xref: /illumos-gate/usr/src/uts/common/io/ib/clients/rdsv3/connection.c (revision f802f1c0f4fae59d1337c3db328d06d9093427e8)
1 /*
2  * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
3  */
4 
5 /*
6  * This file contains code imported from the OFED rds source file connection.c
7  * Oracle elects to have and use the contents of connection.c under and governed
8  * by the OpenIB.org BSD license (see below for full license text). However,
9  * the following notice accompanied the original version of this file:
10  */
11 
12 /*
13  * Copyright (c) 2006 Oracle.  All rights reserved.
14  *
15  * This software is available to you under a choice of one of two
16  * licenses.  You may choose to be licensed under the terms of the GNU
17  * General Public License (GPL) Version 2, available from the file
18  * COPYING in the main directory of this source tree, or the
19  * OpenIB.org BSD license below:
20  *
21  *     Redistribution and use in source and binary forms, with or
22  *     without modification, are permitted provided that the following
23  *     conditions are met:
24  *
25  *      - Redistributions of source code must retain the above
26  *        copyright notice, this list of conditions and the following
27  *        disclaimer.
28  *
29  *      - Redistributions in binary form must reproduce the above
30  *        copyright notice, this list of conditions and the following
31  *        disclaimer in the documentation and/or other materials
32  *        provided with the distribution.
33  *
34  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
35  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
36  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
37  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
38  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
39  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
40  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
41  * SOFTWARE.
42  *
43  */
44 #include <sys/types.h>
45 #include <sys/kmem.h>
46 #include <sys/rds.h>
47 
48 #include <sys/ib/clients/rdsv3/rdsv3.h>
49 #include <sys/ib/clients/rdsv3/loop.h>
50 #include <sys/ib/clients/rdsv3/rdsv3_debug.h>
51 
52 /* converting this to RCU is a chore for another day.. */
53 static krwlock_t rdsv3_conn_lock;
54 struct avl_tree rdsv3_conn_hash;
55 static struct kmem_cache *rdsv3_conn_slab = NULL;
56 
57 #define	rdsv3_conn_info_set(var, test, suffix) do {               \
58 	if (test)                                               \
59 		var |= RDS_INFO_CONNECTION_FLAG_##suffix;     \
60 } while (0)
61 
62 
63 static struct rdsv3_connection *
rdsv3_conn_lookup(uint32_be_t laddr,uint32_be_t faddr,avl_index_t * pos)64 rdsv3_conn_lookup(uint32_be_t laddr, uint32_be_t faddr, avl_index_t *pos)
65 {
66 	struct rdsv3_connection *conn;
67 	struct rdsv3_conn_info_s conn_info;
68 	avl_index_t place = 0;
69 
70 	conn_info.c_laddr = laddr;
71 	conn_info.c_faddr = faddr;
72 
73 	conn = avl_find(&rdsv3_conn_hash, &conn_info, &place);
74 
75 	RDSV3_DPRINTF5("rdsv3_conn_lookup",
76 	    "returning conn %p for %u.%u.%u.%u -> %u.%u.%u.%u",
77 	    conn, NIPQUAD(laddr), NIPQUAD(faddr));
78 
79 	if (pos != NULL)
80 		*pos = place;
81 
82 	return (conn);
83 }
84 
85 /*
86  * This is called by transports as they're bringing down a connection.
87  * It clears partial message state so that the transport can start sending
88  * and receiving over this connection again in the future.  It is up to
89  * the transport to have serialized this call with its send and recv.
90  */
91 void
rdsv3_conn_reset(struct rdsv3_connection * conn)92 rdsv3_conn_reset(struct rdsv3_connection *conn)
93 {
94 	RDSV3_DPRINTF2("rdsv3_conn_reset",
95 	    "connection %u.%u.%u.%u to %u.%u.%u.%u reset",
96 	    NIPQUAD(conn->c_laddr), NIPQUAD(conn->c_faddr));
97 
98 	rdsv3_stats_inc(s_conn_reset);
99 	rdsv3_send_reset(conn);
100 	conn->c_flags = 0;
101 
102 	/*
103 	 * Do not clear next_rx_seq here, else we cannot distinguish
104 	 * retransmitted packets from new packets, and will hand all
105 	 * of them to the application. That is not consistent with the
106 	 * reliability guarantees of RDS.
