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

// SPDX-License-Identifier: GPL-2.0
/* XSKMAP used for AF_XDP sockets
 * Copyright(c) 2018 Intel Corporation.
 */

#include <linux/bpf.h>
#include <linux/capability.h>
#include <net/xdp_sock.h>
#include <linux/slab.h>
#include <linux/sched.h>

int xsk_map_inc(struct xsk_map *map)
{
	bpf_map_inc(&map->map);
	return 0;
}

void xsk_map_put(struct xsk_map *map)
{
	bpf_map_put(&map->map);
}

static struct xsk_map_node *xsk_map_node_alloc(struct xsk_map *map,
					       struct xdp_sock **map_entry)
{
	struct xsk_map_node *node;
	int err;

	node = kzalloc(sizeof(*node), GFP_ATOMIC | __GFP_NOWARN);
	if (!node)
		return ERR_PTR(-ENOMEM);

	err = xsk_map_inc(map);
	if (err) {
		kfree(node);
		return ERR_PTR(err);
	}

	node->map = map;
	node->map_entry = map_entry;
	return node;
}

static void xsk_map_node_free(struct xsk_map_node *node)
{
	xsk_map_put(node->map);
	kfree(node);
}

static void xsk_map_sock_add(struct xdp_sock *xs, struct xsk_map_node *node)
{
	spin_lock_bh(&xs->map_list_lock);
	list_add_tail(&node->node, &xs->map_list);
	spin_unlock_bh(&xs->map_list_lock);
}

static void xsk_map_sock_delete(struct xdp_sock *xs,
				struct xdp_sock **map_entry)
{
	struct xsk_map_node *n, *tmp;

	spin_lock_bh(&xs->map_list_lock);
	list_for_each_entry_safe(n, tmp, &xs->map_list, node) {
		if (map_entry == n->map_entry) {
			list_del(&n->node);
			xsk_map_node_free(n);
		}
	}
	spin_unlock_bh(&xs->map_list_lock);
}

static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
{
	struct bpf_map_memory mem;
	int cpu, err, numa_node;
	struct xsk_map *m;
	u64 cost, size;

	if (!capable(CAP_NET_ADMIN))
		return ERR_PTR(-EPERM);

	if (attr->max_entries == 0 || attr->key_size != 4 ||
	    attr->value_size != 4 ||
	    attr->map_flags & ~(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY))
		return ERR_PTR(-EINVAL);

	numa_node = bpf_map_attr_numa_node(attr);
	size = struct_size(m, xsk_map, attr->max_entries);
	cost = size + array_size(sizeof(*m->flush_list), num_possible_cpus());

	err = bpf_map_charge_init(&mem, cost);
	if (err < 0)
		return ERR_PTR(err);

	m = bpf_map_area_alloc(size, numa_node);
	if (!m) {
		bpf_map_charge_finish(&mem);
		return ERR_PTR(-ENOMEM);
	}

	bpf_map_init_from_attr(&m->map, attr);
	bpf_map_charge_move(&m->map.memory, &mem);
	spin_lock_init(&m->lock);

	m->flush_list = alloc_percpu(struct list_head);
	if (!m->flush_list) {
		bpf_map_charge_finish(&m->map.memory);
		bpf_map_area_free(m);
		return ERR_PTR(-ENOMEM);
	}

	for_each_possible_cpu(cpu)
		INIT_LIST_HEAD(per_cpu_ptr(m->flush_list, cpu));

	return &m->map;
}

static void xsk_map_free(struct bpf_map *map)
{
	struct xsk_map *m = container_of(map, struct xsk_map, map);

	bpf_clear_redirect_map(map);
	synchronize_net();
	free_percpu(m->flush_list);
	bpf_map_area_free(m);
}

static int xsk_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
{
	struct xsk_map *m = container_of(map, struct xsk_map, map);
	u32 index = key ? *(u32 *)key : U32_MAX;
	u32 *next = next_key;

	if (index >= m->map.max_entries) {
		*next = 0;
		return 0;
	}

	if (index == m->map.max_entries - 1)
		return -ENOENT;
	*next = index + 1;
	return 0;
}

static u32 xsk_map_gen_lookup(struct bpf_map *map, struct bpf_insn *insn_buf)
{
	const int ret = BPF_REG_0, mp = BPF_REG_1, index = BPF_REG_2;
	struct bpf_insn *insn = insn_buf;

