diff options
Diffstat (limited to 'arch/x86/kernel/cpu')
64 files changed, 16526 insertions, 2154 deletions
diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile index 25f24dccdcf..7fd54f09b01 100644 --- a/arch/x86/kernel/cpu/Makefile +++ b/arch/x86/kernel/cpu/Makefile @@ -14,8 +14,8 @@ CFLAGS_common.o := $(nostackp) obj-y := intel_cacheinfo.o scattered.o topology.o obj-y += proc.o capflags.o powerflags.o common.o -obj-y += vmware.o hypervisor.o sched.o mshyperv.o obj-y += rdrand.o +obj-y += match.o obj-$(CONFIG_X86_32) += bugs.o obj-$(CONFIG_X86_64) += bugs_64.o @@ -30,20 +30,29 @@ obj-$(CONFIG_CPU_SUP_UMC_32) += umc.o obj-$(CONFIG_PERF_EVENTS) += perf_event.o ifdef CONFIG_PERF_EVENTS -obj-$(CONFIG_CPU_SUP_AMD) += perf_event_amd.o -obj-$(CONFIG_CPU_SUP_INTEL) += perf_event_p6.o perf_event_p4.o perf_event_intel_lbr.o perf_event_intel_ds.o perf_event_intel.o +obj-$(CONFIG_CPU_SUP_AMD) += perf_event_amd.o perf_event_amd_uncore.o +ifdef CONFIG_AMD_IOMMU +obj-$(CONFIG_CPU_SUP_AMD) += perf_event_amd_iommu.o endif +obj-$(CONFIG_CPU_SUP_INTEL) += perf_event_p6.o perf_event_knc.o perf_event_p4.o +obj-$(CONFIG_CPU_SUP_INTEL) += perf_event_intel_lbr.o perf_event_intel_ds.o perf_event_intel.o +obj-$(CONFIG_CPU_SUP_INTEL) += perf_event_intel_uncore.o perf_event_intel_rapl.o +endif + obj-$(CONFIG_X86_MCE) += mcheck/ obj-$(CONFIG_MTRR) += mtrr/ +obj-$(CONFIG_MICROCODE) += microcode/ obj-$(CONFIG_X86_LOCAL_APIC) += perfctr-watchdog.o perf_event_amd_ibs.o +obj-$(CONFIG_HYPERVISOR_GUEST) += vmware.o hypervisor.o mshyperv.o + quiet_cmd_mkcapflags = MKCAP $@ - cmd_mkcapflags = $(PERL) $(srctree)/$(src)/mkcapflags.pl $< $@ + cmd_mkcapflags = $(CONFIG_SHELL) $(srctree)/$(src)/mkcapflags.sh $< $@ cpufeature = $(src)/../../include/asm/cpufeature.h targets += capflags.c -$(obj)/capflags.c: $(cpufeature) $(src)/mkcapflags.pl FORCE +$(obj)/capflags.c: $(cpufeature) $(src)/mkcapflags.sh FORCE $(call if_changed,mkcapflags) diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index f4773f4aae3..ce8b8ff0e0e 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -1,23 +1,55 @@ #include <linux/export.h> -#include <linux/init.h> #include <linux/bitops.h> #include <linux/elf.h> #include <linux/mm.h> #include <linux/io.h> +#include <linux/sched.h> #include <asm/processor.h> #include <asm/apic.h> #include <asm/cpu.h> #include <asm/pci-direct.h> #ifdef CONFIG_X86_64 -# include <asm/numa_64.h> # include <asm/mmconfig.h> # include <asm/cacheflush.h> #endif #include "cpu.h" +static inline int rdmsrl_amd_safe(unsigned msr, unsigned long long *p) +{ + u32 gprs[8] = { 0 }; + int err; + + WARN_ONCE((boot_cpu_data.x86 != 0xf), + "%s should only be used on K8!\n", __func__); + + gprs[1] = msr; + gprs[7] = 0x9c5a203a; + + err = rdmsr_safe_regs(gprs); + + *p = gprs[0] | ((u64)gprs[2] << 32); + + return err; +} + +static inline int wrmsrl_amd_safe(unsigned msr, unsigned long long val) +{ + u32 gprs[8] = { 0 }; + + WARN_ONCE((boot_cpu_data.x86 != 0xf), + "%s should only be used on K8!\n", __func__); + + gprs[0] = (u32)val; + gprs[1] = msr; + gprs[2] = val >> 32; + gprs[7] = 0x9c5a203a; + + return wrmsr_safe_regs(gprs); +} + #ifdef CONFIG_X86_32 /* * B step AMD K6 before B 9730xxxx have hardware bugs that can cause @@ -25,17 +57,18 @@ * contact AMD for precise details and a CPU swap. * * See http://www.multimania.com/poulot/k6bug.html - * http://www.amd.com/K6/k6docs/revgd.html + * and section 2.6.2 of "AMD-K6 Processor Revision Guide - Model 6" + * (Publication # 21266 Issue Date: August 1998) * * The following test is erm.. interesting. AMD neglected to up * the chip setting when fixing the bug but they also tweaked some * performance at the same time.. */ -extern void vide(void); -__asm__(".align 4\nvide: ret"); +extern __visible void vide(void); +__asm__(".globl vide\n\t.align 4\nvide: ret"); -static void __cpuinit init_amd_k5(struct cpuinfo_x86 *c) +static void init_amd_k5(struct cpuinfo_x86 *c) { /* * General Systems BIOSen alias the cpu frequency registers @@ -53,10 +86,10 @@ static void __cpuinit init_amd_k5(struct cpuinfo_x86 *c) } -static void __cpuinit init_amd_k6(struct cpuinfo_x86 *c) +static void init_amd_k6(struct cpuinfo_x86 *c) { u32 l, h; - int mbytes = num_physpages >> (20-PAGE_SHIFT); + int mbytes = get_num_physpages() >> (20-PAGE_SHIFT); if (c->x86_model < 6) { /* Based on AMD doc 20734R - June 2000 */ @@ -93,7 +126,6 @@ static void __cpuinit init_amd_k6(struct cpuinfo_x86 *c) "system stability may be impaired when more than 32 MB are used.\n"); else printk(KERN_CONT "probably OK (after B9730xxxx).\n"); - printk(KERN_INFO "Please see http://membres.lycos.fr/poulot/k6bug.html\n"); } /* K6 with old style WHCR */ @@ -146,7 +178,7 @@ static void __cpuinit init_amd_k6(struct cpuinfo_x86 *c) } } -static void __cpuinit amd_k7_smp_check(struct cpuinfo_x86 *c) +static void amd_k7_smp_check(struct cpuinfo_x86 *c) { /* calling is from identify_secondary_cpu() ? */ if (!c->cpu_index) @@ -159,11 +191,11 @@ static void __cpuinit amd_k7_smp_check(struct cpuinfo_x86 *c) /* Athlon 660/661 is valid. */ if ((c->x86_model == 6) && ((c->x86_mask == 0) || (c->x86_mask == 1))) - goto valid_k7; + return; /* Duron 670 is valid */ if ((c->x86_model == 7) && (c->x86_mask == 0)) - goto valid_k7; + return; /* * Athlon 662, Duron 671, and Athlon >model 7 have capability @@ -176,7 +208,7 @@ static void __cpuinit amd_k7_smp_check(struct cpuinfo_x86 *c) ((c->x86_model == 7) && (c->x86_mask >= 1)) || (c->x86_model > 7)) if (cpu_has_mp) - goto valid_k7; + return; /* If we get here, not a certified SMP capable AMD system. */ @@ -186,14 +218,10 @@ static void __cpuinit amd_k7_smp_check(struct cpuinfo_x86 *c) */ WARN_ONCE(1, "WARNING: This combination of AMD" " processors is not suitable for SMP.\n"); - if (!test_taint(TAINT_UNSAFE_SMP)) - add_taint(TAINT_UNSAFE_SMP); - -valid_k7: - ; + add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_NOW_UNRELIABLE); } -static void __cpuinit init_amd_k7(struct cpuinfo_x86 *c) +static void init_amd_k7(struct cpuinfo_x86 *c) { u32 l, h; @@ -205,9 +233,7 @@ static void __cpuinit init_amd_k7(struct cpuinfo_x86 *c) if (c->x86_model >= 6 && c->x86_model <= 10) { if (!cpu_has(c, X86_FEATURE_XMM)) { printk(KERN_INFO "Enabling disabled K7/SSE Support.\n"); - rdmsr(MSR_K7_HWCR, l, h); - l &= ~0x00008000; - wrmsr(MSR_K7_HWCR, l, h); + msr_clear_bit(MSR_K7_HWCR, 15); set_cpu_cap(c, X86_FEATURE_XMM); } } @@ -238,7 +264,7 @@ static void __cpuinit init_amd_k7(struct cpuinfo_x86 *c) * To workaround broken NUMA config. Read the comment in * srat_detect_node(). */ -static int __cpuinit nearby_node(int apicid) +static int nearby_node(int apicid) { int i, node; @@ -263,14 +289,14 @@ static int __cpuinit nearby_node(int apicid) * (2) AMD processors supporting compute units */ #ifdef CONFIG_X86_HT -static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c) +static void amd_get_topology(struct cpuinfo_x86 *c) { u32 nodes, cores_per_cu = 1; u8 node_id; int cpu = smp_processor_id(); /* get information required for multi-node processors */ - if (cpu_has(c, X86_FEATURE_TOPOEXT)) { + if (cpu_has_topoext) { u32 eax, ebx, ecx, edx; cpuid(0x8000001e, &eax, &ebx, &ecx, &edx); @@ -310,10 +336,10 @@ static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c) #endif /* - * On a AMD dual core setup the lower bits of the APIC id distingush the cores. + * On a AMD dual core setup the lower bits of the APIC id distinguish the cores. * Assumes number of cores is a power of two. */ -static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c) +static void amd_detect_cmp(struct cpuinfo_x86 *c) { #ifdef CONFIG_X86_HT unsigned bits; @@ -330,9 +356,9 @@ static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c) #endif } -int amd_get_nb_id(int cpu) +u16 amd_get_nb_id(int cpu) { - int id = 0; + u16 id = 0; #ifdef CONFIG_SMP id = per_cpu(cpu_llc_id, cpu); #endif @@ -340,7 +366,7 @@ int amd_get_nb_id(int cpu) } EXPORT_SYMBOL_GPL(amd_get_nb_id); -static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) +static void srat_detect_node(struct cpuinfo_x86 *c) { #ifdef CONFIG_NUMA int cpu = smp_processor_id(); @@ -352,10 +378,11 @@ static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) node = per_cpu(cpu_llc_id, cpu); /* - * If core numbers are inconsistent, it's likely a multi-fabric platform, - * so invoke platform-specific handler + * On multi-fabric platform (e.g. Numascale NumaChip) a + * platform-specific handler needs to be called to fixup some + * IDs of the CPU. */ - if (c->phys_proc_id != node) + if (x86_cpuinit.fixup_cpu_id) x86_cpuinit.fixup_cpu_id(c, node); if (!node_online(node)) { @@ -391,7 +418,7 @@ static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) #endif } -static void __cpuinit early_init_amd_mc(struct cpuinfo_x86 *c) +static void early_init_amd_mc(struct cpuinfo_x86 *c) { #ifdef CONFIG_X86_HT unsigned bits, ecx; @@ -417,7 +444,7 @@ static void __cpuinit early_init_amd_mc(struct cpuinfo_x86 *c) #endif } -static void __cpuinit bsp_init_amd(struct cpuinfo_x86 *c) +static void bsp_init_amd(struct cpuinfo_x86 *c) { if (cpu_has(c, X86_FEATURE_CONSTANT_TSC)) { @@ -445,7 +472,7 @@ static void __cpuinit bsp_init_amd(struct cpuinfo_x86 *c) } } -static void __cpuinit early_init_amd(struct cpuinfo_x86 *c) +static void early_init_amd(struct cpuinfo_x86 *c) { early_init_amd_mc(c); @@ -456,6 +483,8 @@ static void __cpuinit early_init_amd(struct cpuinfo_x86 *c) if (c->x86_power & (1 << 8)) { set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC); + if (!check_tsc_unstable()) + set_sched_clock_stable(); } #ifdef CONFIG_X86_64 @@ -476,15 +505,22 @@ static void __cpuinit early_init_amd(struct cpuinfo_x86 *c) set_cpu_cap(c, X86_FEATURE_EXTD_APICID); } #endif + + /* F16h erratum 793, CVE-2013-6885 */ + if (c->x86 == 0x16 && c->x86_model <= 0xf) + msr_set_bit(MSR_AMD64_LS_CFG, 15); } -static void __cpuinit init_amd(struct cpuinfo_x86 *c) +static const int amd_erratum_383[]; +static const int amd_erratum_400[]; +static bool cpu_has_amd_erratum(struct cpuinfo_x86 *cpu, const int *erratum); + +static void init_amd(struct cpuinfo_x86 *c) { u32 dummy; - -#ifdef CONFIG_SMP unsigned long long value; +#ifdef CONFIG_SMP /* * Disable TLB flush filter by setting HWCR.FFDIS on K8 * bit 6 of msr C001_0015 @@ -492,11 +528,8 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) * Errata 63 for SH-B3 steppings * Errata 122 for all steppings (F+ have it disabled by default) */ - if (c->x86 == 0xf) { - rdmsrl(MSR_K7_HWCR, value); - value |= 1 << 6; - wrmsrl(MSR_K7_HWCR, value); - } + if (c->x86 == 0xf) + msr_set_bit(MSR_K7_HWCR, 6); #endif early_init_amd(c); @@ -522,12 +555,10 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) * (AMD Erratum #110, docId: 25759). */ if (c->x86_model < 0x14 && cpu_has(c, X86_FEATURE_LAHF_LM)) { - u64 val; - clear_cpu_cap(c, X86_FEATURE_LAHF_LM); - if (!rdmsrl_amd_safe(0xc001100d, &val)) { - val &= ~(1ULL << 32); - wrmsrl_amd_safe(0xc001100d, val); + if (!rdmsrl_amd_safe(0xc001100d, &value)) { + value &= ~(1ULL << 32); + wrmsrl_amd_safe(0xc001100d, value); } } @@ -576,6 +607,33 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) } } + /* re-enable TopologyExtensions if switched off by BIOS */ + if ((c->x86 == 0x15) && + (c->x86_model >= 0x10) && (c->x86_model <= 0x1f) && + !cpu_has(c, X86_FEATURE_TOPOEXT)) { + + if (msr_set_bit(0xc0011005, 54) > 0) { + rdmsrl(0xc0011005, value); + if (value & BIT_64(54)) { + set_cpu_cap(c, X86_FEATURE_TOPOEXT); + pr_info(FW_INFO "CPU: Re-enabling disabled Topology Extensions Support.\n"); + } + } + } + + /* + * The way access filter has a performance penalty on some workloads. + * Disable it on the affected CPUs. + */ + if ((c->x86 == 0x15) && + (c->x86_model >= 0x02) && (c->x86_model < 0x20)) { + + if (!rdmsrl_safe(0xc0011021, &value) && !(value & 0x1E)) { + value |= 0x1E; + wrmsrl_safe(0xc0011021, value); + } + } + cpu_detect_cache_sizes(c); /* Multi core CPU? */ @@ -588,12 +646,7 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) detect_ht(c); #endif - if (c->extended_cpuid_level >= 0x80000006) { - if (cpuid_edx(0x80000006) & 0xf000) - num_cache_leaves = 4; - else - num_cache_leaves = 3; - } + init_amd_cacheinfo(c); if (c->x86 >= 0xf) set_cpu_cap(c, X86_FEATURE_K8); @@ -621,12 +674,10 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) * benefit in doing so. */ if (!rdmsrl_safe(MSR_K8_TSEG_ADDR, &tseg)) { + unsigned long pfn = tseg >> PAGE_SHIFT; + printk(KERN_DEBUG "tseg: %010llx\n", tseg); - if ((tseg>>PMD_SHIFT) < - (max_low_pfn_mapped>>(PMD_SHIFT-PAGE_SHIFT)) || - ((tseg>>PMD_SHIFT) < - (max_pfn_mapped>>(PMD_SHIFT-PAGE_SHIFT)) && - (tseg>>PMD_SHIFT) >= (1ULL<<(32 - PMD_SHIFT)))) + if (pfn_range_is_mapped(pfn, pfn + 1)) set_memory_4k((unsigned long)__va(tseg), 1); } } @@ -639,34 +690,42 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) if (c->x86 > 0x11) set_cpu_cap(c, X86_FEATURE_ARAT); - /* - * Disable GART TLB Walk Errors on Fam10h. We do this here - * because this is always needed when GART is enabled, even in a - * kernel which has no MCE support built in. - */ if (c->x86 == 0x10) { /* - * BIOS should disable GartTlbWlk Errors themself. If - * it doesn't do it here as suggested by the BKDG. + * Disable GART TLB Walk Errors on Fam10h. We do this here + * because this is always needed when GART is enabled, even in a + * kernel which has no MCE support built in. + * BIOS should disable GartTlbWlk Errors already. If + * it doesn't, do it here as suggested by the BKDG. * * Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=33012 */ - u64 mask; - int err; + msr_set_bit(MSR_AMD64_MCx_MASK(4), 10); - err = rdmsrl_safe(MSR_AMD64_MCx_MASK(4), &mask); - if (err == 0) { - mask |= (1 << 10); - checking_wrmsrl(MSR_AMD64_MCx_MASK(4), mask); - } + /* + * On family 10h BIOS may not have properly enabled WC+ support, + * causing it to be converted to CD memtype. This may result in + * performance degradation for certain nested-paging guests. + * Prevent this conversion by clearing bit 24 in + * MSR_AMD64_BU_CFG2. + * + * NOTE: we want to use the _safe accessors so as not to #GP kvm + * guests on older kvm hosts. + */ + msr_clear_bit(MSR_AMD64_BU_CFG2, 24); + + if (cpu_has_amd_erratum(c, amd_erratum_383)) + set_cpu_bug(c, X86_BUG_AMD_TLB_MMATCH); } + if (cpu_has_amd_erratum(c, amd_erratum_400)) + set_cpu_bug(c, X86_BUG_AMD_APIC_C1E); + rdmsr_safe(MSR_AMD64_PATCH_LEVEL, &c->microcode, &dummy); } #ifdef CONFIG_X86_32 -static unsigned int __cpuinit amd_size_cache(struct cpuinfo_x86 *c, - unsigned int size) +static unsigned int amd_size_cache(struct cpuinfo_x86 *c, unsigned int size) { /* AMD errata T13 (order #21922) */ if ((c->x86 == 6)) { @@ -682,12 +741,68 @@ static unsigned int __cpuinit amd_size_cache(struct cpuinfo_x86 *c, } #endif -static const struct cpu_dev __cpuinitconst amd_cpu_dev = { +static void cpu_set_tlb_flushall_shift(struct cpuinfo_x86 *c) +{ + tlb_flushall_shift = 6; +} + +static void cpu_detect_tlb_amd(struct cpuinfo_x86 *c) +{ + u32 ebx, eax, ecx, edx; + u16 mask = 0xfff; + + if (c->x86 < 0xf) + return; + + if (c->extended_cpuid_level < 0x80000006) + return; + + cpuid(0x80000006, &eax, &ebx, &ecx, &edx); + + tlb_lld_4k[ENTRIES] = (ebx >> 16) & mask; + tlb_lli_4k[ENTRIES] = ebx & mask; + + /* + * K8 doesn't have 2M/4M entries in the L2 TLB so read out the L1 TLB + * characteristics from the CPUID function 0x80000005 instead. + */ + if (c->x86 == 0xf) { + cpuid(0x80000005, &eax, &ebx, &ecx, &edx); + mask = 0xff; + } + + /* Handle DTLB 2M and 4M sizes, fall back to L1 if L2 is disabled */ + if (!((eax >> 16) & mask)) + tlb_lld_2m[ENTRIES] = (cpuid_eax(0x80000005) >> 16) & 0xff; + else + tlb_lld_2m[ENTRIES] = (eax >> 16) & mask; + + /* a 4M entry uses two 2M entries */ + tlb_lld_4m[ENTRIES] = tlb_lld_2m[ENTRIES] >> 1; + + /* Handle ITLB 2M and 4M sizes, fall back to L1 if L2 is disabled */ + if (!(eax & mask)) { + /* Erratum 658 */ + if (c->x86 == 0x15 && c->x86_model <= 0x1f) { + tlb_lli_2m[ENTRIES] = 1024; + } else { + cpuid(0x80000005, &eax, &ebx, &ecx, &edx); + tlb_lli_2m[ENTRIES] = eax & 0xff; + } + } else + tlb_lli_2m[ENTRIES] = eax & mask; + + tlb_lli_4m[ENTRIES] = tlb_lli_2m[ENTRIES] >> 1; + + cpu_set_tlb_flushall_shift(c); +} + +static const struct cpu_dev amd_cpu_dev = { .c_vendor = "AMD", .c_ident = { "AuthenticAMD" }, #ifdef CONFIG_X86_32 - .c_models = { - { .vendor = X86_VENDOR_AMD, .family = 4, .model_names = + .legacy_models = { + { .family = 4, .model_names = { [3] = "486 DX/2", [7] = "486 DX/2-WB", @@ -698,9 +813,10 @@ static const struct cpu_dev __cpuinitconst amd_cpu_dev = { } }, }, - .c_size_cache = amd_size_cache, + .legacy_cache_size = amd_size_cache, #endif .c_early_init = early_init_amd, + .c_detect_tlb = cpu_detect_tlb_amd, .c_bsp_init = bsp_init_amd, .c_init = init_amd, .c_x86_vendor = X86_VENDOR_AMD, @@ -715,8 +831,7 @@ cpu_dev_register(amd_cpu_dev); * 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. + * AMD_MODEL_RANGE(). * * Example: * @@ -726,32 +841,28 @@ cpu_dev_register(amd_cpu_dev); * AMD_MODEL_RANGE(0x10, 0x9, 0x0, 0x9, 0x0)); */ -const int amd_erratum_400[] = +#define AMD_LEGACY_ERRATUM(...) { -1, __VA_ARGS__, 0 } +#define AMD_OSVW_ERRATUM(osvw_id, ...) { osvw_id, __VA_ARGS__, 0 } +#define AMD_MODEL_RANGE(f, m_start, s_start, m_end, s_end) \ + ((f << 24) | (m_start << 16) | (s_start << 12) | (m_end << 4) | (s_end)) +#define AMD_MODEL_RANGE_FAMILY(range) (((range) >> 24) & 0xff) +#define AMD_MODEL_RANGE_START(range) (((range) >> 12) & 0xfff) +#define AMD_MODEL_RANGE_END(range) ((range) & 0xfff) + +static 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[] = +static 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) + +static bool cpu_has_amd_erratum(struct cpuinfo_x86 *cpu, const int *erratum) { - struct cpuinfo_x86 *cpu = __this_cpu_ptr(&cpu_info); 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; @@ -776,5 +887,3 @@ bool cpu_has_amd_erratum(const int *erratum) return false; } - -EXPORT_SYMBOL_GPL(cpu_has_amd_erratum); diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c index 46674fbb62b..03445346ee0 100644 --- a/arch/x86/kernel/cpu/bugs.c +++ b/arch/x86/kernel/cpu/bugs.c @@ -17,24 +17,6 @@ #include <asm/paravirt.h> #include <asm/alternative.h> -static int __init no_halt(char *s) -{ - WARN_ONCE(1, "\"no-hlt\" is deprecated, please use \"idle=poll\"\n"); - boot_cpu_data.hlt_works_ok = 0; - return 1; -} - -__setup("no-hlt", no_halt); - -static int __init no_387(char *s) -{ - boot_cpu_data.hard_math = 0; - write_cr0(X86_CR0_TS | X86_CR0_EM | X86_CR0_MP | read_cr0()); - return 1; -} - -__setup("no387", no_387); - static double __initdata x = 4195835.0; static double __initdata y = 3145727.0; @@ -53,22 +35,13 @@ static void __init check_fpu(void) { s32 fdiv_bug; - if (!boot_cpu_data.hard_math) { -#ifndef CONFIG_MATH_EMULATION - printk(KERN_EMERG "No coprocessor found and no math emulation present.\n"); - printk(KERN_EMERG "Giving up.\n"); - for (;;) ; -#endif - return; - } - kernel_fpu_begin(); /* * trap_init() enabled FXSR and company _before_ testing for FP * problems here. * - * Test for the divl bug.. + * Test for the divl bug: http://en.wikipedia.org/wiki/Fdiv_bug */ __asm__("fninit\n\t" "fldl %1\n\t" @@ -84,91 +57,38 @@ static void __init check_fpu(void) kernel_fpu_end(); - boot_cpu_data.fdiv_bug = fdiv_bug; - if (boot_cpu_data.fdiv_bug) - printk(KERN_WARNING "Hmm, FPU with FDIV bug.\n"); -} - -static void __init check_hlt(void) -{ - if (boot_cpu_data.x86 >= 5 || paravirt_enabled()) - return; - - printk(KERN_INFO "Checking 'hlt' instruction... "); - if (!boot_cpu_data.hlt_works_ok) { - printk("disabled\n"); - return; + if (fdiv_bug) { + set_cpu_bug(&boot_cpu_data, X86_BUG_FDIV); + pr_warn("Hmm, FPU with FDIV bug\n"); } - halt(); - halt(); - halt(); - halt(); - printk(KERN_CONT "OK.\n"); -} - -/* - * Most 386 processors have a bug where a POPAD can lock the - * machine even from user space. - */ - -static void __init check_popad(void) -{ -#ifndef CONFIG_X86_POPAD_OK - int res, inp = (int) &res; - - printk(KERN_INFO "Checking for popad bug... "); - __asm__ __volatile__( - "movl $12345678,%%eax; movl $0,%%edi; pusha; popa; movl (%%edx,%%edi),%%ecx " - : "=&a" (res) - : "d" (inp) - : "ecx", "edi"); - /* - * If this fails, it means that any user program may lock the - * CPU hard. Too bad. - */ - if (res != 12345678) - printk(KERN_CONT "Buggy.\n"); - else - printk(KERN_CONT "OK.\n"); -#endif } -/* - * Check whether we are able to run this kernel safely on SMP. - * - * - In order to run on a i386, we need to be compiled for i386 - * (for due to lack of "invlpg" and working WP on a i386) - * - In order to run on anything without a TSC, we need to be - * compiled for a i486. - */ - -static void __init check_config(void) -{ -/* - * We'd better not be a i386 if we're configured to use some - * i486+ only features! (WP works in supervisor mode and the - * new "invlpg" and "bswap" instructions) - */ -#if defined(CONFIG_X86_WP_WORKS_OK) || defined(CONFIG_X86_INVLPG) || \ - defined(CONFIG_X86_BSWAP) - if (boot_cpu_data.x86 == 3) - panic("Kernel requires i486+ for 'invlpg' and other features"); -#endif -} - - void __init check_bugs(void) { identify_boot_cpu(); #ifndef CONFIG_SMP - printk(KERN_INFO "CPU: "); + pr_info("CPU: "); print_cpu_info(&boot_cpu_data); #endif - check_config(); - check_fpu(); - check_hlt(); - check_popad(); + + /* + * Check whether we are able to run this kernel safely on SMP. + * + * - i386 is no longer supported. + * - In order to run on anything without a TSC, we need to be + * compiled for a i486. + */ + if (boot_cpu_data.x86 < 4) + panic("Kernel requires i486+ for 'invlpg' and other features"); + init_utsname()->machine[1] = '0' + (boot_cpu_data.x86 > 6 ? 6 : boot_cpu_data.x86); alternative_instructions(); + + /* + * kernel_fpu_begin/end() in check_fpu() relies on the patched + * alternative instructions. + */ + if (cpu_has_fpu) + check_fpu(); } diff --git a/arch/x86/kernel/cpu/centaur.c b/arch/x86/kernel/cpu/centaur.c index 159103c0b1f..d8fba5c15fb 100644 --- a/arch/x86/kernel/cpu/centaur.c +++ b/arch/x86/kernel/cpu/centaur.c @@ -1,6 +1,5 @@ #include <linux/bitops.h> #include <linux/kernel.h> -#include <linux/init.h> #include <asm/processor.h> #include <asm/e820.h> @@ -9,236 +8,6 @@ #include "cpu.h" -#ifdef CONFIG_X86_OOSTORE - -static u32 __cpuinit power2(u32 x) -{ - u32 s = 1; - - while (s <= x) - s <<= 1; - - return s >>= 1; -} - - -/* - * Set up an actual MCR - */ -static void __cpuinit centaur_mcr_insert(int reg, u32 base, u32 size, int key) -{ - u32 lo, hi; - - hi = base & ~0xFFF; - lo = ~(size-1); /* Size is a power of 2 so this makes a mask */ - lo &= ~0xFFF; /* Remove the ctrl value bits */ - lo |= key; /* Attribute we wish to set */ - wrmsr(reg+MSR_IDT_MCR0, lo, hi); - mtrr_centaur_report_mcr(reg, lo, hi); /* Tell the mtrr driver */ -} - -/* - * Figure what we can cover with MCR's - * - * Shortcut: We know you can't put 4Gig of RAM on a winchip - */ -static u32 __cpuinit ramtop(void) -{ - u32 clip = 0xFFFFFFFFUL; - u32 top = 0; - int i; - - for (i = 0; i < e820.nr_map; i++) { - unsigned long start, end; - - if (e820.map[i].addr > 0xFFFFFFFFUL) - continue; - /* - * Don't MCR over reserved space. Ignore the ISA hole - * we frob around that catastrophe already - */ - if (e820.map[i].type == E820_RESERVED) { - if (e820.map[i].addr >= 0x100000UL && - e820.map[i].addr < clip) - clip = e820.map[i].addr; - continue; - } - start = e820.map[i].addr; - end = e820.map[i].addr + e820.map[i].size; - if (start >= end) - continue; - if (end > top) - top = end; - } - /* - * Everything below 'top' should be RAM except for the ISA hole. - * Because of the limited MCR's we want to map NV/ACPI into our - * MCR range for gunk in RAM - * - * Clip might cause us to MCR insufficient RAM but that is an - * acceptable failure mode and should only bite obscure boxes with - * a VESA hole at 15Mb - * - * The second case Clip sometimes kicks in is when the EBDA is marked - * as reserved. Again we fail safe with reasonable results - */ - if (top > clip) - top = clip; - - return top; -} - -/* - * Compute a set of MCR's to give maximum coverage - */ -static int __cpuinit centaur_mcr_compute(int nr, int key) -{ - u32 mem = ramtop(); - u32 root = power2(mem); - u32 base = root; - u32 top = root; - u32 floor = 0; - int ct = 0; - - while (ct < nr) { - u32 fspace = 0; - u32 high; - u32 low; - - /* - * Find the largest block we will fill going upwards - */ - high = power2(mem-top); - - /* - * Find the largest block we will fill going downwards - */ - low = base/2; - - /* - * Don't fill below 1Mb going downwards as there - * is an ISA hole in the way. - */ - if (base <= 1024*1024) - low = 0; - - /* - * See how much space we could cover by filling below - * the ISA hole - */ - - if (floor == 0) - fspace = 512*1024; - else if (floor == 512*1024) - fspace = 128*1024; - - /* And forget ROM space */ - - /* - * Now install the largest coverage we get - */ - if (fspace > high && fspace > low) { - centaur_mcr_insert(ct, floor, fspace, key); - floor += fspace; - } else if (high > low) { - centaur_mcr_insert(ct, top, high, key); - top += high; - } else if (low > 0) { - base -= low; - centaur_mcr_insert(ct, base, low, key); - } else - break; - ct++; - } - /* - * We loaded ct values. We now need to set the mask. The caller - * must do this bit. - */ - return ct; -} - -static void __cpuinit centaur_create_optimal_mcr(void) -{ - int used; - int i; - - /* - * Allocate up to 6 mcrs to mark as much of ram as possible - * as write combining and weak write ordered. - * - * To experiment with: Linux never uses stack operations for - * mmio spaces so we could globally enable stack operation wc - * - * Load the registers with type 31 - full write combining, all - * writes weakly ordered. - */ - used = centaur_mcr_compute(6, 31); - - /* - * Wipe unused MCRs - */ - for (i = used; i < 8; i++) - wrmsr(MSR_IDT_MCR0+i, 0, 0); -} - -static void __cpuinit winchip2_create_optimal_mcr(void) -{ - u32 lo, hi; - int used; - int i; - - /* - * Allocate up to 6 mcrs to mark as much of ram as possible - * as write combining, weak store ordered. - * - * Load the registers with type 25 - * 8 - weak write ordering - * 16 - weak read ordering - * 1 - write combining - */ - used = centaur_mcr_compute(6, 25); - - /* - * Mark the registers we are using. - */ - rdmsr(MSR_IDT_MCR_CTRL, lo, hi); - for (i = 0; i < used; i++) - lo |= 1<<(9+i); - wrmsr(MSR_IDT_MCR_CTRL, lo, hi); - - /* - * Wipe unused MCRs - */ - - for (i = used; i < 8; i++) - wrmsr(MSR_IDT_MCR0+i, 0, 0); -} - -/* - * Handle the MCR key on the Winchip 2. - */ -static void __cpuinit winchip2_unprotect_mcr(void) -{ - u32 lo, hi; - u32 key; - - rdmsr(MSR_IDT_MCR_CTRL, lo, hi); - lo &= ~0x1C0; /* blank bits 8-6 */ - key = (lo>>17) & 7; - lo |= key<<6; /* replace with unlock key */ - wrmsr(MSR_IDT_MCR_CTRL, lo, hi); -} - -static void __cpuinit winchip2_protect_mcr(void) -{ - u32 lo, hi; - - rdmsr(MSR_IDT_MCR_CTRL, lo, hi); - lo &= ~0x1C0; /* blank bits 8-6 */ - wrmsr(MSR_IDT_MCR_CTRL, lo, hi); -} -#endif /* CONFIG_X86_OOSTORE */ - #define ACE_PRESENT (1 << 6) #define ACE_ENABLED (1 << 7) #define ACE_FCR (1 << 28) /* MSR_VIA_FCR */ @@ -247,7 +16,7 @@ static void __cpuinit winchip2_protect_mcr(void) #define RNG_ENABLED (1 << 3) #define RNG_ENABLE (1 << 6) /* MSR_VIA_RNG */ -static void __cpuinit init_c3(struct cpuinfo_x86 *c) +static void init_c3(struct cpuinfo_x86 *c) { u32 lo, hi; @@ -318,7 +87,7 @@ enum { EAMD3D = 1<<20, }; -static void __cpuinit early_init_centaur(struct cpuinfo_x86 *c) +static void early_init_centaur(struct cpuinfo_x86 *c) { switch (c->x86) { #ifdef CONFIG_X86_32 @@ -337,7 +106,7 @@ static void __cpuinit early_init_centaur(struct cpuinfo_x86 *c) #endif } -static void __cpuinit init_centaur(struct cpuinfo_x86 *c) +static void init_centaur(struct cpuinfo_x86 *c) { #ifdef CONFIG_X86_32 char *name; @@ -363,20 +132,6 @@ static void __cpuinit init_centaur(struct cpuinfo_x86 *c) fcr_clr = DPDC; printk(KERN_NOTICE "Disabling bugged TSC.\n"); clear_cpu_cap(c, X86_FEATURE_TSC); -#ifdef CONFIG_X86_OOSTORE - centaur_create_optimal_mcr(); - /* - * Enable: - * write combining on non-stack, non-string - * write combining on string, all types - * weak write ordering - * - * The C6 original lacks weak read order - * - * Note 0x120 is write only on Winchip 1 - */ - wrmsr(MSR_IDT_MCR_CTRL, 0x01F0001F, 0); -#endif break; case 8: switch (c->x86_mask) { @@ -393,40 +148,12 @@ static void __cpuinit init_centaur(struct cpuinfo_x86 *c) fcr_set = ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK| E2MMX|EAMD3D; fcr_clr = DPDC; -#ifdef CONFIG_X86_OOSTORE - winchip2_unprotect_mcr(); - winchip2_create_optimal_mcr(); - rdmsr(MSR_IDT_MCR_CTRL, lo, hi); - /* - * Enable: - * write combining on non-stack, non-string - * write combining on string, all types - * weak write ordering - */ - lo |= 31; - wrmsr(MSR_IDT_MCR_CTRL, lo, hi); - winchip2_protect_mcr(); -#endif break; case 9: name = "3"; fcr_set = ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK| E2MMX|EAMD3D; fcr_clr = DPDC; -#ifdef CONFIG_X86_OOSTORE - winchip2_unprotect_mcr(); - winchip2_create_optimal_mcr(); - rdmsr(MSR_IDT_MCR_CTRL, lo, hi); - /* - * Enable: - * write combining on non-stack, non-string - * write combining on string, all types - * weak write ordering - */ - lo |= 31; - wrmsr(MSR_IDT_MCR_CTRL, lo, hi); - winchip2_protect_mcr(); -#endif break; default: name = "??"; @@ -468,10 +195,10 @@ static void __cpuinit init_centaur(struct cpuinfo_x86 *c) #endif } -static unsigned int __cpuinit +#ifdef CONFIG_X86_32 +static unsigned int centaur_size_cache(struct cpuinfo_x86 *c, unsigned int size) { -#ifdef CONFIG_X86_32 /* VIA C3 CPUs (670-68F) need further shifting. */ if ((c->x86 == 6) && ((c->x86_model == 7) || (c->x86_model == 8))) size >>= 8; @@ -484,16 +211,18 @@ centaur_size_cache(struct cpuinfo_x86 *c, unsigned int size) if ((c->x86 == 6) && (c->x86_model == 9) && (c->x86_mask == 1) && (size == 65)) size -= 1; -#endif return size; } +#endif -static const struct cpu_dev __cpuinitconst centaur_cpu_dev = { +static const struct cpu_dev centaur_cpu_dev = { .c_vendor = "Centaur", .c_ident = { "CentaurHauls" }, .c_early_init = early_init_centaur, .c_init = init_centaur, - .c_size_cache = centaur_size_cache, +#ifdef CONFIG_X86_32 + .legacy_cache_size = centaur_size_cache, +#endif .c_x86_vendor = X86_VENDOR_CENTAUR, }; diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index c0f7d68d318..ef1b93f18ed 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -8,6 +8,7 @@ #include <linux/delay.h> #include <linux/sched.h> #include <linux/init.h> +#include <linux/kprobes.h> #include <linux/kgdb.h> #include <linux/smp.h> #include <linux/io.h> @@ -18,7 +19,9 @@ #include <asm/archrandom.h> #include <asm/hypervisor.h> #include <asm/processor.h> +#include <asm/debugreg.h> #include <asm/sections.h> +#include <asm/vsyscall.h> #include <linux/topology.h> #include <linux/cpumask.h> #include <asm/pgtable.h> @@ -28,6 +31,7 @@ #include <asm/apic.h> #include <asm/desc.h> #include <asm/i387.h> +#include <asm/fpu-internal.h> #include <asm/mtrr.h> #include <linux/numa.h> #include <asm/asm.h> @@ -35,6 +39,8 @@ #include <asm/mce.h> #include <asm/msr.h> #include <asm/pat.h> +#include <asm/microcode.h> +#include <asm/microcode_intel.h> #ifdef CONFIG_X86_LOCAL_APIC #include <asm/uv/uv.h> @@ -59,7 +65,7 @@ void __init setup_cpu_local_masks(void) alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask); } -static void __cpuinit default_init(struct cpuinfo_x86 *c) +static void default_init(struct cpuinfo_x86 *c) { #ifdef CONFIG_X86_64 cpu_detect_cache_sizes(c); @@ -76,13 +82,13 @@ static void __cpuinit default_init(struct cpuinfo_x86 *c) #endif } -static const struct cpu_dev __cpuinitconst default_cpu = { +static const struct cpu_dev default_cpu = { .c_init = default_init, .c_vendor = "Unknown", .c_x86_vendor = X86_VENDOR_UNKNOWN, }; -static const struct cpu_dev *this_cpu __cpuinitdata = &default_cpu; +static const struct cpu_dev *this_cpu = &default_cpu; DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = { #ifdef CONFIG_X86_64 @@ -142,6 +148,8 @@ static int __init x86_xsave_setup(char *s) { setup_clear_cpu_cap(X86_FEATURE_XSAVE); setup_clear_cpu_cap(X86_FEATURE_XSAVEOPT); + setup_clear_cpu_cap(X86_FEATURE_AVX); + setup_clear_cpu_cap(X86_FEATURE_AVX2); return 1; } __setup("noxsave", x86_xsave_setup); @@ -154,8 +162,8 @@ static int __init x86_xsaveopt_setup(char *s) __setup("noxsaveopt", x86_xsaveopt_setup); #ifdef CONFIG_X86_32 -static int cachesize_override __cpuinitdata = -1; -static int disable_x86_serial_nr __cpuinitdata = 1; +static int cachesize_override = -1; +static int disable_x86_serial_nr = 1; static int __init cachesize_setup(char *str) { @@ -209,12 +217,12 @@ static inline int flag_is_changeable_p(u32 flag) } /* Probe for the CPUID instruction */ -static int __cpuinit have_cpuid_p(void) +int have_cpuid_p(void) { return flag_is_changeable_p(X86_EFLAGS_ID); } -static void __cpuinit squash_the_stupid_serial_number(struct cpuinfo_x86 *c) +static void squash_the_stupid_serial_number(struct cpuinfo_x86 *c) { unsigned long lo, hi; @@ -245,32 +253,45 @@ static inline int flag_is_changeable_p(u32 flag) { return 1; } -/* Probe for the CPUID instruction */ -static inline int have_cpuid_p(void) -{ - return 1; -} static inline void squash_the_stupid_serial_number(struct cpuinfo_x86 *c) { } #endif -static int disable_smep __cpuinitdata; static __init int setup_disable_smep(char *arg) { - disable_smep = 1; + setup_clear_cpu_cap(X86_FEATURE_SMEP); return 1; } __setup("nosmep", setup_disable_smep); -static __cpuinit void setup_smep(struct cpuinfo_x86 *c) +static __always_inline void setup_smep(struct cpuinfo_x86 *c) +{ + if (cpu_has(c, X86_FEATURE_SMEP)) + set_in_cr4(X86_CR4_SMEP); +} + +static __init int setup_disable_smap(char *arg) +{ + setup_clear_cpu_cap(X86_FEATURE_SMAP); + return 1; +} +__setup("nosmap", setup_disable_smap); + +static __always_inline void setup_smap(struct cpuinfo_x86 *c) { - if (cpu_has(c, X86_FEATURE_SMEP)) { - if (unlikely(disable_smep)) { - setup_clear_cpu_cap(X86_FEATURE_SMEP); - clear_in_cr4(X86_CR4_SMEP); - } else - set_in_cr4(X86_CR4_SMEP); + unsigned long eflags; + + /* This should have been cleared long ago */ + raw_local_save_flags(eflags); + BUG_ON(eflags & X86_EFLAGS_AC); + + if (cpu_has(c, X86_FEATURE_SMAP)) { +#ifdef CONFIG_X86_SMAP + set_in_cr4(X86_CR4_SMAP); +#else + clear_in_cr4(X86_CR4_SMAP); +#endif } } @@ -284,7 +305,7 @@ struct cpuid_dependent_feature { u32 level; }; -static const struct cpuid_dependent_feature __cpuinitconst +static const struct cpuid_dependent_feature cpuid_dependent_features[] = { { X86_FEATURE_MWAIT, 0x00000005 }, { X86_FEATURE_DCA, 0x00000009 }, @@ -292,7 +313,7 @@ cpuid_dependent_features[] = { { 0, 0 } }; -static void __cpuinit filter_cpuid_features(struct cpuinfo_x86 *c, bool warn) +static void filter_cpuid_features(struct cpuinfo_x86 *c, bool warn) { const struct cpuid_dependent_feature *df; @@ -330,9 +351,10 @@ static void __cpuinit filter_cpuid_features(struct cpuinfo_x86 *c, bool warn) */ /* Look up CPU names by table lookup. */ -static const char *__cpuinit table_lookup_model(struct cpuinfo_x86 *c) +static const char *table_lookup_model(struct cpuinfo_x86 *c) { - const struct cpu_model_info *info; +#ifdef CONFIG_X86_32 + const struct legacy_cpu_model_info *info; if (c->x86_model >= 16) return NULL; /* Range check */ @@ -340,18 +362,19 @@ static const char *__cpuinit table_lookup_model(struct cpuinfo_x86 *c) if (!this_cpu) return NULL; - info = this_cpu->c_models; + info = this_cpu->legacy_models; - while (info && info->family) { + while (info->family) { if (info->family == c->x86) return info->model_names[c->x86_model]; info++; } +#endif return NULL; /* Not found */ } -__u32 cpu_caps_cleared[NCAPINTS] __cpuinitdata; -__u32 cpu_caps_set[NCAPINTS] __cpuinitdata; +__u32 cpu_caps_cleared[NCAPINTS]; +__u32 cpu_caps_set[NCAPINTS]; void load_percpu_segment(int cpu) { @@ -380,9 +403,9 @@ void switch_to_new_gdt(int cpu) load_percpu_segment(cpu); } -static const struct cpu_dev *__cpuinitdata cpu_devs[X86_VENDOR_NUM] = {}; +static const struct cpu_dev *cpu_devs[X86_VENDOR_NUM] = {}; -static void __cpuinit get_model_name(struct cpuinfo_x86 *c) +static void get_model_name(struct cpuinfo_x86 *c) { unsigned int *v; char *p, *q; @@ -411,7 +434,7 @@ static void __cpuinit get_model_name(struct cpuinfo_x86 *c) } } -void __cpuinit cpu_detect_cache_sizes(struct cpuinfo_x86 *c) +void cpu_detect_cache_sizes(struct cpuinfo_x86 *c) { unsigned int n, dummy, ebx, ecx, edx, l2size; @@ -436,8 +459,8 @@ void __cpuinit cpu_detect_cache_sizes(struct cpuinfo_x86 *c) c->x86_tlbsize += ((ebx >> 16) & 0xfff) + (ebx & 0xfff); #else /* do processor-specific cache resizing */ - if (this_cpu->c_size_cache) - l2size = this_cpu->c_size_cache(c, l2size); + if (this_cpu->legacy_cache_size) + l2size = this_cpu->legacy_cache_size(c, l2size); /* Allow user to override all this if necessary. */ if (cachesize_override != -1) @@ -450,7 +473,37 @@ void __cpuinit cpu_detect_cache_sizes(struct cpuinfo_x86 *c) c->x86_cache_size = l2size; } -void __cpuinit detect_ht(struct cpuinfo_x86 *c) +u16 __read_mostly tlb_lli_4k[NR_INFO]; +u16 __read_mostly tlb_lli_2m[NR_INFO]; +u16 __read_mostly tlb_lli_4m[NR_INFO]; +u16 __read_mostly tlb_lld_4k[NR_INFO]; +u16 __read_mostly tlb_lld_2m[NR_INFO]; +u16 __read_mostly tlb_lld_4m[NR_INFO]; +u16 __read_mostly tlb_lld_1g[NR_INFO]; + +/* + * tlb_flushall_shift shows the balance point in replacing cr3 write + * with multiple 'invlpg'. It will do this replacement when + * flush_tlb_lines <= active_lines/2^tlb_flushall_shift. + * If tlb_flushall_shift is -1, means the replacement will be disabled. + */ +s8 __read_mostly tlb_flushall_shift = -1; + +void cpu_detect_tlb(struct cpuinfo_x86 *c) +{ + if (this_cpu->c_detect_tlb) + this_cpu->c_detect_tlb(c); + + printk(KERN_INFO "Last level iTLB entries: 4KB %d, 2MB %d, 4MB %d\n" + "Last level dTLB entries: 4KB %d, 2MB %d, 4MB %d, 1GB %d\n" + "tlb_flushall_shift: %d\n", + tlb_lli_4k[ENTRIES], tlb_lli_2m[ENTRIES], + tlb_lli_4m[ENTRIES], tlb_lld_4k[ENTRIES], + tlb_lld_2m[ENTRIES], tlb_lld_4m[ENTRIES], + tlb_lld_1g[ENTRIES], tlb_flushall_shift); +} + +void detect_ht(struct cpuinfo_x86 *c) { #ifdef CONFIG_X86_HT u32 eax, ebx, ecx, edx; @@ -501,7 +554,7 @@ out: #endif } -static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c) +static void get_cpu_vendor(struct cpuinfo_x86 *c) { char *v = c->x86_vendor_id; int i; @@ -528,7 +581,7 @@ static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c) this_cpu = &default_cpu; } -void __cpuinit cpu_detect(struct cpuinfo_x86 *c) +void cpu_detect(struct cpuinfo_x86 *c) { /* Get vendor name */ cpuid(0x00000000, (unsigned int *)&c->cpuid_level, @@ -558,7 +611,7 @@ void __cpuinit cpu_detect(struct cpuinfo_x86 *c) } } -void __cpuinit get_cpu_cap(struct cpuinfo_x86 *c) +void get_cpu_cap(struct cpuinfo_x86 *c) { u32 tfms, xlvl; u32 ebx; @@ -609,7 +662,7 @@ void __cpuinit get_cpu_cap(struct cpuinfo_x86 *c) init_scattered_cpuid_features(c); } -static void __cpuinit identify_cpu_without_cpuid(struct cpuinfo_x86 *c) +static void identify_cpu_without_cpuid(struct cpuinfo_x86 *c) { #ifdef CONFIG_X86_32 int i; @@ -668,10 +721,9 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c) return; cpu_detect(c); - get_cpu_vendor(c); - get_cpu_cap(c); + fpu_detect(c); if (this_cpu->c_early_init) this_cpu->c_early_init(c); @@ -679,10 +731,10 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c) c->cpu_index = 0; filter_cpuid_features(c, false); - setup_smep(c); - if (this_cpu->c_bsp_init) this_cpu->c_bsp_init(c); + + setup_force_cpu_cap(X86_FEATURE_ALWAYS); } void __init early_cpu_init(void) @@ -727,7 +779,7 @@ void __init early_cpu_init(void) * unless we can find a reliable way to detect all the broken cases. * Enable it explicitly on 64-bit for non-constant inputs of cpu_has(). */ -static void __cpuinit detect_nopl(struct cpuinfo_x86 *c) +static void detect_nopl(struct cpuinfo_x86 *c) { #ifdef CONFIG_X86_32 clear_cpu_cap(c, X86_FEATURE_NOPL); @@ -736,7 +788,7 @@ static void __cpuinit detect_nopl(struct cpuinfo_x86 *c) #endif } -static void __cpuinit generic_identify(struct cpuinfo_x86 *c) +static void generic_identify(struct cpuinfo_x86 *c) { c->extended_cpuid_level = 0; @@ -765,8 +817,6 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 *c) c->phys_proc_id = c->initial_apicid; } - setup_smep(c); - get_model_name(c); /* Default name */ detect_nopl(c); @@ -775,7 +825,7 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 *c) /* * This does the hard work of actually picking apart the CPU stuff... */ -static void __cpuinit identify_cpu(struct cpuinfo_x86 *c) +static void identify_cpu(struct cpuinfo_x86 *c) { int i; @@ -831,6 +881,10 @@ static void __cpuinit identify_cpu(struct cpuinfo_x86 *c) /* Disable the PN if appropriate */ squash_the_stupid_serial_number(c); + /* Set up SMEP/SMAP */ + setup_smep(c); + setup_smap(c); + /* * The vendor-specific functions might have changed features. * Now we do "generic changes." @@ -877,6 +931,10 @@ static void __cpuinit identify_cpu(struct cpuinfo_x86 *c) /* AND the already accumulated flags with these */ for (i = 0; i < NCAPINTS; i++) boot_cpu_data.x86_capability[i] &= c->x86_capability[i]; + + /* OR, i.e. replicate the bug flags */ + for (i = NCAPINTS; i < NCAPINTS + NBUGINTS; i++) + c->x86_capability[i] |= boot_cpu_data.x86_capability[i]; } /* Init Machine Check Exception if available. */ @@ -897,6 +955,38 @@ static void vgetcpu_set_mode(void) else vgetcpu_mode = VGETCPU_LSL; } + +/* May not be __init: called during resume */ +static void syscall32_cpu_init(void) +{ + /* Load these always in case some future AMD CPU supports + SYSENTER from compat mode too. */ + wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)__KERNEL_CS); + wrmsrl_safe(MSR_IA32_SYSENTER_ESP, 0ULL); + wrmsrl_safe(MSR_IA32_SYSENTER_EIP, (u64)ia32_sysenter_target); + + wrmsrl(MSR_CSTAR, ia32_cstar_target); +} +#endif + +#ifdef CONFIG_X86_32 +void enable_sep_cpu(void) +{ + int cpu = get_cpu(); + struct tss_struct *tss = &per_cpu(init_tss, cpu); + + if (!boot_cpu_has(X86_FEATURE_SEP)) { + put_cpu(); + return; + } + + tss->x86_tss.ss1 = __KERNEL_CS; + tss->x86_tss.sp1 = sizeof(struct tss_struct) + (unsigned long) tss; + wrmsr(MSR_IA32_SYSENTER_CS, __KERNEL_CS, 0); + wrmsr(MSR_IA32_SYSENTER_ESP, tss->x86_tss.sp1, 0); + wrmsr(MSR_IA32_SYSENTER_EIP, (unsigned long) ia32_sysenter_target, 0); + put_cpu(); +} #endif void __init identify_boot_cpu(void) @@ -909,9 +999,10 @@ void __init identify_boot_cpu(void) #else vgetcpu_set_mode(); #endif + cpu_detect_tlb(&boot_cpu_data); } -void __cpuinit identify_secondary_cpu(struct cpuinfo_x86 *c) +void identify_secondary_cpu(struct cpuinfo_x86 *c) { BUG_ON(c == &boot_cpu_data); identify_cpu(c); @@ -926,14 +1017,14 @@ struct msr_range { unsigned max; }; -static const struct msr_range msr_range_array[] __cpuinitconst = { +static const struct msr_range msr_range_array[] = { { 0x00000000, 0x00000418}, { 0xc0000000, 0xc000040b}, { 0xc0010000, 0xc0010142}, { 0xc0011000, 0xc001103b}, }; -static void __cpuinit print_cpu_msr(void) +static void __print_cpu_msr(void) { unsigned index_min, index_max; unsigned index; @@ -945,14 +1036,14 @@ static void __cpuinit print_cpu_msr(void) index_max = msr_range_array[i].max; for (index = index_min; index < index_max; index++) { - if (rdmsrl_amd_safe(index, &val)) + if (rdmsrl_safe(index, &val)) continue; printk(KERN_INFO " MSR%08x: %016llx\n", index, val); } } } -static int show_msr __cpuinitdata; +static int show_msr; static __init int setup_show_msr(char *arg) { @@ -968,12 +1059,13 @@ __setup("show_msr=", setup_show_msr); static __init int setup_noclflush(char *arg) { - setup_clear_cpu_cap(X86_FEATURE_CLFLSH); + setup_clear_cpu_cap(X86_FEATURE_CLFLUSH); + setup_clear_cpu_cap(X86_FEATURE_CLFLUSHOPT); return 1; } __setup("noclflush", setup_noclflush); -void __cpuinit print_cpu_info(struct cpuinfo_x86 *c) +void print_cpu_info(struct cpuinfo_x86 *c) { const char *vendor = NULL; @@ -988,22 +1080,24 @@ void __cpuinit print_cpu_info(struct cpuinfo_x86 *c) printk(KERN_CONT "%s ", vendor); if (c->x86_model_id[0]) - printk(KERN_CONT "%s", c->x86_model_id); + printk(KERN_CONT "%s", strim(c->x86_model_id)); else printk(KERN_CONT "%d86", c->x86); + printk(KERN_CONT " (fam: %02x, model: %02x", c->x86, c->x86_model); + if (c->x86_mask || c->cpuid_level >= 0) - printk(KERN_CONT " stepping %02x\n", c->x86_mask); + printk(KERN_CONT ", stepping: %02x)\n", c->x86_mask); else - printk(KERN_CONT "\n"); + printk(KERN_CONT ")\n"); -#ifdef CONFIG_SMP + print_cpu_msr(c); +} + +void print_cpu_msr(struct cpuinfo_x86 *c) +{ if (c->cpu_index < show_msr) - print_cpu_msr(); -#else - if (show_msr) - print_cpu_msr(); -#endif + __print_cpu_msr(); } static __init int setup_disablecpuid(char *arg) @@ -1019,13 +1113,17 @@ static __init int setup_disablecpuid(char *arg) } __setup("clearcpuid=", setup_disablecpuid); +DEFINE_PER_CPU(unsigned long, kernel_stack) = + (unsigned long)&init_thread_union - KERNEL_STACK_OFFSET + THREAD_SIZE; +EXPORT_PER_CPU_SYMBOL(kernel_stack); + #ifdef CONFIG_X86_64 struct desc_ptr idt_descr = { NR_VECTORS * 16 - 1, (unsigned long) idt_table }; -struct desc_ptr nmi_idt_descr = { NR_VECTORS * 16 - 1, - (unsigned long) nmi_idt_table }; +struct desc_ptr debug_idt_descr = { NR_VECTORS * 16 - 1, + (unsigned long) debug_idt_table }; DEFINE_PER_CPU_FIRST(union irq_stack_union, - irq_stack_union) __aligned(PAGE_SIZE); + irq_stack_union) __aligned(PAGE_SIZE) __visible; /* * The following four percpu variables are hot. Align current_task to @@ -1035,17 +1133,15 @@ DEFINE_PER_CPU(struct task_struct *, current_task) ____cacheline_aligned = &init_task; EXPORT_PER_CPU_SYMBOL(current_task); -DEFINE_PER_CPU(unsigned long, kernel_stack) = - (unsigned long)&init_thread_union - KERNEL_STACK_OFFSET + THREAD_SIZE; -EXPORT_PER_CPU_SYMBOL(kernel_stack); - DEFINE_PER_CPU(char *, irq_stack_ptr) = init_per_cpu_var(irq_stack_union.irq_stack) + IRQ_STACK_SIZE - 64; -DEFINE_PER_CPU(unsigned int, irq_count) = -1; +DEFINE_PER_CPU(unsigned int, irq_count) __visible = -1; + +DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT; +EXPORT_PER_CPU_SYMBOL(__preempt_count); DEFINE_PER_CPU(struct task_struct *, fpu_owner_task); -EXPORT_PER_CPU_SYMBOL(fpu_owner_task); /* * Special IST stacks which the CPU switches to when it calls @@ -1079,11 +1175,10 @@ void syscall_init(void) /* Flags to clear on syscall */ wrmsrl(MSR_SYSCALL_MASK, - X86_EFLAGS_TF|X86_EFLAGS_DF|X86_EFLAGS_IF|X86_EFLAGS_IOPL); + X86_EFLAGS_TF|X86_EFLAGS_DF|X86_EFLAGS_IF| + X86_EFLAGS_IOPL|X86_EFLAGS_AC); } -unsigned long kernel_eflags; - /* * Copies of the original ist values from the tss are only accessed during * debugging, no special alignment required. @@ -1099,37 +1194,38 @@ int is_debug_stack(unsigned long addr) (addr <= __get_cpu_var(debug_stack_addr) && addr > (__get_cpu_var(debug_stack_addr) - DEBUG_STKSZ)); } +NOKPROBE_SYMBOL(is_debug_stack); + +DEFINE_PER_CPU(u32, debug_idt_ctr); void debug_stack_set_zero(void) { - load_idt((const struct desc_ptr *)&nmi_idt_descr); + this_cpu_inc(debug_idt_ctr); + load_current_idt(); } +NOKPROBE_SYMBOL(debug_stack_set_zero); void debug_stack_reset(void) { - load_idt((const struct desc_ptr *)&idt_descr); + if (WARN_ON(!this_cpu_read(debug_idt_ctr))) + return; + if (this_cpu_dec_return(debug_idt_ctr) == 0) + load_current_idt(); } +NOKPROBE_SYMBOL(debug_stack_reset); #else /* CONFIG_X86_64 */ DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; EXPORT_PER_CPU_SYMBOL(current_task); +DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT; +EXPORT_PER_CPU_SYMBOL(__preempt_count); DEFINE_PER_CPU(struct task_struct *, fpu_owner_task); -EXPORT_PER_CPU_SYMBOL(fpu_owner_task); #ifdef CONFIG_CC_STACKPROTECTOR DEFINE_PER_CPU_ALIGNED(struct stack_canary, stack_canary); #endif -/* Make sure %fs and %gs are initialized properly in idle threads */ -struct pt_regs * __cpuinit idle_regs(struct pt_regs *regs) -{ - memset(regs, 0, sizeof(struct pt_regs)); - regs->fs = __KERNEL_PERCPU; - regs->gs = __KERNEL_STACK_CANARY; - - return regs; -} #endif /* CONFIG_X86_64 */ /* @@ -1163,15 +1259,6 @@ static void dbg_restore_debug_regs(void) #endif /* ! CONFIG_KGDB */ /* - * Prints an error where the NUMA and configured core-number mismatch and the - * platform didn't override this to fix it up - */ -void __cpuinit x86_default_fixup_cpu_id(struct cpuinfo_x86 *c, int node) -{ - pr_err("NUMA core number %d differs from configured core number %d\n", node, c->phys_proc_id); -} - -/* * cpu_init() initializes state that is per-CPU. Some data is already * initialized (naturally) in the bootstrap process, such as the GDT * and IDT. We reload them nevertheless, this function acts as a @@ -1180,7 +1267,7 @@ void __cpuinit x86_default_fixup_cpu_id(struct cpuinfo_x86 *c, int node) */ #ifdef CONFIG_X86_64 -void __cpuinit cpu_init(void) +void cpu_init(void) { struct orig_ist *oist; struct task_struct *me; @@ -1189,12 +1276,18 @@ void __cpuinit cpu_init(void) int cpu; int i; + /* + * Load microcode on this cpu if a valid microcode is available. + * This is early microcode loading procedure. + */ + load_ucode_ap(); + cpu = stack_smp_processor_id(); t = &per_cpu(init_tss, cpu); oist = &per_cpu(orig_ist, cpu); #ifdef CONFIG_NUMA - if (cpu != 0 && percpu_read(numa_node) == 0 && + if (this_cpu_read(numa_node) == 0 && early_cpu_to_node(cpu) != NUMA_NO_NODE) set_numa_node(early_cpu_to_node(cpu)); #endif @@ -1216,7 +1309,7 @@ void __cpuinit cpu_init(void) switch_to_new_gdt(cpu); loadsegment(fs, 0); - load_idt((const struct desc_ptr *)&idt_descr); + load_current_idt(); memset(me->thread.tls_array, 0, GDT_ENTRY_TLS_ENTRIES * 8); syscall_init(); @@ -1226,8 +1319,7 @@ void __cpuinit cpu_init(void) barrier(); x86_configure_nx(); - if (cpu != 0) - enable_x2apic(); + enable_x2apic(); /* * set up and load the per-CPU TSS @@ -1267,9 +1359,6 @@ void __cpuinit cpu_init(void) dbg_restore_debug_regs(); fpu_init(); - xsave_init(); - - raw_local_save_flags(kernel_eflags); if (is_uv_system()) uv_cpu_init(); @@ -1277,13 +1366,15 @@ void __cpuinit cpu_init(void) #else -void __cpuinit cpu_init(void) +void cpu_init(void) { int cpu = smp_processor_id(); struct task_struct *curr = current; struct tss_struct *t = &per_cpu(init_tss, cpu); struct thread_struct *thread = &curr->thread; + show_ucode_info_early(); + if (cpumask_test_and_set_cpu(cpu, cpu_initialized_mask)) { printk(KERN_WARNING "CPU#%d already initialized!\n", cpu); for (;;) @@ -1295,7 +1386,7 @@ void __cpuinit cpu_init(void) if (cpu_has_vme || cpu_has_tsc || cpu_has_de) clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE); - load_idt(&idt_descr); + load_current_idt(); switch_to_new_gdt(cpu); /* @@ -1322,6 +1413,19 @@ void __cpuinit cpu_init(void) dbg_restore_debug_regs(); fpu_init(); - xsave_init(); } #endif + +#ifdef CONFIG_X86_DEBUG_STATIC_CPU_HAS +void warn_pre_alternatives(void) +{ + WARN(1, "You're using static_cpu_has before alternatives have run!\n"); +} +EXPORT_SYMBOL_GPL(warn_pre_alternatives); +#endif + +inline bool __static_cpu_has_safe(u16 bit) +{ + return boot_cpu_has(bit); +} +EXPORT_SYMBOL_GPL(__static_cpu_has_safe); diff --git a/arch/x86/kernel/cpu/cpu.h b/arch/x86/kernel/cpu/cpu.h index 8bacc7826fb..c37dc37e831 100644 --- a/arch/x86/kernel/cpu/cpu.h +++ b/arch/x86/kernel/cpu/cpu.h @@ -1,12 +1,6 @@ #ifndef ARCH_X86_CPU_H #define ARCH_X86_CPU_H -struct cpu_model_info { - int vendor; - int family; - const char *model_names[16]; -}; - /* attempt to consolidate cpu attributes */ struct cpu_dev { const char *c_vendor; @@ -14,14 +8,31 @@ struct cpu_dev { /* some have two possibilities for cpuid string */ const char *c_ident[2]; - struct cpu_model_info c_models[4]; - void (*c_early_init)(struct cpuinfo_x86 *); void (*c_bsp_init)(struct cpuinfo_x86 *); void (*c_init)(struct cpuinfo_x86 *); void (*c_identify)(struct cpuinfo_x86 *); - unsigned int (*c_size_cache)(struct cpuinfo_x86 *, unsigned int); + void (*c_detect_tlb)(struct cpuinfo_x86 *); int c_x86_vendor; +#ifdef CONFIG_X86_32 + /* Optional vendor specific routine to obtain the cache size. */ + unsigned int (*legacy_cache_size)(struct cpuinfo_x86 *, + unsigned int); + + /* Family/stepping-based lookup table for model names. */ + struct legacy_cpu_model_info { + int family; + const char *model_names[16]; + } legacy_models[5]; +#endif +}; + +struct _tlb_table { + unsigned char descriptor; + char tlb_type; + unsigned int entries; + /* unsigned int ways; */ + char info[128]; }; #define cpu_dev_register(cpu_devX) \ diff --git a/arch/x86/kernel/cpu/cyrix.c b/arch/x86/kernel/cpu/cyrix.c index 4fbd384fb64..aaf152e7963 100644 --- a/arch/x86/kernel/cpu/cyrix.c +++ b/arch/x86/kernel/cpu/cyrix.c @@ -1,4 +1,3 @@ -#include <linux/init.h> #include <linux/bitops.h> #include <linux/delay.h> #include <linux/pci.h> @@ -15,7 +14,7 @@ /* * Read NSC/Cyrix DEVID registers (DIR) to get more detailed info. about the CPU */ -static void __cpuinit __do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) +static void __do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) { unsigned char ccr2, ccr3; @@ -44,7 +43,7 @@ static void __cpuinit __do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) } } -static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) +static void do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) { unsigned long flags; @@ -59,25 +58,25 @@ static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) * Actually since bugs.h doesn't even reference this perhaps someone should * fix the documentation ??? */ -static unsigned char Cx86_dir0_msb __cpuinitdata = 0; +static unsigned char Cx86_dir0_msb = 0; -static const char __cpuinitconst Cx86_model[][9] = { +static const char Cx86_model[][9] = { "Cx486", "Cx486", "5x86 ", "6x86", "MediaGX ", "6x86MX ", "M II ", "Unknown" }; -static const char __cpuinitconst Cx486_name[][5] = { +static const char Cx486_name[][5] = { "SLC", "DLC", "SLC2", "DLC2", "SRx", "DRx", "SRx2", "DRx2" }; -static const char __cpuinitconst Cx486S_name[][4] = { +static const char Cx486S_name[][4] = { "S", "S2", "Se", "S2e" }; -static const char __cpuinitconst Cx486D_name[][4] = { +static const char Cx486D_name[][4] = { "DX", "DX2", "?", "?", "?", "DX4" }; -static char Cx86_cb[] __cpuinitdata = "?.5x Core/Bus Clock"; -static const char __cpuinitconst cyrix_model_mult1[] = "12??43"; -static const char __cpuinitconst cyrix_model_mult2[] = "12233445"; +static char Cx86_cb[] = "?.5x Core/Bus Clock"; +static const char cyrix_model_mult1[] = "12??43"; +static const char cyrix_model_mult2[] = "12233445"; /* * Reset the slow-loop (SLOP) bit on the 686(L) which is set by some old @@ -87,7 +86,7 @@ static const char __cpuinitconst cyrix_model_mult2[] = "12233445"; * FIXME: our newer udelay uses the tsc. We don't need to frob with SLOP */ -static void __cpuinit check_cx686_slop(struct cpuinfo_x86 *c) +static void check_cx686_slop(struct cpuinfo_x86 *c) { unsigned long flags; @@ -112,7 +111,7 @@ static void __cpuinit check_cx686_slop(struct cpuinfo_x86 *c) } -static void __cpuinit set_cx86_reorder(void) +static void set_cx86_reorder(void) { u8 ccr3; @@ -127,7 +126,7 @@ static void __cpuinit set_cx86_reorder(void) setCx86(CX86_CCR3, ccr3); } -static void __cpuinit set_cx86_memwb(void) +static void set_cx86_memwb(void) { printk(KERN_INFO "Enable Memory-Write-back mode on Cyrix/NSC processor.\n"); @@ -143,7 +142,7 @@ static void __cpuinit set_cx86_memwb(void) * Configure later MediaGX and/or Geode processor. */ -static void __cpuinit geode_configure(void) +static void geode_configure(void) { unsigned long flags; u8 ccr3; @@ -166,7 +165,7 @@ static void __cpuinit geode_configure(void) local_irq_restore(flags); } -static void __cpuinit early_init_cyrix(struct cpuinfo_x86 *c) +static void early_init_cyrix(struct cpuinfo_x86 *c) { unsigned char dir0, dir0_msn, dir1 = 0; @@ -185,7 +184,7 @@ static void __cpuinit early_init_cyrix(struct cpuinfo_x86 *c) } } -static void __cpuinit init_cyrix(struct cpuinfo_x86 *c) +static void init_cyrix(struct cpuinfo_x86 *c) { unsigned char dir0, dir0_msn, dir0_lsn, dir1 = 0; char *buf = c->x86_model_id; @@ -249,7 +248,7 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c) /* Emulate MTRRs using Cyrix's ARRs. */ set_cpu_cap(c, X86_FEATURE_CYRIX_ARR); /* 6x86's contain this bug */ - c->coma_bug = 1; + set_cpu_bug(c, X86_BUG_COMA); break; case 4: /* MediaGX/GXm or Geode GXM/GXLV/GX1 */ @@ -317,7 +316,8 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c) /* Enable MMX extensions (App note 108) */ setCx86_old(CX86_CCR7, getCx86_old(CX86_CCR7)|1); } else { - c->coma_bug = 1; /* 6x86MX, it has the bug. */ + /* A 6x86MX - it has the bug. */ + set_cpu_bug(c, X86_BUG_COMA); } tmp = (!(dir0_lsn & 7) || dir0_lsn & 1) ? 2 : 0; Cx86_cb[tmp] = cyrix_model_mult2[dir0_lsn & 7]; @@ -332,7 +332,7 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c) switch (dir0_lsn) { case 0xd: /* either a 486SLC or DLC w/o DEVID */ dir0_msn = 0; - p = Cx486_name[(c->hard_math) ? 1 : 0]; + p = Cx486_name[(cpu_has_fpu ? 1 : 0)]; break; case 0xe: /* a 486S A step */ @@ -355,7 +355,7 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c) /* * Handle National Semiconductor branded processors */ -static void __cpuinit init_nsc(struct cpuinfo_x86 *c) +static void init_nsc(struct cpuinfo_x86 *c) { /* * There may be GX1 processors in the wild that are branded @@ -404,7 +404,7 @@ static inline int test_cyrix_52div(void) return (unsigned char) (test >> 8) == 0x02; } -static void __cpuinit cyrix_identify(struct cpuinfo_x86 *c) +static void cyrix_identify(struct cpuinfo_x86 *c) { /* Detect Cyrix with disabled CPUID */ if (c->x86 == 4 && test_cyrix_52div()) { @@ -440,7 +440,7 @@ static void __cpuinit cyrix_identify(struct cpuinfo_x86 *c) } } -static const struct cpu_dev __cpuinitconst cyrix_cpu_dev = { +static const struct cpu_dev cyrix_cpu_dev = { .c_vendor = "Cyrix", .c_ident = { "CyrixInstead" }, .c_early_init = early_init_cyrix, @@ -451,7 +451,7 @@ static const struct cpu_dev __cpuinitconst cyrix_cpu_dev = { cpu_dev_register(cyrix_cpu_dev); -static const struct cpu_dev __cpuinitconst nsc_cpu_dev = { +static const struct cpu_dev nsc_cpu_dev = { .c_vendor = "NSC", .c_ident = { "Geode by NSC" }, .c_init = init_nsc, diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c index 755f64fb074..36ce402a3fa 100644 --- a/arch/x86/kernel/cpu/hypervisor.c +++ b/arch/x86/kernel/cpu/hypervisor.c @@ -25,11 +25,6 @@ #include <asm/processor.h> #include <asm/hypervisor.h> -/* - * Hypervisor detect order. This is specified explicitly here because - * some hypervisors might implement compatibility modes for other - * hypervisors and therefore need to be detected in specific sequence. - */ static const __initconst struct hypervisor_x86 * const hypervisors[] = { #ifdef CONFIG_XEN_PVHVM @@ -37,6 +32,9 @@ static const __initconst struct hypervisor_x86 * const hypervisors[] = #endif &x86_hyper_vmware, &x86_hyper_ms_hyperv, +#ifdef CONFIG_KVM_GUEST + &x86_hyper_kvm, +#endif }; const struct hypervisor_x86 *x86_hyper; @@ -46,18 +44,22 @@ static inline void __init detect_hypervisor_vendor(void) { const struct hypervisor_x86 *h, * const *p; + uint32_t pri, max_pri = 0; for (p = hypervisors; p < hypervisors + ARRAY_SIZE(hypervisors); p++) { h = *p; - if (h->detect()) { + pri = h->detect(); + if (pri != 0 && pri > max_pri) { + max_pri = pri; x86_hyper = h; - printk(KERN_INFO "Hypervisor detected: %s\n", h->name); - break; } } + + if (max_pri) + printk(KERN_INFO "Hypervisor detected: %s\n", x86_hyper->name); } -void __cpuinit init_hypervisor(struct cpuinfo_x86 *c) +void init_hypervisor(struct cpuinfo_x86 *c) { if (x86_hyper && x86_hyper->set_cpu_features) x86_hyper->set_cpu_features(c); @@ -76,3 +78,10 @@ void __init init_hypervisor_platform(void) if (x86_hyper->init_platform) x86_hyper->init_platform(); } + +bool __init hypervisor_x2apic_available(void) +{ + return x86_hyper && + x86_hyper->x2apic_available && + x86_hyper->x2apic_available(); +} diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c index 3e6ff6cbf42..f9e4fdd3b87 100644 --- a/arch/x86/kernel/cpu/intel.c +++ b/arch/x86/kernel/cpu/intel.c @@ -1,4 +1,3 @@ -#include <linux/init.h> #include <linux/kernel.h> #include <linux/string.h> @@ -17,7 +16,6 @@ #ifdef CONFIG_X86_64 #include <linux/topology.h> -#include <asm/numa_64.h> #endif #include "cpu.h" @@ -27,17 +25,14 @@ #include <asm/apic.h> #endif -static void __cpuinit early_init_intel(struct cpuinfo_x86 *c) +static void early_init_intel(struct cpuinfo_x86 *c) { u64 misc_enable; /* Unmask CPUID levels if masked: */ if (c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xd)) { - rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable); - - if (misc_enable & MSR_IA32_MISC_ENABLE_LIMIT_CPUID) { - misc_enable &= ~MSR_IA32_MISC_ENABLE_LIMIT_CPUID; - wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable); + if (msr_clear_bit(MSR_IA32_MISC_ENABLE, + MSR_IA32_MISC_ENABLE_LIMIT_CPUID_BIT) > 0) { c->cpuid_level = cpuid_eax(0); get_cpu_cap(c); } @@ -94,7 +89,19 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c) set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC); if (!check_tsc_unstable()) - sched_clock_stable = 1; + set_sched_clock_stable(); + } + + /* Penwell and Cloverview have the TSC which doesn't sleep on S3 */ + if (c->x86 == 6) { + switch (c->x86_model) { + case 0x27: /* Penwell */ + case 0x35: /* Cloverview */ + set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC_S3); + break; + default: + break; + } } /* @@ -119,16 +126,10 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c) * Ingo Molnar reported a Pentium D (model 6) and a Xeon * (model 2) with the same problem. */ - if (c->x86 == 15) { - rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable); - - if (misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING) { - printk(KERN_INFO "kmemcheck: Disabling fast string operations\n"); - - misc_enable &= ~MSR_IA32_MISC_ENABLE_FAST_STRING; - wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable); - } - } + if (c->x86 == 15) + if (msr_clear_bit(MSR_IA32_MISC_ENABLE, + MSR_IA32_MISC_ENABLE_FAST_STRING_BIT) > 0) + pr_info("kmemcheck: Disabling fast string operations\n"); #endif /* @@ -152,7 +153,7 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c) * This is called before we do cpu ident work */ -int __cpuinit ppro_with_ram_bug(void) +int ppro_with_ram_bug(void) { /* Uses data from early_cpu_detect now */ if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL && @@ -165,21 +166,7 @@ int __cpuinit ppro_with_ram_bug(void) return 0; } -#ifdef CONFIG_X86_F00F_BUG -static void __cpuinit trap_init_f00f_bug(void) -{ - __set_fixmap(FIX_F00F_IDT, __pa(&idt_table), PAGE_KERNEL_RO); - - /* - * Update the IDT descriptor and reload the IDT so that - * it uses the read-only mapped virtual address. - */ - idt_descr.address = fix_to_virt(FIX_F00F_IDT); - load_idt(&idt_descr); -} -#endif - -static void __cpuinit intel_smp_check(struct cpuinfo_x86 *c) +static void intel_smp_check(struct cpuinfo_x86 *c) { /* calling is from identify_secondary_cpu() ? */ if (!c->cpu_index) @@ -199,24 +186,28 @@ static void __cpuinit intel_smp_check(struct cpuinfo_x86 *c) } } -static void __cpuinit intel_workarounds(struct cpuinfo_x86 *c) +static int forcepae; +static int __init forcepae_setup(char *__unused) { - unsigned long lo, hi; + forcepae = 1; + return 1; +} +__setup("forcepae", forcepae_setup); +static void intel_workarounds(struct cpuinfo_x86 *c) +{ #ifdef CONFIG_X86_F00F_BUG /* * All current models of Pentium and Pentium with MMX technology CPUs * have the F0 0F bug, which lets nonprivileged users lock up the - * system. - * Note that the workaround only should be initialized once... + * system. Announce that the fault handler will be checking for it. */ - c->f00f_bug = 0; + clear_cpu_bug(c, X86_BUG_F00F); if (!paravirt_enabled() && c->x86 == 5) { static int f00f_workaround_enabled; - c->f00f_bug = 1; + set_cpu_bug(c, X86_BUG_F00F); if (!f00f_workaround_enabled) { - trap_init_f00f_bug(); printk(KERN_NOTICE "Intel Pentium with F0 0F bug - workaround enabled.\n"); f00f_workaround_enabled = 1; } @@ -231,16 +222,26 @@ static void __cpuinit intel_workarounds(struct cpuinfo_x86 *c) clear_cpu_cap(c, X86_FEATURE_SEP); /* + * PAE CPUID issue: many Pentium M report no PAE but may have a + * functionally usable PAE implementation. + * Forcefully enable PAE if kernel parameter "forcepae" is present. + */ + if (forcepae) { + printk(KERN_WARNING "PAE forced!\n"); + set_cpu_cap(c, X86_FEATURE_PAE); + add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_NOW_UNRELIABLE); + } + + /* * P4 Xeon errata 037 workaround. * Hardware prefetcher may cause stale data to be loaded into the cache. */ if ((c->x86 == 15) && (c->x86_model == 1) && (c->x86_mask == 1)) { - rdmsr(MSR_IA32_MISC_ENABLE, lo, hi); - if ((lo & MSR_IA32_MISC_ENABLE_PREFETCH_DISABLE) == 0) { - printk (KERN_INFO "CPU: C0 stepping P4 Xeon detected.\n"); - printk (KERN_INFO "CPU: Disabling hardware prefetching (Errata 037)\n"); - lo |= MSR_IA32_MISC_ENABLE_PREFETCH_DISABLE; - wrmsr(MSR_IA32_MISC_ENABLE, lo, hi); + if (msr_set_bit(MSR_IA32_MISC_ENABLE, + MSR_IA32_MISC_ENABLE_PREFETCH_DISABLE_BIT) + > 0) { + pr_info("CPU: C0 stepping P4 Xeon detected.\n"); + pr_info("CPU: Disabling hardware prefetching (Errata 037)\n"); } } @@ -273,19 +274,15 @@ static void __cpuinit intel_workarounds(struct cpuinfo_x86 *c) } #endif -#ifdef CONFIG_X86_NUMAQ - numaq_tsc_disable(); -#endif - intel_smp_check(c); } #else -static void __cpuinit intel_workarounds(struct cpuinfo_x86 *c) +static void intel_workarounds(struct cpuinfo_x86 *c) { } #endif -static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) +static void srat_detect_node(struct cpuinfo_x86 *c) { #ifdef CONFIG_NUMA unsigned node; @@ -305,7 +302,7 @@ static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) /* * find out the number of processor cores on the die */ -static int __cpuinit intel_num_cpu_cores(struct cpuinfo_x86 *c) +static int intel_num_cpu_cores(struct cpuinfo_x86 *c) { unsigned int eax, ebx, ecx, edx; @@ -320,7 +317,7 @@ static int __cpuinit intel_num_cpu_cores(struct cpuinfo_x86 *c) return 1; } -static void __cpuinit detect_vmx_virtcap(struct cpuinfo_x86 *c) +static void detect_vmx_virtcap(struct cpuinfo_x86 *c) { /* Intel VMX MSR indicated features */ #define X86_VMX_FEATURE_PROC_CTLS_TPR_SHADOW 0x00200000 @@ -358,7 +355,7 @@ static void __cpuinit detect_vmx_virtcap(struct cpuinfo_x86 *c) } } -static void __cpuinit init_intel(struct cpuinfo_x86 *c) +static void init_intel(struct cpuinfo_x86 *c) { unsigned int l2 = 0; @@ -373,6 +370,17 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c) */ detect_extended_topology(c); + if (!cpu_has(c, X86_FEATURE_XTOPOLOGY)) { + /* + * let's use the legacy cpuid vector 0x1 and 0x4 for topology + * detection. + */ + c->x86_max_cores = intel_num_cpu_cores(c); +#ifdef CONFIG_X86_32 + detect_ht(c); +#endif + } + l2 = init_intel_cacheinfo(c); if (c->cpuid_level > 9) { unsigned eax = cpuid_eax(10); @@ -392,7 +400,8 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c) set_cpu_cap(c, X86_FEATURE_PEBS); } - if (c->x86 == 6 && c->x86_model == 29 && cpu_has_clflush) + if (c->x86 == 6 && cpu_has_clflush && + (c->x86_model == 29 || c->x86_model == 46 || c->x86_model == 47)) set_cpu_cap(c, X86_FEATURE_CLFLUSH_MONITOR); #ifdef CONFIG_X86_64 @@ -440,17 +449,6 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c) set_cpu_cap(c, X86_FEATURE_P3); #endif - if (!cpu_has(c, X86_FEATURE_XTOPOLOGY)) { - /* - * let's use the legacy cpuid vector 0x1 and 0x4 for topology - * detection. - */ - c->x86_max_cores = intel_num_cpu_cores(c); -#ifdef CONFIG_X86_32 - detect_ht(c); -#endif - } - /* Work around errata */ srat_detect_node(c); @@ -477,7 +475,7 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c) } #ifdef CONFIG_X86_32 -static unsigned int __cpuinit intel_size_cache(struct cpuinfo_x86 *c, unsigned int size) +static unsigned int intel_size_cache(struct cpuinfo_x86 *c, unsigned int size) { /* * Intel PIII Tualatin. This comes in two flavours. @@ -491,12 +489,209 @@ static unsigned int __cpuinit intel_size_cache(struct cpuinfo_x86 *c, unsigned i } #endif -static const struct cpu_dev __cpuinitconst intel_cpu_dev = { +#define TLB_INST_4K 0x01 +#define TLB_INST_4M 0x02 +#define TLB_INST_2M_4M 0x03 + +#define TLB_INST_ALL 0x05 +#define TLB_INST_1G 0x06 + +#define TLB_DATA_4K 0x11 +#define TLB_DATA_4M 0x12 +#define TLB_DATA_2M_4M 0x13 +#define TLB_DATA_4K_4M 0x14 + +#define TLB_DATA_1G 0x16 + +#define TLB_DATA0_4K 0x21 +#define TLB_DATA0_4M 0x22 +#define TLB_DATA0_2M_4M 0x23 + +#define STLB_4K 0x41 +#define STLB_4K_2M 0x42 + +static const struct _tlb_table intel_tlb_table[] = { + { 0x01, TLB_INST_4K, 32, " TLB_INST 4 KByte pages, 4-way set associative" }, + { 0x02, TLB_INST_4M, 2, " TLB_INST 4 MByte pages, full associative" }, + { 0x03, TLB_DATA_4K, 64, " TLB_DATA 4 KByte pages, 4-way set associative" }, + { 0x04, TLB_DATA_4M, 8, " TLB_DATA 4 MByte pages, 4-way set associative" }, + { 0x05, TLB_DATA_4M, 32, " TLB_DATA 4 MByte pages, 4-way set associative" }, + { 0x0b, TLB_INST_4M, 4, " TLB_INST 4 MByte pages, 4-way set associative" }, + { 0x4f, TLB_INST_4K, 32, " TLB_INST 4 KByte pages */" }, + { 0x50, TLB_INST_ALL, 64, " TLB_INST 4 KByte and 2-MByte or 4-MByte pages" }, + { 0x51, TLB_INST_ALL, 128, " TLB_INST 4 KByte and 2-MByte or 4-MByte pages" }, + { 0x52, TLB_INST_ALL, 256, " TLB_INST 4 KByte and 2-MByte or 4-MByte pages" }, + { 0x55, TLB_INST_2M_4M, 7, " TLB_INST 2-MByte or 4-MByte pages, fully associative" }, + { 0x56, TLB_DATA0_4M, 16, " TLB_DATA0 4 MByte pages, 4-way set associative" }, + { 0x57, TLB_DATA0_4K, 16, " TLB_DATA0 4 KByte pages, 4-way associative" }, + { 0x59, TLB_DATA0_4K, 16, " TLB_DATA0 4 KByte pages, fully associative" }, + { 0x5a, TLB_DATA0_2M_4M, 32, " TLB_DATA0 2-MByte or 4 MByte pages, 4-way set associative" }, + { 0x5b, TLB_DATA_4K_4M, 64, " TLB_DATA 4 KByte and 4 MByte pages" }, + { 0x5c, TLB_DATA_4K_4M, 128, " TLB_DATA 4 KByte and 4 MByte pages" }, + { 0x5d, TLB_DATA_4K_4M, 256, " TLB_DATA 4 KByte and 4 MByte pages" }, + { 0x61, TLB_INST_4K, 48, " TLB_INST 4 KByte pages, full associative" }, + { 0x63, TLB_DATA_1G, 4, " TLB_DATA 1 GByte pages, 4-way set associative" }, + { 0x76, TLB_INST_2M_4M, 8, " TLB_INST 2-MByte or 4-MByte pages, fully associative" }, + { 0xb0, TLB_INST_4K, 128, " TLB_INST 4 KByte pages, 4-way set associative" }, + { 0xb1, TLB_INST_2M_4M, 4, " TLB_INST 2M pages, 4-way, 8 entries or 4M pages, 4-way entries" }, + { 0xb2, TLB_INST_4K, 64, " TLB_INST 4KByte pages, 4-way set associative" }, + { 0xb3, TLB_DATA_4K, 128, " TLB_DATA 4 KByte pages, 4-way set associative" }, + { 0xb4, TLB_DATA_4K, 256, " TLB_DATA 4 KByte pages, 4-way associative" }, + { 0xb5, TLB_INST_4K, 64, " TLB_INST 4 KByte pages, 8-way set ssociative" }, + { 0xb6, TLB_INST_4K, 128, " TLB_INST 4 KByte pages, 8-way set ssociative" }, + { 0xba, TLB_DATA_4K, 64, " TLB_DATA 4 KByte pages, 4-way associative" }, + { 0xc0, TLB_DATA_4K_4M, 8, " TLB_DATA 4 KByte and 4 MByte pages, 4-way associative" }, + { 0xc1, STLB_4K_2M, 1024, " STLB 4 KByte and 2 MByte pages, 8-way associative" }, + { 0xc2, TLB_DATA_2M_4M, 16, " DTLB 2 MByte/4MByte pages, 4-way associative" }, + { 0xca, STLB_4K, 512, " STLB 4 KByte pages, 4-way associative" }, + { 0x00, 0, 0 } +}; + +static void intel_tlb_lookup(const unsigned char desc) +{ + unsigned char k; + if (desc == 0) + return; + + /* look up this descriptor in the table */ + for (k = 0; intel_tlb_table[k].descriptor != desc && \ + intel_tlb_table[k].descriptor != 0; k++) + ; + + if (intel_tlb_table[k].tlb_type == 0) + return; + + switch (intel_tlb_table[k].tlb_type) { + case STLB_4K: + if (tlb_lli_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4k[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lld_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4k[ENTRIES] = intel_tlb_table[k].entries; + break; + case STLB_4K_2M: + if (tlb_lli_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4k[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lld_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4k[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lli_2m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_2m[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lld_2m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_2m[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lli_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4m[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lld_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4m[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_INST_ALL: + if (tlb_lli_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4k[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lli_2m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_2m[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lli_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4m[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_INST_4K: + if (tlb_lli_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4k[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_INST_4M: + if (tlb_lli_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4m[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_INST_2M_4M: + if (tlb_lli_2m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_2m[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lli_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lli_4m[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_DATA_4K: + case TLB_DATA0_4K: + if (tlb_lld_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4k[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_DATA_4M: + case TLB_DATA0_4M: + if (tlb_lld_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4m[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_DATA_2M_4M: + case TLB_DATA0_2M_4M: + if (tlb_lld_2m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_2m[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lld_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4m[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_DATA_4K_4M: + if (tlb_lld_4k[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4k[ENTRIES] = intel_tlb_table[k].entries; + if (tlb_lld_4m[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_4m[ENTRIES] = intel_tlb_table[k].entries; + break; + case TLB_DATA_1G: + if (tlb_lld_1g[ENTRIES] < intel_tlb_table[k].entries) + tlb_lld_1g[ENTRIES] = intel_tlb_table[k].entries; + break; + } +} + +static void intel_tlb_flushall_shift_set(struct cpuinfo_x86 *c) +{ + switch ((c->x86 << 8) + c->x86_model) { + case 0x60f: /* original 65 nm celeron/pentium/core2/xeon, "Merom"/"Conroe" */ + case 0x616: /* single-core 65 nm celeron/core2solo "Merom-L"/"Conroe-L" */ + case 0x617: /* current 45 nm celeron/core2/xeon "Penryn"/"Wolfdale" */ + case 0x61d: /* six-core 45 nm xeon "Dunnington" */ + tlb_flushall_shift = -1; + break; + case 0x63a: /* Ivybridge */ + tlb_flushall_shift = 2; + break; + case 0x61a: /* 45 nm nehalem, "Bloomfield" */ + case 0x61e: /* 45 nm nehalem, "Lynnfield" */ + case 0x625: /* 32 nm nehalem, "Clarkdale" */ + case 0x62c: /* 32 nm nehalem, "Gulftown" */ + case 0x62e: /* 45 nm nehalem-ex, "Beckton" */ + case 0x62f: /* 32 nm Xeon E7 */ + case 0x62a: /* SandyBridge */ + case 0x62d: /* SandyBridge, "Romely-EP" */ + default: + tlb_flushall_shift = 6; + } +} + +static void intel_detect_tlb(struct cpuinfo_x86 *c) +{ + int i, j, n; + unsigned int regs[4]; + unsigned char *desc = (unsigned char *)regs; + + if (c->cpuid_level < 2) + return; + + /* Number of times to iterate */ + n = cpuid_eax(2) & 0xFF; + + for (i = 0 ; i < n ; i++) { + cpuid(2, ®s[0], ®s[1], ®s[2], ®s[3]); + + /* If bit 31 is set, this is an unknown format */ + for (j = 0 ; j < 3 ; j++) + if (regs[j] & (1 << 31)) + regs[j] = 0; + + /* Byte 0 is level count, not a descriptor */ + for (j = 1 ; j < 16 ; j++) + intel_tlb_lookup(desc[j]); + } + intel_tlb_flushall_shift_set(c); +} + +static const struct cpu_dev intel_cpu_dev = { .c_vendor = "Intel", .c_ident = { "GenuineIntel" }, #ifdef CONFIG_X86_32 - .c_models = { - { .vendor = X86_VENDOR_INTEL, .family = 4, .model_names = + .legacy_models = { + { .family = 4, .model_names = { [0] = "486 DX-25/33", [1] = "486 DX-50", @@ -509,7 +704,7 @@ static const struct cpu_dev __cpuinitconst intel_cpu_dev = { [9] = "486 DX/4-WB" } }, - { .vendor = X86_VENDOR_INTEL, .family = 5, .model_names = + { .family = 5, .model_names = { [0] = "Pentium 60/66 A-step", [1] = "Pentium 60/66", @@ -520,7 +715,7 @@ static const struct cpu_dev __cpuinitconst intel_cpu_dev = { [8] = "Mobile Pentium MMX" } }, - { .vendor = X86_VENDOR_INTEL, .family = 6, .model_names = + { .family = 6, .model_names = { [0] = "Pentium Pro A-step", [1] = "Pentium Pro", @@ -534,7 +729,7 @@ static const struct cpu_dev __cpuinitconst intel_cpu_dev = { [11] = "Pentium III (Tualatin)", } }, - { .vendor = X86_VENDOR_INTEL, .family = 15, .model_names = + { .family = 15, .model_names = { [0] = "Pentium 4 (Unknown)", [1] = "Pentium 4 (Willamette)", @@ -544,8 +739,9 @@ static const struct cpu_dev __cpuinitconst intel_cpu_dev = { } }, }, - .c_size_cache = intel_size_cache, + .legacy_cache_size = intel_size_cache, #endif + .c_detect_tlb = intel_detect_tlb, .c_early_init = early_init_intel, .c_init = init_intel, .c_x86_vendor = X86_VENDOR_INTEL, diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c index 73d08ed98a6..9c8f7394c61 100644 --- a/arch/x86/kernel/cpu/intel_cacheinfo.c +++ b/arch/x86/kernel/cpu/intel_cacheinfo.c @@ -1,5 +1,5 @@ /* - * Routines to indentify caches on Intel CPU. + * Routines to identify caches on Intel CPU. * * Changes: * Venkatesh Pallipadi : Adding cache identification through cpuid(4) @@ -37,7 +37,7 @@ struct _cache_table { /* All the cache descriptor types we care about (no TLB or trace cache entries) */ -static const struct _cache_table __cpuinitconst cache_table[] = +static const struct _cache_table cache_table[] = { { 0x06, LVL_1_INST, 8 }, /* 4-way set assoc, 32 byte line size */ { 0x08, LVL_1_INST, 16 }, /* 4-way set assoc, 32 byte line size */ @@ -203,7 +203,7 @@ union l3_cache { unsigned val; }; -static const unsigned short __cpuinitconst assocs[] = { +static const unsigned short assocs[] = { [1] = 1, [2] = 2, [4] = 4, @@ -217,10 +217,10 @@ static const unsigned short __cpuinitconst assocs[] = { [0xf] = 0xffff /* fully associative - no way to show this currently */ }; -static const unsigned char __cpuinitconst levels[] = { 1, 1, 2, 3 }; -static const unsigned char __cpuinitconst types[] = { 1, 2, 3, 3 }; +static const unsigned char levels[] = { 1, 1, 2, 3 }; +static const unsigned char types[] = { 1, 2, 3, 3 }; -static void __cpuinit +static void amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax, union _cpuid4_leaf_ebx *ebx, union _cpuid4_leaf_ecx *ecx) @@ -298,12 +298,11 @@ struct _cache_attr { unsigned int); }; -#ifdef CONFIG_AMD_NB - +#if defined(CONFIG_AMD_NB) && defined(CONFIG_SYSFS) /* * L3 cache descriptors */ -static void __cpuinit amd_calc_l3_indices(struct amd_northbridge *nb) +static void amd_calc_l3_indices(struct amd_northbridge *nb) { struct amd_l3_cache *l3 = &nb->l3_cache; unsigned int sc0, sc1, sc2, sc3; @@ -326,7 +325,7 @@ static void __cpuinit amd_calc_l3_indices(struct amd_northbridge *nb) l3->indices = (max(max3(sc0, sc1, sc2), sc3) << 10) - 1; } -static void __cpuinit amd_init_l3_cache(struct _cpuid4_info_regs *this_leaf, int index) +static void amd_init_l3_cache(struct _cpuid4_info_regs *this_leaf, int index) { int node; @@ -433,14 +432,14 @@ int amd_set_l3_disable_slot(struct amd_northbridge *nb, int cpu, unsigned slot, /* check if @slot is already used or the index is already disabled */ ret = amd_get_l3_disable_slot(nb, slot); if (ret >= 0) - return -EINVAL; + return -EEXIST; if (index > nb->l3_cache.indices) return -EINVAL; /* check whether the other slot has disabled the same index already */ if (index == amd_get_l3_disable_slot(nb, !slot)) - return -EINVAL; + return -EEXIST; amd_l3_disable_index(nb, cpu, slot, index); @@ -468,8 +467,8 @@ static ssize_t store_cache_disable(struct _cpuid4_info *this_leaf, err = amd_set_l3_disable_slot(this_leaf->base.nb, cpu, slot, val); if (err) { if (err == -EEXIST) - printk(KERN_WARNING "L3 disable slot %d in use!\n", - slot); + pr_warning("L3 slot %d in use/index already disabled!