107 	 */
108 }
109 
110 /*
111  * There is only every one 'conn' for a given pair of addresses in the
112  * system at a time.  They contain messages to be retransmitted and so
113  * span the lifetime of the actual underlying transport connections.
114  *
115  * For now they are not garbage collected once they're created.  They
116  * are torn down as the module is removed, if ever.
117  */
118 static struct rdsv3_connection *
__rdsv3_conn_create(uint32_be_t laddr,uint32_be_t faddr,struct rdsv3_transport * trans,int gfp,int is_outgoing)119 __rdsv3_conn_create(uint32_be_t laddr, uint32_be_t faddr,
120     struct rdsv3_transport *trans, int gfp, int is_outgoing)
121 {
122 	struct rdsv3_connection *conn, *parent = NULL;
123 	avl_index_t pos;
124 	int ret;
125 
126 	rw_enter(&rdsv3_conn_lock, RW_READER);
127 	conn = rdsv3_conn_lookup(laddr, faddr, &pos);
128 	if (conn &&
129 	    conn->c_loopback &&
130 	    conn->c_trans != &rdsv3_loop_transport &&
131 	    !is_outgoing) {
132 		/*
133 		 * This is a looped back IB connection, and we're
134 		 * called by the code handling the incoming connect.
135 		 * We need a second connection object into which we
136 		 * can stick the other QP.
137 		 */
138 		parent = conn;
139 		conn = parent->c_passive;
140 	}
141 	rw_exit(&rdsv3_conn_lock);
142 	if (conn)
143 		goto out;
144 
145 	RDSV3_DPRINTF2("__rdsv3_conn_create", "Enter(%x -> %x)",
146 	    ntohl(laddr), ntohl(faddr));
147 
148 	conn = kmem_cache_alloc(rdsv3_conn_slab, gfp);
149 	if (!conn) {
150 		conn = ERR_PTR(-ENOMEM);
151 		goto out;
152 	}
153 
154 	/* see rdsv3_conn_constructor */
155 	conn->c_laddr = laddr;
156 	conn->c_faddr = faddr;
157 
158 	/*
159 	 * We don't allow sockets to send messages without binding.
160 	 * So, the IP address will already be there in the bind array.
161 	 * Mostly, this is a readonly operation.
162 	 * For now, passing GLOBAL_ZONEID.
163 	 */
164 	conn->c_bucketp = rdsv3_find_ip_bucket(ntohl(laddr), GLOBAL_ZONEID);
165 
166 	ret = rdsv3_cong_get_maps(conn);
167 	if (ret) {
168 		kmem_cache_free(rdsv3_conn_slab, conn);
169 		conn = ERR_PTR(ret);
170 		goto out;
171 	}
172 
173 	/*
174 	 * This is where a connection becomes loopback.  If *any* RDS sockets
175 	 * can bind to the destination address then we'd rather the messages
176 	 * flow through loopback rather than either transport.
177 	 */
178 	if (rdsv3_trans_get_preferred(faddr)) {
179 		conn->c_loopback = 1;
180 		if (is_outgoing && trans->t_prefer_loopback) {
181 			/*
182 			 * "outgoing" connection - and the transport
183 			 * says it wants the connection handled by the
184 			 * loopback transport. This is what TCP does.
