net/ceph/messenger_v1.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 #include <linux/ceph/ceph_debug.h>
   3
   4 #include <linux/bvec.h>
   5 #include <linux/crc32c.h>
   6 #include <linux/net.h>
   7 #include <linux/socket.h>
   8 #include <net/sock.h>
   9
  10 #include <linux/ceph/ceph_features.h>
  11 #include <linux/ceph/decode.h>
  12 #include <linux/ceph/libceph.h>
  13 #include <linux/ceph/messenger.h>
  14
  15 /* static tag bytes (protocol control messages) */
  16 static char tag_msg = CEPH_MSGR_TAG_MSG;
  17 static char tag_ack = CEPH_MSGR_TAG_ACK;
  18 static char tag_keepalive = CEPH_MSGR_TAG_KEEPALIVE;
  19 static char tag_keepalive2 = CEPH_MSGR_TAG_KEEPALIVE2;
  20
  21 /*
  22  * If @buf is NULL, discard up to @len bytes.
  23  */
  24 static int ceph_tcp_recvmsg(struct socket *sock, void *buf, size_t len)
  25 {
  26         struct kvec iov = {buf, len};
  27         struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
  28         int r;
  29
  30         if (!buf)
  31                 msg.msg_flags |= MSG_TRUNC;
  32
  33         iov_iter_kvec(&msg.msg_iter, ITER_DEST, &iov, 1, len);
  34         r = sock_recvmsg(sock, &msg, msg.msg_flags);
  35         if (r == -EAGAIN)
  36                 r = 0;
  37         return r;
  38 }
  39
  40 static int ceph_tcp_recvpage(struct socket *sock, struct page *page,
  41                      int page_offset, size_t length)
  42 {
  43         struct bio_vec bvec;
  44         struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
  45         int r;
  46
  47         BUG_ON(page_offset + length > PAGE_SIZE);
  48         bvec_set_page(&bvec, page, length, page_offset);
  49         iov_iter_bvec(&msg.msg_iter, ITER_DEST, &bvec, 1, length);
  50         r = sock_recvmsg(sock, &msg, msg.msg_flags);
  51         if (r == -EAGAIN)
  52                 r = 0;
  53         return r;
  54 }
  55
  56 /*
  57  * write something.  @more is true if caller will be sending more data
  58  * shortly.
  59  */
  60 static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov,
  61                             size_t kvlen, size_t len, bool more)
  62 {
  63         struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
  64         int r;
  65
  66         if (more)
  67                 msg.msg_flags |= MSG_MORE;
  68         else
  69                 msg.msg_flags |= MSG_EOR;  /* superfluous, but what the hell */
  70
  71         r = kernel_sendmsg(sock, &msg, iov, kvlen, len);
  72         if (r == -EAGAIN)
  73                 r = 0;
  74         return r;
  75 }
  76
  77 /*
  78  * @more: either or both of MSG_MORE and MSG_SENDPAGE_NOTLAST
  79  */
  80 static int ceph_tcp_sendpage(struct socket *sock, struct page *page,
  81                              int offset, size_t size, int more)
  82 {
  83         ssize_t (*sendpage)(struct socket *sock, struct page *page,
  84                             int offset, size_t size, int flags);
  85         int flags = MSG_DONTWAIT | MSG_NOSIGNAL | more;
  86         int ret;
  87
  88         /*
  89          * sendpage cannot properly handle pages with page_count == 0,
  90          * we need to fall back to sendmsg if that's the case.
  91          *
  92          * Same goes for slab pages: skb_can_coalesce() allows
  93          * coalescing neighboring slab objects into a single frag which
  94          * triggers one of hardened usercopy checks.
  95          */
  96         if (sendpage_ok(page))
  97                 sendpage = sock->ops->sendpage;
  98         else
  99                 sendpage = sock_no_sendpage;
 100
 101         ret = sendpage(sock, page, offset, size, flags);
 102         if (ret == -EAGAIN)
 103                 ret = 0;
 104
 105         return ret;
 106 }
 107
 108 static void con_out_kvec_reset(struct ceph_connection *con)
 109 {
 110         BUG_ON(con->v1.out_skip);
 111
 112         con->v1.out_kvec_left = 0;
 113         con->v1.out_kvec_bytes = 0;
 114         con->v1.out_kvec_cur = &con->v1.out_kvec[0];
 115 }
 116
 117 static void con_out_kvec_add(struct ceph_connection *con,
 118                                 size_t size, void *data)
 119 {
 120         int index = con->v1.out_kvec_left;
 121
 122         BUG_ON(con->v1.out_skip);
 123         BUG_ON(index >= ARRAY_SIZE(con->v1.out_kvec));
 124
 125         con->v1.out_kvec[index].iov_len = size;
 126         con->v1.out_kvec[index].iov_base = data;
 127         con->v1.out_kvec_left++;
 128         con->v1.out_kvec_bytes += size;
 129 }
 130
 131 /*
 132  * Chop off a kvec from the end.  Return residual number of bytes for
 133  * that kvec, i.e. how many bytes would have been written if the kvec
 134  * hadn't been nuked.
 135  */
 136 static int con_out_kvec_skip(struct ceph_connection *con)
 137 {
 138         int skip = 0;
 139
 140         if (con->v1.out_kvec_bytes > 0) {
 141                 skip = con->v1.out_kvec_cur[con->v1.out_kvec_left - 1].iov_len;
 142                 BUG_ON(con->v1.out_kvec_bytes < skip);
 143                 BUG_ON(!con->v1.out_kvec_left);
 144                 con->v1.out_kvec_bytes -= skip;
 145                 con->v1.out_kvec_left--;
 146         }
 147
 148         return skip;
 149 }
 150
 151 static size_t sizeof_footer(struct ceph_connection *con)
 152 {
 153         return (con->peer_features & CEPH_FEATURE_MSG_AUTH) ?
 154             sizeof(struct ceph_msg_footer) :
 155             sizeof(struct ceph_msg_footer_old);
 156 }
 157
 158 static void prepare_message_data(struct ceph_msg *msg, u32 data_len)
 159 {
 160         /* Initialize data cursor */
 161
 162         ceph_msg_data_cursor_init(&msg->cursor, msg, data_len);
 163 }
 164
 165 /*
 166  * Prepare footer for currently outgoing message, and finish things
 167  * off.  Assumes out_kvec* are already valid.. we just add on to the end.
 168  */
 169 static void prepare_write_message_footer(struct ceph_connection *con)
 170 {
 171         struct ceph_msg *m = con->out_msg;
 172
 173         m->footer.flags |= CEPH_MSG_FOOTER_COMPLETE;
 174
 175         dout("prepare_write_message_footer %p\n", con);
 176         con_out_kvec_add(con, sizeof_footer(con), &m->footer);
 177         if (con->peer_features & CEPH_FEATURE_MSG_AUTH) {
 178                 if (con->ops->sign_message)
 179                         con->ops->sign_message(m);
 180                 else
 181                         m->footer.sig = 0;
 182         } else {
 183                 m->old_footer.flags = m->footer.flags;
 184         }
 185         con->v1.out_more = m->more_to_follow;
 186         con->v1.out_msg_done = true;
 187 }
 188
 189 /*
 190  * Prepare headers for the next outgoing message.
