Re: [Qemu-devel] [PATCHv2] virtio: verify that all outstanding buffers are flushed

[Michael S. Tsirkin]

On Wed, Dec 12, 2012 at 02:50:50PM +0100, Stefan Hajnoczi wrote:
> On Wed, Dec 12, 2012 at 12:51:01PM +0200, Michael S. Tsirkin wrote:
> > Add sanity check to address the following concern:
> > 
> > During migration, all we pass the index of the request;
> > the rest can be re-read from the ring.
> > 
> > This is not generally enough if any available requests are outstanding.
> > Imagine a ring of size 4.  Below A means available U means used.
> > 
> > A 1
> > A 2
> > U 2
> > A 2
> > U 2
> > A 2
> > U 2
> > A 2
> > U 2
> > 
> > At this point available ring has wrapped around, the only
> > way to know head 1 is outstanding is because backend
> > has stored this info somewhere.
> > 
> > The reason we manage to migrate without tracking this in migration
> > state is because we flush outstanding requests before
> > migration.
> > This flush is device-specific though, let's add
> > a safeguard in virtio core to ensure it's done properly.
> > 
> > Signed-off-by: Michael S. Tsirkin <address@hidden>
> > ---
> > 
> > Changes from v1:
> >     v1 was against the wrong tree, it didn't build against qemu.git
> > 
> >  hw/virtio.c | 2 ++
> >  1 file changed, 2 insertions(+)
> 
> VirtIOBlock->rq can trigger the assertion.
> 
> IIUC hw/virtio-blk.c may handle I/O errors by keeping the request
> pending and on a list (->rq).  This allows the user to restart them
> after, for example, adding more space to the host file system containing
> the disk image file.
> 
> We keep a list of failed requests and we migrate this list.

Could not find it. It needs to be in virtio-blk in order
to know to update the used ring once it completes, right?

>  So I think
> inuse != 0 when migrating with pending failed I/O requests.
> 
> Stefan

Okay but let's make sure this is not a bug.