diff options
Diffstat (limited to 'arch/x86/kernel/cpu')
26 files changed, 1113 insertions, 1211 deletions
diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile index c1f253dac155..68537e957a9b 100644 --- a/arch/x86/kernel/cpu/Makefile +++ b/arch/x86/kernel/cpu/Makefile @@ -13,7 +13,7 @@ CFLAGS_common.o := $(nostackp) obj-y := intel_cacheinfo.o addon_cpuid_features.o obj-y += proc.o capflags.o powerflags.o common.o -obj-y += vmware.o hypervisor.o +obj-y += vmware.o hypervisor.o sched.o obj-$(CONFIG_X86_32) += bugs.o cmpxchg.o obj-$(CONFIG_X86_64) += bugs_64.o @@ -27,7 +27,7 @@ obj-$(CONFIG_CPU_SUP_CENTAUR) += centaur.o obj-$(CONFIG_CPU_SUP_TRANSMETA_32) += transmeta.o obj-$(CONFIG_CPU_SUP_UMC_32) += umc.o -obj-$(CONFIG_PERF_COUNTERS) += perf_counter.o +obj-$(CONFIG_PERF_EVENTS) += perf_event.o obj-$(CONFIG_X86_MCE) += mcheck/ obj-$(CONFIG_MTRR) += mtrr/ diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index 22a47c82f3c0..c910a716a71c 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -184,7 +184,7 @@ static void __cpuinit amd_k7_smp_check(struct cpuinfo_x86 *c) * approved Athlon */ WARN_ONCE(1, "WARNING: This combination of AMD" - "processors is not suitable for SMP.\n"); + " processors is not suitable for SMP.\n"); if (!test_taint(TAINT_UNSAFE_SMP)) add_taint(TAINT_UNSAFE_SMP); @@ -333,6 +333,16 @@ static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c) #endif } +int amd_get_nb_id(int cpu) +{ + int id = 0; +#ifdef CONFIG_SMP + id = per_cpu(cpu_llc_id, cpu); +#endif + return id; +} +EXPORT_SYMBOL_GPL(amd_get_nb_id); + static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) { #if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 2055fc2b2e6b..cc25c2b4a567 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -13,7 +13,7 @@ #include <linux/io.h> #include <asm/stackprotector.h> -#include <asm/perf_counter.h> +#include <asm/perf_event.h> #include <asm/mmu_context.h> #include <asm/hypervisor.h> #include <asm/processor.h> @@ -34,7 +34,6 @@ #include <asm/mce.h> #include <asm/msr.h> #include <asm/pat.h> -#include <linux/smp.h> #ifdef CONFIG_X86_LOCAL_APIC #include <asm/uv/uv.h> @@ -870,7 +869,7 @@ void __init identify_boot_cpu(void) #else vgetcpu_set_mode(); #endif - init_hw_perf_counters(); + init_hw_perf_events(); } void __cpuinit identify_secondary_cpu(struct cpuinfo_x86 *c) diff --git a/arch/x86/kernel/cpu/cpu_debug.c b/arch/x86/kernel/cpu/cpu_debug.c index 6b2a52dd0403..dca325c03999 100644 --- a/arch/x86/kernel/cpu/cpu_debug.c +++ b/arch/x86/kernel/cpu/cpu_debug.c @@ -30,8 +30,8 @@ #include <asm/apic.h> #include <asm/desc.h> -static DEFINE_PER_CPU(struct cpu_cpuX_base, cpu_arr[CPU_REG_ALL_BIT]); -static DEFINE_PER_CPU(struct cpu_private *, priv_arr[MAX_CPU_FILES]); +static DEFINE_PER_CPU(struct cpu_cpuX_base [CPU_REG_ALL_BIT], cpu_arr); +static DEFINE_PER_CPU(struct cpu_private * [MAX_CPU_FILES], priv_arr); static DEFINE_PER_CPU(int, cpu_priv_count); static DEFINE_MUTEX(cpu_debug_lock); diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c index ae9b503220ca..7d5c3b0ea8da 100644 --- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c +++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c @@ -33,7 +33,7 @@ #include <linux/cpufreq.h> #include <linux/compiler.h> #include <linux/dmi.h> -#include <trace/power.h> +#include <trace/events/power.h> #include <linux/acpi.h> #include <linux/io.h> @@ -60,7 +60,6 @@ enum { }; #define INTEL_MSR_RANGE (0xffff) -#define CPUID_6_ECX_APERFMPERF_CAPABILITY (0x1) struct acpi_cpufreq_data { struct acpi_processor_performance *acpi_data; @@ -71,13 +70,7 @@ struct acpi_cpufreq_data { static DEFINE_PER_CPU(struct acpi_cpufreq_data *, drv_data); -struct acpi_msr_data { - u64 saved_aperf, saved_mperf; -}; - -static DEFINE_PER_CPU(struct acpi_msr_data, msr_data); - -DEFINE_TRACE(power_mark); +static DEFINE_PER_CPU(struct aperfmperf, old_perf); /* acpi_perf_data is a pointer to percpu data. */ static struct acpi_processor_performance *acpi_perf_data; @@ -244,23 +237,12 @@ static u32 get_cur_val(const struct cpumask *mask) return cmd.val; } -struct perf_pair { - union { - struct { - u32 lo; - u32 hi; - } split; - u64 whole; - } aperf, mperf; -}; - /* Called via smp_call_function_single(), on the target CPU */ static void read_measured_perf_ctrs(void *_cur) { - struct perf_pair *cur = _cur; + struct aperfmperf *am = _cur; - rdmsr(MSR_IA32_APERF, cur->aperf.split.lo, cur->aperf.split.hi); - rdmsr(MSR_IA32_MPERF, cur->mperf.split.lo, cur->mperf.split.hi); + get_aperfmperf(am); } /* @@ -279,63 +261,17 @@ static void read_measured_perf_ctrs(void *_cur) static unsigned int get_measured_perf(struct cpufreq_policy *policy, unsigned int cpu) { - struct perf_pair readin, cur; - unsigned int perf_percent; + struct aperfmperf perf; + unsigned long ratio; unsigned int retval; - if (smp_call_function_single(cpu, read_measured_perf_ctrs, &readin, 1)) + if (smp_call_function_single(cpu, read_measured_perf_ctrs, &perf, 1)) return 0; - cur.aperf.whole = readin.aperf.whole - - per_cpu(msr_data, cpu).saved_aperf; - cur.mperf.whole = readin.mperf.whole - - per_cpu(msr_data, cpu).saved_mperf; - per_cpu(msr_data, cpu).saved_aperf = readin.aperf.whole; - per_cpu(msr_data, cpu).saved_mperf = readin.mperf.whole; - -#ifdef __i386__ - /* - * We dont want to do 64 bit divide with 32 bit kernel - * Get an approximate value. Return failure in case we cannot get - * an approximate value. - */ - if (unlikely(cur.aperf.split.hi || cur.mperf.split.hi)) { - int shift_count; - u32 h; - - h = max_t(u32, cur.aperf.split.hi, cur.mperf.split.hi); - shift_count = fls(h); - - cur.aperf.whole >>= shift_count; - cur.mperf.whole >>= shift_count; - } - - if (((unsigned long)(-1) / 100) < cur.aperf.split.lo) { - int shift_count = 7; - cur.aperf.split.lo >>= shift_count; - cur.mperf.split.lo >>= shift_count; - } - - if (cur.aperf.split.lo && cur.mperf.split.lo) - perf_percent = (cur.aperf.split.lo * 100) / cur.mperf.split.lo; - else - perf_percent = 0; - -#else - if (unlikely(((unsigned long)(-1) / 100) < cur.aperf.whole)) { - int shift_count = 7; - cur.aperf.whole >>= shift_count; - cur.mperf.whole >>= shift_count; - } - - if (cur.aperf.whole && cur.mperf.whole) - perf_percent = (cur.aperf.whole * 100) / cur.mperf.whole; - else - perf_percent = 0; - -#endif + ratio = calc_aperfmperf_ratio(&per_cpu(old_perf, cpu), &perf); + per_cpu(old_perf, cpu) = perf; - retval = (policy->cpuinfo.max_freq * perf_percent) / 100; + retval = (policy->cpuinfo.max_freq * ratio) >> APERFMPERF_SHIFT; return retval; } @@ -394,7 +330,6 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, unsigned int next_perf_state = 0; /* Index into perf table */ unsigned int i; int result = 0; - struct power_trace it; dprintk("acpi_cpufreq_target %d (%d)\n", target_freq, policy->cpu); @@ -426,7 +361,7 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, } } - trace_power_mark(&it, POWER_PSTATE, next_perf_state); + trace_power_frequency(POWER_PSTATE, data->freq_table[next_state].frequency); switch (data->cpu_feature) { case SYSTEM_INTEL_MSR_CAPABLE: @@ -588,6 +523,21 @@ static const struct dmi_system_id sw_any_bug_dmi_table[] = { }, { } }; + +static int acpi_cpufreq_blacklist(struct cpuinfo_x86 *c) +{ + /* http://www.intel.com/Assets/PDF/specupdate/314554.pdf + * AL30: A Machine Check Exception (MCE) Occurring during an + * Enhanced Intel SpeedStep Technology Ratio Change May Cause + * Both Processor Cores to Lock Up when HT is enabled*/ + if (c->x86_vendor == X86_VENDOR_INTEL) { + if ((c->x86 == 15) && + (c->x86_model == 6) && + (c->x86_mask == 8) && smt_capable()) + return -ENODEV; + } + return 0; +} #endif static int acpi_cpufreq_cpu_init(struct cpufreq_policy *policy) @@ -602,6 +552,12 @@ static int acpi_cpufreq_cpu_init(struct cpufreq_policy *policy) dprintk("acpi_cpufreq_cpu_init\n"); +#ifdef CONFIG_SMP + result = acpi_cpufreq_blacklist(c); + if (result) + return result; +#endif + data = kzalloc(sizeof(struct acpi_cpufreq_data), GFP_KERNEL); if (!data) return -ENOMEM; @@ -731,12 +687,8 @@ static int acpi_cpufreq_cpu_init(struct cpufreq_policy *policy) acpi_processor_notify_smm(THIS_MODULE); /* Check for APERF/MPERF support in hardware */ - if (c->x86_vendor == X86_VENDOR_INTEL && c->cpuid_level >= 6) { - unsigned int ecx; - ecx = cpuid_ecx(6); - if (ecx & CPUID_6_ECX_APERFMPERF_CAPABILITY) - acpi_cpufreq_driver.getavg = get_measured_perf; - } + if (cpu_has(c, X86_FEATURE_APERFMPERF)) + acpi_cpufreq_driver.getavg = get_measured_perf; dprintk("CPU%u - ACPI performance management activated.\n", cpu); for (i = 0; i < perf->state_count; i++) diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c index 2a50ef891000..6394aa5c7985 100644 --- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c +++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c @@ -605,9 +605,10 @@ static int check_pst_table(struct powernow_k8_data *data, struct pst_s *pst, return 0; } -static void invalidate_entry(struct powernow_k8_data *data, unsigned int entry) +static void invalidate_entry(struct cpufreq_frequency_table *powernow_table, + unsigned int entry) { - data->powernow_table[entry].frequency = CPUFREQ_ENTRY_INVALID; + powernow_table[entry].frequency = CPUFREQ_ENTRY_INVALID; } static void print_basics(struct powernow_k8_data *data) @@ -854,6 +855,10 @@ static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data) goto err_out; } + /* fill in data */ + data->numps = data->acpi_data.state_count; + powernow_k8_acpi_pst_values(data, 0); + if (cpu_family == CPU_HW_PSTATE) ret_val = fill_powernow_table_pstate(data, powernow_table); else @@ -866,11 +871,8 @@ static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data) powernow_table[data->acpi_data.state_count].index = 0; data->powernow_table = powernow_table; - /* fill in data */ - data->numps = data->acpi_data.state_count; if (cpumask_first(cpu_core_mask(data->cpu)) == data->cpu) print_basics(data); - powernow_k8_acpi_pst_values(data, 0); /* notify BIOS that we exist */ acpi_processor_notify_smm(THIS_MODULE); @@ -914,13 +916,13 @@ static int fill_powernow_table_pstate(struct powernow_k8_data *data, "bad value %d.\n", i, index); printk(KERN_ERR PFX "Please report to BIOS " "manufacturer\n"); - invalidate_entry(data, i); + invalidate_entry(powernow_table, i); continue; } rdmsr(MSR_PSTATE_DEF_BASE + index, lo, hi); if (!(hi & HW_PSTATE_VALID_MASK)) { dprintk("invalid pstate %d, ignoring\n", index); - invalidate_entry(data, i); + invalidate_entry(powernow_table, i); continue; } @@ -941,7 +943,6 @@ static int fill_powernow_table_fidvid(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table) { int i; - int cntlofreq = 0; for (i = 0; i < data->acpi_data.state_count; i++) { u32 fid; @@ -970,7 +971,7 @@ static int fill_powernow_table_fidvid(struct powernow_k8_data *data, /* verify frequency is OK */ if ((freq > (MAX_FREQ * 1000)) || (freq < (MIN_FREQ * 1000))) { dprintk("invalid freq %u kHz, ignoring\n", freq); - invalidate_entry(data, i); + invalidate_entry(powernow_table, i); continue; } @@ -978,38 +979,17 @@ static int fill_powernow_table_fidvid(struct powernow_k8_data *data, * BIOSs are using "off" to indicate invalid */ if (vid == VID_OFF) { dprintk("invalid vid %u, ignoring\n", vid); - invalidate_entry(data, i); + invalidate_entry(powernow_table, i); continue; } - /* verify only 1 entry from the lo frequency table */ - if (fid < HI_FID_TABLE_BOTTOM) { - if (cntlofreq) { - /* if both entries are the same, - * ignore this one ... */ - if ((freq != powernow_table[cntlofreq].frequency) || - (index != powernow_table[cntlofreq].index)) { - printk(KERN_ERR PFX - "Too many lo freq table " - "entries\n"); - return 1; - } - - dprintk("double low frequency table entry, " - "ignoring it.\n"); - invalidate_entry(data, i); - continue; - } else - cntlofreq = i; - } - if (freq != (data->acpi_data.states[i].core_frequency * 1000)) { printk(KERN_INFO PFX "invalid freq entries " "%u kHz vs. %u kHz\n", freq, (unsigned int) (data->acpi_data.states[i].core_frequency * 1000)); - invalidate_entry(data, i); + invalidate_entry(powernow_table, i); continue; } } diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c index 93ba8eeb100a..08be922de33a 100644 --- a/arch/x86/kernel/cpu/hypervisor.c +++ b/arch/x86/kernel/cpu/hypervisor.c @@ -34,13 +34,6 @@ detect_hypervisor_vendor(struct cpuinfo_x86 *c) c->x86_hyper_vendor = X86_HYPER_VENDOR_NONE; } -unsigned long get_hypervisor_tsc_freq(void) -{ - if (boot_cpu_data.x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE) - return vmware_get_tsc_khz(); - return 0; -} - static inline void __cpuinit hypervisor_set_feature_bits(struct cpuinfo_x86 *c) { @@ -55,3 +48,10 @@ void __cpuinit init_hypervisor(struct cpuinfo_x86 *c) detect_hypervisor_vendor(c); hypervisor_set_feature_bits(c); } + +void __init init_hypervisor_platform(void) +{ + init_hypervisor(&boot_cpu_data); + if (boot_cpu_data.x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE) + vmware_platform_setup(); +} diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c index 80a722a071b5..40e1835b35e8 100644 --- a/arch/x86/kernel/cpu/intel.c +++ b/arch/x86/kernel/cpu/intel.c @@ -350,6 +350,12 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c) set_cpu_cap(c, X86_FEATURE_ARCH_PERFMON); } + if (c->cpuid_level > 6) { + unsigned ecx = cpuid_ecx(6); + if (ecx & 0x01) + set_cpu_cap(c, X86_FEATURE_APERFMPERF); + } + if (cpu_has_xmm2) set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC); if (cpu_has_ds) { diff --git a/arch/x86/kernel/cpu/mcheck/Makefile b/arch/x86/kernel/cpu/mcheck/Makefile index 188a1ca5ad2b..4ac6d48fe11b 100644 --- a/arch/x86/kernel/cpu/mcheck/Makefile +++ b/arch/x86/kernel/cpu/mcheck/Makefile @@ -1,11 +1,8 @@ -obj-y = mce.o +obj-y = mce.o mce-severity.o -obj-$(CONFIG_X86_NEW_MCE) += mce-severity.o -obj-$(CONFIG_X86_OLD_MCE) += k7.o p4.o p6.o obj-$(CONFIG_X86_ANCIENT_MCE) += winchip.o p5.o obj-$(CONFIG_X86_MCE_INTEL) += mce_intel.o obj-$(CONFIG_X86_MCE_AMD) += mce_amd.o -obj-$(CONFIG_X86_MCE_NONFATAL) += non-fatal.o obj-$(CONFIG_X86_MCE_THRESHOLD) += threshold.o obj-$(CONFIG_X86_MCE_INJECT) += mce-inject.o diff --git a/arch/x86/kernel/cpu/mcheck/k7.c b/arch/x86/kernel/cpu/mcheck/k7.c deleted file mode 100644 index b945d5dbc609..000000000000 --- a/arch/x86/kernel/cpu/mcheck/k7.c +++ /dev/null @@ -1,116 +0,0 @@ -/* - * Athlon specific Machine Check Exception Reporting - * (C) Copyright 2002 Dave Jones <davej@redhat.com> - */ -#include <linux/interrupt.h> -#include <linux/kernel.h> -#include <linux/types.h> -#include <linux/init.h> -#include <linux/smp.h> - -#include <asm/processor.h> -#include <asm/system.h> -#include <asm/mce.h> -#include <asm/msr.h> - -/* Machine Check Handler For AMD Athlon/Duron: */ -static void k7_machine_check(struct pt_regs *regs, long error_code) -{ - u32 alow, ahigh, high, low; - u32 mcgstl, mcgsth; - int recover = 1; - int i; - - rdmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth); - if (mcgstl & (1<<0)) /* Recoverable ? */ - recover = 0; - - printk(KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n", - smp_processor_id(), mcgsth, mcgstl); - - for (i = 1; i < nr_mce_banks; i++) { - rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high); - if (high & (1<<31)) { - char misc[20]; - char addr[24]; - - misc[0] = '\0'; - addr[0] = '\0'; - - if (high & (1<<29)) - recover |= 1; - if (high & (1<<25)) - recover |= 2; - high &= ~(1<<31); - - if (high & (1<<27)) { - rdmsr(MSR_IA32_MC0_MISC+i*4, alow, ahigh); - snprintf(misc, 20, "[%08x%08x]", ahigh, alow); - } - if (high & (1<<26)) { - rdmsr(MSR_IA32_MC0_ADDR+i*4, alow, ahigh); - snprintf(addr, 24, " at %08x%08x", ahigh, alow); - } - - printk(KERN_EMERG "CPU %d: Bank %d: %08x%08x%s%s\n", - smp_processor_id(), i, high, low, misc, addr); - - /* Clear it: */ - wrmsr(MSR_IA32_MC0_STATUS+i*4, 0UL, 0UL); - /* Serialize: */ - wmb(); - add_taint(TAINT_MACHINE_CHECK); - } - } - - if (recover & 2) - panic("CPU context corrupt"); - if (recover & 1) - panic("Unable to continue"); - - printk(KERN_EMERG "Attempting to continue.