[PATCH] fix preempt-issues with smp_call_function()

Patch from Thomas Schlichter <schlicht@uni-mannheim.de>

Based on a patch from Dave Jones.

It converts a large number of instances of:

	smp_call_function(foo);
	foo();

into

	on_each_cpu(foo);

and in doing so fixes up the preempt-unsafeness of the first version.

30 files changed, 112 insertions, 148 deletions

diff --git a/arch/alpha/kernel/process.c b/arch/alpha/kernel/process.c
index 85b45eee2868..768e2fb81dee 100644
--- a/arch/alpha/kernel/process.c
+++ b/arch/alpha/kernel/process.c
@@ -155,10 +155,7 @@ common_shutdown(int mode, char *restart_cmd)
 	struct halt_info args;
 	args.mode = mode;
 	args.restart_cmd = restart_cmd;
-#ifdef CONFIG_SMP
-	smp_call_function(common_shutdown_1, &args, 1, 0);
-#endif
-	common_shutdown_1(&args);
+	on_each_cpu(common_shutdown_1, &args, 1, 0);
 }
 
 void
diff --git a/arch/alpha/kernel/smp.c b/arch/alpha/kernel/smp.c
index a1ed17628934..05e25b55e72c 100644
--- a/arch/alpha/kernel/smp.c
+++ b/arch/alpha/kernel/smp.c
@@ -899,10 +899,8 @@ void
 smp_imb(void)
 {
 	/* Must wait other processors to flush their icache before continue. */
-	if (smp_call_function(ipi_imb, NULL, 1, 1))
+	if (on_each_cpu(ipi_imb, NULL, 1, 1))
 		printk(KERN_CRIT "smp_imb: timed out\n");
-
-	imb();
 }
 
 static void
@@ -916,11 +914,9 @@ flush_tlb_all(void)
 {
 	/* Although we don't have any data to pass, we do want to
 	   synchronize with the other processors.  */
-	if (smp_call_function(ipi_flush_tlb_all, NULL, 1, 1)) {
+	if (on_each_cpu(ipi_flush_tlb_all, NULL, 1, 1)) {
 		printk(KERN_CRIT "flush_tlb_all: timed out\n");
 	}
-
-	tbia();
 }
 
 #define asn_locked() (cpu_data[smp_processor_id()].asn_lock)
@@ -938,6 +934,8 @@ ipi_flush_tlb_mm(void *x)
 void
 flush_tlb_mm(struct mm_struct *mm)
 {
+	preempt_disable();
+
 	if (mm == current->active_mm) {
 		flush_tlb_current(mm);
 		if (atomic_read(&mm->mm_users) <= 1) {
@@ -948,6 +946,7 @@ flush_tlb_mm(struct mm_struct *mm)
 				if (mm->context[cpu])
 					mm->context[cpu] = 0;
 			}
+			preempt_enable();
 			return;
 		}
 	}
@@ -955,6 +954,8 @@ flush_tlb_mm(struct mm_struct *mm)
 	if (smp_call_function(ipi_flush_tlb_mm, mm, 1, 1)) {
 		printk(KERN_CRIT "flush_tlb_mm: timed out\n");
 	}
+
+	preempt_enable();
 }
 
 struct flush_tlb_page_struct {
@@ -981,6 +982,8 @@ flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
 	struct flush_tlb_page_struct data;
 	struct mm_struct *mm = vma->vm_mm;
 
+	preempt_disable();
+
 	if (mm == current->active_mm) {
 		flush_tlb_current_page(mm, vma, addr);
 		if (atomic_read(&mm->mm_users) <= 1) {
@@ -991,6 +994,7 @@ flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
 				if (mm->context[cpu])
 					mm->context[cpu] = 0;
 			}
+			preempt_enable();
 			return;
 		}
 	}
@@ -1002,6 +1006,8 @@ flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
 	if (smp_call_function(ipi_flush_tlb_page, &data, 1, 1)) {
 		printk(KERN_CRIT "flush_tlb_page: timed out\n");
 	}
+
+	preempt_enable();
 }
 
 void
@@ -1030,6 +1036,8 @@ flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
 	if ((vma->vm_flags & VM_EXEC) == 0)
 		return;
 
+	preempt_disable();
+
 	if (mm == current->active_mm) {
 		__load_new_mm_context(mm);
 		if (atomic_read(&mm->mm_users) <= 1) {
@@ -1040,6 +1048,7 @@ flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
 				if (mm->context[cpu])
 					mm->context[cpu] = 0;
 			}
+			preempt_enable();
 			return;
 		}
 	}
@@ -1047,6 +1056,8 @@ flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
 	if (smp_call_function(ipi_flush_icache_page, mm, 1, 1)) {
 		printk(KERN_CRIT "flush_icache_page: timed out\n");
 	}
+
+	preempt_enable();
 }
 
