[linux-2.6-block.git] / kernel / bpf / local_storage.c

//SPDX-License-Identifier: GPL-2.0
#include <linux/bpf-cgroup.h>
#include <linux/bpf.h>
#include <linux/btf.h>
#include <linux/bug.h>
#include <linux/filter.h>
#include <linux/mm.h>
#include <linux/rbtree.h>
#include <linux/slab.h>
#include <uapi/linux/btf.h>

DEFINE_PER_CPU(struct bpf_cgroup_storage*, bpf_cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]);

#ifdef CONFIG_CGROUP_BPF

#define LOCAL_STORAGE_CREATE_FLAG_MASK					\
	(BPF_F_NUMA_NODE | BPF_F_ACCESS_MASK)

struct bpf_cgroup_storage_map {
	struct bpf_map map;

	spinlock_t lock;
	struct bpf_prog *prog;
	struct rb_root root;
	struct list_head list;
};

static struct bpf_cgroup_storage_map *map_to_storage(struct bpf_map *map)
{
	return container_of(map, struct bpf_cgroup_storage_map, map);
}

static int bpf_cgroup_storage_key_cmp(
	const struct bpf_cgroup_storage_key *key1,
	const struct bpf_cgroup_storage_key *key2)
{
	if (key1->cgroup_inode_id < key2->cgroup_inode_id)
		return -1;
	else if (key1->cgroup_inode_id > key2->cgroup_inode_id)
		return 1;
	else if (key1->attach_type < key2->attach_type)
		return -1;
	else if (key1->attach_type > key2->attach_type)
		return 1;
	return 0;
}

static struct bpf_cgroup_storage *cgroup_storage_lookup(
	struct bpf_cgroup_storage_map *map, struct bpf_cgroup_storage_key *key,
	bool locked)
{
	struct rb_root *root = &map->root;
	struct rb_node *node;

	if (!locked)
		spin_lock_bh(&map->lock);

	node = root->rb_node;
	while (node) {
		struct bpf_cgroup_storage *storage;

		storage = container_of(node, struct bpf_cgroup_storage, node);

		switch (bpf_cgroup_storage_key_cmp(key, &storage->key)) {
		case -1:
			node = node->rb_left;
			break;
		case 1:
			node = node->rb_right;
			break;
		default:
			if (!locked)
				spin_unlock_bh(&map->lock);
			return storage;
		}
	}

	if (!locked)
		spin_unlock_bh(&map->lock);

	return NULL;
}

static int cgroup_storage_insert(struct bpf_cgroup_storage_map *map,
				 struct bpf_cgroup_storage *storage)
{
	struct rb_root *root = &map->root;
	struct rb_node **new = &(root->rb_node), *parent = NULL;

	while (*new) {
		struct bpf_cgroup_storage *this;

		this = container_of(*new, struct bpf_cgroup_storage, node);

		parent = *new;
		switch (bpf_cgroup_storage_key_cmp(&storage->key, &this->key)) {
		case -1:
			new = &((*new)->rb_left);
			break;
		case 1:
			new = &((*new)->rb_right);
			break;
		default:
			return -EEXIST;
		}
	}

	rb_link_node(&storage->node, parent, new);
	rb_insert_color(&storage->node, root);

	return 0;
}

static void *cgroup_storage_lookup_elem(struct bpf_map *_map, void *_key)
{
	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	struct bpf_cgroup_storage_key *key = _key;
	struct bpf_cgroup_storage *storage;

	storage = cgroup_storage_lookup(map, key, false);
	if (!storage)
		return NULL;

	return &READ_ONCE(storage->buf)->data[0];
}

static int cgroup_storage_update_elem(struct bpf_map *map, void *_key,
				      void *value, u64 flags)
{
	struct bpf_cgroup_storage_key *key = _key;
	struct bpf_cgroup_storage *storage;
	struct bpf_storage_buffer *new;

	if (unlikely(flags & ~(BPF_F_LOCK | BPF_EXIST | BPF_NOEXIST)))
		return -EINVAL;

	if (unlikely(flags & BPF_NOEXIST))
		return -EINVAL;

	if (unlikely((flags & BPF_F_LOCK) &&
		     !map_value_has_spin_lock(map)))
		return -EINVAL;

	storage = cgroup_storage_lookup((struct bpf_cgroup_storage_map *)map,
					key, false);
	if (!storage)
		return -ENOENT;

	if (flags & BPF_F_LOCK) {
		copy_map_value_locked(map, storage->buf->data, value, false);
		return 0;
	}

	new = kmalloc_node(sizeof(struct bpf_storage_buffer) +
			   map->value_size,
			   __GFP_ZERO | GFP_ATOMIC | __GFP_NOWARN,
			   map->numa_node);
	if (!new)
		return -ENOMEM;

	memcpy(&new->data[0], value, map->value_size);
	check_and_init_map_lock(map, new->data);

	new = xchg(&storage->buf, new);
	kfree_rcu(new, rcu);

	return 0;
}

int bpf_percpu_cgroup_storage_copy(struct bpf_map *_map, void *_key,
				   void *value)
{
	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	struct bpf_cgroup_storage_key *key = _key;
	struct bpf_cgroup_storage *storage;
	int cpu, off = 0;
	u32 size;

	rcu_read_lock();
	storage = cgroup_storage_lookup(map, key, false);
	if (!storage) {
		rcu_read_unlock();
		return -ENOENT;
	}

