LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Steen Hegelund <steen.hegelund@microchip.com>
To: "David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>, <UNGLinuxDriver@microchip.com>,
	<netdev@vger.kernel.org>, <linux-arm-kernel@lists.infradead.org>,
	<linux-kernel@vger.kernel.org>
Cc: Steen Hegelund <steen.hegelund@microchip.com>,
	Alexandre Belloni <alexandre.belloni@bootlin.com>
Subject: [PATCH net-next v2 1/2] net: sparx5: switchdev: adding frame DMA functionality
Date: Thu, 19 Aug 2021 09:39:39 +0200	[thread overview]
Message-ID: <20210819073940.1589383-2-steen.hegelund@microchip.com> (raw)
In-Reply-To: <20210819073940.1589383-1-steen.hegelund@microchip.com>

This add frame DMA functionality to the Sparx5 platform.

Ethernet frames can be extracted or injected autonomously to or from the
device’s DDR3/DDR3L memory and/or PCIe memory space. Linked list data
structures in memory are used for injecting or extracting Ethernet frames.
The FDMA generates interrupts when frame extraction or injection is done
and when the linked lists need updating.

The FDMA implements two extraction channels, one per switch core port
towards the VCore CPU system and a total of six injection channels.
Extraction channels are mapped one-to-one to the CPU ports, while injection
channels can be individually assigned to any CPU port.

- FDMA channel 0 through 5 corresponds to CPU port 0 injection direction
  FDMA_CH_CFG[channel].CH_INJ_PORT is set to 0.
- FDMA channel 0 through 5 corresponds to CPU port 1 injection direction when
  FDMA_CH_CFG[channel].CH_INJ_PORT is set to 1.
- FDMA channel 6 corresponds to CPU port 0 extraction direction.
- FDMA channel 7 corresponds to CPU port 1 extraction direction.

The FDMA implements a strict priority scheme among channels. Extraction
channels are prioritized over injection channels and secondarily channels
with higher channel number are prioritized over channels with lower number.
On the other hand, ports are being served on an equal-bandwidth principle
both on injection and extraction directions.  The equal-bandwidth principle
will not force an equal bandwidth. Instead, it ensures that the ports
perform at their best considering the operating conditions.

When more than one injection channel is enabled for injection on the same
CPU port, priority determines which channel can inject data. Ownership
is re-arbitrated on frame boundaries.

The FDMA processes linked lists of DMA Control Block Structures (DCBs). The
DCBs have the same basic structure for both injection and extraction. A DCB
must be placed on a 64-bit word-aligned address in memory. Each DCB has a
per-channel configurable amount of associated data blocks in memory, where
the frame data is stored.

The data blocks that are used by extraction channels must be placed on
64-bit word aligned addresses in memory, and their length must be a
multiple of 128 bytes.

A DCB carries the pointer to the next DCB of the linked list, the INFO word
which holds information for the DCB, and a pair of status word and memory
pointer for every data block that it is associated with.

Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
---
 .../net/ethernet/microchip/sparx5/Makefile    |   2 +-
 .../ethernet/microchip/sparx5/sparx5_fdma.c   | 593 ++++++++++++++++++
 .../ethernet/microchip/sparx5/sparx5_main.c   |  23 +-
 .../ethernet/microchip/sparx5/sparx5_main.h   |  69 ++
 .../ethernet/microchip/sparx5/sparx5_packet.c |  13 +-
 .../ethernet/microchip/sparx5/sparx5_port.c   |   2 +-
 .../ethernet/microchip/sparx5/sparx5_port.h   |   1 +
 7 files changed, 693 insertions(+), 10 deletions(-)
 create mode 100644 drivers/net/ethernet/microchip/sparx5/sparx5_fdma.c

diff --git a/drivers/net/ethernet/microchip/sparx5/Makefile b/drivers/net/ethernet/microchip/sparx5/Makefile
index faa8f07a6b75..c271e86ee292 100644
--- a/drivers/net/ethernet/microchip/sparx5/Makefile
+++ b/drivers/net/ethernet/microchip/sparx5/Makefile
@@ -7,4 +7,4 @@ obj-$(CONFIG_SPARX5_SWITCH) += sparx5-switch.o
 