	*insn++ = BPF_LDX_MEM(BPF_W, ret, index, 0);
	*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 5);
	*insn++ = BPF_ALU64_IMM(BPF_LSH, ret, ilog2(sizeof(struct xsk_sock *)));
	*insn++ = BPF_ALU64_IMM(BPF_ADD, mp, offsetof(struct xsk_map, xsk_map));
	*insn++ = BPF_ALU64_REG(BPF_ADD, ret, mp);
	*insn++ = BPF_LDX_MEM(BPF_SIZEOF(struct xsk_sock *), ret, ret, 0);
	*insn++ = BPF_JMP_IMM(BPF_JA, 0, 0, 1);
	*insn++ = BPF_MOV64_IMM(ret, 0);
	return insn - insn_buf;
}

static void *xsk_map_lookup_elem(struct bpf_map *map, void *key)
{
	WARN_ON_ONCE(!rcu_read_lock_held());
	return __xsk_map_lookup_elem(map, *(u32 *)key);
}

static void *xsk_map_lookup_elem_sys_only(struct bpf_map *map, void *key)
{
	return ERR_PTR(-EOPNOTSUPP);
}

static int xsk_map_update_elem(struct bpf_map *map, void *key, void *value,
			       u64 map_flags)
{
	struct xsk_map *m = container_of(map, struct xsk_map, map);
	struct xdp_sock *xs, *old_xs, **map_entry;
	u32 i = *(u32 *)key, fd = *(u32 *)value;
	struct xsk_map_node *node;
	struct socket *sock;
	int err;

	if (unlikely(map_flags > BPF_EXIST))
		return -EINVAL;
	if (unlikely(i >= m->map.max_entries))
		return -E2BIG;

	sock = sockfd_lookup(fd, &err);
	if (!sock)
		return err;

	if (sock->sk->sk_family != PF_XDP) {
		sockfd_put(sock);
		return -EOPNOTSUPP;
	}

	xs = (struct xdp_sock *)sock->sk;

	if (!xsk_is_setup_for_bpf_map(xs)) {
		sockfd_put(sock);
		return -EOPNOTSUPP;
	}

	map_entry = &m->xsk_map[i];
	node = xsk_map_node_alloc(m, map_entry);
	if (IS_ERR(node)) {
		sockfd_put(sock);
		return PTR_ERR(node);
	}

	spin_lock_bh(&m->lock);
	old_xs = READ_ONCE(*map_entry);
	if (old_xs == xs) {
		err = 0;
		goto out;
	} else if (old_xs && map_flags == BPF_NOEXIST) {
		err = -EEXIST;
		goto out;
	} else if (!old_xs && map_flags == BPF_EXIST) {
		err = -ENOENT;
		goto out;
	}
	xsk_map_sock_add(xs, node);
	WRITE_ONCE(*map_entry, xs);
	if (old_xs)
		xsk_map_sock_delete(old_xs, map_entry);
	spin_unlock_bh(&m->lock);
	sockfd_put(sock);
	return 0;

out:
	spin_unlock_bh(&m->lock);
	sockfd_put(sock);
	xsk_map_node_free(node);
	return err;
}

static int xsk_map_delete_elem(struct bpf_map *map, void *key)
{
	struct xsk_map *m = container_of(map, struct xsk_map, map);
	struct xdp_sock *old_xs, **map_entry;
	int k = *(u32 *)key;

	if (k >= map->max_entries)
		return -EINVAL;

	spin_lock_bh(&m->lock);
	map_entry = &m->xsk_map[k];
	old_xs = xchg(map_entry, NULL);
	if (old_xs)
		xsk_map_sock_delete(old_xs, map_entry);
	spin_unlock_bh(&m->lock);

	return 0;
}

void xsk_map_try_sock_delete(struct xsk_map *map, struct xdp_sock *xs,
			     struct xdp_sock **map_entry)
{
	spin_lock_bh(&map->lock);
	if (READ_ONCE(*map_entry) == xs) {
		WRITE_ONCE(*map_entry, NULL);
		xsk_map_sock_delete(xs, map_entry);
	}
	spin_unlock_bh(&map->lock);
}

