1 /* 2 * Copyright (c) 2006, 2018 Oracle and/or its affiliates. All rights reserved. 3 * 4 * This software is available to you under a choice of one of two 5 * licenses. You may choose to be licensed under the terms of the GNU 6 * General Public License (GPL) Version 2, available from the file 7 * COPYING in the main directory of this source tree, or the 8 * OpenIB.org BSD license below: 9 * 10 * Redistribution and use in source and binary forms, with or 11 * without modification, are permitted provided that the following 12 * conditions are met: 13 * 14 * - Redistributions of source code must retain the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer. 17 * 18 * - Redistributions in binary form must reproduce the above 19 * copyright notice, this list of conditions and the following 20 * disclaimer in the documentation and/or other materials 21 * provided with the distribution. 22 * 23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 30 * SOFTWARE. 31 * 32 */ 33 #include <linux/kernel.h> 34 #include <linux/slab.h> 35 #include <net/sock.h> 36 #include <linux/in.h> 37 #include <linux/export.h> 38 #include <linux/time.h> 39 #include <linux/rds.h> 40 41 #include "rds.h" 42 43 void rds_inc_init(struct rds_incoming *inc, struct rds_connection *conn, 44 struct in6_addr *saddr) 45 { 46 int i; 47 48 refcount_set(&inc->i_refcount, 1); 49 INIT_LIST_HEAD(&inc->i_item); 50 inc->i_conn = conn; 51 inc->i_saddr = *saddr; 52 inc->i_rdma_cookie = 0; 53 inc->i_rx_tstamp.tv_sec = 0; 54 inc->i_rx_tstamp.tv_usec = 0; 55 56 for (i = 0; i < RDS_RX_MAX_TRACES; i++) 57 inc->i_rx_lat_trace[i] = 0; 58 } 59 EXPORT_SYMBOL_GPL(rds_inc_init); 60 61 void rds_inc_path_init(struct rds_incoming *inc, struct rds_conn_path *cp, 62 struct in6_addr *saddr) 63 { 64 refcount_set(&inc->i_refcount, 1); 65 INIT_LIST_HEAD(&inc->i_item); 66 inc->i_conn = cp->cp_conn; 67 inc->i_conn_path = cp; 68 inc->i_saddr = *saddr; 69 inc->i_rdma_cookie = 0; 70 inc->i_rx_tstamp.tv_sec = 0; 71 inc->i_rx_tstamp.tv_usec = 0; 72 } 73 EXPORT_SYMBOL_GPL(rds_inc_path_init); 74 75 static void rds_inc_addref(struct rds_incoming *inc) 76 { 77 rdsdebug("addref inc %p ref %d\n", inc, refcount_read(&inc->i_refcount)); 78 refcount_inc(&inc->i_refcount); 79 } 80 81 void rds_inc_put(struct rds_incoming *inc) 82 { 83 rdsdebug("put inc %p ref %d\n", inc, refcount_read(&inc->i_refcount)); 84 if (refcount_dec_and_test(&inc->i_refcount)) { 85 BUG_ON(!list_empty(&inc->i_item)); 86 87 inc->i_conn->c_trans->inc_free(inc); 88 } 89 } 90 EXPORT_SYMBOL_GPL(rds_inc_put); 91 92 static void rds_recv_rcvbuf_delta(struct rds_sock *rs, struct sock *sk, 93 struct rds_cong_map *map, 94 int delta, __be16 port) 95 { 96 int now_congested; 97 98 if (delta == 0) 99 return; 100 101 rs->rs_rcv_bytes += delta; 102 if (delta > 0) 103 rds_stats_add(s_recv_bytes_added_to_socket, delta); 104 else 105 rds_stats_add(s_recv_bytes_removed_from_socket, -delta); 106 107 /* loop transport doesn't send/recv congestion updates */ 108 if (rs->rs_transport->t_type == RDS_TRANS_LOOP) 109 return; 110 111 now_congested = rs->rs_rcv_bytes > rds_sk_rcvbuf(rs); 112 113 rdsdebug("rs %p (%pI6c:%u) recv bytes %d buf %d " 114 "now_cong %d delta %d\n", 115 rs, &rs->rs_bound_addr, 116 ntohs(rs->rs_bound_port), rs->rs_rcv_bytes, 117 rds_sk_rcvbuf(rs), now_congested, delta); 118 119 /* wasn't -> am congested */ 120 if (!rs->rs_congested && now_congested) { 121 rs->rs_congested = 1; 122 rds_cong_set_bit(map, port); 123 rds_cong_queue_updates(map); 124 } 125 /* was -> aren't congested */ 126 /* Require more free space before reporting uncongested to prevent 127 bouncing cong/uncong state too often */ 128 else if (rs->rs_congested && (rs->rs_rcv_bytes < (rds_sk_rcvbuf(rs)/2))) { 129 rs->rs_congested = 0; 130 rds_cong_clear_bit(map, port); 131 rds_cong_queue_updates(map); 132 } 133 134 /* do nothing if no change in cong state */ 135 } 136 137 static void rds_conn_peer_gen_update(struct rds_connection *conn, 138 u32 peer_gen_num) 139 { 140 int i; 141 struct rds_message *rm, *tmp; 142 unsigned long flags; 143 144 WARN_ON(conn->c_trans->t_type != RDS_TRANS_TCP); 145 if (peer_gen_num != 0) { 146 if (conn->c_peer_gen_num != 0 && 147 peer_gen_num != conn->c_peer_gen_num) { 148 for (i = 0; i < RDS_MPATH_WORKERS; i++) { 149 struct rds_conn_path *cp; 150 151 cp = &conn->c_path[i]; 152 spin_lock_irqsave(&cp->cp_lock, flags); 153 cp->cp_next_tx_seq = 1; 154 cp->cp_next_rx_seq = 0; 155 list_for_each_entry_safe(rm, tmp, 156 &cp->cp_retrans, 157 m_conn_item) { 158 set_bit(RDS_MSG_FLUSH, &rm->m_flags); 159 } 160 spin_unlock_irqrestore(&cp->cp_lock, flags); 161 } 162 } 163 conn->c_peer_gen_num = peer_gen_num; 164 } 165 } 166 167 /* 168 * Process all extension headers that come with this message. 169 */ 170 static void rds_recv_incoming_exthdrs(struct rds_incoming *inc, struct rds_sock *rs) 171 { 172 struct rds_header *hdr = &inc->i_hdr; 173 unsigned int pos = 0, type, len; 174 union { 175 struct rds_ext_header_version version; 176 struct rds_ext_header_rdma rdma; 177 struct rds_ext_header_rdma_dest rdma_dest; 178 } buffer; 179 180 while (1) { 181 len = sizeof(buffer); 182 type = rds_message_next_extension(hdr, &pos, &buffer, &len); 183 if (type == RDS_EXTHDR_NONE) 184 break; 185 /* Process extension header here */ 186 switch (type) { 187 case RDS_EXTHDR_RDMA: 188 rds_rdma_unuse(rs, be32_to_cpu(buffer.rdma.h_rdma_rkey), 0); 189 break; 190 191 case RDS_EXTHDR_RDMA_DEST: 192 /* We ignore the size for now. We could stash it 193 * somewhere and use it for error checking. */ 194 inc->i_rdma_cookie = rds_rdma_make_cookie( 195 be32_to_cpu(buffer.rdma_dest.h_rdma_rkey), 196 be32_to_cpu(buffer.rdma_dest.h_rdma_offset)); 197 198 break; 199 } 200 } 201 } 202 203 static void rds_recv_hs_exthdrs(struct rds_header *hdr, 204 struct rds_connection *conn) 205 { 206 unsigned int pos = 0, type, len; 207 union { 208 struct rds_ext_header_version version; 209 u16 rds_npaths; 210 u32 rds_gen_num; 211 } buffer; 212 u32 new_peer_gen_num = 0; 213 214 while (1) { 215 len = sizeof(buffer); 216 type = rds_message_next_extension(hdr, &pos, &buffer, &len); 217 if (type == RDS_EXTHDR_NONE) 218 break; 219 /* Process