bnx2: Fix bug in bnx2_free_tx_skbs().
authorMichael Chan <mchan@broadcom.com>
Tue, 10 Jul 2012 10:04:40 +0000 (10:04 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 9 Aug 2012 15:27:51 +0000 (08:27 -0700)
[ Upstream commit c1f5163de417dab01fa9daaf09a74bbb19303f3c ]

In rare cases, bnx2x_free_tx_skbs() can unmap the wrong DMA address
when it gets to the last entry of the tx ring.  We were not using
the proper macro to skip the last entry when advancing the tx index.

Reported-by: Zongyun Lai <zlai@vmware.com>
Reviewed-by: Jeffrey Huang <huangjw@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/net/bnx2.c

index 74580bb175f1ec6e0b48e52365582c2521a3c064..c9b123c1608b4d6847c9d8eb9d01147b73db84f6 100644 (file)
@@ -5310,7 +5310,7 @@ bnx2_free_tx_skbs(struct bnx2 *bp)
                        int k, last;
 
                        if (skb == NULL) {
-                               j++;
+                               j = NEXT_TX_BD(j);
                                continue;
                        }
 
@@ -5322,8 +5322,8 @@ bnx2_free_tx_skbs(struct bnx2 *bp)
                        tx_buf->skb = NULL;
 
                        last = tx_buf->nr_frags;
-                       j++;
-                       for (k = 0; k < last; k++, j++) {
+                       j = NEXT_TX_BD(j);
+                       for (k = 0; k < last; k++, j = NEXT_TX_BD(j)) {
                                tx_buf = &txr->tx_buf_ring[TX_RING_IDX(j)];
                                dma_unmap_page(&bp->pdev->dev,
                                        dma_unmap_addr(tx_buf, mapping),