Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755376AbYKZIti (ORCPT ); Wed, 26 Nov 2008 03:49:38 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753961AbYKZInB (ORCPT ); Wed, 26 Nov 2008 03:43:01 -0500 Received: from ey-out-2122.google.com ([74.125.78.26]:45489 "EHLO ey-out-2122.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754129AbYKZIm5 (ORCPT ); Wed, 26 Nov 2008 03:42:57 -0500 DomainKey-Signature: a=rsa-sha1; c=nofws; d=googlemail.com; s=gamma; h=to:cc:subject:from:date:message-id; b=wmYc5yHMUhpjwcV6dUf1/ISmNWvDwYURnCZmgGgWRfk+jQXVqXgDAkZh+quvjd0Dsy 5yblww2yuoiakXW7M9vczZjew9JfZEK/mQiiVhjTGvStjybwIv5e5SA0ssx17xHmI9aI x76gIjaBv72TPXdcy+6JBI2nLCHEEN4nBCRmY= To: linux-kernel@vger.kernel.org Cc: akpm@linux-foundation.org, mingo@elte.hu, x86@kernel.org, andi@firstfloor.org, eranian@gmail.com, sfr@canb.auug.org.au Subject: [patch 21/24] perfmon: Intel architectural PMU support (x86) From: eranian@googlemail.com Date: Wed, 26 Nov 2008 00:42:54 -0800 (PST) Message-ID: <492d0c0e.06e2660a.1583.ffffcc0a@mx.google.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 19708 Lines: 671 This patch adds Intel architectural PMU support (version 1, 2, and 3). Signed-off-by: Stephane Eranian -- Index: o3/arch/x86/perfmon/Makefile =================================================================== --- o3.orig/arch/x86/perfmon/Makefile 2008-11-25 18:09:47.000000000 +0100 +++ o3/arch/x86/perfmon/Makefile 2008-11-25 18:21:33.000000000 +0100 @@ -3,3 +3,4 @@ # Contributed by Stephane Eranian # obj-$(CONFIG_PERFMON) += perfmon.o +obj-$(CONFIG_X86_PERFMON_INTEL_ARCH) += perfmon_intel_arch.o Index: o3/arch/x86/perfmon/perfmon_intel_arch.c =================================================================== --- /dev/null 1970-01-01 00:00:00.000000000 +0000 +++ o3/arch/x86/perfmon/perfmon_intel_arch.c 2008-11-25 18:22:00.000000000 +0100 @@ -0,0 +1,628 @@ +/* + * This file contains the Intel architectural perfmon v1, v2, v3 + * description tables. + * + * Architectural perfmon was introduced with Intel Core Solo/Duo + * processors. + * + * Copyright (c) 2006-2007 Hewlett-Packard Development Company, L.P. + * Contributed by Stephane Eranian + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of version 2 of the GNU General Public + * License as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA + * 02111-1307 USA + */ +#include +#include +#include +#include + +static u64 enable_mask[PFM_MAX_PMCS]; +static u16 max_enable; +static int pfm_intel_arch_version; + +DEFINE_PER_CPU(u64, saved_global_ctrl); + +/* + * - upper 32 bits are reserved + * - INT: APIC enable bit is reserved (forced to 1) + * - bit 21 is reserved + * + * RSVD: reserved bits are 1 + */ +#define PFM_IA_PMC_RSVD ((~((1ULL<<32)-1)) \ + | (1ULL<<20) \ + | (1ULL<<21)) + +/* + * force Local APIC interrupt on overflow + * disable with NO_EMUL64 + */ +#define PFM_IA_PMC_VAL (1ULL<<20) +#define PFM_IA_NO64 (1ULL<<20) + +/* + * architectuture specifies that: + * IA32_PMCx MSR : starts at 0x0c1 & occupy a contiguous block of MSR + * IA32_PERFEVTSELx MSR : starts at 0x186 & occupy a contiguous block of MSR + * MSR_GEN_FIXED_CTR0 : starts at 0x309 & occupy a contiguous block of MSR + */ +#define MSR_GEN_SEL_BASE MSR_P6_EVNTSEL0 +#define MSR_GEN_PMC_BASE