From 1b437c8c73a36daa471dd54a63c426d72af5723d Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:57 +0900
Subject: [PATCH 01/12] x86-64: Move irq stats from PDA to per-cpu and
 consolidate with 32-bit.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/hardirq_64.h | 24 +++++++++++++++++++-----
 arch/x86/include/asm/pda.h        | 10 ----------
 arch/x86/kernel/irq.c             |  6 +-----
 arch/x86/kernel/irq_64.c          |  3 +++
 arch/x86/kernel/nmi.c             | 10 +---------
 arch/x86/xen/smp.c                | 18 +++---------------
 6 files changed, 27 insertions(+), 44 deletions(-)

diff --git a/arch/x86/include/asm/hardirq_64.h b/arch/x86/include/asm/hardirq_64.h
index b5a6b5d56704..a65bab20f6ce 100644
--- a/arch/x86/include/asm/hardirq_64.h
+++ b/arch/x86/include/asm/hardirq_64.h
@@ -3,22 +3,36 @@
 
 #include <linux/threads.h>
 #include <linux/irq.h>
-#include <asm/pda.h>
 #include <asm/apic.h>
 
+typedef struct {
+	unsigned int __softirq_pending;
+	unsigned int __nmi_count;	/* arch dependent */
+	unsigned int apic_timer_irqs;	/* arch dependent */
+	unsigned int irq0_irqs;
+	unsigned int irq_resched_count;
+	unsigned int irq_call_count;
+	unsigned int irq_tlb_count;
+	unsigned int irq_thermal_count;
+	unsigned int irq_spurious_count;
+	unsigned int irq_threshold_count;
+} ____cacheline_aligned irq_cpustat_t;
+
+DECLARE_PER_CPU(irq_cpustat_t, irq_stat);
+
 /* We can have at most NR_VECTORS irqs routed to a cpu at a time */
 #define MAX_HARDIRQS_PER_CPU NR_VECTORS
 
 #define __ARCH_IRQ_STAT 1
 
-#define inc_irq_stat(member)	add_pda(member, 1)
+#define inc_irq_stat(member)	percpu_add(irq_stat.member, 1)
 
-#define local_softirq_pending() read_pda(__softirq_pending)
+#define local_softirq_pending() percpu_read(irq_stat.__softirq_pending)
 
 #define __ARCH_SET_SOFTIRQ_PENDING 1
 
-#define set_softirq_pending(x) write_pda(__softirq_pending, (x))
-#define or_softirq_pending(x)  or_pda(__softirq_pending, (x))
+#define set_softirq_pending(x) percpu_write(irq_stat.__softirq_pending, (x))
+#define or_softirq_pending(x)  percpu_or(irq_stat.__softirq_pending, (x))
 
 extern void ack_bad_irq(unsigned int irq);
 
diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 47f274fe6953..69a40757e217 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -25,19 +25,9 @@ struct x8664_pda {
 	char *irqstackptr;
 	short nodenumber;		/* number of current node (32k max) */
 	short in_bootmem;		/* pda lives in bootmem */
-	unsigned int __softirq_pending;
-	unsigned int __nmi_count;	/* number of NMI on this CPUs */
 	short mmu_state;
 	short isidle;
 	struct mm_struct *active_mm;
-	unsigned apic_timer_irqs;
-	unsigned irq0_irqs;
-	unsigned irq_resched_count;
-	unsigned irq_call_count;
-	unsigned irq_tlb_count;
-	unsigned irq_thermal_count;
-	unsigned irq_threshold_count;
-	unsigned irq_spurious_count;
 } ____cacheline_aligned_in_smp;
 
 DECLARE_PER_CPU(struct x8664_pda, __pda);
diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
index 3973e2df7f87..8b30d0c2512c 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -36,11 +36,7 @@ void ack_bad_irq(unsigned int irq)
 #endif
 }
 
-#ifdef CONFIG_X86_32
-# define irq_stats(x)		(&per_cpu(irq_stat, x))
-#else
-# define irq_stats(x)		cpu_pda(x)
-#endif
+#define irq_stats(x)		(&per_cpu(irq_stat, x))
 /*
  * /proc/interrupts printing:
  */
diff --git a/arch/x86/kernel/irq_64.c b/arch/x86/kernel/irq_64.c
index 0b21cb1ea11f..1db05247b47f 100644
--- a/arch/x86/kernel/irq_64.c
+++ b/arch/x86/kernel/irq_64.c
@@ -19,6 +19,9 @@
 #include <asm/io_apic.h>
 #include <asm/idle.h>
 
+DEFINE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat);
+EXPORT_PER_CPU_SYMBOL(irq_stat);
+
 /*
  * Probabilistic stack overflow check:
  *
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
index 7228979f1e7f..23b6d9e6e4f5 100644
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -61,11 +61,7 @@ static int endflag __initdata;
 
 static inline unsigned int get_nmi_count(int cpu)
 {
-#ifdef CONFIG_X86_64
-	return cpu_pda(cpu)->__nmi_count;
-#else
-	return nmi_count(cpu);
-#endif
+	return per_cpu(irq_stat, cpu).__nmi_count;
 }
 
 static inline int mce_in_progress(void)
@@ -82,12 +78,8 @@ static inline int mce_in_progress(void)
  */
 static inline unsigned int get_timer_irqs(int cpu)
 {
-#ifdef CONFIG_X86_64
-	return read_pda(apic_timer_irqs) + read_pda(irq0_irqs);
-#else
 	return per_cpu(irq_stat, cpu).apic_timer_irqs +
 		per_cpu(irq_stat, cpu).irq0_irqs;
-#endif
 }
 
 #ifdef CONFIG_SMP
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index 3bfd6dd0b47c..9ff3b0999cfb 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -50,11 +50,7 @@ static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id);
  */
 static irqreturn_t xen_reschedule_interrupt(int irq, void *dev_id)
 {
-#ifdef CONFIG_X86_32
-	__get_cpu_var(irq_stat).irq_resched_count++;
-#else
-	add_pda(irq_resched_count, 1);
-#endif
+	inc_irq_stat(irq_resched_count);
 
 	return IRQ_HANDLED;
 }
@@ -435,11 +431,7 @@ static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id)
 {
 	irq_enter();
 	generic_smp_call_function_interrupt();
-#ifdef CONFIG_X86_32
-	__get_cpu_var(irq_stat).irq_call_count++;
-#else
-	add_pda(irq_call_count, 1);
-#endif
+	inc_irq_stat(irq_call_count);
 	irq_exit();
 
 	return IRQ_HANDLED;
@@ -449,11 +441,7 @@ static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id)
 {
 	irq_enter();
 	generic_smp_call_function_single_interrupt();
-#ifdef CONFIG_X86_32
-	__get_cpu_var(irq_stat).irq_call_count++;
-#else
-	add_pda(irq_call_count, 1);
-#endif
+	inc_irq_stat(irq_call_count);
 	irq_exit();
 
 	return IRQ_HANDLED;

From 9eb912d1aa6b8106e06a73ea6702ec3dab0d6a1a Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:57 +0900
Subject: [PATCH 02/12] x86-64: Move TLB state from PDA to per-cpu and
 consolidate with 32-bit.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/mmu_context_64.h | 16 +++++++---------
 arch/x86/include/asm/pda.h            |  2 --
 arch/x86/include/asm/tlbflush.h       |  7 ++-----
 arch/x86/kernel/cpu/common.c          |  2 --
 arch/x86/kernel/tlb_32.c              | 12 ++----------
 arch/x86/kernel/tlb_64.c              | 13 ++++++++-----
 arch/x86/xen/mmu.c                    |  6 +-----
 7 files changed, 20 insertions(+), 38 deletions(-)

diff --git a/arch/x86/include/asm/mmu_context_64.h b/arch/x86/include/asm/mmu_context_64.h
index 677d36e9540a..c4572505ab3e 100644
--- a/arch/x86/include/asm/mmu_context_64.h
+++ b/arch/x86/include/asm/mmu_context_64.h
@@ -1,13 +1,11 @@
 #ifndef _ASM_X86_MMU_CONTEXT_64_H
 #define _ASM_X86_MMU_CONTEXT_64_H
 
-#include <asm/pda.h>
-
 static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk)
 {
 #ifdef CONFIG_SMP
-	if (read_pda(mmu_state) == TLBSTATE_OK)
-		write_pda(mmu_state, TLBSTATE_LAZY);
+	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK)
+		percpu_write(cpu_tlbstate.state, TLBSTATE_LAZY);
 #endif
 }
 
