xref: /openbmc/linux/fs/dlm/lowcomms.c (revision d9fd5a71)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /******************************************************************************
3 *******************************************************************************
4 **
5 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
6 **  Copyright (C) 2004-2009 Red Hat, Inc.  All rights reserved.
7 **
8 **
9 *******************************************************************************
10 ******************************************************************************/
11 
12 /*
13  * lowcomms.c
14  *
15  * This is the "low-level" comms layer.
16  *
17  * It is responsible for sending/receiving messages
18  * from other nodes in the cluster.
19  *
20  * Cluster nodes are referred to by their nodeids. nodeids are
21  * simply 32 bit numbers to the locking module - if they need to
22  * be expanded for the cluster infrastructure then that is its
23  * responsibility. It is this layer's
24  * responsibility to resolve these into IP address or
25  * whatever it needs for inter-node communication.
26  *
27  * The comms level is two kernel threads that deal mainly with
28  * the receiving of messages from other nodes and passing them
29  * up to the mid-level comms layer (which understands the
30  * message format) for execution by the locking core, and
31  * a send thread which does all the setting up of connections
32  * to remote nodes and the sending of data. Threads are not allowed
33  * to send their own data because it may cause them to wait in times
34  * of high load. Also, this way, the sending thread can collect together
35  * messages bound for one node and send them in one block.
36  *
37  * lowcomms will choose to use either TCP or SCTP as its transport layer
38  * depending on the configuration variable 'protocol'. This should be set
39  * to 0 (default) for TCP or 1 for SCTP. It should be configured using a
40  * cluster-wide mechanism as it must be the same on all nodes of the cluster
41  * for the DLM to function.
42  *
43  */
44 
45 #include <asm/ioctls.h>
46 #include <net/sock.h>
47 #include <net/tcp.h>
48 #include <linux/pagemap.h>
49 #include <linux/file.h>
50 #include <linux/mutex.h>
51 #include <linux/sctp.h>
52 #include <linux/slab.h>
53 #include <net/sctp/sctp.h>
54 #include <net/ipv6.h>
55 
56 #include "dlm_internal.h"
57 #include "lowcomms.h"
58 #include "midcomms.h"
59 #include "config.h"
60 
61 #define NEEDED_RMEM (4*1024*1024)
62 #define CONN_HASH_SIZE 32
63 
64 /* Number of messages to send before rescheduling */
65 #define MAX_SEND_MSG_COUNT 25
66 #define DLM_SHUTDOWN_WAIT_TIMEOUT msecs_to_jiffies(10000)
67 
68 struct connection {
69 	struct socket *sock;	/* NULL if not connected */
70 	uint32_t nodeid;	/* So we know who we are in the list */
71 	struct mutex sock_mutex;
72 	unsigned long flags;
73 #define CF_READ_PENDING 1
74 #define CF_WRITE_PENDING 2
75 #define CF_INIT_PENDING 4
76 #define CF_IS_OTHERCON 5
77 #define CF_CLOSE 6
78 #define CF_APP_LIMITED 7
79 #define CF_CLOSING 8
80 #define CF_SHUTDOWN 9
81 #define CF_CONNECTED 10
82 	struct list_head writequeue;  /* List of outgoing writequeue_entries */
83 	spinlock_t writequeue_lock;
84 	void (*connect_action) (struct connection *);	/* What to do to connect */
85 	void (*shutdown_action)(struct connection *con); /* What to do to shutdown */
86 	int retries;
87 #define MAX_CONNECT_RETRIES 3
88 	struct hlist_node list;
89 	struct connection *othercon;
90 	struct work_struct rwork; /* Receive workqueue */
91 	struct work_struct swork; /* Send workqueue */
92 	wait_queue_head_t shutdown_wait; /* wait for graceful shutdown */
93 	unsigned char *rx_buf;
94 	int rx_buflen;
95 	int rx_leftover;
96 	struct rcu_head rcu;
97 };
98 #define sock2con(x) ((struct connection *)(x)->sk_user_data)
99 
100 struct listen_connection {
101 	struct socket *sock;
102 	struct work_struct rwork;
103 };
104 
105 /* An entry waiting to be sent */
106 struct writequeue_entry {
107 	struct list_head list;
108 	struct page *page;
109 	int offset;
110 	int len;
111 	int end;
112 	int users;
113 	struct connection *con;
114 };
115 
116 struct dlm_node_addr {
117 	struct list_head list;
118 	int nodeid;
119 	int addr_count;
120 	int curr_addr_index;
121 	struct sockaddr_storage *addr[DLM_MAX_ADDR_COUNT];
122 };
123 
124 static struct listen_sock_callbacks {
125 	void (*sk_error_report)(struct sock *);
126 	void (*sk_data_ready)(struct sock *);
127 	void (*sk_state_change)(struct sock *);
128 	void (*sk_write_space)(struct sock *);
129 } listen_sock;
130 
131 static LIST_HEAD(dlm_node_addrs);
132 static DEFINE_SPINLOCK(dlm_node_addrs_spin);
133 
134 static struct listen_connection listen_con;
135 static struct sockaddr_storage *dlm_local_addr[DLM_MAX_ADDR_COUNT];
136 static int dlm_local_count;
137 static int dlm_allow_conn;
138 
139 /* Work queues */
140 static struct workqueue_struct *recv_workqueue;
141 static struct workqueue_struct *send_workqueue;
142 
143 static struct hlist_head connection_hash[CONN_HASH_SIZE];
144 static DEFINE_SPINLOCK(connections_lock);
145 DEFINE_STATIC_SRCU(connections_srcu);
146 
147 static void process_recv_sockets(struct work_struct *work);
148 static void process_send_sockets(struct work_struct *work);
149 
150 static void sctp_connect_to_sock(struct connection *con);
151 static void tcp_connect_to_sock(struct connection *con);
152 static void dlm_tcp_shutdown(struct connection *con);
153 
154 /* This is deliberately very simple because most clusters have simple
155    sequential nodeids, so we should be able to go straight to a connection
156    struct in the array */
157 static inline int nodeid_hash(int nodeid)
158 {
159 	return nodeid & (CONN_HASH_SIZE-1);
160 }
161 
162 static struct connection *__find_con(int nodeid)
163 {
164 	int r, idx;
165 	struct connection *con;
166 
167 	r = nodeid_hash(nodeid);
168 
169 	idx = srcu_read_lock(&connections_srcu);
170 	hlist_for_each_entry_rcu(con, &connection_hash[r], list) {
171 		if (con->nodeid == nodeid) {
172 			srcu_read_unlock(&connections_srcu, idx);
173 			return con;
174 		}
175 	}
176 	srcu_read_unlock(&connections_srcu, idx);
177 
178 	return NULL;
179 }
180 
181 static int dlm_con_init(struct connection *con, int nodeid)
182 {
183 	con->rx_buflen = dlm_config.ci_buffer_size;
184 	con->rx_buf = kmalloc(con->rx_buflen, GFP_NOFS);
185 	if (!con->rx_buf)
186 		return -ENOMEM;
187 
188 	con->nodeid = nodeid;
189 	mutex_init(&con->sock_mutex);
190 	INIT_LIST_HEAD(&con->writequeue);
191 	spin_lock_init(&con->writequeue_lock);
192 	INIT_WORK(&con->swork, process_send_sockets);
193 	INIT_WORK(&con->rwork, process_recv_sockets);
194 	init_waitqueue_head(&con->shutdown_wait);
195 
196 	if (dlm_config.ci_protocol == 0) {
197 		con->connect_action = tcp_connect_to_sock;
198 		con->shutdown_action = dlm_tcp_shutdown;
199 	} else {
200 		con->connect_action = sctp_connect_to_sock;
201 	}
202 
203 	return 0;
204 }
205 
206 /*
207  * If 'allocation' is zero then we don't attempt to create a new
208  * connection structure for this node.
