xref: /linux/net/mctp/af_mctp.c (revision aa74c44be19c8b1de38d955c2c45c309991c805a)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Management Component Transport Protocol (MCTP)
4  *
5  * Copyright (c) 2021 Code Construct
6  * Copyright (c) 2021 Google
7  */
8 
9 #include <linux/if_arp.h>
10 #include <linux/net.h>
11 #include <linux/mctp.h>
12 #include <linux/module.h>
13 #include <linux/socket.h>
14 
15 #include <net/mctp.h>
16 #include <net/mctpdevice.h>
17 #include <net/sock.h>
18 
19 #define CREATE_TRACE_POINTS
20 #include <trace/events/mctp.h>
21 
22 /* socket implementation */
23 
24 static int mctp_release(struct socket *sock)
25 {
26 	struct sock *sk = sock->sk;
27 
28 	if (sk) {
29 		sock->sk = NULL;
30 		sk->sk_prot->close(sk, 0);
31 	}
32 
33 	return 0;
34 }
35 
36 /* Generic sockaddr checks, padding checks only so far */
37 static bool mctp_sockaddr_is_ok(const struct sockaddr_mctp *addr)
38 {
39 	return !addr->__smctp_pad0 && !addr->__smctp_pad1;
40 }
41 
42 static bool mctp_sockaddr_ext_is_ok(const struct sockaddr_mctp_ext *addr)
43 {
44 	return !addr->__smctp_pad0[0] &&
45 	       !addr->__smctp_pad0[1] &&
46 	       !addr->__smctp_pad0[2];
47 }
48 
49 static int mctp_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
50 {
51 	struct sock *sk = sock->sk;
52 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
53 	struct sockaddr_mctp *smctp;
54 	int rc;
55 
56 	if (addrlen < sizeof(*smctp))
57 		return -EINVAL;
58 
59 	if (addr->sa_family != AF_MCTP)
60 		return -EAFNOSUPPORT;
61 
62 	if (!capable(CAP_NET_BIND_SERVICE))
63 		return -EACCES;
64 
65 	/* it's a valid sockaddr for MCTP, cast and do protocol checks */
66 	smctp = (struct sockaddr_mctp *)addr;
67 
68 	if (!mctp_sockaddr_is_ok(smctp))
69 		return -EINVAL;
70 
71 	lock_sock(sk);
72 
73 	/* TODO: allow rebind */
74 	if (sk_hashed(sk)) {
75 		rc = -EADDRINUSE;
76 		goto out_release;
77 	}
78 	msk->bind_net = smctp->smctp_network;
79 	msk->bind_addr = smctp->smctp_addr.s_addr;
80 	msk->bind_type = smctp->smctp_type & 0x7f; /* ignore the IC bit */
81 
82 	rc = sk->sk_prot->hash(sk);
83 
84 out_release:
85 	release_sock(sk);
86 
87 	return rc;
88 }
89 
90 static int mctp_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
91 {
92 	DECLARE_SOCKADDR(struct sockaddr_mctp *, addr, msg->msg_name);
93 	const int hlen = MCTP_HEADER_MAXLEN + sizeof(struct mctp_hdr);
94 	int rc, addrlen = msg->msg_namelen;
95 	struct sock *sk = sock->sk;
96 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
97 	struct mctp_skb_cb *cb;
98 	struct mctp_route *rt;
99 	struct sk_buff *skb;
100 
101 	if (addr) {
102 		if (addrlen < sizeof(struct sockaddr_mctp))
103 			return -EINVAL;
104 		if (addr->smctp_family != AF_MCTP)
105 			return -EINVAL;
106 		if (!mctp_sockaddr_is_ok(addr))
107 			return -EINVAL;
108 		if (addr->smctp_tag & ~(MCTP_TAG_MASK | MCTP_TAG_OWNER))
109 			return -EINVAL;
110 
111 	} else {
112 		/* TODO: connect()ed sockets */
113 		return -EDESTADDRREQ;
114 	}
115 
116 	if (!capable(CAP_NET_RAW))
117 		return -EACCES;
118 
119 	if (addr->smctp_network == MCTP_NET_ANY)
120 		addr->smctp_network = mctp_default_net(sock_net(sk));
121 
122 	skb = sock_alloc_send_skb(sk, hlen + 1 + len,
123 				  msg->msg_flags & MSG_DONTWAIT, &rc);
124 	if (!skb)
125 		return rc;
126 
127 	skb_reserve(skb, hlen);
128 
129 	/* set type as fist byte in payload */
130 	*(u8 *)skb_put(skb, 1) = addr->smctp_type;
131 
