diff options
Diffstat (limited to 'arch/x86/kernel/cpu')
21 files changed, 644 insertions, 328 deletions
diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile index 3a785da34b6f..3f0ebe429a01 100644 --- a/arch/x86/kernel/cpu/Makefile +++ b/arch/x86/kernel/cpu/Makefile @@ -12,11 +12,11 @@ endif nostackp := $(call cc-option, -fno-stack-protector) CFLAGS_common.o := $(nostackp) -obj-y := intel_cacheinfo.o addon_cpuid_features.o +obj-y := intel_cacheinfo.o scattered.o topology.o obj-y += proc.o capflags.o powerflags.o common.o obj-y += vmware.o hypervisor.o sched.o mshyperv.o -obj-$(CONFIG_X86_32) += bugs.o cmpxchg.o +obj-$(CONFIG_X86_32) += bugs.o obj-$(CONFIG_X86_64) += bugs_64.o obj-$(CONFIG_CPU_SUP_INTEL) += intel.o diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index e485825130d2..60a57b13082d 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -466,7 +466,7 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) } } - if (c->x86 == 0x10 || c->x86 == 0x11) + if (c->x86 >= 0x10) set_cpu_cap(c, X86_FEATURE_REP_GOOD); /* get apicid instead of initial apic id from cpuid */ @@ -529,7 +529,7 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) num_cache_leaves = 3; } - if (c->x86 >= 0xf && c->x86 <= 0x11) + if (c->x86 >= 0xf) set_cpu_cap(c, X86_FEATURE_K8); if (cpu_has_xmm2) { @@ -546,7 +546,7 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) fam10h_check_enable_mmcfg(); } - if (c == &boot_cpu_data && c->x86 >= 0xf && c->x86 <= 0x11) { + if (c == &boot_cpu_data && c->x86 >= 0xf) { unsigned long long tseg; /* @@ -609,3 +609,74 @@ static const struct cpu_dev __cpuinitconst amd_cpu_dev = { }; cpu_dev_register(amd_cpu_dev); + +/* + * AMD errata checking + * + * Errata are defined as arrays of ints using the AMD_LEGACY_ERRATUM() or + * AMD_OSVW_ERRATUM() macros. The latter is intended for newer errata that + * have an OSVW id assigned, which it takes as first argument. Both take a + * variable number of family-specific model-stepping ranges created by + * AMD_MODEL_RANGE(). Each erratum also has to be declared as extern const + * int[] in arch/x86/include/asm/processor.h. + * + * Example: + * + * const int amd_erratum_319[] = + * AMD_LEGACY_ERRATUM(AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0x4, 0x2), + * AMD_MODEL_RANGE(0x10, 0x8, 0x0, 0x8, 0x0), + * AMD_MODEL_RANGE(0x10, 0x9, 0x0, 0x9, 0x0)); + */ + +const int amd_erratum_400[] = + AMD_OSVW_ERRATUM(1, AMD_MODEL_RANGE(0xf, 0x41, 0x2, 0xff, 0xf), + AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0xff, 0xf)); +EXPORT_SYMBOL_GPL(amd_erratum_400); + +const int amd_erratum_383[] = + AMD_OSVW_ERRATUM(3, AMD_MODEL_RANGE(0x10, 0, 0, 0xff, 0xf)); +EXPORT_SYMBOL_GPL(amd_erratum_383); + +bool cpu_has_amd_erratum(const int *erratum) +{ + struct cpuinfo_x86 *cpu = ¤t_cpu_data; + int osvw_id = *erratum++; + u32 range; + u32 ms; + + /* + * If called early enough that current_cpu_data hasn't been initialized + * yet, fall back to boot_cpu_data. + */ + if (cpu->x86 == 0) + cpu = &boot_cpu_data; + + if (cpu->x86_vendor != X86_VENDOR_AMD) + return false; + + if (osvw_id >= 0 && osvw_id < 65536 && + cpu_has(cpu, X86_FEATURE_OSVW)) { + u64 osvw_len; + + rdmsrl(MSR_AMD64_OSVW_ID_LENGTH, osvw_len); + if (osvw_id < osvw_len) { + u64 osvw_bits; + + rdmsrl(MSR_AMD64_OSVW_STATUS + (osvw_id >> 6), + osvw_bits); + return osvw_bits & (1ULL << (osvw_id & 0x3f)); + } + } + + /* OSVW unavailable or ID unknown, match family-model-stepping range */ + ms = (cpu->x86_model << 8) | cpu->x86_mask; + while ((range = *erratum++)) + if ((cpu->x86 == AMD_MODEL_RANGE_FAMILY(range)) && + (ms >= AMD_MODEL_RANGE_START(range)) && + (ms <= AMD_MODEL_RANGE_END(range))) + return true; + + return false; +} + +EXPORT_SYMBOL_GPL(cpu_has_amd_erratum); diff --git a/arch/x86/kernel/cpu/cmpxchg.c b/arch/x86/kernel/cpu/cmpxchg.c deleted file mode 100644 index 2056ccf572cc..000000000000 --- a/arch/x86/kernel/cpu/cmpxchg.c +++ /dev/null @@ -1,72 +0,0 @@ -/* - * cmpxchg*() fallbacks for CPU not supporting these instructions - */ - -#include <linux/kernel.h> -#include <linux/smp.h> -#include <linux/module.h> - -#ifndef CONFIG_X86_CMPXCHG -unsigned long cmpxchg_386_u8(volatile void *ptr, u8 old, u8 new) -{ - u8 prev; - unsigned long flags; - - /* Poor man's cmpxchg for 386. Unsuitable for SMP */ - local_irq_save(flags); - prev = *(u8 *)ptr; - if (prev == old) - *(u8 *)ptr = new; - local_irq_restore(flags); - return prev; -} -EXPORT_SYMBOL(cmpxchg_386_u8); - -unsigned long cmpxchg_386_u16(volatile void *ptr, u16 old, u16 new) -{ - u16 prev; - unsigned long flags; - - /* Poor man's cmpxchg for 386. Unsuitable for SMP */ - local_irq_save(flags); - prev = *(u16 *)ptr; - if (prev == old) - *(u16 *)ptr = new; - local_irq_restore(flags); - return prev; -} -EXPORT_SYMBOL(cmpxchg_386_u16); - -unsigned long cmpxchg_386_u32(volatile void *ptr, u32 old, u32 new) -{ - u32 prev; - unsigned long flags; - - /* Poor man's cmpxchg for 386. Unsuitable for SMP */ - local_irq_save(flags); - prev = *(u32 *)ptr; - if (prev == old) - *(u32 *)ptr = new; - local_irq_restore(flags); - return prev; -} -EXPORT_SYMBOL(cmpxchg_386_u32); -#endif - -#ifndef CONFIG_X86_CMPXCHG64 -unsigned long long cmpxchg_486_u64(volatile void *ptr, u64 old, u64 new) -{ - u64 prev; - unsigned long flags; - - /* Poor man's cmpxchg8b for 386 and 486. Unsuitable for SMP */ - local_irq_save(flags); - prev = *(u64 *)ptr; - if (prev == old) - *(u64 *)ptr = new; - local_irq_restore(flags); - return prev; -} -EXPORT_SYMBOL(cmpxchg_486_u64); -#endif - diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 68e4a6f2211e..490dac63c2d2 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -140,10 +140,18 @@ EXPORT_PER_CPU_SYMBOL_GPL(gdt_page); static int __init x86_xsave_setup(char *s) { setup_clear_cpu_cap(X86_FEATURE_XSAVE); + setup_clear_cpu_cap(X86_FEATURE_XSAVEOPT); return 1; } __setup("noxsave", x86_xsave_setup); +static int __init x86_xsaveopt_setup(char *s) +{ + setup_clear_cpu_cap(X86_FEATURE_XSAVEOPT); + return 1; +} +__setup("noxsaveopt", x86_xsaveopt_setup); + #ifdef CONFIG_X86_32 static int cachesize_override __cpuinitdata = -1; static int disable_x86_serial_nr __cpuinitdata = 1; @@ -551,6 +559,16 @@ static void __cpuinit get_cpu_cap(struct cpuinfo_x86 *c) c->x86_capability[4] = excap; } + /* Additional Intel-defined flags: level 0x00000007 */ + if (c->cpuid_level >= 0x00000007) { + u32 eax, ebx, ecx, edx; + + cpuid_count(0x00000007, 0, &eax, &ebx, &ecx, &edx); + + if (eax > 0) + c->x86_capability[9] = ebx; + } + /* AMD-defined flags: level 0x80000001 */ xlvl = cpuid_eax(0x80000000); c->extended_cpuid_level = xlvl; @@ -576,6 +594,7 @@ static void __cpuinit get_cpu_cap(struct