xref: /openbmc/linux/kernel/bpf/reuseport_array.c (revision e285d5bf)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2018 Facebook
4  */
5 #include <linux/bpf.h>
6 #include <linux/err.h>
7 #include <linux/sock_diag.h>
8 #include <net/sock_reuseport.h>
9 
10 struct reuseport_array {
11 	struct bpf_map map;
12 	struct sock __rcu *ptrs[];
13 };
14 
15 static struct reuseport_array *reuseport_array(struct bpf_map *map)
16 {
17 	return (struct reuseport_array *)map;
18 }
19 
20 /* The caller must hold the reuseport_lock */
21 void bpf_sk_reuseport_detach(struct sock *sk)
22 {
23 	struct sock __rcu **socks;
24 
25 	write_lock_bh(&sk->sk_callback_lock);
26 	socks = sk->sk_user_data;
27 	if (socks) {
28 		WRITE_ONCE(sk->sk_user_data, NULL);
29 		/*
30 		 * Do not move this NULL assignment outside of
31 		 * sk->sk_callback_lock because there is
32 		 * a race with reuseport_array_free()
33 		 * which does not hold the reuseport_lock.
34 		 */
35 		RCU_INIT_POINTER(*socks, NULL);
36 	}
37 	write_unlock_bh(&sk->sk_callback_lock);
38 }
39 
40 static int reuseport_array_alloc_check(union bpf_attr *attr)
41 {
42 	if (attr->value_size != sizeof(u32) &&
43 	    attr->value_size != sizeof(u64))
44 		return -EINVAL;
45 
46 	return array_map_alloc_check(attr);
47 }
48 
49 static void *reuseport_array_lookup_elem(struct bpf_map *map, void *key)
50 {
51 	struct reuseport_array *array = reuseport_array(map);
52 	u32 index = *(u32 *)key;
53 
54 	if (unlikely(index >= array->map.max_entries))
55 		return NULL;
56 
57 	return rcu_dereference(array->ptrs[index]);
58 }
59 
60 /* Called from syscall only */
61 static int reuseport_array_delete_elem(struct bpf_map *map, void *key)
62 {
63 	struct reuseport_array *array = reuseport_array(map);
64 	u32 index = *(u32 *)key;
65 	struct sock *sk;
66 	int err;
67 
68 	if (index >= map->max_entries)
69 		return -E2BIG;
70 
71 	if (!rcu_access_pointer(array->ptrs[index]))
72 		return -ENOENT;
73 
74 	spin_lock_bh(&reuseport_lock);
75 
76 	sk = rcu_dereference_protected(array->ptrs[index],
77 				       lockdep_is_held(&reuseport_lock));
78 	if (sk) {
79 		write_lock_bh(&sk->sk_callback_lock);
80 		WRITE_ONCE(sk->sk_user_data, NULL);
81 		RCU_INIT_POINTER(array->ptrs[index], NULL);
82 		write_unlock_bh(&sk->sk_callback_lock);
83 		err = 0;
84 	} else {
85 		err = -ENOENT;
86 	}
87 
88 	spin_unlock_bh(&reuseport_lock);
89 
90 	return err;
91 }
92 
93 static void reuseport_array_free(struct bpf_map *map)
94 {
95 	struct reuseport_array *array = reuseport_array(map);
96 	struct sock *sk;
97 	u32 i;
98 
99 	synchronize_rcu();
100 
101 	/*
102 	 * ops->map_*_elem() will not be able to access this
103 	 * array now. Hence, this function only races with
104 	 * bpf_sk_reuseport_detach() which was triggerred by
105 	 * close() or disconnect().
106 	 *
107 	 * This function and bpf_sk_reuseport_detach() are
108 	 * both removing sk from "array".  Who removes it
109 	 * first does not matter.
110 	 *
111 	 * The only concern here is bpf_sk_reuseport_detach()
112 	 * may access "array" which is being freed here.
113 	 * bpf_sk_reuseport_detach() access this "array"
114 	 * through sk->sk_user_data _and_ with sk->sk_callback_lock
115 	 * held which is enough because this "array" is not freed
116 	 * until all sk->sk_user_data has stopped referencing this "array".
117 	 *
118 	 * Hence, due to the above, taking "reuseport_lock" is not
119 	 * needed here.
