Skip to contentSkip to conversation

Regression: Black screen after resume from powersave

DISCUSSION Martin Babutzka 3 participants 4 messages Jul 24, 2017 3 messages not in the archive

lore
Share & tools
lore
#0 lore raw

Hello,

I tracked down another regression affecting my PC with R9 380 GPU, running the amd-staging-4.11 kernel on Xubuntu 17.04:

After resume from powersave, display freezes when the powersave is triggered from the lightdm locked screen or stays completely black if the powersave was triggered from a logged in user.

The regression occurs the first time when I build the kernel including commit 323a1cc8f7d3c40a8e77751d62545de43fa52653. Here is the result from the testing series:
8. Step - crashes
bfd40a54f781899ee5b9ddffc459a6eafe2bba86
11. Step - crashes
dde557132beb2cfb9e98046936cb3565b21d6f9c
12. Step - crashes
323a1cc8f7d3c40a8e77751d62545de43fa52653
10. Step - works
19993e3a86268a30c8b1c31c4518fc403431efe2 3af79f94b85557d26d53909ed85526ec980ffd7f b81aface585aab7262a830aa2f93ff53d4baf3c0
9. Step - works
8d1610a0f3b99e3f950064dc4c476000d3b6ba02 d36719c5ea181601f47e2d056a444d025a5247c6 2247dfee70a96025ba389f363b9088e2d377c7cc 239e79555c7e490823582b9a66390e76ebabb9ab

Maybe this is also linked to:
"iMac 10,1 with Ubuntu 16.04: black screen after suspend (Florian Echtler)"

If you fix this successfully I would be very glad if you push the commit to amd-staging-4.11 again.

Many regards,
Martin Babutzka
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): [email protected]

#2 lore raw
Quoted 35 lines

-----Original Message-----
From: amd-gfx [mailto:[email protected]] On Behalf Of Martin Babutzka
Sent: Monday, July 24, 2017 1:06 PM
To: [email protected]
Subject: Regression: Black screen after resume from powersave

Hello,

I tracked down another regression affecting my PC with R9 380 GPU, running the amd-staging-4.11 kernel on Xubuntu 17.04:

After resume from powersave, display freezes when the powersave is triggered from the lightdm locked screen or stays completely black if the powersave was triggered from a logged in user.

The regression occurs the first time when I build the kernel including commit 323a1cc8f7d3c40a8e77751d62545de43fa52653. Here is the result from the testing series:
8. Step - crashes
bfd40a54f781899ee5b9ddffc459a6eafe2bba86
11. Step - crashes
dde557132beb2cfb9e98046936cb3565b21d6f9c
12. Step - crashes
323a1cc8f7d3c40a8e77751d62545de43fa52653
10. Step - works
19993e3a86268a30c8b1c31c4518fc403431efe2 3af79f94b85557d26d53909ed85526ec980ffd7f b81aface585aab7262a830aa2f93ff53d4baf3c0
9. Step - works
8d1610a0f3b99e3f950064dc4c476000d3b6ba02 d36719c5ea181601f47e2d056a444d025a5247c6 2247dfee70a96025ba389f363b9088e2d377c7cc 239e79555c7e490823582b9a66390e76ebabb9ab

Quoted 3 lines

Maybe this is also linked to:
"iMac 10,1 with Ubuntu 16.04: black screen after suspend (Florian Echtler)"

Not related. That patch is for a different driver (radeon) and different hardware (r7xx).

Quoted 3 lines

If you fix this successfully I would be very glad if you push the commit to amd-staging-4.11 again.

Can you do a proper git bisect?

Alex

Quoted 7 lines

Many regards,
Martin Babutzka
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

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

A message that is not in the archive (sent off-list or not yet received): BN6PR12MB16524C3D042033168AEAC891F7BB0-/b2+HYfkarQqUD6E6FAiowdYzm3356FpvxpqHgZTriW3zl9H0oFU5g@public.gmane.org

author Reviewed-by Alex Deucher
#4 lore raw

Hello,

Didn't know git bisect yet - awesome command. I did basically a binary search manually but grouped some commits logically together. Well this the result with git bisect:

323a1cc8f7d3c40a8e77751d62545de43fa52653 is the first bad commit commit 323a1cc8f7d3c40a8e77751d62545de43fa52653
Author: Christian König <[email protected]> Date:   Fri Jun 30 12:19:42 2017 +0200

    drm/amdgpu: stop mapping BOs to GTT

    No need to map BOs to GTT on eviction and intermediate transfers any more.

