percpu data changes

percpu data changes

[Andrew Morton]

It seems that powerpc has gone and changed their implementation of percpu
data so that there is no memory allocated for not-possible CPUs.  To save a
bit of RAM.

This means that any code which does

	for (i = 0; i < NR_CPUS; i++)
		touch(percpudata(i))

will explode on powerpc.

I have a patch queued up (for 2.6.16) which fixes most users of percpu
data.  Make them use for_each_cpu().  I'll send that in a moment.

I also have a patch queued up (from Eric Dumazet) which will convert
architectures which use asm-generic/percpu.h to not allocate memory for
!possible CPUs.  I don't intend to merge this until after 2.6.16.  Once
this is merged, code which _does_ touch percpu data for !possible CPUs will
explode.

As part of this I've just done a kernel-wide sweep for all instances of
NR_CPUS.  I don't _think_ I found any percpu bugs such as the above, but
sometimes it's not clear.

While I was there I did a general cleanup of quite a lot of open-coded
cruft, make it use the nice macros.



So please review the below post-2.6.16 patch with an eye to the following:

- Any stuff you don't want added to your arch

- Any code which _does_ affect percpu data and which might possibly be
  used on powerpc.  Such changes will need to go into 2.6.16.

Thanks.



From: Andrew Morton <akpm@osdl.org>

When we stop allocating percpu memory for not-possible CPUs we must not touch
the percpu data for not-possible CPUs at all.  The correct way of doing this
is to test cpu_possible() or to use for_each_cpu().

This patch is a kernel-wide sweep of all instances of NR_CPUS.  I found very
few instances of this bug, if any.  But the patch converts lots of open-coded
test to use the preferred helper macros.

Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Kyle McMartin <kyle@parisc-linux.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Andi Kleen <ak@muc.de>
Cc: Christian Zankel <chris@zankel.net>
Cc: Philippe Elie <phil.el@wanadoo.fr>
Cc: Nathan Scott <nathans@sgi.com>
Cc: Jens Axboe <axboe@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

 arch/cris/kernel/irq.c                     |   10 ++----
 arch/frv/kernel/irq.c                      |   10 ++----
 arch/i386/kernel/cpu/cpufreq/powernow-k8.c |    4 --
 arch/i386/kernel/io_apic.c                 |   22 +++++--------
 arch/i386/kernel/nmi.c                     |    4 +-
 arch/i386/oprofile/nmi_int.c               |    7 +---
 arch/m32r/kernel/irq.c                     |   10 ++----
 arch/mips/kernel/irq.c                     |   10 ++----
 arch/mips/kernel/smp.c                     |    4 +-
 arch/mips/sgi-ip27/ip27-irq.c              |    5 ---
 arch/parisc/kernel/smp.c                   |   25 ++++++---------
 arch/powerpc/kernel/irq.c                  |    5 +--
 arch/powerpc/kernel/setup-common.c         |    5 +--
 arch/powerpc/kernel/setup_32.c             |    5 +--
 arch/powerpc/platforms/powermac/smp.c      |    4 --
 arch/ppc/kernel/setup.c                    |   10 ++----
 arch/s390/kernel/smp.c                     |    4 --
 arch/sh/kernel/irq.c                       |    5 +--
 arch/sh/kernel/setup.c                     |    5 +--
 arch/sh64/kernel/irq.c                     |    5 +--
 arch/sparc/kernel/irq.c                    |    5 +--
 arch/sparc/kernel/smp.c                    |   24 ++++++---------
 arch/sparc/kernel/sun4d_irq.c              |    8 +----
 arch/sparc/kernel/sun4d_smp.c              |    8 +----
 arch/sparc/kernel/sun4m_smp.c              |    6 +--
 arch/sparc64/kernel/irq.c                  |    4 --
 arch/sparc64/kernel/setup.c                |   15 +++------
 arch/sparc64/kernel/smp.c                  |   30 +++++++------------
 arch/x86_64/kernel/irq.c                   |   21 +++++--------
 arch/x86_64/kernel/nmi.c                   |    6 +--
 arch/xtensa/kernel/irq.c                   |   15 +++------
 drivers/net/loopback.c                     |    4 --
 drivers/oprofile/cpu_buffer.c              |    3 -
 fs/xfs/linux-2.6/xfs_stats.c               |    7 +---
 fs/xfs/linux-2.6/xfs_sysctl.c              |    3 -
 include/asm-alpha/mmu_context.h            |    5 +--
 include/asm-alpha/topology.h               |    4 +-
 include/asm-generic/percpu.h               |    7 +---
 include/asm-powerpc/percpu.h               |    7 +---
 include/asm-s390/percpu.h                  |    7 +---
 include/asm-sparc64/percpu.h               |    7 +---
 include/asm-x86_64/percpu.h                |    7 +---
 include/linux/genhd.h                      |   14 ++------
 43 files changed, 144 insertions(+), 232 deletions(-)

diff -puN arch/cris/kernel/irq.c~more-for_each_cpu-conversions arch/cris/kernel/irq.c
--- devel/arch/cris/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/cris/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -52,9 +52,8 @@ int show_interrupts(struct seq_file *p, 
 
 	if (i == 0) {
 		seq_printf(p, "           ");
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 		seq_putc(p, '\n');
 	}
 
@@ -67,9 +66,8 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
 #endif
 		seq_printf(p, " %14s", irq_desc[i].handler->typename);
 		seq_printf(p, "  %s", action->name);
diff -puN arch/frv/kernel/irq.c~more-for_each_cpu-conversions arch/frv/kernel/irq.c
--- devel/arch/frv/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/frv/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -75,9 +75,8 @@ int show_interrupts(struct seq_file *p, 
 	switch (i) {
 	case 0:
 		seq_printf(p, "           ");
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 
 		seq_putc(p, '\n');
 		break;
@@ -100,9 +99,8 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", kstat_cpu(j).irqs[i - 1]);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i - 1]);
 #endif
 
 		level = group->sources[ix]->level - frv_irq_levels;
diff -puN arch/i386/kernel/cpu/cpufreq/powernow-k8.c~more-for_each_cpu-conversions arch/i386/kernel/cpu/cpufreq/powernow-k8.c
--- devel/arch/i386/kernel/cpu/cpufreq/powernow-k8.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/i386/kernel/cpu/cpufreq/powernow-k8.c	2006-02-03 01:00:28.000000000 -0800
@@ -1145,9 +1145,7 @@ static int __cpuinit powernowk8_init(voi
 {
 	unsigned int i, supported_cpus = 0;
 
-	for (i=0; i<NR_CPUS; i++) {
-		if (!cpu_online(i))
-			continue;
+	for_each_cpu(i) {
 		if (check_supported_cpu(i))
 			supported_cpus++;
 	}
diff -puN arch/i386/kernel/io_apic.c~more-for_each_cpu-conversions arch/i386/kernel/io_apic.c
--- devel/arch/i386/kernel/io_apic.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/i386/kernel/io_apic.c	2006-02-03 01:00:28.000000000 -0800
@@ -349,8 +349,8 @@ static inline void rotate_irqs_among_cpu
 {
 	int i, j;
 	Dprintk("Rotating IRQs among CPUs.\n");
-	for (i = 0; i < NR_CPUS; i++) {
-		for (j = 0; cpu_online(i) && (j < NR_IRQS); j++) {
+	for_each_online_cpu(i) {
+		for (j = 0; j < NR_IRQS; j++) {
 			if (!irq_desc[j].action)
 				continue;
 			/* Is it a significant load ?  */
@@ -379,7 +379,7 @@ static void do_irq_balance(void)
 	unsigned long imbalance = 0;
 	cpumask_t allowed_mask, target_cpu_mask, tmp;
 