	/* 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);
	for_each_possible_cpu(cpu) {
		bpf_long_memcpy(value + off,
				per_cpu_ptr(storage->percpu_buf, cpu), size);
		off += size;
	}
	rcu_read_unlock();
	return 0;
}

int bpf_percpu_cgroup_storage_update(struct bpf_map *_map, void *_key,
				     void *value, u64 map_flags)
{
	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	struct bpf_cgroup_storage_key *key = _key;
	struct bpf_cgroup_storage *storage;
	int cpu, off = 0;
	u32 size;

	if (map_flags != BPF_ANY && map_flags != BPF_EXIST)
		return -EINVAL;

	rcu_read_lock();
	storage = cgroup_storage_lookup(map, key, false);
	if (!storage) {
		rcu_read_unlock();
		return -ENOENT;
	}

	/* the user space will provide round_up(value_size, 8) bytes that
	 * will be copied into per-cpu area. bpf programs can only access
	 * value_size of it. During lookup the same extra bytes will be
	 * returned or zeros which were zero-filled by percpu_alloc,
	 * so no kernel data leaks possible
	 */
	size = round_up(_map->value_size, 8);
	for_each_possible_cpu(cpu) {
		bpf_long_memcpy(per_cpu_ptr(storage->percpu_buf, cpu),
				value + off, size);
		off += size;
	}
	rcu_read_unlock();
	return 0;
}

static int cgroup_storage_get_next_key(struct bpf_map *_map, void *_key,
				       void *_next_key)
{
	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	struct bpf_cgroup_storage_key *key = _key;
	struct bpf_cgroup_storage_key *next = _next_key;
	struct bpf_cgroup_storage *storage;

	spin_lock_bh(&map->lock);

	if (list_empty(&map->list))
		goto enoent;

	if (key) {
		storage = cgroup_storage_lookup(map, key, true);
		if (!storage)
			goto enoent;

		storage = list_next_entry(storage, list);
		if (!storage)
			goto enoent;
	} else {
		storage = list_first_entry(&map->list,
					 struct bpf_cgroup_storage, list);
	}

	spin_unlock_bh(&map->lock);
	next->attach_type = storage->key.attach_type;
	next->cgroup_inode_id = storage->key.cgroup_inode_id;
	return 0;

enoent:
	spin_unlock_bh(&map->lock);
	return -ENOENT;
}

static struct bpf_map *cgroup_storage_map_alloc(union bpf_attr *attr)
{
	int numa_node = bpf_map_attr_numa_node(attr);
	struct bpf_cgroup_storage_map *map;

	if (attr->key_size != sizeof(struct bpf_cgroup_storage_key))
		return ERR_PTR(-EINVAL);

	if (attr->value_size == 0)
		return ERR_PTR(-EINVAL);

	if (attr->value_size > PAGE_SIZE)
		return ERR_PTR(-E2BIG);

	if (attr->map_flags & ~LOCAL_STORAGE_CREATE_FLAG_MASK ||
	    !bpf_map_flags_access_ok(attr->map_flags))
		return ERR_PTR(-EINVAL);

	if (attr->max_entries)
		/* max_entries is not used and enforced to be 0 */
		return ERR_PTR(-EINVAL);

	map = kmalloc_node(sizeof(struct bpf_cgroup_storage_map),
			   __GFP_ZERO | GFP_USER, numa_node);
	if (!map)
		return ERR_PTR(-ENOMEM);

	map->map.pages = round_up(sizeof(struct bpf_cgroup_storage_map),
				  PAGE_SIZE) >> PAGE_SHIFT;

	/* copy mandatory map attributes */
	bpf_map_init_from_attr(&map->map, attr);

	spin_lock_init(&map->lock);
	map->root = RB_ROOT;
	INIT_LIST_HEAD(&map->list);

	return &map->map;
}

static void cgroup_storage_map_free(struct bpf_map *_map)
{
	struct bpf_cgroup_storage_map *map = map_to_storage(_map);

	WARN_ON(!RB_EMPTY_ROOT(&map->root));
	WARN_ON(!list_empty(&map->list));

	kfree(map);
}

static int cgroup_storage_delete_elem(struct bpf_map *map, void *key)
{
	return -EINVAL;
}

static int cgroup_storage_check_btf(const struct bpf_map *map,
				    const struct btf *btf,
				    const struct btf_type *key_type,
				    const struct btf_type *value_type)
{
	struct btf_member *m;
	u32 offset, size;

	/* Key is expected to be of struct bpf_cgroup_storage_key type,
	 * which is:
	 * struct bpf_cgroup_storage_key {
	 *	__u64	cgroup_inode_id;
	 *	__u32	attach_type;
	 * };
	 */

	/*
	 * Key_type must be a structure with two fields.
	 */
	if (BTF_INFO_KIND(key_type->info) != BTF_KIND_STRUCT ||
	    BTF_INFO_VLEN(key_type->info) != 2)
		return -EINVAL;

