xref: /linux/net/rxrpc/conn_client.c (revision 08f3e0873ac203449465c2b8473d684e2f9f41d1)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Client connection-specific management code.
3  *
4  * Copyright (C) 2016, 2020 Red Hat, Inc. All Rights Reserved.
5  * Written by David Howells (dhowells@redhat.com)
6  *
7  * Client connections need to be cached for a little while after they've made a
8  * call so as to handle retransmitted DATA packets in case the server didn't
9  * receive the final ACK or terminating ABORT we sent it.
10  *
11  * There are flags of relevance to the cache:
12  *
13  *  (2) DONT_REUSE - The connection should be discarded as soon as possible and
14  *      should not be reused.  This is set when an exclusive connection is used
15  *      or a call ID counter overflows.
16  *
17  * The caching state may only be changed if the cache lock is held.
18  *
19  * There are two idle client connection expiry durations.  If the total number
20  * of connections is below the reap threshold, we use the normal duration; if
21  * it's above, we use the fast duration.
22  */
23 
24 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
25 
26 #include <linux/slab.h>
27 #include <linux/idr.h>
28 #include <linux/timer.h>
29 #include <linux/sched/signal.h>
30 
31 #include "ar-internal.h"
32 
33 __read_mostly unsigned int rxrpc_reap_client_connections = 900;
34 __read_mostly unsigned long rxrpc_conn_idle_client_expiry = 2 * 60 * HZ;
35 __read_mostly unsigned long rxrpc_conn_idle_client_fast_expiry = 2 * HZ;
36 
37 /*
38  * We use machine-unique IDs for our client connections.
39  */
40 DEFINE_IDR(rxrpc_client_conn_ids);
41 static DEFINE_SPINLOCK(rxrpc_conn_id_lock);
42 
43 /*
44  * Get a connection ID and epoch for a client connection from the global pool.
45  * The connection struct pointer is then recorded in the idr radix tree.  The
46  * epoch doesn't change until the client is rebooted (or, at least, unless the
47  * module is unloaded).
48  */
49 static int rxrpc_get_client_connection_id(struct rxrpc_connection *conn,
50 					  gfp_t gfp)
51 {
52 	struct rxrpc_net *rxnet = conn->params.local->rxnet;
53 	int id;
54 
55 	_enter("");
56 
57 	idr_preload(gfp);
58 	spin_lock(&rxrpc_conn_id_lock);
59 
60 	id = idr_alloc_cyclic(&rxrpc_client_conn_ids, conn,
61 			      1, 0x40000000, GFP_NOWAIT);
62 	if (id < 0)
63 		goto error;
64 
65 	spin_unlock(&rxrpc_conn_id_lock);
66 	idr_preload_end();
67 
68 	conn->proto.epoch = rxnet->epoch;
69 	conn->proto.cid = id << RXRPC_CIDSHIFT;
70 	set_bit(RXRPC_CONN_HAS_IDR, &conn->flags);
71 	_leave(" [CID %x]", conn->proto.cid);
72 	return 0;
73 
74 error:
75 	spin_unlock(&rxrpc_conn_id_lock);
76 	idr_preload_end();
77 	_leave(" = %d", id);
78 	return id;
79 }
80 
81 /*
82  * Release a connection ID for a client connection from the global pool.
83  */
84 static void rxrpc_put_client_connection_id(struct rxrpc_connection *conn)
85 {
86 	if (test_bit(RXRPC_CONN_HAS_IDR, &conn->flags)) {
87 		spin_lock(&rxrpc_conn_id_lock);
88 		idr_remove(&rxrpc_client_conn_ids,
89 			   conn->proto.cid >> RXRPC_CIDSHIFT);
90 		spin_unlock(&rxrpc_conn_id_lock);
91 	}
92 }
93 
94 /*
95  * Destroy the client connection ID tree.
96  */
97 void rxrpc_destroy_client_conn_ids(void)
98 {
99 	struct rxrpc_connection *conn;
100 	int id;
101 
102 	if (!idr_is_empty(&rxrpc_client_conn_ids)) {
103 		idr_for_each_entry(&rxrpc_client_conn_ids, conn, id) {
104 			pr_err("AF_RXRPC: Leaked client conn %p {%d}\n",
105 			       conn, atomic_read(&conn->usage));
106 		}
107 		BUG();
108 	}
109 
110 	idr_destroy(&rxrpc_client_conn_ids);
111 }
112 
113 /*
114  * Allocate a connection bundle.
115  */
116 static struct rxrpc_bundle *rxrpc_alloc_bundle(struct rxrpc_conn_parameters *cp,
117 					       gfp_t gfp)
118 {
119 	struct rxrpc_bundle *bundle;
120 
121 	bundle = kzalloc(sizeof(*bundle), gfp);
122 	if (bundle) {
123 		bundle->params = *cp;
124 		rxrpc_get_peer(bundle->params.peer);
125 		atomic_set(&bundle->usage, 1);
126 		spin_lock_init(&bundle->channel_lock);
127 		INIT_LIST_HEAD(&bundle->waiting_calls);
128 	}
129 	return bundle;
130 }
131 
132 struct rxrpc_bundle *rxrpc_get_bundle(struct rxrpc_bundle *bundle)
133 {
134 	atomic_inc(&bundle->usage);
135 	return bundle;
136 }
137 
138 void rxrpc_put_bundle(struct rxrpc_bundle *bundle)
139 {
140 	unsigned int d = bundle->debug_id;
141 	unsigned int u = atomic_dec_return(&bundle->usage);
142 
143 	_debug("PUT B=%x %u", d, u);
144 	if (u == 0) {
145 		rxrpc_put_peer(bundle->params.peer);
146 		kfree(bundle);
147 	}
148 }
149 
150 /*
151  * Allocate a client connection.