 #ifdef CONFIG_DEBUG_SPINLOCK
diff --git a/arch/i386/kernel/io_apic.c b/arch/i386/kernel/io_apic.c
index 7f2892e1e35e..8bc3f9fee17c 100644
--- a/arch/i386/kernel/io_apic.c
+++ b/arch/i386/kernel/io_apic.c
@@ -1376,8 +1376,7 @@ void /*__init*/ print_local_APIC(void * dummy)
 
 void print_all_local_APICs (void)
 {
-	smp_call_function(print_local_APIC, NULL, 1, 1);
-	print_local_APIC(NULL);
+	on_each_cpu(print_local_APIC, NULL, 1, 1);
 }
 
 void /*__init*/ print_PIC(void)
@@ -1843,8 +1842,7 @@ static void setup_nmi (void)
 	 */ 
 	printk(KERN_INFO "activating NMI Watchdog ...");
 
-	smp_call_function(enable_NMI_through_LVT0, NULL, 1, 1);
-	enable_NMI_through_LVT0(NULL);
+	on_each_cpu(enable_NMI_through_LVT0, NULL, 1, 1);
 
 	printk(" done.\n");
 }
diff --git a/arch/i386/kernel/ldt.c b/arch/i386/kernel/ldt.c
index 8388fed9f9c2..71a710a64cc8 100644
--- a/arch/i386/kernel/ldt.c
+++ b/arch/i386/kernel/ldt.c
@@ -55,12 +55,14 @@ static int alloc_ldt(mm_context_t *pc, int mincount, int reload)
 	wmb();
 
 	if (reload) {
-		load_LDT(pc);
 #ifdef CONFIG_SMP
 		preempt_disable();
+		load_LDT(pc);
 		if (current->mm->cpu_vm_mask != (1 << smp_processor_id()))
 			smp_call_function(flush_ldt, 0, 1, 1);
 		preempt_enable();
+#else
+		load_LDT(pc);
 #endif
 	}
 	if (oldsize) {
diff --git a/arch/i386/kernel/microcode.c b/arch/i386/kernel/microcode.c
index 3e7b429f7848..0d51b9dc9144 100644
--- a/arch/i386/kernel/microcode.c
+++ b/arch/i386/kernel/microcode.c
@@ -183,11 +183,10 @@ static int do_microcode_update(void)
 	int i, error = 0, err;
 	struct microcode *m;
 
-	if (smp_call_function(do_update_one, NULL, 1, 1) != 0) {
+	if (on_each_cpu(do_update_one, NULL, 1, 1) != 0) {
 		printk(KERN_ERR "microcode: IPI timeout, giving up\n");
 		return -EIO;
 	}
-	do_update_one(NULL);
 
 	for (i=0; i<NR_CPUS; i++) {
 		err = update_req[i].err;
diff --git a/arch/i386/kernel/smp.c b/arch/i386/kernel/smp.c
index 55d21ff02c89..704433220886 100644
--- a/arch/i386/kernel/smp.c
+++ b/arch/i386/kernel/smp.c
@@ -436,7 +436,7 @@ void flush_tlb_page(struct vm_area_struct * vma, unsigned long va)
 	preempt_enable();
 }
 
-static inline void do_flush_tlb_all_local(void)
+static void do_flush_tlb_all(void* info)
 {
 	unsigned long cpu = smp_processor_id();
 
@@ -445,18 +445,9 @@ static inline void do_flush_tlb_all_local(void)
 		leave_mm(cpu);
 }
 
-static void flush_tlb_all_ipi(void* info)
-{
-	do_flush_tlb_all_local();
-}
-
 void flush_tlb_all(void)
 {
-	preempt_disable();
-	smp_call_function (flush_tlb_all_ipi,0,1,1);
-
-	do_flush_tlb_all_local();
-	preempt_enable();
+	on_each_cpu(do_flush_tlb_all, 0, 1, 1);
 }
 
 /*
diff --git a/arch/i386/kernel/sysenter.c b/arch/i386/kernel/sysenter.c
index 5aa698f5447c..2345e5da7d2a 100644
--- a/arch/i386/kernel/sysenter.c
+++ b/arch/i386/kernel/sysenter.c
@@ -95,8 +95,7 @@ static int __init sysenter_setup(void)
 		return 0;
 
 	memcpy((void *) page, sysent, sizeof(sysent));
-	enable_sep_cpu(NULL);
-	smp_call_function(enable_sep_cpu, NULL, 1, 1);
+	on_each_cpu(enable_sep_cpu, NULL, 1, 1);
 	return 0;
 }
 
diff --git a/arch/i386/mach-voyager/voyager_smp.c b/arch/i386/mach-voyager/voyager_smp.c
index 6463639d5b26..6a74600e0b09 100644
--- a/arch/i386/mach-voyager/voyager_smp.c
+++ b/arch/i386/mach-voyager/voyager_smp.c
@@ -1209,8 +1209,8 @@ smp_vic_cpi_interrupt(struct pt_regs regs)
 		smp_call_function_interrupt();
 }
 