 sparx5-switch-objs  := sparx5_main.o sparx5_packet.o \
  sparx5_netdev.o sparx5_phylink.o sparx5_port.o sparx5_mactable.o sparx5_vlan.o \
- sparx5_switchdev.o sparx5_calendar.o sparx5_ethtool.o
+ sparx5_switchdev.o sparx5_calendar.o sparx5_ethtool.o sparx5_fdma.o
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_fdma.c b/drivers/net/ethernet/microchip/sparx5/sparx5_fdma.c
new file mode 100644
index 000000000000..7436f62fa152
--- /dev/null
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_fdma.c
@@ -0,0 +1,593 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* Microchip Sparx5 Switch driver
+ *
+ * Copyright (c) 2021 Microchip Technology Inc. and its subsidiaries.
+ *
+ * The Sparx5 Chip Register Model can be browsed at this location:
+ * https://github.com/microchip-ung/sparx-5_reginfo
+ */
+
+#include <linux/types.h>
+#include <linux/skbuff.h>
+#include <linux/netdevice.h>
+#include <linux/interrupt.h>
+#include <linux/ip.h>
+#include <linux/dma-mapping.h>
+
+#include "sparx5_main_regs.h"
+#include "sparx5_main.h"
+#include "sparx5_port.h"
+
+#define FDMA_XTR_CHANNEL		6
+#define FDMA_INJ_CHANNEL		0
+
+#define FDMA_DCB_INFO_DATAL(x)		((x) & GENMASK(15, 0))
+#define FDMA_DCB_INFO_TOKEN		BIT(17)
+#define FDMA_DCB_INFO_INTR		BIT(18)
+#define FDMA_DCB_INFO_SW(x)		(((x) << 24) & GENMASK(31, 24))
+
+#define FDMA_DCB_STATUS_BLOCKL(x)	((x) & GENMASK(15, 0))
+#define FDMA_DCB_STATUS_SOF		BIT(16)
+#define FDMA_DCB_STATUS_EOF		BIT(17)
+#define FDMA_DCB_STATUS_INTR		BIT(18)
+#define FDMA_DCB_STATUS_DONE		BIT(19)
+#define FDMA_DCB_STATUS_BLOCKO(x)	(((x) << 20) & GENMASK(31, 20))
+#define FDMA_DCB_INVALID_DATA		0x1
+
+#define FDMA_XTR_BUFFER_SIZE		2048
+#define FDMA_WEIGHT			4
+
+/* Frame DMA DCB format
+ *
+ * +---------------------------+
+ * |         Next Ptr          |
+ * +---------------------------+
+ * |   Reserved  |    Info     |
+ * +---------------------------+
+ * |         Data0 Ptr         |
+ * +---------------------------+
+ * |   Reserved  |    Status0  |
+ * +---------------------------+
+ * |         Data1 Ptr         |
+ * +---------------------------+
+ * |   Reserved  |    Status1  |
+ * +---------------------------+
+ * |         Data2 Ptr         |
+ * +---------------------------+
+ * |   Reserved  |    Status2  |
+ * |-------------|-------------|
+ * |                           |
+ * |                           |
+ * |                           |
+ * |                           |
+ * |                           |
+ * |---------------------------|
+ * |         Data14 Ptr        |
+ * +-------------|-------------+
+ * |   Reserved  |    Status14 |
+ * +-------------|-------------+
+ */
+
+/* For each hardware DB there is an entry in this list and when the HW DB
+ * entry is used, this SW DB entry is moved to the back of the list
+ */
+struct sparx5_db {
+	struct list_head list;
+	void *cpu_addr;
+};
+
+static void sparx5_fdma_rx_add_dcb(struct sparx5_rx *rx,
+				   struct sparx5_rx_dcb_hw *dcb,
+				   u64 nextptr)
+{
+	int idx = 0;
+
+	/* Reset the status of the DB */
+	for (idx = 0; idx < FDMA_RX_DCB_MAX_DBS; ++idx) {
+		struct sparx5_db_hw *db = &dcb->db[idx];
+
+		db->status = FDMA_DCB_STATUS_INTR;
+	}
+	dcb->nextptr = FDMA_DCB_INVALID_DATA;
+	dcb->info = FDMA_DCB_INFO_DATAL(FDMA_XTR_BUFFER_SIZE);
+	rx->last_entry->nextptr = nextptr;
+	rx->last_entry = dcb;
+}
+
+static void sparx5_fdma_tx_add_dcb(struct sparx5_tx *tx,
+				   struct sparx5_tx_dcb_hw *dcb,
+				   u64 nextptr)
+{
+	int idx = 0;
+
+	/* Reset the status of the DB */
+	for (idx = 0; idx < FDMA_TX_DCB_MAX_DBS; ++idx) {
+		struct sparx5_db_hw *db = &dcb->db[idx];
+
+		db->status = FDMA_DCB_STATUS_DONE;
+	}
+	dcb->nextptr = FDMA_DCB_INVALID_DATA;
+	dcb->info = FDMA_DCB_INFO_DATAL(FDMA_XTR_BUFFER_SIZE);
+}
+
+static void sparx5_fdma_rx_activate(struct sparx5 *sparx5, struct sparx5_rx *rx)
+{
+	/* Write the buffer address in the LLP and LLP1 regs */
+	spx5_wr(((u64)rx->dma) & GENMASK(31, 0), sparx5,
+		FDMA_DCB_LLP(rx->channel_id));
+	spx5_wr(((u64)rx->dma) >> 32, sparx5, FDMA_DCB_LLP1(rx->channel_id));
+
+	/* Set the number of RX DBs to be used, and DB end-of-frame interrupt */
+	spx5_wr(FDMA_CH_CFG_CH_DCB_DB_CNT_SET(FDMA_RX_DCB_MAX_DBS) |
+		FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY_SET(1) |
+		FDMA_CH_CFG_CH_INJ_PORT_SET(XTR_QUEUE),
+		sparx5, FDMA_CH_CFG(rx->channel_id));
+
+	/* Set the RX Watermark to max */
+	spx5_rmw(FDMA_XTR_CFG_XTR_FIFO_WM_SET(31), FDMA_XTR_CFG_XTR_FIFO_WM,
+		 sparx5,
+		 FDMA_XTR_CFG);
+
+	/* Start RX fdma */
+	spx5_rmw(FDMA_PORT_CTRL_XTR_STOP_SET(0), FDMA_PORT_CTRL_XTR_STOP,
+		 sparx5, FDMA_PORT_CTRL(0));
+
+	/* Enable RX channel DB interrupt */
+	spx5_rmw(BIT(rx->channel_id),
+		 BIT(rx->channel_id) & FDMA_INTR_DB_ENA_INTR_DB_ENA,
+		 sparx5, FDMA_INTR_DB_ENA);
