[PATCH 0/8] NR_CPUS: Various replacements for NR_CPUS

[PATCH 0/8] NR_CPUS: Various replacements for NR_CPUS

[Mike Travis]

I've found one old patch that seems to have been misplaced and
a few minor new instances where usage of NR_CPUS may be incorrect.

Also a fixup for CPUMASK_ALLOC is included.

Signed-off-by: Mike Travis <travis@sgi.com>
---

-- 

[PATCH 1/8] NR_CPUS: Replace NR_CPUS in arch/x86/kernel/cpu/mcheck/mce_64.c

[Mike Travis]

[-- Attachment #1: repl-NR_CPUS-in-arch_x86_kernel_cpu_mcheck_mce_64_c --]
[-- Type: text/plain, Size: 705 bytes --]

  * nr_cpu_ids should be used to allocate arrays based on the number of
    cpu's present.

Applies to both tip/master and linux-next.

Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Cc: Eric W. Biederman <ebiederm@xmission.com>
---
 arch/x86/kernel/cpu/mcheck/mce_64.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- linux-2.6.tip.orig/arch/x86/kernel/cpu/mcheck/mce_64.c
+++ linux-2.6.tip/arch/x86/kernel/cpu/mcheck/mce_64.c
@@ -580,7 +580,7 @@ static ssize_t mce_read(struct file *fil
 	char __user *buf = ubuf;
 	int i, err;
 
-	cpu_tsc = kmalloc(NR_CPUS * sizeof(long), GFP_KERNEL);
+	cpu_tsc = kmalloc(nr_cpu_ids * sizeof(long), GFP_KERNEL);
 	if (!cpu_tsc)
 		return -ENOMEM;
 

-- 

[PATCH 2/8] NR_CPUS: Replace NR_CPUS in arch/x86/kernel/cpu/proc.c

[Mike Travis]

[-- Attachment #1: repl-NR_CPUS-in-arch_x86_kernel_cpu_proc_c --]
[-- Type: text/plain, Size: 739 bytes --]

  * Use nr_cpu_ids instead of NR_CPUS to limit traversal of cpu online map.

Applies to both tip/master and linux-next.

Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
---
 arch/x86/kernel/cpu/proc.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- linux-2.6.tip.orig/arch/x86/kernel/cpu/proc.c
+++ linux-2.6.tip/arch/x86/kernel/cpu/proc.c
@@ -160,7 +160,7 @@ static void *c_start(struct seq_file *m,
 {
 	if (*pos == 0)	/* just in case, cpu 0 is not the first */
 		*pos = first_cpu(cpu_online_map);
-	if ((*pos) < NR_CPUS && cpu_online(*pos))
+	if ((*pos) < nr_cpu_ids && cpu_online(*pos))
 		return &cpu_data(*pos);
 	return NULL;
 }

-- 

[PATCH 3/8] NR_CPUS: Replace NR_CPUS in arch/x86/kernel/genx2apic_uv_x.c

[Mike Travis]

[-- Attachment #1: repl-NR_CPUS-in-arch_x86_kernel_genx2apic_uv_x_c --]
[-- Type: text/plain, Size: 966 bytes --]

  * Replace NR_CPUS loop with for_each_possible_cpu().

  * nr_cpu_ids should be used to determine if a percpu area is
    available for a given cpu.

Applies to both tip/master and linux-next.

Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Jack Steiner <steiner@sgi.com>
---
 arch/x86/kernel/genx2apic_uv_x.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

--- linux-2.6.tip.orig/arch/x86/kernel/genx2apic_uv_x.c
+++ linux-2.6.tip/arch/x86/kernel/genx2apic_uv_x.c
@@ -98,7 +98,7 @@ static void uv_send_IPI_mask(cpumask_t m
 {
 	unsigned int cpu;
 
-	for (cpu = 0; cpu < NR_CPUS; ++cpu)
+	for_each_possible_cpu(cpu)
 		if (cpu_isset(cpu, mask))
 			uv_send_IPI_one(cpu, vector);
 }
@@ -132,7 +132,7 @@ static unsigned int uv_cpu_mask_to_apici
 	 * May as well be the first.
 	 */
 	cpu = first_cpu(cpumask);
-	if ((unsigned)cpu < NR_CPUS)
+	if ((unsigned)cpu < nr_cpu_ids)
 		return per_cpu(x86_cpu_to_apicid, cpu);
 	else
 		return BAD_APICID;

-- 

[PATCH 4/8] NR_CPUS: Replace NR_CPUS in arch/x86/kernel/genapic_flat_64.c

[Mike Travis]

[-- Attachment #1: repl-NR_CPUS-in-arch_x86_kernel_genapic_flat_64_c --]
[-- Type: text/plain, Size: 694 bytes --]

  * nr_cpu_ids should be used to determine if a percpu area is
    available for a given cpu.