	/*
	 * The first field must be a 64 bit integer at 0 offset.
	 */
	m = (struct btf_member *)(key_type + 1);
	size = FIELD_SIZEOF(struct bpf_cgroup_storage_key, cgroup_inode_id);
	if (!btf_member_is_reg_int(btf, key_type, m, 0, size))
		return -EINVAL;

	/*
	 * The second field must be a 32 bit integer at 64 bit offset.
	 */
	m++;
	offset = offsetof(struct bpf_cgroup_storage_key, attach_type);
	size = FIELD_SIZEOF(struct bpf_cgroup_storage_key, attach_type);
	if (!btf_member_is_reg_int(btf, key_type, m, offset, size))
		return -EINVAL;

	return 0;
}

static void cgroup_storage_seq_show_elem(struct bpf_map *map, void *_key,
					 struct seq_file *m)
{
	enum bpf_cgroup_storage_type stype = cgroup_storage_type(map);
	struct bpf_cgroup_storage_key *key = _key;
	struct bpf_cgroup_storage *storage;
	int cpu;

	rcu_read_lock();
	storage = cgroup_storage_lookup(map_to_storage(map), key, false);
	if (!storage) {
		rcu_read_unlock();
		return;
	}

	btf_type_seq_show(map->btf, map->btf_key_type_id, key, m);
	stype = cgroup_storage_type(map);
	if (stype == BPF_CGROUP_STORAGE_SHARED) {
		seq_puts(m, ": ");
		btf_type_seq_show(map->btf, map->btf_value_type_id,
				  &READ_ONCE(storage->buf)->data[0], m);
		seq_puts(m, "\n");
	} else {
		seq_puts(m, ": {\n");
		for_each_possible_cpu(cpu) {
			seq_printf(m, "\tcpu%d: ", cpu);
			btf_type_seq_show(map->btf, map->btf_value_type_id,
					  per_cpu_ptr(storage->percpu_buf, cpu),
					  m);
			seq_puts(m, "\n");
		}
		seq_puts(m, "}\n");
	}
	rcu_read_unlock();
}

const struct bpf_map_ops cgroup_storage_map_ops = {
	.map_alloc = cgroup_storage_map_alloc,
	.map_free = cgroup_storage_map_free,
	.map_get_next_key = cgroup_storage_get_next_key,
	.map_lookup_elem = cgroup_storage_lookup_elem,
	.map_update_elem = cgroup_storage_update_elem,
	.map_delete_elem = cgroup_storage_delete_elem,
	.map_check_btf = cgroup_storage_check_btf,
	.map_seq_show_elem = cgroup_storage_seq_show_elem,
};

int bpf_cgroup_storage_assign(struct bpf_prog *prog, struct bpf_map *_map)
{
	enum bpf_cgroup_storage_type stype = cgroup_storage_type(_map);
	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	int ret = -EBUSY;

	spin_lock_bh(&map->lock);

	if (map->prog && map->prog != prog)
		goto unlock;
	if (prog->aux->cgroup_storage[stype] &&
	    prog->aux->cgroup_storage[stype] != _map)
		goto unlock;

	map->prog = prog;
	prog->aux->cgroup_storage[stype] = _map;
	ret = 0;
unlock:
	spin_unlock_bh(&map->lock);

	return ret;
}

void bpf_cgroup_storage_release(struct bpf_prog *prog, struct bpf_map *_map)
{
	enum bpf_cgroup_storage_type stype = cgroup_storage_type(_map);
	struct bpf_cgroup_storage_map *map = map_to_storage(_map);

	spin_lock_bh(&map->lock);
	if (map->prog == prog) {
		WARN_ON(prog->aux->cgroup_storage[stype] != _map);
		map->prog = NULL;
		prog->aux->cgroup_storage[stype] = NULL;
	}
	spin_unlock_bh(&map->lock);
}

static size_t bpf_cgroup_storage_calculate_size(struct bpf_map *map, u32 *pages)
{
	size_t size;

	if (cgroup_storage_type(map) == BPF_CGROUP_STORAGE_SHARED) {
		size = sizeof(struct bpf_storage_buffer) + map->value_size;
		*pages = round_up(sizeof(struct bpf_cgroup_storage) + size,
				  PAGE_SIZE) >> PAGE_SHIFT;
	} else {
		size = map->value_size;
		*pages = round_up(round_up(size, 8) * num_possible_cpus(),
				  PAGE_SIZE) >> PAGE_SHIFT;
	}

	return size;
}

struct bpf_cgroup_storage *bpf_cgroup_storage_alloc(struct bpf_prog *prog,
					enum bpf_cgroup_storage_type stype)
{
	struct bpf_cgroup_storage *storage;
	struct bpf_map *map;
	gfp_t flags;
	size_t size;
	u32 pages;

	map = prog->aux->cgroup_storage[stype];
	if (!map)
		return NULL;

	size = bpf_cgroup_storage_calculate_size(map, &pages);

	if (bpf_map_charge_memlock(map, pages))
		return ERR_PTR(-EPERM);

	storage = kmalloc_node(sizeof(struct bpf_cgroup_storage),
			       __GFP_ZERO | GFP_USER, map->numa_node);
	if (!storage)
		goto enomem;

	flags = __GFP_ZERO | GFP_USER;

