net/rds/send.c

   1 /*
   2  * Copyright (c) 2006 Oracle.  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/moduleparam.h>
  35 #include <linux/gfp.h>
  36 #include <net/sock.h>
  37 #include <linux/in.h>
  38 #include <linux/list.h>
  39 #include <linux/ratelimit.h>
  40 #include <linux/export.h>
  41 #include <linux/sizes.h>
  42
  43 #include "rds.h"
  44
  45 /* When transmitting messages in rds_send_xmit, we need to emerge from
  46  * time to time and briefly release the CPU. Otherwise the softlock watchdog
  47  * will kick our shin.
  48  * Also, it seems fairer to not let one busy connection stall all the
  49  * others.
  50  *
  51  * send_batch_count is the number of times we'll loop in send_xmit. Setting
  52  * it to 0 will restore the old behavior (where we looped until we had
  53  * drained the queue).
  54  */
  55 static int send_batch_count = SZ_1K;
  56 module_param(send_batch_count, int, 0444);
  57 MODULE_PARM_DESC(send_batch_count, " batch factor when working the send queue");
  58
  59 static void rds_send_remove_from_sock(struct list_head *messages, int status);
  60
  61 /*
  62  * Reset the send state.  Callers must ensure that this doesn't race with
  63  * rds_send_xmit().
  64  */
  65 void rds_send_path_reset(struct rds_conn_path *cp)
  66 {
  67         struct rds_message *rm, *tmp;
  68         unsigned long flags;
  69
  70         if (cp->cp_xmit_rm) {
  71                 rm = cp->cp_xmit_rm;
  72                 cp->cp_xmit_rm = NULL;
  73                 /* Tell the user the RDMA op is no longer mapped by the
  74                  * transport. This isn't entirely true (it's flushed out
  75                  * independently) but as the connection is down, there's
  76                  * no ongoing RDMA to/from that memory */
  77                 rds_message_unmapped(rm);
  78                 rds_message_put(rm);
  79         }
  80
  81         cp->cp_xmit_sg = 0;
  82         cp->cp_xmit_hdr_off = 0;
  83         cp->cp_xmit_data_off = 0;
  84         cp->cp_xmit_atomic_sent = 0;
  85         cp->cp_xmit_rdma_sent = 0;
  86         cp->cp_xmit_data_sent = 0;
  87
  88         cp->cp_conn->c_map_queued = 0;
  89
  90         cp->cp_unacked_packets = rds_sysctl_max_unacked_packets;
  91         cp->cp_unacked_bytes = rds_sysctl_max_unacked_bytes;
  92
  93         /* Mark messages as retransmissions, and move them to the send q */
  94         spin_lock_irqsave(&cp->cp_lock, flags);
  95         list_for_each_entry_safe(rm, tmp, &cp->cp_retrans, m_conn_item) {
  96                 set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags);
  97                 set_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags);
  98         }
  99         list_splice_init(&cp->cp_retrans, &cp->cp_send_queue);
 100         spin_unlock_irqrestore(&cp->cp_lock, flags);
 101 }
 102 EXPORT_SYMBOL_GPL(rds_send_path_reset);
 103
 104 static int acquire_in_xmit(struct rds_conn_path *cp)
 105 {
 106         return test_and_set_bit(RDS_IN_XMIT, &cp->cp_flags) == 0;
 107 }
 108
 109 static void release_in_xmit(struct rds_conn_path *cp)
 110 {
 111         clear_bit(RDS_IN_XMIT, &cp->cp_flags);
 112         smp_mb__after_atomic();
 113         /*
 114          * We don't use wait_on_bit()/wake_up_bit() because our waking is in a
 115          * hot path and finding waiters is very rare.  We don't want to walk
 116          * the system-wide hashed waitqueue buckets in the fast path only to
 117          * almost never find waiters.
 118          */
 119         if (waitqueue_active(&cp->cp_waitq))
 120                 wake_up_all(&cp->cp_waitq);
 121 }
 122
 123 /*
 124  * We're making the conscious trade-off here to only send one message
 125  * down the connection at a time.
 126  *   Pro:
 127  *      - tx queueing is a simple fifo list
 128  *      - reassembly is optional and easily done by transports per conn
 129  *      - no per flow rx lookup at all, straight to the socket
 130  *      - less per-frag memory and wire overhead
 131  *   Con:
 132  *      - queued acks can be delayed behind large messages
 133  *   Depends:
 134  *      - small message latency is higher behind queued large messages
 135  *      - large message latency isn't starved by intervening small sends
 136  */
 137 int rds_send_xmit(struct rds_conn_path *cp)
 138 {
 139         struct rds_connection *conn = cp->cp_conn;
 140         struct rds_message *rm;
 141         unsigned long flags;
 142         unsigned int tmp;
 143         struct scatterlist *sg;
 144         int ret = 0;
 145         LIST_HEAD(to_be_dropped);
 146         int batch_count;
 147         unsigned long send_gen = 0;
 148
 149 restart:
 150         batch_count = 0;
 151
 152         /*
 153          * sendmsg calls here after having queued its message on the send
 154          * queue.  We only have one task feeding the connection at a time.  If
 155          * another thread is already feeding the queue then we back off.  This
 156          * avoids blocking the caller and trading per-connection data between
 157          * caches per message.
 158          */
 159         if (!acquire_in_xmit(cp)) {
 160                 rds_stats_inc(s_send_lock_contention);
 161                 ret = -ENOMEM;
 162                 goto out;
 163         }
 164
 165         /*
 166          * we record the send generation after doing the xmit acquire.
