[RFC,v2,3/4] virtio_net: move tx vq operation under tx queue lock

It's unsafe to operate a vq from multiple threads.
Unfortunately this is exactly what we do when invoking
clean tx poll from rx napi.
As a fix move everything that deals with the vq to under tx lock.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
 drivers/net/virtio_net.c | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)

On Tue, Apr 13, 2021 at 04:54:42PM +0800, Jason Wang wrote:
> 
> 在 2021/4/13 下午1:47, Michael S. Tsirkin 写道:
> > It's unsafe to operate a vq from multiple threads.
> > Unfortunately this is exactly what we do when invoking
> > clean tx poll from rx napi.
> > As a fix move everything that deals with the vq to under tx lock.
> > 
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > ---
> >   drivers/net/virtio_net.c | 22 +++++++++++++++++++++-
> >   1 file changed, 21 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 16d5abed582c..460ccdbb840e 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -1505,6 +1505,8 @@ static int virtnet_poll_tx(struct napi_struct *napi, int budget)
> >   	struct virtnet_info *vi = sq->vq->vdev->priv;
> >   	unsigned int index = vq2txq(sq->vq);
> >   	struct netdev_queue *txq;
> > +	int opaque;
> > +	bool done;
> >   	if (unlikely(is_xdp_raw_buffer_queue(vi, index))) {
> >   		/* We don't need to enable cb for XDP */
> > @@ -1514,10 +1516,28 @@ static int virtnet_poll_tx(struct napi_struct *napi, int budget)
> >   	txq = netdev_get_tx_queue(vi->dev, index);
> >   	__netif_tx_lock(txq, raw_smp_processor_id());
> > +	virtqueue_disable_cb(sq->vq);
> >   	free_old_xmit_skbs(sq, true);
> > +
> > +	opaque = virtqueue_enable_cb_prepare(sq->vq);
> > +
> > +	done = napi_complete_done(napi, 0);
> > +
> > +	if (!done)
> > +		virtqueue_disable_cb(sq->vq);
> > +
> >   	__netif_tx_unlock(txq);
> > -	virtqueue_napi_complete(napi, sq->vq, 0);
> 
> 
> So I wonder why not simply move __netif_tx_unlock() after
> virtqueue_napi_complete()?
> 
> Thanks
> 


Because that calls tx poll which also takes tx lock internally ...


> > +	if (done) {
> > +		if (unlikely(virtqueue_poll(sq->vq, opaque))) {
> > +			if (napi_schedule_prep(napi)) {
> > +				__netif_tx_lock(txq, raw_smp_processor_id());
> > +				virtqueue_disable_cb(sq->vq);
> > +				__netif_tx_unlock(txq);
> > +				__napi_schedule(napi);
> > +			}
> > +		}
> > +	}
> >   	if (sq->vq->num_free >= 2 + MAX_SKB_FRAGS)
> >   		netif_tx_wake_queue(txq);

On Tue, Apr 13, 2021 at 10:03 AM Michael S. Tsirkin <mst@redhat.com> wrote:
>
> On Tue, Apr 13, 2021 at 04:54:42PM +0800, Jason Wang wrote:
> >
> > 在 2021/4/13 下午1:47, Michael S. Tsirkin 写道:
> > > It's unsafe to operate a vq from multiple threads.
> > > Unfortunately this is exactly what we do when invoking
> > > clean tx poll from rx napi.

Actually, the issue goes back to the napi-tx even without the
opportunistic cleaning from the receive interrupt, I think? That races
with processing the vq in start_xmit.

> > > As a fix move everything that deals with the vq to under tx lock.
> > >

If the above is correct:

Fixes: b92f1e6751a6 ("virtio-net: transmit napi")

> > > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > > ---
> > >   drivers/net/virtio_net.c | 22 +++++++++++++++++++++-
> > >   1 file changed, 21 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > > index 16d5abed582c..460ccdbb840e 100644
> > > --- a/drivers/net/virtio_net.c
> > > +++ b/drivers/net/virtio_net.c
> > > @@ -1505,6 +1505,8 @@ static int virtnet_poll_tx(struct napi_struct *napi, int budget)
> > >     struct virtnet_info *vi = sq->vq->vdev->priv;
> > >     unsigned int index = vq2txq(sq->vq);
> > >     struct netdev_queue *txq;
> > > +   int opaque;

nit: virtqueue_napi_complete also stores as int opaque, but
virtqueue_enable_cb_prepare actually returns, and virtqueue_poll
expects, an unsigned int. In the end, conversion works correctly. But
cleaner to use the real type.

> > > +   bool done;
> > >     if (unlikely(is_xdp_raw_buffer_queue(vi, index))) {
> > >             /* We don't need to enable cb for XDP */
> > > @@ -1514,10 +1516,28 @@ static int virtnet_poll_tx(struct napi_struct *napi, int budget)
> > >     txq = netdev_get_tx_queue(vi->dev, index);
> > >     __netif_tx_lock(txq, raw_smp_processor_id());
> > > +   virtqueue_disable_cb(sq->vq);
> > >     free_old_xmit_skbs(sq, true);
> > > +
> > > +   opaque = virtqueue_enable_cb_prepare(sq->vq);
> > > +
> > > +   done = napi_complete_done(napi, 0);
> > > +
> > > +   if (!done)
> > > +           virtqueue_disable_cb(sq->vq);
> > > +
> > >     __netif_tx_unlock(txq);
> > > -   virtqueue_napi_complete(napi, sq->vq, 0);
> >
> >
> > So I wonder why not simply move __netif_tx_unlock() after
> > virtqueue_napi_complete()?
> >
> > Thanks
> >
>
>
> Because that calls tx poll which also takes tx lock internally ...

which tx poll?