\n", + slot); return err; } return count; @@ -524,13 +523,12 @@ store_subcaches(struct _cpuid4_info *this_leaf, const char *buf, size_t count, static struct _cache_attr subcaches = __ATTR(subcaches, 0644, show_subcaches, store_subcaches); -#else /* CONFIG_AMD_NB */ +#else #define amd_init_l3_cache(x, y) -#endif /* CONFIG_AMD_NB */ +#endif /* CONFIG_AMD_NB && CONFIG_SYSFS */ static int -__cpuinit cpuid4_cache_lookup_regs(int index, - struct _cpuid4_info_regs *this_leaf) +cpuid4_cache_lookup_regs(int index, struct _cpuid4_info_regs *this_leaf) { union _cpuid4_leaf_eax eax; union _cpuid4_leaf_ebx ebx; @@ -538,7 +536,11 @@ __cpuinit cpuid4_cache_lookup_regs(int index, unsigned edx; if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) { - amd_cpuid4(index, &eax, &ebx, &ecx); + if (cpu_has_topoext) + cpuid_count(0x8000001d, index, &eax.full, + &ebx.full, &ecx.full, &edx); + else + amd_cpuid4(index, &eax, &ebx, &ecx); amd_init_l3_cache(this_leaf, index); } else { cpuid_count(4, index, &eax.full, &ebx.full, &ecx.full, &edx); @@ -557,22 +559,40 @@ __cpuinit cpuid4_cache_lookup_regs(int index, return 0; } -static int __cpuinit find_num_cache_leaves(void) +static int find_num_cache_leaves(struct cpuinfo_x86 *c) { - unsigned int eax, ebx, ecx, edx; + unsigned int eax, ebx, ecx, edx, op; union _cpuid4_leaf_eax cache_eax; int i = -1; + if (c->x86_vendor == X86_VENDOR_AMD) + op = 0x8000001d; + else + op = 4; + do { ++i; - /* Do cpuid(4) loop to find out num_cache_leaves */ - cpuid_count(4, i, &eax, &ebx, &ecx, &edx); + /* Do cpuid(op) loop to find out num_cache_leaves */ + cpuid_count(op, i, &eax, &ebx, &ecx, &edx); cache_eax.full = eax; } while (cache_eax.split.type != CACHE_TYPE_NULL); return i; } -unsigned int __cpuinit init_intel_cacheinfo(struct cpuinfo_x86 *c) +void init_amd_cacheinfo(struct cpuinfo_x86 *c) +{ + + if (cpu_has_topoext) { + num_cache_leaves = find_num_cache_leaves(c); + } else if (c->extended_cpuid_level >= 0x80000006) { + if (cpuid_edx(0x80000006) & 0xf000) + num_cache_leaves = 4; + else + num_cache_leaves = 3; + } +} + +unsigned int init_intel_cacheinfo(struct cpuinfo_x86 *c) { /* Cache sizes */ unsigned int trace = 0, l1i = 0, l1d = 0, l2 = 0, l3 = 0; @@ -588,7 +608,7 @@ unsigned int __cpuinit init_intel_cacheinfo(struct cpuinfo_x86 *c) if (is_initialized == 0) { /* Init num_cache_leaves from boot CPU */ - num_cache_leaves = find_num_cache_leaves(); + num_cache_leaves = find_num_cache_leaves(c); is_initialized++; } @@ -597,36 +617,34 @@ unsigned int __cpuinit init_intel_cacheinfo(struct cpuinfo_x86 *c) * parameters cpuid leaf to find the cache details */ for (i = 0; i < num_cache_leaves; i++) { - struct _cpuid4_info_regs this_leaf; + struct _cpuid4_info_regs this_leaf = {}; int retval; retval = cpuid4_cache_lookup_regs(i, &this_leaf); - if (retval >= 0) { - switch (this_leaf.eax.split.level) { - case 1: - if (this_leaf.eax.split.type == - CACHE_TYPE_DATA) - new_l1d = this_leaf.size/1024; - else if (this_leaf.eax.split.type == - CACHE_TYPE_INST) - new_l1i = this_leaf.size/1024; - break; - case 2: - new_l2 = this_leaf.size/1024; - num_threads_sharing = 1 + this_leaf.eax.split.num_threads_sharing; - index_msb = get_count_order(num_threads_sharing); - l2_id = c->apicid >> index_msb; - break; - case 3: - new_l3 = this_leaf.size/1024; - num_threads_sharing = 1 + this_leaf.eax.split.num_threads_sharing; - index_msb = get_count_order( - num_threads_sharing); - l3_id = c->apicid >> index_msb; - break; - default: - break; - } + if (retval < 0) + continue; + + switch (this_leaf.eax.split.level) { + case 1: + if (this_leaf.eax.split.type == CACHE_TYPE_DATA) + new_l1d = this_leaf.size/1024; + else if (this_leaf.eax.split.type == CACHE_TYPE_INST) + new_l1i = this_leaf.size/1024; + break; + case 2: + new_l2 = this_leaf.size/1024; + num_threads_sharing = 1 + this_leaf.eax.split.num_threads_sharing; + index_msb = get_count_order(num_threads_sharing); + l2_id = c->apicid & ~((1 << index_msb) - 1); + break; + case 3: + new_l3 = this_leaf.size/1024; + num_threads_sharing = 1 + this_leaf.eax.split.num_threads_sharing; + index_msb = get_count_order(num_threads_sharing); + l3_id = c->apicid & ~((1 << index_msb) - 1); + break; + default: + break; } } } @@ -712,6 +730,18 @@ unsigned int __cpuinit init_intel_cacheinfo(struct cpuinfo_x86 *c) #endif } +#ifdef CONFIG_X86_HT + /* + * If cpu_llc_id is not yet set, this means cpuid_level < 4 which in + * turns means that the only possibility is SMT (as indicated in + * cpuid1). Since cpuid2 doesn't specify shared caches, and we know + * that SMT shares all caches, we can unconditionally set cpu_llc_id to + * c->phys_proc_id. + */ + if (per_cpu(cpu_llc_id, cpu) == BAD_APICID) + per_cpu(cpu_llc_id, cpu) = c->phys_proc_id; +#endif + c->x86_cache_size = l3 ? l3 : (l2 ? l2 : (l1i+l1d)); return l2; @@ -725,43 +755,56 @@ static DEFINE_PER_CPU(struct _cpuid4_info *, ici_cpuid4_info); #ifdef CONFIG_SMP -static int __cpuinit cache_shared_amd_cpu_map_setup(unsigned int cpu, int index) +static int cache_shared_amd_cpu_map_setup(unsigned int cpu, int index) { struct _cpuid4_info *this_leaf; - int ret, i, sibling; - struct cpuinfo_x86 *c = &cpu_data(cpu); + int i, sibling; - ret = 0; - if (index == 3) { - ret = 1; - for_each_cpu(i, cpu_llc_shared_mask(cpu)) { + if (cpu_has_topoext) { + unsigned int apicid, nshared, first, last; + + if (!per_cpu(ici_cpuid4_info, cpu)) + return 0; + + this_leaf = CPUID4_INFO_IDX(cpu, index); + nshared = this_leaf->base.eax.split.num_threads_sharing + 1; + apicid = cpu_data(cpu).apicid; + first = apicid - (apicid % nshared); + last = first + nshared - 1; + + for_each_online_cpu(i) { + apicid = cpu_data(i).apicid; + if ((apicid < first) || (apicid > last)) + continue; if (!per_cpu(ici_cpuid4_info, i)) continue; this_leaf = CPUID4_INFO_IDX(i, index); - for_each_cpu(sibling, cpu_llc_shared_mask(cpu)) { - if (!cpu_online(sibling)) + + for_each_online_cpu(sibling) { + apicid = cpu_data(sibling).apicid; + if ((apicid < first) || (apicid > last)) continue; set_bit(sibling, this_leaf->shared_cpu_map); } } - } else if ((c->x86 == 0x15) && ((index == 1) || (index == 2))) { - ret = 1; - for_each_cpu(i, cpu_sibling_mask(cpu)) { + } else if (index == 3) { + for_each_cpu(i, cpu_llc_shared_mask(cpu)) { if (!per_cpu(ici_cpuid4_info, i)) continue; this_leaf = CPUID4_INFO_IDX(i, index); - for_each_cpu(sibling, cpu_sibling_mask(cpu)) { + for_each_cpu(sibling, cpu_llc_shared_mask(cpu)) { if (!cpu_online(sibling)) continue; set_bit(sibling, this_leaf->shared_cpu_map); } } - } + } else + return 0; - return ret; + return 1; } -static void __cpuinit cache_shared_cpu_map_setup(unsigned int cpu, int index) +static void cache_shared_cpu_map_setup(unsigned int cpu, int index) { struct _cpuid4_info *this_leaf, *sibling_leaf; unsigned long num_threads_sharing; @@ -796,7 +839,7 @@ static void __cpuinit cache_shared_cpu_map_setup(unsigned int cpu, int index) } } } -static void __cpuinit cache_remove_shared_cpu_map(unsigned int cpu, int index) +static void cache_remove_shared_cpu_map(unsigned int cpu, int index) { struct _cpuid4_info *this_leaf, *sibling_leaf; int sibling; @@ -809,16 +852,16 @@ static void __cpuinit cache_remove_shared_cpu_map(unsigned int cpu, int index) } } #else -static void __cpuinit cache_shared_cpu_map_setup(unsigned int cpu, int index) +static void cache_shared_cpu_map_setup(unsigned int cpu, int index) { } -static void __cpuinit cache_remove_shared_cpu_map(unsigned int cpu, int index) +static void cache_remove_shared_cpu_map(unsigned int cpu, int index) { } #endif -static void __cpuinit free_cache_attributes(unsigned int cpu) +static void free_cache_attributes(unsigned int cpu) { int i; @@ -829,7 +872,7 @@ static void __cpuinit free_cache_attributes(unsigned int cpu) per_cpu(ici_cpuid4_info, cpu) = NULL; } -static void __cpuinit get_cpu_leaves(void *_retval) +static void get_cpu_leaves(void *_retval) { int j, *retval = _retval, cpu = smp_processor_id(); @@ -849,7 +892,7 @@ static void __cpuinit get_cpu_leaves(void *_retval) } } -static int __cpuinit detect_cache_attributes(unsigned int cpu) +static int detect_cache_attributes(unsigned int cpu) { int retval; @@ -983,7 +1026,7 @@ static struct attribute *default_attrs[] = { }; #ifdef CONFIG_AMD_NB -static struct attribute ** __cpuinit amd_l3_attrs(void) +static struct attribute **amd_l3_attrs(void) { static struct attribute **attrs; int n; @@ -991,7 +1034,7 @@ static struct attribute ** __cpuinit amd_l3_attrs(void) if (attrs) return attrs; - n = sizeof (default_attrs) / sizeof (struct attribute *); + n = ARRAY_SIZE(default_attrs); if (amd_nb_has_feature(AMD_NB_L3_INDEX_DISABLE)) n += 2; @@ -1059,7 +1102,7 @@ static struct kobj_type ktype_percpu_entry = { .sysfs_ops = &sysfs_ops, }; -static void __cpuinit cpuid4_cache_sysfs_exit(unsigned int cpu) +static void cpuid4_cache_sysfs_exit(unsigned int cpu) { kfree(per_cpu(ici_cache_kobject, cpu)); kfree(per_cpu(ici_index_kobject, cpu)); @@ -1068,7 +1111,7 @@ static void __cpuinit cpuid4_cache_sysfs_exit(unsigned int cpu) free_cache_attributes(cpu); } -static int __cpuinit cpuid4_cache_sysfs_init(unsigned int cpu) +static int cpuid4_cache_sysfs_init(unsigned int cpu) { int err; @@ -1100,7 +1143,7 @@ err_out: static DECLARE_BITMAP(cache_dev_map, NR_CPUS); /* Add/Remove cache interface for CPU device */ -static int __cpuinit cache_add_dev(struct device *dev) +static int cache_add_dev(struct device *dev) { unsigned int cpu = dev->id; unsigned long i, j; @@ -1151,7 +1194,7 @@ static int __cpuinit cache_add_dev(struct device *dev) return 0; } -static void __cpuinit cache_remove_dev(struct device *dev) +static void cache_remove_dev(struct device *dev) { unsigned int cpu = dev->id; unsigned long i; @@ -1168,8 +1211,8 @@ static void __cpuinit cache_remove_dev(struct device *dev) cpuid4_cache_sysfs_exit(cpu); } -static int __cpuinit cacheinfo_cpu_callback(struct notifier_block *nfb, - unsigned long action, void *hcpu) +static int cacheinfo_cpu_callback(struct notifier_block *nfb, + unsigned long action, void *hcpu) { unsigned int cpu = (unsigned long)hcpu; struct device *dev; @@ -1188,27 +1231,30 @@ static int __cpuinit cacheinfo_cpu_callback(struct notifier_block *nfb, return NOTIFY_OK; } -static struct notifier_block __cpuinitdata cacheinfo_cpu_notifier = { +static struct notifier_block cacheinfo_cpu_notifier = { .notifier_call = cacheinfo_cpu_callback, }; -static int __cpuinit cache_sysfs_init(void) +static int __init cache_sysfs_init(void) { - int i; + int i, err = 0; if (num_cache_leaves == 0) return 0; + cpu_notifier_register_begin(); for_each_online_cpu(i) { - int err; struct device *dev = get_cpu_device(i); err = cache_add_dev(dev); if (err) - return err; + goto out; } - register_hotcpu_notifier(&cacheinfo_cpu_notifier); - return 0; + __register_hotcpu_notifier(&cacheinfo_cpu_notifier); + +out: + cpu_notifier_register_done(); + return err; } device_initcall(cache_sysfs_init); diff --git a/arch/x86/kernel/cpu/match.c b/arch/x86/kernel/cpu/match.c new file mode 100644 index 00000000000..afa9f0d487e --- /dev/null +++ b/arch/x86/kernel/cpu/match.c @@ -0,0 +1,49 @@ +#include <asm/cpu_device_id.h> +#include <asm/processor.h> +#include <linux/cpu.h> +#include <linux/module.h> +#include <linux/slab.h> + +/** + * x86_match_cpu - match current CPU again an array of x86_cpu_ids + * @match: Pointer to array of x86_cpu_ids. Last entry terminated with + * {}. + * + * Return the entry if the current CPU matches the entries in the + * passed x86_cpu_id match table. Otherwise NULL. The match table + * contains vendor (X86_VENDOR_*), family, model and feature bits or + * respective wildcard entries. + * + * A typical table entry would be to match a specific CPU + * { X86_VENDOR_INTEL, 6, 0x12 } + * or to match a specific CPU feature + * { X86_FEATURE_MATCH(X86_FEATURE_FOOBAR) } + * + * Fields can be wildcarded with %X86_VENDOR_ANY, %X86_FAMILY_ANY, + * %X86_MODEL_ANY, %X86_FEATURE_ANY or 0 (except for vendor) + * + * Arrays used to match for this should also be declared using + * MODULE_DEVICE_TABLE(x86cpu, ...) + * + * This always matches against the boot cpu, assuming models and features are + * consistent over all CPUs. + */ +const struct x86_cpu_id *x86_match_cpu(const struct x86_cpu_id *match) +{ + const struct x86_cpu_id *m; + struct cpuinfo_x86 *c = &boot_cpu_data; + + for (m = match; m->vendor | m->family | m->model | m->feature; m++) { + if (m->vendor != X86_VENDOR_ANY && c->x86_vendor != m->vendor) + continue; + if (m->family != X86_FAMILY_ANY && c->x86 != m->family) + continue; + if (m->model != X86_MODEL_ANY && c->x86_model != m->model) + continue; + if (m->feature != X86_FEATURE_ANY && !cpu_has(c, m->feature)) + continue; + return m; + } + return NULL; +} +EXPORT_SYMBOL(x86_match_cpu); diff --git a/arch/x86/kernel/cpu/mcheck/mce-apei.c b/arch/x86/kernel/cpu/mcheck/mce-apei.c index 507ea58688e..a1aef953315 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-apei.c +++ b/arch/x86/kernel/cpu/mcheck/mce-apei.c @@ -33,22 +33,28 @@ #include <linux/acpi.h> #include <linux/cper.h> #include <acpi/apei.h> +#include <acpi/ghes.h> #include <asm/mce.h> #include "mce-internal.h" -void apei_mce_report_mem_error(int corrected, struct cper_sec_mem_err *mem_err) +void apei_mce_report_mem_error(int severity, struct cper_sec_mem_err *mem_err) { struct mce m; - /* Only corrected MC is reported */ - if (!corrected) + if (!(mem_err->validation_bits & CPER_MEM_VALID_PA)) return; mce_setup(&m); m.bank = 1; - /* Fake a memory read corrected error with unknown channel */ + /* Fake a memory read error with unknown channel */ m.status = MCI_STATUS_VAL | MCI_STATUS_EN | MCI_STATUS_ADDRV | 0x9f; + + if (severity >= GHES_SEV_RECOVERABLE) + m.status |= MCI_STATUS_UC; + if (severity >= GHES_SEV_PANIC) + m.status |= MCI_STATUS_PCC; + m.addr = mem_err->physical_addr; mce_log(&m); mce_notify_irq(); diff --git a/arch/x86/kernel/cpu/mcheck/mce-inject.c b/arch/x86/kernel/cpu/mcheck/mce-inject.c index fc4beb39357..5ac2d1fb28b 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-inject.c +++ b/arch/x86/kernel/cpu/mcheck/mce-inject.c @@ -78,6 +78,7 @@ static void raise_exception(struct mce *m, struct pt_regs *pregs) } static cpumask_var_t mce_inject_cpumask; +static DEFINE_MUTEX(mce_inject_mutex); static int mce_raise_notify(unsigned int cmd, struct pt_regs *regs) { @@ -152,7 +153,7 @@ static void raise_mce(struct mce *m) return; #ifdef CONFIG_X86_LOCAL_APIC - if (m->inject_flags & (MCJ_IRQ_BRAODCAST | MCJ_NMI_BROADCAST)) { + if (m->inject_flags & (MCJ_IRQ_BROADCAST | MCJ_NMI_BROADCAST)) { unsigned long start; int cpu; @@ -166,7 +167,7 @@ static void raise_mce(struct mce *m) cpumask_clear_cpu(cpu, mce_inject_cpumask); } if (!cpumask_empty(mce_inject_cpumask)) { - if (m->inject_flags & MCJ_IRQ_BRAODCAST) { + if (m->inject_flags & MCJ_IRQ_BROADCAST) { /* * don't wait because mce_irq_ipi is necessary * to be sync with following raise_local @@ -194,7 +195,11 @@ static void raise_mce(struct mce *m) put_online_cpus(); } else #endif + { + preempt_disable(); raise_local(); + preempt_enable(); + } } /* Error injection interface */ @@ -225,7 +230,10 @@ static ssize_t mce_write(struct file *filp, const char __user *ubuf, * so do it a jiffie or two later everywhere. */ schedule_timeout(2); + + mutex_lock(&mce_inject_mutex); raise_mce(&m); + mutex_unlock(&mce_inject_mutex); return usize; } diff --git a/arch/x86/kernel/cpu/mcheck/mce-internal.h b/arch/x86/kernel/cpu/mcheck/mce-internal.h index ed44c8a6585..09edd0b65fe 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-internal.h +++ b/arch/x86/kernel/cpu/mcheck/mce-internal.h @@ -24,9 +24,22 @@ struct mce_bank { 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; +extern mce_banks_t mce_banks_ce_disabled; + +#ifdef CONFIG_X86_MCE_INTEL +unsigned long mce_intel_adjust_timer(unsigned long interval); +void mce_intel_cmci_poll(void); +void mce_intel_hcpu_update(unsigned long cpu); +void cmci_disable_bank(int bank); +#else +# define mce_intel_adjust_timer mce_adjust_timer_default +static inline void mce_intel_cmci_poll(void) { } +static inline void mce_intel_hcpu_update(unsigned long cpu) { } +static inline void cmci_disable_bank(int bank) { } +#endif + +void mce_timer_kick(unsigned long interval); #ifdef CONFIG_ACPI_APEI int apei_write_mce(struct mce *m); diff --git a/arch/x86/kernel/cpu/mcheck/mce-severity.c b/arch/x86/kernel/cpu/mcheck/mce-severity.c index 7395d5f4272..c370e1c4468 100644 --- a/arch/x86/kernel/cpu/mcheck/mce-severity.c +++ b/arch/x86/kernel/cpu/mcheck/mce-severity.c @@ -54,7 +54,7 @@ static struct severity { #define MASK(x, y) .mask = x, .result = y #define MCI_UC_S (MCI_STATUS_UC|MCI_STATUS_S) #define MCI_UC_SAR (MCI_STATUS_UC|MCI_STATUS_S|MCI_STATUS_AR) -#define MCACOD 0xffff +#define MCI_ADDR (MCI_STATUS_ADDRV|MCI_STATUS_MISCV) MCESEV( NO, "Invalid", @@ -102,11 +102,29 @@ static struct severity { SER, BITCLR(MCI_STATUS_S) ), - /* AR add known MCACODs here */ MCESEV( PANIC, "Action required with lost events", SER, BITSET(MCI_STATUS_OVER|MCI_UC_SAR) ), + + /* known AR MCACODs: */ +#ifdef CONFIG_MEMORY_FAILURE + MCESEV( + KEEP, "Action required but unaffected thread is continuable", + SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR, MCI_UC_SAR|MCI_ADDR), + MCGMASK(MCG_STATUS_RIPV|MCG_STATUS_EIPV, MCG_STATUS_RIPV) + ), + MCESEV( + AR, "Action required: data load error in a user process", + SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR|MCACOD, MCI_UC_SAR|MCI_ADDR|MCACOD_DATA), + USER + ), + MCESEV( + AR, "Action required: instruction fetch error in a user process", + SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR|MCACOD, MCI_UC_SAR|MCI_ADDR|MCACOD_INSTR), + USER + ), +#endif MCESEV( PANIC, "Action required: unknown MCACOD", SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_UC_SAR) @@ -115,11 +133,11 @@ static struct severity { /* known AO MCACODs: */ MCESEV( AO, "Action optional: memory scrubbing error", - SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|0xfff0, MCI_UC_S|0x00c0) + SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCACOD_SCRUBMSK, MCI_UC_S|MCACOD_SCRUB) ), MCESEV( AO, "Action optional: last level cache writeback error", - SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCACOD, MCI_UC_S|0x017a) + SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCACOD, MCI_UC_S|MCACOD_L3WB) ), MCESEV( SOME, "Action optional: unknown MCACOD", @@ -145,15 +163,19 @@ static struct severity { }; /* - * If the EIPV bit is set, it means the saved IP is the - * instruction which caused the MCE. + * If mcgstatus indicated that ip/cs on the stack were + * no good, then "m->cs" will be zero and we will have + * to assume the worst case (IN_KERNEL) as we actually + * have no idea what we were executing when the machine + * check hit. + * If we do have a good "m->cs" (or a faked one in the + * case we were executing in VM86 mode) we can use it to + * distinguish an exception taken in user from from one + * taken in the kernel. */ static int error_context(struct mce *m) { - if (m->mcgstatus & MCG_STATUS_EIPV) - return (m->ip && (m->cs & 3) == 3) ? IN_USER : IN_KERNEL; - /* Unknown, assume kernel */ - return IN_KERNEL; + return ((m->cs & 3) == 3) ? IN_USER : IN_KERNEL; } int mce_severity(struct mce *m, int tolerant, char **msg) @@ -166,9 +188,9 @@ int mce_severity(struct mce *m, int tolerant, char **msg) continue; if ((m->mcgstatus & s->mcgmask) != s->mcgres) continue; - if (s->ser == SER_REQUIRED && !mce_ser) + if (s->ser == SER_REQUIRED && !mca_cfg.ser) continue; - if (s->ser == NO_SER && mce_ser) + if (s->ser == NO_SER && mca_cfg.ser) continue; if (s->context && ctx != s->context) continue; diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c index 5a11ae2e9e9..9a79c8dbd8e 100644 --- a/arch/x86/kernel/cpu/mcheck/mce.c +++ b/arch/x86/kernel/cpu/mcheck/mce.c @@ -7,6 +7,9 @@ * Copyright 2008 Intel Corporation * Author: Andi Kleen */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/thread_info.h> #include <linux/capability.h> #include <linux/miscdevice.h> @@ -55,35 +58,24 @@ static DEFINE_MUTEX(mce_chrdev_read_mutex); #define CREATE_TRACE_POINTS #include <trace/events/mce.h> -int mce_disabled __read_mostly; - -#define MISC_MCELOG_MINOR 227 - #define SPINUNIT 100 /* 100ns */ -atomic_t mce_entry; - DEFINE_PER_CPU(unsigned, mce_exception_count); -/* - * Tolerant levels: - * 0: always panic on uncorrected errors, log corrected errors - * 1: panic or SIGBUS on uncorrected errors, log corrected errors - * 2: SIGBUS or log uncorrected errors (if possible), log corrected errors - * 3: never panic or SIGBUS, log all errors (for testing only) - */ -static int tolerant __read_mostly = 1; -static int banks __read_mostly; -static int rip_msr __read_mostly; -static int mce_bootlog __read_mostly = -1; -static int monarch_timeout __read_mostly = -1; -static int mce_panic_timeout __read_mostly; -static int mce_dont_log_ce __read_mostly; -int mce_cmci_disabled __read_mostly; -int mce_ignore_ce __read_mostly; -int mce_ser __read_mostly; - -struct mce_bank *mce_banks __read_mostly; +struct mce_bank *mce_banks __read_mostly; + +struct mca_config mca_cfg __read_mostly = { + .bootlog = -1, + /* + * Tolerant levels: + * 0: always panic on uncorrected errors, log corrected errors + * 1: panic or SIGBUS on uncorrected errors, log corrected errors + * 2: SIGBUS or log uncorrected errors (if possible), log corr. errors + * 3: never panic or SIGBUS, log all errors (for testing only) + */ + .tolerant = 1, + .monarch_timeout = -1 +}; /* User mode helper program triggered by machine check event */ static unsigned long mce_need_notify; @@ -95,13 +87,30 @@ static DECLARE_WAIT_QUEUE_HEAD(mce_chrdev_wait); static DEFINE_PER_CPU(struct mce, mces_seen); static int cpu_missing; -/* MCA banks polled by the period polling timer for corrected events */ +/* CMCI storm detection filter */ +static DEFINE_PER_CPU(unsigned long, mce_polled_error); + +/* + * MCA banks polled by the period polling timer for corrected events. + * With Intel CMCI, this only has MCA banks which do not support CMCI (if any). + */ DEFINE_PER_CPU(mce_banks_t, mce_poll_banks) = { [0 ... BITS_TO_LONGS(MAX_NR_BANKS)-1] = ~0UL }; +/* + * MCA banks controlled through firmware first for corrected errors. + * This is a global list of banks for which we won't enable CMCI and we + * won't poll. Firmware controls these banks and is responsible for + * reporting corrected errors through GHES. Uncorrected/recoverable + * errors are still notified through a machine check. + */ +mce_banks_t mce_banks_ce_disabled; + static DEFINE_PER_CPU(struct work_struct, mce_work); +static void (*quirk_no_way_out)(int bank, struct mce *m, struct pt_regs *regs); + /* * CPU/chipset specific EDAC code can register a notifier call here to print * MCE errors in a human-readable form. @@ -191,7 +200,7 @@ static void drain_mcelog_buffer(void) { unsigned int next, i, prev = 0; - next = rcu_dereference_check_mce(mcelog.next); + next = ACCESS_ONCE(mcelog.next); do { struct mce *m; @@ -210,7 +219,7 @@ static void drain_mcelog_buffer(void) cpu_relax(); if (!m->finished && retries >= 4) { - pr_err("MCE: skipping error being logged currently!\n"); + pr_err("skipping error being logged currently!\n"); break; } } @@ -298,7 +307,7 @@ static void wait_for_panic(void) while (timeout-- > 0) udelay(1); if (panic_timeout == 0) - panic_timeout = mce_panic_timeout; + panic_timeout = mca_cfg.panic_timeout; panic("Panicing machine check CPU died"); } @@ -356,7 +365,7 @@ static void mce_panic(char *msg, struct mce *final, char *exp) pr_emerg(HW_ERR "Machine check: %s\n", exp); if (!fake_panic) { if (panic_timeout == 0) - panic_timeout = mce_panic_timeout; + panic_timeout = mca_cfg.panic_timeout; panic(msg); } else pr_emerg(HW_ERR "Fake kernel panic: %s\n", msg); @@ -368,7 +377,7 @@ static int msr_to_offset(u32 msr) { unsigned bank = __this_cpu_read(injectm.bank); - if (msr == rip_msr) + if (msr == mca_cfg.rip_msr) return offsetof(struct mce, ip); if (msr == MSR_IA32_MCx_STATUS(bank)) return offsetof(struct mce, status); @@ -437,10 +446,18 @@ static inline void mce_gather_info(struct mce *m, struct pt_regs *regs) if (m->mcgstatus & (MCG_STATUS_RIPV|MCG_STATUS_EIPV)) { m->ip = regs->ip; m->cs = regs->cs; + + /* + * When in VM86 mode make the cs look like ring 3 + * always. This is a lie, but it's better than passing + * the additional vm86 bit around everywhere. + */ + if (v8086_mode(regs)) + m->cs |= 3; } /* Use accurate RIP reporting if available. */ - if (rip_msr) - m->ip = mce_rdmsrl(rip_msr); + if (mca_cfg.rip_msr) + m->ip = mce_rdmsrl(mca_cfg.rip_msr); } } @@ -501,18 +518,15 @@ static int mce_ring_add(unsigned long pfn) int mce_available(struct cpuinfo_x86 *c) { - if (mce_disabled) + if (mca_cfg.disabled) return 0; return cpu_has(c, X86_FEATURE_MCE) && cpu_has(c, X86_FEATURE_MCA); } static void mce_schedule_work(void) { - if (!mce_ring_empty()) { - struct work_struct *work = &__get_cpu_var(mce_work); - if (!work_pending(work)) - schedule_work(work); - } + if (!mce_ring_empty()) + schedule_work(&__get_cpu_var(mce_work)); } DEFINE_PER_CPU(struct irq_work, mce_irq_work); @@ -540,6 +554,27 @@ static void mce_report_event(struct pt_regs *regs) irq_work_queue(&__get_cpu_var(mce_irq_work)); } +/* + * Read ADDR and MISC registers. + */ +static void mce_read_aux(struct mce *m, int i) +{ + if (m->status & MCI_STATUS_MISCV) + m->misc = mce_rdmsrl(MSR_IA32_MCx_MISC(i)); + if (m->status & MCI_STATUS_ADDRV) { + m->addr = mce_rdmsrl(MSR_IA32_MCx_ADDR(i)); + + /* + * Mask the reported address by the reported granularity. + */ + if (mca_cfg.ser && (m->status & MCI_STATUS_MISCV)) { + u8 shift = MCI_MISC_ADDR_LSB(m->misc); + m->addr >>= shift; + m->addr <<= shift; + } + } +} + DEFINE_PER_CPU(unsigned, mce_poll_count); /* @@ -562,11 +597,11 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) struct mce m; int i; - percpu_inc(mce_poll_count); + this_cpu_inc(mce_poll_count); mce_gather_info(&m, NULL); - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { if (!mce_banks[i].ctl || !test_bit(i, *b)) continue; @@ -580,6 +615,7 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) if (!(m.status & MCI_STATUS_VAL)) continue; + this_cpu_write(mce_polled_error, 1); /* * Uncorrected or signalled events are handled by the exception * handler when it is enabled, so don't process those here. @@ -587,13 +623,10 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) * TBD do the same check for MCI_STATUS_EN here? */ if (!(flags & MCP_UC) && - (m.status & (mce_ser ? MCI_STATUS_S : MCI_STATUS_UC))) + (m.status & (mca_cfg.ser ? MCI_STATUS_S : MCI_STATUS_UC))) continue; - if (m.status & MCI_STATUS_MISCV) - m.misc = mce_rdmsrl(MSR_IA32_MCx_MISC(i)); - if (m.status & MCI_STATUS_ADDRV) - m.addr = mce_rdmsrl(MSR_IA32_MCx_ADDR(i)); + mce_read_aux(&m, i); if (!(flags & MCP_TIMESTAMP)) m.tsc = 0; @@ -601,7 +634,7 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b) * Don't get the IP here because it's unlikely to * have anything to do with the actual error location. */ - if (!(flags & MCP_DONTLOG) && !mce_dont_log_ce) + if (!(flags & MCP_DONTLOG) && !mca_cfg.dont_log_ce) mce_log(&m); /* @@ -623,16 +656,22 @@ EXPORT_SYMBOL_GPL(machine_check_poll); * Do a quick check if any of the events requires a panic. * This decides if we keep the events around or clear them. */ -static int mce_no_way_out(struct mce *m, char **msg) +static int mce_no_way_out(struct mce *m, char **msg, unsigned long *validp, + struct pt_regs *regs) { - int i; + int i, ret = 0; - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { m->status = mce_rdmsrl(MSR_IA32_MCx_STATUS(i)); - if (mce_severity(m, tolerant, msg) >= MCE_PANIC_SEVERITY) - return 1; + if (m->status & MCI_STATUS_VAL) { + __set_bit(i, validp); + if (quirk_no_way_out) + quirk_no_way_out(i, m, regs); + } + if (mce_severity(m, mca_cfg.tolerant, msg) >= MCE_PANIC_SEVERITY) + ret = 1; } - return 0; + return ret; } /* @@ -660,11 +699,10 @@ static int mce_timed_out(u64 *t) rmb(); if (atomic_read(&mce_paniced)) wait_for_panic(); - if (!monarch_timeout) + if (!mca_cfg.monarch_timeout) goto out; if ((s64)*t < SPINUNIT) { - /* CHECKME: Make panic default for 1 too? */ - if (tolerant < 1) + if (mca_cfg.tolerant <= 1) mce_panic("Timeout synchronizing machine check over CPUs", NULL, NULL); cpu_missing = 1; @@ -714,7 +752,8 @@ static void mce_reign(void) * Grade the severity of the errors of all the CPUs. */ for_each_possible_cpu(cpu) { - int severity = mce_severity(&per_cpu(mces_seen, cpu), tolerant, + int severity = mce_severity(&per_cpu(mces_seen, cpu), + mca_cfg.tolerant, &nmsg); if (severity > global_worst) { msg = nmsg; @@ -728,7 +767,7 @@ static void mce_reign(void) * This dumps all the mces in the log buffer and stops the * other CPUs. */ - if (m && global_worst >= MCE_PANIC_SEVERITY && tolerant < 3) + if (m && global_worst >= MCE_PANIC_SEVERITY && mca_cfg.tolerant < 3) mce_panic("Fatal Machine check", m, msg); /* @@ -741,7 +780,7 @@ static void mce_reign(void) * No machine check event found. Must be some external * source or one CPU is hung. Panic. */ - if (global_worst <= MCE_KEEP_SEVERITY && tolerant < 3) + if (global_worst <= MCE_KEEP_SEVERITY && mca_cfg.tolerant < 3) mce_panic("Machine check from unknown source", NULL, NULL); /* @@ -765,7 +804,7 @@ static int mce_start(int *no_way_out) { int order; int cpus = num_online_cpus(); - u64 timeout = (u64)monarch_timeout * NSEC_PER_USEC; + u64 timeout = (u64)mca_cfg.monarch_timeout * NSEC_PER_USEC; if (!timeout) return -1; @@ -829,7 +868,7 @@ static int mce_start(int *no_way_out) static int mce_end(int order) { int ret = -1; - u64 timeout = (u64)monarch_timeout * NSEC_PER_USEC; + u64 timeout = (u64)mca_cfg.monarch_timeout * NSEC_PER_USEC; if (!timeout) goto reset; @@ -910,13 +949,58 @@ static void mce_clear_state(unsigned long *toclear) { int i; - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { if (test_bit(i, toclear)) mce_wrmsrl(MSR_IA32_MCx_STATUS(i), 0); } } /* + * Need to save faulting physical address associated with a process + * in the machine check handler some place where we can grab it back + * later in mce_notify_process() + */ +#define MCE_INFO_MAX 16 + +struct mce_info { + atomic_t inuse; + struct task_struct *t; + __u64 paddr; + int restartable; +} mce_info[MCE_INFO_MAX]; + +static void mce_save_info(__u64 addr, int c) +{ + struct mce_info *mi; + + for (mi = mce_info; mi < &mce_info[MCE_INFO_MAX]; mi++) { + if (atomic_cmpxchg(&mi->inuse, 0, 1) == 0) { + mi->t = current; + mi->paddr = addr; + mi->restartable = c; + return; + } + } + + mce_panic("Too many concurrent recoverable errors", NULL, NULL); +} + +static struct mce_info *mce_find_info(void) +{ + struct mce_info *mi; + + for (mi = mce_info; mi < &mce_info[MCE_INFO_MAX]; mi++) + if (atomic_read(&mi->inuse) && mi->t == current) + return mi; + return NULL; +} + +static void mce_clear_info(struct mce_info *mi) +{ + atomic_set(&mi->inuse, 0); +} + +/* * The actual machine check handler. This only handles real * exceptions when something got corrupted coming in through int 18. * @@ -930,6 +1014,7 @@ static void mce_clear_state(unsigned long *toclear) */ void do_machine_check(struct pt_regs *regs, long error_code) { + struct mca_config *cfg = &mca_cfg; struct mce m, *final; int i; int worst = 0; @@ -941,7 +1026,7 @@ void do_machine_check(struct pt_regs *regs, long error_code) int order; /* * If no_way_out gets set, there is no safe way to recover from this - * MCE. If tolerant is cranked up, we'll try anyway. + * MCE. If mca_cfg.tolerant is cranked up, we'll try anyway. */ int no_way_out = 0; /* @@ -950,13 +1035,12 @@ void do_machine_check(struct pt_regs *regs, long error_code) */ int kill_it = 0; DECLARE_BITMAP(toclear, MAX_NR_BANKS); + DECLARE_BITMAP(valid_banks, MAX_NR_BANKS); char *msg = "Unknown"; - atomic_inc(&mce_entry); - - percpu_inc(mce_exception_count); + this_cpu_inc(mce_exception_count); - if (!banks) + if (!cfg->banks) goto out; mce_gather_info(&m, regs); @@ -964,12 +1048,15 @@ void do_machine_check(struct pt_regs *regs, long error_code) final = &__get_cpu_var(mces_seen); *final = m; - no_way_out = mce_no_way_out(&m, &msg); + memset(valid_banks, 0, sizeof(valid_banks)); + no_way_out = mce_no_way_out(&m, &msg, valid_banks, regs); barrier(); /* - * When no restart IP must always kill or panic. + * When no restart IP might need to kill or panic. + * Assume the worst for now, but if we find the + * severity is MCE_AR_SEVERITY we have other options. */ if (!(m.mcgstatus & MCG_STATUS_RIPV)) kill_it = 1; @@ -980,8 +1067,10 @@ void do_machine_check(struct pt_regs *regs, long error_code) * because the first one to see it will clear it. */ order = mce_start(&no_way_out); - for (i = 0; i < banks; i++) { + for (i = 0; i < cfg->banks; i++) { __clear_bit(i, toclear); + if (!test_bit(i, valid_banks)) + continue; if (!mce_banks[i].ctl) continue; @@ -997,16 +1086,16 @@ void do_machine_check(struct pt_regs *regs, long error_code) * Non uncorrected or non signaled errors are handled by * machine_check_poll. Leave them alone, unless this panics. */ - if (!(m.status & (mce_ser ? MCI_STATUS_S : MCI_STATUS_UC)) && + if (!(m.status & (cfg->ser ? MCI_STATUS_S : MCI_STATUS_UC)) && !no_way_out) continue; /* * Set taint even when machine check was not enabled. */ - add_taint(TAINT_MACHINE_CHECK); + add_taint(TAINT_MACHINE_CHECK, LOCKDEP_NOW_UNRELIABLE); - severity = mce_severity(&m, tolerant, NULL); + severity = mce_severity(&m, cfg->tolerant, NULL); /* * When machine check was for corrected handler don't touch, @@ -1023,23 +1112,14 @@ void do_machine_check(struct pt_regs *regs, long error_code) continue; } - /* - * Kill on action required. - */ - if (severity == MCE_AR_SEVERITY) - kill_it = 1; - - if (m.status & MCI_STATUS_MISCV) - m.misc = mce_rdmsrl(MSR_IA32_MCx_MISC(i)); - if (m.status & MCI_STATUS_ADDRV) - m.addr = mce_rdmsrl(MSR_IA32_MCx_ADDR(i)); + mce_read_aux(&m, i); /* * Action optional error. Queue address for later processing. * When the ring overflows we just ignore the AO error. * RED-PEN add some logging mechanism when * usable_address or mce_add_ring fails. - * RED-PEN don't ignore overflow for tolerant == 0 + * RED-PEN don't ignore overflow for mca_cfg.tolerant == 0 */ if (severity == MCE_AO_SEVERITY && mce_usable_address(&m)) mce_ring_add(m.addr >> PAGE_SHIFT); @@ -1052,6 +1132,9 @@ void do_machine_check(struct pt_regs *regs, long error_code) } } + /* mce_clear_state will clear *final, save locally for use later */ + m = *final; + if (!no_way_out) mce_clear_state(toclear); @@ -1063,65 +1146,91 @@ void do_machine_check(struct pt_regs *regs, long error_code) no_way_out = worst >= MCE_PANIC_SEVERITY; /* - * If we have decided that we just CAN'T continue, and the user - * has not set tolerant to an insane level, give up and die. - * - * This is mainly used in the case when the system doesn't - * support MCE broadcasting or it has been disabled. + * At insane "tolerant" levels we take no action. Otherwise + * we only die if we have no other choice. For less serious + * issues we try to recover, or limit damage to the current + * process. */ - if (no_way_out && tolerant < 3) - mce_panic("Fatal machine check on current CPU", final, msg); - - /* - * If the error seems to be unrecoverable, something should be - * done. Try to kill as little as possible. If we can kill just - * one task, do that. If the user has set the tolerance very - * high, don't try to do anything at all. - */ - - if (kill_it && tolerant < 3) - force_sig(SIGBUS, current); - - /* notify userspace ASAP */ - set_thread_flag(TIF_MCE_NOTIFY); + if (cfg->tolerant < 3) { + if (no_way_out) + mce_panic("Fatal machine check on current CPU", &m, msg); + if (worst == MCE_AR_SEVERITY) { + /* schedule action before return to userland */ + mce_save_info(m.addr, m.mcgstatus & MCG_STATUS_RIPV); + set_thread_flag(TIF_MCE_NOTIFY); + } else if (kill_it) { + force_sig(SIGBUS, current); + } + } if (worst > 0) mce_report_event(regs); mce_wrmsrl(MSR_IA32_MCG_STATUS, 0); out: - atomic_dec(&mce_entry); sync_core(); } EXPORT_SYMBOL_GPL(do_machine_check); -/* dummy to break dependency. actual code is in mm/memory-failure.c */ -void __attribute__((weak)) memory_failure(unsigned long pfn, int vector) +#ifndef CONFIG_MEMORY_FAILURE +int memory_failure(unsigned long pfn, int vector, int flags) { - printk(KERN_ERR "Action optional memory failure at %lx ignored\n", pfn); + /* mce_severity() should not hand us an ACTION_REQUIRED error */ + BUG_ON(flags & MF_ACTION_REQUIRED); + pr_err("Uncorrected memory error in page 0x%lx ignored\n" + "Rebuild kernel with CONFIG_MEMORY_FAILURE=y for smarter handling\n", + pfn); + + return 0; } +#endif /* - * Called after mce notification in process context. This code - * is allowed to sleep. Call the high level VM handler to process - * any corrupted pages. - * Assume that the work queue code only calls this one at a time - * per CPU. - * Note we don't disable preemption, so this code might run on the wrong - * CPU. In this case the event is picked up by the scheduled work queue. - * This is merely a fast path to expedite processing in some common - * cases. + * Called in process context that interrupted by MCE and marked with + * TIF_MCE_NOTIFY, just before returning to erroneous userland. + * This code is allowed to sleep. + * Attempt possible recovery such as calling the high level VM handler to + * process any corrupted pages, and kill/signal current process if required. + * Action required errors are handled here. */ void mce_notify_process(void) { unsigned long pfn; - mce_notify_irq(); - while (mce_ring_get(&pfn)) - memory_failure(pfn, MCE_VECTOR); + struct mce_info *mi = mce_find_info(); + int flags = MF_ACTION_REQUIRED; + + if (!mi) + mce_panic("Lost physical address for unconsumed uncorrectable error", NULL, NULL); + pfn = mi->paddr >> PAGE_SHIFT; + + clear_thread_flag(TIF_MCE_NOTIFY); + + pr_err("Uncorrected hardware memory error in user-access at %llx", + mi->paddr); + /* + * We must call memory_failure() here even if the current process is + * doomed. We still need to mark the page as poisoned and alert any + * other users of the page. + */ + if (!mi->restartable) + flags |= MF_MUST_KILL; + if (memory_failure(pfn, MCE_VECTOR, flags) < 0) { + pr_err("Memory error not recovered"); + force_sig(SIGBUS, current); + } + mce_clear_info(mi); } +/* + * Action optional processing happens here (picking up + * from the list of faulting pages that do_machine_check() + * placed into the "ring"). + */ static void mce_process_work(struct work_struct *dummy) { - mce_notify_process(); + unsigned long pfn; + + while (mce_ring_get(&pfn)) + memory_failure(pfn, MCE_VECTOR, 0); } #ifdef CONFIG_X86_MCE_INTEL @@ -1154,35 +1263,79 @@ void mce_log_therm_throt_event(__u64 status) * poller finds an MCE, poll 2x faster. When the poller finds no more * errors, poll 2x slower (up to check_interval seconds). */ -static int check_interval = 5 * 60; /* 5 minutes */ +static unsigned long check_interval = 5 * 60; /* 5 minutes */ -static DEFINE_PER_CPU(int, mce_next_interval); /* in jiffies */ +static DEFINE_PER_CPU(unsigned long, mce_next_interval); /* in jiffies */ static DEFINE_PER_CPU(struct timer_list, mce_timer); -static void mce_start_timer(unsigned long data) +static unsigned long mce_adjust_timer_default(unsigned long interval) +{ + return interval; +} + +static unsigned long (*mce_adjust_timer)(unsigned long interval) = + mce_adjust_timer_default; + +static int cmc_error_seen(void) +{ + unsigned long *v = &__get_cpu_var(mce_polled_error); + + return test_and_clear_bit(0, v); +} + +static void mce_timer_fn(unsigned long data) { - struct timer_list *t = &per_cpu(mce_timer, data); - int *n; + struct timer_list *t = &__get_cpu_var(mce_timer); + unsigned long iv; + int notify; WARN_ON(smp_processor_id() != data); if (mce_available(__this_cpu_ptr(&cpu_info))) { machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_poll_banks)); + mce_intel_cmci_poll(); } /* * Alert userspace if needed. If we logged an MCE, reduce the * polling interval, otherwise increase the polling interval. */ - n = &__get_cpu_var(mce_next_interval); - if (mce_notify_irq()) - *n = max(*n/2, HZ/100); - else - *n = min(*n*2, (int)round_jiffies_relative(check_interval*HZ)); + iv = __this_cpu_read(mce_next_interval); + notify = mce_notify_irq(); + notify |= cmc_error_seen(); + if (notify) { + iv = max(iv / 2, (unsigned long) HZ/100); + } else { + iv = min(iv * 2, round_jiffies_relative(check_interval * HZ)); + iv = mce_adjust_timer(iv); + } + __this_cpu_write(mce_next_interval, iv); + /* Might have become 0 after CMCI storm subsided */ + if (iv) { + t->expires = jiffies + iv; + add_timer_on(t, smp_processor_id()); + } +} - t->expires = jiffies + *n; - add_timer_on(t, smp_processor_id()); +/* + * Ensure that the timer is firing in @interval from now. + */ +void mce_timer_kick(unsigned long interval) +{ + struct timer_list *t = &__get_cpu_var(mce_timer); + unsigned long when = jiffies + interval; + unsigned long iv = __this_cpu_read(mce_next_interval); + + if (timer_pending(t)) { + if (time_before(when, t->expires)) + mod_timer_pinned(t, when); + } else { + t->expires = round_jiffies(when); + add_timer_on(t, smp_processor_id()); + } + if (interval < iv) + __this_cpu_write(mce_next_interval, interval); } /* Must not be called in IRQ context where del_timer_sync() can deadlock */ @@ -1211,18 +1364,11 @@ int mce_notify_irq(void) /* Not more than two messages every minute */ static DEFINE_RATELIMIT_STATE(ratelimit, 60*HZ, 2); - clear_thread_flag(TIF_MCE_NOTIFY); - if (test_and_clear_bit(0, &mce_need_notify)) { /* wake processes polling /dev/mcelog */ wake_up_interruptible(&mce_chrdev_wait); - /* - * There is no risk of missing notifications because - * work_pending is always cleared before the function is - * executed. - */ - if (mce_helper[0] && !work_pending(&mce_trigger_work)) + if (mce_helper[0]) schedule_work(&mce_trigger_work); if (__ratelimit(&ratelimit)) @@ -1234,14 +1380,16 @@ int mce_notify_irq(void) } EXPORT_SYMBOL_GPL(mce_notify_irq); -static int __cpuinit __mcheck_cpu_mce_banks_init(void) +static int __mcheck_cpu_mce_banks_init(void) { int i; + u8 num_banks = mca_cfg.banks; - mce_banks = kzalloc(banks * sizeof(struct mce_bank), GFP_KERNEL); + mce_banks = kzalloc(num_banks * sizeof(struct mce_bank), GFP_KERNEL); if (!mce_banks) return -ENOMEM; - for (i = 0; i < banks; i++) { + + for (i = 0; i < num_banks; i++) { struct mce_bank *b = &mce_banks[i]; b->ctl = -1ULL; @@ -1253,7 +1401,7 @@ static int __cpuinit __mcheck_cpu_mce_banks_init(void) /* * Initialize Machine Checks for a CPU. */ -static int __cpuinit __mcheck_cpu_cap_init(void) +static int __mcheck_cpu_cap_init(void) { unsigned b; u64 cap; @@ -1261,19 +1409,19 @@ static int __cpuinit __mcheck_cpu_cap_init(void) rdmsrl(MSR_IA32_MCG_CAP, cap); b = cap & MCG_BANKCNT_MASK; - if (!banks) - printk(KERN_INFO "mce: CPU supports %d MCE banks\n", b); + if (!mca_cfg.banks) + pr_info("CPU supports %d MCE banks\n", b); if (b > MAX_NR_BANKS) { - printk(KERN_WARNING - "MCE: Using only %u machine check banks out of %u\n", + pr_warn("Using only %u machine check banks out of %u\n", MAX_NR_BANKS, b); b = MAX_NR_BANKS; } /* Don't support asymmetric configurations today */ - WARN_ON(banks != 0 && b != banks); - banks = b; + WARN_ON(mca_cfg.banks != 0 && b != mca_cfg.banks); + mca_cfg.banks = b; + if (!mce_banks) { int err = __mcheck_cpu_mce_banks_init(); @@ -1283,25 +1431,29 @@ static int __cpuinit __mcheck_cpu_cap_init(void) /* Use accurate RIP reporting if available. */ if ((cap & MCG_EXT_P) && MCG_EXT_CNT(cap) >= 9) - rip_msr = MSR_IA32_MCG_EIP; + mca_cfg.rip_msr = MSR_IA32_MCG_EIP; if (cap & MCG_SER_P) - mce_ser = 1; + mca_cfg.ser = true; return 0; } static void __mcheck_cpu_init_generic(void) { + enum mcp_flags m_fl = 0; mce_banks_t all_banks; u64 cap; int i; + if (!mca_cfg.bootlog) + m_fl = MCP_DONTLOG; + /* * Log the machine checks left over from the previous reset. */ bitmap_fill(all_banks, MAX_NR_BANKS); - machine_check_poll(MCP_UC|(!mce_bootlog ? MCP_DONTLOG : 0), &all_banks); + machine_check_poll(MCP_UC | m_fl, &all_banks); set_in_cr4(X86_CR4_MCE); @@ -1309,7 +1461,7 @@ static void __mcheck_cpu_init_generic(void) if (cap & MCG_CTL_P) wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff); - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { struct mce_bank *b = &mce_banks[i]; if (!b->init) @@ -1319,17 +1471,47 @@ static void __mcheck_cpu_init_generic(void) } } +/* + * During IFU recovery Sandy Bridge -EP4S processors set the RIPV and + * EIPV bits in MCG_STATUS to zero on the affected logical processor (SDM + * Vol 3B Table 15-20). But this confuses both the code that determines + * whether the machine check occurred in kernel or user mode, and also + * the severity assessment code. Pretend that EIPV was set, and take the + * ip/cs values from the pt_regs that mce_gather_info() ignored earlier. + */ +static void quirk_sandybridge_ifu(int bank, struct mce *m, struct pt_regs *regs) +{ + if (bank != 0) + return; + if ((m->mcgstatus & (MCG_STATUS_EIPV|MCG_STATUS_RIPV)) != 0) + return; + if ((m->status & (MCI_STATUS_OVER|MCI_STATUS_UC| + MCI_STATUS_EN|MCI_STATUS_MISCV|MCI_STATUS_ADDRV| + MCI_STATUS_PCC|MCI_STATUS_S|MCI_STATUS_AR| + MCACOD)) != + (MCI_STATUS_UC|MCI_STATUS_EN| + MCI_STATUS_MISCV|MCI_STATUS_ADDRV|MCI_STATUS_S| + MCI_STATUS_AR|MCACOD_INSTR)) + return; + + m->mcgstatus |= MCG_STATUS_EIPV; + m->ip = regs->ip; + m->cs = regs->cs; +} + /* Add per CPU specific workarounds here */ -static int __cpuinit __mcheck_cpu_apply_quirks(struct cpuinfo_x86 *c) +static int __mcheck_cpu_apply_quirks(struct cpuinfo_x86 *c) { + struct mca_config *cfg = &mca_cfg; + if (c->x86_vendor == X86_VENDOR_UNKNOWN) { - pr_info("MCE: unknown CPU type - not enabling MCE support.\n"); + pr_info("unknown CPU type - not enabling MCE support\n"); return -EOPNOTSUPP; } /* This should be disabled by the BIOS, but isn't always */ if (c->x86_vendor == X86_VENDOR_AMD) { - if (c->x86 == 15 && banks > 4) { + if (c->x86 == 15 && cfg->banks > 4) { /* * disable GART TBL walk error reporting, which * trips off incorrectly with the IOMMU & 3ware @@ -1337,19 +1519,56 @@ static int __cpuinit __mcheck_cpu_apply_quirks(struct cpuinfo_x86 *c) */ clear_bit(10, (unsigned long *)&mce_banks[4].ctl); } - if (c->x86 <= 17 && mce_bootlog < 0) { + if (c->x86 <= 17 && cfg->bootlog < 0) { /* * Lots of broken BIOS around that don't clear them * by default and leave crap in there. Don't log: */ - mce_bootlog = 0; + cfg->bootlog = 0; } /* * Various K7s with broken bank 0 around. Always disable * by default. */ - if (c->x86 == 6 && banks > 0) + if (c->x86 == 6 && cfg->banks > 0) mce_banks[0].ctl = 0; + + /* + * Turn off MC4_MISC thresholding banks on those models since + * they're not supported there. + */ + if (c->x86 == 0x15 && + (c->x86_model >= 0x10 && c->x86_model <= 0x1f)) { + int i; + u64 val, hwcr; + bool need_toggle; + u32 msrs[] = { + 0x00000413, /* MC4_MISC0 */ + 0xc0000408, /* MC4_MISC1 */ + }; + + rdmsrl(MSR_K7_HWCR, hwcr); + + /* McStatusWrEn has to be set */ + need_toggle = !(hwcr & BIT(18)); + + if (need_toggle) + wrmsrl(MSR_K7_HWCR, hwcr | BIT(18)); + + for (i = 0; i < ARRAY_SIZE(msrs); i++) { + rdmsrl(msrs[i], val); + + /* CntP bit set? */ + if (val & BIT_64(62)) { + val &= ~BIT_64(62); + wrmsrl(msrs[i], val); + } + } + + /* restore old settings */ + if (need_toggle) + wrmsrl(MSR_K7_HWCR, hwcr); + } } if (c->x86_vendor == X86_VENDOR_INTEL) { @@ -1362,7 +1581,7 @@ static int __cpuinit __mcheck_cpu_apply_quirks(struct cpuinfo_x86 *c) * valid event later, merely don't write CTL0. */ - if (c->x86 == 6 && c->x86_model < 0x1A && banks > 0) + if (c->x86 == 6 && c->x86_model < 0x1A && cfg->banks > 0) mce_banks[0].init = 0; /* @@ -1370,25 +1589,28 @@ static int __cpuinit __mcheck_cpu_apply_quirks(struct cpuinfo_x86 *c) * synchronization with a one second timeout. */ if ((c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xe)) && - monarch_timeout < 0) - monarch_timeout = USEC_PER_SEC; + cfg->monarch_timeout < 0) + cfg->monarch_timeout = USEC_PER_SEC; /* * There are also broken BIOSes on some Pentium M and * earlier systems: */ - if (c->x86 == 6 && c->x86_model <= 13 && mce_bootlog < 0) - mce_bootlog = 0; + if (c->x86 == 6 && c->x86_model <= 13 && cfg->bootlog < 0) + cfg->bootlog = 0; + + if (c->x86 == 6 && c->x86_model == 45) + quirk_no_way_out = quirk_sandybridge_ifu; } - if (monarch_timeout < 0) - monarch_timeout = 0; - if (mce_bootlog != 0) - mce_panic_timeout = 30; + if (cfg->monarch_timeout < 0) + cfg->monarch_timeout = 0; + if (cfg->bootlog != 0) + cfg->panic_timeout = 30; return 0; } -static int __cpuinit __mcheck_cpu_ancient_init(struct cpuinfo_x86 *c) +static int __mcheck_cpu_ancient_init(struct cpuinfo_x86 *c) { if (c->x86 != 5) return 0; @@ -1412,6 +1634,7 @@ static void __mcheck_cpu_init_vendor(struct cpuinfo_x86 *c) switch (c->x86_vendor) { case X86_VENDOR_INTEL: mce_intel_feature_init(c); + mce_adjust_timer = mce_intel_adjust_timer; break; case X86_VENDOR_AMD: mce_amd_feature_init(c); @@ -1421,27 +1644,32 @@ static void __mcheck_cpu_init_vendor(struct cpuinfo_x86 *c) } } -static void __mcheck_cpu_init_timer(void) +static void mce_start_timer(unsigned int cpu, struct timer_list *t) { - struct timer_list *t = &__get_cpu_var(mce_timer); - int *n = &__get_cpu_var(mce_next_interval); + unsigned long iv = check_interval * HZ; - setup_timer(t, mce_start_timer, smp_processor_id()); - - if (mce_ignore_ce) + if (mca_cfg.ignore_ce || !iv) return; - *n = check_interval * HZ; - if (!*n) - return; - t->expires = round_jiffies(jiffies + *n); - add_timer_on(t, smp_processor_id()); + per_cpu(mce_next_interval, cpu) = iv; + + t->expires = round_jiffies(jiffies + iv); + add_timer_on(t, cpu); +} + +static void __mcheck_cpu_init_timer(void) +{ + struct timer_list *t = &__get_cpu_var(mce_timer); + unsigned int cpu = smp_processor_id(); + + setup_timer(t, mce_timer_fn, cpu); + mce_start_timer(cpu, t); } /* 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", + pr_err("CPU#%d: Unexpected int18 (Machine Check)\n", smp_processor_id()); } @@ -1453,9 +1681,9 @@ void (*machine_check_vector)(struct pt_regs *, long error_code) = * Called for each booted CPU to set up machine checks. * Must be called with preempt off: */ -void __cpuinit mcheck_cpu_init(struct cpuinfo_x86 *c) +void mcheck_cpu_init(struct cpuinfo_x86 *c) { - if (mce_disabled) + if (mca_cfg.disabled) return; if (__mcheck_cpu_ancient_init(c)) @@ -1465,7 +1693,7 @@ void __cpuinit mcheck_cpu_init(struct cpuinfo_x86 *c) return; if (__mcheck_cpu_cap_init() < 0 || __mcheck_cpu_apply_quirks(c) < 0) { - mce_disabled = 1; + mca_cfg.disabled = true; return; } @@ -1541,6 +1769,12 @@ static int __mce_read_apei(char __user **ubuf, size_t usize) /* Error or no more MCE record */ if (rc <= 0) { mce_apei_read_done = 1; + /* + * When ERST is disabled, mce_chrdev_read() should return + * "no record" instead of "no device." + */ + if (rc == -ENODEV) + return 0; return rc; } rc = -EFAULT; @@ -1718,6 +1952,25 @@ static struct miscdevice mce_chrdev_device = { &mce_chrdev_ops, }; +static void __mce_disable_bank(void *arg) +{ + int bank = *((int *)arg); + __clear_bit(bank, __get_cpu_var(mce_poll_banks)); + cmci_disable_bank(bank); +} + +void mce_disable_bank(int bank) +{ + if (bank >= mca_cfg.banks) { + pr_warn(FW_BUG + "Ignoring request to disable invalid MCA bank %d.\n", + bank); + return; + } + set_bit(bank, mce_banks_ce_disabled); + on_each_cpu(__mce_disable_bank, &bank, 1); +} + /* * mce=off Disables machine check * mce=no_cmci Disables CMCI @@ -1728,9 +1981,12 @@ static struct miscdevice mce_chrdev_device = { * check, or 0 to not wait * mce=bootlog Log MCEs from before booting. Disabled by default on AMD. * mce=nobootlog Don't log MCEs from before booting. + * mce=bios_cmci_threshold Don't program the CMCI threshold */ static int __init mcheck_enable(char *str) { + struct mca_config *cfg = &mca_cfg; + if (*str == 0) { enable_p5_mce(); return 1; @@ -1738,24 +1994,25 @@ static int __init mcheck_enable(char *str) if (*str == '=') str++; if (!strcmp(str, "off")) - mce_disabled = 1; + cfg->disabled = true; else if (!strcmp(str, "no_cmci")) - mce_cmci_disabled = 1; + cfg->cmci_disabled = true; else if (!strcmp(str, "dont_log_ce")) - mce_dont_log_ce = 1; + cfg->dont_log_ce = true; else if (!strcmp(str, "ignore_ce")) - mce_ignore_ce = 1; + cfg->ignore_ce = true; else if (!strcmp(str, "bootlog") || !strcmp(str, "nobootlog")) - mce_bootlog = (str[0] == 'b'); + cfg->bootlog = (str[0] == 'b'); + else if (!strcmp(str, "bios_cmci_threshold")) + cfg->bios_cmci_threshold = true; else if (isdigit(str[0])) { - get_option(&str, &tolerant); + get_option(&str, &(cfg->tolerant)); if (*str == ',') { ++str; - get_option(&str, &monarch_timeout); + get_option(&str, &(cfg->monarch_timeout)); } } else { - printk(KERN_INFO "mce argument %s ignored. Please use /sys\n", - str); + pr_info("mce argument %s ignored. Please use /sys\n", str); return 0; } return 1; @@ -1781,7 +2038,7 @@ static int mce_disable_error_reporting(void) { int i; - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { struct mce_bank *b = &mce_banks[i]; if (b->init) @@ -1859,9 +2116,8 @@ static struct bus_type mce_subsys = { .dev_name = "machinecheck", }; -struct device *mce_device[CONFIG_NR_CPUS]; +DEFINE_PER_CPU(struct device *, mce_device); -__cpuinitdata void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu); static inline struct mce_bank *attr_to_bank(struct device_attribute *attr) @@ -1921,15 +2177,15 @@ static ssize_t set_ignore_ce(struct device *s, if (strict_strtoull(buf, 0, &new) < 0) return -EINVAL; - if (mce_ignore_ce ^ !!new) { + if (mca_cfg.ignore_ce ^ !!new) { if (new) { /* disable ce features */ mce_timer_delete_all(); on_each_cpu(mce_disable_cmci, NULL, 1); - mce_ignore_ce = 1; + mca_cfg.ignore_ce = true; } else { /* enable ce features */ - mce_ignore_ce = 0; + mca_cfg.ignore_ce = false; on_each_cpu(mce_enable_ce, (void *)1, 1); } } @@ -1945,14 +2201,14 @@ static ssize_t set_cmci_disabled(struct device *s, if (strict_strtoull(buf, 0, &new) < 0) return -EINVAL; - if (mce_cmci_disabled ^ !!new) { + if (mca_cfg.cmci_disabled ^ !!new) { if (new) { /* disable cmci */ on_each_cpu(mce_disable_cmci, NULL, 1); - mce_cmci_disabled = 1; + mca_cfg.cmci_disabled = true; } else { /* enable cmci */ - mce_cmci_disabled = 0; + mca_cfg.cmci_disabled = false; on_each_cpu(mce_enable_ce, NULL, 1); } } @@ -1969,9 +2225,9 @@ static ssize_t store_int_with_restart(struct device *s, } static DEVICE_ATTR(trigger, 0644, show_trigger, set_trigger); -static DEVICE_INT_ATTR(tolerant, 0644, tolerant); -static DEVICE_INT_ATTR(monarch_timeout, 0644, monarch_timeout); -static DEVICE_INT_ATTR(dont_log_ce, 0644, mce_dont_log_ce); +static DEVICE_INT_ATTR(tolerant, 0644, mca_cfg.tolerant); +static DEVICE_INT_ATTR(monarch_timeout, 0644, mca_cfg.monarch_timeout); +static DEVICE_BOOL_ATTR(dont_log_ce, 0644, mca_cfg.dont_log_ce); static struct dev_ext_attribute dev_attr_check_interval = { __ATTR(check_interval, 0644, device_show_int, store_int_with_restart), @@ -1979,13 +2235,13 @@ static struct dev_ext_attribute dev_attr_check_interval = { }; static struct dev_ext_attribute dev_attr_ignore_ce = { - __ATTR(ignore_ce, 0644, device_show_int, set_ignore_ce), - &mce_ignore_ce + __ATTR(ignore_ce, 0644, device_show_bool, set_ignore_ce), + &mca_cfg.ignore_ce }; static struct dev_ext_attribute dev_attr_cmci_disabled = { - __ATTR(cmci_disabled, 0644, device_show_int, set_cmci_disabled), - &mce_cmci_disabled + __ATTR(cmci_disabled, 0644, device_show_bool, set_cmci_disabled), + &mca_cfg.cmci_disabled }; static struct device_attribute *mce_device_attrs[] = { @@ -2007,7 +2263,7 @@ static void mce_device_release(struct device *dev) } /* Per cpu device init. All of the cpus still share the same ctrl bank: */ -static __cpuinit int mce_device_create(unsigned int cpu) +static int mce_device_create(unsigned int cpu) { struct device *dev; int err; @@ -2024,21 +2280,23 @@ static __cpuinit int mce_device_create(unsigned int cpu) dev->release = &mce_device_release; err = device_register(dev); - if (err) + if (err) { + put_device(dev); return err; + } for (i = 0; mce_device_attrs[i]; i++) { err = device_create_file(dev, mce_device_attrs[i]); if (err) goto error; } - for (j = 0; j < banks; j++) { + for (j = 0; j < mca_cfg.banks; j++) { err = device_create_file(dev, &mce_banks[j].attr); if (err) goto error2; } cpumask_set_cpu(cpu, mce_device_initialized); - mce_device[cpu] = dev; + per_cpu(mce_device, cpu) = dev; return 0; error2: @@ -2053,9 +2311,9 @@ error: return err; } -static __cpuinit void mce_device_remove(unsigned int cpu) +static void mce_device_remove(unsigned int cpu) { - struct device *dev = mce_device[cpu]; + struct device *dev = per_cpu(mce_device, cpu); int i; if (!cpumask_test_cpu(cpu, mce_device_initialized)) @@ -2064,16 +2322,16 @@ static __cpuinit void mce_device_remove(unsigned int cpu) for (i = 0; mce_device_attrs[i]; i++) device_remove_file(dev, mce_device_attrs[i]); - for (i = 0; i < banks; i++) + for (i = 0; i < mca_cfg.banks; i++) device_remove_file(dev, &mce_banks[i].attr); device_unregister(dev); cpumask_clear_cpu(cpu, mce_device_initialized); - mce_device[cpu] = NULL; + per_cpu(mce_device, cpu) = NULL; } /* Make sure there are no machine checks on offlined CPUs. */ -static void __cpuinit mce_disable_cpu(void *h) +static void mce_disable_cpu(void *h) { unsigned long action = *(unsigned long *)h; int i; @@ -2083,7 +2341,7 @@ static void __cpuinit mce_disable_cpu(void *h) if (!(action & CPU_TASKS_FROZEN)) cmci_clear(); - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { struct mce_bank *b = &mce_banks[i]; if (b->init) @@ -2091,7 +2349,7 @@ static void __cpuinit mce_disable_cpu(void *h) } } -static void __cpuinit mce_reenable_cpu(void *h) +static void mce_reenable_cpu(void *h) { unsigned long action = *(unsigned long *)h; int i; @@ -2101,7 +2359,7 @@ static void __cpuinit mce_reenable_cpu(void *h) if (!(action & CPU_TASKS_FROZEN)) cmci_reenable(); - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { struct mce_bank *b = &mce_banks[i]; if (b->init) @@ -2110,48 +2368,43 @@ static void __cpuinit mce_reenable_cpu(void *h) } /* Get notified when a cpu comes on/off. Be hotplug friendly. */ -static int __cpuinit +static int mce_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) { unsigned int cpu = (unsigned long)hcpu; struct timer_list *t = &per_cpu(mce_timer, cpu); - switch (action) { + switch (action & ~CPU_TASKS_FROZEN) { case CPU_ONLINE: - case CPU_ONLINE_FROZEN: mce_device_create(cpu); if (threshold_cpu_callback) threshold_cpu_callback(action, cpu); break; case CPU_DEAD: - case CPU_DEAD_FROZEN: if (threshold_cpu_callback) threshold_cpu_callback(action, cpu); mce_device_remove(cpu); + mce_intel_hcpu_update(cpu); break; case CPU_DOWN_PREPARE: - case CPU_DOWN_PREPARE_FROZEN: - del_timer_sync(t); smp_call_function_single(cpu, mce_disable_cpu, &action, 1); + del_timer_sync(t); break; case CPU_DOWN_FAILED: - case CPU_DOWN_FAILED_FROZEN: - if (!mce_ignore_ce && check_interval) { - t->expires = round_jiffies(jiffies + - __get_cpu_var(mce_next_interval)); - add_timer_on(t, cpu); - } smp_call_function_single(cpu, mce_reenable_cpu, &action, 1); + mce_start_timer(cpu, t); break; - case CPU_POST_DEAD: + } + + if (action == CPU_POST_DEAD) { /* intentionally ignoring frozen here */ - cmci_rediscover(cpu); - break; + cmci_rediscover(); } + return NOTIFY_OK; } -static struct notifier_block mce_cpu_notifier __cpuinitdata = { +static struct notifier_block mce_cpu_notifier = { .notifier_call = mce_cpu_callback, }; @@ -2159,7 +2412,7 @@ static __init void mce_init_banks(void) { int i; - for (i = 0; i < banks; i++) { + for (i = 0; i < mca_cfg.banks; i++) { struct mce_bank *b = &mce_banks[i]; struct device_attribute *a = &b->attr; @@ -2178,39 +2431,78 @@ static __init int mcheck_init_device(void) int err; int i = 0; - if (!mce_available(&boot_cpu_data)) - return -EIO; + if (!mce_available(&boot_cpu_data)) { + err = -EIO; + goto err_out; + } - zalloc_cpumask_var(&mce_device_initialized, GFP_KERNEL); + if (!zalloc_cpumask_var(&mce_device_initialized, GFP_KERNEL)) { + err = -ENOMEM; + goto err_out; + } mce_init_banks(); err = subsys_system_register(&mce_subsys, NULL); if (err) - return err; + goto err_out_mem; + cpu_notifier_register_begin(); for_each_online_cpu(i) { err = mce_device_create(i); - if (err) - return err; + if (err) { + /* + * Register notifier anyway (and do not unreg it) so + * that we don't leave undeleted timers, see notifier + * callback above. + */ + __register_hotcpu_notifier(&mce_cpu_notifier); + cpu_notifier_register_done(); + goto err_device_create; + } } + __register_hotcpu_notifier(&mce_cpu_notifier); + cpu_notifier_register_done(); + register_syscore_ops(&mce_syscore_ops); - register_hotcpu_notifier(&mce_cpu_notifier); /* register character device /dev/mcelog */ - misc_register(&mce_chrdev_device); + err = misc_register(&mce_chrdev_device); + if (err) + goto err_register; + + return 0; + +err_register: + unregister_syscore_ops(&mce_syscore_ops); + +err_device_create: + /* + * We didn't keep track of which devices were created above, but + * even if we had, the set of online cpus might have changed. + * Play safe and remove for every possible cpu, since + * mce_device_remove() will do the right thing. + */ + for_each_possible_cpu(i) + mce_device_remove(i); + +err_out_mem: + free_cpumask_var(mce_device_initialized); + +err_out: + pr_err("Unable to init device /dev/mcelog (rc: %d)\n", err); return err; } -device_initcall(mcheck_init_device); +device_initcall_sync(mcheck_init_device); /* * Old style boot options parsing. Only for compatibility. */ static int __init mcheck_disable(char *str) { - mce_disabled = 1; + mca_cfg.disabled = true; return 1; } __setup("nomce", mcheck_disable); diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd.c b/arch/x86/kernel/cpu/mcheck/mce_amd.c index e4eeaaf58a4..603df4f7464 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_amd.c +++ b/arch/x86/kernel/cpu/mcheck/mce_amd.c @@ -1,15 +1,17 @@ /* - * (c) 2005, 2006 Advanced Micro Devices, Inc. + * (c) 2005-2012 Advanced Micro Devices, Inc. * Your use of this code is subject to the terms and conditions of the * GNU general public license version 2. See "COPYING" or * http://www.gnu.org/licenses/gpl.html * * Written by Jacob Shin - AMD, Inc. * - * Support : jacob.shin@amd.com + * Maintained by: Borislav Petkov <bp@alien8.de> * * April 2006 * - added support for AMD Family 0x10 processors + * May 2012 + * - major scrubbing * * All MC4_MISCi registers are shared between multi-cores */ @@ -25,12 +27,12 @@ #include <linux/cpu.h> #include <linux/smp.h> +#include <asm/amd_nb.h> #include <asm/apic.h> #include <asm/idle.h> #include <asm/mce.h> #include <asm/msr.h> -#define NR_BANKS 6 #define NR_BLOCKS 9 #define THRESHOLD_MAX 0xFFF #define INT_TYPE_APIC 0x00020000 @@ -45,28 +47,16 @@ #define MASK_BLKPTR_LO 0xFF000000 #define MCG_XBLK_ADDR 0xC0000400 -struct threshold_block { - unsigned int block; - unsigned int bank; - unsigned int cpu; - u32 address; - u16 interrupt_enable; - u16 threshold_limit; - struct kobject kobj; - struct list_head miscj; -}; - -struct threshold_bank { - struct kobject *kobj; - struct threshold_block *blocks; - cpumask_var_t cpus; -}; -static DEFINE_PER_CPU(struct threshold_bank * [NR_BANKS], threshold_banks); - -static unsigned char shared_bank[NR_BANKS] = { - 0, 0, 0, 0, 1 +static const char * const th_names[] = { + "load_store", + "insn_fetch", + "combined_unit", + "", + "northbridge", + "execution_unit", }; +static DEFINE_PER_CPU(struct threshold_bank **, threshold_banks); static DEFINE_PER_CPU(unsigned char, bank_map); /* see which banks are on */ static void amd_threshold_interrupt(void); @@ -83,6 +73,47 @@ struct thresh_restart { u16 old_limit; }; +static inline bool is_shared_bank(int bank) +{ + /* Bank 4 is for northbridge reporting and is thus shared */ + return (bank == 4); +} + +static const char * const bank4_names(struct threshold_block *b) +{ + switch (b->address) { + /* MSR4_MISC0 */ + case 0x00000413: + return "dram"; + + case 0xc0000408: + return "ht_links"; + + case 0xc0000409: + return "l3_cache"; + + default: + WARN(1, "Funny MSR: 0x%08x\n", b->address); + return ""; + } +}; + + +static bool lvt_interrupt_supported(unsigned int bank, u32 msr_high_bits) +{ + /* + * bank 4 supports APIC LVT interrupts implicitly since forever. + */ + if (bank == 4) + return true; + + /* + * IntP: interrupt present; if this bit is set, the thresholding + * bank can generate APIC LVT interrupts + */ + return msr_high_bits & BIT(28); +} + static int lvt_off_valid(struct threshold_block *b, int apic, u32 lo, u32 hi) { int msr = (hi & MASK_LVTOFF_HI) >> 20; @@ -104,8 +135,10 @@ static int lvt_off_valid(struct threshold_block *b, int apic, u32 lo, u32 hi) return 1; }; -/* must be called with correct cpu affinity */ -/* Called via smp_call_function_single() */ +/* + * Called via smp_call_function_single(), must be called with correct + * cpu affinity. + */ static void threshold_restart_bank(void *_tr) { struct thresh_restart *tr = _tr; @@ -128,6 +161,12 @@ static void threshold_restart_bank(void *_tr) (new_count & THRESHOLD_MAX); } + /* clear IntType */ + hi &= ~MASK_INT_TYPE_HI; + + if (!tr->b->interrupt_capable) + goto done; + if (tr->set_lvt_off) { if (lvt_off_valid(tr->b, tr->lvt_off, lo, hi)) { /* set new lvt offset */ @@ -136,9 +175,10 @@ static void threshold_restart_bank(void *_tr) } } - tr->b->interrupt_enable ? - (hi = (hi & ~MASK_INT_TYPE_HI) | INT_TYPE_APIC) : - (hi &= ~MASK_INT_TYPE_HI); + if (tr->b->interrupt_enable) + hi |= INT_TYPE_APIC; + + done: hi |= MASK_COUNT_EN_HI; wrmsr(tr->b->address, lo, hi); @@ -174,7 +214,7 @@ void mce_amd_feature_init(struct cpuinfo_x86 *c) unsigned int bank, block; int offset = -1; - for (bank = 0; bank < NR_BANKS; ++bank) { + for (bank = 0; bank < mca_cfg.banks; ++bank) { for (block = 0; block < NR_BLOCKS; ++block) { if (block == 0) address = MSR_IA32_MC0_MISC + bank * 4; @@ -199,17 +239,18 @@ void mce_amd_feature_init(struct cpuinfo_x86 *c) if (!block) per_cpu(bank_map, cpu) |= (1 << bank); - if (shared_bank[bank] && c->cpu_core_id) - break; - - offset = setup_APIC_mce(offset, - (high & MASK_LVTOFF_HI) >> 20); memset(&b, 0, sizeof(b)); - b.cpu = cpu; - b.bank = bank; - b.block = block; - b.address = address; + b.cpu = cpu; + b.bank = bank; + b.block = block; + b.address = address; + b.interrupt_capable = lvt_interrupt_supported(bank, high); + + if (b.interrupt_capable) { + int new = (high & MASK_LVTOFF_HI) >> 20; + offset = setup_APIC_mce(offset, new); + } mce_threshold_block_init(&b, offset); mce_threshold_vector = amd_threshold_interrupt; @@ -235,7 +276,7 @@ static void amd_threshold_interrupt(void) mce_setup(&m); /* assume first bank caused it */ - for (bank = 0; bank < NR_BANKS; ++bank) { + for (bank = 0; bank < mca_cfg.banks; ++bank) { if (!(per_cpu(bank_map, m.cpu) & (1 << bank))) continue; for (block = 0; block < NR_BLOCKS; ++block) { @@ -298,7 +339,7 @@ struct threshold_attr { #define SHOW_FIELDS(name) \ static ssize_t show_ ## name(struct threshold_block *b, char *buf) \ { \ - return sprintf(buf, "%lx\n", (unsigned long) b->name); \ + return sprintf(buf, "%lu\n", (unsigned long) b->name); \ } SHOW_FIELDS(interrupt_enable) SHOW_FIELDS(threshold_limit) @@ -309,6 +350,9 @@ store_interrupt_enable(struct threshold_block *b, const char *buf, size_t size) struct thresh_restart tr; unsigned long new; + if (!b->interrupt_capable) + return -EINVAL; + if (strict_strtoul(buf, 0, &new) < 0) return -EINVAL; @@ -346,38 +390,21 @@ store_threshold_limit(struct threshold_block *b, const char *buf, size_t size) return size; } -struct threshold_block_cross_cpu { - struct threshold_block *tb; - long retval; -}; - -static void local_error_count_handler(void *_tbcc) -{ - struct threshold_block_cross_cpu *tbcc = _tbcc; - struct threshold_block *b = tbcc->tb; - u32 low, high; - - rdmsr(b->address, low, high); - tbcc->retval = (high & 0xFFF) - (THRESHOLD_MAX - b->threshold_limit); -} - static ssize_t show_error_count(struct threshold_block *b, char *buf) { - struct threshold_block_cross_cpu tbcc = { .tb = b, }; + u32 lo, hi; - smp_call_function_single(b->cpu, local_error_count_handler, &tbcc, 1); - return sprintf(buf, "%lx\n", tbcc.retval); -} + rdmsr_on_cpu(b->cpu, b->address, &lo, &hi); -static ssize_t store_error_count(struct threshold_block *b, - const char *buf, size_t count) -{ - struct thresh_restart tr = { .b = b, .reset = 1, .old_limit = 0 }; - - smp_call_function_single(b->cpu, threshold_restart_bank, &tr, 1); - return 1; + return sprintf(buf, "%u\n", ((hi & THRESHOLD_MAX) - + (THRESHOLD_MAX - b->threshold_limit))); } +static struct threshold_attr error_count = { + .attr = {.name = __stringify(error_count), .mode = 0444 }, + .show = show_error_count, +}; + #define RW_ATTR(val) \ static struct threshold_attr val = { \ .attr = {.name = __stringify(val), .mode = 0644 }, \ @@ -387,13 +414,12 @@ static struct threshold_attr val = { \ RW_ATTR(interrupt_enable); RW_ATTR(threshold_limit); -RW_ATTR(error_count); static struct attribute *default_attrs[] = { - &interrupt_enable.attr, &threshold_limit.attr, &error_count.attr, - NULL + NULL, /* possibly interrupt_enable if supported, see below */ + NULL, }; #define to_block(k) container_of(k, struct threshold_block, kobj) @@ -432,16 +458,14 @@ static struct kobj_type threshold_ktype = { .default_attrs = default_attrs, }; -static __cpuinit int allocate_threshold_blocks(unsigned int cpu, - unsigned int bank, - unsigned int block, - u32 address) +static int allocate_threshold_blocks(unsigned int cpu, unsigned int bank, + unsigned int block, u32 address) { struct threshold_block *b = NULL; u32 low, high; int err; - if ((bank >= NR_BANKS) || (block >= NR_BLOCKS)) + if ((bank >= mca_cfg.banks) || (block >= NR_BLOCKS)) return 0; if (rdmsr_safe_on_cpu(cpu, address, &low, &high)) @@ -467,8 +491,14 @@ static __cpuinit int allocate_threshold_blocks(unsigned int cpu, b->cpu = cpu; b->address = address; b->interrupt_enable = 0; + b->interrupt_capable = lvt_interrupt_supported(bank, high); b->threshold_limit = THRESHOLD_MAX; + if (b->interrupt_capable) + threshold_ktype.default_attrs[2] = &interrupt_enable.attr; + else + threshold_ktype.default_attrs[2] = NULL; + INIT_LIST_HEAD(&b->miscj); if (per_cpu(threshold_banks, cpu)[bank]->blocks) { @@ -480,7 +510,7 @@ static __cpuinit int allocate_threshold_blocks(unsigned int cpu, err = kobject_init_and_add(&b->kobj, &threshold_ktype, per_cpu(threshold_banks, cpu)[bank]->kobj, - "misc%i", block); + (bank == 4 ? bank4_names(b) : th_names[bank])); if (err) goto out_free; recurse: @@ -511,108 +541,109 @@ out_free: return err; } -static __cpuinit long -local_allocate_threshold_blocks(int cpu, unsigned int bank) +static int __threshold_add_blocks(struct threshold_bank *b) { - return allocate_threshold_blocks(cpu, bank, 0, - MSR_IA32_MC0_MISC + bank * 4); -} + struct list_head *head = &b->blocks->miscj; + struct threshold_block *pos = NULL; + struct threshold_block *tmp = NULL; + int err = 0; -/* symlinks sibling shared banks to first core. first core owns dir/files. */ -static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) -{ - int i, err = 0; - struct threshold_bank *b = NULL; - struct device *dev = mce_device[cpu]; - char name[32]; + err = kobject_add(&b->blocks->kobj, b->kobj, b->blocks->kobj.name); + if (err) + return err; - sprintf(name, "threshold_bank%i", bank); + list_for_each_entry_safe(pos, tmp, head, miscj) { -#ifdef CONFIG_SMP - if (cpu_data(cpu).cpu_core_id && shared_bank[bank]) { /* symlink */ - i = cpumask_first(cpu_llc_shared_mask(cpu)); + err = kobject_add(&pos->kobj, b->kobj, pos->kobj.name); + if (err) { + list_for_each_entry_safe_reverse(pos, tmp, head, miscj) + kobject_del(&pos->kobj); - /* first core not up yet */ - if (cpu_data(i).cpu_core_id) - goto out; + return err; + } + } + return err; +} - /* already linked */ - if (per_cpu(threshold_banks, cpu)[bank]) - goto out; +static int threshold_create_bank(unsigned int cpu, unsigned int bank) +{ + struct device *dev = per_cpu(mce_device, cpu); + struct amd_northbridge *nb = NULL; + struct threshold_bank *b = NULL; + const char *name = th_names[bank]; + int err = 0; - b = per_cpu(threshold_banks, i)[bank]; + if (is_shared_bank(bank)) { + nb = node_to_amd_nb(amd_get_nb_id(cpu)); - if (!b) - goto out; + /* threshold descriptor already initialized on this node? */ + if (nb && nb->bank4) { + /* yes, use it */ + b = nb->bank4; + err = kobject_add(b->kobj, &dev->kobj, name); + if (err) + goto out; - err = sysfs_create_link(&dev->kobj, b->kobj, name); - if (err) - goto out; + per_cpu(threshold_banks, cpu)[bank] = b; + atomic_inc(&b->cpus); - cpumask_copy(b->cpus, cpu_llc_shared_mask(cpu)); - per_cpu(threshold_banks, cpu)[bank] = b; + err = __threshold_add_blocks(b); - goto out; + goto out; + } } -#endif b = kzalloc(sizeof(struct threshold_bank), GFP_KERNEL); if (!b) { err = -ENOMEM; goto out; } - if (!zalloc_cpumask_var(&b->cpus, GFP_KERNEL)) { - kfree(b); - err = -ENOMEM; - goto out; - } b->kobj = kobject_create_and_add(name, &dev->kobj); - if (!b->kobj) + if (!b->kobj) { + err = -EINVAL; goto out_free; - -#ifndef CONFIG_SMP - cpumask_setall(b->cpus); -#else - cpumask_set_cpu(cpu, b->cpus); -#endif + } per_cpu(threshold_banks, cpu)[bank] = b; - err = local_allocate_threshold_blocks(cpu, bank); - if (err) - goto out_free; - - for_each_cpu(i, b->cpus) { - if (i == cpu) - continue; - - dev = mce_device[i]; - if (dev) - err = sysfs_create_link(&dev->kobj,b->kobj, name); - if (err) - goto out; + if (is_shared_bank(bank)) { + atomic_set(&b->cpus, 1); - per_cpu(threshold_banks, i)[bank] = b; + /* nb is already initialized, see above */ + if (nb) { + WARN_ON(nb->bank4); + nb->bank4 = b; + } } - goto out; + err = allocate_threshold_blocks(cpu, bank, 0, + MSR_IA32_MC0_MISC + bank * 4); + if (!err) + goto out; -out_free: - per_cpu(threshold_banks, cpu)[bank] = NULL; - free_cpumask_var(b->cpus); + out_free: kfree(b); -out: + + out: return err; } /* create dir/files for all valid threshold banks */ -static __cpuinit int threshold_create_device(unsigned int cpu) +static int threshold_create_device(unsigned int cpu) { unsigned int bank; + struct threshold_bank **bp; int err = 0; - for (bank = 0; bank < NR_BANKS; ++bank) { + bp = kzalloc(sizeof(struct threshold_bank *) * mca_cfg.banks, + GFP_KERNEL); + if (!bp) + return -ENOMEM; + + per_cpu(threshold_banks, cpu) = bp; + + for (bank = 0; bank < mca_cfg.banks; ++bank) { if (!(per_cpu(bank_map, cpu) & (1 << bank))) continue; err = threshold_create_bank(cpu, bank); @@ -623,12 +654,6 @@ static __cpuinit int threshold_create_device(unsigned int cpu) return err; } -/* - * let's be hotplug friendly. - * in case of multiple core processors, the first core always takes ownership - * of shared sysfs dir/files, and rest of the cores will be symlinked to it. - */ - static void deallocate_threshold_block(unsigned int cpu, unsigned int bank) { @@ -649,40 +674,42 @@ static void deallocate_threshold_block(unsigned int cpu, per_cpu(threshold_banks, cpu)[bank]->blocks = NULL; } +static void __threshold_remove_blocks(struct threshold_bank *b) +{ + struct threshold_block *pos = NULL; + struct threshold_block *tmp = NULL; + + kobject_del(b->kobj); + + list_for_each_entry_safe(pos, tmp, &b->blocks->miscj, miscj) + kobject_del(&pos->kobj); +} + static void threshold_remove_bank(unsigned int cpu, int bank) { + struct amd_northbridge *nb; struct threshold_bank *b; - struct device *dev; - char name[32]; - int i = 0; b = per_cpu(threshold_banks, cpu)[bank]; if (!b) return; + if (!b->blocks) goto free_out; - sprintf(name, "threshold_bank%i", bank); - -#ifdef CONFIG_SMP - /* sibling symlink */ - if (shared_bank[bank] && b->blocks->cpu != cpu) { - sysfs_remove_link(&mce_device[cpu]->kobj, name); - per_cpu(threshold_banks, cpu)[bank] = NULL; - - return; - } -#endif - - /* remove all sibling symlinks before unregistering */ - for_each_cpu(i, b->cpus) { - if (i == cpu) - continue; - - dev = mce_device[i]; - if (dev) - sysfs_remove_link(&dev->kobj, name); - per_cpu(threshold_banks, i)[bank] = NULL; + if (is_shared_bank(bank)) { + if (!atomic_dec_and_test(&b->cpus)) { + __threshold_remove_blocks(b); + per_cpu(threshold_banks, cpu)[bank] = NULL; + return; + } else { + /* + * the last CPU on this node using the shared bank is + * going away, remove that bank now. + */ + nb = node_to_amd_nb(amd_get_nb_id(cpu)); + nb->bank4 = NULL; + } } deallocate_threshold_block(cpu, bank); @@ -690,7 +717,6 @@ static void threshold_remove_bank(unsigned int cpu, int bank) free_out: kobject_del(b->kobj); kobject_put(b->kobj); - free_cpumask_var(b->cpus); kfree(b); per_cpu(threshold_banks, cpu)[bank] = NULL; } @@ -699,15 +725,16 @@ static void threshold_remove_device(unsigned int cpu) { unsigned int bank; - for (bank = 0; bank < NR_BANKS; ++bank) { + for (bank = 0; bank < mca_cfg.banks; ++bank) { if (!(per_cpu(bank_map, cpu) & (1 << bank))) continue; threshold_remove_bank(cpu, bank); } + kfree(per_cpu(threshold_banks, cpu)); } /* get notified when a cpu comes on/off */ -static void __cpuinit +static void amd_64_threshold_cpu_callback(unsigned long action, unsigned int cpu) { switch (action) { @@ -739,4 +766,24 @@ static __init int threshold_init_device(void) return 0; } -device_initcall(threshold_init_device); +/* + * there are 3 funcs which need to be _initcalled in a logic sequence: + * 1. xen_late_init_mcelog + * 2. mcheck_init_device + * 3. threshold_init_device + * + * xen_late_init_mcelog must register xen_mce_chrdev_device before + * native mce_chrdev_device registration if running under xen platform; + * + * mcheck_init_device should be inited before threshold_init_device to + * initialize mce_device, otherwise a NULL ptr dereference will cause panic. + * + * so we use following _initcalls + * 1. device_initcall(xen_late_init_mcelog); + * 2. device_initcall_sync(mcheck_init_device); + * 3. late_initcall(threshold_init_device); + * + * when running under xen, the initcall order is 1,2,3; + * on baremetal, we skip 1 and we do only 2 and 3. + */ +late_initcall(threshold_init_device); diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel.c b/arch/x86/kernel/cpu/mcheck/mce_intel.c index 38e49bc95ff..9a316b21df8 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_intel.c +++ b/arch/x86/kernel/cpu/mcheck/mce_intel.c @@ -6,15 +6,17 @@ */ #include <linux/gfp.h> -#include <linux/init.h> #include <linux/interrupt.h> #include <linux/percpu.h> #include <linux/sched.h> +#include <linux/cpumask.h> #include <asm/apic.h> #include <asm/processor.h> #include <asm/msr.h> #include <asm/mce.h> +#include "mce-internal.h" + /* * Support for Intel Correct Machine Check Interrupts. This allows * the CPU to raise an interrupt when a corrected machine check happened. @@ -22,21 +24,48 @@ * Also supports reliable discovery of shared banks. */ +/* + * CMCI can be delivered to multiple cpus that share a machine check bank + * so we need to designate a single cpu to process errors logged in each bank + * in the interrupt handler (otherwise we would have many races and potential + * double reporting of the same error). + * Note that this can change when a cpu is offlined or brought online since + * some MCA banks are shared across cpus. When a cpu is offlined, cmci_clear() + * disables CMCI on all banks owned by the cpu and clears this bitfield. At + * this point, cmci_rediscover() kicks in and a different cpu may end up + * taking ownership of some of the shared MCA banks that were previously + * owned by the offlined cpu. + */ static DEFINE_PER_CPU(mce_banks_t, mce_banks_owned); /* * cmci_discover_lock protects against parallel discovery attempts * which could race against each other. */ -static DEFINE_RAW_SPINLOCK(cmci_discover_lock); +static DEFINE_SPINLOCK(cmci_discover_lock); + +#define CMCI_THRESHOLD 1 +#define CMCI_POLL_INTERVAL (30 * HZ) +#define CMCI_STORM_INTERVAL (1 * HZ) +#define CMCI_STORM_THRESHOLD 15 + +static DEFINE_PER_CPU(unsigned long, cmci_time_stamp); +static DEFINE_PER_CPU(unsigned int, cmci_storm_cnt); +static DEFINE_PER_CPU(unsigned int, cmci_storm_state); + +enum { + CMCI_STORM_NONE, + CMCI_STORM_ACTIVE, + CMCI_STORM_SUBSIDED, +}; -#define CMCI_THRESHOLD 1 +static atomic_t cmci_storm_on_cpus; static int cmci_supported(int *banks) { u64 cap; - if (mce_cmci_disabled || mce_ignore_ce) + if (mca_cfg.cmci_disabled || mca_cfg.ignore_ce) return 0; /* @@ -53,6 +82,109 @@ static int cmci_supported(int *banks) return !!(cap & MCG_CMCI_P); } +void mce_intel_cmci_poll(void) +{ + if (__this_cpu_read(cmci_storm_state) == CMCI_STORM_NONE) + return; + machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned)); +} + +void mce_intel_hcpu_update(unsigned long cpu) +{ + if (per_cpu(cmci_storm_state, cpu) == CMCI_STORM_ACTIVE) + atomic_dec(&cmci_storm_on_cpus); + + per_cpu(cmci_storm_state, cpu) = CMCI_STORM_NONE; +} + +unsigned long mce_intel_adjust_timer(unsigned long interval) +{ + int r; + + if (interval < CMCI_POLL_INTERVAL) + return interval; + + switch (__this_cpu_read(cmci_storm_state)) { + case CMCI_STORM_ACTIVE: + /* + * We switch back to interrupt mode once the poll timer has + * silenced itself. That means no events recorded and the + * timer interval is back to our poll interval. + */ + __this_cpu_write(cmci_storm_state, CMCI_STORM_SUBSIDED); + r = atomic_sub_return(1, &cmci_storm_on_cpus); + if (r == 0) + pr_notice("CMCI storm subsided: switching to interrupt mode\n"); + /* FALLTHROUGH */ + + case CMCI_STORM_SUBSIDED: + /* + * We wait for all cpus to go back to SUBSIDED + * state. When that happens we switch back to + * interrupt mode. + */ + if (!atomic_read(&cmci_storm_on_cpus)) { + __this_cpu_write(cmci_storm_state, CMCI_STORM_NONE); + cmci_reenable(); + cmci_recheck(); + } + return CMCI_POLL_INTERVAL; + default: + /* + * We have shiny weather. Let the poll do whatever it + * thinks. + */ + return interval; + } +} + +static void cmci_storm_disable_banks(void) +{ + unsigned long flags, *owned; + int bank; + u64 val; + + spin_lock_irqsave(&cmci_discover_lock, flags); + owned = __get_cpu_var(mce_banks_owned); + for_each_set_bit(bank, owned, MAX_NR_BANKS) { + rdmsrl(MSR_IA32_MCx_CTL2(bank), val); + val &= ~MCI_CTL2_CMCI_EN; + wrmsrl(MSR_IA32_MCx_CTL2(bank), val); + } + spin_unlock_irqrestore(&cmci_discover_lock, flags); +} + +static bool cmci_storm_detect(void) +{ + unsigned int cnt = __this_cpu_read(cmci_storm_cnt); + unsigned long ts = __this_cpu_read(cmci_time_stamp); + unsigned long now = jiffies; + int r; + + if (__this_cpu_read(cmci_storm_state) != CMCI_STORM_NONE) + return true; + + if (time_before_eq(now, ts + CMCI_STORM_INTERVAL)) { + cnt++; + } else { + cnt = 1; + __this_cpu_write(cmci_time_stamp, now); + } + __this_cpu_write(cmci_storm_cnt, cnt); + + if (cnt <= CMCI_STORM_THRESHOLD) + return false; + + cmci_storm_disable_banks(); + __this_cpu_write(cmci_storm_state, CMCI_STORM_ACTIVE); + r = atomic_add_return(1, &cmci_storm_on_cpus); + mce_timer_kick(CMCI_POLL_INTERVAL); + + if (r == 1) + pr_notice("CMCI storm detected: switching to poll mode\n"); + return true; +} + /* * The interrupt handler. This is called on every event. * Just call the poller directly to log any events. @@ -61,64 +193,86 @@ static int cmci_supported(int *banks) */ static void intel_threshold_interrupt(void) { + if (cmci_storm_detect()) + return; machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned)); mce_notify_irq(); } -static void print_update(char *type, int *hdr, int num) -{ - if (*hdr == 0) - printk(KERN_INFO "CPU %d MCA banks", smp_processor_id()); - *hdr = 1; - printk(KERN_CONT " %s:%d", type, num); -} - /* * Enable CMCI (Corrected Machine Check Interrupt) for available MCE banks * on this CPU. Use the algorithm recommended in the SDM to discover shared * banks. */ -static void cmci_discover(int banks, int boot) +static void cmci_discover(int banks) { unsigned long *owned = (void *)&__get_cpu_var(mce_banks_owned); unsigned long flags; - int hdr = 0; int i; + int bios_wrong_thresh = 0; - raw_spin_lock_irqsave(&cmci_discover_lock, flags); + spin_lock_irqsave(&cmci_discover_lock, flags); for (i = 0; i < banks; i++) { u64 val; + int bios_zero_thresh = 0; if (test_bit(i, owned)) continue; + /* Skip banks in firmware first mode */ + if (test_bit(i, mce_banks_ce_disabled)) + continue; + rdmsrl(MSR_IA32_MCx_CTL2(i), val); /* Already owned by someone else? */ if (val & MCI_CTL2_CMCI_EN) { - if (test_and_clear_bit(i, owned) && !boot) - print_update("SHD", &hdr, i); + clear_bit(i, owned); __clear_bit(i, __get_cpu_var(mce_poll_banks)); continue; } - val &= ~MCI_CTL2_CMCI_THRESHOLD_MASK; - val |= MCI_CTL2_CMCI_EN | CMCI_THRESHOLD; + if (!mca_cfg.bios_cmci_threshold) { + val &= ~MCI_CTL2_CMCI_THRESHOLD_MASK; + val |= CMCI_THRESHOLD; + } else if (!