152  */
153 static struct rxrpc_connection *
154 rxrpc_alloc_client_connection(struct rxrpc_bundle *bundle, gfp_t gfp)
155 {
156 	struct rxrpc_connection *conn;
157 	struct rxrpc_net *rxnet = bundle->params.local->rxnet;
158 	int ret;
159 
160 	_enter("");
161 
162 	conn = rxrpc_alloc_connection(gfp);
163 	if (!conn) {
164 		_leave(" = -ENOMEM");
165 		return ERR_PTR(-ENOMEM);
166 	}
167 
168 	atomic_set(&conn->usage, 1);
169 	conn->bundle		= bundle;
170 	conn->params		= bundle->params;
171 	conn->out_clientflag	= RXRPC_CLIENT_INITIATED;
172 	conn->state		= RXRPC_CONN_CLIENT;
173 	conn->service_id	= conn->params.service_id;
174 
175 	ret = rxrpc_get_client_connection_id(conn, gfp);
176 	if (ret < 0)
177 		goto error_0;
178 
179 	ret = rxrpc_init_client_conn_security(conn);
180 	if (ret < 0)
181 		goto error_1;
182 
183 	atomic_inc(&rxnet->nr_conns);
184 	write_lock(&rxnet->conn_lock);
185 	list_add_tail(&conn->proc_link, &rxnet->conn_proc_list);
186 	write_unlock(&rxnet->conn_lock);
187 
188 	rxrpc_get_bundle(bundle);
189 	rxrpc_get_peer(conn->params.peer);
190 	rxrpc_get_local(conn->params.local);
191 	key_get(conn->params.key);
192 
193 	trace_rxrpc_conn(conn->debug_id, rxrpc_conn_new_client,
194 			 atomic_read(&conn->usage),
195 			 __builtin_return_address(0));
196 
197 	atomic_inc(&rxnet->nr_client_conns);
198 	trace_rxrpc_client(conn, -1, rxrpc_client_alloc);
199 	_leave(" = %p", conn);
200 	return conn;
201 
202 error_1:
203 	rxrpc_put_client_connection_id(conn);
204 error_0:
205 	kfree(conn);
206 	_leave(" = %d", ret);
207 	return ERR_PTR(ret);
208 }
209 
210 /*
211  * Determine if a connection may be reused.
212  */
213 static bool rxrpc_may_reuse_conn(struct rxrpc_connection *conn)
214 {
215 	struct rxrpc_net *rxnet;
216 	int id_cursor, id, distance, limit;
217 
218 	if (!conn)
219 		goto dont_reuse;
220 
221 	rxnet = conn->params.local->rxnet;
222 	if (test_bit(RXRPC_CONN_DONT_REUSE, &conn->flags))
223 		goto dont_reuse;
224 
225 	if (conn->state != RXRPC_CONN_CLIENT ||
226 	    conn->proto.epoch != rxnet->epoch)
227 		goto mark_dont_reuse;
228 
229 	/* The IDR tree gets very expensive on memory if the connection IDs are
230 	 * widely scattered throughout the number space, so we shall want to
231 	 * kill off connections that, say, have an ID more than about four
232 	 * times the maximum number of client conns away from the current
233 	 * allocation point to try and keep the IDs concentrated.
234 	 */
235 	id_cursor = idr_get_cursor(&rxrpc_client_conn_ids);
236 	id = conn->proto.cid >> RXRPC_CIDSHIFT;
237 	distance = id - id_cursor;
238 	if (distance < 0)
239 		distance = -distance;
240 	limit = max_t(unsigned long, atomic_read(&rxnet->nr_conns) * 4, 1024);
241 	if (distance > limit)
242 		goto mark_dont_reuse;
243 
244 	return true;
245 
246 mark_dont_reuse:
247 	set_bit(RXRPC_CONN_DONT_REUSE, &conn->flags);
248 dont_reuse:
249 	return false;
250 }
251 
252 /*
253  * Look up the conn bundle that matches the connection parameters, adding it if
254  * it doesn't yet exist.
