Re: [Intel-gfx] [RFC PATCH] drm/i915: Remove all frontbuffer tracking calls from the gem code

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Thu, Jan 27, 2022 at 09:20:14AM +0100, Maarten Lankhorst wrote:
> Op 26-01-2022 om 14:09 schreef Jouni Högander:
> > We should now rely on userspace doing dirtyfb. There is no
> > need to have separate frontbuffer tracking hooks in gem code.
> >
> > This patch is removing all frontbuffer tracking calls from the gem
> > code.
> >
> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > Cc: Jani Nikula <jani.nikula@linux.intel.com>
> > Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
> > Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
> > ---
> >  drivers/gpu/drm/i915/display/intel_overlay.c |  2 --
> >  drivers/gpu/drm/i915/gem/i915_gem_clflush.c  |  2 --
> >  drivers/gpu/drm/i915/gem/i915_gem_domain.c   |  5 ----
> >  drivers/gpu/drm/i915/gem/i915_gem_object.c   | 24 --------------------
> >  drivers/gpu/drm/i915/gem/i915_gem_object.h   | 16 -------------
> >  drivers/gpu/drm/i915/gem/i915_gem_phys.c     |  7 ------
> >  drivers/gpu/drm/i915/i915_gem.c              |  5 ----
> >  7 files changed, 61 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
> > index 5358f03b52db..fc2691dac278 100644
> > --- a/drivers/gpu/drm/i915/display/intel_overlay.c
> > +++ b/drivers/gpu/drm/i915/display/intel_overlay.c
> > @@ -809,8 +809,6 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
> >  		goto out_pin_section;
> >  	}
> >  
> > -	i915_gem_object_flush_frontbuffer(new_bo, ORIGIN_DIRTYFB);
> > -
> >  	if (!overlay->active) {
> >  		const struct intel_crtc_state *crtc_state =
> >  			overlay->crtc->config;
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
> > index 8a248003dfae..115e6d877e38 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
> > @@ -20,8 +20,6 @@ static void __do_clflush(struct drm_i915_gem_object *obj)
> >  {
> >  	GEM_BUG_ON(!i915_gem_object_has_pages(obj));
> >  	drm_clflush_sg(obj->mm.pages);
> > -
> > -	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
> >  }
> >  
> >  static void clflush_work(struct dma_fence_work *base)
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
> > index 26532c07d467..ab1fc2d930e1 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
> > @@ -63,7 +63,6 @@ flush_write_domain(struct drm_i915_gem_object *obj, unsigned int flush_domains)
> >  		}
> >  		spin_unlock(&obj->vma.lock);
> >  
> > -		i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
> >  		break;
> >  
> >  	case I915_GEM_DOMAIN_WC:
> > @@ -615,9 +614,6 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
> >  out_unlock:
> >  	i915_gem_object_unlock(obj);
> >  
> > -	if (!err && write_domain)
> > -		i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
> > -
> >  out:
> >  	i915_gem_object_put(obj);
> >  	return err;
> > @@ -728,7 +724,6 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
> >  	}
> >  
> >  out:
> > -	i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
> >  	obj->mm.dirty = true;
> >  	/* return with the pages pinned */
> >  	return 0;
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
> > index 1a9e1f940a7d..f7ba66deb923 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
> > @@ -394,30 +394,6 @@ static void i915_gem_free_object(struct drm_gem_object *gem_obj)
> >  		queue_delayed_work(i915->wq, &i915->mm.free_work, 0);
> >  }
> >  
> > -void __i915_gem_object_flush_frontbuffer(struct drm_i915_gem_object *obj,
> > -					 enum fb_op_origin origin)
> > -{
> > -	struct intel_frontbuffer *front;
> > -
> > -	front = __intel_frontbuffer_get(obj);
> > -	if (front) {
> > -		intel_frontbuffer_flush(front, origin);
> > -		intel_frontbuffer_put(front);
> > -	}
> > -}
> > -
> > -void __i915_gem_object_invalidate_frontbuffer(struct drm_i915_gem_object *obj,
> > -					      enum fb_op_origin origin)
> > -{
> > -	struct intel_frontbuffer *front;
> > -
> > -	front = __intel_frontbuffer_get(obj);
> > -	if (front) {
> > -		intel_frontbuffer_invalidate(front, origin);
> > -		intel_frontbuffer_put(front);
> > -	}
> > -}
> > -
> >  static void