209  */
210 static struct connection *nodeid2con(int nodeid, gfp_t alloc)
211 {
212 	struct connection *con, *tmp;
213 	int r, ret;
214 
215 	con = __find_con(nodeid);
216 	if (con || !alloc)
217 		return con;
218 
219 	con = kzalloc(sizeof(*con), alloc);
220 	if (!con)
221 		return NULL;
222 
223 	ret = dlm_con_init(con, nodeid);
224 	if (ret) {
225 		kfree(con);
226 		return NULL;
227 	}
228 
229 	r = nodeid_hash(nodeid);
230 
231 	spin_lock(&connections_lock);
232 	/* Because multiple workqueues/threads calls this function it can
233 	 * race on multiple cpu's. Instead of locking hot path __find_con()
234 	 * we just check in rare cases of recently added nodes again
235 	 * under protection of connections_lock. If this is the case we
236 	 * abort our connection creation and return the existing connection.
237 	 */
238 	tmp = __find_con(nodeid);
239 	if (tmp) {
240 		spin_unlock(&connections_lock);
241 		kfree(con->rx_buf);
242 		kfree(con);
243 		return tmp;
244 	}
245 
246 	hlist_add_head_rcu(&con->list, &connection_hash[r]);
247 	spin_unlock(&connections_lock);
248 
249 	return con;
250 }
251 
252 /* Loop round all connections */
253 static void foreach_conn(void (*conn_func)(struct connection *c))
254 {
255 	int i, idx;
256 	struct connection *con;
257 
258 	idx = srcu_read_lock(&connections_srcu);
259 	for (i = 0; i < CONN_HASH_SIZE; i++) {
260 		hlist_for_each_entry_rcu(con, &connection_hash[i], list)
261 			conn_func(con);
262 	}
263 	srcu_read_unlock(&connections_srcu, idx);
264 }
265 
266 static struct dlm_node_addr *find_node_addr(int nodeid)
267 {
268 	struct dlm_node_addr *na;
269 
270 	list_for_each_entry(na, &dlm_node_addrs, list) {
271 		if (na->nodeid == nodeid)
272 			return na;
273 	}
274 	return NULL;
275 }
276 
277 static int addr_compare(const struct sockaddr_storage *x,
278 			const struct sockaddr_storage *y)
279 {
280 	switch (x->ss_family) {
281 	case AF_INET: {
282 		struct sockaddr_in *sinx = (struct sockaddr_in *)x;
283 		struct sockaddr_in *siny = (struct sockaddr_in *)y;
284 		if (sinx->sin_addr.s_addr != siny->sin_addr.s_addr)
285 			return 0;
286 		if (sinx->sin_port != siny->sin_port)
287 			return 0;
288 		break;
289 	}
290 	case AF_INET6: {
291 		struct sockaddr_in6 *sinx = (struct sockaddr_in6 *)x;
292 		struct sockaddr_in6 *siny = (struct sockaddr_in6 *)y;
293 		if (!ipv6_addr_equal(&sinx->sin6_addr, &siny->sin6_addr))
294 			return 0;
295 		if (sinx->sin6_port != siny->sin6_port)
296 			return 0;
297 		break;
298 	}
299 	default:
300 		return 0;
301 	}
302 	return 1;
303 }
304 
305 static int nodeid_to_addr(int nodeid, struct sockaddr_storage *sas_out,
306 			  struct sockaddr *sa_out, bool try_new_addr)
307 {
308 	struct sockaddr_storage sas;
309 	struct dlm_node_addr *na;
310 
311 	if (!dlm_local_count)
312 		return -1;
313 
314 	spin_lock(&dlm_node_addrs_spin);
315 	na = find_node_addr(nodeid);
316 	if (na && na->addr_count) {
317 		memcpy(&sas, na->addr[na->curr_addr_index],
318 		       sizeof(struct sockaddr_storage));
319 
320 		if (try_new_addr) {
321 			na->curr_addr_index++;
322 			if (na->curr_addr_index == na->addr_count)
323 				na->curr_addr_index = 0;
324 		}
325 	}
326 	spin_unlock(&dlm_node_addrs_spin);
327 
328 	if (!na)
329 		return -EEXIST;
330 
331 	if (!na->addr_count)
332 		return -ENOENT;
333 
334 	if (sas_out)
335 		memcpy(sas_out, &sas, sizeof(struct sockaddr_storage));
336 
337 	if (!sa_out)
338 		return 0;
339 
340 	if (dlm_local_addr[0]->ss_family == AF_INET) {
341 		struct sockaddr_in *in4  = (struct sockaddr_in *) &sas;
342 		struct sockaddr_in *ret4 = (struct sockaddr_in *) sa_out;
343 		ret4->sin_addr.s_addr = in4->sin_addr.s_addr;
344 	} else {
345 		struct sockaddr_in6 *in6  = (struct sockaddr_in6 *) &sas;
346 		struct sockaddr_in6 *ret6 = (struct sockaddr_in6 *) sa_out;
347 		ret6->sin6_addr = in6->sin6_addr;
348 	}
349 
350 	return 0;
351 }
352 
353 static int addr_to_nodeid(struct sockaddr_storage *addr, int *nodeid)
354 {
355 	struct dlm_node_addr *na;
356 	int rv = -EEXIST;
357 	int addr_i;
358 
359 	spin_lock(&dlm_node_addrs_spin);
360 	list_for_each_entry(na, &dlm_node_addrs, list) {
361 		if (!na->addr_count)
362 			continue;
363 
364 		for (addr_i = 0; addr_i < na->addr_count; addr_i++) {
365 			if (addr_compare(na->addr[addr_i], addr)) {
366 				*nodeid = na->nodeid;
367 				rv = 0;
368 				goto unlock;
369 			}
370 		}
371 	}
372 unlock:
373 	spin_unlock(&dlm_node_addrs_spin);
374 	return rv;
375 }
376 
377 /* caller need to held dlm_node_addrs_spin lock */
378 static bool dlm_lowcomms_na_has_addr(const struct dlm_node_addr *na,
379 				     const struct sockaddr_storage *addr)
380 {
381 	int i;
382 
383 	for (i = 0; i < na->addr_count; i++) {
384 		if (addr_compare(na->addr[i], addr))
385 			return true;
386 	}
387 
388 	return false;
389 }
390 
391 int dlm_lowcomms_addr(int nodeid, struct sockaddr_storage *addr, int len)
392 {
393 	struct sockaddr_storage *new_addr;
394 	struct dlm_node_addr *new_node, *na;
395 	bool ret;
396 
397 	new_node = kzalloc(sizeof(struct dlm_node_addr), GFP_NOFS);
398 	if (!new_node)
399 		return -ENOMEM;
400 
401 	new_addr = kzalloc(sizeof(struct sockaddr_storage), GFP_NOFS);
402 	if (!new_addr) {
403 		kfree(new_node);
404 		return -ENOMEM;
405 	}
406 
407 	memcpy(new_addr, addr, len);
408 
409 	spin_lock(&dlm_node_addrs_spin);
410 	na = find_node_addr(nodeid);
411 	if (!na) {
412 		new_node->nodeid = nodeid;
413 		new_node->addr[0] = new_addr;
414 		new_node->addr_count = 1;
415 		list_add(&new_node->list, &dlm_node_addrs);
416 		spin_unlock(&dlm_node_addrs_spin);
417 		return 0;
418 	}
419 
420 	ret = dlm_lowcomms_na_has_addr(na, addr);
421 	if (ret) {
422 		spin_unlock(&dlm_node_addrs_spin);
