Re: [PATCH v2 07/13] vfio/ccw: Convert to use vfio_register_group_dev()

[Cornelia Huck <cohuck@redhat.com>]

On Thu, 29 Apr 2021 15:13:47 -0300
Jason Gunthorpe <jgg@nvidia.com> wrote:

> On Thu, Apr 29, 2021 at 01:58:55PM +0200, Cornelia Huck wrote:
> 
> > > This seems like one of these cases where using the mdev GUID API
> > > was not a great fit. The ccs_driver should have just directly
> > > created a vfio_device and not gone into the mdev guid lifecycle
> > > world.  
> > 
> > I don't remember much of the discussion back then, but I don't think
> > the explicit generation of devices was the part we needed, but rather
> > some other kind of mediation -- probably iommu related, as subchannels
> > don't have that concept on their own. Anyway, too late to change now.  
> 
> The mdev part does three significant things:
>  - Provide a lifecycle model based on sysfs and the GUIDs
>  - Hackily inject itself into the VFIO IOMMU code as a special case
>  - Force the creation of a unique iommu group as the group FD is
>    mandatory to get the device FD.
> 
> This is why PASID is such a mess for mdev because it requires even
> more special hacky stuff to link up the dummy IOMMU but still operate
> within the iommu group of the parent device.
> 
> I can see an alternative arrangement using the /dev/ioasid idea that
> is a lot less hacky and does not force the mdev guid lifecycle on
> everyone that wants to create vfio_device.

I have not followed that discussion -- do you have a summary or a
pointer?

> 
> > > I almost did this, but couldn't figure out how the lifetime of the
> > > ccs_driver callbacks are working relative to the lifetime of the mdev
> > > device since they also reach into these structs. Maybe they can't be
> > > called for some css related reason?  
> > 
> > Moving allocations to the mdev driver probe makes sense, I guess. We
> > should also move enabling the subchannel to that point in time (I don't
> > remember why we enable it in the css probe function, and can't think of
> > a good reason for that; obviously needs to be paired with quiescing and
> > disabling the subchannel in the mdev driver remove function); that
> > leaves the uevent dance (which can hopefully also be removed, if some
> > discussed changes are implemented in the common I/O layer) and fencing
> > QDIO.
> > 
> > Regarding the other callbacks,
> > - vfio_ccw_sch_irq should not be invoked if the subchannel is not
> >   enabled; maybe log a message before returning for !private.
> > - vfio_ccw_sch_remove should be able to return 0 for !private (nothing
> >   to quiesce, if the subchannel is not enabled).
> > - vfio_ccw_sch_shutdown has nothing to do for !private (same reason.)
> > - In vfio_ccw_sch_event, we should either skip the fsm_event and the
> >   state change for !private, or return 0 in that case.
> > - vfio_ccw_chp_event already checks for !private. Not sure whether we
> >   should try to update some control blocks and return -ENODEV if the
> >   subchannel is not operational, but it's probably not needed.  
> 
> All the checks for !private need some kind of locking. The driver core
> model is that the 'struct device_driver' callbacks are all called
> under the device_lock (this prevents the driver unbinding during the
> callback). I didn't check if ccs does this or not..

probe/remove/shutdown are basically a forward of the callbacks at the
bus level. The css bus should make sure that we serialize
irq/sch_event/chp_event with probe/remove.

> 
> So if we NULL drvdata under the device_lock everything can be
> quite simple here.
> 
> Jason
>