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 struct list_head writequeue; /* List of outgoing writequeue_entries */
82 spinlock_t writequeue_lock;
83 int (*rx_action) (struct connection *); /* What to do when active */
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 /* An entry waiting to be sent */
101 struct writequeue_entry {
102 struct list_head list;
103 struct page *page;
104 int offset;
105 int len;
106 int end;
107 int users;
108 struct connection *con;
109 };
110
111 struct dlm_node_addr {
112 struct list_head list;
113 int nodeid;
114 int addr_count;
115 int curr_addr_index;
116 struct sockaddr_storage *addr[DLM_MAX_ADDR_COUNT];
117 };
118
119 static struct listen_sock_callbacks {
120 void (*sk_error_report)(struct sock *);
121 void (*sk_data_ready)(struct sock *);
122 void (*sk_state_change)(struct sock *);
123 void (*sk_write_space)(struct sock *);
124 } listen_sock;
125
126 static LIST_HEAD(dlm_node_addrs);
127 static DEFINE_SPINLOCK(dlm_node_addrs_spin);
128
129 static struct sockaddr_storage *dlm_local_addr[DLM_MAX_ADDR_COUNT];
130 static int dlm_local_count;
131 static int dlm_allow_conn;
132
133 /* Work queues */
134 static struct workqueue_struct *recv_workqueue;
135 static struct workqueue_struct *send_workqueue;
136
137 static struct hlist_head connection_hash[CONN_HASH_SIZE];
138 static DEFINE_SPINLOCK(connections_lock);
139 DEFINE_STATIC_SRCU(connections_srcu);
140
141 static void process_recv_sockets(struct work_struct *work);
142 static void process_send_sockets(struct work_struct *work);
143
144
145 /* This is deliberately very simple because most clusters have simple
146 sequential nodeids, so we should be able to go straight to a connection
147 struct in the array */
nodeid_hash(int nodeid)148 static inline int nodeid_hash(int nodeid)
149 {
150 return nodeid & (CONN_HASH_SIZE-1);
151 }
152
__find_con(int nodeid)153 static struct connection *__find_con(int nodeid)
154 {
155 int r, idx;
156 struct connection *con;
157
158 r = nodeid_hash(nodeid);
159
160 idx = srcu_read_lock(&connections_srcu);
161 hlist_for_each_entry_rcu(con, &connection_hash[r], list) {
162 if (con->nodeid == nodeid) {
163 srcu_read_unlock(&connections_srcu, idx);
164 return con;
165 }
166 }
167 srcu_read_unlock(&connections_srcu, idx);
168
169 return NULL;
170 }
171
172 /*
173 * If 'allocation' is zero then we don't attempt to create a new
174 * connection structure for this node.
175 */
nodeid2con(int nodeid,gfp_t alloc)176 static struct connection *nodeid2con(int nodeid, gfp_t alloc)
177 {
178 struct connection *con, *tmp;
179 int r;
180
181 con = __find_con(nodeid);
182 if (con || !alloc)
183 return con;
184
185 con = kzalloc(sizeof(*con), alloc);
186 if (!con)
187 return NULL;
188
189 con->rx_buflen = dlm_config.ci_buffer_size;
190 con->rx_buf = kmalloc(con->rx_buflen, GFP_NOFS);
191 if (!con->rx_buf) {
192 kfree(con);
193 return NULL;
194 }
195
196 con->nodeid = nodeid;
197 mutex_init(&con->sock_mutex);
198 INIT_LIST_HEAD(&con->writequeue);
199 spin_lock_init(&con->writequeue_lock);
200 INIT_WORK(&con->swork, process_send_sockets);
201 INIT_WORK(&con->rwork, process_recv_sockets);
202 init_waitqueue_head(&con->shutdown_wait);
203
204 /* Setup action pointers for child sockets */
205 if (con->nodeid) {
206 struct connection *zerocon = __find_con(0);
207
208 con->connect_action = zerocon->connect_action;
209 if (!con->rx_action)
210 con->rx_action = zerocon->rx_action;
211 }
212
213 r = nodeid_hash(nodeid);
214
215 spin_lock(&connections_lock);
216 /* Because multiple workqueues/threads calls this function it can
217 * race on multiple cpu's. Instead of locking hot path __find_con()
218 * we just check in rare cases of recently added nodes again
219 * under protection of connections_lock. If this is the case we
220 * abort our connection creation and return the existing connection.
221 */
222 tmp = __find_con(nodeid);
223 if (tmp) {
224 spin_unlock(&connections_lock);
225 kfree(con->rx_buf);
226 kfree(con);
227 return tmp;
228 }
229
230 hlist_add_head_rcu(&con->list, &connection_hash[r]);
231 spin_unlock(&connections_lock);
232
233 return con;
234 }
235
236 /* Loop round all connections */
foreach_conn(void (* conn_func)(struct connection * c))237 static void foreach_conn(void (*conn_func)(struct connection *c))
238 {
239 int i, idx;
240 struct connection *con;
241
242 idx = srcu_read_lock(&connections_srcu);
243 for (i = 0; i < CONN_HASH_SIZE; i++) {
244 hlist_for_each_entry_rcu(con, &connection_hash[i], list)
245 conn_func(con);
246 }
247 srcu_read_unlock(&connections_srcu, idx);
248 }
249
find_node_addr(int nodeid)250 static struct dlm_node_addr *find_node_addr(int nodeid)
251 {
252 struct dlm_node_addr *na;
253
254 list_for_each_entry(na, &dlm_node_addrs, list) {
255 if (na->nodeid == nodeid)
256 return na;
257 }
258 return NULL;
259 }
260
addr_compare(struct sockaddr_storage * x,struct sockaddr_storage * y)261 static int addr_compare(struct sockaddr_storage *x, struct sockaddr_storage *y)
262 {
263 switch (x->ss_family) {
264 case AF_INET: {
265 struct sockaddr_in *sinx = (struct sockaddr_in *)x;
