linux-kernel - [RFC PATCH 5/6] perf: Disable PMU around address filter adjustment

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20160923112726.5890-6-alexander.shishkin@linux.intel.com>
Date:   Fri, 23 Sep 2016 14:27:25 +0300
From:   Alexander Shishkin <alexander.shishkin@...ux.intel.com>
To:     Peter Zijlstra <a.p.zijlstra@...llo.nl>
Cc:     Ingo Molnar <mingo@...hat.com>, linux-kernel@...r.kernel.org,
        vince@...ter.net, eranian@...gle.com,
        Arnaldo Carvalho de Melo <acme@...radead.org>,
        tglx@...utronix.de, ak@...ux.intel.com,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>
Subject: [RFC PATCH 5/6] perf: Disable PMU around address filter adjustment

If the PMU is used for AUX data sampling, the hardware event that triggers
it may come in during a critical section of address filters adjustment (if
it's delivered as NMI) and in turn call the address filter sync code,
causing a spinlock recursion.

To prevent this, disable the PMU around these critical sections, so that
AUX sampling won't take place there.

Signed-off-by: Alexander Shishkin <alexander.shishkin@...ux.intel.com>
---
 kernel/events/core.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index fdb20fdeb1..f6582df1c9 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -6377,6 +6377,12 @@ static void perf_event_addr_filters_exec(struct perf_event *event, void *data)
 	if (!has_addr_filter(event))
 		return;
 
+	/*
+	 * if sampling kicks in in the critical section,
+	 * we risk spinlock recursion on the ifh::lock
+	 */
+	perf_pmu_disable(event->pmu);
+
 	raw_spin_lock_irqsave(&ifh->lock, flags);
 	list_for_each_entry(filter, &ifh->list, entry) {
 		if (filter->inode) {
@@ -6387,6 +6393,8 @@ static void perf_event_addr_filters_exec(struct perf_event *event, void *data)
 		count++;
 	}
 
+	perf_pmu_enable(event->pmu);
+
 	if (restart)
 		event->addr_filters_gen++;
 	raw_spin_unlock_irqrestore(&ifh->lock, flags);
@@ -6942,6 +6950,8 @@ static void __perf_addr_filters_adjust(struct perf_event *event, void *data)
 	if (!file)
 		return;
 
+	perf_pmu_disable(event->pmu);
+
 	raw_spin_lock_irqsave(&ifh->lock, flags);
 	list_for_each_entry(filter, &ifh->list, entry) {
 		if (perf_addr_filter_match(filter, file, off,
@@ -6953,6 +6963,8 @@ static void __perf_addr_filters_adjust(struct perf_event *event, void *data)
 		count++;
 	}
 
+	perf_pmu_enable(event->pmu);
+
 	if (restart)
 		event->addr_filters_gen++;
 	raw_spin_unlock_irqrestore(&ifh->lock, flags);
@@ -8126,6 +8138,8 @@ static void perf_event_addr_filters_apply(struct perf_event *event)
 
 	down_read(&mm->mmap_sem);
 
+	perf_pmu_disable(event->pmu);
+
 	raw_spin_lock_irqsave(&ifh->lock, flags);
 	list_for_each_entry(filter, &ifh->list, entry) {
 		event->addr_filters_offs[count] = 0;
@@ -8144,6 +8158,8 @@ static void perf_event_addr_filters_apply(struct perf_event *event)
 	event->addr_filters_gen++;
 	raw_spin_unlock_irqrestore(&ifh->lock, flags);
 
+	perf_pmu_enable(event->pmu);
+
 	up_read(&mm->mmap_sem);
 
 	mmput(mm);
-- 
2.9.3