net/tipc/server.c

   1 /*
   2  * net/tipc/server.c: TIPC server infrastructure
   3  *
   4  * Copyright (c) 2012-2013, Wind River Systems
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions are met:
   9  *
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the names of the copyright holders nor the names of its
  16  *    contributors may be used to endorse or promote products derived from
  17  *    this software without specific prior written permission.
  18  *
  19  * Alternatively, this software may be distributed under the terms of the
  20  * GNU General Public License ("GPL") version 2 as published by the Free
  21  * Software Foundation.
  22  *
  23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  24  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  26  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  27  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  33  * POSSIBILITY OF SUCH DAMAGE.
  34  */
  35
  36 #include "server.h"
  37 #include "core.h"
  38 #include "socket.h"
  39 #include <net/sock.h>
  40 #include <linux/module.h>
  41
  42 /* Number of messages to send before rescheduling */
  43 #define MAX_SEND_MSG_COUNT      25
  44 #define MAX_RECV_MSG_COUNT      25
  45 #define CF_CONNECTED            1
  46 #define CF_SERVER               2
  47
  48 #define sock2con(x) ((struct tipc_conn *)(x)->sk_user_data)
  49
  50 /**
  51  * struct tipc_conn - TIPC connection structure
  52  * @kref: reference counter to connection object
  53  * @conid: connection identifier
  54  * @sock: socket handler associated with connection
  55  * @flags: indicates connection state
  56  * @server: pointer to connected server
  57  * @rwork: receive work item
  58  * @usr_data: user-specified field
  59  * @rx_action: what to do when connection socket is active
  60  * @outqueue: pointer to first outbound message in queue
  61  * @outqueue_lock: control access to the outqueue
  62  * @outqueue: list of connection objects for its server
  63  * @swork: send work item
  64  */
  65 struct tipc_conn {
  66         struct kref kref;
  67         int conid;
  68         struct socket *sock;
  69         unsigned long flags;
  70         struct tipc_server *server;
  71         struct work_struct rwork;
  72         int (*rx_action) (struct tipc_conn *con);
  73         void *usr_data;
  74         struct list_head outqueue;
  75         spinlock_t outqueue_lock;
  76         struct work_struct swork;
  77 };
  78
  79 /* An entry waiting to be sent */
  80 struct outqueue_entry {
  81         struct list_head list;
  82         struct kvec iov;
  83         struct sockaddr_tipc dest;
  84 };
  85
  86 static void tipc_recv_work(struct work_struct *work);
  87 static void tipc_send_work(struct work_struct *work);
  88 static void tipc_clean_outqueues(struct tipc_conn *con);
  89
  90 static void tipc_conn_kref_release(struct kref *kref)
  91 {
  92         struct tipc_conn *con = container_of(kref, struct tipc_conn, kref);
  93         struct sockaddr_tipc *saddr = con->server->saddr;
  94         struct socket *sock = con->sock;
  95         struct sock *sk;
  96
  97         if (sock) {
  98                 sk = sock->sk;
  99                 if (test_bit(CF_SERVER, &con->flags)) {
 100                         __module_get(sock->ops->owner);
 101                         __module_get(sk->sk_prot_creator->owner);
 102                 }
 103                 saddr->scope = -TIPC_NODE_SCOPE;
 104                 kernel_bind(sock, (struct sockaddr *)saddr, sizeof(*saddr));
 105                 sock_release(sock);
 106                 con->sock = NULL;
 107         }
 108
 109         tipc_clean_outqueues(con);
 110         kfree(con);
 111 }
 112
 113 static void conn_put(struct tipc_conn *con)
 114 {
 115         kref_put(&con->kref, tipc_conn_kref_release);
 116 }
 117
 118 static void conn_get(struct tipc_conn *con)
 119 {
 120         kref_get(&con->kref);
 121 }
 122
 123 static struct tipc_conn *tipc_conn_lookup(struct tipc_server *s, int conid)
 124 {
 125         struct tipc_conn *con;
 126
 127         spin_lock_bh(&s->idr_lock);
 128         con = idr_find(&s->conn_idr, conid);
 129         if (con)
 130                 conn_get(con);
 131         spin_unlock_bh(&s->idr_lock);
 132         return con;
