xref: /openbmc/linux/net/rds/connection.c (revision 63705da3)
1 /*
2  * Copyright (c) 2006, 2018 Oracle and/or its affiliates. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * OpenIB.org BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  *
32  */
33 #include <linux/kernel.h>
34 #include <linux/list.h>
35 #include <linux/slab.h>
36 #include <linux/export.h>
37 #include <net/ipv6.h>
38 #include <net/inet6_hashtables.h>
39 #include <net/addrconf.h>
40 
41 #include "rds.h"
42 #include "loop.h"
43 
44 #define RDS_CONNECTION_HASH_BITS 12
45 #define RDS_CONNECTION_HASH_ENTRIES (1 << RDS_CONNECTION_HASH_BITS)
46 #define RDS_CONNECTION_HASH_MASK (RDS_CONNECTION_HASH_ENTRIES - 1)
47 
48 /* converting this to RCU is a chore for another day.. */
49 static DEFINE_SPINLOCK(rds_conn_lock);
50 static unsigned long rds_conn_count;
51 static struct hlist_head rds_conn_hash[RDS_CONNECTION_HASH_ENTRIES];
52 static struct kmem_cache *rds_conn_slab;
53 
54 static struct hlist_head *rds_conn_bucket(const struct in6_addr *laddr,
55 					  const struct in6_addr *faddr)
56 {
57 	static u32 rds6_hash_secret __read_mostly;
58 	static u32 rds_hash_secret __read_mostly;
59 
60 	u32 lhash, fhash, hash;
61 
62 	net_get_random_once(&rds_hash_secret, sizeof(rds_hash_secret));
63 	net_get_random_once(&rds6_hash_secret, sizeof(rds6_hash_secret));
64 
65 	lhash = (__force u32)laddr->s6_addr32[3];
66 #if IS_ENABLED(CONFIG_IPV6)
67 	fhash = __ipv6_addr_jhash(faddr, rds6_hash_secret);
68 #else
69 	fhash = (__force u32)faddr->s6_addr32[3];
70 #endif
71 	hash = __inet_ehashfn(lhash, 0, fhash, 0, rds_hash_secret);
72 
73 	return &rds_conn_hash[hash & RDS_CONNECTION_HASH_MASK];
74 }
75 
76 #define rds_conn_info_set(var, test, suffix) do {		\
77 	if (test)						\
78 		var |= RDS_INFO_CONNECTION_FLAG_##suffix;	\
79 } while (0)
80 
81 /* rcu read lock must be held or the connection spinlock */
82 static struct rds_connection *rds_conn_lookup(struct net *net,
83 					      struct hlist_head *head,
84 					      const struct in6_addr *laddr,
85 					      const struct in6_addr *faddr,
86 					      struct rds_transport *trans,
87 					      u8 tos, int dev_if)
88 {
89 	struct rds_connection *conn, *ret = NULL;
90 
91 	hlist_for_each_entry_rcu(conn, head, c_hash_node) {
92 		if (ipv6_addr_equal(&conn->c_faddr, faddr) &&
93 		    ipv6_addr_equal(&conn->c_laddr, laddr) &&
94 		    conn->c_trans == trans &&
95 		    conn->c_tos == tos &&
96 		    net == rds_conn_net(conn) &&
97 		    conn->c_dev_if == dev_if) {
98 			ret = conn;
99 			break;
100 		}
101 	}
102 	rdsdebug("returning conn %p for %pI6c -> %pI6c\n", ret,
103 		 laddr, faddr);
104 	return ret;
105 }
106 
107 /*
108  * This is called by transports as they're bringing down a connection.
109  * It clears partial message state so that the transport can start sending
110  * and receiving over this connection again in the future.  It is up to
111  * the transport to have serialized this call with its send and recv.