266 struct sockaddr_in *siny = (struct sockaddr_in *)y;
267 if (sinx->sin_addr.s_addr != siny->sin_addr.s_addr)
268 return 0;
269 if (sinx->sin_port != siny->sin_port)
270 return 0;
271 break;
272 }
273 case AF_INET6: {
274 struct sockaddr_in6 *sinx = (struct sockaddr_in6 *)x;
275 struct sockaddr_in6 *siny = (struct sockaddr_in6 *)y;
276 if (!ipv6_addr_equal(&sinx->sin6_addr, &siny->sin6_addr))
277 return 0;
278 if (sinx->sin6_port != siny->sin6_port)
279 return 0;
280 break;
281 }
282 default:
283 return 0;
284 }
285 return 1;
286 }
287
nodeid_to_addr(int nodeid,struct sockaddr_storage * sas_out,struct sockaddr * sa_out,bool try_new_addr)288 static int nodeid_to_addr(int nodeid, struct sockaddr_storage *sas_out,
289 struct sockaddr *sa_out, bool try_new_addr)
290 {
291 struct sockaddr_storage sas;
292 struct dlm_node_addr *na;
293
294 if (!dlm_local_count)
295 return -1;
296
297 spin_lock(&dlm_node_addrs_spin);
298 na = find_node_addr(nodeid);
299 if (na && na->addr_count) {
300 memcpy(&sas, na->addr[na->curr_addr_index],
301 sizeof(struct sockaddr_storage));
302
303 if (try_new_addr) {
304 na->curr_addr_index++;
305 if (na->curr_addr_index == na->addr_count)
306 na->curr_addr_index = 0;
307 }
308 }
309 spin_unlock(&dlm_node_addrs_spin);
310
311 if (!na)
312 return -EEXIST;
313
314 if (!na->addr_count)
315 return -ENOENT;
316
317 if (sas_out)
318 memcpy(sas_out, &sas, sizeof(struct sockaddr_storage));
319
320 if (!sa_out)
321 return 0;
322
323 if (dlm_local_addr[0]->ss_family == AF_INET) {
324 struct sockaddr_in *in4 = (struct sockaddr_in *) &sas;
325 struct sockaddr_in *ret4 = (struct sockaddr_in *) sa_out;
326 ret4->sin_addr.s_addr = in4->sin_addr.s_addr;
327 } else {
328 struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) &sas;
329 struct sockaddr_in6 *ret6 = (struct sockaddr_in6 *) sa_out;
330 ret6->sin6_addr = in6->sin6_addr;
331 }
332
333 return 0;
334 }
335
addr_to_nodeid(struct sockaddr_storage * addr,int * nodeid)336 static int addr_to_nodeid(struct sockaddr_storage *addr, int *nodeid)
337 {
338 struct dlm_node_addr *na;
339 int rv = -EEXIST;
340 int addr_i;
341
342 spin_lock(&dlm_node_addrs_spin);
343 list_for_each_entry(na, &dlm_node_addrs, list) {
344 if (!na->addr_count)
345 continue;
346
347 for (addr_i = 0; addr_i < na->addr_count; addr_i++) {
348 if (addr_compare(na->addr[addr_i], addr)) {
349 *nodeid = na->nodeid;
350 rv = 0;
351 goto unlock;
352 }
353 }
354 }
355 unlock:
356 spin_unlock(&dlm_node_addrs_spin);
357 return rv;
358 }
359
dlm_lowcomms_addr(int nodeid,struct sockaddr_storage * addr,int len)360 int dlm_lowcomms_addr(int nodeid, struct sockaddr_storage *addr, int len)
361 {
362 struct sockaddr_storage *new_addr;
363 struct dlm_node_addr *new_node, *na;
364
365 new_node = kzalloc(sizeof(struct dlm_node_addr), GFP_NOFS);
366 if (!new_node)
367 return -ENOMEM;
368
369 new_addr = kzalloc(sizeof(struct sockaddr_storage), GFP_NOFS);
370 if (!new_addr) {
371 kfree(new_node);
372 return -ENOMEM;
373 }
374
375 memcpy(new_addr, addr, len);
376
377 spin_lock(&dlm_node_addrs_spin);
378 na = find_node_addr(nodeid);
379 if (!na) {
380 new_node->nodeid = nodeid;
381 new_node->addr[0] = new_addr;
382 new_node->addr_count = 1;
383 list_add(&new_node->list, &dlm_node_addrs);
384 spin_unlock(&dlm_node_addrs_spin);
385 return 0;
386 }
387
388 if (na->addr_count >= DLM_MAX_ADDR_COUNT) {
389 spin_unlock(&dlm_node_addrs_spin);
390 kfree(new_addr);
391 kfree(new_node);
392 return -ENOSPC;
393 }
394
395 na->addr[na->addr_count++] = new_addr;
396 spin_unlock(&dlm_node_addrs_spin);
397 kfree(new_node);
398 return 0;
399 }
400
401 /* Data available on socket or listen socket received a connect */
lowcomms_data_ready(struct sock * sk)402 static void lowcomms_data_ready(struct sock *sk)
403 {
404 struct connection *con;
405
406 read_lock_bh(&sk->sk_callback_lock);
407 con = sock2con(sk);
408 if (con && !test_and_set_bit(CF_READ_PENDING, &con->flags))
409 queue_work(recv_workqueue, &con->rwork);
410 read_unlock_bh(&sk->sk_callback_lock);
411 }
412
lowcomms_write_space(struct sock * sk)413 static void lowcomms_write_space(struct sock *sk)
414 {
415 struct connection *con;
416
417 read_lock_bh(&sk->sk_callback_lock);
418 con = sock2con(sk);
419 if (!con)
420 goto out;
421
422 clear_bit(SOCK_NOSPACE, &con->sock->flags);
423
424 if (test_and_clear_bit(CF_APP_LIMITED, &con->flags)) {
425 con->sock->sk->sk_write_pending--;
426 clear_bit(SOCKWQ_ASYNC_NOSPACE, &con->sock->flags);
427 }
428
429 queue_work(send_workqueue, &con->swork);
430 out:
431 read_unlock_bh(&sk->sk_callback_lock);
432 }
433
lowcomms_connect_sock(struct connection * con)434 static inline void lowcomms_connect_sock(struct connection *con)
435 {
436 if (test_bit(CF_CLOSE, &con->flags))
437 return;
438 queue_work(send_workqueue, &con->swork);
439 cond_resched();
440 }
441
lowcomms_state_change(struct sock * sk)442 static void lowcomms_state_change(struct sock *sk)
443 {
444 /* SCTP layer is not calling sk_data_ready when the connection
445 * is done, so we catch the signal through here. Also, it
446 * doesn't switch socket state when entering shutdown, so we
447 * skip the write in that case.
