net/ceph/messenger.c

   1 #include <linux/ceph/ceph_debug.h>
   2
   3 #include <linux/crc32c.h>
   4 #include <linux/ctype.h>
   5 #include <linux/highmem.h>
   6 #include <linux/inet.h>
   7 #include <linux/kthread.h>
   8 #include <linux/net.h>
   9 #include <linux/slab.h>
  10 #include <linux/socket.h>
  11 #include <linux/string.h>
  12 #include <linux/bio.h>
  13 #include <linux/blkdev.h>
  14 #include <net/tcp.h>
  15
  16 #include <linux/ceph/libceph.h>
  17 #include <linux/ceph/messenger.h>
  18 #include <linux/ceph/decode.h>
  19 #include <linux/ceph/pagelist.h>
  20
  21 /*
  22  * Ceph uses the messenger to exchange ceph_msg messages with other
  23  * hosts in the system.  The messenger provides ordered and reliable
  24  * delivery.  We tolerate TCP disconnects by reconnecting (with
  25  * exponential backoff) in the case of a fault (disconnection, bad
  26  * crc, protocol error).  Acks allow sent messages to be discarded by
  27  * the sender.
  28  */
  29
  30 /* static tag bytes (protocol control messages) */
  31 static char tag_msg = CEPH_MSGR_TAG_MSG;
  32 static char tag_ack = CEPH_MSGR_TAG_ACK;
  33 static char tag_keepalive = CEPH_MSGR_TAG_KEEPALIVE;
  34
  35 #ifdef CONFIG_LOCKDEP
  36 static struct lock_class_key socket_class;
  37 #endif
  38
  39
  40 static void queue_con(struct ceph_connection *con);
  41 static void con_work(struct work_struct *);
  42 static void ceph_fault(struct ceph_connection *con);
  43
  44 /*
  45  * nicely render a sockaddr as a string.
  46  */
  47 #define MAX_ADDR_STR 20
  48 #define MAX_ADDR_STR_LEN 60
  49 static char addr_str[MAX_ADDR_STR][MAX_ADDR_STR_LEN];
  50 static DEFINE_SPINLOCK(addr_str_lock);
  51 static int last_addr_str;
  52
  53 const char *ceph_pr_addr(const struct sockaddr_storage *ss)
  54 {
  55         int i;
  56         char *s;
  57         struct sockaddr_in *in4 = (void *)ss;
  58         struct sockaddr_in6 *in6 = (void *)ss;
  59
  60         spin_lock(&addr_str_lock);
  61         i = last_addr_str++;
  62         if (last_addr_str == MAX_ADDR_STR)
  63                 last_addr_str = 0;
  64         spin_unlock(&addr_str_lock);
  65         s = addr_str[i];
  66
  67         switch (ss->ss_family) {
  68         case AF_INET:
  69                 snprintf(s, MAX_ADDR_STR_LEN, "%pI4:%u", &in4->sin_addr,
  70                          (unsigned int)ntohs(in4->sin_port));
  71                 break;
  72
  73         case AF_INET6:
  74                 snprintf(s, MAX_ADDR_STR_LEN, "[%pI6c]:%u", &in6->sin6_addr,
  75                          (unsigned int)ntohs(in6->sin6_port));
  76                 break;
  77
  78         default:
  79                 snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %d)",
  80                          (int)ss->ss_family);
  81         }
  82
  83         return s;
  84 }
  85 EXPORT_SYMBOL(ceph_pr_addr);
  86
  87 static void encode_my_addr(struct ceph_messenger *msgr)
  88 {
  89         memcpy(&msgr->my_enc_addr, &msgr->inst.addr, sizeof(msgr->my_enc_addr));
  90         ceph_encode_addr(&msgr->my_enc_addr);
  91 }
  92
  93 /*
  94  * work queue for all reading and writing to/from the socket.
  95  */
  96 struct workqueue_struct *ceph_msgr_wq;
  97
  98 int ceph_msgr_init(void)
  99 {
 100         ceph_msgr_wq = alloc_workqueue("ceph-msgr", WQ_NON_REENTRANT, 0);
 101         if (!ceph_msgr_wq) {
 102                 pr_err("msgr_init failed to create workqueue\n");
 103                 return -ENOMEM;
 104         }
 105         return 0;
 106 }
 107 EXPORT_SYMBOL(ceph_msgr_init);
 108
 109 void ceph_msgr_exit(void)
 110 {
 111         destroy_workqueue(ceph_msgr_wq);
 112 }
 113 EXPORT_SYMBOL(ceph_msgr_exit);
 114
 115 void ceph_msgr_flush(void)
 116 {
 117         flush_workqueue(ceph_msgr_wq);
 118 }
 119 EXPORT_SYMBOL(ceph_msgr_flush);
 120
 121
 122 /*
 123  * socket callback functions
 124  */
 125
 126 /* data available on socket, or listen socket received a connect */
 127 static void ceph_data_ready(struct sock *sk, int count_unused)
 128 {
 129         struct ceph_connection *con =
 130                 (struct ceph_connection *)sk->sk_user_data;
 131         if (sk->sk_state != TCP_CLOSE_WAIT) {
 132                 dout("ceph_data_ready on %p state = %lu, queueing work\n",
 133                      con, con->state);
 134                 queue_con(con);
 135         }
 136 }
 137
 138 /* socket has buffer space for writing */
 139 static void ceph_write_space(struct sock *sk)
 140 {
 141         struct ceph_connection *con =
 142                 (struct ceph_connection *)sk->sk_user_data;
 143
 144         /* only queue to workqueue if there is data we want to write. */
 145         if (test_bit(WRITE_PENDING, &con->state)) {
 146                 dout("ceph_write_space %p queueing write work\n", con);
 147                 queue_con(con);
 148         } else {
 149                 dout("ceph_write_space %p nothing to write\n", con);
 150         }
 151
 152         /* since we have our own write_space, clear the SOCK_NOSPACE flag */
 153         clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
 154 }
 155
 156 /* socket's state has changed */
 157 static void ceph_state_change(struct sock *sk)
 158 {
 159         struct ceph_connection *con =
 160                 (struct ceph_connection *)sk->sk_user_data;
 161
 162         dout("ceph_state_change %p state = %lu sk_state = %u\n",
 163              con, con->state, sk->sk_state);
 164
 165         if (test_bit(CLOSED, &con->state))
 166                 return;
 167
 168         switch (sk->sk_state) {
 169         case TCP_CLOSE:
 170                 dout("ceph_state_change TCP_CLOSE\n");
 171         case TCP_CLOSE_WAIT:
 172                 dout("ceph_state_change TCP_CLOSE_WAIT\n");
 173                 if (test_and_set_bit(SOCK_CLOSED, &con->state) == 0) {
 174                         if (test_bit(CONNECTING, &con->state))
 175                                 con->error_msg = "connection failed";
 176                         else
 177                                 con->error_msg = "socket closed";
 178                         queue_con(con);
 179                 }
 180                 break;
 181         case TCP_ESTABLISHED:
 182                 dout("ceph_state_change TCP_ESTABLISHED\n");
 183                 queue_con(con);
 184                 break;
 185         }
 186 }
 187
 188 /*
 189  * set up socket callbacks
 190  */
 191 static void set_sock_callbacks(struct socket *sock,
 192                                struct ceph_connection *con)
 193 {
 194         struct sock *sk = sock->sk;
 195         sk->sk_user_data = (void *)con;
 196         sk->sk_data_ready = ceph_data_ready;
 197         sk->sk_write_space = ceph_write_space;
 198         sk->sk_state_change = ceph_state_change;
 199 }
 200
 201
 202 /*
 203  * socket helpers
 204  */
 205
 206 /*
 207  * initiate connection to a remote socket.
 208  */
 209 static struct socket *ceph_tcp_connect(struct ceph_connection *con)
 210 {
 211         struct sockaddr_storage *paddr = &con->peer_addr.in_addr;
 212         struct socket *sock;
 213         int ret;
 214
 215         BUG_ON(con->sock);
 216         ret = sock_create_kern(con->peer_addr.in_addr.ss_family, SOCK_STREAM,
 217                                IPPROTO_TCP, &sock);
 218         if (ret)
 219                 return ERR_PTR(ret);
 220         con->sock = sock;
 221         sock->sk->sk_allocation = GFP_NOFS;
 222
 223 #ifdef CONFIG_LOCKDEP
 224         lockdep_set_class(&sock->sk->sk_lock, &socket_class);
 225 #endif
 226
 227         set_sock_callbacks(sock, con);
 228
 229         dout("connect %s\n", ceph_pr_addr(&con->peer_addr.in_addr));
 230
 231         ret = sock->ops->connect(sock, (struct sockaddr *)paddr, sizeof(*paddr),
 232                                  O_NONBLOCK);
 233         if (ret == -EINPROGRESS) {
 234                 dout("connect %s EINPROGRESS sk_state = %u\n",
 235                      ceph_pr_addr(&con->peer_addr.in_addr),
 236                      sock->sk->sk_state);
 237                 ret = 0;
 238         }
 239         if (ret < 0) {
 240                 pr_err("connect %s error %d\n",
 241                        ceph_pr_addr(&con->peer_addr.in_addr), ret);
 242                 sock_release(sock);
 243                 con->sock = NULL;
 244                 con->error_msg = "connect error";
 245         }
 246
 247         if (ret < 0)
 248                 return ERR_PTR(ret);
 249         return sock;
 250 }
 251
 252 static int ceph_tcp_recvmsg(struct socket *sock, void *buf, size_t len)
 253 {
 254         struct kvec iov = {buf, len};
 255         struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
 256         int r;
 257
 258         r = kernel_recvmsg(sock, &msg, &iov, 1, len, msg.msg_flags);
 259         if (r == -EAGAIN)
 260                 r = 0;
 261         return r;
 262 }
 263
 264 /*
 265  * write something.  @more is true if caller will be sending more data
 266  * shortly.
 267  */
 268 static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov,
 269                      size_t kvlen, size_t len, int more)
 270 {
 271         struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
 272         int r;
 273
 274         if (more)
 275                 msg.msg_flags |= MSG_MORE;
 276         else
 277                 msg.msg_flags |= MSG_EOR;  /* superfluous, but what the hell */
 278
 279         r = kernel_sendmsg(sock, &msg, iov, kvlen, len);
 280         if (r == -EAGAIN)
 281                 r = 0;
 282         return r;
 283 }
 284
 285
 286 /*
 287  * Shutdown/close the socket for the given connection.
 288  */
 289 static int con_close_socket(struct ceph_connection *con)
 290 {
 291         int rc;
 292
 293         dout("con_close_socket on %p sock %p\n", con, con->sock);
 294         if (!con->sock)
 295                 return 0;
 296         set_bit(SOCK_CLOSED, &con->state);
 297         rc = con->sock->ops->shutdown(con->sock, SHUT_RDWR);
 298         sock_release(con->sock);
 299         con->sock = NULL;
 300         clear_bit(SOCK_CLOSED, &con->state);
 301         return rc;
 302 }
 303
 304 /*
 305  * Reset a connection.  Discard all incoming and outgoing messages
 306  * and clear *_seq state.
 307  */
 308 static void ceph_msg_remove(struct ceph_msg *msg)
 309 {
 310         list_del_init(&msg->list_head);
 311         ceph_msg_put(msg);
 312 }
 313 static void ceph_msg_remove_list(struct list_head *head)
 314 {
 315         while (!list_empty(head)) {
 316                 struct ceph_msg *msg = list_first_entry(head, struct ceph_msg,
 317                                                         list_head);
 318                 ceph_msg_remove(msg);
 319         }
 320 }
 321
 322 static void reset_connection(struct ceph_connection *con)
 323 {
 324         /* reset connection, out_queue, msg_ and connect_seq */
 325         /* discard existing out_queue and msg_seq */
 326         ceph_msg_remove_list(&con->out_queue);
 327         ceph_msg_remove_list(&con->out_sent);
 328
 329         if (con->in_msg) {
 330                 ceph_msg_put(con->in_msg);
 331                 con->in_msg = NULL;
 332         }
 333
 334         con->connect_seq = 0;
 335         con->out_seq = 0;
 336         if (con->out_msg) {
 337                 ceph_msg_put(con->out_msg);
 338                 con->out_msg = NULL;
 339         }
 340         con->in_seq = 0;
 341         con->in_seq_acked = 0;
 342 }
 343
 344 /*
 345  * mark a peer down.  drop any open connections.
