From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750882AbVKWO4V (ORCPT ); Wed, 23 Nov 2005 09:56:21 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750885AbVKWO4V (ORCPT ); Wed, 23 Nov 2005 09:56:21 -0500 Received: from ccerelbas03.cce.hp.com ([161.114.21.106]:4060 "EHLO ccerelbas03.cce.hp.com") by vger.kernel.org with ESMTP id S1750880AbVKWO4T (ORCPT ); Wed, 23 Nov 2005 09:56:19 -0500 Date: Wed, 23 Nov 2005 06:55:43 -0800 From: Stephane Eranian To: linux-kernel@vger.kernel.org Cc: Stephane Eranian Subject: Re: [PATCH 3/5] i386 perfmon2 code for review Message-ID: <20051123145543.GD17228@frankl.hpl.hp.com> Reply-To: eranian@hpl.hp.com References: <20051118170055.GF30262@frankl.hpl.hp.com> <20051123145308.GB17228@frankl.hpl.hp.com> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="ABTtc+pdwF7KHXCz" Content-Disposition: inline In-Reply-To: <20051123145308.GB17228@frankl.hpl.hp.com> User-Agent: Mutt/1.4.1i Organisation: HP Labs Palo Alto Address: HP Labs, 1U-17, 1501 Page Mill road, Palo Alto, CA 94304, USA. E-mail: eranian@hpl.hp.com Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org --ABTtc+pdwF7KHXCz Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Hello, For some reasons, the original annoucement from 11/18 for this release never made it to this list. You can download the latest packages from: http://www.sf.net/projects/perfmon2 You need to grab: 2.6.15-rc1-git6 (kernel code) libpfm-3.2-051118 (user library) For convenience, I am also posting the patch directly to the list. At this point, the patch is for review. Please cc me on your replies because due to the volume of traffic I am not a member of the list. I have split the patch into 5 parts. The first part is needed on each platform (common code). All the others are specific to a processor architecture. There is an exception for x86-64 which needs the i386 portion due to cross use of header files. The patch is relative to 2.6.15-rc1-git6. thanks. This E-mail contains Part 2: i386 code (P6, Pentium M, P4/Xeon 32bits) --ABTtc+pdwF7KHXCz Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename="perfmon-new-base-i386-051118.diff" diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/Kconfig linux-2.6.15-rc1-git6/arch/i386/Kconfig --- linux-2.6.15-rc1-git6.orig/arch/i386/Kconfig 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/Kconfig 2005-11-18 05:47:51.000000000 -0800 @@ -684,8 +684,24 @@ depends on HIGHMEM help Generate crash dump after being started by kexec. + +menu "Hardware Performance Monitoring support" + +config PERFMON + bool "Perfmon2 performance monitoring interface" + select X86_LOCAL_APIC + default y + help + include the perfmon2 performance monitoring interface + in the kernel. See for + more details. If you're unsure, say Y. + +source "arch/i386/perfmon/Kconfig" endmenu +endmenu + + menu "Power management options (ACPI, APM)" depends on !X86_VOYAGER diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/Makefile linux-2.6.15-rc1-git6/arch/i386/Makefile --- linux-2.6.15-rc1-git6.orig/arch/i386/Makefile 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/Makefile 2005-11-18 05:47:51.000000000 -0800 @@ -87,6 +87,7 @@ head-y := arch/i386/kernel/head.o arch/i386/kernel/init_task.o libs-y += arch/i386/lib/ +core-$(CONFIG_PERFMON) += arch/i386/perfmon/ core-y += arch/i386/kernel/ \ arch/i386/mm/ \ arch/i386/$(mcore-y)/ \ diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/kernel/apic.c linux-2.6.15-rc1-git6/arch/i386/kernel/apic.c --- linux-2.6.15-rc1-git6.orig/arch/i386/kernel/apic.c 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/kernel/apic.c 2005-11-18 05:47:51.000000000 -0800 @@ -26,6 +26,7 @@ #include #include #include +#include #include #include @@ -1161,6 +1162,8 @@ #endif } + pfm_handle_switch_timeout(); + /* * We take the 'long' return path, and there every subsystem * grabs the apropriate locks (kernel lock/ irq lock). diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/kernel/entry.S linux-2.6.15-rc1-git6/arch/i386/kernel/entry.S --- linux-2.6.15-rc1-git6.orig/arch/i386/kernel/entry.S 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/kernel/entry.S 2005-11-18 05:47:51.000000000 -0800 @@ -430,6 +430,16 @@ /* The include is where all of the SMP etc. interrupts come from */ #include "entry_arch.h" +#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_PERFMON) +ENTRY(pmu_interrupt) + pushl $LOCAL_PERFMON_VECTOR-256 + SAVE_ALL + pushl %esp + call pfm_intr_handler + addl $4, %esp + jmp ret_from_intr +#endif + ENTRY(divide_error) pushl $0 # no error code pushl $do_divide_error diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/kernel/i8259.c linux-2.6.15-rc1-git6/arch/i386/kernel/i8259.c --- linux-2.6.15-rc1-git6.orig/arch/i386/kernel/i8259.c 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/kernel/i8259.c 2005-11-18 05:47:51.000000000 -0800 @@ -11,6 +11,7 @@ #include #include #include +#include #include #include @@ -403,6 +404,7 @@ /* all the set up before the call gates are initialised */ pre_intr_init_hook(); + /* * Cover the whole vector space, no vector can escape * us. (some of these will be overridden and become @@ -427,6 +429,8 @@ */ setup_pit_timer(); + pfm_vector_init(); + /* * External FPU? Set up irq13 if so, for * original braindamaged IBM FERR coupling. diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/kernel/process.c linux-2.6.15-rc1-git6/arch/i386/kernel/process.c --- linux-2.6.15-rc1-git6.orig/arch/i386/kernel/process.c 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/kernel/process.c 2005-11-18 05:47:51.000000000 -0800 @@ -33,6 +33,7 @@ #include #include #include +#include #include #include #include @@ -387,6 +388,7 @@ tss->io_bitmap_base = INVALID_IO_BITMAP_OFFSET; put_cpu(); } + pfm_exit_thread(tsk); } void flush_thread(void) @@ -466,6 +468,8 @@ savesegment(fs,p->thread.fs); savesegment(gs,p->thread.gs); + pfm_copy_thread(p, childregs); + tsk = current; if (unlikely(NULL != tsk->thread.io_bitmap_ptr)) { p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); @@ -727,6 +731,8 @@ disable_tsc(prev_p, next_p); + pfm_ctxswin(next_p); + return prev_p; } diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/kernel/signal.c linux-2.6.15-rc1-git6/arch/i386/kernel/signal.c --- linux-2.6.15-rc1-git6.orig/arch/i386/kernel/signal.c 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/kernel/signal.c 2005-11-18 05:47:51.000000000 -0800 @@ -21,6 +21,7 @@ #include #include #include +#include #include #include #include @@ -667,6 +668,11 @@ regs->eflags |= TF_MASK; clear_thread_flag(TIF_SINGLESTEP); } + /* + * must be done before signals + */ + pfm_handle_work(); + /* deal with pending signal delivery */ if (thread_info_flags & _TIF_SIGPENDING) do_signal(regs,oldset); diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/kernel/syscall_table.S linux-2.6.15-rc1-git6/arch/i386/kernel/syscall_table.S --- linux-2.6.15-rc1-git6.orig/arch/i386/kernel/syscall_table.S 2005-11-18 05:16:42.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/kernel/syscall_table.S 2005-11-18 05:47:51.000000000 -0800 @@ -294,3 +294,15 @@ .long sys_inotify_init .long sys_inotify_add_watch .long sys_inotify_rm_watch + .long sys_pfm_create_context + .long sys_pfm_write_pmcs /* 295 */ + .long sys_pfm_write_pmds + .long sys_pfm_read_pmds + .long sys_pfm_load_context + .long sys_pfm_start + .long sys_pfm_stop /* 300 */ + .long sys_pfm_restart + .long sys_pfm_create_evtsets + .long sys_pfm_getinfo_evtsets + .long sys_pfm_delete_evtsets + .long sys_pfm_unload_context /* 305 */ diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/Kconfig linux-2.6.15-rc1-git6/arch/i386/perfmon/Kconfig --- linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/Kconfig 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/perfmon/Kconfig 2005-11-18 05:47:51.000000000 -0800 @@ -0,0 +1,24 @@ +config PERFMON_P6 + tristate "Support for P6/Pentium M hardware performance counters" + depends on PERFMON + default m + help + Enables support for the P6/Pentium M style hardware performance counters" + If unsure, say M. + +config PERFMON_P4 + tristate "Support for 32-bit P4/Xeon hardware performance counters" + depends on PERFMON + default m + help + Enables support for the 32-bit P4/Xeon style hardware performance counters" + If unsure, say M. + +config PERFMON_P4_PEBS + tristate "Support for Intel P4 PEBS sampling format" + depends on PERFMON_P4 + default m + help + Enables support for Precise Event-Based Sampling (PEBS) on the Intel P4 + processors which support it. Does not work with P6 processors. + If unsure, say m. diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/Makefile linux-2.6.15-rc1-git6/arch/i386/perfmon/Makefile --- linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/Makefile 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/perfmon/Makefile 2005-11-18 05:47:51.000000000 -0800 @@ -0,0 +1,8 @@ +# +# Copyright (c) 2005 Hewlett-Packard Development Company, L.P. +# Contributed by Stephane Eranian +# +obj-$(CONFIG_PERFMON) += perfmon.o +obj-$(CONFIG_PERFMON_P6) += perfmon_p6.o +obj-$(CONFIG_PERFMON_P4) += perfmon_p4.o +obj-$(CONFIG_PERFMON_P4_PEBS) += perfmon_p4_pebs_smpl.o diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon.c linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon.c --- linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon.c 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon.c 2005-11-18 05:47:51.000000000 -0800 @@ -0,0 +1,860 @@ +/* + * This file implements the I386 specific + * support for the