LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH] octeontx2-af: Wait for TX link idle for credits change
@ 2021-08-25  5:46 Geetha sowjanya
  2021-08-25 11:10 ` patchwork-bot+netdevbpf
  0 siblings, 1 reply; 2+ messages in thread
From: Geetha sowjanya @ 2021-08-25  5:46 UTC (permalink / raw)
  To: netdev, linux-kernel
  Cc: kuba, davem, sgoutham, lcherian, gakula, jerinj, sbhatta, hkelam,
	ndabilpuram

From: Nithin Dabilpuram <ndabilpuram@marvell.com>

NIX_AF_TX_LINKX_NORM_CREDIT holds running counter of
tx credits available per link. But, tx credits should be
configured based on MTU config. So MTU change needs tx
credit count update.

An issue exists whereby when both PF & VF are enabled and
PF traffic is flowing, if VF requests for MTU update,
updating the NORM_CREDIT register will lead to corruption
of credit count and subsequent deadlock of tx link as
the NORM_CREDIT register holds running count.

This patch provides workaround by pausing link traffic
using NIX_AF_TL1X_SW_XOFF, waiting for existing packets to
drain, and used credits be returned before updating new
credit count.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
---
 .../net/ethernet/marvell/octeontx2/af/rvu.h   |   1 +
 .../ethernet/marvell/octeontx2/af/rvu_nix.c   | 102 ++++++++++++++++--
 2 files changed, 92 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
index d88f595e63b0..d86c6b366547 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
@@ -314,6 +314,7 @@ struct nix_hw {
 	struct nix_lso lso;
 	struct nix_txvlan txvlan;
 	struct nix_ipolicer *ipolicer;
+	u64    *tx_credits;
 };
 
 /* RVU block's capabilities or functionality,
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
index c5e3f90e562d..a5c8067c8aef 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
@@ -3457,6 +3457,77 @@ static void nix_find_link_frs(struct rvu *rvu,
 		req->minlen = minlen;
 }
 
