blob: a3350e4c8184212d0ad9a5b4f759ec4b3e0a8d55 [file] [log] [blame]
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001/******************************************************************************
2*******************************************************************************
3**
4** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06005** Copyright (C) 2004-2009 Red Hat, Inc. All rights reserved.
Patrick Caulfield6ed72572007-04-17 15:39:57 +01006**
7** This copyrighted material is made available to anyone wishing to use,
8** modify, copy, or redistribute it subject to the terms and conditions
9** of the GNU General Public License v.2.
10**
11*******************************************************************************
12******************************************************************************/
13
14/*
15 * lowcomms.c
16 *
17 * This is the "low-level" comms layer.
18 *
19 * It is responsible for sending/receiving messages
20 * from other nodes in the cluster.
21 *
22 * Cluster nodes are referred to by their nodeids. nodeids are
23 * simply 32 bit numbers to the locking module - if they need to
Joe Perches2cf12c02009-01-22 13:26:47 -080024 * be expanded for the cluster infrastructure then that is its
Patrick Caulfield6ed72572007-04-17 15:39:57 +010025 * responsibility. It is this layer's
26 * responsibility to resolve these into IP address or
27 * whatever it needs for inter-node communication.
28 *
29 * The comms level is two kernel threads that deal mainly with
30 * the receiving of messages from other nodes and passing them
31 * up to the mid-level comms layer (which understands the
32 * message format) for execution by the locking core, and
33 * a send thread which does all the setting up of connections
34 * to remote nodes and the sending of data. Threads are not allowed
35 * to send their own data because it may cause them to wait in times
36 * of high load. Also, this way, the sending thread can collect together
37 * messages bound for one node and send them in one block.
38 *
Joe Perches2cf12c02009-01-22 13:26:47 -080039 * lowcomms will choose to use either TCP or SCTP as its transport layer
Patrick Caulfield6ed72572007-04-17 15:39:57 +010040 * depending on the configuration variable 'protocol'. This should be set
Joe Perches2cf12c02009-01-22 13:26:47 -080041 * to 0 (default) for TCP or 1 for SCTP. It should be configured using a
Patrick Caulfield6ed72572007-04-17 15:39:57 +010042 * cluster-wide mechanism as it must be the same on all nodes of the cluster
43 * for the DLM to function.
44 *
45 */
46
47#include <asm/ioctls.h>
48#include <net/sock.h>
49#include <net/tcp.h>
50#include <linux/pagemap.h>
Patrick Caulfield6ed72572007-04-17 15:39:57 +010051#include <linux/file.h>
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -050052#include <linux/mutex.h>
Patrick Caulfield6ed72572007-04-17 15:39:57 +010053#include <linux/sctp.h>
54#include <net/sctp/user.h>
Joe Perches44ad5322009-01-22 13:24:49 -080055#include <net/ipv6.h>
Patrick Caulfield6ed72572007-04-17 15:39:57 +010056
57#include "dlm_internal.h"
58#include "lowcomms.h"
59#include "midcomms.h"
60#include "config.h"
61
62#define NEEDED_RMEM (4*1024*1024)
Christine Caulfield5e9ccc32009-01-28 12:57:40 -060063#define CONN_HASH_SIZE 32
Patrick Caulfield6ed72572007-04-17 15:39:57 +010064
65struct cbuf {
66 unsigned int base;
67 unsigned int len;
68 unsigned int mask;
69};
70
71static void cbuf_add(struct cbuf *cb, int n)
72{
73 cb->len += n;
74}
75
76static int cbuf_data(struct cbuf *cb)
77{
78 return ((cb->base + cb->len) & cb->mask);
79}
80
81static void cbuf_init(struct cbuf *cb, int size)
82{
83 cb->base = cb->len = 0;
84 cb->mask = size-1;
85}
86
87static void cbuf_eat(struct cbuf *cb, int n)
88{
89 cb->len -= n;
90 cb->base += n;
91 cb->base &= cb->mask;
92}
93
94static bool cbuf_empty(struct cbuf *cb)
95{
96 return cb->len == 0;
97}
98
99struct connection {
100 struct socket *sock; /* NULL if not connected */
101 uint32_t nodeid; /* So we know who we are in the list */
102 struct mutex sock_mutex;
103 unsigned long flags;
104#define CF_READ_PENDING 1
105#define CF_WRITE_PENDING 2
106#define CF_CONNECT_PENDING 3
107#define CF_INIT_PENDING 4
108#define CF_IS_OTHERCON 5
Lars Marowsky-Bree063c4c92009-08-11 16:18:23 -0500109#define CF_CLOSE 6
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100110 struct list_head writequeue; /* List of outgoing writequeue_entries */
111 spinlock_t writequeue_lock;
112 int (*rx_action) (struct connection *); /* What to do when active */
113 void (*connect_action) (struct connection *); /* What to do to connect */
114 struct page *rx_page;
115 struct cbuf cb;
116 int retries;
117#define MAX_CONNECT_RETRIES 3
118 int sctp_assoc;
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600119 struct hlist_node list;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100120 struct connection *othercon;
121 struct work_struct rwork; /* Receive workqueue */
122 struct work_struct swork; /* Send workqueue */
123};
124#define sock2con(x) ((struct connection *)(x)->sk_user_data)
125
126/* An entry waiting to be sent */
127struct writequeue_entry {
128 struct list_head list;
129 struct page *page;
130 int offset;
131 int len;
132 int end;
133 int users;
134 struct connection *con;
135};
136
137static struct sockaddr_storage *dlm_local_addr[DLM_MAX_ADDR_COUNT];
138static int dlm_local_count;
139
140/* Work queues */
141static struct workqueue_struct *recv_workqueue;
142static struct workqueue_struct *send_workqueue;
143
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600144static struct hlist_head connection_hash[CONN_HASH_SIZE];
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -0500145static DEFINE_MUTEX(connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100146static struct kmem_cache *con_cache;
147
148static void process_recv_sockets(struct work_struct *work);
149static void process_send_sockets(struct work_struct *work);
150
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600151
152/* This is deliberately very simple because most clusters have simple
153 sequential nodeids, so we should be able to go straight to a connection
154 struct in the array */
155static inline int nodeid_hash(int nodeid)
156{
157 return nodeid & (CONN_HASH_SIZE-1);
158}
159
160static struct connection *__find_con(int nodeid)
161{
162 int r;
163 struct hlist_node *h;
164 struct connection *con;
165
166 r = nodeid_hash(nodeid);
167
168 hlist_for_each_entry(con, h, &connection_hash[r], list) {
169 if (con->nodeid == nodeid)
170 return con;
171 }
172 return NULL;
173}
174
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100175/*
176 * If 'allocation' is zero then we don't attempt to create a new
177 * connection structure for this node.
