_id = tp->phydev->phy_id;
> +			phydrv_id = tp->phydev->drv->phy_id;
> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
> +
> +			drv_masked = phydrv_id & phydrv_id_mask;
> +			phydev_masked = phydev_id & phydrv_id_mask;
> +
> +#ifdef R8169_PROBE_DEBUG
> +			dev_debug(&pdev->dev,
> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
> +				new_bus->id, phydev_id, phydev_masked,
> +				phydrv_id, drv_masked);
> +#endif
> +
> +			phydev_match = phydev_masked & drv_masked;
> +			phydev_match = phydev_match == phydev_masked;
> +
> +			drv_match = phydev_masked & drv_masked;
> +			drv_match = drv_match == drv_masked;
> +
> +#ifdef R8169_PROBE_DEBUG
> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
> +				  new_bus->id, phydev_match, drv_match);
> +#endif
> +
> +			everything_OK = (phydev_match == drv_match);
> +		}
> +
> +		if (!everything_OK) {
> +			/* Most chip versions fail with the genphy driver.
> +			 * Therefore ensure that the dedicated PHY driver
> +			 * is loaded.
> +			 */
> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> +			mdiobus_unregister(new_bus);
> +			return -EUNATCH;
> +		}
>  	}
>
>  	/* PHY will be woken up in rtl_open() */
> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	u64_stats_init(&tp->rx_stats.syncp);
>  	u64_stats_init(&tp->tx_stats.syncp);
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: MAC\n");
> +#endif
> +
>  	rtl_init_mac_address(tp);
>
>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	dev->hw_features |= NETIF_F_RXFCS;
>
>  	jumbo_max = rtl_jumbo_max(tp);
> +
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
> +#endif
> +
>  	if (jumbo_max)
>  		dev->max_mtu = jumbo_max;
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: irq mask\n");
> +#endif
> +
>  	rtl_set_irq_mask(tp);
>
>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
> +#endif
> +
>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>  					    &tp->counters_phys_addr,
>  					    GFP_KERNEL);
>  	if (!tp->counters)
>  		return -ENOMEM;
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: set driver data\n");
> +#endif
> +
>  	pci_set_drvdata(pdev, dev);
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: register mdio\n");
> +#endif
> +
>  	rc = r8169_mdio_register(tp);
> +
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
> +#endif
> +
>  	if (rc)
>  		return rc;
>
>  	/* chip gets powered up in rtl_open() */
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
> +#endif
> +
>  	rtl_pll_power_down(tp);
>
>  	rc = register_netdev(dev);
> +
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
> +#endif
> +
>  	if (rc)
>  		goto err_mdio_unregister;
>
> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (pci_dev_run_wake(pdev))
>  		pm_runtime_put_sync(&pdev->dev);
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
> +#endif
> +
>  	return 0;
>
>  err_mdio_unregister:
> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
> index 522760c8bca6..41777f379a57 100644
> --- a/drivers/net/phy/mdio_bus.c
> +++ b/drivers/net/phy/mdio_bus.c
> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  {
>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
> -
> +	struct phy_device *rv = NULL;
> +/*
> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
> +*/
>  	if (!mdiodev)
>  		return NULL;
>
>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>  		return NULL;
>
> -	return container_of(mdiodev, struct phy_device, mdio);
> +	rv = container_of(mdiodev, struct phy_device, mdio);
> +/*
> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
> +		 bus->id, addr, mdiodev, rv);
> +*/
> +	return rv;
>  }
>  EXPORT_SYMBOL(mdiobus_get_phy);
>
> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>
>  	bus->state = MDIOBUS_REGISTERED;
> -	pr_info("%s: probed\n", bus->name);
> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>  	return 0;
>
>  error:
> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>  	while (--i >= 0) {
>  		mdiodev = bus->mdio_map[i];
>  		if (!mdiodev)
> --
> 2.26.0
>
>
>
>
> On 2020-04-13 14:34, Leon Romanovsky wrote:
> > On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
> >> Hi,
> >>
> >> Comments inline.
> >>
> >> On 2020-04-13 13:58, Leon Romanovsky wrote:
> >>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
> >>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
> >>>> From: Lauri Jakku <lja@iki.fi>
> >>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
> >>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
> >>>>
> >>>> The driver installation determination made properly by
> >>>> checking PHY vs DRIVER id's.
> >>>> ---
> >>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
> >>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
> >>>>  2 files changed, 72 insertions(+), 9 deletions(-)
> >>>
> >>> I would say that most of the code is debug prints.
> >>>
> >>
> >> I tought that they are helpful to keep, they are using the debug calls, so
> >> they are not visible if user does not like those.
> >
> > You are missing the point of who are your users.
> >
> > Users want to have working device and the code. They don't need or like
> > to debug their kernel.
> >
> > Thanks
> >
>
> --
> Br,
> Lauri J.

> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
> From: Lauri Jakku <lja@iki.fi>
> Date: Mon, 13 Apr 2020 13:18:35 +0300
> Subject: [PATCH] NET: r8168/r8169 identifying fix
>
> The driver installation determination made properly by
> checking PHY vs DRIVER id's.
> ---
>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>  drivers/net/phy/mdio_bus.c                |  15 ++-
>  2 files changed, 119 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..5e992f285527 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -61,6 +61,11 @@
>  #define R8169_MSG_DEFAULT \
>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>
> +
> +/*
> +#define R8169_PROBE_DEBUG
> +*/
> +
>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>  #define	MC_FILTER_LIMIT	32
> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  {
>  	struct pci_dev *pdev = tp->pci_dev;
>  	struct mii_bus *new_bus;
> +	u32 phydev_id = 0;
> +	u32 phydrv_id = 0;
> +	u32 phydrv_id_mask = 0;
>  	int ret;
>
>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  	new_bus->write = r8169_mdio_write_reg;
>
>  	ret = mdiobus_register(new_bus);
> +
> +#ifdef R8169_PROBE_DEBUG
> +	dev_info(&pdev->dev,
> +		 "mdiobus_register: %s, %d\n",
> +		 new_bus->id, ret);
> +#endif
>  	if (ret)
>  		return ret;
>
>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
> +
>  	if (!tp->phydev) {
>  		mdiobus_unregister(new_bus);
>  		return -ENODEV;
> -	} else if (!tp->phydev->drv) {
> -		/* Most chip versions fail with the genphy driver.
> -		 * Therefore ensure that the dedicated PHY driver is loaded.
> -		 */
> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> -		mdiobus_unregister(new_bus);
> -		return -EUNATCH;
> +	} else {
> +		/* tp -> phydev ok */
> +		int everything_OK = 0;
> +
> +		/* Check driver id versus phy */
> +
> +		if (tp->phydev->drv) {
> +			u32 phydev_masked = 0xBEEFDEAD;
> +			u32 drv_masked = ~0;
> +			u32 phydev_match = ~0;
> +			u32 drv_match = 0xDEADBEEF;
> +
> +			phydev_id = tp->phydev->phy_id;
> +			phydrv_id = tp->phydev->drv->phy_id;
> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
> +
> +			drv_masked = phydrv_id & phydrv_id_mask;
> +			phydev_masked = phydev_id & phydrv_id_mask;
> +
> +#ifdef R8169_PROBE_DEBUG
> +			dev_debug(&pdev->dev,
> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
> +				new_bus->id, phydev_id, phydev_masked,
> +				phydrv_id, drv_masked);
> +#endif
> +
> +			phydev_match = phydev_masked & drv_masked;
> +			phydev_match = phydev_match == phydev_masked;
> +
> +			drv_match = phydev_masked & drv_masked;
> +			drv_match = drv_match == drv_masked;
> +
> +#ifdef R8169_PROBE_DEBUG
> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
> +				  new_bus->id, phydev_match, drv_match);
> +#endif
> +
> +			everything_OK = (phydev_match == drv_match);
> +		}
> +
> +		if (!everything_OK) {
> +			/* Most chip versions fail with the genphy driver.
> +			 * Therefore ensure that the dedicated PHY driver
> +			 * is loaded.
> +			 */
> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> +			mdiobus_unregister(new_bus);
> +			return -EUNATCH;
> +		}
>  	}
>
>  	/* PHY will be woken up in rtl_open() */
> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	u64_stats_init(&tp->rx_stats.syncp);
>  	u64_stats_init(&tp->tx_stats.syncp);
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: MAC\n");
> +#endif
> +
>  	rtl_init_mac_address(tp);
>
>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	dev->hw_features |= NETIF_F_RXFCS;
>
>  	jumbo_max = rtl_jumbo_max(tp);
> +
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
> +#endif
> +
>  	if (jumbo_max)
>  		dev->max_mtu = jumbo_max;
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: irq mask\n");
> +#endif
> +
>  	rtl_set_irq_mask(tp);
>
>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
> +#endif
> +
>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>  					    &tp->counters_phys_addr,
>  					    GFP_KERNEL);
>  	if (!tp->counters)
>  		return -ENOMEM;
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: set driver data\n");
> +#endif
> +
>  	pci_set_drvdata(pdev, dev);
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: register mdio\n");
> +#endif
> +
>  	rc = r8169_mdio_register(tp);
> +
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
> +#endif
> +
>  	if (rc)
>  		return rc;
>
>  	/* chip gets powered up in rtl_open() */
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
> +#endif
> +
>  	rtl_pll_power_down(tp);
>
>  	rc = register_netdev(dev);
> +
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
> +#endif
> +
>  	if (rc)
>  		goto err_mdio_unregister;
>
> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (pci_dev_run_wake(pdev))
>  		pm_runtime_put_sync(&pdev->dev);
>
> +#ifdef R816X_PROBE_DEBUG
> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
> +#endif
> +
>  	return 0;
>
>  err_mdio_unregister:
> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
> index 522760c8bca6..41777f379a57 100644
> --- a/drivers/net/phy/mdio_bus.c
> +++ b/drivers/net/phy/mdio_bus.c
> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>  {
>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
> -
> +	struct phy_device *rv = NULL;
> +/*
> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
> +*/
>  	if (!mdiodev)
>  		return NULL;
>
>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>  		return NULL;
>
> -	return container_of(mdiodev, struct phy_device, mdio);
> +	rv = container_of(mdiodev, struct phy_device, mdio);
> +/*
> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
> +		 bus->id, addr, mdiodev, rv);
> +*/
> +	return rv;
>  }
>  EXPORT_SYMBOL(mdiobus_get_phy);
>
> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>
>  	bus->state = MDIOBUS_REGISTERED;
> -	pr_info("%s: probed\n", bus->name);
> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>  	return 0;
>
>  error:
> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>  	while (--i >= 0) {
>  		mdiodev = bus->mdio_map[i];
>  		if (!mdiodev)
> --
> 2.26.0
>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 12:18         ` Leon Romanovsky
@ 2020-04-13 14:44           ` Lauri Jakku
  2020-04-13 15:33             ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 14:44 UTC (permalink / raw)
  To: Leon Romanovsky; +Cc: netdev, Heiner Kallweit, nic_swsd

[-- Attachment #1: Type: text/plain, Size: 19123 bytes --]

Hi,

On 13.4.2020 15.18, Leon Romanovsky wrote:
> On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
>> Hi,
>>
>> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>>
> 
> Please stop to do top-posting.

 I did not realise, sorry. Trying to do better.

> 
>>
>> Latest version:
>>
>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..5e992f285527 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -61,6 +61,11 @@
>>  #define R8169_MSG_DEFAULT \
>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>
>> +
>> +/*
>> +#define R8169_PROBE_DEBUG
>> +*/
> 
> Of course not, it is even worse than before.
> If user recompiles module, he will add prints.
> 
> Thanks
> 

New patch at the end and in attachments.


>> +
>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>  #define	MC_FILTER_LIMIT	32
>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	new_bus->write = r8169_mdio_write_reg;
>>
>>  	ret = mdiobus_register(new_bus);
>> +
>> +#ifdef R8169_PROBE_DEBUG
>> +	dev_info(&pdev->dev,
>> +		 "mdiobus_register: %s, %d\n",
>> +		 new_bus->id, ret);
>> +#endif
>>  	if (ret)
>>  		return ret;
>>
>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>> +
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id = tp->phydev->phy_id;
>> +			phydrv_id = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
>> +
>> +#ifdef R8169_PROBE_DEBUG
>> +			dev_debug(&pdev->dev,
>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>> +				new_bus->id, phydev_id, phydev_masked,
>> +				phydrv_id, drv_masked);
>> +#endif
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match = phydev_masked & drv_masked;
>> +			drv_match = drv_match == drv_masked;
>> +
>> +#ifdef R8169_PROBE_DEBUG
>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>> +				  new_bus->id, phydev_match, drv_match);
>> +#endif
>> +
>> +			everything_OK = (phydev_match == drv_match);
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>
>>  	/* PHY will be woken up in rtl_open() */
>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	u64_stats_init(&tp->rx_stats.syncp);
>>  	u64_stats_init(&tp->tx_stats.syncp);
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>> +#endif
>> +
>>  	rtl_init_mac_address(tp);
>>
>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	dev->hw_features |= NETIF_F_RXFCS;
>>
>>  	jumbo_max = rtl_jumbo_max(tp);
>> +
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>> +#endif
>> +
>>  	if (jumbo_max)
>>  		dev->max_mtu = jumbo_max;
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>> +#endif
>> +
>>  	rtl_set_irq_mask(tp);
>>
>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>> +#endif
>> +
>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>  					    &tp->counters_phys_addr,
>>  					    GFP_KERNEL);
>>  	if (!tp->counters)
>>  		return -ENOMEM;
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>> +#endif
>> +
>>  	pci_set_drvdata(pdev, dev);
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>> +#endif
>> +
>>  	rc = r8169_mdio_register(tp);
>> +
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>> +#endif
>> +
>>  	if (rc)
>>  		return rc;
>>
>>  	/* chip gets powered up in rtl_open() */
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>> +#endif
>> +
>>  	rtl_pll_power_down(tp);
>>
>>  	rc = register_netdev(dev);
>> +
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>> +#endif
>> +
>>  	if (rc)
>>  		goto err_mdio_unregister;
>>
>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (pci_dev_run_wake(pdev))
>>  		pm_runtime_put_sync(&pdev->dev);
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>> +#endif
>> +
>>  	return 0;
>>
>>  err_mdio_unregister:
>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>> index 522760c8bca6..41777f379a57 100644
>> --- a/drivers/net/phy/mdio_bus.c
>> +++ b/drivers/net/phy/mdio_bus.c
>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  {
>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>> -
>> +	struct phy_device *rv = NULL;
>> +/*
>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>> +*/
>>  	if (!mdiodev)
>>  		return NULL;
>>
>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>  		return NULL;
>>
>> -	return container_of(mdiodev, struct phy_device, mdio);
>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>> +/*
>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>> +		 bus->id, addr, mdiodev, rv);
>> +*/
>> +	return rv;
>>  }
>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>
>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>
>>  	bus->state = MDIOBUS_REGISTERED;
>> -	pr_info("%s: probed\n", bus->name);
>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>  	return 0;
>>
>>  error:
>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>  	while (--i >= 0) {
>>  		mdiodev = bus->mdio_map[i];
>>  		if (!mdiodev)
>> --
>> 2.26.0
>>
>>
>>
>>
>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> Comments inline.
>>>>
>>>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>
>>>>>> The driver installation determination made properly by
>>>>>> checking PHY vs DRIVER id's.
>>>>>> ---
>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>>
>>>>> I would say that most of the code is debug prints.
>>>>>
>>>>
>>>> I tought that they are helpful to keep, they are using the debug calls, so
>>>> they are not visible if user does not like those.
>>>
>>> You are missing the point of who are your users.
>>>
>>> Users want to have working device and the code. They don't need or like
>>> to debug their kernel.
>>>
>>> Thanks
>>>
>>
>> --
>> Br,
>> Lauri J.
> 
>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..5e992f285527 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -61,6 +61,11 @@
>>  #define R8169_MSG_DEFAULT \
>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>
>> +
>> +/*
>> +#define R8169_PROBE_DEBUG
>> +*/
>> +
>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>  #define	MC_FILTER_LIMIT	32
>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	new_bus->write = r8169_mdio_write_reg;
>>
>>  	ret = mdiobus_register(new_bus);
>> +
>> +#ifdef R8169_PROBE_DEBUG
>> +	dev_info(&pdev->dev,
>> +		 "mdiobus_register: %s, %d\n",
>> +		 new_bus->id, ret);
>> +#endif
>>  	if (ret)
>>  		return ret;
>>
>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>> +
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id = tp->phydev->phy_id;
>> +			phydrv_id = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
>> +
>> +#ifdef R8169_PROBE_DEBUG
>> +			dev_debug(&pdev->dev,
>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>> +				new_bus->id, phydev_id, phydev_masked,
>> +				phydrv_id, drv_masked);
>> +#endif
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match = phydev_masked & drv_masked;
>> +			drv_match = drv_match == drv_masked;
>> +
>> +#ifdef R8169_PROBE_DEBUG
>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>> +				  new_bus->id, phydev_match, drv_match);
>> +#endif
>> +
>> +			everything_OK = (phydev_match == drv_match);
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>
>>  	/* PHY will be woken up in rtl_open() */
>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	u64_stats_init(&tp->rx_stats.syncp);
>>  	u64_stats_init(&tp->tx_stats.syncp);
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>> +#endif
>> +
>>  	rtl_init_mac_address(tp);
>>
>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	dev->hw_features |= NETIF_F_RXFCS;
>>
>>  	jumbo_max = rtl_jumbo_max(tp);
>> +
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>> +#endif
>> +
>>  	if (jumbo_max)
>>  		dev->max_mtu = jumbo_max;
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>> +#endif
>> +
>>  	rtl_set_irq_mask(tp);
>>
>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>> +#endif
>> +
>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>  					    &tp->counters_phys_addr,
>>  					    GFP_KERNEL);
>>  	if (!tp->counters)
>>  		return -ENOMEM;
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>> +#endif
>> +
>>  	pci_set_drvdata(pdev, dev);
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>> +#endif
>> +
>>  	rc = r8169_mdio_register(tp);
>> +
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>> +#endif
>> +
>>  	if (rc)
>>  		return rc;
>>
>>  	/* chip gets powered up in rtl_open() */
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>> +#endif
>> +
>>  	rtl_pll_power_down(tp);
>>
>>  	rc = register_netdev(dev);
>> +
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>> +#endif
>> +
>>  	if (rc)
>>  		goto err_mdio_unregister;
>>
>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>  	if (pci_dev_run_wake(pdev))
>>  		pm_runtime_put_sync(&pdev->dev);
>>
>> +#ifdef R816X_PROBE_DEBUG
>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>> +#endif
>> +
>>  	return 0;
>>
>>  err_mdio_unregister:
>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>> index 522760c8bca6..41777f379a57 100644
>> --- a/drivers/net/phy/mdio_bus.c
>> +++ b/drivers/net/phy/mdio_bus.c
>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>  {
>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>> -
>> +	struct phy_device *rv = NULL;
>> +/*
>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>> +*/
>>  	if (!mdiodev)
>>  		return NULL;
>>
>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>  		return NULL;
>>
>> -	return container_of(mdiodev, struct phy_device, mdio);
>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>> +/*
>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>> +		 bus->id, addr, mdiodev, rv);
>> +*/
>> +	return rv;
>>  }
>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>
>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>
>>  	bus->state = MDIOBUS_REGISTERED;
>> -	pr_info("%s: probed\n", bus->name);
>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>  	return 0;
>>
>>  error:
>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>  	while (--i >= 0) {
>>  		mdiodev = bus->mdio_map[i];
>>  		if (!mdiodev)
>> --
>> 2.26.0
>>
> 

From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
 1 file changed, 40 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..da08b1b1047c 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id = tp->phydev->phy_id;
+			phydrv_id = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match = phydev_masked & drv_masked;
+			drv_match = drv_match == drv_masked;
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
-- 
2.26.0



-- 
Br,
Lauri J.

[-- Attachment #2: NET-r8169-module-enchansments.patch --]
[-- Type: text/x-patch, Size: 2460 bytes --]

From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Mon, 13 Apr 2020 13:18:35 +0300
Subject: [PATCH] NET: r8168/r8169 identifying fix

The driver installation determination made properly by
checking PHY vs DRIVER id's.
---
 drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
 1 file changed, 40 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..da08b1b1047c 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 {
 	struct pci_dev *pdev = tp->pci_dev;
 	struct mii_bus *new_bus;
+	u32 phydev_id = 0;
+	u32 phydrv_id = 0;
+	u32 phydrv_id_mask = 0;
 	int ret;
 
 	new_bus = devm_mdiobus_alloc(&pdev->dev);
@@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
 	if (!tp->phydev) {
 		mdiobus_unregister(new_bus);
 		return -ENODEV;
-	} else if (!tp->phydev->drv) {
-		/* Most chip versions fail with the genphy driver.
-		 * Therefore ensure that the dedicated PHY driver is loaded.
-		 */
-		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
-		mdiobus_unregister(new_bus);
-		return -EUNATCH;
+	} else {
+		/* tp -> phydev ok */
+		int everything_OK = 0;
+
+		/* Check driver id versus phy */
+
+		if (tp->phydev->drv) {
+			u32 phydev_masked = 0xBEEFDEAD;
+			u32 drv_masked = ~0;
+			u32 phydev_match = ~0;
+			u32 drv_match = 0xDEADBEEF;
+
+			phydev_id = tp->phydev->phy_id;
+			phydrv_id = tp->phydev->drv->phy_id;
+			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
+
+			drv_masked = phydrv_id & phydrv_id_mask;
+			phydev_masked = phydev_id & phydrv_id_mask;
+
+			phydev_match = phydev_masked & drv_masked;
+			phydev_match = phydev_match == phydev_masked;
+
+			drv_match = phydev_masked & drv_masked;
+			drv_match = drv_match == drv_masked;
+
+			everything_OK = (phydev_match == drv_match);
+		}
+
+		if (!everything_OK) {
+			/* Most chip versions fail with the genphy driver.
+			 * Therefore ensure that the dedicated PHY driver
+			 * is loaded.
+			 */
+			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
+			mdiobus_unregister(new_bus);
+			return -EUNATCH;
+		}
 	}
 
 	/* PHY will be woken up in rtl_open() */
-- 
2.26.0


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 14:44           ` Lauri Jakku
@ 2020-04-13 15:33             ` Heiner Kallweit
  2020-04-13 15:50               ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-13 15:33 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 13.04.2020 16:44, Lauri Jakku wrote:
> Hi,
> 
> On 13.4.2020 15.18, Leon Romanovsky wrote:
>> On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
>>> Hi,
>>>
>>> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>>>
>>
>> Please stop to do top-posting.
> 
>  I did not realise, sorry. Trying to do better.
> 
>>
>>>
>>> Latest version:
>>>
>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>> From: Lauri Jakku <lja@iki.fi>
>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>
>>> The driver installation determination made properly by
>>> checking PHY vs DRIVER id's.
>>> ---
>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>> index bf5bf05970a2..5e992f285527 100644
>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>> @@ -61,6 +61,11 @@
>>>  #define R8169_MSG_DEFAULT \
>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>
>>> +
>>> +/*
>>> +#define R8169_PROBE_DEBUG
>>> +*/
>>
>> Of course not, it is even worse than before.
>> If user recompiles module, he will add prints.
>>
>> Thanks
>>
> 
> New patch at the end and in attachments.
> 
Please do as suggested and read through the kernel developer beginners
guides first. The patch itself, and sending patches as attachments
is not acceptable.

Please provide the requested logs and information first so that we can
understand your issue.

> 
>>> +
>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>  #define	MC_FILTER_LIMIT	32
>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  {
>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>  	struct mii_bus *new_bus;
>>> +	u32 phydev_id = 0;
>>> +	u32 phydrv_id = 0;
>>> +	u32 phydrv_id_mask = 0;
>>>  	int ret;
>>>
>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  	new_bus->write = r8169_mdio_write_reg;
>>>
>>>  	ret = mdiobus_register(new_bus);
>>> +
>>> +#ifdef R8169_PROBE_DEBUG
>>> +	dev_info(&pdev->dev,
>>> +		 "mdiobus_register: %s, %d\n",
>>> +		 new_bus->id, ret);
>>> +#endif
>>>  	if (ret)
>>>  		return ret;
>>>
>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>> +
>>>  	if (!tp->phydev) {
>>>  		mdiobus_unregister(new_bus);
>>>  		return -ENODEV;
>>> -	} else if (!tp->phydev->drv) {
>>> -		/* Most chip versions fail with the genphy driver.
>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>> -		 */
>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> -		mdiobus_unregister(new_bus);
>>> -		return -EUNATCH;
>>> +	} else {
>>> +		/* tp -> phydev ok */
>>> +		int everything_OK = 0;
>>> +
>>> +		/* Check driver id versus phy */
>>> +
>>> +		if (tp->phydev->drv) {
>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>> +			u32 drv_masked = ~0;
>>> +			u32 phydev_match = ~0;
>>> +			u32 drv_match = 0xDEADBEEF;
>>> +
>>> +			phydev_id = tp->phydev->phy_id;
>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>> +
>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>> +
>>> +#ifdef R8169_PROBE_DEBUG
>>> +			dev_debug(&pdev->dev,
>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>> +				new_bus->id, phydev_id, phydev_masked,
>>> +				phydrv_id, drv_masked);
>>> +#endif
>>> +
>>> +			phydev_match = phydev_masked & drv_masked;
>>> +			phydev_match = phydev_match == phydev_masked;
>>> +
>>> +			drv_match = phydev_masked & drv_masked;
>>> +			drv_match = drv_match == drv_masked;
>>> +
>>> +#ifdef R8169_PROBE_DEBUG
>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>> +				  new_bus->id, phydev_match, drv_match);
>>> +#endif
>>> +
>>> +			everything_OK = (phydev_match == drv_match);
>>> +		}
>>> +
>>> +		if (!everything_OK) {
>>> +			/* Most chip versions fail with the genphy driver.
>>> +			 * Therefore ensure that the dedicated PHY driver
>>> +			 * is loaded.
>>> +			 */
>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> +			mdiobus_unregister(new_bus);
>>> +			return -EUNATCH;
>>> +		}
>>>  	}
>>>
>>>  	/* PHY will be woken up in rtl_open() */
>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>> +#endif
>>> +
>>>  	rtl_init_mac_address(tp);
>>>
>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>
>>>  	jumbo_max = rtl_jumbo_max(tp);
>>> +
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>> +#endif
>>> +
>>>  	if (jumbo_max)
>>>  		dev->max_mtu = jumbo_max;
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>> +#endif
>>> +
>>>  	rtl_set_irq_mask(tp);
>>>
>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>> +#endif
>>> +
>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>  					    &tp->counters_phys_addr,
>>>  					    GFP_KERNEL);
>>>  	if (!tp->counters)
>>>  		return -ENOMEM;
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>> +#endif
>>> +
>>>  	pci_set_drvdata(pdev, dev);
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>> +#endif
>>> +
>>>  	rc = r8169_mdio_register(tp);
>>> +
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>> +#endif
>>> +
>>>  	if (rc)
>>>  		return rc;
>>>
>>>  	/* chip gets powered up in rtl_open() */
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>> +#endif
>>> +
>>>  	rtl_pll_power_down(tp);
>>>
>>>  	rc = register_netdev(dev);
>>> +
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>> +#endif
>>> +
>>>  	if (rc)
>>>  		goto err_mdio_unregister;
>>>
>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	if (pci_dev_run_wake(pdev))
>>>  		pm_runtime_put_sync(&pdev->dev);
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>> +#endif
>>> +
>>>  	return 0;
>>>
>>>  err_mdio_unregister:
>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>> index 522760c8bca6..41777f379a57 100644
>>> --- a/drivers/net/phy/mdio_bus.c
>>> +++ b/drivers/net/phy/mdio_bus.c
>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>  {
>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>> -
>>> +	struct phy_device *rv = NULL;
>>> +/*
>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>> +*/
>>>  	if (!mdiodev)
>>>  		return NULL;
>>>
>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>  		return NULL;
>>>
>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>> +/*
>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>> +		 bus->id, addr, mdiodev, rv);
>>> +*/
>>> +	return rv;
>>>  }
>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>
>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>
>>>  	bus->state = MDIOBUS_REGISTERED;
>>> -	pr_info("%s: probed\n", bus->name);
>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>  	return 0;
>>>
>>>  error:
>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>  	while (--i >= 0) {
>>>  		mdiodev = bus->mdio_map[i];
>>>  		if (!mdiodev)
>>> --
>>> 2.26.0
>>>
>>>
>>>
>>>
>>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> Comments inline.
>>>>>
>>>>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>>>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>
>>>>>>> The driver installation determination made properly by
>>>>>>> checking PHY vs DRIVER id's.
>>>>>>> ---
>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>>>
>>>>>> I would say that most of the code is debug prints.
>>>>>>
>>>>>
>>>>> I tought that they are helpful to keep, they are using the debug calls, so
>>>>> they are not visible if user does not like those.
>>>>
>>>> You are missing the point of who are your users.
>>>>
>>>> Users want to have working device and the code. They don't need or like
>>>> to debug their kernel.
>>>>
>>>> Thanks
>>>>
>>>
>>> --
>>> Br,
>>> Lauri J.
>>
>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>> From: Lauri Jakku <lja@iki.fi>
>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>
>>> The driver installation determination made properly by
>>> checking PHY vs DRIVER id's.
>>> ---
>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>> index bf5bf05970a2..5e992f285527 100644
>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>> @@ -61,6 +61,11 @@
>>>  #define R8169_MSG_DEFAULT \
>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>
>>> +
>>> +/*
>>> +#define R8169_PROBE_DEBUG
>>> +*/
>>> +
>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>  #define	MC_FILTER_LIMIT	32
>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  {
>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>  	struct mii_bus *new_bus;
>>> +	u32 phydev_id = 0;
>>> +	u32 phydrv_id = 0;
>>> +	u32 phydrv_id_mask = 0;
>>>  	int ret;
>>>
>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  	new_bus->write = r8169_mdio_write_reg;
>>>
>>>  	ret = mdiobus_register(new_bus);
>>> +
>>> +#ifdef R8169_PROBE_DEBUG
>>> +	dev_info(&pdev->dev,
>>> +		 "mdiobus_register: %s, %d\n",
>>> +		 new_bus->id, ret);
>>> +#endif
>>>  	if (ret)
>>>  		return ret;
>>>
>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>> +
>>>  	if (!tp->phydev) {
>>>  		mdiobus_unregister(new_bus);
>>>  		return -ENODEV;
>>> -	} else if (!tp->phydev->drv) {
>>> -		/* Most chip versions fail with the genphy driver.
>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>> -		 */
>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> -		mdiobus_unregister(new_bus);
>>> -		return -EUNATCH;
>>> +	} else {
>>> +		/* tp -> phydev ok */
>>> +		int everything_OK = 0;
>>> +
>>> +		/* Check driver id versus phy */
>>> +
>>> +		if (tp->phydev->drv) {
>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>> +			u32 drv_masked = ~0;
>>> +			u32 phydev_match = ~0;
>>> +			u32 drv_match = 0xDEADBEEF;
>>> +
>>> +			phydev_id = tp->phydev->phy_id;
>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>> +
>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>> +
>>> +#ifdef R8169_PROBE_DEBUG
>>> +			dev_debug(&pdev->dev,
>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>> +				new_bus->id, phydev_id, phydev_masked,
>>> +				phydrv_id, drv_masked);
>>> +#endif
>>> +
>>> +			phydev_match = phydev_masked & drv_masked;
>>> +			phydev_match = phydev_match == phydev_masked;
>>> +
>>> +			drv_match = phydev_masked & drv_masked;
>>> +			drv_match = drv_match == drv_masked;
>>> +
>>> +#ifdef R8169_PROBE_DEBUG
>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>> +				  new_bus->id, phydev_match, drv_match);
>>> +#endif
>>> +
>>> +			everything_OK = (phydev_match == drv_match);
>>> +		}
>>> +
>>> +		if (!everything_OK) {
>>> +			/* Most chip versions fail with the genphy driver.
>>> +			 * Therefore ensure that the dedicated PHY driver
>>> +			 * is loaded.
>>> +			 */
>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> +			mdiobus_unregister(new_bus);
>>> +			return -EUNATCH;
>>> +		}
>>>  	}
>>>
>>>  	/* PHY will be woken up in rtl_open() */
>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>> +#endif
>>> +
>>>  	rtl_init_mac_address(tp);
>>>
>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>
>>>  	jumbo_max = rtl_jumbo_max(tp);
>>> +
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>> +#endif
>>> +
>>>  	if (jumbo_max)
>>>  		dev->max_mtu = jumbo_max;
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>> +#endif
>>> +
>>>  	rtl_set_irq_mask(tp);
>>>
>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>> +#endif
>>> +
>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>  					    &tp->counters_phys_addr,
>>>  					    GFP_KERNEL);
>>>  	if (!tp->counters)
>>>  		return -ENOMEM;
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>> +#endif
>>> +
>>>  	pci_set_drvdata(pdev, dev);
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>> +#endif
>>> +
>>>  	rc = r8169_mdio_register(tp);
>>> +
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>> +#endif
>>> +
>>>  	if (rc)
>>>  		return rc;
>>>
>>>  	/* chip gets powered up in rtl_open() */
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>> +#endif
>>> +
>>>  	rtl_pll_power_down(tp);
>>>
>>>  	rc = register_netdev(dev);
>>> +
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>> +#endif
>>> +
>>>  	if (rc)
>>>  		goto err_mdio_unregister;
>>>
>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>  	if (pci_dev_run_wake(pdev))
>>>  		pm_runtime_put_sync(&pdev->dev);
>>>
>>> +#ifdef R816X_PROBE_DEBUG
>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>> +#endif
>>> +
>>>  	return 0;
>>>
>>>  err_mdio_unregister:
>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>> index 522760c8bca6..41777f379a57 100644
>>> --- a/drivers/net/phy/mdio_bus.c
>>> +++ b/drivers/net/phy/mdio_bus.c
>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>  {
>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>> -
>>> +	struct phy_device *rv = NULL;
>>> +/*
>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>> +*/
>>>  	if (!mdiodev)
>>>  		return NULL;
>>>
>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>  		return NULL;
>>>
>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>> +/*
>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>> +		 bus->id, addr, mdiodev, rv);
>>> +*/
>>> +	return rv;
>>>  }
>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>
>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>
>>>  	bus->state = MDIOBUS_REGISTERED;
>>> -	pr_info("%s: probed\n", bus->name);
>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>  	return 0;
>>>
>>>  error:
>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>  	while (--i >= 0) {
>>>  		mdiodev = bus->mdio_map[i];
>>>  		if (!mdiodev)
>>> --
>>> 2.26.0
>>>
>>
> 
>>From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
> From: Lauri Jakku <lja@iki.fi>
> Date: Mon, 13 Apr 2020 13:18:35 +0300
> Subject: [PATCH] NET: r8168/r8169 identifying fix
> 
> The driver installation determination made properly by
> checking PHY vs DRIVER id's.
> ---
>  drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
>  1 file changed, 40 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..da08b1b1047c 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  {
>  	struct pci_dev *pdev = tp->pci_dev;
>  	struct mii_bus *new_bus;
> +	u32 phydev_id = 0;
> +	u32 phydrv_id = 0;
> +	u32 phydrv_id_mask = 0;
>  	int ret;
>  
>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
> @@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>  	if (!tp->phydev) {
>  		mdiobus_unregister(new_bus);
>  		return -ENODEV;
> -	} else if (!tp->phydev->drv) {
> -		/* Most chip versions fail with the genphy driver.
> -		 * Therefore ensure that the dedicated PHY driver is loaded.
> -		 */
> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> -		mdiobus_unregister(new_bus);
> -		return -EUNATCH;
> +	} else {
> +		/* tp -> phydev ok */
> +		int everything_OK = 0;
> +
> +		/* Check driver id versus phy */
> +
> +		if (tp->phydev->drv) {
> +			u32 phydev_masked = 0xBEEFDEAD;
> +			u32 drv_masked = ~0;
> +			u32 phydev_match = ~0;
> +			u32 drv_match = 0xDEADBEEF;
> +
> +			phydev_id = tp->phydev->phy_id;
> +			phydrv_id = tp->phydev->drv->phy_id;
> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
> +
> +			drv_masked = phydrv_id & phydrv_id_mask;
> +			phydev_masked = phydev_id & phydrv_id_mask;
> +
> +			phydev_match = phydev_masked & drv_masked;
> +			phydev_match = phydev_match == phydev_masked;
> +
> +			drv_match = phydev_masked & drv_masked;
> +			drv_match = drv_match == drv_masked;
> +
> +			everything_OK = (phydev_match == drv_match);
> +		}
> +
> +		if (!everything_OK) {
> +			/* Most chip versions fail with the genphy driver.
> +			 * Therefore ensure that the dedicated PHY driver
> +			 * is loaded.
> +			 */
> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> +			mdiobus_unregister(new_bus);
> +			return -EUNATCH;
> +		}
>  	}
>  
>  	/* PHY will be woken up in rtl_open() */
> 



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 15:33             ` Heiner Kallweit
@ 2020-04-13 15:50               ` Lauri Jakku
  2020-04-13 15:54                 ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 15:50 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 13.4.2020 18.33, Heiner Kallweit wrote:
> On 13.04.2020 16:44, Lauri Jakku wrote:
>> Hi,
>>
>> On 13.4.2020 15.18, Leon Romanovsky wrote:
>>> On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>>>>
>>>
>>> Please stop to do top-posting.
>>
>>  I did not realise, sorry. Trying to do better.
>>
>>>
>>>>
>>>> Latest version:
>>>>
>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>> From: Lauri Jakku <lja@iki.fi>
>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>
>>>> The driver installation determination made properly by
>>>> checking PHY vs DRIVER id's.
>>>> ---
>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>
>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>> index bf5bf05970a2..5e992f285527 100644
>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>> @@ -61,6 +61,11 @@
>>>>  #define R8169_MSG_DEFAULT \
>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>
>>>> +
>>>> +/*
>>>> +#define R8169_PROBE_DEBUG
>>>> +*/
>>>
>>> Of course not, it is even worse than before.
>>> If user recompiles module, he will add prints.
>>>
>>> Thanks
>>>
>>
>> New patch at the end and in attachments.
>>
> Please do as suggested and read through the kernel developer beginners
> guides first. The patch itself, and sending patches as attachments
> is not acceptable.
> 
> Please provide the requested logs and information first so that we can
> understand your issue.
> 

I'm compiling newest patch + 5.6.2-2 and I'll then provide logs from 
5.3, 5.4 and 5.6 (without and with the patch).

steps i take:
1. power off computer properly
2. take output of dmesg 
3. take output of ip link

Initramfs does have libphy, realtek and r8169 modules added.

This wille take some time, i try to provide the logs today.

Do you like if they are provided in tar per configuration, or 
how ?

I make the commit message have log of problem and log of solution
case.


