[GitHub/MotorolaMobilityLLC/kernel-slsi.git] / kernel / smp.c

/*
 * Generic helpers for smp ipi calls
 *
 * (C) Jens Axboe <jens.axboe@oracle.com> 2008
 *
 */
#include <linux/init.h>
#include <linux/module.h>
#include <linux/percpu.h>
#include <linux/rcupdate.h>
#include <linux/rculist.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_WAIT		= 0x01,
	CSD_FLAG_LOCK		= 0x02,
};

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 (!alloc_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_wait/csd_complete are used for synchronous ipi calls
 */
static void csd_wait_prepare(struct call_single_data *data)
{
	data->flags |= CSD_FLAG_WAIT;
}

static void csd_complete(struct call_single_data *data)
{
	if (data->flags & CSD_FLAG_WAIT) {
		/*
		 * ensure we're all done before saying we are
		 */
		smp_mb();
		data->flags &= ~CSD_FLAG_WAIT;
	}
}

static void csd_wait(struct call_single_data *data)
{
	while (data->flags & CSD_FLAG_WAIT)
		cpu_relax();
}

/*
 * 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(struct call_single_data *data)
{
	while (data->flags & CSD_FLAG_LOCK)
		cpu_relax();
	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)
{
	struct call_single_queue *dst = &per_cpu(call_single_queue, cpu);
	int wait = data->flags & CSD_FLAG_WAIT, ipi;
	unsigned long flags;

	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_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_complete(&data->csd);
		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);
	LIST_HEAD(list);
	unsigned int data_flags;

	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);

		if (data_flags & CSD_FLAG_WAIT)
			csd_complete(data);

		/*
		 * 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;
	/* prevent preemption and reschedule on another processor,
	   as well as CPU removal */
	int me = get_cpu();
	int err = 0;

	/* Can deadlock when called with interrupts disabled */
	WARN_ON(irqs_disabled());

	if (cpu == me) {
		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;

		if (!wait) {
			/*
			 * We are calling a function on a single CPU
			 * and we are not going to wait for it to finish.
			 * We use a per cpu data to pass the information to
			 * that CPU. Since all callers of this code will
			 * use the same data, we must synchronize the
			 * callers to prevent a new caller from corrupting
			 * the data before the callee can access it.
			 *
			 * The CSD_FLAG_LOCK is used to let us know when
			 * the IPI handler is done with the data.
			 * The first caller will set it, and the callee
			 * will clear it. The next caller must wait for
			 * it to clear before we set it again. This
			 * will make sure the callee is done with the
			 * data before a new caller will use it.
			 */
			data = &__get_cpu_var(csd_data);
			csd_lock(data);
		} else {
			data = &d;
			csd_wait_prepare(data);
		}

		data->func = func;
		data->info = info;
		generic_exec_single(cpu, data);
	} 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)
{
	/* Can deadlock when called with interrupts disabled */
	WARN_ON((data->flags & CSD_FLAG_WAIT) && irqs_disabled());

	generic_exec_single(cpu, data);
}

/* FIXME: 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, me = smp_processor_id();

	/* Can deadlock when called with interrupts disabled */
	WARN_ON(irqs_disabled());

	/* So, what's a CPU they want?  Ignoring this one. */
	cpu = cpumask_first_and(mask, cpu_online_mask);
	if (cpu == me)
		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 == me)
		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);
	if (wait)
		csd_wait_prepare(&data->csd);

