Re: [virtio-dev] virtio-blk: What's up with VIRTIO_BLK_T_GET_ID (8)?

[Stefan Hajnoczi <stefanha@redhat.com>]

On Sat, Nov 30, 2019 at 06:06:48PM +0100, Jan Kiszka wrote:
> On 28.04.17 14:49, Stefan Hajnoczi wrote:
> > On Fri, Apr 21, 2017 at 11:12:59AM +0200, Cornelia Huck wrote:
> > > I see the request type in $SUBJECT implemented in both Linux and QEMU,
> > > but I can't find it in the standard.
> > > 
> > > If we believe that getting a s/n is useful, we should spec it;
> > > otherwise, if we consider it just a legacy request type, it wouldn't
> > > hurt to mention it in the legacy section.
> > 
> > It's not a legacy request type.
> > 
> > Alexey Zaytev documented it in 2012:
> > 
> > https://lists.linuxfoundation.org/pipermail/virtualization/2012-December/022313.html
> > 
> > It was in the VIRTIO <1.0 spec:
> > 
> > https://github.com/rustyrussell/virtio-spec/blob/master/virtio-spec.lyx
> > 
> > I don't understand why it wasn't part of VIRTIO 1.0... :/
> > 
> > It's worth sending a patch to add the request type to the spec again.
> > 
> 
> What happened with this? I was already searching for weired queue
> corruptions in my implementation when I suddenly got this unspecified
> request type.
> 
> How should this be added? With a feature bit? Or just document the given
> reality (mandatory)?

I have sent a patch to add it.  Let's document it as-is since it has
been widely supported for years.

A new feature bit wouldn't help in this case since existing drivers
already use the command.  Devices that don't support the command respond
with BLK_S_UNSUPP.

Stefan

Attachment: signature.asc
Description: PGP signature