 191  */
 192 static void prepare_write_message(struct ceph_connection *con)
 193 {
 194         struct ceph_msg *m;
 195         u32 crc;
 196
 197         con_out_kvec_reset(con);
 198         con->v1.out_msg_done = false;
 199
 200         /* Sneak an ack in there first?  If we can get it into the same
 201          * TCP packet that's a good thing. */
 202         if (con->in_seq > con->in_seq_acked) {
 203                 con->in_seq_acked = con->in_seq;
 204                 con_out_kvec_add(con, sizeof (tag_ack), &tag_ack);
 205                 con->v1.out_temp_ack = cpu_to_le64(con->in_seq_acked);
 206                 con_out_kvec_add(con, sizeof(con->v1.out_temp_ack),
 207                         &con->v1.out_temp_ack);
 208         }
 209
 210         ceph_con_get_out_msg(con);
 211         m = con->out_msg;
 212
 213         dout("prepare_write_message %p seq %lld type %d len %d+%d+%zd\n",
 214              m, con->out_seq, le16_to_cpu(m->hdr.type),
 215              le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len),
 216              m->data_length);
 217         WARN_ON(m->front.iov_len != le32_to_cpu(m->hdr.front_len));
 218         WARN_ON(m->data_length != le32_to_cpu(m->hdr.data_len));
 219
 220         /* tag + hdr + front + middle */
 221         con_out_kvec_add(con, sizeof (tag_msg), &tag_msg);
 222         con_out_kvec_add(con, sizeof(con->v1.out_hdr), &con->v1.out_hdr);
 223         con_out_kvec_add(con, m->front.iov_len, m->front.iov_base);
 224
 225         if (m->middle)
 226                 con_out_kvec_add(con, m->middle->vec.iov_len,
 227                         m->middle->vec.iov_base);
 228
 229         /* fill in hdr crc and finalize hdr */
 230         crc = crc32c(0, &m->hdr, offsetof(struct ceph_msg_header, crc));
 231         con->out_msg->hdr.crc = cpu_to_le32(crc);
 232         memcpy(&con->v1.out_hdr, &con->out_msg->hdr, sizeof(con->v1.out_hdr));
 233
 234         /* fill in front and middle crc, footer */
 235         crc = crc32c(0, m->front.iov_base, m->front.iov_len);
 236         con->out_msg->footer.front_crc = cpu_to_le32(crc);
 237         if (m->middle) {
 238                 crc = crc32c(0, m->middle->vec.iov_base,
 239                                 m->middle->vec.iov_len);
 240                 con->out_msg->footer.middle_crc = cpu_to_le32(crc);
 241         } else
 242                 con->out_msg->footer.middle_crc = 0;
 243         dout("%s front_crc %u middle_crc %u\n", __func__,
 244              le32_to_cpu(con->out_msg->footer.front_crc),
 245              le32_to_cpu(con->out_msg->footer.middle_crc));
 246         con->out_msg->footer.flags = 0;
 247
 248         /* is there a data payload? */
 249         con->out_msg->footer.data_crc = 0;
 250         if (m->data_length) {
 251                 prepare_message_data(con->out_msg, m->data_length);
 252                 con->v1.out_more = 1;  /* data + footer will follow */
 253         } else {
 254                 /* no, queue up footer too and be done */
 255                 prepare_write_message_footer(con);
 256         }
 257
 258         ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
 259 }
 260
 261 /*
 262  * Prepare an ack.
 263  */
 264 static void prepare_write_ack(struct ceph_connection *con)
 265 {
 266         dout("prepare_write_ack %p %llu -> %llu\n", con,
 267              con->in_seq_acked, con->in_seq);
 268         con->in_seq_acked = con->in_seq;
 269
 270         con_out_kvec_reset(con);
 271
 272         con_out_kvec_add(con, sizeof (tag_ack), &tag_ack);
 273
 274         con->v1.out_temp_ack = cpu_to_le64(con->in_seq_acked);
 275         con_out_kvec_add(con, sizeof(con->v1.out_temp_ack),
 276                          &con->v1.out_temp_ack);
 277
 278         con->v1.out_more = 1;  /* more will follow.. eventually.. */
 279         ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
 280 }
 281
 282 /*
 283  * Prepare to share the seq during handshake
 284  */
 285 static void prepare_write_seq(struct ceph_connection *con)
 286 {
 287         dout("prepare_write_seq %p %llu -> %llu\n", con,
 288              con->in_seq_acked, con->in_seq);
 289         con->in_seq_acked = con->in_seq;
 290
 291         con_out_kvec_reset(con);
 292
 293         con->v1.out_temp_ack = cpu_to_le64(con->in_seq_acked);
 294         con_out_kvec_add(con, sizeof(con->v1.out_temp_ack),
 295                          &con->v1.out_temp_ack);
 296
 297         ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
 298 }
 299
 300 /*
 301  * Prepare to write keepalive byte.
 302  */
 303 static void prepare_write_keepalive(struct ceph_connection *con)
 304 {
 305         dout("prepare_write_keepalive %p\n", con);
 306         con_out_kvec_reset(con);
 307         if (con->peer_features & CEPH_FEATURE_MSGR_KEEPALIVE2) {
 308                 struct timespec64 now;
 309
 310                 ktime_get_real_ts64(&now);
 311                 con_out_kvec_add(con, sizeof(tag_keepalive2), &tag_keepalive2);
 312                 ceph_encode_timespec64(&con->v1.out_temp_keepalive2, &now);
 313                 con_out_kvec_add(con, sizeof(con->v1.out_temp_keepalive2),
 314                                  &con->v1.out_temp_keepalive2);
 315         } else {
 316                 con_out_kvec_add(con, sizeof(tag_keepalive), &tag_keepalive);
 317         }
 318         ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
 319 }
 320
 321 /*
 322  * Connection negotiation.
 323  */
 324
 325 static int get_connect_authorizer(struct ceph_connection *con)
 326 {
 327         struct ceph_auth_handshake *auth;
 328         int auth_proto;
 329
 330         if (!con->ops->get_authorizer) {
 331                 con->v1.auth = NULL;
 332                 con->v1.out_connect.authorizer_protocol = CEPH_AUTH_UNKNOWN;
 333                 con->v1.out_connect.authorizer_len = 0;
 334                 return 0;
 335         }
 336
 337         auth = con->ops->get_authorizer(con, &auth_proto, con->v1.auth_retry);
 338         if (IS_ERR(auth))
 339                 return PTR_ERR(auth);
 340
 341         con->v1.auth = auth;
 342         con->v1.out_connect.authorizer_protocol = cpu_to_le32(auth_proto);
 343         con->v1.out_connect.authorizer_len =
 344                 cpu_to_le32(auth->authorizer_buf_len);
 345         return 0;
 346 }
 347
 348 /*
 349  * We connected to a peer and are saying hello.
