patch 'net/cxgbe: fix Tx queue stuck with mbuf chain coalescing' has been queued to stable release 21.11.2

Kevin Traynor ktraynor at redhat.com
Wed May 25 18:27:58 CEST 2022


Hi,

FYI, your patch has been queued to stable release 21.11.2

Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
It will be pushed if I get no objections before 05/30/22. So please
shout if anyone has objections.

Also note that after the patch there's a diff of the upstream commit vs the
patch applied to the branch. This will indicate if there was any rebasing
needed to apply to the stable branch. If there were code changes for rebasing
(ie: not only metadata diffs), please double check that the rebase was
correctly done.

Queued patches are on a temporary branch at:
https://github.com/kevintraynor/dpdk-stable

This queued commit can be viewed at:
https://github.com/kevintraynor/dpdk-stable/commit/24dabb9d25a99bde83a91335f0bc594fe1cbd2bf

Thanks.

Kevin

---
>From 24dabb9d25a99bde83a91335f0bc594fe1cbd2bf Mon Sep 17 00:00:00 2001
From: Rahul Lakkireddy <rahul.lakkireddy at chelsio.com>
Date: Tue, 19 Apr 2022 03:54:19 +0530
Subject: [PATCH] net/cxgbe: fix Tx queue stuck with mbuf chain coalescing

[ upstream commit 151e828f6427667faf3fdfaa00d14a65c7f57cd6 ]

When trying to coalesce mbufs with chain on Tx side, it is possible
to get stuck during queue wrap around. When coalescing this mbuf
chain fails, the Tx path returns EBUSY and when the same packet
is retried again, it couldn't get coalesced again, and the loop
repeats. Fix by pushing the packet through the normal Tx path.
Also use FW_ETH_TX_PKTS_WR to handle mbufs with chain for FW
to optimize.

Fixes: 6c2809628cd5 ("net/cxgbe: improve latency for slow traffic")

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy at chelsio.com>
---
 drivers/net/cxgbe/sge.c | 38 +++++++++++++++++++++++---------------
 1 file changed, 23 insertions(+), 15 deletions(-)

diff --git a/drivers/net/cxgbe/sge.c b/drivers/net/cxgbe/sge.c
index 5c176004f9..566cd48406 100644
--- a/drivers/net/cxgbe/sge.c
+++ b/drivers/net/cxgbe/sge.c
@@ -790,7 +790,7 @@ static inline void txq_advance(struct sge_txq *q, unsigned int n)
 #define MAX_COALESCE_LEN 64000
 
-static inline int wraps_around(struct sge_txq *q, int ndesc)
+static inline bool wraps_around(struct sge_txq *q, int ndesc)
 {
-	return (q->pidx + ndesc) > q->size ? 1 : 0;
+	return (q->pidx + ndesc) > q->size ? true : false;
 }
 
@@ -843,5 +843,4 @@ static inline void ship_tx_pkt_coalesce_wr(struct adapter *adap,
 	/* fill the pkts WR header */
 	wr = (void *)&q->desc[q->pidx];
-	wr->op_pkd = htonl(V_FW_WR_OP(FW_ETH_TX_PKTS2_WR));
 	vmwr = (void *)&q->desc[q->pidx];
 
@@ -853,6 +852,9 @@ static inline void ship_tx_pkt_coalesce_wr(struct adapter *adap,
 	wr->r3 = 0;
 	if (is_pf4(adap)) {
-		wr->op_pkd = htonl(V_FW_WR_OP(FW_ETH_TX_PKTS2_WR));
 		wr->type = q->coalesce.type;
+		if (likely(wr->type != 0))
+			wr->op_pkd = htonl(V_FW_WR_OP(FW_ETH_TX_PKTS2_WR));
+		else
+			wr->op_pkd = htonl(V_FW_WR_OP(FW_ETH_TX_PKTS_WR));
 	} else {
 		wr->op_pkd = htonl(V_FW_WR_OP(FW_ETH_TX_PKTS_VM_WR));
@@ -933,11 +935,14 @@ static inline int should_tx_packet_coalesce(struct sge_eth_txq *txq,
 		credits = txq_avail(q) - ndesc;
 
+		if (unlikely(wraps_around(q, ndesc)))
+			return 0;
+
 		/* If we are wrapping or this is last mbuf then, send the
 		 * already coalesced mbufs and let the non-coalesce pass
 		 * handle the mbuf.
 		 */
-		if (unlikely(credits < 0 || wraps_around(q, ndesc))) {
+		if (unlikely(credits < 0)) {
 			ship_tx_pkt_coalesce_wr(adap, txq);
-			return 0;
+			return -EBUSY;
 		}
 
@@ -963,6 +968,10 @@ new:
 	credits = txq_avail(q) - ndesc;
 
-	if (unlikely(credits < 0 || wraps_around(q, ndesc)))
+	if (unlikely(wraps_around(q, ndesc)))
 		return 0;
+
+	if (unlikely(credits < 0))
+		return -EBUSY;
+
 	q->coalesce.flits += wr_size / sizeof(__be64);
 	q->coalesce.type = type;
@@ -1107,5 +1116,5 @@ int t4_eth_xmit(struct sge_eth_txq *txq, struct rte_mbuf *mbuf,
 	int l3hdr_len, l4hdr_len, eth_xtra_len;
 	int len, last_desc;
-	int credits;
+	int should_coal, credits;
 	u32 wr_mid;
 	u64 cntrl, *end;
@@ -1139,7 +1148,7 @@ out_free:
 	txq->q.coalesce.max = (8 - (txq->q.pidx & 7)) * 8;
 
-	if (!((m->ol_flags & RTE_MBUF_F_TX_TCP_SEG) ||
-			m->pkt_len > RTE_ETHER_MAX_LEN)) {
-		if (should_tx_packet_coalesce(txq, mbuf, &cflits, adap)) {
+	if ((m->ol_flags & RTE_MBUF_F_TX_TCP_SEG) == 0) {
+		should_coal = should_tx_packet_coalesce(txq, mbuf, &cflits, adap);
+		if (should_coal > 0) {
 			if (unlikely(map_mbuf(mbuf, addr) < 0)) {
 				dev_warn(adap, "%s: mapping err for coalesce\n",
@@ -1150,6 +1159,6 @@ out_free:
 			return tx_do_packet_coalesce(txq, mbuf, cflits, adap,
 						     pi, addr, nb_pkts);
-		} else {
-			return -EBUSY;
+		} else if (should_coal < 0) {
+			return should_coal;
 		}
 	}
@@ -1198,6 +1207,5 @@ out_free:
 	}
 
-	len = 0;
-	len += sizeof(*cpl);
+	len = sizeof(*cpl);
 
 	/* Coalescing skipped and we send through normal path */
-- 
2.34.3

---
  Diff of the applied patch vs upstream commit (please double-check if non-empty:
---
--- -	2022-05-25 17:26:58.787074081 +0100
+++ 0006-net-cxgbe-fix-Tx-queue-stuck-with-mbuf-chain-coalesc.patch	2022-05-25 17:26:58.540828292 +0100
@@ -1 +1 @@
-From 151e828f6427667faf3fdfaa00d14a65c7f57cd6 Mon Sep 17 00:00:00 2001
+From 24dabb9d25a99bde83a91335f0bc594fe1cbd2bf Mon Sep 17 00:00:00 2001
@@ -5,0 +6,2 @@
+[ upstream commit 151e828f6427667faf3fdfaa00d14a65c7f57cd6 ]
+
@@ -15 +16,0 @@
-Cc: stable at dpdk.org



More information about the stable mailing list