 133 }
 134
 135 static void sock_data_ready(struct sock *sk)
 136 {
 137         struct tipc_conn *con;
 138
 139         read_lock(&sk->sk_callback_lock);
 140         con = sock2con(sk);
 141         if (con && test_bit(CF_CONNECTED, &con->flags)) {
 142                 conn_get(con);
 143                 if (!queue_work(con->server->rcv_wq, &con->rwork))
 144                         conn_put(con);
 145         }
 146         read_unlock(&sk->sk_callback_lock);
 147 }
 148
 149 static void sock_write_space(struct sock *sk)
 150 {
 151         struct tipc_conn *con;
 152
 153         read_lock(&sk->sk_callback_lock);
 154         con = sock2con(sk);
 155         if (con && test_bit(CF_CONNECTED, &con->flags)) {
 156                 conn_get(con);
 157                 if (!queue_work(con->server->send_wq, &con->swork))
 158                         conn_put(con);
 159         }
 160         read_unlock(&sk->sk_callback_lock);
 161 }
 162
 163 static void tipc_register_callbacks(struct socket *sock, struct tipc_conn *con)
 164 {
 165         struct sock *sk = sock->sk;
 166
 167         write_lock_bh(&sk->sk_callback_lock);
 168
 169         sk->sk_data_ready = sock_data_ready;
 170         sk->sk_write_space = sock_write_space;
 171         sk->sk_user_data = con;
 172
 173         con->sock = sock;
 174
 175         write_unlock_bh(&sk->sk_callback_lock);
 176 }
 177
 178 static void tipc_unregister_callbacks(struct tipc_conn *con)
 179 {
 180         struct sock *sk = con->sock->sk;
 181
 182         write_lock_bh(&sk->sk_callback_lock);
 183         sk->sk_user_data = NULL;
 184         write_unlock_bh(&sk->sk_callback_lock);
 185 }
 186
 187 static void tipc_close_conn(struct tipc_conn *con)
 188 {
 189         struct tipc_server *s = con->server;
 190
 191         if (test_and_clear_bit(CF_CONNECTED, &con->flags)) {
 192                 if (con->conid)
 193                         s->tipc_conn_shutdown(con->conid, con->usr_data);
 194
 195                 spin_lock_bh(&s->idr_lock);
 196                 idr_remove(&s->conn_idr, con->conid);
 197                 s->idr_in_use--;
 198                 spin_unlock_bh(&s->idr_lock);
 199
 200                 tipc_unregister_callbacks(con);
 201
 202                 /* We shouldn't flush pending works as we may be in the
 203                  * thread. In fact the races with pending rx/tx work structs
 204                  * are harmless for us here as we have already deleted this
 205                  * connection from server connection list and set
 206                  * sk->sk_user_data to 0 before releasing connection object.
 207                  */
 208                 kernel_sock_shutdown(con->sock, SHUT_RDWR);
 209
 210                 conn_put(con);
 211         }
 212 }
 213
 214 static struct tipc_conn *tipc_alloc_conn(struct tipc_server *s)
 215 {
 216         struct tipc_conn *con;
 217         int ret;
 218
 219         con = kzalloc(sizeof(struct tipc_conn), GFP_ATOMIC);
 220         if (!con)
 221                 return ERR_PTR(-ENOMEM);
 222
 223         kref_init(&con->kref);
 224         INIT_LIST_HEAD(&con->outqueue);
 225         spin_lock_init(&con->outqueue_lock);
 226         INIT_WORK(&con->swork, tipc_send_work);
 227         INIT_WORK(&con->rwork, tipc_recv_work);
 228
 229         spin_lock_bh(&s->idr_lock);
 230         ret = idr_alloc(&s->conn_idr, con, 0, 0, GFP_ATOMIC);
 231         if (ret < 0) {
 232                 kfree(con);
 233                 spin_unlock_bh(&s->idr_lock);
 234                 return ERR_PTR(-ENOMEM);
 235         }
 236         con->conid = ret;
 237         s->idr_in_use++;
 238         spin_unlock_bh(&s->idr_lock);
 239
 240         set_bit(CF_CONNECTED, &con->flags);
 241         con->server = s;
 242
 243         return con;
 244 }
 245
 246 static int tipc_receive_from_sock(struct tipc_conn *con)
 247 {
 248         struct msghdr msg = {};
 249         struct tipc_server *s = con->server;
 250         struct sockaddr_tipc addr;
 251         struct kvec iov;
 252         void *buf;
 253         int ret;
 254
 255         buf = kmem_cache_alloc(s->rcvbuf_cache, GFP_ATOMIC);
 256         if (!buf) {
 257                 ret = -ENOMEM;
 258                 goto out_close;
 259         }
 260
 261         iov.iov_base = buf;