(val & MCI_CTL2_CMCI_THRESHOLD_MASK)) { + /* + * If bios_cmci_threshold boot option was specified + * but the threshold is zero, we'll try to initialize + * it to 1. + */ + bios_zero_thresh = 1; + val |= CMCI_THRESHOLD; + } + + val |= MCI_CTL2_CMCI_EN; 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 & MCI_CTL2_CMCI_EN) { - if (!test_and_set_bit(i, owned) && !boot) - print_update("CMCI", &hdr, i); + set_bit(i, owned); __clear_bit(i, __get_cpu_var(mce_poll_banks)); + /* + * We are able to set thresholds for some banks that + * had a threshold of 0. This means the BIOS has not + * set the thresholds properly or does not work with + * this boot option. Note down now and report later. + */ + if (mca_cfg.bios_cmci_threshold && bios_zero_thresh && + (val & MCI_CTL2_CMCI_THRESHOLD_MASK)) + bios_wrong_thresh = 1; } else { WARN_ON(!test_bit(i, __get_cpu_var(mce_poll_banks))); } } - raw_spin_unlock_irqrestore(&cmci_discover_lock, flags); - if (hdr) - printk(KERN_CONT "\n"); + spin_unlock_irqrestore(&cmci_discover_lock, flags); + if (mca_cfg.bios_cmci_threshold && bios_wrong_thresh) { + pr_info_once( + "bios_cmci_threshold: Some banks do not have valid thresholds set\n"); + pr_info_once( + "bios_cmci_threshold: Make sure your BIOS supports this boot option\n"); + } } /* @@ -137,6 +291,19 @@ void cmci_recheck(void) local_irq_restore(flags); } +/* Caller must hold the lock on cmci_discover_lock */ +static void __cmci_disable_bank(int bank) +{ + u64 val; + + if (!test_bit(bank, __get_cpu_var(mce_banks_owned))) + return; + rdmsrl(MSR_IA32_MCx_CTL2(bank), val); + val &= ~MCI_CTL2_CMCI_EN; + wrmsrl(MSR_IA32_MCx_CTL2(bank), val); + __clear_bit(bank, __get_cpu_var(mce_banks_owned)); +} + /* * Disable CMCI on this CPU for all banks it owns when it goes down. * This allows other CPUs to claim the banks on rediscovery. @@ -146,51 +313,33 @@ void cmci_clear(void) unsigned long flags; int i; int banks; - u64 val; if (!cmci_supported(&banks)) return; - raw_spin_lock_irqsave(&cmci_discover_lock, flags); - for (i = 0; i < banks; i++) { - if (!test_bit(i, __get_cpu_var(mce_banks_owned))) - continue; - /* Disable CMCI */ - rdmsrl(MSR_IA32_MCx_CTL2(i), val); - 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)); - } - raw_spin_unlock_irqrestore(&cmci_discover_lock, flags); + spin_lock_irqsave(&cmci_discover_lock, flags); + for (i = 0; i < banks; i++) + __cmci_disable_bank(i); + spin_unlock_irqrestore(&cmci_discover_lock, flags); } -/* - * After a CPU went down cycle through all the others and rediscover - * Must run in process context. - */ -void cmci_rediscover(int dying) +static void cmci_rediscover_work_func(void *arg) +{ + int banks; + + /* Recheck banks in case CPUs don't all have the same */ + if (cmci_supported(&banks)) + cmci_discover(banks); +} + +/* After a CPU went down cycle through all the others and rediscover */ +void cmci_rediscover(void) { int banks; - int cpu; - cpumask_var_t old; if (!cmci_supported(&banks)) return; - if (!alloc_cpumask_var(&old, GFP_KERNEL)) - return; - cpumask_copy(old, ¤t->cpus_allowed); - - for_each_online_cpu(cpu) { - if (cpu == dying) - continue; - if (set_cpus_allowed_ptr(current, cpumask_of(cpu))) - continue; - /* Recheck banks in case CPUs don't all have the same */ - if (cmci_supported(&banks)) - cmci_discover(banks, 0); - } - set_cpus_allowed_ptr(current, old); - free_cpumask_var(old); + on_each_cpu(cmci_rediscover_work_func, NULL, 1); } /* @@ -200,7 +349,20 @@ void cmci_reenable(void) { int banks; if (cmci_supported(&banks)) - cmci_discover(banks, 0); + cmci_discover(banks); +} + +void cmci_disable_bank(int bank) +{ + int banks; + unsigned long flags; + + if (!cmci_supported(&banks)) + return; + + spin_lock_irqsave(&cmci_discover_lock, flags); + __cmci_disable_bank(bank); + spin_unlock_irqrestore(&cmci_discover_lock, flags); } static void intel_init_cmci(void) @@ -211,7 +373,7 @@ static void intel_init_cmci(void) return; mce_threshold_vector = intel_threshold_interrupt; - cmci_discover(banks, 1); + cmci_discover(banks); /* * For CPU #0 this runs with still disabled APIC, but that's * ok because only the vector is set up. We still do another diff --git a/arch/x86/kernel/cpu/mcheck/p5.c b/arch/x86/kernel/cpu/mcheck/p5.c index 5c0e6533d9b..a3042989398 100644 --- a/arch/x86/kernel/cpu/mcheck/p5.c +++ b/arch/x86/kernel/cpu/mcheck/p5.c @@ -5,11 +5,9 @@ #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> @@ -34,7 +32,7 @@ static void pentium_machine_check(struct pt_regs *regs, long error_code) smp_processor_id()); } - add_taint(TAINT_MACHINE_CHECK); + add_taint(TAINT_MACHINE_CHECK, LOCKDEP_NOW_UNRELIABLE); } /* Set up machine check reporting for processors with Intel style MCE: */ diff --git a/arch/x86/kernel/cpu/mcheck/therm_throt.c b/arch/x86/kernel/cpu/mcheck/therm_throt.c index 67bb17a37a0..36a1bb6d1ee 100644 --- a/arch/x86/kernel/cpu/mcheck/therm_throt.c +++ b/arch/x86/kernel/cpu/mcheck/therm_throt.c @@ -25,11 +25,11 @@ #include <linux/cpu.h> #include <asm/processor.h> -#include <asm/system.h> #include <asm/apic.h> #include <asm/idle.h> #include <asm/mce.h> #include <asm/msr.h> +#include <asm/trace/irq_vectors.h> /* How long to wait between reporting thermal events */ #define CHECK_INTERVAL (300 * HZ) @@ -55,12 +55,24 @@ struct thermal_state { struct _thermal_state package_power_limit; struct _thermal_state core_thresh0; struct _thermal_state core_thresh1; + struct _thermal_state pkg_thresh0; + struct _thermal_state pkg_thresh1; }; /* Callback to handle core threshold interrupts */ int (*platform_thermal_notify)(__u64 msr_val); EXPORT_SYMBOL(platform_thermal_notify); +/* Callback to handle core package threshold_interrupts */ +int (*platform_thermal_package_notify)(__u64 msr_val); +EXPORT_SYMBOL_GPL(platform_thermal_package_notify); + +/* Callback support of rate control, return true, if + * callback has rate control */ +bool (*platform_thermal_package_rate_control)(void); +EXPORT_SYMBOL_GPL(platform_thermal_package_rate_control); + + static DEFINE_PER_CPU(struct thermal_state, thermal_state); static atomic_t therm_throt_en = ATOMIC_INIT(0); @@ -182,11 +194,6 @@ static int therm_throt_process(bool new_event, int event, int level) 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); return 1; } if (old_event) { @@ -194,36 +201,46 @@ static int therm_throt_process(bool new_event, int event, int level) 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; } return 0; } -static int thresh_event_valid(int event) +static int thresh_event_valid(int level, int event) { struct _thermal_state *state; unsigned int this_cpu = smp_processor_id(); struct thermal_state *pstate = &per_cpu(thermal_state, this_cpu); u64 now = get_jiffies_64(); - state = (event == 0) ? &pstate->core_thresh0 : &pstate->core_thresh1; + if (level == PACKAGE_LEVEL) + state = (event == 0) ? &pstate->pkg_thresh0 : + &pstate->pkg_thresh1; + else + state = (event == 0) ? &pstate->core_thresh0 : + &pstate->core_thresh1; if (time_before64(now, state->next_check)) return 0; state->next_check = now + CHECK_INTERVAL; + + return 1; +} + +static bool int_pln_enable; +static int __init int_pln_enable_setup(char *s) +{ + int_pln_enable = true; + return 1; } +__setup("int_pln_enable", int_pln_enable_setup); #ifdef CONFIG_SYSFS /* Add/Remove thermal_throttle interface for CPU device: */ -static __cpuinit int thermal_throttle_add_dev(struct device *dev, - unsigned int cpu) +static int thermal_throttle_add_dev(struct device *dev, unsigned int cpu) { int err; struct cpuinfo_x86 *c = &cpu_data(cpu); @@ -232,7 +249,7 @@ static __cpuinit int thermal_throttle_add_dev(struct device *dev, if (err) return err; - if (cpu_has(c, X86_FEATURE_PLN)) + if (cpu_has(c, X86_FEATURE_PLN) && int_pln_enable) err = sysfs_add_file_to_group(&dev->kobj, &dev_attr_core_power_limit_count.attr, thermal_attr_group.name); @@ -240,7 +257,7 @@ static __cpuinit int thermal_throttle_add_dev(struct device *dev, err = sysfs_add_file_to_group(&dev->kobj, &dev_attr_package_throttle_count.attr, thermal_attr_group.name); - if (cpu_has(c, X86_FEATURE_PLN)) + if (cpu_has(c, X86_FEATURE_PLN) && int_pln_enable) err = sysfs_add_file_to_group(&dev->kobj, &dev_attr_package_power_limit_count.attr, thermal_attr_group.name); @@ -249,16 +266,13 @@ static __cpuinit int thermal_throttle_add_dev(struct device *dev, return err; } -static __cpuinit void thermal_throttle_remove_dev(struct device *dev) +static void thermal_throttle_remove_dev(struct device *dev) { sysfs_remove_group(&dev->kobj, &thermal_attr_group); } -/* Mutex protecting device creation against CPU hotplug: */ -static DEFINE_MUTEX(therm_cpu_lock); - /* Get notified when a cpu comes on/off. Be hotplug friendly. */ -static __cpuinit int +static int thermal_throttle_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) @@ -272,24 +286,20 @@ thermal_throttle_cpu_callback(struct notifier_block *nfb, switch (action) { case CPU_UP_PREPARE: case CPU_UP_PREPARE_FROZEN: - mutex_lock(&therm_cpu_lock); err = thermal_throttle_add_dev(dev, cpu); - mutex_unlock(&therm_cpu_lock); WARN_ON(err); break; case CPU_UP_CANCELED: case CPU_UP_CANCELED_FROZEN: case CPU_DEAD: case CPU_DEAD_FROZEN: - mutex_lock(&therm_cpu_lock); thermal_throttle_remove_dev(dev); - mutex_unlock(&therm_cpu_lock); break; } return notifier_from_errno(err); } -static struct notifier_block thermal_throttle_cpu_notifier __cpuinitdata = +static struct notifier_block thermal_throttle_cpu_notifier = { .notifier_call = thermal_throttle_cpu_callback, }; @@ -302,19 +312,16 @@ static __init int thermal_throttle_init_device(void) if (!atomic_read(&therm_throt_en)) return 0; - register_hotcpu_notifier(&thermal_throttle_cpu_notifier); + cpu_notifier_register_begin(); -#ifdef CONFIG_HOTPLUG_CPU - mutex_lock(&therm_cpu_lock); -#endif /* connect live CPUs to sysfs */ for_each_online_cpu(cpu) { err = thermal_throttle_add_dev(get_cpu_device(cpu), cpu); WARN_ON(err); } -#ifdef CONFIG_HOTPLUG_CPU - mutex_unlock(&therm_cpu_lock); -#endif + + __register_hotcpu_notifier(&thermal_throttle_cpu_notifier); + cpu_notifier_register_done(); return 0; } @@ -322,6 +329,39 @@ device_initcall(thermal_throttle_init_device); #endif /* CONFIG_SYSFS */ +static void notify_package_thresholds(__u64 msr_val) +{ + bool notify_thres_0 = false; + bool notify_thres_1 = false; + + if (!platform_thermal_package_notify) + return; + + /* lower threshold check */ + if (msr_val & THERM_LOG_THRESHOLD0) + notify_thres_0 = true; + /* higher threshold check */ + if (msr_val & THERM_LOG_THRESHOLD1) + notify_thres_1 = true; + + if (!notify_thres_0 && !notify_thres_1) + return; + + if (platform_thermal_package_rate_control && + platform_thermal_package_rate_control()) { + /* Rate control is implemented in callback */ + platform_thermal_package_notify(msr_val); + return; + } + + /* lower threshold reached */ + if (notify_thres_0 && thresh_event_valid(PACKAGE_LEVEL, 0)) + platform_thermal_package_notify(msr_val); + /* higher threshold reached */ + if (notify_thres_1 && thresh_event_valid(PACKAGE_LEVEL, 1)) + platform_thermal_package_notify(msr_val); +} + static void notify_thresholds(__u64 msr_val) { /* check whether the interrupt handler is defined; @@ -331,10 +371,12 @@ static void notify_thresholds(__u64 msr_val) return; /* lower threshold reached */ - if ((msr_val & THERM_LOG_THRESHOLD0) && thresh_event_valid(0)) + if ((msr_val & THERM_LOG_THRESHOLD0) && + thresh_event_valid(CORE_LEVEL, 0)) platform_thermal_notify(msr_val); /* higher threshold reached */ - if ((msr_val & THERM_LOG_THRESHOLD1) && thresh_event_valid(1)) + if ((msr_val & THERM_LOG_THRESHOLD1) && + thresh_event_valid(CORE_LEVEL, 1)) platform_thermal_notify(msr_val); } @@ -353,17 +395,19 @@ static void intel_thermal_interrupt(void) CORE_LEVEL) != 0) mce_log_therm_throt_event(msr_val); - if (this_cpu_has(X86_FEATURE_PLN)) + if (this_cpu_has(X86_FEATURE_PLN) && int_pln_enable) therm_throt_process(msr_val & THERM_STATUS_POWER_LIMIT, POWER_LIMIT_EVENT, CORE_LEVEL); if (this_cpu_has(X86_FEATURE_PTS)) { rdmsrl(MSR_IA32_PACKAGE_THERM_STATUS, msr_val); + /* check violations of package thermal thresholds */ + notify_package_thresholds(msr_val); therm_throt_process(msr_val & PACKAGE_THERM_STATUS_PROCHOT, THERMAL_THROTTLING_EVENT, PACKAGE_LEVEL); - if (this_cpu_has(X86_FEATURE_PLN)) + if (this_cpu_has(X86_FEATURE_PLN) && int_pln_enable) therm_throt_process(msr_val & PACKAGE_THERM_STATUS_POWER_LIMIT, POWER_LIMIT_EVENT, @@ -379,15 +423,26 @@ static void unexpected_thermal_interrupt(void) static void (*smp_thermal_vector)(void) = unexpected_thermal_interrupt; -asmlinkage void smp_thermal_interrupt(struct pt_regs *regs) +static inline void __smp_thermal_interrupt(void) { - irq_enter(); - exit_idle(); inc_irq_stat(irq_thermal_count); smp_thermal_vector(); - irq_exit(); - /* Ack only at the end to avoid potential reentry */ - ack_APIC_irq(); +} + +asmlinkage __visible void smp_thermal_interrupt(struct pt_regs *regs) +{ + entering_irq(); + __smp_thermal_interrupt(); + exiting_ack_irq(); +} + +asmlinkage __visible void smp_trace_thermal_interrupt(struct pt_regs *regs) +{ + entering_irq(); + trace_thermal_apic_entry(THERMAL_APIC_VECTOR); + __smp_thermal_interrupt(); + trace_thermal_apic_exit(THERMAL_APIC_VECTOR); + exiting_ack_irq(); } /* Thermal monitoring depends on APIC, ACPI and clock modulation */ @@ -471,9 +526,13 @@ void intel_init_thermal(struct cpuinfo_x86 *c) apic_write(APIC_LVTTHMR, h); rdmsr(MSR_IA32_THERM_INTERRUPT, l, h); - if (cpu_has(c, X86_FEATURE_PLN)) + if (cpu_has(c, X86_FEATURE_PLN) && !int_pln_enable) + wrmsr(MSR_IA32_THERM_INTERRUPT, + (l | (THERM_INT_LOW_ENABLE + | THERM_INT_HIGH_ENABLE)) & ~THERM_INT_PLN_ENABLE, h); + else if (cpu_has(c, X86_FEATURE_PLN) && int_pln_enable) wrmsr(MSR_IA32_THERM_INTERRUPT, - l | (THERM_INT_LOW_ENABLE + l | (THERM_INT_LOW_ENABLE | THERM_INT_HIGH_ENABLE | THERM_INT_PLN_ENABLE), h); else wrmsr(MSR_IA32_THERM_INTERRUPT, @@ -481,9 +540,14 @@ void intel_init_thermal(struct cpuinfo_x86 *c) if (cpu_has(c, X86_FEATURE_PTS)) { rdmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, l, h); - if (cpu_has(c, X86_FEATURE_PLN)) + if (cpu_has(c, X86_FEATURE_PLN) && !int_pln_enable) wrmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, - l | (PACKAGE_THERM_INT_LOW_ENABLE + (l | (PACKAGE_THERM_INT_LOW_ENABLE + | PACKAGE_THERM_INT_HIGH_ENABLE)) + & ~PACKAGE_THERM_INT_PLN_ENABLE, h); + else if (cpu_has(c, X86_FEATURE_PLN) && int_pln_enable) + wrmsr(MSR_IA32_PACKAGE_THERM_INTERRUPT, + l | (PACKAGE_THERM_INT_LOW_ENABLE | PACKAGE_THERM_INT_HIGH_ENABLE | PACKAGE_THERM_INT_PLN_ENABLE), h); else diff --git a/arch/x86/kernel/cpu/mcheck/threshold.c b/arch/x86/kernel/cpu/mcheck/threshold.c index aa578cadb94..7245980186e 100644 --- a/arch/x86/kernel/cpu/mcheck/threshold.c +++ b/arch/x86/kernel/cpu/mcheck/threshold.c @@ -8,6 +8,7 @@ #include <asm/apic.h> #include <asm/idle.h> #include <asm/mce.h> +#include <asm/trace/irq_vectors.h> static void default_threshold_interrupt(void) { @@ -17,13 +18,24 @@ static void default_threshold_interrupt(void) void (*mce_threshold_vector)(void) = default_threshold_interrupt; -asmlinkage void smp_threshold_interrupt(void) +static inline void __smp_threshold_interrupt(void) { - irq_enter(); - exit_idle(); inc_irq_stat(irq_threshold_count); mce_threshold_vector(); - irq_exit(); - /* Ack only at the end to avoid potential reentry */ - ack_APIC_irq(); +} + +asmlinkage __visible void smp_threshold_interrupt(void) +{ + entering_irq(); + __smp_threshold_interrupt(); + exiting_ack_irq(); +} + +asmlinkage __visible void smp_trace_threshold_interrupt(void) +{ + entering_irq(); + trace_threshold_apic_entry(THRESHOLD_APIC_VECTOR); + __smp_threshold_interrupt(); + trace_threshold_apic_exit(THRESHOLD_APIC_VECTOR); + exiting_ack_irq(); } diff --git a/arch/x86/kernel/cpu/mcheck/winchip.c b/arch/x86/kernel/cpu/mcheck/winchip.c index 54060f56597..7dc5564d0cd 100644 --- a/arch/x86/kernel/cpu/mcheck/winchip.c +++ b/arch/x86/kernel/cpu/mcheck/winchip.c @@ -5,10 +5,8 @@ #include <linux/interrupt.h> #include <linux/kernel.h> #include <linux/types.h> -#include <linux/init.h> #include <asm/processor.h> -#include <asm/system.h> #include <asm/mce.h> #include <asm/msr.h> @@ -16,7 +14,7 @@ static void winchip_machine_check(struct pt_regs *regs, long error_code) { printk(KERN_EMERG "CPU0: Machine Check Exception.\n"); - add_taint(TAINT_MACHINE_CHECK); + add_taint(TAINT_MACHINE_CHECK, LOCKDEP_NOW_UNRELIABLE); } /* Set up machine check reporting on the Winchip C6 series */ diff --git a/arch/x86/kernel/cpu/microcode/Makefile b/arch/x86/kernel/cpu/microcode/Makefile new file mode 100644 index 00000000000..285c85427c3 --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/Makefile @@ -0,0 +1,7 @@ +microcode-y := core.o +obj-$(CONFIG_MICROCODE) += microcode.o +microcode-$(CONFIG_MICROCODE_INTEL) += intel.o intel_lib.o +microcode-$(CONFIG_MICROCODE_AMD) += amd.o +obj-$(CONFIG_MICROCODE_EARLY) += core_early.o +obj-$(CONFIG_MICROCODE_INTEL_EARLY) += intel_early.o +obj-$(CONFIG_MICROCODE_AMD_EARLY) += amd_early.o diff --git a/arch/x86/kernel/cpu/microcode/amd.c b/arch/x86/kernel/cpu/microcode/amd.c new file mode 100644 index 00000000000..8fffd845e22 --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/amd.c @@ -0,0 +1,492 @@ +/* + * AMD CPU Microcode Update Driver for Linux + * Copyright (C) 2008-2011 Advanced Micro Devices Inc. + * + * Author: Peter Oruba <peter.oruba@amd.com> + * + * Based on work by: + * Tigran Aivazian <tigran@aivazian.fsnet.co.uk> + * + * Maintainers: + * Andreas Herrmann <herrmann.der.user@googlemail.com> + * Borislav Petkov <bp@alien8.de> + * + * This driver allows to upgrade microcode on F10h AMD + * CPUs and later. + * + * Licensed under the terms of the GNU General Public + * License version 2. See file COPYING for details. + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include <linux/firmware.h> +#include <linux/pci_ids.h> +#include <linux/uaccess.h> +#include <linux/vmalloc.h> +#include <linux/kernel.h> +#include <linux/module.h> +#include <linux/pci.h> + +#include <asm/microcode.h> +#include <asm/processor.h> +#include <asm/msr.h> +#include <asm/microcode_amd.h> + +MODULE_DESCRIPTION("AMD Microcode Update Driver"); +MODULE_AUTHOR("Peter Oruba"); +MODULE_LICENSE("GPL v2"); + +static struct equiv_cpu_entry *equiv_cpu_table; + +struct ucode_patch { + struct list_head plist; + void *data; + u32 patch_id; + u16 equiv_cpu; +}; + +static LIST_HEAD(pcache); + +static u16 __find_equiv_id(unsigned int cpu) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + return find_equiv_id(equiv_cpu_table, uci->cpu_sig.sig); +} + +static u32 find_cpu_family_by_equiv_cpu(u16 equiv_cpu) +{ + int i = 0; + + BUG_ON(!equiv_cpu_table); + + while (equiv_cpu_table[i].equiv_cpu != 0) { + if (equiv_cpu == equiv_cpu_table[i].equiv_cpu) + return equiv_cpu_table[i].installed_cpu; + i++; + } + return 0; +} + +/* + * a small, trivial cache of per-family ucode patches + */ +static struct ucode_patch *cache_find_patch(u16 equiv_cpu) +{ + struct ucode_patch *p; + + list_for_each_entry(p, &pcache, plist) + if (p->equiv_cpu == equiv_cpu) + return p; + return NULL; +} + +static void update_cache(struct ucode_patch *new_patch) +{ + struct ucode_patch *p; + + list_for_each_entry(p, &pcache, plist) { + if (p->equiv_cpu == new_patch->equiv_cpu) { + if (p->patch_id >= new_patch->patch_id) + /* we already have the latest patch */ + return; + + list_replace(&p->plist, &new_patch->plist); + kfree(p->data); + kfree(p); + return; + } + } + /* no patch found, add it */ + list_add_tail(&new_patch->plist, &pcache); +} + +static void free_cache(void) +{ + struct ucode_patch *p, *tmp; + + list_for_each_entry_safe(p, tmp, &pcache, plist) { + __list_del(p->plist.prev, p->plist.next); + kfree(p->data); + kfree(p); + } +} + +static struct ucode_patch *find_patch(unsigned int cpu) +{ + u16 equiv_id; + + equiv_id = __find_equiv_id(cpu); + if (!equiv_id) + return NULL; + + return cache_find_patch(equiv_id); +} + +static int collect_cpu_info_amd(int cpu, struct cpu_signature *csig) +{ + struct cpuinfo_x86 *c = &cpu_data(cpu); + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + struct ucode_patch *p; + + csig->sig = cpuid_eax(0x00000001); + csig->rev = c->microcode; + + /* + * a patch could have been loaded early, set uci->mc so that + * mc_bp_resume() can call apply_microcode() + */ + p = find_patch(cpu); + if (p && (p->patch_id == csig->rev)) + uci->mc = p->data; + + pr_info("CPU%d: patch_level=0x%08x\n", cpu, csig->rev); + + return 0; +} + +static unsigned int verify_patch_size(u8 family, u32 patch_size, + unsigned int size) +{ + u32 max_size; + +#define F1XH_MPB_MAX_SIZE 2048 +#define F14H_MPB_MAX_SIZE 1824 +#define F15H_MPB_MAX_SIZE 4096 +#define F16H_MPB_MAX_SIZE 3458 + + switch (family) { + case 0x14: + max_size = F14H_MPB_MAX_SIZE; + break; + case 0x15: + max_size = F15H_MPB_MAX_SIZE; + break; + case 0x16: + max_size = F16H_MPB_MAX_SIZE; + break; + default: + max_size = F1XH_MPB_MAX_SIZE; + break; + } + + if (patch_size > min_t(u32, size, max_size)) { + pr_err("patch size mismatch\n"); + return 0; + } + + return patch_size; +} + +int __apply_microcode_amd(struct microcode_amd *mc_amd) +{ + u32 rev, dummy; + + native_wrmsrl(MSR_AMD64_PATCH_LOADER, (u64)(long)&mc_amd->hdr.data_code); + + /* verify patch application was successful */ + native_rdmsr(MSR_AMD64_PATCH_LEVEL, rev, dummy); + if (rev != mc_amd->hdr.patch_id) + return -1; + + return 0; +} + +int apply_microcode_amd(int cpu) +{ + struct cpuinfo_x86 *c = &cpu_data(cpu); + struct microcode_amd *mc_amd; + struct ucode_cpu_info *uci; + struct ucode_patch *p; + u32 rev, dummy; + + BUG_ON(raw_smp_processor_id() != cpu); + + uci = ucode_cpu_info + cpu; + + p = find_patch(cpu); + if (!p) + return 0; + + mc_amd = p->data; + uci->mc = p->data; + + rdmsr(MSR_AMD64_PATCH_LEVEL, rev, dummy); + + /* need to apply patch? */ + if (rev >= mc_amd->hdr.patch_id) { + c->microcode = rev; + uci->cpu_sig.rev = rev; + return 0; + } + + if (__apply_microcode_amd(mc_amd)) { + pr_err("CPU%d: update failed for patch_level=0x%08x\n", + cpu, mc_amd->hdr.patch_id); + return -1; + } + pr_info("CPU%d: new patch_level=0x%08x\n", cpu, + mc_amd->hdr.patch_id); + + uci->cpu_sig.rev = mc_amd->hdr.patch_id; + c->microcode = mc_amd->hdr.patch_id; + + return 0; +} + +static int install_equiv_cpu_table(const u8 *buf) +{ + unsigned int *ibuf = (unsigned int *)buf; + unsigned int type = ibuf[1]; + unsigned int size = ibuf[2]; + + if (type != UCODE_EQUIV_CPU_TABLE_TYPE || !size) { + pr_err("empty section/" + "invalid type field in container file section header\n"); + return -EINVAL; + } + + equiv_cpu_table = vmalloc(size); + if (!equiv_cpu_table) { + pr_err("failed to allocate equivalent CPU table\n"); + return -ENOMEM; + } + + memcpy(equiv_cpu_table, buf + CONTAINER_HDR_SZ, size); + + /* add header length */ + return size + CONTAINER_HDR_SZ; +} + +static void free_equiv_cpu_table(void) +{ + vfree(equiv_cpu_table); + equiv_cpu_table = NULL; +} + +static void cleanup(void) +{ + free_equiv_cpu_table(); + free_cache(); +} + +/* + * We return the current size even if some of the checks failed so that + * we can skip over the next patch. If we return a negative value, we + * signal a grave error like a memory allocation has failed and the + * driver cannot continue functioning normally. In such cases, we tear + * down everything we've used up so far and exit. + */ +static int verify_and_add_patch(u8 family, u8 *fw, unsigned int leftover) +{ + struct microcode_header_amd *mc_hdr; + struct ucode_patch *patch; + unsigned int patch_size, crnt_size, ret; + u32 proc_fam; + u16 proc_id; + + patch_size = *(u32 *)(fw + 4); + crnt_size = patch_size + SECTION_HDR_SIZE; + mc_hdr = (struct microcode_header_amd *)(fw + SECTION_HDR_SIZE); + proc_id = mc_hdr->processor_rev_id; + + proc_fam = find_cpu_family_by_equiv_cpu(proc_id); + if (!proc_fam) { + pr_err("No patch family for equiv ID: 0x%04x\n", proc_id); + return crnt_size; + } + + /* check if patch is for the current family */ + proc_fam = ((proc_fam >> 8) & 0xf) + ((proc_fam >> 20) & 0xff); + if (proc_fam != family) + return crnt_size; + + if (mc_hdr->nb_dev_id || mc_hdr->sb_dev_id) { + pr_err("Patch-ID 0x%08x: chipset-specific code unsupported.\n", + mc_hdr->patch_id); + return crnt_size; + } + + ret = verify_patch_size(family, patch_size, leftover); + if (!ret) { + pr_err("Patch-ID 0x%08x: size mismatch.\n", mc_hdr->patch_id); + return crnt_size; + } + + patch = kzalloc(sizeof(*patch), GFP_KERNEL); + if (!patch) { + pr_err("Patch allocation failure.\n"); + return -EINVAL; + } + + patch->data = kzalloc(patch_size, GFP_KERNEL); + if (!patch->data) { + pr_err("Patch data allocation failure.\n"); + kfree(patch); + return -EINVAL; + } + + /* All looks ok, copy patch... */ + memcpy(patch->data, fw + SECTION_HDR_SIZE, patch_size); + INIT_LIST_HEAD(&patch->plist); + patch->patch_id = mc_hdr->patch_id; + patch->equiv_cpu = proc_id; + + pr_debug("%s: Added patch_id: 0x%08x, proc_id: 0x%04x\n", + __func__, patch->patch_id, proc_id); + + /* ... and add to cache. */ + update_cache(patch); + + return crnt_size; +} + +static enum ucode_state __load_microcode_amd(u8 family, const u8 *data, + size_t size) +{ + enum ucode_state ret = UCODE_ERROR; + unsigned int leftover; + u8 *fw = (u8 *)data; + int crnt_size = 0; + int offset; + + offset = install_equiv_cpu_table(data); + if (offset < 0) { + pr_err("failed to create equivalent cpu table\n"); + return ret; + } + fw += offset; + leftover = size - offset; + + if (*(u32 *)fw != UCODE_UCODE_TYPE) { + pr_err("invalid type field in container file section header\n"); + free_equiv_cpu_table(); + return ret; + } + + while (leftover) { + crnt_size = verify_and_add_patch(family, fw, leftover); + if (crnt_size < 0) + return ret; + + fw += crnt_size; + leftover -= crnt_size; + } + + return UCODE_OK; +} + +enum ucode_state load_microcode_amd(u8 family, const u8 *data, size_t size) +{ + enum ucode_state ret; + + /* free old equiv table */ + free_equiv_cpu_table(); + + ret = __load_microcode_amd(family, data, size); + + if (ret != UCODE_OK) + cleanup(); + +#if defined(CONFIG_MICROCODE_AMD_EARLY) && defined(CONFIG_X86_32) + /* save BSP's matching patch for early load */ + if (cpu_data(smp_processor_id()).cpu_index == boot_cpu_data.cpu_index) { + struct ucode_patch *p = find_patch(smp_processor_id()); + if (p) { + memset(amd_ucode_patch, 0, PATCH_MAX_SIZE); + memcpy(amd_ucode_patch, p->data, min_t(u32, ksize(p->data), + PATCH_MAX_SIZE)); + } + } +#endif + return ret; +} + +/* + * AMD microcode firmware naming convention, up to family 15h they are in + * the legacy file: + * + * amd-ucode/microcode_amd.bin + * + * This legacy file is always smaller than 2K in size. + * + * Beginning with family 15h, they are in family-specific firmware files: + * + * amd-ucode/microcode_amd_fam15h.bin + * amd-ucode/microcode_amd_fam16h.bin + * ... + * + * These might be larger than 2K. + */ +static enum ucode_state request_microcode_amd(int cpu, struct device *device, + bool refresh_fw) +{ + char fw_name[36] = "amd-ucode/microcode_amd.bin"; + struct cpuinfo_x86 *c = &cpu_data(cpu); + enum ucode_state ret = UCODE_NFOUND; + const struct firmware *fw; + + /* reload ucode container only on the boot cpu */ + if (!refresh_fw || c->cpu_index != boot_cpu_data.cpu_index) + return UCODE_OK; + + if (c->x86 >= 0x15) + snprintf(fw_name, sizeof(fw_name), "amd-ucode/microcode_amd_fam%.2xh.bin", c->x86); + + if (request_firmware_direct(&fw, (const char *)fw_name, device)) { + pr_debug("failed to load file %s\n", fw_name); + goto out; + } + + ret = UCODE_ERROR; + if (*(u32 *)fw->data != UCODE_MAGIC) { + pr_err("invalid magic value (0x%08x)\n", *(u32 *)fw->data); + goto fw_release; + } + + ret = load_microcode_amd(c->x86, fw->data, fw->size); + + fw_release: + release_firmware(fw); + + out: + return ret; +} + +static enum ucode_state +request_microcode_user(int cpu, const void __user *buf, size_t size) +{ + return UCODE_ERROR; +} + +static void microcode_fini_cpu_amd(int cpu) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + + uci->mc = NULL; +} + +static struct microcode_ops microcode_amd_ops = { + .request_microcode_user = request_microcode_user, + .request_microcode_fw = request_microcode_amd, + .collect_cpu_info = collect_cpu_info_amd, + .apply_microcode = apply_microcode_amd, + .microcode_fini_cpu = microcode_fini_cpu_amd, +}; + +struct microcode_ops * __init init_amd_microcode(void) +{ + struct cpuinfo_x86 *c = &cpu_data(0); + + if (c->x86_vendor != X86_VENDOR_AMD || c->x86 < 0x10) { + pr_warning("AMD CPU family 0x%x not supported\n", c->x86); + return NULL; + } + + return µcode_amd_ops; +} + +void __exit exit_amd_microcode(void) +{ + cleanup(); +} diff --git a/arch/x86/kernel/cpu/microcode/amd_early.c b/arch/x86/kernel/cpu/microcode/amd_early.c new file mode 100644 index 00000000000..617a9e28424 --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/amd_early.c @@ -0,0 +1,395 @@ +/* + * Copyright (C) 2013 Advanced Micro Devices, Inc. + * + * Author: Jacob Shin <jacob.shin@amd.com> + * Fixes: Borislav Petkov <bp@suse.de> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/earlycpio.h> +#include <linux/initrd.h> + +#include <asm/cpu.h> +#include <asm/setup.h> +#include <asm/microcode_amd.h> + +/* + * This points to the current valid container of microcode patches which we will + * save from the initrd before jettisoning its contents. + */ +static u8 *container; +static size_t container_size; + +static u32 ucode_new_rev; +u8 amd_ucode_patch[PATCH_MAX_SIZE]; +static u16 this_equiv_id; + +struct cpio_data ucode_cpio; + +/* + * Microcode patch container file is prepended to the initrd in cpio format. + * See Documentation/x86/early-microcode.txt + */ +static __initdata char ucode_path[] = "kernel/x86/microcode/AuthenticAMD.bin"; + +static struct cpio_data __init find_ucode_in_initrd(void) +{ + long offset = 0; + char *path; + void *start; + size_t size; + +#ifdef CONFIG_X86_32 + struct boot_params *p; + + /* + * On 32-bit, early load occurs before paging is turned on so we need + * to use physical addresses. + */ + p = (struct boot_params *)__pa_nodebug(&boot_params); + path = (char *)__pa_nodebug(ucode_path); + start = (void *)p->hdr.ramdisk_image; + size = p->hdr.ramdisk_size; +#else + path = ucode_path; + start = (void *)(boot_params.hdr.ramdisk_image + PAGE_OFFSET); + size = boot_params.hdr.ramdisk_size; +#endif + + return find_cpio_data(path, start, size, &offset); +} + +static size_t compute_container_size(u8 *data, u32 total_size) +{ + size_t size = 0; + u32 *header = (u32 *)data; + + if (header[0] != UCODE_MAGIC || + header[1] != UCODE_EQUIV_CPU_TABLE_TYPE || /* type */ + header[2] == 0) /* size */ + return size; + + size = header[2] + CONTAINER_HDR_SZ; + total_size -= size; + data += size; + + while (total_size) { + u16 patch_size; + + header = (u32 *)data; + + if (header[0] != UCODE_UCODE_TYPE) + break; + + /* + * Sanity-check patch size. + */ + patch_size = header[1]; + if (patch_size > PATCH_MAX_SIZE) + break; + + size += patch_size + SECTION_HDR_SIZE; + data += patch_size + SECTION_HDR_SIZE; + total_size -= patch_size + SECTION_HDR_SIZE; + } + + return size; +} + +/* + * Early load occurs before we can vmalloc(). So we look for the microcode + * patch container file in initrd, traverse equivalent cpu table, look for a + * matching microcode patch, and update, all in initrd memory in place. + * When vmalloc() is available for use later -- on 64-bit during first AP load, + * and on 32-bit during save_microcode_in_initrd_amd() -- we can call + * load_microcode_amd() to save equivalent cpu table and microcode patches in + * kernel heap memory. + */ +static void apply_ucode_in_initrd(void *ucode, size_t size) +{ + struct equiv_cpu_entry *eq; + size_t *cont_sz; + u32 *header; + u8 *data, **cont; + u16 eq_id = 0; + int offset, left; + u32 rev, eax, ebx, ecx, edx; + u32 *new_rev; + +#ifdef CONFIG_X86_32 + new_rev = (u32 *)__pa_nodebug(&ucode_new_rev); + cont_sz = (size_t *)__pa_nodebug(&container_size); + cont = (u8 **)__pa_nodebug(&container); +#else + new_rev = &ucode_new_rev; + cont_sz = &container_size; + cont = &container; +#endif + + data = ucode; + left = size; + header = (u32 *)data; + + /* find equiv cpu table */ + if (header[0] != UCODE_MAGIC || + header[1] != UCODE_EQUIV_CPU_TABLE_TYPE || /* type */ + header[2] == 0) /* size */ + return; + + eax = 0x00000001; + ecx = 0; + native_cpuid(&eax, &ebx, &ecx, &edx); + + while (left > 0) { + eq = (struct equiv_cpu_entry *)(data + CONTAINER_HDR_SZ); + + *cont = data; + + /* Advance past the container header */ + offset = header[2] + CONTAINER_HDR_SZ; + data += offset; + left -= offset; + + eq_id = find_equiv_id(eq, eax); + if (eq_id) { + this_equiv_id = eq_id; + *cont_sz = compute_container_size(*cont, left + offset); + + /* + * truncate how much we need to iterate over in the + * ucode update loop below + */ + left = *cont_sz - offset; + break; + } + + /* + * support multiple container files appended together. if this + * one does not have a matching equivalent cpu entry, we fast + * forward to the next container file. + */ + while (left > 0) { + header = (u32 *)data; + if (header[0] == UCODE_MAGIC && + header[1] == UCODE_EQUIV_CPU_TABLE_TYPE) + break; + + offset = header[1] + SECTION_HDR_SIZE; + data += offset; + left -= offset; + } + + /* mark where the next microcode container file starts */ + offset = data - (u8 *)ucode; + ucode = data; + } + + if (!eq_id) { + *cont = NULL; + *cont_sz = 0; + return; + } + + /* find ucode and update if needed */ + + native_rdmsr(MSR_AMD64_PATCH_LEVEL, rev, eax); + + while (left > 0) { + struct microcode_amd *mc; + + header = (u32 *)data; + if (header[0] != UCODE_UCODE_TYPE || /* type */ + header[1] == 0) /* size */ + break; + + mc = (struct microcode_amd *)(data + SECTION_HDR_SIZE); + + if (eq_id == mc->hdr.processor_rev_id && rev < mc->hdr.patch_id) { + + if (!__apply_microcode_amd(mc)) { + rev = mc->hdr.patch_id; + *new_rev = rev; + + /* save ucode patch */ + memcpy(amd_ucode_patch, mc, + min_t(u32, header[1], PATCH_MAX_SIZE)); + } + } + + offset = header[1] + SECTION_HDR_SIZE; + data += offset; + left -= offset; + } +} + +void __init load_ucode_amd_bsp(void) +{ + struct cpio_data cp; + void **data; + size_t *size; + +#ifdef CONFIG_X86_32 + data = (void **)__pa_nodebug(&ucode_cpio.data); + size = (size_t *)__pa_nodebug(&ucode_cpio.size); +#else + data = &ucode_cpio.data; + size = &ucode_cpio.size; +#endif + + cp = find_ucode_in_initrd(); + if (!cp.data) + return; + + *data = cp.data; + *size = cp.size; + + apply_ucode_in_initrd(cp.data, cp.size); +} + +#ifdef CONFIG_X86_32 +/* + * On 32-bit, since AP's early load occurs before paging is turned on, we + * cannot traverse cpu_equiv_table and pcache in kernel heap memory. So during + * cold boot, AP will apply_ucode_in_initrd() just like the BSP. During + * save_microcode_in_initrd_amd() BSP's patch is copied to amd_ucode_patch, + * which is used upon resume from suspend. + */ +void load_ucode_amd_ap(void) +{ + struct microcode_amd *mc; + size_t *usize; + void **ucode; + + mc = (struct microcode_amd *)__pa(amd_ucode_patch); + if (mc->hdr.patch_id && mc->hdr.processor_rev_id) { + __apply_microcode_amd(mc); + return; + } + + ucode = (void *)__pa_nodebug(&container); + usize = (size_t *)__pa_nodebug(&container_size); + + if (!*ucode || !*usize) + return; + + apply_ucode_in_initrd(*ucode, *usize); +} + +static void __init collect_cpu_sig_on_bsp(void *arg) +{ + unsigned int cpu = smp_processor_id(); + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + + uci->cpu_sig.sig = cpuid_eax(0x00000001); +} + +static void __init get_bsp_sig(void) +{ + unsigned int bsp = boot_cpu_data.cpu_index; + struct ucode_cpu_info *uci = ucode_cpu_info + bsp; + + if (!uci->cpu_sig.sig) + smp_call_function_single(bsp, collect_cpu_sig_on_bsp, NULL, 1); +} +#else +void load_ucode_amd_ap(void) +{ + unsigned int cpu = smp_processor_id(); + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + struct equiv_cpu_entry *eq; + struct microcode_amd *mc; + u32 rev, eax; + u16 eq_id; + + /* Exit if called on the BSP. */ + if (!cpu) + return; + + if (!container) + return; + + rdmsr(MSR_AMD64_PATCH_LEVEL, rev, eax); + + uci->cpu_sig.rev = rev; + uci->cpu_sig.sig = eax; + + eax = cpuid_eax(0x00000001); + eq = (struct equiv_cpu_entry *)(container + CONTAINER_HDR_SZ); + + eq_id = find_equiv_id(eq, eax); + if (!eq_id) + return; + + if (eq_id == this_equiv_id) { + mc = (struct microcode_amd *)amd_ucode_patch; + + if (mc && rev < mc->hdr.patch_id) { + if (!__apply_microcode_amd(mc)) + ucode_new_rev = mc->hdr.patch_id; + } + + } else { + if (!ucode_cpio.data) + return; + + /* + * AP has a different equivalence ID than BSP, looks like + * mixed-steppings silicon so go through the ucode blob anew. + */ + apply_ucode_in_initrd(ucode_cpio.data, ucode_cpio.size); + } +} +#endif + +int __init save_microcode_in_initrd_amd(void) +{ + unsigned long cont; + enum ucode_state ret; + u32 eax; + + if (!container) + return -EINVAL; + +#ifdef CONFIG_X86_32 + get_bsp_sig(); + cont = (unsigned long)container; +#else + /* + * We need the physical address of the container for both bitness since + * boot_params.hdr.ramdisk_image is a physical address. + */ + cont = __pa(container); +#endif + + /* + * Take into account the fact that the ramdisk might get relocated and + * therefore we need to recompute the container's position in virtual + * memory space. + */ + if (relocated_ramdisk) + container = (u8 *)(__va(relocated_ramdisk) + + (cont - boot_params.hdr.ramdisk_image)); + + if (ucode_new_rev) + pr_info("microcode: updated early to new patch_level=0x%08x\n", + ucode_new_rev); + + eax = cpuid_eax(0x00000001); + eax = ((eax >> 8) & 0xf) + ((eax >> 20) & 0xff); + + ret = load_microcode_amd(eax, container, container_size); + if (ret != UCODE_OK) + return -EINVAL; + + /* + * This will be freed any msec now, stash patches for the current + * family and switch to patch cache for cpu hotplug, etc later. + */ + container = NULL; + container_size = 0; + + return 0; +} diff --git a/arch/x86/kernel/cpu/microcode/core.c b/arch/x86/kernel/cpu/microcode/core.c new file mode 100644 index 00000000000..dd9d6190b08 --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/core.c @@ -0,0 +1,651 @@ +/* + * Intel CPU Microcode Update Driver for Linux + * + * Copyright (C) 2000-2006 Tigran Aivazian <tigran@aivazian.fsnet.co.uk> + * 2006 Shaohua Li <shaohua.li@intel.com> + * + * This driver allows to upgrade microcode on Intel processors + * belonging to IA-32 family - PentiumPro, Pentium II, + * Pentium III, Xeon, Pentium 4, etc. + * + * Reference: Section 8.11 of Volume 3a, IA-32 Intel? Architecture + * Software Developer's Manual + * Order Number 253668 or free download from: + * + * http://developer.intel.com/Assets/PDF/manual/253668.pdf + * + * For more information, go to http://www.urbanmyth.org/microcode + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + * + * 1.0 16 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Initial release. + * 1.01 18 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Added read() support + cleanups. + * 1.02 21 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Added 'device trimming' support. open(O_WRONLY) zeroes + * and frees the saved copy of applied microcode. + * 1.03 29 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Made to use devfs (/dev/cpu/microcode) + cleanups. + * 1.04 06 Jun 2000, Simon Trimmer <simon@veritas.com> + * Added misc device support (now uses both devfs and misc). + * Added MICROCODE_IOCFREE ioctl to clear memory. + * 1.05 09 Jun 2000, Simon Trimmer <simon@veritas.com> + * Messages for error cases (non Intel & no suitable microcode). + * 1.06 03 Aug 2000, Tigran Aivazian <tigran@veritas.com> + * Removed ->release(). Removed exclusive open and status bitmap. + * Added microcode_rwsem to serialize read()/write()/ioctl(). + * Removed global kernel lock usage. + * 1.07 07 Sep 2000, Tigran Aivazian <tigran@veritas.com> + * Write 0 to 0x8B msr and then cpuid before reading revision, + * so that it works even if there were no update done by the + * BIOS. Otherwise, reading from 0x8B gives junk (which happened + * to be 0 on my machine which is why it worked even when I + * disabled update by the BIOS) + * Thanks to Eric W. Biederman <ebiederman@lnxi.com> for the fix. + * 1.08 11 Dec 2000, Richard Schaal <richard.schaal@intel.com> and + * Tigran Aivazian <tigran@veritas.com> + * Intel Pentium 4 processor support and bugfixes. + * 1.09 30 Oct 2001, Tigran Aivazian <tigran@veritas.com> + * Bugfix for HT (Hyper-Threading) enabled processors + * whereby processor resources are shared by all logical processors + * in a single CPU package. + * 1.10 28 Feb 2002 Asit K Mallick <asit.k.mallick@intel.com> and + * Tigran Aivazian <tigran@veritas.com>, + * Serialize updates as required on HT processors due to + * speculative nature of implementation. + * 1.11 22 Mar 2002 Tigran Aivazian <tigran@veritas.com> + * Fix the panic when writing zero-length microcode chunk. + * 1.12 29 Sep 2003 Nitin Kamble <nitin.a.kamble@intel.com>, + * Jun Nakajima <jun.nakajima@intel.com> + * Support for the microcode updates in the new format. + * 1.13 10 Oct 2003 Tigran Aivazian <tigran@veritas.com> + * Removed ->read() method and obsoleted MICROCODE_IOCFREE ioctl + * because we no longer hold a copy of applied microcode + * in kernel memory. + * 1.14 25 Jun 2004 Tigran Aivazian <tigran@veritas.com> + * Fix sigmatch() macro to handle old CPUs with pf == 0. + * Thanks to Stuart Swales for pointing out this bug. + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include <linux/platform_device.h> +#include <linux/miscdevice.h> +#include <linux/capability.h> +#include <linux/kernel.h> +#include <linux/module.h> +#include <linux/mutex.h> +#include <linux/cpu.h> +#include <linux/fs.h> +#include <linux/mm.h> +#include <linux/syscore_ops.h> + +#include <asm/microcode.h> +#include <asm/processor.h> +#include <asm/cpu_device_id.h> +#include <asm/perf_event.h> + +MODULE_DESCRIPTION("Microcode Update Driver"); +MODULE_AUTHOR("Tigran Aivazian <tigran@aivazian.fsnet.co.uk>"); +MODULE_LICENSE("GPL"); + +#define MICROCODE_VERSION "2.00" + +static struct microcode_ops *microcode_ops; + +bool dis_ucode_ldr; +module_param(dis_ucode_ldr, bool, 0); + +/* + * Synchronization. + * + * All non cpu-hotplug-callback call sites use: + * + * - microcode_mutex to synchronize with each other; + * - get/put_online_cpus() to synchronize with + * the cpu-hotplug-callback call sites. + * + * We guarantee that only a single cpu is being + * updated at any particular moment of time. + */ +static DEFINE_MUTEX(microcode_mutex); + +struct ucode_cpu_info ucode_cpu_info[NR_CPUS]; +EXPORT_SYMBOL_GPL(ucode_cpu_info); + +/* + * Operations that are run on a target cpu: + */ + +struct cpu_info_ctx { + struct cpu_signature *cpu_sig; + int err; +}; + +static void collect_cpu_info_local(void *arg) +{ + struct cpu_info_ctx *ctx = arg; + + ctx->err = microcode_ops->collect_cpu_info(smp_processor_id(), + ctx->cpu_sig); +} + +static int collect_cpu_info_on_target(int cpu, struct cpu_signature *cpu_sig) +{ + struct cpu_info_ctx ctx = { .cpu_sig = cpu_sig, .err = 0 }; + int ret; + + ret = smp_call_function_single(cpu, collect_cpu_info_local, &ctx, 1); + if (!ret) + ret = ctx.err; + + return ret; +} + +static int collect_cpu_info(int cpu) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + int ret; + + memset(uci, 0, sizeof(*uci)); + + ret = collect_cpu_info_on_target(cpu, &uci->cpu_sig); + if (!ret) + uci->valid = 1; + + return ret; +} + +struct apply_microcode_ctx { + int err; +}; + +static void apply_microcode_local(void *arg) +{ + struct apply_microcode_ctx *ctx = arg; + + ctx->err = microcode_ops->apply_microcode(smp_processor_id()); +} + +static int apply_microcode_on_target(int cpu) +{ + struct apply_microcode_ctx ctx = { .err = 0 }; + int ret; + + ret = smp_call_function_single(cpu, apply_microcode_local, &ctx, 1); + if (!ret) + ret = ctx.err; + + return ret; +} + +#ifdef CONFIG_MICROCODE_OLD_INTERFACE +static int do_microcode_update(const void __user *buf, size_t size) +{ + int error = 0; + int cpu; + + for_each_online_cpu(cpu) { + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + enum ucode_state ustate; + + if (!uci->valid) + continue; + + ustate = microcode_ops->request_microcode_user(cpu, buf, size); + if (ustate == UCODE_ERROR) { + error = -1; + break; + } else if (ustate == UCODE_OK) + apply_microcode_on_target(cpu); + } + + return error; +} + +static int microcode_open(struct inode *inode, struct file *file) +{ + return capable(CAP_SYS_RAWIO) ? nonseekable_open(inode, file) : -EPERM; +} + +static ssize_t microcode_write(struct file *file, const char __user *buf, + size_t len, loff_t *ppos) +{ + ssize_t ret = -EINVAL; + + if ((len >> PAGE_SHIFT) > totalram_pages) { + pr_err("too much data (max %ld pages)\n", totalram_pages); + return ret; + } + + get_online_cpus(); + mutex_lock(µcode_mutex); + + if (do_microcode_update(buf, len) == 0) + ret = (ssize_t)len; + + if (ret > 0) + perf_check_microcode(); + + mutex_unlock(µcode_mutex); + put_online_cpus(); + + return ret; +} + +static const struct file_operations microcode_fops = { + .owner = THIS_MODULE, + .write = microcode_write, + .open = microcode_open, + .llseek = no_llseek, +}; + +static struct miscdevice microcode_dev = { + .minor = MICROCODE_MINOR, + .name = "microcode", + .nodename = "cpu/microcode", + .fops = µcode_fops, +}; + +static int __init microcode_dev_init(void) +{ + int error; + + error = misc_register(µcode_dev); + if (error) { + pr_err("can't misc_register on minor=%d\n", MICROCODE_MINOR); + return error; + } + + return 0; +} + +static void __exit microcode_dev_exit(void) +{ + misc_deregister(µcode_dev); +} + +MODULE_ALIAS_MISCDEV(MICROCODE_MINOR); +MODULE_ALIAS("devname:cpu/microcode"); +#else +#define microcode_dev_init() 0 +#define microcode_dev_exit() do { } while (0) +#endif + +/* fake device for request_firmware */ +static struct platform_device *microcode_pdev; + +static int reload_for_cpu(int cpu) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + enum ucode_state ustate; + int err = 0; + + if (!uci->valid) + return err; + + ustate = microcode_ops->request_microcode_fw(cpu, µcode_pdev->dev, true); + if (ustate == UCODE_OK) + apply_microcode_on_target(cpu); + else + if (ustate == UCODE_ERROR) + err = -EINVAL; + return err; +} + +static ssize_t reload_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t size) +{ + unsigned long val; + int cpu; + ssize_t ret = 0, tmp_ret; + + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; + + if (val != 1) + return size; + + get_online_cpus(); + mutex_lock(µcode_mutex); + for_each_online_cpu(cpu) { + tmp_ret = reload_for_cpu(cpu); + if (tmp_ret != 0) + pr_warn("Error reloading microcode on CPU %d\n", cpu); + + /* save retval of the first encountered reload error */ + if (!ret) + ret = tmp_ret; + } + if (!ret) + perf_check_microcode(); + mutex_unlock(µcode_mutex); + put_online_cpus(); + + if (!ret) + ret = size; + + return ret; +} + +static ssize_t version_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + dev->id; + + return sprintf(buf, "0x%x\n", uci->cpu_sig.rev); +} + +static ssize_t pf_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + dev->id; + + return sprintf(buf, "0x%x\n", uci->cpu_sig.pf); +} + +static DEVICE_ATTR(reload, 0200, NULL, reload_store); +static DEVICE_ATTR(version, 0400, version_show, NULL); +static DEVICE_ATTR(processor_flags, 0400, pf_show, NULL); + +static struct attribute *mc_default_attrs[] = { + &dev_attr_version.attr, + &dev_attr_processor_flags.attr, + NULL +}; + +static struct attribute_group mc_attr_group = { + .attrs = mc_default_attrs, + .name = "microcode", +}; + +static void microcode_fini_cpu(int cpu) +{ + microcode_ops->microcode_fini_cpu(cpu); +} + +static enum ucode_state microcode_resume_cpu(int cpu) +{ + pr_debug("CPU%d updated upon resume\n", cpu); + + if (apply_microcode_on_target(cpu)) + return UCODE_ERROR; + + return UCODE_OK; +} + +static enum ucode_state microcode_init_cpu(int cpu, bool refresh_fw) +{ + enum ucode_state ustate; + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + + if (uci && uci->valid) + return UCODE_OK; + + if (collect_cpu_info(cpu)) + return UCODE_ERROR; + + /* --dimm. Trigger a delayed update? */ + if (system_state != SYSTEM_RUNNING) + return UCODE_NFOUND; + + ustate = microcode_ops->request_microcode_fw(cpu, µcode_pdev->dev, + refresh_fw); + + if (ustate == UCODE_OK) { + pr_debug("CPU%d updated upon init\n", cpu); + apply_microcode_on_target(cpu); + } + + return ustate; +} + +static enum ucode_state microcode_update_cpu(int cpu) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + + if (uci->valid) + return microcode_resume_cpu(cpu); + + return microcode_init_cpu(cpu, false); +} + +static int mc_device_add(struct device *dev, struct subsys_interface *sif) +{ + int err, cpu = dev->id; + + if (!cpu_online(cpu)) + return 0; + + pr_debug("CPU%d added\n", cpu); + + err = sysfs_create_group(&dev->kobj, &mc_attr_group); + if (err) + return err; + + if (microcode_init_cpu(cpu, true) == UCODE_ERROR) + return -EINVAL; + + return err; +} + +static int mc_device_remove(struct device *dev, struct subsys_interface *sif) +{ + int cpu = dev->id; + + if (!cpu_online(cpu)) + return 0; + + pr_debug("CPU%d removed\n", cpu); + microcode_fini_cpu(cpu); + sysfs_remove_group(&dev->kobj, &mc_attr_group); + return 0; +} + +static struct subsys_interface mc_cpu_interface = { + .name = "microcode", + .subsys = &cpu_subsys, + .add_dev = mc_device_add, + .remove_dev = mc_device_remove, +}; + +/** + * mc_bp_resume - Update boot CPU microcode during resume. + */ +static void mc_bp_resume(void) +{ + int cpu = smp_processor_id(); + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + + if (uci->valid && uci->mc) + microcode_ops->apply_microcode(cpu); +} + +static struct syscore_ops mc_syscore_ops = { + .resume = mc_bp_resume, +}; + +static int +mc_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu) +{ + unsigned int cpu = (unsigned long)hcpu; + struct device *dev; + + dev = get_cpu_device(cpu); + + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_ONLINE: + microcode_update_cpu(cpu); + pr_debug("CPU%d added\n", cpu); + /* + * "break" is missing on purpose here because we want to fall + * through in order to create the sysfs group. + */ + + case CPU_DOWN_FAILED: + if (sysfs_create_group(&dev->kobj, &mc_attr_group)) + pr_err("Failed to create group for CPU%d\n", cpu); + break; + + case CPU_DOWN_PREPARE: + /* Suspend is in progress, only remove the interface */ + sysfs_remove_group(&dev->kobj, &mc_attr_group); + pr_debug("CPU%d removed\n", cpu); + break; + + /* + * case CPU_DEAD: + * + * When a CPU goes offline, don't free up or invalidate the copy of + * the microcode in kernel memory, so that we can reuse it when the + * CPU comes back online without unnecessarily requesting the userspace + * for it again. + */ + } + + /* The CPU refused to come up during a system resume */ + if (action == CPU_UP_CANCELED_FROZEN) + microcode_fini_cpu(cpu); + + return NOTIFY_OK; +} + +static struct notifier_block __refdata mc_cpu_notifier = { + .notifier_call = mc_cpu_callback, +}; + +#ifdef MODULE +/* Autoload on Intel and AMD systems */ +static const struct x86_cpu_id __initconst microcode_id[] = { +#ifdef CONFIG_MICROCODE_INTEL + { X86_VENDOR_INTEL, X86_FAMILY_ANY, X86_MODEL_ANY, }, +#endif +#ifdef CONFIG_MICROCODE_AMD + { X86_VENDOR_AMD, X86_FAMILY_ANY, X86_MODEL_ANY, }, +#endif + {} +}; +MODULE_DEVICE_TABLE(x86cpu, microcode_id); +#endif + +static struct attribute *cpu_root_microcode_attrs[] = { + &dev_attr_reload.attr, + NULL +}; + +static struct attribute_group cpu_root_microcode_group = { + .name = "microcode", + .attrs = cpu_root_microcode_attrs, +}; + +static int __init microcode_init(void) +{ + struct cpuinfo_x86 *c = &cpu_data(0); + int error; + + if (dis_ucode_ldr) + return 0; + + if (c->x86_vendor == X86_VENDOR_INTEL) + microcode_ops = init_intel_microcode(); + else if (c->x86_vendor == X86_VENDOR_AMD) + microcode_ops = init_amd_microcode(); + else + pr_err("no support for this CPU vendor\n"); + + if (!microcode_ops) + return -ENODEV; + + microcode_pdev = platform_device_register_simple("microcode", -1, + NULL, 0); + if (IS_ERR(microcode_pdev)) + return PTR_ERR(microcode_pdev); + + get_online_cpus(); + mutex_lock(µcode_mutex); + + error = subsys_interface_register(&mc_cpu_interface); + if (!error) + perf_check_microcode(); + mutex_unlock(µcode_mutex); + put_online_cpus(); + + if (error) + goto out_pdev; + + error = sysfs_create_group(&cpu_subsys.dev_root->kobj, + &cpu_root_microcode_group); + + if (error) { + pr_err("Error creating microcode group!\n"); + goto out_driver; + } + + error = microcode_dev_init(); + if (error) + goto out_ucode_group; + + register_syscore_ops(&mc_syscore_ops); + register_hotcpu_notifier(&mc_cpu_notifier); + + pr_info("Microcode Update Driver: v" MICROCODE_VERSION + " <tigran@aivazian.fsnet.co.uk>, Peter Oruba\n"); + + return 0; + + out_ucode_group: + sysfs_remove_group(&cpu_subsys.dev_root->kobj, + &cpu_root_microcode_group); + + out_driver: + get_online_cpus(); + mutex_lock(µcode_mutex); + + subsys_interface_unregister(&mc_cpu_interface); + + mutex_unlock(µcode_mutex); + put_online_cpus(); + + out_pdev: + platform_device_unregister(microcode_pdev); + return error; + +} +module_init(microcode_init); + +static void __exit microcode_exit(void) +{ + struct cpuinfo_x86 *c = &cpu_data(0); + + microcode_dev_exit(); + + unregister_hotcpu_notifier(&mc_cpu_notifier); + unregister_syscore_ops(&mc_syscore_ops); + + sysfs_remove_group(&cpu_subsys.dev_root->kobj, + &cpu_root_microcode_group); + + get_online_cpus(); + mutex_lock(µcode_mutex); + + subsys_interface_unregister(&mc_cpu_interface); + + mutex_unlock(µcode_mutex); + put_online_cpus(); + + platform_device_unregister(microcode_pdev); + + microcode_ops = NULL; + + if (c->x86_vendor == X86_VENDOR_AMD) + exit_amd_microcode(); + + pr_info("Microcode Update Driver: v" MICROCODE_VERSION " removed.\n"); +} +module_exit(microcode_exit); diff --git a/arch/x86/kernel/cpu/microcode/core_early.c b/arch/x86/kernel/cpu/microcode/core_early.c new file mode 100644 index 00000000000..5f28a64e71e --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/core_early.c @@ -0,0 +1,178 @@ +/* + * X86 CPU microcode early update for Linux + * + * Copyright (C) 2012 Fenghua Yu <fenghua.yu@intel.com> + * H Peter Anvin" <hpa@zytor.com> + * + * This driver allows to early upgrade microcode on Intel processors + * belonging to IA-32 family - PentiumPro, Pentium II, + * Pentium III, Xeon, Pentium 4, etc. + * + * Reference: Section 9.11 of Volume 3, IA-32 Intel Architecture + * Software Developer's Manual. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include <linux/module.h> +#include <asm/microcode.h> +#include <asm/microcode_intel.h> +#include <asm/microcode_amd.h> +#include <asm/processor.h> +#include <asm/cmdline.h> + +#define QCHAR(a, b, c, d) ((a) + ((b) << 8) + ((c) << 16) + ((d) << 24)) +#define CPUID_INTEL1 QCHAR('G', 'e', 'n', 'u') +#define CPUID_INTEL2 QCHAR('i', 'n', 'e', 'I') +#define CPUID_INTEL3 QCHAR('n', 't', 'e', 'l') +#define CPUID_AMD1 QCHAR('A', 'u', 't', 'h') +#define CPUID_AMD2 QCHAR('e', 'n', 't', 'i') +#define CPUID_AMD3 QCHAR('c', 'A', 'M', 'D') + +#define CPUID_IS(a, b, c, ebx, ecx, edx) \ + (!((ebx ^ (a))|(edx ^ (b))|(ecx ^ (c)))) + +/* + * In early loading microcode phase on BSP, boot_cpu_data is not set up yet. + * x86_vendor() gets vendor id for BSP. + * + * In 32 bit AP case, accessing boot_cpu_data needs linear address. To simplify + * coding, we still use x86_vendor() to get vendor id for AP. + * + * x86_vendor() gets vendor information directly through cpuid. + */ +static int x86_vendor(void) +{ + u32 eax = 0x00000000; + u32 ebx, ecx = 0, edx; + + native_cpuid(&eax, &ebx, &ecx, &edx); + + if (CPUID_IS(CPUID_INTEL1, CPUID_INTEL2, CPUID_INTEL3, ebx, ecx, edx)) + return X86_VENDOR_INTEL; + + if (CPUID_IS(CPUID_AMD1, CPUID_AMD2, CPUID_AMD3, ebx, ecx, edx)) + return X86_VENDOR_AMD; + + return X86_VENDOR_UNKNOWN; +} + +static int x86_family(void) +{ + u32 eax = 0x00000001; + u32 ebx, ecx = 0, edx; + int x86; + + native_cpuid(&eax, &ebx, &ecx, &edx); + + x86 = (eax >> 8) & 0xf; + if (x86 == 15) + x86 += (eax >> 20) & 0xff; + + return x86; +} + +static bool __init check_loader_disabled_bsp(void) +{ +#ifdef CONFIG_X86_32 + const char *cmdline = (const char *)__pa_nodebug(boot_command_line); + const char *opt = "dis_ucode_ldr"; + const char *option = (const char *)__pa_nodebug(opt); + bool *res = (bool *)__pa_nodebug(&dis_ucode_ldr); + +#else /* CONFIG_X86_64 */ + const char *cmdline = boot_command_line; + const char *option = "dis_ucode_ldr"; + bool *res = &dis_ucode_ldr; +#endif + + if (cmdline_find_option_bool(cmdline, option)) + *res = true; + + return *res; +} + +void __init load_ucode_bsp(void) +{ + int vendor, x86; + + if (check_loader_disabled_bsp()) + return; + + if (!have_cpuid_p()) + return; + + vendor = x86_vendor(); + x86 = x86_family(); + + switch (vendor) { + case X86_VENDOR_INTEL: + if (x86 >= 6) + load_ucode_intel_bsp(); + break; + case X86_VENDOR_AMD: + if (x86 >= 0x10) + load_ucode_amd_bsp(); + break; + default: + break; + } +} + +static bool check_loader_disabled_ap(void) +{ +#ifdef CONFIG_X86_32 + return __pa_nodebug(dis_ucode_ldr); +#else + return dis_ucode_ldr; +#endif +} + +void load_ucode_ap(void) +{ + int vendor, x86; + + if (check_loader_disabled_ap()) + return; + + if (!have_cpuid_p()) + return; + + vendor = x86_vendor(); + x86 = x86_family(); + + switch (vendor) { + case X86_VENDOR_INTEL: + if (x86 >= 6) + load_ucode_intel_ap(); + break; + case X86_VENDOR_AMD: + if (x86 >= 0x10) + load_ucode_amd_ap(); + break; + default: + break; + } +} + +int __init save_microcode_in_initrd(void) +{ + struct cpuinfo_x86 *c = &boot_cpu_data; + + switch (c->x86_vendor) { + case X86_VENDOR_INTEL: + if (c->x86 >= 6) + save_microcode_in_initrd_intel(); + break; + case X86_VENDOR_AMD: + if (c->x86 >= 0x10) + save_microcode_in_initrd_amd(); + break; + default: + break; + } + + return 0; +} diff --git a/arch/x86/kernel/cpu/microcode/intel.c b/arch/x86/kernel/cpu/microcode/intel.c new file mode 100644 index 00000000000..a276fa75d9b --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/intel.c @@ -0,0 +1,333 @@ +/* + * Intel CPU Microcode Update Driver for Linux + * + * Copyright (C) 2000-2006 Tigran Aivazian <tigran@aivazian.fsnet.co.uk> + * 2006 Shaohua Li <shaohua.li@intel.com> + * + * This driver allows to upgrade microcode on Intel processors + * belonging to IA-32 family - PentiumPro, Pentium II, + * Pentium III, Xeon, Pentium 4, etc. + * + * Reference: Section 8.11 of Volume 3a, IA-32 Intel? Architecture + * Software Developer's Manual + * Order Number 253668 or free download from: + * + * http://developer.intel.com/Assets/PDF/manual/253668.pdf + * + * For more information, go to http://www.urbanmyth.org/microcode + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + * + * 1.0 16 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Initial release. + * 1.01 18 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Added read() support + cleanups. + * 1.02 21 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Added 'device trimming' support. open(O_WRONLY) zeroes + * and frees the saved copy of applied microcode. + * 1.03 29 Feb 2000, Tigran Aivazian <tigran@sco.com> + * Made to use devfs (/dev/cpu/microcode) + cleanups. + * 1.04 06 Jun 2000, Simon Trimmer <simon@veritas.com> + * Added misc device support (now uses both devfs and misc). + * Added MICROCODE_IOCFREE ioctl to clear memory. + * 1.05 09 Jun 2000, Simon Trimmer <simon@veritas.com> + * Messages for error cases (non Intel & no suitable microcode). + * 1.06 03 Aug 2000, Tigran Aivazian <tigran@veritas.com> + * Removed ->release(). Removed exclusive open and status bitmap. + * Added microcode_rwsem to serialize read()/write()/ioctl(). + * Removed global kernel lock usage. + * 1.07 07 Sep 2000, Tigran Aivazian <tigran@veritas.com> + * Write 0 to 0x8B msr and then cpuid before reading revision, + * so that it works even if there were no update done by the + * BIOS. Otherwise, reading from 0x8B gives junk (which happened + * to be 0 on my machine which is why it worked even when I + * disabled update by the BIOS) + * Thanks to Eric W. Biederman <ebiederman@lnxi.com> for the fix. + * 1.08 11 Dec 2000, Richard Schaal <richard.schaal@intel.com> and + * Tigran Aivazian <tigran@veritas.com> + * Intel Pentium 4 processor support and bugfixes. + * 1.09 30 Oct 2001, Tigran Aivazian <tigran@veritas.com> + * Bugfix for HT (Hyper-Threading) enabled processors + * whereby processor resources are shared by all logical processors + * in a single CPU package. + * 1.10 28 Feb 2002 Asit K Mallick <asit.k.mallick@intel.com> and + * Tigran Aivazian <tigran@veritas.com>, + * Serialize updates as required on HT processors due to + * speculative nature of implementation. + * 1.11 22 Mar 2002 Tigran Aivazian <tigran@veritas.com> + * Fix the panic when writing zero-length microcode chunk. + * 1.12 29 Sep 2003 Nitin Kamble <nitin.a.kamble@intel.com>, + * Jun Nakajima <jun.nakajima@intel.com> + * Support for the microcode updates in the new format. + * 1.13 10 Oct 2003 Tigran Aivazian <tigran@veritas.com> + * Removed ->read() method and obsoleted MICROCODE_IOCFREE ioctl + * because we no longer hold a copy of applied microcode + * in kernel memory. + * 1.14 25 Jun 2004 Tigran Aivazian <tigran@veritas.com> + * Fix sigmatch() macro to handle old CPUs with pf == 0. + * Thanks to Stuart Swales for pointing out this bug. + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include <linux/firmware.h> +#include <linux/uaccess.h> +#include <linux/kernel.h> +#include <linux/module.h> +#include <linux/vmalloc.h> + +#include <asm/microcode_intel.h> +#include <asm/processor.h> +#include <asm/msr.h> + +MODULE_DESCRIPTION("Microcode Update Driver"); +MODULE_AUTHOR("Tigran Aivazian <tigran@aivazian.fsnet.co.uk>"); +MODULE_LICENSE("GPL"); + +static int collect_cpu_info(int cpu_num, struct cpu_signature *csig) +{ + struct cpuinfo_x86 *c = &cpu_data(cpu_num); + unsigned int val[2]; + + memset(csig, 0, sizeof(*csig)); + + csig->sig = cpuid_eax(0x00000001); + + if ((c->x86_model >= 5) || (c->x86 > 6)) { + /* get processor flags from MSR 0x17 */ + rdmsr(MSR_IA32_PLATFORM_ID, val[0], val[1]); + csig->pf = 1 << ((val[1] >> 18) & 7); + } + + csig->rev = c->microcode; + pr_info("CPU%d sig=0x%x, pf=0x%x, revision=0x%x\n", + cpu_num, csig->sig, csig->pf, csig->rev); + + return 0; +} + +/* + * return 0 - no update found + * return 1 - found update + */ +static int get_matching_mc(struct microcode_intel *mc_intel, int cpu) +{ + struct cpu_signature cpu_sig; + unsigned int csig, cpf, crev; + + collect_cpu_info(cpu, &cpu_sig); + + csig = cpu_sig.sig; + cpf = cpu_sig.pf; + crev = cpu_sig.rev; + + return get_matching_microcode(csig, cpf, mc_intel, crev); +} + +int apply_microcode(int cpu) +{ + struct microcode_intel *mc_intel; + struct ucode_cpu_info *uci; + unsigned int val[2]; + int cpu_num = raw_smp_processor_id(); + struct cpuinfo_x86 *c = &cpu_data(cpu_num); + + uci = ucode_cpu_info + cpu; + mc_intel = uci->mc; + + /* We should bind the task to the CPU */ + BUG_ON(cpu_num != cpu); + + if (mc_intel == NULL) + return 0; + + /* + * Microcode on this CPU could be updated earlier. Only apply the + * microcode patch in mc_intel when it is newer than the one on this + * CPU. + */ + if (get_matching_mc(mc_intel, cpu) == 0) + return 0; + + /* write microcode via MSR 0x79 */ + wrmsr(MSR_IA32_UCODE_WRITE, + (unsigned long) mc_intel->bits, + (unsigned long) mc_intel->bits >> 16 >> 16); + wrmsr(MSR_IA32_UCODE_REV, 0, 0); + + /* As documented in the SDM: Do a CPUID 1 here */ + sync_core(); + + /* get the current revision from MSR 0x8B */ + rdmsr(MSR_IA32_UCODE_REV, val[0], val[1]); + + if (val[1] != mc_intel->hdr.rev) { + pr_err("CPU%d update to revision 0x%x failed\n", + cpu_num, mc_intel->hdr.rev); + return -1; + } + pr_info("CPU%d updated to revision 0x%x, date = %04x-%02x-%02x\n", + cpu_num, val[1], + mc_intel->hdr.date & 0xffff, + mc_intel->hdr.date >> 24, + (mc_intel->hdr.date >> 16) & 0xff); + + uci->cpu_sig.rev = val[1]; + c->microcode = val[1]; + + return 0; +} + +static enum ucode_state generic_load_microcode(int cpu, void *data, size_t size, + int (*get_ucode_data)(void *, const void *, size_t)) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + u8 *ucode_ptr = data, *new_mc = NULL, *mc = NULL; + int new_rev = uci->cpu_sig.rev; + unsigned int leftover = size; + enum ucode_state state = UCODE_OK; + unsigned int curr_mc_size = 0; + unsigned int csig, cpf; + + while (leftover) { + struct microcode_header_intel mc_header; + unsigned int mc_size; + + if (get_ucode_data(&mc_header, ucode_ptr, sizeof(mc_header))) + break; + + mc_size = get_totalsize(&mc_header); + if (!mc_size || mc_size > leftover) { + pr_err("error! Bad data in microcode data file\n"); + break; + } + + /* For performance reasons, reuse mc area when possible */ + if (!mc || mc_size > curr_mc_size) { + vfree(mc); + mc = vmalloc(mc_size); + if (!mc) + break; + curr_mc_size = mc_size; + } + + if (get_ucode_data(mc, ucode_ptr, mc_size) || + microcode_sanity_check(mc, 1) < 0) { + break; + } + + csig = uci->cpu_sig.sig; + cpf = uci->cpu_sig.pf; + if (get_matching_microcode(csig, cpf, mc, new_rev)) { + vfree(new_mc); + new_rev = mc_header.rev; + new_mc = mc; + mc = NULL; /* trigger new vmalloc */ + } + + ucode_ptr += mc_size; + leftover -= mc_size; + } + + vfree(mc); + + if (leftover) { + vfree(new_mc); + state = UCODE_ERROR; + goto out; + } + + if (!new_mc) { + state = UCODE_NFOUND; + goto out; + } + + vfree(uci->mc); + uci->mc = (struct microcode_intel *)new_mc; + + /* + * If early loading microcode is supported, save this mc into + * permanent memory. So it will be loaded early when a CPU is hot added + * or resumes. + */ + save_mc_for_early(new_mc); + + pr_debug("CPU%d found a matching microcode update with version 0x%x (current=0x%x)\n", + cpu, new_rev, uci->cpu_sig.rev); +out: + return state; +} + +static int get_ucode_fw(void *to, const void *from, size_t n) +{ + memcpy(to, from, n); + return 0; +} + +static enum ucode_state request_microcode_fw(int cpu, struct device *device, + bool refresh_fw) +{ + char name[30]; + struct cpuinfo_x86 *c = &cpu_data(cpu); + const struct firmware *firmware; + enum ucode_state ret; + + sprintf(name, "intel-ucode/%02x-%02x-%02x", + c->x86, c->x86_model, c->x86_mask); + + if (request_firmware_direct(&firmware, name, device)) { + pr_debug("data file %s load failed\n", name); + return UCODE_NFOUND; + } + + ret = generic_load_microcode(cpu, (void *)firmware->data, + firmware->size, &get_ucode_fw); + + release_firmware(firmware); + + return ret; +} + +static int get_ucode_user(void *to, const void *from, size_t n) +{ + return copy_from_user(to, from, n); +} + +static enum ucode_state +request_microcode_user(int cpu, const void __user *buf, size_t size) +{ + return generic_load_microcode(cpu, (void *)buf, size, &get_ucode_user); +} + +static void microcode_fini_cpu(int cpu) +{ + struct ucode_cpu_info *uci = ucode_cpu_info + cpu; + + vfree(uci->mc); + uci->mc = NULL; +} + +static struct microcode_ops microcode_intel_ops = { + .request_microcode_user = request_microcode_user, + .request_microcode_fw = request_microcode_fw, + .collect_cpu_info = collect_cpu_info, + .apply_microcode = apply_microcode, + .microcode_fini_cpu = microcode_fini_cpu, +}; + +struct microcode_ops * __init init_intel_microcode(void) +{ + struct cpuinfo_x86 *c = &cpu_data(0); + + if (c->x86_vendor != X86_VENDOR_INTEL || c->x86 < 6 || + cpu_has(c, X86_FEATURE_IA64)) { + pr_err("Intel CPU family 0x%x not supported\n", c->x86); + return NULL; + } + + return µcode_intel_ops; +} + diff --git a/arch/x86/kernel/cpu/microcode/intel_early.c b/arch/x86/kernel/cpu/microcode/intel_early.c new file mode 100644 index 00000000000..18f739129e7 --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/intel_early.c @@ -0,0 +1,787 @@ +/* + * Intel CPU microcode early update for Linux + * + * Copyright (C) 2012 Fenghua Yu <fenghua.yu@intel.com> + * H Peter Anvin" <hpa@zytor.com> + * + * This allows to early upgrade microcode on Intel processors + * belonging to IA-32 family - PentiumPro, Pentium II, + * Pentium III, Xeon, Pentium 4, etc. + * + * Reference: Section 9.11 of Volume 3, IA-32 Intel Architecture + * Software Developer's Manual. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include <linux/module.h> +#include <linux/mm.h> +#include <linux/slab.h> +#include <linux/earlycpio.h> +#include <linux/initrd.h> +#include <linux/cpu.h> +#include <asm/msr.h> +#include <asm/microcode_intel.h> +#include <asm/processor.h> +#include <asm/tlbflush.h> +#include <asm/setup.h> + +unsigned long mc_saved_in_initrd[MAX_UCODE_COUNT]; +struct mc_saved_data { + unsigned int mc_saved_count; + struct microcode_intel **mc_saved; +} mc_saved_data; + +static enum ucode_state +generic_load_microcode_early(struct microcode_intel **mc_saved_p, + unsigned int mc_saved_count, + struct ucode_cpu_info *uci) +{ + struct microcode_intel *ucode_ptr, *new_mc = NULL; + int new_rev = uci->cpu_sig.rev; + enum ucode_state state = UCODE_OK; + unsigned int mc_size; + struct microcode_header_intel *mc_header; + unsigned int csig = uci->cpu_sig.sig; + unsigned int cpf = uci->cpu_sig.pf; + int i; + + for (i = 0; i < mc_saved_count; i++) { + ucode_ptr = mc_saved_p[i]; + + mc_header = (struct microcode_header_intel *)ucode_ptr; + mc_size = get_totalsize(mc_header); + if (get_matching_microcode(csig, cpf, ucode_ptr, new_rev)) { + new_rev = mc_header->rev; + new_mc = ucode_ptr; + } + } + + if (!new_mc) { + state = UCODE_NFOUND; + goto out; + } + + uci->mc = (struct microcode_intel *)new_mc; +out: + return state; +} + +static void +microcode_pointer(struct microcode_intel **mc_saved, + unsigned long *mc_saved_in_initrd, + unsigned long initrd_start, int mc_saved_count) +{ + int i; + + for (i = 0; i < mc_saved_count; i++) + mc_saved[i] = (struct microcode_intel *) + (mc_saved_in_initrd[i] + initrd_start); +} + +#ifdef CONFIG_X86_32 +static void +microcode_phys(struct microcode_intel **mc_saved_tmp, + struct mc_saved_data *mc_saved_data) +{ + int i; + struct microcode_intel ***mc_saved; + + mc_saved = (struct microcode_intel ***) + __pa_nodebug(&mc_saved_data->mc_saved); + for (i = 0; i < mc_saved_data->mc_saved_count; i++) { + struct microcode_intel *p; + + p = *(struct microcode_intel **) + __pa_nodebug(mc_saved_data->mc_saved + i); + mc_saved_tmp[i] = (struct microcode_intel *)__pa_nodebug(p); + } +} +#endif + +static enum ucode_state +load_microcode(struct mc_saved_data *mc_saved_data, + unsigned long *mc_saved_in_initrd, + unsigned long initrd_start, + struct ucode_cpu_info *uci) +{ + struct microcode_intel *mc_saved_tmp[MAX_UCODE_COUNT]; + unsigned int count = mc_saved_data->mc_saved_count; + + if (!mc_saved_data->mc_saved) { + microcode_pointer(mc_saved_tmp, mc_saved_in_initrd, + initrd_start, count); + + return generic_load_microcode_early(mc_saved_tmp, count, uci); + } else { +#ifdef CONFIG_X86_32 + microcode_phys(mc_saved_tmp, mc_saved_data); + return generic_load_microcode_early(mc_saved_tmp, count, uci); +#else + return generic_load_microcode_early(mc_saved_data->mc_saved, + count, uci); +#endif + } +} + +static u8 get_x86_family(unsigned long sig) +{ + u8 x86; + + x86 = (sig >> 8) & 0xf; + + if (x86 == 0xf) + x86 += (sig >> 20) & 0xff; + + return x86; +} + +static u8 get_x86_model(unsigned long sig) +{ + u8 x86, x86_model; + + x86 = get_x86_family(sig); + x86_model = (sig >> 4) & 0xf; + + if (x86 == 0x6 || x86 == 0xf) + x86_model += ((sig >> 16) & 0xf) << 4; + + return x86_model; +} + +/* + * Given CPU signature and a microcode patch, this function finds if the + * microcode patch has matching family and model with the CPU. + */ +static enum ucode_state +matching_model_microcode(struct microcode_header_intel *mc_header, + unsigned long sig) +{ + u8 x86, x86_model; + u8 x86_ucode, x86_model_ucode; + struct extended_sigtable *ext_header; + unsigned long total_size = get_totalsize(mc_header); + unsigned long data_size = get_datasize(mc_header); + int ext_sigcount, i; + struct extended_signature *ext_sig; + + x86 = get_x86_family(sig); + x86_model = get_x86_model(sig); + + x86_ucode = get_x86_family(mc_header->sig); + x86_model_ucode = get_x86_model(mc_header->sig); + + if (x86 == x86_ucode && x86_model == x86_model_ucode) + return UCODE_OK; + + /* Look for ext. headers: */ + if (total_size <= data_size + MC_HEADER_SIZE) + return UCODE_NFOUND; + + ext_header = (struct extended_sigtable *) + mc_header + data_size + MC_HEADER_SIZE; + ext_sigcount = ext_header->count; + ext_sig = (void *)ext_header + EXT_HEADER_SIZE; + + for (i = 0; i < ext_sigcount; i++) { + x86_ucode = get_x86_family(ext_sig->sig); + x86_model_ucode = get_x86_model(ext_sig->sig); + + if (x86 == x86_ucode && x86_model == x86_model_ucode) + return UCODE_OK; + + ext_sig++; + } + + return UCODE_NFOUND; +} + +static int +save_microcode(struct mc_saved_data *mc_saved_data, + struct microcode_intel **mc_saved_src, + unsigned int mc_saved_count) +{ + int i, j; + struct microcode_intel **mc_saved_p; + int ret; + + if (!mc_saved_count) + return -EINVAL; + + /* + * Copy new microcode data. + */ + mc_saved_p = kmalloc(mc_saved_count*sizeof(struct microcode_intel *), + GFP_KERNEL); + if (!mc_saved_p) + return -ENOMEM; + + for (i = 0; i < mc_saved_count; i++) { + struct microcode_intel *mc = mc_saved_src[i]; + struct microcode_header_intel *mc_header = &mc->hdr; + unsigned long mc_size = get_totalsize(mc_header); + mc_saved_p[i] = kmalloc(mc_size, GFP_KERNEL); + if (!mc_saved_p[i]) { + ret = -ENOMEM; + goto err; + } + if (!mc_saved_src[i]) { + ret = -EINVAL; + goto err; + } + memcpy(mc_saved_p[i], mc, mc_size); + } + + /* + * Point to newly saved microcode. + */ + mc_saved_data->mc_saved = mc_saved_p; + mc_saved_data->mc_saved_count = mc_saved_count; + + return 0; + +err: + for (j = 0; j <= i; j++) + kfree(mc_saved_p[j]); + kfree(mc_saved_p); + + return ret; +} + +/* + * A microcode patch in ucode_ptr is saved into mc_saved + * - if it has matching signature and newer revision compared to an existing + * patch mc_saved. + * - or if it is a newly discovered microcode patch. + * + * The microcode patch should have matching model with CPU. + */ +static void _save_mc(struct microcode_intel **mc_saved, u8 *ucode_ptr, + unsigned int *mc_saved_count_p) +{ + int i; + int found = 0; + unsigned int mc_saved_count = *mc_saved_count_p; + struct microcode_header_intel *mc_header; + + mc_header = (struct microcode_header_intel *)ucode_ptr; + for (i = 0; i < mc_saved_count; i++) { + unsigned int sig, pf; + unsigned int new_rev; + struct microcode_header_intel *mc_saved_header = + (struct microcode_header_intel *)mc_saved[i]; + sig = mc_saved_header->sig; + pf = mc_saved_header->pf; + new_rev = mc_header->rev; + + if (get_matching_sig(sig, pf, ucode_ptr, new_rev)) { + found = 1; + if (update_match_revision(mc_header, new_rev)) { + /* + * Found an older ucode saved before. + * Replace the older one with this newer + * one. + */ + mc_saved[i] = + (struct microcode_intel *)ucode_ptr; + break; + } + } + } + if (i >= mc_saved_count && !found) + /* + * This ucode is first time discovered in ucode file. + * Save it to memory. + */ + mc_saved[mc_saved_count++] = + (struct microcode_intel *)ucode_ptr; + + *mc_saved_count_p = mc_saved_count; +} + +/* + * Get microcode matching with BSP's model. Only CPUs with the same model as + * BSP can stay in the platform. + */ +static enum ucode_state __init +get_matching_model_microcode(int cpu, unsigned long start, + void *data, size_t size, + struct mc_saved_data *mc_saved_data, + unsigned long *mc_saved_in_initrd, + struct ucode_cpu_info *uci) +{ + u8 *ucode_ptr = data; + unsigned int leftover = size; + enum ucode_state state = UCODE_OK; + unsigned int mc_size; + struct microcode_header_intel *mc_header; + struct microcode_intel *mc_saved_tmp[MAX_UCODE_COUNT]; + unsigned int mc_saved_count = mc_saved_data->mc_saved_count; + int i; + + while (leftover) { + mc_header = (struct microcode_header_intel *)ucode_ptr; + + mc_size = get_totalsize(mc_header); + if (!mc_size || mc_size > leftover || + microcode_sanity_check(ucode_ptr, 0) < 0) + break; + + leftover -= mc_size; + + /* + * Since APs with same family and model as the BSP may boot in + * the platform, we need to find and save microcode patches + * with the same family and model as the BSP. + */ + if (matching_model_microcode(mc_header, uci->cpu_sig.sig) != + UCODE_OK) { + ucode_ptr += mc_size; + continue; + } + + _save_mc(mc_saved_tmp, ucode_ptr, &mc_saved_count); + + ucode_ptr += mc_size; + } + + if (leftover) { + state = UCODE_ERROR; + goto out; + } + + if (mc_saved_count == 0) { + state = UCODE_NFOUND; + goto out; + } + + for (i = 0; i < mc_saved_count; i++) + mc_saved_in_initrd[i] = (unsigned long)mc_saved_tmp[i] - start; + + mc_saved_data->mc_saved_count = mc_saved_count; +out: + return state; +} + +static int collect_cpu_info_early(struct ucode_cpu_info *uci) +{ + unsigned int val[2]; + u8 x86, x86_model; + struct cpu_signature csig; + unsigned int eax, ebx, ecx, edx; + + csig.sig = 0; + csig.pf = 0; + csig.rev = 0; + + memset(uci, 0, sizeof(*uci)); + + eax = 0x00000001; + ecx = 0; + native_cpuid(&eax, &ebx, &ecx, &edx); + csig.sig = eax; + + x86 = get_x86_family(csig.sig); + x86_model = get_x86_model(csig.sig); + + if ((x86_model >= 5) || (x86 > 6)) { + /* get processor flags from MSR 0x17 */ + native_rdmsr(MSR_IA32_PLATFORM_ID, val[0], val[1]); + csig.pf = 1 << ((val[1] >> 18) & 7); + } + native_wrmsr(MSR_IA32_UCODE_REV, 0, 0); + + /* As documented in the SDM: Do a CPUID 1 here */ + sync_core(); + + /* get the current revision from MSR 0x8B */ + native_rdmsr(MSR_IA32_UCODE_REV, val[0], val[1]); + + csig.rev = val[1]; + + uci->cpu_sig = csig; + uci->valid = 1; + + return 0; +} + +#ifdef DEBUG +static void __ref show_saved_mc(void) +{ + int i, j; + unsigned int sig, pf, rev, total_size, data_size, date; + struct ucode_cpu_info uci; + + if (mc_saved_data.mc_saved_count == 0) { + pr_debug("no micorcode data saved.\n"); + return; + } + pr_debug("Total microcode saved: %d\n", mc_saved_data.mc_saved_count); + + collect_cpu_info_early(&uci); + + sig = uci.cpu_sig.sig; + pf = uci.cpu_sig.pf; + rev = uci.cpu_sig.rev; + pr_debug("CPU%d: sig=0x%x, pf=0x%x, rev=0x%x\n", + smp_processor_id(), sig, pf, rev); + + for (i = 0; i < mc_saved_data.mc_saved_count; i++) { + struct microcode_header_intel *mc_saved_header; + struct extended_sigtable *ext_header; + int ext_sigcount; + struct extended_signature *ext_sig; + + mc_saved_header = (struct microcode_header_intel *) + mc_saved_data.mc_saved[i]; + sig = mc_saved_header->sig; + pf = mc_saved_header->pf; + rev = mc_saved_header->rev; + total_size = get_totalsize(mc_saved_header); + data_size = get_datasize(mc_saved_header); + date = mc_saved_header->date; + + pr_debug("mc_saved[%d]: sig=0x%x, pf=0x%x, rev=0x%x, toal size=0x%x, date = %04x-%02x-%02x\n", + i, sig, pf, rev, total_size, + date & 0xffff, + date >> 24, + (date >> 16) & 0xff); + + /* Look for ext. headers: */ + if (total_size <= data_size + MC_HEADER_SIZE) + continue; + + ext_header = (struct extended_sigtable *) + mc_saved_header + data_size + MC_HEADER_SIZE; + ext_sigcount = ext_header->count; + ext_sig = (void *)ext_header + EXT_HEADER_SIZE; + + for (j = 0; j < ext_sigcount; j++) { + sig = ext_sig->sig; + pf = ext_sig->pf; + + pr_debug("\tExtended[%d]: sig=0x%x, pf=0x%x\n", + j, sig, pf); + + ext_sig++; + } + + } +} +#else +static inline void show_saved_mc(void) +{ +} +#endif + +#if defined(CONFIG_MICROCODE_INTEL_EARLY) && defined(CONFIG_HOTPLUG_CPU) +static DEFINE_MUTEX(x86_cpu_microcode_mutex); +/* + * Save this mc into mc_saved_data. So it will be loaded early when a CPU is + * hot added or resumes. + * + * Please make sure this mc should be a valid microcode patch before calling + * this function. + */ +int save_mc_for_early(u8 *mc) +{ + struct microcode_intel *mc_saved_tmp[MAX_UCODE_COUNT]; + unsigned int mc_saved_count_init; + unsigned int mc_saved_count; + struct microcode_intel **mc_saved; + int ret = 0; + int i; + + /* + * Hold hotplug lock so mc_saved_data is not accessed by a CPU in + * hotplug. + */ + mutex_lock(&x86_cpu_microcode_mutex); + + mc_saved_count_init = mc_saved_data.mc_saved_count; + mc_saved_count = mc_saved_data.mc_saved_count; + mc_saved = mc_saved_data.mc_saved; + + if (mc_saved && mc_saved_count) + memcpy(mc_saved_tmp, mc_saved, + mc_saved_count * sizeof(struct mirocode_intel *)); + /* + * Save the microcode patch mc in mc_save_tmp structure if it's a newer + * version. + */ + + _save_mc(mc_saved_tmp, mc, &mc_saved_count); + + /* + * Save the mc_save_tmp in global mc_saved_data. + */ + ret = save_microcode(&mc_saved_data, mc_saved_tmp, mc_saved_count); + if (ret) { + pr_err("Cannot save microcode patch.