448 */
449 if (sk->sk_shutdown) {
450 if (sk->sk_shutdown == RCV_SHUTDOWN)
451 lowcomms_data_ready(sk);
452 } else if (sk->sk_state == TCP_ESTABLISHED) {
453 lowcomms_write_space(sk);
454 }
455 }
456
dlm_lowcomms_connect_node(int nodeid)457 int dlm_lowcomms_connect_node(int nodeid)
458 {
459 struct connection *con;
460
461 if (nodeid == dlm_our_nodeid())
462 return 0;
463
464 con = nodeid2con(nodeid, GFP_NOFS);
465 if (!con)
466 return -ENOMEM;
467 lowcomms_connect_sock(con);
468 return 0;
469 }
470
lowcomms_error_report(struct sock * sk)471 static void lowcomms_error_report(struct sock *sk)
472 {
473 struct connection *con;
474 void (*orig_report)(struct sock *) = NULL;
475 struct inet_sock *inet;
476
477 read_lock_bh(&sk->sk_callback_lock);
478 con = sock2con(sk);
479 if (con == NULL)
480 goto out;
481
482 orig_report = listen_sock.sk_error_report;
483
484 inet = inet_sk(sk);
485 switch (sk->sk_family) {
486 case AF_INET:
487 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
488 "sending to node %d at %pI4, dport %d, "
489 "sk_err=%d/%d\n", dlm_our_nodeid(),
490 con->nodeid, &inet->inet_daddr,
491 ntohs(inet->inet_dport), sk->sk_err,
492 sk->sk_err_soft);
493 break;
494 #if IS_ENABLED(CONFIG_IPV6)
495 case AF_INET6:
496 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
497 "sending to node %d at %pI6c, "
498 "dport %d, sk_err=%d/%d\n", dlm_our_nodeid(),
499 con->nodeid, &sk->sk_v6_daddr,
500 ntohs(inet->inet_dport), sk->sk_err,
501 sk->sk_err_soft);
502 break;
503 #endif
504 default:
505 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
506 "invalid socket family %d set, "
507 "sk_err=%d/%d\n", dlm_our_nodeid(),
508 sk->sk_family, sk->sk_err, sk->sk_err_soft);
509 goto out;
510 }
511 out:
512 read_unlock_bh(&sk->sk_callback_lock);
513 if (orig_report)
514 orig_report(sk);
515 }
516
517 /* Note: sk_callback_lock must be locked before calling this function. */
save_listen_callbacks(struct socket * sock)518 static void save_listen_callbacks(struct socket *sock)
519 {
520 struct sock *sk = sock->sk;
521
522 listen_sock.sk_data_ready = sk->sk_data_ready;
523 listen_sock.sk_state_change = sk->sk_state_change;
524 listen_sock.sk_write_space = sk->sk_write_space;
525 listen_sock.sk_error_report = sk->sk_error_report;
526 }
527
restore_callbacks(struct socket * sock)528 static void restore_callbacks(struct socket *sock)
529 {
530 struct sock *sk = sock->sk;
531
532 write_lock_bh(&sk->sk_callback_lock);
533 sk->sk_user_data = NULL;
534 sk->sk_data_ready = listen_sock.sk_data_ready;
535 sk->sk_state_change = listen_sock.sk_state_change;
536 sk->sk_write_space = listen_sock.sk_write_space;
537 sk->sk_error_report = listen_sock.sk_error_report;
538 write_unlock_bh(&sk->sk_callback_lock);
539 }
540
541 /* Make a socket active */
add_sock(struct socket * sock,struct connection * con)542 static void add_sock(struct socket *sock, struct connection *con)
543 {
544 struct sock *sk = sock->sk;
545
546 write_lock_bh(&sk->sk_callback_lock);
547 con->sock = sock;
548
549 sk->sk_user_data = con;
550 /* Install a data_ready callback */
551 sk->sk_data_ready = lowcomms_data_ready;
552 sk->sk_write_space = lowcomms_write_space;
553 sk->sk_state_change = lowcomms_state_change;
554 sk->sk_allocation = GFP_NOFS;
555 sk->sk_error_report = lowcomms_error_report;
556 write_unlock_bh(&sk->sk_callback_lock);
557 }
558
559 /* Add the port number to an IPv6 or 4 sockaddr and return the address
560 length */
make_sockaddr(struct sockaddr_storage * saddr,uint16_t port,int * addr_len)561 static void make_sockaddr(struct sockaddr_storage *saddr, uint16_t port,
562 int *addr_len)
563 {
564 saddr->ss_family = dlm_local_addr[0]->ss_family;
565 if (saddr->ss_family == AF_INET) {
566 struct sockaddr_in *in4_addr = (struct sockaddr_in *)saddr;
567 in4_addr->sin_port = cpu_to_be16(port);
568 *addr_len = sizeof(struct sockaddr_in);
569 memset(&in4_addr->sin_zero, 0, sizeof(in4_addr->sin_zero));
570 } else {
571 struct sockaddr_in6 *in6_addr = (struct sockaddr_in6 *)saddr;
572 in6_addr->sin6_port = cpu_to_be16(port);
573 *addr_len = sizeof(struct sockaddr_in6);
574 }
575 memset((char *)saddr + *addr_len, 0, sizeof(struct sockaddr_storage) - *addr_len);
576 }
577
578 /* Close a remote connection and tidy up */
close_connection(struct connection * con,bool and_other,bool tx,bool rx)579 static void close_connection(struct connection *con, bool and_other,
580 bool tx, bool rx)
581 {
582 bool closing = test_and_set_bit(CF_CLOSING, &con->flags);
583
584 if (tx && !closing && cancel_work_sync(&con->swork)) {
585 log_print("canceled swork for node %d", con->nodeid);
586 clear_bit(CF_WRITE_PENDING, &con->flags);
587 }
588 if (rx && !closing && cancel_work_sync(&con->rwork)) {
589 log_print("canceled rwork for node %d", con->nodeid);
590 clear_bit(CF_READ_PENDING, &con->flags);
591 }
592
593 mutex_lock(&con->sock_mutex);
594 if (con->sock) {
595 restore_callbacks(con->sock);
596 sock_release(con->sock);
597 con->sock = NULL;
598 }
599 if (con->othercon && and_other) {
600 /* Will only re-enter once. */
601 close_connection(con->othercon, false, tx, rx);
602 }
603
604 con->rx_leftover = 0;
605 con->retries = 0;
606 mutex_unlock(&con->sock_mutex);
607 clear_bit(CF_CLOSING, &con->flags);
608 }
609
shutdown_connection(struct connection * con)610 static void shutdown_connection(struct connection *con)
611 {
612 int ret;
613
614 flush_work(&con->swork);
615
616 mutex_lock(&con->sock_mutex);
617 /* nothing to shutdown */
618 if (!con->sock) {
619 mutex_unlock(&con->sock_mutex);
620 return;
621 }
622
623 set_bit(CF_SHUTDOWN, &con->flags);
624 ret = kernel_sock_shutdown(con->sock, SHUT_WR);
625 mutex_unlock(&con->sock_mutex);
626 if (ret) {
627 log_print("Connection %p failed to shutdown: %d will force close",
628 con, ret);
629 goto force_close;
630 } else {
631 ret = wait_event_timeout(con->shutdown_wait,
632 !test_bit(CF_SHUTDOWN, &con->flags),
633 DLM_SHUTDOWN_WAIT_TIMEOUT);
634 if (ret == 0) {
635 log_print("Connection %p shutdown timed out, will force close",
636 con);
637 goto force_close;
638 }
639 }
640
641 return;
642
643 force_close:
644 clear_bit(CF_SHUTDOWN, &con->flags);
645 close_connection(con, false, true, true);
646 }
647
dlm_tcp_shutdown(struct connection * con)648 static void dlm_tcp_shutdown(struct connection *con)
649 {
650 if (con->othercon)
651 shutdown_connection(con->othercon);
652 shutdown_connection(con);
653 }
654
con_realloc_receive_buf(struct connection * con,int newlen)655 static int con_realloc_receive_buf(struct connection *con, int newlen)
656 {
657 unsigned char *newbuf;
658
659 newbuf = kmalloc(newlen, GFP_NOFS);
660 if (!newbuf)
661 return -ENOMEM;
662
663 /* copy any leftover from last receive */
664 if (con->rx_leftover)
665 memmove(newbuf, con->rx_buf, con->rx_leftover);
666
667 /* swap to new buffer space */
668 kfree(con->rx_buf);
669 con->rx_buflen = newlen;
670 con->rx_buf = newbuf;
671
672 return 0;
673 }
674
675 /* Data received from remote end */
receive_from_sock(struct connection * con)676 static int receive_from_sock(struct connection *con)
677 {
678 int call_again_soon = 0;
679 struct msghdr msg;
680 struct kvec iov;
681 int ret, buflen;
682
683 mutex_lock(&con->sock_mutex);
684
685 if (con->sock == NULL) {
686 ret = -EAGAIN;
687 goto out_close;
688 }
689
690 if (con->nodeid == 0) {
691 ret = -EINVAL;
692 goto out_close;
693 }
694
695 /* realloc if we get new buffer size to read out */
696 buflen = dlm_config.ci_buffer_size;
697 if (con->rx_buflen != buflen && con->rx_leftover <= buflen) {
698 ret = con_realloc_receive_buf(con, buflen);
699 if (ret < 0)
700 goto out_resched;
701 }
702
703 /* calculate new buffer parameter regarding last receive and
704 * possible leftover bytes
705 */
706 iov.iov_base = con->rx_buf + con->rx_leftover;
707 iov.iov_len = con->rx_buflen - con->rx_leftover;
708
709 memset(&msg, 0, sizeof(msg));
710 msg.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
711 ret = kernel_recvmsg(con->sock, &msg, &iov, 1, iov.iov_len,
712 msg.msg_flags);
713 if (ret <= 0)
714 goto out_close;
715 else if (ret == iov.iov_len)
716 call_again_soon = 1;
717
718 /* new buflen according readed bytes and leftover from last receive */
719 buflen = ret + con->rx_leftover;
720 ret = dlm_process_incoming_buffer(con->nodeid, con->rx_buf, buflen);
721 if (ret < 0)
722 goto out_close;
723
724 /* calculate leftover bytes from process and put it into begin of
725 * the receive buffer, so next receive we have the full message
726 * at the start address of the receive buffer.