extension header here */ 220 switch (type) { 221 case RDS_EXTHDR_NPATHS: 222 conn->c_npaths = min_t(int, RDS_MPATH_WORKERS, 223 be16_to_cpu(buffer.rds_npaths)); 224 break; 225 case RDS_EXTHDR_GEN_NUM: 226 new_peer_gen_num = be32_to_cpu(buffer.rds_gen_num); 227 break; 228 default: 229 pr_warn_ratelimited("ignoring unknown exthdr type " 230 "0x%x\n", type); 231 } 232 } 233 /* if RDS_EXTHDR_NPATHS was not found, default to a single-path */ 234 conn->c_npaths = max_t(int, conn->c_npaths, 1); 235 conn->c_ping_triggered = 0; 236 rds_conn_peer_gen_update(conn, new_peer_gen_num); 237 } 238 239 /* rds_start_mprds() will synchronously start multiple paths when appropriate. 240 * The scheme is based on the following rules: 241 * 242 * 1. rds_sendmsg on first connect attempt sends the probe ping, with the 243 * sender's npaths (s_npaths) 244 * 2. rcvr of probe-ping knows the mprds_paths = min(s_npaths, r_npaths). It 245 * sends back a probe-pong with r_npaths. After that, if rcvr is the 246 * smaller ip addr, it starts rds_conn_path_connect_if_down on all 247 * mprds_paths. 248 * 3. sender gets woken up, and can move to rds_conn_path_connect_if_down. 249 * If it is the smaller ipaddr, rds_conn_path_connect_if_down can be 250 * called after reception of the probe-pong on all mprds_paths. 251 * Otherwise (sender of probe-ping is not the smaller ip addr): just call 252 * rds_conn_path_connect_if_down on the hashed path. (see rule 4) 253 * 4. rds_connect_worker must only trigger a connection if laddr < faddr. 254 * 5. sender may end up queuing the packet on the cp. will get sent out later. 255 * when connection is completed. 256 */ 257 static void rds_start_mprds(struct rds_connection *conn) 258 { 259 int i; 260 struct rds_conn_path *cp; 261 262 if (conn->c_npaths > 1 && 263 rds_addr_cmp(&conn->c_laddr, &conn->c_faddr) < 0) { 264 for (i = 0; i < conn->c_npaths; i++) { 265 cp = &conn->c_path[i]; 266 rds_conn_path_connect_if_down(cp); 267 } 268 } 269 } 270 271 /* 272 * The transport must make sure that this is serialized against other 273 * rx and conn reset on this specific conn. 274 * 275 * We currently assert that only one fragmented message will be sent 276 * down a connection at a time. This lets us reassemble in the conn 277 * instead of per-flow which means that we don't have to go digging through 278 * flows to tear down partial reassembly progress on conn failure and 279 * we save flow lookup and locking for each frag arrival. It does mean 280 * that small messages will wait behind large ones. Fragmenting at all 281 * is only to reduce the memory consumption of pre-posted buffers. 282 * 283 * The caller passes in saddr and daddr instead of us getting it from the 284 * conn. This lets loopback, who only has one conn for both directions, 285 * tell us which roles the addrs in the conn are playing for this message. 