Applies to both tip/master and linux-next.

Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Andi Kleen <andi@firstfloor.org>
---
 arch/x86/kernel/genapic_flat_64.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- linux-2.6.tip.orig/arch/x86/kernel/genapic_flat_64.c
+++ linux-2.6.tip/arch/x86/kernel/genapic_flat_64.c
@@ -168,7 +168,7 @@ static unsigned int physflat_cpu_mask_to
 	 * May as well be the first.
 	 */
 	cpu = first_cpu(cpumask);
-	if ((unsigned)cpu < NR_CPUS)
+	if ((unsigned)cpu < nr_cpu_ids)
 		return per_cpu(x86_cpu_to_apicid, cpu);
 	else
 		return BAD_APICID;

-- 

[PATCH 5/8] NR_CPUS: Replace per_cpu(..., smp_processor_id()) with __get_cpu_var

[Mike Travis]

[-- Attachment #1: repl-NR_CPUS-in-include_asm-x86_processor_h --]
[-- Type: text/plain, Size: 716 bytes --]

  * Slight optimization when getting one's own cpu_info percpu data.

Applies to both tip/master and linux-next.

Signed-off-by: Mike Travis <travis@sgi.com>
---
 include/asm-x86/processor.h |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- linux-2.6.tip.orig/include/asm-x86/processor.h
+++ linux-2.6.tip/include/asm-x86/processor.h
@@ -135,7 +135,7 @@ extern __u32			cleared_cpu_caps[NCAPINTS
 #ifdef CONFIG_SMP
 DECLARE_PER_CPU(struct cpuinfo_x86, cpu_info);
 #define cpu_data(cpu)		per_cpu(cpu_info, cpu)
-#define current_cpu_data	cpu_data(smp_processor_id())
+#define current_cpu_data	__get_cpu_var(cpu_info)
 #else
 #define cpu_data(cpu)		boot_cpu_data
 #define current_cpu_data	boot_cpu_data

-- 

[PATCH 6/8] NR_CPUS: Replace NR_CPUS in cpufreq userspace routines

[Mike Travis]

[-- Attachment #1: repl-NR_CPUS-in-cpufreq_userspace --]
[-- Type: text/plain, Size: 5797 bytes --]

  * This patch seems to have been misplaced.  Replace arrays sized by NR_CPUS
    with percpu variables.

    Prior reference: http://marc.info/?l=linux-kernel&m=120251421825989&w=4
    Subject:    [PATCH 1/4] cpufreq: change cpu freq tables to per_cpu variables
    From:       Mike Travis <travis () sgi ! com>
    Date:       2008-02-08 23:37:39

Applies to both tip/master and linux-next.

Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Dave Jones <davej@codemonkey.org.uk>
Cc: cpufreq@lists.linux.org.uk
---
 drivers/cpufreq/cpufreq_userspace.c |   79 +++++++++++++++++++-----------------
 1 file changed, 43 insertions(+), 36 deletions(-)

--- linux-2.6.tip.orig/drivers/cpufreq/cpufreq_userspace.c
+++ linux-2.6.tip/drivers/cpufreq/cpufreq_userspace.c
@@ -30,16 +30,18 @@
 /**
  * A few values needed by the userspace governor
  */
-static unsigned int	cpu_max_freq[NR_CPUS];
-static unsigned int	cpu_min_freq[NR_CPUS];
-static unsigned int	cpu_cur_freq[NR_CPUS]; /* current CPU freq */
-static unsigned int	cpu_set_freq[NR_CPUS]; /* CPU freq desired by userspace */
-static unsigned int	cpu_is_managed[NR_CPUS];
+static DEFINE_PER_CPU(unsigned int, cpu_max_freq);
+static DEFINE_PER_CPU(unsigned int, cpu_min_freq);
+static DEFINE_PER_CPU(unsigned int, cpu_cur_freq); /* current CPU freq */
+static DEFINE_PER_CPU(unsigned int, cpu_set_freq); /* CPU freq desired by
+							userspace */
+static DEFINE_PER_CPU(unsigned int, cpu_is_managed);
 
 static DEFINE_MUTEX	(userspace_mutex);
 static int cpus_using_userspace_governor;
 
