From b76834bc1b6db0a0923eed85c81b1113021b0612 Mon Sep 17 00:00:00 2001
From: Christoph Lameter <cl@linux.com>
Date: Mon, 6 Dec 2010 11:16:25 -0600
Subject: kprobes: Use this_cpu_ops

Use this_cpu ops in various places to optimize per cpu data access.

Cc: Jason Baron <jbaron@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/kernel/kprobes.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

(limited to 'arch')

diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c
index 1cbd54c..572ecc8 100644
--- a/arch/x86/kernel/kprobes.c
+++ b/arch/x86/kernel/kprobes.c
@@ -403,7 +403,7 @@ static void __kprobes save_previous_kprobe(struct kprobe_ctlblk *kcb)
 
 static void __kprobes restore_previous_kprobe(struct kprobe_ctlblk *kcb)
 {
-	__get_cpu_var(current_kprobe) = kcb->prev_kprobe.kp;
+	__this_cpu_write(current_kprobe, kcb->prev_kprobe.kp);
 	kcb->kprobe_status = kcb->prev_kprobe.status;
 	kcb->kprobe_old_flags = kcb->prev_kprobe.old_flags;
 	kcb->kprobe_saved_flags = kcb->prev_kprobe.saved_flags;
@@ -412,7 +412,7 @@ static void __kprobes restore_previous_kprobe(struct kprobe_ctlblk *kcb)
 static void __kprobes set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
 				struct kprobe_ctlblk *kcb)
 {
-	__get_cpu_var(current_kprobe) = p;
+	__this_cpu_write(current_kprobe, p);
 	kcb->kprobe_saved_flags = kcb->kprobe_old_flags
 		= (regs->flags & (X86_EFLAGS_TF | X86_EFLAGS_IF));
 	if (is_IF_modifier(p->ainsn.insn))
@@ -586,7 +586,7 @@ static int __kprobes kprobe_handler(struct pt_regs *regs)
 		preempt_enable_no_resched();
 		return 1;
 	} else if (kprobe_running()) {
-		p = __get_cpu_var(current_kprobe);
+		p = __this_cpu_read(current_kprobe);
 		if (p->break_handler && p->break_handler(p, regs)) {
 			setup_singlestep(p, regs, kcb, 0);
 			return 1;
@@ -759,11 +759,11 @@ static __used __kprobes void *trampoline_handler(struct pt_regs *regs)
 
 		orig_ret_address = (unsigned long)ri->ret_addr;
 		if (ri->rp && ri->rp->handler) {
-			__get_cpu_var(current_kprobe) = &ri->rp->kp;
+			__this_cpu_write(current_kprobe, &ri->rp->kp);
 			get_kprobe_ctlblk()->kprobe_status = KPROBE_HIT_ACTIVE;
 			ri->ret_addr = correct_ret_addr;
 			ri->rp->handler(ri, regs);
-			__get_cpu_var(current_kprobe) = NULL;
+			__this_cpu_write(current_kprobe, NULL);
 		}
 
 		recycle_rp_inst(ri, &empty_rp);
@@ -1198,10 +1198,10 @@ static void __kprobes optimized_callback(struct optimized_kprobe *op,
 		regs->ip = (unsigned long)op->kp.addr + INT3_SIZE;
 		regs->orig_ax = ~0UL;
 
-		__get_cpu_var(current_kprobe) = &op->kp;
+		__this_cpu_write(current_kprobe, &op->kp);
 		kcb->kprobe_status = KPROBE_HIT_ACTIVE;
 		opt_pre_handler(&op->kp, regs);
-		__get_cpu_var(current_kprobe) = NULL;
+		__this_cpu_write(current_kprobe, NULL);
 	}
 	preempt_enable_no_resched();
 }
-- 
cgit v1.1


From 780f36d8b3fa9572f731d4fb85067b2e45e6f993 Mon Sep 17 00:00:00 2001
From: Christoph Lameter <cl@linux.com>
Date: Mon, 6 Dec 2010 11:16:29 -0600
Subject: xen: Use this_cpu_ops

Use this_cpu_ops to reduce code size and simplify things in various places.

V3->V4:
	Move instance of this_cpu_inc_return to a later patchset so that
	this patch can be applied without infrastructure changes.

Cc: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/xen/enlighten.c  | 4 ++--
 arch/x86/xen/multicalls.h | 2 +-
 arch/x86/xen/spinlock.c   | 8 ++++----
 arch/x86/xen/time.c       | 8 ++++----
 4 files changed, 11 insertions(+), 11 deletions(-)

(limited to 'arch')

diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 44dcad4..aa8c89a 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -574,8 +574,8 @@ static void xen_write_idt_entry(gate_desc *dt, int entrynum, const gate_desc *g)
 
 	preempt_disable();
 