255  */
256 static struct rxrpc_bundle *rxrpc_look_up_bundle(struct rxrpc_conn_parameters *cp,
257 						 gfp_t gfp)
258 {
259 	static atomic_t rxrpc_bundle_id;
260 	struct rxrpc_bundle *bundle, *candidate;
261 	struct rxrpc_local *local = cp->local;
262 	struct rb_node *p, **pp, *parent;
263 	long diff;
264 
265 	_enter("{%px,%x,%u,%u}",
266 	       cp->peer, key_serial(cp->key), cp->security_level, cp->upgrade);
267 
268 	if (cp->exclusive)
269 		return rxrpc_alloc_bundle(cp, gfp);
270 
271 	/* First, see if the bundle is already there. */
272 	_debug("search 1");
273 	spin_lock(&local->client_bundles_lock);
274 	p = local->client_bundles.rb_node;
275 	while (p) {
276 		bundle = rb_entry(p, struct rxrpc_bundle, local_node);
277 
278 #define cmp(X) ((long)bundle->params.X - (long)cp->X)
279 		diff = (cmp(peer) ?:
280 			cmp(key) ?:
281 			cmp(security_level) ?:
282 			cmp(upgrade));
283 #undef cmp
284 		if (diff < 0)
285 			p = p->rb_left;
286 		else if (diff > 0)
287 			p = p->rb_right;
288 		else
289 			goto found_bundle;
290 	}
291 	spin_unlock(&local->client_bundles_lock);
292 	_debug("not found");
293 
294 	/* It wasn't.  We need to add one. */
295 	candidate = rxrpc_alloc_bundle(cp, gfp);
296 	if (!candidate)
297 		return NULL;
298 
299 	_debug("search 2");
300 	spin_lock(&local->client_bundles_lock);
301 	pp = &local->client_bundles.rb_node;
302 	parent = NULL;
303 	while (*pp) {
304 		parent = *pp;
305 		bundle = rb_entry(parent, struct rxrpc_bundle, local_node);
306 
307 #define cmp(X) ((long)bundle->params.X - (long)cp->X)
308 		diff = (cmp(peer) ?:
309 			cmp(key) ?:
310 			cmp(security_level) ?:
311 			cmp(upgrade));
312 #undef cmp
313 		if (diff < 0)
314 			pp = &(*pp)->rb_left;
315 		else if (diff > 0)
316 			pp = &(*pp)->rb_right;
317 		else
318 			goto found_bundle_free;
319 	}
320 
321 	_debug("new bundle");
322 	candidate->debug_id = atomic_inc_return(&rxrpc_bundle_id);
323 	rb_link_node(&candidate->local_node, parent, pp);
324 	rb_insert_color(&candidate->local_node, &local->client_bundles);
325 	rxrpc_get_bundle(candidate);
326 	spin_unlock(&local->client_bundles_lock);
327 	_leave(" = %u [new]", candidate->debug_id);
328 	return candidate;
329 
330 found_bundle_free:
331 	kfree(candidate);
332 found_bundle:
333 	rxrpc_get_bundle(bundle);
334 	spin_unlock(&local->client_bundles_lock);
335 	_leave(" = %u [found]", bundle->debug_id);
336 	return bundle;
337 }
338 
339 /*
340  * Create or find a client bundle to use for a call.
341  *
342  * If we return with a connection, the call will be on its waiting list.  It's
343  * left to the caller to assign a channel and wake up the call.
344  */
345 static struct rxrpc_bundle *rxrpc_prep_call(struct rxrpc_sock *rx,
346 					    struct rxrpc_call *call,
347 					    struct rxrpc_conn_parameters *cp,
348 					    struct sockaddr_rxrpc *srx,
349 					    gfp_t gfp)
350 {
351 	struct rxrpc_bundle *bundle;
352 
353 	_enter("{%d,%lx},", call->debug_id, call->user_call_ID);
354 
355 	cp->peer = rxrpc_lookup_peer(rx, cp->local, srx, gfp);
356 	if (!cp->peer)
357 		goto error;
358 
359 	call->cong_cwnd = cp->peer->cong_cwnd;
360 	if (call->cong_cwnd >= call->cong_ssthresh)
361 		call->cong_mode = RXRPC_CALL_CONGEST_AVOIDANCE;
362 	else
363 		call->cong_mode = RXRPC_CALL_SLOW_START;
364 	if (cp->upgrade)
365 		__set_bit(RXRPC_CALL_UPGRADE, &call->flags);
366 
367 	/* Find the client connection bundle. */
368 	bundle = rxrpc_look_up_bundle(cp, gfp);
369 	if (!bundle)
370 		goto error;
371 
372 	/* Get this call queued.  Someone else may activate it whilst we're
373 	 * lining up a new connection, but that's fine.
374 	 */
375 	spin_lock(&bundle->channel_lock);
376 	list_add_tail(&call->chan_wait_link, &bundle->waiting_calls);
377 	spin_unlock(&bundle->channel_lock);
378 
379 	_leave(" = [B=%x]", bundle->debug_id);
380 	return bundle;
381 
382 error:
383 	_leave(" = -ENOMEM");
384 	return ERR_PTR(-ENOMEM);
385 }
386 
387 /*
388  * Allocate a new connection and add it into a bundle.
389  */
390 static void rxrpc_add_conn_to_bundle(struct rxrpc_bundle *bundle, gfp_t gfp)
391 	__releases(bundle->channel_lock)
392 {
393 	struct rxrpc_connection *candidate = NULL, *old = NULL;
394 	bool conflict;
395 	int i;
396 
397 	_enter("");
398 
399 	conflict = bundle->alloc_conn;
400 	if (!conflict)
401 		bundle->alloc_conn = true;
402 	spin_unlock(&bundle->channel_lock);
403 	if (conflict) {
404 		_leave(" [conf]");
405 		return;
406 	}
407 
408 	candidate = rxrpc_alloc_client_connection(bundle, gfp);
409 
410 	spin_lock(&bundle->channel_lock);
411 	bundle->alloc_conn = false;
412 
413 	if (IS_ERR(candidate)) {
414 		bundle->alloc_error = PTR_ERR(candidate);
415 		spin_unlock(&bundle->channel_lock);
416 		_leave(" [err %ld]", PTR_ERR(candidate));
417 		return;
418 	}
419 
420 	bundle->alloc_error = 0;
421 
422 	for (i = 0; i < ARRAY_SIZE(bundle->conns); i++) {
423 		unsigned int shift = i * RXRPC_MAXCALLS;
424 		int j;
425 
426 		old = bundle->conns[i];
427 		if (!rxrpc_may_reuse_conn(old)) {
428 			if (old)
429 				trace_rxrpc_client(old, -1, rxrpc_client_replace);
430 			candidate->bundle_shift = shift;
431 			bundle->conns[i] = candidate;
432 			for (j = 0; j < RXRPC_MAXCALLS; j++)
433 				set_bit(shift + j, &bundle->avail_chans);
434 			candidate = NULL;
435 			break;
436 		}
437 
438 		old = NULL;
439 	}
440 
441 	spin_unlock(&bundle->channel_lock);
442 
443 	if (candidate) {
444 		_debug("discard C=%x", candidate->debug_id);
445 		trace_rxrpc_client(candidate, -1, rxrpc_client_duplicate);
446 		rxrpc_put_connection(candidate);
447 	}
448 
449 	rxrpc_put_connection(old);
450 	_leave("");
451 }
452 
453 /*
454  * Add a connection to a bundle if there are no usable connections or we have
455  * connections waiting for extra capacity.