cpuinfo_x86 *c) if (c->extended_cpuid_level >= 0x80000007) c->x86_power = cpuid_edx(0x80000007); + init_scattered_cpuid_features(c); } static void __cpuinit identify_cpu_without_cpuid(struct cpuinfo_x86 *c) @@ -731,7 +750,6 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 *c) get_model_name(c); /* Default name */ - init_scattered_cpuid_features(c); detect_nopl(c); } @@ -1192,6 +1210,7 @@ void __cpuinit cpu_init(void) dbg_restore_debug_regs(); fpu_init(); + xsave_init(); raw_local_save_flags(kernel_eflags); @@ -1252,12 +1271,7 @@ void __cpuinit cpu_init(void) clear_used_math(); mxcsr_feature_mask_init(); - /* - * Boot processor to setup the FP and extended state context info. - */ - if (smp_processor_id() == boot_cpu_id) - init_thread_xstate(); - + fpu_init(); xsave_init(); } #endif diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c index 246cd3afbb5f..cd8da247dda1 100644 --- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c +++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c @@ -72,7 +72,7 @@ struct acpi_cpufreq_data { static DEFINE_PER_CPU(struct acpi_cpufreq_data *, acfreq_data); /* acpi_perf_data is a pointer to percpu data. */ -static struct acpi_processor_performance *acpi_perf_data; +static struct acpi_processor_performance __percpu *acpi_perf_data; static struct cpufreq_driver acpi_cpufreq_driver; diff --git a/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c index a36de5bbb622..994230d4dc4e 100644 --- a/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c +++ b/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c @@ -110,7 +110,7 @@ struct pcc_cpu { u32 output_offset; }; -static struct pcc_cpu *pcc_cpu_info; +static struct pcc_cpu __percpu *pcc_cpu_info; static int pcc_cpufreq_verify(struct cpufreq_policy *policy) { diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c index dd531cc56a8f..8095f8611f8a 100644 --- a/arch/x86/kernel/cpu/hypervisor.c +++ b/arch/x86/kernel/cpu/hypervisor.c @@ -34,6 +34,9 @@ static const __initconst struct hypervisor_x86 * const hypervisors[] = { &x86_hyper_vmware, &x86_hyper_ms_hyperv, +#ifdef CONFIG_XEN_PVHVM + &x86_hyper_xen_hvm, +#endif }; const struct hypervisor_x86 *x86_hyper; diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c index 33eae2062cf5..898c2f4eab88 100644 --- a/arch/x86/kernel/cpu/intel_cacheinfo.c +++ b/arch/x86/kernel/cpu/intel_cacheinfo.c @@ -347,8 +347,8 @@ static struct amd_l3_cache * __cpuinit amd_init_l3_cache(int node) return l3; } -static void __cpuinit -amd_check_l3_disable(int index, struct _cpuid4_info_regs *this_leaf) +static void __cpuinit amd_check_l3_disable(struct _cpuid4_info_regs *this_leaf, + int index) { int node; @@ -396,20 +396,39 @@ amd_check_l3_disable(int index, struct _cpuid4_info_regs *this_leaf) this_leaf->l3 = l3_caches[node]; } +/* + * check whether a slot used for disabling an L3 index is occupied. + * @l3: L3 cache descriptor + * @slot: slot number (0..1) + * + * @returns: the disabled index if used or negative value if slot free. + */ +int amd_get_l3_disable_slot(struct amd_l3_cache *l3, unsigned slot) +{ + unsigned int reg = 0; + + pci_read_config_dword(l3->dev, 0x1BC + slot * 4, ®); + + /* check whether this slot is activated already */ + if (reg & (3UL << 30)) + return reg & 0xfff; + + return -1; +} + static ssize_t show_cache_disable(struct _cpuid4_info *this_leaf, char *buf, unsigned int slot) { - struct pci_dev *dev = this_leaf->l3->dev; - unsigned int reg = 0; + int index; if (!this_leaf->l3 || !this_leaf->l3->can_disable) return -EINVAL; - if (!dev) - return -EINVAL; + index = amd_get_l3_disable_slot(this_leaf->l3, slot); + if (index >= 0) + return sprintf(buf, "%d\n", index); - pci_read_config_dword(dev, 0x1BC + slot * 4, ®); - return sprintf(buf, "0x%08x\n", reg); + return sprintf(buf, "FREE\n"); } #define SHOW_CACHE_DISABLE(slot) \ @@ -451,37 +470,74 @@ static void amd_l3_disable_index(struct amd_l3_cache *l3, int cpu, } } - -static ssize_t store_cache_disable(struct _cpuid4_info *this_leaf, - const char *buf, size_t count, - unsigned int slot) +/* + * disable a L3 cache index by using a disable-slot + * + * @l3: L3 cache descriptor + * @cpu: A CPU on the node containing the L3 cache + * @slot: slot number (0..1) + * @index: index to disable + * + * @return: 0 on success, error status on failure + */ +int amd_set_l3_disable_slot(struct amd_l3_cache *l3, int cpu, unsigned slot, + unsigned long index) { - struct pci_dev *dev = this_leaf->l3->dev; - int cpu = cpumask_first(to_cpumask(this_leaf->shared_cpu_map)); - unsigned long val = 0; + int ret = 0; #define SUBCACHE_MASK (3UL << 20) #define SUBCACHE_INDEX 0xfff - if (!this_leaf->l3 || !this_leaf->l3->can_disable) + /* + * check whether this slot is already used or + * the index is already disabled + */ + ret = amd_get_l3_disable_slot(l3, slot); + if (ret >= 0) return -EINVAL; + /* + * check whether the other slot has disabled the + * same index already + */ + if (index == amd_get_l3_disable_slot(l3, !slot)) + return -EINVAL; + + /* do not allow writes outside of allowed bits */ + if ((index & ~(SUBCACHE_MASK | SUBCACHE_INDEX)) || + ((index & SUBCACHE_INDEX) > l3->indices)) + return -EINVAL; + + amd_l3_disable_index(l3, cpu, slot, index); + + return 0; +} + +static ssize_t store_cache_disable(struct _cpuid4_info *this_leaf, + const char *buf, size_t count, + unsigned int slot) +{ + unsigned long val = 0; + int cpu, err = 0; + if (!capable(CAP_SYS_ADMIN)) return -EPERM; - if (!dev) + if (!this_leaf->l3 || !this_leaf->l3->can_disable) return -EINVAL; - if (strict_strtoul(buf, 10, &val) < 0) - return -EINVAL; + cpu = cpumask_first(to_cpumask(this_leaf->shared_cpu_map)); - /* do not allow writes outside of allowed bits */ - if ((val & ~(SUBCACHE_MASK | SUBCACHE_INDEX)) || - ((val & SUBCACHE_INDEX) > this_leaf->l3->indices)) + if (strict_strtoul(buf, 10, &val) < 0) return -EINVAL; - amd_l3_disable_index(this_leaf->l3, cpu, slot, val); - + err = amd_set_l3_disable_slot(this_leaf->l3, cpu, slot, val); + if (err) { + if (err == -EEXIST) + printk(KERN_WARNING "L3 disable slot %d in use!