[Intel-gfx] [PATCH 15/15] vfio: Add struct device to vfio_device

[Tian, Kevin]

> From: Alex Williamson <alex.williamson at redhat.com>
> Sent: Thursday, September 1, 2022 1:15 AM
> 
> On Wed, 31 Aug 2022 06:10:51 +0000
> "Tian, Kevin" <kevin.tian at intel.com> wrote:
> 
> > > From: Jason Gunthorpe <jgg at ziepe.ca>
> > > Sent: Wednesday, August 31, 2022 7:53 AM
> > >
> > > On Tue, Aug 30, 2022 at 04:18:38PM -0600, Alex Williamson wrote:
> > > > On Sun, 28 Aug 2022 01:10:37 +0800
> > > > Kevin Tian <kevin.tian at intel.com> wrote:
> > > >
> > > > > From: Yi Liu <yi.l.liu at intel.com>
> > > > >
> > > > > and replace kref. With it a 'vfio-dev/vfioX' node is created under the
> > > > > sysfs path of the parent, indicating the device is bound to a vfio
> > > > > driver, e.g.:
> > > > >
> > > > > /sys/devices/pci0000\:6f/0000\:6f\:01.0/vfio-dev/vfio0
> > > > >
> > > > > It is also a preparatory step toward adding cdev for supporting future
> > > > > device-oriented uAPI.
> > > >
> > > > Shall we start Documentation/ABI/testing/vfio-dev now?  Thanks.
> > >
> > > I always thought that was something to use when adding new custom
> > > sysfs attributes?
> > >
> > > Here we are just creating a standard struct device with its standard
> > > sysfs?
> > >
> >
> > There is nothing special for vfio-dev/vfioX. But from pci device p.o.v
> > this does introduce a custom node in the directory, which is probably
> > what Alex referred to?
> 
> Yup, but not just for pci, we're adding a node into the device
> directory for any device bound to vfio.
> 
> > Anyway if required following can be introduced:
> >
> > diff --git a/Documentation/ABI/testing/sysfs-devices-vfio-dev
> b/Documentation/ABI/testing/sysfs-devices-vfio-dev
> > new file mode 100644
> > index 000000000000..dfe8baaf1ccb
> > --- /dev/null
> > +++ b/Documentation/ABI/testing/sysfs-devices-vfio-dev
> > @@ -0,0 +1,8 @@
> > +What:		 /sys/.../<device>/vfio-dev/vfioX/
> > +Date:		 September 2022
> > +Contact:	 Yi Liu <yi.l.liu at intel.com>
> > +Description:
> > +		 This directory is created when the device is bound to a
> > +		 vfio driver. The layout under this directory matches what
> > +		 exists for a standard 'struct device'. 'X' is a random
> > +		 number marking this device in vfio.
> 
> It's not really random, it's a unique index.  Seems like a good
> starting point.
> 
> >
> > At the start I thought it might make more sense to add it into an
> > existing vfio ABI file. But looks it doesn't exist.
> >
> > Curious why nobody asked for ABI doc for /dev/vfio/vfio, /sys/class/vfio,
> etc...
> 
> Oversight, there should probably be a sysfs-class-vfio file.  Thanks,
> 

I can help add one.

btw I plan to respin v2 tomorrow. Regarding to this ABI thing there are
three options:

1) Just add sysfs-devices-vfio-dev in this series. Later merge to
   sysfs-class-vfio once the latter is introduced in a separate patch.

2) Do sysfs-class-vfio in this series, including both existing vfio ABIs and
   the new vfio-dev.

3) No ABI file in this series. Handle it in a separate patch with
   sysfs-class-vfio.

Which one do  you prefer to?

Thanks
Kevin