Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S936217AbcJTI6a (ORCPT ); Thu, 20 Oct 2016 04:58:30 -0400 Received: from mx1.redhat.com ([209.132.183.28]:50996 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S936030AbcJTI6G (ORCPT ); Thu, 20 Oct 2016 04:58:06 -0400 Date: Thu, 20 Oct 2016 10:58:03 +0200 From: Jiri Olsa To: Peter Zijlstra Cc: CAI Qian , Rob Herring , Kan Liang , Greg Kroah-Hartman , linux-kernel , Ingo Molnar Subject: Re: [4.9-rc1+] intel_uncore builtin + CONFIG_DEBUG_TEST_DRIVER_REMOVE kernel panic Message-ID: <20161020085803.GA31721@krava> References: <907882571.66590.1476113724660.JavaMail.zimbra@redhat.com> <1219480016.67057.1476113847440.JavaMail.zimbra@redhat.com> <20161010172023.GA7148@kroah.com> <1035662571.647973.1476888331396.JavaMail.zimbra@redhat.com> <20161019191943.GA7951@krava> <20161020053944.GQ3102@twins.programming.kicks-ass.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20161020053944.GQ3102@twins.programming.kicks-ass.net> User-Agent: Mutt/1.7.1 (2016-10-04) X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.32]); Thu, 20 Oct 2016 08:58:06 +0000 (UTC) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2258 Lines: 82 On Thu, Oct 20, 2016 at 07:39:44AM +0200, Peter Zijlstra wrote: > On Wed, Oct 19, 2016 at 09:19:43PM +0200, Jiri Olsa wrote: > > I think the reason here is that presume pmu devices are always added, > > but we add them only if pmu_bus_running (in perf_event_sysfs_init) > > is set which might happen after uncore initcall > > > > attached patch fixes the issue for me > > Right, we never expected to be unloaded before userspace runs. > > Strictly speaking we should only read pmu_bus_running while holding > pmus_lock, that way we're serialized against perf_event_sysfs_init() > flipping it while we're being removed etc.. > > With the current setup the introduced race is harmless, but who knows > what other crazy these device people will come up with ;-) > right, did not think of that ;-) also I did not noticed device_remove_file call for pmu->nr_addr_filters and we could save one lock/unlock call later.. I'm testing attached patch now thanks, jirka --- diff --git a/kernel/events/core.c b/kernel/events/core.c index c6e47e97b33f..224dffbc3b9b 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -8581,24 +8581,24 @@ static void update_pmu_context(struct pmu *pmu, struct pmu *old_pmu) } } +/* + * The pmus_lock lock must be taken. + */ static void free_pmu_context(struct pmu *pmu) { struct pmu *i; - mutex_lock(&pmus_lock); /* * Like a real lame refcount. */ list_for_each_entry(i, &pmus, entry) { if (i->pmu_cpu_context == pmu->pmu_cpu_context) { update_pmu_context(i, pmu); - goto out; + return; } } free_percpu(pmu->pmu_cpu_context); -out: - mutex_unlock(&pmus_lock); } /* @@ -8869,11 +8869,15 @@ void perf_pmu_unregister(struct pmu *pmu) free_percpu(pmu->pmu_disable_count); if (pmu->type >= PERF_TYPE_MAX) idr_remove(&pmu_idr, pmu->type); - if (pmu->nr_addr_filters) - device_remove_file(pmu->dev, &dev_attr_nr_addr_filters); - device_del(pmu->dev); - put_device(pmu->dev); + mutex_lock(&pmus_lock); + if (pmu_bus_running) { + if (pmu->nr_addr_filters) + device_remove_file(pmu->dev, &dev_attr_nr_addr_filters); + device_del(pmu->dev); + put_device(pmu->dev); + } free_pmu_context(pmu); + mutex_unlock(&pmus_lock); } EXPORT_SYMBOL_GPL(perf_pmu_unregister);