RE: [PATCH] virtio_net: Use NETDEV_TX_BUSY when has no buf to send

[Angus Chen]

Hi 

> -----Original Message-----
> From: Yunsheng Lin <linyunsheng@xxxxxxxxxx>
> Sent: Thursday, March 9, 2023 6:53 PM
> To: Angus Chen <angus.chen@xxxxxxxxxxxxxxx>; mst@xxxxxxxxxx;
> jasowang@xxxxxxxxxx; davem@xxxxxxxxxxxxx; edumazet@xxxxxxxxxx;
> kuba@xxxxxxxxxx; pabeni@xxxxxxxxxx; ast@xxxxxxxxxx; daniel@xxxxxxxxxxxxx;
> hawk@xxxxxxxxxx; john.fastabend@xxxxxxxxx
> Cc: virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> linux-kernel@xxxxxxxxxxxxxxx; bpf@xxxxxxxxxxxxxxx; Xuan Zhuo
> <xuanzhuo@xxxxxxxxxxxxxxxxx>
> Subject: Re: [PATCH] virtio_net: Use NETDEV_TX_BUSY when has no buf to send
> 
> On 2023/3/9 15:49, Angus Chen wrote:
> > Don't consume skb if virtqueue_add return -ENOSPC.
> 
> Is this fixing the same out of space problem caused by
> xdp xmit as Xuan Zhuo is fixing, or it is fixing a different
> case?
Yes, the same print info.
> 
> https://lore.kernel.org/netdev/20230308071921-mutt-send-email-mst@kernel.
> org/T/#mc4c5766a59fb8be988bb6a4dfa48f49e58df3ea6
> 
En,I haven't noticed this patch before.
> >
> > Signed-off-by: Angus Chen <angus.chen@xxxxxxxxxxxxxxx>
> > ---
> >  drivers/net/virtio_net.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index fb5e68ed3ec2..4096ea3d2eb6 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -1980,7 +1980,7 @@ static netdev_tx_t start_xmit(struct sk_buff *skb,
> struct net_device *dev)
> >  				 qnum, err);
> >  		dev->stats.tx_dropped++;
> >  		dev_kfree_skb_any(skb);
> 
> Returning NETDEV_TX_BUSY will caused stack to requeue the skb
> and send it again when space is available, but you have freed the skb here,
> isn't this cause use-after-free problem?
Yes，you are right, would be crash after I tested more times .I will drop it,thank you.
> 
> > -		return NETDEV_TX_OK;
> > +		return (err == -ENOSPC) ? NETDEV_TX_BUSY : NETDEV_TX_OK;
> >  	}
> >
> >  	/* Don't wait up for transmitted skbs to be freed. */
> >