Re: [PATCH] drm/i915: Fix legacy gamma lut updates in Linux 4.7-rc6

From: Mario Kleiner
Date: Tue Jul 12 2016 - 11:54:14 EST


On 07/12/2016 05:02 PM, Lionel Landwerlin wrote:
On 12/07/16 13:11, Mario Kleiner wrote:
On 07/12/2016 12:50 PM, Lionel Landwerlin wrote:
Hi Mario,


Hi Lionel,

There was a couple of patch to fix this issue :

https://patchwork.freedesktop.org/series/5467/
https://patchwork.freedesktop.org/series/5466/


Looking at them they should fix the issue, but they seem to be stuck
in review?

I tested this late last week on drm-intel-nightly, it seems a series of
revert fixed most of the issues.


You mean something else has fixed legacy gamma updates, as i can't
find above patches applied on drm-intel-nightly?

This revert on drm-intel-nightly seems to have fixed the problem :

https://cgit.freedesktop.org/drm-intel/commit/drivers/gpu/drm/i915?id=e42aeef1237b7c969a77b7f726c50f6cb832185f


Ok, with that intel-nightly looks like drm-next for 4.8 and that indeed has working lut updates in my testing. My own patch was motivated by the way the implementation is done in intel_atomic_commit_tail() from drm-next.



Are those fixes supposed to be already part of 4.7-rc7, the final rc
afaik?

I haven't seen it on 4.7-rc7.


I just checked Linus tree for 4.7-rc7 and there the code in intel_display.c didn't receive any updates since 13 days and looks like the broken code from rc6 which according to my testing doesn't work.

So i'd assume legacy gamma table updates are broken in Linux 4.7 final rc atm. Couldn't test, because for some weird reason 4.7-rc7 doesn't even boot on my laptop :( - However i got that via a quick install from Ubuntu's mainline ppa so it could be some unrelated problem with their ppa builds.

I think either my patch would fix it, but is untested wrt. nuclear pageflip, or those two patches you referenced, which apparently didn't move forward.

What now?
-mario


thanks,
-mario


Cheers,

-
Lionel

On 12/07/16 11:33, Mario Kleiner wrote:
Updating legacy gamma tables, e.g., via RandR doesn't work at all
as of Linux 4.7-rc6.

Reason seems to be that the required call to
drm_atomic_helper_commit_planes_on_crtc is skipped in
intel_atomic_commit after userspace set new gamma tables,
because neither crtc->state->planes_changed nor
update_pipe (= pipe_config->update_pipe) are true.

Removing the check for planes_changed || update_pipe fixes
gamma table updates.

The code for Linux 4.8 drm-next has changed a lot in that area
wrt. 4.7, but the new code for 4.8 also removed those checks
and calls drm_atomic_helper_commit_planes_on_crtc unconditionally,
and legacy gamma lut updates work on drm-next, so this seems to be
the right solution.

Tested also shutdown/reboot, suspend/resume, (un-)plugging displays,
mode switches for resolution/refresh rate, display rotation, and
page-flipping/pageflip timing on Intel HD Ironlake to confirm the
fix apparently doesn't break anything under X11.

Signed-off-by: Mario Kleiner <mario.kleiner.de@xxxxxxxxx>
Cc: Maarten Lankhorst <maarten.lankhorst@xxxxxxxxxxxxxxx>
Cc: Lionel Landwerlin <lionel.g.landwerlin@xxxxxxxxx>
Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
---
drivers/gpu/drm/i915/intel_display.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_display.c
b/drivers/gpu/drm/i915/intel_display.c
index 04452cf..eb8fb36 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -13685,7 +13685,6 @@ static int intel_atomic_commit(struct
drm_device *dev,
bool modeset = needs_modeset(crtc->state);
struct intel_crtc_state *pipe_config =
to_intel_crtc_state(crtc->state);
- bool update_pipe = !modeset && pipe_config->update_pipe;
if (modeset && crtc->state->active) {
update_scanline_offset(to_intel_crtc(crtc));
@@ -13699,8 +13698,7 @@ static int intel_atomic_commit(struct
drm_device *dev,
drm_atomic_get_existing_plane_state(state,
crtc->primary))
intel_fbc_enable(intel_crtc);
- if (crtc->state->active &&
- (crtc->state->planes_changed || update_pipe))
+ if (crtc->state->active)
drm_atomic_helper_commit_planes_on_crtc(old_crtc_state);
if (pipe_config->base.active &&
needs_vblank_wait(pipe_config))