Re: [PATCH net v2] vsock/virtio: fix skb overhead accounting to preserve full buf_alloc

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, May 12, 2026 at 12:03:38PM +0200, Stefano Garzarella wrote:
> On Tue, May 12, 2026 at 04:54:34AM -0400, Michael S. Tsirkin wrote:
> > On Tue, May 12, 2026 at 10:07:37AM +0200, Stefano Garzarella wrote:
> > > From: Stefano Garzarella <sgarzare@redhat.com>
> > > 
> > > After commit 059b7dbd20a6 ("vsock/virtio: fix potential unbounded skb
> > > queue"), virtio_transport_inc_rx_pkt() subtracts per-skb overhead from
> > > buf_alloc when checking whether a new packet fits. This reduces the
> > > effective receive buffer below what the user configured via
> > > SO_VM_SOCKETS_BUFFER_SIZE, causing legitimate data packets to be
> > > silently dropped and applications that rely on the full buffer size
> > > to deadlock.
> > > 
> > > Also, the reduced space is not communicated to the remote peer, so
> > > its credit calculation accounts more credit than the receiver will
> > > actually accept, causing data loss (there is no retransmission).
> > > 
> > > With this approach we currently have failures in
> > > tools/testing/vsock/vsock_test.c. Test 18 sometimes fails, while
> > > test 22 always fails in this way:
> > >     18 - SOCK_STREAM MSG_ZEROCOPY...hash mismatch
> > > 
> > >     22 - SOCK_STREAM virtio credit update + SO_RCVLOWAT...send failed:
> > >     Resource temporarily unavailable
> > > 
> > > Fix this by using `buf_alloc * 2` as the total budget for payload plus
> > > skb overhead in virtio_transport_inc_rx_pkt(), similar to how SO_RCVBUF
> > > is doubled to reserve space for sk_buff metadata. This preserves the
> > > full buf_alloc for payload under normal operation, while still bounding
> > > the skb queue growth.
> > > 
> > > When the total budget (buf_alloc * 2) is exceeded (e.g. under small-packet
> > > flooding where overhead dominates), the connection is reset and local
> > > socket error set to ENOBUFS, so both peers are explicitly notified of
> > > the failure rather than silently losing data.
> > > 
> > > With this patch, all tests in tools/testing/vsock/vsock_test.c are
> > > now passing again.
> > > 
> > > A solution to handle small-packet overhead efficiently also for
> > > SEQPACKET (we already do that for STREAM) is planned as follow-up work.
> > > This patch is needed in any case to prevent silent data loss, because
> > > even if we reduce the overhead, we can't eliminate it entirely.
> > > 
> > > Fixes: 059b7dbd20a6 ("vsock/virtio: fix potential unbounded skb queue")
> > > Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
> > 
> > Thanks for the patch!  I'd like to split this:
> > 1. buf alloc boost
> > 2. reset when out of credits
> 
> Good point, also the reset maybe should have an other fixes tag (i.e.  when
> we introduced that check)
> 
> > 
> > this way we can revert 2 easier later.
> 
> I'm not sure if we should revert them at some point, even though we'll be
> able to handle the overhead better,

I mean we'll prevent the overflow, the condition will never be met.


> but I agree that we should split them.
> 
> I'll wait for a few more comments and then send v3 with the split.
> 
> Thanks,
> Stefano