286 */ 287 void rds_recv_incoming(struct rds_connection *conn, struct in6_addr *saddr, 288 struct in6_addr *daddr, 289 struct rds_incoming *inc, gfp_t gfp) 290 { 291 struct rds_sock *rs = NULL; 292 struct sock *sk; 293 unsigned long flags; 294 struct rds_conn_path *cp; 295 296 inc->i_conn = conn; 297 inc->i_rx_jiffies = jiffies; 298 if (conn->c_trans->t_mp_capable) 299 cp = inc->i_conn_path; 300 else 301 cp = &conn->c_path[0]; 302 303 rdsdebug("conn %p next %llu inc %p seq %llu len %u sport %u dport %u " 304 "flags 0x%x rx_jiffies %lu\n", conn, 305 (unsigned long long)cp->cp_next_rx_seq, 306 inc, 307 (unsigned long long)be64_to_cpu(inc->i_hdr.h_sequence), 308 be32_to_cpu(inc->i_hdr.h_len), 309 be16_to_cpu(inc->i_hdr.h_sport), 310 be16_to_cpu(inc->i_hdr.h_dport), 311 inc->i_hdr.h_flags, 312 inc->i_rx_jiffies); 313 314 /* 315 * Sequence numbers should only increase. Messages get their 316 * sequence number as they're queued in a sending conn. They 317 * can be dropped, though, if the sending socket is closed before 318 * they hit the wire. So sequence numbers can skip forward 319 * under normal operation. They can also drop back in the conn 320 * failover case as previously sent messages are resent down the 321 * new instance of a conn. We drop those, otherwise we have 322 * to assume that the next valid seq does not come after a 323 * hole in the fragment stream. 324 * 325 * The headers don't give us a way to realize if fragments of 326 * a message have been dropped. We assume that frags that arrive 327 * to a flow are part of the current message on the flow that is 328 * being reassembled. This means that senders can't drop messages 329 * from the sending conn until all their frags are sent. 330 * 331 * XXX we could spend more on the wire to get more robust failure 332 * detection, arguably worth it to avoid data corruption. 333 */ 334 if (be64_to_cpu(inc->i_hdr.h_sequence) < cp->cp_next_rx_seq && 335 (inc->i_hdr.h_flags & RDS_FLAG_RETRANSMITTED)) { 336 rds_stats_inc(s_recv_drop_old_seq); 337 goto out; 338 } 339 cp->cp_next_rx_seq = be64_to_cpu(inc->i_hdr.h_sequence) + 1; 340 341 if (rds_sysctl_ping_enable && inc->i_hdr.h_dport == 0) { 342 if (inc->i_hdr.h_sport == 0) { 343 rdsdebug("ignore ping with 0 sport from %pI6c\n", 344 saddr); 345 goto out; 346 } 347 rds_stats_inc(s_recv_ping); 348 rds_send_pong(cp, inc->i_hdr.h_sport); 349 /* if this is a handshake ping, start multipath if necessary */ 350 if (RDS_HS_PROBE(be16_to_cpu(inc->i_hdr.h_sport), 351 be16_to_cpu(inc->i_hdr.h_dport))) { 352 rds_recv_hs_exthdrs(&inc->i_hdr, cp->cp_conn); 353 rds_start_mprds(cp->cp_conn); 354 } 355 goto out; 356 } 357 358 if (be16_to_cpu(inc->i_hdr.h_dport) == RDS_FLAG_PROBE_PORT && 359 inc->i_hdr.h_sport == 0) { 360 rds_recv_hs_exthdrs(&inc->i_hdr, cp->cp_conn); 361 /* if this is a handshake pong, start multipath if necessary */ 362 rds_start_mprds(cp->cp_conn); 363 wake_up(&cp->cp_conn->c_hs_waitq); 364 goto out; 365 } 366 367 rs = rds_find_bound(daddr, inc->i_hdr.h_dport, conn->c_bound_if); 368 if (!rs) { 369 rds_stats_inc(s_recv_drop_no_sock); 370 goto out; 371 } 372 373 /* Process extension headers */ 374 rds_recv_incoming_exthdrs(inc, rs); 375 376 /* We can be racing with rds_release() which marks the socket dead. */ 377 sk = rds_rs_to_sk(rs); 378 379 /* serialize