 346  */
 347 void ceph_con_close(struct ceph_connection *con)
 348 {
 349         dout("con_close %p peer %s\n", con,
 350              ceph_pr_addr(&con->peer_addr.in_addr));
 351         set_bit(CLOSED, &con->state);  /* in case there's queued work */
 352         clear_bit(STANDBY, &con->state);  /* avoid connect_seq bump */
 353         clear_bit(LOSSYTX, &con->state);  /* so we retry next connect */
 354         clear_bit(KEEPALIVE_PENDING, &con->state);
 355         clear_bit(WRITE_PENDING, &con->state);
 356         mutex_lock(&con->mutex);
 357         reset_connection(con);
 358         con->peer_global_seq = 0;
 359         cancel_delayed_work(&con->work);
 360         mutex_unlock(&con->mutex);
 361         queue_con(con);
 362 }
 363 EXPORT_SYMBOL(ceph_con_close);
 364
 365 /*
 366  * Reopen a closed connection, with a new peer address.
 367  */
 368 void ceph_con_open(struct ceph_connection *con, struct ceph_entity_addr *addr)
 369 {
 370         dout("con_open %p %s\n", con, ceph_pr_addr(&addr->in_addr));
 371         set_bit(OPENING, &con->state);
 372         clear_bit(CLOSED, &con->state);
 373         memcpy(&con->peer_addr, addr, sizeof(*addr));
 374         con->delay = 0;      /* reset backoff memory */
 375         queue_con(con);
 376 }
 377 EXPORT_SYMBOL(ceph_con_open);
 378
 379 /*
 380  * return true if this connection ever successfully opened
 381  */
 382 bool ceph_con_opened(struct ceph_connection *con)
 383 {
 384         return con->connect_seq > 0;
 385 }
 386
 387 /*
 388  * generic get/put
 389  */
 390 struct ceph_connection *ceph_con_get(struct ceph_connection *con)
 391 {
 392         dout("con_get %p nref = %d -> %d\n", con,
 393              atomic_read(&con->nref), atomic_read(&con->nref) + 1);
 394         if (atomic_inc_not_zero(&con->nref))
 395                 return con;
 396         return NULL;
 397 }
 398
 399 void ceph_con_put(struct ceph_connection *con)
 400 {
 401         dout("con_put %p nref = %d -> %d\n", con,
 402              atomic_read(&con->nref), atomic_read(&con->nref) - 1);
 403         BUG_ON(atomic_read(&con->nref) == 0);
 404         if (atomic_dec_and_test(&con->nref)) {
 405                 BUG_ON(con->sock);
 406                 kfree(con);
 407         }
 408 }
 409
 410 /*
 411  * initialize a new connection.
 412  */
 413 void ceph_con_init(struct ceph_messenger *msgr, struct ceph_connection *con)
 414 {
 415         dout("con_init %p\n", con);
 416         memset(con, 0, sizeof(*con));
 417         atomic_set(&con->nref, 1);
 418         con->msgr = msgr;
 419         mutex_init(&con->mutex);
 420         INIT_LIST_HEAD(&con->out_queue);
 421         INIT_LIST_HEAD(&con->out_sent);
 422         INIT_DELAYED_WORK(&con->work, con_work);
 423 }
 424 EXPORT_SYMBOL(ceph_con_init);
 425
 426
 427 /*
 428  * We maintain a global counter to order connection attempts.  Get
 429  * a unique seq greater than @gt.
 430  */
 431 static u32 get_global_seq(struct ceph_messenger *msgr, u32 gt)
 432 {
 433         u32 ret;
 434
 435         spin_lock(&msgr->global_seq_lock);
 436         if (msgr->global_seq < gt)
 437                 msgr->global_seq = gt;
 438         ret = ++msgr->global_seq;
 439         spin_unlock(&msgr->global_seq_lock);
 440         return ret;
 441 }
 442
 443
 444 /*
 445  * Prepare footer for currently outgoing message, and finish things
 446  * off.  Assumes out_kvec* are already valid.. we just add on to the end.
 447  */
 448 static void prepare_write_message_footer(struct ceph_connection *con, int v)
 449 {
 450         struct ceph_msg *m = con->out_msg;
 451
 452         dout("prepare_write_message_footer %p\n", con);
 453         con->out_kvec_is_msg = true;
 454         con->out_kvec[v].iov_base = &m->footer;
 455         con->out_kvec[v].iov_len = sizeof(m->footer);
 456         con->out_kvec_bytes += sizeof(m->footer);
 457         con->out_kvec_left++;
 458         con->out_more = m->more_to_follow;
 459         con->out_msg_done = true;
 460 }
 461
 462 /*
 463  * Prepare headers for the next outgoing message.
 464  */
 465 static void prepare_write_message(struct ceph_connection *con)
 466 {
 467         struct ceph_msg *m;
 468         int v = 0;
 469
 470         con->out_kvec_bytes = 0;
 471         con->out_kvec_is_msg = true;
 472         con->out_msg_done = false;
 473
 474         /* Sneak an ack in there first?  If we can get it into the same
 475          * TCP packet that's a good thing. */
 476         if (con->in_seq > con->in_seq_acked) {
 477                 con->in_seq_acked = con->in_seq;
 478                 con->out_kvec[v].iov_base = &tag_ack;
 479                 con->out_kvec[v++].iov_len = 1;
 480                 con->out_temp_ack = cpu_to_le64(con->in_seq_acked);
 481                 con->out_kvec[v].iov_base = &con->out_temp_ack;
 482                 con->out_kvec[v++].iov_len = sizeof(con->out_temp_ack);
 483                 con->out_kvec_bytes = 1 + sizeof(con->out_temp_ack);
 484         }
 485
 486         m = list_first_entry(&con->out_queue,
 487                        struct ceph_msg, list_head);
 488         con->out_msg = m;
 489         if (test_bit(LOSSYTX, &con->state)) {
 490                 list_del_init(&m->list_head);
 491         } else {
 492                 /* put message on sent list */
 493                 ceph_msg_get(m);
 494                 list_move_tail(&m->list_head, &con->out_sent);
 495         }
 496
 497         /*
 498          * only assign outgoing seq # if we haven't sent this message
 499          * yet.  if it is requeued, resend with it's original seq.
 500          */
 501         if (m->needs_out_seq) {
 502                 m->hdr.seq = cpu_to_le64(++con->out_seq);
 503                 m->needs_out_seq = false;
 504         }
 505
 506         dout("prepare_write_message %p seq %lld type %d len %d+%d+%d %d pgs\n",
 507              m, con->out_seq, le16_to_cpu(m->hdr.type),
 508              le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len),
 509              le32_to_cpu(m->hdr.data_len),
 510              m->nr_pages);
 511         BUG_ON(le32_to_cpu(m->hdr.front_len) != m->front.iov_len);
 512
 513         /* tag + hdr + front + middle */
 514         con->out_kvec[v].iov_base = &tag_msg;
 515         con->out_kvec[v++].iov_len = 1;
 516         con->out_kvec[v].iov_base = &m->hdr;
 517         con->out_kvec[v++].iov_len = sizeof(m->hdr);
 518         con->out_kvec[v++] = m->front;
 519         if (m->middle)
 520                 con->out_kvec[v++] = m->middle->vec;
 521         con->out_kvec_left = v;
 522         con->out_kvec_bytes += 1 + sizeof(m->hdr) + m->front.iov_len +
 523                 (m->middle ? m->middle->vec.iov_len : 0);
 524         con->out_kvec_cur = con->out_kvec;
 525
 526         /* fill in crc (except data pages), footer */
 527         con->out_msg->hdr.crc =
 528                 cpu_to_le32(crc32c(0, (void *)&m->hdr,
 529                                       sizeof(m->hdr) - sizeof(m->hdr.crc)));
 530         con->out_msg->footer.flags = CEPH_MSG_FOOTER_COMPLETE;
 531         con->out_msg->footer.front_crc =
 532                 cpu_to_le32(crc32c(0, m->front.iov_base, m->front.iov_len));
 533         if (m->middle)
 534                 con->out_msg->footer.middle_crc =
 535                         cpu_to_le32(crc32c(0, m->middle->vec.iov_base,
 536                                            m->middle->vec.iov_len));
 537         else
 538                 con->out_msg->footer.middle_crc = 0;
 539         con->out_msg->footer.data_crc = 0;
 540         dout("prepare_write_message front_crc %u data_crc %u\n",
 541              le32_to_cpu(con->out_msg->footer.front_crc),
 542              le32_to_cpu(con->out_msg->footer.middle_crc));
 543
 544         /* is there a data payload? */
 545         if (le32_to_cpu(m->hdr.data_len) > 0) {
 546                 /* initialize page iterator */
 547                 con->out_msg_pos.page = 0;
 548                 if (m->pages)
 549                         con->out_msg_pos.page_pos = m->page_alignment;
 550                 else
 551                         con->out_msg_pos.page_pos = 0;
 552                 con->out_msg_pos.data_pos = 0;
 553                 con->out_msg_pos.did_page_crc = 0;
 554                 con->out_more = 1;  /* data + footer will follow */
 555         } else {
 556                 /* no, queue up footer too and be done */
 557                 prepare_write_message_footer(con, v);
 558         }
 559
 560         set_bit(WRITE_PENDING, &con->state);
 561 }
 562
 563 /*
 564  * Prepare an ack.
 565  */
 566 static void prepare_write_ack(struct ceph_connection *con)
 567 {
 568         dout("prepare_write_ack %p %llu -> %llu\n", con,
 569              con->in_seq_acked, con->in_seq);
 570         con->in_seq_acked = con->in_seq;
 571
 572         con->out_kvec[0].iov_base = &tag_ack;
 573         con->out_kvec[0].iov_len = 1;
 574         con->out_temp_ack = cpu_to_le64(con->in_seq_acked);
 575         con->out_kvec[1].iov_base = &con->out_temp_ack;
 576         con->out_kvec[1].iov_len = sizeof(con->out_temp_ack);
 577         con->out_kvec_left = 2;
 578         con->out_kvec_bytes = 1 + sizeof(con->out_temp_ack);
 579         con->out_kvec_cur = con->out_kvec;
 580         con->out_more = 1;  /* more will follow.. eventually.. */
 581         set_bit(WRITE_PENDING, &con->state);
 582 }
 583
 584 /*
 585  * Prepare to write keepalive byte.
 586  */
 587 static void prepare_write_keepalive(struct ceph_connection *con)
 588 {
 589         dout("prepare_write_keepalive %p\n", con);
 590         con->out_kvec[0].iov_base = &tag_keepalive;
 591         con->out_kvec[0].iov_len = 1;
 592         con->out_kvec_left = 1;
 593         con->out_kvec_bytes = 1;
 594         con->out_kvec_cur = con->out_kvec;
 595         set_bit(WRITE_PENDING, &con->state);
 596 }
 597
 598 /*
 599  * Connection negotiation.
