aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-01-02 14:44:09 -0500
committerLinus Torvalds <torvalds@linux-foundation.org>2009-01-02 14:44:09 -0500
commitb840d79631c882786925303c2b0f4fefc31845ed (patch)
treecda60a95d4507fe1321fc285af38982d7eb9693b /arch/x86/kernel
parent597b0d21626da4e6f09f132442caf0cc2b0eb47c (diff)
parentc3d80000e3a812fe5a200d6bde755fbd7fa65481 (diff)
Merge branch 'cpus4096-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'cpus4096-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (66 commits) x86: export vector_used_by_percpu_irq x86: use logical apicid in x2apic_cluster's x2apic_cpu_mask_to_apicid_and() sched: nominate preferred wakeup cpu, fix x86: fix lguest used_vectors breakage, -v2 x86: fix warning in arch/x86/kernel/io_apic.c sched: fix warning in kernel/sched.c sched: move test_sd_parent() to an SMP section of sched.h sched: add SD_BALANCE_NEWIDLE at MC and CPU level for sched_mc>0 sched: activate active load balancing in new idle cpus sched: bias task wakeups to preferred semi-idle packages sched: nominate preferred wakeup cpu sched: favour lower logical cpu number for sched_mc balance sched: framework for sched_mc/smt_power_savings=N sched: convert BALANCE_FOR_xx_POWER to inline functions x86: use possible_cpus=NUM to extend the possible cpus allowed x86: fix cpu_mask_to_apicid_and to include cpu_online_mask x86: update io_apic.c to the new cpumask code x86: Introduce topology_core_cpumask()/topology_thread_cpumask() x86: xen: use smp_call_function_many() x86: use work_on_cpu in x86/kernel/cpu/mcheck/mce_amd_64.c ... Fixed up trivial conflict in kernel/time/tick-sched.c manually
Diffstat (limited to 'arch/x86/kernel')
-rw-r--r--arch/x86/kernel/apic.c34
-rw-r--r--arch/x86/kernel/cpu/intel_cacheinfo.c45
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_amd_64.c108
-rw-r--r--arch/x86/kernel/genapic_flat_64.c107
-rw-r--r--arch/x86/kernel/genx2apic_cluster.c81
-rw-r--r--arch/x86/kernel/genx2apic_phys.c74
-rw-r--r--arch/x86/kernel/genx2apic_uv_x.c61
-rw-r--r--arch/x86/kernel/hpet.c8
-rw-r--r--arch/x86/kernel/i8253.c2
-rw-r--r--arch/x86/kernel/io_apic.c370
-rw-r--r--arch/x86/kernel/ipi.c28
-rw-r--r--arch/x86/kernel/irq.c3
-rw-r--r--arch/x86/kernel/irq_32.c13
-rw-r--r--arch/x86/kernel/irq_64.c15
-rw-r--r--arch/x86/kernel/irqinit_32.c16
-rw-r--r--arch/x86/kernel/irqinit_64.c13
-rw-r--r--arch/x86/kernel/mfgpt_32.c2
-rw-r--r--arch/x86/kernel/reboot.c5
-rw-r--r--arch/x86/kernel/setup_percpu.c19
-rw-r--r--arch/x86/kernel/smp.c8
-rw-r--r--arch/x86/kernel/smpboot.c33
-rw-r--r--arch/x86/kernel/tlb_32.c2
-rw-r--r--arch/x86/kernel/tlb_64.c2
-rw-r--r--arch/x86/kernel/traps.c12
-rw-r--r--arch/x86/kernel/vmiclock_32.c2
25 files changed, 635 insertions, 428 deletions
diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c
index b5229affb953..6b7f824db160 100644
--- a/arch/x86/kernel/apic.c
+++ b/arch/x86/kernel/apic.c
@@ -119,8 +119,6 @@ EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok);
119 119
120int first_system_vector = 0xfe; 120int first_system_vector = 0xfe;
121 121
122char system_vectors[NR_VECTORS] = { [0 ... NR_VECTORS-1] = SYS_VECTOR_FREE};
123
124/* 122/*
125 * Debug level, exported for io_apic.c 123 * Debug level, exported for io_apic.c
126 */ 124 */
@@ -142,7 +140,7 @@ static int lapic_next_event(unsigned long delta,
142 struct clock_event_device *evt); 140 struct clock_event_device *evt);
143static void lapic_timer_setup(enum clock_event_mode mode, 141static void lapic_timer_setup(enum clock_event_mode mode,
144 struct clock_event_device *evt); 142 struct clock_event_device *evt);
145static void lapic_timer_broadcast(cpumask_t mask); 143static void lapic_timer_broadcast(const cpumask_t *mask);
146static void apic_pm_activate(void); 144static void apic_pm_activate(void);
147 145
148/* 146/*
@@ -455,7 +453,7 @@ static void lapic_timer_setup(enum clock_event_mode mode,
455/* 453/*
456 * Local APIC timer broadcast function 454 * Local APIC timer broadcast function
457 */ 455 */
458static void lapic_timer_broadcast(cpumask_t mask) 456static void lapic_timer_broadcast(const cpumask_t *mask)
459{ 457{
460#ifdef CONFIG_SMP 458#ifdef CONFIG_SMP
461 send_IPI_mask(mask, LOCAL_TIMER_VECTOR); 459 send_IPI_mask(mask, LOCAL_TIMER_VECTOR);
@@ -471,7 +469,7 @@ static void __cpuinit setup_APIC_timer(void)
471 struct clock_event_device *levt = &__get_cpu_var(lapic_events); 469 struct clock_event_device *levt = &__get_cpu_var(lapic_events);
472 470
473 memcpy(levt, &lapic_clockevent, sizeof(*levt)); 471 memcpy(levt, &lapic_clockevent, sizeof(*levt));
474 levt->cpumask = cpumask_of_cpu(smp_processor_id()); 472 levt->cpumask = cpumask_of(smp_processor_id());
475 473
476 clockevents_register_device(levt); 474 clockevents_register_device(levt);
477} 475}
@@ -1807,28 +1805,32 @@ void disconnect_bsp_APIC(int virt_wire_setup)
1807void __cpuinit generic_processor_info(int apicid, int version) 1805void __cpuinit generic_processor_info(int apicid, int version)
1808{ 1806{
1809 int cpu; 1807 int cpu;
1810 cpumask_t tmp_map;
1811 1808
1812 /* 1809 /*
1813 * Validate version 1810 * Validate version
1814 */ 1811 */
1815 if (version == 0x0) { 1812 if (version == 0x0) {
1816 pr_warning("BIOS bug, APIC version is 0 for CPU#%d! " 1813 pr_warning("BIOS bug, APIC version is 0 for CPU#%d! "
1817 "fixing up to 0x10. (tell your hw vendor)\n", 1814 "fixing up to 0x10. (tell your hw vendor)\n",
1818 version); 1815 version);
1819 version = 0x10; 1816 version = 0x10;
1820 } 1817 }
1821 apic_version[apicid] = version; 1818 apic_version[apicid] = version;
1822 1819
1823 if (num_processors >= NR_CPUS) { 1820 if (num_processors >= nr_cpu_ids) {
1824 pr_warning("WARNING: NR_CPUS limit of %i reached." 1821 int max = nr_cpu_ids;
1825 " Processor ignored.\n", NR_CPUS); 1822 int thiscpu = max + disabled_cpus;
1823
1824 pr_warning(
1825 "ACPI: NR_CPUS/possible_cpus limit of %i reached."
1826 " Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
1827
1828 disabled_cpus++;
1826 return; 1829 return;
1827 } 1830 }
1828 1831
1829 num_processors++; 1832 num_processors++;
1830 cpus_complement(tmp_map, cpu_present_map); 1833 cpu = cpumask_next_zero(-1, cpu_present_mask);
1831 cpu = first_cpu(tmp_map);
1832 1834
1833 physid_set(apicid, phys_cpu_present_map); 1835 physid_set(apicid, phys_cpu_present_map);
1834 if (apicid == boot_cpu_physical_apicid) { 1836 if (apicid == boot_cpu_physical_apicid) {
@@ -1878,8 +1880,8 @@ void __cpuinit generic_processor_info(int apicid, int version)
1878 } 1880 }
1879#endif 1881#endif
1880 1882
1881 cpu_set(cpu, cpu_possible_map); 1883 set_cpu_possible(cpu, true);
1882 cpu_set(cpu, cpu_present_map); 1884 set_cpu_present(cpu, true);
1883} 1885}
1884 1886
1885#ifdef CONFIG_X86_64 1887#ifdef CONFIG_X86_64
@@ -2081,7 +2083,7 @@ __cpuinit int apic_is_clustered_box(void)
2081 bios_cpu_apicid = early_per_cpu_ptr(x86_bios_cpu_apicid); 2083 bios_cpu_apicid = early_per_cpu_ptr(x86_bios_cpu_apicid);
2082 bitmap_zero(clustermap, NUM_APIC_CLUSTERS); 2084 bitmap_zero(clustermap, NUM_APIC_CLUSTERS);
2083 2085
2084 for (i = 0; i < NR_CPUS; i++) { 2086 for (i = 0; i < nr_cpu_ids; i++) {
2085 /* are we being called early in kernel startup? */ 2087 /* are we being called early in kernel startup? */
2086 if (bios_cpu_apicid) { 2088 if (bios_cpu_apicid) {
2087 id = bios_cpu_apicid[i]; 2089 id = bios_cpu_apicid[i];
diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c
index 68b5d8681cbb..c6ecda64f5f1 100644
--- a/arch/x86/kernel/cpu/intel_cacheinfo.c
+++ b/arch/x86/kernel/cpu/intel_cacheinfo.c
@@ -534,31 +534,16 @@ static void __cpuinit free_cache_attributes(unsigned int cpu)
534 per_cpu(cpuid4_info, cpu) = NULL; 534 per_cpu(cpuid4_info, cpu) = NULL;
535} 535}
536 536
537static int __cpuinit detect_cache_attributes(unsigned int cpu) 537static void get_cpu_leaves(void *_retval)
538{ 538{
539 struct _cpuid4_info *this_leaf; 539 int j, *retval = _retval, cpu = smp_processor_id();
540 unsigned long j;
541 int retval;
542 cpumask_t oldmask;
543
544 if (num_cache_leaves == 0)
545 return -ENOENT;
546
547 per_cpu(cpuid4_info, cpu) = kzalloc(
548 sizeof(struct _cpuid4_info) * num_cache_leaves, GFP_KERNEL);
549 if (per_cpu(cpuid4_info, cpu) == NULL)
550 return -ENOMEM;
551
552 oldmask = current->cpus_allowed;
553 retval = set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
554 if (retval)
555 goto out;
556 540
557 /* Do cpuid and store the results */ 541 /* Do cpuid and store the results */
558 for (j = 0; j < num_cache_leaves; j++) { 542 for (j = 0; j < num_cache_leaves; j++) {
543 struct _cpuid4_info *this_leaf;
559 this_leaf = CPUID4_INFO_IDX(cpu, j); 544 this_leaf = CPUID4_INFO_IDX(cpu, j);
560 retval = cpuid4_cache_lookup(j, this_leaf); 545 *retval = cpuid4_cache_lookup(j, this_leaf);
561 if (unlikely(retval < 0)) { 546 if (unlikely(*retval < 0)) {
562 int i; 547 int i;
563 548
564 for (i = 0; i < j; i++) 549 for (i = 0; i < j; i++)
@@ -567,9 +552,21 @@ static int __cpuinit detect_cache_attributes(unsigned int cpu)
567 } 552 }
568 cache_shared_cpu_map_setup(cpu, j); 553 cache_shared_cpu_map_setup(cpu, j);
569 } 554 }
570 set_cpus_allowed_ptr(current, &oldmask); 555}
556
557static int __cpuinit detect_cache_attributes(unsigned int cpu)
558{
559 int retval;
560
561 if (num_cache_leaves == 0)
562 return -ENOENT;
563
564 per_cpu(cpuid4_info, cpu) = kzalloc(
565 sizeof(struct _cpuid4_info) * num_cache_leaves, GFP_KERNEL);
566 if (per_cpu(cpuid4_info, cpu) == NULL)
567 return -ENOMEM;
571 568
572out: 569 smp_call_function_single(cpu, get_cpu_leaves, &retval, true);
573 if (retval) { 570 if (retval) {
574 kfree(per_cpu(cpuid4_info, cpu)); 571 kfree(per_cpu(cpuid4_info, cpu));
575 per_cpu(cpuid4_info, cpu) = NULL; 572 per_cpu(cpuid4_info, cpu) = NULL;
@@ -626,8 +623,8 @@ static ssize_t show_shared_cpu_map_func(struct _cpuid4_info *this_leaf,
626 cpumask_t *mask = &this_leaf->shared_cpu_map; 623 cpumask_t *mask = &this_leaf->shared_cpu_map;
627 624
628 n = type? 625 n = type?
629 cpulist_scnprintf(buf, len-2, *mask): 626 cpulist_scnprintf(buf, len-2, mask) :
630 cpumask_scnprintf(buf, len-2, *mask); 627 cpumask_scnprintf(buf, len-2, mask);
631 buf[n++] = '\n'; 628 buf[n++] = '\n';
632 buf[n] = '\0'; 629 buf[n] = '\0';
633 } 630 }
diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
index 748c8f9e7a05..a5a5e0530370 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
@@ -83,34 +83,41 @@ static DEFINE_PER_CPU(unsigned char, bank_map); /* see which banks are on */
83 * CPU Initialization 83 * CPU Initialization
84 */ 84 */
85 85
86struct thresh_restart {
87 struct threshold_block *b;
88 int reset;
89 u16 old_limit;
90};
91
86/* must be called with correct cpu affinity */ 92/* must be called with correct cpu affinity */
87static void threshold_restart_bank(struct threshold_block *b, 93static long threshold_restart_bank(void *_tr)
88 int reset, u16 old_limit)
89{ 94{
95 struct thresh_restart *tr = _tr;
90 u32 mci_misc_hi, mci_misc_lo; 96 u32 mci_misc_hi, mci_misc_lo;
91 97
92 rdmsr(b->address, mci_misc_lo, mci_misc_hi); 98 rdmsr(tr->b->address, mci_misc_lo, mci_misc_hi);
93 99
94 if (b->threshold_limit < (mci_misc_hi & THRESHOLD_MAX)) 100 if (tr->b->threshold_limit < (mci_misc_hi & THRESHOLD_MAX))
95 reset = 1; /* limit cannot be lower than err count */ 101 tr->reset = 1; /* limit cannot be lower than err count */
96 102
97 if (reset) { /* reset err count and overflow bit */ 103 if (tr->reset) { /* reset err count and overflow bit */
98 mci_misc_hi = 104 mci_misc_hi =
99 (mci_misc_hi & ~(MASK_ERR_COUNT_HI | MASK_OVERFLOW_HI)) | 105 (mci_misc_hi & ~(MASK_ERR_COUNT_HI | MASK_OVERFLOW_HI)) |
100 (THRESHOLD_MAX - b->threshold_limit); 106 (THRESHOLD_MAX - tr->b->threshold_limit);
101 } else if (old_limit) { /* change limit w/o reset */ 107 } else if (tr->old_limit) { /* change limit w/o reset */
102 int new_count = (mci_misc_hi & THRESHOLD_MAX) + 108 int new_count = (mci_misc_hi & THRESHOLD_MAX) +
103 (old_limit - b->threshold_limit); 109 (tr->old_limit - tr->b->threshold_limit);
104 mci_misc_hi = (mci_misc_hi & ~MASK_ERR_COUNT_HI) | 110 mci_misc_hi = (mci_misc_hi & ~MASK_ERR_COUNT_HI) |
105 (new_count & THRESHOLD_MAX); 111 (new_count & THRESHOLD_MAX);
106 } 112 }
107 113
108 b->interrupt_enable ? 114 tr->b->interrupt_enable ?