\n", + slot); + return err; + } return count; } @@ -502,7 +558,7 @@ static struct _cache_attr cache_disable_1 = __ATTR(cache_disable_1, 0644, #else /* CONFIG_CPU_SUP_AMD */ static void __cpuinit -amd_check_l3_disable(int index, struct _cpuid4_info_regs *this_leaf) +amd_check_l3_disable(struct _cpuid4_info_regs *this_leaf, int index) { }; #endif /* CONFIG_CPU_SUP_AMD */ @@ -518,7 +574,7 @@ __cpuinit cpuid4_cache_lookup_regs(int index, if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) { amd_cpuid4(index, &eax, &ebx, &ecx); - amd_check_l3_disable(index, this_leaf); + amd_check_l3_disable(this_leaf, index); } else { cpuid_count(4, index, &eax.full, &ebx.full, &ecx.full, &edx); } diff --git a/arch/x86/kernel/cpu/mcheck/mce-apei.c b/arch/x86/kernel/cpu/mcheck/mce-apei.c index 745b54f9be89..8209472b27a5 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-apei.c +++ b/arch/x86/kernel/cpu/mcheck/mce-apei.c @@ -80,7 +80,7 @@ int apei_write_mce(struct mce *m) rcd.hdr.revision = CPER_RECORD_REV; rcd.hdr.signature_end = CPER_SIG_END; rcd.hdr.section_count = 1; - rcd.hdr.error_severity = CPER_SER_FATAL; + rcd.hdr.error_severity = CPER_SEV_FATAL; /* timestamp, platform_id, partition_id are all invalid */ rcd.hdr.validation_bits = 0; rcd.hdr.record_length = sizeof(rcd); @@ -96,7 +96,7 @@ int apei_write_mce(struct mce *m) rcd.sec_hdr.validation_bits = 0; rcd.sec_hdr.flags = CPER_SEC_PRIMARY; rcd.sec_hdr.section_type = CPER_SECTION_TYPE_MCE; - rcd.sec_hdr.section_severity = CPER_SER_FATAL; + rcd.sec_hdr.section_severity = CPER_SEV_FATAL; memcpy(&rcd.mce, m, sizeof(*m)); diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c index 1970ef911c99..ed41562909fe 100644 --- a/arch/x86/kernel/cpu/mcheck/mce.c +++ b/arch/x86/kernel/cpu/mcheck/mce.c @@ -51,7 +51,7 @@ static DEFINE_MUTEX(mce_read_mutex); #define rcu_dereference_check_mce(p) \ - rcu_dereference_check((p), \ + rcu_dereference_index_check((p), \ rcu_read_lock_sched_held() || \ lockdep_is_held(&mce_read_mutex)) @@ -107,8 +107,8 @@ EXPORT_SYMBOL_GPL(x86_mce_decoder_chain); static int default_decode_mce(struct notifier_block *nb, unsigned long val, void *data) { - pr_emerg("No human readable MCE decoding support on this CPU type.\n"); - pr_emerg("Run the message through 'mcelog --ascii' to decode.\n"); + pr_emerg(HW_ERR "No human readable MCE decoding support on this CPU type.\n"); + pr_emerg(HW_ERR "Run the message through 'mcelog --ascii' to decode.\n"); return NOTIFY_STOP; } @@ -211,11 +211,11 @@ void mce_log(struct mce *mce) static void print_mce(struct mce *m) { - pr_emerg("CPU %d: Machine Check Exception: %16Lx Bank %d: %016Lx\n", + pr_emerg(HW_ERR "CPU %d: Machine Check Exception: %Lx Bank %d: %016Lx\n", m->extcpu, m->mcgstatus, m->bank, m->status); if (m->ip) { - pr_emerg("RIP%s %02x:<%016Lx> ", + pr_emerg(HW_ERR "RIP%s %02x:<%016Lx> ", !(m->mcgstatus & MCG_STATUS_EIPV) ? " !INEXACT!" : "", m->cs, m->ip); @@ -224,14 +224,14 @@ static void print_mce(struct mce *m) pr_cont("\n"); } - pr_emerg("TSC %llx ", m->tsc); + pr_emerg(HW_ERR "TSC %llx ", m->tsc); if (m->addr) pr_cont("ADDR %llx ", m->addr); if (m->misc) pr_cont("MISC %llx ", m->misc); pr_cont("\n"); - pr_emerg("PROCESSOR %u:%x TIME %llu SOCKET %u APIC %x\n", + pr_emerg(HW_ERR "PROCESSOR %u:%x TIME %llu SOCKET %u APIC %x\n", m->cpuvendor, m->cpuid, m->time, m->socketid, m->apicid); /* @@ -241,16 +241,6 @@ static void print_mce(struct mce *m) atomic_notifier_call_chain(&x86_mce_decoder_chain, 0, m); } -static void print_mce_head(void) -{ - pr_emerg("\nHARDWARE ERROR\n"); -} - -static void print_mce_tail(void) -{ - pr_emerg("This is not a software problem!\n"); -} - #define PANIC_TIMEOUT 5 /* 5 seconds */ static atomic_t mce_paniced; @@ -291,7 +281,6 @@ static void mce_panic(char *msg, struct mce *final, char *exp) 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++) { struct mce *m = &mcelog.entry[i]; @@ -322,16 +311,15 @@ static void mce_panic(char *msg, struct mce *final, char *exp) apei_err = apei_write_mce(final); } if (cpu_missing) - printk(KERN_EMERG "Some CPUs didn't answer in synchronization\n"); - print_mce_tail(); + pr_emerg(HW_ERR "Some CPUs didn't answer in synchronization\n"); if (exp) - printk(KERN_EMERG "Machine check: %s\n", exp); + pr_emerg(HW_ERR "Machine check: %s\n", exp); if (!fake_panic) { if (panic_timeout == 0) panic_timeout = mce_panic_timeout; panic(msg); } else - printk(KERN_EMERG "Fake kernel panic: %s\n", msg); + pr_emerg(HW_ERR "Fake kernel panic: %s\n", msg); } /* Support code for software error injection */ @@ -1221,7 +1209,7 @@ int mce_notify_irq(void) schedule_work(&mce_trigger_work); if (__ratelimit(&ratelimit)) - printk(KERN_INFO "Machine check events logged\n"); + pr_info(HW_ERR "Machine check events logged\n"); return 1; } diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel.c b/arch/x86/kernel/cpu/mcheck/mce_intel.c index 62b48e40920a..6fcd0936194f 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_intel.c +++ b/arch/x86/kernel/cpu/mcheck/mce_intel.c @@ -95,19 +95,20 @@ static void cmci_discover(int banks, int boot) rdmsrl(MSR_IA32_MCx_CTL2(i), val); /* Already owned by someone else? */ - if (val & CMCI_EN) { + if (val & MCI_CTL2_CMCI_EN) { if (test_and_clear_bit(i, owned) && !boot) print_update("SHD", &hdr, i); __clear_bit(i, __get_cpu_var(mce_poll_banks)); continue; } - val |= CMCI_EN | CMCI_THRESHOLD; + val &= ~MCI_CTL2_CMCI_THRESHOLD_MASK; + val |= MCI_CTL2_CMCI_EN | CMCI_THRESHOLD; 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) { + if (val & MCI_CTL2_CMCI_EN) { if (!test_and_set_bit(i, owned) && !boot) print_update("CMCI", &hdr, i); __clear_bit(i, __get_cpu_var(mce_poll_banks)); @@ -155,7 +156,7 @@ void cmci_clear(void) continue; /* Disable CMCI */ rdmsrl(MSR_IA32_MCx_CTL2(i), val); - val &= ~(CMCI_EN|CMCI_THRESHOLD_MASK); + val &= ~(MCI_CTL2_CMCI_EN|MCI_CTL2_CMCI_THRESHOLD_MASK); wrmsrl(MSR_IA32_MCx_CTL2(i), val); __clear_bit(i, __get_cpu_var(mce_banks_owned)); } diff --git a/arch/x86/kernel/cpu/mcheck/therm_throt.c b/arch/x86/kernel/cpu/mcheck/therm_throt.c index e1a0a3bf9716..c2a8b26d4fea 100644 --- a/arch/x86/kernel/cpu/mcheck/therm_throt.c +++ b/arch/x86/kernel/cpu/mcheck/therm_throt.c @@ -34,15 +34,25 @@ /* How long to wait between reporting thermal events */ #define CHECK_INTERVAL (300 * HZ) +#define THERMAL_THROTTLING_EVENT 0 +#define POWER_LIMIT_EVENT 1 + /* - * Current thermal throttling state: + * Current thermal event state: */ -struct thermal_state { - bool is_throttled; - +struct _thermal_state { + bool new_event; + int event; u64 next_check; - unsigned long throttle_count; - unsigned long last_throttle_count; + unsigned long count; + unsigned long last_count; +}; + +struct thermal_state { + struct _thermal_state core_throttle; + struct _thermal_state core_power_limit; + struct _thermal_state package_throttle; + struct _thermal_state package_power_limit; }; static DEFINE_PER_CPU(struct thermal_state, thermal_state); @@ -53,11 +63,13 @@ static u32 lvtthmr_init __read_mostly; #ifdef CONFIG_SYSFS #define define_therm_throt_sysdev_one_ro(_name) \ - static SYSDEV_ATTR(_name, 0444, therm_throt_sysdev_show_##_name, NULL) + static SYSDEV_ATTR(_name, 0444, \ + therm_throt_sysdev_show_##_name, \ + NULL) \ -#define