-	start = __get_cpu_var(idt_desc).address;
-	end = start + __get_cpu_var(idt_desc).size + 1;
+	start = __this_cpu_read(idt_desc.address);
+	end = start + __this_cpu_read(idt_desc.size) + 1;
 
 	xen_mc_flush();
 
diff --git a/arch/x86/xen/multicalls.h b/arch/x86/xen/multicalls.h
index 9e565da..4ec8035 100644
--- a/arch/x86/xen/multicalls.h
+++ b/arch/x86/xen/multicalls.h
@@ -22,7 +22,7 @@ static inline void xen_mc_batch(void)
 	unsigned long flags;
 	/* need to disable interrupts until this entry is complete */
 	local_irq_save(flags);
-	__get_cpu_var(xen_mc_irq_flags) = flags;
+	__this_cpu_write(xen_mc_irq_flags, flags);
 }
 
 static inline struct multicall_space xen_mc_entry(size_t args)
diff --git a/arch/x86/xen/spinlock.c b/arch/x86/xen/spinlock.c
index 23e061b..cc9b1e1 100644
--- a/arch/x86/xen/spinlock.c
+++ b/arch/x86/xen/spinlock.c
@@ -159,8 +159,8 @@ static inline struct xen_spinlock *spinning_lock(struct xen_spinlock *xl)
 {
 	struct xen_spinlock *prev;
 
-	prev = __get_cpu_var(lock_spinners);
-	__get_cpu_var(lock_spinners) = xl;
+	prev = __this_cpu_read(lock_spinners);
+	__this_cpu_write(lock_spinners, xl);
 
 	wmb();			/* set lock of interest before count */
 
@@ -179,14 +179,14 @@ static inline void unspinning_lock(struct xen_spinlock *xl, struct xen_spinlock
 	asm(LOCK_PREFIX " decw %0"
 	    : "+m" (xl->spinners) : : "memory");
 	wmb();			/* decrement count before restoring lock */
-	__get_cpu_var(lock_spinners) = prev;
+	__this_cpu_write(lock_spinners, prev);
 }
 
 static noinline int xen_spin_lock_slow(struct arch_spinlock *lock, bool irq_enable)
 {
 	struct xen_spinlock *xl = (struct xen_spinlock *)lock;
 	struct xen_spinlock *prev;
-	int irq = __get_cpu_var(lock_kicker_irq);
+	int irq = __this_cpu_read(lock_kicker_irq);
 	int ret;
 	u64 start;
 
diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
index b2bb5aa3..ef8930f 100644
--- a/arch/x86/xen/time.c
+++ b/arch/x86/xen/time.c
@@ -135,24 +135,24 @@ static void do_stolen_accounting(void)
 
 	/* Add the appropriate number of ticks of stolen time,
 	   including any left-overs from last time. */
-	stolen = runnable + offline + __get_cpu_var(xen_residual_stolen);
+	stolen = runnable + offline + __this_cpu_read(xen_residual_stolen);
 
 	if (stolen < 0)
 		stolen = 0;
 
 	ticks = iter_div_u64_rem(stolen, NS_PER_TICK, &stolen);
-	__get_cpu_var(xen_residual_stolen) = stolen;
+	__this_cpu_write(xen_residual_stolen, stolen);
 	account_steal_ticks(ticks);
 
 	/* Add the appropriate number of ticks of blocked time,
 	   including any left-overs from last time. */
-	blocked += __get_cpu_var(xen_residual_blocked);
+	blocked += __this_cpu_read(xen_residual_blocked);
 
 	if (blocked < 0)
 		blocked = 0;
 
 	ticks = iter_div_u64_rem(blocked, NS_PER_TICK, &blocked);
-	__get_cpu_var(xen_residual_blocked) = blocked;
+	__this_cpu_write(xen_residual_blocked, blocked);
 	account_idle_ticks(ticks);
 }
 
-- 
cgit v1.1


From 8f1d97c79eb65de1d05799d6b81d79cd94169114 Mon Sep 17 00:00:00 2001
From: Christoph Lameter <cl@linux.com>
Date: Mon, 6 Dec 2010 11:40:00 -0600
Subject: x86: Support for this_cpu_add, sub, dec, inc_return

Supply an implementation for x86 in order to generate more efficient code.

V2->V3:
	- Cleanup
	- Remove strange type checking from percpu_add_return_op.

tj: - Dropped unused typedef from percpu_add_return_op().
    - Renamed ret__ to paro_ret__ in percpu_add_return_op().
    - Minor indentation adjustments.

Acked-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/percpu.h | 43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)

(limited to 'arch')

diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index f899e01..38f9e96 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -177,6 +177,39 @@ do {									\
 	}								\
 } while (0)
 
+/*
+ * 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__;							\
+})
+
 #define percpu_from_op(op, var, constraint)		\
 ({							\
 	typeof(var) pfo_ret__;				\
@@ -300,6 +333,14 @@ do {									\
 #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)
 
+#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_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)
+#endif
 /*
  * Per cpu atomic 64 bit operations are only available under 64 bit.
  * 32 bit must fall back to generic operations.
@@ -324,6 +365,8 @@ do {									\
 #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)
 
+#define __this_cpu_add_return_8(pcp, val) percpu_add_return_op(pcp, val)
+#define this_cpu_add_return_8(pcp, val)	percpu_add_return_op(pcp, val)
 #endif
 
 /* This is not atomic against other CPUs -- CPU preemption needs to be off */
-- 
cgit v1.1


From 403047754cf690b012369b8fb563b738b88086e6 Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Fri, 17 Dec 2010 15:47:04 +0100
Subject: percpu,x86: relocate this_cpu_add_return() and friends

- include/linux/percpu.h: this_cpu_add_return() and friends were
  located next to __this_cpu_add_return().  However, the overall
  organization is to first group by preemption safeness.  Relocate
  this_cpu_add_return() and friends to preemption-safe area.