+
+	/* Activate the RX channel */
+	spx5_wr(BIT(rx->channel_id), sparx5, FDMA_CH_ACTIVATE);
+}
+
+static void sparx5_fdma_rx_deactivate(struct sparx5 *sparx5, struct sparx5_rx *rx)
+{
+	/* Dectivate the RX channel */
+	spx5_rmw(0, BIT(rx->channel_id) & FDMA_CH_ACTIVATE_CH_ACTIVATE,
+		 sparx5, FDMA_CH_ACTIVATE);
+
+	/* Disable RX channel DB interrupt */
+	spx5_rmw(0, BIT(rx->channel_id) & FDMA_INTR_DB_ENA_INTR_DB_ENA,
+		 sparx5, FDMA_INTR_DB_ENA);
+
+	/* Stop RX fdma */
+	spx5_rmw(FDMA_PORT_CTRL_XTR_STOP_SET(1), FDMA_PORT_CTRL_XTR_STOP,
+		 sparx5, FDMA_PORT_CTRL(0));
+}
+
+static void sparx5_fdma_tx_activate(struct sparx5 *sparx5, struct sparx5_tx *tx)
+{
+	/* Write the buffer address in the LLP and LLP1 regs */
+	spx5_wr(((u64)tx->dma) & GENMASK(31, 0), sparx5,
+		FDMA_DCB_LLP(tx->channel_id));
+	spx5_wr(((u64)tx->dma) >> 32, sparx5, FDMA_DCB_LLP1(tx->channel_id));
+
+	/* Set the number of TX DBs to be used, and DB end-of-frame interrupt */
+	spx5_wr(FDMA_CH_CFG_CH_DCB_DB_CNT_SET(FDMA_TX_DCB_MAX_DBS) |
+		FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY_SET(1) |
+		FDMA_CH_CFG_CH_INJ_PORT_SET(INJ_QUEUE),
+		sparx5, FDMA_CH_CFG(tx->channel_id));
+
+	/* Start TX fdma */
+	spx5_rmw(FDMA_PORT_CTRL_INJ_STOP_SET(0), FDMA_PORT_CTRL_INJ_STOP,
+		 sparx5, FDMA_PORT_CTRL(0));
+
+	/* Activate the channel */
+	spx5_wr(BIT(tx->channel_id), sparx5, FDMA_CH_ACTIVATE);
+}
+
+static void sparx5_fdma_tx_deactivate(struct sparx5 *sparx5, struct sparx5_tx *tx)
+{
+	/* Disable the channel */
+	spx5_rmw(0, BIT(tx->channel_id) & FDMA_CH_ACTIVATE_CH_ACTIVATE,
+		 sparx5, FDMA_CH_ACTIVATE);
+}
+
+static void sparx5_fdma_rx_reload(struct sparx5 *sparx5, struct sparx5_rx *rx)
+{
+	/* Reload the RX channel */
+	spx5_wr(BIT(rx->channel_id), sparx5, FDMA_CH_RELOAD);
+}
+
+static void sparx5_fdma_tx_reload(struct sparx5 *sparx5, struct sparx5_tx *tx)
+{
+	/* Reload the TX channel */
+	spx5_wr(BIT(tx->channel_id), sparx5, FDMA_CH_RELOAD);
+}
+
+static struct sk_buff *sparx5_fdma_rx_alloc_skb(struct sparx5_rx *rx)
+{
+	return __netdev_alloc_skb(rx->ndev, FDMA_XTR_BUFFER_SIZE,
+				  GFP_ATOMIC);
+}
+
+static bool sparx5_fdma_rx_get_frame(struct sparx5 *sparx5, struct sparx5_rx *rx)
+{
+	struct sparx5_db_hw *db_hw;
+	unsigned int packet_size;
+	struct sparx5_port *port;
+	struct sk_buff *new_skb;
+	struct frame_info fi;
+	struct sk_buff *skb;
+	dma_addr_t dma_addr;
+
+	/* Check if the DCB is done */
+	db_hw = &rx->dcb_entries[rx->dcb_index].db[rx->db_index];
+	if (unlikely(!(db_hw->status & FDMA_DCB_STATUS_DONE)))
+		return false;
+	skb = rx->skb[rx->dcb_index][rx->db_index];
+	/* Replace the DB entry with a new SKB */
+	new_skb = sparx5_fdma_rx_alloc_skb(rx);
+	if (unlikely(!new_skb))
+		return false;
+	/* Map the new skb data and set the new skb */
+	dma_addr = virt_to_phys(new_skb->data);
+	rx->skb[rx->dcb_index][rx->db_index] = new_skb;
+	db_hw->dataptr = dma_addr;
+	packet_size = FDMA_DCB_STATUS_BLOCKL(db_hw->status);
+	skb_put(skb, packet_size);
+	/* Now do the normal processing of the skb */
+	sparx5_ifh_parse((u32 *)skb->data, &fi);
+	/* Map to port netdev */
+	port = fi.src_port < SPX5_PORTS ?  sparx5->ports[fi.src_port] : NULL;
+	if (!port || !port->ndev) {
+		dev_err(sparx5->dev, "Data on inactive port %d\n", fi.src_port);
+		sparx5_xtr_flush(sparx5, XTR_QUEUE);
+		return false;
+	}
+	skb->dev = port->ndev;
+	skb_pull(skb, IFH_LEN * sizeof(u32));
+	if (likely(!(skb->dev->features & NETIF_F_RXFCS)))
+		skb_trim(skb, skb->len - ETH_FCS_LEN);
+	skb->protocol = eth_type_trans(skb, skb->dev);
+	/* Everything we see on an interface that is in the HW bridge
+	 * has already been forwarded
+	 */
+	if (test_bit(port->portno, sparx5->bridge_mask))
+		skb->offload_fwd_mark = 1;
+	skb->dev->stats.rx_bytes += skb->len;
+	skb->dev->stats.rx_packets++;
+	rx->packets++;
+	netif_receive_skb(skb);
+	return true;
+}
+
+static int sparx5_fdma_napi_callback(struct napi_struct *napi, int weight)
+{
+	struct sparx5_rx *rx = container_of(napi, struct sparx5_rx, napi);
+	struct sparx5 *sparx5 = container_of(rx, struct sparx5, rx);
+	int counter = 0;
+
+	while (counter < weight && sparx5_fdma_rx_get_frame(sparx5, rx)) {
+		struct sparx5_rx_dcb_hw *old_dcb;
+
+		rx->db_index++;
+		counter++;
+		/* Check if the DCB can be reused */
+		if (rx->db_index != FDMA_RX_DCB_MAX_DBS)
+			continue;
+		/* As the DCB  can be reused, just advance the dcb_index
+		 * pointer and set the nextptr in the DCB
+		 */
+		rx->db_index = 0;
+		old_dcb = &rx->dcb_entries[rx->dcb_index];
+		rx->dcb_index++;
+		rx->dcb_index &= FDMA_DCB_MAX - 1;
+		sparx5_fdma_rx_add_dcb(rx, old_dcb,
+				       rx->dma +
+				       ((unsigned long)old_dcb -
+					(unsigned long)rx->dcb_entries));
+	}
+	if (counter < weight) {