define_therm_throt_sysdev_show_func(name) \ +#define define_therm_throt_sysdev_show_func(event, name) \ \ -static ssize_t therm_throt_sysdev_show_##name( \ +static ssize_t therm_throt_sysdev_show_##event##_##name( \ struct sys_device *dev, \ struct sysdev_attribute *attr, \ char *buf) \ @@ -66,30 +78,42 @@ static ssize_t therm_throt_sysdev_show_##name( \ ssize_t ret; \ \ preempt_disable(); /* CPU hotplug */ \ - if (cpu_online(cpu)) \ + if (cpu_online(cpu)) { \ ret = sprintf(buf, "%lu\n", \ - per_cpu(thermal_state, cpu).name); \ - else \ + per_cpu(thermal_state, cpu).event.name); \ + } else \ ret = 0; \ preempt_enable(); \ \ return ret; \ } -define_therm_throt_sysdev_show_func(throttle_count); -define_therm_throt_sysdev_one_ro(throttle_count); +define_therm_throt_sysdev_show_func(core_throttle, count); +define_therm_throt_sysdev_one_ro(core_throttle_count); + +define_therm_throt_sysdev_show_func(core_power_limit, count); +define_therm_throt_sysdev_one_ro(core_power_limit_count); + +define_therm_throt_sysdev_show_func(package_throttle, count); +define_therm_throt_sysdev_one_ro(package_throttle_count); + +define_therm_throt_sysdev_show_func(package_power_limit, count); +define_therm_throt_sysdev_one_ro(package_power_limit_count); static struct attribute *thermal_throttle_attrs[] = { - &attr_throttle_count.attr, + &attr_core_throttle_count.attr, NULL }; -static struct attribute_group thermal_throttle_attr_group = { +static struct attribute_group thermal_attr_group = { .attrs = thermal_throttle_attrs, .name = "thermal_throttle" }; #endif /* CONFIG_SYSFS */ +#define CORE_LEVEL 0 +#define PACKAGE_LEVEL 1 + /*** * therm_throt_process - Process thermal throttling event from interrupt * @curr: Whether the condition is current or not (boolean), since the @@ -106,39 +130,70 @@ 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(bool is_throttled) +static int therm_throt_process(bool new_event, int event, int level) { - struct thermal_state *state; - unsigned int this_cpu; - bool was_throttled; + struct _thermal_state *state; + unsigned int this_cpu = smp_processor_id(); + bool old_event; u64 now; + struct thermal_state *pstate = &per_cpu(thermal_state, this_cpu); - this_cpu = smp_processor_id(); now = get_jiffies_64(); - state = &per_cpu(thermal_state, this_cpu); + if (level == CORE_LEVEL) { + if (event == THERMAL_THROTTLING_EVENT) + state = &pstate->core_throttle; + else if (event == POWER_LIMIT_EVENT) + state = &pstate->core_power_limit; + else + return 0; + } else if (level == PACKAGE_LEVEL) { + if (event == THERMAL_THROTTLING_EVENT) + state = &pstate->package_throttle; + else if (event == POWER_LIMIT_EVENT) + state = &pstate->package_power_limit; + else + return 0; + } else + return 0; - was_throttled = state->is_throttled; - state->is_throttled = is_throttled; + old_event = state->new_event; + state->new_event = new_event; - if (is_throttled) - state->throttle_count++; + if (new_event) + state->count++; if (time_before64(now, state->next_check) && - state->throttle_count != state->last_throttle_count) + state->count != state->last_count) return 0; state->next_check = now + CHECK_INTERVAL; - state->last_throttle_count = state->throttle_count; + state->last_count = state->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", this_cpu, state->throttle_count); + if (new_event) { + if (event == THERMAL_THROTTLING_EVENT) + printk(KERN_CRIT "CPU%d: %s temperature above threshold, cpu clock throttled (total events = %lu)\n", + this_cpu, + level == CORE_LEVEL ? "Core" : "Package", + state->count); + else + printk(KERN_CRIT "CPU%d: %s power limit notification (total events = %lu)\n", + this_cpu, + level == CORE_LEVEL ? "Core" : "Package", + state->count); add_taint(TAINT_MACHINE_CHECK); return 1; } - if (was_throttled) { - printk(KERN_INFO "CPU%d: Temperature/speed normal\n", this_cpu); + if (old_event) { + if (event == THERMAL_THROTTLING_EVENT) + printk(KERN_INFO "CPU%d: %s temperature/speed normal\n", + this_cpu, + level == CORE_LEVEL ? "Core" : "Package"); + else + printk(KERN_INFO "CPU%d: %s power limit normal\n", + this_cpu, + level == CORE_LEVEL ? "Core" : "Package"); return 1; } @@ -149,13 +204,32 @@ static int therm_throt_process(bool is_throttled) /* Add/Remove thermal_throttle interface for CPU device: */ static __cpuinit int thermal_throttle_add_dev(struct sys_device *sys_dev) { - return sysfs_create_group(&sys_dev->kobj, - &thermal_throttle_attr_group); + int err; + struct cpuinfo_x86 *c = &cpu_data(smp_processor_id()); + + err = sysfs_create_group(&sys_dev->kobj, &thermal_attr_group); + if (err) + return err; + + if (cpu_has(c, X86_FEATURE_PLN)) + err = sysfs_add_file_to_group(&sys_dev->kobj, + &attr_core_power_limit_count.attr, + thermal_attr_group.name); + if (cpu_has(c, X86_FEATURE_PTS)) + err = sysfs_add_file_to_group(&sys_dev->kobj, + &attr_package_throttle_count.attr, + thermal_attr_group.name); + if (cpu_has(c, X86_FEATURE_PLN)) + err = sysfs_add_file_to_group(&sys_dev->kobj, + &attr_package_power_limit_count.attr, + thermal_attr_group.name); + + return err; } static __cpuinit void thermal_throttle_remove_dev(struct sys_device *sys_dev) { - sysfs_remove_group(&sys_dev->kobj, &thermal_throttle_attr_group); + sysfs_remove_group(&sys_dev->kobj, &thermal_attr_group); } /* Mutex protecting device creation against CPU hotplug: */ @@ -226,14 +300,50 @@ device_initcall(thermal_throttle_init_device); #endif /* CONFIG_SYSFS */ +/* + * Set up the most two significant bit to notify mce log that this thermal + * event type. + * This is a temp solution. May be changed in the future with mce log + * infrasture. + */ +#define CORE_THROTTLED (0) +#define CORE_POWER_LIMIT ((__u64)1 << 62) +#define PACKAGE_THROTTLED ((__u64)2 << 62) +#define PACKAGE_POWER_LIMIT ((__u64)3 << 62) + /* Thermal transition interrupt handler */ static void intel_thermal_interrupt(void) { __u64 msr_val; + struct cpuinfo_x86 *c = &cpu_data(smp_processor_id()); rdmsrl(MSR_IA32_THERM_STATUS, msr_val); - if (therm_throt_process((msr_val & THERM_STATUS_PROCHOT) != 0)) - mce_log_therm_throt_event(msr_val); + + if (therm_throt_process(msr_val & THERM_STATUS_PROCHOT, + THERMAL_THROTTLING_EVENT, + CORE_LEVEL) != 0) + mce_log_therm_throt_event(CORE_THROTTLED | msr_val); + + if (cpu_has(c, X86_FEATURE_PLN)) + if (therm_throt_process(msr_val & THERM_STATUS_POWER_LIMIT, + POWER_LIMIT_EVENT, + CORE_LEVEL) != 0) + mce_log_therm_throt_event(CORE_POWER_LIMIT | msr_val); + + if (cpu_has(c, X86_FEATURE_PTS)) { + rdmsrl(MSR_IA32_PACKAGE_THERM_STATUS, msr_val); + if (therm_throt_process(msr_val & PACKAGE_THERM_STATUS_PROCHOT, + THERMAL_THROTTLING_EVENT, + PACKAGE_LEVEL) != 0) + mce_log_therm_throt_event(PACKAGE_THROTTLED | msr_val); + if (cpu_has(c, X86_FEATURE_PLN)) + if (therm_throt_process(msr_val & + PACKAGE_THERM_STATUS_POWER_LIMIT, + POWER_LIMIT_EVENT, + PACKAGE_LEVEL) != 0) + mce_log_therm_throt_event(PACKAGE_POWER_LIMIT + | msr_val); + } } static void unexpected_thermal_interrupt(void) @@ -335,8 +445,26 @@ void intel_init_thermal(struct cpuinfo_x86 *c) apic_write(APIC_LVTTHMR, h); rdmsr(MSR_IA32_THERM_INTERRUPT, l, h); - wrmsr(MSR_IA32_THERM_INTERRUPT, - l | (THERM_INT_LOW_ENABLE | THERM_INT_HIGH_ENABLE), h); + if (cpu_has(c, X86_FEATURE_PLN)) + wrmsr(MSR_IA32_THERM_INTERRUPT, + l | (THERM_INT_LOW_ENABLE + | THERM_INT_HIGH_ENABLE | THERM_INT_PLN_ENABLE), h); + else + wrmsr(MSR_IA32_THERM_INTERRUPT, + l | (THERM_INT_LOW_ENABLE | THERM_INT_HIGH_ENABLE), h); + + if (cpu_has(c, X86_FEATURE_PTS)) { + rdmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, l, h); + if (cpu_has(c, X86_FEATURE_PLN)) + wrmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, + l | (PACKAGE_THERM_INT_LOW_ENABLE + | PACKAGE_THERM_INT_HIGH_ENABLE + | PACKAGE_THERM_INT_PLN_ENABLE), h); + else + wrmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, + l | (PACKAGE_THERM_INT_LOW_ENABLE + | PACKAGE_THERM_INT_HIGH_ENABLE), h); + } smp_thermal_vector = intel_thermal_interrupt; diff --git a/arch/x86/kernel/cpu/mshyperv.c b/arch/x86/kernel/cpu/mshyperv.c index 16f41bbe46b6..d944bf6c50e9 100644 --- a/arch/x86/kernel/cpu/mshyperv.c +++ b/arch/x86/kernel/cpu/mshyperv.c @@ -18,6 +18,7 @@ #include <asm/mshyperv.h> struct ms_hyperv_info ms_hyperv; +EXPORT_SYMBOL_GPL(ms_hyperv); static bool __init ms_hyperv_platform(void) { diff --git a/arch/x86/kernel/cpu/mtrr/cleanup.c b/arch/x86/kernel/cpu/mtrr/cleanup.c index 06130b52f012..c5f59d071425 100644 --- a/arch/x86/kernel/cpu/mtrr/cleanup.c +++ b/arch/x86/kernel/cpu/mtrr/cleanup.c @@ -632,9 +632,9 @@ static void __init mtrr_print_out_one_result(int i) unsigned long gran_base, chunk_base, lose_base; char gran_factor, chunk_factor, lose_factor; - gran_base = to_size_factor(result[i].gran_sizek, &gran_factor), - chunk_base = to_size_factor(result[i].chunk_sizek, &chunk_factor), - lose_base = to_size_factor(result[i].lose_cover_sizek, &lose_factor), + gran_base = to_size_factor(result[i].gran_sizek, &gran_factor); + chunk_base = to_size_factor(result[i].chunk_sizek, &chunk_factor); + lose_base = to_size_factor(result[i].lose_cover_sizek, &lose_factor); pr_info("%sgran_size: %ld%c \tchunk_size: %ld%c \t", result[i].bad ? "*BAD*" : " ", diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c index fd31a441c61c..7d28d7d03885 100644 --- a/arch/x86/kernel/cpu/mtrr/generic.c +++ b/arch/x86/kernel/cpu/mtrr/generic.c @@ -433,13 +433,12 @@ static void generic_get_mtrr(unsigned int reg, unsigned long *base, { unsigned int mask_lo, mask_hi, base_lo, base_hi; unsigned int tmp, hi; - int cpu; /* * get_mtrr doesn't need to update mtrr_state, also it could be called * from any cpu, so try to print it out directly. */ - cpu = get_cpu(); + get_cpu(); rdmsr(MTRRphysMask_MSR(reg), mask_lo, mask_hi); diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c index 79556bd9b602..01c0f3ee6cc3 100644 --- a/arch/x86/kernel/cpu/mtrr/main.c +++ b/arch/x86/kernel/cpu/mtrr/main.c @@ -35,6 +35,7 @@ #include <linux/types.h> /* FIXME: kvm_para.h needs this */ +#include <linux/stop_machine.h> #include <linux/kvm_para.h> #include <linux/uaccess.h> #include <linux/module.h> @@ -143,22 +144,28 @@ struct set_mtrr_data { mtrr_type smp_type; }; +static DEFINE_PER_CPU(struct cpu_stop_work, mtrr_work); + /** - * ipi_handler - Synchronisation handler. Executed by "other" CPUs. + * mtrr_work_handler - Synchronisation handler. Executed by "other" CPUs. * @info: pointer to mtrr configuration data * * Returns nothing. */ -static void ipi_handler(void *info) +static int mtrr_work_handler(void *info) { #ifdef CONFIG_SMP struct set_mtrr_data *data = info; unsigned long flags; + atomic_dec(&data->count); + while (!atomic_read(&data->gate)) + cpu_relax(); + local_irq_save(flags); atomic_dec(&data->count); - while (!atomic_read(&data->gate)) + while (atomic_read(&data->gate)) cpu_relax(); /* The master has cleared me to execute */ @@ -173,12 +180,13 @@ static void ipi_handler(void *info) } atomic_dec(&data->count); - while (atomic_read(&data->gate)) + while (!atomic_read(&data->gate)) cpu_relax(); atomic_dec(&data->count); local_irq_restore(flags); #endif + return 0; } static inline int types_compatible(mtrr_type type1, mtrr_type type2) @@ -198,7 +206,7 @@ static inline int types_compatible(mtrr_type type1, mtrr_type type2) * * This is kinda tricky, but fortunately, Intel spelled it out for us cleanly: * - * 1. Send IPI to do the following: + * 1. Queue work to do the following on all processors: * 2. Disable Interrupts * 3. Wait for all procs to do so * 4. Enter no-fill cache mode @@ -215,14 +223,17 @@ static inline int types_compatible(mtrr_type type1, mtrr_type type2) * 15. Enable interrupts. * * What does that mean for us? Well, first we set data.count to the number - * of CPUs. As each CPU disables interrupts, it'll decrement it once. We wait - * until it hits 0 and proceed. We set the data.gate flag and reset data.count. - * Meanwhile, they are waiting for that flag to be set. Once it's set, each + * of CPUs. As each CPU announces that it started the rendezvous handler by + * decrementing the count, We reset data.count and set the data.gate flag + * allowing all the cpu's to proceed with the work. As each cpu disables + * interrupts, it'll decrement data.count once. We wait until it hits 0 and + * proceed. We clear the data.gate flag and reset data.count. Meanwhile, they + * are waiting for that flag to be cleared. Once it's cleared, each * CPU goes through the transition of updating MTRRs. * The CPU vendors may each do it differently, * so we call mtrr_if->set() callback and let them take care of it. * When they're done, they again decrement data->count and wait for data.gate - * to be reset. + * to be set. * When we finish, we wait for data.count to hit 0 and toggle the data.gate flag * Everyone then enables interrupts and we all continue on. * @@ -234,6 +245,9 @@ set_mtrr(unsigned int reg, unsigned long base, unsigned long size, mtrr_type typ { struct set_mtrr_data data; unsigned long flags; + int cpu; + + preempt_disable(); data.smp_reg = reg; data.smp_base = base; @@ -246,10 +260,15 @@ set_mtrr(unsigned int reg, unsigned long base, unsigned long size, mtrr_type typ atomic_set(&data.gate, 0); /* Start the ball rolling on other CPUs */ - if (smp_call_function(ipi_handler, &data, 0) != 0) - panic("mtrr: timed out waiting for other CPUs\n"); + for_each_online_cpu(cpu) { + struct