727 */
728 con->rx_leftover = buflen - ret;
729 if (con->rx_leftover) {
730 memmove(con->rx_buf, con->rx_buf + ret,
731 con->rx_leftover);
732 call_again_soon = true;
733 }
734
735 if (call_again_soon)
736 goto out_resched;
737
738 mutex_unlock(&con->sock_mutex);
739 return 0;
740
741 out_resched:
742 if (!test_and_set_bit(CF_READ_PENDING, &con->flags))
743 queue_work(recv_workqueue, &con->rwork);
744 mutex_unlock(&con->sock_mutex);
745 return -EAGAIN;
746
747 out_close:
748 mutex_unlock(&con->sock_mutex);
749 if (ret != -EAGAIN) {
750 /* Reconnect when there is something to send */
751 close_connection(con, false, true, false);
752 if (ret == 0) {
753 log_print("connection %p got EOF from %d",
754 con, con->nodeid);
755 /* handling for tcp shutdown */
756 clear_bit(CF_SHUTDOWN, &con->flags);
757 wake_up(&con->shutdown_wait);
758 /* signal to breaking receive worker */
759 ret = -1;
760 }
761 }
762 return ret;
763 }
764
765 /* Listening socket is busy, accept a connection */
accept_from_sock(struct connection * con)766 static int accept_from_sock(struct connection *con)
767 {
768 int result;
769 struct sockaddr_storage peeraddr;
770 struct socket *newsock;
771 int len;
772 int nodeid;
773 struct connection *newcon;
774 struct connection *addcon;
775 unsigned int mark;
776
777 if (!dlm_allow_conn) {
778 return -1;
779 }
780
781 mutex_lock_nested(&con->sock_mutex, 0);
782
783 if (!con->sock) {
784 mutex_unlock(&con->sock_mutex);
785 return -ENOTCONN;
786 }
787
788 result = kernel_accept(con->sock, &newsock, O_NONBLOCK);
789 if (result < 0)
790 goto accept_err;
791
792 /* Get the connected socket's peer */
793 memset(&peeraddr, 0, sizeof(peeraddr));
794 len = newsock->ops->getname(newsock, (struct sockaddr *)&peeraddr, 2);
795 if (len < 0) {
796 result = -ECONNABORTED;
797 goto accept_err;
798 }
799
800 /* Get the new node's NODEID */
801 make_sockaddr(&peeraddr, 0, &len);
802 if (addr_to_nodeid(&peeraddr, &nodeid)) {
803 unsigned char *b=(unsigned char *)&peeraddr;
804 log_print("connect from non cluster node");
805 print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
806 b, sizeof(struct sockaddr_storage));
807 sock_release(newsock);
808 mutex_unlock(&con->sock_mutex);
809 return -1;
810 }
811
812 dlm_comm_mark(nodeid, &mark);
813 sock_set_mark(newsock->sk, mark);
814
815 log_print("got connection from %d", nodeid);
816
817 /* Check to see if we already have a connection to this node. This
818 * could happen if the two nodes initiate a connection at roughly
819 * the same time and the connections cross on the wire.
820 * In this case we store the incoming one in "othercon"
821 */
822 newcon = nodeid2con(nodeid, GFP_NOFS);
823 if (!newcon) {
824 result = -ENOMEM;
825 goto accept_err;
826 }
827 mutex_lock_nested(&newcon->sock_mutex, 1);
828 if (newcon->sock) {
829 struct connection *othercon = newcon->othercon;
830
831 if (!othercon) {
832 othercon = kzalloc(sizeof(*othercon), GFP_NOFS);
833 if (!othercon) {
834 log_print("failed to allocate incoming socket");
835 mutex_unlock(&newcon->sock_mutex);
836 result = -ENOMEM;
837 goto accept_err;
838 }
839
840 othercon->rx_buflen = dlm_config.ci_buffer_size;
841 othercon->rx_buf = kmalloc(othercon->rx_buflen, GFP_NOFS);
842 if (!othercon->rx_buf) {
843 mutex_unlock(&newcon->sock_mutex);
844 kfree(othercon);
845 log_print("failed to allocate incoming socket receive buffer");
846 result = -ENOMEM;
847 goto accept_err;
848 }
849
850 othercon->nodeid = nodeid;
851 othercon->rx_action = receive_from_sock;
852 mutex_init(&othercon->sock_mutex);
853 INIT_LIST_HEAD(&othercon->writequeue);
854 spin_lock_init(&othercon->writequeue_lock);
855 INIT_WORK(&othercon->swork, process_send_sockets);
856 INIT_WORK(&othercon->rwork, process_recv_sockets);
857 init_waitqueue_head(&othercon->shutdown_wait);
858 set_bit(CF_IS_OTHERCON, &othercon->flags);
859 } else {
860 /* close other sock con if we have something new */
861 close_connection(othercon, false, true, false);
862 }
863
864 mutex_lock_nested(&othercon->sock_mutex, 2);
865 newcon->othercon = othercon;
866 add_sock(newsock, othercon);
867 addcon = othercon;
868 mutex_unlock(&othercon->sock_mutex);
869 }
870 else {
871 newcon->rx_action = receive_from_sock;
872 /* accept copies the sk after we've saved the callbacks, so we
873 don't want to save them a second time or comm errors will
874 result in calling sk_error_report recursively. */
875 add_sock(newsock, newcon);
876 addcon = newcon;
877 }
878
879 mutex_unlock(&newcon->sock_mutex);
880
881 /*
882 * Add it to the active queue in case we got data
883 * between processing the accept adding the socket
884 * to the read_sockets list
885 */
886 if (!test_and_set_bit(CF_READ_PENDING, &addcon->flags))
887 queue_work(recv_workqueue, &addcon->rwork);
888 mutex_unlock(&con->sock_mutex);
889
890 return 0;
891
892 accept_err:
893 mutex_unlock(&con->sock_mutex);
894 if (newsock)
895 sock_release(newsock);
896
897 if (result != -EAGAIN)
898 log_print("error accepting connection from node: %d", result);
899 return result;
900 }
901
free_entry(struct writequeue_entry * e)902 static void free_entry(struct writequeue_entry *e)
903 {
904 __free_page(e->page);
905 kfree(e);
906 }
907
908 /*
909 * writequeue_entry_complete - try to delete and free write queue entry
910 * @e: write queue entry to try to delete
911 * @completed: bytes completed
912 *
913 * writequeue_lock must be held.