 167          * if someone else manages to jump in and do some work, we'll use
 168          * this to avoid a goto restart farther down.
 169          *
 170          * The acquire_in_xmit() check above ensures that only one
 171          * caller can increment c_send_gen at any time.
 172          */
 173         cp->cp_send_gen++;
 174         send_gen = cp->cp_send_gen;
 175
 176         /*
 177          * rds_conn_shutdown() sets the conn state and then tests RDS_IN_XMIT,
 178          * we do the opposite to avoid races.
 179          */
 180         if (!rds_conn_path_up(cp)) {
 181                 release_in_xmit(cp);
 182                 ret = 0;
 183                 goto out;
 184         }
 185
 186         if (conn->c_trans->xmit_path_prepare)
 187                 conn->c_trans->xmit_path_prepare(cp);
 188
 189         /*
 190          * spin trying to push headers and data down the connection until
 191          * the connection doesn't make forward progress.
 192          */
 193         while (1) {
 194
 195                 rm = cp->cp_xmit_rm;
 196
 197                 /*
 198                  * If between sending messages, we can send a pending congestion
 199                  * map update.
 200                  */
 201                 if (!rm && test_and_clear_bit(0, &conn->c_map_queued)) {
 202                         rm = rds_cong_update_alloc(conn);
 203                         if (IS_ERR(rm)) {
 204                                 ret = PTR_ERR(rm);
 205                                 break;
 206                         }
 207                         rm->data.op_active = 1;
 208                         rm->m_inc.i_conn_path = cp;
 209                         rm->m_inc.i_conn = cp->cp_conn;
 210
 211                         cp->cp_xmit_rm = rm;
 212                 }
 213
 214                 /*
 215                  * If not already working on one, grab the next message.
 216                  *
 217                  * cp_xmit_rm holds a ref while we're sending this message down
 218                  * the connction.  We can use this ref while holding the
 219                  * send_sem.. rds_send_reset() is serialized with it.
 220                  */
 221                 if (!rm) {
 222                         unsigned int len;
 223
 224                         batch_count++;
 225
 226                         /* we want to process as big a batch as we can, but
 227                          * we also want to avoid softlockups.  If we've been
 228                          * through a lot of messages, lets back off and see
 229                          * if anyone else jumps in
 230                          */
 231                         if (batch_count >= send_batch_count)
 232                                 goto over_batch;
 233
 234                         spin_lock_irqsave(&cp->cp_lock, flags);
 235
 236                         if (!list_empty(&cp->cp_send_queue)) {
 237                                 rm = list_entry(cp->cp_send_queue.next,
 238                                                 struct rds_message,
 239                                                 m_conn_item);
 240                                 rds_message_addref(rm);
 241
 242                                 /*
 243                                  * Move the message from the send queue to the retransmit
 244                                  * list right away.
 245                                  */
 246                                 list_move_tail(&rm->m_conn_item,
 247                                                &cp->cp_retrans);
 248                         }
 249
 250                         spin_unlock_irqrestore(&cp->cp_lock, flags);
 251
 252                         if (!rm)
 253                                 break;
 254
 255                         /* Unfortunately, the way Infiniband deals with
 256                          * RDMA to a bad MR key is by moving the entire
 257                          * queue pair to error state. We cold possibly
 258                          * recover from that, but right now we drop the
 259                          * connection.
 260                          * Therefore, we never retransmit messages with RDMA ops.
 261                          */
 262                         if (test_bit(RDS_MSG_FLUSH, &rm->m_flags) ||
 263                             (rm->rdma.op_active &&
 264                             test_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags))) {
 265                                 spin_lock_irqsave(&cp->cp_lock, flags);
 266                                 if (test_and_clear_bit(RDS_MSG_ON_CONN, &rm->m_flags))
 267                                         list_move(&rm->m_conn_item, &to_be_dropped);
 268                                 spin_unlock_irqrestore(&cp->cp_lock, flags);
 269                                 continue;
 270                         }
 271
 272                         /* Require an ACK every once in a while */
 273                         len = ntohl(rm->m_inc.i_hdr.h_len);
 274                         if (cp->cp_unacked_packets == 0 ||
 275                             cp->cp_unacked_bytes < len) {
 276                                 __set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags);
 277
 278                                 cp->cp_unacked_packets =
 279                                         rds_sysctl_max_unacked_packets;
 280                                 cp->cp_unacked_bytes =
 281                                         rds_sysctl_max_unacked_bytes;
 282                                 rds_stats_inc(s_send_ack_required);
 283                         } else {
 284                                 cp->cp_unacked_bytes -= len;
 285                                 cp->cp_unacked_packets--;
 286                         }
 287
 288                         cp->cp_xmit_rm = rm;
 289                 }
 290
 291                 /* The transport either sends the whole rdma or none of it */
 292                 if (rm->rdma.op_active && !cp->cp_xmit_rdma_sent) {
 293                         rm->m_final_op = &rm->rdma;
 294                         /* The transport owns the mapped memory for now.
 295                          * You can't unmap it while it's on the send queue
 296                          */
 297                         set_bit(RDS_MSG_MAPPED, &rm->m_flags);
 298                         ret = conn->c_trans->xmit_rdma(conn, &rm->rdma);
 299                         if (ret) {
 300                                 clear_bit(RDS_MSG_MAPPED, &rm->m_flags);
 301                                 wake_up_interruptible(&rm->m_flush_wait);
 302                                 break;
 303                         }
 304                         cp->cp_xmit_rdma_sent = 1;
 305
 306                 }
 307
 308                 if (rm->atomic.op_active && !cp->cp_xmit_atomic_sent) {
 309                         rm->m_final_op = &rm->atomic;
 310                         /* The transport owns the mapped memory for now.
