[net-next-2.6.git] / kernel / smp.c

/*
 * Generic helpers for smp ipi calls
 *
 * (C) Jens Axboe <jens.axboe@oracle.com> 2008
 */
#include <linux/rcupdate.h>
#include <linux/rculist.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/percpu.h>
#include <linux/init.h>
#include <linux/smp.h>
#include <linux/cpu.h>

static DEFINE_PER_CPU(struct call_single_queue, call_single_queue);

static struct {
	struct list_head	queue;
	spinlock_t		lock;
} call_function __cacheline_aligned_in_smp =
	{
		.queue		= LIST_HEAD_INIT(call_function.queue),
		.lock		= __SPIN_LOCK_UNLOCKED(call_function.lock),
	};

enum {
	CSD_FLAG_LOCK		= 0x01,
};

struct call_function_data {
	struct call_single_data	csd;
	spinlock_t		lock;
	unsigned int		refs;
	cpumask_var_t		cpumask;
};

struct call_single_queue {
	struct list_head	list;
	spinlock_t		lock;
};

static DEFINE_PER_CPU(struct call_function_data, cfd_data) = {
	.lock			= __SPIN_LOCK_UNLOCKED(cfd_data.lock),
};

static int
hotplug_cfd(struct notifier_block *nfb, unsigned long action, void *hcpu)
{
	long cpu = (long)hcpu;
	struct call_function_data *cfd = &per_cpu(cfd_data, cpu);

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
		if (!zalloc_cpumask_var_node(&cfd->cpumask, GFP_KERNEL,
				cpu_to_node(cpu)))
			return NOTIFY_BAD;
		break;

#ifdef CONFIG_CPU_HOTPLUG
	case CPU_UP_CANCELED:
	case CPU_UP_CANCELED_FROZEN:

	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		free_cpumask_var(cfd->cpumask);
		break;
#endif
	};

	return NOTIFY_OK;
}

static struct notifier_block __cpuinitdata hotplug_cfd_notifier = {
	.notifier_call		= hotplug_cfd,
};

static int __cpuinit init_call_single_data(void)
{
	void *cpu = (void *)(long)smp_processor_id();
	int i;

	for_each_possible_cpu(i) {
		struct call_single_queue *q = &per_cpu(call_single_queue, i);

		spin_lock_init(&q->lock);
		INIT_LIST_HEAD(&q->list);
	}

	hotplug_cfd(&hotplug_cfd_notifier, CPU_UP_PREPARE, cpu);
	register_cpu_notifier(&hotplug_cfd_notifier);

	return 0;
}
early_initcall(init_call_single_data);

/*
 * csd_lock/csd_unlock used to serialize access to per-cpu csd resources
 *
 * For non-synchronous ipi calls the csd can still be in use by the
 * previous function call. For multi-cpu calls its even more interesting
 * as we'll have to ensure no other cpu is observing our csd.
 */
static void csd_lock_wait(struct call_single_data *data)
{
	while (data->flags & CSD_FLAG_LOCK)
		cpu_relax();
}

static void csd_lock(struct call_single_data *data)
{
	csd_lock_wait(data);
	data->flags = CSD_FLAG_LOCK;

	/*
	 * prevent CPU from reordering the above assignment
	 * to ->flags with any subsequent assignments to other
	 * fields of the specified call_single_data structure:
	 */
	smp_mb();
}

static void csd_unlock(struct call_single_data *data)
{
	WARN_ON(!(data->flags & CSD_FLAG_LOCK));

	/*
	 * ensure we're all done before releasing data:
	 */
	smp_mb();

	data->flags &= ~CSD_FLAG_LOCK;
}

/*
 * Insert a previously allocated call_single_data element
 * for execution on the given CPU. data must already have
 * ->func, ->info, and ->flags set.
 */
static
void generic_exec_single(int cpu, struct call_single_data *data, int wait)
{
	struct call_single_queue *dst = &per_cpu(call_single_queue, cpu);
	unsigned long flags;
	int ipi;

	spin_lock_irqsave(&dst->lock, flags);
	ipi = list_empty(&dst->list);
	list_add_tail(&data->list, &dst->list);
	spin_unlock_irqrestore(&dst->lock, flags);

