From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Srivatsa S. Bhat" Subject: Re: cpufreq_stats NULL deref on second system suspend Date: Wed, 11 Sep 2013 16:44:37 +0530 Message-ID: <5230509D.6040205@linux.vnet.ibm.com> References: <522E1FEF.6080803@wwwdotorg.org> <1775778.MeiRhuYy7o@vostro.rjw.lan> <522F86AD.6010603@wwwdotorg.org> <2521560.SfeNbV74nj@vostro.rjw.lan> <52304439.3030301@linux.vnet.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: Received: from e28smtp08.in.ibm.com ([122.248.162.8]:47159 "EHLO e28smtp08.in.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752017Ab3IKLSf (ORCPT ); Wed, 11 Sep 2013 07:18:35 -0400 Received: from /spool/local by e28smtp08.in.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Wed, 11 Sep 2013 16:36:38 +0530 In-Reply-To: Sender: linux-pm-owner@vger.kernel.org List-Id: linux-pm@vger.kernel.org To: Viresh Kumar Cc: "Rafael J. Wysocki" , Stephen Warren , "linux-pm@vger.kernel.org" , "linux-kernel@vger.kernel.org" , cpufreq On 09/11/2013 04:15 PM, Viresh Kumar wrote: > On 11 September 2013 16:14, Viresh Kumar wrote: >> But I would have solved it differently :) >> >> We don't really need to call update_policy_cpu() again and again >> as we don't really need to update policy->cpu... >> >> Rather it would be better to just move following inside >> cpufreq_policy_alloc(): > > Half written mail sent :( > > What about moving following to cpufreq_policy_alloc(): > > policy->cpu = cpu; > > ?? > Hmm? The problem is not about merely updating the policy->cpu field; the main issue is that the existing code was not letting the cpufreq-stats code know that we updated the policy->cpu under the hood. It is important for cpufreq-stats to know this because it maintains the reference to its stats structure by associating it with the policy->cpu. So if policy->cpu changes under the hood, it loses track of its reference. So we need to keep that code informed about changes to policy->cpu. Thus, we need to call update_policy_cpu() in the CPU online path (during resume). I don't see how we can skip that. Regards, Srivatsa S. Bhat