netdev - Re: [PATCH net, stable v1 3/3] virtio_net: add checking sq is full inside xdp xmit

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 8 Mar 2023 15:14:07 +0800
From:   Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
To:     Yunsheng Lin <linyunsheng@...wei.com>
Cc:     "Michael S. Tsirkin" <mst@...hat.com>,
        Jason Wang <jasowang@...hat.com>,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Jesper Dangaard Brouer <hawk@...nel.org>,
        John Fastabend <john.fastabend@...il.com>,
        <virtualization@...ts.linux-foundation.org>, <bpf@...r.kernel.org>,
        Yichun Zhang <yichun@...nresty.com>,
        Alexander Duyck <alexanderduyck@...com>,
        <netdev@...r.kernel.org>
Subject: Re: [PATCH net, stable v1 3/3] virtio_net: add checking sq is full inside xdp xmit

On Wed, 8 Mar 2023 14:59:36 +0800, Yunsheng Lin <linyunsheng@...wei.com> wrote:
> On 2023/3/8 10:49, Xuan Zhuo wrote:
> > If the queue of xdp xmit is not an independent queue, then when the xdp
> > xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> > the following error.
> >
> > net ens4: Unexpected TXQ (0) queue failure: -28
> >
> > This patch adds a check whether sq is full in xdp xmit.
> >
> > Fixes: 56434a01b12e ("virtio_net: add XDP_TX support")
> > Reported-by: Yichun Zhang <yichun@...nresty.com>
> > Signed-off-by: Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
> > Reviewed-by: Alexander Duyck <alexanderduyck@...com>
> > Acked-by: Michael S. Tsirkin <mst@...hat.com>
> > ---
> >  drivers/net/virtio_net.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 46bbddaadb0d..1a309cfb4976 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -767,6 +767,9 @@ static int virtnet_xdp_xmit(struct net_device *dev,
> >  	}
> >  	ret = nxmit;
> >
> > +	if (!is_xdp_raw_buffer_queue(vi, sq - vi->sq))
> > +		check_sq_full_and_disable(vi, dev, sq);
> > +
>
> Sorry if I missed something obvious here.
>
> As the comment in start_xmit(), the current skb is added to the sq->vq, so
> NETDEV_TX_BUSY can not be returned.
>
> 	/* If running out of space, stop queue to avoid getting packets that we
> 	 * are then unable to transmit.
> 	 * An alternative would be to force queuing layer to requeue the skb by
> 	 * returning NETDEV_TX_BUSY. However, NETDEV_TX_BUSY should not be
> 	 * returned in a normal path of operation: it means that driver is not
> 	 * maintaining the TX queue stop/start state properly, and causes
> 	 * the stack to do a non-trivial amount of useless work.
> 	 * Since most packets only take 1 or 2 ring slots, stopping the queue
> 	 * early means 16 slots are typically wasted.
> 	 */
>
> It there any reason not to check the sq->vq->num_free at the begin of start_xmit(),
> if the space is not enough for the current skb, TX queue is stopped and NETDEV_TX_BUSY
> is return to the stack to requeue the current skb.
>
> It seems it is the pattern that most network driver follow, and it seems we can avoid
> calling check_sq_full_and_disable() in this patch and not wasting 16 slots as mentioned
> in the comment above.
>



 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
 *                               struct net_device *dev);
 *	Called when a packet needs to be transmitted.
 *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
 *	the queue before that can happen; it's for obsolete devices and weird
 *	corner cases, but the stack really does a non-trivial amount
 *	of useless work if you return NETDEV_TX_BUSY.
 *	Required; cannot be NULL.

It does not affect the XDP TX. It is just that there are some waste on the
path of the protocol stack.

For example, TCP will do some unnecessary work based on the return value here.

Thanks.


>
> >  	if (flags & XDP_XMIT_FLUSH) {
> >  		if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq))
> >  			kicks = 1;
> >