Thread (8 messages) 8 messages, 4 authors, 2025-12-22

Re: [PATCH net v2] virtio-net: enable all napis before scheduling refill work

From: Jason Wang <jasowang@redhat.com>
Date: 2025-12-16 04:17:02
Also in: bpf, lkml, stable, virtualization

On Fri, Dec 12, 2025 at 11:28 PM Bui Quang Minh
[off-list ref] wrote:
quoted hunk ↗ jump to hunk
Calling napi_disable() on an already disabled napi can cause the
deadlock. In commit 4bc12818b363 ("virtio-net: disable delayed refill
when pausing rx"), to avoid the deadlock, when pausing the RX in
virtnet_rx_pause[_all](), we disable and cancel the delayed refill work.
However, in the virtnet_rx_resume_all(), we enable the delayed refill
work too early before enabling all the receive queue napis.

The deadlock can be reproduced by running
selftests/drivers/net/hw/xsk_reconfig.py with multiqueue virtio-net
device and inserting a cond_resched() inside the for loop in
virtnet_rx_resume_all() to increase the success rate. Because the worker
processing the delayed refilled work runs on the same CPU as
virtnet_rx_resume_all(), a reschedule is needed to cause the deadlock.
In real scenario, the contention on netdev_lock can cause the
reschedule.

This fixes the deadlock by ensuring all receive queue's napis are
enabled before we enable the delayed refill work in
virtnet_rx_resume_all() and virtnet_open().

Fixes: 4bc12818b363 ("virtio-net: disable delayed refill when pausing rx")
Reported-by: Paolo Abeni <pabeni@redhat.com>
Closes: https://netdev-ctrl.bots.linux.dev/logs/vmksft/drv-hw-dbg/results/400961/3-xdp-py/stderr
Cc: stable@vger.kernel.org
Signed-off-by: Bui Quang Minh <redacted>
---
Changes in v2:
- Move try_fill_recv() before rx napi_enable()
- Link to v1: https://lore.kernel.org/netdev/20251208153419.18196-1-minhquangbui99@gmail.com/ (local)
---
 drivers/net/virtio_net.c | 71 +++++++++++++++++++++++++---------------
 1 file changed, 45 insertions(+), 26 deletions(-)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 8e04adb57f52..4e08880a9467 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -3214,21 +3214,31 @@ static void virtnet_update_settings(struct virtnet_info *vi)
 static int virtnet_open(struct net_device *dev)
 {
        struct virtnet_info *vi = netdev_priv(dev);
+       bool schedule_refill = false;
        int i, err;

-       enable_delayed_refill(vi);
-
+       /* - We must call try_fill_recv before enabling napi of the same receive
+        * queue so that it doesn't race with the call in virtnet_receive.
+        * - We must enable and schedule delayed refill work only when we have
+        * enabled all the receive queue's napi. Otherwise, in refill_work, we
+        * have a deadlock when calling napi_disable on an already disabled
+        * napi.
+        */
        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);
+                               schedule_refill = true;

                err = virtnet_enable_queue_pair(vi, i);
                if (err < 0)
                        goto err_enable_qp;
        }
So NAPI could be scheduled and it may want to refill but since refill
is not enabled, there would be no refill work.

Is this a problem?

quoted hunk ↗ jump to hunk
+       enable_delayed_refill(vi);
+       if (schedule_refill)
+               schedule_delayed_work(&vi->refill, 0);
+
        if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_STATUS)) {
                if (vi->status & VIRTIO_NET_S_LINK_UP)
                        netif_carrier_on(vi->dev);
@@ -3463,39 +3473,48 @@ static void virtnet_rx_pause(struct virtnet_info *vi, struct receive_queue *rq)
        __virtnet_rx_pause(vi, rq);
 }
Thanks
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help