 311                          * You can't unmap it while it's on the send queue
 312                          */
 313                         set_bit(RDS_MSG_MAPPED, &rm->m_flags);
 314                         ret = conn->c_trans->xmit_atomic(conn, &rm->atomic);
 315                         if (ret) {
 316                                 clear_bit(RDS_MSG_MAPPED, &rm->m_flags);
 317                                 wake_up_interruptible(&rm->m_flush_wait);
 318                                 break;
 319                         }
 320                         cp->cp_xmit_atomic_sent = 1;
 321
 322                 }
 323
 324                 /*
 325                  * A number of cases require an RDS header to be sent
 326                  * even if there is no data.
 327                  * We permit 0-byte sends; rds-ping depends on this.
 328                  * However, if there are exclusively attached silent ops,
 329                  * we skip the hdr/data send, to enable silent operation.
 330                  */
 331                 if (rm->data.op_nents == 0) {
 332                         int ops_present;
 333                         int all_ops_are_silent = 1;
 334
 335                         ops_present = (rm->atomic.op_active || rm->rdma.op_active);
 336                         if (rm->atomic.op_active && !rm->atomic.op_silent)
 337                                 all_ops_are_silent = 0;
 338                         if (rm->rdma.op_active && !rm->rdma.op_silent)
 339                                 all_ops_are_silent = 0;
 340
 341                         if (ops_present && all_ops_are_silent
 342                             && !rm->m_rdma_cookie)
 343                                 rm->data.op_active = 0;
 344                 }
 345
 346                 if (rm->data.op_active && !cp->cp_xmit_data_sent) {
 347                         rm->m_final_op = &rm->data;
 348
 349                         ret = conn->c_trans->xmit(conn, rm,
 350                                                   cp->cp_xmit_hdr_off,
 351                                                   cp->cp_xmit_sg,
 352                                                   cp->cp_xmit_data_off);
 353                         if (ret <= 0)
 354                                 break;
 355
 356                         if (cp->cp_xmit_hdr_off < sizeof(struct rds_header)) {
 357                                 tmp = min_t(int, ret,
 358                                             sizeof(struct rds_header) -
 359                                             cp->cp_xmit_hdr_off);
 360                                 cp->cp_xmit_hdr_off += tmp;
 361                                 ret -= tmp;
 362                         }
 363
 364                         sg = &rm->data.op_sg[cp->cp_xmit_sg];
 365                         while (ret) {
 366                                 tmp = min_t(int, ret, sg->length -
 367                                                       cp->cp_xmit_data_off);
 368                                 cp->cp_xmit_data_off += tmp;
 369                                 ret -= tmp;
 370                                 if (cp->cp_xmit_data_off == sg->length) {
 371                                         cp->cp_xmit_data_off = 0;
 372                                         sg++;
 373                                         cp->cp_xmit_sg++;
 374                                         BUG_ON(ret != 0 && cp->cp_xmit_sg ==
 375                                                rm->data.op_nents);
 376                                 }
 377                         }
 378
 379                         if (cp->cp_xmit_hdr_off == sizeof(struct rds_header) &&
 380                             (cp->cp_xmit_sg == rm->data.op_nents))
 381                                 cp->cp_xmit_data_sent = 1;
 382                 }
 383
 384                 /*
 385                  * A rm will only take multiple times through this loop
 386                  * if there is a data op. Thus, if the data is sent (or there was
 387                  * none), then we're done with the rm.
 388                  */
 389                 if (!rm->data.op_active || cp->cp_xmit_data_sent) {
 390                         cp->cp_xmit_rm = NULL;
 391                         cp->cp_xmit_sg = 0;
 392                         cp->cp_xmit_hdr_off = 0;
 393                         cp->cp_xmit_data_off = 0;
 394                         cp->cp_xmit_rdma_sent = 0;
 395                         cp->cp_xmit_atomic_sent = 0;
 396                         cp->cp_xmit_data_sent = 0;
 397
 398                         rds_message_put(rm);
 399                 }
 400         }
 401
 402 over_batch:
 403         if (conn->c_trans->xmit_path_complete)
 404                 conn->c_trans->xmit_path_complete(cp);
 405         release_in_xmit(cp);
 406
 407         /* Nuke any messages we decided not to retransmit. */
 408         if (!list_empty(&to_be_dropped)) {
 409                 /* irqs on here, so we can put(), unlike above */
 410                 list_for_each_entry(rm, &to_be_dropped, m_conn_item)
 411                         rds_message_put(rm);
 412                 rds_send_remove_from_sock(&to_be_dropped, RDS_RDMA_DROPPED);
 413         }
 414
 415         /*
 416          * Other senders can queue a message after we last test the send queue
 417          * but before we clear RDS_IN_XMIT.  In that case they'd back off and
 418          * not try and send their newly queued message.  We need to check the
 419          * send queue after having cleared RDS_IN_XMIT so that their message
 420          * doesn't get stuck on the send queue.
 421          *
 422          * If the transport cannot continue (i.e ret != 0), then it must
 423          * call us when more room is available, such as from the tx
 424          * completion handler.
