Bug 207581 - [amdgpu] Framebuffer no image CARRIZO and STONEY
Summary: [amdgpu] Framebuffer no image CARRIZO and STONEY
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2020-05-05 13:06 UTC by Jan Burgmeier
Modified: 2020-05-06 07:38 UTC (History)
1 user (show)

See Also:
Kernel Version: 5.4.27
Tree: Mainline
Regression: No


Attachments
Path for 5.4.27 to fix framebuffer (521 bytes, patch)
2020-05-05 13:06 UTC, Jan Burgmeier
Details | Diff
lspci from lenovo 14w (10.00 KB, text/plain)
2020-05-05 13:07 UTC, Jan Burgmeier
Details
lspci from IGEL UD7 (19.01 KB, text/plain)
2020-05-05 13:08 UTC, Jan Burgmeier
Details
git bisect log (2.78 KB, text/plain)
2020-05-05 13:08 UTC, Jan Burgmeier
Details
possible fix (1.21 KB, patch)
2020-05-05 13:46 UTC, Alex Deucher
Details | Diff
possible fix (1.56 KB, patch)
2020-05-05 14:29 UTC, Alex Deucher
Details | Diff

Description Jan Burgmeier 2020-05-05 13:06:02 UTC
Created attachment 288907 [details]
Path for 5.4.27 to fix framebuffer

Hi,

we have some ancient software which uses DirectFB to draw to the framebuffer during installation.
After upgrading from kernel 4.14.99 to 5.4.27 the framebuffer stall the GPU and only a black image was shown. This could be reprocuded on an Igel UD 7 (CARRIZO) and a Lenovo 14w (STONEY) lspci -vv is attached from both devices.

We did some more investigation to find the breaking change please see attached git bisect log:

6c8d74caa2fa33908ecd07fb1cf1b7bc629b367a is the first bad commit

After finding the bad commit we reverted the changes on the 5.4.27 and framebuffer started working again. Patch for reverting is also attached.

Kind regards,
Jan Burgmeier
Comment 1 Jan Burgmeier 2020-05-05 13:07:08 UTC
Created attachment 288909 [details]
lspci from lenovo 14w
Comment 2 Jan Burgmeier 2020-05-05 13:08:21 UTC
Created attachment 288911 [details]
lspci from IGEL UD7
Comment 3 Jan Burgmeier 2020-05-05 13:08:43 UTC
Created attachment 288913 [details]
git bisect log
Comment 4 Alex Deucher 2020-05-05 13:46:50 UTC
Created attachment 288915 [details]
possible fix

Reverting will break all systems with limited vram and large displays.  Does this patch fix the issue?
Comment 5 Alex Deucher 2020-05-05 14:29:36 UTC
Created attachment 288919 [details]
possible fix

This might be a better option.
Comment 6 Jan Burgmeier 2020-05-06 07:38:38 UTC
Thanks for the quick reply.

The first patch from comment 4 works
The second patch from comment 5 does not work

Note You need to log in before you can comment on or make changes to this bug.