- arch/x86/include/asm/percpu.h: Relocate percpu_add_return_op() after
  other more basic operations.  Relocate [__]this_cpu_add_return_8()
  so that they're first grouped by preemption safeness.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
---
 arch/x86/include/asm/percpu.h | 71 +++++++++++++++++++++----------------------
 1 file changed, 35 insertions(+), 36 deletions(-)

(limited to 'arch')

diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index 38f9e96..dd0cd4b 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -177,39 +177,6 @@ do {									\
 	}								\
 } while (0)
 
-/*
- * 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__;							\
-})
-
 #define percpu_from_op(op, var, constraint)		\
 ({							\
 	typeof(var) pfo_ret__;				\
@@ -263,6 +230,39 @@ 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__;							\
+})
+
+/*
  * 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
@@ -352,6 +352,7 @@ 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)
@@ -359,14 +360,12 @@ 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)
-
-#define __this_cpu_add_return_8(pcp, val) percpu_add_return_op(pcp, val)
-#define this_cpu_add_return_8(pcp, val)	percpu_add_return_op(pcp, val)
 #endif
 
 /* This is not atomic against other CPUs -- CPU preemption needs to be off */
-- 
cgit v1.1


From 7296e08abac0a22a2534a4f6e493c764f2c77583 Mon Sep 17 00:00:00 2001
From: Christoph Lameter <cl@linux.com>
Date: Tue, 14 Dec 2010 10:28:44 -0600
Subject: x86: this_cpu_cmpxchg and this_cpu_xchg operations

Provide support as far as the hardware capabilities of the x86 cpus
allow.

Define CONFIG_CMPXCHG_LOCAL in Kconfig.cpu to allow core code to test for
fast cpuops implementations.

V1->V2:
	- Take out the definition for this_cpu_cmpxchg_8 and move it into
	  a separate patch.

tj: - Reordered ops to better follow this_cpu_* organization.
    - Renamed macro temp variables similar to their existing
      neighbours.

Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/Kconfig.cpu          |   3 ++
 arch/x86/include/asm/percpu.h | 107 +++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 109 insertions(+), 1 deletion(-)

(limited to 'arch')

diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu
index 2ac9069..15588a0 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/percpu.h b/arch/x86/include/asm/percpu.h
index dd0cd4b..b85ade5 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -263,6 +263,83 @@ do {									\
 })
 
 /*
+ * Beware: xchg on x86 has an implied lock prefix. There will be the cost of
+ * full lock semantics even though they are not needed.
+ */
+#define percpu_xchg_op(var, nval)					\
+({									\
+	typeof(var) pxo_ret__;						\
+	typeof(var) pxo_new__ = (nval);					\
+	switch (sizeof(var)) {						\
+	case 1:								\
+		asm("xchgb %2, "__percpu_arg(1)				\
+			    : "=a" (pxo_ret__), "+m" (var)		\
+			    : "q" (pxo_new__)				\
+			    : "memory");				\
+		break;							\
+	case 2:								\
+		asm("xchgw %2, "__percpu_arg(1)				\
+			    : "=a" (pxo_ret__), "+m" (var)		\
+			    : "r" (pxo_new__)				\
+			    : "memory");				\
+		break;							\
+	case 4:								\
+		asm("xchgl %2, "__percpu_arg(1)				\
+			    : "=a" (pxo_ret__), "+m" (var)		\
+			    : "r" (pxo_new__)				\
+			    : "memory");				\
+		break;							\
+	case 8:								\
+		asm("xchgq %2, "__percpu_arg(1)				\
+			    : "=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
@@ -300,6 +377,12 @@ 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))
@@ -319,6 +402,11 @@ 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)
@@ -332,15 +420,32 @@ 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)
-#endif
+#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.
  * 32 bit must fall back to generic operations.
-- 
cgit v1.1


From 8270137a0d50507a5b40f880db636527045b8466 Mon Sep 17 00:00:00 2001
From: Christoph Lameter <cl@linux.com>
Date: Tue, 14 Dec 2010 10:28:47 -0600
Subject: cpuops: Use cmpxchg for xchg to avoid lock semantics

Use cmpxchg instead of xchg to realize this_cpu_xchg.

xchg will cause LOCK overhead since LOCK is always implied but cmpxchg
will not.

Baselines:

xchg()		= 18 cycles (no segment prefix, LOCK semantics)
__this_cpu_xchg = 1 cycle

(simulated using this_cpu_read/write, two prefixes. Looks like the
cpu can use loop optimization to get rid of most of the overhead)

Cycles before:

this_cpu_xchg	 = 37 cycles (segment prefix and LOCK (implied by xchg))

After:

this_cpu_xchg	= 11 cycle (using cmpxchg without lock semantics)

Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/percpu.h | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

(limited to 'arch')

diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index b85ade5..8ee4516 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -263,8 +263,9 @@ do {									\
 })
 
 /*
- * Beware: xchg on x86 has an implied lock prefix. There will be the cost of
- * full lock semantics even though they are not needed.
+ * 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)					\
 ({									\
@@ -272,25 +273,33 @@ do {									\
 	typeof(var) pxo_new__ = (nval);					\
 	switch (sizeof(var)) {						\
 	case 1:								\
-		asm("xchgb %2, "__percpu_arg(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("xchgw %2, "__percpu_arg(1)				\
+		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("xchgl %2, "__percpu_arg(1)				\
+		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("xchgq %2, "__percpu_arg(1)				\
+		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");				\
-- 
cgit v1.1


From 0a3aee0da4402aa19b66e458038533c896fb80c6 Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Sat, 18 Dec 2010 16:28:55 +0100
Subject: x86: Use this_cpu_ops to optimize code

Go through x86 code and replace __get_cpu_var and get_cpu_var
instances that refer to a scalar and are not used for address
determinations.

Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/debugreg.h           |  2 +-
 arch/x86/kernel/apic/io_apic.c            |  4 ++--
 arch/x86/kernel/apic/nmi.c                | 24 ++++++++++++------------
 arch/x86/kernel/apic/x2apic_uv_x.c        |  8 ++++----
 arch/x86/kernel/cpu/cpufreq/powernow-k8.c |  2 +-
 arch/x86/kernel/cpu/mcheck/mce.c          |  6 +++---
 arch/x86/kernel/cpu/perf_event.c          | 27 +++++++++++----------------
 arch/x86/kernel/cpu/perf_event_intel.c    |  4 ++--
 arch/x86/kernel/ftrace.c                  |  6 +++---
 arch/x86/kernel/hw_breakpoint.c           | 12 ++++++------
 arch/x86/kernel/irq.c                     |  6 +++---
 arch/x86/kernel/irq_32.c                  |  4 ++--
 arch/x86/kernel/smpboot.c                 |  2 +-
 arch/x86/kernel/tsc.c                     |  2 +-
 arch/x86/kvm/x86.c                        |  8 ++++----
 arch/x86/oprofile/nmi_int.c               |  2 +-
 16 files changed, 57 insertions(+), 62 deletions(-)

(limited to 'arch')

diff --git a/arch/x86/include/asm/debugreg.h b/arch/x86/include/asm/debugreg.h
index b81002f..078ad0c 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/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index 7cc0a72..8d50922 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -2302,7 +2302,7 @@ 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;
@@ -2336,7 +2336,7 @@ 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/nmi.c b/arch/x86/kernel/apic/nmi.c
index c90041c..b387dce 100644
--- a/arch/x86/kernel/apic/nmi.c
+++ b/arch/x86/kernel/apic/nmi.c
@@ -306,12 +306,12 @@ void acpi_nmi_disable(void)
  */
 void cpu_nmi_set_wd_enabled(void)
 {
-	__get_cpu_var(wd_enabled) = 1;
+	__this_cpu_write(wd_enabled, 1);
 }
 
 void setup_apic_nmi_watchdog(void *unused)
 {
-	if (__get_cpu_var(wd_enabled))
+	if (__this_cpu_read(wd_enabled))
 		return;
 
 	/* cheap hack to support suspend/resume */
@@ -322,12 +322,12 @@ void setup_apic_nmi_watchdog(void *unused)
 	switch (nmi_watchdog) {
 	case NMI_LOCAL_APIC:
 		if (lapic_watchdog_init(nmi_hz) < 0) {
-			__get_cpu_var(wd_enabled) = 0;
+			__this_cpu_write(wd_enabled, 0);
 			return;
 		}
 		/* FALL THROUGH */
 	case NMI_IO_APIC:
-		__get_cpu_var(wd_enabled) = 1;
+		__this_cpu_write(wd_enabled, 1);
 		atomic_inc(&nmi_active);
 	}
 }
@@ -337,13 +337,13 @@ void stop_apic_nmi_watchdog(void *unused)
 	/* only support LOCAL and IO APICs for now */
 	if (!nmi_watchdog_active())
 		return;
-	if (__get_cpu_var(wd_enabled) == 0)
+	if (__this_cpu_read(wd_enabled) == 0)
 		return;
 	if (nmi_watchdog == NMI_LOCAL_APIC)
 		lapic_watchdog_stop();
 	else
 		__acpi_nmi_disable(NULL);
-	__get_cpu_var(wd_enabled) = 0;
+	__this_cpu_write(wd_enabled, 0);
 	atomic_dec(&nmi_active);
 }
 
@@ -403,8 +403,8 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason)
 
 	sum = get_timer_irqs(cpu);
 
-	if (__get_cpu_var(nmi_touch)) {
-		__get_cpu_var(nmi_touch) = 0;
+	if (__this_cpu_read(nmi_touch)) {
+		__this_cpu_write(nmi_touch, 0);
 		touched = 1;
 	}
 
@@ -427,7 +427,7 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason)
 		touched = 1;
 
 	/* if the none of the timers isn't firing, this cpu isn't doing much */
-	if (!touched && __get_cpu_var(last_irq_sum) == sum) {
+	if (!touched && __this_cpu_read(last_irq_sum) == sum) {
 		/*
 		 * Ayiee, looks like this CPU is stuck ...
 		 * wait a few IRQs (5 seconds) before doing the oops ...
@@ -440,12 +440,12 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason)
 			die_nmi("BUG: NMI Watchdog detected LOCKUP",
 				regs, panic_on_timeout);
 	} else {
-		__get_cpu_var(last_irq_sum) = sum;
+		__this_cpu_write(last_irq_sum, sum);
 		__this_cpu_write(alert_counter, 0);
 	}
 