-	for (i = 0; i < NR_CPUS; i++) {
+	for_each_cpu(i) {
 		int package_index;
 		CPU_IRQ(i) = 0;
 		if (!cpu_online(i))
@@ -420,9 +420,7 @@ static void do_irq_balance(void)
 		}
 	}
 	/* Find the least loaded processor package */
-	for (i = 0; i < NR_CPUS; i++) {
-		if (!cpu_online(i))
-			continue;
+	for_each_online_cpu(i) {
 		if (i != CPU_TO_PACKAGEINDEX(i))
 			continue;
 		if (min_cpu_irq > CPU_IRQ(i)) {
@@ -439,9 +437,7 @@ tryanothercpu:
 	 */
 	tmp_cpu_irq = 0;
 	tmp_loaded = -1;
-	for (i = 0; i < NR_CPUS; i++) {
-		if (!cpu_online(i))
-			continue;
+	for_each_online_cpu(i) {
 		if (i != CPU_TO_PACKAGEINDEX(i))
 			continue;
 		if (max_cpu_irq <= CPU_IRQ(i)) 
@@ -617,9 +613,7 @@ static int __init balanced_irq_init(void
 	if (smp_num_siblings > 1 && !cpus_empty(tmp))
 		physical_balance = 1;
 
-	for (i = 0; i < NR_CPUS; i++) {
-		if (!cpu_online(i))
-			continue;
+	for_each_online_cpu(i) {
 		irq_cpu_data[i].irq_delta = kmalloc(sizeof(unsigned long) * NR_IRQS, GFP_KERNEL);
 		irq_cpu_data[i].last_irq = kmalloc(sizeof(unsigned long) * NR_IRQS, GFP_KERNEL);
 		if (irq_cpu_data[i].irq_delta == NULL || irq_cpu_data[i].last_irq == NULL) {
@@ -636,9 +630,11 @@ static int __init balanced_irq_init(void
 	else 
 		printk(KERN_ERR "balanced_irq_init: failed to spawn balanced_irq");
 failed:
-	for (i = 0; i < NR_CPUS; i++) {
+	for_each_cpu(i) {
 		kfree(irq_cpu_data[i].irq_delta);
+		irq_cpu_data[i].irq_delta = NULL;
 		kfree(irq_cpu_data[i].last_irq);
+		irq_cpu_data[i].last_irq = NULL;
 	}
 	return 0;
 }
diff -puN arch/i386/kernel/nmi.c~more-for_each_cpu-conversions arch/i386/kernel/nmi.c
--- devel/arch/i386/kernel/nmi.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/i386/kernel/nmi.c	2006-02-03 01:00:28.000000000 -0800
@@ -143,7 +143,7 @@ static int __init check_nmi_watchdog(voi
 	local_irq_enable();
 	mdelay((10*1000)/nmi_hz); // wait 10 ticks
 
-	for (cpu = 0; cpu < NR_CPUS; cpu++) {
+	for_each_cpu(cpu) {
 #ifdef CONFIG_SMP
 		/* Check cpu_callin_map here because that is set
 		   after the timer is started. */
@@ -510,7 +510,7 @@ void touch_nmi_watchdog (void)
 	 * Just reset the alert counters, (other CPUs might be
 	 * spinning on locks we hold):
 	 */
-	for (i = 0; i < NR_CPUS; i++)
+	for_each_cpu(i)
 		alert_counter[i] = 0;
 
 	/*
diff -puN arch/i386/oprofile/nmi_int.c~more-for_each_cpu-conversions arch/i386/oprofile/nmi_int.c
--- devel/arch/i386/oprofile/nmi_int.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/i386/oprofile/nmi_int.c	2006-02-03 01:00:28.000000000 -0800
@@ -122,7 +122,7 @@ static void nmi_save_registers(void * du
 static void free_msrs(void)
 {
 	int i;
-	for (i = 0; i < NR_CPUS; ++i) {
+	for_each_cpu(i) {
 		kfree(cpu_msrs[i].counters);
 		cpu_msrs[i].counters = NULL;
 		kfree(cpu_msrs[i].controls);
@@ -138,10 +138,7 @@ static int allocate_msrs(void)
 	size_t counters_size = sizeof(struct op_msr) * model->num_counters;
 
 	int i;
-	for (i = 0; i < NR_CPUS; ++i) {
-		if (!cpu_online(i))
-			continue;
-
+	for_each_online_cpu(i) {
 		cpu_msrs[i].counters = kmalloc(counters_size, GFP_KERNEL);
 		if (!cpu_msrs[i].counters) {
 			success = 0;
diff -puN arch/m32r/kernel/irq.c~more-for_each_cpu-conversions arch/m32r/kernel/irq.c
--- devel/arch/m32r/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/m32r/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -37,9 +37,8 @@ int show_interrupts(struct seq_file *p, 
 
 	if (i == 0) {
 		seq_printf(p, "           ");
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 		seq_putc(p, '\n');
 	}
 
@@ -52,9 +51,8 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
 #endif
 		seq_printf(p, " %14s", irq_desc[i].handler->typename);
 		seq_printf(p, "  %s", action->name);
diff -puN arch/mips/kernel/irq.c~more-for_each_cpu-conversions arch/mips/kernel/irq.c
--- devel/arch/mips/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/mips/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -68,9 +68,8 @@ int show_interrupts(struct seq_file *p, 
 
 	if (i == 0) {
 		seq_printf(p, "           ");
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 		seq_putc(p, '\n');
 	}
 
@@ -83,9 +82,8 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
 #endif
 		seq_printf(p, " %14s", irq_desc[i].handler->typename);
 		seq_printf(p, "  %s", action->name);
diff -puN arch/mips/kernel/smp.c~more-for_each_cpu-conversions arch/mips/kernel/smp.c
--- devel/arch/mips/kernel/smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/mips/kernel/smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -166,8 +166,8 @@ int smp_call_function (void (*func) (voi
 	mb();
 
 	/* Send a message to all other CPUs and wait for them to respond */
-	for (i = 0; i < NR_CPUS; i++)
-		if (cpu_online(i) && i != cpu)
+	for_each_online_cpu(i)
+		if (i != cpu)
 			core_send_ipi(i, SMP_CALL_FUNCTION);
 
 	/* Wait for response */
diff -puN arch/mips/sgi-ip27/ip27-irq.c~more-for_each_cpu-conversions arch/mips/sgi-ip27/ip27-irq.c
--- devel/arch/mips/sgi-ip27/ip27-irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/mips/sgi-ip27/ip27-irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -88,12 +88,9 @@ static inline int find_level(cpuid_t *cp
 {
 	int cpu, i;
 
-	for (cpu = 0; cpu <= NR_CPUS; cpu++) {
+	for_each_online_cpu(cpu) {
 		struct slice_data *si = cpu_data[cpu].data;
 
-		if (!cpu_online(cpu))
-			continue;
-
 		for (i = BASE_PCI_IRQ; i < LEVELS_PER_SLICE; i++)
 			if (si->level_to_irq[i] == irq) {
 				*cpunum = cpu;
diff -puN arch/parisc/kernel/smp.c~more-for_each_cpu-conversions arch/parisc/kernel/smp.c
--- devel/arch/parisc/kernel/smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/parisc/kernel/smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -298,8 +298,8 @@ send_IPI_allbutself(enum ipi_message_typ
 {
 	int i;
 	
-	for (i = 0; i < NR_CPUS; i++) {
-		if (cpu_online(i) && i != smp_processor_id())
+	for_each_online_cpu(i) {
+		if (i != smp_processor_id())
 			send_IPI_single(i, op);
 	}
 }
@@ -643,14 +643,13 @@ int sys_cpus(int argc, char **argv)
 	if ( argc == 1 ){
 	