-static inline void
-do_flush_tlb_all_local(void)
+static void
+do_flush_tlb_all(void* info)
 {
 	unsigned long cpu = smp_processor_id();
 
@@ -1220,20 +1220,11 @@ do_flush_tlb_all_local(void)
 }
 
 
-static void
-flush_tlb_all_function(void* info)
-{
-	do_flush_tlb_all_local();
-}
-
 /* flush the TLB of every active CPU in the system */
 void
 flush_tlb_all(void)
 {
-	preempt_disable();
-	smp_call_function (flush_tlb_all_function, 0, 1, 1);
-	do_flush_tlb_all_local();
-	preempt_enable();
+	on_each_cpu(do_flush_tlb_all, 0, 1, 1);
 }
 
 /* used to set up the trampoline for other CPUs when the memory manager
diff --git a/arch/i386/mm/pageattr.c b/arch/i386/mm/pageattr.c
index b4e286cec50f..7bb9f7ebe469 100644
--- a/arch/i386/mm/pageattr.c
+++ b/arch/i386/mm/pageattr.c
@@ -130,11 +130,8 @@ __change_page_attr(struct page *page, pgprot_t prot, struct page **oldpage)
 } 
 
 static inline void flush_map(void)
-{	
-#ifdef CONFIG_SMP 
-	smp_call_function(flush_kernel_map, NULL, 1, 1);
-#endif	
-	flush_kernel_map(NULL);
+{
+	on_each_cpu(flush_kernel_map, NULL, 1, 1);
 }
 
 struct deferred_page { 
diff --git a/arch/i386/oprofile/nmi_int.c b/arch/i386/oprofile/nmi_int.c
index e426bbfb3718..c6c30b22a0e2 100644
--- a/arch/i386/oprofile/nmi_int.c
+++ b/arch/i386/oprofile/nmi_int.c
@@ -95,8 +95,7 @@ static int nmi_setup(void)
 	 * without actually triggering any NMIs as this will
 	 * break the core code horrifically.
 	 */
-	smp_call_function(nmi_cpu_setup, NULL, 0, 1);
-	nmi_cpu_setup(0);
+	on_each_cpu(nmi_cpu_setup, NULL, 0, 1);
 	set_nmi_callback(nmi_callback);
 	oprofile_pmdev = set_nmi_pm_callback(oprofile_pm_callback);
 	return 0;
@@ -148,8 +147,7 @@ static void nmi_shutdown(void)
 {
 	unset_nmi_pm_callback(oprofile_pmdev);
 	unset_nmi_callback();
-	smp_call_function(nmi_cpu_shutdown, NULL, 0, 1);
-	nmi_cpu_shutdown(0);
+	on_each_cpu(nmi_cpu_shutdown, NULL, 0, 1);
 }
 
  
@@ -162,8 +160,7 @@ static void nmi_cpu_start(void * dummy)
 
 static int nmi_start(void)
 {
-	smp_call_function(nmi_cpu_start, NULL, 0, 1);
-	nmi_cpu_start(0);
+	on_each_cpu(nmi_cpu_start, NULL, 0, 1);
 	return 0;
 }
  
@@ -177,8 +174,7 @@ static void nmi_cpu_stop(void * dummy)
  
 static void nmi_stop(void)
 {
-	smp_call_function(nmi_cpu_stop, NULL, 0, 1);
-	nmi_cpu_stop(0);
+	on_each_cpu(nmi_cpu_stop, NULL, 0, 1);
 }
 
 
diff --git a/arch/ia64/kernel/smp.c b/arch/ia64/kernel/smp.c
index 9532fe7ff116..a3eff7b9fd66 100644
--- a/arch/ia64/kernel/smp.c
+++ b/arch/ia64/kernel/smp.c
@@ -206,18 +206,18 @@ smp_send_reschedule_all (void)
 void
 smp_flush_tlb_all (void)
 {
-	smp_call_function((void (*)(void *))local_flush_tlb_all, 0, 1, 1);
-	local_flush_tlb_all();
+	on_each_cpu((void (*)(void *))local_flush_tlb_all, 0, 1, 1);
 }
 
 void
 smp_flush_tlb_mm (struct mm_struct *mm)
 {
-	local_finish_flush_tlb_mm(mm);
-
 	/* this happens for the common case of a single-threaded fork():  */
 	if (likely(mm == current->active_mm && atomic_read(&mm->mm_users) == 1))
+	{
+		local_finish_flush_tlb_mm(mm);
 		return;
+	}
 
