Re: adm1026hm

["Frank Wille" <frank%phoenix.owl.de@localhost>]

On 2026-11-13, Julian Coleman <jdc%coris.org.uk@localhost> wrote:

> For my output, I first ran `envstat -c <file>` with this config:
>
> https://ftp.netbsd.org/pub/NetBSD/misc/jdc/sparc64/v240-envsys.conf
>
> However, that expects only 3 chassis fans ...

Yes. I get an error with it, for exactly that reason:

sun# envstat -c v240-envsys.conf
envstat: invalid value for 'warning-min' in `sensor10'
envstat: please fix the configuration file!

Then I tried to remove the sensor properties again, by "envstat -S",
and the command didn't return...

The whole server seemed frozen and didn't respond anymore. After
some time I noticed it had rebooted. So I repeated the test and
monitored the serial console. This happens on "envstat -S":

[ 655.9500067] cpu0: data fault: pc=140b968 rpc=1 addr=0
[ 655.9500067] Skipping crash dump on recursive panic
[ 655.9500067] panic: kernel fault
[ 655.9500067] cpu0: Begin traceback...
[ 655.9500067] cpu0: End traceback...
[ 655.9500067] cpu1: shutting down
[ 655.9500067] cpu0: rebooting

Unfortunately no crash dump.


> ... Yes - the V240 has 3 fans, compared to 4 in the V210, so the output for
> fan 3 here is (sort of) expected. However, I'm confused as to why it appears
> at all and also that the CPU fan speeds are doubled.

Perhaps the accuracy of reported RPM values is quite rough, so two
sensors will sometimes show the same value. I ran envstat multiple
times and eventually got an output where all sensors differ.


> Looking at the code in -current:
>
> https://nxr.netbsd.org/xref/src/sys/dev/i2c/adm1026.c#270
> https://nxr.netbsd.org/xref/src/sys/dev/i2c/adm1026.c#482
>
> we can override the chip configuration to set which fans are present, and
> we do this for the V240:

I see. That's strange indeed.


> I know that set_i2c_bus_props() is called and the machine model is
> SUNW,Sun-Fire-V240, because without those we won't have the adm1026hm nor
> the lmtemp devices. However, I'm puzzled why either we don't call
> set_i2c_dev_props() or we don't match on pcfiic or on SUNW,Sun-Fire-V240.
>
> I'm away from my machines for a couple of weeks, so I can look when I get
> back. If you're inclined, a debug printf in set_i2c_dev_props() to check
> that it's actually called might show the problem.

Might do that. I'm currently checking out the 11-branch, because I
also wanted to build a custom kernel.

-- 
Frank Wille