 425          *
 426          * We have an extra generation check here so that if someone manages
 427          * to jump in after our release_in_xmit, we'll see that they have done
 428          * some work and we will skip our goto
 429          */
 430         if (ret == 0) {
 431                 smp_mb();
 432                 if ((test_bit(0, &conn->c_map_queued) ||
 433                      !list_empty(&cp->cp_send_queue)) &&
 434                     send_gen == cp->cp_send_gen) {
 435                         rds_stats_inc(s_send_lock_queue_raced);
 436                         if (batch_count < send_batch_count)
 437                                 goto restart;
 438                         queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
 439                 }
 440         }
 441 out:
 442         return ret;
 443 }
 444 EXPORT_SYMBOL_GPL(rds_send_xmit);
 445
 446 static void rds_send_sndbuf_remove(struct rds_sock *rs, struct rds_message *rm)
 447 {
 448         u32 len = be32_to_cpu(rm->m_inc.i_hdr.h_len);
 449
 450         assert_spin_locked(&rs->rs_lock);
 451
 452         BUG_ON(rs->rs_snd_bytes < len);
 453         rs->rs_snd_bytes -= len;
 454
 455         if (rs->rs_snd_bytes == 0)
 456                 rds_stats_inc(s_send_queue_empty);
 457 }
 458
 459 static inline int rds_send_is_acked(struct rds_message *rm, u64 ack,
 460                                     is_acked_func is_acked)
 461 {
 462         if (is_acked)
 463                 return is_acked(rm, ack);
 464         return be64_to_cpu(rm->m_inc.i_hdr.h_sequence) <= ack;
 465 }
 466
 467 /*
 468  * This is pretty similar to what happens below in the ACK
 469  * handling code - except that we call here as soon as we get
 470  * the IB send completion on the RDMA op and the accompanying
 471  * message.
 472  */
 473 void rds_rdma_send_complete(struct rds_message *rm, int status)
 474 {
 475         struct rds_sock *rs = NULL;
 476         struct rm_rdma_op *ro;
 477         struct rds_notifier *notifier;
 478         unsigned long flags;
 479
 480         spin_lock_irqsave(&rm->m_rs_lock, flags);
 481
 482         ro = &rm->rdma;
 483         if (test_bit(RDS_MSG_ON_SOCK, &rm->m_flags) &&
 484             ro->op_active && ro->op_notify && ro->op_notifier) {
 485                 notifier = ro->op_notifier;
 486                 rs = rm->m_rs;
 487                 sock_hold(rds_rs_to_sk(rs));
 488
 489                 notifier->n_status = status;
 490                 spin_lock(&rs->rs_lock);
 491                 list_add_tail(&notifier->n_list, &rs->rs_notify_queue);
 492                 spin_unlock(&rs->rs_lock);
 493
 494                 ro->op_notifier = NULL;
 495         }
 496
 497         spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 498
 499         if (rs) {
 500                 rds_wake_sk_sleep(rs);
 501                 sock_put(rds_rs_to_sk(rs));
 502         }
 503 }
 504 EXPORT_SYMBOL_GPL(rds_rdma_send_complete);
 505
 506 /*
 507  * Just like above, except looks at atomic op
 508  */
 509 void rds_atomic_send_complete(struct rds_message *rm, int status)
 510 {
 511         struct rds_sock *rs = NULL;
 512         struct rm_atomic_op *ao;
 513         struct rds_notifier *notifier;
 514         unsigned long flags;
 515
 516         spin_lock_irqsave(&rm->m_rs_lock, flags);
 517
 518         ao = &rm->atomic;
 519         if (test_bit(RDS_MSG_ON_SOCK, &rm->m_flags)
 520             && ao->op_active && ao->op_notify && ao->op_notifier) {
 521                 notifier = ao->op_notifier;
 522                 rs = rm->m_rs;
 523                 sock_hold(rds_rs_to_sk(rs));
 524
 525                 notifier->n_status = status;
 526                 spin_lock(&rs->rs_lock);
 527                 list_add_tail(&notifier->n_list, &rs->rs_notify_queue);
 528                 spin_unlock(&rs->rs_lock);
 529
 530                 ao->op_notifier = NULL;
 531         }
 532
 533         spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 534
 535         if (rs) {
 536                 rds_wake_sk_sleep(rs);
 537                 sock_put(rds_rs_to_sk(rs));
 538         }
 539 }
 540 EXPORT_SYMBOL_GPL(rds_atomic_send_complete);
 541
 542 /*
 543  * This is the same as rds_rdma_send_complete except we
 544  * don't do any locking - we have all the ingredients (message,
 545  * socket, socket lock) and can just move the notifier.
 546  */
 547 static inline void
 548 __rds_send_complete(struct rds_sock *rs, struct rds_message *rm, int status)
 549 {
 550         struct rm_rdma_op *ro;
 551         struct rm_atomic_op *ao;
 552
 553         ro = &rm->rdma;
 554         if (ro->op_active && ro->op_notify && ro->op_notifier) {
 555                 ro->op_notifier->n_status = status;
 556                 list_add_tail(&ro->op_notifier->n_list, &rs->rs_notify_queue);
 557                 ro->op_notifier = NULL;
 558         }
 559
 560         ao = &rm->atomic;
 561         if (ao->op_active && ao->op_notify && ao->op_notifier) {
 562                 ao->op_notifier->n_status = status;
 563                 list_add_tail(&ao->op_notifier->n_list, &rs->rs_notify_queue);
 564                 ao->op_notifier = NULL;
 565         }
 566
 567         /* No need to wake the app - caller does this */
 568 }
 569
 570 /*
 571  * This removes messages from the socket's list if they're on it.  The list
 572  * argument must be private to the caller, we must be able to modify it
 573  * without locks.  The messages must have a reference held for their
 574  * position on the list.  This function will drop that reference after
 575  * removing the messages from the 'messages' list regardless of if it found
 576  * the messages on the socket list or not.
