[deliverable/linux.git] / kernel / bpf / hashtab.c

/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of version 2 of the GNU General Public
 * License as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 * General Public License for more details.
 */
#include <linux/bpf.h>
#include <linux/jhash.h>
#include <linux/filter.h>
#include <linux/vmalloc.h>

struct bucket {
	struct hlist_head head;
	raw_spinlock_t lock;
};

struct bpf_htab {
	struct bpf_map map;
	struct bucket *buckets;
	atomic_t count;	/* number of elements in this hashtable */
	u32 n_buckets;	/* number of hash buckets */
	u32 elem_size;	/* size of each element in bytes */
};

/* each htab element is struct htab_elem + key + value */
struct htab_elem {
	struct hlist_node hash_node;
	struct rcu_head rcu;
	union {
		u32 hash;
		u32 key_size;
	};
	char key[0] __aligned(8);
};

/* Called from syscall */
static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
{
	bool percpu = attr->map_type == BPF_MAP_TYPE_PERCPU_HASH;
	struct bpf_htab *htab;
	int err, i;
	u64 cost;

	htab = kzalloc(sizeof(*htab), GFP_USER);
	if (!htab)
		return ERR_PTR(-ENOMEM);

	/* mandatory map attributes */
	htab->map.map_type = attr->map_type;
	htab->map.key_size = attr->key_size;
	htab->map.value_size = attr->value_size;
	htab->map.max_entries = attr->max_entries;

	/* check sanity of attributes.
	 * value_size == 0 may be allowed in the future to use map as a set
	 */
	err = -EINVAL;
	if (htab->map.max_entries == 0 || htab->map.key_size == 0 ||
	    htab->map.value_size == 0)
		goto free_htab;

	/* hash table size must be power of 2 */
	htab->n_buckets = roundup_pow_of_two(htab->map.max_entries);

	err = -E2BIG;
	if (htab->map.key_size > MAX_BPF_STACK)
		/* eBPF programs initialize keys on stack, so they cannot be
		 * larger than max stack size
		 */
		goto free_htab;

	if (htab->map.value_size >= (1 << (KMALLOC_SHIFT_MAX - 1)) -
	    MAX_BPF_STACK - sizeof(struct htab_elem))
		/* if value_size is bigger, the user space won't be able to
		 * access the elements via bpf syscall. This check also makes
		 * sure that the elem_size doesn't overflow and it's
		 * kmalloc-able later in htab_map_update_elem()
		 */
		goto free_htab;

	if (percpu && round_up(htab->map.value_size, 8) > PCPU_MIN_UNIT_SIZE)
		/* make sure the size for pcpu_alloc() is reasonable */
		goto free_htab;

	htab->elem_size = sizeof(struct htab_elem) +
			  round_up(htab->map.key_size, 8);
	if (percpu)
		htab->elem_size += sizeof(void *);
	else
		htab->elem_size += htab->map.value_size;

	/* prevent zero size kmalloc and check for u32 overflow */
	if (htab->n_buckets == 0 ||
	    htab->n_buckets > U32_MAX / sizeof(struct bucket))
		goto free_htab;

	cost = (u64) htab->n_buckets * sizeof(struct bucket) +
	       (u64) htab->elem_size * htab->map.max_entries;

	if (percpu)
		cost += (u64) round_up(htab->map.value_size, 8) *
			num_possible_cpus() * htab->map.max_entries;

	if (cost >= U32_MAX - PAGE_SIZE)
		/* make sure page count doesn't overflow */
		goto free_htab;

	htab->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;

	err = -ENOMEM;
	htab->buckets = kmalloc_array(htab->n_buckets, sizeof(struct bucket),
				      GFP_USER | __GFP_NOWARN);

	if (!htab->buckets) {
		htab->buckets = vmalloc(htab->n_buckets * sizeof(struct bucket));
		if (!htab->buckets)
			goto free_htab;
	}

	for (i = 0; i < htab->n_buckets; i++) {
		INIT_HLIST_HEAD(&htab->buckets[i].head);
		raw_spin_lock_init(&htab->buckets[i].lock);
	}

	atomic_set(&htab->count, 0);

	return &htab->map;

free_htab:
	kfree(htab);
	return ERR_PTR(err);
}

static inline u32 htab_map_hash(const void *key, u32 key_len)
{
	return jhash(key, key_len, 0);
}

static inline struct bucket *__select_bucket(struct bpf_htab *htab, u32 hash)
{
	return &htab->buckets[hash & (htab->n_buckets - 1)];
}

static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash)
{
	return &__select_bucket(htab, hash)->head;
}

static struct htab_elem *lookup_elem_raw(struct hlist_head *head, u32 hash,
					 void *key, u32 key_size)
{
	struct htab_elem *l;

	hlist_for_each_entry_rcu(l, head, hash_node)
		if (l->hash == hash && !memcmp(&l->key, key, key_size))
			return l;

	return NULL;
}

/* Called from syscall or from eBPF program */
static void *__htab_map_lookup_elem(struct bpf_map *map, void *key)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct hlist_head *head;
	struct htab_elem *l;
	u32 hash, key_size;

