Blame view
net/rds/recv.c
20.7 KB
bdbe6fbc6 RDS: recv.c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 |
/* * Copyright (c) 2006 Oracle. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU * General Public License (GPL) Version 2, available from the file * COPYING in the main directory of this source tree, or the * OpenIB.org BSD license below: * * Redistribution and use in source and binary forms, with or * without modification, are permitted provided that the following * conditions are met: * * - Redistributions of source code must retain the above * copyright notice, this list of conditions and the following * disclaimer. * * - Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. * */ #include <linux/kernel.h> |
5a0e3ad6a include cleanup: ... |
34 |
#include <linux/slab.h> |
bdbe6fbc6 RDS: recv.c |
35 36 |
#include <net/sock.h> #include <linux/in.h> |
bc3b2d7fb net: Add export.h... |
37 |
#include <linux/export.h> |
5711f8b35 RDS: Add support ... |
38 39 |
#include <linux/time.h> #include <linux/rds.h> |
bdbe6fbc6 RDS: recv.c |
40 41 |
#include "rds.h" |
bdbe6fbc6 RDS: recv.c |
42 43 44 45 |
void rds_inc_init(struct rds_incoming *inc, struct rds_connection *conn, __be32 saddr) { |
3289025ae RDS: add receive ... |
46 |
int i; |
b7f029209 net, rds: convert... |
47 |
refcount_set(&inc->i_refcount, 1); |
bdbe6fbc6 RDS: recv.c |
48 49 50 51 |
INIT_LIST_HEAD(&inc->i_item); inc->i_conn = conn; inc->i_saddr = saddr; inc->i_rdma_cookie = 0; |
5711f8b35 RDS: Add support ... |
52 53 |
inc->i_rx_tstamp.tv_sec = 0; inc->i_rx_tstamp.tv_usec = 0; |
3289025ae RDS: add receive ... |
54 55 56 |
for (i = 0; i < RDS_RX_MAX_TRACES; i++) inc->i_rx_lat_trace[i] = 0; |
bdbe6fbc6 RDS: recv.c |
57 |
} |
616b757ae RDS: Export symbo... |
58 |
EXPORT_SYMBOL_GPL(rds_inc_init); |
bdbe6fbc6 RDS: recv.c |
59 |
|
5e833e025 RDS: rds_inc_path... |
60 61 62 |
void rds_inc_path_init(struct rds_incoming *inc, struct rds_conn_path *cp, __be32 saddr) { |
b7f029209 net, rds: convert... |
63 |
refcount_set(&inc->i_refcount, 1); |
5e833e025 RDS: rds_inc_path... |
64 65 66 67 68 69 70 71 72 |
INIT_LIST_HEAD(&inc->i_item); inc->i_conn = cp->cp_conn; inc->i_conn_path = cp; inc->i_saddr = saddr; inc->i_rdma_cookie = 0; inc->i_rx_tstamp.tv_sec = 0; inc->i_rx_tstamp.tv_usec = 0; } EXPORT_SYMBOL_GPL(rds_inc_path_init); |
ff51bf841 rds: make local f... |
73 |
static void rds_inc_addref(struct rds_incoming *inc) |
bdbe6fbc6 RDS: recv.c |
74 |
{ |
b7f029209 net, rds: convert... |
75 76 77 |
rdsdebug("addref inc %p ref %d ", inc, refcount_read(&inc->i_refcount)); refcount_inc(&inc->i_refcount); |
bdbe6fbc6 RDS: recv.c |
78 79 80 81 |
} void rds_inc_put(struct rds_incoming *inc) { |
b7f029209 net, rds: convert... |
82 83 84 |
rdsdebug("put inc %p ref %d ", inc, refcount_read(&inc->i_refcount)); if (refcount_dec_and_test(&inc->i_refcount)) { |
bdbe6fbc6 RDS: recv.c |
85 86 87 88 89 |
BUG_ON(!list_empty(&inc->i_item)); inc->i_conn->c_trans->inc_free(inc); } } |
616b757ae RDS: Export symbo... |
90 |
EXPORT_SYMBOL_GPL(rds_inc_put); |
bdbe6fbc6 RDS: recv.c |
91 92 93 94 95 96 97 98 99 100 101 |
