From d9f2a5ecb2846d0fd368fb4c45182e43f38e4471 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Sat, 20 Jun 2009 13:19:25 +0530 Subject: perf_counter, x8: Fix L1-data-Cache-Store-Referencees for AMD Fix AMD's Data Cache Refills from System event. After this patch : ./tools/perf/perf stat -e l1d -e l1d-misses -e l1d-write -e l1d-prefetch -e l1d-prefetch-miss -e l1i -e l1i-misses -e l1i-prefetch -e l2 -e l2-misses -e l2-write -e dtlb -e dtlb-misses -e itlb -e itlb-misses -e bpu -e bpu-misses ls /dev/ > /dev/null Performance counter stats for 'ls /dev/': 2499484 L1-data-Cache-Load-Referencees (scaled from 3.97%) 70347 L1-data-Cache-Load-Misses (scaled from 7.30%) 9360 L1-data-Cache-Store-Referencees (scaled from 8.64%) 32804 L1-data-Cache-Prefetch-Referencees (scaled from 17.72%) 7693 L1-data-Cache-Prefetch-Misses (scaled from 22.97%) 2180945 L1-instruction-Cache-Load-Referencees (scaled from 28.48%) 14518 L1-instruction-Cache-Load-Misses (scaled from 35.00%) 2405 L1-instruction-Cache-Prefetch-Referencees (scaled from 34.89%) 71387 L2-Cache-Load-Referencees (scaled from 34.94%) 18732 L2-Cache-Load-Misses (scaled from 34.92%) 79918 L2-Cache-Store-Referencees (scaled from 36.02%) 1295294 Data-TLB-Cache-Load-Referencees (scaled from 35.99%) 30896 Data-TLB-Cache-Load-Misses (scaled from 33.36%) 1222030 Instruction-TLB-Cache-Load-Referencees (scaled from 29.46%) 357 Instruction-TLB-Cache-Load-Misses (scaled from 20.46%) 530888 Branch-Cache-Load-Referencees (scaled from 11.48%) 8638 Branch-Cache-Load-Misses (scaled from 5.09%) 0.011295149 seconds time elapsed. Earlier it always shows value 0. Signed-off-by: Jaswinder Singh Rajput LKML-Reference: <1245484165.3102.6.camel@localhost.localdomain> Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 76dfef23f789..22eb3a1d4f9c 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -401,7 +401,7 @@ static const u64 amd_hw_cache_event_ids [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */ }, [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0x0042, /* Data Cache Refills from L2 */ + [ C(RESULT_ACCESS) ] = 0x0142, /* Data Cache Refills :system */ [ C(RESULT_MISS) ] = 0, }, [ C(OP_PREFETCH) ] = { -- cgit v1.2.3 From c14dab5c0782ef632742963a66276a195418a63c Mon Sep 17 00:00:00 2001 From: Yong Wang Date: Wed, 24 Jun 2009 10:13:24 +0800 Subject: perf_counter, x86: Set global control MSR correctly Previous code made an assumption that the power on value of global control MSR has enabled all fixed and general purpose counters properly. However, this is not the case for certain Intel processors, such as Atom - and it might also be firmware dependent. Each enable bit in IA32_PERF_GLOBAL_CTRL is AND'ed with the enable bits for all privilege levels in the respective IA32_PERFEVTSELx or IA32_PERF_FIXED_CTR_CTRL MSRs to start/stop the counting of respective counters. Counting is enabled if the AND'ed results is true; counting is disabled when the result is false. The end result is that all fixed counters are always disabled on Atom processors because the assumption is just invalid. Fix this by not initializing the ctrl-mask out of the global MSR, but setting it to perf_counter_mask. Reported-by: Stephane Eranian Signed-off-by: Yong Wang Cc: Arjan van de Ven Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras LKML-Reference: <20090624021324.GA2788@ywang-moblin2.bj.intel.com> Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 22eb3a1d4f9c..a310d19faca3 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -969,13 +969,6 @@ fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc) if (!x86_pmu.num_counters_fixed) return -1; - /* - * Quirk, IA32_FIXED_CTRs do not work on current Atom processors: - */ - if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL && - boot_cpu_data.x86_model == 28) - return -1; - event = hwc->config & ARCH_PERFMON_EVENT_MASK; if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS))) @@ -1428,8 +1421,6 @@ static int intel_pmu_init(void) */ x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3); - rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl); - /* * Install the hw-cache-events table: */ @@ -1514,6 +1505,7 @@ void __init init_hw_perf_counters(void) perf_counter_mask |= ((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED; + x86_pmu.intel_ctrl = perf_counter_mask; perf_counters_lapic_init(); register_die_notifier(&perf_counter_nmi_notifier); -- cgit v1.2.3 From 194002b274e9169a04beb1b23dcc132159bb566c Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 22 Jun 2009 16:35:24 +0200 Subject: perf_counter, x86: Add mmap counter read support Update the mmap control page with the needed information to use the userspace RDPMC instruction for self monitoring. Signed-off-by: Peter Zijlstra LKML-Reference: Signed-off-by: Ingo Molnar --- arch/powerpc/include/asm/perf_counter.h | 2 ++ arch/x86/include/asm/perf_counter.h | 3 +++ arch/x86/kernel/cpu/perf_counter.c | 6 ++++++ 3 files changed, 11 insertions(+) (limited to 'arch') diff --git a/arch/powerpc/include/asm/perf_counter.h b/arch/powerpc/include/asm/perf_counter.h index 8ccd4e155768..0ea0639fcf75 100644 --- a/arch/powerpc/include/asm/perf_counter.h +++ b/arch/powerpc/include/asm/perf_counter.h @@ -61,6 +61,8 @@ struct pt_regs; extern unsigned long perf_misc_flags(struct pt_regs *regs); extern unsigned long perf_instruction_pointer(struct pt_regs *regs); +#define PERF_COUNTER_INDEX_OFFSET 1 + /* * Only override the default definitions in include/linux/perf_counter.h * if we have hardware PMU support. diff --git a/arch/x86/include/asm/perf_counter.h b/arch/x86/include/asm/perf_counter.h index 5fb33e160ea0..fa64e401589d 100644 --- a/arch/x86/include/asm/perf_counter.h +++ b/arch/x86/include/asm/perf_counter.h @@ -87,6 +87,9 @@ union cpuid10_edx { #ifdef CONFIG_PERF_COUNTERS extern void init_hw_perf_counters(void); extern void perf_counters_lapic_init(void); + +#define PERF_COUNTER_INDEX_OFFSET 0 + #else static inline void init_hw_perf_counters(void) { } static inline void perf_counters_lapic_init(void) { } diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index a310d19faca3..b83474b6021a 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -912,6 +912,8 @@ x86_perf_counter_set_period(struct perf_counter *counter, err = checking_wrmsrl(hwc->counter_base + idx, (u64)(-left) & x86_pmu.counter_mask); + perf_counter_update_userpage(counter); + return ret; } @@ -1034,6 +1036,8 @@ try_generic: x86_perf_counter_set_period(counter, hwc, idx); x86_pmu.enable(hwc, idx); + perf_counter_update_userpage(counter); + return 0; } @@ -1126,6 +1130,8 @@ static void x86_pmu_disable(struct perf_counter *counter) x86_perf_counter_update(counter, hwc, idx); cpuc->counters[idx] = NULL; clear_bit(idx, cpuc->used_mask); + + perf_counter_update_userpage(counter); } /* -- cgit v1.2.3 From 4078c444cf667f018c3fc7ebf141131a2b7c9480 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Mon, 29 Jun 2009 00:41:11 -0700 Subject: perf_counter, x86: Update x86_pmu after WARN() The print out should read the value before changing the value. Signed-off-by: Yinghai Lu Cc: Peter Zijlstra LKML-Reference: <4A487017.4090007@kernel.org> Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index b83474b6021a..d4cf4ce19aac 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -1496,17 +1496,17 @@ void __init init_hw_perf_counters(void) pr_cont("%s PMU driver.\n", x86_pmu.name); if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { - x86_pmu.num_counters = X86_PMC_MAX_GENERIC; WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!", x86_pmu.num_counters, X86_PMC_MAX_GENERIC); + x86_pmu.num_counters = X86_PMC_MAX_GENERIC; } perf_counter_mask = (1 << x86_pmu.num_counters) - 1; perf_max_counters = x86_pmu.num_counters; if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) { - x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED; WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!", x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED); + x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED; } perf_counter_mask |= -- cgit v1.2.3