perfmon2 interface + * + * Copyright (c) 2005 Hewlett-Packard Development Company, L.P. + * Contributed by Stephane Eranian + */ +#include +#include + +#define MSR_IA32_PEBS_ENABLE 0x3f1 /* unique per-thread */ +#define MSR_IA32_DS_AREA 0x600 /* unique per-thread */ + +/* + * Debug Store (DS) management area (for P4/Xeon PEBS) + */ +struct pfm_ds_area { + unsigned long bts_buf_base; + unsigned long bts_index; + unsigned long bts_abs_max; + unsigned long bts_intr_thres; + unsigned long pebs_buf_base; + unsigned long pebs_index; + unsigned long pebs_abs_max; + unsigned long pebs_intr_thres; + u64 pebs_cnt_reset; +}; + +static void (*pfm_stop_active)(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set); + +static inline void pfm_set_pce(void) +{ + write_cr4(read_cr4() | X86_CR4_PCE); +} + +static inline void pfm_clear_pce(void) +{ + write_cr4(read_cr4() & ~X86_CR4_PCE); +} + +static u8 get_smt_id_mask(void) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + u8 mask = 0; + u8 cnt; + + cnt = arch_info->lps_per_core; + + /* TODO : replace this with the algorithm from 7.10.3 as + * soon as I figure out this bizarre GCC/AT&T + * inline assembler + */ + if (likely(cnt <= 2)) + mask = 0x01; + else if (unlikely(cnt<=4)) + mask = 0x03; + else if (unlikely(cnt<=8)) + mask = 0x07; + else if (unlikely(cnt<=16)) + mask = 0x0F; + + return mask; +} + +/* + * returns the logical processor id (0 or 1) if running on an HT system + */ +static u8 get_smt_id(void) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + u8 apic_id ; + + /* + * not HT enabled + */ + if (arch_info->lps_per_core <= 1) + return 0; + + /* + * TODO : add per_cpu caching here to avoid calling cpuid repeatedly + */ + apic_id = (cpuid_ebx(1)&0xff000000) >> 24; + return apic_id & get_smt_id_mask(); +} + +void __pfm_write_reg(const struct pfm_arch_ext_reg *xreg, u64 val) +{ + unsigned long pmi; + int smt_id = get_smt_id(); + + if (unlikely(smt_id > MAX_SMT_ID)) { + printk(KERN_ERR "%s: smt_id (%d) exceeds maximun number of SMTs" + " (%d) supported by perfmon", + __FUNCTION__, smt_id, MAX_SMT_ID); + return; + } + + /* + * LBR TOS is read only + * TODO : need to save the TOS value somewhere and move the LBRs according + * to where ever it is currently pointing + */ + if (xreg->reg_type == PFM_REGT_LBRTOS) + return; + + /* + * Adjust for T1 if necessary: + * + * - move the T0_OS/T0_USR bits into T1 slots + * - move the OVF_PMI_T0 bits into T1 slot + * + * The P4 write checker systematically clears T1 + * fields, i.e., user only works with T0. + */ + if (smt_id > 0) { + if (xreg->reg_type == PFM_REGT_ESCR) { + val |= ((val & 0xc) >> 2); + /* clear bits 3:2 */ + val &= ~0xc; + } else if (xreg->reg_type == PFM_REGT_CCCR) { + pmi = (val >> 26) & 0x1; + if (pmi) { + val &=~(1UL<<26); + val |= 1UL<<27; + } + } + } + + if (xreg->addrs[smt_id]) { + DPRINT(("[0x%lx]=0x%llx\n", xreg->addrs[smt_id], val)); + wrmsrl(xreg->addrs[smt_id], val); + } +} + +void __pfm_read_reg(const struct pfm_arch_ext_reg *xreg, u64 *val) +{ + int smt_id = get_smt_id(); + + *val = 0; + + if (unlikely(smt_id > MAX_SMT_ID)) { + printk(KERN_ERR "%s: smt_id (%d) exceeds maximun number of SMTs" + " (%d) supported by perfmon", + __FUNCTION__, smt_id, MAX_SMT_ID); + return; + } + + if (xreg->addrs[smt_id] != 0) { + rdmsrl(xreg->addrs[smt_id], *val); + /* + * move the Tx_OS and Tx_USR bits into + * T0 slots setting the T1 slots to zero + */ + if (xreg->reg_type == PFM_REGT_ESCR) { + + if (smt_id > 0) + *val |= (((*val) & 0x3) << 2); + + /* + * zero out bits that are reserved + * (including T1_OS and T1_USR) + */ + *val &= PFM_ESCR_PMC_RSVD; + } + DPRINT(("[0x%lx]=0x%llx\n", + xreg->addrs[smt_id], + (unsigned long long)*val)); + } +} + +void pfm_arch_init_percpu(void) +{ + /* + * We initialize APIC with LVTPC vector masked. + * + * this is motivated by the fact that the PMU may be + * in a condition where is has already an interrupt pending + * as with boot. Given that we cannot touch the PMU registers + * at this point, we may not have a way to remove the condition. + * As such, we need to keep the interrupt masked until a PMU + * description is loaded. At that point, we can enable intr. + */ + apic_write(APIC_LVTPC, APIC_LVT_MASKED|LOCAL_PERFMON_VECTOR); + + printk("perfmon: CPU%d APIC LVTPC vector masked\n", smp_processor_id()); +} + +void pfm_arch_context_initialize(struct pfm_context *ctx, u32 ctx_flags) +{ + struct pfm_arch_context *ctx_arch; + + ctx_arch = pfm_ctx_arch(ctx); + ctx_arch->flags = ctx_flags & PFM_I386_FL_INSECURE; +} + +/* + * function called from pfm_load_context_*(). Task is not guaranteed to be + * current task. If not then other task is guaranteed stopped and off any CPU. + * context is locked and interrupts are masked. + * + * On PFM_LOAD_CONTEXT, the interface guarantees monitoring is stopped. + * + * For system-wide task is NULL + */ +int pfm_arch_load_context(struct pfm_context *ctx, struct task_struct *task) +{ + struct pfm_arch_context *ctx_arch; + + ctx_arch = pfm_ctx_arch(ctx); + + /* + * always authorize user level rdpmc for self-monitoring + */ + if (task == current || ctx->ctx_fl_system) { + /* + * for system-wide or self-monitoring, we always enable + * rdpmc at user level + */ + ctx_arch->flags |= PFM_I386_FL_INSECURE; + + pfm_set_pce(); + DPRINT(("setting cr4.pce (rdpmc authorized at user level)\n")); + } + return 0; +} + +/* + * function called from pfm_unload_context_*(). Context is locked. + * interrupts are masked. task is not guaranteed to be current task. + * Access to PMU is not guaranteed. + * + * function must do whatever arch-specific action is required on unload + * of a context. + * + * called for both system-wide and per-thread. task is NULL for ssytem-wide + */ +void pfm_arch_unload_context(struct pfm_context *ctx, struct task_struct *task) +{ + struct pfm_arch_context *ctx_arch; + + ctx_arch = pfm_ctx_arch(ctx); + + if (ctx_arch->flags & PFM_I386_FL_INSECURE) { + pfm_clear_pce(); + DPRINT(("clearing cr4.pce\n")); + } +} + +/* + * called from __pfm_interrupt_handler(). ctx is not NULL. + * ctx is locked. PMU interrupt is masked. + * + * must stop all monitoring to ensure handler has consistent view. + * must collect overflowed PMDs bitmask into set_povfls_pmds and + * set_npend_ovfls. If no interrupt detected then set_npend_ovfls + * must be set to zero. + */ +void pfm_arch_intr_freeze_pmu(struct pfm_context *ctx) +{ + struct pfm_arch_pmu_info *arch_info; + struct pfm_arch_context *ctx_arch; + struct pfm_event_set *set; + struct pfm_ds_area *ds; + + arch_info = pfm_pmu_conf->arch_info; + ctx_arch = pfm_ctx_arch(ctx); + + set = ctx->ctx_active_set; + + /* + * stop active monitoring and collect overflow information + */ + pfm_stop_active(current, ctx, set); + + /* + * PMU is stopped, thus PEBS is stopped already + * on PEBS full interrupt, the IQ_CCCR4 counter does + * not have the OVF bit set. Thus we use the pebs index + * to detect overflow. This is required because we may + * have more than one reason for overflow due to 64-bit + * counter virtualization. + * + * We don't actually keep track of the overflow unless + * IQ_CTR4 is actually used. + * + * With HT enabled, the mappings are such that IQ_CTR4 and IQ_CTR5 + * are mapped onto the same PMD registers. + * + * XXX: remove assumption on IQ_CTR4 and IQ_CTR5 being mapped onto + * the same PMD. + */ + if (ctx_arch->flags & PFM_I386_FL_USE_PEBS) { + ds = ctx_arch->ds_area; + if (ds->pebs_index >= ds->pebs_intr_thres + && pfm_bv_isset(set->set_used_pmds, arch_info->pebs_ctr_idx)) { + pfm_bv_set(set->set_povfl_pmds, arch_info->pebs_ctr_idx); + set->set_npend_ovfls++; + } + } +} + +/* + * unfreeze PMU from pfm_do_interrupt_handler() + * ctx may be NULL for spurious + */ +void pfm_arch_intr_unfreeze_pmu(struct pfm_context *ctx) +{ + struct pfm_arch_context *ctx_arch; + + if (ctx == NULL) + return; + + ctx_arch = pfm_ctx_arch(ctx); + + pfm_arch_restore_pmcs(ctx, ctx->ctx_active_set); + + /* + * reload DS area pointer because it is cleared by + * pfm_stop_active() + */ + if (ctx_arch->flags & PFM_I386_FL_USE_PEBS) { + wrmsr(MSR_IA32_DS_AREA, ctx_arch->ds_area, 0); + DPRINT(("restoring DS\n")); + } +} + +/* + * Called from pfm_ctxswin_*(). Task is guaranteed to be current. + * set cannot be NULL. Context is locked. Interrupts are masked. + * Caller has already restored all PMD and PMC registers. + * + * must reactivate monitoring + */ +void pfm_arch_ctxswin(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set) +{ + struct pfm_arch_context *ctx_arch; + + ctx_arch = pfm_ctx_arch(ctx); + + /* + * nothing to do for system-wide + */ + if (ctx->ctx_fl_system) + return; + + if (ctx_arch->flags & PFM_I386_FL_INSECURE) { + pfm_set_pce(); + DPRINT(("setting cr4.pce\n")); + } + + /* + * reload DS management area pointer. Pointer + * not managed as a PMC hus it is not restored + * with the rest of the registers. + */ + if (ctx_arch->flags & PFM_I386_FL_USE_PEBS) { + wrmsr(MSR_IA32_DS_AREA, ctx_arch->ds_area, 0); + DPRINT(("restored DS\n")); + } +} + +static void __pfm_stop_active_p6(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + struct pfm_arch_ext_reg *xrc, *xrd; + unsigned int i, max; + u64 val, wmask; + + max = pfm_pmu_conf->max_pmc; + xrc = arch_info->pmc_addrs; + xrd = arch_info->pmd_addrs; + wmask = PFM_ONE_64 << pfm_pmu_conf->counter_width; + + /* + * clear enable bits + */ + for (i = 0; i < max; i++) { + if (pfm_bv_isset(set->set_used_pmcs, i)) + wrmsr(xrc[i].addrs[0], 0, 0); + } + + if (set->set_npend_ovfls) + return; + + /* + * check