[deliverable/linux.git] / arch / x86 / kernel / smpboot.c

#include <linux/init.h>
#include <linux/smp.h>
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/percpu.h>
#include <linux/bootmem.h>

#include <asm/nmi.h>
#include <asm/irq.h>
#include <asm/smp.h>
#include <asm/cpu.h>
#include <asm/numa.h>

/* Number of siblings per CPU package */
int smp_num_siblings = 1;
EXPORT_SYMBOL(smp_num_siblings);

/* Last level cache ID of each logical CPU */
DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID;

/* bitmap of online cpus */
cpumask_t cpu_online_map __read_mostly;
EXPORT_SYMBOL(cpu_online_map);

cpumask_t cpu_callin_map;
cpumask_t cpu_callout_map;
cpumask_t cpu_possible_map;
EXPORT_SYMBOL(cpu_possible_map);

/* representing HT siblings of each logical CPU */
DEFINE_PER_CPU(cpumask_t, cpu_sibling_map);
EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);

/* representing HT and core siblings of each logical CPU */
DEFINE_PER_CPU(cpumask_t, cpu_core_map);
EXPORT_PER_CPU_SYMBOL(cpu_core_map);

/* Per CPU bogomips and other parameters */
DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
EXPORT_PER_CPU_SYMBOL(cpu_info);

/* ready for x86_64, no harm for x86, since it will overwrite after alloc */
unsigned char *trampoline_base = __va(SMP_TRAMPOLINE_BASE);

/* representing cpus for which sibling maps can be computed */
static cpumask_t cpu_sibling_setup_map;

#ifdef CONFIG_X86_32
/* Set if we find a B stepping CPU */
int __cpuinitdata smp_b_stepping;
#endif

static void __cpuinit smp_apply_quirks(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_X86_32
	/*
	 * Mask B, Pentium, but not Pentium MMX
	 */
	if (c->x86_vendor == X86_VENDOR_INTEL &&
	    c->x86 == 5 &&
	    c->x86_mask >= 1 && c->x86_mask <= 4 &&
	    c->x86_model <= 3)
		/*
		 * Remember we have B step Pentia with bugs
		 */
		smp_b_stepping = 1;

	/*
	 * Certain Athlons might work (for various values of 'work') in SMP
	 * but they are not certified as MP capable.
	 */
	if ((c->x86_vendor == X86_VENDOR_AMD) && (c->x86 == 6)) {

		if (num_possible_cpus() == 1)
			goto valid_k7;

		/* Athlon 660/661 is valid. */
		if ((c->x86_model == 6) && ((c->x86_mask == 0) ||
		    (c->x86_mask == 1)))
			goto valid_k7;

		/* Duron 670 is valid */
		if ((c->x86_model == 7) && (c->x86_mask == 0))
			goto valid_k7;

		/*
		 * Athlon 662, Duron 671, and Athlon >model 7 have capability
		 * bit. It's worth noting that the A5 stepping (662) of some
		 * Athlon XP's have the MP bit set.
		 * See http://www.heise.de/newsticker/data/jow-18.10.01-000 for
		 * more.
		 */
		if (((c->x86_model == 6) && (c->x86_mask >= 2)) ||
		    ((c->x86_model == 7) && (c->x86_mask >= 1)) ||
		     (c->x86_model > 7))
			if (cpu_has_mp)
				goto valid_k7;

		/* If we get here, not a certified SMP capable AMD system. */
		add_taint(TAINT_UNSAFE_SMP);
	}

valid_k7:
	;
#endif
}

/*
 * The bootstrap kernel entry code has set these up. Save them for
 * a given CPU
 */

void __cpuinit smp_store_cpu_info(int id)
{
	struct cpuinfo_x86 *c = &cpu_data(id);