 577  */
 578 static void rds_send_remove_from_sock(struct list_head *messages, int status)
 579 {
 580         unsigned long flags;
 581         struct rds_sock *rs = NULL;
 582         struct rds_message *rm;
 583
 584         while (!list_empty(messages)) {
 585                 int was_on_sock = 0;
 586
 587                 rm = list_entry(messages->next, struct rds_message,
 588                                 m_conn_item);
 589                 list_del_init(&rm->m_conn_item);
 590
 591                 /*
 592                  * If we see this flag cleared then we're *sure* that someone
 593                  * else beat us to removing it from the sock.  If we race
 594                  * with their flag update we'll get the lock and then really
 595                  * see that the flag has been cleared.
 596                  *
 597                  * The message spinlock makes sure nobody clears rm->m_rs
 598                  * while we're messing with it. It does not prevent the
 599                  * message from being removed from the socket, though.
 600                  */
 601                 spin_lock_irqsave(&rm->m_rs_lock, flags);
 602                 if (!test_bit(RDS_MSG_ON_SOCK, &rm->m_flags))
 603                         goto unlock_and_drop;
 604
 605                 if (rs != rm->m_rs) {
 606                         if (rs) {
 607                                 rds_wake_sk_sleep(rs);
 608                                 sock_put(rds_rs_to_sk(rs));
 609                         }
 610                         rs = rm->m_rs;
 611                         if (rs)
 612                                 sock_hold(rds_rs_to_sk(rs));
 613                 }
 614                 if (!rs)
 615                         goto unlock_and_drop;
 616                 spin_lock(&rs->rs_lock);
 617
 618                 if (test_and_clear_bit(RDS_MSG_ON_SOCK, &rm->m_flags)) {
 619                         struct rm_rdma_op *ro = &rm->rdma;
 620                         struct rds_notifier *notifier;
 621
 622                         list_del_init(&rm->m_sock_item);
 623                         rds_send_sndbuf_remove(rs, rm);
 624
 625                         if (ro->op_active && ro->op_notifier &&
 626                                (ro->op_notify || (ro->op_recverr && status))) {
 627                                 notifier = ro->op_notifier;
 628                                 list_add_tail(&notifier->n_list,
 629                                                 &rs->rs_notify_queue);
 630                                 if (!notifier->n_status)
 631                                         notifier->n_status = status;
 632                                 rm->rdma.op_notifier = NULL;
 633                         }
 634                         was_on_sock = 1;
 635                         rm->m_rs = NULL;
 636                 }
 637                 spin_unlock(&rs->rs_lock);
 638
 639 unlock_and_drop:
 640                 spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 641                 rds_message_put(rm);
 642                 if (was_on_sock)
 643                         rds_message_put(rm);
 644         }
 645
 646         if (rs) {
 647                 rds_wake_sk_sleep(rs);
 648                 sock_put(rds_rs_to_sk(rs));
 649         }
 650 }
 651
 652 /*
 653  * Transports call here when they've determined that the receiver queued
 654  * messages up to, and including, the given sequence number.  Messages are
 655  * moved to the retrans queue when rds_send_xmit picks them off the send
 656  * queue. This means that in the TCP case, the message may not have been
 657  * assigned the m_ack_seq yet - but that's fine as long as tcp_is_acked
 658  * checks the RDS_MSG_HAS_ACK_SEQ bit.
 659  */
 660 void rds_send_path_drop_acked(struct rds_conn_path *cp, u64 ack,
 661                               is_acked_func is_acked)
 662 {
 663         struct rds_message *rm, *tmp;
 664         unsigned long flags;
 665         LIST_HEAD(list);
 666
 667         spin_lock_irqsave(&cp->cp_lock, flags);
 668
 669         list_for_each_entry_safe(rm, tmp, &cp->cp_retrans, m_conn_item) {
 670                 if (!rds_send_is_acked(rm, ack, is_acked))
 671                         break;
 672
 673                 list_move(&rm->m_conn_item, &list);
 674                 clear_bit(RDS_MSG_ON_CONN, &rm->m_flags);
 675         }
 676
 677         /* order flag updates with spin locks */
 678         if (!list_empty(&list))
 679                 smp_mb__after_atomic();
 680
 681         spin_unlock_irqrestore(&cp->cp_lock, flags);
 682
 683         /* now remove the messages from the sock list as needed */
 684         rds_send_remove_from_sock(&list, RDS_RDMA_SUCCESS);
 685 }
 686 EXPORT_SYMBOL_GPL(rds_send_path_drop_acked);
 687
 688 void rds_send_drop_acked(struct rds_connection *conn, u64 ack,
 689                          is_acked_func is_acked)
 690 {
 691         WARN_ON(conn->c_trans->t_mp_capable);
 692         rds_send_path_drop_acked(&conn->c_path[0], ack, is_acked);
 693 }
 694 EXPORT_SYMBOL_GPL(rds_send_drop_acked);
 695
 696 void rds_send_drop_to(struct rds_sock *rs, struct sockaddr_in *dest)
 697 {
 698         struct rds_message *rm, *tmp;
 699         struct rds_connection *conn;
 700         struct rds_conn_path *cp;
 701         unsigned long flags;
 702         LIST_HEAD(list);
 703
 704         /* get all the messages we're dropping under the rs lock */
 705         spin_lock_irqsave(&rs->rs_lock, flags);
 706
 707         list_for_each_entry_safe(rm, tmp, &rs->rs_send_queue, m_sock_item) {
 708                 if (dest && (dest->sin_addr.s_addr != rm->m_daddr ||
 709                              dest->sin_port != rm->m_inc.i_hdr.h_dport))
 710                         continue;
 711
 712                 list_move(&rm->m_sock_item, &list);
 713                 rds_send_sndbuf_remove(rs, rm);
 714                 clear_bit(RDS_MSG_ON_SOCK, &rm->m_flags);
 715         }
 716
 717         /* order flag updates with the rs lock */
 718         smp_mb__after_atomic();
 719
 720         spin_unlock_irqrestore(&rs->rs_lock, flags);
 721
 722         if (list_empty(&list))
 723                 return;
 724
 725         /* Remove the messages from the conn */
 726         list_for_each_entry(rm, &list, m_sock_item) {
 727
 728                 conn = rm->m_inc.i_conn;
 729                 if (conn->c_trans->t_mp_capable)
 730                         cp = rm->m_inc.i_conn_path;
 731                 else
 732                         cp = &conn->c_path[0];
 733
 734                 spin_lock_irqsave(&cp->cp_lock, flags);
 735                 /*
 736                  * Maybe someone else beat us to removing rm from the conn.