	/*
	 * The list addition should be visible before sending the IPI
	 * handler locks the list to pull the entry off it because of
	 * normal cache coherency rules implied by spinlocks.
	 *
	 * If IPIs can go out of order to the cache coherency protocol
	 * in an architecture, sufficient synchronisation should be added
	 * to arch code to make it appear to obey cache coherency WRT
	 * locking and barrier primitives. Generic code isn't really
	 * equipped to do the right thing...
	 */
	if (ipi)
		arch_send_call_function_single_ipi(cpu);

	if (wait)
		csd_lock_wait(data);
}

/*
 * Invoked by arch to handle an IPI for call function. Must be called with
 * interrupts disabled.
 */
void generic_smp_call_function_interrupt(void)
{
	struct call_function_data *data;
	int cpu = get_cpu();

	/*
	 * Ensure entry is visible on call_function_queue after we have
	 * entered the IPI. See comment in smp_call_function_many.
	 * If we don't have this, then we may miss an entry on the list
	 * and never get another IPI to process it.
	 */
	smp_mb();

	/*
	 * It's ok to use list_for_each_rcu() here even though we may
	 * delete 'pos', since list_del_rcu() doesn't clear ->next
	 */
	list_for_each_entry_rcu(data, &call_function.queue, csd.list) {
		int refs;

		spin_lock(&data->lock);
		if (!cpumask_test_cpu(cpu, data->cpumask)) {
			spin_unlock(&data->lock);
			continue;
		}
		cpumask_clear_cpu(cpu, data->cpumask);
		spin_unlock(&data->lock);

		data->csd.func(data->csd.info);

		spin_lock(&data->lock);
		WARN_ON(data->refs == 0);
		refs = --data->refs;
		if (!refs) {
			spin_lock(&call_function.lock);
			list_del_rcu(&data->csd.list);
			spin_unlock(&call_function.lock);
		}
		spin_unlock(&data->lock);

		if (refs)
			continue;

		csd_unlock(&data->csd);
	}

	put_cpu();
}

/*
 * Invoked by arch to handle an IPI for call function single. Must be
 * called from the arch with interrupts disabled.
 */
void generic_smp_call_function_single_interrupt(void)
{
	struct call_single_queue *q = &__get_cpu_var(call_single_queue);
	unsigned int data_flags;
	LIST_HEAD(list);

	spin_lock(&q->lock);
	list_replace_init(&q->list, &list);
	spin_unlock(&q->lock);

	while (!list_empty(&list)) {
		struct call_single_data *data;

		data = list_entry(list.next, struct call_single_data, list);
		list_del(&data->list);

		/*
		 * 'data' can be invalid after this call if flags == 0
		 * (when called through generic_exec_single()),
		 * so save them away before making the call:
		 */
		data_flags = data->flags;

		data->func(data->info);

		/*
		 * Unlocked CSDs are valid through generic_exec_single():
		 */
		if (data_flags & CSD_FLAG_LOCK)
			csd_unlock(data);
	}
}

static DEFINE_PER_CPU(struct call_single_data, csd_data);

/*
 * smp_call_function_single - Run a function on a specific CPU
 * @func: The function to run. This must be fast and non-blocking.
 * @info: An arbitrary pointer to pass to the function.
 * @wait: If true, wait until function has completed on other CPUs.
 *
 * Returns 0 on success, else a negative status code. Note that @wait
 * will be implicitly turned on in case of allocation failures, since
 * we fall back to on-stack allocation.
 */
int smp_call_function_single(int cpu, void (*func) (void *info), void *info,
			     int wait)
{
	struct call_single_data d = {
		.flags = 0,
	};
	unsigned long flags;
	int this_cpu;
	int err = 0;

	/*
	 * prevent preemption and reschedule on another processor,
	 * as well as CPU removal
	 */
	this_cpu = get_cpu();

	/* Can deadlock when called with interrupts disabled */
	WARN_ON_ONCE(irqs_disabled() && !oops_in_progress);

	if (cpu == this_cpu) {
		local_irq_save(flags);
		func(info);
		local_irq_restore(flags);
	} else {
		if ((unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) {
			struct call_single_data *data = &d;

			if (!wait)
				data = &__get_cpu_var(csd_data);

			csd_lock(data);

			data->func = func;
			data->info = info;
			generic_exec_single(cpu, data, wait);
		} else {
			err = -ENXIO;	/* CPU not online */
		}
	}

	put_cpu();

	return err;
}
EXPORT_SYMBOL(smp_call_function_single);

