• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Local endpoint object management
3  *
4  * Copyright (C) 2016 Red Hat, Inc. All Rights Reserved.
5  * Written by David Howells (dhowells@redhat.com)
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/module.h>
11 #include <linux/net.h>
12 #include <linux/skbuff.h>
13 #include <linux/slab.h>
14 #include <linux/udp.h>
15 #include <linux/ip.h>
16 #include <linux/hashtable.h>
17 #include <net/sock.h>
18 #include <net/udp.h>
19 #include <net/af_rxrpc.h>
20 #include "ar-internal.h"
21 
22 static void rxrpc_local_processor(struct work_struct *);
23 static void rxrpc_local_rcu(struct rcu_head *);
24 
25 /*
26  * Compare a local to an address.  Return -ve, 0 or +ve to indicate less than,
27  * same or greater than.
28  *
29  * We explicitly don't compare the RxRPC service ID as we want to reject
30  * conflicting uses by differing services.  Further, we don't want to share
31  * addresses with different options (IPv6), so we don't compare those bits
32  * either.
33  */
rxrpc_local_cmp_key(const struct rxrpc_local * local,const struct sockaddr_rxrpc * srx)34 static long rxrpc_local_cmp_key(const struct rxrpc_local *local,
35 				const struct sockaddr_rxrpc *srx)
36 {
37 	long diff;
38 
39 	diff = ((local->srx.transport_type - srx->transport_type) ?:
40 		(local->srx.transport_len - srx->transport_len) ?:
41 		(local->srx.transport.family - srx->transport.family));
42 	if (diff != 0)
43 		return diff;
44 
45 	switch (srx->transport.family) {
46 	case AF_INET:
47 		/* If the choice of UDP port is left up to the transport, then
48 		 * the endpoint record doesn't match.
49 		 */
50 		return ((u16 __force)local->srx.transport.sin.sin_port -
51 			(u16 __force)srx->transport.sin.sin_port) ?:
52 			memcmp(&local->srx.transport.sin.sin_addr,
53 			       &srx->transport.sin.sin_addr,
54 			       sizeof(struct in_addr));
55 #ifdef CONFIG_AF_RXRPC_IPV6
56 	case AF_INET6:
57 		/* If the choice of UDP6 port is left up to the transport, then
58 		 * the endpoint record doesn't match.
59 		 */
60 		return ((u16 __force)local->srx.transport.sin6.sin6_port -
61 			(u16 __force)srx->transport.sin6.sin6_port) ?:
62 			memcmp(&local->srx.transport.sin6.sin6_addr,
63 			       &srx->transport.sin6.sin6_addr,
64 			       sizeof(struct in6_addr));
65 #endif
66 	default:
67 		BUG();
68 	}
69 }
70 
71 /*
72  * Allocate a new local endpoint.
73  */
rxrpc_alloc_local(struct rxrpc_net * rxnet,const struct sockaddr_rxrpc * srx)74 static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
75 					     const struct sockaddr_rxrpc *srx)
76 {
77 	struct rxrpc_local *local;
78 
79 	local = kzalloc(sizeof(struct rxrpc_local), GFP_KERNEL);
80 	if (local) {
81 		atomic_set(&local->usage, 1);
82 		atomic_set(&local->active_users, 1);
83 		local->rxnet = rxnet;
84 		INIT_LIST_HEAD(&local->link);
85 		INIT_WORK(&local->processor, rxrpc_local_processor);
86 		init_rwsem(&local->defrag_sem);
87 		skb_queue_head_init(&local->reject_queue);
88 		skb_queue_head_init(&local->event_queue);
89 		local->client_conns = RB_ROOT;
90 		spin_lock_init(&local->client_conns_lock);
91 		spin_lock_init(&local->lock);
92 		rwlock_init(&local->services_lock);
93 		local->debug_id = atomic_inc_return(&rxrpc_debug_id);
94 		memcpy(&local->srx, srx, sizeof(*srx));
95 		local->srx.srx_service = 0;
96 		trace_rxrpc_local(local->debug_id, rxrpc_local_new, 1, NULL);
97 	}
98 
99 	_leave(" = %p", local);
100 	return local;
101 }
102 
103 /*
104  * create the local socket
105  * - must be called with rxrpc_local_mutex locked
106  */
rxrpc_open_socket(struct rxrpc_local * local,struct net * net)107 static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
108 {
109 	struct sock *usk;
110 	int ret, opt;
111 
112 	_enter("%p{%d,%d}",
113 	       local, local->srx.transport_type, local->srx.transport.family);
114 
115 	/* create a socket to represent the local endpoint */
116 	ret = sock_create_kern(net, local->srx.transport.family,
117 			       local->srx.transport_type, 0, &local->socket);
118 	if (ret < 0) {
119 		_leave(" = %d [socket]", ret);