	data->csd.func = func;
	data->csd.info = info;
	cpumask_and(data->cpumask, mask, cpu_online_mask);
	cpumask_clear_cpu(me, 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_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
	5	*
	6	*/
	7	#include <linux/init.h>
	8	#include <linux/module.h>
	9	#include <linux/percpu.h>
	10	#include <linux/rcupdate.h>
59190f42	11	#include <linux/rculist.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;
	20	} call_function __cacheline_aligned_in_smp = {
	21	.queue = LIST_HEAD_INIT(call_function.queue),
	22	.lock = __SPIN_LOCK_UNLOCKED(call_function.lock),
	23	};
3d442233 JA	24
	25	enum {
	26	CSD_FLAG_WAIT = 0x01,
8969a5ed	27	CSD_FLAG_LOCK = 0x02,
3d442233 JA	28	};
	29
	30	struct call_function_data {
	31	struct call_single_data csd;
	32	spinlock_t lock;
	33	unsigned int refs;
8969a5ed	34	cpumask_var_t cpumask;
3d442233 JA	35	};
	36
	37	struct call_single_queue {
	38	struct list_head list;
	39	spinlock_t lock;
	40	};
	41
8969a5ed PZ	42	static DEFINE_PER_CPU(struct call_function_data, cfd_data) = {
	43	.lock = __SPIN_LOCK_UNLOCKED(cfd_data.lock),
	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:
	55	if (!alloc_cpumask_var_node(&cfd->cpumask, GFP_KERNEL,
	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 = {
	75	.notifier_call = hotplug_cfd,
	76	};
	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_wait/csd_complete are used for synchronous ipi calls
	99	*/
	100	static void csd_wait_prepare(struct call_single_data *data)
3d442233	101	{
8969a5ed PZ	102	data->flags \|= CSD_FLAG_WAIT;
	103	}
	104
	105	static void csd_complete(struct call_single_data *data)
	106	{
	107	if (data->flags & CSD_FLAG_WAIT) {
	108	/*
	109	* ensure we're all done before saying we are
	110	*/
	111	smp_mb();
	112	data->flags &= ~CSD_FLAG_WAIT;
	113	}
	114	}
	115
	116	static void csd_wait(struct call_single_data *data)
	117	{
	118	while (data->flags & CSD_FLAG_WAIT)
3d442233	119	cpu_relax();
8969a5ed PZ	120	}
	121
	122	/*
	123	* csd_lock/csd_unlock used to serialize access to per-cpu csd resources
	124	*
	125	* For non-synchronous ipi calls the csd can still be in use by the previous
	126	* function call. For multi-cpu calls its even more interesting as we'll have
	127	* to ensure no other cpu is observing our csd.
	128	*/
	129	static void csd_lock(struct call_single_data *data)
	130	{
	131	while (data->flags & CSD_FLAG_LOCK)
	132	cpu_relax();
	133	data->flags = CSD_FLAG_LOCK;
	134
	135	/*
	136	* prevent CPU from reordering the above assignment to ->flags
	137	* with any subsequent assignments to other fields of the
	138	* specified call_single_data structure.
	139	*/
	140
	141	smp_mb();
	142	}
	143
	144	static void csd_unlock(struct call_single_data *data)
	145	{
	146	WARN_ON(!(data->flags & CSD_FLAG_LOCK));
	147	/*
	148	* ensure we're all done before releasing data
	149	*/
	150	smp_mb();
	151	data->flags &= ~CSD_FLAG_LOCK;
3d442233 JA	152	}
	153
	154	/*
	155	* Insert a previously allocated call_single_data element for execution
	156	* on the given CPU. data must already have ->func, ->info, and ->flags set.
	157	*/
	158	static void generic_exec_single(int cpu, struct call_single_data *data)
	159	{
	160	struct call_single_queue *dst = &per_cpu(call_single_queue, cpu);
	161	int wait = data->flags & CSD_FLAG_WAIT, ipi;
	162	unsigned long flags;
	163
	164	spin_lock_irqsave(&dst->lock, flags);
	165	ipi = list_empty(&dst->list);
	166	list_add_tail(&data->list, &dst->list);
	167	spin_unlock_irqrestore(&dst->lock, flags);
	168
561920a0	169	/*
15d0d3b3 NP	170	* The list addition should be visible before sending the IPI
	171	* handler locks the list to pull the entry off it because of
	172	* normal cache coherency rules implied by spinlocks.
	173	*
	174	* If IPIs can go out of order to the cache coherency protocol
	175	* in an architecture, sufficient synchronisation should be added
