Merge branch 'for-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu

* 'for-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu: (30 commits)
  gameport: use this_cpu_read instead of lookup
  x86: udelay: Use this_cpu_read to avoid address calculation
  x86: Use this_cpu_inc_return for nmi counter
  x86: Replace uses of current_cpu_data with this_cpu ops
  x86: Use this_cpu_ops to optimize code
  vmstat: User per cpu atomics to avoid interrupt disable / enable
  irq_work: Use per cpu atomics instead of regular atomics
  cpuops: Use cmpxchg for xchg to avoid lock semantics
  x86: this_cpu_cmpxchg and this_cpu_xchg operations
  percpu: Generic this_cpu_cmpxchg() and this_cpu_xchg support
  percpu,x86: relocate this_cpu_add_return() and friends
  connector: Use this_cpu operations
  xen: Use this_cpu_inc_return
  taskstats: Use this_cpu_ops
  random: Use this_cpu_inc_return
  fs: Use this_cpu_inc_return in buffer.c
  highmem: Use this_cpu_xx_return() operations
  vmstat: Use this_cpu_inc_return for vm statistics
  x86: Support for this_cpu_add, sub, dec, inc_return
  percpu: Generic support for this_cpu_add, sub, dec, inc_return
  ...

Fixed up conflicts: in arch/x86/kernel/{apic/nmi.c, apic/x2apic_uv_x.c, process.c}
as per Tejun.

Author: torvalds

MAINTAINERS

@@ -4653,6 +4653,16 @@ S:	Maintained
 F:	crypto/pcrypt.c
 F:	include/crypto/pcrypt.h
 
+PER-CPU MEMORY ALLOCATOR
+M:	Tejun Heo <[email protected]>
+M:	Christoph Lameter <[email protected]>
+L:	[email protected]
+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 <[email protected]>
 S:	Maintained

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

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);

arch/x86/include/asm/percpu.h

@@ -229,6 +229,125 @@ 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.
@@ -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.
@@ -311,19 +466,20 @@ 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)
 #define this_cpu_add_8(pcp, val)	percpu_add_op((pcp), val)
 #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 */

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;

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(&current_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;

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);
 	}

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));
 }
 
 /*

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 = &current_cpu_data;
+	struct cpuinfo_x86 *cpu = __this_cpu_ptr(&cpu_info);
 	int osvw_id = *erratum++;
 	u32 range;
 	u32 ms;

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);
 }

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)