for pending overflows + */ + max = pfm_pmu_conf->max_cnt_pmd; + for (i = 0; i < max; i++) { + if (pfm_bv_isset(set->set_used_pmds, i)) { + rdmsrl(xrd[i].addrs[0], val); + DPRINT_ovfl(("pmd%d[0x%lx]=0x%llx sw_pmd=0x%llx\n", i, xrd[i].addrs[0], val, set->set_view->set_pmds[i])); + if ((val & wmask) == 0) { + DPRINT_ovfl(("pmd%d overflow\n", i)); + pfm_bv_set(set->set_povfl_pmds, i); + set->set_npend_ovfls++; + } + + } + } +} + +/* + * stop active set only + */ +static void __pfm_stop_active_p4(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + struct pfm_arch_context *ctx_arch; + struct pfm_arch_ext_reg *xrc, *xrd; + unsigned long enable_mask[PFM_PMC_BV]; + unsigned int i, max_pmc; + u64 cccr, ctr1, ctr2; + + ctx_arch = pfm_ctx_arch(ctx); + max_pmc = pfm_pmu_conf->max_pmc; + xrc = arch_info->pmc_addrs; + xrd = arch_info->pmd_addrs; + + bitmap_and(enable_mask, set->set_used_pmcs, arch_info->enable_mask, max_pmc); + + /* + * stop PEBS and clear DS area pointer + */ + if (ctx_arch->flags & PFM_I386_FL_USE_PEBS) { + wrmsr(MSR_IA32_PEBS_ENABLE, 0, 0); + wrmsr(MSR_IA32_DS_AREA, 0, 0); + } + + /* + * ensures we do not destroy information collected + * at ctxswout. This function is called from + * pfm_arch_intr_freeze_pmu() as well. + */ + if (set->set_npend_ovfls) { + for (i = 0; i < max_pmc; i++) { + if (pfm_bv_isset(enable_mask, i)) { + __pfm_write_reg(xrc+i, 0); + } + } + return; + } + + for (i = 0; i < max_pmc; i++) { + + if (pfm_bv_isset(enable_mask, i)) { + + /* read counter controlled by PMC */ + __pfm_read_reg(xrd+(xrc[i].ctr), &ctr1); + + /* read PMC value */ + __pfm_read_reg(xrc+i, &cccr); + + /* clear CCCR value (destroy OVF) */ + __pfm_write_reg(xrc+i, 0); + + /* read counter controlled by CCCR again */ + __pfm_read_reg(xrd+(xrc[i].ctr), &ctr2); + + /* + * there is an overflow if either: + * - CCCR.ovf is set (and we just cleared it) + * - ctr2 < ctr1 + * in that case we set the bit corresponding to the + * overflowed PMD in povfl_pmds. + */ + if ((cccr & (PFM_ONE_64<<31)) || (ctr2 < ctr1)) { + pfm_bv_set(set->set_povfl_pmds, xrc[i].ctr); + set->set_npend_ovfls++; + } + } + } + +} + +/* + * Called from pfm_stop() and pfm_ctxswin_*() when idle + * task and EXCL_IDLE is on. + * + * Interrupts are masked. Context is locked. Set is the active set. + * + * For per-thread: + * task is not necessarily current. If not current task, then + * task is guaranteed stopped and off any cpu. Access to PMU + * is not guaranteed. Interrupts are masked. Context is locked. + * Set is the active set. + * + * For system-wide: + * task is current + * + * must disable active monitoring. + */ +void pfm_arch_stop(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set) +{ + if (task != current) + return; + + pfm_stop_active(task, ctx, set); +} + +/* + * Called from pfm_ctxswout_*(). Task is guaranteed to be current. + * Context is locked. Interrupts are masked. Monitoring is active. + * PMU access is guaranteed. PMC and PMD registers are live in PMU. + * + * for per-thread: + * must stop monitoring for the task + * for system-wide: + * must ensure task has monitoring stopped. But monitoring may continue + * on the current processor + */ +void pfm_arch_ctxswout(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set) +{ + struct pfm_arch_context *ctx_arch; + + ctx_arch = pfm_ctx_arch(ctx); + + /* + * nothing to do in system-wide + */ + if (ctx->ctx_fl_system) + return; + + if (ctx_arch->flags & PFM_I386_FL_INSECURE) { + pfm_clear_pce(); + DPRINT(("clearing cr4.pce\n")); + } + + /* + * disable lazy restore of PMCS on ctxswin because + * we modify some of them. + */ + set->set_priv_flags |= PFM_SETFL_PRIV_MOD_PMCS; + + pfm_stop_active(task, ctx, set); +} + +/* + * called from pfm_start() or pfm_ctxswout_sys() when idle task and + * EXCL_IDLE is on. + * + * Interrupts are masked. Context is locked. Set is the active set. + * + * For per-trhead: + * Task is not necessarily current. If not current task, then task + * is guaranteed stopped and off any cpu. Access to PMU is not guaranteed. + * + * For system-wide: + * task is always current + * + * must enable active monitoring. + */ +static void __pfm_arch_start(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + struct pfm_arch_context *ctx_arch; + struct pfm_arch_ext_reg *xregs; + unsigned long enable_mask[PFM_PMC_BV]; + unsigned int i, max_pmc; + + if (task != current) + return; + + ctx_arch = pfm_ctx_arch(ctx); + max_pmc = pfm_pmu_conf->max_pmc; + xregs = arch_info->pmc_addrs; + + /* + * build bitmask of PMC to enable: intersection of used PMCs with + * implemented PMCs with an enable bit. + */ + bitmap_and(enable_mask, set->set_used_pmcs, arch_info->enable_mask, max_pmc); + + for (i = 0; i < max_pmc; i++) { + if (pfm_bv_isset(enable_mask, i)) { + __pfm_write_reg(xregs+i, set->set_pmcs[i]); + } + } + + /* + * reload DS area pointer. PEBS_ENABLE is restored with the PMCs + * in pfm_restore_pmcs(). PEBS_ENABLE is not considered part of + * the set of PMCs with an enable bit. This is reserved for counter + * PMC, i.e., CCCR. + */ + if (task == current && (ctx_arch->flags & PFM_I386_FL_USE_PEBS)) { + wrmsr(MSR_IA32_DS_AREA, ctx_arch->ds_area, 0); + DPRINT(("restoring DS\n")); + } +} + +void pfm_arch_start(struct task_struct *task, struct pfm_context *ctx, + struct pfm_event_set *set) +{ + /* + * mask/unmask uses start/stop mechanism, so we cannot allow + * while masked. + */ + if (ctx->ctx_state == PFM_CTX_MASKED) + return; + + __pfm_arch_start(task, ctx, set); +} + +/* + * function called from pfm_switch_sets(), pfm_context_load_thread(), + * pfm_context_load_sys(), pfm_ctxswin_*(), pfm_switch_sets() + * context is locked. Interrupts are masked. set cannot be NULL. + * Access to the PMU is guaranteed. + * + * function must restore all PMD registers from set. + */ +void pfm_arch_restore_pmds(struct pfm_context *ctx, struct pfm_event_set *set) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + struct pfm_arch_ext_reg *xregs; + u64 ovfl_mask, val, *pmds; + unsigned long *impl_pmds; + unsigned int i, max_pmd; + + max_pmd = pfm_pmu_conf->max_pmd; + ovfl_mask = pfm_pmu_conf->ovfl_mask; + impl_pmds = pfm_pmu_conf->impl_pmds; + pmds = set->set_view->set_pmds; + xregs = arch_info->pmd_addrs; + + /* + * must restore all pmds to avoid leaking + * information to user. + * + * XXX: should check PFM_I386_FL_INSECURE and use used_pmd instead + */ + for (i = 0; i < max_pmd; i++) { + + if (pfm_bv_isset(impl_pmds, i) == 0) + continue; + + val = pmds[i]; + + /* + * set upper bits for counter to ensure + * overflow will trigger + */ + if (xregs[i].reg_type == PFM_REGT_CTR) + val |= ~ovfl_mask; + + __pfm_write_reg(xregs+i, val); + } +} + +/* + * function called from pfm_switch_sets(), pfm_context_load_thread(), + * pfm_context_load_sys(), pfm_ctxswin_*(). + * Context is locked. Interrupts are masked. set cannot be NULL. + * Access to the PMU is guaranteed. + * + * function must restore all PMC registers from set, if needed. + */ +void pfm_arch_restore_pmcs(struct pfm_context *ctx, struct pfm_event_set *set) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + struct pfm_arch_ext_reg *xregs; + unsigned long *impl_pmcs; + unsigned int i, max_pmc; + + max_pmc = pfm_pmu_conf->max_pmc; + xregs = arch_info->pmc_addrs; + impl_pmcs = pfm_pmu_conf->impl_pmcs; + + /* + * - by default, no PMC measures anything + * - on ctxswout, all used PMCs are disabled (cccr cleared) + * + * we need to restore the PMC (incl enable bits) only if + * not masked and user issued pfm_start() + */ + if (ctx->ctx_state == PFM_CTX_MASKED || ctx->ctx_fl_started == 0) + return; + + /* + * restore all pmcs and not just the ones that are used + * to avoid using stale registers + * + * XXX: could we restrict to just used_pmcs? + */ + for (i = 0; i < max_pmc; i++) + if (pfm_bv_isset(impl_pmcs, i)) + __pfm_write_reg(xregs+i, set->set_pmcs[i]); +} + +/* + * function called from pfm_mask_monitoring(), pfm_switch_sets(), + * pfm_ctxswout_thread(), pfm_flush_pmds(). + * context is locked. interrupts are masked. the set argument cannot + * be NULL. Access to PMU is guaranteed. + * + * function must saved PMD registers into set save area set_pmds[] + */ +void pfm_arch_save_pmds(struct pfm_context *ctx, struct pfm_event_set *set) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + struct pfm_arch_ext_reg *xregs; + u64 val, *pmds, ovfl_mask; + unsigned long *used_pmds; + unsigned int i, max_pmd; + + ovfl_mask = pfm_pmu_conf->ovfl_mask; + max_pmd = pfm_pmu_conf->max_pmd; + used_pmds = set->set_used_pmds; + pmds = set->set_view->set_pmds; + xregs = arch_info->pmd_addrs; + + /* + * save all used pmds + */ + for (i = 0; i < max_pmd; i++) { + if (pfm_bv_isset(used_pmds, i)) { + + __pfm_read_reg(xregs+i, &val); + + if (xregs[i].reg_type == PFM_REGT_CTR) + val = (pmds[i] & ~ovfl_mask) | (val & ovfl_mask); + + pmds[i] = val; + } + } +} + +asmlinkage void pfm_intr_handler(struct pt_regs *regs) +{ + ack_APIC_irq(); + irq_enter(); + pfm_interrupt_handler(LOCAL_PERFMON_VECTOR, NULL, regs); + irq_exit(); + /* + * On Intel P6, Pentium M, P4: + * - it is necessary to clear the MASK field for the LVTPC + * vector. Otherwise interrupts will stay masked. See + * section 8.5.1 + * + * We only clear the mask field, if there is a PMU description + * loaded. Otherwise we would have a problem because without + * PMU description we cannot access PMu registers to clear the + * overfow condition and may end up in a flood of PMU interrupts. + * + * The APIC vector is initialized as masked, but we may