	if (stype == BPF_CGROUP_STORAGE_SHARED) {
		storage->buf = kmalloc_node(size, flags, map->numa_node);
		if (!storage->buf)
			goto enomem;
		check_and_init_map_lock(map, storage->buf->data);
	} else {
		storage->percpu_buf = __alloc_percpu_gfp(size, 8, flags);
		if (!storage->percpu_buf)
			goto enomem;
	}

	storage->map = (struct bpf_cgroup_storage_map *)map;

	return storage;

enomem:
	bpf_map_uncharge_memlock(map, pages);
	kfree(storage);
	return ERR_PTR(-ENOMEM);
}

static void free_shared_cgroup_storage_rcu(struct rcu_head *rcu)
{
	struct bpf_cgroup_storage *storage =
		container_of(rcu, struct bpf_cgroup_storage, rcu);

	kfree(storage->buf);
	kfree(storage);
}

static void free_percpu_cgroup_storage_rcu(struct rcu_head *rcu)
{
	struct bpf_cgroup_storage *storage =
		container_of(rcu, struct bpf_cgroup_storage, rcu);

	free_percpu(storage->percpu_buf);
	kfree(storage);
}

void bpf_cgroup_storage_free(struct bpf_cgroup_storage *storage)
{
	enum bpf_cgroup_storage_type stype;
	struct bpf_map *map;
	u32 pages;

	if (!storage)
		return;

	map = &storage->map->map;

	bpf_cgroup_storage_calculate_size(map, &pages);
	bpf_map_uncharge_memlock(map, pages);

	stype = cgroup_storage_type(map);
	if (stype == BPF_CGROUP_STORAGE_SHARED)
		call_rcu(&storage->rcu, free_shared_cgroup_storage_rcu);
	else
		call_rcu(&storage->rcu, free_percpu_cgroup_storage_rcu);
}

void bpf_cgroup_storage_link(struct bpf_cgroup_storage *storage,
			     struct cgroup *cgroup,
			     enum bpf_attach_type type)
{
	struct bpf_cgroup_storage_map *map;

	if (!storage)
		return;

	storage->key.attach_type = type;
	storage->key.cgroup_inode_id = cgroup->kn->id.id;

	map = storage->map;

	spin_lock_bh(&map->lock);
	WARN_ON(cgroup_storage_insert(map, storage));
	list_add(&storage->list, &map->list);
	spin_unlock_bh(&map->lock);
}

void bpf_cgroup_storage_unlink(struct bpf_cgroup_storage *storage)
{
	struct bpf_cgroup_storage_map *map;
	struct rb_root *root;

	if (!storage)
		return;

	map = storage->map;

	spin_lock_bh(&map->lock);
	root = &map->root;
	rb_erase(&storage->node, root);