>>
>>>> +
>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>  #define	MC_FILTER_LIMIT	32
>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  {
>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>  	struct mii_bus *new_bus;
>>>> +	u32 phydev_id = 0;
>>>> +	u32 phydrv_id = 0;
>>>> +	u32 phydrv_id_mask = 0;
>>>>  	int ret;
>>>>
>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>
>>>>  	ret = mdiobus_register(new_bus);
>>>> +
>>>> +#ifdef R8169_PROBE_DEBUG
>>>> +	dev_info(&pdev->dev,
>>>> +		 "mdiobus_register: %s, %d\n",
>>>> +		 new_bus->id, ret);
>>>> +#endif
>>>>  	if (ret)
>>>>  		return ret;
>>>>
>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>> +
>>>>  	if (!tp->phydev) {
>>>>  		mdiobus_unregister(new_bus);
>>>>  		return -ENODEV;
>>>> -	} else if (!tp->phydev->drv) {
>>>> -		/* Most chip versions fail with the genphy driver.
>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>> -		 */
>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> -		mdiobus_unregister(new_bus);
>>>> -		return -EUNATCH;
>>>> +	} else {
>>>> +		/* tp -> phydev ok */
>>>> +		int everything_OK = 0;
>>>> +
>>>> +		/* Check driver id versus phy */
>>>> +
>>>> +		if (tp->phydev->drv) {
>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>> +			u32 drv_masked = ~0;
>>>> +			u32 phydev_match = ~0;
>>>> +			u32 drv_match = 0xDEADBEEF;
>>>> +
>>>> +			phydev_id = tp->phydev->phy_id;
>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>> +
>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>> +
>>>> +#ifdef R8169_PROBE_DEBUG
>>>> +			dev_debug(&pdev->dev,
>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>> +				phydrv_id, drv_masked);
>>>> +#endif
>>>> +
>>>> +			phydev_match = phydev_masked & drv_masked;
>>>> +			phydev_match = phydev_match == phydev_masked;
>>>> +
>>>> +			drv_match = phydev_masked & drv_masked;
>>>> +			drv_match = drv_match == drv_masked;
>>>> +
>>>> +#ifdef R8169_PROBE_DEBUG
>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>> +				  new_bus->id, phydev_match, drv_match);
>>>> +#endif
>>>> +
>>>> +			everything_OK = (phydev_match == drv_match);
>>>> +		}
>>>> +
>>>> +		if (!everything_OK) {
>>>> +			/* Most chip versions fail with the genphy driver.
>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>> +			 * is loaded.
>>>> +			 */
>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> +			mdiobus_unregister(new_bus);
>>>> +			return -EUNATCH;
>>>> +		}
>>>>  	}
>>>>
>>>>  	/* PHY will be woken up in rtl_open() */
>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>> +#endif
>>>> +
>>>>  	rtl_init_mac_address(tp);
>>>>
>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>
>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>> +
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>> +#endif
>>>> +
>>>>  	if (jumbo_max)
>>>>  		dev->max_mtu = jumbo_max;
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>> +#endif
>>>> +
>>>>  	rtl_set_irq_mask(tp);
>>>>
>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>> +#endif
>>>> +
>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>  					    &tp->counters_phys_addr,
>>>>  					    GFP_KERNEL);
>>>>  	if (!tp->counters)
>>>>  		return -ENOMEM;
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>> +#endif
>>>> +
>>>>  	pci_set_drvdata(pdev, dev);
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>> +#endif
>>>> +
>>>>  	rc = r8169_mdio_register(tp);
>>>> +
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>> +#endif
>>>> +
>>>>  	if (rc)
>>>>  		return rc;
>>>>
>>>>  	/* chip gets powered up in rtl_open() */
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>> +#endif
>>>> +
>>>>  	rtl_pll_power_down(tp);
>>>>
>>>>  	rc = register_netdev(dev);
>>>> +
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>> +#endif
>>>> +
>>>>  	if (rc)
>>>>  		goto err_mdio_unregister;
>>>>
>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	if (pci_dev_run_wake(pdev))
>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>> +#endif
>>>> +
>>>>  	return 0;
>>>>
>>>>  err_mdio_unregister:
>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>> index 522760c8bca6..41777f379a57 100644
>>>> --- a/drivers/net/phy/mdio_bus.c
>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>  {
>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>> -
>>>> +	struct phy_device *rv = NULL;
>>>> +/*
>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>> +*/
>>>>  	if (!mdiodev)
>>>>  		return NULL;
>>>>
>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>  		return NULL;
>>>>
>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>> +/*
>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>> +		 bus->id, addr, mdiodev, rv);
>>>> +*/
>>>> +	return rv;
>>>>  }
>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>
>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>
>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>> -	pr_info("%s: probed\n", bus->name);
>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>  	return 0;
>>>>
>>>>  error:
>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>  	while (--i >= 0) {
>>>>  		mdiodev = bus->mdio_map[i];
>>>>  		if (!mdiodev)
>>>> --
>>>> 2.26.0
>>>>
>>>>
>>>>
>>>>
>>>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> Comments inline.
>>>>>>
>>>>>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>>>>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>>>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>>
>>>>>>>> The driver installation determination made properly by
>>>>>>>> checking PHY vs DRIVER id's.
>>>>>>>> ---
>>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>>>>
>>>>>>> I would say that most of the code is debug prints.
>>>>>>>
>>>>>>
>>>>>> I tought that they are helpful to keep, they are using the debug calls, so
>>>>>> they are not visible if user does not like those.
>>>>>
>>>>> You are missing the point of who are your users.
>>>>>
>>>>> Users want to have working device and the code. They don't need or like
>>>>> to debug their kernel.
>>>>>
>>>>> Thanks
>>>>>
>>>>
>>>> --
>>>> Br,
>>>> Lauri J.
>>>
>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>> From: Lauri Jakku <lja@iki.fi>
>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>
>>>> The driver installation determination made properly by
>>>> checking PHY vs DRIVER id's.
>>>> ---
>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>
>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>> index bf5bf05970a2..5e992f285527 100644
>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>> @@ -61,6 +61,11 @@
>>>>  #define R8169_MSG_DEFAULT \
>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>
>>>> +
>>>> +/*
>>>> +#define R8169_PROBE_DEBUG
>>>> +*/
>>>> +
>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>  #define	MC_FILTER_LIMIT	32
>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  {
>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>  	struct mii_bus *new_bus;
>>>> +	u32 phydev_id = 0;
>>>> +	u32 phydrv_id = 0;
>>>> +	u32 phydrv_id_mask = 0;
>>>>  	int ret;
>>>>
>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>
>>>>  	ret = mdiobus_register(new_bus);
>>>> +
>>>> +#ifdef R8169_PROBE_DEBUG
>>>> +	dev_info(&pdev->dev,
>>>> +		 "mdiobus_register: %s, %d\n",
>>>> +		 new_bus->id, ret);
>>>> +#endif
>>>>  	if (ret)
>>>>  		return ret;
>>>>
>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>> +
>>>>  	if (!tp->phydev) {
>>>>  		mdiobus_unregister(new_bus);
>>>>  		return -ENODEV;
>>>> -	} else if (!tp->phydev->drv) {
>>>> -		/* Most chip versions fail with the genphy driver.
>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>> -		 */
>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> -		mdiobus_unregister(new_bus);
>>>> -		return -EUNATCH;
>>>> +	} else {
>>>> +		/* tp -> phydev ok */
>>>> +		int everything_OK = 0;
>>>> +
>>>> +		/* Check driver id versus phy */
>>>> +
>>>> +		if (tp->phydev->drv) {
>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>> +			u32 drv_masked = ~0;
>>>> +			u32 phydev_match = ~0;
>>>> +			u32 drv_match = 0xDEADBEEF;
>>>> +
>>>> +			phydev_id = tp->phydev->phy_id;
>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>> +
>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>> +
>>>> +#ifdef R8169_PROBE_DEBUG
>>>> +			dev_debug(&pdev->dev,
>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>> +				phydrv_id, drv_masked);
>>>> +#endif
>>>> +
>>>> +			phydev_match = phydev_masked & drv_masked;
>>>> +			phydev_match = phydev_match == phydev_masked;
>>>> +
>>>> +			drv_match = phydev_masked & drv_masked;
>>>> +			drv_match = drv_match == drv_masked;
>>>> +
>>>> +#ifdef R8169_PROBE_DEBUG
>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>> +				  new_bus->id, phydev_match, drv_match);
>>>> +#endif
>>>> +
>>>> +			everything_OK = (phydev_match == drv_match);
>>>> +		}
>>>> +
>>>> +		if (!everything_OK) {
>>>> +			/* Most chip versions fail with the genphy driver.
>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>> +			 * is loaded.
>>>> +			 */
>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> +			mdiobus_unregister(new_bus);
>>>> +			return -EUNATCH;
>>>> +		}
>>>>  	}
>>>>
>>>>  	/* PHY will be woken up in rtl_open() */
>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>> +#endif
>>>> +
>>>>  	rtl_init_mac_address(tp);
>>>>
>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>
>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>> +
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>> +#endif
>>>> +
>>>>  	if (jumbo_max)
>>>>  		dev->max_mtu = jumbo_max;
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>> +#endif
>>>> +
>>>>  	rtl_set_irq_mask(tp);
>>>>
>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>> +#endif
>>>> +
>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>  					    &tp->counters_phys_addr,
>>>>  					    GFP_KERNEL);
>>>>  	if (!tp->counters)
>>>>  		return -ENOMEM;
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>> +#endif
>>>> +
>>>>  	pci_set_drvdata(pdev, dev);
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>> +#endif
>>>> +
>>>>  	rc = r8169_mdio_register(tp);
>>>> +
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>> +#endif
>>>> +
>>>>  	if (rc)
>>>>  		return rc;
>>>>
>>>>  	/* chip gets powered up in rtl_open() */
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>> +#endif
>>>> +
>>>>  	rtl_pll_power_down(tp);
>>>>
>>>>  	rc = register_netdev(dev);
>>>> +
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>> +#endif
>>>> +
>>>>  	if (rc)
>>>>  		goto err_mdio_unregister;
>>>>
>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>  	if (pci_dev_run_wake(pdev))
>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>
>>>> +#ifdef R816X_PROBE_DEBUG
>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>> +#endif
>>>> +
>>>>  	return 0;
>>>>
>>>>  err_mdio_unregister:
>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>> index 522760c8bca6..41777f379a57 100644
>>>> --- a/drivers/net/phy/mdio_bus.c
>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>  {
>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>> -
>>>> +	struct phy_device *rv = NULL;
>>>> +/*
>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>> +*/
>>>>  	if (!mdiodev)
>>>>  		return NULL;
>>>>
>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>  		return NULL;
>>>>
>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>> +/*
>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>> +		 bus->id, addr, mdiodev, rv);
>>>> +*/
>>>> +	return rv;
>>>>  }
>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>
>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>
>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>> -	pr_info("%s: probed\n", bus->name);
>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>  	return 0;
>>>>
>>>>  error:
>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>  	while (--i >= 0) {
>>>>  		mdiodev = bus->mdio_map[i];
>>>>  		if (!mdiodev)
>>>> --
>>>> 2.26.0
>>>>
>>>
>>
>> >From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
>> From: Lauri Jakku <lja@iki.fi>
>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>
>> The driver installation determination made properly by
>> checking PHY vs DRIVER id's.
>> ---
>>  drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
>>  1 file changed, 40 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..da08b1b1047c 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  {
>>  	struct pci_dev *pdev = tp->pci_dev;
>>  	struct mii_bus *new_bus;
>> +	u32 phydev_id = 0;
>> +	u32 phydrv_id = 0;
>> +	u32 phydrv_id_mask = 0;
>>  	int ret;
>>  
>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>> @@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>  	if (!tp->phydev) {
>>  		mdiobus_unregister(new_bus);
>>  		return -ENODEV;
>> -	} else if (!tp->phydev->drv) {
>> -		/* Most chip versions fail with the genphy driver.
>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>> -		 */
>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> -		mdiobus_unregister(new_bus);
>> -		return -EUNATCH;
>> +	} else {
>> +		/* tp -> phydev ok */
>> +		int everything_OK = 0;
>> +
>> +		/* Check driver id versus phy */
>> +
>> +		if (tp->phydev->drv) {
>> +			u32 phydev_masked = 0xBEEFDEAD;
>> +			u32 drv_masked = ~0;
>> +			u32 phydev_match = ~0;
>> +			u32 drv_match = 0xDEADBEEF;
>> +
>> +			phydev_id = tp->phydev->phy_id;
>> +			phydrv_id = tp->phydev->drv->phy_id;
>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>> +
>> +			drv_masked = phydrv_id & phydrv_id_mask;
>> +			phydev_masked = phydev_id & phydrv_id_mask;
>> +
>> +			phydev_match = phydev_masked & drv_masked;
>> +			phydev_match = phydev_match == phydev_masked;
>> +
>> +			drv_match = phydev_masked & drv_masked;
>> +			drv_match = drv_match == drv_masked;
>> +
>> +			everything_OK = (phydev_match == drv_match);
>> +		}
>> +
>> +		if (!everything_OK) {
>> +			/* Most chip versions fail with the genphy driver.
>> +			 * Therefore ensure that the dedicated PHY driver
>> +			 * is loaded.
>> +			 */
>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +			mdiobus_unregister(new_bus);
>> +			return -EUNATCH;
>> +		}
>>  	}
>>  
>>  	/* PHY will be woken up in rtl_open() */
>>
> 
> 

-- 
Br,
Lauri J.

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 15:50               ` Lauri Jakku
@ 2020-04-13 15:54                 ` Heiner Kallweit
  2020-04-13 16:10                   ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-13 15:54 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 13.04.2020 17:50, Lauri Jakku wrote:
> Hi,
> 
> On 13.4.2020 18.33, Heiner Kallweit wrote:
>> On 13.04.2020 16:44, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 13.4.2020 15.18, Leon Romanovsky wrote:
>>>> On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>>>>>
>>>>
>>>> Please stop to do top-posting.
>>>
>>>  I did not realise, sorry. Trying to do better.
>>>
>>>>
>>>>>
>>>>> Latest version:
>>>>>
>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>
>>>>> The driver installation determination made properly by
>>>>> checking PHY vs DRIVER id's.
>>>>> ---
>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>
>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> @@ -61,6 +61,11 @@
>>>>>  #define R8169_MSG_DEFAULT \
>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>
>>>>> +
>>>>> +/*
>>>>> +#define R8169_PROBE_DEBUG
>>>>> +*/
>>>>
>>>> Of course not, it is even worse than before.
>>>> If user recompiles module, he will add prints.
>>>>
>>>> Thanks
>>>>
>>>
>>> New patch at the end and in attachments.
>>>
>> Please do as suggested and read through the kernel developer beginners
>> guides first. The patch itself, and sending patches as attachments
>> is not acceptable.
>>
>> Please provide the requested logs and information first so that we can
>> understand your issue.
>>
> 
> I'm compiling newest patch + 5.6.2-2 and I'll then provide logs from 
> 5.3, 5.4 and 5.6 (without and with the patch).
> 
> steps i take:
> 1. power off computer properly
> 2. take output of dmesg 
> 3. take output of ip link
> 
> Initramfs does have libphy, realtek and r8169 modules added.
> 
Typically you need the network driver in initramfs only when
loading rootfs from network, e.g. via NFS.
How is it if you remove r8169 from initramfs?

> This wille take some time, i try to provide the logs today.
> 
> Do you like if they are provided in tar per configuration, or 
> how ?
> 
The logs you can attach to the mail (name them properly).

> I make the commit message have log of problem and log of solution
> case.
> 
> 
>>>
>>>>> +
>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>  #define	MC_FILTER_LIMIT	32
>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>  {
>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>  	struct mii_bus *new_bus;
>>>>> +	u32 phydev_id = 0;
>>>>> +	u32 phydrv_id = 0;
>>>>> +	u32 phydrv_id_mask = 0;
>>>>>  	int ret;
>>>>>
>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>
>>>>>  	ret = mdiobus_register(new_bus);
>>>>> +
>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>> +	dev_info(&pdev->dev,
>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>> +		 new_bus->id, ret);
>>>>> +#endif
>>>>>  	if (ret)
>>>>>  		return ret;
>>>>>
>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>> +
>>>>>  	if (!tp->phydev) {
>>>>>  		mdiobus_unregister(new_bus);
>>>>>  		return -ENODEV;
>>>>> -	} else if (!tp->phydev->drv) {
>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>> -		 */
>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>> -		mdiobus_unregister(new_bus);
>>>>> -		return -EUNATCH;
>>>>> +	} else {
>>>>> +		/* tp -> phydev ok */
>>>>> +		int everything_OK = 0;
>>>>> +
>>>>> +		/* Check driver id versus phy */
>>>>> +
>>>>> +		if (tp->phydev->drv) {
>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>> +			u32 drv_masked = ~0;
>>>>> +			u32 phydev_match = ~0;
>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>> +
>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>> +
>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>> +
>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>> +			dev_debug(&pdev->dev,
>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>> +				phydrv_id, drv_masked);
>>>>> +#endif
>>>>> +
>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>> +
>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>> +			drv_match = drv_match == drv_masked;
>>>>> +
>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>> +#endif
>>>>> +
>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>> +		}
>>>>> +
>>>>> +		if (!everything_OK) {
>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>> +			 * is loaded.
>>>>> +			 */
>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>> +			mdiobus_unregister(new_bus);
>>>>> +			return -EUNATCH;
>>>>> +		}
>>>>>  	}
>>>>>
>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>> +#endif
>>>>> +
>>>>>  	rtl_init_mac_address(tp);
>>>>>
>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>
>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>> +
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>> +#endif
>>>>> +
>>>>>  	if (jumbo_max)
>>>>>  		dev->max_mtu = jumbo_max;
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>> +#endif
>>>>> +
>>>>>  	rtl_set_irq_mask(tp);
>>>>>
>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>> +#endif
>>>>> +
>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>  					    &tp->counters_phys_addr,
>>>>>  					    GFP_KERNEL);
>>>>>  	if (!tp->counters)
>>>>>  		return -ENOMEM;
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>> +#endif
>>>>> +
>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>> +#endif
>>>>> +
>>>>>  	rc = r8169_mdio_register(tp);
>>>>> +
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>> +#endif
>>>>> +
>>>>>  	if (rc)
>>>>>  		return rc;
>>>>>
>>>>>  	/* chip gets powered up in rtl_open() */
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>> +#endif
>>>>> +
>>>>>  	rtl_pll_power_down(tp);
>>>>>
>>>>>  	rc = register_netdev(dev);
>>>>> +
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>> +#endif
>>>>> +
>>>>>  	if (rc)
>>>>>  		goto err_mdio_unregister;
>>>>>
>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>> +#endif
>>>>> +
>>>>>  	return 0;
>>>>>
>>>>>  err_mdio_unregister:
>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>> index 522760c8bca6..41777f379a57 100644
>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>  {
>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>> -
>>>>> +	struct phy_device *rv = NULL;
>>>>> +/*
>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>> +*/
>>>>>  	if (!mdiodev)
>>>>>  		return NULL;
>>>>>
>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>  		return NULL;
>>>>>
>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>> +/*
>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>> +*/
>>>>> +	return rv;
>>>>>  }
>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>
>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>
>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>  	return 0;
>>>>>
>>>>>  error:
>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>  	while (--i >= 0) {
>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>  		if (!mdiodev)
>>>>> --
>>>>> 2.26.0
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> Comments inline.
>>>>>>>
>>>>>>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>>>>>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>>>>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>>>
>>>>>>>>> The driver installation determination made properly by
>>>>>>>>> checking PHY vs DRIVER id's.
>>>>>>>>> ---
>>>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>>>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>>>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>>>>>
>>>>>>>> I would say that most of the code is debug prints.
>>>>>>>>
>>>>>>>
>>>>>>> I tought that they are helpful to keep, they are using the debug calls, so
>>>>>>> they are not visible if user does not like those.
>>>>>>
>>>>>> You are missing the point of who are your users.
>>>>>>
>>>>>> Users want to have working device and the code. They don't need or like
>>>>>> to debug their kernel.
>>>>>>
>>>>>> Thanks
>>>>>>
>>>>>
>>>>> --
>>>>> Br,
>>>>> Lauri J.
>>>>
>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>
>>>>> The driver installation determination made properly by
>>>>> checking PHY vs DRIVER id's.
>>>>> ---
>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>
>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> @@ -61,6 +61,11 @@
>>>>>  #define R8169_MSG_DEFAULT \
>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>
>>>>> +
>>>>> +/*
>>>>> +#define R8169_PROBE_DEBUG
>>>>> +*/
>>>>> +
>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>  #define	MC_FILTER_LIMIT	32
>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>  {
>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>  	struct mii_bus *new_bus;
>>>>> +	u32 phydev_id = 0;
>>>>> +	u32 phydrv_id = 0;
>>>>> +	u32 phydrv_id_mask = 0;
>>>>>  	int ret;
>>>>>
>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>
>>>>>  	ret = mdiobus_register(new_bus);
>>>>> +
>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>> +	dev_info(&pdev->dev,
>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>> +		 new_bus->id, ret);
>>>>> +#endif
>>>>>  	if (ret)
>>>>>  		return ret;
>>>>>
>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>> +
>>>>>  	if (!tp->phydev) {
>>>>>  		mdiobus_unregister(new_bus);
>>>>>  		return -ENODEV;
>>>>> -	} else if (!tp->phydev->drv) {
>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>> -		 */
>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>> -		mdiobus_unregister(new_bus);
>>>>> -		return -EUNATCH;
>>>>> +	} else {
>>>>> +		/* tp -> phydev ok */
>>>>> +		int everything_OK = 0;
>>>>> +
>>>>> +		/* Check driver id versus phy */
>>>>> +
>>>>> +		if (tp->phydev->drv) {
>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>> +			u32 drv_masked = ~0;
>>>>> +			u32 phydev_match = ~0;
>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>> +
>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>> +
>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>> +
>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>> +			dev_debug(&pdev->dev,
>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>> +				phydrv_id, drv_masked);
>>>>> +#endif
>>>>> +
>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>> +
>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>> +			drv_match = drv_match == drv_masked;
>>>>> +
>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>> +#endif
>>>>> +
>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>> +		}
>>>>> +
>>>>> +		if (!everything_OK) {
>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>> +			 * is loaded.
>>>>> +			 */
>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>> +			mdiobus_unregister(new_bus);
>>>>> +			return -EUNATCH;
>>>>> +		}
>>>>>  	}
>>>>>
>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>> +#endif
>>>>> +
>>>>>  	rtl_init_mac_address(tp);
>>>>>
>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>
>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>> +
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>> +#endif
>>>>> +
>>>>>  	if (jumbo_max)
>>>>>  		dev->max_mtu = jumbo_max;
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>> +#endif
>>>>> +
>>>>>  	rtl_set_irq_mask(tp);
>>>>>
>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>> +#endif
>>>>> +
>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>  					    &tp->counters_phys_addr,
>>>>>  					    GFP_KERNEL);
>>>>>  	if (!tp->counters)
>>>>>  		return -ENOMEM;
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>> +#endif
>>>>> +
>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>> +#endif
>>>>> +
>>>>>  	rc = r8169_mdio_register(tp);
>>>>> +
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>> +#endif
>>>>> +
>>>>>  	if (rc)
>>>>>  		return rc;
>>>>>
>>>>>  	/* chip gets powered up in rtl_open() */
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>> +#endif
>>>>> +
>>>>>  	rtl_pll_power_down(tp);
>>>>>
>>>>>  	rc = register_netdev(dev);
>>>>> +
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>> +#endif
>>>>> +
>>>>>  	if (rc)
>>>>>  		goto err_mdio_unregister;
>>>>>
>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>
>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>> +#endif
>>>>> +
>>>>>  	return 0;
>>>>>
>>>>>  err_mdio_unregister:
>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>> index 522760c8bca6..41777f379a57 100644
>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>  {
>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>> -
>>>>> +	struct phy_device *rv = NULL;
>>>>> +/*
>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>> +*/
>>>>>  	if (!mdiodev)
>>>>>  		return NULL;
>>>>>
>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>  		return NULL;
>>>>>
>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>> +/*
>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>> +*/
>>>>> +	return rv;
>>>>>  }
>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>
>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>
>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>  	return 0;
>>>>>
>>>>>  error:
>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>  	while (--i >= 0) {
>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>  		if (!mdiodev)
>>>>> --
>>>>> 2.26.0
>>>>>
>>>>
>>>
>>> >From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
>>> From: Lauri Jakku <lja@iki.fi>
>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>
>>> The driver installation determination made properly by
>>> checking PHY vs DRIVER id's.
>>> ---
>>>  drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
>>>  1 file changed, 40 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>> index bf5bf05970a2..da08b1b1047c 100644
>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  {
>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>  	struct mii_bus *new_bus;
>>> +	u32 phydev_id = 0;
>>> +	u32 phydrv_id = 0;
>>> +	u32 phydrv_id_mask = 0;
>>>  	int ret;
>>>  
>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>> @@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>  	if (!tp->phydev) {
>>>  		mdiobus_unregister(new_bus);
>>>  		return -ENODEV;
>>> -	} else if (!tp->phydev->drv) {
>>> -		/* Most chip versions fail with the genphy driver.
>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>> -		 */
>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> -		mdiobus_unregister(new_bus);
>>> -		return -EUNATCH;
>>> +	} else {
>>> +		/* tp -> phydev ok */
>>> +		int everything_OK = 0;
>>> +
>>> +		/* Check driver id versus phy */
>>> +
>>> +		if (tp->phydev->drv) {
>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>> +			u32 drv_masked = ~0;
>>> +			u32 phydev_match = ~0;
>>> +			u32 drv_match = 0xDEADBEEF;
>>> +
>>> +			phydev_id = tp->phydev->phy_id;
>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>> +
>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>> +
>>> +			phydev_match = phydev_masked & drv_masked;
>>> +			phydev_match = phydev_match == phydev_masked;
>>> +
>>> +			drv_match = phydev_masked & drv_masked;
>>> +			drv_match = drv_match == drv_masked;
>>> +
>>> +			everything_OK = (phydev_match == drv_match);
>>> +		}
>>> +
>>> +		if (!everything_OK) {
>>> +			/* Most chip versions fail with the genphy driver.
>>> +			 * Therefore ensure that the dedicated PHY driver
>>> +			 * is loaded.
>>> +			 */
>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>> +			mdiobus_unregister(new_bus);
>>> +			return -EUNATCH;
>>> +		}
>>>  	}
>>>  
>>>  	/* PHY will be woken up in rtl_open() */
>>>
>>
>>
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 15:54                 ` Heiner Kallweit
@ 2020-04-13 16:10                   ` Lauri Jakku
  2020-04-13 16:17                     ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 16:10 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd



On 13.4.2020 18.54, Heiner Kallweit wrote:
> On 13.04.2020 17:50, Lauri Jakku wrote:
>> Hi,
>>
>> On 13.4.2020 18.33, Heiner Kallweit wrote:
>>> On 13.04.2020 16:44, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 13.4.2020 15.18, Leon Romanovsky wrote:
>>>>> On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>>>>>>
>>>>>
>>>>> Please stop to do top-posting.
>>>>
>>>>  I did not realise, sorry. Trying to do better.
>>>>
>>>>>
>>>>>>
>>>>>> Latest version:
>>>>>>
>>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>
>>>>>> The driver installation determination made properly by
>>>>>> checking PHY vs DRIVER id's.
>>>>>> ---
>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> @@ -61,6 +61,11 @@
>>>>>>  #define R8169_MSG_DEFAULT \
>>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>>
>>>>>> +
>>>>>> +/*
>>>>>> +#define R8169_PROBE_DEBUG
>>>>>> +*/
>>>>>
>>>>> Of course not, it is even worse than before.
>>>>> If user recompiles module, he will add prints.
>>>>>
>>>>> Thanks
>>>>>
>>>>
>>>> New patch at the end and in attachments.
>>>>
>>> Please do as suggested and read through the kernel developer beginners
>>> guides first. The patch itself, and sending patches as attachments
>>> is not acceptable.
>>>
>>> Please provide the requested logs and information first so that we can
>>> understand your issue.
>>>
>>
>> I'm compiling newest patch + 5.6.2-2 and I'll then provide logs from 
>> 5.3, 5.4 and 5.6 (without and with the patch).
>>
>> steps i take:
>> 1. power off computer properly
>> 2. take output of dmesg 
>> 3. take output of ip link
>>
>> Initramfs does have libphy, realtek and r8169 modules added.
>>
> Typically you need the network driver in initramfs only when
> loading rootfs from network, e.g. via NFS.
> How is it if you remove r8169 from initramfs?
> 

Hmm, I'm using Manjaro's stock packages for kernels. I removed
modules from mkinitcpio.conf and reinstall the kernel packages
now. i check do they have them installed. 

One thing that I've noticed is that when doing reboot, the NIC
does not work ok after bootup. When proper power cycle is done
the NIC works ok, so something in settings etc. does not reset.

I check if adding reset to driver's probe helps, i make diffrent
patch for that.


>> This wille take some time, i try to provide the logs today.
>>
>> Do you like if they are provided in tar per configuration, or 
>> how ?
>>
> The logs you can attach to the mail (name them properly).
> 
>> I make the commit message have log of problem and log of solution
>> case.
>>
>>
>>>>
>>>>>> +
>>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>>  #define	MC_FILTER_LIMIT	32
>>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>  {
>>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>>  	struct mii_bus *new_bus;
>>>>>> +	u32 phydev_id = 0;
>>>>>> +	u32 phydrv_id = 0;
>>>>>> +	u32 phydrv_id_mask = 0;
>>>>>>  	int ret;
>>>>>>
>>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>>
>>>>>>  	ret = mdiobus_register(new_bus);
>>>>>> +
>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>> +	dev_info(&pdev->dev,
>>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>>> +		 new_bus->id, ret);
>>>>>> +#endif
>>>>>>  	if (ret)
>>>>>>  		return ret;
>>>>>>
>>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>>> +
>>>>>>  	if (!tp->phydev) {
>>>>>>  		mdiobus_unregister(new_bus);
>>>>>>  		return -ENODEV;
>>>>>> -	} else if (!tp->phydev->drv) {
>>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>> -		 */
>>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>> -		mdiobus_unregister(new_bus);
>>>>>> -		return -EUNATCH;
>>>>>> +	} else {
>>>>>> +		/* tp -> phydev ok */
>>>>>> +		int everything_OK = 0;
>>>>>> +
>>>>>> +		/* Check driver id versus phy */
>>>>>> +
>>>>>> +		if (tp->phydev->drv) {
>>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>>> +			u32 drv_masked = ~0;
>>>>>> +			u32 phydev_match = ~0;
>>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>>> +
>>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>>> +
>>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>>> +
>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>> +			dev_debug(&pdev->dev,
>>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>>> +				phydrv_id, drv_masked);
>>>>>> +#endif
>>>>>> +
>>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>>> +
>>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>>> +			drv_match = drv_match == drv_masked;
>>>>>> +
>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>>> +#endif
>>>>>> +
>>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>>> +		}
>>>>>> +
>>>>>> +		if (!everything_OK) {
>>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>>> +			 * is loaded.
>>>>>> +			 */
>>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>> +			mdiobus_unregister(new_bus);
>>>>>> +			return -EUNATCH;
>>>>>> +		}
>>>>>>  	}
>>>>>>
>>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rtl_init_mac_address(tp);
>>>>>>
>>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>>
>>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>>> +
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>>> +#endif
>>>>>> +
>>>>>>  	if (jumbo_max)
>>>>>>  		dev->max_mtu = jumbo_max;
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rtl_set_irq_mask(tp);
>>>>>>
>>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>>> +#endif
>>>>>> +
>>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>>  					    &tp->counters_phys_addr,
>>>>>>  					    GFP_KERNEL);
>>>>>>  	if (!tp->counters)
>>>>>>  		return -ENOMEM;
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rc = r8169_mdio_register(tp);
>>>>>> +
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>>> +#endif
>>>>>> +
>>>>>>  	if (rc)
>>>>>>  		return rc;
>>>>>>
>>>>>>  	/* chip gets powered up in rtl_open() */
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rtl_pll_power_down(tp);
>>>>>>
>>>>>>  	rc = register_netdev(dev);
>>>>>> +
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>>> +#endif
>>>>>> +
>>>>>>  	if (rc)
>>>>>>  		goto err_mdio_unregister;
>>>>>>
>>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	return 0;
>>>>>>
>>>>>>  err_mdio_unregister:
>>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>>> index 522760c8bca6..41777f379a57 100644
>>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>>  {
>>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>>> -
>>>>>> +	struct phy_device *rv = NULL;
>>>>>> +/*
>>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>>> +*/
>>>>>>  	if (!mdiodev)
>>>>>>  		return NULL;
>>>>>>
>>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>>  		return NULL;
>>>>>>
>>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>>> +/*
>>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>>> +*/
>>>>>> +	return rv;
>>>>>>  }
>>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>>
>>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>>
>>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>>  	return 0;
>>>>>>
>>>>>>  error:
>>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>>  	while (--i >= 0) {
>>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>>  		if (!mdiodev)
>>>>>> --
>>>>>> 2.26.0
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> Comments inline.
>>>>>>>>
>>>>>>>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>>>>>>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>>>>>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>>>>
>>>>>>>>>> The driver installation determination made properly by
>>>>>>>>>> checking PHY vs DRIVER id's.
>>>>>>>>>> ---
>>>>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>>>>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>>>>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>>>>>>
>>>>>>>>> I would say that most of the code is debug prints.
>>>>>>>>>
>>>>>>>>
>>>>>>>> I tought that they are helpful to keep, they are using the debug calls, so
>>>>>>>> they are not visible if user does not like those.
>>>>>>>
>>>>>>> You are missing the point of who are your users.
>>>>>>>
>>>>>>> Users want to have working device and the code. They don't need or like
>>>>>>> to debug their kernel.
>>>>>>>
>>>>>>> Thanks
>>>>>>>
>>>>>>
>>>>>> --
>>>>>> Br,
>>>>>> Lauri J.
>>>>>
>>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>
>>>>>> The driver installation determination made properly by
>>>>>> checking PHY vs DRIVER id's.
>>>>>> ---
>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> @@ -61,6 +61,11 @@
>>>>>>  #define R8169_MSG_DEFAULT \
>>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>>
>>>>>> +
>>>>>> +/*
>>>>>> +#define R8169_PROBE_DEBUG
>>>>>> +*/
>>>>>> +
>>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>>  #define	MC_FILTER_LIMIT	32
>>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>  {
>>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>>  	struct mii_bus *new_bus;
>>>>>> +	u32 phydev_id = 0;
>>>>>> +	u32 phydrv_id = 0;
>>>>>> +	u32 phydrv_id_mask = 0;
>>>>>>  	int ret;
>>>>>>
>>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>>
>>>>>>  	ret = mdiobus_register(new_bus);
>>>>>> +
>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>> +	dev_info(&pdev->dev,
>>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>>> +		 new_bus->id, ret);
>>>>>> +#endif
>>>>>>  	if (ret)
>>>>>>  		return ret;
>>>>>>
>>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>>> +
>>>>>>  	if (!tp->phydev) {
>>>>>>  		mdiobus_unregister(new_bus);
>>>>>>  		return -ENODEV;
>>>>>> -	} else if (!tp->phydev->drv) {
>>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>> -		 */
>>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>> -		mdiobus_unregister(new_bus);
>>>>>> -		return -EUNATCH;
>>>>>> +	} else {
>>>>>> +		/* tp -> phydev ok */
>>>>>> +		int everything_OK = 0;
>>>>>> +
>>>>>> +		/* Check driver id versus phy */
>>>>>> +
>>>>>> +		if (tp->phydev->drv) {
>>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>>> +			u32 drv_masked = ~0;
>>>>>> +			u32 phydev_match = ~0;
>>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>>> +
>>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>>> +
>>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>>> +
>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>> +			dev_debug(&pdev->dev,
>>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>>> +				phydrv_id, drv_masked);
>>>>>> +#endif
>>>>>> +
>>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>>> +
>>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>>> +			drv_match = drv_match == drv_masked;
>>>>>> +
>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>>> +#endif
>>>>>> +
>>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>>> +		}
>>>>>> +
>>>>>> +		if (!everything_OK) {
>>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>>> +			 * is loaded.
>>>>>> +			 */
>>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>> +			mdiobus_unregister(new_bus);
>>>>>> +			return -EUNATCH;
>>>>>> +		}
>>>>>>  	}
>>>>>>
>>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rtl_init_mac_address(tp);
>>>>>>
>>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>>
>>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>>> +
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>>> +#endif
>>>>>> +
>>>>>>  	if (jumbo_max)
>>>>>>  		dev->max_mtu = jumbo_max;
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rtl_set_irq_mask(tp);
>>>>>>
>>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>>> +#endif
>>>>>> +
>>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>>  					    &tp->counters_phys_addr,
>>>>>>  					    GFP_KERNEL);
>>>>>>  	if (!tp->counters)
>>>>>>  		return -ENOMEM;
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rc = r8169_mdio_register(tp);
>>>>>> +
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>>> +#endif
>>>>>> +
>>>>>>  	if (rc)
>>>>>>  		return rc;
>>>>>>
>>>>>>  	/* chip gets powered up in rtl_open() */
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	rtl_pll_power_down(tp);
>>>>>>
>>>>>>  	rc = register_netdev(dev);
>>>>>> +
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>>> +#endif
>>>>>> +
>>>>>>  	if (rc)
>>>>>>  		goto err_mdio_unregister;
>>>>>>
>>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>>
>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>>> +#endif
>>>>>> +
>>>>>>  	return 0;
>>>>>>
>>>>>>  err_mdio_unregister:
>>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>>> index 522760c8bca6..41777f379a57 100644
>>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>>  {
>>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>>> -
>>>>>> +	struct phy_device *rv = NULL;
>>>>>> +/*
>>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>>> +*/
>>>>>>  	if (!mdiodev)
>>>>>>  		return NULL;
>>>>>>
>>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>>  		return NULL;
>>>>>>
>>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>>> +/*
>>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>>> +*/
>>>>>> +	return rv;
>>>>>>  }
>>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>>
>>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>>
>>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>>  	return 0;
>>>>>>
>>>>>>  error:
>>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>>  	while (--i >= 0) {
>>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>>  		if (!mdiodev)
>>>>>> --
>>>>>> 2.26.0
>>>>>>
>>>>>
>>>>
>>>> >From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
>>>> From: Lauri Jakku <lja@iki.fi>
>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>
>>>> The driver installation determination made properly by
>>>> checking PHY vs DRIVER id's.
>>>> ---
>>>>  drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
>>>>  1 file changed, 40 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>> index bf5bf05970a2..da08b1b1047c 100644
>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  {
>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>  	struct mii_bus *new_bus;
>>>> +	u32 phydev_id = 0;
>>>> +	u32 phydrv_id = 0;
>>>> +	u32 phydrv_id_mask = 0;
>>>>  	int ret;
>>>>  
>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>> @@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>  	if (!tp->phydev) {
>>>>  		mdiobus_unregister(new_bus);
>>>>  		return -ENODEV;
>>>> -	} else if (!tp->phydev->drv) {
>>>> -		/* Most chip versions fail with the genphy driver.
>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>> -		 */
>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> -		mdiobus_unregister(new_bus);
>>>> -		return -EUNATCH;
>>>> +	} else {
>>>> +		/* tp -> phydev ok */
>>>> +		int everything_OK = 0;
>>>> +
>>>> +		/* Check driver id versus phy */
>>>> +
>>>> +		if (tp->phydev->drv) {
>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>> +			u32 drv_masked = ~0;
>>>> +			u32 phydev_match = ~0;
>>>> +			u32 drv_match = 0xDEADBEEF;
>>>> +
>>>> +			phydev_id = tp->phydev->phy_id;
>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>> +
>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>> +
>>>> +			phydev_match = phydev_masked & drv_masked;
>>>> +			phydev_match = phydev_match == phydev_masked;
>>>> +
>>>> +			drv_match = phydev_masked & drv_masked;
>>>> +			drv_match = drv_match == drv_masked;
>>>> +
>>>> +			everything_OK = (phydev_match == drv_match);
>>>> +		}
>>>> +
>>>> +		if (!everything_OK) {
>>>> +			/* Most chip versions fail with the genphy driver.
>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>> +			 * is loaded.
>>>> +			 */
>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> +			mdiobus_unregister(new_bus);
>>>> +			return -EUNATCH;
>>>> +		}
>>>>  	}
>>>>  
>>>>  	/* PHY will be woken up in rtl_open() */
>>>>
>>>
>>>
>>
> 

-- 
Br,
Lauri J.

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 16:10                   ` Lauri Jakku
@ 2020-04-13 16:17                     ` Heiner Kallweit
  2020-04-13 16:37                       ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-13 16:17 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 13.04.2020 18:10, Lauri Jakku wrote:
> 
> 
> On 13.4.2020 18.54, Heiner Kallweit wrote:
>> On 13.04.2020 17:50, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 13.4.2020 18.33, Heiner Kallweit wrote:
>>>> On 13.04.2020 16:44, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 13.4.2020 15.18, Leon Romanovsky wrote:
>>>>>> On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>>>>>>>
>>>>>>
>>>>>> Please stop to do top-posting.
>>>>>
>>>>>  I did not realise, sorry. Trying to do better.
>>>>>
>>>>>>
>>>>>>>
>>>>>>> Latest version:
>>>>>>>
>>>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>
>>>>>>> The driver installation determination made properly by
>>>>>>> checking PHY vs DRIVER id's.
>>>>>>> ---
>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>>>
>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> @@ -61,6 +61,11 @@
>>>>>>>  #define R8169_MSG_DEFAULT \
>>>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>>>
>>>>>>> +
>>>>>>> +/*
>>>>>>> +#define R8169_PROBE_DEBUG
>>>>>>> +*/
>>>>>>
>>>>>> Of course not, it is even worse than before.
>>>>>> If user recompiles module, he will add prints.
>>>>>>
>>>>>> Thanks
>>>>>>
>>>>>
>>>>> New patch at the end and in attachments.
>>>>>
>>>> Please do as suggested and read through the kernel developer beginners
>>>> guides first. The patch itself, and sending patches as attachments
>>>> is not acceptable.
>>>>
>>>> Please provide the requested logs and information first so that we can
>>>> understand your issue.
>>>>
>>>
>>> I'm compiling newest patch + 5.6.2-2 and I'll then provide logs from 
>>> 5.3, 5.4 and 5.6 (without and with the patch).
>>>
>>> steps i take:
>>> 1. power off computer properly
>>> 2. take output of dmesg 
>>> 3. take output of ip link
>>>
>>> Initramfs does have libphy, realtek and r8169 modules added.
>>>
>> Typically you need the network driver in initramfs only when
>> loading rootfs from network, e.g. via NFS.
>> How is it if you remove r8169 from initramfs?
>>
> 
> Hmm, I'm using Manjaro's stock packages for kernels. I removed
> modules from mkinitcpio.conf and reinstall the kernel packages
> now. i check do they have them installed. 
> 
> One thing that I've noticed is that when doing reboot, the NIC
> does not work ok after bootup. When proper power cycle is done
> the NIC works ok, so something in settings etc. does not reset.
> 
> I check if adding reset to driver's probe helps, i make diffrent
> patch for that.
> 
No need for such a patch, probe is resetting the chip already.
Best provide a dmesg log of such a reboot.
The issue can also be caused by a BIOS bug, as I don't see
this behavior on my systems.