 	/*
 	 * We could optimize this further by using mm->cpu_vm_mask to track which CPUs
@@ -226,7 +226,7 @@ smp_flush_tlb_mm (struct mm_struct *mm)
 	 * anyhow, and once a CPU is interrupted, the cost of local_flush_tlb_all() is
 	 * rather trivial.
 	 */
-	smp_call_function((void (*)(void *))local_finish_flush_tlb_mm, mm, 1, 1);
+	on_each_cpu((void (*)(void *))local_finish_flush_tlb_mm, mm, 1, 1);
 }
 
 /*
diff --git a/arch/mips64/kernel/smp.c b/arch/mips64/kernel/smp.c
index 4494b7d96ca6..9ccb54fa72b5 100644
--- a/arch/mips64/kernel/smp.c
+++ b/arch/mips64/kernel/smp.c
@@ -195,8 +195,7 @@ static void flush_tlb_all_ipi(void *info)
 
 void flush_tlb_all(void)
 {
-	smp_call_function(flush_tlb_all_ipi, 0, 1, 1);
-	_flush_tlb_all();
+	on_each_cpu(flush_tlb_all_ipi, 0, 1, 1);
 }
 
 static void flush_tlb_mm_ipi(void *mm)
@@ -219,6 +218,8 @@ static void flush_tlb_mm_ipi(void *mm)
 
 void flush_tlb_mm(struct mm_struct *mm)
 {
+	preempt_disable();
+
 	if ((atomic_read(&mm->mm_users) != 1) || (current->mm != mm)) {
 		smp_call_function(flush_tlb_mm_ipi, (void *)mm, 1, 1);
 	} else {
@@ -228,6 +229,8 @@ void flush_tlb_mm(struct mm_struct *mm)
 				CPU_CONTEXT(i, mm) = 0;
 	}
 	_flush_tlb_mm(mm);
+
+	preempt_enable();
 }
 
 struct flush_tlb_data {
@@ -246,6 +249,8 @@ static void flush_tlb_range_ipi(void *info)
 
 void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned long end)
 {
+	preempt_disable();
+
 	if ((atomic_read(&mm->mm_users) != 1) || (current->mm != mm)) {
 		struct flush_tlb_data fd;
 
@@ -260,6 +265,8 @@ void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned l
 				CPU_CONTEXT(i, mm) = 0;
 	}
 	_flush_tlb_range(mm, start, end);
+
+	preempt_enable();
 }
 
 static void flush_tlb_page_ipi(void *info)
@@ -271,6 +278,8 @@ static void flush_tlb_page_ipi(void *info)
 
 void flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
 {
+	preempt_disable();
+
 	if ((atomic_read(&vma->vm_mm->mm_users) != 1) || (current->mm != vma->vm_mm)) {
 		struct flush_tlb_data fd;
 
@@ -284,5 +293,7 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
 				CPU_CONTEXT(i, vma->vm_mm) = 0;
 	}
 	_flush_tlb_page(vma, page);
+
+	preempt_enable();
 }
 
diff --git a/arch/parisc/kernel/cache.c b/arch/parisc/kernel/cache.c
index 646e4c1a61e3..bf551a9745af 100644
--- a/arch/parisc/kernel/cache.c
+++ b/arch/parisc/kernel/cache.c
@@ -39,8 +39,7 @@ static struct pdc_btlb_info btlb_info;
 void
 flush_data_cache(void)
 {
-	smp_call_function((void (*)(void *))flush_data_cache_local, NULL, 1, 1);
-	flush_data_cache_local();
+	on_each_cpu((void (*)(void *))flush_data_cache_local, NULL, 1, 1);
 }
 #endif
 
diff --git a/arch/parisc/kernel/irq.c b/arch/parisc/kernel/irq.c
index d618a3bec3cc..87d7e7cbf4bf 100644
--- a/arch/parisc/kernel/irq.c
+++ b/arch/parisc/kernel/irq.c
@@ -61,20 +61,17 @@ static volatile unsigned long cpu_eiem = 0;
 
 static spinlock_t irq_lock = SPIN_LOCK_UNLOCKED;  /* protect IRQ regions */
 
-#ifdef CONFIG_SMP
 static void cpu_set_eiem(void *info)
 {
 	set_eiem((unsigned long) info);
 }
-#endif
 
 static inline void disable_cpu_irq(void *unused, int irq)
 {
 	unsigned long eirr_bit = EIEM_MASK(irq);
 