 	/* see if the nmi watchdog went off */
-	if (!__get_cpu_var(wd_enabled))
+	if (!__this_cpu_read(wd_enabled))
 		return rc;
 	switch (nmi_watchdog) {
 	case NMI_LOCAL_APIC:
@@ -467,7 +467,7 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason)
 
 static void enable_ioapic_nmi_watchdog_single(void *unused)
 {
-	__get_cpu_var(wd_enabled) = 1;
+	__this_cpu_write(wd_enabled, 1);
 	atomic_inc(&nmi_active);
 	__acpi_nmi_enable(NULL);
 }
diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c
index c1c52c3..26ec9a7 100644
--- a/arch/x86/kernel/apic/x2apic_uv_x.c
+++ b/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -118,8 +118,8 @@ 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) =
-				nodeid << (uvh_apicid.s.pnode_shift - 1);
+			__this_cpu_write(x2apic_extra_bits,
+				nodeid << (uvh_apicid.s.pnode_shift - 1));
 			uv_system_type = UV_NON_UNIQUE_APIC;
 			uv_set_apicid_hibit();
 			return 1;
@@ -284,7 +284,7 @@ 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;
 }
@@ -376,7 +376,7 @@ 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/cpufreq/powernow-k8.c b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
index 491977b..42a3604 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
@@ -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/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index 7a35b72..0c746af 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)
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 6d75b91..ba85814 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -968,8 +968,7 @@ 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_PERFMON_EVENTSEL_ENABLE);
 }
@@ -1243,7 +1242,7 @@ perf_event_nmi_handler(struct notifier_block *self,
 		break;
 	case DIE_NMIUNKNOWN:
 		this_nmi = percpu_read(irq_stat.__nmi_count);
-		if (this_nmi != __get_cpu_var(pmu_nmi).marked)
+		if (this_nmi != __this_cpu_read(pmu_nmi.marked))
 			/* let the kernel handle the unknown nmi */
 			return NOTIFY_DONE;
 		/*
@@ -1267,8 +1266,8 @@ perf_event_nmi_handler(struct notifier_block *self,
 	this_nmi = percpu_read(irq_stat.__nmi_count);
 	if ((handled > 1) ||
 		/* the next nmi could be a back-to-back nmi */
-	    ((__get_cpu_var(pmu_nmi).marked == this_nmi) &&
-	     (__get_cpu_var(pmu_nmi).handled > 1))) {
+	    ((__this_cpu_read(pmu_nmi.marked) == this_nmi) &&
+	     (__this_cpu_read(pmu_nmi.handled) > 1))) {
 		/*
 		 * We could have two subsequent back-to-back nmis: The
 		 * first handles more than one counter, the 2nd
@@ -1279,8 +1278,8 @@ perf_event_nmi_handler(struct notifier_block *self,
 		 * handling more than one counter. We will mark the
 		 * next (3rd) and then drop it if unhandled.
 		 */
-		__get_cpu_var(pmu_nmi).marked	= this_nmi + 1;
-		__get_cpu_var(pmu_nmi).handled	= handled;
+		__this_cpu_write(pmu_nmi.marked, this_nmi + 1);
+		__this_cpu_write(pmu_nmi.handled, handled);
 	}
 
 	return NOTIFY_STOP;
@@ -1454,11 +1453,9 @@ static inline void x86_pmu_read(struct perf_event *event)
  */
 static void x86_pmu_start_txn(struct pmu *pmu)
 {
-	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
-
 	perf_pmu_disable(pmu);
-	cpuc->group_flag |= PERF_EVENT_TXN;
-	cpuc->n_txn = 0;
+	__this_cpu_or(cpu_hw_events.group_flag, PERF_EVENT_TXN);
+	__this_cpu_write(cpu_hw_events.n_txn, 0);
 }
 
 /*
@@ -1468,14 +1465,12 @@ static void x86_pmu_start_txn(struct pmu *pmu)
  */
 static void x86_pmu_cancel_txn(struct pmu *pmu)
 {
-	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
-
-	cpuc->group_flag &= ~PERF_EVENT_TXN;
+	__this_cpu_and(cpu_hw_events.group_flag, ~PERF_EVENT_TXN);
 	/*
 	 * Truncate the collected events.
 	 */
-	cpuc->n_added -= cpuc->n_txn;
-	cpuc->n_events -= cpuc->n_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));
 	perf_pmu_enable(pmu);
 }
 
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index c8f5c08..4ee59bc 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -649,7 +649,7 @@ static void intel_pmu_enable_event(struct perf_event *event)
 	struct hw_perf_event *hwc = &event->hw;
 
 	if (unlikely(hwc->idx == X86_PMC_IDX_FIXED_BTS)) {
-		if (!__get_cpu_var(cpu_hw_events).enabled)
+		if (!__this_cpu_read(cpu_hw_events.enabled))
 			return;
 
 		intel_pmu_enable_bts(hwc->config);
@@ -679,7 +679,7 @@ static int intel_pmu_save_and_restart(struct perf_event *event)
 
 static void intel_pmu_reset(void)
 {
-	struct debug_store *ds = __get_cpu_var(cpu_hw_events).ds;
+	struct debug_store *ds = __this_cpu_read(cpu_hw_events.ds);
 	unsigned long flags;
 	int idx;
 
diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index 3afb33f..b45246f 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -167,9 +167,9 @@ static void ftrace_mod_code(void)
 
 void ftrace_nmi_enter(void)
 {
-	__get_cpu_var(save_modifying_code) = modifying_code;
+	__this_cpu_write(save_modifying_code, modifying_code);
 
-	if (!__get_cpu_var(save_modifying_code))
+	if (!__this_cpu_read(save_modifying_code))
 		return;
 
 	if (atomic_inc_return(&nmi_running) & MOD_CODE_WRITE_FLAG) {
@@ -183,7 +183,7 @@ void ftrace_nmi_enter(void)
 
 void ftrace_nmi_exit(void)
 {
-	if (!__get_cpu_var(save_modifying_code))
+	if (!__this_cpu_read(save_modifying_code))
 		return;
 