178 */
179static struct connection *__nodeid2con(int nodeid, gfp_t alloc)
180{
181 struct connection *con = NULL;
182 int r;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100183
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600184 con = __find_con(nodeid);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100185 if (con || !alloc)
186 return con;
187
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100188 con = kmem_cache_zalloc(con_cache, alloc);
189 if (!con)
190 return NULL;
191
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600192 r = nodeid_hash(nodeid);
193 hlist_add_head(&con->list, &connection_hash[r]);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100194
195 con->nodeid = nodeid;
196 mutex_init(&con->sock_mutex);
197 INIT_LIST_HEAD(&con->writequeue);
198 spin_lock_init(&con->writequeue_lock);
199 INIT_WORK(&con->swork, process_send_sockets);
200 INIT_WORK(&con->rwork, process_recv_sockets);
201
202 /* Setup action pointers for child sockets */
203 if (con->nodeid) {
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600204 struct connection *zerocon = __find_con(0);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100205
206 con->connect_action = zerocon->connect_action;
207 if (!con->rx_action)
208 con->rx_action = zerocon->rx_action;
209 }
210
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100211 return con;
212}
213
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600214/* Loop round all connections */
215static void foreach_conn(void (*conn_func)(struct connection *c))
216{
217 int i;
218 struct hlist_node *h, *n;
219 struct connection *con;
220
221 for (i = 0; i < CONN_HASH_SIZE; i++) {
222 hlist_for_each_entry_safe(con, h, n, &connection_hash[i], list){
223 conn_func(con);
224 }
225 }
226}
227
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100228static struct connection *nodeid2con(int nodeid, gfp_t allocation)
229{
230 struct connection *con;
231
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -0500232 mutex_lock(&connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100233 con = __nodeid2con(nodeid, allocation);
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -0500234 mutex_unlock(&connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100235
236 return con;
237}
238
239/* This is a bit drastic, but only called when things go wrong */
240static struct connection *assoc2con(int assoc_id)
241{
242 int i;
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600243 struct hlist_node *h;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100244 struct connection *con;
245
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -0500246 mutex_lock(&connections_lock);
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600247
248 for (i = 0 ; i < CONN_HASH_SIZE; i++) {
249 hlist_for_each_entry(con, h, &connection_hash[i], list) {
250 if (con && con->sctp_assoc == assoc_id) {
251 mutex_unlock(&connections_lock);
252 return con;
253 }
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100254 }
255 }
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -0500256 mutex_unlock(&connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100257 return NULL;
258}
259
260static int nodeid_to_addr(int nodeid, struct sockaddr *retaddr)
261{
262 struct sockaddr_storage addr;
263 int error;
264
265 if (!dlm_local_count)
266 return -1;
267
268 error = dlm_nodeid_to_addr(nodeid, &addr);
269 if (error)
270 return error;
271
272 if (dlm_local_addr[0]->ss_family == AF_INET) {
273 struct sockaddr_in *in4 = (struct sockaddr_in *) &addr;
274 struct sockaddr_in *ret4 = (struct sockaddr_in *) retaddr;
275 ret4->sin_addr.s_addr = in4->sin_addr.s_addr;
276 } else {
277 struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) &addr;
278 struct sockaddr_in6 *ret6 = (struct sockaddr_in6 *) retaddr;
Joe Perches44ad5322009-01-22 13:24:49 -0800279 ipv6_addr_copy(&ret6->sin6_addr, &in6->sin6_addr);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100280 }
281
282 return 0;
283}
284
285/* Data available on socket or listen socket received a connect */
286static void lowcomms_data_ready(struct sock *sk, int count_unused)
287{
288 struct connection *con = sock2con(sk);
Patrick Caulfieldafb853f2007-06-01 10:07:26 -0500289 if (con && !test_and_set_bit(CF_READ_PENDING, &con->flags))
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100290 queue_work(recv_workqueue, &con->rwork);
291}
292
293static void lowcomms_write_space(struct sock *sk)
294{
295 struct connection *con = sock2con(sk);
296
Patrick Caulfieldafb853f2007-06-01 10:07:26 -0500297 if (con && !test_and_set_bit(CF_WRITE_PENDING, &con->flags))
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100298 queue_work(send_workqueue, &con->swork);
299}
300
301static inline void lowcomms_connect_sock(struct connection *con)
302{
Lars Marowsky-Bree063c4c92009-08-11 16:18:23 -0500303 if (test_bit(CF_CLOSE, &con->flags))
304 return;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100305 if (!test_and_set_bit(CF_CONNECT_PENDING, &con->flags))
306 queue_work(send_workqueue, &con->swork);
307}
308
309static void lowcomms_state_change(struct sock *sk)
310{
311 if (sk->sk_state == TCP_ESTABLISHED)
312 lowcomms_write_space(sk);
313}
314
Christine Caulfield391fbdc2009-05-07 10:54:16 -0500315int dlm_lowcomms_connect_node(int nodeid)
316{
317 struct connection *con;
318
David Teigland04bedd72009-09-18 14:31:47 -0500319 /* with sctp there's no connecting without sending */
320 if (dlm_config.ci_protocol != 0)
321 return 0;
322
Christine Caulfield391fbdc2009-05-07 10:54:16 -0500323 if (nodeid == dlm_our_nodeid())
324 return 0;
325
326 con = nodeid2con(nodeid, GFP_NOFS);
327 if (!con)
328 return -ENOMEM;
329 lowcomms_connect_sock(con);
330 return 0;
331}
332
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100333/* Make a socket active */
334static int add_sock(struct socket *sock, struct connection *con)
335{
336 con->sock = sock;
337
338 /* Install a data_ready callback */
339 con->sock->sk->sk_data_ready = lowcomms_data_ready;
340 con->sock->sk->sk_write_space = lowcomms_write_space;
341 con->sock->sk->sk_state_change = lowcomms_state_change;
342 con->sock->sk->sk_user_data = con;
Steven Whitehoused6d7b702008-11-12 16:49:48 -0600343 con->sock->sk->sk_allocation = GFP_NOFS;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100344 return 0;
345}
346
347/* Add the port number to an IPv6 or 4 sockaddr and return the address
348 length */
349static void make_sockaddr(struct sockaddr_storage *saddr, uint16_t port,
350 int *addr_len)
351{
352 saddr->ss_family = dlm_local_addr[0]->ss_family;
353 if (saddr->ss_family == AF_INET) {
354 struct sockaddr_in *in4_addr = (struct sockaddr_in *)saddr;
355 in4_addr->sin_port = cpu_to_be16(port);
356 *addr_len = sizeof(struct sockaddr_in);
357 memset(&in4_addr->sin_zero, 0, sizeof(in4_addr->sin_zero));
358 } else {
359 struct sockaddr_in6 *in6_addr = (struct sockaddr_in6 *)saddr;
360 in6_addr->sin6_port = cpu_to_be16(port);
361 *addr_len = sizeof(struct sockaddr_in6);
362 }
Patrick Caulfield01c8cab2007-07-17 16:53:15 +0100363 memset((char *)saddr + *addr_len, 0, sizeof(struct sockaddr_storage) - *addr_len);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100364}
365
366/* Close a remote connection and tidy up */
367static void close_connection(struct connection *con, bool and_other)
368{
369 mutex_lock(&con->sock_mutex);
370
371 if (con->sock) {
372 sock_release(con->sock);