 	cpu_eiem &= ~eirr_bit;
-	set_eiem(cpu_eiem);
-        smp_call_function(cpu_set_eiem, (void *) cpu_eiem, 1, 1);
+        on_each_cpu(cpu_set_eiem, (void *) cpu_eiem, 1, 1);
 }
 
 static void enable_cpu_irq(void *unused, int irq)
@@ -83,8 +80,7 @@ static void enable_cpu_irq(void *unused, int irq)
 
 	mtctl(eirr_bit, 23);	/* clear EIRR bit before unmasking */
 	cpu_eiem |= eirr_bit;
-        smp_call_function(cpu_set_eiem, (void *) cpu_eiem, 1, 1);
-	set_eiem(cpu_eiem);
+        on_each_cpu(cpu_set_eiem, (void *) cpu_eiem, 1, 1);
 }
 
 /* mask and disable are the same at the CPU level
@@ -100,8 +96,7 @@ static inline void unmask_cpu_irq(void *unused, int irq)
 	** handle *any* unmasked pending interrupts.
 	** ie We don't need to check for pending interrupts here.
 	*/
-        smp_call_function(cpu_set_eiem, (void *) cpu_eiem, 1, 1);
-	set_eiem(cpu_eiem);
+        on_each_cpu(cpu_set_eiem, (void *) cpu_eiem, 1, 1);
 }
 
 /*
diff --git a/arch/parisc/kernel/smp.c b/arch/parisc/kernel/smp.c
index f31e00df6bad..84eea09ef6c6 100644
--- a/arch/parisc/kernel/smp.c
+++ b/arch/parisc/kernel/smp.c
@@ -401,7 +401,7 @@ static int __init maxcpus(char *str)
 __setup("maxcpus=", maxcpus);
 
 /*
- * Flush all other CPU's tlb and then mine.  Do this with smp_call_function()
+ * Flush all other CPU's tlb and then mine.  Do this with on_each_cpu()
  * as we want to ensure all TLB's flushed before proceeding.
  */
 
@@ -410,8 +410,7 @@ extern void flush_tlb_all_local(void);
 void
 smp_flush_tlb_all(void)
 {
-	smp_call_function((void (*)(void *))flush_tlb_all_local, NULL, 1, 1);
-	flush_tlb_all_local();
+	on_each_cpu((void (*)(void *))flush_tlb_all_local, NULL, 1, 1);
 }
 
 
diff --git a/arch/parisc/mm/init.c b/arch/parisc/mm/init.c
index 0a78eef7169f..3b382ef901da 100644
--- a/arch/parisc/mm/init.c
+++ b/arch/parisc/mm/init.c
@@ -974,8 +974,7 @@ void flush_tlb_all(void)
 	    do_recycle++;
 	}
 	spin_unlock(&sid_lock);
-	smp_call_function((void (*)(void *))flush_tlb_all_local, NULL, 1, 1);
-	flush_tlb_all_local();
+	on_each_cpu((void (*)(void *))flush_tlb_all_local, NULL, 1, 1);
 	if (do_recycle) {
 	    spin_lock(&sid_lock);
 	    recycle_sids(recycle_ndirty,recycle_dirty_array);
diff --git a/arch/ppc/kernel/temp.c b/arch/ppc/kernel/temp.c
index e096d841767b..fba7fb351831 100644
--- a/arch/ppc/kernel/temp.c
+++ b/arch/ppc/kernel/temp.c
@@ -194,10 +194,7 @@ static void tau_timeout_smp(unsigned long unused)
 
 	/* schedule ourselves to be run again */
 	mod_timer(&tau_timer, jiffies + shrink_timer) ;
-#ifdef CONFIG_SMP
-	smp_call_function(tau_timeout, NULL, 1, 0);
-#endif
-	tau_timeout(NULL);
+	on_each_cpu(tau_timeout, NULL, 1, 0);
 }
 
 /*
@@ -239,10 +236,7 @@ int __init TAU_init(void)
 	tau_timer.expires = jiffies + shrink_timer;
 	add_timer(&tau_timer);
 	
-#ifdef CONFIG_SMP
-	smp_call_function(TAU_init_smp, NULL, 1, 0);
-#endif
-	TAU_init_smp(NULL);
+	on_each_cpu(TAU_init_smp, NULL, 1, 0);
 	
 	printk("Thermal assist unit ");
 #ifdef CONFIG_TAU_INT
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c
index e02e5c8c47fe..2781e9e5ef7d 100644
--- a/arch/s390/kernel/smp.c
+++ b/arch/s390/kernel/smp.c
@@ -228,8 +228,7 @@ static void do_machine_restart(void * __unused)
 void machine_restart_smp(char * __unused) 
 {
 	cpu_restart_map = cpu_online_map;
-        smp_call_function(do_machine_restart, NULL, 0, 0);
-	do_machine_restart(NULL);
+        on_each_cpu(do_machine_restart, NULL, 0, 0);
 }
 