	list_del(&storage->list);
	spin_unlock_bh(&map->lock);
}

#endif
Commit	Line	Data
de9cbbaa RG	1	//SPDX-License-Identifier: GPL-2.0
	2	#include <linux/bpf-cgroup.h>
	3	#include <linux/bpf.h>
9a1126b6	4	#include <linux/btf.h>
de9cbbaa RG	5	#include <linux/bug.h>
	6	#include <linux/filter.h>
	7	#include <linux/mm.h>
	8	#include <linux/rbtree.h>
	9	#include <linux/slab.h>
9a1126b6	10	#include <uapi/linux/btf.h>
de9cbbaa	11
cf0dd411	12	DEFINE_PER_CPU(struct bpf_cgroup_storage*, bpf_cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]);
aa0ad5b0	13
de9cbbaa RG	14	#ifdef CONFIG_CGROUP_BPF
	15
	16	#define LOCAL_STORAGE_CREATE_FLAG_MASK \
591fe988	17	(BPF_F_NUMA_NODE \| BPF_F_ACCESS_MASK)
de9cbbaa RG	18
	19	struct bpf_cgroup_storage_map {
	20	struct bpf_map map;
	21
	22	spinlock_t lock;
	23	struct bpf_prog *prog;
	24	struct rb_root root;
	25	struct list_head list;
	26	};
	27
	28	static struct bpf_cgroup_storage_map map_to_storage(struct bpf_map map)
	29	{
	30	return container_of(map, struct bpf_cgroup_storage_map, map);
	31	}
	32
	33	static int bpf_cgroup_storage_key_cmp(
	34	const struct bpf_cgroup_storage_key *key1,
	35	const struct bpf_cgroup_storage_key *key2)
	36	{
	37	if (key1->cgroup_inode_id < key2->cgroup_inode_id)
	38	return -1;
	39	else if (key1->cgroup_inode_id > key2->cgroup_inode_id)
	40	return 1;
	41	else if (key1->attach_type < key2->attach_type)
	42	return -1;
	43	else if (key1->attach_type > key2->attach_type)
	44	return 1;
	45	return 0;
	46	}
	47
	48	static struct bpf_cgroup_storage *cgroup_storage_lookup(
	49	struct bpf_cgroup_storage_map map, struct bpf_cgroup_storage_key key,
	50	bool locked)
	51	{
	52	struct rb_root *root = &map->root;
	53	struct rb_node *node;
	54
	55	if (!locked)
	56	spin_lock_bh(&map->lock);
	57
	58	node = root->rb_node;
	59	while (node) {
	60	struct bpf_cgroup_storage *storage;
	61
	62	storage = container_of(node, struct bpf_cgroup_storage, node);
	63
	64	switch (bpf_cgroup_storage_key_cmp(key, &storage->key)) {
	65	case -1:
	66	node = node->rb_left;
	67	break;
	68	case 1:
	69	node = node->rb_right;
	70	break;
	71	default:
	72	if (!locked)
	73	spin_unlock_bh(&map->lock);
	74	return storage;
	75	}
	76	}
	77
	78	if (!locked)
	79	spin_unlock_bh(&map->lock);
	80
	81	return NULL;
82	}
83
84	static int cgroup_storage_insert(struct bpf_cgroup_storage_map *map,
85	struct bpf_cgroup_storage *storage)
86	{
87	struct rb_root *root = &map->root;
88	struct rb_node *new = &(root->rb_node), parent = NULL;
89
90	while (*new) {
91	struct bpf_cgroup_storage *this;
92
93	this = container_of(*new, struct bpf_cgroup_storage, node);
94
95	parent = *new;
96	switch (bpf_cgroup_storage_key_cmp(&storage->key, &this->key)) {
97	case -1:
98	new = &((*new)->rb_left);
99	break;
100	case 1:
101	new = &((*new)->rb_right);
102	break;
103	default:
104	return -EEXIST;
105	}
106	}
107
108	rb_link_node(&storage->node, parent, new);
109	rb_insert_color(&storage->node, root);
110
111	return 0;
112	}
113
114	static void cgroup_storage_lookup_elem(struct bpf_map _map, void *_key)
115	{
116	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
117	struct bpf_cgroup_storage_key *key = _key;
118	struct bpf_cgroup_storage *storage;
119
120	storage = cgroup_storage_lookup(map, key, false);
121	if (!storage)
122	return NULL;
123
124	return &READ_ONCE(storage->buf)->data[0];
125	}
126
127	static int cgroup_storage_update_elem(struct bpf_map map, void _key,
128	void *value, u64 flags)
129	{
130	struct bpf_cgroup_storage_key *key = _key;
131	struct bpf_cgroup_storage *storage;
132	struct bpf_storage_buffer *new;
133
96049f3a AS	134	if (unlikely(flags & ~(BPF_F_LOCK \| BPF_EXIST \| BPF_NOEXIST)))
	135	return -EINVAL;
	136
	137	if (unlikely(flags & BPF_NOEXIST))
	138	return -EINVAL;
	139
	140	if (unlikely((flags & BPF_F_LOCK) &&
	141	!map_value_has_spin_lock(map)))
de9cbbaa RG	142	return -EINVAL;
	143
	144	storage = cgroup_storage_lookup((struct bpf_cgroup_storage_map *)map,
	145	key, false);
	146	if (!storage)
	147	return -ENOENT;
	148
96049f3a AS	149	if (flags & BPF_F_LOCK) {
	150	copy_map_value_locked(map, storage->buf->data, value, false);
	151	return 0;
	152	}
	153
de9cbbaa	154	new = kmalloc_node(sizeof(struct bpf_storage_buffer) +
569a933b RG	155	map->value_size,
569a933b RG	156	__GFP_ZERO \| GFP_ATOMIC \| __GFP_NOWARN,
de9cbbaa RG	157	map->numa_node);
