Re: [Qemu-devel] Re: [PATCH] virtio-blk: assign a default serial number if none provided

[Ryan Harper <ryanh@us.ibm.com>]

* Michael S. Tsirkin <mst@redhat.com> [2010-06-02 06:59]:
> On Wed, Jun 02, 2010 at 06:45:46AM -0500, Ryan Harper wrote:
> > * Michael S. Tsirkin <mst@redhat.com> [2010-06-02 04:08]:
> > > On Tue, Jun 01, 2010 at 08:48:54PM -0500, Ryan Harper wrote:
> > > > This patch applies on-top of John's virtio-blk serial patches.
> > > > 
> > > > Generate default serial numbers for virtio drives based on DriveInfo.unit which is
> > > > incremented for each additional virtio-blk device.  This provides a
> > > > per-virtio-blk number to use in the default string: QM%05d that is used in
> > > > hw/ide/core.c.  The resulting serial number looks like: QM00001, etc.
> > > > 
> > > > Signed-off-by: Ryan Harper <ryanh@us.ibm.com>
> > > 
> > > 
> > > I think that there's a problem with this approach in that hot plug A,
> > > hot plug B, hot unplug A is not the same as hot plug B.
> > > So you might get guest boot failures and no easy way to
> > > figure out why. For guests that need S/N, I think they
> > > really must be persistent.
> > 
> > That's true; though I think most boot drives boot via either LVM or UUID
> > which will remain persistent.  That said, if you are relying on the
> > by-id; then of course the user will need to specify serial versus having
> > one auto-generated.
> 
> I guess the question then would be, if you don't rely on the S/N, why
> do you want to set it?

Functional similarity; we have default serial numbers for ide and scsi.
scsi is hotpluggable and would suffer the same issue; and of course ide
doesn't do hotplug.  I don't think most folks will encounter the above
scenario and that having the same default serial numbers being generated
like we do ide and scsi is reasonable.


-- 
Ryan Harper
Software Engineer; Linux Technology Center
IBM Corp., Austin, Tx
ryanh@us.ibm.com