 static void do_machine_halt(void * __unused)
@@ -247,8 +246,7 @@ static void do_machine_halt(void * __unused)
 
 void machine_halt_smp(void)
 {
-        smp_call_function(do_machine_halt, NULL, 0, 0);
-	do_machine_halt(NULL);
+        on_each_cpu(do_machine_halt, NULL, 0, 0);
 }
 
 static void do_machine_power_off(void * __unused)
@@ -266,8 +264,7 @@ static void do_machine_power_off(void * __unused)
 
 void machine_power_off_smp(void)
 {
-        smp_call_function(do_machine_power_off, NULL, 0, 0);
-	do_machine_power_off(NULL);
+        on_each_cpu(do_machine_power_off, NULL, 0, 0);
 }
 
 /*
@@ -339,8 +336,7 @@ void smp_ptlb_callback(void *info)
 
 void smp_ptlb_all(void)
 {
-        smp_call_function(smp_ptlb_callback, NULL, 0, 1);
-	local_flush_tlb();
+        on_each_cpu(smp_ptlb_callback, NULL, 0, 1);
 }
 
 /*
@@ -400,8 +396,10 @@ void smp_ctl_set_bit(int cr, int bit) {
 	parms.end_ctl = cr;
 	parms.orvals[cr] = 1 << bit;
 	parms.andvals[cr] = 0xFFFFFFFF;
+	preempt_disable();
 	smp_call_function(smp_ctl_bit_callback, &parms, 0, 1);
         __ctl_set_bit(cr, bit);
+	preempt_enable();
 }
 
 /*
@@ -414,8 +412,10 @@ void smp_ctl_clear_bit(int cr, int bit) {
 	parms.end_ctl = cr;
 	parms.orvals[cr] = 0x00000000;
 	parms.andvals[cr] = ~(1 << bit);
+	preempt_disable();
 	smp_call_function(smp_ctl_bit_callback, &parms, 0, 1);
         __ctl_clear_bit(cr, bit);
+	preempt_enable();
 }
 
 /*
diff --git a/arch/s390x/kernel/smp.c b/arch/s390x/kernel/smp.c
index 8abfde52cb9a..469a9f647d44 100644
--- a/arch/s390x/kernel/smp.c
+++ b/arch/s390x/kernel/smp.c
@@ -227,8 +227,7 @@ static void do_machine_restart(void * __unused)
 void machine_restart_smp(char * __unused) 
 {
 	cpu_restart_map = cpu_online_map;
-        smp_call_function(do_machine_restart, NULL, 0, 0);
-	do_machine_restart(NULL);
+        on_each_cpu(do_machine_restart, NULL, 0, 0);
 }
 
 static void do_machine_halt(void * __unused)
@@ -246,8 +245,7 @@ static void do_machine_halt(void * __unused)
 
 void machine_halt_smp(void)
 {
-        smp_call_function(do_machine_halt, NULL, 0, 0);
-	do_machine_halt(NULL);
+        on_each_cpu(do_machine_halt, NULL, 0, 0);
 }
 
 static void do_machine_power_off(void * __unused)
@@ -265,8 +263,7 @@ static void do_machine_power_off(void * __unused)
 
 void machine_power_off_smp(void)
 {
-        smp_call_function(do_machine_power_off, NULL, 0, 0);
-	do_machine_power_off(NULL);
+        on_each_cpu(do_machine_power_off, NULL, 0, 0);
 }
 
 /*
@@ -383,8 +380,10 @@ void smp_ctl_set_bit(int cr, int bit) {
 	parms.end_ctl = cr;
 	parms.orvals[cr] = 1 << bit;
 	parms.andvals[cr] = -1L;
+	preempt_disable();
 	smp_call_function(smp_ctl_bit_callback, &parms, 0, 1);
         __ctl_set_bit(cr, bit);
+	preempt_enable();
 }
 
 /*
@@ -397,8 +396,10 @@ void smp_ctl_clear_bit(int cr, int bit) {
 	parms.end_ctl = cr;
 	parms.orvals[cr] = 0;
 	parms.andvals[cr] = ~(1L << bit);
+	preempt_disable();
 	smp_call_function(smp_ctl_bit_callback, &parms, 0, 1);
         __ctl_clear_bit(cr, bit);
+	preempt_enable();
 }
 
 
diff --git a/arch/x86_64/kernel/bluesmoke.c b/arch/x86_64/kernel/bluesmoke.c
index 4a60fa6003a5..831e848a7ace 100644
--- a/arch/x86_64/kernel/bluesmoke.c
+++ b/arch/x86_64/kernel/bluesmoke.c
@@ -111,11 +111,7 @@ static void mce_checkregs (void *info)
 {
 	u32 low, high;
 	int i;
-	unsigned int *cpu = info;
 