 #ifdef DUMP_MORE_STATE
-		for(i=0; i<NR_CPUS; i++) {
+		for_each_online_cpu(i) {
 			int cpus_per_line = 4;
-			if(cpu_online(i)) {
-				if (j++ % cpus_per_line)
-					printk(" %3d",i);
-				else
-					printk("\n %3d",i);
-			}
+
+			if (j++ % cpus_per_line)
+				printk(" %3d",i);
+			else
+				printk("\n %3d",i);
 		}
 		printk("\n"); 
 #else
@@ -659,9 +658,7 @@ int sys_cpus(int argc, char **argv)
 	} else if((argc==2) && !(strcmp(argv[1],"-l"))) {
 		printk("\nCPUSTATE  TASK CPUNUM CPUID HARDCPU(HPA)\n");
 #ifdef DUMP_MORE_STATE
-		for(i=0;i<NR_CPUS;i++) {
-			if (!cpu_online(i))
-				continue;
+		for_each_online_cpu(i) {
 			if (cpu_data[i].cpuid != NO_PROC_ID) {
 				switch(cpu_data[i].state) {
 					case STATE_RENDEZVOUS:
@@ -695,9 +692,7 @@ int sys_cpus(int argc, char **argv)
 	} else if ((argc==2) && !(strcmp(argv[1],"-s"))) { 
 #ifdef DUMP_MORE_STATE
      		printk("\nCPUSTATE   CPUID\n");
-		for (i=0;i<NR_CPUS;i++) {
-			if (!cpu_online(i))
-				continue;
+		for_each_online_cpu(i) {
 			if (cpu_data[i].cpuid != NO_PROC_ID) {
 				switch(cpu_data[i].state) {
 					case STATE_RENDEZVOUS:
diff -puN arch/powerpc/kernel/irq.c~more-for_each_cpu-conversions arch/powerpc/kernel/irq.c
--- devel/arch/powerpc/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/powerpc/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -137,9 +137,8 @@ skip:
 #ifdef CONFIG_TAU_INT
 		if (tau_initialized){
 			seq_puts(p, "TAU: ");
-			for (j = 0; j < NR_CPUS; j++)
-				if (cpu_online(j))
-					seq_printf(p, "%10u ", tau_interrupts(j));
+			for_each_online_cpu(j)
+				seq_printf(p, "%10u ", tau_interrupts(j));
 			seq_puts(p, "  PowerPC             Thermal Assist (cpu temp)\n");
 		}
 #endif
diff -puN arch/powerpc/kernel/setup_32.c~more-for_each_cpu-conversions arch/powerpc/kernel/setup_32.c
--- devel/arch/powerpc/kernel/setup_32.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/powerpc/kernel/setup_32.c	2006-02-03 01:00:28.000000000 -0800
@@ -272,9 +272,8 @@ int __init ppc_init(void)
 	if ( ppc_md.progress ) ppc_md.progress("             ", 0xffff);
 
 	/* register CPU devices */
-	for (i = 0; i < NR_CPUS; i++)
-		if (cpu_possible(i))
-			register_cpu(&cpu_devices[i], i, NULL);
+	for_each_cpu(i)
+		register_cpu(&cpu_devices[i], i, NULL);
 
 	/* call platform init */
 	if (ppc_md.init != NULL) {
diff -puN arch/powerpc/kernel/setup-common.c~more-for_each_cpu-conversions arch/powerpc/kernel/setup-common.c
--- devel/arch/powerpc/kernel/setup-common.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/powerpc/kernel/setup-common.c	2006-02-03 01:00:28.000000000 -0800
@@ -162,9 +162,8 @@ static int show_cpuinfo(struct seq_file 
 #if defined(CONFIG_SMP) && defined(CONFIG_PPC32)
 		unsigned long bogosum = 0;
 		int i;
-		for (i = 0; i < NR_CPUS; ++i)
-			if (cpu_online(i))
-				bogosum += loops_per_jiffy;
+		for_each_online_cpu(i)
+			bogosum += loops_per_jiffy;
 		seq_printf(m, "total bogomips\t: %lu.%02lu\n",
 			   bogosum/(500000/HZ), bogosum/(5000/HZ) % 100);
 #endif /* CONFIG_SMP && CONFIG_PPC32 */
diff -puN arch/powerpc/platforms/powermac/smp.c~more-for_each_cpu-conversions arch/powerpc/platforms/powermac/smp.c
--- devel/arch/powerpc/platforms/powermac/smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/powerpc/platforms/powermac/smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -191,9 +191,7 @@ static void smp_psurge_message_pass(int 
 	if (num_online_cpus() < 2)
 		return;
 
-	for (i = 0; i < NR_CPUS; i++) {
-		if (!cpu_online(i))
-			continue;
+	for_each_online_cpu(i) {
 		if (target == MSG_ALL
 		    || (target == MSG_ALL_BUT_SELF && i != smp_processor_id())
 		    || target == i) {
diff -puN arch/ppc/kernel/setup.c~more-for_each_cpu-conversions arch/ppc/kernel/setup.c
--- devel/arch/ppc/kernel/setup.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/ppc/kernel/setup.c	2006-02-03 01:00:28.000000000 -0800
@@ -168,9 +168,8 @@ int show_cpuinfo(struct seq_file *m, voi
 		/* Show summary information */
 #ifdef CONFIG_SMP
 		unsigned long bogosum = 0;
-		for (i = 0; i < NR_CPUS; ++i)
-			if (cpu_online(i))
-				bogosum += cpu_data[i].loops_per_jiffy;
+		for_each_online_cpu(i)
+			bogosum += cpu_data[i].loops_per_jiffy;
 		seq_printf(m, "total bogomips\t: %lu.%02lu\n",
 			   bogosum/(500000/HZ), bogosum/(5000/HZ) % 100);
 #endif /* CONFIG_SMP */
@@ -712,9 +711,8 @@ int __init ppc_init(void)
 	if ( ppc_md.progress ) ppc_md.progress("             ", 0xffff);
 
 	/* register CPU devices */
-	for (i = 0; i < NR_CPUS; i++)
-		if (cpu_possible(i))
-			register_cpu(&cpu_devices[i], i, NULL);
+	for_each_cpu(i)
+		register_cpu(&cpu_devices[i], i, NULL);
 
 	/* call platform init */
 	if (ppc_md.init != NULL) {
diff -puN arch/s390/kernel/smp.c~more-for_each_cpu-conversions arch/s390/kernel/smp.c
--- devel/arch/s390/kernel/smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/s390/kernel/smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -767,9 +767,7 @@ void __init smp_prepare_cpus(unsigned in
          */
 	print_cpu_info(&S390_lowcore.cpu_data);
 
-        for(i = 0; i < NR_CPUS; i++) {
-		if (!cpu_possible(i))
-			continue;
+        for_each_cpu(i) {
 		lowcore_ptr[i] = (struct _lowcore *)
 			__get_free_pages(GFP_KERNEL|GFP_DMA, 
 					sizeof(void*) == 8 ? 1 : 0);
diff -puN arch/sh64/kernel/irq.c~more-for_each_cpu-conversions arch/sh64/kernel/irq.c
--- devel/arch/sh64/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sh64/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -53,9 +53,8 @@ int show_interrupts(struct seq_file *p, 
 
 	if (i == 0) {
 		seq_puts(p, "           ");
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 		seq_putc(p, '\n');
 	}
 
diff -puN arch/sh/kernel/irq.c~more-for_each_cpu-conversions arch/sh/kernel/irq.c
--- devel/arch/sh/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sh/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -35,9 +35,8 @@ int show_interrupts(struct seq_file *p, 
 
 	if (i == 0) {
 		seq_puts(p, "           ");
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 		seq_putc(p, '\n');
 	}
 
diff -puN arch/sh/kernel/setup.c~more-for_each_cpu-conversions arch/sh/kernel/setup.c
--- devel/arch/sh/kernel/setup.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sh/kernel/setup.c	2006-02-03 01:00:28.000000000 -0800
@@ -404,9 +404,8 @@ static int __init topology_init(void)
 {
 	int cpu_id;
 