static void rds_recv_rcvbuf_delta(struct rds_sock *rs, struct sock *sk, struct rds_cong_map *map, int delta, __be16 port) { int now_congested; if (delta == 0) return; rs->rs_rcv_bytes += delta; |
192a798f5 RDS: add stat for... |
102 103 104 105 |
if (delta > 0) rds_stats_add(s_recv_bytes_added_to_socket, delta); else rds_stats_add(s_recv_bytes_removed_from_socket, -delta); |
a4b57440d rds: avoid unenec... |
106 107 108 109 |
/* loop transport doesn't send/recv congestion updates */ if (rs->rs_transport->t_type == RDS_TRANS_LOOP) return; |
bdbe6fbc6 RDS: recv.c |
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 |
now_congested = rs->rs_rcv_bytes > rds_sk_rcvbuf(rs); rdsdebug("rs %p (%pI4:%u) recv bytes %d buf %d " "now_cong %d delta %d ", rs, &rs->rs_bound_addr, ntohs(rs->rs_bound_port), rs->rs_rcv_bytes, rds_sk_rcvbuf(rs), now_congested, delta); /* wasn't -> am congested */ if (!rs->rs_congested && now_congested) { rs->rs_congested = 1; rds_cong_set_bit(map, port); rds_cong_queue_updates(map); } /* was -> aren't congested */ /* Require more free space before reporting uncongested to prevent bouncing cong/uncong state too often */ else if (rs->rs_congested && (rs->rs_rcv_bytes < (rds_sk_rcvbuf(rs)/2))) { rs->rs_congested = 0; rds_cong_clear_bit(map, port); rds_cong_queue_updates(map); } /* do nothing if no change in cong state */ } |
905dd4184 RDS: TCP: Track p... |
136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 |
static void rds_conn_peer_gen_update(struct rds_connection *conn, u32 peer_gen_num) { int i; struct rds_message *rm, *tmp; unsigned long flags; WARN_ON(conn->c_trans->t_type != RDS_TRANS_TCP); if (peer_gen_num != 0) { if (conn->c_peer_gen_num != 0 && peer_gen_num != conn->c_peer_gen_num) { for (i = 0; i < RDS_MPATH_WORKERS; i++) { struct rds_conn_path *cp; cp = &conn->c_path[i]; spin_lock_irqsave(&cp->cp_lock, flags); cp->cp_next_tx_seq = 1; cp->cp_next_rx_seq = 0; list_for_each_entry_safe(rm, tmp, &cp->cp_retrans, m_conn_item) { set_bit(RDS_MSG_FLUSH, &rm->m_flags); } spin_unlock_irqrestore(&cp->cp_lock, flags); } } conn->c_peer_gen_num = peer_gen_num; } } |
bdbe6fbc6 RDS: recv.c |
165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 |
/* * Process all extension headers that come with this message. */ static void rds_recv_incoming_exthdrs(struct rds_incoming *inc, struct rds_sock *rs) { struct rds_header *hdr = &inc->i_hdr; unsigned int pos = 0, type, len; union { struct rds_ext_header_version version; struct rds_ext_header_rdma rdma; struct rds_ext_header_rdma_dest rdma_dest; } buffer; while (1) { len = sizeof(buffer); type = rds_message_next_extension(hdr, &pos, &buffer, &len); if (type == RDS_EXTHDR_NONE) break; /* Process extension header here */ switch (type) { case RDS_EXTHDR_RDMA: rds_rdma_unuse(rs, be32_to_cpu(buffer.rdma.h_rdma_rkey), 0); break; case RDS_EXTHDR_RDMA_DEST: /* We ignore the size for now. We could stash it * somewhere and use it for error checking. */ inc->i_rdma_cookie = rds_rdma_make_cookie( be32_to_cpu(buffer.rdma_dest.h_rdma_rkey), be32_to_cpu(buffer.rdma_dest.h_rdma_offset)); break; } } } |
5916e2c15 RDS: TCP: Enable ... |
200 201 202 203 204 205 206 |