 	/* Finish all executions before clearing nmi_running */
diff --git a/arch/x86/kernel/hw_breakpoint.c b/arch/x86/kernel/hw_breakpoint.c
index 42c5942..02f0763 100644
--- a/arch/x86/kernel/hw_breakpoint.c
+++ b/arch/x86/kernel/hw_breakpoint.c
@@ -122,7 +122,7 @@ int arch_install_hw_breakpoint(struct perf_event *bp)
 		return -EBUSY;
 
 	set_debugreg(info->address, i);
-	__get_cpu_var(cpu_debugreg[i]) = info->address;
+	__this_cpu_write(cpu_debugreg[i], info->address);
 
 	dr7 = &__get_cpu_var(cpu_dr7);
 	*dr7 |= encode_dr7(i, info->len, info->type);
@@ -397,12 +397,12 @@ void flush_ptrace_hw_breakpoint(struct task_struct *tsk)
 
 void hw_breakpoint_restore(void)
 {
-	set_debugreg(__get_cpu_var(cpu_debugreg[0]), 0);
-	set_debugreg(__get_cpu_var(cpu_debugreg[1]), 1);
-	set_debugreg(__get_cpu_var(cpu_debugreg[2]), 2);
-	set_debugreg(__get_cpu_var(cpu_debugreg[3]), 3);
+	set_debugreg(__this_cpu_read(cpu_debugreg[0]), 0);
+	set_debugreg(__this_cpu_read(cpu_debugreg[1]), 1);
+	set_debugreg(__this_cpu_read(cpu_debugreg[2]), 2);
+	set_debugreg(__this_cpu_read(cpu_debugreg[3]), 3);
 	set_debugreg(current->thread.debugreg6, 6);
-	set_debugreg(__get_cpu_var(cpu_dr7), 7);
+	set_debugreg(__this_cpu_read(cpu_dr7), 7);
 }
 EXPORT_SYMBOL_GPL(hw_breakpoint_restore);
 
diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
index 83ec017..3a43caa 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -234,7 +234,7 @@ unsigned int __irq_entry do_IRQ(struct pt_regs *regs)
 	exit_idle();
 	irq_enter();
 
-	irq = __get_cpu_var(vector_irq)[vector];
+	irq = __this_cpu_read(vector_irq[vector]);
 
 	if (!handle_irq(irq, regs)) {
 		ack_APIC_irq();
@@ -350,12 +350,12 @@ void fixup_irqs(void)
 	for (vector = FIRST_EXTERNAL_VECTOR; vector < NR_VECTORS; vector++) {
 		unsigned int irr;
 
-		if (__get_cpu_var(vector_irq)[vector] < 0)
+		if (__this_cpu_read(vector_irq[vector]) < 0)
 			continue;
 
 		irr = apic_read(APIC_IRR + (vector / 32 * 0x10));
 		if (irr  & (1 << (vector % 32))) {
-			irq = __get_cpu_var(vector_irq)[vector];
+			irq = __this_cpu_read(vector_irq[vector]);
 
 			data = irq_get_irq_data(irq);
 			raw_spin_lock(&desc->lock);
diff --git a/arch/x86/kernel/irq_32.c b/arch/x86/kernel/irq_32.c
index 96656f2..48ff6dc 100644
--- a/arch/x86/kernel/irq_32.c
+++ b/arch/x86/kernel/irq_32.c
@@ -79,7 +79,7 @@ execute_on_irq_stack(int overflow, struct irq_desc *desc, int irq)
 	u32 *isp, arg1, arg2;
 
 	curctx = (union irq_ctx *) current_thread_info();
-	irqctx = __get_cpu_var(hardirq_ctx);
+	irqctx = __this_cpu_read(hardirq_ctx);
 
 	/*
 	 * this is where we switch to the IRQ stack. However, if we are
@@ -166,7 +166,7 @@ asmlinkage void do_softirq(void)
 
 	if (local_softirq_pending()) {
 		curctx = current_thread_info();
-		irqctx = __get_cpu_var(softirq_ctx);
+		irqctx = __this_cpu_read(softirq_ctx);
 		irqctx->tinfo.task = curctx->task;
 		irqctx->tinfo.previous_esp = current_stack_pointer;
 
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 083e99d..ff4e5a1 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -1377,7 +1377,7 @@ void play_dead_common(void)
 
 	mb();
 	/* Ack it */
-	__get_cpu_var(cpu_state) = CPU_DEAD;
+	__this_cpu_write(cpu_state, CPU_DEAD);
 
 	/*
 	 * With physical CPU hotplug, we should halt the cpu
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 0c40d8b..acb08dd 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -659,7 +659,7 @@ void restore_sched_clock_state(void)
 
 	local_irq_save(flags);
 
