Virtualization - Mar 2014 - [PATCH net] virtio-net: correct error handling of virtqueue_kick()

Current error handling of virtqueue_kick() was wrong in two places:
- The skb were freed immediately when virtqueue_kick() fail during
  xmit. This may lead double free since the skb was not detached from
  the virtqueue.
- try_fill_recv() returns false when virtqueue_kick() fail. This will
  lead unnecessary rescheduling of refill work.

Actually, it's safe to just ignore the kick failure in those two
places. So this patch fixes this by partially revert commit
67975901183799af8e93ec60e322f9e2a1940b9b.

Fixes 67975901183799af8e93ec60e322f9e2a1940b9b
(virtio_net: verify if virtqueue_kick() succeeded).

Cc: Heinz Graalfs <graalfs at linux.vnet.ibm.com>
Cc: Rusty Russell <rusty at rustcorp.com.au>
Cc: Michael S. Tsirkin <mst at redhat.com>
Signed-off-by: Jason Wang <jasowang at redhat.com>
---
 drivers/net/virtio_net.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 5632a99..841b608 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -671,8 +671,7 @@ static bool try_fill_recv(struct receive_queue *rq, gfp_t
gfp)
 		if (err)
 			break;
 	} while (rq->vq->num_free);
-	if (unlikely(!virtqueue_kick(rq->vq)))
-		return false;
+	virtqueue_kick(rq->vq);
 	return !oom;
 }
 
@@ -877,7 +876,7 @@ static netdev_tx_t start_xmit(struct sk_buff *skb, struct
net_device *dev)
 	err = xmit_skb(sq, skb);
 
 	/* This should not happen! */
-	if (unlikely(err) || unlikely(!virtqueue_kick(sq->vq))) {
+	if (unlikely(err)) {
 		dev->stats.tx_fifo_errors++;
 		if (net_ratelimit())
 			dev_warn(&dev->dev,
@@ -886,6 +885,7 @@ static netdev_tx_t start_xmit(struct sk_buff *skb, struct
net_device *dev)
 		kfree_skb(skb);
 		return NETDEV_TX_OK;
 	}
+	virtqueue_kick(sq->vq);
 
 	/* Don't wait up for transmitted skbs to be freed. */
 	skb_orphan(skb);
-- 
1.7.1

On Wed, Mar 26, 2014 at 01:03:00PM +0800, Jason Wang
wrote:
> Current error handling of virtqueue_kick() was wrong in two places:
> - The skb were freed immediately when virtqueue_kick() fail during
>   xmit. This may lead double free since the skb was not detached from
>   the virtqueue.
> - try_fill_recv() returns false when virtqueue_kick() fail. This will
>   lead unnecessary rescheduling of refill work.
> 
> Actually, it's safe to just ignore the kick failure in those two
> places. So this patch fixes this by partially revert commit
> 67975901183799af8e93ec60e322f9e2a1940b9b.
> 
> Fixes 67975901183799af8e93ec60e322f9e2a1940b9b
> (virtio_net: verify if virtqueue_kick() succeeded).
> 
> Cc: Heinz Graalfs <graalfs at linux.vnet.ibm.com>
> Cc: Rusty Russell <rusty at rustcorp.com.au>
> Cc: Michael S. Tsirkin <mst at redhat.com>
> Signed-off-by: Jason Wang <jasowang at redhat.com>
Acked-by: Michael S. Tsirkin <mst at redhat.com>
> ---
>  drivers/net/virtio_net.c |    6 +++---
>  1 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 5632a99..841b608 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -671,8 +671,7 @@ static bool try_fill_recv(struct receive_queue *rq,
gfp_t gfp)
>  		if (err)
>  			break;
>  	} while (rq->vq->num_free);
> -	if (unlikely(!virtqueue_kick(rq->vq)))
> -		return false;
> +	virtqueue_kick(rq->vq);
>  	return !oom;
>  }
>  
> @@ -877,7 +876,7 @@ static netdev_tx_t start_xmit(struct sk_buff *skb,
struct net_device *dev)
>  	err = xmit_skb(sq, skb);
>  
>  	/* This should not happen! */
> -	if (unlikely(err) || unlikely(!virtqueue_kick(sq->vq))) {
> +	if (unlikely(err)) {
>  		dev->stats.tx_fifo_errors++;
>  		if (net_ratelimit())
>  			dev_warn(&dev->dev,
> @@ -886,6 +885,7 @@ static netdev_tx_t start_xmit(struct sk_buff *skb,
struct net_device *dev)
>  		kfree_skb(skb);
>  		return NETDEV_TX_OK;
>  	}
> +	virtqueue_kick(sq->vq);
>  
>  	/* Don't wait up for transmitted skbs to be freed. */
>  	skb_orphan(skb);
> -- 
> 1.7.1

From: Jason Wang <jasowang at redhat.com>
Date: Wed, 26 Mar 2014 13:03:00 +0800
> Current error handling of virtqueue_kick() was wrong in two places:
> - The skb were freed immediately when virtqueue_kick() fail during
>   xmit. This may lead double free since the skb was not detached from
>   the virtqueue.
> - try_fill_recv() returns false when virtqueue_kick() fail. This will
>   lead unnecessary rescheduling of refill work.
> 
> Actually, it's safe to just ignore the kick failure in those two
> places. So this patch fixes this by partially revert commit
> 67975901183799af8e93ec60e322f9e2a1940b9b.
> 
> Fixes 67975901183799af8e93ec60e322f9e2a1940b9b
> (virtio_net: verify if virtqueue_kick() succeeded).
> 
> Cc: Heinz Graalfs <graalfs at linux.vnet.ibm.com>
> Cc: Rusty Russell <rusty at rustcorp.com.au>
> Cc: Michael S. Tsirkin <mst at redhat.com>
> Signed-off-by: Jason Wang <jasowang at redhat.com>
Applied and queued up for -stable, thanks.