456  */
457 static void rxrpc_maybe_add_conn(struct rxrpc_bundle *bundle, gfp_t gfp)
458 {
459 	struct rxrpc_call *call;
460 	int i, usable;
461 
462 	_enter("");
463 
464 	spin_lock(&bundle->channel_lock);
465 
466 	/* See if there are any usable connections. */
467 	usable = 0;
468 	for (i = 0; i < ARRAY_SIZE(bundle->conns); i++)
469 		if (rxrpc_may_reuse_conn(bundle->conns[i]))
470 			usable++;
471 
472 	if (!usable && !list_empty(&bundle->waiting_calls)) {
473 		call = list_first_entry(&bundle->waiting_calls,
474 					struct rxrpc_call, chan_wait_link);
475 		if (test_bit(RXRPC_CALL_UPGRADE, &call->flags))
476 			bundle->try_upgrade = true;
477 	}
478 
479 	if (!usable)
480 		goto alloc_conn;
481 
482 	if (!bundle->avail_chans &&
483 	    !bundle->try_upgrade &&
484 	    !list_empty(&bundle->waiting_calls) &&
485 	    usable < ARRAY_SIZE(bundle->conns))
486 		goto alloc_conn;
487 
488 	spin_unlock(&bundle->channel_lock);
489 	_leave("");
490 	return;
491 
492 alloc_conn:
493 	return rxrpc_add_conn_to_bundle(bundle, gfp);
494 }
495 
496 /*
497  * Assign a channel to the call at the front of the queue and wake the call up.
498  * We don't increment the callNumber counter until this number has been exposed
499  * to the world.
500  */
501 static void rxrpc_activate_one_channel(struct rxrpc_connection *conn,
502 				       unsigned int channel)
503 {
504 	struct rxrpc_channel *chan = &conn->channels[channel];
505 	struct rxrpc_bundle *bundle = conn->bundle;
506 	struct rxrpc_call *call = list_entry(bundle->waiting_calls.next,
507 					     struct rxrpc_call, chan_wait_link);
508 	u32 call_id = chan->call_counter + 1;
509 
510 	_enter("C=%x,%u", conn->debug_id, channel);
511 
512 	trace_rxrpc_client(conn, channel, rxrpc_client_chan_activate);
513 
514 	/* Cancel the final ACK on the previous call if it hasn't been sent yet
515 	 * as the DATA packet will implicitly ACK it.
516 	 */
517 	clear_bit(RXRPC_CONN_FINAL_ACK_0 + channel, &conn->flags);
518 	clear_bit(conn->bundle_shift + channel, &bundle->avail_chans);
519 
520 	rxrpc_see_call(call);
521 	list_del_init(&call->chan_wait_link);
522 	call->peer	= rxrpc_get_peer(conn->params.peer);
523 	call->conn	= rxrpc_get_connection(conn);
524 	call->cid	= conn->proto.cid | channel;
525 	call->call_id	= call_id;
526 	call->security	= conn->security;
527 	call->security_ix = conn->security_ix;
528 	call->service_id = conn->service_id;
529 
530 	trace_rxrpc_connect_call(call);
531 	_net("CONNECT call %08x:%08x as call %d on conn %d",
532 	     call->cid, call->call_id, call->debug_id, conn->debug_id);
533 
534 	write_lock_bh(&call->state_lock);
535 	call->state = RXRPC_CALL_CLIENT_SEND_REQUEST;
536 	write_unlock_bh(&call->state_lock);
537 
538 	/* Paired with the read barrier in rxrpc_connect_call().  This orders
539 	 * cid and epoch in the connection wrt to call_id without the need to
540 	 * take the channel_lock.
541 	 *
542 	 * We provisionally assign a callNumber at this point, but we don't
543 	 * confirm it until the call is about to be exposed.
544 	 *
545 	 * TODO: Pair with a barrier in the data_ready handler when that looks
546 	 * at the call ID through a connection channel.
547 	 */
548 	smp_wmb();
549 
550 	chan->call_id		= call_id;
551 	chan->call_debug_id	= call->debug_id;
552 	rcu_assign_pointer(chan->call, call);
553 	wake_up(&call->waitq);
554 }
555 
556 /*
557  * Remove a connection from the idle list if it's on it.
