fs/dlm/lowcomms.c

   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 */
 148 static inline int nodeid_hash(int nodeid)
 149 {
 150         return nodeid & (CONN_HASH_SIZE-1);
 151 }
 152
 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  */
 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 */
 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
 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
 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
 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
 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
 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 */
 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
 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
 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
 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
 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
 471 static void lowcomms_error_report(struct sock *sk)
 472 {
 473         struct connection *con;
 474         struct sockaddr_storage saddr;
 475         void (*orig_report)(struct sock *) = NULL;
 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         if (con->sock == NULL ||
 484             kernel_getpeername(con->sock, (struct sockaddr *)&saddr) < 0) {
 485                 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
 486                                    "sending to node %d, port %d, "
 487                                    "sk_err=%d/%d\n", dlm_our_nodeid(),
 488                                    con->nodeid, dlm_config.ci_tcp_port,
 489                                    sk->sk_err, sk->sk_err_soft);
 490         } else if (saddr.ss_family == AF_INET) {
 491                 struct sockaddr_in *sin4 = (struct sockaddr_in *)&saddr;
 492
 493                 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
 494                                    "sending to node %d at %pI4, port %d, "
 495                                    "sk_err=%d/%d\n", dlm_our_nodeid(),
 496                                    con->nodeid, &sin4->sin_addr.s_addr,
 497                                    dlm_config.ci_tcp_port, sk->sk_err,
 498                                    sk->sk_err_soft);
 499         } else {
 500                 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&saddr;
 501
 502                 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
 503                                    "sending to node %d at %u.%u.%u.%u, "
 504                                    "port %d, sk_err=%d/%d\n", dlm_our_nodeid(),
 505                                    con->nodeid, sin6->sin6_addr.s6_addr32[0],
 506                                    sin6->sin6_addr.s6_addr32[1],
 507                                    sin6->sin6_addr.s6_addr32[2],
 508                                    sin6->sin6_addr.s6_addr32[3],
 509                                    dlm_config.ci_tcp_port, sk->sk_err,
 510                                    sk->sk_err_soft);
 511         }
 512 out:
 513         read_unlock_bh(&sk->sk_callback_lock);
 514         if (orig_report)
 515                 orig_report(sk);
 516 }
 517
 518 /* Note: sk_callback_lock must be locked before calling this function. */
 519 static void save_listen_callbacks(struct socket *sock)
 520 {
 521         struct sock *sk = sock->sk;
 522
 523         listen_sock.sk_data_ready = sk->sk_data_ready;
 524         listen_sock.sk_state_change = sk->sk_state_change;
 525         listen_sock.sk_write_space = sk->sk_write_space;
 526         listen_sock.sk_error_report = sk->sk_error_report;
 527 }
 528
 529 static void restore_callbacks(struct socket *sock)
 530 {
 531         struct sock *sk = sock->sk;
 532
 533         write_lock_bh(&sk->sk_callback_lock);
 534         sk->sk_user_data = NULL;
 535         sk->sk_data_ready = listen_sock.sk_data_ready;
 536         sk->sk_state_change = listen_sock.sk_state_change;
 537         sk->sk_write_space = listen_sock.sk_write_space;
 538         sk->sk_error_report = listen_sock.sk_error_report;
 539         write_unlock_bh(&sk->sk_callback_lock);
 540 }
 541
 542 /* Make a socket active */
 543 static void add_sock(struct socket *sock, struct connection *con)
 544 {
 545         struct sock *sk = sock->sk;
 546
 547         write_lock_bh(&sk->sk_callback_lock);
 548         con->sock = sock;
 549
 550         sk->sk_user_data = con;
 551         /* Install a data_ready callback */
 552         sk->sk_data_ready = lowcomms_data_ready;
 553         sk->sk_write_space = lowcomms_write_space;
 554         sk->sk_state_change = lowcomms_state_change;
 555         sk->sk_allocation = GFP_NOFS;
 556         sk->sk_error_report = lowcomms_error_report;
 557         write_unlock_bh(&sk->sk_callback_lock);
 558 }
 559
 560 /* Add the port number to an IPv6 or 4 sockaddr and return the address
 561    length */
 562 static void make_sockaddr(struct sockaddr_storage *saddr, uint16_t port,
 563                           int *addr_len)
 564 {
 565         saddr->ss_family =  dlm_local_addr[0]->ss_family;
 566         if (saddr->ss_family == AF_INET) {
 567                 struct sockaddr_in *in4_addr = (struct sockaddr_in *)saddr;
 568                 in4_addr->sin_port = cpu_to_be16(port);
 569                 *addr_len = sizeof(struct sockaddr_in);
 570                 memset(&in4_addr->sin_zero, 0, sizeof(in4_addr->sin_zero));
 571         } else {
 572                 struct sockaddr_in6 *in6_addr = (struct sockaddr_in6 *)saddr;
 573                 in6_addr->sin6_port = cpu_to_be16(port);
 574                 *addr_len = sizeof(struct sockaddr_in6);
 575         }
 576         memset((char *)saddr + *addr_len, 0, sizeof(struct sockaddr_storage) - *addr_len);
