From 33cd6211052c5add833150b2eff1bc0729651d3a Mon Sep 17 00:00:00 2001 From: Wei Hu Date: Wed, 17 Jan 2024 09:19:35 +0000 Subject: [PATCH] mana: Fix TX CQE error handling For an unknown TX CQE error type (probably from a newer hardware), still free the mbuf, update the queue tail, etc., otherwise the accounting will be wrong. Also, TX errors can be triggered by injecting corrupted packets, so replace the mana_err to mana_dbg logging. Reported by: NetApp MFC after: 1 week Sponsored by: Microsoft (cherry picked from commit 516b5059705b6b8bbba28821dbe05964c128f9a9) --- sys/dev/mana/mana.h | 2 ++ sys/dev/mana/mana_en.c | 14 ++++++++------ sys/dev/mana/mana_sysctl.c | 6 ++++++ 3 files changed, 16 insertions(+), 6 deletions(-) diff --git a/sys/dev/mana/mana.h b/sys/dev/mana/mana.h index ab394f3203a..906b28eb56b 100644 --- a/sys/dev/mana/mana.h +++ b/sys/dev/mana/mana.h @@ -137,6 +137,8 @@ struct mana_stats { counter_u64_t mbuf_alloc_fail; /* rx */ counter_u64_t alt_chg; /* tx */ counter_u64_t alt_reset; /* tx */ + counter_u64_t cqe_err; /* tx */ + counter_u64_t cqe_unknown_type; /* tx */ }; struct mana_txq { diff --git a/sys/dev/mana/mana_en.c b/sys/dev/mana/mana_en.c index 812a42e51df..09de49a1101 100644 --- a/sys/dev/mana/mana_en.c +++ b/sys/dev/mana/mana_en.c @@ -1487,21 +1487,23 @@ mana_poll_tx_cq(struct mana_cq *cq) case CQE_TX_VPORT_DISABLED: case CQE_TX_VLAN_TAGGING_VIOLATION: sa_drop ++; - mana_err(NULL, + mana_dbg(NULL, "TX: txq %d CQE error %d, ntc = %d, " "pending sends = %d: err ignored.\n", txq_idx, cqe_oob->cqe_hdr.cqe_type, next_to_complete, txq->pending_sends); + counter_u64_add(txq->stats.cqe_err, 1); break; default: - /* If the CQE type is unexpected, log an error, - * and go through the error path. + /* If the CQE type is unknown, log a debug msg, + * and still free the mbuf, etc. */ - mana_err(NULL, - "ERROR: TX: Unexpected CQE type %d: HW BUG?\n", + mana_dbg(NULL, + "ERROR: TX: Unknown CQE type %d\n", cqe_oob->cqe_hdr.cqe_type); - return; + counter_u64_add(txq->stats.cqe_unknown_type, 1); + break; } if (txq->gdma_txq_id != completions[i].wq_num) { mana_dbg(NULL, diff --git a/sys/dev/mana/mana_sysctl.c b/sys/dev/mana/mana_sysctl.c index d03ebe44ab3..844a0504059 100644 --- a/sys/dev/mana/mana_sysctl.c +++ b/sys/dev/mana/mana_sysctl.c @@ -292,6 +292,12 @@ mana_sysctl_add_queues(struct mana_port_context *apc) SYSCTL_ADD_COUNTER_U64(ctx, tx_list, OID_AUTO, "alt_reset", CTLFLAG_RD, &tx_stats->alt_reset, "Reset to self txq"); + SYSCTL_ADD_COUNTER_U64(ctx, tx_list, OID_AUTO, + "cqe_err", CTLFLAG_RD, + &tx_stats->cqe_err, "Error CQE count"); + SYSCTL_ADD_COUNTER_U64(ctx, tx_list, OID_AUTO, + "cqe_unknown_type", CTLFLAG_RD, + &tx_stats->cqe_unknown_type, "Unknown CQE count"); /* RX stats */ rx_node = SYSCTL_ADD_NODE(ctx, queue_list, OID_AUTO,