[linux.git] / lib / percpu_counter.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Fast batching percpu counters.
 */

#include <linux/percpu_counter.h>
#include <linux/mutex.h>
#include <linux/init.h>
#include <linux/cpu.h>
#include <linux/module.h>
#include <linux/debugobjects.h>

#ifdef CONFIG_HOTPLUG_CPU
static LIST_HEAD(percpu_counters);
static DEFINE_SPINLOCK(percpu_counters_lock);
#endif

#ifdef CONFIG_DEBUG_OBJECTS_PERCPU_COUNTER

static const struct debug_obj_descr percpu_counter_debug_descr;

static bool percpu_counter_fixup_free(void *addr, enum debug_obj_state state)
{
	struct percpu_counter *fbc = addr;

	switch (state) {
	case ODEBUG_STATE_ACTIVE:
		percpu_counter_destroy(fbc);
		debug_object_free(fbc, &percpu_counter_debug_descr);
		return true;
	default:
		return false;
	}
}

static const struct debug_obj_descr percpu_counter_debug_descr = {
	.name		= "percpu_counter",
	.fixup_free	= percpu_counter_fixup_free,
};

static inline void debug_percpu_counter_activate(struct percpu_counter *fbc)
{
	debug_object_init(fbc, &percpu_counter_debug_descr);
	debug_object_activate(fbc, &percpu_counter_debug_descr);
}

static inline void debug_percpu_counter_deactivate(struct percpu_counter *fbc)
{
	debug_object_deactivate(fbc, &percpu_counter_debug_descr);
	debug_object_free(fbc, &percpu_counter_debug_descr);
}

#else	/* CONFIG_DEBUG_OBJECTS_PERCPU_COUNTER */
static inline void debug_percpu_counter_activate(struct percpu_counter *fbc)
{ }
static inline void debug_percpu_counter_deactivate(struct percpu_counter *fbc)
{ }
#endif	/* CONFIG_DEBUG_OBJECTS_PERCPU_COUNTER */

void percpu_counter_set(struct percpu_counter *fbc, s64 amount)
{
	int cpu;
	unsigned long flags;

	raw_spin_lock_irqsave(&fbc->lock, flags);
	for_each_possible_cpu(cpu) {
		s32 *pcount = per_cpu_ptr(fbc->counters, cpu);
		*pcount = 0;
	}
	fbc->count = amount;
	raw_spin_unlock_irqrestore(&fbc->lock, flags);
}
EXPORT_SYMBOL(percpu_counter_set);

/*
 * Add to a counter while respecting batch size.
 *
 * There are 2 implementations, both dealing with the following problem:
 *
 * The decision slow path/fast path and the actual update must be atomic.
 * Otherwise a call in process context could check the current values and
 * decide that the fast path can be used. If now an interrupt occurs before
 * the this_cpu_add(), and the interrupt updates this_cpu(*fbc->counters),
 * then the this_cpu_add() that is executed after the interrupt has completed
 * can produce values larger than "batch" or even overflows.
 */
#ifdef CONFIG_HAVE_CMPXCHG_LOCAL
/*
 * Safety against interrupts is achieved in 2 ways:
 * 1. the fast path uses local cmpxchg (note: no lock prefix)
 * 2. the slow path operates with interrupts disabled
 */
void percpu_counter_add_batch(struct percpu_counter *fbc, s64 amount, s32 batch)
{
	s64 count;
	unsigned long flags;

	count = this_cpu_read(*fbc->counters);
	do {
		if (unlikely(abs(count + amount) >= batch)) {
			raw_spin_lock_irqsave(&fbc->lock, flags);
			/*
			 * Note: by now we might have migrated to another CPU
			 * or the value might have changed.
			 */
			count = __this_cpu_read(*fbc->counters);
			fbc->count += count + amount;
			__this_cpu_sub(*fbc->counters, count);
			raw_spin_unlock_irqrestore(&fbc->lock, flags);
			return;
		}
	} while (!this_cpu_try_cmpxchg(*fbc->counters, &count, count + amount));
}
#else
/*
 * local_irq_save() is used to make the function irq safe:
 * - The slow path would be ok as protected by an irq-safe spinlock.
 * - this_cpu_add would be ok as it is irq-safe by definition.
 */
void percpu_counter_add_batch(struct percpu_counter *fbc, s64 amount, s32 batch)
{
	s64 count;
	unsigned long flags;