112  */
113 static void rds_conn_path_reset(struct rds_conn_path *cp)
114 {
115 	struct rds_connection *conn = cp->cp_conn;
116 
117 	rdsdebug("connection %pI6c to %pI6c reset\n",
118 		 &conn->c_laddr, &conn->c_faddr);
119 
120 	rds_stats_inc(s_conn_reset);
121 	rds_send_path_reset(cp);
122 	cp->cp_flags = 0;
123 
124 	/* Do not clear next_rx_seq here, else we cannot distinguish
125 	 * retransmitted packets from new packets, and will hand all
126 	 * of them to the application. That is not consistent with the
127 	 * reliability guarantees of RDS. */
128 }
129 
130 static void __rds_conn_path_init(struct rds_connection *conn,
131 				 struct rds_conn_path *cp, bool is_outgoing)
132 {
133 	spin_lock_init(&cp->cp_lock);
134 	cp->cp_next_tx_seq = 1;
135 	init_waitqueue_head(&cp->cp_waitq);
136 	INIT_LIST_HEAD(&cp->cp_send_queue);
137 	INIT_LIST_HEAD(&cp->cp_retrans);
138 
139 	cp->cp_conn = conn;
140 	atomic_set(&cp->cp_state, RDS_CONN_DOWN);
141 	cp->cp_send_gen = 0;
142 	cp->cp_reconnect_jiffies = 0;
143 	cp->cp_conn->c_proposed_version = RDS_PROTOCOL_VERSION;
144 	INIT_DELAYED_WORK(&cp->cp_send_w, rds_send_worker);
145 	INIT_DELAYED_WORK(&cp->cp_recv_w, rds_recv_worker);
146 	INIT_DELAYED_WORK(&cp->cp_conn_w, rds_connect_worker);
147 	INIT_WORK(&cp->cp_down_w, rds_shutdown_worker);
148 	mutex_init(&cp->cp_cm_lock);
149 	cp->cp_flags = 0;
150 }
151 
152 /*
153  * There is only every one 'conn' for a given pair of addresses in the
154  * system at a time.  They contain messages to be retransmitted and so
155  * span the lifetime of the actual underlying transport connections.
156  *
157  * For now they are not garbage collected once they're created.  They
158  * are torn down as the module is removed, if ever.
159  */
160 static struct rds_connection *__rds_conn_create(struct net *net,
161 						const struct in6_addr *laddr,
162 						const struct in6_addr *faddr,
163 						struct rds_transport *trans,
164 						gfp_t gfp, u8 tos,
165 						int is_outgoing,
166 						int dev_if)
167 {
168 	struct rds_connection *conn, *parent = NULL;
169 	struct hlist_head *head = rds_conn_bucket(laddr, faddr);
170 	struct rds_transport *loop_trans;
171 	unsigned long flags;
172 	int ret, i;
173 	int npaths = (trans->t_mp_capable ? RDS_MPATH_WORKERS : 1);
174 
175 	rcu_read_lock();
176 	conn = rds_conn_lookup(net, head, laddr, faddr, trans, tos, dev_if);
177 	if (conn &&
178 	    conn->c_loopback &&
179 	    conn->c_trans != &rds_loop_transport &&
180 	    ipv6_addr_equal(laddr, faddr) &&
181 	    !is_outgoing) {
182 		/* This is a looped back IB connection, and we're
183 		 * called by the code handling the incoming connect.
184 		 * We need a second connection object into which we
185 		 * can stick the other QP. */
186 		parent = conn;
187 		conn = parent->c_passive;
188 	}
189 	rcu_read_unlock();
190 	if (conn)
191 		goto out;
192 
193 	conn = kmem_cache_zalloc(rds_conn_slab, gfp);
194 	if (!conn) {
195 		conn = ERR_PTR(-ENOMEM);
196 		goto out;
197 	}
198 	conn->c_path = kcalloc(npaths, sizeof(struct rds_conn_path), gfp);
199 	if (!conn->c_path) {
200 		kmem_cache_free(rds_conn_slab, conn);
201 		conn = ERR_PTR(-ENOMEM);
202 		goto out;
203 	}
204 
205 	INIT_HLIST_NODE(&conn->c_hash_node);
206 	conn->c_laddr = *laddr;
207 	conn->c_isv6 = !ipv6_addr_v4mapped(laddr);
208 	conn->c_faddr = *faddr;
209 	conn->c_dev_if = dev_if;
210 	conn->c_tos = tos;
211 
212 #if IS_ENABLED(CONFIG_IPV6)
213 	/* If the local address is link local, set c_bound_if to be the
214 	 * index used for this connection.  Otherwise, set it to 0 as
215 	 * the socket is not bound to an interface.  c_bound_if is used
216 	 * to look up a socket when a packet is received
217 	 */
218 	if (ipv6_addr_type(laddr) & IPV6_ADDR_LINKLOCAL)
219 		conn->c_bound_if = dev_if;
220 	else
221 #endif
222 		conn->c_bound_if = 0;
223 
224 	rds_conn_net_set(conn, net);
225 
226 	ret = rds_cong_get_maps(conn);
227 	if (ret) {
228 		kfree(conn->c_path);
229 		kmem_cache_free(rds_conn_slab, conn);
230 		conn = ERR_PTR(ret);
231 		goto out;
232 	}
233 
234 	/*
235 	 * This is where a connection becomes loopback.  If *any* RDS sockets
236 	 * can bind to the destination address then we'd rather the messages
237 	 * flow through loopback rather than either transport.
238 	 */
239 	loop_trans = rds_trans_get_preferred(net, faddr, conn->c_dev_if);
240 	if (loop_trans) {
241 		rds_trans_put(loop_trans);
242 		conn->c_loopback = 1;
243 		if (trans->t_prefer_loopback) {
244 			if (likely(is_outgoing)) {
245 				/* "outgoing" connection to local address.