-	__get_cpu_var(cyc2ns_offset) = 0;
+	__this_cpu_write(cyc2ns_offset, 0);
 	offset = cyc2ns_suspend - sched_clock();
 
 	for_each_possible_cpu(cpu)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index cdac9e5..79d9606 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -981,7 +981,7 @@ static inline u64 nsec_to_cycles(u64 nsec)
 	if (kvm_tsc_changes_freq())
 		printk_once(KERN_WARNING
 		 "kvm: unreliable cycle conversion on adjustable rate TSC\n");
-	ret = nsec * __get_cpu_var(cpu_tsc_khz);
+	ret = nsec * __this_cpu_read(cpu_tsc_khz);
 	do_div(ret, USEC_PER_SEC);
 	return ret;
 }
@@ -1066,7 +1066,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
 	local_irq_save(flags);
 	kvm_get_msr(v, MSR_IA32_TSC, &tsc_timestamp);
 	kernel_ns = get_kernel_ns();
-	this_tsc_khz = __get_cpu_var(cpu_tsc_khz);
+	this_tsc_khz = __this_cpu_read(cpu_tsc_khz);
 
 	if (unlikely(this_tsc_khz == 0)) {
 		local_irq_restore(flags);
@@ -4432,7 +4432,7 @@ EXPORT_SYMBOL_GPL(kvm_fast_pio_out);
 
 static void tsc_bad(void *info)
 {
-	__get_cpu_var(cpu_tsc_khz) = 0;
+	__this_cpu_write(cpu_tsc_khz, 0);
 }
 
 static void tsc_khz_changed(void *data)
@@ -4446,7 +4446,7 @@ static void tsc_khz_changed(void *data)
 		khz = cpufreq_quick_get(raw_smp_processor_id());
 	if (!khz)
 		khz = tsc_khz;
-	__get_cpu_var(cpu_tsc_khz) = khz;
+	__this_cpu_write(cpu_tsc_khz, khz);
 }
 
 static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c
index 4e8baad..a0cae67 100644
--- a/arch/x86/oprofile/nmi_int.c
+++ b/arch/x86/oprofile/nmi_int.c
@@ -143,7 +143,7 @@ static inline int has_mux(void)
 
 inline int op_x86_phys_to_virt(int phys)
 {
-	return __get_cpu_var(switch_index) + phys;
+	return __this_cpu_read(switch_index) + phys;
 }
 
 inline int op_x86_virt_to_phys(int virt)
-- 
cgit v1.1


From 7b543a5334ff4ea2e3ad3b777fc23cdb8072a988 Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Sat, 18 Dec 2010 16:30:05 +0100
Subject: x86: Replace uses of current_cpu_data with this_cpu ops

Replace all uses of current_cpu_data with this_cpu operations on the
per cpu structure cpu_info.  The scala accesses are replaced with the
matching this_cpu ops which results in smaller and more efficient
code.

In the long run, it might be a good idea to remove cpu_data() macro
too and use per_cpu macro directly.

tj: updated description

Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/processor.h          |  3 +--
 arch/x86/kernel/apic/apic.c               |  2 +-
 arch/x86/kernel/cpu/amd.c                 |  2 +-
 arch/x86/kernel/cpu/cpufreq/powernow-k8.c |  2 +-
 arch/x86/kernel/cpu/intel_cacheinfo.c     |  4 ++--
 arch/x86/kernel/cpu/mcheck/mce.c          | 14 +++++++-------
 arch/x86/kernel/cpu/mcheck/mce_intel.c    |  2 +-
 arch/x86/kernel/process.c                 |  4 ++--
 arch/x86/kernel/smpboot.c                 | 12 ++++++------
 arch/x86/oprofile/op_model_ppro.c         |  8 ++++----
 10 files changed, 26 insertions(+), 27 deletions(-)

(limited to 'arch')

diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index cae9c3c..c6efecf 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 3f838d5..8accfe3 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(&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;
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 9e093f8..7c7bedb 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 = &current_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 42a3604..35c7e65 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);
diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c
index 17ad033..453c616 100644
--- a/arch/x86/kernel/cpu/intel_cacheinfo.c
+++ b/arch/x86/kernel/cpu/intel_cacheinfo.c
@@ -266,7 +266,7 @@ 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)
@@ -288,7 +288,7 @@ 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 0c746af..d916183 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -1159,7 +1159,7 @@ static void mce_start_timer(unsigned long data)
 
 	WARN_ON(smp_processor_id() != data);
 
-	if (mce_available(&current_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(&current_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(&current_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(&current_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(&current_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(&current_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(&current_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 6fcd093..8694ef56 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(&current_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/process.c b/arch/x86/kernel/process.c
index 57d1868..dae1c07 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -445,7 +445,7 @@ void mwait_idle_with_hints(unsigned long ax, unsigned long cx)
 {
 	trace_power_start(POWER_CSTATE, (ax>>4)+1, smp_processor_id());
 	if (!need_resched()) {
-		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))
+		if (cpu_has(__this_cpu_ptr(&cpu_info), X86_FEATURE_CLFLUSH_MONITOR))
 			clflush((void *)&current_thread_info()->flags);
 
 		__monitor((void *)&current_thread_info()->flags, 0, 0);
@@ -460,7 +460,7 @@ static void mwait_idle(void)
 {
 	if (!need_resched()) {
 		trace_power_start(POWER_CSTATE, 1, smp_processor_id());
-		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))
+		if (cpu_has(__this_cpu_ptr(&cpu_info), X86_FEATURE_CLFLUSH_MONITOR))
 			clflush((void *)&current_thread_info()->flags);
 
 		__monitor((void *)&current_thread_info()->flags, 0, 0);
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index ff4e5a1..0720071 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -430,7 +430,7 @@ void __cpuinit set_cpu_sibling_map(int cpu)
 
 	cpumask_set_cpu(cpu, c->llc_shared_map);
 
