Hi Aurelien,
Remember this bug with bullseye and power sensors on the apu2 platform on
idle?
It is fixed with the latest firmware from pcengines.ch (apu2_v4.17.0.3.rom).
Funnily it never appeared on Debian buster nor older releases, only on Debian
bullseye (actually it happened with kernel 5 only).
The good news it is fixed.
root@fake-router:~# while :; do sleep 5; sensors | grep power1; done
power1: 1.63 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.36 W (interval = 0.01 s, crit = 6.00 W)
power1: 1.76 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.19 W (interval = 0.01 s, crit = 6.00 W)
power1: 1.81 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.24 W (interval = 0.01 s, crit = 6.00 W)
power1: 1.86 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.67 W (interval = 0.01 s, crit = 6.00 W)
power1: 1.41 W (interval = 0.01 s, crit = 6.00 W)
power1: 932.00 uW (interval = 0.01 s, crit = 6.00 W)
power1: 687.00 uW (interval = 0.01 s, crit = 6.00 W)
power1: 2.32 W (interval = 0.01 s, crit = 6.00 W)
power1: 1.86 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.28 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.04 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.08 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.24 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.48 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.43 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.54 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.92 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.50 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.47 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.54 W (interval = 0.01 s, crit = 6.00 W)
power1: 2.67 W (interval = 0.01 s, crit = 6.00 W)
Have a nice day.
For reference, my complete old mail:
On Tue Oct 18 16:56:52 2022, schaefer wrote:
Salut Aurélien,
Hi Aurelien,
I am writing to you directly instead of creating a bug report, to have
an idea where to look, first.
The problem
On an apu2 platform, buster:
schaefer@apu2-ds-01:~$ uname -a
Linux apu2-ds-01 4.19.0-22-amd64 #1 SMP Debian 4.19.260-1 (2022-09-29) x86_64 GNU/Linux
schaefer@apu2-ds-01:~$ sensors
fam15h_power-pci-00c4
Adapter: PCI adapter
power1: 6.40 W (interval = 0.01 s, crit = 6.00 W)
k10temp-pci-00c3
Adapter: PCI adapter
temp1: +51.8 C (high = +70.0 C)
(crit = +105.0 C, hyst = +104.0 C)
Those two above values are coherent with direct measurements.
On bullseye:
Linux apu2-test 5.10.0-18-amd64 #1 SMP Debian 5.10.140-1 (2022-09-02) x86_64 GNU/Linux
fam15h_power-pci-00c4
Adapter: PCI adapter
power1: 97.78 W (interval = 0.01 s, crit = 6.00 W)
k10temp-pci-00c3
Adapter: PCI adapter
temp1: +54.6 C (high = +70.0 C)
(crit = +105.0 C, hyst = +104.0 C)
Temperature looks ok, wattage is wrong.
I did not see much in the Debian changelog nor upstream changelog for
lm-sensors (added additional versions of the AMD chips). Also, I diffed
the kernel source for fam15h_power (without Debian patches!) between
those two kernel releases and I saw nothing (except a few defines,
kept their values but were migrated to include files).
Do you have any idea where to look (kernel or lm-sensors)?
Thank you for any pointer.