> 
>>> This wille take some time, i try to provide the logs today.
>>>
>>> Do you like if they are provided in tar per configuration, or 
>>> how ?
>>>
>> The logs you can attach to the mail (name them properly).
>>
>>> I make the commit message have log of problem and log of solution
>>> case.
>>>
>>>
>>>>>
>>>>>>> +
>>>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>>>  #define	MC_FILTER_LIMIT	32
>>>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>  {
>>>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>>>  	struct mii_bus *new_bus;
>>>>>>> +	u32 phydev_id = 0;
>>>>>>> +	u32 phydrv_id = 0;
>>>>>>> +	u32 phydrv_id_mask = 0;
>>>>>>>  	int ret;
>>>>>>>
>>>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>>>
>>>>>>>  	ret = mdiobus_register(new_bus);
>>>>>>> +
>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>> +	dev_info(&pdev->dev,
>>>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>>>> +		 new_bus->id, ret);
>>>>>>> +#endif
>>>>>>>  	if (ret)
>>>>>>>  		return ret;
>>>>>>>
>>>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>>>> +
>>>>>>>  	if (!tp->phydev) {
>>>>>>>  		mdiobus_unregister(new_bus);
>>>>>>>  		return -ENODEV;
>>>>>>> -	} else if (!tp->phydev->drv) {
>>>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>> -		 */
>>>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>> -		mdiobus_unregister(new_bus);
>>>>>>> -		return -EUNATCH;
>>>>>>> +	} else {
>>>>>>> +		/* tp -> phydev ok */
>>>>>>> +		int everything_OK = 0;
>>>>>>> +
>>>>>>> +		/* Check driver id versus phy */
>>>>>>> +
>>>>>>> +		if (tp->phydev->drv) {
>>>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>>>> +			u32 drv_masked = ~0;
>>>>>>> +			u32 phydev_match = ~0;
>>>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>>>> +
>>>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>>>> +
>>>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>>>> +
>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>> +			dev_debug(&pdev->dev,
>>>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>>>> +				phydrv_id, drv_masked);
>>>>>>> +#endif
>>>>>>> +
>>>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>>>> +
>>>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>>>> +			drv_match = drv_match == drv_masked;
>>>>>>> +
>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>>>> +#endif
>>>>>>> +
>>>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>>>> +		}
>>>>>>> +
>>>>>>> +		if (!everything_OK) {
>>>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>>>> +			 * is loaded.
>>>>>>> +			 */
>>>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>> +			mdiobus_unregister(new_bus);
>>>>>>> +			return -EUNATCH;
>>>>>>> +		}
>>>>>>>  	}
>>>>>>>
>>>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rtl_init_mac_address(tp);
>>>>>>>
>>>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>>>
>>>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>>>> +
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	if (jumbo_max)
>>>>>>>  		dev->max_mtu = jumbo_max;
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rtl_set_irq_mask(tp);
>>>>>>>
>>>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>>>  					    &tp->counters_phys_addr,
>>>>>>>  					    GFP_KERNEL);
>>>>>>>  	if (!tp->counters)
>>>>>>>  		return -ENOMEM;
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rc = r8169_mdio_register(tp);
>>>>>>> +
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	if (rc)
>>>>>>>  		return rc;
>>>>>>>
>>>>>>>  	/* chip gets powered up in rtl_open() */
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rtl_pll_power_down(tp);
>>>>>>>
>>>>>>>  	rc = register_netdev(dev);
>>>>>>> +
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	if (rc)
>>>>>>>  		goto err_mdio_unregister;
>>>>>>>
>>>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	return 0;
>>>>>>>
>>>>>>>  err_mdio_unregister:
>>>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>>>> index 522760c8bca6..41777f379a57 100644
>>>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>>>  {
>>>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>>>> -
>>>>>>> +	struct phy_device *rv = NULL;
>>>>>>> +/*
>>>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>>>> +*/
>>>>>>>  	if (!mdiodev)
>>>>>>>  		return NULL;
>>>>>>>
>>>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>>>  		return NULL;
>>>>>>>
>>>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>>>> +/*
>>>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>>>> +*/
>>>>>>> +	return rv;
>>>>>>>  }
>>>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>>>
>>>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>>>
>>>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>>>  	return 0;
>>>>>>>
>>>>>>>  error:
>>>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>>>  	while (--i >= 0) {
>>>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>>>  		if (!mdiodev)
>>>>>>> --
>>>>>>> 2.26.0
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> Comments inline.
>>>>>>>>>
>>>>>>>>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>>>>>>>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>>>>>>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>>>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>>>>>
>>>>>>>>>>> The driver installation determination made properly by
>>>>>>>>>>> checking PHY vs DRIVER id's.
>>>>>>>>>>> ---
>>>>>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>>>>>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>>>>>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>>>>>>>
>>>>>>>>>> I would say that most of the code is debug prints.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> I tought that they are helpful to keep, they are using the debug calls, so
>>>>>>>>> they are not visible if user does not like those.
>>>>>>>>
>>>>>>>> You are missing the point of who are your users.
>>>>>>>>
>>>>>>>> Users want to have working device and the code. They don't need or like
>>>>>>>> to debug their kernel.
>>>>>>>>
>>>>>>>> Thanks
>>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> Br,
>>>>>>> Lauri J.
>>>>>>
>>>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>
>>>>>>> The driver installation determination made properly by
>>>>>>> checking PHY vs DRIVER id's.
>>>>>>> ---
>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>>>
>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> @@ -61,6 +61,11 @@
>>>>>>>  #define R8169_MSG_DEFAULT \
>>>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>>>
>>>>>>> +
>>>>>>> +/*
>>>>>>> +#define R8169_PROBE_DEBUG
>>>>>>> +*/
>>>>>>> +
>>>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>>>  #define	MC_FILTER_LIMIT	32
>>>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>  {
>>>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>>>  	struct mii_bus *new_bus;
>>>>>>> +	u32 phydev_id = 0;
>>>>>>> +	u32 phydrv_id = 0;
>>>>>>> +	u32 phydrv_id_mask = 0;
>>>>>>>  	int ret;
>>>>>>>
>>>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>>>
>>>>>>>  	ret = mdiobus_register(new_bus);
>>>>>>> +
>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>> +	dev_info(&pdev->dev,
>>>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>>>> +		 new_bus->id, ret);
>>>>>>> +#endif
>>>>>>>  	if (ret)
>>>>>>>  		return ret;
>>>>>>>
>>>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>>>> +
>>>>>>>  	if (!tp->phydev) {
>>>>>>>  		mdiobus_unregister(new_bus);
>>>>>>>  		return -ENODEV;
>>>>>>> -	} else if (!tp->phydev->drv) {
>>>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>> -		 */
>>>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>> -		mdiobus_unregister(new_bus);
>>>>>>> -		return -EUNATCH;
>>>>>>> +	} else {
>>>>>>> +		/* tp -> phydev ok */
>>>>>>> +		int everything_OK = 0;
>>>>>>> +
>>>>>>> +		/* Check driver id versus phy */
>>>>>>> +
>>>>>>> +		if (tp->phydev->drv) {
>>>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>>>> +			u32 drv_masked = ~0;
>>>>>>> +			u32 phydev_match = ~0;
>>>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>>>> +
>>>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>>>> +
>>>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>>>> +
>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>> +			dev_debug(&pdev->dev,
>>>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>>>> +				phydrv_id, drv_masked);
>>>>>>> +#endif
>>>>>>> +
>>>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>>>> +
>>>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>>>> +			drv_match = drv_match == drv_masked;
>>>>>>> +
>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>>>> +#endif
>>>>>>> +
>>>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>>>> +		}
>>>>>>> +
>>>>>>> +		if (!everything_OK) {
>>>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>>>> +			 * is loaded.
>>>>>>> +			 */
>>>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>> +			mdiobus_unregister(new_bus);
>>>>>>> +			return -EUNATCH;
>>>>>>> +		}
>>>>>>>  	}
>>>>>>>
>>>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rtl_init_mac_address(tp);
>>>>>>>
>>>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>>>
>>>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>>>> +
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	if (jumbo_max)
>>>>>>>  		dev->max_mtu = jumbo_max;
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rtl_set_irq_mask(tp);
>>>>>>>
>>>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>>>  					    &tp->counters_phys_addr,
>>>>>>>  					    GFP_KERNEL);
>>>>>>>  	if (!tp->counters)
>>>>>>>  		return -ENOMEM;
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rc = r8169_mdio_register(tp);
>>>>>>> +
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	if (rc)
>>>>>>>  		return rc;
>>>>>>>
>>>>>>>  	/* chip gets powered up in rtl_open() */
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	rtl_pll_power_down(tp);
>>>>>>>
>>>>>>>  	rc = register_netdev(dev);
>>>>>>> +
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	if (rc)
>>>>>>>  		goto err_mdio_unregister;
>>>>>>>
>>>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>>>
>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  	return 0;
>>>>>>>
>>>>>>>  err_mdio_unregister:
>>>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>>>> index 522760c8bca6..41777f379a57 100644
>>>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>>>  {
>>>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>>>> -
>>>>>>> +	struct phy_device *rv = NULL;
>>>>>>> +/*
>>>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>>>> +*/
>>>>>>>  	if (!mdiodev)
>>>>>>>  		return NULL;
>>>>>>>
>>>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>>>  		return NULL;
>>>>>>>
>>>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>>>> +/*
>>>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>>>> +*/
>>>>>>> +	return rv;
>>>>>>>  }
>>>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>>>
>>>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>>>
>>>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>>>  	return 0;
>>>>>>>
>>>>>>>  error:
>>>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>>>  	while (--i >= 0) {
>>>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>>>  		if (!mdiodev)
>>>>>>> --
>>>>>>> 2.26.0
>>>>>>>
>>>>>>
>>>>>
>>>>> >From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>
>>>>> The driver installation determination made properly by
>>>>> checking PHY vs DRIVER id's.
>>>>> ---
>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
>>>>>  1 file changed, 40 insertions(+), 7 deletions(-)
>>>>>
>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> index bf5bf05970a2..da08b1b1047c 100644
>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>  {
>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>  	struct mii_bus *new_bus;
>>>>> +	u32 phydev_id = 0;
>>>>> +	u32 phydrv_id = 0;
>>>>> +	u32 phydrv_id_mask = 0;
>>>>>  	int ret;
>>>>>  
>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>> @@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>  	if (!tp->phydev) {
>>>>>  		mdiobus_unregister(new_bus);
>>>>>  		return -ENODEV;
>>>>> -	} else if (!tp->phydev->drv) {
>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>> -		 */
>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>> -		mdiobus_unregister(new_bus);
>>>>> -		return -EUNATCH;
>>>>> +	} else {
>>>>> +		/* tp -> phydev ok */
>>>>> +		int everything_OK = 0;
>>>>> +
>>>>> +		/* Check driver id versus phy */
>>>>> +
>>>>> +		if (tp->phydev->drv) {
>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>> +			u32 drv_masked = ~0;
>>>>> +			u32 phydev_match = ~0;
>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>> +
>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>> +
>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>> +
>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>> +
>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>> +			drv_match = drv_match == drv_masked;
>>>>> +
>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>> +		}
>>>>> +
>>>>> +		if (!everything_OK) {
>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>> +			 * is loaded.
>>>>> +			 */
>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>> +			mdiobus_unregister(new_bus);
>>>>> +			return -EUNATCH;
>>>>> +		}
>>>>>  	}
>>>>>  
>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>
>>>>
>>>>
>>>
>>
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-13 16:17                     ` Heiner Kallweit
@ 2020-04-13 16:37                       ` Lauri Jakku
  0 siblings, 0 replies; 57+ messages in thread
From: Lauri Jakku @ 2020-04-13 16:37 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 13.4.2020 19.17, Heiner Kallweit wrote:
> On 13.04.2020 18:10, Lauri Jakku wrote:
>>
>>
>> On 13.4.2020 18.54, Heiner Kallweit wrote:
>>> On 13.04.2020 17:50, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 13.4.2020 18.33, Heiner Kallweit wrote:
>>>>> On 13.04.2020 16:44, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 13.4.2020 15.18, Leon Romanovsky wrote:
>>>>>>> On Mon, Apr 13, 2020 at 03:01:23PM +0300, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> I've tought that the debug's are worth to save behind an definition/commented out, so they can be enabled if needed.
>>>>>>>>
>>>>>>>
>>>>>>> Please stop to do top-posting.
>>>>>>
>>>>>>  I did not realise, sorry. Trying to do better.
>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>> Latest version:
>>>>>>>>
>>>>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>>
>>>>>>>> The driver installation determination made properly by
>>>>>>>> checking PHY vs DRIVER id's.
>>>>>>>> ---
>>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> @@ -61,6 +61,11 @@
>>>>>>>>  #define R8169_MSG_DEFAULT \
>>>>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>>>>
>>>>>>>> +
>>>>>>>> +/*
>>>>>>>> +#define R8169_PROBE_DEBUG
>>>>>>>> +*/
>>>>>>>
>>>>>>> Of course not, it is even worse than before.
>>>>>>> If user recompiles module, he will add prints.
>>>>>>>
>>>>>>> Thanks
>>>>>>>
>>>>>>
>>>>>> New patch at the end and in attachments.
>>>>>>
>>>>> Please do as suggested and read through the kernel developer beginners
>>>>> guides first. The patch itself, and sending patches as attachments
>>>>> is not acceptable.
>>>>>
>>>>> Please provide the requested logs and information first so that we can
>>>>> understand your issue.
>>>>>
>>>>
>>>> I'm compiling newest patch + 5.6.2-2 and I'll then provide logs from 
>>>> 5.3, 5.4 and 5.6 (without and with the patch).
>>>>
>>>> steps i take:
>>>> 1. power off computer properly
>>>> 2. take output of dmesg 
>>>> 3. take output of ip link
>>>>
>>>> Initramfs does have libphy, realtek and r8169 modules added.
>>>>
>>> Typically you need the network driver in initramfs only when
>>> loading rootfs from network, e.g. via NFS.
>>> How is it if you remove r8169 from initramfs?
>>>
>>
>> Hmm, I'm using Manjaro's stock packages for kernels. I removed
>> modules from mkinitcpio.conf and reinstall the kernel packages
>> now. i check do they have them installed. 
>>
>> One thing that I've noticed is that when doing reboot, the NIC
>> does not work ok after bootup. When proper power cycle is done
>> the NIC works ok, so something in settings etc. does not reset.
>>
>> I check if adding reset to driver's probe helps, i make diffrent
>> patch for that.
>>
> No need for such a patch, probe is resetting the chip already.
> Best provide a dmesg log of such a reboot.
> The issue can also be caused by a BIOS bug, as I don't see
> this behavior on my systems.
> 

Ok, i'll do same runs with soft-reboot and power cycled ones.

Now I dont have r8169 driver in ramfs. 

I'll start testing & log gathering ASAP.

>>
>>>> This wille take some time, i try to provide the logs today.
>>>>
>>>> Do you like if they are provided in tar per configuration, or 
>>>> how ?
>>>>
>>> The logs you can attach to the mail (name them properly).
>>>
>>>> I make the commit message have log of problem and log of solution
>>>> case.
>>>>
>>>>
>>>>>>
>>>>>>>> +
>>>>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>>>>  #define	MC_FILTER_LIMIT	32
>>>>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>  {
>>>>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>>>>  	struct mii_bus *new_bus;
>>>>>>>> +	u32 phydev_id = 0;
>>>>>>>> +	u32 phydrv_id = 0;
>>>>>>>> +	u32 phydrv_id_mask = 0;
>>>>>>>>  	int ret;
>>>>>>>>
>>>>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>>>>
>>>>>>>>  	ret = mdiobus_register(new_bus);
>>>>>>>> +
>>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>>> +	dev_info(&pdev->dev,
>>>>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>>>>> +		 new_bus->id, ret);
>>>>>>>> +#endif
>>>>>>>>  	if (ret)
>>>>>>>>  		return ret;
>>>>>>>>
>>>>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>>>>> +
>>>>>>>>  	if (!tp->phydev) {
>>>>>>>>  		mdiobus_unregister(new_bus);
>>>>>>>>  		return -ENODEV;
>>>>>>>> -	} else if (!tp->phydev->drv) {
>>>>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>> -		 */
>>>>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>> -		mdiobus_unregister(new_bus);
>>>>>>>> -		return -EUNATCH;
>>>>>>>> +	} else {
>>>>>>>> +		/* tp -> phydev ok */
>>>>>>>> +		int everything_OK = 0;
>>>>>>>> +
>>>>>>>> +		/* Check driver id versus phy */
>>>>>>>> +
>>>>>>>> +		if (tp->phydev->drv) {
>>>>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>>>>> +			u32 drv_masked = ~0;
>>>>>>>> +			u32 phydev_match = ~0;
>>>>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>>>>> +
>>>>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>>>>> +
>>>>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>>>>> +
>>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>>> +			dev_debug(&pdev->dev,
>>>>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>>>>> +				phydrv_id, drv_masked);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>>>>> +
>>>>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>>>>> +			drv_match = drv_match == drv_masked;
>>>>>>>> +
>>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>>>>> +		}
>>>>>>>> +
>>>>>>>> +		if (!everything_OK) {
>>>>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>>>>> +			 * is loaded.
>>>>>>>> +			 */
>>>>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>> +			mdiobus_unregister(new_bus);
>>>>>>>> +			return -EUNATCH;
>>>>>>>> +		}
>>>>>>>>  	}
>>>>>>>>
>>>>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rtl_init_mac_address(tp);
>>>>>>>>
>>>>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>>>>
>>>>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>>>>> +
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	if (jumbo_max)
>>>>>>>>  		dev->max_mtu = jumbo_max;
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rtl_set_irq_mask(tp);
>>>>>>>>
>>>>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>>>>  					    &tp->counters_phys_addr,
>>>>>>>>  					    GFP_KERNEL);
>>>>>>>>  	if (!tp->counters)
>>>>>>>>  		return -ENOMEM;
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rc = r8169_mdio_register(tp);
>>>>>>>> +
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	if (rc)
>>>>>>>>  		return rc;
>>>>>>>>
>>>>>>>>  	/* chip gets powered up in rtl_open() */
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rtl_pll_power_down(tp);
>>>>>>>>
>>>>>>>>  	rc = register_netdev(dev);
>>>>>>>> +
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	if (rc)
>>>>>>>>  		goto err_mdio_unregister;
>>>>>>>>
>>>>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	return 0;
>>>>>>>>
>>>>>>>>  err_mdio_unregister:
>>>>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>>>>> index 522760c8bca6..41777f379a57 100644
>>>>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>>>>  {
>>>>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>>>>> -
>>>>>>>> +	struct phy_device *rv = NULL;
>>>>>>>> +/*
>>>>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>>>>> +*/
>>>>>>>>  	if (!mdiodev)
>>>>>>>>  		return NULL;
>>>>>>>>
>>>>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>>>>  		return NULL;
>>>>>>>>
>>>>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>>>>> +/*
>>>>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>>>>> +*/
>>>>>>>> +	return rv;
>>>>>>>>  }
>>>>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>>>>
>>>>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>>>>
>>>>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>>>>  	return 0;
>>>>>>>>
>>>>>>>>  error:
>>>>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>>>>  	while (--i >= 0) {
>>>>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>>>>  		if (!mdiodev)
>>>>>>>> --
>>>>>>>> 2.26.0
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>> On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> Comments inline.
>>>>>>>>>>
>>>>>>>>>> On 2020-04-13 13:58, Leon Romanovsky wrote:
>>>>>>>>>>> On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
>>>>>>>>>>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>>>>>>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>>>>>>
>>>>>>>>>>>> The driver installation determination made properly by
>>>>>>>>>>>> checking PHY vs DRIVER id's.
>>>>>>>>>>>> ---
>>>>>>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>>>>>>>>>>>>  drivers/net/phy/mdio_bus.c                | 11 +++-
>>>>>>>>>>>>  2 files changed, 72 insertions(+), 9 deletions(-)
>>>>>>>>>>>
>>>>>>>>>>> I would say that most of the code is debug prints.
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> I tought that they are helpful to keep, they are using the debug calls, so
>>>>>>>>>> they are not visible if user does not like those.
>>>>>>>>>
>>>>>>>>> You are missing the point of who are your users.
>>>>>>>>>
>>>>>>>>> Users want to have working device and the code. They don't need or like
>>>>>>>>> to debug their kernel.
>>>>>>>>>
>>>>>>>>> Thanks
>>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>> Br,
>>>>>>>> Lauri J.
>>>>>>>
>>>>>>>> From 1a75f6f9065a58180de1fa3c48fd80418af6c347 Mon Sep 17 00:00:00 2001
>>>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>>>
>>>>>>>> The driver installation determination made properly by
>>>>>>>> checking PHY vs DRIVER id's.
>>>>>>>> ---
>>>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 114 ++++++++++++++++++++--
>>>>>>>>  drivers/net/phy/mdio_bus.c                |  15 ++-
>>>>>>>>  2 files changed, 119 insertions(+), 10 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> index bf5bf05970a2..5e992f285527 100644
>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> @@ -61,6 +61,11 @@
>>>>>>>>  #define R8169_MSG_DEFAULT \
>>>>>>>>  	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
>>>>>>>>
>>>>>>>> +
>>>>>>>> +/*
>>>>>>>> +#define R8169_PROBE_DEBUG
>>>>>>>> +*/
>>>>>>>> +
>>>>>>>>  /* Maximum number of multicast addresses to filter (vs. Rx-all-multicast).
>>>>>>>>     The RTL chips use a 64 element hash table based on the Ethernet CRC. */
>>>>>>>>  #define	MC_FILTER_LIMIT	32
>>>>>>>> @@ -5149,6 +5154,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>  {
>>>>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>>>>  	struct mii_bus *new_bus;
>>>>>>>> +	u32 phydev_id = 0;
>>>>>>>> +	u32 phydrv_id = 0;
>>>>>>>> +	u32 phydrv_id_mask = 0;
>>>>>>>>  	int ret;
>>>>>>>>
>>>>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>>>>> @@ -5165,20 +5173,69 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>  	new_bus->write = r8169_mdio_write_reg;
>>>>>>>>
>>>>>>>>  	ret = mdiobus_register(new_bus);
>>>>>>>> +
>>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>>> +	dev_info(&pdev->dev,
>>>>>>>> +		 "mdiobus_register: %s, %d\n",
>>>>>>>> +		 new_bus->id, ret);
>>>>>>>> +#endif
>>>>>>>>  	if (ret)
>>>>>>>>  		return ret;
>>>>>>>>
>>>>>>>>  	tp->phydev = mdiobus_get_phy(new_bus, 0);
>>>>>>>> +
>>>>>>>>  	if (!tp->phydev) {
>>>>>>>>  		mdiobus_unregister(new_bus);
>>>>>>>>  		return -ENODEV;
>>>>>>>> -	} else if (!tp->phydev->drv) {
>>>>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>> -		 */
>>>>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>> -		mdiobus_unregister(new_bus);
>>>>>>>> -		return -EUNATCH;
>>>>>>>> +	} else {
>>>>>>>> +		/* tp -> phydev ok */
>>>>>>>> +		int everything_OK = 0;
>>>>>>>> +
>>>>>>>> +		/* Check driver id versus phy */
>>>>>>>> +
>>>>>>>> +		if (tp->phydev->drv) {
>>>>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>>>>> +			u32 drv_masked = ~0;
>>>>>>>> +			u32 phydev_match = ~0;
>>>>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>>>>> +
>>>>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>>>>> +
>>>>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>>>>> +
>>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>>> +			dev_debug(&pdev->dev,
>>>>>>>> +				  "%s: ID Check: (%x -> %x), drv (%x -> %x)\n",
>>>>>>>> +				new_bus->id, phydev_id, phydev_masked,
>>>>>>>> +				phydrv_id, drv_masked);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>>>>> +
>>>>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>>>>> +			drv_match = drv_match == drv_masked;
>>>>>>>> +
>>>>>>>> +#ifdef R8169_PROBE_DEBUG
>>>>>>>> +			dev_debug(&pdev->dev, "%s: ID Check: %x == %x\n",
>>>>>>>> +				  new_bus->id, phydev_match, drv_match);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>>>>> +		}
>>>>>>>> +
>>>>>>>> +		if (!everything_OK) {
>>>>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>>>>> +			 * is loaded.
>>>>>>>> +			 */
>>>>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>> +			mdiobus_unregister(new_bus);
>>>>>>>> +			return -EUNATCH;
>>>>>>>> +		}
>>>>>>>>  	}
>>>>>>>>
>>>>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>>>> @@ -5435,6 +5492,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>>  	u64_stats_init(&tp->rx_stats.syncp);
>>>>>>>>  	u64_stats_init(&tp->tx_stats.syncp);
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: MAC\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rtl_init_mac_address(tp);
>>>>>>>>
>>>>>>>>  	dev->ethtool_ops = &rtl8169_ethtool_ops;
>>>>>>>> @@ -5483,29 +5544,64 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>>  	dev->hw_features |= NETIF_F_RXFCS;
>>>>>>>>
>>>>>>>>  	jumbo_max = rtl_jumbo_max(tp);
>>>>>>>> +
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: jumbo max: %d\n", jumbo_max);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	if (jumbo_max)
>>>>>>>>  		dev->max_mtu = jumbo_max;
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: irq mask\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rtl_set_irq_mask(tp);
>>>>>>>>
>>>>>>>>  	tp->fw_name = rtl_chip_infos[chipset].fw_name;
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: FW name: %s\n", tp->fw_name);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	tp->counters = dmam_alloc_coherent (&pdev->dev, sizeof(*tp->counters),
>>>>>>>>  					    &tp->counters_phys_addr,
>>>>>>>>  					    GFP_KERNEL);
>>>>>>>>  	if (!tp->counters)
>>>>>>>>  		return -ENOMEM;
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: set driver data\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	pci_set_drvdata(pdev, dev);
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: register mdio\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rc = r8169_mdio_register(tp);
>>>>>>>> +
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: mdio register: %d\n", rc);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	if (rc)
>>>>>>>>  		return rc;
>>>>>>>>
>>>>>>>>  	/* chip gets powered up in rtl_open() */
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: pll pwr down\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	rtl_pll_power_down(tp);
>>>>>>>>
>>>>>>>>  	rc = register_netdev(dev);
>>>>>>>> +
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: netdev register: %d\n", rc);
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	if (rc)
>>>>>>>>  		goto err_mdio_unregister;
>>>>>>>>
>>>>>>>> @@ -5525,6 +5621,10 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>>>  	if (pci_dev_run_wake(pdev))
>>>>>>>>  		pm_runtime_put_sync(&pdev->dev);
>>>>>>>>
>>>>>>>> +#ifdef R816X_PROBE_DEBUG
>>>>>>>> +	dev_dbg(&pdev->dev, "init: ALL DONE!\n");
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>>  	return 0;
>>>>>>>>
>>>>>>>>  err_mdio_unregister:
>>>>>>>> diff --git a/drivers/net/phy/mdio_bus.c b/drivers/net/phy/mdio_bus.c
>>>>>>>> index 522760c8bca6..41777f379a57 100644
>>>>>>>> --- a/drivers/net/phy/mdio_bus.c
>>>>>>>> +++ b/drivers/net/phy/mdio_bus.c
>>>>>>>> @@ -112,14 +112,22 @@ EXPORT_SYMBOL(mdiobus_unregister_device);
>>>>>>>>  struct phy_device *mdiobus_get_phy(struct mii_bus *bus, int addr)
>>>>>>>>  {
>>>>>>>>  	struct mdio_device *mdiodev = bus->mdio_map[addr];
>>>>>>>> -
>>>>>>>> +	struct phy_device *rv = NULL;
>>>>>>>> +/*
>>>>>>>> +	pr_debug("mii_bus %s addr %d, %p\n", bus->id, addr, mdiodev);
>>>>>>>> +*/
>>>>>>>>  	if (!mdiodev)
>>>>>>>>  		return NULL;
>>>>>>>>
>>>>>>>>  	if (!(mdiodev->flags & MDIO_DEVICE_FLAG_PHY))
>>>>>>>>  		return NULL;
>>>>>>>>
>>>>>>>> -	return container_of(mdiodev, struct phy_device, mdio);
>>>>>>>> +	rv = container_of(mdiodev, struct phy_device, mdio);
>>>>>>>> +/*
>>>>>>>> + 	pr_debug("mii_bus OK? %s addr %d, %p -> %p\n",
>>>>>>>> +		 bus->id, addr, mdiodev, rv);
>>>>>>>> +*/
>>>>>>>> +	return rv;
>>>>>>>>  }
>>>>>>>>  EXPORT_SYMBOL(mdiobus_get_phy);
>>>>>>>>
>>>>>>>> @@ -645,10 +653,11 @@ int __mdiobus_register(struct mii_bus *bus, struct module *owner)
>>>>>>>>  	mdiobus_setup_mdiodev_from_board_info(bus, mdiobus_create_device);
>>>>>>>>
>>>>>>>>  	bus->state = MDIOBUS_REGISTERED;
>>>>>>>> -	pr_info("%s: probed\n", bus->name);
>>>>>>>> +	pr_info("%s: probed (mdiobus_register)\n", bus->name);
>>>>>>>>  	return 0;
>>>>>>>>
>>>>>>>>  error:
>>>>>>>> +	pr_err("%s: Error while in mdiobus_register: %d\n", bus->name, err);
>>>>>>>>  	while (--i >= 0) {
>>>>>>>>  		mdiodev = bus->mdio_map[i];
>>>>>>>>  		if (!mdiodev)
>>>>>>>> --
>>>>>>>> 2.26.0
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>> >From 85766c0ab77750faa4488ca7018784e3e8e9eba0 Mon Sep 17 00:00:00 2001
>>>>>> From: Lauri Jakku <lja@iki.fi>
>>>>>> Date: Mon, 13 Apr 2020 13:18:35 +0300
>>>>>> Subject: [PATCH] NET: r8168/r8169 identifying fix
>>>>>>
>>>>>> The driver installation determination made properly by
>>>>>> checking PHY vs DRIVER id's.
>>>>>> ---
>>>>>>  drivers/net/ethernet/realtek/r8169_main.c | 47 +++++++++++++++++++----
>>>>>>  1 file changed, 40 insertions(+), 7 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> index bf5bf05970a2..da08b1b1047c 100644
>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> @@ -5149,6 +5149,9 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>  {
>>>>>>  	struct pci_dev *pdev = tp->pci_dev;
>>>>>>  	struct mii_bus *new_bus;
>>>>>> +	u32 phydev_id = 0;
>>>>>> +	u32 phydrv_id = 0;
>>>>>> +	u32 phydrv_id_mask = 0;
>>>>>>  	int ret;
>>>>>>  
>>>>>>  	new_bus = devm_mdiobus_alloc(&pdev->dev);
>>>>>> @@ -5172,13 +5175,43 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>  	if (!tp->phydev) {
>>>>>>  		mdiobus_unregister(new_bus);
>>>>>>  		return -ENODEV;
>>>>>> -	} else if (!tp->phydev->drv) {
>>>>>> -		/* Most chip versions fail with the genphy driver.
>>>>>> -		 * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>> -		 */
>>>>>> -		dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>> -		mdiobus_unregister(new_bus);
>>>>>> -		return -EUNATCH;
>>>>>> +	} else {
>>>>>> +		/* tp -> phydev ok */
>>>>>> +		int everything_OK = 0;
>>>>>> +
>>>>>> +		/* Check driver id versus phy */
>>>>>> +
>>>>>> +		if (tp->phydev->drv) {
>>>>>> +			u32 phydev_masked = 0xBEEFDEAD;
>>>>>> +			u32 drv_masked = ~0;
>>>>>> +			u32 phydev_match = ~0;
>>>>>> +			u32 drv_match = 0xDEADBEEF;
>>>>>> +
>>>>>> +			phydev_id = tp->phydev->phy_id;
>>>>>> +			phydrv_id = tp->phydev->drv->phy_id;
>>>>>> +			phydrv_id_mask = tp->phydev->drv->phy_id_mask;
>>>>>> +
>>>>>> +			drv_masked = phydrv_id & phydrv_id_mask;
>>>>>> +			phydev_masked = phydev_id & phydrv_id_mask;
>>>>>> +
>>>>>> +			phydev_match = phydev_masked & drv_masked;
>>>>>> +			phydev_match = phydev_match == phydev_masked;
>>>>>> +
>>>>>> +			drv_match = phydev_masked & drv_masked;
>>>>>> +			drv_match = drv_match == drv_masked;
>>>>>> +
>>>>>> +			everything_OK = (phydev_match == drv_match);
>>>>>> +		}
>>>>>> +
>>>>>> +		if (!everything_OK) {
>>>>>> +			/* Most chip versions fail with the genphy driver.
>>>>>> +			 * Therefore ensure that the dedicated PHY driver
>>>>>> +			 * is loaded.
>>>>>> +			 */
>>>>>> +			dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>> +			mdiobus_unregister(new_bus);
>>>>>> +			return -EUNATCH;
>>>>>> +		}
>>>>>>  	}
>>>>>>  
>>>>>>  	/* PHY will be woken up in rtl_open() */
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
> 

-- 
Br,
Lauri J.

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
       [not found]         ` <d3adc7f2-06bb-45bc-ab02-3d443999cefd@gmail.com>
@ 2020-04-15 16:18           ` Heiner Kallweit
       [not found]             ` <4860e57e-93e4-24f5-6103-fa80acbdfa0d@pp.inet.fi>
  2021-03-11 16:00             ` gmail
  0 siblings, 2 replies; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-15 16:18 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 15.04.2020 16:39, Lauri Jakku wrote:
> Hi,
> 
> There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working.

There's no point in working on your patch. W/o proper justification it
isn't acceptable anyway. And so far we still don't know which problem
you actually have.
FIRST please provide the requested logs and explain the actual problem
(incl. the commit that caused the regression).


> 13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti:
> 
>     Hi,
> 
>     Fair enough, i'll strip them.
> 
>     -lja
> 
>     On 2020-04-13 14:34, Leon Romanovsky wrote:
> 
>         On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote:
> 
>             Hi,
> 
>             Comments inline.
> 
>             On 2020-04-13 13:58, Leon Romanovsky wrote:
> 
>                 On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote:
> 
>                     From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001
>                     From: Lauri Jakku <lja@iki.fi>
>                     Date: Mon, 13 Apr 2020 13:18:35 +0300
>                     Subject: [PATCH] NET: r8168/r8169 identifying fix
> 
>                     The driver installation determination made properly by
>                     checking PHY vs DRIVER id's.
>                     ---
>                     drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++---
>                     drivers/net/phy/mdio_bus.c | 11 +++-
>                     2 files changed, 72 insertions(+), 9 deletions(-)
> 
> 
>                 I would say that most of the code is debug prints.
> 
> 
> 
>             I tought that they are helpful to keep, they are using the debug calls, so
>             they are not visible if user does not like those.
> 
> 
>         You are missing the point of who are your users.
> 
>         Users want to have working device and the code. They don't need or like
>         to debug their kernel.
> 
>         Thanks
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
       [not found]             ` <4860e57e-93e4-24f5-6103-fa80acbdfa0d@pp.inet.fi>
@ 2020-04-16 18:26               ` Heiner Kallweit
  2020-04-16 18:37                 ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-16 18:26 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 16.04.2020 13:30, Lauri Jakku wrote:
> Hi,
> 
> 
> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
> 
> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
> 
> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
> 
> All different initcpio's have realtek.ko in them.
> 
Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
a random number as PHY ID, resulting in no PHY driver being found.
Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.


> 
> The problem with old method seems to be, that device does not have had time to attach before the
> PHY driver check.
> 
> The patch:
> 
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..acd122a88d4a 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>         if (!tp->phydev) {
>                 mdiobus_unregister(new_bus);
>                 return -ENODEV;
> -       } else if (!tp->phydev->drv) {
> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>                 /* Most chip versions fail with the genphy driver.
>                  * Therefore ensure that the dedicated PHY driver is loaded.
>                  */
> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>                 mdiobus_unregister(new_bus);
>                 return -EUNATCH;
>         }
> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
> index 66b8c61ca74c..aba2b304b821 100644
> --- a/drivers/net/phy/phy-core.c
> +++ b/drivers/net/phy/phy-core.c
> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>  
>  static int __phy_read_page(struct phy_device *phydev)
>  {
> +       /* If not attached, do nothing (no warning) */
> +       if (!phydev->attached_dev)
> +               return -EOPNOTSUPP;
> +
>         if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>                 return -EOPNOTSUPP;
>  
> @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>  
>  static int __phy_write_page(struct phy_device *phydev, int page)
>  {
> +       /* If not attached, do nothing (no warning) */
> +       if (!phydev->attached_dev)
> +               return -EOPNOTSUPP;
> +
>         if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>                 return -EOPNOTSUPP;
>  
>         return phydev->drv->write_page(phydev, page);
>  }
>  
> +
>  /**
>   * phy_save_page() - take the bus lock and save the current page
>   * @phydev: a pointer to a &struct phy_device
> 
> 
> 
> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
> 
>     On 15.04.2020 16:39, Lauri Jakku wrote:
> 
>         Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working. 
> 
> 
>     There's no point in working on your patch. W/o proper justification it
>     isn't acceptable anyway. And so far we still don't know which problem
>     you actually have.
>     FIRST please provide the requested logs and explain the actual problem
>     (incl. the commit that caused the regression).
> 
> 
> 
> 
>         13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
> 
>         On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks 
> 
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-16 18:26               ` Heiner Kallweit
@ 2020-04-16 18:37                 ` Lauri Jakku
  2020-04-16 19:58                   ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-16 18:37 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 16.4.2020 21.26, Heiner Kallweit wrote:
> On 16.04.2020 13:30, Lauri Jakku wrote:
>> Hi,
>>
>>
>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>
>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>
>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>
>> All different initcpio's have realtek.ko in them.
>>
> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
> a random number as PHY ID, resulting in no PHY driver being found.
> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>
OK, I try that, thank you :)

>> The problem with old method seems to be, that device does not have had time to attach before the
>> PHY driver check.
>>
>> The patch:
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>> index bf5bf05970a2..acd122a88d4a 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>          if (!tp->phydev) {
>>                  mdiobus_unregister(new_bus);
>>                  return -ENODEV;
>> -       } else if (!tp->phydev->drv) {
>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>                  /* Most chip versions fail with the genphy driver.
>>                   * Therefore ensure that the dedicated PHY driver is loaded.
>>                   */
>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>                  mdiobus_unregister(new_bus);
>>                  return -EUNATCH;
>>          }
>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>> index 66b8c61ca74c..aba2b304b821 100644
>> --- a/drivers/net/phy/phy-core.c
>> +++ b/drivers/net/phy/phy-core.c
>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>   
>>   static int __phy_read_page(struct phy_device *phydev)
>>   {
>> +       /* If not attached, do nothing (no warning) */
>> +       if (!phydev->attached_dev)
>> +               return -EOPNOTSUPP;
>> +
>>          if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>                  return -EOPNOTSUPP;
>>   
>> @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>   
>>   static int __phy_write_page(struct phy_device *phydev, int page)
>>   {
>> +       /* If not attached, do nothing (no warning) */
>> +       if (!phydev->attached_dev)
>> +               return -EOPNOTSUPP;
>> +
>>          if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>                  return -EOPNOTSUPP;
>>   
>>          return phydev->drv->write_page(phydev, page);
>>   }
>>   
>> +
>>   /**
>>    * phy_save_page() - take the bus lock and save the current page
>>    * @phydev: a pointer to a &struct phy_device
>>
>>
>>
>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>
>>      On 15.04.2020 16:39, Lauri Jakku wrote:
>>
>>          Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working.
>>
>>
>>      There's no point in working on your patch. W/o proper justification it
>>      isn't acceptable anyway. And so far we still don't know which problem
>>      you actually have.
>>      FIRST please provide the requested logs and explain the actual problem
>>      (incl. the commit that caused the regression).
>>
>>
>>
>>
>>          13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>
>>          On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>
>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-16 18:37                 ` Lauri Jakku
@ 2020-04-16 19:58                   ` Lauri Jakku
  2020-04-16 20:02                     ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-16 19:58 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 16.4.2020 21.37, Lauri Jakku wrote:
> Hi,
>
> On 16.4.2020 21.26, Heiner Kallweit wrote:
>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>> Hi,
>>>
>>>
>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> 
>>> network does not work
>>>
>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods 
>>> only) --> network does not work
>>>
>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> 
>>> devices show up ok, network works
>>>
>>> All different initcpio's have realtek.ko in them.
>>>
>> Thanks for the logs. Based on the logs you're presumable affected by 
>> a known BIOS bug.
>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>> In the first referenced tickets it's about the same mainboard (with 
>> earlier BIOS version).
>> BIOS on this mainboard seems to not initialize the network chip / PHY 
>> correctly, it reports
>> a random number as PHY ID, resulting in no PHY driver being found.
>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be 
>> gone.
>>
> OK, I try that, thank you :)
>

It seems that i DO have the ROM's enabled, i'm now testing some mutex 
guard for phy state and try to use it as indicator

that attach has been done. One thing i've noticed is that driver needs 
to be reloaded to allow traffic (ie. ping works etc.)


>>> The problem with old method seems to be, that device does not have 
>>> had time to attach before the
>>> PHY driver check.
>>>
>>> The patch:
>>>
>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>> index bf5bf05970a2..acd122a88d4a 100644
>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct 
>>> rtl8169_private *tp)
>>>          if (!tp->phydev) {
>>>                  mdiobus_unregister(new_bus);
>>>                  return -ENODEV;
>>> -       } else if (!tp->phydev->drv) {
>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>                  /* Most chip versions fail with the genphy driver.
>>>                   * Therefore ensure that the dedicated PHY driver 
>>> is loaded.
>>>                   */
>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it 
>>> needs to be added to initramfs?\n");
>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>                  mdiobus_unregister(new_bus);
>>>                  return -EUNATCH;
>>>          }
>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>> index 66b8c61ca74c..aba2b304b821 100644
>>> --- a/drivers/net/phy/phy-core.c
>>> +++ b/drivers/net/phy/phy-core.c
>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>     static int __phy_read_page(struct phy_device *phydev)
>>>   {
>>> +       /* If not attached, do nothing (no warning) */
>>> +       if (!phydev->attached_dev)
>>> +               return -EOPNOTSUPP;
>>> +
>>>          if (WARN_ONCE(!phydev->drv->read_page, "read_page callback 
>>> not available, PHY driver not loaded?\n"))
>>>                  return -EOPNOTSUPP;
>>>   @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device 
>>> *phydev)
>>>     static int __phy_write_page(struct phy_device *phydev, int page)
>>>   {
>>> +       /* If not attached, do nothing (no warning) */
>>> +       if (!phydev->attached_dev)
>>> +               return -EOPNOTSUPP;
>>> +
>>>          if (WARN_ONCE(!phydev->drv->write_page, "write_page 
>>> callback not available, PHY driver not loaded?\n"))
>>>                  return -EOPNOTSUPP;
>>>            return phydev->drv->write_page(phydev, page);
>>>   }
>>>   +
>>>   /**
>>>    * phy_save_page() - take the bus lock and save the current page
>>>    * @phydev: a pointer to a &struct phy_device
>>>
>>>
>>>
>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com 
>>> <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>
>>>      On 15.04.2020 16:39, Lauri Jakku wrote:
>>>
>>>          Hi, There seems to he Something odd problem, maybe timing 
>>> related. Stripped version not workingas expected. I get back to you, 
>>> when  i have it working.
>>>
>>>
>>>      There's no point in working on your patch. W/o proper 
>>> justification it
>>>      isn't acceptable anyway. And so far we still don't know which 
>>> problem
>>>      you actually have.
>>>      FIRST please provide the requested logs and explain the actual 
>>> problem
>>>      (incl. the commit that caused the regression).
>>>
>>>
>>>
>>>
>>>          13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com 
>>> <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip 
>>> them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>
>>>          On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku 
>>> wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky 
>>> wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: 
>>> From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 
>>> 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 
>>> +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver 
>>> installation determination made properly by checking PHY vs DRIVER 
>>> id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 
>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files 
>>> changed, 72 insertions(+), 9 deletions(-) I would say that most of 
>>> the code is debug prints. I tought that they are helpful to keep, 
>>> they are using the debug calls, so they are not visible if user does 
>>> not like those. You are missing the point of who are your users. 
>>> Users want to have working device and the code. They don't need or 
>>> like to debug their kernel. Thanks
>>>
>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-16 19:58                   ` Lauri Jakku
@ 2020-04-16 20:02                     ` Heiner Kallweit
  2020-04-16 20:10                       ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-16 20:02 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 16.04.2020 21:58, Lauri Jakku wrote:
> Hi,
> 
> On 16.4.2020 21.37, Lauri Jakku wrote:
>> Hi,
>>
>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>>
>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>
>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>
>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>
>>>> All different initcpio's have realtek.ko in them.
>>>>
>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>> a random number as PHY ID, resulting in no PHY driver being found.
>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>
>> OK, I try that, thank you :)
>>
> 
> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
> 
> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
> 
All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
And what do you mean with "it seems"? Is the option enabled or not?