132 	rc = memcpy_from_msg((void *)skb_put(skb, len), msg, len);
133 	if (rc < 0)
134 		goto err_free;
135 
136 	/* set up cb */
137 	cb = __mctp_cb(skb);
138 	cb->net = addr->smctp_network;
139 
140 	/* direct addressing */
141 	if (msk->addr_ext && addrlen >= sizeof(struct sockaddr_mctp_ext)) {
142 		DECLARE_SOCKADDR(struct sockaddr_mctp_ext *,
143 				 extaddr, msg->msg_name);
144 
145 		if (!mctp_sockaddr_ext_is_ok(extaddr) ||
146 		    extaddr->smctp_halen > sizeof(cb->haddr)) {
147 			rc = -EINVAL;
148 			goto err_free;
149 		}
150 
151 		cb->ifindex = extaddr->smctp_ifindex;
152 		cb->halen = extaddr->smctp_halen;
153 		memcpy(cb->haddr, extaddr->smctp_haddr, cb->halen);
154 
155 		rt = NULL;
156 	} else {
157 		rt = mctp_route_lookup(sock_net(sk), addr->smctp_network,
158 				       addr->smctp_addr.s_addr);
159 		if (!rt) {
160 			rc = -EHOSTUNREACH;
161 			goto err_free;
162 		}
163 	}
164 
165 	rc = mctp_local_output(sk, rt, skb, addr->smctp_addr.s_addr,
166 			       addr->smctp_tag);
167 
168 	return rc ? : len;
169 
170 err_free:
171 	kfree_skb(skb);
172 	return rc;
173 }
174 
175 static int mctp_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
176 			int flags)
177 {
178 	DECLARE_SOCKADDR(struct sockaddr_mctp *, addr, msg->msg_name);
179 	struct sock *sk = sock->sk;
180 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
181 	struct sk_buff *skb;
182 	size_t msglen;
183 	u8 type;
184 	int rc;
185 
186 	if (flags & ~(MSG_DONTWAIT | MSG_TRUNC | MSG_PEEK))
187 		return -EOPNOTSUPP;
188 
189 	skb = skb_recv_datagram(sk, flags, flags & MSG_DONTWAIT, &rc);
190 	if (!skb)
191 		return rc;
192 
193 	if (!skb->len) {
194 		rc = 0;
195 		goto out_free;
196 	}
197 
198 	/* extract message type, remove from data */
199 	type = *((u8 *)skb->data);
200 	msglen = skb->len - 1;
201 
202 	if (len < msglen)
203 		msg->msg_flags |= MSG_TRUNC;
204 	else
205 		len = msglen;
206 
207 	rc = skb_copy_datagram_msg(skb, 1, msg, len);
208 	if (rc < 0)
209 		goto out_free;
210 
211 	sock_recv_ts_and_drops(msg, sk, skb);
212 
213 	if (addr) {
214 		struct mctp_skb_cb *cb = mctp_cb(skb);
215 		/* TODO: expand mctp_skb_cb for header fields? */
216 		struct mctp_hdr *hdr = mctp_hdr(skb);
217 
218 		addr = msg->msg_name;
219 		addr->smctp_family = AF_MCTP;
220 		addr->__smctp_pad0 = 0;
221 		addr->smctp_network = cb->net;
222 		addr->smctp_addr.s_addr = hdr->src;
223 		addr->smctp_type = type;
224 		addr->smctp_tag = hdr->flags_seq_tag &
225 					(MCTP_HDR_TAG_MASK | MCTP_HDR_FLAG_TO);
226 		addr->__smctp_pad1 = 0;
227 		msg->msg_namelen = sizeof(*addr);
228 
229 		if (msk->addr_ext) {
230 			DECLARE_SOCKADDR(struct sockaddr_mctp_ext *, ae,
231 					 msg->msg_name);
232 			msg->msg_namelen = sizeof(*ae);
233 			ae->smctp_ifindex = cb->ifindex;
234 			ae->smctp_halen = cb->halen;
235 			memset(ae->__smctp_pad0, 0x0, sizeof(ae->__smctp_pad0));
236 			memset(ae->smctp_haddr, 0x0, sizeof(ae->smctp_haddr));
237 			memcpy(ae->smctp_haddr, cb->haddr, cb->halen);
238 		}
239 	}
240 
241 	rc = len;
242 
243 	if (flags & MSG_TRUNC)
244 		rc = msglen;
245 
246 out_free:
247 	skb_free_datagram(sk, skb);
248 	return rc;
249 }
250 
251 static int mctp_setsockopt(struct socket *sock, int level, int optname,
252 			   sockptr_t optval, unsigned int optlen)
253 {
254 	struct mctp_sock *msk = container_of(sock->sk, struct mctp_sock, sk);
255 	int val;
256 
257 	if (level != SOL_MCTP)
258 		return -EINVAL;
259 