246 				 * Protocol says it wants the connection
247 				 * handled by the loopback transport.
248 				 * This is what TCP does.
249 				 */
250 				trans = &rds_loop_transport;
251 			} else {
252 				/* No transport currently in use
253 				 * should end up here, but if it
254 				 * does, reset/destroy the connection.
255 				 */
256 				kmem_cache_free(rds_conn_slab, conn);
257 				conn = ERR_PTR(-EOPNOTSUPP);
258 				goto out;
259 			}
260 		}
261 	}
262 
263 	conn->c_trans = trans;
264 
265 	init_waitqueue_head(&conn->c_hs_waitq);
266 	for (i = 0; i < npaths; i++) {
267 		__rds_conn_path_init(conn, &conn->c_path[i],
268 				     is_outgoing);
269 		conn->c_path[i].cp_index = i;
270 	}
271 	rcu_read_lock();
272 	if (rds_destroy_pending(conn))
273 		ret = -ENETDOWN;
274 	else
275 		ret = trans->conn_alloc(conn, GFP_ATOMIC);
276 	if (ret) {
277 		rcu_read_unlock();
278 		kfree(conn->c_path);
279 		kmem_cache_free(rds_conn_slab, conn);
280 		conn = ERR_PTR(ret);
281 		goto out;
282 	}
283 
284 	rdsdebug("allocated conn %p for %pI6c -> %pI6c over %s %s\n",
285 		 conn, laddr, faddr,
286 		 strnlen(trans->t_name, sizeof(trans->t_name)) ?
287 		 trans->t_name : "[unknown]", is_outgoing ? "(outgoing)" : "");
288 
289 	/*
290 	 * Since we ran without holding the conn lock, someone could
291 	 * have created the same conn (either normal or passive) in the
292 	 * interim. We check while holding the lock. If we won, we complete
293 	 * init and return our conn. If we lost, we rollback and return the
294 	 * other one.
295 	 */
296 	spin_lock_irqsave(&rds_conn_lock, flags);
297 	if (parent) {
298 		/* Creating passive conn */
299 		if (parent->c_passive) {
300 			trans->conn_free(conn->c_path[0].cp_transport_data);
301 			kfree(conn->c_path);
302 			kmem_cache_free(rds_conn_slab, conn);
303 			conn = parent->c_passive;
304 		} else {
305 			parent->c_passive = conn;
306 			rds_cong_add_conn(conn);
307 			rds_conn_count++;
308 		}
309 	} else {
310 		/* Creating normal conn */
311 		struct rds_connection *found;
312 
313 		found = rds_conn_lookup(net, head, laddr, faddr, trans,
314 					tos, dev_if);
315 		if (found) {
316 			struct rds_conn_path *cp;
317 			int i;
318 
319 			for (i = 0; i < npaths; i++) {
320 				cp = &conn->c_path[i];
321 				/* The ->conn_alloc invocation may have
322 				 * allocated resource for all paths, so all
323 				 * of them may have to be freed here.
324 				 */
325 				if (cp->cp_transport_data)
326 					trans->conn_free(cp->cp_transport_data);
327 			}
328 			kfree(conn->c_path);
329 			kmem_cache_free(rds_conn_slab, conn);
330 			conn = found;
331 		} else {
332 			conn->c_my_gen_num = rds_gen_num;
333 			conn->c_peer_gen_num = 0;
334 			hlist_add_head_rcu(&conn->c_hash_node, head);
335 			rds_cong_add_conn(conn);
336 			rds_conn_count++;
337 		}
338 	}
339 	spin_unlock_irqrestore(&rds_conn_lock, flags);
340 	rcu_read_unlock();
341 
342 out:
343 	return conn;
344 }
345 
346 struct rds_connection *rds_conn_create(struct net *net,
347 				       const struct in6_addr *laddr,
348 				       const struct in6_addr *faddr,
349 				       struct rds_transport *trans, u8 tos,
350 				       gfp_t gfp, int dev_if)
351 {
352 	return __rds_conn_create(net, laddr, faddr, trans, gfp, tos, 0, dev_if);
353 }
354 EXPORT_SYMBOL_GPL(rds_conn_create);
355 
356 struct rds_connection *rds_conn_create_outgoing(struct net *net,
357 						const struct in6_addr *laddr,
358 						const struct in6_addr *faddr,
359 						struct rds_transport *trans,
360 						u8 tos, gfp_t gfp, int dev_if)
361 {
362 	return __rds_conn_create(net, laddr, faddr, trans, gfp, tos, 1, dev_if);
363 }
364 EXPORT_SYMBOL_GPL(rds_conn_create_outgoing);
365 
366 void rds_conn_shutdown(struct rds_conn_path *cp)
367 {