-#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_GOVERNOR, "userspace", msg)
+#define dprintk(msg...) \
+	cpufreq_debug_printk(CPUFREQ_DEBUG_GOVERNOR, "userspace", msg)
 
 /* keep track of frequency transitions */
 static int
@@ -48,12 +50,12 @@ userspace_cpufreq_notifier(struct notifi
 {
         struct cpufreq_freqs *freq = data;
 
-	if (!cpu_is_managed[freq->cpu])
+	if (!per_cpu(cpu_is_managed, freq->cpu))
 		return 0;
 
 	dprintk("saving cpu_cur_freq of cpu %u to be %u kHz\n",
 			freq->cpu, freq->new);
-	cpu_cur_freq[freq->cpu] = freq->new;
+	per_cpu(cpu_cur_freq, freq->cpu) = freq->new;
 
         return 0;
 }
@@ -77,15 +79,15 @@ static int cpufreq_set(struct cpufreq_po
 	dprintk("cpufreq_set for cpu %u, freq %u kHz\n", policy->cpu, freq);
 
 	mutex_lock(&userspace_mutex);
-	if (!cpu_is_managed[policy->cpu])
+	if (!per_cpu(cpu_is_managed, policy->cpu))
 		goto err;
 
-	cpu_set_freq[policy->cpu] = freq;
+	per_cpu(cpu_set_freq, policy->cpu) = freq;
 
-	if (freq < cpu_min_freq[policy->cpu])
-		freq = cpu_min_freq[policy->cpu];
-	if (freq > cpu_max_freq[policy->cpu])
-		freq = cpu_max_freq[policy->cpu];
+	if (freq < per_cpu(cpu_min_freq, policy->cpu))
+		freq = per_cpu(cpu_min_freq, policy->cpu);
+	if (freq > per_cpu(cpu_max_freq, policy->cpu))
+		freq = per_cpu(cpu_max_freq, policy->cpu);
 
 	/*
 	 * We're safe from concurrent calls to ->target() here
@@ -104,7 +106,7 @@ static int cpufreq_set(struct cpufreq_po
 
 static ssize_t show_speed(struct cpufreq_policy *policy, char *buf)
 {
-	return sprintf(buf, "%u\n", cpu_cur_freq[policy->cpu]);
+	return sprintf(buf, "%u\n", per_cpu(cpu_cur_freq, policy->cpu));
 }
 
 static int cpufreq_governor_userspace(struct cpufreq_policy *policy,
@@ -127,12 +129,17 @@ static int cpufreq_governor_userspace(st
 		}
 		cpus_using_userspace_governor++;
 
-		cpu_is_managed[cpu] = 1;
-		cpu_min_freq[cpu] = policy->min;
-		cpu_max_freq[cpu] = policy->max;
-		cpu_cur_freq[cpu] = policy->cur;
-		cpu_set_freq[cpu] = policy->cur;
-		dprintk("managing cpu %u started (%u - %u kHz, currently %u kHz)\n", cpu, cpu_min_freq[cpu], cpu_max_freq[cpu], cpu_cur_freq[cpu]);
+		per_cpu(cpu_is_managed, cpu) = 1;
+		per_cpu(cpu_min_freq, cpu) = policy->min;
+		per_cpu(cpu_max_freq, cpu) = policy->max;
+		per_cpu(cpu_cur_freq, cpu) = policy->cur;
+		per_cpu(cpu_set_freq, cpu) = policy->cur;
+		dprintk("managing cpu %u started "
+			"(%u - %u kHz, currently %u kHz)\n",
+				cpu,
+				per_cpu(cpu_min_freq, cpu),
+				per_cpu(cpu_max_freq, cpu),
+				per_cpu(cpu_cur_freq, cpu));
 
 		mutex_unlock(&userspace_mutex);
 		break;
@@ -145,34 +152,34 @@ static int cpufreq_governor_userspace(st
 					CPUFREQ_TRANSITION_NOTIFIER);
 		}
 