109 (mci_misc_hi = (mci_misc_hi & ~MASK_INT_TYPE_HI) | INT_TYPE_APIC) : 115 (mci_misc_hi = (mci_misc_hi & ~MASK_INT_TYPE_HI) | INT_TYPE_APIC) :
110 (mci_misc_hi &= ~MASK_INT_TYPE_HI); 116 (mci_misc_hi &= ~MASK_INT_TYPE_HI);
111 117
112 mci_misc_hi |= MASK_COUNT_EN_HI; 118 mci_misc_hi |= MASK_COUNT_EN_HI;
113 wrmsr(b->address, mci_misc_lo, mci_misc_hi); 119 wrmsr(tr->b->address, mci_misc_lo, mci_misc_hi);
120 return 0;
114} 121}
115 122
116/* cpu init entry point, called from mce.c with preempt off */ 123/* cpu init entry point, called from mce.c with preempt off */
@@ -120,6 +127,7 @@ void __cpuinit mce_amd_feature_init(struct cpuinfo_x86 *c)
120 unsigned int cpu = smp_processor_id(); 127 unsigned int cpu = smp_processor_id();
121 u8 lvt_off; 128 u8 lvt_off;
122 u32 low = 0, high = 0, address = 0; 129 u32 low = 0, high = 0, address = 0;
130 struct thresh_restart tr;
123 131
124 for (bank = 0; bank < NR_BANKS; ++bank) { 132 for (bank = 0; bank < NR_BANKS; ++bank) {
125 for (block = 0; block < NR_BLOCKS; ++block) { 133 for (block = 0; block < NR_BLOCKS; ++block) {
@@ -162,7 +170,10 @@ void __cpuinit mce_amd_feature_init(struct cpuinfo_x86 *c)
162 wrmsr(address, low, high); 170 wrmsr(address, low, high);
163 171
164 threshold_defaults.address = address; 172 threshold_defaults.address = address;
165 threshold_restart_bank(&threshold_defaults, 0, 0); 173 tr.b = &threshold_defaults;
174 tr.reset = 0;
175 tr.old_limit = 0;
176 threshold_restart_bank(&tr);
166 } 177 }
167 } 178 }
168} 179}
@@ -251,20 +262,6 @@ struct threshold_attr {
251 ssize_t(*store) (struct threshold_block *, const char *, size_t count); 262 ssize_t(*store) (struct threshold_block *, const char *, size_t count);
252}; 263};
253 264
254static void affinity_set(unsigned int cpu, cpumask_t *oldmask,
255 cpumask_t *newmask)
256{
257 *oldmask = current->cpus_allowed;
258 cpus_clear(*newmask);
259 cpu_set(cpu, *newmask);
260 set_cpus_allowed_ptr(current, newmask);
261}
262
263static void affinity_restore(const cpumask_t *oldmask)
264{
265 set_cpus_allowed_ptr(current, oldmask);
266}
267
268#define SHOW_FIELDS(name) \ 265#define SHOW_FIELDS(name) \
269static ssize_t show_ ## name(struct threshold_block * b, char *buf) \ 266static ssize_t show_ ## name(struct threshold_block * b, char *buf) \
270{ \ 267{ \
@@ -277,15 +274,16 @@ static ssize_t store_interrupt_enable(struct threshold_block *b,
277 const char *buf, size_t count) 274 const char *buf, size_t count)
278{ 275{
279 char *end; 276 char *end;
280 cpumask_t oldmask, newmask; 277 struct thresh_restart tr;
281 unsigned long new = simple_strtoul(buf, &end, 0); 278 unsigned long new = simple_strtoul(buf, &end, 0);
282 if (end == buf) 279 if (end == buf)
283 return -EINVAL; 280 return -EINVAL;
284 b->interrupt_enable = !!new; 281 b->interrupt_enable = !!new;
285 282
286 affinity_set(b->cpu, &oldmask, &newmask); 283 tr.b = b;
287 threshold_restart_bank(b, 0, 0); 284 tr.reset = 0;
288 affinity_restore(&oldmask); 285 tr.old_limit = 0;
286 work_on_cpu(b->cpu, threshold_restart_bank, &tr);
289 287
290 return end - buf; 288 return end - buf;
291} 289}
@@ -294,8 +292,7 @@ static ssize_t store_threshold_limit(struct threshold_block *b,
294 const char *buf, size_t count) 292 const char *buf, size_t count)
295{ 293{
296 char *end; 294 char *end;
297 cpumask_t oldmask, newmask; 295 struct thresh_restart tr;
298 u16 old;
299 unsigned long new = simple_strtoul(buf, &end, 0); 296 unsigned long new = simple_strtoul(buf, &end, 0);
300 if (end == buf) 297 if (end == buf)
301 return -EINVAL; 298 return -EINVAL;
@@ -303,34 +300,36 @@ static ssize_t store_threshold_limit(struct threshold_block *b,
303 new = THRESHOLD_MAX; 300 new = THRESHOLD_MAX;
304 if (new < 1) 301 if (new < 1)
305 new = 1; 302 new = 1;
306 old = b->threshold_limit; 303 tr.old_limit = b->threshold_limit;
307 b->threshold_limit = new; 304 b->threshold_limit = new;
305 tr.b = b;
306 tr.reset = 0;
308 307
309 affinity_set(b->cpu, &oldmask, &newmask); 308 work_on_cpu(b->cpu, threshold_restart_bank, &tr);
310 threshold_restart_bank(b, 0, old);
311 affinity_restore(&oldmask);
312 309
313 return end - buf; 310 return end - buf;
314} 311}
315 312
316static ssize_t show_error_count(struct threshold_block *b, char *buf) 313static long local_error_count(void *_b)
317{ 314{
318 u32 high, low; 315 struct threshold_block *b = _b;
319 cpumask_t oldmask, newmask; 316 u32 low, high;
320 affinity_set(b->cpu, &oldmask, &newmask); 317
321 rdmsr(b->address, low, high); 318 rdmsr(b->address, low, high);
322 affinity_restore(&oldmask); 319 return (high & 0xFFF) - (THRESHOLD_MAX - b->threshold_limit);
323 return sprintf(buf, "%x\n", 320}
324 (high & 0xFFF) - (THRESHOLD_MAX - b->threshold_limit)); 321
322static ssize_t show_error_count(struct threshold_block *b, char *buf)
323{
324 return sprintf(buf, "%lx\n", work_on_cpu(b->cpu, local_error_count, b));
325} 325}
326 326
327static ssize_t store_error_count(struct threshold_block *b, 327static ssize_t store_error_count(struct threshold_block *b,
328 const char *buf, size_t count) 328 const char *buf, size_t count)
329{ 329{
330 cpumask_t oldmask, newmask; 330 struct thresh_restart tr = { .b = b, .reset = 1, .old_limit = 0 };
331 affinity_set(b->cpu, &oldmask, &newmask); 331
332 threshold_restart_bank(b, 1, 0); 332 work_on_cpu(b->cpu, threshold_restart_bank, &tr);
333 affinity_restore(&oldmask);
334 return 1; 333 return 1;
335} 334}
336 335
@@ -463,12 +462,19 @@ out_free:
463 return err; 462 return err;
464} 463}
465 464
465static long local_allocate_threshold_blocks(void *_bank)
466{
467 unsigned int *bank = _bank;
468
469 return allocate_threshold_blocks(smp_processor_id(), *bank, 0,
470 MSR_IA32_MC0_MISC + *bank * 4);
471}
472
466/* symlinks sibling shared banks to first core. first core owns dir/files. */ 473/* symlinks sibling shared banks to first core. first core owns dir/files. */
467static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) 474static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
468{ 475{
469 int i, err = 0; 476 int i, err = 0;
470 struct threshold_bank *b = NULL; 477 struct threshold_bank *b = NULL;
471 cpumask_t oldmask, newmask;
472 char name[32]; 478 char name[32];
473 479
474 sprintf(name, "threshold_bank%i", bank); 480 sprintf(name, "threshold_bank%i", bank);
@@ -519,11 +525,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
519 525
520 per_cpu(threshold_banks, cpu)[bank] = b; 526 per_cpu(threshold_banks, cpu)[bank] = b;
521 527
522 affinity_set(cpu, &oldmask, &newmask); 528 err = work_on_cpu(cpu, local_allocate_threshold_blocks, &bank);
523 err = allocate_threshold_blocks(cpu, bank, 0,
524 MSR_IA32_MC0_MISC + bank * 4);
525 affinity_restore(&oldmask);
526
527 if (err) 529 if (err)
528 goto out_free; 530 goto out_free;
529 531
diff --git a/arch/x86/kernel/genapic_flat_64.c b/arch/x86/kernel/genapic_flat_64.c
index c0262791bda4..34185488e4fb 100644
--- a/arch/x86/kernel/genapic_flat_64.c
+++ b/arch/x86/kernel/genapic_flat_64.c
@@ -30,12 +30,12 @@ static int flat_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
30 return 1; 30 return 1;
31} 31}
32 32
33static cpumask_t flat_target_cpus(void) 33static const struct cpumask *flat_target_cpus(void)
34{ 34{
35 return cpu_online_map; 35 return cpu_online_mask;
36} 36}
37 37
38static cpumask_t flat_vector_allocation_domain(int cpu) 38static void flat_vector_allocation_domain(int cpu, struct cpumask *retmask)
39{ 39{
40 /* Careful. Some cpus do not strictly honor the set of cpus 40 /* Careful. Some cpus do not strictly honor the set of cpus
41 * specified in the interrupt destination when using lowest 41 * specified in the interrupt destination when using lowest
@@ -45,8 +45,8 @@ static cpumask_t flat_vector_allocation_domain(int cpu)
45 * deliver interrupts to the wrong hyperthread when only one 45 * deliver interrupts to the wrong hyperthread when only one
46 * hyperthread was specified in the interrupt desitination. 46 * hyperthread was specified in the interrupt desitination.
47 */ 47 */
48 cpumask_t domain = { { [0] = APIC_ALL_CPUS, } }; 48 cpumask_clear(retmask);
49 return domain; 49 cpumask_bits(retmask)[0] = APIC_ALL_CPUS;
50} 50}
51 51
52/* 52/*
@@ -69,9 +69,8 @@ static void flat_init_apic_ldr(void)
69 apic_write(APIC_LDR, val); 69 apic_write(APIC_LDR, val);
70} 70}
71 71
72static void flat_send_IPI_mask(cpumask_t cpumask, int vector) 72static inline void _flat_send_IPI_mask(unsigned long mask, int vector)
73{ 73{
74 unsigned long mask = cpus_addr(cpumask)[0];
75 unsigned long flags; 74 unsigned long flags;
76 75
77 local_irq_save(flags); 76 local_irq_save(flags);
@@ -79,20 +78,41 @@ static void flat_send_IPI_mask(cpumask_t cpumask, int vector)
79 local_irq_restore(flags); 78 local_irq_restore(flags);
80} 79}
81 80
81static void flat_send_IPI_mask(const struct cpumask *cpumask, int vector)
82{
83 unsigned long mask = cpumask_bits(cpumask)[0];
84
85 _flat_send_IPI_mask(mask, vector);
86}
87
88static void flat_send_IPI_mask_allbutself(const struct cpumask *cpumask,
89 int vector)
90{
91 unsigned long mask = cpumask_bits(cpumask)[0];
92 int cpu = smp_processor_id();
93
94 if (cpu < BITS_PER_LONG)
95 clear_bit(cpu, &mask);
96 _flat_send_IPI_mask(mask, vector);
97}
98
82static void flat_send_IPI_allbutself(int vector) 99static void flat_send_IPI_allbutself(int vector)
83{ 100{
101 int cpu = smp_processor_id();
84#ifdef CONFIG_HOTPLUG_CPU 102#ifdef CONFIG_HOTPLUG_CPU
85 int hotplug = 1; 103 int hotplug = 1;
86#else 104#else
87 int hotplug = 0; 105 int hotplug = 0;
88#endif 106#endif
89 if (hotplug || vector == NMI_VECTOR) { 107 if (hotplug || vector == NMI_VECTOR) {
90 cpumask_t allbutme = cpu_online_map; 108 if (!cpumask_equal(cpu_online_mask, cpumask_of(cpu))) {
109 unsigned long mask = cpumask_bits(cpu_online_mask)[0];
91 110
92 cpu_clear(smp_processor_id(), allbutme); 111 if (cpu < BITS_PER_LONG)
112 clear_bit(cpu, &mask);
93 113
94 if (!cpus_empty(allbutme)) 114 _flat_send_IPI_mask(mask, vector);
95 flat_send_IPI_mask(allbutme, vector); 115 }
96 } else if (num_online_cpus() > 1) { 116 } else if (num_online_cpus() > 1) {
97 __send_IPI_shortcut(APIC_DEST_ALLBUT, vector,APIC_DEST_LOGICAL); 117 __send_IPI_shortcut(APIC_DEST_ALLBUT, vector,APIC_DEST_LOGICAL);
98 } 118 }
@@ -101,7 +121,7 @@ static void flat_send_IPI_allbutself(int vector)
101static void flat_send_IPI_all(int vector) 121static void flat_send_IPI_all(int vector)
102{ 122{
103 if (vector == NMI_VECTOR) 123 if (vector == NMI_VECTOR)
104 flat_send_IPI_mask(cpu_online_map, vector); 124 flat_send_IPI_mask(cpu_online_mask, vector);
105 else 125 else
106 __send_IPI_shortcut(APIC_DEST_ALLINC, vector, APIC_DEST_LOGICAL); 126 __send_IPI_shortcut(APIC_DEST_ALLINC, vector, APIC_DEST_LOGICAL);
107} 127}
@@ -135,9 +155,18 @@ static int flat_apic_id_registered(void)
135 return physid_isset(read_xapic_id(), phys_cpu_present_map); 155 return physid_isset(read_xapic_id(), phys_cpu_present_map);
136} 156}
137 157
138static unsigned int flat_cpu_mask_to_apicid(cpumask_t cpumask) 158static unsigned int flat_cpu_mask_to_apicid(const struct cpumask *cpumask)
159{
160 return cpumask_bits(cpumask)[0] & APIC_ALL_CPUS;
161}
162
163static unsigned int flat_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
164 const struct cpumask *andmask)
139{ 165{
140 return cpus_addr(cpumask)[0] & APIC_ALL_CPUS; 166 unsigned long mask1 = cpumask_bits(cpumask)[0] & APIC_ALL_CPUS;
167 unsigned long mask2 = cpumask_bits(andmask)[0] & APIC_ALL_CPUS;
168
169 return mask1 & mask2;
141} 170}
142 171
143static unsigned int phys_pkg_id(int index_msb) 172static unsigned int phys_pkg_id(int index_msb)
@@ -157,8 +186,10 @@ struct genapic apic_flat = {
157 .send_IPI_all = flat_send_IPI_all, 186 .send_IPI_all = flat_send_IPI_all,
158 .send_IPI_allbutself = flat_send_IPI_allbutself, 187 .send_IPI_allbutself = flat_send_IPI_allbutself,
159 .send_IPI_mask = flat_send_IPI_mask, 188 .send_IPI_mask = flat_send_IPI_mask,
189 .send_IPI_mask_allbutself = flat_send_IPI_mask_allbutself,
160 .send_IPI_self = apic_send_IPI_self, 190 .send_IPI_self = apic_send_IPI_self,
161 .cpu_mask_to_apicid = flat_cpu_mask_to_apicid, 191 .cpu_mask_to_apicid = flat_cpu_mask_to_apicid,
192 .cpu_mask_to_apicid_and = flat_cpu_mask_to_apicid_and,
162 .phys_pkg_id = phys_pkg_id, 193 .phys_pkg_id = phys_pkg_id,
163 .get_apic_id = get_apic_id, 194 .get_apic_id = get_apic_id,
164 .set_apic_id = set_apic_id, 195 .set_apic_id = set_apic_id,
@@ -188,35 +219,39 @@ static int physflat_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
188 return 0; 219 return 0;
189} 220}
190 221
191static cpumask_t physflat_target_cpus(void) 222static const struct cpumask *physflat_target_cpus(void)
192{ 223{
193 return cpu_online_map; 224 return cpu_online_mask;
194} 225}
195 226
196static cpumask_t physflat_vector_allocation_domain(int cpu) 227static void physflat_vector_allocation_domain(int cpu, struct cpumask *retmask)
197{ 228{
198 return cpumask_of_cpu(cpu); 229 cpumask_clear(retmask);
230 cpumask_set_cpu(cpu, retmask);
199} 231}
200 232
201static void physflat_send_IPI_mask(cpumask_t cpumask, int vector) 233static void physflat_send_IPI_mask(const struct cpumask *cpumask, int vector)
202{ 234{
203 send_IPI_mask_sequence(cpumask, vector); 235 send_IPI_mask_sequence(cpumask, vector);
204} 236}
205 237
206static void physflat_send_IPI_allbutself(int vector) 238static void physflat_send_IPI_mask_allbutself(const struct cpumask *cpumask,
239 int vector)
207{ 240{
208 cpumask_t allbutme = cpu_online_map; 241 send_IPI_mask_allbutself(cpumask, vector);
242}
209 243
210 cpu_clear(smp_processor_id(), allbutme); 244static void physflat_send_IPI_allbutself(int vector)
211 physflat_send_IPI_mask(allbutme, vector); 245{
246 send_IPI_mask_allbutself(cpu_online_mask, vector);
212} 247}
213 248
214static void physflat_send_IPI_all(int vector) 249static void physflat_send_IPI_all(int vector)
215{ 250{
216 physflat_send_IPI_mask(cpu_online_map, vector); 251 physflat_send_IPI_mask(cpu_online_mask, vector);
217} 252}
218 253
219static unsigned int physflat_cpu_mask_to_apicid(cpumask_t cpumask) 254static unsigned int physflat_cpu_mask_to_apicid(const struct cpumask *cpumask)
220{ 255{
221 int cpu; 256 int cpu;
222 257
@@ -224,13 +259,31 @@ static unsigned int physflat_cpu_mask_to_apicid(cpumask_t cpumask)
224 * We're using fixed IRQ delivery, can only return one phys APIC ID. 259 * We're using fixed IRQ delivery, can only return one phys APIC ID.
