fs/dlm/lowcomms.c

   1 /******************************************************************************
   2 *******************************************************************************
   3 **
   4 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
   5 **  Copyright (C) 2004-2009 Red Hat, Inc.  All rights reserved.
   6 **
   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
  24  * be expanded for the cluster infrastructure then that is its
  25  * 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  *
  39  * lowcomms will choose to use either TCP or SCTP as its transport layer
  40  * depending on the configuration variable 'protocol'. This should be set
  41  * to 0 (default) for TCP or 1 for SCTP. It should be configured using a
  42  * 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>
  51 #include <linux/file.h>
  52 #include <linux/mutex.h>
  53 #include <linux/sctp.h>
  54 #include <linux/slab.h>
  55 #include <net/sctp/sctp.h>
  56 #include <net/ipv6.h>
  57
  58 #include "dlm_internal.h"
  59 #include "lowcomms.h"
  60 #include "midcomms.h"
  61 #include "config.h"
  62
  63 #define NEEDED_RMEM (4*1024*1024)
  64 #define CONN_HASH_SIZE 32
  65
  66 /* Number of messages to send before rescheduling */
  67 #define MAX_SEND_MSG_COUNT 25
  68
  69 struct cbuf {
  70         unsigned int base;
  71         unsigned int len;
  72         unsigned int mask;
  73 };
  74
  75 static void cbuf_add(struct cbuf *cb, int n)
  76 {
  77         cb->len += n;
  78 }
  79
  80 static int cbuf_data(struct cbuf *cb)
  81 {
  82         return ((cb->base + cb->len) & cb->mask);
  83 }
  84
  85 static void cbuf_init(struct cbuf *cb, int size)
  86 {
  87         cb->base = cb->len = 0;
  88         cb->mask = size-1;
  89 }
  90
  91 static void cbuf_eat(struct cbuf *cb, int n)
  92 {
  93         cb->len  -= n;
  94         cb->base += n;
  95         cb->base &= cb->mask;
  96 }
  97
  98 static bool cbuf_empty(struct cbuf *cb)
  99 {
 100         return cb->len == 0;
 101 }
 102
 103 struct connection {
 104         struct socket *sock;    /* NULL if not connected */
 105         uint32_t nodeid;        /* So we know who we are in the list */
 106         struct mutex sock_mutex;
 107         unsigned long flags;
 108 #define CF_READ_PENDING 1
 109 #define CF_WRITE_PENDING 2
 110 #define CF_CONNECT_PENDING 3
 111 #define CF_INIT_PENDING 4
 112 #define CF_IS_OTHERCON 5
 113 #define CF_CLOSE 6
 114 #define CF_APP_LIMITED 7
 115         struct list_head writequeue;  /* List of outgoing writequeue_entries */
 116         spinlock_t writequeue_lock;
 117         int (*rx_action) (struct connection *); /* What to do when active */
 118         void (*connect_action) (struct connection *);   /* What to do to connect */
 119         struct page *rx_page;
 120         struct cbuf cb;
 121         int retries;
 122 #define MAX_CONNECT_RETRIES 3
 123         struct hlist_node list;
 124         struct connection *othercon;
 125         struct work_struct rwork; /* Receive workqueue */
 126         struct work_struct swork; /* Send workqueue */
 127         void (*orig_error_report)(struct sock *sk);
 128 };
 129 #define sock2con(x) ((struct connection *)(x)->sk_user_data)
 130
 131 /* An entry waiting to be sent */
 132 struct writequeue_entry {
 133         struct list_head list;
 134         struct page *page;
 135         int offset;
 136         int len;
 137         int end;
 138         int users;
 139         struct connection *con;
 140 };
 141
 142 struct dlm_node_addr {
 143         struct list_head list;
 144         int nodeid;
 145         int addr_count;
 146         int curr_addr_index;
 147         struct sockaddr_storage *addr[DLM_MAX_ADDR_COUNT];
 148 };
 149
 150 static LIST_HEAD(dlm_node_addrs);
 151 static DEFINE_SPINLOCK(dlm_node_addrs_spin);
 152
 153 static struct sockaddr_storage *dlm_local_addr[DLM_MAX_ADDR_COUNT];
 154 static int dlm_local_count;
 155 static int dlm_allow_conn;
 156
 157 /* Work queues */
 158 static struct workqueue_struct *recv_workqueue;
 159 static struct workqueue_struct *send_workqueue;
 160
 161 static struct hlist_head connection_hash[CONN_HASH_SIZE];
 162 static DEFINE_MUTEX(connections_lock);
 163 static struct kmem_cache *con_cache;
 164
 165 static void process_recv_sockets(struct work_struct *work);
 166 static void process_send_sockets(struct work_struct *work);
 167
 168
 169 /* This is deliberately very simple because most clusters have simple
 170    sequential nodeids, so we should be able to go straight to a connection
 171    struct in the array */
 172 static inline int nodeid_hash(int nodeid)
 173 {
 174         return nodeid & (CONN_HASH_SIZE-1);
 175 }
 176
 177 static struct connection *__find_con(int nodeid)
 178 {
 179         int r;
 180         struct connection *con;
 181
 182         r = nodeid_hash(nodeid);
 183
 184         hlist_for_each_entry(con, &connection_hash[r], list) {
 185                 if (con->nodeid == nodeid)
 186                         return con;
 187         }
 188         return NULL;
 189 }
 190
 191 /*
 192  * If 'allocation' is zero then we don't attempt to create a new
 193  * connection structure for this node.