558  */
559 static void rxrpc_unidle_conn(struct rxrpc_bundle *bundle, struct rxrpc_connection *conn)
560 {
561 	struct rxrpc_net *rxnet = bundle->params.local->rxnet;
562 	bool drop_ref;
563 
564 	if (!list_empty(&conn->cache_link)) {
565 		drop_ref = false;
566 		spin_lock(&rxnet->client_conn_cache_lock);
567 		if (!list_empty(&conn->cache_link)) {
568 			list_del_init(&conn->cache_link);
569 			drop_ref = true;
570 		}
571 		spin_unlock(&rxnet->client_conn_cache_lock);
572 		if (drop_ref)
573 			rxrpc_put_connection(conn);
574 	}
575 }
576 
577 /*
578  * Assign channels and callNumbers to waiting calls with channel_lock
579  * held by caller.
580  */
581 static void rxrpc_activate_channels_locked(struct rxrpc_bundle *bundle)
582 {
583 	struct rxrpc_connection *conn;
584 	unsigned long avail, mask;
585 	unsigned int channel, slot;
586 
587 	if (bundle->try_upgrade)
588 		mask = 1;
589 	else
590 		mask = ULONG_MAX;
591 
592 	while (!list_empty(&bundle->waiting_calls)) {
593 		avail = bundle->avail_chans & mask;
594 		if (!avail)
595 			break;
596 		channel = __ffs(avail);
597 		clear_bit(channel, &bundle->avail_chans);
598 
599 		slot = channel / RXRPC_MAXCALLS;
600 		conn = bundle->conns[slot];
601 		if (!conn)
602 			break;
603 
604 		if (bundle->try_upgrade)
605 			set_bit(RXRPC_CONN_PROBING_FOR_UPGRADE, &conn->flags);
606 		rxrpc_unidle_conn(bundle, conn);
607 
608 		channel &= (RXRPC_MAXCALLS - 1);
609 		conn->act_chans	|= 1 << channel;
610 		rxrpc_activate_one_channel(conn, channel);
611 	}
612 }
613 
614 /*
615  * Assign channels and callNumbers to waiting calls.
616  */
617 static void rxrpc_activate_channels(struct rxrpc_bundle *bundle)
618 {
619 	_enter("B=%x", bundle->debug_id);
620 
621 	trace_rxrpc_client(NULL, -1, rxrpc_client_activate_chans);
622 
623 	if (!bundle->avail_chans)
624 		return;
625 
626 	spin_lock(&bundle->channel_lock);
627 	rxrpc_activate_channels_locked(bundle);
628 	spin_unlock(&bundle->channel_lock);
629 	_leave("");
630 }
631 
632 /*
633  * Wait for a callNumber and a channel to be granted to a call.
634  */
635 static int rxrpc_wait_for_channel(struct rxrpc_bundle *bundle,
636 				  struct rxrpc_call *call, gfp_t gfp)
637 {
638 	DECLARE_WAITQUEUE(myself, current);
639 	int ret = 0;
640 
641 	_enter("%d", call->debug_id);
642 
643 	if (!gfpflags_allow_blocking(gfp)) {
644 		rxrpc_maybe_add_conn(bundle, gfp);
645 		rxrpc_activate_channels(bundle);
646 		ret = bundle->alloc_error ?: -EAGAIN;
647 		goto out;
648 	}
649 
650 	add_wait_queue_exclusive(&call->waitq, &myself);
651 	for (;;) {
652 		rxrpc_maybe_add_conn(bundle, gfp);
653 		rxrpc_activate_channels(bundle);
654 		ret = bundle->alloc_error;
655 		if (ret < 0)
656 			break;
657 
658 		switch (call->interruptibility) {
659 		case RXRPC_INTERRUPTIBLE:
660 		case RXRPC_PREINTERRUPTIBLE:
661 			set_current_state(TASK_INTERRUPTIBLE);
662 			break;
663 		case RXRPC_UNINTERRUPTIBLE:
664 		default:
665 			set_current_state(TASK_UNINTERRUPTIBLE);
666 			break;
667 		}
668 		if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_AWAIT_CONN)
669 			break;
670 		if ((call->interruptibility == RXRPC_INTERRUPTIBLE ||
671 		     call->interruptibility == RXRPC_PREINTERRUPTIBLE) &&
672 		    signal_pending(current)) {
673 			ret = -ERESTARTSYS;
674 			break;
675 		}
676 		schedule();
677 	}
678 	remove_wait_queue(&call->waitq, &myself);
679 	__set_current_state(TASK_RUNNING);
680 
681 out:
682 	_leave(" = %d", ret);
683 	return ret;
684 }
685 
686 /*
687  * find a connection for a call
688  * - called in process context with IRQs enabled
689  */
690 int rxrpc_connect_call(struct rxrpc_sock *rx,
691 		       struct rxrpc_call *call,
692 		       struct rxrpc_conn_parameters *cp,
693 		       struct sockaddr_rxrpc *srx,
694 		       gfp_t gfp)
695 {
696 	struct rxrpc_bundle *bundle;
697 	struct rxrpc_net *rxnet = cp->local->rxnet;
698 	int ret = 0;
699 
700 	_enter("{%d,%lx},", call->debug_id, call->user_call_ID);
701 
702 	rxrpc_discard_expired_client_conns(&rxnet->client_conn_reaper);
703 
704 	bundle = rxrpc_prep_call(rx, call, cp, srx, gfp);
705 	if (IS_ERR(bundle)) {
706 		ret = PTR_ERR(bundle);
707 		goto out;
708 	}
709 
710 	if (call->state == RXRPC_CALL_CLIENT_AWAIT_CONN) {
711 		ret = rxrpc_wait_for_channel(bundle, call, gfp);
712 		if (ret < 0)
713 			goto wait_failed;
714 	}
715 
716 granted_channel:
717 	/* Paired with the write barrier in rxrpc_activate_one_channel(). */
718 	smp_rmb();
719 
720 out_put_bundle:
721 	rxrpc_put_bundle(bundle);
722 out:
723 	_leave(" = %d", ret);
724 	return ret;
725 
726 wait_failed:
727 	spin_lock(&bundle->channel_lock);
728 	list_del_init(&call->chan_wait_link);
729 	spin_unlock(&bundle->channel_lock);
730 
731 	if (call->state != RXRPC_CALL_CLIENT_AWAIT_CONN) {
732 		ret = 0;
733 		goto granted_channel;
734 	}
735 
736 	trace_rxrpc_client(call->conn, ret, rxrpc_client_chan_wait_failed);
737 	rxrpc_set_call_completion(call, RXRPC_CALL_LOCAL_ERROR, 0, ret);
738 	rxrpc_disconnect_client_call(bundle, call);
739 	goto out_put_bundle;
740 }
741 
742 /*
743  * Note that a call, and thus a connection, is about to be exposed to the
744  * world.