cpu_stop_work *work = &per_cpu(mtrr_work, cpu); + + if (cpu == smp_processor_id()) + continue; + + stop_one_cpu_nowait(cpu, mtrr_work_handler, &data, work); + } - local_irq_save(flags); while (atomic_read(&data.count)) cpu_relax(); @@ -259,6 +278,16 @@ set_mtrr(unsigned int reg, unsigned long base, unsigned long size, mtrr_type typ smp_wmb(); atomic_set(&data.gate, 1); + local_irq_save(flags); + + while (atomic_read(&data.count)) + cpu_relax(); + + /* Ok, reset count and toggle gate */ + atomic_set(&data.count, num_booting_cpus() - 1); + smp_wmb(); + atomic_set(&data.gate, 0); + /* Do our MTRR business */ /* @@ -279,7 +308,7 @@ set_mtrr(unsigned int reg, unsigned long base, unsigned long size, mtrr_type typ atomic_set(&data.count, num_booting_cpus() - 1); smp_wmb(); - atomic_set(&data.gate, 0); + atomic_set(&data.gate, 1); /* * Wait here for everyone to have seen the gate change @@ -289,6 +318,7 @@ set_mtrr(unsigned int reg, unsigned long base, unsigned long size, mtrr_type typ cpu_relax(); local_irq_restore(flags); + preempt_enable(); } /** diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 5db5b7d65a18..f2da20fda02d 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -220,6 +220,7 @@ struct x86_pmu { struct perf_event *event); struct event_constraint *event_constraints; void (*quirks)(void); + int perfctr_second_write; int (*cpu_prepare)(int cpu); void (*cpu_starting)(int cpu); @@ -295,10 +296,10 @@ x86_perf_event_update(struct perf_event *event) * count to the generic event atomically: */ again: - prev_raw_count = atomic64_read(&hwc->prev_count); + prev_raw_count = local64_read(&hwc->prev_count); rdmsrl(hwc->event_base + idx, new_raw_count); - if (atomic64_cmpxchg(&hwc->prev_count, prev_raw_count, + if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, new_raw_count) != prev_raw_count) goto again; @@ -313,8 +314,8 @@ again: delta = (new_raw_count << shift) - (prev_raw_count << shift); delta >>= shift; - atomic64_add(delta, &event->count); - atomic64_sub(delta, &hwc->period_left); + local64_add(delta, &event->count); + local64_sub(delta, &hwc->period_left); return new_raw_count; } @@ -438,7 +439,7 @@ static int x86_setup_perfctr(struct perf_event *event) if (!hwc->sample_period) { hwc->sample_period = x86_pmu.max_period; hwc->last_period = hwc->sample_period; - atomic64_set(&hwc->period_left, hwc->sample_period); + local64_set(&hwc->period_left, hwc->sample_period); } else { /* * If we have a PMU initialized but no APIC @@ -885,7 +886,7 @@ static int x86_perf_event_set_period(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; - s64 left = atomic64_read(&hwc->period_left); + s64 left = local64_read(&hwc->period_left); s64 period = hwc->sample_period; int ret = 0, idx = hwc->idx; @@ -897,14 +898,14 @@ x86_perf_event_set_period(struct perf_event *event) */ if (unlikely(left <= -period)) { left = period; - atomic64_set(&hwc->period_left, left); + local64_set(&hwc->period_left, left); hwc->last_period = period; ret = 1; } if (unlikely(left <= 0)) { left += period; - atomic64_set(&hwc->period_left, left); + local64_set(&hwc->period_left, left); hwc->last_period = period; ret = 1; } @@ -923,10 +924,19 @@ x86_perf_event_set_period(struct perf_event *event) * 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); + local64_set(&hwc->prev_count, (u64)-left); - wrmsrl(hwc->event_base + idx, + wrmsrl(hwc->event_base + idx, (u64)(-left) & x86_pmu.cntval_mask); + + /* + * Due to erratum on certan cpu we need + * a second write to be sure the register + * is updated properly + */ + if (x86_pmu.perfctr_second_write) { + wrmsrl(hwc->event_base + idx, (u64)(-left) & x86_pmu.cntval_mask); + } perf_event_update_userpage(event); @@ -969,7 +979,7 @@ static int x86_pmu_enable(struct perf_event *event) * skip the schedulability test here, it will be peformed * at commit time(->commit_txn) as a whole */ - if (cpuc->group_flag & PERF_EVENT_TXN_STARTED) + if (cpuc->group_flag & PERF_EVENT_TXN) goto out; ret = x86_pmu.schedule_events(cpuc, n, assign); @@ -1096,7 +1106,7 @@ static void x86_pmu_disable(struct perf_event *event) * The events never got scheduled and ->cancel_txn will truncate * the event_list. */ - if (cpuc->group_flag & PERF_EVENT_TXN_STARTED) + if (cpuc->group_flag & PERF_EVENT_TXN) return; x86_pmu_stop(event); @@ -1388,7 +1398,7 @@ static void x86_pmu_start_txn(const struct pmu *pmu) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - cpuc->group_flag |= PERF_EVENT_TXN_STARTED; + cpuc->group_flag |= PERF_EVENT_TXN; cpuc->n_txn = 0; } @@ -1401,7 +1411,7 @@ static void x86_pmu_cancel_txn(const struct pmu *pmu) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - cpuc->group_flag &= ~PERF_EVENT_TXN_STARTED; + cpuc->group_flag &= ~PERF_EVENT_TXN; /* * Truncate the collected events. */ @@ -1435,11 +1445,7 @@ static int x86_pmu_commit_txn(const struct pmu *pmu) */ memcpy(cpuc->assign, assign, n*sizeof(int)); - /* - * Clear out the txn count so that ->cancel_txn() which gets - * run after ->commit_txn() doesn't undo things. - */ - cpuc->n_txn = 0; + cpuc->group_flag &= ~PERF_EVENT_TXN; return 0; } @@ -1607,8 +1613,6 @@ static const struct stacktrace_ops backtrace_ops = { .walk_stack = print_context_stack_bp, }; -#include "../dumpstack.h" - static void perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry) { @@ -1730,22 +1734,6 @@ struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) return entry; } -void perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int skip) -{ - regs->ip = ip; - /* - * perf_arch_fetch_caller_regs adds another call, we need to increment - * the skip level - */ - regs->bp = rewind_frame_pointer(skip + 1); - regs->cs = __KERNEL_CS; - /* - * We abuse bit 3 to pass exact information, see perf_misc_flags - * and the comment with PERF_EFLAGS_EXACT. - */ - regs->flags = 0; -} - unsigned long perf_instruction_pointer(struct pt_regs *regs) { unsigned long ip; diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c index ae85d69644d1..febb12cea795 100644 --- a/arch/x86/kernel/cpu/perf_event_p4.c +++ b/arch/x86/kernel/cpu/perf_event_p4.c @@ -21,22 +21,36 @@ struct p4_event_bind { char cntr[2][P4_CNTR_LIMIT]; /* counter index (offset), -1 on abscence */ }; -struct p4_cache_event_bind { +struct p4_pebs_bind { unsigned int metric_pebs; unsigned int metric_vert; }; -#define P4_GEN_CACHE_EVENT_BIND(name) \ - [P4_CACHE__##name] = { \ - .metric_pebs = P4_PEBS__##name, \ - .metric_vert = P4_VERT__##name, \ +/* it sets P4_PEBS_ENABLE_UOP_TAG as well */ +#define P4_GEN_PEBS_BIND(name, pebs, vert) \ + [P4_PEBS_METRIC__##name] = { \ + .metric_pebs = pebs | P4_PEBS_ENABLE_UOP_TAG, \ + .metric_vert = vert, \ } -static struct p4_cache_event_bind p4_cache_event_bind_map[] = { - P4_GEN_CACHE_EVENT_BIND(1stl_cache_load_miss_retired), - P4_GEN_CACHE_EVENT_BIND(2ndl_cache_load_miss_retired), - P4_GEN_CACHE_EVENT_BIND(dtlb_load_miss_retired), - P4_GEN_CACHE_EVENT_BIND(dtlb_store_miss_retired), +/* + * note we have P4_PEBS_ENABLE_UOP_TAG always set here + * + * it's needed for mapping P4_PEBS_CONFIG_METRIC_MASK bits of + * event configuration to find out which values are to be + * written into MSR_IA32_PEBS_ENABLE and MSR_P4_PEBS_MATRIX_VERT + * resgisters + */ +static struct p4_pebs_bind p4_pebs_bind_map[] = { + P4_GEN_PEBS_BIND(1stl_cache_load_miss_retired, 0x0000001, 0x0000001), + P4_GEN_PEBS_BIND(2ndl_cache_load_miss_retired, 0x0000002, 0x0000001), + P4_GEN_PEBS_BIND(dtlb_load_miss_retired, 0x0000004, 0x0000001), + P4_GEN_PEBS_BIND(dtlb_store_miss_retired, 0x0000004, 0x0000002), + P4_GEN_PEBS_BIND(dtlb_all_miss_retired, 0x0000004, 0x0000003), + P4_GEN_PEBS_BIND(tagged_mispred_branch, 0x0018000, 0x0000010), + P4_GEN_PEBS_BIND(mob_load_replay_retired, 0x0000200, 0x0000001), + P4_GEN_PEBS_BIND(split_load_retired, 0x0000400, 0x0000001), + P4_GEN_PEBS_BIND(split_store_retired, 0x0000400, 0x0000002), }; /* @@ -281,10 +295,10 @@ static struct p4_event_bind p4_event_bind_map[] = { }, }; -#define P4_GEN_CACHE_EVENT(event, bit, cache_event) \ +#define P4_GEN_CACHE_EVENT(event, bit, metric) \ p4_config_pack_escr(P4_ESCR_EVENT(event) | \ P4_ESCR_EMASK_BIT(event, bit)) | \ - p4_config_pack_cccr(cache_event | \ + p4_config_pack_cccr(metric | \ P4_CCCR_ESEL(P4_OPCODE_ESEL(P4_OPCODE(event)))) static __initconst const u64 p4_hw_cache_event_ids @@ -296,34 +310,34 @@ static __initconst const u64 p4_hw_cache_event_ids [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0x0, [ C(RESULT_MISS) ] = P4_GEN_CACHE_EVENT(P4_EVENT_REPLAY_EVENT, NBOGUS, - P4_CACHE__1stl_cache_load_miss_retired), + P4_PEBS_METRIC__1stl_cache_load_miss_retired), }, }, [ C(LL ) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0x0, [ C(RESULT_MISS) ] = P4_GEN_CACHE_EVENT(P4_EVENT_REPLAY_EVENT, NBOGUS, - P4_CACHE__2ndl_cache_load_miss_retired), + P4_PEBS_METRIC__2ndl_cache_load_miss_retired), }, }, [ C(DTLB) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0x0, [ C(RESULT_MISS) ] = P4_GEN_CACHE_EVENT(P4_EVENT_REPLAY_EVENT, NBOGUS, - P4_CACHE__dtlb_load_miss_retired), + P4_PEBS_METRIC__dtlb_load_miss_retired), }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = 0x0, [ C(RESULT_MISS) ] = P4_GEN_CACHE_EVENT(P4_EVENT_REPLAY_EVENT, NBOGUS, - P4_CACHE__dtlb_store_miss_retired), + P4_PEBS_METRIC__dtlb_store_miss_retired), }, }, [ C(ITLB) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = P4_GEN_CACHE_EVENT(P4_EVENT_ITLB_REFERENCE, HIT, - P4_CACHE__itlb_reference_hit), + P4_PEBS_METRIC__none), [ C(RESULT_MISS) ] = P4_GEN_CACHE_EVENT(P4_EVENT_ITLB_REFERENCE, MISS, - P4_CACHE__itlb_reference_miss), + P4_PEBS_METRIC__none), }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = -1, @@ -414,11 +428,37 @@ static u64 p4_pmu_event_map(int hw_event) return config; } +static int p4_validate_raw_event(struct perf_event *event) +{ + unsigned int v; + + /* user data may have out-of-bound event index */ + v = p4_config_unpack_event(event->attr.config); + if (v >= ARRAY_SIZE(p4_event_bind_map)) { + pr_warning("P4 PMU: Unknown event code: %d\n", v); + return -EINVAL; + } + + /* + * it may have some screwed PEBS bits + */ + if (p4_config_pebs_has(event->attr.config, P4_PEBS_CONFIG_ENABLE)) { + pr_warning("P4 PMU: PEBS are not supported yet\n"); + return -EINVAL; + } + v = p4_config_unpack_metric(event->attr.config); + if (v >= ARRAY_SIZE(p4_pebs_bind_map)) { + pr_warning("P4 PMU: Unknown metric code: %d\n", v); + return -EINVAL; + } + + return 0; +} + static int p4_hw_config(struct perf_event *event) { int cpu = get_cpu(); int rc = 0; - unsigned int evnt; u32 escr, cccr; /* @@ -438,12 +478,9 @@ static int p4_hw_config(struct perf_event *event) if (event->attr.type == PERF_TYPE_RAW) { - /* user data may have out-of-bound event index */ - evnt = p4_config_unpack_event(event->attr.config); - if (evnt >= ARRAY_SIZE(p4_event_bind_map)) { - rc = -EINVAL; + rc = p4_validate_raw_event(event); + if (rc) goto out; - } /* * We don't control raw events so it's up to the caller @@ -451,12 +488,15 @@ static int p4_hw_config(struct perf_event *event) * on HT machine but allow HT-compatible specifics to be * passed on) * + * Note that for RAW events we allow user to use P4_CCCR_RESERVED + * bits since we keep additional info here (for cache events and etc) + * * XXX: HT wide things should check perf_paranoid_cpu() && * CAP_SYS_ADMIN */ event->hw.config |= event->attr.config & (p4_config_pack_escr(P4_ESCR_MASK_HT) | - p4_config_pack_cccr(P4_CCCR_MASK_HT)); + p4_config_pack_cccr(P4_CCCR_MASK_HT | P4_CCCR_RESERVED)); } rc = x86_setup_perfctr(event); @@ -482,6 +522,29 @@ static inline int p4_pmu_clear_cccr_ovf(struct hw_perf_event *hwc) return overflow; } +static void p4_pmu_disable_pebs(void) +{ + /* + * FIXME + * + * It's still allowed that two threads setup same cache + * events so we can't simply clear metrics until we knew + * noone is depending on us, so we need kind of counter + * for "ReplayEvent" users. + * + * What is more complex -- RAW events, if user (for some + * reason) will pass some cache event metric with improper + * event opcode -- it's fine from hardware point of view + * but completely nonsence from "meaning" of such action. + * + * So at moment let leave metrics turned on forever -- it's + * ok for now but need to be revisited! + * + * (void)checking_wrmsrl(MSR_IA32_PEBS_ENABLE, (u64)0); + * (void)checking_wrmsrl(MSR_P4_PEBS_MATRIX_VERT, (u64)0); + */ +} + static inline void p4_pmu_disable_event(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; @@ -507,6 +570,26 @@ static void p4_pmu_disable_all(void) continue; p4_pmu_disable_event(event); } + + p4_pmu_disable_pebs(); +} + +/* configuration must be valid */ +static void p4_pmu_enable_pebs(u64 config) +{ + struct p4_pebs_bind *bind; + unsigned int idx; + + BUILD_BUG_ON(P4_PEBS_METRIC__max > P4_PEBS_CONFIG_METRIC_MASK); + + idx = p4_config_unpack_metric(config); + if (idx == P4_PEBS_METRIC__none) + return; + + bind = &p4_pebs_bind_map[idx]; + + (void)checking_wrmsrl(MSR_IA32_PEBS_ENABLE, (u64)bind->metric_pebs); + (void)checking_wrmsrl(MSR_P4_PEBS_MATRIX_VERT, (u64)bind->metric_vert); } static void p4_pmu_enable_event(struct perf_event *event) @@ -515,9 +598,7 @@ static void p4_pmu_enable_event(struct perf_event *event) int thread = p4_ht_config_thread(hwc->config); u64 escr_conf = p4_config_unpack_escr(p4_clear_ht_bit(hwc->config)); unsigned int