On 2/17/2024 03:39, Hans de Goede wrote:
Hi,
On 2/17/24 09:13, Linux regression tracking (Thorsten Leemhuis) wrote:
[/me among others adds Shyam (author of the culprit) and Hans (committed
it) to the list of recipients]
For the newly joined, this thread starts here:
https://lore.kernel.org/all/ed2226ff-257b-4cfd-afd6-bf3be9785474@localhost/
Thanks (1)
Trolli, there are 2 PMF fixes pending for 6.8, one of which is
suspend-resume related:
https://patchwork.kernel.org/project/platform-driver-x86/patch/20240216064112.962582-1-Shyam-sundar.S-k@xxxxxxx/
https://patchwork.kernel.org/project/platform-driver-x86/patch/20240216064112.962582-2-Shyam-sundar.S-k@xxxxxxx/
I plan to merge these and send them in their way to Linus next week.
Can you build a 6.8-c4 with those 2 included and see if that fixes things ?
You can download mbox files from patchwork and then do "git am <file>"
to apply them.
Regards,
Hans
I could reproduce the hang myself as well. I've tried those patches but
they're not enough.
Here's a fix:
https://lore.kernel.org/platform-driver-x86/20240217005216.113408-1-mario.limonciello@xxxxxxx/T/#u
1) Note I'm subscribed to the regressions list so I was already following this.
On 17.02.24 00:26, Trolli Schmittlauch wrote:
Still, I had the time to bisect the vanilla kernel and landed at
7c45534afa4435c9fceeeb8ca33c0fdc269c2240 as the first bad commit.
Reminder, that's 7c45534afa4435 ("platform/x86/amd/pmf: Add support for
PMF Policy Binary") [v6.8-rc1]
Could be a red herring though[]
If this turns out to be accurate, it's actually quite interesting.
You can prove it's accurate by doing a module blacklist for the
amd-pmf driver (even on 6.8-rc4+).
If that works, it's a great hint at the problem scope and we need to
pull Shyam into the conversation.
I just checked this with Linux version 6.8.0-rc4 (nixbld@localhost) (gcc
(GCC) 12.3.0, GNU ld (GNU Binutils) 2.40) #1-NixOS SMP PREEMPT_DYNAMIC
Sun Feb 11 20:18:13 UTC 2024.
Blacklisting the "amd-pmf" module indeed resolves the issue, good catch.
I've attached to s2idle logs of 4 successful cycles without amd-pmf. To
be honest, I don't know why the script considers the kernel to be
tainted. I have compiled the kernel via the NixOS packaging
infrastracture though without checking whether the distro applies any
patches, but a guix user had also confirmed the issue in the Framework
forum so I guess this issue is vanilla enough.
Looking forward to resolving this. I guess for most use cases I can just
keep the module blacklisted for now?
Ciao, Thorsten
P.S.: 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 7c45534afa4435 /
#regzbot title platform/x86/amd/pmf: system freezes after resuming from
suspend
#regzbot ignore-activity