 194  */
 195 static struct connection *__nodeid2con(int nodeid, gfp_t alloc)
 196 {
 197         struct connection *con = NULL;
 198         int r;
 199
 200         con = __find_con(nodeid);
 201         if (con || !alloc)
 202                 return con;
 203
 204         con = kmem_cache_zalloc(con_cache, alloc);
 205         if (!con)
 206                 return NULL;
 207
 208         r = nodeid_hash(nodeid);
 209         hlist_add_head(&con->list, &connection_hash[r]);
 210
 211         con->nodeid = nodeid;
 212         mutex_init(&con->sock_mutex);
 213         INIT_LIST_HEAD(&con->writequeue);
 214         spin_lock_init(&con->writequeue_lock);
 215         INIT_WORK(&con->swork, process_send_sockets);
 216         INIT_WORK(&con->rwork, process_recv_sockets);
 217
 218         /* Setup action pointers for child sockets */
 219         if (con->nodeid) {
 220                 struct connection *zerocon = __find_con(0);
 221
 222                 con->connect_action = zerocon->connect_action;
 223                 if (!con->rx_action)
 224                         con->rx_action = zerocon->rx_action;
 225         }
 226
 227         return con;
 228 }
 229
 230 /* Loop round all connections */
 231 static void foreach_conn(void (*conn_func)(struct connection *c))
 232 {
 233         int i;
 234         struct hlist_node *n;
 235         struct connection *con;
 236
 237         for (i = 0; i < CONN_HASH_SIZE; i++) {
 238                 hlist_for_each_entry_safe(con, n, &connection_hash[i], list)
 239                         conn_func(con);
 240         }
 241 }
 242
 243 static struct connection *nodeid2con(int nodeid, gfp_t allocation)
 244 {
 245         struct connection *con;
 246
 247         mutex_lock(&connections_lock);
 248         con = __nodeid2con(nodeid, allocation);
 249         mutex_unlock(&connections_lock);
 250
 251         return con;
 252 }
 253
 254 static struct dlm_node_addr *find_node_addr(int nodeid)
 255 {
 256         struct dlm_node_addr *na;
 257
 258         list_for_each_entry(na, &dlm_node_addrs, list) {
 259                 if (na->nodeid == nodeid)
 260                         return na;
 261         }
 262         return NULL;
 263 }
 264
 265 static int addr_compare(struct sockaddr_storage *x, struct sockaddr_storage *y)
 266 {
 267         switch (x->ss_family) {
 268         case AF_INET: {
 269                 struct sockaddr_in *sinx = (struct sockaddr_in *)x;
 270                 struct sockaddr_in *siny = (struct sockaddr_in *)y;
 271                 if (sinx->sin_addr.s_addr != siny->sin_addr.s_addr)
 272                         return 0;
 273                 if (sinx->sin_port != siny->sin_port)
 274                         return 0;
 275                 break;
 276         }
 277         case AF_INET6: {
 278                 struct sockaddr_in6 *sinx = (struct sockaddr_in6 *)x;
 279                 struct sockaddr_in6 *siny = (struct sockaddr_in6 *)y;
 280                 if (!ipv6_addr_equal(&sinx->sin6_addr, &siny->sin6_addr))
 281                         return 0;
 282                 if (sinx->sin6_port != siny->sin6_port)
 283                         return 0;
 284                 break;
 285         }
 286         default:
 287                 return 0;
 288         }
 289         return 1;
 290 }
 291
 292 static int nodeid_to_addr(int nodeid, struct sockaddr_storage *sas_out,
 293                           struct sockaddr *sa_out, bool try_new_addr)
 294 {
 295         struct sockaddr_storage sas;
 296         struct dlm_node_addr *na;
 297
 298         if (!dlm_local_count)
 299                 return -1;
 300
 301         spin_lock(&dlm_node_addrs_spin);
 302         na = find_node_addr(nodeid);
 303         if (na && na->addr_count) {
 304                 memcpy(&sas, na->addr[na->curr_addr_index],
 305                        sizeof(struct sockaddr_storage));
 306
 307                 if (try_new_addr) {
 308                         na->curr_addr_index++;
 309                         if (na->curr_addr_index == na->addr_count)
 310                                 na->curr_addr_index = 0;
 311                 }
 312         }
 313         spin_unlock(&dlm_node_addrs_spin);
 314
 315         if (!na)
 316                 return -EEXIST;
 317
 318         if (!na->addr_count)
 319                 return -ENOENT;
 320
 321         if (sas_out)
 322                 memcpy(sas_out, &sas, sizeof(struct sockaddr_storage));
 323
 324         if (!sa_out)
 325                 return 0;
 326
 327         if (dlm_local_addr[0]->ss_family == AF_INET) {
 328                 struct sockaddr_in *in4  = (struct sockaddr_in *) &sas;
 329                 struct sockaddr_in *ret4 = (struct sockaddr_in *) sa_out;
 330                 ret4->sin_addr.s_addr = in4->sin_addr.s_addr;
 331         } else {
 332                 struct sockaddr_in6 *in6  = (struct sockaddr_in6 *) &sas;
 333                 struct sockaddr_in6 *ret6 = (struct sockaddr_in6 *) sa_out;
 334                 ret6->sin6_addr = in6->sin6_addr;
 335         }
 336
 337         return 0;
 338 }
 339
 340 static int addr_to_nodeid(struct sockaddr_storage *addr, int *nodeid)
 341 {
 342         struct dlm_node_addr *na;
 343         int rv = -EEXIST;
 344         int addr_i;
 345
 346         spin_lock(&dlm_node_addrs_spin);
 347         list_for_each_entry(na, &dlm_node_addrs, list) {
 348                 if (!na->addr_count)
 349                         continue;
 350
 351                 for (addr_i = 0; addr_i < na->addr_count; addr_i++) {
 352                         if (addr_compare(na->addr[addr_i], addr)) {
 353                                 *nodeid = na->nodeid;
 354                                 rv = 0;
 355                                 goto unlock;
 356                         }
 357                 }
 358         }
 359 unlock:
 360         spin_unlock(&dlm_node_addrs_spin);
 361         return rv;
 362 }
 363
 364 int dlm_lowcomms_addr(int nodeid, struct sockaddr_storage *addr, int len)
 365 {
 366         struct sockaddr_storage *new_addr;
 367         struct dlm_node_addr *new_node, *na;
 368
 369         new_node = kzalloc(sizeof(struct dlm_node_addr), GFP_NOFS);
 370         if (!new_node)
 371                 return -ENOMEM;
 372
 373         new_addr = kzalloc(sizeof(struct sockaddr_storage), GFP_NOFS);
 374         if (!new_addr) {
 375                 kfree(new_node);
 376                 return -ENOMEM;
 377         }
 378
 379         memcpy(new_addr, addr, len);
 380
 381         spin_lock(&dlm_node_addrs_spin);
 382         na = find_node_addr(nodeid);
 383         if (!na) {
 384                 new_node->nodeid = nodeid;
 385                 new_node->addr[0] = new_addr;
 386                 new_node->addr_count = 1;
 387                 list_add(&new_node->list, &dlm_node_addrs);
 388                 spin_unlock(&dlm_node_addrs_spin);
 389                 return 0;
 390         }
 391
 392         if (na->addr_count >= DLM_MAX_ADDR_COUNT) {
 393                 spin_unlock(&dlm_node_addrs_spin);
 394                 kfree(new_addr);
 395                 kfree(new_node);
 396                 return -ENOSPC;
 397         }
 398
 399         na->addr[na->addr_count++] = new_addr;
 400         spin_unlock(&dlm_node_addrs_spin);
 401         kfree(new_node);
 402         return 0;
 403 }
 404
 405 /* Data available on socket or listen socket received a connect */
 406 static void lowcomms_data_ready(struct sock *sk)
 407 {
 408         struct connection *con = sock2con(sk);
 409         if (con && !test_and_set_bit(CF_READ_PENDING, &con->flags))
 410                 queue_work(recv_workqueue, &con->rwork);
 411 }
 412
 413 static void lowcomms_write_space(struct sock *sk)
 414 {
 415         struct connection *con = sock2con(sk);
 416
 417         if (!con)
 418                 return;
 419
 420         clear_bit(SOCK_NOSPACE, &con->sock->flags);
 421
 422         if (test_and_clear_bit(CF_APP_LIMITED, &con->flags)) {
 423                 con->sock->sk->sk_write_pending--;
 424                 clear_bit(SOCK_ASYNC_NOSPACE, &con->sock->flags);
 425         }
 426
 427         if (!test_and_set_bit(CF_WRITE_PENDING, &con->flags))
 428                 queue_work(send_workqueue, &con->swork);
 429 }
 430
 431 static inline void lowcomms_connect_sock(struct connection *con)
 432 {
 433         if (test_bit(CF_CLOSE, &con->flags))
 434                 return;
 435         if (!test_and_set_bit(CF_CONNECT_PENDING, &con->flags))
 436                 queue_work(send_workqueue, &con->swork);
 437 }
 438
 439 static void lowcomms_state_change(struct sock *sk)
 440 {
 441         /* SCTP layer is not calling sk_data_ready when the connection
 442          * is done, so we catch the signal through here. Also, it
 443          * doesn't switch socket state when entering shutdown, so we
 444          * skip the write in that case.