225 * May as well be the first. 260 * May as well be the first.
226 */ 261 */
227 cpu = first_cpu(cpumask); 262 cpu = cpumask_first(cpumask);
228 if ((unsigned)cpu < nr_cpu_ids) 263 if ((unsigned)cpu < nr_cpu_ids)
229 return per_cpu(x86_cpu_to_apicid, cpu); 264 return per_cpu(x86_cpu_to_apicid, cpu);
230 else 265 else
231 return BAD_APICID; 266 return BAD_APICID;
232} 267}
233 268
269static unsigned int
270physflat_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
271 const struct cpumask *andmask)
272{
273 int cpu;
274
275 /*
276 * We're using fixed IRQ delivery, can only return one phys APIC ID.
277 * May as well be the first.
278 */
279 for_each_cpu_and(cpu, cpumask, andmask)
280 if (cpumask_test_cpu(cpu, cpu_online_mask))
281 break;
282 if (cpu < nr_cpu_ids)
283 return per_cpu(x86_cpu_to_apicid, cpu);
284 return BAD_APICID;
285}
286
234struct genapic apic_physflat = { 287struct genapic apic_physflat = {
235 .name = "physical flat", 288 .name = "physical flat",
236 .acpi_madt_oem_check = physflat_acpi_madt_oem_check, 289 .acpi_madt_oem_check = physflat_acpi_madt_oem_check,
@@ -243,8 +296,10 @@ struct genapic apic_physflat = {
243 .send_IPI_all = physflat_send_IPI_all, 296 .send_IPI_all = physflat_send_IPI_all,
244 .send_IPI_allbutself = physflat_send_IPI_allbutself, 297 .send_IPI_allbutself = physflat_send_IPI_allbutself,
245 .send_IPI_mask = physflat_send_IPI_mask, 298 .send_IPI_mask = physflat_send_IPI_mask,
299 .send_IPI_mask_allbutself = physflat_send_IPI_mask_allbutself,
246 .send_IPI_self = apic_send_IPI_self, 300 .send_IPI_self = apic_send_IPI_self,
247 .cpu_mask_to_apicid = physflat_cpu_mask_to_apicid, 301 .cpu_mask_to_apicid = physflat_cpu_mask_to_apicid,
302 .cpu_mask_to_apicid_and = physflat_cpu_mask_to_apicid_and,
248 .phys_pkg_id = phys_pkg_id, 303 .phys_pkg_id = phys_pkg_id,
249 .get_apic_id = get_apic_id, 304 .get_apic_id = get_apic_id,
250 .set_apic_id = set_apic_id, 305 .set_apic_id = set_apic_id,
diff --git a/arch/x86/kernel/genx2apic_cluster.c b/arch/x86/kernel/genx2apic_cluster.c
index f6a2c8eb48a6..6ce497cc372d 100644
--- a/arch/x86/kernel/genx2apic_cluster.c
+++ b/arch/x86/kernel/genx2apic_cluster.c
@@ -22,19 +22,18 @@ static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
22 22
23/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */ 23/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */
24 24
25static cpumask_t x2apic_target_cpus(void) 25static const struct cpumask *x2apic_target_cpus(void)
26{ 26{
27 return cpumask_of_cpu(0); 27 return cpumask_of(0);
28} 28}
29 29
30/* 30/*
31 * for now each logical cpu is in its own vector allocation domain. 31 * for now each logical cpu is in its own vector allocation domain.
32 */ 32 */
33static cpumask_t x2apic_vector_allocation_domain(int cpu) 33static void x2apic_vector_allocation_domain(int cpu, struct cpumask *retmask)
34{ 34{
35 cpumask_t domain = CPU_MASK_NONE; 35 cpumask_clear(retmask);
36 cpu_set(cpu, domain); 36 cpumask_set_cpu(cpu, retmask);
37 return domain;
38} 37}
39 38
40static void __x2apic_send_IPI_dest(unsigned int apicid, int vector, 39static void __x2apic_send_IPI_dest(unsigned int apicid, int vector,
@@ -56,32 +55,53 @@ static void __x2apic_send_IPI_dest(unsigned int apicid, int vector,
56 * at once. We have 16 cpu's in a cluster. This will minimize IPI register 55 * at once. We have 16 cpu's in a cluster. This will minimize IPI register
57 * writes. 56 * writes.
58 */ 57 */
59static void x2apic_send_IPI_mask(cpumask_t mask, int vector) 58static void x2apic_send_IPI_mask(const struct cpumask *mask, int vector)
60{ 59{
61 unsigned long flags; 60 unsigned long flags;
62 unsigned long query_cpu; 61 unsigned long query_cpu;
63 62
64 local_irq_save(flags); 63 local_irq_save(flags);
65 for_each_cpu_mask(query_cpu, mask) { 64 for_each_cpu(query_cpu, mask)
66 __x2apic_send_IPI_dest(per_cpu(x86_cpu_to_logical_apicid, query_cpu), 65 __x2apic_send_IPI_dest(
67 vector, APIC_DEST_LOGICAL); 66 per_cpu(x86_cpu_to_logical_apicid, query_cpu),
68 } 67 vector, APIC_DEST_LOGICAL);
69 local_irq_restore(flags); 68 local_irq_restore(flags);
70} 69}
71 70
72static void x2apic_send_IPI_allbutself(int vector) 71static void x2apic_send_IPI_mask_allbutself(const struct cpumask *mask,
72 int vector)
73{ 73{
74 cpumask_t mask = cpu_online_map; 74 unsigned long flags;
75 unsigned long query_cpu;
76 unsigned long this_cpu = smp_processor_id();
75 77
76 cpu_clear(smp_processor_id(), mask); 78 local_irq_save(flags);
79 for_each_cpu(query_cpu, mask)
80 if (query_cpu != this_cpu)
81 __x2apic_send_IPI_dest(
82 per_cpu(x86_cpu_to_logical_apicid, query_cpu),
83 vector, APIC_DEST_LOGICAL);
84 local_irq_restore(flags);
85}
86
87static void x2apic_send_IPI_allbutself(int vector)
88{
89 unsigned long flags;
90 unsigned long query_cpu;
91 unsigned long this_cpu = smp_processor_id();
77 92
78 if (!cpus_empty(mask)) 93 local_irq_save(flags);
79 x2apic_send_IPI_mask(mask, vector); 94 for_each_online_cpu(query_cpu)
95 if (query_cpu != this_cpu)
96 __x2apic_send_IPI_dest(
97 per_cpu(x86_cpu_to_logical_apicid, query_cpu),
98 vector, APIC_DEST_LOGICAL);
99 local_irq_restore(flags);
80} 100}
81 101
82static void x2apic_send_IPI_all(int vector) 102static void x2apic_send_IPI_all(int vector)
83{ 103{
84 x2apic_send_IPI_mask(cpu_online_map, vector); 104 x2apic_send_IPI_mask(cpu_online_mask, vector);
85} 105}
86 106
87static int x2apic_apic_id_registered(void) 107static int x2apic_apic_id_registered(void)
@@ -89,21 +109,38 @@ static int x2apic_apic_id_registered(void)
89 return 1; 109 return 1;
90} 110}
91 111
92static unsigned int x2apic_cpu_mask_to_apicid(cpumask_t cpumask) 112static unsigned int x2apic_cpu_mask_to_apicid(const struct cpumask *cpumask)
93{ 113{
94 int cpu; 114 int cpu;
95 115
96 /* 116 /*
97 * We're using fixed IRQ delivery, can only return one phys APIC ID. 117 * We're using fixed IRQ delivery, can only return one logical APIC ID.
98 * May as well be the first. 118 * May as well be the first.
99 */ 119 */
100 cpu = first_cpu(cpumask); 120 cpu = cpumask_first(cpumask);
101 if ((unsigned)cpu < NR_CPUS) 121 if ((unsigned)cpu < nr_cpu_ids)
102 return per_cpu(x86_cpu_to_logical_apicid, cpu); 122 return per_cpu(x86_cpu_to_logical_apicid, cpu);
103 else 123 else
104 return BAD_APICID; 124 return BAD_APICID;
105} 125}
106 126
127static unsigned int x2apic_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
128 const struct cpumask *andmask)
129{
130 int cpu;
131
132 /*
133 * We're using fixed IRQ delivery, can only return one logical APIC ID.
134 * May as well be the first.
135 */
136 for_each_cpu_and(cpu, cpumask, andmask)
137 if (cpumask_test_cpu(cpu, cpu_online_mask))
138 break;
139 if (cpu < nr_cpu_ids)
140 return per_cpu(x86_cpu_to_logical_apicid, cpu);
141 return BAD_APICID;
142}
143
107static unsigned int get_apic_id(unsigned long x) 144static unsigned int get_apic_id(unsigned long x)
108{ 145{
109 unsigned int id; 146 unsigned int id;
@@ -150,8 +187,10 @@ struct genapic apic_x2apic_cluster = {
150 .send_IPI_all = x2apic_send_IPI_all, 187 .send_IPI_all = x2apic_send_IPI_all,
151 .send_IPI_allbutself = x2apic_send_IPI_allbutself, 188 .send_IPI_allbutself = x2apic_send_IPI_allbutself,
152 .send_IPI_mask = x2apic_send_IPI_mask, 189 .send_IPI_mask = x2apic_send_IPI_mask,
190 .send_IPI_mask_allbutself = x2apic_send_IPI_mask_allbutself,
153 .send_IPI_self = x2apic_send_IPI_self, 191 .send_IPI_self = x2apic_send_IPI_self,
154 .cpu_mask_to_apicid = x2apic_cpu_mask_to_apicid, 192 .cpu_mask_to_apicid = x2apic_cpu_mask_to_apicid,
193 .cpu_mask_to_apicid_and = x2apic_cpu_mask_to_apicid_and,
155 .phys_pkg_id = phys_pkg_id, 194 .phys_pkg_id = phys_pkg_id,
156 .get_apic_id = get_apic_id, 195 .get_apic_id = get_apic_id,
157 .set_apic_id = set_apic_id, 196 .set_apic_id = set_apic_id,
diff --git a/arch/x86/kernel/genx2apic_phys.c b/arch/x86/kernel/genx2apic_phys.c
index d042211768b7..62895cf315ff 100644
--- a/arch/x86/kernel/genx2apic_phys.c
+++ b/arch/x86/kernel/genx2apic_phys.c
@@ -29,16 +29,15 @@ static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
29 29
30/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */ 30/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */
31 31
32static cpumask_t x2apic_target_cpus(void) 32static const struct cpumask *x2apic_target_cpus(void)
33{ 33{
34 return cpumask_of_cpu(0); 34 return cpumask_of(0);
35} 35}
36 36
37static cpumask_t x2apic_vector_allocation_domain(int cpu) 37static void x2apic_vector_allocation_domain(int cpu, struct cpumask *retmask)
38{ 38{
39 cpumask_t domain = CPU_MASK_NONE; 39 cpumask_clear(retmask);
40 cpu_set(cpu, domain); 40 cpumask_set_cpu(cpu, retmask);
41 return domain;
42} 41}
43 42
44static void __x2apic_send_IPI_dest(unsigned int apicid, int vector, 43static void __x2apic_send_IPI_dest(unsigned int apicid, int vector,
@@ -54,32 +53,54 @@ static void __x2apic_send_IPI_dest(unsigned int apicid, int vector,
54 x2apic_icr_write(cfg, apicid); 53 x2apic_icr_write(cfg, apicid);
55} 54}
56 55
57static void x2apic_send_IPI_mask(cpumask_t mask, int vector) 56static void x2apic_send_IPI_mask(const struct cpumask *mask, int vector)
58{ 57{
59 unsigned long flags; 58 unsigned long flags;
60 unsigned long query_cpu; 59 unsigned long query_cpu;
61 60
62 local_irq_save(flags); 61 local_irq_save(flags);
63 for_each_cpu_mask(query_cpu, mask) { 62 for_each_cpu(query_cpu, mask) {
64 __x2apic_send_IPI_dest(per_cpu(x86_cpu_to_apicid, query_cpu), 63 __x2apic_send_IPI_dest(per_cpu(x86_cpu_to_apicid, query_cpu),
65 vector, APIC_DEST_PHYSICAL); 64 vector, APIC_DEST_PHYSICAL);
66 } 65 }
67 local_irq_restore(flags); 66 local_irq_restore(flags);
68} 67}
69 68
70static void x2apic_send_IPI_allbutself(int vector) 69static void x2apic_send_IPI_mask_allbutself(const struct cpumask *mask,
70 int vector)
71{ 71{
72 cpumask_t mask = cpu_online_map; 72 unsigned long flags;
73 unsigned long query_cpu;
74 unsigned long this_cpu = smp_processor_id();
75
76 local_irq_save(flags);
77 for_each_cpu(query_cpu, mask) {
78 if (query_cpu != this_cpu)
79 __x2apic_send_IPI_dest(
80 per_cpu(x86_cpu_to_apicid, query_cpu),
81 vector, APIC_DEST_PHYSICAL);
82 }
83 local_irq_restore(flags);
84}
73 85
74 cpu_clear(smp_processor_id(), mask); 86static void x2apic_send_IPI_allbutself(int vector)
87{
88 unsigned long flags;
89 unsigned long query_cpu;
90 unsigned long this_cpu = smp_processor_id();
75 91
76 if (!cpus_empty(mask)) 92 local_irq_save(flags);
77 x2apic_send_IPI_mask(mask, vector); 93 for_each_online_cpu(query_cpu)
94 if (query_cpu != this_cpu)
95 __x2apic_send_IPI_dest(
96 per_cpu(x86_cpu_to_apicid, query_cpu),
97 vector, APIC_DEST_PHYSICAL);
98 local_irq_restore(flags);
78} 99}
79 100
80static void x2apic_send_IPI_all(int vector) 101static void x2apic_send_IPI_all(int vector)
81{ 102{
82 x2apic_send_IPI_mask(cpu_online_map, vector); 103 x2apic_send_IPI_mask(cpu_online_mask, vector);
83} 104}
84 105
85static int x2apic_apic_id_registered(void) 106static int x2apic_apic_id_registered(void)
@@ -87,7 +108,7 @@ static int x2apic_apic_id_registered(void)
87 return 1; 108 return 1;
88} 109}
89 110
90static unsigned int x2apic_cpu_mask_to_apicid(cpumask_t cpumask) 111static unsigned int x2apic_cpu_mask_to_apicid(const struct cpumask *cpumask)
91{ 112{
92 int cpu; 113 int cpu;
93 114
@@ -95,13 +116,30 @@ static unsigned int x2apic_cpu_mask_to_apicid(cpumask_t cpumask)
95 * We're using fixed IRQ delivery, can only return one phys APIC ID. 116 * We're using fixed IRQ delivery, can only return one phys APIC ID.
96 * May as well be the first. 117 * May as well be the first.
97 */ 118 */
98 cpu = first_cpu(cpumask); 119 cpu = cpumask_first(cpumask);
99 if ((unsigned)cpu < NR_CPUS) 120 if ((unsigned)cpu < nr_cpu_ids)
100 return per_cpu(x86_cpu_to_apicid, cpu); 121 return per_cpu(x86_cpu_to_apicid, cpu);
101 else 122 else
102 return BAD_APICID; 123 return BAD_APICID;
103} 124}
104 125
126static unsigned int x2apic_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
127 const struct cpumask *andmask)
128{
129 int cpu;
130
131 /*
132 * We're using fixed IRQ delivery, can only return one phys APIC ID.
133 * May as well be the first.
