xref: /linux/net/mctp/af_mctp.c (revision d69eb204c255c35abd9e8cb621484e8074c75eaa)
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/compat.h>
10 #include <linux/if_arp.h>
11 #include <linux/net.h>
12 #include <linux/mctp.h>
13 #include <linux/module.h>
14 #include <linux/socket.h>
15 
16 #include <net/mctp.h>
17 #include <net/mctpdevice.h>
18 #include <net/sock.h>
19 
20 #define CREATE_TRACE_POINTS
21 #include <trace/events/mctp.h>
22 
23 /* socket implementation */
24 
25 static void mctp_sk_expire_keys(struct timer_list *timer);
26 
mctp_release(struct socket * sock)27 static int mctp_release(struct socket *sock)
28 {
29 	struct sock *sk = sock->sk;
30 
31 	if (sk) {
32 		sock->sk = NULL;
33 		sk->sk_prot->close(sk, 0);
34 	}
35 
36 	return 0;
37 }
38 
39 /* Generic sockaddr checks, padding checks only so far */
mctp_sockaddr_is_ok(const struct sockaddr_mctp * addr)40 static bool mctp_sockaddr_is_ok(const struct sockaddr_mctp *addr)
41 {
42 	return !addr->__smctp_pad0 && !addr->__smctp_pad1;
43 }
44 
mctp_sockaddr_ext_is_ok(const struct sockaddr_mctp_ext * addr)45 static bool mctp_sockaddr_ext_is_ok(const struct sockaddr_mctp_ext *addr)
46 {
47 	return !addr->__smctp_pad0[0] &&
48 	       !addr->__smctp_pad0[1] &&
49 	       !addr->__smctp_pad0[2];
50 }
51 
mctp_bind(struct socket * sock,struct sockaddr * addr,int addrlen)52 static int mctp_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
53 {
54 	struct sock *sk = sock->sk;
55 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
56 	struct net *net = sock_net(&msk->sk);
57 	struct sockaddr_mctp *smctp;
58 	int rc;
59 
60 	if (addrlen < sizeof(*smctp))
61 		return -EINVAL;
62 
63 	if (addr->sa_family != AF_MCTP)
64 		return -EAFNOSUPPORT;
65 
66 	if (!capable(CAP_NET_BIND_SERVICE))
67 		return -EACCES;
68 
69 	/* it's a valid sockaddr for MCTP, cast and do protocol checks */
70 	smctp = (struct sockaddr_mctp *)addr;
71 
72 	if (!mctp_sockaddr_is_ok(smctp))
73 		return -EINVAL;
74 
75 	lock_sock(sk);
76 
77 	if (sk_hashed(sk)) {
78 		rc = -EADDRINUSE;
79 		goto out_release;
80 	}
81 
82 	msk->bind_local_addr = smctp->smctp_addr.s_addr;
83 
84 	/* MCTP_NET_ANY with a specific EID is resolved to the default net
85 	 * at bind() time.
86 	 * For bind_addr=MCTP_ADDR_ANY it is handled specially at route
87 	 * lookup time.
88 	 */
89 	if (smctp->smctp_network == MCTP_NET_ANY &&
90 	    msk->bind_local_addr != MCTP_ADDR_ANY) {
91 		msk->bind_net = mctp_default_net(net);
92 	} else {
93 		msk->bind_net = smctp->smctp_network;
94 	}
95 
96 	/* ignore the IC bit */
97 	smctp->smctp_type &= 0x7f;
98 
99 	if (msk->bind_peer_set) {
100 		if (msk->bind_type != smctp->smctp_type) {
101 			/* Prior connect() had a different type */
102 			rc = -EINVAL;
103 			goto out_release;
104 		}
105 
106 		if (msk->bind_net == MCTP_NET_ANY) {
107 			/* Restrict to the network passed to connect() */
108 			msk->bind_net = msk->bind_peer_net;
109 		}
110 
111 		if (msk->bind_net != msk->bind_peer_net) {
112 			/* connect() had a different net to bind() */
113 			rc = -EINVAL;
114 			goto out_release;
115 		}
116 	} else {
117 		msk->bind_type = smctp->smctp_type;
118 	}
119 
120 	rc = sk->sk_prot->hash(sk);
121 
122 out_release:
123 	release_sock(sk);
124 
125 	return rc;
126 }
127 
128 /* Used to set a specific peer prior to bind. Not used for outbound
129  * connections (Tag Owner set) since MCTP is a datagram protocol.
