Re: [PATCH 0/2] s390/vfio-ap: do not open code locks for VFIO_GROUP_NOTIFY_SET_KVM notification

[Tony Krowiak]

On 8/19/21 11:30 AM, Cornelia Huck wrote:
> On Wed, Aug 18 2021, Alex Williamson <alex.williamson@xxxxxxxxxx> wrote:
>
> > On Wed, 18 Aug 2021 17:59:51 +0200
> > Christian Borntraeger <borntraeger@xxxxxxxxxx> wrote:
> >
> > > On 02.08.21 18:32, Tony Krowiak wrote:
> > > > On 8/2/21 9:53 AM, Halil Pasic wrote:
> > > > > On Mon, 2 Aug 2021 09:10:26 -0400
> > > > > Tony Krowiak <akrowiak@xxxxxxxxxxxxx> wrote:
> > > > >   
> > > > > > PING!
> > > > > >
> > > > > > This patch will pre-req version 17 of a patch series I have waiting in
> > > > > > the wings,
> > > > > > so I'd like to get this one merged ASAP. In particular, if a KVM
> > > > > > maintainer can
> > > > > > take a look at the comments concerning the taking of the kvm->lock
> > > > > > before the
> > > > > > matrix_mdev->lock it would be greatly appreciated. Those comments begin with
> > > > > > Message ID <20210727004329.3bcc7d4f.pasic@xxxxxxxxxxxxx> from Halil Pasic.
> > > > > As far as I'm concerned, we can move forward with this. Was this
> > > > > supposed to go in via Alex's tree?
> > > > I am not certain, Christian queued the previous patches related to
> > > > this on:
> > > >
> > > >
> > > > https://git.kernel.org/pub/scm/linux/kernel/git/s390/linux.git/log/?h=fixes
> > > >
> > > > Jason G., since this will need to be integrated with your other patches,
> > > > where should this be queued?
> > >
> > > This previous patch (s390/vfio-ap: clean up mdev resources when remove callback invoked) is
> > > already in master.
> > > Can you respin the series with all Acks and RBs?
> > >
> > > Alex, can you then take these 2 patches via your tree? Thanks
> > >
> > > Acked-by: Christian Borntraeger <borntraeger@xxxxxxxxxx>
> > > for this series.
> >
> > I see some review feedback that seems to suggest a new version would be
> > posted:
> >
> > https://lore.kernel.org/linux-s390/0f03ab0b-2dfd-e1c1-fe43-be2a59030a71@xxxxxxxxxxxxx/
> Yeah, I thought so as well. But it also looks like something that could
> be a fixup on top.

I will post the new patch today. I was waiting for the remainder of
the feedback and frankly forgot to post the patch incorporating
the changes precipitated by the previous comments.

> > I also see in this thread:
> >
> > https://lore.kernel.org/linux-s390/20210721164550.5402fe1c.pasic@xxxxxxxxxxxxx/
> >
> > that Halil's concern's around open/close races are addressed by Jason's
> > device_open/close series that's already in my next branch and he
> > provided an Ack, but there's still the above question regarding the
> > kvm->lock that was looking for a review from... I'm not sure, maybe
> > Connie or Paolo.  Christian, is this specifically what you're ack'ing?
> I'm also unsure about the kvm->lock thing. Is taking the lock buried
> somewhere deep in the code that will ultimately trigger the release?
> I would at least like a pointer.

I'm not quite sure what you're asking here, but if you follow the
thread starting with the link above it may reveal the answer to
what you are asking here.


> > It can ultimately go in through my tree, but not being familiar with
> > this code I'd hope for more closure.  Thanks,
> >
> > Alex