/**
 * __smp_call_function_single(): Run a function on another CPU
 * @cpu: The CPU to run on.
 * @data: Pre-allocated and setup data structure
 *
 * Like smp_call_function_single(), but allow caller to pass in a
 * pre-allocated data structure. Useful for embedding @data inside
 * other structures, for instance.
 */
void __smp_call_function_single(int cpu, struct call_single_data *data,
				int wait)
{
	csd_lock(data);

	/* Can deadlock when called with interrupts disabled */
	WARN_ON_ONCE(wait && irqs_disabled() && !oops_in_progress);

	generic_exec_single(cpu, data, wait);
}

/* Deprecated: shim for archs using old arch_send_call_function_ipi API. */

#ifndef arch_send_call_function_ipi_mask
# define arch_send_call_function_ipi_mask(maskp) \
	 arch_send_call_function_ipi(*(maskp))
#endif

/**
 * smp_call_function_many(): Run a function on a set of other CPUs.
 * @mask: The set of cpus to run on (only runs on online subset).
 * @func: The function to run. This must be fast and non-blocking.
 * @info: An arbitrary pointer to pass to the function.
 * @wait: If true, wait (atomically) until function has completed
 *        on other CPUs.
 *
 * If @wait is true, then returns once @func has returned. Note that @wait
 * will be implicitly turned on in case of allocation failures, since
 * we fall back to on-stack allocation.
 *
 * You must not call this function with disabled interrupts or from a
 * hardware interrupt handler or from a bottom half handler. Preemption
 * must be disabled when calling this function.
 */
void smp_call_function_many(const struct cpumask *mask,
			    void (*func)(void *), void *info, bool wait)
{
	struct call_function_data *data;
	unsigned long flags;
	int cpu, next_cpu, this_cpu = smp_processor_id();

	/* Can deadlock when called with interrupts disabled */
	WARN_ON_ONCE(irqs_disabled() && !oops_in_progress);

	/* So, what's a CPU they want? Ignoring this one. */
	cpu = cpumask_first_and(mask, cpu_online_mask);
	if (cpu == this_cpu)
		cpu = cpumask_next_and(cpu, mask, cpu_online_mask);

	/* No online cpus?  We're done. */
	if (cpu >= nr_cpu_ids)
		return;

	/* Do we have another CPU which isn't us? */
	next_cpu = cpumask_next_and(cpu, mask, cpu_online_mask);
	if (next_cpu == this_cpu)
		next_cpu = cpumask_next_and(next_cpu, mask, cpu_online_mask);

	/* Fastpath: do that cpu by itself. */
	if (next_cpu >= nr_cpu_ids) {
		smp_call_function_single(cpu, func, info, wait);
		return;
	}

	data = &__get_cpu_var(cfd_data);
	csd_lock(&data->csd);

	spin_lock_irqsave(&data->lock, flags);
	data->csd.func = func;
	data->csd.info = info;
	cpumask_and(data->cpumask, mask, cpu_online_mask);
	cpumask_clear_cpu(this_cpu, data->cpumask);
	data->refs = cpumask_weight(data->cpumask);

	spin_lock(&call_function.lock);
	/*
	 * Place entry at the _HEAD_ of the list, so that any cpu still
	 * observing the entry in generic_smp_call_function_interrupt()
	 * will not miss any other list entries:
	 */
	list_add_rcu(&data->csd.list, &call_function.queue);
	spin_unlock(&call_function.lock);

	spin_unlock_irqrestore(&data->lock, flags);

	/*
	 * Make the list addition visible before sending the ipi.
	 * (IPIs must obey or appear to obey normal Linux cache
	 * coherency rules -- see comment in generic_exec_single).
	 */
	smp_mb();

	/* Send a message to all CPUs in the map */
	arch_send_call_function_ipi_mask(data->cpumask);

	/* Optionally wait for the CPUs to complete */
	if (wait)
		csd_lock_wait(&data->csd);
}
EXPORT_SYMBOL(smp_call_function_many);

