Hi Guys,
sorry for the silence from my side. I had a lot of things to take care
of after returning from vacation. Also I had to wait on the zfs modules
to be updated to support kernel 6.3 for further testing.
The bad news is that I am still experiencing issues. I have been able to
get a reproducible trigger for the buggy behavior. The moment I take a
screenshot or any other program like `wdisplays` accesses the screen
buffer the screen starts flickering. The only way to reset it is to
reboot the machine or log out of the desktop.
With this I did a bisection to figure out which commit is responsible
for this. I attached the logs to the mail. The short version is that I
identified commit 81d0bcf9900932633d270d5bc4a54ff599c6ebdb as the
culprit. Seems that there are side effects of having more flexible
buffer placement for the case of the internal GPU. To verify that this
actually is the cause of the issue I built the current archlinux kernel
with an extra patch to revert the commit:
https://github.com/ju6ge/linux/tree/v6.3.5-ju6ge. The result is that be
bug is fixed!
Now if this is the desired long term fix I do not know …
Kind regards,
Felix Richter
On 02.05.23 16:12, Linux regression tracking (Thorsten Leemhuis) wrote:
On 02.05.23 15:48, Felix Richter wrote:
On 5/2/23 15:34, Linux regression tracking (Thorsten Leemhuis) wrote:
On 02.05.23 15:13, Alex Deucher wrote:
On Tue, May 2, 2023 at 7:45 AM Linux regression tracking (Thorsten
Leemhuis)<regressions@xxxxxxxxxxxxx> wrote:
On 30.04.23 13:44, Felix Richter wrote:
Hi,
I am running into an issue with the integrated GPU of the Ryzen 9
7950X. It seems to be a regression from kernel version 6.1 to 6.2.
The bug materializes in from of my monitor blinking, meaning it
turns full white shortly. This happens very often so that the
system becomes unpleasant to use.
I am running the Archlinux Kernel:
The Issue happens on the bleeding edge kernel: 6.2.13
Switching back to the LTS kernel resolves the issue: 6.1.26
I have two monitors attached to the system. One 42 inch 4k Display
and a 24 inch 1080p Display and am running sway as my desktop.
Let me know if there is more information I could provide to help
narrow down the issue.
Thanks for the report. To be sure the issue doesn't fall through the
cracks unnoticed, I'm adding it to regzbot, the Linux kernel regression
tracking bot:
#regzbot ^introduced v6.1..v6.2
#regzbot title drm: amdgpu: system becomes unpleasant to use after
monitor starts blinking and turns full white
#regzbot ignore-activity
This isn't a regression? This issue or a fix for it are already
discussed somewhere else? It was fixed already? You want to clarify
when
the regression started to happen? Or point out I got the title or
something else totally wrong? Then just reply and tell me -- ideally
while also telling regzbot about it, as explained by the page listed in
the footer of this mail.
Developers: When fixing the issue, remember to add 'Link:' tags
pointing
to the report (the parent of this mail). See page linked in footer for
details.
This sounds exactly like the issue that was fixed in this patch which
is already on it's way to Linus:
https://gitlab.freedesktop.org/agd5f/linux/-/commit/08da182175db4c7f80850354849d95f2670e8cd9
FWIW, you in the flood of emails likely missed that this is the same
thread where you yesterday replied "If the module parameter didn't help
then perhaps you are seeing some other issue. Can you bisect?". That's
why I decided to add this to the tracking. Or am I missing something
obvious here?
/me looks around again and can't see anything, but that doesn't have to
mean anything...
Felix, btw, this guide might help you with the bisection, even if it's
just for kernel compilation:
https://docs.kernel.org/next/admin-guide/quickly-build-trimmed-linux.html
And to indirectly reply to your mail from yesterday[1]. You might want
to ignore the arch linux kernel git repo and just do a bisection between
6.1 and the latest 6.2.y kernel using upstream repos; and if I were you
I'd also try 6.3 or even mainline before that, in case the issue was
fixed already.
[1]
https://lore.kernel.org/all/04749ee4-0728-92fe-bcb0-a7320279eaac@xxxxxxxxxxxxxxxxx/
Thanks for the pointers, I'll do a bisection on my desktop from 6.1 to
the newest commit.
FWIW, I wonder what you actually mean with "newest commit" here: a
bisection between 6.1 and mainline HEAD might be a waste of time, *if*
this is something that only happens in 6.2.y (say due to a broken or
incomplete backport)
That was the part I was mostly unsure about … where
to start from.