-		cpu_is_managed[cpu] = 0;
-		cpu_min_freq[cpu] = 0;
-		cpu_max_freq[cpu] = 0;
-		cpu_set_freq[cpu] = 0;
+		per_cpu(cpu_is_managed, cpu) = 0;
+		per_cpu(cpu_min_freq, cpu) = 0;
+		per_cpu(cpu_max_freq, cpu) = 0;
+		per_cpu(cpu_set_freq, cpu) = 0;
 		dprintk("managing cpu %u stopped\n", cpu);
 		mutex_unlock(&userspace_mutex);
 		break;
 	case CPUFREQ_GOV_LIMITS:
 		mutex_lock(&userspace_mutex);
-		dprintk("limit event for cpu %u: %u - %u kHz,"
+		dprintk("limit event for cpu %u: %u - %u kHz, "
 			"currently %u kHz, last set to %u kHz\n",
 			cpu, policy->min, policy->max,
-			cpu_cur_freq[cpu], cpu_set_freq[cpu]);
-		if (policy->max < cpu_set_freq[cpu]) {
+			per_cpu(cpu_cur_freq, cpu),
+			per_cpu(cpu_set_freq, cpu));
+		if (policy->max < per_cpu(cpu_set_freq, cpu)) {
 			__cpufreq_driver_target(policy, policy->max,
 						CPUFREQ_RELATION_H);
-		}
-		else if (policy->min > cpu_set_freq[cpu]) {
+		} else if (policy->min > per_cpu(cpu_set_freq, cpu)) {
 			__cpufreq_driver_target(policy, policy->min,
 						CPUFREQ_RELATION_L);
-		}
-		else {
-			__cpufreq_driver_target(policy, cpu_set_freq[cpu],
+		} else {
+			__cpufreq_driver_target(policy,
+						per_cpu(cpu_set_freq, cpu),
 						CPUFREQ_RELATION_L);
 		}
-		cpu_min_freq[cpu] = policy->min;
-		cpu_max_freq[cpu] = policy->max;
-		cpu_cur_freq[cpu] = policy->cur;
+		per_cpu(cpu_min_freq, cpu) = policy->min;
+		per_cpu(cpu_max_freq, cpu) = policy->max;
+		per_cpu(cpu_cur_freq, cpu) = policy->cur;
 		mutex_unlock(&userspace_mutex);
 		break;
 	}

-- 

[PATCH 7/8] cpumask: Provide a generic set of CPUMASK_ALLOC macros, FIXUP

[Mike Travis]

[-- Attachment #1: fix-CPUMASK_ALLOC --]
[-- Type: text/plain, Size: 3204 bytes --]

  * Rename CPUMASK_VAR --> CPUMASK_PTR (and simplify)

  * Fix a semantic error in CPUMASK_ALLOC

  * Add a bit of commentry to cpumask.h

Based on linux-2.6.tip/master with the following commit:

    commit 77586c2bdad0798cb24e35de5a878e7c6b200574
    Author: Mike Travis <travis@sgi.com>  2008-07-15 14:14:37
    Committer: Ingo Molnar <mingo@elte.hu>  2008-07-18 13:03:00
    Parent: 77586c2bdad0798cb24e35de5a878e7c6b200574 (cpumask: Provide a generic set of CPUMASK_ALLOC macros)
    Child:  06f8d00e9eecb738c99b737ac38a585ea7583ad5 (cpumask: Optimize cpumask_of_cpu in lib/smp_processor_id.c, fix)
    Child:  32650cb6a8665a7635d025e52d9bf1905981ed11 (Merge branch 'cpus4096')
    Branches: -, devel, remotes/tip/cpus4096, remotes/tip/master, tip-latest

    cpumask: Provide a generic set of CPUMASK_ALLOC macros


Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Bert Wesarg <bert.wesarg@googlemail.com>
---
 include/linux/cpumask.h |   33 ++++++++++++++++++++++++++-------
 1 file changed, 26 insertions(+), 7 deletions(-)

