perf/x86: Fix leak in uncore_type_init failure paths
authorDave Jones <davej@redhat.com>
Thu, 6 Mar 2014 17:20:28 +0000 (12:20 -0500)
committerIngo Molnar <mingo@kernel.org>
Tue, 11 Mar 2014 10:59:34 +0000 (11:59 +0100)
The error path of uncore_type_init() frees up any allocations
that were made along the way, but it relies upon type->pmus
being set, which only happens if the function succeeds. As
type->pmus remains null in this case, the call to
uncore_type_exit will do nothing.

Moving the assignment earlier will allow us to actually free
those allocations should something go awry.

Signed-off-by: Dave Jones <davej@fedoraproject.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140306172028.GA552@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/kernel/cpu/perf_event_intel_uncore.c

index c88f7f4b03ee063ba090ec1e53c864fa4279fdec..047f540cf3f71cfe03f69d796a1957b8a58caebc 100644 (file)
@@ -3334,6 +3334,8 @@ static int __init uncore_type_init(struct intel_uncore_type *type)
        if (!pmus)
                return -ENOMEM;
 
+       type->pmus = pmus;
+
        type->unconstrainted = (struct event_constraint)
                __EVENT_CONSTRAINT(0, (1ULL << type->num_counters) - 1,
                                0, type->num_counters, 0, 0);
@@ -3369,7 +3371,6 @@ static int __init uncore_type_init(struct intel_uncore_type *type)
        }
 
        type->pmu_group = &uncore_pmu_attr_group;
-       type->pmus = pmus;
        return 0;
 fail:
        uncore_type_exit(type);