 600  */
 601
 602 static int prepare_connect_authorizer(struct ceph_connection *con)
 603 {
 604         void *auth_buf;
 605         int auth_len = 0;
 606         int auth_protocol = 0;
 607
 608         mutex_unlock(&con->mutex);
 609         if (con->ops->get_authorizer)
 610                 con->ops->get_authorizer(con, &auth_buf, &auth_len,
 611                                          &auth_protocol, &con->auth_reply_buf,
 612                                          &con->auth_reply_buf_len,
 613                                          con->auth_retry);
 614         mutex_lock(&con->mutex);
 615
 616         if (test_bit(CLOSED, &con->state) ||
 617             test_bit(OPENING, &con->state))
 618                 return -EAGAIN;
 619
 620         con->out_connect.authorizer_protocol = cpu_to_le32(auth_protocol);
 621         con->out_connect.authorizer_len = cpu_to_le32(auth_len);
 622
 623         if (auth_len) {
 624                 con->out_kvec[con->out_kvec_left].iov_base = auth_buf;
 625                 con->out_kvec[con->out_kvec_left].iov_len = auth_len;
 626                 con->out_kvec_left++;
 627                 con->out_kvec_bytes += auth_len;
 628         }
 629         return 0;
 630 }
 631
 632 /*
 633  * We connected to a peer and are saying hello.
 634  */
 635 static void prepare_write_banner(struct ceph_messenger *msgr,
 636                                  struct ceph_connection *con)
 637 {
 638         int len = strlen(CEPH_BANNER);
 639
 640         con->out_kvec[0].iov_base = CEPH_BANNER;
 641         con->out_kvec[0].iov_len = len;
 642         con->out_kvec[1].iov_base = &msgr->my_enc_addr;
 643         con->out_kvec[1].iov_len = sizeof(msgr->my_enc_addr);
 644         con->out_kvec_left = 2;
 645         con->out_kvec_bytes = len + sizeof(msgr->my_enc_addr);
 646         con->out_kvec_cur = con->out_kvec;
 647         con->out_more = 0;
 648         set_bit(WRITE_PENDING, &con->state);
 649 }
 650
 651 static int prepare_write_connect(struct ceph_messenger *msgr,
 652                                  struct ceph_connection *con,
 653                                  int after_banner)
 654 {
 655         unsigned global_seq = get_global_seq(con->msgr, 0);
 656         int proto;
 657
 658         switch (con->peer_name.type) {
 659         case CEPH_ENTITY_TYPE_MON:
 660                 proto = CEPH_MONC_PROTOCOL;
 661                 break;
 662         case CEPH_ENTITY_TYPE_OSD:
 663                 proto = CEPH_OSDC_PROTOCOL;
 664                 break;
 665         case CEPH_ENTITY_TYPE_MDS:
 666                 proto = CEPH_MDSC_PROTOCOL;
 667                 break;
 668         default:
 669                 BUG();
 670         }
 671
 672         dout("prepare_write_connect %p cseq=%d gseq=%d proto=%d\n", con,
 673              con->connect_seq, global_seq, proto);
 674
 675         con->out_connect.features = cpu_to_le64(msgr->supported_features);
 676         con->out_connect.host_type = cpu_to_le32(CEPH_ENTITY_TYPE_CLIENT);
 677         con->out_connect.connect_seq = cpu_to_le32(con->connect_seq);
 678         con->out_connect.global_seq = cpu_to_le32(global_seq);
 679         con->out_connect.protocol_version = cpu_to_le32(proto);
 680         con->out_connect.flags = 0;
 681
 682         if (!after_banner) {
 683                 con->out_kvec_left = 0;
 684                 con->out_kvec_bytes = 0;
 685         }
 686         con->out_kvec[con->out_kvec_left].iov_base = &con->out_connect;
 687         con->out_kvec[con->out_kvec_left].iov_len = sizeof(con->out_connect);
 688         con->out_kvec_left++;
 689         con->out_kvec_bytes += sizeof(con->out_connect);
 690         con->out_kvec_cur = con->out_kvec;
 691         con->out_more = 0;
 692         set_bit(WRITE_PENDING, &con->state);
 693
 694         return prepare_connect_authorizer(con);
 695 }
 696
 697
 698 /*
 699  * write as much of pending kvecs to the socket as we can.
 700  *  1 -> done
 701  *  0 -> socket full, but more to do
 702  * <0 -> error
 703  */
 704 static int write_partial_kvec(struct ceph_connection *con)
 705 {
 706         int ret;
 707
 708         dout("write_partial_kvec %p %d left\n", con, con->out_kvec_bytes);
 709         while (con->out_kvec_bytes > 0) {
 710                 ret = ceph_tcp_sendmsg(con->sock, con->out_kvec_cur,
 711                                        con->out_kvec_left, con->out_kvec_bytes,
 712                                        con->out_more);
 713                 if (ret <= 0)
 714                         goto out;
 715                 con->out_kvec_bytes -= ret;
 716                 if (con->out_kvec_bytes == 0)
 717                         break;            /* done */
 718                 while (ret > 0) {
 719                         if (ret >= con->out_kvec_cur->iov_len) {
 720                                 ret -= con->out_kvec_cur->iov_len;
 721                                 con->out_kvec_cur++;
 722                                 con->out_kvec_left--;
 723                         } else {
 724                                 con->out_kvec_cur->iov_len -= ret;
 725                                 con->out_kvec_cur->iov_base += ret;
 726                                 ret = 0;
 727                                 break;
 728                         }
 729                 }
 730         }
 731         con->out_kvec_left = 0;
 732         con->out_kvec_is_msg = false;
 733         ret = 1;
 734 out:
 735         dout("write_partial_kvec %p %d left in %d kvecs ret = %d\n", con,
 736              con->out_kvec_bytes, con->out_kvec_left, ret);
 737         return ret;  /* done! */
 738 }
 739
 740 #ifdef CONFIG_BLOCK
 741 static void init_bio_iter(struct bio *bio, struct bio **iter, int *seg)
 742 {
 743         if (!bio) {
 744                 *iter = NULL;
 745                 *seg = 0;
 746                 return;
 747         }
 748         *iter = bio;
 749         *seg = bio->bi_idx;
 750 }
 751
 752 static void iter_bio_next(struct bio **bio_iter, int *seg)
 753 {
 754         if (*bio_iter == NULL)
 755                 return;
 756
 757         BUG_ON(*seg >= (*bio_iter)->bi_vcnt);
 758
 759         (*seg)++;
 760         if (*seg == (*bio_iter)->bi_vcnt)
 761                 init_bio_iter((*bio_iter)->bi_next, bio_iter, seg);
 762 }
 763 #endif
 764
 765 /*
 766  * Write as much message data payload as we can.  If we finish, queue
 767  * up the footer.
 768  *  1 -> done, footer is now queued in out_kvec[].
 769  *  0 -> socket full, but more to do
 770  * <0 -> error
 771  */
 772 static int write_partial_msg_pages(struct ceph_connection *con)
 773 {
 774         struct ceph_msg *msg = con->out_msg;
 775         unsigned data_len = le32_to_cpu(msg->hdr.data_len);
 776         size_t len;
 777         int crc = con->msgr->nocrc;
 778         int ret;
 779         int total_max_write;
 780         int in_trail = 0;
 781         size_t trail_len = (msg->trail ? msg->trail->length : 0);
 782
 783         dout("write_partial_msg_pages %p msg %p page %d/%d offset %d\n",
 784              con, con->out_msg, con->out_msg_pos.page, con->out_msg->nr_pages,
 785              con->out_msg_pos.page_pos);
 786
 787 #ifdef CONFIG_BLOCK
 788         if (msg->bio && !msg->bio_iter)
 789                 init_bio_iter(msg->bio, &msg->bio_iter, &msg->bio_seg);
 790 #endif
 791
 792         while (data_len > con->out_msg_pos.data_pos) {
 793                 struct page *page = NULL;
 794                 void *kaddr = NULL;
 795                 int max_write = PAGE_SIZE;
 796                 int page_shift = 0;
 797
 798                 total_max_write = data_len - trail_len -
 799                         con->out_msg_pos.data_pos;
 800
 801                 /*
 802                  * if we are calculating the data crc (the default), we need
 803                  * to map the page.  if our pages[] has been revoked, use the
 804                  * zero page.
 805                  */
 806
 807                 /* have we reached the trail part of the data? */
 808                 if (con->out_msg_pos.data_pos >= data_len - trail_len) {
 809                         in_trail = 1;
 810
 811                         total_max_write = data_len - con->out_msg_pos.data_pos;
 812
 813                         page = list_first_entry(&msg->trail->head,
 814                                                 struct page, lru);
 815                         if (crc)
 816                                 kaddr = kmap(page);
 817                         max_write = PAGE_SIZE;
 818                 } else if (msg->pages) {
 819                         page = msg->pages[con->out_msg_pos.page];
 820                         if (crc)
 821                                 kaddr = kmap(page);
 822                 } else if (msg->pagelist) {
 823                         page = list_first_entry(&msg->pagelist->head,
 824                                                 struct page, lru);
 825                         if (crc)
 826                                 kaddr = kmap(page);
 827 #ifdef CONFIG_BLOCK
 828                 } else if (msg->bio) {
 829                         struct bio_vec *bv;
 830
 831                         bv = bio_iovec_idx(msg->bio_iter, msg->bio_seg);
 832                         page = bv->bv_page;
 833                         page_shift = bv->bv_offset;
 834                         if (crc)
 835                                 kaddr = kmap(page) + page_shift;
 836                         max_write = bv->bv_len;
 837 #endif
 838                 } else {
 839                         page = con->msgr->zero_page;
 840                         if (crc)
 841                                 kaddr = page_address(con->msgr->zero_page);
 842                 }
 843                 len = min_t(int, max_write - con->out_msg_pos.page_pos,
 844                             total_max_write);
 845
 846                 if (crc && !con->out_msg_pos.did_page_crc) {
 847                         void *base = kaddr + con->out_msg_pos.page_pos;
 848                         u32 tmpcrc = le32_to_cpu(con->out_msg->footer.data_crc);
 849
 850                         BUG_ON(kaddr == NULL);
 851                         con->out_msg->footer.data_crc =
 852                                 cpu_to_le32(crc32c(tmpcrc, base, len));
 853                         con->out_msg_pos.did_page_crc = 1;
 854                 }
 855                 ret = kernel_sendpage(con->sock, page,
 856                                       con->out_msg_pos.page_pos + page_shift,
 857                                       len,
 858                                       MSG_DONTWAIT | MSG_NOSIGNAL |
 859                                       MSG_MORE);
 860
 861                 if (crc &&
 862                     (msg->pages || msg->pagelist || msg->bio || in_trail))
 863                         kunmap(page);
 864
 865                 if (ret == -EAGAIN)
 866                         ret = 0;
 867                 if (ret <= 0)
 868                         goto out;
 869
 870                 con->out_msg_pos.data_pos += ret;
 871                 con->out_msg_pos.page_pos += ret;
 872                 if (ret == len) {
 873                         con->out_msg_pos.page_pos = 0;
 874                         con->out_msg_pos.page++;
 875                         con->out_msg_pos.did_page_crc = 0;
 876                         if (in_trail)
 877                                 list_move_tail(&page->lru,
 878                                                &msg->trail->head);
 879                         else if (msg->pagelist)
 880                                 list_move_tail(&page->lru,
 881                                                &msg->pagelist->head);
 882 #ifdef CONFIG_BLOCK
 883                         else if (msg->bio)
 884                                 iter_bio_next(&msg->bio_iter, &msg->bio_seg);
 885 #endif
 886                 }
 887         }
 888
 889         dout("write_partial_msg_pages %p msg %p done\n", con, msg);
 890
 891         /* prepare and queue up footer, too */
 892         if (!crc)
 893                 con->out_msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC;
 894         con->out_kvec_bytes = 0;
 895         con->out_kvec_left = 0;
 896         con->out_kvec_cur = con->out_kvec;
 897         prepare_write_message_footer(con, 0);
 898         ret = 1;
 899 out:
 900         return ret;
 901 }
 902
 903 /*
 904  * write some zeros
 905  */
 906 static int write_partial_skip(struct ceph_connection *con)
 907 {
 908         int ret;
 909
 910         while (con->out_skip > 0) {
 911                 struct kvec iov = {
 912                         .iov_base = page_address(con->msgr->zero_page),
 913                         .iov_len = min(con->out_skip, (int)PAGE_CACHE_SIZE)
 914                 };
 915
 916                 ret = ceph_tcp_sendmsg(con->sock, &iov, 1, iov.iov_len, 1);
 917                 if (ret <= 0)
 918                         goto out;
 919                 con->out_skip -= ret;
 920         }
 921         ret = 1;
 922 out:
 923         return ret;
 924 }
 925
 926 /*
 927  * Prepare to read connection handshake, or an ack.