 577 }
 578
 579 /* Close a remote connection and tidy up */
 580 static void close_connection(struct connection *con, bool and_other,
 581                              bool tx, bool rx)
 582 {
 583         bool closing = test_and_set_bit(CF_CLOSING, &con->flags);
 584
 585         if (tx && !closing && cancel_work_sync(&con->swork)) {
 586                 log_print("canceled swork for node %d", con->nodeid);
 587                 clear_bit(CF_WRITE_PENDING, &con->flags);
 588         }
 589         if (rx && !closing && cancel_work_sync(&con->rwork)) {
 590                 log_print("canceled rwork for node %d", con->nodeid);
 591                 clear_bit(CF_READ_PENDING, &con->flags);
 592         }
 593
 594         mutex_lock(&con->sock_mutex);
 595         if (con->sock) {
 596                 restore_callbacks(con->sock);
 597                 sock_release(con->sock);
 598                 con->sock = NULL;
 599         }
 600         if (con->othercon && and_other) {
 601                 /* Will only re-enter once. */
 602                 close_connection(con->othercon, false, true, true);
 603         }
 604
 605         con->rx_leftover = 0;
 606         con->retries = 0;
 607         mutex_unlock(&con->sock_mutex);
 608         clear_bit(CF_CLOSING, &con->flags);
 609 }
 610
 611 static void shutdown_connection(struct connection *con)
 612 {
 613         int ret;
 614
 615         if (cancel_work_sync(&con->swork)) {
 616                 log_print("canceled swork for node %d", con->nodeid);
 617                 clear_bit(CF_WRITE_PENDING, &con->flags);
 618         }
 619
 620         mutex_lock(&con->sock_mutex);
 621         /* nothing to shutdown */
 622         if (!con->sock) {
 623                 mutex_unlock(&con->sock_mutex);
 624                 return;
 625         }
 626
 627         set_bit(CF_SHUTDOWN, &con->flags);
 628         ret = kernel_sock_shutdown(con->sock, SHUT_WR);
 629         mutex_unlock(&con->sock_mutex);
 630         if (ret) {
 631                 log_print("Connection %p failed to shutdown: %d will force close",
 632                           con, ret);
 633                 goto force_close;
 634         } else {
 635                 ret = wait_event_timeout(con->shutdown_wait,
 636                                          !test_bit(CF_SHUTDOWN, &con->flags),
 637                                          DLM_SHUTDOWN_WAIT_TIMEOUT);
 638                 if (ret == 0) {
 639                         log_print("Connection %p shutdown timed out, will force close",
 640                                   con);
 641                         goto force_close;
 642                 }
 643         }
 644
 645         return;
 646
 647 force_close:
 648         clear_bit(CF_SHUTDOWN, &con->flags);
 649         close_connection(con, false, true, true);
 650 }
 651
 652 static void dlm_tcp_shutdown(struct connection *con)
 653 {
 654         if (con->othercon)
 655                 shutdown_connection(con->othercon);
 656         shutdown_connection(con);
 657 }
 658
 659 static int con_realloc_receive_buf(struct connection *con, int newlen)
 660 {
 661         unsigned char *newbuf;
 662
 663         newbuf = kmalloc(newlen, GFP_NOFS);
 664         if (!newbuf)
 665                 return -ENOMEM;
 666
 667         /* copy any leftover from last receive */
 668         if (con->rx_leftover)
 669                 memmove(newbuf, con->rx_buf, con->rx_leftover);
 670
 671         /* swap to new buffer space */
 672         kfree(con->rx_buf);
 673         con->rx_buflen = newlen;
 674         con->rx_buf = newbuf;
 675
 676         return 0;
 677 }
 678
 679 /* Data received from remote end */
 680 static int receive_from_sock(struct connection *con)
 681 {
 682         int call_again_soon = 0;
 683         struct msghdr msg;
 684         struct kvec iov;
 685         int ret, buflen;
 686
 687         mutex_lock(&con->sock_mutex);
 688
 689         if (con->sock == NULL) {
 690                 ret = -EAGAIN;
 691                 goto out_close;
 692         }
 693
 694         if (con->nodeid == 0) {
 695                 ret = -EINVAL;
 696                 goto out_close;
 697         }
 698
 699         /* realloc if we get new buffer size to read out */
 700         buflen = dlm_config.ci_buffer_size;
 701         if (con->rx_buflen != buflen && con->rx_leftover <= buflen) {
 702                 ret = con_realloc_receive_buf(con, buflen);
 703                 if (ret < 0)
 704                         goto out_resched;
 705         }
 706
 707         /* calculate new buffer parameter regarding last receive and
 708          * possible leftover bytes
 709          */
 710         iov.iov_base = con->rx_buf + con->rx_leftover;
 711         iov.iov_len = con->rx_buflen - con->rx_leftover;
 712
 713         memset(&msg, 0, sizeof(msg));
 714         msg.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
 715         ret = kernel_recvmsg(con->sock, &msg, &iov, 1, iov.iov_len,
 716                              msg.msg_flags);
 717         if (ret <= 0)
 718                 goto out_close;
 719         else if (ret == iov.iov_len)
 720                 call_again_soon = 1;
 721
 722         /* new buflen according readed bytes and leftover from last receive */
 723         buflen = ret + con->rx_leftover;
 724         ret = dlm_process_incoming_buffer(con->nodeid, con->rx_buf, buflen);
 725         if (ret < 0)
 726                 goto out_close;
 727
 728         /* calculate leftover bytes from process and put it into begin of
 729          * the receive buffer, so next receive we have the full message
 730          * at the start address of the receive buffer.