static void rds_recv_hs_exthdrs(struct rds_header *hdr, struct rds_connection *conn) { unsigned int pos = 0, type, len; union { struct rds_ext_header_version version; u16 rds_npaths; |
905dd4184 RDS: TCP: Track p... |
207 |
u32 rds_gen_num; |
5916e2c15 RDS: TCP: Enable ... |
208 |
} buffer; |
905dd4184 RDS: TCP: Track p... |
209 |
u32 new_peer_gen_num = 0; |
5916e2c15 RDS: TCP: Enable ... |
210 211 212 213 214 215 216 217 218 219 |
while (1) { len = sizeof(buffer); type = rds_message_next_extension(hdr, &pos, &buffer, &len); if (type == RDS_EXTHDR_NONE) break; /* Process extension header here */ switch (type) { case RDS_EXTHDR_NPATHS: conn->c_npaths = min_t(int, RDS_MPATH_WORKERS, |
00354de57 rds: tcp: various... |
220 |
be16_to_cpu(buffer.rds_npaths)); |
5916e2c15 RDS: TCP: Enable ... |
221 |
break; |
905dd4184 RDS: TCP: Track p... |
222 |
case RDS_EXTHDR_GEN_NUM: |
00354de57 rds: tcp: various... |
223 |
new_peer_gen_num = be32_to_cpu(buffer.rds_gen_num); |
905dd4184 RDS: TCP: Track p... |
224 |
break; |
5916e2c15 RDS: TCP: Enable ... |
225 226 227 228 229 230 231 232 |
default: pr_warn_ratelimited("ignoring unknown exthdr type " "0x%x ", type); } } /* if RDS_EXTHDR_NPATHS was not found, default to a single-path */ conn->c_npaths = max_t(int, conn->c_npaths, 1); |
69b92b5b7 rds: tcp: send ha... |
233 |
conn->c_ping_triggered = 0; |
905dd4184 RDS: TCP: Track p... |
234 |
rds_conn_peer_gen_update(conn, new_peer_gen_num); |
5916e2c15 RDS: TCP: Enable ... |
235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 |
} /* rds_start_mprds() will synchronously start multiple paths when appropriate. * The scheme is based on the following rules: * * 1. rds_sendmsg on first connect attempt sends the probe ping, with the * sender's npaths (s_npaths) * 2. rcvr of probe-ping knows the mprds_paths = min(s_npaths, r_npaths). It * sends back a probe-pong with r_npaths. After that, if rcvr is the * smaller ip addr, it starts rds_conn_path_connect_if_down on all * mprds_paths. * 3. sender gets woken up, and can move to rds_conn_path_connect_if_down. * If it is the smaller ipaddr, rds_conn_path_connect_if_down can be * called after reception of the probe-pong on all mprds_paths. * Otherwise (sender of probe-ping is not the smaller ip addr): just call * rds_conn_path_connect_if_down on the hashed path. (see rule 4) |
69b92b5b7 rds: tcp: send ha... |
251 |
* 4. rds_connect_worker must only trigger a connection if laddr < faddr. |
5916e2c15 RDS: TCP: Enable ... |
252 253 254 255 256 257 258 |
* 5. sender may end up queuing the packet on the cp. will get sent out later. * when connection is completed. */ static void rds_start_mprds(struct rds_connection *conn) { int i; struct rds_conn_path *cp; |
00354de57 rds: tcp: various... |
259 260 |
if (conn->c_npaths > 1 && IS_CANONICAL(conn->c_laddr, conn->c_faddr)) { |
69b92b5b7 rds: tcp: send ha... |
261 |
for (i = 0; i < conn->c_npaths; i++) { |
5916e2c15 RDS: TCP: Enable ... |
262 263 264 265 266 |
cp = &conn->c_path[i]; rds_conn_path_connect_if_down(cp); } } } |
bdbe6fbc6 RDS: recv.c |
267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 |