745  */
746 void rxrpc_expose_client_call(struct rxrpc_call *call)
747 {
748 	unsigned int channel = call->cid & RXRPC_CHANNELMASK;
749 	struct rxrpc_connection *conn = call->conn;
750 	struct rxrpc_channel *chan = &conn->channels[channel];
751 
752 	if (!test_and_set_bit(RXRPC_CALL_EXPOSED, &call->flags)) {
753 		/* Mark the call ID as being used.  If the callNumber counter
754 		 * exceeds ~2 billion, we kill the connection after its
755 		 * outstanding calls have finished so that the counter doesn't
756 		 * wrap.
757 		 */
758 		chan->call_counter++;
759 		if (chan->call_counter >= INT_MAX)
760 			set_bit(RXRPC_CONN_DONT_REUSE, &conn->flags);
761 		trace_rxrpc_client(conn, channel, rxrpc_client_exposed);
762 	}
763 }
764 
765 /*
766  * Set the reap timer.
767  */
768 static void rxrpc_set_client_reap_timer(struct rxrpc_net *rxnet)
769 {
770 	if (!rxnet->kill_all_client_conns) {
771 		unsigned long now = jiffies;
772 		unsigned long reap_at = now + rxrpc_conn_idle_client_expiry;
773 
774 		if (rxnet->live)
775 			timer_reduce(&rxnet->client_conn_reap_timer, reap_at);
776 	}
777 }
778 
779 /*
780  * Disconnect a client call.
781  */
782 void rxrpc_disconnect_client_call(struct rxrpc_bundle *bundle, struct rxrpc_call *call)
783 {
784 	struct rxrpc_connection *conn;
785 	struct rxrpc_channel *chan = NULL;
786 	struct rxrpc_net *rxnet = bundle->params.local->rxnet;
787 	unsigned int channel;
788 	bool may_reuse;
789 	u32 cid;
790 
791 	_enter("c=%x", call->debug_id);
792 
793 	spin_lock(&bundle->channel_lock);
794 	set_bit(RXRPC_CALL_DISCONNECTED, &call->flags);
795 
796 	/* Calls that have never actually been assigned a channel can simply be
797 	 * discarded.
798 	 */
799 	conn = call->conn;
800 	if (!conn) {
801 		_debug("call is waiting");
802 		ASSERTCMP(call->call_id, ==, 0);
803 		ASSERT(!test_bit(RXRPC_CALL_EXPOSED, &call->flags));
804 		list_del_init(&call->chan_wait_link);
805 		goto out;
806 	}
807 
808 	cid = call->cid;
809 	channel = cid & RXRPC_CHANNELMASK;
810 	chan = &conn->channels[channel];
811 	trace_rxrpc_client(conn, channel, rxrpc_client_chan_disconnect);
812 
813 	if (rcu_access_pointer(chan->call) != call) {
814 		spin_unlock(&bundle->channel_lock);
815 		BUG();
816 	}
817 
818 	may_reuse = rxrpc_may_reuse_conn(conn);
819 
820 	/* If a client call was exposed to the world, we save the result for
821 	 * retransmission.
822 	 *
823 	 * We use a barrier here so that the call number and abort code can be
824 	 * read without needing to take a lock.
825 	 *
826 	 * TODO: Make the incoming packet handler check this and handle
827 	 * terminal retransmission without requiring access to the call.
828 	 */
829 	if (test_bit(RXRPC_CALL_EXPOSED, &call->flags)) {
830 		_debug("exposed %u,%u", call->call_id, call->abort_code);
831 		__rxrpc_disconnect_call(conn, call);
832 
833 		if (test_and_clear_bit(RXRPC_CONN_PROBING_FOR_UPGRADE, &conn->flags)) {
834 			trace_rxrpc_client(conn, channel, rxrpc_client_to_active);
835 			bundle->try_upgrade = false;
836 			if (may_reuse)
837 				rxrpc_activate_channels_locked(bundle);
838 		}
839 
840 	}
841 
842 	/* See if we can pass the channel directly to another call. */
843 	if (may_reuse && !list_empty(&bundle->waiting_calls)) {
844 		trace_rxrpc_client(conn, channel, rxrpc_client_chan_pass);
845 		rxrpc_activate_one_channel(conn, channel);
846 		goto out;
847 	}
848 
849 	/* Schedule the final ACK to be transmitted in a short while so that it
850 	 * can be skipped if we find a follow-on call.  The first DATA packet
851 	 * of the follow on call will implicitly ACK this call.