@@ -19,8 +17,8 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		/* stop flush ipis for the previous mm */
 		cpu_clear(cpu, prev->cpu_vm_mask);
 #ifdef CONFIG_SMP
-		write_pda(mmu_state, TLBSTATE_OK);
-		write_pda(active_mm, next);
+		percpu_write(cpu_tlbstate.state, TLBSTATE_OK);
+		percpu_write(cpu_tlbstate.active_mm, next);
 #endif
 		cpu_set(cpu, next->cpu_vm_mask);
 		load_cr3(next->pgd);
@@ -30,9 +28,9 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	}
 #ifdef CONFIG_SMP
 	else {
-		write_pda(mmu_state, TLBSTATE_OK);
-		if (read_pda(active_mm) != next)
-			BUG();
+		percpu_write(cpu_tlbstate.state, TLBSTATE_OK);
+		BUG_ON(percpu_read(cpu_tlbstate.active_mm) != next);
+
 		if (!cpu_test_and_set(cpu, next->cpu_vm_mask)) {
 			/* We were in lazy tlb mode and leave_mm disabled
 			 * tlb flush IPI delivery. We must reload CR3
diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 69a40757e217..8ee835ed10e1 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -25,9 +25,7 @@ struct x8664_pda {
 	char *irqstackptr;
 	short nodenumber;		/* number of current node (32k max) */
 	short in_bootmem;		/* pda lives in bootmem */
-	short mmu_state;
 	short isidle;
-	struct mm_struct *active_mm;
 } ____cacheline_aligned_in_smp;
 
 DECLARE_PER_CPU(struct x8664_pda, __pda);
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index 17feaa9c7e76..d3539f998f88 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -148,20 +148,17 @@ void native_flush_tlb_others(const struct cpumask *cpumask,
 #define TLBSTATE_OK	1
 #define TLBSTATE_LAZY	2
 
-#ifdef CONFIG_X86_32
 struct tlb_state {
 	struct mm_struct *active_mm;
 	int state;
-	char __cacheline_padding[L1_CACHE_BYTES-8];
 };
 DECLARE_PER_CPU(struct tlb_state, cpu_tlbstate);
 
-void reset_lazy_tlbstate(void);
-#else
 static inline void reset_lazy_tlbstate(void)
 {
+	percpu_write(cpu_tlbstate.state, 0);
+	percpu_write(cpu_tlbstate.active_mm, &init_mm);
 }
-#endif
 
 #endif	/* SMP */
 
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index c49498d40830..3d0cc6f17116 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -897,8 +897,6 @@ void __cpuinit pda_init(int cpu)
 	pda->irqcount = -1;
 	pda->kernelstack = (unsigned long)stack_thread_info() -
 				 PDA_STACKOFFSET + THREAD_SIZE;
-	pda->active_mm = &init_mm;
-	pda->mmu_state = 0;
 
 	if (cpu == 0) {
 		/* others are initialized in smpboot.c */
diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c
index e65449d0f7d9..abf0808d6fc4 100644
--- a/arch/x86/kernel/tlb_32.c
+++ b/arch/x86/kernel/tlb_32.c
@@ -4,8 +4,8 @@
 
 #include <asm/tlbflush.h>
 
-DEFINE_PER_CPU(struct tlb_state, cpu_tlbstate)
-			____cacheline_aligned = { &init_mm, 0, };
+DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate)
+			= { &init_mm, 0, };
 
 /* must come after the send_IPI functions above for inlining */
 #include <mach_ipi.h>
@@ -231,14 +231,6 @@ void flush_tlb_all(void)
 	on_each_cpu(do_flush_tlb_all, NULL, 1);
 }
 
-void reset_lazy_tlbstate(void)
-{
-	int cpu = raw_smp_processor_id();
-
-	per_cpu(cpu_tlbstate, cpu).state = 0;
-	per_cpu(cpu_tlbstate, cpu).active_mm = &init_mm;
-}
-
 static int init_flush_cpumask(void)
 {
 	alloc_cpumask_var(&flush_cpumask, GFP_KERNEL);
diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c
index 7f4141d3b661..e64a32c48825 100644
--- a/arch/x86/kernel/tlb_64.c
+++ b/arch/x86/kernel/tlb_64.c
@@ -18,6 +18,9 @@
 #include <asm/uv/uv_hub.h>
 #include <asm/uv/uv_bau.h>
 
+DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate)
+			= { &init_mm, 0, };
+
 #include <mach_ipi.h>
 /*
  *	Smarter SMP flushing macros.
@@ -62,9 +65,9 @@ static DEFINE_PER_CPU(union smp_flush_state, flush_state);
  */
 void leave_mm(int cpu)
 {
-	if (read_pda(mmu_state) == TLBSTATE_OK)
+	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK)
 		BUG();
-	cpu_clear(cpu, read_pda(active_mm)->cpu_vm_mask);
+	cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask);
 	load_cr3(swapper_pg_dir);
 }
 EXPORT_SYMBOL_GPL(leave_mm);
@@ -142,8 +145,8 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs)
 		 * BUG();
 		 */
 
-	if (f->flush_mm == read_pda(active_mm)) {
-		if (read_pda(mmu_state) == TLBSTATE_OK) {
+	if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) {
+		if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) {
 			if (f->flush_va == TLB_FLUSH_ALL)
 				local_flush_tlb();
 			else
@@ -281,7 +284,7 @@ static void do_flush_tlb_all(void *info)
 	unsigned long cpu = smp_processor_id();
 
 	__flush_tlb_all();
-	if (read_pda(mmu_state) == TLBSTATE_LAZY)
+	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY)
 		leave_mm(cpu);
 }
 
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
index 7bc7852cc5c4..98cb9869eb24 100644
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -1063,11 +1063,7 @@ static void drop_other_mm_ref(void *info)
 	struct mm_struct *mm = info;
 	struct mm_struct *active_mm;
 
-#ifdef CONFIG_X86_64
-	active_mm = read_pda(active_mm);
-#else
-	active_mm = __get_cpu_var(cpu_tlbstate).active_mm;
-#endif
+	active_mm = percpu_read(cpu_tlbstate.active_mm);
 
 	if (active_mm == mm)
 		leave_mm(smp_processor_id());

From 26f80bd6a9ab17bc8a60b6092e7c0d05c5927ce5 Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:58 +0900
Subject: [PATCH 03/12] x86-64: Convert irqstacks to per-cpu

Move the irqstackptr variable from the PDA to per-cpu.  Make the
stacks themselves per-cpu, removing some specific allocation code.
Add a seperate flag (is_boot_cpu) to simplify the per-cpu boot
adjustments.

tj: * sprinkle some underbars around.

    * irq_stack_ptr is not used till traps_init(), no reason to
      initialize it early.  On SMP, just leaving it NULL till proper
      initialization in setup_per_cpu_areas() works.  Dropped
      is_boot_cpu and early irq_stack_ptr initialization.

    * do DECLARE/DEFINE_PER_CPU(char[IRQ_STACK_SIZE], irq_stack)
      instead of (char, irq_stack[IRQ_STACK_SIZE]).

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/page_64.h   |  4 ++--
 arch/x86/include/asm/pda.h       |  1 -
 arch/x86/include/asm/processor.h |  3 +++
 arch/x86/kernel/asm-offsets_64.c |  1 -
 arch/x86/kernel/cpu/common.c     | 19 +++++++-----------
 arch/x86/kernel/dumpstack_64.c   | 33 ++++++++++++++++----------------
 arch/x86/kernel/entry_64.S       |  6 +++---
 arch/x86/kernel/setup_percpu.c   |  4 +++-
 8 files changed, 35 insertions(+), 36 deletions(-)

diff --git a/arch/x86/include/asm/page_64.h b/arch/x86/include/asm/page_64.h
index 5ebca29f44f0..e27fdbe5f9e4 100644
--- a/arch/x86/include/asm/page_64.h
+++ b/arch/x86/include/asm/page_64.h
@@ -13,8 +13,8 @@
 #define DEBUG_STACK_ORDER (EXCEPTION_STACK_ORDER + 1)
 #define DEBUG_STKSZ (PAGE_SIZE << DEBUG_STACK_ORDER)
 
-#define IRQSTACK_ORDER 2
-#define IRQSTACKSIZE (PAGE_SIZE << IRQSTACK_ORDER)
+#define IRQ_STACK_ORDER 2
+#define IRQ_STACK_SIZE (PAGE_SIZE << IRQ_STACK_ORDER)
 
 #define STACKFAULT_STACK 1
 #define DOUBLEFAULT_STACK 2
diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 8ee835ed10e1..09965f7a2165 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -22,7 +22,6 @@ struct x8664_pda {
 					/* gcc-ABI: this canary MUST be at
 					   offset 40!!! */
 #endif
-	char *irqstackptr;
 	short nodenumber;		/* number of current node (32k max) */
 	short in_bootmem;		/* pda lives in bootmem */
 	short isidle;
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 091cd8855f2e..f511246fa6cd 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -378,6 +378,9 @@ union thread_xstate {
 
 #ifdef CONFIG_X86_64
 DECLARE_PER_CPU(struct orig_ist, orig_ist);
+
+DECLARE_PER_CPU(char[IRQ_STACK_SIZE], irq_stack);
+DECLARE_PER_CPU(char *, irq_stack_ptr);
 #endif
 
 extern void print_cpu_info(struct cpuinfo_x86 *);
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index f4cc81bfbf89..5b821fbdaf7b 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -54,7 +54,6 @@ int main(void)
 	ENTRY(pcurrent); 
 	ENTRY(irqcount);
 	ENTRY(cpunumber);
-	ENTRY(irqstackptr);
 	DEFINE(pda_size, sizeof(struct x8664_pda));
 	BLANK();
 #undef ENTRY
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 3d0cc6f17116..496f0a01919b 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -881,7 +881,13 @@ __setup("clearcpuid=", setup_disablecpuid);
 #ifdef CONFIG_X86_64
 struct desc_ptr idt_descr = { 256 * 16 - 1, (unsigned long) idt_table };
 