\n"); - - mcgstl &= ~(1<<2); - wrmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth); -} - - -/* AMD K7 machine check is Intel like: */ -void amd_mcheck_init(struct cpuinfo_x86 *c) -{ - u32 l, h; - int i; - - if (!cpu_has(c, X86_FEATURE_MCE)) - return; - - machine_check_vector = k7_machine_check; - /* Make sure the vector pointer is visible before we enable MCEs: */ - wmb(); - - printk(KERN_INFO "Intel machine check architecture supported.\n"); - - rdmsr(MSR_IA32_MCG_CAP, l, h); - if (l & (1<<8)) /* Control register present ? */ - wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff); - nr_mce_banks = l & 0xff; - - /* - * Clear status for MC index 0 separately, we don't touch CTL, - * as some K7 Athlons cause spurious MCEs when its enabled: - */ - if (boot_cpu_data.x86 == 6) { - wrmsr(MSR_IA32_MC0_STATUS, 0x0, 0x0); - i = 1; - } else - i = 0; - - for (; i < nr_mce_banks; i++) { - wrmsr(MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff); - wrmsr(MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0); - } - - set_in_cr4(X86_CR4_MCE); - printk(KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n", - smp_processor_id()); -} diff --git a/arch/x86/kernel/cpu/mcheck/mce-inject.c b/arch/x86/kernel/cpu/mcheck/mce-inject.c index a3a235a53f09..472763d92098 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-inject.c +++ b/arch/x86/kernel/cpu/mcheck/mce-inject.c @@ -18,7 +18,12 @@ #include <linux/string.h> #include <linux/fs.h> #include <linux/smp.h> +#include <linux/notifier.h> +#include <linux/kdebug.h> +#include <linux/cpu.h> +#include <linux/sched.h> #include <asm/mce.h> +#include <asm/apic.h> /* Update fake mce registers on current CPU. */ static void inject_mce(struct mce *m) @@ -39,44 +44,142 @@ static void inject_mce(struct mce *m) i->finished = 1; } -struct delayed_mce { - struct timer_list timer; - struct mce m; -}; +static void raise_poll(struct mce *m) +{ + unsigned long flags; + mce_banks_t b; -/* Inject mce on current CPU */ -static void raise_mce(unsigned long data) + memset(&b, 0xff, sizeof(mce_banks_t)); + local_irq_save(flags); + machine_check_poll(0, &b); + local_irq_restore(flags); + m->finished = 0; +} + +static void raise_exception(struct mce *m, struct pt_regs *pregs) { - struct delayed_mce *dm = (struct delayed_mce *)data; - struct mce *m = &dm->m; - int cpu = m->extcpu; + struct pt_regs regs; + unsigned long flags; - inject_mce(m); - if (m->status & MCI_STATUS_UC) { - struct pt_regs regs; + if (!pregs) { memset(®s, 0, sizeof(struct pt_regs)); regs.ip = m->ip; regs.cs = m->cs; + pregs = ®s; + } + /* in mcheck exeception handler, irq will be disabled */ + local_irq_save(flags); + do_machine_check(pregs, 0); + local_irq_restore(flags); + m->finished = 0; +} + +static cpumask_t mce_inject_cpumask; + +static int mce_raise_notify(struct notifier_block *self, + unsigned long val, void *data) +{ + struct die_args *args = (struct die_args *)data; + int cpu = smp_processor_id(); + struct mce *m = &__get_cpu_var(injectm); + if (val != DIE_NMI_IPI || !cpu_isset(cpu, mce_inject_cpumask)) + return NOTIFY_DONE; + cpu_clear(cpu, mce_inject_cpumask); + if (m->inject_flags & MCJ_EXCEPTION) + raise_exception(m, args->regs); + else if (m->status) + raise_poll(m); + return NOTIFY_STOP; +} + +static struct notifier_block mce_raise_nb = { + .notifier_call = mce_raise_notify, + .priority = 1000, +}; + +/* Inject mce on current CPU */ +static int raise_local(void) +{ + struct mce *m = &__get_cpu_var(injectm); + int context = MCJ_CTX(m->inject_flags); + int ret = 0; + int cpu = m->extcpu; + + if (m->inject_flags & MCJ_EXCEPTION) { printk(KERN_INFO "Triggering MCE exception on CPU %d\n", cpu); - do_machine_check(®s, 0); + switch (context) { + case MCJ_CTX_IRQ: + /* + * Could do more to fake interrupts like + * calling irq_enter, but the necessary + * machinery isn't exported currently. + */ + /*FALL THROUGH*/ + case MCJ_CTX_PROCESS: + raise_exception(m, NULL); + break; + default: + printk(KERN_INFO "Invalid MCE context\n"); + ret = -EINVAL; + } printk(KERN_INFO "MCE exception done on CPU %d\n", cpu); - } else { - mce_banks_t b; - memset(&b, 0xff, sizeof(mce_banks_t)); + } else if (m->status) { printk(KERN_INFO "Starting machine check poll CPU %d\n", cpu); - machine_check_poll(0, &b); + raise_poll(m); mce_notify_irq(); - printk(KERN_INFO "Finished machine check poll on CPU %d\n", - cpu); - } - kfree(dm); + printk(KERN_INFO "Machine check poll done on CPU %d\n", cpu); + } else + m->finished = 0; + + return ret; +} + +static void raise_mce(struct mce *m) +{ + int context = MCJ_CTX(m->inject_flags); + + inject_mce(m); + + if (context == MCJ_CTX_RANDOM) + return; + +#ifdef CONFIG_X86_LOCAL_APIC + if (m->inject_flags & MCJ_NMI_BROADCAST) { + unsigned long start; + int cpu; + get_online_cpus(); + mce_inject_cpumask = cpu_online_map; + cpu_clear(get_cpu(), mce_inject_cpumask); + for_each_online_cpu(cpu) { + struct mce *mcpu = &per_cpu(injectm, cpu); + if (!mcpu->finished || + MCJ_CTX(mcpu->inject_flags) != MCJ_CTX_RANDOM) + cpu_clear(cpu, mce_inject_cpumask); + } + if (!cpus_empty(mce_inject_cpumask)) + apic->send_IPI_mask(&mce_inject_cpumask, NMI_VECTOR); + start = jiffies; + while (!cpus_empty(mce_inject_cpumask)) { + if (!time_before(jiffies, start + 2*HZ)) { + printk(KERN_ERR + "Timeout waiting for mce inject NMI %lx\n", + *cpus_addr(mce_inject_cpumask)); + break; + } + cpu_relax(); + } + raise_local(); + put_cpu(); + put_online_cpus(); + } else +#endif + raise_local(); } /* Error injection interface */ static ssize_t mce_write(struct file *filp, const char __user *ubuf, size_t usize, loff_t *off) { - struct delayed_mce *dm; struct mce m; if (!capable(CAP_SYS_ADMIN)) @@ -96,19 +199,12 @@ static ssize_t mce_write(struct file *filp, const char __user *ubuf, if (m.extcpu >= num_possible_cpus() || !cpu_online(m.extcpu)) return -EINVAL; - dm = kmalloc(sizeof(struct delayed_mce), GFP_KERNEL); - if (!dm) - return -ENOMEM; - /* * Need to give user space some time to set everything up, * so do it a jiffie or two later everywhere. - * Should we use a hrtimer here for better synchronization? */ - memcpy(&dm->m, &m, sizeof(struct mce)); - setup_timer(&dm->timer, raise_mce, (unsigned long)dm); - dm->timer.expires = jiffies + 2; - add_timer_on(&dm->timer, m.extcpu); + schedule_timeout(2); + raise_mce(&m); return usize; } @@ -116,6 +212,7 @@ static int inject_init(void) { printk(KERN_INFO "Machine check injector initialized\n"); mce_chrdev_ops.write = mce_write; + register_die_notifier(&mce_raise_nb); return 0; } diff --git a/arch/x86/kernel/cpu/mcheck/mce-internal.h b/arch/x86/kernel/cpu/mcheck/mce-internal.h index 54dcb8ff12e5..32996f9fab67 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-internal.h +++ b/arch/x86/kernel/cpu/mcheck/mce-internal.h @@ -1,3 +1,4 @@ +#include <linux/sysdev.h> #include <asm/mce.h> enum severity_level { @@ -10,6 +11,20 @@ enum severity_level { MCE_PANIC_SEVERITY, }; +#define ATTR_LEN 16 + +/* One object for each MCE bank, shared by all CPUs */ +struct mce_bank { + u64 ctl; /* subevents to enable */ + unsigned char init; /* initialise bank? */ + struct sysdev_attribute attr; /* sysdev attribute */ + char attrname[ATTR_LEN]; /* attribute name */ +}; + int mce_severity(struct mce *a, int tolerant, char **msg); +struct dentry *mce_get_debugfs_dir(void); extern int mce_ser; + +extern struct mce_bank *mce_banks; + diff --git a/arch/x86/kernel/cpu/mcheck/mce-severity.c b/arch/x86/kernel/cpu/mcheck/mce-severity.c index ff0807f97056..8a85dd1b1aa1 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-severity.c +++ b/arch/x86/kernel/cpu/mcheck/mce-severity.c @@ -139,6 +139,7 @@ int mce_severity(struct mce *a, int tolerant, char **msg) } } +#ifdef CONFIG_DEBUG_FS static void *s_start(struct seq_file *f, loff_t *pos) { if (*pos >= ARRAY_SIZE(severities)) @@ -197,7 +198,7 @@ static int __init severities_debugfs_init(void) { struct dentry *dmce = NULL, *fseverities_coverage = NULL; - dmce = debugfs_create_dir("mce", NULL); + dmce = mce_get_debugfs_dir(); if (dmce == NULL) goto err_out; fseverities_coverage = debugfs_create_file("severities-coverage", @@ -209,10 +210,7 @@ static int __init severities_debugfs_init(void) return 0; err_out: - if (fseverities_coverage) - debugfs_remove(fseverities_coverage); - if (dmce) - debugfs_remove(dmce); return -ENOMEM; } late_initcall(severities_debugfs_init); +#endif diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c index 9bfe9d2ea615..b1598a9436d0 100644 --- a/arch/x86/kernel/cpu/mcheck/mce.c +++ b/arch/x86/kernel/cpu/mcheck/mce.c @@ -34,6 +34,7 @@ #include <linux/smp.h> #include <linux/fs.h> #include <linux/mm.h> +#include <linux/debugfs.h> #include <asm/processor.h> #include <asm/hw_irq.h> @@ -45,21 +46,8 @@ #include "mce-internal.h" -/* Handle unconfigured int18 (should never happen) */ -static void unexpected_machine_check(struct pt_regs *regs, long error_code) -{ - printk(KERN_ERR "CPU#%d: Unexpected int18 (Machine Check).\n", - smp_processor_id()); -} - -/* Call the installed machine check handler for this CPU setup. */ -void (*machine_check_vector)(struct pt_regs *, long error_code) = - unexpected_machine_check; - int mce_disabled __read_mostly; -#ifdef CONFIG_X86_NEW_MCE - #define MISC_MCELOG_MINOR 227 #define SPINUNIT 100 /* 100ns */ @@ -77,7 +65,6 @@ DEFINE_PER_CPU(unsigned, mce_exception_count); */ static int tolerant __read_mostly = 1; static int banks __read_mostly; -static u64 *bank __read_mostly; static int rip_msr __read_mostly; static int mce_bootlog __read_mostly = -1; static int monarch_timeout __read_mostly = -1; @@ -87,28 +74,35 @@ int mce_cmci_disabled __read_mostly; int mce_ignore_ce __read_mostly; int mce_ser __read_mostly; +struct mce_bank *mce_banks __read_mostly; + /* User mode helper program triggered by machine check event */ static unsigned long mce_need_notify; static char mce_helper[128]; static char *mce_helper_argv[2] = { mce_helper, NULL }; -static unsigned long dont_init_banks; - static DECLARE_WAIT_QUEUE_HEAD(mce_wait); static DEFINE_PER_CPU(struct mce, mces_seen); static int cpu_missing; +static void default_decode_mce(struct mce *m) +{ + pr_emerg("No human readable MCE decoding support on this CPU type.\n"); + pr_emerg("Run the message through 'mcelog --ascii' to decode.\n"); +} + +/* + * CPU/chipset specific EDAC code can register a callback here to print + * MCE errors in a human-readable form: + */ +void (*x86_mce_decode_callback)(struct mce *m) = default_decode_mce; +EXPORT_SYMBOL(x86_mce_decode_callback); /* MCA banks polled by the period polling timer for corrected events */ DEFINE_PER_CPU(mce_banks_t, mce_poll_banks) = { [0 ... BITS_TO_LONGS(MAX_NR_BANKS)-1] = ~0UL }; -static inline int skip_bank_init(int i) -{ - return i < BITS_PER_LONG && test_bit(i, &dont_init_banks); -} - static DEFINE_PER_CPU(struct work_struct, mce_work); /* Do initial initialization of a struct mce */ @@ -183,59 +177,60 @@ void mce_log(struct mce *mce) set_bit(0, &mce_need_notify); } -void __weak decode_mce(struct mce *m) -{ - return; -} - static void print_mce(struct mce *m) { - printk(KERN_EMERG - "CPU %d: Machine Check Exception: %16Lx Bank %d: %016Lx\n", + pr_emerg("CPU %d: Machine Check Exception: %16Lx Bank %d: %016Lx\n", m->extcpu, m->mcgstatus, m->bank, m->status); + if (m->ip) { - printk(KERN_EMERG "RIP%s %02x:<%016Lx> ", - !(m->mcgstatus & MCG_STATUS_EIPV) ? " !INEXACT!" : "", - m->cs, m->ip); + pr_emerg("RIP%s %02x:<%016Lx> ", + !(m->mcgstatus & MCG_STATUS_EIPV) ? " !INEXACT!" : "", + m->cs, m->ip); + if (m->cs == __KERNEL_CS) print_symbol("{%s}", m->ip); - printk(KERN_CONT "\n"); + pr_cont("\n"); } - printk(KERN_EMERG "TSC %llx ", m->tsc); + + pr_emerg("TSC %llx ", m->tsc); if (m->addr) - printk(KERN_CONT "ADDR %llx ", m->addr); + pr_cont("ADDR %llx ", m->addr); if (m->misc) - printk(KERN_CONT "MISC %llx ", m->misc); - printk(KERN_CONT "\n"); - printk(KERN_EMERG "PROCESSOR %u:%x TIME %llu SOCKET %u APIC %x\n", - m->cpuvendor, m->cpuid, m->time, m->socketid, - m->apicid); + pr_cont("MISC %llx ", m->misc); - decode_mce(m); + pr_cont("\n"); + pr_emerg("PROCESSOR %u:%x TIME %llu SOCKET %u APIC %x\n", + m->cpuvendor, m->cpuid, m->time, m->socketid, m->apicid); + + /* + * Print out human-readable details about the MCE error, + * (if the CPU has an implementation for that): + */ + x86_mce_decode_callback(m); } static void print_mce_head(void) { - printk(KERN_EMERG "\nHARDWARE ERROR\n"); + pr_emerg("\nHARDWARE ERROR\n"); } static void print_mce_tail(void) { - printk(KERN_EMERG "This is not a software problem!\n" -#if (!defined(CONFIG_EDAC) || !defined(CONFIG_CPU_SUP_AMD)) - "Run through mcelog --ascii to decode and contact your hardware vendor\n" -#endif - ); + pr_emerg("This is not a software problem!\n"); } #define PANIC_TIMEOUT 5 /* 5 seconds */ static atomic_t mce_paniced; +static int fake_panic; +static atomic_t mce_fake_paniced; + /* Panic in progress. Enable interrupts and wait for final IPI */ static void wait_for_panic(void) { long timeout = PANIC_TIMEOUT*USEC_PER_SEC; + preempt_disable(); local_irq_enable(); while (timeout-- > 0) @@ -249,15 +244,21 @@ static void mce_panic(char *msg, struct mce *final, char *exp) { int i; - /* - * Make sure only one CPU runs in machine check panic - */ - if (atomic_add_return(1, &mce_paniced) > 1) - wait_for_panic(); - barrier(); + if (!fake_panic) { + /* + * Make sure only one CPU runs in machine check panic + */ + if (atomic_inc_return(&mce_paniced) > 1) + wait_for_panic(); + barrier(); - bust_spinlocks(1); - console_verbose(); + bust_spinlocks(1); + console_verbose(); + } else { + /* Don't log too much for fake panic */ + if (atomic_inc_return(&mce_fake_paniced) > 1) + return; + } print_mce_head(); /* First print corrected ones that are still unlogged */ for (i = 0; i < MCE_LOG_LEN; i++) { @@ -284,9 +285,12 @@ static void mce_panic(char *msg, struct mce *final, char *exp) print_mce_tail(); if (exp) printk(KERN_EMERG "Machine check: %s\n", exp); - if (panic_timeout == 0) - panic_timeout = mce_panic_timeout; - panic(msg); + if (!fake_panic) { + if (panic_timeout == 0) + panic_timeout = mce_panic_timeout; + panic(msg); + } else + printk(KERN_EMERG "Fake kernel panic: %s\n", msg); } /* Support code for software error injection */ @@ -294,13 +298,14 @@ static void mce_panic(char *msg, struct mce *final, char *exp) static int msr_to_offset(u32 msr) { unsigned bank = __get_cpu_var(injectm.bank); + if (msr == rip_msr) return offsetof(struct mce, ip); - if (msr == MSR_IA32_MC0_STATUS + bank*4) + if (msr == MSR_IA32_MCx_STATUS(bank)) return offsetof(struct mce, status); - if (msr == MSR_IA32_MC0_ADDR + bank*4) + if (msr == MSR_IA32_MCx_ADDR(bank)) return offsetof(struct mce, addr); - if (msr == MSR_IA32_MC0_MISC + bank*4) + if (msr == MSR_IA32_MCx_MISC(bank)) return offsetof(struct mce, misc); if (msr == MSR_IA32_MCG_STATUS) return offsetof(struct mce, mcgstatus); @@ -311,13 +316,25 @@ static int msr_to_offset(u32 msr) static u64 mce_rdmsrl(u32 msr) { u64 v; + if (__get_cpu_var(injectm).finished) { int offset = msr_to_offset(msr); + if (offset < 0) return 0; return *(u64 *)((char *)&__get_cpu_var(injectm) + offset); } - rdmsrl(msr, v); + + if (rdmsrl_safe(msr, &v)) { + WARN_ONCE(1, "mce: Unable to read msr %d!