120 		return ret;
121 	}
122 
123 	/* set the socket up */
124 	usk = local->socket->sk;
125 	inet_sk(usk)->mc_loop = 0;
126 
127 	/* Enable CHECKSUM_UNNECESSARY to CHECKSUM_COMPLETE conversion */
128 	inet_inc_convert_csum(usk);
129 
130 	rcu_assign_sk_user_data(usk, local);
131 
132 	udp_sk(usk)->encap_type = UDP_ENCAP_RXRPC;
133 	udp_sk(usk)->encap_rcv = rxrpc_input_packet;
134 	udp_sk(usk)->encap_destroy = NULL;
135 	udp_sk(usk)->gro_receive = NULL;
136 	udp_sk(usk)->gro_complete = NULL;
137 
138 	udp_encap_enable();
139 #if IS_ENABLED(CONFIG_AF_RXRPC_IPV6)
140 	if (local->srx.transport.family == AF_INET6)
141 		udpv6_encap_enable();
142 #endif
143 	usk->sk_error_report = rxrpc_error_report;
144 
145 	/* if a local address was supplied then bind it */
146 	if (local->srx.transport_len > sizeof(sa_family_t)) {
147 		_debug("bind");
148 		ret = kernel_bind(local->socket,
149 				  (struct sockaddr *)&local->srx.transport,
150 				  local->srx.transport_len);
151 		if (ret < 0) {
152 			_debug("bind failed %d", ret);
153 			goto error;
154 		}
155 	}
156 
157 	switch (local->srx.transport.family) {
158 	case AF_INET6:
159 		/* we want to receive ICMPv6 errors */
160 		opt = 1;
161 		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_RECVERR,
162 					(char *) &opt, sizeof(opt));
163 		if (ret < 0) {
164 			_debug("setsockopt failed");
165 			goto error;
166 		}
167 
168 		/* we want to set the don't fragment bit */
169 		opt = IPV6_PMTUDISC_DO;
170 		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_MTU_DISCOVER,
171 					(char *) &opt, sizeof(opt));
172 		if (ret < 0) {
173 			_debug("setsockopt failed");
174 			goto error;
175 		}
176 
177 		/* Fall through and set IPv4 options too otherwise we don't get
178 		 * errors from IPv4 packets sent through the IPv6 socket.
179 		 */
180 		/* Fall through */
181 	case AF_INET:
182 		/* we want to receive ICMP errors */
183 		opt = 1;
184 		ret = kernel_setsockopt(local->socket, SOL_IP, IP_RECVERR,
185 					(char *) &opt, sizeof(opt));
186 		if (ret < 0) {
187 			_debug("setsockopt failed");
188 			goto error;
189 		}
190 
191 		/* we want to set the don't fragment bit */
192 		opt = IP_PMTUDISC_DO;
193 		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
194 					(char *) &opt, sizeof(opt));
195 		if (ret < 0) {
196 			_debug("setsockopt failed");
197 			goto error;
198 		}
199 
200 		/* We want receive timestamps. */
201 		opt = 1;
202 		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
203 					(char *)&opt, sizeof(opt));
204 		if (ret < 0) {
205 			_debug("setsockopt failed");
206 			goto error;
207 		}
208 		break;
209 
210 	default:
211 		BUG();
212 	}
213 
214 	_leave(" = 0");
215 	return 0;
216 
217 error:
218 	kernel_sock_shutdown(local->socket, SHUT_RDWR);
219 	local->socket->sk->sk_user_data = NULL;
220 	sock_release(local->socket);
221 	local->socket = NULL;
222 
223 	_leave(" = %d", ret);
224 	return ret;
225 }
226 
227 /*
228  * Look up or create a new local endpoint using the specified local address.
229  */
rxrpc_lookup_local(struct net * net,const struct sockaddr_rxrpc * srx)230 struct rxrpc_local *rxrpc_lookup_local(struct net *net,
231 				       const struct sockaddr_rxrpc *srx)
232 {
233 	struct rxrpc_local *local;
234 	struct rxrpc_net *rxnet = rxrpc_net(net);
235 	struct list_head *cursor;
236 	const char *age;
237 	long diff;
238 	int ret;
239 
240 	_enter("{%d,%d,%pISp}",
241 	       srx->transport_type, srx->transport.family, &srx->transport);
242 
243 	mutex_lock(&rxnet->local_mutex);
244 
245 	for (cursor = rxnet->local_endpoints.next;
246 	     cursor != &rxnet->local_endpoints;
247 	     cursor = cursor->next) {
248 		local = list_entry(cursor, struct rxrpc_local, link);
249 
250 		diff = rxrpc_local_cmp_key(local, srx);
251 		if (diff < 0)
252 			continue;
253 		if (diff > 0)
254 			break;
255 
256 		/* Services aren't allowed to share transport sockets, so
257 		 * reject that here.  It is possible that the object is dying -
258 		 * but it may also still have the local transport address that
259 		 * we want bound.