130  */
mctp_connect(struct socket * sock,struct sockaddr * addr,int addrlen,int flags)131 static int mctp_connect(struct socket *sock, struct sockaddr *addr,
132 			int addrlen, int flags)
133 {
134 	struct sock *sk = sock->sk;
135 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
136 	struct net *net = sock_net(&msk->sk);
137 	struct sockaddr_mctp *smctp;
138 	int rc;
139 
140 	if (addrlen != sizeof(*smctp))
141 		return -EINVAL;
142 
143 	if (addr->sa_family != AF_MCTP)
144 		return -EAFNOSUPPORT;
145 
146 	/* It's a valid sockaddr for MCTP, cast and do protocol checks */
147 	smctp = (struct sockaddr_mctp *)addr;
148 
149 	if (!mctp_sockaddr_is_ok(smctp))
150 		return -EINVAL;
151 
152 	/* Can't bind by tag */
153 	if (smctp->smctp_tag)
154 		return -EINVAL;
155 
156 	/* IC bit must be unset */
157 	if (smctp->smctp_type & 0x80)
158 		return -EINVAL;
159 
160 	lock_sock(sk);
161 
162 	if (sk_hashed(sk)) {
163 		/* bind() already */
164 		rc = -EADDRINUSE;
165 		goto out_release;
166 	}
167 
168 	if (msk->bind_peer_set) {
169 		/* connect() already */
170 		rc = -EADDRINUSE;
171 		goto out_release;
172 	}
173 
174 	msk->bind_peer_set = true;
175 	msk->bind_peer_addr = smctp->smctp_addr.s_addr;
176 	msk->bind_type = smctp->smctp_type;
177 	if (smctp->smctp_network == MCTP_NET_ANY)
178 		msk->bind_peer_net = mctp_default_net(net);
179 	else
180 		msk->bind_peer_net = smctp->smctp_network;
181 
182 	rc = 0;
183 
184 out_release:
185 	release_sock(sk);
186 	return rc;
187 }
188 
mctp_sendmsg(struct socket * sock,struct msghdr * msg,size_t len)189 static int mctp_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
190 {
191 	DECLARE_SOCKADDR(struct sockaddr_mctp *, addr, msg->msg_name);
192 	int rc, addrlen = msg->msg_namelen;
193 	struct sock *sk = sock->sk;
194 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
195 	struct mctp_skb_cb *cb;
196 	struct sk_buff *skb = NULL;
197 	struct mctp_dst dst;
198 	int hlen;
199 
200 	if (addr) {
201 		const u8 tagbits = MCTP_TAG_MASK | MCTP_TAG_OWNER |
202 			MCTP_TAG_PREALLOC;
203 
204 		if (addrlen < sizeof(struct sockaddr_mctp))
205 			return -EINVAL;
206 		if (addr->smctp_family != AF_MCTP)
207 			return -EINVAL;
208 		if (!mctp_sockaddr_is_ok(addr))
209 			return -EINVAL;
210 		if (addr->smctp_tag & ~tagbits)
211 			return -EINVAL;
212 		/* can't preallocate a non-owned tag */
213 		if (addr->smctp_tag & MCTP_TAG_PREALLOC &&
214 		    !(addr->smctp_tag & MCTP_TAG_OWNER))
215 			return -EINVAL;
216 
217 	} else {
218 		/* TODO: connect()ed sockets */
219 		return -EDESTADDRREQ;
220 	}
221 
222 	if (!capable(CAP_NET_RAW))
223 		return -EACCES;
224 
225 	if (addr->smctp_network == MCTP_NET_ANY)
226 		addr->smctp_network = mctp_default_net(sock_net(sk));
227 
228 	/* direct addressing */
229 	if (msk->addr_ext && addrlen >= sizeof(struct sockaddr_mctp_ext)) {
230 		DECLARE_SOCKADDR(struct sockaddr_mctp_ext *,
231 				 extaddr, msg->msg_name);
232 
233 		if (!mctp_sockaddr_ext_is_ok(extaddr))
234 			return -EINVAL;
235 
236 		rc = mctp_dst_from_extaddr(&dst, sock_net(sk),
237 					   extaddr->smctp_ifindex,
238 					   extaddr->smctp_halen,
239 					   extaddr->smctp_haddr);
240 		if (rc)