185 			 */
186 			trans = &rdsv3_loop_transport;
187 		}
188 	}
189 
190 	conn->c_trans = trans;
191 
192 	ret = trans->conn_alloc(conn, gfp);
193 	if (ret) {
194 		kmem_cache_free(rdsv3_conn_slab, conn);
195 		conn = ERR_PTR(ret);
196 		goto out;
197 	}
198 
199 	conn->c_state = RDSV3_CONN_DOWN;
200 	conn->c_reconnect_jiffies = 0;
201 	RDSV3_INIT_DELAYED_WORK(&conn->c_send_w, rdsv3_send_worker);
202 	RDSV3_INIT_DELAYED_WORK(&conn->c_recv_w, rdsv3_recv_worker);
203 	RDSV3_INIT_DELAYED_WORK(&conn->c_conn_w, rdsv3_connect_worker);
204 	RDSV3_INIT_DELAYED_WORK(&conn->c_reap_w, rdsv3_reaper_worker);
205 	RDSV3_INIT_WORK(&conn->c_down_w, rdsv3_shutdown_worker);
206 	mutex_init(&conn->c_cm_lock, NULL, MUTEX_DRIVER, NULL);
207 	conn->c_flags = 0;
208 
209 	RDSV3_DPRINTF2("__rdsv3_conn_create",
210 	    "allocated conn %p for %u.%u.%u.%u -> %u.%u.%u.%u over %s %s",
211 	    conn, NIPQUAD(laddr), NIPQUAD(faddr),
212 	    *trans->t_name != '\0' ? trans->t_name : "[unknown]",
213 	    is_outgoing ? "(outgoing)" : "");
214 
215 	/*
216 	 * Since we ran without holding the conn lock, someone could
217 	 * have created the same conn (either normal or passive) in the
218 	 * interim. We check while holding the lock. If we won, we complete
219 	 * init and return our conn. If we lost, we rollback and return the
220 	 * other one.
221 	 */
222 	rw_enter(&rdsv3_conn_lock, RW_WRITER);
223 	if (parent) {
224 		/* Creating passive conn */
225 		if (parent->c_passive) {
226 			trans->conn_free(conn->c_transport_data);
227 			kmem_cache_free(rdsv3_conn_slab, conn);
228 			conn = parent->c_passive;
229 		} else {
230 			parent->c_passive = conn;
231 			rdsv3_cong_add_conn(conn);
232 		}
233 	} else {
234 		/* Creating normal conn */
235 		struct rdsv3_connection *found;
236 
237 		found = rdsv3_conn_lookup(laddr, faddr, &pos);
238 		if (found) {
239 			trans->conn_free(conn->c_transport_data);
240 			kmem_cache_free(rdsv3_conn_slab, conn);
241 			conn = found;
242 		} else {
243 			avl_insert(&rdsv3_conn_hash, conn, pos);
244 			rdsv3_cong_add_conn(conn);
245 			rdsv3_queue_delayed_work(rdsv3_wq, &conn->c_reap_w,
246 			    RDSV3_REAPER_WAIT_JIFFIES);
247 		}
248 	}
249 
250 	rw_exit(&rdsv3_conn_lock);
251 
252 	RDSV3_DPRINTF2("__rdsv3_conn_create", "Return(conn: %p)", conn);
253 
254 out:
255 	return (conn);
256 }
257 
258 struct rdsv3_connection *
rdsv3_conn_create(uint32_be_t laddr,uint32_be_t faddr,struct rdsv3_transport * trans,int gfp)259 rdsv3_conn_create(uint32_be_t laddr, uint32_be_t faddr,
260     struct rdsv3_transport *trans, int gfp)
261 {
262 	return (__rdsv3_conn_create(laddr, faddr, trans, gfp, 0));
263 }
264 
265 struct rdsv3_connection *
rdsv3_conn_create_outgoing(uint32_be_t laddr,uint32_be_t faddr,struct rdsv3_transport * trans,int gfp)266 rdsv3_conn_create_outgoing(uint32_be_t laddr, uint32_be_t faddr,
267     struct rdsv3_transport *trans, int gfp)
268 {