260 		 */
261 		if (srx->srx_service) {
262 			local = NULL;
263 			goto addr_in_use;
264 		}
265 
266 		/* Found a match.  We replace a dying object.  Attempting to
267 		 * bind the transport socket may still fail if we're attempting
268 		 * to use a local address that the dying object is still using.
269 		 */
270 		if (!rxrpc_use_local(local))
271 			break;
272 
273 		age = "old";
274 		goto found;
275 	}
276 
277 	local = rxrpc_alloc_local(rxnet, srx);
278 	if (!local)
279 		goto nomem;
280 
281 	ret = rxrpc_open_socket(local, net);
282 	if (ret < 0)
283 		goto sock_error;
284 
285 	if (cursor != &rxnet->local_endpoints)
286 		list_replace_init(cursor, &local->link);
287 	else
288 		list_add_tail(&local->link, cursor);
289 	age = "new";
290 
291 found:
292 	mutex_unlock(&rxnet->local_mutex);
293 
294 	_net("LOCAL %s %d {%pISp}",
295 	     age, local->debug_id, &local->srx.transport);
296 
297 	_leave(" = %p", local);
298 	return local;
299 
300 nomem:
301 	ret = -ENOMEM;
302 sock_error:
303 	mutex_unlock(&rxnet->local_mutex);
304 	if (local)
305 		call_rcu(&local->rcu, rxrpc_local_rcu);
306 	_leave(" = %d", ret);
307 	return ERR_PTR(ret);
308 
309 addr_in_use:
310 	mutex_unlock(&rxnet->local_mutex);
311 	_leave(" = -EADDRINUSE");
312 	return ERR_PTR(-EADDRINUSE);
313 }
314 
315 /*
316  * Get a ref on a local endpoint.
317  */
rxrpc_get_local(struct rxrpc_local * local)318 struct rxrpc_local *rxrpc_get_local(struct rxrpc_local *local)
319 {
320 	const void *here = __builtin_return_address(0);
321 	int n;
322 
323 	n = atomic_inc_return(&local->usage);
324 	trace_rxrpc_local(local->debug_id, rxrpc_local_got, n, here);
325 	return local;
326 }
327 
328 /*
329  * Get a ref on a local endpoint unless its usage has already reached 0.
330  */
rxrpc_get_local_maybe(struct rxrpc_local * local)331 struct rxrpc_local *rxrpc_get_local_maybe(struct rxrpc_local *local)
332 {
333 	const void *here = __builtin_return_address(0);
334 
335 	if (local) {
336 		int n = atomic_fetch_add_unless(&local->usage, 1, 0);
337 		if (n > 0)
338 			trace_rxrpc_local(local->debug_id, rxrpc_local_got,
339 					  n + 1, here);
340 		else
341 			local = NULL;
342 	}
343 	return local;
344 }
345 
346 /*
347  * Queue a local endpoint and pass the caller's reference to the work item.
348  */
rxrpc_queue_local(struct rxrpc_local * local)349 void rxrpc_queue_local(struct rxrpc_local *local)
350 {
351 	const void *here = __builtin_return_address(0);
352 	unsigned int debug_id = local->debug_id;
353 	int n = atomic_read(&local->usage);
354 
355 	if (rxrpc_queue_work(&local->processor))
356 		trace_rxrpc_local(debug_id, rxrpc_local_queued, n, here);
357 	else
358 		rxrpc_put_local(local);
359 }
360 
361 /*
362  * Drop a ref on a local endpoint.
363  */
rxrpc_put_local(struct rxrpc_local * local)364 void rxrpc_put_local(struct rxrpc_local *local)
365 {
366 	const void *here = __builtin_return_address(0);
367 	int n;
368 
369 	if (local) {
370 		n = atomic_dec_return(&local->usage);
371 		trace_rxrpc_local(local->debug_id, rxrpc_local_put, n, here);
372 
373 		if (n == 0)
374 			call_rcu(&local->rcu, rxrpc_local_rcu);
375 	}
376 }
377 
378 /*
379  * Start using a local endpoint.