 731          */
 732         con->rx_leftover = buflen - ret;
 733         if (con->rx_leftover) {
 734                 memmove(con->rx_buf, con->rx_buf + ret,
 735                         con->rx_leftover);
 736                 call_again_soon = true;
 737         }
 738
 739         if (call_again_soon)
 740                 goto out_resched;
 741
 742         mutex_unlock(&con->sock_mutex);
 743         return 0;
 744
 745 out_resched:
 746         if (!test_and_set_bit(CF_READ_PENDING, &con->flags))
 747                 queue_work(recv_workqueue, &con->rwork);
 748         mutex_unlock(&con->sock_mutex);
 749         return -EAGAIN;
 750
 751 out_close:
 752         mutex_unlock(&con->sock_mutex);
 753         if (ret != -EAGAIN) {
 754                 /* Reconnect when there is something to send */
 755                 close_connection(con, false, true, false);
 756                 if (ret == 0) {
 757                         log_print("connection %p got EOF from %d",
 758                                   con, con->nodeid);
 759                         /* handling for tcp shutdown */
 760                         clear_bit(CF_SHUTDOWN, &con->flags);
 761                         wake_up(&con->shutdown_wait);
 762                         /* signal to breaking receive worker */
 763                         ret = -1;
 764                 }
 765         }
 766         return ret;
 767 }
 768
 769 /* Listening socket is busy, accept a connection */
 770 static int accept_from_sock(struct connection *con)
 771 {
 772         int result;
 773         struct sockaddr_storage peeraddr;
 774         struct socket *newsock;
 775         int len;
 776         int nodeid;
 777         struct connection *newcon;
 778         struct connection *addcon;
 779         unsigned int mark;
 780
 781         if (!dlm_allow_conn) {
 782                 return -1;
 783         }
 784
 785         mutex_lock_nested(&con->sock_mutex, 0);
 786
 787         if (!con->sock) {
 788                 mutex_unlock(&con->sock_mutex);
 789                 return -ENOTCONN;
 790         }
 791
 792         result = kernel_accept(con->sock, &newsock, O_NONBLOCK);
 793         if (result < 0)
 794                 goto accept_err;
 795
 796         /* Get the connected socket's peer */
 797         memset(&peeraddr, 0, sizeof(peeraddr));
 798         len = newsock->ops->getname(newsock, (struct sockaddr *)&peeraddr, 2);
 799         if (len < 0) {
 800                 result = -ECONNABORTED;
 801                 goto accept_err;
 802         }
 803
 804         /* Get the new node's NODEID */
 805         make_sockaddr(&peeraddr, 0, &len);
 806         if (addr_to_nodeid(&peeraddr, &nodeid)) {
 807                 unsigned char *b=(unsigned char *)&peeraddr;
 808                 log_print("connect from non cluster node");
 809                 print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
 810                                      b, sizeof(struct sockaddr_storage));
 811                 sock_release(newsock);
 812                 mutex_unlock(&con->sock_mutex);
 813                 return -1;
 814         }
 815
 816         dlm_comm_mark(nodeid, &mark);
 817         sock_set_mark(newsock->sk, mark);
 818
 819         log_print("got connection from %d", nodeid);
 820
 821         /*  Check to see if we already have a connection to this node. This
 822          *  could happen if the two nodes initiate a connection at roughly
 823          *  the same time and the connections cross on the wire.
 824          *  In this case we store the incoming one in "othercon"
 825          */
 826         newcon = nodeid2con(nodeid, GFP_NOFS);
 827         if (!newcon) {
 828                 result = -ENOMEM;
 829                 goto accept_err;
 830         }
 831         mutex_lock_nested(&newcon->sock_mutex, 1);
 832         if (newcon->sock) {
 833                 struct connection *othercon = newcon->othercon;
 834
 835                 if (!othercon) {
 836                         othercon = kzalloc(sizeof(*othercon), GFP_NOFS);
 837                         if (!othercon) {
 838                                 log_print("failed to allocate incoming socket");
 839                                 mutex_unlock(&newcon->sock_mutex);
 840                                 result = -ENOMEM;
 841                                 goto accept_err;
 842                         }
 843
 844                         othercon->rx_buflen = dlm_config.ci_buffer_size;
 845                         othercon->rx_buf = kmalloc(othercon->rx_buflen, GFP_NOFS);
 846                         if (!othercon->rx_buf) {
 847                                 mutex_unlock(&newcon->sock_mutex);
 848                                 kfree(othercon);
 849                                 log_print("failed to allocate incoming socket receive buffer");
 850                                 result = -ENOMEM;
 851                                 goto accept_err;
 852                         }
 853
 854                         othercon->nodeid = nodeid;
 855                         othercon->rx_action = receive_from_sock;
 856                         mutex_init(&othercon->sock_mutex);
 857                         INIT_LIST_HEAD(&othercon->writequeue);
 858                         spin_lock_init(&othercon->writequeue_lock);
 859                         INIT_WORK(&othercon->swork, process_send_sockets);
 860                         INIT_WORK(&othercon->rwork, process_recv_sockets);
 861                         init_waitqueue_head(&othercon->shutdown_wait);
 862                         set_bit(CF_IS_OTHERCON, &othercon->flags);
 863                 } else {
 864                         /* close other sock con if we have something new */
 865                         close_connection(othercon, false, true, false);
 866                 }
 867
 868                 mutex_lock_nested(&othercon->sock_mutex, 2);
 869                 newcon->othercon = othercon;
 870                 add_sock(newsock, othercon);
 871                 addcon = othercon;
 872                 mutex_unlock(&othercon->sock_mutex);
 873         }
 874         else {
 875                 newcon->rx_action = receive_from_sock;
 876                 /* accept copies the sk after we've saved the callbacks, so we
 877                    don't want to save them a second time or comm errors will
 878                    result in calling sk_error_report recursively. */
 879                 add_sock(newsock, newcon);
 880                 addcon = newcon;
 881         }
 882
 883         mutex_unlock(&newcon->sock_mutex);
 884
 885         /*
 886          * Add it to the active queue in case we got data
 887          * between processing the accept adding the socket
 888          * to the read_sockets list
 889          */
 890         if (!test_and_set_bit(CF_READ_PENDING, &addcon->flags))
 891                 queue_work(recv_workqueue, &addcon->rwork);
 892         mutex_unlock(&con->sock_mutex);
 893
 894         return 0;
 895
 896 accept_err:
 897         mutex_unlock(&con->sock_mutex);
 898         if (newsock)
 899                 sock_release(newsock);
 900
 901         if (result != -EAGAIN)
 902                 log_print("error accepting connection from node: %d", result);
 903         return result;
 904 }
 905
 906 static void free_entry(struct writequeue_entry *e)
 907 {
 908         __free_page(e->page);
 909         kfree(e);
 910 }
 911
 912 /*
 913  * writequeue_entry_complete - try to delete and free write queue entry
 914  * @e: write queue entry to try to delete
 915  * @completed: bytes completed
 916  *
 917  * writequeue_lock must be held.