269 	return (__rdsv3_conn_create(laddr, faddr, trans, gfp, 1));
270 }
271 
272 extern struct avl_tree	rdsv3_conn_hash;
273 
274 void
rdsv3_conn_shutdown(struct rdsv3_connection * conn)275 rdsv3_conn_shutdown(struct rdsv3_connection *conn)
276 {
277 	RDSV3_DPRINTF2("rdsv3_conn_shutdown", "Enter(conn: %p)", conn);
278 
279 	/* shut it down unless it's down already */
280 	if (!rdsv3_conn_transition(conn, RDSV3_CONN_DOWN, RDSV3_CONN_DOWN)) {
281 		/*
282 		 * Quiesce the connection mgmt handlers before we start tearing
283 		 * things down. We don't hold the mutex for the entire
284 		 * duration of the shutdown operation, else we may be
285 		 * deadlocking with the CM handler. Instead, the CM event
286 		 * handler is supposed to check for state DISCONNECTING
287 		 */
288 		mutex_enter(&conn->c_cm_lock);
289 		if (!rdsv3_conn_transition(conn, RDSV3_CONN_UP,
290 		    RDSV3_CONN_DISCONNECTING) &&
291 		    !rdsv3_conn_transition(conn, RDSV3_CONN_ERROR,
292 		    RDSV3_CONN_DISCONNECTING)) {
293 			RDSV3_DPRINTF2("rdsv3_conn_shutdown",
294 			    "shutdown called in state %d",
295 			    atomic_get(&conn->c_state));
296 			rdsv3_conn_drop(conn);
297 			mutex_exit(&conn->c_cm_lock);
298 			return;
299 		}
300 		mutex_exit(&conn->c_cm_lock);
301 
302 		/* verify everybody's out of rds_send_xmit() */
303 		mutex_enter(&conn->c_send_lock);
304 		while (atomic_get(&conn->c_senders)) {
305 			mutex_exit(&conn->c_send_lock);
306 			delay(1);
307 			mutex_enter(&conn->c_send_lock);
308 		}
309 
310 		conn->c_trans->conn_shutdown(conn);
311 		rdsv3_conn_reset(conn);
312 		mutex_exit(&conn->c_send_lock);
313 
314 		if (!rdsv3_conn_transition(conn, RDSV3_CONN_DISCONNECTING,
315 		    RDSV3_CONN_DOWN)) {
316 			/*
317 			 * This can happen - eg when we're in the middle of
318 			 * tearing down the connection, and someone unloads
319 			 * the rds module.
320 			 * Quite reproduceable with loopback connections.
321 			 * Mostly harmless.
322 			 */
323 #ifndef __lock_lint
324 			RDSV3_DPRINTF2("rdsv3_conn_shutdown",
325 			    "failed to transition to state DOWN, "
326 			    "current statis is: %d",
327 			    atomic_get(&conn->c_state));
328 			rdsv3_conn_drop(conn);
329 #endif
330 			return;
331 		}
332 	}
333 
334 	/*
335 	 * Then reconnect if it's still live.
336 	 * The passive side of an IB loopback connection is never added
337 	 * to the conn hash, so we never trigger a reconnect on this
338 	 * conn - the reconnect is always triggered by the active peer.
339 	 */
340 	rdsv3_cancel_delayed_work(&conn->c_conn_w);
341 
342 	{
343 		struct rdsv3_conn_info_s conn_info;
344 
345 		conn_info.c_laddr = conn->c_laddr;
346 		conn_info.c_faddr = conn->c_faddr;
347 		if (avl_find(&rdsv3_conn_hash, &conn_info, NULL) == conn)
348 			rdsv3_queue_reconnect(conn);
349 	}
350 	RDSV3_DPRINTF2("rdsv3_conn_shutdown", "Exit");
351 }
352 
353 /*
354  * Stop and free a connection.