const struct bpf_map_ops xsk_map_ops = {
	.map_alloc = xsk_map_alloc,
	.map_free = xsk_map_free,
	.map_get_next_key = xsk_map_get_next_key,
	.map_lookup_elem = xsk_map_lookup_elem,
	.map_gen_lookup = xsk_map_gen_lookup,
	.map_lookup_elem_sys_only = xsk_map_lookup_elem_sys_only,
	.map_update_elem = xsk_map_update_elem,
	.map_delete_elem = xsk_map_delete_elem,
	.map_check_btf = map_check_no_btf,
};
Commit	Line	Data
fbfc504a BT	1	// SPDX-License-Identifier: GPL-2.0
	2	/* XSKMAP used for AF_XDP sockets
	3	* Copyright(c) 2018 Intel Corporation.
fbfc504a BT	4	*/
	5
	6	#include <linux/bpf.h>
	7	#include <linux/capability.h>
	8	#include <net/xdp_sock.h>
	9	#include <linux/slab.h>
	10	#include <linux/sched.h>
	11
0402acd6 BT	12	int xsk_map_inc(struct xsk_map *map)
0402acd6 BT	13	{
1e0bd5a0 AN	14	bpf_map_inc(&map->map);
1e0bd5a0 AN	15	return 0;
0402acd6 BT	16	}
	17
	18	void xsk_map_put(struct xsk_map *map)
	19	{
	20	bpf_map_put(&map->map);
	21	}
	22
	23	static struct xsk_map_node xsk_map_node_alloc(struct xsk_map map,
	24	struct xdp_sock **map_entry)
	25	{
	26	struct xsk_map_node *node;
	27	int err;
	28
	29	node = kzalloc(sizeof(*node), GFP_ATOMIC \| __GFP_NOWARN);
	30	if (!node)
fcd30ae0	31	return ERR_PTR(-ENOMEM);
0402acd6 BT	32
	33	err = xsk_map_inc(map);
	34	if (err) {
	35	kfree(node);
	36	return ERR_PTR(err);
	37	}
	38
	39	node->map = map;
	40	node->map_entry = map_entry;
	41	return node;
	42	}
	43
	44	static void xsk_map_node_free(struct xsk_map_node *node)
	45	{
	46	xsk_map_put(node->map);
	47	kfree(node);
	48	}
	49
	50	static void xsk_map_sock_add(struct xdp_sock xs, struct xsk_map_node node)
	51	{
	52	spin_lock_bh(&xs->map_list_lock);
	53	list_add_tail(&node->node, &xs->map_list);
	54	spin_unlock_bh(&xs->map_list_lock);
	55	}
	56
	57	static void xsk_map_sock_delete(struct xdp_sock *xs,
	58	struct xdp_sock **map_entry)
	59	{
	60	struct xsk_map_node n, tmp;
	61
	62	spin_lock_bh(&xs->map_list_lock);
	63	list_for_each_entry_safe(n, tmp, &xs->map_list, node) {
	64	if (map_entry == n->map_entry) {
	65	list_del(&n->node);
	66	xsk_map_node_free(n);
	67	}
	68	}
	69	spin_unlock_bh(&xs->map_list_lock);
	70	}
	71
fbfc504a BT	72	static struct bpf_map xsk_map_alloc(union bpf_attr attr)
fbfc504a BT	73	{
64fe8c06 BT	74	struct bpf_map_memory mem;
64fe8c06 BT	75	int cpu, err, numa_node;
fbfc504a	76	struct xsk_map *m;
64fe8c06	77	u64 cost, size;
fbfc504a BT	78
	79	if (!capable(CAP_NET_ADMIN))
	80	return ERR_PTR(-EPERM);
	81
	82	if (attr->max_entries == 0 \|\| attr->key_size != 4 \|\|
	83	attr->value_size != 4 \|\|
	84	attr->map_flags & ~(BPF_F_NUMA_NODE \| BPF_F_RDONLY \| BPF_F_WRONLY))
	85	return ERR_PTR(-EINVAL);
	86
64fe8c06 BT	87	numa_node = bpf_map_attr_numa_node(attr);
	88	size = struct_size(m, xsk_map, attr->max_entries);
	89	cost = size + array_size(sizeof(*m->flush_list), num_possible_cpus());
	90
	91	err = bpf_map_charge_init(&mem, cost);