852 	 */
853 	if (call->completion == RXRPC_CALL_SUCCEEDED &&
854 	    test_bit(RXRPC_CALL_EXPOSED, &call->flags)) {
855 		unsigned long final_ack_at = jiffies + 2;
856 
857 		WRITE_ONCE(chan->final_ack_at, final_ack_at);
858 		smp_wmb(); /* vs rxrpc_process_delayed_final_acks() */
859 		set_bit(RXRPC_CONN_FINAL_ACK_0 + channel, &conn->flags);
860 		rxrpc_reduce_conn_timer(conn, final_ack_at);
861 	}
862 
863 	/* Deactivate the channel. */
864 	rcu_assign_pointer(chan->call, NULL);
865 	set_bit(conn->bundle_shift + channel, &conn->bundle->avail_chans);
866 	conn->act_chans	&= ~(1 << channel);
867 
868 	/* If no channels remain active, then put the connection on the idle
869 	 * list for a short while.  Give it a ref to stop it going away if it
870 	 * becomes unbundled.
871 	 */
872 	if (!conn->act_chans) {
873 		trace_rxrpc_client(conn, channel, rxrpc_client_to_idle);
874 		conn->idle_timestamp = jiffies;
875 
876 		rxrpc_get_connection(conn);
877 		spin_lock(&rxnet->client_conn_cache_lock);
878 		list_move_tail(&conn->cache_link, &rxnet->idle_client_conns);
879 		spin_unlock(&rxnet->client_conn_cache_lock);
880 
881 		rxrpc_set_client_reap_timer(rxnet);
882 	}
883 
884 out:
885 	spin_unlock(&bundle->channel_lock);
886 	_leave("");
887 	return;
888 }
889 
890 /*
891  * Remove a connection from a bundle.
892  */
893 static void rxrpc_unbundle_conn(struct rxrpc_connection *conn)
894 {
895 	struct rxrpc_bundle *bundle = conn->bundle;
896 	struct rxrpc_local *local = bundle->params.local;
897 	unsigned int bindex;
898 	bool need_drop = false, need_put = false;
899 	int i;
900 
901 	_enter("C=%x", conn->debug_id);
902 
903 	if (conn->flags & RXRPC_CONN_FINAL_ACK_MASK)
904 		rxrpc_process_delayed_final_acks(conn, true);
905 
906 	spin_lock(&bundle->channel_lock);
907 	bindex = conn->bundle_shift / RXRPC_MAXCALLS;
908 	if (bundle->conns[bindex] == conn) {
909 		_debug("clear slot %u", bindex);
910 		bundle->conns[bindex] = NULL;
911 		for (i = 0; i < RXRPC_MAXCALLS; i++)
912 			clear_bit(conn->bundle_shift + i, &bundle->avail_chans);
913 		need_drop = true;
914 	}
915 	spin_unlock(&bundle->channel_lock);
916 
917 	/* If there are no more connections, remove the bundle */
918 	if (!bundle->avail_chans) {
919 		_debug("maybe unbundle");
920 		spin_lock(&local->client_bundles_lock);
921 
922 		for (i = 0; i < ARRAY_SIZE(bundle->conns); i++)
923 			if (bundle->conns[i])
924 				break;
925 		if (i == ARRAY_SIZE(bundle->conns) && !bundle->params.exclusive) {
926 			_debug("erase bundle");
927 			rb_erase(&bundle->local_node, &local->client_bundles);
928 			need_put = true;
929 		}
930 
931 		spin_unlock(&local->client_bundles_lock);
932 		if (need_put)
933 			rxrpc_put_bundle(bundle);
934 	}
935 
936 	if (need_drop)
937 		rxrpc_put_connection(conn);
938 	_leave("");
939 }
940 
941 /*
942  * Clean up a dead client connection.
943  */
944 static void rxrpc_kill_client_conn(struct rxrpc_connection *conn)
945 {
946 	struct rxrpc_local *local = conn->params.local;
947 	struct rxrpc_net *rxnet = local->rxnet;
948 
949 	_enter("C=%x", conn->debug_id);
950 
951 	trace_rxrpc_client(conn, -1, rxrpc_client_cleanup);
952 	atomic_dec(&rxnet->nr_client_conns);
953 
954 	rxrpc_put_client_connection_id(conn);
955 	rxrpc_kill_connection(conn);
956 }
957 
958 /*
959  * Clean up a dead client connections.
960  */
961 void rxrpc_put_client_conn(struct rxrpc_connection *conn)
962 {
963 	const void *here = __builtin_return_address(0);
964 	unsigned int debug_id = conn->debug_id;
965 	int n;
966 
967 	n = atomic_dec_return(&conn->usage);
968 	trace_rxrpc_conn(debug_id, rxrpc_conn_put_client, n, here);
969 	if (n <= 0) {
970 		ASSERTCMP(n, >=, 0);
971 		rxrpc_kill_client_conn(conn);
972 	}
973 }
974 
975 /*
976  * Discard expired client connections from the idle list.  Each conn in the
977  * idle list has been exposed and holds an extra ref because of that.
978  *
979  * This may be called from conn setup or from a work item so cannot be
980  * considered non-reentrant.