	176	* to arch code to make it appear to obey cache coherency WRT
	177	* locking and barrier primitives. Generic code isn't really equipped
	178	* to do the right thing...
561920a0	179	*/
561920a0	180
3d442233 JA	181	if (ipi)
	182	arch_send_call_function_single_ipi(cpu);
	183
	184	if (wait)
8969a5ed	185	csd_wait(data);
3d442233 JA	186	}
	187
	188	/*
	189	* Invoked by arch to handle an IPI for call function. Must be called with
	190	* interrupts disabled.
	191	*/
	192	void generic_smp_call_function_interrupt(void)
	193	{
	194	struct call_function_data *data;
	195	int cpu = get_cpu();
	196
15d0d3b3 NP	197	/*
	198	* Ensure entry is visible on call_function_queue after we have
	199	* entered the IPI. See comment in smp_call_function_many.
	200	* If we don't have this, then we may miss an entry on the list
	201	* and never get another IPI to process it.
	202	*/
	203	smp_mb();
	204
3d442233 JA	205	/*
	206	* It's ok to use list_for_each_rcu() here even though we may delete
	207	* 'pos', since list_del_rcu() doesn't clear ->next
	208	*/
8969a5ed	209	list_for_each_entry_rcu(data, &call_function.queue, csd.list) {
3d442233 JA	210	int refs;
3d442233 JA	211
8969a5ed PZ	212	spin_lock(&data->lock);
	213	if (!cpumask_test_cpu(cpu, data->cpumask)) {
	214	spin_unlock(&data->lock);
3d442233	215	continue;
8969a5ed PZ	216	}
	217	cpumask_clear_cpu(cpu, data->cpumask);
	218	spin_unlock(&data->lock);
3d442233 JA	219
	220	data->csd.func(data->csd.info);
	221
	222	spin_lock(&data->lock);
3d442233	223	WARN_ON(data->refs == 0);
8969a5ed PZ	224	refs = --data->refs;
	225	if (!refs) {
	226	spin_lock(&call_function.lock);
	227	list_del_rcu(&data->csd.list);
	228	spin_unlock(&call_function.lock);
	229	}
3d442233 JA	230	spin_unlock(&data->lock);
	231
	232	if (refs)
	233	continue;
	234
8969a5ed PZ	235	csd_complete(&data->csd);
8969a5ed PZ	236	csd_unlock(&data->csd);
3d442233	237	}
3d442233 JA	238
	239	put_cpu();
	240	}
	241
	242	/*
	243	* Invoked by arch to handle an IPI for call function single. Must be called
	244	* from the arch with interrupts disabled.
	245	*/
	246	void generic_smp_call_function_single_interrupt(void)
	247	{
	248	struct call_single_queue *q = &__get_cpu_var(call_single_queue);
	249	LIST_HEAD(list);
15d0d3b3	250	unsigned int data_flags;
3d442233	251
15d0d3b3 NP	252	spin_lock(&q->lock);
	253	list_replace_init(&q->list, &list);
	254	spin_unlock(&q->lock);
3d442233	255
15d0d3b3 NP	256	while (!list_empty(&list)) {
15d0d3b3 NP	257	struct call_single_data *data;
3d442233	258
15d0d3b3 NP	259	data = list_entry(list.next, struct call_single_data,
	260	list);
	261	list_del(&data->list);
3d442233	262
3d442233	263	/*
15d0d3b3 NP	264	* 'data' can be invalid after this call if
	265	* flags == 0 (when called through
	266	* generic_exec_single(), so save them away before
	267	* making the call.
3d442233	268	*/
15d0d3b3 NP	269	data_flags = data->flags;
	270
	271	data->func(data->info);
	272
8969a5ed PZ	273	if (data_flags & CSD_FLAG_WAIT)
	274	csd_complete(data);
	275
	276	/*
	277	* Unlocked CSDs are valid through generic_exec_single()
	278	*/
	279	if (data_flags & CSD_FLAG_LOCK)
	280	csd_unlock(data);
3d442233 JA	281	}
	282	}
	283
d7240b98 SR	284	static DEFINE_PER_CPU(struct call_single_data, csd_data);
d7240b98 SR	285
3d442233 JA	286	/*
	287	* smp_call_function_single - Run a function on a specific CPU
	288	* @func: The function to run. This must be fast and non-blocking.
	289	* @info: An arbitrary pointer to pass to the function.
3d442233 JA	290	* @wait: If true, wait until function has completed on other CPUs.
	291	*
	292	* Returns 0 on success, else a negative status code. Note that @wait
	293	* will be implicitly turned on in case of allocation failures, since
	294	* we fall back to on-stack allocation.
	295	*/
	296	int smp_call_function_single(int cpu, void (func) (void info), void *info,
8691e5a8	297	int wait)