-	for (cpu_id = 0; cpu_id < NR_CPUS; cpu_id++)
-		if (cpu_possible(cpu_id))
-			register_cpu(&cpu[cpu_id], cpu_id, NULL);
+	for_each_cpu(cpu_id)
+		register_cpu(&cpu[cpu_id], cpu_id, NULL);
 
 	return 0;
 }
diff -puN arch/sparc64/kernel/irq.c~more-for_each_cpu-conversions arch/sparc64/kernel/irq.c
--- devel/arch/sparc64/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc64/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -115,9 +115,7 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j = 0; j < NR_CPUS; j++) {
-			if (!cpu_online(j))
-				continue;
+		for_each_online_cpu(j) {
 			seq_printf(p, "%10u ",
 				   kstat_cpu(j).irqs[i]);
 		}
diff -puN arch/sparc64/kernel/setup.c~more-for_each_cpu-conversions arch/sparc64/kernel/setup.c
--- devel/arch/sparc64/kernel/setup.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc64/kernel/setup.c	2006-02-03 01:00:28.000000000 -0800
@@ -690,15 +690,12 @@ static int __init topology_init(void)
 	while (!cpu_find_by_instance(ncpus_probed, NULL, NULL))
 		ncpus_probed++;
 
-	for (i = 0; i < NR_CPUS; i++) {
-		if (cpu_possible(i)) {
-			struct cpu *p = kmalloc(sizeof(*p), GFP_KERNEL);
-
-			if (p) {
-				memset(p, 0, sizeof(*p));
-				register_cpu(p, i, NULL);
-				err = 0;
-			}
+	for_each_cpu(i) {
+		struct cpu *p = kmalloc(sizeof(*p), GFP_KERNEL);
+		if (p) {
+			memset(p, 0, sizeof(*p));
+			register_cpu(p, i, NULL);
+			err = 0;
 		}
 	}
 
diff -puN arch/sparc64/kernel/smp.c~more-for_each_cpu-conversions arch/sparc64/kernel/smp.c
--- devel/arch/sparc64/kernel/smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc64/kernel/smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -54,25 +54,21 @@ void smp_info(struct seq_file *m)
 	int i;
 	
 	seq_printf(m, "State:\n");
-	for (i = 0; i < NR_CPUS; i++) {
-		if (cpu_online(i))
-			seq_printf(m,
-				   "CPU%d:\t\tonline\n", i);
-	}
+	for_each_online_cpu(i)
+		seq_printf(m, "CPU%d:\t\tonline\n", i);
 }
 
 void smp_bogo(struct seq_file *m)
 {
 	int i;
 	
-	for (i = 0; i < NR_CPUS; i++)
-		if (cpu_online(i))
-			seq_printf(m,
-				   "Cpu%dBogo\t: %lu.%02lu\n"
-				   "Cpu%dClkTck\t: %016lx\n",
-				   i, cpu_data(i).udelay_val / (500000/HZ),
-				   (cpu_data(i).udelay_val / (5000/HZ)) % 100,
-				   i, cpu_data(i).clock_tick);
+	for_each_online_cpu(i)
+		seq_printf(m,
+			   "Cpu%dBogo\t: %lu.%02lu\n"
+			   "Cpu%dClkTck\t: %016lx\n",
+			   i, cpu_data(i).udelay_val / (500000/HZ),
+			   (cpu_data(i).udelay_val / (5000/HZ)) % 100,
+			   i, cpu_data(i).clock_tick);
 }
 
 void __init smp_store_cpu_info(int id)
@@ -1071,7 +1067,7 @@ int setup_profiling_timer(unsigned int m
 		return -EINVAL;
 
 	spin_lock_irqsave(&prof_setup_lock, flags);
-	for (i = 0; i < NR_CPUS; i++)
+	for_each_cpu(i)
 		prof_multiplier(i) = multiplier;
 	current_tick_offset = (timer_tick_offset / multiplier);
 	spin_unlock_irqrestore(&prof_setup_lock, flags);
@@ -1140,10 +1136,8 @@ void __init smp_cpus_done(unsigned int m
 	unsigned long bogosum = 0;
 	int i;
 
-	for (i = 0; i < NR_CPUS; i++) {
-		if (cpu_online(i))
-			bogosum += cpu_data(i).udelay_val;
-	}
+	for_each_online_cpu(i)
+		bogosum += cpu_data(i).udelay_val;
 	printk("Total of %ld processors activated "
 	       "(%lu.%02lu BogoMIPS).\n",
 	       (long) num_online_cpus(),
diff -puN arch/sparc/kernel/irq.c~more-for_each_cpu-conversions arch/sparc/kernel/irq.c
--- devel/arch/sparc/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -184,9 +184,8 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j = 0; j < NR_CPUS; j++) {
-			if (cpu_online(j))
-				seq_printf(p, "%10u ",
+		for_each_online_cpu(j) {
+			seq_printf(p, "%10u ",
 				    kstat_cpu(cpu_logical_map(j)).irqs[i]);
 		}
 #endif
diff -puN arch/sparc/kernel/smp.c~more-for_each_cpu-conversions arch/sparc/kernel/smp.c
--- devel/arch/sparc/kernel/smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc/kernel/smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -243,9 +243,8 @@ int setup_profiling_timer(unsigned int m
 		return -EINVAL;
 
 	spin_lock_irqsave(&prof_setup_lock, flags);
-	for(i = 0; i < NR_CPUS; i++) {
-		if (cpu_possible(i))
-			load_profile_irq(i, lvl14_resolution / multiplier);
+	for_each_cpu(i) {
+		load_profile_irq(i, lvl14_resolution / multiplier);
 		prof_multiplier(i) = multiplier;
 	}
 	spin_unlock_irqrestore(&prof_setup_lock, flags);
@@ -273,13 +272,12 @@ void smp_bogo(struct seq_file *m)
 {
 	int i;
 	
-	for (i = 0; i < NR_CPUS; i++) {
-		if (cpu_online(i))
-			seq_printf(m,
-				   "Cpu%dBogo\t: %lu.%02lu\n", 
-				   i,
-				   cpu_data(i).udelay_val/(500000/HZ),
-				   (cpu_data(i).udelay_val/(5000/HZ))%100);
+	for_each_online_cpu(i) {
+		seq_printf(m,
+			   "Cpu%dBogo\t: %lu.%02lu\n",
+			   i,
+			   cpu_data(i).udelay_val/(500000/HZ),
+			   (cpu_data(i).udelay_val/(5000/HZ))%100);
 	}
 }
 
@@ -288,8 +286,6 @@ void smp_info(struct seq_file *m)
 	int i;
 
 	seq_printf(m, "State:\n");
-	for (i = 0; i < NR_CPUS; i++) {
-		if (cpu_online(i))
-			seq_printf(m, "CPU%d\t\t: online\n", i);
-	}
+	for_each_online_cpu(i)
+		seq_printf(m, "CPU%d\t\t: online\n", i);
 }
diff -puN arch/sparc/kernel/sun4d_irq.c~more-for_each_cpu-conversions arch/sparc/kernel/sun4d_irq.c
--- devel/arch/sparc/kernel/sun4d_irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc/kernel/sun4d_irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -103,11 +103,9 @@ found_it:	seq_printf(p, "%3d: ", i);
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (x = 0; x < NR_CPUS; x++) {
-			if (cpu_online(x))
-				seq_printf(p, "%10u ",
-				       kstat_cpu(cpu_logical_map(x)).irqs[i]);
-		}
+		for_each_online_cpu(x)
+			seq_printf(p, "%10u ",
+			       kstat_cpu(cpu_logical_map(x)).irqs[i]);
 #endif
 		seq_printf(p, "%c %s",
 			(action->flags & SA_INTERRUPT) ? '+' : ' ',
diff -puN arch/sparc/kernel/sun4d_smp.c~more-for_each_cpu-conversions arch/sparc/kernel/sun4d_smp.c
--- devel/arch/sparc/kernel/sun4d_smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc/kernel/sun4d_smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -249,11 +249,9 @@ void __init smp4d_boot_cpus(void)
 	} else {
 		unsigned long bogosum = 0;
 		
