| Message ID | 20240520152410.1098393-1-vidya.srinivas@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/i915/dpt: Make DPT object unshrinkable | expand |
On Mon, May 20, 2024 at 08:54:10PM +0530, Srinivas, Vidya wrote: > In some scenarios, the DPT object gets shrunk but > the actual framebuffer did not and thus its still > there on the DPT's vm->bound_list. Then it tries to > rewrite the PTEs via a stale CPU mapping. This causes panic. > > Credits-to: Ville Syrjala <ville.syrjala@linux.intel.com> > Shawn Lee <shawn.c.lee@intel.com> > > Signed-off-by: Srinivas, Vidya <vidya.srinivas@intel.com> The format should be "first_name last_name <email>" We also probably want Cc: stable@vger.kernel.org Fixes: 0dc987b699ce ("drm/i915/display: Add smem fallback allocation for dpt") Although the patch won't actually build unless we also have commit 779cb5ba64ec ("drm/i915/dpt: Treat the DPT BO as a framebuffer") but that hast the same fixes tag, so should be fine even if someone backports things that far back. > --- > drivers/gpu/drm/i915/gem/i915_gem_object.h | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h > index 3560a062d287..e6b485fc54d4 100644 > --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h > +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h > @@ -284,7 +284,8 @@ bool i915_gem_object_has_iomem(const struct drm_i915_gem_object *obj); > static inline bool > i915_gem_object_is_shrinkable(const struct drm_i915_gem_object *obj) > { Maybe toss something like this here: /* TODO: make DPT shrinkable when it has no bound vmas */ DPTs aren't necessarily so small that shrinking them wouldn't have any benefits. But actually implementing that would require some actual work, so not suitable for a quick fix. I can add all that stuff when applying the patch, no need to resend for this. > - return i915_gem_object_type_has(obj, I915_GEM_OBJECT_IS_SHRINKABLE); > + return i915_gem_object_type_has(obj, I915_GEM_OBJECT_IS_SHRINKABLE) && > + !obj->is_dpt; > } > > static inline bool > -- > 2.34.1
> -----Original Message----- > From: Ville Syrjälä <ville.syrjala@linux.intel.com> > Sent: Monday, May 20, 2024 10:10 PM > To: Srinivas, Vidya <vidya.srinivas@intel.com> > Cc: intel-gfx@lists.freedesktop.org; Syrjala, Ville <ville.syrjala@intel.com>; Lee, > Shawn C <shawn.c.lee@intel.com>; Srinivas@freedesktop.org > Subject: Re: [PATCH] drm/i915/dpt: Make DPT object unshrinkable > > On Mon, May 20, 2024 at 08:54:10PM +0530, Srinivas, Vidya wrote: > > In some scenarios, the DPT object gets shrunk but the actual > > framebuffer did not and thus its still there on the DPT's > > vm->bound_list. Then it tries to rewrite the PTEs via a stale CPU > > mapping. This causes panic. > > > > Credits-to: Ville Syrjala <ville.syrjala@linux.intel.com> > > Shawn Lee <shawn.c.lee@intel.com> > > > > Signed-off-by: Srinivas, Vidya <vidya.srinivas@intel.com> > > The format should be "first_name last_name <email>" Apologies for the mistake. My gitconfig got messed up. > > We also probably want > Cc: stable@vger.kernel.org > Fixes: 0dc987b699ce ("drm/i915/display: Add smem fallback allocation for > dpt") > Thank you so much. Will float new patch with this added. > Although the patch won't actually build unless we also have commit > 779cb5ba64ec ("drm/i915/dpt: Treat the DPT BO as a framebuffer") but that > hast the same fixes tag, so should be fine even if someone backports things > that far back. > > > --- > > drivers/gpu/drm/i915/gem/i915_gem_object.h | 3 ++- > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h > > b/drivers/gpu/drm/i915/gem/i915_gem_object.h > > index 3560a062d287..e6b485fc54d4 100644 > > --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h > > +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h > > @@ -284,7 +284,8 @@ bool i915_gem_object_has_iomem(const struct > > drm_i915_gem_object *obj); static inline bool > > i915_gem_object_is_shrinkable(const struct drm_i915_gem_object *obj) > > { > > Maybe toss something like this here: > /* TODO: make DPT shrinkable when it has no bound vmas */ > > DPTs aren't necessarily so small that shrinking them wouldn't have any > benefits. But actually implementing that would require some actual work, so > not suitable for a quick fix. > > I can add all that stuff when applying the patch, no need to resend for this. > > > - return i915_gem_object_type_has(obj, > I915_GEM_OBJECT_IS_SHRINKABLE); > > + return i915_gem_object_type_has(obj, > I915_GEM_OBJECT_IS_SHRINKABLE) && > > + !obj->is_dpt; > > } > > > > static inline bool > > -- > > 2.34.1 > > -- > Ville Syrjälä > Intel
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index 3560a062d287..e6b485fc54d4 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -284,7 +284,8 @@ bool i915_gem_object_has_iomem(const struct drm_i915_gem_object *obj); static inline bool i915_gem_object_is_shrinkable(const struct drm_i915_gem_object *obj) { - return i915_gem_object_type_has(obj, I915_GEM_OBJECT_IS_SHRINKABLE); + return i915_gem_object_type_has(obj, I915_GEM_OBJECT_IS_SHRINKABLE) && + !obj->is_dpt; } static inline bool
In some scenarios, the DPT object gets shrunk but the actual framebuffer did not and thus its still there on the DPT's vm->bound_list. Then it tries to rewrite the PTEs via a stale CPU mapping. This causes panic. Credits-to: Ville Syrjala <ville.syrjala@linux.intel.com> Shawn Lee <shawn.c.lee@intel.com> Signed-off-by: Srinivas, Vidya <vidya.srinivas@intel.com> --- drivers/gpu/drm/i915/gem/i915_gem_object.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)