/* * The transport must make sure that this is serialized against other * rx and conn reset on this specific conn. * * We currently assert that only one fragmented message will be sent * down a connection at a time. This lets us reassemble in the conn * instead of per-flow which means that we don't have to go digging through * flows to tear down partial reassembly progress on conn failure and * we save flow lookup and locking for each frag arrival. It does mean * that small messages will wait behind large ones. Fragmenting at all * is only to reduce the memory consumption of pre-posted buffers. * * The caller passes in saddr and daddr instead of us getting it from the * conn. This lets loopback, who only has one conn for both directions, * tell us which roles the addrs in the conn are playing for this message. */ void rds_recv_incoming(struct rds_connection *conn, __be32 saddr, __be32 daddr, |
6114eab53 rds: remove the s... |
284 |
struct rds_incoming *inc, gfp_t gfp) |
bdbe6fbc6 RDS: recv.c |
285 286 287 288 |
{ struct rds_sock *rs = NULL; struct sock *sk; unsigned long flags; |
ef9e62c2e RDS: recv path ge... |
289 |
struct rds_conn_path *cp; |
bdbe6fbc6 RDS: recv.c |
290 291 292 |
inc->i_conn = conn; inc->i_rx_jiffies = jiffies; |
ef9e62c2e RDS: recv path ge... |
293 294 295 296 |
if (conn->c_trans->t_mp_capable) cp = inc->i_conn_path; else cp = &conn->c_path[0]; |
bdbe6fbc6 RDS: recv.c |
297 298 299 300 |
rdsdebug("conn %p next %llu inc %p seq %llu len %u sport %u dport %u " "flags 0x%x rx_jiffies %lu ", conn, |
ef9e62c2e RDS: recv path ge... |
301 |
(unsigned long long)cp->cp_next_rx_seq, |
bdbe6fbc6 RDS: recv.c |
302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 |
inc, (unsigned long long)be64_to_cpu(inc->i_hdr.h_sequence), be32_to_cpu(inc->i_hdr.h_len), be16_to_cpu(inc->i_hdr.h_sport), be16_to_cpu(inc->i_hdr.h_dport), inc->i_hdr.h_flags, inc->i_rx_jiffies); /* * Sequence numbers should only increase. Messages get their * sequence number as they're queued in a sending conn. They * can be dropped, though, if the sending socket is closed before * they hit the wire. So sequence numbers can skip forward * under normal operation. They can also drop back in the conn * failover case as previously sent messages are resent down the * new instance of a conn. We drop those, otherwise we have * to assume that the next valid seq does not come after a * hole in the fragment stream. * * The headers don't give us a way to realize if fragments of * a message have been dropped. We assume that frags that arrive * to a flow are part of the current message on the flow that is * being reassembled. This means that senders can't drop messages * from the sending conn until all their frags are sent. * * XXX we could spend more on the wire to get more robust failure * detection, arguably worth it to avoid data corruption. */ |
ef9e62c2e RDS: recv path ge... |
330 |
if (be64_to_cpu(inc->i_hdr.h_sequence) < cp->cp_next_rx_seq && |
f64f9e719 net: Move && and ... |
331 |
(inc->i_hdr.h_flags & RDS_FLAG_RETRANSMITTED)) { |
bdbe6fbc6 RDS: recv.c |
332 333 334 |
rds_stats_inc(s_recv_drop_old_seq); goto out; } |