 928  */
 929 static void prepare_read_banner(struct ceph_connection *con)
 930 {
 931         dout("prepare_read_banner %p\n", con);
 932         con->in_base_pos = 0;
 933 }
 934
 935 static void prepare_read_connect(struct ceph_connection *con)
 936 {
 937         dout("prepare_read_connect %p\n", con);
 938         con->in_base_pos = 0;
 939 }
 940
 941 static void prepare_read_ack(struct ceph_connection *con)
 942 {
 943         dout("prepare_read_ack %p\n", con);
 944         con->in_base_pos = 0;
 945 }
 946
 947 static void prepare_read_tag(struct ceph_connection *con)
 948 {
 949         dout("prepare_read_tag %p\n", con);
 950         con->in_base_pos = 0;
 951         con->in_tag = CEPH_MSGR_TAG_READY;
 952 }
 953
 954 /*
 955  * Prepare to read a message.
 956  */
 957 static int prepare_read_message(struct ceph_connection *con)
 958 {
 959         dout("prepare_read_message %p\n", con);
 960         BUG_ON(con->in_msg != NULL);
 961         con->in_base_pos = 0;
 962         con->in_front_crc = con->in_middle_crc = con->in_data_crc = 0;
 963         return 0;
 964 }
 965
 966
 967 static int read_partial(struct ceph_connection *con,
 968                         int *to, int size, void *object)
 969 {
 970         *to += size;
 971         while (con->in_base_pos < *to) {
 972                 int left = *to - con->in_base_pos;
 973                 int have = size - left;
 974                 int ret = ceph_tcp_recvmsg(con->sock, object + have, left);
 975                 if (ret <= 0)
 976                         return ret;
 977                 con->in_base_pos += ret;
 978         }
 979         return 1;
 980 }
 981
 982
 983 /*
 984  * Read all or part of the connect-side handshake on a new connection
 985  */
 986 static int read_partial_banner(struct ceph_connection *con)
 987 {
 988         int ret, to = 0;
 989
 990         dout("read_partial_banner %p at %d\n", con, con->in_base_pos);
 991
 992         /* peer's banner */
 993         ret = read_partial(con, &to, strlen(CEPH_BANNER), con->in_banner);
 994         if (ret <= 0)
 995                 goto out;
 996         ret = read_partial(con, &to, sizeof(con->actual_peer_addr),
 997                            &con->actual_peer_addr);
 998         if (ret <= 0)
 999                 goto out;
1000         ret = read_partial(con, &to, sizeof(con->peer_addr_for_me),
1001                            &con->peer_addr_for_me);
1002         if (ret <= 0)
1003                 goto out;
1004 out:
1005         return ret;
1006 }
1007
1008 static int read_partial_connect(struct ceph_connection *con)
1009 {
1010         int ret, to = 0;
1011
1012         dout("read_partial_connect %p at %d\n", con, con->in_base_pos);
1013
1014         ret = read_partial(con, &to, sizeof(con->in_reply), &con->in_reply);
1015         if (ret <= 0)
1016                 goto out;
1017         ret = read_partial(con, &to, le32_to_cpu(con->in_reply.authorizer_len),
1018                            con->auth_reply_buf);
1019         if (ret <= 0)
1020                 goto out;
1021
1022         dout("read_partial_connect %p tag %d, con_seq = %u, g_seq = %u\n",
1023              con, (int)con->in_reply.tag,
1024              le32_to_cpu(con->in_reply.connect_seq),
1025              le32_to_cpu(con->in_reply.global_seq));
1026 out:
1027         return ret;
1028
1029 }
1030
1031 /*
1032  * Verify the hello banner looks okay.
1033  */
1034 static int verify_hello(struct ceph_connection *con)
1035 {
1036         if (memcmp(con->in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) {
1037                 pr_err("connect to %s got bad banner\n",
1038                        ceph_pr_addr(&con->peer_addr.in_addr));
1039                 con->error_msg = "protocol error, bad banner";
1040                 return -1;
1041         }
1042         return 0;
1043 }
1044
1045 static bool addr_is_blank(struct sockaddr_storage *ss)
1046 {
1047         switch (ss->ss_family) {
1048         case AF_INET:
1049                 return ((struct sockaddr_in *)ss)->sin_addr.s_addr == 0;
1050         case AF_INET6:
1051                 return
1052                      ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[0] == 0 &&
1053                      ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[1] == 0 &&
1054                      ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[2] == 0 &&
1055                      ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[3] == 0;
1056         }
1057         return false;
1058 }
1059
1060 static int addr_port(struct sockaddr_storage *ss)
1061 {
1062         switch (ss->ss_family) {
1063         case AF_INET:
1064                 return ntohs(((struct sockaddr_in *)ss)->sin_port);
1065         case AF_INET6:
1066                 return ntohs(((struct sockaddr_in6 *)ss)->sin6_port);
1067         }
1068         return 0;
1069 }
1070
1071 static void addr_set_port(struct sockaddr_storage *ss, int p)
1072 {
1073         switch (ss->ss_family) {
1074         case AF_INET:
1075                 ((struct sockaddr_in *)ss)->sin_port = htons(p);
1076                 break;
1077         case AF_INET6:
1078                 ((struct sockaddr_in6 *)ss)->sin6_port = htons(p);
1079                 break;
1080         }
1081 }
1082
1083 /*
1084  * Parse an ip[:port] list into an addr array.  Use the default
1085  * monitor port if a port isn't specified.
1086  */
1087 int ceph_parse_ips(const char *c, const char *end,
1088                    struct ceph_entity_addr *addr,
1089                    int max_count, int *count)
1090 {
1091         int i;
1092         const char *p = c;
1093
1094         dout("parse_ips on '%.*s'\n", (int)(end-c), c);
1095         for (i = 0; i < max_count; i++) {
1096                 const char *ipend;
1097                 struct sockaddr_storage *ss = &addr[i].in_addr;
1098                 struct sockaddr_in *in4 = (void *)ss;
1099                 struct sockaddr_in6 *in6 = (void *)ss;
1100                 int port;
1101                 char delim = ',';
1102
1103                 if (*p == '[') {
1104                         delim = ']';
1105                         p++;
1106                 }
1107
1108                 memset(ss, 0, sizeof(*ss));
1109                 if (in4_pton(p, end - p, (u8 *)&in4->sin_addr.s_addr,
1110                              delim, &ipend))
1111                         ss->ss_family = AF_INET;
1112                 else if (in6_pton(p, end - p, (u8 *)&in6->sin6_addr.s6_addr,
1113                                   delim, &ipend))
1114                         ss->ss_family = AF_INET6;
1115                 else
1116                         goto bad;
1117                 p = ipend;
1118
1119                 if (delim == ']') {
1120                         if (*p != ']') {
1121                                 dout("missing matching ']'\n");
1122                                 goto bad;
1123                         }
1124                         p++;
1125                 }
1126
1127                 /* port? */
1128                 if (p < end && *p == ':') {
1129                         port = 0;
1130                         p++;
1131                         while (p < end && *p >= '0' && *p <= '9') {
1132                                 port = (port * 10) + (*p - '0');
1133                                 p++;
1134                         }
1135                         if (port > 65535 || port == 0)
1136                                 goto bad;
1137                 } else {
1138                         port = CEPH_MON_PORT;
1139                 }
1140
1141                 addr_set_port(ss, port);
1142
1143                 dout("parse_ips got %s\n", ceph_pr_addr(ss));
1144
1145                 if (p == end)
1146                         break;
1147                 if (*p != ',')
1148                         goto bad;
1149                 p++;
1150         }
1151
1152         if (p != end)
1153                 goto bad;
1154
1155         if (count)
1156                 *count = i + 1;
1157         return 0;
1158
1159 bad:
1160         pr_err("parse_ips bad ip '%.*s'\n", (int)(end - c), c);
1161         return -EINVAL;
1162 }
1163 EXPORT_SYMBOL(ceph_parse_ips);
1164
1165 static int process_banner(struct ceph_connection *con)
1166 {
1167         dout("process_banner on %p\n", con);
1168
1169         if (verify_hello(con) < 0)
1170                 return -1;
1171
1172         ceph_decode_addr(&con->actual_peer_addr);
1173         ceph_decode_addr(&con->peer_addr_for_me);
1174
1175         /*
1176          * Make sure the other end is who we wanted.  note that the other
1177          * end may not yet know their ip address, so if it's 0.0.0.0, give
1178          * them the benefit of the doubt.
1179          */
1180         if (memcmp(&con->peer_addr, &con->actual_peer_addr,
1181                    sizeof(con->peer_addr)) != 0 &&
1182             !(addr_is_blank(&con->actual_peer_addr.in_addr) &&
1183               con->actual_peer_addr.nonce == con->peer_addr.nonce)) {
1184                 pr_warning("wrong peer, want %s/%d, got %s/%d\n",
1185                            ceph_pr_addr(&con->peer_addr.in_addr),
1186                            (int)le32_to_cpu(con->peer_addr.nonce),
1187                            ceph_pr_addr(&con->actual_peer_addr.in_addr),
1188                            (int)le32_to_cpu(con->actual_peer_addr.nonce));
1189                 con->error_msg = "wrong peer at address";
1190                 return -1;
1191         }
1192
1193         /*
1194          * did we learn our address?
1195          */
1196         if (addr_is_blank(&con->msgr->inst.addr.in_addr)) {
1197                 int port = addr_port(&con->msgr->inst.addr.in_addr);
1198
1199                 memcpy(&con->msgr->inst.addr.in_addr,
1200                        &con->peer_addr_for_me.in_addr,
1201                        sizeof(con->peer_addr_for_me.in_addr));
1202                 addr_set_port(&con->msgr->inst.addr.in_addr, port);
1203                 encode_my_addr(con->msgr);
1204                 dout("process_banner learned my addr is %s\n",
1205                      ceph_pr_addr(&con->msgr->inst.addr.in_addr));
1206         }
1207
1208         set_bit(NEGOTIATING, &con->state);
1209         prepare_read_connect(con);
1210         return 0;
1211 }
1212
1213 static void fail_protocol(struct ceph_connection *con)
1214 {
1215         reset_connection(con);
1216         set_bit(CLOSED, &con->state);  /* in case there's queued work */
1217
1218         mutex_unlock(&con->mutex);
1219         if (con->ops->bad_proto)
1220                 con->ops->bad_proto(con);
1221         mutex_lock(&con->mutex);
1222 }
1223
1224 static int process_connect(struct ceph_connection *con)
1225 {
1226         u64 sup_feat = con->msgr->supported_features;
1227         u64 req_feat = con->msgr->required_features;
1228         u64 server_feat = le64_to_cpu(con->in_reply.features);
1229         int ret;
1230
1231         dout("process_connect on %p tag %d\n", con, (int)con->in_tag);
1232
1233         switch (con->in_reply.tag) {
1234         case CEPH_MSGR_TAG_FEATURES:
1235                 pr_err("%s%lld %s feature set mismatch,"
1236                        " my %llx < server's %llx, missing %llx\n",
1237                        ENTITY_NAME(con->peer_name),
1238                        ceph_pr_addr(&con->peer_addr.in_addr),
1239                        sup_feat, server_feat, server_feat & ~sup_feat);
1240                 con->error_msg = "missing required protocol features";
1241                 fail_protocol(con);
1242                 return -1;
1243
1244         case CEPH_MSGR_TAG_BADPROTOVER:
1245                 pr_err("%s%lld %s protocol version mismatch,"
1246                        " my %d != server's %d\n",
1247                        ENTITY_NAME(con->peer_name),
1248                        ceph_pr_addr(&con->peer_addr.in_addr),
1249                        le32_to_cpu(con->out_connect.protocol_version),
1250                        le32_to_cpu(con->in_reply.protocol_version));
1251                 con->error_msg = "protocol version mismatch";
1252                 fail_protocol(con);
1253                 return -1;
1254
1255         case CEPH_MSGR_TAG_BADAUTHORIZER:
1256                 con->auth_retry++;
1257                 dout("process_connect %p got BADAUTHORIZER attempt %d\n", con,
1258                      con->auth_retry);
1259                 if (con->auth_retry == 2) {
1260                         con->error_msg = "connect authorization failure";
1261                         return -1;
1262                 }
1263                 con->auth_retry = 1;
1264                 ret = prepare_write_connect(con->msgr, con, 0);
1265                 if (ret < 0)
1266                         return ret;
1267                 prepare_read_connect(con);
1268                 break;
1269
1270         case CEPH_MSGR_TAG_RESETSESSION:
1271                 /*
1272                  * If we connected with a large connect_seq but the peer
1273                  * has no record of a session with us (no connection, or
1274                  * connect_seq == 0), they will send RESETSESION to indicate
1275                  * that they must have reset their session, and may have
1276                  * dropped messages.