914 */
writequeue_entry_complete(struct writequeue_entry * e,int completed)915 static void writequeue_entry_complete(struct writequeue_entry *e, int completed)
916 {
917 e->offset += completed;
918 e->len -= completed;
919
920 if (e->len == 0 && e->users == 0) {
921 list_del(&e->list);
922 free_entry(e);
923 }
924 }
925
926 /*
927 * sctp_bind_addrs - bind a SCTP socket to all our addresses
928 */
sctp_bind_addrs(struct connection * con,uint16_t port)929 static int sctp_bind_addrs(struct connection *con, uint16_t port)
930 {
931 struct sockaddr_storage localaddr;
932 struct sockaddr *addr = (struct sockaddr *)&localaddr;
933 int i, addr_len, result = 0;
934
935 for (i = 0; i < dlm_local_count; i++) {
936 memcpy(&localaddr, dlm_local_addr[i], sizeof(localaddr));
937 make_sockaddr(&localaddr, port, &addr_len);
938
939 if (!i)
940 result = kernel_bind(con->sock, addr, addr_len);
941 else
942 result = sock_bind_add(con->sock->sk, addr, addr_len);
943
944 if (result < 0) {
945 log_print("Can't bind to %d addr number %d, %d.\n",
946 port, i + 1, result);
947 break;
948 }
949 }
950 return result;
951 }
952
953 /* Initiate an SCTP association.
954 This is a special case of send_to_sock() in that we don't yet have a
955 peeled-off socket for this association, so we use the listening socket
956 and add the primary IP address of the remote node.
957 */
sctp_connect_to_sock(struct connection * con)958 static void sctp_connect_to_sock(struct connection *con)
959 {
960 struct sockaddr_storage daddr;
961 int result;
962 int addr_len;
963 struct socket *sock;
964 unsigned int mark;
965
966 if (con->nodeid == 0) {
967 log_print("attempt to connect sock 0 foiled");
968 return;
969 }
970
971 dlm_comm_mark(con->nodeid, &mark);
972
973 mutex_lock(&con->sock_mutex);
974
975 /* Some odd races can cause double-connects, ignore them */
976 if (con->retries++ > MAX_CONNECT_RETRIES)
977 goto out;
978
979 if (con->sock) {
980 log_print("node %d already connected.", con->nodeid);
981 goto out;
982 }
983
984 memset(&daddr, 0, sizeof(daddr));
985 result = nodeid_to_addr(con->nodeid, &daddr, NULL, true);
986 if (result < 0) {
987 log_print("no address for nodeid %d", con->nodeid);
988 goto out;
989 }
990
991 /* Create a socket to communicate with */
992 result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
993 SOCK_STREAM, IPPROTO_SCTP, &sock);
994 if (result < 0)
995 goto socket_err;
996
997 sock_set_mark(sock->sk, mark);
998
999 con->rx_action = receive_from_sock;
1000 con->connect_action = sctp_connect_to_sock;
1001 add_sock(sock, con);
1002
1003 /* Bind to all addresses. */
1004 if (sctp_bind_addrs(con, 0))
1005 goto bind_err;
1006
1007 make_sockaddr(&daddr, dlm_config.ci_tcp_port, &addr_len);
1008
1009 log_print("connecting to %d", con->nodeid);
1010
1011 /* Turn off Nagle's algorithm */
1012 sctp_sock_set_nodelay(sock->sk);
1013
1014 /*
1015 * Make sock->ops->connect() function return in specified time,
1016 * since O_NONBLOCK argument in connect() function does not work here,
1017 * then, we should restore the default value of this attribute.
1018 */
1019 sock_set_sndtimeo(sock->sk, 5);
1020 result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
1021 0);
1022 sock_set_sndtimeo(sock->sk, 0);
1023
1024 if (result == -EINPROGRESS)
1025 result = 0;
1026 if (result == 0)
1027 goto out;
1028
1029 bind_err:
1030 con->sock = NULL;
1031 sock_release(sock);
1032
1033 socket_err:
1034 /*
1035 * Some errors are fatal and this list might need adjusting. For other
1036 * errors we try again until the max number of retries is reached.
1037 */
1038 if (result != -EHOSTUNREACH &&
1039 result != -ENETUNREACH &&
1040 result != -ENETDOWN &&
1041 result != -EINVAL &&
1042 result != -EPROTONOSUPPORT) {
1043 log_print("connect %d try %d error %d", con->nodeid,
1044 con->retries, result);
1045 mutex_unlock(&con->sock_mutex);
1046 msleep(1000);
1047 lowcomms_connect_sock(con);
1048 return;
1049 }
1050
1051 out:
1052 mutex_unlock(&con->sock_mutex);
1053 }
1054
1055 /* Connect a new socket to its peer */
tcp_connect_to_sock(struct connection * con)1056 static void tcp_connect_to_sock(struct connection *con)
1057 {
1058 struct sockaddr_storage saddr, src_addr;
1059 int addr_len;
1060 struct socket *sock = NULL;
1061 unsigned int mark;
1062 int result;
1063
1064 if (con->nodeid == 0) {
1065 log_print("attempt to connect sock 0 foiled");
1066 return;
1067 }
1068
1069 dlm_comm_mark(con->nodeid, &mark);
1070
1071 mutex_lock(&con->sock_mutex);
1072 if (con->retries++ > MAX_CONNECT_RETRIES)
1073 goto out;
1074
1075 /* Some odd races can cause double-connects, ignore them */
1076 if (con->sock)
1077 goto out;
1078
1079 /* Create a socket to communicate with */
1080 result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1081 SOCK_STREAM, IPPROTO_TCP, &sock);
1082 if (result < 0)
1083 goto out_err;
1084
1085 sock_set_mark(sock->sk, mark);
1086
1087 memset(&saddr, 0, sizeof(saddr));
1088 result = nodeid_to_addr(con->nodeid, &saddr, NULL, false);
1089 if (result < 0) {
1090 log_print("no address for nodeid %d", con->nodeid);
1091 goto out_err;
1092 }
1093
1094 con->rx_action = receive_from_sock;
1095 con->connect_action = tcp_connect_to_sock;
1096 con->shutdown_action = dlm_tcp_shutdown;
1097 add_sock(sock, con);
1098
1099 /* Bind to our cluster-known address connecting to avoid
1100 routing problems */
1101 memcpy(&src_addr, dlm_local_addr[0], sizeof(src_addr));
1102 make_sockaddr(&src_addr, 0, &addr_len);
1103 result = sock->ops->bind(sock, (struct sockaddr *) &src_addr,
1104 addr_len);
1105 if (result < 0) {
1106 log_print("could not bind for connect: %d", result);
1107 /* This *may* not indicate a critical error */
1108 }
1109
1110 make_sockaddr(&saddr, dlm_config.ci_tcp_port, &addr_len);
1111
1112 log_print("connecting to %d", con->nodeid);
1113
1114 /* Turn off Nagle's algorithm */
1115 tcp_sock_set_nodelay(sock->sk);
1116
1117 result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
1118 O_NONBLOCK);
1119 if (result == -EINPROGRESS)
1120 result = 0;
1121 if (result == 0)
1122 goto out;
1123
1124 out_err:
1125 if (con->sock) {
1126 sock_release(con->sock);
1127 con->sock = NULL;
1128 } else if (sock) {
1129 sock_release(sock);
1130 }
1131 /*
1132 * Some errors are fatal and this list might need adjusting. For other
1133 * errors we try again until the max number of retries is reached.