	local_irq_save(flags);
	count = __this_cpu_read(*fbc->counters) + amount;
	if (abs(count) >= batch) {
		raw_spin_lock(&fbc->lock);
		fbc->count += count;
		__this_cpu_sub(*fbc->counters, count - amount);
		raw_spin_unlock(&fbc->lock);
	} else {
		this_cpu_add(*fbc->counters, amount);
	}
	local_irq_restore(flags);
}
#endif
EXPORT_SYMBOL(percpu_counter_add_batch);

/*
 * For percpu_counter with a big batch, the devication of its count could
 * be big, and there is requirement to reduce the deviation, like when the
 * counter's batch could be runtime decreased to get a better accuracy,
 * which can be achieved by running this sync function on each CPU.
 */
void percpu_counter_sync(struct percpu_counter *fbc)
{
	unsigned long flags;
	s64 count;

	raw_spin_lock_irqsave(&fbc->lock, flags);
	count = __this_cpu_read(*fbc->counters);
	fbc->count += count;
	__this_cpu_sub(*fbc->counters, count);
	raw_spin_unlock_irqrestore(&fbc->lock, flags);
}
EXPORT_SYMBOL(percpu_counter_sync);

/*
 * Add up all the per-cpu counts, return the result.  This is a more accurate
 * but much slower version of percpu_counter_read_positive().
 *
 * We use the cpu mask of (cpu_online_mask | cpu_dying_mask) to capture sums
 * from CPUs that are in the process of being taken offline. Dying cpus have
 * been removed from the online mask, but may not have had the hotplug dead
 * notifier called to fold the percpu count back into the global counter sum.
 * By including dying CPUs in the iteration mask, we avoid this race condition
 * so __percpu_counter_sum() just does the right thing when CPUs are being taken
 * offline.
 */
s64 __percpu_counter_sum(struct percpu_counter *fbc)
{
	s64 ret;
	int cpu;
	unsigned long flags;

	raw_spin_lock_irqsave(&fbc->lock, flags);
	ret = fbc->count;
	for_each_cpu_or(cpu, cpu_online_mask, cpu_dying_mask) {
		s32 *pcount = per_cpu_ptr(fbc->counters, cpu);
		ret += *pcount;
	}
	raw_spin_unlock_irqrestore(&fbc->lock, flags);
	return ret;
}
EXPORT_SYMBOL(__percpu_counter_sum);

int __percpu_counter_init_many(struct percpu_counter *fbc, s64 amount,
			       gfp_t gfp, u32 nr_counters,
			       struct lock_class_key *key)
{
	unsigned long flags __maybe_unused;
	size_t counter_size;
	s32 __percpu *counters;
	u32 i;

	counter_size = ALIGN(sizeof(*counters), __alignof__(*counters));
	counters = __alloc_percpu_gfp(nr_counters * counter_size,
				      __alignof__(*counters), gfp);
	if (!counters) {
		fbc[0].counters = NULL;
		return -ENOMEM;
	}

	for (i = 0; i < nr_counters; i++) {
		raw_spin_lock_init(&fbc[i].lock);
		lockdep_set_class(&fbc[i].lock, key);
#ifdef CONFIG_HOTPLUG_CPU
		INIT_LIST_HEAD(&fbc[i].list);
#endif
		fbc[i].count = amount;
		fbc[i].counters = (void __percpu *)counters + i * counter_size;

		debug_percpu_counter_activate(&fbc[i]);
	}

#ifdef CONFIG_HOTPLUG_CPU
	spin_lock_irqsave(&percpu_counters_lock, flags);
	for (i = 0; i < nr_counters; i++)
		list_add(&fbc[i].list, &percpu_counters);
	spin_unlock_irqrestore(&percpu_counters_lock, flags);
#endif
	return 0;
}
EXPORT_SYMBOL(__percpu_counter_init_many);

void percpu_counter_destroy_many(struct percpu_counter *fbc, u32 nr_counters)
{
	unsigned long flags __maybe_unused;
	u32 i;

	if (WARN_ON_ONCE(!fbc))
		return;

	if (!fbc[0].counters)
		return;

	for (i = 0; i < nr_counters; i++)
		debug_percpu_counter_deactivate(&fbc[i]);

