Bug 206973 - [AMD/ATI] RV200 [Radeon 7500/7500 LE] [1002:5157] overwhelming video corruption in Xorg
Summary: [AMD/ATI] RV200 [Radeon 7500/7500 LE] [1002:5157] overwhelming video corrupti...
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(AGP) (show other bugs)
Hardware: i386 Linux
: P1 normal
Assignee: Dave Airlie
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2020-03-26 06:02 UTC by Felix Miata
Modified: 2020-09-18 01:57 UTC (History)
1 user (show)

See Also:
Kernel Version: 5.5.4
Subsystem:
Regression: No
Bisected commit-id:


Attachments
corruption screenshot on Tumbleweed (91.39 KB, image/png)
2020-03-26 06:02 UTC, Felix Miata
Details
dmesg from (corrupting) 5.5.4 Mageia 7 kernel (83.00 KB, text/plain)
2020-03-26 08:29 UTC, Felix Miata
Details
revert (2.97 KB, patch)
2020-08-17 15:12 UTC, Alex Deucher
Details | Diff

Description Felix Miata 2020-03-26 06:02:11 UTC
Created attachment 288067 [details]
corruption screenshot on Tumbleweed

Original summary:
[AMD/ATI] RV200 [Radeon 7500/7500 LE] [1002:5157] overwhelming video corruption in Xorg

Downstream bug report https://bugs.mageia.org/show_bug.cgi?id=26237 against 5.5.4 fixed in 5.5.6, is not reproducible in 5.4.17 or 5.5.9. Excerpts from Mageia kernel-desktop changelogs:
[quote]
update to 5.5.6 20200312:
- work around radeon regression introduced in 5.4 series (mga#26237)
  * Revert "drm/radeon: simplify and cleanup setting the dma mask"
  * Revert "drm/radeon: handle PCIe root ports with addressing limitations"
update to 5.4.6 20200215:
- Revert "drm/radeon: simplify and cleanup setting the dma mask"
- Revert "drm/radeon: handle PCIe root ports with addressing limitations"
[/quote]

Mailing list thread starter:
https://lists.freedesktop.org/archives/amd-gfx/2020-March/047703.html

Not reproducible in Debian testing/11 Bullseye 5.4.x.

Corruption appears in openSUSE Tumbleweed post-5.3.12 (e.g. in 5.4.17 & 5.5.9).

Corruption appears in Fedora 30 post-5.3.7 (e.g. in 5.4.19 & 5.5.11).

Workaround in Fedora and Tumbleweed: radeon.agpmode=-1

Not reproducible with Radeon AGP rv250 (32bit) or PCIe rv380 (64bit).
Comment 1 Felix Miata 2020-03-26 08:11:33 UTC
(In reply to Felix Miata from comment #0)
> Not reproducible in Debian testing/11 Bullseye 5.4.x.

That statement was the result of a mixup with my rv250 and rv380 testing. Corruption is present using linux-image-5.4.0-4-686 aka  5.4.19.
Comment 2 Felix Miata 2020-03-26 08:29:13 UTC
Created attachment 288071 [details]
dmesg from (corrupting) 5.5.4 Mageia 7 kernel
Comment 3 Alex Deucher 2020-08-17 15:12:44 UTC
Created attachment 290951 [details]
revert

Does this revert fix the issue?
Comment 4 Felix Miata 2020-08-17 18:42:42 UTC
Does that revert show up in a ready-to-install package for Debian, Tumbleweed or Mageia? I'm not a builder.
Comment 5 Alex Deucher 2020-09-16 20:49:55 UTC
Does it work correctly with 5.9-rc1 or newer?
Comment 6 Alex Deucher 2020-09-16 22:06:10 UTC
Another thing to try, does setting radeon.agpmode=-1 fix the issue?
Comment 7 Felix Miata 2020-09-18 01:57:03 UTC
No apparent persistent corruption in KDM3, Konsole3, Kicker3 or KDE3 desktop with RV200 in Tumbleweed 20200916 with kernel-default-5.9.rc5-1.1.ge921ea1.i586, without need to employ radeon.agpmode=-1. With 'Option "Composite" "Disable"' removed from config, I do observe a full screen width, fleeting flash of corruption on KDE session startup that doesn't reappear when it's restored. The full screen with corruption resembles the flash of corruption I've come to expect from old Radeons during init without Plymouth installed or enabled on tty1 when initial video mode switches to the configured video mode, e.g. video=1440x900 from 80x25 text or vga=791.

radeon.agpmode=-1 does avoid the corruption used with TW's kernel-default-5.7.11.

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