 445          */
 446         if (sk->sk_shutdown) {
 447                 if (sk->sk_shutdown == RCV_SHUTDOWN)
 448                         lowcomms_data_ready(sk);
 449         } else if (sk->sk_state == TCP_ESTABLISHED) {
 450                 lowcomms_write_space(sk);
 451         }
 452 }
 453
 454 int dlm_lowcomms_connect_node(int nodeid)
 455 {
 456         struct connection *con;
 457
 458         if (nodeid == dlm_our_nodeid())
 459                 return 0;
 460
 461         con = nodeid2con(nodeid, GFP_NOFS);
 462         if (!con)
 463                 return -ENOMEM;
 464         lowcomms_connect_sock(con);
 465         return 0;
 466 }
 467
 468 static void lowcomms_error_report(struct sock *sk)
 469 {
 470         struct connection *con = sock2con(sk);
 471         struct sockaddr_storage saddr;
 472
 473         if (nodeid_to_addr(con->nodeid, &saddr, NULL, false)) {
 474                 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
 475                                    "sending to node %d, port %d, "
 476                                    "sk_err=%d/%d\n", dlm_our_nodeid(),
 477                                    con->nodeid, dlm_config.ci_tcp_port,
 478                                    sk->sk_err, sk->sk_err_soft);
 479                 return;
 480         } else if (saddr.ss_family == AF_INET) {
 481                 struct sockaddr_in *sin4 = (struct sockaddr_in *)&saddr;
 482
 483                 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
 484                                    "sending to node %d at %pI4, port %d, "
 485                                    "sk_err=%d/%d\n", dlm_our_nodeid(),
 486                                    con->nodeid, &sin4->sin_addr.s_addr,
 487                                    dlm_config.ci_tcp_port, sk->sk_err,
 488                                    sk->sk_err_soft);
 489         } else {
 490                 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&saddr;
 491
 492                 printk_ratelimited(KERN_ERR "dlm: node %d: socket error "
 493                                    "sending to node %d at %u.%u.%u.%u, "
 494                                    "port %d, sk_err=%d/%d\n", dlm_our_nodeid(),
 495                                    con->nodeid, sin6->sin6_addr.s6_addr32[0],
 496                                    sin6->sin6_addr.s6_addr32[1],
 497                                    sin6->sin6_addr.s6_addr32[2],
 498                                    sin6->sin6_addr.s6_addr32[3],
 499                                    dlm_config.ci_tcp_port, sk->sk_err,
 500                                    sk->sk_err_soft);
 501         }
 502         con->orig_error_report(sk);
 503 }
 504
 505 /* Make a socket active */
 506 static void add_sock(struct socket *sock, struct connection *con)
 507 {
 508         con->sock = sock;
 509
 510         /* Install a data_ready callback */
 511         con->sock->sk->sk_data_ready = lowcomms_data_ready;
 512         con->sock->sk->sk_write_space = lowcomms_write_space;
 513         con->sock->sk->sk_state_change = lowcomms_state_change;
 514         con->sock->sk->sk_user_data = con;
 515         con->sock->sk->sk_allocation = GFP_NOFS;
 516         con->orig_error_report = con->sock->sk->sk_error_report;
 517         con->sock->sk->sk_error_report = lowcomms_error_report;
 518 }
 519
 520 /* Add the port number to an IPv6 or 4 sockaddr and return the address
 521    length */
 522 static void make_sockaddr(struct sockaddr_storage *saddr, uint16_t port,
 523                           int *addr_len)
 524 {
 525         saddr->ss_family =  dlm_local_addr[0]->ss_family;
 526         if (saddr->ss_family == AF_INET) {
 527                 struct sockaddr_in *in4_addr = (struct sockaddr_in *)saddr;
 528                 in4_addr->sin_port = cpu_to_be16(port);
 529                 *addr_len = sizeof(struct sockaddr_in);
 530                 memset(&in4_addr->sin_zero, 0, sizeof(in4_addr->sin_zero));
 531         } else {
 532                 struct sockaddr_in6 *in6_addr = (struct sockaddr_in6 *)saddr;
 533                 in6_addr->sin6_port = cpu_to_be16(port);
 534                 *addr_len = sizeof(struct sockaddr_in6);
 535         }
 536         memset((char *)saddr + *addr_len, 0, sizeof(struct sockaddr_storage) - *addr_len);
 537 }
 538
 539 /* Close a remote connection and tidy up */
 540 static void close_connection(struct connection *con, bool and_other,
 541                              bool tx, bool rx)
 542 {
 543         clear_bit(CF_CONNECT_PENDING, &con->flags);
 544         clear_bit(CF_WRITE_PENDING, &con->flags);
 545         if (tx && cancel_work_sync(&con->swork))
 546                 log_print("canceled swork for node %d", con->nodeid);
 547         if (rx && cancel_work_sync(&con->rwork))
 548                 log_print("canceled rwork for node %d", con->nodeid);
 549
 550         mutex_lock(&con->sock_mutex);
 551         if (con->sock) {
 552                 sock_release(con->sock);
 553                 con->sock = NULL;
 554         }
 555         if (con->othercon && and_other) {
 556                 /* Will only re-enter once. */
 557                 close_connection(con->othercon, false, true, true);
 558         }
 559         if (con->rx_page) {
 560                 __free_page(con->rx_page);
 561                 con->rx_page = NULL;
 562         }
 563
 564         con->retries = 0;
 565         mutex_unlock(&con->sock_mutex);
 566 }
 567
 568 /* Data received from remote end */
 569 static int receive_from_sock(struct connection *con)
 570 {
 571         int ret = 0;
 572         struct msghdr msg = {};
 573         struct kvec iov[2];
 574         unsigned len;
 575         int r;
 576         int call_again_soon = 0;
 577         int nvec;
 578
 579         mutex_lock(&con->sock_mutex);
 580
 581         if (con->sock == NULL) {
 582                 ret = -EAGAIN;
 583                 goto out_close;
 584         }
 585         if (con->nodeid == 0) {
 586                 ret = -EINVAL;
 587                 goto out_close;
 588         }
 589
 590         if (con->rx_page == NULL) {
 591                 /*
 592                  * This doesn't need to be atomic, but I think it should
 593                  * improve performance if it is.
 594                  */
 595                 con->rx_page = alloc_page(GFP_ATOMIC);
 596                 if (con->rx_page == NULL)
 597                         goto out_resched;
 598                 cbuf_init(&con->cb, PAGE_CACHE_SIZE);
 599         }
 600
 601         /*
 602          * iov[0] is the bit of the circular buffer between the current end
 603          * point (cb.base + cb.len) and the end of the buffer.