 918  */
 919 static void writequeue_entry_complete(struct writequeue_entry *e, int completed)
 920 {
 921         e->offset += completed;
 922         e->len -= completed;
 923
 924         if (e->len == 0 && e->users == 0) {
 925                 list_del(&e->list);
 926                 free_entry(e);
 927         }
 928 }
 929
 930 /*
 931  * sctp_bind_addrs - bind a SCTP socket to all our addresses
 932  */
 933 static int sctp_bind_addrs(struct connection *con, uint16_t port)
 934 {
 935         struct sockaddr_storage localaddr;
 936         struct sockaddr *addr = (struct sockaddr *)&localaddr;
 937         int i, addr_len, result = 0;
 938
 939         for (i = 0; i < dlm_local_count; i++) {
 940                 memcpy(&localaddr, dlm_local_addr[i], sizeof(localaddr));
 941                 make_sockaddr(&localaddr, port, &addr_len);
 942
 943                 if (!i)
 944                         result = kernel_bind(con->sock, addr, addr_len);
 945                 else
 946                         result = sock_bind_add(con->sock->sk, addr, addr_len);
 947
 948                 if (result < 0) {
 949                         log_print("Can't bind to %d addr number %d, %d.\n",
 950                                   port, i + 1, result);
 951                         break;
 952                 }
 953         }
 954         return result;
 955 }
 956
 957 /* Initiate an SCTP association.
 958    This is a special case of send_to_sock() in that we don't yet have a
 959    peeled-off socket for this association, so we use the listening socket
 960    and add the primary IP address of the remote node.
 961  */
 962 static void sctp_connect_to_sock(struct connection *con)
 963 {
 964         struct sockaddr_storage daddr;
 965         int result;
 966         int addr_len;
 967         struct socket *sock;
 968         unsigned int mark;
 969
 970         if (con->nodeid == 0) {
 971                 log_print("attempt to connect sock 0 foiled");
 972                 return;
 973         }
 974
 975         dlm_comm_mark(con->nodeid, &mark);
 976
 977         mutex_lock(&con->sock_mutex);
 978
 979         /* Some odd races can cause double-connects, ignore them */
 980         if (con->retries++ > MAX_CONNECT_RETRIES)
 981                 goto out;
 982
 983         if (con->sock) {
 984                 log_print("node %d already connected.", con->nodeid);
 985                 goto out;
 986         }
 987
 988         memset(&daddr, 0, sizeof(daddr));
 989         result = nodeid_to_addr(con->nodeid, &daddr, NULL, true);
 990         if (result < 0) {
 991                 log_print("no address for nodeid %d", con->nodeid);
 992                 goto out;
 993         }
 994
 995         /* Create a socket to communicate with */
 996         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
 997                                   SOCK_STREAM, IPPROTO_SCTP, &sock);
 998         if (result < 0)
 999                 goto socket_err;
1000
1001         sock_set_mark(sock->sk, mark);
1002
1003         con->rx_action = receive_from_sock;
1004         con->connect_action = sctp_connect_to_sock;
1005         add_sock(sock, con);
1006
1007         /* Bind to all addresses. */
1008         if (sctp_bind_addrs(con, 0))
1009                 goto bind_err;
1010
1011         make_sockaddr(&daddr, dlm_config.ci_tcp_port, &addr_len);
1012
1013         log_print("connecting to %d", con->nodeid);
1014
1015         /* Turn off Nagle's algorithm */
1016         sctp_sock_set_nodelay(sock->sk);
1017
1018         /*
1019          * Make sock->ops->connect() function return in specified time,
1020          * since O_NONBLOCK argument in connect() function does not work here,
1021          * then, we should restore the default value of this attribute.
1022          */
1023         sock_set_sndtimeo(sock->sk, 5);
1024         result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
1025                                    0);
1026         sock_set_sndtimeo(sock->sk, 0);
1027
1028         if (result == -EINPROGRESS)
1029                 result = 0;
1030         if (result == 0)
1031                 goto out;
1032
1033 bind_err:
1034         con->sock = NULL;
1035         sock_release(sock);
1036
1037 socket_err:
1038         /*
1039          * Some errors are fatal and this list might need adjusting. For other
1040          * errors we try again until the max number of retries is reached.