241 			return rc;
242 
243 	} else {
244 		rc = mctp_route_lookup(sock_net(sk), addr->smctp_network,
245 				       addr->smctp_addr.s_addr, &dst);
246 		if (rc)
247 			return rc;
248 	}
249 
250 	hlen = LL_RESERVED_SPACE(dst.dev->dev) + sizeof(struct mctp_hdr);
251 
252 	skb = sock_alloc_send_skb(sk, hlen + 1 + len,
253 				  msg->msg_flags & MSG_DONTWAIT, &rc);
254 	if (!skb)
255 		goto err_release_dst;
256 
257 	skb_reserve(skb, hlen);
258 
259 	/* set type as fist byte in payload */
260 	*(u8 *)skb_put(skb, 1) = addr->smctp_type;
261 
262 	rc = memcpy_from_msg((void *)skb_put(skb, len), msg, len);
263 	if (rc < 0)
264 		goto err_free;
265 
266 	/* set up cb */
267 	cb = __mctp_cb(skb);
268 	cb->net = addr->smctp_network;
269 
270 	rc = mctp_local_output(sk, &dst, skb, addr->smctp_addr.s_addr,
271 			       addr->smctp_tag);
272 
273 	mctp_dst_release(&dst);
274 	return rc ? : len;
275 
276 err_free:
277 	kfree_skb(skb);
278 err_release_dst:
279 	mctp_dst_release(&dst);
280 	return rc;
281 }
282 
mctp_recvmsg(struct socket * sock,struct msghdr * msg,size_t len,int flags)283 static int mctp_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
284 			int flags)
285 {
286 	DECLARE_SOCKADDR(struct sockaddr_mctp *, addr, msg->msg_name);
287 	struct sock *sk = sock->sk;
288 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
289 	struct sk_buff *skb;
290 	size_t msglen;
291 	u8 type;
292 	int rc;
293 
294 	if (flags & ~(MSG_DONTWAIT | MSG_TRUNC | MSG_PEEK))
295 		return -EOPNOTSUPP;
296 
297 	skb = skb_recv_datagram(sk, flags, &rc);
298 	if (!skb)
299 		return rc;
300 
301 	if (!skb->len) {
302 		rc = 0;
303 		goto out_free;
304 	}
305 
306 	/* extract message type, remove from data */
307 	type = *((u8 *)skb->data);
308 	msglen = skb->len - 1;
309 
310 	if (len < msglen)
311 		msg->msg_flags |= MSG_TRUNC;
312 	else
313 		len = msglen;
314 
315 	rc = skb_copy_datagram_msg(skb, 1, msg, len);
316 	if (rc < 0)
317 		goto out_free;
318 
319 	sock_recv_cmsgs(msg, sk, skb);
320 
321 	if (addr) {
322 		struct mctp_skb_cb *cb = mctp_cb(skb);
323 		/* TODO: expand mctp_skb_cb for header fields? */
324 		struct mctp_hdr *hdr = mctp_hdr(skb);
325 
326 		addr = msg->msg_name;
327 		addr->smctp_family = AF_MCTP;
328 		addr->__smctp_pad0 = 0;
329 		addr->smctp_network = cb->net;
330 		addr->smctp_addr.s_addr = hdr->src;
331 		addr->smctp_type = type;
332 		addr->smctp_tag = hdr->flags_seq_tag &
333 					(MCTP_HDR_TAG_MASK | MCTP_HDR_FLAG_TO);
334 		addr->__smctp_pad1 = 0;
335 		msg->msg_namelen = sizeof(*addr);
336 
337 		if (msk->addr_ext) {
338 			DECLARE_SOCKADDR(struct sockaddr_mctp_ext *, ae,
339 					 msg->msg_name);
340 			msg->msg_namelen = sizeof(*ae);
341 			ae->smctp_ifindex = cb->ifindex;
342 			ae->smctp_halen = cb->halen;
343 			memset(ae->__smctp_pad0, 0x0, sizeof(ae->__smctp_pad0));
344 			memset(ae->smctp_haddr, 0x0, sizeof(ae->smctp_haddr));
345 			memcpy(ae->smctp_haddr, cb->haddr, cb->halen);
346 		}
347 	}
348 
349 	rc = len;
350 
351 	if (flags & MSG_TRUNC)
352 		rc = msglen;
353 
354 out_free:
355 	skb_free_datagram(sk, skb);
356 	return rc;
357 }
358 
359 /* We're done with the key; invalidate, stop reassembly, and remove from lists.