already + * have a pending PMU overflow by the time we get to + * pfm_arch_init_percpu(). Such interrupt would generate a call + * to this function, which would undo the masking and would + * cause a flood. + */ + if (pfm_pmu_conf) apic_write(APIC_LVTPC, LOCAL_PERFMON_VECTOR); +} + +asmlinkage void pmu_interrupt(void); +void pfm_vector_init(void) +{ + set_intr_gate(LOCAL_PERFMON_VECTOR, (void *)pmu_interrupt); + printk("CPU%d installed perfmon gate\n", smp_processor_id()); +} + +static void __pfm_stop_one_pmu(void *data) +{ + struct pfm_pmu_config *cfg = data; + struct pfm_arch_pmu_info *arch_info = cfg->arch_info; + struct pfm_arch_ext_reg *xregs; + unsigned int i, num_pmcs; + + DPRINT(("stopping on CPU%d: LVT=0x%x\n", + smp_processor_id(), + (unsigned int)apic_read(APIC_LVTPC))); + + num_pmcs = cfg->num_pmcs; + xregs = arch_info->pmc_addrs; + + for (i = 0; i < num_pmcs; i++) + if (pfm_bv_isset(arch_info->enable_mask, i)) + __pfm_write_reg(xregs+i, 0); + + /* + * now that we have a PMU description we can deal with spurious + * interrupts, thus we can safely re-enable the LVTPC vector + * by clearing the mask field + */ + apic_write(APIC_LVTPC, LOCAL_PERFMON_VECTOR); +} + +/* + * called from pfm_register_pmu_config() after the new + * config has been validated and installed. The pfm_session_lock + * is held. + * + * Must sanity check the arch-specific config information + * + * return: + * < 0 : if error + * 0 : if success + */ +int pfm_arch_pmu_config_check(struct pfm_pmu_config *cfg) +{ + struct pfm_arch_pmu_info *arch_info = cfg->arch_info; + + switch(arch_info->pmu_style) { + case PFM_I386_PMU_P4: + pfm_stop_active = __pfm_stop_active_p4; + break; + case PFM_I386_PMU_P6: + pfm_stop_active = __pfm_stop_active_p6; + break; + default: + printk(KERN_ERR"unknown pmu_style=%d\n", arch_info->pmu_style); + return -1; + } + return 0; +} + +/* + * called from pfm_register_pmu_config() after the new + * config has been validated and installed. No lock + * is held. Interrupts are not masked. + * + * The role of the function is, based on the PMU description, to + * put the PMU into a quiet state on each CPU. This function is + * not necessary if there is an architected way of doing this + * for a processor family. + */ +void pfm_arch_pmu_config_init(struct pfm_pmu_config *cfg) +{ + on_each_cpu(__pfm_stop_one_pmu, cfg, 1, 1); +} + +int pfm_arch_initialize(void) +{ + return 0; +} + +void pfm_arch_mask_monitoring(struct pfm_context *ctx) +{ + /* + * on IA-32 masking/unmasking uses start/stop + * mechanism + */ + pfm_arch_stop(current, ctx, ctx->ctx_active_set); +} + +void pfm_arch_unmask_monitoring(struct pfm_context *ctx) +{ + /* + * on IA-32 masking/unmasking uses start/stop + * mechanism + */ + __pfm_arch_start(current, ctx, ctx->ctx_active_set); +} diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon_p4.c linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon_p4.c --- linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon_p4.c 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon_p4.c 2005-11-18 05:54:51.000000000 -0800 @@ -0,0 +1,442 @@ +/* + * This file contains the P4/Xeon 32-bit PMU register description tables + * and pmc checker used by perfmon.c. + * + * Copyright (c) 2005 Intel Corporation + * Contributed by Bryan Wilkerson + */ +#include +#include +#include +#include + +MODULE_AUTHOR("Bryan Wilkerson "); +MODULE_DESCRIPTION("32-bit P4/Xeon PMU description tables"); +MODULE_LICENSE("GPL"); + +/* + * CCCR default value: + * - OVF_PMI_T0=1 (bit 26) + * - OVF_PMI_T1=0 (bit 27) (set if necessary in pfm_write_reg()) + * - all other bits are zero + * + * OVF_PMI is force to zero if PFM_REGFL_NO_EMUL64 is set on CCCR + */ +#define PFM_CCCR_DFL (1ULL<<26) + +/* + * CCCR reserved fields: + * - bits 0-11, 25-29, 31-63 + * - OVF_PMI (26-27), override with REGFL_NO_EMUL64 + */ +#define PFM_CCCR_PMC_RSVD 0x0000000041fff000 + +#define PMC_NOHT_IDX 32 /* index of no HT PMC namespace */ +#define PMD_NOHT_IDX 9 /* index of no HT PMD namespace */ + +/* + * With HyperThreading enabled: + * + * The ESCRs and CCCRs are divided in half with the top half + * belonging to logical processor 0 and the bottom half going to + * logical processor 1. Thus only half of the PMU resources are + * accessible to applications. + * + * PEBS is not available due to the fact that: + * - MSR_PEBS_MATRIX_VERT is shared between the threads + * - IA32_PEBS_ENABLE is shared between the threads + * + * With HyperThreading disabled: + * + * The full set of PMU resources is exposed to applications. + * + * The mapping is chosen such that PMCxx -> MSR is the same + * in HT and non HT mode, if register is present in HT mode. + * + */ +static struct pfm_arch_pmu_info pfm_p4_pmu_info={ + .pmc_addrs = { + /*pmc 0 */ {{0x3b2, 0x3b3}, 0, PFM_REGT_ESCR}, /* BPU_ESCR0,1 */ + /*pmc 1 */ {{0x3b4, 0x3b5}, 0, PFM_REGT_ESCR}, /* IS_ESCR0,1 */ + /*pmc 2 */ {{0x3aa, 0x3ab}, 0, PFM_REGT_ESCR}, /* MOB_ESCR0,1 */ + /*pmc 3 */ {{0x3b6, 0x3b7}, 0, PFM_REGT_ESCR}, /* ITLB_ESCR0,1 */ + /*pmc 4 */ {{0x3ac, 0x3ad}, 0, PFM_REGT_ESCR}, /* PMH_ESCR0,1 */ + /*pmc 5 */ {{0x3c8, 0x3c9}, 0, PFM_REGT_ESCR}, /* IX_ESCR0,1 */ + /*pmc 6 */ {{0x3a2, 0x3a3}, 0, PFM_REGT_ESCR}, /* FSB_ESCR0,1 */ + /*pmc 7 */ {{0x3a0, 0x3a1}, 0, PFM_REGT_ESCR}, /* BSU_ESCR0,1 */ + /*pmc 8 */ {{0x3c0, 0x3c1}, 0, PFM_REGT_ESCR}, /* MS_ESCR0,1 */ + /*pmc 9 */ {{0x3c4, 0x3c5}, 0, PFM_REGT_ESCR}, /* TC_ESCR0,1 */ + /*pmc 10*/ {{0x3c2, 0x3c3}, 0, PFM_REGT_ESCR}, /* TBPU_ESCR0,1 */ + /*pmc 11*/ {{0x3a6, 0x3a7}, 0, PFM_REGT_ESCR}, /* FLAME_ESCR0,1 */ + /*pmc 12*/ {{0x3a4, 0x3a5}, 0, PFM_REGT_ESCR}, /* FIRM_ESCR0,1 */ + /*pmc 13*/ {{0x3ae, 0x3af}, 0, PFM_REGT_ESCR}, /* SAAT_ESCR0,1 */ + /*pmc 14*/ {{0x3b0, 0x3b1}, 0, PFM_REGT_ESCR}, /* U2L_ESCR0,1 */ + /*pmc 15*/ {{0x3a8, 0x3a9}, 0, PFM_REGT_ESCR}, /* DAC_ESCR0,1 */ + /*pmc 16*/ {{0x3ba, 0x3bb}, 0, PFM_REGT_ESCR}, /* IQ_ESCR0,1 */ + /*pmc 17*/ {{0x3ca, 0x3cb}, 0, PFM_REGT_ESCR}, /* ALF_ESCR0,1 */ + /*pmc 18*/ {{0x3bc, 0x3bd}, 0, PFM_REGT_ESCR}, /* RAT_ESCR0,1 */ + /*pmc 19*/ {{0x3be, 0}, 0, PFM_REGT_ESCR}, /* SSU_ESCR0 */ + /*pmc 20*/ {{0x3b8, 0x3b9}, 0, PFM_REGT_ESCR}, /* CRU_ESCR0,1 */ + /*pmc 21*/ {{0x3cc, 0x3cd}, 0, PFM_REGT_ESCR}, /* CRU_ESCR2,3 */ + /*pmc 22*/ {{0x3e0, 0x3e1}, 0, PFM_REGT_ESCR}, /* CRU_ESCR4,5 */ + + /*pmc 23*/ {{0x360, 0x362}, 0, PFM_REGT_CCCR}, /* BPU_CCCR0,1 */ + /*pmc 24*/ {{0x361, 0x363}, 1, PFM_REGT_CCCR}, /* BPU_CCCR2,3 */ + /*pmc 25*/ {{0x364, 0x366}, 2, PFM_REGT_CCCR}, /* MS_CCCR0,1 */ + /*pmc 26*/ {{0x365, 0x367}, 3, PFM_REGT_CCCR}, /* MS_CCCR2,3 */ + /*pmc 27*/ {{0x368, 0x36a}, 4, PFM_REGT_CCCR}, /* FLAME_CCCR0,1 */ + /*pmc 28*/ {{0x369, 0x36b}, 5, PFM_REGT_CCCR}, /* FLAME_CCCR2,3 */ + /*pmc 29*/ {{0x36c, 0x36e}, 6, PFM_REGT_CCCR}, /* IQ_CCCR0,1 */ + /*pmc 30*/ {{0x36d, 0x36f}, 7, PFM_REGT_CCCR}, /* IQ_CCCR2,3 */ + /*pmc 31*/ {{0x370, 0x371}, 8, PFM_REGT_CCCR}, /* IQ_CCCR4,5 */ + /* non HT extensions */ + /*pmc 32*/ {{0x3b3, 0}, 0, PFM_REGT_ESCR}, /* BPU_ESCR1 */ + /*pmc 33*/ {{0x3b5, 0}, 0, PFM_REGT_ESCR}, /* IS_ESCR1 */ + /*pmc 34*/ {{0x3ab, 0}, 0, PFM_REGT_ESCR}, /* MOB_ESCR1 */ + /*pmc 35*/ {{0x3b7, 0}, 0, PFM_REGT_ESCR}, /* ITLB_ESCR1 */ + /*pmc 36*/ {{0x3ad, 0}, 0, PFM_REGT_ESCR}, /* PMH_ESCR1 */ + /*pmc 37*/ {{0x3c9, 0}, 0, PFM_REGT_ESCR}, /* IX_ESCR1 */ + /*pmc 38*/ {{0x3a3, 0}, 0, PFM_REGT_ESCR}, /* FSB_ESCR1 */ + /*pmc 39*/ {{0x3a1, 0}, 0, PFM_REGT_ESCR}, /* BSU_ESCR1 */ + /*pmc 40*/ {{0x3c1, 0}, 0, PFM_REGT_ESCR}, /* MS_ESCR1 */ + /*pmc 41*/ {{0x3c5, 0}, 0, PFM_REGT_ESCR}, /* TC_ESCR1 */ + /*pmc 42*/ {{0x3c3, 0}, 0, PFM_REGT_ESCR}, /* TBPU_ESCR1 */ + /*pmc 43*/ {{0x3a7, 0}, 0, PFM_REGT_ESCR}, /* FLAME_ESCR1 */ + /*pmc 44*/ {{0x3a5, 0}, 0, PFM_REGT_ESCR}, /* FIRM_ESCR1 */ + /*pmc 45*/ {{0x3af, 0}, 0, PFM_REGT_ESCR}, /* SAAT_ESCR1 */ + /*pmc 46*/ {{0x3b1, 0}, 0, PFM_REGT_ESCR}, /* U2L_ESCR1 */ + /*pmc 47*/ {{0x3a9, 0}, 0, PFM_REGT_ESCR}, /* DAC_ESCR1 */ + /*pmc 48*/ {{0x3bb, 0}, 0, PFM_REGT_ESCR}, /* IQ_ESCR1 */ + /*pmc 49*/ {{0x3cb, 0}, 0, PFM_REGT_ESCR}, /* ALF_ESCR1 */ + /*pmc 50*/ {{0x3bd, 0}, 0, PFM_REGT_ESCR}, /* RAT_ESCR1 */ + /*pmc 51*/ {{0x3b9, 0}, 0, PFM_REGT_ESCR}, /* CRU_ESCR1 */ + /*pmc 52*/ {{0x3cd, 0}, 0, PFM_REGT_ESCR}, /* CRU_ESCR3 */ + /*pmc 53*/ {{0x3e1, 0}, 0, PFM_REGT_ESCR}, /* CRU_ESCR5 */ + /*pmc 54*/ {{0x362, 0}, 9, PFM_REGT_CCCR}, /* BPU_CCCR1 */ + /*pmc 55*/ {{0x363, 0},10, PFM_REGT_CCCR}, /* BPU_CCCR3 */ + /*pmc 56*/ {{0x366, 0},11, PFM_REGT_CCCR}, /* MS_CCCR1 */ + /*pmc 57*/ {{0x367, 0},12, PFM_REGT_CCCR}, /* MS_CCCR3 */ + /*pmc 58*/ {{0x36a, 0},13, PFM_REGT_CCCR}, /* FLAME_CCCR1 */ + /*pmc 59*/ {{0x36b, 0},14, PFM_REGT_CCCR}, /* FLAME_CCCR3 */ + /*pmc 60*/ {{0x36e, 0},15, PFM_REGT_CCCR}, /* IQ_CCCR1 */ + /*pmc 61*/ {{0x36f, 0},16, PFM_REGT_CCCR}, /* IQ_CCCR3 */ + /*pmc 62*/ {{0x371, 0},17, PFM_REGT_CCCR}, /* IQ_CCCR5 */ + /*pmc 63*/ {{0x3f2, 0}, 0, PFM_REGT_DEBUG},/* PEBS_MATRIX_VERT */ + /*pmc 64*/ {{0x3f1, 0}, 0, PFM_REGT_DEBUG} /* PEBS_ENABLE */ + }, + + .pmd_addrs = { + /*pmd 0 */ {{0x300, 0x302}, 0, PFM_REGT_CTR}, /* BPU_CRT0,1 */ + /*pmd 1 */ {{0x301, 0x303}, 0, PFM_REGT_CTR}, /* BPU_CTR2,3 */ + /*pmd 2 */ {{0x304, 0x306}, 0, PFM_REGT_CTR}, /* MS_CRT0,1 */ + /*pmd 3 */ {{0x305, 0x307}, 0, PFM_REGT_CTR}, /* MS_CRT2,3 */ + /*pmd 4 */ {{0x308, 0x30a}, 0, PFM_REGT_CTR}, /* FLAME_CRT0,1 */ + /*pmd 5 */ {{0x309, 0x30b}, 0, PFM_REGT_CTR}, /* FLAME_CRT2,3 */ + /*pmd 6 */ {{0x30c, 0x30e}, 0, PFM_REGT_CTR}, /* IQ_CRT0,1 */ + /*pmd 7 */ {{0x30d, 0x30f}, 0, PFM_REGT_CTR}, /* IQ_CRT2,3 */ + /*pmd 8 */ {{0x310, 0x311}, 