355  */
356 void
rdsv3_conn_destroy(struct rdsv3_connection * conn)357 rdsv3_conn_destroy(struct rdsv3_connection *conn)
358 {
359 	struct rdsv3_message *rm, *rtmp;
360 	list_t to_be_dropped;
361 
362 	RDSV3_DPRINTF4("rdsv3_conn_destroy",
363 	    "freeing conn %p for %u.%u.%u.%u -> %u.%u.%u.%u",
364 	    conn, NIPQUAD(conn->c_laddr), NIPQUAD(conn->c_faddr));
365 
366 	avl_remove(&rdsv3_conn_hash, conn);
367 
368 	rdsv3_cancel_delayed_work(&conn->c_reap_w);
369 	rdsv3_cancel_delayed_work(&conn->c_send_w);
370 	rdsv3_cancel_delayed_work(&conn->c_recv_w);
371 
372 	rdsv3_conn_shutdown(conn);
373 
374 	/* tear down queued messages */
375 
376 	list_create(&to_be_dropped, sizeof (struct rdsv3_message),
377 	    offsetof(struct rdsv3_message, m_conn_item));
378 
379 	RDSV3_FOR_EACH_LIST_NODE_SAFE(rm, rtmp, &conn->c_retrans, m_conn_item) {
380 		list_remove_node(&rm->m_conn_item);
381 		list_insert_tail(&to_be_dropped, rm);
382 	}
383 
384 	RDSV3_FOR_EACH_LIST_NODE_SAFE(rm, rtmp, &conn->c_send_queue,
385 	    m_conn_item) {
386 		list_remove_node(&rm->m_conn_item);
387 		list_insert_tail(&to_be_dropped, rm);
388 	}
389 
390 	RDSV3_FOR_EACH_LIST_NODE_SAFE(rm, rtmp, &to_be_dropped, m_conn_item) {
391 		clear_bit(RDSV3_MSG_ON_CONN, &rm->m_flags);
392 		list_remove_node(&rm->m_conn_item);
393 		rdsv3_message_put(rm);
394 	}
395 
396 	if (conn->c_xmit_rm)
397 		rdsv3_message_put(conn->c_xmit_rm);
398 
399 	conn->c_trans->conn_free(conn->c_transport_data);
400 
401 	/*
402 	 * The congestion maps aren't freed up here.  They're
403 	 * freed by rdsv3_cong_exit() after all the connections
404 	 * have been freed.
405 	 */
406 	rdsv3_cong_remove_conn(conn);
407 
408 	ASSERT(list_is_empty(&conn->c_retrans));
409 	kmem_cache_free(rdsv3_conn_slab, conn);
410 
411 }
412 
413 /* ARGSUSED */
414 static void
rdsv3_conn_message_info(struct rsock * sock,unsigned int len,struct rdsv3_info_iterator * iter,struct rdsv3_info_lengths * lens,int want_send)415 rdsv3_conn_message_info(struct rsock *sock, unsigned int len,
416     struct rdsv3_info_iterator *iter,
417     struct rdsv3_info_lengths *lens,
418     int want_send)
419 {
420 	struct list *list;
421 	struct rdsv3_connection *conn;
422 	struct rdsv3_message *rm;
423 	unsigned int total = 0;
424 
425 	RDSV3_DPRINTF4("rdsv3_conn_message_info", "Enter");
426 
427 	len /= sizeof (struct rds_info_message);
428 
429 	rw_enter(&rdsv3_conn_lock, RW_READER);
430 
431 	if (avl_is_empty(&rdsv3_conn_hash)) {
432 		/* no connections */
433 		rw_exit(&rdsv3_conn_lock);
434 		return;
435 	}
436 
437 	conn = (struct rdsv3_connection *)avl_first(&rdsv3_conn_hash);
438 
439 	do {
440 		if (want_send)
441 			list = &conn->c_send_queue;
442 		else
443 			list = &conn->c_retrans;
444 
445 		mutex_enter(&conn->c_lock);
446 
447 		/* XXX too lazy to maintain counts.. */
448 		RDSV3_FOR_EACH_LIST_NODE(rm, list, m_conn_item) {
449 			total++;
450 			if (total <= len)
451 				rdsv3_inc_info_copy(&rm->m_inc, iter,
452 				    conn->c_laddr, conn->c_faddr, 0);
453 		}
454 
455 		mutex_exit(&conn->c_lock);
456 
457 		conn = AVL_NEXT(&rdsv3_conn_hash, conn);
458 	} while (conn != NULL);
459 	rw_exit(&rdsv3_conn_lock);
460 
461 	lens->nr = total;
462 	lens->each = sizeof (struct rds_info_message);
463 
464 	RDSV3_DPRINTF4("rdsv3_conn_message_info", "Return");
465 }
466 
467 static void
rdsv3_conn_message_info_send(struct rsock * sock,unsigned int len,struct rdsv3_info_iterator * iter,struct rdsv3_info_lengths * lens)468 rdsv3_conn_message_info_send(struct rsock *sock, unsigned int len,
469     struct rdsv3_info_iterator *iter,
470     struct rdsv3_info_lengths *lens)
471 {
472 	rdsv3_conn_message_info(sock, len, iter, lens, 1);
473 }
474 
475 static void
rdsv3_conn_message_info_retrans(struct rsock * sock,unsigned int len,struct rdsv3_info_iterator * iter,struct rdsv3_info_lengths * lens)476 rdsv3_conn_message_info_retrans(struct rsock *sock,
477     unsigned int len,
478     struct rdsv3_info_iterator *iter,
479     struct rdsv3_info_lengths *lens)
480 {
481 	rdsv3_conn_message_info(sock, len, iter, lens, 0);
482 }
483 
484 /* ARGSUSED */
485 void
rdsv3_for_each_conn_info(struct rsock * sock,unsigned int len,struct rdsv3_info_iterator * iter,struct rdsv3_info_lengths * lens,int (* visitor)(struct rdsv3_connection *,void *),size_t item_len)486 rdsv3_for_each_conn_info(struct rsock *sock, unsigned int len,
487     struct rdsv3_info_iterator *iter,
488     struct rdsv3_info_lengths *lens,
489     int (*visitor)(struct rdsv3_connection *, void *),
490     size_t item_len)
491 {
492 	uint8_t *buffer;
493 	struct rdsv3_connection *conn;
494 
495 	rw_enter(&rdsv3_conn_lock, RW_READER);
496 
497 	lens->nr = 0;
498 	lens->each = item_len;
499 
500 	if (avl_is_empty(&rdsv3_conn_hash)) {
501 		/* no connections */
502 		rw_exit(&rdsv3_conn_lock);
503 		return;
504 	}
505 
506 	/* allocate a little extra as this can get cast to a uint64_t */
507 	buffer = kmem_zalloc(item_len + 8, KM_SLEEP);
508 
509 	conn = (struct rdsv3_connection *)avl_first(&rdsv3_conn_hash);
510 
511 	do {
512 		/* XXX no c_lock usage.. */
513 		if (visitor(conn, buffer)) {
514 			/*
515 			 * We copy as much as we can fit in the buffer,
516 			 * but we count all items so that the caller
517 			 * can resize the buffer.
