[linux.git] / lib / objpool.c

// SPDX-License-Identifier: GPL-2.0

#include <linux/objpool.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
#include <linux/atomic.h>
#include <linux/irqflags.h>
#include <linux/cpumask.h>
#include <linux/log2.h>

/*
 * objpool: ring-array based lockless MPMC/FIFO queues
 *
 * Copyright: wuqiang.matt@bytedance.com,mhiramat@kernel.org
 */

/* initialize percpu objpool_slot */
static int
objpool_init_percpu_slot(struct objpool_head *pool,
			 struct objpool_slot *slot,
			 int nodes, void *context,
			 objpool_init_obj_cb objinit)
{
	void *obj = (void *)&slot->entries[pool->capacity];
	int i;

	/* initialize elements of percpu objpool_slot */
	slot->mask = pool->capacity - 1;

	for (i = 0; i < nodes; i++) {
		if (objinit) {
			int rc = objinit(obj, context);
			if (rc)
				return rc;
		}
		slot->entries[slot->tail & slot->mask] = obj;
		obj = obj + pool->obj_size;
		slot->tail++;
		slot->last = slot->tail;
		pool->nr_objs++;
	}

	return 0;
}

/* allocate and initialize percpu slots */
static int
objpool_init_percpu_slots(struct objpool_head *pool, int nr_objs,
			  void *context, objpool_init_obj_cb objinit)
{
	int i, cpu_count = 0;

	for (i = 0; i < nr_cpu_ids; i++) {

		struct objpool_slot *slot;
		int nodes, size, rc;

		/* skip the cpu node which could never be present */
		if (!cpu_possible(i))
			continue;

		/* compute how many objects to be allocated with this slot */
		nodes = nr_objs / pool->nr_possible_cpus;
		if (cpu_count < (nr_objs % pool->nr_possible_cpus))
			nodes++;
		cpu_count++;

		size = struct_size(slot, entries, pool->capacity) +
			pool->obj_size * nodes;

		/*
		 * here we allocate percpu-slot & objs together in a single
		 * allocation to make it more compact, taking advantage of
		 * warm caches and TLB hits. in default vmalloc is used to
		 * reduce the pressure of kernel slab system. as we know,
		 * mimimal size of vmalloc is one page since vmalloc would
		 * always align the requested size to page size
		 */
		if (pool->gfp & GFP_ATOMIC)
			slot = kmalloc_node(size, pool->gfp, cpu_to_node(i));
		else
			slot = __vmalloc_node(size, sizeof(void *), pool->gfp,
				cpu_to_node(i), __builtin_return_address(0));
		if (!slot)
			return -ENOMEM;
		memset(slot, 0, size);
		pool->cpu_slots[i] = slot;

		/* initialize the objpool_slot of cpu node i */
		rc = objpool_init_percpu_slot(pool, slot, nodes, context, objinit);
		if (rc)
			return rc;
	}

	return 0;
}

/* cleanup all percpu slots of the object pool */
static void objpool_fini_percpu_slots(struct objpool_head *pool)
{
	int i;

	if (!pool->cpu_slots)
		return;

	for (i = 0; i < nr_cpu_ids; i++)
		kvfree(pool->cpu_slots[i]);
	kfree(pool->cpu_slots);
}

/* initialize object pool and pre-allocate objects */
int objpool_init(struct objpool_head *pool, int nr_objs, int object_size,
		gfp_t gfp, void *context, objpool_init_obj_cb objinit,
		objpool_fini_cb release)
{
	int rc, capacity, slot_size;

	/* check input parameters */
	if (nr_objs <= 0 || nr_objs > OBJPOOL_NR_OBJECT_MAX ||
	    object_size <= 0 || object_size > OBJPOOL_OBJECT_SIZE_MAX)
		return -EINVAL;

	/* align up to unsigned long size */
	object_size = ALIGN(object_size, sizeof(long));

	/* calculate capacity of percpu objpool_slot */
	capacity = roundup_pow_of_two(nr_objs);
	if (!capacity)
		return -EINVAL;

	/* initialize objpool pool */
	memset(pool, 0, sizeof(struct objpool_head));
	pool->nr_possible_cpus = num_possible_cpus();
	pool->obj_size = object_size;
	pool->capacity = capacity;
	pool->gfp = gfp & ~__GFP_ZERO;
	pool->context = context;
	pool->release = release;
	slot_size = nr_cpu_ids * sizeof(struct objpool_slot);
	pool->cpu_slots = kzalloc(slot_size, pool->gfp);
	if (!pool->cpu_slots)
		return -ENOMEM;

