https://bugzilla.kernel.org/show_bug.cgi?id=220287
Bug ID: 220287
Summary: [REGRESSION] Rare hangs during suspend to RAM since
around Linux 6.9
Product: ACPI
Version: 2.5
Hardware: Intel
OS: Linux
Status: NEW
Severity: normal
Priority: P3
Component: Power-Sleep-Wake
Assignee: acpi_power-sleep-w...@kernel-bugs.osdl.org
Reporter: mat.jonc...@o2.pl
Regression: No
Created attachment 308324
--> https://bugzilla.kernel.org/attachment.cgi?id=308324&action=edit
DRAFT: power/trace: use correct year in case RTC has a century register (resp.
Hello,
Since some time, I have been observing sporadic hangs during suspend to
RAM on my HP 17-by0001nw laptop. This laptop was bought in 2019 and has
an Intel Kaby Lake CPU.
I usually suspend the laptop to RAM most of the time and make a full
restart infrequently. These hangs happen around once 5-7 days on average
with my normal usage, which involves many suspend to RAM cycles.
The symptoms are:
- the laptop's power LED stays on (doesn't blink as in suspend),
- no reaction to keyboard or a short press on the power button,
- SysRQ doesn't work (made sure it is enabled via sysctl).
There's nothing in the pstore after a reset.
I tried to debug this with pm-trace [1] (which stores some debugging
information in the RTC clock) and determined it is happening no earlier
then after syscore_suspend() / pm_wakeup_pending() have been executed by
suspend_enter() (see attached for a diff).
Also, this happens during a deep suspend, the laptop does not support
s2idle well. I think I have ruled out the s2idle fallback as a culprit,
as this hang happens also when it is forcefully disabled in
pm_suspend():
if (state == PM_SUSPEND_TO_IDLE)
return -EINVAL;
Not able to reproduce on demand with sleepgraph, at least not with 20
second delays.
Additionally, it is possible that it has happened once during a normal
laptop shutdown (as in "shutdown -h now"), but it's quite possible that
it was caused by something other.
This issue does not happen on kernel 5.15 from Ubuntu 20.04, Linux
6.6.93 also appears solid (5 + 13 days without a hang). On the other
hand, it happened to me twice on 6.10.14. My old journalctl logs suggest
somewhat that it was broken by Linux 6.9, but it is in no way
definitive.
Bisection would take a long time, as I need around 2-3 weeks to
determine that a kernel is solid. However, I'll do it if necessary.
I use Ubuntu 20.04.
[1] I had to fix pm-trace as my BIOS was not happy with RTC dates before
2000, which pm-trace normally uses. See attached.
--
You may reply to this email to add a comment.
You are receiving this mail because:
You are watching the assignee of the bug.
_______________________________________________
acpi-bugzilla mailing list
acpi-bugzilla@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/acpi-bugzilla