	*c = boot_cpu_data;
	c->cpu_index = id;
	if (id != 0)
		identify_secondary_cpu(c);
	smp_apply_quirks(c);
}


void __cpuinit set_cpu_sibling_map(int cpu)
{
	int i;
	struct cpuinfo_x86 *c = &cpu_data(cpu);

	cpu_set(cpu, cpu_sibling_setup_map);

	if (smp_num_siblings > 1) {
		for_each_cpu_mask(i, cpu_sibling_setup_map) {
			if (c->phys_proc_id == cpu_data(i).phys_proc_id &&
			    c->cpu_core_id == cpu_data(i).cpu_core_id) {
				cpu_set(i, per_cpu(cpu_sibling_map, cpu));
				cpu_set(cpu, per_cpu(cpu_sibling_map, i));
				cpu_set(i, per_cpu(cpu_core_map, cpu));
				cpu_set(cpu, per_cpu(cpu_core_map, i));
				cpu_set(i, c->llc_shared_map);
				cpu_set(cpu, cpu_data(i).llc_shared_map);
			}
		}
	} else {
		cpu_set(cpu, per_cpu(cpu_sibling_map, cpu));
	}

	cpu_set(cpu, c->llc_shared_map);

	if (current_cpu_data.x86_max_cores == 1) {
		per_cpu(cpu_core_map, cpu) = per_cpu(cpu_sibling_map, cpu);
		c->booted_cores = 1;
		return;
	}

	for_each_cpu_mask(i, cpu_sibling_setup_map) {
		if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
		    per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
			cpu_set(i, c->llc_shared_map);
			cpu_set(cpu, cpu_data(i).llc_shared_map);
		}
		if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
			cpu_set(i, per_cpu(cpu_core_map, cpu));
			cpu_set(cpu, per_cpu(cpu_core_map, i));
			/*
			 *  Does this new cpu bringup a new core?
			 */
			if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1) {
				/*
				 * for each core in package, increment
				 * the booted_cores for this new cpu
				 */
				if (first_cpu(per_cpu(cpu_sibling_map, i)) == i)
					c->booted_cores++;
				/*
				 * increment the core count for all
				 * the other cpus in this package
				 */
				if (i != cpu)
					cpu_data(i).booted_cores++;
			} else if (i != cpu && !c->booted_cores)
				c->booted_cores = cpu_data(i).booted_cores;
		}
	}
}

/* maps the cpu to the sched domain representing multi-core */
cpumask_t cpu_coregroup_map(int cpu)
{
	struct cpuinfo_x86 *c = &cpu_data(cpu);
	/*
	 * For perf, we return last level cache shared map.
	 * And for power savings, we return cpu_core_map
	 */
	if (sched_mc_power_savings || sched_smt_power_savings)
		return per_cpu(cpu_core_map, cpu);
	else
		return c->llc_shared_map;
}

/*
 * Currently trivial. Write the real->protected mode
 * bootstrap into the page concerned. The caller
 * has made sure it's suitably aligned.
 */

unsigned long __cpuinit setup_trampoline(void)
{
	memcpy(trampoline_base, trampoline_data,
	       trampoline_end - trampoline_data);
	return virt_to_phys(trampoline_base);
}

#ifdef CONFIG_X86_32
/*
 * We are called very early to get the low memory for the
 * SMP bootup trampoline page.
 */
void __init smp_alloc_memory(void)
{
	trampoline_base = alloc_bootmem_low_pages(PAGE_SIZE);
	/*
	 * Has to be in very low memory so we can execute
	 * real-mode AP code.
	 */
	if (__pa(trampoline_base) >= 0x9F000)
		BUG();
}
#endif

void impress_friends(void)
{
	int cpu;
	unsigned long bogosum = 0;
	/*
	 * Allow the user to impress friends.
	 */
	Dprintk("Before bogomips.\n");
	for_each_possible_cpu(cpu)
		if (cpu_isset(cpu, cpu_callout_map))
			bogosum += cpu_data(cpu).loops_per_jiffy;
	printk(KERN_INFO
		"Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
		cpus_weight(cpu_present_map),
		bogosum/(500000/HZ),
		(bogosum/(5000/HZ))%100);