	158	if (!new)
	159	return -ENOMEM;
	160
	161	memcpy(&new->data[0], value, map->value_size);
e16d2f1a	162	check_and_init_map_lock(map, new->data);
de9cbbaa RG	163
	164	new = xchg(&storage->buf, new);
	165	kfree_rcu(new, rcu);
	166
	167	return 0;
	168	}
	169
b741f163 RG	170	int bpf_percpu_cgroup_storage_copy(struct bpf_map _map, void _key,
	171	void *value)
	172	{
	173	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	174	struct bpf_cgroup_storage_key *key = _key;
	175	struct bpf_cgroup_storage *storage;
	176	int cpu, off = 0;
	177	u32 size;
	178
	179	rcu_read_lock();
	180	storage = cgroup_storage_lookup(map, key, false);
	181	if (!storage) {
	182	rcu_read_unlock();
	183	return -ENOENT;
	184	}
	185
	186	/* per_cpu areas are zero-filled and bpf programs can only
	187	* access 'value_size' of them, so copying rounded areas
	188	* will not leak any kernel data
	189	*/
	190	size = round_up(_map->value_size, 8);
	191	for_each_possible_cpu(cpu) {
	192	bpf_long_memcpy(value + off,
	193	per_cpu_ptr(storage->percpu_buf, cpu), size);
	194	off += size;
	195	}
	196	rcu_read_unlock();
	197	return 0;
	198	}
	199
	200	int bpf_percpu_cgroup_storage_update(struct bpf_map _map, void _key,
	201	void *value, u64 map_flags)
	202	{
	203	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	204	struct bpf_cgroup_storage_key *key = _key;
	205	struct bpf_cgroup_storage *storage;
	206	int cpu, off = 0;
	207	u32 size;
	208
	209	if (map_flags != BPF_ANY && map_flags != BPF_EXIST)
	210	return -EINVAL;
	211
	212	rcu_read_lock();
	213	storage = cgroup_storage_lookup(map, key, false);
	214	if (!storage) {
	215	rcu_read_unlock();
	216	return -ENOENT;
	217	}
	218
	219	/* the user space will provide round_up(value_size, 8) bytes that
	220	* will be copied into per-cpu area. bpf programs can only access
	221	* value_size of it. During lookup the same extra bytes will be
	222	* returned or zeros which were zero-filled by percpu_alloc,
	223	* so no kernel data leaks possible
	224	*/
	225	size = round_up(_map->value_size, 8);
	226	for_each_possible_cpu(cpu) {
	227	bpf_long_memcpy(per_cpu_ptr(storage->percpu_buf, cpu),
	228	value + off, size);
	229	off += size;
	230	}
	231	rcu_read_unlock();
	232	return 0;
	233	}
234
de9cbbaa RG	235	static int cgroup_storage_get_next_key(struct bpf_map _map, void _key,
	236	void *_next_key)
	237	{
	238	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	239	struct bpf_cgroup_storage_key *key = _key;
	240	struct bpf_cgroup_storage_key *next = _next_key;
	241	struct bpf_cgroup_storage *storage;
	242
	243	spin_lock_bh(&map->lock);
	244
	245	if (list_empty(&map->list))
	246	goto enoent;
	247
	248	if (key) {
	249	storage = cgroup_storage_lookup(map, key, true);
	250	if (!storage)
	251	goto enoent;
	252
	253	storage = list_next_entry(storage, list);
	254	if (!storage)
	255	goto enoent;
	256	} else {
	257	storage = list_first_entry(&map->list,
	258	struct bpf_cgroup_storage, list);
	259	}
	260
	261	spin_unlock_bh(&map->lock);
	262	next->attach_type = storage->key.attach_type;
	263	next->cgroup_inode_id = storage->key.cgroup_inode_id;
	264	return 0;
	265
	266	enoent:
	267	spin_unlock_bh(&map->lock);
	268	return -ENOENT;
	269	}
	270
	271	static struct bpf_map cgroup_storage_map_alloc(union bpf_attr attr)
	272	{
	273	int numa_node = bpf_map_attr_numa_node(attr);
	274	struct bpf_cgroup_storage_map *map;
	275
	276	if (attr->key_size != sizeof(struct bpf_cgroup_storage_key))
	277	return ERR_PTR(-EINVAL);
	278
b0584ea6 RG	279	if (attr->value_size == 0)
	280	return ERR_PTR(-EINVAL);
	281
de9cbbaa RG	282	if (attr->value_size > PAGE_SIZE)
	283	return ERR_PTR(-E2BIG);
	284
591fe988 DB	285	if (attr->map_flags & ~LOCAL_STORAGE_CREATE_FLAG_MASK \|\|
591fe988 DB	286	!bpf_map_flags_access_ok(attr->map_flags))
de9cbbaa RG	287	return ERR_PTR(-EINVAL);
	288
	289	if (attr->max_entries)
	290	/* max_entries is not used and enforced to be 0 */
	291	return ERR_PTR(-EINVAL);
	292
	293	map = kmalloc_node(sizeof(struct bpf_cgroup_storage_map),
	294	__GFP_ZERO \| GFP_USER, numa_node);
	295	if (!map)
	296	return ERR_PTR(-ENOMEM);
	297
	298	map->map.pages = round_up(sizeof(struct bpf_cgroup_storage_map),
	299	PAGE_SIZE) >> PAGE_SHIFT;
	300
	301	/* copy mandatory map attributes */
	302	bpf_map_init_from_attr(&map->map, attr);
	303
	304	spin_lock_init(&map->lock);
	305	map->root = RB_ROOT;