\n"); + goto out; + } + + show_saved_mc(); + + /* + * Free old saved microcod data. + */ + if (mc_saved) { + for (i = 0; i < mc_saved_count_init; i++) + kfree(mc_saved[i]); + kfree(mc_saved); + } + +out: + mutex_unlock(&x86_cpu_microcode_mutex); + + return ret; +} +EXPORT_SYMBOL_GPL(save_mc_for_early); +#endif + +static __initdata char ucode_name[] = "kernel/x86/microcode/GenuineIntel.bin"; +static __init enum ucode_state +scan_microcode(unsigned long start, unsigned long end, + struct mc_saved_data *mc_saved_data, + unsigned long *mc_saved_in_initrd, + struct ucode_cpu_info *uci) +{ + unsigned int size = end - start + 1; + struct cpio_data cd; + long offset = 0; +#ifdef CONFIG_X86_32 + char *p = (char *)__pa_nodebug(ucode_name); +#else + char *p = ucode_name; +#endif + + cd.data = NULL; + cd.size = 0; + + cd = find_cpio_data(p, (void *)start, size, &offset); + if (!cd.data) + return UCODE_ERROR; + + + return get_matching_model_microcode(0, start, cd.data, cd.size, + mc_saved_data, mc_saved_in_initrd, + uci); +} + +/* + * Print ucode update info. + */ +static void +print_ucode_info(struct ucode_cpu_info *uci, unsigned int date) +{ + int cpu = smp_processor_id(); + + pr_info("CPU%d microcode updated early to revision 0x%x, date = %04x-%02x-%02x\n", + cpu, + uci->cpu_sig.rev, + date & 0xffff, + date >> 24, + (date >> 16) & 0xff); +} + +#ifdef CONFIG_X86_32 + +static int delay_ucode_info; +static int current_mc_date; + +/* + * Print early updated ucode info after printk works. This is delayed info dump. + */ +void show_ucode_info_early(void) +{ + struct ucode_cpu_info uci; + + if (delay_ucode_info) { + collect_cpu_info_early(&uci); + print_ucode_info(&uci, current_mc_date); + delay_ucode_info = 0; + } +} + +/* + * At this point, we can not call printk() yet. Keep microcode patch number in + * mc_saved_data.mc_saved and delay printing microcode info in + * show_ucode_info_early() until printk() works. + */ +static void print_ucode(struct ucode_cpu_info *uci) +{ + struct microcode_intel *mc_intel; + int *delay_ucode_info_p; + int *current_mc_date_p; + + mc_intel = uci->mc; + if (mc_intel == NULL) + return; + + delay_ucode_info_p = (int *)__pa_nodebug(&delay_ucode_info); + current_mc_date_p = (int *)__pa_nodebug(¤t_mc_date); + + *delay_ucode_info_p = 1; + *current_mc_date_p = mc_intel->hdr.date; +} +#else + +/* + * Flush global tlb. We only do this in x86_64 where paging has been enabled + * already and PGE should be enabled as well. + */ +static inline void flush_tlb_early(void) +{ + __native_flush_tlb_global_irq_disabled(); +} + +static inline void print_ucode(struct ucode_cpu_info *uci) +{ + struct microcode_intel *mc_intel; + + mc_intel = uci->mc; + if (mc_intel == NULL) + return; + + print_ucode_info(uci, mc_intel->hdr.date); +} +#endif + +static int apply_microcode_early(struct mc_saved_data *mc_saved_data, + struct ucode_cpu_info *uci) +{ + struct microcode_intel *mc_intel; + unsigned int val[2]; + + mc_intel = uci->mc; + if (mc_intel == NULL) + return 0; + + /* write microcode via MSR 0x79 */ + native_wrmsr(MSR_IA32_UCODE_WRITE, + (unsigned long) mc_intel->bits, + (unsigned long) mc_intel->bits >> 16 >> 16); + native_wrmsr(MSR_IA32_UCODE_REV, 0, 0); + + /* As documented in the SDM: Do a CPUID 1 here */ + sync_core(); + + /* get the current revision from MSR 0x8B */ + native_rdmsr(MSR_IA32_UCODE_REV, val[0], val[1]); + if (val[1] != mc_intel->hdr.rev) + return -1; + +#ifdef CONFIG_X86_64 + /* Flush global tlb. This is precaution. */ + flush_tlb_early(); +#endif + uci->cpu_sig.rev = val[1]; + + print_ucode(uci); + + return 0; +} + +/* + * This function converts microcode patch offsets previously stored in + * mc_saved_in_initrd to pointers and stores the pointers in mc_saved_data. + */ +int __init save_microcode_in_initrd_intel(void) +{ + unsigned int count = mc_saved_data.mc_saved_count; + struct microcode_intel *mc_saved[MAX_UCODE_COUNT]; + int ret = 0; + + if (count == 0) + return ret; + + microcode_pointer(mc_saved, mc_saved_in_initrd, initrd_start, count); + ret = save_microcode(&mc_saved_data, mc_saved, count); + if (ret) + pr_err("Cannot save microcode patches from initrd.\n"); + + show_saved_mc(); + + return ret; +} + +static void __init +_load_ucode_intel_bsp(struct mc_saved_data *mc_saved_data, + unsigned long *mc_saved_in_initrd, + unsigned long initrd_start_early, + unsigned long initrd_end_early, + struct ucode_cpu_info *uci) +{ + collect_cpu_info_early(uci); + scan_microcode(initrd_start_early, initrd_end_early, mc_saved_data, + mc_saved_in_initrd, uci); + load_microcode(mc_saved_data, mc_saved_in_initrd, + initrd_start_early, uci); + apply_microcode_early(mc_saved_data, uci); +} + +void __init +load_ucode_intel_bsp(void) +{ + u64 ramdisk_image, ramdisk_size; + unsigned long initrd_start_early, initrd_end_early; + struct ucode_cpu_info uci; +#ifdef CONFIG_X86_32 + struct boot_params *boot_params_p; + + boot_params_p = (struct boot_params *)__pa_nodebug(&boot_params); + ramdisk_image = boot_params_p->hdr.ramdisk_image; + ramdisk_size = boot_params_p->hdr.ramdisk_size; + initrd_start_early = ramdisk_image; + initrd_end_early = initrd_start_early + ramdisk_size; + + _load_ucode_intel_bsp( + (struct mc_saved_data *)__pa_nodebug(&mc_saved_data), + (unsigned long *)__pa_nodebug(&mc_saved_in_initrd), + initrd_start_early, initrd_end_early, &uci); +#else + ramdisk_image = boot_params.hdr.ramdisk_image; + ramdisk_size = boot_params.hdr.ramdisk_size; + initrd_start_early = ramdisk_image + PAGE_OFFSET; + initrd_end_early = initrd_start_early + ramdisk_size; + + _load_ucode_intel_bsp(&mc_saved_data, mc_saved_in_initrd, + initrd_start_early, initrd_end_early, &uci); +#endif +} + +void load_ucode_intel_ap(void) +{ + struct mc_saved_data *mc_saved_data_p; + struct ucode_cpu_info uci; + unsigned long *mc_saved_in_initrd_p; + unsigned long initrd_start_addr; +#ifdef CONFIG_X86_32 + unsigned long *initrd_start_p; + + mc_saved_in_initrd_p = + (unsigned long *)__pa_nodebug(mc_saved_in_initrd); + mc_saved_data_p = (struct mc_saved_data *)__pa_nodebug(&mc_saved_data); + initrd_start_p = (unsigned long *)__pa_nodebug(&initrd_start); + initrd_start_addr = (unsigned long)__pa_nodebug(*initrd_start_p); +#else + mc_saved_data_p = &mc_saved_data; + mc_saved_in_initrd_p = mc_saved_in_initrd; + initrd_start_addr = initrd_start; +#endif + + /* + * If there is no valid ucode previously saved in memory, no need to + * update ucode on this AP. + */ + if (mc_saved_data_p->mc_saved_count == 0) + return; + + collect_cpu_info_early(&uci); + load_microcode(mc_saved_data_p, mc_saved_in_initrd_p, + initrd_start_addr, &uci); + apply_microcode_early(mc_saved_data_p, &uci); +} diff --git a/arch/x86/kernel/cpu/microcode/intel_lib.c b/arch/x86/kernel/cpu/microcode/intel_lib.c new file mode 100644 index 00000000000..ce69320d017 --- /dev/null +++ b/arch/x86/kernel/cpu/microcode/intel_lib.c @@ -0,0 +1,174 @@ +/* + * Intel CPU Microcode Update Driver for Linux + * + * Copyright (C) 2012 Fenghua Yu <fenghua.yu@intel.com> + * H Peter Anvin" <hpa@zytor.com> + * + * This driver allows to upgrade microcode on Intel processors + * belonging to IA-32 family - PentiumPro, Pentium II, + * Pentium III, Xeon, Pentium 4, etc. + * + * Reference: Section 8.11 of Volume 3a, IA-32 Intel? Architecture + * Software Developer's Manual + * Order Number 253668 or free download from: + * + * http://developer.intel.com/Assets/PDF/manual/253668.pdf + * + * For more information, go to http://www.urbanmyth.org/microcode + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + * + */ +#include <linux/firmware.h> +#include <linux/uaccess.h> +#include <linux/kernel.h> +#include <linux/module.h> + +#include <asm/microcode_intel.h> +#include <asm/processor.h> +#include <asm/msr.h> + +static inline int +update_match_cpu(unsigned int csig, unsigned int cpf, + unsigned int sig, unsigned int pf) +{ + return (!sigmatch(sig, csig, pf, cpf)) ? 0 : 1; +} + +int +update_match_revision(struct microcode_header_intel *mc_header, int rev) +{ + return (mc_header->rev <= rev) ? 0 : 1; +} + +int microcode_sanity_check(void *mc, int print_err) +{ + unsigned long total_size, data_size, ext_table_size; + struct microcode_header_intel *mc_header = mc; + struct extended_sigtable *ext_header = NULL; + int sum, orig_sum, ext_sigcount = 0, i; + struct extended_signature *ext_sig; + + total_size = get_totalsize(mc_header); + data_size = get_datasize(mc_header); + + if (data_size + MC_HEADER_SIZE > total_size) { + if (print_err) + pr_err("error! Bad data size in microcode data file\n"); + return -EINVAL; + } + + if (mc_header->ldrver != 1 || mc_header->hdrver != 1) { + if (print_err) + pr_err("error! Unknown microcode update format\n"); + return -EINVAL; + } + ext_table_size = total_size - (MC_HEADER_SIZE + data_size); + if (ext_table_size) { + if ((ext_table_size < EXT_HEADER_SIZE) + || ((ext_table_size - EXT_HEADER_SIZE) % EXT_SIGNATURE_SIZE)) { + if (print_err) + pr_err("error! Small exttable size in microcode data file\n"); + return -EINVAL; + } + ext_header = mc + MC_HEADER_SIZE + data_size; + if (ext_table_size != exttable_size(ext_header)) { + if (print_err) + pr_err("error! Bad exttable size in microcode data file\n"); + return -EFAULT; + } + ext_sigcount = ext_header->count; + } + + /* check extended table checksum */ + if (ext_table_size) { + int ext_table_sum = 0; + int *ext_tablep = (int *)ext_header; + + i = ext_table_size / DWSIZE; + while (i--) + ext_table_sum += ext_tablep[i]; + if (ext_table_sum) { + if (print_err) + pr_warn("aborting, bad extended signature table checksum\n"); + return -EINVAL; + } + } + + /* calculate the checksum */ + orig_sum = 0; + i = (MC_HEADER_SIZE + data_size) / DWSIZE; + while (i--) + orig_sum += ((int *)mc)[i]; + if (orig_sum) { + if (print_err) + pr_err("aborting, bad checksum\n"); + return -EINVAL; + } + if (!ext_table_size) + return 0; + /* check extended signature checksum */ + for (i = 0; i < ext_sigcount; i++) { + ext_sig = (void *)ext_header + EXT_HEADER_SIZE + + EXT_SIGNATURE_SIZE * i; + sum = orig_sum + - (mc_header->sig + mc_header->pf + mc_header->cksum) + + (ext_sig->sig + ext_sig->pf + ext_sig->cksum); + if (sum) { + if (print_err) + pr_err("aborting, bad checksum\n"); + return -EINVAL; + } + } + return 0; +} +EXPORT_SYMBOL_GPL(microcode_sanity_check); + +/* + * return 0 - no update found + * return 1 - found update + */ +int get_matching_sig(unsigned int csig, int cpf, void *mc, int rev) +{ + struct microcode_header_intel *mc_header = mc; + struct extended_sigtable *ext_header; + unsigned long total_size = get_totalsize(mc_header); + int ext_sigcount, i; + struct extended_signature *ext_sig; + + if (update_match_cpu(csig, cpf, mc_header->sig, mc_header->pf)) + return 1; + + /* Look for ext. headers: */ + if (total_size <= get_datasize(mc_header) + MC_HEADER_SIZE) + return 0; + + ext_header = mc + get_datasize(mc_header) + MC_HEADER_SIZE; + ext_sigcount = ext_header->count; + ext_sig = (void *)ext_header + EXT_HEADER_SIZE; + + for (i = 0; i < ext_sigcount; i++) { + if (update_match_cpu(csig, cpf, ext_sig->sig, ext_sig->pf)) + return 1; + ext_sig++; + } + return 0; +} + +/* + * return 0 - no update found + * return 1 - found update + */ +int get_matching_microcode(unsigned int csig, int cpf, void *mc, int rev) +{ + struct microcode_header_intel *mc_header = mc; + + if (!update_match_revision(mc_header, rev)) + return 0; + + return get_matching_sig(csig, cpf, mc, rev); +} +EXPORT_SYMBOL_GPL(get_matching_microcode); diff --git a/arch/x86/kernel/cpu/mkcapflags.pl b/arch/x86/kernel/cpu/mkcapflags.pl deleted file mode 100644 index dfea390e160..00000000000 --- a/arch/x86/kernel/cpu/mkcapflags.pl +++ /dev/null @@ -1,32 +0,0 @@ -#!/usr/bin/perl -# -# Generate the x86_cap_flags[] array from include/asm-x86/cpufeature.h -# - -($in, $out) = @ARGV; - -open(IN, "< $in\0") or die "$0: cannot open: $in: $!\n"; -open(OUT, "> $out\0") or die "$0: cannot create: $out: $!\n"; - -print OUT "#include <asm/cpufeature.h>\n\n"; -print OUT "const char * const x86_cap_flags[NCAPINTS*32] = {\n"; - -while (defined($line = <IN>)) { - if ($line =~ /^\s*\#\s*define\s+(X86_FEATURE_(\S+))\s+(.*)$/) { - $macro = $1; - $feature = $2; - $tail = $3; - if ($tail =~ /\/\*\s*\"([^"]*)\".*\*\//) { - $feature = $1; - } - - if ($feature ne '') { - printf OUT "\t%-32s = \"%s\",\n", - "[$macro]", "\L$feature"; - } - } -} -print OUT "};\n"; - -close(IN); -close(OUT); diff --git a/arch/x86/kernel/cpu/mkcapflags.sh b/arch/x86/kernel/cpu/mkcapflags.sh new file mode 100644 index 00000000000..2bf61650549 --- /dev/null +++ b/arch/x86/kernel/cpu/mkcapflags.sh @@ -0,0 +1,41 @@ +#!/bin/sh +# +# Generate the x86_cap_flags[] array from include/asm/cpufeature.h +# + +IN=$1 +OUT=$2 + +TABS="$(printf '\t\t\t\t\t')" +trap 'rm "$OUT"' EXIT + +( + echo "#ifndef _ASM_X86_CPUFEATURE_H" + echo "#include <asm/cpufeature.h>" + echo "#endif" + echo "" + echo "const char * const x86_cap_flags[NCAPINTS*32] = {" + + # Iterate through any input lines starting with #define X86_FEATURE_ + sed -n -e 's/\t/ /g' -e 's/^ *# *define *X86_FEATURE_//p' $IN | + while read i + do + # Name is everything up to the first whitespace + NAME="$(echo "$i" | sed 's/ .*//')" + + # If the /* comment */ starts with a quote string, grab that. + VALUE="$(echo "$i" | sed -n 's@.*/\* *\("[^"]*"\).*\*/@\1@p')" + [ -z "$VALUE" ] && VALUE="\"$NAME\"" + [ "$VALUE" == '""' ] && continue + + # Name is uppercase, VALUE is all lowercase + VALUE="$(echo "$VALUE" | tr A-Z a-z)" + + TABCOUNT=$(( ( 5*8 - 14 - $(echo "$NAME" | wc -c) ) / 8 )) + printf "\t[%s]%.*s = %s,\n" \ + "X86_FEATURE_$NAME" "$TABCOUNT" "$TABS" "$VALUE" + done + echo "};" +) > $OUT + +trap - EXIT diff --git a/arch/x86/kernel/cpu/mshyperv.c b/arch/x86/kernel/cpu/mshyperv.c index 0a630dd4b62..a450373e8e9 100644 --- a/arch/x86/kernel/cpu/mshyperv.c +++ b/arch/x86/kernel/cpu/mshyperv.c @@ -14,28 +14,80 @@ #include <linux/time.h> #include <linux/clocksource.h> #include <linux/module.h> +#include <linux/hardirq.h> +#include <linux/efi.h> +#include <linux/interrupt.h> +#include <linux/irq.h> #include <asm/processor.h> #include <asm/hypervisor.h> #include <asm/hyperv.h> #include <asm/mshyperv.h> +#include <asm/desc.h> +#include <asm/idle.h> +#include <asm/irq_regs.h> +#include <asm/i8259.h> +#include <asm/apic.h> +#include <asm/timer.h> struct ms_hyperv_info ms_hyperv; EXPORT_SYMBOL_GPL(ms_hyperv); -static bool __init ms_hyperv_platform(void) +#if IS_ENABLED(CONFIG_HYPERV) +static void (*vmbus_handler)(void); + +void hyperv_vector_handler(struct pt_regs *regs) +{ + struct pt_regs *old_regs = set_irq_regs(regs); + + irq_enter(); + exit_idle(); + + inc_irq_stat(irq_hv_callback_count); + if (vmbus_handler) + vmbus_handler(); + + irq_exit(); + set_irq_regs(old_regs); +} + +void hv_setup_vmbus_irq(void (*handler)(void)) +{ + vmbus_handler = handler; + /* + * Setup the IDT for hypervisor callback. Prevent reallocation + * at module reload. + */ + if (!test_bit(HYPERVISOR_CALLBACK_VECTOR, used_vectors)) + alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR, + hyperv_callback_vector); +} + +void hv_remove_vmbus_irq(void) +{ + /* We have no way to deallocate the interrupt gate */ + vmbus_handler = NULL; +} +EXPORT_SYMBOL_GPL(hv_setup_vmbus_irq); +EXPORT_SYMBOL_GPL(hv_remove_vmbus_irq); +#endif + +static uint32_t __init ms_hyperv_platform(void) { u32 eax; u32 hyp_signature[3]; if (!boot_cpu_has(X86_FEATURE_HYPERVISOR)) - return false; + return 0; cpuid(HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS, &eax, &hyp_signature[0], &hyp_signature[1], &hyp_signature[2]); - return eax >= HYPERV_CPUID_MIN && - eax <= HYPERV_CPUID_MAX && - !memcmp("Microsoft Hv", hyp_signature, 12); + if (eax >= HYPERV_CPUID_MIN && + eax <= HYPERV_CPUID_MAX && + !memcmp("Microsoft Hv", hyp_signature, 12)) + return HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS; + + return 0; } static cycle_t read_hv_clock(struct clocksource *arg) @@ -68,7 +120,28 @@ static void __init ms_hyperv_init_platform(void) printk(KERN_INFO "HyperV: features 0x%x, hints 0x%x\n", ms_hyperv.features, ms_hyperv.hints); - clocksource_register_hz(&hyperv_cs, NSEC_PER_SEC/100); +#ifdef CONFIG_X86_LOCAL_APIC + if (ms_hyperv.features & HV_X64_MSR_APIC_FREQUENCY_AVAILABLE) { + /* + * Get the APIC frequency. + */ + u64 hv_lapic_frequency; + + rdmsrl(HV_X64_MSR_APIC_FREQUENCY, hv_lapic_frequency); + hv_lapic_frequency = div_u64(hv_lapic_frequency, HZ); + lapic_timer_frequency = hv_lapic_frequency; + printk(KERN_INFO "HyperV: LAPIC Timer Frequency: %#x\n", + lapic_timer_frequency); + } +#endif + + if (ms_hyperv.features & HV_X64_MSR_TIME_REF_COUNT_AVAILABLE) + clocksource_register_hz(&hyperv_cs, NSEC_PER_SEC/100); + +#ifdef CONFIG_X86_IO_APIC + no_timer_check = 1; +#endif + } const __refconst struct hypervisor_x86 x86_hyper_ms_hyperv = { diff --git a/arch/x86/kernel/cpu/mtrr/cleanup.c b/arch/x86/kernel/cpu/mtrr/cleanup.c index ac140c7be39..5f90b85ff22 100644 --- a/arch/x86/kernel/cpu/mtrr/cleanup.c +++ b/arch/x86/kernel/cpu/mtrr/cleanup.c @@ -258,15 +258,15 @@ range_to_mtrr(unsigned int reg, unsigned long range_startk, /* Compute the maximum size with which we can make a range: */ if (range_startk) - max_align = ffs(range_startk) - 1; + max_align = __ffs(range_startk); else - max_align = 32; + max_align = BITS_PER_LONG - 1; - align = fls(range_sizek) - 1; + align = __fls(range_sizek); if (align > max_align) align = max_align; - sizek = 1 << align; + sizek = 1UL << align; if (debug_print) { char start_factor = 'K', size_factor = 'K'; unsigned long start_base, size_base; @@ -714,15 +714,15 @@ int __init mtrr_cleanup(unsigned address_bits) if (mtrr_tom2) x_remove_size = (mtrr_tom2 >> PAGE_SHIFT) - x_remove_base; - nr_range = x86_get_mtrr_mem_range(range, 0, x_remove_base, x_remove_size); /* * [0, 1M) should always be covered by var mtrr with WB * and fixed mtrrs should take effect before var mtrr for it: */ - nr_range = add_range_with_merge(range, RANGE_NUM, nr_range, 0, + nr_range = add_range_with_merge(range, RANGE_NUM, 0, 0, 1ULL<<(20 - PAGE_SHIFT)); - /* Sort the ranges: */ - sort_range(range, nr_range); + /* add from var mtrr at last */ + nr_range = x86_get_mtrr_mem_range(range, nr_range, + x_remove_base, x_remove_size); range_sums = sum_ranges(range, nr_range); printk(KERN_INFO "total RAM covered: %ldM\n", diff --git a/arch/x86/kernel/cpu/mtrr/cyrix.c b/arch/x86/kernel/cpu/mtrr/cyrix.c index 68a3343e579..9e451b0876b 100644 --- a/arch/x86/kernel/cpu/mtrr/cyrix.c +++ b/arch/x86/kernel/cpu/mtrr/cyrix.c @@ -167,7 +167,7 @@ static void post_set(void) setCx86(CX86_CCR3, ccr3); /* Enable caches */ - write_cr0(read_cr0() & 0xbfffffff); + write_cr0(read_cr0() & ~X86_CR0_CD); /* Restore value of CR4 */ if (cpu_has_pge) diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c index 97b26356e9e..0e25a1bc5ab 100644 --- a/arch/x86/kernel/cpu/mtrr/generic.c +++ b/arch/x86/kernel/cpu/mtrr/generic.c @@ -12,7 +12,6 @@ #include <asm/processor-flags.h> #include <asm/cpufeature.h> #include <asm/tlbflush.h> -#include <asm/system.h> #include <asm/mtrr.h> #include <asm/msr.h> #include <asm/pat.h> @@ -362,11 +361,7 @@ static void __init print_mtrr_state(void) } pr_debug("MTRR variable ranges %sabled:\n", mtrr_state.enabled & 2 ? "en" : "dis"); - if (size_or_mask & 0xffffffffUL) - high_width = ffs(size_or_mask & 0xffffffffUL) - 1; - else - high_width = ffs(size_or_mask>>32) + 32 - 1; - high_width = (high_width - (32 - PAGE_SHIFT) + 3) / 4; + high_width = (__ffs64(size_or_mask) - (32 - PAGE_SHIFT) + 3) / 4; for (i = 0; i < num_var_ranges; ++i) { if (mtrr_state.var_ranges[i].mask_lo & (1 << 11)) @@ -515,8 +510,9 @@ generic_get_free_region(unsigned long base, unsigned long size, int replace_reg) static void generic_get_mtrr(unsigned int reg, unsigned long *base, unsigned long *size, mtrr_type *type) { - unsigned int mask_lo, mask_hi, base_lo, base_hi; - unsigned int tmp, hi; + u32 mask_lo, mask_hi, base_lo, base_hi; + unsigned int hi; + u64 tmp, mask; /* * get_mtrr doesn't need to update mtrr_state, also it could be called @@ -537,18 +533,18 @@ static void generic_get_mtrr(unsigned int reg, unsigned long *base, rdmsr(MTRRphysBase_MSR(reg), base_lo, base_hi); /* Work out the shifted address mask: */ - tmp = mask_hi << (32 - PAGE_SHIFT) | mask_lo >> PAGE_SHIFT; - mask_lo = size_or_mask | tmp; + tmp = (u64)mask_hi << (32 - PAGE_SHIFT) | mask_lo >> PAGE_SHIFT; + mask = size_or_mask | tmp; /* Expand tmp with high bits to all 1s: */ - hi = fls(tmp); + hi = fls64(tmp); if (hi > 0) { - tmp |= ~((1<<(hi - 1)) - 1); + tmp |= ~((1ULL<<(hi - 1)) - 1); - if (tmp != mask_lo) { + if (tmp != mask) { printk(KERN_WARNING "mtrr: your BIOS has configured an incorrect mask, fixing it.\n"); - add_taint(TAINT_FIRMWARE_WORKAROUND); - mask_lo = tmp; + add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK); + mask = tmp; } } @@ -556,8 +552,8 @@ static void generic_get_mtrr(unsigned int reg, unsigned long *base, * This works correctly if size is a power of two, i.e. a * contiguous range: */ - *size = -mask_lo; - *base = base_hi << (32 - PAGE_SHIFT) | base_lo >> PAGE_SHIFT; + *size = -mask; + *base = (u64)base_hi << (32 - PAGE_SHIFT) | base_lo >> PAGE_SHIFT; *type = base_lo & 0xff; out_put_cpu: @@ -687,6 +683,7 @@ static void prepare_set(void) __acquires(set_atomicity_lock) } /* Flush all TLBs via a mov %cr3, %reg; mov %reg, %cr3 */ + count_vm_tlb_event(NR_TLB_LOCAL_FLUSH_ALL); __flush_tlb(); /* Save MTRR state */ @@ -700,13 +697,14 @@ static void prepare_set(void) __acquires(set_atomicity_lock) static void post_set(void) __releases(set_atomicity_lock) { /* Flush TLBs (no need to flush caches - they are disabled) */ + count_vm_tlb_event(NR_TLB_LOCAL_FLUSH_ALL); __flush_tlb(); /* Intel (P6) standard MTRRs */ mtrr_wrmsr(MSR_MTRRdefType, deftype_lo, deftype_hi); /* Enable caches */ - write_cr0(read_cr0() & 0xbfffffff); + write_cr0(read_cr0() & ~X86_CR0_CD); /* Restore value of CR4 */ if (cpu_has_pge) diff --git a/arch/x86/kernel/cpu/mtrr/if.c b/arch/x86/kernel/cpu/mtrr/if.c index 79289632cb2..a041e094b8b 100644 --- a/arch/x86/kernel/cpu/mtrr/if.c +++ b/arch/x86/kernel/cpu/mtrr/if.c @@ -167,6 +167,7 @@ mtrr_ioctl(struct file *file, unsigned int cmd, unsigned long __arg) { int err = 0; mtrr_type type; + unsigned long base; unsigned long size; struct mtrr_sentry sentry; struct mtrr_gentry gentry; @@ -267,14 +268,14 @@ mtrr_ioctl(struct file *file, unsigned int cmd, unsigned long __arg) #endif if (gentry.regnum >= num_var_ranges) return -EINVAL; - mtrr_if->get(gentry.regnum, &gentry.base, &size, &type); + mtrr_if->get(gentry.regnum, &base, &size, &type); /* Hide entries that go above 4GB */ - if (gentry.base + size - 1 >= (1UL << (8 * sizeof(gentry.size) - PAGE_SHIFT)) + if (base + size - 1 >= (1UL << (8 * sizeof(gentry.size) - PAGE_SHIFT)) || size >= (1UL << (8 * sizeof(gentry.size) - PAGE_SHIFT))) gentry.base = gentry.size = gentry.type = 0; else { - gentry.base <<= PAGE_SHIFT; + gentry.base = base << PAGE_SHIFT; gentry.size = size << PAGE_SHIFT; gentry.type = type; } @@ -321,11 +322,12 @@ mtrr_ioctl(struct file *file, unsigned int cmd, unsigned long __arg) #endif if (gentry.regnum >= num_var_ranges) return -EINVAL; - mtrr_if->get(gentry.regnum, &gentry.base, &size, &type); + mtrr_if->get(gentry.regnum, &base, &size, &type); /* Hide entries that would overflow */ if (size != (__typeof__(gentry.size))size) gentry.base = gentry.size = gentry.type = 0; else { + gentry.base = base; gentry.size = size; gentry.type = type; } diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c index 6b96110bb0c..f961de9964c 100644 --- a/arch/x86/kernel/cpu/mtrr/main.c +++ b/arch/x86/kernel/cpu/mtrr/main.c @@ -51,9 +51,13 @@ #include <asm/e820.h> #include <asm/mtrr.h> #include <asm/msr.h> +#include <asm/pat.h> #include "mtrr.h" +/* arch_phys_wc_add returns an MTRR register index plus this offset. */ +#define MTRR_TO_PHYS_WC_OFFSET 1000 + u32 num_var_ranges; unsigned int mtrr_usage_table[MTRR_MAX_VAR_RANGES]; @@ -305,7 +309,8 @@ int mtrr_add_page(unsigned long base, unsigned long size, return -EINVAL; } - if (base & size_or_mask || size & size_or_mask) { + if ((base | (base + size - 1)) >> + (boot_cpu_data.x86_phys_bits - PAGE_SHIFT)) { pr_warning("mtrr: base or size exceeds the MTRR width\n"); return -EINVAL; } @@ -524,6 +529,73 @@ int mtrr_del(int reg, unsigned long base, unsigned long size) } EXPORT_SYMBOL(mtrr_del); +/** + * arch_phys_wc_add - add a WC MTRR and handle errors if PAT is unavailable + * @base: Physical base address + * @size: Size of region + * + * If PAT is available, this does nothing. If PAT is unavailable, it + * attempts to add a WC MTRR covering size bytes starting at base and + * logs an error if this fails. + * + * Drivers must store the return value to pass to mtrr_del_wc_if_needed, + * but drivers should not try to interpret that return value. + */ +int arch_phys_wc_add(unsigned long base, unsigned long size) +{ + int ret; + + if (pat_enabled) + return 0; /* Success! (We don't need to do anything.) */ + + ret = mtrr_add(base, size, MTRR_TYPE_WRCOMB, true); + if (ret < 0) { + pr_warn("Failed to add WC MTRR for [%p-%p]; performance may suffer.", + (void *)base, (void *)(base + size - 1)); + return ret; + } + return ret + MTRR_TO_PHYS_WC_OFFSET; +} +EXPORT_SYMBOL(arch_phys_wc_add); + +/* + * arch_phys_wc_del - undoes arch_phys_wc_add + * @handle: Return value from arch_phys_wc_add + * + * This cleans up after mtrr_add_wc_if_needed. + * + * The API guarantees that mtrr_del_wc_if_needed(error code) and + * mtrr_del_wc_if_needed(0) do nothing. + */ +void arch_phys_wc_del(int handle) +{ + if (handle >= 1) { + WARN_ON(handle < MTRR_TO_PHYS_WC_OFFSET); + mtrr_del(handle - MTRR_TO_PHYS_WC_OFFSET, 0, 0); + } +} +EXPORT_SYMBOL(arch_phys_wc_del); + +/* + * phys_wc_to_mtrr_index - translates arch_phys_wc_add's return value + * @handle: Return value from arch_phys_wc_add + * + * This will turn the return value from arch_phys_wc_add into an mtrr + * index suitable for debugging. + * + * Note: There is no legitimate use for this function, except possibly + * in printk line. Alas there is an illegitimate use in some ancient + * drm ioctls. + */ +int phys_wc_to_mtrr_index(int handle) +{ + if (handle < MTRR_TO_PHYS_WC_OFFSET) + return -1; + else + return handle - MTRR_TO_PHYS_WC_OFFSET; +} +EXPORT_SYMBOL_GPL(phys_wc_to_mtrr_index); + /* * HACK ALERT! * These should be called implicitly, but we can't yet until all the initcall @@ -583,6 +655,7 @@ static struct syscore_ops mtrr_syscore_ops = { int __initdata changed_by_mtrr_cleanup; +#define SIZE_OR_MASK_BITS(n) (~((1ULL << ((n) - PAGE_SHIFT)) - 1)) /** * mtrr_bp_init - initialize mtrrs on the boot CPU * @@ -600,13 +673,13 @@ void __init mtrr_bp_init(void) if (cpu_has_mtrr) { mtrr_if = &generic_mtrr_ops; - size_or_mask = 0xff000000; /* 36 bits */ + size_or_mask = SIZE_OR_MASK_BITS(36); size_and_mask = 0x00f00000; phys_addr = 36; /* * This is an AMD specific MSR, but we assume(hope?) that - * Intel will implement it to when they extend the address + * Intel will implement it too when they extend the address * bus of the Xeon. */ if (cpuid_eax(0x80000000) >= 0x80000008) { @@ -619,7 +692,7 @@ void __init mtrr_bp_init(void) boot_cpu_data.x86_mask == 0x4)) phys_addr = 36; - size_or_mask = ~((1ULL << (phys_addr - PAGE_SHIFT)) - 1); + size_or_mask = SIZE_OR_MASK_BITS(phys_addr); size_and_mask = ~size_or_mask & 0xfffff00000ULL; } else if (boot_cpu_data.x86_vendor == X86_VENDOR_CENTAUR && boot_cpu_data.x86 == 6) { @@ -627,7 +700,7 @@ void __init mtrr_bp_init(void) * VIA C* family have Intel style MTRRs, * but don't support PAE */ - size_or_mask = 0xfff00000; /* 32 bits */ + size_or_mask = SIZE_OR_MASK_BITS(32); size_and_mask = 0; phys_addr = 32; } @@ -637,21 +710,21 @@ void __init mtrr_bp_init(void) if (cpu_has_k6_mtrr) { /* Pre-Athlon (K6) AMD CPU MTRRs */ mtrr_if = mtrr_ops[X86_VENDOR_AMD]; - size_or_mask = 0xfff00000; /* 32 bits */ + size_or_mask = SIZE_OR_MASK_BITS(32); size_and_mask = 0; } break; case X86_VENDOR_CENTAUR: if (cpu_has_centaur_mcr) { mtrr_if = mtrr_ops[X86_VENDOR_CENTAUR]; - size_or_mask = 0xfff00000; /* 32 bits */ + size_or_mask = SIZE_OR_MASK_BITS(32); size_and_mask = 0; } break; case X86_VENDOR_CYRIX: if (cpu_has_cyrix_arr) { mtrr_if = mtrr_ops[X86_VENDOR_CYRIX]; - size_or_mask = 0xfff00000; /* 32 bits */ + size_or_mask = SIZE_OR_MASK_BITS(32); size_and_mask = 0; } break; @@ -695,11 +768,16 @@ void mtrr_ap_init(void) } /** - * Save current fixed-range MTRR state of the BSP + * Save current fixed-range MTRR state of the first cpu in cpu_online_mask. */ void mtrr_save_state(void) { - smp_call_function_single(0, mtrr_save_fixed_ranges, NULL, 1); + int first_cpu; + + get_online_cpus(); + first_cpu = cpumask_first(cpu_online_mask); + smp_call_function_single(first_cpu, mtrr_save_fixed_ranges, NULL, 1); + put_online_cpus(); } void set_mtrr_aps_delayed_init(void) diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 5adce1040b1..2879ecdaac4 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -24,27 +24,19 @@ #include <linux/slab.h> #include <linux/cpu.h> #include <linux/bitops.h> +#include <linux/device.h> #include <asm/apic.h> #include <asm/stacktrace.h> #include <asm/nmi.h> -#include <asm/compat.h> #include <asm/smp.h> #include <asm/alternative.h> +#include <asm/timer.h> +#include <asm/desc.h> +#include <asm/ldt.h> #include "perf_event.h" -#if 0 -#undef wrmsrl -#define wrmsrl(msr, val) \ -do { \ - trace_printk("wrmsrl(%lx, %lx)\n", (unsigned long)(msr),\ - (unsigned long)(val)); \ - native_write_msr((msr), (u32)((u64)(val)), \ - (u32)((u64)(val) >> 32)); \ -} while (0) -#endif - struct x86_pmu x86_pmu __read_mostly; DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events) = { @@ -73,7 +65,7 @@ u64 x86_perf_event_update(struct perf_event *event) int idx = hwc->idx; s64 delta; - if (idx == X86_PMC_IDX_FIXED_BTS) + if (idx == INTEL_PMC_IDX_FIXED_BTS) return 0; /* @@ -85,7 +77,7 @@ u64 x86_perf_event_update(struct perf_event *event) */ again: prev_raw_count = local64_read(&hwc->prev_count); - rdmsrl(hwc->event_base, new_raw_count); + rdpmcl(hwc->event_base_rdpmc, new_raw_count); if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, new_raw_count) != prev_raw_count) @@ -126,6 +118,9 @@ static int x86_pmu_extra_regs(u64 config, struct perf_event *event) continue; if (event->attr.config1 & ~er->valid_mask) return -EINVAL; + /* Check if the extra msrs can be safely accessed*/ + if (!er->extra_msr_access) + return -ENXIO; reg->idx = er->idx; reg->config = event->attr.config1; @@ -188,8 +183,9 @@ static void release_pmc_hardware(void) {} static bool check_hw_exists(void) { - u64 val, val_new = 0; - int i, reg, ret = 0; + u64 val, val_fail, val_new= ~0; + int i, reg, reg_fail, ret = 0; + int bios_fail = 0; /* * Check to see if the BIOS enabled any of the counters, if so @@ -200,8 +196,11 @@ static bool check_hw_exists(void) ret = rdmsrl_safe(reg, &val); if (ret) goto msr_fail; - if (val & ARCH_PERFMON_EVENTSEL_ENABLE) - goto bios_fail; + if (val & ARCH_PERFMON_EVENTSEL_ENABLE) { + bios_fail = 1; + val_fail = val; + reg_fail = reg; + } } if (x86_pmu.num_counters_fixed) { @@ -210,35 +209,41 @@ static bool check_hw_exists(void) if (ret) goto msr_fail; for (i = 0; i < x86_pmu.num_counters_fixed; i++) { - if (val & (0x03 << i*4)) - goto bios_fail; + if (val & (0x03 << i*4)) { + bios_fail = 1; + val_fail = val; + reg_fail = reg; + } } } /* - * Now write a value and read it back to see if it matches, - * this is needed to detect certain hardware emulators (qemu/kvm) - * that don't trap on the MSR access and always return 0s. + * Read the current value, change it and read it back to see if it + * matches, this is needed to detect certain hardware emulators + * (qemu/kvm) that don't trap on the MSR access and always return 0s. */ - val = 0xabcdUL; - ret = checking_wrmsrl(x86_pmu_event_addr(0), val); - ret |= rdmsrl_safe(x86_pmu_event_addr(0), &val_new); + reg = x86_pmu_event_addr(0); + if (rdmsrl_safe(reg, &val)) + goto msr_fail; + val ^= 0xffffUL; + ret = wrmsrl_safe(reg, val); + ret |= rdmsrl_safe(reg, &val_new); if (ret || val != val_new) goto msr_fail; - return true; - -bios_fail: /* * We still allow the PMU driver to operate: */ - printk(KERN_CONT "Broken BIOS detected, complain to your hardware vendor.\n"); - printk(KERN_ERR FW_BUG "the BIOS has corrupted hw-PMU resources (MSR %x is %Lx)\n", reg, val); + if (bios_fail) { + printk(KERN_CONT "Broken BIOS detected, complain to your hardware vendor.\n"); + printk(KERN_ERR FW_BUG "the BIOS has corrupted hw-PMU resources (MSR %x is %Lx)\n", reg_fail, val_fail); + } return true; msr_fail: printk(KERN_CONT "Broken PMU hardware detected, using software events only.\n"); + printk(KERN_ERR "Failed to access perfctr msr (MSR %x is %Lx)\n", reg, val_new); return false; } @@ -301,15 +306,6 @@ int x86_setup_perfctr(struct perf_event *event) hwc->sample_period = x86_pmu.max_period; hwc->last_period = hwc->sample_period; local64_set(&hwc->period_left, hwc->sample_period); - } else { - /* - * If we have a PMU initialized but no APIC - * interrupts, we cannot sample hardware - * events (user-space has to fall back and - * sample via a hrtimer based software event): - */ - if (!x86_pmu.apic) - return -EOPNOTSUPP; } if (attr->type == PERF_TYPE_RAW) @@ -351,13 +347,43 @@ int x86_setup_perfctr(struct perf_event *event) return 0; } +/* + * check that branch_sample_type is compatible with + * settings needed for precise_ip > 1 which implies + * using the LBR to capture ALL taken branches at the + * priv levels of the measurement + */ +static inline int precise_br_compat(struct perf_event *event) +{ + u64 m = event->attr.branch_sample_type; + u64 b = 0; + + /* must capture all branches */ + if (!(m & PERF_SAMPLE_BRANCH_ANY)) + return 0; + + m &= PERF_SAMPLE_BRANCH_KERNEL | PERF_SAMPLE_BRANCH_USER; + + if (!event->attr.exclude_user) + b |= PERF_SAMPLE_BRANCH_USER; + + if (!event->attr.exclude_kernel) + b |= PERF_SAMPLE_BRANCH_KERNEL; + + /* + * ignore PERF_SAMPLE_BRANCH_HV, not supported on x86 + */ + + return m == b; +} + int x86_pmu_hw_config(struct perf_event *event) { if (event->attr.precise_ip) { int precise = 0; /* Support for constant skid */ - if (x86_pmu.pebs_active) { + if (x86_pmu.pebs_active && !x86_pmu.pebs_broken) { precise++; /* Support for IP fixup */ @@ -367,6 +393,37 @@ int x86_pmu_hw_config(struct perf_event *event) if (event->attr.precise_ip > precise) return -EOPNOTSUPP; + /* + * check that PEBS LBR correction does not conflict with + * whatever the user is asking with attr->branch_sample_type + */ + if (event->attr.precise_ip > 1 && + x86_pmu.intel_cap.pebs_format < 2) { + u64 *br_type = &event->attr.branch_sample_type; + + if (has_branch_stack(event)) { + if (!precise_br_compat(event)) + return -EOPNOTSUPP; + + /* branch_sample_type is compatible */ + + } else { + /* + * user did not specify branch_sample_type + * + * For PEBS fixups, we capture all + * the branches at the priv level of the + * event. + */ + *br_type = PERF_SAMPLE_BRANCH_ANY; + + if (!event->attr.exclude_user) + *br_type |= PERF_SAMPLE_BRANCH_USER; + + if (!event->attr.exclude_kernel) + *br_type |= PERF_SAMPLE_BRANCH_KERNEL; + } + } } /* @@ -423,6 +480,7 @@ static int __x86_pmu_event_init(struct perf_event *event) /* mark unused */ event->hw.extra_reg.idx = EXTRA_REG_NONE; + event->hw.branch_reg.idx = EXTRA_REG_NONE; return x86_pmu.hw_config(event); } @@ -505,7 +563,7 @@ struct sched_state { struct perf_sched { int max_weight; int max_events; - struct event_constraint **constraints; + struct perf_event **events; struct sched_state state; int saved_states; struct sched_state saved[SCHED_STATES_MAX]; @@ -514,7 +572,7 @@ struct perf_sched { /* * Initialize interator that runs through all events and counters. */ -static void perf_sched_init(struct perf_sched *sched, struct event_constraint **c, +static void perf_sched_init(struct perf_sched *sched, struct perf_event **events, int num, int wmin, int wmax) { int idx; @@ -522,10 +580,10 @@ static void perf_sched_init(struct perf_sched *sched, struct event_constraint ** memset(sched, 0, sizeof(*sched)); sched->max_events = num; sched->max_weight = wmax; - sched->constraints = c; + sched->events = events; for (idx = 0; idx < num; idx++) { - if (c[idx]->weight == wmin) + if (events[idx]->hw.constraint->weight == wmin) break; } @@ -572,19 +630,18 @@ static bool __perf_sched_find_counter(struct perf_sched *sched) if (sched->state.event >= sched->max_events) return false; - c = sched->constraints[sched->state.event]; - + c = sched->events[sched->state.event]->hw.constraint; /* Prefer fixed purpose counters */ - if (x86_pmu.num_counters_fixed) { - idx = X86_PMC_IDX_FIXED; - for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_MAX) { + if (c->idxmsk64 & (~0ULL << INTEL_PMC_IDX_FIXED)) { + idx = INTEL_PMC_IDX_FIXED; + for_each_set_bit_from(idx, c->idxmsk, X86_PMC_IDX_MAX) { if (!__test_and_set_bit(idx, sched->state.used)) goto done; } } /* Grab the first unused counter starting with idx */ idx = sched->state.counter; - for_each_set_bit_cont(idx, c->idxmsk, X86_PMC_IDX_FIXED) { + for_each_set_bit_from(idx, c->idxmsk, INTEL_PMC_IDX_FIXED) { if (!__test_and_set_bit(idx, sched->state.used)) goto done; } @@ -631,7 +688,7 @@ static bool perf_sched_next_event(struct perf_sched *sched) if (sched->state.weight > sched->max_weight) return false; } - c = sched->constraints[sched->state.event]; + c = sched->events[sched->state.event]->hw.constraint; } while (c->weight != sched->state.weight); sched->state.counter = 0; /* start with first counter */ @@ -642,12 +699,12 @@ static bool perf_sched_next_event(struct perf_sched *sched) /* * Assign a counter for each event. */ -static int perf_assign_events(struct event_constraint **constraints, int n, - int wmin, int wmax, int *assign) +int perf_assign_events(struct perf_event **events, int n, + int wmin, int wmax, int *assign) { struct perf_sched sched; - perf_sched_init(&sched, constraints, n, wmin, wmax); + perf_sched_init(&sched, events, n, wmin, wmax); do { if (!perf_sched_find_counter(&sched)) @@ -658,19 +715,23 @@ static int perf_assign_events(struct event_constraint **constraints, int n, return sched.state.unassigned; } +EXPORT_SYMBOL_GPL(perf_assign_events); int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign) { - struct event_constraint *c, *constraints[X86_PMC_IDX_MAX]; + struct event_constraint *c; unsigned long used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; + struct perf_event *e; int i, wmin, wmax, num = 0; struct hw_perf_event *hwc; bitmap_zero(used_mask, X86_PMC_IDX_MAX); for (i = 0, wmin = X86_PMC_IDX_MAX, wmax = 0; i < n; i++) { + hwc = &cpuc->event_list[i]->hw; c = x86_pmu.get_event_constraints(cpuc, cpuc->event_list[i]); - constraints[i] = c; + hwc->constraint = c; + wmin = min(wmin, c->weight); wmax = max(wmax, c->weight); } @@ -680,7 +741,7 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign) */ for (i = 0; i < n; i++) { hwc = &cpuc->event_list[i]->hw; - c = constraints[i]; + c = hwc->constraint; /* never assigned */ if (hwc->idx == -1) @@ -701,16 +762,35 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign) /* slow path */ if (i != n) - num = perf_assign_events(constraints, n, wmin, wmax, assign); + num = perf_assign_events(cpuc->event_list, n, wmin, + wmax, assign); /* + * Mark the event as committed, so we do not put_constraint() + * in case new events are added and fail scheduling. + */ + if (!num && assign) { + for (i = 0; i < n; i++) { + e = cpuc->event_list[i]; + e->hw.flags |= PERF_X86_EVENT_COMMITTED; + } + } + /* * scheduling failed or is just a simulation, * free resources if necessary */ if (!assign || num) { for (i = 0; i < n; i++) { + e = cpuc->event_list[i]; + /* + * do not put_constraint() on comitted events, + * because they are good to go + */ + if ((e->hw.flags & PERF_X86_EVENT_COMMITTED)) + continue; + if (x86_pmu.put_event_constraints) - x86_pmu.put_event_constraints(cpuc, cpuc->event_list[i]); + x86_pmu.put_event_constraints(cpuc, e); } } return num ? -EINVAL : 0; @@ -762,15 +842,17 @@ static inline void x86_assign_hw_event(struct perf_event *event, hwc->last_cpu = smp_processor_id(); hwc->last_tag = ++cpuc->tags[i]; - if (hwc->idx == X86_PMC_IDX_FIXED_BTS) { + if (hwc->idx == INTEL_PMC_IDX_FIXED_BTS) { hwc->config_base = 0; hwc->event_base = 0; - } else if (hwc->idx >= X86_PMC_IDX_FIXED) { + } else if (hwc->idx >= INTEL_PMC_IDX_FIXED) { hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; - hwc->event_base = MSR_ARCH_PERFMON_FIXED_CTR0 + (hwc->idx - X86_PMC_IDX_FIXED); + hwc->event_base = MSR_ARCH_PERFMON_FIXED_CTR0 + (hwc->idx - INTEL_PMC_IDX_FIXED); + hwc->event_base_rdpmc = (hwc->idx - INTEL_PMC_IDX_FIXED) | 1<<30; } else { hwc->config_base = x86_pmu_config_addr(hwc->idx); hwc->event_base = x86_pmu_event_addr(hwc->idx); + hwc->event_base_rdpmc = x86_pmu_rdpmc_index(hwc->idx); } } @@ -805,7 +887,6 @@ static void x86_pmu_enable(struct pmu *pmu) * hw_perf_group_sched_in() or x86_pmu_enable() * * step1: save events moving to new counters - * step2: reprogram moved events into new counters */ for (i = 0; i < n_running; i++) { event = cpuc->event_list[i]; @@ -831,6 +912,9 @@ static void x86_pmu_enable(struct pmu *pmu) x86_pmu_stop(event, PERF_EF_UPDATE); } + /* + * step2: reprogram moved events into new counters + */ for (i = 0; i < cpuc->n_events; i++) { event = cpuc->event_list[i]; hwc = &event->hw; @@ -868,7 +952,7 @@ int x86_perf_event_set_period(struct perf_event *event) s64 period = hwc->sample_period; int ret = 0, idx = hwc->idx; - if (idx == X86_PMC_IDX_FIXED_BTS) + if (idx == INTEL_PMC_IDX_FIXED_BTS) return 0; /* @@ -956,7 +1040,7 @@ static int x86_pmu_add(struct perf_event *event, int flags) /* * If group events scheduling transaction was started, * skip the schedulability test here, it will be performed - * at commit time (->commit_txn) as a whole + * at commit time (->commit_txn) as a whole. */ if (cpuc->group_flag & PERF_EVENT_TXN) goto done_collect; @@ -971,6 +1055,10 @@ static int x86_pmu_add(struct perf_event *event, int flags) memcpy(cpuc->assign, assign, n*sizeof(int)); done_collect: + /* + * Commit the collect_events() state. See x86_pmu_del() and + * x86_pmu_*_txn(). + */ cpuc->n_events = n; cpuc->n_added += n - n0; cpuc->n_txn += n - n0; @@ -1088,28 +1176,46 @@ static void x86_pmu_del(struct perf_event *event, int flags) int i; /* + * event is descheduled + */ + event->hw.flags &= ~PERF_X86_EVENT_COMMITTED; + + /* * If we're called during a txn, we don't need to do anything. * The events never got scheduled and ->cancel_txn will truncate * the event_list. + * + * XXX assumes any ->del() called during a TXN will only be on + * an event added during that same TXN. */ if (cpuc->group_flag & PERF_EVENT_TXN) return; + /* + * Not a TXN, therefore cleanup properly. + */ x86_pmu_stop(event, PERF_EF_UPDATE); for (i = 0; i < cpuc->n_events; i++) { - if (event == cpuc->event_list[i]) { + if (event == cpuc->event_list[i]) + break; + } - if (x86_pmu.put_event_constraints) - x86_pmu.put_event_constraints(cpuc, event); + if (WARN_ON_ONCE(i == cpuc->n_events)) /* called ->del() without ->add() ? */ + return; - while (++i < cpuc->n_events) - cpuc->event_list[i-1] = cpuc->event_list[i]; + /* If we have a newly added event; make sure to decrease n_added. */ + if (i >= cpuc->n_events - cpuc->n_added) + --cpuc->n_added; + + if (x86_pmu.put_event_constraints) + x86_pmu.put_event_constraints(cpuc, event); + + /* Delete the array entry. */ + while (++i < cpuc->n_events) + cpuc->event_list[i-1] = cpuc->event_list[i]; + --cpuc->n_events; - --cpuc->n_events; - break; - } - } perf_event_update_userpage(event); } @@ -1121,8 +1227,6 @@ int x86_pmu_handle_irq(struct pt_regs *regs) int idx, handled = 0; u64 val; - perf_sample_data_init(&data, 0); - cpuc = &__get_cpu_var(cpu_hw_events); /* @@ -1157,7 +1261,7 @@ int x86_pmu_handle_irq(struct pt_regs *regs) * event overflow */ handled++; - data.period = event->hw.last_period; + perf_sample_data_init(&data, 0, event->hw.last_period); if (!x86_perf_event_set_period(event)) continue; @@ -1183,19 +1287,30 @@ void perf_events_lapic_init(void) apic_write(APIC_LVTPC, APIC_DM_NMI); } -static int __kprobes +static int perf_event_nmi_handler(unsigned int cmd, struct pt_regs *regs) { + u64 start_clock; + u64 finish_clock; + int ret; + if (!atomic_read(&active_events)) return NMI_DONE; - return x86_pmu.handle_irq(regs); + start_clock = sched_clock(); + ret = x86_pmu.handle_irq(regs); + finish_clock = sched_clock(); + + perf_sample_event_took(finish_clock - start_clock); + + return ret; } +NOKPROBE_SYMBOL(perf_event_nmi_handler); struct event_constraint emptyconstraint; struct event_constraint unconstrained; -static int __cpuinit +static int x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu) { unsigned int cpu = (long)hcpu; @@ -1210,6 +1325,8 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu) break; case CPU_STARTING: + if (x86_pmu.attr_rdpmc) + set_in_cr4(X86_CR4_PCE); if (x86_pmu.cpu_starting) x86_pmu.cpu_starting(cpu); break; @@ -1244,12 +1361,163 @@ static void __init pmu_check_apic(void) x86_pmu.apic = 0; pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n"); pr_info("no hardware sampling interrupt available.\n"); + + /* + * If we have a PMU initialized but no APIC + * interrupts, we cannot sample hardware + * events (user-space has to fall back and + * sample via a hrtimer based software event): + */ + pmu.capabilities |= PERF_PMU_CAP_NO_INTERRUPT; + +} + +static struct attribute_group x86_pmu_format_group = { + .name = "format", + .attrs = NULL, +}; + +/* + * Remove all undefined events (x86_pmu.event_map(id) == 0) + * out of events_attr attributes. + */ +static void __init filter_events(struct attribute **attrs) +{ + struct device_attribute *d; + struct perf_pmu_events_attr *pmu_attr; + int i, j; + + for (i = 0; attrs[i]; i++) { + d = (struct device_attribute *)attrs[i]; + pmu_attr = container_of(d, struct perf_pmu_events_attr, attr); + /* str trumps id */ + if (pmu_attr->event_str) + continue; + if (x86_pmu.event_map(i)) + continue; + + for (j = i; attrs[j]; j++) + attrs[j] = attrs[j + 1]; + + /* Check the shifted attr. */ + i--; + } +} + +/* Merge two pointer arrays */ +static __init struct attribute **merge_attr(struct attribute **a, struct attribute **b) +{ + struct attribute **new; + int j, i; + + for (j = 0; a[j]; j++) + ; + for (i = 0; b[i]; i++) + j++; + j++; + + new = kmalloc(sizeof(struct attribute *) * j, GFP_KERNEL); + if (!new) + return NULL; + + j = 0; + for (i = 0; a[i]; i++) + new[j++] = a[i]; + for (i = 0; b[i]; i++) + new[j++] = b[i]; + new[j] = NULL; + + return new; +} + +ssize_t events_sysfs_show(struct device *dev, struct device_attribute *attr, + char *page) +{ + struct perf_pmu_events_attr *pmu_attr = \ + container_of(attr, struct perf_pmu_events_attr, attr); + u64 config = x86_pmu.event_map(pmu_attr->id); + + /* string trumps id */ + if (pmu_attr->event_str) + return sprintf(page, "%s", pmu_attr->event_str); + + return x86_pmu.events_sysfs_show(page, config); +} + +EVENT_ATTR(cpu-cycles, CPU_CYCLES ); +EVENT_ATTR(instructions, INSTRUCTIONS ); +EVENT_ATTR(cache-references, CACHE_REFERENCES ); +EVENT_ATTR(cache-misses, CACHE_MISSES ); +EVENT_ATTR(branch-instructions, BRANCH_INSTRUCTIONS ); +EVENT_ATTR(branch-misses, BRANCH_MISSES ); +EVENT_ATTR(bus-cycles, BUS_CYCLES ); +EVENT_ATTR(stalled-cycles-frontend, STALLED_CYCLES_FRONTEND ); +EVENT_ATTR(stalled-cycles-backend, STALLED_CYCLES_BACKEND ); +EVENT_ATTR(ref-cycles, REF_CPU_CYCLES ); + +static struct attribute *empty_attrs; + +static struct attribute *events_attr[] = { + EVENT_PTR(CPU_CYCLES), + EVENT_PTR(INSTRUCTIONS), + EVENT_PTR(CACHE_REFERENCES), + EVENT_PTR(CACHE_MISSES), + EVENT_PTR(BRANCH_INSTRUCTIONS), + EVENT_PTR(BRANCH_MISSES), + EVENT_PTR(BUS_CYCLES), + EVENT_PTR(STALLED_CYCLES_FRONTEND), + EVENT_PTR(STALLED_CYCLES_BACKEND), + EVENT_PTR(REF_CPU_CYCLES), + NULL, +}; + +static struct attribute_group x86_pmu_events_group = { + .name = "events", + .attrs = events_attr, +}; + +ssize_t x86_event_sysfs_show(char *page, u64 config, u64 event) +{ + u64 umask = (config & ARCH_PERFMON_EVENTSEL_UMASK) >> 8; + u64 cmask = (config & ARCH_PERFMON_EVENTSEL_CMASK) >> 24; + bool edge = (config & ARCH_PERFMON_EVENTSEL_EDGE); + bool pc = (config & ARCH_PERFMON_EVENTSEL_PIN_CONTROL); + bool any = (config & ARCH_PERFMON_EVENTSEL_ANY); + bool inv = (config & ARCH_PERFMON_EVENTSEL_INV); + ssize_t ret; + + /* + * We have whole page size to spend and just little data + * to write, so we can safely use sprintf. + */ + ret = sprintf(page, "event=0x%02llx", event); + + if (umask) + ret += sprintf(page + ret, ",umask=0x%02llx", umask); + + if (edge) + ret += sprintf(page + ret, ",edge"); + + if (pc) + ret += sprintf(page + ret, ",pc"); + + if (any) + ret += sprintf(page + ret, ",any"); + + if (inv) + ret += sprintf(page + ret, ",inv"); + + if (cmask) + ret += sprintf(page + ret, ",cmask=0x%02llx", cmask); + + ret += sprintf(page + ret, "\n"); + + return ret; } static int __init init_hw_perf_events(void) { struct x86_pmu_quirk *quirk; - struct event_constraint *c; int err; pr_info("Performance Events: "); @@ -1262,7 +1530,7 @@ static int __init init_hw_perf_events(void) err = amd_pmu_init(); break; default: - return 0; + err = -ENOTSUPP; } if (err != 0) { pr_cont("no PMU driver, software events only.\n"); @@ -1277,46 +1545,37 @@ static int __init init_hw_perf_events(void) pr_cont("%s PMU driver.\n", x86_pmu.name); + x86_pmu.attr_rdpmc = 1; /* enable userspace RDPMC usage by default */ + for (quirk = x86_pmu.quirks; quirk; quirk = quirk->next) quirk->func(); - if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { - WARN(1, KERN_ERR "hw perf events %d > max(%d), clipping!", - x86_pmu.num_counters, X86_PMC_MAX_GENERIC); - x86_pmu.num_counters = X86_PMC_MAX_GENERIC; - } - x86_pmu.intel_ctrl = (1 << x86_pmu.num_counters) - 1; - - if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) { - WARN(1, KERN_ERR "hw perf events fixed %d > max(%d), clipping!", - x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED); - x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED; - } - - x86_pmu.intel_ctrl |= - ((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED; + if (!x86_pmu.intel_ctrl) + x86_pmu.intel_ctrl = (1 << x86_pmu.num_counters) - 1; perf_events_lapic_init(); register_nmi_handler(NMI_LOCAL, perf_event_nmi_handler, 0, "PMI"); unconstrained = (struct event_constraint) __EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_counters) - 1, - 0, x86_pmu.num_counters, 0); + 0, x86_pmu.num_counters, 0, 0); - if (x86_pmu.event_constraints) { - /* - * event on fixed counter2 (REF_CYCLES) only works on this - * counter, so do not extend mask to generic counters - */ - for_each_event_constraint(c, x86_pmu.event_constraints) { - if (c->cmask != X86_RAW_EVENT_MASK - || c->idxmsk64 == X86_PMC_MSK_FIXED_REF_CYCLES) { - continue; - } + x86_pmu_format_group.attrs = x86_pmu.format_attrs; - c->idxmsk64 |= (1ULL << x86_pmu.num_counters) - 1; - c->weight += x86_pmu.num_counters; - } + if (x86_pmu.event_attrs) + x86_pmu_events_group.attrs = x86_pmu.event_attrs; + + if (!x86_pmu.events_sysfs_show) + x86_pmu_events_group.attrs = &empty_attrs; + else + filter_events(x86_pmu_events_group.attrs); + + if (x86_pmu.cpu_events) { + struct attribute **tmp; + + tmp = merge_attr(x86_pmu_events_group.attrs, x86_pmu.cpu_events); + if (!WARN_ON(!tmp)) + x86_pmu_events_group.attrs = tmp; } pr_info("... version: %d\n", x86_pmu.version); @@ -1360,7 +1619,8 @@ static void x86_pmu_cancel_txn(struct pmu *pmu) { __this_cpu_and(cpu_hw_events.group_flag, ~PERF_EVENT_TXN); /* - * Truncate the collected events. + * Truncate collected array by the number of events added in this + * transaction. See x86_pmu_add() and x86_pmu_*_txn(). */ __this_cpu_sub(cpu_hw_events.n_added, __this_cpu_read(cpu_hw_events.n_txn)); __this_cpu_sub(cpu_hw_events.n_events, __this_cpu_read(cpu_hw_events.n_txn)); @@ -1371,6 +1631,8 @@ static void x86_pmu_cancel_txn(struct pmu *pmu) * Commit group events scheduling transaction * Perform the group schedulability test as a whole * Return 0 if success + * + * Does not cancel the transaction on failure; expects the caller to do this. */ static int x86_pmu_commit_txn(struct pmu *pmu) { @@ -1426,6 +1688,7 @@ static struct cpu_hw_events *allocate_fake_cpuc(void) if (!cpuc->shared_regs) goto error; } + cpuc->is_fake = 1; return cpuc; error: free_fake_cpuc(cpuc); @@ -1542,23 +1805,138 @@ static int x86_pmu_event_init(struct perf_event *event) return err; } +static int x86_pmu_event_idx(struct perf_event *event) +{ + int idx = event->hw.idx; + + if (!x86_pmu.attr_rdpmc) + return 0; + + if (x86_pmu.num_counters_fixed && idx >= INTEL_PMC_IDX_FIXED) { + idx -= INTEL_PMC_IDX_FIXED; + idx |= 1 << 30; + } + + return idx + 1; +} + +static ssize_t get_attr_rdpmc(struct device *cdev, + struct device_attribute *attr, + char *buf) +{ + return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc); +} + +static void change_rdpmc(void *info) +{ + bool enable = !!(unsigned long)info; + + if (enable) + set_in_cr4(X86_CR4_PCE); + else + clear_in_cr4(X86_CR4_PCE); +} + +static ssize_t set_attr_rdpmc(struct device *cdev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + unsigned long val; + ssize_t ret; + + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; + + if (x86_pmu.attr_rdpmc_broken) + return -ENOTSUPP; + + if (!!val != !!x86_pmu.attr_rdpmc) { + x86_pmu.attr_rdpmc = !!val; + on_each_cpu(change_rdpmc, (void *)val, 1); + } + + return count; +} + +static DEVICE_ATTR(rdpmc, S_IRUSR | S_IWUSR, get_attr_rdpmc, set_attr_rdpmc); + +static struct attribute *x86_pmu_attrs[] = { + &dev_attr_rdpmc.attr, + NULL, +}; + +static struct attribute_group x86_pmu_attr_group = { + .attrs = x86_pmu_attrs, +}; + +static const struct attribute_group *x86_pmu_attr_groups[] = { + &x86_pmu_attr_group, + &x86_pmu_format_group, + &x86_pmu_events_group, + NULL, +}; + +static void x86_pmu_flush_branch_stack(void) +{ + if (x86_pmu.flush_branch_stack) + x86_pmu.flush_branch_stack(); +} + +void perf_check_microcode(void) +{ + if (x86_pmu.check_microcode) + x86_pmu.check_microcode(); +} +EXPORT_SYMBOL_GPL(perf_check_microcode); + static struct pmu pmu = { - .pmu_enable = x86_pmu_enable, - .pmu_disable = x86_pmu_disable, + .pmu_enable = x86_pmu_enable, + .pmu_disable = x86_pmu_disable, + + .attr_groups = x86_pmu_attr_groups, + + .event_init = x86_pmu_event_init, - .event_init = x86_pmu_event_init, + .add = x86_pmu_add, + .del = x86_pmu_del, + .start = x86_pmu_start, + .stop = x86_pmu_stop, + .read = x86_pmu_read, - .add = x86_pmu_add, - .del = x86_pmu_del, - .start = x86_pmu_start, - .stop = x86_pmu_stop, - .read = x86_pmu_read, + .start_txn = x86_pmu_start_txn, + .cancel_txn = x86_pmu_cancel_txn, + .commit_txn = x86_pmu_commit_txn, - .start_txn = x86_pmu_start_txn, - .cancel_txn = x86_pmu_cancel_txn, - .commit_txn = x86_pmu_commit_txn, + .event_idx = x86_pmu_event_idx, + .flush_branch_stack = x86_pmu_flush_branch_stack, }; +void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now) +{ + struct cyc2ns_data *data; + + userpg->cap_user_time = 0; + userpg->cap_user_time_zero = 0; + userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc; + userpg->pmc_width = x86_pmu.cntval_bits; + + if (!sched_clock_stable()) + return; + + data = cyc2ns_read_begin(); + + userpg->cap_user_time = 1; + userpg->time_mult = data->cyc2ns_mul; + userpg->time_shift = data->cyc2ns_shift; + userpg->time_offset = data->cyc2ns_offset - now; + + userpg->cap_user_time_zero = 1; + userpg->time_zero = data->cyc2ns_offset; + + cyc2ns_read_end(data); +} + /* * callchain support */ @@ -1594,32 +1972,68 @@ perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs) dump_trace(NULL, regs, NULL, 0, &backtrace_ops, entry); } +static inline int +valid_user_frame(const void __user *fp, unsigned long size) +{ + return (__range_not_ok(fp, size, TASK_SIZE) == 0); +} + +static unsigned long get_segment_base(unsigned int segment) +{ + struct desc_struct *desc; + int idx = segment >> 3; + + if ((segment & SEGMENT_TI_MASK) == SEGMENT_LDT) { + if (idx > LDT_ENTRIES) + return 0; + + if (idx > current->active_mm->context.size) + return 0; + + desc = current->active_mm->context.ldt; + } else { + if (idx > GDT_ENTRIES) + return 0; + + desc = __this_cpu_ptr(&gdt_page.gdt[0]); + } + + return get_desc_base(desc + idx); +} + #ifdef CONFIG_COMPAT + +#include <asm/compat.h> + static inline int perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry) { /* 32-bit process in 64-bit kernel. */ + unsigned long ss_base, cs_base; struct stack_frame_ia32 frame; const void __user *fp; if (!test_thread_flag(TIF_IA32)) return 0; - fp = compat_ptr(regs->bp); + cs_base = get_segment_base(regs->cs); + ss_base = get_segment_base(regs->ss); + + fp = compat_ptr(ss_base + regs->bp); while (entry->nr < PERF_MAX_STACK_DEPTH) { unsigned long bytes; frame.next_frame = 0; frame.return_address = 0; bytes = copy_from_user_nmi(&frame, fp, sizeof(frame)); - if (bytes != sizeof(frame)) + if (bytes != 0) break; - if (fp < compat_ptr(regs->sp)) + if (!valid_user_frame(fp, sizeof(frame))) break; - perf_callchain_store(entry, frame.return_address); - fp = compat_ptr(frame.next_frame); + perf_callchain_store(entry, cs_base + frame.return_address); + fp = compat_ptr(ss_base + frame.next_frame); } return 1; } @@ -1642,6 +2056,12 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs) return; } + /* + * We don't know what to do with VM86 stacks.. ignore them for now. + */ + if (regs->flags & (X86_VM_MASK | PERF_EFLAGS_VM)) + return; + fp = (void __user *)regs->bp; perf_callchain_store(entry, regs->ip); @@ -1658,10 +2078,10 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs) frame.return_address = 0; bytes = copy_from_user_nmi(&frame, fp, sizeof(frame)); - if (bytes != sizeof(frame)) + if (bytes != 0) break; - if ((unsigned long)fp < regs->sp) + if (!valid_user_frame(fp, sizeof(frame))) break; perf_callchain_store(entry, frame.return_address); @@ -1669,16 +2089,50 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs) } } -unsigned long perf_instruction_pointer(struct pt_regs *regs) +/* + * Deal with code segment offsets for the various execution modes: + * + * VM86 - the good olde 16 bit days, where the linear address is + * 20 bits and we use regs->ip + 0x10 * regs->cs. + * + * IA32 - Where we need to look at GDT/LDT segment descriptor tables + * to figure out what the 32bit base address is. + * + * X32 - has TIF_X32 set, but is running in x86_64 + * + * X86_64 - CS,DS,SS,ES are all zero based. + */ +static unsigned long code_segment_base(struct pt_regs *regs) { - unsigned long ip; + /* + * If we are in VM86 mode, add the segment offset to convert to a + * linear address. + */ + if (regs->flags & X86_VM_MASK) + return 0x10 * regs->cs; + + /* + * For IA32 we look at the GDT/LDT segment base to convert the + * effective IP to a linear address. + */ +#ifdef CONFIG_X86_32 + if (user_mode(regs) && regs->cs != __USER_CS) + return get_segment_base(regs->cs); +#else + if (test_thread_flag(TIF_IA32)) { + if (user_mode(regs) && regs->cs != __USER32_CS) + return get_segment_base(regs->cs); + } +#endif + return 0; +} +unsigned long perf_instruction_pointer(struct pt_regs *regs) +{ if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) - ip = perf_guest_cbs->get_guest_ip(); - else - ip = instruction_pointer(regs); + return perf_guest_cbs->get_guest_ip(); - return ip; + return regs->ip + code_segment_base(regs); } unsigned long perf_misc_flags(struct pt_regs *regs) diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h index c30c807ddc7..8ade93111e0 100644 --- a/arch/x86/kernel/cpu/perf_event.h +++ b/arch/x86/kernel/cpu/perf_event.h @@ -14,6 +14,18 @@ #include <linux/perf_event.h> +#if 0 +#undef wrmsrl +#define wrmsrl(msr, val) \ +do { \ + unsigned int _msr = (msr); \ + u64 _val = (val); \ + trace_printk("wrmsrl(%x, %Lx)\n", (unsigned int)(_msr), \ + (unsigned long long)(_val)); \ + native_write_msr((_msr), (u32)(_val), (u32)(_val >> 32)); \ +} while (0) +#endif + /* * | NHM/WSM | SNB | * register ------------------------------- @@ -33,6 +45,8 @@ enum extra_reg_type { EXTRA_REG_RSP_0 = 0, /* offcore_response_0 */ EXTRA_REG_RSP_1 = 1, /* offcore_response_1 */ + EXTRA_REG_LBR = 2, /* lbr_select */ + EXTRA_REG_LDLAT = 3, /* ld_lat_threshold */ EXTRA_REG_MAX /* number of entries needed */ }; @@ -46,7 +60,15 @@ struct event_constraint { u64 cmask; int weight; int overlap; + int flags; }; +/* + * struct hw_perf_event.flags flags + */ +#define PERF_X86_EVENT_PEBS_LDLAT 0x1 /* ld+ldlat data address sampling */ +#define PERF_X86_EVENT_PEBS_ST 0x2 /* st data address sampling */ +#define PERF_X86_EVENT_PEBS_ST_HSW 0x4 /* haswell style st data sampling */ +#define PERF_X86_EVENT_COMMITTED 0x8 /* event passed commit_txn */ struct amd_nb { int nb_id; /* NorthBridge id */ @@ -56,7 +78,7 @@ struct amd_nb { }; /* The maximal number of PEBS events: */ -#define MAX_PEBS_EVENTS 4 +#define MAX_PEBS_EVENTS 8 /* * A debug store configuration. @@ -108,14 +130,17 @@ struct cpu_hw_events { unsigned long running[BITS_TO_LONGS(X86_PMC_IDX_MAX)]; int enabled; - int n_events; - int n_added; - int n_txn; + int n_events; /* the # of events in the below arrays */ + int n_added; /* the # last events in the below arrays; + they've never been enabled yet */ + int n_txn; /* the # last events in the below arrays; + added in the current transaction */ int assign[X86_PMC_IDX_MAX]; /* event to counter assignment */ u64 tags[X86_PMC_IDX_MAX]; struct perf_event *event_list[X86_PMC_IDX_MAX]; /* in enabled order */ unsigned int group_flag; + int is_fake; /* * Intel DebugStore bits @@ -130,6 +155,8 @@ struct cpu_hw_events { void *lbr_context; struct perf_branch_stack lbr_stack; struct perf_branch_entry lbr_entries[MAX_LBR_ENTRIES]; + struct er_account *lbr_sel; + u64 br_sel; /* * Intel host/guest exclude bits @@ -139,6 +166,11 @@ struct cpu_hw_events { struct perf_guest_switch_msr guest_switch_msrs[X86_PMC_IDX_MAX]; /* + * Intel checkpoint mask + */ + u64 intel_cp_status; + + /* * manage shared (per-core, per-cpu) registers * used on Intel NHM/WSM/SNB */ @@ -154,16 +186,17 @@ struct cpu_hw_events { void *kfree_on_online; }; -#define __EVENT_CONSTRAINT(c, n, m, w, o) {\ +#define __EVENT_CONSTRAINT(c, n, m, w, o, f) {\ { .idxmsk64 = (n) }, \ .code = (c), \ .cmask = (m), \ .weight = (w), \ .overlap = (o), \ + .flags = f, \ } #define EVENT_CONSTRAINT(c, n, m) \ - __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n), 0) + __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n), 0, 0) /* * The overlap flag marks event constraints with overlapping counter @@ -187,7 +220,7 @@ struct cpu_hw_events { * and its counter masks must be kept at a minimum. */ #define EVENT_CONSTRAINT_OVERLAP(c, n, m) \ - __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n), 1) + __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n), 1, 0) /* * Constraint on the Event code. @@ -203,11 +236,14 @@ struct cpu_hw_events { * - inv * - edge * - cnt-mask + * - in_tx + * - in_tx_checkpointed * The other filters are supported by fixed counters. * The any-thread option is supported starting with v3. */ +#define FIXED_EVENT_FLAGS (X86_RAW_EVENT_MASK|HSW_IN_TX|HSW_IN_TX_CHECKPOINTED) #define FIXED_EVENT_CONSTRAINT(c, n) \ - EVENT_CONSTRAINT(c, (1ULL << (32+n)), X86_RAW_EVENT_MASK) + EVENT_CONSTRAINT(c, (1ULL << (32+n)), FIXED_EVENT_FLAGS) /* * Constraint on the Event code + UMask @@ -215,11 +251,33 @@ struct cpu_hw_events { #define INTEL_UEVENT_CONSTRAINT(c, n) \ EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK) -#define EVENT_CONSTRAINT_END \ - EVENT_CONSTRAINT(0, 0, 0) +#define INTEL_PLD_CONSTRAINT(c, n) \ + __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \ + HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_LDLAT) +#define INTEL_PST_CONSTRAINT(c, n) \ + __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \ + HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST) + +/* DataLA version of store sampling without extra enable bit. */ +#define INTEL_PST_HSW_CONSTRAINT(c, n) \ + __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \ + HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST_HSW) + +/* + * We define the end marker as having a weight of -1 + * to enable blacklisting of events using a counter bitmask + * of zero and thus a weight of zero. + * The end marker has a weight that cannot possibly be + * obtained from counting the bits in the bitmask. + */ +#define EVENT_CONSTRAINT_END { .weight = -1 } + +/* + * Check for end marker with weight == -1 + */ #define for_each_event_constraint(e, c) \ - for ((e) = (c); (e)->weight; (e)++) + for ((e) = (c); (e)->weight != -1; (e)++) /* * Extra registers for specific events. @@ -237,19 +295,31 @@ struct extra_reg { u64 config_mask; u64 valid_mask; int idx; /* per_xxx->regs[] reg index */ + bool extra_msr_access; }; #define EVENT_EXTRA_REG(e, ms, m, vm, i) { \ - .event = (e), \ - .msr = (ms), \ - .config_mask = (m), \ - .valid_mask = (vm), \ - .idx = EXTRA_REG_##i \ + .event = (e), \ + .msr = (ms), \ + .config_mask = (m), \ + .valid_mask = (vm), \ + .idx = EXTRA_REG_##i, \ + .extra_msr_access = true, \ } #define INTEL_EVENT_EXTRA_REG(event, msr, vm, idx) \ EVENT_EXTRA_REG(event, msr, ARCH_PERFMON_EVENTSEL_EVENT, vm, idx) +#define INTEL_UEVENT_EXTRA_REG(event, msr, vm, idx) \ + EVENT_EXTRA_REG(event, msr, ARCH_PERFMON_EVENTSEL_EVENT | \ + ARCH_PERFMON_EVENTSEL_UMASK, vm, idx) + +#define INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG(c) \ + INTEL_UEVENT_EXTRA_REG(c, \ + MSR_PEBS_LD_LAT_THRESHOLD, \ + 0xffff, \ + LDLAT) + #define EVENT_EXTRA_END EVENT_EXTRA_REG(0, 0, 0, 0, RSP_0) union perf_capabilities { @@ -259,6 +329,11 @@ union perf_capabilities { u64 pebs_arch_reg:1; u64 pebs_format:4; u64 smm_freeze:1; + /* + * PMU supports separate counter range for writing + * values > 32bit. + */ + u64 full_width_write:1; }; u64 capabilities; }; @@ -268,6 +343,29 @@ struct x86_pmu_quirk { void (*func)(void); }; +union x86_pmu_config { + struct { + u64 event:8, + umask:8, + usr:1, + os:1, + edge:1, + pc:1, + interrupt:1, + __reserved1:1, + en:1, + inv:1, + cmask:8, + event2:4, + __reserved2:4, + go:1, + ho:1; + } bits; + u64 value; +}; + +#define X86_CONFIG(args...) ((union x86_pmu_config){.bits = {args}}).value + /* * struct x86_pmu - generic x86 pmu */ @@ -286,6 +384,8 @@ struct x86_pmu { int (*schedule_events)(struct cpu_hw_events *cpuc, int n, int *assign); unsigned eventsel; unsigned perfctr; + int (*addr_offset)(int index, bool eventsel); + int (*rdpmc_index)(int index); u64 (*event_map)(int); int max_events; int num_counters; @@ -308,12 +408,30 @@ struct x86_pmu { struct event_constraint *event_constraints; struct x86_pmu_quirk *quirks; int perfctr_second_write; + bool late_ack; + + /* + * sysfs attrs + */ + int attr_rdpmc_broken; + int attr_rdpmc; + struct attribute **format_attrs; + struct attribute **event_attrs; + ssize_t (*events_sysfs_show)(char *page, u64 config); + struct attribute **cpu_events; + + /* + * CPU Hotplug hooks + */ int (*cpu_prepare)(int cpu); void (*cpu_starting)(int cpu); void (*cpu_dying)(int cpu); void (*cpu_dead)(int cpu); + void (*check_microcode)(void); + void (*flush_branch_stack)(void); + /* * Intel Arch Perfmon v2+ */ @@ -323,17 +441,25 @@ struct x86_pmu { /* * Intel DebugStore bits */ - int bts, pebs; - int bts_active, pebs_active; + unsigned int bts :1, + bts_active :1, + pebs :1, + pebs_active :1, + pebs_broken :1; int pebs_record_size; void (*drain_pebs)(struct pt_regs *regs); struct event_constraint *pebs_constraints; + void (*pebs_aliases)(struct perf_event *event); + int max_pebs_events; /* * Intel LBR */ unsigned long lbr_tos, lbr_from, lbr_to; /* MSR base regs */ int lbr_nr; /* hardware stack size */ + u64 lbr_sel_mask; /* LBR_SELECT valid bits */ + const int *lbr_sel_map; /* lbr_select mappings */ + bool lbr_double_abort; /* duplicated lbr aborts */ /* * Extra registers for events @@ -359,6 +485,23 @@ do { \ #define ERF_NO_HT_SHARING 1 #define ERF_HAS_RSP_1 2 +#define EVENT_VAR(_id) event_attr_##_id +#define EVENT_PTR(_id) &event_attr_##_id.attr.attr + +#define EVENT_ATTR(_name, _id) \ +static struct perf_pmu_events_attr EVENT_VAR(_id) = { \ + .attr = __ATTR(_name, 0444, events_sysfs_show, NULL), \ + .id = PERF_COUNT_HW_##_id, \ + .event_str = NULL, \ +}; + +#define EVENT_ATTR_STR(_name, v, str) \ +static struct perf_pmu_events_attr event_attr_##v = { \ + .attr = __ATTR(_name, 0444, events_sysfs_show, NULL), \ + .id = 0, \ + .event_str = str, \ +}; + extern struct x86_pmu x86_pmu __read_mostly; DECLARE_PER_CPU(struct cpu_hw_events, cpu_hw_events); @@ -386,28 +529,21 @@ extern u64 __read_mostly hw_cache_extra_regs u64 x86_perf_event_update(struct perf_event *event); -static inline int x86_pmu_addr_offset(int index) +static inline unsigned int x86_pmu_config_addr(int index) { - int offset; - - /* offset = X86_FEATURE_PERFCTR_CORE ? index << 1 : index */ - alternative_io(ASM_NOP2, - "shll $1, %%eax", - X86_FEATURE_PERFCTR_CORE, - "=a" (offset), - "a" (index)); - - return offset; + return x86_pmu.eventsel + (x86_pmu.addr_offset ? + x86_pmu.addr_offset(index, true) : index); } -static inline unsigned int x86_pmu_config_addr(int index) +static inline unsigned int x86_pmu_event_addr(int index) { - return x86_pmu.eventsel + x86_pmu_addr_offset(index); + return x86_pmu.perfctr + (x86_pmu.addr_offset ? + x86_pmu.addr_offset(index, false) : index); } -static inline unsigned int x86_pmu_event_addr(int index) +static inline int x86_pmu_rdpmc_index(int index) { - return x86_pmu.perfctr + x86_pmu_addr_offset(index); + return x86_pmu.rdpmc_index ? x86_pmu.rdpmc_index(index) : index; } int x86_setup_perfctr(struct perf_event *event); @@ -428,6 +564,8 @@ static inline void __x86_pmu_enable_event(struct hw_perf_event *hwc, void x86_pmu_enable_all(int added); +int perf_assign_events(struct perf_event **events, int n, + int wmin, int wmax, int *assign); int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign); void x86_pmu_stop(struct perf_event *event, int flags); @@ -447,6 +585,38 @@ extern struct event_constraint emptyconstraint; extern struct event_constraint unconstrained; +static inline bool kernel_ip(unsigned long ip) +{ +#ifdef CONFIG_X86_32 + return ip > PAGE_OFFSET; +#else + return (long)ip < 0; +#endif +} + +/* + * Not all PMUs provide the right context information to place the reported IP + * into full context. Specifically segment registers are typically not + * supplied. + * + * Assuming the address is a linear address (it is for IBS), we fake the CS and + * vm86 mode using the known zero-based code segment and 'fix up' the registers + * to reflect this. + * + * Intel PEBS/LBR appear to typically provide the effective address, nothing + * much we can do about that but pray and treat it like a linear address. + */ +static inline void set_linear_ip(struct pt_regs *regs, unsigned long ip) +{ + regs->cs = kernel_ip(ip) ? __KERNEL_CS : __USER_CS; + if (regs->flags & X86_VM_MASK) + regs->flags ^= (PERF_EFLAGS_VM | X86_VM_MASK); + regs->ip = ip; +} + +ssize_t x86_event_sysfs_show(char *page, u64 config, u64 event); +ssize_t intel_event_sysfs_show(char *page, u64 config); + #ifdef CONFIG_CPU_SUP_AMD int amd_pmu_init(void); @@ -491,12 +661,18 @@ extern struct event_constraint intel_core2_pebs_event_constraints[]; extern struct event_constraint intel_atom_pebs_event_constraints[]; +extern struct event_constraint intel_slm_pebs_event_constraints[]; + extern struct event_constraint intel_nehalem_pebs_event_constraints[]; extern struct event_constraint intel_westmere_pebs_event_constraints[]; extern struct event_constraint intel_snb_pebs_event_constraints[]; +extern struct event_constraint intel_ivb_pebs_event_constraints[]; + +extern struct event_constraint intel_hsw_pebs_event_constraints[]; + struct event_constraint *intel_pebs_constraints(struct perf_event *event); void intel_pmu_pebs_enable(struct perf_event *event); @@ -527,10 +703,19 @@ void intel_pmu_lbr_init_nhm(void); void intel_pmu_lbr_init_atom(void); +void intel_pmu_lbr_init_snb(void); + +int intel_pmu_setup_lbr_filter(struct perf_event *event); + int p4_pmu_init(void); int p6_pmu_init(void); +int knc_pmu_init(void); + +ssize_t events_sysfs_show(struct device *dev, struct device_attribute *attr, + char *page); + #else /* CONFIG_CPU_SUP_INTEL */ static inline void reserve_ds_buffers(void) diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c index 67250a52430..beeb7cc0704 100644 --- a/arch/x86/kernel/cpu/perf_event_amd.c +++ b/arch/x86/kernel/cpu/perf_event_amd.c @@ -132,13 +132,49 @@ static u64 amd_pmu_event_map(int hw_event) return amd_perfmon_event_map[hw_event]; } -static int amd_pmu_hw_config(struct perf_event *event) +/* + * Previously calculated offsets + */ +static unsigned int event_offsets[X86_PMC_IDX_MAX] __read_mostly; +static unsigned int count_offsets[X86_PMC_IDX_MAX] __read_mostly; + +/* + * Legacy CPUs: + * 4 counters starting at 0xc0010000 each offset by 1 + * + * CPUs with core performance counter extensions: + * 6 counters starting at 0xc0010200 each offset by 2 + */ +static inline int amd_pmu_addr_offset(int index, bool eventsel) { - int ret = x86_pmu_hw_config(event); + int offset; - if (ret) - return ret; + if (!index) + return index; + if (eventsel) + offset = event_offsets[index]; + else + offset = count_offsets[index]; + + if (offset) + return offset; + + if (!cpu_has_perfctr_core) + offset = index; + else + offset = index << 1; + + if (eventsel) + event_offsets[index] = offset; + else + count_offsets[index] = offset; + + return offset; +} + +static int amd_core_hw_config(struct perf_event *event) +{ if (event->attr.exclude_host && event->attr.exclude_guest) /* * When HO == GO == 1 the hardware treats that as GO == HO == 0 @@ -148,14 +184,9 @@ static int amd_pmu_hw_config(struct perf_event *event) event->hw.config &= ~(ARCH_PERFMON_EVENTSEL_USR | ARCH_PERFMON_EVENTSEL_OS); else if (event->attr.exclude_host) - event->hw.config |= AMD_PERFMON_EVENTSEL_GUESTONLY; + event->hw.config |= AMD64_EVENTSEL_GUESTONLY; else if (event->attr.exclude_guest) - event->hw.config |= AMD_PERFMON_EVENTSEL_HOSTONLY; - - if (event->attr.type != PERF_TYPE_RAW) - return 0; - - event->hw.config |= event->attr.config & AMD64_RAW_EVENT_MASK; + event->hw.config |= AMD64_EVENTSEL_HOSTONLY; return 0; } @@ -180,20 +211,34 @@ static inline int amd_has_nb(struct cpu_hw_events *cpuc) return nb && nb->nb_id != -1; } -static void amd_put_event_constraints(struct cpu_hw_events *cpuc, - struct perf_event *event) +static int amd_pmu_hw_config(struct perf_event *event) +{ + int ret; + + /* pass precise event sampling to ibs: */ + if (event->attr.precise_ip && get_ibs_caps()) + return -ENOENT; + + if (has_branch_stack(event)) + return -EOPNOTSUPP; + + ret = x86_pmu_hw_config(event); + if (ret) + return ret; + + if (event->attr.type == PERF_TYPE_RAW) + event->hw.config |= event->attr.config & AMD64_RAW_EVENT_MASK; + + return amd_core_hw_config(event); +} + +static void __amd_put_nb_event_constraints(struct cpu_hw_events *cpuc, + struct perf_event *event) { - struct hw_perf_event *hwc = &event->hw; struct amd_nb *nb = cpuc->amd_nb; int i; /* - * only care about NB events - */ - if (!(amd_has_nb(cpuc) && amd_is_nb_event(hwc))) - return; - - /* * need to scan whole list because event may not have * been assigned during scheduling * @@ -202,10 +247,8 @@ static void amd_put_event_constraints(struct cpu_hw_events *cpuc, * when we come here */ for (i = 0; i < x86_pmu.num_counters; i++) { - if (nb->owners[i] == event) { - cmpxchg(nb->owners+i, event, NULL); + if (cmpxchg(nb->owners + i, event, NULL) == event) break; - } } } @@ -241,24 +284,24 @@ static void amd_put_event_constraints(struct cpu_hw_events *cpuc, * * Given that resources are allocated (cmpxchg), they must be * eventually freed for others to use. This is accomplished by - * calling amd_put_event_constraints(). + * calling __amd_put_nb_event_constraints() * * Non NB events are not impacted by this restriction. */ static struct event_constraint * -amd_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) +__amd_get_nb_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event, + struct event_constraint *c) { struct hw_perf_event *hwc = &event->hw; struct amd_nb *nb = cpuc->amd_nb; - struct perf_event *old = NULL; - int max = x86_pmu.num_counters; - int i, j, k = -1; + struct perf_event *old; + int idx, new = -1; - /* - * if not NB event or no NB, then no constraints - */ - if (!(amd_has_nb(cpuc) && amd_is_nb_event(hwc))) - return &unconstrained; + if (!c) + c = &unconstrained; + + if (cpuc->is_fake) + return c; /* * detect if already present, if so reuse @@ -270,48 +313,33 @@ amd_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) * because of successive calls to x86_schedule_events() from * hw_perf_group_sched_in() without hw_perf_enable() */ - for (i = 0; i < max; i++) { - /* - * keep track of first free slot - */ - if (k == -1 && !nb->owners[i]) - k = i; + for_each_set_bit(idx, c->idxmsk, x86_pmu.num_counters) { + if (new == -1 || hwc->idx == idx) + /* assign free slot, prefer hwc->idx */ + old = cmpxchg(nb->owners + idx, NULL, event); + else if (nb->owners[idx] == event) + /* event already present */ + old = event; + else + continue; + + if (old && old != event) + continue; + + /* reassign to this slot */ + if (new != -1) + cmpxchg(nb->owners + new, event, NULL); + new = idx; /* already present, reuse */ - if (nb->owners[i] == event) - goto done; - } - /* - * not present, so grab a new slot - * starting either at: - */ - if (hwc->idx != -1) { - /* previous assignment */ - i = hwc->idx; - } else if (k != -1) { - /* start from free slot found */ - i = k; - } else { - /* - * event not found, no slot found in - * first pass, try again from the - * beginning - */ - i = 0; - } - j = i; - do { - old = cmpxchg(nb->owners+i, NULL, event); - if (!old) + if (old == event) break; - if (++i == max) - i = 0; - } while (i != j); -done: - if (!old) - return &nb->event_constraints[i]; - - return &emptyconstraint; + } + + if (new == -1) + return &emptyconstraint; + + return &nb->event_constraints[new]; } static struct amd_nb *amd_alloc_nb(int cpu) @@ -319,8 +347,7 @@ static struct amd_nb *amd_alloc_nb(int cpu) struct amd_nb *nb; int i; - nb = kmalloc_node(sizeof(struct amd_nb), GFP_KERNEL | __GFP_ZERO, - cpu_to_node(cpu)); + nb = kzalloc_node(sizeof(struct amd_nb), GFP_KERNEL, cpu_to_node(cpu)); if (!nb) return NULL; @@ -358,9 +385,9 @@ static void amd_pmu_cpu_starting(int cpu) struct amd_nb *nb; int i, nb_id; - cpuc->perf_ctr_virt_mask = AMD_PERFMON_EVENTSEL_HOSTONLY; + cpuc->perf_ctr_virt_mask = AMD64_EVENTSEL_HOSTONLY; - if (boot_cpu_data.x86_max_cores < 2 || boot_cpu_data.x86 == 0x15) + if (boot_cpu_data.x86_max_cores < 2) return; nb_id = amd_get_nb_id(cpu); @@ -401,31 +428,38 @@ static void amd_pmu_cpu_dead(int cpu) } } -static __initconst const struct x86_pmu amd_pmu = { - .name = "AMD", - .handle_irq = x86_pmu_handle_irq, - .disable_all = x86_pmu_disable_all, - .enable_all = x86_pmu_enable_all, - .enable = x86_pmu_enable_event, - .disable = x86_pmu_disable_event, - .hw_config = amd_pmu_hw_config, - .schedule_events = x86_schedule_events, - .eventsel = MSR_K7_EVNTSEL0, - .perfctr = MSR_K7_PERFCTR0, - .event_map = amd_pmu_event_map, - .max_events = ARRAY_SIZE(amd_perfmon_event_map), - .num_counters = AMD64_NUM_COUNTERS, - .cntval_bits = 48, - .cntval_mask = (1ULL << 48) - 1, - .apic = 1, - /* use highest bit to detect overflow */ - .max_period = (1ULL << 47) - 1, - .get_event_constraints = amd_get_event_constraints, - .put_event_constraints = amd_put_event_constraints, +static struct event_constraint * +amd_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) +{ + /* + * if not NB event or no NB, then no constraints + */ + if (!(amd_has_nb(cpuc) && amd_is_nb_event(&event->hw))) + return &unconstrained; - .cpu_prepare = amd_pmu_cpu_prepare, - .cpu_starting = amd_pmu_cpu_starting, - .cpu_dead = amd_pmu_cpu_dead, + return __amd_get_nb_event_constraints(cpuc, event, NULL); +} + +static void amd_put_event_constraints(struct cpu_hw_events *cpuc, + struct perf_event *event) +{ + if (amd_has_nb(cpuc) && amd_is_nb_event(&event->hw)) + __amd_put_nb_event_constraints(cpuc, event); +} + +PMU_FORMAT_ATTR(event, "config:0-7,32-35"); +PMU_FORMAT_ATTR(umask, "config:8-15" ); +PMU_FORMAT_ATTR(edge, "config:18" ); +PMU_FORMAT_ATTR(inv, "config:23" ); +PMU_FORMAT_ATTR(cmask, "config:24-31" ); + +static struct attribute *amd_format_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_cmask.attr, + NULL, }; /* AMD Family 15h */ @@ -473,6 +507,7 @@ static __initconst const struct x86_pmu amd_pmu = { * 0x023 DE PERF_CTL[2:0] * 0x02D LS PERF_CTL[3] * 0x02E LS PERF_CTL[3,0] + * 0x031 LS PERF_CTL[2:0] (**) * 0x043 CU PERF_CTL[2:0] * 0x045 CU PERF_CTL[2:0] * 0x046 CU PERF_CTL[2:0] @@ -486,10 +521,12 @@ static __initconst const struct x86_pmu amd_pmu = { * 0x0DD LS PERF_CTL[5:0] * 0x0DE LS PERF_CTL[5:0] * 0x0DF LS PERF_CTL[5:0] + * 0x1C0 EX PERF_CTL[5:3] * 0x1D6 EX PERF_CTL[5:0] * 0x1D8 EX PERF_CTL[5:0] * - * (*) depending on the umask all FPU counters may be used + * (*) depending on the umask all FPU counters may be used + * (**) only one unitmask enabled at a time */ static struct event_constraint amd_f15_PMC0 = EVENT_CONSTRAINT(0, 0x01, 0); @@ -539,6 +576,12 @@ amd_get_event_constraints_f15h(struct cpu_hw_events *cpuc, struct perf_event *ev return &amd_f15_PMC3; case 0x02E: return &amd_f15_PMC30; + case 0x031: + if (hweight_long(hwc->config & ARCH_PERFMON_EVENTSEL_UMASK) <= 1) + return &amd_f15_PMC20; + return &emptyconstraint; + case 0x1C0: + return &amd_f15_PMC53; default: return &amd_f15_PMC50; } @@ -558,15 +601,23 @@ amd_get_event_constraints_f15h(struct cpu_hw_events *cpuc, struct perf_event *ev return &amd_f15_PMC20; } case AMD_EVENT_NB: - /* not yet implemented */ + /* moved to perf_event_amd_uncore.c */ return &emptyconstraint; default: return &emptyconstraint; } } -static __initconst const struct x86_pmu amd_pmu_f15h = { - .name = "AMD Family 15h", +static ssize_t amd_event_sysfs_show(char *page, u64 config) +{ + u64 event = (config & ARCH_PERFMON_EVENTSEL_EVENT) | + (config & AMD64_EVENTSEL_EVENT) >> 24; + + return x86_event_sysfs_show(page, config, event); +} + +static __initconst const struct x86_pmu amd_pmu = { + .name = "AMD", .handle_irq = x86_pmu_handle_irq, .disable_all = x86_pmu_disable_all, .enable_all = x86_pmu_enable_all, @@ -574,50 +625,71 @@ static __initconst const struct x86_pmu amd_pmu_f15h = { .disable = x86_pmu_disable_event, .hw_config = amd_pmu_hw_config, .schedule_events = x86_schedule_events, - .eventsel = MSR_F15H_PERF_CTL, - .perfctr = MSR_F15H_PERF_CTR, + .eventsel = MSR_K7_EVNTSEL0, + .perfctr = MSR_K7_PERFCTR0, + .addr_offset = amd_pmu_addr_offset, .event_map = amd_pmu_event_map, .max_events = ARRAY_SIZE(amd_perfmon_event_map), - .num_counters = AMD64_NUM_COUNTERS_F15H, + .num_counters = AMD64_NUM_COUNTERS, .cntval_bits = 48, .cntval_mask = (1ULL << 48) - 1, .apic = 1, /* use highest bit to detect overflow */ .max_period = (1ULL << 47) - 1, - .get_event_constraints = amd_get_event_constraints_f15h, - /* nortbridge counters not yet implemented: */ -#if 0 + .get_event_constraints = amd_get_event_constraints, .put_event_constraints = amd_put_event_constraints, + .format_attrs = amd_format_attr, + .events_sysfs_show = amd_event_sysfs_show, + .cpu_prepare = amd_pmu_cpu_prepare, - .cpu_dead = amd_pmu_cpu_dead, -#endif .cpu_starting = amd_pmu_cpu_starting, + .cpu_dead = amd_pmu_cpu_dead, }; -__init int amd_pmu_init(void) +static int __init amd_core_pmu_init(void) { - /* Performance-monitoring supported from K7 and later: */ - if (boot_cpu_data.x86 < 6) - return -ENODEV; + if (!cpu_has_perfctr_core) + return 0; - /* - * If core performance counter extensions exists, it must be - * family 15h, otherwise fail. See x86_pmu_addr_offset(). - */ switch (boot_cpu_data.x86) { case 0x15: - if (!cpu_has_perfctr_core) - return -ENODEV; - x86_pmu = amd_pmu_f15h; + pr_cont("Fam15h "); + x86_pmu.get_event_constraints = amd_get_event_constraints_f15h; break; + default: - if (cpu_has_perfctr_core) - return -ENODEV; - x86_pmu = amd_pmu; - break; + pr_err("core perfctr but no constraints; unknown hardware!