:040000 040000 962cffe27f66877559c78b69ba7f0e2d503e12d4
cd0a1c7e073dbbcc6497d4e6a76ba41ad248ea30 M drivers

git bisect log
git bisect start
# bad: [f44148a55c29decda4ce3156426008282feb846e] drm/amdgpu: Try evicting from CPU visible to invisible VRAM first
git bisect bad f44148a55c29decda4ce3156426008282feb846e
# good: [66e26593fbe5523965ccb547fb18c50b7becf91e] drm/amd/powerplay:
fix bug fail to remove sysfs when rmmod amdgpu.
git bisect good 66e26593fbe5523965ccb547fb18c50b7becf91e
# bad: [dde557132beb2cfb9e98046936cb3565b21d6f9c] drm/amdgpu: remove maximum BO size limitation v2
git bisect bad dde557132beb2cfb9e98046936cb3565b21d6f9c
# good: [69112fb54676b519aedb3f15069ac0273fc051b4] drm/amd/powerplay:
move VI common AVFS code to smu7_smumgr.c
git bisect good 69112fb54676b519aedb3f15069ac0273fc051b4
# good: [d36719c5ea181601f47e2d056a444d025a5247c6] drm/amdgpu/gmc9: get vram width from atom for Raven
git bisect good d36719c5ea181601f47e2d056a444d025a5247c6
# good: [3af79f94b85557d26d53909ed85526ec980ffd7f] drm/amdgpu: add amdgpu_gart_map function v2
git bisect good 3af79f94b85557d26d53909ed85526ec980ffd7f
# bad: [323a1cc8f7d3c40a8e77751d62545de43fa52653] drm/amdgpu: stop mapping BOs to GTT
git bisect bad 323a1cc8f7d3c40a8e77751d62545de43fa52653
# good: [19993e3a86268a30c8b1c31c4518fc403431efe2] drm/amdgpu: use the GTT windows for BO moves v2
git bisect good 19993e3a86268a30c8b1c31c4518fc403431efe2

# first bad commit: [323a1cc8f7d3c40a8e77751d62545de43fa52653]
drm/amdgpu: stop mapping BOs to GTT

Many regards,
Martin Babutzka

Am Montag, den 24.07.2017, 18:12 +0000 schrieb Deucher, Alexander:

Quoted 46 lines
Quoted 38 lines

-----Original Message-----
From: amd-gfx [mailto:[email protected]] On Behalf
Of Martin Babutzka
Sent: Monday, July 24, 2017 1:06 PM
To: [email protected]
Subject: Regression: Black screen after resume from powersave

Hello,

I tracked down another regression affecting my PC with R9 380 GPU, running the amd-staging-4.11 kernel on Xubuntu 17.04:

After resume from powersave, display freezes when the powersave is triggered from the lightdm locked screen or stays completely black if
the powersave was triggered from a logged in user.

The regression occurs the first time when I build the kernel including
commit 323a1cc8f7d3c40a8e77751d62545de43fa52653. Here is the result from the testing series:
8. Step - crashes bfd40a54f781899ee5b9ddffc459a6eafe2bba86
11. Step - crashes dde557132beb2cfb9e98046936cb3565b21d6f9c
12. Step - crashes 323a1cc8f7d3c40a8e77751d62545de43fa52653
10. Step - works 19993e3a86268a30c8b1c31c4518fc403431efe2 3af79f94b85557d26d53909ed85526ec980ffd7f b81aface585aab7262a830aa2f93ff53d4baf3c0
9. Step - works
8d1610a0f3b99e3f950064dc4c476000d3b6ba02 d36719c5ea181601f47e2d056a444d025a5247c6 2247dfee70a96025ba389f363b9088e2d377c7cc 239e79555c7e490823582b9a66390e76ebabb9ab

Quoted 3 lines

Maybe this is also linked to:
"iMac 10,1 with Ubuntu 16.04: black screen after suspend (Florian Echtler)"

Not related.  That patch is for a different driver (radeon) and different hardware (r7xx).

Quoted 1 line
Quoted 14 lines
Quoted 2 lines

If you fix this successfully I would be very glad if you push the commit to amd-staging-4.11 again.

Can you do a proper git bisect?

Alex

Quoted 7 lines

Many regards,
Martin Babutzka
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

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

A message that is not in the archive (sent off-list or not yet received): [email protected]

#6 lore raw

Hi Martin,

Thanks, this was also reported internally and I'm fairly certain Christian pushed a fix to the amd-staging-4.11 tree two weeks ago.

I'm fairly certain this is the fix:

d6729592ba0ed766bfe634d766ea8b58dbdd487d

Which is in the public tree:

https://cgit.freedesktop.org/~agd5f/linux/commit/?h=amd-staging-4.11&id=d6729592ba0ed766bfe634d766ea8b58dbdd487d

I had a suspend/resume + load issues on my Carrizo that this series seemed to have fixed.

