Skip to contentSkip to conversation

perf_counter, x86: Fix APIC NMI programming

tip:perfcounters/core tip-bot for Ingo Molnar 1 participant 1 message May 26, 2009 x86

lore
Share & tools
lore
#0 lore raw
Commit-ID:  79202ba9ff8cf570a75596f42e011167734d1c4b
Gitweb:     http://git.kernel.org/tip/79202ba9ff8cf570a75596f42e011167734d1c4b
Author:     Ingo Molnar <[email protected]>
AuthorDate: Tue, 26 May 2009 08:10:00 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 26 May 2009 09:49:28 +0200

perf_counter, x86: Fix APIC NMI programming

My Nehalem box locks up in certain situations (with an always-asserted NMI causing a lockup) if the PMU LVT entry is programmed between NMI and IRQ mode with a high frequency.

Standardize exlusively on NMIs instead.

[ Impact: fix lockup ]

Cc: Peter Zijlstra <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Corey Ashford <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Kacur <[email protected]>
LKML-Reference: <new-submission>

---
 arch/x86/kernel/cpu/perf_counter.c |   16 +++-------------
 1 files changed, 3 insertions(+), 13 deletions(-)
arch/x86/kernel/cpu/perf_counter.c +3 -13
index 189bf9d..ece3813 100644
@@ -285,14 +285,10 @@ static int __hw_perf_counter_init(struct perf_counter *counter)
 		hwc->config |= ARCH_PERFMON_EVENTSEL_OS;
 
 	/*
-	 * If privileged enough, allow NMI events:
+	 * Use NMI events all the time:
 	 */
-	hwc->nmi = 0;
-	if (hw_event->nmi) {
-		if (sysctl_perf_counter_priv && !capable(CAP_SYS_ADMIN))
-			return -EACCES;
-		hwc->nmi = 1;
-	}
+	hwc->nmi	= 1;
+	hw_event->nmi	= 1;
 
 	if (!hwc->irq_period)
 		hwc->irq_period = x86_pmu.max_period;
@@ -553,9 +549,6 @@ fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc)
 	if (!x86_pmu.num_counters_fixed)
 		return -1;
 
-	if (unlikely(hwc->nmi))
-		return -1;
-
 	event = hwc->config & ARCH_PERFMON_EVENT_MASK;
 
 	if (unlikely(event == x86_pmu.event_map(PERF_COUNT_INSTRUCTIONS)))
@@ -806,9 +799,6 @@ static int amd_pmu_handle_irq(struct pt_regs *regs, int nmi)
 		counter = cpuc->counters[idx];
 		hwc = &counter->hw;
 
-		if (counter->hw_event.nmi != nmi)
-			continue;
-
 		val = x86_perf_counter_update(counter, hwc, idx);
 		if (val & (1ULL << (x86_pmu.counter_bits - 1)))
 			continue;