 604          */
 605         iov[0].iov_len = con->cb.base - cbuf_data(&con->cb);
 606         iov[0].iov_base = page_address(con->rx_page) + cbuf_data(&con->cb);
 607         iov[1].iov_len = 0;
 608         nvec = 1;
 609
 610         /*
 611          * iov[1] is the bit of the circular buffer between the start of the
 612          * buffer and the start of the currently used section (cb.base)
 613          */
 614         if (cbuf_data(&con->cb) >= con->cb.base) {
 615                 iov[0].iov_len = PAGE_CACHE_SIZE - cbuf_data(&con->cb);
 616                 iov[1].iov_len = con->cb.base;
 617                 iov[1].iov_base = page_address(con->rx_page);
 618                 nvec = 2;
 619         }
 620         len = iov[0].iov_len + iov[1].iov_len;
 621
 622         r = ret = kernel_recvmsg(con->sock, &msg, iov, nvec, len,
 623                                MSG_DONTWAIT | MSG_NOSIGNAL);
 624         if (ret <= 0)
 625                 goto out_close;
 626         else if (ret == len)
 627                 call_again_soon = 1;
 628
 629         cbuf_add(&con->cb, ret);
 630         ret = dlm_process_incoming_buffer(con->nodeid,
 631                                           page_address(con->rx_page),
 632                                           con->cb.base, con->cb.len,
 633                                           PAGE_CACHE_SIZE);
 634         if (ret == -EBADMSG) {
 635                 log_print("lowcomms: addr=%p, base=%u, len=%u, read=%d",
 636                           page_address(con->rx_page), con->cb.base,
 637                           con->cb.len, r);
 638         }
 639         if (ret < 0)
 640                 goto out_close;
 641         cbuf_eat(&con->cb, ret);
 642
 643         if (cbuf_empty(&con->cb) && !call_again_soon) {
 644                 __free_page(con->rx_page);
 645                 con->rx_page = NULL;
 646         }
 647
 648         if (call_again_soon)
 649                 goto out_resched;
 650         mutex_unlock(&con->sock_mutex);
 651         return 0;
 652
 653 out_resched:
 654         if (!test_and_set_bit(CF_READ_PENDING, &con->flags))
 655                 queue_work(recv_workqueue, &con->rwork);
 656         mutex_unlock(&con->sock_mutex);
 657         return -EAGAIN;
 658
 659 out_close:
 660         mutex_unlock(&con->sock_mutex);
 661         if (ret != -EAGAIN) {
 662                 close_connection(con, false, true, false);
 663                 /* Reconnect when there is something to send */
 664         }
 665         /* Don't return success if we really got EOF */
 666         if (ret == 0)
 667                 ret = -EAGAIN;
 668
 669         return ret;
 670 }
 671
 672 /* Listening socket is busy, accept a connection */
 673 static int tcp_accept_from_sock(struct connection *con)
 674 {
 675         int result;
 676         struct sockaddr_storage peeraddr;
 677         struct socket *newsock;
 678         int len;
 679         int nodeid;
 680         struct connection *newcon;
 681         struct connection *addcon;
 682
 683         mutex_lock(&connections_lock);
 684         if (!dlm_allow_conn) {
 685                 mutex_unlock(&connections_lock);
 686                 return -1;
 687         }
 688         mutex_unlock(&connections_lock);
 689
 690         memset(&peeraddr, 0, sizeof(peeraddr));
 691         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
 692                                   SOCK_STREAM, IPPROTO_TCP, &newsock);
 693         if (result < 0)
 694                 return -ENOMEM;
 695
 696         mutex_lock_nested(&con->sock_mutex, 0);
 697
 698         result = -ENOTCONN;
 699         if (con->sock == NULL)
 700                 goto accept_err;
 701
 702         newsock->type = con->sock->type;
 703         newsock->ops = con->sock->ops;
 704
 705         result = con->sock->ops->accept(con->sock, newsock, O_NONBLOCK);
 706         if (result < 0)
 707                 goto accept_err;
 708
 709         /* Get the connected socket's peer */
 710         memset(&peeraddr, 0, sizeof(peeraddr));
 711         if (newsock->ops->getname(newsock, (struct sockaddr *)&peeraddr,
 712                                   &len, 2)) {
 713                 result = -ECONNABORTED;
 714                 goto accept_err;
 715         }
 716
 717         /* Get the new node's NODEID */
 718         make_sockaddr(&peeraddr, 0, &len);
 719         if (addr_to_nodeid(&peeraddr, &nodeid)) {
 720                 unsigned char *b=(unsigned char *)&peeraddr;
 721                 log_print("connect from non cluster node");
 722                 print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
 723                                      b, sizeof(struct sockaddr_storage));
 724                 sock_release(newsock);
 725                 mutex_unlock(&con->sock_mutex);
 726                 return -1;
 727         }
 728
 729         log_print("got connection from %d", nodeid);
 730
 731         /*  Check to see if we already have a connection to this node. This
 732          *  could happen if the two nodes initiate a connection at roughly
 733          *  the same time and the connections cross on the wire.
 734          *  In this case we store the incoming one in "othercon"
 735          */
 736         newcon = nodeid2con(nodeid, GFP_NOFS);
 737         if (!newcon) {
 738                 result = -ENOMEM;
 739                 goto accept_err;
 740         }
 741         mutex_lock_nested(&newcon->sock_mutex, 1);
 742         if (newcon->sock) {
 743                 struct connection *othercon = newcon->othercon;
 744
 745                 if (!othercon) {
 746                         othercon = kmem_cache_zalloc(con_cache, GFP_NOFS);
 747                         if (!othercon) {
 748                                 log_print("failed to allocate incoming socket");
 749                                 mutex_unlock(&newcon->sock_mutex);
 750                                 result = -ENOMEM;
 751                                 goto accept_err;
 752                         }
 753                         othercon->nodeid = nodeid;
 754                         othercon->rx_action = receive_from_sock;
 755                         mutex_init(&othercon->sock_mutex);
 756                         INIT_WORK(&othercon->swork, process_send_sockets);
 757                         INIT_WORK(&othercon->rwork, process_recv_sockets);
 758                         set_bit(CF_IS_OTHERCON, &othercon->flags);
 759                 }
 760                 if (!othercon->sock) {
 761                         newcon->othercon = othercon;
 762                         othercon->sock = newsock;
 763                         newsock->sk->sk_user_data = othercon;
 764                         add_sock(newsock, othercon);
 765                         addcon = othercon;
 766                 }
 767                 else {
 768                         printk("Extra connection from node %d attempted\n", nodeid);
 769                         result = -EAGAIN;
 770                         mutex_unlock(&newcon->sock_mutex);
 771                         goto accept_err;
 772                 }
 773         }
 774         else {
 775                 newsock->sk->sk_user_data = newcon;
 776                 newcon->rx_action = receive_from_sock;
 777                 add_sock(newsock, newcon);
 778                 addcon = newcon;
 779         }
 780
 781         mutex_unlock(&newcon->sock_mutex);
 782
 783         /*
 784          * Add it to the active queue in case we got data
 785          * between processing the accept adding the socket
 786          * to the read_sockets list
 787          */
 788         if (!test_and_set_bit(CF_READ_PENDING, &addcon->flags))
 789                 queue_work(recv_workqueue, &addcon->rwork);
 790         mutex_unlock(&con->sock_mutex);
 791
 792         return 0;
 793
 794 accept_err:
 795         mutex_unlock(&con->sock_mutex);
 796         sock_release(newsock);
 797
 798         if (result != -EAGAIN)
 799                 log_print("error accepting connection from node: %d", result);
 800         return result;
 801 }
 802
 803 static int sctp_accept_from_sock(struct connection *con)
 804 {
 805         /* Check that the new node is in the lockspace */
 806         struct sctp_prim prim;
 807         int nodeid;
 808         int prim_len, ret;
 809         int addr_len;
 810         struct connection *newcon;
 811         struct connection *addcon;
 812         struct socket *newsock;
 813
 814         mutex_lock(&connections_lock);
 815         if (!dlm_allow_conn) {
 816                 mutex_unlock(&connections_lock);
 817                 return -1;
 818         }