 350  */
 351 static void prepare_write_banner(struct ceph_connection *con)
 352 {
 353         con_out_kvec_add(con, strlen(CEPH_BANNER), CEPH_BANNER);
 354         con_out_kvec_add(con, sizeof (con->msgr->my_enc_addr),
 355                                         &con->msgr->my_enc_addr);
 356
 357         con->v1.out_more = 0;
 358         ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
 359 }
 360
 361 static void __prepare_write_connect(struct ceph_connection *con)
 362 {
 363         con_out_kvec_add(con, sizeof(con->v1.out_connect),
 364                          &con->v1.out_connect);
 365         if (con->v1.auth)
 366                 con_out_kvec_add(con, con->v1.auth->authorizer_buf_len,
 367                                  con->v1.auth->authorizer_buf);
 368
 369         con->v1.out_more = 0;
 370         ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
 371 }
 372
 373 static int prepare_write_connect(struct ceph_connection *con)
 374 {
 375         unsigned int global_seq = ceph_get_global_seq(con->msgr, 0);
 376         int proto;
 377         int ret;
 378
 379         switch (con->peer_name.type) {
 380         case CEPH_ENTITY_TYPE_MON:
 381                 proto = CEPH_MONC_PROTOCOL;
 382                 break;
 383         case CEPH_ENTITY_TYPE_OSD:
 384                 proto = CEPH_OSDC_PROTOCOL;
 385                 break;
 386         case CEPH_ENTITY_TYPE_MDS:
 387                 proto = CEPH_MDSC_PROTOCOL;
 388                 break;
 389         default:
 390                 BUG();
 391         }
 392
 393         dout("prepare_write_connect %p cseq=%d gseq=%d proto=%d\n", con,
 394              con->v1.connect_seq, global_seq, proto);
 395
 396         con->v1.out_connect.features =
 397                 cpu_to_le64(from_msgr(con->msgr)->supported_features);
 398         con->v1.out_connect.host_type = cpu_to_le32(CEPH_ENTITY_TYPE_CLIENT);
 399         con->v1.out_connect.connect_seq = cpu_to_le32(con->v1.connect_seq);
 400         con->v1.out_connect.global_seq = cpu_to_le32(global_seq);
 401         con->v1.out_connect.protocol_version = cpu_to_le32(proto);
 402         con->v1.out_connect.flags = 0;
 403
 404         ret = get_connect_authorizer(con);
 405         if (ret)
 406                 return ret;
 407
 408         __prepare_write_connect(con);
 409         return 0;
 410 }
 411
 412 /*
 413  * write as much of pending kvecs to the socket as we can.
 414  *  1 -> done
 415  *  0 -> socket full, but more to do
 416  * <0 -> error
 417  */
 418 static int write_partial_kvec(struct ceph_connection *con)
 419 {
 420         int ret;
 421
 422         dout("write_partial_kvec %p %d left\n", con, con->v1.out_kvec_bytes);
 423         while (con->v1.out_kvec_bytes > 0) {
 424                 ret = ceph_tcp_sendmsg(con->sock, con->v1.out_kvec_cur,
 425                                        con->v1.out_kvec_left,
 426                                        con->v1.out_kvec_bytes,
 427                                        con->v1.out_more);
 428                 if (ret <= 0)
 429                         goto out;
 430                 con->v1.out_kvec_bytes -= ret;
 431                 if (!con->v1.out_kvec_bytes)
 432                         break;            /* done */
 433
 434                 /* account for full iov entries consumed */
 435                 while (ret >= con->v1.out_kvec_cur->iov_len) {
 436                         BUG_ON(!con->v1.out_kvec_left);
 437                         ret -= con->v1.out_kvec_cur->iov_len;
 438                         con->v1.out_kvec_cur++;
 439                         con->v1.out_kvec_left--;
 440                 }
 441                 /* and for a partially-consumed entry */
 442                 if (ret) {
 443                         con->v1.out_kvec_cur->iov_len -= ret;
 444                         con->v1.out_kvec_cur->iov_base += ret;
 445                 }
 446         }
 447         con->v1.out_kvec_left = 0;
 448         ret = 1;
 449 out:
 450         dout("write_partial_kvec %p %d left in %d kvecs ret = %d\n", con,
 451              con->v1.out_kvec_bytes, con->v1.out_kvec_left, ret);
 452         return ret;  /* done! */
 453 }
 454
 455 /*
 456  * Write as much message data payload as we can.  If we finish, queue
 457  * up the footer.
 458  *  1 -> done, footer is now queued in out_kvec[].
 459  *  0 -> socket full, but more to do
 460  * <0 -> error
 461  */
 462 static int write_partial_message_data(struct ceph_connection *con)
 463 {
 464         struct ceph_msg *msg = con->out_msg;
 465         struct ceph_msg_data_cursor *cursor = &msg->cursor;
 466         bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
 467         int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
 468         u32 crc;
 469
 470         dout("%s %p msg %p\n", __func__, con, msg);
 471
 472         if (!msg->num_data_items)
 473                 return -EINVAL;
 474
 475         /*
 476          * Iterate through each page that contains data to be
 477          * written, and send as much as possible for each.
 478          *
 479          * If we are calculating the data crc (the default), we will
 480          * need to map the page.  If we have no pages, they have
 481          * been revoked, so use the zero page.
 482          */
 483         crc = do_datacrc ? le32_to_cpu(msg->footer.data_crc) : 0;
 484         while (cursor->total_resid) {
 485                 struct page *page;
 486                 size_t page_offset;
 487                 size_t length;
 488                 int ret;
 489
 490                 if (!cursor->resid) {
 491                         ceph_msg_data_advance(cursor, 0);
 492                         continue;
 493                 }
 494
 495                 page = ceph_msg_data_next(cursor, &page_offset, &length);
 496                 if (length == cursor->total_resid)
 497                         more = MSG_MORE;
 498                 ret = ceph_tcp_sendpage(con->sock, page, page_offset, length,
 499                                         more);
 500                 if (ret <= 0) {
 501                         if (do_datacrc)
 502                                 msg->footer.data_crc = cpu_to_le32(crc);
 503
 504                         return ret;
 505                 }
 506                 if (do_datacrc && cursor->need_crc)
 507                         crc = ceph_crc32c_page(crc, page, page_offset, length);
 508                 ceph_msg_data_advance(cursor, (size_t)ret);
 509         }
 510
 511         dout("%s %p msg %p done\n", __func__, con, msg);
 512
 513         /* prepare and queue up footer, too */
 514         if (do_datacrc)
 515                 msg->footer.data_crc = cpu_to_le32(crc);
 516         else
 517                 msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC;
 518         con_out_kvec_reset(con);
 519         prepare_write_message_footer(con);
 520
 521         return 1;       /* must return > 0 to indicate success */
 522 }
 523
 524 /*
 525  * write some zeros
 526  */
 527 static int write_partial_skip(struct ceph_connection *con)
 528 {
 529         int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
 530         int ret;
 531
 532         dout("%s %p %d left\n", __func__, con, con->v1.out_skip);
 533         while (con->v1.out_skip > 0) {
 534                 size_t size = min(con->v1.out_skip, (int)PAGE_SIZE);
 535
 536                 if (size == con->v1.out_skip)
 537                         more = MSG_MORE;
 538                 ret = ceph_tcp_sendpage(con->sock, ceph_zero_page, 0, size,
 539                                         more);
 540                 if (ret <= 0)
 541                         goto out;
 542                 con->v1.out_skip -= ret;
 543         }
 544         ret = 1;
 545 out:
 546         return ret;
 547 }
 548
 549 /*
 550  * Prepare to read connection handshake, or an ack.
 551  */
 552 static void prepare_read_banner(struct ceph_connection *con)
 553 {
 554         dout("prepare_read_banner %p\n", con);
 555         con->v1.in_base_pos = 0;
 556 }
 557
 558 static void prepare_read_connect(struct ceph_connection *con)
 559 {
 560         dout("prepare_read_connect %p\n", con);
 561         con->v1.in_base_pos = 0;
 562 }
 563
 564 static void prepare_read_ack(struct ceph_connection *con)
 565 {
 566         dout("prepare_read_ack %p\n", con);
 567         con->v1.in_base_pos = 0;
 568 }
 569
 570 static void prepare_read_seq(struct ceph_connection *con)
 571 {
 572         dout("prepare_read_seq %p\n", con);
 573         con->v1.in_base_pos = 0;
 574         con->v1.in_tag = CEPH_MSGR_TAG_SEQ;
 575 }
 576
 577 static void prepare_read_tag(struct ceph_connection *con)
 578 {
 579         dout("prepare_read_tag %p\n", con);
 580         con->v1.in_base_pos = 0;
 581         con->v1.in_tag = CEPH_MSGR_TAG_READY;
 582 }
 583
 584 static void prepare_read_keepalive_ack(struct ceph_connection *con)
 585 {
 586         dout("prepare_read_keepalive_ack %p\n", con);
 587         con->v1.in_base_pos = 0;
 588 }
 589
 590 /*
 591  * Prepare to read a message.
