[linux-block.git] / net / xdp / 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/filter.h>
#include <net/xdp_sock.h>
#include <linux/slab.h>
#include <linux/sched.h>
#include <linux/btf_ids.h>

#include "xsk.h"

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

	node = bpf_map_kzalloc(&map->map, sizeof(*node),
			       GFP_ATOMIC | __GFP_NOWARN);
	if (!node)
		return ERR_PTR(-ENOMEM);

	bpf_map_inc(&map->map);
	atomic_inc(&map->count);

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

static void xsk_map_node_free(struct xsk_map_node *node)
{
	struct xsk_map *map = node->map;

	bpf_map_put(&node->map->map);
	kfree(node);
	atomic_dec(&map->count);
}

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 __rcu **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 xsk_map *m;
	int numa_node;
	u64 size;

	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);

	m = bpf_map_area_alloc(size, numa_node);
	if (!m)
		return ERR_PTR(-ENOMEM);

	bpf_map_init_from_attr(&m->map, attr);
	spin_lock_init(&m->lock);

	return &m->map;
}

static u64 xsk_map_mem_usage(const struct bpf_map *map)
{
	struct xsk_map *m = container_of(map, struct xsk_map, map);

	return struct_size(m, xsk_map, map->max_entries) +
		   (u64)atomic_read(&m->count) * sizeof(struct xsk_map_node);
}

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

	synchronize_net();
	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 int 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;
}

/* Elements are kept alive by RCU; either by rcu_read_lock() (from syscall) or
 * by local_bh_disable() (from XDP calls inside NAPI). The
 * rcu_read_lock_bh_held() below makes lockdep accept both.
 */
static void *__xsk_map_lookup_elem(struct bpf_map *map, u32 key)
{
	struct xsk_map *m = container_of(map, struct xsk_map, map);

	if (key >= map->max_entries)
		return NULL;

	return rcu_dereference_check(m->xsk_map[key], rcu_read_lock_bh_held());
}

static void *xsk_map_lookup_elem(struct bpf_map *map, void *key)
{
	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 long 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 __rcu **map_entry;
	struct xdp_sock *xs, *old_xs;
	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;

	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 = rcu_dereference_protected(*map_entry, lockdep_is_held(&m->lock));
	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);
	rcu_assign_pointer(*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 long xsk_map_delete_elem(struct bpf_map *map, void *key)
{
	struct xsk_map *m = container_of(map, struct xsk_map, map);
	struct xdp_sock __rcu **map_entry;
	struct xdp_sock *old_xs;
	int k = *(u32 *)key;

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

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

	return 0;
}

static long xsk_map_redirect(struct bpf_map *map, u64 index, u64 flags)
{
	return __bpf_xdp_redirect_map(map, index, flags, 0,
				      __xsk_map_lookup_elem);
}

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

static bool xsk_map_meta_equal(const struct bpf_map *meta0,
			       const struct bpf_map *meta1)
{
	return meta0->max_entries == meta1->max_entries &&
		bpf_map_meta_equal(meta0, meta1);
}