	/* Must be called with rcu_read_lock. */
	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);

	head = select_bucket(htab, hash);

	l = lookup_elem_raw(head, hash, key, key_size);

	return l;
}

static void *htab_map_lookup_elem(struct bpf_map *map, void *key)
{
	struct htab_elem *l = __htab_map_lookup_elem(map, key);

	if (l)
		return l->key + round_up(map->key_size, 8);

	return NULL;
}

/* Called from syscall */
static int htab_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct hlist_head *head;
	struct htab_elem *l, *next_l;
	u32 hash, key_size;
	int i;

	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);

	head = select_bucket(htab, hash);

	/* lookup the key */
	l = lookup_elem_raw(head, hash, key, key_size);

	if (!l) {
		i = 0;
		goto find_first_elem;
	}

	/* key was found, get next key in the same bucket */
	next_l = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&l->hash_node)),
				  struct htab_elem, hash_node);

	if (next_l) {
		/* if next elem in this hash list is non-zero, just return it */
		memcpy(next_key, next_l->key, key_size);
		return 0;
	}

	/* no more elements in this hash list, go to the next bucket */
	i = hash & (htab->n_buckets - 1);
	i++;

find_first_elem:
	/* iterate over buckets */
	for (; i < htab->n_buckets; i++) {
		head = select_bucket(htab, i);

		/* pick first element in the bucket */
		next_l = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
					  struct htab_elem, hash_node);
		if (next_l) {
			/* if it's not empty, just return it */
			memcpy(next_key, next_l->key, key_size);
			return 0;
		}
	}

	/* itereated over all buckets and all elements */
	return -ENOENT;
}


static inline void htab_elem_set_ptr(struct htab_elem *l, u32 key_size,
				     void __percpu *pptr)
{
	*(void __percpu **)(l->key + key_size) = pptr;
}

static inline void __percpu *htab_elem_get_ptr(struct htab_elem *l, u32 key_size)
{
	return *(void __percpu **)(l->key + key_size);
}

static void htab_percpu_elem_free(struct htab_elem *l)
{
	free_percpu(htab_elem_get_ptr(l, l->key_size));
	kfree(l);
}

static void htab_percpu_elem_free_rcu(struct rcu_head *head)
{
	struct htab_elem *l = container_of(head, struct htab_elem, rcu);

	htab_percpu_elem_free(l);
}

static void free_htab_elem(struct htab_elem *l, bool percpu, u32 key_size)
{
	if (percpu) {
		l->key_size = key_size;
		call_rcu(&l->rcu, htab_percpu_elem_free_rcu);
	} else {
		kfree_rcu(l, rcu);
	}
}

static struct htab_elem *alloc_htab_elem(struct bpf_htab *htab, void *key,
					 void *value, u32 key_size, u32 hash,
					 bool percpu, bool onallcpus)
{
	u32 size = htab->map.value_size;
	struct htab_elem *l_new;
	void __percpu *pptr;

	l_new = kmalloc(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN);
	if (!l_new)
		return NULL;

	memcpy(l_new->key, key, key_size);
	if (percpu) {
		/* round up value_size to 8 bytes */
		size = round_up(size, 8);

		/* alloc_percpu zero-fills */
		pptr = __alloc_percpu_gfp(size, 8, GFP_ATOMIC | __GFP_NOWARN);
		if (!pptr) {
			kfree(l_new);
			return NULL;
		}

		if (!onallcpus) {
			/* copy true value_size bytes */
			memcpy(this_cpu_ptr(pptr), value, htab->map.value_size);
		} else {
			int off = 0, cpu;

			for_each_possible_cpu(cpu) {
				bpf_long_memcpy(per_cpu_ptr(pptr, cpu),
						value + off, size);
				off += size;
			}
		}
		htab_elem_set_ptr(l_new, key_size, pptr);
	} else {
		memcpy(l_new->key + round_up(key_size, 8), value, size);
	}

	l_new->hash = hash;
	return l_new;
}

static int check_flags(struct bpf_htab *htab, struct htab_elem *l_old,
		       u64 map_flags)
{
	if (!l_old && unlikely(atomic_read(&htab->count) >= htab->map.max_entries))
		/* if elem with this 'key' doesn't exist and we've reached
		 * max_entries limit, fail insertion of new elem
		 */
		return -E2BIG;

	if (l_old && map_flags == BPF_NOEXIST)
		/* elem already exists */
		return -EEXIST;

	if (!l_old && map_flags == BPF_EXIST)
		/* elem doesn't exist, cannot update it */
		return -ENOENT;

	return 0;
}

/* Called from syscall or from eBPF program */
static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
				u64 map_flags)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct htab_elem *l_new = NULL, *l_old;
	struct hlist_head *head;
	unsigned long flags;
	struct bucket *b;
	u32 key_size, hash;
	int ret;

	if (unlikely(map_flags > BPF_EXIST))
		/* unknown flags */
		return -EINVAL;

	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);

	/* allocate new element outside of the lock, since
	 * we're most likley going to insert it
	 */
	l_new = alloc_htab_elem(htab, key, value, key_size, hash, false, false);
	if (!l_new)
		return -ENOMEM;

	b = __select_bucket(htab, hash);
	head = &b->head;

	/* bpf_map_update_elem() can be called in_irq() */
	raw_spin_lock_irqsave(&b->lock, flags);

	l_old = lookup_elem_raw(head, hash, key, key_size);

	ret = check_flags(htab, l_old, map_flags);
	if (ret)
		goto err;