--- linux-2.6.tip.orig/include/linux/cpumask.h
+++ linux-2.6.tip/include/linux/cpumask.h
@@ -75,16 +75,36 @@
  * CPU_MASK_NONE			Initializer - no bits set
  * unsigned long *cpus_addr(mask)	Array of unsigned long's in mask
  *
+ * CPUMASK_ALLOC kmalloc's a structure that is a composite of many cpumask_t
+ * variables, and CPUMASK_PTR provides pointers to each field.
+ *
+ * The structure should be defined something like this:
+ * struct my_cpumasks {
+ *	cpumask_t mask1;
+ *	cpumask_t mask2;
+ * };
+ *
+ * Usage is then:
+ *	CPUMASK_ALLOC(my_cpumasks);
+ *	CPUMASK_PTR(mask1, my_cpumasks);
+ *	CPUMASK_PTR(mask2, my_cpumasks);
+ *
+ *	--- DO NOT reference cpumask_t pointers until this check ---
+ *	if (my_cpumasks == NULL)
+ *		"kmalloc failed"...
+ *
+ * References are now pointers to the cpumask_t variables (*mask1, ...)
+ *
  *if NR_CPUS > BITS_PER_LONG
  *   CPUMASK_ALLOC(m)			Declares and allocates struct m *m =
- *					   (struct m *)kmalloc(sizeof(*m), ...)
- *   CPUMASK_FREE(m)			Macro for kfree(v)
+ *						kmalloc(sizeof(*m), GFP_KERNEL)
+ *   CPUMASK_FREE(m)			Macro for kfree(m)
  *else
  *   CPUMASK_ALLOC(m)			Declares struct m _m, *m = &_m
  *   CPUMASK_FREE(m)			Nop
  *endif
- *   CPUMASK_VAR(v, m)			Declares cpumask_t *v =
- *						m + offset(struct m, v)
+ *   CPUMASK_PTR(v, m)			Declares cpumask_t *v = &(m->v)
+ * ------------------------------------------------------------------------
  *
  * int cpumask_scnprintf(buf, len, mask) Format cpumask for printing
  * int cpumask_parse_user(ubuf, ulen, mask)	Parse ascii string as cpumask
@@ -326,11 +346,10 @@ extern cpumask_t cpu_mask_all;
 #define	CPUMASK_ALLOC(m)	struct m *m = kmalloc(sizeof(*m), GFP_KERNEL)
 #define	CPUMASK_FREE(m)		kfree(m)
 #else
-#define	CPUMASK_ALLOC(m)	struct allmasks _m, *m = &_m
+#define	CPUMASK_ALLOC(m)	struct m _m, *m = &_m
 #define	CPUMASK_FREE(m)
 #endif
-#define	CPUMASK_VAR(v, m) 	cpumask_t *v = (cpumask_t *)		\
-				((unsigned long)(m) + offsetof(struct m, v))
+#define	CPUMASK_PTR(v, m) 	cpumask_t *v = &(m->v)
 
 #define cpumask_scnprintf(buf, len, src) \
 			__cpumask_scnprintf((buf), (len), &(src), NR_CPUS)

-- 

[PATCH 8/8] NR_CPUS: Replace NR_CPUS in speedstep-centrino.c

[Mike Travis]

[-- Attachment #1: repl-NR_CPUS-in-arch_x86_kernel_cpu_cpufreq_speedstep-centrino_c --]
[-- Type: text/plain, Size: 8472 bytes --]

Some cleanups in speedstep-centrino.c for NR_CPUS=4096.

  * Use new CPUMASK_PTR (instead of old CPUMASK_VAR).

  * Replace arrays sized by NR_CPUS with percpu variables.

  * Cleanup some formatting problems (>80 chars per line)
    and other checkpatch complaints.