+		napi_complete_done(&rx->napi, counter);
+		spx5_rmw(BIT(rx->channel_id),
+			 BIT(rx->channel_id) & FDMA_INTR_DB_ENA_INTR_DB_ENA,
+			 sparx5, FDMA_INTR_DB_ENA);
+	}
+	if (counter)
+		sparx5_fdma_rx_reload(sparx5, rx);
+	return counter;
+}
+
+static struct sparx5_tx_dcb_hw *sparx5_fdma_next_dcb(struct sparx5_tx *tx,
+						     struct sparx5_tx_dcb_hw *dcb)
+{
+	struct sparx5_tx_dcb_hw *next_dcb;
+
+	next_dcb = dcb;
+	next_dcb++;
+	/* Handle wrap-around */
+	if ((unsigned long)next_dcb >=
+	    ((unsigned long)tx->first_entry + FDMA_DCB_MAX * sizeof(*dcb)))
+		next_dcb = tx->first_entry;
+	return next_dcb;
+}
+
+int sparx5_fdma_xmit(struct sparx5 *sparx5, u32 *ifh, struct sk_buff *skb)
+{
+	struct sparx5_tx_dcb_hw *next_dcb_hw;
+	struct sparx5_tx *tx = &sparx5->tx;
+	static bool first_time = true;
+	struct sparx5_db_hw *db_hw;
+	struct sparx5_db *db;
+
+	next_dcb_hw = sparx5_fdma_next_dcb(tx, tx->curr_entry);
+	db_hw = &next_dcb_hw->db[0];
+	if (!(db_hw->status & FDMA_DCB_STATUS_DONE))
+		tx->dropped++;
+	db = list_first_entry(&tx->db_list, struct sparx5_db, list);
+	list_move_tail(&db->list, &tx->db_list);
+	next_dcb_hw->nextptr = FDMA_DCB_INVALID_DATA;
+	tx->curr_entry->nextptr = tx->dma +
+		((unsigned long)next_dcb_hw -
+		 (unsigned long)tx->first_entry);
+	tx->curr_entry = next_dcb_hw;
+	memset(db->cpu_addr, 0, FDMA_XTR_BUFFER_SIZE);
+	memcpy(db->cpu_addr, ifh, IFH_LEN * 4);
+	memcpy(db->cpu_addr + IFH_LEN * 4, skb->data, skb->len);
+	db_hw->status = FDMA_DCB_STATUS_SOF |
+			FDMA_DCB_STATUS_EOF |
+			FDMA_DCB_STATUS_BLOCKO(0) |
+			FDMA_DCB_STATUS_BLOCKL(skb->len + IFH_LEN * 4 + 4);
+	if (first_time) {
+		sparx5_fdma_tx_activate(sparx5, tx);
+		first_time = false;
+	} else {
+		sparx5_fdma_tx_reload(sparx5, tx);
+	}
+	return NETDEV_TX_OK;
+}
+
+static int sparx5_fdma_rx_alloc(struct sparx5 *sparx5)
+{
+	struct sparx5_rx *rx = &sparx5->rx;
+	struct sparx5_rx_dcb_hw *dcb;
+	int idx, jdx;
+	int size;
+
+	size = sizeof(struct sparx5_rx_dcb_hw) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+	rx->dcb_entries = devm_kzalloc(sparx5->dev, size, GFP_KERNEL);
+	if (!rx->dcb_entries)
+		return -ENOMEM;
+	rx->dma = virt_to_phys(rx->dcb_entries);
+	rx->last_entry = rx->dcb_entries;
+	rx->db_index = 0;
+	rx->dcb_index = 0;
+	/* Now for each dcb allocate the db */
+	for (idx = 0; idx < FDMA_DCB_MAX; ++idx) {
+		dcb = &rx->dcb_entries[idx];
+		dcb->info = 0;
+		/* For each db allocate an skb and map skb data pointer to the DB
+		 * dataptr. In this way when the frame is received the skb->data
+		 * will contain the frame, so no memcpy is needed
+		 */
+		for (jdx = 0; jdx < FDMA_RX_DCB_MAX_DBS; ++jdx) {
+			struct sparx5_db_hw *db_hw = &dcb->db[jdx];
+			dma_addr_t dma_addr;
+			struct sk_buff *skb;
+
+			skb = sparx5_fdma_rx_alloc_skb(rx);
+			if (!skb)
+				return -ENOMEM;
+
+			dma_addr = virt_to_phys(skb->data);
+			db_hw->dataptr = dma_addr;
+			db_hw->status = 0;
+			rx->skb[idx][jdx] = skb;
+		}
+		sparx5_fdma_rx_add_dcb(rx, dcb, rx->dma + sizeof(*dcb) * idx);
+	}
+	netif_napi_add(rx->ndev, &rx->napi, sparx5_fdma_napi_callback, FDMA_WEIGHT);
+	napi_enable(&rx->napi);
+	sparx5_fdma_rx_activate(sparx5, rx);
+	return 0;
+}
+
+static int sparx5_fdma_tx_alloc(struct sparx5 *sparx5)
+{
+	struct sparx5_tx *tx = &sparx5->tx;
+	struct sparx5_tx_dcb_hw *dcb;
+	int idx, jdx;
+	int size;
+
+	size = sizeof(struct sparx5_tx_dcb_hw) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+	tx->curr_entry = devm_kzalloc(sparx5->dev, size, GFP_KERNEL);
+	if (!tx->curr_entry)
+		return -ENOMEM;
+	tx->dma = virt_to_phys(tx->curr_entry);
+	tx->first_entry = tx->curr_entry;
+	INIT_LIST_HEAD(&tx->db_list);
+	/* Now for each dcb allocate the db */
+	for (idx = 0; idx < FDMA_DCB_MAX; ++idx) {
+		dcb = &tx->curr_entry[idx];
+		dcb->info = 0;
+		/* TX databuffers must be 16byte aligned */
+		for (jdx = 0; jdx < FDMA_TX_DCB_MAX_DBS; ++jdx) {
+			struct sparx5_db_hw *db_hw = &dcb->db[jdx];
+			struct sparx5_db *db;
+			dma_addr_t phys;
+			void *cpu_addr;
+
+			cpu_addr = devm_kzalloc(sparx5->dev,
+						FDMA_XTR_BUFFER_SIZE,
+						GFP_KERNEL);
+			if (!cpu_addr)
+				return -ENOMEM;
+			phys = virt_to_phys(cpu_addr);
+			db_hw->dataptr = phys;
+			db_hw->status = 0;
+			db = devm_kzalloc(sparx5->dev, sizeof(*db), GFP_KERNEL);
+			db->cpu_addr = cpu_addr;
+			list_add_tail(&db->list, &tx->db_list);
+		}
+		sparx5_fdma_tx_add_dcb(tx, dcb, tx->dma + sizeof(*dcb) * idx);
+		/* Let the curr_entry to point to the last allocated entry */
+		if (idx == FDMA_DCB_MAX - 1)
+			tx->curr_entry = dcb;
+	}
+	return 0;
+}
+
+static void sparx5_fdma_rx_init(struct sparx5 *sparx5,
+				struct sparx5_rx *rx, int channel)
+{
+	int idx;
+
+	rx->channel_id = channel;
+	/* Fetch a netdev for SKB and NAPI use, any will do */