ef9e62c2e RDS: recv path ge... |
335 |
cp->cp_next_rx_seq = be64_to_cpu(inc->i_hdr.h_sequence) + 1; |
bdbe6fbc6 RDS: recv.c |
336 337 |
if (rds_sysctl_ping_enable && inc->i_hdr.h_dport == 0) { |
11bb62f7c RDS: Do not send ... |
338 339 340 341 342 |
if (inc->i_hdr.h_sport == 0) { rdsdebug("ignore ping with 0 sport from 0x%x ", saddr); goto out; } |
bdbe6fbc6 RDS: recv.c |
343 |
rds_stats_inc(s_recv_ping); |
45997e9e2 RDS: Make rds_sen... |
344 |
rds_send_pong(cp, inc->i_hdr.h_sport); |
5916e2c15 RDS: TCP: Enable ... |
345 |
/* if this is a handshake ping, start multipath if necessary */ |
00354de57 rds: tcp: various... |
346 347 |
if (RDS_HS_PROBE(be16_to_cpu(inc->i_hdr.h_sport), be16_to_cpu(inc->i_hdr.h_dport))) { |
5916e2c15 RDS: TCP: Enable ... |
348 349 350 351 352 |
rds_recv_hs_exthdrs(&inc->i_hdr, cp->cp_conn); rds_start_mprds(cp->cp_conn); } goto out; } |
00354de57 rds: tcp: various... |
353 |
if (be16_to_cpu(inc->i_hdr.h_dport) == RDS_FLAG_PROBE_PORT && |
5916e2c15 RDS: TCP: Enable ... |
354 355 356 357 358 |
inc->i_hdr.h_sport == 0) { rds_recv_hs_exthdrs(&inc->i_hdr, cp->cp_conn); /* if this is a handshake pong, start multipath if necessary */ rds_start_mprds(cp->cp_conn); wake_up(&cp->cp_conn->c_hs_waitq); |
bdbe6fbc6 RDS: recv.c |
359 360 361 362 |
goto out; } rs = rds_find_bound(daddr, inc->i_hdr.h_dport); |
8690bfa17 RDS: cleanup: rem... |
363 |
if (!rs) { |
bdbe6fbc6 RDS: recv.c |
364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 |
rds_stats_inc(s_recv_drop_no_sock); goto out; } /* Process extension headers */ rds_recv_incoming_exthdrs(inc, rs); /* We can be racing with rds_release() which marks the socket dead. */ sk = rds_rs_to_sk(rs); /* serialize with rds_release -> sock_orphan */ write_lock_irqsave(&rs->rs_recv_lock, flags); if (!sock_flag(sk, SOCK_DEAD)) { rdsdebug("adding inc %p to rs %p's recv queue ", inc, rs); rds_stats_inc(s_recv_queued); rds_recv_rcvbuf_delta(rs, sk, inc->i_conn->c_lcong, be32_to_cpu(inc->i_hdr.h_len), inc->i_hdr.h_dport); |
5711f8b35 RDS: Add support ... |
383 384 |
if (sock_flag(sk, SOCK_RCVTSTAMP)) do_gettimeofday(&inc->i_rx_tstamp); |
bdbe6fbc6 RDS: recv.c |
385 |
rds_inc_addref(inc); |
3289025ae RDS: add receive ... |
386 |
inc->i_rx_lat_trace[RDS_MSG_RX_END] = local_clock(); |
bdbe6fbc6 RDS: recv.c |
387 388 389 390 391 392 393 394 395 396 397 |
list_add_tail(&inc->i_item, &rs->rs_recv_queue); __rds_wake_sk_sleep(sk); } else { rds_stats_inc(s_recv_drop_dead_sock); } write_unlock_irqrestore(&rs->rs_recv_lock, flags); out: if (rs) rds_sock_put(rs); } |
616b757ae RDS: Export symbo... |
398 |
EXPORT_SYMBOL_GPL(rds_recv_incoming); |
bdbe6fbc6 RDS: recv.c |
399 400 401 402 403 404 405 406 |
/* * be very careful here. This is being called as the condition in * wait_event_*() needs to cope with being called many times. */ static int rds_next_incoming(struct rds_sock *rs, struct rds_incoming **inc) { unsigned long flags; |
8690bfa17 RDS: cleanup: rem... |
407 |
if (!*inc) { |
bdbe6fbc6 RDS: recv.c |
408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 |