-	BUG_ON (*cpu != smp_processor_id());
-
-	preempt_disable();
 	for (i=0; i<banks; i++) {
 		rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high);
 
@@ -130,20 +126,12 @@ static void mce_checkregs (void *info)
 			wmb();
 		}
 	}
-	preempt_enable();
 }
 
 
 static void mce_timerfunc (unsigned long data)
 {
-	unsigned int i;
-
-	for (i=0; i<smp_num_cpus; i++) {
-		if (i == smp_processor_id())
-			mce_checkregs(&i);
-		else
-			smp_call_function (mce_checkregs, &i, 1, 1);
-	}
+	on_each_cpu (mce_checkregs, NULL, 1, 1);
 
 	/* Refresh the timer. */
 	mce_timer.expires = jiffies + MCE_RATE;
diff --git a/arch/x86_64/kernel/io_apic.c b/arch/x86_64/kernel/io_apic.c
index 02be6caeecc8..06783f127a3d 100644
--- a/arch/x86_64/kernel/io_apic.c
+++ b/arch/x86_64/kernel/io_apic.c
@@ -928,8 +928,7 @@ void /*__init*/ print_local_APIC(void * dummy)
 
 void print_all_local_APICs (void)
 {
-	smp_call_function(print_local_APIC, NULL, 1, 1);
-	print_local_APIC(NULL);
+	on_each_cpu(print_local_APIC, NULL, 1, 1);
 }
 
 void /*__init*/ print_PIC(void)
diff --git a/arch/x86_64/kernel/ldt.c b/arch/x86_64/kernel/ldt.c
index e4d13043c168..91abd555afd3 100644
--- a/arch/x86_64/kernel/ldt.c
+++ b/arch/x86_64/kernel/ldt.c
@@ -60,12 +60,14 @@ static int alloc_ldt(mm_context_t *pc, int mincount, int reload)
 	pc->size = mincount;
 	wmb();
 	if (reload) {
-		load_LDT(pc);
 #ifdef CONFIG_SMP
 		preempt_disable();
+		load_LDT(pc);
 		if (current->mm->cpu_vm_mask != (1<<smp_processor_id()))
 			smp_call_function(flush_ldt, 0, 1, 1);
 		preempt_enable();
+#else
+		load_LDT(pc);
 #endif
 	}
 	if (oldsize) {
diff --git a/arch/x86_64/kernel/smp.c b/arch/x86_64/kernel/smp.c
index 557bfa365399..c5d1957b1f9f 100644
--- a/arch/x86_64/kernel/smp.c
+++ b/arch/x86_64/kernel/smp.c
@@ -328,7 +328,7 @@ void flush_tlb_page(struct vm_area_struct * vma, unsigned long va)
 	preempt_enable();
 }
 
-static inline void do_flush_tlb_all_local(void)
+static void do_flush_tlb_all(void* info)
 {
 	unsigned long cpu = smp_processor_id();
 
@@ -337,18 +337,9 @@ static inline void do_flush_tlb_all_local(void)
 		leave_mm(cpu);
 }
 
-static void flush_tlb_all_ipi(void* info)
-{
-	do_flush_tlb_all_local();
-}
-
 void flush_tlb_all(void)
 {
-	preempt_disable();
-	smp_call_function (flush_tlb_all_ipi,0,1,1);
-
-	do_flush_tlb_all_local();
-	preempt_enable();
+	on_each_cpu(do_flush_tlb_all, 0, 1, 1);
 }
 
 void smp_kdb_stop(void)
diff --git a/arch/x86_64/mm/pageattr.c b/arch/x86_64/mm/pageattr.c
index fabd1f60560d..4eaf4e5df48d 100644
--- a/arch/x86_64/mm/pageattr.c
+++ b/arch/x86_64/mm/pageattr.c
@@ -123,12 +123,7 @@ __change_page_attr(unsigned long address, struct page *page, pgprot_t prot,
 
 static inline void flush_map(unsigned long address)
 {	
-	preempt_disable();
-#ifdef CONFIG_SMP 
-	smp_call_function(flush_kernel_map, (void *)address, 1, 1);
-#endif	
-	flush_kernel_map((void *)address);
-	preempt_enable();
+	on_each_cpu(flush_kernel_map, (void *)address, 1, 1);
 }
 
 struct deferred_page { 
diff --git a/drivers/char/agp/agp.h b/drivers/char/agp/agp.h
index c1ab6b0a10cd..3f7aaa76a5bb 100644
--- a/drivers/char/agp/agp.h
+++ b/drivers/char/agp/agp.h
@@ -42,9 +42,8 @@ static void ipi_handler(void *null)
 