with rds_release -> sock_orphan */ 380 write_lock_irqsave(&rs->rs_recv_lock, flags); 381 if (!sock_flag(sk, SOCK_DEAD)) { 382 rdsdebug("adding inc %p to rs %p's recv queue\n", inc, rs); 383 rds_stats_inc(s_recv_queued); 384 rds_recv_rcvbuf_delta(rs, sk, inc->i_conn->c_lcong, 385 be32_to_cpu(inc->i_hdr.h_len), 386 inc->i_hdr.h_dport); 387 if (sock_flag(sk, SOCK_RCVTSTAMP)) 388 do_gettimeofday(&inc->i_rx_tstamp); 389 rds_inc_addref(inc); 390 inc->i_rx_lat_trace[RDS_MSG_RX_END] = local_clock(); 391 list_add_tail(&inc->i_item, &rs->rs_recv_queue); 392 __rds_wake_sk_sleep(sk); 393 } else { 394 rds_stats_inc(s_recv_drop_dead_sock); 395 } 396 write_unlock_irqrestore(&rs->rs_recv_lock, flags); 397 398 out: 399 if (rs) 400 rds_sock_put(rs); 401 } 402 EXPORT_SYMBOL_GPL(rds_recv_incoming); 403 404 /* 405 * be very careful here. This is being called as the condition in 406 * wait_event_*() needs to cope with being called many times. 407 */ 408 static int rds_next_incoming(struct rds_sock *rs, struct rds_incoming **inc) 409 { 410 unsigned long flags; 411 412 if (!*inc) { 413 read_lock_irqsave(&rs->rs_recv_lock, flags); 414 if (!list_empty(&rs->rs_recv_queue)) { 415 *inc = list_entry(rs->rs_recv_queue.next, 416 struct rds_incoming, 417 i_item); 418 rds_inc_addref(*inc); 419 } 420 read_unlock_irqrestore(&rs->rs_recv_lock, flags); 421 } 422 423 return *inc != NULL; 424 } 425 426 static int rds_still_queued(struct rds_sock *rs, struct rds_incoming *inc, 427 int drop) 428 { 429 struct sock *sk = rds_rs_to_sk(rs); 430 int ret = 0; 431 unsigned long flags; 432 433 write_lock_irqsave(&rs->rs_recv_lock, flags); 434 if (!list_empty(&inc->i_item)) { 435 ret = 1; 436 if (drop) { 437 /* XXX make sure this i_conn is reliable */ 438 rds_recv_rcvbuf_delta(rs, sk, inc->i_conn->c_lcong, 439 -be32_to_cpu(inc->i_hdr.h_len), 440 inc->i_hdr.h_dport); 441 list_del_init(&inc->i_item); 442 rds_inc_put(inc); 443 } 444 } 445 write_unlock_irqrestore(&rs->rs_recv_lock, flags); 446 447 rdsdebug("inc %p rs %p still %d dropped %d\n", inc, rs, ret, drop); 448 return ret; 449 } 450 451 /* 452 * Pull errors off the error queue. 453 * If msghdr is NULL, we will just purge the error queue. 454 */ 455 int rds_notify_queue_get(struct rds_sock *rs, struct msghdr *msghdr) 456 { 457 struct rds_notifier *notifier; 458 struct rds_rdma_notify cmsg = { 0 }; /* fill holes with zero */ 459 unsigned int count = 0, max_messages = ~0U; 460 unsigned long flags; 461 LIST_HEAD(copy); 462 int err = 0; 463 464 465 /* put_cmsg copies to user space and thus may sleep. We can't do this 466 * with rs_lock held, so first grab as many notifications as we can stuff 467 * in the user provided cmsg buffer. We don't try to copy more, to avoid 468 * losing notifications - except when the buffer is so small that it wouldn't 469 * even hold a single notification. Then we give him as much of this single 470 * msg as we can squeeze in, and set MSG_CTRUNC. 