-		for(i = 0; i < NR_CPUS; i++) {
-			if (cpu_isset(i, cpu_present_map)) {
-				bogosum += cpu_data(i).udelay_val;
-				smp_highest_cpu = i;
-			}
+		for_each_present_cpu(i) {
+			bogosum += cpu_data(i).udelay_val;
+			smp_highest_cpu = i;
 		}
 		SMP_PRINTK(("Total of %d Processors activated (%lu.%02lu BogoMIPS).\n", cpucount + 1, bogosum/(500000/HZ), (bogosum/(5000/HZ))%100));
 		printk("Total of %d Processors activated (%lu.%02lu BogoMIPS).\n",
diff -puN arch/sparc/kernel/sun4m_smp.c~more-for_each_cpu-conversions arch/sparc/kernel/sun4m_smp.c
--- devel/arch/sparc/kernel/sun4m_smp.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/sparc/kernel/sun4m_smp.c	2006-02-03 01:00:28.000000000 -0800
@@ -218,10 +218,8 @@ void __init smp4m_boot_cpus(void)
 		cpu_present_map = cpumask_of_cpu(smp_processor_id());
 	} else {
 		unsigned long bogosum = 0;
-		for(i = 0; i < NR_CPUS; i++) {
-			if (cpu_isset(i, cpu_present_map))
-				bogosum += cpu_data(i).udelay_val;
-		}
+		for_each_present_cpu(i)
+			bogosum += cpu_data(i).udelay_val;
 		printk("Total of %d Processors activated (%lu.%02lu BogoMIPS).\n",
 		       cpucount + 1,
 		       bogosum/(500000/HZ),
diff -puN arch/x86_64/kernel/irq.c~more-for_each_cpu-conversions arch/x86_64/kernel/irq.c
--- devel/arch/x86_64/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/x86_64/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -38,9 +38,8 @@ int show_interrupts(struct seq_file *p, 
 
 	if (i == 0) {
 		seq_printf(p, "           ");
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 		seq_putc(p, '\n');
 	}
 
@@ -53,10 +52,8 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-			seq_printf(p, "%10u ",
-				kstat_cpu(j).irqs[i]);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
 #endif
 		seq_printf(p, " %14s", irq_desc[i].handler->typename);
 
@@ -68,15 +65,13 @@ skip:
 		spin_unlock_irqrestore(&irq_desc[i].lock, flags);
 	} else if (i == NR_IRQS) {
 		seq_printf(p, "NMI: ");
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", cpu_pda(j)->__nmi_count);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", cpu_pda(j)->__nmi_count);
 		seq_putc(p, '\n');
 #ifdef CONFIG_X86_LOCAL_APIC
 		seq_printf(p, "LOC: ");
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", cpu_pda(j)->apic_timer_irqs);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", cpu_pda(j)->apic_timer_irqs);
 		seq_putc(p, '\n');
 #endif
 		seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
diff -puN arch/x86_64/kernel/nmi.c~more-for_each_cpu-conversions arch/x86_64/kernel/nmi.c
--- devel/arch/x86_64/kernel/nmi.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/x86_64/kernel/nmi.c	2006-02-03 01:00:28.000000000 -0800
@@ -161,9 +161,7 @@ int __init check_nmi_watchdog (void)
 	local_irq_enable();
 	mdelay((10*1000)/nmi_hz); // wait 10 ticks
 
-	for (cpu = 0; cpu < NR_CPUS; cpu++) {
-		if (!cpu_online(cpu))
-			continue;
+	for_each_online_cpu(cpu) {
 		if (cpu_pda(cpu)->__nmi_count - counts[cpu] <= 5) {
 			endflag = 1;
 			printk("CPU#%d: NMI appears to be stuck (%d->%d)!\n",
@@ -462,7 +460,7 @@ void touch_nmi_watchdog (void)
 	 * do it ourselves because the alert count increase is not
 	 * atomic.
 	 */
-	for (i = 0; i < NR_CPUS; i++)
+	for_each_cpu(i)
 		per_cpu(nmi_touch, i) = 1;
 
  	touch_softlockup_watchdog();
diff -puN arch/xtensa/kernel/irq.c~more-for_each_cpu-conversions arch/xtensa/kernel/irq.c
--- devel/arch/xtensa/kernel/irq.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/arch/xtensa/kernel/irq.c	2006-02-03 01:00:28.000000000 -0800
@@ -83,9 +83,8 @@ int show_interrupts(struct seq_file *p, 
 
 	if (i == 0) {
 		seq_printf(p, "           ");
-		for (j=0; j<NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "CPU%d       ",j);
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
 		seq_putc(p, '\n');
 	}
 
@@ -98,9 +97,8 @@ int show_interrupts(struct seq_file *p, 
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
 #endif
 		seq_printf(p, " %14s", irq_desc[i].handler->typename);
 		seq_printf(p, "  %s", action->name);
@@ -113,9 +111,8 @@ skip:
 		spin_unlock_irqrestore(&irq_desc[i].lock, flags);
 	} else if (i == NR_IRQS) {
 		seq_printf(p, "NMI: ");
-		for (j = 0; j < NR_CPUS; j++)
-			if (cpu_online(j))
-				seq_printf(p, "%10u ", nmi_count(j));
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", nmi_count(j));
 		seq_putc(p, '\n');
 		seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
 	}
diff -puN drivers/net/loopback.c~more-for_each_cpu-conversions drivers/net/loopback.c
--- devel/drivers/net/loopback.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/drivers/net/loopback.c	2006-02-03 01:00:28.000000000 -0800
@@ -172,11 +172,9 @@ static struct net_device_stats *get_stat
 
 	memset(stats, 0, sizeof(struct net_device_stats));
 
-	for (i=0; i < NR_CPUS; i++) {
+	for_each_cpu(i) {
 		struct net_device_stats *lb_stats;
 
-		if (!cpu_possible(i)) 
-			continue;
 		lb_stats = &per_cpu(loopback_stats, i);
 		stats->rx_bytes   += lb_stats->rx_bytes;
 		stats->tx_bytes   += lb_stats->tx_bytes;
diff -puN drivers/oprofile/cpu_buffer.c~more-for_each_cpu-conversions drivers/oprofile/cpu_buffer.c
--- devel/drivers/oprofile/cpu_buffer.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/drivers/oprofile/cpu_buffer.c	2006-02-03 01:00:28.000000000 -0800
@@ -38,9 +38,8 @@ void free_cpu_buffers(void)
 {
 	int i;
  