134 */
135 for_each_cpu_and(cpu, cpumask, andmask)
136 if (cpumask_test_cpu(cpu, cpu_online_mask))
137 break;
138 if (cpu < nr_cpu_ids)
139 return per_cpu(x86_cpu_to_apicid, cpu);
140 return BAD_APICID;
141}
142
105static unsigned int get_apic_id(unsigned long x) 143static unsigned int get_apic_id(unsigned long x)
106{ 144{
107 unsigned int id; 145 unsigned int id;
@@ -145,8 +183,10 @@ struct genapic apic_x2apic_phys = {
145 .send_IPI_all = x2apic_send_IPI_all, 183 .send_IPI_all = x2apic_send_IPI_all,
146 .send_IPI_allbutself = x2apic_send_IPI_allbutself, 184 .send_IPI_allbutself = x2apic_send_IPI_allbutself,
147 .send_IPI_mask = x2apic_send_IPI_mask, 185 .send_IPI_mask = x2apic_send_IPI_mask,
186 .send_IPI_mask_allbutself = x2apic_send_IPI_mask_allbutself,
148 .send_IPI_self = x2apic_send_IPI_self, 187 .send_IPI_self = x2apic_send_IPI_self,
149 .cpu_mask_to_apicid = x2apic_cpu_mask_to_apicid, 188 .cpu_mask_to_apicid = x2apic_cpu_mask_to_apicid,
189 .cpu_mask_to_apicid_and = x2apic_cpu_mask_to_apicid_and,
150 .phys_pkg_id = phys_pkg_id, 190 .phys_pkg_id = phys_pkg_id,
151 .get_apic_id = get_apic_id, 191 .get_apic_id = get_apic_id,
152 .set_apic_id = set_apic_id, 192 .set_apic_id = set_apic_id,
diff --git a/arch/x86/kernel/genx2apic_uv_x.c b/arch/x86/kernel/genx2apic_uv_x.c
index dece17289731..b193e082f6ce 100644
--- a/arch/x86/kernel/genx2apic_uv_x.c
+++ b/arch/x86/kernel/genx2apic_uv_x.c
@@ -79,16 +79,15 @@ EXPORT_SYMBOL(sn_rtc_cycles_per_second);
79 79
80/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */ 80/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */
81 81
82static cpumask_t uv_target_cpus(void) 82static const struct cpumask *uv_target_cpus(void)
83{ 83{
84 return cpumask_of_cpu(0); 84 return cpumask_of(0);
85} 85}
86 86
87static cpumask_t uv_vector_allocation_domain(int cpu) 87static void uv_vector_allocation_domain(int cpu, struct cpumask *retmask)
88{ 88{
89 cpumask_t domain = CPU_MASK_NONE; 89 cpumask_clear(retmask);
90 cpu_set(cpu, domain); 90 cpumask_set_cpu(cpu, retmask);
91 return domain;
92} 91}
93 92
94int uv_wakeup_secondary(int phys_apicid, unsigned int start_rip) 93int uv_wakeup_secondary(int phys_apicid, unsigned int start_rip)
@@ -127,28 +126,37 @@ static void uv_send_IPI_one(int cpu, int vector)
127 uv_write_global_mmr64(pnode, UVH_IPI_INT, val); 126 uv_write_global_mmr64(pnode, UVH_IPI_INT, val);
128} 127}
129 128
130static void uv_send_IPI_mask(cpumask_t mask, int vector) 129static void uv_send_IPI_mask(const struct cpumask *mask, int vector)
131{ 130{
132 unsigned int cpu; 131 unsigned int cpu;
133 132
134 for_each_possible_cpu(cpu) 133 for_each_cpu(cpu, mask)
135 if (cpu_isset(cpu, mask)) 134 uv_send_IPI_one(cpu, vector);
135}
136
137static void uv_send_IPI_mask_allbutself(const struct cpumask *mask, int vector)
138{
139 unsigned int cpu;
140 unsigned int this_cpu = smp_processor_id();
141
142 for_each_cpu(cpu, mask)
143 if (cpu != this_cpu)
136 uv_send_IPI_one(cpu, vector); 144 uv_send_IPI_one(cpu, vector);
137} 145}
138 146
139static void uv_send_IPI_allbutself(int vector) 147static void uv_send_IPI_allbutself(int vector)
140{ 148{
141 cpumask_t mask = cpu_online_map; 149 unsigned int cpu;
142 150 unsigned int this_cpu = smp_processor_id();
143 cpu_clear(smp_processor_id(), mask);
144 151
145 if (!cpus_empty(mask)) 152 for_each_online_cpu(cpu)
146 uv_send_IPI_mask(mask, vector); 153 if (cpu != this_cpu)
154 uv_send_IPI_one(cpu, vector);
147} 155}
148 156
149static void uv_send_IPI_all(int vector) 157static void uv_send_IPI_all(int vector)
150{ 158{
151 uv_send_IPI_mask(cpu_online_map, vector); 159 uv_send_IPI_mask(cpu_online_mask, vector);
152} 160}
153 161
154static int uv_apic_id_registered(void) 162static int uv_apic_id_registered(void)
@@ -160,7 +168,7 @@ static void uv_init_apic_ldr(void)
160{ 168{
161} 169}
162 170
163static unsigned int uv_cpu_mask_to_apicid(cpumask_t cpumask) 171static unsigned int uv_cpu_mask_to_apicid(const struct cpumask *cpumask)
164{ 172{
165 int cpu; 173 int cpu;
166 174
@@ -168,13 +176,30 @@ static unsigned int uv_cpu_mask_to_apicid(cpumask_t cpumask)
168 * We're using fixed IRQ delivery, can only return one phys APIC ID. 176 * We're using fixed IRQ delivery, can only return one phys APIC ID.
169 * May as well be the first. 177 * May as well be the first.
170 */ 178 */
171 cpu = first_cpu(cpumask); 179 cpu = cpumask_first(cpumask);
172 if ((unsigned)cpu < nr_cpu_ids) 180 if ((unsigned)cpu < nr_cpu_ids)
173 return per_cpu(x86_cpu_to_apicid, cpu); 181 return per_cpu(x86_cpu_to_apicid, cpu);
174 else 182 else
175 return BAD_APICID; 183 return BAD_APICID;
176} 184}
177 185
186static unsigned int uv_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
187 const struct cpumask *andmask)
188{
189 int cpu;
190
191 /*
192 * We're using fixed IRQ delivery, can only return one phys APIC ID.
193 * May as well be the first.
194 */
195 for_each_cpu_and(cpu, cpumask, andmask)
196 if (cpumask_test_cpu(cpu, cpu_online_mask))
197 break;
198 if (cpu < nr_cpu_ids)
199 return per_cpu(x86_cpu_to_apicid, cpu);
200 return BAD_APICID;
201}
202
178static unsigned int get_apic_id(unsigned long x) 203static unsigned int get_apic_id(unsigned long x)
179{ 204{
180 unsigned int id; 205 unsigned int id;
@@ -222,8 +247,10 @@ struct genapic apic_x2apic_uv_x = {
222 .send_IPI_all = uv_send_IPI_all, 247 .send_IPI_all = uv_send_IPI_all,
223 .send_IPI_allbutself = uv_send_IPI_allbutself, 248 .send_IPI_allbutself = uv_send_IPI_allbutself,
224 .send_IPI_mask = uv_send_IPI_mask, 249 .send_IPI_mask = uv_send_IPI_mask,
250 .send_IPI_mask_allbutself = uv_send_IPI_mask_allbutself,
225 .send_IPI_self = uv_send_IPI_self, 251 .send_IPI_self = uv_send_IPI_self,
226 .cpu_mask_to_apicid = uv_cpu_mask_to_apicid, 252 .cpu_mask_to_apicid = uv_cpu_mask_to_apicid,
253 .cpu_mask_to_apicid_and = uv_cpu_mask_to_apicid_and,
227 .phys_pkg_id = phys_pkg_id, 254 .phys_pkg_id = phys_pkg_id,
228 .get_apic_id = get_apic_id, 255 .get_apic_id = get_apic_id,
229 .set_apic_id = set_apic_id, 256 .set_apic_id = set_apic_id,
diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c
index 845ea097383e..cd759ad90690 100644
--- a/arch/x86/kernel/hpet.c
+++ b/arch/x86/kernel/hpet.c
@@ -248,7 +248,7 @@ static void hpet_legacy_clockevent_register(void)
248 * Start hpet with the boot cpu mask and make it 248 * Start hpet with the boot cpu mask and make it
249 * global after the IO_APIC has been initialized. 249 * global after the IO_APIC has been initialized.
250 */ 250 */
251 hpet_clockevent.cpumask = cpumask_of_cpu(smp_processor_id()); 251 hpet_clockevent.cpumask = cpumask_of(smp_processor_id());
252 clockevents_register_device(&hpet_clockevent); 252 clockevents_register_device(&hpet_clockevent);
253 global_clock_event = &hpet_clockevent; 253 global_clock_event = &hpet_clockevent;
254 printk(KERN_DEBUG "hpet clockevent registered\n"); 254 printk(KERN_DEBUG "hpet clockevent registered\n");
@@ -303,7 +303,7 @@ static void hpet_set_mode(enum clock_event_mode mode,
303 struct hpet_dev *hdev = EVT_TO_HPET_DEV(evt); 303 struct hpet_dev *hdev = EVT_TO_HPET_DEV(evt);
304 hpet_setup_msi_irq(hdev->irq); 304 hpet_setup_msi_irq(hdev->irq);
305 disable_irq(hdev->irq); 305 disable_irq(hdev->irq);
306 irq_set_affinity(hdev->irq, cpumask_of_cpu(hdev->cpu)); 306 irq_set_affinity(hdev->irq, cpumask_of(hdev->cpu));
307 enable_irq(hdev->irq); 307 enable_irq(hdev->irq);
308 } 308 }
309 break; 309 break;
@@ -451,7 +451,7 @@ static int hpet_setup_irq(struct hpet_dev *dev)
451 return -1; 451 return -1;
452 452
453 disable_irq(dev->irq); 453 disable_irq(dev->irq);
454 irq_set_affinity(dev->irq, cpumask_of_cpu(dev->cpu)); 454 irq_set_affinity(dev->irq, cpumask_of(dev->cpu));
455 enable_irq(dev->irq); 455 enable_irq(dev->irq);
456 456
457 printk(KERN_DEBUG "hpet: %s irq %d for MSI\n", 457 printk(KERN_DEBUG "hpet: %s irq %d for MSI\n",
@@ -502,7 +502,7 @@ static void init_one_hpet_msi_clockevent(struct hpet_dev *hdev, int cpu)
502 /* 5 usec minimum reprogramming delta. */ 502 /* 5 usec minimum reprogramming delta. */
503 evt->min_delta_ns = 5000; 503 evt->min_delta_ns = 5000;
504 504
505 evt->cpumask = cpumask_of_cpu(hdev->cpu); 505 evt->cpumask = cpumask_of(hdev->cpu);
506 clockevents_register_device(evt); 506 clockevents_register_device(evt);
507} 507}
508 508
diff --git a/arch/x86/kernel/i8253.c b/arch/x86/kernel/i8253.c
index c1b5e3ece1f2..10f92fb532f3 100644
--- a/arch/x86/kernel/i8253.c
+++ b/arch/x86/kernel/i8253.c
@@ -114,7 +114,7 @@ void __init setup_pit_timer(void)
114 * Start pit with the boot cpu mask and make it global after the 114 * Start pit with the boot cpu mask and make it global after the
115 * IO_APIC has been initialized. 115 * IO_APIC has been initialized.
116 */ 116 */
117 pit_clockevent.cpumask = cpumask_of_cpu(smp_processor_id()); 117 pit_clockevent.cpumask = cpumask_of(smp_processor_id());
118 pit_clockevent.mult = div_sc(CLOCK_TICK_RATE, NSEC_PER_SEC, 118 pit_clockevent.mult = div_sc(CLOCK_TICK_RATE, NSEC_PER_SEC,
119 pit_clockevent.shift); 119 pit_clockevent.shift);
120 pit_clockevent.max_delta_ns = 120 pit_clockevent.max_delta_ns =
diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c
index 74917658b004..62ecfc991e1e 100644
--- a/arch/x86/kernel/io_apic.c
+++ b/arch/x86/kernel/io_apic.c
@@ -136,8 +136,8 @@ static struct irq_pin_list *get_one_free_irq_2_pin(int cpu)
136 136
137struct irq_cfg { 137struct irq_cfg {
138 struct irq_pin_list *irq_2_pin; 138 struct irq_pin_list *irq_2_pin;
139 cpumask_t domain; 139 cpumask_var_t domain;
140 cpumask_t old_domain; 140 cpumask_var_t old_domain;
141 unsigned move_cleanup_count; 141 unsigned move_cleanup_count;
142 u8 vector; 142 u8 vector;
143 u8 move_in_progress : 1; 143 u8 move_in_progress : 1;
@@ -152,22 +152,22 @@ static struct irq_cfg irq_cfgx[] = {
152#else 152#else
153static struct irq_cfg irq_cfgx[NR_IRQS] = { 153static struct irq_cfg irq_cfgx[NR_IRQS] = {
154#endif 154#endif
155 [0] = { .domain = CPU_MASK_ALL, .vector = IRQ0_VECTOR, }, 155 [0] = { .vector = IRQ0_VECTOR, },
156 [1] = { .domain = CPU_MASK_ALL, .vector = IRQ1_VECTOR, }, 156 [1] = { .vector = IRQ1_VECTOR, },
157 [2] = { .domain = CPU_MASK_ALL, .vector = IRQ2_VECTOR, }, 157 [2] = { .vector = IRQ2_VECTOR, },
158 [3] = { .domain = CPU_MASK_ALL, .vector = IRQ3_VECTOR, }, 158 [3] = { .vector = IRQ3_VECTOR, },
159 [4] = { .domain = CPU_MASK_ALL, .vector = IRQ4_VECTOR, }, 159 [4] = { .vector = IRQ4_VECTOR, },
160 [5] = { .domain = CPU_MASK_ALL, .vector = IRQ5_VECTOR, }, 160 [5] = { .vector = IRQ5_VECTOR, },
161 [6] = { .domain = CPU_MASK_ALL, .vector = IRQ6_VECTOR, }, 161 [6] = { .vector = IRQ6_VECTOR, },
162 [7] = { .domain = CPU_MASK_ALL, .vector = IRQ7_VECTOR, }, 162 [7] = { .vector = IRQ7_VECTOR, },
163 [8] = { .domain = CPU_MASK_ALL, .vector = IRQ8_VECTOR, }, 163 [8] = { .vector = IRQ8_VECTOR, },
164 [9] = { .domain = CPU_MASK_ALL, .vector = IRQ9_VECTOR, }, 164 [9] = { .vector = IRQ9_VECTOR, },
165 [10] = { .domain = CPU_MASK_ALL, .vector = IRQ10_VECTOR, }, 165 [10] = { .vector = IRQ10_VECTOR, },
166 [11] = { .domain = CPU_MASK_ALL, .vector = IRQ11_VECTOR, }, 166 [11] = { .vector = IRQ11_VECTOR, },
167 [12] = { .domain = CPU_MASK_ALL, .vector = IRQ12_VECTOR, }, 167 [12] = { .vector = IRQ12_VECTOR, },
168 [13] = { .domain = CPU_MASK_ALL, .vector = IRQ13_VECTOR, }, 168 [13] = { .vector = IRQ13_VECTOR, },
169 [14] = { .domain = CPU_MASK_ALL, .vector = IRQ14_VECTOR, }, 169 [14] = { .vector = IRQ14_VECTOR, },
170 [15] = { .domain = CPU_MASK_ALL, .vector = IRQ15_VECTOR, }, 170 [15] = { .vector = IRQ15_VECTOR, },
171}; 171};
172 172
173int __init arch_early_irq_init(void) 173int __init arch_early_irq_init(void)
@@ -183,6 +183,10 @@ int __init arch_early_irq_init(void)
183 for (i = 0; i < count; i++) { 183 for (i = 0; i < count; i++) {
184 desc = irq_to_desc(i); 184 desc = irq_to_desc(i);
185 desc->chip_data = &cfg[i]; 185 desc->chip_data = &cfg[i];
186 alloc_bootmem_cpumask_var(&cfg[i].domain);
187 alloc_bootmem_cpumask_var(&cfg[i].old_domain);
188 if (i < NR_IRQS_LEGACY)
189 cpumask_setall(cfg[i].domain);
186 } 190 }
187 191
188 return 0; 192 return 0;
@@ -209,6 +213,20 @@ static struct irq_cfg *get_one_free_irq_cfg(int cpu)
209 node = cpu_to_node(cpu); 213 node = cpu_to_node(cpu);
210 214
211 cfg = kzalloc_node(sizeof(*cfg), GFP_ATOMIC, node); 215 cfg = kzalloc_node(sizeof(*cfg), GFP_ATOMIC, node);
216 if (cfg) {
217 /* FIXME: needs alloc_cpumask_var_node() */
218 if (!alloc_cpumask_var(&cfg->domain, GFP_ATOMIC)) {
219 kfree(cfg);
220 cfg = NULL;
221 } else if (!alloc_cpumask_var(&cfg->old_domain, GFP_ATOMIC)) {
222 free_cpumask_var(cfg->domain);
223 kfree(cfg);
224 cfg = NULL;
225 } else {
226 cpumask_clear(cfg->domain);
227 cpumask_clear(cfg->old_domain);
228 }
229 }
212 printk(KERN_DEBUG " alloc irq_cfg on cpu %d node %d\n", cpu, node); 230 printk(KERN_DEBUG " alloc irq_cfg on cpu %d node %d\n", cpu, node);
213 231
214 return cfg; 232 return cfg;
@@ -333,13 +351,14 @@ void arch_free_chip_data(struct irq_desc *old_desc, struct irq_desc *desc)
333 } 351 }
334} 352}
335 353
336static void set_extra_move_desc(struct irq_desc *desc, cpumask_t mask) 354static void
355set_extra_move_desc(struct irq_desc *desc, const struct cpumask *mask)
337{ 356{
338 struct irq_cfg *cfg = desc->chip_data; 357 struct irq_cfg *cfg = desc->chip_data;
339 358
340 if (!cfg->move_in_progress) { 359 if (!cfg->move_in_progress) {
341 /* it means that domain is not changed */ 360 /* it means that domain is not changed */
342 if (!cpus_intersects(desc->affinity, mask)) 361 if (!cpumask_intersects(&desc->affinity, mask))
343 cfg->move_desc_pending = 1; 362 cfg->move_desc_pending = 1;
344 } 363 }
345} 364}
@@ -354,7 +373,8 @@ static struct irq_cfg *irq_cfg(unsigned int irq)
354#endif 373#endif
355 374
356#ifndef CONFIG_NUMA_MIGRATE_IRQ_DESC 375#ifndef CONFIG_NUMA_MIGRATE_IRQ_DESC
357static inline void set_extra_move_desc(struct irq_desc *desc, cpumask_t mask) 376static inline void
377set_extra_move_desc(struct irq_desc *desc, const struct cpumask *mask)
358{ 378{
359} 379}
360#endif 380#endif
@@ -485,6 +505,26 @@ static void ioapic_mask_entry(int apic, int pin)
485} 505}
486 506
487#ifdef CONFIG_SMP 507#ifdef CONFIG_SMP
508static void send_cleanup_vector(struct irq_cfg *cfg)
509{
510 cpumask_var_t cleanup_mask;
511
512 if (unlikely(!alloc_cpumask_var(&cleanup_mask, GFP_ATOMIC))) {
513 unsigned int i;
514 cfg->move_cleanup_count = 0;
515 for_each_cpu_and(i, cfg->old_domain, cpu_online_mask)
516 cfg->move_cleanup_count++;
517 for_each_cpu_and(i, cfg->old_domain, cpu_online_mask)
518 send_IPI_mask(cpumask_of(i), IRQ_MOVE_CLEANUP_VECTOR);
519 } else {
520 cpumask_and(cleanup_mask, cfg->old_domain, cpu_online_mask);
521 cfg->move_cleanup_count = cpumask_weight(cleanup_mask);
522 send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
523 free_cpumask_var(cleanup_mask);
524 }
525 cfg->move_in_progress = 0;
526}
527
488static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, struct irq_cfg *cfg) 528static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, struct irq_cfg *cfg)
489{ 529{
490 int apic, pin; 530 int apic, pin;
@@ -520,41 +560,55 @@ static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, struct irq
520 } 560 }
521} 561}
522 562
523static int assign_irq_vector(int irq, struct irq_cfg *cfg, cpumask_t mask); 563static int
564assign_irq_vector(int irq, struct irq_cfg *cfg, const struct cpumask *mask);
524 565
525static void set_ioapic_affinity_irq_desc(struct irq_desc *desc, cpumask_t mask) 566/*
567 * Either sets desc->affinity to a valid value, and returns cpu_mask_to_apicid
568 * of that, or returns BAD_APICID and leaves desc->affinity untouched.