#ifdef CONFIG_HOTPLUG_CPU
	spin_lock_irqsave(&percpu_counters_lock, flags);
	for (i = 0; i < nr_counters; i++)
		list_del(&fbc[i].list);
	spin_unlock_irqrestore(&percpu_counters_lock, flags);
#endif

	free_percpu(fbc[0].counters);

	for (i = 0; i < nr_counters; i++)
		fbc[i].counters = NULL;
}
EXPORT_SYMBOL(percpu_counter_destroy_many);

int percpu_counter_batch __read_mostly = 32;
EXPORT_SYMBOL(percpu_counter_batch);

static int compute_batch_value(unsigned int cpu)
{
	int nr = num_online_cpus();

	percpu_counter_batch = max(32, nr*2);
	return 0;
}

static int percpu_counter_cpu_dead(unsigned int cpu)
{
#ifdef CONFIG_HOTPLUG_CPU
	struct percpu_counter *fbc;

	compute_batch_value(cpu);

	spin_lock_irq(&percpu_counters_lock);
	list_for_each_entry(fbc, &percpu_counters, list) {
		s32 *pcount;

		raw_spin_lock(&fbc->lock);
		pcount = per_cpu_ptr(fbc->counters, cpu);
		fbc->count += *pcount;
		*pcount = 0;
		raw_spin_unlock(&fbc->lock);
	}
	spin_unlock_irq(&percpu_counters_lock);
#endif
	return 0;
}

/*
 * Compare counter against given value.
 * Return 1 if greater, 0 if equal and -1 if less
 */
int __percpu_counter_compare(struct percpu_counter *fbc, s64 rhs, s32 batch)
{
	s64	count;

	count = percpu_counter_read(fbc);
	/* Check to see if rough count will be sufficient for comparison */
	if (abs(count - rhs) > (batch * num_online_cpus())) {
		if (count > rhs)
			return 1;
		else
			return -1;
	}
	/* Need to use precise count */
	count = percpu_counter_sum(fbc);
	if (count > rhs)
		return 1;
	else if (count < rhs)
		return -1;
	else
		return 0;
}
EXPORT_SYMBOL(__percpu_counter_compare);

/*
 * Compare counter, and add amount if total is: less than or equal to limit if
 * amount is positive, or greater than or equal to limit if amount is negative.
 * Return true if amount is added, or false if total would be beyond the limit.
 *
 * Negative limit is allowed, but unusual.
 * When negative amounts (subs) are given to percpu_counter_limited_add(),
 * the limit would most naturally be 0 - but other limits are also allowed.
 *
 * Overflow beyond S64_MAX is not allowed for: counter, limit and amount
 * are all assumed to be sane (far from S64_MIN and S64_MAX).
 */
bool __percpu_counter_limited_add(struct percpu_counter *fbc,
				  s64 limit, s64 amount, s32 batch)
{
	s64 count;
	s64 unknown;
	unsigned long flags;
	bool good = false;

	if (amount == 0)
		return true;

	local_irq_save(flags);
	unknown = batch * num_online_cpus();
	count = __this_cpu_read(*fbc->counters);

	/* Skip taking the lock when safe */
	if (abs(count + amount) <= batch &&
	    ((amount > 0 && fbc->count + unknown <= limit) ||
	     (amount < 0 && fbc->count - unknown >= limit))) {
		this_cpu_add(*fbc->counters, amount);
		local_irq_restore(flags);
		return true;
	}

	raw_spin_lock(&fbc->lock);
	count = fbc->count + amount;