\n", msr); + /* + * Return zero in case the access faulted. This should + * not happen normally but can happen if the CPU does + * something weird, or if the code is buggy. + */ + v = 0; + } + return v; } @@ -325,6 +342,7 @@ static void mce_wrmsrl(u32 msr, u64 v) { if (__get_cpu_var(injectm).finished) { int offset = msr_to_offset(msr); + if (offset >= 0) *(u64 *)((char *)&__get_cpu_var(injectm) + offset) = v; return; @@ -421,7 +439,7 @@ static inline void mce_get_rip(struct mce *m, struct pt_regs *regs) m->ip = mce_rdmsrl(rip_msr); } -#ifdef CONFIG_X86_LOCAL_APIC +#ifdef CONFIG_X86_LOCAL_APIC /* * Called after interrupts have been reenabled again * when a MCE happened during an interrupts off region @@ -505,7 +523,7 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) m.mcgstatus = mce_rdmsrl(MSR_IA32_MCG_STATUS); for (i = 0; i < banks; i++) { - if (!bank[i] || !test_bit(i, *b)) + if (!mce_banks[i].ctl || !test_bit(i, *b)) continue; m.misc = 0; @@ -514,7 +532,7 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) m.tsc = 0; barrier(); - m.status = mce_rdmsrl(MSR_IA32_MC0_STATUS + i*4); + m.status = mce_rdmsrl(MSR_IA32_MCx_STATUS(i)); if (!(m.status & MCI_STATUS_VAL)) continue; @@ -529,9 +547,9 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) continue; if (m.status & MCI_STATUS_MISCV) - m.misc = mce_rdmsrl(MSR_IA32_MC0_MISC + i*4); + m.misc = mce_rdmsrl(MSR_IA32_MCx_MISC(i)); if (m.status & MCI_STATUS_ADDRV) - m.addr = mce_rdmsrl(MSR_IA32_MC0_ADDR + i*4); + m.addr = mce_rdmsrl(MSR_IA32_MCx_ADDR(i)); if (!(flags & MCP_TIMESTAMP)) m.tsc = 0; @@ -547,7 +565,7 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) /* * Clear state for this bank. */ - mce_wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0); + mce_wrmsrl(MSR_IA32_MCx_STATUS(i), 0); } /* @@ -568,7 +586,7 @@ static int mce_no_way_out(struct mce *m, char **msg) int i; for (i = 0; i < banks; i++) { - m->status = mce_rdmsrl(MSR_IA32_MC0_STATUS + i*4); + m->status = mce_rdmsrl(MSR_IA32_MCx_STATUS(i)); if (mce_severity(m, tolerant, msg) >= MCE_PANIC_SEVERITY) return 1; } @@ -628,7 +646,7 @@ out: * This way we prevent any potential data corruption in a unrecoverable case * and also makes sure always all CPU's errors are examined. * - * Also this detects the case of an machine check event coming from outer + * Also this detects the case of a machine check event coming from outer * space (not detected by any CPUs) In this case some external agent wants * us to shut down, so panic too. * @@ -681,7 +699,7 @@ static void mce_reign(void) * No machine check event found. Must be some external * source or one CPU is hung. Panic. */ - if (!m && tolerant < 3) + if (global_worst <= MCE_KEEP_SEVERITY && tolerant < 3) mce_panic("Machine check from unknown source", NULL, NULL); /* @@ -715,7 +733,7 @@ static int mce_start(int *no_way_out) * global_nwo should be updated before mce_callin */ smp_wmb(); - order = atomic_add_return(1, &mce_callin); + order = atomic_inc_return(&mce_callin); /* * Wait for everyone. @@ -852,7 +870,7 @@ static void mce_clear_state(unsigned long *toclear) for (i = 0; i < banks; i++) { if (test_bit(i, toclear)) - mce_wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0); + mce_wrmsrl(MSR_IA32_MCx_STATUS(i), 0); } } @@ -905,11 +923,11 @@ void do_machine_check(struct pt_regs *regs, long error_code) mce_setup(&m); m.mcgstatus = mce_rdmsrl(MSR_IA32_MCG_STATUS); - no_way_out = mce_no_way_out(&m, &msg); - final = &__get_cpu_var(mces_seen); *final = m; + no_way_out = mce_no_way_out(&m, &msg); + barrier(); /* @@ -926,14 +944,14 @@ void do_machine_check(struct pt_regs *regs, long error_code) order = mce_start(&no_way_out); for (i = 0; i < banks; i++) { __clear_bit(i, toclear); - if (!bank[i]) + if (!mce_banks[i].ctl) continue; m.misc = 0; m.addr = 0; m.bank = i; - m.status = mce_rdmsrl(MSR_IA32_MC0_STATUS + i*4); + m.status = mce_rdmsrl(MSR_IA32_MCx_STATUS(i)); if ((m.status & MCI_STATUS_VAL) == 0) continue; @@ -974,9 +992,9 @@ void do_machine_check(struct pt_regs *regs, long error_code) kill_it = 1; if (m.status & MCI_STATUS_MISCV) - m.misc = mce_rdmsrl(MSR_IA32_MC0_MISC + i*4); + m.misc = mce_rdmsrl(MSR_IA32_MCx_MISC(i)); if (m.status & MCI_STATUS_ADDRV) - m.addr = mce_rdmsrl(MSR_IA32_MC0_ADDR + i*4); + m.addr = mce_rdmsrl(MSR_IA32_MCx_ADDR(i)); /* * Action optional error. Queue address for later processing. @@ -1101,7 +1119,7 @@ void mce_log_therm_throt_event(__u64 status) */ static int check_interval = 5 * 60; /* 5 minutes */ -static DEFINE_PER_CPU(int, next_interval); /* in jiffies */ +static DEFINE_PER_CPU(int, mce_next_interval); /* in jiffies */ static DEFINE_PER_CPU(struct timer_list, mce_timer); static void mcheck_timer(unsigned long data) @@ -1120,7 +1138,7 @@ static void mcheck_timer(unsigned long data) * Alert userspace if needed. If we logged an MCE, reduce the * polling interval, otherwise increase the polling interval. */ - n = &__get_cpu_var(next_interval); + n = &__get_cpu_var(mce_next_interval); if (mce_notify_irq()) *n = max(*n/2, HZ/100); else @@ -1169,10 +1187,26 @@ int mce_notify_irq(void) } EXPORT_SYMBOL_GPL(mce_notify_irq); +static int mce_banks_init(void) +{ + int i; + + mce_banks = kzalloc(banks * sizeof(struct mce_bank), GFP_KERNEL); + if (!mce_banks) + return -ENOMEM; + for (i = 0; i < banks; i++) { + struct mce_bank *b = &mce_banks[i]; + + b->ctl = -1ULL; + b->init = 1; + } + return 0; +} + /* * Initialize Machine Checks for a CPU. */ -static int mce_cap_init(void) +static int __cpuinit mce_cap_init(void) { unsigned b; u64 cap; @@ -1192,11 +1226,11 @@ static int mce_cap_init(void) /* Don't support asymmetric configurations today */ WARN_ON(banks != 0 && b != banks); banks = b; - if (!bank) { - bank = kmalloc(banks * sizeof(u64), GFP_KERNEL); - if (!bank) - return -ENOMEM; - memset(bank, 0xff, banks * sizeof(u64)); + if (!mce_banks) { + int err = mce_banks_init(); + + if (err) + return err; } /* Use accurate RIP reporting if available. */ @@ -1228,15 +1262,17 @@ static void mce_init(void) wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff); for (i = 0; i < banks; i++) { - if (skip_bank_init(i)) + struct mce_bank *b = &mce_banks[i]; + + if (!b->init) continue; - wrmsrl(MSR_IA32_MC0_CTL+4*i, bank[i]); - wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0); + wrmsrl(MSR_IA32_MCx_CTL(i), b->ctl); + wrmsrl(MSR_IA32_MCx_STATUS(i), 0); } } /* Add per CPU specific workarounds here */ -static int mce_cpu_quirks(struct cpuinfo_x86 *c) +static int __cpuinit mce_cpu_quirks(struct cpuinfo_x86 *c) { if (c->x86_vendor == X86_VENDOR_UNKNOWN) { pr_info("MCE: unknown CPU type - not enabling MCE support.\n"); @@ -1251,7 +1287,7 @@ static int mce_cpu_quirks(struct cpuinfo_x86 *c) * trips off incorrectly with the IOMMU & 3ware * & Cerberus: */ - clear_bit(10, (unsigned long *)&bank[4]); + clear_bit(10, (unsigned long *)&mce_banks[4].ctl); } if (c->x86 <= 17 && mce_bootlog < 0) { /* @@ -1265,7 +1301,7 @@ static int mce_cpu_quirks(struct cpuinfo_x86 *c) * by default. */ if (c->x86 == 6 && banks > 0) - bank[0] = 0; + mce_banks[0].ctl = 0; } if (c->x86_vendor == X86_VENDOR_INTEL) { @@ -1278,8 +1314,8 @@ static int mce_cpu_quirks(struct cpuinfo_x86 *c) * valid event later, merely don't write CTL0. */ - if (c->x86 == 6 && c->x86_model < 0x1A) - __set_bit(0, &dont_init_banks); + if (c->x86 == 6 && c->x86_model < 0x1A && banks > 0) + mce_banks[0].init = 0; /* * All newer Intel systems support MCE broadcasting. Enable @@ -1335,7 +1371,7 @@ static void mce_cpu_features(struct cpuinfo_x86 *c) static void mce_init_timer(void) { struct timer_list *t = &__get_cpu_var(mce_timer); - int *n = &__get_cpu_var(next_interval); + int *n = &__get_cpu_var(mce_next_interval); if (mce_ignore_ce) return; @@ -1348,6 +1384,17 @@ static void mce_init_timer(void) add_timer_on(t, smp_processor_id()); } +/* Handle unconfigured int18 (should never happen) */ +static void unexpected_machine_check(struct pt_regs *regs, long error_code) +{ + printk(KERN_ERR "CPU#%d: Unexpected int18 (Machine Check).\n", + smp_processor_id()); +} + +/* Call the installed machine check handler for this CPU setup. */ +void (*machine_check_vector)(struct pt_regs *, long error_code) = + unexpected_machine_check; + /* * Called for each booted CPU to set up machine checks. * Must be called with preempt off: @@ -1561,8 +1608,10 @@ static struct miscdevice mce_log_device = { */ static int __init mcheck_enable(char *str) { - if (*str == 0) + if (*str == 0) { enable_p5_mce(); + return 1; + } if (*str == '=') str++; if (!strcmp(str, "off")) @@ -1603,8 +1652,10 @@ static int mce_disable(void) int i; for (i = 0; i < banks; i++) { - if (!skip_bank_init(i)) - wrmsrl(MSR_IA32_MC0_CTL + i*4, 0); + struct mce_bank *b = &mce_banks[i]; + + if (b->init) + wrmsrl(MSR_IA32_MCx_CTL(i), 0); } return 0; } @@ -1679,14 +1730,15 @@ DEFINE_PER_CPU(struct sys_device, mce_dev); __cpuinitdata void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu); -static struct sysdev_attribute *bank_attrs; +static inline struct mce_bank *attr_to_bank(struct sysdev_attribute *attr) +{ + return container_of(attr, struct mce_bank, attr); +} static ssize_t show_bank(struct sys_device *s, struct sysdev_attribute *attr, char *buf) { - u64 b = bank[attr - bank_attrs]; - - return sprintf(buf, "%llx\n", b); + return sprintf(buf, "%llx\n", attr_to_bank(attr)->ctl); } static ssize_t set_bank(struct sys_device *s, struct sysdev_attribute *attr, @@ -1697,7 +1749,7 @@ static ssize_t set_bank(struct sys_device *s, struct sysdev_attribute *attr, if (strict_strtoull(buf, 0, &new) < 0) return -EINVAL; - bank[attr - bank_attrs] = new; + attr_to_bank(attr)->ctl = new; mce_restart(); return size; @@ -1839,7 +1891,7 @@ static __cpuinit int mce_create_device(unsigned int cpu) } for (j = 0; j < banks; j++) { err = sysdev_create_file(&per_cpu(mce_dev, cpu), - &bank_attrs[j]); + &mce_banks[j].attr); if (err) goto error2; } @@ -1848,10 +1900,10 @@ static __cpuinit int mce_create_device(unsigned int cpu) return 0; error2: while (--j >= 0) - sysdev_remove_file(&per_cpu(mce_dev, cpu), &bank_attrs[j]); + sysdev_remove_file(&per_cpu(mce_dev, cpu), &mce_banks[j].attr); error: while (--i >= 0) - sysdev_remove_file(&per_cpu(mce_dev, cpu), mce_attrs[i]); + sysdev_remove_file(&per_cpu(mce_dev, cpu), &mce_banks[i].attr); sysdev_unregister(&per_cpu(mce_dev, cpu)); @@ -1869,7 +1921,7 @@ static __cpuinit void mce_remove_device(unsigned int cpu) sysdev_remove_file(&per_cpu(mce_dev, cpu), mce_attrs[i]); for (i = 0; i < banks; i++) - sysdev_remove_file(&per_cpu(mce_dev, cpu), &bank_attrs[i]); + sysdev_remove_file(&per_cpu(mce_dev, cpu), &mce_banks[i].attr); sysdev_unregister(&per_cpu(mce_dev, cpu)); cpumask_clear_cpu(cpu, mce_dev_initialized); @@ -1886,8 +1938,10 @@ static void mce_disable_cpu(void *h) if (!(action & CPU_TASKS_FROZEN)) cmci_clear(); for (i = 0; i < banks; i++) { - if (!skip_bank_init(i)) - wrmsrl(MSR_IA32_MC0_CTL + i*4, 0); + struct mce_bank *b = &mce_banks[i]; + + if (b->init) + wrmsrl(MSR_IA32_MCx_CTL(i), 0); } } @@ -1902,8 +1956,10 @@ static void mce_reenable_cpu(void *h) if (!(action & CPU_TASKS_FROZEN)) cmci_reenable(); for (i = 0; i < banks; i++) { - if (!skip_bank_init(i)) - wrmsrl(MSR_IA32_MC0_CTL + i*4, bank[i]); + struct mce_bank *b = &mce_banks[i]; + + if (b->init) + wrmsrl(MSR_IA32_MCx_CTL(i), b->ctl); } } @@ -1935,7 +1991,7 @@ mce_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) case CPU_DOWN_FAILED: case CPU_DOWN_FAILED_FROZEN: t->expires = round_jiffies(jiffies + - __get_cpu_var(next_interval)); + __get_cpu_var(mce_next_interval)); add_timer_on(t, cpu); smp_call_function_single(cpu, mce_reenable_cpu, &action, 1); break; @@ -1951,35 +2007,21 @@ static struct notifier_block mce_cpu_notifier __cpuinitdata = { .notifier_call = mce_cpu_callback, }; -static __init int mce_init_banks(void) +static __init void mce_init_banks(void) { int i; - bank_attrs = kzalloc(sizeof(struct sysdev_attribute) * banks, - GFP_KERNEL); - if (!bank_attrs) - return -ENOMEM; - for (i = 0; i < banks; i++) { - struct sysdev_attribute *a = &bank_attrs[i]; + struct mce_bank *b = &mce_banks[i]; + struct sysdev_attribute *a = &b->attr; - a->attr.name = kasprintf(GFP_KERNEL, "bank%d", i); - if (!a->attr.name) - goto nomem; + a->attr.name = b->attrname; + snprintf(b->attrname, ATTR_LEN, "bank%d", i); a->attr.mode = 0644; a->show = show_bank; a->store = set_bank; } - return 0; - -nomem: - while (--i >= 0) - kfree(bank_attrs[i].attr.name); - kfree(bank_attrs); - bank_attrs = NULL; - - return -ENOMEM; } static __init int mce_init_device(void) @@ -1992,9 +2034,7 @@ static __init int mce_init_device(void) zalloc_cpumask_var(&mce_dev_initialized, GFP_KERNEL); - err = mce_init_banks(); - if (err) - return err; + mce_init_banks(); err = sysdev_class_register(&mce_sysclass); if (err) @@ -2014,57 +2054,65 @@ static __init int mce_init_device(void) device_initcall(mce_init_device); -#else /* CONFIG_X86_OLD_MCE: */ - -int nr_mce_banks; -EXPORT_SYMBOL_GPL(nr_mce_banks); /* non-fatal.o */ +/* + * Old style boot options parsing. Only for compatibility. + */ +static int __init mcheck_disable(char *str) +{ + mce_disabled = 1; + return 1; +} +__setup("nomce", mcheck_disable); -/* This has to be run for each processor */ -void mcheck_init(struct cpuinfo_x86 *c) +#ifdef CONFIG_DEBUG_FS +struct dentry *mce_get_debugfs_dir(void) { - if (mce_disabled) - return; + static struct dentry *dmce; - switch (c->x86_vendor) { - case X86_VENDOR_AMD: - amd_mcheck_init(c); - break; + if (!dmce) + dmce = debugfs_create_dir("mce", NULL); - case X86_VENDOR_INTEL: - if (c->x86 == 5) - intel_p5_mcheck_init(c); - if (c->x86 == 6) - intel_p6_mcheck_init(c); - if (c->x86 == 15) - intel_p4_mcheck_init(c); - break; + return dmce; +} - case X86_VENDOR_CENTAUR: - if (c->x86 == 5) - winchip_mcheck_init(c); - break; +static void mce_reset(void) +{ + cpu_missing = 0; + atomic_set(&mce_fake_paniced, 0); + atomic_set(&mce_executing, 