423 		kfree(new_addr);
424 		kfree(new_node);
425 		return -EEXIST;
426 	}
427 
428 	if (na->addr_count >= DLM_MAX_ADDR_COUNT) {
429 		spin_unlock(&dlm_node_addrs_spin);
430 		kfree(new_addr);
431 		kfree(new_node);
432 		return -ENOSPC;
433 	}
434 
435 	na->addr[na->addr_count++] = new_addr;
436 	spin_unlock(&dlm_node_addrs_spin);
437 	kfree(new_node);
438 	return 0;
439 }
440 
441 /* Data available on socket or listen socket received a connect */
442 static void lowcomms_data_ready(struct sock *sk)
443 {
444 	struct connection *con;
445 
446 	read_lock_bh(&sk->sk_callback_lock);
447 	con = sock2con(sk);
448 	if (con && !test_and_set_bit(CF_READ_PENDING, &con->flags))
449 		queue_work(recv_workqueue, &con->rwork);
450 	read_unlock_bh(&sk->sk_callback_lock);
451 }
452 
453 static void lowcomms_listen_data_ready(struct sock *sk)
454 {
455 	queue_work(recv_workqueue, &listen_con.rwork);
456 }
457 
458 static void lowcomms_write_space(struct sock *sk)
459 {
460 	struct connection *con;
461 
462 	read_lock_bh(&sk->sk_callback_lock);
463 	con = sock2con(sk);
464 	if (!con)
465 		goto out;
466 
467 	if (!test_and_set_bit(CF_CONNECTED, &con->flags)) {
468 		log_print("successful connected to node %d", con->nodeid);
469 		queue_work(send_workqueue, &con->swork);
470 		goto out;
471 	}
472 
473 	clear_bit(SOCK_NOSPACE, &con->sock->flags);
474 
475 	if (test_and_clear_bit(CF_APP_LIMITED, &con->flags)) {
476 		con->sock->sk->sk_write_pending--;
477 		clear_bit(SOCKWQ_ASYNC_NOSPACE, &con->sock->flags);
478 	}
479 
480 	queue_work(send_workqueue, &con->swork);
481 out:
482 	read_unlock_bh(&sk->sk_callback_lock);
483 }
484 
485 static inline void lowcomms_connect_sock(struct connection *con)
486 {
487 	if (test_bit(CF_CLOSE, &con->flags))
488 		return;
489 	queue_work(send_workqueue, &con->swork);
490 	cond_resched();
491 }
492 
493 static void lowcomms_state_change(struct sock *sk)
494 {
495 	/* SCTP layer is not calling sk_data_ready when the connection
496 	 * is done, so we catch the signal through here. Also, it
497 	 * doesn't switch socket state when entering shutdown, so we
498 	 * skip the write in that case.
499 	 */
500 	if (sk->sk_shutdown) {
501 		if (sk->sk_shutdown == RCV_SHUTDOWN)
502 			lowcomms_data_ready(sk);
503 	} else if (sk->sk_state == TCP_ESTABLISHED) {
504 		lowcomms_write_space(sk);
505 	}
506 }
507 
508 int dlm_lowcomms_connect_node(int nodeid)
509 {
510 	struct connection *con;
511 
512 	if (nodeid == dlm_our_nodeid())
513 		return 0;
514 
515 	con = nodeid2con(nodeid, GFP_NOFS);
516 	if (!con)
517 		return -ENOMEM;
518 	lowcomms_connect_sock(con);
519 	return 0;
520 }
521 
522 static void lowcomms_error_report(struct sock *sk)
523 {
524 	struct connection *con;
525 	struct sockaddr_storage saddr;
526 	void (*orig_report)(struct sock *) = NULL;
527 
528 	read_lock_bh(&sk->sk_callback_lock);
529 	con = sock2con(sk);
530 	if (con == NULL)
531 		goto out;
532 
533 	orig_report = listen_sock.sk_error_report;
534 	if (con->sock == NULL ||
535 	    kernel_getpeername(con->sock, (struct sockaddr *)&saddr) < 0) {
536 		printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
537 				   "sending to node %d, port %d, "
538 				   "sk_err=%d/%d\n", dlm_our_nodeid(),
539 				   con->nodeid, dlm_config.ci_tcp_port,
540 				   sk->sk_err, sk->sk_err_soft);
541 	} else if (saddr.ss_family == AF_INET) {
542 		struct sockaddr_in *sin4 = (struct sockaddr_in *)&saddr;
543 
544 		printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
545 				   "sending to node %d at %pI4, port %d, "
546 				   "sk_err=%d/%d\n", dlm_our_nodeid(),
547 				   con->nodeid, &sin4->sin_addr.s_addr,
548 				   dlm_config.ci_tcp_port, sk->sk_err,
549 				   sk->sk_err_soft);
550 	} else {
551 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&saddr;
552 
553 		printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
554 				   "sending to node %d at %u.%u.%u.%u, "
555 				   "port %d, sk_err=%d/%d\n", dlm_our_nodeid(),
556 				   con->nodeid, sin6->sin6_addr.s6_addr32[0],
557 				   sin6->sin6_addr.s6_addr32[1],
558 				   sin6->sin6_addr.s6_addr32[2],
559 				   sin6->sin6_addr.s6_addr32[3],
560 				   dlm_config.ci_tcp_port, sk->sk_err,
561 				   sk->sk_err_soft);
562 	}
563 out:
564 	read_unlock_bh(&sk->sk_callback_lock);
565 	if (orig_report)
566 		orig_report(sk);
567 }
568 
569 /* Note: sk_callback_lock must be locked before calling this function. */
570 static void save_listen_callbacks(struct socket *sock)
571 {
572 	struct sock *sk = sock->sk;
573 
574 	listen_sock.sk_data_ready = sk->sk_data_ready;
575 	listen_sock.sk_state_change = sk->sk_state_change;
576 	listen_sock.sk_write_space = sk->sk_write_space;
577 	listen_sock.sk_error_report = sk->sk_error_report;
578 }
579 
580 static void restore_callbacks(struct socket *sock)
581 {
582 	struct sock *sk = sock->sk;
583 
584 	write_lock_bh(&sk->sk_callback_lock);
585 	sk->sk_user_data = NULL;
586 	sk->sk_data_ready = listen_sock.sk_data_ready;
587 	sk->sk_state_change = listen_sock.sk_state_change;
588 	sk->sk_write_space = listen_sock.sk_write_space;
589 	sk->sk_error_report = listen_sock.sk_error_report;
590 	write_unlock_bh(&sk->sk_callback_lock);
591 }
592 
593 static void add_listen_sock(struct socket *sock, struct listen_connection *con)
594 {
595 	struct sock *sk = sock->sk;
596 
597 	write_lock_bh(&sk->sk_callback_lock);
598 	save_listen_callbacks(sock);
599 	con->sock = sock;
600 
601 	sk->sk_user_data = con;
602 	sk->sk_allocation = GFP_NOFS;
603 	/* Install a data_ready callback */
604 	sk->sk_data_ready = lowcomms_listen_data_ready;
605 	write_unlock_bh(&sk->sk_callback_lock);
606 }
607 
608 /* Make a socket active */
609 static void add_sock(struct socket *sock, struct connection *con)