-static char boot_cpu_stack[IRQSTACKSIZE] __page_aligned_bss;
+DEFINE_PER_CPU_PAGE_ALIGNED(char[IRQ_STACK_SIZE], irq_stack);
+#ifdef CONFIG_SMP
+DEFINE_PER_CPU(char *, irq_stack_ptr);	/* will be set during per cpu init */
+#else
+DEFINE_PER_CPU(char *, irq_stack_ptr) =
+	per_cpu_var(irq_stack) + IRQ_STACK_SIZE - 64;
+#endif
 
 void __cpuinit pda_init(int cpu)
 {
@@ -901,18 +907,7 @@ void __cpuinit pda_init(int cpu)
 	if (cpu == 0) {
 		/* others are initialized in smpboot.c */
 		pda->pcurrent = &init_task;
-		pda->irqstackptr = boot_cpu_stack;
-		pda->irqstackptr += IRQSTACKSIZE - 64;
 	} else {
-		if (!pda->irqstackptr) {
-			pda->irqstackptr = (char *)
-				__get_free_pages(GFP_ATOMIC, IRQSTACK_ORDER);
-			if (!pda->irqstackptr)
-				panic("cannot allocate irqstack for cpu %d",
-				      cpu);
-			pda->irqstackptr += IRQSTACKSIZE - 64;
-		}
-
 		if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE)
 			pda->nodenumber = cpu_to_node(cpu);
 	}
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c
index c302d0707048..28e26a4315df 100644
--- a/arch/x86/kernel/dumpstack_64.c
+++ b/arch/x86/kernel/dumpstack_64.c
@@ -106,7 +106,8 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
 		const struct stacktrace_ops *ops, void *data)
 {
 	const unsigned cpu = get_cpu();
-	unsigned long *irqstack_end = (unsigned long *)cpu_pda(cpu)->irqstackptr;
+	unsigned long *irq_stack_end =
+		(unsigned long *)per_cpu(irq_stack_ptr, cpu);
 	unsigned used = 0;
 	struct thread_info *tinfo;
 	int graph = 0;
@@ -160,23 +161,23 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
 			stack = (unsigned long *) estack_end[-2];
 			continue;
 		}
-		if (irqstack_end) {
-			unsigned long *irqstack;
-			irqstack = irqstack_end -
-				(IRQSTACKSIZE - 64) / sizeof(*irqstack);
+		if (irq_stack_end) {
+			unsigned long *irq_stack;
+			irq_stack = irq_stack_end -
+				(IRQ_STACK_SIZE - 64) / sizeof(*irq_stack);
 
-			if (stack >= irqstack && stack < irqstack_end) {
+			if (stack >= irq_stack && stack < irq_stack_end) {
 				if (ops->stack(data, "IRQ") < 0)
 					break;
 				bp = print_context_stack(tinfo, stack, bp,
-					ops, data, irqstack_end, &graph);
+					ops, data, irq_stack_end, &graph);
 				/*
 				 * We link to the next stack (which would be
 				 * the process stack normally) the last
 				 * pointer (index -1 to end) in the IRQ stack:
 				 */
-				stack = (unsigned long *) (irqstack_end[-1]);
-				irqstack_end = NULL;
+				stack = (unsigned long *) (irq_stack_end[-1]);
+				irq_stack_end = NULL;
 				ops->stack(data, "EOI");
 				continue;
 			}
@@ -199,10 +200,10 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
 	unsigned long *stack;
 	int i;
 	const int cpu = smp_processor_id();
-	unsigned long *irqstack_end =
-		(unsigned long *) (cpu_pda(cpu)->irqstackptr);
-	unsigned long *irqstack =
-		(unsigned long *) (cpu_pda(cpu)->irqstackptr - IRQSTACKSIZE);
+	unsigned long *irq_stack_end =
+		(unsigned long *)(per_cpu(irq_stack_ptr, cpu));
+	unsigned long *irq_stack =
+		(unsigned long *)(per_cpu(irq_stack_ptr, cpu) - IRQ_STACK_SIZE);
 
 	/*
 	 * debugging aid: "show_stack(NULL, NULL);" prints the
@@ -218,9 +219,9 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
 
 	stack = sp;
 	for (i = 0; i < kstack_depth_to_print; i++) {
-		if (stack >= irqstack && stack <= irqstack_end) {
-			if (stack == irqstack_end) {
-				stack = (unsigned long *) (irqstack_end[-1]);
+		if (stack >= irq_stack && stack <= irq_stack_end) {
+			if (stack == irq_stack_end) {
+				stack = (unsigned long *) (irq_stack_end[-1]);
 				printk(" <EOI> ");
 			}
 		} else {
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index 4833f3a19650..d22677a66438 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -345,7 +345,7 @@ ENTRY(save_args)
 1:	incl %gs:pda_irqcount
 	jne 2f
 	popq_cfi %rax			/* move return address... */
-	mov %gs:pda_irqstackptr,%rsp
+	mov PER_CPU_VAR(irq_stack_ptr),%rsp
 	EMPTY_FRAME 0
 	pushq_cfi %rax			/* ... to the new stack */
 	/*
@@ -1261,7 +1261,7 @@ ENTRY(call_softirq)
 	mov  %rsp,%rbp
 	CFI_DEF_CFA_REGISTER rbp
 	incl %gs:pda_irqcount
-	cmove %gs:pda_irqstackptr,%rsp
+	cmove PER_CPU_VAR(irq_stack_ptr),%rsp
 	push  %rbp			# backlink for old unwinder
 	call __do_softirq
 	leaveq
@@ -1300,7 +1300,7 @@ ENTRY(xen_do_hypervisor_callback)   # do_hypervisor_callback(struct *pt_regs)
 11:	incl %gs:pda_irqcount
 	movq %rsp,%rbp
 	CFI_DEF_CFA_REGISTER rbp
-	cmovzq %gs:pda_irqstackptr,%rsp
+	cmovzq PER_CPU_VAR(irq_stack_ptr),%rsp
 	pushq %rbp			# backlink for old unwinder
 	call xen_evtchn_do_upcall
 	popq %rsp
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index b5c35af2011d..8b53ef83c611 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -192,7 +192,10 @@ void __init setup_per_cpu_areas(void)
 
 		memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start);
 		per_cpu_offset(cpu) = ptr - __per_cpu_start;
+		per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu);
 #ifdef CONFIG_X86_64
+		per_cpu(irq_stack_ptr, cpu) =
+			(char *)per_cpu(irq_stack, cpu) + IRQ_STACK_SIZE - 64;
 		/*
 		 * CPU0 modified pda in the init data area, reload pda
 		 * offset for CPU0 and clear the area for others.
@@ -202,7 +205,6 @@ void __init setup_per_cpu_areas(void)
 		else
 			memset(cpu_pda(cpu), 0, sizeof(*cpu_pda(cpu)));
 #endif
-		per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu);
 
 		DBG("PERCPU: cpu %4d %p\n", cpu, ptr);
 	}

From 92d65b2371d86d40807e1dbfdccadc4d501edcde Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:58 +0900
Subject: [PATCH 04/12] x86-64: Convert exception stacks to per-cpu

Move the exception stacks to per-cpu, removing specific allocation code.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/kernel/cpu/common.c | 23 ++++++++---------------
 1 file changed, 8 insertions(+), 15 deletions(-)

diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 496f0a01919b..b6d7eec0be77 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -913,8 +913,9 @@ void __cpuinit pda_init(int cpu)
 	}
 }
 
-static char boot_exception_stacks[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ +
-				  DEBUG_STKSZ] __page_aligned_bss;
+static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks
+	[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + DEBUG_STKSZ])
+	__aligned(PAGE_SIZE);
 
 extern asmlinkage void ignore_sysret(void);
 
@@ -972,15 +973,12 @@ void __cpuinit cpu_init(void)
 	struct tss_struct *t = &per_cpu(init_tss, cpu);
 	struct orig_ist *orig_ist = &per_cpu(orig_ist, cpu);
 	unsigned long v;
-	char *estacks = NULL;
 	struct task_struct *me;
 	int i;
 
 	/* CPU 0 is initialised in head64.c */
 	if (cpu != 0)
 		pda_init(cpu);
-	else
-		estacks = boot_exception_stacks;
 
 	me = current;
 
@@ -1014,18 +1012,13 @@ void __cpuinit cpu_init(void)
 	 * set up and load the per-CPU TSS
 	 */
 	if (!orig_ist->ist[0]) {
-		static const unsigned int order[N_EXCEPTION_STACKS] = {
-		  [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER,
-		  [DEBUG_STACK - 1] = DEBUG_STACK_ORDER
+		static const unsigned int sizes[N_EXCEPTION_STACKS] = {
+		  [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STKSZ,
+		  [DEBUG_STACK - 1] = DEBUG_STKSZ
 		};
+		char *estacks = per_cpu(exception_stacks, cpu);
 		for (v = 0; v < N_EXCEPTION_STACKS; v++) {
-			if (cpu) {
-				estacks = (char *)__get_free_pages(GFP_ATOMIC, order[v]);
-				if (!estacks)
-					panic("Cannot allocate exception "
-					      "stack %ld %d\n", v, cpu);
-			}
-			estacks += PAGE_SIZE << order[v];
+			estacks += sizes[v];
 			orig_ist->ist[v] = t->x86_tss.ist[v] =
 					(unsigned long)estacks;
 		}