	/* Skip percpu_counter_sum() when safe */
	if (amount > 0) {
		if (count - unknown > limit)
			goto out;
		if (count + unknown <= limit)
			good = true;
	} else {
		if (count + unknown < limit)
			goto out;
		if (count - unknown >= limit)
			good = true;
	}

	if (!good) {
		s32 *pcount;
		int cpu;

		for_each_cpu_or(cpu, cpu_online_mask, cpu_dying_mask) {
			pcount = per_cpu_ptr(fbc->counters, cpu);
			count += *pcount;
		}
		if (amount > 0) {
			if (count > limit)
				goto out;
		} else {
			if (count < limit)
				goto out;
		}
		good = true;
	}

	count = __this_cpu_read(*fbc->counters);
	fbc->count += count + amount;
	__this_cpu_sub(*fbc->counters, count);
out:
	raw_spin_unlock(&fbc->lock);
	local_irq_restore(flags);
	return good;
}

static int __init percpu_counter_startup(void)
{
	int ret;

	ret = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "lib/percpu_cnt:online",
				compute_batch_value, NULL);
	WARN_ON(ret < 0);
	ret = cpuhp_setup_state_nocalls(CPUHP_PERCPU_CNT_DEAD,
					"lib/percpu_cnt:dead", NULL,
					percpu_counter_cpu_dead);
	WARN_ON(ret < 0);
	return 0;
}
module_init(percpu_counter_startup);
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
3cbc5640 RT	2	/*
	3	* Fast batching percpu counters.
	4	*/
	5
	6	#include <linux/percpu_counter.h>
c67ad917 AM	7	#include <linux/mutex.h>
	8	#include <linux/init.h>
	9	#include <linux/cpu.h>
3cbc5640	10	#include <linux/module.h>
e2852ae8	11	#include <linux/debugobjects.h>
3cbc5640	12
3a8495c7	13	#ifdef CONFIG_HOTPLUG_CPU
c67ad917	14	static LIST_HEAD(percpu_counters);
d87aae2f	15	static DEFINE_SPINLOCK(percpu_counters_lock);
3a8495c7	16	#endif
c67ad917	17
e2852ae8 TH	18	#ifdef CONFIG_DEBUG_OBJECTS_PERCPU_COUNTER
e2852ae8 TH	19
f9e62f31	20	static const struct debug_obj_descr percpu_counter_debug_descr;
e2852ae8	21
d99b1d89	22	static bool percpu_counter_fixup_free(void *addr, enum debug_obj_state state)
e2852ae8 TH	23	{
	24	struct percpu_counter *fbc = addr;
	25
	26	switch (state) {
	27	case ODEBUG_STATE_ACTIVE:
	28	percpu_counter_destroy(fbc);
	29	debug_object_free(fbc, &percpu_counter_debug_descr);
d99b1d89	30	return true;
e2852ae8	31	default:
d99b1d89	32	return false;
e2852ae8 TH	33	}
	34	}
	35
f9e62f31	36	static const struct debug_obj_descr percpu_counter_debug_descr = {
e2852ae8 TH	37	.name = "percpu_counter",
	38	.fixup_free = percpu_counter_fixup_free,
	39	};
	40
	41	static inline void debug_percpu_counter_activate(struct percpu_counter *fbc)
	42	{
	43	debug_object_init(fbc, &percpu_counter_debug_descr);
	44	debug_object_activate(fbc, &percpu_counter_debug_descr);
	45	}
	46
	47	static inline void debug_percpu_counter_deactivate(struct percpu_counter *fbc)
	48	{
	49	debug_object_deactivate(fbc, &percpu_counter_debug_descr);
	50	debug_object_free(fbc, &percpu_counter_debug_descr);
	51	}
	52
	53	#else /* CONFIG_DEBUG_OBJECTS_PERCPU_COUNTER */
	54	static inline void debug_percpu_counter_activate(struct percpu_counter *fbc)
	55	{ }
	56	static inline void debug_percpu_counter_deactivate(struct percpu_counter *fbc)