0); + atomic_set(&mce_callin, 0); + atomic_set(&global_nwo, 0); +} - default: - break; - } - printk(KERN_INFO "mce: CPU supports %d MCE banks\n", nr_mce_banks); +static int fake_panic_get(void *data, u64 *val) +{ + *val = fake_panic; + return 0; } -static int __init mcheck_enable(char *str) +static int fake_panic_set(void *data, u64 val) { - mce_p5_enabled = 1; - return 1; + mce_reset(); + fake_panic = val; + return 0; } -__setup("mce", mcheck_enable); -#endif /* CONFIG_X86_OLD_MCE */ +DEFINE_SIMPLE_ATTRIBUTE(fake_panic_fops, fake_panic_get, + fake_panic_set, "%llu\n"); -/* - * Old style boot options parsing. Only for compatibility. - */ -static int __init mcheck_disable(char *str) +static int __init mce_debugfs_init(void) { - mce_disabled = 1; - return 1; + struct dentry *dmce, *ffake_panic; + + dmce = mce_get_debugfs_dir(); + if (!dmce) + return -ENOMEM; + ffake_panic = debugfs_create_file("fake_panic", 0444, dmce, NULL, + &fake_panic_fops); + if (!ffake_panic) + return -ENOMEM; + + return 0; } -__setup("nomce", mcheck_disable); +late_initcall(mce_debugfs_init); +#endif diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd.c b/arch/x86/kernel/cpu/mcheck/mce_amd.c index 1fecba404fd8..83a3d1f4efca 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_amd.c +++ b/arch/x86/kernel/cpu/mcheck/mce_amd.c @@ -69,7 +69,7 @@ struct threshold_bank { struct threshold_block *blocks; cpumask_var_t cpus; }; -static DEFINE_PER_CPU(struct threshold_bank *, threshold_banks[NR_BANKS]); +static DEFINE_PER_CPU(struct threshold_bank * [NR_BANKS], threshold_banks); #ifdef CONFIG_SMP static unsigned char shared_bank[NR_BANKS] = { @@ -489,8 +489,9 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) int i, err = 0; struct threshold_bank *b = NULL; char name[32]; +#ifdef CONFIG_SMP struct cpuinfo_x86 *c = &cpu_data(cpu); - +#endif sprintf(name, "threshold_bank%i", bank); diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel.c b/arch/x86/kernel/cpu/mcheck/mce_intel.c index e1acec0f7a32..7c785634af2b 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_intel.c +++ b/arch/x86/kernel/cpu/mcheck/mce_intel.c @@ -8,6 +8,7 @@ #include <linux/init.h> #include <linux/interrupt.h> #include <linux/percpu.h> +#include <linux/sched.h> #include <asm/apic.h> #include <asm/processor.h> #include <asm/msr.h> @@ -90,7 +91,7 @@ static void cmci_discover(int banks, int boot) if (test_bit(i, owned)) continue; - rdmsrl(MSR_IA32_MC0_CTL2 + i, val); + rdmsrl(MSR_IA32_MCx_CTL2(i), val); /* Already owned by someone else? */ if (val & CMCI_EN) { @@ -101,8 +102,8 @@ static void cmci_discover(int banks, int boot) } val |= CMCI_EN | CMCI_THRESHOLD; - wrmsrl(MSR_IA32_MC0_CTL2 + i, val); - rdmsrl(MSR_IA32_MC0_CTL2 + i, val); + wrmsrl(MSR_IA32_MCx_CTL2(i), val); + rdmsrl(MSR_IA32_MCx_CTL2(i), val); /* Did the enable bit stick? -- the bank supports CMCI */ if (val & CMCI_EN) { @@ -152,9 +153,9 @@ void cmci_clear(void) if (!test_bit(i, __get_cpu_var(mce_banks_owned))) continue; /* Disable CMCI */ - rdmsrl(MSR_IA32_MC0_CTL2 + i, val); + rdmsrl(MSR_IA32_MCx_CTL2(i), val); val &= ~(CMCI_EN|CMCI_THRESHOLD_MASK); - wrmsrl(MSR_IA32_MC0_CTL2 + i, val); + wrmsrl(MSR_IA32_MCx_CTL2(i), val); __clear_bit(i, __get_cpu_var(mce_banks_owned)); } spin_unlock_irqrestore(&cmci_discover_lock, flags); diff --git a/arch/x86/kernel/cpu/mcheck/non-fatal.c b/arch/x86/kernel/cpu/mcheck/non-fatal.c deleted file mode 100644 index f5f2d6f71fb6..000000000000 --- a/arch/x86/kernel/cpu/mcheck/non-fatal.c +++ /dev/null @@ -1,94 +0,0 @@ -/* - * Non Fatal Machine Check Exception Reporting - * - * (C) Copyright 2002 Dave Jones. <davej@redhat.com> - * - * This file contains routines to check for non-fatal MCEs every 15s - * - */ -#include <linux/interrupt.h> -#include <linux/workqueue.h> -#include <linux/jiffies.h> -#include <linux/kernel.h> -#include <linux/module.h> -#include <linux/types.h> -#include <linux/init.h> -#include <linux/smp.h> - -#include <asm/processor.h> -#include <asm/system.h> -#include <asm/mce.h> -#include <asm/msr.h> - -static int firstbank; - -#define MCE_RATE (15*HZ) /* timer rate is 15s */ - -static void mce_checkregs(void *info) -{ - u32 low, high; - int i; - - for (i = firstbank; i < nr_mce_banks; i++) { - rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high); - - if (!(high & (1<<31))) - continue; - - printk(KERN_INFO "MCE: The hardware reports a non fatal, " - "correctable incident occurred on CPU %d.\n", - smp_processor_id()); - - printk(KERN_INFO "Bank %d: %08x%08x\n", i, high, low); - - /* - * Scrub the error so we don't pick it up in MCE_RATE - * seconds time: - */ - wrmsr(MSR_IA32_MC0_STATUS+i*4, 0UL, 0UL); - - /* Serialize: */ - wmb(); - add_taint(TAINT_MACHINE_CHECK); - } -} - -static void mce_work_fn(struct work_struct *work); -static DECLARE_DELAYED_WORK(mce_work, mce_work_fn); - -static void mce_work_fn(struct work_struct *work) -{ - on_each_cpu(mce_checkregs, NULL, 1); - schedule_delayed_work(&mce_work, round_jiffies_relative(MCE_RATE)); -} - -static int __init init_nonfatal_mce_checker(void) -{ - struct cpuinfo_x86 *c = &boot_cpu_data; - - /* Check for MCE support */ - if (!cpu_has(c, X86_FEATURE_MCE)) - return -ENODEV; - - /* Check for PPro style MCA */ - if (!cpu_has(c, X86_FEATURE_MCA)) - return -ENODEV; - - /* Some Athlons misbehave when we frob bank 0 */ - if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD && - boot_cpu_data.x86 == 6) - firstbank = 1; - else - firstbank = 0; - - /* - * Check for non-fatal errors every MCE_RATE s - */ - schedule_delayed_work(&mce_work, round_jiffies_relative(MCE_RATE)); - printk(KERN_INFO "Machine check exception polling timer started.\n"); - - return 0; -} -module_init(init_nonfatal_mce_checker); - -MODULE_LICENSE("GPL"); diff --git a/arch/x86/kernel/cpu/mcheck/p4.c b/arch/x86/kernel/cpu/mcheck/p4.c deleted file mode 100644 index 4482aea9aa2e..000000000000 --- a/arch/x86/kernel/cpu/mcheck/p4.c +++ /dev/null @@ -1,163 +0,0 @@ -/* - * P4 specific Machine Check Exception Reporting - */ -#include <linux/kernel.h> -#include <linux/types.h> -#include <linux/init.h> -#include <linux/smp.h> - -#include <asm/processor.h> -#include <asm/mce.h> -#include <asm/msr.h> - -/* as supported by the P4/Xeon family */ -struct intel_mce_extended_msrs { - u32 eax; - u32 ebx; - u32 ecx; - u32 edx; - u32 esi; - u32 edi; - u32 ebp; - u32 esp; - u32 eflags; - u32 eip; - /* u32 *reserved[]; */ -}; - -static int mce_num_extended_msrs; - -/* P4/Xeon Extended MCE MSR retrieval, return 0 if unsupported */ -static void intel_get_extended_msrs(struct intel_mce_extended_msrs *r) -{ - u32 h; - - rdmsr(MSR_IA32_MCG_EAX, r->eax, h); - rdmsr(MSR_IA32_MCG_EBX, r->ebx, h); - rdmsr(MSR_IA32_MCG_ECX, r->ecx, h); - rdmsr(MSR_IA32_MCG_EDX, r->edx, h); - rdmsr(MSR_IA32_MCG_ESI, r->esi, h); - rdmsr(MSR_IA32_MCG_EDI, r->edi, h); - rdmsr(MSR_IA32_MCG_EBP, r->ebp, h); - rdmsr(MSR_IA32_MCG_ESP, r->esp, h); - rdmsr(MSR_IA32_MCG_EFLAGS, r->eflags, h); - rdmsr(MSR_IA32_MCG_EIP, r->eip, h); -} - -static void intel_machine_check(struct pt_regs *regs, long error_code) -{ - u32 alow, ahigh, high, low; - u32 mcgstl, mcgsth; - int recover = 1; - int i; - - rdmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth); - if (mcgstl & (1<<0)) /* Recoverable ? */ - recover = 0; - - printk(KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n", - smp_processor_id(), mcgsth, mcgstl); - - if (mce_num_extended_msrs > 0) { - struct intel_mce_extended_msrs dbg; - - intel_get_extended_msrs(&dbg); - - printk(KERN_DEBUG "CPU %d: EIP: %08x EFLAGS: %08x\n" - "\teax: %08x ebx: %08x ecx: %08x edx: %08x\n" - "\tesi: %08x edi: %08x ebp: %08x esp: %08x\n", - smp_processor_id(), dbg.eip, dbg.eflags, - dbg.eax, dbg.ebx, dbg.ecx, dbg.edx, - dbg.esi, dbg.edi, dbg.ebp, dbg.esp); - } - - for (i = 0; i < nr_mce_banks; i++) { - rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high); - if (high & (1<<31)) { - char misc[20]; - char addr[24]; - - misc[0] = addr[0] = '\0'; - if (high & (1<<29)) - recover |= 1; - if (high & (1<<25)) - recover |= 2; - high &= ~(1<<31); - if (high & (1<<27)) { - rdmsr(MSR_IA32_MC0_MISC+i*4, alow, ahigh); - snprintf(misc, 20, "[%08x%08x]", ahigh, alow); - } - if (high & (1<<26)) { - rdmsr(MSR_IA32_MC0_ADDR+i*4, alow, ahigh); - snprintf(addr, 24, " at %08x%08x", ahigh, alow); - } - printk(KERN_EMERG "CPU %d: Bank %d: %08x%08x%s%s\n", - smp_processor_id(), i, high, low, misc, addr); - } - } - - if (recover & 2) - panic("CPU context corrupt"); - if (recover & 1) - panic("Unable to continue"); - - printk(KERN_EMERG "Attempting to continue.\n"); - - /* - * Do not clear the MSR_IA32_MCi_STATUS if the error is not - * recoverable/continuable.This will allow BIOS to look at the MSRs - * for errors if the OS could not log the error. - */ - for (i = 0; i < nr_mce_banks; i++) { - u32 msr; - msr = MSR_IA32_MC0_STATUS+i*4; - rdmsr(msr, low, high); - if (high&(1<<31)) { - /* Clear it */ - wrmsr(msr, 0UL, 0UL); - /* Serialize */ - wmb(); - add_taint(TAINT_MACHINE_CHECK); - } - } - mcgstl &= ~(1<<2); - wrmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth); -} - -void intel_p4_mcheck_init(struct cpuinfo_x86 *c) -{ - u32 l, h; - int i; - - machine_check_vector = intel_machine_check; - wmb(); - - printk(KERN_INFO "Intel machine check architecture supported.\n"); - rdmsr(MSR_IA32_MCG_CAP, l, h); - if (l & (1<<8)) /* Control register present ? */ - wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff); - nr_mce_banks = l & 0xff; - - for (i = 0; i < nr_mce_banks; i++) { - wrmsr(MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff); - wrmsr(MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0); - } - - set_in_cr4(X86_CR4_MCE); - printk(KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n", - smp_processor_id()); - - /* Check for P4/Xeon extended MCE MSRs */ - rdmsr(MSR_IA32_MCG_CAP, l, h); - if (l & (1<<9)) {/* MCG_EXT_P */ - mce_num_extended_msrs = (l >> 16) & 0xff; - printk(KERN_INFO "CPU%d: Intel P4/Xeon Extended MCE MSRs (%d)" - " available\n", - smp_processor_id(), mce_num_extended_msrs); - -#ifdef CONFIG_X86_MCE_P4THERMAL - /* Check for P4/Xeon Thermal monitor */ - intel_init_thermal(c); -#endif - } -} diff --git a/arch/x86/kernel/cpu/mcheck/p6.c b/arch/x86/kernel/cpu/mcheck/p6.c deleted file mode 100644 index 01e4f8178183..000000000000 --- a/arch/x86/kernel/cpu/mcheck/p6.c +++ /dev/null @@ -1,127 +0,0 @@ -/* - * P6 specific Machine Check Exception Reporting - * (C) Copyright 2002 Alan Cox <alan@lxorguk.ukuu.org.uk> - */ -#include <linux/interrupt.h> -#include <linux/kernel.h> -#include <linux/types.h> -#include <linux/init.h> -#include <linux/smp.h> - -#include <asm/processor.h> -#include <asm/system.h> -#include <asm/mce.h> -#include <asm/msr.h> - -/* Machine Check Handler For PII/PIII */ -static void intel_machine_check(struct pt_regs *regs, long error_code) -{ - u32 alow, ahigh, high, low; - u32 mcgstl, mcgsth; - int recover = 1; - int i; - - rdmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth); - if (mcgstl & (1<<0)) /* Recoverable ? */ - recover = 0; - - printk(KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n", - smp_processor_id(), mcgsth, mcgstl); - - for (i = 0; i < nr_mce_banks; i++) { - rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high); - if (high & (1<<31)) { - char misc[20]; - char addr[24]; - - misc[0] = '\0'; - addr[0] = '\0'; - - if (high & (1<<29)) - recover |= 1; - if (high & (1<<25)) - recover |= 2; - high &= ~(1<<31); - - if (high & (1<<27)) { - rdmsr(MSR_IA32_MC0_MISC+i*4, alow, ahigh); - snprintf(misc, 20, "[%08x%08x]", ahigh, alow); - } - if (high & (1<<26)) { - rdmsr(MSR_IA32_MC0_ADDR+i*4, alow, ahigh); - snprintf(addr, 24, " at %08x%08x", ahigh, alow); - } - - printk(KERN_EMERG "CPU %d: Bank %d: %08x%08x%s%s\n", - smp_processor_id(), i, high, low, misc, addr); - } - } - - if (recover & 2) - panic("CPU context corrupt"); - if (recover & 1) - panic("Unable to continue"); - - printk(KERN_EMERG "Attempting to continue.\n"); - /* - * Do not clear the MSR_IA32_MCi_STATUS if the error is not - * recoverable/continuable.This will allow BIOS to look at the MSRs - * for errors if the OS could not log the error: - */ - for (i = 0; i < nr_mce_banks; i++) { - unsigned int msr; - - msr = MSR_IA32_MC0_STATUS+i*4; - rdmsr(msr, low, high); - if (high & (1<<31)) { - /* Clear it: */ - wrmsr(msr, 0UL, 0UL); - /* Serialize: */ - wmb(); - add_taint(TAINT_MACHINE_CHECK); - } - } - mcgstl &= ~(1<<2); - wrmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth); -} - -/* Set up machine check reporting for processors with Intel style MCE: */ -void intel_p6_mcheck_init(struct cpuinfo_x86 *c) -{ - u32 l, h; - int i; - - /* Check for MCE support */ - if (!cpu_has(c, X86_FEATURE_MCE)) - return; - - /* Check for PPro style MCA */ - if (!cpu_has(c, X86_FEATURE_MCA)) - return; - - /* Ok machine check is available */ - machine_check_vector = intel_machine_check; - /* Make sure the vector pointer is visible before we enable MCEs: */ - wmb(); - - printk(KERN_INFO "Intel machine check architecture supported.\n"); - rdmsr(MSR_IA32_MCG_CAP, l, h); - if (l & (1<<8)) /* Control register present ? */ - wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff); - nr_mce_banks = l & 0xff; - - /* - * Following the example in IA-32 SDM Vol 3: - * - MC0_CTL should not be written - * - Status registers on all banks should be cleared on reset - */ - for (i = 1; i < nr_mce_banks; i++) - wrmsr(MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff); - - for (i = 0; i < nr_mce_banks; i++) - wrmsr(MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0); - - set_in_cr4(X86_CR4_MCE); - printk(KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n", - smp_processor_id()); -} diff --git a/arch/x86/kernel/cpu/mcheck/therm_throt.c b/arch/x86/kernel/cpu/mcheck/therm_throt.c index 5957a93e5173..b3a1dba75330 100644 --- a/arch/x86/kernel/cpu/mcheck/therm_throt.c +++ b/arch/x86/kernel/cpu/mcheck/therm_throt.c @@ -34,20 +34,31 @@ /* How long to wait between reporting thermal events */ #define CHECK_INTERVAL (300 * HZ) -static DEFINE_PER_CPU(__u64, next_check) = INITIAL_JIFFIES; -static DEFINE_PER_CPU(unsigned long, thermal_throttle_count); -static DEFINE_PER_CPU(bool, thermal_throttle_active); +/* + * Current thermal throttling state: + */ +struct thermal_state { + bool is_throttled; + + u64 next_check; + unsigned long throttle_count; + unsigned long last_throttle_count; +}; + +static DEFINE_PER_CPU(struct thermal_state, thermal_state); -static atomic_t therm_throt_en = ATOMIC_INIT(0); +static atomic_t therm_throt_en = ATOMIC_INIT(0); #ifdef CONFIG_SYSFS #define define_therm_throt_sysdev_one_ro(_name) \ static SYSDEV_ATTR(_name, 0444, therm_throt_sysdev_show_##_name, NULL) #define define_therm_throt_sysdev_show_func(name) \ -static ssize_t therm_throt_sysdev_show_##name(struct sys_device *dev, \ - struct sysdev_attribute *attr, \ - char *buf) \ + \ +static ssize_t therm_throt_sysdev_show_##name( \ + struct sys_device *dev, \ + struct sysdev_attribute *attr, \ + char *buf) \ { \ unsigned int cpu = dev->id; \ ssize_t ret; \ @@ -55,7 +66,7 @@ static ssize_t therm_throt_sysdev_show_##name(struct sys_device *dev, \ preempt_disable(); /* CPU hotplug */ \ if (cpu_online(cpu)) \ ret = sprintf(buf, "%lu\n", \ - per_cpu(thermal_throttle_##name, cpu)); \ + per_cpu(thermal_state, cpu).name); \ else \ ret = 0; \ preempt_enable(); \ @@ -63,11 +74,11 @@ static ssize_t therm_throt_sysdev_show_##name(struct sys_device *dev, \ return ret; \ } -define_therm_throt_sysdev_show_func(count); -define_therm_throt_sysdev_one_ro(count); +define_therm_throt_sysdev_show_func(throttle_count); +define_therm_throt_sysdev_one_ro(throttle_count); static struct attribute *thermal_throttle_attrs[] = { - &attr_count.attr, + &attr_throttle_count.attr, NULL }; @@ -93,33 +104,39 @@ static struct attribute_group thermal_throttle_attr_group = { * 1 : Event should be logged further, and a message has been * printed to the syslog. */ -static int therm_throt_process(int curr) +static int therm_throt_process(bool is_throttled) { - unsigned int cpu = smp_processor_id(); - __u64 tmp_jiffs = get_jiffies_64(); - bool was_throttled = __get_cpu_var(thermal_throttle_active); - bool is_throttled = __get_cpu_var(thermal_throttle_active) = curr; + struct thermal_state *state; + unsigned int this_cpu; + bool was_throttled; + u64 now; + + this_cpu = smp_processor_id(); + now = get_jiffies_64(); + state = &per_cpu(thermal_state, this_cpu); + + was_throttled = state->is_throttled; + state->is_throttled = is_throttled; if (is_throttled) - __get_cpu_var(thermal_throttle_count)++; + state->throttle_count++; - if (!