Skip to contentSkip to conversation

Screen flicker regression on Baytrail

DISCUSSION Pierre-Louis Bossart 2 participants 2 messages Oct 27, 2016 – Oct 28, 2016

lore
Share & tools
lore
#0 lore raw

While testing our upcoming HDMI audio patches, I experienced pretty bad screen flicker regressions on my Baytrail compute stick. This happens with both v09-rc2 and drm-intel-nightly.

A quick bisect with all the audio patches removed points to the following commit:

2efb813d5388e18255c54afac77bd91acd586908 is the first bad commit commit 2efb813d5388e18255c54afac77bd91acd586908

Author: Chris Wilson <[email protected]>
Date:   Thu Aug 18 17:17:06 2016 +0100

drm/i915: Fallback to using unmappable memory for scanout

More logs below. I don't have the background to fix this myself and I don't see an obvious link with Baytrail but I hope someone smarter than me can help fix this

-Pierre

git bisect start
# bad: [d8923dcfa53d59886d432a3fc430e26cb92ce86a] drm/i915: Track display alignment on VMA
git bisect bad d8923dcfa53d59886d432a3fc430e26cb92ce86a
# good: [03af84fe7f48937941fbb4dcd6cf66c68ae0edd4] drm/i915: Choose partial chunksize based on tile row size git bisect good 03af84fe7f48937941fbb4dcd6cf66c68ae0edd4
# good: [50349247ea807ad0950bbcedb1abb576e6a785db] drm/i915: Drop ORIGIN_GTT for untracked GTT writes git bisect good 50349247ea807ad0950bbcedb1abb576e6a785db
# bad: [2efb813d5388e18255c54afac77bd91acd586908] drm/i915: Fallback to using unmappable memory for scanout git bisect bad 2efb813d5388e18255c54afac77bd91acd586908
# good: [821188778b9be2050d45490c4b2b009d51f041e0] drm/i915: Choose not to evict faultable objects from the GGTT git bisect good 821188778b9be2050d45490c4b2b009d51f041e0

# first bad commit: [2efb813d5388e18255c54afac77bd91acd586908] drm/i915: Fallback to using unmappable memory for scanout

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#1 lore raw
Quoted 9 lines · Pierre-Louis Bossart

While testing our upcoming HDMI audio patches, I experienced pretty bad screen flicker regressions on my Baytrail compute stick. This happens with both v09-rc2 and drm-intel-nightly.

A quick bisect with all the audio patches removed points to the following commit:

2efb813d5388e18255c54afac77bd91acd586908 is the first bad commit commit 2efb813d5388e18255c54afac77bd91acd586908

Quoted 8 lines
Author: Chris Wilson <[email protected]>
Date:   Thu Aug 18 17:17:06 2016 +0100

drm/i915: Fallback to using unmappable memory for scanout

More logs below. I don't have the background to fix this myself and I don't see an obvious link with Baytrail but I hope someone smarter than me can help fix this

Looking around, there's bug [1] with same culprit. Please follow up there.

BR,
Jani.

[1] https://bugs.freedesktop.org/show_bug.cgi?id=98036

Quoted 27 lines

-Pierre

git bisect start
# bad: [d8923dcfa53d59886d432a3fc430e26cb92ce86a] drm/i915: Track display alignment on VMA
git bisect bad d8923dcfa53d59886d432a3fc430e26cb92ce86a
# good: [03af84fe7f48937941fbb4dcd6cf66c68ae0edd4] drm/i915: Choose partial chunksize based on tile row size
git bisect good 03af84fe7f48937941fbb4dcd6cf66c68ae0edd4
# good: [50349247ea807ad0950bbcedb1abb576e6a785db] drm/i915: Drop ORIGIN_GTT for untracked GTT writes
git bisect good 50349247ea807ad0950bbcedb1abb576e6a785db
# bad: [2efb813d5388e18255c54afac77bd91acd586908] drm/i915: Fallback to using unmappable memory for scanout
git bisect bad 2efb813d5388e18255c54afac77bd91acd586908
# good: [821188778b9be2050d45490c4b2b009d51f041e0] drm/i915: Choose not to evict faultable objects from the GGTT
git bisect good 821188778b9be2050d45490c4b2b009d51f041e0

# first bad commit: [2efb813d5388e18255c54afac77bd91acd586908] drm/i915: 
Fallback to using unmappable memory for scanout

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx