thinkfan
thinkfan copied to clipboard
vmatare
Autostart fails, and restart is needed because hwmon is not found
Hello, For the last few days, I have been experiencing issues with thinkfan failing to start, because it cannot find hwmon.
OS: Manjaro Linux x86_64 - testing branch Host: ThinkPad T490 Kernel: 5.9.12-1-MANJARO CPU: Intel i7-8565U
● thinkfan.service - simple and lightweight fan control program
Loaded: loaded (/usr/lib/systemd/system/thinkfan.service; enabled; vendor preset: disabled)
Drop-In: /etc/systemd/system/thinkfan.service.d
└─override.conf
Active: failed (Result: exit-code) since Tue 2020-12-15 21:58:42 CET; 1h 25min ago
Process: 520 ExecStart=/usr/bin/thinkfan $THINKFAN_ARGS (code=exited, status=1/FAILURE)
Dec 15 21:58:42 miroslav-20n2cto1ww systemd[1]: Starting simple and lightweight fan control program...
Dec 15 21:58:42 miroslav-20n2cto1ww thinkfan[520]: ERROR: Error scanning /sys/devices/platform/coretemp.0/hwmon: No such file or directory
Dec 15 21:58:42 miroslav-20n2cto1ww systemd[1]: thinkfan.service: Control process exited, code=exited, status=1/FAILURE
Dec 15 21:58:42 miroslav-20n2cto1ww systemd[1]: thinkfan.service: Failed with result 'exit-code'.
Dec 15 21:58:42 miroslav-20n2cto1ww systemd[1]: Failed to start simple and lightweight fan control program.
thinkfan.yaml:
sensors:
- hwmon: /sys/devices/platform/coretemp.0/hwmon
indices: [2, 3, 4, 5]
levels:
- [0, 0 , 55]
- [1, 50, 60]
- [2, 55, 65]
- [3, 60, 70]
- [4, 65, 75]
- [5, 70, 80]
- [6, 75, 85]
- [7, 80, 32767]
Since restarting the service fixes the issue, I assume that hwmon is initialized after thinkfan. Is this a more widespread issue or just me having messed up configuration? Thanks for the help!
I've made a small change, can you uninstall thinkfan and re-try with the latest master built from github?
Unfortunately no improvement. I double-checked that got rid of the old version from the AUR + that I pulled the latest master. I tried it compiled for both release and debug.
● thinkfan.service - Thinkfan, the minimalist fan control program
Loaded: loaded (/usr/local/lib/systemd/system/thinkfan.service; enabled; vendor preset: disabled)
Drop-In: /etc/systemd/system/thinkfan.service.d
└─default.conf, override.conf
Active: failed (Result: exit-code) since Wed 2020-12-16 01:02:42 CET; 43s ago
Process: 490 ExecStart=/usr/local/sbin/thinkfan $THINKFAN_ARGS (code=exited, status=1/FAILURE)
Dec 16 01:02:42 miroslav-20n2cto1ww systemd[1]: Starting Thinkfan, the minimalist fan control program...
Dec 16 01:02:42 miroslav-20n2cto1ww thinkfan[490]: ERROR: Error scanning /sys/devices/platform/coretemp.0/hwmon: No s>
Dec 16 01:02:42 miroslav-20n2cto1ww systemd[1]: thinkfan.service: Control process exited, code=exited, status=1/FAILU>
Dec 16 01:02:42 miroslav-20n2cto1ww systemd[1]: thinkfan.service: Failed with result 'exit-code'.
Dec 16 01:02:42 miroslav-20n2cto1ww systemd[1]: Failed to start Thinkfan, the minimalist fan control program.
Restarting the service still (temporarily, until the next restart) fixes the issue.
Here it is.
>>> systemctl list-dependencies thinkfan [130] thinkfan.service ● ├─system.slice ● └─sysinit.target ● ├─dev-hugepages.mount ● ├─dev-mqueue.mount ● ├─kmod-static-nodes.service ● ├─ldconfig.service ● ├─lvm2-lvmetad.socket ● ├─lvm2-lvmpolld.socket ● ├─lvm2-monitor.service ● ├─proc-sys-fs-binfmt_misc.automount ● ├─sys-fs-fuse-connections.mount ● ├─sys-kernel-config.mount ● ├─sys-kernel-debug.mount ● ├─sys-kernel-tracing.mount ● ├─systemd-ask-password-console.path ● ├─systemd-binfmt.service ● ├─systemd-boot-system-token.service ● ├─systemd-firstboot.service ● ├─systemd-hwdb-update.service ● ├─systemd-journal-catalog-update.service ● ├─systemd-journal-flush.service ● ├─systemd-journald.service ● ├─systemd-machine-id-commit.service ● ├─systemd-modules-load.service ● ├─systemd-random-seed.service ● ├─systemd-repart.service ● ├─systemd-sysctl.service ● ├─systemd-sysusers.service ● ├─systemd-timesyncd.service ● ├─systemd-tmpfiles-setup-dev.service ● ├─systemd-tmpfiles-setup.service ● ├─systemd-udev-trigger.service ● ├─systemd-udevd.service ● ├─systemd-update-done.service ● ├─systemd-update-utmp.service ● ├─cryptsetup.target ● ├─local-fs.target ● │ ├─-.mount ● │ ├─boot-efi.mount ● │ ├─systemd-fsck-root.service ● │ ├─systemd-remount-fs.service ● │ └─tmp.mount ● └─swap.target
OK, so I was looking for a dependency on systemd-modules-load.service. I added it to the thinkfan service file, but it also seems to be part of sysinit.target on which thinkfan has always depended. So thinkfan should start only after all modules have been initialized. Assuming systemd dependency handling works correctly, the only explanation for this issue would be that a module takes very long to initialize its sysfs files after being loaded.
