[PATCH stable] virtio_net: don't leak memory or block when too many frags
From: "Michael S. Tsirkin" <mst@redhat.com>
Date: 2013-12-25 20:09:34
Also in:
lkml, netdev
Subsystem:
networking drivers, the rest, virtio net driver · Maintainers:
Andrew Lunn, "David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, Linus Torvalds, "Michael S. Tsirkin", Jason Wang, Eugenio Pérez
We leak an skb when there are too many frags, we also stop processing the packet in the middle, the result is almost sure to be loss of networking. Reported-by: Michael Dalton <redacted> Signed-off-by: Michael S. Tsirkin <mst@redhat.com> --- Note: this path is gone upstream, so this patch is for stable kernel only. This is on top of series: virtio-net: backport error handling bugfix that I sent previously (and that this is in reply to). drivers/net/virtio_net.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index c0ed6d5..86fe1e7 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c@@ -344,7 +344,7 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, if (i >= MAX_SKB_FRAGS) { pr_debug("%s: packet too long\n", skb->dev->name); skb->dev->stats.rx_length_errors++; - return NULL; + goto err_frags; } page = virtqueue_get_buf(rq->vq, &len); if (!page) {
@@ -364,6 +364,7 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, return skb; err_skb: give_pages(rq, page); while (--num_buf) { +err_frags: buf = virtqueue_get_buf(rq->vq, &len); if (unlikely(!buf)) {
--
MST