+static int
+nix_config_link_credits(struct rvu *rvu, int blkaddr, int link,
+			u16 pcifunc, u64 tx_credits)
+{
+	struct rvu_hwinfo *hw = rvu->hw;
+	int pf = rvu_get_pf(pcifunc);
+	u8 cgx_id = 0, lmac_id = 0;
+	unsigned long poll_tmo;
+	bool restore_tx_en = 0;
+	struct nix_hw *nix_hw;
+	u64 cfg, sw_xoff = 0;
+	u32 schq = 0;
+	u32 credits;
+	int rc;
+
+	nix_hw = get_nix_hw(rvu->hw, blkaddr);
+	if (!nix_hw)
+		return NIX_AF_ERR_INVALID_NIXBLK;
+
+	if (tx_credits == nix_hw->tx_credits[link])
+		return 0;
+
+	/* Enable cgx tx if disabled for credits to be back */
+	if (is_pf_cgxmapped(rvu, pf)) {
+		rvu_get_cgx_lmac_id(rvu->pf2cgxlmac_map[pf], &cgx_id, &lmac_id);
+		restore_tx_en = !cgx_lmac_tx_enable(rvu_cgx_pdata(cgx_id, rvu),
+						    lmac_id, true);
+	}
+
+	mutex_lock(&rvu->rsrc_lock);
+	/* Disable new traffic to link */
+	if (hw->cap.nix_shaping) {
+		schq = nix_get_tx_link(rvu, pcifunc);
+		sw_xoff = rvu_read64(rvu, blkaddr, NIX_AF_TL1X_SW_XOFF(schq));
+		rvu_write64(rvu, blkaddr,
+			    NIX_AF_TL1X_SW_XOFF(schq), BIT_ULL(0));
+	}
+
+	rc = -EBUSY;
+	poll_tmo = jiffies + usecs_to_jiffies(10000);
+	/* Wait for credits to return */
+	do {
+		if (time_after(jiffies, poll_tmo))
+			goto exit;
+		usleep_range(100, 200);
+
+		cfg = rvu_read64(rvu, blkaddr,
+				 NIX_AF_TX_LINKX_NORM_CREDIT(link));
+		credits = (cfg >> 12) & 0xFFFFFULL;
+	} while (credits != nix_hw->tx_credits[link]);
+
+	cfg &= ~(0xFFFFFULL << 12);
+	cfg |= (tx_credits << 12);
+	rvu_write64(rvu, blkaddr, NIX_AF_TX_LINKX_NORM_CREDIT(link), cfg);
+	rc = 0;
+
+	nix_hw->tx_credits[link] = tx_credits;
+
+exit:
+	/* Enable traffic back */
+	if (hw->cap.nix_shaping && !sw_xoff)
+		rvu_write64(rvu, blkaddr, NIX_AF_TL1X_SW_XOFF(schq), 0);
+
+	/* Restore state of cgx tx */
+	if (restore_tx_en)
+		cgx_lmac_tx_enable(rvu_cgx_pdata(cgx_id, rvu), lmac_id, false);
+
+	mutex_unlock(&rvu->rsrc_lock);
+	return rc;
+}
+
 int rvu_mbox_handler_nix_set_hw_frs(struct rvu *rvu, struct nix_frs_cfg *req,
 				    struct msg_rsp *rsp)
 {
@@ -3545,11 +3616,8 @@ int rvu_mbox_handler_nix_set_hw_frs(struct rvu *rvu, struct nix_frs_cfg *req,
 	lmac_fifo_len =
 		rvu_cgx_get_fifolen(rvu) /
 		cgx_get_lmac_cnt(rvu_cgx_pdata(cgx, rvu));
-	cfg = rvu_read64(rvu, blkaddr, NIX_AF_TX_LINKX_NORM_CREDIT(link));
-	cfg &= ~(0xFFFFFULL << 12);
-	cfg |=  ((lmac_fifo_len - req->maxlen) / 16) << 12;
-	rvu_write64(rvu, blkaddr, NIX_AF_TX_LINKX_NORM_CREDIT(link), cfg);
-	return 0;
+	return nix_config_link_credits(rvu, blkaddr, link, pcifunc,
+				       (lmac_fifo_len - req->maxlen) / 16);
 }
 
 int rvu_mbox_handler_nix_set_rx_cfg(struct rvu *rvu, struct nix_rx_cfg *req,
@@ -3593,12 +3661,13 @@ static u64 rvu_get_lbk_link_credits(struct rvu *rvu, u16 lbk_max_frs)
 	return 1600; /* 16 * max LBK datarate = 16 * 100Gbps */
 }
 
-static void nix_link_config(struct rvu *rvu, int blkaddr)
+static void nix_link_config(struct rvu *rvu, int blkaddr,
+			    struct nix_hw *nix_hw)
 {
 	struct rvu_hwinfo *hw = rvu->hw;
 	int cgx, lmac_cnt, slink, link;
 	u16 lbk_max_frs, lmac_max_frs;
-	u64 tx_credits;
+	u64 tx_credits, cfg;
 
 	rvu_get_lbk_link_max_frs(rvu, &lbk_max_frs);
 	rvu_get_lmac_link_max_frs(rvu, &lmac_max_frs);
@@ -3629,15 +3698,18 @@ static void nix_link_config(struct rvu *rvu, int blkaddr)
 	 */
 	for (cgx = 0; cgx < hw->cgx; cgx++) {
 		lmac_cnt = cgx_get_lmac_cnt(rvu_cgx_pdata(cgx, rvu));
+		/* Skip when cgx is not available or lmac cnt is zero */
+		if (lmac_cnt <= 0)
+			continue;
 		tx_credits = ((rvu_cgx_get_fifolen(rvu) / lmac_cnt) -
 			       lmac_max_frs) / 16;
 		/* Enable credits and set credit pkt count to max allowed */
-		tx_credits =  (tx_credits << 12) | (0x1FF << 2) | BIT_ULL(1);
+		cfg =  (tx_credits << 12) | (0x1FF << 2) | BIT_ULL(1);
 		slink = cgx * hw->lmac_per_cgx;
 		for (link = slink; link < (slink + lmac_cnt); link++) {
+			nix_hw->tx_credits[link] = tx_credits;
 			rvu_write64(rvu, blkaddr,
-				    NIX_AF_TX_LINKX_NORM_CREDIT(link),
-				    tx_credits);
+				    NIX_AF_TX_LINKX_NORM_CREDIT(link), cfg);
 		}
 	}
 
