Hi,

This series is based on the observation that the policy rwsem is used (or rather not
used) inconsistently in cpufreq_remove_dev() (in summary, it does things without
holding the policy rwsem that are done under that rwsem elsewhere).

The first two patches are preparatory (but patch [1/3] is a good enough improvement
by itself IMO) and patch [3/3] makes the essential change.

Please refer to the patch changelogs for details.

Thanks!