1134 */
1135 if (result != -EHOSTUNREACH &&
1136 result != -ENETUNREACH &&
1137 result != -ENETDOWN &&
1138 result != -EINVAL &&
1139 result != -EPROTONOSUPPORT) {
1140 log_print("connect %d try %d error %d", con->nodeid,
1141 con->retries, result);
1142 mutex_unlock(&con->sock_mutex);
1143 msleep(1000);
1144 lowcomms_connect_sock(con);
1145 return;
1146 }
1147 out:
1148 mutex_unlock(&con->sock_mutex);
1149 return;
1150 }
1151
tcp_create_listen_sock(struct connection * con,struct sockaddr_storage * saddr)1152 static struct socket *tcp_create_listen_sock(struct connection *con,
1153 struct sockaddr_storage *saddr)
1154 {
1155 struct socket *sock = NULL;
1156 int result = 0;
1157 int addr_len;
1158
1159 if (dlm_local_addr[0]->ss_family == AF_INET)
1160 addr_len = sizeof(struct sockaddr_in);
1161 else
1162 addr_len = sizeof(struct sockaddr_in6);
1163
1164 /* Create a socket to communicate with */
1165 result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1166 SOCK_STREAM, IPPROTO_TCP, &sock);
1167 if (result < 0) {
1168 log_print("Can't create listening comms socket");
1169 goto create_out;
1170 }
1171
1172 sock_set_mark(sock->sk, dlm_config.ci_mark);
1173
1174 /* Turn off Nagle's algorithm */
1175 tcp_sock_set_nodelay(sock->sk);
1176
1177 sock_set_reuseaddr(sock->sk);
1178
1179 write_lock_bh(&sock->sk->sk_callback_lock);
1180 sock->sk->sk_user_data = con;
1181 save_listen_callbacks(sock);
1182 con->rx_action = accept_from_sock;
1183 con->connect_action = tcp_connect_to_sock;
1184 write_unlock_bh(&sock->sk->sk_callback_lock);
1185
1186 /* Bind to our port */
1187 make_sockaddr(saddr, dlm_config.ci_tcp_port, &addr_len);
1188 result = sock->ops->bind(sock, (struct sockaddr *) saddr, addr_len);
1189 if (result < 0) {
1190 log_print("Can't bind to port %d", dlm_config.ci_tcp_port);
1191 sock_release(sock);
1192 sock = NULL;
1193 con->sock = NULL;
1194 goto create_out;
1195 }
1196 sock_set_keepalive(sock->sk);
1197
1198 result = sock->ops->listen(sock, 5);
1199 if (result < 0) {
1200 log_print("Can't listen on port %d", dlm_config.ci_tcp_port);
1201 sock_release(sock);
1202 sock = NULL;
1203 goto create_out;
1204 }
1205
1206 create_out:
1207 return sock;
1208 }
1209
1210 /* Get local addresses */
init_local(void)1211 static void init_local(void)
1212 {
1213 struct sockaddr_storage sas, *addr;
1214 int i;
1215
1216 dlm_local_count = 0;
1217 for (i = 0; i < DLM_MAX_ADDR_COUNT; i++) {
1218 if (dlm_our_addr(&sas, i))
1219 break;
1220
1221 addr = kmemdup(&sas, sizeof(*addr), GFP_NOFS);
1222 if (!addr)
1223 break;
1224 dlm_local_addr[dlm_local_count++] = addr;
1225 }
1226 }
1227
deinit_local(void)1228 static void deinit_local(void)
1229 {
1230 int i;
1231
1232 for (i = 0; i < dlm_local_count; i++)
1233 kfree(dlm_local_addr[i]);
1234 }
1235
1236 /* Initialise SCTP socket and bind to all interfaces */
sctp_listen_for_all(void)1237 static int sctp_listen_for_all(void)
1238 {
1239 struct socket *sock = NULL;
1240 int result = -EINVAL;
1241 struct connection *con = nodeid2con(0, GFP_NOFS);
1242
1243 if (!con)
1244 return -ENOMEM;
1245
1246 log_print("Using SCTP for communications");
1247
1248 result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1249 SOCK_STREAM, IPPROTO_SCTP, &sock);
1250 if (result < 0) {
1251 log_print("Can't create comms socket, check SCTP is loaded");
1252 goto out;
1253 }
1254
1255 sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
1256 sock_set_mark(sock->sk, dlm_config.ci_mark);
1257 sctp_sock_set_nodelay(sock->sk);
1258
1259 write_lock_bh(&sock->sk->sk_callback_lock);
1260 /* Init con struct */
1261 sock->sk->sk_user_data = con;
1262 save_listen_callbacks(sock);
1263 con->sock = sock;
1264 con->sock->sk->sk_data_ready = lowcomms_data_ready;
1265 con->rx_action = accept_from_sock;
1266 con->connect_action = sctp_connect_to_sock;
1267
1268 write_unlock_bh(&sock->sk->sk_callback_lock);
1269
1270 /* Bind to all addresses. */
1271 if (sctp_bind_addrs(con, dlm_config.ci_tcp_port))
1272 goto create_delsock;
1273
1274 result = sock->ops->listen(sock, 5);
1275 if (result < 0) {
1276 log_print("Can't set socket listening");
1277 goto create_delsock;
1278 }
1279
1280 return 0;
1281
1282 create_delsock:
1283 sock_release(sock);
1284 con->sock = NULL;
1285 out:
1286 return result;
1287 }
1288
tcp_listen_for_all(void)1289 static int tcp_listen_for_all(void)
1290 {
1291 struct socket *sock = NULL;
1292 struct connection *con = nodeid2con(0, GFP_NOFS);
1293 int result = -EINVAL;
1294
1295 if (!con)
1296 return -ENOMEM;
1297
1298 /* We don't support multi-homed hosts */
1299 if (dlm_local_addr[1] != NULL) {
1300 log_print("TCP protocol can't handle multi-homed hosts, "
1301 "try SCTP");
1302 return -EINVAL;
1303 }
1304
1305 log_print("Using TCP for communications");
1306
1307 sock = tcp_create_listen_sock(con, dlm_local_addr[0]);
1308 if (sock) {
1309 add_sock(sock, con);
1310 result = 0;
1311 }
1312 else {
1313 result = -EADDRINUSE;
1314 }
1315
1316 return result;
1317 }
1318
1319
1320
new_writequeue_entry(struct connection * con,gfp_t allocation)1321 static struct writequeue_entry *new_writequeue_entry(struct connection *con,
1322 gfp_t allocation)
1323 {
1324 struct writequeue_entry *entry;
1325
1326 entry = kmalloc(sizeof(struct writequeue_entry), allocation);
1327 if (!entry)
1328 return NULL;
1329