	Dprintk("Before bogocount - setting activated=1.\n");
}

#ifdef CONFIG_HOTPLUG_CPU
void remove_siblinginfo(int cpu)
{
	int sibling;
	struct cpuinfo_x86 *c = &cpu_data(cpu);

	for_each_cpu_mask(sibling, per_cpu(cpu_core_map, cpu)) {
		cpu_clear(cpu, per_cpu(cpu_core_map, sibling));
		/*/
		 * last thread sibling in this cpu core going down
		 */
		if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1)
			cpu_data(sibling).booted_cores--;
	}

	for_each_cpu_mask(sibling, per_cpu(cpu_sibling_map, cpu))
		cpu_clear(cpu, per_cpu(cpu_sibling_map, sibling));
	cpus_clear(per_cpu(cpu_sibling_map, cpu));
	cpus_clear(per_cpu(cpu_core_map, cpu));
	c->phys_proc_id = 0;
	c->cpu_core_id = 0;
	cpu_clear(cpu, cpu_sibling_setup_map);
}

int additional_cpus __initdata = -1;

static __init int setup_additional_cpus(char *s)
{
	return s && get_option(&s, &additional_cpus) ? 0 : -EINVAL;
}
early_param("additional_cpus", setup_additional_cpus);

/*
 * cpu_possible_map should be static, it cannot change as cpu's
 * are onlined, or offlined. The reason is per-cpu data-structures
 * are allocated by some modules at init time, and dont expect to
 * do this dynamically on cpu arrival/departure.
 * cpu_present_map on the other hand can change dynamically.
 * In case when cpu_hotplug is not compiled, then we resort to current
 * behaviour, which is cpu_possible == cpu_present.
 * - Ashok Raj
 *
 * Three ways to find out the number of additional hotplug CPUs:
 * - If the BIOS specified disabled CPUs in ACPI/mptables use that.
 * - The user can overwrite it with additional_cpus=NUM
 * - Otherwise don't reserve additional CPUs.
 * We do this because additional CPUs waste a lot of memory.
 * -AK
 */
__init void prefill_possible_map(void)
{
	int i;
	int possible;

	if (additional_cpus == -1) {
		if (disabled_cpus > 0)
			additional_cpus = disabled_cpus;
		else
			additional_cpus = 0;
	}
	possible = num_processors + additional_cpus;
	if (possible > NR_CPUS)
		possible = NR_CPUS;

	printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n",
		possible, max_t(int, possible - num_processors, 0));

	for (i = 0; i < possible; i++)
		cpu_set(i, cpu_possible_map);
}

static void __ref remove_cpu_from_maps(int cpu)
{
	cpu_clear(cpu, cpu_online_map);
#ifdef CONFIG_X86_64
	cpu_clear(cpu, cpu_callout_map);
	cpu_clear(cpu, cpu_callin_map);
	/* was set by cpu_init() */
	clear_bit(cpu, (unsigned long *)&cpu_initialized);
	clear_node_cpumask(cpu);
#endif
}

int __cpu_disable(void)
{
	int cpu = smp_processor_id();

	/*
	 * Perhaps use cpufreq to drop frequency, but that could go
	 * into generic code.
	 *
	 * We won't take down the boot processor on i386 due to some
	 * interrupts only being able to be serviced by the BSP.
	 * Especially so if we're not using an IOAPIC	-zwane
	 */
	if (cpu == 0)
		return -EBUSY;

	if (nmi_watchdog == NMI_LOCAL_APIC)
		stop_apic_nmi_watchdog(NULL);
	clear_local_APIC();

	/*
	 * HACK:
	 * Allow any queued timer interrupts to get serviced
	 * This is only a temporary solution until we cleanup
	 * fixup_irqs as we do for IA64.
	 */
	local_irq_enable();
	mdelay(1);

	local_irq_disable();
	remove_siblinginfo(cpu);

	/* It's now safe to remove this processor from the online map */
	remove_cpu_from_maps(cpu);
	fixup_irqs(cpu_online_map);
	return 0;
}

void __cpu_die(unsigned int cpu)
{
	/* We don't do anything here: idle task is faking death itself. */
	unsigned int i;

	for (i = 0; i < 10; i++) {
		/* They ack this in play_dead by setting CPU_DEAD */
		if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
			printk(KERN_INFO "CPU %d is now offline\n", cpu);
			if (1 == num_online_cpus())
				alternatives_smp_switch(0);
			return;
		}
		msleep(100);
	}
	printk(KERN_ERR "CPU %u didn't die...\n", cpu);
}
#else /* ... !CONFIG_HOTPLUG_CPU */
int __cpu_disable(void)
{
	return -ENOSYS;
}

void __cpu_die(unsigned int cpu)
{
	/* We said "no" in __cpu_disable */
	BUG();
}
#endif