-	for_each_online_cpu(i) {
+	for_each_online_cpu(i)
 		vfree(cpu_buffer[i].buffer);
-	}
 }
 
 int alloc_cpu_buffers(void)
diff -puN fs/xfs/linux-2.6/xfs_stats.c~more-for_each_cpu-conversions fs/xfs/linux-2.6/xfs_stats.c
--- devel/fs/xfs/linux-2.6/xfs_stats.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/fs/xfs/linux-2.6/xfs_stats.c	2006-02-03 01:00:28.000000000 -0800
@@ -62,18 +62,15 @@ xfs_read_xfsstats(
 		while (j < xstats[i].endpoint) {
 			val = 0;
 			/* sum over all cpus */
-			for (c = 0; c < NR_CPUS; c++) {
-				if (!cpu_possible(c)) continue;
+			for_each_cpu(c)
 				val += *(((__u32*)&per_cpu(xfsstats, c) + j));
-			}
 			len += sprintf(buffer + len, " %u", val);
 			j++;
 		}
 		buffer[len++] = '\n';
 	}
 	/* extra precision counters */
-	for (i = 0; i < NR_CPUS; i++) {
-		if (!cpu_possible(i)) continue;
+	for_each_cpu(i) {
 		xs_xstrat_bytes += per_cpu(xfsstats, i).xs_xstrat_bytes;
 		xs_write_bytes += per_cpu(xfsstats, i).xs_write_bytes;
 		xs_read_bytes += per_cpu(xfsstats, i).xs_read_bytes;
diff -puN fs/xfs/linux-2.6/xfs_sysctl.c~more-for_each_cpu-conversions fs/xfs/linux-2.6/xfs_sysctl.c
--- devel/fs/xfs/linux-2.6/xfs_sysctl.c~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/fs/xfs/linux-2.6/xfs_sysctl.c	2006-02-03 01:00:28.000000000 -0800
@@ -38,8 +38,7 @@ xfs_stats_clear_proc_handler(
 
 	if (!ret && write && *valp) {
 		printk("XFS Clearing xfsstats\n");
-		for (c = 0; c < NR_CPUS; c++) {
-			if (!cpu_possible(c)) continue;
+		for_each_cpu(c) {
 			preempt_disable();
 			/* save vn_active, it's a universal truth! */
 			vn_active = per_cpu(xfsstats, c).vn_active;
diff -puN include/asm-alpha/mmu_context.h~more-for_each_cpu-conversions include/asm-alpha/mmu_context.h
--- devel/include/asm-alpha/mmu_context.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/asm-alpha/mmu_context.h	2006-02-03 01:00:28.000000000 -0800
@@ -231,9 +231,8 @@ init_new_context(struct task_struct *tsk
 {
 	int i;
 
-	for (i = 0; i < NR_CPUS; i++)
-		if (cpu_online(i))
-			mm->context[i] = 0;
+	for_each_online_cpu(i)
+		mm->context[i] = 0;
 	if (tsk != current)
 		task_thread_info(tsk)->pcb.ptbr
 		  = ((unsigned long)mm->pgd - IDENT_ADDR) >> PAGE_SHIFT;
diff -puN include/asm-alpha/topology.h~more-for_each_cpu-conversions include/asm-alpha/topology.h
--- devel/include/asm-alpha/topology.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/asm-alpha/topology.h	2006-02-03 01:00:28.000000000 -0800
@@ -27,8 +27,8 @@ static inline cpumask_t node_to_cpumask(
 	cpumask_t node_cpu_mask = CPU_MASK_NONE;
 	int cpu;
 
-	for(cpu = 0; cpu < NR_CPUS; cpu++) {
-		if (cpu_online(cpu) && (cpu_to_node(cpu) == node))
+	for_each_online_cpu(cpu) {
+		if (cpu_to_node(cpu) == node)
 			cpu_set(cpu, node_cpu_mask);
 	}
 
diff -puN include/asm-generic/percpu.h~more-for_each_cpu-conversions include/asm-generic/percpu.h
--- devel/include/asm-generic/percpu.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/asm-generic/percpu.h	2006-02-03 01:00:28.000000000 -0800
@@ -19,10 +19,9 @@ extern unsigned long __per_cpu_offset[NR
 #define percpu_modcopy(pcpudst, src, size)			\
 do {								\
 	unsigned int __i;					\
-	for (__i = 0; __i < NR_CPUS; __i++)			\
-		if (cpu_possible(__i))				\
-			memcpy((pcpudst)+__per_cpu_offset[__i],	\
-			       (src), (size));			\
+	for_each_cpu(__i)					\
+		memcpy((pcpudst)+__per_cpu_offset[__i],		\
+		       (src), (size));				\
 } while (0)
 #else /* ! SMP */
 
diff -puN include/asm-powerpc/percpu.h~more-for_each_cpu-conversions include/asm-powerpc/percpu.h
--- devel/include/asm-powerpc/percpu.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/asm-powerpc/percpu.h	2006-02-03 01:00:28.000000000 -0800
@@ -27,10 +27,9 @@
 #define percpu_modcopy(pcpudst, src, size)			\
 do {								\
 	unsigned int __i;					\
-	for (__i = 0; __i < NR_CPUS; __i++)			\
-		if (cpu_possible(__i))				\
-			memcpy((pcpudst)+__per_cpu_offset(__i),	\
-			       (src), (size));			\
+	for_each_cpu(__i)					\
+		memcpy((pcpudst)+__per_cpu_offset(__i),		\
+		       (src), (size));				\
 } while (0)
 
 extern void setup_per_cpu_areas(void);
diff -puN include/asm-s390/percpu.h~more-for_each_cpu-conversions include/asm-s390/percpu.h
--- devel/include/asm-s390/percpu.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/asm-s390/percpu.h	2006-02-03 01:00:28.000000000 -0800
@@ -46,10 +46,9 @@ extern unsigned long __per_cpu_offset[NR
 #define percpu_modcopy(pcpudst, src, size)			\
 do {								\
 	unsigned int __i;					\
-	for (__i = 0; __i < NR_CPUS; __i++)			\
-		if (cpu_possible(__i))				\
-			memcpy((pcpudst)+__per_cpu_offset[__i],	\
-			       (src), (size));			\
+	for_each_cpu(__i)					\
+		memcpy((pcpudst)+__per_cpu_offset[__i],		\
+		       (src), (size));				\
 } while (0)
 
 #else /* ! SMP */
diff -puN include/asm-sparc64/percpu.h~more-for_each_cpu-conversions include/asm-sparc64/percpu.h
--- devel/include/asm-sparc64/percpu.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/asm-sparc64/percpu.h	2006-02-03 01:00:28.000000000 -0800
@@ -26,10 +26,9 @@ register unsigned long __local_per_cpu_o
 #define percpu_modcopy(pcpudst, src, size)			\
 do {								\
 	unsigned int __i;					\
-	for (__i = 0; __i < NR_CPUS; __i++)			\
-		if (cpu_possible(__i))				\
-			memcpy((pcpudst)+__per_cpu_offset(__i),	\
-			       (src), (size));			\
+	for_each_cpu(__i)					\
+		memcpy((pcpudst)+__per_cpu_offset(__i),		\
+		       (src), (size));				\
 } while (0)
 #else /* ! SMP */
 
diff -puN include/asm-x86_64/percpu.h~more-for_each_cpu-conversions include/asm-x86_64/percpu.h
--- devel/include/asm-x86_64/percpu.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/asm-x86_64/percpu.h	2006-02-03 01:00:28.000000000 -0800
@@ -26,10 +26,9 @@
 #define percpu_modcopy(pcpudst, src, size)			\
 do {								\
 	unsigned int __i;					\
-	for (__i = 0; __i < NR_CPUS; __i++)			\
-		if (cpu_possible(__i))				\
-			memcpy((pcpudst)+__per_cpu_offset(__i),	\
-			       (src), (size));			\
+	for_each_cpu(__i)					\
+		memcpy((pcpudst)+__per_cpu_offset(__i),		\
+		       (src), (size));				\
 } while (0)
 
 extern void setup_per_cpu_areas(void);
diff -puN include/linux/genhd.h~more-for_each_cpu-conversions include/linux/genhd.h
--- devel/include/linux/genhd.h~more-for_each_cpu-conversions	2006-02-03 01:00:28.000000000 -0800
+++ devel-akpm/include/linux/genhd.h	2006-02-03 01:00:28.000000000 -0800
@@ -149,22 +149,16 @@ struct disk_attribute {
 ({									\
 	typeof(gendiskp->dkstats->field) res = 0;			\
 	int i;								\
-	for (i=0; i < NR_CPUS; i++) {					\
-		if (!cpu_possible(i))					\
-			continue;					\
+	for_each_cpu(i)							\
 		res += per_cpu_ptr(gendiskp->dkstats, i)->field;	\
-	}								\
 	res;								\
 })
 
 static inline void disk_stat_set_all(struct gendisk *gendiskp, int value)	{
 	int i;
-	for (i=0; i < NR_CPUS; i++) {
-		if (cpu_possible(i)) {
-			memset(per_cpu_ptr(gendiskp->dkstats, i), value,	
-					sizeof (struct disk_stats));
-		}
-	}
+	for_each_cpu(i)
+		memset(per_cpu_ptr(gendiskp->dkstats, i), value,
+				sizeof (struct disk_stats));
 }		
 				
 #else
_

Re: percpu data changes

[Andrew Morton]

Andrew Morton <akpm@osdl.org> wrote:
>
> I have a patch queued up (for 2.6.16) which fixes most users of percpu
>  data.  Make them use for_each_cpu().  I'll send that in a moment.


