Re: [Qemu-devel] [PATCH v14 06/14] block: Add backing_blocker in BlockDriverState

[Jeff Cody <jcody@redhat.com>]

On Thu, Feb 20, 2014 at 04:28:56PM +0800, Fam Zheng wrote:
> On Thu, 02/20 00:08, Jeff Cody wrote:
> > On Thu, Feb 20, 2014 at 01:01:38PM +0800, Fam Zheng wrote:
> > > On Wed, 02/19 16:17, Jeff Cody wrote:
> > > > On Wed, Feb 19, 2014 at 09:42:23PM +0800, Fam Zheng wrote:
> > > > > This makes use of op_blocker and blocks all the operations except for
> > > > > commit target, on each BlockDriverState->backing_hd.
> > > > > 
> > > > > The asserts for op_blocker in bdrv_swap are removed because with this
> > > > > change, the target of block commit has at least the backing blocker of
> > > > > its child, so the assertion is not true. Callers should do their check.
> > > > > 
> > > > > Signed-off-by: Fam Zheng <famz@redhat.com>
> > > > > ---
> > > > >  block.c                   | 19 +++++++++++++++----
> > > > >  include/block/block_int.h |  3 +++
> > > > >  2 files changed, 18 insertions(+), 4 deletions(-)
> > > > > 
> > > > > diff --git a/block.c b/block.c
> > > > > index dec44d4..95d8c1f 100644
> > > > > --- a/block.c
> > > > > +++ b/block.c
> > > > > @@ -1044,19 +1044,33 @@ fail:
> > > > >  void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd)
> > > > >  {
> > > > >      if (bs->backing_hd) {
> > > > > +        assert(error_is_set(&bs->backing_blocker));
> > > > 
> > > > When I run block-commit, on either the active or non-active layer, I
> > > > get an assertion here.  The qemu-iotests do not catch it, and I
> > > > presume it is because happens a couple of seconds or so after the
> > > > success message is returned over QMP.
> > > > 
> > > 
> > > I can't reproduce this, could you give some specific steps? Thanks.
> > >
> > 
> > Sure - I am guessing the key is performing some live block snapshots
> > first.  Here is what I did (this is from memory, but I think the steps
> > are right):
> > 
> > Nothing special really about the cmdline:
> > qemu-system-x86_64 -drive file=/home/jtc/test.qcow2,if=virtio -qmp stdio ...
> > 
> > The QMP commands:
> > 
> > For the non-active layer case:
> > 
> > { "execute": "qmp_capabilities" }
> > { "execute": "blockdev-snapshot-sync", "arguments": { "device": "virtio0","snapshot-file":"/tmp/snap1.qcow2","format": "qcow2" } }
> > { "execute": "blockdev-snapshot-sync", "arguments": { "device": "virtio0","snapshot-file":"/tmp/snap2.qcow2","format": "qcow2" } }
> > { "execute": "block-commit", "arguments": { "device": "virtio0", "top": "/tmp/snap1.qcow2" } }
> > 
> > 
> > For the active layer case (I think I still had 2 snapshots here, not
> > entirely positive):
> > 
> > { "execute": "qmp_capabilities" }
> > { "execute": "blockdev-snapshot-sync", "arguments": { "device": "virtio0","snapshot-file":"/tmp/snap1.qcow2","format": "qcow2" } }
> > { "execute": "blockdev-snapshot-sync", "arguments": { "device": "virtio0","snapshot-file":"/tmp/snap2.qcow2","format": "qcow2" } }
> > { "execute": "block-commit", "arguments": { "device": "virtio0", "top": "/tmp/snap2.qcow2" } }
> > { "execute": "block-job-complete", "arguments": { "device": "virtio0" }}
> > 
> 
> Yes. I forgot to use bdrv_set_backing_hd in bdrv_append.
> 
> Could you try if the below patch fixes it? Thanks.
> 

Yep, this seems to fix it.

> Fam
> 
> ---
> 
> diff --git a/block.c b/block.c
> index 1af43b9..66a8e35 100644
> --- a/block.c
> +++ b/block.c
> @@ -1978,7 +1978,7 @@ void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top)
> 
>      /* The contents of 'tmp' will become bs_top, as we are
>       * swapping bs_new and bs_top contents. */
> -    bs_top->backing_hd = bs_new;
> +    bdrv_set_backing_hd(bs_top, bs_new);
>      bs_top->open_flags &= ~BDRV_O_NO_BACKING;
>      pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file),
>              bs_new->filename);