1277                  */
1278                 dout("process_connect got RESET peer seq %u\n",
1279                      le32_to_cpu(con->in_connect.connect_seq));
1280                 pr_err("%s%lld %s connection reset\n",
1281                        ENTITY_NAME(con->peer_name),
1282                        ceph_pr_addr(&con->peer_addr.in_addr));
1283                 reset_connection(con);
1284                 prepare_write_connect(con->msgr, con, 0);
1285                 prepare_read_connect(con);
1286
1287                 /* Tell ceph about it. */
1288                 mutex_unlock(&con->mutex);
1289                 pr_info("reset on %s%lld\n", ENTITY_NAME(con->peer_name));
1290                 if (con->ops->peer_reset)
1291                         con->ops->peer_reset(con);
1292                 mutex_lock(&con->mutex);
1293                 if (test_bit(CLOSED, &con->state) ||
1294                     test_bit(OPENING, &con->state))
1295                         return -EAGAIN;
1296                 break;
1297
1298         case CEPH_MSGR_TAG_RETRY_SESSION:
1299                 /*
1300                  * If we sent a smaller connect_seq than the peer has, try
1301                  * again with a larger value.
1302                  */
1303                 dout("process_connect got RETRY my seq = %u, peer_seq = %u\n",
1304                      le32_to_cpu(con->out_connect.connect_seq),
1305                      le32_to_cpu(con->in_connect.connect_seq));
1306                 con->connect_seq = le32_to_cpu(con->in_connect.connect_seq);
1307                 prepare_write_connect(con->msgr, con, 0);
1308                 prepare_read_connect(con);
1309                 break;
1310
1311         case CEPH_MSGR_TAG_RETRY_GLOBAL:
1312                 /*
1313                  * If we sent a smaller global_seq than the peer has, try
1314                  * again with a larger value.
1315                  */
1316                 dout("process_connect got RETRY_GLOBAL my %u peer_gseq %u\n",
1317                      con->peer_global_seq,
1318                      le32_to_cpu(con->in_connect.global_seq));
1319                 get_global_seq(con->msgr,
1320                                le32_to_cpu(con->in_connect.global_seq));
1321                 prepare_write_connect(con->msgr, con, 0);
1322                 prepare_read_connect(con);
1323                 break;
1324
1325         case CEPH_MSGR_TAG_READY:
1326                 if (req_feat & ~server_feat) {
1327                         pr_err("%s%lld %s protocol feature mismatch,"
1328                                " my required %llx > server's %llx, need %llx\n",
1329                                ENTITY_NAME(con->peer_name),
1330                                ceph_pr_addr(&con->peer_addr.in_addr),
1331                                req_feat, server_feat, req_feat & ~server_feat);
1332                         con->error_msg = "missing required protocol features";
1333                         fail_protocol(con);
1334                         return -1;
1335                 }
1336                 clear_bit(CONNECTING, &con->state);
1337                 con->peer_global_seq = le32_to_cpu(con->in_reply.global_seq);
1338                 con->connect_seq++;
1339                 con->peer_features = server_feat;
1340                 dout("process_connect got READY gseq %d cseq %d (%d)\n",
1341                      con->peer_global_seq,
1342                      le32_to_cpu(con->in_reply.connect_seq),
1343                      con->connect_seq);
1344                 WARN_ON(con->connect_seq !=
1345                         le32_to_cpu(con->in_reply.connect_seq));
1346
1347                 if (con->in_reply.flags & CEPH_MSG_CONNECT_LOSSY)
1348                         set_bit(LOSSYTX, &con->state);
1349
1350                 prepare_read_tag(con);
1351                 break;
1352
1353         case CEPH_MSGR_TAG_WAIT:
1354                 /*
1355                  * If there is a connection race (we are opening
1356                  * connections to each other), one of us may just have
1357                  * to WAIT.  This shouldn't happen if we are the
1358                  * client.
1359                  */
1360                 pr_err("process_connect got WAIT as client\n");
1361                 con->error_msg = "protocol error, got WAIT as client";
1362                 return -1;
1363
1364         default:
1365                 pr_err("connect protocol error, will retry\n");
1366                 con->error_msg = "protocol error, garbage tag during connect";
1367                 return -1;
1368         }
1369         return 0;
1370 }
1371
1372
1373 /*
1374  * read (part of) an ack
1375  */
1376 static int read_partial_ack(struct ceph_connection *con)
1377 {
1378         int to = 0;
1379
1380         return read_partial(con, &to, sizeof(con->in_temp_ack),
1381                             &con->in_temp_ack);
1382 }
1383
1384
1385 /*
1386  * We can finally discard anything that's been acked.
1387  */
1388 static void process_ack(struct ceph_connection *con)
1389 {
1390         struct ceph_msg *m;
1391         u64 ack = le64_to_cpu(con->in_temp_ack);
1392         u64 seq;
1393
1394         while (!list_empty(&con->out_sent)) {
1395                 m = list_first_entry(&con->out_sent, struct ceph_msg,
1396                                      list_head);
1397                 seq = le64_to_cpu(m->hdr.seq);
1398                 if (seq > ack)
1399                         break;
1400                 dout("got ack for seq %llu type %d at %p\n", seq,
1401                      le16_to_cpu(m->hdr.type), m);
1402                 ceph_msg_remove(m);
1403         }
1404         prepare_read_tag(con);
1405 }
1406
1407
1408
1409
1410 static int read_partial_message_section(struct ceph_connection *con,
1411                                         struct kvec *section,
1412                                         unsigned int sec_len, u32 *crc)
1413 {
1414         int ret, left;
1415
1416         BUG_ON(!section);
1417
1418         while (section->iov_len < sec_len) {
1419                 BUG_ON(section->iov_base == NULL);
1420                 left = sec_len - section->iov_len;
1421                 ret = ceph_tcp_recvmsg(con->sock, (char *)section->iov_base +
1422                                        section->iov_len, left);
1423                 if (ret <= 0)
1424                         return ret;
1425                 section->iov_len += ret;
1426                 if (section->iov_len == sec_len)
1427                         *crc = crc32c(0, section->iov_base,
1428                                       section->iov_len);
1429         }
1430
1431         return 1;
1432 }
1433
1434 static struct ceph_msg *ceph_alloc_msg(struct ceph_connection *con,
1435                                 struct ceph_msg_header *hdr,
1436                                 int *skip);
1437
1438
1439 static int read_partial_message_pages(struct ceph_connection *con,
1440                                       struct page **pages,
1441                                       unsigned data_len, int datacrc)
1442 {
1443         void *p;
1444         int ret;
1445         int left;
1446
1447         left = min((int)(data_len - con->in_msg_pos.data_pos),
1448                    (int)(PAGE_SIZE - con->in_msg_pos.page_pos));
1449         /* (page) data */
1450         BUG_ON(pages == NULL);
1451         p = kmap(pages[con->in_msg_pos.page]);
1452         ret = ceph_tcp_recvmsg(con->sock, p + con->in_msg_pos.page_pos,
1453                                left);
1454         if (ret > 0 && datacrc)
1455                 con->in_data_crc =
1456                         crc32c(con->in_data_crc,
1457                                   p + con->in_msg_pos.page_pos, ret);
1458         kunmap(pages[con->in_msg_pos.page]);
1459         if (ret <= 0)
1460                 return ret;
1461         con->in_msg_pos.data_pos += ret;
1462         con->in_msg_pos.page_pos += ret;
1463         if (con->in_msg_pos.page_pos == PAGE_SIZE) {
1464                 con->in_msg_pos.page_pos = 0;
1465                 con->in_msg_pos.page++;
1466         }
1467
1468         return ret;
1469 }
1470
1471 #ifdef CONFIG_BLOCK
1472 static int read_partial_message_bio(struct ceph_connection *con,
1473                                     struct bio **bio_iter, int *bio_seg,
1474                                     unsigned data_len, int datacrc)
1475 {
1476         struct bio_vec *bv = bio_iovec_idx(*bio_iter, *bio_seg);
1477         void *p;
1478         int ret, left;
1479
1480         if (IS_ERR(bv))
1481                 return PTR_ERR(bv);
1482
1483         left = min((int)(data_len - con->in_msg_pos.data_pos),
1484                    (int)(bv->bv_len - con->in_msg_pos.page_pos));
1485
1486         p = kmap(bv->bv_page) + bv->bv_offset;
1487
1488         ret = ceph_tcp_recvmsg(con->sock, p + con->in_msg_pos.page_pos,
1489                                left);
1490         if (ret > 0 && datacrc)
1491                 con->in_data_crc =
1492                         crc32c(con->in_data_crc,
1493                                   p + con->in_msg_pos.page_pos, ret);
1494         kunmap(bv->bv_page);
1495         if (ret <= 0)
1496                 return ret;
1497         con->in_msg_pos.data_pos += ret;
1498         con->in_msg_pos.page_pos += ret;
1499         if (con->in_msg_pos.page_pos == bv->bv_len) {
1500                 con->in_msg_pos.page_pos = 0;
1501                 iter_bio_next(bio_iter, bio_seg);
1502         }
1503
1504         return ret;
1505 }
1506 #endif
1507
1508 /*
1509  * read (part of) a message.