0, PFM_REGT_CTR}, /* IQ_CRT4,5 */ + /* + * non HT extensions + */ + /*pmd 9 */ {{0x302, 0}, 0, PFM_REGT_CTR}, /* BPU_CRT1 */ + /*pmd 10*/ {{0x303, 0}, 0, PFM_REGT_CTR}, /* BPU_CTR3 */ + /*pmd 11*/ {{0x306, 0}, 0, PFM_REGT_CTR}, /* MS_CRT1 */ + /*pmd 12*/ {{0x307, 0}, 0, PFM_REGT_CTR}, /* MS_CRT3 */ + /*pmd 13*/ {{0x30a, 0}, 0, PFM_REGT_CTR}, /* FLAME_CRT1 */ + /*pmd 14*/ {{0x30b, 0}, 0, PFM_REGT_CTR}, /* FLAME_CRT3 */ + /*pmd 15*/ {{0x30e, 0}, 0, PFM_REGT_CTR}, /* IQ_CRT1 */ + /*pmd 16*/ {{0x30f, 0}, 0, PFM_REGT_CTR}, /* IQ_CRT3 */ + /*pmd 17*/ {{0x311, 0}, 0, PFM_REGT_CTR}, /* IQ_CRT5 */ + }, + .pebs_ctr_idx = 8, /* thread0: IQ_CTR4, thread1: IQ_CTR5 */ + .pmu_style = PFM_I386_PMU_P4, + .lps_per_core = 1 +}; + +/* + * dependency rules: + * CTR -> CCCR: one-to-one dependency + * CCCR -> CTR : one-to-one dependency + * CTR -> ESCR: none until CCCR is accessed + * CCCR -> ESCR: dynamic dependency (escr_select field) + * ESCR -> ESCR: unrelated until CCCR is accessed + * + * The CCCR -> ESCR dependency is dynamic and cannot be determined + * without inspection of the CCCR.escr_select field. However, nothing useful + * gets recorded unless the ESCR is explicitly programmed. The converse is + * true if only ESCR is accessed and no CCCR, then nothing is measured. + * + * As a consequence, the tables only need to encoded the CTR <-> CCCR + * dependencies. + * + * Because of counter cascading, we have CCCR <-> CCCR dependencies + */ +static struct pfm_reg_desc pfm_p4_pmc_desc[PFM_MAX_PMCS+1]={ +/* pmc0 */ { PFM_REG_W, "BPU_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc1 */ { PFM_REG_W, "IS_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc2 */ { PFM_REG_W, "MOB_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc3 */ { PFM_REG_W, "ITLB_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc4 */ { PFM_REG_W, "PMH_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc5 */ { PFM_REG_W, "IX_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc6 */ { PFM_REG_W, "FSB_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc7 */ { PFM_REG_W, "BSU_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc8 */ { PFM_REG_W, "MS_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc9 */ { PFM_REG_W, "TC_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc10 */ { PFM_REG_W, "TBPU_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc11 */ { PFM_REG_W, "FLAME_ESCR0", 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc12 */ { PFM_REG_W, "FIRM_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc13 */ { PFM_REG_W, "SAAT_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc14 */ { PFM_REG_W, "U2L_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc15 */ { PFM_REG_W, "DAC_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc16 */ { PFM_REG_W, "IQ_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc17 */ { PFM_REG_W, "ALF_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc18 */ { PFM_REG_W, "RAT_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc19 */ { PFM_REG_W, "SSU_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc20 */ { PFM_REG_W, "CRU_ESCR0" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc21 */ { PFM_REG_W, "CRU_ESCR2" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc22 */ { PFM_REG_W, "CRU_ESCR4" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc23 */ { PFM_REG_W, "BPU_CCCR0" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(0)|RDEP(1)}}, +/* pmc24 */ { PFM_REG_W, "BPU_CCCR2" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(0)|RDEP(1)}}, +/* pmc25 */ { PFM_REG_W, "MS_CCCR0" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(2)|RDEP(3)}}, +/* pmc26 */ { PFM_REG_W, "MS_CCCR2" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(2)|RDEP(3)}}, +/* pmc27 */ { PFM_REG_W, "FLAME_CCCR0", PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(4)|RDEP(5)}}, +/* pmc28 */ { PFM_REG_W, "FLAME_CCCR2", PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(4)|RDEP(5)}}, +/* pmc29 */ { PFM_REG_W, "IQ_CCCR0" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(6)|RDEP(7)}}, +/* pmc30 */ { PFM_REG_W, "IQ_CCCR2" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(6)|RDEP(7)}}, +/* pmc31 */ { PFM_REG_W, "IQ_CCCR4" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(8)}}, + /* No HT extension */ +/* pmc32 */ { PFM_REG_W, "BPU_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc33 */ { PFM_REG_W, "IS_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc34 */ { PFM_REG_W, "MOB_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc35 */ { PFM_REG_W, "ITLB_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc36 */ { PFM_REG_W, "PMH_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc37 */ { PFM_REG_W, "IX_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc38 */ { PFM_REG_W, "FSB_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc39 */ { PFM_REG_W, "BSU_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc40 */ { PFM_REG_W, "MS_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc41 */ { PFM_REG_W, "TC_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc42 */ { PFM_REG_W, "TBPU_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc43 */ { PFM_REG_W, "FLAME_ESCR1", 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc44 */ { PFM_REG_W, "FIRM_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc45 */ { PFM_REG_W, "SAAT_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc46 */ { PFM_REG_W, "U2L_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc47 */ { PFM_REG_W, "DAC_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc48 */ { PFM_REG_W, "IQ_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc49 */ { PFM_REG_W, "ALF_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc50 */ { PFM_REG_W, "RAT_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc51 */ { PFM_REG_W, "CRU_ESCR1" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc52 */ { PFM_REG_W, "CRU_ESCR3" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc53 */ { PFM_REG_W, "CRU_ESCR5" , 0x0, PFM_ESCR_PMC_RSVD, {0}}, +/* pmc54 */ { PFM_REG_W, "BPU_CCCR1" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(9)}}, +/* pmc55 */ { PFM_REG_W, "BPU_CCCR3" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(10)}}, +/* pmc56 */ { PFM_REG_W, "MS_CCCR1" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(11)}}, +/* pmc57 */ { PFM_REG_W, "MS_CCCR3" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(12)}}, +/* pmc58 */ { PFM_REG_W, "FLAME_CCCR1", PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(13)}}, +/* pmc59 */ { PFM_REG_W, "FLAME_CCCR3", PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(14)}}, +/* pmc60 */ { PFM_REG_W, "IQ_CCCR1" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(15)}}, +/* pmc61 */ { PFM_REG_W, "IQ_CCCR3" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(16)}}, +/* pmc62 */ { PFM_REG_W, "IQ_CCCR5" , PFM_CCCR_DFL, PFM_CCCR_PMC_RSVD, {RDEP(17)}}, +/* pmc63 */ { PFM_REG_W, "PEBS_MATRIX_VERT", 0, 0x13, {0}}, +/* pmc64 */ { PFM_REG_W, "PEBS_ENABLE", 0, 0x3000fff, {0}}, + { PFM_REG_END, } /* end marker */ +}; + +/* + * Because of counter cascading, there is a dependency between the various + * counters, i.e., PMD <-> PMD + * + * See section 15.10.6.6 for details about the IQ block + */ +static struct pfm_reg_desc pfm_p4_pmd_desc[PFM_MAX_PMDS+1]={ +/* pmd0 */ { PFM_REG_C, "BPU_CTR0" , 0x0, -1, {RDEP(1)}}, +/* pmd1 */ { PFM_REG_C, "BPU_CTR2" , 0x0, -1, {RDEP(0)}}, +/* pmd2 */ { PFM_REG_C, "MS_CTR0" , 0x0, -1, {RDEP(3)}}, +/* pmd3 */ { PFM_REG_C, "MS_CTR2" , 0x0, -1, {RDEP(2)}}, +/* pmd4 */ { PFM_REG_C, "FLAME_CTR0" , 0x0, -1, {RDEP(5)}}, +/* pmd5 */ { PFM_REG_C, "FLAME_CTR2" , 0x0, -1, {RDEP(4)}}, +/* pmd6 */ { PFM_REG_C, "IQ_CTR0" , 0x0, -1, {RDEP(7)}}, +/* pmd7 */ { PFM_REG_C, "IQ_CTR2" , 0x0, -1, {RDEP(8)}}, +/* pmd8 */ { PFM_REG_C, "IQ_CTR4" , 0x0, -1, {0}}, + /* no HT extension */ +/* pmd9 */ { PFM_REG_C, "BPU_CTR1" , 0x0, -1, {RDEP(10)}}, +/* pmd10 */ { PFM_REG_C, "BPU_CTR3" , 0x0, -1, {RDEP(9)}}, +/* pmd11 */ { PFM_REG_C, "MS_CTR1" , 0x0, -1, {RDEP(12)}}, +/* pmd12 */ { PFM_REG_C, "MS_CTR3" , 0x0, -1, {RDEP(11)}}, +/* pmd13 */ { PFM_REG_C, "FLAME_CTR1" , 0x0, -1, {RDEP(14)}}, +/* pmd14 */ { PFM_REG_C, "FLAME_CTR3" , 0x0, -1, {RDEP(13)}}, +/* pmd15 */ { PFM_REG_C, "IQ_CTR1" , 0x0, -1, {RDEP(16)}}, +/* pmd16 */ { PFM_REG_C, "IQ_CTR3" , 0x0, -1, {RDEP(17)}}, +/* pmd17 */ { PFM_REG_C, "IQ_CTR5" , 0x0, -1, {0}}, + { PFM_REG_END, } /* end marker */ +}; + +static int pfm_p4_pmc_check(struct pfm_context *ctx, struct pfm_event_set *set, + u16 cnum, u32 flags, u64 *val) +{ +#define IS_CCCR(c) (pfm_p4_pmu_info.pmc_addrs[(c)].reg_type == PFM_REGT_CCCR) + u64 tmpval, tmp1, tmp2; + u64 rsvd_mask, dfl_value, ival; + + tmpval = ival = *val; + rsvd_mask = pfm_p4_pmc_desc[cnum].reserved_mask; + dfl_value = pfm_p4_pmc_desc[cnum].default_value; + + + /* remove reserved areas from user value */ + tmp1 = tmpval & rsvd_mask; + + /* + * clear OVF_PMI if NO_EMUL64 is set + */ + if (IS_CCCR(cnum) && (flags & PFM_REGFL_NO_EMUL64)) { + if (pfm_p4_pmu_info.lps_per_core > 1) + dfl_value &= ~(3ULL << 26); + else + dfl_value &= ~(1ULL << 26); + } + + /* get reserved fields values */ + tmp2 = dfl_value & ~rsvd_mask; + + *val = tmp1 | tmp2; + + DPRINT(("pmc%u: %s ival=0x%llx rsvd=0x%llx, def=0x%llx, val=0x%llx\n", + cnum, + pfm_p4_pmc_desc[cnum].desc, + (unsigned long long)ival, + (unsigned long long)rsvd_mask, + (unsigned long long)dfl_value, + (unsigned long long)*val)); + + return 0; +} + + +#define MSR_IA32_MISC_ENABLE_PEBS_UNAVAIL (1<<12) /* PEBS unavailable */ +#define