read_lock_irqsave(&rs->rs_recv_lock, flags); if (!list_empty(&rs->rs_recv_queue)) { *inc = list_entry(rs->rs_recv_queue.next, struct rds_incoming, i_item); rds_inc_addref(*inc); } read_unlock_irqrestore(&rs->rs_recv_lock, flags); } return *inc != NULL; } static int rds_still_queued(struct rds_sock *rs, struct rds_incoming *inc, int drop) { struct sock *sk = rds_rs_to_sk(rs); int ret = 0; unsigned long flags; write_lock_irqsave(&rs->rs_recv_lock, flags); if (!list_empty(&inc->i_item)) { ret = 1; if (drop) { /* XXX make sure this i_conn is reliable */ rds_recv_rcvbuf_delta(rs, sk, inc->i_conn->c_lcong, -be32_to_cpu(inc->i_hdr.h_len), inc->i_hdr.h_dport); list_del_init(&inc->i_item); rds_inc_put(inc); } } write_unlock_irqrestore(&rs->rs_recv_lock, flags); rdsdebug("inc %p rs %p still %d dropped %d ", inc, rs, ret, drop); return ret; } /* * Pull errors off the error queue. * If msghdr is NULL, we will just purge the error queue. */ int rds_notify_queue_get(struct rds_sock *rs, struct msghdr *msghdr) { struct rds_notifier *notifier; |
f037590ff rds: fix a leak o... |
454 |
struct rds_rdma_notify cmsg = { 0 }; /* fill holes with zero */ |
bdbe6fbc6 RDS: recv.c |
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 |
unsigned int count = 0, max_messages = ~0U; unsigned long flags; LIST_HEAD(copy); int err = 0; /* put_cmsg copies to user space and thus may sleep. We can't do this * with rs_lock held, so first grab as many notifications as we can stuff * in the user provided cmsg buffer. We don't try to copy more, to avoid * losing notifications - except when the buffer is so small that it wouldn't * even hold a single notification. Then we give him as much of this single * msg as we can squeeze in, and set MSG_CTRUNC. */ if (msghdr) { max_messages = msghdr->msg_controllen / CMSG_SPACE(sizeof(cmsg)); if (!max_messages) max_messages = 1; } spin_lock_irqsave(&rs->rs_lock, flags); while (!list_empty(&rs->rs_notify_queue) && count < max_messages) { notifier = list_entry(rs->rs_notify_queue.next, struct rds_notifier, n_list); list_move(¬ifier->n_list, ©); count++; } spin_unlock_irqrestore(&rs->rs_lock, flags); if (!count) return 0; while (!list_empty(©)) { notifier = list_entry(copy.next, struct rds_notifier, n_list); if (msghdr) { cmsg.user_token = notifier->n_user_token; |
6200ed779 RDS: Whitespace |
491 |
cmsg.status = notifier->n_status; |
bdbe6fbc6 RDS: recv.c |
492 493 |
err = put_cmsg(msghdr, SOL_RDS, RDS_CMSG_RDMA_STATUS, |
6200ed779 RDS: Whitespace |
494 |
sizeof(cmsg), &cmsg); |
bdbe6fbc6 RDS: recv.c |
495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 |
if (err) break; } list_del_init(¬ifier->n_list); kfree(notifier); } /* If we bailed out because of an error in put_cmsg, * we may be left with one or more notifications that we * didn't process. Return them to the head of the list. */ if (!list_empty(©)) { spin_lock_irqsave(&rs->rs_lock, flags); list_splice(©, &rs->rs_notify_queue); spin_unlock_irqrestore(&rs->rs_lock, flags); } return err; } /* * Queue a congestion notification */ static int rds_notify_cong(struct rds_sock *rs, struct msghdr *msghdr) { uint64_t notify = rs->rs_cong_notify; unsigned long flags; int err; err = put_cmsg(msghdr, SOL_RDS, RDS_CMSG_CONG_UPDATE, sizeof(notify), ¬ify); if (err) return err; spin_lock_irqsave(&rs->rs_lock, flags); rs->rs_cong_notify &= ~notify; spin_unlock_irqrestore(&rs->rs_lock, flags); return 0; } /* * Receive any control messages. */ |