569 */
570static unsigned int
571set_desc_affinity(struct irq_desc *desc, const struct cpumask *mask)
526{ 572{
527 struct irq_cfg *cfg; 573 struct irq_cfg *cfg;
528 unsigned long flags;
529 unsigned int dest;
530 cpumask_t tmp;
531 unsigned int irq; 574 unsigned int irq;
532 575
533 cpus_and(tmp, mask, cpu_online_map); 576 if (!cpumask_intersects(mask, cpu_online_mask))
534 if (cpus_empty(tmp)) 577 return BAD_APICID;
535 return;
536 578
537 irq = desc->irq; 579 irq = desc->irq;
538 cfg = desc->chip_data; 580 cfg = desc->chip_data;
539 if (assign_irq_vector(irq, cfg, mask)) 581 if (assign_irq_vector(irq, cfg, mask))
540 return; 582 return BAD_APICID;
541 583
584 cpumask_and(&desc->affinity, cfg->domain, mask);
542 set_extra_move_desc(desc, mask); 585 set_extra_move_desc(desc, mask);
586 return cpu_mask_to_apicid_and(&desc->affinity, cpu_online_mask);
587}
543 588
544 cpus_and(tmp, cfg->domain, mask); 589static void
545 dest = cpu_mask_to_apicid(tmp); 590set_ioapic_affinity_irq_desc(struct irq_desc *desc, const struct cpumask *mask)
546 /* 591{
547 * Only the high 8 bits are valid. 592 struct irq_cfg *cfg;
548 */ 593 unsigned long flags;
549 dest = SET_APIC_LOGICAL_ID(dest); 594 unsigned int dest;
595 unsigned int irq;
596
597 irq = desc->irq;
598 cfg = desc->chip_data;
550 599
551 spin_lock_irqsave(&ioapic_lock, flags); 600 spin_lock_irqsave(&ioapic_lock, flags);
552 __target_IO_APIC_irq(irq, dest, cfg); 601 dest = set_desc_affinity(desc, mask);
553 desc->affinity = mask; 602 if (dest != BAD_APICID) {
603 /* Only the high 8 bits are valid. */
604 dest = SET_APIC_LOGICAL_ID(dest);
605 __target_IO_APIC_irq(irq, dest, cfg);
606 }
554 spin_unlock_irqrestore(&ioapic_lock, flags); 607 spin_unlock_irqrestore(&ioapic_lock, flags);
555} 608}
556 609
557static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask) 610static void
611set_ioapic_affinity_irq(unsigned int irq, const struct cpumask *mask)
558{ 612{
559 struct irq_desc *desc; 613 struct irq_desc *desc;
560 614
@@ -1222,7 +1276,8 @@ void unlock_vector_lock(void)
1222 spin_unlock(&vector_lock); 1276 spin_unlock(&vector_lock);
1223} 1277}
1224 1278
1225static int __assign_irq_vector(int irq, struct irq_cfg *cfg, cpumask_t mask) 1279static int
1280__assign_irq_vector(int irq, struct irq_cfg *cfg, const struct cpumask *mask)
1226{ 1281{
1227 /* 1282 /*
1228 * NOTE! The local APIC isn't very good at handling 1283 * NOTE! The local APIC isn't very good at handling
@@ -1237,49 +1292,49 @@ static int __assign_irq_vector(int irq, struct irq_cfg *cfg, cpumask_t mask)
1237 */ 1292 */
1238 static int current_vector = FIRST_DEVICE_VECTOR, current_offset = 0; 1293 static int current_vector = FIRST_DEVICE_VECTOR, current_offset = 0;
1239 unsigned int old_vector; 1294 unsigned int old_vector;
1240 int cpu; 1295 int cpu, err;
1296 cpumask_var_t tmp_mask;
1241 1297
1242 if ((cfg->move_in_progress) || cfg->move_cleanup_count) 1298 if ((cfg->move_in_progress) || cfg->move_cleanup_count)
1243 return -EBUSY; 1299 return -EBUSY;
1244 1300
1245 /* Only try and allocate irqs on cpus that are present */ 1301 if (!alloc_cpumask_var(&tmp_mask, GFP_ATOMIC))
1246 cpus_and(mask, mask, cpu_online_map); 1302 return -ENOMEM;
1247 1303
1248 old_vector = cfg->vector; 1304 old_vector = cfg->vector;
1249 if (old_vector) { 1305 if (old_vector) {
1250 cpumask_t tmp; 1306 cpumask_and(tmp_mask, mask, cpu_online_mask);
1251 cpus_and(tmp, cfg->domain, mask); 1307 cpumask_and(tmp_mask, cfg->domain, tmp_mask);
1252 if (!cpus_empty(tmp)) 1308 if (!cpumask_empty(tmp_mask)) {
1309 free_cpumask_var(tmp_mask);
1253 return 0; 1310 return 0;
1311 }
1254 } 1312 }
1255 1313
1256 for_each_cpu_mask_nr(cpu, mask) { 1314 /* Only try and allocate irqs on cpus that are present */
1257 cpumask_t domain, new_mask; 1315 err = -ENOSPC;
1316 for_each_cpu_and(cpu, mask, cpu_online_mask) {
1258 int new_cpu; 1317 int new_cpu;
1259 int vector, offset; 1318 int vector, offset;
1260 1319
1261 domain = vector_allocation_domain(cpu); 1320 vector_allocation_domain(cpu, tmp_mask);
1262 cpus_and(new_mask, domain, cpu_online_map);
1263 1321
1264 vector = current_vector; 1322 vector = current_vector;
1265 offset = current_offset; 1323 offset = current_offset;
1266next: 1324next:
1267 vector += 8; 1325 vector += 8;
1268 if (vector >= first_system_vector) { 1326 if (vector >= first_system_vector) {
1269 /* If we run out of vectors on large boxen, must share them. */ 1327 /* If out of vectors on large boxen, must share them. */
1270 offset = (offset + 1) % 8; 1328 offset = (offset + 1) % 8;
1271 vector = FIRST_DEVICE_VECTOR + offset; 1329 vector = FIRST_DEVICE_VECTOR + offset;
1272 } 1330 }
1273 if (unlikely(current_vector == vector)) 1331 if (unlikely(current_vector == vector))
1274 continue; 1332 continue;
1275#ifdef CONFIG_X86_64 1333
1276 if (vector == IA32_SYSCALL_VECTOR) 1334 if (test_bit(vector, used_vectors))
1277 goto next;
1278#else
1279 if (vector == SYSCALL_VECTOR)
1280 goto next; 1335 goto next;
1281#endif 1336
1282 for_each_cpu_mask_nr(new_cpu, new_mask) 1337 for_each_cpu_and(new_cpu, tmp_mask, cpu_online_mask)
1283 if (per_cpu(vector_irq, new_cpu)[vector] != -1) 1338 if (per_cpu(vector_irq, new_cpu)[vector] != -1)
1284 goto next; 1339 goto next;
1285 /* Found one! */ 1340 /* Found one! */
@@ -1287,18 +1342,21 @@ next:
1287 current_offset = offset; 1342 current_offset = offset;
1288 if (old_vector) { 1343 if (old_vector) {
1289 cfg->move_in_progress = 1; 1344 cfg->move_in_progress = 1;
1290 cfg->old_domain = cfg->domain; 1345 cpumask_copy(cfg->old_domain, cfg->domain);
1291 } 1346 }
1292 for_each_cpu_mask_nr(new_cpu, new_mask) 1347 for_each_cpu_and(new_cpu, tmp_mask, cpu_online_mask)
1293 per_cpu(vector_irq, new_cpu)[vector] = irq; 1348 per_cpu(vector_irq, new_cpu)[vector] = irq;
1294 cfg->vector = vector; 1349 cfg->vector = vector;
1295 cfg->domain = domain; 1350 cpumask_copy(cfg->domain, tmp_mask);
1296 return 0; 1351 err = 0;
1352 break;
1297 } 1353 }
1298 return -ENOSPC; 1354 free_cpumask_var(tmp_mask);
1355 return err;
1299} 1356}
1300 1357
1301static int assign_irq_vector(int irq, struct irq_cfg *cfg, cpumask_t mask) 1358static int
1359assign_irq_vector(int irq, struct irq_cfg *cfg, const struct cpumask *mask)
1302{ 1360{
1303 int err; 1361 int err;
1304 unsigned long flags; 1362 unsigned long flags;
@@ -1311,23 +1369,20 @@ static int assign_irq_vector(int irq, struct irq_cfg *cfg, cpumask_t mask)
1311 1369
1312static void __clear_irq_vector(int irq, struct irq_cfg *cfg) 1370static void __clear_irq_vector(int irq, struct irq_cfg *cfg)
1313{ 1371{
1314 cpumask_t mask;
1315 int cpu, vector; 1372 int cpu, vector;
1316 1373
1317 BUG_ON(!cfg->vector); 1374 BUG_ON(!cfg->vector);
1318 1375
1319 vector = cfg->vector; 1376 vector = cfg->vector;
1320 cpus_and(mask, cfg->domain, cpu_online_map); 1377 for_each_cpu_and(cpu, cfg->domain, cpu_online_mask)
1321 for_each_cpu_mask_nr(cpu, mask)
1322 per_cpu(vector_irq, cpu)[vector] = -1; 1378 per_cpu(vector_irq, cpu)[vector] = -1;
1323 1379
1324 cfg->vector = 0; 1380 cfg->vector = 0;
1325 cpus_clear(cfg->domain); 1381 cpumask_clear(cfg->domain);
1326 1382
1327 if (likely(!cfg->move_in_progress)) 1383 if (likely(!cfg->move_in_progress))
1328 return; 1384 return;
1329 cpus_and(mask, cfg->old_domain, cpu_online_map); 1385 for_each_cpu_and(cpu, cfg->old_domain, cpu_online_mask) {
1330 for_each_cpu_mask_nr(cpu, mask) {
1331 for (vector = FIRST_EXTERNAL_VECTOR; vector < NR_VECTORS; 1386 for (vector = FIRST_EXTERNAL_VECTOR; vector < NR_VECTORS;
1332 vector++) { 1387 vector++) {
1333 if (per_cpu(vector_irq, cpu)[vector] != irq) 1388 if (per_cpu(vector_irq, cpu)[vector] != irq)
@@ -1350,7 +1405,7 @@ void __setup_vector_irq(int cpu)
1350 /* Mark the inuse vectors */ 1405 /* Mark the inuse vectors */
1351 for_each_irq_desc(irq, desc) { 1406 for_each_irq_desc(irq, desc) {
1352 cfg = desc->chip_data; 1407 cfg = desc->chip_data;
1353 if (!cpu_isset(cpu, cfg->domain)) 1408 if (!cpumask_test_cpu(cpu, cfg->domain))
1354 continue; 1409 continue;
1355 vector = cfg->vector; 1410 vector = cfg->vector;
1356 per_cpu(vector_irq, cpu)[vector] = irq; 1411 per_cpu(vector_irq, cpu)[vector] = irq;
@@ -1362,7 +1417,7 @@ void __setup_vector_irq(int cpu)
1362 continue; 1417 continue;
1363 1418
1364 cfg = irq_cfg(irq); 1419 cfg = irq_cfg(irq);
1365 if (!cpu_isset(cpu, cfg->domain)) 1420 if (!cpumask_test_cpu(cpu, cfg->domain))
1366 per_cpu(vector_irq, cpu)[vector] = -1; 1421 per_cpu(vector_irq, cpu)[vector] = -1;
1367 } 1422 }
1368} 1423}
@@ -1498,18 +1553,17 @@ static void setup_IO_APIC_irq(int apic, int pin, unsigned int irq, struct irq_de
1498{ 1553{
1499 struct irq_cfg *cfg; 1554 struct irq_cfg *cfg;
1500 struct IO_APIC_route_entry entry; 1555 struct IO_APIC_route_entry entry;
1501 cpumask_t mask; 1556 unsigned int dest;
1502 1557
1503 if (!IO_APIC_IRQ(irq)) 1558 if (!IO_APIC_IRQ(irq))
1504 return; 1559 return;
1505 1560
1506 cfg = desc->chip_data; 1561 cfg = desc->chip_data;
1507 1562
1508 mask = TARGET_CPUS; 1563 if (assign_irq_vector(irq, cfg, TARGET_CPUS))
1509 if (assign_irq_vector(irq, cfg, mask))
1510 return; 1564 return;
1511 1565
1512 cpus_and(mask, cfg->domain, mask); 1566 dest = cpu_mask_to_apicid_and(cfg->domain, TARGET_CPUS);
1513 1567
1514 apic_printk(APIC_VERBOSE,KERN_DEBUG 1568 apic_printk(APIC_VERBOSE,KERN_DEBUG
1515 "IOAPIC[%d]: Set routing entry (%d-%d -> 0x%x -> " 1569 "IOAPIC[%d]: Set routing entry (%d-%d -> 0x%x -> "
@@ -1519,8 +1573,7 @@ static void setup_IO_APIC_irq(int apic, int pin, unsigned int irq, struct irq_de
1519 1573
1520 1574
1521 if (setup_ioapic_entry(mp_ioapics[apic].mp_apicid, irq, &entry, 1575 if (setup_ioapic_entry(mp_ioapics[apic].mp_apicid, irq, &entry,
1522 cpu_mask_to_apicid(mask), trigger, polarity, 1576 dest, trigger, polarity, cfg->vector)) {
1523 cfg->vector)) {
1524 printk("Failed to setup ioapic entry for ioapic %d, pin %d\n", 1577 printk("Failed to setup ioapic entry for ioapic %d, pin %d\n",
1525 mp_ioapics[apic].mp_apicid, pin); 1578 mp_ioapics[apic].mp_apicid, pin);
1526 __clear_irq_vector(irq, cfg); 1579 __clear_irq_vector(irq, cfg);
@@ -2240,7 +2293,7 @@ static int ioapic_retrigger_irq(unsigned int irq)
2240 unsigned long flags; 2293 unsigned long flags;
2241 2294
2242 spin_lock_irqsave(&vector_lock, flags); 2295 spin_lock_irqsave(&vector_lock, flags);
2243 send_IPI_mask(cpumask_of_cpu(first_cpu(cfg->domain)), cfg->vector); 2296 send_IPI_mask(cpumask_of(cpumask_first(cfg->domain)), cfg->vector);
2244 spin_unlock_irqrestore(&vector_lock, flags); 2297 spin_unlock_irqrestore(&vector_lock, flags);
2245 2298
2246 return 1; 2299 return 1;
@@ -2289,18 +2342,17 @@ static DECLARE_DELAYED_WORK(ir_migration_work, ir_irq_migration);
2289 * as simple as edge triggered migration and we can do the irq migration 2342 * as simple as edge triggered migration and we can do the irq migration
2290 * with a simple atomic update to IO-APIC RTE. 2343 * with a simple atomic update to IO-APIC RTE.