+	for (idx = 0; idx < SPX5_PORTS; ++idx) {
+		struct sparx5_port *port = sparx5->ports[idx];
+
+		if (port && port->ndev) {
+			rx->ndev = port->ndev;
+			break;
+		}
+	}
+}
+
+static void sparx5_fdma_tx_init(struct sparx5 *sparx5,
+				struct sparx5_tx *tx, int channel)
+{
+	tx->channel_id = channel;
+}
+
+irqreturn_t sparx5_fdma_handler(int irq, void *args)
+{
+	struct sparx5 *sparx5 = args;
+	u32 db = 0, err = 0;
+
+	db = spx5_rd(sparx5, FDMA_INTR_DB);
+	err = spx5_rd(sparx5, FDMA_INTR_ERR);
+	/* Clear interrupt */
+	if (db) {
+		spx5_wr(0, sparx5, FDMA_INTR_DB_ENA);
+		spx5_wr(db, sparx5, FDMA_INTR_DB);
+		napi_schedule(&sparx5->rx.napi);
+	}
+	if (err) {
+		u32 err_type = spx5_rd(sparx5, FDMA_ERRORS);
+
+		dev_err_ratelimited(sparx5->dev,
+				    "ERR: int: %#x, type: %#x\n",
+				    err, err_type);
+		spx5_wr(err, sparx5, FDMA_INTR_ERR);
+		spx5_wr(err_type, sparx5, FDMA_ERRORS);
+	}
+	return IRQ_HANDLED;
+}
+
+static void sparx5_fdma_injection_mode(struct sparx5 *sparx5)
+{
+	const int byte_swap = 1;
+	int portno;
+	int urgency;
+
+	/* Change mode to fdma extraction and injection */
+	spx5_wr(QS_XTR_GRP_CFG_MODE_SET(2) |
+		QS_XTR_GRP_CFG_STATUS_WORD_POS_SET(1) |
+		QS_XTR_GRP_CFG_BYTE_SWAP_SET(byte_swap),
+		sparx5, QS_XTR_GRP_CFG(XTR_QUEUE));
+	spx5_wr(QS_INJ_GRP_CFG_MODE_SET(2) |
+		QS_INJ_GRP_CFG_BYTE_SWAP_SET(byte_swap),
+		sparx5, QS_INJ_GRP_CFG(INJ_QUEUE));
+
+	/* CPU ports capture setup */
+	for (portno = SPX5_PORT_CPU_0; portno <= SPX5_PORT_CPU_1; portno++) {
+		/* ASM CPU port: No preamble, IFH, enable padding */
+		spx5_wr(ASM_PORT_CFG_PAD_ENA_SET(1) |
+			ASM_PORT_CFG_NO_PREAMBLE_ENA_SET(1) |
+			ASM_PORT_CFG_INJ_FORMAT_CFG_SET(1), /* 1 = IFH */
+			sparx5, ASM_PORT_CFG(portno));
+
+		/* Reset WM cnt to unclog queued frames */
+		spx5_rmw(DSM_DEV_TX_STOP_WM_CFG_DEV_TX_CNT_CLR_SET(1),
+			 DSM_DEV_TX_STOP_WM_CFG_DEV_TX_CNT_CLR,
+			 sparx5,
+			 DSM_DEV_TX_STOP_WM_CFG(portno));
+
+		/* Set Disassembler Stop Watermark level */
+		spx5_rmw(DSM_DEV_TX_STOP_WM_CFG_DEV_TX_STOP_WM_SET(100),
+			 DSM_DEV_TX_STOP_WM_CFG_DEV_TX_STOP_WM,
+			 sparx5,
+			 DSM_DEV_TX_STOP_WM_CFG(portno));
+
+		/* Enable port in queue system */
+		urgency = sparx5_port_fwd_urg(sparx5, SPEED_2500);
+		spx5_rmw(QFWD_SWITCH_PORT_MODE_PORT_ENA_SET(1) |
+			 QFWD_SWITCH_PORT_MODE_FWD_URGENCY_SET(urgency),
+			 QFWD_SWITCH_PORT_MODE_PORT_ENA |
+			 QFWD_SWITCH_PORT_MODE_FWD_URGENCY,
+			 sparx5,
+			 QFWD_SWITCH_PORT_MODE(portno));
+
+		/* Disable Disassembler buffer underrun watchdog
+		 * to avoid truncated packets in XTR
+		 */
+		spx5_rmw(DSM_BUF_CFG_UNDERFLOW_WATCHDOG_DIS_SET(1),
+			 DSM_BUF_CFG_UNDERFLOW_WATCHDOG_DIS,
+			 sparx5,
+			 DSM_BUF_CFG(portno));
+
+		/* Disabling frame aging */
+		spx5_rmw(HSCH_PORT_MODE_AGE_DIS_SET(1),
+			 HSCH_PORT_MODE_AGE_DIS,
+			 sparx5,
+			 HSCH_PORT_MODE(portno));
+	}
+}
+
+int sparx5_fdma_start(struct sparx5 *sparx5)
+{
+	int err;
+
+	/* Reset FDMA state */
+	spx5_wr(FDMA_CTRL_NRESET_SET(0), sparx5, FDMA_CTRL);
+	spx5_wr(FDMA_CTRL_NRESET_SET(1), sparx5, FDMA_CTRL);
+
+	/* Force ACP caching but disable read/write allocation */
+	spx5_rmw(CPU_PROC_CTRL_ACP_CACHE_FORCE_ENA_SET(1) |
+		 CPU_PROC_CTRL_ACP_AWCACHE_SET(0) |
+		 CPU_PROC_CTRL_ACP_ARCACHE_SET(0),
+		 CPU_PROC_CTRL_ACP_CACHE_FORCE_ENA |
+		 CPU_PROC_CTRL_ACP_AWCACHE |
+		 CPU_PROC_CTRL_ACP_ARCACHE,
+		 sparx5, CPU_PROC_CTRL);
+
+	sparx5_fdma_injection_mode(sparx5);
+	sparx5_fdma_rx_init(sparx5, &sparx5->rx, FDMA_XTR_CHANNEL);
+	sparx5_fdma_tx_init(sparx5, &sparx5->tx, FDMA_INJ_CHANNEL);
+	err = sparx5_fdma_rx_alloc(sparx5);
+	if (err) {
+		dev_err(sparx5->dev, "Could not allocate RX buffers: %d\n", err);
+		return err;
+	}
+	err = sparx5_fdma_tx_alloc(sparx5);
+	if (err) {
+		dev_err(sparx5->dev, "Could not allocate TX buffers: %d\n", err);
+		return err;
+	}
+	return err;
+}
+
+static u32 sparx5_fdma_port_ctrl(struct sparx5 *sparx5)
+{
+	return spx5_rd(sparx5, FDMA_PORT_CTRL(0));
+}
+
+int sparx5_fdma_stop(struct sparx5 *sparx5)
+{
+	u32 val;
+
+	napi_disable(&sparx5->rx.napi);
+	/* Stop the fdma and channel interrupts */
+	sparx5_fdma_rx_deactivate(sparx5, &sparx5->rx);
+	sparx5_fdma_tx_deactivate(sparx5, &sparx5->tx);
+	/* Wait for the RX channel to stop */
+	read_poll_timeout(sparx5_fdma_port_ctrl, val,
+			  FDMA_PORT_CTRL_XTR_BUF_IS_EMPTY_GET(val) == 0,
+			  500, 10000, 0, sparx5);
+	return 0;
+}
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_main.c b/drivers/net/ethernet/microchip/sparx5/sparx5_main.c
index f666133a15de..cbece6e9bff2 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_main.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_main.c
@@ -640,8 +640,23 @@ static int sparx5_start(struct sparx5 *sparx5)
 	sparx5_board_init(sparx5);
 	err = sparx5_register_notifier_blocks(sparx5);
 