5711f8b35 RDS: Add support ... |
539 540 |
static int rds_cmsg_recv(struct rds_incoming *inc, struct msghdr *msg, struct rds_sock *rs) |
bdbe6fbc6 RDS: recv.c |
541 542 543 544 545 546 547 |
{ int ret = 0; if (inc->i_rdma_cookie) { ret = put_cmsg(msg, SOL_RDS, RDS_CMSG_RDMA_DEST, sizeof(inc->i_rdma_cookie), &inc->i_rdma_cookie); if (ret) |
3289025ae RDS: add receive ... |
548 |
goto out; |
bdbe6fbc6 RDS: recv.c |
549 |
} |
5711f8b35 RDS: Add support ... |
550 551 552 553 554 555 |
if ((inc->i_rx_tstamp.tv_sec != 0) && sock_flag(rds_rs_to_sk(rs), SOCK_RCVTSTAMP)) { ret = put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(struct timeval), &inc->i_rx_tstamp); if (ret) |
3289025ae RDS: add receive ... |
556 |
goto out; |
5711f8b35 RDS: Add support ... |
557 |
} |
3289025ae RDS: add receive ... |
558 559 560 |
if (rs->rs_rx_traces) { struct rds_cmsg_rx_trace t; int i, j; |
30ffa967a rds: do not leak ... |
561 |
memset(&t, 0, sizeof(t)); |
3289025ae RDS: add receive ... |
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 |
inc->i_rx_lat_trace[RDS_MSG_RX_CMSG] = local_clock(); t.rx_traces = rs->rs_rx_traces; for (i = 0; i < rs->rs_rx_traces; i++) { j = rs->rs_rx_trace[i]; t.rx_trace_pos[i] = j; t.rx_trace[i] = inc->i_rx_lat_trace[j + 1] - inc->i_rx_lat_trace[j]; } ret = put_cmsg(msg, SOL_RDS, RDS_CMSG_RXPATH_LATENCY, sizeof(t), &t); if (ret) goto out; } out: return ret; |
bdbe6fbc6 RDS: recv.c |
579 |
} |
1b7841404 net: Remove iocb ... |
580 581 |
int rds_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, int msg_flags) |
bdbe6fbc6 RDS: recv.c |
582 583 584 585 586 |
{ struct sock *sk = sock->sk; struct rds_sock *rs = rds_sk_to_rs(sk); long timeo; int ret = 0, nonblock = msg_flags & MSG_DONTWAIT; |
342dfc306 net: add build-ti... |
587 |
DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); |
bdbe6fbc6 RDS: recv.c |
588 589 590 591 592 593 594 595 596 597 |
struct rds_incoming *inc = NULL; /* udp_recvmsg()->sock_recvtimeo() gets away without locking too.. */ timeo = sock_rcvtimeo(sk, nonblock); rdsdebug("size %zu flags 0x%x timeo %ld ", size, msg_flags, timeo); if (msg_flags & MSG_OOB) goto out; |
edacaeae5 RDS: Fix completi... |
598 599 600 601 602 603 |
while (1) { /* If there are pending notifications, do those - and nothing else */ if (!list_empty(&rs->rs_notify_queue)) { ret = rds_notify_queue_get(rs, msg); break; } |
bdbe6fbc6 RDS: recv.c |
604 |
|
edacaeae5 RDS: Fix completi... |
605 606 607 608 |
if (rs->rs_cong_notify) { ret = rds_notify_cong(rs, msg); break; } |
bdbe6fbc6 RDS: recv.c |
609 |
|
bdbe6fbc6 RDS: recv.c |
610 611 612 613 614 |
if (!rds_next_incoming(rs, &inc)) { if (nonblock) { ret = -EAGAIN; break; } |
aa3951451 net: sk_sleep() h... |
615 |
timeo = wait_event_interruptible_timeout(*sk_sleep(sk), |
f64f9e719 net: Move && and ... |
616 617 618 |
(!list_empty(&rs->rs_notify_queue) || rs->rs_cong_notify || rds_next_incoming(rs, &inc)), timeo); |