360  */
__mctp_key_remove(struct mctp_sk_key * key,struct net * net,unsigned long flags,unsigned long reason)361 static void __mctp_key_remove(struct mctp_sk_key *key, struct net *net,
362 			      unsigned long flags, unsigned long reason)
363 __releases(&key->lock)
364 __must_hold(&net->mctp.keys_lock)
365 {
366 	struct sk_buff *skb;
367 
368 	trace_mctp_key_release(key, reason);
369 	skb = key->reasm_head;
370 	key->reasm_head = NULL;
371 	key->reasm_dead = true;
372 	key->valid = false;
373 	mctp_dev_release_key(key->dev, key);
374 	spin_unlock_irqrestore(&key->lock, flags);
375 
376 	if (!hlist_unhashed(&key->hlist)) {
377 		hlist_del_init(&key->hlist);
378 		hlist_del_init(&key->sklist);
379 		/* unref for the lists */
380 		mctp_key_unref(key);
381 	}
382 
383 	kfree_skb(skb);
384 }
385 
mctp_setsockopt(struct socket * sock,int level,int optname,sockptr_t optval,unsigned int optlen)386 static int mctp_setsockopt(struct socket *sock, int level, int optname,
387 			   sockptr_t optval, unsigned int optlen)
388 {
389 	struct mctp_sock *msk = container_of(sock->sk, struct mctp_sock, sk);
390 	int val;
391 
392 	if (level != SOL_MCTP)
393 		return -EINVAL;
394 
395 	if (optname == MCTP_OPT_ADDR_EXT) {
396 		if (optlen != sizeof(int))
397 			return -EINVAL;
398 		if (copy_from_sockptr(&val, optval, sizeof(int)))
399 			return -EFAULT;
400 		msk->addr_ext = val;
401 		return 0;
402 	}
403 
404 	return -ENOPROTOOPT;
405 }
406 
mctp_getsockopt(struct socket * sock,int level,int optname,char __user * optval,int __user * optlen)407 static int mctp_getsockopt(struct socket *sock, int level, int optname,
408 			   char __user *optval, int __user *optlen)
409 {
410 	struct mctp_sock *msk = container_of(sock->sk, struct mctp_sock, sk);
411 	int len, val;
412 
413 	if (level != SOL_MCTP)
414 		return -EINVAL;
415 
416 	if (get_user(len, optlen))
417 		return -EFAULT;
418 
419 	if (optname == MCTP_OPT_ADDR_EXT) {
420 		if (len != sizeof(int))
421 			return -EINVAL;
422 		val = !!msk->addr_ext;
423 		if (copy_to_user(optval, &val, len))
424 			return -EFAULT;
425 		return 0;
426 	}
427 
428 	return -ENOPROTOOPT;
429 }
430 
431 /* helpers for reading/writing the tag ioc, handling compatibility across the
432  * two versions, and some basic API error checking
433  */
mctp_ioctl_tag_copy_from_user(unsigned long arg,struct mctp_ioc_tag_ctl2 * ctl,bool tagv2)434 static int mctp_ioctl_tag_copy_from_user(unsigned long arg,
435 					 struct mctp_ioc_tag_ctl2 *ctl,
436 					 bool tagv2)
437 {
438 	struct mctp_ioc_tag_ctl ctl_compat;
439 	unsigned long size;
440 	void *ptr;
441 	int rc;
442 
443 	if (tagv2) {
444 		size = sizeof(*ctl);
445 		ptr = ctl;
446 	} else {
447 		size = sizeof(ctl_compat);
448 		ptr = &ctl_compat;
449 	}
450 
451 	rc = copy_from_user(ptr, (void __user *)arg, size);
452 	if (rc)
453 		return -EFAULT;
454 
455 	if (!tagv2) {
456 		/* compat, using defaults for new fields */
457 		ctl->net = MCTP_INITIAL_DEFAULT_NET;
458 		ctl->peer_addr = ctl_compat.peer_addr;
459 		ctl->local_addr = MCTP_ADDR_ANY;
460 		ctl->flags = ctl_compat.flags;