(was_throttled ^ is_throttled) && - time_before64(tmp_jiffs, __get_cpu_var(next_check))) + if (time_before64(now, state->next_check) && + state->throttle_count != state->last_throttle_count) return 0; - __get_cpu_var(next_check) = tmp_jiffs + CHECK_INTERVAL; + state->next_check = now + CHECK_INTERVAL; + state->last_throttle_count = state->throttle_count; /* if we just entered the thermal event */ if (is_throttled) { - printk(KERN_CRIT "CPU%d: Temperature above threshold, " - "cpu clock throttled (total events = %lu)\n", - cpu, __get_cpu_var(thermal_throttle_count)); + printk(KERN_CRIT "CPU%d: Temperature above threshold, cpu clock throttled (total events = %lu)\n", this_cpu, state->throttle_count); add_taint(TAINT_MACHINE_CHECK); return 1; } if (was_throttled) { - printk(KERN_INFO "CPU%d: Temperature/speed normal\n", cpu); + printk(KERN_INFO "CPU%d: Temperature/speed normal\n", this_cpu); return 1; } @@ -213,7 +230,7 @@ static void intel_thermal_interrupt(void) __u64 msr_val; rdmsrl(MSR_IA32_THERM_STATUS, msr_val); - if (therm_throt_process(msr_val & THERM_STATUS_PROCHOT)) + if (therm_throt_process((msr_val & THERM_STATUS_PROCHOT) != 0)) mce_log_therm_throt_event(msr_val); } @@ -260,9 +277,6 @@ void intel_init_thermal(struct cpuinfo_x86 *c) return; } - if (cpu_has(c, X86_FEATURE_TM2) && (l & MSR_IA32_MISC_ENABLE_TM2)) - tm2 = 1; - /* Check whether a vector already exists */ if (h & APIC_VECTOR_MASK) { printk(KERN_DEBUG @@ -271,6 +285,16 @@ void intel_init_thermal(struct cpuinfo_x86 *c) return; } + /* early Pentium M models use different method for enabling TM2 */ + if (cpu_has(c, X86_FEATURE_TM2)) { + if (c->x86 == 6 && (c->x86_model == 9 || c->x86_model == 13)) { + rdmsr(MSR_THERM2_CTL, l, h); + if (l & MSR_THERM2_CTL_TM_SELECT) + tm2 = 1; + } else if (l & MSR_IA32_MISC_ENABLE_TM2) + tm2 = 1; + } + /* We'll mask the thermal vector in the lapic till we're ready: */ h = THERMAL_APIC_VECTOR | APIC_DM_FIXED | APIC_LVT_MASKED; apic_write(APIC_LVTTHMR, h); diff --git a/arch/x86/kernel/cpu/mtrr/if.c b/arch/x86/kernel/cpu/mtrr/if.c index 08b6ea4c62b4..3c1b12d461d1 100644 --- a/arch/x86/kernel/cpu/mtrr/if.c +++ b/arch/x86/kernel/cpu/mtrr/if.c @@ -96,17 +96,24 @@ mtrr_write(struct file *file, const char __user *buf, size_t len, loff_t * ppos) unsigned long long base, size; char *ptr; char line[LINE_SIZE]; + int length; size_t linelen; if (!capable(CAP_SYS_ADMIN)) return -EPERM; - if (!len) - return -EINVAL; memset(line, 0, LINE_SIZE); - if (len > LINE_SIZE) - len = LINE_SIZE; - if (copy_from_user(line, buf, len - 1)) + + length = len; + length--; + + if (length > LINE_SIZE - 1) + length = LINE_SIZE - 1; + + if (length < 0) + return -EINVAL; + + if (copy_from_user(line, buf, length)) return -EFAULT; linelen = strlen(line); @@ -126,8 +133,8 @@ mtrr_write(struct file *file, const char __user *buf, size_t len, loff_t * ppos) return -EINVAL; base = simple_strtoull(line + 5, &ptr, 0); - for (; isspace(*ptr); ++ptr) - ; + while (isspace(*ptr)) + ptr++; if (strncmp(ptr, "size=", 5)) return -EINVAL; @@ -135,14 +142,14 @@ mtrr_write(struct file *file, const char __user *buf, size_t len, loff_t * ppos) size = simple_strtoull(ptr + 5, &ptr, 0); if ((base & 0xfff) || (size & 0xfff)) return -EINVAL; - for (; isspace(*ptr); ++ptr) - ; + while (isspace(*ptr)) + ptr++; if (strncmp(ptr, "type=", 5)) return -EINVAL; ptr += 5; - for (; isspace(*ptr); ++ptr) - ; + while (isspace(*ptr)) + ptr++; for (i = 0; i < MTRR_NUM_TYPES; ++i) { if (strcmp(ptr, mtrr_strings[i])) diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c index 7af0f88a4163..84e83de54575 100644 --- a/arch/x86/kernel/cpu/mtrr/main.c +++ b/arch/x86/kernel/cpu/mtrr/main.c @@ -58,6 +58,7 @@ unsigned int mtrr_usage_table[MTRR_MAX_VAR_RANGES]; static DEFINE_MUTEX(mtrr_mutex); u64 size_or_mask, size_and_mask; +static bool mtrr_aps_delayed_init; static struct mtrr_ops *mtrr_ops[X86_VENDOR_NUM]; @@ -163,7 +164,10 @@ static void ipi_handler(void *info) if (data->smp_reg != ~0U) { mtrr_if->set(data->smp_reg, data->smp_base, data->smp_size, data->smp_type); - } else { + } else if (mtrr_aps_delayed_init) { + /* + * Initialize the MTRRs inaddition to the synchronisation. + */ mtrr_if->set_all(); } @@ -265,6 +269,8 @@ set_mtrr(unsigned int reg, unsigned long base, unsigned long size, mtrr_type typ */ if (reg != ~0U) mtrr_if->set(reg, base, size, type); + else if (!mtrr_aps_delayed_init) + mtrr_if->set_all(); /* Wait for the others */ while (atomic_read(&data.count)) @@ -721,9 +727,7 @@ void __init mtrr_bp_init(void) void mtrr_ap_init(void) { - unsigned long flags; - - if (!mtrr_if || !use_intel()) + if (!use_intel() || mtrr_aps_delayed_init) return; /* * Ideally we should hold mtrr_mutex here to avoid mtrr entries @@ -738,11 +742,7 @@ void mtrr_ap_init(void) * 2. cpu hotadd time. We let mtrr_add/del_page hold cpuhotplug * lock to prevent mtrr entry changes */ - local_irq_save(flags); - - mtrr_if->set_all(); - - local_irq_restore(flags); + set_mtrr(~0U, 0, 0, 0); } /** @@ -753,6 +753,34 @@ void mtrr_save_state(void) smp_call_function_single(0, mtrr_save_fixed_ranges, NULL, 1); } +void set_mtrr_aps_delayed_init(void) +{ + if (!use_intel()) + return; + + mtrr_aps_delayed_init = true; +} + +/* + * MTRR initialization for all AP's + */ +void mtrr_aps_init(void) +{ + if (!use_intel()) + return; + + set_mtrr(~0U, 0, 0, 0); + mtrr_aps_delayed_init = false; +} + +void mtrr_bp_restore(void) +{ + if (!use_intel()) + return; + + mtrr_if->set_all(); +} + static int __init mtrr_init_finialize(void) { if (!mtrr_if) diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_event.c index f9cd0849bd42..2e20bca3cca1 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1,5 +1,5 @@ /* - * Performance counter x86 architecture code + * Performance events x86 architecture code * * Copyright (C) 2008 Thomas Gleixner <tglx@linutronix.de> * Copyright (C) 2008-2009 Red Hat, Inc., Ingo Molnar @@ -11,7 +11,7 @@ * For licencing details see kernel-base/COPYING */ -#include <linux/perf_counter.h> +#include <linux/perf_event.h> #include <linux/capability.h> #include <linux/notifier.h> #include <linux/hardirq.h> @@ -27,19 +27,19 @@ #include <asm/stacktrace.h> #include <asm/nmi.h> -static u64 perf_counter_mask __read_mostly; +static u64 perf_event_mask __read_mostly; -/* The maximal number of PEBS counters: */ -#define MAX_PEBS_COUNTERS 4 +/* The maximal number of PEBS events: */ +#define MAX_PEBS_EVENTS 4 /* The size of a BTS record in bytes: */ #define BTS_RECORD_SIZE 24 /* The size of a per-cpu BTS buffer in bytes: */ -#define BTS_BUFFER_SIZE (BTS_RECORD_SIZE * 1024) +#define BTS_BUFFER_SIZE (BTS_RECORD_SIZE * 2048) /* The BTS overflow threshold in bytes from the end of the buffer: */ -#define BTS_OVFL_TH (BTS_RECORD_SIZE * 64) +#define BTS_OVFL_TH (BTS_RECORD_SIZE * 128) /* @@ -65,11 +65,11 @@ struct debug_store { u64 pebs_index; u64 pebs_absolute_maximum; u64 pebs_interrupt_threshold; - u64 pebs_counter_reset[MAX_PEBS_COUNTERS]; + u64 pebs_event_reset[MAX_PEBS_EVENTS]; }; -struct cpu_hw_counters { - struct perf_counter *counters[X86_PMC_IDX_MAX]; +struct cpu_hw_events { + struct perf_event *events[X86_PMC_IDX_MAX]; unsigned long used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; unsigned long active_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; unsigned long interrupts; @@ -77,6 +77,18 @@ struct cpu_hw_counters { struct debug_store *ds; }; +struct event_constraint { + unsigned long idxmsk[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; + int code; +}; + +#define EVENT_CONSTRAINT(c, m) { .code = (c), .idxmsk[0] = (m) } +#define EVENT_CONSTRAINT_END { .code = 0, .idxmsk[0] = 0 } + +#define for_each_event_constraint(e, c) \ + for ((e) = (c); (e)->idxmsk[0]; (e)++) + + /* * struct x86_pmu - generic x86 pmu */ @@ -86,30 +98,34 @@ struct x86_pmu { int (*handle_irq)(struct pt_regs *); void (*disable_all)(void); void (*enable_all)(void); - void (*enable)(struct hw_perf_counter *, int); - void (*disable)(struct hw_perf_counter *, int); + void (*enable)(struct hw_perf_event *, int); + void (*disable)(struct hw_perf_event *, int); unsigned eventsel; unsigned perfctr; u64 (*event_map)(int); u64 (*raw_event)(u64); int max_events; - int num_counters; - int num_counters_fixed; - int counter_bits; - u64 counter_mask; + int num_events; + int num_events_fixed; + int event_bits; + u64 event_mask; int apic; u64 max_period; u64 intel_ctrl; void (*enable_bts)(u64 config); void (*disable_bts)(void); + int (*get_event_idx)(struct cpu_hw_events *cpuc, + struct hw_perf_event *hwc); }; static struct x86_pmu x86_pmu __read_mostly; -static DEFINE_PER_CPU(struct cpu_hw_counters, cpu_hw_counters) = { +static DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events) = { .enabled = 1, }; +static const struct event_constraint *event_constraints; + /* * Not sure about some of these */ @@ -124,37 +140,47 @@ static const u64 p6_perfmon_event_map[] = [PERF_COUNT_HW_BUS_CYCLES] = 0x0062, }; -static u64 p6_pmu_event_map(int event) +static u64 p6_pmu_event_map(int hw_event) { - return p6_perfmon_event_map[event]; + return p6_perfmon_event_map[hw_event]; } /* - * Counter setting that is specified not to count anything. + * Event setting that is specified not to count anything. * We use this to effectively disable a counter. * * L2_RQSTS with 0 MESI unit mask. */ -#define P6_NOP_COUNTER 0x0000002EULL +#define P6_NOP_EVENT 0x0000002EULL -static u64 p6_pmu_raw_event(u64 event) +static u64 p6_pmu_raw_event(u64 hw_event) { #define P6_EVNTSEL_EVENT_MASK 0x000000FFULL #define P6_EVNTSEL_UNIT_MASK 0x0000FF00ULL #define P6_EVNTSEL_EDGE_MASK 0x00040000ULL #define P6_EVNTSEL_INV_MASK 0x00800000ULL -#define P6_EVNTSEL_COUNTER_MASK 0xFF000000ULL +#define P6_EVNTSEL_REG_MASK 0xFF000000ULL #define P6_EVNTSEL_MASK \ (P6_EVNTSEL_EVENT_MASK | \ P6_EVNTSEL_UNIT_MASK | \ P6_EVNTSEL_EDGE_MASK | \ P6_EVNTSEL_INV_MASK | \ - P6_EVNTSEL_COUNTER_MASK) + P6_EVNTSEL_REG_MASK) - return event & P6_EVNTSEL_MASK; + return hw_event & P6_EVNTSEL_MASK; } +static const struct event_constraint intel_p6_event_constraints[] = +{ + EVENT_CONSTRAINT(0xc1, 0x1), /* FLOPS */ + EVENT_CONSTRAINT(0x10, 0x1), /* FP_COMP_OPS_EXE */ + EVENT_CONSTRAINT(0x11, 0x1), /* FP_ASSIST */ + EVENT_CONSTRAINT(0x12, 0x2), /* MUL */ + EVENT_CONSTRAINT(0x13, 0x2), /* DIV */ + EVENT_CONSTRAINT(0x14, 0x1), /* CYCLES_DIV_BUSY */ + EVENT_CONSTRAINT_END +}; /* * Intel PerfMon v3. Used on Core2 and later. @@ -170,16 +196,45 @@ static const u64 intel_perfmon_event_map[] = [PERF_COUNT_HW_BUS_CYCLES] = 0x013c, }; -static u64 intel_pmu_event_map(int event) +static const struct event_constraint intel_core_event_constraints[] = +{ + EVENT_CONSTRAINT(0x10, 0x1), /* FP_COMP_OPS_EXE */ + EVENT_CONSTRAINT(0x11, 0x2), /* FP_ASSIST */ + EVENT_CONSTRAINT(0x12, 0x2), /* MUL */ + EVENT_CONSTRAINT(0x13, 0x2), /* DIV */ + EVENT_CONSTRAINT(0x14, 0x1), /* CYCLES_DIV_BUSY */ + EVENT_CONSTRAINT(0x18, 0x1), /* IDLE_DURING_DIV */ + EVENT_CONSTRAINT(0x19, 0x2), /* DELAYED_BYPASS */ + EVENT_CONSTRAINT(0xa1, 0x1), /* RS_UOPS_DISPATCH_CYCLES */ + EVENT_CONSTRAINT(0xcb, 0x1), /* MEM_LOAD_RETIRED */ + EVENT_CONSTRAINT_END +}; + +static const struct event_constraint intel_nehalem_event_constraints[] = +{ + EVENT_CONSTRAINT(0x40, 0x3), /* L1D_CACHE_LD */ + EVENT_CONSTRAINT(0x41, 0x3), /* L1D_CACHE_ST */ + EVENT_CONSTRAINT(0x42, 0x3), /* L1D_CACHE_LOCK */ + EVENT_CONSTRAINT(0x43, 0x3), /* L1D_ALL_REF */ + EVENT_CONSTRAINT(0x4e, 0x3), /* L1D_PREFETCH */ + EVENT_CONSTRAINT(0x4c, 0x3), /* LOAD_HIT_PRE */ + EVENT_CONSTRAINT(0x51, 0x3), /* L1D */ + EVENT_CONSTRAINT(0x52, 0x3), /* L1D_CACHE_PREFETCH_LOCK_FB_HIT */ + EVENT_CONSTRAINT(0x53, 0x3), /* L1D_CACHE_LOCK_FB_HIT */ + EVENT_CONSTRAINT(0xc5, 0x3), /* CACHE_LOCK_CYCLES */ + EVENT_CONSTRAINT_END +}; + +static u64 intel_pmu_event_map(int hw_event) { - return intel_perfmon_event_map[event]; + return intel_perfmon_event_map[hw_event]; } /* - * Generalized hw caching related event table, filled + * Generalized hw caching related hw_event table, filled * in on a per model basis. A value of 0 means - * 'not supported', -1 means 'event makes no sense on - * this CPU', any other value means the raw event + * 'not supported', -1 means 'hw_event makes no sense on + * this CPU', any other value means the raw hw_event * ID. */ @@ -463,22 +518,22 @@ static const u64 atom_hw_cache_event_ids }, }; -static u64 intel_pmu_raw_event(u64 event) +static u64 intel_pmu_raw_event(u64 hw_event) { #define CORE_EVNTSEL_EVENT_MASK 0x000000FFULL #define CORE_EVNTSEL_UNIT_MASK 0x0000FF00ULL #define CORE_EVNTSEL_EDGE_MASK 0x00040000ULL #define CORE_EVNTSEL_INV_MASK 0x00800000ULL -#define CORE_EVNTSEL_COUNTER_MASK 0xFF000000ULL +#define CORE_EVNTSEL_REG_MASK 0xFF000000ULL #define CORE_EVNTSEL_MASK \ (CORE_EVNTSEL_EVENT_MASK | \ CORE_EVNTSEL_UNIT_MASK | \ CORE_EVNTSEL_EDGE_MASK | \ CORE_EVNTSEL_INV_MASK | \ - CORE_EVNTSEL_COUNTER_MASK) + CORE_EVNTSEL_REG_MASK) - return event & CORE_EVNTSEL_MASK; + return hw_event & CORE_EVNTSEL_MASK; } static const u64 amd_hw_cache_event_ids @@ -585,39 +640,39 @@ static const u64 amd_perfmon_event_map[] = [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, }; -static u64 amd_pmu_event_map(int event) +static u64 amd_pmu_event_map(int hw_event) { - return amd_perfmon_event_map[event]; + return amd_perfmon_event_map[hw_event]; } -static u64 amd_pmu_raw_event(u64 event) +static u64 amd_pmu_raw_event(u64 hw_event) { #define K7_EVNTSEL_EVENT_MASK 0x7000000FFULL #define K7_EVNTSEL_UNIT_MASK 0x00000FF00ULL #define K7_EVNTSEL_EDGE_MASK 0x000040000ULL #define K7_EVNTSEL_INV_MASK 0x000800000ULL -#define K7_EVNTSEL_COUNTER_MASK 0x0FF000000ULL +#define K7_EVNTSEL_REG_MASK 0x0FF000000ULL #define K7_EVNTSEL_MASK \ (K7_EVNTSEL_EVENT_MASK | \ K7_EVNTSEL_UNIT_MASK | \ K7_EVNTSEL_EDGE_MASK | \ K7_EVNTSEL_INV_MASK | \ - K7_EVNTSEL_COUNTER_MASK) + K7_EVNTSEL_REG_MASK) - return event & K7_EVNTSEL_MASK; + return hw_event & K7_EVNTSEL_MASK; } /* - * Propagate counter elapsed time into the generic counter. - * Can only be executed on the CPU where the counter is active. + * Propagate event elapsed time into the generic event. + * Can only be executed on the CPU where the event is active. * Returns the delta events processed. */ static u64 -x86_perf_counter_update(struct perf_counter *counter, - struct hw_perf_counter *hwc, int idx) +x86_perf_event_update(struct perf_event *event, + struct hw_perf_event *hwc, int idx) { - int shift = 64 - x86_pmu.counter_bits; + int shift = 64 - x86_pmu.event_bits; u64 prev_raw_count, new_raw_count; s64 delta; @@ -625,15 +680,15 @@ x86_perf_counter_update(struct perf_counter *counter, return 0; /* - * Careful: an NMI might modify the previous counter value. + * Careful: an NMI might modify the previous event value. * * Our tactic to handle this is to first atomically read and * exchange a new raw count - then add that new-prev delta - * count to the generic counter atomically: + * count to the generic event atomically: */ again: prev_raw_count = atomic64_read(&hwc->prev_count); - rdmsrl(hwc->counter_base + idx, new_raw_count); + rdmsrl(hwc->event_base + idx, new_raw_count); if (atomic64_cmpxchg(&hwc->prev_count, prev_raw_count, new_raw_count) != prev_raw_count) @@ -642,7 +697,7 @@ again: /* * Now we have the new raw value and have updated the prev * timestamp already. We can now calculate the elapsed delta - * (counter-)time and add that to the generic counter. + * (event-)time and add that to the generic event. * * Careful, not all hw sign-extends above the physical width * of the count. @@ -650,13 +705,13 @@ again: delta = (new_raw_count << shift) - (prev_raw_count << shift); delta >>= shift; - atomic64_add(delta, &counter->count); + atomic64_add(delta, &event->count); atomic64_sub(delta, &hwc->period_left); return new_raw_count; } -static atomic_t active_counters; +static atomic_t active_events; static DEFINE_MUTEX(pmc_reserve_mutex); static bool reserve_pmc_hardware(void) @@ -667,12 +722,12 @@ static bool reserve_pmc_hardware(void) if (nmi_watchdog == NMI_LOCAL_APIC) disable_lapic_nmi_watchdog(); - for (i = 0; i < x86_pmu.num_counters; i++) { + for (i = 0; i < x86_pmu.num_events; i++) { if (!reserve_perfctr_nmi(x86_pmu.perfctr + i)) goto perfctr_fail; } - for (i = 0; i < x86_pmu.num_counters; i++) { + for (i = 0; i < x86_pmu.num_events; i++) { if (!reserve_evntsel_nmi(x86_pmu.eventsel + i)) goto eventsel_fail; } @@ -685,7 +740,7 @@ eventsel_fail: for (i--; i >= 0; i--) release_evntsel_nmi(x86_pmu.eventsel + i); - i = x86_pmu.num_counters; + i = x86_pmu.num_events; perfctr_fail: for (i--; i >= 0; i--) @@ -703,7 +758,7 @@ static void release_pmc_hardware(void) #ifdef CONFIG_X86_LOCAL_APIC int i; - for (i = 0; i < x86_pmu.num_counters; i++) { + for (i = 0; i < x86_pmu.num_events; i++) { release_perfctr_nmi(x86_pmu.perfctr + i); release_evntsel_nmi(x86_pmu.eventsel + i); } @@ -720,7 +775,7 @@ static inline bool bts_available(void) static inline void init_debug_store_on_cpu(int cpu) { - struct debug_store *ds = per_cpu(cpu_hw_counters, cpu).ds; + struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds; if (!ds) return; @@ -732,7 +787,7 @@ static inline void init_debug_store_on_cpu(int cpu) static inline void fini_debug_store_on_cpu(int cpu) { - if (!per_cpu(cpu_hw_counters, cpu).ds) + if (!per_cpu(cpu_hw_events, cpu).ds) return; wrmsr_on_cpu(cpu, MSR_IA32_DS_AREA, 0, 0); @@ -751,12 +806,12 @@ static void release_bts_hardware(void) fini_debug_store_on_cpu(cpu); for_each_possible_cpu(cpu) { - struct debug_store *ds = per_cpu(cpu_hw_counters, cpu).ds; + struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds; if (!ds) continue; - per_cpu(cpu_hw_counters, cpu).ds = NULL; + per_cpu(cpu_hw_events, cpu).ds = NULL; kfree((void *)(unsigned long)ds->bts_buffer_base); kfree(ds); @@ -796,7 +851,7 @@ static int reserve_bts_hardware(void) ds->bts_interrupt_threshold = ds->bts_absolute_maximum - BTS_OVFL_TH; - per_cpu(cpu_hw_counters, cpu).ds = ds; + per_cpu(cpu_hw_events, cpu).ds = ds; err = 0; } @@ -812,9 +867,9 @@ static int reserve_bts_hardware(void) return err; } -static void hw_perf_counter_destroy(struct perf_counter *counter) +static void hw_perf_event_destroy(struct perf_event *event) { - if (atomic_dec_and_mutex_lock(&active_counters, &pmc_reserve_mutex)) { + if (atomic_dec_and_mutex_lock(&active_events, &pmc_reserve_mutex)) { release_pmc_hardware(); release_bts_hardware(); mutex_unlock(&pmc_reserve_mutex); @@ -827,7 +882,7 @@ static inline int x86_pmu_initialized(void) } static inline int -set_ext_hw_attr(struct hw_perf_counter *hwc, struct perf_counter_attr *attr) +set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr) { unsigned int cache_type, cache_op, cache_result; u64 config, val; @@ -880,7 +935,7 @@ static void intel_pmu_enable_bts(u64 config) static void intel_pmu_disable_bts(void) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); unsigned long debugctlmsr; if (!cpuc->ds) @@ -898,10 +953,10 @@ static void intel_pmu_disable_bts(void) /* * Setup the hardware configuration for a given attr_type */ -static int __hw_perf_counter_init(struct perf_counter *counter) +static int __hw_perf_event_init(struct perf_event *event) { - struct perf_counter_attr *attr = &counter->attr; - struct hw_perf_counter *hwc = &counter->hw; + struct perf_event_attr *attr = &event->attr; + struct hw_perf_event *hwc = &event->hw; u64 config; int err; @@ -909,27 +964,31 @@ static int __hw_perf_counter_init(struct perf_counter *counter) return -ENODEV; err = 0; - if (!atomic_inc_not_zero(&active_counters)) { + if (!atomic_inc_not_zero(&active_events)) { mutex_lock(&pmc_reserve_mutex); - if (atomic_read(&active_counters) == 0) { + if (atomic_read(&active_events) == 0) { if (!reserve_pmc_hardware()) err = -EBUSY; else err = reserve_bts_hardware(); } if (!err) - atomic_inc(&active_counters); + atomic_inc(&active_events); mutex_unlock(&pmc_reserve_mutex); } if (err) return err; + event->destroy = hw_perf_event_destroy; + /* * Generate PMC IRQs: * (keep 'enabled' bit clear for now) */ hwc->config = ARCH_PERFMON_EVENTSEL_INT; + hwc->idx = -1; + /* * Count user and OS events unless requested not to. */ @@ -946,17 +1005,15 @@ static int __hw_perf_counter_init(struct perf_counter *counter) /* * If we have a PMU initialized but no APIC * interrupts, we cannot sample hardware - * counters (user-space has to fall back and - * sample via a hrtimer based software counter): + * events (user-space has to fall back and + * sample via a hrtimer based software event): */ if (!x86_pmu.apic) return -EOPNOTSUPP; } - counter->destroy = hw_perf_counter_destroy; - /* - * Raw event type provide the config in the event structure + * Raw hw_event type provide the config in the hw_event structure */ if (attr->type == PERF_TYPE_RAW) { hwc->config |= x86_pmu.raw_event(attr->config); @@ -1001,7 +1058,7 @@ static int __hw_perf_counter_init(struct perf_counter *counter) static void p6_pmu_disable_all(void) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); u64 val; if (!cpuc->enabled) @@ -1018,7 +1075,7 @@ static void p6_pmu_disable_all(void) static void intel_pmu_disable_all(void) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); if (!cpuc->enabled) return; @@ -1034,7 +1091,7 @@ static void intel_pmu_disable_all(void) static void amd_pmu_disable_all(void) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); int idx; if (!cpuc->enabled) @@ -1043,12 +1100,12 @@ static void amd_pmu_disable_all(void) cpuc->enabled = 0; /* * ensure we write the disable before we start disabling the - * counters proper, so that amd_pmu_enable_counter() does the + * events proper, so that amd_pmu_enable_event() does the * right thing. */ barrier(); - for (idx = 0; idx < x86_pmu.num_counters; idx++) { + for (idx = 0; idx < x86_pmu.num_events; idx++) { u64 val; if (!test_bit(idx, cpuc->active_mask)) @@ -1070,7 +1127,7 @@ void hw_perf_disable(void) static void p6_pmu_enable_all(void) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); unsigned long val; if (cpuc->enabled) @@ -1087,7 +1144,7 @@ static void p6_pmu_enable_all(void) static void intel_pmu_enable_all(void) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); if (cpuc->enabled) return; @@ -1098,19 +1155,19 @@ static void intel_pmu_enable_all(void) wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl); if (test_bit(X86_PMC_IDX_FIXED_BTS, cpuc->active_mask)) { - struct perf_counter *counter = - cpuc->counters[X86_PMC_IDX_FIXED_BTS]; + struct perf_event *event = + cpuc->events[X86_PMC_IDX_FIXED_BTS]; - if (WARN_ON_ONCE(!counter)) + if (WARN_ON_ONCE(!event)) return; - intel_pmu_enable_bts(counter->hw.config); + intel_pmu_enable_bts(event->hw.config); } } static void amd_pmu_enable_all(void) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); int idx; if (cpuc->enabled) @@ -1119,14 +1176,14 @@ static void amd_pmu_enable_all(void) cpuc->enabled = 1; barrier(); - for (idx = 0; idx < x86_pmu.num_counters; idx++) { - struct perf_counter *counter = cpuc->counters[idx]; + for (idx = 0; idx < x86_pmu.num_events; idx++) { + struct perf_event *event = cpuc->events[idx]; u64 val; if (!test_bit(idx, cpuc->active_mask)) continue; - val = counter->hw.config; + val = event->hw.config; val |= ARCH_PERFMON_EVENTSEL0_ENABLE; wrmsrl(MSR_K7_EVNTSEL0 + idx, val); } @@ -1153,19 +1210,19 @@ static inline void intel_pmu_ack_status(u64 ack) wrmsrl(MSR_CORE_PERF_GLOBAL_OVF_CTRL, ack); } -static inline void x86_pmu_enable_counter(struct hw_perf_counter *hwc, int idx) +static inline void x86_pmu_enable_event(struct hw_perf_event *hwc, int idx) { (void)checking_wrmsrl(hwc->config_base + idx, hwc->config | ARCH_PERFMON_EVENTSEL0_ENABLE); } -static inline void x86_pmu_disable_counter(struct hw_perf_counter *hwc, int idx) +static inline void x86_pmu_disable_event(struct hw_perf_event *hwc, int idx) { (void)checking_wrmsrl(hwc->config_base + idx, hwc->config); } static inline void -intel_pmu_disable_fixed(struct hw_perf_counter *hwc, int __idx) +intel_pmu_disable_fixed(struct hw_perf_event *hwc, int __idx) { int idx = __idx - X86_PMC_IDX_FIXED; u64 ctrl_val, mask; @@ -1178,10 +1235,10 @@ intel_pmu_disable_fixed(struct hw_perf_counter *hwc, int __idx) } static inline void -p6_pmu_disable_counter(struct hw_perf_counter *hwc, int idx) +p6_pmu_disable_event(struct hw_perf_event *hwc, int idx) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); - u64 val = P6_NOP_COUNTER; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + u64 val = P6_NOP_EVENT; if (cpuc->enabled) val |= ARCH_PERFMON_EVENTSEL0_ENABLE; @@ -1190,7 +1247,7 @@ p6_pmu_disable_counter(struct hw_perf_counter *hwc, int idx) } static inline void -intel_pmu_disable_counter(struct hw_perf_counter *hwc, int idx) +intel_pmu_disable_event(struct hw_perf_event *hwc, int idx) { if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) { intel_pmu_disable_bts(); @@ -1202,24 +1259,24 @@ intel_pmu_disable_counter(struct hw_perf_counter *hwc, int idx) return; } - x86_pmu_disable_counter(hwc, idx); + x86_pmu_disable_event(hwc, idx); } static inline void -amd_pmu_disable_counter(struct hw_perf_counter *hwc, int idx) +amd_pmu_disable_event(struct hw_perf_event *hwc, int idx) { - x86_pmu_disable_counter(hwc, idx); + x86_pmu_disable_event(hwc, idx); } -static DEFINE_PER_CPU(u64, prev_left[X86_PMC_IDX_MAX]); +static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left); /* * Set the next IRQ period, based on the hwc->period_left value. - * To be called with the counter disabled in hw: + * To be called with the event disabled in hw: */ static int -x86_perf_counter_set_period(struct perf_counter *counter, - struct hw_perf_counter *hwc, int idx) +x86_perf_event_set_period(struct perf_event *event, + struct hw_perf_event *hwc, int idx) { s64 left = atomic64_read(&hwc->period_left); s64 period = hwc->sample_period; @@ -1245,7 +1302,7 @@ x86_perf_counter_set_period(struct perf_counter *counter, ret = 1; } /* - * Quirk: certain CPUs dont like it if just 1 event is left: + * Quirk: certain CPUs dont like it if just 1 hw_event is left: */ if (unlikely(left < 2)) left = 2; @@ -1253,24 +1310,24 @@ x86_perf_counter_set_period(struct perf_counter *counter, if (left > x86_pmu.max_period) left = x86_pmu.max_period; - per_cpu(prev_left[idx], smp_processor_id()) = left; + per_cpu(pmc_prev_left[idx], smp_processor_id()) = left; /* - * The hw counter starts counting from this counter offset, + * The hw event starts counting from this event offset, * mark it to be able to extra future deltas: */ atomic64_set(&hwc->prev_count, (u64)-left); - err = checking_wrmsrl(hwc->counter_base + idx, - (u64)(-left) & x86_pmu.counter_mask); + err = checking_wrmsrl(hwc->event_base + idx, + (u64)(-left) & x86_pmu.event_mask); - perf_counter_update_userpage(counter); + perf_event_update_userpage(event); return ret; } static inline void -intel_pmu_enable_fixed(struct hw_perf_counter *hwc, int __idx) +intel_pmu_enable_fixed(struct hw_perf_event *hwc, int __idx) { int idx = __idx - X86_PMC_IDX_FIXED; u64 ctrl_val, bits, mask; @@ -1295,9 +1352,9 @@ intel_pmu_enable_fixed(struct hw_perf_counter *hwc, int __idx) err = checking_wrmsrl(hwc->config_base, ctrl_val); } -static void p6_pmu_enable_counter(struct hw_perf_counter *hwc, int idx) +static void p6_pmu_enable_event(struct hw_perf_event *hwc, int idx) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); u64 val; val = hwc->config; @@ -1308,10 +1365,10 @@ static void p6_pmu_enable_counter(struct hw_perf_counter *hwc, int idx) } -static void intel_pmu_enable_counter(struct hw_perf_counter *hwc, int idx) +static void intel_pmu_enable_event(struct hw_perf_event *hwc, int idx) { if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) { - if (!__get_cpu_var(cpu_hw_counters).enabled) + if (!