[Mesa-dev] [Intel-gfx] [PATCH i-g-t] lib/i915_pciids.h: synchronize with kernel header

[Rodrigo Vivi]

Hi all,

restarting a very old thread since Dmitry is interested in tackle
the issue that we have to propagate and keep all IDs in sync across
the stack.

He has few ideas of i915 ioctls in a way that it wouldn't necessarily
tie build dependencies etc..

Since I believe this place is the best one to discuss ideas like that
let me invite him for the discussion in a way he can explain his
ideas here.

Also cc'ing some relevant mailing lists.

Thanks,
Rodrigo.

On Wed, Dec 20, 2017 at 12:06:24PM -0800, De Marchi, Lucas wrote:
> On Tue, 2017-12-19 at 15:28 -0800, Rodrigo Vivi wrote:
> > On Tue, Dec 19, 2017 at 10:07:30PM +0000, Chris Wilson wrote:
> > > Quoting Rodrigo Vivi (2017-12-19 21:59:43)
> > > > On Fri, Dec 08, 2017 at 10:06:46PM +0000, Lucas De Marchi wrote:
> > > > > This copies include/drm/i915_pciids.h from kernel as of drm-tip:
> > > > > drm-tip: 2017y-12m-08d-21h-06m-35s UTC + patch adding INTEL_CFL_IDS
> > > > > that
> > > > > was missing there[1].
> > > > 
> > > > Since this tip name is not easily found maybe it would be good to
> > > > mention latest kernel commit that touched this file.
> > > > 
> > > > > The goal is to keep track of the PCI IDs in a
> > > > > single place (kernel).
> > > > 
> > > > good idea.
> > > > 
> > > > > 
> > > > > Right now a simple copy is done to catch up with latest changes there,
> > > > > although in future it could be more sofisticated pointing the build
> > > > > system to the external header.
> > > > 
> > > > Yeap, a real single place would be awesome.
> > > 
> > > Whilst maintaining independence of the igt build itself?
> > 
> > yeap... nevermind...
> > I was just in another discussion about detection of cnl on vaapi driver
> > and I realized that the userspace individual detection has its advantages...
> 
> I get the benefit of not tying i-g-t to particular kernel versions, but we may
> need to think in a way to alleviate the tedious work of updating the IDs
> across the projects.
> 
> Some alternatives:
> 
> 1) Move header to e.g. include/external/i915_pciids.h and add a
> scripts/i915-update-ids that receives the kernel directory as argument.  The
> script does whateve is needed to update the header, which may or may not
> include adding a verbiage on top of the header
> 
> 	/* This header is automatically generated, do not edit it */
> 
> 2) Just create a policy that changes to this header needs to actually be a
> copy from the kernel header with the commit hash in the commit message
> 
> 3) A mix or partial implementation of (1) and (2).
> 
> What do you think?
> 
> Lucas De Marchi