[ofa-general] [PATCH] Corrects a race in ipoib_cm_post_receive_nonsrq()

David J. Wilder dwilder at us.ibm.com
Wed Jun 25 15:24:47 PDT 2008


This patch corrects a race condition in ipoib_cm_post_receive_nonsrq()
which allows wqes from one QP context to be post_recv
to another QP context. Thanks to Hong-Nam Nguyen for finding this bug

Signed-off-by: Pradeep Satyanarayana <pradeep at us.ibm.com>
Signed-off-by: David Wilder <dwilder at us.ibm.com>
---
 drivers/infiniband/ulp/ipoib/ipoib_cm.c |   18 ++++++++++++++----
 1 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index 97e67d3..8a1f92b 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -116,15 +116,25 @@ static int ipoib_cm_post_receive_nonsrq(struct net_device *dev,
 					struct ipoib_cm_rx *rx, int id)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
-	struct ib_recv_wr *bad_wr;
+	struct ib_recv_wr *bad_wr, rx_wr;
+	struct ib_sge	   rx_sge[IPOIB_CM_RX_SG];
 	int i, ret;
 
-	priv->cm.rx_wr.wr_id = id | IPOIB_OP_CM | IPOIB_OP_RECV;
+	rx_sge[0].length = IPOIB_CM_HEAD_SIZE;
+	rx_sge[0].lkey	 = priv->mr->lkey;
+	for (i = 1; i < IPOIB_CM_RX_SG; ++i) {
+		rx_sge[i].length = PAGE_SIZE;
+		rx_sge[i].lkey	 = priv->mr->lkey;
+	}
+	rx_wr.next 	= NULL;
+	rx_wr.sg_list	= rx_sge;
+	rx_wr.num_sge	= priv->cm.num_frags;
+	rx_wr.wr_id 	= id | IPOIB_OP_CM | IPOIB_OP_RECV;
 
 	for (i = 0; i < IPOIB_CM_RX_SG; ++i)
-		priv->cm.rx_sge[i].addr = rx->rx_ring[id].mapping[i];
+		rx_sge[i].addr = rx->rx_ring[id].mapping[i];
 
-	ret = ib_post_recv(rx->qp, &priv->cm.rx_wr, &bad_wr);
+	ret = ib_post_recv(rx->qp, &rx_wr, &bad_wr);
 	if (unlikely(ret)) {
 		ipoib_warn(priv, "post recv failed for buf %d (%d)\n", id, ret);
 		ipoib_cm_dma_unmap_rx(priv, IPOIB_CM_RX_SG - 1,





More information about the general mailing list