Re: [PATCH 2/6] xen-blkfront: handle backend CLOSED without CLOSING

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Fri, Oct 05, 2012 at 04:57:36PM +0100, David Vrabel wrote:
> On 05/10/12 12:55, Jan Beulich wrote:
> >>>> On 05.10.12 at 13:42, David Vrabel <david.vrabel@citrix.com> wrote:
> >> On 02/10/12 21:02, Konrad Rzeszutek Wilk wrote:
> >>> On Mon, Oct 01, 2012 at 06:19:19PM +0100, David Vrabel wrote:
> >>>> On 25/09/12 18:53, David Vrabel wrote:
> >>>>> On 21/09/12 17:04, David Vrabel wrote:
> >>>>>> From: David Vrabel <david.vrabel@citrix.com>
> >>>>>>
> >>>>>> Backend drivers shouldn't transistion to CLOSED unless the frontend is
> >>>>>> CLOSED.  If a backend does transition to CLOSED too soon then the
> >>>>>> frontend may not see the CLOSING state and will not properly shutdown.
> >>>>>>
> >>>>>> So, treat an unexpected backend CLOSED state the same as CLOSING.
> >>>>>
> >>>>> Didn't handle the frontend block device being mounted. Updated patch here.
> >>>>
> >>>> Konrad, can you ack this updated patch if you're happy with it.
> >>>
> >>> Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> >>>
> >>> Or should I just carry it in my for-jens-3.7 bug-fixes queue and ask
> >>> Jen to pull it once rc0 is out?
> >>
> >> This seems easiest, if Jan is happy with the patch.
> > 
> > I see the point of your explanation yesterday, but am still not
> > convinced that the early cleanup in spite of active users of the
> > disk can't cause any problems (like dangling pointers or NULL
> > dereferences).
> 
> I tested it some more and it is broken.  Please apply the first version
> instead.

OK, can we do this once more - can you repost the patches, but CC the
invidiual maintainers? The FB and KBD need to go through Florian I think?

> 
> blkfront needs to cancel and complete with an error requests that are
> are still on the ring after the backend has disconnected and it needs to
> complete with an error all requests submitted while disconnected.
> Otherwise, if there is outstanding requests or new requests then
> del_gendisk() blocks waiting for the I/O to complete.
> 
> David