/*
 * If the BIOS enumerates physical processors before logical,
 * maxcpus=N at enumeration-time can be used to disable HT.
 */
static int __init parse_maxcpus(char *arg)
{
	extern unsigned int maxcpus;

	maxcpus = simple_strtoul(arg, NULL, 0);
	return 0;
}
early_param("maxcpus", parse_maxcpus);
Commit	Line	Data
68a1c3f8 GC	1	#include <linux/init.h>
68a1c3f8 GC	2	#include <linux/smp.h>
a355352b	3	#include <linux/module.h>
70708a18	4	#include <linux/sched.h>
69c18c15	5	#include <linux/percpu.h>
91718e8d	6	#include <linux/bootmem.h>
69c18c15 GC	7
	8	#include <asm/nmi.h>
	9	#include <asm/irq.h>
	10	#include <asm/smp.h>
	11	#include <asm/cpu.h>
	12	#include <asm/numa.h>
68a1c3f8	13
a355352b GC	14	/* Number of siblings per CPU package */
	15	int smp_num_siblings = 1;
	16	EXPORT_SYMBOL(smp_num_siblings);
	17
	18	/* Last level cache ID of each logical CPU */
	19	DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID;
	20
	21	/* bitmap of online cpus */
	22	cpumask_t cpu_online_map __read_mostly;
	23	EXPORT_SYMBOL(cpu_online_map);
	24
	25	cpumask_t cpu_callin_map;
	26	cpumask_t cpu_callout_map;
	27	cpumask_t cpu_possible_map;
	28	EXPORT_SYMBOL(cpu_possible_map);
	29
	30	/* representing HT siblings of each logical CPU */
	31	DEFINE_PER_CPU(cpumask_t, cpu_sibling_map);
	32	EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
	33
	34	/* representing HT and core siblings of each logical CPU */
	35	DEFINE_PER_CPU(cpumask_t, cpu_core_map);
	36	EXPORT_PER_CPU_SYMBOL(cpu_core_map);
	37
	38	/* Per CPU bogomips and other parameters */
	39	DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
	40	EXPORT_PER_CPU_SYMBOL(cpu_info);
768d9505	41
91718e8d GC	42	/* ready for x86_64, no harm for x86, since it will overwrite after alloc */
	43	unsigned char *trampoline_base = __va(SMP_TRAMPOLINE_BASE);
	44
768d9505 GC	45	/* representing cpus for which sibling maps can be computed */
	46	static cpumask_t cpu_sibling_setup_map;
	47
1d89a7f0 GOC	48	#ifdef CONFIG_X86_32
	49	/* Set if we find a B stepping CPU */
	50	int __cpuinitdata smp_b_stepping;
	51	#endif
	52
	53	static void __cpuinit smp_apply_quirks(struct cpuinfo_x86 *c)
	54	{
	55	#ifdef CONFIG_X86_32
	56	/*
	57	* Mask B, Pentium, but not Pentium MMX
	58	*/
	59	if (c->x86_vendor == X86_VENDOR_INTEL &&
	60	c->x86 == 5 &&
	61	c->x86_mask >= 1 && c->x86_mask <= 4 &&
	62	c->x86_model <= 3)
	63	/*
	64	* Remember we have B step Pentia with bugs
	65	*/
	66	smp_b_stepping = 1;
	67
	68	/*
	69	* Certain Athlons might work (for various values of 'work') in SMP
	70	* but they are not certified as MP capable.
	71	*/
	72	if ((c->x86_vendor == X86_VENDOR_AMD) && (c->x86 == 6)) {
	73
	74	if (num_possible_cpus() == 1)
	75	goto valid_k7;
	76
	77	/* Athlon 660/661 is valid. */
	78	if ((c->x86_model == 6) && ((c->x86_mask == 0) \|\|
	79	(c->x86_mask == 1)))
	80	goto valid_k7;
	81
	82	/* Duron 670 is valid */
	83	if ((c->x86_model == 7) && (c->x86_mask == 0))
	84	goto valid_k7;
	85
	86	/*
	87	* Athlon 662, Duron 671, and Athlon >model 7 have capability
	88	* bit. It's worth noting that the A5 stepping (662) of some
	89	* Athlon XP's have the MP bit set.