From: Eric Dumazet <dada1@cosmosbay.com>

percpu_data blindly allocates bootmem memory to store NR_CPUS instances of
cpudata, instead of allocating memory only for possible cpus.

As a preparation for changing that, we need to convert various 0 -> NR_CPUS
loops to use for_each_cpu().

(The above only applies to users of asm-generic/percpu.h.  powerpc has gone it
alone and is presently only allocating memory for present CPUs, so t's
currently corrupting memory).

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: Jens Axboe <axboe@suse.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Jens Axboe <axboe@suse.de>
Cc: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

 arch/i386/kernel/nmi.c |    2 +-
 block/ll_rw_blk.c      |    2 +-
 drivers/scsi/scsi.c    |    2 +-
 fs/file.c              |    3 +--
 kernel/sched.c         |    2 +-
 net/core/utils.c       |    2 +-
 net/ipv4/proc.c        |    2 +-
 net/ipv6/proc.c        |    2 +-
 net/socket.c           |    2 +-
 main.c                 |    0 
 10 files changed, 9 insertions(+), 10 deletions(-)

diff -puN arch/i386/kernel/nmi.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject arch/i386/kernel/nmi.c
--- devel/arch/i386/kernel/nmi.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/arch/i386/kernel/nmi.c	2006-02-02 23:53:35.000000000 -0800
@@ -138,7 +138,7 @@ static int __init check_nmi_watchdog(voi
 	if (nmi_watchdog == NMI_LOCAL_APIC)
 		smp_call_function(nmi_cpu_busy, (void *)&endflag, 0, 0);
 
-	for (cpu = 0; cpu < NR_CPUS; cpu++)
+	for_each_cpu(cpu)
 		prev_nmi_count[cpu] = per_cpu(irq_stat, cpu).__nmi_count;
 	local_irq_enable();
 	mdelay((10*1000)/nmi_hz); // wait 10 ticks
diff -puN block/ll_rw_blk.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject block/ll_rw_blk.c
--- devel/block/ll_rw_blk.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/block/ll_rw_blk.c	2006-02-02 23:53:34.000000000 -0800
@@ -3453,7 +3453,7 @@ int __init blk_dev_init(void)
 	iocontext_cachep = kmem_cache_create("blkdev_ioc",
 			sizeof(struct io_context), 0, SLAB_PANIC, NULL, NULL);
 
-	for (i = 0; i < NR_CPUS; i++)
+	for_each_cpu(i)
 		INIT_LIST_HEAD(&per_cpu(blk_cpu_done, i));
 
 	open_softirq(BLOCK_SOFTIRQ, blk_done_softirq, NULL);
diff -puN drivers/scsi/scsi.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject drivers/scsi/scsi.c
--- devel/drivers/scsi/scsi.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/drivers/scsi/scsi.c	2006-02-02 23:53:34.000000000 -0800
@@ -1245,7 +1245,7 @@ static int __init init_scsi(void)
 	if (error)
 		goto cleanup_sysctl;
 
-	for (i = 0; i < NR_CPUS; i++)
+	for_each_cpu(i)
 		INIT_LIST_HEAD(&per_cpu(scsi_done_q, i));
 
 	devfs_mk_dir("scsi");
diff -puN fs/file.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject fs/file.c
--- devel/fs/file.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/fs/file.c	2006-02-02 23:53:35.000000000 -0800
@@ -379,7 +379,6 @@ static void __devinit fdtable_defer_list
 void __init files_defer_init(void)
 {
 	int i;
-	/* Really early - can't use for_each_cpu */
-	for (i = 0; i < NR_CPUS; i++)
+	for_each_cpu(i)
 		fdtable_defer_list_init(i);
 }
diff -puN init/main.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject init/main.c
diff -puN kernel/sched.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject kernel/sched.c
--- devel/kernel/sched.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/kernel/sched.c	2006-02-02 23:53:35.000000000 -0800
@@ -6109,7 +6109,7 @@ void __init sched_init(void)
 	runqueue_t *rq;
 	int i, j, k;
 
-	for (i = 0; i < NR_CPUS; i++) {
+	for_each_cpu(i) {
 		prio_array_t *array;
 
 		rq = cpu_rq(i);
diff -puN net/core/utils.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject net/core/utils.c
--- devel/net/core/utils.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/net/core/utils.c	2006-02-02 23:53:34.000000000 -0800
@@ -121,7 +121,7 @@ void __init net_random_init(void)
 {
 	int i;
 
-	for (i = 0; i < NR_CPUS; i++) {
+	for_each_cpu(i) {
 		struct nrnd_state *state = &per_cpu(net_rand_state,i);
 		__net_srandom(state, i+jiffies);
 	}
diff -puN net/ipv4/proc.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject net/ipv4/proc.c
--- devel/net/ipv4/proc.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/net/ipv4/proc.c	2006-02-02 23:53:34.000000000 -0800
@@ -49,7 +49,7 @@ static int fold_prot_inuse(struct proto 
 	int res = 0;
 	int cpu;
 
-	for (cpu = 0; cpu < NR_CPUS; cpu++)
+	for_each_cpu(cpu)
 		res += proto->stats[cpu].inuse;
 
 	return res;
diff -puN net/ipv6/proc.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject net/ipv6/proc.c
--- devel/net/ipv6/proc.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/net/ipv6/proc.c	2006-02-02 23:53:34.000000000 -0800
@@ -38,7 +38,7 @@ static int fold_prot_inuse(struct proto 
 	int res = 0;
 	int cpu;
 
-	for (cpu=0; cpu<NR_CPUS; cpu++)
+	for_each_cpu(cpu)
 		res += proto->stats[cpu].inuse;
 
 	return res;
diff -puN net/socket.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject net/socket.c
--- devel/net/socket.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject	2006-02-02 23:53:34.000000000 -0800
+++ devel-akpm/net/socket.c	2006-02-02 23:53:34.000000000 -0800
@@ -2078,7 +2078,7 @@ void socket_seq_show(struct seq_file *se
 	int cpu;
 	int counter = 0;
 
-	for (cpu = 0; cpu < NR_CPUS; cpu++)
+	for_each_cpu(cpu)
 		counter += per_cpu(sockets_in_use, cpu);
 
 	/* It can be negative, by the way. 8) */
_

Re: percpu data changes

[Andrew Morton]

Andrew Morton <akpm@osdl.org> wrote:
>
> Andrew Morton <akpm@osdl.org> wrote:
>  >
>  > I have a patch queued up (for 2.6.16) which fixes most users of percpu
>  >  data.  Make them use for_each_cpu().  I'll send that in a moment.