373 con->sock = NULL;
374 }
375 if (con->othercon && and_other) {
376 /* Will only re-enter once. */
377 close_connection(con->othercon, false);
378 }
379 if (con->rx_page) {
380 __free_page(con->rx_page);
381 con->rx_page = NULL;
382 }
Patrick Caulfield9e5f2822007-08-02 14:58:14 +0100383
Patrick Caulfield61d96be2007-08-20 15:13:38 +0100384 con->retries = 0;
385 mutex_unlock(&con->sock_mutex);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100386}
387
388/* We only send shutdown messages to nodes that are not part of the cluster */
389static void sctp_send_shutdown(sctp_assoc_t associd)
390{
391 static char outcmsg[CMSG_SPACE(sizeof(struct sctp_sndrcvinfo))];
392 struct msghdr outmessage;
393 struct cmsghdr *cmsg;
394 struct sctp_sndrcvinfo *sinfo;
395 int ret;
396 struct connection *con;
397
398 con = nodeid2con(0,0);
399 BUG_ON(con == NULL);
400
401 outmessage.msg_name = NULL;
402 outmessage.msg_namelen = 0;
403 outmessage.msg_control = outcmsg;
404 outmessage.msg_controllen = sizeof(outcmsg);
405 outmessage.msg_flags = MSG_EOR;
406
407 cmsg = CMSG_FIRSTHDR(&outmessage);
408 cmsg->cmsg_level = IPPROTO_SCTP;
409 cmsg->cmsg_type = SCTP_SNDRCV;
410 cmsg->cmsg_len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo));
411 outmessage.msg_controllen = cmsg->cmsg_len;
412 sinfo = CMSG_DATA(cmsg);
413 memset(sinfo, 0x00, sizeof(struct sctp_sndrcvinfo));
414
415 sinfo->sinfo_flags |= MSG_EOF;
416 sinfo->sinfo_assoc_id = associd;
417
418 ret = kernel_sendmsg(con->sock, &outmessage, NULL, 0, 0);
419
420 if (ret != 0)
421 log_print("send EOF to node failed: %d", ret);
422}
423
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600424static void sctp_init_failed_foreach(struct connection *con)
425{
426 con->sctp_assoc = 0;
427 if (test_and_clear_bit(CF_CONNECT_PENDING, &con->flags)) {
428 if (!test_and_set_bit(CF_WRITE_PENDING, &con->flags))
429 queue_work(send_workqueue, &con->swork);
430 }
431}
432
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100433/* INIT failed but we don't know which node...
434 restart INIT on all pending nodes */
435static void sctp_init_failed(void)
436{
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -0500437 mutex_lock(&connections_lock);
Christine Caulfield5e9ccc32009-01-28 12:57:40 -0600438
439 foreach_conn(sctp_init_failed_foreach);
440
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -0500441 mutex_unlock(&connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100442}
443
444/* Something happened to an association */
David Teigland617e82e2007-04-26 13:46:49 -0500445static void process_sctp_notification(struct connection *con,
446 struct msghdr *msg, char *buf)
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100447{
448 union sctp_notification *sn = (union sctp_notification *)buf;
449
450 if (sn->sn_header.sn_type == SCTP_ASSOC_CHANGE) {
451 switch (sn->sn_assoc_change.sac_state) {
452
453 case SCTP_COMM_UP:
454 case SCTP_RESTART:
455 {
456 /* Check that the new node is in the lockspace */
457 struct sctp_prim prim;
458 int nodeid;
459 int prim_len, ret;
460 int addr_len;
461 struct connection *new_con;
462 struct file *file;
463 sctp_peeloff_arg_t parg;
464 int parglen = sizeof(parg);
465
466 /*
467 * We get this before any data for an association.
468 * We verify that the node is in the cluster and
469 * then peel off a socket for it.
470 */
471 if ((int)sn->sn_assoc_change.sac_assoc_id <= 0) {
472 log_print("COMM_UP for invalid assoc ID %d",
David Teigland617e82e2007-04-26 13:46:49 -0500473 (int)sn->sn_assoc_change.sac_assoc_id);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100474 sctp_init_failed();
475 return;
476 }
477 memset(&prim, 0, sizeof(struct sctp_prim));
478 prim_len = sizeof(struct sctp_prim);
479 prim.ssp_assoc_id = sn->sn_assoc_change.sac_assoc_id;
480
481 ret = kernel_getsockopt(con->sock,
482 IPPROTO_SCTP,
483 SCTP_PRIMARY_ADDR,
484 (char*)&prim,
485 &prim_len);
486 if (ret < 0) {
487 log_print("getsockopt/sctp_primary_addr on "
488 "new assoc %d failed : %d",
489 (int)sn->sn_assoc_change.sac_assoc_id,
490 ret);
491
492 /* Retry INIT later */
493 new_con = assoc2con(sn->sn_assoc_change.sac_assoc_id);
494 if (new_con)
495 clear_bit(CF_CONNECT_PENDING, &con->flags);
496 return;
497 }
498 make_sockaddr(&prim.ssp_addr, 0, &addr_len);
499 if (dlm_addr_to_nodeid(&prim.ssp_addr, &nodeid)) {
500 int i;
501 unsigned char *b=(unsigned char *)&prim.ssp_addr;
502 log_print("reject connect from unknown addr");
503 for (i=0; i<sizeof(struct sockaddr_storage);i++)
504 printk("%02x ", b[i]);
505 printk("\n");
506 sctp_send_shutdown(prim.ssp_assoc_id);
507 return;
508 }
509
David Teigland748285c2009-05-15 10:50:57 -0500510 new_con = nodeid2con(nodeid, GFP_NOFS);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100511 if (!new_con)
512 return;
513
514 /* Peel off a new sock */
515 parg.associd = sn->sn_assoc_change.sac_assoc_id;
David Teigland617e82e2007-04-26 13:46:49 -0500516 ret = kernel_getsockopt(con->sock, IPPROTO_SCTP,
517 SCTP_SOCKOPT_PEELOFF,
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100518 (void *)&parg, &parglen);
Patrick Caulfield30d3a232007-04-23 16:26:21 +0100519 if (ret) {
David Teigland617e82e2007-04-26 13:46:49 -0500520 log_print("Can't peel off a socket for "
521 "connection %d to node %d: err=%d\n",
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100522 parg.associd, nodeid, ret);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100523 }
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100524 file = fget(parg.sd);
525 new_con->sock = SOCKET_I(file->f_dentry->d_inode);
526 add_sock(new_con->sock, new_con);
527 fput(file);
528 put_unused_fd(parg.sd);
529
530 log_print("got new/restarted association %d nodeid %d",
David Teigland617e82e2007-04-26 13:46:49 -0500531 (int)sn->sn_assoc_change.sac_assoc_id, nodeid);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100532
533 /* Send any pending writes */
534 clear_bit(CF_CONNECT_PENDING, &new_con->flags);
535 clear_bit(CF_INIT_PENDING, &con->flags);
536 if (!test_and_set_bit(CF_WRITE_PENDING, &new_con->flags)) {
537 queue_work(send_workqueue, &new_con->swork);
538 }
539 if (!test_and_set_bit(CF_READ_PENDING, &new_con->flags))
540 queue_work(recv_workqueue, &new_con->rwork);
541 }
542 break;
543
544 case SCTP_COMM_LOST:
545 case SCTP_SHUTDOWN_COMP:
546 {
547 con = assoc2con(sn->sn_assoc_change.sac_assoc_id);
548 if (con) {
549 con->sctp_assoc = 0;
550 }
551 }
552 break;
553
554 /* We don't know which INIT failed, so clear the PENDING flags
555 * on them all. if assoc_id is zero then it will then try
556 * again */
557
558 case SCTP_CANT_STR_ASSOC:
559 {
560 log_print("Can't start SCTP association - retrying");
561 sctp_init_failed();
562 }
563 break;
564
565 default:
566 log_print("unexpected SCTP assoc change id=%d state=%d",
567 (int)sn->sn_assoc_change.sac_assoc_id,
568 sn->sn_assoc_change.sac_state);
569 }
570 }
571}
572
573/* Data received from remote end */
574static int receive_from_sock(struct connection *con)
575{
576 int ret = 0;
577 struct msghdr msg = {};
578 struct kvec iov[2];
579 unsigned len;
580 int r;
581 int call_again_soon = 0;
582 int nvec;
583 char incmsg[CMSG_SPACE(sizeof(struct sctp_sndrcvinfo))];
584
585 mutex_lock(&con->sock_mutex);
586
587 if (con->sock == NULL) {
588 ret = -EAGAIN;
589 goto out_close;
590 }
591
592 if (con->rx_page == NULL) {
593 /*
594 * This doesn't need to be atomic, but I think it should
595 * improve performance if it is.