 819         mutex_unlock(&connections_lock);
 820
 821         mutex_lock_nested(&con->sock_mutex, 0);
 822
 823         ret = kernel_accept(con->sock, &newsock, O_NONBLOCK);
 824         if (ret < 0)
 825                 goto accept_err;
 826
 827         memset(&prim, 0, sizeof(struct sctp_prim));
 828         prim_len = sizeof(struct sctp_prim);
 829
 830         ret = kernel_getsockopt(newsock, IPPROTO_SCTP, SCTP_PRIMARY_ADDR,
 831                                 (char *)&prim, &prim_len);
 832         if (ret < 0) {
 833                 log_print("getsockopt/sctp_primary_addr failed: %d", ret);
 834                 goto accept_err;
 835         }
 836
 837         make_sockaddr(&prim.ssp_addr, 0, &addr_len);
 838         if (addr_to_nodeid(&prim.ssp_addr, &nodeid)) {
 839                 unsigned char *b = (unsigned char *)&prim.ssp_addr;
 840
 841                 log_print("reject connect from unknown addr");
 842                 print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
 843                                      b, sizeof(struct sockaddr_storage));
 844                 goto accept_err;
 845         }
 846
 847         newcon = nodeid2con(nodeid, GFP_NOFS);
 848         if (!newcon) {
 849                 ret = -ENOMEM;
 850                 goto accept_err;
 851         }
 852
 853         mutex_lock_nested(&newcon->sock_mutex, 1);
 854
 855         if (newcon->sock) {
 856                 struct connection *othercon = newcon->othercon;
 857
 858                 if (!othercon) {
 859                         othercon = kmem_cache_zalloc(con_cache, GFP_NOFS);
 860                         if (!othercon) {
 861                                 log_print("failed to allocate incoming socket");
 862                                 mutex_unlock(&newcon->sock_mutex);
 863                                 ret = -ENOMEM;
 864                                 goto accept_err;
 865                         }
 866                         othercon->nodeid = nodeid;
 867                         othercon->rx_action = receive_from_sock;
 868                         mutex_init(&othercon->sock_mutex);
 869                         INIT_WORK(&othercon->swork, process_send_sockets);
 870                         INIT_WORK(&othercon->rwork, process_recv_sockets);
 871                         set_bit(CF_IS_OTHERCON, &othercon->flags);
 872                 }
 873                 if (!othercon->sock) {
 874                         newcon->othercon = othercon;
 875                         othercon->sock = newsock;
 876                         newsock->sk->sk_user_data = othercon;
 877                         add_sock(newsock, othercon);
 878                         addcon = othercon;
 879                 } else {
 880                         printk("Extra connection from node %d attempted\n", nodeid);
 881                         ret = -EAGAIN;
 882                         mutex_unlock(&newcon->sock_mutex);
 883                         goto accept_err;
 884                 }
 885         } else {
 886                 newsock->sk->sk_user_data = newcon;
 887                 newcon->rx_action = receive_from_sock;
 888                 add_sock(newsock, newcon);
 889                 addcon = newcon;
 890         }
 891
 892         log_print("connected to %d", nodeid);
 893
 894         mutex_unlock(&newcon->sock_mutex);
 895
 896         /*
 897          * Add it to the active queue in case we got data
 898          * between processing the accept adding the socket
 899          * to the read_sockets list
 900          */
 901         if (!test_and_set_bit(CF_READ_PENDING, &addcon->flags))
 902                 queue_work(recv_workqueue, &addcon->rwork);
 903         mutex_unlock(&con->sock_mutex);
 904
 905         return 0;
 906
 907 accept_err:
 908         mutex_unlock(&con->sock_mutex);
 909         if (newsock)
 910                 sock_release(newsock);
 911         if (ret != -EAGAIN)
 912                 log_print("error accepting connection from node: %d", ret);
 913
 914         return ret;
 915 }
 916
 917 static void free_entry(struct writequeue_entry *e)
 918 {
 919         __free_page(e->page);
 920         kfree(e);
 921 }
 922
 923 /*
 924  * writequeue_entry_complete - try to delete and free write queue entry
 925  * @e: write queue entry to try to delete
 926  * @completed: bytes completed
 927  *
 928  * writequeue_lock must be held.
 929  */
 930 static void writequeue_entry_complete(struct writequeue_entry *e, int completed)
 931 {
 932         e->offset += completed;
 933         e->len -= completed;
 934
 935         if (e->len == 0 && e->users == 0) {
 936                 list_del(&e->list);
 937                 free_entry(e);
 938         }
 939 }
 940
 941 /*
 942  * sctp_bind_addrs - bind a SCTP socket to all our addresses
 943  */
 944 static int sctp_bind_addrs(struct connection *con, uint16_t port)
 945 {
 946         struct sockaddr_storage localaddr;
 947         int i, addr_len, result = 0;
 948
 949         for (i = 0; i < dlm_local_count; i++) {
 950                 memcpy(&localaddr, dlm_local_addr[i], sizeof(localaddr));
 951                 make_sockaddr(&localaddr, port, &addr_len);
 952
 953                 if (!i)
 954                         result = kernel_bind(con->sock,
 955                                              (struct sockaddr *)&localaddr,
 956                                              addr_len);
 957                 else
 958                         result = kernel_setsockopt(con->sock, SOL_SCTP,
 959                                                    SCTP_SOCKOPT_BINDX_ADD,
 960                                                    (char *)&localaddr, addr_len);
 961
 962                 if (result < 0) {
 963                         log_print("Can't bind to %d addr number %d, %d.\n",
 964                                   port, i + 1, result);
 965                         break;
 966                 }
 967         }
 968         return result;
 969 }
 970
 971 /* Initiate an SCTP association.
 972    This is a special case of send_to_sock() in that we don't yet have a
 973    peeled-off socket for this association, so we use the listening socket
 974    and add the primary IP address of the remote node.
 975  */
 976 static void sctp_connect_to_sock(struct connection *con)
 977 {
 978         struct sockaddr_storage daddr;
 979         int one = 1;
 980         int result;
 981         int addr_len;
 982         struct socket *sock;
 983
 984         if (con->nodeid == 0) {
 985                 log_print("attempt to connect sock 0 foiled");
 986                 return;
 987         }
 988
 989         mutex_lock(&con->sock_mutex);
 990
 991         /* Some odd races can cause double-connects, ignore them */
 992         if (con->retries++ > MAX_CONNECT_RETRIES)
 993                 goto out;
 994
 995         if (con->sock) {
 996                 log_print("node %d already connected.", con->nodeid);
 997                 goto out;
 998         }
 999
1000         memset(&daddr, 0, sizeof(daddr));
1001         result = nodeid_to_addr(con->nodeid, &daddr, NULL, true);
1002         if (result < 0) {
1003                 log_print("no address for nodeid %d", con->nodeid);
1004                 goto out;
1005         }
1006
1007         /* Create a socket to communicate with */
1008         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1009                                   SOCK_STREAM, IPPROTO_SCTP, &sock);
1010         if (result < 0)
1011                 goto socket_err;
1012
1013         sock->sk->sk_user_data = con;
1014         con->rx_action = receive_from_sock;
1015         con->connect_action = sctp_connect_to_sock;
1016         add_sock(sock, con);
1017
1018         /* Bind to all addresses. */
1019         if (sctp_bind_addrs(con, 0))
1020                 goto bind_err;
1021
1022         make_sockaddr(&daddr, dlm_config.ci_tcp_port, &addr_len);
1023
1024         log_print("connecting to %d", con->nodeid);
1025
1026         /* Turn off Nagle's algorithm */
1027         kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
1028                           sizeof(one));
1029
1030         result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
1031                                    O_NONBLOCK);
1032         if (result == -EINPROGRESS)
1033                 result = 0;
1034         if (result == 0)
1035                 goto out;
1036
1037
1038 bind_err:
1039         con->sock = NULL;
1040         sock_release(sock);
1041
1042 socket_err:
1043         /*
1044          * Some errors are fatal and this list might need adjusting. For other
1045          * errors we try again until the max number of retries is reached.
