perf/core: Disable the event on a truncated AUX record
authorAlexander Shishkin <alexander.shishkin@linux.intel.com>
Tue, 10 May 2016 13:18:33 +0000 (16:18 +0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 19 May 2016 00:06:48 +0000 (17:06 -0700)
commit 9f448cd3cbcec8995935e60b27802ae56aac8cc0 upstream.

When the PMU driver reports a truncated AUX record, it effectively means
that there is no more usable room in the event's AUX buffer (even though
there may still be some room, so that perf_aux_output_begin() doesn't take
action). At this point the consumer still has to be woken up and the event
has to be disabled, otherwise the event will just keep spinning between
perf_aux_output_begin() and perf_aux_output_end() until its context gets
unscheduled.

Again, for cpu-wide events this means never, so once in this condition,
they will be forever losing data.

Fix this by disabling the event and waking up the consumer in case of a
truncated AUX record.

Reported-by: Markus Metzger <markus.t.metzger@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1462886313-13660-3-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernel/events/ring_buffer.c

index adfdc0536117c1f10bf8f0dd8014798011e1f855..014b6952819463e165a8f273839af53424c258a3 100644 (file)
@@ -347,6 +347,7 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size,
                         bool truncated)
 {
        struct ring_buffer *rb = handle->rb;
+       bool wakeup = truncated;
        unsigned long aux_head;
        u64 flags = 0;
 
@@ -375,9 +376,16 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size,
        aux_head = rb->user_page->aux_head = local_read(&rb->aux_head);
 
        if (aux_head - local_read(&rb->aux_wakeup) >= rb->aux_watermark) {
-               perf_output_wakeup(handle);
+               wakeup = true;
                local_add(rb->aux_watermark, &rb->aux_wakeup);
        }
+
+       if (wakeup) {
+               if (truncated)
+                       handle->event->pending_disable = 1;
+               perf_output_wakeup(handle);
+       }
+
        handle->event = NULL;
 
        local_set(&rb->aux_nest, 0);