610 {
611 	struct sock *sk = sock->sk;
612 
613 	write_lock_bh(&sk->sk_callback_lock);
614 	con->sock = sock;
615 
616 	sk->sk_user_data = con;
617 	/* Install a data_ready callback */
618 	sk->sk_data_ready = lowcomms_data_ready;
619 	sk->sk_write_space = lowcomms_write_space;
620 	sk->sk_state_change = lowcomms_state_change;
621 	sk->sk_allocation = GFP_NOFS;
622 	sk->sk_error_report = lowcomms_error_report;
623 	write_unlock_bh(&sk->sk_callback_lock);
624 }
625 
626 /* Add the port number to an IPv6 or 4 sockaddr and return the address
627    length */
628 static void make_sockaddr(struct sockaddr_storage *saddr, uint16_t port,
629 			  int *addr_len)
630 {
631 	saddr->ss_family =  dlm_local_addr[0]->ss_family;
632 	if (saddr->ss_family == AF_INET) {
633 		struct sockaddr_in *in4_addr = (struct sockaddr_in *)saddr;
634 		in4_addr->sin_port = cpu_to_be16(port);
635 		*addr_len = sizeof(struct sockaddr_in);
636 		memset(&in4_addr->sin_zero, 0, sizeof(in4_addr->sin_zero));
637 	} else {
638 		struct sockaddr_in6 *in6_addr = (struct sockaddr_in6 *)saddr;
639 		in6_addr->sin6_port = cpu_to_be16(port);
640 		*addr_len = sizeof(struct sockaddr_in6);
641 	}
642 	memset((char *)saddr + *addr_len, 0, sizeof(struct sockaddr_storage) - *addr_len);
643 }
644 
645 static void dlm_close_sock(struct socket **sock)
646 {
647 	if (*sock) {
648 		restore_callbacks(*sock);
649 		sock_release(*sock);
650 		*sock = NULL;
651 	}
652 }
653 
654 /* Close a remote connection and tidy up */
655 static void close_connection(struct connection *con, bool and_other,
656 			     bool tx, bool rx)
657 {
658 	bool closing = test_and_set_bit(CF_CLOSING, &con->flags);
659 
660 	if (tx && !closing && cancel_work_sync(&con->swork)) {
661 		log_print("canceled swork for node %d", con->nodeid);
662 		clear_bit(CF_WRITE_PENDING, &con->flags);
663 	}
664 	if (rx && !closing && cancel_work_sync(&con->rwork)) {
665 		log_print("canceled rwork for node %d", con->nodeid);
666 		clear_bit(CF_READ_PENDING, &con->flags);
667 	}
668 
669 	mutex_lock(&con->sock_mutex);
670 	dlm_close_sock(&con->sock);
671 
672 	if (con->othercon && and_other) {
673 		/* Will only re-enter once. */
674 		close_connection(con->othercon, false, true, true);
675 	}
676 
677 	con->rx_leftover = 0;
678 	con->retries = 0;
679 	clear_bit(CF_CONNECTED, &con->flags);
680 	mutex_unlock(&con->sock_mutex);
681 	clear_bit(CF_CLOSING, &con->flags);
682 }
683 
684 static void shutdown_connection(struct connection *con)
685 {
686 	int ret;
687 
688 	if (cancel_work_sync(&con->swork)) {
689 		log_print("canceled swork for node %d", con->nodeid);
690 		clear_bit(CF_WRITE_PENDING, &con->flags);
691 	}
692 
693 	mutex_lock(&con->sock_mutex);
694 	/* nothing to shutdown */
695 	if (!con->sock) {
696 		mutex_unlock(&con->sock_mutex);
697 		return;
698 	}
699 
700 	set_bit(CF_SHUTDOWN, &con->flags);
701 	ret = kernel_sock_shutdown(con->sock, SHUT_WR);
702 	mutex_unlock(&con->sock_mutex);
703 	if (ret) {
704 		log_print("Connection %p failed to shutdown: %d will force close",
705 			  con, ret);
706 		goto force_close;
707 	} else {
708 		ret = wait_event_timeout(con->shutdown_wait,
709 					 !test_bit(CF_SHUTDOWN, &con->flags),
710 					 DLM_SHUTDOWN_WAIT_TIMEOUT);
711 		if (ret == 0) {
712 			log_print("Connection %p shutdown timed out, will force close",
713 				  con);
714 			goto force_close;
715 		}
716 	}
717 
718 	return;
719 
720 force_close:
721 	clear_bit(CF_SHUTDOWN, &con->flags);
722 	close_connection(con, false, true, true);
723 }
724 
725 static void dlm_tcp_shutdown(struct connection *con)
726 {
727 	if (con->othercon)
728 		shutdown_connection(con->othercon);
729 	shutdown_connection(con);
730 }
731 
732 static int con_realloc_receive_buf(struct connection *con, int newlen)
733 {
734 	unsigned char *newbuf;
735 
736 	newbuf = kmalloc(newlen, GFP_NOFS);
737 	if (!newbuf)
738 		return -ENOMEM;
739 
740 	/* copy any leftover from last receive */
741 	if (con->rx_leftover)
742 		memmove(newbuf, con->rx_buf, con->rx_leftover);
743 
744 	/* swap to new buffer space */
745 	kfree(con->rx_buf);
746 	con->rx_buflen = newlen;
747 	con->rx_buf = newbuf;
748 
749 	return 0;
750 }
751 
752 /* Data received from remote end */
753 static int receive_from_sock(struct connection *con)
754 {
755 	int call_again_soon = 0;
756 	struct msghdr msg;
757 	struct kvec iov;
758 	int ret, buflen;
759 
760 	mutex_lock(&con->sock_mutex);
761 
762 	if (con->sock == NULL) {
763 		ret = -EAGAIN;
764 		goto out_close;
765 	}
766 
767 	/* realloc if we get new buffer size to read out */
768 	buflen = dlm_config.ci_buffer_size;
769 	if (con->rx_buflen != buflen && con->rx_leftover <= buflen) {
770 		ret = con_realloc_receive_buf(con, buflen);
771 		if (ret < 0)
772 			goto out_resched;
773 	}
774 
775 	/* calculate new buffer parameter regarding last receive and
776 	 * possible leftover bytes
777 	 */
778 	iov.iov_base = con->rx_buf + con->rx_leftover;
779 	iov.iov_len = con->rx_buflen - con->rx_leftover;
780 
781 	memset(&msg, 0, sizeof(msg));
782 	msg.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
783 	ret = kernel_recvmsg(con->sock, &msg, &iov, 1, iov.iov_len,
784 			     msg.msg_flags);
785 	if (ret <= 0)
786 		goto out_close;
787 	else if (ret == iov.iov_len)
788 		call_again_soon = 1;
789 
790 	/* new buflen according readed bytes and leftover from last receive */
791 	buflen = ret + con->rx_leftover;
792 	ret = dlm_process_incoming_buffer(con->nodeid, con->rx_buf, buflen);
793 	if (ret < 0)
794 		goto out_close;
795 
796 	/* calculate leftover bytes from process and put it into begin of
797 	 * the receive buffer, so next receive we have the full message
798 	 * at the start address of the receive buffer.