MSR_P6_PERFCTR0 +#define MSR_GEN_FIXED_PMC_BASE MSR_CORE_PERF_FIXED_CTR0 + +/* + * layout of EAX for CPUID.0xa leaf function + */ +struct pmu_eax { + unsigned int version:8; /* architectural perfmon version */ + unsigned int num_cnt:8; /* number of generic counters */ + unsigned int cnt_width:8; /* width of generic counters */ + unsigned int ebx_length:8; /* number of architected events */ +}; + +/* + * layout of EDX for CPUID.0xa leaf function when perfmon v2 is detected + */ +struct pmu_edx { + unsigned int num_cnt:5; /* number of fixed counters */ + unsigned int cnt_width:8; /* width of fixed counters */ + unsigned int reserved:19; +}; + +static void pfm_intel_arch_acquire_pmu_percpu(void); +static void pfm_intel_arch_release_pmu_percpu(void); +static int pfm_intel_arch_stop_save(struct pfm_context *ctx, + struct pfm_event_set *set); +static int pfm_intel_arch_has_ovfls(struct pfm_context *ctx); +static void __kprobes pfm_intel_arch_quiesce(void); + +/* + * physical addresses of MSR controlling the perfevtsel and counter registers + */ +struct pfm_arch_pmu_info pfm_intel_arch_pmu_info = { + .stop_save = pfm_intel_arch_stop_save, + .has_ovfls = pfm_intel_arch_has_ovfls, + .quiesce = pfm_intel_arch_quiesce, + .acquire_pmu_percpu = pfm_intel_arch_acquire_pmu_percpu, + .release_pmu_percpu = pfm_intel_arch_release_pmu_percpu +}; + +#define PFM_IA_C(n) { \ + .type = PFM_REG_I64, \ + .desc = "PERFEVTSEL"#n, \ + .dfl_val = PFM_IA_PMC_VAL, \ + .rsvd_msk = PFM_IA_PMC_RSVD, \ + .no_emul64_msk = PFM_IA_NO64, \ + .hw_addr = MSR_GEN_SEL_BASE+(n) \ + } + +#define PFM_IA_D(n) \ + { .type = PFM_REG_C, \ + .desc = "PMC"#n, \ + .hw_addr = MSR_P6_PERFCTR0+n, \ + .dep_pmcs[0] = 1ULL << n \ + } + +#define PFM_IA_FD(n) \ + { .type = PFM_REG_C, \ + .desc = "FIXED_CTR"#n, \ + .hw_addr = MSR_CORE_PERF_FIXED_CTR0+n,\ + .dep_pmcs[0] = 1ULL << 16 \ + } + + +static struct pfm_regmap_desc pfm_intel_arch_pmc_desc[] = { +/* pmc0 */ PFM_IA_C(0), PFM_IA_C(1), PFM_IA_C(2), PFM_IA_C(3), +/* pmc4 */ PFM_IA_C(4), PFM_IA_C(5), PFM_IA_C(6), PFM_IA_C(7), +/* pmc8 */ PFM_IA_C(8), PFM_IA_C(9), PFM_IA_C(10), PFM_IA_C(11), +/* pmc12 */ PFM_IA_C(12), PFM_IA_C(13), PFM_IA_C(14), PFM_IA_C(15), + +/* pmc16 */ { .type = PFM_REG_I, + .desc = "FIXED_CTRL", + .dfl_val = 0x8888888888888888ULL, /* force PMI */ + .rsvd_msk = 0, /* set dynamically */ + .no_emul64_msk = 0, + .hw_addr = MSR_CORE_PERF_FIXED_CTR_CTRL + }, +}; +#define PFM_IA_MAX_PMCS ARRAY_SIZE(pfm_intel_arch_pmc_desc) + +static struct pfm_regmap_desc pfm_intel_arch_pmd_desc[] = { +/* pmd0 */ PFM_IA_D(0), PFM_IA_D(1), PFM_IA_D(2), PFM_IA_D(3), +/* pmd4 */ PFM_IA_D(4), PFM_IA_D(5), PFM_IA_D(6), PFM_IA_D(7), +/* pmd8 */ PFM_IA_D(8), PFM_IA_D(9), PFM_IA_D(10), PFM_IA_D(11), +/* pmd12 */ PFM_IA_D(12), PFM_IA_D(13), PFM_IA_D(14), PFM_IA_D(15), + +/* pmd16 */ PFM_IA_FD(0), PFM_IA_FD(1), PFM_IA_FD(2), PFM_IA_FD(3), +/* pmd20 */ PFM_IA_FD(4), PFM_IA_FD(5), PFM_IA_FD(6), PFM_IA_FD(7), +/* pmd24 */ PFM_IA_FD(8), PFM_IA_FD(9), PFM_IA_FD(10), PFM_IA_FD(11), +/* pmd28 */ PFM_IA_FD(16), PFM_IA_FD(17), PFM_IA_FD(18), PFM_IA_FD(19) +}; +#define PFM_IA_MAX_PMDS ARRAY_SIZE(pfm_intel_arch_pmd_desc) + +#define PFM_IA_MAX_CNT 16 /* # generic counters in mapping table */ +#define PFM_IA_MAX_FCNT 16 /* # of fixed counters in