596 */
597 con->rx_page = alloc_page(GFP_ATOMIC);
598 if (con->rx_page == NULL)
599 goto out_resched;
600 cbuf_init(&con->cb, PAGE_CACHE_SIZE);
601 }
602
603 /* Only SCTP needs these really */
604 memset(&incmsg, 0, sizeof(incmsg));
605 msg.msg_control = incmsg;
606 msg.msg_controllen = sizeof(incmsg);
607
608 /*
609 * iov[0] is the bit of the circular buffer between the current end
610 * point (cb.base + cb.len) and the end of the buffer.
611 */
612 iov[0].iov_len = con->cb.base - cbuf_data(&con->cb);
613 iov[0].iov_base = page_address(con->rx_page) + cbuf_data(&con->cb);
614 iov[1].iov_len = 0;
615 nvec = 1;
616
617 /*
618 * iov[1] is the bit of the circular buffer between the start of the
619 * buffer and the start of the currently used section (cb.base)
620 */
621 if (cbuf_data(&con->cb) >= con->cb.base) {
622 iov[0].iov_len = PAGE_CACHE_SIZE - cbuf_data(&con->cb);
623 iov[1].iov_len = con->cb.base;
624 iov[1].iov_base = page_address(con->rx_page);
625 nvec = 2;
626 }
627 len = iov[0].iov_len + iov[1].iov_len;
628
629 r = ret = kernel_recvmsg(con->sock, &msg, iov, nvec, len,
630 MSG_DONTWAIT | MSG_NOSIGNAL);
631 if (ret <= 0)
632 goto out_close;
633
634 /* Process SCTP notifications */
635 if (msg.msg_flags & MSG_NOTIFICATION) {
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100636 msg.msg_control = incmsg;
637 msg.msg_controllen = sizeof(incmsg);
638
639 process_sctp_notification(con, &msg,
David Teigland617e82e2007-04-26 13:46:49 -0500640 page_address(con->rx_page) + con->cb.base);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100641 mutex_unlock(&con->sock_mutex);
642 return 0;
643 }
644 BUG_ON(con->nodeid == 0);
645
646 if (ret == len)
647 call_again_soon = 1;
648 cbuf_add(&con->cb, ret);
649 ret = dlm_process_incoming_buffer(con->nodeid,
650 page_address(con->rx_page),
651 con->cb.base, con->cb.len,
652 PAGE_CACHE_SIZE);
653 if (ret == -EBADMSG) {
David Teigland617e82e2007-04-26 13:46:49 -0500654 log_print("lowcomms: addr=%p, base=%u, len=%u, "
655 "iov_len=%u, iov_base[0]=%p, read=%d",
656 page_address(con->rx_page), con->cb.base, con->cb.len,
657 len, iov[0].iov_base, r);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100658 }
659 if (ret < 0)
660 goto out_close;
661 cbuf_eat(&con->cb, ret);
662
663 if (cbuf_empty(&con->cb) && !call_again_soon) {
664 __free_page(con->rx_page);
665 con->rx_page = NULL;
666 }
667
668 if (call_again_soon)
669 goto out_resched;
670 mutex_unlock(&con->sock_mutex);
671 return 0;
672
673out_resched:
674 if (!test_and_set_bit(CF_READ_PENDING, &con->flags))
675 queue_work(recv_workqueue, &con->rwork);
676 mutex_unlock(&con->sock_mutex);
677 return -EAGAIN;
678
679out_close:
680 mutex_unlock(&con->sock_mutex);
Patrick Caulfield9e5f2822007-08-02 14:58:14 +0100681 if (ret != -EAGAIN) {
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100682 close_connection(con, false);
683 /* Reconnect when there is something to send */
684 }
685 /* Don't return success if we really got EOF */
686 if (ret == 0)
687 ret = -EAGAIN;
688
689 return ret;
690}
691
692/* Listening socket is busy, accept a connection */
693static int tcp_accept_from_sock(struct connection *con)
694{
695 int result;
696 struct sockaddr_storage peeraddr;
697 struct socket *newsock;
698 int len;
699 int nodeid;
700 struct connection *newcon;
701 struct connection *addcon;
702
703 memset(&peeraddr, 0, sizeof(peeraddr));
704 result = sock_create_kern(dlm_local_addr[0]->ss_family, SOCK_STREAM,
705 IPPROTO_TCP, &newsock);
706 if (result < 0)
707 return -ENOMEM;
708
709 mutex_lock_nested(&con->sock_mutex, 0);
710
711 result = -ENOTCONN;
712 if (con->sock == NULL)
713 goto accept_err;
714
715 newsock->type = con->sock->type;
716 newsock->ops = con->sock->ops;
717
718 result = con->sock->ops->accept(con->sock, newsock, O_NONBLOCK);
719 if (result < 0)
720 goto accept_err;
721
722 /* Get the connected socket's peer */
723 memset(&peeraddr, 0, sizeof(peeraddr));
724 if (newsock->ops->getname(newsock, (struct sockaddr *)&peeraddr,
725 &len, 2)) {
726 result = -ECONNABORTED;
727 goto accept_err;
728 }
729
730 /* Get the new node's NODEID */
731 make_sockaddr(&peeraddr, 0, &len);
732 if (dlm_addr_to_nodeid(&peeraddr, &nodeid)) {
David Teigland617e82e2007-04-26 13:46:49 -0500733 log_print("connect from non cluster node");
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100734 sock_release(newsock);
735 mutex_unlock(&con->sock_mutex);
736 return -1;
737 }
738
739 log_print("got connection from %d", nodeid);
740
741 /* Check to see if we already have a connection to this node. This
742 * could happen if the two nodes initiate a connection at roughly
743 * the same time and the connections cross on the wire.