	57	{ }
	58	#endif /* CONFIG_DEBUG_OBJECTS_PERCPU_COUNTER */
	59
3a587f47 PZ	60	void percpu_counter_set(struct percpu_counter *fbc, s64 amount)
	61	{
	62	int cpu;
098faf58	63	unsigned long flags;
3a587f47	64
098faf58	65	raw_spin_lock_irqsave(&fbc->lock, flags);
3a587f47 PZ	66	for_each_possible_cpu(cpu) {
	67	s32 *pcount = per_cpu_ptr(fbc->counters, cpu);
	68	*pcount = 0;
	69	}
	70	fbc->count = amount;
098faf58	71	raw_spin_unlock_irqrestore(&fbc->lock, flags);
3a587f47 PZ	72	}
	73	EXPORT_SYMBOL(percpu_counter_set);
	74
db65a867	75	/*
51d82165 MG	76	* Add to a counter while respecting batch size.
	77	*
	78	* There are 2 implementations, both dealing with the following problem:
	79	*
	80	* The decision slow path/fast path and the actual update must be atomic.
805afd83 MS	81	* Otherwise a call in process context could check the current values and
	82	* decide that the fast path can be used. If now an interrupt occurs before
	83	* the this_cpu_add(), and the interrupt updates this_cpu(*fbc->counters),
	84	* then the this_cpu_add() that is executed after the interrupt has completed
	85	* can produce values larger than "batch" or even overflows.
3e8f399d	86	*/
51d82165 MG	87	#ifdef CONFIG_HAVE_CMPXCHG_LOCAL
	88	/*
	89	* Safety against interrupts is achieved in 2 ways:
	90	* 1. the fast path uses local cmpxchg (note: no lock prefix)
	91	* 2. the slow path operates with interrupts disabled
	92	*/
	93	void percpu_counter_add_batch(struct percpu_counter *fbc, s64 amount, s32 batch)
	94	{
	95	s64 count;
	96	unsigned long flags;
	97
	98	count = this_cpu_read(*fbc->counters);
	99	do {
	100	if (unlikely(abs(count + amount) >= batch)) {
	101	raw_spin_lock_irqsave(&fbc->lock, flags);
	102	/*
	103	* Note: by now we might have migrated to another CPU
	104	* or the value might have changed.
	105	*/
	106	count = __this_cpu_read(*fbc->counters);
	107	fbc->count += count + amount;
	108	__this_cpu_sub(*fbc->counters, count);
	109	raw_spin_unlock_irqrestore(&fbc->lock, flags);
	110	return;
	111	}
	112	} while (!this_cpu_try_cmpxchg(*fbc->counters, &count, count + amount));
	113	}
	114	#else
	115	/*
	116	* local_irq_save() is used to make the function irq safe:
	117	* - The slow path would be ok as protected by an irq-safe spinlock.
	118	* - this_cpu_add would be ok as it is irq-safe by definition.
	119	*/
104b4e51	120	void percpu_counter_add_batch(struct percpu_counter *fbc, s64 amount, s32 batch)
3cbc5640	121	{
20e89767	122	s64 count;
805afd83	123	unsigned long flags;
3cbc5640	124
805afd83	125	local_irq_save(flags);
819a72af	126	count = __this_cpu_read(*fbc->counters) + amount;
1d339638	127	if (abs(count) >= batch) {
805afd83	128	raw_spin_lock(&fbc->lock);
3cbc5640	129	fbc->count += count;
d1969a84	130	__this_cpu_sub(*fbc->counters, count - amount);
805afd83	131	raw_spin_unlock(&fbc->lock);
3cbc5640	132	} else {
74e72f89	133	this_cpu_add(*fbc->counters, amount);
3cbc5640	134	}
805afd83	135	local_irq_restore(flags);
3cbc5640	136	}
51d82165	137	#endif