120 	 */
121 
122 	/*
123 	 * Since reuseport_lock is not taken, sk is accessed under
124 	 * rcu_read_lock()
125 	 */
126 	rcu_read_lock();
127 	for (i = 0; i < map->max_entries; i++) {
128 		sk = rcu_dereference(array->ptrs[i]);
129 		if (sk) {
130 			write_lock_bh(&sk->sk_callback_lock);
131 			/*
132 			 * No need for WRITE_ONCE(). At this point,
133 			 * no one is reading it without taking the
134 			 * sk->sk_callback_lock.
135 			 */
136 			sk->sk_user_data = NULL;
137 			write_unlock_bh(&sk->sk_callback_lock);
138 			RCU_INIT_POINTER(array->ptrs[i], NULL);
139 		}
140 	}
141 	rcu_read_unlock();
142 
143 	/*
144 	 * Once reaching here, all sk->sk_user_data is not
145 	 * referenceing this "array".  "array" can be freed now.
146 	 */
147 	bpf_map_area_free(array);
148 }
149 
150 static struct bpf_map *reuseport_array_alloc(union bpf_attr *attr)
151 {
152 	int err, numa_node = bpf_map_attr_numa_node(attr);
153 	struct reuseport_array *array;
154 	u64 cost, array_size;
155 
156 	if (!capable(CAP_SYS_ADMIN))
157 		return ERR_PTR(-EPERM);
158 
159 	array_size = sizeof(*array);
160 	array_size += (u64)attr->max_entries * sizeof(struct sock *);
161 
162 	/* make sure there is no u32 overflow later in round_up() */
163 	cost = array_size;
164 	if (cost >= U32_MAX - PAGE_SIZE)
165 		return ERR_PTR(-ENOMEM);
166 	cost = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
167 
168 	err = bpf_map_precharge_memlock(cost);
169 	if (err)
170 		return ERR_PTR(err);
171 
172 	/* allocate all map elements and zero-initialize them */
173 	array = bpf_map_area_alloc(array_size, numa_node);
174 	if (!array)
175 		return ERR_PTR(-ENOMEM);
176 
177 	/* copy mandatory map attributes */
178 	bpf_map_init_from_attr(&array->map, attr);
179 	array->map.pages = cost;
180 
181 	return &array->map;
182 }
183 
184 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
185 				       void *value)
186 {
187 	struct sock *sk;
188 	int err;
189 
190 	if (map->value_size != sizeof(u64))
191 		return -ENOSPC;
192 
193 	rcu_read_lock();
194 	sk = reuseport_array_lookup_elem(map, key);
195 	if (sk) {
196 		*(u64 *)value = sock_gen_cookie(sk);
197 		err = 0;
198 	} else {
199 		err = -ENOENT;
200 	}
201 	rcu_read_unlock();
202 
203 	return err;
204 }
205 
206 static int
207 reuseport_array_update_check(const struct reuseport_array *array,
208 			     const struct sock *nsk,
209 			     const struct sock *osk,
210 			     const struct sock_reuseport *nsk_reuse,
211 			     u32 map_flags)
212 {
213 	if (osk && map_flags == BPF_NOEXIST)
214 		return -EEXIST;
215 
216 	if (!osk && map_flags == BPF_EXIST)
217 		return -ENOENT;
218 
219 	if (nsk->sk_protocol != IPPROTO_UDP && nsk->sk_protocol != IPPROTO_TCP)
220 		return -ENOTSUPP;
221 
222 	if (nsk->sk_family != AF_INET && nsk->sk_family != AF_INET6)
223 		return -ENOTSUPP;
224 
225 	if (nsk->sk_type != SOCK_STREAM && nsk->sk_type != SOCK_DGRAM)
226 		return -ENOTSUPP;
227 
228 	/*
229 	 * sk must be hashed (i.e. listening in the TCP case or binded
230 	 * in the UDP case) and
231 	 * it must also be a SO_REUSEPORT sk (i.e. reuse cannot be NULL).
232 	 *
233 	 * Also, sk will be used in bpf helper that is protected by
234 	 * rcu_read_lock().
235 	 */
236 	if (!sock_flag(nsk, SOCK_RCU_FREE) || !sk_hashed(nsk) || !nsk_reuse)
237 		return -EINVAL;
238 
239 	/* READ_ONCE because the sk->sk_callback_lock may not be held here */
240 	if (READ_ONCE(nsk->sk_user_data))
241 		return -EBUSY;
242 
243 	return 0;
244 }
245 
246 /*
247  * Called from syscall only.