	90	* See http://www.heise.de/newsticker/data/jow-18.10.01-000 for
	91	* more.
	92	*/
	93	if (((c->x86_model == 6) && (c->x86_mask >= 2)) \|\|
	94	((c->x86_model == 7) && (c->x86_mask >= 1)) \|\|
	95	(c->x86_model > 7))
	96	if (cpu_has_mp)
	97	goto valid_k7;
	98
	99	/* If we get here, not a certified SMP capable AMD system. */
	100	add_taint(TAINT_UNSAFE_SMP);
	101	}
	102
	103	valid_k7:
	104	;
	105	#endif
	106	}
	107
	108	/*
	109	* The bootstrap kernel entry code has set these up. Save them for
	110	* a given CPU
	111	*/
112
113	void __cpuinit smp_store_cpu_info(int id)
114	{
115	struct cpuinfo_x86 *c = &cpu_data(id);
116
117	*c = boot_cpu_data;
118	c->cpu_index = id;
119	if (id != 0)
120	identify_secondary_cpu(c);
121	smp_apply_quirks(c);
122	}
123
124
768d9505 GC	125	void __cpuinit set_cpu_sibling_map(int cpu)
	126	{
	127	int i;
	128	struct cpuinfo_x86 *c = &cpu_data(cpu);
	129
	130	cpu_set(cpu, cpu_sibling_setup_map);
	131
	132	if (smp_num_siblings > 1) {
	133	for_each_cpu_mask(i, cpu_sibling_setup_map) {
	134	if (c->phys_proc_id == cpu_data(i).phys_proc_id &&
	135	c->cpu_core_id == cpu_data(i).cpu_core_id) {
	136	cpu_set(i, per_cpu(cpu_sibling_map, cpu));
	137	cpu_set(cpu, per_cpu(cpu_sibling_map, i));
	138	cpu_set(i, per_cpu(cpu_core_map, cpu));
	139	cpu_set(cpu, per_cpu(cpu_core_map, i));
	140	cpu_set(i, c->llc_shared_map);
	141	cpu_set(cpu, cpu_data(i).llc_shared_map);
	142	}
	143	}
	144	} else {
	145	cpu_set(cpu, per_cpu(cpu_sibling_map, cpu));
	146	}
	147
	148	cpu_set(cpu, c->llc_shared_map);
	149
	150	if (current_cpu_data.x86_max_cores == 1) {
	151	per_cpu(cpu_core_map, cpu) = per_cpu(cpu_sibling_map, cpu);
	152	c->booted_cores = 1;
	153	return;
	154	}
	155
	156	for_each_cpu_mask(i, cpu_sibling_setup_map) {
	157	if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
	158	per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
	159	cpu_set(i, c->llc_shared_map);
	160	cpu_set(cpu, cpu_data(i).llc_shared_map);
	161	}
	162	if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
	163	cpu_set(i, per_cpu(cpu_core_map, cpu));
	164	cpu_set(cpu, per_cpu(cpu_core_map, i));
	165	/*
	166	* Does this new cpu bringup a new core?
	167	*/
	168	if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1) {
	169	/*
	170	* for each core in package, increment
	171	* the booted_cores for this new cpu
	172	*/
	173	if (first_cpu(per_cpu(cpu_sibling_map, i)) == i)
	174	c->booted_cores++;
	175	/*
	176	* increment the core count for all
	177	* the other cpus in this package
	178	*/
	179	if (i != cpu)
	180	cpu_data(i).booted_cores++;
	181	} else if (i != cpu && !c->booted_cores)
	182	c->booted_cores = cpu_data(i).booted_cores;
	183	}
	184	}
	185	}
	186
70708a18 GC	187	/* maps the cpu to the sched domain representing multi-core */
	188	cpumask_t cpu_coregroup_map(int cpu)
	189	{
	190	struct cpuinfo_x86 *c = &cpu_data(cpu);
	191	/*
	192	* For perf, we return last level cache shared map.
	193	* And for power savings, we return cpu_core_map
	194	*/
	195	if (sched_mc_power_savings \|\| sched_smt_power_savings)
	196	return per_cpu(cpu_core_map, cpu);
	197	else
	198	return c->llc_shared_map;