1510  */
1511 static int read_partial_message(struct ceph_connection *con)
1512 {
1513         struct ceph_msg *m = con->in_msg;
1514         int ret;
1515         int to, left;
1516         unsigned front_len, middle_len, data_len;
1517         int datacrc = con->msgr->nocrc;
1518         int skip;
1519         u64 seq;
1520
1521         dout("read_partial_message con %p msg %p\n", con, m);
1522
1523         /* header */
1524         while (con->in_base_pos < sizeof(con->in_hdr)) {
1525                 left = sizeof(con->in_hdr) - con->in_base_pos;
1526                 ret = ceph_tcp_recvmsg(con->sock,
1527                                        (char *)&con->in_hdr + con->in_base_pos,
1528                                        left);
1529                 if (ret <= 0)
1530                         return ret;
1531                 con->in_base_pos += ret;
1532                 if (con->in_base_pos == sizeof(con->in_hdr)) {
1533                         u32 crc = crc32c(0, (void *)&con->in_hdr,
1534                                  sizeof(con->in_hdr) - sizeof(con->in_hdr.crc));
1535                         if (crc != le32_to_cpu(con->in_hdr.crc)) {
1536                                 pr_err("read_partial_message bad hdr "
1537                                        " crc %u != expected %u\n",
1538                                        crc, con->in_hdr.crc);
1539                                 return -EBADMSG;
1540                         }
1541                 }
1542         }
1543         front_len = le32_to_cpu(con->in_hdr.front_len);
1544         if (front_len > CEPH_MSG_MAX_FRONT_LEN)
1545                 return -EIO;
1546         middle_len = le32_to_cpu(con->in_hdr.middle_len);
1547         if (middle_len > CEPH_MSG_MAX_DATA_LEN)
1548                 return -EIO;
1549         data_len = le32_to_cpu(con->in_hdr.data_len);
1550         if (data_len > CEPH_MSG_MAX_DATA_LEN)
1551                 return -EIO;
1552
1553         /* verify seq# */
1554         seq = le64_to_cpu(con->in_hdr.seq);
1555         if ((s64)seq - (s64)con->in_seq < 1) {
1556                 pr_info("skipping %s%lld %s seq %lld expected %lld\n",
1557                         ENTITY_NAME(con->peer_name),
1558                         ceph_pr_addr(&con->peer_addr.in_addr),
1559                         seq, con->in_seq + 1);
1560                 con->in_base_pos = -front_len - middle_len - data_len -
1561                         sizeof(m->footer);
1562                 con->in_tag = CEPH_MSGR_TAG_READY;
1563                 return 0;
1564         } else if ((s64)seq - (s64)con->in_seq > 1) {
1565                 pr_err("read_partial_message bad seq %lld expected %lld\n",
1566                        seq, con->in_seq + 1);
1567                 con->error_msg = "bad message sequence # for incoming message";
1568                 return -EBADMSG;
1569         }
1570
1571         /* allocate message? */
1572         if (!con->in_msg) {
1573                 dout("got hdr type %d front %d data %d\n", con->in_hdr.type,
1574                      con->in_hdr.front_len, con->in_hdr.data_len);
1575                 skip = 0;
1576                 con->in_msg = ceph_alloc_msg(con, &con->in_hdr, &skip);
1577                 if (skip) {
1578                         /* skip this message */
1579                         dout("alloc_msg said skip message\n");
1580                         BUG_ON(con->in_msg);
1581                         con->in_base_pos = -front_len - middle_len - data_len -
1582                                 sizeof(m->footer);
1583                         con->in_tag = CEPH_MSGR_TAG_READY;
1584                         con->in_seq++;
1585                         return 0;
1586                 }
1587                 if (!con->in_msg) {
1588                         con->error_msg =
1589                                 "error allocating memory for incoming message";
1590                         return -ENOMEM;
1591                 }
1592                 m = con->in_msg;
1593                 m->front.iov_len = 0;    /* haven't read it yet */
1594                 if (m->middle)
1595                         m->middle->vec.iov_len = 0;
1596
1597                 con->in_msg_pos.page = 0;
1598                 if (m->pages)
1599                         con->in_msg_pos.page_pos = m->page_alignment;
1600                 else
1601                         con->in_msg_pos.page_pos = 0;
1602                 con->in_msg_pos.data_pos = 0;
1603         }
1604
1605         /* front */
1606         ret = read_partial_message_section(con, &m->front, front_len,
1607                                            &con->in_front_crc);
1608         if (ret <= 0)
1609                 return ret;
1610
1611         /* middle */
1612         if (m->middle) {
1613                 ret = read_partial_message_section(con, &m->middle->vec,
1614                                                    middle_len,
1615                                                    &con->in_middle_crc);
1616                 if (ret <= 0)
1617                         return ret;
1618         }
1619 #ifdef CONFIG_BLOCK
1620         if (m->bio && !m->bio_iter)
1621                 init_bio_iter(m->bio, &m->bio_iter, &m->bio_seg);
1622 #endif
1623
1624         /* (page) data */
1625         while (con->in_msg_pos.data_pos < data_len) {
1626                 if (m->pages) {
1627                         ret = read_partial_message_pages(con, m->pages,
1628                                                  data_len, datacrc);
1629                         if (ret <= 0)
1630                                 return ret;
1631 #ifdef CONFIG_BLOCK
1632                 } else if (m->bio) {
1633
1634                         ret = read_partial_message_bio(con,
1635                                                  &m->bio_iter, &m->bio_seg,
1636                                                  data_len, datacrc);
1637                         if (ret <= 0)
1638                                 return ret;
1639 #endif
1640                 } else {
1641                         BUG_ON(1);
1642                 }
1643         }
1644
1645         /* footer */
1646         to = sizeof(m->hdr) + sizeof(m->footer);
1647         while (con->in_base_pos < to) {
1648                 left = to - con->in_base_pos;
1649                 ret = ceph_tcp_recvmsg(con->sock, (char *)&m->footer +
1650                                        (con->in_base_pos - sizeof(m->hdr)),
1651                                        left);
1652                 if (ret <= 0)
1653                         return ret;
1654                 con->in_base_pos += ret;
1655         }
1656         dout("read_partial_message got msg %p %d (%u) + %d (%u) + %d (%u)\n",
1657              m, front_len, m->footer.front_crc, middle_len,
1658              m->footer.middle_crc, data_len, m->footer.data_crc);
1659
1660         /* crc ok? */
1661         if (con->in_front_crc != le32_to_cpu(m->footer.front_crc)) {
1662                 pr_err("read_partial_message %p front crc %u != exp. %u\n",
1663                        m, con->in_front_crc, m->footer.front_crc);
1664                 return -EBADMSG;
1665         }
1666         if (con->in_middle_crc != le32_to_cpu(m->footer.middle_crc)) {
1667                 pr_err("read_partial_message %p middle crc %u != exp %u\n",
1668                        m, con->in_middle_crc, m->footer.middle_crc);
1669                 return -EBADMSG;
1670         }
1671         if (datacrc &&
1672             (m->footer.flags & CEPH_MSG_FOOTER_NOCRC) == 0 &&
1673             con->in_data_crc != le32_to_cpu(m->footer.data_crc)) {
1674                 pr_err("read_partial_message %p data crc %u != exp. %u\n", m,
1675                        con->in_data_crc, le32_to_cpu(m->footer.data_crc));
1676                 return -EBADMSG;
1677         }
1678
1679         return 1; /* done! */
1680 }
1681
1682 /*
1683  * Process message.  This happens in the worker thread.  The callback should
1684  * be careful not to do anything that waits on other incoming messages or it
1685  * may deadlock.
1686  */
1687 static void process_message(struct ceph_connection *con)
1688 {
1689         struct ceph_msg *msg;
1690
1691         msg = con->in_msg;
1692         con->in_msg = NULL;
1693
1694         /* if first message, set peer_name */
1695         if (con->peer_name.type == 0)
1696                 con->peer_name = msg->hdr.src;
1697
1698         con->in_seq++;
1699         mutex_unlock(&con->mutex);
1700
1701         dout("===== %p %llu from %s%lld %d=%s len %d+%d (%u %u %u) =====\n",
1702              msg, le64_to_cpu(msg->hdr.seq),
1703              ENTITY_NAME(msg->hdr.src),
1704              le16_to_cpu(msg->hdr.type),
1705              ceph_msg_type_name(le16_to_cpu(msg->hdr.type)),
1706              le32_to_cpu(msg->hdr.front_len),
1707              le32_to_cpu(msg->hdr.data_len),
1708              con->in_front_crc, con->in_middle_crc, con->in_data_crc);
1709         con->ops->dispatch(con, msg);
1710
1711         mutex_lock(&con->mutex);
1712         prepare_read_tag(con);
1713 }
1714
1715
1716 /*
1717  * Write something to the socket.  Called in a worker thread when the
1718  * socket appears to be writeable and we have something ready to send.
1719  */
1720 static int try_write(struct ceph_connection *con)
1721 {
1722         struct ceph_messenger *msgr = con->msgr;
1723         int ret = 1;
1724
1725         dout("try_write start %p state %lu nref %d\n", con, con->state,
1726              atomic_read(&con->nref));
1727
1728 more:
1729         dout("try_write out_kvec_bytes %d\n", con->out_kvec_bytes);
1730
1731         /* open the socket first? */
1732         if (con->sock == NULL) {
1733                 prepare_write_banner(msgr, con);
1734                 prepare_write_connect(msgr, con, 1);
1735                 prepare_read_banner(con);
1736                 set_bit(CONNECTING, &con->state);
1737                 clear_bit(NEGOTIATING, &con->state);
1738
1739                 BUG_ON(con->in_msg);
1740                 con->in_tag = CEPH_MSGR_TAG_READY;
1741                 dout("try_write initiating connect on %p new state %lu\n",
1742                      con, con->state);
1743                 con->sock = ceph_tcp_connect(con);
1744                 if (IS_ERR(con->sock)) {
1745                         con->sock = NULL;
1746                         con->error_msg = "connect error";
1747                         ret = -1;
1748                         goto out;
1749                 }
1750         }
1751
1752 more_kvec:
1753         /* kvec data queued? */
1754         if (con->out_skip) {
1755                 ret = write_partial_skip(con);
1756                 if (ret <= 0)
1757                         goto out;
1758         }
1759         if (con->out_kvec_left) {
1760                 ret = write_partial_kvec(con);
1761                 if (ret <= 0)
1762                         goto out;
1763         }
1764
1765         /* msg pages? */
1766         if (con->out_msg) {
1767                 if (con->out_msg_done) {
1768                         ceph_msg_put(con->out_msg);
1769                         con->out_msg = NULL;   /* we're done with this one */
1770                         goto do_next;
1771                 }
1772
1773                 ret = write_partial_msg_pages(con);
1774                 if (ret == 1)
1775                         goto more_kvec;  /* we need to send the footer, too! */
1776                 if (ret == 0)
1777                         goto out;
1778                 if (ret < 0) {
1779                         dout("try_write write_partial_msg_pages err %d\n",
1780                              ret);
1781                         goto out;
1782                 }
1783         }
1784
1785 do_next:
1786         if (!test_bit(CONNECTING, &con->state)) {
1787                 /* is anything else pending? */
1788                 if (!list_empty(&con->out_queue)) {
1789                         prepare_write_message(con);
1790                         goto more;
1791                 }
1792                 if (con->in_seq > con->in_seq_acked) {
1793                         prepare_write_ack(con);
1794                         goto more;
1795                 }
1796                 if (test_and_clear_bit(KEEPALIVE_PENDING, &con->state)) {
1797                         prepare_write_keepalive(con);
1798                         goto more;
1799                 }
1800         }
1801
1802         /* Nothing to do! */
1803         clear_bit(WRITE_PENDING, &con->state);
1804         dout("try_write nothing else to write.\n");
1805         ret = 0;
1806 out:
1807         dout("try_write done on %p ret %d\n", con, ret);
1808         return ret;
1809 }
1810
1811
1812
1813 /*
1814  * Read what we can from the socket.
1815  */
1816 static int try_read(struct ceph_connection *con)
1817 {
1818         int ret = -1;
1819
1820         if (!con->sock)
1821                 return 0;
1822
1823         if (test_bit(STANDBY, &con->state))
1824                 return 0;
1825
1826         dout("try_read start on %p\n", con);
1827
1828 more:
1829         dout("try_read tag %d in_base_pos %d\n", (int)con->in_tag,
1830              con->in_base_pos);
1831
1832         /*
1833          * process_connect and process_message drop and re-take
1834          * con->mutex.  make sure we handle a racing close or reopen.
1835          */
1836         if (test_bit(CLOSED, &con->state) ||
1837             test_bit(OPENING, &con->state)) {
1838                 ret = -EAGAIN;
1839                 goto out;
1840         }
1841
1842         if (test_bit(CONNECTING, &con->state)) {
1843                 if (!test_bit(NEGOTIATING, &con->state)) {
1844                         dout("try_read connecting\n");
1845                         ret = read_partial_banner(con);
1846                         if (ret <= 0)
1847                                 goto out;
1848                         ret = process_banner(con);
1849                         if (ret < 0)
1850                                 goto out;
1851                 }
1852                 ret = read_partial_connect(con);
1853                 if (ret <= 0)
1854                         goto out;
1855                 ret = process_connect(con);
1856                 if (ret < 0)
1857                         goto out;
1858                 goto more;
1859         }
1860
1861         if (con->in_base_pos < 0) {
1862                 /*
1863                  * skipping + discarding content.
