Re: [BUG] vhost_net: livelock in handle_rx() when GRO packet exceeds virtqueue capacity

["ShuangYu" <shuangyu@yunyoo.cc>]

> From: "Michael S. Tsirkin"<mst@redhat.com>
> Date:  Mon, Mar 2, 2026, 08:45
> Subject:  Re: [BUG] vhost_net: livelock in handle_rx() when GRO packet exceeds virtqueue capacity
> To: "ShuangYu"<shuangyu@yunyoo.cc>
> Cc: "jasowang@redhat.com"<jasowang@redhat.com>, "virtualization@lists.linux.dev"<virtualization@lists.linux.dev>, "netdev@vger.kernel.org"<netdev@vger.kernel.org>, "linux-kernel@vger.kernel.org"<linux-kernel@vger.kernel.org>, "kvm@vger.kernel.org"<kvm@vger.kernel.org>
> On Sun, Mar 01, 2026 at 07:39:30PM -0500, Michael S. Tsirkin wrote:
> > On Sun, Mar 01, 2026 at 07:10:06PM -0500, Michael S. Tsirkin wrote:
> > > On Sun, Mar 01, 2026 at 10:36:39PM +0000, ShuangYu wrote:
> > > > Hi,
> > > > 
> > > > We have hit a severe livelock in vhost_net on 6.18.x. The vhost
> > > > kernel thread spins at 100% CPU indefinitely in handle_rx(), and
> > > > QEMU becomes unkillable (stuck in D state).
> > > > [This is a text/plain messages]
> > > > 
> > > > Environment
> > > > -----------
> > > >   Kernel:  6.18.10-1.el8.elrepo.x86_64
> > > >   QEMU:    7.2.19
> > > >   Virtio:  VIRTIO_F_IN_ORDER is negotiated
> > > >   Backend: vhost (kernel)
> > > > 
> > > > Symptoms
> > > > --------
> > > >   - vhost-<pid> kernel thread at 100% CPU (R state, never yields)
> > > >   - QEMU stuck in D state at vhost_dev_flush() after receiving SIGTERM
> > > >   - kill -9 has no effect on the QEMU process
> > > >   - libvirt management plane deadlocks ("cannot acquire state change lock")
> > > > 
> > > > Root Cause
> > > > ----------
> > > > The livelock is triggered when a GRO-merged packet on the host TAP
> > > > interface (e.g., ~60KB) exceeds the remaining free capacity of the
> > > > guest's RX virtqueue (e.g., ~40KB of available buffers).
> > > > 
> > > > The loop in handle_rx() (drivers/vhost/net.c) proceeds as follows:
> > > > 
> > > >   1. get_rx_bufs() calls vhost_get_vq_desc_n() to fetch descriptors.
> > > >     It advances vq->last_avail_idx and vq->next_avail_head as it
> > > >     consumes buffers, but runs out before satisfying datalen.
> > > > 
> > > >   2. get_rx_bufs() jumps to err: and calls
> > > >     vhost_discard_vq_desc(vq, headcount, n), which rolls back
> > > >     vq->last_avail_idx and vq->next_avail_head.
> > > > 
> > > >     Critically, vq->avail_idx (the cached copy of the guest's
> > > >     avail->idx) is NOT rolled back. This is correct behavior in
> > > >     isolation, but creates a persistent mismatch:
> > > > 
> > > >       vq->avail_idx      = 108  (cached, unchanged)
> > > >       vq->last_avail_idx = 104  (rolled back)
> > > > 
> > > >   3. handle_rx() sees headcount == 0 and calls vhost_enable_notify().
> > > >     Inside, vhost_get_avail_idx() finds:
> > > > 
> > > >       vq->avail_idx (108) != vq->last_avail_idx (104)
> > > > 
> > > >     It returns 1 (true), indicating "new buffers available."
> > > >     But these are the SAME buffers that were just discarded.
> > > > 
> > > >   4. handle_rx() hits `continue`, restarting the loop.
> > > > 
> > > >   5. In the next iteration, vhost_get_vq_desc_n() checks:
> > > > 
> > > >       if (vq->avail_idx == vq->last_avail_idx)
> > > > 
> > > >     This is FALSE (108 != 104), so it skips re-reading the guest's
> > > >     actual avail->idx and directly fetches the same descriptors.
> > > > 
> > > >   6. The exact same sequence repeats: fetch -> too small -> discard
> > > >     -> rollback -> "new buffers!" -> continue. Indefinitely.
> > > > 
> > > > This appears to be a regression introduced by the VIRTIO_F_IN_ORDER
> > > > support, which added vhost_get_vq_desc_n() with the cached avail_idx
> > > > short-circuit check, and the two-argument vhost_discard_vq_desc()
> > > > with next_avail_head rollback. The mismatch between the rollback
> > > > scope (last_avail_idx, next_avail_head) and the check scope
> > > > (avail_idx vs last_avail_idx) was not present before this change.
> > > > 
> > > > bpftrace Evidence
> > > > -----------------
> > > > During the 100% CPU lockup, we traced:
> > > > 
> > > >   @get_rx_ret[0]:      4468052   // get_rx_bufs() returns 0 every time