368 	struct rds_connection *conn = cp->cp_conn;
369 
370 	/* shut it down unless it's down already */
371 	if (!rds_conn_path_transition(cp, RDS_CONN_DOWN, RDS_CONN_DOWN)) {
372 		/*
373 		 * Quiesce the connection mgmt handlers before we start tearing
374 		 * things down. We don't hold the mutex for the entire
375 		 * duration of the shutdown operation, else we may be
376 		 * deadlocking with the CM handler. Instead, the CM event
377 		 * handler is supposed to check for state DISCONNECTING
378 		 */
379 		mutex_lock(&cp->cp_cm_lock);
380 		if (!rds_conn_path_transition(cp, RDS_CONN_UP,
381 					      RDS_CONN_DISCONNECTING) &&
382 		    !rds_conn_path_transition(cp, RDS_CONN_ERROR,
383 					      RDS_CONN_DISCONNECTING)) {
384 			rds_conn_path_error(cp,
385 					    "shutdown called in state %d\n",
386 					    atomic_read(&cp->cp_state));
387 			mutex_unlock(&cp->cp_cm_lock);
388 			return;
389 		}
390 		mutex_unlock(&cp->cp_cm_lock);
391 
392 		wait_event(cp->cp_waitq,
393 			   !test_bit(RDS_IN_XMIT, &cp->cp_flags));
394 		wait_event(cp->cp_waitq,
395 			   !test_bit(RDS_RECV_REFILL, &cp->cp_flags));
396 
397 		conn->c_trans->conn_path_shutdown(cp);
398 		rds_conn_path_reset(cp);
399 
400 		if (!rds_conn_path_transition(cp, RDS_CONN_DISCONNECTING,
401 					      RDS_CONN_DOWN) &&
402 		    !rds_conn_path_transition(cp, RDS_CONN_ERROR,
403 					      RDS_CONN_DOWN)) {
404 			/* This can happen - eg when we're in the middle of tearing
405 			 * down the connection, and someone unloads the rds module.
406 			 * Quite reproducible with loopback connections.
407 			 * Mostly harmless.
408 			 *
409 			 * Note that this also happens with rds-tcp because
410 			 * we could have triggered rds_conn_path_drop in irq
411 			 * mode from rds_tcp_state change on the receipt of
412 			 * a FIN, thus we need to recheck for RDS_CONN_ERROR
413 			 * here.
414 			 */
415 			rds_conn_path_error(cp, "%s: failed to transition "
416 					    "to state DOWN, current state "
417 					    "is %d\n", __func__,
418 					    atomic_read(&cp->cp_state));
419 			return;
420 		}
421 	}
422 
423 	/* Then reconnect if it's still live.
424 	 * The passive side of an IB loopback connection is never added
425 	 * to the conn hash, so we never trigger a reconnect on this
426 	 * conn - the reconnect is always triggered by the active peer. */
427 	cancel_delayed_work_sync(&cp->cp_conn_w);
428 	rcu_read_lock();
429 	if (!hlist_unhashed(&conn->c_hash_node)) {
430 		rcu_read_unlock();
431 		rds_queue_reconnect(cp);
432 	} else {
433 		rcu_read_unlock();
434 	}
435 }
436 
437 /* destroy a single rds_conn_path. rds_conn_destroy() iterates over
438  * all paths using rds_conn_path_destroy()
439  */
440 static void rds_conn_path_destroy(struct rds_conn_path *cp)
441 {
442 	struct rds_message *rm, *rtmp;
443 
444 	if (!cp->cp_transport_data)
445 		return;
446 
447 	/* make sure lingering queued work won't try to ref the conn */
448 	cancel_delayed_work_sync(&cp->cp_send_w);
449 	cancel_delayed_work_sync(&cp->cp_recv_w);
450 
451 	rds_conn_path_drop(cp, true);
452 	flush_work(&cp->cp_down_w);
453 
454 	/* tear down queued messages */
455 	list_for_each_entry_safe(rm, rtmp,
456 				 &cp->cp_send_queue,
457 				 m_conn_item) {
458 		list_del_init(&rm->m_conn_item);
459 		BUG_ON(!list_empty(&rm->m_sock_item));
460 		rds_message_put(rm);
461 	}
462 	if (cp->cp_xmit_rm)
463 		rds_message_put(cp->cp_xmit_rm);
464 
465 	WARN_ON(delayed_work_pending(&cp->cp_send_w));
466 	WARN_ON(delayed_work_pending(&cp->cp_recv_w));
467 	WARN_ON(delayed_work_pending(&cp->cp_conn_w));
468 	WARN_ON(work_pending(&cp->cp_down_w));
469 
470 	cp->cp_conn->c_trans->conn_free(cp->cp_transport_data);
471 }
472 
473 /*
474  * Stop and free a connection.
