1 // SPDX-License-Identifier: GPL-2.0
2 /* XSKMAP used for AF_XDP sockets
3 * Copyright(c) 2018 Intel Corporation.
5 * This program is free software; you can redistribute it and/or modify it
6 * under the terms and conditions of the GNU General Public License,
7 * version 2, as published by the Free Software Foundation.
9 * This program is distributed in the hope it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 #include <linux/bpf.h>
16 #include <linux/capability.h>
17 #include <net/xdp_sock.h>
18 #include <linux/slab.h>
19 #include <linux/sched.h>
23 struct xdp_sock **xsk_map;
24 struct list_head __percpu *flush_list;
27 static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
29 int cpu, err = -EINVAL;
33 if (!capable(CAP_NET_ADMIN))
34 return ERR_PTR(-EPERM);
36 if (attr->max_entries == 0 || attr->key_size != 4 ||
37 attr->value_size != 4 ||
38 attr->map_flags & ~(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY))
39 return ERR_PTR(-EINVAL);
41 m = kzalloc(sizeof(*m), GFP_USER);
43 return ERR_PTR(-ENOMEM);
45 bpf_map_init_from_attr(&m->map, attr);
47 cost = (u64)m->map.max_entries * sizeof(struct xdp_sock *);
48 cost += sizeof(struct list_head) * num_possible_cpus();
49 if (cost >= U32_MAX - PAGE_SIZE)
52 m->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
54 /* Notice returns -EPERM on if map size is larger than memlock limit */
55 err = bpf_map_precharge_memlock(m->map.pages);
59 m->flush_list = alloc_percpu(struct list_head);
63 for_each_possible_cpu(cpu)
64 INIT_LIST_HEAD(per_cpu_ptr(m->flush_list, cpu));
66 m->xsk_map = bpf_map_area_alloc(m->map.max_entries *
67 sizeof(struct xdp_sock *),
74 free_percpu(m->flush_list);
80 static void xsk_map_free(struct bpf_map *map)
82 struct xsk_map *m = container_of(map, struct xsk_map, map);
87 for (i = 0; i < map->max_entries; i++) {
94 sock_put((struct sock *)xs);
97 free_percpu(m->flush_list);
98 bpf_map_area_free(m->xsk_map);
102 static int xsk_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
104 struct xsk_map *m = container_of(map, struct xsk_map, map);
105 u32 index = key ? *(u32 *)key : U32_MAX;
106 u32 *next = next_key;
108 if (index >= m->map.max_entries) {
113 if (index == m->map.max_entries - 1)
119 struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, u32 key)
121 struct xsk_map *m = container_of(map, struct xsk_map, map);
124 if (key >= map->max_entries)
127 xs = READ_ONCE(m->xsk_map[key]);
131 int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp,
134 struct xsk_map *m = container_of(map, struct xsk_map, map);
135 struct list_head *flush_list = this_cpu_ptr(m->flush_list);
138 err = xsk_rcv(xs, xdp);
142 if (!xs->flush_node.prev)
143 list_add(&xs->flush_node, flush_list);
148 void __xsk_map_flush(struct bpf_map *map)
150 struct xsk_map *m = container_of(map, struct xsk_map, map);
151 struct list_head *flush_list = this_cpu_ptr(m->flush_list);
152 struct xdp_sock *xs, *tmp;
154 list_for_each_entry_safe(xs, tmp, flush_list, flush_node) {
156 __list_del(xs->flush_node.prev, xs->flush_node.next);
157 xs->flush_node.prev = NULL;
161 static void *xsk_map_lookup_elem(struct bpf_map *map, void *key)
166 static int xsk_map_update_elem(struct bpf_map *map, void *key, void *value,
169 struct xsk_map *m = container_of(map, struct xsk_map, map);
170 u32 i = *(u32 *)key, fd = *(u32 *)value;
171 struct xdp_sock *xs, *old_xs;
175 if (unlikely(map_flags > BPF_EXIST))
177 if (unlikely(i >= m->map.max_entries))
179 if (unlikely(map_flags == BPF_NOEXIST))
182 sock = sockfd_lookup(fd, &err);
186 if (sock->sk->sk_family != PF_XDP) {
191 xs = (struct xdp_sock *)sock->sk;
193 if (!xsk_is_setup_for_bpf_map(xs)) {
200 old_xs = xchg(&m->xsk_map[i], xs);
202 /* Make sure we've flushed everything. */
204 sock_put((struct sock *)old_xs);
211 static int xsk_map_delete_elem(struct bpf_map *map, void *key)
213 struct xsk_map *m = container_of(map, struct xsk_map, map);
214 struct xdp_sock *old_xs;
217 if (k >= map->max_entries)
220 old_xs = xchg(&m->xsk_map[k], NULL);
222 /* Make sure we've flushed everything. */
224 sock_put((struct sock *)old_xs);
230 const struct bpf_map_ops xsk_map_ops = {
231 .map_alloc = xsk_map_alloc,
232 .map_free = xsk_map_free,
233 .map_get_next_key = xsk_map_get_next_key,
234 .map_lookup_elem = xsk_map_lookup_elem,
235 .map_update_elem = xsk_map_update_elem,
236 .map_delete_elem = xsk_map_delete_elem,