799 	 */
800 	con->rx_leftover = buflen - ret;
801 	if (con->rx_leftover) {
802 		memmove(con->rx_buf, con->rx_buf + ret,
803 			con->rx_leftover);
804 		call_again_soon = true;
805 	}
806 
807 	if (call_again_soon)
808 		goto out_resched;
809 
810 	mutex_unlock(&con->sock_mutex);
811 	return 0;
812 
813 out_resched:
814 	if (!test_and_set_bit(CF_READ_PENDING, &con->flags))
815 		queue_work(recv_workqueue, &con->rwork);
816 	mutex_unlock(&con->sock_mutex);
817 	return -EAGAIN;
818 
819 out_close:
820 	mutex_unlock(&con->sock_mutex);
821 	if (ret != -EAGAIN) {
822 		/* Reconnect when there is something to send */
823 		close_connection(con, false, true, false);
824 		if (ret == 0) {
825 			log_print("connection %p got EOF from %d",
826 				  con, con->nodeid);
827 			/* handling for tcp shutdown */
828 			clear_bit(CF_SHUTDOWN, &con->flags);
829 			wake_up(&con->shutdown_wait);
830 			/* signal to breaking receive worker */
831 			ret = -1;
832 		}
833 	}
834 	return ret;
835 }
836 
837 /* Listening socket is busy, accept a connection */
838 static int accept_from_sock(struct listen_connection *con)
839 {
840 	int result;
841 	struct sockaddr_storage peeraddr;
842 	struct socket *newsock;
843 	int len;
844 	int nodeid;
845 	struct connection *newcon;
846 	struct connection *addcon;
847 	unsigned int mark;
848 
849 	if (!dlm_allow_conn) {
850 		return -1;
851 	}
852 
853 	if (!con->sock)
854 		return -ENOTCONN;
855 
856 	result = kernel_accept(con->sock, &newsock, O_NONBLOCK);
857 	if (result < 0)
858 		goto accept_err;
859 
860 	/* Get the connected socket's peer */
861 	memset(&peeraddr, 0, sizeof(peeraddr));
862 	len = newsock->ops->getname(newsock, (struct sockaddr *)&peeraddr, 2);
863 	if (len < 0) {
864 		result = -ECONNABORTED;
865 		goto accept_err;
866 	}
867 
868 	/* Get the new node's NODEID */
869 	make_sockaddr(&peeraddr, 0, &len);
870 	if (addr_to_nodeid(&peeraddr, &nodeid)) {
871 		unsigned char *b=(unsigned char *)&peeraddr;
872 		log_print("connect from non cluster node");
873 		print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
874 				     b, sizeof(struct sockaddr_storage));
875 		sock_release(newsock);
876 		return -1;
877 	}
878 
879 	dlm_comm_mark(nodeid, &mark);
880 	sock_set_mark(newsock->sk, mark);
881 
882 	log_print("got connection from %d", nodeid);
883 
884 	/*  Check to see if we already have a connection to this node. This
885 	 *  could happen if the two nodes initiate a connection at roughly
886 	 *  the same time and the connections cross on the wire.
887 	 *  In this case we store the incoming one in "othercon"
888 	 */
889 	newcon = nodeid2con(nodeid, GFP_NOFS);
890 	if (!newcon) {
891 		result = -ENOMEM;
892 		goto accept_err;
893 	}
894 
895 	mutex_lock(&newcon->sock_mutex);
896 	if (newcon->sock) {
897 		struct connection *othercon = newcon->othercon;
898 
899 		if (!othercon) {
900 			othercon = kzalloc(sizeof(*othercon), GFP_NOFS);
901 			if (!othercon) {
902 				log_print("failed to allocate incoming socket");
903 				mutex_unlock(&newcon->sock_mutex);
904 				result = -ENOMEM;
905 				goto accept_err;
906 			}
907 
908 			result = dlm_con_init(othercon, nodeid);
909 			if (result < 0) {
910 				kfree(othercon);
911 				goto accept_err;
912 			}
913 
914 			newcon->othercon = othercon;
915 		} else {
916 			/* close other sock con if we have something new */
917 			close_connection(othercon, false, true, false);
918 		}
919 
920 		mutex_lock_nested(&othercon->sock_mutex, 1);
921 		add_sock(newsock, othercon);
922 		addcon = othercon;
923 		mutex_unlock(&othercon->sock_mutex);
924 	}
925 	else {
926 		/* accept copies the sk after we've saved the callbacks, so we
927 		   don't want to save them a second time or comm errors will
928 		   result in calling sk_error_report recursively. */
929 		add_sock(newsock, newcon);
930 		addcon = newcon;
931 	}
932 
933 	mutex_unlock(&newcon->sock_mutex);
934 
935 	/*
936 	 * Add it to the active queue in case we got data
937 	 * between processing the accept adding the socket
938 	 * to the read_sockets list
939 	 */
940 	if (!test_and_set_bit(CF_READ_PENDING, &addcon->flags))
941 		queue_work(recv_workqueue, &addcon->rwork);
942 
943 	return 0;
944 
945 accept_err:
946 	if (newsock)
947 		sock_release(newsock);
948 
949 	if (result != -EAGAIN)
950 		log_print("error accepting connection from node: %d", result);
951 	return result;
952 }
953 
954 static void free_entry(struct writequeue_entry *e)
955 {
956 	__free_page(e->page);
957 	kfree(e);
958 }
959 
960 /*
961  * writequeue_entry_complete - try to delete and free write queue entry
962  * @e: write queue entry to try to delete
963  * @completed: bytes completed
964  *
965  * writequeue_lock must be held.
966  */
967 static void writequeue_entry_complete(struct writequeue_entry *e, int completed)
968 {
969 	e->offset += completed;
970 	e->len -= completed;
971 
972 	if (e->len == 0 && e->users == 0) {
973 		list_del(&e->list);
974 		free_entry(e);
975 	}
976 }
977 
978 /*
979  * sctp_bind_addrs - bind a SCTP socket to all our addresses
980  */
981 static int sctp_bind_addrs(struct socket *sock, uint16_t port)
982 {
983 	struct sockaddr_storage localaddr;
984 	struct sockaddr *addr = (struct sockaddr *)&localaddr;
985 	int i, addr_len, result = 0;
986 
987 	for (i = 0; i < dlm_local_count; i++) {
988 		memcpy(&localaddr, dlm_local_addr[i], sizeof(localaddr));
989 		make_sockaddr(&localaddr, port, &addr_len);
990 
991 		if (!i)
992 			result = kernel_bind(sock, addr, addr_len);
993 		else
994 			result = sock_bind_add(sock->sk, addr, addr_len);
995 
996 		if (result < 0) {
997 			log_print("Can't bind to %d addr number %d, %d.\n",
998 				  port, i + 1, result);
999 			break;
1000 		}
1001 	}
1002 	return result;
1003 }
1004 
1005 /* Initiate an SCTP association.
1006    This is a special case of send_to_sock() in that we don't yet have a
1007    peeled-off socket for this association, so we use the listening socket
1008    and add the primary IP address of the remote node.
1009  */
1010 static void sctp_connect_to_sock(struct connection *con)
1011 {
1012 	struct sockaddr_storage daddr;
1013 	int result;
1014 	int addr_len;
1015 	struct socket *sock;
1016 	unsigned int mark;
1017 
1018 	dlm_comm_mark(con->nodeid, &mark);
1019 
1020 	mutex_lock(&con->sock_mutex);
1021 
1022 	/* Some odd races can cause double-connects, ignore them */
1023 	if (con->retries++ > MAX_CONNECT_RETRIES)
1024 		goto out;
1025 
1026 	if (con->sock) {
1027 		log_print("node %d already connected.", con->nodeid);
1028 		goto out;
1029 	}
1030 
1031 	memset(&daddr, 0, sizeof(daddr));
1032 	result = nodeid_to_addr(con->nodeid, &daddr, NULL, true);
1033 	if (result < 0) {
1034 		log_print("no address for nodeid %d", con->nodeid);
1035 		goto out;
1036 	}
1037 
1038 	/* Create a socket to communicate with */
1039 	result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1040 				  SOCK_STREAM, IPPROTO_SCTP, &sock);
1041 	if (result < 0)
1042 		goto socket_err;
1043 
1044 	sock_set_mark(sock->sk, mark);
1045 
1046 	add_sock(sock, con);
1047 
1048 	/* Bind to all addresses. */
1049 	if (sctp_bind_addrs(con->sock, 0))
1050 		goto bind_err;
1051 
1052 	make_sockaddr(&daddr, dlm_config.ci_tcp_port, &addr_len);
1053 
1054 	log_print("connecting to %d", con->nodeid);
1055 
1056 	/* Turn off Nagle's algorithm */
1057 	sctp_sock_set_nodelay(sock->sk);
1058 
1059 	/*
1060 	 * Make sock->ops->connect() function return in specified time,
1061 	 * since O_NONBLOCK argument in connect() function does not work here,
1062 	 * then, we should restore the default value of this attribute.