mapping table */ +#define PFM_IA_FCNT_BASE 16 /* base index of fixed counters PMD */ + +static struct pfm_pmu_config pfm_intel_arch_pmu_conf; + +static void pfm_intel_arch_check_errata(void) +{ + /* + * Core Duo errata AE49 (no fix). Both counters share a single + * enable bit in PERFEVTSEL0 + */ + if (current_cpu_data.x86 == 6 && current_cpu_data.x86_model == 14) + pfm_intel_arch_pmu_info.flags |= PFM_X86_FL_NO_SHARING; +} + +static inline void set_enable_mask(unsigned int i) +{ + pfm_arch_bv_set_bit(i, enable_mask); + + /* max_enable = highest + 1 */ + if ((i+1) > max_enable) + max_enable = i+ 1; +} + +static void pfm_intel_arch_setup_generic(unsigned int version, + unsigned int width, + unsigned int count) +{ + u64 rsvd; + unsigned int i; + + /* + * first we handle the generic counters: + * + * - ensure HW does not have more registers than hardcoded in the tables + * - adjust rsvd_msk to actual counter width + * - initialize enable_mask (list of PMC with start/stop capability) + * - mark unused hardcoded generic counters as unimplemented + */ + + /* + * min of number of Hw counters and hardcoded in the tables + */ + if (count >= PFM_IA_MAX_CNT) { + printk(KERN_INFO "perfmon: Limiting number of generic counters" + " to %u, HW supports %u", + PFM_IA_MAX_CNT, count); + count = PFM_IA_MAX_CNT; + } + + /* + * adjust rsvd_msk for generic counters based on actual width + * initialize enable_mask (1 per pmd) + */ + rsvd = ~((1ULL << width)-1); + for (i = 0; i < count; i++) { + pfm_intel_arch_pmd_desc[i].rsvd_msk = rsvd; + set_enable_mask(i); + } + + /* + * handle version 3 new anythread bit (21) + */ + if (version == 3) { + for (i = 0; i < count; i++) + pfm_intel_arch_pmc_desc[i].rsvd_msk &= ~(1ULL << 21); + } + + + /* + * mark unused generic counters as not available + */ + for (i = count ; i < PFM_IA_MAX_CNT; i++) { + pfm_intel_arch_pmd_desc[i].type = PFM_REG_NA; + pfm_intel_arch_pmc_desc[i].type = PFM_REG_NA; + } +} + +static void pfm_intel_arch_setup_fixed(unsigned int version, + unsigned int width, + unsigned int count) +{ + u64 rsvd, dfl; + unsigned int i; + + /* + * handle the fixed counters (if any): + * + * - ensure HW does not have more registers than hardcoded in the tables + * - adjust rsvd_msk to actual counter width + * - initialize enable_mask (list of PMC with start/stop capability) + * - mark unused hardcoded generic counters as unimplemented + */ + if (count >= PFM_IA_MAX_FCNT) { + printk(KERN_INFO "perfmon: Limiting number of fixed counters" + " to %u, HW supports %u", + PFM_IA_MAX_FCNT, count); + count = PFM_IA_MAX_FCNT; + } + /* + * adjust rsvd_msk for fixed counters based on actual width + */ + rsvd = ~((1ULL << width)-1); + for (i = 0; i < count; i++) + pfm_intel_arch_pmd_desc[PFM_IA_FCNT_BASE+i].rsvd_msk = rsvd; + + /* + * handle version new anythread bit (bit 2) + */ + if (version == 3) + rsvd = 1ULL << 3; + else + rsvd = 3ULL << 2; + + pfm_intel_arch_pmc_desc[16].rsvd_msk = 0; + for (i = 0; i < count; i++) + pfm_intel_arch_pmc_desc[16].rsvd_msk |= rsvd << (i<<2); + + /* + * mark unused fixed counters as unimplemented + * + * update the rsvd_msk, dfl_val in FIXED_CTRL: + * - rsvd_msk: set all 4 bits + * - dfl_val : clear all 4 bits + */ + dfl = pfm_intel_arch_pmc_desc[16].dfl_val; + rsvd = pfm_intel_arch_pmc_desc[16].rsvd_msk; + + for (i = count ; i < PFM_IA_MAX_FCNT; i++) { + pfm_intel_arch_pmd_desc[PFM_IA_FCNT_BASE+i].type = PFM_REG_NA; + rsvd |= 0xfULL << (i<<2); + dfl &= ~(0xfULL << (i<<2)); + } + + /* + * FIXED_CTR_CTRL unavailable when no fixed counters are defined + */ + if (!count) { + pfm_intel_arch_pmc_desc[16].type = PFM_REG_NA; + } else { + /* update rsvd_mask and dfl_val */ + pfm_intel_arch_pmc_desc[16].rsvd_msk = rsvd; + pfm_intel_arch_pmc_desc[16].dfl_val = dfl; + set_enable_mask(16); + } +} + +static int pfm_intel_arch_probe_pmu(void) +{ + union { + unsigned int val; + struct pmu_eax eax; + struct pmu_edx edx; + } eax, edx; + unsigned int ebx, ecx; + unsigned int width = 0; + + edx.val = 0; + + if (!cpu_has_arch_perfmon) { + PFM_INFO("no support for Intel architectural PMU"); + return -1; + } + + if (!cpu_has_apic) { + PFM_INFO("no Local APIC, try rebooting with lapic option"); + return -1; + } + + /* cpuid() call protected by cpu_has_arch_perfmon */ + cpuid(0xa, &eax.val, &ebx, &ecx, &edx.val); + + /* + * some 6/15 models have buggy BIOS + */ + if (eax.eax.version == 0 + && current_cpu_data.x86 == 6 && current_cpu_data.x86_model == 15) { + PFM_INFO("buggy v2 BIOS, adjusting for 2 generic counters"); + eax.eax.version = 2; + eax.eax.num_cnt = 2; + eax.eax.cnt_width = 40; + } + + /* + * some v2 BIOSes are incomplete + */ + if (eax.eax.version == 2 && !edx.edx.num_cnt) { + PFM_INFO("buggy v2 BIOS, adjusting for 3 fixed counters"); + edx.edx.num_cnt = 3; + edx.edx.cnt_width = 40; + } + + /* + * no fixed counters on earlier versions + */ + if (eax.eax.version < 2) { + edx.val = 0; + } else { + /* + * use the min value of both widths until we support + * variable width counters + */ + width = eax.eax.cnt_width < edx.edx.cnt_width ? + eax.eax.cnt_width : edx.edx.cnt_width; + } + + /* + * Intel Atom processors have a buggy firmware which does not report + * the correct number of fixed counters + */ + if (eax.eax.version == 3 && edx.edx.num_cnt < 3 + && current_cpu_data.x86 == 6 && current_cpu_data.x86_model == 28) { + PFM_INFO("buggy v3 BIOS, adjusting for 3 fixed counters"); + edx.edx.num_cnt = 3; + } + + PFM_INFO("detected architecural perfmon v%d", eax.eax.version); + PFM_INFO("num_gen=%d width=%d num_fixed=%d width=%d", + eax.eax.num_cnt, + eax.eax.cnt_width, + edx.edx.num_cnt, + edx.edx.cnt_width); + + pfm_intel_arch_setup_generic(eax.eax.version, + width, + eax.eax.num_cnt); + + pfm_intel_arch_setup_fixed(eax.eax.version, + width, + edx.edx.num_cnt); + + pfm_intel_arch_check_errata(); + + pfm_intel_arch_version = eax.eax.version; + + return 0; +} + +/** + * pfm_intel_arch_has_ovfls - check for pending overflow condition + * @ctx: context to work on + * + * detect if counters have overflowed. + * return: + * 0 : no overflow + * 1 : at least one overflow + */ +static int __kprobes pfm_intel_arch_has_ovfls(struct pfm_context *ctx) +{ + u64 *cnt_mask; + u64 wmask, val; + u16 i, num; + + cnt_mask = ctx->regs.cnt_pmds; + num = ctx->regs.num_counters; + wmask = 1ULL << pfm_pmu_conf->counter_width; + + /* + * we can leverage the fact that we know the mapping + * to hardcode the MSR address and avoid accessing + * more cachelines + * + * We need to check cnt_mask because not all registers + * may be available. + */ + for (i = 0; num; i++) { + if (pfm_arch_bv_test_bit(i, cnt_mask)) { + rdmsrl(pfm_intel_arch_pmd_desc[i].hw_addr, val); + if (!