Re: [PATCH v3 00/18] Add VFIO mediated device support and DEV-MSI support for the idxd driver

From: Alex Williamson
Date: Thu Sep 17 2020 - 13:32:50 EST

On Thu, 17 Sep 2020 10:15:24 -0700
Dave Jiang <dave.jiang@xxxxxxxxx> wrote:

> On 9/17/2020 8:06 AM, Jason Gunthorpe wrote:
> > On Tue, Sep 15, 2020 at 04:27:35PM -0700, Dave Jiang wrote:
> >> drivers/dma/idxd/idxd.h | 65 +
> >> drivers/dma/idxd/init.c | 100 ++
> >> drivers/dma/idxd/irq.c | 6
> >> drivers/dma/idxd/mdev.c | 1089 ++++++++++++++++++++
> >> drivers/dma/idxd/mdev.h | 118 ++
> >
> > It is common that drivers of a subsystem will be under that
> > subsystem's directory tree. This allows the subsystem community to
> > manage pages related to their subsystem and it's drivers.
> >
> > Should the mdev parts be moved there?
> I personally don't have a preference. I'll defer to Alex or Kirti to provide
> that guidance. It may make certains things like dealing with dma fault regions
> and etc easier using vfio calls from vfio_pci_private.h later on for vSVM
> support. It also may be the better code review and maintenance domain and
> alleviate Vinod having to deal with that portion since it's not dmaengine domain.

TBH, I'd expect an mdev driver to be co-located with the remainder of
its parent driver. It's hopefully sharing more code with that driver
than anything in mdev of vfio (either of which should be exported for
the vendor driver). mdev support is largely expected to be a feature of
a driver, much like it is for i915, not necessarily its sole purpose for
existing (see the rejected fpga mdev driver). Also being nearby
vfio_pci_private shouldn't invite anyone other than vfio_pci to make
use of that header. Thanks,