cpu_has_dts boot_cpu_has(X86_FEATURE_DTES) + +static int pfm_p4_probe_pmu(void) +{ + int high, low; + + printk(KERN_INFO"perfmon_p4: family=%d x86_model=%d\n", + cpu_data->x86, + cpu_data->x86_model); + /* + * must be family 15 + */ + if (cpu_data->x86 != 15) { + printk(KERN_INFO"perfmon_p4: unsupported family=%d\n", cpu_data->x86); + return -1; + } + + /* + * only works on Intel processors + */ + if (cpu_data->x86_vendor != X86_VENDOR_INTEL) { + printk(KERN_INFO"perfmon_p4: not running on Intel processor\n"); + return -1; + } + + switch(cpu_data->x86_model) { + case 2: + printk(KERN_INFO"perfmon_p4: Northwood P4 detected\n"); + break; + default: + /* + * do not know if they all work the same, so reject for now + */ + printk(KERN_INFO"perfmon_p4: unknown model %d\n", cpu_data->x86_model); + return -1; + } + + /* + * does not work with EM64T processors + */ + if (cpuid_eax(0x80000000) < 0x80000001 + || (cpuid_edx(0x80000001) & (1<<29))) { + printk(KERN_INFO"perfmon_p4: does not support EM64T mode\n"); + return -1; + } + + /* + * check for local APIC (required) + */ + if (!cpu_has_apic) { + printk(KERN_INFO"perfmon_p4: no local APIC, unsupported\n"); + return -1; + } + rdmsr(MSR_IA32_APICBASE, low, high); + if ((low & MSR_IA32_APICBASE_ENABLE) == 0) + printk(KERN_INFO"Local APIC in 3-wire mode\n"); + +#ifdef CONFIG_SMP + pfm_p4_pmu_info.lps_per_core = cpus_weight(cpu_sibling_map[0]); +#endif + + printk(KERN_INFO"perfmon_p4: cores/package=%d threads/core=%d\n", + cpu_data->x86_max_cores, + pfm_p4_pmu_info.lps_per_core); + + if (cpu_has_ht) { + + printk(KERN_INFO"HyperThreading supported: %s\n", + pfm_p4_pmu_info.lps_per_core > 1 ? "on": "off"); + /* + * disable sections not supporting HT + */ + if (pfm_p4_pmu_info.lps_per_core > 1) { + pfm_p4_pmc_desc[PMC_NOHT_IDX].type = PFM_REG_END; + pfm_p4_pmd_desc[PMD_NOHT_IDX].type = PFM_REG_END; + } + } + + if (cpu_has_dts) { + printk("Data Save Area (DS) supported\n"); + pfm_p4_pmu_info.flags = PFM_I386_PMU_DS; + rdmsr(MSR_IA32_MISC_ENABLE, low, high); + if ((low & MSR_IA32_MISC_ENABLE_PEBS_UNAVAIL)==0) { + printk(KERN_INFO"PEBS supported\n"); + pfm_p4_pmu_info.flags |= PFM_I386_PMU_PEBS; + } + } + return 0; +} + +/* + * impl_pmcs, impl_pmds are computed at runtime to minimize errors! + */ +static struct pfm_pmu_config pfm_p4_pmu_conf={ + .pmu_name = "Intel 32-bit P4/Xeon", + .counter_width = 40, + .pmd_desc = pfm_p4_pmd_desc, + .pmc_desc = pfm_p4_pmc_desc, + .pmc_write_check = pfm_p4_pmc_check, + .probe_pmu = pfm_p4_probe_pmu, + .version = "1.0", + .arch_info = &pfm_p4_pmu_info +}; + +static int __init pfm_p4_pmu_init_module(void) +{ + unsigned int i; + + /* + * compute enable bitmask + */ + bitmap_zero(pfm_p4_pmu_info.enable_mask, PFM_MAX_HW_PMCS); + for(i=0; i < PFM_MAX_HW_PMCS; i++) { + if (pfm_p4_pmu_info.pmc_addrs[i].reg_type == PFM_REGT_CCCR) { + pfm_bv_set(pfm_p4_pmu_info.enable_mask, i); + } + } + return pfm_register_pmu_config(&pfm_p4_pmu_conf); +} + +static void __exit pfm_p4_pmu_cleanup_module(void) +{ + pfm_unregister_pmu_config(&pfm_p4_pmu_conf); +} + +module_init(pfm_p4_pmu_init_module); +module_exit(pfm_p4_pmu_cleanup_module); diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon_p4_pebs_smpl.c linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon_p4_pebs_smpl.c --- linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon_p4_pebs_smpl.c 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon_p4_pebs_smpl.c 2005-11-18 05:47:51.000000000 -0800 @@ -0,0 +1,216 @@ +/* + * Copyright (c) 2005 Hewlett-Packard Development Company, L.P. + * Contributed by Stephane Eranian + * + * This file implements the PEBS sampling format for 32-bit + * Pentium 4/Xeon processors. it does not work with Intel EM64T + * processors. + */ +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +#ifndef __i386__ +#error "this module is for the 32-bit Pentium 4 processors" +#endif + +MODULE_AUTHOR("Stephane Eranian "); +MODULE_DESCRIPTION("Intel 32-bit P4/Xeon PEBS sampling"); +MODULE_LICENSE("GPL"); + +static int pfm_pebs_fmt_validate(u32 flags, u16 npmds, void *data) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + pfm_p4_pebs_smpl_arg_t *arg = data; + size_t min_buf_size; + + /* + * host CPU does not have PEBS support + */ + if ((arch_info->flags & PFM_I386_PMU_PEBS) == 0) { + DPRINT(("host PMU does not support PEBS sampling\n")); + return -EINVAL; + } + + /* + * need to define at least the size of the buffer + */ + if (data == NULL) { + DPRINT(("no argument passed\n")); + return -EINVAL; + } + + /* + * compute min buf size. npmds is the maximum number + * of implemented PMD registers. + */ + min_buf_size = sizeof(pfm_p4_pebs_smpl_hdr_t) + sizeof(pfm_p4_pebs_smpl_entry_t); + + DPRINT(("validate flags=0x%x min_buf_size=%zu buf_size=%zu\n", + flags, + min_buf_size, + arg->buf_size)); + + /* + * must hold at least the buffer header + one minimally sized entry + */ + if (arg->buf_size < min_buf_size) + return -EINVAL; + + return 0; +} + +static int pfm_pebs_fmt_get_size(unsigned int flags, void *data, size_t *size) +{ + pfm_p4_pebs_smpl_arg_t *arg = data; + + /* + * size has been validated in pebs_fmt_validate() + */ + *size = arg->buf_size + 256; + + return 0; +} + +static int pfm_pebs_fmt_init(struct pfm_context *ctx, void *buf, + u32 flags, u16 npmds, void *data) +{ + struct pfm_arch_context *ctx_arch; + pfm_p4_pebs_smpl_hdr_t *hdr; + pfm_p4_pebs_smpl_arg_t *arg = data; + unsigned long base; + pfm_p4_ds_area_t *ds; + + ctx_arch = pfm_ctx_arch(ctx); + + hdr = buf; + ds = &hdr->hdr_ds; + + hdr->hdr_version = PFM_P4_PEBS_SMPL_VERSION; + hdr->hdr_buf_size = arg->buf_size; + hdr->hdr_overflows = 0; + + /* + * align base + */ + base = ((unsigned long)(hdr+1) + 256) & ~0xffUL; + hdr->hdr_start_offs = base - (unsigned long)(hdr+1); + ds->pebs_buf_base = base; + ds->pebs_abs_max = base + arg->buf_size + 1; + ds->pebs_intr_thres = base + arg->intr_thres*sizeof(pfm_p4_pebs_smpl_entry_t); + ds->pebs_index = base; + + /* + * save counter reset value for IQ_CCCR4 (thread0) or IQ_CCCR5 (thread1) + */ + ds->pebs_cnt_reset = arg->cnt_reset; + + /* + * ensure proper PEBS handling in arch/.../perfmon/perfmon.c + */ + ctx_arch->flags |= PFM_I386_FL_USE_PEBS; + ctx_arch->ds_area = ds; + + DPRINT(("buffer=%p buf_size=%zu ctx_flags=0x%x" + "pebs_base=0x%x pebs_max=0x%x pebs_thres=0x%x cnt_reset=0x%llx\n", + buf, + hdr->hdr_buf_size, + ctx_arch->flags, + ds->pebs_buf_base, + ds->pebs_abs_max, + ds->pebs_intr_thres, + (unsigned long long)ds->pebs_cnt_reset)); + + return 0; +} + +static int pfm_pebs_fmt_handler(void *buf, struct pfm_ovfl_arg *arg, + unsigned long ip, u64 tstamp, void *data) +{ + pfm_p4_pebs_smpl_hdr_t *hdr; + + hdr = buf; + + DPRINT_ovfl(("buffer full\n")); + /* + * increment number of buffer overflows. + * important to detect duplicate set of samples. + */ + hdr->hdr_overflows++; + + /* + * request notification and masking of monitoring. + * Notification is still subject to the overflowed + * register having the FL_NOTIFY flag set. + */ + arg->ovfl_ctrl = PFM_OVFL_CTRL_NOTIFY| PFM_OVFL_CTRL_MASK; + + return -ENOBUFS; /* we are full, sorry */ +} + +static int pfm_pebs_fmt_restart(int is_active, u32 *ovfl_ctrl, void *buf) +{ + pfm_p4_pebs_smpl_hdr_t *hdr = buf; + + /* + * reset index to base of buffer + */ + hdr->hdr_ds.pebs_index = hdr->hdr_ds.pebs_buf_base; + + *ovfl_ctrl = PFM_OVFL_CTRL_RESET; + + return 0; +} + +static int pfm_pebs_fmt_exit(void *buf) +{ + return 0; +} + +static struct pfm_smpl_fmt pebs_fmt={ + .fmt_name = "32-bit P4/Xeon PEBS", + .fmt_uuid = PFM_P4_PEBS_SMPL_UUID, + .fmt_arg_size = sizeof(pfm_p4_pebs_smpl_arg_t), + .fmt_validate = pfm_pebs_fmt_validate, + .fmt_getsize = pfm_pebs_fmt_get_size, + .fmt_init = pfm_pebs_fmt_init, + .fmt_handler = pfm_pebs_fmt_handler, + .fmt_restart = pfm_pebs_fmt_restart, + .fmt_exit = pfm_pebs_fmt_exit, +}; + +#define MSR_IA32_MISC_ENABLE_PEBS_UNAVAIL (1<<12) /* PEBS unavailable */ +#define cpu_has_dts boot_cpu_has(X86_FEATURE_DTES) + +static int __init pfm_pebs_fmt_init_module(void) +{ + int low, high; + + if (!cpu_has_dts) { + printk(KERN_INFO"Processor does not have Data Save Area (DS)\n"); + return -1; + } + rdmsr(MSR_IA32_MISC_ENABLE, low, high); + + if (low & MSR_IA32_MISC_ENABLE_PEBS_UNAVAIL) { + printk(KERN_INFO"Processor does not support PEBS\n"); + return -1; + } + return pfm_register_smpl_fmt(&pebs_fmt); +} + +static void __exit pfm_pebs_fmt_cleanup_module(void) +{ + pfm_unregister_smpl_fmt(pebs_fmt.fmt_uuid); +} + +module_init(pfm_pebs_fmt_init_module); +module_exit(pfm_pebs_fmt_cleanup_module); diff -urN linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon_p6.c linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon_p6.c --- linux-2.6.15-rc1-git6.orig/arch/i386/perfmon/perfmon_p6.c 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/arch/i386/perfmon/perfmon_p6.c 2005-11-18 05:47:51.000000000 -0800 @@ -0,0 +1,176 @@ +/* + * This file contains the P6 family processor PMU register description tables + * and pmc checker used by perfmon.c. + * + * Copyright (c) 2005 Hewlett-Packard Development Company, L.P. + * Contributed by Stephane Eranian + */ +#include +#include +#include +#include + +MODULE_AUTHOR("Stephane Eranian "); +MODULE_DESCRIPTION("P6/Pentium M description tables"); +MODULE_LICENSE("GPL"); + +/* + * - upper 32 bits are reserved + * - INT: APIC enable bit is reserved (forced to 1) + * - bit 21 is reserved + */ +#define PFM_P6_PMC_RSVD ~((0xffffffffULL<<32) \ + | (PFM_ONE_64<<20) \ + | (PFM_ONE_64<<21)) + +/* + * force Local APIC interrupt