adjusting broken ACPI thermal trip points

adjusting broken ACPI thermal trip points

[Olaf Hering]

My HP ProBook 6555b has broken trip points. As a result the fan is
either off, or running at full speed until it cooled down, the its off
again.

How is this supposed to be handled?


And the long version is:

Initially I reported https://bugzilla.kernel.org/show_bug.cgi?id=67101
and provided some details about the machine and I was pointed to some
patches in https://bugzilla.kernel.org/show_bug.cgi?id=78201. At some
point the mainboard was replaced by HP, and it got a new firmware.  This
one is appearently (still?) broken, the trip points are wrong. I used
the patches from #78201 and added a hack ontop which forced sane values
in ACPI code. This worked well enough for me to not be constantly
annoyed by the noise changes.

This week I checked the state of #78201 again and noticed some or all of
the suggested changes are in mainline, which is what runs right now.
This allows me to manually start the fan by writing to sysfs, just like
I have done it in the last months. But I have not yet ported my hack for
the ACPI core (acpi_thermal_trips_update) to adjust the hardware trip points.

Thats because I saw during make oldconfig a new config option which is
supposed to avoid such hack. But so far I have not seen the
corresponding sysfs files to tweak the trip points at runtime.

So what should be done for this machine? Fix the firmware during boot
with some custom DSDT as suggested in comment #35, use that new
trip-adjustment interface, or something else?


Another question: how does one control the fan speed anyway? I think the
firmware provides no info about the current state of the fan, if and how
fast it runs. It looks like the only way to decide which of the three
states should be activated is to poll the current temperature and enable
one of the states. Are there already GUI tools to control fan behaviour?

Olaf

Re: adjusting broken ACPI thermal trip points

[Olaf Hering]

On Thu, Oct 01, Olaf Hering wrote:

> Thats because I saw during make oldconfig a new config option which is
> supposed to avoid such hack. But so far I have not seen the
> corresponding sysfs files to tweak the trip points at runtime.

Why does acpi_thermal_register_thermal_zone call
thermal_zone_device_register with mask=0? IMO the core should remember
that trip_point_temp_store was called and reused that value instead of
reusing broken firmware values again and again...

Olaf

Re: adjusting broken ACPI thermal trip points

[Olaf Hering]

On Thu, Oct 01, Olaf Hering wrote:

> Initially I reported https://bugzilla.kernel.org/show_bug.cgi?id=67101

And related to this:

why would the first command enable the fan after boot only sometimes,
while the second one will succeed every time? Is this a kernel or
firmware bug?

# while :;do for i in 2 4 ; do echo 1 > /sys/class/thermal/cooling_device$i/cur_state ; cat /sys/class/thermal/cooling_device$i/cur_state > /dev/null ; done ; sleep 2 ; done

# while :;do for i in 2 4 ; do echo 1 > /sys/class/thermal/cooling_device$i/cur_state ; done ; sleep 2 ; done


Appearently cur_state must be read right after writing to it to have an
effect when selecting fan speeds for 0, 1 or 2.

Olaf