 592  */
 593 static int prepare_read_message(struct ceph_connection *con)
 594 {
 595         dout("prepare_read_message %p\n", con);
 596         BUG_ON(con->in_msg != NULL);
 597         con->v1.in_base_pos = 0;
 598         con->in_front_crc = con->in_middle_crc = con->in_data_crc = 0;
 599         return 0;
 600 }
 601
 602 static int read_partial(struct ceph_connection *con,
 603                         int end, int size, void *object)
 604 {
 605         while (con->v1.in_base_pos < end) {
 606                 int left = end - con->v1.in_base_pos;
 607                 int have = size - left;
 608                 int ret = ceph_tcp_recvmsg(con->sock, object + have, left);
 609                 if (ret <= 0)
 610                         return ret;
 611                 con->v1.in_base_pos += ret;
 612         }
 613         return 1;
 614 }
 615
 616 /*
 617  * Read all or part of the connect-side handshake on a new connection
 618  */
 619 static int read_partial_banner(struct ceph_connection *con)
 620 {
 621         int size;
 622         int end;
 623         int ret;
 624
 625         dout("read_partial_banner %p at %d\n", con, con->v1.in_base_pos);
 626
 627         /* peer's banner */
 628         size = strlen(CEPH_BANNER);
 629         end = size;
 630         ret = read_partial(con, end, size, con->v1.in_banner);
 631         if (ret <= 0)
 632                 goto out;
 633
 634         size = sizeof(con->v1.actual_peer_addr);
 635         end += size;
 636         ret = read_partial(con, end, size, &con->v1.actual_peer_addr);
 637         if (ret <= 0)
 638                 goto out;
 639         ceph_decode_banner_addr(&con->v1.actual_peer_addr);
 640
 641         size = sizeof(con->v1.peer_addr_for_me);
 642         end += size;
 643         ret = read_partial(con, end, size, &con->v1.peer_addr_for_me);
 644         if (ret <= 0)
 645                 goto out;
 646         ceph_decode_banner_addr(&con->v1.peer_addr_for_me);
 647
 648 out:
 649         return ret;
 650 }
 651
 652 static int read_partial_connect(struct ceph_connection *con)
 653 {
 654         int size;
 655         int end;
 656         int ret;
 657
 658         dout("read_partial_connect %p at %d\n", con, con->v1.in_base_pos);
 659
 660         size = sizeof(con->v1.in_reply);
 661         end = size;
 662         ret = read_partial(con, end, size, &con->v1.in_reply);
 663         if (ret <= 0)
 664                 goto out;
 665
 666         if (con->v1.auth) {
 667                 size = le32_to_cpu(con->v1.in_reply.authorizer_len);
 668                 if (size > con->v1.auth->authorizer_reply_buf_len) {
 669                         pr_err("authorizer reply too big: %d > %zu\n", size,
 670                                con->v1.auth->authorizer_reply_buf_len);
 671                         ret = -EINVAL;
 672                         goto out;
 673                 }
 674
 675                 end += size;
 676                 ret = read_partial(con, end, size,
 677                                    con->v1.auth->authorizer_reply_buf);
 678                 if (ret <= 0)
 679                         goto out;
 680         }
 681
 682         dout("read_partial_connect %p tag %d, con_seq = %u, g_seq = %u\n",
 683              con, con->v1.in_reply.tag,
 684              le32_to_cpu(con->v1.in_reply.connect_seq),
 685              le32_to_cpu(con->v1.in_reply.global_seq));
 686 out:
 687         return ret;
 688 }
 689
 690 /*
 691  * Verify the hello banner looks okay.
 692  */
 693 static int verify_hello(struct ceph_connection *con)
 694 {
 695         if (memcmp(con->v1.in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) {
 696                 pr_err("connect to %s got bad banner\n",
 697                        ceph_pr_addr(&con->peer_addr));
 698                 con->error_msg = "protocol error, bad banner";
 699                 return -1;
 700         }
 701         return 0;
 702 }
 703
 704 static int process_banner(struct ceph_connection *con)
 705 {
 706         struct ceph_entity_addr *my_addr = &con->msgr->inst.addr;
 707
 708         dout("process_banner on %p\n", con);
 709
 710         if (verify_hello(con) < 0)
 711                 return -1;
 712
 713         /*
 714          * Make sure the other end is who we wanted.  note that the other
 715          * end may not yet know their ip address, so if it's 0.0.0.0, give
 716          * them the benefit of the doubt.
 717          */
 718         if (memcmp(&con->peer_addr, &con->v1.actual_peer_addr,
 719                    sizeof(con->peer_addr)) != 0 &&
 720             !(ceph_addr_is_blank(&con->v1.actual_peer_addr) &&
 721               con->v1.actual_peer_addr.nonce == con->peer_addr.nonce)) {
 722                 pr_warn("wrong peer, want %s/%u, got %s/%u\n",
 723                         ceph_pr_addr(&con->peer_addr),
 724                         le32_to_cpu(con->peer_addr.nonce),
 725                         ceph_pr_addr(&con->v1.actual_peer_addr),
 726                         le32_to_cpu(con->v1.actual_peer_addr.nonce));
 727                 con->error_msg = "wrong peer at address";
 728                 return -1;
 729         }
 730
 731         /*
 732          * did we learn our address?
 733          */
 734         if (ceph_addr_is_blank(my_addr)) {
 735                 memcpy(&my_addr->in_addr,
 736                        &con->v1.peer_addr_for_me.in_addr,
 737                        sizeof(con->v1.peer_addr_for_me.in_addr));
 738                 ceph_addr_set_port(my_addr, 0);
 739                 ceph_encode_my_addr(con->msgr);
 740                 dout("process_banner learned my addr is %s\n",
 741                      ceph_pr_addr(my_addr));
 742         }
 743
 744         return 0;
 745 }
 746
 747 static int process_connect(struct ceph_connection *con)
 748 {
 749         u64 sup_feat = from_msgr(con->msgr)->supported_features;
 750         u64 req_feat = from_msgr(con->msgr)->required_features;
 751         u64 server_feat = le64_to_cpu(con->v1.in_reply.features);
 752         int ret;
 753
 754         dout("process_connect on %p tag %d\n", con, con->v1.in_tag);
 755
 756         if (con->v1.auth) {
 757                 int len = le32_to_cpu(con->v1.in_reply.authorizer_len);
 758
 759                 /*
 760                  * Any connection that defines ->get_authorizer()
 761                  * should also define ->add_authorizer_challenge() and
 762                  * ->verify_authorizer_reply().
 763                  *
 764                  * See get_connect_authorizer().