 737                  * If we race with their flag update we'll get the lock and
 738                  * then really see that the flag has been cleared.
 739                  */
 740                 if (!test_and_clear_bit(RDS_MSG_ON_CONN, &rm->m_flags)) {
 741                         spin_unlock_irqrestore(&cp->cp_lock, flags);
 742                         spin_lock_irqsave(&rm->m_rs_lock, flags);
 743                         rm->m_rs = NULL;
 744                         spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 745                         continue;
 746                 }
 747                 list_del_init(&rm->m_conn_item);
 748                 spin_unlock_irqrestore(&cp->cp_lock, flags);
 749
 750                 /*
 751                  * Couldn't grab m_rs_lock in top loop (lock ordering),
 752                  * but we can now.
 753                  */
 754                 spin_lock_irqsave(&rm->m_rs_lock, flags);
 755
 756                 spin_lock(&rs->rs_lock);
 757                 __rds_send_complete(rs, rm, RDS_RDMA_CANCELED);
 758                 spin_unlock(&rs->rs_lock);
 759
 760                 rm->m_rs = NULL;
 761                 spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 762
 763                 rds_message_put(rm);
 764         }
 765
 766         rds_wake_sk_sleep(rs);
 767
 768         while (!list_empty(&list)) {
 769                 rm = list_entry(list.next, struct rds_message, m_sock_item);
 770                 list_del_init(&rm->m_sock_item);
 771                 rds_message_wait(rm);
 772
 773                 /* just in case the code above skipped this message
 774                  * because RDS_MSG_ON_CONN wasn't set, run it again here
 775                  * taking m_rs_lock is the only thing that keeps us
 776                  * from racing with ack processing.
 777                  */
 778                 spin_lock_irqsave(&rm->m_rs_lock, flags);
 779
 780                 spin_lock(&rs->rs_lock);
 781                 __rds_send_complete(rs, rm, RDS_RDMA_CANCELED);
 782                 spin_unlock(&rs->rs_lock);
 783
 784                 rm->m_rs = NULL;
 785                 spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 786
 787                 rds_message_put(rm);
 788         }
 789 }
 790
 791 /*
 792  * we only want this to fire once so we use the callers 'queued'.  It's
 793  * possible that another thread can race with us and remove the
 794  * message from the flow with RDS_CANCEL_SENT_TO.
 795  */
 796 static int rds_send_queue_rm(struct rds_sock *rs, struct rds_connection *conn,
 797                              struct rds_conn_path *cp,
 798                              struct rds_message *rm, __be16 sport,
 799                              __be16 dport, int *queued)
 800 {
 801         unsigned long flags;
 802         u32 len;
 803
 804         if (*queued)
 805                 goto out;
 806
 807         len = be32_to_cpu(rm->m_inc.i_hdr.h_len);
 808
 809         /* this is the only place which holds both the socket's rs_lock
 810          * and the connection's c_lock */
 811         spin_lock_irqsave(&rs->rs_lock, flags);
 812
 813         /*
 814          * If there is a little space in sndbuf, we don't queue anything,
 815          * and userspace gets -EAGAIN. But poll() indicates there's send
 816          * room. This can lead to bad behavior (spinning) if snd_bytes isn't
 817          * freed up by incoming acks. So we check the *old* value of
 818          * rs_snd_bytes here to allow the last msg to exceed the buffer,
 819          * and poll() now knows no more data can be sent.
 820          */
 821         if (rs->rs_snd_bytes < rds_sk_sndbuf(rs)) {
 822                 rs->rs_snd_bytes += len;
 823
 824                 /* let recv side know we are close to send space exhaustion.
 825                  * This is probably not the optimal way to do it, as this
 826                  * means we set the flag on *all* messages as soon as our
 827                  * throughput hits a certain threshold.
 828                  */
 829                 if (rs->rs_snd_bytes >= rds_sk_sndbuf(rs) / 2)
 830                         __set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags);
 831
 832                 list_add_tail(&rm->m_sock_item, &rs->rs_send_queue);
 833                 set_bit(RDS_MSG_ON_SOCK, &rm->m_flags);
 834                 rds_message_addref(rm);
 835                 rm->m_rs = rs;
 836
 837                 /* The code ordering is a little weird, but we're
 838                    trying to minimize the time we hold c_lock */
 839                 rds_message_populate_header(&rm->m_inc.i_hdr, sport, dport, 0);
 840                 rm->m_inc.i_conn = conn;
 841                 rm->m_inc.i_conn_path = cp;
 842                 rds_message_addref(rm);
 843
 844                 spin_lock(&cp->cp_lock);
 845                 rm->m_inc.i_hdr.h_sequence = cpu_to_be64(cp->cp_next_tx_seq++);
 846                 list_add_tail(&rm->m_conn_item, &cp->cp_send_queue);
 847                 set_bit(RDS_MSG_ON_CONN, &rm->m_flags);
 848                 spin_unlock(&cp->cp_lock);
 849
 850                 rdsdebug("queued msg %p len %d, rs %p bytes %d seq %llu\n",
 851                          rm, len, rs, rs->rs_snd_bytes,
 852                          (unsigned long long)be64_to_cpu(rm->m_inc.i_hdr.h_sequence));
 853
 854                 *queued = 1;
 855         }
 856
 857         spin_unlock_irqrestore(&rs->rs_lock, flags);
 858 out:
 859         return *queued;
 860 }
 861
 862 /*
 863  * rds_message is getting to be quite complicated, and we'd like to allocate
 864  * it all in one go. This figures out how big it needs to be up front.