Applies to tip/master with following commit:

	commit 06f8d00e9eecb738c99b737ac38a585ea7583ad5
	Author: Mike Travis <travis@sgi.com>  2008-07-15 14:14:37
	Committer: Ingo Molnar <mingo@elte.hu>  2008-07-18 13:03:00
	Parent: 77586c2bdad0798cb24e35de5a878e7c6b200574 (cpumask: Provide a generic set of CPUMASK_ALLOC macros)
	Child:  06f8d00e9eecb738c99b737ac38a585ea7583ad5 (cpumask: Optimize cpumask_of_cpu in lib/smp_processor_id.c, fix)
	Child:  32650cb6a8665a7635d025e52d9bf1905981ed11 (Merge branch 'cpus4096')
	Branches: -, devel, remotes/tip/cpus4096, remotes/tip/master, tip-latest

	cpumask: Use optimized CPUMASK_ALLOC macros in the centrino_target


Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Venki Pallipadi <venkatesh.pallipadi@intel.com>
Cc: Dave Jones <davej@codemonkey.org.uk>
Cc: cpufreq@lists.linux.org.uk
---
 arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c |   85 +++++++++++++----------
 1 file changed, 51 insertions(+), 34 deletions(-)

--- linux-2.6.tip.orig/arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c
+++ linux-2.6.tip/arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c
@@ -28,7 +28,8 @@
 #define PFX		"speedstep-centrino: "
 #define MAINTAINER	"cpufreq@lists.linux.org.uk"
 
-#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "speedstep-centrino", msg)
+#define dprintk(msg...) \
+	cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "speedstep-centrino", msg)
 
 #define INTEL_MSR_RANGE	(0xffff)
 
@@ -68,11 +69,12 @@ struct cpu_model
 
 	struct cpufreq_frequency_table *op_points; /* clock/voltage pairs */
 };
-static int centrino_verify_cpu_id(const struct cpuinfo_x86 *c, const struct cpu_id *x);
+static int centrino_verify_cpu_id(const struct cpuinfo_x86 *c,
+				  const struct cpu_id *x);
 
 /* Operating points for current CPU */
-static struct cpu_model *centrino_model[NR_CPUS];
-static const struct cpu_id *centrino_cpu[NR_CPUS];
+static DEFINE_PER_CPU(struct cpu_model *, centrino_model);
+static DEFINE_PER_CPU(const struct cpu_id *, centrino_cpu);
 
 static struct cpufreq_driver centrino_driver;
 
@@ -348,7 +350,7 @@ static int centrino_cpu_init_table(struc
 		return -ENOENT;
 	}
 
-	centrino_model[policy->cpu] = model;
+	per_cpu(centrino_model, policy->cpu) = model;
 
 	dprintk("found \"%s\": max frequency: %dkHz\n",
 	       model->model_name, model->max_freq);
@@ -357,10 +359,14 @@ static int centrino_cpu_init_table(struc
 }
 
 #else
-static inline int centrino_cpu_init_table(struct cpufreq_policy *policy) { return -ENODEV; }
+static inline int centrino_cpu_init_table(struct cpufreq_policy *policy)
+{
+	return -ENODEV;
+}
 #endif /* CONFIG_X86_SPEEDSTEP_CENTRINO_TABLE */
 
-static int centrino_verify_cpu_id(const struct cpuinfo_x86 *c, const struct cpu_id *x)
+static int centrino_verify_cpu_id(const struct cpuinfo_x86 *c,
+				  const struct cpu_id *x)
 {
 	if ((c->x86 == x->x86) &&
 	    (c->x86_model == x->x86_model) &&
@@ -379,23 +385,28 @@ static unsigned extract_clock(unsigned m
 	 * for centrino, as some DSDTs are buggy.
 	 * Ideally, this can be done using the acpi_data structure.
 	 */
-	if ((centrino_cpu[cpu] == &cpu_ids[CPU_BANIAS]) ||
-	    (centrino_cpu[cpu] == &cpu_ids[CPU_DOTHAN_A1]) ||
-	    (centrino_cpu[cpu] == &cpu_ids[CPU_DOTHAN_B0])) {
+	if ((per_cpu(centrino_cpu, cpu) == &cpu_ids[CPU_BANIAS]) ||
+	    (per_cpu(centrino_cpu, cpu) == &cpu_ids[CPU_DOTHAN_A1]) ||
+	    (per_cpu(centrino_cpu, cpu) == &cpu_ids[CPU_DOTHAN_B0])) {
 		msr = (msr >> 8) & 0xff;
 		return msr * 100000;
 	}
 