\n"); + return -ENODEV; } + /* + * If core performance counter extensions exists, we must use + * MSR_F15H_PERF_CTL/MSR_F15H_PERF_CTR msrs. See also + * amd_pmu_addr_offset(). + */ + x86_pmu.eventsel = MSR_F15H_PERF_CTL; + x86_pmu.perfctr = MSR_F15H_PERF_CTR; + x86_pmu.num_counters = AMD64_NUM_COUNTERS_CORE; + + pr_cont("core perfctr, "); + return 0; +} + +__init int amd_pmu_init(void) +{ + int ret; + + /* Performance-monitoring supported from K7 and later: */ + if (boot_cpu_data.x86 < 6) + return -ENODEV; + + x86_pmu = amd_pmu; + + ret = amd_core_pmu_init(); + if (ret) + return ret; + /* Events are common for all AMDs */ memcpy(hw_cache_event_ids, amd_hw_cache_event_ids, sizeof(hw_cache_event_ids)); @@ -647,7 +719,7 @@ void amd_pmu_disable_virt(void) * SVM is disabled the Guest-only bits still gets set and the counter * will not count anything. */ - cpuc->perf_ctr_virt_mask = AMD_PERFMON_EVENTSEL_HOSTONLY; + cpuc->perf_ctr_virt_mask = AMD64_EVENTSEL_HOSTONLY; /* Reload all events */ x86_pmu_disable_all(); diff --git a/arch/x86/kernel/cpu/perf_event_amd_ibs.c b/arch/x86/kernel/cpu/perf_event_amd_ibs.c index 3b8a2d30d14..cbb1be3ed9e 100644 --- a/arch/x86/kernel/cpu/perf_event_amd_ibs.c +++ b/arch/x86/kernel/cpu/perf_event_amd_ibs.c @@ -9,43 +9,652 @@ #include <linux/perf_event.h> #include <linux/module.h> #include <linux/pci.h> +#include <linux/ptrace.h> +#include <linux/syscore_ops.h> #include <asm/apic.h> +#include "perf_event.h" + static u32 ibs_caps; #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD) -static struct pmu perf_ibs; +#include <linux/kprobes.h> +#include <linux/hardirq.h> + +#include <asm/nmi.h> + +#define IBS_FETCH_CONFIG_MASK (IBS_FETCH_RAND_EN | IBS_FETCH_MAX_CNT) +#define IBS_OP_CONFIG_MASK IBS_OP_MAX_CNT + +enum ibs_states { + IBS_ENABLED = 0, + IBS_STARTED = 1, + IBS_STOPPING = 2, + + IBS_MAX_STATES, +}; + +struct cpu_perf_ibs { + struct perf_event *event; + unsigned long state[BITS_TO_LONGS(IBS_MAX_STATES)]; +}; + +struct perf_ibs { + struct pmu pmu; + unsigned int msr; + u64 config_mask; + u64 cnt_mask; + u64 enable_mask; + u64 valid_mask; + u64 max_period; + unsigned long offset_mask[1]; + int offset_max; + struct cpu_perf_ibs __percpu *pcpu; + + struct attribute **format_attrs; + struct attribute_group format_group; + const struct attribute_group *attr_groups[2]; + + u64 (*get_count)(u64 config); +}; + +struct perf_ibs_data { + u32 size; + union { + u32 data[0]; /* data buffer starts here */ + u32 caps; + }; + u64 regs[MSR_AMD64_IBS_REG_COUNT_MAX]; +}; + +static int +perf_event_set_period(struct hw_perf_event *hwc, u64 min, u64 max, u64 *hw_period) +{ + s64 left = local64_read(&hwc->period_left); + s64 period = hwc->sample_period; + int overflow = 0; + + /* + * If we are way outside a reasonable range then just skip forward: + */ + if (unlikely(left <= -period)) { + left = period; + local64_set(&hwc->period_left, left); + hwc->last_period = period; + overflow = 1; + } + + if (unlikely(left < (s64)min)) { + left += period; + local64_set(&hwc->period_left, left); + hwc->last_period = period; + overflow = 1; + } + + /* + * If the hw period that triggers the sw overflow is too short + * we might hit the irq handler. This biases the results. + * Thus we shorten the next-to-last period and set the last + * period to the max period. + */ + if (left > max) { + left -= max; + if (left > max) + left = max; + else if (left < min) + left = min; + } + + *hw_period = (u64)left; + + return overflow; +} + +static int +perf_event_try_update(struct perf_event *event, u64 new_raw_count, int width) +{ + struct hw_perf_event *hwc = &event->hw; + int shift = 64 - width; + u64 prev_raw_count; + u64 delta; + + /* + * 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 event atomically: + */ + prev_raw_count = local64_read(&hwc->prev_count); + if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, + new_raw_count) != prev_raw_count) + return 0; + + /* + * Now we have the new raw value and have updated the prev + * timestamp already. We can now calculate the elapsed delta + * (event-)time and add that to the generic event. + * + * Careful, not all hw sign-extends above the physical width + * of the count. + */ + delta = (new_raw_count << shift) - (prev_raw_count << shift); + delta >>= shift; + + local64_add(delta, &event->count); + local64_sub(delta, &hwc->period_left); + + return 1; +} + +static struct perf_ibs perf_ibs_fetch; +static struct perf_ibs perf_ibs_op; + +static struct perf_ibs *get_ibs_pmu(int type) +{ + if (perf_ibs_fetch.pmu.type == type) + return &perf_ibs_fetch; + if (perf_ibs_op.pmu.type == type) + return &perf_ibs_op; + return NULL; +} + +/* + * Use IBS for precise event sampling: + * + * perf record -a -e cpu-cycles:p ... # use ibs op counting cycle count + * perf record -a -e r076:p ... # same as -e cpu-cycles:p + * perf record -a -e r0C1:p ... # use ibs op counting micro-ops + * + * IbsOpCntCtl (bit 19) of IBS Execution Control Register (IbsOpCtl, + * MSRC001_1033) is used to select either cycle or micro-ops counting + * mode. + * + * The rip of IBS samples has skid 0. Thus, IBS supports precise + * levels 1 and 2 and the PERF_EFLAGS_EXACT is set. In rare cases the + * rip is invalid when IBS was not able to record the rip correctly. + * We clear PERF_EFLAGS_EXACT and take the rip from pt_regs then. + * + */ +static int perf_ibs_precise_event(struct perf_event *event, u64 *config) +{ + switch (event->attr.precise_ip) { + case 0: + return -ENOENT; + case 1: + case 2: + break; + default: + return -EOPNOTSUPP; + } + + switch (event->attr.type) { + case PERF_TYPE_HARDWARE: + switch (event->attr.config) { + case PERF_COUNT_HW_CPU_CYCLES: + *config = 0; + return 0; + } + break; + case PERF_TYPE_RAW: + switch (event->attr.config) { + case 0x0076: + *config = 0; + return 0; + case 0x00C1: + *config = IBS_OP_CNT_CTL; + return 0; + } + break; + default: + return -ENOENT; + } + + return -EOPNOTSUPP; +} + +static const struct perf_event_attr ibs_notsupp = { + .exclude_user = 1, + .exclude_kernel = 1, + .exclude_hv = 1, + .exclude_idle = 1, + .exclude_host = 1, + .exclude_guest = 1, +}; static int perf_ibs_init(struct perf_event *event) { - if (perf_ibs.type != event->attr.type) + struct hw_perf_event *hwc = &event->hw; + struct perf_ibs *perf_ibs; + u64 max_cnt, config; + int ret; + + perf_ibs = get_ibs_pmu(event->attr.type); + if (perf_ibs) { + config = event->attr.config; + } else { + perf_ibs = &perf_ibs_op; + ret = perf_ibs_precise_event(event, &config); + if (ret) + return ret; + } + + if (event->pmu != &perf_ibs->pmu) return -ENOENT; + + if (perf_flags(&event->attr) & perf_flags(&ibs_notsupp)) + return -EINVAL; + + if (config & ~perf_ibs->config_mask) + return -EINVAL; + + if (hwc->sample_period) { + if (config & perf_ibs->cnt_mask) + /* raw max_cnt may not be set */ + return -EINVAL; + if (!event->attr.sample_freq && hwc->sample_period & 0x0f) + /* + * lower 4 bits can not be set in ibs max cnt, + * but allowing it in case we adjust the + * sample period to set a frequency. + */ + return -EINVAL; + hwc->sample_period &= ~0x0FULL; + if (!hwc->sample_period) + hwc->sample_period = 0x10; + } else { + max_cnt = config & perf_ibs->cnt_mask; + config &= ~perf_ibs->cnt_mask; + event->attr.sample_period = max_cnt << 4; + hwc->sample_period = event->attr.sample_period; + } + + if (!hwc->sample_period) + return -EINVAL; + + /* + * If we modify hwc->sample_period, we also need to update + * hwc->last_period and hwc->period_left. + */ + hwc->last_period = hwc->sample_period; + local64_set(&hwc->period_left, hwc->sample_period); + + hwc->config_base = perf_ibs->msr; + hwc->config = config; + return 0; } +static int perf_ibs_set_period(struct perf_ibs *perf_ibs, + struct hw_perf_event *hwc, u64 *period) +{ + int overflow; + + /* ignore lower 4 bits in min count: */ + overflow = perf_event_set_period(hwc, 1<<4, perf_ibs->max_period, period); + local64_set(&hwc->prev_count, 0); + + return overflow; +} + +static u64 get_ibs_fetch_count(u64 config) +{ + return (config & IBS_FETCH_CNT) >> 12; +} + +static u64 get_ibs_op_count(u64 config) +{ + u64 count = 0; + + if (config & IBS_OP_VAL) + count += (config & IBS_OP_MAX_CNT) << 4; /* cnt rolled over */ + + if (ibs_caps & IBS_CAPS_RDWROPCNT) + count += (config & IBS_OP_CUR_CNT) >> 32; + + return count; +} + +static void +perf_ibs_event_update(struct perf_ibs *perf_ibs, struct perf_event *event, + u64 *config) +{ + u64 count = perf_ibs->get_count(*config); + + /* + * Set width to 64 since we do not overflow on max width but + * instead on max count. In perf_ibs_set_period() we clear + * prev count manually on overflow. + */ + while (!perf_event_try_update(event, count, 64)) { + rdmsrl(event->hw.config_base, *config); + count = perf_ibs->get_count(*config); + } +} + +static inline void perf_ibs_enable_event(struct perf_ibs *perf_ibs, + struct hw_perf_event *hwc, u64 config) +{ + wrmsrl(hwc->config_base, hwc->config | config | perf_ibs->enable_mask); +} + +/* + * Erratum #420 Instruction-Based Sampling Engine May Generate + * Interrupt that Cannot Be Cleared: + * + * Must clear counter mask first, then clear the enable bit. See + * Revision Guide for AMD Family 10h Processors, Publication #41322. + */ +static inline void perf_ibs_disable_event(struct perf_ibs *perf_ibs, + struct hw_perf_event *hwc, u64 config) +{ + config &= ~perf_ibs->cnt_mask; + wrmsrl(hwc->config_base, config); + config &= ~perf_ibs->enable_mask; + wrmsrl(hwc->config_base, config); +} + +/* + * We cannot restore the ibs pmu state, so we always needs to update + * the event while stopping it and then reset the state when starting + * again. Thus, ignoring PERF_EF_RELOAD and PERF_EF_UPDATE flags in + * perf_ibs_start()/perf_ibs_stop() and instead always do it. + */ +static void perf_ibs_start(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + struct perf_ibs *perf_ibs = container_of(event->pmu, struct perf_ibs, pmu); + struct cpu_perf_ibs *pcpu = this_cpu_ptr(perf_ibs->pcpu); + u64 period; + + if (WARN_ON_ONCE(!(hwc->state & PERF_HES_STOPPED))) + return; + + WARN_ON_ONCE(!(hwc->state & PERF_HES_UPTODATE)); + hwc->state = 0; + + perf_ibs_set_period(perf_ibs, hwc, &period); + set_bit(IBS_STARTED, pcpu->state); + perf_ibs_enable_event(perf_ibs, hwc, period >> 4); + + perf_event_update_userpage(event); +} + +static void perf_ibs_stop(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + struct perf_ibs *perf_ibs = container_of(event->pmu, struct perf_ibs, pmu); + struct cpu_perf_ibs *pcpu = this_cpu_ptr(perf_ibs->pcpu); + u64 config; + int stopping; + + stopping = test_and_clear_bit(IBS_STARTED, pcpu->state); + + if (!stopping && (hwc->state & PERF_HES_UPTODATE)) + return; + + rdmsrl(hwc->config_base, config); + + if (stopping) { + set_bit(IBS_STOPPING, pcpu->state); + perf_ibs_disable_event(perf_ibs, hwc, config); + WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED); + hwc->state |= PERF_HES_STOPPED; + } + + if (hwc->state & PERF_HES_UPTODATE) + return; + + /* + * Clear valid bit to not count rollovers on update, rollovers + * are only updated in the irq handler. + */ + config &= ~perf_ibs->valid_mask; + + perf_ibs_event_update(perf_ibs, event, &config); + hwc->state |= PERF_HES_UPTODATE; +} + static int perf_ibs_add(struct perf_event *event, int flags) { + struct perf_ibs *perf_ibs = container_of(event->pmu, struct perf_ibs, pmu); + struct cpu_perf_ibs *pcpu = this_cpu_ptr(perf_ibs->pcpu); + + if (test_and_set_bit(IBS_ENABLED, pcpu->state)) + return -ENOSPC; + + event->hw.state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + + pcpu->event = event; + + if (flags & PERF_EF_START) + perf_ibs_start(event, PERF_EF_RELOAD); + return 0; } static void perf_ibs_del(struct perf_event *event, int flags) { + struct perf_ibs *perf_ibs = container_of(event->pmu, struct perf_ibs, pmu); + struct cpu_perf_ibs *pcpu = this_cpu_ptr(perf_ibs->pcpu); + + if (!test_and_clear_bit(IBS_ENABLED, pcpu->state)) + return; + + perf_ibs_stop(event, PERF_EF_UPDATE); + + pcpu->event = NULL; + + perf_event_update_userpage(event); } -static struct pmu perf_ibs = { - .event_init= perf_ibs_init, - .add= perf_ibs_add, - .del= perf_ibs_del, +static void perf_ibs_read(struct perf_event *event) { } + +PMU_FORMAT_ATTR(rand_en, "config:57"); +PMU_FORMAT_ATTR(cnt_ctl, "config:19"); + +static struct attribute *ibs_fetch_format_attrs[] = { + &format_attr_rand_en.attr, + NULL, }; +static struct attribute *ibs_op_format_attrs[] = { + NULL, /* &format_attr_cnt_ctl.attr if IBS_CAPS_OPCNT */ + NULL, +}; + +static struct perf_ibs perf_ibs_fetch = { + .pmu = { + .task_ctx_nr = perf_invalid_context, + + .event_init = perf_ibs_init, + .add = perf_ibs_add, + .del = perf_ibs_del, + .start = perf_ibs_start, + .stop = perf_ibs_stop, + .read = perf_ibs_read, + }, + .msr = MSR_AMD64_IBSFETCHCTL, + .config_mask = IBS_FETCH_CONFIG_MASK, + .cnt_mask = IBS_FETCH_MAX_CNT, + .enable_mask = IBS_FETCH_ENABLE, + .valid_mask = IBS_FETCH_VAL, + .max_period = IBS_FETCH_MAX_CNT << 4, + .offset_mask = { MSR_AMD64_IBSFETCH_REG_MASK }, + .offset_max = MSR_AMD64_IBSFETCH_REG_COUNT, + .format_attrs = ibs_fetch_format_attrs, + + .get_count = get_ibs_fetch_count, +}; + +static struct perf_ibs perf_ibs_op = { + .pmu = { + .task_ctx_nr = perf_invalid_context, + + .event_init = perf_ibs_init, + .add = perf_ibs_add, + .del = perf_ibs_del, + .start = perf_ibs_start, + .stop = perf_ibs_stop, + .read = perf_ibs_read, + }, + .msr = MSR_AMD64_IBSOPCTL, + .config_mask = IBS_OP_CONFIG_MASK, + .cnt_mask = IBS_OP_MAX_CNT, + .enable_mask = IBS_OP_ENABLE, + .valid_mask = IBS_OP_VAL, + .max_period = IBS_OP_MAX_CNT << 4, + .offset_mask = { MSR_AMD64_IBSOP_REG_MASK }, + .offset_max = MSR_AMD64_IBSOP_REG_COUNT, + .format_attrs = ibs_op_format_attrs, + + .get_count = get_ibs_op_count, +}; + +static int perf_ibs_handle_irq(struct perf_ibs *perf_ibs, struct pt_regs *iregs) +{ + struct cpu_perf_ibs *pcpu = this_cpu_ptr(perf_ibs->pcpu); + struct perf_event *event = pcpu->event; + struct hw_perf_event *hwc = &event->hw; + struct perf_sample_data data; + struct perf_raw_record raw; + struct pt_regs regs; + struct perf_ibs_data ibs_data; + int offset, size, check_rip, offset_max, throttle = 0; + unsigned int msr; + u64 *buf, *config, period; + + if (!test_bit(IBS_STARTED, pcpu->state)) { + /* + * Catch spurious interrupts after stopping IBS: After + * disabling IBS there could be still incoming NMIs + * with samples that even have the valid bit cleared. + * Mark all this NMIs as handled. + */ + return test_and_clear_bit(IBS_STOPPING, pcpu->state) ? 1 : 0; + } + + msr = hwc->config_base; + buf = ibs_data.regs; + rdmsrl(msr, *buf); + if (!(*buf++ & perf_ibs->valid_mask)) + return 0; + + config = &ibs_data.regs[0]; + perf_ibs_event_update(perf_ibs, event, config); + perf_sample_data_init(&data, 0, hwc->last_period); + if (!perf_ibs_set_period(perf_ibs, hwc, &period)) + goto out; /* no sw counter overflow */ + + ibs_data.caps = ibs_caps; + size = 1; + offset = 1; + check_rip = (perf_ibs == &perf_ibs_op && (ibs_caps & IBS_CAPS_RIPINVALIDCHK)); + if (event->attr.sample_type & PERF_SAMPLE_RAW) + offset_max = perf_ibs->offset_max; + else if (check_rip) + offset_max = 2; + else + offset_max = 1; + do { + rdmsrl(msr + offset, *buf++); + size++; + offset = find_next_bit(perf_ibs->offset_mask, + perf_ibs->offset_max, + offset + 1); + } while (offset < offset_max); + ibs_data.size = sizeof(u64) * size; + + regs = *iregs; + if (check_rip && (ibs_data.regs[2] & IBS_RIP_INVALID)) { + regs.flags &= ~PERF_EFLAGS_EXACT; + } else { + set_linear_ip(®s, ibs_data.regs[1]); + regs.flags |= PERF_EFLAGS_EXACT; + } + + if (event->attr.sample_type & PERF_SAMPLE_RAW) { + raw.size = sizeof(u32) + ibs_data.size; + raw.data = ibs_data.data; + data.raw = &raw; + } + + throttle = perf_event_overflow(event, &data, ®s); +out: + if (throttle) + perf_ibs_disable_event(perf_ibs, hwc, *config); + else + perf_ibs_enable_event(perf_ibs, hwc, period >> 4); + + perf_event_update_userpage(event); + + return 1; +} + +static int +perf_ibs_nmi_handler(unsigned int cmd, struct pt_regs *regs) +{ + int handled = 0; + + handled += perf_ibs_handle_irq(&perf_ibs_fetch, regs); + handled += perf_ibs_handle_irq(&perf_ibs_op, regs); + + if (handled) + inc_irq_stat(apic_perf_irqs); + + return handled; +} +NOKPROBE_SYMBOL(perf_ibs_nmi_handler); + +static __init int perf_ibs_pmu_init(struct perf_ibs *perf_ibs, char *name) +{ + struct cpu_perf_ibs __percpu *pcpu; + int ret; + + pcpu = alloc_percpu(struct cpu_perf_ibs); + if (!pcpu) + return -ENOMEM; + + perf_ibs->pcpu = pcpu; + + /* register attributes */ + if (perf_ibs->format_attrs[0]) { + memset(&perf_ibs->format_group, 0, sizeof(perf_ibs->format_group)); + perf_ibs->format_group.name = "format"; + perf_ibs->format_group.attrs = perf_ibs->format_attrs; + + memset(&perf_ibs->attr_groups, 0, sizeof(perf_ibs->attr_groups)); + perf_ibs->attr_groups[0] = &perf_ibs->format_group; + perf_ibs->pmu.attr_groups = perf_ibs->attr_groups; + } + + ret = perf_pmu_register(&perf_ibs->pmu, name, -1); + if (ret) { + perf_ibs->pcpu = NULL; + free_percpu(pcpu); + } + + return ret; +} + static __init int perf_event_ibs_init(void) { + struct attribute **attr = ibs_op_format_attrs; + if (!ibs_caps) return -ENODEV; /* ibs not supported by the cpu */ - perf_pmu_register(&perf_ibs, "ibs", -1); + perf_ibs_pmu_init(&perf_ibs_fetch, "ibs_fetch"); + + if (ibs_caps & IBS_CAPS_OPCNT) { + perf_ibs_op.config_mask |= IBS_OP_CNT_CTL; + *attr++ = &format_attr_cnt_ctl.attr; + } + perf_ibs_pmu_init(&perf_ibs_op, "ibs_op"); + + register_nmi_handler(NMI_LOCAL, perf_ibs_nmi_handler, 0, "perf_ibs"); printk(KERN_INFO "perf: AMD IBS detected (0x%08x)\n", ibs_caps); return 0; @@ -209,6 +818,18 @@ out: return ret; } +static void ibs_eilvt_setup(void) +{ + /* + * Force LVT offset assignment for family 10h: The offsets are + * not assigned by the BIOS for this family, so the OS is + * responsible for doing it. If the OS assignment fails, fall + * back to BIOS settings and try to setup this. + */ + if (boot_cpu_data.x86 == 0x10) + force_ibs_eilvt_setup(); +} + static inline int get_ibs_lvt_offset(void) { u64 val; @@ -244,7 +865,37 @@ static void clear_APIC_ibs(void *dummy) setup_APIC_eilvt(offset, 0, APIC_EILVT_MSG_FIX, 1); } -static int __cpuinit +#ifdef CONFIG_PM + +static int perf_ibs_suspend(void) +{ + clear_APIC_ibs(NULL); + return 0; +} + +static void perf_ibs_resume(void) +{ + ibs_eilvt_setup(); + setup_APIC_ibs(NULL); +} + +static struct syscore_ops perf_ibs_syscore_ops = { + .resume = perf_ibs_resume, + .suspend = perf_ibs_suspend, +}; + +static void perf_ibs_pm_init(void) +{ + register_syscore_ops(&perf_ibs_syscore_ops); +} + +#else + +static inline void perf_ibs_pm_init(void) { } + +#endif + +static int perf_ibs_cpu_notifier(struct notifier_block *self, unsigned long action, void *hcpu) { switch (action & ~CPU_TASKS_FROZEN) { @@ -270,25 +921,19 @@ static __init int amd_ibs_init(void) if (!caps) return -ENODEV; /* ibs not supported by the cpu */ - /* - * Force LVT offset assignment for family 10h: The offsets are - * not assigned by the BIOS for this family, so the OS is - * responsible for doing it. If the OS assignment fails, fall - * back to BIOS settings and try to setup this. - */ - if (boot_cpu_data.x86 == 0x10) - force_ibs_eilvt_setup(); + ibs_eilvt_setup(); if (!ibs_eilvt_valid()) goto out; - get_online_cpus(); + perf_ibs_pm_init(); + cpu_notifier_register_begin(); ibs_caps = caps; /* make ibs_caps visible to other cpus: */ smp_mb(); - perf_cpu_notifier(perf_ibs_cpu_notifier); smp_call_function(setup_APIC_ibs, NULL, 1); - put_online_cpus(); + __perf_cpu_notifier(perf_ibs_cpu_notifier); + cpu_notifier_register_done(); ret = perf_event_ibs_init(); out: diff --git a/arch/x86/kernel/cpu/perf_event_amd_iommu.c b/arch/x86/kernel/cpu/perf_event_amd_iommu.c new file mode 100644 index 00000000000..639d1289b1b --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_amd_iommu.c @@ -0,0 +1,502 @@ +/* + * Copyright (C) 2013 Advanced Micro Devices, Inc. + * + * Author: Steven Kinney <Steven.Kinney@amd.com> + * Author: Suravee Suthikulpanit <Suraveee.Suthikulpanit@amd.com> + * + * Perf: amd_iommu - AMD IOMMU Performance Counter PMU implementation + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/perf_event.h> +#include <linux/module.h> +#include <linux/cpumask.h> +#include <linux/slab.h> + +#include "perf_event.h" +#include "perf_event_amd_iommu.h" + +#define COUNTER_SHIFT 16 + +#define _GET_BANK(ev) ((u8)(ev->hw.extra_reg.reg >> 8)) +#define _GET_CNTR(ev) ((u8)(ev->hw.extra_reg.reg)) + +/* iommu pmu config masks */ +#define _GET_CSOURCE(ev) ((ev->hw.config & 0xFFULL)) +#define _GET_DEVID(ev) ((ev->hw.config >> 8) & 0xFFFFULL) +#define _GET_PASID(ev) ((ev->hw.config >> 24) & 0xFFFFULL) +#define _GET_DOMID(ev) ((ev->hw.config >> 40) & 0xFFFFULL) +#define _GET_DEVID_MASK(ev) ((ev->hw.extra_reg.config) & 0xFFFFULL) +#define _GET_PASID_MASK(ev) ((ev->hw.extra_reg.config >> 16) & 0xFFFFULL) +#define _GET_DOMID_MASK(ev) ((ev->hw.extra_reg.config >> 32) & 0xFFFFULL) + +static struct perf_amd_iommu __perf_iommu; + +struct perf_amd_iommu { + struct pmu pmu; + u8 max_banks; + u8 max_counters; + u64 cntr_assign_mask; + raw_spinlock_t lock; + const struct attribute_group *attr_groups[4]; +}; + +#define format_group attr_groups[0] +#define cpumask_group attr_groups[1] +#define events_group attr_groups[2] +#define null_group attr_groups[3] + +/*--------------------------------------------- + * sysfs format attributes + *---------------------------------------------*/ +PMU_FORMAT_ATTR(csource, "config:0-7"); +PMU_FORMAT_ATTR(devid, "config:8-23"); +PMU_FORMAT_ATTR(pasid, "config:24-39"); +PMU_FORMAT_ATTR(domid, "config:40-55"); +PMU_FORMAT_ATTR(devid_mask, "config1:0-15"); +PMU_FORMAT_ATTR(pasid_mask, "config1:16-31"); +PMU_FORMAT_ATTR(domid_mask, "config1:32-47"); + +static struct attribute *iommu_format_attrs[] = { + &format_attr_csource.attr, + &format_attr_devid.attr, + &format_attr_pasid.attr, + &format_attr_domid.attr, + &format_attr_devid_mask.attr, + &format_attr_pasid_mask.attr, + &format_attr_domid_mask.attr, + NULL, +}; + +static struct attribute_group amd_iommu_format_group = { + .name = "format", + .attrs = iommu_format_attrs, +}; + +/*--------------------------------------------- + * sysfs events attributes + *---------------------------------------------*/ +struct amd_iommu_event_desc { + struct kobj_attribute attr; + const char *event; +}; + +static ssize_t _iommu_event_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct amd_iommu_event_desc *event = + container_of(attr, struct amd_iommu_event_desc, attr); + return sprintf(buf, "%s\n", event->event); +} + +#define AMD_IOMMU_EVENT_DESC(_name, _event) \ +{ \ + .attr = __ATTR(_name, 0444, _iommu_event_show, NULL), \ + .event = _event, \ +} + +static struct amd_iommu_event_desc amd_iommu_v2_event_descs[] = { + AMD_IOMMU_EVENT_DESC(mem_pass_untrans, "csource=0x01"), + AMD_IOMMU_EVENT_DESC(mem_pass_pretrans, "csource=0x02"), + AMD_IOMMU_EVENT_DESC(mem_pass_excl, "csource=0x03"), + AMD_IOMMU_EVENT_DESC(mem_target_abort, "csource=0x04"), + AMD_IOMMU_EVENT_DESC(mem_trans_total, "csource=0x05"), + AMD_IOMMU_EVENT_DESC(mem_iommu_tlb_pte_hit, "csource=0x06"), + AMD_IOMMU_EVENT_DESC(mem_iommu_tlb_pte_mis, "csource=0x07"), + AMD_IOMMU_EVENT_DESC(mem_iommu_tlb_pde_hit, "csource=0x08"), + AMD_IOMMU_EVENT_DESC(mem_iommu_tlb_pde_mis, "csource=0x09"), + AMD_IOMMU_EVENT_DESC(mem_dte_hit, "csource=0x0a"), + AMD_IOMMU_EVENT_DESC(mem_dte_mis, "csource=0x0b"), + AMD_IOMMU_EVENT_DESC(page_tbl_read_tot, "csource=0x0c"), + AMD_IOMMU_EVENT_DESC(page_tbl_read_nst, "csource=0x0d"), + AMD_IOMMU_EVENT_DESC(page_tbl_read_gst, "csource=0x0e"), + AMD_IOMMU_EVENT_DESC(int_dte_hit, "csource=0x0f"), + AMD_IOMMU_EVENT_DESC(int_dte_mis, "csource=0x10"), + AMD_IOMMU_EVENT_DESC(cmd_processed, "csource=0x11"), + AMD_IOMMU_EVENT_DESC(cmd_processed_inv, "csource=0x12"), + AMD_IOMMU_EVENT_DESC(tlb_inv, "csource=0x13"), + { /* end: all zeroes */ }, +}; + +/*--------------------------------------------- + * sysfs cpumask attributes + *---------------------------------------------*/ +static cpumask_t iommu_cpumask; + +static ssize_t _iommu_cpumask_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + int n = cpulist_scnprintf(buf, PAGE_SIZE - 2, &iommu_cpumask); + buf[n++] = '\n'; + buf[n] = '\0'; + return n; +} +static DEVICE_ATTR(cpumask, S_IRUGO, _iommu_cpumask_show, NULL); + +static struct attribute *iommu_cpumask_attrs[] = { + &dev_attr_cpumask.attr, + NULL, +}; + +static struct attribute_group amd_iommu_cpumask_group = { + .attrs = iommu_cpumask_attrs, +}; + +/*---------------------------------------------*/ + +static int get_next_avail_iommu_bnk_cntr(struct perf_amd_iommu *perf_iommu) +{ + unsigned long flags; + int shift, bank, cntr, retval; + int max_banks = perf_iommu->max_banks; + int max_cntrs = perf_iommu->max_counters; + + raw_spin_lock_irqsave(&perf_iommu->lock, flags); + + for (bank = 0, shift = 0; bank < max_banks; bank++) { + for (cntr = 0; cntr < max_cntrs; cntr++) { + shift = bank + (bank*3) + cntr; + if (perf_iommu->cntr_assign_mask & (1ULL<<shift)) { + continue; + } else { + perf_iommu->cntr_assign_mask |= (1ULL<<shift); + retval = ((u16)((u16)bank<<8) | (u8)(cntr)); + goto out; + } + } + } + retval = -ENOSPC; +out: + raw_spin_unlock_irqrestore(&perf_iommu->lock, flags); + return retval; +} + +static int clear_avail_iommu_bnk_cntr(struct perf_amd_iommu *perf_iommu, + u8 bank, u8 cntr) +{ + unsigned long flags; + int max_banks, max_cntrs; + int shift = 0; + + max_banks = perf_iommu->max_banks; + max_cntrs = perf_iommu->max_counters; + + if ((bank > max_banks) || (cntr > max_cntrs)) + return -EINVAL; + + shift = bank + cntr + (bank*3); + + raw_spin_lock_irqsave(&perf_iommu->lock, flags); + perf_iommu->cntr_assign_mask &= ~(1ULL<<shift); + raw_spin_unlock_irqrestore(&perf_iommu->lock, flags); + + return 0; +} + +static int perf_iommu_event_init(struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct perf_amd_iommu *perf_iommu; + u64 config, config1; + + /* test the event attr type check for PMU enumeration */ + if (event->attr.type != event->pmu->type) + return -ENOENT; + + /* + * IOMMU counters are shared across all cores. + * Therefore, it does not support per-process mode. + * Also, it does not support event sampling mode. + */ + if (is_sampling_event(event) || event->attach_state & PERF_ATTACH_TASK) + return -EINVAL; + + /* IOMMU counters do not have usr/os/guest/host bits */ + if (event->attr.exclude_user || event->attr.exclude_kernel || + event->attr.exclude_host || event->attr.exclude_guest) + return -EINVAL; + + if (event->cpu < 0) + return -EINVAL; + + perf_iommu = &__perf_iommu; + + if (event->pmu != &perf_iommu->pmu) + return -ENOENT; + + if (perf_iommu) { + config = event->attr.config; + config1 = event->attr.config1; + } else { + return -EINVAL; + } + + /* integrate with iommu base devid (0000), assume one iommu */ + perf_iommu->max_banks = + amd_iommu_pc_get_max_banks(IOMMU_BASE_DEVID); + perf_iommu->max_counters = + amd_iommu_pc_get_max_counters(IOMMU_BASE_DEVID); + if ((perf_iommu->max_banks == 0) || (perf_iommu->max_counters == 0)) + return -EINVAL; + + /* update the hw_perf_event struct with the iommu config data */ + hwc->config = config; + hwc->extra_reg.config = config1; + + return 0; +} + +static void perf_iommu_enable_event(struct perf_event *ev) +{ + u8 csource = _GET_CSOURCE(ev); + u16 devid = _GET_DEVID(ev); + u64 reg = 0ULL; + + reg = csource; + amd_iommu_pc_get_set_reg_val(devid, + _GET_BANK(ev), _GET_CNTR(ev) , + IOMMU_PC_COUNTER_SRC_REG, ®, true); + + reg = 0ULL | devid | (_GET_DEVID_MASK(ev) << 32); + if (reg) + reg |= (1UL << 31); + amd_iommu_pc_get_set_reg_val(devid, + _GET_BANK(ev), _GET_CNTR(ev) , + IOMMU_PC_DEVID_MATCH_REG, ®, true); + + reg = 0ULL | _GET_PASID(ev) | (_GET_PASID_MASK(ev) << 32); + if (reg) + reg |= (1UL << 31); + amd_iommu_pc_get_set_reg_val(devid, + _GET_BANK(ev), _GET_CNTR(ev) , + IOMMU_PC_PASID_MATCH_REG, ®, true); + + reg = 0ULL | _GET_DOMID(ev) | (_GET_DOMID_MASK(ev) << 32); + if (reg) + reg |= (1UL << 31); + amd_iommu_pc_get_set_reg_val(devid, + _GET_BANK(ev), _GET_CNTR(ev) , + IOMMU_PC_DOMID_MATCH_REG, ®, true); +} + +static void perf_iommu_disable_event(struct perf_event *event) +{ + u64 reg = 0ULL; + + amd_iommu_pc_get_set_reg_val(_GET_DEVID(event), + _GET_BANK(event), _GET_CNTR(event), + IOMMU_PC_COUNTER_SRC_REG, ®, true); +} + +static void perf_iommu_start(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + + pr_debug("perf: amd_iommu:perf_iommu_start\n"); + if (WARN_ON_ONCE(!(hwc->state & PERF_HES_STOPPED))) + return; + + WARN_ON_ONCE(!(hwc->state & PERF_HES_UPTODATE)); + hwc->state = 0; + + if (flags & PERF_EF_RELOAD) { + u64 prev_raw_count = local64_read(&hwc->prev_count); + amd_iommu_pc_get_set_reg_val(_GET_DEVID(event), + _GET_BANK(event), _GET_CNTR(event), + IOMMU_PC_COUNTER_REG, &prev_raw_count, true); + } + + perf_iommu_enable_event(event); + perf_event_update_userpage(event); + +} + +static void perf_iommu_read(struct perf_event *event) +{ + u64 count = 0ULL; + u64 prev_raw_count = 0ULL; + u64 delta = 0ULL; + struct hw_perf_event *hwc = &event->hw; + pr_debug("perf: amd_iommu:perf_iommu_read\n"); + + amd_iommu_pc_get_set_reg_val(_GET_DEVID(event), + _GET_BANK(event), _GET_CNTR(event), + IOMMU_PC_COUNTER_REG, &count, false); + + /* IOMMU pc counter register is only 48 bits */ + count &= 0xFFFFFFFFFFFFULL; + + prev_raw_count = local64_read(&hwc->prev_count); + if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, + count) != prev_raw_count) + return; + + /* Handling 48-bit counter overflowing */ + delta = (count << COUNTER_SHIFT) - (prev_raw_count << COUNTER_SHIFT); + delta >>= COUNTER_SHIFT; + local64_add(delta, &event->count); + +} + +static void perf_iommu_stop(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + u64 config; + + pr_debug("perf: amd_iommu:perf_iommu_stop\n"); + + if (hwc->state & PERF_HES_UPTODATE) + return; + + perf_iommu_disable_event(event); + WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED); + hwc->state |= PERF_HES_STOPPED; + + if (hwc->state & PERF_HES_UPTODATE) + return; + + config = hwc->config; + perf_iommu_read(event); + hwc->state |= PERF_HES_UPTODATE; +} + +static int perf_iommu_add(struct perf_event *event, int flags) +{ + int retval; + struct perf_amd_iommu *perf_iommu = + container_of(event->pmu, struct perf_amd_iommu, pmu); + + pr_debug("perf: amd_iommu:perf_iommu_add\n"); + event->hw.state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + + /* request an iommu bank/counter */ + retval = get_next_avail_iommu_bnk_cntr(perf_iommu); + if (retval != -ENOSPC) + event->hw.extra_reg.reg = (u16)retval; + else + return retval; + + if (flags & PERF_EF_START) + perf_iommu_start(event, PERF_EF_RELOAD); + + return 0; +} + +static void perf_iommu_del(struct perf_event *event, int flags) +{ + struct perf_amd_iommu *perf_iommu = + container_of(event->pmu, struct perf_amd_iommu, pmu); + + pr_debug("perf: amd_iommu:perf_iommu_del\n"); + perf_iommu_stop(event, PERF_EF_UPDATE); + + /* clear the assigned iommu bank/counter */ + clear_avail_iommu_bnk_cntr(perf_iommu, + _GET_BANK(event), + _GET_CNTR(event)); + + perf_event_update_userpage(event); +} + +static __init int _init_events_attrs(struct perf_amd_iommu *perf_iommu) +{ + struct attribute **attrs; + struct attribute_group *attr_group; + int i = 0, j; + + while (amd_iommu_v2_event_descs[i].attr.attr.name) + i++; + + attr_group = kzalloc(sizeof(struct attribute *) + * (i + 1) + sizeof(*attr_group), GFP_KERNEL); + if (!attr_group) + return -ENOMEM; + + attrs = (struct attribute **)(attr_group + 1); + for (j = 0; j < i; j++) + attrs[j] = &amd_iommu_v2_event_descs[j].attr.attr; + + attr_group->name = "events"; + attr_group->attrs = attrs; + perf_iommu->events_group = attr_group; + + return 0; +} + +static __init void amd_iommu_pc_exit(void) +{ + if (__perf_iommu.events_group != NULL) { + kfree(__perf_iommu.events_group); + __perf_iommu.events_group = NULL; + } +} + +static __init int _init_perf_amd_iommu( + struct perf_amd_iommu *perf_iommu, char *name) +{ + int ret; + + raw_spin_lock_init(&perf_iommu->lock); + + /* Init format attributes */ + perf_iommu->format_group = &amd_iommu_format_group; + + /* Init cpumask attributes to only core 0 */ + cpumask_set_cpu(0, &iommu_cpumask); + perf_iommu->cpumask_group = &amd_iommu_cpumask_group; + + /* Init events attributes */ + if (_init_events_attrs(perf_iommu) != 0) + pr_err("perf: amd_iommu: Only support raw events.\n"); + + /* Init null attributes */ + perf_iommu->null_group = NULL; + perf_iommu->pmu.attr_groups = perf_iommu->attr_groups; + + ret = perf_pmu_register(&perf_iommu->pmu, name, -1); + if (ret) { + pr_err("perf: amd_iommu: Failed to initialized.\n"); + amd_iommu_pc_exit(); + } else { + pr_info("perf: amd_iommu: Detected. (%d banks, %d counters/bank)\n", + amd_iommu_pc_get_max_banks(IOMMU_BASE_DEVID), + amd_iommu_pc_get_max_counters(IOMMU_BASE_DEVID)); + } + + return ret; +} + +static struct perf_amd_iommu __perf_iommu = { + .pmu = { + .event_init = perf_iommu_event_init, + .add = perf_iommu_add, + .del = perf_iommu_del, + .start = perf_iommu_start, + .stop = perf_iommu_stop, + .read = perf_iommu_read, + }, + .max_banks = 0x00, + .max_counters = 0x00, + .cntr_assign_mask = 0ULL, + .format_group = NULL, + .cpumask_group = NULL, + .events_group = NULL, + .null_group = NULL, +}; + +static __init int amd_iommu_pc_init(void) +{ + /* Make sure the IOMMU PC resource is available */ + if (!amd_iommu_pc_supported()) + return -ENODEV; + + _init_perf_amd_iommu(&__perf_iommu, "amd_iommu"); + + return 0; +} + +device_initcall(amd_iommu_pc_init); diff --git a/arch/x86/kernel/cpu/perf_event_amd_iommu.h b/arch/x86/kernel/cpu/perf_event_amd_iommu.h new file mode 100644 index 00000000000..845d173278e --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_amd_iommu.h @@ -0,0 +1,40 @@ +/* + * Copyright (C) 2013 Advanced Micro Devices, Inc. + * + * Author: Steven Kinney <Steven.Kinney@amd.com> + * Author: Suravee Suthikulpanit <Suraveee.Suthikulpanit@amd.com> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#ifndef _PERF_EVENT_AMD_IOMMU_H_ +#define _PERF_EVENT_AMD_IOMMU_H_ + +/* iommu pc mmio region register indexes */ +#define IOMMU_PC_COUNTER_REG 0x00 +#define IOMMU_PC_COUNTER_SRC_REG 0x08 +#define IOMMU_PC_PASID_MATCH_REG 0x10 +#define IOMMU_PC_DOMID_MATCH_REG 0x18 +#define IOMMU_PC_DEVID_MATCH_REG 0x20 +#define IOMMU_PC_COUNTER_REPORT_REG 0x28 + +/* maximun specified bank/counters */ +#define PC_MAX_SPEC_BNKS 64 +#define PC_MAX_SPEC_CNTRS 16 + +/* iommu pc reg masks*/ +#define IOMMU_BASE_DEVID 0x0000 + +/* amd_iommu_init.c external support functions */ +extern bool amd_iommu_pc_supported(void); + +extern u8 amd_iommu_pc_get_max_banks(u16 devid); + +extern u8 amd_iommu_pc_get_max_counters(u16 devid); + +extern int amd_iommu_pc_get_set_reg_val(u16 devid, u8 bank, u8 cntr, + u8 fxn, u64 *value, bool is_write); + +#endif /*_PERF_EVENT_AMD_IOMMU_H_*/ diff --git a/arch/x86/kernel/cpu/perf_event_amd_uncore.c b/arch/x86/kernel/cpu/perf_event_amd_uncore.c new file mode 100644 index 00000000000..3bbdf4cd38b --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_amd_uncore.c @@ -0,0 +1,547 @@ +/* + * Copyright (C) 2013 Advanced Micro Devices, Inc. + * + * Author: Jacob Shin <jacob.shin@amd.com> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/perf_event.h> +#include <linux/percpu.h> +#include <linux/types.h> +#include <linux/slab.h> +#include <linux/init.h> +#include <linux/cpu.h> +#include <linux/cpumask.h> + +#include <asm/cpufeature.h> +#include <asm/perf_event.h> +#include <asm/msr.h> + +#define NUM_COUNTERS_NB 4 +#define NUM_COUNTERS_L2 4 +#define MAX_COUNTERS NUM_COUNTERS_NB + +#define RDPMC_BASE_NB 6 +#define RDPMC_BASE_L2 10 + +#define COUNTER_SHIFT 16 + +struct amd_uncore { + int id; + int refcnt; + int cpu; + int num_counters; + int rdpmc_base; + u32 msr_base; + cpumask_t *active_mask; + struct pmu *pmu; + struct perf_event *events[MAX_COUNTERS]; + struct amd_uncore *free_when_cpu_online; +}; + +static struct amd_uncore * __percpu *amd_uncore_nb; +static struct amd_uncore * __percpu *amd_uncore_l2; + +static struct pmu amd_nb_pmu; +static struct pmu amd_l2_pmu; + +static cpumask_t amd_nb_active_mask; +static cpumask_t amd_l2_active_mask; + +static bool is_nb_event(struct perf_event *event) +{ + return event->pmu->type == amd_nb_pmu.type; +} + +static bool is_l2_event(struct perf_event *event) +{ + return event->pmu->type == amd_l2_pmu.type; +} + +static struct amd_uncore *event_to_amd_uncore(struct perf_event *event) +{ + if (is_nb_event(event) && amd_uncore_nb) + return *per_cpu_ptr(amd_uncore_nb, event->cpu); + else if (is_l2_event(event) && amd_uncore_l2) + return *per_cpu_ptr(amd_uncore_l2, event->cpu); + + return NULL; +} + +static void amd_uncore_read(struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + u64 prev, new; + s64 delta; + + /* + * since we do not enable counter overflow interrupts, + * we do not have to worry about prev_count changing on us + */ + + prev = local64_read(&hwc->prev_count); + rdpmcl(hwc->event_base_rdpmc, new); + local64_set(&hwc->prev_count, new); + delta = (new << COUNTER_SHIFT) - (prev << COUNTER_SHIFT); + delta >>= COUNTER_SHIFT; + local64_add(delta, &event->count); +} + +static void amd_uncore_start(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + + if (flags & PERF_EF_RELOAD) + wrmsrl(hwc->event_base, (u64)local64_read(&hwc->prev_count)); + + hwc->state = 0; + wrmsrl(hwc->config_base, (hwc->config | ARCH_PERFMON_EVENTSEL_ENABLE)); + perf_event_update_userpage(event); +} + +static void amd_uncore_stop(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + + wrmsrl(hwc->config_base, hwc->config); + hwc->state |= PERF_HES_STOPPED; + + if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) { + amd_uncore_read(event); + hwc->state |= PERF_HES_UPTODATE; + } +} + +static int amd_uncore_add(struct perf_event *event, int flags) +{ + int i; + struct amd_uncore *uncore = event_to_amd_uncore(event); + struct hw_perf_event *hwc = &event->hw; + + /* are we already assigned? */ + if (hwc->idx != -1 && uncore->events[hwc->idx] == event) + goto out; + + for (i = 0; i < uncore->num_counters; i++) { + if (uncore->events[i] == event) { + hwc->idx = i; + goto out; + } + } + + /* if not, take the first available counter */ + hwc->idx = -1; + for (i = 0; i < uncore->num_counters; i++) { + if (cmpxchg(&uncore->events[i], NULL, event) == NULL) { + hwc->idx = i; + break; + } + } + +out: + if (hwc->idx == -1) + return -EBUSY; + + hwc->config_base = uncore->msr_base + (2 * hwc->idx); + hwc->event_base = uncore->msr_base + 1 + (2 * hwc->idx); + hwc->event_base_rdpmc = uncore->rdpmc_base + hwc->idx; + hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + + if (flags & PERF_EF_START) + amd_uncore_start(event, PERF_EF_RELOAD); + + return 0; +} + +static void amd_uncore_del(struct perf_event *event, int flags) +{ + int i; + struct amd_uncore *uncore = event_to_amd_uncore(event); + struct hw_perf_event *hwc = &event->hw; + + amd_uncore_stop(event, PERF_EF_UPDATE); + + for (i = 0; i < uncore->num_counters; i++) { + if (cmpxchg(&uncore->events[i], event, NULL) == event) + break; + } + + hwc->idx = -1; +} + +static int amd_uncore_event_init(struct perf_event *event) +{ + struct amd_uncore *uncore; + struct hw_perf_event *hwc = &event->hw; + + if (event->attr.type != event->pmu->type) + return -ENOENT; + + /* + * NB and L2 counters (MSRs) are shared across all cores that share the + * same NB / L2 cache. Interrupts can be directed to a single target + * core, however, event counts generated by processes running on other + * cores cannot be masked out. So we do not support sampling and + * per-thread events. + */ + if (is_sampling_event(event) || event->attach_state & PERF_ATTACH_TASK) + return -EINVAL; + + /* NB and L2 counters do not have usr/os/guest/host bits */ + if (event->attr.exclude_user || event->attr.exclude_kernel || + event->attr.exclude_host || event->attr.exclude_guest) + return -EINVAL; + + /* and we do not enable counter overflow interrupts */ + hwc->config = event->attr.config & AMD64_RAW_EVENT_MASK_NB; + hwc->idx = -1; + + if (event->cpu < 0) + return -EINVAL; + + uncore = event_to_amd_uncore(event); + if (!uncore) + return -ENODEV; + + /* + * since request can come in to any of the shared cores, we will remap + * to a single common cpu. + */ + event->cpu = uncore->cpu; + + return 0; +} + +static ssize_t amd_uncore_attr_show_cpumask(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + int n; + cpumask_t *active_mask; + struct pmu *pmu = dev_get_drvdata(dev); + + if (pmu->type == amd_nb_pmu.type) + active_mask = &amd_nb_active_mask; + else if (pmu->type == amd_l2_pmu.type) + active_mask = &amd_l2_active_mask; + else + return 0; + + n = cpulist_scnprintf(buf, PAGE_SIZE - 2, active_mask); + buf[n++] = '\n'; + buf[n] = '\0'; + return n; +} +static DEVICE_ATTR(cpumask, S_IRUGO, amd_uncore_attr_show_cpumask, NULL); + +static struct attribute *amd_uncore_attrs[] = { + &dev_attr_cpumask.attr, + NULL, +}; + +static struct attribute_group amd_uncore_attr_group = { + .attrs = amd_uncore_attrs, +}; + +PMU_FORMAT_ATTR(event, "config:0-7,32-35"); +PMU_FORMAT_ATTR(umask, "config:8-15"); + +static struct attribute *amd_uncore_format_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + NULL, +}; + +static struct attribute_group amd_uncore_format_group = { + .name = "format", + .attrs = amd_uncore_format_attr, +}; + +static const struct attribute_group *amd_uncore_attr_groups[] = { + &amd_uncore_attr_group, + &amd_uncore_format_group, + NULL, +}; + +static struct pmu amd_nb_pmu = { + .attr_groups = amd_uncore_attr_groups, + .name = "amd_nb", + .event_init = amd_uncore_event_init, + .add = amd_uncore_add, + .del = amd_uncore_del, + .start = amd_uncore_start, + .stop = amd_uncore_stop, + .read = amd_uncore_read, +}; + +static struct pmu amd_l2_pmu = { + .attr_groups = amd_uncore_attr_groups, + .name = "amd_l2", + .event_init = amd_uncore_event_init, + .add = amd_uncore_add, + .del = amd_uncore_del, + .start = amd_uncore_start, + .stop = amd_uncore_stop, + .read = amd_uncore_read, +}; + +static struct amd_uncore *amd_uncore_alloc(unsigned int cpu) +{ + return kzalloc_node(sizeof(struct amd_uncore), GFP_KERNEL, + cpu_to_node(cpu)); +} + +static void amd_uncore_cpu_up_prepare(unsigned int cpu) +{ + struct amd_uncore *uncore; + + if (amd_uncore_nb) { + uncore = amd_uncore_alloc(cpu); + uncore->cpu = cpu; + uncore->num_counters = NUM_COUNTERS_NB; + uncore->rdpmc_base = RDPMC_BASE_NB; + uncore->msr_base = MSR_F15H_NB_PERF_CTL; + uncore->active_mask = &amd_nb_active_mask; + uncore->pmu = &amd_nb_pmu; + *per_cpu_ptr(amd_uncore_nb, cpu) = uncore; + } + + if (amd_uncore_l2) { + uncore = amd_uncore_alloc(cpu); + uncore->cpu = cpu; + uncore->num_counters = NUM_COUNTERS_L2; + uncore->rdpmc_base = RDPMC_BASE_L2; + uncore->msr_base = MSR_F16H_L2I_PERF_CTL; + uncore->active_mask = &amd_l2_active_mask; + uncore->pmu = &amd_l2_pmu; + *per_cpu_ptr(amd_uncore_l2, cpu) = uncore; + } +} + +static struct amd_uncore * +amd_uncore_find_online_sibling(struct amd_uncore *this, + struct amd_uncore * __percpu *uncores) +{ + unsigned int cpu; + struct amd_uncore *that; + + for_each_online_cpu(cpu) { + that = *per_cpu_ptr(uncores, cpu); + + if (!that) + continue; + + if (this == that) + continue; + + if (this->id == that->id) { + that->free_when_cpu_online = this; + this = that; + break; + } + } + + this->refcnt++; + return this; +} + +static void amd_uncore_cpu_starting(unsigned int cpu) +{ + unsigned int eax, ebx, ecx, edx; + struct amd_uncore *uncore; + + if (amd_uncore_nb) { + uncore = *per_cpu_ptr(amd_uncore_nb, cpu); + cpuid(0x8000001e, &eax, &ebx, &ecx, &edx); + uncore->id = ecx & 0xff; + + uncore = amd_uncore_find_online_sibling(uncore, amd_uncore_nb); + *per_cpu_ptr(amd_uncore_nb, cpu) = uncore; + } + + if (amd_uncore_l2) { + unsigned int apicid = cpu_data(cpu).apicid; + unsigned int nshared; + + uncore = *per_cpu_ptr(amd_uncore_l2, cpu); + cpuid_count(0x8000001d, 2, &eax, &ebx, &ecx, &edx); + nshared = ((eax >> 14) & 0xfff) + 1; + uncore->id = apicid - (apicid % nshared); + + uncore = amd_uncore_find_online_sibling(uncore, amd_uncore_l2); + *per_cpu_ptr(amd_uncore_l2, cpu) = uncore; + } +} + +static void uncore_online(unsigned int cpu, + struct amd_uncore * __percpu *uncores) +{ + struct amd_uncore *uncore = *per_cpu_ptr(uncores, cpu); + + kfree(uncore->free_when_cpu_online); + uncore->free_when_cpu_online = NULL; + + if (cpu == uncore->cpu) + cpumask_set_cpu(cpu, uncore->active_mask); +} + +static void amd_uncore_cpu_online(unsigned int cpu) +{ + if (amd_uncore_nb) + uncore_online(cpu, amd_uncore_nb); + + if (amd_uncore_l2) + uncore_online(cpu, amd_uncore_l2); +} + +static void uncore_down_prepare(unsigned int cpu, + struct amd_uncore * __percpu *uncores) +{ + unsigned int i; + struct amd_uncore *this = *per_cpu_ptr(uncores, cpu); + + if (this->cpu != cpu) + return; + + /* this cpu is going down, migrate to a shared sibling if possible */ + for_each_online_cpu(i) { + struct amd_uncore *that = *per_cpu_ptr(uncores, i); + + if (cpu == i) + continue; + + if (this == that) { + perf_pmu_migrate_context(this->pmu, cpu, i); + cpumask_clear_cpu(cpu, that->active_mask); + cpumask_set_cpu(i, that->active_mask); + that->cpu = i; + break; + } + } +} + +static void amd_uncore_cpu_down_prepare(unsigned int cpu) +{ + if (amd_uncore_nb) + uncore_down_prepare(cpu, amd_uncore_nb); + + if (amd_uncore_l2) + uncore_down_prepare(cpu, amd_uncore_l2); +} + +static void uncore_dead(unsigned int cpu, struct amd_uncore * __percpu *uncores) +{ + struct amd_uncore *uncore = *per_cpu_ptr(uncores, cpu); + + if (cpu == uncore->cpu) + cpumask_clear_cpu(cpu, uncore->active_mask); + + if (!--uncore->refcnt) + kfree(uncore); + *per_cpu_ptr(amd_uncore_nb, cpu) = NULL; +} + +static void amd_uncore_cpu_dead(unsigned int cpu) +{ + if (amd_uncore_nb) + uncore_dead(cpu, amd_uncore_nb); + + if (amd_uncore_l2) + uncore_dead(cpu, amd_uncore_l2); +} + +static int +amd_uncore_cpu_notifier(struct notifier_block *self, unsigned long action, + void *hcpu) +{ + unsigned int cpu = (long)hcpu; + + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_UP_PREPARE: + amd_uncore_cpu_up_prepare(cpu); + break; + + case CPU_STARTING: + amd_uncore_cpu_starting(cpu); + break; + + case CPU_ONLINE: + amd_uncore_cpu_online(cpu); + break; + + case CPU_DOWN_PREPARE: + amd_uncore_cpu_down_prepare(cpu); + break; + + case CPU_UP_CANCELED: + case CPU_DEAD: + amd_uncore_cpu_dead(cpu); + break; + + default: + break; + } + + return NOTIFY_OK; +} + +static struct notifier_block amd_uncore_cpu_notifier_block = { + .notifier_call = amd_uncore_cpu_notifier, + .priority = CPU_PRI_PERF + 1, +}; + +static void __init init_cpu_already_online(void *dummy) +{ + unsigned int cpu = smp_processor_id(); + + amd_uncore_cpu_starting(cpu); + amd_uncore_cpu_online(cpu); +} + +static int __init amd_uncore_init(void) +{ + unsigned int cpu; + int ret = -ENODEV; + + if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD) + return -ENODEV; + + if (!cpu_has_topoext) + return -ENODEV; + + if (cpu_has_perfctr_nb) { + amd_uncore_nb = alloc_percpu(struct amd_uncore *); + perf_pmu_register(&amd_nb_pmu, amd_nb_pmu.name, -1); + + printk(KERN_INFO "perf: AMD NB counters detected\n"); + ret = 0; + } + + if (cpu_has_perfctr_l2) { + amd_uncore_l2 = alloc_percpu(struct amd_uncore *); + perf_pmu_register(&amd_l2_pmu, amd_l2_pmu.name, -1); + + printk(KERN_INFO "perf: AMD L2I counters detected\n"); + ret = 0; + } + + if (ret) + return -ENODEV; + + cpu_notifier_register_begin(); + + /* init cpus already online before registering for hotplug notifier */ + for_each_online_cpu(cpu) { + amd_uncore_cpu_up_prepare(cpu); + smp_call_function_single(cpu, init_cpu_already_online, NULL, 1); + } + + __register_cpu_notifier(&amd_uncore_cpu_notifier_block); + cpu_notifier_register_done(); + + return 0; +} +device_initcall(amd_uncore_init); diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 61d4f79a550..2502d0d9d24 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -5,12 +5,15 @@ * among events on a single PMU. */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/stddef.h> #include <linux/types.h> #include <linux/init.h> #include <linux/slab.h> #include <linux/export.h> +#include <asm/cpufeature.h> #include <asm/hardirq.h> #include <asm/apic.h> @@ -21,14 +24,14 @@ */ static u64 intel_perfmon_event_map[PERF_COUNT_HW_MAX] __read_mostly = { - [PERF_COUNT_HW_CPU_CYCLES] = 0x003c, - [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, - [PERF_COUNT_HW_CACHE_REFERENCES] = 0x4f2e, - [PERF_COUNT_HW_CACHE_MISSES] = 0x412e, - [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, - [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, - [PERF_COUNT_HW_BUS_CYCLES] = 0x013c, - [PERF_COUNT_HW_REF_CPU_CYCLES] = 0x0300, /* pseudo-encoding */ + [PERF_COUNT_HW_CPU_CYCLES] = 0x003c, + [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, + [PERF_COUNT_HW_CACHE_REFERENCES] = 0x4f2e, + [PERF_COUNT_HW_CACHE_MISSES] = 0x412e, + [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, + [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, + [PERF_COUNT_HW_BUS_CYCLES] = 0x013c, + [PERF_COUNT_HW_REF_CPU_CYCLES] = 0x0300, /* pseudo-encoding */ }; static struct event_constraint intel_core_event_constraints[] __read_mostly = @@ -78,7 +81,9 @@ static struct event_constraint intel_nehalem_event_constraints[] __read_mostly = static struct extra_reg intel_nehalem_extra_regs[] __read_mostly = { - INTEL_EVENT_EXTRA_REG(0xb7, MSR_OFFCORE_RSP_0, 0xffff, RSP_0), + /* must define OFFCORE_RSP_X first, see intel_fixup_er() */ + INTEL_UEVENT_EXTRA_REG(0x01b7, MSR_OFFCORE_RSP_0, 0xffff, RSP_0), + INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG(0x100b), EVENT_EXTRA_END }; @@ -99,16 +104,51 @@ static struct event_constraint intel_snb_event_constraints[] __read_mostly = FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ + INTEL_UEVENT_CONSTRAINT(0x04a3, 0xf), /* CYCLE_ACTIVITY.CYCLES_NO_DISPATCH */ + INTEL_UEVENT_CONSTRAINT(0x05a3, 0xf), /* CYCLE_ACTIVITY.STALLS_L2_PENDING */ + INTEL_UEVENT_CONSTRAINT(0x02a3, 0x4), /* CYCLE_ACTIVITY.CYCLES_L1D_PENDING */ + INTEL_UEVENT_CONSTRAINT(0x06a3, 0x4), /* CYCLE_ACTIVITY.STALLS_L1D_PENDING */ INTEL_EVENT_CONSTRAINT(0x48, 0x4), /* L1D_PEND_MISS.PENDING */ INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PREC_DIST */ INTEL_EVENT_CONSTRAINT(0xcd, 0x8), /* MEM_TRANS_RETIRED.LOAD_LATENCY */ + INTEL_UEVENT_CONSTRAINT(0x04a3, 0xf), /* CYCLE_ACTIVITY.CYCLES_NO_DISPATCH */ + INTEL_UEVENT_CONSTRAINT(0x02a3, 0x4), /* CYCLE_ACTIVITY.CYCLES_L1D_PENDING */ + EVENT_CONSTRAINT_END +}; + +static struct event_constraint intel_ivb_event_constraints[] __read_mostly = +{ + FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ + FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ + INTEL_UEVENT_CONSTRAINT(0x0148, 0x4), /* L1D_PEND_MISS.PENDING */ + INTEL_UEVENT_CONSTRAINT(0x0279, 0xf), /* IDQ.EMTPY */ + INTEL_UEVENT_CONSTRAINT(0x019c, 0xf), /* IDQ_UOPS_NOT_DELIVERED.CORE */ + INTEL_UEVENT_CONSTRAINT(0x02a3, 0xf), /* CYCLE_ACTIVITY.CYCLES_LDM_PENDING */ + INTEL_UEVENT_CONSTRAINT(0x04a3, 0xf), /* CYCLE_ACTIVITY.CYCLES_NO_EXECUTE */ + INTEL_UEVENT_CONSTRAINT(0x05a3, 0xf), /* CYCLE_ACTIVITY.STALLS_L2_PENDING */ + INTEL_UEVENT_CONSTRAINT(0x06a3, 0xf), /* CYCLE_ACTIVITY.STALLS_LDM_PENDING */ + INTEL_UEVENT_CONSTRAINT(0x08a3, 0x4), /* CYCLE_ACTIVITY.CYCLES_L1D_PENDING */ + INTEL_UEVENT_CONSTRAINT(0x0ca3, 0x4), /* CYCLE_ACTIVITY.STALLS_L1D_PENDING */ + INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PREC_DIST */ + /* + * Errata BV98 -- MEM_*_RETIRED events can leak between counters of SMT + * siblings; disable these events because they can corrupt unrelated + * counters. + */ + INTEL_EVENT_CONSTRAINT(0xd0, 0x0), /* MEM_UOPS_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xd1, 0x0), /* MEM_LOAD_UOPS_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xd2, 0x0), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xd3, 0x0), /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.* */ EVENT_CONSTRAINT_END }; static struct extra_reg intel_westmere_extra_regs[] __read_mostly = { - INTEL_EVENT_EXTRA_REG(0xb7, MSR_OFFCORE_RSP_0, 0xffff, RSP_0), - INTEL_EVENT_EXTRA_REG(0xbb, MSR_OFFCORE_RSP_1, 0xffff, RSP_1), + /* must define OFFCORE_RSP_X first, see intel_fixup_er() */ + INTEL_UEVENT_EXTRA_REG(0x01b7, MSR_OFFCORE_RSP_0, 0xffff, RSP_0), + INTEL_UEVENT_EXTRA_REG(0x01bb, MSR_OFFCORE_RSP_1, 0xffff, RSP_1), + INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG(0x100b), EVENT_EXTRA_END }; @@ -125,17 +165,144 @@ static struct event_constraint intel_gen_event_constraints[] __read_mostly = EVENT_CONSTRAINT_END }; +static struct event_constraint intel_slm_event_constraints[] __read_mostly = +{ + FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ + FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* pseudo CPU_CLK_UNHALTED.REF */ + EVENT_CONSTRAINT_END +}; + static struct extra_reg intel_snb_extra_regs[] __read_mostly = { - INTEL_EVENT_EXTRA_REG(0xb7, MSR_OFFCORE_RSP_0, 0x3fffffffffull, RSP_0), - INTEL_EVENT_EXTRA_REG(0xbb, MSR_OFFCORE_RSP_1, 0x3fffffffffull, RSP_1), + /* must define OFFCORE_RSP_X first, see intel_fixup_er() */ + INTEL_UEVENT_EXTRA_REG(0x01b7, MSR_OFFCORE_RSP_0, 0x3f807f8fffull, RSP_0), + INTEL_UEVENT_EXTRA_REG(0x01bb, MSR_OFFCORE_RSP_1, 0x3f807f8fffull, RSP_1), + INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG(0x01cd), EVENT_EXTRA_END }; +static struct extra_reg intel_snbep_extra_regs[] __read_mostly = { + /* must define OFFCORE_RSP_X first, see intel_fixup_er() */ + INTEL_UEVENT_EXTRA_REG(0x01b7, MSR_OFFCORE_RSP_0, 0x3fffff8fffull, RSP_0), + INTEL_UEVENT_EXTRA_REG(0x01bb, MSR_OFFCORE_RSP_1, 0x3fffff8fffull, RSP_1), + INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG(0x01cd), + EVENT_EXTRA_END +}; + +EVENT_ATTR_STR(mem-loads, mem_ld_nhm, "event=0x0b,umask=0x10,ldlat=3"); +EVENT_ATTR_STR(mem-loads, mem_ld_snb, "event=0xcd,umask=0x1,ldlat=3"); +EVENT_ATTR_STR(mem-stores, mem_st_snb, "event=0xcd,umask=0x2"); + +struct attribute *nhm_events_attrs[] = { + EVENT_PTR(mem_ld_nhm), + NULL, +}; + +struct attribute *snb_events_attrs[] = { + EVENT_PTR(mem_ld_snb), + EVENT_PTR(mem_st_snb), + NULL, +}; + +static struct event_constraint intel_hsw_event_constraints[] = { + FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ + FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */ + FIXED_EVENT_CONSTRAINT(0x0300, 2), /* CPU_CLK_UNHALTED.REF */ + INTEL_EVENT_CONSTRAINT(0x48, 0x4), /* L1D_PEND_MISS.* */ + INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PREC_DIST */ + INTEL_EVENT_CONSTRAINT(0xcd, 0x8), /* MEM_TRANS_RETIRED.LOAD_LATENCY */ + /* CYCLE_ACTIVITY.CYCLES_L1D_PENDING */ + INTEL_EVENT_CONSTRAINT(0x08a3, 0x4), + /* CYCLE_ACTIVITY.STALLS_L1D_PENDING */ + INTEL_EVENT_CONSTRAINT(0x0ca3, 0x4), + /* CYCLE_ACTIVITY.CYCLES_NO_EXECUTE */ + INTEL_EVENT_CONSTRAINT(0x04a3, 0xf), + EVENT_CONSTRAINT_END +}; + static u64 intel_pmu_event_map(int hw_event) { return intel_perfmon_event_map[hw_event]; } +#define SNB_DMND_DATA_RD (1ULL << 0) +#define SNB_DMND_RFO (1ULL << 1) +#define SNB_DMND_IFETCH (1ULL << 2) +#define SNB_DMND_WB (1ULL << 3) +#define SNB_PF_DATA_RD (1ULL << 4) +#define SNB_PF_RFO (1ULL << 5) +#define SNB_PF_IFETCH (1ULL << 6) +#define SNB_LLC_DATA_RD (1ULL << 7) +#define SNB_LLC_RFO (1ULL << 8) +#define SNB_LLC_IFETCH (1ULL << 9) +#define SNB_BUS_LOCKS (1ULL << 10) +#define SNB_STRM_ST (1ULL << 11) +#define SNB_OTHER (1ULL << 15) +#define SNB_RESP_ANY (1ULL << 16) +#define SNB_NO_SUPP (1ULL << 17) +#define SNB_LLC_HITM (1ULL << 18) +#define SNB_LLC_HITE (1ULL << 19) +#define SNB_LLC_HITS (1ULL << 20) +#define SNB_LLC_HITF (1ULL << 21) +#define SNB_LOCAL (1ULL << 22) +#define SNB_REMOTE (0xffULL << 23) +#define SNB_SNP_NONE (1ULL << 31) +#define SNB_SNP_NOT_NEEDED (1ULL << 32) +#define SNB_SNP_MISS (1ULL << 33) +#define SNB_NO_FWD (1ULL << 34) +#define SNB_SNP_FWD (1ULL << 35) +#define SNB_HITM (1ULL << 36) +#define SNB_NON_DRAM (1ULL << 37) + +#define SNB_DMND_READ (SNB_DMND_DATA_RD|SNB_LLC_DATA_RD) +#define SNB_DMND_WRITE (SNB_DMND_RFO|SNB_LLC_RFO) +#define SNB_DMND_PREFETCH (SNB_PF_DATA_RD|SNB_PF_RFO) + +#define SNB_SNP_ANY (SNB_SNP_NONE|SNB_SNP_NOT_NEEDED| \ + SNB_SNP_MISS|SNB_NO_FWD|SNB_SNP_FWD| \ + SNB_HITM) + +#define SNB_DRAM_ANY (SNB_LOCAL|SNB_REMOTE|SNB_SNP_ANY) +#define SNB_DRAM_REMOTE (SNB_REMOTE|SNB_SNP_ANY) + +#define SNB_L3_ACCESS SNB_RESP_ANY +#define SNB_L3_MISS (SNB_DRAM_ANY|SNB_NON_DRAM) + +static __initconst const u64 snb_hw_cache_extra_regs + [PERF_COUNT_HW_CACHE_MAX] + [PERF_COUNT_HW_CACHE_OP_MAX] + [PERF_COUNT_HW_CACHE_RESULT_MAX] = +{ + [ C(LL ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = SNB_DMND_READ|SNB_L3_ACCESS, + [ C(RESULT_MISS) ] = SNB_DMND_READ|SNB_L3_MISS, + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = SNB_DMND_WRITE|SNB_L3_ACCESS, + [ C(RESULT_MISS) ] = SNB_DMND_WRITE|SNB_L3_MISS, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = SNB_DMND_PREFETCH|SNB_L3_ACCESS, + [ C(RESULT_MISS) ] = SNB_DMND_PREFETCH|SNB_L3_MISS, + }, + }, + [ C(NODE) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = SNB_DMND_READ|SNB_DRAM_ANY, + [ C(RESULT_MISS) ] = SNB_DMND_READ|SNB_DRAM_REMOTE, + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = SNB_DMND_WRITE|SNB_DRAM_ANY, + [ C(RESULT_MISS) ] = SNB_DMND_WRITE|SNB_DRAM_REMOTE, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = SNB_DMND_PREFETCH|SNB_DRAM_ANY, + [ C(RESULT_MISS) ] = SNB_DMND_PREFETCH|SNB_DRAM_REMOTE, + }, + }, +}; + static __initconst const u64 snb_hw_cache_event_ids [PERF_COUNT_HW_CACHE_MAX] [PERF_COUNT_HW_CACHE_OP_MAX] @@ -233,16 +400,16 @@ static __initconst const u64 snb_hw_cache_event_ids }, [ C(NODE) ] = { [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, + [ C(RESULT_ACCESS) ] = 0x01b7, + [ C(RESULT_MISS) ] = 0x01b7, }, [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, + [ C(RESULT_ACCESS) ] = 0x01b7, + [ C(RESULT_MISS) ] = 0x01b7, }, [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = -1, - [ C(RESULT_MISS) ] = -1, + [ C(RESULT_ACCESS) ] = 0x01b7, + [ C(RESULT_MISS) ] = 0x01b7, }, }, @@ -728,13 +895,161 @@ static __initconst const u64 atom_hw_cache_event_ids }, }; +static struct extra_reg intel_slm_extra_regs[] __read_mostly = +{ + /* must define OFFCORE_RSP_X first, see intel_fixup_er() */ + INTEL_UEVENT_EXTRA_REG(0x01b7, MSR_OFFCORE_RSP_0, 0x768005ffffull, RSP_0), + INTEL_UEVENT_EXTRA_REG(0x02b7, MSR_OFFCORE_RSP_1, 0x768005ffffull, RSP_1), + EVENT_EXTRA_END +}; + +#define SLM_DMND_READ SNB_DMND_DATA_RD +#define SLM_DMND_WRITE SNB_DMND_RFO +#define SLM_DMND_PREFETCH (SNB_PF_DATA_RD|SNB_PF_RFO) + +#define SLM_SNP_ANY (SNB_SNP_NONE|SNB_SNP_MISS|SNB_NO_FWD|SNB_HITM) +#define SLM_LLC_ACCESS SNB_RESP_ANY +#define SLM_LLC_MISS (SLM_SNP_ANY|SNB_NON_DRAM) + +static __initconst const u64 slm_hw_cache_extra_regs + [PERF_COUNT_HW_CACHE_MAX] + [PERF_COUNT_HW_CACHE_OP_MAX] + [PERF_COUNT_HW_CACHE_RESULT_MAX] = +{ + [ C(LL ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = SLM_DMND_READ|SLM_LLC_ACCESS, + [ C(RESULT_MISS) ] = SLM_DMND_READ|SLM_LLC_MISS, + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = SLM_DMND_WRITE|SLM_LLC_ACCESS, + [ C(RESULT_MISS) ] = SLM_DMND_WRITE|SLM_LLC_MISS, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = SLM_DMND_PREFETCH|SLM_LLC_ACCESS, + [ C(RESULT_MISS) ] = SLM_DMND_PREFETCH|SLM_LLC_MISS, + }, + }, +}; + +static __initconst const u64 slm_hw_cache_event_ids + [PERF_COUNT_HW_CACHE_MAX] + [PERF_COUNT_HW_CACHE_OP_MAX] + [PERF_COUNT_HW_CACHE_RESULT_MAX] = +{ + [ C(L1D) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0x0104, /* LD_DCU_MISS */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + }, + [ C(L1I ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x0380, /* ICACHE.ACCESSES */ + [ C(RESULT_MISS) ] = 0x0280, /* ICACGE.MISSES */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + }, + [ C(LL ) ] = { + [ C(OP_READ) ] = { + /* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x01b7, + /* OFFCORE_RESPONSE.ANY_DATA.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x01b7, + }, + [ C(OP_WRITE) ] = { + /* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x01b7, + /* OFFCORE_RESPONSE.ANY_RFO.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x01b7, + }, + [ C(OP_PREFETCH) ] = { + /* OFFCORE_RESPONSE.PREFETCH.LOCAL_CACHE */ + [ C(RESULT_ACCESS) ] = 0x01b7, + /* OFFCORE_RESPONSE.PREFETCH.ANY_LLC_MISS */ + [ C(RESULT_MISS) ] = 0x01b7, + }, + }, + [ C(DTLB) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0x0804, /* LD_DTLB_MISS */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + }, + [ C(ITLB) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x00c0, /* INST_RETIRED.ANY_P */ + [ C(RESULT_MISS) ] = 0x0282, /* ITLB.MISSES */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + }, + [ C(BPU ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED.ANY */ + [ C(RESULT_MISS) ] = 0x00c5, /* BP_INST_RETIRED.MISPRED */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + }, +}; + +static inline bool intel_pmu_needs_lbr_smpl(struct perf_event *event) +{ + /* user explicitly requested branch sampling */ + if (has_branch_stack(event)) + return true; + + /* implicit branch sampling to correct PEBS skid */ + if (x86_pmu.intel_cap.pebs_trap && event->attr.precise_ip > 1 && + x86_pmu.intel_cap.pebs_format < 2) + return true; + + return false; +} + static void intel_pmu_disable_all(void) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0); - if (test_bit(X86_PMC_IDX_FIXED_BTS, cpuc->active_mask)) + if (test_bit(INTEL_PMC_IDX_FIXED_BTS, cpuc->active_mask)) intel_pmu_disable_bts(); intel_pmu_pebs_disable_all(); @@ -750,9 +1065,9 @@ static void intel_pmu_enable_all(int added) wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_guest_mask); - if (test_bit(X86_PMC_IDX_FIXED_BTS, cpuc->active_mask)) { + if (test_bit(INTEL_PMC_IDX_FIXED_BTS, cpuc->active_mask)) { struct perf_event *event = - cpuc->events[X86_PMC_IDX_FIXED_BTS]; + cpuc->events[INTEL_PMC_IDX_FIXED_BTS]; if (WARN_ON_ONCE(!event)) return; @@ -858,7 +1173,7 @@ static inline void intel_pmu_ack_status(u64 ack) static void intel_pmu_disable_fixed(struct hw_perf_event *hwc) { - int idx = hwc->idx - X86_PMC_IDX_FIXED; + int idx = hwc->idx - INTEL_PMC_IDX_FIXED; u64 ctrl_val, mask; mask = 0xfULL << (idx * 4); @@ -868,12 +1183,17 @@ static void intel_pmu_disable_fixed(struct hw_perf_event *hwc) wrmsrl(hwc->config_base, ctrl_val); } +static inline bool event_is_checkpointed(struct perf_event *event) +{ + return (event->hw.config & HSW_IN_TX_CHECKPOINTED) != 0; +} + static void intel_pmu_disable_event(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - if (unlikely(hwc->idx == X86_PMC_IDX_FIXED_BTS)) { + if (unlikely(hwc->idx == INTEL_PMC_IDX_FIXED_BTS)) { intel_pmu_disable_bts(); intel_pmu_drain_bts_buffer(); return; @@ -881,6 +1201,14 @@ static void intel_pmu_disable_event(struct perf_event *event) cpuc->intel_ctrl_guest_mask &= ~(1ull << hwc->idx); cpuc->intel_ctrl_host_mask &= ~(1ull << hwc->idx); + cpuc->intel_cp_status &= ~(1ull << hwc->idx); + + /* + * must disable before any actual event + * because any event may be combined with LBR + */ + if (intel_pmu_needs_lbr_smpl(event)) + intel_pmu_lbr_disable(event); if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL)) { intel_pmu_disable_fixed(hwc); @@ -895,7 +1223,7 @@ static void intel_pmu_disable_event(struct perf_event *event) static void intel_pmu_enable_fixed(struct hw_perf_event *hwc) { - int idx = hwc->idx - X86_PMC_IDX_FIXED; + int idx = hwc->idx - INTEL_PMC_IDX_FIXED; u64 ctrl_val, bits, mask; /* @@ -929,19 +1257,28 @@ static void intel_pmu_enable_event(struct perf_event *event) struct hw_perf_event *hwc = &event->hw; struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - if (unlikely(hwc->idx == X86_PMC_IDX_FIXED_BTS)) { + if (unlikely(hwc->idx == INTEL_PMC_IDX_FIXED_BTS)) { if (!__this_cpu_read(cpu_hw_events.enabled)) return; intel_pmu_enable_bts(hwc->config); return; } + /* + * must enabled before any actual event + * because any event may be combined with LBR + */ + if (intel_pmu_needs_lbr_smpl(event)) + intel_pmu_lbr_enable(event); if (event->attr.exclude_host) cpuc->intel_ctrl_guest_mask |= (1ull << hwc->idx); if (event->attr.exclude_guest) cpuc->intel_ctrl_host_mask |= (1ull << hwc->idx); + if (unlikely(event_is_checkpointed(event))) + cpuc->intel_cp_status |= (1ull << hwc->idx); + if (unlikely(hwc->config_base == MSR_ARCH_PERFMON_FIXED_CTR_CTRL)) { intel_pmu_enable_fixed(hwc); return; @@ -960,6 +1297,17 @@ static void intel_pmu_enable_event(struct perf_event *event) int intel_pmu_save_and_restart(struct perf_event *event) { x86_perf_event_update(event); + /* + * For a checkpointed counter always reset back to 0. This + * avoids a situation where the counter overflows, aborts the + * transaction and is then set back to shortly before the + * overflow, and overflows and aborts again. + */ + if (unlikely(event_is_checkpointed(event))) { + /* No race with NMIs because the counter should not be armed */ + wrmsrl(event->hw.event_base, 0); + local64_set(&event->hw.prev_count, 0); + } return x86_perf_event_set_period(event); } @@ -974,14 +1322,14 @@ static void intel_pmu_reset(void) local_irq_save(flags); - printk("clearing PMU state on CPU#%d\n", smp_processor_id()); + pr_info("clearing PMU state on CPU#%d\n", smp_processor_id()); for (idx = 0; idx < x86_pmu.num_counters; idx++) { - checking_wrmsrl(x86_pmu_config_addr(idx), 0ull); - checking_wrmsrl(x86_pmu_event_addr(idx), 0ull); + wrmsrl_safe(x86_pmu_config_addr(idx), 0ull); + wrmsrl_safe(x86_pmu_event_addr(idx), 0ull); } for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) - checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull); + wrmsrl_safe(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull); if (ds) ds->bts_index = ds->bts_buffer_base; @@ -1001,34 +1349,30 @@ static int intel_pmu_handle_irq(struct pt_regs *regs) u64 status; int handled; - perf_sample_data_init(&data, 0); - cpuc = &__get_cpu_var(cpu_hw_events); /* - * Some chipsets need to unmask the LVTPC in a particular spot - * inside the nmi handler. As a result, the unmasking was pushed - * into all the nmi handlers. - * - * This handler doesn't seem to have any issues with the unmasking - * so it was left at the top. + * No known reason to not always do late ACK, + * but just in case do it opt-in. */ - apic_write(APIC_LVTPC, APIC_DM_NMI); - + if (!x86_pmu.late_ack) + apic_write(APIC_LVTPC, APIC_DM_NMI); intel_pmu_disable_all(); handled = intel_pmu_drain_bts_buffer(); status = intel_pmu_get_status(); - if (!status) { - intel_pmu_enable_all(0); - return handled; - } + if (!status) + goto done; loops = 0; again: intel_pmu_ack_status(status); if (++loops > 100) { - WARN_ONCE(1, "perfevents: irq loop stuck!\n"); - perf_event_print_debug(); + static bool warned = false; + if (!warned) { + WARN(1, "perfevents: irq loop stuck!\n"); + perf_event_print_debug(); + warned = true; + } intel_pmu_reset(); goto done; } @@ -1038,6 +1382,15 @@ again: intel_pmu_lbr_read(); /* + * CondChgd bit 63 doesn't mean any overflow status. Ignore + * and clear the bit. + */ + if (__test_and_clear_bit(63, (unsigned long *)&status)) { + if (!status) + goto done; + } + + /* * PEBS overflow sets bit 62 in the global status register */ if (__test_and_clear_bit(62, (unsigned long *)&status)) { @@ -1045,6 +1398,13 @@ again: x86_pmu.drain_pebs(regs); } + /* + * Checkpointed counters can lead to 'spurious' PMIs because the + * rollback caused by the PMI will have cleared the overflow status + * bit. Therefore always force probe these counters. + */ + status |= cpuc->intel_cp_status; + for_each_set_bit(bit, (unsigned long *)&status, X86_PMC_IDX_MAX) { struct perf_event *event = cpuc->events[bit]; @@ -1056,7 +1416,10 @@ again: if (!intel_pmu_save_and_restart(event)) continue; - data.period = event->hw.last_period; + perf_sample_data_init(&data, 0, event->hw.last_period); + + if (has_branch_stack(event)) + data.br_stack = &cpuc->lbr_stack; if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); @@ -1071,6 +1434,13 @@ again: done: intel_pmu_enable_all(0); + /* + * Only unmask the NMI after the overflow counters + * have been reset. This avoids spurious NMIs on + * Haswell CPUs. + */ + if (x86_pmu.late_ack) + apic_write(APIC_LVTPC, APIC_DM_NMI); return handled; } @@ -1092,27 +1462,33 @@ intel_bts_constraints(struct perf_event *event) return NULL; } -static bool intel_try_alt_er(struct perf_event *event, int orig_idx) +static int intel_alt_er(int idx) { if (!(x86_pmu.er_flags & ERF_HAS_RSP_1)) - return false; + return idx; - if (event->hw.extra_reg.idx == EXTRA_REG_RSP_0) { - event->hw.config &= ~INTEL_ARCH_EVENT_MASK; - event->hw.config |= 0x01bb; - event->hw.extra_reg.idx = EXTRA_REG_RSP_1; - event->hw.extra_reg.reg = MSR_OFFCORE_RSP_1; - } else if (event->hw.extra_reg.idx == EXTRA_REG_RSP_1) { + if (idx == EXTRA_REG_RSP_0) + return EXTRA_REG_RSP_1; + + if (idx == EXTRA_REG_RSP_1) + return EXTRA_REG_RSP_0; + + return idx; +} + +static void intel_fixup_er(struct perf_event *event, int idx) +{ + event->hw.extra_reg.idx = idx; + + if (idx == EXTRA_REG_RSP_0) { event->hw.config &= ~INTEL_ARCH_EVENT_MASK; - event->hw.config |= 0x01b7; - event->hw.extra_reg.idx = EXTRA_REG_RSP_0; + event->hw.config |= x86_pmu.extra_regs[EXTRA_REG_RSP_0].event; event->hw.extra_reg.reg = MSR_OFFCORE_RSP_0; + } else if (idx == EXTRA_REG_RSP_1) { + event->hw.config &= ~INTEL_ARCH_EVENT_MASK; + event->hw.config |= x86_pmu.extra_regs[EXTRA_REG_RSP_1].event; + event->hw.extra_reg.reg = MSR_OFFCORE_RSP_1; } - - if (event->hw.extra_reg.idx == orig_idx) - return false; - - return true; } /* @@ -1124,20 +1500,24 @@ static bool intel_try_alt_er(struct perf_event *event, int orig_idx) */ static struct event_constraint * __intel_shared_reg_get_constraints(struct cpu_hw_events *cpuc, - struct perf_event *event) + struct perf_event *event, + struct hw_perf_event_extra *reg) { struct event_constraint *c = &emptyconstraint; - struct hw_perf_event_extra *reg = &event->hw.extra_reg; struct er_account *era; unsigned long flags; - int orig_idx = reg->idx; + int idx = reg->idx; - /* already allocated shared msr */ - if (reg->alloc) - return &unconstrained; + /* + * reg->alloc can be set due to existing state, so for fake cpuc we + * need to ignore this, otherwise we might fail to allocate proper fake + * state for this extra reg constraint. Also see the comment below. + */ + if (reg->alloc && !cpuc->is_fake) + return NULL; /* call x86_get_event_constraint() */ again: - era = &cpuc->shared_regs->regs[reg->idx]; + era = &cpuc->shared_regs->regs[idx]; /* * we use spin_lock_irqsave() to avoid lockdep issues when * passing a fake cpuc @@ -1146,6 +1526,29 @@ again: if (!atomic_read(&era->ref) || era->config == reg->config) { + /* + * If its a fake cpuc -- as per validate_{group,event}() we + * shouldn't touch event state and we can avoid doing so + * since both will only call get_event_constraints() once + * on each event, this avoids the need for reg->alloc. + * + * Not doing the ER fixup will only result in era->reg being + * wrong, but since we won't actually try and program hardware + * this isn't a problem either. + */ + if (!cpuc->is_fake) { + if (idx != reg->idx) + intel_fixup_er(event, idx); + + /* + * x86_schedule_events() can call get_event_constraints() + * multiple times on events in the case of incremental + * scheduling(). reg->alloc ensures we only do the ER + * allocation once. + */ + reg->alloc = 1; + } + /* lock in msr value */ era->config = reg->config; era->reg = reg->reg; @@ -1153,21 +1556,17 @@ again: /* one more user */ atomic_inc(&era->ref); - /* no need to reallocate during incremental event scheduling */ - reg->alloc = 1; - /* - * All events using extra_reg are unconstrained. - * Avoids calling x86_get_event_constraints() - * - * Must revisit if extra_reg controlling events - * ever have constraints. Worst case we go through - * the regular event constraint table. + * need to call x86_get_event_constraint() + * to check if associated event has constraints */ - c = &unconstrained; - } else if (intel_try_alt_er(event, orig_idx)) { - raw_spin_unlock_irqrestore(&era->lock, flags); - goto again; + c = NULL; + } else { + idx = intel_alt_er(idx); + if (idx != reg->idx) { + raw_spin_unlock_irqrestore(&era->lock, flags); + goto again; + } } raw_spin_unlock_irqrestore(&era->lock, flags); @@ -1181,11 +1580,14 @@ __intel_shared_reg_put_constraints(struct cpu_hw_events *cpuc, struct er_account *era; /* - * only put constraint if extra reg was actually - * allocated. Also takes care of event which do - * not use an extra shared reg + * Only put constraint if extra reg was actually allocated. Also takes + * care of event which do not use an extra shared reg. + * + * Also, if this is a fake cpuc we shouldn't touch any event state + * (reg->alloc) and we don't care about leaving inconsistent cpuc state + * either since it'll be thrown out. */ - if (!reg->alloc) + if (!reg->alloc || cpuc->is_fake) return; era = &cpuc->shared_regs->regs[reg->idx]; @@ -1201,11 +1603,23 @@ static struct event_constraint * intel_shared_regs_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) { - struct event_constraint *c = NULL; - - if (event->hw.extra_reg.idx != EXTRA_REG_NONE) - c = __intel_shared_reg_get_constraints(cpuc, event); - + struct event_constraint *c = NULL, *d; + struct hw_perf_event_extra *xreg, *breg; + + xreg = &event->hw.extra_reg; + if (xreg->idx != EXTRA_REG_NONE) { + c = __intel_shared_reg_get_constraints(cpuc, event, xreg); + if (c == &emptyconstraint) + return c; + } + breg = &event->hw.branch_reg; + if (breg->idx != EXTRA_REG_NONE) { + d = __intel_shared_reg_get_constraints(cpuc, event, breg); + if (d == &emptyconstraint) { + __intel_shared_reg_put_constraints(cpuc, xreg); + c = d; + } + } return c; } @@ -1216,8 +1630,10 @@ x86_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) if (x86_pmu.event_constraints) { for_each_event_constraint(c, x86_pmu.event_constraints) { - if ((event->hw.config & c->cmask) == c->code) + if ((event->hw.config & c->cmask) == c->code) { + event->hw.flags |= c->flags; return c; + } } } @@ -1253,6 +1669,10 @@ intel_put_shared_regs_event_constraints(struct cpu_hw_events *cpuc, reg = &event->hw.extra_reg; if (reg->idx != EXTRA_REG_NONE) __intel_shared_reg_put_constraints(cpuc, reg); + + reg = &event->hw.branch_reg; + if (reg->idx != EXTRA_REG_NONE) + __intel_shared_reg_put_constraints(cpuc, reg); } static void intel_put_event_constraints(struct cpu_hw_events *cpuc, @@ -1261,15 +1681,9 @@ static void intel_put_event_constraints(struct cpu_hw_events *cpuc, intel_put_shared_regs_event_constraints(cpuc, event); } -static int intel_pmu_hw_config(struct perf_event *event) +static void intel_pebs_aliases_core2(struct perf_event *event) { - int ret = x86_pmu_hw_config(event); - - if (ret) - return ret; - - if (event->attr.precise_ip && - (event->hw.config & X86_RAW_EVENT_MASK) == 0x003c) { + if ((event->hw.config & X86_RAW_EVENT_MASK) == 0x003c) { /* * Use an alternative encoding for CPU_CLK_UNHALTED.THREAD_P * (0x003c) so that we can use it with PEBS. @@ -1288,11 +1702,56 @@ static int intel_pmu_hw_config(struct perf_event *event) * * Thereby we gain a PEBS capable cycle counter. */ - u64 alt_config = 0x108000c0; /* INST_RETIRED.TOTAL_CYCLES */ + u64 alt_config = X86_CONFIG(.event=0xc0, .inv=1, .cmask=16); + + alt_config |= (event->hw.config & ~X86_RAW_EVENT_MASK); + event->hw.config = alt_config; + } +} + +static void intel_pebs_aliases_snb(struct perf_event *event) +{ + if ((event->hw.config & X86_RAW_EVENT_MASK) == 0x003c) { + /* + * Use an alternative encoding for CPU_CLK_UNHALTED.THREAD_P + * (0x003c) so that we can use it with PEBS. + * + * The regular CPU_CLK_UNHALTED.THREAD_P event (0x003c) isn't + * PEBS capable. However we can use UOPS_RETIRED.ALL + * (0x01c2), which is a PEBS capable event, to get the same + * count. + * + * UOPS_RETIRED.ALL counts the number of cycles that retires + * CNTMASK micro-ops. By setting CNTMASK to a value (16) + * larger than the maximum number of micro-ops that can be + * retired per cycle (4) and then inverting the condition, we + * count all cycles that retire 16 or less micro-ops, which + * is every cycle. + * + * Thereby we gain a PEBS capable cycle counter. + */ + u64 alt_config = X86_CONFIG(.event=0xc2, .umask=0x01, .inv=1, .cmask=16); alt_config |= (event->hw.config & ~X86_RAW_EVENT_MASK); event->hw.config = alt_config; } +} + +static int intel_pmu_hw_config(struct perf_event *event) +{ + int ret = x86_pmu_hw_config(event); + + if (ret) + return ret; + + if (event->attr.precise_ip && x86_pmu.pebs_aliases) + x86_pmu.pebs_aliases(event); + + if (intel_pmu_needs_lbr_smpl(event)) { + ret = intel_pmu_setup_lbr_filter(event); + if (ret) + return ret; + } if (event->attr.type != PERF_TYPE_RAW) return 0; @@ -1328,8 +1787,16 @@ static struct perf_guest_switch_msr *intel_guest_get_msrs(int *nr) arr[0].msr = MSR_CORE_PERF_GLOBAL_CTRL; arr[0].host = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_guest_mask; arr[0].guest = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_host_mask; + /* + * If PMU counter has PEBS enabled it is not enough to disable counter + * on a guest entry since PEBS memory write can overshoot guest entry + * and corrupt guest memory. Disabling PEBS solves the problem. + */ + arr[1].msr = MSR_IA32_PEBS_ENABLE; + arr[1].host = cpuc->pebs_enabled; + arr[1].guest = 0; - *nr = 1; + *nr = 2; return arr; } @@ -1383,6 +1850,88 @@ static void core_pmu_enable_all(int added) } } +static int hsw_hw_config(struct perf_event *event) +{ + int ret = intel_pmu_hw_config(event); + + if (ret) + return ret; + if (!boot_cpu_has(X86_FEATURE_RTM) && !boot_cpu_has(X86_FEATURE_HLE)) + return 0; + event->hw.config |= event->attr.config & (HSW_IN_TX|HSW_IN_TX_CHECKPOINTED); + + /* + * IN_TX/IN_TX-CP filters are not supported by the Haswell PMU with + * PEBS or in ANY thread mode. Since the results are non-sensical forbid + * this combination. + */ + if ((event->hw.config & (HSW_IN_TX|HSW_IN_TX_CHECKPOINTED)) && + ((event->hw.config & ARCH_PERFMON_EVENTSEL_ANY) || + event->attr.precise_ip > 0)) + return -EOPNOTSUPP; + + if (event_is_checkpointed(event)) { + /* + * Sampling of checkpointed events can cause situations where + * the CPU constantly aborts because of a overflow, which is + * then checkpointed back and ignored. Forbid checkpointing + * for sampling. + * + * But still allow a long sampling period, so that perf stat + * from KVM works. + */ + if (event->attr.sample_period > 0 && + event->attr.sample_period < 0x7fffffff) + return -EOPNOTSUPP; + } + return 0; +} + +static struct event_constraint counter2_constraint = + EVENT_CONSTRAINT(0, 0x4, 0); + +static struct event_constraint * +hsw_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) +{ + struct event_constraint *c = intel_get_event_constraints(cpuc, event); + + /* Handle special quirk on in_tx_checkpointed only in counter 2 */ + if (event->hw.config & HSW_IN_TX_CHECKPOINTED) { + if (c->idxmsk64 & (1U << 2)) + return &counter2_constraint; + return &emptyconstraint; + } + + return c; +} + +PMU_FORMAT_ATTR(event, "config:0-7" ); +PMU_FORMAT_ATTR(umask, "config:8-15" ); +PMU_FORMAT_ATTR(edge, "config:18" ); +PMU_FORMAT_ATTR(pc, "config:19" ); +PMU_FORMAT_ATTR(any, "config:21" ); /* v3 + */ +PMU_FORMAT_ATTR(inv, "config:23" ); +PMU_FORMAT_ATTR(cmask, "config:24-31" ); +PMU_FORMAT_ATTR(in_tx, "config:32"); +PMU_FORMAT_ATTR(in_tx_cp, "config:33"); + +static struct attribute *intel_arch_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_pc.attr, + &format_attr_inv.attr, + &format_attr_cmask.attr, + NULL, +}; + +ssize_t intel_event_sysfs_show(char *page, u64 config) +{ + u64 event = (config & ARCH_PERFMON_EVENTSEL_EVENT); + + return x86_event_sysfs_show(page, config, event); +} + static __initconst const struct x86_pmu core_pmu = { .name = "core", .handle_irq = x86_pmu_handle_irq, @@ -1407,6 +1956,8 @@ static __initconst const struct x86_pmu core_pmu = { .put_event_constraints = intel_put_event_constraints, .event_constraints = intel_core_event_constraints, .guest_get_msrs = core_guest_get_msrs, + .format_attrs = intel_arch_formats_attr, + .events_sysfs_show = intel_event_sysfs_show, }; struct intel_shared_regs *allocate_shared_regs(int cpu) @@ -1432,7 +1983,7 @@ static int intel_pmu_cpu_prepare(int cpu) { struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu); - if (!x86_pmu.extra_regs) + if (!