 865  */
 866 static int rds_rm_size(struct msghdr *msg, int data_len)
 867 {
 868         struct cmsghdr *cmsg;
 869         int size = 0;
 870         int cmsg_groups = 0;
 871         int retval;
 872
 873         for_each_cmsghdr(cmsg, msg) {
 874                 if (!CMSG_OK(msg, cmsg))
 875                         return -EINVAL;
 876
 877                 if (cmsg->cmsg_level != SOL_RDS)
 878                         continue;
 879
 880                 switch (cmsg->cmsg_type) {
 881                 case RDS_CMSG_RDMA_ARGS:
 882                         cmsg_groups |= 1;
 883                         retval = rds_rdma_extra_size(CMSG_DATA(cmsg));
 884                         if (retval < 0)
 885                                 return retval;
 886                         size += retval;
 887
 888                         break;
 889
 890                 case RDS_CMSG_RDMA_DEST:
 891                 case RDS_CMSG_RDMA_MAP:
 892                         cmsg_groups |= 2;
 893                         /* these are valid but do no add any size */
 894                         break;
 895
 896                 case RDS_CMSG_ATOMIC_CSWP:
 897                 case RDS_CMSG_ATOMIC_FADD:
 898                 case RDS_CMSG_MASKED_ATOMIC_CSWP:
 899                 case RDS_CMSG_MASKED_ATOMIC_FADD:
 900                         cmsg_groups |= 1;
 901                         size += sizeof(struct scatterlist);
 902                         break;
 903
 904                 default:
 905                         return -EINVAL;
 906                 }
 907
 908         }
 909
 910         size += ceil(data_len, PAGE_SIZE) * sizeof(struct scatterlist);
 911
 912         /* Ensure (DEST, MAP) are never used with (ARGS, ATOMIC) */
 913         if (cmsg_groups == 3)
 914                 return -EINVAL;
 915
 916         return size;
 917 }
 918
 919 static int rds_cmsg_send(struct rds_sock *rs, struct rds_message *rm,
 920                          struct msghdr *msg, int *allocated_mr)
 921 {
 922         struct cmsghdr *cmsg;
 923         int ret = 0;
 924
 925         for_each_cmsghdr(cmsg, msg) {
 926                 if (!CMSG_OK(msg, cmsg))
 927                         return -EINVAL;
 928
 929                 if (cmsg->cmsg_level != SOL_RDS)
 930                         continue;
 931
 932                 /* As a side effect, RDMA_DEST and RDMA_MAP will set
 933                  * rm->rdma.m_rdma_cookie and rm->rdma.m_rdma_mr.
 934                  */
 935                 switch (cmsg->cmsg_type) {
 936                 case RDS_CMSG_RDMA_ARGS:
 937                         ret = rds_cmsg_rdma_args(rs, rm, cmsg);
 938                         break;
 939
 940                 case RDS_CMSG_RDMA_DEST:
 941                         ret = rds_cmsg_rdma_dest(rs, rm, cmsg);
 942                         break;
 943
 944                 case RDS_CMSG_RDMA_MAP:
 945                         ret = rds_cmsg_rdma_map(rs, rm, cmsg);
 946                         if (!ret)
 947                                 *allocated_mr = 1;
 948                         break;
 949                 case RDS_CMSG_ATOMIC_CSWP:
 950                 case RDS_CMSG_ATOMIC_FADD:
 951                 case RDS_CMSG_MASKED_ATOMIC_CSWP:
 952                 case RDS_CMSG_MASKED_ATOMIC_FADD:
 953                         ret = rds_cmsg_atomic(rs, rm, cmsg);
 954                         break;
 955
 956                 default:
 957                         return -EINVAL;
 958                 }
 959
 960                 if (ret)
 961                         break;
 962         }
 963
 964         return ret;
 965 }
 966
 967 static void rds_send_ping(struct rds_connection *conn);
 968
 969 static int rds_send_mprds_hash(struct rds_sock *rs, struct rds_connection *conn)
 970 {
 971         int hash;
 972
 973         if (conn->c_npaths == 0)
 974                 hash = RDS_MPATH_HASH(rs, RDS_MPATH_WORKERS);
 975         else
 976                 hash = RDS_MPATH_HASH(rs, conn->c_npaths);
 977         if (conn->c_npaths == 0 && hash != 0) {
 978                 rds_send_ping(conn);
 979
 980                 if (conn->c_npaths == 0) {
 981                         wait_event_interruptible(conn->c_hs_waitq,
 982                                                  (conn->c_npaths != 0));
 983                 }
 984                 if (conn->c_npaths == 1)
 985                         hash = 0;
 986         }
 987         return hash;
 988 }
 989
 990 int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len)
 991 {
 992         struct sock *sk = sock->sk;
 993         struct rds_sock *rs = rds_sk_to_rs(sk);
 994         DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 995         __be32 daddr;
 996         __be16 dport;
 997         struct rds_message *rm = NULL;
 998         struct rds_connection *conn;
 999         int ret = 0;
1000         int queued = 0, allocated_mr = 0;
1001         int nonblock = msg->msg_flags & MSG_DONTWAIT;
1002         long timeo = sock_sndtimeo(sk, nonblock);
1003         struct rds_conn_path *cpath;
1004
1005         /* Mirror Linux UDP mirror of BSD error message compatibility */
