When on level "0": Unexpected short fan activation.

[kolibril13]

I have a Thinkpad T495 Ubuntu 20.4 and want to use Thinkfan, as I don't want to have fan noises when e.g. watching videos. I installed thinkfan 0.9.1, and it workes more or less, except that about every two minutes, the fan is starting to spin but also stopping immediately (starting+stopping takes maybe 3 seconds) at low temperatures (~45°C), where I would expect level 0. My file /etc/thinkfan.conf looks like this:

hwmon /sys/devices/pci0000:00/0000:00:18.3/hwmon/hwmon2/temp1_input

(0, 0,  85)
(1, 85, 86)
(2, 86, 87)
(3, 87, 88)
(4, 88, 89)
(5, 89, 90)
(6, 90, 91)
(7, 91, 32767)

Can someone maybe give me any advice? My theory is that the Ubuntu internal and the think fan scripts to control the fan are interfering somehow. Thanks a lot!

[vmatare]

Hi, that sounds like a bug that has been fixed for a very, very long time now. I think it should already be fixed in the 0.9.3 release, and that would be from 2016. Since then thinkfan has undergone a complete redesign and we're currently at 1.2.1.

Someone (preferrably a Ubuntu user, cough cough) should ping the package maintainer and tell them to update ;-)

[vmatare]

oh, and are you actually serious about keeping the fan off until your temperature has reached 85°C? I don't know what you're (not) cooling there, but 85°C sounds awfully hot for any device...

[kolibril13]

Thanks a lot, @vmatare ! When googling thinkfan, there are really few hints that there is a newer version. Maybe one can also note in the latest release in the wiki, https://github.com/vmatare/thinkfan/wiki , as there are already the other releases mentioned. I can ping the package maintainer, can you tell me where to find them? Is there an email address, or a website? At the moment, when one types sudo apt-get install thinkfan it will install thinkfan 0.9.1

[kolibril13]

serious about keeping the fan off until your temperature has reached 85°C?

That was only a test script, I don't intend to use it further. This is how my current script looks:

sensors:
  - hwmon: /sys/devices/pci0000:00/0000:00:18.3/hwmon/hwmon2/temp1_input

levels:
  - [0, 0, 65]
  - ["level auto", 55, 75]

However, I still have the problem of unexpected short fan activation. I just measured the time, exactly every 112 seconds the fan starts to spin, and stops then again, which is also quite annoying. Any ideas of how I can suppress that?

[vmatare]

However, I still have the problem of unexpected short fan activation. I just measured the time, exactly every 112 seconds the fan starts to spin, and stops then again, which is also quite annoying. Any ideas of how I can suppress that?

Yes, upgrade at least to 0.9.3. Even Debian have it in their stable repos: https://packages.debian.org/stable/thinkfan In fact, you might be able to just install the debian package.

And then, to wake up the Ubuntu people you should report it as a Ubuntu bug: https://bugs.launchpad.net/ubuntu

[kolibril13]

Yes, upgrade at least to 0.9.3.

What I forgot to say, the problem is at the moment with thinkfan version 1.2 , so downgrading to 0.9.3 might fix it? :)

In fact, you might be able to just install the debian package.

Ok, I will try this then. Thanks a lot for your help, I will get in contact with the ubuntu people! :)

[vmatare]

Sorry, I just saw that thinkfan 0.9.3 incorrectly reports its version number as 0.9.1. So it's likely that your package database says 0.9.3, while thinkfan -h will say 0.9.1. In that case, you probably do in fact have 0.9.3. But as I said, all of the 0.x versions are ANCIENT and I don't support them any more. I might think of publishing a Debian/Ubuntu package myself but right now I don't have the time. So for the time being I'd ask you to uninstall the Ubuntu package and manually upgrade to thinkfan 1.2.1.

Oh, and please don't downgrade, I don't believe that will fix your problem.

If you actually observed this problem with 1.2 or 1.2.1, it might be something I need to look at. So are you saying the behavior is exactly the same as with the old 0.9 version?

[kolibril13]

If you actually observed this problem with 1.2 or 1.2.1, it might be something I need to look at. So are you saying the behaviour is exactly the same as with the old 0.9 version?

sudo thinkfan -h gives me the following output:

thinkfan 1.2: A minimalist fan control program

So for the time being I'd ask you to uninstall the Ubuntu package and manually upgrade to thinkfan 1.2.1.