BTF_ID_LIST_SINGLE(xsk_map_btf_ids, struct, xsk_map)
const struct bpf_map_ops xsk_map_ops = {
	.map_meta_equal = xsk_map_meta_equal,
	.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,
	.map_mem_usage = xsk_map_mem_usage,
	.map_btf_id = &xsk_map_btf_ids[0],
	.map_redirect = xsk_map_redirect,
};
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>
b6459415	7	#include <linux/filter.h>
fbfc504a BT	8	#include <net/xdp_sock.h>
	9	#include <linux/slab.h>
	10	#include <linux/sched.h>
c317ab71	11	#include <linux/btf_ids.h>
fbfc504a	12
d20a1676 BT	13	#include "xsk.h"
d20a1676 BT	14
0402acd6	15	static struct xsk_map_node xsk_map_node_alloc(struct xsk_map map,
782347b6	16	struct xdp_sock __rcu **map_entry)
0402acd6 BT	17	{
0402acd6 BT	18	struct xsk_map_node *node;
0402acd6	19
28e1dcde RG	20	node = bpf_map_kzalloc(&map->map, sizeof(*node),
28e1dcde RG	21	GFP_ATOMIC \| __GFP_NOWARN);
0402acd6	22	if (!node)
fcd30ae0	23	return ERR_PTR(-ENOMEM);
0402acd6	24
bb1b25ca	25	bpf_map_inc(&map->map);
b4fd0d67	26	atomic_inc(&map->count);
0402acd6 BT	27
	28	node->map = map;
	29	node->map_entry = map_entry;
	30	return node;
	31	}
	32
	33	static void xsk_map_node_free(struct xsk_map_node *node)
	34	{
b4fd0d67 YS	35	struct xsk_map *map = node->map;
b4fd0d67 YS	36
bb1b25ca	37	bpf_map_put(&node->map->map);
0402acd6	38	kfree(node);
b4fd0d67	39	atomic_dec(&map->count);
0402acd6 BT	40	}
	41
	42	static void xsk_map_sock_add(struct xdp_sock xs, struct xsk_map_node node)
	43	{
	44	spin_lock_bh(&xs->map_list_lock);
	45	list_add_tail(&node->node, &xs->map_list);
	46	spin_unlock_bh(&xs->map_list_lock);
	47	}
	48
	49	static void xsk_map_sock_delete(struct xdp_sock *xs,
782347b6	50	struct xdp_sock __rcu **map_entry)
0402acd6 BT	51	{
	52	struct xsk_map_node n, tmp;
	53
	54	spin_lock_bh(&xs->map_list_lock);
	55	list_for_each_entry_safe(n, tmp, &xs->map_list, node) {
	56	if (map_entry == n->map_entry) {
	57	list_del(&n->node);
	58	xsk_map_node_free(n);
	59	}
	60	}
	61	spin_unlock_bh(&xs->map_list_lock);
	62	}
	63
fbfc504a BT	64	static struct bpf_map xsk_map_alloc(union bpf_attr attr)
fbfc504a BT	65	{
fbfc504a	66	struct xsk_map *m;
819a4f32	67	int numa_node;
e312b9e7	68	u64 size;
fbfc504a	69
fbfc504a BT	70	if (attr->max_entries == 0 \|\| attr->key_size != 4 \|\|
	71	attr->value_size != 4 \|\|
	72	attr->map_flags & ~(BPF_F_NUMA_NODE \| BPF_F_RDONLY \| BPF_F_WRONLY))
	73	return ERR_PTR(-EINVAL);
	74
64fe8c06 BT	75	numa_node = bpf_map_attr_numa_node(attr);
64fe8c06 BT	76	size = struct_size(m, xsk_map, attr->max_entries);
64fe8c06	77
64fe8c06	78	m = bpf_map_area_alloc(size, numa_node);
819a4f32	79	if (!m)
fbfc504a BT	80	return ERR_PTR(-ENOMEM);
	81
	82	bpf_map_init_from_attr(&m->map, attr);
0402acd6	83	spin_lock_init(&m->lock);
fbfc504a	84
fbfc504a	85	return &m->map;
fbfc504a BT	86	}
fbfc504a BT	87
b4fd0d67 YS	88	static u64 xsk_map_mem_usage(const struct bpf_map *map)
	89	{
	90	struct xsk_map *m = container_of(map, struct xsk_map, map);
	91
	92	return struct_size(m, xsk_map, map->max_entries) +
	93	(u64)atomic_read(&m->count) * sizeof(struct xsk_map_node);
	94	}
	95
fbfc504a BT	96	static void xsk_map_free(struct bpf_map *map)
	97	{