From ea9279066de44053d0c20ea855bc9f4706652d84 Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:58 +0900
Subject: [PATCH 05/12] x86-64: Move cpu number from PDA to per-cpu and
 consolidate with 32-bit.

tj: moved cpu_number definition out of CONFIG_HAVE_SETUP_PER_CPU_AREA
    for voyager.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/pda.h       |  2 +-
 arch/x86/include/asm/smp.h       |  4 +---
 arch/x86/kernel/asm-offsets_64.c |  1 -
 arch/x86/kernel/cpu/common.c     |  1 -
 arch/x86/kernel/process_32.c     |  3 ---
 arch/x86/kernel/setup_percpu.c   | 10 ++++++++++
 arch/x86/kernel/smpcommon.c      |  2 --
 7 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 09965f7a2165..668d5a5b6f70 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -16,7 +16,7 @@ struct x8664_pda {
 	unsigned long kernelstack;	/* 16 top of kernel stack for current */
 	unsigned long oldrsp;		/* 24 user rsp for system call */
 	int irqcount;			/* 32 Irq nesting counter. Starts -1 */
-	unsigned int cpunumber;		/* 36 Logical CPU number */
+	unsigned int unused6;		/* 36 was cpunumber */
 #ifdef CONFIG_CC_STACKPROTECTOR
 	unsigned long stack_canary;	/* 40 stack canary value */
 					/* gcc-ABI: this canary MUST be at
diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
index c7bbbbe65d3f..68636e767a91 100644
--- a/arch/x86/include/asm/smp.h
+++ b/arch/x86/include/asm/smp.h
@@ -25,9 +25,7 @@ extern unsigned int num_processors;
 DECLARE_PER_CPU(cpumask_t, cpu_sibling_map);
 DECLARE_PER_CPU(cpumask_t, cpu_core_map);
 DECLARE_PER_CPU(u16, cpu_llc_id);
-#ifdef CONFIG_X86_32
 DECLARE_PER_CPU(int, cpu_number);
-#endif
 
 static inline struct cpumask *cpu_sibling_mask(int cpu)
 {
@@ -164,7 +162,7 @@ extern unsigned disabled_cpus __cpuinitdata;
 extern int safe_smp_processor_id(void);
 
 #elif defined(CONFIG_X86_64_SMP)
-#define raw_smp_processor_id()	read_pda(cpunumber)
+#define raw_smp_processor_id() (percpu_read(cpu_number))
 
 #define stack_smp_processor_id()					\
 ({								\
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index 5b821fbdaf7b..cae6697c0991 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -53,7 +53,6 @@ int main(void)
 	ENTRY(oldrsp); 
 	ENTRY(pcurrent); 
 	ENTRY(irqcount);
-	ENTRY(cpunumber);
 	DEFINE(pda_size, sizeof(struct x8664_pda));
 	BLANK();
 #undef ENTRY
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index b6d7eec0be77..4221e920886d 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -899,7 +899,6 @@ void __cpuinit pda_init(int cpu)
 
 	load_pda_offset(cpu);
 
-	pda->cpunumber = cpu;
 	pda->irqcount = -1;
 	pda->kernelstack = (unsigned long)stack_thread_info() -
 				 PDA_STACKOFFSET + THREAD_SIZE;
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 77d546817d94..2c00a57ccb90 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -66,9 +66,6 @@ asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
 DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
 EXPORT_PER_CPU_SYMBOL(current_task);
 
-DEFINE_PER_CPU(int, cpu_number);
-EXPORT_PER_CPU_SYMBOL(cpu_number);
-
 /*
  * Return saved PC of a blocked thread.
  */
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index 8b53ef83c611..258497f93f4d 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -22,6 +22,15 @@
 # define DBG(x...)
 #endif
 
+/*
+ * Could be inside CONFIG_HAVE_SETUP_PER_CPU_AREA with other stuff but
+ * voyager wants cpu_number too.
+ */
+#ifdef CONFIG_SMP
+DEFINE_PER_CPU(int, cpu_number);
+EXPORT_PER_CPU_SYMBOL(cpu_number);
+#endif
+
 #ifdef CONFIG_X86_LOCAL_APIC
 unsigned int num_processors;
 unsigned disabled_cpus __cpuinitdata;
@@ -193,6 +202,7 @@ void __init setup_per_cpu_areas(void)
 		memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start);
 		per_cpu_offset(cpu) = ptr - __per_cpu_start;
 		per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu);
+		per_cpu(cpu_number, cpu) = cpu;
 #ifdef CONFIG_X86_64
 		per_cpu(irq_stack_ptr, cpu) =
 			(char *)per_cpu(irq_stack, cpu) + IRQ_STACK_SIZE - 64;
diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c
index 7e157810062f..add36b4e37c9 100644
--- a/arch/x86/kernel/smpcommon.c
+++ b/arch/x86/kernel/smpcommon.c
@@ -28,7 +28,5 @@ __cpuinit void init_gdt(int cpu)
 
 	write_gdt_entry(get_cpu_gdt_table(cpu),
 			GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S);
-
-	per_cpu(cpu_number, cpu) = cpu;
 }
 #endif

From c6f5e0acd5d12ee23f701f15889872e67b47caa6 Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:58 +0900
Subject: [PATCH 06/12] x86-64: Move current task from PDA to per-cpu and
 consolidate with 32-bit.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/current.h   | 24 +++---------------------
 arch/x86/include/asm/pda.h       |  4 ++--
 arch/x86/include/asm/system.h    |  4 ++--
 arch/x86/kernel/asm-offsets_64.c |  1 -
 arch/x86/kernel/cpu/common.c     |  5 +----
 arch/x86/kernel/dumpstack_64.c   |  2 +-
 arch/x86/kernel/process_64.c     |  5 ++++-
 arch/x86/kernel/smpboot.c        |  3 +--
 arch/x86/xen/smp.c               |  3 +--
 9 files changed, 15 insertions(+), 36 deletions(-)

diff --git a/arch/x86/include/asm/current.h b/arch/x86/include/asm/current.h
index 0728480f5c56..c68c361697e1 100644
--- a/arch/x86/include/asm/current.h
+++ b/arch/x86/include/asm/current.h
@@ -1,39 +1,21 @@
 #ifndef _ASM_X86_CURRENT_H
 #define _ASM_X86_CURRENT_H
 
-#ifdef CONFIG_X86_32
 #include <linux/compiler.h>
 #include <asm/percpu.h>
 
+#ifndef __ASSEMBLY__
 struct task_struct;
 
 DECLARE_PER_CPU(struct task_struct *, current_task);
+
 static __always_inline struct task_struct *get_current(void)
 {
 	return percpu_read(current_task);
 }
 
-#else /* X86_32 */
-
-#ifndef __ASSEMBLY__
-#include <asm/pda.h>
-
-struct task_struct;
-
-static __always_inline struct task_struct *get_current(void)
-{
-	return read_pda(pcurrent);
-}
-
-#else /* __ASSEMBLY__ */
-
-#include <asm/asm-offsets.h>
-#define GET_CURRENT(reg) movq %gs:(pda_pcurrent),reg
+#define current get_current()
 
 #endif /* __ASSEMBLY__ */
 
-#endif /* X86_32 */
-
-#define current get_current()
-
 #endif /* _ASM_X86_CURRENT_H */
diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 668d5a5b6f70..7209302d9227 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -11,8 +11,8 @@
 
 /* Per processor datastructure. %gs points to it while the kernel runs */
 struct x8664_pda {
-	struct task_struct *pcurrent;	/* 0  Current process */
-	unsigned long dummy;
+	unsigned long unused1;
+	unsigned long unused2;
 	unsigned long kernelstack;	/* 16 top of kernel stack for current */
 	unsigned long oldrsp;		/* 24 user rsp for system call */
 	int irqcount;			/* 32 Irq nesting counter. Starts -1 */
diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h
index 8e626ea33a1a..4399aac680e9 100644
--- a/arch/x86/include/asm/system.h
+++ b/arch/x86/include/asm/system.h
@@ -94,7 +94,7 @@ do {									\
 	     "call __switch_to\n\t"					  \
 	     ".globl thread_return\n"					  \
 	     "thread_return:\n\t"					  \
-	     "movq %%gs:%P[pda_pcurrent],%%rsi\n\t"			  \
+	     "movq "__percpu_seg_str"%P[current_task],%%rsi\n\t"	  \
 	     "movq %P[thread_info](%%rsi),%%r8\n\t"			  \
 	     LOCK_PREFIX "btr  %[tif_fork],%P[ti_flags](%%r8)\n\t"	  \
 	     "movq %%rax,%%rdi\n\t" 					  \
@@ -106,7 +106,7 @@ do {									\
 	       [ti_flags] "i" (offsetof(struct thread_info, flags)),	  \
 	       [tif_fork] "i" (TIF_FORK),			  	  \
 	       [thread_info] "i" (offsetof(struct task_struct, stack)),   \
-	       [pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent))  \
+	       [current_task] "m" (per_cpu_var(current_task))		  \
 	     : "memory", "cc" __EXTRA_CLOBBER)
 #endif
 