-	if (current_cpu_data.x86_max_cores == 1) {
+	if (__this_cpu_read(cpu_info.x86_max_cores) == 1) {
 		cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu));
 		c->booted_cores = 1;
 		return;
@@ -1094,7 +1094,7 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
 
 	preempt_disable();
 	smp_cpu_index_default();
-	current_cpu_data = boot_cpu_data;
+	memcpy(__this_cpu_ptr(&cpu_info), &boot_cpu_data, sizeof(cpu_info));
 	cpumask_copy(cpu_callin_mask, cpumask_of(0));
 	mb();
 	/*
@@ -1397,11 +1397,11 @@ static inline void mwait_play_dead(void)
 	int i;
 	void *mwait_ptr;
 
-	if (!cpu_has(&current_cpu_data, X86_FEATURE_MWAIT))
+	if (!cpu_has(__this_cpu_ptr(&cpu_info), X86_FEATURE_MWAIT))
 		return;
-	if (!cpu_has(&current_cpu_data, X86_FEATURE_CLFLSH))
+	if (!cpu_has(__this_cpu_ptr(&cpu_info), X86_FEATURE_CLFLSH))
 		return;
-	if (current_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
+	if (__this_cpu_read(cpu_info.cpuid_level) < CPUID_MWAIT_LEAF)
 		return;
 
 	eax = CPUID_MWAIT_LEAF;
@@ -1452,7 +1452,7 @@ static inline void mwait_play_dead(void)
 
 static inline void hlt_play_dead(void)
 {
-	if (current_cpu_data.x86 >= 4)
+	if (__this_cpu_read(cpu_info.x86) >= 4)
 		wbinvd();
 
 	while (1) {
diff --git a/arch/x86/oprofile/op_model_ppro.c b/arch/x86/oprofile/op_model_ppro.c
index d769cda..94b7450 100644
--- a/arch/x86/oprofile/op_model_ppro.c
+++ b/arch/x86/oprofile/op_model_ppro.c
@@ -95,8 +95,8 @@ static void ppro_setup_ctrs(struct op_x86_model_spec const *model,
 		 * counter width:
 		 */
 		if (!(eax.split.version_id == 0 &&
-			current_cpu_data.x86 == 6 &&
-				current_cpu_data.x86_model == 15)) {
+			__this_cpu_read(cpu_info.x86) == 6 &&
+				__this_cpu_read(cpu_info.x86_model) == 15)) {
 
 			if (counter_width < eax.split.bit_width)
 				counter_width = eax.split.bit_width;
@@ -235,8 +235,8 @@ static void arch_perfmon_setup_counters(void)
 	eax.full = cpuid_eax(0xa);
 
 	/* Workaround for BIOS bugs in 6/15. Taken from perfmon2 */
-	if (eax.split.version_id == 0 && current_cpu_data.x86 == 6 &&
-		current_cpu_data.x86_model == 15) {
+	if (eax.split.version_id == 0 && __this_cpu_read(cpu_info.x86) == 6 &&
+		__this_cpu_read(cpu_info.x86_model) == 15) {
 		eax.split.version_id = 2;
 		eax.split.num_counters = 2;
 		eax.split.bit_width = 40;
-- 
cgit v1.1


From c1955b5f3a95717ce1f5235f6e9968da068e3183 Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Sat, 18 Dec 2010 16:30:48 +0100
Subject: x86: Use this_cpu_inc_return for nmi counter

this_cpu_inc_return() saves us a memory access there.

Reviewed-by: Pekka Enberg <penberg@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/kernel/apic/nmi.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

(limited to 'arch')

diff --git a/arch/x86/kernel/apic/nmi.c b/arch/x86/kernel/apic/nmi.c
index b387dce..37769cc 100644
--- a/arch/x86/kernel/apic/nmi.c
+++ b/arch/x86/kernel/apic/nmi.c
@@ -432,8 +432,7 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason)
 		 * Ayiee, looks like this CPU is stuck ...
 		 * wait a few IRQs (5 seconds) before doing the oops ...
 		 */
-		__this_cpu_inc(alert_counter);
-		if (__this_cpu_read(alert_counter) == 5 * nmi_hz)
+		if (__this_cpu_inc_return(alert_counter) == 5 * nmi_hz)
 			/*
 			 * die_nmi will return ONLY if NOTIFY_STOP happens..
 			 */
-- 
cgit v1.1


From 357089fca91f639dd005ae0721f5f932b4f276ab Mon Sep 17 00:00:00 2001
From: Christoph Lameter <cl@linux.com>
Date: Thu, 16 Dec 2010 12:14:43 -0600
Subject: x86: udelay: Use this_cpu_read to avoid address calculation

The code will use a segment prefix instead of doing the lookup and
calculation.

Signed-off-by: Christoph Lameter <cl@linux.com>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/lib/delay.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'arch')

diff --git a/arch/x86/lib/delay.c b/arch/x86/lib/delay.c
index ff485d3..fc45ba8 100644
--- a/arch/x86/lib/delay.c
+++ b/arch/x86/lib/delay.c
@@ -121,7 +121,7 @@ inline void __const_udelay(unsigned long xloops)
 	asm("mull %%edx"
 		:"=d" (xloops), "=&a" (d0)
 		:"1" (xloops), "0"
-		(cpu_data(raw_smp_processor_id()).loops_per_jiffy * (HZ/4)));
+		(this_cpu_read(cpu_info.loops_per_jiffy) * (HZ/4)));
 
 	__delay(++xloops);
 }
-- 
cgit v1.1