From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?utf-8?Q?Bj=C3=B8rn_Mork?= Subject: Re: [PATCH 1/2] cpufreq: try to resume policies which failed on last resume Date: Fri, 03 Jan 2014 12:55:29 +0100 Message-ID: <87wqihmg9a.fsf@nemi.mork.no> References: <5562479.pVWRuDL0y6@vostro.rjw.lan> <87zjne7f75.fsf@nemi.mork.no> <2302938.b8tymqrMEz@vostro.rjw.lan> <878uuxquxu.fsf@nemi.mork.no> <871u0po0gx.fsf@nemi.mork.no> Mime-Version: 1.0 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: (Viresh Kumar's message of "Fri, 3 Jan 2014 16:49:17 +0530") Sender: linux-kernel-owner@vger.kernel.org List-ID: Content-Type: text/plain; charset="utf-8" To: Viresh Kumar Cc: "Rafael J. Wysocki" , "cpufreq@vger.kernel.org" , "linux-pm@vger.kernel.org" , Linux Kernel Mailing List Viresh Kumar writes: > On 3 January 2014 15:23, Bj=C3=B8rn Mork wrote: >> Note that "ondemand" and "1401000" are the default vaules, so I don'= t >> actually change anything here. The write is causing the problem, no= t >> the value. As expected, I guess. >> >> Also note that boot vs non-boot cpu doesn't seem to matter. Nor doe= s >> cancelling the hibernation. The warning appears on hibernate - not = on >> resume. > > Hmm... I spent quite some time understanding whats going on and reall= y > couldn't get across anything as of now. I haven't tried reproducing i= t though. > > Few things that I can make out of this mail chain so far: > - Apart from the log, everything is working fine. i.e. system is back= in > working condition. Correct. And users not running a lock debugging kernel will of course not even see the warning. > - It only happens when cpufreq_add_dev() fails during hibernation whi= le > we enable non-boot CPUs again to save image to disk. So, isn't a prob= lem > for a system which doesn't have any issues with add_dev() failing on > hibernation Wrong. This was my initial assumption but I later found out that the issue is unrelated to hibernation failures. Sorry about the confusion. > - There is a contention of locks in the order they are taken. And the= contention > looks to be between, hotplug lock taken by cpu_online_cpus() and s_ac= tive > lock for sysfs files. Don't know what's the role of previous write to > sysfs files. > As that should finish before hibernation starts and so all locks shou= ld be back > in place. Yes, that seems logical. But I guess this is where it fails? Bj=C3=B8rn