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index cae6697c0991..4f7a210e1e58 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -51,7 +51,6 @@ int main(void)
 #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry))
 	ENTRY(kernelstack); 
 	ENTRY(oldrsp); 
-	ENTRY(pcurrent); 
 	ENTRY(irqcount);
 	DEFINE(pda_size, sizeof(struct x8664_pda));
 	BLANK();
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 4221e920886d..b50e38d16888 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -903,10 +903,7 @@ void __cpuinit pda_init(int cpu)
 	pda->kernelstack = (unsigned long)stack_thread_info() -
 				 PDA_STACKOFFSET + THREAD_SIZE;
 
-	if (cpu == 0) {
-		/* others are initialized in smpboot.c */
-		pda->pcurrent = &init_task;
-	} else {
+	if (cpu != 0) {
 		if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE)
 			pda->nodenumber = cpu_to_node(cpu);
 	}
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c
index 28e26a4315df..d35db5993fd6 100644
--- a/arch/x86/kernel/dumpstack_64.c
+++ b/arch/x86/kernel/dumpstack_64.c
@@ -242,7 +242,7 @@ void show_registers(struct pt_regs *regs)
 	int i;
 	unsigned long sp;
 	const int cpu = smp_processor_id();
-	struct task_struct *cur = cpu_pda(cpu)->pcurrent;
+	struct task_struct *cur = current;
 
 	sp = regs->sp;
 	printk("CPU %d ", cpu);
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 416fb9282f4f..e00c31a4b3c0 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -57,6 +57,9 @@
 
 asmlinkage extern void ret_from_fork(void);
 
+DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
+EXPORT_PER_CPU_SYMBOL(current_task);
+
 unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED;
 
 static ATOMIC_NOTIFIER_HEAD(idle_notifier);
@@ -615,7 +618,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 	 */
 	prev->usersp = read_pda(oldrsp);
 	write_pda(oldrsp, next->usersp);
-	write_pda(pcurrent, next_p);
+	percpu_write(current_task, next_p);
 
 	write_pda(kernelstack,
 		  (unsigned long)task_stack_page(next_p) +
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 2f0e0f1090f6..5854be0fb804 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -790,13 +790,12 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu)
 
 	set_idle_for_cpu(cpu, c_idle.idle);
 do_rest:
-#ifdef CONFIG_X86_32
 	per_cpu(current_task, cpu) = c_idle.idle;
+#ifdef CONFIG_X86_32
 	init_gdt(cpu);
 	/* Stack for startup_32 can be just as for start_secondary onwards */
 	irq_ctx_init(cpu);
 #else
-	cpu_pda(cpu)->pcurrent = c_idle.idle;
 	clear_tsk_thread_flag(c_idle.idle, TIF_FORK);
 	initial_gs = per_cpu_offset(cpu);
 #endif
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index 9ff3b0999cfb..72c2eb9b64cd 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -279,12 +279,11 @@ static int __cpuinit xen_cpu_up(unsigned int cpu)
 	struct task_struct *idle = idle_task(cpu);
 	int rc;
 
+	per_cpu(current_task, cpu) = idle;
 #ifdef CONFIG_X86_32
 	init_gdt(cpu);
-	per_cpu(current_task, cpu) = idle;
 	irq_ctx_init(cpu);
 #else
-	cpu_pda(cpu)->pcurrent = idle;
 	clear_tsk_thread_flag(idle, TIF_FORK);
 #endif
 	xen_setup_timer(cpu);

From 9af45651f1f7c89942e016a1a00a7ebddfa727f8 Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:58 +0900
Subject: [PATCH 07/12] x86-64: Move kernelstack from PDA to per-cpu.

Also clean up PER_CPU_VAR usage in xen-asm_64.S

tj: * remove now unused stack_thread_info()
    * s/kernelstack/kernel_stack/
    * added FIXME comment in xen-asm_64.S

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/ia32/ia32entry.S          |  8 ++++----
 arch/x86/include/asm/pda.h         |  4 +---
 arch/x86/include/asm/thread_info.h | 20 ++++++++------------
 arch/x86/kernel/asm-offsets_64.c   |  1 -
 arch/x86/kernel/cpu/common.c       |  6 ++++--
 arch/x86/kernel/entry_64.S         |  4 ++--
 arch/x86/kernel/process_64.c       |  4 ++--
 arch/x86/kernel/smpboot.c          |  3 +++
 arch/x86/xen/xen-asm_64.S          | 23 +++++++++++------------
 9 files changed, 35 insertions(+), 38 deletions(-)

diff --git a/arch/x86/ia32/ia32entry.S b/arch/x86/ia32/ia32entry.S
index 256b00b61892..9c79b2477008 100644
--- a/arch/x86/ia32/ia32entry.S
+++ b/arch/x86/ia32/ia32entry.S
@@ -112,8 +112,8 @@ ENTRY(ia32_sysenter_target)
 	CFI_DEF_CFA	rsp,0
 	CFI_REGISTER	rsp,rbp
 	SWAPGS_UNSAFE_STACK
-	movq	%gs:pda_kernelstack, %rsp
-	addq	$(PDA_STACKOFFSET),%rsp	
+	movq	PER_CPU_VAR(kernel_stack), %rsp
+	addq	$(KERNEL_STACK_OFFSET),%rsp
 	/*
 	 * No need to follow this irqs on/off section: the syscall
 	 * disabled irqs, here we enable it straight after entry:
@@ -273,13 +273,13 @@ ENDPROC(ia32_sysenter_target)
 ENTRY(ia32_cstar_target)
 	CFI_STARTPROC32	simple
 	CFI_SIGNAL_FRAME
-	CFI_DEF_CFA	rsp,PDA_STACKOFFSET
+	CFI_DEF_CFA	rsp,KERNEL_STACK_OFFSET
 	CFI_REGISTER	rip,rcx
 	/*CFI_REGISTER	rflags,r11*/
 	SWAPGS_UNSAFE_STACK
 	movl	%esp,%r8d
 	CFI_REGISTER	rsp,r8
-	movq	%gs:pda_kernelstack,%rsp
+	movq	PER_CPU_VAR(kernel_stack),%rsp
 	/*
 	 * No need to follow this irqs on/off section: the syscall
 	 * disabled irqs and here we enable it straight after entry:
diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 7209302d9227..4d28ffba6e1b 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -13,7 +13,7 @@
 struct x8664_pda {
 	unsigned long unused1;
 	unsigned long unused2;
-	unsigned long kernelstack;	/* 16 top of kernel stack for current */
+	unsigned long unused3;
 	unsigned long oldrsp;		/* 24 user rsp for system call */
 	int irqcount;			/* 32 Irq nesting counter. Starts -1 */
 	unsigned int unused6;		/* 36 was cpunumber */
@@ -44,6 +44,4 @@ extern void pda_init(int);
 
 #endif
 
-#define PDA_STACKOFFSET (5*8)
-
 #endif /* _ASM_X86_PDA_H */
diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h
index 98789647baa9..b46f8ca007b5 100644
--- a/arch/x86/include/asm/thread_info.h
+++ b/arch/x86/include/asm/thread_info.h
@@ -194,25 +194,21 @@ static inline struct thread_info *current_thread_info(void)
 
 #else /* X86_32 */
 
-#include <asm/pda.h>
+#include <asm/percpu.h>
+#define KERNEL_STACK_OFFSET (5*8)
 
 /*
  * macros/functions for gaining access to the thread information structure
  * preempt_count needs to be 1 initially, until the scheduler is functional.
  */
 #ifndef __ASSEMBLY__
+DECLARE_PER_CPU(unsigned long, kernel_stack);
+
 static inline struct thread_info *current_thread_info(void)
 {
 	struct thread_info *ti;
-	ti = (void *)(read_pda(kernelstack) + PDA_STACKOFFSET - THREAD_SIZE);
-	return ti;
-}
-
-/* do not use in interrupt context */
-static inline struct thread_info *stack_thread_info(void)
-{
-	struct thread_info *ti;
-	asm("andq %%rsp,%0; " : "=r" (ti) : "0" (~(THREAD_SIZE - 1)));
+	ti = (void *)(percpu_read(kernel_stack) +
+		      KERNEL_STACK_OFFSET - THREAD_SIZE);
 	return ti;
 }
 
@@ -220,8 +216,8 @@ static inline struct thread_info *stack_thread_info(void)
 
 /* how to get the thread information struct from ASM */
 #define GET_THREAD_INFO(reg) \
-	movq %gs:pda_kernelstack,reg ; \
-	subq $(THREAD_SIZE-PDA_STACKOFFSET),reg
+	movq PER_CPU_VAR(kernel_stack),reg ; \
+	subq $(THREAD_SIZE-KERNEL_STACK_OFFSET),reg
 