-	if ((!centrino_model[cpu]) || (!centrino_model[cpu]->op_points))
+	if ((!per_cpu(centrino_model, cpu)) ||
+	    (!per_cpu(centrino_model, cpu)->op_points))
 		return 0;
 
 	msr &= 0xffff;
-	for (i=0;centrino_model[cpu]->op_points[i].frequency != CPUFREQ_TABLE_END; i++) {
-		if (msr == centrino_model[cpu]->op_points[i].index)
-			return centrino_model[cpu]->op_points[i].frequency;
+	for (i = 0;
+		per_cpu(centrino_model, cpu)->op_points[i].frequency
+							!= CPUFREQ_TABLE_END;
+	     i++) {
+		if (msr == per_cpu(centrino_model, cpu)->op_points[i].index)
+			return per_cpu(centrino_model, cpu)->
+							op_points[i].frequency;
 	}
 	if (failsafe)
-		return centrino_model[cpu]->op_points[i-1].frequency;
+		return per_cpu(centrino_model, cpu)->op_points[i-1].frequency;
 	else
 		return 0;
 }
@@ -441,7 +452,8 @@ static int centrino_cpu_init(struct cpuf
 	int i;
 
 	/* Only Intel makes Enhanced Speedstep-capable CPUs */
-	if (cpu->x86_vendor != X86_VENDOR_INTEL || !cpu_has(cpu, X86_FEATURE_EST))
+	if (cpu->x86_vendor != X86_VENDOR_INTEL ||
+	    !cpu_has(cpu, X86_FEATURE_EST))
 		return -ENODEV;
 
 	if (cpu_has(cpu, X86_FEATURE_CONSTANT_TSC))
@@ -455,9 +467,9 @@ static int centrino_cpu_init(struct cpuf
 			break;
 
 	if (i != N_IDS)
-		centrino_cpu[policy->cpu] = &cpu_ids[i];
+		per_cpu(centrino_cpu, policy->cpu) = &cpu_ids[i];
 
-	if (!centrino_cpu[policy->cpu]) {
+	if (!per_cpu(centrino_cpu, policy->cpu)) {
 		dprintk("found unsupported CPU with "
 		"Enhanced SpeedStep: send /proc/cpuinfo to "
 		MAINTAINER "\n");
@@ -480,23 +492,26 @@ static int centrino_cpu_init(struct cpuf
 		/* check to see if it stuck */
 		rdmsr(MSR_IA32_MISC_ENABLE, l, h);
 		if (!(l & (1<<16))) {
-			printk(KERN_INFO PFX "couldn't enable Enhanced SpeedStep\n");
+			printk(KERN_INFO PFX
+				"couldn't enable Enhanced SpeedStep\n");
 			return -ENODEV;
 		}
 	}
 
 	freq = get_cur_freq(policy->cpu);
-
-	policy->cpuinfo.transition_latency = 10000; /* 10uS transition latency */
+	policy->cpuinfo.transition_latency = 10000;
+						/* 10uS transition latency */
 	policy->cur = freq;
 
 	dprintk("centrino_cpu_init: cur=%dkHz\n", policy->cur);
 
-	ret = cpufreq_frequency_table_cpuinfo(policy, centrino_model[policy->cpu]->op_points);
+	ret = cpufreq_frequency_table_cpuinfo(policy,
+		per_cpu(centrino_model, policy->cpu)->op_points);
 	if (ret)
 		return (ret);
 
-	cpufreq_frequency_table_get_attr(centrino_model[policy->cpu]->op_points, policy->cpu);
+	cpufreq_frequency_table_get_attr(
+		per_cpu(centrino_model, policy->cpu)->op_points, policy->cpu);
 
 	return 0;
 }
@@ -505,12 +520,12 @@ static int centrino_cpu_exit(struct cpuf
 {
 	unsigned int cpu = policy->cpu;
 
-	if (!centrino_model[cpu])
+	if (!per_cpu(centrino_model, cpu))
 		return -ENODEV;
 
 	cpufreq_frequency_table_put_attr(cpu);
 