/**
 * smp_call_function(): Run a function on all other CPUs.
 * @func: The function to run. This must be fast and non-blocking.
 * @info: An arbitrary pointer to pass to the function.
 * @wait: If true, wait (atomically) until function has completed
 *        on other CPUs.
 *
 * Returns 0.
 *
 * If @wait is true, then returns once @func has returned; otherwise
 * it returns just before the target cpu calls @func. In case of allocation
 * failure, @wait will be implicitly turned on.
 *
 * You must not call this function with disabled interrupts or from a
 * hardware interrupt handler or from a bottom half handler.
 */
int smp_call_function(void (*func)(void *), void *info, int wait)
{
	preempt_disable();
	smp_call_function_many(cpu_online_mask, func, info, wait);
	preempt_enable();

	return 0;
}
EXPORT_SYMBOL(smp_call_function);

void ipi_call_lock(void)
{
	spin_lock(&call_function.lock);
}

void ipi_call_unlock(void)
{
	spin_unlock(&call_function.lock);
}

void ipi_call_lock_irq(void)
{
	spin_lock_irq(&call_function.lock);
}

void ipi_call_unlock_irq(void)
{
	spin_unlock_irq(&call_function.lock);
}
Commit	Line	Data
3d442233 JA	1	/*
	2	* Generic helpers for smp ipi calls
	3	*
	4	* (C) Jens Axboe <jens.axboe@oracle.com> 2008
3d442233	5	*/
3d442233	6	#include <linux/rcupdate.h>
59190f42	7	#include <linux/rculist.h>
641cd4cf	8	#include <linux/kernel.h>
0b13fda1 IM	9	#include <linux/module.h>
	10	#include <linux/percpu.h>
	11	#include <linux/init.h>
3d442233	12	#include <linux/smp.h>
8969a5ed	13	#include <linux/cpu.h>
3d442233 JA	14
3d442233 JA	15	static DEFINE_PER_CPU(struct call_single_queue, call_single_queue);
8969a5ed PZ	16
	17	static struct {
	18	struct list_head queue;
	19	spinlock_t lock;
0b13fda1 IM	20	} call_function __cacheline_aligned_in_smp =
	21	{
	22	.queue = LIST_HEAD_INIT(call_function.queue),
	23	.lock = __SPIN_LOCK_UNLOCKED(call_function.lock),
	24	};
3d442233 JA	25
3d442233 JA	26	enum {
6e275637	27	CSD_FLAG_LOCK = 0x01,
3d442233 JA	28	};
	29
	30	struct call_function_data {
0b13fda1 IM	31	struct call_single_data csd;
	32	spinlock_t lock;
	33	unsigned int refs;
	34	cpumask_var_t cpumask;
3d442233 JA	35	};
	36
	37	struct call_single_queue {
0b13fda1 IM	38	struct list_head list;
0b13fda1 IM	39	spinlock_t lock;
3d442233 JA	40	};
3d442233 JA	41
8969a5ed	42	static DEFINE_PER_CPU(struct call_function_data, cfd_data) = {
0b13fda1	43	.lock = __SPIN_LOCK_UNLOCKED(cfd_data.lock),
8969a5ed PZ	44	};
	45
	46	static int
	47	hotplug_cfd(struct notifier_block nfb, unsigned long action, void hcpu)
	48	{
	49	long cpu = (long)hcpu;
	50	struct call_function_data *cfd = &per_cpu(cfd_data, cpu);
	51
	52	switch (action) {
	53	case CPU_UP_PREPARE:
	54	case CPU_UP_PREPARE_FROZEN:
eaa95840	55	if (!zalloc_cpumask_var_node(&cfd->cpumask, GFP_KERNEL,
8969a5ed PZ	56	cpu_to_node(cpu)))
	57	return NOTIFY_BAD;
	58	break;
	59
	60	#ifdef CONFIG_CPU_HOTPLUG
	61	case CPU_UP_CANCELED:
	62	case CPU_UP_CANCELED_FROZEN:
	63
	64	case CPU_DEAD:
	65	case CPU_DEAD_FROZEN:
	66	free_cpumask_var(cfd->cpumask);
	67	break;
	68	#endif
	69	};
	70
	71	return NOTIFY_OK;
	72	}
	73
	74	static struct notifier_block __cpuinitdata hotplug_cfd_notifier = {
0b13fda1	75	.notifier_call = hotplug_cfd,
8969a5ed PZ	76	};
8969a5ed PZ	77
7babe8db	78	static int __cpuinit init_call_single_data(void)
3d442233	79	{
8969a5ed	80	void cpu = (void )(long)smp_processor_id();
3d442233 JA	81	int i;
	82
	83	for_each_possible_cpu(i) {
