diff options
| author | Peter Zijlstra <a.p.zijlstra@chello.nl> | 2009-05-14 14:52:17 +0200 |
|---|---|---|
| committer | Ingo Molnar <mingo@elte.hu> | 2009-05-15 09:47:03 +0200 |
| commit | a4016a79fcbd139e7378944c0d86a39fdbc70ecc (patch) | |
| tree | 5cb378852a6714d84c6e5db1936bfff14c5d9c0d /arch/x86 | |
| parent | perf_counter: Rework the perf counter disable/enable (diff) | |
| download | linux-a4016a79fcbd139e7378944c0d86a39fdbc70ecc.tar.xz linux-a4016a79fcbd139e7378944c0d86a39fdbc70ecc.zip | |
perf_counter: x86: Robustify interrupt handling
Two consecutive NMIs could daze and confuse the machine when the
first would handle the overflow of both counters.
[ Impact: fix false-positive syslog messages under multi-session profiling ]
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'arch/x86')
| -rw-r--r-- | arch/x86/kernel/cpu/perf_counter.c | 16 |
1 files changed, 13 insertions, 3 deletions
diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 313638cecbb5..1dcf67057f16 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -783,6 +783,10 @@ 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) + goto next; + val = x86_perf_counter_update(counter, hwc, idx); if (val & (1ULL << (x86_pmu.counter_bits - 1))) goto next; @@ -869,7 +873,6 @@ perf_counter_nmi_handler(struct notifier_block *self, { struct die_args *args = __args; struct pt_regs *regs; - int ret; if (!atomic_read(&active_counters)) return NOTIFY_DONE; @@ -886,9 +889,16 @@ perf_counter_nmi_handler(struct notifier_block *self, regs = args->regs; apic_write(APIC_LVTPC, APIC_DM_NMI); - ret = x86_pmu.handle_irq(regs, 1); + /* + * Can't rely on the handled return value to say it was our NMI, two + * counters could trigger 'simultaneously' raising two back-to-back NMIs. + * + * If the first NMI handles both, the latter will be empty and daze + * the CPU. + */ + x86_pmu.handle_irq(regs, 1); - return ret ? NOTIFY_STOP : NOTIFY_OK; + return NOTIFY_STOP; } static __read_mostly struct notifier_block perf_counter_nmi_notifier = { |