1041          */
1042         if (result != -EHOSTUNREACH &&
1043             result != -ENETUNREACH &&
1044             result != -ENETDOWN &&
1045             result != -EINVAL &&
1046             result != -EPROTONOSUPPORT) {
1047                 log_print("connect %d try %d error %d", con->nodeid,
1048                           con->retries, result);
1049                 mutex_unlock(&con->sock_mutex);
1050                 msleep(1000);
1051                 lowcomms_connect_sock(con);
1052                 return;
1053         }
1054
1055 out:
1056         mutex_unlock(&con->sock_mutex);
1057 }
1058
1059 /* Connect a new socket to its peer */
1060 static void tcp_connect_to_sock(struct connection *con)
1061 {
1062         struct sockaddr_storage saddr, src_addr;
1063         int addr_len;
1064         struct socket *sock = NULL;
1065         unsigned int mark;
1066         int result;
1067
1068         if (con->nodeid == 0) {
1069                 log_print("attempt to connect sock 0 foiled");
1070                 return;
1071         }
1072
1073         dlm_comm_mark(con->nodeid, &mark);
1074
1075         mutex_lock(&con->sock_mutex);
1076         if (con->retries++ > MAX_CONNECT_RETRIES)
1077                 goto out;
1078
1079         /* Some odd races can cause double-connects, ignore them */
1080         if (con->sock)
1081                 goto out;
1082
1083         /* Create a socket to communicate with */
1084         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1085                                   SOCK_STREAM, IPPROTO_TCP, &sock);
1086         if (result < 0)
1087                 goto out_err;
1088
1089         sock_set_mark(sock->sk, mark);
1090
1091         memset(&saddr, 0, sizeof(saddr));
1092         result = nodeid_to_addr(con->nodeid, &saddr, NULL, false);
1093         if (result < 0) {
1094                 log_print("no address for nodeid %d", con->nodeid);
1095                 goto out_err;
1096         }
1097
1098         con->rx_action = receive_from_sock;
1099         con->connect_action = tcp_connect_to_sock;
1100         con->shutdown_action = dlm_tcp_shutdown;
1101         add_sock(sock, con);
1102
1103         /* Bind to our cluster-known address connecting to avoid
1104            routing problems */
1105         memcpy(&src_addr, dlm_local_addr[0], sizeof(src_addr));
1106         make_sockaddr(&src_addr, 0, &addr_len);
1107         result = sock->ops->bind(sock, (struct sockaddr *) &src_addr,
1108                                  addr_len);
1109         if (result < 0) {
1110                 log_print("could not bind for connect: %d", result);
1111                 /* This *may* not indicate a critical error */
1112         }
1113
1114         make_sockaddr(&saddr, dlm_config.ci_tcp_port, &addr_len);
1115
1116         log_print("connecting to %d", con->nodeid);
1117
1118         /* Turn off Nagle's algorithm */
1119         tcp_sock_set_nodelay(sock->sk);
1120
1121         result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
1122                                    O_NONBLOCK);
1123         if (result == -EINPROGRESS)
1124                 result = 0;
1125         if (result == 0)
1126                 goto out;
1127
1128 out_err:
1129         if (con->sock) {
1130                 sock_release(con->sock);
1131                 con->sock = NULL;
1132         } else if (sock) {
1133                 sock_release(sock);
1134         }
1135         /*
1136          * Some errors are fatal and this list might need adjusting. For other
1137          * errors we try again until the max number of retries is reached.
1138          */
1139         if (result != -EHOSTUNREACH &&
1140             result != -ENETUNREACH &&
1141             result != -ENETDOWN &&
1142             result != -EINVAL &&
1143             result != -EPROTONOSUPPORT) {
1144                 log_print("connect %d try %d error %d", con->nodeid,
1145                           con->retries, result);
1146                 mutex_unlock(&con->sock_mutex);
1147                 msleep(1000);
1148                 lowcomms_connect_sock(con);
1149                 return;
1150         }
1151 out:
1152         mutex_unlock(&con->sock_mutex);
1153         return;
1154 }
1155
1156 static struct socket *tcp_create_listen_sock(struct connection *con,
1157                                              struct sockaddr_storage *saddr)
1158 {
1159         struct socket *sock = NULL;
1160         int result = 0;
1161         int addr_len;
1162
1163         if (dlm_local_addr[0]->ss_family == AF_INET)
1164                 addr_len = sizeof(struct sockaddr_in);
1165         else
1166                 addr_len = sizeof(struct sockaddr_in6);
1167
1168         /* Create a socket to communicate with */
1169         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1170                                   SOCK_STREAM, IPPROTO_TCP, &sock);
1171         if (result < 0) {
1172                 log_print("Can't create listening comms socket");
1173                 goto create_out;
1174         }
1175
1176         sock_set_mark(sock->sk, dlm_config.ci_mark);
1177
1178         /* Turn off Nagle's algorithm */
1179         tcp_sock_set_nodelay(sock->sk);
1180
1181         sock_set_reuseaddr(sock->sk);
1182
1183         write_lock_bh(&sock->sk->sk_callback_lock);
1184         sock->sk->sk_user_data = con;
1185         save_listen_callbacks(sock);
1186         con->rx_action = accept_from_sock;
1187         con->connect_action = tcp_connect_to_sock;
1188         write_unlock_bh(&sock->sk->sk_callback_lock);
1189
1190         /* Bind to our port */
1191         make_sockaddr(saddr, dlm_config.ci_tcp_port, &addr_len);
1192         result = sock->ops->bind(sock, (struct sockaddr *) saddr, addr_len);
1193         if (result < 0) {
1194                 log_print("Can't bind to port %d", dlm_config.ci_tcp_port);
1195                 sock_release(sock);
1196                 sock = NULL;
1197                 con->sock = NULL;
1198                 goto create_out;