 765                  */
 766                 if (con->v1.in_reply.tag ==
 767                                 CEPH_MSGR_TAG_CHALLENGE_AUTHORIZER) {
 768                         ret = con->ops->add_authorizer_challenge(
 769                                 con, con->v1.auth->authorizer_reply_buf, len);
 770                         if (ret < 0)
 771                                 return ret;
 772
 773                         con_out_kvec_reset(con);
 774                         __prepare_write_connect(con);
 775                         prepare_read_connect(con);
 776                         return 0;
 777                 }
 778
 779                 if (len) {
 780                         ret = con->ops->verify_authorizer_reply(con);
 781                         if (ret < 0) {
 782                                 con->error_msg = "bad authorize reply";
 783                                 return ret;
 784                         }
 785                 }
 786         }
 787
 788         switch (con->v1.in_reply.tag) {
 789         case CEPH_MSGR_TAG_FEATURES:
 790                 pr_err("%s%lld %s feature set mismatch,"
 791                        " my %llx < server's %llx, missing %llx\n",
 792                        ENTITY_NAME(con->peer_name),
 793                        ceph_pr_addr(&con->peer_addr),
 794                        sup_feat, server_feat, server_feat & ~sup_feat);
 795                 con->error_msg = "missing required protocol features";
 796                 return -1;
 797
 798         case CEPH_MSGR_TAG_BADPROTOVER:
 799                 pr_err("%s%lld %s protocol version mismatch,"
 800                        " my %d != server's %d\n",
 801                        ENTITY_NAME(con->peer_name),
 802                        ceph_pr_addr(&con->peer_addr),
 803                        le32_to_cpu(con->v1.out_connect.protocol_version),
 804                        le32_to_cpu(con->v1.in_reply.protocol_version));
 805                 con->error_msg = "protocol version mismatch";
 806                 return -1;
 807
 808         case CEPH_MSGR_TAG_BADAUTHORIZER:
 809                 con->v1.auth_retry++;
 810                 dout("process_connect %p got BADAUTHORIZER attempt %d\n", con,
 811                      con->v1.auth_retry);
 812                 if (con->v1.auth_retry == 2) {
 813                         con->error_msg = "connect authorization failure";
 814                         return -1;
 815                 }
 816                 con_out_kvec_reset(con);
 817                 ret = prepare_write_connect(con);
 818                 if (ret < 0)
 819                         return ret;
 820                 prepare_read_connect(con);
 821                 break;
 822
 823         case CEPH_MSGR_TAG_RESETSESSION:
 824                 /*
 825                  * If we connected with a large connect_seq but the peer
 826                  * has no record of a session with us (no connection, or
 827                  * connect_seq == 0), they will send RESETSESION to indicate
 828                  * that they must have reset their session, and may have
 829                  * dropped messages.
 830                  */
 831                 dout("process_connect got RESET peer seq %u\n",
 832                      le32_to_cpu(con->v1.in_reply.connect_seq));
 833                 pr_info("%s%lld %s session reset\n",
 834                         ENTITY_NAME(con->peer_name),
 835                         ceph_pr_addr(&con->peer_addr));
 836                 ceph_con_reset_session(con);
 837                 con_out_kvec_reset(con);
 838                 ret = prepare_write_connect(con);
 839                 if (ret < 0)
 840                         return ret;
 841                 prepare_read_connect(con);
 842
 843                 /* Tell ceph about it. */
 844                 mutex_unlock(&con->mutex);
 845                 if (con->ops->peer_reset)
 846                         con->ops->peer_reset(con);
 847                 mutex_lock(&con->mutex);
 848                 if (con->state != CEPH_CON_S_V1_CONNECT_MSG)
 849                         return -EAGAIN;
 850                 break;
 851
 852         case CEPH_MSGR_TAG_RETRY_SESSION:
 853                 /*
 854                  * If we sent a smaller connect_seq than the peer has, try
 855                  * again with a larger value.
 856                  */
 857                 dout("process_connect got RETRY_SESSION my seq %u, peer %u\n",
 858                      le32_to_cpu(con->v1.out_connect.connect_seq),
 859                      le32_to_cpu(con->v1.in_reply.connect_seq));
 860                 con->v1.connect_seq = le32_to_cpu(con->v1.in_reply.connect_seq);
 861                 con_out_kvec_reset(con);
 862                 ret = prepare_write_connect(con);
 863                 if (ret < 0)
 864                         return ret;
 865                 prepare_read_connect(con);
 866                 break;
 867
 868         case CEPH_MSGR_TAG_RETRY_GLOBAL:
 869                 /*
 870                  * If we sent a smaller global_seq than the peer has, try
 871                  * again with a larger value.
 872                  */
 873                 dout("process_connect got RETRY_GLOBAL my %u peer_gseq %u\n",
 874                      con->v1.peer_global_seq,
 875                      le32_to_cpu(con->v1.in_reply.global_seq));
 876                 ceph_get_global_seq(con->msgr,
 877                                     le32_to_cpu(con->v1.in_reply.global_seq));
 878                 con_out_kvec_reset(con);
 879                 ret = prepare_write_connect(con);
 880                 if (ret < 0)
 881                         return ret;
 882                 prepare_read_connect(con);
 883                 break;
 884
 885         case CEPH_MSGR_TAG_SEQ:
 886         case CEPH_MSGR_TAG_READY:
 887                 if (req_feat & ~server_feat) {
 888                         pr_err("%s%lld %s protocol feature mismatch,"
 889                                " my required %llx > server's %llx, need %llx\n",
 890                                ENTITY_NAME(con->peer_name),
 891                                ceph_pr_addr(&con->peer_addr),
 892                                req_feat, server_feat, req_feat & ~server_feat);
 893                         con->error_msg = "missing required protocol features";
 894                         return -1;
 895                 }
 896
 897                 WARN_ON(con->state != CEPH_CON_S_V1_CONNECT_MSG);
 898                 con->state = CEPH_CON_S_OPEN;
 899                 con->v1.auth_retry = 0;    /* we authenticated; clear flag */
 900                 con->v1.peer_global_seq =
 901                         le32_to_cpu(con->v1.in_reply.global_seq);
 902                 con->v1.connect_seq++;
 903                 con->peer_features = server_feat;
 904                 dout("process_connect got READY gseq %d cseq %d (%d)\n",
 905                      con->v1.peer_global_seq,
 906                      le32_to_cpu(con->v1.in_reply.connect_seq),
 907                      con->v1.connect_seq);
 908                 WARN_ON(con->v1.connect_seq !=
 909                         le32_to_cpu(con->v1.in_reply.connect_seq));
 910
 911                 if (con->v1.in_reply.flags & CEPH_MSG_CONNECT_LOSSY)
 912                         ceph_con_flag_set(con, CEPH_CON_F_LOSSYTX);
 913
 914                 con->delay = 0;      /* reset backoff memory */
 915
 916                 if (con->v1.in_reply.tag == CEPH_MSGR_TAG_SEQ) {
 917                         prepare_write_seq(con);
 918                         prepare_read_seq(con);
 919                 } else {
 920                         prepare_read_tag(con);
 921                 }
 922                 break;
 923
 924         case CEPH_MSGR_TAG_WAIT:
 925                 /*
 926                  * If there is a connection race (we are opening
 927                  * connections to each other), one of us may just have
 928                  * to WAIT.  This shouldn't happen if we are the
 929                  * client.
 930                  */
 931                 con->error_msg = "protocol error, got WAIT as client";
 932                 return -1;
 933
 934         default:
 935                 con->error_msg = "protocol error, garbage tag during connect";
 936                 return -1;
 937         }
 938         return 0;
 939 }
 940
 941 /*
 942  * read (part of) an ack
 943  */
 944 static int read_partial_ack(struct ceph_connection *con)
 945 {
 946         int size = sizeof(con->v1.in_temp_ack);
 947         int end = size;
 948
 949         return read_partial(con, end, size, &con->v1.in_temp_ack);
 950 }
 951
 952 /*
 953  * We can finally discard anything that's been acked.
