On Fri, Jul 26, 2024 at 10:12:31AM +0800, Jason Wang wrote: > On Fri, Jul 26, 2024 at 6:19 AM Peter Xu <pet...@redhat.com> wrote: > > > > On Tue, Aug 01, 2023 at 01:31:48AM +0300, Yuri Benditovich wrote: > > > USO features of virtio-net device depend on kernel ability > > > to support them, for backward compatibility by default the > > > features are disabled on 8.0 and earlier. > > > > > > Signed-off-by: Yuri Benditovich <yuri.benditov...@daynix.com> > > > Signed-off-by: Andrew Melnychecnko <and...@daynix.com> > > > > Looks like this patch broke migration when the VM starts on a host that has > > USO supported, to another host that doesn't.. > > > > Yuri, would it be possible we always keep all the USO* features off by > > default (so this feature bit never affects migration ABI), but then: > > > > - only enable them when the user specified ON > > > > - meanwhile, if detecting host feature doesn't support USO*, it could > > fail qemu from boot, rather than silently turning it from ON->OFF > > > > ? > > I agree, I have raised the same issue several times in the past. > > > > > Silently flipping the bit may cause migration issues like this. > > Looking at virtio_net_get_features(), it silently clears a lot of features...
Yes.. :-( I saw that too when looking at this. Is it because most of those features are supported on most of the kernels, so we're good until now by chance? While it looks like e.g. TUN_F_USO4 was supported only since ~1.5 years ago so it looks relatively new. Thanks, -- Peter Xu