	98	struct xsk_map *m = container_of(map, struct xsk_map, map);
fbfc504a BT	99
fbfc504a BT	100	synchronize_net();
64fe8c06	101	bpf_map_area_free(m);
fbfc504a BT	102	}
	103
	104	static int xsk_map_get_next_key(struct bpf_map map, void key, void *next_key)
	105	{
	106	struct xsk_map *m = container_of(map, struct xsk_map, map);
	107	u32 index = key ? (u32 )key : U32_MAX;
	108	u32 *next = next_key;
	109
	110	if (index >= m->map.max_entries) {
	111	*next = 0;
	112	return 0;
	113	}
	114
	115	if (index == m->map.max_entries - 1)
	116	return -ENOENT;
	117	*next = index + 1;
	118	return 0;
	119	}
	120
4a8f87e6	121	static int xsk_map_gen_lookup(struct bpf_map map, struct bpf_insn insn_buf)
e65650f2 MF	122	{
	123	const int ret = BPF_REG_0, mp = BPF_REG_1, index = BPF_REG_2;
	124	struct bpf_insn *insn = insn_buf;
	125
	126	*insn++ = BPF_LDX_MEM(BPF_W, ret, index, 0);
	127	*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 5);
	128	insn++ = BPF_ALU64_IMM(BPF_LSH, ret, ilog2(sizeof(struct xsk_sock )));
	129	*insn++ = BPF_ALU64_IMM(BPF_ADD, mp, offsetof(struct xsk_map, xsk_map));
	130	*insn++ = BPF_ALU64_REG(BPF_ADD, ret, mp);
	131	insn++ = BPF_LDX_MEM(BPF_SIZEOF(struct xsk_sock ), ret, ret, 0);
	132	*insn++ = BPF_JMP_IMM(BPF_JA, 0, 0, 1);
	133	*insn++ = BPF_MOV64_IMM(ret, 0);
	134	return insn - insn_buf;
	135	}
	136
782347b6 THJ	137	/* Elements are kept alive by RCU; either by rcu_read_lock() (from syscall) or
	138	* by local_bh_disable() (from XDP calls inside NAPI). The
	139	* rcu_read_lock_bh_held() below makes lockdep accept both.
	140	*/
e6a4750f BT	141	static void __xsk_map_lookup_elem(struct bpf_map map, u32 key)
	142	{
	143	struct xsk_map *m = container_of(map, struct xsk_map, map);
	144
	145	if (key >= map->max_entries)
	146	return NULL;
	147
782347b6	148	return rcu_dereference_check(m->xsk_map[key], rcu_read_lock_bh_held());
e6a4750f BT	149	}
e6a4750f BT	150
fbfc504a	151	static void xsk_map_lookup_elem(struct bpf_map map, void *key)
fada7fdc	152	{
fada7fdc JL	153	return __xsk_map_lookup_elem(map, (u32 )key);
	154	}
	155
	156	static void xsk_map_lookup_elem_sys_only(struct bpf_map map, void *key)
fbfc504a	157	{
3b4a63f6	158	return ERR_PTR(-EOPNOTSUPP);
fbfc504a BT	159	}
fbfc504a BT	160
d7ba4cc9 JK	161	static long xsk_map_update_elem(struct bpf_map map, void key, void *value,
d7ba4cc9 JK	162	u64 map_flags)
fbfc504a BT	163	{
fbfc504a BT	164	struct xsk_map *m = container_of(map, struct xsk_map, map);
782347b6 THJ	165	struct xdp_sock __rcu **map_entry;
782347b6 THJ	166	struct xdp_sock xs, old_xs;
fbfc504a	167	u32 i = (u32 )key, fd = (u32 )value;
0402acd6	168	struct xsk_map_node *node;
fbfc504a BT	169	struct socket *sock;
	170	int err;
	171
	172	if (unlikely(map_flags > BPF_EXIST))
	173	return -EINVAL;
	174	if (unlikely(i >= m->map.max_entries))
	175	return -E2BIG;
fbfc504a BT	176
	177	sock = sockfd_lookup(fd, &err);
	178	if (!sock)
	179	return err;
	180
	181	if (sock->sk->sk_family != PF_XDP) {
	182	sockfd_put(sock);
	183	return -EOPNOTSUPP;
	184	}
	185
	186	xs = (struct xdp_sock *)sock->sk;
	187
0402acd6 BT	188	map_entry = &m->xsk_map[i];
	189	node = xsk_map_node_alloc(m, map_entry);
	190	if (IS_ERR(node)) {
	191	sockfd_put(sock);
	192	return PTR_ERR(node);
