Am 24.08.2021 um 17:25 hat Peter Maydell geschrieben: > On Tue, 24 Aug 2021 at 16:21, Kevin Wolf <kw...@redhat.com> wrote: > > > > Am 17.08.2021 um 20:48 hat Peter Maydell geschrieben: > > > On Tue, 17 Aug 2021 at 17:27, Dr. David Alan Gilbert > > > <dgilb...@redhat.com> wrote: > > > > > > > > * Gabriel Southern (gsout...@qti.qualcomm.com) wrote: > > > > > Hi, > > > > > > > > > > Are there plans to support using -loadvm with -snapshot? > > > > > > > > > > I saw some past discussion on mailing list including bug that was > > > > > closed earlier this year but no recent activity: > > > > > > > > > > https://lore.kernel.org/qemu-devel/162424905685.11837.7303570061857383641.mal...@loganberry.canonical.com/ > > > > > > > > > > Testing with latest qemu it looks like -loadvm still does not work > > > > > when combined with -snapshot. > > > > > > > > > > I'm curious how complex it would be to implement this feature and if > > > > > it may show up on QEMU roadmap in the future. Or if there is > > > > > alterative command that can be used to save the state of a running VM > > > > > and then use the same qcow to run multiple QEMU instances loading > > > > > this VM state running in snapshot mode. > > > > > > Do you know why -loadvm and -snapshot don't work together? > > > It doesn't on the face of it seem like they would be incompatible... > > > > Probably because -snapshot first adds a temporary qcow2 overlay to the > > image, and then -loadvm fails because the newly created temporary > > overlay doesn't contain the requested snapshot. > > That sounds like the bug could be fixed by having "find the > snapshot" look not just in the overlay but also down into > the underlying image?
In this specific case yes, because we know that the overlay was just created and is still empty. In the general case, loading internal snapshots on a backing file would be wrong because the data in the overlay would be from a different time than the data in the backing file, probably causing data corruption. Kevin
