> > It is not crash in user space. Execution of the user space
binary
> > sensors-detect is causing kernel space OOPS.
> >
> > This is what is printed on the terminal on executing this program:
>
> I don't believe sensors-detect should be used on the RPi, in fact on
> any ARM device as it's very much focused on servers/x86 and accessing
> things via DMI/SMbus and other pretty x86 methods, it uses rudimentary
> techniques to just bash random locations where there might be sensors.
>
> What are you actually trying to achieve with it? The Pi OOTB has no
> onboard sensors of note, and the addons should likely be
> detected/accessed via GPIO or the IIO stack.
I don't believe that any binary provided by distribution packages should be
able to generate OOPS :)
Yep, that's a fair point, but also lmsensors is also "special" (IE
garbage), I'll report it upstream but ultimately it's won't be fixed
in the short term and I bet if you run lmsensors on most ARM devices
you'll get another crash and I don't have the time or inclination to
report them all upstream.
Initializing sensors was part of my standard template system
initialization
after installation which I'm using on normal/prod systems.
You can call it coincidence that I found it :)
It's not the first time I've seen something like that from lmsensors,
in the short term I suggest adding a check for arch and not doing
lmsensors on ARM devices (tmon in the kernel-tools package works on
most ARM devices for temp/thermal)
PS. BTW sensors under Linux. At the moment it is total mess in Linux
now.
Thing like temp sensors are exposed over three completely independent kernel
interfaces: hwmon, thermal and yet another one which I do not remember now.
After man years of the development Linux still is decade if not more behind
for example Solaris FMA.
You're not telling me anything I don't already know. You actually miss
at least one, I think I know of 5 interfaces, the one I think you're
thinking of is IIO (Industrial I/O). I can so a massive rant on that,
and when you then get into things like PWM, which should be part of
the IIO framework, it goes even further. Most of the client libraries
are absolute garbage, and there's even some that choose to
re-implement all the already existing kernel space drivers in
userspace and just use i2c/gpio kernel drivers to speak to them. I can
rant on this for a long time....