[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250430012856-mutt-send-email-mst@kernel.org>
Date: Wed, 30 Apr 2025 01:29:06 -0400
From: "Michael S. Tsirkin" <mst@...hat.com>
To: Jason Wang <jasowang@...hat.com>
Cc: Jakub Kicinski <kuba@...nel.org>, davem@...emloft.net,
netdev@...r.kernel.org, edumazet@...gle.com, pabeni@...hat.com,
andrew+netdev@...n.ch, horms@...nel.org, xuanzhuo@...ux.alibaba.com,
eperezma@...hat.com, minhquangbui99@...il.com, romieu@...zoreil.com,
kuniyu@...zon.com, virtualization@...ts.linux.dev
Subject: Re: [PATCH net] virtio-net: don't re-enable refill work too early
when NAPI is disabled
On Wed, Apr 30, 2025 at 11:49:15AM +0800, Jason Wang wrote:
> On Tue, Apr 29, 2025 at 10:31 PM Jakub Kicinski <kuba@...nel.org> wrote:
> >
> > Commit 4bc12818b363 ("virtio-net: disable delayed refill when pausing rx")
> > fixed a deadlock between reconfig paths and refill work trying to disable
> > the same NAPI instance. The refill work can't run in parallel with reconfig
> > because trying to double-disable a NAPI instance causes a stall under the
> > instance lock, which the reconfig path needs to re-enable the NAPI and
> > therefore unblock the stalled thread.
> >
> > There are two cases where we re-enable refill too early. One is in the
> > virtnet_set_queues() handler. We call it when installing XDP:
> >
> > virtnet_rx_pause_all(vi);
> > ...
> > virtnet_napi_tx_disable(..);
> > ...
> > virtnet_set_queues(..);
> > ...
> > virtnet_rx_resume_all(..);
> >
> > We want the work to be disabled until we call virtnet_rx_resume_all(),
> > but virtnet_set_queues() kicks it before NAPIs were re-enabled.
> >
> > The other case is a more trivial case of mis-ordering in
> > __virtnet_rx_resume() found by code inspection.
> >
> > Fixes: 4bc12818b363 ("virtio-net: disable delayed refill when pausing rx")
> > Fixes: 413f0271f396 ("net: protect NAPI enablement with netdev_lock()")
> > Signed-off-by: Jakub Kicinski <kuba@...nel.org>
> > ---
> > CC: mst@...hat.com
> > CC: jasowang@...hat.com
> > CC: xuanzhuo@...ux.alibaba.com
> > CC: eperezma@...hat.com
> > CC: minhquangbui99@...il.com
> > CC: romieu@...zoreil.com
> > CC: kuniyu@...zon.com
> > CC: virtualization@...ts.linux.dev
> > ---
> > drivers/net/virtio_net.c | 9 ++++++---
> > 1 file changed, 6 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 848fab51dfa1..4c904e176495 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -3383,12 +3383,15 @@ 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_delayed_work(&vi->refill, 0);
> > -
> > + schedule_refill = true;
> > if (running)
> > virtnet_napi_enable(rq);
> > +
> > + if (schedule_refill)
> > + schedule_delayed_work(&vi->refill, 0);
> > }
> >
> > static void virtnet_rx_resume_all(struct virtnet_info *vi)
> > @@ -3728,7 +3731,7 @@ 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. */
> > - if (dev->flags & IFF_UP)
> > + if (dev->flags & IFF_UP && vi->refill_enabled)
> > schedule_delayed_work(&vi->refill, 0);
>
> This has the assumption that the toggle of the refill_enabled is under
> RTNL. Though it's true now but it looks to me it's better to protect
> it against refill_lock.
>
> Thanks
Good point.
> >
> > return 0;
> > --
> > 2.49.0
> >
Powered by blists - more mailing lists