I was planning to use PKGBUILD scripts from arch to achieve the same
configuration as I would when installing
the package and just rewrite the script to use a local copy of the
source code instead of the repository.
That way I can just use the bisect command, rebuild the package and test
again.
In my experience trying to deal with Linux distro's package managers
creates more trouble than it's worth.
But I probably won't be able to finish it this week, since I am on
vacation starting tomorrow and will not have access to the computer in
question. I will be back next week, by that time the patch Alex is
talking about might
already be in mainline. So if that fixes it, I will notice and let you
know. If not I will do the bisection to figure out what the actual issue
is.
Enjoy your vacation!
Ciao, Thorsten
git bisect start
# Status: warte auf guten und schlechten Commit
# bad: [55c7d6a91d42ad98cbfb10da077ce8bb7084dc0e] Merge tag 'drm-next-2022-12-23' of git://anongit.freedesktop.org/drm/drm
git bisect bad 55c7d6a91d42ad98cbfb10da077ce8bb7084dc0e
# Status: warte auf gute(n) Commit(s), schlechter Commit bekannt
# good: [1eb206208b0f3f707c67134ef6ba394410effb67] block, bfq: only do counting of pending-request for BFQ_GROUP_IOSCHED
git bisect good 1eb206208b0f3f707c67134ef6ba394410effb67
# good: [dd6f9b17cd7af68b6a5090deedf1f5e84f66f4e6] Merge tag 'tty-6.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
git bisect good dd6f9b17cd7af68b6a5090deedf1f5e84f66f4e6
# good: [5f6e430f931d245da838db3e10e918681207029b] Merge tag 'powerpc-6.2-1' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
git bisect good 5f6e430f931d245da838db3e10e918681207029b
# good: [609d3bc6230514a8ca79b377775b17e8c3d9ac93] Merge tag 'net-6.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
git bisect good 609d3bc6230514a8ca79b377775b17e8c3d9ac93
# good: [5461e079009ae2732c833281c4b50dfb58d15ba5] Merge tag 'media/v6.2-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media
git bisect good 5461e079009ae2732c833281c4b50dfb58d15ba5
# good: [9d2f6060fe4c3b49d0cdc1dce1c99296f33379c8] Merge tag 'trace-v6.2-1' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
git bisect good 9d2f6060fe4c3b49d0cdc1dce1c99296f33379c8
# good: [d1ac1a2b14264e98c24db6f8c2bd452e695c7238] Merge tag 'perf-tools-for-v6.2-2-2022-12-22' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux
git bisect good d1ac1a2b14264e98c24db6f8c2bd452e695c7238
# bad: [d118b18fb1da02b41df2da78cb2794b3638d89cd] drm/amd/pm: avoid large variable on kernel stack
git bisect bad d118b18fb1da02b41df2da78cb2794b3638d89cd
# bad: [3273f11675ef11959d25a56df3279f712bcd41b7] drm/amdgpu: Remove unnecessary domain argument
git bisect bad 3273f11675ef11959d25a56df3279f712bcd41b7
# bad: [f95f51a4c3357eabf74fe14ab7daa5b5c0422b27] drm/amdgpu: Add notifier lock for KFD userptrs
git bisect bad f95f51a4c3357eabf74fe14ab7daa5b5c0422b27
# bad: [47ea20762bb7875a62e10433a3cd5d34e9133f47] drm/amdgpu: Add an extra evict_resource call during device_suspend.
git bisect bad 47ea20762bb7875a62e10433a3cd5d34e9133f47
# bad: [81d0bcf9900932633d270d5bc4a54ff599c6ebdb] drm/amdgpu: make display pinning more flexible (v2)
git bisect bad 81d0bcf9900932633d270d5bc4a54ff599c6ebdb
81d0bcf9900932633d270d5bc4a54ff599c6ebdb is the first bad commit
commit 81d0bcf9900932633d270d5bc4a54ff599c6ebdb
Author: Alex Deucher <alexander.deucher@xxxxxxx>
Date: Wed Dec 7 11:08:53 2022 -0500
drm/amdgpu: make display pinning more flexible (v2)
Only apply the static threshold for Stoney and Carrizo.
This hardware has certain requirements that don't allow
mixing of GTT and VRAM. Newer asics do not have these
requirements so we should be able to be more flexible
with where buffers end up.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2270
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2291
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2255
Acked-by: Luben Tuikov <luben.tuikov@xxxxxxx>
Reviewed-by: Christian König <christian.koenig@xxxxxxx>
Signed-off-by: Alex Deucher <alexander.deucher@xxxxxxx>
Cc: stable@xxxxxxxxxxxxxxx
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