744 * In this case we store the incoming one in "othercon"
745 */
David Teigland748285c2009-05-15 10:50:57 -0500746 newcon = nodeid2con(nodeid, GFP_NOFS);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100747 if (!newcon) {
748 result = -ENOMEM;
749 goto accept_err;
750 }
751 mutex_lock_nested(&newcon->sock_mutex, 1);
752 if (newcon->sock) {
753 struct connection *othercon = newcon->othercon;
754
755 if (!othercon) {
David Teigland748285c2009-05-15 10:50:57 -0500756 othercon = kmem_cache_zalloc(con_cache, GFP_NOFS);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100757 if (!othercon) {
David Teigland617e82e2007-04-26 13:46:49 -0500758 log_print("failed to allocate incoming socket");
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100759 mutex_unlock(&newcon->sock_mutex);
760 result = -ENOMEM;
761 goto accept_err;
762 }
763 othercon->nodeid = nodeid;
764 othercon->rx_action = receive_from_sock;
765 mutex_init(&othercon->sock_mutex);
766 INIT_WORK(&othercon->swork, process_send_sockets);
767 INIT_WORK(&othercon->rwork, process_recv_sockets);
768 set_bit(CF_IS_OTHERCON, &othercon->flags);
Patrick Caulfield61d96be2007-08-20 15:13:38 +0100769 }
770 if (!othercon->sock) {
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100771 newcon->othercon = othercon;
Patrick Caulfield97d84832007-06-27 11:36:23 +0100772 othercon->sock = newsock;
773 newsock->sk->sk_user_data = othercon;
774 add_sock(newsock, othercon);
775 addcon = othercon;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100776 }
Patrick Caulfield97d84832007-06-27 11:36:23 +0100777 else {
778 printk("Extra connection from node %d attempted\n", nodeid);
779 result = -EAGAIN;
akpm@linux-foundation.orgf4fadb23c2007-06-27 14:43:37 -0700780 mutex_unlock(&newcon->sock_mutex);
Patrick Caulfield97d84832007-06-27 11:36:23 +0100781 goto accept_err;
782 }
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100783 }
784 else {
785 newsock->sk->sk_user_data = newcon;
786 newcon->rx_action = receive_from_sock;
787 add_sock(newsock, newcon);
788 addcon = newcon;
789 }
790
791 mutex_unlock(&newcon->sock_mutex);
792
793 /*
794 * Add it to the active queue in case we got data
795 * beween processing the accept adding the socket
796 * to the read_sockets list
797 */
798 if (!test_and_set_bit(CF_READ_PENDING, &addcon->flags))
799 queue_work(recv_workqueue, &addcon->rwork);
800 mutex_unlock(&con->sock_mutex);
801
802 return 0;
803
804accept_err:
805 mutex_unlock(&con->sock_mutex);
806 sock_release(newsock);
807
808 if (result != -EAGAIN)
David Teigland617e82e2007-04-26 13:46:49 -0500809 log_print("error accepting connection from node: %d", result);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100810 return result;
811}
812
813static void free_entry(struct writequeue_entry *e)
814{
815 __free_page(e->page);
816 kfree(e);
817}
818
819/* Initiate an SCTP association.
820 This is a special case of send_to_sock() in that we don't yet have a
821 peeled-off socket for this association, so we use the listening socket
822 and add the primary IP address of the remote node.
823 */
824static void sctp_init_assoc(struct connection *con)
825{
826 struct sockaddr_storage rem_addr;
827 char outcmsg[CMSG_SPACE(sizeof(struct sctp_sndrcvinfo))];
828 struct msghdr outmessage;
829 struct cmsghdr *cmsg;
830 struct sctp_sndrcvinfo *sinfo;
831 struct connection *base_con;
832 struct writequeue_entry *e;
833 int len, offset;
834 int ret;
835 int addrlen;
836 struct kvec iov[1];
837
838 if (test_and_set_bit(CF_INIT_PENDING, &con->flags))
839 return;
840
841 if (con->retries++ > MAX_CONNECT_RETRIES)
842 return;
843
844 log_print("Initiating association with node %d", con->nodeid);
845
846 if (nodeid_to_addr(con->nodeid, (struct sockaddr *)&rem_addr)) {
847 log_print("no address for nodeid %d", con->nodeid);
848 return;
849 }
850 base_con = nodeid2con(0, 0);
851 BUG_ON(base_con == NULL);
852
853 make_sockaddr(&rem_addr, dlm_config.ci_tcp_port, &addrlen);
854
855 outmessage.msg_name = &rem_addr;
856 outmessage.msg_namelen = addrlen;
857 outmessage.msg_control = outcmsg;
858 outmessage.msg_controllen = sizeof(outcmsg);
859 outmessage.msg_flags = MSG_EOR;
860
861 spin_lock(&con->writequeue_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100862
David Teigland04bedd72009-09-18 14:31:47 -0500863 if (list_empty(&con->writequeue)) {
864 spin_unlock(&con->writequeue_lock);
865 log_print("writequeue empty for nodeid %d", con->nodeid);
866 return;
867 }
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100868
David Teigland04bedd72009-09-18 14:31:47 -0500869 e = list_first_entry(&con->writequeue, struct writequeue_entry, list);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100870 len = e->len;
871 offset = e->offset;
872 spin_unlock(&con->writequeue_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100873
874 /* Send the first block off the write queue */
875 iov[0].iov_base = page_address(e->page)+offset;
876 iov[0].iov_len = len;
877
878 cmsg = CMSG_FIRSTHDR(&outmessage);
879 cmsg->cmsg_level = IPPROTO_SCTP;
880 cmsg->cmsg_type = SCTP_SNDRCV;
881 cmsg->cmsg_len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo));
882 sinfo = CMSG_DATA(cmsg);
883 memset(sinfo, 0x00, sizeof(struct sctp_sndrcvinfo));
884 sinfo->sinfo_ppid = cpu_to_le32(dlm_our_nodeid());
885 outmessage.msg_controllen = cmsg->cmsg_len;
886
887 ret = kernel_sendmsg(base_con->sock, &outmessage, iov, 1, len);
888 if (ret < 0) {
David Teigland617e82e2007-04-26 13:46:49 -0500889 log_print("Send first packet to node %d failed: %d",
890 con->nodeid, ret);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100891
892 /* Try again later */
893 clear_bit(CF_CONNECT_PENDING, &con->flags);
894 clear_bit(CF_INIT_PENDING, &con->flags);
895 }
896 else {
897 spin_lock(&con->writequeue_lock);
898 e->offset += ret;
899 e->len -= ret;
900
901 if (e->len == 0 && e->users == 0) {
902 list_del(&e->list);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100903 free_entry(e);
904 }
905 spin_unlock(&con->writequeue_lock);
906 }
907}
908
909/* Connect a new socket to its peer */
910static void tcp_connect_to_sock(struct connection *con)
911{
912 int result = -EHOSTUNREACH;
Lon Hohberger6bd8fed2007-10-25 18:51:54 -0400913 struct sockaddr_storage saddr, src_addr;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100914 int addr_len;
Casey Dahlina89d63a2009-07-14 12:17:51 -0500915 struct socket *sock = NULL;
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100916
917 if (con->nodeid == 0) {
918 log_print("attempt to connect sock 0 foiled");
919 return;
920 }
921
922 mutex_lock(&con->sock_mutex);
923 if (con->retries++ > MAX_CONNECT_RETRIES)
924 goto out;
925
926 /* Some odd races can cause double-connects, ignore them */
927 if (con->sock) {
928 result = 0;
929 goto out;
930 }
931
932 /* Create a socket to communicate with */
933 result = sock_create_kern(dlm_local_addr[0]->ss_family, SOCK_STREAM,
934 IPPROTO_TCP, &sock);
935 if (result < 0)
936 goto out_err;
937
938 memset(&saddr, 0, sizeof(saddr));
Casey Dahlinb5711b82009-07-28 12:29:05 -0500939 if (dlm_nodeid_to_addr(con->nodeid, &saddr))
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100940 goto out_err;
941
942 sock->sk->sk_user_data = con;
943 con->rx_action = receive_from_sock;
944 con->connect_action = tcp_connect_to_sock;
945 add_sock(sock, con);
946
Lon Hohberger6bd8fed2007-10-25 18:51:54 -0400947 /* Bind to our cluster-known address connecting to avoid
948 routing problems */
949 memcpy(&src_addr, dlm_local_addr[0], sizeof(src_addr));
950 make_sockaddr(&src_addr, 0, &addr_len);
951 result = sock->ops->bind(sock, (struct sockaddr *) &src_addr,
952 addr_len);
953 if (result < 0) {
954 log_print("could not bind for connect: %d", result);
955 /* This *may* not indicate a critical error */
956 }
957
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100958 make_sockaddr(&saddr, dlm_config.ci_tcp_port, &addr_len);
959
960 log_print("connecting to %d", con->nodeid);
961 result =
962 sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
963 O_NONBLOCK);
964 if (result == -EINPROGRESS)
965 result = 0;
966 if (result == 0)
967 goto out;
968
969out_err:
970 if (con->sock) {
971 sock_release(con->sock);
972 con->sock = NULL;
Casey Dahlina89d63a2009-07-14 12:17:51 -0500973 } else if (sock) {
974 sock_release(sock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100975 }
976 /*
977 * Some errors are fatal and this list might need adjusting. For other
978 * errors we try again until the max number of retries is reached.
