Hi,
while trying on the global ondemand sysfs and machine wide
governor switching I run into two bugs introduced lately.
This patch series fixes the two bugs.
Also the set_speed changes should be rather harmless.
The last three patches are about the global sysfs directory:
/sys/devices/system/cpu/cpufreq
- One patch for creating it
- One patch for ondemand using it
- One patch for machine wide governor switching
-> different governors per core will not be allowed anymore then.
Still all current sysfs files are usable the same way
(beside that chaning a governor of a cpu will change all and a slight
modification of set_speed file -> will now be created when the userspace
governor gets activated and will be removed when switching to another
governor).
The last patch is the most intrusive one, especially store_scaling_governor().
I did a parallel online/offlining, governor switching and showing stress test
which identified the first two bugs. It even found a NULL pointer in
2.6.27 when I wanted to find out whether this works at all.
It now seem to work, beside a little, IMO not sever side effect which
is described in the patch itself.
Further cleanups possible:
- Provide a global (not per core) governor start/exit entry func
- Rip out all the deprecated things in a while
- I forgot the conservative governor..., no time right now, I will
send something if this gets accepted
Comments?
Can this make it into next?
Thanks,
Thomas
--
To unsubscribe from this list: send the line "unsubscribe cpufreq" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