475  *
476  * This can only be used in very limited circumstances.  It assumes that once
477  * the conn has been shutdown that no one else is referencing the connection.
478  * We can only ensure this in the rmmod path in the current code.
479  */
480 void rds_conn_destroy(struct rds_connection *conn)
481 {
482 	unsigned long flags;
483 	int i;
484 	struct rds_conn_path *cp;
485 	int npaths = (conn->c_trans->t_mp_capable ? RDS_MPATH_WORKERS : 1);
486 
487 	rdsdebug("freeing conn %p for %pI4 -> "
488 		 "%pI4\n", conn, &conn->c_laddr,
489 		 &conn->c_faddr);
490 
491 	/* Ensure conn will not be scheduled for reconnect */
492 	spin_lock_irq(&rds_conn_lock);
493 	hlist_del_init_rcu(&conn->c_hash_node);
494 	spin_unlock_irq(&rds_conn_lock);
495 	synchronize_rcu();
496 
497 	/* shut the connection down */
498 	for (i = 0; i < npaths; i++) {
499 		cp = &conn->c_path[i];
500 		rds_conn_path_destroy(cp);
501 		BUG_ON(!list_empty(&cp->cp_retrans));
502 	}
503 
504 	/*
505 	 * The congestion maps aren't freed up here.  They're
506 	 * freed by rds_cong_exit() after all the connections
507 	 * have been freed.
508 	 */
509 	rds_cong_remove_conn(conn);
510 
511 	kfree(conn->c_path);
512 	kmem_cache_free(rds_conn_slab, conn);
513 
514 	spin_lock_irqsave(&rds_conn_lock, flags);
515 	rds_conn_count--;
516 	spin_unlock_irqrestore(&rds_conn_lock, flags);
517 }
518 EXPORT_SYMBOL_GPL(rds_conn_destroy);
519 
520 static void __rds_inc_msg_cp(struct rds_incoming *inc,
521 			     struct rds_info_iterator *iter,
522 			     void *saddr, void *daddr, int flip, bool isv6)
523 {
524 #if IS_ENABLED(CONFIG_IPV6)
525 	if (isv6)
526 		rds6_inc_info_copy(inc, iter, saddr, daddr, flip);
527 	else
528 #endif
529 		rds_inc_info_copy(inc, iter, *(__be32 *)saddr,
530 				  *(__be32 *)daddr, flip);
531 }
532 
533 static void rds_conn_message_info_cmn(struct socket *sock, unsigned int len,
534 				      struct rds_info_iterator *iter,
535 				      struct rds_info_lengths *lens,
536 				      int want_send, bool isv6)
537 {
538 	struct hlist_head *head;
539 	struct list_head *list;
540 	struct rds_connection *conn;
541 	struct rds_message *rm;
542 	unsigned int total = 0;
543 	unsigned long flags;
544 	size_t i;
545 	int j;
546 
547 	if (isv6)
548 		len /= sizeof(struct rds6_info_message);
549 	else
550 		len /= sizeof(struct rds_info_message);
551 
552 	rcu_read_lock();
553 
554 	for (i = 0, head = rds_conn_hash; i < ARRAY_SIZE(rds_conn_hash);
555 	     i++, head++) {
556 		hlist_for_each_entry_rcu(conn, head, c_hash_node) {
557 			struct rds_conn_path *cp;
558 			int npaths;
559 
560 			if (!isv6 && conn->c_isv6)
561 				continue;
562 
563 			npaths = (conn->c_trans->t_mp_capable ?