__get_cpu_var(cpu_hw_events).enabled) return; intel_pmu_enable_bts(hwc->config); @@ -1323,134 +1380,189 @@ static void intel_pmu_enable_counter(struct hw_perf_counter *hwc, int idx) return; } - x86_pmu_enable_counter(hwc, idx); + x86_pmu_enable_event(hwc, idx); } -static void amd_pmu_enable_counter(struct hw_perf_counter *hwc, int idx) +static void amd_pmu_enable_event(struct hw_perf_event *hwc, int idx) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); if (cpuc->enabled) - x86_pmu_enable_counter(hwc, idx); + x86_pmu_enable_event(hwc, idx); } -static int -fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc) +static int fixed_mode_idx(struct hw_perf_event *hwc) { - unsigned int event; + unsigned int hw_event; - event = hwc->config & ARCH_PERFMON_EVENT_MASK; + hw_event = hwc->config & ARCH_PERFMON_EVENT_MASK; - if (unlikely((event == + if (unlikely((hw_event == x86_pmu.event_map(PERF_COUNT_HW_BRANCH_INSTRUCTIONS)) && (hwc->sample_period == 1))) return X86_PMC_IDX_FIXED_BTS; - if (!x86_pmu.num_counters_fixed) + if (!x86_pmu.num_events_fixed) return -1; - if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS))) + /* + * fixed counters do not take all possible filters + */ + if (hwc->config & ARCH_PERFMON_EVENT_FILTER_MASK) + return -1; + + if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS))) return X86_PMC_IDX_FIXED_INSTRUCTIONS; - if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_CPU_CYCLES))) + if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_CPU_CYCLES))) return X86_PMC_IDX_FIXED_CPU_CYCLES; - if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_BUS_CYCLES))) + if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_BUS_CYCLES))) return X86_PMC_IDX_FIXED_BUS_CYCLES; return -1; } /* - * Find a PMC slot for the freshly enabled / scheduled in counter: + * generic counter allocator: get next free counter */ -static int x86_pmu_enable(struct perf_counter *counter) +static int +gen_get_event_idx(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); - struct hw_perf_counter *hwc = &counter->hw; int idx; - idx = fixed_mode_idx(counter, hwc); + idx = find_first_zero_bit(cpuc->used_mask, x86_pmu.num_events); + return idx == x86_pmu.num_events ? -1 : idx; +} + +/* + * intel-specific counter allocator: check event constraints + */ +static int +intel_get_event_idx(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc) +{ + const struct event_constraint *event_constraint; + int i, code; + + if (!event_constraints) + goto skip; + + code = hwc->config & CORE_EVNTSEL_EVENT_MASK; + + for_each_event_constraint(event_constraint, event_constraints) { + if (code == event_constraint->code) { + for_each_bit(i, event_constraint->idxmsk, X86_PMC_IDX_MAX) { + if (!test_and_set_bit(i, cpuc->used_mask)) + return i; + } + return -1; + } + } +skip: + return gen_get_event_idx(cpuc, hwc); +} + +static int +x86_schedule_event(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc) +{ + int idx; + + idx = fixed_mode_idx(hwc); if (idx == X86_PMC_IDX_FIXED_BTS) { /* BTS is already occupied. */ if (test_and_set_bit(idx, cpuc->used_mask)) return -EAGAIN; hwc->config_base = 0; - hwc->counter_base = 0; + hwc->event_base = 0; hwc->idx = idx; } else if (idx >= 0) { /* - * Try to get the fixed counter, if that is already taken - * then try to get a generic counter: + * Try to get the fixed event, if that is already taken + * then try to get a generic event: */ if (test_and_set_bit(idx, cpuc->used_mask)) goto try_generic; hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; /* - * We set it so that counter_base + idx in wrmsr/rdmsr maps to + * We set it so that event_base + idx in wrmsr/rdmsr maps to * MSR_ARCH_PERFMON_FIXED_CTR0 ... CTR2: */ - hwc->counter_base = + hwc->event_base = MSR_ARCH_PERFMON_FIXED_CTR0 - X86_PMC_IDX_FIXED; hwc->idx = idx; } else { idx = hwc->idx; - /* Try to get the previous generic counter again */ - if (test_and_set_bit(idx, cpuc->used_mask)) { + /* Try to get the previous generic event again */ + if (idx == -1 || test_and_set_bit(idx, cpuc->used_mask)) { try_generic: - idx = find_first_zero_bit(cpuc->used_mask, - x86_pmu.num_counters); - if (idx == x86_pmu.num_counters) + idx = x86_pmu.get_event_idx(cpuc, hwc); + if (idx == -1) return -EAGAIN; set_bit(idx, cpuc->used_mask); hwc->idx = idx; } - hwc->config_base = x86_pmu.eventsel; - hwc->counter_base = x86_pmu.perfctr; + hwc->config_base = x86_pmu.eventsel; + hwc->event_base = x86_pmu.perfctr; } - perf_counters_lapic_init(); + return idx; +} + +/* + * Find a PMC slot for the freshly enabled / scheduled in event: + */ +static int x86_pmu_enable(struct perf_event *event) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct hw_perf_event *hwc = &event->hw; + int idx; + + idx = x86_schedule_event(cpuc, hwc); + if (idx < 0) + return idx; + + perf_events_lapic_init(); x86_pmu.disable(hwc, idx); - cpuc->counters[idx] = counter; + cpuc->events[idx] = event; set_bit(idx, cpuc->active_mask); - x86_perf_counter_set_period(counter, hwc, idx); + x86_perf_event_set_period(event, hwc, idx); x86_pmu.enable(hwc, idx); - perf_counter_update_userpage(counter); + perf_event_update_userpage(event); return 0; } -static void x86_pmu_unthrottle(struct perf_counter *counter) +static void x86_pmu_unthrottle(struct perf_event *event) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); - struct hw_perf_counter *hwc = &counter->hw; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct hw_perf_event *hwc = &event->hw; if (WARN_ON_ONCE(hwc->idx >= X86_PMC_IDX_MAX || - cpuc->counters[hwc->idx] != counter)) + cpuc->events[hwc->idx] != event)) return; x86_pmu.enable(hwc, hwc->idx); } -void perf_counter_print_debug(void) +void perf_event_print_debug(void) { u64 ctrl, status, overflow, pmc_ctrl, pmc_count, prev_left, fixed; - struct cpu_hw_counters *cpuc; + struct cpu_hw_events *cpuc; unsigned long flags; int cpu, idx; - if (!x86_pmu.num_counters) + if (!x86_pmu.num_events) return; local_irq_save(flags); cpu = smp_processor_id(); - cpuc = &per_cpu(cpu_hw_counters, cpu); + cpuc = &per_cpu(cpu_hw_events, cpu); if (x86_pmu.version >= 2) { rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, ctrl); @@ -1466,11 +1578,11 @@ void perf_counter_print_debug(void) } pr_info("CPU#%d: used: %016llx\n", cpu, *(u64 *)cpuc->used_mask); - for (idx = 0; idx < x86_pmu.num_counters; idx++) { + for (idx = 0; idx < x86_pmu.num_events; idx++) { rdmsrl(x86_pmu.eventsel + idx, pmc_ctrl); rdmsrl(x86_pmu.perfctr + idx, pmc_count); - prev_left = per_cpu(prev_left[idx], cpu); + prev_left = per_cpu(pmc_prev_left[idx], cpu); pr_info("CPU#%d: gen-PMC%d ctrl: %016llx\n", cpu, idx, pmc_ctrl); @@ -1479,7 +1591,7 @@ void perf_counter_print_debug(void) pr_info("CPU#%d: gen-PMC%d left: %016llx\n", cpu, idx, prev_left); } - for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) { + for (idx = 0; idx < x86_pmu.num_events_fixed; idx++) { rdmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, pmc_count); pr_info("CPU#%d: fixed-PMC%d count: %016llx\n", @@ -1488,8 +1600,7 @@ void perf_counter_print_debug(void) local_irq_restore(flags); } -static void intel_pmu_drain_bts_buffer(struct cpu_hw_counters *cpuc, - struct perf_sample_data *data) +static void intel_pmu_drain_bts_buffer(struct cpu_hw_events *cpuc) { struct debug_store *ds = cpuc->ds; struct bts_record { @@ -1497,11 +1608,14 @@ static void intel_pmu_drain_bts_buffer(struct cpu_hw_counters *cpuc, u64 to; u64 flags; }; - struct perf_counter *counter = cpuc->counters[X86_PMC_IDX_FIXED_BTS]; - unsigned long orig_ip = data->regs->ip; + struct perf_event *event = cpuc->events[X86_PMC_IDX_FIXED_BTS]; struct bts_record *at, *top; + struct perf_output_handle handle; + struct perf_event_header header; + struct perf_sample_data data; + struct pt_regs regs; - if (!counter) + if (!event) return; if (!ds) @@ -1510,26 +1624,45 @@ static void intel_pmu_drain_bts_buffer(struct cpu_hw_counters *cpuc, at = (struct bts_record *)(unsigned long)ds->bts_buffer_base; top = (struct bts_record *)(unsigned long)ds->bts_index; + if (top <= at) + return; + ds->bts_index = ds->bts_buffer_base; + + data.period = event->hw.last_period; + data.addr = 0; + regs.ip = 0; + + /* + * Prepare a generic sample, i.e. fill in the invariant fields. + * We will overwrite the from and to address before we output + * the sample. + */ + perf_prepare_sample(&header, &data, event, ®s); + + if (perf_output_begin(&handle, event, + header.size * (top - at), 1, 1)) + return; + for (; at < top; at++) { - data->regs->ip = at->from; - data->addr = at->to; + data.ip = at->from; + data.addr = at->to; - perf_counter_output(counter, 1, data); + perf_output_sample(&handle, &header, &data, event); } - data->regs->ip = orig_ip; - data->addr = 0; + perf_output_end(&handle); /* There's new data available. */ - counter->pending_kill = POLL_IN; + event->hw.interrupts++; + event->pending_kill = POLL_IN; } -static void x86_pmu_disable(struct perf_counter *counter) +static void x86_pmu_disable(struct perf_event *event) { - struct cpu_hw_counters *cpuc = &__get_cpu_var(cpu_hw_counters); - struct hw_perf_counter *hwc = &counter->hw; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct hw_perf_event *hwc = &event->hw; int idx = hwc->idx; /* @@ -1541,67 +1674,63 @@ static void x86_pmu_disable(struct perf_counter *counter) /* * Make sure the cleared pointer becomes visible before we - * (potentially) free the counter: + * (potentially) free the event: */ barrier(); /* - * Drain the remaining delta count out of a counter + * Drain the remaining delta count out of a event * that we are disabling: */ - x86_perf_counter_update(counter, hwc, idx); + x86_perf_event_update(event, hwc, idx); /* Drain the remaining BTS records. */ - if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) { - struct perf_sample_data data; - struct pt_regs regs; + if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) + intel_pmu_drain_bts_buffer(cpuc); - data.regs = ®s; - intel_pmu_drain_bts_buffer(cpuc, &data); - } - cpuc->counters[idx] = NULL; + cpuc->events[idx] = NULL; clear_bit(idx, cpuc->used_mask); - perf_counter_update_userpage(counter); + perf_event_update_userpage(event); } /* - * Save and restart an expired counter. Called by NMI contexts, - * so it has to be careful about preempting normal counter ops: + * Save and restart an expired event. Called by NMI contexts, + * so it has to be careful about preempting normal event ops: */ -static int intel_pmu_save_and_restart(struct perf_counter *counter) +static int intel_pmu_save_and_restart(struct perf_event *event) { - struct hw_perf_counter *hwc = &counter->hw; + struct hw_perf_event *hwc = &event->hw; int idx = hwc->idx; int ret; - x86_perf_counter_update(counter, hwc, idx); - ret = x86_perf_counter_set_period(counter, hwc, idx); + x86_perf_event_update(event, hwc, idx); + ret = x86_perf_event_set_period(event, hwc, idx); - if (counter->state == PERF_COUNTER_STATE_ACTIVE) - intel_pmu_enable_counter(hwc, idx); + if (event->state == PERF_EVENT_STATE_ACTIVE) + intel_pmu_enable_event(hwc, idx); return ret; } static void intel_pmu_reset(void) { - struct debug_store *ds = __get_cpu_var(cpu_hw_counters).ds; + struct debug_store *ds = __get_cpu_var(cpu_hw_events).ds; unsigned long flags; int idx; - if (!x86_pmu.num_counters) + if (!x86_pmu.num_events) return; local_irq_save(flags); printk("clearing PMU state on CPU#%d\n", smp_processor_id()); - for (idx = 0; idx < x86_pmu.num_counters; idx++) { + for (idx = 0; idx < x86_pmu.num_events; idx++) { checking_wrmsrl(x86_pmu.eventsel + idx, 0ull); checking_wrmsrl(x86_pmu.perfctr + idx, 0ull); } - for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) { + for (idx = 0; idx < x86_pmu.num_events_fixed; idx++) { checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull); } if (ds) @@ -1613,39 +1742,38 @@ static void intel_pmu_reset(void) static int p6_pmu_handle_irq(struct pt_regs *regs) { struct perf_sample_data data; - struct cpu_hw_counters *cpuc; - struct perf_counter *counter; - struct hw_perf_counter *hwc; + struct cpu_hw_events *cpuc; + struct perf_event *event; + struct hw_perf_event *hwc; int idx, handled = 0; u64 val; - data.regs = regs; data.addr = 0; - cpuc = &__get_cpu_var(cpu_hw_counters); + cpuc = &__get_cpu_var(cpu_hw_events); - for (idx = 0; idx < x86_pmu.num_counters; idx++) { + for (idx = 0; idx < x86_pmu.num_events; idx++) { if (!test_bit(idx, cpuc->active_mask)) continue; - counter = cpuc->counters[idx]; - hwc = &counter->hw; + event = cpuc->events[idx]; + hwc = &event->hw; - val = x86_perf_counter_update(counter, hwc, idx); - if (val & (1ULL << (x86_pmu.counter_bits - 1))) + val = x86_perf_event_update(event, hwc, idx); + if (val & (1ULL << (x86_pmu.event_bits - 1))) continue; /* - * counter overflow + * event overflow */ handled = 1; - data.period = counter->hw.last_period; + data.period = event->hw.last_period; - if (!x86_perf_counter_set_period(counter, hwc, idx)) + if (!x86_perf_event_set_period(event, hwc, idx)) continue; - if (perf_counter_overflow(counter, 1, &data)) - p6_pmu_disable_counter(hwc, idx); + if (perf_event_overflow(event, 1, &data, regs)) + p6_pmu_disable_event(hwc, idx); } if (handled) @@ -1661,17 +1789,16 @@ static int p6_pmu_handle_irq(struct pt_regs *regs) static int intel_pmu_handle_irq(struct pt_regs *regs) { struct perf_sample_data data; - struct cpu_hw_counters *cpuc; + struct cpu_hw_events *cpuc; int bit, loops; u64 ack, status; - data.regs = regs; data.addr = 0; - cpuc = &__get_cpu_var(cpu_hw_counters); + cpuc = &__get_cpu_var(cpu_hw_events); perf_disable(); - intel_pmu_drain_bts_buffer(cpuc, &data); + intel_pmu_drain_bts_buffer(cpuc); status = intel_pmu_get_status(); if (!status) { perf_enable(); @@ -1681,8 +1808,8 @@ static int intel_pmu_handle_irq(struct pt_regs *regs) loops = 0; again: if (++loops > 100) { - WARN_ONCE(1, "perfcounters: irq loop stuck!\n"); - perf_counter_print_debug(); + WARN_ONCE(1, "perfevents: irq loop stuck!