-	centrino_model[cpu] = NULL;
+	per_cpu(centrino_model, cpu) = NULL;
 
 	return 0;
 }
@@ -524,14 +539,16 @@ static int centrino_cpu_exit(struct cpuf
  */
 static int centrino_verify (struct cpufreq_policy *policy)
 {
-	return cpufreq_frequency_table_verify(policy, centrino_model[policy->cpu]->op_points);
+	return cpufreq_frequency_table_verify(policy,
+			per_cpu(centrino_model, policy->cpu)->op_points);
 }
 
 /**
  * centrino_setpolicy - set a new CPUFreq policy
  * @policy: new policy
  * @target_freq: the target frequency
- * @relation: how that frequency relates to achieved frequency (CPUFREQ_RELATION_L or CPUFREQ_RELATION_H)
+ * @relation: how that frequency relates to achieved frequency
+ *	(CPUFREQ_RELATION_L or CPUFREQ_RELATION_H)
  *
  * Sets a new CPUFreq policy.
  */
@@ -552,21 +569,21 @@ static int centrino_target (struct cpufr
 	int			retval = 0;
 	unsigned int		j, k, first_cpu, tmp;
 	CPUMASK_ALLOC(allmasks);
-	CPUMASK_VAR(online_policy_cpus, allmasks);
-	CPUMASK_VAR(saved_mask, allmasks);
-	CPUMASK_VAR(set_mask, allmasks);
-	CPUMASK_VAR(covered_cpus, allmasks);
+	CPUMASK_PTR(online_policy_cpus, allmasks);
+	CPUMASK_PTR(saved_mask, allmasks);
+	CPUMASK_PTR(set_mask, allmasks);
+	CPUMASK_PTR(covered_cpus, allmasks);
 
 	if (unlikely(allmasks == NULL))
 		return -ENOMEM;
 
-	if (unlikely(centrino_model[cpu] == NULL)) {
+	if (unlikely(per_cpu(centrino_model, cpu) == NULL)) {
 		retval = -ENODEV;
 		goto out;
 	}
 
 	if (unlikely(cpufreq_frequency_table_target(policy,
-			centrino_model[cpu]->op_points,
+			per_cpu(centrino_model, cpu)->op_points,
 			target_freq,
 			relation,
 			&newstate))) {
@@ -608,7 +625,7 @@ static int centrino_target (struct cpufr
 			break;
 		}
 
-		msr = centrino_model[cpu]->op_points[newstate].index;
+		msr = per_cpu(centrino_model, cpu)->op_points[newstate].index;
 
 		if (first_cpu) {
 			rdmsr(MSR_IA32_PERF_CTL, oldmsr, h);

-- 

Re: [PATCH 0/8] NR_CPUS: Various replacements for NR_CPUS

[Ingo Molnar]

* Mike Travis <travis@sgi.com> wrote:

> I've found one old patch that seems to have been misplaced [...]

How did you notice that i missed them - actual crashes, or measurable 
performance impact?

> [...] and a few minor new instances where usage of NR_CPUS may be 
> incorrect.
>
> Also a fixup for CPUMASK_ALLOC is included.

applied to tip/cpus4096, thanks Mike.

	Ingo

Re: [PATCH 0/8] NR_CPUS: Various replacements for NR_CPUS

[Mike Travis]

Ingo Molnar wrote:
> * Mike Travis <travis@sgi.com> wrote:
> 
>> I've found one old patch that seems to have been misplaced [...]
> 
> How did you notice that i missed them - actual crashes, or measurable 
> performance impact?

It might not have been you.  I was just doing a survey again of NR_CPUS
usages and noticed something that looked familiar.  (In fact there were
two patches missing but when I updated the source one of them *magically*
appeared... ;-)

I have three more cases that can be changed coming real soon now.

> 
>> [...] and a few minor new instances where usage of NR_CPUS may be 
>> incorrect.
>>
>> Also a fixup for CPUMASK_ALLOC is included.
> 
> applied to tip/cpus4096, thanks Mike.

No, no, thank you!  ;-)

Mike

Re: [PATCH 0/8] NR_CPUS: Various replacements for NR_CPUS

[Mike Travis]

Mike Travis wrote:
...
> I have three more cases that can be changed coming real soon now.

Turns out the changes I was going to make are in linux-next already.

Thanks,
Mike