	/* add new element to the head of the list, so that
	 * concurrent search will find it before old elem
	 */
	hlist_add_head_rcu(&l_new->hash_node, head);
	if (l_old) {
		hlist_del_rcu(&l_old->hash_node);
		kfree_rcu(l_old, rcu);
	} else {
		atomic_inc(&htab->count);
	}
	raw_spin_unlock_irqrestore(&b->lock, flags);
	return 0;
err:
	raw_spin_unlock_irqrestore(&b->lock, flags);
	kfree(l_new);
	return ret;
}

static int __htab_percpu_map_update_elem(struct bpf_map *map, void *key,
					 void *value, u64 map_flags,
					 bool onallcpus)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct htab_elem *l_new = NULL, *l_old;
	struct hlist_head *head;
	unsigned long flags;
	struct bucket *b;
	u32 key_size, hash;
	int ret;

	if (unlikely(map_flags > BPF_EXIST))
		/* unknown flags */
		return -EINVAL;

	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);

	b = __select_bucket(htab, hash);
	head = &b->head;

	/* bpf_map_update_elem() can be called in_irq() */
	raw_spin_lock_irqsave(&b->lock, flags);

	l_old = lookup_elem_raw(head, hash, key, key_size);

	ret = check_flags(htab, l_old, map_flags);
	if (ret)
		goto err;

	if (l_old) {
		void __percpu *pptr = htab_elem_get_ptr(l_old, key_size);
		u32 size = htab->map.value_size;

		/* per-cpu hash map can update value in-place */
		if (!onallcpus) {
			memcpy(this_cpu_ptr(pptr), value, size);
		} else {
			int off = 0, cpu;

			size = round_up(size, 8);
			for_each_possible_cpu(cpu) {
				bpf_long_memcpy(per_cpu_ptr(pptr, cpu),
						value + off, size);
				off += size;
			}
		}
	} else {
		l_new = alloc_htab_elem(htab, key, value, key_size,
					hash, true, onallcpus);
		if (!l_new) {
			ret = -ENOMEM;
			goto err;
		}
		hlist_add_head_rcu(&l_new->hash_node, head);
		atomic_inc(&htab->count);
	}
	ret = 0;
err:
	raw_spin_unlock_irqrestore(&b->lock, flags);
	return ret;
}

static int htab_percpu_map_update_elem(struct bpf_map *map, void *key,
				       void *value, u64 map_flags)
{
	return __htab_percpu_map_update_elem(map, key, value, map_flags, false);
}

/* Called from syscall or from eBPF program */
static int htab_map_delete_elem(struct bpf_map *map, void *key)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	bool percpu = map->map_type == BPF_MAP_TYPE_PERCPU_HASH;
	struct hlist_head *head;
	struct bucket *b;
	struct htab_elem *l;
	unsigned long flags;
	u32 hash, key_size;
	int ret = -ENOENT;

	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);
	b = __select_bucket(htab, hash);
	head = &b->head;

	raw_spin_lock_irqsave(&b->lock, flags);

	l = lookup_elem_raw(head, hash, key, key_size);

	if (l) {
		hlist_del_rcu(&l->hash_node);
		atomic_dec(&htab->count);
		free_htab_elem(l, percpu, key_size);
		ret = 0;
	}

	raw_spin_unlock_irqrestore(&b->lock, flags);
	return ret;
}

static void delete_all_elements(struct bpf_htab *htab)
{
	int i;

	for (i = 0; i < htab->n_buckets; i++) {
		struct hlist_head *head = select_bucket(htab, i);
		struct hlist_node *n;
		struct htab_elem *l;

		hlist_for_each_entry_safe(l, n, head, hash_node) {
			hlist_del_rcu(&l->hash_node);
			atomic_dec(&htab->count);
			if (htab->map.map_type == BPF_MAP_TYPE_PERCPU_HASH) {
				l->key_size = htab->map.key_size;
				htab_percpu_elem_free(l);
			} else {
				kfree(l);
			}
		}
	}
}

/* Called when map->refcnt goes to zero, either from workqueue or from syscall */
static void htab_map_free(struct bpf_map *map)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);

	/* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
	 * so the programs (can be more than one that used this map) were
	 * disconnected from events. Wait for outstanding critical sections in
	 * these programs to complete
	 */
	synchronize_rcu();

	/* some of kfree_rcu() callbacks for elements of this map may not have
	 * executed. It's ok. Proceed to free residual elements and map itself
	 */
	delete_all_elements(htab);
	kvfree(htab->buckets);
	kfree(htab);
}

static const struct bpf_map_ops htab_ops = {
	.map_alloc = htab_map_alloc,
	.map_free = htab_map_free,
	.map_get_next_key = htab_map_get_next_key,
	.map_lookup_elem = htab_map_lookup_elem,
	.map_update_elem = htab_map_update_elem,
	.map_delete_elem = htab_map_delete_elem,
};

static struct bpf_map_type_list htab_type __read_mostly = {
	.ops = &htab_ops,
	.type = BPF_MAP_TYPE_HASH,
};

/* Called from eBPF program */
static void *htab_percpu_map_lookup_elem(struct bpf_map *map, void *key)
{
	struct htab_elem *l = __htab_map_lookup_elem(map, key);

	if (l)
		return this_cpu_ptr(htab_elem_get_ptr(l, map->key_size));
	else
		return NULL;
}

int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value)
{
	struct htab_elem *l;
	void __percpu *pptr;
	int ret = -ENOENT;
	int cpu, off = 0;
	u32 size;