3d442233	298	{
8969a5ed PZ	299	struct call_single_data d = {
	300	.flags = 0,
	301	};
3d442233	302	unsigned long flags;
f73be6de PA	303	/* prevent preemption and reschedule on another processor,
f73be6de PA	304	as well as CPU removal */
3d442233	305	int me = get_cpu();
f73be6de	306	int err = 0;
3d442233 JA	307
	308	/* Can deadlock when called with interrupts disabled */
	309	WARN_ON(irqs_disabled());
	310
	311	if (cpu == me) {
	312	local_irq_save(flags);
	313	func(info);
	314	local_irq_restore(flags);
4f4b6c1a	315	} else if ((unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) {
d7240b98	316	struct call_single_data *data;
3d442233 JA	317
3d442233 JA	318	if (!wait) {
d7240b98 SR	319	/*
	320	* We are calling a function on a single CPU
	321	* and we are not going to wait for it to finish.
8969a5ed PZ	322	* We use a per cpu data to pass the information to
	323	* that CPU. Since all callers of this code will
	324	* use the same data, we must synchronize the
	325	* callers to prevent a new caller from corrupting
	326	* the data before the callee can access it.
d7240b98 SR	327	*
	328	* The CSD_FLAG_LOCK is used to let us know when
	329	* the IPI handler is done with the data.
	330	* The first caller will set it, and the callee
	331	* will clear it. The next caller must wait for
	332	* it to clear before we set it again. This
	333	* will make sure the callee is done with the
	334	* data before a new caller will use it.
	335	*/
8969a5ed PZ	336	data = &__get_cpu_var(csd_data);
8969a5ed PZ	337	csd_lock(data);
d7240b98	338	} else {
3d442233	339	data = &d;
8969a5ed	340	csd_wait_prepare(data);
3d442233 JA	341	}
	342
	343	data->func = func;
	344	data->info = info;
	345	generic_exec_single(cpu, data);
f73be6de PA	346	} else {
f73be6de PA	347	err = -ENXIO; /* CPU not online */
3d442233 JA	348	}
	349
	350	put_cpu();
f73be6de	351	return err;
3d442233 JA	352	}
	353	EXPORT_SYMBOL(smp_call_function_single);
	354
	355	/**
	356	* __smp_call_function_single(): Run a function on another CPU
	357	* @cpu: The CPU to run on.
	358	* @data: Pre-allocated and setup data structure
	359	*
	360	* Like smp_call_function_single(), but allow caller to pass in a pre-allocated
	361	* data structure. Useful for embedding @data inside other structures, for
	362	* instance.
	363	*
	364	*/
	365	void __smp_call_function_single(int cpu, struct call_single_data *data)
	366	{
	367	/* Can deadlock when called with interrupts disabled */
	368	WARN_ON((data->flags & CSD_FLAG_WAIT) && irqs_disabled());
	369
	370	generic_exec_single(cpu, data);
	371	}
	372
ce47d974 RR	373	/* FIXME: Shim for archs using old arch_send_call_function_ipi API. */
	374	#ifndef arch_send_call_function_ipi_mask
	375	#define arch_send_call_function_ipi_mask(maskp) \
	376	arch_send_call_function_ipi(*(maskp))
	377	#endif
	378
3d442233	379	/**
54b11e6d RR	380	* smp_call_function_many(): Run a function on a set of other CPUs.
54b11e6d RR	381	* @mask: The set of cpus to run on (only runs on online subset).
3d442233 JA	382	* @func: The function to run. This must be fast and non-blocking.
	383	* @info: An arbitrary pointer to pass to the function.
	384	* @wait: If true, wait (atomically) until function has completed on other CPUs.
	385	*
3d442233 JA	386	* If @wait is true, then returns once @func has returned. Note that @wait
	387	* will be implicitly turned on in case of allocation failures, since
	388	* we fall back to on-stack allocation.
	389	*
	390	* You must not call this function with disabled interrupts or from a
	391	* hardware interrupt handler or from a bottom half handler. Preemption
	392	* must be disabled when calling this function.
	393	*/
54b11e6d RR	394	void smp_call_function_many(const struct cpumask *mask,
	395	void (func)(void ), void *info,
	396	bool wait)
3d442233	397	{
54b11e6d	398	struct call_function_data *data;
3d442233	399	unsigned long flags;
