[RESPONDED] Arch hibernation woes on AMD 13

Framework Laptop 13 Linux
21

As a data-point, hibernation still works on mine, latest arch, latest bios on the 7840U framework 13.

1 Like
22

@Adrian_Joachim can you get a high number of cycles using the stress test script I posted above?

23

Interesting it crashed after at least 8 cycles (didn’t pay attention to it after 8 and came back to it on the encryption password prompt).

Never had that happen in actual use before.

Edit: interesting, second run ended with frozen system after 10 cycles

1 Like
24

Yeah, you’ve just been lucky I guess :grimacing:

25

I am now 21 cycles into a run, all I changed was increase the wait between cycles fro 5 to 30s, This may be a going down when not quite up issue or something.

Edit: and just as I post this it freezes XD

1 Like
26

Linux 6.8.4 now includes possibly relevant changes:
"Greg Kroah-Hartman (12):
Revert “workqueue: Shorten events_freezable_power_efficient name”
Revert “workqueue: Don’t call cpumask_test_cpu() with -1 CPU in wq_update_node_max_active()”
Revert “workqueue: Implement system-wide nr_active enforcement for unbound workqueues”
Revert “workqueue: Introduce struct wq_node_nr_active”
Revert “workqueue: RCU protect wq->dfl_pwq and implement accessors for it”
Revert “workqueue: Make wq_adjust_max_active() round-robin pwqs while activating”
Revert “workqueue: Move nr_active handling into helpers”
Revert “workqueue: Replace pwq_activate_inactive_work() with [__]pwq_activate_work()”
Revert “workqueue: Factor out pwq_is_empty()”
Revert “workqueue: Move pwq->max_active to wq->max_active”
Revert “workqueue.c: Increase workqueue name length”
Linux 6.8.4
"
Testing suspend/hibernate now.
suspend works, hibernate wouldn’t start with systemctl hibernate–that is, screen blanked, power led stayed on, and screen/system resumed.
And after attempting to hibernate, returning from suspend takes ~30sec.

OK, a major hint: suspend and hibernate work fine using kernel 6.6.25-1. Kernel 6.8.4 breaks hibernate.
However, even using 6.6.25, resume from suspend after hibernating takes many seconds. Resume from hibernate is quick.
Manjaro Linux, KDE latest.

27

6.8.4 didn’t make a difference for me.

28

Looks like a relevant kernel (6.8.x) error has been patched upstream: [PATCH] PM: s2idle: Make sure CPUs will wakeup directly on resume - Anna-Maria Behnsen

30

looks like that patch fixes a bug that could end up in not reneabling clock(event) for all cores after a resume from suspend to idle, but isn’t this thread about problems with suspend to disk (hibernate) and suspend to ram?

31

Can confirm 6.8.4 is still doing it after 10ish cycles

1 Like
32

Hibernation still fails with 6.8.5 :frowning: 6.6.26 works fine.

33

Normal using it fine or doing more than 21 cycles of the stress-test fine?

Cause mine does the first but not the second.

34

Try this patch:
https://lore.kernel.org/amd-gfx/20240328041720.276684-1-tim.huang@amd.com/

That’s not planned for stable but if it does help for you we can make a case to bring it to stable.

2 Likes
35

hibernation doesn’t work at all for me using 6.8.5, but works normally in 6.6.26 .

36

Thank you. I’m not compiling my own kernel, so will have to rely on other people testing the patch :frowning:

37

On 6.8.4 it mostly works, unless you run the stresstest then it freezes after 10-20 cycles. Never happened in actual use but quite reproducable

38 
Starting cycle 48
Ending cycle 48
Starting cycle 49
Ending cycle 49
Starting cycle 50
Ending cycle 50

Yusss! 50 cycles!

39

Cool! I just sent a request out for it to be brought to stable from your testing.
https://lore.kernel.org/stable/CA+EcB1OKkBSj-VoJpyAgTxPEuj9EOBz-B6Li6fcByYjME7QxDQ@mail.gmail.com/T/#u

3 Likes
40

Thank you! Kernel 6.8.6 does not seem to include it, as expected.

41

What did you do?