564 				 RDS_MPATH_WORKERS : 1);
565 
566 			for (j = 0; j < npaths; j++) {
567 				cp = &conn->c_path[j];
568 				if (want_send)
569 					list = &cp->cp_send_queue;
570 				else
571 					list = &cp->cp_retrans;
572 
573 				spin_lock_irqsave(&cp->cp_lock, flags);
574 
575 				/* XXX too lazy to maintain counts.. */
576 				list_for_each_entry(rm, list, m_conn_item) {
577 					total++;
578 					if (total <= len)
579 						__rds_inc_msg_cp(&rm->m_inc,
580 								 iter,
581 								 &conn->c_laddr,
582 								 &conn->c_faddr,
583 								 0, isv6);
584 				}
585 
586 				spin_unlock_irqrestore(&cp->cp_lock, flags);
587 			}
588 		}
589 	}
590 	rcu_read_unlock();
591 
592 	lens->nr = total;
593 	if (isv6)
594 		lens->each = sizeof(struct rds6_info_message);
595 	else
596 		lens->each = sizeof(struct rds_info_message);
597 }
598 
599 static void rds_conn_message_info(struct socket *sock, unsigned int len,
600 				  struct rds_info_iterator *iter,
601 				  struct rds_info_lengths *lens,
602 				  int want_send)
603 {
604 	rds_conn_message_info_cmn(sock, len, iter, lens, want_send, false);
605 }
606 
607 #if IS_ENABLED(CONFIG_IPV6)
608 static void rds6_conn_message_info(struct socket *sock, unsigned int len,
609 				   struct rds_info_iterator *iter,
610 				   struct rds_info_lengths *lens,
611 				   int want_send)
612 {
613 	rds_conn_message_info_cmn(sock, len, iter, lens, want_send, true);
614 }
615 #endif
616 
617 static void rds_conn_message_info_send(struct socket *sock, unsigned int len,
618 				       struct rds_info_iterator *iter,
619 				       struct rds_info_lengths *lens)
620 {
621 	rds_conn_message_info(sock, len, iter, lens, 1);
622 }
623 
624 #if IS_ENABLED(CONFIG_IPV6)
625 static void rds6_conn_message_info_send(struct socket *sock, unsigned int len,
626 					struct rds_info_iterator *iter,
627 					struct rds_info_lengths *lens)
628 {
629 	rds6_conn_message_info(sock, len, iter, lens, 1);
630 }
631 #endif
632 
633 static void rds_conn_message_info_retrans(struct socket *sock,
634 					  unsigned int len,
635 					  struct rds_info_iterator *iter,
636 					  struct rds_info_lengths *lens)
637 {
638 	rds_conn_message_info(sock, len, iter, lens, 0);
639 }
640 
641 #if IS_ENABLED(CONFIG_IPV6)
642 static void rds6_conn_message_info_retrans(struct socket *sock,
643 					   unsigned int len,
644 					   struct rds_info_iterator *iter,
645 					   struct rds_info_lengths *lens)
646 {
647 	rds6_conn_message_info(sock, len, iter, lens, 0);
648 }
649 #endif
650 
651 void rds_for_each_conn_info(struct socket *sock, unsigned int len,
652 			  struct rds_info_iterator *iter,
653 			  struct rds_info_lengths *lens,
654 			  int (*visitor)(struct rds_connection *, void *),
655 			  u64 *buffer,
656 			  size_t item_len)
657 {
658 	struct hlist_head *head;
659 	struct rds_connection *conn;
660 	size_t i;
661 
662 	rcu_read_lock();
663 
664 	lens->nr = 0;
665 	lens->each = item_len;
666 
667 	for (i = 0, head = rds_conn_hash; i < ARRAY_SIZE(rds_conn_hash);
668 	     i++, head++) {
669 		hlist_for_each_entry_rcu(conn, head, c_hash_node) {
670 
671 			/* XXX no c_lock usage.. */
672 			if (!visitor(conn, buffer))
673 				continue;
674 
675 			/* We copy as much as we can fit in the buffer,
676 			 * but we count all items so that the caller
677 			 * can resize the buffer. */
678 			if (len >= item_len) {
679 				rds_info_copy(iter, buffer, item_len);
680 				len -= item_len;
681 			}
682 			lens->nr++;
683 		}
684 	}
685 	rcu_read_unlock();
686 }
687 EXPORT_SYMBOL_GPL(rds_for_each_conn_info);
688 
689 static void rds_walk_conn_path_info(struct socket *sock, unsigned int len,
690 				    struct rds_info_iterator *iter,
691 				    struct rds_info_lengths *lens,
692 				    int (*visitor)(struct rds_conn_path *, void *),
693 				    u64 *buffer,
694 				    size_t item_len)
695 {
696 	struct hlist_head *head;
697 	struct rds_connection *conn;
698 	size_t i;
699 
700 	rcu_read_lock();
701 
702 	lens->nr = 0;
703 	lens->each = item_len;
704 
705 	for (i = 0, head = rds_conn_hash; i < ARRAY_SIZE(rds_conn_hash);
706 	     i++, head++) {
707 		hlist_for_each_entry_rcu(conn, head, c_hash_node) {
708 			struct rds_conn_path *cp;
709 
710 			/* XXX We only copy the information from the first
711 			 * path for now.  The problem is that if there are
712 			 * more than one underlying paths, we cannot report
713 			 * information of all of them using the existing
714 			 * API.  For example, there is only one next_tx_seq,
715 			 * which path's next_tx_seq should we report?  It is
716 			 * a bug in the design of MPRDS.
717 			 */
718 			cp = conn->c_path;
719 
720 			/* XXX no cp_lock usage.. */
721 			if (!visitor(cp, buffer))
722 				continue;
723 
724 			/* We copy as much as we can fit in the buffer,
725 			 * but we count all items so that the caller
726 			 * can resize the buffer.