471 */ 472 if (msghdr) { 473 max_messages = msghdr->msg_controllen / CMSG_SPACE(sizeof(cmsg)); 474 if (!max_messages) 475 max_messages = 1; 476 } 477 478 spin_lock_irqsave(&rs->rs_lock, flags); 479 while (!list_empty(&rs->rs_notify_queue) && count < max_messages) { 480 notifier = list_entry(rs->rs_notify_queue.next, 481 struct rds_notifier, n_list); 482 list_move(¬ifier->n_list, ©); 483 count++; 484 } 485 spin_unlock_irqrestore(&rs->rs_lock, flags); 486 487 if (!count) 488 return 0; 489 490 while (!list_empty(©)) { 491 notifier = list_entry(copy.next, struct rds_notifier, n_list); 492 493 if (msghdr) { 494 cmsg.user_token = notifier->n_user_token; 495 cmsg.status = notifier->n_status; 496 497 err = put_cmsg(msghdr, SOL_RDS, RDS_CMSG_RDMA_STATUS, 498 sizeof(cmsg), &cmsg); 499 if (err) 500 break; 501 } 502 503 list_del_init(¬ifier->n_list); 504 kfree(notifier); 505 } 506 507 /* If we bailed out because of an error in put_cmsg, 508 * we may be left with one or more notifications that we 509 * didn't process. Return them to the head of the list. */ 510 if (!list_empty(©)) { 511 spin_lock_irqsave(&rs->rs_lock, flags); 512 list_splice(©, &rs->rs_notify_queue); 513 spin_unlock_irqrestore(&rs->rs_lock, flags); 514 } 515 516 return err; 517 } 518 519 /* 520 * Queue a congestion notification 521 */ 522 static int rds_notify_cong(struct rds_sock *rs, struct msghdr *msghdr) 523 { 524 uint64_t notify = rs->rs_cong_notify; 525 unsigned long flags; 526 int err; 527 528 err = put_cmsg(msghdr, SOL_RDS, RDS_CMSG_CONG_UPDATE, 529 sizeof(notify), ¬ify); 530 if (err) 531 return err; 532 533 spin_lock_irqsave(&rs->rs_lock, flags); 534 rs->rs_cong_notify &= ~notify; 535 spin_unlock_irqrestore(&rs->rs_lock, flags); 536 537 return 0; 538 } 539 540 /* 541 * Receive any control messages. 542 */ 543 static int rds_cmsg_recv(struct rds_incoming *inc, struct msghdr *msg, 544 struct rds_sock *rs) 545 { 546 int ret = 0; 547 548 if (inc->i_rdma_cookie) { 549 ret = put_cmsg(msg, SOL_RDS, RDS_CMSG_RDMA_DEST, 550 sizeof(inc->i_rdma_cookie), &inc->i_rdma_cookie); 551 if (ret) 552 goto out; 553 } 554 555 if ((inc->i_rx_tstamp.tv_sec != 0) && 556 sock_flag(rds_rs_to_sk(rs), SOCK_RCVTSTAMP)) { 557 ret = put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, 558 sizeof(struct timeval), 559 &inc->i_rx_tstamp); 560 if (ret) 561 goto out; 562 } 563 564 if (rs->rs_rx_traces) { 565 struct rds_cmsg_rx_trace t; 566 int i, j; 567 568 memset(&t, 0, sizeof(t)); 569 inc->i_rx_lat_trace[RDS_MSG_RX_CMSG] = local_clock(); 570 t.rx_traces = rs->rs_rx_traces; 571 for (i = 0; i < rs->rs_rx_traces; i++) { 572 j = rs->rs_rx_trace[i]; 573 t.rx_trace_pos[i] = j; 574 t.rx_trace[i] = inc->i_rx_lat_trace[j + 1] - 575 inc->i_rx_lat_trace[j]; 576 } 577 578 ret = put_cmsg(msg, SOL_RDS, RDS_CMSG_RXPATH_LATENCY, 579 sizeof(t), &t); 580 if (ret) 581 goto out; 582 } 583 584 out: 585 return ret; 586 } 587 588 static bool rds_recvmsg_zcookie(struct rds_sock *rs, struct msghdr *msg) 589 { 590 struct rds_msg_zcopy_queue *q = &rs->rs_zcookie_queue; 591 struct rds_msg_zcopy_info *info = NULL; 592 struct rds_zcopy_cookies *done; 593 unsigned long flags; 594 595 if (!msg->msg_control) 596 return false; 597 598 if (!sock_flag(rds_rs_to_sk(rs), SOCK_ZEROCOPY) || 