1864                  *
1865                  * FIXME: there must be a better way to do this!
1866                  */
1867                 static char buf[1024];
1868                 int skip = min(1024, -con->in_base_pos);
1869                 dout("skipping %d / %d bytes\n", skip, -con->in_base_pos);
1870                 ret = ceph_tcp_recvmsg(con->sock, buf, skip);
1871                 if (ret <= 0)
1872                         goto out;
1873                 con->in_base_pos += ret;
1874                 if (con->in_base_pos)
1875                         goto more;
1876         }
1877         if (con->in_tag == CEPH_MSGR_TAG_READY) {
1878                 /*
1879                  * what's next?
1880                  */
1881                 ret = ceph_tcp_recvmsg(con->sock, &con->in_tag, 1);
1882                 if (ret <= 0)
1883                         goto out;
1884                 dout("try_read got tag %d\n", (int)con->in_tag);
1885                 switch (con->in_tag) {
1886                 case CEPH_MSGR_TAG_MSG:
1887                         prepare_read_message(con);
1888                         break;
1889                 case CEPH_MSGR_TAG_ACK:
1890                         prepare_read_ack(con);
1891                         break;
1892                 case CEPH_MSGR_TAG_CLOSE:
1893                         set_bit(CLOSED, &con->state);   /* fixme */
1894                         goto out;
1895                 default:
1896                         goto bad_tag;
1897                 }
1898         }
1899         if (con->in_tag == CEPH_MSGR_TAG_MSG) {
1900                 ret = read_partial_message(con);
1901                 if (ret <= 0) {
1902                         switch (ret) {
1903                         case -EBADMSG:
1904                                 con->error_msg = "bad crc";
1905                                 ret = -EIO;
1906                                 break;
1907                         case -EIO:
1908                                 con->error_msg = "io error";
1909                                 break;
1910                         }
1911                         goto out;
1912                 }
1913                 if (con->in_tag == CEPH_MSGR_TAG_READY)
1914                         goto more;
1915                 process_message(con);
1916                 goto more;
1917         }
1918         if (con->in_tag == CEPH_MSGR_TAG_ACK) {
1919                 ret = read_partial_ack(con);
1920                 if (ret <= 0)
1921                         goto out;
1922                 process_ack(con);
1923                 goto more;
1924         }
1925
1926 out:
1927         dout("try_read done on %p ret %d\n", con, ret);
1928         return ret;
1929
1930 bad_tag:
1931         pr_err("try_read bad con->in_tag = %d\n", (int)con->in_tag);
1932         con->error_msg = "protocol error, garbage tag";
1933         ret = -1;
1934         goto out;
1935 }
1936
1937
1938 /*
1939  * Atomically queue work on a connection.  Bump @con reference to
1940  * avoid races with connection teardown.
1941  */
1942 static void queue_con(struct ceph_connection *con)
1943 {
1944         if (test_bit(DEAD, &con->state)) {
1945                 dout("queue_con %p ignoring: DEAD\n",
1946                      con);
1947                 return;
1948         }
1949
1950         if (!con->ops->get(con)) {
1951                 dout("queue_con %p ref count 0\n", con);
1952                 return;
1953         }
1954
1955         if (!queue_delayed_work(ceph_msgr_wq, &con->work, 0)) {
1956                 dout("queue_con %p - already queued\n", con);
1957                 con->ops->put(con);
1958         } else {
1959                 dout("queue_con %p\n", con);
1960         }
1961 }
1962
1963 /*
1964  * Do some work on a connection.  Drop a connection ref when we're done.
1965  */
1966 static void con_work(struct work_struct *work)
1967 {
1968         struct ceph_connection *con = container_of(work, struct ceph_connection,
1969                                                    work.work);
1970         int ret;
1971
1972         mutex_lock(&con->mutex);
1973 restart:
1974         if (test_and_clear_bit(BACKOFF, &con->state)) {
1975                 dout("con_work %p backing off\n", con);
1976                 if (queue_delayed_work(ceph_msgr_wq, &con->work,
1977                                        round_jiffies_relative(con->delay))) {
1978                         dout("con_work %p backoff %lu\n", con, con->delay);
1979                         mutex_unlock(&con->mutex);
1980                         return;
1981                 } else {
1982                         con->ops->put(con);
1983                         dout("con_work %p FAILED to back off %lu\n", con,
1984                              con->delay);
1985                 }
1986         }
1987
1988         if (test_bit(STANDBY, &con->state)) {
1989                 dout("con_work %p STANDBY\n", con);
1990                 goto done;
1991         }
1992         if (test_bit(CLOSED, &con->state)) { /* e.g. if we are replaced */
1993                 dout("con_work CLOSED\n");
1994                 con_close_socket(con);
1995                 goto done;
1996         }
1997         if (test_and_clear_bit(OPENING, &con->state)) {
1998                 /* reopen w/ new peer */
1999                 dout("con_work OPENING\n");
2000                 con_close_socket(con);
2001         }
2002
2003         if (test_and_clear_bit(SOCK_CLOSED, &con->state))
2004                 goto fault;
2005
2006         ret = try_read(con);
2007         if (ret == -EAGAIN)
2008                 goto restart;
2009         if (ret < 0)
2010                 goto fault;
2011
2012         ret = try_write(con);
2013         if (ret == -EAGAIN)
2014                 goto restart;
2015         if (ret < 0)
2016                 goto fault;
2017
2018 done:
2019         mutex_unlock(&con->mutex);
2020 done_unlocked:
2021         con->ops->put(con);
2022         return;
2023
2024 fault:
2025         mutex_unlock(&con->mutex);
2026         ceph_fault(con);     /* error/fault path */
2027         goto done_unlocked;
2028 }
2029
2030
2031 /*
2032  * Generic error/fault handler.  A retry mechanism is used with
2033  * exponential backoff
2034  */
2035 static void ceph_fault(struct ceph_connection *con)
2036 {
2037         pr_err("%s%lld %s %s\n", ENTITY_NAME(con->peer_name),
2038                ceph_pr_addr(&con->peer_addr.in_addr), con->error_msg);
2039         dout("fault %p state %lu to peer %s\n",
2040              con, con->state, ceph_pr_addr(&con->peer_addr.in_addr));
2041
2042         if (test_bit(LOSSYTX, &con->state)) {
2043                 dout("fault on LOSSYTX channel\n");
2044                 goto out;
2045         }
2046
2047         mutex_lock(&con->mutex);
2048         if (test_bit(CLOSED, &con->state))
2049                 goto out_unlock;
2050
2051         con_close_socket(con);
2052
2053         if (con->in_msg) {
2054                 ceph_msg_put(con->in_msg);
2055                 con->in_msg = NULL;
2056         }
2057
2058         /* Requeue anything that hasn't been acked */
2059         list_splice_init(&con->out_sent, &con->out_queue);
2060
2061         /* If there are no messages queued or keepalive pending, place
2062          * the connection in a STANDBY state */
2063         if (list_empty(&con->out_queue) &&
2064             !test_bit(KEEPALIVE_PENDING, &con->state)) {
2065                 dout("fault %p setting STANDBY clearing WRITE_PENDING\n", con);
2066                 clear_bit(WRITE_PENDING, &con->state);
2067                 set_bit(STANDBY, &con->state);
2068         } else {
2069                 /* retry after a delay. */
2070                 if (con->delay == 0)
2071                         con->delay = BASE_DELAY_INTERVAL;
2072                 else if (con->delay < MAX_DELAY_INTERVAL)
2073                         con->delay *= 2;
2074                 con->ops->get(con);
2075                 if (queue_delayed_work(ceph_msgr_wq, &con->work,
2076                                        round_jiffies_relative(con->delay))) {
2077                         dout("fault queued %p delay %lu\n", con, con->delay);
2078                 } else {
2079                         con->ops->put(con);
2080                         dout("fault failed to queue %p delay %lu, backoff\n",
2081                              con, con->delay);
2082                         /*
2083                          * In many cases we see a socket state change
2084                          * while con_work is running and end up
2085                          * queuing (non-delayed) work, such that we
2086                          * can't backoff with a delay.  Set a flag so
2087                          * that when con_work restarts we schedule the
2088                          * delay then.
2089                          */
2090                         set_bit(BACKOFF, &con->state);
2091                 }
2092         }
2093
2094 out_unlock:
2095         mutex_unlock(&con->mutex);
2096 out:
2097         /*
2098          * in case we faulted due to authentication, invalidate our
2099          * current tickets so that we can get new ones.
2100          */
2101         if (con->auth_retry && con->ops->invalidate_authorizer) {
2102                 dout("calling invalidate_authorizer()\n");
2103                 con->ops->invalidate_authorizer(con);
2104         }
2105
2106         if (con->ops->fault)
2107                 con->ops->fault(con);
2108 }
2109
2110
2111
2112 /*
2113  * create a new messenger instance
2114  */
2115 struct ceph_messenger *ceph_messenger_create(struct ceph_entity_addr *myaddr,
2116                                              u32 supported_features,
2117                                              u32 required_features)
2118 {
2119         struct ceph_messenger *msgr;
2120
2121         msgr = kzalloc(sizeof(*msgr), GFP_KERNEL);
2122         if (msgr == NULL)
2123                 return ERR_PTR(-ENOMEM);
2124
2125         msgr->supported_features = supported_features;
2126         msgr->required_features = required_features;
2127
2128         spin_lock_init(&msgr->global_seq_lock);
2129
2130         /* the zero page is needed if a request is "canceled" while the message
2131          * is being written over the socket */
2132         msgr->zero_page = __page_cache_alloc(GFP_KERNEL | __GFP_ZERO);
2133         if (!msgr->zero_page) {
2134                 kfree(msgr);
2135                 return ERR_PTR(-ENOMEM);
2136         }
2137         kmap(msgr->zero_page);
2138
2139         if (myaddr)
2140                 msgr->inst.addr = *myaddr;
2141
2142         /* select a random nonce */
2143         msgr->inst.addr.type = 0;
2144         get_random_bytes(&msgr->inst.addr.nonce, sizeof(msgr->inst.addr.nonce));
2145         encode_my_addr(msgr);
2146
2147         dout("messenger_create %p\n", msgr);
2148         return msgr;
2149 }
2150 EXPORT_SYMBOL(ceph_messenger_create);
2151
2152 void ceph_messenger_destroy(struct ceph_messenger *msgr)
2153 {
2154         dout("destroy %p\n", msgr);
2155         kunmap(msgr->zero_page);
2156         __free_page(msgr->zero_page);
2157         kfree(msgr);
2158         dout("destroyed messenger %p\n", msgr);
2159 }
2160 EXPORT_SYMBOL(ceph_messenger_destroy);
2161
2162 static void clear_standby(struct ceph_connection *con)
2163 {
2164         /* come back from STANDBY? */
2165         if (test_and_clear_bit(STANDBY, &con->state)) {
2166                 mutex_lock(&con->mutex);
2167                 dout("clear_standby %p and ++connect_seq\n", con);
2168                 con->connect_seq++;
2169                 WARN_ON(test_bit(WRITE_PENDING, &con->state));
2170                 WARN_ON(test_bit(KEEPALIVE_PENDING, &con->state));
2171                 mutex_unlock(&con->mutex);
2172         }
2173 }
2174
2175 /*
2176  * Queue up an outgoing message on the given connection.
