[tip:perf/core] perf: Fix 'perf sched record' deadlock

From: tip-bot for Mike Galbraith
Date: Fri Apr 02 2010 - 15:08:27 EST

Commit-ID: 8bb39f9aa068262732fe44b965d7a6eb5a5a7d67
Gitweb: http://git.kernel.org/tip/8bb39f9aa068262732fe44b965d7a6eb5a5a7d67
Author: Mike Galbraith <efault@xxxxxx>
AuthorDate: Fri, 26 Mar 2010 11:11:33 +0100
Committer: Ingo Molnar <mingo@xxxxxxx>
CommitDate: Fri, 2 Apr 2010 19:30:05 +0200

perf: Fix 'perf sched record' deadlock

perf sched record can deadlock a box should the holder of
handle->data->lock take an interrupt, and then attempt to
acquire an rq lock held by a CPU trying to acquire the
same lock. Disable interrupts.

sched event with rq->lock held
grab handle->data->lock
spin on handle->data->lock
try to grab rq->lock

Reported-by: Li Zefan <lizf@xxxxxxxxxxxxxx>
Signed-off-by: Mike Galbraith <efault@xxxxxx>
Tested-by: Li Zefan <lizf@xxxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Cc: Frederic Weisbecker <fweisbec@xxxxxxxxx>
LKML-Reference: <1269598293.6174.8.camel@xxxxxxxxxxxxxxxx>
Signed-off-by: Ingo Molnar <mingo@xxxxxxx>
kernel/perf_event.c | 15 ++++++++++++---
1 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index b0feb47..96aae13 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -3376,15 +3376,23 @@ static void perf_event_task_output(struct perf_event *event,
struct perf_task_event *task_event)
struct perf_output_handle handle;
- int size;
struct task_struct *task = task_event->task;
- int ret;
+ unsigned long flags;
+ int size, ret;
+ /*
+ * If this CPU attempts to acquire an rq lock held by a CPU spinning
+ * in perf_output_lock() from interrupt context, it's game over.
+ */
+ local_irq_save(flags);

size = task_event->event_id.header.size;
ret = perf_output_begin(&handle, event, size, 0, 0);

- if (ret)
+ if (ret) {
+ local_irq_restore(flags);
+ }

task_event->event_id.pid = perf_event_pid(event, task);
task_event->event_id.ppid = perf_event_pid(event, current);
@@ -3395,6 +3403,7 @@ static void perf_event_task_output(struct perf_event *event,
perf_output_put(&handle, task_event->event_id);

+ local_irq_restore(flags);

static int perf_event_task_match(struct perf_event *event)
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/