1046          */
1047         if (result != -EHOSTUNREACH &&
1048             result != -ENETUNREACH &&
1049             result != -ENETDOWN &&
1050             result != -EINVAL &&
1051             result != -EPROTONOSUPPORT) {
1052                 log_print("connect %d try %d error %d", con->nodeid,
1053                           con->retries, result);
1054                 mutex_unlock(&con->sock_mutex);
1055                 msleep(1000);
1056                 clear_bit(CF_CONNECT_PENDING, &con->flags);
1057                 lowcomms_connect_sock(con);
1058                 return;
1059         }
1060
1061 out:
1062         mutex_unlock(&con->sock_mutex);
1063         set_bit(CF_WRITE_PENDING, &con->flags);
1064 }
1065
1066 /* Connect a new socket to its peer */
1067 static void tcp_connect_to_sock(struct connection *con)
1068 {
1069         struct sockaddr_storage saddr, src_addr;
1070         int addr_len;
1071         struct socket *sock = NULL;
1072         int one = 1;
1073         int result;
1074
1075         if (con->nodeid == 0) {
1076                 log_print("attempt to connect sock 0 foiled");
1077                 return;
1078         }
1079
1080         mutex_lock(&con->sock_mutex);
1081         if (con->retries++ > MAX_CONNECT_RETRIES)
1082                 goto out;
1083
1084         /* Some odd races can cause double-connects, ignore them */
1085         if (con->sock)
1086                 goto out;
1087
1088         /* Create a socket to communicate with */
1089         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1090                                   SOCK_STREAM, IPPROTO_TCP, &sock);
1091         if (result < 0)
1092                 goto out_err;
1093
1094         memset(&saddr, 0, sizeof(saddr));
1095         result = nodeid_to_addr(con->nodeid, &saddr, NULL, false);
1096         if (result < 0) {
1097                 log_print("no address for nodeid %d", con->nodeid);
1098                 goto out_err;
1099         }
1100
1101         sock->sk->sk_user_data = con;
1102         con->rx_action = receive_from_sock;
1103         con->connect_action = tcp_connect_to_sock;
1104         add_sock(sock, con);
1105
1106         /* Bind to our cluster-known address connecting to avoid
1107            routing problems */
1108         memcpy(&src_addr, dlm_local_addr[0], sizeof(src_addr));
1109         make_sockaddr(&src_addr, 0, &addr_len);
1110         result = sock->ops->bind(sock, (struct sockaddr *) &src_addr,
1111                                  addr_len);
1112         if (result < 0) {
1113                 log_print("could not bind for connect: %d", result);
1114                 /* This *may* not indicate a critical error */
1115         }
1116
1117         make_sockaddr(&saddr, dlm_config.ci_tcp_port, &addr_len);
1118
1119         log_print("connecting to %d", con->nodeid);
1120
1121         /* Turn off Nagle's algorithm */
1122         kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
1123                           sizeof(one));
1124
1125         result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
1126                                    O_NONBLOCK);
1127         if (result == -EINPROGRESS)
1128                 result = 0;
1129         if (result == 0)
1130                 goto out;
1131
1132 out_err:
1133         if (con->sock) {
1134                 sock_release(con->sock);
1135                 con->sock = NULL;
1136         } else if (sock) {
1137                 sock_release(sock);
1138         }
1139         /*
1140          * Some errors are fatal and this list might need adjusting. For other
1141          * errors we try again until the max number of retries is reached.
1142          */
1143         if (result != -EHOSTUNREACH &&
1144             result != -ENETUNREACH &&
1145             result != -ENETDOWN &&
1146             result != -EINVAL &&
1147             result != -EPROTONOSUPPORT) {
1148                 log_print("connect %d try %d error %d", con->nodeid,
1149                           con->retries, result);
1150                 mutex_unlock(&con->sock_mutex);
1151                 msleep(1000);
1152                 clear_bit(CF_CONNECT_PENDING, &con->flags);
1153                 lowcomms_connect_sock(con);
1154                 return;
1155         }
1156 out:
1157         mutex_unlock(&con->sock_mutex);
1158         set_bit(CF_WRITE_PENDING, &con->flags);
1159         return;
1160 }
1161
1162 static struct socket *tcp_create_listen_sock(struct connection *con,
1163                                              struct sockaddr_storage *saddr)
1164 {
1165         struct socket *sock = NULL;
1166         int result = 0;
1167         int one = 1;
1168         int addr_len;
1169
1170         if (dlm_local_addr[0]->ss_family == AF_INET)
1171                 addr_len = sizeof(struct sockaddr_in);
1172         else
1173                 addr_len = sizeof(struct sockaddr_in6);
1174
1175         /* Create a socket to communicate with */
1176         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1177                                   SOCK_STREAM, IPPROTO_TCP, &sock);
1178         if (result < 0) {
1179                 log_print("Can't create listening comms socket");
1180                 goto create_out;
1181         }
1182
1183         /* Turn off Nagle's algorithm */
1184         kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
1185                           sizeof(one));
1186
1187         result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
1188                                    (char *)&one, sizeof(one));
1189
1190         if (result < 0) {
1191                 log_print("Failed to set SO_REUSEADDR on socket: %d", result);
1192         }
1193         con->rx_action = tcp_accept_from_sock;
1194         con->connect_action = tcp_connect_to_sock;
1195
1196         /* Bind to our port */
1197         make_sockaddr(saddr, dlm_config.ci_tcp_port, &addr_len);
1198         result = sock->ops->bind(sock, (struct sockaddr *) saddr, addr_len);
1199         if (result < 0) {
1200                 log_print("Can't bind to port %d", dlm_config.ci_tcp_port);
1201                 sock_release(sock);
1202                 sock = NULL;
1203                 con->sock = NULL;
1204                 goto create_out;
1205         }
1206         result = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
1207                                  (char *)&one, sizeof(one));
1208         if (result < 0) {
1209                 log_print("Set keepalive failed: %d", result);
1210         }
1211
1212         result = sock->ops->listen(sock, 5);
1213         if (result < 0) {
1214                 log_print("Can't listen on port %d", dlm_config.ci_tcp_port);
1215                 sock_release(sock);
1216                 sock = NULL;
1217                 goto create_out;
1218         }
1219
1220 create_out:
1221         return sock;
1222 }
1223
1224 /* Get local addresses */
1225 static void init_local(void)
1226 {
1227         struct sockaddr_storage sas, *addr;
1228         int i;
1229
1230         dlm_local_count = 0;
1231         for (i = 0; i < DLM_MAX_ADDR_COUNT; i++) {
1232                 if (dlm_our_addr(&sas, i))
1233                         break;
1234
1235                 addr = kmalloc(sizeof(*addr), GFP_NOFS);
1236                 if (!addr)
1237                         break;
1238                 memcpy(addr, &sas, sizeof(*addr));
