[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);

/*
 * local_irq_save() is needed 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.
 * But:
 * The decision slow path/fast path and the actual update must be atomic, too.
 * 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.
 */
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);
}
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 *)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	/*
805afd83 MS	76	* local_irq_save() is needed to make the function irq safe:
	77	* - The slow path would be ok as protected by an irq-safe spinlock.
	78	* - this_cpu_add would be ok as it is irq-safe by definition.
	79	* But:
	80	* The decision slow path/fast path and the actual update must be atomic, too.
	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	*/
104b4e51	87	void percpu_counter_add_batch(struct percpu_counter *fbc, s64 amount, s32 batch)
3cbc5640	88	{
20e89767	89	s64 count;
805afd83	90	unsigned long flags;
3cbc5640	91
805afd83	92	local_irq_save(flags);
819a72af	93	count = __this_cpu_read(*fbc->counters) + amount;
1d339638	94	if (abs(count) >= batch) {
805afd83	95	raw_spin_lock(&fbc->lock);
3cbc5640	96	fbc->count += count;
d1969a84	97	__this_cpu_sub(*fbc->counters, count - amount);
805afd83	98	raw_spin_unlock(&fbc->lock);
3cbc5640	99	} else {
74e72f89	100	this_cpu_add(*fbc->counters, amount);
3cbc5640	101	}
805afd83	102	local_irq_restore(flags);
3cbc5640	103	}
104b4e51	104	EXPORT_SYMBOL(percpu_counter_add_batch);
3cbc5640	105
0a4954a8 FT	106	/*
	107	* For percpu_counter with a big batch, the devication of its count could
	108	* be big, and there is requirement to reduce the deviation, like when the
	109	* counter's batch could be runtime decreased to get a better accuracy,
	110	* which can be achieved by running this sync function on each CPU.
	111	*/
	112	void percpu_counter_sync(struct percpu_counter *fbc)
	113	{
	114	unsigned long flags;
	115	s64 count;
	116
	117	raw_spin_lock_irqsave(&fbc->lock, flags);
	118	count = __this_cpu_read(*fbc->counters);
	119	fbc->count += count;
	120	__this_cpu_sub(*fbc->counters, count);
	121	raw_spin_unlock_irqrestore(&fbc->lock, flags);
	122	}
	123	EXPORT_SYMBOL(percpu_counter_sync);
	124
f689054a SB	125	/*
f689054a SB	126	* Add up all the per-cpu counts, return the result. This is a more accurate
8b57b11c DC	127	* but much slower version of percpu_counter_read_positive().
	128	*
	129	* We use the cpu mask of (cpu_online_mask \| cpu_dying_mask) to capture sums
	130	* from CPUs that are in the process of being taken offline. Dying cpus have
	131	* been removed from the online mask, but may not have had the hotplug dead
	132	* notifier called to fold the percpu count back into the global counter sum.
	133	* By including dying CPUs in the iteration mask, we avoid this race condition
	134	* so __percpu_counter_sum() just does the right thing when CPUs are being taken
	135	* offline.
f689054a SB	136	*/
	137	s64 __percpu_counter_sum(struct percpu_counter *fbc)
	138	{
e9b60c7f DC	139	s64 ret;
	140	int cpu;
	141	unsigned long flags;
8b57b11c	142
e9b60c7f DC	143	raw_spin_lock_irqsave(&fbc->lock, flags);
	144	ret = fbc->count;
	145	for_each_cpu_or(cpu, cpu_online_mask, cpu_dying_mask) {
	146	s32 *pcount = per_cpu_ptr(fbc->counters, cpu);
	147	ret += *pcount;
	148	}
	149	raw_spin_unlock_irqrestore(&fbc->lock, flags);
	150	return ret;
f689054a	151	}