(x86_pmu.extra_regs || x86_pmu.lbr_sel_map)) return NOTIFY_OK; cpuc->shared_regs = allocate_shared_regs(cpu); @@ -1454,22 +2005,28 @@ static void intel_pmu_cpu_starting(int cpu) */ intel_pmu_lbr_reset(); - if (!cpuc->shared_regs || (x86_pmu.er_flags & ERF_NO_HT_SHARING)) + cpuc->lbr_sel = NULL; + + if (!cpuc->shared_regs) return; - for_each_cpu(i, topology_thread_cpumask(cpu)) { - struct intel_shared_regs *pc; + if (!(x86_pmu.er_flags & ERF_NO_HT_SHARING)) { + for_each_cpu(i, topology_thread_cpumask(cpu)) { + struct intel_shared_regs *pc; - pc = per_cpu(cpu_hw_events, i).shared_regs; - if (pc && pc->core_id == core_id) { - cpuc->kfree_on_online = cpuc->shared_regs; - cpuc->shared_regs = pc; - break; + pc = per_cpu(cpu_hw_events, i).shared_regs; + if (pc && pc->core_id == core_id) { + cpuc->kfree_on_online = cpuc->shared_regs; + cpuc->shared_regs = pc; + break; + } } + cpuc->shared_regs->core_id = core_id; + cpuc->shared_regs->refcnt++; } - cpuc->shared_regs->core_id = core_id; - cpuc->shared_regs->refcnt++; + if (x86_pmu.lbr_sel_map) + cpuc->lbr_sel = &cpuc->shared_regs->regs[EXTRA_REG_LBR]; } static void intel_pmu_cpu_dying(int cpu) @@ -1487,6 +2044,38 @@ static void intel_pmu_cpu_dying(int cpu) fini_debug_store_on_cpu(cpu); } +static void intel_pmu_flush_branch_stack(void) +{ + /* + * Intel LBR does not tag entries with the + * PID of the current task, then we need to + * flush it on ctxsw + * For now, we simply reset it + */ + if (x86_pmu.lbr_nr) + intel_pmu_lbr_reset(); +} + +PMU_FORMAT_ATTR(offcore_rsp, "config1:0-63"); + +PMU_FORMAT_ATTR(ldlat, "config1:0-15"); + +static struct attribute *intel_arch3_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_pc.attr, + &format_attr_any.attr, + &format_attr_inv.attr, + &format_attr_cmask.attr, + &format_attr_in_tx.attr, + &format_attr_in_tx_cp.attr, + + &format_attr_offcore_rsp.attr, /* XXX do NHM/WSM + SNB breakout */ + &format_attr_ldlat.attr, /* PEBS load latency */ + NULL, +}; + static __initconst const struct x86_pmu intel_pmu = { .name = "Intel", .handle_irq = intel_pmu_handle_irq, @@ -1509,11 +2098,16 @@ static __initconst const struct x86_pmu intel_pmu = { .max_period = (1ULL << 31) - 1, .get_event_constraints = intel_get_event_constraints, .put_event_constraints = intel_put_event_constraints, + .pebs_aliases = intel_pebs_aliases_core2, + + .format_attrs = intel_arch3_formats_attr, + .events_sysfs_show = intel_event_sysfs_show, .cpu_prepare = intel_pmu_cpu_prepare, .cpu_starting = intel_pmu_cpu_starting, .cpu_dying = intel_pmu_cpu_dying, .guest_get_msrs = intel_guest_get_msrs, + .flush_branch_stack = intel_pmu_flush_branch_stack, }; static __init void intel_clovertown_quirk(void) @@ -1537,16 +2131,96 @@ static __init void intel_clovertown_quirk(void) * But taken together it might just make sense to not enable PEBS on * these chips. */ - printk(KERN_WARNING "PEBS disabled due to CPU errata.\n"); + pr_warn("PEBS disabled due to CPU errata\n"); x86_pmu.pebs = 0; x86_pmu.pebs_constraints = NULL; } +static int intel_snb_pebs_broken(int cpu) +{ + u32 rev = UINT_MAX; /* default to broken for unknown models */ + + switch (cpu_data(cpu).x86_model) { + case 42: /* SNB */ + rev = 0x28; + break; + + case 45: /* SNB-EP */ + switch (cpu_data(cpu).x86_mask) { + case 6: rev = 0x618; break; + case 7: rev = 0x70c; break; + } + } + + return (cpu_data(cpu).microcode < rev); +} + +static void intel_snb_check_microcode(void) +{ + int pebs_broken = 0; + int cpu; + + get_online_cpus(); + for_each_online_cpu(cpu) { + if ((pebs_broken = intel_snb_pebs_broken(cpu))) + break; + } + put_online_cpus(); + + if (pebs_broken == x86_pmu.pebs_broken) + return; + + /* + * Serialized by the microcode lock.. + */ + if (x86_pmu.pebs_broken) { + pr_info("PEBS enabled due to microcode update\n"); + x86_pmu.pebs_broken = 0; + } else { + pr_info("PEBS disabled due to CPU errata, please upgrade microcode\n"); + x86_pmu.pebs_broken = 1; + } +} + +/* + * Under certain circumstances, access certain MSR may cause #GP. + * The function tests if the input MSR can be safely accessed. + */ +static bool check_msr(unsigned long msr, u64 mask) +{ + u64 val_old, val_new, val_tmp; + + /* + * Read the current value, change it and read it back to see if it + * matches, this is needed to detect certain hardware emulators + * (qemu/kvm) that don't trap on the MSR access and always return 0s. + */ + if (rdmsrl_safe(msr, &val_old)) + return false; + + /* + * Only change the bits which can be updated by wrmsrl. + */ + val_tmp = val_old ^ mask; + if (wrmsrl_safe(msr, val_tmp) || + rdmsrl_safe(msr, &val_new)) + return false; + + if (val_new != val_tmp) + return false; + + /* Here it's sure that the MSR can be safely accessed. + * Restore the old value and return. + */ + wrmsrl(msr, val_old); + + return true; +} + static __init void intel_sandybridge_quirk(void) { - printk(KERN_WARNING "PEBS disabled due to CPU errata.\n"); - x86_pmu.pebs = 0; - x86_pmu.pebs_constraints = NULL; + x86_pmu.check_microcode = intel_snb_check_microcode; + intel_snb_check_microcode(); } static const struct { int id; char *name; } intel_arch_events_map[] __initconst = { @@ -1566,8 +2240,8 @@ static __init void intel_arch_events_quirk(void) /* disable event that reported as not presend by cpuid */ for_each_set_bit(bit, x86_pmu.events_mask, ARRAY_SIZE(intel_arch_events_map)) { intel_perfmon_event_map[intel_arch_events_map[bit].id] = 0; - printk(KERN_WARNING "CPUID marked event: \'%s\' unavailable\n", - intel_arch_events_map[bit].name); + pr_warn("CPUID marked event: \'%s\' unavailable\n", + intel_arch_events_map[bit].name); } } @@ -1586,22 +2260,61 @@ static __init void intel_nehalem_quirk(void) intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89; ebx.split.no_branch_misses_retired = 0; x86_pmu.events_maskl = ebx.full; - printk(KERN_INFO "CPU erratum AAJ80 worked around\n"); + pr_info("CPU erratum AAJ80 worked around\n"); } } +EVENT_ATTR_STR(mem-loads, mem_ld_hsw, "event=0xcd,umask=0x1,ldlat=3"); +EVENT_ATTR_STR(mem-stores, mem_st_hsw, "event=0xd0,umask=0x82") + +/* Haswell special events */ +EVENT_ATTR_STR(tx-start, tx_start, "event=0xc9,umask=0x1"); +EVENT_ATTR_STR(tx-commit, tx_commit, "event=0xc9,umask=0x2"); +EVENT_ATTR_STR(tx-abort, tx_abort, "event=0xc9,umask=0x4"); +EVENT_ATTR_STR(tx-capacity, tx_capacity, "event=0x54,umask=0x2"); +EVENT_ATTR_STR(tx-conflict, tx_conflict, "event=0x54,umask=0x1"); +EVENT_ATTR_STR(el-start, el_start, "event=0xc8,umask=0x1"); +EVENT_ATTR_STR(el-commit, el_commit, "event=0xc8,umask=0x2"); +EVENT_ATTR_STR(el-abort, el_abort, "event=0xc8,umask=0x4"); +EVENT_ATTR_STR(el-capacity, el_capacity, "event=0x54,umask=0x2"); +EVENT_ATTR_STR(el-conflict, el_conflict, "event=0x54,umask=0x1"); +EVENT_ATTR_STR(cycles-t, cycles_t, "event=0x3c,in_tx=1"); +EVENT_ATTR_STR(cycles-ct, cycles_ct, "event=0x3c,in_tx=1,in_tx_cp=1"); + +static struct attribute *hsw_events_attrs[] = { + EVENT_PTR(tx_start), + EVENT_PTR(tx_commit), + EVENT_PTR(tx_abort), + EVENT_PTR(tx_capacity), + EVENT_PTR(tx_conflict), + EVENT_PTR(el_start), + EVENT_PTR(el_commit), + EVENT_PTR(el_abort), + EVENT_PTR(el_capacity), + EVENT_PTR(el_conflict), + EVENT_PTR(cycles_t), + EVENT_PTR(cycles_ct), + EVENT_PTR(mem_ld_hsw), + EVENT_PTR(mem_st_hsw), + NULL +}; + __init int intel_pmu_init(void) { union cpuid10_edx edx; union cpuid10_eax eax; union cpuid10_ebx ebx; + struct event_constraint *c; unsigned int unused; - int version; + struct extra_reg *er; + int version, i; if (!cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) { switch (boot_cpu_data.x86) { case 0x6: return p6_pmu_init(); + case 0xb: + return knc_pmu_init(); case 0xf: return p4_pmu_init(); } @@ -1630,6 +2343,8 @@ __init int intel_pmu_init(void) x86_pmu.events_maskl = ebx.full; x86_pmu.events_mask_len = eax.split.mask_length; + x86_pmu.max_pebs_events = min_t(unsigned, MAX_PEBS_EVENTS, x86_pmu.num_counters); + /* * Quirk: v2 perfmon does not report fixed-purpose events, so * assume at least 3 events: @@ -1637,10 +2352,7 @@ __init int intel_pmu_init(void) if (version > 1) x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3); - /* - * v2 and above have a perf capabilities MSR - */ - if (version > 1) { + if (boot_cpu_has(X86_FEATURE_PDCM)) { u64 capabilities; rdmsrl(MSR_IA32_PERF_CAPABILITIES, capabilities); @@ -1689,10 +2401,14 @@ __init int intel_pmu_init(void) x86_pmu.enable_all = intel_pmu_nhm_enable_all; x86_pmu.extra_regs = intel_nehalem_extra_regs; + x86_pmu.cpu_events = nhm_events_attrs; + /* UOPS_ISSUED.STALLED_CYCLES */ - intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = 0x180010e; + intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = + X86_CONFIG(.event=0x0e, .umask=0x01, .inv=1, .cmask=1); /* UOPS_EXECUTED.CORE_ACTIVE_CYCLES,c=1,i=1 */ - intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = 0x1803fb1; + intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = + X86_CONFIG(.event=0xb1, .umask=0x3f, .inv=1, .cmask=1); x86_add_quirk(intel_nehalem_quirk); @@ -1700,6 +2416,10 @@ __init int intel_pmu_init(void) break; case 28: /* Atom */ + case 38: /* Lincroft */ + case 39: /* Penwell */ + case 53: /* Cloverview */ + case 54: /* Cedarview */ memcpy(hw_cache_event_ids, atom_hw_cache_event_ids, sizeof(hw_cache_event_ids)); @@ -1710,6 +2430,22 @@ __init int intel_pmu_init(void) pr_cont("Atom events, "); break; + case 55: /* Atom 22nm "Silvermont" */ + case 77: /* Avoton "Silvermont" */ + memcpy(hw_cache_event_ids, slm_hw_cache_event_ids, + sizeof(hw_cache_event_ids)); + memcpy(hw_cache_extra_regs, slm_hw_cache_extra_regs, + sizeof(hw_cache_extra_regs)); + + intel_pmu_lbr_init_atom(); + + x86_pmu.event_constraints = intel_slm_event_constraints; + x86_pmu.pebs_constraints = intel_slm_pebs_event_constraints; + x86_pmu.extra_regs = intel_slm_extra_regs; + x86_pmu.er_flags |= ERF_HAS_RSP_1; + pr_cont("Silvermont events, "); + break; + case 37: /* 32 nm nehalem, "Clarkdale" */ case 44: /* 32 nm nehalem, "Gulftown" */ case 47: /* 32 nm Xeon E7 */ @@ -1726,36 +2462,108 @@ __init int intel_pmu_init(void) x86_pmu.extra_regs = intel_westmere_extra_regs; x86_pmu.er_flags |= ERF_HAS_RSP_1; + x86_pmu.cpu_events = nhm_events_attrs; + /* UOPS_ISSUED.STALLED_CYCLES */ - intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = 0x180010e; + intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = + X86_CONFIG(.event=0x0e, .umask=0x01, .inv=1, .cmask=1); /* UOPS_EXECUTED.CORE_ACTIVE_CYCLES,c=1,i=1 */ - intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = 0x1803fb1; + intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = + X86_CONFIG(.event=0xb1, .umask=0x3f, .inv=1, .cmask=1); pr_cont("Westmere events, "); break; case 42: /* SandyBridge */ - x86_add_quirk(intel_sandybridge_quirk); case 45: /* SandyBridge, "Romely-EP" */ + x86_add_quirk(intel_sandybridge_quirk); memcpy(hw_cache_event_ids, snb_hw_cache_event_ids, sizeof(hw_cache_event_ids)); + memcpy(hw_cache_extra_regs, snb_hw_cache_extra_regs, + sizeof(hw_cache_extra_regs)); - intel_pmu_lbr_init_nhm(); + intel_pmu_lbr_init_snb(); x86_pmu.event_constraints = intel_snb_event_constraints; x86_pmu.pebs_constraints = intel_snb_pebs_event_constraints; - x86_pmu.extra_regs = intel_snb_extra_regs; + x86_pmu.pebs_aliases = intel_pebs_aliases_snb; + if (boot_cpu_data.x86_model == 45) + x86_pmu.extra_regs = intel_snbep_extra_regs; + else + x86_pmu.extra_regs = intel_snb_extra_regs; /* all extra regs are per-cpu when HT is on */ x86_pmu.er_flags |= ERF_HAS_RSP_1; x86_pmu.er_flags |= ERF_NO_HT_SHARING; + x86_pmu.cpu_events = snb_events_attrs; + /* UOPS_ISSUED.ANY,c=1,i=1 to count stall cycles */ - intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = 0x180010e; + intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = + X86_CONFIG(.event=0x0e, .umask=0x01, .inv=1, .cmask=1); /* UOPS_DISPATCHED.THREAD,c=1,i=1 to count stall cycles*/ - intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = 0x18001b1; + intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = + X86_CONFIG(.event=0xb1, .umask=0x01, .inv=1, .cmask=1); pr_cont("SandyBridge events, "); break; + case 58: /* IvyBridge */ + case 62: /* IvyBridge EP */ + memcpy(hw_cache_event_ids, snb_hw_cache_event_ids, + sizeof(hw_cache_event_ids)); + /* dTLB-load-misses on IVB is different than SNB */ + hw_cache_event_ids[C(DTLB)][C(OP_READ)][C(RESULT_MISS)] = 0x8108; /* DTLB_LOAD_MISSES.DEMAND_LD_MISS_CAUSES_A_WALK */ + + memcpy(hw_cache_extra_regs, snb_hw_cache_extra_regs, + sizeof(hw_cache_extra_regs)); + + intel_pmu_lbr_init_snb(); + + x86_pmu.event_constraints = intel_ivb_event_constraints; + x86_pmu.pebs_constraints = intel_ivb_pebs_event_constraints; + x86_pmu.pebs_aliases = intel_pebs_aliases_snb; + if (boot_cpu_data.x86_model == 62) + x86_pmu.extra_regs = intel_snbep_extra_regs; + else + x86_pmu.extra_regs = intel_snb_extra_regs; + /* all extra regs are per-cpu when HT is on */ + x86_pmu.er_flags |= ERF_HAS_RSP_1; + x86_pmu.er_flags |= ERF_NO_HT_SHARING; + + x86_pmu.cpu_events = snb_events_attrs; + + /* UOPS_ISSUED.ANY,c=1,i=1 to count stall cycles */ + intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = + X86_CONFIG(.event=0x0e, .umask=0x01, .inv=1, .cmask=1); + + pr_cont("IvyBridge events, "); + break; + + + case 60: /* Haswell Client */ + case 70: + case 71: + case 63: + case 69: + x86_pmu.late_ack = true; + memcpy(hw_cache_event_ids, snb_hw_cache_event_ids, sizeof(hw_cache_event_ids)); + memcpy(hw_cache_extra_regs, snb_hw_cache_extra_regs, sizeof(hw_cache_extra_regs)); + + intel_pmu_lbr_init_snb(); + + x86_pmu.event_constraints = intel_hsw_event_constraints; + x86_pmu.pebs_constraints = intel_hsw_pebs_event_constraints; + x86_pmu.extra_regs = intel_snb_extra_regs; + x86_pmu.pebs_aliases = intel_pebs_aliases_snb; + /* all extra regs are per-cpu when HT is on */ + x86_pmu.er_flags |= ERF_HAS_RSP_1; + x86_pmu.er_flags |= ERF_NO_HT_SHARING; + + x86_pmu.hw_config = hsw_hw_config; + x86_pmu.get_event_constraints = hsw_get_event_constraints; + x86_pmu.cpu_events = hsw_events_attrs; + x86_pmu.lbr_double_abort = true; + pr_cont("Haswell events, "); + break; default: switch (x86_pmu.version) { @@ -1773,5 +2581,72 @@ __init int intel_pmu_init(void) } } + if (x86_pmu.num_counters > INTEL_PMC_MAX_GENERIC) { + WARN(1, KERN_ERR "hw perf events %d > max(%d), clipping!", + x86_pmu.num_counters, INTEL_PMC_MAX_GENERIC); + x86_pmu.num_counters = INTEL_PMC_MAX_GENERIC; + } + x86_pmu.intel_ctrl = (1 << x86_pmu.num_counters) - 1; + + if (x86_pmu.num_counters_fixed > INTEL_PMC_MAX_FIXED) { + WARN(1, KERN_ERR "hw perf events fixed %d > max(%d), clipping!", + x86_pmu.num_counters_fixed, INTEL_PMC_MAX_FIXED); + x86_pmu.num_counters_fixed = INTEL_PMC_MAX_FIXED; + } + + x86_pmu.intel_ctrl |= + ((1LL << x86_pmu.num_counters_fixed)-1) << INTEL_PMC_IDX_FIXED; + + if (x86_pmu.event_constraints) { + /* + * event on fixed counter2 (REF_CYCLES) only works on this + * counter, so do not extend mask to generic counters + */ + for_each_event_constraint(c, x86_pmu.event_constraints) { + if (c->cmask != FIXED_EVENT_FLAGS + || c->idxmsk64 == INTEL_PMC_MSK_FIXED_REF_CYCLES) { + continue; + } + + c->idxmsk64 |= (1ULL << x86_pmu.num_counters) - 1; + c->weight += x86_pmu.num_counters; + } + } + + /* + * Access LBR MSR may cause #GP under certain circumstances. + * E.g. KVM doesn't support LBR MSR + * Check all LBT MSR here. + * Disable LBR access if any LBR MSRs can not be accessed. + */ + if (x86_pmu.lbr_nr && !check_msr(x86_pmu.lbr_tos, 0x3UL)) + x86_pmu.lbr_nr = 0; + for (i = 0; i < x86_pmu.lbr_nr; i++) { + if (!(check_msr(x86_pmu.lbr_from + i, 0xffffUL) && + check_msr(x86_pmu.lbr_to + i, 0xffffUL))) + x86_pmu.lbr_nr = 0; + } + + /* + * Access extra MSR may cause #GP under certain circumstances. + * E.g. KVM doesn't support offcore event + * Check all extra_regs here. + */ + if (x86_pmu.extra_regs) { + for (er = x86_pmu.extra_regs; er->msr; er++) { + er->extra_msr_access = check_msr(er->msr, 0x1ffUL); + /* Disable LBR select mapping */ + if ((er->idx == EXTRA_REG_LBR) && !er->extra_msr_access) + x86_pmu.lbr_sel_map = NULL; + } + } + + /* Support full width counters using alternative MSR range */ + if (x86_pmu.intel_cap.full_width_write) { + x86_pmu.max_period = x86_pmu.cntval_mask; + x86_pmu.perfctr = MSR_IA32_PMC0; + pr_cont("full-width counters, "); + } + return 0; } diff --git a/arch/x86/kernel/cpu/perf_event_intel_ds.c b/arch/x86/kernel/cpu/perf_event_intel_ds.c index d6bd49faa40..696ade311de 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_ds.c +++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c @@ -3,6 +3,7 @@ #include <linux/slab.h> #include <asm/perf_event.h> +#include <asm/insn.h> #include "perf_event.h" @@ -11,6 +12,7 @@ #define BTS_BUFFER_SIZE (PAGE_SIZE << 4) #define PEBS_BUFFER_SIZE PAGE_SIZE +#define PEBS_FIXUP_SIZE PAGE_SIZE /* * pebs_record_32 for p4 and core not supported @@ -23,6 +25,159 @@ struct pebs_record_32 { */ +union intel_x86_pebs_dse { + u64 val; + struct { + unsigned int ld_dse:4; + unsigned int ld_stlb_miss:1; + unsigned int ld_locked:1; + unsigned int ld_reserved:26; + }; + struct { + unsigned int st_l1d_hit:1; + unsigned int st_reserved1:3; + unsigned int st_stlb_miss:1; + unsigned int st_locked:1; + unsigned int st_reserved2:26; + }; +}; + + +/* + * Map PEBS Load Latency Data Source encodings to generic + * memory data source information + */ +#define P(a, b) PERF_MEM_S(a, b) +#define OP_LH (P(OP, LOAD) | P(LVL, HIT)) +#define SNOOP_NONE_MISS (P(SNOOP, NONE) | P(SNOOP, MISS)) + +static const u64 pebs_data_source[] = { + P(OP, LOAD) | P(LVL, MISS) | P(LVL, L3) | P(SNOOP, NA),/* 0x00:ukn L3 */ + OP_LH | P(LVL, L1) | P(SNOOP, NONE), /* 0x01: L1 local */ + OP_LH | P(LVL, LFB) | P(SNOOP, NONE), /* 0x02: LFB hit */ + OP_LH | P(LVL, L2) | P(SNOOP, NONE), /* 0x03: L2 hit */ + OP_LH | P(LVL, L3) | P(SNOOP, NONE), /* 0x04: L3 hit */ + OP_LH | P(LVL, L3) | P(SNOOP, MISS), /* 0x05: L3 hit, snoop miss */ + OP_LH | P(LVL, L3) | P(SNOOP, HIT), /* 0x06: L3 hit, snoop hit */ + OP_LH | P(LVL, L3) | P(SNOOP, HITM), /* 0x07: L3 hit, snoop hitm */ + OP_LH | P(LVL, REM_CCE1) | P(SNOOP, HIT), /* 0x08: L3 miss snoop hit */ + OP_LH | P(LVL, REM_CCE1) | P(SNOOP, HITM), /* 0x09: L3 miss snoop hitm*/ + OP_LH | P(LVL, LOC_RAM) | P(SNOOP, HIT), /* 0x0a: L3 miss, shared */ + OP_LH | P(LVL, REM_RAM1) | P(SNOOP, HIT), /* 0x0b: L3 miss, shared */ + OP_LH | P(LVL, LOC_RAM) | SNOOP_NONE_MISS,/* 0x0c: L3 miss, excl */ + OP_LH | P(LVL, REM_RAM1) | SNOOP_NONE_MISS,/* 0x0d: L3 miss, excl */ + OP_LH | P(LVL, IO) | P(SNOOP, NONE), /* 0x0e: I/O */ + OP_LH | P(LVL, UNC) | P(SNOOP, NONE), /* 0x0f: uncached */ +}; + +static u64 precise_store_data(u64 status) +{ + union intel_x86_pebs_dse dse; + u64 val = P(OP, STORE) | P(SNOOP, NA) | P(LVL, L1) | P(TLB, L2); + + dse.val = status; + + /* + * bit 4: TLB access + * 1 = stored missed 2nd level TLB + * + * so it either hit the walker or the OS + * otherwise hit 2nd level TLB + */ + if (dse.st_stlb_miss) + val |= P(TLB, MISS); + else + val |= P(TLB, HIT); + + /* + * bit 0: hit L1 data cache + * if not set, then all we know is that + * it missed L1D + */ + if (dse.st_l1d_hit) + val |= P(LVL, HIT); + else + val |= P(LVL, MISS); + + /* + * bit 5: Locked prefix + */ + if (dse.st_locked) + val |= P(LOCK, LOCKED); + + return val; +} + +static u64 precise_store_data_hsw(struct perf_event *event, u64 status) +{ + union perf_mem_data_src dse; + u64 cfg = event->hw.config & INTEL_ARCH_EVENT_MASK; + + dse.val = 0; + dse.mem_op = PERF_MEM_OP_STORE; + dse.mem_lvl = PERF_MEM_LVL_NA; + + /* + * L1 info only valid for following events: + * + * MEM_UOPS_RETIRED.STLB_MISS_STORES + * MEM_UOPS_RETIRED.LOCK_STORES + * MEM_UOPS_RETIRED.SPLIT_STORES + * MEM_UOPS_RETIRED.ALL_STORES + */ + if (cfg != 0x12d0 && cfg != 0x22d0 && cfg != 0x42d0 && cfg != 0x82d0) + return dse.mem_lvl; + + if (status & 1) + dse.mem_lvl = PERF_MEM_LVL_L1 | PERF_MEM_LVL_HIT; + else + dse.mem_lvl = PERF_MEM_LVL_L1 | PERF_MEM_LVL_MISS; + + /* Nothing else supported. Sorry. */ + return dse.val; +} + +static u64 load_latency_data(u64 status) +{ + union intel_x86_pebs_dse dse; + u64 val; + int model = boot_cpu_data.x86_model; + int fam = boot_cpu_data.x86; + + dse.val = status; + + /* + * use the mapping table for bit 0-3 + */ + val = pebs_data_source[dse.ld_dse]; + + /* + * Nehalem models do not support TLB, Lock infos + */ + if (fam == 0x6 && (model == 26 || model == 30 + || model == 31 || model == 46)) { + val |= P(TLB, NA) | P(LOCK, NA); + return val; + } + /* + * bit 4: TLB access + * 0 = did not miss 2nd level TLB + * 1 = missed 2nd level TLB + */ + if (dse.ld_stlb_miss) + val |= P(TLB, MISS) | P(TLB, L2); + else + val |= P(TLB, HIT) | P(TLB, L1) | P(TLB, L2); + + /* + * bit 5: locked prefix + */ + if (dse.ld_locked) + val |= P(LOCK, LOCKED); + + return val; +} + struct pebs_record_core { u64 flags, ip; u64 ax, bx, cx, dx; @@ -40,6 +195,36 @@ struct pebs_record_nhm { u64 status, dla, dse, lat; }; +/* + * Same as pebs_record_nhm, with two additional fields. + */ +struct pebs_record_hsw { + u64 flags, ip; + u64 ax, bx, cx, dx; + u64 si, di, bp, sp; + u64 r8, r9, r10, r11; + u64 r12, r13, r14, r15; + u64 status, dla, dse, lat; + u64 real_ip, tsx_tuning; +}; + +union hsw_tsx_tuning { + struct { + u32 cycles_last_block : 32, + hle_abort : 1, + rtm_abort : 1, + instruction_abort : 1, + non_instruction_abort : 1, + retry : 1, + data_conflict : 1, + capacity_writes : 1, + capacity_reads : 1; + }; + u64 value; +}; + +#define PEBS_HSW_TSX_FLAGS 0xff00000000ULL + void init_debug_store_on_cpu(int cpu) { struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds; @@ -60,20 +245,35 @@ void fini_debug_store_on_cpu(int cpu) wrmsr_on_cpu(cpu, MSR_IA32_DS_AREA, 0, 0); } +static DEFINE_PER_CPU(void *, insn_buffer); + static int alloc_pebs_buffer(int cpu) { struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds; int node = cpu_to_node(cpu); int max, thresh = 1; /* always use a single PEBS record */ - void *buffer; + void *buffer, *ibuffer; if (!x86_pmu.pebs) return 0; - buffer = kmalloc_node(PEBS_BUFFER_SIZE, GFP_KERNEL | __GFP_ZERO, node); + buffer = kzalloc_node(PEBS_BUFFER_SIZE, GFP_KERNEL, node); if (unlikely(!buffer)) return -ENOMEM; + /* + * HSW+ already provides us the eventing ip; no need to allocate this + * buffer then. + */ + if (x86_pmu.intel_cap.pebs_format < 2) { + ibuffer = kzalloc_node(PEBS_FIXUP_SIZE, GFP_KERNEL, node); + if (!ibuffer) { + kfree(buffer); + return -ENOMEM; + } + per_cpu(insn_buffer, cpu) = ibuffer; + } + max = PEBS_BUFFER_SIZE / x86_pmu.pebs_record_size; ds->pebs_buffer_base = (u64)(unsigned long)buffer; @@ -94,6 +294,9 @@ static void release_pebs_buffer(int cpu) if (!ds || !x86_pmu.pebs) return; + kfree(per_cpu(insn_buffer, cpu)); + per_cpu(insn_buffer, cpu) = NULL; + kfree((void *)(unsigned long)ds->pebs_buffer_base); ds->pebs_buffer_base = 0; } @@ -108,9 +311,11 @@ static int alloc_bts_buffer(int cpu) if (!x86_pmu.bts) return 0; - buffer = kmalloc_node(BTS_BUFFER_SIZE, GFP_KERNEL | __GFP_ZERO, node); - if (unlikely(!buffer)) + buffer = kzalloc_node(BTS_BUFFER_SIZE, GFP_KERNEL | __GFP_NOWARN, node); + if (unlikely(!buffer)) { + WARN_ONCE(1, "%s: BTS buffer allocation failure\n", __func__); return -ENOMEM; + } max = BTS_BUFFER_SIZE / BTS_RECORD_SIZE; thresh = max / 16; @@ -141,7 +346,7 @@ static int alloc_ds_buffer(int cpu) int node = cpu_to_node(cpu); struct debug_store *ds; - ds = kmalloc_node(sizeof(*ds), GFP_KERNEL | __GFP_ZERO, node); + ds = kzalloc_node(sizeof(*ds), GFP_KERNEL, node); if (unlikely(!ds)) return -ENOMEM; @@ -247,7 +452,7 @@ void reserve_ds_buffers(void) */ struct event_constraint bts_constraint = - EVENT_CONSTRAINT(0, 1ULL << X86_PMC_IDX_FIXED_BTS, 0); + EVENT_CONSTRAINT(0, 1ULL << INTEL_PMC_IDX_FIXED_BTS, 0); void intel_pmu_enable_bts(u64 config) { @@ -294,7 +499,7 @@ int intel_pmu_drain_bts_buffer(void) u64 to; u64 flags; }; - struct perf_event *event = cpuc->events[X86_PMC_IDX_FIXED_BTS]; + struct perf_event *event = cpuc->events[INTEL_PMC_IDX_FIXED_BTS]; struct bts_record *at, *top; struct perf_output_handle handle; struct perf_event_header header; @@ -313,11 +518,11 @@ int intel_pmu_drain_bts_buffer(void) if (top <= at) return 0; + memset(®s, 0, sizeof(regs)); + ds->bts_index = ds->bts_buffer_base; - perf_sample_data_init(&data, 0); - data.period = event->hw.last_period; - regs.ip = 0; + perf_sample_data_init(&data, 0, event->hw.last_period); /* * Prepare a generic sample, i.e. fill in the invariant fields. @@ -363,8 +568,34 @@ struct event_constraint intel_atom_pebs_event_constraints[] = { EVENT_CONSTRAINT_END }; +struct event_constraint intel_slm_pebs_event_constraints[] = { + INTEL_UEVENT_CONSTRAINT(0x0103, 0x1), /* REHABQ.LD_BLOCK_ST_FORWARD_PS */ + INTEL_UEVENT_CONSTRAINT(0x0803, 0x1), /* REHABQ.LD_SPLITS_PS */ + INTEL_UEVENT_CONSTRAINT(0x0204, 0x1), /* MEM_UOPS_RETIRED.L2_HIT_LOADS_PS */ + INTEL_UEVENT_CONSTRAINT(0x0404, 0x1), /* MEM_UOPS_RETIRED.L2_MISS_LOADS_PS */ + INTEL_UEVENT_CONSTRAINT(0x0804, 0x1), /* MEM_UOPS_RETIRED.DTLB_MISS_LOADS_PS */ + INTEL_UEVENT_CONSTRAINT(0x2004, 0x1), /* MEM_UOPS_RETIRED.HITM_PS */ + INTEL_UEVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY_PS */ + INTEL_UEVENT_CONSTRAINT(0x00c4, 0x1), /* BR_INST_RETIRED.ALL_BRANCHES_PS */ + INTEL_UEVENT_CONSTRAINT(0x7ec4, 0x1), /* BR_INST_RETIRED.JCC_PS */ + INTEL_UEVENT_CONSTRAINT(0xbfc4, 0x1), /* BR_INST_RETIRED.FAR_BRANCH_PS */ + INTEL_UEVENT_CONSTRAINT(0xebc4, 0x1), /* BR_INST_RETIRED.NON_RETURN_IND_PS */ + INTEL_UEVENT_CONSTRAINT(0xf7c4, 0x1), /* BR_INST_RETIRED.RETURN_PS */ + INTEL_UEVENT_CONSTRAINT(0xf9c4, 0x1), /* BR_INST_RETIRED.CALL_PS */ + INTEL_UEVENT_CONSTRAINT(0xfbc4, 0x1), /* BR_INST_RETIRED.IND_CALL_PS */ + INTEL_UEVENT_CONSTRAINT(0xfdc4, 0x1), /* BR_INST_RETIRED.REL_CALL_PS */ + INTEL_UEVENT_CONSTRAINT(0xfec4, 0x1), /* BR_INST_RETIRED.TAKEN_JCC_PS */ + INTEL_UEVENT_CONSTRAINT(0x00c5, 0x1), /* BR_INST_MISP_RETIRED.ALL_BRANCHES_PS */ + INTEL_UEVENT_CONSTRAINT(0x7ec5, 0x1), /* BR_INST_MISP_RETIRED.JCC_PS */ + INTEL_UEVENT_CONSTRAINT(0xebc5, 0x1), /* BR_INST_MISP_RETIRED.NON_RETURN_IND_PS */ + INTEL_UEVENT_CONSTRAINT(0xf7c5, 0x1), /* BR_INST_MISP_RETIRED.RETURN_PS */ + INTEL_UEVENT_CONSTRAINT(0xfbc5, 0x1), /* BR_INST_MISP_RETIRED.IND_CALL_PS */ + INTEL_UEVENT_CONSTRAINT(0xfec5, 0x1), /* BR_INST_MISP_RETIRED.TAKEN_JCC_PS */ + EVENT_CONSTRAINT_END +}; + struct event_constraint intel_nehalem_pebs_event_constraints[] = { - INTEL_EVENT_CONSTRAINT(0x0b, 0xf), /* MEM_INST_RETIRED.* */ + INTEL_PLD_CONSTRAINT(0x100b, 0xf), /* MEM_INST_RETIRED.* */ INTEL_EVENT_CONSTRAINT(0x0f, 0xf), /* MEM_UNCORE_RETIRED.* */ INTEL_UEVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */ INTEL_EVENT_CONSTRAINT(0xc0, 0xf), /* INST_RETIRED.ANY */ @@ -379,7 +610,7 @@ struct event_constraint intel_nehalem_pebs_event_constraints[] = { }; struct event_constraint intel_westmere_pebs_event_constraints[] = { - INTEL_EVENT_CONSTRAINT(0x0b, 0xf), /* MEM_INST_RETIRED.* */ + INTEL_PLD_CONSTRAINT(0x100b, 0xf), /* MEM_INST_RETIRED.* */ INTEL_EVENT_CONSTRAINT(0x0f, 0xf), /* MEM_UNCORE_RETIRED.* */ INTEL_UEVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */ INTEL_EVENT_CONSTRAINT(0xc0, 0xf), /* INSTR_RETIRED.* */ @@ -399,21 +630,67 @@ struct event_constraint intel_snb_pebs_event_constraints[] = { INTEL_UEVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */ INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */ INTEL_EVENT_CONSTRAINT(0xc5, 0xf), /* BR_MISP_RETIRED.* */ - INTEL_EVENT_CONSTRAINT(0xcd, 0x8), /* MEM_TRANS_RETIRED.* */ - INTEL_UEVENT_CONSTRAINT(0x11d0, 0xf), /* MEM_UOP_RETIRED.STLB_MISS_LOADS */ - INTEL_UEVENT_CONSTRAINT(0x12d0, 0xf), /* MEM_UOP_RETIRED.STLB_MISS_STORES */ - INTEL_UEVENT_CONSTRAINT(0x21d0, 0xf), /* MEM_UOP_RETIRED.LOCK_LOADS */ - INTEL_UEVENT_CONSTRAINT(0x22d0, 0xf), /* MEM_UOP_RETIRED.LOCK_STORES */ - INTEL_UEVENT_CONSTRAINT(0x41d0, 0xf), /* MEM_UOP_RETIRED.SPLIT_LOADS */ - INTEL_UEVENT_CONSTRAINT(0x42d0, 0xf), /* MEM_UOP_RETIRED.SPLIT_STORES */ - INTEL_UEVENT_CONSTRAINT(0x81d0, 0xf), /* MEM_UOP_RETIRED.ANY_LOADS */ - INTEL_UEVENT_CONSTRAINT(0x82d0, 0xf), /* MEM_UOP_RETIRED.ANY_STORES */ + INTEL_PLD_CONSTRAINT(0x01cd, 0x8), /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */ + INTEL_PST_CONSTRAINT(0x02cd, 0x8), /* MEM_TRANS_RETIRED.PRECISE_STORES */ + INTEL_EVENT_CONSTRAINT(0xd0, 0xf), /* MEM_UOP_RETIRED.* */ INTEL_EVENT_CONSTRAINT(0xd1, 0xf), /* MEM_LOAD_UOPS_RETIRED.* */ INTEL_EVENT_CONSTRAINT(0xd2, 0xf), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xd3, 0xf), /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.* */ INTEL_UEVENT_CONSTRAINT(0x02d4, 0xf), /* MEM_LOAD_UOPS_MISC_RETIRED.LLC_MISS */ EVENT_CONSTRAINT_END }; +struct event_constraint intel_ivb_pebs_event_constraints[] = { + INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */ + INTEL_UEVENT_CONSTRAINT(0x01c2, 0xf), /* UOPS_RETIRED.ALL */ + INTEL_UEVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */ + INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xc5, 0xf), /* BR_MISP_RETIRED.* */ + INTEL_PLD_CONSTRAINT(0x01cd, 0x8), /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */ + INTEL_PST_CONSTRAINT(0x02cd, 0x8), /* MEM_TRANS_RETIRED.PRECISE_STORES */ + INTEL_EVENT_CONSTRAINT(0xd0, 0xf), /* MEM_UOP_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xd1, 0xf), /* MEM_LOAD_UOPS_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xd2, 0xf), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */ + INTEL_EVENT_CONSTRAINT(0xd3, 0xf), /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.* */ + EVENT_CONSTRAINT_END +}; + +struct event_constraint intel_hsw_pebs_event_constraints[] = { + INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */ + INTEL_PST_HSW_CONSTRAINT(0x01c2, 0xf), /* UOPS_RETIRED.ALL */ + INTEL_UEVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */ + INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */ + INTEL_UEVENT_CONSTRAINT(0x01c5, 0xf), /* BR_MISP_RETIRED.CONDITIONAL */ + INTEL_UEVENT_CONSTRAINT(0x04c5, 0xf), /* BR_MISP_RETIRED.ALL_BRANCHES */ + INTEL_UEVENT_CONSTRAINT(0x20c5, 0xf), /* BR_MISP_RETIRED.NEAR_TAKEN */ + INTEL_PLD_CONSTRAINT(0x01cd, 0x8), /* MEM_TRANS_RETIRED.* */ + /* MEM_UOPS_RETIRED.STLB_MISS_LOADS */ + INTEL_UEVENT_CONSTRAINT(0x11d0, 0xf), + /* MEM_UOPS_RETIRED.STLB_MISS_STORES */ + INTEL_UEVENT_CONSTRAINT(0x12d0, 0xf), + INTEL_UEVENT_CONSTRAINT(0x21d0, 0xf), /* MEM_UOPS_RETIRED.LOCK_LOADS */ + INTEL_UEVENT_CONSTRAINT(0x41d0, 0xf), /* MEM_UOPS_RETIRED.SPLIT_LOADS */ + /* MEM_UOPS_RETIRED.SPLIT_STORES */ + INTEL_UEVENT_CONSTRAINT(0x42d0, 0xf), + INTEL_UEVENT_CONSTRAINT(0x81d0, 0xf), /* MEM_UOPS_RETIRED.ALL_LOADS */ + INTEL_PST_HSW_CONSTRAINT(0x82d0, 0xf), /* MEM_UOPS_RETIRED.ALL_STORES */ + INTEL_UEVENT_CONSTRAINT(0x01d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L1_HIT */ + INTEL_UEVENT_CONSTRAINT(0x02d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L2_HIT */ + INTEL_UEVENT_CONSTRAINT(0x04d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L3_HIT */ + /* MEM_LOAD_UOPS_RETIRED.HIT_LFB */ + INTEL_UEVENT_CONSTRAINT(0x40d1, 0xf), + /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS */ + INTEL_UEVENT_CONSTRAINT(0x01d2, 0xf), + /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT */ + INTEL_UEVENT_CONSTRAINT(0x02d2, 0xf), + /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM */ + INTEL_UEVENT_CONSTRAINT(0x01d3, 0xf), + INTEL_UEVENT_CONSTRAINT(0x04c8, 0xf), /* HLE_RETIRED.Abort */ + INTEL_UEVENT_CONSTRAINT(0x04c9, 0xf), /* RTM_RETIRED.Abort */ + + EVENT_CONSTRAINT_END +}; + struct event_constraint *intel_pebs_constraints(struct perf_event *event) { struct event_constraint *c; @@ -423,8 +700,10 @@ struct event_constraint *intel_pebs_constraints(struct perf_event *event) if (x86_pmu.pebs_constraints) { for_each_event_constraint(c, x86_pmu.pebs_constraints) { - if ((event->hw.config & c->cmask) == c->code) + if ((event->hw.config & c->cmask) == c->code) { + event->hw.flags |= c->flags; return c; + } } } @@ -440,8 +719,10 @@ void intel_pmu_pebs_enable(struct perf_event *event) cpuc->pebs_enabled |= 1ULL << hwc->idx; - if (x86_pmu.intel_cap.pebs_trap && event->attr.precise_ip > 1) - intel_pmu_lbr_enable(event); + if (event->hw.flags & PERF_X86_EVENT_PEBS_LDLAT) + cpuc->pebs_enabled |= 1ULL << (hwc->idx + 32); + else if (event->hw.flags & PERF_X86_EVENT_PEBS_ST) + cpuc->pebs_enabled |= 1ULL << 63; } void intel_pmu_pebs_disable(struct perf_event *event) @@ -450,13 +731,16 @@ void intel_pmu_pebs_disable(struct perf_event *event) struct hw_perf_event *hwc = &event->hw; cpuc->pebs_enabled &= ~(1ULL << hwc->idx); + + if (event->hw.constraint->flags & PERF_X86_EVENT_PEBS_LDLAT) + cpuc->pebs_enabled &= ~(1ULL << (hwc->idx + 32)); + else if (event->hw.constraint->flags & PERF_X86_EVENT_PEBS_ST) + cpuc->pebs_enabled &= ~(1ULL << 63); + if (cpuc->enabled) wrmsrl(MSR_IA32_PEBS_ENABLE, cpuc->pebs_enabled); hwc->config |= ARCH_PERFMON_EVENTSEL_INT; - - if (x86_pmu.intel_cap.pebs_trap && event->attr.precise_ip > 1) - intel_pmu_lbr_disable(event); } void intel_pmu_pebs_enable_all(void) @@ -475,17 +759,6 @@ void intel_pmu_pebs_disable_all(void) wrmsrl(MSR_IA32_PEBS_ENABLE, 0); } -#include <asm/insn.h> - -static inline bool kernel_ip(unsigned long ip) -{ -#ifdef CONFIG_X86_32 - return ip > PAGE_OFFSET; -#else - return (long)ip < 0; -#endif -} - static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); @@ -493,6 +766,7 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs) unsigned long old_to, to = cpuc->lbr_entries[0].to; unsigned long ip = regs->ip; int is_64bit = 0; + void *kaddr; /* * We don't need to fixup if the PEBS assist is fault like @@ -516,44 +790,48 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs) * unsigned math, either ip is before the start (impossible) or * the basic block is larger than 1 page (sanity) */ - if ((ip - to) > PAGE_SIZE) + if ((ip - to) > PEBS_FIXUP_SIZE) return 0; /* * We sampled a branch insn, rewind using the LBR stack */ if (ip == to) { - regs->ip = from; + set_linear_ip(regs, from); return 1; } + if (!kernel_ip(ip)) { + int size, bytes; + u8 *buf = this_cpu_read(insn_buffer); + + size = ip - to; /* Must fit our buffer, see above */ + bytes = copy_from_user_nmi(buf, (void __user *)to, size); + if (bytes != 0) + return 0; + + kaddr = buf; + } else { + kaddr = (void *)to; + } + do { struct insn insn; - u8 buf[MAX_INSN_SIZE]; - void *kaddr; old_to = to; - if (!kernel_ip(ip)) { - int bytes, size = MAX_INSN_SIZE; - - bytes = copy_from_user_nmi(buf, (void __user *)to, size); - if (bytes != size) - return 0; - - kaddr = buf; - } else - kaddr = (void *)to; #ifdef CONFIG_X86_64 is_64bit = kernel_ip(to) || !test_thread_flag(TIF_IA32); #endif insn_init(&insn, kaddr, is_64bit); insn_get_length(&insn); + to += insn.length; + kaddr += insn.length; } while (to < ip); if (to == ip) { - regs->ip = old_to; + set_linear_ip(regs, old_to); return 1; } @@ -564,23 +842,74 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs) return 0; } +static inline u64 intel_hsw_weight(struct pebs_record_hsw *pebs) +{ + if (pebs->tsx_tuning) { + union hsw_tsx_tuning tsx = { .value = pebs->tsx_tuning }; + return tsx.cycles_last_block; + } + return 0; +} + +static inline u64 intel_hsw_transaction(struct pebs_record_hsw *pebs) +{ + u64 txn = (pebs->tsx_tuning & PEBS_HSW_TSX_FLAGS) >> 32; + + /* For RTM XABORTs also log the abort code from AX */ + if ((txn & PERF_TXN_TRANSACTION) && (pebs->ax & 1)) + txn |= ((pebs->ax >> 24) & 0xff) << PERF_TXN_ABORT_SHIFT; + return txn; +} + static void __intel_pmu_pebs_event(struct perf_event *event, struct pt_regs *iregs, void *__pebs) { /* - * We cast to pebs_record_core since that is a subset of - * both formats and we don't use the other fields in this - * routine. + * We cast to the biggest pebs_record but are careful not to + * unconditionally access the 'extra' entries. */ - struct pebs_record_core *pebs = __pebs; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct pebs_record_hsw *pebs = __pebs; struct perf_sample_data data; struct pt_regs regs; + u64 sample_type; + int fll, fst; if (!intel_pmu_save_and_restart(event)) return; - perf_sample_data_init(&data, 0); + fll = event->hw.flags & PERF_X86_EVENT_PEBS_LDLAT; + fst = event->hw.flags & (PERF_X86_EVENT_PEBS_ST | + PERF_X86_EVENT_PEBS_ST_HSW); + + perf_sample_data_init(&data, 0, event->hw.last_period); + data.period = event->hw.last_period; + sample_type = event->attr.sample_type; + + /* + * if PEBS-LL or PreciseStore + */ + if (fll || fst) { + /* + * Use latency for weight (only avail with PEBS-LL) + */ + if (fll && (sample_type & PERF_SAMPLE_WEIGHT)) + data.weight = pebs->lat; + + /* + * data.data_src encodes the data source + */ + if (sample_type & PERF_SAMPLE_DATA_SRC) { + if (fll) + data.data_src.val = load_latency_data(pebs->dse); + else if (event->hw.flags & PERF_X86_EVENT_PEBS_ST_HSW) + data.data_src.val = + precise_store_data_hsw(event, pebs->dse); + else + data.data_src.val = precise_store_data(pebs->dse); + } + } /* * We use the interrupt regs as a base because the PEBS record @@ -593,15 +922,35 @@ static void __intel_pmu_pebs_event(struct perf_event *event, * A possible PERF_SAMPLE_REGS will have to transfer all regs. */ regs = *iregs; - regs.ip = pebs->ip; + regs.flags = pebs->flags; + set_linear_ip(®s, pebs->ip); regs.bp = pebs->bp; regs.sp = pebs->sp; - if (event->attr.precise_ip > 1 && intel_pmu_pebs_fixup_ip(®s)) + if (event->attr.precise_ip > 1 && x86_pmu.intel_cap.pebs_format >= 2) { + regs.ip = pebs->real_ip; + regs.flags |= PERF_EFLAGS_EXACT; + } else if (event->attr.precise_ip > 1 && intel_pmu_pebs_fixup_ip(®s)) regs.flags |= PERF_EFLAGS_EXACT; else regs.flags &= ~PERF_EFLAGS_EXACT; + if ((event->attr.sample_type & PERF_SAMPLE_ADDR) && + x86_pmu.intel_cap.pebs_format >= 1) + data.addr = pebs->dla; + + if (x86_pmu.intel_cap.pebs_format >= 2) { + /* Only set the TSX weight when no memory weight. */ + if ((event->attr.sample_type & PERF_SAMPLE_WEIGHT) && !fll) + data.weight = intel_hsw_weight(pebs); + + if (event->attr.sample_type & PERF_SAMPLE_TRANSACTION) + data.txn = intel_hsw_transaction(pebs); + } + + if (has_branch_stack(event)) + data.br_stack = &cpuc->lbr_stack; + if (perf_event_overflow(event, &data, ®s)) x86_pmu_stop(event, 0); } @@ -641,7 +990,7 @@ static void intel_pmu_drain_pebs_core(struct pt_regs *iregs) * Should not happen, we program the threshold at 1 and do not * set a reset value. */ - WARN_ON_ONCE(n > 1); + WARN_ONCE(n > 1, "bad leftover pebs %d\n", n); at += n - 1; __intel_pmu_pebs_event(event, iregs, at); @@ -651,10 +1000,10 @@ static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); struct debug_store *ds = cpuc->ds; - struct pebs_record_nhm *at, *top; struct perf_event *event = NULL; + void *at, *top; u64 status = 0; - int bit, n; + int bit; if (!x86_pmu.pebs_active) return; @@ -664,18 +1013,22 @@ static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs) ds->pebs_index = ds->pebs_buffer_base; - n = top - at; - if (n <= 0) + if (unlikely(at > top)) return; /* * Should not happen, we program the threshold at 1 and do not * set a reset value. */ - WARN_ON_ONCE(n > MAX_PEBS_EVENTS); + WARN_ONCE(top - at > x86_pmu.max_pebs_events * x86_pmu.pebs_record_size, + "Unexpected number of pebs records %ld\n", + (long)(top - at) / x86_pmu.pebs_record_size); - for ( ; at < top; at++) { - for_each_set_bit(bit, (unsigned long *)&at->status, MAX_PEBS_EVENTS) { + for (; at < top; at += x86_pmu.pebs_record_size) { + struct pebs_record_nhm *p = at; + + for_each_set_bit(bit, (unsigned long *)&p->status, + x86_pmu.max_pebs_events) { event = cpuc->events[bit]; if (!test_bit(bit, cpuc->active_mask)) continue; @@ -691,7 +1044,7 @@ static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs) break; } - if (!event || bit >= MAX_PEBS_EVENTS) + if (!event || bit >= x86_pmu.max_pebs_events) continue; __intel_pmu_pebs_event(event, iregs, at); @@ -729,9 +1082,25 @@ void intel_ds_init(void) x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm; break; + case 2: + pr_cont("PEBS fmt2%c, ", pebs_type); + x86_pmu.pebs_record_size = sizeof(struct pebs_record_hsw); + x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm; + break; + default: printk(KERN_CONT "no PEBS fmt%d%c, ", format, pebs_type); x86_pmu.pebs = 0; } } } + +void perf_restore_debug_store(void) +{ + struct debug_store *ds = __this_cpu_read(cpu_hw_events.ds); + + if (!x86_pmu.bts && !x86_pmu.pebs) + return; + + wrmsrl(MSR_IA32_DS_AREA, (unsigned long)ds); +} diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c index 47a7e63bfe5..9dd2459a4c7 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c +++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c @@ -3,6 +3,7 @@ #include <asm/perf_event.h> #include <asm/msr.h> +#include <asm/insn.h> #include "perf_event.h" @@ -11,8 +12,119 @@ enum { LBR_FORMAT_LIP = 0x01, LBR_FORMAT_EIP = 0x02, LBR_FORMAT_EIP_FLAGS = 0x03, + LBR_FORMAT_EIP_FLAGS2 = 0x04, + LBR_FORMAT_MAX_KNOWN = LBR_FORMAT_EIP_FLAGS2, }; +static enum { + LBR_EIP_FLAGS = 1, + LBR_TSX = 2, +} lbr_desc[LBR_FORMAT_MAX_KNOWN + 1] = { + [LBR_FORMAT_EIP_FLAGS] = LBR_EIP_FLAGS, + [LBR_FORMAT_EIP_FLAGS2] = LBR_EIP_FLAGS | LBR_TSX, +}; + +/* + * Intel LBR_SELECT bits + * Intel Vol3a, April 2011, Section 16.7 Table 16-10 + * + * Hardware branch filter (not available on all CPUs) + */ +#define LBR_KERNEL_BIT 0 /* do not capture at ring0 */ +#define LBR_USER_BIT 1 /* do not capture at ring > 0 */ +#define LBR_JCC_BIT 2 /* do not capture conditional branches */ +#define LBR_REL_CALL_BIT 3 /* do not capture relative calls */ +#define LBR_IND_CALL_BIT 4 /* do not capture indirect calls */ +#define LBR_RETURN_BIT 5 /* do not capture near returns */ +#define LBR_IND_JMP_BIT 6 /* do not capture indirect jumps */ +#define LBR_REL_JMP_BIT 7 /* do not capture relative jumps */ +#define LBR_FAR_BIT 8 /* do not capture far branches */ + +#define LBR_KERNEL (1 << LBR_KERNEL_BIT) +#define LBR_USER (1 << LBR_USER_BIT) +#define LBR_JCC (1 << LBR_JCC_BIT) +#define LBR_REL_CALL (1 << LBR_REL_CALL_BIT) +#define LBR_IND_CALL (1 << LBR_IND_CALL_BIT) +#define LBR_RETURN (1 << LBR_RETURN_BIT) +#define LBR_REL_JMP (1 << LBR_REL_JMP_BIT) +#define LBR_IND_JMP (1 << LBR_IND_JMP_BIT) +#define LBR_FAR (1 << LBR_FAR_BIT) + +#define LBR_PLM (LBR_KERNEL | LBR_USER) + +#define LBR_SEL_MASK 0x1ff /* valid bits in LBR_SELECT */ +#define LBR_NOT_SUPP -1 /* LBR filter not supported */ +#define LBR_IGN 0 /* ignored */ + +#define LBR_ANY \ + (LBR_JCC |\ + LBR_REL_CALL |\ + LBR_IND_CALL |\ + LBR_RETURN |\ + LBR_REL_JMP |\ + LBR_IND_JMP |\ + LBR_FAR) + +#define LBR_FROM_FLAG_MISPRED (1ULL << 63) +#define LBR_FROM_FLAG_IN_TX (1ULL << 62) +#define LBR_FROM_FLAG_ABORT (1ULL << 61) + +#define for_each_branch_sample_type(x) \ + for ((x) = PERF_SAMPLE_BRANCH_USER; \ + (x) < PERF_SAMPLE_BRANCH_MAX; (x) <<= 1) + +/* + * x86control flow change classification + * x86control flow changes include branches, interrupts, traps, faults + */ +enum { + X86_BR_NONE = 0, /* unknown */ + + X86_BR_USER = 1 << 0, /* branch target is user */ + X86_BR_KERNEL = 1 << 1, /* branch target is kernel */ + + X86_BR_CALL = 1 << 2, /* call */ + X86_BR_RET = 1 << 3, /* return */ + X86_BR_SYSCALL = 1 << 4, /* syscall */ + X86_BR_SYSRET = 1 << 5, /* syscall return */ + X86_BR_INT = 1 << 6, /* sw interrupt */ + X86_BR_IRET = 1 << 7, /* return from interrupt */ + X86_BR_JCC = 1 << 8, /* conditional */ + X86_BR_JMP = 1 << 9, /* jump */ + X86_BR_IRQ = 1 << 10,/* hw interrupt or trap or fault */ + X86_BR_IND_CALL = 1 << 11,/* indirect calls */ + X86_BR_ABORT = 1 << 12,/* transaction abort */ + X86_BR_IN_TX = 1 << 13,/* in transaction */ + X86_BR_NO_TX = 1 << 14,/* not in transaction */ +}; + +#define X86_BR_PLM (X86_BR_USER | X86_BR_KERNEL) +#define X86_BR_ANYTX (X86_BR_NO_TX | X86_BR_IN_TX) + +#define X86_BR_ANY \ + (X86_BR_CALL |\ + X86_BR_RET |\ + X86_BR_SYSCALL |\ + X86_BR_SYSRET |\ + X86_BR_INT |\ + X86_BR_IRET |\ + X86_BR_JCC |\ + X86_BR_JMP |\ + X86_BR_IRQ |\ + X86_BR_ABORT |\ + X86_BR_IND_CALL) + +#define X86_BR_ALL (X86_BR_PLM | X86_BR_ANY) + +#define X86_BR_ANY_CALL \ + (X86_BR_CALL |\ + X86_BR_IND_CALL |\ + X86_BR_SYSCALL |\ + X86_BR_IRQ |\ + X86_BR_INT) + +static void intel_pmu_lbr_filter(struct cpu_hw_events *cpuc); + /* * We only support LBR implementations that have FREEZE_LBRS_ON_PMI * otherwise it becomes near impossible to get a reliable stack. @@ -21,6 +133,10 @@ enum { static void __intel_pmu_lbr_enable(void) { u64 debugctl; + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + + if (cpuc->lbr_sel) + wrmsrl(MSR_LBR_SELECT, cpuc->lbr_sel->config); rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctl); debugctl |= (DEBUGCTLMSR_LBR | DEBUGCTLMSR_FREEZE_LBRS_ON_PMI); @@ -76,11 +192,11 @@ void intel_pmu_lbr_enable(struct perf_event *event) * Reset the LBR stack if we changed task context to * avoid data leaks. */ - if (event->ctx->task && cpuc->lbr_context != event->ctx) { intel_pmu_lbr_reset(); cpuc->lbr_context = event->ctx; } + cpuc->br_sel = event->hw.branch_reg.reg; cpuc->lbr_users++; } @@ -95,8 +211,11 @@ void intel_pmu_lbr_disable(struct perf_event *event) cpuc->lbr_users--; WARN_ON_ONCE(cpuc->lbr_users < 0); - if (cpuc->enabled && !cpuc->lbr_users) + if (cpuc->enabled && !cpuc->lbr_users) { __intel_pmu_lbr_disable(); + /* avoid stale pointer */ + cpuc->lbr_context = NULL; + } } void intel_pmu_lbr_enable_all(void) @@ -115,6 +234,9 @@ void intel_pmu_lbr_disable_all(void) __intel_pmu_lbr_disable(); } +/* + * TOS = most recently recorded branch + */ static inline u64 intel_pmu_lbr_tos(void) { u64 tos; @@ -142,15 +264,15 @@ static void intel_pmu_lbr_read_32(struct cpu_hw_events *cpuc) rdmsrl(x86_pmu.lbr_from + lbr_idx, msr_lastbranch.lbr); - cpuc->lbr_entries[i].from = msr_lastbranch.from; - cpuc->lbr_entries[i].to = msr_lastbranch.to; - cpuc->lbr_entries[i].flags = 0; + cpuc->lbr_entries[i].from = msr_lastbranch.from; + cpuc->lbr_entries[i].to = msr_lastbranch.to; + cpuc->lbr_entries[i].mispred = 0; + cpuc->lbr_entries[i].predicted = 0; + cpuc->lbr_entries[i].reserved = 0; } cpuc->lbr_stack.nr = i; } -#define LBR_FROM_FLAG_MISPRED (1ULL << 63) - /* * Due to lack of segmentation in Linux the effective address (offset) * is the same as the linear address, allowing us to merge the LIP and EIP @@ -162,24 +284,50 @@ static void intel_pmu_lbr_read_64(struct cpu_hw_events *cpuc) int lbr_format = x86_pmu.intel_cap.lbr_format; u64 tos = intel_pmu_lbr_tos(); int i; + int out = 0; for (i = 0; i < x86_pmu.lbr_nr; i++) { unsigned long lbr_idx = (tos - i) & mask; - u64 from, to, flags = 0; + u64 from, to, mis = 0, pred = 0, in_tx = 0, abort = 0; + int skip = 0; + int lbr_flags = lbr_desc[lbr_format]; rdmsrl(x86_pmu.lbr_from + lbr_idx, from); rdmsrl(x86_pmu.lbr_to + lbr_idx, to); - if (lbr_format == LBR_FORMAT_EIP_FLAGS) { - flags = !!(from & LBR_FROM_FLAG_MISPRED); - from = (u64)((((s64)from) << 1) >> 1); + if (lbr_flags & LBR_EIP_FLAGS) { + mis = !!(from & LBR_FROM_FLAG_MISPRED); + pred = !mis; + skip = 1; } - - cpuc->lbr_entries[i].from = from; - cpuc->lbr_entries[i].to = to; - cpuc->lbr_entries[i].flags = flags; + if (lbr_flags & LBR_TSX) { + in_tx = !!(from & LBR_FROM_FLAG_IN_TX); + abort = !!(from & LBR_FROM_FLAG_ABORT); + skip = 3; + } + from = (u64)((((s64)from) << skip) >> skip); + + /* + * Some CPUs report duplicated abort records, + * with the second entry not having an abort bit set. + * Skip them here. This loop runs backwards, + * so we need to undo the previous record. + * If the abort just happened outside the window + * the extra entry cannot be removed. + */ + if (abort && x86_pmu.lbr_double_abort && out > 0) + out--; + + cpuc->lbr_entries[out].from = from; + cpuc->lbr_entries[out].to = to; + cpuc->lbr_entries[out].mispred = mis; + cpuc->lbr_entries[out].predicted = pred; + cpuc->lbr_entries[out].in_tx = in_tx; + cpuc->lbr_entries[out].abort = abort; + cpuc->lbr_entries[out].reserved = 0; + out++; } - cpuc->lbr_stack.nr = i; + cpuc->lbr_stack.nr = out; } void intel_pmu_lbr_read(void) @@ -193,28 +341,439 @@ void intel_pmu_lbr_read(void) intel_pmu_lbr_read_32(cpuc); else intel_pmu_lbr_read_64(cpuc); + + intel_pmu_lbr_filter(cpuc); +} + +/* + * SW filter is used: + * - in case there is no HW filter + * - in case the HW filter has errata or limitations + */ +static void intel_pmu_setup_sw_lbr_filter(struct perf_event *event) +{ + u64 br_type = event->attr.branch_sample_type; + int mask = 0; + + if (br_type & PERF_SAMPLE_BRANCH_USER) + mask |= X86_BR_USER; + + if (br_type & PERF_SAMPLE_BRANCH_KERNEL) + mask |= X86_BR_KERNEL; + + /* we ignore BRANCH_HV here */ + + if (br_type & PERF_SAMPLE_BRANCH_ANY) + mask |= X86_BR_ANY; + + if (br_type & PERF_SAMPLE_BRANCH_ANY_CALL) + mask |= X86_BR_ANY_CALL; + + if (br_type & PERF_SAMPLE_BRANCH_ANY_RETURN) + mask |= X86_BR_RET | X86_BR_IRET | X86_BR_SYSRET; + + if (br_type & PERF_SAMPLE_BRANCH_IND_CALL) + mask |= X86_BR_IND_CALL; + + if (br_type & PERF_SAMPLE_BRANCH_ABORT_TX) + mask |= X86_BR_ABORT; + + if (br_type & PERF_SAMPLE_BRANCH_IN_TX) + mask |= X86_BR_IN_TX; + + if (br_type & PERF_SAMPLE_BRANCH_NO_TX) + mask |= X86_BR_NO_TX; + + if (br_type & PERF_SAMPLE_BRANCH_COND) + mask |= X86_BR_JCC; + + /* + * stash actual user request into reg, it may + * be used by fixup code for some CPU + */ + event->hw.branch_reg.reg = mask; +} + +/* + * setup the HW LBR filter + * Used only when available, may not be enough to disambiguate + * all branches, may need the help of the SW filter + */ +static int intel_pmu_setup_hw_lbr_filter(struct perf_event *event) +{ + struct hw_perf_event_extra *reg; + u64 br_type = event->attr.branch_sample_type; + u64 mask = 0, m; + u64 v; + + for_each_branch_sample_type(m) { + if (!(br_type & m)) + continue; + + v = x86_pmu.lbr_sel_map[m]; + if (v == LBR_NOT_SUPP) + return -EOPNOTSUPP; + + if (v != LBR_IGN) + mask |= v; + } + reg = &event->hw.branch_reg; + reg->idx = EXTRA_REG_LBR; + + /* LBR_SELECT operates in suppress mode so invert mask */ + reg->config = ~mask & x86_pmu.lbr_sel_mask; + + return 0; +} + +int intel_pmu_setup_lbr_filter(struct perf_event *event) +{ + int ret = 0; + + /* + * no LBR on this PMU + */ + if (!x86_pmu.lbr_nr) + return -EOPNOTSUPP; + + /* + * setup SW LBR filter + */ + intel_pmu_setup_sw_lbr_filter(event); + + /* + * setup HW LBR filter, if any + */ + if (x86_pmu.lbr_sel_map) + ret = intel_pmu_setup_hw_lbr_filter(event); + + return ret; } +/* + * return the type of control flow change at address "from" + * intruction is not necessarily a branch (in case of interrupt). + * + * The branch type returned also includes the priv level of the + * target of the control flow change (X86_BR_USER, X86_BR_KERNEL). + * + * If a branch type is unknown OR the instruction cannot be + * decoded (e.g., text page not present), then X86_BR_NONE is + * returned. + */ +static int branch_type(unsigned long from, unsigned long to, int abort) +{ + struct insn insn; + void *addr; + int bytes, size = MAX_INSN_SIZE; + int ret = X86_BR_NONE; + int ext, to_plm, from_plm; + u8 buf[MAX_INSN_SIZE]; + int is64 = 0; + + to_plm = kernel_ip(to) ? X86_BR_KERNEL : X86_BR_USER; + from_plm = kernel_ip(from) ? X86_BR_KERNEL : X86_BR_USER; + + /* + * maybe zero if lbr did not fill up after a reset by the time + * we get a PMU interrupt + */ + if (from == 0 || to == 0) + return X86_BR_NONE; + + if (abort) + return X86_BR_ABORT | to_plm; + + if (from_plm == X86_BR_USER) { + /* + * can happen if measuring at the user level only + * and we interrupt in a kernel thread, e.g., idle. + */ + if (!current->mm) + return X86_BR_NONE; + + /* may fail if text not present */ + bytes = copy_from_user_nmi(buf, (void __user *)from, size); + if (bytes != 0) + return X86_BR_NONE; + + addr = buf; + } else { + /* + * The LBR logs any address in the IP, even if the IP just + * faulted. This means userspace can control the from address. + * Ensure we don't blindy read any address by validating it is + * a known text address. + */ + if (kernel_text_address(from)) + addr = (void *)from; + else + return X86_BR_NONE; + } + + /* + * decoder needs to know the ABI especially + * on 64-bit systems running 32-bit apps + */ +#ifdef CONFIG_X86_64 + is64 = kernel_ip((unsigned long)addr) || !test_thread_flag(TIF_IA32); +#endif + insn_init(&insn, addr, is64); + insn_get_opcode(&insn); + + switch (insn.opcode.bytes[0]) { + case 0xf: + switch (insn.opcode.bytes[1]) { + case 0x05: /* syscall */ + case 0x34: /* sysenter */ + ret = X86_BR_SYSCALL; + break; + case 0x07: /* sysret */ + case 0x35: /* sysexit */ + ret = X86_BR_SYSRET; + break; + case 0x80 ... 0x8f: /* conditional */ + ret = X86_BR_JCC; + break; + default: + ret = X86_BR_NONE; + } + break; + case 0x70 ... 0x7f: /* conditional */ + ret = X86_BR_JCC; + break; + case 0xc2: /* near ret */ + case 0xc3: /* near ret */ + case 0xca: /* far ret */ + case 0xcb: /* far ret */ + ret = X86_BR_RET; + break; + case 0xcf: /* iret */ + ret = X86_BR_IRET; + break; + case 0xcc ... 0xce: /* int */ + ret = X86_BR_INT; + break; + case 0xe8: /* call near rel */ + case 0x9a: /* call far absolute */ + ret = X86_BR_CALL; + break; + case 0xe0 ... 0xe3: /* loop jmp */ + ret = X86_BR_JCC; + break; + case 0xe9 ... 0xeb: /* jmp */ + ret = X86_BR_JMP; + break; + case 0xff: /* call near absolute, call far absolute ind */ + insn_get_modrm(&insn); + ext = (insn.modrm.bytes[0] >> 3) & 0x7; + switch (ext) { + case 2: /* near ind call */ + case 3: /* far ind call */ + ret = X86_BR_IND_CALL; + break; + case 4: + case 5: + ret = X86_BR_JMP; + break; + } + break; + default: + ret = X86_BR_NONE; + } + /* + * interrupts, traps, faults (and thus ring transition) may + * occur on any instructions. Thus, to classify them correctly, + * we need to first look at the from and to priv levels. If they + * are different and to is in the kernel, then it indicates + * a ring transition. If the from instruction is not a ring + * transition instr (syscall, systenter, int), then it means + * it was a irq, trap or fault. + * + * we have no way of detecting kernel to kernel faults. + */ + if (from_plm == X86_BR_USER && to_plm == X86_BR_KERNEL + && ret != X86_BR_SYSCALL && ret != X86_BR_INT) + ret = X86_BR_IRQ; + + /* + * branch priv level determined by target as + * is done by HW when LBR_SELECT is implemented + */ + if (ret != X86_BR_NONE) + ret |= to_plm; + + return ret; +} + +/* + * implement actual branch filter based on user demand. + * Hardware may not exactly satisfy that request, thus + * we need to inspect opcodes. Mismatched branches are + * discarded. Therefore, the number of branches returned + * in PERF_SAMPLE_BRANCH_STACK sample may vary. + */ +static void +intel_pmu_lbr_filter(struct cpu_hw_events *cpuc) +{ + u64 from, to; + int br_sel = cpuc->br_sel; + int i, j, type; + bool compress = false; + + /* if sampling all branches, then nothing to filter */ + if ((br_sel & X86_BR_ALL) == X86_BR_ALL) + return; + + for (i = 0; i < cpuc->lbr_stack.nr; i++) { + + from = cpuc->lbr_entries[i].from; + to = cpuc->lbr_entries[i].to; + + type = branch_type(from, to, cpuc->lbr_entries[i].abort); + if (type != X86_BR_NONE && (br_sel & X86_BR_ANYTX)) { + if (cpuc->lbr_entries[i].in_tx) + type |= X86_BR_IN_TX; + else + type |= X86_BR_NO_TX; + } + + /* if type does not correspond, then discard */ + if (type == X86_BR_NONE || (br_sel & type) != type) { + cpuc->lbr_entries[i].from = 0; + compress = true; + } + } + + if (!compress) + return; + + /* remove all entries with from=0 */ + for (i = 0; i < cpuc->lbr_stack.nr; ) { + if (!cpuc->lbr_entries[i].from) { + j = i; + while (++j < cpuc->lbr_stack.nr) + cpuc->lbr_entries[j-1] = cpuc->lbr_entries[j]; + cpuc->lbr_stack.nr--; + if (!cpuc->lbr_entries[i].from) + continue; + } + i++; + } +} + +/* + * Map interface branch filters onto LBR filters + */ +static const int nhm_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX] = { + [PERF_SAMPLE_BRANCH_ANY] = LBR_ANY, + [PERF_SAMPLE_BRANCH_USER] = LBR_USER, + [PERF_SAMPLE_BRANCH_KERNEL] = LBR_KERNEL, + [PERF_SAMPLE_BRANCH_HV] = LBR_IGN, + [PERF_SAMPLE_BRANCH_ANY_RETURN] = LBR_RETURN | LBR_REL_JMP + | LBR_IND_JMP | LBR_FAR, + /* + * NHM/WSM erratum: must include REL_JMP+IND_JMP to get CALL branches + */ + [PERF_SAMPLE_BRANCH_ANY_CALL] = + LBR_REL_CALL | LBR_IND_CALL | LBR_REL_JMP | LBR_IND_JMP | LBR_FAR, + /* + * NHM/WSM erratum: must include IND_JMP to capture IND_CALL + */ + [PERF_SAMPLE_BRANCH_IND_CALL] = LBR_IND_CALL | LBR_IND_JMP, + [PERF_SAMPLE_BRANCH_COND] = LBR_JCC, +}; + +static const int snb_lbr_sel_map[PERF_SAMPLE_BRANCH_MAX] = { + [PERF_SAMPLE_BRANCH_ANY] = LBR_ANY, + [PERF_SAMPLE_BRANCH_USER] = LBR_USER, + [PERF_SAMPLE_BRANCH_KERNEL] = LBR_KERNEL, + [PERF_SAMPLE_BRANCH_HV] = LBR_IGN, + [PERF_SAMPLE_BRANCH_ANY_RETURN] = LBR_RETURN | LBR_FAR, + [PERF_SAMPLE_BRANCH_ANY_CALL] = LBR_REL_CALL | LBR_IND_CALL + | LBR_FAR, + [PERF_SAMPLE_BRANCH_IND_CALL] = LBR_IND_CALL, + [PERF_SAMPLE_BRANCH_COND] = LBR_JCC, +}; + +/* core */ void intel_pmu_lbr_init_core(void) { x86_pmu.lbr_nr = 4; - x86_pmu.lbr_tos = 0x01c9; - x86_pmu.lbr_from = 0x40; - x86_pmu.lbr_to = 0x60; + x86_pmu.lbr_tos = MSR_LBR_TOS; + x86_pmu.lbr_from = MSR_LBR_CORE_FROM; + x86_pmu.lbr_to = MSR_LBR_CORE_TO; + + /* + * SW branch filter usage: + * - compensate for lack of HW filter + */ + pr_cont("4-deep LBR, "); } +/* nehalem/westmere */ void intel_pmu_lbr_init_nhm(void) { x86_pmu.lbr_nr = 16; - x86_pmu.lbr_tos = 0x01c9; - x86_pmu.lbr_from = 0x680; - x86_pmu.lbr_to = 0x6c0; + x86_pmu.lbr_tos = MSR_LBR_TOS; + x86_pmu.lbr_from = MSR_LBR_NHM_FROM; + x86_pmu.lbr_to = MSR_LBR_NHM_TO; + + x86_pmu.lbr_sel_mask = LBR_SEL_MASK; + x86_pmu.lbr_sel_map = nhm_lbr_sel_map; + + /* + * SW branch filter usage: + * - workaround LBR_SEL errata (see above) + * - support syscall, sysret capture. + * That requires LBR_FAR but that means far + * jmp need to be filtered out + */ + pr_cont("16-deep LBR, "); +} + +/* sandy bridge */ +void intel_pmu_lbr_init_snb(void) +{ + x86_pmu.lbr_nr = 16; + x86_pmu.lbr_tos = MSR_LBR_TOS; + x86_pmu.lbr_from = MSR_LBR_NHM_FROM; + x86_pmu.lbr_to = MSR_LBR_NHM_TO; + + x86_pmu.lbr_sel_mask = LBR_SEL_MASK; + x86_pmu.lbr_sel_map = snb_lbr_sel_map; + + /* + * SW branch filter usage: + * - support syscall, sysret capture. + * That requires LBR_FAR but that means far + * jmp need to be filtered out + */ + pr_cont("16-deep LBR, "); } +/* atom */ void intel_pmu_lbr_init_atom(void) { + /* + * only models starting at stepping 10 seems + * to have an operational LBR which can freeze + * on PMU interrupt + */ + if (boot_cpu_data.x86_model == 28 + && boot_cpu_data.x86_mask < 10) { + pr_cont("LBR disabled due to erratum"); + return; + } + x86_pmu.lbr_nr = 8; - x86_pmu.lbr_tos = 0x01c9; - x86_pmu.lbr_from = 0x40; - x86_pmu.lbr_to = 0x60; + x86_pmu.lbr_tos = MSR_LBR_TOS; + x86_pmu.lbr_from = MSR_LBR_CORE_FROM; + x86_pmu.lbr_to = MSR_LBR_CORE_TO; + + /* + * SW branch filter usage: + * - compensate for lack of HW filter + */ + pr_cont("8-deep LBR, "); } diff --git a/arch/x86/kernel/cpu/perf_event_intel_rapl.c b/arch/x86/kernel/cpu/perf_event_intel_rapl.c new file mode 100644 index 00000000000..619f7699487 --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_intel_rapl.c @@ -0,0 +1,714 @@ +/* + * perf_event_intel_rapl.c: support Intel RAPL energy consumption counters + * Copyright (C) 2013 Google, Inc., Stephane Eranian + * + * Intel RAPL interface is specified in the IA-32 Manual Vol3b + * section 14.7.1 (September 2013) + * + * RAPL provides more controls than just reporting energy consumption + * however here we only expose the 3 energy consumption free running + * counters (pp0, pkg, dram). + * + * Each of those counters increments in a power unit defined by the + * RAPL_POWER_UNIT MSR. On SandyBridge, this unit is 1/(2^16) Joules + * but it can vary. + * + * Counter to rapl events mappings: + * + * pp0 counter: consumption of all physical cores (power plane 0) + * event: rapl_energy_cores + * perf code: 0x1 + * + * pkg counter: consumption of the whole processor package + * event: rapl_energy_pkg + * perf code: 0x2 + * + * dram counter: consumption of the dram domain (servers only) + * event: rapl_energy_dram + * perf code: 0x3 + * + * dram counter: consumption of the builtin-gpu domain (client only) + * event: rapl_energy_gpu + * perf code: 0x4 + * + * We manage those counters as free running (read-only). They may be + * use simultaneously by other tools, such as turbostat. + * + * The events only support system-wide mode counting. There is no + * sampling support because it does not make sense and is not + * supported by the RAPL hardware. + * + * Because we want to avoid floating-point operations in the kernel, + * the events are all reported in fixed point arithmetic (32.32). + * Tools must adjust the counts to convert them to Watts using + * the duration of the measurement. Tools may use a function such as + * ldexp(raw_count, -32); + */ +#include <linux/module.h> +#include <linux/slab.h> +#include <linux/perf_event.h> +#include <asm/cpu_device_id.h> +#include "perf_event.h" + +/* + * RAPL energy status counters + */ +#define RAPL_IDX_PP0_NRG_STAT 0 /* all cores */ +#define INTEL_RAPL_PP0 0x1 /* pseudo-encoding */ +#define RAPL_IDX_PKG_NRG_STAT 1 /* entire package */ +#define INTEL_RAPL_PKG 0x2 /* pseudo-encoding */ +#define RAPL_IDX_RAM_NRG_STAT 2 /* DRAM */ +#define INTEL_RAPL_RAM 0x3 /* pseudo-encoding */ +#define RAPL_IDX_PP1_NRG_STAT 3 /* gpu */ +#define INTEL_RAPL_PP1 0x4 /* pseudo-encoding */ + +/* Clients have PP0, PKG */ +#define RAPL_IDX_CLN (1<<RAPL_IDX_PP0_NRG_STAT|\ + 1<<RAPL_IDX_PKG_NRG_STAT|\ + 1<<RAPL_IDX_PP1_NRG_STAT) + +/* Servers have PP0, PKG, RAM */ +#define RAPL_IDX_SRV (1<<RAPL_IDX_PP0_NRG_STAT|\ + 1<<RAPL_IDX_PKG_NRG_STAT|\ + 1<<RAPL_IDX_RAM_NRG_STAT) + +/* Servers have PP0, PKG, RAM, PP1 */ +#define RAPL_IDX_HSW (1<<RAPL_IDX_PP0_NRG_STAT|\ + 1<<RAPL_IDX_PKG_NRG_STAT|\ + 1<<RAPL_IDX_RAM_NRG_STAT|\ + 1<<RAPL_IDX_PP1_NRG_STAT) + +/* + * event code: LSB 8 bits, passed in attr->config + * any other bit is reserved + */ +#define RAPL_EVENT_MASK 0xFFULL + +#define DEFINE_RAPL_FORMAT_ATTR(_var, _name, _format) \ +static ssize_t __rapl_##_var##_show(struct kobject *kobj, \ + struct kobj_attribute *attr, \ + char *page) \ +{ \ + BUILD_BUG_ON(sizeof(_format) >= PAGE_SIZE); \ + return sprintf(page, _format "\n"); \ +} \ +static struct kobj_attribute format_attr_##_var = \ + __ATTR(_name, 0444, __rapl_##_var##_show, NULL) + +#define RAPL_EVENT_DESC(_name, _config) \ +{ \ + .attr = __ATTR(_name, 0444, rapl_event_show, NULL), \ + .config = _config, \ +} + +#define RAPL_CNTR_WIDTH 32 /* 32-bit rapl counters */ + +struct rapl_pmu { + spinlock_t lock; + int hw_unit; /* 1/2^hw_unit Joule */ + int n_active; /* number of active events */ + struct list_head active_list; + struct pmu *pmu; /* pointer to rapl_pmu_class */ + ktime_t timer_interval; /* in ktime_t unit */ + struct hrtimer hrtimer; +}; + +static struct pmu rapl_pmu_class; +static cpumask_t rapl_cpu_mask; +static int rapl_cntr_mask; + +static DEFINE_PER_CPU(struct rapl_pmu *, rapl_pmu); +static DEFINE_PER_CPU(struct rapl_pmu *, rapl_pmu_to_free); + +static inline u64 rapl_read_counter(struct perf_event *event) +{ + u64 raw; + rdmsrl(event->hw.event_base, raw); + return raw; +} + +static inline u64 rapl_scale(u64 v) +{ + /* + * scale delta to smallest unit (1/2^32) + * users must then scale back: count * 1/(1e9*2^32) to get Joules + * or use ldexp(count, -32). + * Watts = Joules/Time delta + */ + return v << (32 - __get_cpu_var(rapl_pmu)->hw_unit); +} + +static u64 rapl_event_update(struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + u64 prev_raw_count, new_raw_count; + s64 delta, sdelta; + int shift = RAPL_CNTR_WIDTH; + +again: + prev_raw_count = local64_read(&hwc->prev_count); + rdmsrl(event->hw.event_base, new_raw_count); + + if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, + new_raw_count) != prev_raw_count) { + cpu_relax(); + goto again; + } + + /* + * Now we have the new raw value and have updated the prev + * timestamp already. We can now calculate the elapsed delta + * (event-)time and add that to the generic event. + * + * Careful, not all hw sign-extends above the physical width + * of the count. + */ + delta = (new_raw_count << shift) - (prev_raw_count << shift); + delta >>= shift; + + sdelta = rapl_scale(delta); + + local64_add(sdelta, &event->count); + + return new_raw_count; +} + +static void rapl_start_hrtimer(struct rapl_pmu *pmu) +{ + __hrtimer_start_range_ns(&pmu->hrtimer, + pmu->timer_interval, 0, + HRTIMER_MODE_REL_PINNED, 0); +} + +static void rapl_stop_hrtimer(struct rapl_pmu *pmu) +{ + hrtimer_cancel(&pmu->hrtimer); +} + +static enum hrtimer_restart rapl_hrtimer_handle(struct hrtimer *hrtimer) +{ + struct rapl_pmu *pmu = __get_cpu_var(rapl_pmu); + struct perf_event *event; + unsigned long flags; + + if (!pmu->n_active) + return HRTIMER_NORESTART; + + spin_lock_irqsave(&pmu->lock, flags); + + list_for_each_entry(event, &pmu->active_list, active_entry) { + rapl_event_update(event); + } + + spin_unlock_irqrestore(&pmu->lock, flags); + + hrtimer_forward_now(hrtimer, pmu->timer_interval); + + return HRTIMER_RESTART; +} + +static void rapl_hrtimer_init(struct rapl_pmu *pmu) +{ + struct hrtimer *hr = &pmu->hrtimer; + + hrtimer_init(hr, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + hr->function = rapl_hrtimer_handle; +} + +static void __rapl_pmu_event_start(struct rapl_pmu *pmu, + struct perf_event *event) +{ + if (WARN_ON_ONCE(!(event->hw.state & PERF_HES_STOPPED))) + return; + + event->hw.state = 0; + + list_add_tail(&event->active_entry, &pmu->active_list); + + local64_set(&event->hw.prev_count, rapl_read_counter(event)); + + pmu->n_active++; + if (pmu->n_active == 1) + rapl_start_hrtimer(pmu); +} + +static void rapl_pmu_event_start(struct perf_event *event, int mode) +{ + struct rapl_pmu *pmu = __get_cpu_var(rapl_pmu); + unsigned long flags; + + spin_lock_irqsave(&pmu->lock, flags); + __rapl_pmu_event_start(pmu, event); + spin_unlock_irqrestore(&pmu->lock, flags); +} + +static void rapl_pmu_event_stop(struct perf_event *event, int mode) +{ + struct rapl_pmu *pmu = __get_cpu_var(rapl_pmu); + struct hw_perf_event *hwc = &event->hw; + unsigned long flags; + + spin_lock_irqsave(&pmu->lock, flags); + + /* mark event as deactivated and stopped */ + if (!(hwc->state & PERF_HES_STOPPED)) { + WARN_ON_ONCE(pmu->n_active <= 0); + pmu->n_active--; + if (pmu->n_active == 0) + rapl_stop_hrtimer(pmu); + + list_del(&event->active_entry); + + WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED); + hwc->state |= PERF_HES_STOPPED; + } + + /* check if update of sw counter is necessary */ + if ((mode & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) { + /* + * Drain the remaining delta count out of a event + * that we are disabling: + */ + rapl_event_update(event); + hwc->state |= PERF_HES_UPTODATE; + } + + spin_unlock_irqrestore(&pmu->lock, flags); +} + +static int rapl_pmu_event_add(struct perf_event *event, int mode) +{ + struct rapl_pmu *pmu = __get_cpu_var(rapl_pmu); + struct hw_perf_event *hwc = &event->hw; + unsigned long flags; + + spin_lock_irqsave(&pmu->lock, flags); + + hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + + if (mode & PERF_EF_START) + __rapl_pmu_event_start(pmu, event); + + spin_unlock_irqrestore(&pmu->lock, flags); + + return 0; +} + +static void rapl_pmu_event_del(struct perf_event *event, int flags) +{ + rapl_pmu_event_stop(event, PERF_EF_UPDATE); +} + +static int rapl_pmu_event_init(struct perf_event *event) +{ + u64 cfg = event->attr.config & RAPL_EVENT_MASK; + int bit, msr, ret = 0; + + /* only look at RAPL events */ + if (event->attr.type != rapl_pmu_class.type) + return -ENOENT; + + /* check only supported bits are set */ + if (event->attr.config & ~RAPL_EVENT_MASK) + return -EINVAL; + + /* + * check event is known (determines counter) + */ + switch (cfg) { + case INTEL_RAPL_PP0: + bit = RAPL_IDX_PP0_NRG_STAT; + msr = MSR_PP0_ENERGY_STATUS; + break; + case INTEL_RAPL_PKG: + bit = RAPL_IDX_PKG_NRG_STAT; + msr = MSR_PKG_ENERGY_STATUS; + break; + case INTEL_RAPL_RAM: + bit = RAPL_IDX_RAM_NRG_STAT; + msr = MSR_DRAM_ENERGY_STATUS; + break; + case INTEL_RAPL_PP1: + bit = RAPL_IDX_PP1_NRG_STAT; + msr = MSR_PP1_ENERGY_STATUS; + break; + default: + return -EINVAL; + } + /* check event supported */ + if (!(rapl_cntr_mask & (1 << bit))) + return -EINVAL; + + /* unsupported modes and filters */ + if (event->attr.exclude_user || + event->attr.exclude_kernel || + event->attr.exclude_hv || + event->attr.exclude_idle || + event->attr.exclude_host || + event->attr.exclude_guest || + event->attr.sample_period) /* no sampling */ + return -EINVAL; + + /* must be done before validate_group */ + event->hw.event_base = msr; + event->hw.config = cfg; + event->hw.idx = bit; + + return ret; +} + +static void rapl_pmu_event_read(struct perf_event *event) +{ + rapl_event_update(event); +} + +static ssize_t rapl_get_attr_cpumask(struct device *dev, + struct device_attribute *attr, char *buf) +{ + int n = cpulist_scnprintf(buf, PAGE_SIZE - 2, &rapl_cpu_mask); + + buf[n++] = '\n'; + buf[n] = '\0'; + return n; +} + +static DEVICE_ATTR(cpumask, S_IRUGO, rapl_get_attr_cpumask, NULL); + +static struct attribute *rapl_pmu_attrs[] = { + &dev_attr_cpumask.attr, + NULL, +}; + +static struct attribute_group rapl_pmu_attr_group = { + .attrs = rapl_pmu_attrs, +}; + +EVENT_ATTR_STR(energy-cores, rapl_cores, "event=0x01"); +EVENT_ATTR_STR(energy-pkg , rapl_pkg, "event=0x02"); +EVENT_ATTR_STR(energy-ram , rapl_ram, "event=0x03"); +EVENT_ATTR_STR(energy-gpu , rapl_gpu, "event=0x04"); + +EVENT_ATTR_STR(energy-cores.unit, rapl_cores_unit, "Joules"); +EVENT_ATTR_STR(energy-pkg.unit , rapl_pkg_unit, "Joules"); +EVENT_ATTR_STR(energy-ram.unit , rapl_ram_unit, "Joules"); +EVENT_ATTR_STR(energy-gpu.unit , rapl_gpu_unit, "Joules"); + +/* + * we compute in 0.23 nJ increments regardless of MSR + */ +EVENT_ATTR_STR(energy-cores.scale, rapl_cores_scale, "2.3283064365386962890625e-10"); +EVENT_ATTR_STR(energy-pkg.scale, rapl_pkg_scale, "2.3283064365386962890625e-10"); +EVENT_ATTR_STR(energy-ram.scale, rapl_ram_scale, "2.3283064365386962890625e-10"); +EVENT_ATTR_STR(energy-gpu.scale, rapl_gpu_scale, "2.3283064365386962890625e-10"); + +static struct attribute *rapl_events_srv_attr[] = { + EVENT_PTR(rapl_cores), + EVENT_PTR(rapl_pkg), + EVENT_PTR(rapl_ram), + + EVENT_PTR(rapl_cores_unit), + EVENT_PTR(rapl_pkg_unit), + EVENT_PTR(rapl_ram_unit), + + EVENT_PTR(rapl_cores_scale), + EVENT_PTR(rapl_pkg_scale), + EVENT_PTR(rapl_ram_scale), + NULL, +}; + +static struct attribute *rapl_events_cln_attr[] = { + EVENT_PTR(rapl_cores), + EVENT_PTR(rapl_pkg), + EVENT_PTR(rapl_gpu), + + EVENT_PTR(rapl_cores_unit), + EVENT_PTR(rapl_pkg_unit), + EVENT_PTR(rapl_gpu_unit), + + EVENT_PTR(rapl_cores_scale), + EVENT_PTR(rapl_pkg_scale), + EVENT_PTR(rapl_gpu_scale), + NULL, +}; + +static struct attribute *rapl_events_hsw_attr[] = { + EVENT_PTR(rapl_cores), + EVENT_PTR(rapl_pkg), + EVENT_PTR(rapl_gpu), + EVENT_PTR(rapl_ram), + + EVENT_PTR(rapl_cores_unit), + EVENT_PTR(rapl_pkg_unit), + EVENT_PTR(rapl_gpu_unit), + EVENT_PTR(rapl_ram_unit), + + EVENT_PTR(rapl_cores_scale), + EVENT_PTR(rapl_pkg_scale), + EVENT_PTR(rapl_gpu_scale), + EVENT_PTR(rapl_ram_scale), + NULL, +}; + +static struct attribute_group rapl_pmu_events_group = { + .name = "events", + .attrs = NULL, /* patched at runtime */ +}; + +DEFINE_RAPL_FORMAT_ATTR(event, event, "config:0-7"); +static struct attribute *rapl_formats_attr[] = { + &format_attr_event.attr, + NULL, +}; + +static struct attribute_group rapl_pmu_format_group = { + .name = "format", + .attrs = rapl_formats_attr, +}; + +const struct attribute_group *rapl_attr_groups[] = { + &rapl_pmu_attr_group, + &rapl_pmu_format_group, + &rapl_pmu_events_group, + NULL, +}; + +static struct pmu rapl_pmu_class = { + .attr_groups = rapl_attr_groups, + .task_ctx_nr = perf_invalid_context, /* system-wide only */ + .event_init = rapl_pmu_event_init, + .add = rapl_pmu_event_add, /* must have */ + .del = rapl_pmu_event_del, /* must have */ + .start = rapl_pmu_event_start, + .stop = rapl_pmu_event_stop, + .read = rapl_pmu_event_read, +}; + +static void rapl_cpu_exit(int cpu) +{ + struct rapl_pmu *pmu = per_cpu(rapl_pmu, cpu); + int i, phys_id = topology_physical_package_id(cpu); + int target = -1; + + /* find a new cpu on same package */ + for_each_online_cpu(i) { + if (i == cpu) + continue; + if (phys_id == topology_physical_package_id(i)) { + target = i; + break; + } + } + /* + * clear cpu from cpumask + * if was set in cpumask and still some cpu on package, + * then move to new cpu + */ + if (cpumask_test_and_clear_cpu(cpu, &rapl_cpu_mask) && target >= 0) + cpumask_set_cpu(target, &rapl_cpu_mask); + + WARN_ON(cpumask_empty(&rapl_cpu_mask)); + /* + * migrate events and context to new cpu + */ + if (target >= 0) + perf_pmu_migrate_context(pmu->pmu, cpu, target); + + /* cancel overflow polling timer for CPU */ + rapl_stop_hrtimer(pmu); +} + +static void rapl_cpu_init(int cpu) +{ + int i, phys_id = topology_physical_package_id(cpu); + + /* check if phys_is is already covered */ + for_each_cpu(i, &rapl_cpu_mask) { + if (phys_id == topology_physical_package_id(i)) + return; + } + /* was not found, so add it */ + cpumask_set_cpu(cpu, &rapl_cpu_mask); +} + +static int rapl_cpu_prepare(int cpu) +{ + struct rapl_pmu *pmu = per_cpu(rapl_pmu, cpu); + int phys_id = topology_physical_package_id(cpu); + u64 ms; + u64 msr_rapl_power_unit_bits; + + if (pmu) + return 0; + + if (phys_id < 0) + return -1; + + /* protect rdmsrl() to handle virtualization */ + if (rdmsrl_safe(MSR_RAPL_POWER_UNIT, &msr_rapl_power_unit_bits)) + return -1; + + pmu = kzalloc_node(sizeof(*pmu), GFP_KERNEL, cpu_to_node(cpu)); + if (!pmu) + return -1; + + spin_lock_init(&pmu->lock); + + INIT_LIST_HEAD(&pmu->active_list); + + /* + * grab power unit as: 1/2^unit Joules + * + * we cache in local PMU instance + */ + pmu->hw_unit = (msr_rapl_power_unit_bits >> 8) & 0x1FULL; + pmu->pmu = &rapl_pmu_class; + + /* + * use reference of 200W for scaling the timeout + * to avoid missing counter overflows. + * 200W = 200 Joules/sec + * divide interval by 2 to avoid lockstep (2 * 100) + * if hw unit is 32, then we use 2 ms 1/200/2 + */ + if (pmu->hw_unit < 32) + ms = (1000 / (2 * 100)) * (1ULL << (32 - pmu->hw_unit - 1)); + else + ms = 2; + + pmu->timer_interval = ms_to_ktime(ms); + + rapl_hrtimer_init(pmu); + + /* set RAPL pmu for this cpu for now */ + per_cpu(rapl_pmu, cpu) = pmu; + per_cpu(rapl_pmu_to_free, cpu) = NULL; + + return 0; +} + +static void rapl_cpu_kfree(int cpu) +{ + struct rapl_pmu *pmu = per_cpu(rapl_pmu_to_free, cpu); + + kfree(pmu); + + per_cpu(rapl_pmu_to_free, cpu) = NULL; +} + +static int rapl_cpu_dying(int cpu) +{ + struct rapl_pmu *pmu = per_cpu(rapl_pmu, cpu); + + if (!pmu) + return 0; + + per_cpu(rapl_pmu, cpu) = NULL; + + per_cpu(rapl_pmu_to_free, cpu) = pmu; + + return 0; +} + +static int rapl_cpu_notifier(struct notifier_block *self, + unsigned long action, void *hcpu) +{ + unsigned int cpu = (long)hcpu; + + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_UP_PREPARE: + rapl_cpu_prepare(cpu); + break; + case CPU_STARTING: + rapl_cpu_init(cpu); + break; + case CPU_UP_CANCELED: + case CPU_DYING: + rapl_cpu_dying(cpu); + break; + case CPU_ONLINE: + case CPU_DEAD: + rapl_cpu_kfree(cpu); + break; + case CPU_DOWN_PREPARE: + rapl_cpu_exit(cpu); + break; + default: + break; + } + + return NOTIFY_OK; +} + +static const struct x86_cpu_id rapl_cpu_match[] = { + [0] = { .vendor = X86_VENDOR_INTEL, .family = 6 }, + [1] = {}, +}; + +static int __init rapl_pmu_init(void) +{ + struct rapl_pmu *pmu; + int cpu, ret; + + /* + * check for Intel processor family 6 + */ + if (!x86_match_cpu(rapl_cpu_match)) + return 0; + + /* check supported CPU */ + switch (boot_cpu_data.x86_model) { + case 42: /* Sandy Bridge */ + case 58: /* Ivy Bridge */ + rapl_cntr_mask = RAPL_IDX_CLN; + rapl_pmu_events_group.attrs = rapl_events_cln_attr; + break; + case 60: /* Haswell */ + case 69: /* Haswell-Celeron */ + rapl_cntr_mask = RAPL_IDX_HSW; + rapl_pmu_events_group.attrs = rapl_events_hsw_attr; + break; + case 45: /* Sandy Bridge-EP */ + case 62: /* IvyTown */ + rapl_cntr_mask = RAPL_IDX_SRV; + rapl_pmu_events_group.attrs = rapl_events_srv_attr; + break; + + default: + /* unsupported */ + return 0; + } + + cpu_notifier_register_begin(); + + for_each_online_cpu(cpu) { + ret = rapl_cpu_prepare(cpu); + if (ret) + goto out; + rapl_cpu_init(cpu); + } + + __perf_cpu_notifier(rapl_cpu_notifier); + + ret = perf_pmu_register(&rapl_pmu_class, "power", -1); + if (WARN_ON(ret)) { + pr_info("RAPL PMU detected, registration failed (%d), RAPL PMU disabled\n", ret); + cpu_notifier_register_done(); + return -1; + } + + pmu = __get_cpu_var(rapl_pmu); + + pr_info("RAPL PMU detected, hw unit 2^-%d Joules," + " API unit is 2^-32 Joules," + " %d fixed counters" + " %llu ms ovfl timer\n", + pmu->hw_unit, + hweight32(rapl_cntr_mask), + ktime_to_ms(pmu->timer_interval)); + +out: + cpu_notifier_register_done(); + + return 0; +} +device_initcall(rapl_pmu_init); diff --git a/arch/x86/kernel/cpu/perf_event_intel_uncore.c b/arch/x86/kernel/cpu/perf_event_intel_uncore.c new file mode 100644 index 00000000000..ae6552a0701 --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_intel_uncore.c @@ -0,0 +1,4298 @@ +#include "perf_event_intel_uncore.h" + +static struct intel_uncore_type *empty_uncore[] = { NULL, }; +static struct intel_uncore_type **msr_uncores = empty_uncore; +static struct intel_uncore_type **pci_uncores = empty_uncore; +/* pci bus to socket mapping */ +static int pcibus_to_physid[256] = { [0 ... 