1330 entry->page = alloc_page(allocation);
1331 if (!entry->page) {
1332 kfree(entry);
1333 return NULL;
1334 }
1335
1336 entry->offset = 0;
1337 entry->len = 0;
1338 entry->end = 0;
1339 entry->users = 0;
1340 entry->con = con;
1341
1342 return entry;
1343 }
1344
dlm_lowcomms_get_buffer(int nodeid,int len,gfp_t allocation,char ** ppc)1345 void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
1346 {
1347 struct connection *con;
1348 struct writequeue_entry *e;
1349 int offset = 0;
1350
1351 con = nodeid2con(nodeid, allocation);
1352 if (!con)
1353 return NULL;
1354
1355 spin_lock(&con->writequeue_lock);
1356 e = list_entry(con->writequeue.prev, struct writequeue_entry, list);
1357 if ((&e->list == &con->writequeue) ||
1358 (PAGE_SIZE - e->end < len)) {
1359 e = NULL;
1360 } else {
1361 offset = e->end;
1362 e->end += len;
1363 e->users++;
1364 }
1365 spin_unlock(&con->writequeue_lock);
1366
1367 if (e) {
1368 got_one:
1369 *ppc = page_address(e->page) + offset;
1370 return e;
1371 }
1372
1373 e = new_writequeue_entry(con, allocation);
1374 if (e) {
1375 spin_lock(&con->writequeue_lock);
1376 offset = e->end;
1377 e->end += len;
1378 e->users++;
1379 list_add_tail(&e->list, &con->writequeue);
1380 spin_unlock(&con->writequeue_lock);
1381 goto got_one;
1382 }
1383 return NULL;
1384 }
1385
dlm_lowcomms_commit_buffer(void * mh)1386 void dlm_lowcomms_commit_buffer(void *mh)
1387 {
1388 struct writequeue_entry *e = (struct writequeue_entry *)mh;
1389 struct connection *con = e->con;
1390 int users;
1391
1392 spin_lock(&con->writequeue_lock);
1393 users = --e->users;
1394 if (users)
1395 goto out;
1396 e->len = e->end - e->offset;
1397 spin_unlock(&con->writequeue_lock);
1398
1399 queue_work(send_workqueue, &con->swork);
1400 return;
1401
1402 out:
1403 spin_unlock(&con->writequeue_lock);
1404 return;
1405 }
1406
1407 /* Send a message */
send_to_sock(struct connection * con)1408 static void send_to_sock(struct connection *con)
1409 {
1410 int ret = 0;
1411 const int msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
1412 struct writequeue_entry *e;
1413 int len, offset;
1414 int count = 0;
1415
1416 mutex_lock(&con->sock_mutex);
1417 if (con->sock == NULL)
1418 goto out_connect;
1419
1420 spin_lock(&con->writequeue_lock);
1421 for (;;) {
1422 e = list_entry(con->writequeue.next, struct writequeue_entry,
1423 list);
1424 if ((struct list_head *) e == &con->writequeue)
1425 break;
1426
1427 len = e->len;
1428 offset = e->offset;
1429 BUG_ON(len == 0 && e->users == 0);
1430 spin_unlock(&con->writequeue_lock);
1431
1432 ret = 0;
1433 if (len) {
1434 ret = kernel_sendpage(con->sock, e->page, offset, len,
1435 msg_flags);
1436 if (ret == -EAGAIN || ret == 0) {
1437 if (ret == -EAGAIN &&
1438 test_bit(SOCKWQ_ASYNC_NOSPACE, &con->sock->flags) &&
1439 !test_and_set_bit(CF_APP_LIMITED, &con->flags)) {
1440 /* Notify TCP that we're limited by the
1441 * application window size.
1442 */
1443 set_bit(SOCK_NOSPACE, &con->sock->flags);
1444 con->sock->sk->sk_write_pending++;
1445 }
1446 cond_resched();
1447 goto out;
1448 } else if (ret < 0)
1449 goto send_error;
1450 }
1451
1452 /* Don't starve people filling buffers */
1453 if (++count >= MAX_SEND_MSG_COUNT) {
1454 cond_resched();
1455 count = 0;
1456 }
1457
1458 spin_lock(&con->writequeue_lock);
1459 writequeue_entry_complete(e, ret);
1460 }
1461 spin_unlock(&con->writequeue_lock);
1462 out:
1463 mutex_unlock(&con->sock_mutex);
1464 return;
1465
1466 send_error:
1467 mutex_unlock(&con->sock_mutex);
1468 close_connection(con, false, false, true);
1469 /* Requeue the send work. When the work daemon runs again, it will try
1470 a new connection, then call this function again. */
1471 queue_work(send_workqueue, &con->swork);
1472 return;
1473
1474 out_connect:
1475 mutex_unlock(&con->sock_mutex);
1476 queue_work(send_workqueue, &con->swork);
1477 cond_resched();
1478 }
1479
clean_one_writequeue(struct connection * con)1480 static void clean_one_writequeue(struct connection *con)
1481 {
1482 struct writequeue_entry *e, *safe;
1483
1484 spin_lock(&con->writequeue_lock);
1485 list_for_each_entry_safe(e, safe, &con->writequeue, list) {
1486 list_del(&e->list);
1487 free_entry(e);
1488 }
1489 spin_unlock(&con->writequeue_lock);
1490 }
1491
1492 /* Called from recovery when it knows that a node has
1493 left the cluster */
dlm_lowcomms_close(int nodeid)1494 int dlm_lowcomms_close(int nodeid)
1495 {
1496 struct connection *con;
1497 struct dlm_node_addr *na;
1498
1499 log_print("closing connection to node %d", nodeid);
1500 con = nodeid2con(nodeid, 0);
1501 if (con) {
1502 set_bit(CF_CLOSE, &con->flags);
1503 close_connection(con, true, true, true);
1504 clean_one_writequeue(con);
1505 }
1506
1507 spin_lock(&dlm_node_addrs_spin);
1508 na = find_node_addr(nodeid);
1509 if (na) {
1510 list_del(&na->list);
1511 while (na->addr_count--)
1512 kfree(na->addr[na->addr_count]);
1513 kfree(na);
1514 }
1515 spin_unlock(&dlm_node_addrs_spin);
1516
1517 return 0;
1518 }
1519
1520 /* Receive workqueue function */
process_recv_sockets(struct work_struct * work)1521 static void process_recv_sockets(struct work_struct *work)
1522 {
1523 struct connection *con = container_of(work, struct connection, rwork);
1524 int err;
1525
1526 clear_bit(CF_READ_PENDING, &con->flags);
1527 do {
1528 err = con->rx_action(con);