\n"); + perf_event_print_debug(); intel_pmu_reset(); perf_enable(); return 1; @@ -1691,19 +1818,19 @@ again: inc_irq_stat(apic_perf_irqs); ack = status; for_each_bit(bit, (unsigned long *)&status, X86_PMC_IDX_MAX) { - struct perf_counter *counter = cpuc->counters[bit]; + struct perf_event *event = cpuc->events[bit]; clear_bit(bit, (unsigned long *) &status); if (!test_bit(bit, cpuc->active_mask)) continue; - if (!intel_pmu_save_and_restart(counter)) + if (!intel_pmu_save_and_restart(event)) continue; - data.period = counter->hw.last_period; + data.period = event->hw.last_period; - if (perf_counter_overflow(counter, 1, &data)) - intel_pmu_disable_counter(&counter->hw, bit); + if (perf_event_overflow(event, 1, &data, regs)) + intel_pmu_disable_event(&event->hw, bit); } intel_pmu_ack_status(ack); @@ -1723,39 +1850,38 @@ again: static int amd_pmu_handle_irq(struct pt_regs *regs) { struct perf_sample_data data; - struct cpu_hw_counters *cpuc; - struct perf_counter *counter; - struct hw_perf_counter *hwc; + struct cpu_hw_events *cpuc; + struct perf_event *event; + struct hw_perf_event *hwc; int idx, handled = 0; u64 val; - data.regs = regs; data.addr = 0; - cpuc = &__get_cpu_var(cpu_hw_counters); + cpuc = &__get_cpu_var(cpu_hw_events); - for (idx = 0; idx < x86_pmu.num_counters; idx++) { + for (idx = 0; idx < x86_pmu.num_events; idx++) { if (!test_bit(idx, cpuc->active_mask)) continue; - counter = cpuc->counters[idx]; - hwc = &counter->hw; + event = cpuc->events[idx]; + hwc = &event->hw; - val = x86_perf_counter_update(counter, hwc, idx); - if (val & (1ULL << (x86_pmu.counter_bits - 1))) + val = x86_perf_event_update(event, hwc, idx); + if (val & (1ULL << (x86_pmu.event_bits - 1))) continue; /* - * counter overflow + * event overflow */ handled = 1; - data.period = counter->hw.last_period; + data.period = event->hw.last_period; - if (!x86_perf_counter_set_period(counter, hwc, idx)) + if (!x86_perf_event_set_period(event, hwc, idx)) continue; - if (perf_counter_overflow(counter, 1, &data)) - amd_pmu_disable_counter(hwc, idx); + if (perf_event_overflow(event, 1, &data, regs)) + amd_pmu_disable_event(hwc, idx); } if (handled) @@ -1769,18 +1895,21 @@ void smp_perf_pending_interrupt(struct pt_regs *regs) irq_enter(); ack_APIC_irq(); inc_irq_stat(apic_pending_irqs); - perf_counter_do_pending(); + perf_event_do_pending(); irq_exit(); } -void set_perf_counter_pending(void) +void set_perf_event_pending(void) { #ifdef CONFIG_X86_LOCAL_APIC + if (!x86_pmu.apic || !x86_pmu_initialized()) + return; + apic->send_IPI_self(LOCAL_PENDING_VECTOR); #endif } -void perf_counters_lapic_init(void) +void perf_events_lapic_init(void) { #ifdef CONFIG_X86_LOCAL_APIC if (!x86_pmu.apic || !x86_pmu_initialized()) @@ -1794,13 +1923,13 @@ void perf_counters_lapic_init(void) } static int __kprobes -perf_counter_nmi_handler(struct notifier_block *self, +perf_event_nmi_handler(struct notifier_block *self, unsigned long cmd, void *__args) { struct die_args *args = __args; struct pt_regs *regs; - if (!atomic_read(&active_counters)) + if (!atomic_read(&active_events)) return NOTIFY_DONE; switch (cmd) { @@ -1819,7 +1948,7 @@ perf_counter_nmi_handler(struct notifier_block *self, #endif /* * Can't rely on the handled return value to say it was our NMI, two - * counters could trigger 'simultaneously' raising two back-to-back NMIs. + * events could trigger 'simultaneously' raising two back-to-back NMIs. * * If the first NMI handles both, the latter will be empty and daze * the CPU. @@ -1829,8 +1958,8 @@ perf_counter_nmi_handler(struct notifier_block *self, return NOTIFY_STOP; } -static __read_mostly struct notifier_block perf_counter_nmi_notifier = { - .notifier_call = perf_counter_nmi_handler, +static __read_mostly struct notifier_block perf_event_nmi_notifier = { + .notifier_call = perf_event_nmi_handler, .next = NULL, .priority = 1 }; @@ -1840,8 +1969,8 @@ static struct x86_pmu p6_pmu = { .handle_irq = p6_pmu_handle_irq, .disable_all = p6_pmu_disable_all, .enable_all = p6_pmu_enable_all, - .enable = p6_pmu_enable_counter, - .disable = p6_pmu_disable_counter, + .enable = p6_pmu_enable_event, + .disable = p6_pmu_disable_event, .eventsel = MSR_P6_EVNTSEL0, .perfctr = MSR_P6_PERFCTR0, .event_map = p6_pmu_event_map, @@ -1850,16 +1979,17 @@ static struct x86_pmu p6_pmu = { .apic = 1, .max_period = (1ULL << 31) - 1, .version = 0, - .num_counters = 2, + .num_events = 2, /* - * Counters have 40 bits implemented. However they are designed such + * Events 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 32 bits only. + * effective width of a event for P6-like PMU is 32 bits only. * * See IA-32 Intel Architecture Software developer manual Vol 3B */ - .counter_bits = 32, - .counter_mask = (1ULL << 32) - 1, + .event_bits = 32, + .event_mask = (1ULL << 32) - 1, + .get_event_idx = intel_get_event_idx, }; static struct x86_pmu intel_pmu = { @@ -1867,8 +1997,8 @@ static struct x86_pmu intel_pmu = { .handle_irq = intel_pmu_handle_irq, .disable_all = intel_pmu_disable_all, .enable_all = intel_pmu_enable_all, - .enable = intel_pmu_enable_counter, - .disable = intel_pmu_disable_counter, + .enable = intel_pmu_enable_event, + .disable = intel_pmu_disable_event, .eventsel = MSR_ARCH_PERFMON_EVENTSEL0, .perfctr = MSR_ARCH_PERFMON_PERFCTR0, .event_map = intel_pmu_event_map, @@ -1878,11 +2008,12 @@ static struct x86_pmu intel_pmu = { /* * Intel PMCs cannot be accessed sanely above 32 bit width, * so we install an artificial 1<<31 period regardless of - * the generic counter period: + * the generic event period: */ .max_period = (1ULL << 31) - 1, .enable_bts = intel_pmu_enable_bts, .disable_bts = intel_pmu_disable_bts, + .get_event_idx = intel_get_event_idx, }; static struct x86_pmu amd_pmu = { @@ -1890,19 +2021,20 @@ static struct x86_pmu amd_pmu = { .handle_irq = amd_pmu_handle_irq, .disable_all = amd_pmu_disable_all, .enable_all = amd_pmu_enable_all, - .enable = amd_pmu_enable_counter, - .disable = amd_pmu_disable_counter, + .enable = amd_pmu_enable_event, + .disable = amd_pmu_disable_event, .eventsel = MSR_K7_EVNTSEL0, .perfctr = MSR_K7_PERFCTR0, .event_map = amd_pmu_event_map, .raw_event = amd_pmu_raw_event, .max_events = ARRAY_SIZE(amd_perfmon_event_map), - .num_counters = 4, - .counter_bits = 48, - .counter_mask = (1ULL << 48) - 1, + .num_events = 4, + .event_bits = 48, + .event_mask = (1ULL << 48) - 1, .apic = 1, /* use highest bit to detect overflow */ .max_period = (1ULL << 47) - 1, + .get_event_idx = gen_get_event_idx, }; static int p6_pmu_init(void) @@ -1915,10 +2047,12 @@ static int p6_pmu_init(void) case 7: case 8: case 11: /* Pentium III */ + event_constraints = intel_p6_event_constraints; break; case 9: case 13: /* Pentium M */ + event_constraints = intel_p6_event_constraints; break; default: pr_cont("unsupported p6 CPU model %d ", @@ -1956,7 +2090,7 @@ static int intel_pmu_init(void) /* * Check whether the Architectural PerfMon supports - * Branch Misses Retired Event or not. + * Branch Misses Retired hw_event or not. */ cpuid(10, &eax.full, &ebx, &unused, &edx.full); if (eax.split.mask_length <= ARCH_PERFMON_BRANCH_MISSES_RETIRED) @@ -1968,15 +2102,15 @@ static int intel_pmu_init(void) x86_pmu = intel_pmu; x86_pmu.version = version; - x86_pmu.num_counters = eax.split.num_counters; - x86_pmu.counter_bits = eax.split.bit_width; - x86_pmu.counter_mask = (1ULL << eax.split.bit_width) - 1; + x86_pmu.num_events = eax.split.num_events; + x86_pmu.event_bits = eax.split.bit_width; + x86_pmu.event_mask = (1ULL << eax.split.bit_width) - 1; /* - * Quirk: v2 perfmon does not report fixed-purpose counters, so - * assume at least 3 counters: + * Quirk: v2 perfmon does not report fixed-purpose events, so + * assume at least 3 events: */ - x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3); + x86_pmu.num_events_fixed = max((int)edx.split.num_events_fixed, 3); /* * Install the hw-cache-events table: @@ -1990,12 +2124,14 @@ static int intel_pmu_init(void) sizeof(hw_cache_event_ids)); pr_cont("Core2 events, "); + event_constraints = intel_core_event_constraints; break; default: case 26: memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids, sizeof(hw_cache_event_ids)); + event_constraints = intel_nehalem_event_constraints; pr_cont("Nehalem/Corei7 events, "); break; case 28: @@ -2023,11 +2159,11 @@ static int amd_pmu_init(void) return 0; } -void __init init_hw_perf_counters(void) +void __init init_hw_perf_events(void) { int err; - pr_info("Performance Counters: "); + pr_info("Performance Events: "); switch (boot_cpu_data.x86_vendor) { case X86_VENDOR_INTEL: @@ -2040,45 +2176,45 @@ void __init init_hw_perf_counters(void) return; } if (err != 0) { - pr_cont("no PMU driver, software counters only.\n"); + pr_cont("no PMU driver, software events only.\n"); return; } pr_cont("%s PMU driver.\n", x86_pmu.name); - if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { - WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!", - x86_pmu.num_counters, X86_PMC_MAX_GENERIC); - x86_pmu.num_counters = X86_PMC_MAX_GENERIC; + if (x86_pmu.num_events > X86_PMC_MAX_GENERIC) { + WARN(1, KERN_ERR "hw perf events %d > max(%d), clipping!", + x86_pmu.num_events, X86_PMC_MAX_GENERIC); + x86_pmu.num_events = X86_PMC_MAX_GENERIC; } - perf_counter_mask = (1 << x86_pmu.num_counters) - 1; - perf_max_counters = x86_pmu.num_counters; + perf_event_mask = (1 << x86_pmu.num_events) - 1; + perf_max_events = x86_pmu.num_events; - if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) { - WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!", - x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED); - x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED; + if (x86_pmu.num_events_fixed > X86_PMC_MAX_FIXED) { + WARN(1, KERN_ERR "hw perf events fixed %d > max(%d), clipping!", + x86_pmu.num_events_fixed, X86_PMC_MAX_FIXED); + x86_pmu.num_events_fixed = X86_PMC_MAX_FIXED; } - perf_counter_mask |= - ((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED; - x86_pmu.intel_ctrl = perf_counter_mask; + perf_event_mask |= + ((1LL << x86_pmu.num_events_fixed)-1) << X86_PMC_IDX_FIXED; + x86_pmu.intel_ctrl = perf_event_mask; - perf_counters_lapic_init(); - register_die_notifier(&perf_counter_nmi_notifier); + perf_events_lapic_init(); + register_die_notifier(&perf_event_nmi_notifier); - pr_info("... version: %d\n", x86_pmu.version); - pr_info("... bit width: %d\n", x86_pmu.counter_bits); - pr_info("... generic counters: %d\n", x86_pmu.num_counters); - pr_info("... value mask: %016Lx\n", x86_pmu.counter_mask); - pr_info("... max period: %016Lx\n", x86_pmu.max_period); - pr_info("... fixed-purpose counters: %d\n", x86_pmu.num_counters_fixed); - pr_info("... counter mask: %016Lx\n", perf_counter_mask); + pr_info("... version: %d\n", x86_pmu.version); + pr_info("... bit width: %d\n", x86_pmu.event_bits); + pr_info("... generic registers: %d\n", x86_pmu.num_events); + pr_info("... value mask: %016Lx\n", x86_pmu.event_mask); + pr_info("... max period: %016Lx\n", x86_pmu.max_period); + pr_info("... fixed-purpose events: %d\n", x86_pmu.num_events_fixed); + pr_info("... event mask: %016Lx\n", perf_event_mask); } -static inline void x86_pmu_read(struct perf_counter *counter) +static inline void x86_pmu_read(struct perf_event *event) { - x86_perf_counter_update(counter, &counter->hw, counter->hw.idx); + x86_perf_event_update(event, &event->hw, event->hw.idx); } static const struct pmu pmu = { @@ -2088,13 +2224,52 @@ static const struct pmu pmu = { .unthrottle = x86_pmu_unthrottle, }; -const struct pmu *hw_perf_counter_init(struct perf_counter *counter) +static int +validate_event(struct cpu_hw_events *cpuc, struct perf_event *event) +{ + struct hw_perf_event fake_event = event->hw; + + if (event->pmu != &pmu) + return 0; + + return x86_schedule_event(cpuc, &fake_event); +} + +static int validate_group(struct perf_event *event) +{ + struct perf_event *sibling, *leader = event->group_leader; + struct cpu_hw_events fake_pmu; + + memset(&fake_pmu, 0, sizeof(fake_pmu)); + + if (!validate_event(&fake_pmu, leader)) + return -ENOSPC; + + list_for_each_entry(sibling, &leader->sibling_list, group_entry) { + if (!validate_event(&fake_pmu, sibling)) + return -ENOSPC; + } + + if (!validate_event(&fake_pmu, event)) + return -ENOSPC; + + return 0; +} + +const struct pmu *hw_perf_event_init(struct perf_event *event) { int err; - err = __hw_perf_counter_init(counter); - if (err) + err = __hw_perf_event_init(event); + if (!err) { + if (event->group_leader != event) + err = validate_group(event); + } + if (err) { + if (event->destroy) + event->destroy(event); return ERR_PTR(err); + } return &pmu; } @@ -2110,8 +2285,8 @@ void callchain_store(struct perf_callchain_entry *entry, u64 ip) entry->ip[entry->nr++] = ip; } -static DEFINE_PER_CPU(struct perf_callchain_entry, irq_entry); -static DEFINE_PER_CPU(struct perf_callchain_entry, nmi_entry); +static DEFINE_PER_CPU(struct perf_callchain_entry, pmc_irq_entry); +static DEFINE_PER_CPU(struct perf_callchain_entry, pmc_nmi_entry); static DEFINE_PER_CPU(int, in_nmi_frame); @@ -2264,9 +2439,9 @@ struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) struct perf_callchain_entry *entry; if (in_nmi()) - entry = &__get_cpu_var(nmi_entry); + entry = &__get_cpu_var(pmc_nmi_entry); else - entry = &__get_cpu_var(irq_entry); + entry = &__get_cpu_var(pmc_irq_entry); entry->nr = 0; @@ -2275,7 +2450,7 @@ struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) return entry; } -void hw_perf_counter_setup_online(int cpu) +void hw_perf_event_setup_online(int cpu) { init_debug_store_on_cpu(cpu); } diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c index 392bea43b890..fab786f60ed6 100644 --- a/arch/x86/kernel/cpu/perfctr-watchdog.c +++ b/arch/x86/kernel/cpu/perfctr-watchdog.c @@ -20,7 +20,7 @@ #include <linux/kprobes.h> #include <asm/apic.h> -#include <asm/perf_counter.h> +#include <asm/perf_event.h> struct nmi_watchdog_ctlblk { unsigned int cccr_msr; diff --git a/arch/x86/kernel/cpu/sched.c b/arch/x86/kernel/cpu/sched.c new file mode 100644 index 000000000000..a640ae5ad201 --- /dev/null +++ b/arch/x86/kernel/cpu/sched.c @@ -0,0 +1,55 @@ +#include <linux/sched.h> +#include <linux/math64.h> +#include <linux/percpu.h> +#include <linux/irqflags.h> + +#include <asm/cpufeature.h> +#include <asm/processor.h> + +#ifdef CONFIG_SMP + +static DEFINE_PER_CPU(struct aperfmperf, old_perf_sched); + +static unsigned long scale_aperfmperf(void) +{ + struct aperfmperf val, *old = &__get_cpu_var(old_perf_sched); + unsigned long ratio, flags; + + local_irq_save(flags); + get_aperfmperf(&val); + local_irq_restore(flags); + + ratio = calc_aperfmperf_ratio(old, &val); + *old = val; + + return ratio; +} + +unsigned long arch_scale_freq_power(struct sched_domain *sd, int cpu) +{ + /* + * do aperf/mperf on the cpu level because it includes things + * like turbo mode, which are relevant to full cores. + */ + if (boot_cpu_has(X86_FEATURE_APERFMPERF)) + return scale_aperfmperf(); + + /* + * maybe have something cpufreq here + */ + + return default_scale_freq_power(sd, cpu); +} + +unsigned long arch_scale_smt_power(struct sched_domain *sd, int cpu) +{ + /* + * aperf/mperf already includes the smt gain + */ + if (boot_cpu_has(X86_FEATURE_APERFMPERF)) + return SCHED_LOAD_SCALE; + + return default_scale_smt_power(sd, cpu); +} + +#endif diff --git a/arch/x86/kernel/cpu/vmware.c b/arch/x86/kernel/cpu/vmware.c index bc24f514ec93..1cbed97b59cf 100644 --- a/arch/x86/kernel/cpu/vmware.c +++ b/arch/x86/kernel/cpu/vmware.c @@ -24,6 +24,7 @@ #include <linux/dmi.h> #include <asm/div64.h> #include <asm/vmware.h> +#include <asm/x86_init.h> #define CPUID_VMWARE_INFO_LEAF 0x40000000 #define VMWARE_HYPERVISOR_MAGIC 0x564D5868 @@ -47,21 +48,35 @@ static inline int __vmware_platform(void) return eax != (uint32_t)-1 && ebx == VMWARE_HYPERVISOR_MAGIC; } -static unsigned long __vmware_get_tsc_khz(void) +static unsigned long vmware_get_tsc_khz(void) { uint64_t tsc_hz; uint32_t eax, ebx, ecx, edx; VMWARE_PORT(GETHZ, eax, ebx, ecx, edx); - if (ebx == UINT_MAX) - return 0; tsc_hz = eax | (((uint64_t)ebx) << 32); do_div(tsc_hz, 1000); BUG_ON(tsc_hz >> 32); + printk(KERN_INFO "TSC freq read from hypervisor : %lu.%03lu MHz\n", + (unsigned long) tsc_hz / 1000, + (unsigned long) tsc_hz % 1000); return tsc_hz; } +void __init vmware_platform_setup(void) +{ + uint32_t eax, ebx, ecx, edx; + + VMWARE_PORT(GETHZ, eax, ebx, ecx, edx); + + if (ebx != UINT_MAX) + x86_platform.calibrate_tsc = vmware_get_tsc_khz; + else + printk(KERN_WARNING + "Failed to get TSC freq from the hypervisor\n"); +} + /* * While checking the dmi string infomation, just checking the product * serial key should be enough, as this will always have a VMware @@ -87,12 +102,6 @@ int vmware_platform(void) return 0; } -unsigned long vmware_get_tsc_khz(void) -{ - BUG_ON(!vmware_platform()); - return __vmware_get_tsc_khz(); -} - /* * VMware hypervisor takes care of exporting a reliable TSC to the guest. * Still, due to timing difference when running on virtual cpus, the TSC can |