Here's the list discussion:

https://lists.freedesktop.org/archives/amd-gfx/2017-July/011092.html

Tom

Quoted 117 lines · Martin Babutzka

Hello,

Didn't know git bisect yet - awesome command. I did basically a binary search manually but grouped some commits logically together. Well this the result with git bisect:

323a1cc8f7d3c40a8e77751d62545de43fa52653 is the first bad commit commit 323a1cc8f7d3c40a8e77751d62545de43fa52653

Author: Christian König <[email protected]>
Date:   Fri Jun 30 12:19:42 2017 +0200

drm/amdgpu: stop mapping BOs to GTT

No need to map BOs to GTT on eviction and intermediate transfers
any more.

:040000 040000 962cffe27f66877559c78b69ba7f0e2d503e12d4
cd0a1c7e073dbbcc6497d4e6a76ba41ad248ea30 M drivers

git bisect log
git bisect start
# bad: [f44148a55c29decda4ce3156426008282feb846e] drm/amdgpu: Try evicting from CPU visible to invisible VRAM first
git bisect bad f44148a55c29decda4ce3156426008282feb846e
# good: [66e26593fbe5523965ccb547fb18c50b7becf91e] drm/amd/powerplay:
fix bug fail to remove sysfs when rmmod amdgpu.
git bisect good 66e26593fbe5523965ccb547fb18c50b7becf91e
# bad: [dde557132beb2cfb9e98046936cb3565b21d6f9c] drm/amdgpu: remove maximum BO size limitation v2
git bisect bad dde557132beb2cfb9e98046936cb3565b21d6f9c
# good: [69112fb54676b519aedb3f15069ac0273fc051b4] drm/amd/powerplay:
move VI common AVFS code to smu7_smumgr.c
git bisect good 69112fb54676b519aedb3f15069ac0273fc051b4
# good: [d36719c5ea181601f47e2d056a444d025a5247c6] drm/amdgpu/gmc9: get vram width from atom for Raven
git bisect good d36719c5ea181601f47e2d056a444d025a5247c6
# good: [3af79f94b85557d26d53909ed85526ec980ffd7f] drm/amdgpu: add amdgpu_gart_map function v2
git bisect good 3af79f94b85557d26d53909ed85526ec980ffd7f
# bad: [323a1cc8f7d3c40a8e77751d62545de43fa52653] drm/amdgpu: stop mapping BOs to GTT
git bisect bad 323a1cc8f7d3c40a8e77751d62545de43fa52653
# good: [19993e3a86268a30c8b1c31c4518fc403431efe2] drm/amdgpu: use the GTT windows for BO moves v2
git bisect good 19993e3a86268a30c8b1c31c4518fc403431efe2

# first bad commit: [323a1cc8f7d3c40a8e77751d62545de43fa52653]
drm/amdgpu: stop mapping BOs to GTT

Many regards,
Martin Babutzka

Am Montag, den 24.07.2017, 18:12 +0000 schrieb Deucher, Alexander:

Quoted 46 lines
Quoted 38 lines

-----Original Message-----
From: amd-gfx [mailto:[email protected]] On Behalf
Of Martin Babutzka
Sent: Monday, July 24, 2017 1:06 PM
To: [email protected]
Subject: Regression: Black screen after resume from powersave

Hello,

I tracked down another regression affecting my PC with R9 380 GPU, running the amd-staging-4.11 kernel on Xubuntu 17.04:

After resume from powersave, display freezes when the powersave is triggered from the lightdm locked screen or stays completely black if
the powersave was triggered from a logged in user.

The regression occurs the first time when I build the kernel including
commit 323a1cc8f7d3c40a8e77751d62545de43fa52653. Here is the result from the testing series:
8. Step - crashes bfd40a54f781899ee5b9ddffc459a6eafe2bba86
11. Step - crashes dde557132beb2cfb9e98046936cb3565b21d6f9c
12. Step - crashes 323a1cc8f7d3c40a8e77751d62545de43fa52653
10. Step - works 19993e3a86268a30c8b1c31c4518fc403431efe2 3af79f94b85557d26d53909ed85526ec980ffd7f b81aface585aab7262a830aa2f93ff53d4baf3c0
9. Step - works
8d1610a0f3b99e3f950064dc4c476000d3b6ba02 d36719c5ea181601f47e2d056a444d025a5247c6 2247dfee70a96025ba389f363b9088e2d377c7cc 239e79555c7e490823582b9a66390e76ebabb9ab

Quoted 3 lines

Maybe this is also linked to:
"iMac 10,1 with Ubuntu 16.04: black screen after suspend (Florian Echtler)"

Not related. That patch is for a different driver (radeon) and different hardware (r7xx).

Quoted 1 line
Quoted 14 lines
Quoted 2 lines

If you fix this successfully I would be very glad if you push the commit to amd-staging-4.11 again.

Can you do a proper git bisect?

Alex

Quoted 7 lines

Many regards,
Martin Babutzka
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Quoted 5 lines

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

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