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.

   CPU0                            CPU1
   sched event with rq->lock held
                                   grab handle->data->lock
   spin on handle->data->lock
                                   interrupt
                                   try to grab rq->lock

Reported-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Tested-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1269598293.6174.8.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 12 insertions, 3 deletions

diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index b0feb4795af..96aae13c796 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);
                 return;
+        }
 
         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);
 
         perf_output_end(&handle);
+        local_irq_restore(flags);
 }
 
 static int perf_event_task_match(struct perf_event *event)