regression: amdgpu: pci_pm_freeze returns -121, hibernate no longer works

I've noticed that as of the last week, my laptop no longer hibernates. Typically, the battery is almost full after sitting for a week, but this weekend, I noticed it was at 50%. I am tempted to say this did not happen on whatever linux release was current as of the week before last, but I cannot be sure that I updated Arch that week.

uname -a
Linux REDACTED 6.13.7-arch1-1 #1 SMP PREEMPT_DYNAMIC Thu, 13 Mar 2025 18:12:00 +0000 x86_64 GNU/Linux

My laptop is a Framework 16 with the Radeon RX 7700S expansion module and 780M integrated graphics.

I performed an experiment to manually trigger a hibernate and see what would happen:

[root@REDACTED ~]# echo "######ABOUT TO SYSTEMCTL HIBERNATE######" > /dev/kmsg
[root@REDACTED ~]# systemctl hibernate
[root@REDACTED ~]# echo "######RESUMED######" > /dev/kmsg

I saw this (complete logs attached):

[89772.220713] amdgpu 0000:03:00.0: amdgpu: MODE1 reset
[89772.220716] amdgpu 0000:03:00.0: amdgpu: GPU mode1 reset
[89772.222512] amdgpu 0000:03:00.0: amdgpu: GPU smu mode1 reset
[89772.222515] amdgpu 0000:03:00.0: amdgpu: SMU: response:0xFFFFFFFF for index:47 param:0x00000000 message:Mode1Reset?
[89772.222522] amdgpu 0000:03:00.0: amdgpu: GPU mode1 reset failed
[89772.222526] amdgpu 0000:03:00.0: PM: pci_pm_freeze(): amdgpu_pmops_freeze [amdgpu] returns -121
[89772.222680] amdgpu 0000:03:00.0: PM: dpm_run_callback(): pci_pm_freeze returns -121
[89772.222686] amdgpu 0000:03:00.0: PM: failed to freeze async: error -121

And I noticed that instead of the power button light turning off as I would typically expect, the system instead appeared to only be sleeping.

Looking at the logs, they're just full of amdgpu warnings and stack traces. Attaching the full logs here: dmesg_for_forums.log

changed the description

I am also seeing the same issue on my Framework 16 laptop. I am running Linux REDACTED 6.13.7-3-cachyos #1 SMP PREEMPT_DYNAMIC Wed, 19 Mar 2025 16:02:53 +0000 x86_64 GNU/Linux I have attached output from my journalctl starting from when I tried to hibernate like @nicholas2 showed above: errors log. I can also confirm the power button light stays on for me too.

I notice when opening the laptop lid after it has been sleeping, I encounter a lock up where no mouse/keyboard input is registered for several seconds, after which it does seem to return to a usable state. However sometimes it then stutters a bit after a second or two for an additional moment.

Yes, I have also started having freezing when I open the laptop lid after sleeping, and it resolves after several seconds.

Possibly a duplicate of #3634. Do the patches proposed on that report help?

I cannot get those patches to apply on torvalds/linux v6.14 , but this might also not be the same issue, since the discussion on that bug goes back to kernel 6.11, which did not have this issue for me.

I also have this problem on a Framework 16 laptop with the Radeon RX 7700S expansion module (Arch Linux). The problem that occurs is fine for me on 6.12.21 lts and breaks on version 6.13.8 (maybe a lower version). It also fixes itself on 6.13.8 when downgrading linux-firmware to 20250210. This also may be related to #4083 (closed) as I also have this issue and it started occurring for me when this occurred too.

can you try the patches from #4083 (closed) (https://lore.kernel.org/amd-gfx/20250328130857.4071486-1-alexander.deucher@amd.com/)?

FWIW, this may be a separate issue. When I run dmesg | grep "MES" after experiencing this issue, I only see "MES ring buffer is full" messages, none of the logs reported on that #4083 (closed).

@superm1 I don't know how to apply the patches myself. Sorry.

added S4 label