2291 */ 2344 */
2292static void migrate_ioapic_irq_desc(struct irq_desc *desc, cpumask_t mask) 2345static void
2346migrate_ioapic_irq_desc(struct irq_desc *desc, const struct cpumask *mask)
2293{ 2347{
2294 struct irq_cfg *cfg; 2348 struct irq_cfg *cfg;
2295 cpumask_t tmp, cleanup_mask;
2296 struct irte irte; 2349 struct irte irte;
2297 int modify_ioapic_rte; 2350 int modify_ioapic_rte;
2298 unsigned int dest; 2351 unsigned int dest;
2299 unsigned long flags; 2352 unsigned long flags;
2300 unsigned int irq; 2353 unsigned int irq;
2301 2354
2302 cpus_and(tmp, mask, cpu_online_map); 2355 if (!cpumask_intersects(mask, cpu_online_mask))
2303 if (cpus_empty(tmp))
2304 return; 2356 return;
2305 2357
2306 irq = desc->irq; 2358 irq = desc->irq;
@@ -2313,8 +2365,7 @@ static void migrate_ioapic_irq_desc(struct irq_desc *desc, cpumask_t mask)
2313 2365
2314 set_extra_move_desc(desc, mask); 2366 set_extra_move_desc(desc, mask);
2315 2367
2316 cpus_and(tmp, cfg->domain, mask); 2368 dest = cpu_mask_to_apicid_and(cfg->domain, mask);
2317 dest = cpu_mask_to_apicid(tmp);
2318 2369
2319 modify_ioapic_rte = desc->status & IRQ_LEVEL; 2370 modify_ioapic_rte = desc->status & IRQ_LEVEL;
2320 if (modify_ioapic_rte) { 2371 if (modify_ioapic_rte) {
@@ -2331,14 +2382,10 @@ static void migrate_ioapic_irq_desc(struct irq_desc *desc, cpumask_t mask)
2331 */ 2382 */
2332 modify_irte(irq, &irte); 2383 modify_irte(irq, &irte);
2333 2384
2334 if (cfg->move_in_progress) { 2385 if (cfg->move_in_progress)
2335 cpus_and(cleanup_mask, cfg->old_domain, cpu_online_map); 2386 send_cleanup_vector(cfg);
2336 cfg->move_cleanup_count = cpus_weight(cleanup_mask);
2337 send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
2338 cfg->move_in_progress = 0;
2339 }
2340 2387
2341 desc->affinity = mask; 2388 cpumask_copy(&desc->affinity, mask);
2342} 2389}
2343 2390
2344static int migrate_irq_remapped_level_desc(struct irq_desc *desc) 2391static int migrate_irq_remapped_level_desc(struct irq_desc *desc)
@@ -2360,11 +2407,11 @@ static int migrate_irq_remapped_level_desc(struct irq_desc *desc)
2360 } 2407 }
2361 2408
2362 /* everthing is clear. we have right of way */ 2409 /* everthing is clear. we have right of way */
2363 migrate_ioapic_irq_desc(desc, desc->pending_mask); 2410 migrate_ioapic_irq_desc(desc, &desc->pending_mask);
2364 2411
2365 ret = 0; 2412 ret = 0;
2366 desc->status &= ~IRQ_MOVE_PENDING; 2413 desc->status &= ~IRQ_MOVE_PENDING;
2367 cpus_clear(desc->pending_mask); 2414 cpumask_clear(&desc->pending_mask);
2368 2415
2369unmask: 2416unmask:
2370 unmask_IO_APIC_irq_desc(desc); 2417 unmask_IO_APIC_irq_desc(desc);
@@ -2389,7 +2436,7 @@ static void ir_irq_migration(struct work_struct *work)
2389 continue; 2436 continue;
2390 } 2437 }
2391 2438
2392 desc->chip->set_affinity(irq, desc->pending_mask); 2439 desc->chip->set_affinity(irq, &desc->pending_mask);
2393 spin_unlock_irqrestore(&desc->lock, flags); 2440 spin_unlock_irqrestore(&desc->lock, flags);
2394 } 2441 }
2395 } 2442 }
@@ -2398,18 +2445,20 @@ static void ir_irq_migration(struct work_struct *work)
2398/* 2445/*
2399 * Migrates the IRQ destination in the process context. 2446 * Migrates the IRQ destination in the process context.
2400 */ 2447 */
2401static void set_ir_ioapic_affinity_irq_desc(struct irq_desc *desc, cpumask_t mask) 2448static void set_ir_ioapic_affinity_irq_desc(struct irq_desc *desc,
2449 const struct cpumask *mask)
2402{ 2450{
2403 if (desc->status & IRQ_LEVEL) { 2451 if (desc->status & IRQ_LEVEL) {
2404 desc->status |= IRQ_MOVE_PENDING; 2452 desc->status |= IRQ_MOVE_PENDING;
2405 desc->pending_mask = mask; 2453 cpumask_copy(&desc->pending_mask, mask);
2406 migrate_irq_remapped_level_desc(desc); 2454 migrate_irq_remapped_level_desc(desc);
2407 return; 2455 return;
2408 } 2456 }
2409 2457
2410 migrate_ioapic_irq_desc(desc, mask); 2458 migrate_ioapic_irq_desc(desc, mask);
2411} 2459}
2412static void set_ir_ioapic_affinity_irq(unsigned int irq, cpumask_t mask) 2460static void set_ir_ioapic_affinity_irq(unsigned int irq,
2461 const struct cpumask *mask)
2413{ 2462{
2414 struct irq_desc *desc = irq_to_desc(irq); 2463 struct irq_desc *desc = irq_to_desc(irq);
2415 2464
@@ -2444,7 +2493,7 @@ asmlinkage void smp_irq_move_cleanup_interrupt(void)
2444 if (!cfg->move_cleanup_count) 2493 if (!cfg->move_cleanup_count)
2445 goto unlock; 2494 goto unlock;
2446 2495
2447 if ((vector == cfg->vector) && cpu_isset(me, cfg->domain)) 2496 if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain))
2448 goto unlock; 2497 goto unlock;
2449 2498
2450 __get_cpu_var(vector_irq)[vector] = -1; 2499 __get_cpu_var(vector_irq)[vector] = -1;
@@ -2481,20 +2530,14 @@ static void irq_complete_move(struct irq_desc **descp)
2481 2530
2482 vector = ~get_irq_regs()->orig_ax; 2531 vector = ~get_irq_regs()->orig_ax;
2483 me = smp_processor_id(); 2532 me = smp_processor_id();
2484 if ((vector == cfg->vector) && cpu_isset(me, cfg->domain)) {
2485 cpumask_t cleanup_mask;
2486
2487#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC 2533#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
2488 *descp = desc = move_irq_desc(desc, me); 2534 *descp = desc = move_irq_desc(desc, me);
2489 /* get the new one */ 2535 /* get the new one */
2490 cfg = desc->chip_data; 2536 cfg = desc->chip_data;
2491#endif 2537#endif
2492 2538
2493 cpus_and(cleanup_mask, cfg->old_domain, cpu_online_map); 2539 if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain))
2494 cfg->move_cleanup_count = cpus_weight(cleanup_mask); 2540 send_cleanup_vector(cfg);
2495 send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
2496 cfg->move_in_progress = 0;
2497 }
2498} 2541}
2499#else 2542#else
2500static inline void irq_complete_move(struct irq_desc **descp) {} 2543static inline void irq_complete_move(struct irq_desc **descp) {}
@@ -3216,16 +3259,13 @@ static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_ms
3216 struct irq_cfg *cfg; 3259 struct irq_cfg *cfg;
3217 int err; 3260 int err;
3218 unsigned dest; 3261 unsigned dest;
3219 cpumask_t tmp;
3220 3262
3221 cfg = irq_cfg(irq); 3263 cfg = irq_cfg(irq);
3222 tmp = TARGET_CPUS; 3264 err = assign_irq_vector(irq, cfg, TARGET_CPUS);
3223 err = assign_irq_vector(irq, cfg, tmp);
3224 if (err) 3265 if (err)
3225 return err; 3266 return err;
3226 3267
3227 cpus_and(tmp, cfg->domain, tmp); 3268 dest = cpu_mask_to_apicid_and(cfg->domain, TARGET_CPUS);
3228 dest = cpu_mask_to_apicid(tmp);
3229 3269
3230#ifdef CONFIG_INTR_REMAP 3270#ifdef CONFIG_INTR_REMAP
3231 if (irq_remapped(irq)) { 3271 if (irq_remapped(irq)) {
@@ -3279,26 +3319,18 @@ static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_ms
3279} 3319}
3280 3320
3281#ifdef CONFIG_SMP 3321#ifdef CONFIG_SMP
3282static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask) 3322static void set_msi_irq_affinity(unsigned int irq, const struct cpumask *mask)
3283{ 3323{
3284 struct irq_desc *desc = irq_to_desc(irq); 3324 struct irq_desc *desc = irq_to_desc(irq);
3285 struct irq_cfg *cfg; 3325 struct irq_cfg *cfg;
3286 struct msi_msg msg; 3326 struct msi_msg msg;
3287 unsigned int dest; 3327 unsigned int dest;
3288 cpumask_t tmp;
3289 3328
3290 cpus_and(tmp, mask, cpu_online_map); 3329 dest = set_desc_affinity(desc, mask);
3291 if (cpus_empty(tmp)) 3330 if (dest == BAD_APICID)
3292 return; 3331 return;
3293 3332
3294 cfg = desc->chip_data; 3333 cfg = desc->chip_data;
3295 if (assign_irq_vector(irq, cfg, mask))
3296 return;
3297
3298 set_extra_move_desc(desc, mask);
3299
3300 cpus_and(tmp, cfg->domain, mask);
3301 dest = cpu_mask_to_apicid(tmp);
3302 3334
3303 read_msi_msg_desc(desc, &msg); 3335 read_msi_msg_desc(desc, &msg);
3304 3336
@@ -3308,37 +3340,27 @@ static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
3308 msg.address_lo |= MSI_ADDR_DEST_ID(dest); 3340 msg.address_lo |= MSI_ADDR_DEST_ID(dest);
3309 3341
3310 write_msi_msg_desc(desc, &msg); 3342 write_msi_msg_desc(desc, &msg);
3311 desc->affinity = mask;
3312} 3343}
3313#ifdef CONFIG_INTR_REMAP 3344#ifdef CONFIG_INTR_REMAP
3314/* 3345/*
3315 * Migrate the MSI irq to another cpumask. This migration is 3346 * Migrate the MSI irq to another cpumask. This migration is
3316 * done in the process context using interrupt-remapping hardware. 3347 * done in the process context using interrupt-remapping hardware.
3317 */ 3348 */
3318static void ir_set_msi_irq_affinity(unsigned int irq, cpumask_t mask) 3349static void
3350ir_set_msi_irq_affinity(unsigned int irq, const struct cpumask *mask)
3319{ 3351{
3320 struct irq_desc *desc = irq_to_desc(irq); 3352 struct irq_desc *desc = irq_to_desc(irq);
3321 struct irq_cfg *cfg; 3353 struct irq_cfg *cfg = desc->chip_data;
3322 unsigned int dest; 3354 unsigned int dest;
3323 cpumask_t tmp, cleanup_mask;
3324 struct irte irte; 3355 struct irte irte;
3325 3356
3326 cpus_and(tmp, mask, cpu_online_map);
3327 if (cpus_empty(tmp))
3328 return;
3329
3330 if (get_irte(irq, &irte)) 3357 if (get_irte(irq, &irte))
3331 return; 3358 return;
3332 3359
3333 cfg = desc->chip_data; 3360 dest = set_desc_affinity(desc, mask);
3334 if (assign_irq_vector(irq, cfg, mask)) 3361 if (dest == BAD_APICID)
3335 return; 3362 return;
3336 3363
3337 set_extra_move_desc(desc, mask);
3338
3339 cpus_and(tmp, cfg->domain, mask);
3340 dest = cpu_mask_to_apicid(tmp);
3341
3342 irte.vector = cfg->vector; 3364 irte.vector = cfg->vector;
3343 irte.dest_id = IRTE_DEST(dest); 3365 irte.dest_id = IRTE_DEST(dest);
3344 3366
@@ -3352,14 +3374,8 @@ static void ir_set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
3352 * at the new destination. So, time to cleanup the previous 3374 * at the new destination. So, time to cleanup the previous
3353 * vector allocation. 3375 * vector allocation.
