Re: [PATCH v4 5/5] iotests: add backup-discard-source

[Vladimir Sementsov-Ogievskiy]

On 13.06.24 11:02, Kevin Wolf wrote:
> Am 12.06.2024 um 21:21 hat Vladimir Sementsov-Ogievskiy geschrieben:
> > On 11.06.24 20:49, Kevin Wolf wrote:
> > > Am 13.03.2024 um 16:28 hat Vladimir Sementsov-Ogievskiy geschrieben:
> > > > Add test for a new backup option: discard-source.
> > > >
> > > > Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
> > > > Reviewed-by: Fiona Ebner <f.ebner@proxmox.com>
> > > > Tested-by: Fiona Ebner <f.ebner@proxmox.com>
> > >
> > > This test fails for me, and it already does so after this commit that
> > > introduced it. I haven't checked what get_actual_size(), but I'm running
> > > on XFS, so its preallocation could be causing this. We generally avoid
> > > checking the number of allocated blocks in image files for this reason.
> >
> > Hmm right, I see that relying on allocated size is bad thing. Better
> > is to check block status, to see how many qcow2 clusters are
> > allocated.
> >
> > Do we have some qmp command to get such information? The simplest way
> > I see is to add dirty to temp block-node, and then check its dirty
> > count through query-named-block-nodes
>
> Hm, does it have to be QMP in a running QEMU process?

hmm, yes, seems in test_discard_written() we do want to examine running 
process. I'll try to go with bitmap

> I'm not sure what
> the best way would be there.
>
> Otherwise, my approach would be 'qemu-img check' or even 'qemu-img map',
> depending on what you want to verify with it.
>
> Kevin

--
Best regards,
Vladimir