599 msg->msg_controllen < CMSG_SPACE(sizeof(*done))) 600 return false; 601 602 spin_lock_irqsave(&q->lock, flags); 603 if (!list_empty(&q->zcookie_head)) { 604 info = list_entry(q->zcookie_head.next, 605 struct rds_msg_zcopy_info, rs_zcookie_next); 606 list_del(&info->rs_zcookie_next); 607 } 608 spin_unlock_irqrestore(&q->lock, flags); 609 if (!info) 610 return false; 611 done = &info->zcookies; 612 if (put_cmsg(msg, SOL_RDS, RDS_CMSG_ZCOPY_COMPLETION, sizeof(*done), 613 done)) { 614 spin_lock_irqsave(&q->lock, flags); 615 list_add(&info->rs_zcookie_next, &q->zcookie_head); 616 spin_unlock_irqrestore(&q->lock, flags); 617 return false; 618 } 619 kfree(info); 620 return true; 621 } 622 623 int rds_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 624 int msg_flags) 625 { 626 struct sock *sk = sock->sk; 627 struct rds_sock *rs = rds_sk_to_rs(sk); 628 long timeo; 629 int ret = 0, nonblock = msg_flags & MSG_DONTWAIT; 630 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 631 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 632 struct rds_incoming *inc = NULL; 633 634 /* udp_recvmsg()->sock_recvtimeo() gets away without locking too.. */ 635 timeo = sock_rcvtimeo(sk, nonblock); 636 637 rdsdebug("size %zu flags 0x%x timeo %ld\n", size, msg_flags, timeo); 638 639 if (msg_flags & MSG_OOB) 640 goto out; 641 if (msg_flags & MSG_ERRQUEUE) 642 return sock_recv_errqueue(sk, msg, size, SOL_IP, IP_RECVERR); 643 644 while (1) { 645 /* If there are pending notifications, do those - and nothing else */ 646 if (!list_empty(&rs->rs_notify_queue)) { 647 ret = rds_notify_queue_get(rs, msg); 648 break; 649 } 650 651 if (rs->rs_cong_notify) { 652 ret = rds_notify_cong(rs, msg); 653 break; 654 } 655 656 if (!rds_next_incoming(rs, &inc)) { 657 if (nonblock) { 658 bool reaped = rds_recvmsg_zcookie(rs, msg); 659 660 ret = reaped ? 0 : -EAGAIN; 661 break; 662 } 663 664 timeo = wait_event_interruptible_timeout(*sk_sleep(sk), 665 (!list_empty(&rs->rs_notify_queue) || 666 rs->rs_cong_notify || 667 rds_next_incoming(rs, &inc)), timeo); 668 rdsdebug("recvmsg woke inc %p timeo %ld\n", inc, 669 timeo); 670 if (timeo > 0 || timeo == MAX_SCHEDULE_TIMEOUT) 671 continue; 672 673 ret = timeo; 674 if (ret == 0) 675 ret = -ETIMEDOUT; 676 break; 677 } 678 679 rdsdebug("copying inc %p from %pI6c:%u to user\n", inc, 680 &inc->i_conn->c_faddr, 681 ntohs(inc->i_hdr.h_sport)); 682 ret = inc->i_conn->c_trans->inc_copy_to_user(inc, &msg->msg_iter); 683 if (ret < 0) 684 break; 685 686 /* 687 * if the message we just copied isn't at the head of the 688 * recv queue then someone else raced us to return it, try 689 * to get the next message. 690 */ 691 if (!rds_still_queued(rs, inc, !