	/* per_cpu areas are zero-filled and bpf programs can only
	 * access 'value_size' of them, so copying rounded areas
	 * will not leak any kernel data
	 */
	size = round_up(map->value_size, 8);
	rcu_read_lock();
	l = __htab_map_lookup_elem(map, key);
	if (!l)
		goto out;
	pptr = htab_elem_get_ptr(l, map->key_size);
	for_each_possible_cpu(cpu) {
		bpf_long_memcpy(value + off,
				per_cpu_ptr(pptr, cpu), size);
		off += size;
	}
	ret = 0;
out:
	rcu_read_unlock();
	return ret;
}

int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
			   u64 map_flags)
{
	return __htab_percpu_map_update_elem(map, key, value, map_flags, true);
}

static const struct bpf_map_ops htab_percpu_ops = {
	.map_alloc = htab_map_alloc,
	.map_free = htab_map_free,
	.map_get_next_key = htab_map_get_next_key,
	.map_lookup_elem = htab_percpu_map_lookup_elem,
	.map_update_elem = htab_percpu_map_update_elem,
	.map_delete_elem = htab_map_delete_elem,
};

static struct bpf_map_type_list htab_percpu_type __read_mostly = {
	.ops = &htab_percpu_ops,
	.type = BPF_MAP_TYPE_PERCPU_HASH,
};