8969a5ed	400	int cpu, next_cpu, me = smp_processor_id();
3d442233 JA	401
	402	/* Can deadlock when called with interrupts disabled */
	403	WARN_ON(irqs_disabled());
	404
54b11e6d RR	405	/* So, what's a CPU they want? Ignoring this one. */
54b11e6d RR	406	cpu = cpumask_first_and(mask, cpu_online_mask);
8969a5ed	407	if (cpu == me)
54b11e6d RR	408	cpu = cpumask_next_and(cpu, mask, cpu_online_mask);
	409	/* No online cpus? We're done. */
	410	if (cpu >= nr_cpu_ids)
	411	return;
	412
	413	/* Do we have another CPU which isn't us? */
	414	next_cpu = cpumask_next_and(cpu, mask, cpu_online_mask);
8969a5ed	415	if (next_cpu == me)
54b11e6d RR	416	next_cpu = cpumask_next_and(next_cpu, mask, cpu_online_mask);
	417
	418	/* Fastpath: do that cpu by itself. */
	419	if (next_cpu >= nr_cpu_ids) {
	420	smp_call_function_single(cpu, func, info, wait);
	421	return;
3d442233 JA	422	}
3d442233 JA	423
8969a5ed PZ	424	data = &__get_cpu_var(cfd_data);
8969a5ed PZ	425	csd_lock(&data->csd);
3d442233	426
8969a5ed	427	spin_lock_irqsave(&data->lock, flags);
54b11e6d	428	if (wait)
8969a5ed PZ	429	csd_wait_prepare(&data->csd);
8969a5ed PZ	430
3d442233 JA	431	data->csd.func = func;
3d442233 JA	432	data->csd.info = info;
8969a5ed PZ	433	cpumask_and(data->cpumask, mask, cpu_online_mask);
	434	cpumask_clear_cpu(me, data->cpumask);
	435	data->refs = cpumask_weight(data->cpumask);
3d442233	436
8969a5ed PZ	437	spin_lock(&call_function.lock);
	438	/*
	439	* Place entry at the _HEAD_ of the list, so that any cpu still
	440	* observing the entry in generic_smp_call_function_interrupt() will
	441	* not miss any other list entries.
	442	*/
	443	list_add_rcu(&data->csd.list, &call_function.queue);
	444	spin_unlock(&call_function.lock);
	445	spin_unlock_irqrestore(&data->lock, flags);
3d442233	446
561920a0 SS	447	/*
561920a0 SS	448	* Make the list addition visible before sending the ipi.
15d0d3b3 NP	449	* (IPIs must obey or appear to obey normal Linux cache coherency
15d0d3b3 NP	450	* rules -- see comment in generic_exec_single).
561920a0 SS	451	*/
	452	smp_mb();
	453
3d442233	454	/* Send a message to all CPUs in the map */
8969a5ed	455	arch_send_call_function_ipi_mask(data->cpumask);
3d442233 JA	456
3d442233 JA	457	/* optionally wait for the CPUs to complete */
54b11e6d	458	if (wait)
8969a5ed	459	csd_wait(&data->csd);
3d442233	460	}
54b11e6d	461	EXPORT_SYMBOL(smp_call_function_many);
3d442233 JA	462
	463	/**
	464	* smp_call_function(): Run a function on all other CPUs.
	465	* @func: The function to run. This must be fast and non-blocking.
	466	* @info: An arbitrary pointer to pass to the function.
3d442233 JA	467	* @wait: If true, wait (atomically) until function has completed on other CPUs.
3d442233 JA	468	*
54b11e6d	469	* Returns 0.
3d442233 JA	470	*
	471	* If @wait is true, then returns once @func has returned; otherwise
	472	* it returns just before the target cpu calls @func. In case of allocation
	473	* failure, @wait will be implicitly turned on.
	474	*
	475	* You must not call this function with disabled interrupts or from a
	476	* hardware interrupt handler or from a bottom half handler.
	477	*/
8691e5a8	478	int smp_call_function(void (func)(void ), void *info, int wait)
3d442233	479	{
3d442233	480	preempt_disable();
54b11e6d	481	smp_call_function_many(cpu_online_mask, func, info, wait);
3d442233	482	preempt_enable();
54b11e6d	483	return 0;
3d442233 JA	484	}
	485	EXPORT_SYMBOL(smp_call_function);
	486
	487	void ipi_call_lock(void)
	488	{
8969a5ed	489	spin_lock(&call_function.lock);
3d442233 JA	490	}
	491
	492	void ipi_call_unlock(void)
	493	{
8969a5ed	494	spin_unlock(&call_function.lock);
3d442233 JA	495	}
	496
	497	void ipi_call_lock_irq(void)
	498	{
8969a5ed	499	spin_lock_irq(&call_function.lock);
3d442233 JA	500	}
	501
	502	void ipi_call_unlock_irq(void)
	503	{
8969a5ed	504	spin_unlock_irq(&call_function.lock);
3d442233	505	}