It seems that since 5.16-rc1, I am unable to change CPU frequency scaling drivers. My system just hangs, with the process blocking. It also takes very long to reboot for the various timeouts to occur. I wonder if anyone reading this knows how to shorten the timeouts so that I can debug faster and/or bisect the kernel faster? Extract from "journalctl --since=today", and was after more than 5 minutes:
Code:
Nov 25 14:21:18 s19 systemd[1]: session-1.scope: Stopping timed out. Killing.
Nov 25 14:21:18 s19 systemd[1]: session-1.scope: Killing process 1192 (sudo) with signal SIGKILL.
Nov 25 14:21:18 s19 systemd[1]: session-1.scope: Killing process 1193 (tee) with signal SIGKILL.
Nov 25 14:21:18 s19 systemd[1]: session-1.scope: Failed with result 'timeout'.
Nov 25 14:21:18 s19 systemd[1]: Stopped Session 1 of user doug.
Nov 25 14:21:18 s19 systemd[1]: Stopping Login Service...
By the way I was trying to change from the intel_cpufreq CPU frequency scaling driver (A.K.A. intel_pstate driver in passive mode) to the intel_pstate driver (A.K.A. active mode) via:
Code:
$ echo active | sudo tee /sys/devices/system/cpu/intel_pstate/status
The hang occurs in either direction with or without HWP: passive to active; active to passive;
I am able to launch a "sudo shutdown -r now" command via another terminal.
EDIT 1: I will try shorter hung task timeout:
Code:
doug@s19:/var/log$ grep . /proc/sys/kernel/hung*
/proc/sys/kernel/hung_task_all_cpu_backtrace:0
/proc/sys/kernel/hung_task_check_count:4194304
/proc/sys/kernel/hung_task_check_interval_secs:0
/proc/sys/kernel/hung_task_panic:0
/proc/sys/kernel/hung_task_timeout_secs:120
/proc/sys/kernel/hung_task_warnings:10
EDIT 2: "sudo reboot --force" seems to work faster.
EDIT3: Results from kernel bisection:
Code:
# first bad commit: [7a89d7eacf8e84f2afb94db5ae9d9f9faa93f01c] powercap/drivers/dtpm: Simplify the dtpm table
commit 7a89d7eacf8e84f2afb94db5ae9d9f9faa93f01c
Author: Daniel Lezcano <daniel.lezcano@linaro.org>
Date: Fri Mar 12 14:04:09 2021 +0100
powercap/drivers/dtpm: Simplify the dtpm table
The dtpm table is an array of pointers, that forces the user of the
table to define initdata along with the declaration of the table
entry. It is more efficient to create an array of dtpm structure, so
the declaration of the table entry can be done by initializing the
different fields.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Link: https://lore.kernel.org/r/20210312130411.29833-3-daniel.lezcano@linaro.org
diff --git a/drivers/powercap/dtpm.c b/drivers/powercap/dtpm.c
index 58433b8ef9a1..8c032398f6ce 100644
--- a/drivers/powercap/dtpm.c
+++ b/drivers/powercap/dtpm.c
@@ -467,7 +467,7 @@ int dtpm_register(const char *name, struct dtpm *dtpm, struct dtpm *parent)
...
Upstream email thread.
Tested workarounds : Revert the commit; disable dtpm in the kernel configuration (it's still experimental anyhow, and shouldn't actually be enabled by default); Apply the temporary upstream patch, which missed -rc3, but hopefully will make -rc4.
