From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754560AbbAHMKE (ORCPT ); Thu, 8 Jan 2015 07:10:04 -0500 Received: from cam-admin0.cambridge.arm.com ([217.140.96.50]:61158 "EHLO cam-admin0.cambridge.arm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750744AbbAHMKD (ORCPT ); Thu, 8 Jan 2015 07:10:03 -0500 Date: Thu, 8 Jan 2015 12:09:34 +0000 From: Mark Rutland To: Will Deacon Cc: "linux-kernel@vger.kernel.org" , Peter Zijlstra , Paul Mackerras , Ingo Molnar , Arnaldo Carvalho de Melo Subject: Re: [PATCH] perf: drop module reference on event init failure Message-ID: <20150108120934.GB10537@leverpostej> References: <1420642611-22667-1-git-send-email-mark.rutland@arm.com> <20150108113212.GE11583@arm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150108113212.GE11583@arm.com> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Jan 08, 2015 at 11:32:13AM +0000, Will Deacon wrote: > On Wed, Jan 07, 2015 at 02:56:51PM +0000, Mark Rutland wrote: > > When initialising an event, perf_init_event will call try_module_get to > > ensure that the PMU's module cannot be removed for the lifetime of the > > event, with __free_event dropping the reference when the event is > > finally destroyed. If something fails after the event has been > > initialised, but before the event is installed, perf_event_alloc will > > drop the reference on the module. > > > > However, if we fail to initialise an event for some reason (e.g. we ask > > an uncore PMU to perform sampling, and it refuses to initialise the > > event), we do not drop the refcount. If we try to open such a bogus > > event without a precise IDR type, we will loop over each PMU in the pmus > > list, incrementing each of their refcounts without decrementing them. > > > > This patch adds a module_put when pmu->event_init(event) fails, ensuring > > that the refcounts are balanced in failure cases. As the innards of the > > precise and search based initialisation look very similar, this logic is > > hoisted out into a new helper function. While the early return for the > > failed try_module_get is removed from the search case, this is handled > > by the remaining return when ret is not -ENOENT. > > > > Signed-off-by: Mark Rutland > > Cc: Will Deacon > > Cc: Peter Zijlstra > > Cc: Paul Mackerras > > Cc: Ingo Molnar > > Cc: Arnaldo Carvalho de Melo > > --- > > kernel/events/core.c | 28 ++++++++++++++++------------ > > 1 file changed, 16 insertions(+), 12 deletions(-) > > Looks like a straightforward fix to me -- does it need to go to stable too? Assuming people are using PMU drivers built as modules on stable kernels, then yes. Otherwise it's not strictly necessary as far as I can see. The only such driver I'm immediately aware of is for the ARM CCN, but it's entirely possible I missed another. I guess this fixes c464c76eec4be587 (perf: Allow building PMU drivers as modules). > Acked-by: Will Deacon Cheers. Mark. > > Will > > > diff --git a/kernel/events/core.c b/kernel/events/core.c > > index 4c1ee7f..4faccf3 100644 > > --- a/kernel/events/core.c > > +++ b/kernel/events/core.c > > @@ -6857,6 +6857,20 @@ void perf_pmu_unregister(struct pmu *pmu) > > } > > EXPORT_SYMBOL_GPL(perf_pmu_unregister); > > > > +static int perf_try_init_event(struct pmu *pmu, struct perf_event *event) > > +{ > > + int ret; > > + > > + if (!try_module_get(pmu->module)) > > + return -ENODEV; > > + event->pmu = pmu; > > + ret = pmu->event_init(event); > > + if (ret) > > + module_put(pmu->module); > > + > > + return ret; > > +} > > + > > struct pmu *perf_init_event(struct perf_event *event) > > { > > struct pmu *pmu = NULL; > > @@ -6869,24 +6883,14 @@ struct pmu *perf_init_event(struct perf_event *event) > > pmu = idr_find(&pmu_idr, event->attr.type); > > rcu_read_unlock(); > > if (pmu) { > > - if (!try_module_get(pmu->module)) { > > - pmu = ERR_PTR(-ENODEV); > > - goto unlock; > > - } > > - event->pmu = pmu; > > - ret = pmu->event_init(event); > > + ret = perf_try_init_event(pmu, event); > > if (ret) > > pmu = ERR_PTR(ret); > > goto unlock; > > } > > > > list_for_each_entry_rcu(pmu, &pmus, entry) { > > - if (!try_module_get(pmu->module)) { > > - pmu = ERR_PTR(-ENODEV); > > - goto unlock; > > - } > > - event->pmu = pmu; > > - ret = pmu->event_init(event); > > + ret = perf_try_init_event(pmu, event); > > if (!ret) > > goto unlock; > > > > -- > > 1.9.1 > > >