380  */
rxrpc_use_local(struct rxrpc_local * local)381 struct rxrpc_local *rxrpc_use_local(struct rxrpc_local *local)
382 {
383 	unsigned int au;
384 
385 	local = rxrpc_get_local_maybe(local);
386 	if (!local)
387 		return NULL;
388 
389 	au = atomic_fetch_add_unless(&local->active_users, 1, 0);
390 	if (au == 0) {
391 		rxrpc_put_local(local);
392 		return NULL;
393 	}
394 
395 	return local;
396 }
397 
398 /*
399  * Cease using a local endpoint.  Once the number of active users reaches 0, we
400  * start the closure of the transport in the work processor.
401  */
rxrpc_unuse_local(struct rxrpc_local * local)402 void rxrpc_unuse_local(struct rxrpc_local *local)
403 {
404 	unsigned int au;
405 
406 	if (local) {
407 		au = atomic_dec_return(&local->active_users);
408 		if (au == 0)
409 			rxrpc_queue_local(local);
410 		else
411 			rxrpc_put_local(local);
412 	}
413 }
414 
415 /*
416  * Destroy a local endpoint's socket and then hand the record to RCU to dispose
417  * of.
418  *
419  * Closing the socket cannot be done from bottom half context or RCU callback
420  * context because it might sleep.
421  */
rxrpc_local_destroyer(struct rxrpc_local * local)422 static void rxrpc_local_destroyer(struct rxrpc_local *local)
423 {
424 	struct socket *socket = local->socket;
425 	struct rxrpc_net *rxnet = local->rxnet;
426 
427 	_enter("%d", local->debug_id);
428 
429 	local->dead = true;
430 
431 	mutex_lock(&rxnet->local_mutex);
432 	list_del_init(&local->link);
433 	mutex_unlock(&rxnet->local_mutex);
434 
435 	rxrpc_clean_up_local_conns(local);
436 	rxrpc_service_connection_reaper(&rxnet->service_conn_reaper);
437 	ASSERT(!local->service);
438 
439 	if (socket) {
440 		local->socket = NULL;
441 		kernel_sock_shutdown(socket, SHUT_RDWR);
442 		socket->sk->sk_user_data = NULL;
443 		sock_release(socket);
444 	}
445 
446 	/* At this point, there should be no more packets coming in to the
447 	 * local endpoint.
448 	 */
449 	rxrpc_purge_queue(&local->reject_queue);
450 	rxrpc_purge_queue(&local->event_queue);
451 }
452 
453 /*
454  * Process events on an endpoint.  The work item carries a ref which
455  * we must release.
456  */
rxrpc_local_processor(struct work_struct * work)457 static void rxrpc_local_processor(struct work_struct *work)
458 {
459 	struct rxrpc_local *local =
460 		container_of(work, struct rxrpc_local, processor);
461 	bool again;
462 
463 	trace_rxrpc_local(local->debug_id, rxrpc_local_processing,
464 			  atomic_read(&local->usage), NULL);
465 
466 	do {
467 		again = false;
468 		if (atomic_read(&local->active_users) == 0) {
469 			rxrpc_local_destroyer(local);
470 			break;
471 		}
472 
473 		if (!skb_queue_empty(&local->reject_queue)) {
474 			rxrpc_reject_packets(local);
475 			again = true;
476 		}
477 
478 		if (!skb_queue_empty(&local->event_queue)) {
479 			rxrpc_process_local_events(local);
480 			again = true;
481 		}
482 	} while (again);
483 
484 	rxrpc_put_local(local);
485 }
486 
487 /*
488  * Destroy a local endpoint after the RCU grace period expires.
489  */
rxrpc_local_rcu(struct rcu_head * rcu)490 static void rxrpc_local_rcu(struct rcu_head *rcu)
491 {
492 	struct rxrpc_local *local = container_of(rcu, struct rxrpc_local, rcu);
493 
494 	_enter("%d", local->debug_id);
495 
496 	ASSERT(!work_pending(&local->processor));
497 
498 	_net("DESTROY LOCAL %d", local->debug_id);
499 	kfree(local);
500 	_leave("");
501 }
502 
503 /*
504  * Verify the local endpoint list is empty by this point.
505  */
rxrpc_destroy_all_locals(struct rxrpc_net * rxnet)506 void rxrpc_destroy_all_locals(struct rxrpc_net *rxnet)
507 {
508 	struct rxrpc_local *local;
509 
510 	_enter("");
511 
512 	flush_workqueue(rxrpc_workqueue);
513 
514 	if (!list_empty(&rxnet->local_endpoints)) {
515 		mutex_lock(&rxnet->local_mutex);
516 		list_for_each_entry(local, &rxnet->local_endpoints, link) {
517 			pr_err("AF_RXRPC: Leaked local %p {%d}\n",
518 			       local, atomic_read(&local->usage));
519 		}
520 		mutex_unlock(&rxnet->local_mutex);
521 		BUG();
522 	}
523 }
524