1239                 dlm_local_addr[dlm_local_count++] = addr;
1240         }
1241 }
1242
1243 /* Initialise SCTP socket and bind to all interfaces */
1244 static int sctp_listen_for_all(void)
1245 {
1246         struct socket *sock = NULL;
1247         int result = -EINVAL;
1248         struct connection *con = nodeid2con(0, GFP_NOFS);
1249         int bufsize = NEEDED_RMEM;
1250         int one = 1;
1251
1252         if (!con)
1253                 return -ENOMEM;
1254
1255         log_print("Using SCTP for communications");
1256
1257         result = sock_create_kern(&init_net, dlm_local_addr[0]->ss_family,
1258                                   SOCK_STREAM, IPPROTO_SCTP, &sock);
1259         if (result < 0) {
1260                 log_print("Can't create comms socket, check SCTP is loaded");
1261                 goto out;
1262         }
1263
1264         result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
1265                                  (char *)&bufsize, sizeof(bufsize));
1266         if (result)
1267                 log_print("Error increasing buffer space on socket %d", result);
1268
1269         result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
1270                                    sizeof(one));
1271         if (result < 0)
1272                 log_print("Could not set SCTP NODELAY error %d\n", result);
1273
1274         /* Init con struct */
1275         sock->sk->sk_user_data = con;
1276         con->sock = sock;
1277         con->sock->sk->sk_data_ready = lowcomms_data_ready;
1278         con->rx_action = sctp_accept_from_sock;
1279         con->connect_action = sctp_connect_to_sock;
1280
1281         /* Bind to all addresses. */
1282         if (sctp_bind_addrs(con, dlm_config.ci_tcp_port))
1283                 goto create_delsock;
1284
1285         result = sock->ops->listen(sock, 5);
1286         if (result < 0) {
1287                 log_print("Can't set socket listening");
1288                 goto create_delsock;
1289         }
1290
1291         return 0;
1292
1293 create_delsock:
1294         sock_release(sock);
1295         con->sock = NULL;
1296 out:
1297         return result;
1298 }
1299
1300 static int tcp_listen_for_all(void)
1301 {
1302         struct socket *sock = NULL;
1303         struct connection *con = nodeid2con(0, GFP_NOFS);
1304         int result = -EINVAL;
1305
1306         if (!con)
1307                 return -ENOMEM;
1308
1309         /* We don't support multi-homed hosts */
1310         if (dlm_local_addr[1] != NULL) {
1311                 log_print("TCP protocol can't handle multi-homed hosts, "
1312                           "try SCTP");
1313                 return -EINVAL;
1314         }
1315
1316         log_print("Using TCP for communications");
1317
1318         sock = tcp_create_listen_sock(con, dlm_local_addr[0]);
1319         if (sock) {
1320                 add_sock(sock, con);
1321                 result = 0;
1322         }
1323         else {
1324                 result = -EADDRINUSE;
1325         }
1326
1327         return result;
1328 }
1329
1330
1331
1332 static struct writequeue_entry *new_writequeue_entry(struct connection *con,
1333                                                      gfp_t allocation)
1334 {
1335         struct writequeue_entry *entry;
1336
1337         entry = kmalloc(sizeof(struct writequeue_entry), allocation);
1338         if (!entry)
1339                 return NULL;
1340
1341         entry->page = alloc_page(allocation);
1342         if (!entry->page) {
1343                 kfree(entry);
1344                 return NULL;
1345         }
1346
1347         entry->offset = 0;
1348         entry->len = 0;
1349         entry->end = 0;
1350         entry->users = 0;
1351         entry->con = con;
1352
1353         return entry;
1354 }
1355
1356 void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
1357 {
1358         struct connection *con;
1359         struct writequeue_entry *e;
1360         int offset = 0;
1361
1362         con = nodeid2con(nodeid, allocation);
1363         if (!con)
1364                 return NULL;
1365
1366         spin_lock(&con->writequeue_lock);
1367         e = list_entry(con->writequeue.prev, struct writequeue_entry, list);
1368         if ((&e->list == &con->writequeue) ||
1369             (PAGE_CACHE_SIZE - e->end < len)) {
1370                 e = NULL;
1371         } else {
1372                 offset = e->end;
1373                 e->end += len;
1374                 e->users++;
1375         }
1376         spin_unlock(&con->writequeue_lock);
1377
1378         if (e) {
1379         got_one:
1380                 *ppc = page_address(e->page) + offset;
1381                 return e;
1382         }
1383
1384         e = new_writequeue_entry(con, allocation);
1385         if (e) {
1386                 spin_lock(&con->writequeue_lock);
1387                 offset = e->end;
1388                 e->end += len;
1389                 e->users++;
1390                 list_add_tail(&e->list, &con->writequeue);
1391                 spin_unlock(&con->writequeue_lock);
1392                 goto got_one;
1393         }
1394         return NULL;
1395 }
1396
1397 void dlm_lowcomms_commit_buffer(void *mh)
1398 {
1399         struct writequeue_entry *e = (struct writequeue_entry *)mh;
1400         struct connection *con = e->con;
1401         int users;
1402
1403         spin_lock(&con->writequeue_lock);
1404         users = --e->users;
1405         if (users)
1406                 goto out;
1407         e->len = e->end - e->offset;
1408         spin_unlock(&con->writequeue_lock);
1409
1410         if (!test_and_set_bit(CF_WRITE_PENDING, &con->flags)) {
1411                 queue_work(send_workqueue, &con->swork);
1412         }
1413         return;
1414
1415 out:
1416         spin_unlock(&con->writequeue_lock);
1417         return;
1418 }
1419
1420 /* Send a message */
1421 static void send_to_sock(struct connection *con)
1422 {
1423         int ret = 0;
1424         const int msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
1425         struct writequeue_entry *e;
1426         int len, offset;
1427         int count = 0;
1428
1429         mutex_lock(&con->sock_mutex);
1430         if (con->sock == NULL)
1431                 goto out_connect;
1432
1433         spin_lock(&con->writequeue_lock);
1434         for (;;) {
1435                 e = list_entry(con->writequeue.next, struct writequeue_entry,
1436                                list);
1437                 if ((struct list_head *) e == &con->writequeue)
1438                         break;
1439
1440                 len = e->len;
1441                 offset = e->offset;
1442                 BUG_ON(len == 0 && e->users == 0);
1443                 spin_unlock(&con->writequeue_lock);
1444
1445                 ret = 0;
1446                 if (len) {
1447                         ret = kernel_sendpage(con->sock, e->page, offset, len,
1448                                               msg_flags);
1449                         if (ret == -EAGAIN || ret == 0) {
1450                                 if (ret == -EAGAIN &&
1451                                     test_bit(SOCK_ASYNC_NOSPACE, &con->sock->flags) &&
1452                                     !test_and_set_bit(CF_APP_LIMITED, &con->flags)) {
1453                                         /* Notify TCP that we're limited by the
1454                                          * application window size.