idx = p4_config_unpack_event(hwc->config); - unsigned int idx_cache = p4_config_unpack_cache_event(hwc->config); struct p4_event_bind *bind; - struct p4_cache_event_bind *bind_cache; u64 escr_addr, cccr; bind = &p4_event_bind_map[idx]; @@ -537,16 +618,10 @@ static void p4_pmu_enable_event(struct perf_event *event) cccr = p4_config_unpack_cccr(hwc->config); /* - * it could be Cache event so that we need to - * set metrics into additional MSRs + * it could be Cache event so we need to write metrics + * into additional MSRs */ - BUILD_BUG_ON(P4_CACHE__MAX > P4_CCCR_CACHE_OPS_MASK); - if (idx_cache > P4_CACHE__NONE && - idx_cache < ARRAY_SIZE(p4_cache_event_bind_map)) { - bind_cache = &p4_cache_event_bind_map[idx_cache]; - (void)checking_wrmsrl(MSR_IA32_PEBS_ENABLE, (u64)bind_cache->metric_pebs); - (void)checking_wrmsrl(MSR_P4_PEBS_MATRIX_VERT, (u64)bind_cache->metric_vert); - } + p4_pmu_enable_pebs(hwc->config); (void)checking_wrmsrl(escr_addr, escr_conf); (void)checking_wrmsrl(hwc->config_base + hwc->idx, @@ -581,6 +656,7 @@ static int p4_pmu_handle_irq(struct pt_regs *regs) cpuc = &__get_cpu_var(cpu_hw_events); for (idx = 0; idx < x86_pmu.num_counters; idx++) { + int overflow; if (!test_bit(idx, cpuc->active_mask)) continue; @@ -591,12 +667,14 @@ static int p4_pmu_handle_irq(struct pt_regs *regs) WARN_ON_ONCE(hwc->idx != idx); /* it might be unflagged overflow */ - handled = p4_pmu_clear_cccr_ovf(hwc); + overflow = p4_pmu_clear_cccr_ovf(hwc); val = x86_perf_event_update(event); - if (!handled && (val & (1ULL << (x86_pmu.cntval_bits - 1)))) + if (!overflow && (val & (1ULL << (x86_pmu.cntval_bits - 1)))) continue; + handled += overflow; + /* event overflow for sure */ data.period = event->hw.last_period; @@ -612,7 +690,7 @@ static int p4_pmu_handle_irq(struct pt_regs *regs) inc_irq_stat(apic_perf_irqs); } - return handled; + return handled > 0; } /* @@ -829,6 +907,15 @@ static __initconst const struct x86_pmu p4_pmu = { .max_period = (1ULL << 39) - 1, .hw_config = p4_hw_config, .schedule_events = p4_pmu_schedule_events, + /* + * This handles erratum N15 in intel doc 249199-029, + * the counter may not be updated correctly on write + * so we need a second write operation to do the trick + * (the official workaround didn't work) + * + * the former idea is taken from OProfile code + */ + .perfctr_second_write = 1, }; static __init int p4_pmu_init(void) diff --git a/arch/x86/kernel/cpu/scattered.c b/arch/x86/kernel/cpu/scattered.c new file mode 100644 index 000000000000..34b4dad6f0b8 --- /dev/null +++ b/arch/x86/kernel/cpu/scattered.c @@ -0,0 +1,63 @@ +/* + * Routines to indentify additional cpu features that are scattered in + * cpuid space. + */ +#include <linux/cpu.h> + +#include <asm/pat.h> +#include <asm/processor.h> + +#include <asm/apic.h> + +struct cpuid_bit { + u16 feature; + u8 reg; + u8 bit; + u32 level; + u32 sub_leaf; +}; + +enum cpuid_regs { + CR_EAX = 0, + CR_ECX, + CR_EDX, + CR_EBX +}; + +void __cpuinit init_scattered_cpuid_features(struct cpuinfo_x86 *c) +{ + u32 max_level; + u32 regs[4]; + const struct cpuid_bit *cb; + + static const struct cpuid_bit __cpuinitconst cpuid_bits[] = { + { X86_FEATURE_IDA, CR_EAX, 1, 0x00000006, 0 }, + { X86_FEATURE_ARAT, CR_EAX, 2, 0x00000006, 0 }, + { X86_FEATURE_PLN, CR_EAX, 4, 0x00000006, 0 }, + { X86_FEATURE_PTS, CR_EAX, 6, 0x00000006, 0 }, + { X86_FEATURE_APERFMPERF, CR_ECX, 0, 0x00000006, 0 }, + { X86_FEATURE_EPB, CR_ECX, 3, 0x00000006, 0 }, + { X86_FEATURE_XSAVEOPT, CR_EAX, 0, 0x0000000d, 1 }, + { X86_FEATURE_CPB, CR_EDX, 9, 0x80000007, 0 }, + { X86_FEATURE_NPT, CR_EDX, 0, 0x8000000a, 0 }, + { X86_FEATURE_LBRV, CR_EDX, 1, 0x8000000a, 0 }, + { X86_FEATURE_SVML, CR_EDX, 2, 0x8000000a, 0 }, + { X86_FEATURE_NRIPS, CR_EDX, 3, 0x8000000a, 0 }, + { 0, 0, 0, 0, 0 } + }; + + for (cb = cpuid_bits; cb->feature; cb++) { + + /* Verify that the level is valid */ + max_level = cpuid_eax(cb->level & 0xffff0000); + if (max_level < cb->level || + max_level > (cb->level | 0xffff)) + continue; + + cpuid_count(cb->level, cb->sub_leaf, ®s[CR_EAX], + ®s[CR_EBX], ®s[CR_ECX], ®s[CR_EDX]); + + if (regs[cb->reg] & (1 << cb->bit)) + set_cpu_cap(c, cb->feature); + } +} diff --git a/arch/x86/kernel/cpu/addon_cpuid_features.c b/arch/x86/kernel/cpu/topology.c index 10fa5684a662..4397e987a1cf 100644 --- a/arch/x86/kernel/cpu/addon_cpuid_features.c +++ b/arch/x86/kernel/cpu/topology.c @@ -1,62 +1,14 @@ /* - * Routines to indentify additional cpu features that are scattered in - * cpuid space. + * Check for extended topology enumeration cpuid leaf 0xb and if it + * exists, use it for populating initial_apicid and cpu topology + * detection. */ -#include <linux/cpu.h> +#include <linux/cpu.h> +#include <asm/apic.h> #include <asm/pat.h> #include <asm/processor.h> -#include <asm/apic.h> - -struct cpuid_bit { - u16 feature; - u8 reg; - u8 bit; - u32 level; -}; - -enum cpuid_regs { - CR_EAX = 0, - CR_ECX, - CR_EDX, - CR_EBX -}; - -void __cpuinit init_scattered_cpuid_features(struct cpuinfo_x86 *c) -{ - u32 max_level; - u32 regs[4]; - const struct cpuid_bit *cb; - - static const struct cpuid_bit __cpuinitconst cpuid_bits[] = { - { X86_FEATURE_IDA, CR_EAX, 1, 0x00000006 }, - { X86_FEATURE_ARAT, CR_EAX, 2, 0x00000006 }, - { X86_FEATURE_APERFMPERF, CR_ECX, 0, 0x00000006 }, - { X86_FEATURE_CPB, CR_EDX, 9, 0x80000007 }, - { X86_FEATURE_NPT, CR_EDX, 0, 0x8000000a }, - { X86_FEATURE_LBRV, CR_EDX, 1, 0x8000000a }, - { X86_FEATURE_SVML, CR_EDX, 2, 0x8000000a }, - { X86_FEATURE_NRIPS, CR_EDX, 3, 0x8000000a }, - { 0, 0, 0, 0 } - }; - - for (cb = cpuid_bits; cb->feature; cb++) { - - /* Verify that the level is valid */ - max_level = cpuid_eax(cb->level & 0xffff0000); - if (max_level < cb->level || - max_level > (cb->level | 0xffff)) - continue; - - cpuid(cb->level, ®s[CR_EAX], ®s[CR_EBX], - ®s[CR_ECX], ®s[CR_EDX]); - - if (regs[cb->reg] & (1 << cb->bit)) - set_cpu_cap(c, cb->feature); - } -} - /* leaf 0xb SMT level */ #define SMT_LEVEL 0 diff --git a/arch/x86/kernel/cpu/vmware.c b/arch/x86/kernel/cpu/vmware.c index b9d1ff588445..227b0448960d 100644 --- a/arch/x86/kernel/cpu/vmware.c +++ b/arch/x86/kernel/cpu/vmware.c @@ -51,7 +51,7 @@ static inline int __vmware_platform(void) static unsigned long vmware_get_tsc_khz(void) { - uint64_t tsc_hz; + uint64_t tsc_hz, lpj; uint32_t eax, ebx, ecx, edx; VMWARE_PORT(GETHZ, eax, ebx, ecx, edx); @@ -62,6 +62,13 @@ static unsigned long vmware_get_tsc_khz(void) printk(KERN_INFO "TSC freq read from hypervisor : %lu.%03lu MHz\n", (unsigned long) tsc_hz / 1000, (unsigned long) tsc_hz % 1000); + + if (!preset_lpj) { + lpj = ((u64)tsc_hz * 1000); + do_div(lpj, HZ); + preset_lpj = lpj; + } + return tsc_hz; } |