461 		ctl->tag = ctl_compat.tag;
462 	}
463 
464 	if (ctl->flags)
465 		return -EINVAL;
466 
467 	if (ctl->local_addr != MCTP_ADDR_ANY &&
468 	    ctl->local_addr != MCTP_ADDR_NULL)
469 		return -EINVAL;
470 
471 	return 0;
472 }
473 
mctp_ioctl_tag_copy_to_user(unsigned long arg,struct mctp_ioc_tag_ctl2 * ctl,bool tagv2)474 static int mctp_ioctl_tag_copy_to_user(unsigned long arg,
475 				       struct mctp_ioc_tag_ctl2 *ctl,
476 				       bool tagv2)
477 {
478 	struct mctp_ioc_tag_ctl ctl_compat;
479 	unsigned long size;
480 	void *ptr;
481 	int rc;
482 
483 	if (tagv2) {
484 		ptr = ctl;
485 		size = sizeof(*ctl);
486 	} else {
487 		ctl_compat.peer_addr = ctl->peer_addr;
488 		ctl_compat.tag = ctl->tag;
489 		ctl_compat.flags = ctl->flags;
490 
491 		ptr = &ctl_compat;
492 		size = sizeof(ctl_compat);
493 	}
494 
495 	rc = copy_to_user((void __user *)arg, ptr, size);
496 	if (rc)
497 		return -EFAULT;
498 
499 	return 0;
500 }
501 
mctp_ioctl_alloctag(struct mctp_sock * msk,bool tagv2,unsigned long arg)502 static int mctp_ioctl_alloctag(struct mctp_sock *msk, bool tagv2,
503 			       unsigned long arg)
504 {
505 	struct net *net = sock_net(&msk->sk);
506 	struct mctp_sk_key *key = NULL;
507 	struct mctp_ioc_tag_ctl2 ctl;
508 	unsigned long flags;
509 	u8 tag;
510 	int rc;
511 
512 	rc = mctp_ioctl_tag_copy_from_user(arg, &ctl, tagv2);
513 	if (rc)
514 		return rc;
515 
516 	if (ctl.tag)
517 		return -EINVAL;
518 
519 	key = mctp_alloc_local_tag(msk, ctl.net, MCTP_ADDR_ANY,
520 				   ctl.peer_addr, true, &tag);
521 	if (IS_ERR(key))
522 		return PTR_ERR(key);
523 
524 	ctl.tag = tag | MCTP_TAG_OWNER | MCTP_TAG_PREALLOC;
525 	rc = mctp_ioctl_tag_copy_to_user(arg, &ctl, tagv2);
526 	if (rc) {
527 		unsigned long fl2;
528 		/* Unwind our key allocation: the keys list lock needs to be
529 		 * taken before the individual key locks, and we need a valid
530 		 * flags value (fl2) to pass to __mctp_key_remove, hence the
531 		 * second spin_lock_irqsave() rather than a plain spin_lock().
532 		 */
533 		spin_lock_irqsave(&net->mctp.keys_lock, flags);
534 		spin_lock_irqsave(&key->lock, fl2);
535 		__mctp_key_remove(key, net, fl2, MCTP_TRACE_KEY_DROPPED);
536 		mctp_key_unref(key);
537 		spin_unlock_irqrestore(&net->mctp.keys_lock, flags);
538 		return rc;
539 	}
540 
541 	mctp_key_unref(key);
542 	return 0;
543 }
544 
mctp_ioctl_droptag(struct mctp_sock * msk,bool tagv2,unsigned long arg)545 static int mctp_ioctl_droptag(struct mctp_sock *msk, bool tagv2,
546 			      unsigned long arg)
547 {
548 	struct net *net = sock_net(&msk->sk);
549 	struct mctp_ioc_tag_ctl2 ctl;
550 	unsigned long flags, fl2;
551 	struct mctp_sk_key *key;
552 	struct hlist_node *tmp;
553 	int rc;
554 	u8 tag;
555 
556 	rc = mctp_ioctl_tag_copy_from_user(arg, &ctl, tagv2);
557 	if (rc)
558 		return rc;
559 
560 	/* Must be a local tag, TO set, preallocated */
561 	if ((ctl.tag & ~MCTP_TAG_MASK) != (MCTP_TAG_OWNER | MCTP_TAG_PREALLOC))
562 		return -EINVAL;
563 
564 	tag = ctl.tag & MCTP_TAG_MASK;
565 	rc = -EINVAL;
566 
567 	if (ctl.peer_addr == MCTP_ADDR_NULL)
