[ltp] Re: Temperature sensors on X201s

Stefan Monnier linux-thinkpad@linux-thinkpad.org
Tue, 11 May 2010 16:26:15 -0400

>> - cat /proc/acpi/ibm/thermal gives me just:
>> temperatures:   43 0 0 0 0 0 0 0
>> now 43� is fine, but how come there's only 1 value there.  Are there

> Pester Lenovo in their forums about it...  Either there are no sensors
> anymore, or they decided to hide them because of the windows fan-control
> crowd or some other random reason.

Maybe Lenovo's idea is that we should use the "standard" sensors rather
than the thinkpad-acpi ones.  I.e. tp-smapi still provides sensors for
the batteries, hddtemp/smartctl still provides sensors for the drives,

> Note that the chipsets, memory (at least on servers), and a lot of other
> stuff has thermal sensors embedded nowadays.  It is just that getting to
> them is often quite hard.

Right, it looks like the i7 sensors aren't supported by my current
lm-sensors but that it's already supported upstream.

> There is a firmware thermal alarm now, though, so at least you will get
> an alert if things are about to catch fire.  Useless for any sort of
> runtime tunning...

In all fairness, this X201s seems to be nice and cool (and quiet, partly
as a consequence) at least the way I use it.

>> - /sys/class/power_supply/BAT0/current_now seems to sometimes report
>> power rather than current (i.e. sometimes it's 1000 times
>> /sys/devices/platform/smapi/BAT0/current_avg and sometimes it's 1000
>> times /sys/devices/platform/smapi/BAT0/power_avg).  Not clear when it
>> changes from one to the other.  powertop also seems confused about it
>> (when current_now correctly gives current, powertop seems to interpret
>> it as power).
> Report this as a bug in bugzilla.kernel.org or the distro, please.  But
> make sure you're using the very latest firmware, first.

I did update the firmware to the latest version already, as it is needed
for reliable suspend under GNU/Linux.

>> - Xorg only works with KMS.  Without it, it crashes the whole system
>> (black screen, typically with blinking caps-lock led, tho not always).
> Report it through the distro, I think.  KMS pratically requires
> bleeding-edge backports in the kernel and userspace...

OK, will do.

>> - hibernate tends to corrupt some part of the memory (I have the same
>> thing on my MSI Wind, where disabling KMS works around the problem).
> Yikes.  Report to distro.

it's actually already at https://bugs.freedesktop.org/show_bug.cgi?id=26825