static int __init register_htab_map(void)
{
	bpf_register_map_type(&htab_type);
	bpf_register_map_type(&htab_percpu_type);
	return 0;
}
late_initcall(register_htab_map);
Commit	Line	Data
0f8e4bd8 AS	1	/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
	2	*
	3	* This program is free software; you can redistribute it and/or
	4	* modify it under the terms of version 2 of the GNU General Public
	5	* License as published by the Free Software Foundation.
	6	*
	7	* This program is distributed in the hope that it will be useful, but
	8	* WITHOUT ANY WARRANTY; without even the implied warranty of
	9	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	10	* General Public License for more details.
	11	*/
	12	#include <linux/bpf.h>
	13	#include <linux/jhash.h>
	14	#include <linux/filter.h>
	15	#include <linux/vmalloc.h>
	16
688ecfe6	17	struct bucket {
	18	struct hlist_head head;
	19	raw_spinlock_t lock;
	20	};
	21
0f8e4bd8 AS	22	struct bpf_htab {
0f8e4bd8 AS	23	struct bpf_map map;
688ecfe6	24	struct bucket *buckets;
6591f1e6	25	atomic_t count; /* number of elements in this hashtable */
0f8e4bd8 AS	26	u32 n_buckets; /* number of hash buckets */
	27	u32 elem_size; /* size of each element in bytes */
	28	};
	29
	30	/* each htab element is struct htab_elem + key + value */
	31	struct htab_elem {
	32	struct hlist_node hash_node;
	33	struct rcu_head rcu;
824bd0ce AS	34	union {
	35	u32 hash;
	36	u32 key_size;
	37	};
0f8e4bd8 AS	38	char key[0] __aligned(8);
	39	};
	40
	41	/* Called from syscall */
	42	static struct bpf_map htab_map_alloc(union bpf_attr attr)
	43	{
824bd0ce	44	bool percpu = attr->map_type == BPF_MAP_TYPE_PERCPU_HASH;
0f8e4bd8 AS	45	struct bpf_htab *htab;
0f8e4bd8 AS	46	int err, i;
824bd0ce	47	u64 cost;
0f8e4bd8 AS	48
	49	htab = kzalloc(sizeof(*htab), GFP_USER);
	50	if (!htab)
	51	return ERR_PTR(-ENOMEM);
	52
	53	/* mandatory map attributes */
824bd0ce	54	htab->map.map_type = attr->map_type;
0f8e4bd8 AS	55	htab->map.key_size = attr->key_size;
	56	htab->map.value_size = attr->value_size;
	57	htab->map.max_entries = attr->max_entries;
	58
	59	/* check sanity of attributes.
	60	* value_size == 0 may be allowed in the future to use map as a set
	61	*/
	62	err = -EINVAL;
	63	if (htab->map.max_entries == 0 \|\| htab->map.key_size == 0 \|\|
	64	htab->map.value_size == 0)
	65	goto free_htab;
	66
	67	/* hash table size must be power of 2 */
	68	htab->n_buckets = roundup_pow_of_two(htab->map.max_entries);
	69
	70	err = -E2BIG;
	71	if (htab->map.key_size > MAX_BPF_STACK)
	72	/* eBPF programs initialize keys on stack, so they cannot be
	73	* larger than max stack size
	74	*/
	75	goto free_htab;
	76
01b3f521 AS	77	if (htab->map.value_size >= (1 << (KMALLOC_SHIFT_MAX - 1)) -
	78	MAX_BPF_STACK - sizeof(struct htab_elem))
	79	/* if value_size is bigger, the user space won't be able to
	80	* access the elements via bpf syscall. This check also makes
	81	* sure that the elem_size doesn't overflow and it's
	82	* kmalloc-able later in htab_map_update_elem()
	83	*/
	84	goto free_htab;
	85
824bd0ce AS	86	if (percpu && round_up(htab->map.value_size, 8) > PCPU_MIN_UNIT_SIZE)
	87	/* make sure the size for pcpu_alloc() is reasonable */
	88	goto free_htab;
	89
01b3f521	90	htab->elem_size = sizeof(struct htab_elem) +
824bd0ce AS	91	round_up(htab->map.key_size, 8);
	92	if (percpu)
	93	htab->elem_size += sizeof(void *);
	94	else
	95	htab->elem_size += htab->map.value_size;
01b3f521	96
daaf427c AS	97	/* prevent zero size kmalloc and check for u32 overflow */
daaf427c AS	98	if (htab->n_buckets == 0 \|\|
688ecfe6	99	htab->n_buckets > U32_MAX / sizeof(struct bucket))
daaf427c AS	100	goto free_htab;
daaf427c AS	101
824bd0ce AS	102	cost = (u64) htab->n_buckets * sizeof(struct bucket) +
	103	(u64) htab->elem_size * htab->map.max_entries;
	104
	105	if (percpu)
	106	cost += (u64) round_up(htab->map.value_size, 8) *
	107	num_possible_cpus() * htab->map.max_entries;
	108
	109	if (cost >= U32_MAX - PAGE_SIZE)
01b3f521 AS	110	/* make sure page count doesn't overflow */
	111	goto free_htab;
	112
824bd0ce	113	htab->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
01b3f521 AS	114
01b3f521 AS	115	err = -ENOMEM;
688ecfe6	116	htab->buckets = kmalloc_array(htab->n_buckets, sizeof(struct bucket),
0f8e4bd8 AS	117	GFP_USER \| __GFP_NOWARN);
	118
	119	if (!htab->buckets) {
688ecfe6	120	htab->buckets = vmalloc(htab->n_buckets * sizeof(struct bucket));
0f8e4bd8 AS	121	if (!htab->buckets)
	122	goto free_htab;
	123	}
	124
688ecfe6	125	for (i = 0; i < htab->n_buckets; i++) {
	126	INIT_HLIST_HEAD(&htab->buckets[i].head);
	127	raw_spin_lock_init(&htab->buckets[i].lock);
	128	}
0f8e4bd8	129
6591f1e6	130	atomic_set(&htab->count, 0);
0f8e4bd8	131
0f8e4bd8 AS	132	return &htab->map;
	133
	134	free_htab:
	135	kfree(htab);
	136	return ERR_PTR(err);
	137	}
	138
	139	static inline u32 htab_map_hash(const void *key, u32 key_len)
	140	{
	141	return jhash(key, key_len, 0);
	142	}
	143