 262         iov.iov_len = s->max_rcvbuf_size;
 263         msg.msg_name = &addr;
 264         ret = kernel_recvmsg(con->sock, &msg, &iov, 1, iov.iov_len,
 265                              MSG_DONTWAIT);
 266         if (ret <= 0) {
 267                 kmem_cache_free(s->rcvbuf_cache, buf);
 268                 goto out_close;
 269         }
 270
 271         s->tipc_conn_recvmsg(sock_net(con->sock->sk), con->conid, &addr,
 272                              con->usr_data, buf, ret);
 273
 274         kmem_cache_free(s->rcvbuf_cache, buf);
 275
 276         return 0;
 277
 278 out_close:
 279         if (ret != -EWOULDBLOCK)
 280                 tipc_close_conn(con);
 281         else if (ret == 0)
 282                 /* Don't return success if we really got EOF */
 283                 ret = -EAGAIN;
 284
 285         return ret;
 286 }
 287
 288 static int tipc_accept_from_sock(struct tipc_conn *con)
 289 {
 290         struct tipc_server *s = con->server;
 291         struct socket *sock = con->sock;
 292         struct socket *newsock;
 293         struct tipc_conn *newcon;
 294         int ret;
 295
 296         ret = kernel_accept(sock, &newsock, O_NONBLOCK);
 297         if (ret < 0)
 298                 return ret;
 299
 300         newcon = tipc_alloc_conn(con->server);
 301         if (IS_ERR(newcon)) {
 302                 ret = PTR_ERR(newcon);
 303                 sock_release(newsock);
 304                 return ret;
 305         }
 306
 307         newcon->rx_action = tipc_receive_from_sock;
 308         tipc_register_callbacks(newsock, newcon);
 309
 310         /* Notify that new connection is incoming */
 311         newcon->usr_data = s->tipc_conn_new(newcon->conid);
 312         if (!newcon->usr_data) {
 313                 sock_release(newsock);
 314                 return -ENOMEM;
 315         }
 316
 317         /* Wake up receive process in case of 'SYN+' message */
 318         newsock->sk->sk_data_ready(newsock->sk);
 319         return ret;
 320 }
 321
 322 static struct socket *tipc_create_listen_sock(struct tipc_conn *con)
 323 {
 324         struct tipc_server *s = con->server;
 325         struct socket *sock = NULL;
 326         int ret;
 327
 328         ret = sock_create_kern(s->net, AF_TIPC, SOCK_SEQPACKET, 0, &sock);
 329         if (ret < 0)
 330                 return NULL;
 331         ret = kernel_setsockopt(sock, SOL_TIPC, TIPC_IMPORTANCE,
 332                                 (char *)&s->imp, sizeof(s->imp));
 333         if (ret < 0)
 334                 goto create_err;
 335         ret = kernel_bind(sock, (struct sockaddr *)s->saddr, sizeof(*s->saddr));
 336         if (ret < 0)
 337                 goto create_err;
 338
 339         switch (s->type) {
 340         case SOCK_STREAM:
 341         case SOCK_SEQPACKET:
 342                 con->rx_action = tipc_accept_from_sock;
 343
 344                 ret = kernel_listen(sock, 0);
 345                 if (ret < 0)
 346                         goto create_err;
 347                 break;
 348         case SOCK_DGRAM:
 349         case SOCK_RDM:
 350                 con->rx_action = tipc_receive_from_sock;
 351                 break;
 352         default:
 353                 pr_err("Unknown socket type %d\n", s->type);
 354                 goto create_err;
 355         }
 356
 357         /* As server's listening socket owner and creator is the same module,
 358          * we have to decrease TIPC module reference count to guarantee that
 359          * it remains zero after the server socket is created, otherwise,
 360          * executing "rmmod" command is unable to make TIPC module deleted
 361          * after TIPC module is inserted successfully.
 362          *
 363          * However, the reference count is ever increased twice in
 364          * sock_create_kern(): one is to increase the reference count of owner
 365          * of TIPC socket's proto_ops struct; another is to increment the
 366          * reference count of owner of TIPC proto struct. Therefore, we must
 367          * decrement the module reference count twice to ensure that it keeps
 368          * zero after server's listening socket is created. Of course, we
 369          * must bump the module reference count twice as well before the socket
 370          * is closed.
