[SRU][N:linux-bluefield][PATCH v1 1/1] IB/mad: Check available slots before posting receive WRs

[Stav Aviram]

From: Maher Sanalla <msanalla at nvidia.com>

BugLink: https://bugs.launchpad.net/bugs/2126876

The ib_post_receive_mads() function handles posting receive work
requests (WRs) to MAD QPs and is called in two cases:
1) When a MAD port is opened.
2) When a receive WQE is consumed upon receiving a new MAD.

Whereas, if MADs arrive during the port open phase, a race condition
might cause an extra WR to be posted, exceeding the QP’s capacity.
This leads to failures such as:
infiniband mlx5_0: ib_post_recv failed: -12
infiniband mlx5_0: Couldn't post receive WRs
infiniband mlx5_0: Couldn't start port
infiniband mlx5_0: Couldn't open port 1

Fix this by checking the current receive count before posting a new WR.
If the QP’s receive queue is full, do not post additional WRs.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Maher Sanalla <msanalla at nvidia.com>
Link: https://patch.msgid.link/c4984ba3c3a98a5711a558bccefcad789587ecf1.1741875592.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon at kernel.org>
(cherry picked from commit 37826f0a8c2f6b6add5179003b8597e32a445362)
Signed-off-by: Stav Aviram <saviram at nvidia.com>
---
 drivers/infiniband/core/mad.c | 38 ++++++++++++++++++-----------------
 1 file changed, 20 insertions(+), 18 deletions(-)

diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index 58befbaaf0ad..242434c09e8d 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -2671,11 +2671,11 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 				    struct ib_mad_private *mad)
 {
 	unsigned long flags;
-	int post, ret;
 	struct ib_mad_private *mad_priv;
 	struct ib_sge sg_list;
 	struct ib_recv_wr recv_wr;
 	struct ib_mad_queue *recv_queue = &qp_info->recv_queue;
+	int ret = 0;
 
 	/* Initialize common scatter list fields */
 	sg_list.lkey = qp_info->port_priv->pd->local_dma_lkey;
@@ -2685,7 +2685,7 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 	recv_wr.sg_list = &sg_list;
 	recv_wr.num_sge = 1;
 
-	do {
+	while (true) {
 		/* Allocate and map receive buffer */
 		if (mad) {
 			mad_priv = mad;
@@ -2693,10 +2693,8 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 		} else {
 			mad_priv = alloc_mad_private(port_mad_size(qp_info->port_priv),
 						     GFP_ATOMIC);
-			if (!mad_priv) {
-				ret = -ENOMEM;
-				break;
-			}
+			if (!mad_priv)
+				return -ENOMEM;
 		}
 		sg_list.length = mad_priv_dma_size(mad_priv);
 		sg_list.addr = ib_dma_map_single(qp_info->port_priv->device,
@@ -2705,37 +2703,41 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 						 DMA_FROM_DEVICE);
 		if (unlikely(ib_dma_mapping_error(qp_info->port_priv->device,
 						  sg_list.addr))) {
-			kfree(mad_priv);
 			ret = -ENOMEM;
-			break;
+			goto free_mad_priv;
 		}
 		mad_priv->header.mapping = sg_list.addr;
 		mad_priv->header.mad_list.mad_queue = recv_queue;
 		mad_priv->header.mad_list.cqe.done = ib_mad_recv_done;
 		recv_wr.wr_cqe = &mad_priv->header.mad_list.cqe;
-
-		/* Post receive WR */
 		spin_lock_irqsave(&recv_queue->lock, flags);
-		post = (++recv_queue->count < recv_queue->max_active);
-		list_add_tail(&mad_priv->header.mad_list.list, &recv_queue->list);
+		if (recv_queue->count >= recv_queue->max_active) {
+			/* Fully populated the receive queue */
+			spin_unlock_irqrestore(&recv_queue->lock, flags);
+			break;
+		}
+		recv_queue->count++;
+		list_add_tail(&mad_priv->header.mad_list.list,
+			      &recv_queue->list);
 		spin_unlock_irqrestore(&recv_queue->lock, flags);
+
 		ret = ib_post_recv(qp_info->qp, &recv_wr, NULL);
 		if (ret) {
 			spin_lock_irqsave(&recv_queue->lock, flags);
 			list_del(&mad_priv->header.mad_list.list);
 			recv_queue->count--;
 			spin_unlock_irqrestore(&recv_queue->lock, flags);
-			ib_dma_unmap_single(qp_info->port_priv->device,
-					    mad_priv->header.mapping,
-					    mad_priv_dma_size(mad_priv),
-					    DMA_FROM_DEVICE);
-			kfree(mad_priv);
 			dev_err(&qp_info->port_priv->device->dev,
 				"ib_post_recv failed: %d\n", ret);
 			break;
 		}
-	} while (post);
+	}
 
+	ib_dma_unmap_single(qp_info->port_priv->device,
+			    mad_priv->header.mapping,
+			    mad_priv_dma_size(mad_priv), DMA_FROM_DEVICE);
+free_mad_priv:
+	kfree(mad_priv);
 	return ret;
 }
 
-- 
2.38.1