On Sun, Jan 04, 2026 at 03:34:52PM +0700, Bui Quang Minh wrote:
> On 1/4/26 13:09, Jason Wang wrote:
> > On Fri, Jan 2, 2026 at 11:20 PM Bui Quang Minh <[email protected]>
> > wrote:
> > > When we fail to refill the receive buffers, we schedule a delayed worker
> > > to retry later. However, this worker creates some concurrency issues
> > > such as races and deadlocks. To simplify the logic and avoid further
> > > problems, we will instead retry refilling in the next NAPI poll.
> > >
> > > Fixes: 4bc12818b363 ("virtio-net: disable delayed refill when pausing rx")
> > > Reported-by: Paolo Abeni <[email protected]>
> > > Closes:
> > > https://netdev-ctrl.bots.linux.dev/logs/vmksft/drv-hw-dbg/results/400961/3-xdp-py/stderr
> > > Cc: [email protected]
> > > Suggested-by: Xuan Zhuo <[email protected]>
> > > Signed-off-by: Bui Quang Minh <[email protected]>
> > > ---
> > > drivers/net/virtio_net.c | 55 ++++++++++++++++++++++------------------
> > > 1 file changed, 30 insertions(+), 25 deletions(-)
> > >
> > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > > index 1bb3aeca66c6..ac514c9383ae 100644
> > > --- a/drivers/net/virtio_net.c
> > > +++ b/drivers/net/virtio_net.c
> > > @@ -3035,7 +3035,7 @@ static int virtnet_receive_packets(struct
> > > virtnet_info *vi,
> > > }
> > >
> > > static int virtnet_receive(struct receive_queue *rq, int budget,
> > > - unsigned int *xdp_xmit)
> > > + unsigned int *xdp_xmit, bool *retry_refill)
> > > {
> > > struct virtnet_info *vi = rq->vq->vdev->priv;
> > > struct virtnet_rq_stats stats = {};
> > > @@ -3047,12 +3047,8 @@ static int virtnet_receive(struct receive_queue
> > > *rq, int budget,
> > > packets = virtnet_receive_packets(vi, rq, budget,
> > > xdp_xmit, &stats);
> > >
> > > if (rq->vq->num_free > min((unsigned int)budget,
> > > virtqueue_get_vring_size(rq->vq)) / 2) {
> > > - if (!try_fill_recv(vi, rq, GFP_ATOMIC)) {
> > > - spin_lock(&vi->refill_lock);
> > > - if (vi->refill_enabled)
> > > - schedule_delayed_work(&vi->refill, 0);
> > > - spin_unlock(&vi->refill_lock);
> > > - }
> > > + if (!try_fill_recv(vi, rq, GFP_ATOMIC))
> > > + *retry_refill = true;
> > > }
> > >
> > > u64_stats_set(&stats.packets, packets);
> > > @@ -3129,18 +3125,18 @@ static int virtnet_poll(struct napi_struct *napi,
> > > int budget)
> > > struct send_queue *sq;
> > > unsigned int received;
> > > unsigned int xdp_xmit = 0;
> > > - bool napi_complete;
> > > + bool napi_complete, retry_refill = false;
> > >
> > > virtnet_poll_cleantx(rq, budget);
> > >
> > > - received = virtnet_receive(rq, budget, &xdp_xmit);
> > > + received = virtnet_receive(rq, budget, &xdp_xmit, &retry_refill);
> > I think we can simply let virtnet_receive() to return the budget when
> > reill fails.
>
> That makes sense, I'll change it.
>
> >
> > > rq->packets_in_napi += received;
> > >
> > > if (xdp_xmit & VIRTIO_XDP_REDIR)
> > > xdp_do_flush();
> > >
> > > /* Out of packets? */
> > > - if (received < budget) {
> > > + if (received < budget && !retry_refill) {
> > > napi_complete = virtqueue_napi_complete(napi, rq->vq,
> > > received);
> > > /* Intentionally not taking dim_lock here. This may
> > > result in a
> > > * spurious net_dim call. But if that happens
> > > virtnet_rx_dim_work
> > > @@ -3160,7 +3156,7 @@ static int virtnet_poll(struct napi_struct *napi,
> > > int budget)
> > > virtnet_xdp_put_sq(vi, sq);
> > > }
> > >
> > > - return received;
> > > + return retry_refill ? budget : received;
> > > }
> > >
> > > static void virtnet_disable_queue_pair(struct virtnet_info *vi, int
> > > qp_index)
> > > @@ -3230,9 +3226,11 @@ static int virtnet_open(struct net_device *dev)
> > >
> > > for (i = 0; i < vi->max_queue_pairs; i++) {
> > > if (i < vi->curr_queue_pairs)
> > > - /* Make sure we have some buffers: if oom use wq.