255] = -1, }; + +static struct pci_dev *extra_pci_dev[UNCORE_SOCKET_MAX][UNCORE_EXTRA_PCI_DEV_MAX]; + +static DEFINE_RAW_SPINLOCK(uncore_box_lock); + +/* mask of cpus that collect uncore events */ +static cpumask_t uncore_cpu_mask; + +/* constraint for the fixed counter */ +static struct event_constraint constraint_fixed = + EVENT_CONSTRAINT(~0ULL, 1 << UNCORE_PMC_IDX_FIXED, ~0ULL); +static struct event_constraint constraint_empty = + EVENT_CONSTRAINT(0, 0, 0); + +#define __BITS_VALUE(x, i, n) ((typeof(x))(((x) >> ((i) * (n))) & \ + ((1ULL << (n)) - 1))) + +DEFINE_UNCORE_FORMAT_ATTR(event, event, "config:0-7"); +DEFINE_UNCORE_FORMAT_ATTR(event_ext, event, "config:0-7,21"); +DEFINE_UNCORE_FORMAT_ATTR(umask, umask, "config:8-15"); +DEFINE_UNCORE_FORMAT_ATTR(edge, edge, "config:18"); +DEFINE_UNCORE_FORMAT_ATTR(tid_en, tid_en, "config:19"); +DEFINE_UNCORE_FORMAT_ATTR(inv, inv, "config:23"); +DEFINE_UNCORE_FORMAT_ATTR(cmask5, cmask, "config:24-28"); +DEFINE_UNCORE_FORMAT_ATTR(cmask8, cmask, "config:24-31"); +DEFINE_UNCORE_FORMAT_ATTR(thresh8, thresh, "config:24-31"); +DEFINE_UNCORE_FORMAT_ATTR(thresh5, thresh, "config:24-28"); +DEFINE_UNCORE_FORMAT_ATTR(occ_sel, occ_sel, "config:14-15"); +DEFINE_UNCORE_FORMAT_ATTR(occ_invert, occ_invert, "config:30"); +DEFINE_UNCORE_FORMAT_ATTR(occ_edge, occ_edge, "config:14-51"); +DEFINE_UNCORE_FORMAT_ATTR(filter_tid, filter_tid, "config1:0-4"); +DEFINE_UNCORE_FORMAT_ATTR(filter_link, filter_link, "config1:5-8"); +DEFINE_UNCORE_FORMAT_ATTR(filter_nid, filter_nid, "config1:10-17"); +DEFINE_UNCORE_FORMAT_ATTR(filter_nid2, filter_nid, "config1:32-47"); +DEFINE_UNCORE_FORMAT_ATTR(filter_state, filter_state, "config1:18-22"); +DEFINE_UNCORE_FORMAT_ATTR(filter_state2, filter_state, "config1:17-22"); +DEFINE_UNCORE_FORMAT_ATTR(filter_opc, filter_opc, "config1:23-31"); +DEFINE_UNCORE_FORMAT_ATTR(filter_opc2, filter_opc, "config1:52-60"); +DEFINE_UNCORE_FORMAT_ATTR(filter_band0, filter_band0, "config1:0-7"); +DEFINE_UNCORE_FORMAT_ATTR(filter_band1, filter_band1, "config1:8-15"); +DEFINE_UNCORE_FORMAT_ATTR(filter_band2, filter_band2, "config1:16-23"); +DEFINE_UNCORE_FORMAT_ATTR(filter_band3, filter_band3, "config1:24-31"); +DEFINE_UNCORE_FORMAT_ATTR(match_rds, match_rds, "config1:48-51"); +DEFINE_UNCORE_FORMAT_ATTR(match_rnid30, match_rnid30, "config1:32-35"); +DEFINE_UNCORE_FORMAT_ATTR(match_rnid4, match_rnid4, "config1:31"); +DEFINE_UNCORE_FORMAT_ATTR(match_dnid, match_dnid, "config1:13-17"); +DEFINE_UNCORE_FORMAT_ATTR(match_mc, match_mc, "config1:9-12"); +DEFINE_UNCORE_FORMAT_ATTR(match_opc, match_opc, "config1:5-8"); +DEFINE_UNCORE_FORMAT_ATTR(match_vnw, match_vnw, "config1:3-4"); +DEFINE_UNCORE_FORMAT_ATTR(match0, match0, "config1:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(match1, match1, "config1:32-63"); +DEFINE_UNCORE_FORMAT_ATTR(mask_rds, mask_rds, "config2:48-51"); +DEFINE_UNCORE_FORMAT_ATTR(mask_rnid30, mask_rnid30, "config2:32-35"); +DEFINE_UNCORE_FORMAT_ATTR(mask_rnid4, mask_rnid4, "config2:31"); +DEFINE_UNCORE_FORMAT_ATTR(mask_dnid, mask_dnid, "config2:13-17"); +DEFINE_UNCORE_FORMAT_ATTR(mask_mc, mask_mc, "config2:9-12"); +DEFINE_UNCORE_FORMAT_ATTR(mask_opc, mask_opc, "config2:5-8"); +DEFINE_UNCORE_FORMAT_ATTR(mask_vnw, mask_vnw, "config2:3-4"); +DEFINE_UNCORE_FORMAT_ATTR(mask0, mask0, "config2:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(mask1, mask1, "config2:32-63"); + +static void uncore_pmu_start_hrtimer(struct intel_uncore_box *box); +static void uncore_pmu_cancel_hrtimer(struct intel_uncore_box *box); +static void uncore_perf_event_update(struct intel_uncore_box *box, struct perf_event *event); +static void uncore_pmu_event_read(struct perf_event *event); + +static struct intel_uncore_pmu *uncore_event_to_pmu(struct perf_event *event) +{ + return container_of(event->pmu, struct intel_uncore_pmu, pmu); +} + +static struct intel_uncore_box * +uncore_pmu_to_box(struct intel_uncore_pmu *pmu, int cpu) +{ + struct intel_uncore_box *box; + + box = *per_cpu_ptr(pmu->box, cpu); + if (box) + return box; + + raw_spin_lock(&uncore_box_lock); + list_for_each_entry(box, &pmu->box_list, list) { + if (box->phys_id == topology_physical_package_id(cpu)) { + atomic_inc(&box->refcnt); + *per_cpu_ptr(pmu->box, cpu) = box; + break; + } + } + raw_spin_unlock(&uncore_box_lock); + + return *per_cpu_ptr(pmu->box, cpu); +} + +static struct intel_uncore_box *uncore_event_to_box(struct perf_event *event) +{ + /* + * perf core schedules event on the basis of cpu, uncore events are + * collected by one of the cpus inside a physical package. + */ + return uncore_pmu_to_box(uncore_event_to_pmu(event), smp_processor_id()); +} + +static u64 uncore_msr_read_counter(struct intel_uncore_box *box, struct perf_event *event) +{ + u64 count; + + rdmsrl(event->hw.event_base, count); + + return count; +} + +/* + * generic get constraint function for shared match/mask registers. + */ +static struct event_constraint * +uncore_get_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct intel_uncore_extra_reg *er; + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct hw_perf_event_extra *reg2 = &event->hw.branch_reg; + unsigned long flags; + bool ok = false; + + /* + * reg->alloc can be set due to existing state, so for fake box we + * need to ignore this, otherwise we might fail to allocate proper + * fake state for this extra reg constraint. + */ + if (reg1->idx == EXTRA_REG_NONE || + (!uncore_box_is_fake(box) && reg1->alloc)) + return NULL; + + er = &box->shared_regs[reg1->idx]; + raw_spin_lock_irqsave(&er->lock, flags); + if (!atomic_read(&er->ref) || + (er->config1 == reg1->config && er->config2 == reg2->config)) { + atomic_inc(&er->ref); + er->config1 = reg1->config; + er->config2 = reg2->config; + ok = true; + } + raw_spin_unlock_irqrestore(&er->lock, flags); + + if (ok) { + if (!uncore_box_is_fake(box)) + reg1->alloc = 1; + return NULL; + } + + return &constraint_empty; +} + +static void uncore_put_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct intel_uncore_extra_reg *er; + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + + /* + * Only put constraint if extra reg was actually allocated. Also + * takes care of event which do not use an extra shared reg. + * + * Also, if this is a fake box we shouldn't touch any event state + * (reg->alloc) and we don't care about leaving inconsistent box + * state either since it will be thrown out. + */ + if (uncore_box_is_fake(box) || !reg1->alloc) + return; + + er = &box->shared_regs[reg1->idx]; + atomic_dec(&er->ref); + reg1->alloc = 0; +} + +static u64 uncore_shared_reg_config(struct intel_uncore_box *box, int idx) +{ + struct intel_uncore_extra_reg *er; + unsigned long flags; + u64 config; + + er = &box->shared_regs[idx]; + + raw_spin_lock_irqsave(&er->lock, flags); + config = er->config; + raw_spin_unlock_irqrestore(&er->lock, flags); + + return config; +} + +/* Sandy Bridge-EP uncore support */ +static struct intel_uncore_type snbep_uncore_cbox; +static struct intel_uncore_type snbep_uncore_pcu; + +static void snbep_uncore_pci_disable_box(struct intel_uncore_box *box) +{ + struct pci_dev *pdev = box->pci_dev; + int box_ctl = uncore_pci_box_ctl(box); + u32 config = 0; + + if (!pci_read_config_dword(pdev, box_ctl, &config)) { + config |= SNBEP_PMON_BOX_CTL_FRZ; + pci_write_config_dword(pdev, box_ctl, config); + } +} + +static void snbep_uncore_pci_enable_box(struct intel_uncore_box *box) +{ + struct pci_dev *pdev = box->pci_dev; + int box_ctl = uncore_pci_box_ctl(box); + u32 config = 0; + + if (!pci_read_config_dword(pdev, box_ctl, &config)) { + config &= ~SNBEP_PMON_BOX_CTL_FRZ; + pci_write_config_dword(pdev, box_ctl, config); + } +} + +static void snbep_uncore_pci_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct pci_dev *pdev = box->pci_dev; + struct hw_perf_event *hwc = &event->hw; + + pci_write_config_dword(pdev, hwc->config_base, hwc->config | SNBEP_PMON_CTL_EN); +} + +static void snbep_uncore_pci_disable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct pci_dev *pdev = box->pci_dev; + struct hw_perf_event *hwc = &event->hw; + + pci_write_config_dword(pdev, hwc->config_base, hwc->config); +} + +static u64 snbep_uncore_pci_read_counter(struct intel_uncore_box *box, struct perf_event *event) +{ + struct pci_dev *pdev = box->pci_dev; + struct hw_perf_event *hwc = &event->hw; + u64 count = 0; + + pci_read_config_dword(pdev, hwc->event_base, (u32 *)&count); + pci_read_config_dword(pdev, hwc->event_base + 4, (u32 *)&count + 1); + + return count; +} + +static void snbep_uncore_pci_init_box(struct intel_uncore_box *box) +{ + struct pci_dev *pdev = box->pci_dev; + + pci_write_config_dword(pdev, SNBEP_PCI_PMON_BOX_CTL, SNBEP_PMON_BOX_CTL_INT); +} + +static void snbep_uncore_msr_disable_box(struct intel_uncore_box *box) +{ + u64 config; + unsigned msr; + + msr = uncore_msr_box_ctl(box); + if (msr) { + rdmsrl(msr, config); + config |= SNBEP_PMON_BOX_CTL_FRZ; + wrmsrl(msr, config); + } +} + +static void snbep_uncore_msr_enable_box(struct intel_uncore_box *box) +{ + u64 config; + unsigned msr; + + msr = uncore_msr_box_ctl(box); + if (msr) { + rdmsrl(msr, config); + config &= ~SNBEP_PMON_BOX_CTL_FRZ; + wrmsrl(msr, config); + } +} + +static void snbep_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + + if (reg1->idx != EXTRA_REG_NONE) + wrmsrl(reg1->reg, uncore_shared_reg_config(box, 0)); + + wrmsrl(hwc->config_base, hwc->config | SNBEP_PMON_CTL_EN); +} + +static void snbep_uncore_msr_disable_event(struct intel_uncore_box *box, + struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + + wrmsrl(hwc->config_base, hwc->config); +} + +static void snbep_uncore_msr_init_box(struct intel_uncore_box *box) +{ + unsigned msr = uncore_msr_box_ctl(box); + + if (msr) + wrmsrl(msr, SNBEP_PMON_BOX_CTL_INT); +} + +static struct attribute *snbep_uncore_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh8.attr, + NULL, +}; + +static struct attribute *snbep_uncore_ubox_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh5.attr, + NULL, +}; + +static struct attribute *snbep_uncore_cbox_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_tid_en.attr, + &format_attr_inv.attr, + &format_attr_thresh8.attr, + &format_attr_filter_tid.attr, + &format_attr_filter_nid.attr, + &format_attr_filter_state.attr, + &format_attr_filter_opc.attr, + NULL, +}; + +static struct attribute *snbep_uncore_pcu_formats_attr[] = { + &format_attr_event_ext.attr, + &format_attr_occ_sel.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh5.attr, + &format_attr_occ_invert.attr, + &format_attr_occ_edge.attr, + &format_attr_filter_band0.attr, + &format_attr_filter_band1.attr, + &format_attr_filter_band2.attr, + &format_attr_filter_band3.attr, + NULL, +}; + +static struct attribute *snbep_uncore_qpi_formats_attr[] = { + &format_attr_event_ext.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh8.attr, + &format_attr_match_rds.attr, + &format_attr_match_rnid30.attr, + &format_attr_match_rnid4.attr, + &format_attr_match_dnid.attr, + &format_attr_match_mc.attr, + &format_attr_match_opc.attr, + &format_attr_match_vnw.attr, + &format_attr_match0.attr, + &format_attr_match1.attr, + &format_attr_mask_rds.attr, + &format_attr_mask_rnid30.attr, + &format_attr_mask_rnid4.attr, + &format_attr_mask_dnid.attr, + &format_attr_mask_mc.attr, + &format_attr_mask_opc.attr, + &format_attr_mask_vnw.attr, + &format_attr_mask0.attr, + &format_attr_mask1.attr, + NULL, +}; + +static struct uncore_event_desc snbep_uncore_imc_events[] = { + INTEL_UNCORE_EVENT_DESC(clockticks, "event=0xff,umask=0x00"), + INTEL_UNCORE_EVENT_DESC(cas_count_read, "event=0x04,umask=0x03"), + INTEL_UNCORE_EVENT_DESC(cas_count_write, "event=0x04,umask=0x0c"), + { /* end: all zeroes */ }, +}; + +static struct uncore_event_desc snbep_uncore_qpi_events[] = { + INTEL_UNCORE_EVENT_DESC(clockticks, "event=0x14"), + INTEL_UNCORE_EVENT_DESC(txl_flits_active, "event=0x00,umask=0x06"), + INTEL_UNCORE_EVENT_DESC(drs_data, "event=0x102,umask=0x08"), + INTEL_UNCORE_EVENT_DESC(ncb_data, "event=0x103,umask=0x04"), + { /* end: all zeroes */ }, +}; + +static struct attribute_group snbep_uncore_format_group = { + .name = "format", + .attrs = snbep_uncore_formats_attr, +}; + +static struct attribute_group snbep_uncore_ubox_format_group = { + .name = "format", + .attrs = snbep_uncore_ubox_formats_attr, +}; + +static struct attribute_group snbep_uncore_cbox_format_group = { + .name = "format", + .attrs = snbep_uncore_cbox_formats_attr, +}; + +static struct attribute_group snbep_uncore_pcu_format_group = { + .name = "format", + .attrs = snbep_uncore_pcu_formats_attr, +}; + +static struct attribute_group snbep_uncore_qpi_format_group = { + .name = "format", + .attrs = snbep_uncore_qpi_formats_attr, +}; + +#define SNBEP_UNCORE_MSR_OPS_COMMON_INIT() \ + .init_box = snbep_uncore_msr_init_box, \ + .disable_box = snbep_uncore_msr_disable_box, \ + .enable_box = snbep_uncore_msr_enable_box, \ + .disable_event = snbep_uncore_msr_disable_event, \ + .enable_event = snbep_uncore_msr_enable_event, \ + .read_counter = uncore_msr_read_counter + +static struct intel_uncore_ops snbep_uncore_msr_ops = { + SNBEP_UNCORE_MSR_OPS_COMMON_INIT(), +}; + +#define SNBEP_UNCORE_PCI_OPS_COMMON_INIT() \ + .init_box = snbep_uncore_pci_init_box, \ + .disable_box = snbep_uncore_pci_disable_box, \ + .enable_box = snbep_uncore_pci_enable_box, \ + .disable_event = snbep_uncore_pci_disable_event, \ + .read_counter = snbep_uncore_pci_read_counter + +static struct intel_uncore_ops snbep_uncore_pci_ops = { + SNBEP_UNCORE_PCI_OPS_COMMON_INIT(), + .enable_event = snbep_uncore_pci_enable_event, \ +}; + +static struct event_constraint snbep_uncore_cbox_constraints[] = { + UNCORE_EVENT_CONSTRAINT(0x01, 0x1), + UNCORE_EVENT_CONSTRAINT(0x02, 0x3), + UNCORE_EVENT_CONSTRAINT(0x04, 0x3), + UNCORE_EVENT_CONSTRAINT(0x05, 0x3), + UNCORE_EVENT_CONSTRAINT(0x07, 0x3), + UNCORE_EVENT_CONSTRAINT(0x09, 0x3), + UNCORE_EVENT_CONSTRAINT(0x11, 0x1), + UNCORE_EVENT_CONSTRAINT(0x12, 0x3), + UNCORE_EVENT_CONSTRAINT(0x13, 0x3), + UNCORE_EVENT_CONSTRAINT(0x1b, 0xc), + UNCORE_EVENT_CONSTRAINT(0x1c, 0xc), + UNCORE_EVENT_CONSTRAINT(0x1d, 0xc), + UNCORE_EVENT_CONSTRAINT(0x1e, 0xc), + EVENT_CONSTRAINT_OVERLAP(0x1f, 0xe, 0xff), + UNCORE_EVENT_CONSTRAINT(0x21, 0x3), + UNCORE_EVENT_CONSTRAINT(0x23, 0x3), + UNCORE_EVENT_CONSTRAINT(0x31, 0x3), + UNCORE_EVENT_CONSTRAINT(0x32, 0x3), + UNCORE_EVENT_CONSTRAINT(0x33, 0x3), + UNCORE_EVENT_CONSTRAINT(0x34, 0x3), + UNCORE_EVENT_CONSTRAINT(0x35, 0x3), + UNCORE_EVENT_CONSTRAINT(0x36, 0x1), + UNCORE_EVENT_CONSTRAINT(0x37, 0x3), + UNCORE_EVENT_CONSTRAINT(0x38, 0x3), + UNCORE_EVENT_CONSTRAINT(0x39, 0x3), + UNCORE_EVENT_CONSTRAINT(0x3b, 0x1), + EVENT_CONSTRAINT_END +}; + +static struct event_constraint snbep_uncore_r2pcie_constraints[] = { + UNCORE_EVENT_CONSTRAINT(0x10, 0x3), + UNCORE_EVENT_CONSTRAINT(0x11, 0x3), + UNCORE_EVENT_CONSTRAINT(0x12, 0x1), + UNCORE_EVENT_CONSTRAINT(0x23, 0x3), + UNCORE_EVENT_CONSTRAINT(0x24, 0x3), + UNCORE_EVENT_CONSTRAINT(0x25, 0x3), + UNCORE_EVENT_CONSTRAINT(0x26, 0x3), + UNCORE_EVENT_CONSTRAINT(0x32, 0x3), + UNCORE_EVENT_CONSTRAINT(0x33, 0x3), + UNCORE_EVENT_CONSTRAINT(0x34, 0x3), + EVENT_CONSTRAINT_END +}; + +static struct event_constraint snbep_uncore_r3qpi_constraints[] = { + UNCORE_EVENT_CONSTRAINT(0x10, 0x3), + UNCORE_EVENT_CONSTRAINT(0x11, 0x3), + UNCORE_EVENT_CONSTRAINT(0x12, 0x3), + UNCORE_EVENT_CONSTRAINT(0x13, 0x1), + UNCORE_EVENT_CONSTRAINT(0x20, 0x3), + UNCORE_EVENT_CONSTRAINT(0x21, 0x3), + UNCORE_EVENT_CONSTRAINT(0x22, 0x3), + UNCORE_EVENT_CONSTRAINT(0x23, 0x3), + UNCORE_EVENT_CONSTRAINT(0x24, 0x3), + UNCORE_EVENT_CONSTRAINT(0x25, 0x3), + UNCORE_EVENT_CONSTRAINT(0x26, 0x3), + UNCORE_EVENT_CONSTRAINT(0x28, 0x3), + UNCORE_EVENT_CONSTRAINT(0x29, 0x3), + UNCORE_EVENT_CONSTRAINT(0x2a, 0x3), + UNCORE_EVENT_CONSTRAINT(0x2b, 0x3), + UNCORE_EVENT_CONSTRAINT(0x2c, 0x3), + UNCORE_EVENT_CONSTRAINT(0x2d, 0x3), + UNCORE_EVENT_CONSTRAINT(0x2e, 0x3), + UNCORE_EVENT_CONSTRAINT(0x2f, 0x3), + UNCORE_EVENT_CONSTRAINT(0x30, 0x3), + UNCORE_EVENT_CONSTRAINT(0x31, 0x3), + UNCORE_EVENT_CONSTRAINT(0x32, 0x3), + UNCORE_EVENT_CONSTRAINT(0x33, 0x3), + UNCORE_EVENT_CONSTRAINT(0x34, 0x3), + UNCORE_EVENT_CONSTRAINT(0x36, 0x3), + UNCORE_EVENT_CONSTRAINT(0x37, 0x3), + UNCORE_EVENT_CONSTRAINT(0x38, 0x3), + UNCORE_EVENT_CONSTRAINT(0x39, 0x3), + EVENT_CONSTRAINT_END +}; + +static struct intel_uncore_type snbep_uncore_ubox = { + .name = "ubox", + .num_counters = 2, + .num_boxes = 1, + .perf_ctr_bits = 44, + .fixed_ctr_bits = 48, + .perf_ctr = SNBEP_U_MSR_PMON_CTR0, + .event_ctl = SNBEP_U_MSR_PMON_CTL0, + .event_mask = SNBEP_U_MSR_PMON_RAW_EVENT_MASK, + .fixed_ctr = SNBEP_U_MSR_PMON_UCLK_FIXED_CTR, + .fixed_ctl = SNBEP_U_MSR_PMON_UCLK_FIXED_CTL, + .ops = &snbep_uncore_msr_ops, + .format_group = &snbep_uncore_ubox_format_group, +}; + +static struct extra_reg snbep_uncore_cbox_extra_regs[] = { + SNBEP_CBO_EVENT_EXTRA_REG(SNBEP_CBO_PMON_CTL_TID_EN, + SNBEP_CBO_PMON_CTL_TID_EN, 0x1), + SNBEP_CBO_EVENT_EXTRA_REG(0x0334, 0xffff, 0x4), + SNBEP_CBO_EVENT_EXTRA_REG(0x4334, 0xffff, 0x6), + SNBEP_CBO_EVENT_EXTRA_REG(0x0534, 0xffff, 0x4), + SNBEP_CBO_EVENT_EXTRA_REG(0x4534, 0xffff, 0x6), + SNBEP_CBO_EVENT_EXTRA_REG(0x0934, 0xffff, 0x4), + SNBEP_CBO_EVENT_EXTRA_REG(0x4934, 0xffff, 0x6), + SNBEP_CBO_EVENT_EXTRA_REG(0x4134, 0xffff, 0x6), + SNBEP_CBO_EVENT_EXTRA_REG(0x0135, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x0335, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x4135, 0xffff, 0xa), + SNBEP_CBO_EVENT_EXTRA_REG(0x4335, 0xffff, 0xa), + SNBEP_CBO_EVENT_EXTRA_REG(0x4435, 0xffff, 0x2), + SNBEP_CBO_EVENT_EXTRA_REG(0x4835, 0xffff, 0x2), + SNBEP_CBO_EVENT_EXTRA_REG(0x4a35, 0xffff, 0x2), + SNBEP_CBO_EVENT_EXTRA_REG(0x5035, 0xffff, 0x2), + SNBEP_CBO_EVENT_EXTRA_REG(0x0136, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x0336, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x4136, 0xffff, 0xa), + SNBEP_CBO_EVENT_EXTRA_REG(0x4336, 0xffff, 0xa), + SNBEP_CBO_EVENT_EXTRA_REG(0x4436, 0xffff, 0x2), + SNBEP_CBO_EVENT_EXTRA_REG(0x4836, 0xffff, 0x2), + SNBEP_CBO_EVENT_EXTRA_REG(0x4a36, 0xffff, 0x2), + SNBEP_CBO_EVENT_EXTRA_REG(0x4037, 0x40ff, 0x2), + EVENT_EXTRA_END +}; + +static void snbep_cbox_put_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct intel_uncore_extra_reg *er = &box->shared_regs[0]; + int i; + + if (uncore_box_is_fake(box)) + return; + + for (i = 0; i < 5; i++) { + if (reg1->alloc & (0x1 << i)) + atomic_sub(1 << (i * 6), &er->ref); + } + reg1->alloc = 0; +} + +static struct event_constraint * +__snbep_cbox_get_constraint(struct intel_uncore_box *box, struct perf_event *event, + u64 (*cbox_filter_mask)(int fields)) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct intel_uncore_extra_reg *er = &box->shared_regs[0]; + int i, alloc = 0; + unsigned long flags; + u64 mask; + + if (reg1->idx == EXTRA_REG_NONE) + return NULL; + + raw_spin_lock_irqsave(&er->lock, flags); + for (i = 0; i < 5; i++) { + if (!(reg1->idx & (0x1 << i))) + continue; + if (!uncore_box_is_fake(box) && (reg1->alloc & (0x1 << i))) + continue; + + mask = cbox_filter_mask(0x1 << i); + if (!__BITS_VALUE(atomic_read(&er->ref), i, 6) || + !((reg1->config ^ er->config) & mask)) { + atomic_add(1 << (i * 6), &er->ref); + er->config &= ~mask; + er->config |= reg1->config & mask; + alloc |= (0x1 << i); + } else { + break; + } + } + raw_spin_unlock_irqrestore(&er->lock, flags); + if (i < 5) + goto fail; + + if (!uncore_box_is_fake(box)) + reg1->alloc |= alloc; + + return NULL; +fail: + for (; i >= 0; i--) { + if (alloc & (0x1 << i)) + atomic_sub(1 << (i * 6), &er->ref); + } + return &constraint_empty; +} + +static u64 snbep_cbox_filter_mask(int fields) +{ + u64 mask = 0; + + if (fields & 0x1) + mask |= SNBEP_CB0_MSR_PMON_BOX_FILTER_TID; + if (fields & 0x2) + mask |= SNBEP_CB0_MSR_PMON_BOX_FILTER_NID; + if (fields & 0x4) + mask |= SNBEP_CB0_MSR_PMON_BOX_FILTER_STATE; + if (fields & 0x8) + mask |= SNBEP_CB0_MSR_PMON_BOX_FILTER_OPC; + + return mask; +} + +static struct event_constraint * +snbep_cbox_get_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + return __snbep_cbox_get_constraint(box, event, snbep_cbox_filter_mask); +} + +static int snbep_cbox_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct extra_reg *er; + int idx = 0; + + for (er = snbep_uncore_cbox_extra_regs; er->msr; er++) { + if (er->event != (event->hw.config & er->config_mask)) + continue; + idx |= er->idx; + } + + if (idx) { + reg1->reg = SNBEP_C0_MSR_PMON_BOX_FILTER + + SNBEP_CBO_MSR_OFFSET * box->pmu->pmu_idx; + reg1->config = event->attr.config1 & snbep_cbox_filter_mask(idx); + reg1->idx = idx; + } + return 0; +} + +static struct intel_uncore_ops snbep_uncore_cbox_ops = { + SNBEP_UNCORE_MSR_OPS_COMMON_INIT(), + .hw_config = snbep_cbox_hw_config, + .get_constraint = snbep_cbox_get_constraint, + .put_constraint = snbep_cbox_put_constraint, +}; + +static struct intel_uncore_type snbep_uncore_cbox = { + .name = "cbox", + .num_counters = 4, + .num_boxes = 8, + .perf_ctr_bits = 44, + .event_ctl = SNBEP_C0_MSR_PMON_CTL0, + .perf_ctr = SNBEP_C0_MSR_PMON_CTR0, + .event_mask = SNBEP_CBO_MSR_PMON_RAW_EVENT_MASK, + .box_ctl = SNBEP_C0_MSR_PMON_BOX_CTL, + .msr_offset = SNBEP_CBO_MSR_OFFSET, + .num_shared_regs = 1, + .constraints = snbep_uncore_cbox_constraints, + .ops = &snbep_uncore_cbox_ops, + .format_group = &snbep_uncore_cbox_format_group, +}; + +static u64 snbep_pcu_alter_er(struct perf_event *event, int new_idx, bool modify) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + u64 config = reg1->config; + + if (new_idx > reg1->idx) + config <<= 8 * (new_idx - reg1->idx); + else + config >>= 8 * (reg1->idx - new_idx); + + if (modify) { + hwc->config += new_idx - reg1->idx; + reg1->config = config; + reg1->idx = new_idx; + } + return config; +} + +static struct event_constraint * +snbep_pcu_get_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct intel_uncore_extra_reg *er = &box->shared_regs[0]; + unsigned long flags; + int idx = reg1->idx; + u64 mask, config1 = reg1->config; + bool ok = false; + + if (reg1->idx == EXTRA_REG_NONE || + (!uncore_box_is_fake(box) && reg1->alloc)) + return NULL; +again: + mask = 0xffULL << (idx * 8); + raw_spin_lock_irqsave(&er->lock, flags); + if (!__BITS_VALUE(atomic_read(&er->ref), idx, 8) || + !((config1 ^ er->config) & mask)) { + atomic_add(1 << (idx * 8), &er->ref); + er->config &= ~mask; + er->config |= config1 & mask; + ok = true; + } + raw_spin_unlock_irqrestore(&er->lock, flags); + + if (!ok) { + idx = (idx + 1) % 4; + if (idx != reg1->idx) { + config1 = snbep_pcu_alter_er(event, idx, false); + goto again; + } + return &constraint_empty; + } + + if (!uncore_box_is_fake(box)) { + if (idx != reg1->idx) + snbep_pcu_alter_er(event, idx, true); + reg1->alloc = 1; + } + return NULL; +} + +static void snbep_pcu_put_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct intel_uncore_extra_reg *er = &box->shared_regs[0]; + + if (uncore_box_is_fake(box) || !reg1->alloc) + return; + + atomic_sub(1 << (reg1->idx * 8), &er->ref); + reg1->alloc = 0; +} + +static int snbep_pcu_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + int ev_sel = hwc->config & SNBEP_PMON_CTL_EV_SEL_MASK; + + if (ev_sel >= 0xb && ev_sel <= 0xe) { + reg1->reg = SNBEP_PCU_MSR_PMON_BOX_FILTER; + reg1->idx = ev_sel - 0xb; + reg1->config = event->attr.config1 & (0xff << reg1->idx); + } + return 0; +} + +static struct intel_uncore_ops snbep_uncore_pcu_ops = { + SNBEP_UNCORE_MSR_OPS_COMMON_INIT(), + .hw_config = snbep_pcu_hw_config, + .get_constraint = snbep_pcu_get_constraint, + .put_constraint = snbep_pcu_put_constraint, +}; + +static struct intel_uncore_type snbep_uncore_pcu = { + .name = "pcu", + .num_counters = 4, + .num_boxes = 1, + .perf_ctr_bits = 48, + .perf_ctr = SNBEP_PCU_MSR_PMON_CTR0, + .event_ctl = SNBEP_PCU_MSR_PMON_CTL0, + .event_mask = SNBEP_PCU_MSR_PMON_RAW_EVENT_MASK, + .box_ctl = SNBEP_PCU_MSR_PMON_BOX_CTL, + .num_shared_regs = 1, + .ops = &snbep_uncore_pcu_ops, + .format_group = &snbep_uncore_pcu_format_group, +}; + +static struct intel_uncore_type *snbep_msr_uncores[] = { + &snbep_uncore_ubox, + &snbep_uncore_cbox, + &snbep_uncore_pcu, + NULL, +}; + +enum { + SNBEP_PCI_QPI_PORT0_FILTER, + SNBEP_PCI_QPI_PORT1_FILTER, +}; + +static int snbep_qpi_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + + if ((hwc->config & SNBEP_PMON_CTL_EV_SEL_MASK) == 0x38) { + reg1->idx = 0; + reg1->reg = SNBEP_Q_Py_PCI_PMON_PKT_MATCH0; + reg1->config = event->attr.config1; + reg2->reg = SNBEP_Q_Py_PCI_PMON_PKT_MASK0; + reg2->config = event->attr.config2; + } + return 0; +} + +static void snbep_qpi_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct pci_dev *pdev = box->pci_dev; + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + + if (reg1->idx != EXTRA_REG_NONE) { + int idx = box->pmu->pmu_idx + SNBEP_PCI_QPI_PORT0_FILTER; + struct pci_dev *filter_pdev = extra_pci_dev[box->phys_id][idx]; + WARN_ON_ONCE(!filter_pdev); + if (filter_pdev) { + pci_write_config_dword(filter_pdev, reg1->reg, + (u32)reg1->config); + pci_write_config_dword(filter_pdev, reg1->reg + 4, + (u32)(reg1->config >> 32)); + pci_write_config_dword(filter_pdev, reg2->reg, + (u32)reg2->config); + pci_write_config_dword(filter_pdev, reg2->reg + 4, + (u32)(reg2->config >> 32)); + } + } + + pci_write_config_dword(pdev, hwc->config_base, hwc->config | SNBEP_PMON_CTL_EN); +} + +static struct intel_uncore_ops snbep_uncore_qpi_ops = { + SNBEP_UNCORE_PCI_OPS_COMMON_INIT(), + .enable_event = snbep_qpi_enable_event, + .hw_config = snbep_qpi_hw_config, + .get_constraint = uncore_get_constraint, + .put_constraint = uncore_put_constraint, +}; + +#define SNBEP_UNCORE_PCI_COMMON_INIT() \ + .perf_ctr = SNBEP_PCI_PMON_CTR0, \ + .event_ctl = SNBEP_PCI_PMON_CTL0, \ + .event_mask = SNBEP_PMON_RAW_EVENT_MASK, \ + .box_ctl = SNBEP_PCI_PMON_BOX_CTL, \ + .ops = &snbep_uncore_pci_ops, \ + .format_group = &snbep_uncore_format_group + +static struct intel_uncore_type snbep_uncore_ha = { + .name = "ha", + .num_counters = 4, + .num_boxes = 1, + .perf_ctr_bits = 48, + SNBEP_UNCORE_PCI_COMMON_INIT(), +}; + +static struct intel_uncore_type snbep_uncore_imc = { + .name = "imc", + .num_counters = 4, + .num_boxes = 4, + .perf_ctr_bits = 48, + .fixed_ctr_bits = 48, + .fixed_ctr = SNBEP_MC_CHy_PCI_PMON_FIXED_CTR, + .fixed_ctl = SNBEP_MC_CHy_PCI_PMON_FIXED_CTL, + .event_descs = snbep_uncore_imc_events, + SNBEP_UNCORE_PCI_COMMON_INIT(), +}; + +static struct intel_uncore_type snbep_uncore_qpi = { + .name = "qpi", + .num_counters = 4, + .num_boxes = 2, + .perf_ctr_bits = 48, + .perf_ctr = SNBEP_PCI_PMON_CTR0, + .event_ctl = SNBEP_PCI_PMON_CTL0, + .event_mask = SNBEP_QPI_PCI_PMON_RAW_EVENT_MASK, + .box_ctl = SNBEP_PCI_PMON_BOX_CTL, + .num_shared_regs = 1, + .ops = &snbep_uncore_qpi_ops, + .event_descs = snbep_uncore_qpi_events, + .format_group = &snbep_uncore_qpi_format_group, +}; + + +static struct intel_uncore_type snbep_uncore_r2pcie = { + .name = "r2pcie", + .num_counters = 4, + .num_boxes = 1, + .perf_ctr_bits = 44, + .constraints = snbep_uncore_r2pcie_constraints, + SNBEP_UNCORE_PCI_COMMON_INIT(), +}; + +static struct intel_uncore_type snbep_uncore_r3qpi = { + .name = "r3qpi", + .num_counters = 3, + .num_boxes = 2, + .perf_ctr_bits = 44, + .constraints = snbep_uncore_r3qpi_constraints, + SNBEP_UNCORE_PCI_COMMON_INIT(), +}; + +enum { + SNBEP_PCI_UNCORE_HA, + SNBEP_PCI_UNCORE_IMC, + SNBEP_PCI_UNCORE_QPI, + SNBEP_PCI_UNCORE_R2PCIE, + SNBEP_PCI_UNCORE_R3QPI, +}; + +static struct intel_uncore_type *snbep_pci_uncores[] = { + [SNBEP_PCI_UNCORE_HA] = &snbep_uncore_ha, + [SNBEP_PCI_UNCORE_IMC] = &snbep_uncore_imc, + [SNBEP_PCI_UNCORE_QPI] = &snbep_uncore_qpi, + [SNBEP_PCI_UNCORE_R2PCIE] = &snbep_uncore_r2pcie, + [SNBEP_PCI_UNCORE_R3QPI] = &snbep_uncore_r3qpi, + NULL, +}; + +static DEFINE_PCI_DEVICE_TABLE(snbep_uncore_pci_ids) = { + { /* Home Agent */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_HA), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_HA, 0), + }, + { /* MC Channel 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_IMC0), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_IMC, 0), + }, + { /* MC Channel 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_IMC1), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_IMC, 1), + }, + { /* MC Channel 2 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_IMC2), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_IMC, 2), + }, + { /* MC Channel 3 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_IMC3), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_IMC, 3), + }, + { /* QPI Port 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_QPI0), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_QPI, 0), + }, + { /* QPI Port 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_QPI1), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_QPI, 1), + }, + { /* R2PCIe */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_R2PCIE), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_R2PCIE, 0), + }, + { /* R3QPI Link 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_R3QPI0), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_R3QPI, 0), + }, + { /* R3QPI Link 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_UNC_R3QPI1), + .driver_data = UNCORE_PCI_DEV_DATA(SNBEP_PCI_UNCORE_R3QPI, 1), + }, + { /* QPI Port 0 filter */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x3c86), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, + SNBEP_PCI_QPI_PORT0_FILTER), + }, + { /* QPI Port 0 filter */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x3c96), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, + SNBEP_PCI_QPI_PORT1_FILTER), + }, + { /* end: all zeroes */ } +}; + +static struct pci_driver snbep_uncore_pci_driver = { + .name = "snbep_uncore", + .id_table = snbep_uncore_pci_ids, +}; + +/* + * build pci bus to socket mapping + */ +static int snbep_pci2phy_map_init(int devid) +{ + struct pci_dev *ubox_dev = NULL; + int i, bus, nodeid; + int err = 0; + u32 config = 0; + + while (1) { + /* find the UBOX device */ + ubox_dev = pci_get_device(PCI_VENDOR_ID_INTEL, devid, ubox_dev); + if (!ubox_dev) + break; + bus = ubox_dev->bus->number; + /* get the Node ID of the local register */ + err = pci_read_config_dword(ubox_dev, 0x40, &config); + if (err) + break; + nodeid = config; + /* get the Node ID mapping */ + err = pci_read_config_dword(ubox_dev, 0x54, &config); + if (err) + break; + /* + * every three bits in the Node ID mapping register maps + * to a particular node. + */ + for (i = 0; i < 8; i++) { + if (nodeid == ((config >> (3 * i)) & 0x7)) { + pcibus_to_physid[bus] = i; + break; + } + } + } + + if (!err) { + /* + * For PCI bus with no UBOX device, find the next bus + * that has UBOX device and use its mapping. + */ + i = -1; + for (bus = 255; bus >= 0; bus--) { + if (pcibus_to_physid[bus] >= 0) + i = pcibus_to_physid[bus]; + else + pcibus_to_physid[bus] = i; + } + } + + if (ubox_dev) + pci_dev_put(ubox_dev); + + return err ? pcibios_err_to_errno(err) : 0; +} +/* end of Sandy Bridge-EP uncore support */ + +/* IvyTown uncore support */ +static void ivt_uncore_msr_init_box(struct intel_uncore_box *box) +{ + unsigned msr = uncore_msr_box_ctl(box); + if (msr) + wrmsrl(msr, IVT_PMON_BOX_CTL_INT); +} + +static void ivt_uncore_pci_init_box(struct intel_uncore_box *box) +{ + struct pci_dev *pdev = box->pci_dev; + + pci_write_config_dword(pdev, SNBEP_PCI_PMON_BOX_CTL, IVT_PMON_BOX_CTL_INT); +} + +#define IVT_UNCORE_MSR_OPS_COMMON_INIT() \ + .init_box = ivt_uncore_msr_init_box, \ + .disable_box = snbep_uncore_msr_disable_box, \ + .enable_box = snbep_uncore_msr_enable_box, \ + .disable_event = snbep_uncore_msr_disable_event, \ + .enable_event = snbep_uncore_msr_enable_event, \ + .read_counter = uncore_msr_read_counter + +static struct intel_uncore_ops ivt_uncore_msr_ops = { + IVT_UNCORE_MSR_OPS_COMMON_INIT(), +}; + +static struct intel_uncore_ops ivt_uncore_pci_ops = { + .init_box = ivt_uncore_pci_init_box, + .disable_box = snbep_uncore_pci_disable_box, + .enable_box = snbep_uncore_pci_enable_box, + .disable_event = snbep_uncore_pci_disable_event, + .enable_event = snbep_uncore_pci_enable_event, + .read_counter = snbep_uncore_pci_read_counter, +}; + +#define IVT_UNCORE_PCI_COMMON_INIT() \ + .perf_ctr = SNBEP_PCI_PMON_CTR0, \ + .event_ctl = SNBEP_PCI_PMON_CTL0, \ + .event_mask = IVT_PMON_RAW_EVENT_MASK, \ + .box_ctl = SNBEP_PCI_PMON_BOX_CTL, \ + .ops = &ivt_uncore_pci_ops, \ + .format_group = &ivt_uncore_format_group + +static struct attribute *ivt_uncore_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh8.attr, + NULL, +}; + +static struct attribute *ivt_uncore_ubox_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh5.attr, + NULL, +}; + +static struct attribute *ivt_uncore_cbox_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_tid_en.attr, + &format_attr_thresh8.attr, + &format_attr_filter_tid.attr, + &format_attr_filter_link.attr, + &format_attr_filter_state2.attr, + &format_attr_filter_nid2.attr, + &format_attr_filter_opc2.attr, + NULL, +}; + +static struct attribute *ivt_uncore_pcu_formats_attr[] = { + &format_attr_event_ext.attr, + &format_attr_occ_sel.attr, + &format_attr_edge.attr, + &format_attr_thresh5.attr, + &format_attr_occ_invert.attr, + &format_attr_occ_edge.attr, + &format_attr_filter_band0.attr, + &format_attr_filter_band1.attr, + &format_attr_filter_band2.attr, + &format_attr_filter_band3.attr, + NULL, +}; + +static struct attribute *ivt_uncore_qpi_formats_attr[] = { + &format_attr_event_ext.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_thresh8.attr, + &format_attr_match_rds.attr, + &format_attr_match_rnid30.attr, + &format_attr_match_rnid4.attr, + &format_attr_match_dnid.attr, + &format_attr_match_mc.attr, + &format_attr_match_opc.attr, + &format_attr_match_vnw.attr, + &format_attr_match0.attr, + &format_attr_match1.attr, + &format_attr_mask_rds.attr, + &format_attr_mask_rnid30.attr, + &format_attr_mask_rnid4.attr, + &format_attr_mask_dnid.attr, + &format_attr_mask_mc.attr, + &format_attr_mask_opc.attr, + &format_attr_mask_vnw.attr, + &format_attr_mask0.attr, + &format_attr_mask1.attr, + NULL, +}; + +static struct attribute_group ivt_uncore_format_group = { + .name = "format", + .attrs = ivt_uncore_formats_attr, +}; + +static struct attribute_group ivt_uncore_ubox_format_group = { + .name = "format", + .attrs = ivt_uncore_ubox_formats_attr, +}; + +static struct attribute_group ivt_uncore_cbox_format_group = { + .name = "format", + .attrs = ivt_uncore_cbox_formats_attr, +}; + +static struct attribute_group ivt_uncore_pcu_format_group = { + .name = "format", + .attrs = ivt_uncore_pcu_formats_attr, +}; + +static struct attribute_group ivt_uncore_qpi_format_group = { + .name = "format", + .attrs = ivt_uncore_qpi_formats_attr, +}; + +static struct intel_uncore_type ivt_uncore_ubox = { + .name = "ubox", + .num_counters = 2, + .num_boxes = 1, + .perf_ctr_bits = 44, + .fixed_ctr_bits = 48, + .perf_ctr = SNBEP_U_MSR_PMON_CTR0, + .event_ctl = SNBEP_U_MSR_PMON_CTL0, + .event_mask = IVT_U_MSR_PMON_RAW_EVENT_MASK, + .fixed_ctr = SNBEP_U_MSR_PMON_UCLK_FIXED_CTR, + .fixed_ctl = SNBEP_U_MSR_PMON_UCLK_FIXED_CTL, + .ops = &ivt_uncore_msr_ops, + .format_group = &ivt_uncore_ubox_format_group, +}; + +static struct extra_reg ivt_uncore_cbox_extra_regs[] = { + SNBEP_CBO_EVENT_EXTRA_REG(SNBEP_CBO_PMON_CTL_TID_EN, + SNBEP_CBO_PMON_CTL_TID_EN, 0x1), + SNBEP_CBO_EVENT_EXTRA_REG(0x1031, 0x10ff, 0x2), + + SNBEP_CBO_EVENT_EXTRA_REG(0x1134, 0xffff, 0x4), + SNBEP_CBO_EVENT_EXTRA_REG(0x4134, 0xffff, 0xc), + SNBEP_CBO_EVENT_EXTRA_REG(0x5134, 0xffff, 0xc), + SNBEP_CBO_EVENT_EXTRA_REG(0x0334, 0xffff, 0x4), + SNBEP_CBO_EVENT_EXTRA_REG(0x4334, 0xffff, 0xc), + SNBEP_CBO_EVENT_EXTRA_REG(0x0534, 0xffff, 0x4), + SNBEP_CBO_EVENT_EXTRA_REG(0x4534, 0xffff, 0xc), + SNBEP_CBO_EVENT_EXTRA_REG(0x0934, 0xffff, 0x4), + SNBEP_CBO_EVENT_EXTRA_REG(0x4934, 0xffff, 0xc), + SNBEP_CBO_EVENT_EXTRA_REG(0x0135, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x0335, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x2135, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x2335, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x4135, 0xffff, 0x18), + SNBEP_CBO_EVENT_EXTRA_REG(0x4335, 0xffff, 0x18), + SNBEP_CBO_EVENT_EXTRA_REG(0x4435, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x4835, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x4a35, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x5035, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x8135, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x8335, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x0136, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x0336, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x2136, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x2336, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x4136, 0xffff, 0x18), + SNBEP_CBO_EVENT_EXTRA_REG(0x4336, 0xffff, 0x18), + SNBEP_CBO_EVENT_EXTRA_REG(0x4436, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x4836, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x4a36, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x5036, 0xffff, 0x8), + SNBEP_CBO_EVENT_EXTRA_REG(0x8136, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x8336, 0xffff, 0x10), + SNBEP_CBO_EVENT_EXTRA_REG(0x4037, 0x40ff, 0x8), + EVENT_EXTRA_END +}; + +static u64 ivt_cbox_filter_mask(int fields) +{ + u64 mask = 0; + + if (fields & 0x1) + mask |= IVT_CB0_MSR_PMON_BOX_FILTER_TID; + if (fields & 0x2) + mask |= IVT_CB0_MSR_PMON_BOX_FILTER_LINK; + if (fields & 0x4) + mask |= IVT_CB0_MSR_PMON_BOX_FILTER_STATE; + if (fields & 0x8) + mask |= IVT_CB0_MSR_PMON_BOX_FILTER_NID; + if (fields & 0x10) + mask |= IVT_CB0_MSR_PMON_BOX_FILTER_OPC; + + return mask; +} + +static struct event_constraint * +ivt_cbox_get_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + return __snbep_cbox_get_constraint(box, event, ivt_cbox_filter_mask); +} + +static int ivt_cbox_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct extra_reg *er; + int idx = 0; + + for (er = ivt_uncore_cbox_extra_regs; er->msr; er++) { + if (er->event != (event->hw.config & er->config_mask)) + continue; + idx |= er->idx; + } + + if (idx) { + reg1->reg = SNBEP_C0_MSR_PMON_BOX_FILTER + + SNBEP_CBO_MSR_OFFSET * box->pmu->pmu_idx; + reg1->config = event->attr.config1 & ivt_cbox_filter_mask(idx); + reg1->idx = idx; + } + return 0; +} + +static void ivt_cbox_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + + if (reg1->idx != EXTRA_REG_NONE) { + u64 filter = uncore_shared_reg_config(box, 0); + wrmsrl(reg1->reg, filter & 0xffffffff); + wrmsrl(reg1->reg + 6, filter >> 32); + } + + wrmsrl(hwc->config_base, hwc->config | SNBEP_PMON_CTL_EN); +} + +static struct intel_uncore_ops ivt_uncore_cbox_ops = { + .init_box = ivt_uncore_msr_init_box, + .disable_box = snbep_uncore_msr_disable_box, + .enable_box = snbep_uncore_msr_enable_box, + .disable_event = snbep_uncore_msr_disable_event, + .enable_event = ivt_cbox_enable_event, + .read_counter = uncore_msr_read_counter, + .hw_config = ivt_cbox_hw_config, + .get_constraint = ivt_cbox_get_constraint, + .put_constraint = snbep_cbox_put_constraint, +}; + +static struct intel_uncore_type ivt_uncore_cbox = { + .name = "cbox", + .num_counters = 4, + .num_boxes = 15, + .perf_ctr_bits = 44, + .event_ctl = SNBEP_C0_MSR_PMON_CTL0, + .perf_ctr = SNBEP_C0_MSR_PMON_CTR0, + .event_mask = IVT_CBO_MSR_PMON_RAW_EVENT_MASK, + .box_ctl = SNBEP_C0_MSR_PMON_BOX_CTL, + .msr_offset = SNBEP_CBO_MSR_OFFSET, + .num_shared_regs = 1, + .constraints = snbep_uncore_cbox_constraints, + .ops = &ivt_uncore_cbox_ops, + .format_group = &ivt_uncore_cbox_format_group, +}; + +static struct intel_uncore_ops ivt_uncore_pcu_ops = { + IVT_UNCORE_MSR_OPS_COMMON_INIT(), + .hw_config = snbep_pcu_hw_config, + .get_constraint = snbep_pcu_get_constraint, + .put_constraint = snbep_pcu_put_constraint, +}; + +static struct intel_uncore_type ivt_uncore_pcu = { + .name = "pcu", + .num_counters = 4, + .num_boxes = 1, + .perf_ctr_bits = 48, + .perf_ctr = SNBEP_PCU_MSR_PMON_CTR0, + .event_ctl = SNBEP_PCU_MSR_PMON_CTL0, + .event_mask = IVT_PCU_MSR_PMON_RAW_EVENT_MASK, + .box_ctl = SNBEP_PCU_MSR_PMON_BOX_CTL, + .num_shared_regs = 1, + .ops = &ivt_uncore_pcu_ops, + .format_group = &ivt_uncore_pcu_format_group, +}; + +static struct intel_uncore_type *ivt_msr_uncores[] = { + &ivt_uncore_ubox, + &ivt_uncore_cbox, + &ivt_uncore_pcu, + NULL, +}; + +static struct intel_uncore_type ivt_uncore_ha = { + .name = "ha", + .num_counters = 4, + .num_boxes = 2, + .perf_ctr_bits = 48, + IVT_UNCORE_PCI_COMMON_INIT(), +}; + +static struct intel_uncore_type ivt_uncore_imc = { + .name = "imc", + .num_counters = 4, + .num_boxes = 8, + .perf_ctr_bits = 48, + .fixed_ctr_bits = 48, + .fixed_ctr = SNBEP_MC_CHy_PCI_PMON_FIXED_CTR, + .fixed_ctl = SNBEP_MC_CHy_PCI_PMON_FIXED_CTL, + IVT_UNCORE_PCI_COMMON_INIT(), +}; + +/* registers in IRP boxes are not properly aligned */ +static unsigned ivt_uncore_irp_ctls[] = {0xd8, 0xdc, 0xe0, 0xe4}; +static unsigned ivt_uncore_irp_ctrs[] = {0xa0, 0xb0, 0xb8, 0xc0}; + +static void ivt_uncore_irp_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct pci_dev *pdev = box->pci_dev; + struct hw_perf_event *hwc = &event->hw; + + pci_write_config_dword(pdev, ivt_uncore_irp_ctls[hwc->idx], + hwc->config | SNBEP_PMON_CTL_EN); +} + +static void ivt_uncore_irp_disable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct pci_dev *pdev = box->pci_dev; + struct hw_perf_event *hwc = &event->hw; + + pci_write_config_dword(pdev, ivt_uncore_irp_ctls[hwc->idx], hwc->config); +} + +static u64 ivt_uncore_irp_read_counter(struct intel_uncore_box *box, struct perf_event *event) +{ + struct pci_dev *pdev = box->pci_dev; + struct hw_perf_event *hwc = &event->hw; + u64 count = 0; + + pci_read_config_dword(pdev, ivt_uncore_irp_ctrs[hwc->idx], (u32 *)&count); + pci_read_config_dword(pdev, ivt_uncore_irp_ctrs[hwc->idx] + 4, (u32 *)&count + 1); + + return count; +} + +static struct intel_uncore_ops ivt_uncore_irp_ops = { + .init_box = ivt_uncore_pci_init_box, + .disable_box = snbep_uncore_pci_disable_box, + .enable_box = snbep_uncore_pci_enable_box, + .disable_event = ivt_uncore_irp_disable_event, + .enable_event = ivt_uncore_irp_enable_event, + .read_counter = ivt_uncore_irp_read_counter, +}; + +static struct intel_uncore_type ivt_uncore_irp = { + .name = "irp", + .num_counters = 4, + .num_boxes = 1, + .perf_ctr_bits = 48, + .event_mask = IVT_PMON_RAW_EVENT_MASK, + .box_ctl = SNBEP_PCI_PMON_BOX_CTL, + .ops = &ivt_uncore_irp_ops, + .format_group = &ivt_uncore_format_group, +}; + +static struct intel_uncore_ops ivt_uncore_qpi_ops = { + .init_box = ivt_uncore_pci_init_box, + .disable_box = snbep_uncore_pci_disable_box, + .enable_box = snbep_uncore_pci_enable_box, + .disable_event = snbep_uncore_pci_disable_event, + .enable_event = snbep_qpi_enable_event, + .read_counter = snbep_uncore_pci_read_counter, + .hw_config = snbep_qpi_hw_config, + .get_constraint = uncore_get_constraint, + .put_constraint = uncore_put_constraint, +}; + +static struct intel_uncore_type ivt_uncore_qpi = { + .name = "qpi", + .num_counters = 4, + .num_boxes = 3, + .perf_ctr_bits = 48, + .perf_ctr = SNBEP_PCI_PMON_CTR0, + .event_ctl = SNBEP_PCI_PMON_CTL0, + .event_mask = IVT_QPI_PCI_PMON_RAW_EVENT_MASK, + .box_ctl = SNBEP_PCI_PMON_BOX_CTL, + .num_shared_regs = 1, + .ops = &ivt_uncore_qpi_ops, + .format_group = &ivt_uncore_qpi_format_group, +}; + +static struct intel_uncore_type ivt_uncore_r2pcie = { + .name = "r2pcie", + .num_counters = 4, + .num_boxes = 1, + .perf_ctr_bits = 44, + .constraints = snbep_uncore_r2pcie_constraints, + IVT_UNCORE_PCI_COMMON_INIT(), +}; + +static struct intel_uncore_type ivt_uncore_r3qpi = { + .name = "r3qpi", + .num_counters = 3, + .num_boxes = 2, + .perf_ctr_bits = 44, + .constraints = snbep_uncore_r3qpi_constraints, + IVT_UNCORE_PCI_COMMON_INIT(), +}; + +enum { + IVT_PCI_UNCORE_HA, + IVT_PCI_UNCORE_IMC, + IVT_PCI_UNCORE_IRP, + IVT_PCI_UNCORE_QPI, + IVT_PCI_UNCORE_R2PCIE, + IVT_PCI_UNCORE_R3QPI, +}; + +static struct intel_uncore_type *ivt_pci_uncores[] = { + [IVT_PCI_UNCORE_HA] = &ivt_uncore_ha, + [IVT_PCI_UNCORE_IMC] = &ivt_uncore_imc, + [IVT_PCI_UNCORE_IRP] = &ivt_uncore_irp, + [IVT_PCI_UNCORE_QPI] = &ivt_uncore_qpi, + [IVT_PCI_UNCORE_R2PCIE] = &ivt_uncore_r2pcie, + [IVT_PCI_UNCORE_R3QPI] = &ivt_uncore_r3qpi, + NULL, +}; + +static DEFINE_PCI_DEVICE_TABLE(ivt_uncore_pci_ids) = { + { /* Home Agent 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe30), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_HA, 0), + }, + { /* Home Agent 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe38), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_HA, 1), + }, + { /* MC0 Channel 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xeb4), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 0), + }, + { /* MC0 Channel 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xeb5), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 1), + }, + { /* MC0 Channel 3 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xeb0), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 2), + }, + { /* MC0 Channel 4 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xeb1), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 3), + }, + { /* MC1 Channel 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xef4), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 4), + }, + { /* MC1 Channel 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xef5), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 5), + }, + { /* MC1 Channel 3 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xef0), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 6), + }, + { /* MC1 Channel 4 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xef1), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IMC, 7), + }, + { /* IRP */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe39), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_IRP, 0), + }, + { /* QPI0 Port 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe32), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_QPI, 0), + }, + { /* QPI0 Port 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe33), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_QPI, 1), + }, + { /* QPI1 Port 2 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe3a), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_QPI, 2), + }, + { /* R2PCIe */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe34), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_R2PCIE, 0), + }, + { /* R3QPI0 Link 0 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe36), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_R3QPI, 0), + }, + { /* R3QPI0 Link 1 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe37), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_R3QPI, 1), + }, + { /* R3QPI1 Link 2 */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe3e), + .driver_data = UNCORE_PCI_DEV_DATA(IVT_PCI_UNCORE_R3QPI, 2), + }, + { /* QPI Port 0 filter */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe86), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, + SNBEP_PCI_QPI_PORT0_FILTER), + }, + { /* QPI Port 0 filter */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0xe96), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, + SNBEP_PCI_QPI_PORT1_FILTER), + }, + { /* end: all zeroes */ } +}; + +static struct pci_driver ivt_uncore_pci_driver = { + .name = "ivt_uncore", + .id_table = ivt_uncore_pci_ids, +}; +/* end of IvyTown uncore support */ + +/* Sandy Bridge uncore support */ +static void snb_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + + if (hwc->idx < UNCORE_PMC_IDX_FIXED) + wrmsrl(hwc->config_base, hwc->config | SNB_UNC_CTL_EN); + else + wrmsrl(hwc->config_base, SNB_UNC_CTL_EN); +} + +static void snb_uncore_msr_disable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + wrmsrl(event->hw.config_base, 0); +} + +static void snb_uncore_msr_init_box(struct intel_uncore_box *box) +{ + if (box->pmu->pmu_idx == 0) { + wrmsrl(SNB_UNC_PERF_GLOBAL_CTL, + SNB_UNC_GLOBAL_CTL_EN | SNB_UNC_GLOBAL_CTL_CORE_ALL); + } +} + +static struct uncore_event_desc snb_uncore_events[] = { + INTEL_UNCORE_EVENT_DESC(clockticks, "event=0xff,umask=0x00"), + { /* end: all zeroes */ }, +}; + +static struct attribute *snb_uncore_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_cmask5.attr, + NULL, +}; + +static struct attribute_group snb_uncore_format_group = { + .name = "format", + .attrs = snb_uncore_formats_attr, +}; + +static struct intel_uncore_ops snb_uncore_msr_ops = { + .init_box = snb_uncore_msr_init_box, + .disable_event = snb_uncore_msr_disable_event, + .enable_event = snb_uncore_msr_enable_event, + .read_counter = uncore_msr_read_counter, +}; + +static struct event_constraint snb_uncore_cbox_constraints[] = { + UNCORE_EVENT_CONSTRAINT(0x80, 0x1), + UNCORE_EVENT_CONSTRAINT(0x83, 0x1), + EVENT_CONSTRAINT_END +}; + +static struct intel_uncore_type snb_uncore_cbox = { + .name = "cbox", + .num_counters = 2, + .num_boxes = 4, + .perf_ctr_bits = 44, + .fixed_ctr_bits = 48, + .perf_ctr = SNB_UNC_CBO_0_PER_CTR0, + .event_ctl = SNB_UNC_CBO_0_PERFEVTSEL0, + .fixed_ctr = SNB_UNC_FIXED_CTR, + .fixed_ctl = SNB_UNC_FIXED_CTR_CTRL, + .single_fixed = 1, + .event_mask = SNB_UNC_RAW_EVENT_MASK, + .msr_offset = SNB_UNC_CBO_MSR_OFFSET, + .constraints = snb_uncore_cbox_constraints, + .ops = &snb_uncore_msr_ops, + .format_group = &snb_uncore_format_group, + .event_descs = snb_uncore_events, +}; + +static struct intel_uncore_type *snb_msr_uncores[] = { + &snb_uncore_cbox, + NULL, +}; + +enum { + SNB_PCI_UNCORE_IMC, +}; + +static struct uncore_event_desc snb_uncore_imc_events[] = { + INTEL_UNCORE_EVENT_DESC(data_reads, "event=0x01"), + INTEL_UNCORE_EVENT_DESC(data_reads.scale, "6.103515625e-5"), + INTEL_UNCORE_EVENT_DESC(data_reads.unit, "MiB"), + + INTEL_UNCORE_EVENT_DESC(data_writes, "event=0x02"), + INTEL_UNCORE_EVENT_DESC(data_writes.scale, "6.103515625e-5"), + INTEL_UNCORE_EVENT_DESC(data_writes.unit, "MiB"), + + { /* end: all zeroes */ }, +}; + +#define SNB_UNCORE_PCI_IMC_EVENT_MASK 0xff +#define SNB_UNCORE_PCI_IMC_BAR_OFFSET 0x48 + +/* page size multiple covering all config regs */ +#define SNB_UNCORE_PCI_IMC_MAP_SIZE 0x6000 + +#define SNB_UNCORE_PCI_IMC_DATA_READS 0x1 +#define SNB_UNCORE_PCI_IMC_DATA_READS_BASE 0x5050 +#define SNB_UNCORE_PCI_IMC_DATA_WRITES 0x2 +#define SNB_UNCORE_PCI_IMC_DATA_WRITES_BASE 0x5054 +#define SNB_UNCORE_PCI_IMC_CTR_BASE SNB_UNCORE_PCI_IMC_DATA_READS_BASE + +static struct attribute *snb_uncore_imc_formats_attr[] = { + &format_attr_event.attr, + NULL, +}; + +static struct attribute_group snb_uncore_imc_format_group = { + .name = "format", + .attrs = snb_uncore_imc_formats_attr, +}; + +static void snb_uncore_imc_init_box(struct intel_uncore_box *box) +{ + struct pci_dev *pdev = box->pci_dev; + int where = SNB_UNCORE_PCI_IMC_BAR_OFFSET; + resource_size_t addr; + u32 pci_dword; + + pci_read_config_dword(pdev, where, &pci_dword); + addr = pci_dword; + +#ifdef CONFIG_PHYS_ADDR_T_64BIT + pci_read_config_dword(pdev, where + 4, &pci_dword); + addr |= ((resource_size_t)pci_dword << 32); +#endif + + addr &= ~(PAGE_SIZE - 1); + + box->io_addr = ioremap(addr, SNB_UNCORE_PCI_IMC_MAP_SIZE); + box->hrtimer_duration = UNCORE_SNB_IMC_HRTIMER_INTERVAL; +} + +static void snb_uncore_imc_enable_box(struct intel_uncore_box *box) +{} + +static void snb_uncore_imc_disable_box(struct intel_uncore_box *box) +{} + +static void snb_uncore_imc_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{} + +static void snb_uncore_imc_disable_event(struct intel_uncore_box *box, struct perf_event *event) +{} + +static u64 snb_uncore_imc_read_counter(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + + return (u64)*(unsigned int *)(box->io_addr + hwc->event_base); +} + +/* + * custom event_init() function because we define our own fixed, free + * running counters, so we do not want to conflict with generic uncore + * logic. Also simplifies processing + */ +static int snb_uncore_imc_event_init(struct perf_event *event) +{ + struct intel_uncore_pmu *pmu; + struct intel_uncore_box *box; + struct hw_perf_event *hwc = &event->hw; + u64 cfg = event->attr.config & SNB_UNCORE_PCI_IMC_EVENT_MASK; + int idx, base; + + if (event->attr.type != event->pmu->type) + return -ENOENT; + + pmu = uncore_event_to_pmu(event); + /* no device found for this pmu */ + if (pmu->func_id < 0) + return -ENOENT; + + /* Sampling not supported yet */ + if (hwc->sample_period) + return -EINVAL; + + /* unsupported modes and filters */ + if (event->attr.exclude_user || + event->attr.exclude_kernel || + event->attr.exclude_hv || + event->attr.exclude_idle || + event->attr.exclude_host || + event->attr.exclude_guest || + event->attr.sample_period) /* no sampling */ + return -EINVAL; + + /* + * Place all uncore events for a particular physical package + * onto a single cpu + */ + if (event->cpu < 0) + return -EINVAL; + + /* check only supported bits are set */ + if (event->attr.config & ~SNB_UNCORE_PCI_IMC_EVENT_MASK) + return -EINVAL; + + box = uncore_pmu_to_box(pmu, event->cpu); + if (!box || box->cpu < 0) + return -EINVAL; + + event->cpu = box->cpu; + + event->hw.idx = -1; + event->hw.last_tag = ~0ULL; + event->hw.extra_reg.idx = EXTRA_REG_NONE; + event->hw.branch_reg.idx = EXTRA_REG_NONE; + /* + * check event is known (whitelist, determines counter) + */ + switch (cfg) { + case SNB_UNCORE_PCI_IMC_DATA_READS: + base = SNB_UNCORE_PCI_IMC_DATA_READS_BASE; + idx = UNCORE_PMC_IDX_FIXED; + break; + case SNB_UNCORE_PCI_IMC_DATA_WRITES: + base = SNB_UNCORE_PCI_IMC_DATA_WRITES_BASE; + idx = UNCORE_PMC_IDX_FIXED + 1; + break; + default: + return -EINVAL; + } + + /* must be done before validate_group */ + event->hw.event_base = base; + event->hw.config = cfg; + event->hw.idx = idx; + + /* no group validation needed, we have free running counters */ + + return 0; +} + +static int snb_uncore_imc_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + return 0; +} + +static void snb_uncore_imc_event_start(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + u64 count; + + if (WARN_ON_ONCE(!(event->hw.state & PERF_HES_STOPPED))) + return; + + event->hw.state = 0; + box->n_active++; + + list_add_tail(&event->active_entry, &box->active_list); + + count = snb_uncore_imc_read_counter(box, event); + local64_set(&event->hw.prev_count, count); + + if (box->n_active == 1) + uncore_pmu_start_hrtimer(box); +} + +static void snb_uncore_imc_event_stop(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + struct hw_perf_event *hwc = &event->hw; + + if (!(hwc->state & PERF_HES_STOPPED)) { + box->n_active--; + + WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED); + hwc->state |= PERF_HES_STOPPED; + + list_del(&event->active_entry); + + if (box->n_active == 0) + uncore_pmu_cancel_hrtimer(box); + } + + if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) { + /* + * Drain the remaining delta count out of a event + * that we are disabling: + */ + uncore_perf_event_update(box, event); + hwc->state |= PERF_HES_UPTODATE; + } +} + +static int snb_uncore_imc_event_add(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + struct hw_perf_event *hwc = &event->hw; + + if (!box) + return -ENODEV; + + hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + if (!(flags & PERF_EF_START)) + hwc->state |= PERF_HES_ARCH; + + snb_uncore_imc_event_start(event, 0); + + box->n_events++; + + return 0; +} + +static void snb_uncore_imc_event_del(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + int i; + + snb_uncore_imc_event_stop(event, PERF_EF_UPDATE); + + for (i = 0; i < box->n_events; i++) { + if (event == box->event_list[i]) { + --box->n_events; + break; + } + } +} + +static int snb_pci2phy_map_init(int devid) +{ + struct pci_dev *dev = NULL; + int bus; + + dev = pci_get_device(PCI_VENDOR_ID_INTEL, devid, dev); + if (!dev) + return -ENOTTY; + + bus = dev->bus->number; + + pcibus_to_physid[bus] = 0; + + pci_dev_put(dev); + + return 0; +} + +static struct pmu snb_uncore_imc_pmu = { + .task_ctx_nr = perf_invalid_context, + .event_init = snb_uncore_imc_event_init, + .add = snb_uncore_imc_event_add, + .del = snb_uncore_imc_event_del, + .start = snb_uncore_imc_event_start, + .stop = snb_uncore_imc_event_stop, + .read = uncore_pmu_event_read, +}; + +static struct intel_uncore_ops snb_uncore_imc_ops = { + .init_box = snb_uncore_imc_init_box, + .enable_box = snb_uncore_imc_enable_box, + .disable_box = snb_uncore_imc_disable_box, + .disable_event = snb_uncore_imc_disable_event, + .enable_event = snb_uncore_imc_enable_event, + .hw_config = snb_uncore_imc_hw_config, + .read_counter = snb_uncore_imc_read_counter, +}; + +static struct intel_uncore_type snb_uncore_imc = { + .name = "imc", + .num_counters = 2, + .num_boxes = 1, + .fixed_ctr_bits = 32, + .fixed_ctr = SNB_UNCORE_PCI_IMC_CTR_BASE, + .event_descs = snb_uncore_imc_events, + .format_group = &snb_uncore_imc_format_group, + .perf_ctr = SNB_UNCORE_PCI_IMC_DATA_READS_BASE, + .event_mask = SNB_UNCORE_PCI_IMC_EVENT_MASK, + .ops = &snb_uncore_imc_ops, + .pmu = &snb_uncore_imc_pmu, +}; + +static struct intel_uncore_type *snb_pci_uncores[] = { + [SNB_PCI_UNCORE_IMC] = &snb_uncore_imc, + NULL, +}; + +static DEFINE_PCI_DEVICE_TABLE(snb_uncore_pci_ids) = { + { /* IMC */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_SNB_IMC), + .driver_data = UNCORE_PCI_DEV_DATA(SNB_PCI_UNCORE_IMC, 0), + }, + { /* end: all zeroes */ }, +}; + +static DEFINE_PCI_DEVICE_TABLE(ivb_uncore_pci_ids) = { + { /* IMC */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IVB_IMC), + .driver_data = UNCORE_PCI_DEV_DATA(SNB_PCI_UNCORE_IMC, 0), + }, + { /* end: all zeroes */ }, +}; + +static DEFINE_PCI_DEVICE_TABLE(hsw_uncore_pci_ids) = { + { /* IMC */ + PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_HSW_IMC), + .driver_data = UNCORE_PCI_DEV_DATA(SNB_PCI_UNCORE_IMC, 0), + }, + { /* end: all zeroes */ }, +}; + +static struct pci_driver snb_uncore_pci_driver = { + .name = "snb_uncore", + .id_table = snb_uncore_pci_ids, +}; + +static struct pci_driver ivb_uncore_pci_driver = { + .name = "ivb_uncore", + .id_table = ivb_uncore_pci_ids, +}; + +static struct pci_driver hsw_uncore_pci_driver = { + .name = "hsw_uncore", + .id_table = hsw_uncore_pci_ids, +}; + +/* end of Sandy Bridge uncore support */ + +/* Nehalem uncore support */ +static void nhm_uncore_msr_disable_box(struct intel_uncore_box *box) +{ + wrmsrl(NHM_UNC_PERF_GLOBAL_CTL, 0); +} + +static void nhm_uncore_msr_enable_box(struct intel_uncore_box *box) +{ + wrmsrl(NHM_UNC_PERF_GLOBAL_CTL, NHM_UNC_GLOBAL_CTL_EN_PC_ALL | NHM_UNC_GLOBAL_CTL_EN_FC); +} + +static void nhm_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + + if (hwc->idx < UNCORE_PMC_IDX_FIXED) + wrmsrl(hwc->config_base, hwc->config | SNB_UNC_CTL_EN); + else + wrmsrl(hwc->config_base, NHM_UNC_FIXED_CTR_CTL_EN); +} + +static struct attribute *nhm_uncore_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_cmask8.attr, + NULL, +}; + +static struct attribute_group nhm_uncore_format_group = { + .name = "format", + .attrs = nhm_uncore_formats_attr, +}; + +static struct uncore_event_desc nhm_uncore_events[] = { + INTEL_UNCORE_EVENT_DESC(clockticks, "event=0xff,umask=0x00"), + INTEL_UNCORE_EVENT_DESC(qmc_writes_full_any, "event=0x2f,umask=0x0f"), + INTEL_UNCORE_EVENT_DESC(qmc_normal_reads_any, "event=0x2c,umask=0x0f"), + INTEL_UNCORE_EVENT_DESC(qhl_request_ioh_reads, "event=0x20,umask=0x01"), + INTEL_UNCORE_EVENT_DESC(qhl_request_ioh_writes, "event=0x20,umask=0x02"), + INTEL_UNCORE_EVENT_DESC(qhl_request_remote_reads, "event=0x20,umask=0x04"), + INTEL_UNCORE_EVENT_DESC(qhl_request_remote_writes, "event=0x20,umask=0x08"), + INTEL_UNCORE_EVENT_DESC(qhl_request_local_reads, "event=0x20,umask=0x10"), + INTEL_UNCORE_EVENT_DESC(qhl_request_local_writes, "event=0x20,umask=0x20"), + { /* end: all zeroes */ }, +}; + +static struct intel_uncore_ops nhm_uncore_msr_ops = { + .disable_box = nhm_uncore_msr_disable_box, + .enable_box = nhm_uncore_msr_enable_box, + .disable_event = snb_uncore_msr_disable_event, + .enable_event = nhm_uncore_msr_enable_event, + .read_counter = uncore_msr_read_counter, +}; + +static struct intel_uncore_type nhm_uncore = { + .name = "", + .num_counters = 8, + .num_boxes = 1, + .perf_ctr_bits = 48, + .fixed_ctr_bits = 48, + .event_ctl = NHM_UNC_PERFEVTSEL0, + .perf_ctr = NHM_UNC_UNCORE_PMC0, + .fixed_ctr = NHM_UNC_FIXED_CTR, + .fixed_ctl = NHM_UNC_FIXED_CTR_CTRL, + .event_mask = NHM_UNC_RAW_EVENT_MASK, + .event_descs = nhm_uncore_events, + .ops = &nhm_uncore_msr_ops, + .format_group = &nhm_uncore_format_group, +}; + +static struct intel_uncore_type *nhm_msr_uncores[] = { + &nhm_uncore, + NULL, +}; +/* end of Nehalem uncore support */ + +/* Nehalem-EX uncore support */ +DEFINE_UNCORE_FORMAT_ATTR(event5, event, "config:1-5"); +DEFINE_UNCORE_FORMAT_ATTR(counter, counter, "config:6-7"); +DEFINE_UNCORE_FORMAT_ATTR(match, match, "config1:0-63"); +DEFINE_UNCORE_FORMAT_ATTR(mask, mask, "config2:0-63"); + +static void nhmex_uncore_msr_init_box(struct intel_uncore_box *box) +{ + wrmsrl(NHMEX_U_MSR_PMON_GLOBAL_CTL, NHMEX_U_PMON_GLOBAL_EN_ALL); +} + +static void nhmex_uncore_msr_disable_box(struct intel_uncore_box *box) +{ + unsigned msr = uncore_msr_box_ctl(box); + u64 config; + + if (msr) { + rdmsrl(msr, config); + config &= ~((1ULL << uncore_num_counters(box)) - 1); + /* WBox has a fixed counter */ + if (uncore_msr_fixed_ctl(box)) + config &= ~NHMEX_W_PMON_GLOBAL_FIXED_EN; + wrmsrl(msr, config); + } +} + +static void nhmex_uncore_msr_enable_box(struct intel_uncore_box *box) +{ + unsigned msr = uncore_msr_box_ctl(box); + u64 config; + + if (msr) { + rdmsrl(msr, config); + config |= (1ULL << uncore_num_counters(box)) - 1; + /* WBox has a fixed counter */ + if (uncore_msr_fixed_ctl(box)) + config |= NHMEX_W_PMON_GLOBAL_FIXED_EN; + wrmsrl(msr, config); + } +} + +static void nhmex_uncore_msr_disable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + wrmsrl(event->hw.config_base, 0); +} + +static void nhmex_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + + if (hwc->idx >= UNCORE_PMC_IDX_FIXED) + wrmsrl(hwc->config_base, NHMEX_PMON_CTL_EN_BIT0); + else if (box->pmu->type->event_mask & NHMEX_PMON_CTL_EN_BIT0) + wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT22); + else + wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT0); +} + +#define NHMEX_UNCORE_OPS_COMMON_INIT() \ + .init_box = nhmex_uncore_msr_init_box, \ + .disable_box = nhmex_uncore_msr_disable_box, \ + .enable_box = nhmex_uncore_msr_enable_box, \ + .disable_event = nhmex_uncore_msr_disable_event, \ + .read_counter = uncore_msr_read_counter + +static struct intel_uncore_ops nhmex_uncore_ops = { + NHMEX_UNCORE_OPS_COMMON_INIT(), + .enable_event = nhmex_uncore_msr_enable_event, +}; + +static struct attribute *nhmex_uncore_ubox_formats_attr[] = { + &format_attr_event.attr, + &format_attr_edge.attr, + NULL, +}; + +static struct attribute_group nhmex_uncore_ubox_format_group = { + .name = "format", + .attrs = nhmex_uncore_ubox_formats_attr, +}; + +static struct intel_uncore_type nhmex_uncore_ubox = { + .name = "ubox", + .num_counters = 1, + .num_boxes = 1, + .perf_ctr_bits = 48, + .event_ctl = NHMEX_U_MSR_PMON_EV_SEL, + .perf_ctr = NHMEX_U_MSR_PMON_CTR, + .event_mask = NHMEX_U_PMON_RAW_EVENT_MASK, + .box_ctl = NHMEX_U_MSR_PMON_GLOBAL_CTL, + .ops = &nhmex_uncore_ops, + .format_group = &nhmex_uncore_ubox_format_group +}; + +static struct attribute *nhmex_uncore_cbox_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh8.attr, + NULL, +}; + +static struct attribute_group nhmex_uncore_cbox_format_group = { + .name = "format", + .attrs = nhmex_uncore_cbox_formats_attr, +}; + +/* msr offset for each instance of cbox */ +static unsigned nhmex_cbox_msr_offsets[] = { + 0x0, 0x80, 0x40, 0xc0, 0x20, 0xa0, 0x60, 0xe0, 0x240, 0x2c0, +}; + +static struct intel_uncore_type nhmex_uncore_cbox = { + .name = "cbox", + .num_counters = 6, + .num_boxes = 10, + .perf_ctr_bits = 48, + .event_ctl = NHMEX_C0_MSR_PMON_EV_SEL0, + .perf_ctr = NHMEX_C0_MSR_PMON_CTR0, + .event_mask = NHMEX_PMON_RAW_EVENT_MASK, + .box_ctl = NHMEX_C0_MSR_PMON_GLOBAL_CTL, + .msr_offsets = nhmex_cbox_msr_offsets, + .pair_ctr_ctl = 1, + .ops = &nhmex_uncore_ops, + .format_group = &nhmex_uncore_cbox_format_group +}; + +static struct uncore_event_desc nhmex_uncore_wbox_events[] = { + INTEL_UNCORE_EVENT_DESC(clockticks, "event=0xff,umask=0"), + { /* end: all zeroes */ }, +}; + +static struct intel_uncore_type nhmex_uncore_wbox = { + .name = "wbox", + .num_counters = 4, + .num_boxes = 1, + .perf_ctr_bits = 48, + .event_ctl = NHMEX_W_MSR_PMON_CNT0, + .perf_ctr = NHMEX_W_MSR_PMON_EVT_SEL0, + .fixed_ctr = NHMEX_W_MSR_PMON_FIXED_CTR, + .fixed_ctl = NHMEX_W_MSR_PMON_FIXED_CTL, + .event_mask = NHMEX_PMON_RAW_EVENT_MASK, + .box_ctl = NHMEX_W_MSR_GLOBAL_CTL, + .pair_ctr_ctl = 1, + .event_descs = nhmex_uncore_wbox_events, + .ops = &nhmex_uncore_ops, + .format_group = &nhmex_uncore_cbox_format_group +}; + +static int nhmex_bbox_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + int ctr, ev_sel; + + ctr = (hwc->config & NHMEX_B_PMON_CTR_MASK) >> + NHMEX_B_PMON_CTR_SHIFT; + ev_sel = (hwc->config & NHMEX_B_PMON_CTL_EV_SEL_MASK) >> + NHMEX_B_PMON_CTL_EV_SEL_SHIFT; + + /* events that do not use the match/mask registers */ + if ((ctr == 0 && ev_sel > 0x3) || (ctr == 1 && ev_sel > 0x6) || + (ctr == 2 && ev_sel != 0x4) || ctr == 3) + return 0; + + if (box->pmu->pmu_idx == 0) + reg1->reg = NHMEX_B0_MSR_MATCH; + else + reg1->reg = NHMEX_B1_MSR_MATCH; + reg1->idx = 0; + reg1->config = event->attr.config1; + reg2->config = event->attr.config2; + return 0; +} + +static void nhmex_bbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + + if (reg1->idx != EXTRA_REG_NONE) { + wrmsrl(reg1->reg, reg1->config); + wrmsrl(reg1->reg + 1, reg2->config); + } + wrmsrl(hwc->config_base, NHMEX_PMON_CTL_EN_BIT0 | + (hwc->config & NHMEX_B_PMON_CTL_EV_SEL_MASK)); +} + +/* + * The Bbox has 4 counters, but each counter monitors different events. + * Use bits 6-7 in the event config to select counter. + */ +static struct event_constraint nhmex_uncore_bbox_constraints[] = { + EVENT_CONSTRAINT(0 , 1, 0xc0), + EVENT_CONSTRAINT(0x40, 2, 0xc0), + EVENT_CONSTRAINT(0x80, 4, 0xc0), + EVENT_CONSTRAINT(0xc0, 8, 0xc0), + EVENT_CONSTRAINT_END, +}; + +static struct attribute *nhmex_uncore_bbox_formats_attr[] = { + &format_attr_event5.attr, + &format_attr_counter.attr, + &format_attr_match.attr, + &format_attr_mask.attr, + NULL, +}; + +static struct attribute_group nhmex_uncore_bbox_format_group = { + .name = "format", + .attrs = nhmex_uncore_bbox_formats_attr, +}; + +static struct intel_uncore_ops nhmex_uncore_bbox_ops = { + NHMEX_UNCORE_OPS_COMMON_INIT(), + .enable_event = nhmex_bbox_msr_enable_event, + .hw_config = nhmex_bbox_hw_config, + .get_constraint = uncore_get_constraint, + .put_constraint = uncore_put_constraint, +}; + +static struct intel_uncore_type nhmex_uncore_bbox = { + .name = "bbox", + .num_counters = 4, + .num_boxes = 2, + .perf_ctr_bits = 48, + .event_ctl = NHMEX_B0_MSR_PMON_CTL0, + .perf_ctr = NHMEX_B0_MSR_PMON_CTR0, + .event_mask = NHMEX_B_PMON_RAW_EVENT_MASK, + .box_ctl = NHMEX_B0_MSR_PMON_GLOBAL_CTL, + .msr_offset = NHMEX_B_MSR_OFFSET, + .pair_ctr_ctl = 1, + .num_shared_regs = 1, + .constraints = nhmex_uncore_bbox_constraints, + .ops = &nhmex_uncore_bbox_ops, + .format_group = &nhmex_uncore_bbox_format_group +}; + +static int nhmex_sbox_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + + /* only TO_R_PROG_EV event uses the match/mask register */ + if ((hwc->config & NHMEX_PMON_CTL_EV_SEL_MASK) != + NHMEX_S_EVENT_TO_R_PROG_EV) + return 0; + + if (box->pmu->pmu_idx == 0) + reg1->reg = NHMEX_S0_MSR_MM_CFG; + else + reg1->reg = NHMEX_S1_MSR_MM_CFG; + reg1->idx = 0; + reg1->config = event->attr.config1; + reg2->config = event->attr.config2; + return 0; +} + +static void nhmex_sbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + + if (reg1->idx != EXTRA_REG_NONE) { + wrmsrl(reg1->reg, 0); + wrmsrl(reg1->reg + 1, reg1->config); + wrmsrl(reg1->reg + 2, reg2->config); + wrmsrl(reg1->reg, NHMEX_S_PMON_MM_CFG_EN); + } + wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT22); +} + +static struct attribute *nhmex_uncore_sbox_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_thresh8.attr, + &format_attr_match.attr, + &format_attr_mask.attr, + NULL, +}; + +static struct attribute_group nhmex_uncore_sbox_format_group = { + .name = "format", + .attrs = nhmex_uncore_sbox_formats_attr, +}; + +static struct intel_uncore_ops nhmex_uncore_sbox_ops = { + NHMEX_UNCORE_OPS_COMMON_INIT(), + .enable_event = nhmex_sbox_msr_enable_event, + .hw_config = nhmex_sbox_hw_config, + .get_constraint = uncore_get_constraint, + .put_constraint = uncore_put_constraint, +}; + +static struct intel_uncore_type nhmex_uncore_sbox = { + .name = "sbox", + .num_counters = 4, + .num_boxes = 2, + .perf_ctr_bits = 48, + .event_ctl = NHMEX_S0_MSR_PMON_CTL0, + .perf_ctr = NHMEX_S0_MSR_PMON_CTR0, + .event_mask = NHMEX_PMON_RAW_EVENT_MASK, + .box_ctl = NHMEX_S0_MSR_PMON_GLOBAL_CTL, + .msr_offset = NHMEX_S_MSR_OFFSET, + .pair_ctr_ctl = 1, + .num_shared_regs = 1, + .ops = &nhmex_uncore_sbox_ops, + .format_group = &nhmex_uncore_sbox_format_group +}; + +enum { + EXTRA_REG_NHMEX_M_FILTER, + EXTRA_REG_NHMEX_M_DSP, + EXTRA_REG_NHMEX_M_ISS, + EXTRA_REG_NHMEX_M_MAP, + EXTRA_REG_NHMEX_M_MSC_THR, + EXTRA_REG_NHMEX_M_PGT, + EXTRA_REG_NHMEX_M_PLD, + EXTRA_REG_NHMEX_M_ZDP_CTL_FVC, +}; + +static struct extra_reg nhmex_uncore_mbox_extra_regs[] = { + MBOX_INC_SEL_EXTAR_REG(0x0, DSP), + MBOX_INC_SEL_EXTAR_REG(0x4, MSC_THR), + MBOX_INC_SEL_EXTAR_REG(0x5, MSC_THR), + MBOX_INC_SEL_EXTAR_REG(0x9, ISS), + /* event 0xa uses two extra registers */ + MBOX_INC_SEL_EXTAR_REG(0xa, ISS), + MBOX_INC_SEL_EXTAR_REG(0xa, PLD), + MBOX_INC_SEL_EXTAR_REG(0xb, PLD), + /* events 0xd ~ 0x10 use the same extra register */ + MBOX_INC_SEL_EXTAR_REG(0xd, ZDP_CTL_FVC), + MBOX_INC_SEL_EXTAR_REG(0xe, ZDP_CTL_FVC), + MBOX_INC_SEL_EXTAR_REG(0xf, ZDP_CTL_FVC), + MBOX_INC_SEL_EXTAR_REG(0x10, ZDP_CTL_FVC), + MBOX_INC_SEL_EXTAR_REG(0x16, PGT), + MBOX_SET_FLAG_SEL_EXTRA_REG(0x0, DSP), + MBOX_SET_FLAG_SEL_EXTRA_REG(0x1, ISS), + MBOX_SET_FLAG_SEL_EXTRA_REG(0x5, PGT), + MBOX_SET_FLAG_SEL_EXTRA_REG(0x6, MAP), + EVENT_EXTRA_END +}; + +/* Nehalem-EX or Westmere-EX ? */ +static bool uncore_nhmex; + +static bool nhmex_mbox_get_shared_reg(struct intel_uncore_box *box, int idx, u64 config) +{ + struct intel_uncore_extra_reg *er; + unsigned long flags; + bool ret = false; + u64 mask; + + if (idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) { + er = &box->shared_regs[idx]; + raw_spin_lock_irqsave(&er->lock, flags); + if (!atomic_read(&er->ref) || er->config == config) { + atomic_inc(&er->ref); + er->config = config; + ret = true; + } + raw_spin_unlock_irqrestore(&er->lock, flags); + + return ret; + } + /* + * The ZDP_CTL_FVC MSR has 4 fields which are used to control + * events 0xd ~ 0x10. Besides these 4 fields, there are additional + * fields which are shared. + */ + idx -= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC; + if (WARN_ON_ONCE(idx >= 4)) + return false; + + /* mask of the shared fields */ + if (uncore_nhmex) + mask = NHMEX_M_PMON_ZDP_CTL_FVC_MASK; + else + mask = WSMEX_M_PMON_ZDP_CTL_FVC_MASK; + er = &box->shared_regs[EXTRA_REG_NHMEX_M_ZDP_CTL_FVC]; + + raw_spin_lock_irqsave(&er->lock, flags); + /* add mask of the non-shared field if it's in use */ + if (__BITS_VALUE(atomic_read(&er->ref), idx, 8)) { + if (uncore_nhmex) + mask |= NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx); + else + mask |= WSMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx); + } + + if (!atomic_read(&er->ref) || !((er->config ^ config) & mask)) { + atomic_add(1 << (idx * 8), &er->ref); + if (uncore_nhmex) + mask = NHMEX_M_PMON_ZDP_CTL_FVC_MASK | + NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx); + else + mask = WSMEX_M_PMON_ZDP_CTL_FVC_MASK | + WSMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx); + er->config &= ~mask; + er->config |= (config & mask); + ret = true; + } + raw_spin_unlock_irqrestore(&er->lock, flags); + + return ret; +} + +static void nhmex_mbox_put_shared_reg(struct intel_uncore_box *box, int idx) +{ + struct intel_uncore_extra_reg *er; + + if (idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) { + er = &box->shared_regs[idx]; + atomic_dec(&er->ref); + return; + } + + idx -= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC; + er = &box->shared_regs[EXTRA_REG_NHMEX_M_ZDP_CTL_FVC]; + atomic_sub(1 << (idx * 8), &er->ref); +} + +static u64 nhmex_mbox_alter_er(struct perf_event *event, int new_idx, bool modify) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + u64 idx, orig_idx = __BITS_VALUE(reg1->idx, 0, 8); + u64 config = reg1->config; + + /* get the non-shared control bits and shift them */ + idx = orig_idx - EXTRA_REG_NHMEX_M_ZDP_CTL_FVC; + if (uncore_nhmex) + config &= NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx); + else + config &= WSMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx); + if (new_idx > orig_idx) { + idx = new_idx - orig_idx; + config <<= 3 * idx; + } else { + idx = orig_idx - new_idx; + config >>= 3 * idx; + } + + /* add the shared control bits back */ + if (uncore_nhmex) + config |= NHMEX_M_PMON_ZDP_CTL_FVC_MASK & reg1->config; + else + config |= WSMEX_M_PMON_ZDP_CTL_FVC_MASK & reg1->config; + config |= NHMEX_M_PMON_ZDP_CTL_FVC_MASK & reg1->config; + if (modify) { + /* adjust the main event selector */ + if (new_idx > orig_idx) + hwc->config += idx << NHMEX_M_PMON_CTL_INC_SEL_SHIFT; + else + hwc->config -= idx << NHMEX_M_PMON_CTL_INC_SEL_SHIFT; + reg1->config = config; + reg1->idx = ~0xff | new_idx; + } + return config; +} + +static struct event_constraint * +nhmex_mbox_get_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct hw_perf_event_extra *reg2 = &event->hw.branch_reg; + int i, idx[2], alloc = 0; + u64 config1 = reg1->config; + + idx[0] = __BITS_VALUE(reg1->idx, 0, 8); + idx[1] = __BITS_VALUE(reg1->idx, 1, 8); +again: + for (i = 0; i < 2; i++) { + if (!uncore_box_is_fake(box) && (reg1->alloc & (0x1 << i))) + idx[i] = 0xff; + + if (idx[i] == 0xff) + continue; + + if (!nhmex_mbox_get_shared_reg(box, idx[i], + __BITS_VALUE(config1, i, 32))) + goto fail; + alloc |= (0x1 << i); + } + + /* for the match/mask registers */ + if (reg2->idx != EXTRA_REG_NONE && + (uncore_box_is_fake(box) || !reg2->alloc) && + !nhmex_mbox_get_shared_reg(box, reg2->idx, reg2->config)) + goto fail; + + /* + * If it's a fake box -- as per validate_{group,event}() we + * shouldn't touch event state and we can avoid doing so + * since both will only call get_event_constraints() once + * on each event, this avoids the need for reg->alloc. + */ + if (!uncore_box_is_fake(box)) { + if (idx[0] != 0xff && idx[0] != __BITS_VALUE(reg1->idx, 0, 8)) + nhmex_mbox_alter_er(event, idx[0], true); + reg1->alloc |= alloc; + if (reg2->idx != EXTRA_REG_NONE) + reg2->alloc = 1; + } + return NULL; +fail: + if (idx[0] != 0xff && !(alloc & 0x1) && + idx[0] >= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) { + /* + * events 0xd ~ 0x10 are functional identical, but are + * controlled by different fields in the ZDP_CTL_FVC + * register. If we failed to take one field, try the + * rest 3 choices. + */ + BUG_ON(__BITS_VALUE(reg1->idx, 1, 8) != 0xff); + idx[0] -= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC; + idx[0] = (idx[0] + 1) % 4; + idx[0] += EXTRA_REG_NHMEX_M_ZDP_CTL_FVC; + if (idx[0] != __BITS_VALUE(reg1->idx, 0, 8)) { + config1 = nhmex_mbox_alter_er(event, idx[0], false); + goto again; + } + } + + if (alloc & 0x1) + nhmex_mbox_put_shared_reg(box, idx[0]); + if (alloc & 0x2) + nhmex_mbox_put_shared_reg(box, idx[1]); + return &constraint_empty; +} + +static void nhmex_mbox_put_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct hw_perf_event_extra *reg2 = &event->hw.branch_reg; + + if (uncore_box_is_fake(box)) + return; + + if (reg1->alloc & 0x1) + nhmex_mbox_put_shared_reg(box, __BITS_VALUE(reg1->idx, 0, 8)); + if (reg1->alloc & 0x2) + nhmex_mbox_put_shared_reg(box, __BITS_VALUE(reg1->idx, 1, 8)); + reg1->alloc = 0; + + if (reg2->alloc) { + nhmex_mbox_put_shared_reg(box, reg2->idx); + reg2->alloc = 0; + } +} + +static int nhmex_mbox_extra_reg_idx(struct extra_reg *er) +{ + if (er->idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) + return er->idx; + return er->idx + (er->event >> NHMEX_M_PMON_CTL_INC_SEL_SHIFT) - 0xd; +} + +static int nhmex_mbox_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct intel_uncore_type *type = box->pmu->type; + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct hw_perf_event_extra *reg2 = &event->hw.branch_reg; + struct extra_reg *er; + unsigned msr; + int reg_idx = 0; + /* + * The mbox events may require 2 extra MSRs at the most. But only + * the lower 32 bits in these MSRs are significant, so we can use + * config1 to pass two MSRs' config. + */ + for (er = nhmex_uncore_mbox_extra_regs; er->msr; er++) { + if (er->event != (event->hw.config & er->config_mask)) + continue; + if (event->attr.config1 & ~er->valid_mask) + return -EINVAL; + + msr = er->msr + type->msr_offset * box->pmu->pmu_idx; + if (WARN_ON_ONCE(msr >= 0xffff || er->idx >= 0xff)) + return -EINVAL; + + /* always use the 32~63 bits to pass the PLD config */ + if (er->idx == EXTRA_REG_NHMEX_M_PLD) + reg_idx = 1; + else if (WARN_ON_ONCE(reg_idx > 0)) + return -EINVAL; + + reg1->idx &= ~(0xff << (reg_idx * 8)); + reg1->reg &= ~(0xffff << (reg_idx * 16)); + reg1->idx |= nhmex_mbox_extra_reg_idx(er) << (reg_idx * 8); + reg1->reg |= msr << (reg_idx * 16); + reg1->config = event->attr.config1; + reg_idx++; + } + /* + * The mbox only provides ability to perform address matching + * for the PLD events. + */ + if (reg_idx == 2) { + reg2->idx = EXTRA_REG_NHMEX_M_FILTER; + if (event->attr.config2 & NHMEX_M_PMON_MM_CFG_EN) + reg2->config = event->attr.config2; + else + reg2->config = ~0ULL; + if (box->pmu->pmu_idx == 0) + reg2->reg = NHMEX_M0_MSR_PMU_MM_CFG; + else + reg2->reg = NHMEX_M1_MSR_PMU_MM_CFG; + } + return 0; +} + +static u64 nhmex_mbox_shared_reg_config(struct intel_uncore_box *box, int idx) +{ + struct intel_uncore_extra_reg *er; + unsigned long flags; + u64 config; + + if (idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) + return box->shared_regs[idx].config; + + er = &box->shared_regs[EXTRA_REG_NHMEX_M_ZDP_CTL_FVC]; + raw_spin_lock_irqsave(&er->lock, flags); + config = er->config; + raw_spin_unlock_irqrestore(&er->lock, flags); + return config; +} + +static void nhmex_mbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + int idx; + + idx = __BITS_VALUE(reg1->idx, 0, 8); + if (idx != 0xff) + wrmsrl(__BITS_VALUE(reg1->reg, 0, 16), + nhmex_mbox_shared_reg_config(box, idx)); + idx = __BITS_VALUE(reg1->idx, 1, 8); + if (idx != 0xff) + wrmsrl(__BITS_VALUE(reg1->reg, 1, 16), + nhmex_mbox_shared_reg_config(box, idx)); + + if (reg2->idx != EXTRA_REG_NONE) { + wrmsrl(reg2->reg, 0); + if (reg2->config != ~0ULL) { + wrmsrl(reg2->reg + 1, + reg2->config & NHMEX_M_PMON_ADDR_MATCH_MASK); + wrmsrl(reg2->reg + 2, NHMEX_M_PMON_ADDR_MASK_MASK & + (reg2->config >> NHMEX_M_PMON_ADDR_MASK_SHIFT)); + wrmsrl(reg2->reg, NHMEX_M_PMON_MM_CFG_EN); + } + } + + wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT0); +} + +DEFINE_UNCORE_FORMAT_ATTR(count_mode, count_mode, "config:2-3"); +DEFINE_UNCORE_FORMAT_ATTR(storage_mode, storage_mode, "config:4-5"); +DEFINE_UNCORE_FORMAT_ATTR(wrap_mode, wrap_mode, "config:6"); +DEFINE_UNCORE_FORMAT_ATTR(flag_mode, flag_mode, "config:7"); +DEFINE_UNCORE_FORMAT_ATTR(inc_sel, inc_sel, "config:9-13"); +DEFINE_UNCORE_FORMAT_ATTR(set_flag_sel, set_flag_sel, "config:19-21"); +DEFINE_UNCORE_FORMAT_ATTR(filter_cfg_en, filter_cfg_en, "config2:63"); +DEFINE_UNCORE_FORMAT_ATTR(filter_match, filter_match, "config2:0-33"); +DEFINE_UNCORE_FORMAT_ATTR(filter_mask, filter_mask, "config2:34-61"); +DEFINE_UNCORE_FORMAT_ATTR(dsp, dsp, "config1:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(thr, thr, "config1:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(fvc, fvc, "config1:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(pgt, pgt, "config1:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(map, map, "config1:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(iss, iss, "config1:0-31"); +DEFINE_UNCORE_FORMAT_ATTR(pld, pld, "config1:32-63"); + +static struct attribute *nhmex_uncore_mbox_formats_attr[] = { + &format_attr_count_mode.attr, + &format_attr_storage_mode.attr, + &format_attr_wrap_mode.attr, + &format_attr_flag_mode.attr, + &format_attr_inc_sel.attr, + &format_attr_set_flag_sel.attr, + &format_attr_filter_cfg_en.attr, + &format_attr_filter_match.attr, + &format_attr_filter_mask.attr, + &format_attr_dsp.attr, + &format_attr_thr.attr, + &format_attr_fvc.attr, + &format_attr_pgt.attr, + &format_attr_map.attr, + &format_attr_iss.attr, + &format_attr_pld.attr, + NULL, +}; + +static struct attribute_group nhmex_uncore_mbox_format_group = { + .name = "format", + .attrs = nhmex_uncore_mbox_formats_attr, +}; + +static struct uncore_event_desc nhmex_uncore_mbox_events[] = { + INTEL_UNCORE_EVENT_DESC(bbox_cmds_read, "inc_sel=0xd,fvc=0x2800"), + INTEL_UNCORE_EVENT_DESC(bbox_cmds_write, "inc_sel=0xd,fvc=0x2820"), + { /* end: all zeroes */ }, +}; + +static struct uncore_event_desc wsmex_uncore_mbox_events[] = { + INTEL_UNCORE_EVENT_DESC(bbox_cmds_read, "inc_sel=0xd,fvc=0x5000"), + INTEL_UNCORE_EVENT_DESC(bbox_cmds_write, "inc_sel=0xd,fvc=0x5040"), + { /* end: all zeroes */ }, +}; + +static struct intel_uncore_ops nhmex_uncore_mbox_ops = { + NHMEX_UNCORE_OPS_COMMON_INIT(), + .enable_event = nhmex_mbox_msr_enable_event, + .hw_config = nhmex_mbox_hw_config, + .get_constraint = nhmex_mbox_get_constraint, + .put_constraint = nhmex_mbox_put_constraint, +}; + +static struct intel_uncore_type nhmex_uncore_mbox = { + .name = "mbox", + .num_counters = 6, + .num_boxes = 2, + .perf_ctr_bits = 48, + .event_ctl = NHMEX_M0_MSR_PMU_CTL0, + .perf_ctr = NHMEX_M0_MSR_PMU_CNT0, + .event_mask = NHMEX_M_PMON_RAW_EVENT_MASK, + .box_ctl = NHMEX_M0_MSR_GLOBAL_CTL, + .msr_offset = NHMEX_M_MSR_OFFSET, + .pair_ctr_ctl = 1, + .num_shared_regs = 8, + .event_descs = nhmex_uncore_mbox_events, + .ops = &nhmex_uncore_mbox_ops, + .format_group = &nhmex_uncore_mbox_format_group, +}; + +static void nhmex_rbox_alter_er(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + + /* adjust the main event selector and extra register index */ + if (reg1->idx % 2) { + reg1->idx--; + hwc->config -= 1 << NHMEX_R_PMON_CTL_EV_SEL_SHIFT; + } else { + reg1->idx++; + hwc->config += 1 << NHMEX_R_PMON_CTL_EV_SEL_SHIFT; + } + + /* adjust extra register config */ + switch (reg1->idx % 6) { + case 2: + /* shift the 8~15 bits to the 0~7 bits */ + reg1->config >>= 8; + break; + case 3: + /* shift the 0~7 bits to the 8~15 bits */ + reg1->config <<= 8; + break; + }; +} + +/* + * Each rbox has 4 event set which monitor PQI port 0~3 or 4~7. + * An event set consists of 6 events, the 3rd and 4th events in + * an event set use the same extra register. So an event set uses + * 5 extra registers. + */ +static struct event_constraint * +nhmex_rbox_get_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + struct intel_uncore_extra_reg *er; + unsigned long flags; + int idx, er_idx; + u64 config1; + bool ok = false; + + if (!uncore_box_is_fake(box) && reg1->alloc) + return NULL; + + idx = reg1->idx % 6; + config1 = reg1->config; +again: + er_idx = idx; + /* the 3rd and 4th events use the same extra register */ + if (er_idx > 2) + er_idx--; + er_idx += (reg1->idx / 6) * 5; + + er = &box->shared_regs[er_idx]; + raw_spin_lock_irqsave(&er->lock, flags); + if (idx < 2) { + if (!atomic_read(&er->ref) || er->config == reg1->config) { + atomic_inc(&er->ref); + er->config = reg1->config; + ok = true; + } + } else if (idx == 2 || idx == 3) { + /* + * these two events use different fields in a extra register, + * the 0~7 bits and the 8~15 bits respectively. + */ + u64 mask = 0xff << ((idx - 2) * 8); + if (!__BITS_VALUE(atomic_read(&er->ref), idx - 2, 8) || + !