568 		ctl.peer_addr = MCTP_ADDR_ANY;
569 
570 	spin_lock_irqsave(&net->mctp.keys_lock, flags);
571 	hlist_for_each_entry_safe(key, tmp, &msk->keys, sklist) {
572 		/* we do an irqsave here, even though we know the irq state,
573 		 * so we have the flags to pass to __mctp_key_remove
574 		 */
575 		spin_lock_irqsave(&key->lock, fl2);
576 		if (key->manual_alloc &&
577 		    ctl.net == key->net &&
578 		    ctl.peer_addr == key->peer_addr &&
579 		    tag == key->tag) {
580 			__mctp_key_remove(key, net, fl2,
581 					  MCTP_TRACE_KEY_DROPPED);
582 			rc = 0;
583 		} else {
584 			spin_unlock_irqrestore(&key->lock, fl2);
585 		}
586 	}
587 	spin_unlock_irqrestore(&net->mctp.keys_lock, flags);
588 
589 	return rc;
590 }
591 
mctp_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)592 static int mctp_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
593 {
594 	struct mctp_sock *msk = container_of(sock->sk, struct mctp_sock, sk);
595 	bool tagv2 = false;
596 
597 	switch (cmd) {
598 	case SIOCMCTPALLOCTAG2:
599 	case SIOCMCTPALLOCTAG:
600 		tagv2 = cmd == SIOCMCTPALLOCTAG2;
601 		return mctp_ioctl_alloctag(msk, tagv2, arg);
602 	case SIOCMCTPDROPTAG:
603 	case SIOCMCTPDROPTAG2:
604 		tagv2 = cmd == SIOCMCTPDROPTAG2;
605 		return mctp_ioctl_droptag(msk, tagv2, arg);
606 	}
607 
608 	return -EINVAL;
609 }
610 
611 #ifdef CONFIG_COMPAT
mctp_compat_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)612 static int mctp_compat_ioctl(struct socket *sock, unsigned int cmd,
613 			     unsigned long arg)
614 {
615 	void __user *argp = compat_ptr(arg);
616 
617 	switch (cmd) {
618 	/* These have compatible ptr layouts */
619 	case SIOCMCTPALLOCTAG:
620 	case SIOCMCTPDROPTAG:
621 		return mctp_ioctl(sock, cmd, (unsigned long)argp);
622 	}
623 
624 	return -ENOIOCTLCMD;
625 }
626 #endif
627 
628 static const struct proto_ops mctp_dgram_ops = {
629 	.family		= PF_MCTP,
630 	.release	= mctp_release,
631 	.bind		= mctp_bind,
632 	.connect	= mctp_connect,
633 	.socketpair	= sock_no_socketpair,
634 	.accept		= sock_no_accept,
635 	.getname	= sock_no_getname,
636 	.poll		= datagram_poll,
637 	.ioctl		= mctp_ioctl,
638 	.gettstamp	= sock_gettstamp,
639 	.listen		= sock_no_listen,
640 	.shutdown	= sock_no_shutdown,
641 	.setsockopt	= mctp_setsockopt,
642 	.getsockopt	= mctp_getsockopt,
643 	.sendmsg	= mctp_sendmsg,
644 	.recvmsg	= mctp_recvmsg,
645 	.mmap		= sock_no_mmap,
646 #ifdef CONFIG_COMPAT
647 	.compat_ioctl	= mctp_compat_ioctl,
648 #endif
649 };
650 
mctp_sk_expire_keys(struct timer_list * timer)651 static void mctp_sk_expire_keys(struct timer_list *timer)
652 {
653 	struct mctp_sock *msk = container_of(timer, struct mctp_sock,
654 					     key_expiry);
655 	struct net *net = sock_net(&msk->sk);
656 	unsigned long next_expiry, flags, fl2;
657 	struct mctp_sk_key *key;
658 	struct hlist_node *tmp;
659 	bool next_expiry_valid = false;
660 
661 	spin_lock_irqsave(&net->mctp.keys_lock, flags);
662 
663 	hlist_for_each_entry_safe(key, tmp, &msk->keys, sklist) {
664 		/* don't expire. manual_alloc is immutable, no locking
665 		 * required.