	193	}
fbfc504a	194
0402acd6	195	spin_lock_bh(&m->lock);
782347b6	196	old_xs = rcu_dereference_protected(*map_entry, lockdep_is_held(&m->lock));
0402acd6 BT	197	if (old_xs == xs) {
	198	err = 0;
	199	goto out;
36cc3435 BT	200	} else if (old_xs && map_flags == BPF_NOEXIST) {
	201	err = -EEXIST;
	202	goto out;
	203	} else if (!old_xs && map_flags == BPF_EXIST) {
	204	err = -ENOENT;
	205	goto out;
0402acd6 BT	206	}
0402acd6 BT	207	xsk_map_sock_add(xs, node);
782347b6	208	rcu_assign_pointer(*map_entry, xs);
cee27167	209	if (old_xs)
0402acd6 BT	210	xsk_map_sock_delete(old_xs, map_entry);
0402acd6 BT	211	spin_unlock_bh(&m->lock);
fbfc504a BT	212	sockfd_put(sock);
fbfc504a BT	213	return 0;
0402acd6 BT	214
	215	out:
	216	spin_unlock_bh(&m->lock);
	217	sockfd_put(sock);
	218	xsk_map_node_free(node);
	219	return err;
fbfc504a BT	220	}
fbfc504a BT	221
d7ba4cc9	222	static long xsk_map_delete_elem(struct bpf_map map, void key)
fbfc504a BT	223	{
fbfc504a BT	224	struct xsk_map *m = container_of(map, struct xsk_map, map);
782347b6 THJ	225	struct xdp_sock __rcu **map_entry;
782347b6 THJ	226	struct xdp_sock *old_xs;
fbfc504a BT	227	int k = (u32 )key;
	228
	229	if (k >= map->max_entries)
	230	return -EINVAL;
	231
0402acd6 BT	232	spin_lock_bh(&m->lock);
0402acd6 BT	233	map_entry = &m->xsk_map[k];
782347b6	234	old_xs = unrcu_pointer(xchg(map_entry, NULL));
cee27167	235	if (old_xs)
0402acd6 BT	236	xsk_map_sock_delete(old_xs, map_entry);
0402acd6 BT	237	spin_unlock_bh(&m->lock);
fbfc504a BT	238
	239	return 0;
	240	}
	241
d7ba4cc9	242	static long xsk_map_redirect(struct bpf_map *map, u64 index, u64 flags)
e6a4750f	243	{
32637e33	244	return __bpf_xdp_redirect_map(map, index, flags, 0,
e624d4ed	245	__xsk_map_lookup_elem);
e6a4750f BT	246	}
e6a4750f BT	247
0402acd6	248	void xsk_map_try_sock_delete(struct xsk_map map, struct xdp_sock xs,
782347b6	249	struct xdp_sock __rcu **map_entry)
0402acd6 BT	250	{
0402acd6 BT	251	spin_lock_bh(&map->lock);
782347b6 THJ	252	if (rcu_access_pointer(*map_entry) == xs) {
782347b6 THJ	253	rcu_assign_pointer(*map_entry, NULL);
0402acd6 BT	254	xsk_map_sock_delete(xs, map_entry);
	255	}
	256	spin_unlock_bh(&map->lock);
	257	}
	258
134fede4 MKL	259	static bool xsk_map_meta_equal(const struct bpf_map *meta0,
	260	const struct bpf_map *meta1)
	261	{
	262	return meta0->max_entries == meta1->max_entries &&
	263	bpf_map_meta_equal(meta0, meta1);
	264	}
	265
c317ab71	266	BTF_ID_LIST_SINGLE(xsk_map_btf_ids, struct, xsk_map)
fbfc504a	267	const struct bpf_map_ops xsk_map_ops = {
134fede4	268	.map_meta_equal = xsk_map_meta_equal,
fbfc504a BT	269	.map_alloc = xsk_map_alloc,
	270	.map_free = xsk_map_free,
	271	.map_get_next_key = xsk_map_get_next_key,
	272	.map_lookup_elem = xsk_map_lookup_elem,
e65650f2	273	.map_gen_lookup = xsk_map_gen_lookup,
fada7fdc	274	.map_lookup_elem_sys_only = xsk_map_lookup_elem_sys_only,
fbfc504a BT	275	.map_update_elem = xsk_map_update_elem,
fbfc504a BT	276	.map_delete_elem = xsk_map_delete_elem,
e8d2bec0	277	.map_check_btf = map_check_no_btf,
b4fd0d67	278	.map_mem_usage = xsk_map_mem_usage,
c317ab71	279	.map_btf_id = &xsk_map_btf_ids[0],
e6a4750f	280	.map_redirect = xsk_map_redirect,
fbfc504a	281	};