bf1d89c8	152	EXPORT_SYMBOL(__percpu_counter_sum);
c67ad917	153
c439d5e8 MG	154	int __percpu_counter_init_many(struct percpu_counter *fbc, s64 amount,
	155	gfp_t gfp, u32 nr_counters,
	156	struct lock_class_key *key)
c67ad917	157	{
ebd8fef3	158	unsigned long flags __maybe_unused;
c439d5e8 MG	159	size_t counter_size;
	160	s32 __percpu *counters;
	161	u32 i;
	162
	163	counter_size = ALIGN(sizeof(counters), __alignof__(counters));
	164	counters = __alloc_percpu_gfp(nr_counters * counter_size,
	165	__alignof__(*counters), gfp);
	166	if (!counters) {
	167	fbc[0].counters = NULL;
833f4077	168	return -ENOMEM;
c439d5e8	169	}
e2852ae8	170
c439d5e8 MG	171	for (i = 0; i < nr_counters; i++) {
	172	raw_spin_lock_init(&fbc[i].lock);
	173	lockdep_set_class(&fbc[i].lock, key);
	174	#ifdef CONFIG_HOTPLUG_CPU
	175	INIT_LIST_HEAD(&fbc[i].list);
	176	#endif
	177	fbc[i].count = amount;
	178	fbc[i].counters = (void )counters + (i counter_size);
	179
	180	debug_percpu_counter_activate(&fbc[i]);
	181	}
e2852ae8	182
c67ad917	183	#ifdef CONFIG_HOTPLUG_CPU
ebd8fef3	184	spin_lock_irqsave(&percpu_counters_lock, flags);
c439d5e8 MG	185	for (i = 0; i < nr_counters; i++)
c439d5e8 MG	186	list_add(&fbc[i].list, &percpu_counters);
ebd8fef3	187	spin_unlock_irqrestore(&percpu_counters_lock, flags);
c67ad917	188	#endif
833f4077	189	return 0;
c67ad917	190	}
c439d5e8	191	EXPORT_SYMBOL(__percpu_counter_init_many);
c67ad917	192
c439d5e8	193	void percpu_counter_destroy_many(struct percpu_counter *fbc, u32 nr_counters)
c67ad917	194	{
ebd8fef3	195	unsigned long flags __maybe_unused;
c439d5e8 MG	196	u32 i;
	197
	198	if (WARN_ON_ONCE(!fbc))
	199	return;
ebd8fef3	200
c439d5e8	201	if (!fbc[0].counters)
833f4077 PZ	202	return;
833f4077 PZ	203
c439d5e8 MG	204	for (i = 0; i < nr_counters; i++)
c439d5e8 MG	205	debug_percpu_counter_deactivate(&fbc[i]);
e2852ae8	206
c67ad917	207	#ifdef CONFIG_HOTPLUG_CPU
ebd8fef3	208	spin_lock_irqsave(&percpu_counters_lock, flags);
c439d5e8 MG	209	for (i = 0; i < nr_counters; i++)
c439d5e8 MG	210	list_del(&fbc[i].list);
ebd8fef3	211	spin_unlock_irqrestore(&percpu_counters_lock, flags);
c67ad917	212	#endif
c439d5e8 MG	213
	214	free_percpu(fbc[0].counters);
	215
	216	for (i = 0; i < nr_counters; i++)
	217	fbc[i].counters = NULL;
c67ad917	218	}
c439d5e8	219	EXPORT_SYMBOL(percpu_counter_destroy_many);
c67ad917	220
179f7ebf ED	221	int percpu_counter_batch __read_mostly = 32;
	222	EXPORT_SYMBOL(percpu_counter_batch);
	223
5588f5af	224	static int compute_batch_value(unsigned int cpu)
179f7ebf ED	225	{
	226	int nr = num_online_cpus();
	227
	228	percpu_counter_batch = max(32, nr*2);
5588f5af	229	return 0;
179f7ebf ED	230	}
179f7ebf ED	231
5588f5af	232	static int percpu_counter_cpu_dead(unsigned int cpu)
c67ad917	233	{
179f7ebf	234	#ifdef CONFIG_HOTPLUG_CPU
c67ad917 AM	235	struct percpu_counter *fbc;
c67ad917 AM	236
5588f5af	237	compute_batch_value(cpu);
c67ad917	238
ebd8fef3	239	spin_lock_irq(&percpu_counters_lock);
c67ad917 AM	240	list_for_each_entry(fbc, &percpu_counters, list) {
	241	s32 *pcount;
	242
aaf0f2fa	243	raw_spin_lock(&fbc->lock);
c67ad917 AM	244	pcount = per_cpu_ptr(fbc->counters, cpu);
	245	fbc->count += *pcount;
	246	*pcount = 0;
aaf0f2fa	247	raw_spin_unlock(&fbc->lock);
c67ad917	248	}
ebd8fef3	249	spin_unlock_irq(&percpu_counters_lock);
179f7ebf	250	#endif
5588f5af	251	return 0;
c67ad917 AM	252	}
c67ad917 AM	253
27f5e0f6 TC	254	/*
	255	* Compare counter against given value.