	84	struct call_single_queue *q = &per_cpu(call_single_queue, i);
	85
	86	spin_lock_init(&q->lock);
	87	INIT_LIST_HEAD(&q->list);
	88	}
8969a5ed PZ	89
	90	hotplug_cfd(&hotplug_cfd_notifier, CPU_UP_PREPARE, cpu);
	91	register_cpu_notifier(&hotplug_cfd_notifier);
	92
7babe8db	93	return 0;
3d442233	94	}
7babe8db	95	early_initcall(init_call_single_data);
3d442233	96
8969a5ed PZ	97	/*
	98	* csd_lock/csd_unlock used to serialize access to per-cpu csd resources
	99	*
0b13fda1 IM	100	* For non-synchronous ipi calls the csd can still be in use by the
	101	* previous function call. For multi-cpu calls its even more interesting
	102	* as we'll have to ensure no other cpu is observing our csd.
8969a5ed	103	*/
6e275637	104	static void csd_lock_wait(struct call_single_data *data)
8969a5ed PZ	105	{
	106	while (data->flags & CSD_FLAG_LOCK)
	107	cpu_relax();
6e275637 PZ	108	}
	109
	110	static void csd_lock(struct call_single_data *data)
	111	{
	112	csd_lock_wait(data);
8969a5ed PZ	113	data->flags = CSD_FLAG_LOCK;
	114
	115	/*
0b13fda1 IM	116	* prevent CPU from reordering the above assignment
	117	* to ->flags with any subsequent assignments to other
	118	* fields of the specified call_single_data structure:
8969a5ed	119	*/
8969a5ed PZ	120	smp_mb();
	121	}
	122
	123	static void csd_unlock(struct call_single_data *data)
	124	{
	125	WARN_ON(!(data->flags & CSD_FLAG_LOCK));
0b13fda1	126
8969a5ed	127	/*
0b13fda1	128	* ensure we're all done before releasing data:
8969a5ed PZ	129	*/
8969a5ed PZ	130	smp_mb();
0b13fda1	131
8969a5ed	132	data->flags &= ~CSD_FLAG_LOCK;
3d442233 JA	133	}
	134
	135	/*
0b13fda1 IM	136	* Insert a previously allocated call_single_data element
	137	* for execution on the given CPU. data must already have
	138	* ->func, ->info, and ->flags set.
3d442233	139	*/
6e275637 PZ	140	static
6e275637 PZ	141	void generic_exec_single(int cpu, struct call_single_data *data, int wait)
3d442233 JA	142	{
3d442233 JA	143	struct call_single_queue *dst = &per_cpu(call_single_queue, cpu);
3d442233	144	unsigned long flags;
6e275637	145	int ipi;
3d442233 JA	146
	147	spin_lock_irqsave(&dst->lock, flags);
	148	ipi = list_empty(&dst->list);
	149	list_add_tail(&data->list, &dst->list);
	150	spin_unlock_irqrestore(&dst->lock, flags);
	151
561920a0	152	/*
15d0d3b3 NP	153	* The list addition should be visible before sending the IPI
	154	* handler locks the list to pull the entry off it because of
	155	* normal cache coherency rules implied by spinlocks.
	156	*
	157	* If IPIs can go out of order to the cache coherency protocol
	158	* in an architecture, sufficient synchronisation should be added
	159	* to arch code to make it appear to obey cache coherency WRT
0b13fda1 IM	160	* locking and barrier primitives. Generic code isn't really
0b13fda1 IM	161	* equipped to do the right thing...
561920a0	162	*/
3d442233 JA	163	if (ipi)
	164	arch_send_call_function_single_ipi(cpu);
	165
	166	if (wait)
6e275637	167	csd_lock_wait(data);
3d442233 JA	168	}
	169
	170	/*
	171	* Invoked by arch to handle an IPI for call function. Must be called with
	172	* interrupts disabled.
	173	*/
	174	void generic_smp_call_function_interrupt(void)
	175	{
	176	struct call_function_data *data;
	177	int cpu = get_cpu();
	178
15d0d3b3 NP	179	/*
	180	* Ensure entry is visible on call_function_queue after we have
	181	* entered the IPI. See comment in smp_call_function_many.
	182	* If we don't have this, then we may miss an entry on the list
	183	* and never get another IPI to process it.