This is exactly what I did, and the problem is still there. If you want to have a look at that, that would be really great and I would appreciate this a lot! Feel free to contact me if you want to have support.

EDIT: And yes, the behaviour is the exact same as from version 0.9.

[vmatare]

Which command line options are you using when you launch thinkfan? In particular, are you using the -p or -s options?

[kolibril13]

Hi @vmatare , sorry, I somehow overlooked your last message. Tanks again for helping me with this. I tried now both options, -p or -s, both times still the same behaviour with the short fan activation. Here my console output:

$ sudo thinkfan -s 10
/proc/acpi/ibm/fan: Restoring initial state: auto.
Daemon PID: 30651

$ sudo thinkfan -p
Disengaging the fan controller for 5 seconds every 0.500000 seconds
/proc/acpi/ibm/fan: Restoring initial state: auto.
Daemon PID: 30207

[kolibril13]

@vmatare : any ideas maybe? Or will switching from ubuntu to Debian (which I already thought of in another context) solve this problem?

[mhhayashi]

I have the same issue on my Thinkpad T14s (AMD) running Arch.

Whenever /proc/acpi/ibm/fan is written to (even writing 0), the fan will spin for half a second. Since thinkfan seems to write to this every 120 seconds, the fan spins every 120 seconds.

Changing the number on line 84 of drivers.cpp for watchdog_timeout changes how often the fan spins https://github.com/vmatare/thinkfan/blob/eab6f5bf53fc93b7677f541cb5dc92b13aa4243e/src/drivers.cpp#L84

One possible fix could be to disable watchdog. Setting timeout to 0 in code does not seem to work because it will cause thinkfan to repeatedly write to /proc/acpi/ibm/fan, running the fan constantly.

[buck2202]

@vmatare I'm not personally affected by this (the fan in my T430 never gets to zero...), but came here via a reddit post about this issue. The user claims that writing watchdog xx periodically (instead of re-setting the current speed) is enough to both keep the watchdog happy, and doesn't result in momentary fan spin up if the current level is zero. AMD T14 https://www.reddit.com/r/thinkpad/comments/rw2xur/is_fan_control_supposed_to_be_persistent_with/

Thought it was worth mentioning, so maybe someone affected can test/confirm that here.

[pennae]

affected and can confirm. then AMD T14 Gen2 spins the fan up very briefly even when turning it off. this patch is sufficient to stop the 2-minutely spinup while preventing watchdog reset (on 1.3.1)

diff --git a/src/fans.cpp b/src/fans.cpp
index 56a15d7..08c9b4a 100644
--- a/src/fans.cpp
+++ b/src/fans.cpp
@@ -123,7 +123,11 @@ void TpFanDriver::ping_watchdog_and_depulse(const Level &level)
    }
    else if (last_watchdog_ping_ + watchdog_ - sleeptime <= std::chrono::system_clock::now()) {
        log(TF_DBG) << "Watchdog ping" << flush;
-       set_speed(level);
+       std::fstream f(path_);
+       if (!(f.is_open() && f.good()))
+           throw IOerror(MSG_FAN_INIT(path_), errno);
+       if (!(f << "watchdog " << watchdog_.count() << std::flush))
+           throw IOerror(MSG_FAN_INIT(path_), errno);
    }
 }

[peter-stoll]

I'm affected and can confirm that the patch works with 1.3.1 on a T495 (actually a T490 with AMD CPU). The patch is not in the current HEAD. Is there any reason why the patch has not found it's way into HEAD after 1 1/2 years? Or is it just, that nobody found the time to create a pull request or whatever?

[tova-cup]

I'm affected and can confirm that the patch works with 1.3.1 on a T495 (actually a T490 with AMD CPU). The patch is not in the current HEAD. Is there any reason why the patch has not found it's way into HEAD after 1 1/2 years? Or is it just, that nobody found the time to create a pull request or whatever?

@peter-stoll, I stumbled upon this thread while facing the issue myself and I took the matter into my own hands.

A pull request (see above) was submitted ; it is pending review and merging by @vmatare.