> 
>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>> PHY driver check.
>>>>
>>>> The patch:
>>>>
>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>          if (!tp->phydev) {
>>>>                  mdiobus_unregister(new_bus);
>>>>                  return -ENODEV;
>>>> -       } else if (!tp->phydev->drv) {
>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>                  /* Most chip versions fail with the genphy driver.
>>>>                   * Therefore ensure that the dedicated PHY driver is loaded.
>>>>                   */
>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>                  mdiobus_unregister(new_bus);
>>>>                  return -EUNATCH;
>>>>          }
>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>> --- a/drivers/net/phy/phy-core.c
>>>> +++ b/drivers/net/phy/phy-core.c
>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>     static int __phy_read_page(struct phy_device *phydev)
>>>>   {
>>>> +       /* If not attached, do nothing (no warning) */
>>>> +       if (!phydev->attached_dev)
>>>> +               return -EOPNOTSUPP;
>>>> +
>>>>          if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>                  return -EOPNOTSUPP;
>>>>   @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>     static int __phy_write_page(struct phy_device *phydev, int page)
>>>>   {
>>>> +       /* If not attached, do nothing (no warning) */
>>>> +       if (!phydev->attached_dev)
>>>> +               return -EOPNOTSUPP;
>>>> +
>>>>          if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>                  return -EOPNOTSUPP;
>>>>            return phydev->drv->write_page(phydev, page);
>>>>   }
>>>>   +
>>>>   /**
>>>>    * phy_save_page() - take the bus lock and save the current page
>>>>    * @phydev: a pointer to a &struct phy_device
>>>>
>>>>
>>>>
>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>
>>>>      On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>
>>>>          Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working.
>>>>
>>>>
>>>>      There's no point in working on your patch. W/o proper justification it
>>>>      isn't acceptable anyway. And so far we still don't know which problem
>>>>      you actually have.
>>>>      FIRST please provide the requested logs and explain the actual problem
>>>>      (incl. the commit that caused the regression).
>>>>
>>>>
>>>>
>>>>
>>>>          13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>
>>>>          On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>
>>>>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-16 20:02                     ` Heiner Kallweit
@ 2020-04-16 20:10                       ` Lauri Jakku
  2020-04-16 20:38                         ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-16 20:10 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd


On 16.4.2020 23.02, Heiner Kallweit wrote:
> On 16.04.2020 21:58, Lauri Jakku wrote:
>> Hi,
>>
>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>>
>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>
>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>
>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>
>>>>> All different initcpio's have realtek.ko in them.
>>>>>
>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>
>>> OK, I try that, thank you :)
>>>
>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>
>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>
> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
> And what do you mean with "it seems"? Is the option enabled or not?
>
I do have ROM's enabled, and it does not help with my issue.
>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>> PHY driver check.
>>>>>
>>>>> The patch:
>>>>>
>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>           if (!tp->phydev) {
>>>>>                   mdiobus_unregister(new_bus);
>>>>>                   return -ENODEV;
>>>>> -       } else if (!tp->phydev->drv) {
>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>                   /* Most chip versions fail with the genphy driver.
>>>>>                    * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>                    */
>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>                   mdiobus_unregister(new_bus);
>>>>>                   return -EUNATCH;
>>>>>           }
>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>> --- a/drivers/net/phy/phy-core.c
>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>      static int __phy_read_page(struct phy_device *phydev)
>>>>>    {
>>>>> +       /* If not attached, do nothing (no warning) */
>>>>> +       if (!phydev->attached_dev)
>>>>> +               return -EOPNOTSUPP;
>>>>> +
>>>>>           if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>                   return -EOPNOTSUPP;
>>>>>    @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>      static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>    {
>>>>> +       /* If not attached, do nothing (no warning) */
>>>>> +       if (!phydev->attached_dev)
>>>>> +               return -EOPNOTSUPP;
>>>>> +
>>>>>           if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>                   return -EOPNOTSUPP;
>>>>>             return phydev->drv->write_page(phydev, page);
>>>>>    }
>>>>>    +
>>>>>    /**
>>>>>     * phy_save_page() - take the bus lock and save the current page
>>>>>     * @phydev: a pointer to a &struct phy_device
>>>>>
>>>>>
>>>>>
>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>
>>>>>       On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>
>>>>>           Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working.
>>>>>
>>>>>
>>>>>       There's no point in working on your patch. W/o proper justification it
>>>>>       isn't acceptable anyway. And so far we still don't know which problem
>>>>>       you actually have.
>>>>>       FIRST please provide the requested logs and explain the actual problem
>>>>>       (incl. the commit that caused the regression).
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>           13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>
>>>>>           On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>
>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-16 20:10                       ` Lauri Jakku
@ 2020-04-16 20:38                         ` Lauri Jakku
  2020-04-16 20:50                           ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-16 20:38 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi

On 16.4.2020 23.10, Lauri Jakku wrote:
>
> On 16.4.2020 23.02, Heiner Kallweit wrote:
>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>>
>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> 
>>>>>> network does not work
>>>>>>
>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods 
>>>>>> only) --> network does not work
>>>>>>
>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> 
>>>>>> devices show up ok, network works
>>>>>>
>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>
>>>>> Thanks for the logs. Based on the logs you're presumable affected 
>>>>> by a known BIOS bug.
>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>> In the first referenced tickets it's about the same mainboard 
>>>>> (with earlier BIOS version).
>>>>> BIOS on this mainboard seems to not initialize the network chip / 
>>>>> PHY correctly, it reports
>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should 
>>>>> be gone.
>>>>>
>>>> OK, I try that, thank you :)
>>>>
>>> It seems that i DO have the ROM's enabled, i'm now testing some 
>>> mutex guard for phy state and try to use it as indicator
>>>
>>> that attach has been done. One thing i've noticed is that driver 
>>> needs to be reloaded to allow traffic (ie. ping works etc.)
>>>
>> All that shouldn't be needed. Just check with which PHY ID the PHY 
>> comes up.
>> And what do you mean with "it seems"? Is the option enabled or not?
>>
> I do have ROM's enabled, and it does not help with my issue.

I check the ID, and revert all other changes, and check how it is 
working after adding the PHY id to list.

>>>>>> The problem with old method seems to be, that device does not 
>>>>>> have had time to attach before the
>>>>>> PHY driver check.
>>>>>>
>>>>>> The patch:
>>>>>>
>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct 
>>>>>> rtl8169_private *tp)
>>>>>>           if (!tp->phydev) {
>>>>>>                   mdiobus_unregister(new_bus);
>>>>>>                   return -ENODEV;
>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>                   /* Most chip versions fail with the genphy driver.
>>>>>>                    * Therefore ensure that the dedicated PHY 
>>>>>> driver is loaded.
>>>>>>                    */
>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe 
>>>>>> it needs to be added to initramfs?\n");
>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>                   mdiobus_unregister(new_bus);
>>>>>>                   return -EUNATCH;
>>>>>>           }
>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>      static int __phy_read_page(struct phy_device *phydev)
>>>>>>    {
>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>> +       if (!phydev->attached_dev)
>>>>>> +               return -EOPNOTSUPP;
>>>>>> +
>>>>>>           if (WARN_ONCE(!phydev->drv->read_page, "read_page 
>>>>>> callback not available, PHY driver not loaded?\n"))
>>>>>>                   return -EOPNOTSUPP;
>>>>>>    @@ -712,12 +716,17 @@ static int __phy_read_page(struct 
>>>>>> phy_device *phydev)
>>>>>>      static int __phy_write_page(struct phy_device *phydev, int 
>>>>>> page)
>>>>>>    {
>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>> +       if (!phydev->attached_dev)
>>>>>> +               return -EOPNOTSUPP;
>>>>>> +
>>>>>>           if (WARN_ONCE(!phydev->drv->write_page, "write_page 
>>>>>> callback not available, PHY driver not loaded?\n"))
>>>>>>                   return -EOPNOTSUPP;
>>>>>>             return phydev->drv->write_page(phydev, page);
>>>>>>    }
>>>>>>    +
>>>>>>    /**
>>>>>>     * phy_save_page() - take the bus lock and save the current page
>>>>>>     * @phydev: a pointer to a &struct phy_device
>>>>>>
>>>>>>
>>>>>>
>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com 
>>>>>> <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>
>>>>>>       On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>
>>>>>>           Hi, There seems to he Something odd problem, maybe 
>>>>>> timing related. Stripped version not workingas expected. I get 
>>>>>> back to you, when  i have it working.
>>>>>>
>>>>>>
>>>>>>       There's no point in working on your patch. W/o proper 
>>>>>> justification it
>>>>>>       isn't acceptable anyway. And so far we still don't know 
>>>>>> which problem
>>>>>>       you actually have.
>>>>>>       FIRST please provide the requested logs and explain the 
>>>>>> actual problem
>>>>>>       (incl. the commit that caused the regression).
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>           13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, 
>>>>>> Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon 
>>>>>> Romanovsky wrote:
>>>>>>
>>>>>>           On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku 
>>>>>> wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky 
>>>>>> wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku 
>>>>>> wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 
>>>>>> 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 
>>>>>> 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying 
>>>>>> fix The driver installation determination made properly by 
>>>>>> checking PHY vs DRIVER id's. --- 
>>>>>> drivers/net/ethernet/realtek/r8169_main.c | 70 
>>>>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 
>>>>>> files changed, 72 insertions(+), 9 deletions(-) I would say that 
>>>>>> most of the code is debug prints. I tought that they are helpful 
>>>>>> to keep, they are using the debug calls, so they are not visible 
>>>>>> if user does not like those. You are missing the point of who are 
>>>>>> your users. Users want to have working device and the code. They 
>>>>>> don't need or like to debug their kernel. Thanks
>>>>>>
>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-16 20:38                         ` Lauri Jakku
@ 2020-04-16 20:50                           ` Heiner Kallweit
  2020-04-17  6:23                             ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-16 20:50 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 16.04.2020 22:38, Lauri Jakku wrote:
> Hi
> 
> On 16.4.2020 23.10, Lauri Jakku wrote:
>>
>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>>
>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>
>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>
>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>
>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>
>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>
>>>>> OK, I try that, thank you :)
>>>>>
>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>
>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>
>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>
>> I do have ROM's enabled, and it does not help with my issue.
> 
Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
mainboard + BIOS as the user who opened bug ticket 202275.

> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
> 
>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>> PHY driver check.
>>>>>>>
>>>>>>> The patch:
>>>>>>>
>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>           if (!tp->phydev) {
>>>>>>>                   mdiobus_unregister(new_bus);
>>>>>>>                   return -ENODEV;
>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>                   /* Most chip versions fail with the genphy driver.
>>>>>>>                    * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>                    */
>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>                   mdiobus_unregister(new_bus);
>>>>>>>                   return -EUNATCH;
>>>>>>>           }
>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>      static int __phy_read_page(struct phy_device *phydev)
>>>>>>>    {
>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>> +       if (!phydev->attached_dev)
>>>>>>> +               return -EOPNOTSUPP;
>>>>>>> +
>>>>>>>           if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>                   return -EOPNOTSUPP;
>>>>>>>    @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>      static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>    {
>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>> +       if (!phydev->attached_dev)
>>>>>>> +               return -EOPNOTSUPP;
>>>>>>> +
>>>>>>>           if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>                   return -EOPNOTSUPP;
>>>>>>>             return phydev->drv->write_page(phydev, page);
>>>>>>>    }
>>>>>>>    +
>>>>>>>    /**
>>>>>>>     * phy_save_page() - take the bus lock and save the current page
>>>>>>>     * @phydev: a pointer to a &struct phy_device
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>
>>>>>>>       On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>
>>>>>>>           Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working.
>>>>>>>
>>>>>>>
>>>>>>>       There's no point in working on your patch. W/o proper justification it
>>>>>>>       isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>       you actually have.
>>>>>>>       FIRST please provide the requested logs and explain the actual problem
>>>>>>>       (incl. the commit that caused the regression).
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>           13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>
>>>>>>>           On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>
>>>>>>>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-16 20:50                           ` Heiner Kallweit
@ 2020-04-17  6:23                             ` Lauri Jakku
  2020-04-17  7:30                               ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-17  6:23 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd


On 16.4.2020 23.50, Heiner Kallweit wrote:
> On 16.04.2020 22:38, Lauri Jakku wrote:
>> Hi
>>
>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>>
>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>>
>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>>
>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>>
>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>
>>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>>
>>>>>> OK, I try that, thank you :)
>>>>>>
>>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>>
>>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>
>>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>
>>> I do have ROM's enabled, and it does not help with my issue.
> Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
> mainboard + BIOS as the user who opened bug ticket 202275.
>
huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0xc2077002
huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC 
version: 23

....

huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0x1cc912

huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC 
version: 23

.. after module unload & load cycle:

huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0x1cc912
huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC 
version: 23


it seem to be the case that the phy_id chances onetime, then stays the 
same. I'll do few shutdowns and see

is there a pattern at all .. next i'm going to try how it behaves, if i 
read mac/phy versions twice on MAC version 23.


The BIOS downgrade: I'd like to solve this without downgrading BIOS. If 
I can't, then I'll do systemd-service that

reloads r8169 driver at boot, cause then network is just fine.


>> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
>>
>>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>>> PHY driver check.
>>>>>>>>
>>>>>>>> The patch:
>>>>>>>>
>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>            if (!tp->phydev) {
>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>                    return -ENODEV;
>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>                    /* Most chip versions fail with the genphy driver.
>>>>>>>>                     * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>>                     */
>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>                    return -EUNATCH;
>>>>>>>>            }
>>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>     {
>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>> +
>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>     @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>       static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>>     {
>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>> +
>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>     }
>>>>>>>>     +
>>>>>>>>     /**
>>>>>>>>      * phy_save_page() - take the bus lock and save the current page
>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>
>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>
>>>>>>>>            Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working.
>>>>>>>>
>>>>>>>>
>>>>>>>>        There's no point in working on your patch. W/o proper justification it
>>>>>>>>        isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>>        you actually have.
>>>>>>>>        FIRST please provide the requested logs and explain the actual problem
>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>
>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>>
>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-17  6:23                             ` Lauri Jakku
@ 2020-04-17  7:30                               ` Lauri Jakku
  2020-04-17  8:57                                 ` Heiner Kallweit
  2020-04-18 11:06                                 ` Lauri Jakku
  0 siblings, 2 replies; 57+ messages in thread
From: Lauri Jakku @ 2020-04-17  7:30 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 17.4.2020 9.23, Lauri Jakku wrote:
>
> On 16.4.2020 23.50, Heiner Kallweit wrote:
>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>> Hi
>>>
>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications 
>>>>>>>>> --> network does not work
>>>>>>>>>
>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 
>>>>>>>>> mods only) --> network does not work
>>>>>>>>>
>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods 
>>>>>>>>> -> devices show up ok, network works
>>>>>>>>>
>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>
>>>>>>>> Thanks for the logs. Based on the logs you're presumable 
>>>>>>>> affected by a known BIOS bug.
>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>> In the first referenced tickets it's about the same mainboard 
>>>>>>>> (with earlier BIOS version).
>>>>>>>> BIOS on this mainboard seems to not initialize the network chip 
>>>>>>>> / PHY correctly, it reports
>>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem 
>>>>>>>> should be gone.
>>>>>>>>
>>>>>>> OK, I try that, thank you :)
>>>>>>>
>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some 
>>>>>> mutex guard for phy state and try to use it as indicator
>>>>>>
>>>>>> that attach has been done. One thing i've noticed is that driver 
>>>>>> needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>
>>>>> All that shouldn't be needed. Just check with which PHY ID the PHY 
>>>>> comes up.
>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>
>>>> I do have ROM's enabled, and it does not help with my issue.
>> Your BIOS is a beta version, downgrading to F7 may help. Then you 
>> have the same
>> mainboard + BIOS as the user who opened bug ticket 202275.
>>
> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0xc2077002
> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC 
> version: 23
>
> ....
>
> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0x1cc912
>
> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC 
> version: 23
>
> .. after module unload & load cycle:
>
> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0x1cc912
> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC 
> version: 23
>
>
> it seem to be the case that the phy_id chances onetime, then stays the 
> same. I'll do few shutdowns and see
>
> is there a pattern at all .. next i'm going to try how it behaves, if 
> i read mac/phy versions twice on MAC version 23.
>
>
> The BIOS downgrade: I'd like to solve this without downgrading BIOS. 
> If I can't, then I'll do systemd-service that
>
> reloads r8169 driver at boot, cause then network is just fine.
>
>
What i've gathered samples now, there is three values for PHY ID:

[sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY ver"
huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0xc2077002
huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0xc2077002
huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0x1cc912
huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0x1cc912
huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0x1cc912
huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0x1cc912
huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0xc1071002
huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0xc1071002
huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0x1cc912
huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0x1cc912
huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0xc1071002
huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0xc1071002
huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0x1cc912
huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0x1cc912
huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0xc1071002
huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0xc1071002
huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
version: 0x1cc912
huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
version: 0x1cc912

I dont know are those hard coded or what, and are they device specific 
how much.

i haven't coldbooted things up, that may be that something to check do 
they vary how per coldboot.

>>> I check the ID, and revert all other changes, and check how it is 
>>> working after adding the PHY id to list.
>>>
>>>>>>>>> The problem with old method seems to be, that device does not 
>>>>>>>>> have had time to attach before the
>>>>>>>>> PHY driver check.
>>>>>>>>>
>>>>>>>>> The patch:
>>>>>>>>>
>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct 
>>>>>>>>> rtl8169_private *tp)
>>>>>>>>>            if (!tp->phydev) {
>>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>>                    return -ENODEV;
>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>                    /* Most chip versions fail with the genphy 
>>>>>>>>> driver.
>>>>>>>>>                     * Therefore ensure that the dedicated PHY 
>>>>>>>>> driver is loaded.
>>>>>>>>>                     */
>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, 
>>>>>>>>> maybe it needs to be added to initramfs?\n");
>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>>                    return -EUNATCH;
>>>>>>>>>            }
>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>     {
>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>> +
>>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, "read_page 
>>>>>>>>> callback not available, PHY driver not loaded?\n"))
>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>     @@ -712,12 +716,17 @@ static int __phy_read_page(struct 
>>>>>>>>> phy_device *phydev)
>>>>>>>>>       static int __phy_write_page(struct phy_device *phydev, 
>>>>>>>>> int page)
>>>>>>>>>     {
>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>> +
>>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, "write_page 
>>>>>>>>> callback not available, PHY driver not loaded?\n"))
>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>>     }
>>>>>>>>>     +
>>>>>>>>>     /**
>>>>>>>>>      * phy_save_page() - take the bus lock and save the 
>>>>>>>>> current page
>>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com 
>>>>>>>>> <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>
>>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>
>>>>>>>>>            Hi, There seems to he Something odd problem, maybe 
>>>>>>>>> timing related. Stripped version not workingas expected. I get 
>>>>>>>>> back to you, when  i have it working.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>        There's no point in working on your patch. W/o proper 
>>>>>>>>> justification it
>>>>>>>>>        isn't acceptable anyway. And so far we still don't know 
>>>>>>>>> which problem
>>>>>>>>>        you actually have.
>>>>>>>>>        FIRST please provide the requested logs and explain the 
>>>>>>>>> actual problem
>>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: 
>>>>>>>>> Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, 
>>>>>>>>> Leon Romanovsky wrote:
>>>>>>>>>
>>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri 
>>>>>>>>> Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon 
>>>>>>>>> Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, 
>>>>>>>>> Lauri Jakku wrote: From 
>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 
>>>>>>>>> 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 
>>>>>>>>> 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying 
>>>>>>>>> fix The driver installation determination made properly by 
>>>>>>>>> checking PHY vs DRIVER id's. --- 
>>>>>>>>> drivers/net/ethernet/realtek/r8169_main.c | 70 
>>>>>>>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 
>>>>>>>>> files changed, 72 insertions(+), 9 deletions(-) I would say 
>>>>>>>>> that most of the code is debug prints. I tought that they are 
>>>>>>>>> helpful to keep, they are using the debug calls, so they are 
>>>>>>>>> not visible if user does not like those. You are missing the 
>>>>>>>>> point of who are your users. Users want to have working device 
>>>>>>>>> and the code. They don't need or like to debug their kernel. 
>>>>>>>>> Thanks
>>>>>>>>>
>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-17  7:30                               ` Lauri Jakku
@ 2020-04-17  8:57                                 ` Heiner Kallweit
  2020-04-18 11:06                                 ` Lauri Jakku
  1 sibling, 0 replies; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-17  8:57 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 17.04.2020 09:30, Lauri Jakku wrote:
> Hi,
> 
> On 17.4.2020 9.23, Lauri Jakku wrote:
>>
>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>> Hi
>>>>
>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>>>>
>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>>>>
>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>>>>
>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>
>>>>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>>>>
>>>>>>>> OK, I try that, thank you :)
>>>>>>>>
>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>>>>
>>>>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>
>>>>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>
>>>>> I do have ROM's enabled, and it does not help with my issue.
>>> Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>
>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC version: 23
>>
>> ....
>>
>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>
>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>
>> .. after module unload & load cycle:
>>
>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>
>>
>> it seem to be the case that the phy_id chances onetime, then stays the same. I'll do few shutdowns and see
>>
>> is there a pattern at all .. next i'm going to try how it behaves, if i read mac/phy versions twice on MAC version 23.
>>
>>
>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. If I can't, then I'll do systemd-service that
>>
>> reloads r8169 driver at boot, cause then network is just fine.
>>
>>
> What i've gathered samples now, there is three values for PHY ID:
> 
> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY ver"
> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc2077002
> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
> 
> I dont know are those hard coded or what, and are they device specific how much.
> 
0x001cc912 is the correct value, the other ones are more or less random values
caused by the BIOS bug.

> i haven't coldbooted things up, that may be that something to check do they vary how per coldboot.
> 
>>>> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
>>>>
>>>>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>>>>> PHY driver check.
>>>>>>>>>>
>>>>>>>>>> The patch:
>>>>>>>>>>
>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>            if (!tp->phydev) {
>>>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>>>                    return -ENODEV;
>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>                    /* Most chip versions fail with the genphy driver.
>>>>>>>>>>                     * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>>>>                     */
>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>>>                    return -EUNATCH;
>>>>>>>>>>            }
>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>     {
>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>> +
>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>     @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>       static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>>>>     {
>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>> +
>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>>>     }
>>>>>>>>>>     +
>>>>>>>>>>     /**
>>>>>>>>>>      * phy_save_page() - take the bus lock and save the current page
>>>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>
>>>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>
>>>>>>>>>>            Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when  i have it working.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>        There's no point in working on your patch. W/o proper justification it
>>>>>>>>>>        isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>>>>        you actually have.
>>>>>>>>>>        FIRST please provide the requested logs and explain the actual problem
>>>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>
>>>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>>>>
>>>>>>>>>>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-17  7:30                               ` Lauri Jakku
  2020-04-17  8:57                                 ` Heiner Kallweit
@ 2020-04-18 11:06                                 ` Lauri Jakku
  2020-04-18 18:46                                   ` Lauri Jakku
  1 sibling, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-18 11:06 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

[-- Attachment #1: Type: text/plain, Size: 11497 bytes --]

Hi,

On 17.4.2020 10.30, Lauri Jakku wrote:
> Hi,
>
> On 17.4.2020 9.23, Lauri Jakku wrote:
>>
>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>> Hi
>>>>
>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications 
>>>>>>>>>> --> network does not work
>>>>>>>>>>
>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 
>>>>>>>>>> mods only) --> network does not work
>>>>>>>>>>
>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 
>>>>>>>>>> mods -> devices show up ok, network works
>>>>>>>>>>
>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>
>>>>>>>>> Thanks for the logs. Based on the logs you're presumable 
>>>>>>>>> affected by a known BIOS bug.
>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>> In the first referenced tickets it's about the same mainboard 
>>>>>>>>> (with earlier BIOS version).
>>>>>>>>> BIOS on this mainboard seems to not initialize the network 
>>>>>>>>> chip / PHY correctly, it reports
>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being 
>>>>>>>>> found.
>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem 
>>>>>>>>> should be gone.
>>>>>>>>>
>>>>>>>> OK, I try that, thank you :)
>>>>>>>>
>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some 
>>>>>>> mutex guard for phy state and try to use it as indicator
>>>>>>>
>>>>>>> that attach has been done. One thing i've noticed is that driver 
>>>>>>> needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>
>>>>>> All that shouldn't be needed. Just check with which PHY ID the 
>>>>>> PHY comes up.
>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>
>>>>> I do have ROM's enabled, and it does not help with my issue.
>>> Your BIOS is a beta version, downgrading to F7 may help. Then you 
>>> have the same
>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>
>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0xc2077002
>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC 
>> version: 23
>>
>> ....
>>
>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0x1cc912
>>
>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC 
>> version: 23
>>
>> .. after module unload & load cycle:
>>
>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC 
>> version: 23
>>
>>
>> it seem to be the case that the phy_id chances onetime, then stays 
>> the same. I'll do few shutdowns and see
>>
>> is there a pattern at all .. next i'm going to try how it behaves, if 
>> i read mac/phy versions twice on MAC version 23.
>>
>>
>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. 
>> If I can't, then I'll do systemd-service that
>>
>> reloads r8169 driver at boot, cause then network is just fine.
>>
>>
> What i've gathered samples now, there is three values for PHY ID:
>
> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY 
> ver"
> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0xc2077002
> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0xc2077002
> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0x1cc912
> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0x1cc912
> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0x1cc912
> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0x1cc912
> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0xc1071002
> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0xc1071002
> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0x1cc912
> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0x1cc912
> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0xc1071002
> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0xc1071002
> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0x1cc912
> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0x1cc912
> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0xc1071002
> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0xc1071002
> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
> version: 0x1cc912
> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
> version: 0x1cc912
>
> I dont know are those hard coded or what, and are they device specific 
> how much.
>
> i haven't coldbooted things up, that may be that something to check do 
> they vary how per coldboot.
>
>>>> I check the ID, and revert all other changes, and check how it is 
>>>> working after adding the PHY id to list.
>>>>
What i've now learned: the patch + script + journald services -> Results 
working network, but it is still a workaround.


>>>>>>>>>> The problem with old method seems to be, that device does not 
>>>>>>>>>> have had time to attach before the
>>>>>>>>>> PHY driver check.
>>>>>>>>>>
>>>>>>>>>> The patch:
>>>>>>>>>>
>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct 
>>>>>>>>>> rtl8169_private *tp)
>>>>>>>>>>            if (!tp->phydev) {
>>>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>>>                    return -ENODEV;
>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>                    /* Most chip versions fail with the genphy 
>>>>>>>>>> driver.
>>>>>>>>>>                     * Therefore ensure that the dedicated PHY 
>>>>>>>>>> driver is loaded.
>>>>>>>>>>                     */
>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, 
>>>>>>>>>> maybe it needs to be added to initramfs?\n");
>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>                    mdiobus_unregister(new_bus);
>>>>>>>>>>                    return -EUNATCH;
>>>>>>>>>>            }
>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>     {
>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>> +
>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, "read_page 
>>>>>>>>>> callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>     @@ -712,12 +716,17 @@ static int __phy_read_page(struct 
>>>>>>>>>> phy_device *phydev)
>>>>>>>>>>       static int __phy_write_page(struct phy_device *phydev, 
>>>>>>>>>> int page)
>>>>>>>>>>     {
>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>> +
>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, 
>>>>>>>>>> "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>>>     }
>>>>>>>>>>     +
>>>>>>>>>>     /**
>>>>>>>>>>      * phy_save_page() - take the bus lock and save the 
>>>>>>>>>> current page
>>>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>> <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>
>>>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>
>>>>>>>>>>            Hi, There seems to he Something odd problem, maybe 
>>>>>>>>>> timing related. Stripped version not workingas expected. I 
>>>>>>>>>> get back to you, when i have it working.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>        There's no point in working on your patch. W/o proper 
>>>>>>>>>> justification it
>>>>>>>>>>        isn't acceptable anyway. And so far we still don't 
>>>>>>>>>> know which problem
>>>>>>>>>>        you actually have.
>>>>>>>>>>        FIRST please provide the requested logs and explain 
>>>>>>>>>> the actual problem
>>>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: 
>>>>>>>>>> Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, 
>>>>>>>>>> Leon Romanovsky wrote:
>>>>>>>>>>
>>>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri 
>>>>>>>>>> Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon 
>>>>>>>>>> Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, 
>>>>>>>>>> Lauri Jakku wrote: From 
>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 
>>>>>>>>>> 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 
>>>>>>>>>> 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying 
>>>>>>>>>> fix The driver installation determination made properly by 
>>>>>>>>>> checking PHY vs DRIVER id's. --- 
>>>>>>>>>> drivers/net/ethernet/realtek/r8169_main.c | 70 
>>>>>>>>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 
>>>>>>>>>> 2 files changed, 72 insertions(+), 9 deletions(-) I would say 
>>>>>>>>>> that most of the code is debug prints. I tought that they are 
>>>>>>>>>> helpful to keep, they are using the debug calls, so they are 
>>>>>>>>>> not visible if user does not like those. You are missing the 
>>>>>>>>>> point of who are your users. Users want to have working 
>>>>>>>>>> device and the code. They don't need or like to debug their 
>>>>>>>>>> kernel. Thanks
>>>>>>>>>>
>>>>>>>>>>

[-- Attachment #2: refresh-network.service --]
[-- Type: text/x-dbus-service, Size: 546 bytes --]

#  SPDX-License-Identifier: LGPL-2.1+
#
#  This file is part of systemd.
#
#  systemd is free software; you can redistribute it and/or modify it
#  under the terms of the GNU Lesser General Public License as published by
#  the Free Software Foundation; either version 2.1 of the License, or
#  (at your option) any later version.
[Unit]
Description=Refresh network
After=refresh-network.target
Before=multi-user.target

[Service]
Type=oneshot
ExecStart=/usr/local/bin/refresh-network.sh
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target

[-- Attachment #3: refresh-network.target --]
[-- Type: text/x-systemd-unit, Size: 523 bytes --]

#  SPDX-License-Identifier: LGPL-2.1+
#
#  This file is part of systemd.
#
#  systemd is free software; you can redistribute it and/or modify it
#  under the terms of the GNU Lesser General Public License as published by
#  the Free Software Foundation; either version 2.1 of the License, or
#  (at your option) any later version.

[Unit]
Description=Refresh network
After=network.target refresh-network.service
Before=multi-user.target
Requires=network.target

[Install]
WantedBy=refresh-network.service multi-user.target

[-- Attachment #4: refresh-network.sh --]
[-- Type: application/x-shellscript, Size: 459 bytes --]

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-18 11:06                                 ` Lauri Jakku
@ 2020-04-18 18:46                                   ` Lauri Jakku
  2020-04-19 15:09                                     ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-18 18:46 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 18.4.2020 14.06, Lauri Jakku wrote:
> Hi,
>
> On 17.4.2020 10.30, Lauri Jakku wrote:
>> Hi,
>>
>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>
>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>> Hi
>>>>>
>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications 
>>>>>>>>>>> --> network does not work
>>>>>>>>>>>
>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 
>>>>>>>>>>> mods only) --> network does not work
>>>>>>>>>>>
>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 
>>>>>>>>>>> mods -> devices show up ok, network works
>>>>>>>>>>>
>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>
>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable 
>>>>>>>>>> affected by a known BIOS bug.
>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>> In the first referenced tickets it's about the same mainboard 
>>>>>>>>>> (with earlier BIOS version).
>>>>>>>>>> BIOS on this mainboard seems to not initialize the network 
>>>>>>>>>> chip / PHY correctly, it reports
>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being 
>>>>>>>>>> found.
>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem 
>>>>>>>>>> should be gone.
>>>>>>>>>>
>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>
>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some 
>>>>>>>> mutex guard for phy state and try to use it as indicator
>>>>>>>>
>>>>>>>> that attach has been done. One thing i've noticed is that 
>>>>>>>> driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>>
>>>>>>> All that shouldn't be needed. Just check with which PHY ID the 
>>>>>>> PHY comes up.
>>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>>
>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you 
>>>> have the same
>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>
>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0xc2077002
>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> MAC version: 23
>>>
>>> ....
>>>
>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0x1cc912
>>>
>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> MAC version: 23
>>>
>>> .. after module unload & load cycle:
>>>
>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> MAC version: 23
>>>
>>>
>>> it seem to be the case that the phy_id chances onetime, then stays 
>>> the same. I'll do few shutdowns and see
>>>
>>> is there a pattern at all .. next i'm going to try how it behaves, 
>>> if i read mac/phy versions twice on MAC version 23.
>>>
>>>
>>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. 
>>> If I can't, then I'll do systemd-service that
>>>
>>> reloads r8169 driver at boot, cause then network is just fine.
>>>
>>>
>> What i've gathered samples now, there is three values for PHY ID:
>>
>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY 
>> ver"
>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0xc2077002
>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0xc2077002
>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0xc1071002
>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0xc1071002
>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0xc1071002
>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0xc1071002
>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0xc1071002
>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0xc1071002
>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY 
>> version: 0x1cc912
>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY 
>> version: 0x1cc912
>>
>> I dont know are those hard coded or what, and are they device 
>> specific how much.
>>
>> i haven't coldbooted things up, that may be that something to check 
>> do they vary how per coldboot.
>>
>>>>> I check the ID, and revert all other changes, and check how it is 
>>>>> working after adding the PHY id to list.
>>>>>
> What i've now learned: the patch + script + journald services -> 
> Results working network, but it is still a workaround.
>
Following patch trusts the MAC version, another thing witch could help 
is to derive method to do 2dn pass of the probeing:

if specific MAC version is found.

diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
b/drivers/net/ethernet/realtek/r8169_main.c
index acd122a88d4a..62b37a1abc24 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
rtl8169_private *tp)
         if (!tp->phydev) {
                 mdiobus_unregister(new_bus);
                 return -ENODEV;
-       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
-               /* Most chip versions fail with the genphy driver.
-                * Therefore ensure that the dedicated PHY driver is loaded.
-                */
-               dev_err(&pdev->dev, "Not known MAC version.\n");
-               mdiobus_unregister(new_bus);
-               return -EUNATCH;
+       } else {
+               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
tp->phydev->phy_id);
+               dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
+
+               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
+                       /* Most chip versions fail with the genphy driver.
+                        * Therefore ensure that the dedicated PHY 
driver is loaded.
+                        */
+                       dev_err(&pdev->dev, "Not known MAC/PHY 
version.\n", tp->phydev->phy_id);
+                       mdiobus_unregister(new_bus);
+                       return -EUNATCH;
+               }
         }

         /* PHY will be woken up in rtl_open() */

>
>>>>>>>>>>> The problem with old method seems to be, that device does 
>>>>>>>>>>> not have had time to attach before the
>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>
>>>>>>>>>>> The patch:
>>>>>>>>>>>
>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int 
>>>>>>>>>>> r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>            if (!tp->phydev) {
>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>                    return -ENODEV;
>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>                    /* Most chip versions fail with the 
>>>>>>>>>>> genphy driver.
>>>>>>>>>>>                     * Therefore ensure that the dedicated 
>>>>>>>>>>> PHY driver is loaded.
>>>>>>>>>>>                     */
>>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, 
>>>>>>>>>>> maybe it needs to be added to initramfs?\n");
>>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC 
>>>>>>>>>>> version.\n");
>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>                    return -EUNATCH;
>>>>>>>>>>>            }
>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>     {
>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>> +
>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, "read_page 
>>>>>>>>>>> callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>     @@ -712,12 +716,17 @@ static int __phy_read_page(struct 
>>>>>>>>>>> phy_device *phydev)
>>>>>>>>>>>       static int __phy_write_page(struct phy_device *phydev, 
>>>>>>>>>>> int page)
>>>>>>>>>>>     {
>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>> +
>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, 
>>>>>>>>>>> "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>     }
>>>>>>>>>>>     +
>>>>>>>>>>>     /**
>>>>>>>>>>>      * phy_save_page() - take the bus lock and save the 
>>>>>>>>>>> current page
>>>>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>>> <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>>
>>>>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>
>>>>>>>>>>>            Hi, There seems to he Something odd problem, 
>>>>>>>>>>> maybe timing related. Stripped version not workingas 
>>>>>>>>>>> expected. I get back to you, when i have it working.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>        There's no point in working on your patch. W/o proper 
>>>>>>>>>>> justification it
>>>>>>>>>>>        isn't acceptable anyway. And so far we still don't 
>>>>>>>>>>> know which problem
>>>>>>>>>>>        you actually have.
>>>>>>>>>>>        FIRST please provide the requested logs and explain 
>>>>>>>>>>> the actual problem
>>>>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: 
>>>>>>>>>>> Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, 
>>>>>>>>>>> Leon Romanovsky wrote:
>>>>>>>>>>>
>>>>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri 
>>>>>>>>>>> Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon 
>>>>>>>>>>> Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, 
>>>>>>>>>>> Lauri Jakku wrote: From 
>>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 
>>>>>>>>>>> 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 
>>>>>>>>>>> 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying 
>>>>>>>>>>> fix The driver installation determination made properly by 
>>>>>>>>>>> checking PHY vs DRIVER id's. --- 
>>>>>>>>>>> drivers/net/ethernet/realtek/r8169_main.c | 70 
>>>>>>>>>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 
>>>>>>>>>>> 2 files changed, 72 insertions(+), 9 deletions(-) I would 
>>>>>>>>>>> say that most of the code is debug prints. I tought that 
>>>>>>>>>>> they are helpful to keep, they are using the debug calls, so 
>>>>>>>>>>> they are not visible if user does not like those. You are 
>>>>>>>>>>> missing the point of who are your users. Users want to have 
>>>>>>>>>>> working device and the code. They don't need or like to 
>>>>>>>>>>> debug their kernel. Thanks
>>>>>>>>>>>
>>>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-18 18:46                                   ` Lauri Jakku
@ 2020-04-19 15:09                                     ` Lauri Jakku
  2020-04-19 16:49                                       ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-19 15:09 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 18.4.2020 21.46, Lauri Jakku wrote:

> Hi,
>
> On 18.4.2020 14.06, Lauri Jakku wrote:
>> Hi,
>>
>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>
>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>> Hi
>>>>>>
>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>> Hi,
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without 
>>>>>>>>>>>> modifications --> network does not work
>>>>>>>>>>>>
>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel 
>>>>>>>>>>>> (r8169 mods only) --> network does not work
>>>>>>>>>>>>
>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 
>>>>>>>>>>>> mods -> devices show up ok, network works
>>>>>>>>>>>>
>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>
>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable 
>>>>>>>>>>> affected by a known BIOS bug.
>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>> In the first referenced tickets it's about the same 
>>>>>>>>>>> mainboard (with earlier BIOS version).
>>>>>>>>>>> BIOS on this mainboard seems to not initialize the network 
>>>>>>>>>>> chip / PHY correctly, it reports
>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being 
>>>>>>>>>>> found.
>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem 
>>>>>>>>>>> should be gone.
>>>>>>>>>>>
>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>
>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing 
>>>>>>>>> some mutex guard for phy state and try to use it as indicator
>>>>>>>>>
>>>>>>>>> that attach has been done. One thing i've noticed is that 
>>>>>>>>> driver needs to be reloaded to allow traffic (ie. ping works 
>>>>>>>>> etc.)
>>>>>>>>>
>>>>>>>> All that shouldn't be needed. Just check with which PHY ID the 
>>>>>>>> PHY comes up.
>>>>>>>> And what do you mean with "it seems"? Is the option enabled or 
>>>>>>>> not?
>>>>>>>>
>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you 
>>>>> have the same
>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>
>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0xc2077002
>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> MAC version: 23
>>>>
>>>> ....
>>>>
>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0x1cc912
>>>>
>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> MAC version: 23
>>>>
>>>> .. after module unload & load cycle:
>>>>
>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> MAC version: 23
>>>>
>>>>
>>>> it seem to be the case that the phy_id chances onetime, then stays 
>>>> the same. I'll do few shutdowns and see
>>>>
>>>> is there a pattern at all .. next i'm going to try how it behaves, 
>>>> if i read mac/phy versions twice on MAC version 23.
>>>>
>>>>
>>>> The BIOS downgrade: I'd like to solve this without downgrading 
>>>> BIOS. If I can't, then I'll do systemd-service that
>>>>
>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>
>>>>
>>> What i've gathered samples now, there is three values for PHY ID:
>>>
>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep 
>>> "PHY ver"
>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0xc2077002
>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0xc2077002
>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0xc1071002
>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0xc1071002
>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0xc1071002
>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0xc1071002
>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0xc1071002
>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0xc1071002
>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>> PHY version: 0x1cc912
>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>> PHY version: 0x1cc912
>>>
>>> I dont know are those hard coded or what, and are they device 
>>> specific how much.
>>>
>>> i haven't coldbooted things up, that may be that something to check 
>>> do they vary how per coldboot.
>>>
>>>>>> I check the ID, and revert all other changes, and check how it is 
>>>>>> working after adding the PHY id to list.
>>>>>>
>> What i've now learned: the patch + script + journald services -> 
>> Results working network, but it is still a workaround.
>>
> Following patch trusts the MAC version, another thing witch could help 
> is to derive method to do 2dn pass of the probeing:
>
> if specific MAC version is found.
>
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
> b/drivers/net/ethernet/realtek/r8169_main.c
> index acd122a88d4a..62b37a1abc24 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
> rtl8169_private *tp)
>         if (!tp->phydev) {
>                 mdiobus_unregister(new_bus);
>                 return -ENODEV;
> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
> -               /* Most chip versions fail with the genphy driver.
> -                * Therefore ensure that the dedicated PHY driver is 
> loaded.
> -                */
> -               dev_err(&pdev->dev, "Not known MAC version.\n");
> -               mdiobus_unregister(new_bus);
> -               return -EUNATCH;
> +       } else {
> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
> tp->phydev->phy_id);
> +               dev_info(&pdev->dev, "MAC version: %d\n", 
> tp->mac_version);
> +
> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
> +                       /* Most chip versions fail with the genphy 
> driver.
> +                        * Therefore ensure that the dedicated PHY 
> driver is loaded.
> +                        */
> +                       dev_err(&pdev->dev, "Not known MAC/PHY 
> version.\n", tp->phydev->phy_id);
> +                       mdiobus_unregister(new_bus);
> +                       return -EUNATCH;
> +               }
>         }
>
>         /* PHY will be woken up in rtl_open() */
>

I just got bleeding edge 5.7.0-1 kernel compiled + firmware's updated.. 
and  now up'n'running. There is one (WARN_ONCE) stack trace coming from 
driver, i think i tinker with it next, with above patch the network 
devices shows up and they can be configured.