1199         }
1200         sock_set_keepalive(sock->sk);
1201
1202         result = sock->ops->listen(sock, 5);
1203         if (result < 0) {
1204                 log_print("Can't listen on port %d", dlm_config.ci_tcp_port);
1205                 sock_release(sock);
1206                 sock = NULL;
1207                 goto create_out;
1208         }
1209
1210 create_out:
1211         return sock;
1212 }
1213
1214 /* Get local addresses */
1215 static void init_local(void)
1216 {
1217         struct sockaddr_storage sas, *addr;
1218         int i;
1219
1220         dlm_local_count = 0;
1221         for (i = 0; i < DLM_MAX_ADDR_COUNT; i++) {
1222                 if (dlm_our_addr(&sas, i))
1223                         break;
1224
1225                 addr = kmemdup(&sas, sizeof(*addr), GFP_NOFS);
1226                 if (!addr)
1227                         break;
1228                 dlm_local_addr[dlm_local_count++] = addr;
1229         }
1230 }
1231
1232 static void deinit_local(void)
1233 {
1234         int i;
1235
1236         for (i = 0; i < dlm_local_count; i++)
1237                 kfree(dlm_local_addr[i]);
1238 }
1239
1240 /* Initialise SCTP socket and bind to all interfaces */
1241 static int sctp_listen_for_all(void)
1242 {
1243         struct socket *sock = NULL;
1244         int result = -EINVAL;
1245         struct connection *con = nodeid2con(0, GFP_NOFS);
1246
1247         if (!con)
1248                 return -ENOMEM;
1249
1250         log_print("Using SCTP for communications");
1251
1252         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1253                                   SOCK_STREAM, IPPROTO_SCTP, &sock);
1254         if (result < 0) {
1255                 log_print("Can't create comms socket, check SCTP is loaded");
1256                 goto out;
1257         }
1258
1259         sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
1260         sock_set_mark(sock->sk, dlm_config.ci_mark);
1261         sctp_sock_set_nodelay(sock->sk);
1262
1263         write_lock_bh(&sock->sk->sk_callback_lock);
1264         /* Init con struct */
1265         sock->sk->sk_user_data = con;
1266         save_listen_callbacks(sock);
1267         con->sock = sock;
1268         con->sock->sk->sk_data_ready = lowcomms_data_ready;
1269         con->rx_action = accept_from_sock;
1270         con->connect_action = sctp_connect_to_sock;
1271
1272         write_unlock_bh(&sock->sk->sk_callback_lock);
1273
1274         /* Bind to all addresses. */
1275         if (sctp_bind_addrs(con, dlm_config.ci_tcp_port))
1276                 goto create_delsock;
1277
1278         result = sock->ops->listen(sock, 5);
1279         if (result < 0) {
1280                 log_print("Can't set socket listening");
1281                 goto create_delsock;
1282         }
1283
1284         return 0;
1285
1286 create_delsock:
1287         sock_release(sock);
1288         con->sock = NULL;
1289 out:
1290         return result;
1291 }
1292
1293 static int tcp_listen_for_all(void)
1294 {
1295         struct socket *sock = NULL;
1296         struct connection *con = nodeid2con(0, GFP_NOFS);
1297         int result = -EINVAL;
1298
1299         if (!con)
1300                 return -ENOMEM;
1301
1302         /* We don't support multi-homed hosts */
1303         if (dlm_local_addr[1] != NULL) {
1304                 log_print("TCP protocol can't handle multi-homed hosts, "
1305                           "try SCTP");
1306                 return -EINVAL;
1307         }
1308
1309         log_print("Using TCP for communications");
1310
1311         sock = tcp_create_listen_sock(con, dlm_local_addr[0]);
1312         if (sock) {
1313                 add_sock(sock, con);
1314                 result = 0;
1315         }
1316         else {
1317                 result = -EADDRINUSE;
1318         }
1319
1320         return result;
1321 }
1322
1323
1324
1325 static struct writequeue_entry *new_writequeue_entry(struct connection *con,
1326                                                      gfp_t allocation)
1327 {
1328         struct writequeue_entry *entry;
1329
1330         entry = kmalloc(sizeof(struct writequeue_entry), allocation);
1331         if (!entry)
1332                 return NULL;
1333
1334         entry->page = alloc_page(allocation);
1335         if (!entry->page) {
1336                 kfree(entry);
1337                 return NULL;
1338         }
1339
1340         entry->offset = 0;
1341         entry->len = 0;
1342         entry->end = 0;
1343         entry->users = 0;
1344         entry->con = con;
1345
1346         return entry;
1347 }
1348
1349 void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
1350 {
1351         struct connection *con;
1352         struct writequeue_entry *e;
1353         int offset = 0;
1354
1355         con = nodeid2con(nodeid, allocation);
1356         if (!con)
1357                 return NULL;
1358
1359         spin_lock(&con->writequeue_lock);
1360         e = list_entry(con->writequeue.prev, struct writequeue_entry, list);
1361         if ((&e->list == &con->writequeue) ||
1362             (PAGE_SIZE - e->end < len)) {
1363                 e = NULL;
1364         } else {
1365                 offset = e->end;
1366                 e->end += len;
1367                 e->users++;
1368         }
1369         spin_unlock(&con->writequeue_lock);
1370
1371         if (e) {
1372         got_one:
1373                 *ppc = page_address(e->page) + offset;
1374                 return e;
1375         }
1376
1377         e = new_writequeue_entry(con, allocation);
1378         if (e) {
1379                 spin_lock(&con->writequeue_lock);
1380                 offset = e->end;
1381                 e->end += len;
1382                 e->users++;
1383                 list_add_tail(&e->list, &con->writequeue);
1384                 spin_unlock(&con->writequeue_lock);
1385                 goto got_one;
1386         }
1387         return NULL;
1388 }
1389
1390 void dlm_lowcomms_commit_buffer(void *mh)
