Re: [patch 0/4] Take care of cpufreq lockdep issues (take 2)

[Ingo Molnar <mingo@elte.hu>]

* Pallipadi, Venkatesh <venkatesh.pallipadi@intel.com> wrote:

>  
> 
> >-----Original Message-----
> >From: Ingo Molnar [mailto:mingo@elte.hu] 
> >Sent: Thursday, July 02, 2009 11:54 PM
> >To: Pallipadi, Venkatesh
> >Cc: Dave Jones; linux-kernel@vger.kernel.org; 
> >cpufreq@vger.kernel.org; kernel-testers@vger.kernel.org; 
> >Rafael J. Wysocki; Dave Young; Pekka Enberg; Mathieu 
> >Desnoyers; Thomas Renninger
> >Subject: Re: [patch 0/4] Take care of cpufreq lockdep issues (take 2)
> >
> >
> >* venkatesh.pallipadi@intel.com <venkatesh.pallipadi@intel.com> wrote:
> >
> >> Since recent chanegs to ondemand and conservative governor, there 
> >> have been multiple reports of lockdep issues in cpufreq. Patch 
> >> series takes care of these problems.
> >> 
> >> This is the next attempt following the one here, which was not a 
> >> complete fix. 
> >> http://lkml.indiana.edu/hypermail/linux/kernel/0906.3/01073.html
> >> 
> >> I am currently running some stress tests to make sure there are no 
> >> issues with these patches. But, wanted to send them out for 
> >> review/comments/testing before I head out for the long weekend.
> >> 
> >> If this patchset seems sane, the first patch in the patchset 
> >> should also get into 30.stable.
> >
> >Btw., FYI, because my test-systems were frequently triggering those 
> >bugs, i kept testing the following series from you and Mathieu in 
> >-tip:
> >
> > ecf8b04: cpufreq: Define dbs_mutex purpose and cleanup its 
> >usage conservative gov
> > b08c597: cpufreq: Define dbs_mutex purpose and cleanup its usage
> > 0807e30: cpufreq: remove rwsem lock from CPUFREQ_GOV_STOP 
> >call (second call site)
> >
> >So that fix-series, while probably not complete (given that you sent 
> >a v2 series), worked well in practice and gets my:
> >
> > Tested-by: Ingo Molnar <mingo@elte.hu>
> >
> >Is the delta between this (tested) series and your v2 version 
> >significant? If not it might make sense to shape it as a delta patch 
> >to the v1 series, if that looks clean enough - to preserve testing 
> >results.
> >
> 
> Thanks for testing. That earlier version even though it took care 
> of lockdep complaints, did not address all the race conditions 
> properly. The delta is significant as I had to change the approach 
> compared to first patchset. So, diff will not be very clean.

Fair enough - these cases are when it makes sense to do a clean 
rebase.

	Ingo