perf/core: Fix perf_sched_count derailment
authorAlexander Shishkin <alexander.shishkin@linux.intel.com>
Wed, 2 Mar 2016 11:24:14 +0000 (13:24 +0200)
committerIngo Molnar <mingo@kernel.org>
Tue, 8 Mar 2016 11:18:31 +0000 (12:18 +0100)
The error path in perf_event_open() is such that asking for a sampling
event on a PMU that doesn't generate interrupts will end up in dropping
the perf_sched_count even though it hasn't been incremented for this
event yet.

Given a sufficient amount of these calls, we'll end up disabling
scheduler's jump label even though we'd still have active events in the
system, thereby facilitating the arrival of the infernal regions upon us.

I'm fixing this by moving account_event() inside perf_event_alloc().

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
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/1456917854-29427-1-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/events/core.c

index 5dcc0bd08d11ffc5914e99c5ae8adeb447cfa581..b7231498de47f79787e0e46b95cf99fb93737b0a 100644 (file)
@@ -8000,6 +8000,9 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
                }
        }
 
+       /* symmetric to unaccount_event() in _free_event() */
+       account_event(event);
+
        return event;
 
 err_per_task:
@@ -8363,8 +8366,6 @@ SYSCALL_DEFINE5(perf_event_open,
                }
        }
 
-       account_event(event);
-
        /*
         * Special case software events and allow them to be part of
         * any hardware group.
@@ -8661,8 +8662,6 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
        /* Mark owner so we could distinguish it from user events. */
        event->owner = TASK_TOMBSTONE;
 
-       account_event(event);
-
        ctx = find_get_context(event->pmu, task, event);
        if (IS_ERR(ctx)) {
                err = PTR_ERR(ctx);