666 		 */
667 		if (key->manual_alloc)
668 			continue;
669 
670 		spin_lock_irqsave(&key->lock, fl2);
671 		if (!time_after_eq(key->expiry, jiffies)) {
672 			__mctp_key_remove(key, net, fl2,
673 					  MCTP_TRACE_KEY_TIMEOUT);
674 			continue;
675 		}
676 
677 		if (next_expiry_valid) {
678 			if (time_before(key->expiry, next_expiry))
679 				next_expiry = key->expiry;
680 		} else {
681 			next_expiry = key->expiry;
682 			next_expiry_valid = true;
683 		}
684 		spin_unlock_irqrestore(&key->lock, fl2);
685 	}
686 
687 	spin_unlock_irqrestore(&net->mctp.keys_lock, flags);
688 
689 	if (next_expiry_valid)
690 		mod_timer(timer, next_expiry);
691 }
692 
mctp_sk_init(struct sock * sk)693 static int mctp_sk_init(struct sock *sk)
694 {
695 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
696 
697 	INIT_HLIST_HEAD(&msk->keys);
698 	timer_setup(&msk->key_expiry, mctp_sk_expire_keys, 0);
699 	msk->bind_peer_set = false;
700 	return 0;
701 }
702 
mctp_sk_close(struct sock * sk,long timeout)703 static void mctp_sk_close(struct sock *sk, long timeout)
704 {
705 	sk_common_release(sk);
706 }
707 
mctp_sk_hash(struct sock * sk)708 static int mctp_sk_hash(struct sock *sk)
709 {
710 	struct net *net = sock_net(sk);
711 	struct sock *existing;
712 	struct mctp_sock *msk;
713 	mctp_eid_t remote;
714 	u32 hash;
715 	int rc;
716 
717 	msk = container_of(sk, struct mctp_sock, sk);
718 
719 	if (msk->bind_peer_set)
720 		remote = msk->bind_peer_addr;
721 	else
722 		remote = MCTP_ADDR_ANY;
723 	hash = mctp_bind_hash(msk->bind_type, msk->bind_local_addr, remote);
724 
725 	mutex_lock(&net->mctp.bind_lock);
726 
727 	/* Prevent duplicate binds. */
728 	sk_for_each(existing, &net->mctp.binds[hash]) {
729 		struct mctp_sock *mex =
730 			container_of(existing, struct mctp_sock, sk);
731 
732 		bool same_peer = (mex->bind_peer_set && msk->bind_peer_set &&
733 				  mex->bind_peer_addr == msk->bind_peer_addr) ||
734 				 (!mex->bind_peer_set && !msk->bind_peer_set);
735 
736 		if (mex->bind_type == msk->bind_type &&
737 		    mex->bind_local_addr == msk->bind_local_addr && same_peer &&
738 		    mex->bind_net == msk->bind_net) {
739 			rc = -EADDRINUSE;
740 			goto out;
741 		}
742 	}
743 
744 	/* Bind lookup runs under RCU, remain live during that. */
745 	sock_set_flag(sk, SOCK_RCU_FREE);
746 
747 	sk_add_node_rcu(sk, &net->mctp.binds[hash]);
748 	rc = 0;
749 
750 out:
751 	mutex_unlock(&net->mctp.bind_lock);
752 	return rc;
753 }
754 
mctp_sk_unhash(struct sock * sk)755 static void mctp_sk_unhash(struct sock *sk)
756 {
757 	struct mctp_sock *msk = container_of(sk, struct mctp_sock, sk);
758 	struct net *net = sock_net(sk);
759 	unsigned long flags, fl2;
760 	struct mctp_sk_key *key;
761 	struct hlist_node *tmp;
762 
763 	/* remove from any type-based binds */
764 	mutex_lock(&net->mctp.bind_lock);
765 	sk_del_node_init_rcu(sk);
766 	mutex_unlock(&net->mctp.bind_lock);
767 
768 	/* remove tag allocations */
769 	spin_lock_irqsave(&net->mctp.keys_lock, flags);
770 	hlist_for_each_entry_safe(key, tmp, &msk->keys, sklist) {
771 		spin_lock_irqsave(&key->lock, fl2);
772 		__mctp_key_remove(key, net, fl2, MCTP_TRACE_KEY_CLOSED);
773 	}
774 	sock_set_flag(sk, SOCK_DEAD);
775 	spin_unlock_irqrestore(&net->mctp.keys_lock, flags);