on overflow + */ +#define PFM_P6_PMC_VAL (PFM_ONE_64<<20) + +/* + * physical addresses of MSR control the perfself and perfctr registers + */ +struct pfm_arch_pmu_info pfm_p6_pmu_info={ + .pmc_addrs = { + {{MSR_P6_EVNTSEL0, 0}, 0, PFM_REGT_PERFSEL}, + {{MSR_P6_EVNTSEL1, 0}, 1, PFM_REGT_PERFSEL} + }, + .pmd_addrs = { + {{MSR_P6_PERFCTR0, 0}, 0, PFM_REGT_CTR}, + {{MSR_P6_PERFCTR1, 0}, 0, PFM_REGT_CTR} + }, + .pmu_style = PFM_I386_PMU_P6, + .lps_per_core = 1 +}; + +static struct pfm_reg_desc pfm_p6_pmc_desc[PFM_MAX_PMCS+1]={ +/* pmc0 */ { PFM_REG_W, "PERFSEL0", PFM_P6_PMC_VAL, PFM_P6_PMC_RSVD, {RDEP(0)}}, +/* pmc1 */ { PFM_REG_W, "PERFSEL1", PFM_P6_PMC_VAL, PFM_P6_PMC_RSVD, {RDEP(1)}}, + { PFM_REG_END, } /* end marker */ +}; + +static struct pfm_reg_desc pfm_p6_pmd_desc[PFM_MAX_PMDS+1]={ +/* pmd0 */ { PFM_REG_C , "PERFCTR0", 0x0, -1, {0}}, +/* pmd1 */ { PFM_REG_C , "PERFCTR1", 0x0, -1, {0}}, + { PFM_REG_END, } /* end marker */ +}; + +static int pfm_p6_pmc_check(struct pfm_context *ctx, struct pfm_event_set *set, + u16 cnum, u32 flags, u64 *val) +{ + u64 tmpval, tmp1, tmp2; + u64 rsvd_mask, dfl_value; + + tmpval = *val; + rsvd_mask = pfm_p6_pmc_desc[cnum].reserved_mask; + dfl_value = pfm_p6_pmc_desc[cnum].default_value; + + /* remove reserved areas from user value */ + tmp1 = tmpval & rsvd_mask; + + /* get reserved fields values */ + tmp2 = dfl_value & ~rsvd_mask; + *val = tmp1 | tmp2; + + DPRINT(("pmc%u: %s rsvd=0x%llx, def=0x%llx, oval=0x%llx\n", + cnum, + pfm_p6_pmc_desc[cnum].desc, + (unsigned long long)rsvd_mask, + (unsigned long long)dfl_value, + (unsigned long long)*val)); + + return 0; +} + +#define MSR_IA32_MISC_ENABLE_PERF_AVAIL (1<<7) /* read-only status bit */ + +static int pfm_p6_probe_pmu(void) +{ + int high, low; + + printk(KERN_INFO"perfmon_p6: family=%d x86_model=%d\n", + cpu_data->x86, + cpu_data->x86_model); + /* + * not from P6 processor family + */ + if (cpu_data->x86 != 6 && cpu_data->x86 != 15) { + printk(KERN_INFO"perfmon_p6: unsupported family=%d\n", cpu_data->x86); + return -1; + } + + switch(cpu_data->x86_model) { + case 8: + printk(KERN_INFO"P6 core PMU detected\n"); + break; + case 9: + case 13: + rdmsr(MSR_IA32_MISC_ENABLE, low, high); + if ((low & MSR_IA32_MISC_ENABLE_PERF_AVAIL) == 0) { + printk(KERN_INFO"Pentium M without PMU\n"); + return -1; + } + printk(KERN_INFO"Pentium M PMU detected\n"); + break; + default: + printk(KERN_INFO"unsupported CPU model %d\n", + cpu_data->x86_model); + return -1; + + } + + if (!cpu_has_apic) { + printk(KERN_INFO"perfmon_p6: no Local APIC, unsupported\n"); + return -1; + } + + rdmsr(MSR_IA32_APICBASE, low, high); + if ((low & MSR_IA32_APICBASE_ENABLE) == 0) { + printk(KERN_INFO"perfmon_p6: local APIC disabled, you must enable with lapic kernel command line option\n"); + return -1; + } + return 0; +} + +/* + * Counters have 40 bits implemented. However they are designed such + * that bits [32-39] are sign extensions of bit 31. As such the + * effective width of a counter for P6-like PMU is 31 bits only. + * + * See IA-32 Intel Architecture Software developer manual Vol 3: + * system programming and section 15.14.2 in particular. + */ +static struct pfm_pmu_config pfm_p6_pmu_conf={ + .pmu_name = "Intel P6 Family Processor", + .counter_width = 31, + .pmd_desc = pfm_p6_pmd_desc, + .pmc_desc = pfm_p6_pmc_desc, + .pmc_write_check = pfm_p6_pmc_check, + .probe_pmu = pfm_p6_probe_pmu, + .version = "1.0", + .arch_info = &pfm_p6_pmu_info +}; + +static int __init pfm_p6_pmu_init_module(void) +{ + unsigned int i; + /* + * XXX: could be hardcoded for this PMU model + */ + bitmap_zero(pfm_p6_pmu_info.enable_mask, PFM_MAX_HW_PMCS); + for(i=0; i < PFM_MAX_HW_PMCS; i++) { + if (pfm_p6_pmu_info.pmc_addrs[i].reg_type == PFM_REGT_PERFSEL) { + pfm_bv_set(pfm_p6_pmu_info.enable_mask, i); + } + } + return pfm_register_pmu_config(&pfm_p6_pmu_conf); +} + +static void __exit pfm_p6_pmu_cleanup_module(void) +{ + pfm_unregister_pmu_config(&pfm_p6_pmu_conf); +} + +module_init(pfm_p6_pmu_init_module); +module_exit(pfm_p6_pmu_cleanup_module); diff -urN linux-2.6.15-rc1-git6.orig/include/asm-i386/mach-default/irq_vectors.h linux-2.6.15-rc1-git6/include/asm-i386/mach-default/irq_vectors.h --- linux-2.6.15-rc1-git6.orig/include/asm-i386/mach-default/irq_vectors.h 2005-11-18 05:16:49.000000000 -0800 +++ linux-2.6.15-rc1-git6/include/asm-i386/mach-default/irq_vectors.h 2005-11-18 05:47:51.000000000 -0800 @@ -56,6 +56,7 @@ * sources per level' errata. */ #define LOCAL_TIMER_VECTOR 0xef +#define LOCAL_PERFMON_VECTOR 0xee /* * First APIC vector available to drivers: (vectors 0x30-0xee) @@ -63,7 +64,7 @@ * levels. (0x80 is the syscall vector) */ #define FIRST_DEVICE_VECTOR 0x31 -#define FIRST_SYSTEM_VECTOR 0xef +#define FIRST_SYSTEM_VECTOR 0xee #define TIMER_IRQ 0 diff -urN linux-2.6.15-rc1-git6.orig/include/asm-i386/perfmon.h linux-2.6.15-rc1-git6/include/asm-i386/perfmon.h --- linux-2.6.15-rc1-git6.orig/include/asm-i386/perfmon.h 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/include/asm-i386/perfmon.h 2005-11-18 05:47:51.000000000 -0800 @@ -0,0 +1,256 @@ +/* + * Copyright (c) 2005 Hewlett-Packard Development Company, L.P. + * Contributed by Stephane Eranian + * + * This file contains i386 Processor Family specific definitions + * for the perfmon interface. + * + * This file MUST never be included directly. Use linux/perfmon.h. + */ +#ifndef _ASM_I386_PERFMON_H_ +#define _ASM_I386_PERFMON_H_ + +/* + * I386 specific context flags + */ +#define PFM_I386_FL_INSECURE 0x10000 /* allow rdpmc at user level */ + +#ifdef __KERNEL__ + +#include +#include + +/* + * - bits 31 - 63 reserved + * - T1_OS and T1_USR bits are reserved - set depending on logical proc + * user mode application should use T0_OS and T0_USR to indicate + */ +#define PFM_ESCR_PMC_RSVD 0x000000007ffffffc + +#define PFM_REGT_PERFSEL 1 /* cannot be zero due to ext_reg init */ +#define PFM_REGT_ESCR 2 +#define PFM_REGT_CCCR 3 /* contains the enable bit */ +#define PFM_REGT_CTR 4 /* is a counter */ +#define PFM_REGT_LBR 5 +#define PFM_REGT_LBRTOS 6 +#define PFM_REGT_DEBUG 7 + +/* + * This design and the partitioning of resources for SMT (hyper threads) + * is very static and limited due to limitations in the number of ESCRs + * and CCCRs per group. + */ +#define MAX_SMT_ID 1 +/* + * For extended register information in addition to address that is used + * at runtime to figure out the mapping of reg addresses to logical procs + * and association of registers to hardware specific features + */ +struct pfm_arch_ext_reg { + /* + * one each for the logical CPUs. Index 0 corresponds to T0 and + * index 1 corresponds to T1. Index 1 can be zero if no T1 + * complement reg exists. + */ + unsigned long addrs[MAX_SMT_ID+1]; + int ctr; /* for CCCR/PERFSEL, associated counter */ + int reg_type; +}; + +struct pfm_arch_pmu_info { + struct pfm_arch_ext_reg pmc_addrs[PFM_MAX_HW_PMCS]; + struct pfm_arch_ext_reg pmd_addrs[PFM_MAX_HW_PMDS]; + unsigned long enable_mask[PFM_PMC_BV]; /* PMC registers with enable bit */ + + u16 lps_per_core; /* # of logical processors per core */ + u8 pmu_style; /* type of PMU interface (P4, P6) */ + u16 pebs_ctr_idx; /* index of PEBS IQ_CTR4 counter (for overflow) */ + u16 flags; /* PMU feature flags */ +}; +/* + * I386 PMU style + */ + +#define PFM_I386_PMU_P4 1 /* Intel: P4/Xeon processor PMU */ +#define PFM_I386_PMU_P6 2 /* Intel: P6 processor PMU */ + +/* + * PMU feature flags + */ +#define PFM_I386_PMU_DS 0x1 /* Intel: support for Data Save Area (DS) */ +#define PFM_I386_PMU_PEBS 0x2 /* Intel: support PEBS (implies DS) */ + +extern void __pfm_read_reg(const struct pfm_arch_ext_reg *xreg, u64 *val); +extern void __pfm_write_reg(const struct pfm_arch_ext_reg *xreg, u64 val); + +static inline void pfm_arch_resend_irq(void) +{ + unsigned long val, dest; + /* + * we cannot use hw_resend_irq() because it goes to + * the I/O APIC. We need to go to the Local Apic. + * + * The "int vec" is not the right solution either + * because it triggers a software intr. We need + * to regenerate the intr. and have it pended until + * we unmask interrupts. + * + * Instead we send ourself an IPI on the perfmon + * vector. + */ + val = APIC_DEST_SELF|APIC_INT_ASSERT| + APIC_DM_FIXED|LOCAL_PERFMON_VECTOR; + dest = apic_read(APIC_ID); + apic_write(APIC_ICR2, dest); + apic_write(APIC_ICR, val); + +} + +#define pfm_arch_serialize() /* nothing */ + +static inline u64 pfm_arch_get_itc(void) +{ + u64 tmp; + rdtscll(tmp); + return tmp; +} + +static inline void pfm_arch_write_pmc(struct pfm_context *ctx, unsigned int cnum, u64 value) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + /* + * we only write to the actual register when monitoring is + * active (pfm_start was issued) + */ + if (ctx && ctx->ctx_fl_started == 0) return; + + __pfm_write_reg(&arch_info->pmc_addrs[cnum], value); +} + +static inline void pfm_arch_write_pmd(struct pfm_context *ctx, unsigned int cnum, u64 value) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + + /* + * force upper bit set for counter to ensure overflow + */ + if (arch_info->pmd_addrs[cnum].reg_type == PFM_REGT_CTR) + value |=~pfm_pmu_conf->ovfl_mask; + + __pfm_write_reg(&arch_info->pmd_addrs[cnum], value); +} + +static inline u64 pfm_arch_read_pmd(struct pfm_context *ctx, unsigned int cnum) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + u64 tmp; + __pfm_read_reg(&arch_info->pmd_addrs[cnum], &tmp); + return tmp; +} + +static inline u64 pfm_arch_read_pmc(struct pfm_context *ctx, unsigned int cnum) +{ + struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; + u64 tmp; + __pfm_read_reg(&arch_info->pmc_addrs[cnum], &tmp); + return tmp; +} +/* + * At certain points, perfmon needs to know if monitoring has been + * explicitely started/stopped by user via pfm_start/pfm_stop. The + * information is tracked in ctx_fl_started. However on certain + * architectures, it may be possible to start/stop directly from + * user level with a single assembly instruction bypassing + * the kernel. This function must be used to determine by + * an arch-specific mean if monitoring is actually started/stopped. + * If there is no other way but to go through pfm_start/pfm_stop + * then this function can simply return 0 + */ +static inline int pfm_arch_is_active(struct pfm_context *ctx) +{ + return 0; +} + + +extern void pfm_arch_init_percpu(void); +extern void pfm_arch_ctxswout(struct task_struct *task, + struct pfm_context *ctx, struct pfm_event_set *set); +extern void pfm_arch_ctxswin(struct task_struct *task, + struct pfm_context *ctx, struct pfm_event_set *set); +extern void pfm_arch_stop(struct task_struct *task, + struct pfm_context *ctx, struct pfm_event_set *set); +extern void pfm_arch_start(struct task_struct *task, + struct pfm_context *ctx, struct pfm_event_set *set); +extern void pfm_arch_restore_pmds(struct pfm_context *ctx, struct pfm_event_set *set); +extern void pfm_arch_save_pmds(struct pfm_context *ctx, struct pfm_event_set *set); +extern void pfm_arch_restore_pmcs(struct pfm_context *ctx, struct pfm_event_set *set); +extern void pfm_arch_intr_freeze_pmu(struct pfm_context *ctx); +extern void pfm_arch_intr_unfreeze_pmu(struct pfm_context *ctx); +extern int pfm_arch_pmu_config_check(struct pfm_pmu_config *cfg); +extern void pfm_arch_pmu_config_init(struct pfm_pmu_config *cfg); +extern int pfm_arch_initialize(void); +extern void pfm_arch_mask_monitoring(struct pfm_context *ctx); +extern void pfm_arch_unmask_monitoring(struct pfm_context *ctx); +extern void pfm_arch_context_initialize(struct pfm_context *ctx, u32 ctx_flags); +extern void pfm_arch_unload_context(struct pfm_context *ctx, + struct task_struct *task); +extern int pfm_arch_load_context(struct pfm_context *ctx, + struct task_struct *task); + +static inline int pfm_arch_reserve_session(struct pfm_context *ctx, + unsigned int cpu) +{ + return 0; +} + +//static inline void pfm_arch_unreserve_session(struct pfm_context *ctx, unsigned int cpu) +#define pfm_arch_unreserve_session(ctx, cpu) /* nothing */ + +/* + * function called from pfm_setfl_sane(). Context is locked + * and interrupts are masked. + * The value of flags is the value of ctx_flags as passed by + * user. + * + * function must check arch-specific set flags. + * Return: + * 1 when flags are valid + * 0 on error + */ +static inline int pfm_arch_setfl_sane(struct pfm_context *ctx, u32 flags) +{ + return 0; +} + +//static inline void pfm_arch_show_session(struct seq_file *m); +#define pfm_arch_show_session(m) /* nothing */ + +/* + * on P6, P4, the upper bits of a counter must be set in order + * for the overflow interrupt to happen. On overflow, the counter + * has wrapped around, and the upper bits are now cleared. This + * function set them back. + * + * The current version loses whatever is remaining in the counter, + * which is usually not zero but has a small count. In order not + * to loose this count, we do a read-modify-write to set the upper + * bits while preserving the low-order bits. This is slow but + * works. + */ +static inline void pfm_arch_ovfl_reset_pmd(struct pfm_context *ctx, unsigned int cnum) +{ + u64 val; + val = pfm_arch_read_pmd(ctx, cnum); + pfm_arch_write_pmd(ctx, cnum, val); +} + +#define PFM_I386_FL_USE_PEBS 0x20000 /* Intel: context uses PEBS */ +struct pfm_arch_context { + void *ds_area; /* pointer to DS management area */ + u32 flags; /* arch-specific flags */ +}; + +#define PFM_ARCH_CTX_SIZE (sizeof(struct pfm_arch_context)) + +#endif /* __KERNEL__ */ +#endif /* _ASM_I386_PERFMON_H_ */ diff -urN linux-2.6.15-rc1-git6.orig/include/asm-i386/perfmon_p4_pebs_smpl.h linux-2.6.15-rc1-git6/include/asm-i386/perfmon_p4_pebs_smpl.h --- linux-2.6.15-rc1-git6.orig/include/asm-i386/perfmon_p4_pebs_smpl.h 1969-12-31 16:00:00.000000000 -0800 +++ linux-2.6.15-rc1-git6/include/asm-i386/perfmon_p4_pebs_smpl.h 2005-11-18 05:47:51.000000000 -0800 @@ -0,0 +1,96 @@ +/* + * Copyright (c) 2005 Hewlett-Packard Development Company, L.P. + * Contributed by Stephane Eranian + * + * This file implements the PEBS sampling format for 32-bit + * Intel Pentium 4/Xeon processors. Not to be used with Intel EM64T + * processors. + */ +#ifndef __PERFMON_P4_PEBS_SMPL_H__ +#define __PERFMON_P4_PEBS_SMPL_H__ 1 + +#define PFM_P4_PEBS_SMPL_UUID { \ + 0x0d, 0x85, 0x91, 0xe7, 0x49, 0x3f, 0x49, 0xae,\ + 0x8c, 0xfc, 0xe8, 0xb9, 0x33, 0xe4, 0xeb, 0x8b} + +/* + * format specific parameters (passed at context creation) + * + * intr_thres: index from start of buffer of entry where the + * PMU interrupt must be triggered. It must be several samples + * short of the end of the buffer. + */ +typedef struct { + size_t buf_size; /* size of the buffer in bytes */ + size_t intr_thres; /* index of interrupt threshold entry */ + u32 flags; /* buffer specific flags */ + u64 cnt_reset; /* counter reset value */ + u32 res1; /* for future use */ + u64 reserved[2]; /* for future use */ +} pfm_p4_pebs_smpl_arg_t; + +/* + * combined context+format specific structure. Can be passed + * to pfm_context_create() + */ +typedef struct { + pfarg_ctx_t ctx_arg; + pfm_p4_pebs_smpl_arg_t buf_arg; +} pfm_p4_pebs_smpl_ctx_arg_t; + +/* + * DS save area as described in section 15.10.5 + */ +typedef struct { + u32 bts_buf_base; + u32 bts_index; + u32 bts_abs_max; + u32 bts_intr_thres; + u32 pebs_buf_base; + u32 pebs_index; + u32 pebs_abs_max; + u32 pebs_intr_thres; + u64 pebs_cnt_reset; +} pfm_p4_ds_area_t; + +/* + * This header is at the beginning of the sampling buffer returned to the user. + * + * Because of PEBS alignement constraints, the actual PEBS buffer area does + * not necessarily begin right after the header. The hdr_start_offs must be + * used to compute the first byte of the buffer. The offset is defined as + * the number of bytes between the end of the header and the beginning of + * the buffer. As such the formula is: + * actual_buffer = (unsigned long)(hdr+1)+hdr->hdr_start_offs + */ +typedef struct { + u64 hdr_overflows; /* #overflows for buffer */ + size_t hdr_buf_size; /* bytes in the buffer */ + size_t hdr_start_offs; /* actual buffer start offset */ + u32 hdr_version; /* smpl format version */ + u64 hdr_res[3]; /* for future use */ + pfm_p4_ds_area_t hdr_ds; /* DS management Area */ +} pfm_p4_pebs_smpl_hdr_t; + +/* + * PEBS record format as described in section 15.10.6 + */ +typedef struct { + u32 eflags; + u32 ip; + u32 eax; + u32 ebx; + u32 ecx; + u32 edx; + u32 esi; + u32 edi; + u32 ebp; + u32 esp; +} pfm_p4_pebs_smpl_entry_t; + +#define PFM_P4_PEBS_SMPL_VERSION_MAJ 1U +#define PFM_P4_PEBS_SMPL_VERSION_MIN 0U +#define PFM_P4_PEBS_SMPL_VERSION (((PFM_P4_PEBS_SMPL_VERSION_MAJ&0xffff)<<16)|\ + (PFM_P4_PEBS_SMPL_VERSION_MIN & 0xffff)) + +#endif /* __PERFMON_P4_PEBS_SMPL_H__ */ diff -urN linux-2.6.15-rc1-git6.orig/include/asm-i386/processor.h linux-2.6.15-rc1-git6/include/asm-i386/processor.h --- linux-2.6.15-rc1-git6.orig/include/asm-i386/processor.h 2005-11-18 05:16:49.000000000 -0800 +++ linux-2.6.15-rc1-git6/include/asm-i386/processor.h 2005-11-18 05:47:51.000000000 -0800 @@ -423,6 +423,7 @@ */ unsigned long io_bitmap_max; struct thread_struct *io_bitmap_owner; + /* * pads the TSS to be cacheline-aligned (size is 0x100) */ @@ -460,6 +461,7 @@ unsigned long iopl; /* max allowed port in the bitmap, in bytes: */ unsigned long io_bitmap_max; + void *pfm_context; }; #define INIT_THREAD { \ diff -urN linux-2.6.15-rc1-git6.orig/include/asm-i386/system.h linux-2.6.15-rc1-git6/include/asm-i386/system.h --- linux-2.6.15-rc1-git6.orig/include/asm-i386/system.h 2005-11-18 05:16:49.000000000 -0800 +++ linux-2.6.15-rc1-git6/include/asm-i386/system.h 2005-11-18 05:47:51.000000000 -0800 @@ -14,6 +14,7 @@ #define switch_to(prev,next,last) do { \ unsigned long esi,edi; \ + pfm_ctxswout(prev); \ asm volatile("pushl %%ebp\n\t" \ "movl %%esp,%0\n\t" /* save ESP */ \ "movl %5,%%esp\n\t" /* restore ESP */ \ diff -urN linux-2.6.15-rc1-git6.orig/include/asm-i386/unistd.h linux-2.6.15-rc1-git6/include/asm-i386/unistd.h --- linux-2.6.15-rc1-git6.orig/include/asm-i386/unistd.h 2005-11-18 05:16:49.000000000 -0800 +++ linux-2.6.15-rc1-git6/include/asm-i386/unistd.h 2005-11-18 05:47:51.000000000 -0800 @@ -299,8 +299,20 @@ #define __NR_inotify_init 291 #define __NR_inotify_add_watch 292 #define __NR_inotify_rm_watch 293 - -#define NR_syscalls 294 +#define __NR_pfm_create_context 294 +#define __NR_pfm_write_pmcs (__NR_pfm_create_context+1) +#define __NR_pfm_write_pmds (__NR_pfm_create_context+2) +#define __NR_pfm_read_pmds (__NR_pfm_create_context+3) +#define __NR_pfm_load_context (__NR_pfm_create_context+4) +#define __NR_pfm_start (__NR_pfm_create_context+5) +#define __NR_pfm_stop (__NR_pfm_create_context+6) +#define __NR_pfm_restart (__NR_pfm_create_context+7) +#define __NR_pfm_create_evtsets (__NR_pfm_create_context+8) +#define __NR_pfm_getinfo_evtsets (__NR_pfm_create_context+9) +#define __NR_pfm_delete_evtsets (__NR_pfm_create_context+10) +#define __NR_pfm_unload_context (__NR_pfm_create_context+11) + +#define NR_syscalls 305 /* * user-visible error numbers are in the range -1 - -128: see --ABTtc+pdwF7KHXCz--