	199	}
	200
91718e8d GC	201	/*
	202	* Currently trivial. Write the real->protected mode
	203	* bootstrap into the page concerned. The caller
	204	* has made sure it's suitably aligned.
	205	*/
	206
	207	unsigned long __cpuinit setup_trampoline(void)
	208	{
	209	memcpy(trampoline_base, trampoline_data,
	210	trampoline_end - trampoline_data);
	211	return virt_to_phys(trampoline_base);
	212	}
	213
	214	#ifdef CONFIG_X86_32
	215	/*
	216	* We are called very early to get the low memory for the
	217	* SMP bootup trampoline page.
	218	*/
	219	void __init smp_alloc_memory(void)
	220	{
	221	trampoline_base = alloc_bootmem_low_pages(PAGE_SIZE);
	222	/*
	223	* Has to be in very low memory so we can execute
	224	* real-mode AP code.
	225	*/
	226	if (__pa(trampoline_base) >= 0x9F000)
	227	BUG();
	228	}
	229	#endif
70708a18	230
904541e2 GOC	231	void impress_friends(void)
	232	{
	233	int cpu;
	234	unsigned long bogosum = 0;
	235	/*
	236	* Allow the user to impress friends.
	237	*/
	238	Dprintk("Before bogomips.\n");
	239	for_each_possible_cpu(cpu)
	240	if (cpu_isset(cpu, cpu_callout_map))
	241	bogosum += cpu_data(cpu).loops_per_jiffy;
	242	printk(KERN_INFO
	243	"Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
	244	cpus_weight(cpu_present_map),
	245	bogosum/(500000/HZ),
	246	(bogosum/(5000/HZ))%100);
	247
	248	Dprintk("Before bogocount - setting activated=1.\n");
	249	}
	250
68a1c3f8	251	#ifdef CONFIG_HOTPLUG_CPU
768d9505 GC	252	void remove_siblinginfo(int cpu)
	253	{
	254	int sibling;
	255	struct cpuinfo_x86 *c = &cpu_data(cpu);
	256
	257	for_each_cpu_mask(sibling, per_cpu(cpu_core_map, cpu)) {
	258	cpu_clear(cpu, per_cpu(cpu_core_map, sibling));
	259	/*/
	260	* last thread sibling in this cpu core going down
	261	*/
	262	if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1)
	263	cpu_data(sibling).booted_cores--;
	264	}
	265
	266	for_each_cpu_mask(sibling, per_cpu(cpu_sibling_map, cpu))
	267	cpu_clear(cpu, per_cpu(cpu_sibling_map, sibling));
	268	cpus_clear(per_cpu(cpu_sibling_map, cpu));
	269	cpus_clear(per_cpu(cpu_core_map, cpu));
	270	c->phys_proc_id = 0;
	271	c->cpu_core_id = 0;
	272	cpu_clear(cpu, cpu_sibling_setup_map);
	273	}
68a1c3f8 GC	274
	275	int additional_cpus __initdata = -1;
	276
	277	static __init int setup_additional_cpus(char *s)
	278	{
	279	return s && get_option(&s, &additional_cpus) ? 0 : -EINVAL;
	280	}
	281	early_param("additional_cpus", setup_additional_cpus);
	282
	283	/*
	284	* cpu_possible_map should be static, it cannot change as cpu's
	285	* are onlined, or offlined. The reason is per-cpu data-structures
	286	* are allocated by some modules at init time, and dont expect to
	287	* do this dynamically on cpu arrival/departure.
	288	* cpu_present_map on the other hand can change dynamically.
	289	* In case when cpu_hotplug is not compiled, then we resort to current
	290	* behaviour, which is cpu_possible == cpu_present.
	291	* - Ashok Raj
	292	*
	293	* Three ways to find out the number of additional hotplug CPUs:
	294	* - If the BIOS specified disabled CPUs in ACPI/mptables use that.
	295	* - The user can overwrite it with additional_cpus=NUM