3354 */ 3376 */
3355 if (cfg->move_in_progress) { 3377 if (cfg->move_in_progress)
3356 cpus_and(cleanup_mask, cfg->old_domain, cpu_online_map); 3378 send_cleanup_vector(cfg);
3357 cfg->move_cleanup_count = cpus_weight(cleanup_mask);
3358 send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
3359 cfg->move_in_progress = 0;
3360 }
3361
3362 desc->affinity = mask;
3363} 3379}
3364 3380
3365#endif 3381#endif
@@ -3550,26 +3566,18 @@ void arch_teardown_msi_irq(unsigned int irq)
3550 3566
3551#ifdef CONFIG_DMAR 3567#ifdef CONFIG_DMAR
3552#ifdef CONFIG_SMP 3568#ifdef CONFIG_SMP
3553static void dmar_msi_set_affinity(unsigned int irq, cpumask_t mask) 3569static void dmar_msi_set_affinity(unsigned int irq, const struct cpumask *mask)
3554{ 3570{
3555 struct irq_desc *desc = irq_to_desc(irq); 3571 struct irq_desc *desc = irq_to_desc(irq);
3556 struct irq_cfg *cfg; 3572 struct irq_cfg *cfg;
3557 struct msi_msg msg; 3573 struct msi_msg msg;
3558 unsigned int dest; 3574 unsigned int dest;
3559 cpumask_t tmp;
3560 3575
3561 cpus_and(tmp, mask, cpu_online_map); 3576 dest = set_desc_affinity(desc, mask);
3562 if (cpus_empty(tmp)) 3577 if (dest == BAD_APICID)
3563 return; 3578 return;
3564 3579
3565 cfg = desc->chip_data; 3580 cfg = desc->chip_data;
3566 if (assign_irq_vector(irq, cfg, mask))
3567 return;
3568
3569 set_extra_move_desc(desc, mask);
3570
3571 cpus_and(tmp, cfg->domain, mask);
3572 dest = cpu_mask_to_apicid(tmp);
3573 3581
3574 dmar_msi_read(irq, &msg); 3582 dmar_msi_read(irq, &msg);
3575 3583
@@ -3579,7 +3587,6 @@ static void dmar_msi_set_affinity(unsigned int irq, cpumask_t mask)
3579 msg.address_lo |= MSI_ADDR_DEST_ID(dest); 3587 msg.address_lo |= MSI_ADDR_DEST_ID(dest);
3580 3588
3581 dmar_msi_write(irq, &msg); 3589 dmar_msi_write(irq, &msg);
3582 desc->affinity = mask;
3583} 3590}
3584 3591
3585#endif /* CONFIG_SMP */ 3592#endif /* CONFIG_SMP */
@@ -3613,26 +3620,18 @@ int arch_setup_dmar_msi(unsigned int irq)
3613#ifdef CONFIG_HPET_TIMER 3620#ifdef CONFIG_HPET_TIMER
3614 3621
3615#ifdef CONFIG_SMP 3622#ifdef CONFIG_SMP
3616static void hpet_msi_set_affinity(unsigned int irq, cpumask_t mask) 3623static void hpet_msi_set_affinity(unsigned int irq, const struct cpumask *mask)
3617{ 3624{
3618 struct irq_desc *desc = irq_to_desc(irq); 3625 struct irq_desc *desc = irq_to_desc(irq);
3619 struct irq_cfg *cfg; 3626 struct irq_cfg *cfg;
3620 struct msi_msg msg; 3627 struct msi_msg msg;
3621 unsigned int dest; 3628 unsigned int dest;
3622 cpumask_t tmp;
3623 3629
3624 cpus_and(tmp, mask, cpu_online_map); 3630 dest = set_desc_affinity(desc, mask);
3625 if (cpus_empty(tmp)) 3631 if (dest == BAD_APICID)
3626 return; 3632 return;
3627 3633
3628 cfg = desc->chip_data; 3634 cfg = desc->chip_data;
3629 if (assign_irq_vector(irq, cfg, mask))
3630 return;
3631
3632 set_extra_move_desc(desc, mask);
3633
3634 cpus_and(tmp, cfg->domain, mask);
3635 dest = cpu_mask_to_apicid(tmp);
3636 3635
3637 hpet_msi_read(irq, &msg); 3636 hpet_msi_read(irq, &msg);
3638 3637
@@ -3642,7 +3641,6 @@ static void hpet_msi_set_affinity(unsigned int irq, cpumask_t mask)
3642 msg.address_lo |= MSI_ADDR_DEST_ID(dest); 3641 msg.address_lo |= MSI_ADDR_DEST_ID(dest);
3643 3642
3644 hpet_msi_write(irq, &msg); 3643 hpet_msi_write(irq, &msg);
3645 desc->affinity = mask;
3646} 3644}
3647 3645
3648#endif /* CONFIG_SMP */ 3646#endif /* CONFIG_SMP */
@@ -3697,28 +3695,19 @@ static void target_ht_irq(unsigned int irq, unsigned int dest, u8 vector)
3697 write_ht_irq_msg(irq, &msg); 3695 write_ht_irq_msg(irq, &msg);
3698} 3696}
3699 3697
3700static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask) 3698static void set_ht_irq_affinity(unsigned int irq, const struct cpumask *mask)
3701{ 3699{
3702 struct irq_desc *desc = irq_to_desc(irq); 3700 struct irq_desc *desc = irq_to_desc(irq);
3703 struct irq_cfg *cfg; 3701 struct irq_cfg *cfg;
3704 unsigned int dest; 3702 unsigned int dest;
3705 cpumask_t tmp;
3706 3703
3707 cpus_and(tmp, mask, cpu_online_map); 3704 dest = set_desc_affinity(desc, mask);
3708 if (cpus_empty(tmp)) 3705 if (dest == BAD_APICID)
3709 return; 3706 return;
3710 3707
3711 cfg = desc->chip_data; 3708 cfg = desc->chip_data;
3712 if (assign_irq_vector(irq, cfg, mask))
3713 return;
3714
3715 set_extra_move_desc(desc, mask);
3716
3717 cpus_and(tmp, cfg->domain, mask);
3718 dest = cpu_mask_to_apicid(tmp);
3719 3709
3720 target_ht_irq(irq, dest, cfg->vector); 3710 target_ht_irq(irq, dest, cfg->vector);
3721 desc->affinity = mask;
3722} 3711}
3723 3712
3724#endif 3713#endif
@@ -3738,17 +3727,14 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
3738{ 3727{
3739 struct irq_cfg *cfg; 3728 struct irq_cfg *cfg;
3740 int err; 3729 int err;
3741 cpumask_t tmp;
3742 3730
3743 cfg = irq_cfg(irq); 3731 cfg = irq_cfg(irq);
3744 tmp = TARGET_CPUS; 3732 err = assign_irq_vector(irq, cfg, TARGET_CPUS);
3745 err = assign_irq_vector(irq, cfg, tmp);
3746 if (!err) { 3733 if (!err) {
3747 struct ht_irq_msg msg; 3734 struct ht_irq_msg msg;
3748 unsigned dest; 3735 unsigned dest;
3749 3736
3750 cpus_and(tmp, cfg->domain, tmp); 3737 dest = cpu_mask_to_apicid_and(cfg->domain, TARGET_CPUS);
3751 dest = cpu_mask_to_apicid(tmp);
3752 3738
3753 msg.address_hi = HT_IRQ_HIGH_DEST_ID(dest); 3739 msg.address_hi = HT_IRQ_HIGH_DEST_ID(dest);
3754 3740
@@ -3784,7 +3770,7 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
3784int arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade, 3770int arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
3785 unsigned long mmr_offset) 3771 unsigned long mmr_offset)
3786{ 3772{
3787 const cpumask_t *eligible_cpu = get_cpu_mask(cpu); 3773 const struct cpumask *eligible_cpu = cpumask_of(cpu);
3788 struct irq_cfg *cfg; 3774 struct irq_cfg *cfg;
3789 int mmr_pnode; 3775 int mmr_pnode;
3790 unsigned long mmr_value; 3776 unsigned long mmr_value;
@@ -3794,7 +3780,7 @@ int arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
3794 3780
3795 cfg = irq_cfg(irq); 3781 cfg = irq_cfg(irq);
3796 3782
3797 err = assign_irq_vector(irq, cfg, *eligible_cpu); 3783 err = assign_irq_vector(irq, cfg, eligible_cpu);
3798 if (err != 0) 3784 if (err != 0)
3799 return err; 3785 return err;
3800 3786
@@ -3813,7 +3799,7 @@ int arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
3813 entry->polarity = 0; 3799 entry->polarity = 0;
3814 entry->trigger = 0; 3800 entry->trigger = 0;
3815 entry->mask = 0; 3801 entry->mask = 0;
3816 entry->dest = cpu_mask_to_apicid(*eligible_cpu); 3802 entry->dest = cpu_mask_to_apicid(eligible_cpu);
3817 3803
3818 mmr_pnode = uv_blade_to_pnode(mmr_blade); 3804 mmr_pnode = uv_blade_to_pnode(mmr_blade);
3819 uv_write_global_mmr64(mmr_pnode, mmr_offset, mmr_value); 3805 uv_write_global_mmr64(mmr_pnode, mmr_offset, mmr_value);
@@ -4024,7 +4010,7 @@ void __init setup_ioapic_dest(void)
4024 int pin, ioapic, irq, irq_entry; 4010 int pin, ioapic, irq, irq_entry;
4025 struct irq_desc *desc; 4011 struct irq_desc *desc;
4026 struct irq_cfg *cfg; 4012 struct irq_cfg *cfg;
4027 cpumask_t mask; 4013 const struct cpumask *mask;
4028 4014
4029 if (skip_ioapic_setup == 1) 4015 if (skip_ioapic_setup == 1)
4030 return; 4016 return;
@@ -4055,7 +4041,7 @@ void __init setup_ioapic_dest(void)
4055 */ 4041 */
4056 if (desc->status & 4042 if (desc->status &
4057 (IRQ_NO_BALANCING | IRQ_AFFINITY_SET)) 4043 (IRQ_NO_BALANCING | IRQ_AFFINITY_SET))
4058 mask = desc->affinity; 4044 mask = &desc->affinity;
4059 else 4045 else
4060 mask = TARGET_CPUS; 4046 mask = TARGET_CPUS;
4061 4047
diff --git a/arch/x86/kernel/ipi.c b/arch/x86/kernel/ipi.c
index f1c688e46f35..285bbf8831fa 100644
--- a/arch/x86/kernel/ipi.c
+++ b/arch/x86/kernel/ipi.c
@@ -116,18 +116,18 @@ static inline void __send_IPI_dest_field(unsigned long mask, int vector)
116/* 116/*
117 * This is only used on smaller machines. 117 * This is only used on smaller machines.
118 */ 118 */
119void send_IPI_mask_bitmask(cpumask_t cpumask, int vector) 119void send_IPI_mask_bitmask(const struct cpumask *cpumask, int vector)
120{ 120{
121 unsigned long mask = cpus_addr(cpumask)[0]; 121 unsigned long mask = cpumask_bits(cpumask)[0];
122 unsigned long flags; 122 unsigned long flags;
123 123
124 local_irq_save(flags); 124 local_irq_save(flags);
125 WARN_ON(mask & ~cpus_addr(cpu_online_map)[0]); 125 WARN_ON(mask & ~cpumask_bits(cpu_online_mask)[0]);
126 __send_IPI_dest_field(mask, vector); 126 __send_IPI_dest_field(mask, vector);
127 local_irq_restore(flags); 127 local_irq_restore(flags);
128} 128}
129 129
130void send_IPI_mask_sequence(cpumask_t mask, int vector) 130void send_IPI_mask_sequence(const struct cpumask *mask, int vector)
131{ 131{
132 unsigned long flags; 132 unsigned long flags;
133 unsigned int query_cpu; 133 unsigned int query_cpu;
@@ -139,12 +139,24 @@ void send_IPI_mask_sequence(cpumask_t mask, int vector)
139 */ 139 */
140 140
141 local_irq_save(flags); 141 local_irq_save(flags);
142 for_each_possible_cpu(query_cpu) { 142 for_each_cpu(query_cpu, mask)
143 if (cpu_isset(query_cpu, mask)) { 143 __send_IPI_dest_field(cpu_to_logical_apicid(query_cpu), vector);
144 local_irq_restore(flags);
145}
146
147void send_IPI_mask_allbutself(const struct cpumask *mask, int vector)
148{
149 unsigned long flags;
150 unsigned int query_cpu;
151 unsigned int this_cpu = smp_processor_id();
152
153 /* See Hack comment above */
154
155 local_irq_save(flags);
156 for_each_cpu(query_cpu, mask)
157 if (query_cpu != this_cpu)
144 __send_IPI_dest_field(cpu_to_logical_apicid(query_cpu), 158 __send_IPI_dest_field(cpu_to_logical_apicid(query_cpu),
145 vector); 159 vector);
146 }
147 }
148 local_irq_restore(flags); 160 local_irq_restore(flags);
149} 161}
150 162
diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
index 3f1d9d18df67..bce53e1352a0 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -9,6 +9,7 @@
9#include <asm/apic.h> 9#include <asm/apic.h>
10#include <asm/io_apic.h> 10#include <asm/io_apic.h>
11#include <asm/smp.h> 11#include <asm/smp.h>
12#include <asm/irq.h>
12 13
13atomic_t irq_err_count; 14atomic_t irq_err_count;
14 15
@@ -190,3 +191,5 @@ u64 arch_irq_stat(void)
190#endif 191#endif
191 return sum; 192 return sum;
192} 193}
194
195EXPORT_SYMBOL_GPL(vector_used_by_percpu_irq);
diff --git a/arch/x86/kernel/irq_32.c b/arch/x86/kernel/irq_32.c
index 119fc9c8ff7f..9dc5588f336a 100644
--- a/arch/x86/kernel/irq_32.c
+++ b/arch/x86/kernel/irq_32.c
@@ -233,27 +233,28 @@ unsigned int do_IRQ(struct pt_regs *regs)
233#ifdef CONFIG_HOTPLUG_CPU 233#ifdef CONFIG_HOTPLUG_CPU
234#include <mach_apic.h> 234#include <mach_apic.h>
235 235
236void fixup_irqs(cpumask_t map) 236/* A cpu has been removed from cpu_online_mask. Reset irq affinities. */
237void fixup_irqs(void)
237{ 238{
238 unsigned int irq; 239 unsigned int irq;
239 static int warned; 240 static int warned;
240 struct irq_desc *desc; 241 struct irq_desc *desc;
241 242
242 for_each_irq_desc(irq, desc) { 243 for_each_irq_desc(irq, desc) {
243 cpumask_t mask; 244 const struct cpumask *affinity;
244 245
245 if (!desc) 246 if (!desc)
246 continue; 247 continue;
247 if (irq == 2) 248 if (irq == 2)
248 continue; 249 continue;
249 250
250 cpus_and(mask, desc->affinity, map); 251 affinity = &desc->affinity;
251 if (any_online_cpu(mask) == NR_CPUS) { 252 if (cpumask_any_and(affinity, cpu_online_mask) >= nr_cpu_ids) {
252 printk("Breaking affinity for irq %i\n", irq); 253 printk("Breaking affinity for irq %i\n", irq);
253 mask = map; 254 affinity = cpu_all_mask;
254 } 255 }
255 if (desc->chip->set_affinity) 256 if (desc->chip->set_affinity)
256 desc->chip->set_affinity(irq, mask); 257 desc->chip->set_affinity(irq, affinity);
257 else if (desc->action && !(warned++)) 258 else if (desc->action && !(warned++))
258 printk("Cannot set affinity for irq %i\n", irq); 259 printk("Cannot set affinity for irq %i\n", irq);
259 } 260 }
diff --git a/arch/x86/kernel/irq_64.c b/arch/x86/kernel/irq_64.c
index a174a217eb1a..6383d50f82ea 100644
--- a/arch/x86/kernel/irq_64.c
+++ b/arch/x86/kernel/irq_64.c
@@ -80,16 +80,17 @@ asmlinkage unsigned int __irq_entry do_IRQ(struct pt_regs *regs)
80} 80}
81 81
82#ifdef CONFIG_HOTPLUG_CPU 82#ifdef CONFIG_HOTPLUG_CPU
83void fixup_irqs(cpumask_t map) 83/* A cpu has been removed from cpu_online_mask. Reset irq affinities. */
84void fixup_irqs(void)
84{ 85{
85 unsigned int irq; 86 unsigned int irq;
86 static int warned; 87 static int warned;
87 struct irq_desc *desc; 88 struct irq_desc *desc;
88 89
89 for_each_irq_desc(irq, desc) { 90 for_each_irq_desc(irq, desc) {
90 cpumask_t mask;
91 int break_affinity = 0; 91 int break_affinity = 0;
92 int set_affinity = 1; 92 int set_affinity = 1;
93 const struct cpumask *affinity;
93 94
94 if (!desc) 95 if (!desc)
95 continue; 96 continue;
@@ -99,23 +100,23 @@ void fixup_irqs(cpumask_t map)
99 /* interrupt's are disabled at this point */ 100 /* interrupt's are disabled at this point */
100 spin_lock(&desc->lock); 101 spin_lock(&desc->lock);
101 102
103 affinity = &desc->affinity;
102 if (!irq_has_action(irq) || 104 if (!irq_has_action(irq) ||
103 cpus_equal(desc->affinity, map)) { 105 cpumask_equal(affinity, cpu_online_mask)) {
104 spin_unlock(&desc->lock); 106 spin_unlock(&desc->lock);
105 continue; 107 continue;
106 } 108 }
107 109
108 cpus_and(mask, desc->affinity, map); 110 if (cpumask_any_and(affinity, cpu_online_mask) >= nr_cpu_ids) {
109 if (cpus_empty(mask)) {
110 break_affinity = 1; 111 break_affinity = 1;
111 mask = map; 112 affinity = cpu_all_mask;
112 } 113 }
113 114
114 if (desc->chip->mask) 115 if (desc->chip->mask)
115 desc->chip->mask(irq); 116 desc->chip->mask(irq);
116 117
117 if (desc->chip->set_affinity) 118 if (desc->chip->set_affinity)
118 desc->chip->set_affinity(irq, mask); 119 desc->chip->set_affinity(irq, affinity);
119 else if (!(warned++)) 120 else if (!(warned++))
120 set_affinity = 0; 121 set_affinity = 0;
121 122
diff --git a/arch/x86/kernel/irqinit_32.c b/arch/x86/kernel/irqinit_32.c
index 203384ed2b5d..84723295f88a 100644
--- a/arch/x86/kernel/irqinit_32.c
+++ b/arch/x86/kernel/irqinit_32.c
@@ -110,6 +110,18 @@ DEFINE_PER_CPU(vector_irq_t, vector_irq) = {
110 [IRQ15_VECTOR + 1 ... NR_VECTORS - 1] = -1 110 [IRQ15_VECTOR + 1 ... NR_VECTORS - 1] = -1
111}; 111};
112 112
113int vector_used_by_percpu_irq(unsigned int vector)
114{
115 int cpu;
116
117 for_each_online_cpu(cpu) {
118 if (per_cpu(vector_irq, cpu)[vector] != -1)
119 return 1;
120 }
121
122 return 0;
123}
124
113/* Overridden in paravirt.c */ 125/* Overridden in paravirt.c */
114void init_IRQ(void) __attribute__((weak, alias("native_init_IRQ"))); 126void init_IRQ(void) __attribute__((weak, alias("native_init_IRQ")));
115 127
@@ -146,10 +158,12 @@ void __init native_init_IRQ(void)
146 alloc_intr_gate(CALL_FUNCTION_VECTOR, call_function_interrupt); 158 alloc_intr_gate(CALL_FUNCTION_VECTOR, call_function_interrupt);