1063 	 */
1064 	sock_set_sndtimeo(sock->sk, 5);
1065 	result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
1066 				   0);
1067 	sock_set_sndtimeo(sock->sk, 0);
1068 
1069 	if (result == -EINPROGRESS)
1070 		result = 0;
1071 	if (result == 0) {
1072 		if (!test_and_set_bit(CF_CONNECTED, &con->flags))
1073 			log_print("successful connected to node %d", con->nodeid);
1074 		goto out;
1075 	}
1076 
1077 bind_err:
1078 	con->sock = NULL;
1079 	sock_release(sock);
1080 
1081 socket_err:
1082 	/*
1083 	 * Some errors are fatal and this list might need adjusting. For other
1084 	 * errors we try again until the max number of retries is reached.
1085 	 */
1086 	if (result != -EHOSTUNREACH &&
1087 	    result != -ENETUNREACH &&
1088 	    result != -ENETDOWN &&
1089 	    result != -EINVAL &&
1090 	    result != -EPROTONOSUPPORT) {
1091 		log_print("connect %d try %d error %d", con->nodeid,
1092 			  con->retries, result);
1093 		mutex_unlock(&con->sock_mutex);
1094 		msleep(1000);
1095 		lowcomms_connect_sock(con);
1096 		return;
1097 	}
1098 
1099 out:
1100 	mutex_unlock(&con->sock_mutex);
1101 }
1102 
1103 /* Connect a new socket to its peer */
1104 static void tcp_connect_to_sock(struct connection *con)
1105 {
1106 	struct sockaddr_storage saddr, src_addr;
1107 	int addr_len;
1108 	struct socket *sock = NULL;
1109 	unsigned int mark;
1110 	int result;
1111 
1112 	dlm_comm_mark(con->nodeid, &mark);
1113 
1114 	mutex_lock(&con->sock_mutex);
1115 	if (con->retries++ > MAX_CONNECT_RETRIES)
1116 		goto out;
1117 
1118 	/* Some odd races can cause double-connects, ignore them */
1119 	if (con->sock)
1120 		goto out;
1121 
1122 	/* Create a socket to communicate with */
1123 	result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1124 				  SOCK_STREAM, IPPROTO_TCP, &sock);
1125 	if (result < 0)
1126 		goto out_err;
1127 
1128 	sock_set_mark(sock->sk, mark);
1129 
1130 	memset(&saddr, 0, sizeof(saddr));
1131 	result = nodeid_to_addr(con->nodeid, &saddr, NULL, false);
1132 	if (result < 0) {
1133 		log_print("no address for nodeid %d", con->nodeid);
1134 		goto out_err;
1135 	}
1136 
1137 	add_sock(sock, con);
1138 
1139 	/* Bind to our cluster-known address connecting to avoid
1140 	   routing problems */
1141 	memcpy(&src_addr, dlm_local_addr[0], sizeof(src_addr));
1142 	make_sockaddr(&src_addr, 0, &addr_len);
1143 	result = sock->ops->bind(sock, (struct sockaddr *) &src_addr,
1144 				 addr_len);
1145 	if (result < 0) {
1146 		log_print("could not bind for connect: %d", result);
1147 		/* This *may* not indicate a critical error */
1148 	}
1149 
1150 	make_sockaddr(&saddr, dlm_config.ci_tcp_port, &addr_len);
1151 
1152 	log_print("connecting to %d", con->nodeid);
1153 
1154 	/* Turn off Nagle's algorithm */
1155 	tcp_sock_set_nodelay(sock->sk);
1156 
1157 	result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
1158 				   O_NONBLOCK);
1159 	if (result == -EINPROGRESS)
1160 		result = 0;
1161 	if (result == 0)
1162 		goto out;
1163 
1164 out_err:
1165 	if (con->sock) {
1166 		sock_release(con->sock);
1167 		con->sock = NULL;
1168 	} else if (sock) {
1169 		sock_release(sock);
1170 	}
1171 	/*
1172 	 * Some errors are fatal and this list might need adjusting. For other
1173 	 * errors we try again until the max number of retries is reached.
1174 	 */
1175 	if (result != -EHOSTUNREACH &&
1176 	    result != -ENETUNREACH &&
1177 	    result != -ENETDOWN &&
1178 	    result != -EINVAL &&
1179 	    result != -EPROTONOSUPPORT) {
1180 		log_print("connect %d try %d error %d", con->nodeid,
1181 			  con->retries, result);
1182 		mutex_unlock(&con->sock_mutex);
1183 		msleep(1000);
1184 		lowcomms_connect_sock(con);
1185 		return;
1186 	}
1187 out:
1188 	mutex_unlock(&con->sock_mutex);
1189 	return;
1190 }
1191 
1192 /* On error caller must run dlm_close_sock() for the
1193  * listen connection socket.
1194  */
1195 static int tcp_create_listen_sock(struct listen_connection *con,
1196 				  struct sockaddr_storage *saddr)
1197 {
1198 	struct socket *sock = NULL;
1199 	int result = 0;
1200 	int addr_len;
1201 
1202 	if (dlm_local_addr[0]->ss_family == AF_INET)
1203 		addr_len = sizeof(struct sockaddr_in);
1204 	else
1205 		addr_len = sizeof(struct sockaddr_in6);
1206 
1207 	/* Create a socket to communicate with */
1208 	result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1209 				  SOCK_STREAM, IPPROTO_TCP, &sock);
1210 	if (result < 0) {
1211 		log_print("Can't create listening comms socket");
1212 		goto create_out;
1213 	}
1214 
1215 	sock_set_mark(sock->sk, dlm_config.ci_mark);
1216 
1217 	/* Turn off Nagle's algorithm */
1218 	tcp_sock_set_nodelay(sock->sk);
1219 
1220 	sock_set_reuseaddr(sock->sk);
1221 
1222 	add_listen_sock(sock, con);
1223 
1224 	/* Bind to our port */
1225 	make_sockaddr(saddr, dlm_config.ci_tcp_port, &addr_len);
1226 	result = sock->ops->bind(sock, (struct sockaddr *) saddr, addr_len);
1227 	if (result < 0) {
1228 		log_print("Can't bind to port %d", dlm_config.ci_tcp_port);
1229 		goto create_out;
1230 	}
1231 	sock_set_keepalive(sock->sk);
1232 
1233 	result = sock->ops->listen(sock, 5);
1234 	if (result < 0) {
1235 		log_print("Can't listen on port %d", dlm_config.ci_tcp_port);
1236 		goto create_out;
1237 	}
1238 
1239 	return 0;
1240 
1241 create_out:
1242 	return result;
1243 }
1244 
1245 /* Get local addresses */
1246 static void init_local(void)
1247 {
1248 	struct sockaddr_storage sas, *addr;
1249 	int i;
1250 
1251 	dlm_local_count = 0;
1252 	for (i = 0; i < DLM_MAX_ADDR_COUNT; i++) {
1253 		if (dlm_our_addr(&sas, i))
1254 			break;
1255 
1256 		addr = kmemdup(&sas, sizeof(*addr), GFP_NOFS);
1257 		if (!addr)
1258 			break;
1259 		dlm_local_addr[dlm_local_count++] = addr;
1260 	}
1261 }
1262 
1263 static void deinit_local(void)
1264 {
1265 	int i;
1266 
1267 	for (i = 0; i < dlm_local_count; i++)
1268 		kfree(dlm_local_addr[i]);
1269 }
1270 
1271 /* Initialise SCTP socket and bind to all interfaces
1272  * On error caller must run dlm_close_sock() for the
1273  * listen connection socket.