	306	INIT_LIST_HEAD(&map->list);
	307
	308	return &map->map;
	309	}
	310
	311	static void cgroup_storage_map_free(struct bpf_map *_map)
	312	{
	313	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	314
	315	WARN_ON(!RB_EMPTY_ROOT(&map->root));
	316	WARN_ON(!list_empty(&map->list));
	317
	318	kfree(map);
	319	}
	320
	321	static int cgroup_storage_delete_elem(struct bpf_map map, void key)
	322	{
	323	return -EINVAL;
	324	}
	325
9a1126b6 RG	326	static int cgroup_storage_check_btf(const struct bpf_map *map,
	327	const struct btf *btf,
	328	const struct btf_type *key_type,
	329	const struct btf_type *value_type)
	330	{
9a1126b6	331	struct btf_member *m;
ffa0c1cf	332	u32 offset, size;
9a1126b6 RG	333
	334	/* Key is expected to be of struct bpf_cgroup_storage_key type,
	335	* which is:
	336	* struct bpf_cgroup_storage_key {
	337	* __u64 cgroup_inode_id;
	338	* __u32 attach_type;
	339	* };
	340	*/
	341
	342	/*
	343	* Key_type must be a structure with two fields.
	344	*/
	345	if (BTF_INFO_KIND(key_type->info) != BTF_KIND_STRUCT \|\|
	346	BTF_INFO_VLEN(key_type->info) != 2)
	347	return -EINVAL;
	348
	349	/*
	350	* The first field must be a 64 bit integer at 0 offset.
	351	*/
	352	m = (struct btf_member *)(key_type + 1);
9a1126b6	353	size = FIELD_SIZEOF(struct bpf_cgroup_storage_key, cgroup_inode_id);
ffa0c1cf	354	if (!btf_member_is_reg_int(btf, key_type, m, 0, size))
9a1126b6 RG	355	return -EINVAL;
	356
	357	/*
	358	* The second field must be a 32 bit integer at 64 bit offset.
	359	*/
	360	m++;
ffa0c1cf	361	offset = offsetof(struct bpf_cgroup_storage_key, attach_type);
9a1126b6	362	size = FIELD_SIZEOF(struct bpf_cgroup_storage_key, attach_type);
ffa0c1cf	363	if (!btf_member_is_reg_int(btf, key_type, m, offset, size))
9a1126b6 RG	364	return -EINVAL;
	365
	366	return 0;
	367	}
	368
	369	static void cgroup_storage_seq_show_elem(struct bpf_map map, void _key,
	370	struct seq_file *m)
	371	{
	372	enum bpf_cgroup_storage_type stype = cgroup_storage_type(map);
	373	struct bpf_cgroup_storage_key *key = _key;
	374	struct bpf_cgroup_storage *storage;
	375	int cpu;
	376
	377	rcu_read_lock();
	378	storage = cgroup_storage_lookup(map_to_storage(map), key, false);
	379	if (!storage) {
	380	rcu_read_unlock();
	381	return;
	382	}
	383
	384	btf_type_seq_show(map->btf, map->btf_key_type_id, key, m);
	385	stype = cgroup_storage_type(map);
	386	if (stype == BPF_CGROUP_STORAGE_SHARED) {
	387	seq_puts(m, ": ");
	388	btf_type_seq_show(map->btf, map->btf_value_type_id,
	389	&READ_ONCE(storage->buf)->data[0], m);
	390	seq_puts(m, "\n");
	391	} else {
	392	seq_puts(m, ": {\n");
	393	for_each_possible_cpu(cpu) {
	394	seq_printf(m, "\tcpu%d: ", cpu);
	395	btf_type_seq_show(map->btf, map->btf_value_type_id,
	396	per_cpu_ptr(storage->percpu_buf, cpu),
	397	m);
	398	seq_puts(m, "\n");
	399	}
	400	seq_puts(m, "}\n");
	401	}
	402	rcu_read_unlock();
	403	}
	404
de9cbbaa RG	405	const struct bpf_map_ops cgroup_storage_map_ops = {
	406	.map_alloc = cgroup_storage_map_alloc,
	407	.map_free = cgroup_storage_map_free,
	408	.map_get_next_key = cgroup_storage_get_next_key,
	409	.map_lookup_elem = cgroup_storage_lookup_elem,
	410	.map_update_elem = cgroup_storage_update_elem,
	411	.map_delete_elem = cgroup_storage_delete_elem,
9a1126b6 RG	412	.map_check_btf = cgroup_storage_check_btf,
9a1126b6 RG	413	.map_seq_show_elem = cgroup_storage_seq_show_elem,
de9cbbaa RG	414	};
	415
	416	int bpf_cgroup_storage_assign(struct bpf_prog prog, struct bpf_map _map)
	417	{
8bad74f9	418	enum bpf_cgroup_storage_type stype = cgroup_storage_type(_map);
de9cbbaa RG	419	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	420	int ret = -EBUSY;
	421
	422	spin_lock_bh(&map->lock);
	423
	424	if (map->prog && map->prog != prog)
	425	goto unlock;
8bad74f9 RG	426	if (prog->aux->cgroup_storage[stype] &&
8bad74f9 RG	427	prog->aux->cgroup_storage[stype] != _map)
de9cbbaa RG	428	goto unlock;
	429
	430	map->prog = prog;
8bad74f9	431	prog->aux->cgroup_storage[stype] = _map;
de9cbbaa RG	432	ret = 0;
	433	unlock:
	434	spin_unlock_bh(&map->lock);
	435
	436	return ret;
	437	}
	438
	439	void bpf_cgroup_storage_release(struct bpf_prog prog, struct bpf_map _map)
	440	{
8bad74f9	441	enum bpf_cgroup_storage_type stype = cgroup_storage_type(_map);
de9cbbaa RG	442	struct bpf_cgroup_storage_map *map = map_to_storage(_map);
	443
