Blob Blame History Raw
From 015ba6091e225914db8eeb9774fe3ddea1482f5c Mon Sep 17 00:00:00 2001
From: Edwin Peer <edwin.peer@broadcom.com>
Date: Sun, 29 Aug 2021 03:35:00 -0400
Subject: [PATCH 16/22] bnxt_en: discard out of sequence HWRM responses
Patch-mainline: v5.14
Git-commit: 02b9aa1068682319508f9a1678e993ad958a8a4f
References: jsc#SLE-19704

During firmware crash recovery, it is possible for firmware to respond
to stale HWRM commands that have already timed out. Because response
buffers may be reused, any out of sequence responses need to be ignored
and only the matching seq_id should be accepted.

Also, READ_ONCE should be used for the reads from the DMA buffer to
ensure that the necessary loads are scheduled.

Reviewed-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Denis Kirjanov <denis.kirjanov@suse.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt_hwrm.c | 21 +++++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_hwrm.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_hwrm.c
index a45f2a619086..621daf687a00 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_hwrm.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_hwrm.c
@@ -399,9 +399,10 @@ static int __hwrm_send(struct bnxt *bp, struct bnxt_hwrm_ctx *ctx)
 					   le16_to_cpu(ctx->req->req_type));
 			goto exit;
 		}
-		len = le16_to_cpu(ctx->resp->resp_len);
+		len = le16_to_cpu(READ_ONCE(ctx->resp->resp_len));
 		valid = ((u8 *)ctx->resp) + len - 1;
 	} else {
+		__le16 seen_out_of_seq = ctx->req->seq_id; /* will never see */
 		int j;
 
 		/* Check if response len is updated */
@@ -411,9 +412,21 @@ static int __hwrm_send(struct bnxt *bp, struct bnxt_hwrm_ctx *ctx)
 			 */
 			if (test_bit(BNXT_STATE_FW_FATAL_COND, &bp->state))
 				goto exit;
-			len = le16_to_cpu(ctx->resp->resp_len);
-			if (len)
-				break;
+			len = le16_to_cpu(READ_ONCE(ctx->resp->resp_len));
+			if (len) {
+				__le16 resp_seq = READ_ONCE(ctx->resp->seq_id);
+
+				if (resp_seq == ctx->req->seq_id)
+					break;
+				if (resp_seq != seen_out_of_seq) {
+					netdev_warn(bp->dev, "Discarding out of seq response: 0x%x for msg {0x%x 0x%x}\n",
+						    le16_to_cpu(resp_seq),
+						    le16_to_cpu(ctx->req->req_type),
+						    le16_to_cpu(ctx->req->seq_id));
+					seen_out_of_seq = resp_seq;
+				}
+			}
+
 			/* on first few passes, just barely sleep */
 			if (i < HWRM_SHORT_TIMEOUT_COUNTER) {
 				usleep_range(HWRM_SHORT_MIN_TIMEOUT,
-- 
2.16.4