perf, x86: Less disastrous PEBS/BTS buffer allocation failure

Currently PEBS/BTS buffers are allocated when we instantiate the first
event, when this fails everything fails.

This is a problem because esp. BTS tries to allocate a rather large
buffer (64K), which can easily fail.

This patch changes the logic such that when either buffer allocation
fails, we simply don't allow events that would use these facilities,
but continue functioning for all other events.

This logic comes from a much larger patch proposed by Stephane.

Suggested-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Stephane Eranian <eranian@google.com>
LKML-Reference: <20101019134808.354429461@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 3 insertions, 2 deletions

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index f369c53315a..61e78f65106 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -238,6 +238,7 @@ struct x86_pmu {
          * Intel DebugStore bits
          */
         int             bts, pebs;
+        int             bts_active, pebs_active;
         int             pebs_record_size;
         void            (*drain_pebs)(struct pt_regs *regs);
         struct event_constraint *pebs_constraints;
@@ -478,7 +479,7 @@ static int x86_setup_perfctr(struct perf_event *event)
         if ((attr->config == PERF_COUNT_HW_BRANCH_INSTRUCTIONS) &&
             (hwc->sample_period == 1)) {
                 /* BTS is not supported by this architecture. */
-                if (!x86_pmu.bts)
+                if (!x86_pmu.bts_active)
                         return -EOPNOTSUPP;
 
                 /* BTS is currently only allowed for user-mode. */
@@ -497,7 +498,7 @@ static int x86_pmu_hw_config(struct perf_event *event)
                 int precise = 0;
 
                 /* Support for constant skid */
-                if (x86_pmu.pebs) {
+                if (x86_pmu.pebs_active) {
                         precise++;
 
                         /* Support for IP fixup */