 371          */
 372         module_put(sock->ops->owner);
 373         module_put(sock->sk->sk_prot_creator->owner);
 374         set_bit(CF_SERVER, &con->flags);
 375
 376         return sock;
 377
 378 create_err:
 379         kernel_sock_shutdown(sock, SHUT_RDWR);
 380         sock_release(sock);
 381         return NULL;
 382 }
 383
 384 static int tipc_open_listening_sock(struct tipc_server *s)
 385 {
 386         struct socket *sock;
 387         struct tipc_conn *con;
 388
 389         con = tipc_alloc_conn(s);
 390         if (IS_ERR(con))
 391                 return PTR_ERR(con);
 392
 393         sock = tipc_create_listen_sock(con);
 394         if (!sock) {
 395                 idr_remove(&s->conn_idr, con->conid);
 396                 s->idr_in_use--;
 397                 kfree(con);
 398                 return -EINVAL;
 399         }
 400
 401         tipc_register_callbacks(sock, con);
 402         return 0;
 403 }
 404
 405 static struct outqueue_entry *tipc_alloc_entry(void *data, int len)
 406 {
 407         struct outqueue_entry *entry;
 408         void *buf;
 409
 410         entry = kmalloc(sizeof(struct outqueue_entry), GFP_ATOMIC);
 411         if (!entry)
 412                 return NULL;
 413
 414         buf = kmalloc(len, GFP_ATOMIC);
 415         if (!buf) {
 416                 kfree(entry);
 417                 return NULL;
 418         }
 419
 420         memcpy(buf, data, len);
 421         entry->iov.iov_base = buf;
 422         entry->iov.iov_len = len;
 423
 424         return entry;
 425 }
 426
 427 static void tipc_free_entry(struct outqueue_entry *e)
 428 {
 429         kfree(e->iov.iov_base);
 430         kfree(e);
 431 }
 432
 433 static void tipc_clean_outqueues(struct tipc_conn *con)
 434 {
 435         struct outqueue_entry *e, *safe;
 436
 437         spin_lock_bh(&con->outqueue_lock);
 438         list_for_each_entry_safe(e, safe, &con->outqueue, list) {
 439                 list_del(&e->list);
 440                 tipc_free_entry(e);
 441         }
 442         spin_unlock_bh(&con->outqueue_lock);
 443 }
 444
 445 int tipc_conn_sendmsg(struct tipc_server *s, int conid,
 446                       struct sockaddr_tipc *addr, void *data, size_t len)
 447 {
 448         struct outqueue_entry *e;
 449         struct tipc_conn *con;
 450
 451         con = tipc_conn_lookup(s, conid);
 452         if (!con)
 453                 return -EINVAL;
 454
 455         e = tipc_alloc_entry(data, len);
 456         if (!e) {
 457                 conn_put(con);
 458                 return -ENOMEM;
 459         }
 460
 461         if (addr)
 462                 memcpy(&e->dest, addr, sizeof(struct sockaddr_tipc));
 463
 464         spin_lock_bh(&con->outqueue_lock);
 465         list_add_tail(&e->list, &con->outqueue);
 466         spin_unlock_bh(&con->outqueue_lock);
 467
 468         if (test_bit(CF_CONNECTED, &con->flags)) {
 469                 if (!queue_work(s->send_wq, &con->swork))
 470                         conn_put(con);
 471         } else {
 472                 conn_put(con);
 473         }
 474         return 0;
 475 }
 476
 477 void tipc_conn_terminate(struct tipc_server *s, int conid)
 478 {
 479         struct tipc_conn *con;
 480
 481         con = tipc_conn_lookup(s, conid);
 482         if (con) {
 483                 tipc_close_conn(con);
 484                 conn_put(con);
 485         }
 486 }
 487
 488 static void tipc_send_to_sock(struct tipc_conn *con)
 489 {
 490         int count = 0;
 491         struct tipc_server *s = con->server;
 492         struct outqueue_entry *e;
 493         struct msghdr msg;
 494         int ret;
 495
 496         spin_lock_bh(&con->outqueue_lock);
 497         while (1) {
 498                 e = list_entry(con->outqueue.next, struct outqueue_entry,
 499                                list);
 500                 if ((struct list_head *) e == &con->outqueue)
 501                         break;
 502                 spin_unlock_bh(&con->outqueue_lock);
 503
 504                 memset(&msg, 0, sizeof(msg));
 505                 msg.msg_flags = MSG_DONTWAIT;
 506
 507                 if (s->type == SOCK_DGRAM || s->type == SOCK_RDM) {
 508                         msg.msg_name = &e->dest;