 #endif
 
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index 4f7a210e1e58..cafff5f4a031 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -49,7 +49,6 @@ int main(void)
 	BLANK();
 #undef ENTRY
 #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry))
-	ENTRY(kernelstack); 
 	ENTRY(oldrsp); 
 	ENTRY(irqcount);
 	DEFINE(pda_size, sizeof(struct x8664_pda));
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index b50e38d16888..06b6290088f4 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -889,6 +889,10 @@ DEFINE_PER_CPU(char *, irq_stack_ptr) =
 	per_cpu_var(irq_stack) + IRQ_STACK_SIZE - 64;
 #endif
 
+DEFINE_PER_CPU(unsigned long, kernel_stack) =
+	(unsigned long)&init_thread_union - KERNEL_STACK_OFFSET + THREAD_SIZE;
+EXPORT_PER_CPU_SYMBOL(kernel_stack);
+
 void __cpuinit pda_init(int cpu)
 {
 	struct x8664_pda *pda = cpu_pda(cpu);
@@ -900,8 +904,6 @@ void __cpuinit pda_init(int cpu)
 	load_pda_offset(cpu);
 
 	pda->irqcount = -1;
-	pda->kernelstack = (unsigned long)stack_thread_info() -
-				 PDA_STACKOFFSET + THREAD_SIZE;
 
 	if (cpu != 0) {
 		if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE)
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index d22677a66438..0dd45859a7a8 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -468,7 +468,7 @@ END(ret_from_fork)
 ENTRY(system_call)
 	CFI_STARTPROC	simple
 	CFI_SIGNAL_FRAME
-	CFI_DEF_CFA	rsp,PDA_STACKOFFSET
+	CFI_DEF_CFA	rsp,KERNEL_STACK_OFFSET
 	CFI_REGISTER	rip,rcx
 	/*CFI_REGISTER	rflags,r11*/
 	SWAPGS_UNSAFE_STACK
@@ -480,7 +480,7 @@ ENTRY(system_call)
 ENTRY(system_call_after_swapgs)
 
 	movq	%rsp,%gs:pda_oldrsp
-	movq	%gs:pda_kernelstack,%rsp
+	movq	PER_CPU_VAR(kernel_stack),%rsp
 	/*
 	 * No need to follow this irqs off/on section - it's straight
 	 * and short:
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index e00c31a4b3c0..6c5f57602108 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -620,9 +620,9 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 	write_pda(oldrsp, next->usersp);
 	percpu_write(current_task, next_p);
 
-	write_pda(kernelstack,
+	percpu_write(kernel_stack,
 		  (unsigned long)task_stack_page(next_p) +
-		  THREAD_SIZE - PDA_STACKOFFSET);
+		  THREAD_SIZE - KERNEL_STACK_OFFSET);
 #ifdef CONFIG_CC_STACKPROTECTOR
 	write_pda(stack_canary, next_p->stack_canary);
 	/*
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 5854be0fb804..869b98840fd0 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -798,6 +798,9 @@ do_rest:
 #else
 	clear_tsk_thread_flag(c_idle.idle, TIF_FORK);
 	initial_gs = per_cpu_offset(cpu);
+	per_cpu(kernel_stack, cpu) =
+		(unsigned long)task_stack_page(c_idle.idle) -
+		KERNEL_STACK_OFFSET + THREAD_SIZE;
 #endif
 	early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu);
 	initial_code = (unsigned long)start_secondary;
diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S
index 05794c566e87..5a23e8993678 100644
--- a/arch/x86/xen/xen-asm_64.S
+++ b/arch/x86/xen/xen-asm_64.S
@@ -17,6 +17,7 @@
 #include <asm/processor-flags.h>
 #include <asm/errno.h>
 #include <asm/segment.h>
+#include <asm/percpu.h>
 
 #include <xen/interface/xen.h>
 
@@ -28,12 +29,10 @@
 
 #if 1
 /*
-	x86-64 does not yet support direct access to percpu variables
-	via a segment override, so we just need to make sure this code
-	never gets used
+	FIXME: x86_64 now can support direct access to percpu variables
+	via a segment override.  Update xen accordingly.
  */
 #define BUG			ud2a
-#define PER_CPU_VAR(var, off)	0xdeadbeef
 #endif
 
 /*
@@ -45,14 +44,14 @@ ENTRY(xen_irq_enable_direct)
 	BUG
 
 	/* Unmask events */
-	movb $0, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask)
+	movb $0, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask
 
 	/* Preempt here doesn't matter because that will deal with
 	   any pending interrupts.  The pending check may end up being
 	   run on the wrong CPU, but that doesn't hurt. */
 
 	/* Test for pending */
-	testb $0xff, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_pending)
+	testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending
 	jz 1f
 
 2:	call check_events
@@ -69,7 +68,7 @@ ENDPATCH(xen_irq_enable_direct)
 ENTRY(xen_irq_disable_direct)
 	BUG
 
-	movb $1, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask)
+	movb $1, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask
 ENDPATCH(xen_irq_disable_direct)
 	ret
 	ENDPROC(xen_irq_disable_direct)
@@ -87,7 +86,7 @@ ENDPATCH(xen_irq_disable_direct)
 ENTRY(xen_save_fl_direct)
 	BUG
 
-	testb $0xff, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask)
+	testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask
 	setz %ah
 	addb %ah,%ah
 ENDPATCH(xen_save_fl_direct)
@@ -107,13 +106,13 @@ ENTRY(xen_restore_fl_direct)
 	BUG
 
 	testb $X86_EFLAGS_IF>>8, %ah
-	setz PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask)
+	setz PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask
 	/* Preempt here doesn't matter because that will deal with
 	   any pending interrupts.  The pending check may end up being
 	   run on the wrong CPU, but that doesn't hurt. */
 
 	/* check for unmasked and pending */
-	cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_pending)
+	cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending
 	jz 1f
 2:	call check_events
 1:
@@ -196,7 +195,7 @@ ENTRY(xen_sysret64)
 	/* We're already on the usermode stack at this point, but still
 	   with the kernel gs, so we can easily switch back */
 	movq %rsp, %gs:pda_oldrsp
-	movq %gs:pda_kernelstack,%rsp
+	movq PER_CPU_VAR(kernel_stack),%rsp
 
 	pushq $__USER_DS
 	pushq %gs:pda_oldrsp
@@ -213,7 +212,7 @@ ENTRY(xen_sysret32)
 	/* We're already on the usermode stack at this point, but still
 	   with the kernel gs, so we can easily switch back */
 	movq %rsp, %gs:pda_oldrsp
-	movq %gs:pda_kernelstack, %rsp
+	movq PER_CPU_VAR(kernel_stack), %rsp
 
 	pushq $__USER32_DS
 	pushq %gs:pda_oldrsp

From 3d1e42a7cf945e289d6ba26159aa0e2b0645401b Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:58 +0900
Subject: [PATCH 08/12] x86-64: Move oldrsp from PDA to per-cpu.

tj: * in asm-offsets_64.c, pda.h inclusion shouldn't be removed as pda
      is still referenced in the file
    * s/oldrsp/old_rsp/

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/pda.h       |  2 +-
 arch/x86/kernel/asm-offsets_64.c |  1 -
 arch/x86/kernel/entry_64.S       | 10 +++++-----
 arch/x86/kernel/process_64.c     |  8 +++++---
 arch/x86/xen/xen-asm_64.S        |  8 ++++----
 5 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 4d28ffba6e1b..ae23deb99559 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -14,7 +14,7 @@ struct x8664_pda {
 	unsigned long unused1;
 	unsigned long unused2;
 	unsigned long unused3;
-	unsigned long oldrsp;		/* 24 user rsp for system call */
+	unsigned long unused4;
 	int irqcount;			/* 32 Irq nesting counter. Starts -1 */
 	unsigned int unused6;		/* 36 was cpunumber */
 #ifdef CONFIG_CC_STACKPROTECTOR
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index cafff5f4a031..afda6deb8515 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -49,7 +49,6 @@ int main(void)
 	BLANK();
 #undef ENTRY
 #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry))
-	ENTRY(oldrsp); 
 	ENTRY(irqcount);
 	DEFINE(pda_size, sizeof(struct x8664_pda));
 	BLANK();
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index 0dd45859a7a8..7c27da407da7 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -210,7 +210,7 @@ ENTRY(native_usergs_sysret64)
 
 	/* %rsp:at FRAMEEND */
 	.macro FIXUP_TOP_OF_STACK tmp offset=0
-	movq %gs:pda_oldrsp,\tmp
+	movq PER_CPU_VAR(old_rsp),\tmp
 	movq \tmp,RSP+\offset(%rsp)
 	movq $__USER_DS,SS+\offset(%rsp)
 	movq $__USER_CS,CS+\offset(%rsp)
@@ -221,7 +221,7 @@ ENTRY(native_usergs_sysret64)
 
 	.macro RESTORE_TOP_OF_STACK tmp offset=0
 	movq RSP+\offset(%rsp),\tmp
-	movq \tmp,%gs:pda_oldrsp
+	movq \tmp,PER_CPU_VAR(old_rsp)
 	movq EFLAGS+\offset(%rsp),\tmp
 	movq \tmp,R11+\offset(%rsp)
 	.endm
@@ -479,7 +479,7 @@ ENTRY(system_call)
 	 */
 ENTRY(system_call_after_swapgs)
 