981  */
982 void rxrpc_discard_expired_client_conns(struct work_struct *work)
983 {
984 	struct rxrpc_connection *conn;
985 	struct rxrpc_net *rxnet =
986 		container_of(work, struct rxrpc_net, client_conn_reaper);
987 	unsigned long expiry, conn_expires_at, now;
988 	unsigned int nr_conns;
989 
990 	_enter("");
991 
992 	if (list_empty(&rxnet->idle_client_conns)) {
993 		_leave(" [empty]");
994 		return;
995 	}
996 
997 	/* Don't double up on the discarding */
998 	if (!spin_trylock(&rxnet->client_conn_discard_lock)) {
999 		_leave(" [already]");
1000 		return;
1001 	}
1002 
1003 	/* We keep an estimate of what the number of conns ought to be after
1004 	 * we've discarded some so that we don't overdo the discarding.
1005 	 */
1006 	nr_conns = atomic_read(&rxnet->nr_client_conns);
1007 
1008 next:
1009 	spin_lock(&rxnet->client_conn_cache_lock);
1010 
1011 	if (list_empty(&rxnet->idle_client_conns))
1012 		goto out;
1013 
1014 	conn = list_entry(rxnet->idle_client_conns.next,
1015 			  struct rxrpc_connection, cache_link);
1016 
1017 	if (!rxnet->kill_all_client_conns) {
1018 		/* If the number of connections is over the reap limit, we
1019 		 * expedite discard by reducing the expiry timeout.  We must,
1020 		 * however, have at least a short grace period to be able to do
1021 		 * final-ACK or ABORT retransmission.
1022 		 */
1023 		expiry = rxrpc_conn_idle_client_expiry;
1024 		if (nr_conns > rxrpc_reap_client_connections)
1025 			expiry = rxrpc_conn_idle_client_fast_expiry;
1026 		if (conn->params.local->service_closed)
1027 			expiry = rxrpc_closed_conn_expiry * HZ;
1028 
1029 		conn_expires_at = conn->idle_timestamp + expiry;
1030 
1031 		now = READ_ONCE(jiffies);
1032 		if (time_after(conn_expires_at, now))
1033 			goto not_yet_expired;
1034 	}
1035 
1036 	trace_rxrpc_client(conn, -1, rxrpc_client_discard);
1037 	list_del_init(&conn->cache_link);
1038 
1039 	spin_unlock(&rxnet->client_conn_cache_lock);
1040 
1041 	rxrpc_unbundle_conn(conn);
1042 	rxrpc_put_connection(conn); /* Drop the ->cache_link ref */
1043 
1044 	nr_conns--;
1045 	goto next;
1046 
1047 not_yet_expired:
1048 	/* The connection at the front of the queue hasn't yet expired, so
1049 	 * schedule the work item for that point if we discarded something.
1050 	 *
1051 	 * We don't worry if the work item is already scheduled - it can look
1052 	 * after rescheduling itself at a later time.  We could cancel it, but
1053 	 * then things get messier.
1054 	 */
1055 	_debug("not yet");
1056 	if (!rxnet->kill_all_client_conns)
1057 		timer_reduce(&rxnet->client_conn_reap_timer, conn_expires_at);
1058 
1059 out:
1060 	spin_unlock(&rxnet->client_conn_cache_lock);
1061 	spin_unlock(&rxnet->client_conn_discard_lock);
1062 	_leave("");
1063 }
1064 
1065 /*
1066  * Preemptively destroy all the client connection records rather than waiting
1067  * for them to time out
1068  */
1069 void rxrpc_destroy_all_client_connections(struct rxrpc_net *rxnet)
1070 {
1071 	_enter("");
1072 
1073 	spin_lock(&rxnet->client_conn_cache_lock);
1074 	rxnet->kill_all_client_conns = true;
1075 	spin_unlock(&rxnet->client_conn_cache_lock);
1076 
1077 	del_timer_sync(&rxnet->client_conn_reap_timer);
1078 
1079 	if (!rxrpc_queue_work(&rxnet->client_conn_reaper))
1080 		_debug("destroy: queue failed");
1081 
1082 	_leave("");
1083 }
1084 
1085 /*
1086  * Clean up the client connections on a local endpoint.
1087  */
1088 void rxrpc_clean_up_local_conns(struct rxrpc_local *local)
1089 {
1090 	struct rxrpc_connection *conn, *tmp;
1091 	struct rxrpc_net *rxnet = local->rxnet;
1092 	LIST_HEAD(graveyard);
1093 
1094 	_enter("");
1095 
1096 	spin_lock(&rxnet->client_conn_cache_lock);
1097 
1098 	list_for_each_entry_safe(conn, tmp, &rxnet->idle_client_conns,
1099 				 cache_link) {
1100 		if (conn->params.local == local) {
1101 			trace_rxrpc_client(conn, -1, rxrpc_client_discard);
1102 			list_move(&conn->cache_link, &graveyard);
1103 		}
1104 	}
1105 
1106 	spin_unlock(&rxnet->client_conn_cache_lock);
1107 
1108 	while (!list_empty(&graveyard)) {
1109 		conn = list_entry(graveyard.next,
1110 				  struct rxrpc_connection, cache_link);
1111 		list_del_init(&conn->cache_link);
1112 		rxrpc_unbundle_conn(conn);
1113 		rxrpc_put_connection(conn);
1114 	}
1115 
1116 	_leave(" [culled]");
1117 }
1118