-	/* Start register based INJ/XTR */
+	/* Start Frame DMA with fallback to register based INJ/XTR */
 	err = -ENXIO;
+	if (sparx5->fdma_irq >= 0) {
+		if (GCB_CHIP_ID_REV_ID_GET(sparx5->chip_id) > 0)
+			err = devm_request_threaded_irq(sparx5->dev,
+							sparx5->fdma_irq,
+							NULL,
+							sparx5_fdma_handler,
+							IRQF_ONESHOT,
+							"sparx5-fdma", sparx5);
+		if (!err)
+			err = sparx5_fdma_start(sparx5);
+		if (err)
+			sparx5->fdma_irq = -ENXIO;
+	} else {
+		sparx5->fdma_irq = -ENXIO;
+	}
 	if (err && sparx5->xtr_irq >= 0) {
 		err = devm_request_irq(sparx5->dev, sparx5->xtr_irq,
 				       sparx5_xtr_handler, IRQF_SHARED,
@@ -766,6 +781,7 @@ static int mchp_sparx5_probe(struct platform_device *pdev)
 		sparx5->base_mac[5] = 0;
 	}
 
+	sparx5->fdma_irq = platform_get_irq_byname(sparx5->pdev, "fdma");
 	sparx5->xtr_irq = platform_get_irq_byname(sparx5->pdev, "xtr");
 