1391 {
1392         struct writequeue_entry *e = (struct writequeue_entry *)mh;
1393         struct connection *con = e->con;
1394         int users;
1395
1396         spin_lock(&con->writequeue_lock);
1397         users = --e->users;
1398         if (users)
1399                 goto out;
1400         e->len = e->end - e->offset;
1401         spin_unlock(&con->writequeue_lock);
1402
1403         queue_work(send_workqueue, &con->swork);
1404         return;
1405
1406 out:
1407         spin_unlock(&con->writequeue_lock);
1408         return;
1409 }
1410
1411 /* Send a message */
1412 static void send_to_sock(struct connection *con)
1413 {
1414         int ret = 0;
1415         const int msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
1416         struct writequeue_entry *e;
1417         int len, offset;
1418         int count = 0;
1419
1420         mutex_lock(&con->sock_mutex);
1421         if (con->sock == NULL)
1422                 goto out_connect;
1423
1424         spin_lock(&con->writequeue_lock);
1425         for (;;) {
1426                 e = list_entry(con->writequeue.next, struct writequeue_entry,
1427                                list);
1428                 if ((struct list_head *) e == &con->writequeue)
1429                         break;
1430
1431                 len = e->len;
1432                 offset = e->offset;
1433                 BUG_ON(len == 0 && e->users == 0);
1434                 spin_unlock(&con->writequeue_lock);
1435
1436                 ret = 0;
1437                 if (len) {
1438                         ret = kernel_sendpage(con->sock, e->page, offset, len,
1439                                               msg_flags);
1440                         if (ret == -EAGAIN || ret == 0) {
1441                                 if (ret == -EAGAIN &&
1442                                     test_bit(SOCKWQ_ASYNC_NOSPACE, &con->sock->flags) &&
1443                                     !test_and_set_bit(CF_APP_LIMITED, &con->flags)) {
1444                                         /* Notify TCP that we're limited by the
1445                                          * application window size.
1446                                          */
1447                                         set_bit(SOCK_NOSPACE, &con->sock->flags);
1448                                         con->sock->sk->sk_write_pending++;
1449                                 }
1450                                 cond_resched();
1451                                 goto out;
1452                         } else if (ret < 0)
1453                                 goto send_error;
1454                 }
1455
1456                 /* Don't starve people filling buffers */
1457                 if (++count >= MAX_SEND_MSG_COUNT) {
1458                         cond_resched();
1459                         count = 0;
1460                 }
1461
1462                 spin_lock(&con->writequeue_lock);
1463                 writequeue_entry_complete(e, ret);
1464         }
1465         spin_unlock(&con->writequeue_lock);
1466 out:
1467         mutex_unlock(&con->sock_mutex);
1468         return;
1469
1470 send_error:
1471         mutex_unlock(&con->sock_mutex);
1472         close_connection(con, false, false, true);
1473         /* Requeue the send work. When the work daemon runs again, it will try
1474            a new connection, then call this function again. */
1475         queue_work(send_workqueue, &con->swork);
1476         return;
1477
1478 out_connect:
1479         mutex_unlock(&con->sock_mutex);
1480         queue_work(send_workqueue, &con->swork);
1481         cond_resched();
1482 }
1483
1484 static void clean_one_writequeue(struct connection *con)
1485 {
1486         struct writequeue_entry *e, *safe;
1487
1488         spin_lock(&con->writequeue_lock);
1489         list_for_each_entry_safe(e, safe, &con->writequeue, list) {
1490                 list_del(&e->list);
1491                 free_entry(e);
1492         }
1493         spin_unlock(&con->writequeue_lock);
1494 }
1495
1496 /* Called from recovery when it knows that a node has
1497    left the cluster */
1498 int dlm_lowcomms_close(int nodeid)
1499 {
1500         struct connection *con;
1501         struct dlm_node_addr *na;
1502
1503         log_print("closing connection to node %d", nodeid);
1504         con = nodeid2con(nodeid, 0);
1505         if (con) {
1506                 set_bit(CF_CLOSE, &con->flags);
1507                 close_connection(con, true, true, true);
1508                 clean_one_writequeue(con);
1509         }
1510
1511         spin_lock(&dlm_node_addrs_spin);
1512         na = find_node_addr(nodeid);
1513         if (na) {
1514                 list_del(&na->list);
1515                 while (na->addr_count--)
1516                         kfree(na->addr[na->addr_count]);
1517                 kfree(na);
1518         }
1519         spin_unlock(&dlm_node_addrs_spin);
1520
1521         return 0;
1522 }
1523
1524 /* Receive workqueue function */
1525 static void process_recv_sockets(struct work_struct *work)
1526 {
1527         struct connection *con = container_of(work, struct connection, rwork);
1528         int err;
1529
1530         clear_bit(CF_READ_PENDING, &con->flags);
1531         do {
1532                 err = con->rx_action(con);
1533         } while (!err);
1534 }
1535
1536 /* Send workqueue function */
1537 static void process_send_sockets(struct work_struct *work)
1538 {
1539         struct connection *con = container_of(work, struct connection, swork);
1540
1541         clear_bit(CF_WRITE_PENDING, &con->flags);
1542         if (con->sock == NULL) /* not mutex protected so check it inside too */
1543                 con->connect_action(con);
1544         if (!list_empty(&con->writequeue))
1545                 send_to_sock(con);
1546 }
1547
1548 static void work_stop(void)
1549 {
1550         if (recv_workqueue)
1551                 destroy_workqueue(recv_workqueue);
1552         if (send_workqueue)
1553                 destroy_workqueue(send_workqueue);