104b4e51	138	EXPORT_SYMBOL(percpu_counter_add_batch);
3cbc5640	139
0a4954a8 FT	140	/*
	141	* For percpu_counter with a big batch, the devication of its count could
	142	* be big, and there is requirement to reduce the deviation, like when the
	143	* counter's batch could be runtime decreased to get a better accuracy,
	144	* which can be achieved by running this sync function on each CPU.
	145	*/
	146	void percpu_counter_sync(struct percpu_counter *fbc)
	147	{
	148	unsigned long flags;
	149	s64 count;
	150
	151	raw_spin_lock_irqsave(&fbc->lock, flags);
	152	count = __this_cpu_read(*fbc->counters);
	153	fbc->count += count;
	154	__this_cpu_sub(*fbc->counters, count);
	155	raw_spin_unlock_irqrestore(&fbc->lock, flags);
	156	}
	157	EXPORT_SYMBOL(percpu_counter_sync);
	158
f689054a SB	159	/*
f689054a SB	160	* Add up all the per-cpu counts, return the result. This is a more accurate
8b57b11c DC	161	* but much slower version of percpu_counter_read_positive().
	162	*
	163	* We use the cpu mask of (cpu_online_mask \| cpu_dying_mask) to capture sums
	164	* from CPUs that are in the process of being taken offline. Dying cpus have
	165	* been removed from the online mask, but may not have had the hotplug dead
	166	* notifier called to fold the percpu count back into the global counter sum.
	167	* By including dying CPUs in the iteration mask, we avoid this race condition
	168	* so __percpu_counter_sum() just does the right thing when CPUs are being taken
	169	* offline.
f689054a SB	170	*/
	171	s64 __percpu_counter_sum(struct percpu_counter *fbc)
	172	{
e9b60c7f DC	173	s64 ret;
	174	int cpu;
	175	unsigned long flags;
8b57b11c	176
e9b60c7f DC	177	raw_spin_lock_irqsave(&fbc->lock, flags);
	178	ret = fbc->count;
	179	for_each_cpu_or(cpu, cpu_online_mask, cpu_dying_mask) {
	180	s32 *pcount = per_cpu_ptr(fbc->counters, cpu);
	181	ret += *pcount;
	182	}
	183	raw_spin_unlock_irqrestore(&fbc->lock, flags);
	184	return ret;
f689054a	185	}
bf1d89c8	186	EXPORT_SYMBOL(__percpu_counter_sum);
c67ad917	187
c439d5e8 MG	188	int __percpu_counter_init_many(struct percpu_counter *fbc, s64 amount,
	189	gfp_t gfp, u32 nr_counters,
	190	struct lock_class_key *key)
c67ad917	191	{
ebd8fef3	192	unsigned long flags __maybe_unused;
c439d5e8 MG	193	size_t counter_size;
	194	s32 __percpu *counters;
	195	u32 i;
	196
	197	counter_size = ALIGN(sizeof(counters), __alignof__(counters));
	198	counters = __alloc_percpu_gfp(nr_counters * counter_size,
	199	__alignof__(*counters), gfp);
	200	if (!counters) {
	201	fbc[0].counters = NULL;
833f4077	202	return -ENOMEM;
c439d5e8	203	}
e2852ae8	204
c439d5e8 MG	205	for (i = 0; i < nr_counters; i++) {
	206	raw_spin_lock_init(&fbc[i].lock);
	207	lockdep_set_class(&fbc[i].lock, key);
	208	#ifdef CONFIG_HOTPLUG_CPU
	209	INIT_LIST_HEAD(&fbc[i].list);
	210	#endif
	211	fbc[i].count = amount;
16d9691a	212	fbc[i].counters = (void __percpu )counters + i counter_size;
c439d5e8 MG	213
	214	debug_percpu_counter_activate(&fbc[i]);
	215	}
e2852ae8	216
c67ad917	217	#ifdef CONFIG_HOTPLUG_CPU