	/* initialize per-cpu slots */
	rc = objpool_init_percpu_slots(pool, nr_objs, context, objinit);
	if (rc)
		objpool_fini_percpu_slots(pool);
	else
		refcount_set(&pool->ref, pool->nr_objs + 1);

	return rc;
}
EXPORT_SYMBOL_GPL(objpool_init);

/* release whole objpool forcely */
void objpool_free(struct objpool_head *pool)
{
	if (!pool->cpu_slots)
		return;

	/* release percpu slots */
	objpool_fini_percpu_slots(pool);

	/* call user's cleanup callback if provided */
	if (pool->release)
		pool->release(pool, pool->context);
}
EXPORT_SYMBOL_GPL(objpool_free);

/* drop the allocated object, rather reclaim it to objpool */
int objpool_drop(void *obj, struct objpool_head *pool)
{
	if (!obj || !pool)
		return -EINVAL;

	if (refcount_dec_and_test(&pool->ref)) {
		objpool_free(pool);
		return 0;
	}

	return -EAGAIN;
}
EXPORT_SYMBOL_GPL(objpool_drop);

/* drop unused objects and defref objpool for releasing */
void objpool_fini(struct objpool_head *pool)
{
	int count = 1; /* extra ref for objpool itself */

	/* drop all remained objects from objpool */
	while (objpool_pop(pool))
		count++;