688ecfe6	144	static inline struct bucket __select_bucket(struct bpf_htab htab, u32 hash)
0f8e4bd8 AS	145	{
	146	return &htab->buckets[hash & (htab->n_buckets - 1)];
	147	}
	148
688ecfe6	149	static inline struct hlist_head select_bucket(struct bpf_htab htab, u32 hash)
	150	{
	151	return &__select_bucket(htab, hash)->head;
	152	}
	153
0f8e4bd8 AS	154	static struct htab_elem lookup_elem_raw(struct hlist_head head, u32 hash,
	155	void *key, u32 key_size)
	156	{
	157	struct htab_elem *l;
	158
	159	hlist_for_each_entry_rcu(l, head, hash_node)
	160	if (l->hash == hash && !memcmp(&l->key, key, key_size))
	161	return l;
	162
	163	return NULL;
	164	}
	165
	166	/* Called from syscall or from eBPF program */
824bd0ce	167	static void __htab_map_lookup_elem(struct bpf_map map, void *key)
0f8e4bd8 AS	168	{
	169	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	170	struct hlist_head *head;
	171	struct htab_elem *l;
	172	u32 hash, key_size;
	173
	174	/* Must be called with rcu_read_lock. */
	175	WARN_ON_ONCE(!rcu_read_lock_held());
	176
	177	key_size = map->key_size;
	178
	179	hash = htab_map_hash(key, key_size);
	180
	181	head = select_bucket(htab, hash);
	182
	183	l = lookup_elem_raw(head, hash, key, key_size);
	184
824bd0ce AS	185	return l;
	186	}
	187
	188	static void htab_map_lookup_elem(struct bpf_map map, void *key)
	189	{
	190	struct htab_elem *l = __htab_map_lookup_elem(map, key);
	191
0f8e4bd8 AS	192	if (l)
	193	return l->key + round_up(map->key_size, 8);
	194
	195	return NULL;
	196	}
	197
	198	/* Called from syscall */
	199	static int htab_map_get_next_key(struct bpf_map map, void key, void *next_key)
	200	{
	201	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	202	struct hlist_head *head;
	203	struct htab_elem l, next_l;
	204	u32 hash, key_size;
	205	int i;
	206
	207	WARN_ON_ONCE(!rcu_read_lock_held());
	208
	209	key_size = map->key_size;
	210
	211	hash = htab_map_hash(key, key_size);
	212
	213	head = select_bucket(htab, hash);
	214
	215	/* lookup the key */
	216	l = lookup_elem_raw(head, hash, key, key_size);
	217
	218	if (!l) {
	219	i = 0;
	220	goto find_first_elem;
	221	}
	222
	223	/* key was found, get next key in the same bucket */
	224	next_l = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&l->hash_node)),
	225	struct htab_elem, hash_node);
	226
	227	if (next_l) {
	228	/* if next elem in this hash list is non-zero, just return it */
	229	memcpy(next_key, next_l->key, key_size);
	230	return 0;
	231	}
	232
	233	/* no more elements in this hash list, go to the next bucket */
	234	i = hash & (htab->n_buckets - 1);
	235	i++;
	236
	237	find_first_elem:
	238	/* iterate over buckets */
	239	for (; i < htab->n_buckets; i++) {
	240	head = select_bucket(htab, i);
	241
	242	/* pick first element in the bucket */
	243	next_l = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
	244	struct htab_elem, hash_node);
	245	if (next_l) {
	246	/* if it's not empty, just return it */
	247	memcpy(next_key, next_l->key, key_size);
	248	return 0;
	249	}
	250	}
	251
	252	/* itereated over all buckets and all elements */
	253	return -ENOENT;
	254	}
	255
824bd0ce AS	256
	257	static inline void htab_elem_set_ptr(struct htab_elem *l, u32 key_size,
	258	void __percpu *pptr)
	259	{
	260	(void __percpu *)(l->key + key_size) = pptr;
	261	}
	262
	263	static inline void __percpu htab_elem_get_ptr(struct htab_elem l, u32 key_size)
	264	{
	265	return (void __percpu *)(l->key + key_size);
	266	}
	267
	268	static void htab_percpu_elem_free(struct htab_elem *l)
	269	{
	270	free_percpu(htab_elem_get_ptr(l, l->key_size));
	271	kfree(l);
	272	}
	273
	274	static void htab_percpu_elem_free_rcu(struct rcu_head *head)
	275	{
	276	struct htab_elem *l = container_of(head, struct htab_elem, rcu);
	277
	278	htab_percpu_elem_free(l);
	279	}
	280
	281	static void free_htab_elem(struct htab_elem *l, bool percpu, u32 key_size)
	282	{
	283	if (percpu) {
	284	l->key_size = key_size;
	285	call_rcu(&l->rcu, htab_percpu_elem_free_rcu);
	286	} else {
	287	kfree_rcu(l, rcu);
	288	}
	289	}
	290
	291	static struct htab_elem alloc_htab_elem(struct bpf_htab htab, void *key,
	292	void *value, u32 key_size, u32 hash,
15a07b33	293	bool percpu, bool onallcpus)
824bd0ce AS	294	{
	295	u32 size = htab->map.value_size;
	296	struct htab_elem *l_new;
	297	void __percpu *pptr;
	298
	299	l_new = kmalloc(htab->elem_size, GFP_ATOMIC \| __GFP_NOWARN);
	300	if (!l_new)
	301	return NULL;
	302
	303	memcpy(l_new->key, key, key_size);
	304	if (percpu) {
	305	/* round up value_size to 8 bytes */
	306	size = round_up(size, 8);
	307
	308	/* alloc_percpu zero-fills */
	309	pptr = __alloc_percpu_gfp(size, 8, GFP_ATOMIC \| __GFP_NOWARN);
	310	if (!pptr) {
	311	kfree(l_new);
	312	return NULL;
	313	}
	314
15a07b33 AS	315	if (!onallcpus) {
	316	/* copy true value_size bytes */
	317	memcpy(this_cpu_ptr(pptr), value, htab->map.value_size);
	318	} else {
	319	int off = 0, cpu;
	320
	321	for_each_possible_cpu(cpu) {