-	movq	%rsp,%gs:pda_oldrsp
+	movq	%rsp,PER_CPU_VAR(old_rsp)
 	movq	PER_CPU_VAR(kernel_stack),%rsp
 	/*
 	 * No need to follow this irqs off/on section - it's straight
@@ -523,7 +523,7 @@ sysret_check:
 	CFI_REGISTER	rip,rcx
 	RESTORE_ARGS 0,-ARG_SKIP,1
 	/*CFI_REGISTER	rflags,r11*/
-	movq	%gs:pda_oldrsp, %rsp
+	movq	PER_CPU_VAR(old_rsp), %rsp
 	USERGS_SYSRET64
 
 	CFI_RESTORE_STATE
@@ -833,7 +833,7 @@ common_interrupt:
 	XCPT_FRAME
 	addq $-0x80,(%rsp)		/* Adjust vector to [-256,-1] range */
 	interrupt do_IRQ
-	/* 0(%rsp): oldrsp-ARGOFFSET */
+	/* 0(%rsp): old_rsp-ARGOFFSET */
 ret_from_intr:
 	DISABLE_INTERRUPTS(CLBR_NONE)
 	TRACE_IRQS_OFF
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 6c5f57602108..480128918926 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -60,6 +60,8 @@ asmlinkage extern void ret_from_fork(void);
 DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
 EXPORT_PER_CPU_SYMBOL(current_task);
 
+DEFINE_PER_CPU(unsigned long, old_rsp);
+
 unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED;
 
 static ATOMIC_NOTIFIER_HEAD(idle_notifier);
@@ -395,7 +397,7 @@ start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
 	load_gs_index(0);
 	regs->ip		= new_ip;
 	regs->sp		= new_sp;
-	write_pda(oldrsp, new_sp);
+	percpu_write(old_rsp, new_sp);
 	regs->cs		= __USER_CS;
 	regs->ss		= __USER_DS;
 	regs->flags		= 0x200;
@@ -616,8 +618,8 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 	/*
 	 * Switch the PDA and FPU contexts.
 	 */
-	prev->usersp = read_pda(oldrsp);
-	write_pda(oldrsp, next->usersp);
+	prev->usersp = percpu_read(old_rsp);
+	percpu_write(old_rsp, next->usersp);
 	percpu_write(current_task, next_p);
 
 	percpu_write(kernel_stack,
diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S
index 5a23e8993678..d6fc51f4ce85 100644
--- a/arch/x86/xen/xen-asm_64.S
+++ b/arch/x86/xen/xen-asm_64.S
@@ -194,11 +194,11 @@ RELOC(xen_sysexit, 1b+1)
 ENTRY(xen_sysret64)
 	/* We're already on the usermode stack at this point, but still
 	   with the kernel gs, so we can easily switch back */
-	movq %rsp, %gs:pda_oldrsp
+	movq %rsp, PER_CPU_VAR(old_rsp)
 	movq PER_CPU_VAR(kernel_stack),%rsp
 
 	pushq $__USER_DS
-	pushq %gs:pda_oldrsp
+	pushq PER_CPU_VAR(old_rsp)
 	pushq %r11
 	pushq $__USER_CS
 	pushq %rcx
@@ -211,11 +211,11 @@ RELOC(xen_sysret64, 1b+1)
 ENTRY(xen_sysret32)
 	/* We're already on the usermode stack at this point, but still
 	   with the kernel gs, so we can easily switch back */
-	movq %rsp, %gs:pda_oldrsp
+	movq %rsp, PER_CPU_VAR(old_rsp)
 	movq PER_CPU_VAR(kernel_stack), %rsp
 
 	pushq $__USER32_DS
-	pushq %gs:pda_oldrsp
+	pushq PER_CPU_VAR(old_rsp)
 	pushq %r11
 	pushq $__USER32_CS
 	pushq %rcx

From 5689553076c4a67b83426b076082c63085b7567a Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:58 +0900
Subject: [PATCH 09/12] x86-64: Move irqcount from PDA to per-cpu.

tj: s/irqcount/irq_count/

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/pda.h       |  2 +-
 arch/x86/kernel/asm-offsets_64.c |  1 -
 arch/x86/kernel/cpu/common.c     |  4 ++--
 arch/x86/kernel/entry_64.S       | 14 +++++++-------
 4 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index ae23deb99559..4527d70314d4 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -15,7 +15,7 @@ struct x8664_pda {
 	unsigned long unused2;
 	unsigned long unused3;
 	unsigned long unused4;
-	int irqcount;			/* 32 Irq nesting counter. Starts -1 */
+	int unused5;
 	unsigned int unused6;		/* 36 was cpunumber */
 #ifdef CONFIG_CC_STACKPROTECTOR
 	unsigned long stack_canary;	/* 40 stack canary value */
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index afda6deb8515..64c834a39aa8 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -49,7 +49,6 @@ int main(void)
 	BLANK();
 #undef ENTRY
 #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry))
-	ENTRY(irqcount);
 	DEFINE(pda_size, sizeof(struct x8664_pda));
 	BLANK();
 #undef ENTRY
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 06b6290088f4..e2323ecce1d3 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -893,6 +893,8 @@ 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(unsigned int, irq_count) = -1;
+
 void __cpuinit pda_init(int cpu)
 {
 	struct x8664_pda *pda = cpu_pda(cpu);
@@ -903,8 +905,6 @@ void __cpuinit pda_init(int cpu)
 
 	load_pda_offset(cpu);
 
-	pda->irqcount = -1;
-
 	if (cpu != 0) {
 		if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE)
 			pda->nodenumber = cpu_to_node(cpu);
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index 7c27da407da7..c52b60919163 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -337,12 +337,12 @@ ENTRY(save_args)
 	je 1f
 	SWAPGS
 	/*
-	 * irqcount is used to check if a CPU is already on an interrupt stack
+	 * irq_count is used to check if a CPU is already on an interrupt stack
 	 * or not. While this is essentially redundant with preempt_count it is
 	 * a little cheaper to use a separate counter in the PDA (short of
 	 * moving irq_enter into assembly, which would be too much work)
 	 */
-1:	incl %gs:pda_irqcount
+1:	incl PER_CPU_VAR(irq_count)
 	jne 2f
 	popq_cfi %rax			/* move return address... */
 	mov PER_CPU_VAR(irq_stack_ptr),%rsp
@@ -837,7 +837,7 @@ common_interrupt:
 ret_from_intr:
 	DISABLE_INTERRUPTS(CLBR_NONE)
 	TRACE_IRQS_OFF
-	decl %gs:pda_irqcount
+	decl PER_CPU_VAR(irq_count)
 	leaveq
 	CFI_DEF_CFA_REGISTER	rsp
 	CFI_ADJUST_CFA_OFFSET	-8
@@ -1260,14 +1260,14 @@ ENTRY(call_softirq)
 	CFI_REL_OFFSET rbp,0
 	mov  %rsp,%rbp
 	CFI_DEF_CFA_REGISTER rbp
-	incl %gs:pda_irqcount
+	incl PER_CPU_VAR(irq_count)
 	cmove PER_CPU_VAR(irq_stack_ptr),%rsp
 	push  %rbp			# backlink for old unwinder
 	call __do_softirq
 	leaveq
 	CFI_DEF_CFA_REGISTER	rsp
 	CFI_ADJUST_CFA_OFFSET   -8
-	decl %gs:pda_irqcount
+	decl PER_CPU_VAR(irq_count)
 	ret
 	CFI_ENDPROC
 END(call_softirq)
@@ -1297,7 +1297,7 @@ ENTRY(xen_do_hypervisor_callback)   # do_hypervisor_callback(struct *pt_regs)
 	movq %rdi, %rsp            # we don't return, adjust the stack frame
 	CFI_ENDPROC
 	DEFAULT_FRAME
-11:	incl %gs:pda_irqcount
+11:	incl PER_CPU_VAR(irq_count)
 	movq %rsp,%rbp
 	CFI_DEF_CFA_REGISTER rbp
 	cmovzq PER_CPU_VAR(irq_stack_ptr),%rsp
@@ -1305,7 +1305,7 @@ ENTRY(xen_do_hypervisor_callback)   # do_hypervisor_callback(struct *pt_regs)
 	call xen_evtchn_do_upcall
 	popq %rsp
 	CFI_DEF_CFA_REGISTER rsp
-	decl %gs:pda_irqcount
+	decl PER_CPU_VAR(irq_count)
 	jmp  error_exit
 	CFI_ENDPROC
 END(do_hypervisor_callback)

From e7a22c1ebcc1caa8178df1819d05128bb5b45ab9 Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:59 +0900
Subject: [PATCH 10/12] x86-64: Move nodenumber from PDA to per-cpu.

tj: * s/nodenumber/node_number/
    * removed now unused pda variable from pda_init()

