Re: [RFC PATCH] s390: vfio-ap: Register the vfio_ap module for the "ap" parent bus

[Thomas Huth]

On 14/12/2021 22.55, Tony Krowiak wrote:
> On 12/13/21 11:11, Cornelia Huck wrote:
> > On Mon, Dec 13 2021, Harald Freudenberger <freude@xxxxxxxxxxxxx> wrote:
> >
> > > On 01.12.21 15:11, Thomas Huth wrote:
> > > > The crypto devices that we can use with the vfio_ap module are sitting
> > > > on the "ap" bus, not on the "vfio_ap" bus that the module defines
> > > > itself. With this change, the vfio_ap module now gets automatically
> > > > loaded if a supported crypto adapter is available in the host.
> > > >
> > > > Signed-off-by: Thomas Huth <thuth@xxxxxxxxxx>
> > > > ---
> > > >   Note: Marked as "RFC" since I'm not 100% sure about it ...
> > > >         please review carefully!
> > > >
> > > >   drivers/s390/crypto/vfio_ap_drv.c | 2 +-
> > > >   1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/s390/crypto/vfio_ap_drv.c 
> > > > b/drivers/s390/crypto/vfio_ap_drv.c
> > > > index 4d2556bc7fe5..5580e40608a4 100644
> > > > --- a/drivers/s390/crypto/vfio_ap_drv.c
> > > > +++ b/drivers/s390/crypto/vfio_ap_drv.c
> > > > @@ -39,7 +39,7 @@ static struct ap_device_id ap_queue_ids[] = {
> > > >       { /* end of sibling */ },
> > > >   };
> > > > -MODULE_DEVICE_TABLE(vfio_ap, ap_queue_ids);
> > > > +MODULE_DEVICE_TABLE(ap, ap_queue_ids);
> > > >   /**
> > > >    * vfio_ap_queue_dev_probe:
> > > I had a chance to check this now.
> > > First I have to apologize about the dispute with vfio devices appearing 
> > > on the ap bus.
> > > That's not the case with this patch. As Connie states the 
> > > MODULE_DEVICE_TABLE() does not
> > > change the parent of a device and vfio_ap_drv is a driver for ap devices 
> > > and thus
> > > belongs to the ap bus anyway.
> > > So what's left is that with this change the vfio_ap kernel module is 
> > > automatically loaded
> > > when an ap device type 10-13 is recognized by the ap bus. So the 
> > > intention of the patch
> > > is fulfilled.
> > > Yet another kernel module which may occupy memory but will never get used 
> > > by most customers.
> > > This may not be a problem but I had a glance at the list of kernel 
> > > modules loaded on my
> > > LPAR with and without the patch and the difference is:
> > > ...
> > > kvm                   512000  1 vfio_ap
> > > vfio_ap                28672  0
> > > ...
> > > So the vfio_ap module has a dependency to the biggest kernel module ever 
> > > - kvm.
> > > Do I need to say something more?
> > >
> > > If this dependency is removed then I would not hesitate to accept this 
> > > patch. However
> > > this is up to Tony as he is the maintainer of the vfio ap device driver.
> > I don't think you can drop the kvm reference, as the code in vfio-ap
> > obviously depends on it...
> >
> > One possibility is simply blocking autoload of the module in userspace by
> > default, and only allow it to be loaded automatically when e.g. qemu-kvm
> > is installed on the system. This is obviously something that needs to be
> > decided by the distros.
> >
> > (kvm might actually be autoloaded already, so autoloading vfio-ap would
> > not really make it worse.)
>
> Of the vfio_ccw module is automatically loaded, then the kvm
> module will also get loaded. I startup up a RHEL8.3 system and
> sure enough, the vfio_ccw module is loaded along with the
> kvm, vfio and mdev modules. If this is true for all distros, then
> it wouldn't make much difference if the vfio_ap module is
> autoloaded too.

I think I don't mind too much if we auto-load vfio-ap or not - but I think 
we should make it consistent with vfio-ccw. So either auto-load both modules 
(if the corresponding devices are available), or remove the 
MODULE_DEVICE_TABLE() entries from both modules?

 Thomas