diff options
62 files changed, 703 insertions, 275 deletions
diff --git a/MAINTAINERS b/MAINTAINERS index 78d317c545b9..23d04363a195 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -4653,6 +4653,16 @@ S: Maintained F: crypto/pcrypt.c F: include/crypto/pcrypt.h +PER-CPU MEMORY ALLOCATOR +M: Tejun Heo <tj@kernel.org> +M: Christoph Lameter <cl@linux-foundation.org> +L: linux-kernel@vger.kernel.org +T: git git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu.git +S: Maintained +F: include/linux/percpu*.h +F: mm/percpu*.c +F: arch/*/include/asm/percpu.h + PER-TASK DELAY ACCOUNTING M: Balbir Singh <balbir@linux.vnet.ibm.com> S: Maintained diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu index 2ac9069890cd..15588a0ef466 100644 --- a/arch/x86/Kconfig.cpu +++ b/arch/x86/Kconfig.cpu @@ -310,6 +310,9 @@ config X86_INTERNODE_CACHE_SHIFT config X86_CMPXCHG def_bool X86_64 || (X86_32 && !M386) +config CMPXCHG_LOCAL + def_bool X86_64 || (X86_32 && !M386) + config X86_L1_CACHE_SHIFT int default "7" if MPENTIUM4 || MPSC diff --git a/arch/x86/include/asm/debugreg.h b/arch/x86/include/asm/debugreg.h index b81002f23614..078ad0caefc6 100644 --- a/arch/x86/include/asm/debugreg.h +++ b/arch/x86/include/asm/debugreg.h @@ -94,7 +94,7 @@ static inline void hw_breakpoint_disable(void) static inline int hw_breakpoint_active(void) { - return __get_cpu_var(cpu_dr7) & DR_GLOBAL_ENABLE_MASK; + return __this_cpu_read(cpu_dr7) & DR_GLOBAL_ENABLE_MASK; } extern void aout_dump_debugregs(struct user *dump); diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h index f899e01a8ac9..8ee45167e817 100644 --- a/arch/x86/include/asm/percpu.h +++ b/arch/x86/include/asm/percpu.h @@ -230,6 +230,125 @@ do { \ }) /* + * Add return operation + */ +#define percpu_add_return_op(var, val) \ +({ \ + typeof(var) paro_ret__ = val; \ + switch (sizeof(var)) { \ + case 1: \ + asm("xaddb %0, "__percpu_arg(1) \ + : "+q" (paro_ret__), "+m" (var) \ + : : "memory"); \ + break; \ + case 2: \ + asm("xaddw %0, "__percpu_arg(1) \ + : "+r" (paro_ret__), "+m" (var) \ + : : "memory"); \ + break; \ + case 4: \ + asm("xaddl %0, "__percpu_arg(1) \ + : "+r" (paro_ret__), "+m" (var) \ + : : "memory"); \ + break; \ + case 8: \ + asm("xaddq %0, "__percpu_arg(1) \ + : "+re" (paro_ret__), "+m" (var) \ + : : "memory"); \ + break; \ + default: __bad_percpu_size(); \ + } \ + paro_ret__ += val; \ + paro_ret__; \ +}) + +/* + * xchg is implemented using cmpxchg without a lock prefix. xchg is + * expensive due to the implied lock prefix. The processor cannot prefetch + * cachelines if xchg is used. + */ +#define percpu_xchg_op(var, nval) \ +({ \ + typeof(var) pxo_ret__; \ + typeof(var) pxo_new__ = (nval); \ + switch (sizeof(var)) { \ + case 1: \ + asm("\n1:mov "__percpu_arg(1)",%%al" \ + "\n\tcmpxchgb %2, "__percpu_arg(1) \ + "\n\tjnz 1b" \ + : "=a" (pxo_ret__), "+m" (var) \ + : "q" (pxo_new__) \ + : "memory"); \ + break; \ + case 2: \ + asm("\n1:mov "__percpu_arg(1)",%%ax" \ + "\n\tcmpxchgw %2, "__percpu_arg(1) \ + "\n\tjnz 1b" \ + : "=a" (pxo_ret__), "+m" (var) \ + : "r" (pxo_new__) \ + : "memory"); \ + break; \ + case 4: \ + asm("\n1:mov "__percpu_arg(1)",%%eax" \ + "\n\tcmpxchgl %2, "__percpu_arg(1) \ + "\n\tjnz 1b" \ + : "=a" (pxo_ret__), "+m" (var) \ + : "r" (pxo_new__) \ + : "memory"); \ + break; \ + case 8: \ + asm("\n1:mov "__percpu_arg(1)",%%rax" \ + "\n\tcmpxchgq %2, "__percpu_arg(1) \ + "\n\tjnz 1b" \ + : "=a" (pxo_ret__), "+m" (var) \ + : "r" (pxo_new__) \ + : "memory"); \ + break; \ + default: __bad_percpu_size(); \ + } \ + pxo_ret__; \ +}) + +/* + * cmpxchg has no such implied lock semantics as a result it is much + * more efficient for cpu local operations. + */ +#define percpu_cmpxchg_op(var, oval, nval) \ +({ \ + typeof(var) pco_ret__; \ + typeof(var) pco_old__ = (oval); \ + typeof(var) pco_new__ = (nval); \ + switch (sizeof(var)) { \ + case 1: \ + asm("cmpxchgb %2, "__percpu_arg(1) \ + : "=a" (pco_ret__), "+m" (var) \ + : "q" (pco_new__), "0" (pco_old__) \ + : "memory"); \ + break; \ + case 2: \ + asm("cmpxchgw %2, "__percpu_arg(1) \ + : "=a" (pco_ret__), "+m" (var) \ + : "r" (pco_new__), "0" (pco_old__) \ + : "memory"); \ + break; \ + case 4: \ + asm("cmpxchgl %2, "__percpu_arg(1) \ + : "=a" (pco_ret__), "+m" (var) \ + : "r" (pco_new__), "0" (pco_old__) \ + : "memory"); \ + break; \ + case 8: \ + asm("cmpxchgq %2, "__percpu_arg(1) \ + : "=a" (pco_ret__), "+m" (var) \ + : "r" (pco_new__), "0" (pco_old__) \ + : "memory"); \ + break; \ + default: __bad_percpu_size(); \ + } \ + pco_ret__; \ +}) + +/* * percpu_read() makes gcc load the percpu variable every time it is * accessed while percpu_read_stable() allows the value to be cached. * percpu_read_stable() is more efficient and can be used if its value @@ -267,6 +386,12 @@ do { \ #define __this_cpu_xor_1(pcp, val) percpu_to_op("xor", (pcp), val) #define __this_cpu_xor_2(pcp, val) percpu_to_op("xor", (pcp), val) #define __this_cpu_xor_4(pcp, val) percpu_to_op("xor", (pcp), val) +/* + * Generic fallback operations for __this_cpu_xchg_[1-4] are okay and much + * faster than an xchg with forced lock semantics. + */ +#define __this_cpu_xchg_8(pcp, nval) percpu_xchg_op(pcp, nval) +#define __this_cpu_cmpxchg_8(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) #define this_cpu_read_1(pcp) percpu_from_op("mov", (pcp), "m"(pcp)) #define this_cpu_read_2(pcp) percpu_from_op("mov", (pcp), "m"(pcp)) @@ -286,6 +411,11 @@ do { \ #define this_cpu_xor_1(pcp, val) percpu_to_op("xor", (pcp), val) #define this_cpu_xor_2(pcp, val) percpu_to_op("xor", (pcp), val) #define this_cpu_xor_4(pcp, val) percpu_to_op("xor", (pcp), val) +#define this_cpu_xchg_1(pcp, nval) percpu_xchg_op(pcp, nval) +#define this_cpu_xchg_2(pcp, nval) percpu_xchg_op(pcp, nval) +#define this_cpu_xchg_4(pcp, nval) percpu_xchg_op(pcp, nval) +#define this_cpu_xchg_8(pcp, nval) percpu_xchg_op(pcp, nval) +#define this_cpu_cmpxchg_8(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) #define irqsafe_cpu_add_1(pcp, val) percpu_add_op((pcp), val) #define irqsafe_cpu_add_2(pcp, val) percpu_add_op((pcp), val) @@ -299,6 +429,31 @@ do { \ #define irqsafe_cpu_xor_1(pcp, val) percpu_to_op("xor", (pcp), val) #define irqsafe_cpu_xor_2(pcp, val) percpu_to_op("xor", (pcp), val) #define irqsafe_cpu_xor_4(pcp, val) percpu_to_op("xor", (pcp), val) +#define irqsafe_cpu_xchg_1(pcp, nval) percpu_xchg_op(pcp, nval) +#define irqsafe_cpu_xchg_2(pcp, nval) percpu_xchg_op(pcp, nval) +#define irqsafe_cpu_xchg_4(pcp, nval) percpu_xchg_op(pcp, nval) +#define irqsafe_cpu_xchg_8(pcp, nval) percpu_xchg_op(pcp, nval) +#define irqsafe_cpu_cmpxchg_8(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) + +#ifndef CONFIG_M386 +#define __this_cpu_add_return_1(pcp, val) percpu_add_return_op(pcp, val) +#define __this_cpu_add_return_2(pcp, val) percpu_add_return_op(pcp, val) +#define __this_cpu_add_return_4(pcp, val) percpu_add_return_op(pcp, val) +#define __this_cpu_cmpxchg_1(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) +#define __this_cpu_cmpxchg_2(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) +#define __this_cpu_cmpxchg_4(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) + +#define this_cpu_add_return_1(pcp, val) percpu_add_return_op(pcp, val) +#define this_cpu_add_return_2(pcp, val) percpu_add_return_op(pcp, val) +#define this_cpu_add_return_4(pcp, val) percpu_add_return_op(pcp, val) +#define this_cpu_cmpxchg_1(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) +#define this_cpu_cmpxchg_2(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) +#define this_cpu_cmpxchg_4(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) + +#define irqsafe_cpu_cmpxchg_1(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) +#define irqsafe_cpu_cmpxchg_2(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) +#define irqsafe_cpu_cmpxchg_4(pcp, oval, nval) percpu_cmpxchg_op(pcp, oval, nval) +#endif /* !CONFIG_M386 */ /* * Per cpu atomic 64 bit operations are only available under 64 bit. @@ -311,6 +466,7 @@ do { \ #define __this_cpu_and_8(pcp, val) percpu_to_op("and", (pcp), val) #define __this_cpu_or_8(pcp, val) percpu_to_op("or", (pcp), val) #define __this_cpu_xor_8(pcp, val) percpu_to_op("xor", (pcp), val) +#define __this_cpu_add_return_8(pcp, val) percpu_add_return_op(pcp, val) #define this_cpu_read_8(pcp) percpu_from_op("mov", (pcp), "m"(pcp)) #define this_cpu_write_8(pcp, val) percpu_to_op("mov", (pcp), val) @@ -318,12 +474,12 @@ do { \ #define this_cpu_and_8(pcp, val) percpu_to_op("and", (pcp), val) #define this_cpu_or_8(pcp, val) percpu_to_op("or", (pcp), val) #define this_cpu_xor_8(pcp, val) percpu_to_op("xor", (pcp), val) +#define this_cpu_add_return_8(pcp, val) percpu_add_return_op(pcp, val) #define irqsafe_cpu_add_8(pcp, val) percpu_add_op((pcp), val) #define irqsafe_cpu_and_8(pcp, val) percpu_to_op("and", (pcp), val) #define irqsafe_cpu_or_8(pcp, val) percpu_to_op("or", (pcp), val) #define irqsafe_cpu_xor_8(pcp, val) percpu_to_op("xor", (pcp), val) - #endif /* This is not atomic against other CPUs -- CPU preemption needs to be off */ diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h index cae9c3cb95cf..c6efecf85a6a 100644 --- a/arch/x86/include/asm/processor.h +++ b/arch/x86/include/asm/processor.h @@ -141,10 +141,9 @@ extern __u32 cpu_caps_set[NCAPINTS]; #ifdef CONFIG_SMP DECLARE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info); #define cpu_data(cpu) per_cpu(cpu_info, cpu) -#define current_cpu_data __get_cpu_var(cpu_info) #else +#define cpu_info boot_cpu_data #define cpu_data(cpu) boot_cpu_data -#define current_cpu_data boot_cpu_data #endif extern const struct seq_operations cpuinfo_op; diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c index ce65d449b750..79e6baa8aa0a 100644 --- a/arch/x86/kernel/apic/apic.c +++ b/arch/x86/kernel/apic/apic.c @@ -516,7 +516,7 @@ static void __cpuinit setup_APIC_timer(void) { struct clock_event_device *levt = &__get_cpu_var(lapic_events); - if (cpu_has(¤t_cpu_data, X86_FEATURE_ARAT)) { + if (cpu_has(__this_cpu_ptr(&cpu_info), X86_FEATURE_ARAT)) { lapic_clockevent.features &= ~CLOCK_EVT_FEAT_C3STOP; /* Make LAPIC timer preferrable over percpu HPET */ lapic_clockevent.rating = 150; diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c index 52735a710c30..697dc34b7b87 100644 --- a/arch/x86/kernel/apic/io_apic.c +++ b/arch/x86/kernel/apic/io_apic.c @@ -2329,7 +2329,7 @@ asmlinkage void smp_irq_move_cleanup_interrupt(void) unsigned int irr; struct irq_desc *desc; struct irq_cfg *cfg; - irq = __get_cpu_var(vector_irq)[vector]; + irq = __this_cpu_read(vector_irq[vector]); if (irq == -1) continue; @@ -2363,7 +2363,7 @@ asmlinkage void smp_irq_move_cleanup_interrupt(void) apic->send_IPI_self(IRQ_MOVE_CLEANUP_VECTOR); goto unlock; } - __get_cpu_var(vector_irq)[vector] = -1; + __this_cpu_write(vector_irq[vector], -1); unlock: raw_spin_unlock(&desc->lock); } diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c index 2a3f2a7db243..ecca5f41ad2c 100644 --- a/arch/x86/kernel/apic/x2apic_uv_x.c +++ b/arch/x86/kernel/apic/x2apic_uv_x.c @@ -120,8 +120,8 @@ static int __init uv_acpi_madt_oem_check(char *oem_id, char *oem_table_id) else if (!strcmp(oem_table_id, "UVX")) uv_system_type = UV_X2APIC; else if (!strcmp(oem_table_id, "UVH")) { - __get_cpu_var(x2apic_extra_bits) = - pnodeid << uvh_apicid.s.pnode_shift; + __this_cpu_write(x2apic_extra_bits, + pnodeid << uvh_apicid.s.pnode_shift); uv_system_type = UV_NON_UNIQUE_APIC; uv_set_apicid_hibit(); return 1; @@ -286,7 +286,7 @@ static unsigned int x2apic_get_apic_id(unsigned long x) unsigned int id; WARN_ON(preemptible() && num_online_cpus() > 1); - id = x | __get_cpu_var(x2apic_extra_bits); + id = x | __this_cpu_read(x2apic_extra_bits); return id; } @@ -378,7 +378,7 @@ struct apic __refdata apic_x2apic_uv_x = { static __cpuinit void set_x2apic_extra_bits(int pnode) { - __get_cpu_var(x2apic_extra_bits) = (pnode << 6); + __this_cpu_write(x2apic_extra_bits, (pnode << 6)); } /* diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index 9e093f8fe78c..7c7bedb83c5a 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -668,7 +668,7 @@ EXPORT_SYMBOL_GPL(amd_erratum_383); bool cpu_has_amd_erratum(const int *erratum) { - struct cpuinfo_x86 *cpu = ¤t_cpu_data; + struct cpuinfo_x86 *cpu = __this_cpu_ptr(&cpu_info); int osvw_id = *erratum++; u32 range; u32 ms; diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c index 491977baf6c0..35c7e65e59be 100644 --- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c +++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c @@ -521,7 +521,7 @@ static void check_supported_cpu(void *_rc) *rc = -ENODEV; - if (current_cpu_data.x86_vendor != X86_VENDOR_AMD) + if (__this_cpu_read(cpu_info.x86_vendor) != X86_VENDOR_AMD) return; eax = cpuid_eax(CPUID_PROCESSOR_SIGNATURE); @@ -1377,7 +1377,7 @@ static int __devexit powernowk8_cpu_exit(struct cpufreq_policy *pol) static void query_values_on_cpu(void *_err) { int *err = _err; - struct powernow_k8_data *data = __get_cpu_var(powernow_data); + struct powernow_k8_data *data = __this_cpu_read(powernow_data); *err = query_current_values_with_pending_wait(data); } diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c index 9ecf81f9b90f..7283e98deaae 100644 --- a/arch/x86/kernel/cpu/intel_cacheinfo.c +++ b/arch/x86/kernel/cpu/intel_cacheinfo.c @@ -265,7 +265,7 @@ amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax, line_size = l2.line_size; lines_per_tag = l2.lines_per_tag; /* cpu_data has errata corrections for K7 applied */ - size_in_kb = current_cpu_data.x86_cache_size; + size_in_kb = __this_cpu_read(cpu_info.x86_cache_size); break; case 3: if (!l3.val) @@ -287,7 +287,7 @@ amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax, eax->split.type = types[leaf]; eax->split.level = levels[leaf]; eax->split.num_threads_sharing = 0; - eax->split.num_cores_on_die = current_cpu_data.x86_max_cores - 1; + eax->split.num_cores_on_die = __this_cpu_read(cpu_info.x86_max_cores) - 1; if (assoc == 0xffff) diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c index 7a35b72d7c03..d916183b7f9c 100644 --- a/arch/x86/kernel/cpu/mcheck/mce.c +++ b/arch/x86/kernel/cpu/mcheck/mce.c @@ -326,7 +326,7 @@ static void mce_panic(char *msg, struct mce *final, char *exp) static int msr_to_offset(u32 msr) { - unsigned bank = __get_cpu_var(injectm.bank); + unsigned bank = __this_cpu_read(injectm.bank); if (msr == rip_msr) return offsetof(struct mce, ip); @@ -346,7 +346,7 @@ static u64 mce_rdmsrl(u32 msr) { u64 v; - if (__get_cpu_var(injectm).finished) { + if (__this_cpu_read(injectm.finished)) { int offset = msr_to_offset(msr); if (offset < 0) @@ -369,7 +369,7 @@ static u64 mce_rdmsrl(u32 msr) static void mce_wrmsrl(u32 msr, u64 v) { - if (__get_cpu_var(injectm).finished) { + if (__this_cpu_read(injectm.finished)) { int offset = msr_to_offset(msr); if (offset >= 0) @@ -1159,7 +1159,7 @@ static void mce_start_timer(unsigned long data) WARN_ON(smp_processor_id() != data); - if (mce_available(¤t_cpu_data)) { + if (mce_available(__this_cpu_ptr(&cpu_info))) { machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_poll_banks)); } @@ -1767,7 +1767,7 @@ static int mce_shutdown(struct sys_device *dev) static int mce_resume(struct sys_device *dev) { __mcheck_cpu_init_generic(); - __mcheck_cpu_init_vendor(¤t_cpu_data); + __mcheck_cpu_init_vendor(__this_cpu_ptr(&cpu_info)); return 0; } @@ -1775,7 +1775,7 @@ static int mce_resume(struct sys_device *dev) static void mce_cpu_restart(void *data) { del_timer_sync(&__get_cpu_var(mce_timer)); - if (!mce_available(¤t_cpu_data)) + if (!mce_available(__this_cpu_ptr(&cpu_info))) return; __mcheck_cpu_init_generic(); __mcheck_cpu_init_timer(); @@ -1790,7 +1790,7 @@ static void mce_restart(void) /* Toggle features for corrected errors */ static void mce_disable_ce(void *all) { - if (!mce_available(¤t_cpu_data)) + if (!mce_available(__this_cpu_ptr(&cpu_info))) return; if (all) del_timer_sync(&__get_cpu_var(mce_timer)); @@ -1799,7 +1799,7 @@ static void mce_disable_ce(void *all) static void mce_enable_ce(void *all) { - if (!mce_available(¤t_cpu_data)) + if (!mce_available(__this_cpu_ptr(&cpu_info))) return; cmci_reenable(); cmci_recheck(); @@ -2022,7 +2022,7 @@ static void __cpuinit mce_disable_cpu(void *h) unsigned long action = *(unsigned long *)h; int i; - if (!mce_available(¤t_cpu_data)) + if (!mce_available(__this_cpu_ptr(&cpu_info))) return; if (!(action & CPU_TASKS_FROZEN)) @@ -2040,7 +2040,7 @@ static void __cpuinit mce_reenable_cpu(void *h) unsigned long action = *(unsigned long *)h; int i; - if (!mce_available(¤t_cpu_data)) + if (!mce_available(__this_cpu_ptr(&cpu_info))) return; if (!(action & CPU_TASKS_FROZEN)) diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel.c b/arch/x86/kernel/cpu/mcheck/mce_intel.c index 6fcd0936194f..8694ef56459d 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_intel.c +++ b/arch/x86/kernel/cpu/mcheck/mce_intel.c @@ -130,7 +130,7 @@ void cmci_recheck(void) unsigned long flags; int banks; - if (!mce_available(¤t_cpu_data) || !cmci_supported(&banks)) + if (!mce_available(__this_cpu_ptr(&cpu_info)) || !cmci_supported(&banks)) return; local_irq_save(flags); machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned)); diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 0a360d146596..04921017abe0 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -997,8 +997,7 @@ x86_perf_event_set_period(struct perf_event *event) static void x86_pmu_enable_event(struct perf_event *event) { - struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); - if (cpuc->enabled) + if (__this_cpu_read(cpu_hw_events.enabled)) __x86_pmu_enable_event(&event->hw, ARCH_PE |