bdbe6fbc6 RDS: recv.c |
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 |
rdsdebug("recvmsg woke inc %p timeo %ld ", inc, timeo); if (timeo > 0 || timeo == MAX_SCHEDULE_TIMEOUT) continue; ret = timeo; if (ret == 0) ret = -ETIMEDOUT; break; } rdsdebug("copying inc %p from %pI4:%u to user ", inc, &inc->i_conn->c_faddr, ntohs(inc->i_hdr.h_sport)); |
c0371da60 put iov_iter into... |
635 |
ret = inc->i_conn->c_trans->inc_copy_to_user(inc, &msg->msg_iter); |
bdbe6fbc6 RDS: recv.c |
636 637 638 639 640 641 642 643 644 645 646 647 |
if (ret < 0) break; /* * if the message we just copied isn't at the head of the * recv queue then someone else raced us to return it, try * to get the next message. */ if (!rds_still_queued(rs, inc, !(msg_flags & MSG_PEEK))) { rds_inc_put(inc); inc = NULL; rds_stats_inc(s_recv_deliver_raced); |
dc88e3b4c rds: make use of ... |
648 |
iov_iter_revert(&msg->msg_iter, ret); |
bdbe6fbc6 RDS: recv.c |
649 650 651 652 653 654 655 656 |
continue; } if (ret < be32_to_cpu(inc->i_hdr.h_len)) { if (msg_flags & MSG_TRUNC) ret = be32_to_cpu(inc->i_hdr.h_len); msg->msg_flags |= MSG_TRUNC; } |
5711f8b35 RDS: Add support ... |
657 |
if (rds_cmsg_recv(inc, msg, rs)) { |
bdbe6fbc6 RDS: recv.c |
658 659 660 661 662 |
ret = -EFAULT; goto out; } rds_stats_inc(s_recv_delivered); |
bdbe6fbc6 RDS: recv.c |
663 664 665 666 667 |
if (sin) { sin->sin_family = AF_INET; sin->sin_port = inc->i_hdr.h_sport; sin->sin_addr.s_addr = inc->i_saddr; memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); |
06b6a1cf6 rds: set correct ... |
668 |
msg->msg_namelen = sizeof(*sin); |
bdbe6fbc6 RDS: recv.c |
669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 |
} break; } if (inc) rds_inc_put(inc); out: return ret; } /* * The socket is being shut down and we're asked to drop messages that were * queued for recvmsg. The caller has unbound the socket so the receive path * won't queue any more incoming fragments or messages on the socket. */ void rds_clear_recv_queue(struct rds_sock *rs) { struct sock *sk = rds_rs_to_sk(rs); struct rds_incoming *inc, *tmp; unsigned long flags; write_lock_irqsave(&rs->rs_recv_lock, flags); list_for_each_entry_safe(inc, tmp, &rs->rs_recv_queue, i_item) { rds_recv_rcvbuf_delta(rs, sk, inc->i_conn->c_lcong, -be32_to_cpu(inc->i_hdr.h_len), inc->i_hdr.h_dport); list_del_init(&inc->i_item); rds_inc_put(inc); } write_unlock_irqrestore(&rs->rs_recv_lock, flags); } /* * inc->i_saddr isn't used here because it is only set in the receive * path. */ void rds_inc_info_copy(struct rds_incoming *inc, struct rds_info_iterator *iter, __be32 saddr, __be32 daddr, int flip) { struct rds_info_message minfo; minfo.seq = be64_to_cpu(inc->i_hdr.h_sequence); minfo.len = be32_to_cpu(inc->i_hdr.h_len); if (flip) { minfo.laddr = daddr; minfo.faddr = saddr; minfo.lport = inc->i_hdr.h_dport; minfo.fport = inc->i_hdr.h_sport; } else { minfo.laddr = saddr; minfo.faddr = daddr; minfo.lport = inc->i_hdr.h_sport; minfo.fport = inc->i_hdr.h_dport; } |
4116def23 rds: fix an infol... |
726 |
minfo.flags = 0; |
bdbe6fbc6 RDS: recv.c |
727 728 |
rds_info_copy(iter, &minfo, sizeof(minfo)); } |