> > > */
> > > - if (!try_fill_recv(vi, &vi->rq[i], GFP_KERNEL))
> > > - schedule_delayed_work(&vi->refill, 0);
> > > + /* If this fails, we will retry later in
> > > + * NAPI poll, which is scheduled in the below
> > > + * virtnet_enable_queue_pair
> > > + */
> > > + try_fill_recv(vi, &vi->rq[i], GFP_KERNEL);
> > Consider NAPI will be eventually scheduled, I wonder if it's still
> > worth to refill here.
>
> With GFP_KERNEL here, I think it's more likely to succeed than GFP_ATOMIC in
> NAPI poll. Another small benefit is that the actual packet can happen
> earlier. In case the receive buffer is empty and we don't refill here, the
> #1 NAPI poll refill the buffer and the #2 NAPI poll can receive packets. The
> #2 NAPI poll is scheduled in the interrupt handler because the #1 NAPI poll
> will deschedule the NAPI and enable the device interrupt. In case we
> successfully refill here, the #1 NAPI poll can receive packets right away.
>
Right. But I think this is a part that needs elucidating, not
error handling.
/* Pre-fill rq agressively, to make sure we are ready to get packets
* immediately.
* */
> >
> > > err = virtnet_enable_queue_pair(vi, i);
> > > if (err < 0)
> > > @@ -3473,15 +3471,15 @@ static void __virtnet_rx_resume(struct
> > > virtnet_info *vi,
> > > bool refill)
> > > {
> > > bool running = netif_running(vi->dev);
> > > - bool schedule_refill = false;
> > >
> > > - if (refill && !try_fill_recv(vi, rq, GFP_KERNEL))
> > > - schedule_refill = true;
> > > + if (refill)
> > > + /* If this fails, we will retry later in NAPI poll, which
> > > is
> > > + * scheduled in the below virtnet_napi_enable
> > > + */
> > > + try_fill_recv(vi, rq, GFP_KERNEL);
> > and here.
> >
> > > +
> > > if (running)
> > > virtnet_napi_enable(rq);
here the part that isn't clear is why are we refilling if !running
and what handles failures in that case.
> > > -
> > > - if (schedule_refill)
> > > - schedule_delayed_work(&vi->refill, 0);
> > > }
> > >
> > > static void virtnet_rx_resume_all(struct virtnet_info *vi)
> > > @@ -3777,6 +3775,7 @@ static int virtnet_set_queues(struct virtnet_info
> > > *vi, u16 queue_pairs)
> > > struct virtio_net_rss_config_trailer old_rss_trailer;
> > > struct net_device *dev = vi->dev;
> > > struct scatterlist sg;
> > > + int i;
> > >
> > > if (!vi->has_cvq || !virtio_has_feature(vi->vdev,
> > > VIRTIO_NET_F_MQ))
> > > return 0;
> > > @@ -3829,11 +3828,17 @@ static int virtnet_set_queues(struct virtnet_info
> > > *vi, u16 queue_pairs)
> > > }
> > > succ:
> > > vi->curr_queue_pairs = queue_pairs;
> > > - /* virtnet_open() will refill when device is going to up. */
> > > - spin_lock_bh(&vi->refill_lock);
> > > - if (dev->flags & IFF_UP && vi->refill_enabled)
> > > - schedule_delayed_work(&vi->refill, 0);
> > > - spin_unlock_bh(&vi->refill_lock);
> > > + if (dev->flags & IFF_UP) {
> > > + /* Let the NAPI poll refill the receive buffer for us. We
> > > can't
> > > + * safely call try_fill_recv() here because the NAPI
> > > might be
> > > + * enabled already.
> > > + */
> > > + local_bh_disable();
> > > + for (i = 0; i < vi->curr_queue_pairs; i++)
> > > + virtqueue_napi_schedule(&vi->rq[i].napi,
> > > vi->rq[i].vq);
> > > +
> > > + local_bh_enable();
> > > + }
> > >
> > > return 0;
> > > }
> > > --
> > > 2.43.0
> > >
> > Thanks
> >
>
> Thanks,
> Quang Minh.