	322	bpf_long_memcpy(per_cpu_ptr(pptr, cpu),
	323	value + off, size);
	324	off += size;
	325	}
	326	}
824bd0ce AS	327	htab_elem_set_ptr(l_new, key_size, pptr);
	328	} else {
	329	memcpy(l_new->key + round_up(key_size, 8), value, size);
	330	}
	331
	332	l_new->hash = hash;
	333	return l_new;
	334	}
	335
	336	static int check_flags(struct bpf_htab htab, struct htab_elem l_old,
	337	u64 map_flags)
	338	{
	339	if (!l_old && unlikely(atomic_read(&htab->count) >= htab->map.max_entries))
	340	/* if elem with this 'key' doesn't exist and we've reached
	341	* max_entries limit, fail insertion of new elem
	342	*/
	343	return -E2BIG;
	344
	345	if (l_old && map_flags == BPF_NOEXIST)
	346	/* elem already exists */
	347	return -EEXIST;
	348
	349	if (!l_old && map_flags == BPF_EXIST)
	350	/* elem doesn't exist, cannot update it */
	351	return -ENOENT;
	352
	353	return 0;
	354	}
	355
0f8e4bd8 AS	356	/* Called from syscall or from eBPF program */
	357	static int htab_map_update_elem(struct bpf_map map, void key, void *value,
	358	u64 map_flags)
	359	{
	360	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
824bd0ce	361	struct htab_elem l_new = NULL, l_old;
0f8e4bd8 AS	362	struct hlist_head *head;
0f8e4bd8 AS	363	unsigned long flags;
824bd0ce AS	364	struct bucket *b;
824bd0ce AS	365	u32 key_size, hash;
0f8e4bd8 AS	366	int ret;
0f8e4bd8 AS	367
824bd0ce	368	if (unlikely(map_flags > BPF_EXIST))
0f8e4bd8 AS	369	/* unknown flags */
	370	return -EINVAL;
	371
	372	WARN_ON_ONCE(!rcu_read_lock_held());
	373
0f8e4bd8 AS	374	key_size = map->key_size;
0f8e4bd8 AS	375
824bd0ce AS	376	hash = htab_map_hash(key, key_size);
	377
	378	/* allocate new element outside of the lock, since
	379	* we're most likley going to insert it
	380	*/
15a07b33	381	l_new = alloc_htab_elem(htab, key, value, key_size, hash, false, false);
824bd0ce AS	382	if (!l_new)
824bd0ce AS	383	return -ENOMEM;
0f8e4bd8	384
824bd0ce	385	b = __select_bucket(htab, hash);
688ecfe6	386	head = &b->head;
0f8e4bd8 AS	387
0f8e4bd8 AS	388	/* bpf_map_update_elem() can be called in_irq() */
688ecfe6	389	raw_spin_lock_irqsave(&b->lock, flags);
0f8e4bd8	390
824bd0ce	391	l_old = lookup_elem_raw(head, hash, key, key_size);
0f8e4bd8	392
824bd0ce AS	393	ret = check_flags(htab, l_old, map_flags);
824bd0ce AS	394	if (ret)
0f8e4bd8	395	goto err;
0f8e4bd8	396
824bd0ce AS	397	/* add new element to the head of the list, so that
824bd0ce AS	398	* concurrent search will find it before old elem
0f8e4bd8 AS	399	*/
	400	hlist_add_head_rcu(&l_new->hash_node, head);
	401	if (l_old) {
	402	hlist_del_rcu(&l_old->hash_node);
	403	kfree_rcu(l_old, rcu);
	404	} else {
6591f1e6	405	atomic_inc(&htab->count);
0f8e4bd8	406	}
688ecfe6	407	raw_spin_unlock_irqrestore(&b->lock, flags);
0f8e4bd8 AS	408	return 0;
0f8e4bd8 AS	409	err:
688ecfe6	410	raw_spin_unlock_irqrestore(&b->lock, flags);
0f8e4bd8 AS	411	kfree(l_new);
	412	return ret;
	413	}
	414
15a07b33 AS	415	static int __htab_percpu_map_update_elem(struct bpf_map map, void key,
	416	void *value, u64 map_flags,
	417	bool onallcpus)
824bd0ce AS	418	{
	419	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	420	struct htab_elem l_new = NULL, l_old;
	421	struct hlist_head *head;
	422	unsigned long flags;
	423	struct bucket *b;
	424	u32 key_size, hash;
	425	int ret;
	426
	427	if (unlikely(map_flags > BPF_EXIST))
	428	/* unknown flags */
	429	return -EINVAL;
	430
	431	WARN_ON_ONCE(!rcu_read_lock_held());
	432
	433	key_size = map->key_size;
	434
	435	hash = htab_map_hash(key, key_size);
	436
	437	b = __select_bucket(htab, hash);
	438	head = &b->head;
	439
	440	/* bpf_map_update_elem() can be called in_irq() */
	441	raw_spin_lock_irqsave(&b->lock, flags);
	442
	443	l_old = lookup_elem_raw(head, hash, key, key_size);
	444
	445	ret = check_flags(htab, l_old, map_flags);
	446	if (ret)
	447	goto err;
	448
	449	if (l_old) {
15a07b33 AS	450	void __percpu *pptr = htab_elem_get_ptr(l_old, key_size);
	451	u32 size = htab->map.value_size;
	452
824bd0ce	453	/* per-cpu hash map can update value in-place */
15a07b33 AS	454	if (!onallcpus) {
	455	memcpy(this_cpu_ptr(pptr), value, size);
	456	} else {
	457	int off = 0, cpu;
	458
	459	size = round_up(size, 8);
	460	for_each_possible_cpu(cpu) {
	461	bpf_long_memcpy(per_cpu_ptr(pptr, cpu),
	462	value + off, size);
	463	off += size;
	464	}
	465	}
824bd0ce AS	466	} else {
824bd0ce AS	467	l_new = alloc_htab_elem(htab, key, value, key_size,
15a07b33	468	hash, true, onallcpus);
824bd0ce AS	469	if (!l_new) {
	470	ret = -ENOMEM;
	471	goto err;
	472	}
	473	hlist_add_head_rcu(&l_new->hash_node, head);
	474	atomic_inc(&htab->count);
	475	}
	476	ret = 0;
	477	err:
	478	raw_spin_unlock_irqrestore(&b->lock, flags);
	479	return ret;
	480	}
	481
15a07b33 AS	482	static int htab_percpu_map_update_elem(struct bpf_map map, void key,
	483	void *value, u64 map_flags)
	484	{
	485	return __htab_percpu_map_update_elem(map, key, value, map_flags, false);