>>
>>>>>>>>>>>> The problem with old method seems to be, that device does 
>>>>>>>>>>>> not have had time to attach before the
>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>
>>>>>>>>>>>> The patch:
>>>>>>>>>>>>
>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int 
>>>>>>>>>>>> r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>            if (!tp->phydev) {
>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>                    return -ENODEV;
>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>                    /* Most chip versions fail with the 
>>>>>>>>>>>> genphy driver.
>>>>>>>>>>>>                     * Therefore ensure that the dedicated 
>>>>>>>>>>>> PHY driver is loaded.
>>>>>>>>>>>>                     */
>>>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, 
>>>>>>>>>>>> maybe it needs to be added to initramfs?\n");
>>>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC 
>>>>>>>>>>>> version.\n");
>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>                    return -EUNATCH;
>>>>>>>>>>>>            }
>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>     {
>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>> +
>>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, 
>>>>>>>>>>>> "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>>     @@ -712,12 +716,17 @@ static int __phy_read_page(struct 
>>>>>>>>>>>> phy_device *phydev)
>>>>>>>>>>>>       static int __phy_write_page(struct phy_device 
>>>>>>>>>>>> *phydev, int page)
>>>>>>>>>>>>     {
>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>> +
>>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, 
>>>>>>>>>>>> "write_page callback not available, PHY driver not 
>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>     }
>>>>>>>>>>>>     +
>>>>>>>>>>>>     /**
>>>>>>>>>>>>      * phy_save_page() - take the bus lock and save the 
>>>>>>>>>>>> current page
>>>>>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>>>> <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> 
>>>>>>>>>>>> kirjoitti:
>>>>>>>>>>>>
>>>>>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>            Hi, There seems to he Something odd problem, 
>>>>>>>>>>>> maybe timing related. Stripped version not workingas 
>>>>>>>>>>>> expected. I get back to you, when i have it working.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>        There's no point in working on your patch. W/o 
>>>>>>>>>>>> proper justification it
>>>>>>>>>>>>        isn't acceptable anyway. And so far we still don't 
>>>>>>>>>>>> know which problem
>>>>>>>>>>>>        you actually have.
>>>>>>>>>>>>        FIRST please provide the requested logs and explain 
>>>>>>>>>>>> the actual problem
>>>>>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: 
>>>>>>>>>>>> Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, 
>>>>>>>>>>>> Leon Romanovsky wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri 
>>>>>>>>>>>> Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon 
>>>>>>>>>>>> Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, 
>>>>>>>>>>>> Lauri Jakku wrote: From 
>>>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 
>>>>>>>>>>>> 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 
>>>>>>>>>>>> Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 
>>>>>>>>>>>> identifying fix The driver installation determination made 
>>>>>>>>>>>> properly by checking PHY vs DRIVER id's. --- 
>>>>>>>>>>>> drivers/net/ethernet/realtek/r8169_main.c | 70 
>>>>>>>>>>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 
>>>>>>>>>>>> +++- 2 files changed, 72 insertions(+), 9 deletions(-) I 
>>>>>>>>>>>> would say that most of the code is debug prints. I tought 
>>>>>>>>>>>> that they are helpful to keep, they are using the debug 
>>>>>>>>>>>> calls, so they are not visible if user does not like those. 
>>>>>>>>>>>> You are missing the point of who are your users. Users want 
>>>>>>>>>>>> to have working device and the code. They don't need or 
>>>>>>>>>>>> like to debug their kernel. Thanks
>>>>>>>>>>>>
>>>>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-19 16:49                                       ` Lauri Jakku
@ 2020-04-19 16:00                                         ` Heiner Kallweit
  2020-04-19 21:00                                           ` Lauri Jakku
  2020-05-01 19:12                                           ` Lauri Jakku
  0 siblings, 2 replies; 57+ messages in thread
From: Heiner Kallweit @ 2020-04-19 16:00 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 19.04.2020 18:49, Lauri Jakku wrote:
> Hi,
> 
> On 19.4.2020 18.09, Lauri Jakku wrote:
>> Hi,
>>
>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>
>>> Hi,
>>>
>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>
>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>> Hi
>>>>>>>>
>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>> Hi,
>>>>>>>>>>>>
>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>>>>>>>>
>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>
>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>>
>>>>>>>>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>>>>>
>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>>>>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>>>>>
>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>
>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC version: 23
>>>>>>
>>>>>> ....
>>>>>>
>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>
>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>
>>>>>> .. after module unload & load cycle:
>>>>>>
>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>
>>>>>>
>>>>>> it seem to be the case that the phy_id chances onetime, then stays the same. I'll do few shutdowns and see
>>>>>>
>>>>>> is there a pattern at all .. next i'm going to try how it behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>
>>>>>>
>>>>>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. If I can't, then I'll do systemd-service that
>>>>>>
>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>
>>>>>>
>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>
>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY ver"
>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc2077002
>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>
>>>>> I dont know are those hard coded or what, and are they device specific how much.
>>>>>
>>>>> i haven't coldbooted things up, that may be that something to check do they vary how per coldboot.
>>>>>
>>>>>>>> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
>>>>>>>>
>>>> What i've now learned: the patch + script + journald services -> Results working network, but it is still a workaround.
>>>>
>>> Following patch trusts the MAC version, another thing witch could help is to derive method to do 2dn pass of the probeing:
>>>
>>> if specific MAC version is found.
>>>
>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>> index acd122a88d4a..62b37a1abc24 100644
>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>         if (!tp->phydev) {
>>>                 mdiobus_unregister(new_bus);
>>>                 return -ENODEV;
>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>> -               /* Most chip versions fail with the genphy driver.
>>> -                * Therefore ensure that the dedicated PHY driver is loaded.
>>> -                */
>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>> -               mdiobus_unregister(new_bus);
>>> -               return -EUNATCH;
>>> +       } else {
>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
>>> +               dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
>>> +
>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>> +                       /* Most chip versions fail with the genphy driver.
>>> +                        * Therefore ensure that the dedicated PHY driver is loaded.
>>> +                        */
>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY version.\n", tp->phydev->phy_id);
>>> +                       mdiobus_unregister(new_bus);
>>> +                       return -EUNATCH;
>>> +               }
>>>         }
>>>
>>>         /* PHY will be woken up in rtl_open() */
>>>
>>
>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's updated.. and  now up'n'running. There is one (WARN_ONCE) stack trace coming from driver, i think i tinker with it next, with above patch the network devices shows up and they can be configured.
>>
> 
> I tought to ask first, before going to make new probe_type for errorneus hw (propetype + retry counter) to do re-probe if requested, N times. Or should the r8169_main.c return deferred probe on error on some MAC enums ? Which approach is design-wise sound ?
> 
> I just tought that the DEFERRED probe may just do the trick i'm looking ways to implement the re-probeing... hmm. I try the deferred thing and check would that help.
> 
Playing with options to work around the issue is of course a great way to
learn about the kernel. However it's questionable whether a workaround in
the driver is acceptable for dealing with the broken BIOS of exactly one
> 10 yrs old board (for which even a userspace workaround exists).

>>>>
>>>>>>>>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>            if (!tp->phydev) {
>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>                    return -ENODEV;
>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>                    /* Most chip versions fail with the genphy driver.
>>>>>>>>>>>>>>                     * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>>>>>>>>                     */
>>>>>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>                    return -EUNATCH;
>>>>>>>>>>>>>>            }
>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>     {
>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>>>>     @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>       static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>>>>>>>>     {
>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>>>     }
>>>>>>>>>>>>>>     +
>>>>>>>>>>>>>>     /**
>>>>>>>>>>>>>>      * phy_save_page() - take the bus lock and save the current page
>>>>>>>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>            Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when i have it working.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>        There's no point in working on your patch. W/o proper justification it
>>>>>>>>>>>>>>        isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>>>>>>>>        you actually have.
>>>>>>>>>>>>>>        FIRST please provide the requested logs and explain the actual problem
>>>>>>>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-19 15:09                                     ` Lauri Jakku
@ 2020-04-19 16:49                                       ` Lauri Jakku
  2020-04-19 16:00                                         ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-19 16:49 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 19.4.2020 18.09, Lauri Jakku wrote:
> Hi,
>
> On 18.4.2020 21.46, Lauri Jakku wrote:
>
>> Hi,
>>
>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>
>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>> Hi
>>>>>>>
>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without 
>>>>>>>>>>>>> modifications --> network does not work
>>>>>>>>>>>>>
>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel 
>>>>>>>>>>>>> (r8169 mods only) --> network does not work
>>>>>>>>>>>>>
>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 
>>>>>>>>>>>>> mods -> devices show up ok, network works
>>>>>>>>>>>>>
>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>
>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable 
>>>>>>>>>>>> affected by a known BIOS bug.
>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>> In the first referenced tickets it's about the same 
>>>>>>>>>>>> mainboard (with earlier BIOS version).
>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the network 
>>>>>>>>>>>> chip / PHY correctly, it reports
>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being 
>>>>>>>>>>>> found.
>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem 
>>>>>>>>>>>> should be gone.
>>>>>>>>>>>>
>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>
>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing 
>>>>>>>>>> some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>
>>>>>>>>>> that attach has been done. One thing i've noticed is that 
>>>>>>>>>> driver needs to be reloaded to allow traffic (ie. ping works 
>>>>>>>>>> etc.)
>>>>>>>>>>
>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID the 
>>>>>>>>> PHY comes up.
>>>>>>>>> And what do you mean with "it seems"? Is the option enabled or 
>>>>>>>>> not?
>>>>>>>>>
>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you 
>>>>>> have the same
>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>
>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>>> PHY version: 0xc2077002
>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>>> MAC version: 23
>>>>>
>>>>> ....
>>>>>
>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>>> PHY version: 0x1cc912
>>>>>
>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>>> MAC version: 23
>>>>>
>>>>> .. after module unload & load cycle:
>>>>>
>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>>> PHY version: 0x1cc912
>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>>> MAC version: 23
>>>>>
>>>>>
>>>>> it seem to be the case that the phy_id chances onetime, then stays 
>>>>> the same. I'll do few shutdowns and see
>>>>>
>>>>> is there a pattern at all .. next i'm going to try how it behaves, 
>>>>> if i read mac/phy versions twice on MAC version 23.
>>>>>
>>>>>
>>>>> The BIOS downgrade: I'd like to solve this without downgrading 
>>>>> BIOS. If I can't, then I'll do systemd-service that
>>>>>
>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>
>>>>>
>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>
>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep 
>>>> "PHY ver"
>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0xc2077002
>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0xc2077002
>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0xc1071002
>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0xc1071002
>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0xc1071002
>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0xc1071002
>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0xc1071002
>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0xc1071002
>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: 
>>>> PHY version: 0x1cc912
>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: 
>>>> PHY version: 0x1cc912
>>>>
>>>> I dont know are those hard coded or what, and are they device 
>>>> specific how much.
>>>>
>>>> i haven't coldbooted things up, that may be that something to check 
>>>> do they vary how per coldboot.
>>>>
>>>>>>> I check the ID, and revert all other changes, and check how it 
>>>>>>> is working after adding the PHY id to list.
>>>>>>>
>>> What i've now learned: the patch + script + journald services -> 
>>> Results working network, but it is still a workaround.
>>>
>> Following patch trusts the MAC version, another thing witch could 
>> help is to derive method to do 2dn pass of the probeing:
>>
>> if specific MAC version is found.
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>> b/drivers/net/ethernet/realtek/r8169_main.c
>> index acd122a88d4a..62b37a1abc24 100644
>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
>> rtl8169_private *tp)
>>         if (!tp->phydev) {
>>                 mdiobus_unregister(new_bus);
>>                 return -ENODEV;
>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>> -               /* Most chip versions fail with the genphy driver.
>> -                * Therefore ensure that the dedicated PHY driver is 
>> loaded.
>> -                */
>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>> -               mdiobus_unregister(new_bus);
>> -               return -EUNATCH;
>> +       } else {
>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
>> tp->phydev->phy_id);
>> +               dev_info(&pdev->dev, "MAC version: %d\n", 
>> tp->mac_version);
>> +
>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>> +                       /* Most chip versions fail with the genphy 
>> driver.
>> +                        * Therefore ensure that the dedicated PHY 
>> driver is loaded.
>> +                        */
>> +                       dev_err(&pdev->dev, "Not known MAC/PHY 
>> version.\n", tp->phydev->phy_id);
>> +                       mdiobus_unregister(new_bus);
>> +                       return -EUNATCH;
>> +               }
>>         }
>>
>>         /* PHY will be woken up in rtl_open() */
>>
>
> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's 
> updated.. and  now up'n'running. There is one (WARN_ONCE) stack trace 
> coming from driver, i think i tinker with it next, with above patch 
> the network devices shows up and they can be configured.
>

I tought to ask first, before going to make new probe_type for errorneus 
hw (propetype + retry counter) to do re-probe if requested, N times. Or 
should the r8169_main.c return deferred probe on error on some MAC enums 
? Which approach is design-wise sound ?

I just tought that the DEFERRED probe may just do the trick i'm looking 
ways to implement the re-probeing... hmm. I try the deferred thing and 
check would that help.

>>>
>>>>>>>>>>>>> The problem with old method seems to be, that device does 
>>>>>>>>>>>>> not have had time to attach before the
>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>
>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>
>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int 
>>>>>>>>>>>>> r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>            if (!tp->phydev) {
>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>                    return -ENODEV;
>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>                    /* Most chip versions fail with the 
>>>>>>>>>>>>> genphy driver.
>>>>>>>>>>>>>                     * Therefore ensure that the dedicated 
>>>>>>>>>>>>> PHY driver is loaded.
>>>>>>>>>>>>>                     */
>>>>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not 
>>>>>>>>>>>>> loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC 
>>>>>>>>>>>>> version.\n");
>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>                    return -EUNATCH;
>>>>>>>>>>>>>            }
>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>       static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>     {
>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>> +
>>>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->read_page, 
>>>>>>>>>>>>> "read_page callback not available, PHY driver not 
>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>>>     @@ -712,12 +716,17 @@ static int 
>>>>>>>>>>>>> __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>       static int __phy_write_page(struct phy_device 
>>>>>>>>>>>>> *phydev, int page)
>>>>>>>>>>>>>     {
>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>> +
>>>>>>>>>>>>>            if (WARN_ONCE(!phydev->drv->write_page, 
>>>>>>>>>>>>> "write_page callback not available, PHY driver not 
>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>                    return -EOPNOTSUPP;
>>>>>>>>>>>>>              return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>>     }
>>>>>>>>>>>>>     +
>>>>>>>>>>>>>     /**
>>>>>>>>>>>>>      * phy_save_page() - take the bus lock and save the 
>>>>>>>>>>>>> current page
>>>>>>>>>>>>>      * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>>>>> <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> 
>>>>>>>>>>>>> kirjoitti:
>>>>>>>>>>>>>
>>>>>>>>>>>>>        On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>>            Hi, There seems to he Something odd problem, 
>>>>>>>>>>>>> maybe timing related. Stripped version not workingas 
>>>>>>>>>>>>> expected. I get back to you, when i have it working.
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>        There's no point in working on your patch. W/o 
>>>>>>>>>>>>> proper justification it
>>>>>>>>>>>>>        isn't acceptable anyway. And so far we still don't 
>>>>>>>>>>>>> know which problem
>>>>>>>>>>>>>        you actually have.
>>>>>>>>>>>>>        FIRST please provide the requested logs and explain 
>>>>>>>>>>>>> the actual problem
>>>>>>>>>>>>>        (incl. the commit that caused the regression).
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>            13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> 
>>>>>>>>>>>>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 
>>>>>>>>>>>>> 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>>            On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri 
>>>>>>>>>>>>> Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, 
>>>>>>>>>>>>> Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM 
>>>>>>>>>>>>> +0300, Lauri Jakku wrote: From 
>>>>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 
>>>>>>>>>>>>> 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 
>>>>>>>>>>>>> Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 
>>>>>>>>>>>>> identifying fix The driver installation determination made 
>>>>>>>>>>>>> properly by checking PHY vs DRIVER id's. --- 
>>>>>>>>>>>>> drivers/net/ethernet/realtek/r8169_main.c | 70 
>>>>>>>>>>>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 
>>>>>>>>>>>>> +++- 2 files changed, 72 insertions(+), 9 deletions(-) I 
>>>>>>>>>>>>> would say that most of the code is debug prints. I tought 
>>>>>>>>>>>>> that they are helpful to keep, they are using the debug 
>>>>>>>>>>>>> calls, so they are not visible if user does not like 
>>>>>>>>>>>>> those. You are missing the point of who are your users. 
>>>>>>>>>>>>> Users want to have working device and the code. They don't 
>>>>>>>>>>>>> need or like to debug their kernel. Thanks
>>>>>>>>>>>>>
>>>>>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-19 16:00                                         ` Heiner Kallweit
@ 2020-04-19 21:00                                           ` Lauri Jakku
  2020-04-20 19:56                                             ` Lauri Jakku
  2020-05-01 19:12                                           ` Lauri Jakku
  1 sibling, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-04-19 21:00 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd


On 19.4.2020 19.00, Heiner Kallweit wrote:
> On 19.04.2020 18:49, Lauri Jakku wrote:
>> Hi,
>>
>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>
>>>> Hi,
>>>>
>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>> Hi
>>>>>>>>>
>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>> Hi,
>>>>>>>>>>>>
>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>>>>>>>>>
>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>
>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>>>
>>>>>>>>>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>>>>>>
>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>>>>>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>>>>>>
>>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC version: 23
>>>>>>>
>>>>>>> ....
>>>>>>>
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>
>>>>>>> .. after module unload & load cycle:
>>>>>>>
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>
>>>>>>>
>>>>>>> it seem to be the case that the phy_id chances onetime, then stays the same. I'll do few shutdowns and see
>>>>>>>
>>>>>>> is there a pattern at all .. next i'm going to try how it behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>
>>>>>>>
>>>>>>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. If I can't, then I'll do systemd-service that
>>>>>>>
>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>
>>>>>>>
>>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>>
>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY ver"
>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc2077002
>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>
>>>>>> I dont know are those hard coded or what, and are they device specific how much.
>>>>>>
>>>>>> i haven't coldbooted things up, that may be that something to check do they vary how per coldboot.
>>>>>>
>>>>>>>>> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
>>>>>>>>>
>>>>> What i've now learned: the patch + script + journald services -> Results working network, but it is still a workaround.
>>>>>
>>>> Following patch trusts the MAC version, another thing witch could help is to derive method to do 2dn pass of the probeing:
>>>>
>>>> if specific MAC version is found.
>>>>
>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>> index acd122a88d4a..62b37a1abc24 100644
>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>          if (!tp->phydev) {
>>>>                  mdiobus_unregister(new_bus);
>>>>                  return -ENODEV;
>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>> -               /* Most chip versions fail with the genphy driver.
>>>> -                * Therefore ensure that the dedicated PHY driver is loaded.
>>>> -                */
>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>> -               mdiobus_unregister(new_bus);
>>>> -               return -EUNATCH;
>>>> +       } else {
>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
>>>> +
>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>> +                       /* Most chip versions fail with the genphy driver.
>>>> +                        * Therefore ensure that the dedicated PHY driver is loaded.
>>>> +                        */
>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY version.\n", tp->phydev->phy_id);
>>>> +                       mdiobus_unregister(new_bus);
>>>> +                       return -EUNATCH;
>>>> +               }
>>>>          }
>>>>
>>>>          /* PHY will be woken up in rtl_open() */
>>>>
>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's updated.. and  now up'n'running. There is one (WARN_ONCE) stack trace coming from driver, i think i tinker with it next, with above patch the network devices shows up and they can be configured.
>>>
>> I tought to ask first, before going to make new probe_type for errorneus hw (propetype + retry counter) to do re-probe if requested, N times. Or should the r8169_main.c return deferred probe on error on some MAC enums ? Which approach is design-wise sound ?
>>
>> I just tought that the DEFERRED probe may just do the trick i'm looking ways to implement the re-probeing... hmm. I try the deferred thing and check would that help.
>>
> Playing with options to work around the issue is of course a great way to
> learn about the kernel. However it's questionable whether a workaround in
> the driver is acceptable for dealing with the broken BIOS of exactly one

Hmm, I think that it is better to have devices supported, even if the HW 
is old, and cause it has had support.

There are even older harddisk & network drivers, so the argument is not 
good, thus i see no reason why not

to take patch in.


>> 10 yrs old board (for which even a userspace workaround exists).
The age of driver has never been the thing .. there are very old 
drivers, HDD and Network etc.
>>>>>>>>>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>             if (!tp->phydev) {
>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>                     return -ENODEV;
>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>                     /* Most chip versions fail with the genphy driver.
>>>>>>>>>>>>>>>                      * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>                      */
>>>>>>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>                     return -EUNATCH;
>>>>>>>>>>>>>>>             }
>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>        static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>      @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>        static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>               return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>>>>      }
>>>>>>>>>>>>>>>      +
>>>>>>>>>>>>>>>      /**
>>>>>>>>>>>>>>>       * phy_save_page() - take the bus lock and save the current page
>>>>>>>>>>>>>>>       * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>         On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>             Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when i have it working.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>         There's no point in working on your patch. W/o proper justification it
>>>>>>>>>>>>>>>         isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>>>>>>>>>         you actually have.
>>>>>>>>>>>>>>>         FIRST please provide the requested logs and explain the actual problem
>>>>>>>>>>>>>>>         (incl. the commit that caused the regression).
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>             13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>             On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-19 21:00                                           ` Lauri Jakku
@ 2020-04-20 19:56                                             ` Lauri Jakku
  0 siblings, 0 replies; 57+ messages in thread
From: Lauri Jakku @ 2020-04-20 19:56 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,


On 20.4.2020 0.00, Lauri Jakku wrote:
>
> On 19.4.2020 19.00, Heiner Kallweit wrote:
>> On 19.04.2020 18:49, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>>> Hi
>>>>>>>>>>
>>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without 
>>>>>>>>>>>>>>>> modifications --> network does not work
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel 
>>>>>>>>>>>>>>>> (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + 
>>>>>>>>>>>>>>>> r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable 
>>>>>>>>>>>>>>> affected by a known BIOS bug.
>>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>>>> In the first referenced tickets it's about the same 
>>>>>>>>>>>>>>> mainboard (with earlier BIOS version).
>>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the 
>>>>>>>>>>>>>>> network chip / PHY correctly, it reports
>>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver 
>>>>>>>>>>>>>>> being found.
>>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your 
>>>>>>>>>>>>>>> problem should be gone.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>>
>>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing 
>>>>>>>>>>>>> some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>>>>
>>>>>>>>>>>>> that attach has been done. One thing i've noticed is that 
>>>>>>>>>>>>> driver needs to be reloaded to allow traffic (ie. ping 
>>>>>>>>>>>>> works etc.)
>>>>>>>>>>>>>
>>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID 
>>>>>>>>>>>> the PHY comes up.
>>>>>>>>>>>> And what do you mean with "it seems"? Is the option enabled 
>>>>>>>>>>>> or not?
>>>>>>>>>>>>
>>>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then 
>>>>>>>>> you have the same
>>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>>
>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: MAC version: 23
>>>>>>>>
>>>>>>>> ....
>>>>>>>>
>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>
>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>
>>>>>>>> .. after module unload & load cycle:
>>>>>>>>
>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>
>>>>>>>>
>>>>>>>> it seem to be the case that the phy_id chances onetime, then 
>>>>>>>> stays the same. I'll do few shutdowns and see
>>>>>>>>
>>>>>>>> is there a pattern at all .. next i'm going to try how it 
>>>>>>>> behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>>
>>>>>>>>
>>>>>>>> The BIOS downgrade: I'd like to solve this without downgrading 
>>>>>>>> BIOS. If I can't, then I'll do systemd-service that
>>>>>>>>
>>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>>
>>>>>>>>
>>>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>>>
>>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep 
>>>>>>> "PHY ver"
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc2077002
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>
>>>>>>> I dont know are those hard coded or what, and are they device 
>>>>>>> specific how much.
>>>>>>>
>>>>>>> i haven't coldbooted things up, that may be that something to 
>>>>>>> check do they vary how per coldboot.
>>>>>>>
>>>>>>>>>> I check the ID, and revert all other changes, and check how 
>>>>>>>>>> it is working after adding the PHY id to list.
>>>>>>>>>>
>>>>>> What i've now learned: the patch + script + journald services -> 
>>>>>> Results working network, but it is still a workaround.
>>>>>>
>>>>> Following patch trusts the MAC version, another thing witch could 
>>>>> help is to derive method to do 2dn pass of the probeing:
>>>>>
>>>>> if specific MAC version is found.
>>>>>
>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> index acd122a88d4a..62b37a1abc24 100644
>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
>>>>> rtl8169_private *tp)
>>>>>          if (!tp->phydev) {
>>>>>                  mdiobus_unregister(new_bus);
>>>>>                  return -ENODEV;
>>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>> -               /* Most chip versions fail with the genphy driver.
>>>>> -                * Therefore ensure that the dedicated PHY driver 
>>>>> is loaded.
>>>>> -                */
>>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>> -               mdiobus_unregister(new_bus);
>>>>> -               return -EUNATCH;
>>>>> +       } else {
>>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
>>>>> tp->phydev->phy_id);
>>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", 
>>>>> tp->mac_version);
>>>>> +
>>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>> +                       /* Most chip versions fail with the genphy 
>>>>> driver.
>>>>> +                        * Therefore ensure that the dedicated PHY 
>>>>> driver is loaded.
>>>>> +                        */
>>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY 
>>>>> version.\n", tp->phydev->phy_id);
>>>>> +                       mdiobus_unregister(new_bus);
>>>>> +                       return -EUNATCH;
>>>>> +               }
>>>>>          }
>>>>>
>>>>>          /* PHY will be woken up in rtl_open() */
>>>>>
>>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's 
>>>> updated.. and  now up'n'running. There is one (WARN_ONCE) stack 
>>>> trace coming from driver, i think i tinker with it next, with above 
>>>> patch the network devices shows up and they can be configured.
>>>>
>>> I tought to ask first, before going to make new probe_type for 
>>> errorneus hw (propetype + retry counter) to do re-probe if 
>>> requested, N times. Or should the r8169_main.c return deferred probe 
>>> on error on some MAC enums ? Which approach is design-wise sound ?
>>>
>>> I just tought that the DEFERRED probe may just do the trick i'm 
>>> looking ways to implement the re-probeing... hmm. I try the deferred 
>>> thing and check would that help.
>>>
>> Playing with options to work around the issue is of course a great 
>> way to
>> learn about the kernel. However it's questionable whether a 
>> workaround in
>> the driver is acceptable for dealing with the broken BIOS of exactly one
>
> Hmm, I think that it is better to have devices supported, even if the 
> HW is old, and cause it has had support.
>
> There are even older harddisk & network drivers, so the argument is 
> not good, thus i see no reason why not
>
> to take patch in.
>
>
>>> 10 yrs old board (for which even a userspace workaround exists).
> The age of driver has never been the thing .. there are very old 
> drivers, HDD and Network etc.
>>>>>>>>>>>>>>>> The problem with old method seems to be, that device 
>>>>>>>>>>>>>>>> does not have had time to attach before the

I think i have now good combination of driver, that does the old trick 
on real non realtek NICs + userspace added configuration. See at Manjaro 
forum https://forum.manjaro.org/t/r8168-kernel-5-6-3-driver-broken/135325/10

for them for meantime. I think the 'refresh' within kernel, is doable, 
but there is quite much i don't still know how kernel works, so my 
solution might be bit more hacky and slashy,

i try to do some elegant ish solution, but the kernel patch + those 
systemd files are for those by-all-means-hack-fixers.

>>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int 
>>>>>>>>>>>>>>>> r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>>             if (!tp->phydev) {
>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>                     return -ENODEV;
>>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>>                     /* Most chip versions fail with the 
>>>>>>>>>>>>>>>> genphy driver.
>>>>>>>>>>>>>>>>                      * Therefore ensure that the 
>>>>>>>>>>>>>>>> dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>>                      */
>>>>>>>>>>>>>>>> - dev_err(&pdev->dev, "realtek.ko not loaded, maybe it 
>>>>>>>>>>>>>>>> needs to be added to initramfs?\n");
>>>>>>>>>>>>>>>> + dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>                     return -EUNATCH;
>>>>>>>>>>>>>>>>             }
>>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>>        static int __phy_read_page(struct phy_device 
>>>>>>>>>>>>>>>> *phydev)
>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->read_page, 
>>>>>>>>>>>>>>>> "read_page callback not available, PHY driver not 
>>>>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>      @@ -712,12 +716,17 @@ static int 
>>>>>>>>>>>>>>>> __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>        static int __phy_write_page(struct phy_device 
>>>>>>>>>>>>>>>> *phydev, int page)
>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->write_page, 
>>>>>>>>>>>>>>>> "write_page callback not available, PHY driver not 
>>>>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>               return phydev->drv->write_page(phydev, 
>>>>>>>>>>>>>>>> page);
>>>>>>>>>>>>>>>>      }
>>>>>>>>>>>>>>>>      +
>>>>>>>>>>>>>>>>      /**
>>>>>>>>>>>>>>>>       * phy_save_page() - take the bus lock and save 
>>>>>>>>>>>>>>>> the current page
>>>>>>>>>>>>>>>>       * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>>>>>>>> <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> 
>>>>>>>>>>>>>>>> kirjoitti:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>         On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>             Hi, There seems to he Something odd 
>>>>>>>>>>>>>>>> problem, maybe timing related. Stripped version not 
>>>>>>>>>>>>>>>> workingas expected. I get back to you, when i have it 
>>>>>>>>>>>>>>>> working.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>         There's no point in working on your patch. W/o 
>>>>>>>>>>>>>>>> proper justification it
>>>>>>>>>>>>>>>>         isn't acceptable anyway. And so far we still 
>>>>>>>>>>>>>>>> don't know which problem
>>>>>>>>>>>>>>>>         you actually have.
>>>>>>>>>>>>>>>>         FIRST please provide the requested logs and 
>>>>>>>>>>>>>>>> explain the actual problem
>>>>>>>>>>>>>>>>         (incl. the commit that caused the regression).
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>             13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> 
>>>>>>>>>>>>>>>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 
>>>>>>>>>>>>>>>> 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>             On Mon, Apr 13, 2020 at 02:02:01PM +0300, 
>>>>>>>>>>>>>>>> Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 
>>>>>>>>>>>>>>>> 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 
>>>>>>>>>>>>>>>> 01:30:13PM +0300, Lauri Jakku wrote: From 
>>>>>>>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 
>>>>>>>>>>>>>>>> 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 
>>>>>>>>>>>>>>>> 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: 
>>>>>>>>>>>>>>>> r8168/r8169 identifying fix The driver installation 
>>>>>>>>>>>>>>>> determination made properly by checking PHY vs DRIVER 
>>>>>>>>>>>>>>>> id's. --- drivers/net/ethernet/realtek/r8169_main.c | 
>>>>>>>>>>>>>>>> 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 
>>>>>>>>>>>>>>>> 11 +++- 2 files changed, 72 insertions(+), 9 
>>>>>>>>>>>>>>>> deletions(-) I would say that most of the code is debug 
>>>>>>>>>>>>>>>> prints. I tought that they are helpful to keep, they 
>>>>>>>>>>>>>>>> are using the debug calls, so they are not visible if 
>>>>>>>>>>>>>>>> user does not like those. You are missing the point of 
>>>>>>>>>>>>>>>> who are your users. Users want to have working device 
>>>>>>>>>>>>>>>> and the code. They don't need or like to debug their 
>>>>>>>>>>>>>>>> kernel. Thanks
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-04-19 16:00                                         ` Heiner Kallweit
  2020-04-19 21:00                                           ` Lauri Jakku
@ 2020-05-01 19:12                                           ` Lauri Jakku
  2020-05-02 17:56                                             ` Lauri Jakku
  1 sibling, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-05-01 19:12 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,


On 19.4.2020 19.00, Heiner Kallweit wrote:
> On 19.04.2020 18:49, Lauri Jakku wrote:
>> Hi,
>>
>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>
>>>> Hi,
>>>>
>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>> Hi
>>>>>>>>>
>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>> Hi,
>>>>>>>>>>>>
>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>>>>>>>>>
>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>
>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>>>
>>>>>>>>>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>>>>>>
>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>>>>>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>>>>>>
>>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC version: 23
>>>>>>>
>>>>>>> ....
>>>>>>>
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>
>>>>>>> .. after module unload & load cycle:
>>>>>>>
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>
>>>>>>>
>>>>>>> it seem to be the case that the phy_id chances onetime, then stays the same. I'll do few shutdowns and see
>>>>>>>
>>>>>>> is there a pattern at all .. next i'm going to try how it behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>
>>>>>>>
>>>>>>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. If I can't, then I'll do systemd-service that
>>>>>>>
>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>
>>>>>>>
>>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>>
>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY ver"
>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc2077002
>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>
>>>>>> I dont know are those hard coded or what, and are they device specific how much.
>>>>>>
>>>>>> i haven't coldbooted things up, that may be that something to check do they vary how per coldboot.
>>>>>>
>>>>>>>>> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
>>>>>>>>>
>>>>> What i've now learned: the patch + script + journald services -> Results working network, but it is still a workaround.
>>>>>
>>>> Following patch trusts the MAC version, another thing witch could help is to derive method to do 2dn pass of the probeing:
>>>>
>>>> if specific MAC version is found.
>>>>
>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>> index acd122a88d4a..62b37a1abc24 100644
>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>          if (!tp->phydev) {
>>>>                  mdiobus_unregister(new_bus);
>>>>                  return -ENODEV;
>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>> -               /* Most chip versions fail with the genphy driver.
>>>> -                * Therefore ensure that the dedicated PHY driver is loaded.
>>>> -                */
>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>> -               mdiobus_unregister(new_bus);
>>>> -               return -EUNATCH;
>>>> +       } else {
>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
>>>> +
>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>> +                       /* Most chip versions fail with the genphy driver.
>>>> +                        * Therefore ensure that the dedicated PHY driver is loaded.
>>>> +                        */
>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY version.\n", tp->phydev->phy_id);
>>>> +                       mdiobus_unregister(new_bus);
>>>> +                       return -EUNATCH;
>>>> +               }
>>>>          }
>>>>
>>>>          /* PHY will be woken up in rtl_open() */
>>>>
>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's updated.. and  now up'n'running. There is one (WARN_ONCE) stack trace coming from driver, i think i tinker with it next, with above patch the network devices shows up and they can be configured.
>>>
>> I tought to ask first, before going to make new probe_type for errorneus hw (propetype + retry counter) to do re-probe if requested, N times. Or should the r8169_main.c return deferred probe on error on some MAC enums ? Which approach is design-wise sound ?
>>
>> I just tought that the DEFERRED probe may just do the trick i'm looking ways to implement the re-probeing... hmm. I try the deferred thing and check would that help.
>>
> Playing with options to work around the issue is of course a great way to
> learn about the kernel. However it's questionable whether a workaround in
> the driver is acceptable for dealing with the broken BIOS of exactly one
>> 10 yrs old board (for which even a userspace workaround exists).

problem recognized: libphy-module get's unloaded for some reason before 
r8169 driver loads -> missing lowlevel functionality -> not working 
driver. This only occurs at 1st load of module.. seeking solution.


There is [last unloaded: libphy] entries in log BEFORE r8169 is probed 
first time.


Any clue what is responsible for unloading to occur ?


>>>>>>>>>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>             if (!tp->phydev) {
>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>                     return -ENODEV;
>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>                     /* Most chip versions fail with the genphy driver.
>>>>>>>>>>>>>>>                      * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>                      */
>>>>>>>>>>>>>>> -               dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>>>>>>>>> +               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>                     return -EUNATCH;
>>>>>>>>>>>>>>>             }
>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>        static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>      @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>        static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>               return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>>>>      }
>>>>>>>>>>>>>>>      +
>>>>>>>>>>>>>>>      /**
>>>>>>>>>>>>>>>       * phy_save_page() - take the bus lock and save the current page
>>>>>>>>>>>>>>>       * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>         On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>             Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when i have it working.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>         There's no point in working on your patch. W/o proper justification it
>>>>>>>>>>>>>>>         isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>>>>>>>>>         you actually have.
>>>>>>>>>>>>>>>         FIRST please provide the requested logs and explain the actual problem
>>>>>>>>>>>>>>>         (incl. the commit that caused the regression).
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>             13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>             On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-05-01 19:12                                           ` Lauri Jakku
@ 2020-05-02 17:56                                             ` Lauri Jakku
  2020-05-02 22:48                                               ` Lauri Jakku
       [not found]                                               ` <5cdc7f73-b109-2a37-8473-12889506b6a9@pp.inet.fi>
  0 siblings, 2 replies; 57+ messages in thread
From: Lauri Jakku @ 2020-05-02 17:56 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 1.5.2020 22.12, Lauri Jakku wrote:
> Hi,
>
>
> On 19.4.2020 19.00, Heiner Kallweit wrote:
>> On 19.04.2020 18:49, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>>> Hi
>>>>>>>>>>
>>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without 
>>>>>>>>>>>>>>>> modifications --> network does not work
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel 
>>>>>>>>>>>>>>>> (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + 
>>>>>>>>>>>>>>>> r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable 
>>>>>>>>>>>>>>> affected by a known BIOS bug.
>>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>>>> In the first referenced tickets it's about the same 
>>>>>>>>>>>>>>> mainboard (with earlier BIOS version).
>>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the 
>>>>>>>>>>>>>>> network chip / PHY correctly, it reports
>>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver 
>>>>>>>>>>>>>>> being found.
>>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your 
>>>>>>>>>>>>>>> problem should be gone.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>>
>>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing 
>>>>>>>>>>>>> some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>>>>
>>>>>>>>>>>>> that attach has been done. One thing i've noticed is that 
>>>>>>>>>>>>> driver needs to be reloaded to allow traffic (ie. ping 
>>>>>>>>>>>>> works etc.)
>>>>>>>>>>>>>
>>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID 
>>>>>>>>>>>> the PHY comes up.
>>>>>>>>>>>> And what do you mean with "it seems"? Is the option enabled 
>>>>>>>>>>>> or not?
>>>>>>>>>>>>
>>>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then 
>>>>>>>>> you have the same
>>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>>
>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: MAC version: 23
>>>>>>>>
>>>>>>>> ....
>>>>>>>>
>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>
>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>
>>>>>>>> .. after module unload & load cycle:
>>>>>>>>
>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>
>>>>>>>>
>>>>>>>> it seem to be the case that the phy_id chances onetime, then 
>>>>>>>> stays the same. I'll do few shutdowns and see
>>>>>>>>
>>>>>>>> is there a pattern at all .. next i'm going to try how it 
>>>>>>>> behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>>
>>>>>>>>
>>>>>>>> The BIOS downgrade: I'd like to solve this without downgrading 
>>>>>>>> BIOS. If I can't, then I'll do systemd-service that
>>>>>>>>
>>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>>
>>>>>>>>
>>>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>>>
>>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep 
>>>>>>> "PHY ver"
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc2077002
>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>
>>>>>>> I dont know are those hard coded or what, and are they device 
>>>>>>> specific how much.
>>>>>>>
>>>>>>> i haven't coldbooted things up, that may be that something to 
>>>>>>> check do they vary how per coldboot.
>>>>>>>
>>>>>>>>>> I check the ID, and revert all other changes, and check how 
>>>>>>>>>> it is working after adding the PHY id to list.
>>>>>>>>>>
>>>>>> What i've now learned: the patch + script + journald services -> 
>>>>>> Results working network, but it is still a workaround.
>>>>>>
>>>>> Following patch trusts the MAC version, another thing witch could 
>>>>> help is to derive method to do 2dn pass of the probeing:
>>>>>
>>>>> if specific MAC version is found.
>>>>>
>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> index acd122a88d4a..62b37a1abc24 100644
>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
>>>>> rtl8169_private *tp)
>>>>>          if (!tp->phydev) {
>>>>>                  mdiobus_unregister(new_bus);
>>>>>                  return -ENODEV;
>>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>> -               /* Most chip versions fail with the genphy driver.
>>>>> -                * Therefore ensure that the dedicated PHY driver 
>>>>> is loaded.
>>>>> -                */
>>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>> -               mdiobus_unregister(new_bus);
>>>>> -               return -EUNATCH;
>>>>> +       } else {
>>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
>>>>> tp->phydev->phy_id);
>>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", 
>>>>> tp->mac_version);
>>>>> +
>>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>> +                       /* Most chip versions fail with the genphy 
>>>>> driver.
>>>>> +                        * Therefore ensure that the dedicated PHY 
>>>>> driver is loaded.
>>>>> +                        */
>>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY 
>>>>> version.\n", tp->phydev->phy_id);
>>>>> +                       mdiobus_unregister(new_bus);
>>>>> +                       return -EUNATCH;
>>>>> +               }
>>>>>          }
>>>>>
>>>>>          /* PHY will be woken up in rtl_open() */
>>>>>
>>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's 
>>>> updated.. and  now up'n'running. There is one (WARN_ONCE) stack 
>>>> trace coming from driver, i think i tinker with it next, with above 
>>>> patch the network devices shows up and they can be configured.
>>>>
>>> I tought to ask first, before going to make new probe_type for 
>>> errorneus hw (propetype + retry counter) to do re-probe if 
>>> requested, N times. Or should the r8169_main.c return deferred probe 
>>> on error on some MAC enums ? Which approach is design-wise sound ?
>>>
>>> I just tought that the DEFERRED probe may just do the trick i'm 
>>> looking ways to implement the re-probeing... hmm. I try the deferred 
>>> thing and check would that help.
>>>
>> Playing with options to work around the issue is of course a great 
>> way to
>> learn about the kernel. However it's questionable whether a 
>> workaround in
>> the driver is acceptable for dealing with the broken BIOS of exactly one
>>> 10 yrs old board (for which even a userspace workaround exists).
>
> problem recognized: libphy-module get's unloaded for some reason 
> before r8169 driver loads -> missing lowlevel functionality -> not 
> working driver. This only occurs at 1st load of module.. seeking 
> solution.
>
>
> There is [last unloaded: libphy] entries in log BEFORE r8169 is probed 
> first time.
>
>
> Any clue what is responsible for unloading to occur ?
>
>
Right now I'm debugging what is the reason, behind that the module 
starts to work properly only when

unload & reload cycle is done.