248  * The "nsk" in the fd refcnt.
249  * The "osk" and "reuse" are protected by reuseport_lock.
250  */
251 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
252 				       void *value, u64 map_flags)
253 {
254 	struct reuseport_array *array = reuseport_array(map);
255 	struct sock *free_osk = NULL, *osk, *nsk;
256 	struct sock_reuseport *reuse;
257 	u32 index = *(u32 *)key;
258 	struct socket *socket;
259 	int err, fd;
260 
261 	if (map_flags > BPF_EXIST)
262 		return -EINVAL;
263 
264 	if (index >= map->max_entries)
265 		return -E2BIG;
266 
267 	if (map->value_size == sizeof(u64)) {
268 		u64 fd64 = *(u64 *)value;
269 
270 		if (fd64 > S32_MAX)
271 			return -EINVAL;
272 		fd = fd64;
273 	} else {
274 		fd = *(int *)value;
275 	}
276 
277 	socket = sockfd_lookup(fd, &err);
278 	if (!socket)
279 		return err;
280 
281 	nsk = socket->sk;
282 	if (!nsk) {
283 		err = -EINVAL;
284 		goto put_file;
285 	}
286 
287 	/* Quick checks before taking reuseport_lock */
288 	err = reuseport_array_update_check(array, nsk,
289 					   rcu_access_pointer(array->ptrs[index]),
290 					   rcu_access_pointer(nsk->sk_reuseport_cb),
291 					   map_flags);
292 	if (err)
293 		goto put_file;
294 
295 	spin_lock_bh(&reuseport_lock);
296 	/*
297 	 * Some of the checks only need reuseport_lock
298 	 * but it is done under sk_callback_lock also
299 	 * for simplicity reason.
300 	 */
301 	write_lock_bh(&nsk->sk_callback_lock);
302 
303 	osk = rcu_dereference_protected(array->ptrs[index],
304 					lockdep_is_held(&reuseport_lock));
305 	reuse = rcu_dereference_protected(nsk->sk_reuseport_cb,
306 					  lockdep_is_held(&reuseport_lock));
307 	err = reuseport_array_update_check(array, nsk, osk, reuse, map_flags);
308 	if (err)
309 		goto put_file_unlock;
310 
311 	/* Ensure reuse->reuseport_id is set */
312 	err = reuseport_get_id(reuse);
313 	if (err < 0)
314 		goto put_file_unlock;
315 
316 	WRITE_ONCE(nsk->sk_user_data, &array->ptrs[index]);
317 	rcu_assign_pointer(array->ptrs[index], nsk);
318 	free_osk = osk;
319 	err = 0;
320 
321 put_file_unlock:
322 	write_unlock_bh(&nsk->sk_callback_lock);
323 
324 	if (free_osk) {
325 		write_lock_bh(&free_osk->sk_callback_lock);
326 		WRITE_ONCE(free_osk->sk_user_data, NULL);
327 		write_unlock_bh(&free_osk->sk_callback_lock);
328 	}
329 
330 	spin_unlock_bh(&reuseport_lock);
331 put_file:
332 	fput(socket->file);
333 	return err;
334 }
335 
336 /* Called from syscall */
337 static int reuseport_array_get_next_key(struct bpf_map *map, void *key,
338 					void *next_key)
339 {
340 	struct reuseport_array *array = reuseport_array(map);
341 	u32 index = key ? *(u32 *)key : U32_MAX;
342 	u32 *next = (u32 *)next_key;
343 
344 	if (index >= array->map.max_entries) {
345 		*next = 0;
346 		return 0;
347 	}
348 
349 	if (index == array->map.max_entries - 1)
350 		return -ENOENT;
351 
352 	*next = index + 1;
353 	return 0;
354 }
355 
356 const struct bpf_map_ops reuseport_array_ops = {
357 	.map_alloc_check = reuseport_array_alloc_check,
358 	.map_alloc = reuseport_array_alloc,
359 	.map_free = reuseport_array_free,
360 	.map_lookup_elem = reuseport_array_lookup_elem,
361 	.map_get_next_key = reuseport_array_get_next_key,
362 	.map_delete_elem = reuseport_array_delete_elem,
363 };
364