 	/* Read chip ID to check CPU interface */
@@ -824,6 +840,11 @@ static int mchp_sparx5_remove(struct platform_device *pdev)
 		disable_irq(sparx5->xtr_irq);
 		sparx5->xtr_irq = -ENXIO;
 	}
+	if (sparx5->fdma_irq) {
+		disable_irq(sparx5->fdma_irq);
+		sparx5->fdma_irq = -ENXIO;
+	}
+	sparx5_fdma_stop(sparx5);
 	sparx5_cleanup_ports(sparx5);
 	/* Unregister netdevs */
 	sparx5_unregister_notifier_blocks(sparx5);
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
index 4d5f44c3a421..a1acc9b461f2 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
@@ -73,8 +73,61 @@ enum sparx5_vlan_port_type {
 #define XTR_QUEUE     0
 #define INJ_QUEUE     0
 
+#define FDMA_DCB_MAX			64
+#define FDMA_RX_DCB_MAX_DBS		15
+#define FDMA_TX_DCB_MAX_DBS		1
+
 struct sparx5;
 
+struct sparx5_db_hw {
+	u64 dataptr;
+	u64 status;
+};
+
+struct sparx5_rx_dcb_hw {
+	u64 nextptr;
+	u64 info;
+	struct sparx5_db_hw db[FDMA_RX_DCB_MAX_DBS];
+};
+
+struct sparx5_tx_dcb_hw {
+	u64 nextptr;
+	u64 info;
+	struct sparx5_db_hw db[FDMA_TX_DCB_MAX_DBS];
+};
+
+/* Frame DMA receive state:
+ * For each DB, there is a SKB, and the skb data pointer is mapped in
+ * the DB. Once a frame is received the skb is given to the upper layers
+ * and a new skb is added to the dcb.
+ * When the db_index reached FDMA_RX_DCB_MAX_DBS the DB is reused.
+ */
+struct sparx5_rx {
+	struct sparx5_rx_dcb_hw *dcb_entries;
+	struct sparx5_rx_dcb_hw *last_entry;
+	struct sk_buff *skb[FDMA_DCB_MAX][FDMA_RX_DCB_MAX_DBS];
+	int db_index;
+	int dcb_index;
+	dma_addr_t dma;
+	struct napi_struct napi;
+	u32 channel_id;
+	struct net_device *ndev;
+	u64 packets;
+};
+
+/* Frame DMA transmit state:
+ * DCBs are chained using the DCBs nextptr field.
+ */
+struct sparx5_tx {
+	struct sparx5_tx_dcb_hw *curr_entry;
+	struct sparx5_tx_dcb_hw *first_entry;
+	struct list_head db_list;
+	dma_addr_t dma;
+	u32 channel_id;
+	u64 packets;
+	u64 dropped;
+};
+
 struct sparx5_port_config {
 	phy_interface_t portmode;
 	u32 bandwidth;
@@ -167,6 +220,10 @@ struct sparx5 {
 	bool sd_sgpio_remapping;
 	/* Register based inj/xtr */
 	int xtr_irq;
+	/* Frame DMA */
+	int fdma_irq;
+	struct sparx5_rx rx;
+	struct sparx5_tx tx;
 };
 