2177  */
2178 void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg)
2179 {
2180         if (test_bit(CLOSED, &con->state)) {
2181                 dout("con_send %p closed, dropping %p\n", con, msg);
2182                 ceph_msg_put(msg);
2183                 return;
2184         }
2185
2186         /* set src+dst */
2187         msg->hdr.src = con->msgr->inst.name;
2188
2189         BUG_ON(msg->front.iov_len != le32_to_cpu(msg->hdr.front_len));
2190
2191         msg->needs_out_seq = true;
2192
2193         /* queue */
2194         mutex_lock(&con->mutex);
2195         BUG_ON(!list_empty(&msg->list_head));
2196         list_add_tail(&msg->list_head, &con->out_queue);
2197         dout("----- %p to %s%lld %d=%s len %d+%d+%d -----\n", msg,
2198              ENTITY_NAME(con->peer_name), le16_to_cpu(msg->hdr.type),
2199              ceph_msg_type_name(le16_to_cpu(msg->hdr.type)),
2200              le32_to_cpu(msg->hdr.front_len),
2201              le32_to_cpu(msg->hdr.middle_len),
2202              le32_to_cpu(msg->hdr.data_len));
2203         mutex_unlock(&con->mutex);
2204
2205         /* if there wasn't anything waiting to send before, queue
2206          * new work */
2207         clear_standby(con);
2208         if (test_and_set_bit(WRITE_PENDING, &con->state) == 0)
2209                 queue_con(con);
2210 }
2211 EXPORT_SYMBOL(ceph_con_send);
2212
2213 /*
2214  * Revoke a message that was previously queued for send
2215  */
2216 void ceph_con_revoke(struct ceph_connection *con, struct ceph_msg *msg)
2217 {
2218         mutex_lock(&con->mutex);
2219         if (!list_empty(&msg->list_head)) {
2220                 dout("con_revoke %p msg %p - was on queue\n", con, msg);
2221                 list_del_init(&msg->list_head);
2222                 ceph_msg_put(msg);
2223                 msg->hdr.seq = 0;
2224         }
2225         if (con->out_msg == msg) {
2226                 dout("con_revoke %p msg %p - was sending\n", con, msg);
2227                 con->out_msg = NULL;
2228                 if (con->out_kvec_is_msg) {
2229                         con->out_skip = con->out_kvec_bytes;
2230                         con->out_kvec_is_msg = false;
2231                 }
2232                 ceph_msg_put(msg);
2233                 msg->hdr.seq = 0;
2234         }
2235         mutex_unlock(&con->mutex);
2236 }
2237
2238 /*
2239  * Revoke a message that we may be reading data into
2240  */
2241 void ceph_con_revoke_message(struct ceph_connection *con, struct ceph_msg *msg)
2242 {
2243         mutex_lock(&con->mutex);
2244         if (con->in_msg && con->in_msg == msg) {
2245                 unsigned front_len = le32_to_cpu(con->in_hdr.front_len);
2246                 unsigned middle_len = le32_to_cpu(con->in_hdr.middle_len);
2247                 unsigned data_len = le32_to_cpu(con->in_hdr.data_len);
2248
2249                 /* skip rest of message */
2250                 dout("con_revoke_pages %p msg %p revoked\n", con, msg);
2251                         con->in_base_pos = con->in_base_pos -
2252                                 sizeof(struct ceph_msg_header) -
2253                                 front_len -
2254                                 middle_len -
2255                                 data_len -
2256                                 sizeof(struct ceph_msg_footer);
2257                 ceph_msg_put(con->in_msg);
2258                 con->in_msg = NULL;
2259                 con->in_tag = CEPH_MSGR_TAG_READY;
2260                 con->in_seq++;
2261         } else {
2262                 dout("con_revoke_pages %p msg %p pages %p no-op\n",
2263                      con, con->in_msg, msg);
2264         }
2265         mutex_unlock(&con->mutex);
2266 }
2267
2268 /*
2269  * Queue a keepalive byte to ensure the tcp connection is alive.
2270  */
2271 void ceph_con_keepalive(struct ceph_connection *con)
2272 {
2273         dout("con_keepalive %p\n", con);
2274         clear_standby(con);
2275         if (test_and_set_bit(KEEPALIVE_PENDING, &con->state) == 0 &&
2276             test_and_set_bit(WRITE_PENDING, &con->state) == 0)
2277                 queue_con(con);
2278 }
2279 EXPORT_SYMBOL(ceph_con_keepalive);
2280
2281
2282 /*
2283  * construct a new message with given type, size
2284  * the new msg has a ref count of 1.
2285  */
2286 struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags)
2287 {
2288         struct ceph_msg *m;
2289
2290         m = kmalloc(sizeof(*m), flags);
2291         if (m == NULL)
2292                 goto out;
2293         kref_init(&m->kref);
2294         INIT_LIST_HEAD(&m->list_head);
2295
2296         m->hdr.tid = 0;
2297         m->hdr.type = cpu_to_le16(type);
2298         m->hdr.priority = cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);
2299         m->hdr.version = 0;
2300         m->hdr.front_len = cpu_to_le32(front_len);
2301         m->hdr.middle_len = 0;
2302         m->hdr.data_len = 0;
2303         m->hdr.data_off = 0;
2304         m->hdr.reserved = 0;
2305         m->footer.front_crc = 0;
2306         m->footer.middle_crc = 0;
2307         m->footer.data_crc = 0;
2308         m->footer.flags = 0;
2309         m->front_max = front_len;
2310         m->front_is_vmalloc = false;
2311         m->more_to_follow = false;
2312         m->pool = NULL;
2313
2314         /* middle */
2315         m->middle = NULL;
2316
2317         /* data */
2318         m->nr_pages = 0;
2319         m->page_alignment = 0;
2320         m->pages = NULL;
2321         m->pagelist = NULL;
2322         m->bio = NULL;
2323         m->bio_iter = NULL;
2324         m->bio_seg = 0;
2325         m->trail = NULL;
2326
2327         /* front */
2328         if (front_len) {
2329                 if (front_len > PAGE_CACHE_SIZE) {
2330                         m->front.iov_base = __vmalloc(front_len, flags,
2331                                                       PAGE_KERNEL);
2332                         m->front_is_vmalloc = true;
2333                 } else {
2334                         m->front.iov_base = kmalloc(front_len, flags);
2335                 }
2336                 if (m->front.iov_base == NULL) {
2337                         pr_err("msg_new can't allocate %d bytes\n",
2338                              front_len);
2339                         goto out2;
2340                 }
2341         } else {
2342                 m->front.iov_base = NULL;
2343         }
2344         m->front.iov_len = front_len;
2345
2346         dout("ceph_msg_new %p front %d\n", m, front_len);
2347         return m;
2348
2349 out2:
2350         ceph_msg_put(m);
2351 out:
2352         pr_err("msg_new can't create type %d front %d\n", type, front_len);
2353         return NULL;
2354 }
2355 EXPORT_SYMBOL(ceph_msg_new);
2356
2357 /*
2358  * Allocate "middle" portion of a message, if it is needed and wasn't
2359  * allocated by alloc_msg.  This allows us to read a small fixed-size
2360  * per-type header in the front and then gracefully fail (i.e.,
2361  * propagate the error to the caller based on info in the front) when
2362  * the middle is too large.
2363  */
2364 static int ceph_alloc_middle(struct ceph_connection *con, struct ceph_msg *msg)
2365 {
2366         int type = le16_to_cpu(msg->hdr.type);
2367         int middle_len = le32_to_cpu(msg->hdr.middle_len);
2368
2369         dout("alloc_middle %p type %d %s middle_len %d\n", msg, type,
2370              ceph_msg_type_name(type), middle_len);
2371         BUG_ON(!middle_len);
2372         BUG_ON(msg->middle);
2373
2374         msg->middle = ceph_buffer_new(middle_len, GFP_NOFS);
2375         if (!msg->middle)
2376                 return -ENOMEM;
2377         return 0;
2378 }
2379
2380 /*
2381  * Generic message allocator, for incoming messages.
2382  */
2383 static struct ceph_msg *ceph_alloc_msg(struct ceph_connection *con,
2384                                 struct ceph_msg_header *hdr,
2385                                 int *skip)
2386 {
2387         int type = le16_to_cpu(hdr->type);
2388         int front_len = le32_to_cpu(hdr->front_len);
2389         int middle_len = le32_to_cpu(hdr->middle_len);
2390         struct ceph_msg *msg = NULL;
2391         int ret;
2392
2393         if (con->ops->alloc_msg) {
2394                 mutex_unlock(&con->mutex);
2395                 msg = con->ops->alloc_msg(con, hdr, skip);
2396                 mutex_lock(&con->mutex);
2397                 if (!msg || *skip)
2398                         return NULL;
2399         }
2400         if (!msg) {
2401                 *skip = 0;
2402                 msg = ceph_msg_new(type, front_len, GFP_NOFS);
2403                 if (!msg) {
2404                         pr_err("unable to allocate msg type %d len %d\n",
2405                                type, front_len);
2406                         return NULL;
2407                 }
2408                 msg->page_alignment = le16_to_cpu(hdr->data_off);
2409         }
2410         memcpy(&msg->hdr, &con->in_hdr, sizeof(con->in_hdr));
2411
2412         if (middle_len && !msg->middle) {
2413                 ret = ceph_alloc_middle(con, msg);
2414                 if (ret < 0) {
2415                         ceph_msg_put(msg);
2416                         return NULL;
2417                 }
2418         }
2419
2420         return msg;
2421 }
2422
2423
2424 /*
2425  * Free a generically kmalloc'd message.
2426  */
2427 void ceph_msg_kfree(struct ceph_msg *m)
2428 {
2429         dout("msg_kfree %p\n", m);
2430         if (m->front_is_vmalloc)
2431                 vfree(m->front.iov_base);
2432         else
2433                 kfree(m->front.iov_base);
2434         kfree(m);
2435 }
2436
2437 /*
2438  * Drop a msg ref.  Destroy as needed.
2439  */
2440 void ceph_msg_last_put(struct kref *kref)
2441 {
2442         struct ceph_msg *m = container_of(kref, struct ceph_msg, kref);
2443
2444         dout("ceph_msg_put last one on %p\n", m);
2445         WARN_ON(!list_empty(&m->list_head));
2446
2447         /* drop middle, data, if any */
2448         if (m->middle) {
2449                 ceph_buffer_put(m->middle);
2450                 m->middle = NULL;
2451         }
2452         m->nr_pages = 0;
2453         m->pages = NULL;
2454
2455         if (m->pagelist) {
2456                 ceph_pagelist_release(m->pagelist);
2457                 kfree(m->pagelist);
2458                 m->pagelist = NULL;
2459         }
2460
2461         m->trail = NULL;
2462
2463         if (m->pool)
2464                 ceph_msgpool_put(m->pool, m);
2465         else
2466                 ceph_msg_kfree(m);
2467 }
2468 EXPORT_SYMBOL(ceph_msg_last_put);
2469
2470 void ceph_msg_dump(struct ceph_msg *msg)
2471 {
2472         pr_debug("msg_dump %p (front_max %d nr_pages %d)\n", msg,
2473                  msg->front_max, msg->nr_pages);
2474         print_hex_dump(KERN_DEBUG, "header: ",
2475                        DUMP_PREFIX_OFFSET, 16, 1,
2476                        &msg->hdr, sizeof(msg->hdr), true);
2477         print_hex_dump(KERN_DEBUG, " front: ",
2478                        DUMP_PREFIX_OFFSET, 16, 1,
2479                        msg->front.iov_base, msg->front.iov_len, true);
2480         if (msg->middle)
2481                 print_hex_dump(KERN_DEBUG, "middle: ",
2482                                DUMP_PREFIX_OFFSET, 16, 1,
2483                                msg->middle->vec.iov_base,
2484                                msg->middle->vec.iov_len, true);
2485         print_hex_dump(KERN_DEBUG, "footer: ",
2486                        DUMP_PREFIX_OFFSET, 16, 1,
2487                        &msg->footer, sizeof(msg->footer), true);
2488 }
2489 EXPORT_SYMBOL(ceph_msg_dump);