1274  */
1275 static int sctp_listen_for_all(struct listen_connection *con)
1276 {
1277 	struct socket *sock = NULL;
1278 	int result = -EINVAL;
1279 
1280 	log_print("Using SCTP for communications");
1281 
1282 	result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1283 				  SOCK_STREAM, IPPROTO_SCTP, &sock);
1284 	if (result < 0) {
1285 		log_print("Can't create comms socket, check SCTP is loaded");
1286 		goto out;
1287 	}
1288 
1289 	sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
1290 	sock_set_mark(sock->sk, dlm_config.ci_mark);
1291 	sctp_sock_set_nodelay(sock->sk);
1292 
1293 	add_listen_sock(sock, con);
1294 
1295 	/* Bind to all addresses. */
1296 	result = sctp_bind_addrs(con->sock, dlm_config.ci_tcp_port);
1297 	if (result < 0)
1298 		goto out;
1299 
1300 	result = sock->ops->listen(sock, 5);
1301 	if (result < 0) {
1302 		log_print("Can't set socket listening");
1303 		goto out;
1304 	}
1305 
1306 	return 0;
1307 
1308 out:
1309 	return result;
1310 }
1311 
1312 static int tcp_listen_for_all(void)
1313 {
1314 	/* We don't support multi-homed hosts */
1315 	if (dlm_local_count > 1) {
1316 		log_print("TCP protocol can't handle multi-homed hosts, "
1317 			  "try SCTP");
1318 		return -EINVAL;
1319 	}
1320 
1321 	log_print("Using TCP for communications");
1322 
1323 	return tcp_create_listen_sock(&listen_con, dlm_local_addr[0]);
1324 }
1325 
1326 
1327 
1328 static struct writequeue_entry *new_writequeue_entry(struct connection *con,
1329 						     gfp_t allocation)
1330 {
1331 	struct writequeue_entry *entry;
1332 
1333 	entry = kmalloc(sizeof(struct writequeue_entry), allocation);
1334 	if (!entry)
1335 		return NULL;
1336 
1337 	entry->page = alloc_page(allocation);
1338 	if (!entry->page) {
1339 		kfree(entry);
1340 		return NULL;
1341 	}
1342 
1343 	entry->offset = 0;
1344 	entry->len = 0;
1345 	entry->end = 0;
1346 	entry->users = 0;
1347 	entry->con = con;
1348 
1349 	return entry;
1350 }
1351 
1352 void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
1353 {
1354 	struct connection *con;
1355 	struct writequeue_entry *e;
1356 	int offset = 0;
1357 
1358 	if (len > LOWCOMMS_MAX_TX_BUFFER_LEN) {
1359 		BUILD_BUG_ON(PAGE_SIZE < LOWCOMMS_MAX_TX_BUFFER_LEN);
1360 		log_print("failed to allocate a buffer of size %d", len);
1361 		return NULL;
1362 	}
1363 
1364 	con = nodeid2con(nodeid, allocation);
1365 	if (!con)
1366 		return NULL;
1367 
1368 	spin_lock(&con->writequeue_lock);
1369 	e = list_entry(con->writequeue.prev, struct writequeue_entry, list);
1370 	if ((&e->list == &con->writequeue) ||
1371 	    (PAGE_SIZE - e->end < len)) {
1372 		e = NULL;
1373 	} else {
1374 		offset = e->end;
1375 		e->end += len;
1376 		e->users++;
1377 	}
1378 	spin_unlock(&con->writequeue_lock);
1379 
1380 	if (e) {
1381 	got_one:
1382 		*ppc = page_address(e->page) + offset;
1383 		return e;
1384 	}
1385 
1386 	e = new_writequeue_entry(con, allocation);
1387 	if (e) {
1388 		spin_lock(&con->writequeue_lock);
1389 		offset = e->end;
1390 		e->end += len;
1391 		e->users++;
1392 		list_add_tail(&e->list, &con->writequeue);
1393 		spin_unlock(&con->writequeue_lock);
1394 		goto got_one;
1395 	}
1396 	return NULL;
1397 }
1398 
1399 void dlm_lowcomms_commit_buffer(void *mh)
1400 {
1401 	struct writequeue_entry *e = (struct writequeue_entry *)mh;
1402 	struct connection *con = e->con;
1403 	int users;
1404 
1405 	spin_lock(&con->writequeue_lock);
1406 	users = --e->users;
1407 	if (users)
1408 		goto out;
1409 	e->len = e->end - e->offset;
1410 	spin_unlock(&con->writequeue_lock);
1411 
1412 	queue_work(send_workqueue, &con->swork);
1413 	return;
1414 
1415 out:
1416 	spin_unlock(&con->writequeue_lock);
1417 	return;
1418 }
1419 
1420 /* Send a message */
1421 static void send_to_sock(struct connection *con)
1422 {
1423 	int ret = 0;
1424 	const int msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
1425 	struct writequeue_entry *e;
1426 	int len, offset;
1427 	int count = 0;
1428 
1429 	mutex_lock(&con->sock_mutex);
1430 	if (con->sock == NULL)
1431 		goto out_connect;
1432 
1433 	spin_lock(&con->writequeue_lock);
1434 	for (;;) {
1435 		e = list_entry(con->writequeue.next, struct writequeue_entry,
1436 			       list);
1437 		if ((struct list_head *) e == &con->writequeue)
1438 			break;
1439 
1440 		len = e->len;
1441 		offset = e->offset;
1442 		BUG_ON(len == 0 && e->users == 0);
1443 		spin_unlock(&con->writequeue_lock);
1444 
1445 		ret = 0;
1446 		if (len) {
1447 			ret = kernel_sendpage(con->sock, e->page, offset, len,
1448 					      msg_flags);
1449 			if (ret == -EAGAIN || ret == 0) {
1450 				if (ret == -EAGAIN &&
1451 				    test_bit(SOCKWQ_ASYNC_NOSPACE, &con->sock->flags) &&
1452 				    !test_and_set_bit(CF_APP_LIMITED, &con->flags)) {
1453 					/* Notify TCP that we're limited by the
1454 					 * application window size.