 954  */
 955 static void process_ack(struct ceph_connection *con)
 956 {
 957         u64 ack = le64_to_cpu(con->v1.in_temp_ack);
 958
 959         if (con->v1.in_tag == CEPH_MSGR_TAG_ACK)
 960                 ceph_con_discard_sent(con, ack);
 961         else
 962                 ceph_con_discard_requeued(con, ack);
 963
 964         prepare_read_tag(con);
 965 }
 966
 967 static int read_partial_message_section(struct ceph_connection *con,
 968                                         struct kvec *section,
 969                                         unsigned int sec_len, u32 *crc)
 970 {
 971         int ret, left;
 972
 973         BUG_ON(!section);
 974
 975         while (section->iov_len < sec_len) {
 976                 BUG_ON(section->iov_base == NULL);
 977                 left = sec_len - section->iov_len;
 978                 ret = ceph_tcp_recvmsg(con->sock, (char *)section->iov_base +
 979                                        section->iov_len, left);
 980                 if (ret <= 0)
 981                         return ret;
 982                 section->iov_len += ret;
 983         }
 984         if (section->iov_len == sec_len)
 985                 *crc = crc32c(0, section->iov_base, section->iov_len);
 986
 987         return 1;
 988 }
 989
 990 static int read_partial_msg_data(struct ceph_connection *con)
 991 {
 992         struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor;
 993         bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
 994         struct page *page;
 995         size_t page_offset;
 996         size_t length;
 997         u32 crc = 0;
 998         int ret;
 999
1000         if (do_datacrc)
1001                 crc = con->in_data_crc;
1002         while (cursor->total_resid) {
1003                 if (!cursor->resid) {
1004                         ceph_msg_data_advance(cursor, 0);
1005                         continue;
1006                 }
1007
1008                 page = ceph_msg_data_next(cursor, &page_offset, &length);
1009                 ret = ceph_tcp_recvpage(con->sock, page, page_offset, length);
1010                 if (ret <= 0) {
1011                         if (do_datacrc)
1012                                 con->in_data_crc = crc;
1013
1014                         return ret;
1015                 }
1016
1017                 if (do_datacrc)
1018                         crc = ceph_crc32c_page(crc, page, page_offset, ret);
1019                 ceph_msg_data_advance(cursor, (size_t)ret);
1020         }
1021         if (do_datacrc)
1022                 con->in_data_crc = crc;
1023
1024         return 1;       /* must return > 0 to indicate success */
1025 }
1026
1027 static int read_partial_msg_data_bounce(struct ceph_connection *con)
1028 {
1029         struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor;
1030         struct page *page;
1031         size_t off, len;
1032         u32 crc;
1033         int ret;
1034
1035         if (unlikely(!con->bounce_page)) {
1036                 con->bounce_page = alloc_page(GFP_NOIO);
1037                 if (!con->bounce_page) {
1038                         pr_err("failed to allocate bounce page\n");
1039                         return -ENOMEM;
1040                 }
1041         }
1042
1043         crc = con->in_data_crc;
1044         while (cursor->total_resid) {
1045                 if (!cursor->resid) {
1046                         ceph_msg_data_advance(cursor, 0);
1047                         continue;
1048                 }
1049
1050                 page = ceph_msg_data_next(cursor, &off, &len);
1051                 ret = ceph_tcp_recvpage(con->sock, con->bounce_page, 0, len);
1052                 if (ret <= 0) {
1053                         con->in_data_crc = crc;
1054                         return ret;
1055                 }
1056
1057                 crc = crc32c(crc, page_address(con->bounce_page), ret);
1058                 memcpy_to_page(page, off, page_address(con->bounce_page), ret);
1059
1060                 ceph_msg_data_advance(cursor, ret);
1061         }
1062         con->in_data_crc = crc;
1063
1064         return 1;       /* must return > 0 to indicate success */
1065 }
1066
1067 /*
1068  * read (part of) a message.
1069  */
1070 static int read_partial_message(struct ceph_connection *con)
1071 {
1072         struct ceph_msg *m = con->in_msg;
1073         int size;
1074         int end;
1075         int ret;
1076         unsigned int front_len, middle_len, data_len;
1077         bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
1078         bool need_sign = (con->peer_features & CEPH_FEATURE_MSG_AUTH);
1079         u64 seq;
1080         u32 crc;
1081
1082         dout("read_partial_message con %p msg %p\n", con, m);
1083
1084         /* header */
1085         size = sizeof(con->v1.in_hdr);
1086         end = size;
1087         ret = read_partial(con, end, size, &con->v1.in_hdr);
1088         if (ret <= 0)
1089                 return ret;
1090
1091         crc = crc32c(0, &con->v1.in_hdr, offsetof(struct ceph_msg_header, crc));
1092         if (cpu_to_le32(crc) != con->v1.in_hdr.crc) {
1093                 pr_err("read_partial_message bad hdr crc %u != expected %u\n",
1094                        crc, con->v1.in_hdr.crc);
1095                 return -EBADMSG;
1096         }
1097
1098         front_len = le32_to_cpu(con->v1.in_hdr.front_len);
1099         if (front_len > CEPH_MSG_MAX_FRONT_LEN)
1100                 return -EIO;
1101         middle_len = le32_to_cpu(con->v1.in_hdr.middle_len);
1102         if (middle_len > CEPH_MSG_MAX_MIDDLE_LEN)
1103                 return -EIO;
1104         data_len = le32_to_cpu(con->v1.in_hdr.data_len);
1105         if (data_len > CEPH_MSG_MAX_DATA_LEN)
1106                 return -EIO;
1107
1108         /* verify seq# */
1109         seq = le64_to_cpu(con->v1.in_hdr.seq);
1110         if ((s64)seq - (s64)con->in_seq < 1) {
1111                 pr_info("skipping %s%lld %s seq %lld expected %lld\n",
1112                         ENTITY_NAME(con->peer_name),
1113                         ceph_pr_addr(&con->peer_addr),
1114                         seq, con->in_seq + 1);
1115                 con->v1.in_base_pos = -front_len - middle_len - data_len -
1116                                       sizeof_footer(con);
1117                 con->v1.in_tag = CEPH_MSGR_TAG_READY;
1118                 return 1;
1119         } else if ((s64)seq - (s64)con->in_seq > 1) {
1120                 pr_err("read_partial_message bad seq %lld expected %lld\n",
1121                        seq, con->in_seq + 1);
1122                 con->error_msg = "bad message sequence # for incoming message";
1123                 return -EBADE;
1124         }
1125
1126         /* allocate message? */
1127         if (!con->in_msg) {
1128                 int skip = 0;
1129
1130                 dout("got hdr type %d front %d data %d\n", con->v1.in_hdr.type,
1131                      front_len, data_len);
1132                 ret = ceph_con_in_msg_alloc(con, &con->v1.in_hdr, &skip);
1133                 if (ret < 0)