((er->config ^ config1) & mask)) { + atomic_add(1 << ((idx - 2) * 8), &er->ref); + er->config &= ~mask; + er->config |= config1 & mask; + ok = true; + } + } else { + if (!atomic_read(&er->ref) || + (er->config == (hwc->config >> 32) && + er->config1 == reg1->config && + er->config2 == reg2->config)) { + atomic_inc(&er->ref); + er->config = (hwc->config >> 32); + er->config1 = reg1->config; + er->config2 = reg2->config; + ok = true; + } + } + raw_spin_unlock_irqrestore(&er->lock, flags); + + if (!ok) { + /* + * The Rbox events are always in pairs. The paired + * events are functional identical, but use different + * extra registers. If we failed to take an extra + * register, try the alternative. + */ + if (idx % 2) + idx--; + else + idx++; + if (idx != reg1->idx % 6) { + if (idx == 2) + config1 >>= 8; + else if (idx == 3) + config1 <<= 8; + goto again; + } + } else { + if (!uncore_box_is_fake(box)) { + if (idx != reg1->idx % 6) + nhmex_rbox_alter_er(box, event); + reg1->alloc = 1; + } + return NULL; + } + return &constraint_empty; +} + +static void nhmex_rbox_put_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct intel_uncore_extra_reg *er; + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + int idx, er_idx; + + if (uncore_box_is_fake(box) || !reg1->alloc) + return; + + idx = reg1->idx % 6; + er_idx = idx; + if (er_idx > 2) + er_idx--; + er_idx += (reg1->idx / 6) * 5; + + er = &box->shared_regs[er_idx]; + if (idx == 2 || idx == 3) + atomic_sub(1 << ((idx - 2) * 8), &er->ref); + else + atomic_dec(&er->ref); + + reg1->alloc = 0; +} + +static int nhmex_rbox_hw_config(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &event->hw.extra_reg; + struct hw_perf_event_extra *reg2 = &event->hw.branch_reg; + int idx; + + idx = (event->hw.config & NHMEX_R_PMON_CTL_EV_SEL_MASK) >> + NHMEX_R_PMON_CTL_EV_SEL_SHIFT; + if (idx >= 0x18) + return -EINVAL; + + reg1->idx = idx; + reg1->config = event->attr.config1; + + switch (idx % 6) { + case 4: + case 5: + hwc->config |= event->attr.config & (~0ULL << 32); + reg2->config = event->attr.config2; + break; + }; + return 0; +} + +static void nhmex_rbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + struct hw_perf_event_extra *reg1 = &hwc->extra_reg; + struct hw_perf_event_extra *reg2 = &hwc->branch_reg; + int idx, port; + + idx = reg1->idx; + port = idx / 6 + box->pmu->pmu_idx * 4; + + switch (idx % 6) { + case 0: + wrmsrl(NHMEX_R_MSR_PORTN_IPERF_CFG0(port), reg1->config); + break; + case 1: + wrmsrl(NHMEX_R_MSR_PORTN_IPERF_CFG1(port), reg1->config); + break; + case 2: + case 3: + wrmsrl(NHMEX_R_MSR_PORTN_QLX_CFG(port), + uncore_shared_reg_config(box, 2 + (idx / 6) * 5)); + break; + case 4: + wrmsrl(NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(port), + hwc->config >> 32); + wrmsrl(NHMEX_R_MSR_PORTN_XBR_SET1_MATCH(port), reg1->config); + wrmsrl(NHMEX_R_MSR_PORTN_XBR_SET1_MASK(port), reg2->config); + break; + case 5: + wrmsrl(NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(port), + hwc->config >> 32); + wrmsrl(NHMEX_R_MSR_PORTN_XBR_SET2_MATCH(port), reg1->config); + wrmsrl(NHMEX_R_MSR_PORTN_XBR_SET2_MASK(port), reg2->config); + break; + }; + + wrmsrl(hwc->config_base, NHMEX_PMON_CTL_EN_BIT0 | + (hwc->config & NHMEX_R_PMON_CTL_EV_SEL_MASK)); +} + +DEFINE_UNCORE_FORMAT_ATTR(xbr_mm_cfg, xbr_mm_cfg, "config:32-63"); +DEFINE_UNCORE_FORMAT_ATTR(xbr_match, xbr_match, "config1:0-63"); +DEFINE_UNCORE_FORMAT_ATTR(xbr_mask, xbr_mask, "config2:0-63"); +DEFINE_UNCORE_FORMAT_ATTR(qlx_cfg, qlx_cfg, "config1:0-15"); +DEFINE_UNCORE_FORMAT_ATTR(iperf_cfg, iperf_cfg, "config1:0-31"); + +static struct attribute *nhmex_uncore_rbox_formats_attr[] = { + &format_attr_event5.attr, + &format_attr_xbr_mm_cfg.attr, + &format_attr_xbr_match.attr, + &format_attr_xbr_mask.attr, + &format_attr_qlx_cfg.attr, + &format_attr_iperf_cfg.attr, + NULL, +}; + +static struct attribute_group nhmex_uncore_rbox_format_group = { + .name = "format", + .attrs = nhmex_uncore_rbox_formats_attr, +}; + +static struct uncore_event_desc nhmex_uncore_rbox_events[] = { + INTEL_UNCORE_EVENT_DESC(qpi0_flit_send, "event=0x0,iperf_cfg=0x80000000"), + INTEL_UNCORE_EVENT_DESC(qpi1_filt_send, "event=0x6,iperf_cfg=0x80000000"), + INTEL_UNCORE_EVENT_DESC(qpi0_idle_filt, "event=0x0,iperf_cfg=0x40000000"), + INTEL_UNCORE_EVENT_DESC(qpi1_idle_filt, "event=0x6,iperf_cfg=0x40000000"), + INTEL_UNCORE_EVENT_DESC(qpi0_date_response, "event=0x0,iperf_cfg=0xc4"), + INTEL_UNCORE_EVENT_DESC(qpi1_date_response, "event=0x6,iperf_cfg=0xc4"), + { /* end: all zeroes */ }, +}; + +static struct intel_uncore_ops nhmex_uncore_rbox_ops = { + NHMEX_UNCORE_OPS_COMMON_INIT(), + .enable_event = nhmex_rbox_msr_enable_event, + .hw_config = nhmex_rbox_hw_config, + .get_constraint = nhmex_rbox_get_constraint, + .put_constraint = nhmex_rbox_put_constraint, +}; + +static struct intel_uncore_type nhmex_uncore_rbox = { + .name = "rbox", + .num_counters = 8, + .num_boxes = 2, + .perf_ctr_bits = 48, + .event_ctl = NHMEX_R_MSR_PMON_CTL0, + .perf_ctr = NHMEX_R_MSR_PMON_CNT0, + .event_mask = NHMEX_R_PMON_RAW_EVENT_MASK, + .box_ctl = NHMEX_R_MSR_GLOBAL_CTL, + .msr_offset = NHMEX_R_MSR_OFFSET, + .pair_ctr_ctl = 1, + .num_shared_regs = 20, + .event_descs = nhmex_uncore_rbox_events, + .ops = &nhmex_uncore_rbox_ops, + .format_group = &nhmex_uncore_rbox_format_group +}; + +static struct intel_uncore_type *nhmex_msr_uncores[] = { + &nhmex_uncore_ubox, + &nhmex_uncore_cbox, + &nhmex_uncore_bbox, + &nhmex_uncore_sbox, + &nhmex_uncore_mbox, + &nhmex_uncore_rbox, + &nhmex_uncore_wbox, + NULL, +}; +/* end of Nehalem-EX uncore support */ + +static void uncore_assign_hw_event(struct intel_uncore_box *box, struct perf_event *event, int idx) +{ + struct hw_perf_event *hwc = &event->hw; + + hwc->idx = idx; + hwc->last_tag = ++box->tags[idx]; + + if (hwc->idx == UNCORE_PMC_IDX_FIXED) { + hwc->event_base = uncore_fixed_ctr(box); + hwc->config_base = uncore_fixed_ctl(box); + return; + } + + hwc->config_base = uncore_event_ctl(box, hwc->idx); + hwc->event_base = uncore_perf_ctr(box, hwc->idx); +} + +static void uncore_perf_event_update(struct intel_uncore_box *box, struct perf_event *event) +{ + u64 prev_count, new_count, delta; + int shift; + + if (event->hw.idx >= UNCORE_PMC_IDX_FIXED) + shift = 64 - uncore_fixed_ctr_bits(box); + else + shift = 64 - uncore_perf_ctr_bits(box); + + /* the hrtimer might modify the previous event value */ +again: + prev_count = local64_read(&event->hw.prev_count); + new_count = uncore_read_counter(box, event); + if (local64_xchg(&event->hw.prev_count, new_count) != prev_count) + goto again; + + delta = (new_count << shift) - (prev_count << shift); + delta >>= shift; + + local64_add(delta, &event->count); +} + +/* + * The overflow interrupt is unavailable for SandyBridge-EP, is broken + * for SandyBridge. So we use hrtimer to periodically poll the counter + * to avoid overflow. + */ +static enum hrtimer_restart uncore_pmu_hrtimer(struct hrtimer *hrtimer) +{ + struct intel_uncore_box *box; + struct perf_event *event; + unsigned long flags; + int bit; + + box = container_of(hrtimer, struct intel_uncore_box, hrtimer); + if (!box->n_active || box->cpu != smp_processor_id()) + return HRTIMER_NORESTART; + /* + * disable local interrupt to prevent uncore_pmu_event_start/stop + * to interrupt the update process + */ + local_irq_save(flags); + + /* + * handle boxes with an active event list as opposed to active + * counters + */ + list_for_each_entry(event, &box->active_list, active_entry) { + uncore_perf_event_update(box, event); + } + + for_each_set_bit(bit, box->active_mask, UNCORE_PMC_IDX_MAX) + uncore_perf_event_update(box, box->events[bit]); + + local_irq_restore(flags); + + hrtimer_forward_now(hrtimer, ns_to_ktime(box->hrtimer_duration)); + return HRTIMER_RESTART; +} + +static void uncore_pmu_start_hrtimer(struct intel_uncore_box *box) +{ + __hrtimer_start_range_ns(&box->hrtimer, + ns_to_ktime(box->hrtimer_duration), 0, + HRTIMER_MODE_REL_PINNED, 0); +} + +static void uncore_pmu_cancel_hrtimer(struct intel_uncore_box *box) +{ + hrtimer_cancel(&box->hrtimer); +} + +static void uncore_pmu_init_hrtimer(struct intel_uncore_box *box) +{ + hrtimer_init(&box->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + box->hrtimer.function = uncore_pmu_hrtimer; +} + +static struct intel_uncore_box *uncore_alloc_box(struct intel_uncore_type *type, int node) +{ + struct intel_uncore_box *box; + int i, size; + + size = sizeof(*box) + type->num_shared_regs * sizeof(struct intel_uncore_extra_reg); + + box = kzalloc_node(size, GFP_KERNEL, node); + if (!box) + return NULL; + + for (i = 0; i < type->num_shared_regs; i++) + raw_spin_lock_init(&box->shared_regs[i].lock); + + uncore_pmu_init_hrtimer(box); + atomic_set(&box->refcnt, 1); + box->cpu = -1; + box->phys_id = -1; + + /* set default hrtimer timeout */ + box->hrtimer_duration = UNCORE_PMU_HRTIMER_INTERVAL; + + INIT_LIST_HEAD(&box->active_list); + + return box; +} + +static int +uncore_collect_events(struct intel_uncore_box *box, struct perf_event *leader, bool dogrp) +{ + struct perf_event *event; + int n, max_count; + + max_count = box->pmu->type->num_counters; + if (box->pmu->type->fixed_ctl) + max_count++; + + if (box->n_events >= max_count) + return -EINVAL; + + n = box->n_events; + box->event_list[n] = leader; + n++; + if (!dogrp) + return n; + + list_for_each_entry(event, &leader->sibling_list, group_entry) { + if (event->state <= PERF_EVENT_STATE_OFF) + continue; + + if (n >= max_count) + return -EINVAL; + + box->event_list[n] = event; + n++; + } + return n; +} + +static struct event_constraint * +uncore_get_event_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + struct intel_uncore_type *type = box->pmu->type; + struct event_constraint *c; + + if (type->ops->get_constraint) { + c = type->ops->get_constraint(box, event); + if (c) + return c; + } + + if (event->attr.config == UNCORE_FIXED_EVENT) + return &constraint_fixed; + + if (type->constraints) { + for_each_event_constraint(c, type->constraints) { + if ((event->hw.config & c->cmask) == c->code) + return c; + } + } + + return &type->unconstrainted; +} + +static void uncore_put_event_constraint(struct intel_uncore_box *box, struct perf_event *event) +{ + if (box->pmu->type->ops->put_constraint) + box->pmu->type->ops->put_constraint(box, event); +} + +static int uncore_assign_events(struct intel_uncore_box *box, int assign[], int n) +{ + unsigned long used_mask[BITS_TO_LONGS(UNCORE_PMC_IDX_MAX)]; + struct event_constraint *c; + int i, wmin, wmax, ret = 0; + struct hw_perf_event *hwc; + + bitmap_zero(used_mask, UNCORE_PMC_IDX_MAX); + + for (i = 0, wmin = UNCORE_PMC_IDX_MAX, wmax = 0; i < n; i++) { + hwc = &box->event_list[i]->hw; + c = uncore_get_event_constraint(box, box->event_list[i]); + hwc->constraint = c; + wmin = min(wmin, c->weight); + wmax = max(wmax, c->weight); + } + + /* fastpath, try to reuse previous register */ + for (i = 0; i < n; i++) { + hwc = &box->event_list[i]->hw; + c = hwc->constraint; + + /* never assigned */ + if (hwc->idx == -1) + break; + + /* constraint still honored */ + if (!test_bit(hwc->idx, c->idxmsk)) + break; + + /* not already used */ + if (test_bit(hwc->idx, used_mask)) + break; + + __set_bit(hwc->idx, used_mask); + if (assign) + assign[i] = hwc->idx; + } + /* slow path */ + if (i != n) + ret = perf_assign_events(box->event_list, n, + wmin, wmax, assign); + + if (!assign || ret) { + for (i = 0; i < n; i++) + uncore_put_event_constraint(box, box->event_list[i]); + } + return ret ? -EINVAL : 0; +} + +static void uncore_pmu_event_start(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + int idx = event->hw.idx; + + if (WARN_ON_ONCE(!(event->hw.state & PERF_HES_STOPPED))) + return; + + if (WARN_ON_ONCE(idx == -1 || idx >= UNCORE_PMC_IDX_MAX)) + return; + + event->hw.state = 0; + box->events[idx] = event; + box->n_active++; + __set_bit(idx, box->active_mask); + + local64_set(&event->hw.prev_count, uncore_read_counter(box, event)); + uncore_enable_event(box, event); + + if (box->n_active == 1) { + uncore_enable_box(box); + uncore_pmu_start_hrtimer(box); + } +} + +static void uncore_pmu_event_stop(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + struct hw_perf_event *hwc = &event->hw; + + if (__test_and_clear_bit(hwc->idx, box->active_mask)) { + uncore_disable_event(box, event); + box->n_active--; + box->events[hwc->idx] = NULL; + WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED); + hwc->state |= PERF_HES_STOPPED; + + if (box->n_active == 0) { + uncore_disable_box(box); + uncore_pmu_cancel_hrtimer(box); + } + } + + if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) { + /* + * Drain the remaining delta count out of a event + * that we are disabling: + */ + uncore_perf_event_update(box, event); + hwc->state |= PERF_HES_UPTODATE; + } +} + +static int uncore_pmu_event_add(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + struct hw_perf_event *hwc = &event->hw; + int assign[UNCORE_PMC_IDX_MAX]; + int i, n, ret; + + if (!box) + return -ENODEV; + + ret = n = uncore_collect_events(box, event, false); + if (ret < 0) + return ret; + + hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + if (!(flags & PERF_EF_START)) + hwc->state |= PERF_HES_ARCH; + + ret = uncore_assign_events(box, assign, n); + if (ret) + return ret; + + /* save events moving to new counters */ + for (i = 0; i < box->n_events; i++) { + event = box->event_list[i]; + hwc = &event->hw; + + if (hwc->idx == assign[i] && + hwc->last_tag == box->tags[assign[i]]) + continue; + /* + * Ensure we don't accidentally enable a stopped + * counter simply because we rescheduled. + */ + if (hwc->state & PERF_HES_STOPPED) + hwc->state |= PERF_HES_ARCH; + + uncore_pmu_event_stop(event, PERF_EF_UPDATE); + } + + /* reprogram moved events into new counters */ + for (i = 0; i < n; i++) { + event = box->event_list[i]; + hwc = &event->hw; + + if (hwc->idx != assign[i] || + hwc->last_tag != box->tags[assign[i]]) + uncore_assign_hw_event(box, event, assign[i]); + else if (i < box->n_events) + continue; + + if (hwc->state & PERF_HES_ARCH) + continue; + + uncore_pmu_event_start(event, 0); + } + box->n_events = n; + + return 0; +} + +static void uncore_pmu_event_del(struct perf_event *event, int flags) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + int i; + + uncore_pmu_event_stop(event, PERF_EF_UPDATE); + + for (i = 0; i < box->n_events; i++) { + if (event == box->event_list[i]) { + uncore_put_event_constraint(box, event); + + while (++i < box->n_events) + box->event_list[i - 1] = box->event_list[i]; + + --box->n_events; + break; + } + } + + event->hw.idx = -1; + event->hw.last_tag = ~0ULL; +} + +static void uncore_pmu_event_read(struct perf_event *event) +{ + struct intel_uncore_box *box = uncore_event_to_box(event); + uncore_perf_event_update(box, event); +} + +/* + * validation ensures the group can be loaded onto the + * PMU if it was the only group available. + */ +static int uncore_validate_group(struct intel_uncore_pmu *pmu, + struct perf_event *event) +{ + struct perf_event *leader = event->group_leader; + struct intel_uncore_box *fake_box; + int ret = -EINVAL, n; + + fake_box = uncore_alloc_box(pmu->type, NUMA_NO_NODE); + if (!fake_box) + return -ENOMEM; + + fake_box->pmu = pmu; + /* + * the event is not yet connected with its + * siblings therefore we must first collect + * existing siblings, then add the new event + * before we can simulate the scheduling + */ + n = uncore_collect_events(fake_box, leader, true); + if (n < 0) + goto out; + + fake_box->n_events = n; + n = uncore_collect_events(fake_box, event, false); + if (n < 0) + goto out; + + fake_box->n_events = n; + + ret = uncore_assign_events(fake_box, NULL, n); +out: + kfree(fake_box); + return ret; +} + +static int uncore_pmu_event_init(struct perf_event *event) +{ + struct intel_uncore_pmu *pmu; + struct intel_uncore_box *box; + struct hw_perf_event *hwc = &event->hw; + int ret; + + if (event->attr.type != event->pmu->type) + return -ENOENT; + + pmu = uncore_event_to_pmu(event); + /* no device found for this pmu */ + if (pmu->func_id < 0) + return -ENOENT; + + /* + * Uncore PMU does measure at all privilege level all the time. + * So it doesn't make sense to specify any exclude bits. + */ + if (event->attr.exclude_user || event->attr.exclude_kernel || + event->attr.exclude_hv || event->attr.exclude_idle) + return -EINVAL; + + /* Sampling not supported yet */ + if (hwc->sample_period) + return -EINVAL; + + /* + * Place all uncore events for a particular physical package + * onto a single cpu + */ + if (event->cpu < 0) + return -EINVAL; + box = uncore_pmu_to_box(pmu, event->cpu); + if (!box || box->cpu < 0) + return -EINVAL; + event->cpu = box->cpu; + + event->hw.idx = -1; + event->hw.last_tag = ~0ULL; + event->hw.extra_reg.idx = EXTRA_REG_NONE; + event->hw.branch_reg.idx = EXTRA_REG_NONE; + + if (event->attr.config == UNCORE_FIXED_EVENT) { + /* no fixed counter */ + if (!pmu->type->fixed_ctl) + return -EINVAL; + /* + * if there is only one fixed counter, only the first pmu + * can access the fixed counter + */ + if (pmu->type->single_fixed && pmu->pmu_idx > 0) + return -EINVAL; + + /* fixed counters have event field hardcoded to zero */ + hwc->config = 0ULL; + } else { + hwc->config = event->attr.config & pmu->type->event_mask; + if (pmu->type->ops->hw_config) { + ret = pmu->type->ops->hw_config(box, event); + if (ret) + return ret; + } + } + + if (event->group_leader != event) + ret = uncore_validate_group(pmu, event); + else + ret = 0; + + return ret; +} + +static ssize_t uncore_get_attr_cpumask(struct device *dev, + struct device_attribute *attr, char *buf) +{ + int n = cpulist_scnprintf(buf, PAGE_SIZE - 2, &uncore_cpu_mask); + + buf[n++] = '\n'; + buf[n] = '\0'; + return n; +} + +static DEVICE_ATTR(cpumask, S_IRUGO, uncore_get_attr_cpumask, NULL); + +static struct attribute *uncore_pmu_attrs[] = { + &dev_attr_cpumask.attr, + NULL, +}; + +static struct attribute_group uncore_pmu_attr_group = { + .attrs = uncore_pmu_attrs, +}; + +static int __init uncore_pmu_register(struct intel_uncore_pmu *pmu) +{ + int ret; + + if (!pmu->type->pmu) { + pmu->pmu = (struct pmu) { + .attr_groups = pmu->type->attr_groups, + .task_ctx_nr = perf_invalid_context, + .event_init = uncore_pmu_event_init, + .add = uncore_pmu_event_add, + .del = uncore_pmu_event_del, + .start = uncore_pmu_event_start, + .stop = uncore_pmu_event_stop, + .read = uncore_pmu_event_read, + }; + } else { + pmu->pmu = *pmu->type->pmu; + pmu->pmu.attr_groups = pmu->type->attr_groups; + } + + if (pmu->type->num_boxes == 1) { + if (strlen(pmu->type->name) > 0) + sprintf(pmu->name, "uncore_%s", pmu->type->name); + else + sprintf(pmu->name, "uncore"); + } else { + sprintf(pmu->name, "uncore_%s_%d", pmu->type->name, + pmu->pmu_idx); + } + + ret = perf_pmu_register(&pmu->pmu, pmu->name, -1); + return ret; +} + +static void __init uncore_type_exit(struct intel_uncore_type *type) +{ + int i; + + for (i = 0; i < type->num_boxes; i++) + free_percpu(type->pmus[i].box); + kfree(type->pmus); + type->pmus = NULL; + kfree(type->events_group); + type->events_group = NULL; +} + +static void __init uncore_types_exit(struct intel_uncore_type **types) +{ + int i; + for (i = 0; types[i]; i++) + uncore_type_exit(types[i]); +} + +static int __init uncore_type_init(struct intel_uncore_type *type) +{ + struct intel_uncore_pmu *pmus; + struct attribute_group *attr_group; + struct attribute **attrs; + int i, j; + + pmus = kzalloc(sizeof(*pmus) * type->num_boxes, GFP_KERNEL); + if (!pmus) + return -ENOMEM; + + type->pmus = pmus; + + type->unconstrainted = (struct event_constraint) + __EVENT_CONSTRAINT(0, (1ULL << type->num_counters) - 1, + 0, type->num_counters, 0, 0); + + for (i = 0; i < type->num_boxes; i++) { + pmus[i].func_id = -1; + pmus[i].pmu_idx = i; + pmus[i].type = type; + INIT_LIST_HEAD(&pmus[i].box_list); + pmus[i].box = alloc_percpu(struct intel_uncore_box *); + if (!pmus[i].box) + goto fail; + } + + if (type->event_descs) { + i = 0; + while (type->event_descs[i].attr.attr.name) + i++; + + attr_group = kzalloc(sizeof(struct attribute *) * (i + 1) + + sizeof(*attr_group), GFP_KERNEL); + if (!attr_group) + goto fail; + + attrs = (struct attribute **)(attr_group + 1); + attr_group->name = "events"; + attr_group->attrs = attrs; + + for (j = 0; j < i; j++) + attrs[j] = &type->event_descs[j].attr.attr; + + type->events_group = attr_group; + } + + type->pmu_group = &uncore_pmu_attr_group; + return 0; +fail: + uncore_type_exit(type); + return -ENOMEM; +} + +static int __init uncore_types_init(struct intel_uncore_type **types) +{ + int i, ret; + + for (i = 0; types[i]; i++) { + ret = uncore_type_init(types[i]); + if (ret) + goto fail; + } + return 0; +fail: + while (--i >= 0) + uncore_type_exit(types[i]); + return ret; +} + +static struct pci_driver *uncore_pci_driver; +static bool pcidrv_registered; + +/* + * add a pci uncore device + */ +static int uncore_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + struct intel_uncore_pmu *pmu; + struct intel_uncore_box *box; + struct intel_uncore_type *type; + int phys_id; + + phys_id = pcibus_to_physid[pdev->bus->number]; + if (phys_id < 0) + return -ENODEV; + + if (UNCORE_PCI_DEV_TYPE(id->driver_data) == UNCORE_EXTRA_PCI_DEV) { + extra_pci_dev[phys_id][UNCORE_PCI_DEV_IDX(id->driver_data)] = pdev; + pci_set_drvdata(pdev, NULL); + return 0; + } + + type = pci_uncores[UNCORE_PCI_DEV_TYPE(id->driver_data)]; + box = uncore_alloc_box(type, NUMA_NO_NODE); + if (!box) + return -ENOMEM; + + /* + * for performance monitoring unit with multiple boxes, + * each box has a different function id. + */ + pmu = &type->pmus[UNCORE_PCI_DEV_IDX(id->driver_data)]; + if (pmu->func_id < 0) + pmu->func_id = pdev->devfn; + else + WARN_ON_ONCE(pmu->func_id != pdev->devfn); + + box->phys_id = phys_id; + box->pci_dev = pdev; + box->pmu = pmu; + uncore_box_init(box); + pci_set_drvdata(pdev, box); + + raw_spin_lock(&uncore_box_lock); + list_add_tail(&box->list, &pmu->box_list); + raw_spin_unlock(&uncore_box_lock); + + return 0; +} + +static void uncore_pci_remove(struct pci_dev *pdev) +{ + struct intel_uncore_box *box = pci_get_drvdata(pdev); + struct intel_uncore_pmu *pmu; + int i, cpu, phys_id = pcibus_to_physid[pdev->bus->number]; + + box = pci_get_drvdata(pdev); + if (!box) { + for (i = 0; i < UNCORE_EXTRA_PCI_DEV_MAX; i++) { + if (extra_pci_dev[phys_id][i] == pdev) { + extra_pci_dev[phys_id][i] = NULL; + break; + } + } + WARN_ON_ONCE(i >= UNCORE_EXTRA_PCI_DEV_MAX); + return; + } + + pmu = box->pmu; + if (WARN_ON_ONCE(phys_id != box->phys_id)) + return; + + pci_set_drvdata(pdev, NULL); + + raw_spin_lock(&uncore_box_lock); + list_del(&box->list); + raw_spin_unlock(&uncore_box_lock); + + for_each_possible_cpu(cpu) { + if (*per_cpu_ptr(pmu->box, cpu) == box) { + *per_cpu_ptr(pmu->box, cpu) = NULL; + atomic_dec(&box->refcnt); + } + } + + WARN_ON_ONCE(atomic_read(&box->refcnt) != 1); + kfree(box); +} + +static int __init uncore_pci_init(void) +{ + int ret; + + switch (boot_cpu_data.x86_model) { + case 45: /* Sandy Bridge-EP */ + ret = snbep_pci2phy_map_init(0x3ce0); + if (ret) + return ret; + pci_uncores = snbep_pci_uncores; + uncore_pci_driver = &snbep_uncore_pci_driver; + break; + case 62: /* IvyTown */ + ret = snbep_pci2phy_map_init(0x0e1e); + if (ret) + return ret; + pci_uncores = ivt_pci_uncores; + uncore_pci_driver = &ivt_uncore_pci_driver; + break; + case 42: /* Sandy Bridge */ + ret = snb_pci2phy_map_init(PCI_DEVICE_ID_INTEL_SNB_IMC); + if (ret) + return ret; + pci_uncores = snb_pci_uncores; + uncore_pci_driver = &snb_uncore_pci_driver; + break; + case 58: /* Ivy Bridge */ + ret = snb_pci2phy_map_init(PCI_DEVICE_ID_INTEL_IVB_IMC); + if (ret) + return ret; + pci_uncores = snb_pci_uncores; + uncore_pci_driver = &ivb_uncore_pci_driver; + break; + case 60: /* Haswell */ + case 69: /* Haswell Celeron */ + ret = snb_pci2phy_map_init(PCI_DEVICE_ID_INTEL_HSW_IMC); + if (ret) + return ret; + pci_uncores = snb_pci_uncores; + uncore_pci_driver = &hsw_uncore_pci_driver; + break; + default: + return 0; + } + + ret = uncore_types_init(pci_uncores); + if (ret) + return ret; + + uncore_pci_driver->probe = uncore_pci_probe; + uncore_pci_driver->remove = uncore_pci_remove; + + ret = pci_register_driver(uncore_pci_driver); + if (ret == 0) + pcidrv_registered = true; + else + uncore_types_exit(pci_uncores); + + return ret; +} + +static void __init uncore_pci_exit(void) +{ + if (pcidrv_registered) { + pcidrv_registered = false; + pci_unregister_driver(uncore_pci_driver); + uncore_types_exit(pci_uncores); + } +} + +/* CPU hot plug/unplug are serialized by cpu_add_remove_lock mutex */ +static LIST_HEAD(boxes_to_free); + +static void uncore_kfree_boxes(void) +{ + struct intel_uncore_box *box; + + while (!list_empty(&boxes_to_free)) { + box = list_entry(boxes_to_free.next, + struct intel_uncore_box, list); + list_del(&box->list); + kfree(box); + } +} + +static void uncore_cpu_dying(int cpu) +{ + struct intel_uncore_type *type; + struct intel_uncore_pmu *pmu; + struct intel_uncore_box *box; + int i, j; + + for (i = 0; msr_uncores[i]; i++) { + type = msr_uncores[i]; + for (j = 0; j < type->num_boxes; j++) { + pmu = &type->pmus[j]; + box = *per_cpu_ptr(pmu->box, cpu); + *per_cpu_ptr(pmu->box, cpu) = NULL; + if (box && atomic_dec_and_test(&box->refcnt)) + list_add(&box->list, &boxes_to_free); + } + } +} + +static int uncore_cpu_starting(int cpu) +{ + struct intel_uncore_type *type; + struct intel_uncore_pmu *pmu; + struct intel_uncore_box *box, *exist; + int i, j, k, phys_id; + + phys_id = topology_physical_package_id(cpu); + + for (i = 0; msr_uncores[i]; i++) { + type = msr_uncores[i]; + for (j = 0; j < type->num_boxes; j++) { + pmu = &type->pmus[j]; + box = *per_cpu_ptr(pmu->box, cpu); + /* called by uncore_cpu_init? */ + if (box && box->phys_id >= 0) { + uncore_box_init(box); + continue; + } + + for_each_online_cpu(k) { + exist = *per_cpu_ptr(pmu->box, k); + if (exist && exist->phys_id == phys_id) { + atomic_inc(&exist->refcnt); + *per_cpu_ptr(pmu->box, cpu) = exist; + if (box) { + list_add(&box->list, + &boxes_to_free); + box = NULL; + } + break; + } + } + + if (box) { + box->phys_id = phys_id; + uncore_box_init(box); + } + } + } + return 0; +} + +static int uncore_cpu_prepare(int cpu, int phys_id) +{ + struct intel_uncore_type *type; + struct intel_uncore_pmu *pmu; + struct intel_uncore_box *box; + int i, j; + + for (i = 0; msr_uncores[i]; i++) { + type = msr_uncores[i]; + for (j = 0; j < type->num_boxes; j++) { + pmu = &type->pmus[j]; + if (pmu->func_id < 0) + pmu->func_id = j; + + box = uncore_alloc_box(type, cpu_to_node(cpu)); + if (!box) + return -ENOMEM; + + box->pmu = pmu; + box->phys_id = phys_id; + *per_cpu_ptr(pmu->box, cpu) = box; + } + } + return 0; +} + +static void +uncore_change_context(struct intel_uncore_type **uncores, int old_cpu, int new_cpu) +{ + struct intel_uncore_type *type; + struct intel_uncore_pmu *pmu; + struct intel_uncore_box *box; + int i, j; + + for (i = 0; uncores[i]; i++) { + type = uncores[i]; + for (j = 0; j < type->num_boxes; j++) { + pmu = &type->pmus[j]; + if (old_cpu < 0) + box = uncore_pmu_to_box(pmu, new_cpu); + else + box = uncore_pmu_to_box(pmu, old_cpu); + if (!box) + continue; + + if (old_cpu < 0) { + WARN_ON_ONCE(box->cpu != -1); + box->cpu = new_cpu; + continue; + } + + WARN_ON_ONCE(box->cpu != old_cpu); + if (new_cpu >= 0) { + uncore_pmu_cancel_hrtimer(box); + perf_pmu_migrate_context(&pmu->pmu, + old_cpu, new_cpu); + box->cpu = new_cpu; + } else { + box->cpu = -1; + } + } + } +} + +static void uncore_event_exit_cpu(int cpu) +{ + int i, phys_id, target; + + /* if exiting cpu is used for collecting uncore events */ + if (!cpumask_test_and_clear_cpu(cpu, &uncore_cpu_mask)) + return; + + /* find a new cpu to collect uncore events */ + phys_id = topology_physical_package_id(cpu); + target = -1; + for_each_online_cpu(i) { + if (i == cpu) + continue; + if (phys_id == topology_physical_package_id(i)) { + target = i; + break; + } + } + + /* migrate uncore events to the new cpu */ + if (target >= 0) + cpumask_set_cpu(target, &uncore_cpu_mask); + + uncore_change_context(msr_uncores, cpu, target); + uncore_change_context(pci_uncores, cpu, target); +} + +static void uncore_event_init_cpu(int cpu) +{ + int i, phys_id; + + phys_id = topology_physical_package_id(cpu); + for_each_cpu(i, &uncore_cpu_mask) { + if (phys_id == topology_physical_package_id(i)) + return; + } + + cpumask_set_cpu(cpu, &uncore_cpu_mask); + + uncore_change_context(msr_uncores, -1, cpu); + uncore_change_context(pci_uncores, -1, cpu); +} + +static int uncore_cpu_notifier(struct notifier_block *self, + unsigned long action, void *hcpu) +{ + unsigned int cpu = (long)hcpu; + + /* allocate/free data structure for uncore box */ + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_UP_PREPARE: + uncore_cpu_prepare(cpu, -1); + break; + case CPU_STARTING: + uncore_cpu_starting(cpu); + break; + case CPU_UP_CANCELED: + case CPU_DYING: + uncore_cpu_dying(cpu); + break; + case CPU_ONLINE: + case CPU_DEAD: + uncore_kfree_boxes(); + break; + default: + break; + } + + /* select the cpu that collects uncore events */ + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_DOWN_FAILED: + case CPU_STARTING: + uncore_event_init_cpu(cpu); + break; + case CPU_DOWN_PREPARE: + uncore_event_exit_cpu(cpu); + break; + default: + break; + } + + return NOTIFY_OK; +} + +static struct notifier_block uncore_cpu_nb = { + .notifier_call = uncore_cpu_notifier, + /* + * to migrate uncore events, our notifier should be executed + * before perf core's notifier. + */ + .priority = CPU_PRI_PERF + 1, +}; + +static void __init uncore_cpu_setup(void *dummy) +{ + uncore_cpu_starting(smp_processor_id()); +} + +static int __init uncore_cpu_init(void) +{ + int ret, max_cores; + + max_cores = boot_cpu_data.x86_max_cores; + switch (boot_cpu_data.x86_model) { + case 26: /* Nehalem */ + case 30: + case 37: /* Westmere */ + case 44: + msr_uncores = nhm_msr_uncores; + break; + case 42: /* Sandy Bridge */ + case 58: /* Ivy Bridge */ + if (snb_uncore_cbox.num_boxes > max_cores) + snb_uncore_cbox.num_boxes = max_cores; + msr_uncores = snb_msr_uncores; + break; + case 45: /* Sandy Bridge-EP */ + if (snbep_uncore_cbox.num_boxes > max_cores) + snbep_uncore_cbox.num_boxes = max_cores; + msr_uncores = snbep_msr_uncores; + break; + case 46: /* Nehalem-EX */ + uncore_nhmex = true; + case 47: /* Westmere-EX aka. Xeon E7 */ + if (!uncore_nhmex) + nhmex_uncore_mbox.event_descs = wsmex_uncore_mbox_events; + if (nhmex_uncore_cbox.num_boxes > max_cores) + nhmex_uncore_cbox.num_boxes = max_cores; + msr_uncores = nhmex_msr_uncores; + break; + case 62: /* IvyTown */ + if (ivt_uncore_cbox.num_boxes > max_cores) + ivt_uncore_cbox.num_boxes = max_cores; + msr_uncores = ivt_msr_uncores; + break; + + default: + return 0; + } + + ret = uncore_types_init(msr_uncores); + if (ret) + return ret; + + return 0; +} + +static int __init uncore_pmus_register(void) +{ + struct intel_uncore_pmu *pmu; + struct intel_uncore_type *type; + int i, j; + + for (i = 0; msr_uncores[i]; i++) { + type = msr_uncores[i]; + for (j = 0; j < type->num_boxes; j++) { + pmu = &type->pmus[j]; + uncore_pmu_register(pmu); + } + } + + for (i = 0; pci_uncores[i]; i++) { + type = pci_uncores[i]; + for (j = 0; j < type->num_boxes; j++) { + pmu = &type->pmus[j]; + uncore_pmu_register(pmu); + } + } + + return 0; +} + +static void __init uncore_cpumask_init(void) +{ + int cpu; + + /* + * ony invoke once from msr or pci init code + */ + if (!cpumask_empty(&uncore_cpu_mask)) + return; + + cpu_notifier_register_begin(); + + for_each_online_cpu(cpu) { + int i, phys_id = topology_physical_package_id(cpu); + + for_each_cpu(i, &uncore_cpu_mask) { + if (phys_id == topology_physical_package_id(i)) { + phys_id = -1; + break; + } + } + if (phys_id < 0) + continue; + + uncore_cpu_prepare(cpu, phys_id); + uncore_event_init_cpu(cpu); + } + on_each_cpu(uncore_cpu_setup, NULL, 1); + + __register_cpu_notifier(&uncore_cpu_nb); + + cpu_notifier_register_done(); +} + + +static int __init intel_uncore_init(void) +{ + int ret; + + if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL) + return -ENODEV; + + if (cpu_has_hypervisor) + return -ENODEV; + + ret = uncore_pci_init(); + if (ret) + goto fail; + ret = uncore_cpu_init(); + if (ret) { + uncore_pci_exit(); + goto fail; + } + uncore_cpumask_init(); + + uncore_pmus_register(); + return 0; +fail: + return ret; +} +device_initcall(intel_uncore_init); diff --git a/arch/x86/kernel/cpu/perf_event_intel_uncore.h b/arch/x86/kernel/cpu/perf_event_intel_uncore.h new file mode 100644 index 00000000000..90236f0c94a --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_intel_uncore.h @@ -0,0 +1,696 @@ +#include <linux/module.h> +#include <linux/slab.h> +#include <linux/pci.h> +#include <linux/perf_event.h> +#include "perf_event.h" + +#define UNCORE_PMU_NAME_LEN 32 +#define UNCORE_PMU_HRTIMER_INTERVAL (60LL * NSEC_PER_SEC) +#define UNCORE_SNB_IMC_HRTIMER_INTERVAL (5ULL * NSEC_PER_SEC) + +#define UNCORE_FIXED_EVENT 0xff +#define UNCORE_PMC_IDX_MAX_GENERIC 8 +#define UNCORE_PMC_IDX_FIXED UNCORE_PMC_IDX_MAX_GENERIC +#define UNCORE_PMC_IDX_MAX (UNCORE_PMC_IDX_FIXED + 1) + +#define UNCORE_PCI_DEV_DATA(type, idx) ((type << 8) | idx) +#define UNCORE_PCI_DEV_TYPE(data) ((data >> 8) & 0xff) +#define UNCORE_PCI_DEV_IDX(data) (data & 0xff) +#define UNCORE_EXTRA_PCI_DEV 0xff +#define UNCORE_EXTRA_PCI_DEV_MAX 2 + +/* support up to 8 sockets */ +#define UNCORE_SOCKET_MAX 8 + +#define UNCORE_EVENT_CONSTRAINT(c, n) EVENT_CONSTRAINT(c, n, 0xff) + +/* SNB event control */ +#define SNB_UNC_CTL_EV_SEL_MASK 0x000000ff +#define SNB_UNC_CTL_UMASK_MASK 0x0000ff00 +#define SNB_UNC_CTL_EDGE_DET (1 << 18) +#define SNB_UNC_CTL_EN (1 << 22) +#define SNB_UNC_CTL_INVERT (1 << 23) +#define SNB_UNC_CTL_CMASK_MASK 0x1f000000 +#define NHM_UNC_CTL_CMASK_MASK 0xff000000 +#define NHM_UNC_FIXED_CTR_CTL_EN (1 << 0) + +#define SNB_UNC_RAW_EVENT_MASK (SNB_UNC_CTL_EV_SEL_MASK | \ + SNB_UNC_CTL_UMASK_MASK | \ + SNB_UNC_CTL_EDGE_DET | \ + SNB_UNC_CTL_INVERT | \ + SNB_UNC_CTL_CMASK_MASK) + +#define NHM_UNC_RAW_EVENT_MASK (SNB_UNC_CTL_EV_SEL_MASK | \ + SNB_UNC_CTL_UMASK_MASK | \ + SNB_UNC_CTL_EDGE_DET | \ + SNB_UNC_CTL_INVERT | \ + NHM_UNC_CTL_CMASK_MASK) + +/* SNB global control register */ +#define SNB_UNC_PERF_GLOBAL_CTL 0x391 +#define SNB_UNC_FIXED_CTR_CTRL 0x394 +#define SNB_UNC_FIXED_CTR 0x395 + +/* SNB uncore global control */ +#define SNB_UNC_GLOBAL_CTL_CORE_ALL ((1 << 4) - 1) +#define SNB_UNC_GLOBAL_CTL_EN (1 << 29) + +/* SNB Cbo register */ +#define SNB_UNC_CBO_0_PERFEVTSEL0 0x700 +#define SNB_UNC_CBO_0_PER_CTR0 0x706 +#define SNB_UNC_CBO_MSR_OFFSET 0x10 + +/* NHM global control register */ +#define NHM_UNC_PERF_GLOBAL_CTL 0x391 +#define NHM_UNC_FIXED_CTR 0x394 +#define NHM_UNC_FIXED_CTR_CTRL 0x395 + +/* NHM uncore global control */ +#define NHM_UNC_GLOBAL_CTL_EN_PC_ALL ((1ULL << 8) - 1) +#define NHM_UNC_GLOBAL_CTL_EN_FC (1ULL << 32) + +/* NHM uncore register */ +#define NHM_UNC_PERFEVTSEL0 0x3c0 +#define NHM_UNC_UNCORE_PMC0 0x3b0 + +/* SNB-EP Box level control */ +#define SNBEP_PMON_BOX_CTL_RST_CTRL (1 << 0) +#define SNBEP_PMON_BOX_CTL_RST_CTRS (1 << 1) +#define SNBEP_PMON_BOX_CTL_FRZ (1 << 8) +#define SNBEP_PMON_BOX_CTL_FRZ_EN (1 << 16) +#define SNBEP_PMON_BOX_CTL_INT (SNBEP_PMON_BOX_CTL_RST_CTRL | \ + SNBEP_PMON_BOX_CTL_RST_CTRS | \ + SNBEP_PMON_BOX_CTL_FRZ_EN) +/* SNB-EP event control */ +#define SNBEP_PMON_CTL_EV_SEL_MASK 0x000000ff +#define SNBEP_PMON_CTL_UMASK_MASK 0x0000ff00 +#define SNBEP_PMON_CTL_RST (1 << 17) +#define SNBEP_PMON_CTL_EDGE_DET (1 << 18) +#define SNBEP_PMON_CTL_EV_SEL_EXT (1 << 21) +#define SNBEP_PMON_CTL_EN (1 << 22) +#define SNBEP_PMON_CTL_INVERT (1 << 23) +#define SNBEP_PMON_CTL_TRESH_MASK 0xff000000 +#define SNBEP_PMON_RAW_EVENT_MASK (SNBEP_PMON_CTL_EV_SEL_MASK | \ + SNBEP_PMON_CTL_UMASK_MASK | \ + SNBEP_PMON_CTL_EDGE_DET | \ + SNBEP_PMON_CTL_INVERT | \ + SNBEP_PMON_CTL_TRESH_MASK) + +/* SNB-EP Ubox event control */ +#define SNBEP_U_MSR_PMON_CTL_TRESH_MASK 0x1f000000 +#define SNBEP_U_MSR_PMON_RAW_EVENT_MASK \ + (SNBEP_PMON_CTL_EV_SEL_MASK | \ + SNBEP_PMON_CTL_UMASK_MASK | \ + SNBEP_PMON_CTL_EDGE_DET | \ + SNBEP_PMON_CTL_INVERT | \ + SNBEP_U_MSR_PMON_CTL_TRESH_MASK) + +#define SNBEP_CBO_PMON_CTL_TID_EN (1 << 19) +#define SNBEP_CBO_MSR_PMON_RAW_EVENT_MASK (SNBEP_PMON_RAW_EVENT_MASK | \ + SNBEP_CBO_PMON_CTL_TID_EN) + +/* SNB-EP PCU event control */ +#define SNBEP_PCU_MSR_PMON_CTL_OCC_SEL_MASK 0x0000c000 +#define SNBEP_PCU_MSR_PMON_CTL_TRESH_MASK 0x1f000000 +#define SNBEP_PCU_MSR_PMON_CTL_OCC_INVERT (1 << 30) +#define SNBEP_PCU_MSR_PMON_CTL_OCC_EDGE_DET (1 << 31) +#define SNBEP_PCU_MSR_PMON_RAW_EVENT_MASK \ + (SNBEP_PMON_CTL_EV_SEL_MASK | \ + SNBEP_PCU_MSR_PMON_CTL_OCC_SEL_MASK | \ + SNBEP_PMON_CTL_EDGE_DET | \ + SNBEP_PMON_CTL_EV_SEL_EXT | \ + SNBEP_PMON_CTL_INVERT | \ + SNBEP_PCU_MSR_PMON_CTL_TRESH_MASK | \ + SNBEP_PCU_MSR_PMON_CTL_OCC_INVERT | \ + SNBEP_PCU_MSR_PMON_CTL_OCC_EDGE_DET) + +#define SNBEP_QPI_PCI_PMON_RAW_EVENT_MASK \ + (SNBEP_PMON_RAW_EVENT_MASK | \ + SNBEP_PMON_CTL_EV_SEL_EXT) + +/* SNB-EP pci control register */ +#define SNBEP_PCI_PMON_BOX_CTL 0xf4 +#define SNBEP_PCI_PMON_CTL0 0xd8 +/* SNB-EP pci counter register */ +#define SNBEP_PCI_PMON_CTR0 0xa0 + +/* SNB-EP home agent register */ +#define SNBEP_HA_PCI_PMON_BOX_ADDRMATCH0 0x40 +#define SNBEP_HA_PCI_PMON_BOX_ADDRMATCH1 0x44 +#define SNBEP_HA_PCI_PMON_BOX_OPCODEMATCH 0x48 +/* SNB-EP memory controller register */ +#define SNBEP_MC_CHy_PCI_PMON_FIXED_CTL 0xf0 +#define SNBEP_MC_CHy_PCI_PMON_FIXED_CTR 0xd0 +/* SNB-EP QPI register */ +#define SNBEP_Q_Py_PCI_PMON_PKT_MATCH0 0x228 +#define SNBEP_Q_Py_PCI_PMON_PKT_MATCH1 0x22c +#define SNBEP_Q_Py_PCI_PMON_PKT_MASK0 0x238 +#define SNBEP_Q_Py_PCI_PMON_PKT_MASK1 0x23c + +/* SNB-EP Ubox register */ +#define SNBEP_U_MSR_PMON_CTR0 0xc16 +#define SNBEP_U_MSR_PMON_CTL0 0xc10 + +#define SNBEP_U_MSR_PMON_UCLK_FIXED_CTL 0xc08 +#define SNBEP_U_MSR_PMON_UCLK_FIXED_CTR 0xc09 + +/* SNB-EP Cbo register */ +#define SNBEP_C0_MSR_PMON_CTR0 0xd16 +#define SNBEP_C0_MSR_PMON_CTL0 0xd10 +#define SNBEP_C0_MSR_PMON_BOX_CTL 0xd04 +#define SNBEP_C0_MSR_PMON_BOX_FILTER 0xd14 +#define SNBEP_CBO_MSR_OFFSET 0x20 + +#define SNBEP_CB0_MSR_PMON_BOX_FILTER_TID 0x1f +#define SNBEP_CB0_MSR_PMON_BOX_FILTER_NID 0x3fc00 +#define SNBEP_CB0_MSR_PMON_BOX_FILTER_STATE 0x7c0000 +#define SNBEP_CB0_MSR_PMON_BOX_FILTER_OPC 0xff800000 + +#define SNBEP_CBO_EVENT_EXTRA_REG(e, m, i) { \ + .event = (e), \ + .msr = SNBEP_C0_MSR_PMON_BOX_FILTER, \ + .config_mask = (m), \ + .idx = (i) \ +} + +/* SNB-EP PCU register */ +#define SNBEP_PCU_MSR_PMON_CTR0 0xc36 +#define SNBEP_PCU_MSR_PMON_CTL0 0xc30 +#define SNBEP_PCU_MSR_PMON_BOX_CTL 0xc24 +#define SNBEP_PCU_MSR_PMON_BOX_FILTER 0xc34 +#define SNBEP_PCU_MSR_PMON_BOX_FILTER_MASK 0xffffffff +#define SNBEP_PCU_MSR_CORE_C3_CTR 0x3fc +#define SNBEP_PCU_MSR_CORE_C6_CTR 0x3fd + +/* IVT event control */ +#define IVT_PMON_BOX_CTL_INT (SNBEP_PMON_BOX_CTL_RST_CTRL | \ + SNBEP_PMON_BOX_CTL_RST_CTRS) +#define IVT_PMON_RAW_EVENT_MASK (SNBEP_PMON_CTL_EV_SEL_MASK | \ + SNBEP_PMON_CTL_UMASK_MASK | \ + SNBEP_PMON_CTL_EDGE_DET | \ + SNBEP_PMON_CTL_TRESH_MASK) +/* IVT Ubox */ +#define IVT_U_MSR_PMON_GLOBAL_CTL 0xc00 +#define IVT_U_PMON_GLOBAL_FRZ_ALL (1 << 31) +#define IVT_U_PMON_GLOBAL_UNFRZ_ALL (1 << 29) + +#define IVT_U_MSR_PMON_RAW_EVENT_MASK \ + (SNBEP_PMON_CTL_EV_SEL_MASK | \ + SNBEP_PMON_CTL_UMASK_MASK | \ + SNBEP_PMON_CTL_EDGE_DET | \ + SNBEP_U_MSR_PMON_CTL_TRESH_MASK) +/* IVT Cbo */ +#define IVT_CBO_MSR_PMON_RAW_EVENT_MASK (IVT_PMON_RAW_EVENT_MASK | \ + SNBEP_CBO_PMON_CTL_TID_EN) + +#define IVT_CB0_MSR_PMON_BOX_FILTER_TID (0x1fULL << 0) +#define IVT_CB0_MSR_PMON_BOX_FILTER_LINK (0xfULL << 5) +#define IVT_CB0_MSR_PMON_BOX_FILTER_STATE (0x3fULL << 17) +#define IVT_CB0_MSR_PMON_BOX_FILTER_NID (0xffffULL << 32) +#define IVT_CB0_MSR_PMON_BOX_FILTER_OPC (0x1ffULL << 52) +#define IVT_CB0_MSR_PMON_BOX_FILTER_C6 (0x1ULL << 61) +#define IVT_CB0_MSR_PMON_BOX_FILTER_NC (0x1ULL << 62) +#define IVT_CB0_MSR_PMON_BOX_FILTER_IOSC (0x1ULL << 63) + +/* IVT home agent */ +#define IVT_HA_PCI_PMON_CTL_Q_OCC_RST (1 << 16) +#define IVT_HA_PCI_PMON_RAW_EVENT_MASK \ + (IVT_PMON_RAW_EVENT_MASK | \ + IVT_HA_PCI_PMON_CTL_Q_OCC_RST) +/* IVT PCU */ +#define IVT_PCU_MSR_PMON_RAW_EVENT_MASK \ + (SNBEP_PMON_CTL_EV_SEL_MASK | \ + SNBEP_PMON_CTL_EV_SEL_EXT | \ + SNBEP_PCU_MSR_PMON_CTL_OCC_SEL_MASK | \ + SNBEP_PMON_CTL_EDGE_DET | \ + SNBEP_PCU_MSR_PMON_CTL_TRESH_MASK | \ + SNBEP_PCU_MSR_PMON_CTL_OCC_INVERT | \ + SNBEP_PCU_MSR_PMON_CTL_OCC_EDGE_DET) +/* IVT QPI */ +#define IVT_QPI_PCI_PMON_RAW_EVENT_MASK \ + (IVT_PMON_RAW_EVENT_MASK | \ + SNBEP_PMON_CTL_EV_SEL_EXT) + +/* NHM-EX event control */ +#define NHMEX_PMON_CTL_EV_SEL_MASK 0x000000ff +#define NHMEX_PMON_CTL_UMASK_MASK 0x0000ff00 +#define NHMEX_PMON_CTL_EN_BIT0 (1 << 0) +#define NHMEX_PMON_CTL_EDGE_DET (1 << 18) +#define NHMEX_PMON_CTL_PMI_EN (1 << 20) +#define NHMEX_PMON_CTL_EN_BIT22 (1 << 22) +#define NHMEX_PMON_CTL_INVERT (1 << 23) +#define NHMEX_PMON_CTL_TRESH_MASK 0xff000000 +#define NHMEX_PMON_RAW_EVENT_MASK (NHMEX_PMON_CTL_EV_SEL_MASK | \ + NHMEX_PMON_CTL_UMASK_MASK | \ + NHMEX_PMON_CTL_EDGE_DET | \ + NHMEX_PMON_CTL_INVERT | \ + NHMEX_PMON_CTL_TRESH_MASK) + +/* NHM-EX Ubox */ +#define NHMEX_U_MSR_PMON_GLOBAL_CTL 0xc00 +#define NHMEX_U_MSR_PMON_CTR 0xc11 +#define NHMEX_U_MSR_PMON_EV_SEL 0xc10 + +#define NHMEX_U_PMON_GLOBAL_EN (1 << 0) +#define NHMEX_U_PMON_GLOBAL_PMI_CORE_SEL 0x0000001e +#define NHMEX_U_PMON_GLOBAL_EN_ALL (1 << 28) +#define NHMEX_U_PMON_GLOBAL_RST_ALL (1 << 29) +#define NHMEX_U_PMON_GLOBAL_FRZ_ALL (1 << 31) + +#define NHMEX_U_PMON_RAW_EVENT_MASK \ + (NHMEX_PMON_CTL_EV_SEL_MASK | \ + NHMEX_PMON_CTL_EDGE_DET) + +/* NHM-EX Cbox */ +#define NHMEX_C0_MSR_PMON_GLOBAL_CTL 0xd00 +#define NHMEX_C0_MSR_PMON_CTR0 0xd11 +#define NHMEX_C0_MSR_PMON_EV_SEL0 0xd10 +#define NHMEX_C_MSR_OFFSET 0x20 + +/* NHM-EX Bbox */ +#define NHMEX_B0_MSR_PMON_GLOBAL_CTL 0xc20 +#define NHMEX_B0_MSR_PMON_CTR0 0xc31 +#define NHMEX_B0_MSR_PMON_CTL0 0xc30 +#define NHMEX_B_MSR_OFFSET 0x40 +#define NHMEX_B0_MSR_MATCH 0xe45 +#define NHMEX_B0_MSR_MASK 0xe46 +#define NHMEX_B1_MSR_MATCH 0xe4d +#define NHMEX_B1_MSR_MASK 0xe4e + +#define NHMEX_B_PMON_CTL_EN (1 << 0) +#define NHMEX_B_PMON_CTL_EV_SEL_SHIFT 1 +#define NHMEX_B_PMON_CTL_EV_SEL_MASK \ + (0x1f << NHMEX_B_PMON_CTL_EV_SEL_SHIFT) +#define NHMEX_B_PMON_CTR_SHIFT 6 +#define NHMEX_B_PMON_CTR_MASK \ + (0x3 << NHMEX_B_PMON_CTR_SHIFT) +#define NHMEX_B_PMON_RAW_EVENT_MASK \ + (NHMEX_B_PMON_CTL_EV_SEL_MASK | \ + NHMEX_B_PMON_CTR_MASK) + +/* NHM-EX Sbox */ +#define NHMEX_S0_MSR_PMON_GLOBAL_CTL 0xc40 +#define NHMEX_S0_MSR_PMON_CTR0 0xc51 +#define NHMEX_S0_MSR_PMON_CTL0 0xc50 +#define NHMEX_S_MSR_OFFSET 0x80 +#define NHMEX_S0_MSR_MM_CFG 0xe48 +#define NHMEX_S0_MSR_MATCH 0xe49 +#define NHMEX_S0_MSR_MASK 0xe4a +#define NHMEX_S1_MSR_MM_CFG 0xe58 +#define NHMEX_S1_MSR_MATCH 0xe59 +#define NHMEX_S1_MSR_MASK 0xe5a + +#define NHMEX_S_PMON_MM_CFG_EN (0x1ULL << 63) +#define NHMEX_S_EVENT_TO_R_PROG_EV 0 + +/* NHM-EX Mbox */ +#define NHMEX_M0_MSR_GLOBAL_CTL 0xca0 +#define NHMEX_M0_MSR_PMU_DSP 0xca5 +#define NHMEX_M0_MSR_PMU_ISS 0xca6 +#define NHMEX_M0_MSR_PMU_MAP 0xca7 +#define NHMEX_M0_MSR_PMU_MSC_THR 0xca8 +#define NHMEX_M0_MSR_PMU_PGT 0xca9 +#define NHMEX_M0_MSR_PMU_PLD 0xcaa +#define NHMEX_M0_MSR_PMU_ZDP_CTL_FVC 0xcab +#define NHMEX_M0_MSR_PMU_CTL0 0xcb0 +#define NHMEX_M0_MSR_PMU_CNT0 0xcb1 +#define NHMEX_M_MSR_OFFSET 0x40 +#define NHMEX_M0_MSR_PMU_MM_CFG 0xe54 +#define NHMEX_M1_MSR_PMU_MM_CFG 0xe5c + +#define NHMEX_M_PMON_MM_CFG_EN (1ULL << 63) +#define NHMEX_M_PMON_ADDR_MATCH_MASK 0x3ffffffffULL +#define NHMEX_M_PMON_ADDR_MASK_MASK 0x7ffffffULL +#define NHMEX_M_PMON_ADDR_MASK_SHIFT 34 + +#define NHMEX_M_PMON_CTL_EN (1 << 0) +#define NHMEX_M_PMON_CTL_PMI_EN (1 << 1) +#define NHMEX_M_PMON_CTL_COUNT_MODE_SHIFT 2 +#define NHMEX_M_PMON_CTL_COUNT_MODE_MASK \ + (0x3 << NHMEX_M_PMON_CTL_COUNT_MODE_SHIFT) +#define NHMEX_M_PMON_CTL_STORAGE_MODE_SHIFT 4 +#define NHMEX_M_PMON_CTL_STORAGE_MODE_MASK \ + (0x3 << NHMEX_M_PMON_CTL_STORAGE_MODE_SHIFT) +#define NHMEX_M_PMON_CTL_WRAP_MODE (1 << 6) +#define NHMEX_M_PMON_CTL_FLAG_MODE (1 << 7) +#define NHMEX_M_PMON_CTL_INC_SEL_SHIFT 9 +#define NHMEX_M_PMON_CTL_INC_SEL_MASK \ + (0x1f << NHMEX_M_PMON_CTL_INC_SEL_SHIFT) +#define NHMEX_M_PMON_CTL_SET_FLAG_SEL_SHIFT 19 +#define NHMEX_M_PMON_CTL_SET_FLAG_SEL_MASK \ + (0x7 << NHMEX_M_PMON_CTL_SET_FLAG_SEL_SHIFT) +#define NHMEX_M_PMON_RAW_EVENT_MASK \ + (NHMEX_M_PMON_CTL_COUNT_MODE_MASK | \ + NHMEX_M_PMON_CTL_STORAGE_MODE_MASK | \ + NHMEX_M_PMON_CTL_WRAP_MODE | \ + NHMEX_M_PMON_CTL_FLAG_MODE | \ + NHMEX_M_PMON_CTL_INC_SEL_MASK | \ + NHMEX_M_PMON_CTL_SET_FLAG_SEL_MASK) + +#define NHMEX_M_PMON_ZDP_CTL_FVC_MASK (((1 << 11) - 1) | (1 << 23)) +#define NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(n) (0x7ULL << (11 + 3 * (n))) + +#define WSMEX_M_PMON_ZDP_CTL_FVC_MASK (((1 << 12) - 1) | (1 << 24)) +#define WSMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(n) (0x7ULL << (12 + 3 * (n))) + +/* + * use the 9~13 bits to select event If the 7th bit is not set, + * otherwise use the 19~21 bits to select event. + */ +#define MBOX_INC_SEL(x) ((x) << NHMEX_M_PMON_CTL_INC_SEL_SHIFT) +#define MBOX_SET_FLAG_SEL(x) (((x) << NHMEX_M_PMON_CTL_SET_FLAG_SEL_SHIFT) | \ + NHMEX_M_PMON_CTL_FLAG_MODE) +#define MBOX_INC_SEL_MASK (NHMEX_M_PMON_CTL_INC_SEL_MASK | \ + NHMEX_M_PMON_CTL_FLAG_MODE) +#define MBOX_SET_FLAG_SEL_MASK (NHMEX_M_PMON_CTL_SET_FLAG_SEL_MASK | \ + NHMEX_M_PMON_CTL_FLAG_MODE) +#define MBOX_INC_SEL_EXTAR_REG(c, r) \ + EVENT_EXTRA_REG(MBOX_INC_SEL(c), NHMEX_M0_MSR_PMU_##r, \ + MBOX_INC_SEL_MASK, (u64)-1, NHMEX_M_##r) +#define MBOX_SET_FLAG_SEL_EXTRA_REG(c, r) \ + EVENT_EXTRA_REG(MBOX_SET_FLAG_SEL(c), NHMEX_M0_MSR_PMU_##r, \ + MBOX_SET_FLAG_SEL_MASK, \ + (u64)-1, NHMEX_M_##r) + +/* NHM-EX Rbox */ +#define NHMEX_R_MSR_GLOBAL_CTL 0xe00 +#define NHMEX_R_MSR_PMON_CTL0 0xe10 +#define NHMEX_R_MSR_PMON_CNT0 0xe11 +#define NHMEX_R_MSR_OFFSET 0x20 + +#define NHMEX_R_MSR_PORTN_QLX_CFG(n) \ + ((n) < 4 ? (0xe0c + (n)) : (0xe2c + (n) - 4)) +#define NHMEX_R_MSR_PORTN_IPERF_CFG0(n) (0xe04 + (n)) +#define NHMEX_R_MSR_PORTN_IPERF_CFG1(n) (0xe24 + (n)) +#define NHMEX_R_MSR_PORTN_XBR_OFFSET(n) \ + (((n) < 4 ? 0 : 0x10) + (n) * 4) +#define NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(n) \ + (0xe60 + NHMEX_R_MSR_PORTN_XBR_OFFSET(n)) +#define NHMEX_R_MSR_PORTN_XBR_SET1_MATCH(n) \ + (NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(n) + 1) +#define NHMEX_R_MSR_PORTN_XBR_SET1_MASK(n) \ + (NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(n) + 2) +#define NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(n) \ + (0xe70 + NHMEX_R_MSR_PORTN_XBR_OFFSET(n)) +#define NHMEX_R_MSR_PORTN_XBR_SET2_MATCH(n) \ + (NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(n) + 1) +#define NHMEX_R_MSR_PORTN_XBR_SET2_MASK(n) \ + (NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(n) + 2) + +#define NHMEX_R_PMON_CTL_EN (1 << 0) +#define NHMEX_R_PMON_CTL_EV_SEL_SHIFT 1 +#define NHMEX_R_PMON_CTL_EV_SEL_MASK \ + (0x1f << NHMEX_R_PMON_CTL_EV_SEL_SHIFT) +#define NHMEX_R_PMON_CTL_PMI_EN (1 << 6) +#define NHMEX_R_PMON_RAW_EVENT_MASK NHMEX_R_PMON_CTL_EV_SEL_MASK + +/* NHM-EX Wbox */ +#define NHMEX_W_MSR_GLOBAL_CTL 0xc80 +#define NHMEX_W_MSR_PMON_CNT0 0xc90 +#define NHMEX_W_MSR_PMON_EVT_SEL0 0xc91 +#define NHMEX_W_MSR_PMON_FIXED_CTR 0x394 +#define NHMEX_W_MSR_PMON_FIXED_CTL 0x395 + +#define NHMEX_W_PMON_GLOBAL_FIXED_EN (1ULL << 31) + +struct intel_uncore_ops; +struct intel_uncore_pmu; +struct intel_uncore_box; +struct uncore_event_desc; + +struct intel_uncore_type { + const char *name; + int num_counters; + int num_boxes; + int perf_ctr_bits; + int fixed_ctr_bits; + unsigned perf_ctr; + unsigned event_ctl; + unsigned event_mask; + unsigned fixed_ctr; + unsigned fixed_ctl; + unsigned box_ctl; + unsigned msr_offset; + unsigned num_shared_regs:8; + unsigned single_fixed:1; + unsigned pair_ctr_ctl:1; + unsigned *msr_offsets; + struct event_constraint unconstrainted; + struct event_constraint *constraints; + struct intel_uncore_pmu *pmus; + struct intel_uncore_ops *ops; + struct uncore_event_desc *event_descs; + const struct attribute_group *attr_groups[4]; + struct pmu *pmu; /* for custom pmu ops */ +}; + +#define pmu_group attr_groups[0] +#define format_group attr_groups[1] +#define events_group attr_groups[2] + +struct intel_uncore_ops { + void (*init_box)(struct intel_uncore_box *); + void (*disable_box)(struct intel_uncore_box *); + void (*enable_box)(struct intel_uncore_box *); + void (*disable_event)(struct intel_uncore_box *, struct perf_event *); + void (*enable_event)(struct intel_uncore_box *, struct perf_event *); + u64 (*read_counter)(struct intel_uncore_box *, struct perf_event *); + int (*hw_config)(struct intel_uncore_box *, struct perf_event *); + struct event_constraint *(*get_constraint)(struct intel_uncore_box *, + struct perf_event *); + void (*put_constraint)(struct intel_uncore_box *, struct perf_event *); +}; + +struct intel_uncore_pmu { + struct pmu pmu; + char name[UNCORE_PMU_NAME_LEN]; + int pmu_idx; + int func_id; + struct intel_uncore_type *type; + struct intel_uncore_box ** __percpu box; + struct list_head box_list; +}; + +struct intel_uncore_extra_reg { + raw_spinlock_t lock; + u64 config, config1, config2; + atomic_t ref; +}; + +struct intel_uncore_box { + int phys_id; + int n_active; /* number of active events */ + int n_events; + int cpu; /* cpu to collect events */ + unsigned long flags; + atomic_t refcnt; + struct perf_event *events[UNCORE_PMC_IDX_MAX]; + struct perf_event *event_list[UNCORE_PMC_IDX_MAX]; + unsigned long active_mask[BITS_TO_LONGS(UNCORE_PMC_IDX_MAX)]; + u64 tags[UNCORE_PMC_IDX_MAX]; + struct pci_dev *pci_dev; + struct intel_uncore_pmu *pmu; + u64 hrtimer_duration; /* hrtimer timeout for this box */ + struct hrtimer hrtimer; + struct list_head list; + struct list_head active_list; + void *io_addr; + struct intel_uncore_extra_reg shared_regs[0]; +}; + +#define UNCORE_BOX_FLAG_INITIATED 0 + +struct uncore_event_desc { + struct kobj_attribute attr; + const char *config; +}; + +#define INTEL_UNCORE_EVENT_DESC(_name, _config) \ +{ \ + .attr = __ATTR(_name, 0444, uncore_event_show, NULL), \ + .config = _config, \ +} + +#define DEFINE_UNCORE_FORMAT_ATTR(_var, _name, _format) \ +static ssize_t __uncore_##_var##_show(struct kobject *kobj, \ + struct kobj_attribute *attr, \ + char *page) \ +{ \ + BUILD_BUG_ON(sizeof(_format) >= PAGE_SIZE); \ + return sprintf(page, _format "\n"); \ +} \ +static struct kobj_attribute format_attr_##_var = \ + __ATTR(_name, 0444, __uncore_##_var##_show, NULL) + + +static ssize_t uncore_event_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct uncore_event_desc *event = + container_of(attr, struct uncore_event_desc, attr); + return sprintf(buf, "%s", event->config); +} + +static inline unsigned uncore_pci_box_ctl(struct intel_uncore_box *box) +{ + return box->pmu->type->box_ctl; +} + +static inline unsigned uncore_pci_fixed_ctl(struct intel_uncore_box *box) +{ + return box->pmu->type->fixed_ctl; +} + +static inline unsigned uncore_pci_fixed_ctr(struct intel_uncore_box *box) +{ + return box->pmu->type->fixed_ctr; +} + +static inline +unsigned uncore_pci_event_ctl(struct intel_uncore_box *box, int idx) +{ + return idx * 4 + box->pmu->type->event_ctl; +} + +static inline +unsigned uncore_pci_perf_ctr(struct intel_uncore_box *box, int idx) +{ + return idx * 8 + box->pmu->type->perf_ctr; +} + +static inline unsigned uncore_msr_box_offset(struct intel_uncore_box *box) +{ + struct intel_uncore_pmu *pmu = box->pmu; + return pmu->type->msr_offsets ? + pmu->type->msr_offsets[pmu->pmu_idx] : + pmu->type->msr_offset * pmu->pmu_idx; +} + +static inline unsigned uncore_msr_box_ctl(struct intel_uncore_box *box) +{ + if (!box->pmu->type->box_ctl) + return 0; + return box->pmu->type->box_ctl + uncore_msr_box_offset(box); +} + +static inline unsigned uncore_msr_fixed_ctl(struct intel_uncore_box *box) +{ + if (!box->pmu->type->fixed_ctl) + return 0; + return box->pmu->type->fixed_ctl + uncore_msr_box_offset(box); +} + +static inline unsigned uncore_msr_fixed_ctr(struct intel_uncore_box *box) +{ + return box->pmu->type->fixed_ctr + uncore_msr_box_offset(box); +} + +static inline +unsigned uncore_msr_event_ctl(struct intel_uncore_box *box, int idx) +{ + return box->pmu->type->event_ctl + + (box->pmu->type->pair_ctr_ctl ? 2 * idx : idx) + + uncore_msr_box_offset(box); +} + +static inline +unsigned uncore_msr_perf_ctr(struct intel_uncore_box *box, int idx) +{ + return box->pmu->type->perf_ctr + + (box->pmu->type->pair_ctr_ctl ? 2 * idx : idx) + + uncore_msr_box_offset(box); +} + +static inline +unsigned uncore_fixed_ctl(struct intel_uncore_box *box) +{ + if (box->pci_dev) + return uncore_pci_fixed_ctl(box); + else + return uncore_msr_fixed_ctl(box); +} + +static inline +unsigned uncore_fixed_ctr(struct intel_uncore_box *box) +{ + if (box->pci_dev) + return uncore_pci_fixed_ctr(box); + else + return uncore_msr_fixed_ctr(box); +} + +static inline +unsigned uncore_event_ctl(struct intel_uncore_box *box, int idx) +{ + if (box->pci_dev) + return uncore_pci_event_ctl(box, idx); + else + return uncore_msr_event_ctl(box, idx); +} + +static inline +unsigned uncore_perf_ctr(struct intel_uncore_box *box, int idx) +{ + if (box->pci_dev) + return uncore_pci_perf_ctr(box, idx); + else + return uncore_msr_perf_ctr(box, idx); +} + +static inline int uncore_perf_ctr_bits(struct intel_uncore_box *box) +{ + return box->pmu->type->perf_ctr_bits; +} + +static inline int uncore_fixed_ctr_bits(struct intel_uncore_box *box) +{ + return box->pmu->type->fixed_ctr_bits; +} + +static inline int uncore_num_counters(struct intel_uncore_box *box) +{ + return box->pmu->type->num_counters; +} + +static inline void uncore_disable_box(struct intel_uncore_box *box) +{ + if (box->pmu->type->ops->disable_box) + box->pmu->type->ops->disable_box(box); +} + +static inline void uncore_enable_box(struct intel_uncore_box *box) +{ + if (box->pmu->type->ops->enable_box) + box->pmu->type->ops->enable_box(box); +} + +static inline void uncore_disable_event(struct intel_uncore_box *box, + struct perf_event *event) +{ + box->pmu->type->ops->disable_event(box, event); +} + +static inline void uncore_enable_event(struct intel_uncore_box *box, + struct perf_event *event) +{ + box->pmu->type->ops->enable_event(box, event); +} + +static inline u64 uncore_read_counter(struct intel_uncore_box *box, + struct perf_event *event) +{ + return box->pmu->type->ops->read_counter(box, event); +} + +static inline void uncore_box_init(struct intel_uncore_box *box) +{ + if (!test_and_set_bit(UNCORE_BOX_FLAG_INITIATED, &box->flags)) { + if (box->pmu->type->ops->init_box) + box->pmu->type->ops->init_box(box); + } +} + +static inline bool uncore_box_is_fake(struct intel_uncore_box *box) +{ + return (box->phys_id < 0); +} diff --git a/arch/x86/kernel/cpu/perf_event_knc.c b/arch/x86/kernel/cpu/perf_event_knc.c new file mode 100644 index 00000000000..838fa8772c6 --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_knc.c @@ -0,0 +1,319 @@ +/* Driver for Intel Xeon Phi "Knights Corner" PMU */ + +#include <linux/perf_event.h> +#include <linux/types.h> + +#include <asm/hardirq.h> + +#include "perf_event.h" + +static const u64 knc_perfmon_event_map[] = +{ + [PERF_COUNT_HW_CPU_CYCLES] = 0x002a, + [PERF_COUNT_HW_INSTRUCTIONS] = 0x0016, + [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0028, + [PERF_COUNT_HW_CACHE_MISSES] = 0x0029, + [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x0012, + [PERF_COUNT_HW_BRANCH_MISSES] = 0x002b, +}; + +static const u64 __initconst knc_hw_cache_event_ids + [PERF_COUNT_HW_CACHE_MAX] + [PERF_COUNT_HW_CACHE_OP_MAX] + [PERF_COUNT_HW_CACHE_RESULT_MAX] = +{ + [ C(L1D) ] = { + [ C(OP_READ) ] = { + /* On Xeon Phi event "0" is a valid DATA_READ */ + /* (L1 Data Cache Reads) Instruction. */ + /* We code this as ARCH_PERFMON_EVENTSEL_INT as this */ + /* bit will always be set in x86_pmu_hw_config(). */ + [ C(RESULT_ACCESS) ] = ARCH_PERFMON_EVENTSEL_INT, + /* DATA_READ */ + [ C(RESULT_MISS) ] = 0x0003, /* DATA_READ_MISS */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0x0001, /* DATA_WRITE */ + [ C(RESULT_MISS) ] = 0x0004, /* DATA_WRITE_MISS */ + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0x0011, /* L1_DATA_PF1 */ + [ C(RESULT_MISS) ] = 0x001c, /* L1_DATA_PF1_MISS */ + }, + }, + [ C(L1I ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x000c, /* CODE_READ */ + [ C(RESULT_MISS) ] = 0x000e, /* CODE_CACHE_MISS */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0x0, + [ C(RESULT_MISS) ] = 0x0, + }, + }, + [ C(LL ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0x10cb, /* L2_READ_MISS */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0x10cc, /* L2_WRITE_HIT */ + [ C(RESULT_MISS) ] = 0, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0x10fc, /* L2_DATA_PF2 */ + [ C(RESULT_MISS) ] = 0x10fe, /* L2_DATA_PF2_MISS */ + }, + }, + [ C(DTLB) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = ARCH_PERFMON_EVENTSEL_INT, + /* DATA_READ */ + /* see note on L1 OP_READ */ + [ C(RESULT_MISS) ] = 0x0002, /* DATA_PAGE_WALK */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0x0001, /* DATA_WRITE */ + [ C(RESULT_MISS) ] = 0x0002, /* DATA_PAGE_WALK */ + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0x0, + [ C(RESULT_MISS) ] = 0x0, + }, + }, + [ C(ITLB) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x000c, /* CODE_READ */ + [ C(RESULT_MISS) ] = 0x000d, /* CODE_PAGE_WALK */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + }, + [ C(BPU ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x0012, /* BRANCHES */ + [ C(RESULT_MISS) ] = 0x002b, /* BRANCHES_MISPREDICTED */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + }, +}; + + +static u64 knc_pmu_event_map(int hw_event) +{ + return knc_perfmon_event_map[hw_event]; +} + +static struct event_constraint knc_event_constraints[] = +{ + INTEL_EVENT_CONSTRAINT(0xc3, 0x1), /* HWP_L2HIT */ + INTEL_EVENT_CONSTRAINT(0xc4, 0x1), /* HWP_L2MISS */ + INTEL_EVENT_CONSTRAINT(0xc8, 0x1), /* L2_READ_HIT_E */ + INTEL_EVENT_CONSTRAINT(0xc9, 0x1), /* L2_READ_HIT_M */ + INTEL_EVENT_CONSTRAINT(0xca, 0x1), /* L2_READ_HIT_S */ + INTEL_EVENT_CONSTRAINT(0xcb, 0x1), /* L2_READ_MISS */ + INTEL_EVENT_CONSTRAINT(0xcc, 0x1), /* L2_WRITE_HIT */ + INTEL_EVENT_CONSTRAINT(0xce, 0x1), /* L2_STRONGLY_ORDERED_STREAMING_VSTORES_MISS */ + INTEL_EVENT_CONSTRAINT(0xcf, 0x1), /* L2_WEAKLY_ORDERED_STREAMING_VSTORE_MISS */ + INTEL_EVENT_CONSTRAINT(0xd7, 0x1), /* L2_VICTIM_REQ_WITH_DATA */ + INTEL_EVENT_CONSTRAINT(0xe3, 0x1), /* SNP_HITM_BUNIT */ + INTEL_EVENT_CONSTRAINT(0xe6, 0x1), /* SNP_HIT_L2 */ + INTEL_EVENT_CONSTRAINT(0xe7, 0x1), /* SNP_HITM_L2 */ + INTEL_EVENT_CONSTRAINT(0xf1, 0x1), /* L2_DATA_READ_MISS_CACHE_FILL */ + INTEL_EVENT_CONSTRAINT(0xf2, 0x1), /* L2_DATA_WRITE_MISS_CACHE_FILL */ + INTEL_EVENT_CONSTRAINT(0xf6, 0x1), /* L2_DATA_READ_MISS_MEM_FILL */ + INTEL_EVENT_CONSTRAINT(0xf7, 0x1), /* L2_DATA_WRITE_MISS_MEM_FILL */ + INTEL_EVENT_CONSTRAINT(0xfc, 0x1), /* L2_DATA_PF2 */ + INTEL_EVENT_CONSTRAINT(0xfd, 0x1), /* L2_DATA_PF2_DROP */ + INTEL_EVENT_CONSTRAINT(0xfe, 0x1), /* L2_DATA_PF2_MISS */ + INTEL_EVENT_CONSTRAINT(0xff, 0x1), /* L2_DATA_HIT_INFLIGHT_PF2 */ + EVENT_CONSTRAINT_END +}; + +#define MSR_KNC_IA32_PERF_GLOBAL_STATUS 0x0000002d +#define MSR_KNC_IA32_PERF_GLOBAL_OVF_CONTROL 0x0000002e +#define MSR_KNC_IA32_PERF_GLOBAL_CTRL 0x0000002f + +#define KNC_ENABLE_COUNTER0 0x00000001 +#define KNC_ENABLE_COUNTER1 0x00000002 + +static void knc_pmu_disable_all(void) +{ + u64 val; + + rdmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val); + val &= ~(KNC_ENABLE_COUNTER0|KNC_ENABLE_COUNTER1); + wrmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val); +} + +static void knc_pmu_enable_all(int added) +{ + u64 val; + + rdmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val); + val |= (KNC_ENABLE_COUNTER0|KNC_ENABLE_COUNTER1); + wrmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val); +} + +static inline void +knc_pmu_disable_event(struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + u64 val; + + val = hwc->config; + val &= ~ARCH_PERFMON_EVENTSEL_ENABLE; + + (void)wrmsrl_safe(hwc->config_base + hwc->idx, val); +} + +static void knc_pmu_enable_event(struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + u64 val; + + val = hwc->config; + val |= ARCH_PERFMON_EVENTSEL_ENABLE; + + (void)wrmsrl_safe(hwc->config_base + hwc->idx, val); +} + +static inline u64 knc_pmu_get_status(void) +{ + u64 status; + + rdmsrl(MSR_KNC_IA32_PERF_GLOBAL_STATUS, status); + + return status; +} + +static inline void knc_pmu_ack_status(u64 ack) +{ + wrmsrl(MSR_KNC_IA32_PERF_GLOBAL_OVF_CONTROL, ack); +} + +static int knc_pmu_handle_irq(struct pt_regs *regs) +{ + struct perf_sample_data data; + struct cpu_hw_events *cpuc; + int handled = 0; + int bit, loops; + u64 status; + + cpuc = &__get_cpu_var(cpu_hw_events); + + knc_pmu_disable_all(); + + status = knc_pmu_get_status(); + if (!status) { + knc_pmu_enable_all(0); + return handled; + } + + loops = 0; +again: + knc_pmu_ack_status(status); + if (++loops > 100) { + WARN_ONCE(1, "perf: irq loop stuck!\n"); + perf_event_print_debug(); + goto done; + } + + inc_irq_stat(apic_perf_irqs); + + for_each_set_bit(bit, (unsigned long *)&status, X86_PMC_IDX_MAX) { + struct perf_event *event = cpuc->events[bit]; + + handled++; + + if (!test_bit(bit, cpuc->active_mask)) + continue; + + if (!intel_pmu_save_and_restart(event)) + continue; + + perf_sample_data_init(&data, 0, event->hw.last_period); + + if (perf_event_overflow(event, &data, regs)) + x86_pmu_stop(event, 0); + } + + /* + * Repeat if there is more work to be done: + */ + status = knc_pmu_get_status(); + if (status) + goto again; + +done: + knc_pmu_enable_all(0); + + return handled; +} + + +PMU_FORMAT_ATTR(event, "config:0-7" ); +PMU_FORMAT_ATTR(umask, "config:8-15" ); +PMU_FORMAT_ATTR(edge, "config:18" ); +PMU_FORMAT_ATTR(inv, "config:23" ); +PMU_FORMAT_ATTR(cmask, "config:24-31" ); + +static struct attribute *intel_knc_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_inv.attr, + &format_attr_cmask.attr, + NULL, +}; + +static const struct x86_pmu knc_pmu __initconst = { + .name = "knc", + .handle_irq = knc_pmu_handle_irq, + .disable_all = knc_pmu_disable_all, + .enable_all = knc_pmu_enable_all, + .enable = knc_pmu_enable_event, + .disable = knc_pmu_disable_event, + .hw_config = x86_pmu_hw_config, + .schedule_events = x86_schedule_events, + .eventsel = MSR_KNC_EVNTSEL0, + .perfctr = MSR_KNC_PERFCTR0, + .event_map = knc_pmu_event_map, + .max_events = ARRAY_SIZE(knc_perfmon_event_map), + .apic = 1, + .max_period = (1ULL << 39) - 1, + .version = 0, + .num_counters = 2, + .cntval_bits = 40, + .cntval_mask = (1ULL << 40) - 1, + .get_event_constraints = x86_get_event_constraints, + .event_constraints = knc_event_constraints, + .format_attrs = intel_knc_formats_attr, +}; + +__init int knc_pmu_init(void) +{ + x86_pmu = knc_pmu; + + memcpy(hw_cache_event_ids, knc_hw_cache_event_ids, + sizeof(hw_cache_event_ids)); + + return 0; +} diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c index ef484d9d0a2..5d466b7d860 100644 --- a/arch/x86/kernel/cpu/perf_event_p4.c +++ b/arch/x86/kernel/cpu/perf_event_p4.c @@ -895,8 +895,8 @@ static void p4_pmu_disable_pebs(void) * 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); + * (void)wrmsrl_safe(MSR_IA32_PEBS_ENABLE, 0); + * (void)wrmsrl_safe(MSR_P4_PEBS_MATRIX_VERT, 0); */ } @@ -909,9 +909,8 @@ static inline void p4_pmu_disable_event(struct perf_event *event) * state we need to clear P4_CCCR_OVF, otherwise interrupt get * asserted again and again */ - (void)checking_wrmsrl(hwc->config_base, - (u64)(p4_config_unpack_cccr(hwc->config)) & - ~P4_CCCR_ENABLE & ~P4_CCCR_OVF & ~P4_CCCR_RESERVED); + (void)wrmsrl_safe(hwc->config_base, + p4_config_unpack_cccr(hwc->config) & ~P4_CCCR_ENABLE & ~P4_CCCR_OVF & ~P4_CCCR_RESERVED); } static void p4_pmu_disable_all(void) @@ -943,8 +942,8 @@ static void p4_pmu_enable_pebs(u64 config) 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); + (void)wrmsrl_safe(MSR_IA32_PEBS_ENABLE, (u64)bind->metric_pebs); + (void)wrmsrl_safe(MSR_P4_PEBS_MATRIX_VERT, (u64)bind->metric_vert); } static void p4_pmu_enable_event(struct perf_event *event) @@ -957,7 +956,7 @@ static void p4_pmu_enable_event(struct perf_event *event) u64 escr_addr, cccr; bind = &p4_event_bind_map[idx]; - escr_addr = (u64)bind->escr_msr[thread]; + escr_addr = bind->escr_msr[thread]; /* * - we dont support cascaded counters yet @@ -978,8 +977,8 @@ static void p4_pmu_enable_event(struct perf_event *event) */ p4_pmu_enable_pebs(hwc->config); - (void)checking_wrmsrl(escr_addr, escr_conf); - (void)checking_wrmsrl(hwc->config_base, + (void)wrmsrl_safe(escr_addr, escr_conf); + (void)wrmsrl_safe(hwc->config_base, (cccr & ~P4_CCCR_RESERVED) | P4_CCCR_ENABLE); } @@ -1005,8 +1004,6 @@ static int p4_pmu_handle_irq(struct pt_regs *regs) int idx, handled = 0; u64 val; - perf_sample_data_init(&data, 0); - cpuc = &__get_cpu_var(cpu_hw_events); for (idx = 0; idx < x86_pmu.num_counters; idx++) { @@ -1034,10 +1031,12 @@ static int p4_pmu_handle_irq(struct pt_regs *regs) handled += overflow; /* event overflow for sure */ - data.period = event->hw.last_period; + perf_sample_data_init(&data, 0, hwc->last_period); if (!x86_perf_event_set_period(event)) continue; + + if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); } @@ -1258,7 +1257,24 @@ again: pass++; goto again; } - + /* + * Perf does test runs to see if a whole group can be assigned + * together succesfully. There can be multiple rounds of this. + * Unfortunately, p4_pmu_swap_config_ts touches the hwc->config + * bits, such that the next round of group assignments will + * cause the above p4_should_swap_ts to pass instead of fail. + * This leads to counters exclusive to thread0 being used by + * thread1. + * + * Solve this with a cheap hack, reset the idx back to -1 to + * force a new lookup (p4_next_cntr) to get the right counter + * for the right thread. + * + * This probably doesn't comply with the general spirit of how + * perf wants to work, but P4 is special. :-( + */ + if (p4_should_swap_ts(hwc->config, cpu)) + hwc->idx = -1; p4_pmu_swap_config_ts(hwc, cpu); if (assign) assign[i] = cntr_idx; @@ -1271,6 +1287,17 @@ done: return num ? -EINVAL : 0; } +PMU_FORMAT_ATTR(cccr, "config:0-31" ); +PMU_FORMAT_ATTR(escr, "config:32-62"); +PMU_FORMAT_ATTR(ht, "config:63" ); + +static struct attribute *intel_p4_formats_attr[] = { + &format_attr_cccr.attr, + &format_attr_escr.attr, + &format_attr_ht.attr, + NULL, +}; + static __initconst const struct x86_pmu p4_pmu = { .name = "Netburst P4/Xeon", .handle_irq = p4_pmu_handle_irq, @@ -1305,14 +1332,17 @@ static __initconst const struct x86_pmu p4_pmu = { * the former idea is taken from OProfile code */ .perfctr_second_write = 1, + + .format_attrs = intel_p4_formats_attr, }; __init int p4_pmu_init(void) { unsigned int low, high; + int i, reg; /* If we get stripped -- indexing fails */ - BUILD_BUG_ON(ARCH_P4_MAX_CCCR > X86_PMC_MAX_GENERIC); + BUILD_BUG_ON(ARCH_P4_MAX_CCCR > INTEL_PMC_MAX_GENERIC); rdmsr(MSR_IA32_MISC_ENABLE, low, high); if (!(low & (1 << 7))) { @@ -1328,5 +1358,19 @@ __init int p4_pmu_init(void) x86_pmu = p4_pmu; + /* + * Even though the counters are configured to interrupt a particular + * logical processor when an overflow happens, testing has shown that + * on kdump kernels (which uses a single cpu), thread1's counter + * continues to run and will report an NMI on thread0. Due to the + * overflow bug, this leads to a stream of unknown NMIs. + * + * Solve this by zero'ing out the registers to mimic a reset. + */ + for (i = 0; i < x86_pmu.num_counters; i++) { + reg = x86_pmu_config_addr(i); + wrmsrl_safe(reg, 0ULL); + } + return 0; } diff --git a/arch/x86/kernel/cpu/perf_event_p6.c b/arch/x86/kernel/cpu/perf_event_p6.c index c7181befecd..7c1a0c07b60 100644 --- a/arch/x86/kernel/cpu/perf_event_p6.c +++ b/arch/x86/kernel/cpu/perf_event_p6.c @@ -8,13 +8,106 @@ */ static const u64 p6_perfmon_event_map[] = { - [PERF_COUNT_HW_CPU_CYCLES] = 0x0079, - [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, - [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0f2e, - [PERF_COUNT_HW_CACHE_MISSES] = 0x012e, - [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, - [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, - [PERF_COUNT_HW_BUS_CYCLES] = 0x0062, + [PERF_COUNT_HW_CPU_CYCLES] = 0x0079, /* CPU_CLK_UNHALTED */ + [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, /* INST_RETIRED */ + [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0f2e, /* L2_RQSTS:M:E:S:I */ + [PERF_COUNT_HW_CACHE_MISSES] = 0x012e, /* L2_RQSTS:I */ + [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, /* BR_INST_RETIRED */ + [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, /* BR_MISS_PRED_RETIRED */ + [PERF_COUNT_HW_BUS_CYCLES] = 0x0062, /* BUS_DRDY_CLOCKS */ + [PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = 0x00a2, /* RESOURCE_STALLS */ + +}; + +static const u64 __initconst p6_hw_cache_event_ids + [PERF_COUNT_HW_CACHE_MAX] + [PERF_COUNT_HW_CACHE_OP_MAX] + [PERF_COUNT_HW_CACHE_RESULT_MAX] = +{ + [ C(L1D) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x0043, /* DATA_MEM_REFS */ + [ C(RESULT_MISS) ] = 0x0045, /* DCU_LINES_IN */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0x0f29, /* L2_LD:M:E:S:I */ + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + }, + [ C(L1I ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x0080, /* IFU_IFETCH */ + [ C(RESULT_MISS) ] = 0x0f28, /* L2_IFETCH:M:E:S:I */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + }, + [ C(LL ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0x0025, /* L2_M_LINES_INM */ + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + }, + [ C(DTLB) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x0043, /* DATA_MEM_REFS */ + [ C(RESULT_MISS) ] = 0, + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_MISS) ] = 0, + }, + }, + [ C(ITLB) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x0080, /* IFU_IFETCH */ + [ C(RESULT_MISS) ] = 0x0085, /* ITLB_MISS */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + }, + [ C(BPU ) ] = { + [ C(OP_READ) ] = { + [ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED */ + [ C(RESULT_MISS) ] = 0x00c5, /* BR_MISS_PRED_RETIRED */ + }, + [ C(OP_WRITE) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + [ C(OP_PREFETCH) ] = { + [ C(RESULT_ACCESS) ] = -1, + [ C(RESULT_MISS) ] = -1, + }, + }, }; static u64 p6_pmu_event_map(int hw_event) @@ -34,7 +127,7 @@ static struct event_constraint p6_event_constraints[] = { INTEL_EVENT_CONSTRAINT(0xc1, 0x1), /* FLOPS */ INTEL_EVENT_CONSTRAINT(0x10, 0x1), /* FP_COMP_OPS_EXE */ - INTEL_EVENT_CONSTRAINT(0x11, 0x1), /* FP_ASSIST */ + INTEL_EVENT_CONSTRAINT(0x11, 0x2), /* FP_ASSIST */ INTEL_EVENT_CONSTRAINT(0x12, 0x2), /* MUL */ INTEL_EVENT_CONSTRAINT(0x13, 0x2), /* DIV */ INTEL_EVENT_CONSTRAINT(0x14, 0x1), /* CYCLES_DIV_BUSY */ @@ -64,29 +157,46 @@ static void p6_pmu_enable_all(int added) static inline void p6_pmu_disable_event(struct perf_event *event) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); struct hw_perf_event *hwc = &event->hw; u64 val = P6_NOP_EVENT; - if (cpuc->enabled) - val |= ARCH_PERFMON_EVENTSEL_ENABLE; - - (void)checking_wrmsrl(hwc->config_base, val); + (void)wrmsrl_safe(hwc->config_base, val); } static void p6_pmu_enable_event(struct perf_event *event) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); struct hw_perf_event *hwc = &event->hw; u64 val; val = hwc->config; - if (cpuc->enabled) - val |= ARCH_PERFMON_EVENTSEL_ENABLE; - (void)checking_wrmsrl(hwc->config_base, val); + /* + * p6 only has a global event enable, set on PerfEvtSel0 + * We "disable" events by programming P6_NOP_EVENT + * and we rely on p6_pmu_enable_all() being called + * to actually enable the events. + */ + + (void)wrmsrl_safe(hwc->config_base, val); } +PMU_FORMAT_ATTR(event, "config:0-7" ); +PMU_FORMAT_ATTR(umask, "config:8-15" ); +PMU_FORMAT_ATTR(edge, "config:18" ); +PMU_FORMAT_ATTR(pc, "config:19" ); +PMU_FORMAT_ATTR(inv, "config:23" ); +PMU_FORMAT_ATTR(cmask, "config:24-31" ); + +static struct attribute *intel_p6_formats_attr[] = { + &format_attr_event.attr, + &format_attr_umask.attr, + &format_attr_edge.attr, + &format_attr_pc.attr, + &format_attr_inv.attr, + &format_attr_cmask.attr, + NULL, +}; + static __initconst const struct x86_pmu p6_pmu = { .name = "p6", .handle_irq = x86_pmu_handle_irq, @@ -115,29 +225,55 @@ static __initconst const struct x86_pmu p6_pmu = { .cntval_mask = (1ULL << 32) - 1, .get_event_constraints = x86_get_event_constraints, .event_constraints = p6_event_constraints, + + .format_attrs = intel_p6_formats_attr, + .events_sysfs_show = intel_event_sysfs_show, + }; +static __init void p6_pmu_rdpmc_quirk(void) +{ + if (boot_cpu_data.x86_mask < 9) { + /* + * PPro erratum 26; fixed in stepping 9 and above. + */ + pr_warn("Userspace RDPMC support disabled due to a CPU erratum\n"); + x86_pmu.attr_rdpmc_broken = 1; + x86_pmu.attr_rdpmc = 0; + } +} + __init int p6_pmu_init(void) { + x86_pmu = p6_pmu; + switch (boot_cpu_data.x86_model) { - case 1: - case 3: /* Pentium Pro */ - case 5: - case 6: /* Pentium II */ - case 7: - case 8: - case 11: /* Pentium III */ - case 9: - case 13: - /* Pentium M */ + case 1: /* Pentium Pro */ + x86_add_quirk(p6_pmu_rdpmc_quirk); + break; + + case 3: /* Pentium II - Klamath */ + case 5: /* Pentium II - Deschutes */ + case 6: /* Pentium II - Mendocino */ break; + + case 7: /* Pentium III - Katmai */ + case 8: /* Pentium III - Coppermine */ + case 10: /* Pentium III Xeon */ + case 11: /* Pentium III - Tualatin */ + break; + + case 9: /* Pentium M - Banias */ + case 13: /* Pentium M - Dothan */ + break; + default: - pr_cont("unsupported p6 CPU model %d ", - boot_cpu_data.x86_model); + pr_cont("unsupported p6 CPU model %d ", boot_cpu_data.x86_model); return -ENODEV; } - x86_pmu = p6_pmu; + memcpy(hw_cache_event_ids, p6_hw_cache_event_ids, + sizeof(hw_cache_event_ids)); return 0; } diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c index 966512b2cac..2e8caf03f59 100644 --- a/arch/x86/kernel/cpu/perfctr-watchdog.c +++ b/arch/x86/kernel/cpu/perfctr-watchdog.c @@ -56,6 +56,8 @@ static inline unsigned int nmi_perfctr_msr_to_bit(unsigned int msr) switch (boot_cpu_data.x86) { case 6: return msr - MSR_P6_PERFCTR0; + case 11: + return msr - MSR_KNC_PERFCTR0; case 15: return msr - MSR_P4_BPU_PERFCTR0; } @@ -82,6 +84,8 @@ static inline unsigned int nmi_evntsel_msr_to_bit(unsigned int msr) switch (boot_cpu_data.x86) { case 6: return msr - MSR_P6_EVNTSEL0; + case 11: + return msr - MSR_KNC_EVNTSEL0; case 15: return msr - MSR_P4_BSU_ESCR0; } diff --git a/arch/x86/kernel/cpu/powerflags.c b/arch/x86/kernel/cpu/powerflags.c index 7b3fe56b1c2..31f0f335ed2 100644 --- a/arch/x86/kernel/cpu/powerflags.c +++ b/arch/x86/kernel/cpu/powerflags.c @@ -11,10 +11,10 @@ const char *const x86_power_flags[32] = { "fid", /* frequency id control */ "vid", /* voltage id control */ "ttp", /* thermal trip */ - "tm", - "stc", - "100mhzsteps", - "hwpstate", + "tm", /* hardware thermal control */ + "stc", /* software thermal control */ + "100mhzsteps", /* 100 MHz multiplier control */ + "hwpstate", /* hardware P-state control */ "", /* tsc invariant mapped to constant_tsc */ "cpb", /* core performance boost */ "eff_freq_ro", /* Readonly aperf/mperf */ diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c index 8022c668148..06fe3ed8b85 100644 --- a/arch/x86/kernel/cpu/proc.c +++ b/arch/x86/kernel/cpu/proc.c @@ -11,41 +11,31 @@ static void show_cpuinfo_core(struct seq_file *m, struct cpuinfo_x86 *c, unsigned int cpu) { #ifdef CONFIG_SMP - if (c->x86_max_cores * smp_num_siblings > 1) { - seq_printf(m, "physical id\t: %d\n", c->phys_proc_id); - seq_printf(m, "siblings\t: %d\n", - cpumask_weight(cpu_core_mask(cpu))); - seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id); - seq_printf(m, "cpu cores\t: %d\n", c->booted_cores); - seq_printf(m, "apicid\t\t: %d\n", c->apicid); - seq_printf(m, "initial apicid\t: %d\n", c->initial_apicid); - } + seq_printf(m, "physical id\t: %d\n", c->phys_proc_id); + seq_printf(m, "siblings\t: %d\n", cpumask_weight(cpu_core_mask(cpu))); + seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id); + seq_printf(m, "cpu cores\t: %d\n", c->booted_cores); + seq_printf(m, "apicid\t\t: %d\n", c->apicid); + seq_printf(m, "initial apicid\t: %d\n", c->initial_apicid); #endif } #ifdef CONFIG_X86_32 static void show_cpuinfo_misc(struct seq_file *m, struct cpuinfo_x86 *c) { - /* - * We use exception 16 if we have hardware math and we've either seen - * it or the CPU claims it is internal - */ - int fpu_exception = c->hard_math && (ignore_fpu_irq || cpu_has_fpu); seq_printf(m, "fdiv_bug\t: %s\n" - "hlt_bug\t\t: %s\n" "f00f_bug\t: %s\n" "coma_bug\t: %s\n" "fpu\t\t: %s\n" "fpu_exception\t: %s\n" "cpuid level\t: %d\n" "wp\t\t: %s\n", - c->fdiv_bug ? "yes" : "no", - c->hlt_works_ok ? "no" : "yes", - c->f00f_bug ? "yes" : "no", - c->coma_bug ? "yes" : "no", - c->hard_math ? "yes" : "no", - fpu_exception ? "yes" : "no", + static_cpu_has_bug(X86_BUG_FDIV) ? "yes" : "no", + static_cpu_has_bug(X86_BUG_F00F) ? "yes" : "no", + static_cpu_has_bug(X86_BUG_COMA) ? "yes" : "no", + static_cpu_has(X86_FEATURE_FPU) ? "yes" : "no", + static_cpu_has(X86_FEATURE_FPU) ? "yes" : "no", c->cpuid_level, c->wp_works_ok ? "yes" : "no"); } @@ -140,10 +130,7 @@ static int show_cpuinfo(struct seq_file *m, void *v) static void *c_start(struct seq_file *m, loff_t *pos) { - if (*pos == 0) /* just in case, cpu 0 is not the first */ - *pos = cpumask_first(cpu_online_mask); - else - *pos = cpumask_next(*pos - 1, cpu_online_mask); + *pos = cpumask_next(*pos - 1, cpu_online_mask); if ((*pos) < nr_cpu_ids) return &cpu_data(*pos); return NULL; diff --git a/arch/x86/kernel/cpu/rdrand.c b/arch/x86/kernel/cpu/rdrand.c index feca286c2bb..136ac74dee8 100644 --- a/arch/x86/kernel/cpu/rdrand.c +++ b/arch/x86/kernel/cpu/rdrand.c @@ -27,24 +27,11 @@ static int __init x86_rdrand_setup(char *s) { setup_clear_cpu_cap(X86_FEATURE_RDRAND); + setup_clear_cpu_cap(X86_FEATURE_RDSEED); return 1; } __setup("nordrand", x86_rdrand_setup); -/* We can't use arch_get_random_long() here since alternatives haven't run */ -static inline int rdrand_long(unsigned long *v) -{ - int ok; - asm volatile("1: " RDRAND_LONG "\n\t" - "jc 2f\n\t" - "decl %0\n\t" - "jnz 1b\n\t" - "2:" - : "=r" (ok), "=a" (*v) - : "0" (RDRAND_RETRY_LOOPS)); - return ok; -} - /* * Force a reseed cycle; we are architecturally guaranteed a reseed * after no more than 512 128-bit chunks of random data. This also @@ -52,7 +39,7 @@ static inline int rdrand_long(unsigned long *v) */ #define RESEED_LOOP ((512*128)/sizeof(unsigned long)) -void __cpuinit x86_init_rdrand(struct cpuinfo_x86 *c) +void x86_init_rdrand(struct cpuinfo_x86 *c) { #ifdef CONFIG_ARCH_RANDOM unsigned long tmp; diff --git a/arch/x86/kernel/cpu/scattered.c b/arch/x86/kernel/cpu/scattered.c index c7f64e6f537..b6f794aa169 100644 --- a/arch/x86/kernel/cpu/scattered.c +++ b/arch/x86/kernel/cpu/scattered.c @@ -1,5 +1,5 @@ /* - * Routines to indentify additional cpu features that are scattered in + * Routines to identify additional cpu features that are scattered in * cpuid space. */ #include <linux/cpu.h> @@ -24,14 +24,14 @@ enum cpuid_regs { CR_EBX }; -void __cpuinit init_scattered_cpuid_features(struct cpuinfo_x86 *c) +void 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_DTS, CR_EAX, 0, 0x00000006, 0 }, + static const struct cpuid_bit cpuid_bits[] = { + { X86_FEATURE_DTHERM, CR_EAX, 0, 0x00000006, 0 }, { X86_FEATURE_IDA, CR_EAX, 1, 0x00000006, 0 }, { X86_FEATURE_ARAT, CR_EAX, 2, 0x00000006, 0 }, { X86_FEATURE_PLN, CR_EAX, 4, 0x00000006, 0 }, @@ -39,7 +39,9 @@ void __cpuinit init_scattered_cpuid_features(struct cpuinfo_x86 *c) { 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_HW_PSTATE, CR_EDX, 7, 0x80000007, 0 }, { X86_FEATURE_CPB, CR_EDX, 9, 0x80000007, 0 }, + { X86_FEATURE_PROC_FEEDBACK, CR_EDX,11, 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 }, diff --git a/arch/x86/kernel/cpu/sched.c b/arch/x86/kernel/cpu/sched.c deleted file mode 100644 index a640ae5ad20..00000000000 --- a/arch/x86/kernel/cpu/sched.c +++ /dev/null @@ -1,55 +0,0 @@ -#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/topology.c b/arch/x86/kernel/cpu/topology.c index 4397e987a1c..4c60eaf0571 100644 --- a/arch/x86/kernel/cpu/topology.c +++ b/arch/x86/kernel/cpu/topology.c @@ -26,7 +26,7 @@ * exists, use it for populating initial_apicid and cpu topology * detection. */ -void __cpuinit detect_extended_topology(struct cpuinfo_x86 *c) +void detect_extended_topology(struct cpuinfo_x86 *c) { #ifdef CONFIG_SMP unsigned int eax, ebx, ecx, edx, sub_index; diff --git a/arch/x86/kernel/cpu/transmeta.c b/arch/x86/kernel/cpu/transmeta.c index 28000743bbb..3fa0e5ad86b 100644 --- a/arch/x86/kernel/cpu/transmeta.c +++ b/arch/x86/kernel/cpu/transmeta.c @@ -1,11 +1,10 @@ #include <linux/kernel.h> #include <linux/mm.h> -#include <linux/init.h> #include <asm/processor.h> #include <asm/msr.h> #include "cpu.h" -static void __cpuinit early_init_transmeta(struct cpuinfo_x86 *c) +static void early_init_transmeta(struct cpuinfo_x86 *c) { u32 xlvl; @@ -17,7 +16,7 @@ static void __cpuinit early_init_transmeta(struct cpuinfo_x86 *c) } } -static void __cpuinit init_transmeta(struct cpuinfo_x86 *c) +static void init_transmeta(struct cpuinfo_x86 *c) { unsigned int cap_mask, uk, max, dummy; unsigned int cms_rev1, cms_rev2; @@ -98,7 +97,7 @@ static void __cpuinit init_transmeta(struct cpuinfo_x86 *c) #endif } -static const struct cpu_dev __cpuinitconst transmeta_cpu_dev = { +static const struct cpu_dev transmeta_cpu_dev = { .c_vendor = "Transmeta", .c_ident = { "GenuineTMx86", "TransmetaCPU" }, .c_early_init = early_init_transmeta, diff --git a/arch/x86/kernel/cpu/umc.c b/arch/x86/kernel/cpu/umc.c index fd2c37bf7ac..ef9c2a0078b 100644 --- a/arch/x86/kernel/cpu/umc.c +++ b/arch/x86/kernel/cpu/umc.c @@ -1,5 +1,4 @@ #include <linux/kernel.h> -#include <linux/init.h> #include <asm/processor.h> #include "cpu.h" @@ -8,11 +7,11 @@ * so no special init takes place. */ -static const struct cpu_dev __cpuinitconst umc_cpu_dev = { +static const struct cpu_dev umc_cpu_dev = { .c_vendor = "UMC", .c_ident = { "UMC UMC UMC" }, - .c_models = { - { .vendor = X86_VENDOR_UMC, .family = 4, .model_names = + .legacy_models = { + { .family = 4, .model_names = { [1] = "U5D", [2] = "U5S", diff --git a/arch/x86/kernel/cpu/vmware.c b/arch/x86/kernel/cpu/vmware.c index d22d0c4edcf..628a059a9a0 100644 --- a/arch/x86/kernel/cpu/vmware.c +++ b/arch/x86/kernel/cpu/vmware.c @@ -33,6 +33,9 @@ #define VMWARE_PORT_CMD_GETVERSION 10 #define VMWARE_PORT_CMD_GETHZ 45 +#define VMWARE_PORT_CMD_GETVCPU_INFO 68 +#define VMWARE_PORT_CMD_LEGACY_X2APIC 3 +#define VMWARE_PORT_CMD_VCPU_RESERVED 31 #define VMWARE_PORT(cmd, eax, ebx, ecx, edx) \ __asm__("inl (%%dx)" : \ @@ -90,7 +93,7 @@ static void __init vmware_platform_setup(void) * serial key should be enough, as this will always have a VMware * specific string when running under VMware hypervisor. */ -static bool __init vmware_platform(void) +static uint32_t __init vmware_platform(void) { if (cpu_has_hypervisor) { unsigned int eax; @@ -99,12 +102,12 @@ static bool __init vmware_platform(void) cpuid(CPUID_VMWARE_INFO_LEAF, &eax, &hyper_vendor_id[0], &hyper_vendor_id[1], &hyper_vendor_id[2]); if (!memcmp(hyper_vendor_id, "VMwareVMware", 12)) - return true; + return CPUID_VMWARE_INFO_LEAF; } else if (dmi_available && dmi_name_in_serial("VMware") && __vmware_platform()) - return true; + return 1; - return false; + return 0; } /* @@ -119,16 +122,26 @@ static bool __init vmware_platform(void) * so that the kernel could just trust the hypervisor with providing a * reliable virtual TSC that is suitable for timekeeping. */ -static void __cpuinit vmware_set_cpu_features(struct cpuinfo_x86 *c) +static void vmware_set_cpu_features(struct cpuinfo_x86 *c) { set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); set_cpu_cap(c, X86_FEATURE_TSC_RELIABLE); } +/* Checks if hypervisor supports x2apic without VT-D interrupt remapping. */ +static bool __init vmware_legacy_x2apic_available(void) +{ + uint32_t eax, ebx, ecx, edx; + VMWARE_PORT(GETVCPU_INFO, eax, ebx, ecx, edx); + return (eax & (1 << VMWARE_PORT_CMD_VCPU_RESERVED)) == 0 && + (eax & (1 << VMWARE_PORT_CMD_LEGACY_X2APIC)) != 0; +} + const __refconst struct hypervisor_x86 x86_hyper_vmware = { .name = "VMware", .detect = vmware_platform, .set_cpu_features = vmware_set_cpu_features, .init_platform = vmware_platform_setup, + .x2apic_available = vmware_legacy_x2apic_available, }; EXPORT_SYMBOL(x86_hyper_vmware); |