Quoting Eric Blake (2014-08-07 10:50:41) > On 08/07/2014 03:19 AM, Michael Roth wrote: > >> > >> Libvirt could support active commit against qemu 2.0.1 if you backport > >> these patches: > >> > >> Jeff Cody > >> 7676e2c597 block: make 'top' argument to block-commit optional > >> > >> Fam Zheng > >> 9e48b02540 mirror: Go through ready -> complete process for 0 len image > > > > Actually ended up needing the following with a few fix-ups: > > > > 7676e2c->98103fa block: make 'top' argument to block-commit optional > > 8b9a30c->e5f0eb0 qemu-iotests: Test BLOCK_JOB_READY event for 0Kb image > > active commit > > 9e48b02->43ac708 mirror: Go through ready -> complete process for 0 len > > image > > dc71ce4->8e09e20 blockjob: Add block_job_yield() > > 373df5b->520b341 mirror: Fix resource leak when bdrv_getlength fails > > > > I've gone ahead and pushed them, but please test as we generally don't > > do features (even backward-compatible ones) for stable, and this wasn't > > as trivial as I was hoping. > > Yes, I'll test and report back. However, I don't think this is a
Thanks! > Yes, I'll test and report back. However, I don't think this is a > feature addition, so much as a bug fix for an existing feature (all the > hard work for active commit was already in 2.0, all that was missing was > a way for libvirt to introspect that it existed, and some corner case > bugs with 0-length images). Well, I think it would be hard to draw the line if you start getting into things like new command-line options and such, but the distinction does seem reasonable in this case. > > -- > Eric Blake eblake redhat com +1-919-301-3266 > Libvirt virtualization library http://libvirt.org