776 
777 	/* Since there are no more tag allocations (we have removed all of the
778 	 * keys), stop any pending expiry events. the timer cannot be re-queued
779 	 * as the sk is no longer observable
780 	 */
781 	timer_delete_sync(&msk->key_expiry);
782 }
783 
mctp_sk_destruct(struct sock * sk)784 static void mctp_sk_destruct(struct sock *sk)
785 {
786 	skb_queue_purge(&sk->sk_receive_queue);
787 }
788 
789 static struct proto mctp_proto = {
790 	.name		= "MCTP",
791 	.owner		= THIS_MODULE,
792 	.obj_size	= sizeof(struct mctp_sock),
793 	.init		= mctp_sk_init,
794 	.close		= mctp_sk_close,
795 	.hash		= mctp_sk_hash,
796 	.unhash		= mctp_sk_unhash,
797 };
798 
mctp_pf_create(struct net * net,struct socket * sock,int protocol,int kern)799 static int mctp_pf_create(struct net *net, struct socket *sock,
800 			  int protocol, int kern)
801 {
802 	const struct proto_ops *ops;
803 	struct proto *proto;
804 	struct sock *sk;
805 	int rc;
806 
807 	if (protocol)
808 		return -EPROTONOSUPPORT;
809 
810 	/* only datagram sockets are supported */
811 	if (sock->type != SOCK_DGRAM)
812 		return -ESOCKTNOSUPPORT;
813 
814 	proto = &mctp_proto;
815 	ops = &mctp_dgram_ops;
816 
817 	sock->state = SS_UNCONNECTED;
818 	sock->ops = ops;
819 
820 	sk = sk_alloc(net, PF_MCTP, GFP_KERNEL, proto, kern);
821 	if (!sk)
822 		return -ENOMEM;
823 
824 	sock_init_data(sock, sk);
825 	sk->sk_destruct = mctp_sk_destruct;
826 
827 	rc = 0;
828 	if (sk->sk_prot->init)
829 		rc = sk->sk_prot->init(sk);
830 
831 	if (rc)
832 		goto err_sk_put;
833 
834 	return 0;
835 
836 err_sk_put:
837 	sock_orphan(sk);
838 	sock_put(sk);
839 	return rc;
840 }
841 
842 static struct net_proto_family mctp_pf = {
843 	.family = PF_MCTP,
844 	.create = mctp_pf_create,
845 	.owner = THIS_MODULE,
846 };
847 
mctp_init(void)848 static __init int mctp_init(void)
849 {
850 	int rc;
851 
852 	/* ensure our uapi tag definitions match the header format */
853 	BUILD_BUG_ON(MCTP_TAG_OWNER != MCTP_HDR_FLAG_TO);
854 	BUILD_BUG_ON(MCTP_TAG_MASK != MCTP_HDR_TAG_MASK);
855 
856 	pr_info("mctp: management component transport protocol core\n");
857 
858 	rc = sock_register(&mctp_pf);
859 	if (rc)
860 		return rc;
861 
862 	rc = proto_register(&mctp_proto, 0);
863 	if (rc)
864 		goto err_unreg_sock;
865 
866 	rc = mctp_routes_init();
867 	if (rc)
868 		goto err_unreg_proto;
869 
870 	rc = mctp_neigh_init();
871 	if (rc)
872 		goto err_unreg_routes;
873 
874 	rc = mctp_device_init();
875 	if (rc)
876 		goto err_unreg_neigh;
877 
878 	return 0;
879 
880 err_unreg_neigh:
881 	mctp_neigh_exit();
882 err_unreg_routes:
883 	mctp_routes_exit();
884 err_unreg_proto:
885 	proto_unregister(&mctp_proto);
886 err_unreg_sock:
887 	sock_unregister(PF_MCTP);
888 
889 	return rc;
890 }
891 
mctp_exit(void)892 static __exit void mctp_exit(void)
893 {
894 	mctp_device_exit();
895 	mctp_neigh_exit();
896 	mctp_routes_exit();
897 	proto_unregister(&mctp_proto);
898 	sock_unregister(PF_MCTP);
899 }
900 
901 subsys_initcall(mctp_init);
902 module_exit(mctp_exit);
903 
904 MODULE_DESCRIPTION("MCTP core");
905 MODULE_AUTHOR("Jeremy Kerr <jk@codeconstruct.com.au>");
906 
907 MODULE_ALIAS_NETPROTO(PF_MCTP);
908 
909 #if IS_ENABLED(CONFIG_MCTP_TEST)
910 #include "test/sock-test.c"
911 #endif
912