1455 					 */
1456 					set_bit(SOCK_NOSPACE, &con->sock->flags);
1457 					con->sock->sk->sk_write_pending++;
1458 				}
1459 				cond_resched();
1460 				goto out;
1461 			} else if (ret < 0)
1462 				goto send_error;
1463 		}
1464 
1465 		/* Don't starve people filling buffers */
1466 		if (++count >= MAX_SEND_MSG_COUNT) {
1467 			cond_resched();
1468 			count = 0;
1469 		}
1470 
1471 		spin_lock(&con->writequeue_lock);
1472 		writequeue_entry_complete(e, ret);
1473 	}
1474 	spin_unlock(&con->writequeue_lock);
1475 out:
1476 	mutex_unlock(&con->sock_mutex);
1477 	return;
1478 
1479 send_error:
1480 	mutex_unlock(&con->sock_mutex);
1481 	close_connection(con, false, false, true);
1482 	/* Requeue the send work. When the work daemon runs again, it will try
1483 	   a new connection, then call this function again. */
1484 	queue_work(send_workqueue, &con->swork);
1485 	return;
1486 
1487 out_connect:
1488 	mutex_unlock(&con->sock_mutex);
1489 	queue_work(send_workqueue, &con->swork);
1490 	cond_resched();
1491 }
1492 
1493 static void clean_one_writequeue(struct connection *con)
1494 {
1495 	struct writequeue_entry *e, *safe;
1496 
1497 	spin_lock(&con->writequeue_lock);
1498 	list_for_each_entry_safe(e, safe, &con->writequeue, list) {
1499 		list_del(&e->list);
1500 		free_entry(e);
1501 	}
1502 	spin_unlock(&con->writequeue_lock);
1503 }
1504 
1505 /* Called from recovery when it knows that a node has
1506    left the cluster */
1507 int dlm_lowcomms_close(int nodeid)
1508 {
1509 	struct connection *con;
1510 	struct dlm_node_addr *na;
1511 
1512 	log_print("closing connection to node %d", nodeid);
1513 	con = nodeid2con(nodeid, 0);
1514 	if (con) {
1515 		set_bit(CF_CLOSE, &con->flags);
1516 		close_connection(con, true, true, true);
1517 		clean_one_writequeue(con);
1518 		if (con->othercon)
1519 			clean_one_writequeue(con->othercon);
1520 	}
1521 
1522 	spin_lock(&dlm_node_addrs_spin);
1523 	na = find_node_addr(nodeid);
1524 	if (na) {
1525 		list_del(&na->list);
1526 		while (na->addr_count--)
1527 			kfree(na->addr[na->addr_count]);
1528 		kfree(na);
1529 	}
1530 	spin_unlock(&dlm_node_addrs_spin);
1531 
1532 	return 0;
1533 }
1534 
1535 /* Receive workqueue function */
1536 static void process_recv_sockets(struct work_struct *work)
1537 {
1538 	struct connection *con = container_of(work, struct connection, rwork);
1539 	int err;
1540 
1541 	clear_bit(CF_READ_PENDING, &con->flags);
1542 	do {
1543 		err = receive_from_sock(con);
1544 	} while (!err);
1545 }
1546 
1547 static void process_listen_recv_socket(struct work_struct *work)
1548 {
1549 	accept_from_sock(&listen_con);
1550 }
1551 
1552 /* Send workqueue function */
1553 static void process_send_sockets(struct work_struct *work)
1554 {
1555 	struct connection *con = container_of(work, struct connection, swork);
1556 
1557 	clear_bit(CF_WRITE_PENDING, &con->flags);
1558 	if (con->sock == NULL) /* not mutex protected so check it inside too */
1559 		con->connect_action(con);
1560 	if (!list_empty(&con->writequeue))
1561 		send_to_sock(con);
1562 }
1563 
1564 static void work_stop(void)
1565 {
1566 	if (recv_workqueue)
1567 		destroy_workqueue(recv_workqueue);
1568 	if (send_workqueue)
1569 		destroy_workqueue(send_workqueue);
1570 }
1571 
1572 static int work_start(void)
1573 {
1574 	recv_workqueue = alloc_workqueue("dlm_recv",
1575 					 WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1576 	if (!recv_workqueue) {
1577 		log_print("can't start dlm_recv");
1578 		return -ENOMEM;
1579 	}
1580 
1581 	send_workqueue = alloc_workqueue("dlm_send",
1582 					 WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1583 	if (!send_workqueue) {
1584 		log_print("can't start dlm_send");
1585 		destroy_workqueue(recv_workqueue);
1586 		return -ENOMEM;
1587 	}
1588 
1589 	return 0;
1590 }
1591 
1592 static void _stop_conn(struct connection *con, bool and_other)
1593 {
1594 	mutex_lock(&con->sock_mutex);
1595 	set_bit(CF_CLOSE, &con->flags);
1596 	set_bit(CF_READ_PENDING, &con->flags);
1597 	set_bit(CF_WRITE_PENDING, &con->flags);
1598 	if (con->sock && con->sock->sk) {
1599 		write_lock_bh(&con->sock->sk->sk_callback_lock);
1600 		con->sock->sk->sk_user_data = NULL;
1601 		write_unlock_bh(&con->sock->sk->sk_callback_lock);
1602 	}
1603 	if (con->othercon && and_other)
1604 		_stop_conn(con->othercon, false);
1605 	mutex_unlock(&con->sock_mutex);
1606 }
1607 
1608 static void stop_conn(struct connection *con)
1609 {
1610 	_stop_conn(con, true);
1611 }
1612 
1613 static void shutdown_conn(struct connection *con)
1614 {
1615 	if (con->shutdown_action)
1616 		con->shutdown_action(con);
1617 }
1618 
1619 static void connection_release(struct rcu_head *rcu)
1620 {
1621 	struct connection *con = container_of(rcu, struct connection, rcu);
1622 
1623 	kfree(con->rx_buf);
1624 	kfree(con);
1625 }
1626 
1627 static void free_conn(struct connection *con)
1628 {
1629 	close_connection(con, true, true, true);
1630 	spin_lock(&connections_lock);
1631 	hlist_del_rcu(&con->list);
1632 	spin_unlock(&connections_lock);
1633 	if (con->othercon) {
1634 		clean_one_writequeue(con->othercon);
1635 		call_srcu(&connections_srcu, &con->othercon->rcu,
1636 			  connection_release);
1637 	}
1638 	clean_one_writequeue(con);
1639 	call_srcu(&connections_srcu, &con->rcu, connection_release);
1640 }
1641 
1642 static void work_flush(void)
1643 {
1644 	int ok, idx;
1645 	int i;
1646 	struct connection *con;
1647 
1648 	do {
1649 		ok = 1;
1650 		foreach_conn(stop_conn);
1651 		if (recv_workqueue)
1652 			flush_workqueue(recv_workqueue);
1653 		if (send_workqueue)
1654 			flush_workqueue(send_workqueue);
1655 		idx = srcu_read_lock(&connections_srcu);
1656 		for (i = 0; i < CONN_HASH_SIZE && ok; i++) {
1657 			hlist_for_each_entry_rcu(con, &connection_hash[i],
1658 						 list) {
1659 				ok &= test_bit(CF_READ_PENDING, &con->flags);
1660 				ok &= test_bit(CF_WRITE_PENDING, &con->flags);
1661 				if (con->othercon) {
1662 					ok &= test_bit(CF_READ_PENDING,
1663 						       &con->othercon->flags);
1664 					ok &= test_bit(CF_WRITE_PENDING,
1665 						       &con->othercon->flags);
1666 				}
1667 			}
1668 		}
1669 		srcu_read_unlock(&connections_srcu, idx);
1670 	} while (!ok);
1671 }
1672 
1673 void dlm_lowcomms_stop(void)
1674 {
1675 	/* Set all the flags to prevent any
1676 	   socket activity.
1677 	*/
1678 	dlm_allow_conn = 0;
1679 
1680 	if (recv_workqueue)
1681 		flush_workqueue(recv_workqueue);
1682 	if (send_workqueue)
1683 		flush_workqueue(send_workqueue);
1684 
1685 	dlm_close_sock(&listen_con.sock);
1686 
1687 	foreach_conn(shutdown_conn);
1688 	work_flush();
1689 	foreach_conn(free_conn);
1690 	work_stop();
1691 	deinit_local();
1692 }
1693 
1694 int dlm_lowcomms_start(void)
1695 {
1696 	int error = -EINVAL;
1697 	int i;
1698 
1699 	for (i = 0; i < CONN_HASH_SIZE; i++)
1700 		INIT_HLIST_HEAD(&connection_hash[i]);
1701 
1702 	init_local();
1703 	if (!dlm_local_count) {
1704 		error = -ENOTCONN;
1705 		log_print("no local IP address has been set");
1706 		goto fail;
1707 	}
1708 
1709 	INIT_WORK(&listen_con.rwork, process_listen_recv_socket);
1710 
1711 	error = work_start();
1712 	if (error)
1713 		goto fail;
1714 
1715 	dlm_allow_conn = 1;
1716 
1717 	/* Start listening */
1718 	if (dlm_config.ci_protocol == 0)
1719 		error = tcp_listen_for_all();
1720 	else
1721 		error = sctp_listen_for_all(&listen_con);
1722 	if (error)
1723 		goto fail_unlisten;
1724 
1725 	return 0;
1726 
1727 fail_unlisten:
1728 	dlm_allow_conn = 0;
1729 	dlm_close_sock(&listen_con.sock);
1730 fail:
1731 	return error;
1732 }
1733 
1734 void dlm_lowcomms_exit(void)
1735 {
1736 	struct dlm_node_addr *na, *safe;
1737 
1738 	spin_lock(&dlm_node_addrs_spin);
1739 	list_for_each_entry_safe(na, safe, &dlm_node_addrs, list) {
1740 		list_del(&na->list);
1741 		while (na->addr_count--)
1742 			kfree(na->addr[na->addr_count]);
1743 		kfree(na);
1744 	}
1745 	spin_unlock(&dlm_node_addrs_spin);
1746 }
1747