1134                         return ret;
1135
1136                 BUG_ON((!con->in_msg) ^ skip);
1137                 if (skip) {
1138                         /* skip this message */
1139                         dout("alloc_msg said skip message\n");
1140                         con->v1.in_base_pos = -front_len - middle_len -
1141                                               data_len - sizeof_footer(con);
1142                         con->v1.in_tag = CEPH_MSGR_TAG_READY;
1143                         con->in_seq++;
1144                         return 1;
1145                 }
1146
1147                 BUG_ON(!con->in_msg);
1148                 BUG_ON(con->in_msg->con != con);
1149                 m = con->in_msg;
1150                 m->front.iov_len = 0;    /* haven't read it yet */
1151                 if (m->middle)
1152                         m->middle->vec.iov_len = 0;
1153
1154                 /* prepare for data payload, if any */
1155
1156                 if (data_len)
1157                         prepare_message_data(con->in_msg, data_len);
1158         }
1159
1160         /* front */
1161         ret = read_partial_message_section(con, &m->front, front_len,
1162                                            &con->in_front_crc);
1163         if (ret <= 0)
1164                 return ret;
1165
1166         /* middle */
1167         if (m->middle) {
1168                 ret = read_partial_message_section(con, &m->middle->vec,
1169                                                    middle_len,
1170                                                    &con->in_middle_crc);
1171                 if (ret <= 0)
1172                         return ret;
1173         }
1174
1175         /* (page) data */
1176         if (data_len) {
1177                 if (!m->num_data_items)
1178                         return -EIO;
1179
1180                 if (ceph_test_opt(from_msgr(con->msgr), RXBOUNCE))
1181                         ret = read_partial_msg_data_bounce(con);
1182                 else
1183                         ret = read_partial_msg_data(con);
1184                 if (ret <= 0)
1185                         return ret;
1186         }
1187
1188         /* footer */
1189         size = sizeof_footer(con);
1190         end += size;
1191         ret = read_partial(con, end, size, &m->footer);
1192         if (ret <= 0)
1193                 return ret;
1194
1195         if (!need_sign) {
1196                 m->footer.flags = m->old_footer.flags;
1197                 m->footer.sig = 0;
1198         }
1199
1200         dout("read_partial_message got msg %p %d (%u) + %d (%u) + %d (%u)\n",
1201              m, front_len, m->footer.front_crc, middle_len,
1202              m->footer.middle_crc, data_len, m->footer.data_crc);
1203
1204         /* crc ok? */
1205         if (con->in_front_crc != le32_to_cpu(m->footer.front_crc)) {
1206                 pr_err("read_partial_message %p front crc %u != exp. %u\n",
1207                        m, con->in_front_crc, m->footer.front_crc);
1208                 return -EBADMSG;
1209         }
1210         if (con->in_middle_crc != le32_to_cpu(m->footer.middle_crc)) {
1211                 pr_err("read_partial_message %p middle crc %u != exp %u\n",
1212                        m, con->in_middle_crc, m->footer.middle_crc);
1213                 return -EBADMSG;
1214         }
1215         if (do_datacrc &&
1216             (m->footer.flags & CEPH_MSG_FOOTER_NOCRC) == 0 &&
1217             con->in_data_crc != le32_to_cpu(m->footer.data_crc)) {
1218                 pr_err("read_partial_message %p data crc %u != exp. %u\n", m,
1219                        con->in_data_crc, le32_to_cpu(m->footer.data_crc));
1220                 return -EBADMSG;
1221         }
1222
1223         if (need_sign && con->ops->check_message_signature &&
1224             con->ops->check_message_signature(m)) {
1225                 pr_err("read_partial_message %p signature check failed\n", m);
1226                 return -EBADMSG;
1227         }
1228
1229         return 1; /* done! */
1230 }
1231
1232 static int read_keepalive_ack(struct ceph_connection *con)
1233 {
1234         struct ceph_timespec ceph_ts;
1235         size_t size = sizeof(ceph_ts);
1236         int ret = read_partial(con, size, size, &ceph_ts);
1237         if (ret <= 0)
1238                 return ret;
1239         ceph_decode_timespec64(&con->last_keepalive_ack, &ceph_ts);
1240         prepare_read_tag(con);
1241         return 1;
1242 }
1243
1244 /*
1245  * Read what we can from the socket.
1246  */
1247 int ceph_con_v1_try_read(struct ceph_connection *con)
1248 {
1249         int ret = -1;
1250
1251 more:
1252         dout("try_read start %p state %d\n", con, con->state);
1253         if (con->state != CEPH_CON_S_V1_BANNER &&
1254             con->state != CEPH_CON_S_V1_CONNECT_MSG &&
1255             con->state != CEPH_CON_S_OPEN)
1256                 return 0;
1257
1258         BUG_ON(!con->sock);
1259
1260         dout("try_read tag %d in_base_pos %d\n", con->v1.in_tag,
1261              con->v1.in_base_pos);
1262
1263         if (con->state == CEPH_CON_S_V1_BANNER) {
1264                 ret = read_partial_banner(con);
1265                 if (ret <= 0)
1266                         goto out;
1267                 ret = process_banner(con);
1268                 if (ret < 0)
1269                         goto out;
1270
1271                 con->state = CEPH_CON_S_V1_CONNECT_MSG;
1272
1273                 /*
1274                  * Received banner is good, exchange connection info.
1275                  * Do not reset out_kvec, as sending our banner raced
1276                  * with receiving peer banner after connect completed.
1277                  */
1278                 ret = prepare_write_connect(con);
1279                 if (ret < 0)
1280                         goto out;
1281                 prepare_read_connect(con);
1282
1283                 /* Send connection info before awaiting response */
1284                 goto out;
1285         }
1286
1287         if (con->state == CEPH_CON_S_V1_CONNECT_MSG) {
1288                 ret = read_partial_connect(con);
1289                 if (ret <= 0)
1290                         goto out;
1291                 ret = process_connect(con);
1292                 if (ret < 0)
1293                         goto out;
1294                 goto more;
1295         }
1296
1297         WARN_ON(con->state != CEPH_CON_S_OPEN);
1298
1299         if (con->v1.in_base_pos < 0) {
1300                 /*
1301                  * skipping + discarding content.
1302                  */
1303                 ret = ceph_tcp_recvmsg(con->sock, NULL, -con->v1.in_base_pos);
1304                 if (ret <= 0)
1305                         goto out;
1306                 dout("skipped %d / %d bytes\n", ret, -con->v1.in_base_pos);
1307                 con->v1.in_base_pos += ret;
1308                 if (con->v1.in_base_pos)
1309                         goto more;
1310         }
1311         if (con->v1.in_tag == CEPH_MSGR_TAG_READY) {
1312                 /*
1313                  * what's next?
