QE tested this patch's v2 with virtio-net regression tests, everything
works fine.
Tested-by: Lei Yang <leiy...@redhat.com>
On Thu, Apr 24, 2025 at 11:49 PM Antoine Damhet <adam...@scaleway.com> wrote:
>
> On Thu, Apr 24, 2025 at 06:49:57PM +0900, Akihiko Odaki wrote:
> > Commit e28fbd1c525d ("Revert "virtio-net: Copy received header to
> > buffer"") reverted commit 7987d2be5a8b, which attempted to remove the
> > need to patch the (const) input buffer.
> >
> > Achieve the original goal by copying the header or the entire packet to
> > a writable buffer as necessary. Copy the virtio-net header when patching
> > it. Copy the entire packet when filling the UDP checksum as required by
> > net_checksum_calculate().
> >
> > Signed-off-by: Akihiko Odaki <akihiko.od...@daynix.com>
> > ---
> > Supersedes: <20250405-mtu-v1-1-08c5910fa...@daynix.com>
> > ("[PATCH] virtio-net: Copy all for dhclient workaround")
> >
> > This reapplies commit 7987d2be5a8b ("virtio-net: Copy all for dhclient
> > workaround"), which was reverted by commit e28fbd1c525d ("Revert
> > "virtio-net: Copy received header to buffer""), with a fix in the
> > superseded patch. It also renames identifiers according to the
> > discussion with Antoine Damhet.
> > ---
> > Changes in v2:
> > - Rewrote the message avoiding archeology as suggested by
> > Michael S. Tsirkin.
> > - Link to v1:
> > https://lore.kernel.org/qemu-devel/20250423-reapply-v1-1-6f4fc3027...@daynix.com
> > ---
> > hw/net/virtio-net.c | 91
> > ++++++++++++++++++++++++++++-------------------------
> > 1 file changed, 48 insertions(+), 43 deletions(-)
>
> Reviewed-by: Antoine Damhet <adam...@scaleway.com>
> Tested-by: Antoine Damhet <adam...@scaleway.com>
>
> --
> Antoine 'xdbob' Damhet
> Engineer @scaleway