 509                         msg.msg_namelen = sizeof(struct sockaddr_tipc);
 510                 }
 511                 ret = kernel_sendmsg(con->sock, &msg, &e->iov, 1,
 512                                      e->iov.iov_len);
 513                 if (ret == -EWOULDBLOCK || ret == 0) {
 514                         cond_resched();
 515                         goto out;
 516                 } else if (ret < 0) {
 517                         goto send_err;
 518                 }
 519
 520                 /* Don't starve users filling buffers */
 521                 if (++count >= MAX_SEND_MSG_COUNT) {
 522                         cond_resched();
 523                         count = 0;
 524                 }
 525
 526                 spin_lock_bh(&con->outqueue_lock);
 527                 list_del(&e->list);
 528                 tipc_free_entry(e);
 529         }
 530         spin_unlock_bh(&con->outqueue_lock);
 531 out:
 532         return;
 533
 534 send_err:
 535         tipc_close_conn(con);
 536 }
 537
 538 static void tipc_recv_work(struct work_struct *work)
 539 {
 540         struct tipc_conn *con = container_of(work, struct tipc_conn, rwork);
 541         int count = 0;
 542
 543         while (test_bit(CF_CONNECTED, &con->flags)) {
 544                 if (con->rx_action(con))
 545                         break;
 546
 547                 /* Don't flood Rx machine */
 548                 if (++count >= MAX_RECV_MSG_COUNT) {
 549                         cond_resched();
 550                         count = 0;
 551                 }
 552         }
 553         conn_put(con);
 554 }
 555
 556 static void tipc_send_work(struct work_struct *work)
 557 {
 558         struct tipc_conn *con = container_of(work, struct tipc_conn, swork);
 559
 560         if (test_bit(CF_CONNECTED, &con->flags))
 561                 tipc_send_to_sock(con);
 562
 563         conn_put(con);
 564 }
 565
 566 static void tipc_work_stop(struct tipc_server *s)
 567 {
 568         destroy_workqueue(s->rcv_wq);
 569         destroy_workqueue(s->send_wq);
 570 }
 571
 572 static int tipc_work_start(struct tipc_server *s)
 573 {
 574         s->rcv_wq = alloc_ordered_workqueue("tipc_rcv", 0);
 575         if (!s->rcv_wq) {
 576                 pr_err("can't start tipc receive workqueue\n");
 577                 return -ENOMEM;
 578         }
 579
 580         s->send_wq = alloc_ordered_workqueue("tipc_send", 0);
 581         if (!s->send_wq) {
 582                 pr_err("can't start tipc send workqueue\n");
 583                 destroy_workqueue(s->rcv_wq);
 584                 return -ENOMEM;
 585         }
 586
 587         return 0;
 588 }
 589
 590 int tipc_server_start(struct tipc_server *s)
 591 {
 592         int ret;
 593
 594         spin_lock_init(&s->idr_lock);
 595         idr_init(&s->conn_idr);
 596         s->idr_in_use = 0;
 597
 598         s->rcvbuf_cache = kmem_cache_create(s->name, s->max_rcvbuf_size,
 599                                             0, SLAB_HWCACHE_ALIGN, NULL);
 600         if (!s->rcvbuf_cache)
 601                 return -ENOMEM;
 602
 603         ret = tipc_work_start(s);
 604         if (ret < 0) {
 605                 kmem_cache_destroy(s->rcvbuf_cache);
 606                 return ret;
 607         }
 608         ret = tipc_open_listening_sock(s);
 609         if (ret < 0) {
 610                 tipc_work_stop(s);
 611                 kmem_cache_destroy(s->rcvbuf_cache);
 612                 return ret;
 613         }
 614         return ret;
 615 }
 616
 617 void tipc_server_stop(struct tipc_server *s)
 618 {
 619         struct tipc_conn *con;
 620         int total = 0;
 621         int id;
 622
 623         spin_lock_bh(&s->idr_lock);
 624         for (id = 0; total < s->idr_in_use; id++) {
 625                 con = idr_find(&s->conn_idr, id);
 626                 if (con) {
 627                         total++;
 628                         spin_unlock_bh(&s->idr_lock);
 629                         tipc_close_conn(con);
 630                         spin_lock_bh(&s->idr_lock);
 631                 }
 632         }
 633         spin_unlock_bh(&s->idr_lock);
 634
 635         tipc_work_stop(s);
 636         kmem_cache_destroy(s->rcvbuf_cache);
 637         idr_destroy(&s->conn_idr);
 638 }