(val & wmask)) + return 1; + num--; + } + } + return 0; +} + +static int pfm_intel_arch_stop_save(struct pfm_context *ctx, + struct pfm_event_set *set) +{ + u64 used_mask[PFM_PMC_BV]; + u64 val, wmask, ovfl_mask; + u32 i, count; + + wmask = 1ULL << pfm_pmu_conf->counter_width; + + pfm_arch_bv_and(used_mask, + set->used_pmcs, + enable_mask, + max_enable); + + count = pfm_arch_bv_weight(used_mask, max_enable); + + /* + * stop monitoring + * Unfortunately, this is very expensive! + * wrmsrl() is serializing. + */ + for (i = 0; count; i++) { + if (pfm_arch_bv_test_bit(i, used_mask)) { + wrmsrl(pfm_pmu_conf->pmc_desc[i].hw_addr, 0); + count--; + } + } + + /* + * if we already having a pending overflow condition, we simply + * return to take care of this first. + */ + if (set->npend_ovfls) + return 1; + + ovfl_mask = pfm_pmu_conf->ovfl_mask; + + /* + * check for pending overflows and save PMDs (combo) + * we employ used_pmds because we also need to save + * and not just check for pending interrupts. + * + * all pmds are counters + */ + count = set->nused_pmds; + for (i = 0; count; i++) { + if (pfm_arch_bv_test_bit(i, set->used_pmds)) { + val = pfm_arch_read_pmd(ctx, i); + if (!(val & wmask)) { + pfm_arch_bv_set_bit(i, set->povfl_pmds); + set->npend_ovfls++; + } + val = (set->pmds[i] & ~ovfl_mask) + | (val & ovfl_mask); + set->pmds[i] = val; + count--; + } + } + /* 0 means: no need to save PMDs at upper level */ + return 0; +} + +/** + * pfm_intel_arch_quiesce - stop monitoring without grabbing any lock + * + * called from NMI interrupt handler to immediately stop monitoring + * cannot grab any lock, including perfmon related locks + */ +static void __kprobes pfm_intel_arch_quiesce(void) +{ + u16 i; + + /* + * PMC16 is the fixed control register so it has a + * distinct MSR address + * + * We do not use the hw_addr field in the table to avoid touching + * too many cachelines + */ + for (i = 0; i < pfm_pmu_conf->regs_all.max_pmc; i++) { + if (pfm_arch_bv_test_bit(i, pfm_pmu_conf->regs_all.pmcs)) { + if (i == 16) + wrmsrl(MSR_CORE_PERF_FIXED_CTR_CTRL, 0); + else + wrmsrl(MSR_P6_EVNTSEL0+i, 0); + } + } +} +/** +* pfm_intel_arch_release_pmu_percpu - release PMU resource per CPU +* +* Since v2, there exists global control MSR, to start/stop and +* also collect overflow status information. In particular, +* GLOBAL_CTRL controls start/stop and has one bit per counter. +* To maintain backward compatibility with v1, the power-on value +* of GLOBAL_CTRL should be such that generic counters are enabled +* but fixed counters are disabled (true on Penryn and Atom currently). +* +* Here, we simply make sure that all available counters are enabled. +* After that, start/stop is controlled on a per-counter basis. +*/ +static void pfm_intel_arch_acquire_pmu_percpu(void) +{ + struct pfm_regmap_desc *d; + u64 mask = 0; + unsigned int i; + + /* nothing to do for v1 */ + if (pfm_intel_arch_version < 2) + return; + + /* + * build bitmask of registers that are available to + * us. In some cases, there may be fewer registers than + * what the PMU supports due to sharing with