1529 } while (!err);
1530 }
1531
1532 /* Send workqueue function */
process_send_sockets(struct work_struct * work)1533 static void process_send_sockets(struct work_struct *work)
1534 {
1535 struct connection *con = container_of(work, struct connection, swork);
1536
1537 clear_bit(CF_WRITE_PENDING, &con->flags);
1538 if (con->sock == NULL) /* not mutex protected so check it inside too */
1539 con->connect_action(con);
1540 if (!list_empty(&con->writequeue))
1541 send_to_sock(con);
1542 }
1543
work_stop(void)1544 static void work_stop(void)
1545 {
1546 if (recv_workqueue)
1547 destroy_workqueue(recv_workqueue);
1548 if (send_workqueue)
1549 destroy_workqueue(send_workqueue);
1550 }
1551
work_start(void)1552 static int work_start(void)
1553 {
1554 recv_workqueue = alloc_workqueue("dlm_recv",
1555 WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1556 if (!recv_workqueue) {
1557 log_print("can't start dlm_recv");
1558 return -ENOMEM;
1559 }
1560
1561 send_workqueue = alloc_workqueue("dlm_send",
1562 WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1563 if (!send_workqueue) {
1564 log_print("can't start dlm_send");
1565 destroy_workqueue(recv_workqueue);
1566 return -ENOMEM;
1567 }
1568
1569 return 0;
1570 }
1571
_stop_conn(struct connection * con,bool and_other)1572 static void _stop_conn(struct connection *con, bool and_other)
1573 {
1574 mutex_lock(&con->sock_mutex);
1575 set_bit(CF_CLOSE, &con->flags);
1576 set_bit(CF_READ_PENDING, &con->flags);
1577 set_bit(CF_WRITE_PENDING, &con->flags);
1578 if (con->sock && con->sock->sk) {
1579 write_lock_bh(&con->sock->sk->sk_callback_lock);
1580 con->sock->sk->sk_user_data = NULL;
1581 write_unlock_bh(&con->sock->sk->sk_callback_lock);
1582 }
1583 if (con->othercon && and_other)
1584 _stop_conn(con->othercon, false);
1585 mutex_unlock(&con->sock_mutex);
1586 }
1587
stop_conn(struct connection * con)1588 static void stop_conn(struct connection *con)
1589 {
1590 _stop_conn(con, true);
1591 }
1592
shutdown_conn(struct connection * con)1593 static void shutdown_conn(struct connection *con)
1594 {
1595 if (con->shutdown_action)
1596 con->shutdown_action(con);
1597 }
1598
connection_release(struct rcu_head * rcu)1599 static void connection_release(struct rcu_head *rcu)
1600 {
1601 struct connection *con = container_of(rcu, struct connection, rcu);
1602
1603 kfree(con->rx_buf);
1604 kfree(con);
1605 }
1606
free_conn(struct connection * con)1607 static void free_conn(struct connection *con)
1608 {
1609 close_connection(con, true, true, true);
1610 spin_lock(&connections_lock);
1611 hlist_del_rcu(&con->list);
1612 spin_unlock(&connections_lock);
1613 if (con->othercon) {
1614 clean_one_writequeue(con->othercon);
1615 call_rcu(&con->othercon->rcu, connection_release);
1616 }
1617 clean_one_writequeue(con);
1618 call_rcu(&con->rcu, connection_release);
1619 }
1620
work_flush(void)1621 static void work_flush(void)
1622 {
1623 int ok, idx;
1624 int i;
1625 struct connection *con;
1626
1627 do {
1628 ok = 1;
1629 foreach_conn(stop_conn);
1630 if (recv_workqueue)
1631 flush_workqueue(recv_workqueue);
1632 if (send_workqueue)
1633 flush_workqueue(send_workqueue);
1634 idx = srcu_read_lock(&connections_srcu);
1635 for (i = 0; i < CONN_HASH_SIZE && ok; i++) {
1636 hlist_for_each_entry_rcu(con, &connection_hash[i],
1637 list) {
1638 ok &= test_bit(CF_READ_PENDING, &con->flags);
1639 ok &= test_bit(CF_WRITE_PENDING, &con->flags);
1640 if (con->othercon) {
1641 ok &= test_bit(CF_READ_PENDING,
1642 &con->othercon->flags);
1643 ok &= test_bit(CF_WRITE_PENDING,
1644 &con->othercon->flags);
1645 }
1646 }
1647 }
1648 srcu_read_unlock(&connections_srcu, idx);
1649 } while (!ok);
1650 }
1651
dlm_lowcomms_stop(void)1652 void dlm_lowcomms_stop(void)
1653 {
1654 /* Set all the flags to prevent any
1655 socket activity.
1656 */
1657 dlm_allow_conn = 0;
1658
1659 if (recv_workqueue)
1660 flush_workqueue(recv_workqueue);
1661 if (send_workqueue)
1662 flush_workqueue(send_workqueue);
1663
1664 foreach_conn(shutdown_conn);
1665 work_flush();
1666 foreach_conn(free_conn);
1667 work_stop();
1668 deinit_local();
1669 }
1670
dlm_lowcomms_start(void)1671 int dlm_lowcomms_start(void)
1672 {
1673 int error = -EINVAL;
1674 struct connection *con;
1675 int i;
1676
1677 for (i = 0; i < CONN_HASH_SIZE; i++)
1678 INIT_HLIST_HEAD(&connection_hash[i]);
1679
1680 init_local();
1681 if (!dlm_local_count) {
1682 error = -ENOTCONN;
1683 log_print("no local IP address has been set");
1684 goto fail;
1685 }
1686
1687 error = work_start();
1688 if (error)
1689 goto fail;
1690
1691 dlm_allow_conn = 1;
1692
1693 /* Start listening */
1694 if (dlm_config.ci_protocol == 0)
1695 error = tcp_listen_for_all();
1696 else
1697 error = sctp_listen_for_all();
1698 if (error)
1699 goto fail_unlisten;
1700
1701 return 0;
1702
1703 fail_unlisten:
1704 dlm_allow_conn = 0;
1705 con = nodeid2con(0,0);
1706 if (con)
1707 free_conn(con);
1708 fail:
1709 return error;
1710 }
1711
dlm_lowcomms_exit(void)1712 void dlm_lowcomms_exit(void)
1713 {
1714 struct dlm_node_addr *na, *safe;
1715
1716 spin_lock(&dlm_node_addrs_spin);
1717 list_for_each_entry_safe(na, safe, &dlm_node_addrs, list) {
1718 list_del(&na->list);
1719 while (na->addr_count--)
1720 kfree(na->addr[na->addr_count]);
1721 kfree(na);
1722 }
1723 spin_unlock(&dlm_node_addrs_spin);
1724 }
1725