	444	spin_lock_bh(&map->lock);
	445	if (map->prog == prog) {
8bad74f9	446	WARN_ON(prog->aux->cgroup_storage[stype] != _map);
de9cbbaa	447	map->prog = NULL;
8bad74f9	448	prog->aux->cgroup_storage[stype] = NULL;
de9cbbaa RG	449	}
	450	spin_unlock_bh(&map->lock);
	451	}
	452
b741f163 RG	453	static size_t bpf_cgroup_storage_calculate_size(struct bpf_map map, u32 pages)
	454	{
	455	size_t size;
	456
	457	if (cgroup_storage_type(map) == BPF_CGROUP_STORAGE_SHARED) {
	458	size = sizeof(struct bpf_storage_buffer) + map->value_size;
	459	*pages = round_up(sizeof(struct bpf_cgroup_storage) + size,
	460	PAGE_SIZE) >> PAGE_SHIFT;
	461	} else {
	462	size = map->value_size;
	463	pages = round_up(round_up(size, 8) num_possible_cpus(),
	464	PAGE_SIZE) >> PAGE_SHIFT;
	465	}
	466
	467	return size;
	468	}
	469
8bad74f9 RG	470	struct bpf_cgroup_storage bpf_cgroup_storage_alloc(struct bpf_prog prog,
8bad74f9 RG	471	enum bpf_cgroup_storage_type stype)
de9cbbaa RG	472	{
	473	struct bpf_cgroup_storage *storage;
	474	struct bpf_map *map;
b741f163 RG	475	gfp_t flags;
b741f163 RG	476	size_t size;
de9cbbaa RG	477	u32 pages;
de9cbbaa RG	478
8bad74f9	479	map = prog->aux->cgroup_storage[stype];
de9cbbaa RG	480	if (!map)
	481	return NULL;
	482
b741f163 RG	483	size = bpf_cgroup_storage_calculate_size(map, &pages);
b741f163 RG	484
de9cbbaa RG	485	if (bpf_map_charge_memlock(map, pages))
	486	return ERR_PTR(-EPERM);
	487
	488	storage = kmalloc_node(sizeof(struct bpf_cgroup_storage),
	489	__GFP_ZERO \| GFP_USER, map->numa_node);
b741f163 RG	490	if (!storage)
b741f163 RG	491	goto enomem;
de9cbbaa	492
b741f163 RG	493	flags = __GFP_ZERO \| GFP_USER;
	494
	495	if (stype == BPF_CGROUP_STORAGE_SHARED) {
	496	storage->buf = kmalloc_node(size, flags, map->numa_node);
	497	if (!storage->buf)
	498	goto enomem;
e16d2f1a	499	check_and_init_map_lock(map, storage->buf->data);
b741f163 RG	500	} else {
	501	storage->percpu_buf = __alloc_percpu_gfp(size, 8, flags);
	502	if (!storage->percpu_buf)
	503	goto enomem;
de9cbbaa RG	504	}
	505
	506	storage->map = (struct bpf_cgroup_storage_map *)map;
	507
	508	return storage;
b741f163 RG	509
	510	enomem:
	511	bpf_map_uncharge_memlock(map, pages);
	512	kfree(storage);
	513	return ERR_PTR(-ENOMEM);
	514	}
	515
	516	static void free_shared_cgroup_storage_rcu(struct rcu_head *rcu)
	517	{
	518	struct bpf_cgroup_storage *storage =
	519	container_of(rcu, struct bpf_cgroup_storage, rcu);
	520
	521	kfree(storage->buf);
	522	kfree(storage);
	523	}
	524
	525	static void free_percpu_cgroup_storage_rcu(struct rcu_head *rcu)
	526	{
	527	struct bpf_cgroup_storage *storage =
	528	container_of(rcu, struct bpf_cgroup_storage, rcu);
	529
	530	free_percpu(storage->percpu_buf);
	531	kfree(storage);
de9cbbaa RG	532	}
	533
	534	void bpf_cgroup_storage_free(struct bpf_cgroup_storage *storage)
	535	{
b741f163	536	enum bpf_cgroup_storage_type stype;
de9cbbaa	537	struct bpf_map *map;
b741f163	538	u32 pages;
de9cbbaa RG	539
	540	if (!storage)
	541	return;
	542
	543	map = &storage->map->map;
b741f163 RG	544
b741f163 RG	545	bpf_cgroup_storage_calculate_size(map, &pages);
de9cbbaa RG	546	bpf_map_uncharge_memlock(map, pages);
de9cbbaa RG	547
b741f163 RG	548	stype = cgroup_storage_type(map);
	549	if (stype == BPF_CGROUP_STORAGE_SHARED)
	550	call_rcu(&storage->rcu, free_shared_cgroup_storage_rcu);
	551	else
	552	call_rcu(&storage->rcu, free_percpu_cgroup_storage_rcu);
de9cbbaa RG	553	}
	554
	555	void bpf_cgroup_storage_link(struct bpf_cgroup_storage *storage,
	556	struct cgroup *cgroup,
	557	enum bpf_attach_type type)
	558	{
	559	struct bpf_cgroup_storage_map *map;
	560
	561	if (!storage)
	562	return;
	563
	564	storage->key.attach_type = type;
	565	storage->key.cgroup_inode_id = cgroup->kn->id.id;
	566
	567	map = storage->map;
	568
	569	spin_lock_bh(&map->lock);
	570	WARN_ON(cgroup_storage_insert(map, storage));
	571	list_add(&storage->list, &map->list);
	572	spin_unlock_bh(&map->lock);
	573	}
	574
	575	void bpf_cgroup_storage_unlink(struct bpf_cgroup_storage *storage)
	576	{
	577	struct bpf_cgroup_storage_map *map;
	578	struct rb_root *root;
	579
	580	if (!storage)
	581	return;
	582
	583	map = storage->map;
	584
	585	spin_lock_bh(&map->lock);
	586	root = &map->root;
	587	rb_erase(&storage->node, root);
	588
	589	list_del(&storage->list);
	590	spin_unlock_bh(&map->lock);
	591	}
	592
	593	#endif