ebd8fef3	218	spin_lock_irqsave(&percpu_counters_lock, flags);
c439d5e8 MG	219	for (i = 0; i < nr_counters; i++)
c439d5e8 MG	220	list_add(&fbc[i].list, &percpu_counters);
ebd8fef3	221	spin_unlock_irqrestore(&percpu_counters_lock, flags);
c67ad917	222	#endif
833f4077	223	return 0;
c67ad917	224	}
c439d5e8	225	EXPORT_SYMBOL(__percpu_counter_init_many);
c67ad917	226
c439d5e8	227	void percpu_counter_destroy_many(struct percpu_counter *fbc, u32 nr_counters)
c67ad917	228	{
ebd8fef3	229	unsigned long flags __maybe_unused;
c439d5e8 MG	230	u32 i;
	231
	232	if (WARN_ON_ONCE(!fbc))
	233	return;
ebd8fef3	234
c439d5e8	235	if (!fbc[0].counters)
833f4077 PZ	236	return;
833f4077 PZ	237
c439d5e8 MG	238	for (i = 0; i < nr_counters; i++)
c439d5e8 MG	239	debug_percpu_counter_deactivate(&fbc[i]);
e2852ae8	240
c67ad917	241	#ifdef CONFIG_HOTPLUG_CPU
ebd8fef3	242	spin_lock_irqsave(&percpu_counters_lock, flags);
c439d5e8 MG	243	for (i = 0; i < nr_counters; i++)
c439d5e8 MG	244	list_del(&fbc[i].list);
ebd8fef3	245	spin_unlock_irqrestore(&percpu_counters_lock, flags);
c67ad917	246	#endif
c439d5e8 MG	247
	248	free_percpu(fbc[0].counters);
	249
	250	for (i = 0; i < nr_counters; i++)
	251	fbc[i].counters = NULL;
c67ad917	252	}
c439d5e8	253	EXPORT_SYMBOL(percpu_counter_destroy_many);
c67ad917	254
179f7ebf ED	255	int percpu_counter_batch __read_mostly = 32;
	256	EXPORT_SYMBOL(percpu_counter_batch);
	257
5588f5af	258	static int compute_batch_value(unsigned int cpu)
179f7ebf ED	259	{
	260	int nr = num_online_cpus();
	261
	262	percpu_counter_batch = max(32, nr*2);
5588f5af	263	return 0;
179f7ebf ED	264	}
179f7ebf ED	265
5588f5af	266	static int percpu_counter_cpu_dead(unsigned int cpu)
c67ad917	267	{
179f7ebf	268	#ifdef CONFIG_HOTPLUG_CPU
c67ad917 AM	269	struct percpu_counter *fbc;
c67ad917 AM	270
5588f5af	271	compute_batch_value(cpu);
c67ad917	272
ebd8fef3	273	spin_lock_irq(&percpu_counters_lock);
c67ad917 AM	274	list_for_each_entry(fbc, &percpu_counters, list) {
	275	s32 *pcount;
	276
aaf0f2fa	277	raw_spin_lock(&fbc->lock);
c67ad917 AM	278	pcount = per_cpu_ptr(fbc->counters, cpu);
	279	fbc->count += *pcount;
	280	*pcount = 0;
aaf0f2fa	281	raw_spin_unlock(&fbc->lock);
c67ad917	282	}
ebd8fef3	283	spin_unlock_irq(&percpu_counters_lock);
179f7ebf	284	#endif
5588f5af	285	return 0;
c67ad917 AM	286	}
c67ad917 AM	287
27f5e0f6 TC	288	/*
	289	* Compare counter against given value.
	290	* Return 1 if greater, 0 if equal and -1 if less
	291	*/
80188b0d	292	int __percpu_counter_compare(struct percpu_counter *fbc, s64 rhs, s32 batch)
27f5e0f6 TC	293	{
	294	s64 count;
	295
	296	count = percpu_counter_read(fbc);
	297	/* Check to see if rough count will be sufficient for comparison */
80188b0d	298	if (abs(count - rhs) > (batch * num_online_cpus())) {
27f5e0f6 TC	299	if (count > rhs)
	300	return 1;
	301	else
	302	return -1;
	303	}
	304	/* Need to use precise count */
	305	count = percpu_counter_sum(fbc);
	306	if (count > rhs)
	307	return 1;
	308	else if (count < rhs)
	309	return -1;