147 159
148 /* IPI for single call function */ 160 /* IPI for single call function */
149 set_intr_gate(CALL_FUNCTION_SINGLE_VECTOR, call_function_single_interrupt); 161 alloc_intr_gate(CALL_FUNCTION_SINGLE_VECTOR,
162 call_function_single_interrupt);
150 163
151 /* Low priority IPI to cleanup after moving an irq */ 164 /* Low priority IPI to cleanup after moving an irq */
152 set_intr_gate(IRQ_MOVE_CLEANUP_VECTOR, irq_move_cleanup_interrupt); 165 set_intr_gate(IRQ_MOVE_CLEANUP_VECTOR, irq_move_cleanup_interrupt);
166 set_bit(IRQ_MOVE_CLEANUP_VECTOR, used_vectors);
153#endif 167#endif
154 168
155#ifdef CONFIG_X86_LOCAL_APIC 169#ifdef CONFIG_X86_LOCAL_APIC
diff --git a/arch/x86/kernel/irqinit_64.c b/arch/x86/kernel/irqinit_64.c
index 6190e6ef546c..31ebfe38e96c 100644
--- a/arch/x86/kernel/irqinit_64.c
+++ b/arch/x86/kernel/irqinit_64.c
@@ -69,6 +69,18 @@ DEFINE_PER_CPU(vector_irq_t, vector_irq) = {
69 [IRQ15_VECTOR + 1 ... NR_VECTORS - 1] = -1 69 [IRQ15_VECTOR + 1 ... NR_VECTORS - 1] = -1
70}; 70};
71 71
72int vector_used_by_percpu_irq(unsigned int vector)
73{
74 int cpu;
75
76 for_each_online_cpu(cpu) {
77 if (per_cpu(vector_irq, cpu)[vector] != -1)
78 return 1;
79 }
80
81 return 0;
82}
83
72void __init init_ISA_irqs(void) 84void __init init_ISA_irqs(void)
73{ 85{
74 int i; 86 int i;
@@ -121,6 +133,7 @@ static void __init smp_intr_init(void)
121 133
122 /* Low priority IPI to cleanup after moving an irq */ 134 /* Low priority IPI to cleanup after moving an irq */
123 set_intr_gate(IRQ_MOVE_CLEANUP_VECTOR, irq_move_cleanup_interrupt); 135 set_intr_gate(IRQ_MOVE_CLEANUP_VECTOR, irq_move_cleanup_interrupt);
136 set_bit(IRQ_MOVE_CLEANUP_VECTOR, used_vectors);
124#endif 137#endif
125} 138}
126 139
diff --git a/arch/x86/kernel/mfgpt_32.c b/arch/x86/kernel/mfgpt_32.c
index 3b599518c322..c12314c9e86f 100644
--- a/arch/x86/kernel/mfgpt_32.c
+++ b/arch/x86/kernel/mfgpt_32.c
@@ -287,7 +287,7 @@ static struct clock_event_device mfgpt_clockevent = {
287 .set_mode = mfgpt_set_mode, 287 .set_mode = mfgpt_set_mode,
288 .set_next_event = mfgpt_next_event, 288 .set_next_event = mfgpt_next_event,
289 .rating = 250, 289 .rating = 250,
290 .cpumask = CPU_MASK_ALL, 290 .cpumask = cpu_all_mask,
291 .shift = 32 291 .shift = 32
292}; 292};
293 293
diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c
index 72e0e4e712d6..39643b1df061 100644
--- a/arch/x86/kernel/reboot.c
+++ b/arch/x86/kernel/reboot.c
@@ -650,10 +650,7 @@ static int crash_nmi_callback(struct notifier_block *self,
650 650
651static void smp_send_nmi_allbutself(void) 651static void smp_send_nmi_allbutself(void)
652{ 652{
653 cpumask_t mask = cpu_online_map; 653 send_IPI_allbutself(NMI_VECTOR);
654 cpu_clear(safe_smp_processor_id(), mask);
655 if (!cpus_empty(mask))
656 send_IPI_mask(mask, NMI_VECTOR);
657} 654}
658 655
659static struct notifier_block crash_nmi_nb = { 656static struct notifier_block crash_nmi_nb = {
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index ae0c0d3bb770..0b63b08e7530 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -152,6 +152,11 @@ void __init setup_per_cpu_areas(void)
152 old_size = PERCPU_ENOUGH_ROOM; 152 old_size = PERCPU_ENOUGH_ROOM;
153 align = max_t(unsigned long, PAGE_SIZE, align); 153 align = max_t(unsigned long, PAGE_SIZE, align);
154 size = roundup(old_size, align); 154 size = roundup(old_size, align);
155
156 printk(KERN_INFO
157 "NR_CPUS:%d nr_cpumask_bits:%d nr_cpu_ids:%d nr_node_ids:%d\n",
158 NR_CPUS, nr_cpumask_bits, nr_cpu_ids, nr_node_ids);
159
155 printk(KERN_INFO "PERCPU: Allocating %zd bytes of per cpu data\n", 160 printk(KERN_INFO "PERCPU: Allocating %zd bytes of per cpu data\n",
156 size); 161 size);
157 162
@@ -168,24 +173,24 @@ void __init setup_per_cpu_areas(void)
168 "cpu %d has no node %d or node-local memory\n", 173 "cpu %d has no node %d or node-local memory\n",
169 cpu, node); 174 cpu, node);
170 if (ptr) 175 if (ptr)
171 printk(KERN_DEBUG "per cpu data for cpu%d at %016lx\n", 176 printk(KERN_DEBUG
177 "per cpu data for cpu%d at %016lx\n",
172 cpu, __pa(ptr)); 178 cpu, __pa(ptr));
173 } 179 }
174 else { 180 else {
175 ptr = __alloc_bootmem_node(NODE_DATA(node), size, align, 181 ptr = __alloc_bootmem_node(NODE_DATA(node), size, align,
176 __pa(MAX_DMA_ADDRESS)); 182 __pa(MAX_DMA_ADDRESS));
177 if (ptr) 183 if (ptr)
178 printk(KERN_DEBUG "per cpu data for cpu%d on node%d at %016lx\n", 184 printk(KERN_DEBUG
179 cpu, node, __pa(ptr)); 185 "per cpu data for cpu%d on node%d "
186 "at %016lx\n",
187 cpu, node, __pa(ptr));
180 } 188 }
181#endif 189#endif
182 per_cpu_offset(cpu) = ptr - __per_cpu_start; 190 per_cpu_offset(cpu) = ptr - __per_cpu_start;
183 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start); 191 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
184 } 192 }
185 193
186 printk(KERN_DEBUG "NR_CPUS: %d, nr_cpu_ids: %d, nr_node_ids %d\n",
187 NR_CPUS, nr_cpu_ids, nr_node_ids);
188
189 /* Setup percpu data maps */ 194 /* Setup percpu data maps */
190 setup_per_cpu_maps(); 195 setup_per_cpu_maps();
191 196
@@ -282,7 +287,7 @@ static void __cpuinit numa_set_cpumask(int cpu, int enable)
282 else 287 else
283 cpu_clear(cpu, *mask); 288 cpu_clear(cpu, *mask);
284 289
285 cpulist_scnprintf(buf, sizeof(buf), *mask); 290 cpulist_scnprintf(buf, sizeof(buf), mask);
286 printk(KERN_DEBUG "%s cpu %d node %d: mask now %s\n", 291 printk(KERN_DEBUG "%s cpu %d node %d: mask now %s\n",
287 enable? "numa_add_cpu":"numa_remove_cpu", cpu, node, buf); 292 enable? "numa_add_cpu":"numa_remove_cpu", cpu, node, buf);
288 } 293 }
diff --git a/arch/x86/kernel/smp.c b/arch/x86/kernel/smp.c
index 7e558db362c1..beea2649a240 100644
--- a/arch/x86/kernel/smp.c
+++ b/arch/x86/kernel/smp.c
@@ -118,22 +118,22 @@ static void native_smp_send_reschedule(int cpu)
118 WARN_ON(1); 118 WARN_ON(1);
119 return; 119 return;
120 } 120 }
121 send_IPI_mask(cpumask_of_cpu(cpu), RESCHEDULE_VECTOR); 121 send_IPI_mask(cpumask_of(cpu), RESCHEDULE_VECTOR);
122} 122}
123 123
124void native_send_call_func_single_ipi(int cpu) 124void native_send_call_func_single_ipi(int cpu)
125{ 125{
126 send_IPI_mask(cpumask_of_cpu(cpu), CALL_FUNCTION_SINGLE_VECTOR); 126 send_IPI_mask(cpumask_of(cpu), CALL_FUNCTION_SINGLE_VECTOR);
127} 127}
128 128
129void native_send_call_func_ipi(cpumask_t mask) 129void native_send_call_func_ipi(const struct cpumask *mask)
130{ 130{
131 cpumask_t allbutself; 131 cpumask_t allbutself;
132 132
133 allbutself = cpu_online_map; 133 allbutself = cpu_online_map;
134 cpu_clear(smp_processor_id(), allbutself); 134 cpu_clear(smp_processor_id(), allbutself);
135 135
136 if (cpus_equal(mask, allbutself) && 136 if (cpus_equal(*mask, allbutself) &&
137 cpus_equal(cpu_online_map, cpu_callout_map)) 137 cpus_equal(cpu_online_map, cpu_callout_map))
138 send_IPI_allbutself(CALL_FUNCTION_VECTOR); 138 send_IPI_allbutself(CALL_FUNCTION_VECTOR);
139 else 139 else
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index f8500c969442..31869bf5fabd 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -102,14 +102,8 @@ EXPORT_SYMBOL(smp_num_siblings);
102/* Last level cache ID of each logical CPU */ 102/* Last level cache ID of each logical CPU */
103DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID; 103DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID;
104 104
105/* bitmap of online cpus */
106cpumask_t cpu_online_map __read_mostly;
107EXPORT_SYMBOL(cpu_online_map);
108
109cpumask_t cpu_callin_map; 105cpumask_t cpu_callin_map;
110cpumask_t cpu_callout_map; 106cpumask_t cpu_callout_map;
111cpumask_t cpu_possible_map;
112EXPORT_SYMBOL(cpu_possible_map);
113 107
114/* representing HT siblings of each logical CPU */ 108/* representing HT siblings of each logical CPU */
115DEFINE_PER_CPU(cpumask_t, cpu_sibling_map); 109DEFINE_PER_CPU(cpumask_t, cpu_sibling_map);
@@ -1260,6 +1254,15 @@ void __init native_smp_cpus_done(unsigned int max_cpus)
1260 check_nmi_watchdog(); 1254 check_nmi_watchdog();
1261} 1255}
1262 1256
1257static int __initdata setup_possible_cpus = -1;
1258static int __init _setup_possible_cpus(char *str)
1259{
1260 get_option(&str, &setup_possible_cpus);
1261 return 0;
1262}
1263early_param("possible_cpus", _setup_possible_cpus);
1264
1265
1263/* 1266/*
1264 * cpu_possible_map should be static, it cannot change as cpu's 1267 * cpu_possible_map should be static, it cannot change as cpu's
1265 * are onlined, or offlined. The reason is per-cpu data-structures 1268 * are onlined, or offlined. The reason is per-cpu data-structures
@@ -1272,7 +1275,7 @@ void __init native_smp_cpus_done(unsigned int max_cpus)
1272 * 1275 *
1273 * Three ways to find out the number of additional hotplug CPUs: 1276 * Three ways to find out the number of additional hotplug CPUs:
1274 * - If the BIOS specified disabled CPUs in ACPI/mptables use that. 1277 * - If the BIOS specified disabled CPUs in ACPI/mptables use that.
1275 * - The user can overwrite it with additional_cpus=NUM 1278 * - The user can overwrite it with possible_cpus=NUM
1276 * - Otherwise don't reserve additional CPUs. 1279 * - Otherwise don't reserve additional CPUs.
1277 * We do this because additional CPUs waste a lot of memory. 1280 * We do this because additional CPUs waste a lot of memory.
1278 * -AK 1281 * -AK
@@ -1285,9 +1288,17 @@ __init void prefill_possible_map(void)
1285 if (!num_processors) 1288 if (!num_processors)
1286 num_processors = 1; 1289 num_processors = 1;
1287 1290
1288 possible = num_processors + disabled_cpus; 1291 if (setup_possible_cpus == -1)
1289 if (possible > NR_CPUS) 1292 possible = num_processors + disabled_cpus;
1290 possible = NR_CPUS; 1293 else
1294 possible = setup_possible_cpus;
1295
1296 if (possible > CONFIG_NR_CPUS) {
1297 printk(KERN_WARNING
1298 "%d Processors exceeds NR_CPUS limit of %d\n",
1299 possible, CONFIG_NR_CPUS);
1300 possible = CONFIG_NR_CPUS;
1301 }
1291 1302
1292 printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n", 1303 printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n",
1293 possible, max_t(int, possible - num_processors, 0)); 1304 possible, max_t(int, possible - num_processors, 0));
@@ -1352,7 +1363,7 @@ void cpu_disable_common(void)
1352 lock_vector_lock(); 1363 lock_vector_lock();
1353 remove_cpu_from_maps(cpu); 1364 remove_cpu_from_maps(cpu);
1354 unlock_vector_lock(); 1365 unlock_vector_lock();
1355 fixup_irqs(cpu_online_map); 1366 fixup_irqs();
1356} 1367}
1357 1368
1358int native_cpu_disable(void) 1369int native_cpu_disable(void)
diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c
index 8da059f949be..ce5054642247 100644
--- a/arch/x86/kernel/tlb_32.c
+++ b/arch/x86/kernel/tlb_32.c
@@ -163,7 +163,7 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm,
163 * We have to send the IPI only to 163 * We have to send the IPI only to
164 * CPUs affected. 164 * CPUs affected.
165 */ 165 */
166 send_IPI_mask(cpumask, INVALIDATE_TLB_VECTOR); 166 send_IPI_mask(&cpumask, INVALIDATE_TLB_VECTOR);
167 167
168 while (!cpus_empty(flush_cpumask)) 168 while (!cpus_empty(flush_cpumask))
169 /* nothing. lockup detection does not belong here */ 169 /* nothing. lockup detection does not belong here */
diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c
index 29887d7081a9..f8be6f1d2e48 100644
--- a/arch/x86/kernel/tlb_64.c
+++ b/arch/x86/kernel/tlb_64.c
@@ -191,7 +191,7 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm,
191 * We have to send the IPI only to 191 * We have to send the IPI only to
192 * CPUs affected. 192 * CPUs affected.
193 */ 193 */
194 send_IPI_mask(cpumask, INVALIDATE_TLB_VECTOR_START + sender); 194 send_IPI_mask(&cpumask, INVALIDATE_TLB_VECTOR_START + sender);
195 195
196 while (!cpus_empty(f->flush_cpumask)) 196 while (!cpus_empty(f->flush_cpumask))
197 cpu_relax(); 197 cpu_relax();
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index 141907ab6e22..2d1f4c7e4052 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -72,9 +72,6 @@
72 72
73#include "cpu/mcheck/mce.h" 73#include "cpu/mcheck/mce.h"
74 74
75DECLARE_BITMAP(used_vectors, NR_VECTORS);
76EXPORT_SYMBOL_GPL(used_vectors);
77
78asmlinkage int system_call(void); 75asmlinkage int system_call(void);
79 76
80/* Do we ignore FPU interrupts ? */ 77/* Do we ignore FPU interrupts ? */
@@ -89,6 +86,9 @@ gate_desc idt_table[256]
89 __attribute__((__section__(".data.idt"))) = { { { { 0, 0 } } }, }; 86 __attribute__((__section__(".data.idt"))) = { { { { 0, 0 } } }, };
90#endif 87#endif
91 88
89DECLARE_BITMAP(used_vectors, NR_VECTORS);
90EXPORT_SYMBOL_GPL(used_vectors);
91
92static int ignore_nmis; 92static int ignore_nmis;
93 93
94static inline void conditional_sti(struct pt_regs *regs) 94static inline void conditional_sti(struct pt_regs *regs)
@@ -941,9 +941,7 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
941 941
942void __init trap_init(void) 942void __init trap_init(void)
943{ 943{
944#ifdef CONFIG_X86_32
945 int i; 944 int i;
946#endif
947 945
948#ifdef CONFIG_EISA 946#ifdef CONFIG_EISA
949 void __iomem *p = early_ioremap(0x0FFFD9, 4); 947 void __iomem *p = early_ioremap(0x0FFFD9, 4);
@@ -1000,11 +998,15 @@ void __init trap_init(void)
1000 } 998 }
1001 999
1002 set_system_trap_gate(SYSCALL_VECTOR, &system_call); 1000 set_system_trap_gate(SYSCALL_VECTOR, &system_call);
1001#endif
1003 1002
1004 /* Reserve all the builtin and the syscall vector: */ 1003 /* Reserve all the builtin and the syscall vector: */
1005 for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++) 1004 for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
1006 set_bit(i, used_vectors); 1005 set_bit(i, used_vectors);
1007 1006
1007#ifdef CONFIG_X86_64
1008 set_bit(IA32_SYSCALL_VECTOR, used_vectors);
1009#else
1008 set_bit(SYSCALL_VECTOR, used_vectors); 1010 set_bit(SYSCALL_VECTOR, used_vectors);
1009#endif 1011#endif
1010 /* 1012 /*
diff --git a/arch/x86/kernel/vmiclock_32.c b/arch/x86/kernel/vmiclock_32.c
index 254ee07f8635..c4c1f9e09402 100644
--- a/arch/x86/kernel/vmiclock_32.c
+++ b/arch/x86/kernel/vmiclock_32.c
@@ -226,7 +226,7 @@ static void __devinit vmi_time_init_clockevent(void)
226 /* Upper bound is clockevent's use of ulong for cycle deltas. */ 226 /* Upper bound is clockevent's use of ulong for cycle deltas. */
227 evt->max_delta_ns = clockevent_delta2ns(ULONG_MAX, evt); 227 evt->max_delta_ns = clockevent_delta2ns(ULONG_MAX, evt);
228 evt->min_delta_ns = clockevent_delta2ns(1, evt); 228 evt->min_delta_ns = clockevent_delta2ns(1, evt);
229 evt->cpumask = cpumask_of_cpu(cpu); 229 evt->cpumask = cpumask_of(cpu);
230 230
231 printk(KERN_WARNING "vmi: registering clock event %s. mult=%lu shift=%u\n", 231 printk(KERN_WARNING "vmi: registering clock event %s. mult=%lu shift=%u\n",
232 evt->name, evt->mult, evt->shift); 232 evt->name, evt->mult, evt->shift);