diff options
| author | Ingo Molnar <mingo@elte.hu> | 2011-04-27 11:51:41 +0200 |
|---|---|---|
| committer | Ingo Molnar <mingo@elte.hu> | 2011-04-26 19:34:34 +0200 |
| commit | ec75a71634dabe439db91c1ef51d5099f4493808 (patch) | |
| tree | bf62d417bd365f62d4a8afca02a93d37a106a339 /arch/x86/kernel/cpu/perf_event_intel.c | |
| parent | 18a073a3acd3a47fbb5e23333df7fad28d576345 (diff) | |
| download | linux-ec75a71634dabe439db91c1ef51d5099f4493808.tar.gz linux-ec75a71634dabe439db91c1ef51d5099f4493808.tar.xz linux-ec75a71634dabe439db91c1ef51d5099f4493808.zip | |
perf events, x86: Work around the Nehalem AAJ80 erratum
On Nehalem CPUs the retired branch-misses event can be completely bogus,
when there are no branch-misses occuring. When there are a lot of branch
misses then the count is pretty accurate. Still, this leaves us with an
event that over-counts a lot.
Detect this erratum and work it around by using BR_MISP_EXEC.ANY events.
These will also count speculated branches but still it's a lot more
precise in practice than the architectural event.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-yyfg0bxo9jsqxd6a0ovfny27@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'arch/x86/kernel/cpu/perf_event_intel.c')
| -rw-r--r-- | arch/x86/kernel/cpu/perf_event_intel.c | 16 |
1 files changed, 14 insertions, 2 deletions
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 9194b0698d63..9ae4a2aa7398 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -25,7 +25,7 @@ struct intel_percore { /* * Intel PerfMon, used on Core and later. */ -static const u64 intel_perfmon_event_map[] = +static u64 intel_perfmon_event_map[PERF_COUNT_HW_MAX] __read_mostly = { [PERF_COUNT_HW_CPU_CYCLES] = 0x003c, [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, @@ -1308,7 +1308,7 @@ static void intel_clovertown_quirks(void) * AJ106 could possibly be worked around by not allowing LBR * usage from PEBS, including the fixup. * AJ68 could possibly be worked around by always programming - * a pebs_event_reset[0] value and coping with the lost events. + * a pebs_event_reset[0] value and coping with the lost events. * * But taken together it might just make sense to not enable PEBS on * these chips. @@ -1412,6 +1412,18 @@ static __init int intel_pmu_init(void) x86_pmu.percore_constraints = intel_nehalem_percore_constraints; x86_pmu.enable_all = intel_pmu_nhm_enable_all; x86_pmu.extra_regs = intel_nehalem_extra_regs; + + if (ebx & 0x40) { + /* + * Erratum AAJ80 detected, we work it around by using + * the BR_MISP_EXEC.ANY event. This will over-count + * branch-misses, but it's still much better than the + * architectural event which is often completely bogus: + */ + intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89; + + pr_cont("erratum AAJ80 worked around, "); + } pr_cont("Nehalem events, "); break; |