- Fix the handling of the scaling_cur_freq cpufreq policy attribute
on x86 systems with the MPERF/APERF registers present to make it
behave more as expected after recent changes (Rafael Wysocki).
- Drop a leftover callback from the intel_pstate driver which also
prevents the cpuinfo_cur_freq cpufreq policy attribute from being
incorrectly exposed when intel_pstate works in the active mode
(Rafael Wysocki).
- Add a missing piece describing the cpuinfo_cur_freq policy
attribute to cpufreq documentation (Rafael Wysocki).
- Fix up a recently added part of the Thunderbolt driver to avoid
aborting system suspends if its mailbox commands time out (Rafael
Wysocki).
- Update device runtime PM framework documentation to reflect the
current behavior of the code (Johan Hovold).