	184	*/
	185	smp_mb();
	186
3d442233	187	/*
0b13fda1 IM	188	* It's ok to use list_for_each_rcu() here even though we may
0b13fda1 IM	189	* delete 'pos', since list_del_rcu() doesn't clear ->next
3d442233	190	*/
8969a5ed	191	list_for_each_entry_rcu(data, &call_function.queue, csd.list) {
3d442233 JA	192	int refs;
3d442233 JA	193
8969a5ed PZ	194	spin_lock(&data->lock);
	195	if (!cpumask_test_cpu(cpu, data->cpumask)) {
	196	spin_unlock(&data->lock);
3d442233	197	continue;
8969a5ed PZ	198	}
	199	cpumask_clear_cpu(cpu, data->cpumask);
	200	spin_unlock(&data->lock);
3d442233 JA	201
	202	data->csd.func(data->csd.info);
	203
	204	spin_lock(&data->lock);
3d442233	205	WARN_ON(data->refs == 0);
8969a5ed PZ	206	refs = --data->refs;
	207	if (!refs) {
	208	spin_lock(&call_function.lock);
	209	list_del_rcu(&data->csd.list);
	210	spin_unlock(&call_function.lock);
	211	}
3d442233 JA	212	spin_unlock(&data->lock);
	213
	214	if (refs)
	215	continue;
	216
8969a5ed	217	csd_unlock(&data->csd);
3d442233	218	}
3d442233 JA	219
	220	put_cpu();
	221	}
	222
	223	/*
0b13fda1 IM	224	* Invoked by arch to handle an IPI for call function single. Must be
0b13fda1 IM	225	* called from the arch with interrupts disabled.
3d442233 JA	226	*/
	227	void generic_smp_call_function_single_interrupt(void)
	228	{
	229	struct call_single_queue *q = &__get_cpu_var(call_single_queue);
15d0d3b3	230	unsigned int data_flags;
0b13fda1	231	LIST_HEAD(list);
3d442233	232
15d0d3b3 NP	233	spin_lock(&q->lock);
	234	list_replace_init(&q->list, &list);
	235	spin_unlock(&q->lock);
3d442233	236
15d0d3b3 NP	237	while (!list_empty(&list)) {
15d0d3b3 NP	238	struct call_single_data *data;
3d442233	239
0b13fda1	240	data = list_entry(list.next, struct call_single_data, list);
15d0d3b3	241	list_del(&data->list);
3d442233	242
3d442233	243	/*
0b13fda1 IM	244	* 'data' can be invalid after this call if flags == 0
	245	* (when called through generic_exec_single()),
	246	* so save them away before making the call:
3d442233	247	*/
15d0d3b3 NP	248	data_flags = data->flags;
	249
	250	data->func(data->info);
	251
8969a5ed	252	/*
0b13fda1	253	* Unlocked CSDs are valid through generic_exec_single():
8969a5ed PZ	254	*/
	255	if (data_flags & CSD_FLAG_LOCK)
	256	csd_unlock(data);
3d442233 JA	257	}
	258	}
	259
d7240b98 SR	260	static DEFINE_PER_CPU(struct call_single_data, csd_data);
d7240b98 SR	261
3d442233 JA	262	/*
	263	* smp_call_function_single - Run a function on a specific CPU
	264	* @func: The function to run. This must be fast and non-blocking.
	265	* @info: An arbitrary pointer to pass to the function.
3d442233 JA	266	* @wait: If true, wait until function has completed on other CPUs.
	267	*
	268	* Returns 0 on success, else a negative status code. Note that @wait
	269	* will be implicitly turned on in case of allocation failures, since
	270	* we fall back to on-stack allocation.
	271	*/
	272	int smp_call_function_single(int cpu, void (func) (void info), void *info,
8691e5a8	273	int wait)
3d442233	274	{
8969a5ed PZ	275	struct call_single_data d = {
	276	.flags = 0,
	277	};
3d442233	278	unsigned long flags;
0b13fda1	279	int this_cpu;
f73be6de	280	int err = 0;
3d442233	281
0b13fda1 IM	282	/*
	283	* prevent preemption and reschedule on another processor,
	284	* as well as CPU removal
	285	*/
	286	this_cpu = get_cpu();
	287
3d442233	288	/* Can deadlock when called with interrupts disabled */
641cd4cf	289	WARN_ON_ONCE(irqs_disabled() && !oops_in_progress);