1314                  */
1315                 ret = ceph_tcp_recvmsg(con->sock, &con->v1.in_tag, 1);
1316                 if (ret <= 0)
1317                         goto out;
1318                 dout("try_read got tag %d\n", con->v1.in_tag);
1319                 switch (con->v1.in_tag) {
1320                 case CEPH_MSGR_TAG_MSG:
1321                         prepare_read_message(con);
1322                         break;
1323                 case CEPH_MSGR_TAG_ACK:
1324                         prepare_read_ack(con);
1325                         break;
1326                 case CEPH_MSGR_TAG_KEEPALIVE2_ACK:
1327                         prepare_read_keepalive_ack(con);
1328                         break;
1329                 case CEPH_MSGR_TAG_CLOSE:
1330                         ceph_con_close_socket(con);
1331                         con->state = CEPH_CON_S_CLOSED;
1332                         goto out;
1333                 default:
1334                         goto bad_tag;
1335                 }
1336         }
1337         if (con->v1.in_tag == CEPH_MSGR_TAG_MSG) {
1338                 ret = read_partial_message(con);
1339                 if (ret <= 0) {
1340                         switch (ret) {
1341                         case -EBADMSG:
1342                                 con->error_msg = "bad crc/signature";
1343                                 fallthrough;
1344                         case -EBADE:
1345                                 ret = -EIO;
1346                                 break;
1347                         case -EIO:
1348                                 con->error_msg = "io error";
1349                                 break;
1350                         }
1351                         goto out;
1352                 }
1353                 if (con->v1.in_tag == CEPH_MSGR_TAG_READY)
1354                         goto more;
1355                 ceph_con_process_message(con);
1356                 if (con->state == CEPH_CON_S_OPEN)
1357                         prepare_read_tag(con);
1358                 goto more;
1359         }
1360         if (con->v1.in_tag == CEPH_MSGR_TAG_ACK ||
1361             con->v1.in_tag == CEPH_MSGR_TAG_SEQ) {
1362                 /*
1363                  * the final handshake seq exchange is semantically
1364                  * equivalent to an ACK
1365                  */
1366                 ret = read_partial_ack(con);
1367                 if (ret <= 0)
1368                         goto out;
1369                 process_ack(con);
1370                 goto more;
1371         }
1372         if (con->v1.in_tag == CEPH_MSGR_TAG_KEEPALIVE2_ACK) {
1373                 ret = read_keepalive_ack(con);
1374                 if (ret <= 0)
1375                         goto out;
1376                 goto more;
1377         }
1378
1379 out:
1380         dout("try_read done on %p ret %d\n", con, ret);
1381         return ret;
1382
1383 bad_tag:
1384         pr_err("try_read bad tag %d\n", con->v1.in_tag);
1385         con->error_msg = "protocol error, garbage tag";
1386         ret = -1;
1387         goto out;
1388 }
1389
1390 /*
1391  * Write something to the socket.  Called in a worker thread when the
1392  * socket appears to be writeable and we have something ready to send.
1393  */
1394 int ceph_con_v1_try_write(struct ceph_connection *con)
1395 {
1396         int ret = 1;
1397
1398         dout("try_write start %p state %d\n", con, con->state);
1399         if (con->state != CEPH_CON_S_PREOPEN &&
1400             con->state != CEPH_CON_S_V1_BANNER &&
1401             con->state != CEPH_CON_S_V1_CONNECT_MSG &&
1402             con->state != CEPH_CON_S_OPEN)
1403                 return 0;
1404
1405         /* open the socket first? */
1406         if (con->state == CEPH_CON_S_PREOPEN) {
1407                 BUG_ON(con->sock);
1408                 con->state = CEPH_CON_S_V1_BANNER;
1409
1410                 con_out_kvec_reset(con);
1411                 prepare_write_banner(con);
1412                 prepare_read_banner(con);
1413
1414                 BUG_ON(con->in_msg);
1415                 con->v1.in_tag = CEPH_MSGR_TAG_READY;
1416                 dout("try_write initiating connect on %p new state %d\n",
1417                      con, con->state);
1418                 ret = ceph_tcp_connect(con);
1419                 if (ret < 0) {
1420                         con->error_msg = "connect error";
1421                         goto out;
1422                 }
1423         }
1424
1425 more:
1426         dout("try_write out_kvec_bytes %d\n", con->v1.out_kvec_bytes);
1427         BUG_ON(!con->sock);
1428
1429         /* kvec data queued? */
1430         if (con->v1.out_kvec_left) {
1431                 ret = write_partial_kvec(con);
1432                 if (ret <= 0)
1433                         goto out;
1434         }
1435         if (con->v1.out_skip) {
1436                 ret = write_partial_skip(con);
1437                 if (ret <= 0)
1438                         goto out;
1439         }
1440
1441         /* msg pages? */
1442         if (con->out_msg) {
1443                 if (con->v1.out_msg_done) {
1444                         ceph_msg_put(con->out_msg);
1445                         con->out_msg = NULL;   /* we're done with this one */
1446                         goto do_next;
1447                 }
1448
1449                 ret = write_partial_message_data(con);
1450                 if (ret == 1)
1451                         goto more;  /* we need to send the footer, too! */
1452                 if (ret == 0)
1453                         goto out;
1454                 if (ret < 0) {
1455                         dout("try_write write_partial_message_data err %d\n",
1456                              ret);
1457                         goto out;
1458                 }
1459         }
1460
1461 do_next:
1462         if (con->state == CEPH_CON_S_OPEN) {
1463                 if (ceph_con_flag_test_and_clear(con,
1464                                 CEPH_CON_F_KEEPALIVE_PENDING)) {
1465                         prepare_write_keepalive(con);
1466                         goto more;
1467                 }
1468                 /* is anything else pending? */
1469                 if (!list_empty(&con->out_queue)) {
1470                         prepare_write_message(con);
1471                         goto more;
1472                 }
1473                 if (con->in_seq > con->in_seq_acked) {
1474                         prepare_write_ack(con);
1475                         goto more;
1476                 }
1477         }
1478
1479         /* Nothing to do! */
1480         ceph_con_flag_clear(con, CEPH_CON_F_WRITE_PENDING);
1481         dout("try_write nothing else to write.\n");
1482         ret = 0;
1483 out:
1484         dout("try_write done on %p ret %d\n", con, ret);
1485         return ret;
1486 }
1487
1488 void ceph_con_v1_revoke(struct ceph_connection *con)
1489 {
1490         struct ceph_msg *msg = con->out_msg;
1491
1492         WARN_ON(con->v1.out_skip);
1493         /* footer */
1494         if (con->v1.out_msg_done) {
1495                 con->v1.out_skip += con_out_kvec_skip(con);
1496         } else {
1497                 WARN_ON(!msg->data_length);
1498                 con->v1.out_skip += sizeof_footer(con);
1499         }
1500         /* data, middle, front */
1501         if (msg->data_length)
1502                 con->v1.out_skip += msg->cursor.total_resid;
1503         if (msg->middle)
1504                 con->v1.out_skip += con_out_kvec_skip(con);
1505         con->v1.out_skip += con_out_kvec_skip(con);
1506
1507         dout("%s con %p out_kvec_bytes %d out_skip %d\n", __func__, con,
1508              con->v1.out_kvec_bytes, con->v1.out_skip);
1509 }
1510
1511 void ceph_con_v1_revoke_incoming(struct ceph_connection *con)
1512 {
1513         unsigned int front_len = le32_to_cpu(con->v1.in_hdr.front_len);
1514         unsigned int middle_len = le32_to_cpu(con->v1.in_hdr.middle_len);
1515         unsigned int data_len = le32_to_cpu(con->v1.in_hdr.data_len);
1516
1517         /* skip rest of message */
1518         con->v1.in_base_pos = con->v1.in_base_pos -
1519                         sizeof(struct ceph_msg_header) -
1520                         front_len -
1521                         middle_len -
1522                         data_len -
1523                         sizeof(struct ceph_msg_footer);
1524
1525         con->v1.in_tag = CEPH_MSGR_TAG_READY;
1526         con->in_seq++;
1527
1528         dout("%s con %p in_base_pos %d\n", __func__, con, con->v1.in_base_pos);
1529 }
1530
1531 bool ceph_con_v1_opened(struct ceph_connection *con)
1532 {
1533         return con->v1.connect_seq;
1534 }
1535
1536 void ceph_con_v1_reset_session(struct ceph_connection *con)
1537 {
1538         con->v1.connect_seq = 0;
1539         con->v1.peer_global_seq = 0;
1540 }
1541
1542 void ceph_con_v1_reset_protocol(struct ceph_connection *con)
1543 {
1544         con->v1.out_skip = 0;
1545 }