	92	if (err < 0)
	93	return ERR_PTR(err);
	94
	95	m = bpf_map_area_alloc(size, numa_node);
	96	if (!m) {
	97	bpf_map_charge_finish(&mem);
fbfc504a	98	return ERR_PTR(-ENOMEM);
64fe8c06	99	}
fbfc504a BT	100
fbfc504a BT	101	bpf_map_init_from_attr(&m->map, attr);
64fe8c06	102	bpf_map_charge_move(&m->map.memory, &mem);
0402acd6	103	spin_lock_init(&m->lock);
fbfc504a	104
fbfc504a	105	m->flush_list = alloc_percpu(struct list_head);
64fe8c06 BT	106	if (!m->flush_list) {
	107	bpf_map_charge_finish(&m->map.memory);
	108	bpf_map_area_free(m);
	109	return ERR_PTR(-ENOMEM);
	110	}
fbfc504a BT	111
	112	for_each_possible_cpu(cpu)
	113	INIT_LIST_HEAD(per_cpu_ptr(m->flush_list, cpu));
	114
fbfc504a	115	return &m->map;
fbfc504a BT	116	}
	117
	118	static void xsk_map_free(struct bpf_map *map)
	119	{
	120	struct xsk_map *m = container_of(map, struct xsk_map, map);
fbfc504a	121
f6069b9a	122	bpf_clear_redirect_map(map);
fbfc504a	123	synchronize_net();
fbfc504a	124	free_percpu(m->flush_list);
64fe8c06	125	bpf_map_area_free(m);
fbfc504a BT	126	}
	127
	128	static int xsk_map_get_next_key(struct bpf_map map, void key, void *next_key)
	129	{
	130	struct xsk_map *m = container_of(map, struct xsk_map, map);
	131	u32 index = key ? (u32 )key : U32_MAX;
	132	u32 *next = next_key;
	133
	134	if (index >= m->map.max_entries) {
	135	*next = 0;
	136	return 0;
	137	}
	138
	139	if (index == m->map.max_entries - 1)
	140	return -ENOENT;
	141	*next = index + 1;
	142	return 0;
	143	}
	144
e65650f2 MF	145	static u32 xsk_map_gen_lookup(struct bpf_map map, struct bpf_insn insn_buf)
	146	{
	147	const int ret = BPF_REG_0, mp = BPF_REG_1, index = BPF_REG_2;
	148	struct bpf_insn *insn = insn_buf;
	149
	150	*insn++ = BPF_LDX_MEM(BPF_W, ret, index, 0);
	151	*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 5);
	152	insn++ = BPF_ALU64_IMM(BPF_LSH, ret, ilog2(sizeof(struct xsk_sock )));
	153	*insn++ = BPF_ALU64_IMM(BPF_ADD, mp, offsetof(struct xsk_map, xsk_map));
	154	*insn++ = BPF_ALU64_REG(BPF_ADD, ret, mp);
	155	insn++ = BPF_LDX_MEM(BPF_SIZEOF(struct xsk_sock ), ret, ret, 0);
	156	*insn++ = BPF_JMP_IMM(BPF_JA, 0, 0, 1);
	157	*insn++ = BPF_MOV64_IMM(ret, 0);
	158	return insn - insn_buf;
	159	}
	160
fbfc504a	161	static void xsk_map_lookup_elem(struct bpf_map map, void *key)
fada7fdc JL	162	{
	163	WARN_ON_ONCE(!rcu_read_lock_held());
	164	return __xsk_map_lookup_elem(map, (u32 )key);
	165	}
	166
	167	static void xsk_map_lookup_elem_sys_only(struct bpf_map map, void *key)
fbfc504a	168	{
3b4a63f6	169	return ERR_PTR(-EOPNOTSUPP);
fbfc504a BT	170	}
	171
	172	static int xsk_map_update_elem(struct bpf_map map, void key, void *value,
	173	u64 map_flags)
	174	{
	175	struct xsk_map *m = container_of(map, struct xsk_map, map);
0402acd6	176	struct xdp_sock xs, old_xs, **map_entry;
fbfc504a	177	u32 i = (u32 )key, fd = (u32 )value;