(msg_flags & MSG_PEEK))) { 692 rds_inc_put(inc); 693 inc = NULL; 694 rds_stats_inc(s_recv_deliver_raced); 695 iov_iter_revert(&msg->msg_iter, ret); 696 continue; 697 } 698 699 if (ret < be32_to_cpu(inc->i_hdr.h_len)) { 700 if (msg_flags & MSG_TRUNC) 701 ret = be32_to_cpu(inc->i_hdr.h_len); 702 msg->msg_flags |= MSG_TRUNC; 703 } 704 705 if (rds_cmsg_recv(inc, msg, rs)) { 706 ret = -EFAULT; 707 goto out; 708 } 709 rds_recvmsg_zcookie(rs, msg); 710 711 rds_stats_inc(s_recv_delivered); 712 713 if (msg->msg_name) { 714 if (ipv6_addr_v4mapped(&inc->i_saddr)) { 715 sin = (struct sockaddr_in *)msg->msg_name; 716 717 sin->sin_family = AF_INET; 718 sin->sin_port = inc->i_hdr.h_sport; 719 sin->sin_addr.s_addr = 720 inc->i_saddr.s6_addr32[3]; 721 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 722 msg->msg_namelen = sizeof(*sin); 723 } else { 724 sin6 = (struct sockaddr_in6 *)msg->msg_name; 725 726 sin6->sin6_family = AF_INET6; 727 sin6->sin6_port = inc->i_hdr.h_sport; 728 sin6->sin6_addr = inc->i_saddr; 729 sin6->sin6_flowinfo = 0; 730 sin6->sin6_scope_id = rs->rs_bound_scope_id; 731 msg->msg_namelen = sizeof(*sin6); 732 } 733 } 734 break; 735 } 736 737 if (inc) 738 rds_inc_put(inc); 739 740 out: 741 return ret; 742 } 743 744 /* 745 * The socket is being shut down and we're asked to drop messages that were 746 * queued for recvmsg. The caller has unbound the socket so the receive path 747 * won't queue any more incoming fragments or messages on the socket. 748 */ 749 void rds_clear_recv_queue(struct rds_sock *rs) 750 { 751 struct sock *sk = rds_rs_to_sk(rs); 752 struct rds_incoming *inc, *tmp; 753 unsigned long flags; 754 755 write_lock_irqsave(&rs->rs_recv_lock, flags); 756 list_for_each_entry_safe(inc, tmp, &rs->rs_recv_queue, i_item) { 757 rds_recv_rcvbuf_delta(rs, sk, inc->i_conn->c_lcong, 758 -be32_to_cpu(inc->i_hdr.h_len), 759 inc->i_hdr.h_dport); 760 list_del_init(&inc->i_item); 761 rds_inc_put(inc); 762 } 763 write_unlock_irqrestore(&rs->rs_recv_lock, flags); 764 } 765 766 /* 767 * inc->i_saddr isn't used here because it is only set in the receive 768 * path. 769 */ 770 void rds_inc_info_copy(struct rds_incoming *inc, 771 struct rds_info_iterator *iter, 772 __be32 saddr, __be32 daddr, int flip) 773 { 774 struct rds_info_message minfo; 775 776 minfo.seq = be64_to_cpu(inc->i_hdr.h_sequence); 777 minfo.len = be32_to_cpu(inc->i_hdr.h_len); 778 779 if (flip) { 780 minfo.laddr = daddr; 781 minfo.faddr = saddr; 782 minfo.lport = inc->i_hdr.h_dport; 783 minfo.fport = inc->i_hdr.h_sport; 784 } else { 785 minfo.laddr = saddr; 786 minfo.faddr = daddr; 787 minfo.lport = inc->i_hdr.h_sport; 788 minfo.fport = inc->i_hdr.h_dport; 789 } 790 791 minfo.flags = 0; 792 793 rds_info_copy(iter, &minfo, sizeof(minfo)); 794 } 795 796 #if IS_ENABLED(CONFIG_IPV6) 797 void rds6_inc_info_copy(struct rds_incoming *inc, 798 struct rds_info_iterator *iter, 799 struct in6_addr *saddr, struct in6_addr *daddr, 800 int flip) 801 { 802 struct rds6_info_message minfo6; 803 804 minfo6.seq = be64_to_cpu(inc->i_hdr.h_sequence); 805 minfo6.len = be32_to_cpu(inc->i_hdr.h_len); 806 807 if (flip) { 808 minfo6.laddr = *daddr; 809 minfo6.faddr = *saddr; 810 minfo6.lport = inc->i_hdr.h_dport; 811 minfo6.fport = inc->i_hdr.h_sport; 812 } else { 813 minfo6.laddr = *saddr; 814 minfo6.faddr = *daddr; 815 minfo6.lport = inc->i_hdr.h_sport; 816 minfo6.fport = inc->i_hdr.h_dport; 817 } 818 819 rds_info_copy(iter, &minfo6, sizeof(minfo6)); 820 } 821 #endif 822