The libphy is listed as loaded, but the check for low level read/write 
function is not set -> r8169 modules rlt_open() fails.

See here:

touko 02 20:40:04 MinistryOfSillyWalk kernel: ------------[ cut here 
]------------
touko 02 20:40:04 MinistryOfSillyWalk kernel: read_page callback not 
available, PHY driver not loaded?
touko 02 20:40:04 MinistryOfSillyWalk kernel: WARNING: CPU: 3 PID: 787 
at drivers/net/phy/phy-core.c:750 phy_save_page+0xb1/0xe3 [libph
y]
touko 02 20:40:04 MinistryOfSillyWalk kernel: Modules linked in: cmac 
algif_hash algif_skcipher af_alg bnep btusb btrtl btbcm btintel b
luetooth ecdh_generic rfkill ecc uvcvideo videobuf2_vmalloc 
videobuf2_memops snd_usb_audio videobuf2_v4l2 videobuf2_common snd_usbmidi_
lib videodev snd_rawmidi snd_seq_device mc input_leds joydev mousedev 
squashfs loop amdgpu snd_hda_codec_realtek gpu_sched i2c_algo_bit
  ttm snd_hda_codec_generic drm_kms_helper edac_mce_amd ledtrig_audio 
cec rc_core kvm_amd drm ccp snd_hda_codec_hdmi agpgart rng_core kv
m snd_hda_intel syscopyarea sysfillrect ppdev sysimgblt snd_intel_dspcfg 
fb_sys_fops snd_hda_codec snd_hda_core snd_hwdep irqbypass wmi
_bmof snd_pcm snd_timer snd evdev pcspkr soundcore parport_pc parport 
sp5100_tco mac_hid i2c_piix4 k10temp acpi_cpufreq uinput crypto_u
ser ip_tables x_tables hid_generic usbhid hid ohci_pci virtio_net 
net_failover failover firewire_ohci firewire_core crc_itu_t pata_atii
xp ehci_pci ehci_hcd sr_mod cdrom ohci_hcd ata_generic pata_acpi 
pata_jmicron wmi floppy
touko 02 20:40:04 MinistryOfSillyWalk kernel:  r8169 realtek libphy
touko 02 20:40:04 MinistryOfSillyWalk kernel: CPU: 3 PID: 787 Comm: 
NetworkManager Not tainted 5.7.0-1-raw #12
touko 02 20:40:04 MinistryOfSillyWalk kernel: Hardware name: Gigabyte 
Technology Co., Ltd. GA-MA790FXT-UD5P/GA-MA790FXT-UD5P, BIOS F8l 07/15/2010
touko 02 20:40:04 MinistryOfSillyWalk kernel: RIP: 
0010:phy_save_page+0xb1/0xe3 [libphy]
touko 02 20:40:04 MinistryOfSillyWalk kernel: Code: c8 82 11 c0 e8 06 28 
ff cc 85 db 74 47 48 8b 85 48 03 00 00 48 83 b8 68 01 00 00 00 75 10 48 
c7 c7 e8 82 11 c0 e8 a9 dd f7 cc <0f> 0b eb 26 48 c7 c7 52 78 11 c0 e8 
99 dd f7 cc 0f 0b 48 8b 85 48
touko 02 20:40:04 MinistryOfSillyWalk kernel: RSP: 0018:ffff962c408ef370 
EFLAGS: 00010282
touko 02 20:40:04 MinistryOfSillyWalk kernel: RAX: 0000000000000000 RBX: 
0000000000000001 RCX: 0000000000000000
touko 02 20:40:04 MinistryOfSillyWalk kernel: RDX: 0000000000000001 RSI: 
0000000000000092 RDI: 00000000ffffffff
touko 02 20:40:05 MinistryOfSillyWalk kernel: RBP: ffff8b1af3eb8800 R08: 
00000000000004b3 R09: 0000000000000004
touko 02 20:40:05 MinistryOfSillyWalk kernel: R10: 0000000000000000 R11: 
0000000000000001 R12: 00000000ffffffa1
touko 02 20:40:05 MinistryOfSillyWalk kernel: R13: 0000000000000002 R14: 
0000000000000002 R15: ffff8b1af3eb8800
touko 02 20:40:05 MinistryOfSillyWalk kernel: FS: 00007f07be5d4d80(0000) 
GS:ffff8b1af7cc0000(0000) knlGS:0000000000000000
touko 02 20:40:05 MinistryOfSillyWalk kernel: CS:  0010 DS: 0000 ES: 
0000 CR0: 0000000080050033
touko 02 20:40:05 MinistryOfSillyWalk kernel: CR2: 000055b83aecb008 CR3: 
00000002246b0000 CR4: 00000000000006e0
touko 02 20:40:05 MinistryOfSillyWalk kernel: Call Trace:
touko 02 20:40:05 MinistryOfSillyWalk kernel: phy_select_page+0x53/0x7a 
[libphy]
touko 02 20:40:05 MinistryOfSillyWalk kernel: phy_write_paged+0x5c/0xa0 
[libphy]
touko 02 20:40:05 MinistryOfSillyWalk kernel: 
rtl8168d_1_hw_phy_config+0x9d/0x210 [r8169]
touko 02 20:40:05 MinistryOfSillyWalk kernel: 
rtl8169_init_phy+0x19/0x110 [r8169]
touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl_open+0x354/0x4d0 [r8169]
touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_open+0xe0/0x170
touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_change_flags+0x188/0x1e0
touko 02 20:40:05 MinistryOfSillyWalk kernel: dev_change_flags+0x21/0x60
touko 02 20:40:05 MinistryOfSillyWalk kernel: do_setlink+0x78a/0xfd0


Something does not setup/register properly at first the way it should.


>>>>>>>>>>>>>>>> The problem with old method seems to be, that device 
>>>>>>>>>>>>>>>> does not have had time to attach before the
>>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int 
>>>>>>>>>>>>>>>> r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>>             if (!tp->phydev) {
>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>                     return -ENODEV;
>>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>>                     /* Most chip versions fail with the 
>>>>>>>>>>>>>>>> genphy driver.
>>>>>>>>>>>>>>>>                      * Therefore ensure that the 
>>>>>>>>>>>>>>>> dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>>                      */
>>>>>>>>>>>>>>>> - dev_err(&pdev->dev, "realtek.ko not loaded, maybe it 
>>>>>>>>>>>>>>>> needs to be added to initramfs?\n");
>>>>>>>>>>>>>>>> + dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>                     return -EUNATCH;
>>>>>>>>>>>>>>>>             }
>>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>>        static int __phy_read_page(struct phy_device 
>>>>>>>>>>>>>>>> *phydev)
>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->read_page, 
>>>>>>>>>>>>>>>> "read_page callback not available, PHY driver not 
>>>>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>      @@ -712,12 +716,17 @@ static int 
>>>>>>>>>>>>>>>> __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>        static int __phy_write_page(struct phy_device 
>>>>>>>>>>>>>>>> *phydev, int page)
>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->write_page, 
>>>>>>>>>>>>>>>> "write_page callback not available, PHY driver not 
>>>>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>               return phydev->drv->write_page(phydev, 
>>>>>>>>>>>>>>>> page);
>>>>>>>>>>>>>>>>      }
>>>>>>>>>>>>>>>>      +
>>>>>>>>>>>>>>>>      /**
>>>>>>>>>>>>>>>>       * phy_save_page() - take the bus lock and save 
>>>>>>>>>>>>>>>> the current page
>>>>>>>>>>>>>>>>       * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>>>>>>>> <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> 
>>>>>>>>>>>>>>>> kirjoitti:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>         On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>             Hi, There seems to he Something odd 
>>>>>>>>>>>>>>>> problem, maybe timing related. Stripped version not 
>>>>>>>>>>>>>>>> workingas expected. I get back to you, when i have it 
>>>>>>>>>>>>>>>> working.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>         There's no point in working on your patch. W/o 
>>>>>>>>>>>>>>>> proper justification it
>>>>>>>>>>>>>>>>         isn't acceptable anyway. And so far we still 
>>>>>>>>>>>>>>>> don't know which problem
>>>>>>>>>>>>>>>>         you actually have.
>>>>>>>>>>>>>>>>         FIRST please provide the requested logs and 
>>>>>>>>>>>>>>>> explain the actual problem
>>>>>>>>>>>>>>>>         (incl. the commit that caused the regression).
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>             13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> 
>>>>>>>>>>>>>>>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 
>>>>>>>>>>>>>>>> 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>             On Mon, Apr 13, 2020 at 02:02:01PM +0300, 
>>>>>>>>>>>>>>>> Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 
>>>>>>>>>>>>>>>> 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 
>>>>>>>>>>>>>>>> 01:30:13PM +0300, Lauri Jakku wrote: From 
>>>>>>>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 
>>>>>>>>>>>>>>>> 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 
>>>>>>>>>>>>>>>> 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: 
>>>>>>>>>>>>>>>> r8168/r8169 identifying fix The driver installation 
>>>>>>>>>>>>>>>> determination made properly by checking PHY vs DRIVER 
>>>>>>>>>>>>>>>> id's. --- drivers/net/ethernet/realtek/r8169_main.c | 
>>>>>>>>>>>>>>>> 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 
>>>>>>>>>>>>>>>> 11 +++- 2 files changed, 72 insertions(+), 9 
>>>>>>>>>>>>>>>> deletions(-) I would say that most of the code is debug 
>>>>>>>>>>>>>>>> prints. I tought that they are helpful to keep, they 
>>>>>>>>>>>>>>>> are using the debug calls, so they are not visible if 
>>>>>>>>>>>>>>>> user does not like those. You are missing the point of 
>>>>>>>>>>>>>>>> who are your users. Users want to have working device 
>>>>>>>>>>>>>>>> and the code. They don't need or like to debug their 
>>>>>>>>>>>>>>>> kernel. Thanks
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-05-02 17:56                                             ` Lauri Jakku
@ 2020-05-02 22:48                                               ` Lauri Jakku
       [not found]                                               ` <5cdc7f73-b109-2a37-8473-12889506b6a9@pp.inet.fi>
  1 sibling, 0 replies; 57+ messages in thread
From: Lauri Jakku @ 2020-05-02 22:48 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,



On 2.5.2020 20.56, Lauri Jakku wrote:
> Hi,
>
> On 1.5.2020 22.12, Lauri Jakku wrote:
>> Hi,
>>
>>
>> On 19.4.2020 19.00, Heiner Kallweit wrote:
>>> On 19.04.2020 18:49, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>>>> Hi
>>>>>>>>>>>
>>>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without 
>>>>>>>>>>>>>>>>> modifications --> network does not work
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel 
>>>>>>>>>>>>>>>>> (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + 
>>>>>>>>>>>>>>>>> r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're 
>>>>>>>>>>>>>>>> presumable affected by a known BIOS bug.
>>>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>>>>> In the first referenced tickets it's about the same 
>>>>>>>>>>>>>>>> mainboard (with earlier BIOS version).
>>>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the 
>>>>>>>>>>>>>>>> network chip / PHY correctly, it reports
>>>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver 
>>>>>>>>>>>>>>>> being found.
>>>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your 
>>>>>>>>>>>>>>>> problem should be gone.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now 
>>>>>>>>>>>>>> testing some mutex guard for phy state and try to use it 
>>>>>>>>>>>>>> as indicator
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> that attach has been done. One thing i've noticed is that 
>>>>>>>>>>>>>> driver needs to be reloaded to allow traffic (ie. ping 
>>>>>>>>>>>>>> works etc.)
>>>>>>>>>>>>>>
>>>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID 
>>>>>>>>>>>>> the PHY comes up.
>>>>>>>>>>>>> And what do you mean with "it seems"? Is the option 
>>>>>>>>>>>>> enabled or not?
>>>>>>>>>>>>>
>>>>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then 
>>>>>>>>>> you have the same
>>>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>>>
>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>> 0000:03:00.0: MAC version: 23
>>>>>>>>>
>>>>>>>>> ....
>>>>>>>>>
>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>
>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>>
>>>>>>>>> .. after module unload & load cycle:
>>>>>>>>>
>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> it seem to be the case that the phy_id chances onetime, then 
>>>>>>>>> stays the same. I'll do few shutdowns and see
>>>>>>>>>
>>>>>>>>> is there a pattern at all .. next i'm going to try how it 
>>>>>>>>> behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> The BIOS downgrade: I'd like to solve this without downgrading 
>>>>>>>>> BIOS. If I can't, then I'll do systemd-service that
>>>>>>>>>
>>>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>>>
>>>>>>>>>
>>>>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>>>>
>>>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl 
>>>>>>>> |grep "PHY ver"
>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0xc2077002
>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>
>>>>>>>> I dont know are those hard coded or what, and are they device 
>>>>>>>> specific how much.
>>>>>>>>
>>>>>>>> i haven't coldbooted things up, that may be that something to 
>>>>>>>> check do they vary how per coldboot.
>>>>>>>>
>>>>>>>>>>> I check the ID, and revert all other changes, and check how 
>>>>>>>>>>> it is working after adding the PHY id to list.
>>>>>>>>>>>
>>>>>>> What i've now learned: the patch + script + journald services -> 
>>>>>>> Results working network, but it is still a workaround.
>>>>>>>
>>>>>> Following patch trusts the MAC version, another thing witch could 
>>>>>> help is to derive method to do 2dn pass of the probeing:
>>>>>>
>>>>>> if specific MAC version is found.
>>>>>>
>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> index acd122a88d4a..62b37a1abc24 100644
>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
>>>>>> rtl8169_private *tp)
>>>>>>          if (!tp->phydev) {
>>>>>>                  mdiobus_unregister(new_bus);
>>>>>>                  return -ENODEV;
>>>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>> -               /* Most chip versions fail with the genphy driver.
>>>>>> -                * Therefore ensure that the dedicated PHY driver 
>>>>>> is loaded.
>>>>>> -                */
>>>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>> -               mdiobus_unregister(new_bus);
>>>>>> -               return -EUNATCH;
>>>>>> +       } else {
>>>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
>>>>>> tp->phydev->phy_id);
>>>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", 
>>>>>> tp->mac_version);
>>>>>> +
>>>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>> +                       /* Most chip versions fail with the 
>>>>>> genphy driver.
>>>>>> +                        * Therefore ensure that the dedicated 
>>>>>> PHY driver is loaded.
>>>>>> +                        */
>>>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY 
>>>>>> version.\n", tp->phydev->phy_id);
>>>>>> +                       mdiobus_unregister(new_bus);
>>>>>> +                       return -EUNATCH;
>>>>>> +               }
>>>>>>          }
>>>>>>
>>>>>>          /* PHY will be woken up in rtl_open() */
>>>>>>
>>>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's 
>>>>> updated.. and  now up'n'running. There is one (WARN_ONCE) stack 
>>>>> trace coming from driver, i think i tinker with it next, with 
>>>>> above patch the network devices shows up and they can be configured.
>>>>>
>>>> I tought to ask first, before going to make new probe_type for 
>>>> errorneus hw (propetype + retry counter) to do re-probe if 
>>>> requested, N times. Or should the r8169_main.c return deferred 
>>>> probe on error on some MAC enums ? Which approach is design-wise 
>>>> sound ?
>>>>
>>>> I just tought that the DEFERRED probe may just do the trick i'm 
>>>> looking ways to implement the re-probeing... hmm. I try the 
>>>> deferred thing and check would that help.
>>>>
>>> Playing with options to work around the issue is of course a great 
>>> way to
>>> learn about the kernel. However it's questionable whether a 
>>> workaround in
>>> the driver is acceptable for dealing with the broken BIOS of exactly one
>>>> 10 yrs old board (for which even a userspace workaround exists).
>>
>> problem recognized: libphy-module get's unloaded for some reason 
>> before r8169 driver loads -> missing lowlevel functionality -> not 
>> working driver. This only occurs at 1st load of module.. seeking 
>> solution.
>>
>>
>> There is [last unloaded: libphy] entries in log BEFORE r8169 is 
>> probed first time.
>>
>>
>> Any clue what is responsible for unloading to occur ?
>>
>>
> Right now I'm debugging what is the reason, behind that the module 
> starts to work properly only when
>
> unload & reload cycle is done.
>
>
> The libphy is listed as loaded, but the check for low level read/write 
> function is not set -> r8169 modules rlt_open() fails.
>
> See here:
>
> touko 02 20:40:04 MinistryOfSillyWalk kernel: ------------[ cut here 
> ]------------
> touko 02 20:40:04 MinistryOfSillyWalk kernel: read_page callback not 
> available, PHY driver not loaded?
> touko 02 20:40:04 MinistryOfSillyWalk kernel: WARNING: CPU: 3 PID: 787 
> at drivers/net/phy/phy-core.c:750 phy_save_page+0xb1/0xe3 [libph
> y]
> touko 02 20:40:04 MinistryOfSillyWalk kernel: Modules linked in: cmac 
> algif_hash algif_skcipher af_alg bnep btusb btrtl btbcm btintel b
> luetooth ecdh_generic rfkill ecc uvcvideo videobuf2_vmalloc 
> videobuf2_memops snd_usb_audio videobuf2_v4l2 videobuf2_common 
> snd_usbmidi_
> lib videodev snd_rawmidi snd_seq_device mc input_leds joydev mousedev 
> squashfs loop amdgpu snd_hda_codec_realtek gpu_sched i2c_algo_bit
>  ttm snd_hda_codec_generic drm_kms_helper edac_mce_amd ledtrig_audio 
> cec rc_core kvm_amd drm ccp snd_hda_codec_hdmi agpgart rng_core kv
> m snd_hda_intel syscopyarea sysfillrect ppdev sysimgblt 
> snd_intel_dspcfg fb_sys_fops snd_hda_codec snd_hda_core snd_hwdep 
> irqbypass wmi
> _bmof snd_pcm snd_timer snd evdev pcspkr soundcore parport_pc parport 
> sp5100_tco mac_hid i2c_piix4 k10temp acpi_cpufreq uinput crypto_u
> ser ip_tables x_tables hid_generic usbhid hid ohci_pci virtio_net 
> net_failover failover firewire_ohci firewire_core crc_itu_t pata_atii
> xp ehci_pci ehci_hcd sr_mod cdrom ohci_hcd ata_generic pata_acpi 
> pata_jmicron wmi floppy
> touko 02 20:40:04 MinistryOfSillyWalk kernel:  r8169 realtek libphy
> touko 02 20:40:04 MinistryOfSillyWalk kernel: CPU: 3 PID: 787 Comm: 
> NetworkManager Not tainted 5.7.0-1-raw #12
> touko 02 20:40:04 MinistryOfSillyWalk kernel: Hardware name: Gigabyte 
> Technology Co., Ltd. GA-MA790FXT-UD5P/GA-MA790FXT-UD5P, BIOS F8l 
> 07/15/2010
> touko 02 20:40:04 MinistryOfSillyWalk kernel: RIP: 
> 0010:phy_save_page+0xb1/0xe3 [libphy]
> touko 02 20:40:04 MinistryOfSillyWalk kernel: Code: c8 82 11 c0 e8 06 
> 28 ff cc 85 db 74 47 48 8b 85 48 03 00 00 48 83 b8 68 01 00 00 00 75 
> 10 48 c7 c7 e8 82 11 c0 e8 a9 dd f7 cc <0f> 0b eb 26 48 c7 c7 52 78 11 
> c0 e8 99 dd f7 cc 0f 0b 48 8b 85 48
> touko 02 20:40:04 MinistryOfSillyWalk kernel: RSP: 
> 0018:ffff962c408ef370 EFLAGS: 00010282
> touko 02 20:40:04 MinistryOfSillyWalk kernel: RAX: 0000000000000000 
> RBX: 0000000000000001 RCX: 0000000000000000
> touko 02 20:40:04 MinistryOfSillyWalk kernel: RDX: 0000000000000001 
> RSI: 0000000000000092 RDI: 00000000ffffffff
> touko 02 20:40:05 MinistryOfSillyWalk kernel: RBP: ffff8b1af3eb8800 
> R08: 00000000000004b3 R09: 0000000000000004
> touko 02 20:40:05 MinistryOfSillyWalk kernel: R10: 0000000000000000 
> R11: 0000000000000001 R12: 00000000ffffffa1
> touko 02 20:40:05 MinistryOfSillyWalk kernel: R13: 0000000000000002 
> R14: 0000000000000002 R15: ffff8b1af3eb8800
> touko 02 20:40:05 MinistryOfSillyWalk kernel: FS: 
> 00007f07be5d4d80(0000) GS:ffff8b1af7cc0000(0000) knlGS:0000000000000000
> touko 02 20:40:05 MinistryOfSillyWalk kernel: CS:  0010 DS: 0000 ES: 
> 0000 CR0: 0000000080050033
> touko 02 20:40:05 MinistryOfSillyWalk kernel: CR2: 000055b83aecb008 
> CR3: 00000002246b0000 CR4: 00000000000006e0
> touko 02 20:40:05 MinistryOfSillyWalk kernel: Call Trace:
> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
> phy_select_page+0x53/0x7a [libphy]
> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
> phy_write_paged+0x5c/0xa0 [libphy]
> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
> rtl8168d_1_hw_phy_config+0x9d/0x210 [r8169]
> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
> rtl8169_init_phy+0x19/0x110 [r8169]
> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl_open+0x354/0x4d0 [r8169]
> touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_open+0xe0/0x170
> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
> __dev_change_flags+0x188/0x1e0
> touko 02 20:40:05 MinistryOfSillyWalk kernel: dev_change_flags+0x21/0x60
> touko 02 20:40:05 MinistryOfSillyWalk kernel: do_setlink+0x78a/0xfd0
>
>
> Something does not setup/register properly at first the way it should.
>
>

I think I solved it: realtek (the phy driver) was missing device entry 
for the PHY ID reported by the NIC to match -> read_page and write_page 
function pointers should now be set. The generic PHY does not fill the 
driver's functionality to read or write pages. It happens to be so that 
the drivers for RTL8211B Gigabit Ethernet seems to work just fine for my 
NIC's.

I've documented current state to Manjaro's forum: 
https://forum.manjaro.org/t/your-post-in-r8168-kernel-5-6-3-driver-broken/139869/3 
.. after i cleanup debugs away and do speedtests.


Could the fix be something to considerate to upstream, the fix I'm 
proposing is not the reload hack, but just the kernel modifications?



>>>>>>>>>>>>>>>>> The problem with old method seems to be, that device 
>>>>>>>>>>>>>>>>> does not have had time to attach before the
>>>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int 
>>>>>>>>>>>>>>>>> r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>>>             if (!tp->phydev) {
>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>                     return -ENODEV;
>>>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>>>                     /* Most chip versions fail with 
>>>>>>>>>>>>>>>>> the genphy driver.
>>>>>>>>>>>>>>>>>                      * Therefore ensure that the 
>>>>>>>>>>>>>>>>> dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>>>                      */
>>>>>>>>>>>>>>>>> - dev_err(&pdev->dev, "realtek.ko not loaded, maybe it 
>>>>>>>>>>>>>>>>> needs to be added to initramfs?\n");
>>>>>>>>>>>>>>>>> + dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>                     return -EUNATCH;
>>>>>>>>>>>>>>>>>             }
>>>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>>>        static int __phy_read_page(struct phy_device 
>>>>>>>>>>>>>>>>> *phydev)
>>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->read_page, 
>>>>>>>>>>>>>>>>> "read_page callback not available, PHY driver not 
>>>>>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>      @@ -712,12 +716,17 @@ static int 
>>>>>>>>>>>>>>>>> __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>>        static int __phy_write_page(struct phy_device 
>>>>>>>>>>>>>>>>> *phydev, int page)
>>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->write_page, 
>>>>>>>>>>>>>>>>> "write_page callback not available, PHY driver not 
>>>>>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>               return phydev->drv->write_page(phydev, 
>>>>>>>>>>>>>>>>> page);
>>>>>>>>>>>>>>>>>      }
>>>>>>>>>>>>>>>>>      +
>>>>>>>>>>>>>>>>>      /**
>>>>>>>>>>>>>>>>>       * phy_save_page() - take the bus lock and save 
>>>>>>>>>>>>>>>>> the current page
>>>>>>>>>>>>>>>>>       * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>>>>>>>>> <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> 
>>>>>>>>>>>>>>>>> kirjoitti:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>         On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>             Hi, There seems to he Something odd 
>>>>>>>>>>>>>>>>> problem, maybe timing related. Stripped version not 
>>>>>>>>>>>>>>>>> workingas expected. I get back to you, when i have it 
>>>>>>>>>>>>>>>>> working.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>         There's no point in working on your patch. W/o 
>>>>>>>>>>>>>>>>> proper justification it
>>>>>>>>>>>>>>>>>         isn't acceptable anyway. And so far we still 
>>>>>>>>>>>>>>>>> don't know which problem
>>>>>>>>>>>>>>>>>         you actually have.
>>>>>>>>>>>>>>>>>         FIRST please provide the requested logs and 
>>>>>>>>>>>>>>>>> explain the actual problem
>>>>>>>>>>>>>>>>>         (incl. the commit that caused the regression).
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>             13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> 
>>>>>>>>>>>>>>>>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 
>>>>>>>>>>>>>>>>> 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>             On Mon, Apr 13, 2020 at 02:02:01PM +0300, 
>>>>>>>>>>>>>>>>> Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 
>>>>>>>>>>>>>>>>> 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 
>>>>>>>>>>>>>>>>> 01:30:13PM +0300, Lauri Jakku wrote: From 
>>>>>>>>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 
>>>>>>>>>>>>>>>>> 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: 
>>>>>>>>>>>>>>>>> Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: 
>>>>>>>>>>>>>>>>> r8168/r8169 identifying fix The driver installation 
>>>>>>>>>>>>>>>>> determination made properly by checking PHY vs DRIVER 
>>>>>>>>>>>>>>>>> id's. --- drivers/net/ethernet/realtek/r8169_main.c | 
>>>>>>>>>>>>>>>>> 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c 
>>>>>>>>>>>>>>>>> | 11 +++- 2 files changed, 72 insertions(+), 9 
>>>>>>>>>>>>>>>>> deletions(-) I would say that most of the code is 
>>>>>>>>>>>>>>>>> debug prints. I tought that they are helpful to keep, 
>>>>>>>>>>>>>>>>> they are using the debug calls, so they are not 
>>>>>>>>>>>>>>>>> visible if user does not like those. You are missing 
>>>>>>>>>>>>>>>>> the point of who are your users. Users want to have 
>>>>>>>>>>>>>>>>> working device and the code. They don't need or like 
>>>>>>>>>>>>>>>>> to debug their kernel. Thanks
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
       [not found]                                               ` <5cdc7f73-b109-2a37-8473-12889506b6a9@pp.inet.fi>
@ 2020-05-02 23:15                                                 ` Heiner Kallweit
  2020-05-03  0:11                                                   ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Heiner Kallweit @ 2020-05-02 23:15 UTC (permalink / raw)
  To: Lauri Jakku; +Cc: Leon Romanovsky, netdev, nic_swsd

On 03.05.2020 00:42, Lauri Jakku wrote:
> Hi,
> 
> 
> 
> On 2.5.2020 20.56, Lauri Jakku wrote:
>> Hi,
>>
>> On 1.5.2020 22.12, Lauri Jakku wrote:
>>> Hi,
>>>
>>>
>>> On 19.4.2020 19.00, Heiner Kallweit wrote:
>>>> On 19.04.2020 18:49, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>>>>> Hi
>>>>>>>>>>>>
>>>>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>>>>>>>>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>>>>>>>>>
>>>>>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
>>>>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>>>>
>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC version: 23
>>>>>>>>>>
>>>>>>>>>> ....
>>>>>>>>>>
>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>
>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>>>>
>>>>>>>>>> .. after module unload & load cycle:
>>>>>>>>>>
>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> it seem to be the case that the phy_id chances onetime, then stays the same. I'll do few shutdowns and see
>>>>>>>>>>
>>>>>>>>>> is there a pattern at all .. next i'm going to try how it behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. If I can't, then I'll do systemd-service that
>>>>>>>>>>
>>>>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>>>>>
>>>>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY ver"
>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc2077002
>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>
>>>>>>>>> I dont know are those hard coded or what, and are they device specific how much.
>>>>>>>>>
>>>>>>>>> i haven't coldbooted things up, that may be that something to check do they vary how per coldboot.
>>>>>>>>>
>>>>>>>>>>>> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
>>>>>>>>>>>>
>>>>>>>> What i've now learned: the patch + script + journald services -> Results working network, but it is still a workaround.
>>>>>>>>
>>>>>>> Following patch trusts the MAC version, another thing witch could help is to derive method to do 2dn pass of the probeing:
>>>>>>>
>>>>>>> if specific MAC version is found.
>>>>>>>
>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> index acd122a88d4a..62b37a1abc24 100644
>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>          if (!tp->phydev) {
>>>>>>>                  mdiobus_unregister(new_bus);
>>>>>>>                  return -ENODEV;
>>>>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>> -               /* Most chip versions fail with the genphy driver.
>>>>>>> -                * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>> -                */
>>>>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>> -               mdiobus_unregister(new_bus);
>>>>>>> -               return -EUNATCH;
>>>>>>> +       } else {
>>>>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
>>>>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
>>>>>>> +
>>>>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>> +                       /* Most chip versions fail with the genphy driver.
>>>>>>> +                        * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>> +                        */
>>>>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY version.\n", tp->phydev->phy_id);
>>>>>>> +                       mdiobus_unregister(new_bus);
>>>>>>> +                       return -EUNATCH;
>>>>>>> +               }
>>>>>>>          }
>>>>>>>
>>>>>>>          /* PHY will be woken up in rtl_open() */
>>>>>>>
>>>>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's updated.. and  now up'n'running. There is one (WARN_ONCE) stack trace coming from driver, i think i tinker with it next, with above patch the network devices shows up and they can be configured.
>>>>>>
>>>>> I tought to ask first, before going to make new probe_type for errorneus hw (propetype + retry counter) to do re-probe if requested, N times. Or should the r8169_main.c return deferred probe on error on some MAC enums ? Which approach is design-wise sound ?
>>>>>
>>>>> I just tought that the DEFERRED probe may just do the trick i'm looking ways to implement the re-probeing... hmm. I try the deferred thing and check would that help.
>>>>>
>>>> Playing with options to work around the issue is of course a great way to
>>>> learn about the kernel. However it's questionable whether a workaround in
>>>> the driver is acceptable for dealing with the broken BIOS of exactly one
>>>>> 10 yrs old board (for which even a userspace workaround exists).
>>>
>>> problem recognized: libphy-module get's unloaded for some reason before r8169 driver loads -> missing lowlevel functionality -> not working driver. This only occurs at 1st load of module.. seeking solution.
>>>
>>>
>>> There is [last unloaded: libphy] entries in log BEFORE r8169 is probed first time.
>>>
>>>
>>> Any clue what is responsible for unloading to occur ?
>>>
>>>
>> Right now I'm debugging what is the reason, behind that the module starts to work properly only when
>>
>> unload & reload cycle is done.
>>
>>
>> The libphy is listed as loaded, but the check for low level read/write function is not set -> r8169 modules rlt_open() fails.
>>
>> See here:
>>
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: ------------[ cut here ]------------
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: read_page callback not available, PHY driver not loaded?
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: WARNING: CPU: 3 PID: 787 at drivers/net/phy/phy-core.c:750 phy_save_page+0xb1/0xe3 [libph
>> y]
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Modules linked in: cmac algif_hash algif_skcipher af_alg bnep btusb btrtl btbcm btintel b
>> luetooth ecdh_generic rfkill ecc uvcvideo videobuf2_vmalloc videobuf2_memops snd_usb_audio videobuf2_v4l2 videobuf2_common snd_usbmidi_
>> lib videodev snd_rawmidi snd_seq_device mc input_leds joydev mousedev squashfs loop amdgpu snd_hda_codec_realtek gpu_sched i2c_algo_bit
>>  ttm snd_hda_codec_generic drm_kms_helper edac_mce_amd ledtrig_audio cec rc_core kvm_amd drm ccp snd_hda_codec_hdmi agpgart rng_core kv
>> m snd_hda_intel syscopyarea sysfillrect ppdev sysimgblt snd_intel_dspcfg fb_sys_fops snd_hda_codec snd_hda_core snd_hwdep irqbypass wmi
>> _bmof snd_pcm snd_timer snd evdev pcspkr soundcore parport_pc parport sp5100_tco mac_hid i2c_piix4 k10temp acpi_cpufreq uinput crypto_u
>> ser ip_tables x_tables hid_generic usbhid hid ohci_pci virtio_net net_failover failover firewire_ohci firewire_core crc_itu_t pata_atii
>> xp ehci_pci ehci_hcd sr_mod cdrom ohci_hcd ata_generic pata_acpi pata_jmicron wmi floppy
>> touko 02 20:40:04 MinistryOfSillyWalk kernel:  r8169 realtek libphy
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: CPU: 3 PID: 787 Comm: NetworkManager Not tainted 5.7.0-1-raw #12
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Hardware name: Gigabyte Technology Co., Ltd. GA-MA790FXT-UD5P/GA-MA790FXT-UD5P, BIOS F8l 07/15/2010
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RIP: 0010:phy_save_page+0xb1/0xe3 [libphy]
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Code: c8 82 11 c0 e8 06 28 ff cc 85 db 74 47 48 8b 85 48 03 00 00 48 83 b8 68 01 00 00 00 75 10 48 c7 c7 e8 82 11 c0 e8 a9 dd f7 cc <0f> 0b eb 26 48 c7 c7 52 78 11 c0 e8 99 dd f7 cc 0f 0b 48 8b 85 48
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RSP: 0018:ffff962c408ef370 EFLAGS: 00010282
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RAX: 0000000000000000 RBX: 0000000000000001 RCX: 0000000000000000
>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RDX: 0000000000000001 RSI: 0000000000000092 RDI: 00000000ffffffff
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: RBP: ffff8b1af3eb8800 R08: 00000000000004b3 R09: 0000000000000004
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R10: 0000000000000000 R11: 0000000000000001 R12: 00000000ffffffa1
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R13: 0000000000000002 R14: 0000000000000002 R15: ffff8b1af3eb8800
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: FS: 00007f07be5d4d80(0000) GS:ffff8b1af7cc0000(0000) knlGS:0000000000000000
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CR2: 000055b83aecb008 CR3: 00000002246b0000 CR4: 00000000000006e0
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: Call Trace:
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: phy_select_page+0x53/0x7a [libphy]
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: phy_write_paged+0x5c/0xa0 [libphy]
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl8168d_1_hw_phy_config+0x9d/0x210 [r8169]
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl8169_init_phy+0x19/0x110 [r8169]
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl_open+0x354/0x4d0 [r8169]
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_open+0xe0/0x170
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_change_flags+0x188/0x1e0
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: dev_change_flags+0x21/0x60
>> touko 02 20:40:05 MinistryOfSillyWalk kernel: do_setlink+0x78a/0xfd0
>>
>>
>> Something does not setup/register properly at first the way it should.
>>
>>
> 
> I think i solved it: realtek (the phy driver) was missing device entry for the PHY ID reported by the NIC to match -> read_page and write_page function pointers should now be set. The generic PHY does not fill
> 
> the driver's functionality to read or write pages. It happens to be so that the drivers for |RTL8211B Gigabit Ethernet seems to work just fine for my NIC's.

The analysis is wrong. The incorrect PHY ID is not root cause of the
problem, it's caused by a BIOS bug. It's not a valid PHY ID. If you want
to do something, then you could try to inject a PHY soft reset before
the MII bus is registered. This should be board-specific, e.g. using
dmi_check_system().

> |
> 
> |
> |
> 
> |I've documented current state to Manjaro's forum: https://forum.manjaro.org/t/your-post-in-r8168-kernel-5-6-3-driver-broken/139869/3 .. after i cleanup debugs away and do speedtests. Could the fix be something to considerate to upstream ?|
> 
> |
> |
> 
> |The fix I'm proposing is not the reload hack, but just kernel modifications. 
> |
> 
> 
>>>>>>>>>>>>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>>>>             if (!tp->phydev) {
>>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>>                     return -ENODEV;
>>>>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>>>>                     /* Most chip versions fail with the genphy driver.
>>>>>>>>>>>>>>>>>>                      * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>>>>                      */
>>>>>>>>>>>>>>>>>> - dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>>>>>>>>>>>> + dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>>                     return -EUNATCH;
>>>>>>>>>>>>>>>>>>             }
>>>>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>>>>        static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>      @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>>>        static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>>>>>>>>>>>>      {
>>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>>             if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>>>>                     return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>               return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>>>>>>>      }
>>>>>>>>>>>>>>>>>>      +
>>>>>>>>>>>>>>>>>>      /**
>>>>>>>>>>>>>>>>>>       * phy_save_page() - take the bus lock and save the current page
>>>>>>>>>>>>>>>>>>       * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>         On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>             Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when i have it working.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>         There's no point in working on your patch. W/o proper justification it
>>>>>>>>>>>>>>>>>>         isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>>>>>>>>>>>>         you actually have.
>>>>>>>>>>>>>>>>>>         FIRST please provide the requested logs and explain the actual problem
>>>>>>>>>>>>>>>>>>         (incl. the commit that caused the regression).
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>             13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>             On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-05-02 23:15                                                 ` Heiner Kallweit
@ 2020-05-03  0:11                                                   ` Lauri Jakku
  2020-05-03  1:34                                                     ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-05-03  0:11 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd


On 3.5.2020 2.15, Heiner Kallweit wrote:
> On 03.05.2020 00:42, Lauri Jakku wrote:
>> Hi,
>>
>>
>>
>> On 2.5.2020 20.56, Lauri Jakku wrote:
>>> Hi,
>>>
>>> On 1.5.2020 22.12, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>>
>>>> On 19.4.2020 19.00, Heiner Kallweit wrote:
>>>>> On 19.04.2020 18:49, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>>>>>> Hi
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without modifications --> network does not work
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified kernel (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're presumable affected by a known BIOS bug.
>>>>>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at bugzilla.kernel.org.
>>>>>>>>>>>>>>>>>> In the first referenced tickets it's about the same mainboard (with earlier BIOS version).
>>>>>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the network chip / PHY correctly, it reports
>>>>>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY driver being found.
>>>>>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your problem should be gone.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now testing some mutex guard for phy state and try to use it as indicator
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> that attach has been done. One thing i've noticed is that driver needs to be reloaded to allow traffic (ie. ping works etc.)
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY ID the PHY comes up.
>>>>>>>>>>>>>>> And what do you mean with "it seems"? Is the option enabled or not?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I do have ROM's enabled, and it does not help with my issue.
>>>>>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. Then you have the same
>>>>>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>>>>>
>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: MAC version: 23
>>>>>>>>>>>
>>>>>>>>>>> ....
>>>>>>>>>>>
>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>
>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>>>>>
>>>>>>>>>>> .. after module unload & load cycle:
>>>>>>>>>>>
>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: MAC version: 23
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> it seem to be the case that the phy_id chances onetime, then stays the same. I'll do few shutdowns and see
>>>>>>>>>>>
>>>>>>>>>>> is there a pattern at all .. next i'm going to try how it behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> The BIOS downgrade: I'd like to solve this without downgrading BIOS. If I can't, then I'll do systemd-service that
>>>>>>>>>>>
>>>>>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>> What i've gathered samples now, there is three values for PHY ID:
>>>>>>>>>>
>>>>>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl |grep "PHY ver"
>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc2077002
>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>
>>>>>>>>>> I dont know are those hard coded or what, and are they device specific how much.
>>>>>>>>>>
>>>>>>>>>> i haven't coldbooted things up, that may be that something to check do they vary how per coldboot.
>>>>>>>>>>
>>>>>>>>>>>>> I check the ID, and revert all other changes, and check how it is working after adding the PHY id to list.
>>>>>>>>>>>>>
>>>>>>>>> What i've now learned: the patch + script + journald services -> Results working network, but it is still a workaround.
>>>>>>>>>
>>>>>>>> Following patch trusts the MAC version, another thing witch could help is to derive method to do 2dn pass of the probeing:
>>>>>>>>
>>>>>>>> if specific MAC version is found.
>>>>>>>>
>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> index acd122a88d4a..62b37a1abc24 100644
>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>           if (!tp->phydev) {
>>>>>>>>                   mdiobus_unregister(new_bus);
>>>>>>>>                   return -ENODEV;
>>>>>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>> -               /* Most chip versions fail with the genphy driver.
>>>>>>>> -                * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>> -                */
>>>>>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>> -               mdiobus_unregister(new_bus);
>>>>>>>> -               return -EUNATCH;
>>>>>>>> +       } else {
>>>>>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
>>>>>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
>>>>>>>> +
>>>>>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>> +                       /* Most chip versions fail with the genphy driver.
>>>>>>>> +                        * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>> +                        */
>>>>>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY version.\n", tp->phydev->phy_id);
>>>>>>>> +                       mdiobus_unregister(new_bus);
>>>>>>>> +                       return -EUNATCH;
>>>>>>>> +               }
>>>>>>>>           }
>>>>>>>>
>>>>>>>>           /* PHY will be woken up in rtl_open() */
>>>>>>>>
>>>>>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's updated.. and  now up'n'running. There is one (WARN_ONCE) stack trace coming from driver, i think i tinker with it next, with above patch the network devices shows up and they can be configured.
>>>>>>>
>>>>>> I tought to ask first, before going to make new probe_type for errorneus hw (propetype + retry counter) to do re-probe if requested, N times. Or should the r8169_main.c return deferred probe on error on some MAC enums ? Which approach is design-wise sound ?
>>>>>>
>>>>>> I just tought that the DEFERRED probe may just do the trick i'm looking ways to implement the re-probeing... hmm. I try the deferred thing and check would that help.
>>>>>>
>>>>> Playing with options to work around the issue is of course a great way to
>>>>> learn about the kernel. However it's questionable whether a workaround in
>>>>> the driver is acceptable for dealing with the broken BIOS of exactly one
>>>>>> 10 yrs old board (for which even a userspace workaround exists).
>>>> problem recognized: libphy-module get's unloaded for some reason before r8169 driver loads -> missing lowlevel functionality -> not working driver. This only occurs at 1st load of module.. seeking solution.
>>>>
>>>>
>>>> There is [last unloaded: libphy] entries in log BEFORE r8169 is probed first time.
>>>>
>>>>
>>>> Any clue what is responsible for unloading to occur ?
>>>>
>>>>
>>> Right now I'm debugging what is the reason, behind that the module starts to work properly only when
>>>
>>> unload & reload cycle is done.
>>>
>>>
>>> The libphy is listed as loaded, but the check for low level read/write function is not set -> r8169 modules rlt_open() fails.
>>>
>>> See here:
>>>
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: ------------[ cut here ]------------
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: read_page callback not available, PHY driver not loaded?
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: WARNING: CPU: 3 PID: 787 at drivers/net/phy/phy-core.c:750 phy_save_page+0xb1/0xe3 [libph
>>> y]
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Modules linked in: cmac algif_hash algif_skcipher af_alg bnep btusb btrtl btbcm btintel b
>>> luetooth ecdh_generic rfkill ecc uvcvideo videobuf2_vmalloc videobuf2_memops snd_usb_audio videobuf2_v4l2 videobuf2_common snd_usbmidi_
>>> lib videodev snd_rawmidi snd_seq_device mc input_leds joydev mousedev squashfs loop amdgpu snd_hda_codec_realtek gpu_sched i2c_algo_bit
>>>   ttm snd_hda_codec_generic drm_kms_helper edac_mce_amd ledtrig_audio cec rc_core kvm_amd drm ccp snd_hda_codec_hdmi agpgart rng_core kv
>>> m snd_hda_intel syscopyarea sysfillrect ppdev sysimgblt snd_intel_dspcfg fb_sys_fops snd_hda_codec snd_hda_core snd_hwdep irqbypass wmi
>>> _bmof snd_pcm snd_timer snd evdev pcspkr soundcore parport_pc parport sp5100_tco mac_hid i2c_piix4 k10temp acpi_cpufreq uinput crypto_u
>>> ser ip_tables x_tables hid_generic usbhid hid ohci_pci virtio_net net_failover failover firewire_ohci firewire_core crc_itu_t pata_atii
>>> xp ehci_pci ehci_hcd sr_mod cdrom ohci_hcd ata_generic pata_acpi pata_jmicron wmi floppy
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel:  r8169 realtek libphy
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: CPU: 3 PID: 787 Comm: NetworkManager Not tainted 5.7.0-1-raw #12
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Hardware name: Gigabyte Technology Co., Ltd. GA-MA790FXT-UD5P/GA-MA790FXT-UD5P, BIOS F8l 07/15/2010
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RIP: 0010:phy_save_page+0xb1/0xe3 [libphy]
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Code: c8 82 11 c0 e8 06 28 ff cc 85 db 74 47 48 8b 85 48 03 00 00 48 83 b8 68 01 00 00 00 75 10 48 c7 c7 e8 82 11 c0 e8 a9 dd f7 cc <0f> 0b eb 26 48 c7 c7 52 78 11 c0 e8 99 dd f7 cc 0f 0b 48 8b 85 48
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RSP: 0018:ffff962c408ef370 EFLAGS: 00010282
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RAX: 0000000000000000 RBX: 0000000000000001 RCX: 0000000000000000
>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RDX: 0000000000000001 RSI: 0000000000000092 RDI: 00000000ffffffff
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: RBP: ffff8b1af3eb8800 R08: 00000000000004b3 R09: 0000000000000004
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R10: 0000000000000000 R11: 0000000000000001 R12: 00000000ffffffa1
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R13: 0000000000000002 R14: 0000000000000002 R15: ffff8b1af3eb8800
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: FS: 00007f07be5d4d80(0000) GS:ffff8b1af7cc0000(0000) knlGS:0000000000000000
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CR2: 000055b83aecb008 CR3: 00000002246b0000 CR4: 00000000000006e0
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: Call Trace:
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: phy_select_page+0x53/0x7a [libphy]
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: phy_write_paged+0x5c/0xa0 [libphy]
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl8168d_1_hw_phy_config+0x9d/0x210 [r8169]
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl8169_init_phy+0x19/0x110 [r8169]
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl_open+0x354/0x4d0 [r8169]
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_open+0xe0/0x170
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_change_flags+0x188/0x1e0
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: dev_change_flags+0x21/0x60
>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: do_setlink+0x78a/0xfd0
>>>
>>>
>>> Something does not setup/register properly at first the way it should.
>>>
>>>
>> I think i solved it: realtek (the phy driver) was missing device entry for the PHY ID reported by the NIC to match -> read_page and write_page function pointers should now be set. The generic PHY does not fill
>>
>> the driver's functionality to read or write pages. It happens to be so that the drivers for |RTL8211B Gigabit Ethernet seems to work just fine for my NIC's.
> The analysis is wrong. The incorrect PHY ID is not root cause of the
> problem, it's caused by a BIOS bug. It's not a valid PHY ID. If you want
> to do something, then you could try to inject a PHY soft reset before
> the MII bus is registered. This should be board-specific, e.g. using
> dmi_check_system().