> > > >   @peek_ret[60366]:    4385533   // same 60KB packet seen every iteration
> > > >   @sock_err[recvmsg]:        0   // tun_recvmsg() is never reached
> > > > 
> > > > vhost_get_vq_desc_n() was observed iterating over the exact same 11
> > > > descriptor addresses millions of times per second.
> > > > 
> > > > Workaround
> > > > ----------
> > > > Either of the following avoids the livelock:
> > > > 
> > > >   - Disable GRO/GSO on the TAP interface:
> > > >      ethtool -K <tap> gro off gso off
> > > > 
> > > >   - Switch from kernel vhost to userspace QEMU backend:
> > > >      <driver name='qemu'/> in libvirt XML
> > > > 
> > > > Bisect
> > > > ------
> > > > We have not yet completed a full git bisect, but the issue does not
> > > > occur on 6.17.x kernels which lack the VIRTIO_F_IN_ORDER vhost
> > > > support. We will follow up with a Fixes: tag if we can identify the
> > > > exact commit.
> > > > 
> > > > Suggested Fix Direction
> > > > -----------------------
> > > > In handle_rx(), when get_rx_bufs() returns 0 (headcount == 0) due to
> > > > insufficient buffers (not because the queue is truly empty), the code
> > > > should break out of the loop rather than relying on
> > > > vhost_enable_notify() to make that determination. For example, when
> > > > get_rx_bufs() returns r == 0 with datalen still > 0, this indicates a
> > > > "packet too large" condition, not a "queue empty" condition, and
> > > > should be handled differently.
> > > > 
> > > > Thanks,
> > > > ShuangYu
> > > 
> > > Hmm. On a hunch, does the following help? completely untested,
> > > it is night here, sorry.
> > > 
> > > 
> > > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> > > index 2f2c45d20883..aafae15d5156 100644
> > > --- a/drivers/vhost/vhost.c
> > > +++ b/drivers/vhost/vhost.c
> > > @@ -1522,6 +1522,7 @@ static void vhost_dev_unlock_vqs(struct vhost_dev *d)
> > >  static inline int vhost_get_avail_idx(struct vhost_virtqueue *vq)
> > >  {
> > >          __virtio16 idx;
> > > +        u16 avail_idx;
> > >          int r;
> > >  
> > >          r = vhost_get_avail(vq, idx, &vq->avail->idx);
> > > @@ -1532,17 +1533,19 @@ static inline int vhost_get_avail_idx(struct vhost_virtqueue *vq)
> > >          }
> > >  
> > >          /* Check it isn't doing very strange thing with available indexes */
> > > -        vq->avail_idx = vhost16_to_cpu(vq, idx);
> > > -        if (unlikely((u16)(vq->avail_idx - vq->last_avail_idx) > vq->num)) {
> > > +        avail_idx = vhost16_to_cpu(vq, idx);
> > > +        if (unlikely((u16)(avail_idx - vq->last_avail_idx) > vq->num)) {
> > >                  vq_err(vq, "Invalid available index change from %u to %u",
> > >                         vq->last_avail_idx, vq->avail_idx);
> > >                  return -EINVAL;
> > >          }
> > >  
> > >          /* We're done if there is nothing new */
> > > -        if (vq->avail_idx == vq->last_avail_idx)
> > > +        if (avail_idx == vq->avail_idx)
> > >                  return 0;
> > >  
> > > +        vq->avail_idx == avail_idx;
> > > +
> > 
> > meaning 
> >         vq->avail_idx = avail_idx; 
> > of course
> > 
> > >          /*
> > >           * We updated vq->avail_idx so we need a memory barrier between
> > >           * the index read above and the caller reading avail ring entries.
> 
> 
> and the change this is fixing was done in d3bb267bbdcba199568f1325743d9d501dea0560
> 
> -- 
> MST
> 

Thank you for the quick fix and for identifying the root commit.

I've reviewed the patch and I believe the logic is correct — changing
the "nothing new" check in vhost_get_avail_idx() from comparing
against vq->last_avail_idx to comparing against the cached
vq->avail_idx makes it immune to the rollback done by
vhost_discard_vq_desc(), which is exactly what breaks the loop.

One minor nit: the vq_err message on the sanity check path still
references vq->avail_idx before it has been updated:

        vq_err(vq, "Invalid available index change from %u to %u",
-              vq->last_avail_idx, vq->avail_idx);
+              vq->last_avail_idx, avail_idx);


Since this issue was found in production, I need some time to prepare a test setup to verify the patch

Thanks,
ShuangYu