I was always under the impression that an application like thinkfan shouldn't be responsible for this problem. This is dependency management and therefore it's the job of the system's service manager. So technically, this should be taken up with either systemd or the kernel itself.
However for the sake of practicality, I might implement a workaround thinkfan. For the time being, affected users should find some other way to delay thinkfan startup. For examle:
Do systemctl edit thinkfan and add the following:
[Unit]
After=graphical.target
This should delay thinkfan startup until after the graphical login manager has started, which should hopefully be enough.
Thanks for your help! If you want, I can try to make a bug report, however, my knowledge of systemd is very limited, so I might not be able to answer their questions.
I tried your suggested workaround, and it resulted in a service being dead (inactive), not starting up.
[Unit] After=graphical.target
so I tried instead basic.target and that just meant that the issue continued. Using gdm.service fixed the issue.
Printed sudo systemd-analyze critical-chain
graphical.target @2.120s
└─multi-user.target @2.120s
└─thinkfan.service @2.060s +60ms
└─gdm.service @2.052s +6ms
└─optimus-manager.service @1.051s +999ms
└─basic.target @1.043s
└─sockets.target @1.043s
└─snapd.socket @1.043s +534us
└─sysinit.target @1.041s
└─systemd-backlight@backlight:intel_backlight.service @1.646s +2.195s
└─system-systemd\x2dbacklight.slice @1.445s
└─system.slice @231ms
└─-.slice @231ms
Anyways, thanks for this helpful report. Since the error says Error scanning /sys/devices/platform/coretemp.0/hwmon, I would think that it's related to the coretemp module. It could also be an issue with the module, e.g. if it completes its initialization procedure before having set up all sysfs symlinks.
Oh, and there's also the possibility that something else is actually un-loading and re-loading modules during bootup. If that happens during thinkfan startup, there's bound to be trouble.
So I think the precise cause of the issue should be investigated more closely before reporting bugs anywhere else.
I have had what I think is the same issue on OpenSUSE for a few days. I guess it appeared after a recent kernel update to 5.9.14, but it might have crept in earlier. I have coretemp, acpitz and pch-skylake hwmons in my /etc/thinkfan.conf. It systematically fails on bootup with error messages like this:
> sudo systemctl status thinkfan
● thinkfan.service - Thinkfan, the minimalist fan control program
Loaded: loaded (/usr/lib/systemd/system/thinkfan.service; enabled; vendor preset: disabled)
Drop-In: /etc/systemd/system/thinkfan.service.d
└─override.conf
Active: failed (Result: exit-code) since Sun 2020-12-20 13:21:04 CET; 43s ago
Process: 890 ExecStart=/usr/sbin/thinkfan $THINKFAN_ARGS (code=exited, status=1/FAILURE)
Dez 20 13:21:04 calypso systemd[1]: Starting Thinkfan, the minimalist fan control program...
Dez 20 13:21:04 calypso thinkfan[890]: ERROR: /etc/thinkfan.conf:52:
name: coretemp
^
Could not find a hwmon with this name.
Dez 20 13:21:04 calypso systemd[1]: thinkfan.service: Control process exited, code=exited, status=1/FAILURE
Dez 20 13:21:04 calypso systemd[1]: thinkfan.service: Failed with result 'exit-code'.
Dez 20 13:21:04 calypso systemd[1]: Failed to start Thinkfan, the minimalist fan control program.
Note that this is not limited to the coretemp module; if I disable that in /etc/thinkfan.conf it will show the same message with one of the other hwmons.
On my OpenSUSE system, to fix the problem it seems to be enough to delay loading a bit by doing sudo systemctl edit thinkfan and add
[Unit]
After=basic.target
On my laptop basic.target is reached after about 2 seconds, which seems to be enough to bring up the sysfs symlinks. If I try graphical.target, I get the same error as @lakotamm with an inactive (dead) service.
Just thought of another thing: The coretemp module is probably loaded on-demand and I'm not sure exactly how/when that is triggered. So you could try loading it explicitly at bootup via systemd-modules-load.service:
echo coretemp > /etc/modules-load.d/coretemp.conf
I had the same issue, solved it by making the service file restartable on failure:
[Service]
Restart=on-failure
RestartSec=5s
I just encountered the same issue on Fedora Silverblue 35. Adding 10s startup delay is still an easy fix.
Fix is currently in development at https://github.com/vmatare/thinkfan/tree/init_retries. For now, the branch simply makes the optional keyword work at startup, too, providing a workaround for this problem. Soon I'll add another option max_errors which will allow a certain number of consecutive init or read errors to be ignored on a given sensor. That'll become the designated way to handle the startup issues in the 2.0 release.
@lakotamm @ceigh @broccoli5 @phrxmd, this should be fixed in master now with a new config setting: https://github.com/vmatare/thinkfan/issues/189#issuecomment-1196011925