727 			 */
728 			if (len >= item_len) {
729 				rds_info_copy(iter, buffer, item_len);
730 				len -= item_len;
731 			}
732 			lens->nr++;
733 		}
734 	}
735 	rcu_read_unlock();
736 }
737 
738 static int rds_conn_info_visitor(struct rds_conn_path *cp, void *buffer)
739 {
740 	struct rds_info_connection *cinfo = buffer;
741 	struct rds_connection *conn = cp->cp_conn;
742 
743 	if (conn->c_isv6)
744 		return 0;
745 
746 	cinfo->next_tx_seq = cp->cp_next_tx_seq;
747 	cinfo->next_rx_seq = cp->cp_next_rx_seq;
748 	cinfo->laddr = conn->c_laddr.s6_addr32[3];
749 	cinfo->faddr = conn->c_faddr.s6_addr32[3];
750 	cinfo->tos = conn->c_tos;
751 	strncpy(cinfo->transport, conn->c_trans->t_name,
752 		sizeof(cinfo->transport));
753 	cinfo->flags = 0;
754 
755 	rds_conn_info_set(cinfo->flags, test_bit(RDS_IN_XMIT, &cp->cp_flags),
756 			  SENDING);
757 	/* XXX Future: return the state rather than these funky bits */
758 	rds_conn_info_set(cinfo->flags,
759 			  atomic_read(&cp->cp_state) == RDS_CONN_CONNECTING,
760 			  CONNECTING);
761 	rds_conn_info_set(cinfo->flags,
762 			  atomic_read(&cp->cp_state) == RDS_CONN_UP,
763 			  CONNECTED);
764 	return 1;
765 }
766 
767 #if IS_ENABLED(CONFIG_IPV6)
768 static int rds6_conn_info_visitor(struct rds_conn_path *cp, void *buffer)
769 {
770 	struct rds6_info_connection *cinfo6 = buffer;
771 	struct rds_connection *conn = cp->cp_conn;
772 
773 	cinfo6->next_tx_seq = cp->cp_next_tx_seq;
774 	cinfo6->next_rx_seq = cp->cp_next_rx_seq;
775 	cinfo6->laddr = conn->c_laddr;
776 	cinfo6->faddr = conn->c_faddr;
777 	strncpy(cinfo6->transport, conn->c_trans->t_name,
778 		sizeof(cinfo6->transport));
779 	cinfo6->flags = 0;
780 
781 	rds_conn_info_set(cinfo6->flags, test_bit(RDS_IN_XMIT, &cp->cp_flags),
782 			  SENDING);
783 	/* XXX Future: return the state rather than these funky bits */
784 	rds_conn_info_set(cinfo6->flags,
785 			  atomic_read(&cp->cp_state) == RDS_CONN_CONNECTING,
786 			  CONNECTING);
787 	rds_conn_info_set(cinfo6->flags,
788 			  atomic_read(&cp->cp_state) == RDS_CONN_UP,
789 			  CONNECTED);
790 	/* Just return 1 as there is no error case. This is a helper function
791 	 * for rds_walk_conn_path_info() and it wants a return value.
792 	 */
793 	return 1;
794 }
795 #endif
796 
797 static void rds_conn_info(struct socket *sock, unsigned int len,
798 			  struct rds_info_iterator *iter,
799 			  struct rds_info_lengths *lens)
800 {
801 	u64 buffer[(sizeof(struct rds_info_connection) + 7) / 8];
802 
803 	rds_walk_conn_path_info(sock, len, iter, lens,
804 				rds_conn_info_visitor,
805 				buffer,
806 				sizeof(struct rds_info_connection));
807 }
808 
809 #if IS_ENABLED(CONFIG_IPV6)
810 static void rds6_conn_info(struct socket *sock, unsigned int len,
811 			   struct rds_info_iterator *iter,
812 			   struct rds_info_lengths *lens)
813 {
814 	u64 buffer[(sizeof(struct rds6_info_connection) + 7) / 8];
815 
816 	rds_walk_conn_path_info(sock, len, iter, lens,
817 				rds6_conn_info_visitor,
818 				buffer,
819 				sizeof(struct rds6_info_connection));
820 }
821 #endif
822 
823 int rds_conn_init(void)
824 {
825 	int ret;
826 
827 	ret = rds_loop_net_init(); /* register pernet callback */
828 	if (ret)
829 		return ret;
830 
831 	rds_conn_slab = kmem_cache_create("rds_connection",
832 					  sizeof(struct rds_connection),
833 					  0, 0, NULL);
834 	if (!rds_conn_slab) {