Ok, i'll try that approach, thanks for the tip. Hmm, I'll do similar 
function mapping what the HW init/config function does determine

function to use.


I'll get back to you when I got something new developed.

>> |
>>
>> |
>> |
>>
>> |I've documented current state to Manjaro's forum: https://forum.manjaro.org/t/your-post-in-r8168-kernel-5-6-3-driver-broken/139869/3 .. after i cleanup debugs away and do speedtests. Could the fix be something to considerate to upstream ?|
>>
>> |
>> |
>>
>> |The fix I'm proposing is not the reload hack, but just kernel modifications.
>> |
>>
>>
>>>>>>>>>>>>>>>>>>> The problem with old method seems to be, that device does not have had time to attach before the
>>>>>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>>>>>              if (!tp->phydev) {
>>>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>>>                      return -ENODEV;
>>>>>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>>>>>> +       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>>>>>                      /* Most chip versions fail with the genphy driver.
>>>>>>>>>>>>>>>>>>>                       * Therefore ensure that the dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>>>>>                       */
>>>>>>>>>>>>>>>>>>> - dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n");
>>>>>>>>>>>>>>>>>>> + dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>>>                      return -EUNATCH;
>>>>>>>>>>>>>>>>>>>              }
>>>>>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>>>>>         static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>>>>       {
>>>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>>>              if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>>>>>                      return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>>       @@ -712,12 +716,17 @@ static int __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>>>>         static int __phy_write_page(struct phy_device *phydev, int page)
>>>>>>>>>>>>>>>>>>>       {
>>>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>>>              if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>>>>>                      return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>>                return phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>>>>>>>>       }
>>>>>>>>>>>>>>>>>>>       +
>>>>>>>>>>>>>>>>>>>       /**
>>>>>>>>>>>>>>>>>>>        * phy_save_page() - take the bus lock and save the current page
>>>>>>>>>>>>>>>>>>>        * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>          On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>              Hi, There seems to he Something odd problem, maybe timing related. Stripped version not workingas expected. I get back to you, when i have it working.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>          There's no point in working on your patch. W/o proper justification it
>>>>>>>>>>>>>>>>>>>          isn't acceptable anyway. And so far we still don't know which problem
>>>>>>>>>>>>>>>>>>>          you actually have.
>>>>>>>>>>>>>>>>>>>          FIRST please provide the requested logs and explain the actual problem
>>>>>>>>>>>>>>>>>>>          (incl. the commit that caused the regression).
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>              13. huhtik. 2020, 14.46, Lauri Jakku <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> kirjoitti: Hi, Fair enough, i'll strip them. -lja On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>              On Mon, Apr 13, 2020 at 02:02:01PM +0300, Lauri Jakku wrote: Hi, Comments inline. On 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku wrote: From 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] NET: r8168/r8169 identifying fix The driver installation determination made properly by checking PHY vs DRIVER id's. --- drivers/net/ethernet/realtek/r8169_main.c | 70 ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c | 11 +++- 2 files changed, 72 insertions(+), 9 deletions(-) I would say that most of the code is debug prints. I tought that they are helpful to keep, they are using the debug calls, so they are not visible if user does not like those. You are missing the point of who are your users. Users want to have working device and the code. They don't need or like to debug their kernel. Thanks
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-05-03  0:11                                                   ` Lauri Jakku
@ 2020-05-03  1:34                                                     ` Lauri Jakku
  2020-05-03  2:28                                                       ` Lauri Jakku
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-05-03  1:34 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,


On 3.5.2020 3.11, Lauri Jakku wrote:
>
> On 3.5.2020 2.15, Heiner Kallweit wrote:
>> On 03.05.2020 00:42, Lauri Jakku wrote:
>>> Hi,
>>>
>>>
>>>
>>> On 2.5.2020 20.56, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>> On 1.5.2020 22.12, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>>
>>>>> On 19.4.2020 19.00, Heiner Kallweit wrote:
>>>>>> On 19.04.2020 18:49, Lauri Jakku wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>>>>>>
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>>>>>>> Hi
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without 
>>>>>>>>>>>>>>>>>>>> modifications --> network does not work
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified 
>>>>>>>>>>>>>>>>>>>> kernel (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched + 
>>>>>>>>>>>>>>>>>>>> r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're 
>>>>>>>>>>>>>>>>>>> presumable affected by a known BIOS bug.
>>>>>>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at 
>>>>>>>>>>>>>>>>>>> bugzilla.kernel.org.
>>>>>>>>>>>>>>>>>>> In the first referenced tickets it's about the same 
>>>>>>>>>>>>>>>>>>> mainboard (with earlier BIOS version).
>>>>>>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the 
>>>>>>>>>>>>>>>>>>> network chip / PHY correctly, it reports
>>>>>>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY 
>>>>>>>>>>>>>>>>>>> driver being found.
>>>>>>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your 
>>>>>>>>>>>>>>>>>>> problem should be gone.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now 
>>>>>>>>>>>>>>>>> testing some mutex guard for phy state and try to use 
>>>>>>>>>>>>>>>>> it as indicator
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> that attach has been done. One thing i've noticed is 
>>>>>>>>>>>>>>>>> that driver needs to be reloaded to allow traffic (ie. 
>>>>>>>>>>>>>>>>> ping works etc.)
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> All that shouldn't be needed. Just check with which PHY 
>>>>>>>>>>>>>>>> ID the PHY comes up.
>>>>>>>>>>>>>>>> And what do you mean with "it seems"? Is the option 
>>>>>>>>>>>>>>>> enabled or not?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> I do have ROM's enabled, and it does not help with my 
>>>>>>>>>>>>>>> issue.
>>>>>>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. 
>>>>>>>>>>>>> Then you have the same
>>>>>>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>>>>>>
>>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: MAC version: 23
>>>>>>>>>>>>
>>>>>>>>>>>> ....
>>>>>>>>>>>>
>>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>>
>>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>>>>>
>>>>>>>>>>>> .. after module unload & load cycle:
>>>>>>>>>>>>
>>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> it seem to be the case that the phy_id chances onetime, 
>>>>>>>>>>>> then stays the same. I'll do few shutdowns and see
>>>>>>>>>>>>
>>>>>>>>>>>> is there a pattern at all .. next i'm going to try how it 
>>>>>>>>>>>> behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> The BIOS downgrade: I'd like to solve this without 
>>>>>>>>>>>> downgrading BIOS. If I can't, then I'll do systemd-service 
>>>>>>>>>>>> that
>>>>>>>>>>>>
>>>>>>>>>>>> reloads r8169 driver at boot, cause then network is just fine.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>> What i've gathered samples now, there is three values for 
>>>>>>>>>>> PHY ID:
>>>>>>>>>>>
>>>>>>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl 
>>>>>>>>>>> |grep "PHY ver"
>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc2077002
>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>>
>>>>>>>>>>> I dont know are those hard coded or what, and are they 
>>>>>>>>>>> device specific how much.
>>>>>>>>>>>
>>>>>>>>>>> i haven't coldbooted things up, that may be that something 
>>>>>>>>>>> to check do they vary how per coldboot.
>>>>>>>>>>>
>>>>>>>>>>>>>> I check the ID, and revert all other changes, and check 
>>>>>>>>>>>>>> how it is working after adding the PHY id to list.
>>>>>>>>>>>>>>
>>>>>>>>>> What i've now learned: the patch + script + journald services 
>>>>>>>>>> -> Results working network, but it is still a workaround.
>>>>>>>>>>
>>>>>>>>> Following patch trusts the MAC version, another thing witch 
>>>>>>>>> could help is to derive method to do 2dn pass of the probeing:
>>>>>>>>>
>>>>>>>>> if specific MAC version is found.
>>>>>>>>>
>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>> index acd122a88d4a..62b37a1abc24 100644
>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
>>>>>>>>> rtl8169_private *tp)
>>>>>>>>>           if (!tp->phydev) {
>>>>>>>>>                   mdiobus_unregister(new_bus);
>>>>>>>>>                   return -ENODEV;
>>>>>>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>> -               /* Most chip versions fail with the genphy 
>>>>>>>>> driver.
>>>>>>>>> -                * Therefore ensure that the dedicated PHY 
>>>>>>>>> driver is loaded.
>>>>>>>>> -                */
>>>>>>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>> -               mdiobus_unregister(new_bus);
>>>>>>>>> -               return -EUNATCH;
>>>>>>>>> +       } else {
>>>>>>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
>>>>>>>>> tp->phydev->phy_id);
>>>>>>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", 
>>>>>>>>> tp->mac_version);
>>>>>>>>> +
>>>>>>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>> +                       /* Most chip versions fail with the 
>>>>>>>>> genphy driver.
>>>>>>>>> +                        * Therefore ensure that the dedicated 
>>>>>>>>> PHY driver is loaded.
>>>>>>>>> +                        */
>>>>>>>>> +                       dev_err(&pdev->dev, "Not known MAC/PHY 
>>>>>>>>> version.\n", tp->phydev->phy_id);
>>>>>>>>> + mdiobus_unregister(new_bus);
>>>>>>>>> +                       return -EUNATCH;
>>>>>>>>> +               }
>>>>>>>>>           }
>>>>>>>>>
>>>>>>>>>           /* PHY will be woken up in rtl_open() */
>>>>>>>>>
>>>>>>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's 
>>>>>>>> updated.. and  now up'n'running. There is one (WARN_ONCE) stack 
>>>>>>>> trace coming from driver, i think i tinker with it next, with 
>>>>>>>> above patch the network devices shows up and they can be 
>>>>>>>> configured.
>>>>>>>>
>>>>>>> I tought to ask first, before going to make new probe_type for 
>>>>>>> errorneus hw (propetype + retry counter) to do re-probe if 
>>>>>>> requested, N times. Or should the r8169_main.c return deferred 
>>>>>>> probe on error on some MAC enums ? Which approach is design-wise 
>>>>>>> sound ?
>>>>>>>
>>>>>>> I just tought that the DEFERRED probe may just do the trick i'm 
>>>>>>> looking ways to implement the re-probeing... hmm. I try the 
>>>>>>> deferred thing and check would that help.
>>>>>>>
>>>>>> Playing with options to work around the issue is of course a 
>>>>>> great way to
>>>>>> learn about the kernel. However it's questionable whether a 
>>>>>> workaround in
>>>>>> the driver is acceptable for dealing with the broken BIOS of 
>>>>>> exactly one
>>>>>>> 10 yrs old board (for which even a userspace workaround exists).
>>>>> problem recognized: libphy-module get's unloaded for some reason 
>>>>> before r8169 driver loads -> missing lowlevel functionality -> not 
>>>>> working driver. This only occurs at 1st load of module.. seeking 
>>>>> solution.
>>>>>
>>>>>
>>>>> There is [last unloaded: libphy] entries in log BEFORE r8169 is 
>>>>> probed first time.
>>>>>
>>>>>
>>>>> Any clue what is responsible for unloading to occur ?
>>>>>
>>>>>
>>>> Right now I'm debugging what is the reason, behind that the module 
>>>> starts to work properly only when
>>>>
>>>> unload & reload cycle is done.
>>>>
>>>>
>>>> The libphy is listed as loaded, but the check for low level 
>>>> read/write function is not set -> r8169 modules rlt_open() fails.
>>>>
>>>> See here:
>>>>
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: ------------[ cut 
>>>> here ]------------
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: read_page callback 
>>>> not available, PHY driver not loaded?
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: WARNING: CPU: 3 PID: 
>>>> 787 at drivers/net/phy/phy-core.c:750 phy_save_page+0xb1/0xe3 [libph
>>>> y]
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Modules linked in: 
>>>> cmac algif_hash algif_skcipher af_alg bnep btusb btrtl btbcm btintel b
>>>> luetooth ecdh_generic rfkill ecc uvcvideo videobuf2_vmalloc 
>>>> videobuf2_memops snd_usb_audio videobuf2_v4l2 videobuf2_common 
>>>> snd_usbmidi_
>>>> lib videodev snd_rawmidi snd_seq_device mc input_leds joydev 
>>>> mousedev squashfs loop amdgpu snd_hda_codec_realtek gpu_sched 
>>>> i2c_algo_bit
>>>>   ttm snd_hda_codec_generic drm_kms_helper edac_mce_amd 
>>>> ledtrig_audio cec rc_core kvm_amd drm ccp snd_hda_codec_hdmi 
>>>> agpgart rng_core kv
>>>> m snd_hda_intel syscopyarea sysfillrect ppdev sysimgblt 
>>>> snd_intel_dspcfg fb_sys_fops snd_hda_codec snd_hda_core snd_hwdep 
>>>> irqbypass wmi
>>>> _bmof snd_pcm snd_timer snd evdev pcspkr soundcore parport_pc 
>>>> parport sp5100_tco mac_hid i2c_piix4 k10temp acpi_cpufreq uinput 
>>>> crypto_u
>>>> ser ip_tables x_tables hid_generic usbhid hid ohci_pci virtio_net 
>>>> net_failover failover firewire_ohci firewire_core crc_itu_t pata_atii
>>>> xp ehci_pci ehci_hcd sr_mod cdrom ohci_hcd ata_generic pata_acpi 
>>>> pata_jmicron wmi floppy
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel:  r8169 realtek libphy
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: CPU: 3 PID: 787 Comm: 
>>>> NetworkManager Not tainted 5.7.0-1-raw #12
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Hardware name: 
>>>> Gigabyte Technology Co., Ltd. GA-MA790FXT-UD5P/GA-MA790FXT-UD5P, 
>>>> BIOS F8l 07/15/2010
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RIP: 
>>>> 0010:phy_save_page+0xb1/0xe3 [libphy]
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Code: c8 82 11 c0 e8 
>>>> 06 28 ff cc 85 db 74 47 48 8b 85 48 03 00 00 48 83 b8 68 01 00 00 
>>>> 00 75 10 48 c7 c7 e8 82 11 c0 e8 a9 dd f7 cc <0f> 0b eb 26 48 c7 c7 
>>>> 52 78 11 c0 e8 99 dd f7 cc 0f 0b 48 8b 85 48
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RSP: 
>>>> 0018:ffff962c408ef370 EFLAGS: 00010282
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RAX: 0000000000000000 
>>>> RBX: 0000000000000001 RCX: 0000000000000000
>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RDX: 0000000000000001 
>>>> RSI: 0000000000000092 RDI: 00000000ffffffff
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: RBP: ffff8b1af3eb8800 
>>>> R08: 00000000000004b3 R09: 0000000000000004
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R10: 0000000000000000 
>>>> R11: 0000000000000001 R12: 00000000ffffffa1
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R13: 0000000000000002 
>>>> R14: 0000000000000002 R15: ffff8b1af3eb8800
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: FS: 
>>>> 00007f07be5d4d80(0000) GS:ffff8b1af7cc0000(0000) 
>>>> knlGS:0000000000000000
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CS:  0010 DS: 0000 
>>>> ES: 0000 CR0: 0000000080050033
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CR2: 000055b83aecb008 
>>>> CR3: 00000002246b0000 CR4: 00000000000006e0
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: Call Trace:
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>> phy_select_page+0x53/0x7a [libphy]
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>> phy_write_paged+0x5c/0xa0 [libphy]
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>> rtl8168d_1_hw_phy_config+0x9d/0x210 [r8169]
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>> rtl8169_init_phy+0x19/0x110 [r8169]
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl_open+0x354/0x4d0 
>>>> [r8169]
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_open+0xe0/0x170
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>> __dev_change_flags+0x188/0x1e0
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>> dev_change_flags+0x21/0x60
>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: do_setlink+0x78a/0xfd0
>>>>
>>>>
>>>> Something does not setup/register properly at first the way it should.
>>>>
>>>>
>>> I think i solved it: realtek (the phy driver) was missing device 
>>> entry for the PHY ID reported by the NIC to match -> read_page and 
>>> write_page function pointers should now be set. The generic PHY does 
>>> not fill
>>>
>>> the driver's functionality to read or write pages. It happens to be 
>>> so that the drivers for |RTL8211B Gigabit Ethernet seems to work 
>>> just fine for my NIC's.
>> The analysis is wrong. The incorrect PHY ID is not root cause of the
>> problem, it's caused by a BIOS bug. It's not a valid PHY ID. If you want
>> to do something, then you could try to inject a PHY soft reset before
>> the MII bus is registered. This should be board-specific, e.g. using
>> dmi_check_system().
>
> Ok, i'll try that approach, thanks for the tip. Hmm, I'll do similar 
> function mapping what the HW init/config function does determine
>
> function to use.
>
>
> I'll get back to you when I got something new developed.
>
Oujeah, now it works OK.. the soft reset was all it needed.


Here's the patch:

 From 06c5deacf3ca9f9258431756a41ff0ba1792f1f7 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Thu, 16 Apr 2020 00:38:51 +0300
Subject: [PATCH] NET: r8169 driver identifying improvement.

Trust device MAC enum + r8169d NIC soft reset
before configuration.

This commit adds enumeration check and allows
driver to be slow to attach.

Signed-off-by: Lauri Jakku <lja@iki.fi>
---
  drivers/net/ethernet/realtek/r8169_main.c     | 24 +++--
  .../net/ethernet/realtek/r8169_phy_config.c   | 80 ++++++++++++++++
  drivers/net/phy/phy-core.c                    | 91 ++++++++++++++++++-
  3 files changed, 182 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..a85764f6e448 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -640,7 +640,6 @@ MODULE_AUTHOR("Realtek and the Linux r8169 crew 
<netdev@vger.kernel.org>");
  MODULE_DESCRIPTION("RealTek RTL-8169 Gigabit Ethernet driver");
  module_param_named(debug, debug.msg_enable, int, 0);
  MODULE_PARM_DESC(debug, "Debug verbosity level (0=none, ..., 16=all)");
-MODULE_SOFTDEP("pre: realtek");
  MODULE_LICENSE("GPL");
  MODULE_FIRMWARE(FIRMWARE_8168D_1);
  MODULE_FIRMWARE(FIRMWARE_8168D_2);
@@ -5172,13 +5171,18 @@ static int r8169_mdio_register(struct 
rtl8169_private *tp)
      if (!tp->phydev) {
          mdiobus_unregister(new_bus);
          return -ENODEV;
-    } else if (!tp->phydev->drv) {
-        /* Most chip versions fail with the genphy driver.
-         * Therefore ensure that the dedicated PHY driver is loaded.
-         */
-        dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to 
be added to initramfs?\n");
-        mdiobus_unregister(new_bus);
-        return -EUNATCH;
+    } else {
+        dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
+        dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
+
+        if (tp->mac_version == RTL_GIGA_MAC_NONE) {
+            /* Most chip versions fail with the genphy driver.
+             * Therefore ensure that the dedicated PHY driver is loaded.
+             */
+            dev_err(&pdev->dev, "Not known MAC/PHY version.\n");
+            mdiobus_unregister(new_bus);
+            return -EUNATCH;
+        }
      }

      /* PHY will be woken up in rtl_open() */
@@ -5513,6 +5517,9 @@ static int rtl_init_one(struct pci_dev *pdev, 
const struct pci_device_id *ent)
             rtl_chip_infos[chipset].name, dev->dev_addr, xid,
             pci_irq_vector(pdev, 0));

+    dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
+    dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
+
      if (jumbo_max)
          netif_info(tp, probe, dev,
                 "jumbo features [frames: %d bytes, tx checksumming: %s]\n",
@@ -5532,6 +5539,7 @@ static int rtl_init_one(struct pci_dev *pdev, 
const struct pci_device_id *ent)
      return rc;
  }

+
  static struct pci_driver rtl8169_pci_driver = {
      .name        = MODULENAME,
      .id_table    = rtl8169_pci_tbl,
diff --git a/drivers/net/ethernet/realtek/r8169_phy_config.c 
b/drivers/net/ethernet/realtek/r8169_phy_config.c
index b73f7d023e99..73d604c7a8a8 100644
--- a/drivers/net/ethernet/realtek/r8169_phy_config.c
+++ b/drivers/net/ethernet/realtek/r8169_phy_config.c
@@ -1243,6 +1243,84 @@ static void rtl8125_2_hw_phy_config(struct 
rtl8169_private *tp,
      rtl8125_config_eee_phy(phydev);
  }

+
+static void rtl8168d_hw_phy_pre_config_actions(struct rtl8169_private *tp,
+                            struct phy_device *phydev)
+{
+    /* Reset the PHY before configuration. There is BIOS bug that gives
+     * random PHY ID when device is not soft resetted first. --lja
+     */
+    genphy_soft_reset(phydev);
+}
+
+
+
+void r8169_hw_phy_pre_config_actions(struct rtl8169_private *tp, struct 
phy_device *phydev,
+                    enum mac_version ver)
+{
+    static const rtl_phy_cfg_fct phy_pre_config_actions[] = {
+        /* PCI devices. */
+        [RTL_GIGA_MAC_VER_02] = NULL,
+        [RTL_GIGA_MAC_VER_03] = NULL,
+        [RTL_GIGA_MAC_VER_04] = NULL,
+        [RTL_GIGA_MAC_VER_05] = NULL,
+        [RTL_GIGA_MAC_VER_06] = NULL,
+        /* PCI-E devices. */
+        [RTL_GIGA_MAC_VER_07] = NULL,
+        [RTL_GIGA_MAC_VER_08] = NULL,
+        [RTL_GIGA_MAC_VER_09] = NULL,
+        [RTL_GIGA_MAC_VER_10] = NULL,
+        [RTL_GIGA_MAC_VER_11] = NULL,
+        [RTL_GIGA_MAC_VER_12] = NULL,
+        [RTL_GIGA_MAC_VER_13] = NULL,
+        [RTL_GIGA_MAC_VER_14] = NULL,
+        [RTL_GIGA_MAC_VER_15] = NULL,
+        [RTL_GIGA_MAC_VER_16] = NULL,
+        [RTL_GIGA_MAC_VER_17] = NULL,
+        [RTL_GIGA_MAC_VER_18] = NULL,
+        [RTL_GIGA_MAC_VER_19] = NULL,
+        [RTL_GIGA_MAC_VER_20] = NULL,
+        [RTL_GIGA_MAC_VER_21] = NULL,
+        [RTL_GIGA_MAC_VER_22] = NULL,
+        [RTL_GIGA_MAC_VER_23] = NULL,
+        [RTL_GIGA_MAC_VER_24] = NULL,
+        [RTL_GIGA_MAC_VER_25] = rtl8168d_hw_phy_pre_config_actions,
+        [RTL_GIGA_MAC_VER_26] = rtl8168d_hw_phy_pre_config_actions,
+        [RTL_GIGA_MAC_VER_27] = rtl8168d_hw_phy_pre_config_actions,
+        [RTL_GIGA_MAC_VER_28] = rtl8168d_hw_phy_pre_config_actions,
+        [RTL_GIGA_MAC_VER_29] = NULL,
+        [RTL_GIGA_MAC_VER_30] = NULL,
+        [RTL_GIGA_MAC_VER_31] = NULL,
+        [RTL_GIGA_MAC_VER_32] = NULL,
+        [RTL_GIGA_MAC_VER_33] = NULL,
+        [RTL_GIGA_MAC_VER_34] = NULL,
+        [RTL_GIGA_MAC_VER_35] = NULL,
+        [RTL_GIGA_MAC_VER_36] = NULL,
+        [RTL_GIGA_MAC_VER_37] = NULL,
+        [RTL_GIGA_MAC_VER_38] = NULL,
+        [RTL_GIGA_MAC_VER_39] = NULL,
+        [RTL_GIGA_MAC_VER_40] = NULL,
+        [RTL_GIGA_MAC_VER_41] = NULL,
+        [RTL_GIGA_MAC_VER_42] = NULL,
+        [RTL_GIGA_MAC_VER_43] = NULL,
+        [RTL_GIGA_MAC_VER_44] = NULL,
+        [RTL_GIGA_MAC_VER_45] = NULL,
+        [RTL_GIGA_MAC_VER_46] = NULL,
+        [RTL_GIGA_MAC_VER_47] = NULL,
+        [RTL_GIGA_MAC_VER_48] = NULL,
+        [RTL_GIGA_MAC_VER_49] = NULL,
+        [RTL_GIGA_MAC_VER_50] = NULL,
+        [RTL_GIGA_MAC_VER_51] = NULL,
+        [RTL_GIGA_MAC_VER_52] = NULL,
+        [RTL_GIGA_MAC_VER_60] = NULL,
+        [RTL_GIGA_MAC_VER_61] = NULL,
+    };
+
+    if (phy_pre_config_actions[ver])
+        phy_pre_config_actions[ver](tp, phydev);
+}
+
+
  void r8169_hw_phy_config(struct rtl8169_private *tp, struct phy_device 
*phydev,
               enum mac_version ver)
  {
@@ -1303,6 +1381,8 @@ void r8169_hw_phy_config(struct rtl8169_private 
*tp, struct phy_device *phydev,
          [RTL_GIGA_MAC_VER_60] = rtl8125_1_hw_phy_config,
          [RTL_GIGA_MAC_VER_61] = rtl8125_2_hw_phy_config,
      };
+
+    r8169_hw_phy_pre_config_actions(tp, phydev);

      if (phy_configs[ver])
          phy_configs[ver](tp, phydev);
diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
index 66b8c61ca74c..b170185a1ed1 100644
--- a/drivers/net/phy/phy-core.c
+++ b/drivers/net/phy/phy-core.c
@@ -151,6 +151,25 @@ static const struct phy_setting settings[] = {
  };
  #undef PHY_SETTING

+#ifdef DEBUG
+#define R8169_ATTACHED_PRINTK \
+    printk("Phy is attached check %s @ %d: %d\n", \
+    __FUNC__, __LINE__, is_attached_check);
+
+#else
+#define R8169_ATTACHED_PRINTK
+#endif
+
+#define PHY_NOT_ATTACHED_CHECK \
+    { \
+        void *attached_dev_ptr = (phydev) ? phydev->attached_dev : NULL; \
+        int is_attached_check = (attached_dev_ptr != NULL) && \
+                                ((phydev) && (phydev->state >= 
PHY_READY)); \
+        R8169_ATTACHED_PRINTK \
+        if (! is_attached_check ) return -EOPNOTSUPP; \
+    };
+
+
  /**
   * phy_lookup_setting - lookup a PHY setting
   * @speed: speed to match
@@ -457,6 +476,9 @@ int phy_read_mmd(struct phy_device *phydev, int 
devad, u32 regnum)
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      phy_lock_mdio_bus(phydev);
      ret = __phy_read_mmd(phydev, devad, regnum);
      phy_unlock_mdio_bus(phydev);
@@ -479,6 +501,9 @@ int __phy_write_mmd(struct phy_device *phydev, int 
devad, u32 regnum, u16 val)
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      if (regnum > (u16)~0 || devad > 32)
          return -EINVAL;

@@ -518,6 +543,9 @@ int phy_write_mmd(struct phy_device *phydev, int 
devad, u32 regnum, u16 val)
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      phy_lock_mdio_bus(phydev);
      ret = __phy_write_mmd(phydev, devad, regnum, val);
      phy_unlock_mdio_bus(phydev);
@@ -543,6 +571,10 @@ int phy_modify_changed(struct phy_device *phydev, 
u32 regnum, u16 mask, u16 set)
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+
      phy_lock_mdio_bus(phydev);
      ret = __phy_modify_changed(phydev, regnum, mask, set);
      phy_unlock_mdio_bus(phydev);
@@ -587,6 +619,9 @@ int phy_modify(struct phy_device *phydev, u32 
regnum, u16 mask, u16 set)
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      phy_lock_mdio_bus(phydev);
      ret = __phy_modify(phydev, regnum, mask, set);
      phy_unlock_mdio_bus(phydev);
@@ -613,6 +648,9 @@ int __phy_modify_mmd_changed(struct phy_device 
*phydev, int devad, u32 regnum,
  {
      int new, ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      ret = __phy_read_mmd(phydev, devad, regnum);
      if (ret < 0)
          return ret;
@@ -646,6 +684,10 @@ int phy_modify_mmd_changed(struct phy_device 
*phydev, int devad, u32 regnum,
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+
      phy_lock_mdio_bus(phydev);
      ret = __phy_modify_mmd_changed(phydev, devad, regnum, mask, set);
      phy_unlock_mdio_bus(phydev);
@@ -671,6 +713,9 @@ int __phy_modify_mmd(struct phy_device *phydev, int 
devad, u32 regnum,
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      ret = __phy_modify_mmd_changed(phydev, devad, regnum, mask, set);

      return ret < 0 ? ret : 0;
@@ -694,6 +739,9 @@ int phy_modify_mmd(struct phy_device *phydev, int 
devad, u32 regnum,
  {
      int ret;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      phy_lock_mdio_bus(phydev);
      ret = __phy_modify_mmd(phydev, devad, regnum, mask, set);
      phy_unlock_mdio_bus(phydev);
@@ -704,7 +752,11 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);

  static int __phy_read_page(struct phy_device *phydev)
  {
-    if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not 
available, PHY driver not loaded?\n"))
+
+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+    if (WARN(!phydev->drv->read_page, "read_page callback not 
available, PHY driver not loaded?\n"))
          return -EOPNOTSUPP;

      return phydev->drv->read_page(phydev);
@@ -712,12 +764,16 @@ static int __phy_read_page(struct phy_device *phydev)

  static int __phy_write_page(struct phy_device *phydev, int page)
  {
-    if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not 
available, PHY driver not loaded?\n"))
+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+    if (WARN(!phydev->drv->write_page, "write_page callback not 
available, PHY driver not loaded?\n"))
          return -EOPNOTSUPP;

      return phydev->drv->write_page(phydev, page);
  }

+
  /**
   * phy_save_page() - take the bus lock and save the current page
   * @phydev: a pointer to a &struct phy_device
@@ -728,7 +784,11 @@ static int __phy_write_page(struct phy_device 
*phydev, int page)
   */
  int phy_save_page(struct phy_device *phydev)
  {
-    phy_lock_mdio_bus(phydev);
+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+
+    phy_lock_mdio_bus(phydev);
      return __phy_read_page(phydev);
  }
  EXPORT_SYMBOL_GPL(phy_save_page);
@@ -748,7 +808,10 @@ int phy_select_page(struct phy_device *phydev, int 
page)
  {
      int ret, oldpage;

-    oldpage = ret = phy_save_page(phydev);
+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+    oldpage = ret = phy_save_page(phydev);
      if (ret < 0)
          return ret;

@@ -782,6 +845,9 @@ int phy_restore_page(struct phy_device *phydev, int 
oldpage, int ret)
  {
      int r;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
      if (oldpage >= 0) {
          r = __phy_write_page(phydev, oldpage);

@@ -812,6 +878,9 @@ EXPORT_SYMBOL_GPL(phy_restore_page);
  int phy_read_paged(struct phy_device *phydev, int page, u32 regnum)
  {
      int ret = 0, oldpage;
+
+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;

      oldpage = phy_select_page(phydev, page);
      if (oldpage >= 0)
@@ -834,6 +903,10 @@ int phy_write_paged(struct phy_device *phydev, int 
page, u32 regnum, u16 val)
  {
      int ret = 0, oldpage;

+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+
      oldpage = phy_select_page(phydev, page);
      if (oldpage >= 0)
          ret = __phy_write(phydev, regnum, val);
@@ -856,6 +929,9 @@ int phy_modify_paged_changed(struct phy_device 
*phydev, int page, u32 regnum,
                   u16 mask, u16 set)
  {
      int ret = 0, oldpage;
+
+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;

      oldpage = phy_select_page(phydev, page);
      if (oldpage >= 0)
@@ -878,7 +954,12 @@ EXPORT_SYMBOL(phy_modify_paged_changed);
  int phy_modify_paged(struct phy_device *phydev, int page, u32 regnum,
               u16 mask, u16 set)
  {
-    int ret = phy_modify_paged_changed(phydev, page, regnum, mask, set);
+    int ret = -1;
+
+    /* If not attached, do nothing. */
+    PHY_NOT_ATTACHED_CHECK;
+
+    ret = phy_modify_paged_changed(phydev, page, regnum, mask, set);

      return ret < 0 ? ret : 0;
  }
-- 
2.26.2




>>> |
>>>
>>> |
>>> |
>>>
>>> |I've documented current state to Manjaro's forum: 
>>> https://forum.manjaro.org/t/your-post-in-r8168-kernel-5-6-3-driver-broken/139869/3 
>>> .. after i cleanup debugs away and do speedtests. Could the fix be 
>>> something to considerate to upstream ?|
>>>
>>> |
>>> |
>>>
>>> |The fix I'm proposing is not the reload hack, but just kernel 
>>> modifications.
>>> |
>>>
>>>
>>>>>>>>>>>>>>>>>>>> The problem with old method seems to be, that 
>>>>>>>>>>>>>>>>>>>> device does not have had time to attach before the
>>>>>>>>>>>>>>>>>>>> PHY driver check.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> The patch:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> diff --git 
>>>>>>>>>>>>>>>>>>>> a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>>>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>>>> index bf5bf05970a2..acd122a88d4a 100644
>>>>>>>>>>>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>>>>>>>>>>>> @@ -5172,11 +5172,11 @@ static int 
>>>>>>>>>>>>>>>>>>>> r8169_mdio_register(struct rtl8169_private *tp)
>>>>>>>>>>>>>>>>>>>>              if (!tp->phydev) {
>>>>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>>>>                      return -ENODEV;
>>>>>>>>>>>>>>>>>>>> -       } else if (!tp->phydev->drv) {
>>>>>>>>>>>>>>>>>>>> +       } else if (tp->mac_version == 
>>>>>>>>>>>>>>>>>>>> RTL_GIGA_MAC_NONE) {
>>>>>>>>>>>>>>>>>>>>                      /* Most chip versions fail 
>>>>>>>>>>>>>>>>>>>> with the genphy driver.
>>>>>>>>>>>>>>>>>>>>                       * Therefore ensure that the 
>>>>>>>>>>>>>>>>>>>> dedicated PHY driver is loaded.
>>>>>>>>>>>>>>>>>>>>                       */
>>>>>>>>>>>>>>>>>>>> - dev_err(&pdev->dev, "realtek.ko not loaded, maybe 
>>>>>>>>>>>>>>>>>>>> it needs to be added to initramfs?\n");
>>>>>>>>>>>>>>>>>>>> + dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>>>>>>>>>>>> mdiobus_unregister(new_bus);
>>>>>>>>>>>>>>>>>>>>                      return -EUNATCH;
>>>>>>>>>>>>>>>>>>>>              }
>>>>>>>>>>>>>>>>>>>> diff --git a/drivers/net/phy/phy-core.c 
>>>>>>>>>>>>>>>>>>>> b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>>>> index 66b8c61ca74c..aba2b304b821 100644
>>>>>>>>>>>>>>>>>>>> --- a/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>>>> +++ b/drivers/net/phy/phy-core.c
>>>>>>>>>>>>>>>>>>>> @@ -704,6 +704,10 @@ 
>>>>>>>>>>>>>>>>>>>> EXPORT_SYMBOL_GPL(phy_modify_mmd);
>>>>>>>>>>>>>>>>>>>>         static int __phy_read_page(struct 
>>>>>>>>>>>>>>>>>>>> phy_device *phydev)
>>>>>>>>>>>>>>>>>>>>       {
>>>>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>>>>              if (WARN_ONCE(!phydev->drv->read_page, 
>>>>>>>>>>>>>>>>>>>> "read_page callback not available, PHY driver not 
>>>>>>>>>>>>>>>>>>>> loaded?\n"))
>>>>>>>>>>>>>>>>>>>>                      return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>>>       @@ -712,12 +716,17 @@ static int 
>>>>>>>>>>>>>>>>>>>> __phy_read_page(struct phy_device *phydev)
>>>>>>>>>>>>>>>>>>>>         static int __phy_write_page(struct 
>>>>>>>>>>>>>>>>>>>> phy_device *phydev, int page)
>>>>>>>>>>>>>>>>>>>>       {
>>>>>>>>>>>>>>>>>>>> +       /* If not attached, do nothing (no warning) */
>>>>>>>>>>>>>>>>>>>> +       if (!phydev->attached_dev)
>>>>>>>>>>>>>>>>>>>> +               return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>>> +
>>>>>>>>>>>>>>>>>>>>              if 
>>>>>>>>>>>>>>>>>>>> (WARN_ONCE(!phydev->drv->write_page, "write_page 
>>>>>>>>>>>>>>>>>>>> callback not available, PHY driver not loaded?\n"))
>>>>>>>>>>>>>>>>>>>>                      return -EOPNOTSUPP;
>>>>>>>>>>>>>>>>>>>>                return 
>>>>>>>>>>>>>>>>>>>> phydev->drv->write_page(phydev, page);
>>>>>>>>>>>>>>>>>>>>       }
>>>>>>>>>>>>>>>>>>>>       +
>>>>>>>>>>>>>>>>>>>>       /**
>>>>>>>>>>>>>>>>>>>>        * phy_save_page() - take the bus lock and 
>>>>>>>>>>>>>>>>>>>> save the current page
>>>>>>>>>>>>>>>>>>>>        * @phydev: a pointer to a &struct phy_device
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> 15. huhtik. 2020, 19.18, Heiner Kallweit 
>>>>>>>>>>>>>>>>>>>> <hkallweit1@gmail.com 
>>>>>>>>>>>>>>>>>>>> <mailto:hkallweit1@gmail.com>> kirjoitti:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>          On 15.04.2020 16:39, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>              Hi, There seems to he Something odd 
>>>>>>>>>>>>>>>>>>>> problem, maybe timing related. Stripped version not 
>>>>>>>>>>>>>>>>>>>> workingas expected. I get back to you, when i have 
>>>>>>>>>>>>>>>>>>>> it working.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>          There's no point in working on your patch. 
>>>>>>>>>>>>>>>>>>>> W/o proper justification it
>>>>>>>>>>>>>>>>>>>>          isn't acceptable anyway. And so far we 
>>>>>>>>>>>>>>>>>>>> still don't know which problem
>>>>>>>>>>>>>>>>>>>>          you actually have.
>>>>>>>>>>>>>>>>>>>>          FIRST please provide the requested logs 
>>>>>>>>>>>>>>>>>>>> and explain the actual problem
>>>>>>>>>>>>>>>>>>>>          (incl. the commit that caused the 
>>>>>>>>>>>>>>>>>>>> regression).
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>              13. huhtik. 2020, 14.46, Lauri Jakku 
>>>>>>>>>>>>>>>>>>>> <ljakku77@gmail.com <mailto:ljakku77@gmail.com>> 
>>>>>>>>>>>>>>>>>>>> kirjoitti: Hi, Fair enough, i'll strip them. -lja 
>>>>>>>>>>>>>>>>>>>> On 2020-04-13 14:34, Leon Romanovsky wrote:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>              On Mon, Apr 13, 2020 at 02:02:01PM 
>>>>>>>>>>>>>>>>>>>> +0300, Lauri Jakku wrote: Hi, Comments inline. On 
>>>>>>>>>>>>>>>>>>>> 2020-04-13 13:58, Leon Romanovsky wrote: On Mon, 
>>>>>>>>>>>>>>>>>>>> Apr 13, 2020 at 01:30:13PM +0300, Lauri Jakku 
>>>>>>>>>>>>>>>>>>>> wrote: From 
>>>>>>>>>>>>>>>>>>>> 2d41edd4e6455187094f3a13d58c46eeee35aa31 Mon Sep 17 
>>>>>>>>>>>>>>>>>>>> 00:00:00 2001 From: Lauri Jakku <lja@iki.fi> Date: 
>>>>>>>>>>>>>>>>>>>> Mon, 13 Apr 2020 13:18:35 +0300 Subject: [PATCH] 
>>>>>>>>>>>>>>>>>>>> NET: r8168/r8169 identifying fix The driver 
>>>>>>>>>>>>>>>>>>>> installation determination made properly by 
>>>>>>>>>>>>>>>>>>>> checking PHY vs DRIVER id's. --- 
>>>>>>>>>>>>>>>>>>>> drivers/net/ethernet/realtek/r8169_main.c | 70 
>>>>>>>>>>>>>>>>>>>> ++++++++++++++++++++--- drivers/net/phy/mdio_bus.c 
>>>>>>>>>>>>>>>>>>>> | 11 +++- 2 files changed, 72 insertions(+), 9 
>>>>>>>>>>>>>>>>>>>> deletions(-) I would say that most of the code is 
>>>>>>>>>>>>>>>>>>>> debug prints. I tought that they are helpful to 
>>>>>>>>>>>>>>>>>>>> keep, they are using the debug calls, so they are 
>>>>>>>>>>>>>>>>>>>> not visible if user does not like those. You are 
>>>>>>>>>>>>>>>>>>>> missing the point of who are your users. Users want 
>>>>>>>>>>>>>>>>>>>> to have working device and the code. They don't 
>>>>>>>>>>>>>>>>>>>> need or like to debug their kernel. Thanks
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>

