perf, x86: P4 PMU -- handle unflagged events

It might happen that an event can overflow without
the proper overflow flag set. Check the sign bit in
the raw counter value to solve this problem.

Tested-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: fweisbec@gmail.com
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <1274083984.6540.15.camel@minggr.sh.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c
index 424fc8d..02f0728 100644 (file)
--- a/arch/x86/kernel/cpu/perf_event_p4.c
+++ b/arch/x86/kernel/cpu/perf_event_p4.c
@@ -465,15 +465,21 @@ out:
        return rc;
 }
 
-static inline void p4_pmu_clear_cccr_ovf(struct hw_perf_event *hwc)
+static inline int p4_pmu_clear_cccr_ovf(struct hw_perf_event *hwc)
 {
-       unsigned long dummy;
+       int overflow = 0;
+       u32 low, high;
 
-       rdmsrl(hwc->config_base + hwc->idx, dummy);
-       if (dummy & P4_CCCR_OVF) {
+       rdmsr(hwc->config_base + hwc->idx, low, high);
+
+       /* we need to check high bit for unflagged overflows */
+       if ((low & P4_CCCR_OVF) || (high & (1 << 31))) {
+               overflow = 1;
                (void)checking_wrmsrl(hwc->config_base + hwc->idx,
-                       ((u64)dummy) & ~P4_CCCR_OVF);
+                       ((u64)low) & ~P4_CCCR_OVF);
        }
+
+       return overflow;
 }
 
 static inline void p4_pmu_disable_event(struct perf_event *event)
@@ -584,21 +590,15 @@ static int p4_pmu_handle_irq(struct pt_regs *regs)
 
                WARN_ON_ONCE(hwc->idx != idx);
 
-               /*
-                * FIXME: Redundant call, actually not needed
-                * but just to check if we're screwed
-                */
-               p4_pmu_clear_cccr_ovf(hwc);
+               /* it might be unflagged overflow */
+               handled = p4_pmu_clear_cccr_ovf(hwc);
 
                val = x86_perf_event_update(event);
-               if (val & (1ULL << (x86_pmu.cntval_bits - 1)))
+               if (!handled && (val & (1ULL << (x86_pmu.cntval_bits - 1))))
                        continue;
 
-               /*
-                * event overflow
-                */
-               handled         = 1;
-               data.period     = event->hw.last_period;
+               /* event overflow for sure */
+               data.period = event->hw.last_period;
 
                if (!x86_perf_event_set_period(event))
                        continue;