979 */
980 if (result != -EHOSTUNREACH && result != -ENETUNREACH &&
Marcin Slusarz0035a4b2008-05-11 22:01:29 +0200981 result != -ENETDOWN && result != -EINVAL
Patrick Caulfield6ed72572007-04-17 15:39:57 +0100982 && result != -EPROTONOSUPPORT) {
983 lowcomms_connect_sock(con);
984 result = 0;
985 }
986out:
987 mutex_unlock(&con->sock_mutex);
988 return;
989}
990
991static struct socket *tcp_create_listen_sock(struct connection *con,
992 struct sockaddr_storage *saddr)
993{
994 struct socket *sock = NULL;
995 int result = 0;
996 int one = 1;
997 int addr_len;
998
999 if (dlm_local_addr[0]->ss_family == AF_INET)
1000 addr_len = sizeof(struct sockaddr_in);
1001 else
1002 addr_len = sizeof(struct sockaddr_in6);
1003
1004 /* Create a socket to communicate with */
David Teigland617e82e2007-04-26 13:46:49 -05001005 result = sock_create_kern(dlm_local_addr[0]->ss_family, SOCK_STREAM,
1006 IPPROTO_TCP, &sock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001007 if (result < 0) {
David Teigland617e82e2007-04-26 13:46:49 -05001008 log_print("Can't create listening comms socket");
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001009 goto create_out;
1010 }
1011
1012 result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
1013 (char *)&one, sizeof(one));
1014
1015 if (result < 0) {
David Teigland617e82e2007-04-26 13:46:49 -05001016 log_print("Failed to set SO_REUSEADDR on socket: %d", result);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001017 }
1018 sock->sk->sk_user_data = con;
1019 con->rx_action = tcp_accept_from_sock;
1020 con->connect_action = tcp_connect_to_sock;
1021 con->sock = sock;
1022
1023 /* Bind to our port */
1024 make_sockaddr(saddr, dlm_config.ci_tcp_port, &addr_len);
1025 result = sock->ops->bind(sock, (struct sockaddr *) saddr, addr_len);
1026 if (result < 0) {
David Teigland617e82e2007-04-26 13:46:49 -05001027 log_print("Can't bind to port %d", dlm_config.ci_tcp_port);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001028 sock_release(sock);
1029 sock = NULL;
1030 con->sock = NULL;
1031 goto create_out;
1032 }
1033 result = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
1034 (char *)&one, sizeof(one));
1035 if (result < 0) {
David Teigland617e82e2007-04-26 13:46:49 -05001036 log_print("Set keepalive failed: %d", result);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001037 }
1038
1039 result = sock->ops->listen(sock, 5);
1040 if (result < 0) {
David Teigland617e82e2007-04-26 13:46:49 -05001041 log_print("Can't listen on port %d", dlm_config.ci_tcp_port);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001042 sock_release(sock);
1043 sock = NULL;
1044 goto create_out;
1045 }
1046
1047create_out:
1048 return sock;
1049}
1050
1051/* Get local addresses */
1052static void init_local(void)
1053{
1054 struct sockaddr_storage sas, *addr;
1055 int i;
1056
Patrick Caulfield30d3a232007-04-23 16:26:21 +01001057 dlm_local_count = 0;
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001058 for (i = 0; i < DLM_MAX_ADDR_COUNT - 1; i++) {
1059 if (dlm_our_addr(&sas, i))
1060 break;
1061
1062 addr = kmalloc(sizeof(*addr), GFP_KERNEL);
1063 if (!addr)
1064 break;
1065 memcpy(addr, &sas, sizeof(*addr));
1066 dlm_local_addr[dlm_local_count++] = addr;
1067 }
1068}
1069
David Teigland617e82e2007-04-26 13:46:49 -05001070/* Bind to an IP address. SCTP allows multiple address so it can do
1071 multi-homing */
1072static int add_sctp_bind_addr(struct connection *sctp_con,
1073 struct sockaddr_storage *addr,
1074 int addr_len, int num)
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001075{
1076 int result = 0;
1077
1078 if (num == 1)
1079 result = kernel_bind(sctp_con->sock,
1080 (struct sockaddr *) addr,
1081 addr_len);
1082 else
1083 result = kernel_setsockopt(sctp_con->sock, SOL_SCTP,
1084 SCTP_SOCKOPT_BINDX_ADD,
1085 (char *)addr, addr_len);
1086
1087 if (result < 0)
1088 log_print("Can't bind to port %d addr number %d",
1089 dlm_config.ci_tcp_port, num);
1090
1091 return result;
1092}
1093
1094/* Initialise SCTP socket and bind to all interfaces */
1095static int sctp_listen_for_all(void)
1096{
1097 struct socket *sock = NULL;
1098 struct sockaddr_storage localaddr;
1099 struct sctp_event_subscribe subscribe;
1100 int result = -EINVAL, num = 1, i, addr_len;
1101 struct connection *con = nodeid2con(0, GFP_KERNEL);
1102 int bufsize = NEEDED_RMEM;
1103
1104 if (!con)
1105 return -ENOMEM;
1106
1107 log_print("Using SCTP for communications");
1108
1109 result = sock_create_kern(dlm_local_addr[0]->ss_family, SOCK_SEQPACKET,
1110 IPPROTO_SCTP, &sock);
1111 if (result < 0) {
1112 log_print("Can't create comms socket, check SCTP is loaded");
1113 goto out;
1114 }
1115
1116 /* Listen for events */
1117 memset(&subscribe, 0, sizeof(subscribe));
1118 subscribe.sctp_data_io_event = 1;
1119 subscribe.sctp_association_event = 1;
1120 subscribe.sctp_send_failure_event = 1;
1121 subscribe.sctp_shutdown_event = 1;
1122 subscribe.sctp_partial_delivery_event = 1;
1123
David S. Millerdf61c952007-11-06 23:48:57 -08001124 result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001125 (char *)&bufsize, sizeof(bufsize));
1126 if (result)
David Teigland617e82e2007-04-26 13:46:49 -05001127 log_print("Error increasing buffer space on socket %d", result);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001128
1129 result = kernel_setsockopt(sock, SOL_SCTP, SCTP_EVENTS,
David Teigland617e82e2007-04-26 13:46:49 -05001130 (char *)&subscribe, sizeof(subscribe));
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001131 if (result < 0) {
1132 log_print("Failed to set SCTP_EVENTS on socket: result=%d",
1133 result);
1134 goto create_delsock;
1135 }
1136
1137 /* Init con struct */
1138 sock->sk->sk_user_data = con;
1139 con->sock = sock;
1140 con->sock->sk->sk_data_ready = lowcomms_data_ready;
1141 con->rx_action = receive_from_sock;
1142 con->connect_action = sctp_init_assoc;
1143
1144 /* Bind to all interfaces. */
1145 for (i = 0; i < dlm_local_count; i++) {