518 			 */
519 			if (len >= item_len) {
520 				RDSV3_DPRINTF4("rdsv3_for_each_conn_info",
521 				    "buffer: %p iter: %p bytes: %d", buffer,
522 				    iter->addr + iter->offset, item_len);
523 				rdsv3_info_copy(iter, buffer, item_len);
524 				len -= item_len;
525 			}
526 			lens->nr++;
527 		}
528 		conn = AVL_NEXT(&rdsv3_conn_hash, conn);
529 	} while (conn != NULL);
530 	rw_exit(&rdsv3_conn_lock);
531 
532 	kmem_free(buffer, item_len + 8);
533 }
534 
535 static int
rdsv3_conn_info_visitor(struct rdsv3_connection * conn,void * buffer)536 rdsv3_conn_info_visitor(struct rdsv3_connection *conn, void *buffer)
537 {
538 	struct rds_info_connection *cinfo = buffer;
539 
540 	cinfo->next_tx_seq = conn->c_next_tx_seq;
541 	cinfo->next_rx_seq = conn->c_next_rx_seq;
542 	cinfo->laddr = conn->c_laddr;
543 	cinfo->faddr = conn->c_faddr;
544 	(void) strncpy((char *)cinfo->transport, conn->c_trans->t_name,
545 	    sizeof (cinfo->transport));
546 	cinfo->flags = 0;
547 
548 	rdsv3_conn_info_set(cinfo->flags,
549 	    MUTEX_HELD(&conn->c_send_lock), SENDING);
550 
551 	/* XXX Future: return the state rather than these funky bits */
552 	rdsv3_conn_info_set(cinfo->flags,
553 	    atomic_get(&conn->c_state) == RDSV3_CONN_CONNECTING,
554 	    CONNECTING);
555 	rdsv3_conn_info_set(cinfo->flags,
556 	    atomic_get(&conn->c_state) == RDSV3_CONN_UP,
557 	    CONNECTED);
558 	return (1);
559 }
560 
561 static void
rdsv3_conn_info(struct rsock * sock,unsigned int len,struct rdsv3_info_iterator * iter,struct rdsv3_info_lengths * lens)562 rdsv3_conn_info(struct rsock *sock, unsigned int len,
563     struct rdsv3_info_iterator *iter, struct rdsv3_info_lengths *lens)
564 {
565 	rdsv3_for_each_conn_info(sock, len, iter, lens,
566 	    rdsv3_conn_info_visitor, sizeof (struct rds_info_connection));
567 }
568 
569 int
rdsv3_conn_init()570 rdsv3_conn_init()
571 {
572 	RDSV3_DPRINTF4("rdsv3_conn_init", "Enter");
573 
574 	rdsv3_conn_slab = kmem_cache_create("rdsv3_connection",
575 	    sizeof (struct rdsv3_connection), 0, rdsv3_conn_constructor,
576 	    rdsv3_conn_destructor, NULL, NULL, NULL, 0);
577 	if (!rdsv3_conn_slab) {
578 		RDSV3_DPRINTF2("rdsv3_conn_init",
579 		    "kmem_cache_create(rdsv3_conn_slab) failed");
580 		return (-ENOMEM);
581 	}
582 
583 	avl_create(&rdsv3_conn_hash, rdsv3_conn_compare,
584 	    sizeof (struct rdsv3_connection), offsetof(struct rdsv3_connection,
585 	    c_hash_node));
586 
587 	rw_init(&rdsv3_conn_lock, NULL, RW_DRIVER, NULL);
588 
589 	rdsv3_loop_init();
590 
591 	rdsv3_info_register_func(RDS_INFO_CONNECTIONS, rdsv3_conn_info);
592 	rdsv3_info_register_func(RDS_INFO_SEND_MESSAGES,
593 	    rdsv3_conn_message_info_send);
594 	rdsv3_info_register_func(RDS_INFO_RETRANS_MESSAGES,
595 	    rdsv3_conn_message_info_retrans);
596 
597 	RDSV3_DPRINTF4("rdsv3_conn_init", "Return");
598 
599 	return (0);
600 }
601 
602 void
rdsv3_conn_exit()603 rdsv3_conn_exit()
604 {
605 	RDSV3_DPRINTF4("rdsv3_conn_exit", "Enter");
606 
607 	rdsv3_loop_exit();
608 
609 	rw_destroy(&rdsv3_conn_lock);
610 	avl_destroy(&rdsv3_conn_hash);
611 
612 	ASSERT(rdsv3_conn_slab);
613 	kmem_cache_destroy(rdsv3_conn_slab);
614 
615 	RDSV3_DPRINTF4("rdsv3_conn_exit", "Return");
616 }
617 
618 /*
619  * Force a disconnect
620  */
621 void
rdsv3_conn_drop(struct rdsv3_connection * conn)622 rdsv3_conn_drop(struct rdsv3_connection *conn)
623 {
624 	conn->c_state = RDSV3_CONN_ERROR;
625 	rdsv3_queue_work(rdsv3_wq, &conn->c_down_w);
626 }
627