1554 }
1555
1556 static int work_start(void)
1557 {
1558         recv_workqueue = alloc_workqueue("dlm_recv",
1559                                          WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1560         if (!recv_workqueue) {
1561                 log_print("can't start dlm_recv");
1562                 return -ENOMEM;
1563         }
1564
1565         send_workqueue = alloc_workqueue("dlm_send",
1566                                          WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1567         if (!send_workqueue) {
1568                 log_print("can't start dlm_send");
1569                 destroy_workqueue(recv_workqueue);
1570                 return -ENOMEM;
1571         }
1572
1573         return 0;
1574 }
1575
1576 static void _stop_conn(struct connection *con, bool and_other)
1577 {
1578         mutex_lock(&con->sock_mutex);
1579         set_bit(CF_CLOSE, &con->flags);
1580         set_bit(CF_READ_PENDING, &con->flags);
1581         set_bit(CF_WRITE_PENDING, &con->flags);
1582         if (con->sock && con->sock->sk) {
1583                 write_lock_bh(&con->sock->sk->sk_callback_lock);
1584                 con->sock->sk->sk_user_data = NULL;
1585                 write_unlock_bh(&con->sock->sk->sk_callback_lock);
1586         }
1587         if (con->othercon && and_other)
1588                 _stop_conn(con->othercon, false);
1589         mutex_unlock(&con->sock_mutex);
1590 }
1591
1592 static void stop_conn(struct connection *con)
1593 {
1594         _stop_conn(con, true);
1595 }
1596
1597 static void shutdown_conn(struct connection *con)
1598 {
1599         if (con->shutdown_action)
1600                 con->shutdown_action(con);
1601 }
1602
1603 static void connection_release(struct rcu_head *rcu)
1604 {
1605         struct connection *con = container_of(rcu, struct connection, rcu);
1606
1607         kfree(con->rx_buf);
1608         kfree(con);
1609 }
1610
1611 static void free_conn(struct connection *con)
1612 {
1613         close_connection(con, true, true, true);
1614         spin_lock(&connections_lock);
1615         hlist_del_rcu(&con->list);
1616         spin_unlock(&connections_lock);
1617         if (con->othercon) {
1618                 clean_one_writequeue(con->othercon);
1619                 call_rcu(&con->othercon->rcu, connection_release);
1620         }
1621         clean_one_writequeue(con);
1622         call_rcu(&con->rcu, connection_release);
1623 }
1624
1625 static void work_flush(void)
1626 {
1627         int ok, idx;
1628         int i;
1629         struct connection *con;
1630
1631         do {
1632                 ok = 1;
1633                 foreach_conn(stop_conn);
1634                 if (recv_workqueue)
1635                         flush_workqueue(recv_workqueue);
1636                 if (send_workqueue)
1637                         flush_workqueue(send_workqueue);
1638                 idx = srcu_read_lock(&connections_srcu);
1639                 for (i = 0; i < CONN_HASH_SIZE && ok; i++) {
1640                         hlist_for_each_entry_rcu(con, &connection_hash[i],
1641                                                  list) {
1642                                 ok &= test_bit(CF_READ_PENDING, &con->flags);
1643                                 ok &= test_bit(CF_WRITE_PENDING, &con->flags);
1644                                 if (con->othercon) {
1645                                         ok &= test_bit(CF_READ_PENDING,
1646                                                        &con->othercon->flags);
1647                                         ok &= test_bit(CF_WRITE_PENDING,
1648                                                        &con->othercon->flags);
1649                                 }
1650                         }
1651                 }
1652                 srcu_read_unlock(&connections_srcu, idx);
1653         } while (!ok);
1654 }
1655
1656 void dlm_lowcomms_stop(void)
1657 {
1658         /* Set all the flags to prevent any
1659            socket activity.
1660         */
1661         dlm_allow_conn = 0;
1662
1663         if (recv_workqueue)
1664                 flush_workqueue(recv_workqueue);
1665         if (send_workqueue)
1666                 flush_workqueue(send_workqueue);
1667
1668         foreach_conn(shutdown_conn);
1669         work_flush();
1670         foreach_conn(free_conn);
1671         work_stop();
1672         deinit_local();
1673 }
1674
1675 int dlm_lowcomms_start(void)
1676 {
1677         int error = -EINVAL;
1678         struct connection *con;
1679         int i;
1680
1681         for (i = 0; i < CONN_HASH_SIZE; i++)
1682                 INIT_HLIST_HEAD(&connection_hash[i]);
1683
1684         init_local();
1685         if (!dlm_local_count) {
1686                 error = -ENOTCONN;
1687                 log_print("no local IP address has been set");
1688                 goto fail;
1689         }
1690
1691         error = work_start();
1692         if (error)
1693                 goto fail;
1694
1695         dlm_allow_conn = 1;
1696
1697         /* Start listening */
1698         if (dlm_config.ci_protocol == 0)
1699                 error = tcp_listen_for_all();
1700         else
1701                 error = sctp_listen_for_all();
1702         if (error)
1703                 goto fail_unlisten;
1704
1705         return 0;
1706
1707 fail_unlisten:
1708         dlm_allow_conn = 0;
1709         con = nodeid2con(0,0);
1710         if (con)
1711                 free_conn(con);
1712 fail:
1713         return error;
1714 }
1715
1716 void dlm_lowcomms_exit(void)
1717 {
1718         struct dlm_node_addr *na, *safe;
1719
1720         spin_lock(&dlm_node_addrs_spin);
1721         list_for_each_entry_safe(na, safe, &dlm_node_addrs, list) {
1722                 list_del(&na->list);
1723                 while (na->addr_count--)
1724                         kfree(na->addr[na->addr_count]);
1725                 kfree(na);
1726         }
1727         spin_unlock(&dlm_node_addrs_spin);
1728 }