1146 memcpy(&localaddr, dlm_local_addr[i], sizeof(localaddr));
1147 make_sockaddr(&localaddr, dlm_config.ci_tcp_port, &addr_len);
1148
1149 result = add_sctp_bind_addr(con, &localaddr, addr_len, num);
1150 if (result)
1151 goto create_delsock;
1152 ++num;
1153 }
1154
1155 result = sock->ops->listen(sock, 5);
1156 if (result < 0) {
1157 log_print("Can't set socket listening");
1158 goto create_delsock;
1159 }
1160
1161 return 0;
1162
1163create_delsock:
1164 sock_release(sock);
1165 con->sock = NULL;
1166out:
1167 return result;
1168}
1169
1170static int tcp_listen_for_all(void)
1171{
1172 struct socket *sock = NULL;
1173 struct connection *con = nodeid2con(0, GFP_KERNEL);
1174 int result = -EINVAL;
1175
1176 if (!con)
1177 return -ENOMEM;
1178
1179 /* We don't support multi-homed hosts */
1180 if (dlm_local_addr[1] != NULL) {
David Teigland617e82e2007-04-26 13:46:49 -05001181 log_print("TCP protocol can't handle multi-homed hosts, "
1182 "try SCTP");
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001183 return -EINVAL;
1184 }
1185
1186 log_print("Using TCP for communications");
1187
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001188 sock = tcp_create_listen_sock(con, dlm_local_addr[0]);
1189 if (sock) {
1190 add_sock(sock, con);
1191 result = 0;
1192 }
1193 else {
1194 result = -EADDRINUSE;
1195 }
1196
1197 return result;
1198}
1199
1200
1201
1202static struct writequeue_entry *new_writequeue_entry(struct connection *con,
1203 gfp_t allocation)
1204{
1205 struct writequeue_entry *entry;
1206
1207 entry = kmalloc(sizeof(struct writequeue_entry), allocation);
1208 if (!entry)
1209 return NULL;
1210
1211 entry->page = alloc_page(allocation);
1212 if (!entry->page) {
1213 kfree(entry);
1214 return NULL;
1215 }
1216
1217 entry->offset = 0;
1218 entry->len = 0;
1219 entry->end = 0;
1220 entry->users = 0;
1221 entry->con = con;
1222
1223 return entry;
1224}
1225
David Teigland617e82e2007-04-26 13:46:49 -05001226void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001227{
1228 struct connection *con;
1229 struct writequeue_entry *e;
1230 int offset = 0;
1231 int users = 0;
1232
1233 con = nodeid2con(nodeid, allocation);
1234 if (!con)
1235 return NULL;
1236
1237 spin_lock(&con->writequeue_lock);
1238 e = list_entry(con->writequeue.prev, struct writequeue_entry, list);
1239 if ((&e->list == &con->writequeue) ||
1240 (PAGE_CACHE_SIZE - e->end < len)) {
1241 e = NULL;
1242 } else {
1243 offset = e->end;
1244 e->end += len;
1245 users = e->users++;
1246 }
1247 spin_unlock(&con->writequeue_lock);
1248
1249 if (e) {
1250 got_one:
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001251 *ppc = page_address(e->page) + offset;
1252 return e;
1253 }
1254
1255 e = new_writequeue_entry(con, allocation);
1256 if (e) {
1257 spin_lock(&con->writequeue_lock);
1258 offset = e->end;
1259 e->end += len;
1260 users = e->users++;
1261 list_add_tail(&e->list, &con->writequeue);
1262 spin_unlock(&con->writequeue_lock);
1263 goto got_one;
1264 }
1265 return NULL;
1266}
1267
1268void dlm_lowcomms_commit_buffer(void *mh)
1269{
1270 struct writequeue_entry *e = (struct writequeue_entry *)mh;
1271 struct connection *con = e->con;
1272 int users;
1273
1274 spin_lock(&con->writequeue_lock);
1275 users = --e->users;
1276 if (users)
1277 goto out;
1278 e->len = e->end - e->offset;
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001279 spin_unlock(&con->writequeue_lock);
1280
1281 if (!test_and_set_bit(CF_WRITE_PENDING, &con->flags)) {
1282 queue_work(send_workqueue, &con->swork);
1283 }
1284 return;
1285
1286out:
1287 spin_unlock(&con->writequeue_lock);
1288 return;
1289}
1290
1291/* Send a message */
1292static void send_to_sock(struct connection *con)
1293{
1294 int ret = 0;
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001295 const int msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
1296 struct writequeue_entry *e;
1297 int len, offset;
1298
1299 mutex_lock(&con->sock_mutex);
1300 if (con->sock == NULL)
1301 goto out_connect;
1302
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001303 spin_lock(&con->writequeue_lock);
1304 for (;;) {
1305 e = list_entry(con->writequeue.next, struct writequeue_entry,
1306 list);
1307 if ((struct list_head *) e == &con->writequeue)
1308 break;
1309
1310 len = e->len;
1311 offset = e->offset;
1312 BUG_ON(len == 0 && e->users == 0);
1313 spin_unlock(&con->writequeue_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001314
1315 ret = 0;
1316 if (len) {
Paolo Bonzini1329e3f2009-08-24 13:18:04 -05001317 ret = kernel_sendpage(con->sock, e->page, offset, len,
1318 msg_flags);
Patrick Caulfieldd66f8272007-09-14 08:49:21 +01001319 if (ret == -EAGAIN || ret == 0) {
1320 cond_resched();
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001321 goto out;
Patrick Caulfieldd66f8272007-09-14 08:49:21 +01001322 }
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001323 if (ret <= 0)
1324 goto send_error;
Patrick Caulfieldd66f8272007-09-14 08:49:21 +01001325 }
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001326 /* Don't starve people filling buffers */
1327 cond_resched();
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001328
1329 spin_lock(&con->writequeue_lock);
1330 e->offset += ret;
1331 e->len -= ret;
1332
1333 if (e->len == 0 && e->users == 0) {
1334 list_del(&e->list);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001335 free_entry(e);
1336 continue;
1337 }
1338 }
1339 spin_unlock(&con->writequeue_lock);
1340out:
1341 mutex_unlock(&con->sock_mutex);
1342 return;
1343
1344send_error:
1345 mutex_unlock(&con->sock_mutex);
1346 close_connection(con, false);
1347 lowcomms_connect_sock(con);
1348 return;
1349
1350out_connect:
1351 mutex_unlock(&con->sock_mutex);
1352 if (!test_bit(CF_INIT_PENDING, &con->flags))
1353 lowcomms_connect_sock(con);
1354 return;
1355}
1356
1357static void clean_one_writequeue(struct connection *con)
1358{
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001359 struct writequeue_entry *e, *safe;
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001360
