k/x/lubuntu 18.04.5 text unreadable on old thinkpads (radeon, HWE 5.4 kernel only)

[Chris Guiver]

** Summary
Screen is unreadable much of the time with black badly-tuned-digital-tv type fragments covering image. As you action things (use a menu, click on a different window etc) the black fragments move (though often only parts of screen; not all of screen)

** Background

This was looked into because of a number of support requests and I detected a pattern. All users are using the HWE kernel, and issue didn't occur using 18.04.4, though I personally only tested the live daily (during the development cycle).

- thinkpad x31
https://discourse.lubuntu.me/t/reviving-thinkpad-x31-problems-with-messed-up-screen/1457/
- thinkpad t42
https://ubuntuforums.org/showthread.php?t=2448800&p=13979514#post13979514
- thinkpad t41
https://askubuntu.com/questions/1266918/latest-updates-to-lubuntu-18-04-4-lts-32-bit-create-almost-impossible-graphics-i

(and I believe others before I recognized the pattern & started noting; I ignored initially as I had no issues with t43, not having used the t42p or r50p in QA testing for awhile - a scan of my QA-testing live will show when I tested with each box in first line of comments). I've also seen this on very different boxes to old thinkpads too.

My boxes impacted

- ibm thinkpad t42p (pentium m, 2gb ram, amd/ati mobility firegl t2; rv350/m10 gl)
- ibm thinkpad r50p (pentium m, 1gb ram, amd/ati mobility firegl t2; rv350/m10 gl)

** 20.04/HWE kernel only

If I live QA-test using the 18.04.4 ISO or released image, no issues occur (using 5.3 kernel & it's packages). Likewise I have no issues using the GA (4.15) kernel in testing (the r50p was used more regularly in testing in prior bionic cycles, alas not much in 18.04.5 development/recently; expecting same results with thinkpad t43 which does not have this issue)

This report should possibly be filed on xserver-xorg. I've only experienced it with old radeon thinkpads (but screen on box makes it difficult to explore)

(Filed online as I've been struggling to file this on actual box)

I get the same issue when using Kubuntu 18.04.5 live on r50p

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This bug is missing log files that will aid in diagnosing the problem. While running an Ubuntu kernel (not a mainline or third-party kernel) please enter the following command in a terminal window:

apport-collect 1891790

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

[Chris Guiver (guiverc)]

Photo of live (Lubuntu 18.04.5) running on r50p

ibm thinkpad r50p (pentium m, 1gb ram, amd/ati mobility firegl t2; rv350/m10 gl)

https://photos.google.com/share/AF1QipMVvJkfURew9B14g4-wmVSUoPMmwjrWCBY1GS9ZDxTa_NPvEwJTEngUnnZAOmsqkQ/photo/AF1QipMtgzUJJE3mZ4IdUXWAW4_UNB7SH3hCPlt-XbEr?key=dkU2T2c0WUJzVi1xNVI3cURUTG9TZlJyYnNNSnFn

Photo of live (Lubuntu 18.04.5) running on t42p

ibm thinkpad t42p (pentium m, 2gb ram, amd/ati mobility firegl t2; rv350/m10 gl)

https://photos.google.com/share/AF1QipMVvJkfURew9B14g4-wmVSUoPMmwjrWCBY1GS9ZDxTa_NPvEwJTEngUnnZAOmsqkQ/photo/AF1QipMXR61U1xJK6gzkoKoatHXgUcok8_8ZgxghKtYv?key=dkU2T2c0WUJzVi1xNVI3cURUTG9TZlJyYnNNSnFn

[Chris Guiver (guiverc)]

I doubt I'm going to be able to `apport-collect 1891790` as I tried for >3 hours on the r50p, and the t42p is now having the same issue (it's just slower to be impacted; maybe because it has more RAM)

Checks for squashfs errors are clean (ie. `dmesg|grep squash`) with only program title appearing once, no errors

Both boxes crash though, at first slow to respond till the eventually pause. If I try `ubuntu-bug linux` (last done on t42p) I get a crash file in /var/crash, but it's hard to read, and it won't file using `ubuntu-bug`.

The closest I got was authorization is required, but firefox never gets to show anything allowing me to authorize it. The laptops appear to get slower & slower they run, and trying to read anything useful on the screen isn't really possible.

I even tried `sudo apt install openssh-server` hoping I could read more from a different box, but that seems to speed up the slowing nature of whatever this box doesn't like in the 5.4 kernel/stack. This will likely be a different issue, but why I filed on the kernel, not xserver-xorg

(apport-collect has now run >30 mins on t42p and I've not been able to authorize it, can't move mouse anymore, but I'm giving up... SysRq caused reboot, so kernel is still running, even if I could not use ctrl+alt+f4 to switch to text terminal; fyi: text terminals are easily read)

[Chris Guiver (guiverc)]

FYI: I could run `apport-collect` on either box using Lubuntu 18.04.4 ISO, or earlier ISO, but I've really tried using 18.04.5 & neither laptop is usable on those boxes (likely two issues, laptop screen unreadable & it slows & freezes due a possibly different issue; this is the real killed as I haven't been able to authorize upload)

[Chris Guiver (guiverc)]

I've finally managed to file
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1892011
over the locking-up/crashing I've mentioned in this report..

These boxes (r50p & t42p) are used regularly by me (r50p is ALWAYS on & up) and run Lubuntu 18.04.4 LTS normally in test (both actually have debian installed and not a Ubuntu flavor)

Also note: I booted Kubuntu 18.04.5 on r50p and it has the same video issue described in this report. The issue is not specific to Lubuntu & LXDE.

[Chris Guiver (guiverc)]

Beyond comment #5 where I confirmed Kubuntu is identical.

In https://ubuntuforums.org/showthread.php?t=2448800&p=13979514#post13979514 a user found the same with Xubuntu 18.04.5 (only that was an installed 18.04.4 upgraded to 18.04.5 rather than my 'live' testing)

[Chris Guiver (guiverc)]

I've booted Xubuntu 18.04.5 (live) on t42p (at least last RC pre-release) and the issue occurs (as reported by user mentioned in comment #6)

[Chris Guiver (guiverc)]

This isn't related to kernels 5.4 or above as Debian Bullseye (5.10 kernel) is giving me no issues on

- ibm thinkpad r50p (pentium m, 1gb ram, amd/ati mobility firegl t2; rv350/m10 gl)

I expect the same on t42p as well (I just haven't got to testing that yet).

In support I was just telling people to stick to the GA kernel stack, as I didn't have issues with that on my kernels (nor 18.04.4 or earlier with HWE). Lubuntu 18.04 LTS however is EOL for the Lubuntu team.