	486	}
	487
0f8e4bd8 AS	488	/* Called from syscall or from eBPF program */
	489	static int htab_map_delete_elem(struct bpf_map map, void key)
	490	{
	491	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
824bd0ce	492	bool percpu = map->map_type == BPF_MAP_TYPE_PERCPU_HASH;
0f8e4bd8	493	struct hlist_head *head;
688ecfe6	494	struct bucket *b;
0f8e4bd8 AS	495	struct htab_elem *l;
	496	unsigned long flags;
	497	u32 hash, key_size;
	498	int ret = -ENOENT;
	499
	500	WARN_ON_ONCE(!rcu_read_lock_held());
	501
	502	key_size = map->key_size;
	503
	504	hash = htab_map_hash(key, key_size);
688ecfe6	505	b = __select_bucket(htab, hash);
688ecfe6	506	head = &b->head;
0f8e4bd8	507
688ecfe6	508	raw_spin_lock_irqsave(&b->lock, flags);
0f8e4bd8	509
0f8e4bd8 AS	510	l = lookup_elem_raw(head, hash, key, key_size);
	511
	512	if (l) {
	513	hlist_del_rcu(&l->hash_node);
6591f1e6	514	atomic_dec(&htab->count);
824bd0ce	515	free_htab_elem(l, percpu, key_size);
0f8e4bd8 AS	516	ret = 0;
	517	}
	518
688ecfe6	519	raw_spin_unlock_irqrestore(&b->lock, flags);
0f8e4bd8 AS	520	return ret;
	521	}
	522
	523	static void delete_all_elements(struct bpf_htab *htab)
	524	{
	525	int i;
	526
	527	for (i = 0; i < htab->n_buckets; i++) {
	528	struct hlist_head *head = select_bucket(htab, i);
	529	struct hlist_node *n;
	530	struct htab_elem *l;
	531
	532	hlist_for_each_entry_safe(l, n, head, hash_node) {
	533	hlist_del_rcu(&l->hash_node);
6591f1e6	534	atomic_dec(&htab->count);
824bd0ce AS	535	if (htab->map.map_type == BPF_MAP_TYPE_PERCPU_HASH) {
	536	l->key_size = htab->map.key_size;
	537	htab_percpu_elem_free(l);
	538	} else {
	539	kfree(l);
	540	}
0f8e4bd8 AS	541	}
	542	}
	543	}
	544
	545	/* Called when map->refcnt goes to zero, either from workqueue or from syscall */
	546	static void htab_map_free(struct bpf_map *map)
	547	{
	548	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	549
	550	/* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
	551	* so the programs (can be more than one that used this map) were
	552	* disconnected from events. Wait for outstanding critical sections in
	553	* these programs to complete
	554	*/
	555	synchronize_rcu();
	556
	557	/* some of kfree_rcu() callbacks for elements of this map may not have
	558	* executed. It's ok. Proceed to free residual elements and map itself
	559	*/
	560	delete_all_elements(htab);
	561	kvfree(htab->buckets);
	562	kfree(htab);
	563	}
	564
a2c83fff	565	static const struct bpf_map_ops htab_ops = {
0f8e4bd8 AS	566	.map_alloc = htab_map_alloc,
	567	.map_free = htab_map_free,
	568	.map_get_next_key = htab_map_get_next_key,
	569	.map_lookup_elem = htab_map_lookup_elem,
	570	.map_update_elem = htab_map_update_elem,
	571	.map_delete_elem = htab_map_delete_elem,
	572	};
	573
a2c83fff	574	static struct bpf_map_type_list htab_type __read_mostly = {
0f8e4bd8 AS	575	.ops = &htab_ops,
	576	.type = BPF_MAP_TYPE_HASH,
	577	};
	578
824bd0ce AS	579	/* Called from eBPF program */
	580	static void htab_percpu_map_lookup_elem(struct bpf_map map, void *key)
	581	{
	582	struct htab_elem *l = __htab_map_lookup_elem(map, key);
	583
	584	if (l)
	585	return this_cpu_ptr(htab_elem_get_ptr(l, map->key_size));
	586	else
	587	return NULL;
	588	}
	589
15a07b33 AS	590	int bpf_percpu_hash_copy(struct bpf_map map, void key, void *value)
	591	{
	592	struct htab_elem *l;
	593	void __percpu *pptr;
	594	int ret = -ENOENT;
	595	int cpu, off = 0;
	596	u32 size;
	597
	598	/* per_cpu areas are zero-filled and bpf programs can only
	599	* access 'value_size' of them, so copying rounded areas
	600	* will not leak any kernel data
	601	*/
	602	size = round_up(map->value_size, 8);
	603	rcu_read_lock();
	604	l = __htab_map_lookup_elem(map, key);
	605	if (!l)
	606	goto out;
	607	pptr = htab_elem_get_ptr(l, map->key_size);
	608	for_each_possible_cpu(cpu) {
	609	bpf_long_memcpy(value + off,
	610	per_cpu_ptr(pptr, cpu), size);
	611	off += size;
	612	}
	613	ret = 0;
	614	out:
	615	rcu_read_unlock();
	616	return ret;
	617	}
	618
	619	int bpf_percpu_hash_update(struct bpf_map map, void key, void *value,
	620	u64 map_flags)
	621	{
	622	return __htab_percpu_map_update_elem(map, key, value, map_flags, true);
	623	}
	624
824bd0ce AS	625	static const struct bpf_map_ops htab_percpu_ops = {
	626	.map_alloc = htab_map_alloc,
	627	.map_free = htab_map_free,
	628	.map_get_next_key = htab_map_get_next_key,
	629	.map_lookup_elem = htab_percpu_map_lookup_elem,
	630	.map_update_elem = htab_percpu_map_update_elem,
	631	.map_delete_elem = htab_map_delete_elem,
	632	};
	633
	634	static struct bpf_map_type_list htab_percpu_type __read_mostly = {
	635	.ops = &htab_percpu_ops,
	636	.type = BPF_MAP_TYPE_PERCPU_HASH,
	637	};
	638
0f8e4bd8 AS	639	static int __init register_htab_map(void)
0f8e4bd8 AS	640	{
a2c83fff	641	bpf_register_map_type(&htab_type);
824bd0ce	642	bpf_register_map_type(&htab_percpu_type);
0f8e4bd8 AS	643	return 0;
	644	}
	645	late_initcall(register_htab_map);