1361 spin_lock(&con->writequeue_lock);
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001362 list_for_each_entry_safe(e, safe, &con->writequeue, list) {
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001363 list_del(&e->list);
1364 free_entry(e);
1365 }
1366 spin_unlock(&con->writequeue_lock);
1367}
1368
1369/* Called from recovery when it knows that a node has
1370 left the cluster */
1371int dlm_lowcomms_close(int nodeid)
1372{
1373 struct connection *con;
1374
1375 log_print("closing connection to node %d", nodeid);
1376 con = nodeid2con(nodeid, 0);
1377 if (con) {
Lars Marowsky-Bree063c4c92009-08-11 16:18:23 -05001378 clear_bit(CF_CONNECT_PENDING, &con->flags);
1379 clear_bit(CF_WRITE_PENDING, &con->flags);
1380 set_bit(CF_CLOSE, &con->flags);
1381 if (cancel_work_sync(&con->swork))
1382 log_print("canceled swork for node %d", nodeid);
1383 if (cancel_work_sync(&con->rwork))
1384 log_print("canceled rwork for node %d", nodeid);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001385 clean_one_writequeue(con);
1386 close_connection(con, true);
1387 }
1388 return 0;
1389}
1390
1391/* Receive workqueue function */
1392static void process_recv_sockets(struct work_struct *work)
1393{
1394 struct connection *con = container_of(work, struct connection, rwork);
1395 int err;
1396
1397 clear_bit(CF_READ_PENDING, &con->flags);
1398 do {
1399 err = con->rx_action(con);
1400 } while (!err);
1401}
1402
1403/* Send workqueue function */
1404static void process_send_sockets(struct work_struct *work)
1405{
1406 struct connection *con = container_of(work, struct connection, swork);
1407
1408 if (test_and_clear_bit(CF_CONNECT_PENDING, &con->flags)) {
1409 con->connect_action(con);
Lars Marowsky-Bree063c4c92009-08-11 16:18:23 -05001410 set_bit(CF_WRITE_PENDING, &con->flags);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001411 }
Lars Marowsky-Bree063c4c92009-08-11 16:18:23 -05001412 if (test_and_clear_bit(CF_WRITE_PENDING, &con->flags))
1413 send_to_sock(con);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001414}
1415
1416
1417/* Discard all entries on the write queues */
1418static void clean_writequeues(void)
1419{
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001420 foreach_conn(clean_one_writequeue);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001421}
1422
1423static void work_stop(void)
1424{
1425 destroy_workqueue(recv_workqueue);
1426 destroy_workqueue(send_workqueue);
1427}
1428
1429static int work_start(void)
1430{
1431 int error;
1432 recv_workqueue = create_workqueue("dlm_recv");
1433 error = IS_ERR(recv_workqueue);
1434 if (error) {
1435 log_print("can't start dlm_recv %d", error);
1436 return error;
1437 }
1438
1439 send_workqueue = create_singlethread_workqueue("dlm_send");
1440 error = IS_ERR(send_workqueue);
1441 if (error) {
1442 log_print("can't start dlm_send %d", error);
1443 destroy_workqueue(recv_workqueue);
1444 return error;
1445 }
1446
1447 return 0;
1448}
1449
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001450static void stop_conn(struct connection *con)
1451{
1452 con->flags |= 0x0F;
Christine Caulfield391fbdc2009-05-07 10:54:16 -05001453 if (con->sock && con->sock->sk)
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001454 con->sock->sk->sk_user_data = NULL;
1455}
1456
1457static void free_conn(struct connection *con)
1458{
1459 close_connection(con, true);
1460 if (con->othercon)
1461 kmem_cache_free(con_cache, con->othercon);
1462 hlist_del(&con->list);
1463 kmem_cache_free(con_cache, con);
1464}
1465
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001466void dlm_lowcomms_stop(void)
1467{
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001468 /* Set all the flags to prevent any
1469 socket activity.
1470 */
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -05001471 mutex_lock(&connections_lock);
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001472 foreach_conn(stop_conn);
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -05001473 mutex_unlock(&connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001474
1475 work_stop();
1476
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -05001477 mutex_lock(&connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001478 clean_writequeues();
1479
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001480 foreach_conn(free_conn);
1481
Matthias Kaehlcke7a936ce2008-05-12 10:04:51 -05001482 mutex_unlock(&connections_lock);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001483 kmem_cache_destroy(con_cache);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001484}
1485
1486int dlm_lowcomms_start(void)
1487{
1488 int error = -EINVAL;
1489 struct connection *con;
Christine Caulfield5e9ccc32009-01-28 12:57:40 -06001490 int i;
1491
1492 for (i = 0; i < CONN_HASH_SIZE; i++)
1493 INIT_HLIST_HEAD(&connection_hash[i]);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001494
1495 init_local();
1496 if (!dlm_local_count) {
David Teigland617e82e2007-04-26 13:46:49 -05001497 error = -ENOTCONN;
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001498 log_print("no local IP address has been set");
1499 goto out;
1500 }
1501
1502 error = -ENOMEM;
1503 con_cache = kmem_cache_create("dlm_conn", sizeof(struct connection),
1504 __alignof__(struct connection), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001505 NULL);
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001506 if (!con_cache)
1507 goto out;
1508
Patrick Caulfield6ed72572007-04-17 15:39:57 +01001509 /* Start listening */
1510 if (dlm_config.ci_protocol == 0)
1511 error = tcp_listen_for_all();
1512 else
1513 error = sctp_listen_for_all();
1514 if (error)
1515 goto fail_unlisten;
1516
1517 error = work_start();
1518 if (error)
1519 goto fail_unlisten;
1520
1521 return 0;
1522
1523fail_unlisten:
1524 con = nodeid2con(0,0);
1525 if (con) {
1526 close_connection(con, false);
1527 kmem_cache_free(con_cache, con);
1528 }
1529 kmem_cache_destroy(con_cache);
1530
1531out:
1532 return error;
1533}