	if (refcount_sub_and_test(count, &pool->ref))
		objpool_free(pool);
}
EXPORT_SYMBOL_GPL(objpool_fini);
Commit	Line	Data
b4edb8d2	1	// SPDX-License-Identifier: GPL-2.0
	2
	3	#include <linux/objpool.h>
	4	#include <linux/slab.h>
	5	#include <linux/vmalloc.h>
	6	#include <linux/atomic.h>
	7	#include <linux/irqflags.h>
	8	#include <linux/cpumask.h>
	9	#include <linux/log2.h>
	10
	11	/*
	12	* objpool: ring-array based lockless MPMC/FIFO queues
	13	*
	14	* Copyright: wuqiang.matt@bytedance.com,mhiramat@kernel.org
	15	*/
	16
	17	/* initialize percpu objpool_slot */
	18	static int
	19	objpool_init_percpu_slot(struct objpool_head *pool,
	20	struct objpool_slot *slot,
	21	int nodes, void *context,
	22	objpool_init_obj_cb objinit)
	23	{
	24	void obj = (void )&slot->entries[pool->capacity];
	25	int i;
	26
	27	/* initialize elements of percpu objpool_slot */
	28	slot->mask = pool->capacity - 1;
	29
	30	for (i = 0; i < nodes; i++) {
	31	if (objinit) {
	32	int rc = objinit(obj, context);
	33	if (rc)
	34	return rc;
	35	}
	36	slot->entries[slot->tail & slot->mask] = obj;
	37	obj = obj + pool->obj_size;
	38	slot->tail++;
	39	slot->last = slot->tail;
	40	pool->nr_objs++;
	41	}
	42
	43	return 0;
	44	}
	45
	46	/* allocate and initialize percpu slots */
	47	static int
	48	objpool_init_percpu_slots(struct objpool_head *pool, int nr_objs,
	49	void *context, objpool_init_obj_cb objinit)
	50	{
	51	int i, cpu_count = 0;
	52
78d0b161	53	for (i = 0; i < nr_cpu_ids; i++) {
b4edb8d2	54
	55	struct objpool_slot *slot;
	56	int nodes, size, rc;
	57
	58	/* skip the cpu node which could never be present */
	59	if (!cpu_possible(i))
	60	continue;
	61
	62	/* compute how many objects to be allocated with this slot */
78d0b161 AN	63	nodes = nr_objs / pool->nr_possible_cpus;
78d0b161 AN	64	if (cpu_count < (nr_objs % pool->nr_possible_cpus))
b4edb8d2	65	nodes++;
	66	cpu_count++;
	67
	68	size = struct_size(slot, entries, pool->capacity) +
	69	pool->obj_size * nodes;
	70
	71	/*
	72	* here we allocate percpu-slot & objs together in a single
	73	* allocation to make it more compact, taking advantage of
	74	* warm caches and TLB hits. in default vmalloc is used to
	75	* reduce the pressure of kernel slab system. as we know,
	76	* mimimal size of vmalloc is one page since vmalloc would
	77	* always align the requested size to page size
	78	*/
	79	if (pool->gfp & GFP_ATOMIC)
	80	slot = kmalloc_node(size, pool->gfp, cpu_to_node(i));
	81	else
	82	slot = __vmalloc_node(size, sizeof(void *), pool->gfp,
	83	cpu_to_node(i), __builtin_return_address(0));
	84	if (!slot)
	85	return -ENOMEM;
	86	memset(slot, 0, size);
	87	pool->cpu_slots[i] = slot;
	88
	89	/* initialize the objpool_slot of cpu node i */
	90	rc = objpool_init_percpu_slot(pool, slot, nodes, context, objinit);
	91	if (rc)
	92	return rc;
	93	}
	94
	95	return 0;
	96	}
	97
	98	/* cleanup all percpu slots of the object pool */
	99	static void objpool_fini_percpu_slots(struct objpool_head *pool)
	100	{
	101	int i;
	102
	103	if (!pool->cpu_slots)
	104	return;
	105
78d0b161	106	for (i = 0; i < nr_cpu_ids; i++)
b4edb8d2	107	kvfree(pool->cpu_slots[i]);
	108	kfree(pool->cpu_slots);
	109	}
	110
	111	/* initialize object pool and pre-allocate objects */
	112	int objpool_init(struct objpool_head *pool, int nr_objs, int object_size,
	113	gfp_t gfp, void *context, objpool_init_obj_cb objinit,
	114	objpool_fini_cb release)
	115	{
	116	int rc, capacity, slot_size;
	117
	118	/* check input parameters */
	119	if (nr_objs <= 0 \|\| nr_objs > OBJPOOL_NR_OBJECT_MAX \|\|
	120	object_size <= 0 \|\| object_size > OBJPOOL_OBJECT_SIZE_MAX)
	121	return -EINVAL;
	122
	123	/* align up to unsigned long size */
	124	object_size = ALIGN(object_size, sizeof(long));
	125
	126	/* calculate capacity of percpu objpool_slot */
	127	capacity = roundup_pow_of_two(nr_objs);
	128	if (!capacity)
	129	return -EINVAL;
	130
	131	/* initialize objpool pool */
	132	memset(pool, 0, sizeof(struct objpool_head));
78d0b161	133	pool->nr_possible_cpus = num_possible_cpus();
b4edb8d2	134	pool->obj_size = object_size;
	135	pool->capacity = capacity;
	136	pool->gfp = gfp & ~__GFP_ZERO;
	137	pool->context = context;
	138	pool->release = release;
78d0b161	139	slot_size = nr_cpu_ids * sizeof(struct objpool_slot);
b4edb8d2	140	pool->cpu_slots = kzalloc(slot_size, pool->gfp);
	141	if (!pool->cpu_slots)
	142	return -ENOMEM;
	143
	144	/* initialize per-cpu slots */
	145	rc = objpool_init_percpu_slots(pool, nr_objs, context, objinit);
	146	if (rc)
	147	objpool_fini_percpu_slots(pool);
	148	else
	149	refcount_set(&pool->ref, pool->nr_objs + 1);
	150
	151	return rc;
	152	}
	153	EXPORT_SYMBOL_GPL(objpool_init);
	154
b4edb8d2	155	/* release whole objpool forcely */
	156	void objpool_free(struct objpool_head *pool)
	157	{
	158	if (!pool->cpu_slots)
	159	return;
	160
	161	/* release percpu slots */
	162	objpool_fini_percpu_slots(pool);
	163
	164	/* call user's cleanup callback if provided */
	165	if (pool->release)
	166	pool->release(pool, pool->context);
	167	}
	168	EXPORT_SYMBOL_GPL(objpool_free);
	169
	170	/* drop the allocated object, rather reclaim it to objpool */
	171	int objpool_drop(void obj, struct objpool_head pool)
	172	{
	173	if (!obj \|\| !pool)
	174	return -EINVAL;
	175
	176	if (refcount_dec_and_test(&pool->ref)) {
	177	objpool_free(pool);
	178	return 0;
	179	}
	180
	181	return -EAGAIN;
	182	}
	183	EXPORT_SYMBOL_GPL(objpool_drop);
	184
	185	/* drop unused objects and defref objpool for releasing */
	186	void objpool_fini(struct objpool_head *pool)
	187	{
	188	int count = 1; /* extra ref for objpool itself */
	189
	190	/* drop all remained objects from objpool */
	191	while (objpool_pop(pool))
	192	count++;
	193
	194	if (refcount_sub_and_test(count, &pool->ref))
	195	objpool_free(pool);
	196	}
	197	EXPORT_SYMBOL_GPL(objpool_fini);