	310	else
	311	return 0;
	312	}
80188b0d	313	EXPORT_SYMBOL(__percpu_counter_compare);
27f5e0f6	314
beb98686	315	/*
1431996b HD	316	* Compare counter, and add amount if total is: less than or equal to limit if
	317	* amount is positive, or greater than or equal to limit if amount is negative.
	318	* Return true if amount is added, or false if total would be beyond the limit.
	319	*
	320	* Negative limit is allowed, but unusual.
	321	* When negative amounts (subs) are given to percpu_counter_limited_add(),
	322	* the limit would most naturally be 0 - but other limits are also allowed.
	323	*
	324	* Overflow beyond S64_MAX is not allowed for: counter, limit and amount
	325	* are all assumed to be sane (far from S64_MIN and S64_MAX).
beb98686 HD	326	*/
	327	bool __percpu_counter_limited_add(struct percpu_counter *fbc,
	328	s64 limit, s64 amount, s32 batch)
	329	{
	330	s64 count;
	331	s64 unknown;
	332	unsigned long flags;
1431996b	333	bool good = false;
beb98686	334
1431996b HD	335	if (amount == 0)
1431996b HD	336	return true;
beb98686 HD	337
	338	local_irq_save(flags);
	339	unknown = batch * num_online_cpus();
	340	count = __this_cpu_read(*fbc->counters);
	341
	342	/* Skip taking the lock when safe */
	343	if (abs(count + amount) <= batch &&
1431996b HD	344	((amount > 0 && fbc->count + unknown <= limit) \|\|
1431996b HD	345	(amount < 0 && fbc->count - unknown >= limit))) {
beb98686 HD	346	this_cpu_add(*fbc->counters, amount);
	347	local_irq_restore(flags);
	348	return true;
	349	}
	350
	351	raw_spin_lock(&fbc->lock);
	352	count = fbc->count + amount;
	353
	354	/* Skip percpu_counter_sum() when safe */
1431996b HD	355	if (amount > 0) {
	356	if (count - unknown > limit)
	357	goto out;
	358	if (count + unknown <= limit)
	359	good = true;
	360	} else {
	361	if (count + unknown < limit)
	362	goto out;
	363	if (count - unknown >= limit)
	364	good = true;
	365	}
	366
	367	if (!good) {
beb98686 HD	368	s32 *pcount;
	369	int cpu;
	370
	371	for_each_cpu_or(cpu, cpu_online_mask, cpu_dying_mask) {
	372	pcount = per_cpu_ptr(fbc->counters, cpu);
	373	count += *pcount;
	374	}
1431996b HD	375	if (amount > 0) {
	376	if (count > limit)
	377	goto out;
	378	} else {
	379	if (count < limit)
	380	goto out;
	381	}
	382	good = true;
beb98686 HD	383	}
beb98686 HD	384
1431996b HD	385	count = __this_cpu_read(*fbc->counters);
	386	fbc->count += count + amount;
	387	__this_cpu_sub(*fbc->counters, count);
	388	out:
beb98686 HD	389	raw_spin_unlock(&fbc->lock);
	390	local_irq_restore(flags);
	391	return good;
	392	}
	393
c67ad917 AM	394	static int __init percpu_counter_startup(void)
c67ad917 AM	395	{
5588f5af SAS	396	int ret;
	397
	398	ret = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "lib/percpu_cnt:online",
	399	compute_batch_value, NULL);
	400	WARN_ON(ret < 0);
	401	ret = cpuhp_setup_state_nocalls(CPUHP_PERCPU_CNT_DEAD,
	402	"lib/percpu_cnt:dead", NULL,
	403	percpu_counter_cpu_dead);
	404	WARN_ON(ret < 0);
c67ad917 AM	405	return 0;
	406	}
	407	module_init(percpu_counter_startup);