Re: [RFC PATCH] vfio-ccw: Permit missing IRQs

[Eric Farman]

On Wed, 2021-04-21 at 12:01 +0200, Cornelia Huck wrote:
> On Mon, 19 Apr 2021 20:49:06 +0200
> Eric Farman <farman@linux.ibm.com> wrote:
> 
> > Commit 690e29b91102 ("vfio-ccw: Refactor ccw irq handler") changed
> > one of the checks for the IRQ notifier registration from saying
> > "the host needs to recognize the only IRQ that exists" to saying
> > "the host needs to recognize ANY IRQ that exists."
> > 
> > And this worked fine, because the subsequent change to support the
> > CRW IRQ notifier doesn't get into this code when running on an
> > older
> > kernel, thanks to a guard by a capability region. The later
> > addition
> > of the REQ(uest) IRQ by commit b2f96f9e4f5f ("vfio-ccw: Connect the
> > device request notifier") broke this assumption because there is no
> > matching capability region. Thus, running new QEMU on an older
> > kernel fails with:
> > 
> >   vfio: unexpected number of irqs 2
> > 
> > Let's simply remove the check (and the less-than-helpful message),
> > and make the VFIO_DEVICE_GET_IRQ_INFO ioctl request for the IRQ
> > being processed. If it returns with EINVAL, we can treat it as
> > an unfortunate mismatch but not a fatal error for the guest.
> > 
> > Fixes: 690e29b91102 ("vfio-ccw: Refactor ccw irq handler")
> > Fixes: b2f96f9e4f5f ("vfio-ccw: Connect the device request
> > notifier")
> > Signed-off-by: Eric Farman <farman@linux.ibm.com>
> > ---
> >  hw/vfio/ccw.c | 15 +++++++--------
> >  1 file changed, 7 insertions(+), 8 deletions(-)
> > 
> > diff --git a/hw/vfio/ccw.c b/hw/vfio/ccw.c
> > index b2df708e4b..cfbfc3d1a2 100644
> > --- a/hw/vfio/ccw.c
> > +++ b/hw/vfio/ccw.c
> > @@ -411,20 +411,19 @@ static void
> > vfio_ccw_register_irq_notifier(VFIOCCWDevice *vcdev,
> >          return;
> >      }
> >  
> > -    if (vdev->num_irqs < irq + 1) {
> > -        error_setg(errp, "vfio: unexpected number of irqs %u",
> > -                   vdev->num_irqs);
> 
> Alternative proposal: Change this message to
> 
> "vfio: IRQ %u not available (number of irqs %u)"

> and still fail this function, while treating a failure of
> vfio_ccw_register_irq_notifier(vcdev, VFIO_CCW_REQ_IRQ_INDEX, &err);
> in
> vfio_ccw_realize() as a non-fatal error (maybe log a message).

This all sounds fine to me. I'll send a v2 as such.

> 
> This allows us to skip doing an ioctl call, of which we already know
> that it would fail. 

True, though as this is at the configuration time it's not as critical.

> Still, we can catch cases where a broken kernel e.g.
> provides the crw region, but not the matching irq (I believe
> something
> like that should indeed be a fatal error.)

Well they shouldn't do THAT. :)

> 
> > -        return;
> > -    }
> > -
> >      argsz = sizeof(*irq_info);
> >      irq_info = g_malloc0(argsz);
> >      irq_info->index = irq;
> >      irq_info->argsz = argsz;
> >      if (ioctl(vdev->fd, VFIO_DEVICE_GET_IRQ_INFO,
> >                irq_info) < 0 || irq_info->count < 1) {
> > -        error_setg_errno(errp, errno, "vfio: Error getting irq
> > info");
> > -        goto out_free_info;
> > +        if (errno == EINVAL) {
> > +            warn_report("Unable to get information about IRQ %u",
> > irq);
> > +            goto out_free_info;
> > +        } else {
> > +            error_setg_errno(errp, errno, "vfio: Error getting irq
> > info");
> > +            goto out_free_info;
> > +        }
> >      }
> >  
> >      if (event_notifier_init(notifier, 0)) {