^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: NET: r8168/r8169 identifying fix
  2020-05-03  1:34                                                     ` Lauri Jakku
@ 2020-05-03  2:28                                                       ` Lauri Jakku
  2020-05-03  8:33                                                         ` Heiner Kallweit
  0 siblings, 1 reply; 57+ messages in thread
From: Lauri Jakku @ 2020-05-03  2:28 UTC (permalink / raw)
  To: Heiner Kallweit; +Cc: Leon Romanovsky, netdev, nic_swsd

Hi,

On 3.5.2020 4.34, Lauri Jakku wrote:
> Hi,
>
>
> On 3.5.2020 3.11, Lauri Jakku wrote:
>>
>> On 3.5.2020 2.15, Heiner Kallweit wrote:
>>> On 03.05.2020 00:42, Lauri Jakku wrote:
>>>> Hi,
>>>>
>>>>
>>>>
>>>> On 2.5.2020 20.56, Lauri Jakku wrote:
>>>>> Hi,
>>>>>
>>>>> On 1.5.2020 22.12, Lauri Jakku wrote:
>>>>>> Hi,
>>>>>>
>>>>>>
>>>>>> On 19.4.2020 19.00, Heiner Kallweit wrote:
>>>>>>> On 19.04.2020 18:49, Lauri Jakku wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On 19.4.2020 18.09, Lauri Jakku wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> On 18.4.2020 21.46, Lauri Jakku wrote:
>>>>>>>>>
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> On 18.4.2020 14.06, Lauri Jakku wrote:
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>> On 17.4.2020 10.30, Lauri Jakku wrote:
>>>>>>>>>>>> Hi,
>>>>>>>>>>>>
>>>>>>>>>>>> On 17.4.2020 9.23, Lauri Jakku wrote:
>>>>>>>>>>>>> On 16.4.2020 23.50, Heiner Kallweit wrote:
>>>>>>>>>>>>>> On 16.04.2020 22:38, Lauri Jakku wrote:
>>>>>>>>>>>>>>> Hi
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 16.4.2020 23.10, Lauri Jakku wrote:
>>>>>>>>>>>>>>>> On 16.4.2020 23.02, Heiner Kallweit wrote:
>>>>>>>>>>>>>>>>> On 16.04.2020 21:58, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> On 16.4.2020 21.37, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> On 16.4.2020 21.26, Heiner Kallweit wrote:
>>>>>>>>>>>>>>>>>>>> On 16.04.2020 13:30, Lauri Jakku wrote:
>>>>>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO: stock manjaro kernel, without 
>>>>>>>>>>>>>>>>>>>>> modifications --> network does not work
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-lja: No attach check, modified 
>>>>>>>>>>>>>>>>>>>>> kernel (r8169 mods only) --> network does not work
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> 5.6.3-2-MANJARO-with-the-r8169-patch: phy patched 
>>>>>>>>>>>>>>>>>>>>> + r8169 mods -> devices show up ok, network works
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> All different initcpio's have realtek.ko in them.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> Thanks for the logs. Based on the logs you're 
>>>>>>>>>>>>>>>>>>>> presumable affected by a known BIOS bug.
>>>>>>>>>>>>>>>>>>>> Check bug tickets 202275 and 207203 at 
>>>>>>>>>>>>>>>>>>>> bugzilla.kernel.org.
>>>>>>>>>>>>>>>>>>>> In the first referenced tickets it's about the same 
>>>>>>>>>>>>>>>>>>>> mainboard (with earlier BIOS version).
>>>>>>>>>>>>>>>>>>>> BIOS on this mainboard seems to not initialize the 
>>>>>>>>>>>>>>>>>>>> network chip / PHY correctly, it reports
>>>>>>>>>>>>>>>>>>>> a random number as PHY ID, resulting in no PHY 
>>>>>>>>>>>>>>>>>>>> driver being found.
>>>>>>>>>>>>>>>>>>>> Enable "Onboard LAN Boot ROM" in the BIOS, and your 
>>>>>>>>>>>>>>>>>>>> problem should be gone.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> OK, I try that, thank you :)
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> It seems that i DO have the ROM's enabled, i'm now 
>>>>>>>>>>>>>>>>>> testing some mutex guard for phy state and try to use 
>>>>>>>>>>>>>>>>>> it as indicator
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> that attach has been done. One thing i've noticed is 
>>>>>>>>>>>>>>>>>> that driver needs to be reloaded to allow traffic 
>>>>>>>>>>>>>>>>>> (ie. ping works etc.)
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> All that shouldn't be needed. Just check with which 
>>>>>>>>>>>>>>>>> PHY ID the PHY comes up.
>>>>>>>>>>>>>>>>> And what do you mean with "it seems"? Is the option 
>>>>>>>>>>>>>>>>> enabled or not?
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> I do have ROM's enabled, and it does not help with my 
>>>>>>>>>>>>>>>> issue.
>>>>>>>>>>>>>> Your BIOS is a beta version, downgrading to F7 may help. 
>>>>>>>>>>>>>> Then you have the same
>>>>>>>>>>>>>> mainboard + BIOS as the user who opened bug ticket 202275.
>>>>>>>>>>>>>>
>>>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>>> 0000:03:00.0: MAC version: 23
>>>>>>>>>>>>>
>>>>>>>>>>>>> ....
>>>>>>>>>>>>>
>>>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>>>
>>>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>>>>>>
>>>>>>>>>>>>> .. after module unload & load cycle:
>>>>>>>>>>>>>
>>>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>>> 0000:02:00.0: MAC version: 23
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> it seem to be the case that the phy_id chances onetime, 
>>>>>>>>>>>>> then stays the same. I'll do few shutdowns and see
>>>>>>>>>>>>>
>>>>>>>>>>>>> is there a pattern at all .. next i'm going to try how it 
>>>>>>>>>>>>> behaves, if i read mac/phy versions twice on MAC version 23.
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> The BIOS downgrade: I'd like to solve this without 
>>>>>>>>>>>>> downgrading BIOS. If I can't, then I'll do systemd-service 
>>>>>>>>>>>>> that
>>>>>>>>>>>>>
>>>>>>>>>>>>> reloads r8169 driver at boot, cause then network is just 
>>>>>>>>>>>>> fine.
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>> What i've gathered samples now, there is three values for 
>>>>>>>>>>>> PHY ID:
>>>>>>>>>>>>
>>>>>>>>>>>> [sillyme@MinistryOfSillyWalk KernelStuff]$ sudo journalctl 
>>>>>>>>>>>> |grep "PHY ver"
>>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc2077002
>>>>>>>>>>>> huhti 17 09:01:49 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc2077002
>>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 09:03:29 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 09:17:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>>>> huhti 17 09:24:53 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 09:27:59 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>>>> huhti 17 10:08:42 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 10:12:07 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0xc1071002
>>>>>>>>>>>> huhti 17 10:20:35 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0xc1071002
>>>>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:02:00.0: PHY version: 0x1cc912
>>>>>>>>>>>> huhti 17 10:23:46 MinistryOfSillyWalk kernel: r8169 
>>>>>>>>>>>> 0000:03:00.0: PHY version: 0x1cc912
>>>>>>>>>>>>
>>>>>>>>>>>> I dont know are those hard coded or what, and are they 
>>>>>>>>>>>> device specific how much.
>>>>>>>>>>>>
>>>>>>>>>>>> i haven't coldbooted things up, that may be that something 
>>>>>>>>>>>> to check do they vary how per coldboot.
>>>>>>>>>>>>
>>>>>>>>>>>>>>> I check the ID, and revert all other changes, and check 
>>>>>>>>>>>>>>> how it is working after adding the PHY id to list.
>>>>>>>>>>>>>>>
>>>>>>>>>>> What i've now learned: the patch + script + journald 
>>>>>>>>>>> services -> Results working network, but it is still a 
>>>>>>>>>>> workaround.
>>>>>>>>>>>
>>>>>>>>>> Following patch trusts the MAC version, another thing witch 
>>>>>>>>>> could help is to derive method to do 2dn pass of the probeing:
>>>>>>>>>>
>>>>>>>>>> if specific MAC version is found.
>>>>>>>>>>
>>>>>>>>>> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
>>>>>>>>>> b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> index acd122a88d4a..62b37a1abc24 100644
>>>>>>>>>> --- a/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> +++ b/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>>>>> @@ -5172,13 +5172,18 @@ static int r8169_mdio_register(struct 
>>>>>>>>>> rtl8169_private *tp)
>>>>>>>>>>           if (!tp->phydev) {
>>>>>>>>>>                   mdiobus_unregister(new_bus);
>>>>>>>>>>                   return -ENODEV;
>>>>>>>>>> -       } else if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>> -               /* Most chip versions fail with the genphy 
>>>>>>>>>> driver.
>>>>>>>>>> -                * Therefore ensure that the dedicated PHY 
>>>>>>>>>> driver is loaded.
>>>>>>>>>> -                */
>>>>>>>>>> -               dev_err(&pdev->dev, "Not known MAC version.\n");
>>>>>>>>>> -               mdiobus_unregister(new_bus);
>>>>>>>>>> -               return -EUNATCH;
>>>>>>>>>> +       } else {
>>>>>>>>>> +               dev_info(&pdev->dev, "PHY version: 0x%x\n", 
>>>>>>>>>> tp->phydev->phy_id);
>>>>>>>>>> +               dev_info(&pdev->dev, "MAC version: %d\n", 
>>>>>>>>>> tp->mac_version);
>>>>>>>>>> +
>>>>>>>>>> +               if (tp->mac_version == RTL_GIGA_MAC_NONE) {
>>>>>>>>>> +                       /* Most chip versions fail with the 
>>>>>>>>>> genphy driver.
>>>>>>>>>> +                        * Therefore ensure that the 
>>>>>>>>>> dedicated PHY driver is loaded.
>>>>>>>>>> +                        */
>>>>>>>>>> + dev_err(&pdev->dev, "Not known MAC/PHY version.\n", 
>>>>>>>>>> tp->phydev->phy_id);
>>>>>>>>>> + mdiobus_unregister(new_bus);
>>>>>>>>>> +                       return -EUNATCH;
>>>>>>>>>> +               }
>>>>>>>>>>           }
>>>>>>>>>>
>>>>>>>>>>           /* PHY will be woken up in rtl_open() */
>>>>>>>>>>
>>>>>>>>> I just got bleeding edge 5.7.0-1 kernel compiled + firmware's 
>>>>>>>>> updated.. and  now up'n'running. There is one (WARN_ONCE) 
>>>>>>>>> stack trace coming from driver, i think i tinker with it next, 
>>>>>>>>> with above patch the network devices shows up and they can be 
>>>>>>>>> configured.
>>>>>>>>>
>>>>>>>> I tought to ask first, before going to make new probe_type for 
>>>>>>>> errorneus hw (propetype + retry counter) to do re-probe if 
>>>>>>>> requested, N times. Or should the r8169_main.c return deferred 
>>>>>>>> probe on error on some MAC enums ? Which approach is 
>>>>>>>> design-wise sound ?
>>>>>>>>
>>>>>>>> I just tought that the DEFERRED probe may just do the trick i'm 
>>>>>>>> looking ways to implement the re-probeing... hmm. I try the 
>>>>>>>> deferred thing and check would that help.
>>>>>>>>
>>>>>>> Playing with options to work around the issue is of course a 
>>>>>>> great way to
>>>>>>> learn about the kernel. However it's questionable whether a 
>>>>>>> workaround in
>>>>>>> the driver is acceptable for dealing with the broken BIOS of 
>>>>>>> exactly one
>>>>>>>> 10 yrs old board (for which even a userspace workaround exists).
>>>>>> problem recognized: libphy-module get's unloaded for some reason 
>>>>>> before r8169 driver loads -> missing lowlevel functionality -> 
>>>>>> not working driver. This only occurs at 1st load of module.. 
>>>>>> seeking solution.
>>>>>>
>>>>>>
>>>>>> There is [last unloaded: libphy] entries in log BEFORE r8169 is 
>>>>>> probed first time.
>>>>>>
>>>>>>
>>>>>> Any clue what is responsible for unloading to occur ?
>>>>>>
>>>>>>
>>>>> Right now I'm debugging what is the reason, behind that the module 
>>>>> starts to work properly only when
>>>>>
>>>>> unload & reload cycle is done.
>>>>>
>>>>>
>>>>> The libphy is listed as loaded, but the check for low level 
>>>>> read/write function is not set -> r8169 modules rlt_open() fails.
>>>>>
>>>>> See here:
>>>>>
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: ------------[ cut 
>>>>> here ]------------
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: read_page callback 
>>>>> not available, PHY driver not loaded?
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: WARNING: CPU: 3 PID: 
>>>>> 787 at drivers/net/phy/phy-core.c:750 phy_save_page+0xb1/0xe3 [libph
>>>>> y]
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Modules linked in: 
>>>>> cmac algif_hash algif_skcipher af_alg bnep btusb btrtl btbcm 
>>>>> btintel b
>>>>> luetooth ecdh_generic rfkill ecc uvcvideo videobuf2_vmalloc 
>>>>> videobuf2_memops snd_usb_audio videobuf2_v4l2 videobuf2_common 
>>>>> snd_usbmidi_
>>>>> lib videodev snd_rawmidi snd_seq_device mc input_leds joydev 
>>>>> mousedev squashfs loop amdgpu snd_hda_codec_realtek gpu_sched 
>>>>> i2c_algo_bit
>>>>>   ttm snd_hda_codec_generic drm_kms_helper edac_mce_amd 
>>>>> ledtrig_audio cec rc_core kvm_amd drm ccp snd_hda_codec_hdmi 
>>>>> agpgart rng_core kv
>>>>> m snd_hda_intel syscopyarea sysfillrect ppdev sysimgblt 
>>>>> snd_intel_dspcfg fb_sys_fops snd_hda_codec snd_hda_core snd_hwdep 
>>>>> irqbypass wmi
>>>>> _bmof snd_pcm snd_timer snd evdev pcspkr soundcore parport_pc 
>>>>> parport sp5100_tco mac_hid i2c_piix4 k10temp acpi_cpufreq uinput 
>>>>> crypto_u
>>>>> ser ip_tables x_tables hid_generic usbhid hid ohci_pci virtio_net 
>>>>> net_failover failover firewire_ohci firewire_core crc_itu_t pata_atii
>>>>> xp ehci_pci ehci_hcd sr_mod cdrom ohci_hcd ata_generic pata_acpi 
>>>>> pata_jmicron wmi floppy
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel:  r8169 realtek libphy
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: CPU: 3 PID: 787 
>>>>> Comm: NetworkManager Not tainted 5.7.0-1-raw #12
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Hardware name: 
>>>>> Gigabyte Technology Co., Ltd. GA-MA790FXT-UD5P/GA-MA790FXT-UD5P, 
>>>>> BIOS F8l 07/15/2010
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RIP: 
>>>>> 0010:phy_save_page+0xb1/0xe3 [libphy]
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: Code: c8 82 11 c0 e8 
>>>>> 06 28 ff cc 85 db 74 47 48 8b 85 48 03 00 00 48 83 b8 68 01 00 00 
>>>>> 00 75 10 48 c7 c7 e8 82 11 c0 e8 a9 dd f7 cc <0f> 0b eb 26 48 c7 
>>>>> c7 52 78 11 c0 e8 99 dd f7 cc 0f 0b 48 8b 85 48
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RSP: 
>>>>> 0018:ffff962c408ef370 EFLAGS: 00010282
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RAX: 
>>>>> 0000000000000000 RBX: 0000000000000001 RCX: 0000000000000000
>>>>> touko 02 20:40:04 MinistryOfSillyWalk kernel: RDX: 
>>>>> 0000000000000001 RSI: 0000000000000092 RDI: 00000000ffffffff
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: RBP: 
>>>>> ffff8b1af3eb8800 R08: 00000000000004b3 R09: 0000000000000004
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R10: 
>>>>> 0000000000000000 R11: 0000000000000001 R12: 00000000ffffffa1
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: R13: 
>>>>> 0000000000000002 R14: 0000000000000002 R15: ffff8b1af3eb8800
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: FS: 
>>>>> 00007f07be5d4d80(0000) GS:ffff8b1af7cc0000(0000) 
>>>>> knlGS:0000000000000000
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CS:  0010 DS: 0000 
>>>>> ES: 0000 CR0: 0000000080050033
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: CR2: 
>>>>> 000055b83aecb008 CR3: 00000002246b0000 CR4: 00000000000006e0
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: Call Trace:
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>>> phy_select_page+0x53/0x7a [libphy]
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>>> phy_write_paged+0x5c/0xa0 [libphy]
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>>> rtl8168d_1_hw_phy_config+0x9d/0x210 [r8169]
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>>> rtl8169_init_phy+0x19/0x110 [r8169]
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: rtl_open+0x354/0x4d0 
>>>>> [r8169]
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: __dev_open+0xe0/0x170
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>>> __dev_change_flags+0x188/0x1e0
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: 
>>>>> dev_change_flags+0x21/0x60
>>>>> touko 02 20:40:05 MinistryOfSillyWalk kernel: do_setlink+0x78a/0xfd0
>>>>>
>>>>>
>>>>> Something does not setup/register properly at first the way it 
>>>>> should.
>>>>>
>>>>>
>>>> I think i solved it: realtek (the phy driver) was missing device 
>>>> entry for the PHY ID reported by the NIC to match -> read_page and 
>>>> write_page function pointers should now be set. The generic PHY 
>>>> does not fill
>>>>
>>>> the driver's functionality to read or write pages. It happens to be 
>>>> so that the drivers for |RTL8211B Gigabit Ethernet seems to work 
>>>> just fine for my NIC's.
>>> The analysis is wrong. The incorrect PHY ID is not root cause of the
>>> problem, it's caused by a BIOS bug. It's not a valid PHY ID. If you 
>>> want
>>> to do something, then you could try to inject a PHY soft reset before
>>> the MII bus is registered. This should be board-specific, e.g. using
>>> dmi_check_system().
>>
>> Ok, i'll try that approach, thanks for the tip. Hmm, I'll do similar 
>> function mapping what the HW init/config function does determine
>>
>> function to use.
>>
>>
>> I'll get back to you when I got something new developed.
>>
> Oujeah, now it works OK.. the soft reset was all it needed.
>
>
> Here's the patch:
>
> From 06c5deacf3ca9f9258431756a41ff0ba1792f1f7 Mon Sep 17 00:00:00 2001
> From: Lauri Jakku <lja@iki.fi>
> Date: Thu, 16 Apr 2020 00:38:51 +0300
> Subject: [PATCH] NET: r8169 driver identifying improvement.
>
> Trust device MAC enum + r8169d NIC soft reset
> before configuration.
>
> This commit adds enumeration check and allows
> driver to be slow to attach.
>
> Signed-off-by: Lauri Jakku <lja@iki.fi>
> ---
>  drivers/net/ethernet/realtek/r8169_main.c     | 24 +++--
>  .../net/ethernet/realtek/r8169_phy_config.c   | 80 ++++++++++++++++
>  drivers/net/phy/phy-core.c                    | 91 ++++++++++++++++++-
>  3 files changed, 182 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
> b/drivers/net/ethernet/realtek/r8169_main.c
> index bf5bf05970a2..a85764f6e448 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -640,7 +640,6 @@ MODULE_AUTHOR("Realtek and the Linux r8169 crew 
> <netdev@vger.kernel.org>");
>  MODULE_DESCRIPTION("RealTek RTL-8169 Gigabit Ethernet driver");
>  module_param_named(debug, debug.msg_enable, int, 0);
>  MODULE_PARM_DESC(debug, "Debug verbosity level (0=none, ..., 16=all)");
> -MODULE_SOFTDEP("pre: realtek");
>  MODULE_LICENSE("GPL");
>  MODULE_FIRMWARE(FIRMWARE_8168D_1);
>  MODULE_FIRMWARE(FIRMWARE_8168D_2);
> @@ -5172,13 +5171,18 @@ static int r8169_mdio_register(struct 
> rtl8169_private *tp)
>      if (!tp->phydev) {
>          mdiobus_unregister(new_bus);
>          return -ENODEV;
> -    } else if (!tp->phydev->drv) {
> -        /* Most chip versions fail with the genphy driver.
> -         * Therefore ensure that the dedicated PHY driver is loaded.
> -         */
> -        dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to 
> be added to initramfs?\n");
> -        mdiobus_unregister(new_bus);
> -        return -EUNATCH;
> +    } else {
> +        dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
> +        dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
> +
> +        if (tp->mac_version == RTL_GIGA_MAC_NONE) {
> +            /* Most chip versions fail with the genphy driver.
> +             * Therefore ensure that the dedicated PHY driver is loaded.
> +             */
> +            dev_err(&pdev->dev, "Not known MAC/PHY version.\n");
> +            mdiobus_unregister(new_bus);
> +            return -EUNATCH;
> +        }
>      }
>
>      /* PHY will be woken up in rtl_open() */
> @@ -5513,6 +5517,9 @@ static int rtl_init_one(struct pci_dev *pdev, 
> const struct pci_device_id *ent)
>             rtl_chip_infos[chipset].name, dev->dev_addr, xid,
>             pci_irq_vector(pdev, 0));
>
> +    dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
> +    dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
> +
>      if (jumbo_max)
>          netif_info(tp, probe, dev,
>                 "jumbo features [frames: %d bytes, tx checksumming: 
> %s]\n",
> @@ -5532,6 +5539,7 @@ static int rtl_init_one(struct pci_dev *pdev, 
> const struct pci_device_id *ent)
>      return rc;
>  }
>
> +
>  static struct pci_driver rtl8169_pci_driver = {
>      .name        = MODULENAME,
>      .id_table    = rtl8169_pci_tbl,
> diff --git a/drivers/net/ethernet/realtek/r8169_phy_config.c 
> b/drivers/net/ethernet/realtek/r8169_phy_config.c
> index b73f7d023e99..73d604c7a8a8 100644
> --- a/drivers/net/ethernet/realtek/r8169_phy_config.c
> +++ b/drivers/net/ethernet/realtek/r8169_phy_config.c
> @@ -1243,6 +1243,84 @@ static void rtl8125_2_hw_phy_config(struct 
> rtl8169_private *tp,
>      rtl8125_config_eee_phy(phydev);
>  }
>
> +
> +static void rtl8168d_hw_phy_pre_config_actions(struct rtl8169_private 
> *tp,
> +                            struct phy_device *phydev)
> +{
> +    /* Reset the PHY before configuration. There is BIOS bug that gives
> +     * random PHY ID when device is not soft resetted first. --lja
> +     */
> +    genphy_soft_reset(phydev);
> +}
> +
> +
> +
> +void r8169_hw_phy_pre_config_actions(struct rtl8169_private *tp, 
> struct phy_device *phydev,
> +                    enum mac_version ver)
> +{
> +    static const rtl_phy_cfg_fct phy_pre_config_actions[] = {
> +        /* PCI devices. */
> +        [RTL_GIGA_MAC_VER_02] = NULL,
> +        [RTL_GIGA_MAC_VER_03] = NULL,
> +        [RTL_GIGA_MAC_VER_04] = NULL,
> +        [RTL_GIGA_MAC_VER_05] = NULL,
> +        [RTL_GIGA_MAC_VER_06] = NULL,
> +        /* PCI-E devices. */
> +        [RTL_GIGA_MAC_VER_07] = NULL,
> +        [RTL_GIGA_MAC_VER_08] = NULL,
> +        [RTL_GIGA_MAC_VER_09] = NULL,
> +        [RTL_GIGA_MAC_VER_10] = NULL,
> +        [RTL_GIGA_MAC_VER_11] = NULL,
> +        [RTL_GIGA_MAC_VER_12] = NULL,
> +        [RTL_GIGA_MAC_VER_13] = NULL,
> +        [RTL_GIGA_MAC_VER_14] = NULL,
> +        [RTL_GIGA_MAC_VER_15] = NULL,
> +        [RTL_GIGA_MAC_VER_16] = NULL,
> +        [RTL_GIGA_MAC_VER_17] = NULL,
> +        [RTL_GIGA_MAC_VER_18] = NULL,
> +        [RTL_GIGA_MAC_VER_19] = NULL,
> +        [RTL_GIGA_MAC_VER_20] = NULL,
> +        [RTL_GIGA_MAC_VER_21] = NULL,
> +        [RTL_GIGA_MAC_VER_22] = NULL,
> +        [RTL_GIGA_MAC_VER_23] = NULL,
> +        [RTL_GIGA_MAC_VER_24] = NULL,
> +        [RTL_GIGA_MAC_VER_25] = rtl8168d_hw_phy_pre_config_actions,
> +        [RTL_GIGA_MAC_VER_26] = rtl8168d_hw_phy_pre_config_actions,
> +        [RTL_GIGA_MAC_VER_27] = rtl8168d_hw_phy_pre_config_actions,
> +        [RTL_GIGA_MAC_VER_28] = rtl8168d_hw_phy_pre_config_actions,
> +        [RTL_GIGA_MAC_VER_29] = NULL,
> +        [RTL_GIGA_MAC_VER_30] = NULL,
> +        [RTL_GIGA_MAC_VER_31] = NULL,
> +        [RTL_GIGA_MAC_VER_32] = NULL,
> +        [RTL_GIGA_MAC_VER_33] = NULL,
> +        [RTL_GIGA_MAC_VER_34] = NULL,
> +        [RTL_GIGA_MAC_VER_35] = NULL,
> +        [RTL_GIGA_MAC_VER_36] = NULL,
> +        [RTL_GIGA_MAC_VER_37] = NULL,
> +        [RTL_GIGA_MAC_VER_38] = NULL,
> +        [RTL_GIGA_MAC_VER_39] = NULL,
> +        [RTL_GIGA_MAC_VER_40] = NULL,
> +        [RTL_GIGA_MAC_VER_41] = NULL,
> +        [RTL_GIGA_MAC_VER_42] = NULL,
> +        [RTL_GIGA_MAC_VER_43] = NULL,
> +        [RTL_GIGA_MAC_VER_44] = NULL,
> +        [RTL_GIGA_MAC_VER_45] = NULL,
> +        [RTL_GIGA_MAC_VER_46] = NULL,
> +        [RTL_GIGA_MAC_VER_47] = NULL,
> +        [RTL_GIGA_MAC_VER_48] = NULL,
> +        [RTL_GIGA_MAC_VER_49] = NULL,
> +        [RTL_GIGA_MAC_VER_50] = NULL,
> +        [RTL_GIGA_MAC_VER_51] = NULL,
> +        [RTL_GIGA_MAC_VER_52] = NULL,
> +        [RTL_GIGA_MAC_VER_60] = NULL,
> +        [RTL_GIGA_MAC_VER_61] = NULL,
> +    };
> +
> +    if (phy_pre_config_actions[ver])
> +        phy_pre_config_actions[ver](tp, phydev);
> +}
> +
> +
>  void r8169_hw_phy_config(struct rtl8169_private *tp, struct 
> phy_device *phydev,
>               enum mac_version ver)
>  {
> @@ -1303,6 +1381,8 @@ void r8169_hw_phy_config(struct rtl8169_private 
> *tp, struct phy_device *phydev,
>          [RTL_GIGA_MAC_VER_60] = rtl8125_1_hw_phy_config,
>          [RTL_GIGA_MAC_VER_61] = rtl8125_2_hw_phy_config,
>      };
> +
> +    r8169_hw_phy_pre_config_actions(tp, phydev);
>
>      if (phy_configs[ver])
>          phy_configs[ver](tp, phydev);
> diff --git a/drivers/net/phy/phy-core.c b/drivers/net/phy/phy-core.c
> index 66b8c61ca74c..b170185a1ed1 100644
> --- a/drivers/net/phy/phy-core.c
> +++ b/drivers/net/phy/phy-core.c
> @@ -151,6 +151,25 @@ static const struct phy_setting settings[] = {
>  };
>  #undef PHY_SETTING
>
> +#ifdef DEBUG
> +#define R8169_ATTACHED_PRINTK \
> +    printk("Phy is attached check %s @ %d: %d\n", \
> +    __FUNC__, __LINE__, is_attached_check);
> +
> +#else
> +#define R8169_ATTACHED_PRINTK
> +#endif
> +
> +#define PHY_NOT_ATTACHED_CHECK \
> +    { \
> +        void *attached_dev_ptr = (phydev) ? phydev->attached_dev : 
> NULL; \
> +        int is_attached_check = (attached_dev_ptr != NULL) && \
> +                                ((phydev) && (phydev->state >= 
> PHY_READY)); \
> +        R8169_ATTACHED_PRINTK \
> +        if (! is_attached_check ) return -EOPNOTSUPP; \
> +    };
> +
> +
>  /**
>   * phy_lookup_setting - lookup a PHY setting
>   * @speed: speed to match
> @@ -457,6 +476,9 @@ int phy_read_mmd(struct phy_device *phydev, int 
> devad, u32 regnum)
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      phy_lock_mdio_bus(phydev);
>      ret = __phy_read_mmd(phydev, devad, regnum);
>      phy_unlock_mdio_bus(phydev);
> @@ -479,6 +501,9 @@ int __phy_write_mmd(struct phy_device *phydev, int 
> devad, u32 regnum, u16 val)
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      if (regnum > (u16)~0 || devad > 32)
>          return -EINVAL;
>
> @@ -518,6 +543,9 @@ int phy_write_mmd(struct phy_device *phydev, int 
> devad, u32 regnum, u16 val)
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      phy_lock_mdio_bus(phydev);
>      ret = __phy_write_mmd(phydev, devad, regnum, val);
>      phy_unlock_mdio_bus(phydev);
> @@ -543,6 +571,10 @@ int phy_modify_changed(struct phy_device *phydev, 
> u32 regnum, u16 mask, u16 set)
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +
>      phy_lock_mdio_bus(phydev);
>      ret = __phy_modify_changed(phydev, regnum, mask, set);
>      phy_unlock_mdio_bus(phydev);
> @@ -587,6 +619,9 @@ int phy_modify(struct phy_device *phydev, u32 
> regnum, u16 mask, u16 set)
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      phy_lock_mdio_bus(phydev);
>      ret = __phy_modify(phydev, regnum, mask, set);
>      phy_unlock_mdio_bus(phydev);
> @@ -613,6 +648,9 @@ int __phy_modify_mmd_changed(struct phy_device 
> *phydev, int devad, u32 regnum,
>  {
>      int new, ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      ret = __phy_read_mmd(phydev, devad, regnum);
>      if (ret < 0)
>          return ret;
> @@ -646,6 +684,10 @@ int phy_modify_mmd_changed(struct phy_device 
> *phydev, int devad, u32 regnum,
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +
>      phy_lock_mdio_bus(phydev);
>      ret = __phy_modify_mmd_changed(phydev, devad, regnum, mask, set);
>      phy_unlock_mdio_bus(phydev);
> @@ -671,6 +713,9 @@ int __phy_modify_mmd(struct phy_device *phydev, 
> int devad, u32 regnum,
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      ret = __phy_modify_mmd_changed(phydev, devad, regnum, mask, set);
>
>      return ret < 0 ? ret : 0;
> @@ -694,6 +739,9 @@ int phy_modify_mmd(struct phy_device *phydev, int 
> devad, u32 regnum,
>  {
>      int ret;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      phy_lock_mdio_bus(phydev);
>      ret = __phy_modify_mmd(phydev, devad, regnum, mask, set);
>      phy_unlock_mdio_bus(phydev);
> @@ -704,7 +752,11 @@ EXPORT_SYMBOL_GPL(phy_modify_mmd);
>
>  static int __phy_read_page(struct phy_device *phydev)
>  {
> -    if (WARN_ONCE(!phydev->drv->read_page, "read_page callback not 
> available, PHY driver not loaded?\n"))
> +
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +    if (WARN(!phydev->drv->read_page, "read_page callback not 
> available, PHY driver not loaded?\n"))
>          return -EOPNOTSUPP;
>
>      return phydev->drv->read_page(phydev);
> @@ -712,12 +764,16 @@ static int __phy_read_page(struct phy_device 
> *phydev)
>
>  static int __phy_write_page(struct phy_device *phydev, int page)
>  {
> -    if (WARN_ONCE(!phydev->drv->write_page, "write_page callback not 
> available, PHY driver not loaded?\n"))
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +    if (WARN(!phydev->drv->write_page, "write_page callback not 
> available, PHY driver not loaded?\n"))
>          return -EOPNOTSUPP;
>
>      return phydev->drv->write_page(phydev, page);
>  }
>
> +
>  /**
>   * phy_save_page() - take the bus lock and save the current page
>   * @phydev: a pointer to a &struct phy_device
> @@ -728,7 +784,11 @@ static int __phy_write_page(struct phy_device 
> *phydev, int page)
>   */
>  int phy_save_page(struct phy_device *phydev)
>  {
> -    phy_lock_mdio_bus(phydev);
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +
> +    phy_lock_mdio_bus(phydev);
>      return __phy_read_page(phydev);
>  }
>  EXPORT_SYMBOL_GPL(phy_save_page);
> @@ -748,7 +808,10 @@ int phy_select_page(struct phy_device *phydev, 
> int page)
>  {
>      int ret, oldpage;
>
> -    oldpage = ret = phy_save_page(phydev);
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +    oldpage = ret = phy_save_page(phydev);
>      if (ret < 0)
>          return ret;
>
> @@ -782,6 +845,9 @@ int phy_restore_page(struct phy_device *phydev, 
> int oldpage, int ret)
>  {
>      int r;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
>      if (oldpage >= 0) {
>          r = __phy_write_page(phydev, oldpage);
>
> @@ -812,6 +878,9 @@ EXPORT_SYMBOL_GPL(phy_restore_page);
>  int phy_read_paged(struct phy_device *phydev, int page, u32 regnum)
>  {
>      int ret = 0, oldpage;
> +
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
>
>      oldpage = phy_select_page(phydev, page);
>      if (oldpage >= 0)
> @@ -834,6 +903,10 @@ int phy_write_paged(struct phy_device *phydev, 
> int page, u32 regnum, u16 val)
>  {
>      int ret = 0, oldpage;
>
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +
>      oldpage = phy_select_page(phydev, page);
>      if (oldpage >= 0)
>          ret = __phy_write(phydev, regnum, val);
> @@ -856,6 +929,9 @@ int phy_modify_paged_changed(struct phy_device 
> *phydev, int page, u32 regnum,
>                   u16 mask, u16 set)
>  {
>      int ret = 0, oldpage;
> +
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
>
>      oldpage = phy_select_page(phydev, page);
>      if (oldpage >= 0)
> @@ -878,7 +954,12 @@ EXPORT_SYMBOL(phy_modify_paged_changed);
>  int phy_modify_paged(struct phy_device *phydev, int page, u32 regnum,
>               u16 mask, u16 set)
>  {
> -    int ret = phy_modify_paged_changed(phydev, page, regnum, mask, set);
> +    int ret = -1;
> +
> +    /* If not attached, do nothing. */
> +    PHY_NOT_ATTACHED_CHECK;
> +
> +    ret = phy_modify_paged_changed(phydev, page, regnum, mask, set);
>
>      return ret < 0 ? ret : 0;
>  }

here is the polished patch:

 From 6dd6f2813c543dc728efb8dca796bbbe870bd031 Mon Sep 17 00:00:00 2001
From: Lauri Jakku <lja@iki.fi>
Date: Thu, 16 Apr 2020 00:38:51 +0300
Subject: [PATCH] NET: r8169 driver identifying improvement.

Trust device MAC enum + r8169d NIC soft reset
before configuration.

This commit adds enumeration check and allows
driver to be slow to attach.

Signed-off-by: Lauri Jakku <lja@iki.fi>
---
  drivers/net/ethernet/realtek/r8169_main.c     | 21 +++--
  .../net/ethernet/realtek/r8169_phy_config.c   | 80 +++++++++++++++++++
  2 files changed, 93 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c 
b/drivers/net/ethernet/realtek/r8169_main.c
index bf5bf05970a2..6828e755a460 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -640,7 +640,6 @@ MODULE_AUTHOR("Realtek and the Linux r8169 crew 
<netdev@vger.kernel.org>");
  MODULE_DESCRIPTION("RealTek RTL-8169 Gigabit Ethernet driver");
  module_param_named(debug, debug.msg_enable, int, 0);
  MODULE_PARM_DESC(debug, "Debug verbosity level (0=none, ..., 16=all)");
-MODULE_SOFTDEP("pre: realtek");
  MODULE_LICENSE("GPL");
  MODULE_FIRMWARE(FIRMWARE_8168D_1);
  MODULE_FIRMWARE(FIRMWARE_8168D_2);
@@ -5172,13 +5171,18 @@ static int r8169_mdio_register(struct 
rtl8169_private *tp)
      if (!tp->phydev) {
          mdiobus_unregister(new_bus);
          return -ENODEV;
-    } else if (!tp->phydev->drv) {
-        /* Most chip versions fail with the genphy driver.
-         * Therefore ensure that the dedicated PHY driver is loaded.
-         */
-        dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to 
be added to initramfs?\n");
-        mdiobus_unregister(new_bus);
-        return -EUNATCH;
+    } else {
+        dev_info(&pdev->dev, "PHY version: 0x%x\n", tp->phydev->phy_id);
+        dev_info(&pdev->dev, "MAC version: %d\n", tp->mac_version);
+
+        if (tp->mac_version == RTL_GIGA_MAC_NONE) {
+            /* Most chip versions fail with the genphy driver.
+             * Therefore ensure that the dedicated PHY driver is loaded.
+             */
+            dev_err(&pdev->dev, "Not known MAC/PHY version.\n");
+            mdiobus_unregister(new_bus);
+            return -EUNATCH;
+        }
      }

      /* PHY will be woken up in rtl_open() */
@@ -5532,6 +5536,7 @@ static int rtl_init_one(struct pci_dev *pdev, 
const struct pci_device_id *ent)
      return rc;
  }

+
  static struct pci_driver rtl8169_pci_driver = {
      .name        = MODULENAME,
      .id_table    = rtl8169_pci_tbl,
diff --git a/drivers/net/ethernet/realtek/r8169_phy_config.c 
b/drivers/net/ethernet/realtek/r8169_phy_config.c
index b73f7d023e99..f13f68b79a92 100644
--- a/drivers/net/ethernet/realtek/r8169_phy_config.c
+++ b/drivers/net/ethernet/realtek/r8169_phy_config.c
@@ -1243,6 +1243,84 @@ static void rtl8125_2_hw_phy_config(struct 
rtl8169_private *tp,
      rtl8125_config_eee_phy(phydev);
  }

+
+static void rtl8168d_hw_phy_pre_config_actions(struct rtl8169_private *tp,
+                            struct phy_device *phydev)
+{
+    /* Reset the PHY before configuration. There is BIOS bug that gives
+     * random PHY ID when device is not soft resetted first. --lja
+     */
+    genphy_soft_reset(phydev);
+}
+
+
+
+void r8169_hw_phy_pre_config_actions(struct rtl8169_private *tp, struct 
phy_device *phydev,