@@ -3645,6 +3717,7 @@ static void nix_link_config(struct rvu *rvu, int blkaddr)
 	slink = hw->cgx_links;
 	for (link = slink; link < (slink + hw->lbk_links); link++) {
 		tx_credits = rvu_get_lbk_link_credits(rvu, lbk_max_frs);
+		nix_hw->tx_credits[link] = tx_credits;
 		/* Enable credits and set credit pkt count to max allowed */
 		tx_credits =  (tx_credits << 12) | (0x1FF << 2) | BIT_ULL(1);
 		rvu_write64(rvu, blkaddr,
@@ -3908,8 +3981,13 @@ static int rvu_nix_block_init(struct rvu *rvu, struct nix_hw *nix_hw)
 		if (err)
 			return err;
 
+		nix_hw->tx_credits = kcalloc(hw->cgx_links + hw->lbk_links,
+					     sizeof(u64), GFP_KERNEL);
+		if (!nix_hw->tx_credits)
+			return -ENOMEM;
+
 		/* Initialize CGX/LBK/SDP link credits, min/max pkt lengths */
-		nix_link_config(rvu, blkaddr);
+		nix_link_config(rvu, blkaddr, nix_hw);
 
 		/* Enable Channel backpressure */
 		rvu_write64(rvu, blkaddr, NIX_AF_RX_CFG, BIT_ULL(0));
@@ -3965,6 +4043,8 @@ static void rvu_nix_block_freemem(struct rvu *rvu, int blkaddr,
 			kfree(txsch->schq.bmap);
 		}
 
+		kfree(nix_hw->tx_credits);
+
 		nix_ipolicer_freemem(rvu, nix_hw);
 
 		vlan = &nix_hw->txvlan;
-- 
2.17.1


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] octeontx2-af: Wait for TX link idle for credits change
  2021-08-25  5:46 [PATCH] octeontx2-af: Wait for TX link idle for credits change Geetha sowjanya
@ 2021-08-25 11:10 ` patchwork-bot+netdevbpf
  0 siblings, 0 replies; 2+ messages in thread
From: patchwork-bot+netdevbpf @ 2021-08-25 11:10 UTC (permalink / raw)
  To: Geetha sowjanya
  Cc: netdev, linux-kernel, kuba, davem, sgoutham, lcherian, jerinj,
	sbhatta, hkelam, ndabilpuram

Hello:

This patch was applied to netdev/net-next.git (refs/heads/master):

On Wed, 25 Aug 2021 11:16:21 +0530 you wrote:
> From: Nithin Dabilpuram <ndabilpuram@marvell.com>
> 
> NIX_AF_TX_LINKX_NORM_CREDIT holds running counter of
> tx credits available per link. But, tx credits should be
> configured based on MTU config. So MTU change needs tx
> credit count update.
> 
> [...]

Here is the summary with links:
  - octeontx2-af: Wait for TX link idle for credits change
    https://git.kernel.org/netdev/net-next/c/1c74b89171c3

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html



^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-08-25 11:10 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-08-25  5:46 [PATCH] octeontx2-af: Wait for TX link idle for credits change Geetha sowjanya
2021-08-25 11:10 ` patchwork-bot+netdevbpf

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).