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/pda.h      |  1 -
 arch/x86/include/asm/topology.h |  3 ++-
 arch/x86/kernel/cpu/common.c    | 13 ++++++-------
 arch/x86/kernel/setup_percpu.c  |  4 +++-
 4 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 4527d70314d4..b30ef6bddc43 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -22,7 +22,6 @@ struct x8664_pda {
 					/* gcc-ABI: this canary MUST be at
 					   offset 40!!! */
 #endif
-	short nodenumber;		/* number of current node (32k max) */
 	short in_bootmem;		/* pda lives in bootmem */
 	short isidle;
 } ____cacheline_aligned_in_smp;
diff --git a/arch/x86/include/asm/topology.h b/arch/x86/include/asm/topology.h
index 87ca3fd86e88..ffea1fe03a99 100644
--- a/arch/x86/include/asm/topology.h
+++ b/arch/x86/include/asm/topology.h
@@ -83,7 +83,8 @@ extern cpumask_t *node_to_cpumask_map;
 DECLARE_EARLY_PER_CPU(int, x86_cpu_to_node_map);
 
 /* Returns the number of the current Node. */
-#define numa_node_id()		read_pda(nodenumber)
+DECLARE_PER_CPU(int, node_number);
+#define numa_node_id()		percpu_read(node_number)
 
 #ifdef CONFIG_DEBUG_PER_CPU_MAPS
 extern int cpu_to_node(int cpu);
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index e2323ecce1d3..7976a6a0f65c 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -897,18 +897,11 @@ DEFINE_PER_CPU(unsigned int, irq_count) = -1;
 
 void __cpuinit pda_init(int cpu)
 {
-	struct x8664_pda *pda = cpu_pda(cpu);
-
 	/* Setup up data that may be needed in __get_free_pages early */
 	loadsegment(fs, 0);
 	loadsegment(gs, 0);
 
 	load_pda_offset(cpu);
-
-	if (cpu != 0) {
-		if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE)
-			pda->nodenumber = cpu_to_node(cpu);
-	}
 }
 
 static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks
@@ -978,6 +971,12 @@ void __cpuinit cpu_init(void)
 	if (cpu != 0)
 		pda_init(cpu);
 
+#ifdef CONFIG_NUMA
+	if (cpu != 0 && percpu_read(node_number) == 0 &&
+	    cpu_to_node(cpu) != NUMA_NO_NODE)
+		percpu_write(node_number, cpu_to_node(cpu));
+#endif
+
 	me = current;
 
 	if (cpumask_test_and_set_cpu(cpu, cpu_initialized_mask))
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index 258497f93f4d..efbafbbff584 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -53,6 +53,8 @@ EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
 
 #if defined(CONFIG_NUMA) && defined(CONFIG_X86_64)
 #define	X86_64_NUMA	1	/* (used later) */
+DEFINE_PER_CPU(int, node_number) = 0;
+EXPORT_PER_CPU_SYMBOL(node_number);
 
 /*
  * Map cpu index to node index
@@ -283,7 +285,7 @@ void __cpuinit numa_set_node(int cpu, int node)
 	per_cpu(x86_cpu_to_node_map, cpu) = node;
 
 	if (node != NUMA_NO_NODE)
-		cpu_pda(cpu)->nodenumber = node;
+		per_cpu(node_number, cpu) = node;
 }
 
 void __cpuinit numa_clear_node(int cpu)

From c2558e0eba66b49993e619da66c95a50a97830a3 Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:59 +0900
Subject: [PATCH 11/12] x86-64: Move isidle from PDA to per-cpu.

tj: s/isidle/is_idle/

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/pda.h   | 1 -
 arch/x86/kernel/process_64.c | 5 +++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index b30ef6bddc43..c31ca048a901 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -23,7 +23,6 @@ struct x8664_pda {
 					   offset 40!!! */
 #endif
 	short in_bootmem;		/* pda lives in bootmem */
-	short isidle;
 } ____cacheline_aligned_in_smp;
 
 DECLARE_PER_CPU(struct x8664_pda, __pda);
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 480128918926..4523ff88a69d 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -61,6 +61,7 @@ DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
 EXPORT_PER_CPU_SYMBOL(current_task);
 
 DEFINE_PER_CPU(unsigned long, old_rsp);
+static DEFINE_PER_CPU(unsigned char, is_idle);
 
 unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED;
 
@@ -80,13 +81,13 @@ EXPORT_SYMBOL_GPL(idle_notifier_unregister);
 
 void enter_idle(void)
 {
-	write_pda(isidle, 1);
+	percpu_write(is_idle, 1);
 	atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL);
 }
 
 static void __exit_idle(void)
 {
-	if (test_and_clear_bit_pda(0, isidle) == 0)
+	if (x86_test_and_clear_bit_percpu(0, is_idle) == 0)
 		return;
 	atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL);
 }

From 87b264065880fa696c121dad8498a60524e0f6de Mon Sep 17 00:00:00 2001
From: Brian Gerst <brgerst@gmail.com>
Date: Mon, 19 Jan 2009 00:38:59 +0900
Subject: [PATCH 12/12] x86-64: Use absolute displacements for per-cpu
 accesses.

Accessing memory through %gs should not use rip-relative addressing.
Adding a P prefix for the argument tells gcc to not add (%rip) to
the memory references.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 arch/x86/include/asm/percpu.h | 26 +++++++++++++-------------
 arch/x86/include/asm/system.h |  2 +-
 2 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index 03aa4b00a1c3..165d5272ece1 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -39,10 +39,10 @@
 #include <linux/stringify.h>
 
 #ifdef CONFIG_SMP
-#define __percpu_seg_str	"%%"__stringify(__percpu_seg)":"
+#define __percpu_arg(x)		"%%"__stringify(__percpu_seg)":%P" #x
 #define __my_cpu_offset		percpu_read(this_cpu_off)
 #else
-#define __percpu_seg_str
+#define __percpu_arg(x)		"%" #x
 #endif
 
 /* For arch-specific code, we can use direct single-insn ops (they
@@ -58,22 +58,22 @@ do {							\
 	}						\
 	switch (sizeof(var)) {				\
 	case 1:						\
-		asm(op "b %1,"__percpu_seg_str"%0"	\
+		asm(op "b %1,"__percpu_arg(0)		\
 		    : "+m" (var)			\
 		    : "ri" ((T__)val));			\
 		break;					\
 	case 2:						\
-		asm(op "w %1,"__percpu_seg_str"%0"	\
+		asm(op "w %1,"__percpu_arg(0)		\
 		    : "+m" (var)			\
 		    : "ri" ((T__)val));			\
 		break;					\
 	case 4:						\
-		asm(op "l %1,"__percpu_seg_str"%0"	\
+		asm(op "l %1,"__percpu_arg(0)		\
 		    : "+m" (var)			\
 		    : "ri" ((T__)val));			\
 		break;					\
 	case 8:						\
-		asm(op "q %1,"__percpu_seg_str"%0"	\
+		asm(op "q %1,"__percpu_arg(0)		\
 		    : "+m" (var)			\
 		    : "r" ((T__)val));			\
 		break;					\
@@ -86,22 +86,22 @@ do {							\
 	typeof(var) ret__;				\
 	switch (sizeof(var)) {				\
 	case 1:						\
-		asm(op "b "__percpu_seg_str"%1,%0"	\
+		asm(op "b "__percpu_arg(1)",%0"		\
 		    : "=r" (ret__)			\
 		    : "m" (var));			\
 		break;					\
 	case 2:						\
-		asm(op "w "__percpu_seg_str"%1,%0"	\
+		asm(op "w "__percpu_arg(1)",%0"		\
 		    : "=r" (ret__)			\
 		    : "m" (var));			\
 		break;					\
 	case 4:						\
-		asm(op "l "__percpu_seg_str"%1,%0"	\
+		asm(op "l "__percpu_arg(1)",%0"		\
 		    : "=r" (ret__)			\
 		    : "m" (var));			\
 		break;					\
 	case 8:						\
-		asm(op "q "__percpu_seg_str"%1,%0"	\
+		asm(op "q "__percpu_arg(1)",%0"		\
 		    : "=r" (ret__)			\
 		    : "m" (var));			\
 		break;					\
@@ -122,9 +122,9 @@ do {							\
 #define x86_test_and_clear_bit_percpu(bit, var)				\
 ({									\
 	int old__;							\
-	asm volatile("btr %1,"__percpu_seg_str"%c2\n\tsbbl %0,%0"	\
-		     : "=r" (old__)					\
-		     : "dIr" (bit), "i" (&per_cpu__##var) : "memory");	\
+	asm volatile("btr %2,"__percpu_arg(1)"\n\tsbbl %0,%0"		\
+		     : "=r" (old__), "+m" (per_cpu__##var)		\
+		     : "dIr" (bit));					\
 	old__;								\
 })
 
diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h
index 4399aac680e9..d1dc27dba36d 100644
--- a/arch/x86/include/asm/system.h
+++ b/arch/x86/include/asm/system.h
@@ -94,7 +94,7 @@ do {									\
 	     "call __switch_to\n\t"					  \
 	     ".globl thread_return\n"					  \
 	     "thread_return:\n\t"					  \
-	     "movq "__percpu_seg_str"%P[current_task],%%rsi\n\t"	  \
+	     "movq "__percpu_arg([current_task])",%%rsi\n\t"		  \
 	     "movq %P[thread_info](%%rsi),%%r8\n\t"			  \
 	     LOCK_PREFIX "btr  %[tif_fork],%P[ti_flags](%%r8)\n\t"	  \
 	     "movq %%rax,%%rdi\n\t" 					  \