Stability problems with 6.1 kernel

Stability problems with 6.1 kernel - Printable Version

+- PINE64 (https://forum.pine64.org)
+-- Forum: PinePhone (https://forum.pine64.org/forumdisplay.php?fid=120)
+--- Forum: PinePhone Software (https://forum.pine64.org/forumdisplay.php?fid=121)
+---- Forum: Mobian on PinePhone (https://forum.pine64.org/forumdisplay.php?fid=139)
+---- Thread: Stability problems with 6.1 kernel (/showthread.php?tid=18175)

Stability problems with 6.1 kernel - Zebulon Walton - 04-27-2023

Anyone else seeing stability problems with their Pinephone running Mobian and the 6.1 kernel? I'm finding that with 6.1 the phone experiences frequent random crashes, spontaneous reboots, and user interface lockups. On the other hand the phone is very stable if reverted back to the 5.15 kernel. I'm not sure if it's a hardware issue with this particular phone or if maybe the 6.1 kernel is not quite fully-baked yet for this platform. (I've been retrying 6.1 each time it's updated but so far it just won't run reliably.)

RE: Stability problems with 6.1 kernel - tiol - 05-01-2023

Hi,
I have also seen from time to time spontaneous reboots and (much more frequently) lockups of the user interface on my OG PP with Mobian.

IMHO it would be helpful to collect logs and report issues to [1] so that the maintainer of the -sunxi64 kernel package gets aware of those issues with the package they're maintaining. Guess what, if issues do not get reported, it's unlikely they get solved... Big Grin

You have already tracked down that the culprit is some kernel regression between 5.15 and 6.1, then you are halfway there.

E.g. for UI lockups, it seems we've found the root cause, please check issue #74 [2]. There's a workaround reported that, at least on my device, got rid of lockups: maybe you can also help and test that. When more users confirm it to be the solution, it can be fixed for everyone and we can look on to the next issue Cool

[1] https://salsa.debian.org/Mobian-team/devices/kernels/sunxi64-linux/-/issues
[2] https://salsa.debian.org/Mobian-team/devices/kernels/sunxi64-linux/-/issues/74

RE: Stability problems with 6.1 kernel - zetabeta - 05-01-2023

just in case, do you have 1.2a hw version or 1.2b version?

i got few crashes on manjaro, which sounds very similar. not in mobian though. i think this is separate bug than flipping frames, both powersaving and frequency change. however, i only used powersaving prevention and not frequency change prevention in manjaro. so it could be distantly related to frequenccy change bug.

log from manjaro:

Code:
Apr 27 09:40:05 manjaro-arm kernel: lima 1c40000.gpu: mmu resume

Apr 27 09:40:05 manjaro-arm kernel: lima 1c40000.gpu: pp task error 1 int_state=0 status=5

Apr 27 09:40:05 manjaro-arm kernel: lima 1c40000.gpu: pp task error 0 int_state=0 status=0

Apr 27 09:40:05 manjaro-arm kernel: lima 1c40000.gpu: mmu page fault at 0x2e01a40 from bus id 0 of type read on ppmmu1

Apr 27 09:39:47 manjaro-arm kernel: lima 1c40000.gpu: pp task error 1 int_state=0 status=5

Apr 27 09:39:47 manjaro-arm kernel: lima 1c40000.gpu: pp task error 0 int_state=0 status=0

Apr 27 09:39:47 manjaro-arm kernel: [drm:lima_sched_timedout_job] *ERROR* lima job timeout

if i recall correctly, mobian kernel 5.15 has static gpu frequency into 432MHz. whereas, later kernel has variable frequencies, 432, 312 120 Mhz.

still look these:
https://forum.pine64.org/showthread.php?tid=17771&highlight=flipping+frames+bug
https://gitlab.freedesktop.org/mesa/mesa/-/issues/8415
https://gitlab.freedesktop.org/mesa/mesa/-/issues/8410

RE: Stability problems with 6.1 kernel - Zebulon Walton - 05-01-2023

I have the 1.2b hardware revision. It may well be the variable GPU frequency causing problems. I just ran updates on the phone and see that the 6.1 kernel was updated so I'll see if anything has improved. I've added the workaround suggested in one of the links provided by @tiol (udev rule to fix gpu frequency), and will also look over the others...

RE: Stability problems with 6.1 kernel - treebeard - 05-01-2023

For me, 6.1 remains more stable than 5.15. The fail-to-wake problems are greatly reduced but I have lately seen more instances of fail-to-sleep. (1.2b hardware)

RE: Stability problems with 6.1 kernel - Zebulon Walton - 05-04-2023

Well it's been a few days now and the phone has been stable so far with the latest 6.1 kernel and GPU frequency scaling disabled. Previously I would have started seeing problems by now but I'll keep an eye on it and see how it works longer term.

RE: Stability problems with 6.1 kernel - zetabeta - 05-07-2023

i think i found yet another bug in manjaro, maybe other o.s.'s. i don't know is it related to flipping frames bug(s). my guess it is not. so basically i go ... bug hunting.

very premilinary things tell, that using audio, gpu, mobile data and cpu at the same time crashes system.

as a example, this bug https://gitlab.freedesktop.org/mesa/mesa/-/issues/8198 is not related to the flipping frames bug(s).

RE: Stability problems with 6.1 kernel - Zebulon Walton - 05-08-2023

(05-07-2023, 01:26 PM)zetabeta Wrote: i think i found yet another bug in manjaro, maybe other o.s.'s. i don't know is it related to flipping frames bug(s). my guess it is not. so basically i go ... bug hunting.

very premilinary things tell, that using audio, gpu, mobile data and cpu at the same time crashes system.

as a example, this bug https://gitlab.freedesktop.org/mesa/mesa/-/issues/8198 is not related to the flipping frames bug(s).

Interesting. So far my Mobian installation is still stable running the 6.1 kernel so that problem appears to be solved albeit with a hack to disable GPU frequency scaling. If this new bug is triggered partly by mobile data use I wouldn't see it though, even if present in Mobian, since I don't use mobile data.

RE: Stability problems with 6.1 kernel - zetabeta - 05-11-2023

seems that there is new bug going on it affects mobian as well https://salsa.debian.org/Mobian-team/devices/kernels/sunxi64-linux/-/issues/74#note_398491 .

for me mobian works fine with workarounds. but manjaro is buggy, crashes, reboot and jams. seems to be related with pipewire audio.

RE: Stability problems with 6.1 kernel - zetabeta - 05-12-2023

i think there is new bug going on, now i have some kind of clue what is going on.

i used fedora phosh with flipping frames bug preventions, both frequency and powersaving.

bug is caused by pipewire/wireplumber and/or ram memory hog. note conditions and and or. could be both or either one. still, this assessment is more or less inaccurate.

there is one possible way to reduce this bug, it is kernel parameter "lima.sched_timeout_ms=3000". however, it won't fix underlying issues.