other kernel + * subsystems, such as NMI + */ + d = pfm_pmu_conf->pmd_desc; + for (i=0; i < 16; i++) { + if ((d[i].type & PFM_REG_I) == 0) + continue; + mask |= 1ull << i; + } + for (i=16; i < PFM_IA_MAX_PMDS; i++) { + if ((d[i].type & PFM_REG_I) == 0) + continue; + mask |= 1ull << (32+i-16); + } + /* + * keep a local copy of the current MSR_CORE_PERF_GLOBAL_CTRL + */ + rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, __get_cpu_var(saved_global_ctrl)); + + PFM_DBG("global=0x%llx set to 0x%llx", + __get_cpu_var(saved_global_ctrl), + mask); + /* + * enable all registers + * + * No need to quiesce PMU. If there is a overflow, it will be + * treated as spurious by the handler + */ + wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, mask); +} + +/** +* pfm_intel_arch_release_pmu_percpu - release PMU resource per CPU +* +* Since v2, there exists global control MSR, to start/stop and +* also collect overflow status information. In particular, +* GLOBAL_CTRL controls start/stop and has one bit per counter. +* To maintain backward compatibility with v1, the power-on value +* of GLOBAL_CTRL should be such that generic counters are enabled +* but fixed counters are disabled (true on Penryn and Atom currently). +* +* Here, we are done using the PMU. so we restore the power-on value. +*/ +static void pfm_intel_arch_release_pmu_percpu(void) +{ + /* nothing to do for v1 */ + if (pfm_intel_arch_version < 2) + return; + + PFM_DBG("global_ctrl restored to 0x%llx\n", + __get_cpu_var(saved_global_ctrl)); + + wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, __get_cpu_var(saved_global_ctrl)); +} + +/* + * Counters may have model-specific width. Yet the documentation says + * that only the lower 32 bits can be written to due to the specification + * of wrmsr. bits [32-(w-1)] are sign extensions of bit 31. Bits [w-63] must + * not be set (see rsvd_msk for PMDs). As such the effective width of a + * counter is 31 bits only regardless of what CPUID.0xa returns. + * + * See IA-32 Intel Architecture Software developer manual Vol 3B chapter 18 + */ +static struct pfm_pmu_config pfm_intel_arch_pmu_conf = { + .pmu_name = "Intel architectural", + .pmd_desc = pfm_intel_arch_pmd_desc, + .counter_width = 31, + .num_pmc_entries = PFM_IA_MAX_PMCS, + .num_pmd_entries = PFM_IA_MAX_PMDS, + .pmc_desc = pfm_intel_arch_pmc_desc, + .version = "1.0", + .pmu_info = &pfm_intel_arch_pmu_info +}; + +static int __init pfm_intel_arch_pmu_init_module(void) +{ + if (pfm_intel_arch_probe_pmu()) + return -ENOSYS; + + return pfm_pmu_register(&pfm_intel_arch_pmu_conf); +} + +device_initcall(pfm_intel_arch_pmu_init_module); Index: o3/arch/x86/perfmon/Kconfig =================================================================== --- o3.orig/arch/x86/perfmon/Kconfig 2008-11-25 18:09:47.000000000 +0100 +++ o3/arch/x86/perfmon/Kconfig 2008-11-25 18:21:33.000000000 +0100 @@ -15,4 +15,11 @@ help Enables perfmon debugging support +config X86_PERFMON_INTEL_ARCH + bool "Support for Intel architectural perfmon v1/v2/v3" + depends on PERFMON + default n + help + Enables support for Intel architectural performance counters. + This feature was introduced with Intel Core Solo/Core Duo processors. endmenu -- -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/