	296	* - Otherwise don't reserve additional CPUs.
	297	* We do this because additional CPUs waste a lot of memory.
	298	* -AK
	299	*/
	300	__init void prefill_possible_map(void)
	301	{
	302	int i;
	303	int possible;
	304
	305	if (additional_cpus == -1) {
	306	if (disabled_cpus > 0)
	307	additional_cpus = disabled_cpus;
	308	else
	309	additional_cpus = 0;
	310	}
	311	possible = num_processors + additional_cpus;
	312	if (possible > NR_CPUS)
	313	possible = NR_CPUS;
	314
	315	printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n",
	316	possible, max_t(int, possible - num_processors, 0));
	317
	318	for (i = 0; i < possible; i++)
	319	cpu_set(i, cpu_possible_map);
	320	}
69c18c15 GC	321
	322	static void __ref remove_cpu_from_maps(int cpu)
	323	{
	324	cpu_clear(cpu, cpu_online_map);
	325	#ifdef CONFIG_X86_64
	326	cpu_clear(cpu, cpu_callout_map);
	327	cpu_clear(cpu, cpu_callin_map);
	328	/* was set by cpu_init() */
	329	clear_bit(cpu, (unsigned long *)&cpu_initialized);
	330	clear_node_cpumask(cpu);
	331	#endif
	332	}
	333
	334	int __cpu_disable(void)
	335	{
	336	int cpu = smp_processor_id();
	337
	338	/*
	339	* Perhaps use cpufreq to drop frequency, but that could go
	340	* into generic code.
	341	*
	342	* We won't take down the boot processor on i386 due to some
	343	* interrupts only being able to be serviced by the BSP.
	344	* Especially so if we're not using an IOAPIC -zwane
	345	*/
	346	if (cpu == 0)
	347	return -EBUSY;
	348
	349	if (nmi_watchdog == NMI_LOCAL_APIC)
	350	stop_apic_nmi_watchdog(NULL);
	351	clear_local_APIC();
	352
	353	/*
	354	* HACK:
	355	* Allow any queued timer interrupts to get serviced
	356	* This is only a temporary solution until we cleanup
	357	* fixup_irqs as we do for IA64.
	358	*/
	359	local_irq_enable();
	360	mdelay(1);
	361
	362	local_irq_disable();
	363	remove_siblinginfo(cpu);
	364
	365	/* It's now safe to remove this processor from the online map */
	366	remove_cpu_from_maps(cpu);
	367	fixup_irqs(cpu_online_map);
	368	return 0;
	369	}
	370
	371	void __cpu_die(unsigned int cpu)
	372	{
	373	/* We don't do anything here: idle task is faking death itself. */
	374	unsigned int i;
	375
	376	for (i = 0; i < 10; i++) {
	377	/* They ack this in play_dead by setting CPU_DEAD */
	378	if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
	379	printk(KERN_INFO "CPU %d is now offline\n", cpu);
	380	if (1 == num_online_cpus())
	381	alternatives_smp_switch(0);
	382	return;
	383	}
	384	msleep(100);
385	}
386	printk(KERN_ERR "CPU %u didn't die...\n", cpu);
387	}
388	#else /* ... !CONFIG_HOTPLUG_CPU */
389	int __cpu_disable(void)
390	{
391	return -ENOSYS;
392	}
393
394	void __cpu_die(unsigned int cpu)
395	{
396	/* We said "no" in __cpu_disable */
397	BUG();
398	}
68a1c3f8 GC	399	#endif
68a1c3f8 GC	400
89b08200 GC	401	/*
	402	* If the BIOS enumerates physical processors before logical,
	403	* maxcpus=N at enumeration-time can be used to disable HT.
	404	*/
	405	static int __init parse_maxcpus(char *arg)
	406	{
	407	extern unsigned int maxcpus;
	408
	409	maxcpus = simple_strtoul(arg, NULL, 0);
	410	return 0;
	411	}
	412	early_param("maxcpus", parse_maxcpus);