Red Hat Bugzilla – Bug 172655
cpuspeed needs some doc
Last modified: 2007-11-30 17:11:16 EST
Description of problem:
On this system cpuspeed starts but crashes at once :
[root@rousalka nim]# /etc/init.d/cpuspeed start
DÃ©marrage de cpuspeed : [ OK ]
[root@rousalka nim]# /etc/init.d/cpuspeed status
cpuspeed est mort mais subsys est verrouillÃ©
Probably because the default parameters are
OPTS="$OPTS -t /proc/acpi/thermal_zone/*/temperature 75"
but there is no thermal zone on this system
Version-Release number of selected component (if applicable):
Steps to Reproduce:
[root@rousalka xorg]# rpm -q cpuspeed kernel
Created attachment 120796 [details]
Created attachment 120797 [details]
[root@rousalka nim]# find /proc/acpi/
Gigabyte GA-K8N Ultra-9 motherboard
(would probably be easier if there was some doc in the package in the first
place so I would know what the black magic options actually mean)
The system is able to change speed :
powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version 1.50.4)
powernow-k8: 0 : fid 0xe (2200 MHz), vid 0x6 (1400 mV)
powernow-k8: 1 : fid 0x2 (1000 MHz), vid 0x12 (1100 mV)
I'm puzzled. The OPTS line is commented out in the cpuspeed.conf as shipped in
the rpm, so it shouldn't be trying to use it.
ah, I babelfish'd the error msg you get, and it makes a bit more sense.
the thermal_zone is unrelated, its a problem with /var/lock/subsys/cpuspeed
I probably commented out the line when editing the file to set powernow-k8 (or
maybe it was this way at the time, don't remember). Then noticed cpuspeed was
silently crashing a month later. Commenting it again cpuspeed no longuer crashes
(/var/lock/subsys/cpuspeed is the lock leftover after a crash, this is the
standard error message when a service crashed)
I'd probably have found the problem myself if cpuspeed options where documented
somewhere. As it was I could only rely on the file comments, and since I wanted
thermal management I uncommented the wrong line (is there a way to do thermal
management without acpi zones ? This system has working temperature sensors)
So the two problems are :
- cpuspeed is lacking doc, it hasn't got even a README
- when cpuspeed crashes the service still writes a green OK and there's no error
... and my memory is hopeless
thermal stuff without acpi is a bit of a nightmare, as you need chipset specific
drivers (basically lm-sensors), which isn't always reliable.
As for the lack of cpuspeed docs, there's nothing in the upstream tarball, as is
common with far too many other small projects. There is another bugzilla open
on for cpuspeed docs iirc, but it's one of those things that there's always
something else more important to fix, so I never get around to it.
I'll look into why it thought it succeeded after the crash.
lm-sensors works fine here. Owing probably to the fact it's documented, so I
could figure how to make it work :)
cpuspeed --help is really about all there is. I'll work on converting that
output (along with some other bits) into a proper man page, time permitting...
*** Bug 177562 has been marked as a duplicate of this bug. ***
Okay, time permitted today. Pending rawhide build includes both documentation
files that are now in the upstream tarball and a conversion of cpuspeed --help
to a man page (along with some other bits added to it).
Thanks a lot, that's awesome!
Rawhide builds now have a man page and a few docs. Will push to FC5 and FC6 after they've baked in
rawhide for a bit...