Re: [PATCH v2 0/5] Cpufreq frequency serialization fixes

["Rafael J. Wysocki" <rjw@rjwysocki.net>]

On Tuesday, April 29, 2014 12:23:54 AM Srivatsa S. Bhat wrote:
> 
> Hi,
> 
> Meelis Roos reported hangs during boot in the longhaul cpufreq driver, after
> commit 12478cf0c55 (cpufreq: Make sure frequency transitions are serialized).
> The root-cause of this issue is the extra invocation of the
> cpufreq_freq_transition_begin() and cpufreq_freq_transition_end() APIs in the
> longhaul driver. I found similar issues in the powernow-k6 and powernow-k7
> drivers as well. This patchset fixes the issue in all the 3 drivers and also
> adds a debug infrastructure to catch such issues easily.
> 
> Patches 1-4 fix the regression in longhaul, powernow-k6 and powernow-k7
> drivers. (Patch 2 fixes a different bug in powernow-k6, and it is kept as a
> separate patch instead of merging it with patch 3, because I felt that it was
> a bit subtle and needed attention in a separate patch).
> 
> Patch 5 adds a debug infrastructure to the cpufreq core to catch such problems
> more easily in the future.

I've queued up patches [1-4/5] for 3.15, the last one I need to have another
look at tomorrow.

BTW, when you fix regressions, please always add a Fixes: tag to the changelog.

Thanks!


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.