[PATCH 0/1][SRU][oem-5.14/impish/jammy] AMD Cezanne/Barcelo taking cores offline causes S0i3 failures
You-Sheng Yang
vicamo.yang at canonical.com
Wed Dec 15 16:40:16 UTC 2021
BugLink: https://bugs.launchpad.net/bugs/1954930
[Impact]
Detailed in https://gitlab.freedesktop.org/drm/amd/-/issues/1708, taking
some cpu cores offline using cpufreq gadgets or via sysfs may hang the
system.
[Fix]
In v5.16-rc1 commit d6b88ce2eb9d ("ACPI: processor idle: Allow playing
dead in C3 state") fixes this issue.
[Test Case]
As stated in aforementioned bug url, setup cpufreq extention to take
down a few cpu cores, and trigger system suspend. There are ~50% chances
that networking/input/... would hang and the user can only reboot by
sysrq keys.
[Where problems could occur]
According to the patch discussion thread in
https://lore.kernel.org/linux-acpi/20210922133116.102-1-richard.gong@amd.com/,
the limitation to allow enter_dead in no more than ACPI_STATE_C2 might
not have a practical meaning, but simply C2 was the deepest supported
then.
[Other Info]
While this is currently only available in v5.16-rc1 and affects AMD
Cezanne/Barcelo, oem-5.14/impish and jammy are nominated.
Richard Gong (1):
ACPI: processor idle: Allow playing dead in C3 state
drivers/acpi/processor_idle.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
--
2.32.0
More information about the kernel-team
mailing list