 /* sparx5_switchdev.c */
@@ -174,11 +231,23 @@ int sparx5_register_notifier_blocks(struct sparx5 *sparx5);
 void sparx5_unregister_notifier_blocks(struct sparx5 *sparx5);
 
 /* sparx5_packet.c */
+struct frame_info {
+	int src_port;
+};
+
+void sparx5_xtr_flush(struct sparx5 *sparx5, u8 grp);
+void sparx5_ifh_parse(u32 *ifh, struct frame_info *info);
 irqreturn_t sparx5_xtr_handler(int irq, void *_priv);
 int sparx5_port_xmit_impl(struct sk_buff *skb, struct net_device *dev);
 int sparx5_manual_injection_mode(struct sparx5 *sparx5);
 void sparx5_port_inj_timer_setup(struct sparx5_port *port);
 
+/* sparx5_fdma.c */
+int sparx5_fdma_start(struct sparx5 *sparx5);
+int sparx5_fdma_stop(struct sparx5 *sparx5);
+int sparx5_fdma_xmit(struct sparx5 *sparx5, u32 *ifh, struct sk_buff *skb);
+irqreturn_t sparx5_fdma_handler(int irq, void *args);
+
 /* sparx5_mactable.c */
 void sparx5_mact_pull_work(struct work_struct *work);
 int sparx5_mact_learn(struct sparx5 *sparx5, int port,
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_packet.c b/drivers/net/ethernet/microchip/sparx5/sparx5_packet.c
index 09ca7a3bafdc..dc7e5ea6ec15 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_packet.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_packet.c
@@ -20,11 +20,7 @@
 
 #define INJ_TIMEOUT_NS 50000
 
-struct frame_info {
-	int src_port;
-};
-
-static void sparx5_xtr_flush(struct sparx5 *sparx5, u8 grp)
+void sparx5_xtr_flush(struct sparx5 *sparx5, u8 grp)
 {
 	/* Start flush */
 	spx5_wr(QS_XTR_FLUSH_FLUSH_SET(BIT(grp)), sparx5, QS_XTR_FLUSH);
@@ -36,7 +32,7 @@ static void sparx5_xtr_flush(struct sparx5 *sparx5, u8 grp)
 	spx5_wr(0, sparx5, QS_XTR_FLUSH);
 }
 
-static void sparx5_ifh_parse(u32 *ifh, struct frame_info *info)
+void sparx5_ifh_parse(u32 *ifh, struct frame_info *info)
 {
 	u8 *xtr_hdr = (u8 *)ifh;
 
@@ -224,7 +220,10 @@ int sparx5_port_xmit_impl(struct sk_buff *skb, struct net_device *dev)
 	struct sparx5 *sparx5 = port->sparx5;
 	int ret;
 
-	ret = sparx5_inject(sparx5, port->ifh, skb, dev);
+	if (sparx5->fdma_irq > 0)
+		ret = sparx5_fdma_xmit(sparx5, port->ifh, skb);
+	else
+		ret = sparx5_inject(sparx5, port->ifh, skb, dev);
 
 	if (ret == NETDEV_TX_OK) {
 		stats->tx_bytes += skb->len;
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_port.c b/drivers/net/ethernet/microchip/sparx5/sparx5_port.c
index d2e3250928bf..189a6a0a2e08 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_port.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_port.c
@@ -596,7 +596,7 @@ static int sparx5_port_max_tags_set(struct sparx5 *sparx5,
 	return 0;
 }
 
-static int sparx5_port_fwd_urg(struct sparx5 *sparx5, u32 speed)
+int sparx5_port_fwd_urg(struct sparx5 *sparx5, u32 speed)
 {
 	u32 clk_period_ps = 1600; /* 625Mhz for now */
 	u32 urg = 672000;
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_port.h b/drivers/net/ethernet/microchip/sparx5/sparx5_port.h
index fd05ab6436d1..2f8043eac71b 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_port.h
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_port.h
@@ -89,5 +89,6 @@ int sparx5_get_port_status(struct sparx5 *sparx5,
 			   struct sparx5_port_status *status);
 
 void sparx5_port_enable(struct sparx5_port *port, bool enable);
+int sparx5_port_fwd_urg(struct sparx5 *sparx5, u32 speed);
 
 #endif	/* __SPARX5_PORT_H__ */
-- 
2.32.0


  reply	other threads:[~2021-08-19  7:39 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-19  7:39 [PATCH net-next v2 0/2] Adding Frame DMA functionality to Sparx5 Steen Hegelund
2021-08-19  7:39 ` Steen Hegelund [this message]
2021-08-19  7:39 ` [PATCH net-next v2 2/2] arm64: dts: sparx5: Add the Sparx5 switch frame DMA support Steen Hegelund
2021-08-20 13:40 ` [PATCH net-next v2 0/2] Adding Frame DMA functionality to Sparx5 patchwork-bot+netdevbpf

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210819073940.1589383-2-steen.hegelund@microchip.com \
    --to=steen.hegelund@microchip.com \
    --cc=UNGLinuxDriver@microchip.com \
    --cc=alexandre.belloni@bootlin.com \
    --cc=davem@davemloft.net \
    --cc=kuba@kernel.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=netdev@vger.kernel.org \
    --subject='Re: [PATCH net-next v2 1/2] net: sparx5: switchdev: adding frame DMA functionality' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).