Re: [Qemu-devel] [PATCH 1/3] qmp: Add command 'blockdev-backup'

[Fam Zheng]

On Tue, 11/04 07:47, Markus Armbruster wrote:
> Fam Zheng <address@hidden> writes:
> 
> > On Mon, 11/03 15:32, Kevin Wolf wrote:
> >> Am 03.11.2014 um 02:46 hat Fam Zheng geschrieben:
> >> > On Fri, 10/31 10:01, Kevin Wolf wrote:
> >> > > Am 11.09.2014 um 07:05 hat Fam Zheng geschrieben:
> >> > > > Similar to drive-backup, but this command uses a device id as target
> >> > > > instead of creating/opening an image file.
> >> > > > 
> >> > > > Also add blocker on target bs, since the target is also a named 
> >> > > > device
> >> > > > now.
> >> > > > 
> >> > > > Add check and report error for bs == target which became possible 
> >> > > > but is
> >> > > > an illegal case with introduction of blockdev-backup.
> >> > > > 
> >> > > > Signed-off-by: Fam Zheng <address@hidden>
> >> > > 
> >> > > > diff --git a/qapi/block-core.json b/qapi/block-core.json
> >> > > > index a685d02..b953c7b 100644
> >> > > > --- a/qapi/block-core.json
> >> > > > +++ b/qapi/block-core.json
> >> > > > @@ -669,6 +669,40 @@
> >> > > >              '*on-target-error': 'BlockdevOnError' } }
> >> > > >  
> >> > > >  ##
> >> > > > +# @BlockdevBackup
> >> > > > +#
> >> > > > +# @device: the name of the device which should be copied.
> >> > > > +#
> >> > > > +# @target: the name of the backup target device.
> >> > > 
> >> > > Both of these are either a BlockBackend ID or a BDS node-name, right? 
> >> > > Do
> >> > > we have a standard way of expressing this? "name of the device" isn't
> >> > > quite clear.
> >> > 
> >> > "name of the device" is used everywhere to document the "device"
> >> > parameters in
> >> > our json schema. Since we have BlockBackend now, device-name and 
> >> > node-name
> >> > could be better distinguished. All we have to do is giving a
> >> > beautiful name to
> >> > both.
> >> > 
> >> > [This patch is only a copy&paste and is consistent with the rest part of 
> >> > the
> >> > file. So I'll leave it for now :]
> >> 
> >> The rest of the file doesn't accept node names. But looking at your
> >> actual code, it seems that you are doing the same (by usign bdrv_find()
> >> instead of bdrv_lookup_bs()).
> >
> > Yes, to be consistent with drive-backup.
> >
> >> 
> >> Shouldn't a proper blockdev-* command accept node names as well?
> >> 
> >
> > I am not sure, it's still blockdev-backup, not blocknode-backup.
> >
> > I think that may be another thing, to changed drive-*'s @device parameter, 
> > and
> > blockdev-*'s @device and @target to accept node names, altogether.
> 
> We have many commands identifying nodes by some name: root nodes by BDS
> device_name[] (now BB name), inner nodes by "file name" (ugh), and
> arbitrary nodes by BDS name_name[].
> 
> An obvious task is deprecating "file name" in favor of node names.
> 
> Less obvious is where to accept a node name instead of / in addition to
> a device name.
> 
> Want me to start a thread on this?

Yes please, it's good to review the situation and make some concrete
convention and plan for the transition.

Fam