Re: [PATCH v1] drm/xe/drm_ras: Move has_drm_ras check to drm_ras layer

[Raag Jadav <raag.jadav@intel.com>]

On Mon, Aug 03, 2026 at 11:03:11AM +0200, Michal Wajdeczko wrote:
> On 8/3/2026 10:08 AM, Raag Jadav wrote:
> > has_drm_ras flag is meant to facilitate drm_ras feature. Move it to the
> > correct layer where it belogs.
> 
> typo: belongs

Good catch.

> > Fixes: 63dfab5786ca ("drm/xe/xe_ras: Add drm_ras feature flag")
> > Signed-off-by: Raag Jadav <raag.jadav@intel.com>
> > ---
> >  drivers/gpu/drm/xe/xe_drm_ras.c | 3 +++
> >  drivers/gpu/drm/xe/xe_ras.c     | 3 ---
> >  2 files changed, 3 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_drm_ras.c b/drivers/gpu/drm/xe/xe_drm_ras.c
> > index 4afa2ad98300..a23cfb49361b 100644
> > --- a/drivers/gpu/drm/xe/xe_drm_ras.c
> > +++ b/drivers/gpu/drm/xe/xe_drm_ras.c
> > @@ -233,6 +233,9 @@ int xe_drm_ras_init(struct xe_device *xe)
> >  	struct drm_ras_node *node;
> >  	int err;
> >  
> > +	if (!xe->info.has_drm_ras)
> > +		return 0;
> > +
> >  	node = drmm_kcalloc(&xe->drm, DRM_XE_RAS_ERR_SEV_MAX, sizeof(*node), GFP_KERNEL);
> >  	if (!node)
> >  		return -ENOMEM;
> > diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
> > index 28123f08e513..f7af4da8172c 100644
> > --- a/drivers/gpu/drm/xe/xe_ras.c
> > +++ b/drivers/gpu/drm/xe/xe_ras.c
> > @@ -816,9 +816,6 @@ void xe_ras_init(struct xe_device *xe)
> >  {
> >  	int ret;
> >  
> > -	if (!xe->info.has_drm_ras)
> > -		return;
> > -
> >  	xe_drm_ras_init(xe);
> 
> btw, shouldn't we handle xe_drm_ras_init() errors here?
> 
> hmm, and it looks that xe_ras_init() ignores all errors,
> which is kind of against the "all-or-nothing" driver rule

The discussion[1] at the time was that we shouldn't let the driver suffer
for the failures of non-critical features.

I don't mind refactoring now, but this'll require some consensus.

[1] https://lore.kernel.org/intel-xe/ae9a0aee-d81b-4532-ab36-8025c50ae69a@intel.com/

Raag