260 	if (optname == MCTP_OPT_ADDR_EXT) {
261 		if (optlen != sizeof(int))
262 			return -EINVAL;
263 		if (copy_from_sockptr(&val, optval, sizeof(int)))
264 			return -EFAULT;
265 		msk->addr_ext = val;
266 		return 0;
267 	}
268 
269 	return -ENOPROTOOPT;
270 }
271 
272 static int mctp_getsockopt(struct socket *sock, int level, int optname,
273 			   char __user *optval, int __user *optlen)
274 {
275 	struct mctp_sock *msk = container_of(sock->sk, struct mctp_sock, sk);
276 	int len, val;
277 
278 	if (level != SOL_MCTP)
279 		return -EINVAL;
280 
281 	if (get_user(len, optlen))
282 		return -EFAULT;
283 
284 	if (optname == MCTP_OPT_ADDR_EXT) {
285 		if (len != sizeof(int))
286 			return -EINVAL;
287 		val = !!msk->addr_ext;
288 		if (copy_to_user(optval, &val, len))
289 			return -EFAULT;
290 		return 0;
291 	}
292 
293 	return -EINVAL;
294 }
295 
296 static const struct proto_ops mctp_dgram_ops = {
297 	.family		= PF_MCTP,
298 	.release	= mctp_release,
299 	.bind		= mctp_bind,
300 	.connect	= sock_no_connect,
301 	.socketpair	= sock_no_socketpair,
302 	.accept		= sock_no_accept,
303 	.getname	= sock_no_getname,
304 	.poll		= datagram_poll,
305 	.ioctl		= sock_no_ioctl,
306 	.gettstamp	= sock_gettstamp,
307 	.listen		= sock_no_listen,
308 	.shutdown	= sock_no_shutdown,
309 	.setsockopt	= mctp_setsockopt,
310 	.getsockopt	= mctp_getsockopt,
311 	.sendmsg	= mctp_sendmsg,
312 	.recvmsg	= mctp_recvmsg,
313 	.mmap		= sock_no_mmap,
314 	.sendpage	= sock_no_sendpage,
315 };
316 
317 static void mctp_sk_expire_keys(struct timer_list *timer)
318 {
319 	struct mctp_sock *msk = container_of(timer, struct mctp_sock,
320 					     key_expiry);
321 	struct net *net = sock_net(&msk->sk);
322 	unsigned long next_expiry, flags;
323 	struct mctp_sk_key *key;
324 	struct hlist_node *tmp;
325 	bool next_expiry_valid = false;
326 
327 	spin_lock_irqsave(&net->mctp.keys_lock, flags);
328 
329 	hlist_for_each_entry_safe(key, tmp, &msk->keys, sklist) {
330 		spin_lock(&key->lock);
331 
332 		if (!time_after_eq(key->expiry, jiffies)) {
333 			trace_mctp_key_release(key, MCTP_TRACE_KEY_TIMEOUT);
334 			key->valid = false;
335 			hlist_del_rcu(&key->hlist);
336 			hlist_del_rcu(&key->sklist);
337 			spin_unlock(&key->lock);
338 			mctp_key_unref(key);
339 			continue;
340 		}
341 
342 		if (next_expiry_valid) {
343 			if (time_before(key->expiry, next_expiry))
344 				next_expiry = key->expiry;
345 		} else {
346 			next_expiry = key->expiry;
347 			next_expiry_valid = true;
348 		}
349 		spin_unlock(&key->lock);
350 	}
351 
352 	spin_unlock_irqrestore(&net->mctp.keys_lock, flags);
353 
354 	if (next_expiry_valid)
355 		mod_timer(timer, next_expiry);
356 }
357 
358 static int mctp_sk_init(struct sock *sk)
359 {
360 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
361 
362 	INIT_HLIST_HEAD(&msk->keys);
363 	timer_setup(&msk->key_expiry, mctp_sk_expire_keys, 0);
364 	return 0;
365 }
366 
367 static void mctp_sk_close(struct sock *sk, long timeout)
368 {
369 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
370 
371 	del_timer_sync(&msk->key_expiry);
372 	sk_common_release(sk);
373 }
374 
375 static int mctp_sk_hash(struct sock *sk)
376 {
377 	struct net *net = sock_net(sk);
378 
379 	mutex_lock(&net->mctp.bind_lock);
380 	sk_add_node_rcu(sk, &net->mctp.binds);
381 	mutex_unlock(&net->mctp.bind_lock);
382 
383 	return 0;
384 }
385 
386 static void mctp_sk_unhash(struct sock *sk)
387 {
388 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