0402acd6	178	struct xsk_map_node *node;
fbfc504a BT	179	struct socket *sock;
	180	int err;
	181
	182	if (unlikely(map_flags > BPF_EXIST))
	183	return -EINVAL;
	184	if (unlikely(i >= m->map.max_entries))
	185	return -E2BIG;
fbfc504a BT	186
	187	sock = sockfd_lookup(fd, &err);
	188	if (!sock)
	189	return err;
	190
	191	if (sock->sk->sk_family != PF_XDP) {
	192	sockfd_put(sock);
	193	return -EOPNOTSUPP;
	194	}
	195
	196	xs = (struct xdp_sock *)sock->sk;
	197
	198	if (!xsk_is_setup_for_bpf_map(xs)) {
	199	sockfd_put(sock);
	200	return -EOPNOTSUPP;
	201	}
	202
0402acd6 BT	203	map_entry = &m->xsk_map[i];
	204	node = xsk_map_node_alloc(m, map_entry);
	205	if (IS_ERR(node)) {
	206	sockfd_put(sock);
	207	return PTR_ERR(node);
	208	}
fbfc504a	209
0402acd6 BT	210	spin_lock_bh(&m->lock);
	211	old_xs = READ_ONCE(*map_entry);
	212	if (old_xs == xs) {
	213	err = 0;
	214	goto out;
36cc3435 BT	215	} else if (old_xs && map_flags == BPF_NOEXIST) {
	216	err = -EEXIST;
	217	goto out;
	218	} else if (!old_xs && map_flags == BPF_EXIST) {
	219	err = -ENOENT;
	220	goto out;
0402acd6 BT	221	}
	222	xsk_map_sock_add(xs, node);
	223	WRITE_ONCE(*map_entry, xs);
cee27167	224	if (old_xs)
0402acd6 BT	225	xsk_map_sock_delete(old_xs, map_entry);
0402acd6 BT	226	spin_unlock_bh(&m->lock);
fbfc504a BT	227	sockfd_put(sock);
fbfc504a BT	228	return 0;
0402acd6 BT	229
	230	out:
	231	spin_unlock_bh(&m->lock);
	232	sockfd_put(sock);
	233	xsk_map_node_free(node);
	234	return err;
fbfc504a BT	235	}
	236
	237	static int xsk_map_delete_elem(struct bpf_map map, void key)
	238	{
	239	struct xsk_map *m = container_of(map, struct xsk_map, map);
0402acd6	240	struct xdp_sock old_xs, *map_entry;
fbfc504a BT	241	int k = (u32 )key;
	242
	243	if (k >= map->max_entries)
	244	return -EINVAL;
	245
0402acd6 BT	246	spin_lock_bh(&m->lock);
	247	map_entry = &m->xsk_map[k];
	248	old_xs = xchg(map_entry, NULL);
cee27167	249	if (old_xs)
0402acd6 BT	250	xsk_map_sock_delete(old_xs, map_entry);
0402acd6 BT	251	spin_unlock_bh(&m->lock);
fbfc504a BT	252
	253	return 0;
	254	}
	255
0402acd6 BT	256	void xsk_map_try_sock_delete(struct xsk_map map, struct xdp_sock xs,
	257	struct xdp_sock **map_entry)
	258	{
	259	spin_lock_bh(&map->lock);
	260	if (READ_ONCE(*map_entry) == xs) {
	261	WRITE_ONCE(*map_entry, NULL);
	262	xsk_map_sock_delete(xs, map_entry);
	263	}
	264	spin_unlock_bh(&map->lock);
	265	}
	266
fbfc504a BT	267	const struct bpf_map_ops xsk_map_ops = {
	268	.map_alloc = xsk_map_alloc,
	269	.map_free = xsk_map_free,
	270	.map_get_next_key = xsk_map_get_next_key,
	271	.map_lookup_elem = xsk_map_lookup_elem,
e65650f2	272	.map_gen_lookup = xsk_map_gen_lookup,
fada7fdc	273	.map_lookup_elem_sys_only = xsk_map_lookup_elem_sys_only,
fbfc504a BT	274	.map_update_elem = xsk_map_update_elem,
fbfc504a BT	275	.map_delete_elem = xsk_map_delete_elem,
e8d2bec0	276	.map_check_btf = map_check_no_btf,
fbfc504a	277	};