Re: [PATCH v4 06/12] virtio: blk: Add freeze, restore handlers to support S4

["Michael S. Tsirkin" <mst@redhat.com>]

On Wed, Dec 07, 2011 at 04:26:47PM +0530, Amit Shah wrote:
> On (Wed) 07 Dec 2011 [12:37:02], Michael S. Tsirkin wrote:
> > On Wed, Dec 07, 2011 at 01:18:44AM +0530, Amit Shah wrote:
> > > Delete the vq and flush any pending requests from the block queue on the
> > > freeze callback to prepare for hibernation.
> > > 
> > > Re-create the vq in the restore callback to resume normal function.
> > > 
> > > Signed-off-by: Amit Shah <amit.shah@redhat.com>
> > > ---
> > >  drivers/block/virtio_blk.c |   38 ++++++++++++++++++++++++++++++++++++++
> > >  1 files changed, 38 insertions(+), 0 deletions(-)
> > > 
> > > diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> > > index 467f218..a9147a6 100644
> > > --- a/drivers/block/virtio_blk.c
> > > +++ b/drivers/block/virtio_blk.c
> > > @@ -568,6 +568,40 @@ static void __devexit virtblk_remove(struct virtio_device *vdev)
> > >  	ida_simple_remove(&vd_index_ida, index);
> > >  }
> > >  
> > > +#ifdef CONFIG_PM
> > > +static int virtblk_freeze(struct virtio_device *vdev)
> > > +{
> > > +	struct virtio_blk *vblk = vdev->priv;
> > > +
> > > +	/* Ensure we don't receive any more interrupts */
> > > +	vdev->config->reset(vdev);
> > > +
> > > +	flush_work(&vblk->config_work);
> > 
> > It bothers me that config work can be running
> > after reset here. If it does, it will not get sane
> > values from reading config.
> 
> Why so?
> 
> The reset only ensures the host doesn't write anything more, isn't it?
> Why would the values be affected?

Generally, not only that. Reset also clears configuration to the
reset value :) As since accesses are done byte
by byte you might get a value that is different from
*both* old and new one as a result.

But that is a general comment, specifically for block,
I don't know if there is a problem with this.

Same for console.

> > Also, can there be stuff in the reqs list?
> > If yes is this a problem?
> 
> Should be all cleared by the two commands below.  At least that's my
> expectation.  If not, let me know!
> 
> > > +	spin_lock_irq(vblk->disk->queue->queue_lock);
> > > +	blk_stop_queue(vblk->disk->queue);
> > > +	spin_unlock_irq(vblk->disk->queue->queue_lock);
> > > +	blk_sync_queue(vblk->disk->queue);
> > > +
> > > +	vdev->config->del_vqs(vdev);
> > > +	return 0;
> > > +}
> > > +
> > 
> > Thinking about it, looks like there's a bug in
> > virtblk_remove: if we get a config change after
> > flush_work we schedule another work.
> > That's a problem for sure as structure is removed.
> 
> Yep, it is a potential issue.
> 
> 		Amit

Sent a patch.