1455                                          */
1456                                         set_bit(SOCK_NOSPACE, &con->sock->flags);
1457                                         con->sock->sk->sk_write_pending++;
1458                                 }
1459                                 cond_resched();
1460                                 goto out;
1461                         } else if (ret < 0)
1462                                 goto send_error;
1463                 }
1464
1465                 /* Don't starve people filling buffers */
1466                 if (++count >= MAX_SEND_MSG_COUNT) {
1467                         cond_resched();
1468                         count = 0;
1469                 }
1470
1471                 spin_lock(&con->writequeue_lock);
1472                 writequeue_entry_complete(e, ret);
1473         }
1474         spin_unlock(&con->writequeue_lock);
1475 out:
1476         mutex_unlock(&con->sock_mutex);
1477         return;
1478
1479 send_error:
1480         mutex_unlock(&con->sock_mutex);
1481         close_connection(con, false, false, true);
1482         lowcomms_connect_sock(con);
1483         return;
1484
1485 out_connect:
1486         mutex_unlock(&con->sock_mutex);
1487         lowcomms_connect_sock(con);
1488 }
1489
1490 static void clean_one_writequeue(struct connection *con)
1491 {
1492         struct writequeue_entry *e, *safe;
1493
1494         spin_lock(&con->writequeue_lock);
1495         list_for_each_entry_safe(e, safe, &con->writequeue, list) {
1496                 list_del(&e->list);
1497                 free_entry(e);
1498         }
1499         spin_unlock(&con->writequeue_lock);
1500 }
1501
1502 /* Called from recovery when it knows that a node has
1503    left the cluster */
1504 int dlm_lowcomms_close(int nodeid)
1505 {
1506         struct connection *con;
1507         struct dlm_node_addr *na;
1508
1509         log_print("closing connection to node %d", nodeid);
1510         con = nodeid2con(nodeid, 0);
1511         if (con) {
1512                 set_bit(CF_CLOSE, &con->flags);
1513                 close_connection(con, true, true, true);
1514                 clean_one_writequeue(con);
1515         }
1516
1517         spin_lock(&dlm_node_addrs_spin);
1518         na = find_node_addr(nodeid);
1519         if (na) {
1520                 list_del(&na->list);
1521                 while (na->addr_count--)
1522                         kfree(na->addr[na->addr_count]);
1523                 kfree(na);
1524         }
1525         spin_unlock(&dlm_node_addrs_spin);
1526
1527         return 0;
1528 }
1529
1530 /* Receive workqueue function */
1531 static void process_recv_sockets(struct work_struct *work)
1532 {
1533         struct connection *con = container_of(work, struct connection, rwork);
1534         int err;
1535
1536         clear_bit(CF_READ_PENDING, &con->flags);
1537         do {
1538                 err = con->rx_action(con);
1539         } while (!err);
1540 }
1541
1542 /* Send workqueue function */
1543 static void process_send_sockets(struct work_struct *work)
1544 {
1545         struct connection *con = container_of(work, struct connection, swork);
1546
1547         if (test_and_clear_bit(CF_CONNECT_PENDING, &con->flags))
1548                 con->connect_action(con);
1549         if (test_and_clear_bit(CF_WRITE_PENDING, &con->flags))
1550                 send_to_sock(con);
1551 }
1552
1553
1554 /* Discard all entries on the write queues */
1555 static void clean_writequeues(void)
1556 {
1557         foreach_conn(clean_one_writequeue);
1558 }
1559
1560 static void work_stop(void)
1561 {
1562         destroy_workqueue(recv_workqueue);
1563         destroy_workqueue(send_workqueue);
1564 }
1565
1566 static int work_start(void)
1567 {
1568         recv_workqueue = alloc_workqueue("dlm_recv",
1569                                          WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1570         if (!recv_workqueue) {
1571                 log_print("can't start dlm_recv");
1572                 return -ENOMEM;
1573         }
1574
1575         send_workqueue = alloc_workqueue("dlm_send",
1576                                          WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
1577         if (!send_workqueue) {
1578                 log_print("can't start dlm_send");
1579                 destroy_workqueue(recv_workqueue);
1580                 return -ENOMEM;
1581         }
1582
1583         return 0;
1584 }
1585
1586 static void stop_conn(struct connection *con)
1587 {
1588         con->flags |= 0x0F;
1589         if (con->sock && con->sock->sk)
1590                 con->sock->sk->sk_user_data = NULL;
1591 }
1592
1593 static void free_conn(struct connection *con)
1594 {
1595         close_connection(con, true, true, true);
1596         if (con->othercon)
1597                 kmem_cache_free(con_cache, con->othercon);
1598         hlist_del(&con->list);
1599         kmem_cache_free(con_cache, con);
1600 }
1601
1602 void dlm_lowcomms_stop(void)
1603 {
1604         /* Set all the flags to prevent any
1605            socket activity.
1606         */
1607         mutex_lock(&connections_lock);
1608         dlm_allow_conn = 0;
1609         foreach_conn(stop_conn);
1610         mutex_unlock(&connections_lock);
1611
1612         work_stop();
1613
1614         mutex_lock(&connections_lock);
1615         clean_writequeues();
1616
1617         foreach_conn(free_conn);
1618
1619         mutex_unlock(&connections_lock);
1620         kmem_cache_destroy(con_cache);
1621 }
1622
1623 int dlm_lowcomms_start(void)
1624 {
1625         int error = -EINVAL;
1626         struct connection *con;
1627         int i;
1628
1629         for (i = 0; i < CONN_HASH_SIZE; i++)
1630                 INIT_HLIST_HEAD(&connection_hash[i]);
1631
1632         init_local();
1633         if (!dlm_local_count) {
1634                 error = -ENOTCONN;
1635                 log_print("no local IP address has been set");
1636                 goto fail;
1637         }
1638
1639         error = -ENOMEM;
1640         con_cache = kmem_cache_create("dlm_conn", sizeof(struct connection),
1641                                       __alignof__(struct connection), 0,
1642                                       NULL);
1643         if (!con_cache)
1644                 goto fail;
1645
1646         error = work_start();
1647         if (error)
1648                 goto fail_destroy;
1649
1650         dlm_allow_conn = 1;
1651
1652         /* Start listening */
1653         if (dlm_config.ci_protocol == 0)
1654                 error = tcp_listen_for_all();
1655         else
1656                 error = sctp_listen_for_all();
1657         if (error)
1658                 goto fail_unlisten;
1659
1660         return 0;
1661
1662 fail_unlisten:
1663         dlm_allow_conn = 0;
1664         con = nodeid2con(0,0);
1665         if (con) {
1666                 close_connection(con, false, true, true);
1667                 kmem_cache_free(con_cache, con);
1668         }
1669 fail_destroy:
1670         kmem_cache_destroy(con_cache);
1671 fail:
1672         return error;
1673 }
1674
1675 void dlm_lowcomms_exit(void)
1676 {
1677         struct dlm_node_addr *na, *safe;
1678
1679         spin_lock(&dlm_node_addrs_spin);
1680         list_for_each_entry_safe(na, safe, &dlm_node_addrs, list) {
1681                 list_del(&na->list);
1682                 while (na->addr_count--)
1683                         kfree(na->addr[na->addr_count]);
1684                 kfree(na);
1685         }
1686         spin_unlock(&dlm_node_addrs_spin);
1687 }