3d442233	290
0b13fda1	291	if (cpu == this_cpu) {
3d442233 JA	292	local_irq_save(flags);
	293	func(info);
	294	local_irq_restore(flags);
0b13fda1 IM	295	} else {
	296	if ((unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) {
	297	struct call_single_data *data = &d;
3d442233	298
0b13fda1 IM	299	if (!wait)
0b13fda1 IM	300	data = &__get_cpu_var(csd_data);
6e275637	301
0b13fda1	302	csd_lock(data);
3d442233	303
0b13fda1 IM	304	data->func = func;
	305	data->info = info;
	306	generic_exec_single(cpu, data, wait);
	307	} else {
	308	err = -ENXIO; /* CPU not online */
	309	}
3d442233 JA	310	}
	311
	312	put_cpu();
0b13fda1	313
f73be6de	314	return err;
3d442233 JA	315	}
	316	EXPORT_SYMBOL(smp_call_function_single);
	317
	318	/**
	319	* __smp_call_function_single(): Run a function on another CPU
	320	* @cpu: The CPU to run on.
	321	* @data: Pre-allocated and setup data structure
	322	*
0b13fda1 IM	323	* Like smp_call_function_single(), but allow caller to pass in a
	324	* pre-allocated data structure. Useful for embedding @data inside
	325	* other structures, for instance.
3d442233	326	*/
6e275637 PZ	327	void __smp_call_function_single(int cpu, struct call_single_data *data,
6e275637 PZ	328	int wait)
3d442233	329	{
6e275637 PZ	330	csd_lock(data);
6e275637 PZ	331
3d442233	332	/* Can deadlock when called with interrupts disabled */
641cd4cf	333	WARN_ON_ONCE(wait && irqs_disabled() && !oops_in_progress);
3d442233	334
6e275637	335	generic_exec_single(cpu, data, wait);
3d442233 JA	336	}
3d442233 JA	337
0b13fda1 IM	338	/* Deprecated: shim for archs using old arch_send_call_function_ipi API. */
0b13fda1 IM	339
ce47d974	340	#ifndef arch_send_call_function_ipi_mask
0b13fda1 IM	341	# define arch_send_call_function_ipi_mask(maskp) \
0b13fda1 IM	342	arch_send_call_function_ipi(*(maskp))
ce47d974 RR	343	#endif
ce47d974 RR	344
3d442233	345	/**
54b11e6d RR	346	* smp_call_function_many(): Run a function on a set of other CPUs.
54b11e6d RR	347	* @mask: The set of cpus to run on (only runs on online subset).
3d442233 JA	348	* @func: The function to run. This must be fast and non-blocking.
3d442233 JA	349	* @info: An arbitrary pointer to pass to the function.
0b13fda1 IM	350	* @wait: If true, wait (atomically) until function has completed
0b13fda1 IM	351	* on other CPUs.
3d442233	352	*
3d442233 JA	353	* If @wait is true, then returns once @func has returned. Note that @wait
	354	* will be implicitly turned on in case of allocation failures, since
	355	* we fall back to on-stack allocation.
	356	*
	357	* You must not call this function with disabled interrupts or from a
	358	* hardware interrupt handler or from a bottom half handler. Preemption
	359	* must be disabled when calling this function.
	360	*/
54b11e6d	361	void smp_call_function_many(const struct cpumask *mask,
0b13fda1	362	void (func)(void ), void *info, bool wait)
3d442233	363	{
54b11e6d	364	struct call_function_data *data;
3d442233	365	unsigned long flags;
0b13fda1	366	int cpu, next_cpu, this_cpu = smp_processor_id();
3d442233 JA	367
3d442233 JA	368	/* Can deadlock when called with interrupts disabled */
641cd4cf	369	WARN_ON_ONCE(irqs_disabled() && !oops_in_progress);
3d442233	370
0b13fda1	371	/* So, what's a CPU they want? Ignoring this one. */
54b11e6d	372	cpu = cpumask_first_and(mask, cpu_online_mask);
0b13fda1	373	if (cpu == this_cpu)
54b11e6d	374	cpu = cpumask_next_and(cpu, mask, cpu_online_mask);
0b13fda1	375
54b11e6d RR	376	/* No online cpus? We're done. */