A couple more.


 net/core/dev.c   |    2 +-
 net/core/utils.c |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff -puN net/core/dev.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject-update net/core/dev.c
--- devel/net/core/dev.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject-update	2006-02-03 01:33:56.000000000 -0800
+++ devel-akpm/net/core/dev.c	2006-02-03 01:34:16.000000000 -0800
@@ -3237,7 +3237,7 @@ static int __init net_dev_init(void)
 	 *	Initialise the packet receive queues.
 	 */
 
-	for (i = 0; i < NR_CPUS; i++) {
+	for_each_cpu(i) {
 		struct softnet_data *queue;
 
 		queue = &per_cpu(softnet_data, i);
diff -puN net/core/utils.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject-update net/core/utils.c
--- devel/net/core/utils.c~reduce-size-of-percpudata-and-make-sure-per_cpuobject-update	2006-02-03 01:33:56.000000000 -0800
+++ devel-akpm/net/core/utils.c	2006-02-03 01:34:31.000000000 -0800
@@ -133,7 +133,7 @@ static int net_random_reseed(void)
 	unsigned long seed[NR_CPUS];
 
 	get_random_bytes(seed, sizeof(seed));
-	for (i = 0; i < NR_CPUS; i++) {
+	for_each_cpu(i) {
 		struct nrnd_state *state = &per_cpu(net_rand_state,i);
 		__net_srandom(state, seed[i]);
 	}
_

Re: percpu data changes

[William Lee Irwin III]

On Fri, Feb 03, 2006 at 01:17:48AM -0800, Andrew Morton wrote:
> It seems that powerpc has gone and changed their implementation of percpu
> data so that there is no memory allocated for not-possible CPUs.  To save a
> bit of RAM.
> This means that any code which does
> 
> 	for (i = 0; i < NR_CPUS; i++)
> 		touch(percpudata(i))
> will explode on powerpc.
> I have a patch queued up (for 2.6.16) which fixes most users of percpu
> data.  Make them use for_each_cpu().  I'll send that in a moment.
...

Blanket Acked-by: William Irwin <wli@holomorphy.com>


-- wli

Re: percpu data changes

[David S. Miller]

From: Andrew Morton <akpm@osdl.org>
Date: Fri, 3 Feb 2006 01:35:21 -0800

> A couple more.
> 
>  net/core/dev.c   |    2 +-
>  net/core/utils.c |    2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)

Looks good to me, please merge this in Andrew.

Re: percpu data changes

[David S. Miller]

From: Andrew Morton <akpm@osdl.org>
Date: Fri, 3 Feb 2006 01:17:48 -0800

> So please review the below post-2.6.16 patch with an eye to the following:
> 
> - Any stuff you don't want added to your arch
> 
> - Any code which _does_ affect percpu data and which might possibly be
>   used on powerpc.  Such changes will need to go into 2.6.16.
> 
> Thanks.

Sparc64 bits look fine to me.

Re: percpu data changes

[Andi Kleen]

On Friday 03 February 2006 10:17, Andrew Morton wrote:
> 
> It seems that powerpc has gone and changed their implementation of percpu
> data so that there is no memory allocated for not-possible CPUs.  To save a
> bit of RAM.
> 
> This means that any code which does
> 
> 	for (i = 0; i < NR_CPUS; i++)
> 		touch(percpudata(i))
> 
> will explode on powerpc.

It also explodes since some time on x86-64.

But I added a workaround now (or rather sent one and Linus dropped it)
to point the not possible CPUs to the reference data and not free it.
With that violating that protocol is mostly harmless.

Later the plan was to point it to unmapped data to catch all users.

-Andi

Re: percpu data changes

[Eric Dumazet]

[-- Attachment #1: Type: text/plain, Size: 898 bytes --]

Andi Kleen a écrit :
> On Friday 03 February 2006 10:17, Andrew Morton wrote:
>> It seems that powerpc has gone and changed their implementation of percpu
>> data so that there is no memory allocated for not-possible CPUs.  To save a
>> bit of RAM.
>>
>> This means that any code which does
>>
>> 	for (i = 0; i < NR_CPUS; i++)
>> 		touch(percpudata(i))
>>
>> will explode on powerpc.
> 
> It also explodes since some time on x86-64.
> 
> But I added a workaround now (or rather sent one and Linus dropped it)
> to point the not possible CPUs to the reference data and not free it.
> With that violating that protocol is mostly harmless.
> 
> Later the plan was to point it to unmapped data to catch all users.

Maybe you can port the following i386 patch to x86_64 ? (I dont have a x86_64 
test machine at this moment)

(This was sent to Andrew Jan 29th, and is included in 2.6.16-rc1-mm5)

Eric


[-- Attachment #2: i386_mm_init.patch --]
[-- Type: text/plain, Size: 1511 bytes --]

--- a/arch/i386/Kconfig.debug	2006-01-29 22:30:10.000000000 +0100
+++ b/arch/i386/Kconfig.debug	2006-01-29 22:35:54.000000000 +0100
@@ -61,6 +61,18 @@
 	  portion of the kernel code won't be covered by a 2MB TLB anymore.
 	  If in doubt, say "N".
 
+config DEBUG_INITDATA
+	bool "Read/Write protect kernel init data structures"
+	depends on DEBUG_KERNEL
+	help
+	  The init data is normally freed when kernel has booted.
+	  Some code may still try to read or write to data in this area.
+	  If you say Y here, the kernel will mark this zone as not readable
+	  or writeable at all. Buggy code will then fault.
+	  This option may have a slight performance impact because a
+	  portion of the kernel code won't be covered by a 2MB TLB anymore.
+	  If in doubt, say "N".
+
 config 4KSTACKS
 	bool "Use 4Kb + 4Kb for kernel stacks instead of 8Kb" if DEBUG_KERNEL
 	default y
--- a/arch/i386/mm/init.c	2006-01-25 10:17:24.000000000 +0100
+++ b/arch/i386/mm/init.c	2006-01-29 22:38:53.000000000 +0100
@@ -750,11 +750,18 @@
 	for (addr = begin; addr < end; addr += PAGE_SIZE) {
 		ClearPageReserved(virt_to_page(addr));
 		set_page_count(virt_to_page(addr), 1);
+#ifdef CONFIG_DEBUG_INITDATA
+		change_page_attr(virt_to_page(addr), 1, __pgprot(0));
+#else
 		memset((void *)addr, 0xcc, PAGE_SIZE);
+#endif
 		free_page(addr);
 		totalram_pages++;
 	}
 	printk(KERN_INFO "Freeing %s: %ldk freed\n", what, (end - begin) >> 10);
+#ifdef CONFIG_DEBUG_INITDATA
+	global_flush_tlb();
+#endif
 }
 
 void free_initmem(void)

Re: percpu data changes

[Andi Kleen]

On Friday 03 February 2006 12:01, Eric Dumazet wrote:
> Andi Kleen a écrit :
> > On Friday 03 February 2006 10:17, Andrew Morton wrote:
> >> It seems that powerpc has gone and changed their implementation of percpu
> >> data so that there is no memory allocated for not-possible CPUs.  To save a
> >> bit of RAM.
> >>
> >> This means that any code which does
> >>
> >> 	for (i = 0; i < NR_CPUS; i++)
> >> 		touch(percpudata(i))
> >>
> >> will explode on powerpc.
> > 
> > It also explodes since some time on x86-64.
> > 
> > But I added a workaround now (or rather sent one and Linus dropped it)
> > to point the not possible CPUs to the reference data and not free it.
> > With that violating that protocol is mostly harmless.
> > 
> > Later the plan was to point it to unmapped data to catch all users.
> 
> Maybe you can port the following i386 patch to x86_64 ? (I dont have a x86_64 
> test machine at this moment)

Yes good idea. I will add it after 2.6.16.

-Andi

Re: percpu data changes

[Kyle McMartin]

On Fri, Feb 03, 2006 at 01:17:48AM -0800, Andrew Morton wrote:
> Cc: Kyle McMartin <kyle@parisc-linux.org>
>

Looks good.

Acked-by: Kyle McMartin <kyle@parisc-linux.org>

Cheers!
	Kyle