835 		rds_loop_net_exit();
836 		return -ENOMEM;
837 	}
838 
839 	rds_info_register_func(RDS_INFO_CONNECTIONS, rds_conn_info);
840 	rds_info_register_func(RDS_INFO_SEND_MESSAGES,
841 			       rds_conn_message_info_send);
842 	rds_info_register_func(RDS_INFO_RETRANS_MESSAGES,
843 			       rds_conn_message_info_retrans);
844 #if IS_ENABLED(CONFIG_IPV6)
845 	rds_info_register_func(RDS6_INFO_CONNECTIONS, rds6_conn_info);
846 	rds_info_register_func(RDS6_INFO_SEND_MESSAGES,
847 			       rds6_conn_message_info_send);
848 	rds_info_register_func(RDS6_INFO_RETRANS_MESSAGES,
849 			       rds6_conn_message_info_retrans);
850 #endif
851 	return 0;
852 }
853 
854 void rds_conn_exit(void)
855 {
856 	rds_loop_net_exit(); /* unregister pernet callback */
857 	rds_loop_exit();
858 
859 	WARN_ON(!hlist_empty(rds_conn_hash));
860 
861 	kmem_cache_destroy(rds_conn_slab);
862 
863 	rds_info_deregister_func(RDS_INFO_CONNECTIONS, rds_conn_info);
864 	rds_info_deregister_func(RDS_INFO_SEND_MESSAGES,
865 				 rds_conn_message_info_send);
866 	rds_info_deregister_func(RDS_INFO_RETRANS_MESSAGES,
867 				 rds_conn_message_info_retrans);
868 #if IS_ENABLED(CONFIG_IPV6)
869 	rds_info_deregister_func(RDS6_INFO_CONNECTIONS, rds6_conn_info);
870 	rds_info_deregister_func(RDS6_INFO_SEND_MESSAGES,
871 				 rds6_conn_message_info_send);
872 	rds_info_deregister_func(RDS6_INFO_RETRANS_MESSAGES,
873 				 rds6_conn_message_info_retrans);
874 #endif
875 }
876 
877 /*
878  * Force a disconnect
879  */
880 void rds_conn_path_drop(struct rds_conn_path *cp, bool destroy)
881 {
882 	atomic_set(&cp->cp_state, RDS_CONN_ERROR);
883 
884 	rcu_read_lock();
885 	if (!destroy && rds_destroy_pending(cp->cp_conn)) {
886 		rcu_read_unlock();
887 		return;
888 	}
889 	queue_work(rds_wq, &cp->cp_down_w);
890 	rcu_read_unlock();
891 }
892 EXPORT_SYMBOL_GPL(rds_conn_path_drop);
893 
894 void rds_conn_drop(struct rds_connection *conn)
895 {
896 	WARN_ON(conn->c_trans->t_mp_capable);
897 	rds_conn_path_drop(&conn->c_path[0], false);
898 }
899 EXPORT_SYMBOL_GPL(rds_conn_drop);
900 
901 /*
902  * If the connection is down, trigger a connect. We may have scheduled a
903  * delayed reconnect however - in this case we should not interfere.
904  */
905 void rds_conn_path_connect_if_down(struct rds_conn_path *cp)
906 {
907 	rcu_read_lock();
908 	if (rds_destroy_pending(cp->cp_conn)) {
909 		rcu_read_unlock();
910 		return;
911 	}
912 	if (rds_conn_path_state(cp) == RDS_CONN_DOWN &&
913 	    !test_and_set_bit(RDS_RECONNECT_PENDING, &cp->cp_flags))
914 		queue_delayed_work(rds_wq, &cp->cp_conn_w, 0);
915 	rcu_read_unlock();
916 }
917 EXPORT_SYMBOL_GPL(rds_conn_path_connect_if_down);
918 
919 /* Check connectivity of all paths
920  */
921 void rds_check_all_paths(struct rds_connection *conn)
922 {
923 	int i = 0;
924 
925 	do {
926 		rds_conn_path_connect_if_down(&conn->c_path[i]);
927 	} while (++i < conn->c_npaths);
928 }
929 
930 void rds_conn_connect_if_down(struct rds_connection *conn)
931 {
932 	WARN_ON(conn->c_trans->t_mp_capable);
933 	rds_conn_path_connect_if_down(&conn->c_path[0]);
934 }
935 EXPORT_SYMBOL_GPL(rds_conn_connect_if_down);
936 
937 void
938 __rds_conn_path_error(struct rds_conn_path *cp, const char *fmt, ...)
939 {
940 	va_list ap;
941 
942 	va_start(ap, fmt);
943 	vprintk(fmt, ap);
944 	va_end(ap);
945 
946 	rds_conn_path_drop(cp, false);
947 }
948