linux.git/drivers/gpu/drm/msm/msm_atomic.c, branch v6.0

drm/msm: Avoid dirtyfb stalls on video mode displays (v2)

2022-02-25T15:59:58+00:00

Someone on IRC once asked an innocent enough sounding question:  Why
with xf86-video-modesetting is es2gears limited at 120fps.

So I broke out the perfetto tracing mesa MR and took a look.  It turns
out the problem was drm_atomic_helper_dirtyfb(), which would end up
waiting for vblank.. es2gears would rapidly push two frames to Xorg,
which would blit them to screen and in idle hook (I assume) call the
DIRTYFB ioctl.  Which in turn would do an atomic update to flush the
dirty rects, which would stall until the next vblank.  And then the
whole process would repeat.

But this is a bit silly, we only need dirtyfb for command mode DSI
panels.  So track in plane state whether dirtyfb is required, and
track in the fb how many attached planes require dirtyfb so that we
can skip it when not required.  (Note, mdp4 does not actually have
cmd mode support.)

Signed-off-by: Rob Clark 
Reviewed-by: Dmitry Baryshkov 
Link: https://lore.kernel.org/r/20220223191118.881321-1-robdclark@gmail.com
Signed-off-by: Rob Clark

drm/msm: Add hrtimer + kthread_work helper

2021-10-15T23:35:39+00:00

Before open-coding this a 2nd time, add a helper.

Signed-off-by: Rob Clark 
Link: https://lore.kernel.org/r/20210927230455.1066297-1-robdclark@gmail.com
Signed-off-by: Rob Clark

Merge drm/drm-fixes into drm-next

2021-04-13T21:15:09+00:00

msm-next pull request has a baseline with stuff from -fixes, roll
forward first.

Some simple conflicts in amdgpu, ttm and one in i915 where git gets
confused and tries to add the same function twice.

Signed-off-by: Daniel Vetter

drm/msm/kms: Use nested locking for crtc lock instead of custom classes

2021-02-23T20:39:29+00:00

We don't need to make up custom lock classes here, we can simply use
mutex_lock_nested() and pass in the index of the crtc to the locking
APIs instead. This helps lockdep understand that these are really
different locks while avoiding having to allocate custom lockdep
classes.

Cc: Krishna Manikandan 
Suggested-by: Daniel Vetter 
Fixes: b3d91800d9ac ("drm/msm: Fix race condition in msm driver with async layer updates")
Signed-off-by: Stephen Boyd 
Signed-off-by: Rob Clark

drm/gem: Move drm_gem_fb_prepare_fb() to GEM atomic helpers

2021-02-23T07:54:22+00:00

The function drm_gem_fb_prepare_fb() is a helper for atomic modesetting,
but currently located next to framebuffer helpers. Move it to GEM atomic
helpers, rename it slightly and adopt the drivers. Same for the rsp
simple-pipe helper.

Compile-tested with x86-64, aarch64 and arm. The patch is fairly large,
but there are no functional changes.

v3:
	* remove out-comented line in drm_gem_framebuffer_helper.h
	  (Maxime)
v2:
	* rename to drm_gem_plane_helper_prepare_fb() (Daniel)
	* add tutorial-style documentation

Signed-off-by: Thomas Zimmermann 
Acked-by: Maxime Ripard 
Link: https://patchwork.freedesktop.org/patch/msgid/20210222141756.7864-1-tzimmermann@suse.de

drm/msm/atomic: Convert to per-CRTC kthread_work

2020-11-01T18:13:32+00:00

Use a SCHED_FIFO kthread_worker for async atomic commits.  We have a
hard deadline if we don't want to miss a frame.

Signed-off-by: Rob Clark

drm/msm/atomic: Drop per-CRTC locks in reverse order

2020-11-01T18:13:09+00:00

lockdep dislikes seeing locks unwound in a non-nested fashion.

Fixes: b3d91800d9ac ("drm/msm: Fix race condition in msm driver with async layer updates")
Signed-off-by: Rob Clark 
Reviewed-by: Abhinav Kumar

drm/msm: Fix race condition in msm driver with async layer updates

2020-11-01T18:11:59+00:00

When there are back to back commits with async cursor update,
there is a case where second commit can program the DPU hw
blocks while first didn't complete flushing config to HW.

Synchronize the compositions such that second commit waits
until first commit flushes the composition.

This change also introduces per crtc commit lock, such that
commits on different crtcs are not blocked by each other.

Changes in v2:
	- Use an array of mutexes in kms to handle commit
	  lock per crtc. (Rob Clark)

Changes in v3:
	- Add wrapper functions to handle lock and unlock of
	  commit_lock for each crtc. (Rob Clark)

Signed-off-by: Krishna Manikandan 
Reviewed-by: Rob Clark 
Signed-off-by: Rob Clark

drm/msm: enable vblank during atomic commits

2020-08-22T05:30:17+00:00

This has roughly the same effect as drm_atomic_helper_wait_for_vblanks(),
basically just ensuring that vblank accounting is enabled so that we get
valid timestamp/seqn on pageflip events.

Signed-off-by: Rob Clark 
Tested-by: Stephen Boyd 
Signed-off-by: Rob Clark

drm/msm: add atomic traces

2019-09-03T23:17:02+00:00

This was useful for debugging fps drops.  I suspect it will be useful
again.

Signed-off-by: Rob Clark 
Reviewed-by: Sean Paul