Re: [PATCH v5 05/14] vfio/mdev: idxd: add basic mdev registration and helper functions

From: Jason Gunthorpe <jgg@nvidia.com>
To: Dave Jiang <dave.jiang@intel.com>
Cc: <alex.williamson@redhat.com>, <kwankhede@nvidia.com>,
	<tglx@linutronix.de>, <vkoul@kernel.org>, <megha.dey@intel.com>,
	<jacob.jun.pan@intel.com>, <ashok.raj@intel.com>,
	<yi.l.liu@intel.com>, <baolu.lu@intel.com>,
	<kevin.tian@intel.com>, <sanjay.k.kumar@intel.com>,
	<tony.luck@intel.com>, <dan.j.williams@intel.com>,
	<eric.auger@redhat.com>, <parav@mellanox.com>,
	<netanelg@mellanox.com>, <shahafs@mellanox.com>,
	<pbonzini@redhat.com>, <dmaengine@vger.kernel.org>,
	<linux-kernel@vger.kernel.org>, <kvm@vger.kernel.org>
Subject: Re: [PATCH v5 05/14] vfio/mdev: idxd: add basic mdev registration and helper functions
Date: Mon, 1 Mar 2021 20:50:46 -0400	[thread overview]
Message-ID: <20210302005046.GD4247@nvidia.com> (raw)
In-Reply-To: <3dd57aa0-3f64-d8cd-8f61-d91d7b1a1bdd@intel.com>

On Mon, Mar 01, 2021 at 05:48:00PM -0700, Dave Jiang wrote:
> 
> On 3/1/2021 5:29 PM, Jason Gunthorpe wrote:
> > On Mon, Mar 01, 2021 at 05:23:47PM -0700, Dave Jiang wrote:
> > > So after looking at the code in vfio_pci_intrs.c, I agree that the set_irqs
> > > code between VFIO_PCI and this driver can be made in common. Given that Alex
> > > doesn't want a vfio_pci device embedded in the driver,
> > idxd isn't a vfio_pci so it would be improper to do something like
> > that here anyhow.
> > 
> > > I think we'll need some sort of generic VFIO device that can be used
> > > from the vfio_pci side and vfio_mdev side to pass down in order to
> > > have common support library functions.
> > Why do you need more layers?
> > 
> > Just make some helper functions to manage this and build them into
> > their own struct and function family. All this needs is some callback
> > to for the end driver to hook in the raw device programming and some
> > entry points to direct the emulation access to the module.
> > 
> > It should be fully self contained and completely unrelated to vfio_pci
> > 
> Maybe I'm looking at this wrong. I see a some code in vfio_pci_intrs.c that
> we can reuse with some changes here and there. But, I think see where you
> are getting at with just common functions for mdev side. Let me create it
> just for IMS emulation and then we can go from there trying to figure if
> that's the right path to go down or if we need to share code with vfio_pci.

If it really is very common it could all be consolidated in a
vfio_utils.c kind of thing that all the places can use.

There is nothing wrong with splitting pieces of vfio_pci out.

Jason