> >  i915_gem_object_read_from_page_kmap(struct drm_i915_gem_object *obj, u64 offset, void *dst, int size)
> >  {
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
> > index 02c37fe4a535..d7a08172b239 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
> > @@ -578,22 +578,6 @@ void __i915_gem_object_flush_frontbuffer(struct drm_i915_gem_object *obj,
> >  void __i915_gem_object_invalidate_frontbuffer(struct drm_i915_gem_object *obj,
> >  					      enum fb_op_origin origin);
> >  
> > -static inline void
> > -i915_gem_object_flush_frontbuffer(struct drm_i915_gem_object *obj,
> > -				  enum fb_op_origin origin)
> > -{
> > -	if (unlikely(rcu_access_pointer(obj->frontbuffer)))
> > -		__i915_gem_object_flush_frontbuffer(obj, origin);
> > -}
> > -
> > -static inline void
> > -i915_gem_object_invalidate_frontbuffer(struct drm_i915_gem_object *obj,
> > -				       enum fb_op_origin origin)
> > -{
> > -	if (unlikely(rcu_access_pointer(obj->frontbuffer)))
> > -		__i915_gem_object_invalidate_frontbuffer(obj, origin);
> > -}
> > -
> >  int i915_gem_object_read_from_page(struct drm_i915_gem_object *obj, u64 offset, void *dst, int size);
> >  
> >  bool i915_gem_object_is_shmem(const struct drm_i915_gem_object *obj);
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
> > index ca6faffcc496..e98a9884cf5a 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_phys.c
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
> > @@ -151,19 +151,12 @@ int i915_gem_object_pwrite_phys(struct drm_i915_gem_object *obj,
> >  	if (err)
> >  		return err;
> >  
> > -	/*
> > -	 * We manually control the domain here and pretend that it
> > -	 * remains coherent i.e. in the GTT domain, like shmem_pwrite.
> > -	 */
> > -	i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
> > -
> >  	if (copy_from_user(vaddr, user_data, args->size))
> >  		return -EFAULT;
> >  
> >  	drm_clflush_virt_range(vaddr, args->size);
> >  	intel_gt_chipset_flush(to_gt(i915));
> >  
> > -	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
> >  	return 0;
> >  }
> >  
> > diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> > index e3a2c2a0e156..60838209f9cd 100644
> > --- a/drivers/gpu/drm/i915/i915_gem.c
> > +++ b/drivers/gpu/drm/i915/i915_gem.c
> > @@ -563,8 +563,6 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
> >  		goto out_rpm;
> >  	}
> >  
> > -	i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
> > -
> >  	user_data = u64_to_user_ptr(args->data_ptr);
> >  	offset = args->offset;
> >  	remain = args->size;
> > @@ -607,7 +605,6 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
> >  	}
> >  
> >  	intel_gt_flush_ggtt_writes(ggtt->vm.gt);
> > -	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
> >  
> >  	i915_gem_gtt_cleanup(obj, &node, vma);
> >  out_rpm:
> > @@ -694,8 +691,6 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
> >  		offset = 0;
> >  	}
> >  
> > -	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
> > -
> >  	i915_gem_object_unpin_pages(obj);
> >  	return ret;
> >  
> 
> We should get rid of frontbuffer tracking completely still, this should definitely happen. I've looked at it before, but at that the time we didn't do it yet. Mostly out of concerns of breaking old userspace.
> 
> The people you cc'd are not part of the cc here. I added them.
> 
> I see i915_pm_dc failing on dc5-psr, something to look into?
> 
> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> 
> With the caveat that you wil need someone else to ack the changes, as the last time I proposed this, there was pushback out of fear of breaking userspace.

I think at the very least we'd need something along the lines of
https://patchwork.freedesktop.org/series/70636/
to make intel ddx use dirtyfb when needed, and a bit of time to
let it spread before we pull the trigger.

Though the status quo is already broken (maybe with tearfree enabled
it works, not sure) if you enable fbc since wc mmaps don't trip the
hw gtt tracking and so you get all kinds of stale junk on screen.
So dirtyfb also very much needed before I can enable fbc across the
board (which I think we could now do otherwise).

-- 
Ville Syrjälä
Intel