From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756691Ab0KRKsC (ORCPT ); Thu, 18 Nov 2010 05:48:02 -0500 Received: from one.firstfloor.org ([213.235.205.2]:42124 "EHLO one.firstfloor.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755832Ab0KRKrs (ORCPT ); Thu, 18 Nov 2010 05:47:48 -0500 From: Andi Kleen To: a.p.zijlstra@chello.nl Cc: eranian@google.com, linux-kernel@vger.kernel.org, x86@kernel.org, Andi Kleen Subject: [PATCH 4/4] perf-events: Fix LLC-* events on Intel Nehalem/Westmere v2 Date: Thu, 18 Nov 2010 11:47:34 +0100 Message-Id: <1290077254-12165-5-git-send-email-andi@firstfloor.org> X-Mailer: git-send-email 1.7.1 In-Reply-To: <1290077254-12165-1-git-send-email-andi@firstfloor.org> References: <1290077254-12165-1-git-send-email-andi@firstfloor.org> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Andi Kleen The generic perf LLC-* events do count the L2 caches, not the real L3 LLC on Intel Nehalem and Westmere. This lead to quite some confusion. Fixing this properly requires use of the special OFFCORE_RESPONSE events which need a separate mask register. This has been implemented in a earlier patch. Now use this infrastructure to set correct events for the LLC-* on Nehalem and Westmere Signed-off-by: Andi Kleen --- arch/x86/kernel/cpu/perf_event.c | 12 ++++---- arch/x86/kernel/cpu/perf_event_intel.c | 46 +++++++++++++++++++++++--------- 2 files changed, 39 insertions(+), 19 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 346006a..b1abe89 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -467,8 +467,9 @@ static inline int x86_pmu_initialized(void) } static inline int -set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr) +set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event *event) { + struct perf_event_attr *attr = &event->attr; unsigned int cache_type, cache_op, cache_result; u64 config, val; @@ -494,9 +495,8 @@ set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr) if (val == -1) return -EINVAL; - hwc->config |= val; - - return 0; + hwc->config |= val & X86_RAW_EVENT_MASK; + return x86_pmu_extra_regs(val, event); } static int x86_setup_perfctr(struct perf_event *event) @@ -521,10 +521,10 @@ static int x86_setup_perfctr(struct perf_event *event) } if (attr->type == PERF_TYPE_RAW) - return 0; + return x86_pmu_extra_regs(event->attr.config, event); if (attr->type == PERF_TYPE_HW_CACHE) - return set_ext_hw_attr(hwc, attr); + return set_ext_hw_attr(hwc, event); if (attr->config >= x86_pmu.max_events) return -EINVAL; diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 675a666..7017ef7 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -86,7 +86,7 @@ static struct event_constraint intel_nehalem_event_constraints[] = static struct extra_reg intel_nehalem_extra_regs[] = { - INTEL_EVENT_EXTRA_REG(0xb7, 0x1a6, 0xffff, 32), /* OFFCORE_RESPONSE_0 */ + INTEL_EVENT_EXTRA_REG(0xb7, 0x1a6, 0xffff, 32), /* OFFCORE_RESPONSE */ EVENT_EXTRA_END }; @@ -170,16 +170,26 @@ static __initconst const u64 westmere_hw_cache_event_ids }, [ C(LL ) ] = { [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS */ - [ C(RESULT_MISS) ] = 0x0224, /* L2_RQSTS.LD_MISS */ + /* OFFCORE_RESPONSE_0.ANY_DATA.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x00000711004301b7, + /* OFFCORE_RESPONSE_1.ANY_DATA.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x0000f811004301bb, }, + /* + * Use RFO, not WRITEBACK, because a write miss would typically occur + * on RFO. + */ [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS */ - [ C(RESULT_MISS) ] = 0x0824, /* L2_RQSTS.RFO_MISS */ + /* OFFCORE_RESPONSE_1.ANY_RFO.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x00000722004301bb, + /* OFFCORE_RESPONSE_0.ANY_RFO.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x0000f822004301b7, }, [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference */ - [ C(RESULT_MISS) ] = 0x412e, /* LLC Misses */ + /* OFFCORE_RESPONSE_0.PREFETCH.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x00000770004301b7, + /* OFFCORE_RESPONSE_1.PREFETCH.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x0000f870004301bb, }, }, [ C(DTLB) ] = { @@ -261,16 +271,26 @@ static __initconst const u64 nehalem_hw_cache_event_ids }, [ C(LL ) ] = { [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS */ - [ C(RESULT_MISS) ] = 0x0224, /* L2_RQSTS.LD_MISS */ + /* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x00000711004301b7, + /* OFFCORE_RESPONSE.ANY_DATA.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x0000f811004301b7, }, + /* + * Use RFO, not WRITEBACK, because a write miss would typically occur + * on RFO. + */ [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS */ - [ C(RESULT_MISS) ] = 0x0824, /* L2_RQSTS.RFO_MISS */ + /* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x00000722004301b7, + /* OFFCORE_RESPONSE.ANY_RFO.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x0000f822004301b7, }, [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference */ - [ C(RESULT_MISS) ] = 0x412e, /* LLC Misses */ + /* OFFCORE_RESPONSE.PREFETCH.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x00000770004301b7, + /* OFFCORE_RESPONSE.PREFETCH.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x0000f870004301b7, }, }, [ C(DTLB) ] = { -- 1.7.1