389 	struct net *net = sock_net(sk);
390 	struct mctp_sk_key *key;
391 	struct hlist_node *tmp;
392 	unsigned long flags;
393 
394 	/* remove from any type-based binds */
395 	mutex_lock(&net->mctp.bind_lock);
396 	sk_del_node_init_rcu(sk);
397 	mutex_unlock(&net->mctp.bind_lock);
398 
399 	/* remove tag allocations */
400 	spin_lock_irqsave(&net->mctp.keys_lock, flags);
401 	hlist_for_each_entry_safe(key, tmp, &msk->keys, sklist) {
402 		hlist_del(&key->sklist);
403 		hlist_del(&key->hlist);
404 
405 		trace_mctp_key_release(key, MCTP_TRACE_KEY_CLOSED);
406 
407 		spin_lock(&key->lock);
408 		kfree_skb(key->reasm_head);
409 		key->reasm_head = NULL;
410 		key->reasm_dead = true;
411 		key->valid = false;
412 		spin_unlock(&key->lock);
413 
414 		/* key is no longer on the lookup lists, unref */
415 		mctp_key_unref(key);
416 	}
417 	spin_unlock_irqrestore(&net->mctp.keys_lock, flags);
418 }
419 
420 static struct proto mctp_proto = {
421 	.name		= "MCTP",
422 	.owner		= THIS_MODULE,
423 	.obj_size	= sizeof(struct mctp_sock),
424 	.init		= mctp_sk_init,
425 	.close		= mctp_sk_close,
426 	.hash		= mctp_sk_hash,
427 	.unhash		= mctp_sk_unhash,
428 };
429 
430 static int mctp_pf_create(struct net *net, struct socket *sock,
431 			  int protocol, int kern)
432 {
433 	const struct proto_ops *ops;
434 	struct proto *proto;
435 	struct sock *sk;
436 	int rc;
437 
438 	if (protocol)
439 		return -EPROTONOSUPPORT;
440 
441 	/* only datagram sockets are supported */
442 	if (sock->type != SOCK_DGRAM)
443 		return -ESOCKTNOSUPPORT;
444 
445 	proto = &mctp_proto;
446 	ops = &mctp_dgram_ops;
447 
448 	sock->state = SS_UNCONNECTED;
449 	sock->ops = ops;
450 
451 	sk = sk_alloc(net, PF_MCTP, GFP_KERNEL, proto, kern);
452 	if (!sk)
453 		return -ENOMEM;
454 
455 	sock_init_data(sock, sk);
456 
457 	rc = 0;
458 	if (sk->sk_prot->init)
459 		rc = sk->sk_prot->init(sk);
460 
461 	if (rc)
462 		goto err_sk_put;
463 
464 	return 0;
465 
466 err_sk_put:
467 	sock_orphan(sk);
468 	sock_put(sk);
469 	return rc;
470 }
471 
472 static struct net_proto_family mctp_pf = {
473 	.family = PF_MCTP,
474 	.create = mctp_pf_create,
475 	.owner = THIS_MODULE,
476 };
477 
478 static __init int mctp_init(void)
479 {
480 	int rc;
481 
482 	/* ensure our uapi tag definitions match the header format */
483 	BUILD_BUG_ON(MCTP_TAG_OWNER != MCTP_HDR_FLAG_TO);
484 	BUILD_BUG_ON(MCTP_TAG_MASK != MCTP_HDR_TAG_MASK);
485 
486 	pr_info("mctp: management component transport protocol core\n");
487 
488 	rc = sock_register(&mctp_pf);
489 	if (rc)
490 		return rc;
491 
492 	rc = proto_register(&mctp_proto, 0);
493 	if (rc)
494 		goto err_unreg_sock;
495 
496 	rc = mctp_routes_init();
497 	if (rc)
498 		goto err_unreg_proto;
499 
500 	rc = mctp_neigh_init();
501 	if (rc)
502 		goto err_unreg_proto;
503 
504 	mctp_device_init();
505 
506 	return 0;
507 
508 err_unreg_proto:
509 	proto_unregister(&mctp_proto);
510 err_unreg_sock:
511 	sock_unregister(PF_MCTP);
512 
513 	return rc;
514 }
515 
516 static __exit void mctp_exit(void)
517 {
518 	mctp_device_exit();
519 	mctp_neigh_exit();
520 	mctp_routes_exit();
521 	proto_unregister(&mctp_proto);
522 	sock_unregister(PF_MCTP);
523 }
524 
525 subsys_initcall(mctp_init);
526 module_exit(mctp_exit);
527 
528 MODULE_DESCRIPTION("MCTP core");
529 MODULE_LICENSE("GPL v2");
530 MODULE_AUTHOR("Jeremy Kerr <jk@codeconstruct.com.au>");
531 
532 MODULE_ALIAS_NETPROTO(PF_MCTP);
533