	377	if (cpu >= nr_cpu_ids)
	378	return;
	379
	380	/* Do we have another CPU which isn't us? */
	381	next_cpu = cpumask_next_and(cpu, mask, cpu_online_mask);
0b13fda1	382	if (next_cpu == this_cpu)
54b11e6d RR	383	next_cpu = cpumask_next_and(next_cpu, mask, cpu_online_mask);
	384
	385	/* Fastpath: do that cpu by itself. */
	386	if (next_cpu >= nr_cpu_ids) {
	387	smp_call_function_single(cpu, func, info, wait);
	388	return;
3d442233 JA	389	}
3d442233 JA	390
8969a5ed PZ	391	data = &__get_cpu_var(cfd_data);
8969a5ed PZ	392	csd_lock(&data->csd);
3d442233	393
8969a5ed	394	spin_lock_irqsave(&data->lock, flags);
3d442233 JA	395	data->csd.func = func;
3d442233 JA	396	data->csd.info = info;
8969a5ed	397	cpumask_and(data->cpumask, mask, cpu_online_mask);
0b13fda1	398	cpumask_clear_cpu(this_cpu, data->cpumask);
8969a5ed	399	data->refs = cpumask_weight(data->cpumask);
3d442233	400
8969a5ed PZ	401	spin_lock(&call_function.lock);
	402	/*
	403	* Place entry at the _HEAD_ of the list, so that any cpu still
0b13fda1 IM	404	* observing the entry in generic_smp_call_function_interrupt()
0b13fda1 IM	405	* will not miss any other list entries:
8969a5ed PZ	406	*/
	407	list_add_rcu(&data->csd.list, &call_function.queue);
	408	spin_unlock(&call_function.lock);
0b13fda1	409
8969a5ed	410	spin_unlock_irqrestore(&data->lock, flags);
3d442233	411
561920a0 SS	412	/*
561920a0 SS	413	* Make the list addition visible before sending the ipi.
0b13fda1 IM	414	* (IPIs must obey or appear to obey normal Linux cache
0b13fda1 IM	415	* coherency rules -- see comment in generic_exec_single).
561920a0 SS	416	*/
	417	smp_mb();
	418
3d442233	419	/* Send a message to all CPUs in the map */
8969a5ed	420	arch_send_call_function_ipi_mask(data->cpumask);
3d442233	421
0b13fda1	422	/* Optionally wait for the CPUs to complete */
54b11e6d	423	if (wait)
6e275637	424	csd_lock_wait(&data->csd);
3d442233	425	}
54b11e6d	426	EXPORT_SYMBOL(smp_call_function_many);
3d442233 JA	427
	428	/**
	429	* smp_call_function(): Run a function on all other CPUs.
	430	* @func: The function to run. This must be fast and non-blocking.
	431	* @info: An arbitrary pointer to pass to the function.
0b13fda1 IM	432	* @wait: If true, wait (atomically) until function has completed
0b13fda1 IM	433	* on other CPUs.
3d442233	434	*
54b11e6d	435	* Returns 0.
3d442233 JA	436	*
	437	* If @wait is true, then returns once @func has returned; otherwise
	438	* it returns just before the target cpu calls @func. In case of allocation
	439	* failure, @wait will be implicitly turned on.
	440	*
	441	* You must not call this function with disabled interrupts or from a
	442	* hardware interrupt handler or from a bottom half handler.
	443	*/
8691e5a8	444	int smp_call_function(void (func)(void ), void *info, int wait)
3d442233	445	{
3d442233	446	preempt_disable();
54b11e6d	447	smp_call_function_many(cpu_online_mask, func, info, wait);
3d442233	448	preempt_enable();
0b13fda1	449
54b11e6d	450	return 0;
3d442233 JA	451	}
	452	EXPORT_SYMBOL(smp_call_function);
	453
	454	void ipi_call_lock(void)
	455	{
8969a5ed	456	spin_lock(&call_function.lock);
3d442233 JA	457	}
	458
	459	void ipi_call_unlock(void)
	460	{
8969a5ed	461	spin_unlock(&call_function.lock);
3d442233 JA	462	}
	463
	464	void ipi_call_lock_irq(void)
	465	{
8969a5ed	466	spin_lock_irq(&call_function.lock);
3d442233 JA	467	}
	468
	469	void ipi_call_unlock_irq(void)
	470	{
8969a5ed	471	spin_unlock_irq(&call_function.lock);
3d442233	472	}