 static void __attribute__((unused)) global_cache_flush(void)
 {
-	if (smp_call_function(ipi_handler, NULL, 1, 1) != 0)
+	if (on_each_cpu(ipi_handler, NULL, 1, 1) != 0)
 		panic(PFX "timed out waiting for the other CPUs!\n");
-	flush_agp_cache();
 }
 #else
 static inline void global_cache_flush(void)
diff --git a/drivers/s390/char/sclp.c b/drivers/s390/char/sclp.c
index 60f5a3bd4fe6..cfef881fb52f 100644
--- a/drivers/s390/char/sclp.c
+++ b/drivers/s390/char/sclp.c
@@ -481,8 +481,7 @@ static void
 do_machine_quiesce(void)
 {
 	cpu_quiesce_map = cpu_online_map;
-	smp_call_function(do_load_quiesce_psw, NULL, 0, 0);
-	do_load_quiesce_psw(NULL);
+	on_each_cpu(do_load_quiesce_psw, NULL, 0, 0);
 }
 #else
 static void
diff --git a/fs/buffer.c b/fs/buffer.c
index 9258e2d6b6fe..7a04adf9c174 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -1403,10 +1403,7 @@ static void invalidate_bh_lru(void *arg)
 	
 static void invalidate_bh_lrus(void)
 {
-	preempt_disable();
-	invalidate_bh_lru(NULL);
-	smp_call_function(invalidate_bh_lru, NULL, 1, 1);
-	preempt_enable();
+	on_each_cpu(invalidate_bh_lru, NULL, 1, 1);
 }
 
 void set_bh_page(struct buffer_head *bh,
diff --git a/include/asm-parisc/cacheflush.h b/include/asm-parisc/cacheflush.h
index 325dd36a2849..b7012daae860 100644
--- a/include/asm-parisc/cacheflush.h
+++ b/include/asm-parisc/cacheflush.h
@@ -25,16 +25,10 @@ flush_page_to_ram(struct page *page)
 
 extern void flush_cache_all_local(void);
 
-#ifdef CONFIG_SMP
 static inline void flush_cache_all(void)
 {
-	smp_call_function((void (*)(void *))flush_cache_all_local, NULL, 1, 1);
-	flush_cache_all_local();
+	on_each_cpu((void (*)(void *))flush_cache_all_local, NULL, 1, 1);
 }
-#else
-#define flush_cache_all flush_cache_all_local
-#endif
-
 
 /* The following value needs to be tuned and probably scaled with the
  * cache size.
diff --git a/include/linux/smp.h b/include/linux/smp.h
index ce016a544cd4..81a906924efe 100644
--- a/include/linux/smp.h
+++ b/include/linux/smp.h
@@ -10,9 +10,10 @@
 
 #ifdef CONFIG_SMP
 
+#include <linux/preempt.h>
 #include <linux/kernel.h>
 #include <linux/compiler.h>
-#include <linux/threads.h>
+#include <linux/thread_info.h>
 #include <asm/smp.h>
 #include <asm/bug.h>
 
@@ -54,6 +55,21 @@ extern int smp_call_function (void (*func) (void *info), void *info,
 			      int retry, int wait);
 
 /*
+ * Call a function on all processors
+ */
+static inline int on_each_cpu(void (*func) (void *info), void *info,
+			      int retry, int wait)
+{
+	int ret = 0;
+
+	preempt_disable();
+	ret = smp_call_function(func, info, retry, wait);
+	func(info);
+	preempt_enable();
+	return ret;
+}
+
+/*
  * True once the per process idle is forked
  */
 extern int smp_threads_ready;
@@ -96,6 +112,7 @@ void smp_prepare_boot_cpu(void);
 #define hard_smp_processor_id()			0
 #define smp_threads_ready			1
 #define smp_call_function(func,info,retry,wait)	({ 0; })
+#define on_each_cpu(func,info,retry,wait)	({ func(info); 0; })
 static inline void smp_send_reschedule(int cpu) { }
 static inline void smp_send_reschedule_all(void) { }
 #define cpu_online_map				1
diff --git a/mm/slab.c b/mm/slab.c
index c199cf890d58..d5f5a178f633 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -1116,12 +1116,16 @@ static inline void check_spinlock_acquired(kmem_cache_t *cachep)
 static void smp_call_function_all_cpus(void (*func) (void *arg), void *arg)
 {
 	check_irq_on();
+	preempt_disable();
+
 	local_irq_disable();
 	func(arg);
 	local_irq_enable();
 
 	if (smp_call_function(func, arg, 1, 1))
 		BUG();
+
+	preempt_enable();
 }
 
 static void free_block (kmem_cache_t* cachep, void** objpp, int len);