	256	* Return 1 if greater, 0 if equal and -1 if less
	257	*/
80188b0d	258	int __percpu_counter_compare(struct percpu_counter *fbc, s64 rhs, s32 batch)
27f5e0f6 TC	259	{
	260	s64 count;
	261
	262	count = percpu_counter_read(fbc);
	263	/* Check to see if rough count will be sufficient for comparison */
80188b0d	264	if (abs(count - rhs) > (batch * num_online_cpus())) {
27f5e0f6 TC	265	if (count > rhs)
	266	return 1;
	267	else
	268	return -1;
	269	}
	270	/* Need to use precise count */
	271	count = percpu_counter_sum(fbc);
	272	if (count > rhs)
	273	return 1;
	274	else if (count < rhs)
	275	return -1;
	276	else
	277	return 0;
	278	}
80188b0d	279	EXPORT_SYMBOL(__percpu_counter_compare);
27f5e0f6	280
beb98686	281	/*
1431996b HD	282	* Compare counter, and add amount if total is: less than or equal to limit if
	283	* amount is positive, or greater than or equal to limit if amount is negative.
	284	* Return true if amount is added, or false if total would be beyond the limit.
	285	*
	286	* Negative limit is allowed, but unusual.
	287	* When negative amounts (subs) are given to percpu_counter_limited_add(),
	288	* the limit would most naturally be 0 - but other limits are also allowed.
	289	*
	290	* Overflow beyond S64_MAX is not allowed for: counter, limit and amount
	291	* are all assumed to be sane (far from S64_MIN and S64_MAX).
beb98686 HD	292	*/
	293	bool __percpu_counter_limited_add(struct percpu_counter *fbc,
	294	s64 limit, s64 amount, s32 batch)
	295	{
	296	s64 count;
	297	s64 unknown;
	298	unsigned long flags;
1431996b	299	bool good = false;
beb98686	300
1431996b HD	301	if (amount == 0)
1431996b HD	302	return true;
beb98686 HD	303
	304	local_irq_save(flags);
	305	unknown = batch * num_online_cpus();
	306	count = __this_cpu_read(*fbc->counters);
	307
	308	/* Skip taking the lock when safe */
	309	if (abs(count + amount) <= batch &&
1431996b HD	310	((amount > 0 && fbc->count + unknown <= limit) \|\|
1431996b HD	311	(amount < 0 && fbc->count - unknown >= limit))) {
beb98686 HD	312	this_cpu_add(*fbc->counters, amount);
	313	local_irq_restore(flags);
	314	return true;
	315	}
	316
	317	raw_spin_lock(&fbc->lock);
	318	count = fbc->count + amount;
	319
	320	/* Skip percpu_counter_sum() when safe */
1431996b HD	321	if (amount > 0) {
	322	if (count - unknown > limit)
	323	goto out;
	324	if (count + unknown <= limit)
	325	good = true;
	326	} else {
	327	if (count + unknown < limit)
	328	goto out;
	329	if (count - unknown >= limit)
	330	good = true;
	331	}
	332
	333	if (!good) {
beb98686 HD	334	s32 *pcount;
	335	int cpu;
	336
	337	for_each_cpu_or(cpu, cpu_online_mask, cpu_dying_mask) {
	338	pcount = per_cpu_ptr(fbc->counters, cpu);
	339	count += *pcount;
	340	}
1431996b HD	341	if (amount > 0) {
	342	if (count > limit)
	343	goto out;
	344	} else {
	345	if (count < limit)
	346	goto out;
	347	}
	348	good = true;
beb98686 HD	349	}
beb98686 HD	350
1431996b HD	351	count = __this_cpu_read(*fbc->counters);
	352	fbc->count += count + amount;
	353	__this_cpu_sub(*fbc->counters, count);
	354	out:
beb98686 HD	355	raw_spin_unlock(&fbc->lock);
	356	local_irq_restore(flags);
	357	return good;
	358	}
	359
c67ad917 AM	360	static int __init percpu_counter_startup(void)
c67ad917 AM	361	{
5588f5af SAS	362	int ret;
	363
	364	ret = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "lib/percpu_cnt:online",
	365	compute_batch_value, NULL);
	366	WARN_ON(ret < 0);
	367	ret = cpuhp_setup_state_nocalls(CPUHP_PERCPU_CNT_DEAD,
	368	"lib/percpu_cnt:dead", NULL,
	369	percpu_counter_cpu_dead);
	370	WARN_ON(ret < 0);
c67ad917 AM	371	return 0;
	372	}
	373	module_init(percpu_counter_startup);