1006         /* XXX: Perhaps MSG_MORE someday */
1007         if (msg->msg_flags & ~(MSG_DONTWAIT | MSG_CMSG_COMPAT)) {
1008                 ret = -EOPNOTSUPP;
1009                 goto out;
1010         }
1011
1012         if (msg->msg_namelen) {
1013                 /* XXX fail non-unicast destination IPs? */
1014                 if (msg->msg_namelen < sizeof(*usin) || usin->sin_family != AF_INET) {
1015                         ret = -EINVAL;
1016                         goto out;
1017                 }
1018                 daddr = usin->sin_addr.s_addr;
1019                 dport = usin->sin_port;
1020         } else {
1021                 /* We only care about consistency with ->connect() */
1022                 lock_sock(sk);
1023                 daddr = rs->rs_conn_addr;
1024                 dport = rs->rs_conn_port;
1025                 release_sock(sk);
1026         }
1027
1028         lock_sock(sk);
1029         if (daddr == 0 || rs->rs_bound_addr == 0) {
1030                 release_sock(sk);
1031                 ret = -ENOTCONN; /* XXX not a great errno */
1032                 goto out;
1033         }
1034         release_sock(sk);
1035
1036         if (payload_len > rds_sk_sndbuf(rs)) {
1037                 ret = -EMSGSIZE;
1038                 goto out;
1039         }
1040
1041         /* size of rm including all sgs */
1042         ret = rds_rm_size(msg, payload_len);
1043         if (ret < 0)
1044                 goto out;
1045
1046         rm = rds_message_alloc(ret, GFP_KERNEL);
1047         if (!rm) {
1048                 ret = -ENOMEM;
1049                 goto out;
1050         }
1051
1052         /* Attach data to the rm */
1053         if (payload_len) {
1054                 rm->data.op_sg = rds_message_alloc_sgs(rm, ceil(payload_len, PAGE_SIZE));
1055                 if (!rm->data.op_sg) {
1056                         ret = -ENOMEM;
1057                         goto out;
1058                 }
1059                 ret = rds_message_copy_from_user(rm, &msg->msg_iter);
1060                 if (ret)
1061                         goto out;
1062         }
1063         rm->data.op_active = 1;
1064
1065         rm->m_daddr = daddr;
1066
1067         /* rds_conn_create has a spinlock that runs with IRQ off.
1068          * Caching the conn in the socket helps a lot. */
1069         if (rs->rs_conn && rs->rs_conn->c_faddr == daddr)
1070                 conn = rs->rs_conn;
1071         else {
1072                 conn = rds_conn_create_outgoing(sock_net(sock->sk),
1073                                                 rs->rs_bound_addr, daddr,
1074                                         rs->rs_transport,
1075                                         sock->sk->sk_allocation);
1076                 if (IS_ERR(conn)) {
1077                         ret = PTR_ERR(conn);
1078                         goto out;
1079                 }
1080                 rs->rs_conn = conn;
1081         }
1082
1083         /* Parse any control messages the user may have included. */
1084         ret = rds_cmsg_send(rs, rm, msg, &allocated_mr);
1085         if (ret)
1086                 goto out;
1087
1088         if (rm->rdma.op_active && !conn->c_trans->xmit_rdma) {
1089                 printk_ratelimited(KERN_NOTICE "rdma_op %p conn xmit_rdma %p\n",
1090                                &rm->rdma, conn->c_trans->xmit_rdma);
1091                 ret = -EOPNOTSUPP;
1092                 goto out;
1093         }
1094
1095         if (rm->atomic.op_active && !conn->c_trans->xmit_atomic) {
1096                 printk_ratelimited(KERN_NOTICE "atomic_op %p conn xmit_atomic %p\n",
1097                                &rm->atomic, conn->c_trans->xmit_atomic);
1098                 ret = -EOPNOTSUPP;
1099                 goto out;
1100         }
1101
1102         if (conn->c_trans->t_mp_capable)
1103                 cpath = &conn->c_path[rds_send_mprds_hash(rs, conn)];
1104         else
1105                 cpath = &conn->c_path[0];
1106
1107         rds_conn_path_connect_if_down(cpath);
1108
1109         ret = rds_cong_wait(conn->c_fcong, dport, nonblock, rs);
1110         if (ret) {
1111                 rs->rs_seen_congestion = 1;
1112                 goto out;
1113         }
1114         while (!rds_send_queue_rm(rs, conn, cpath, rm, rs->rs_bound_port,
1115                                   dport, &queued)) {
1116                 rds_stats_inc(s_send_queue_full);
1117
1118                 if (nonblock) {
1119                         ret = -EAGAIN;
1120                         goto out;
1121                 }
1122
1123                 timeo = wait_event_interruptible_timeout(*sk_sleep(sk),
1124                                         rds_send_queue_rm(rs, conn, cpath, rm,
1125                                                           rs->rs_bound_port,
1126                                                           dport,
1127                                                           &queued),
1128                                         timeo);
1129                 rdsdebug("sendmsg woke queued %d timeo %ld\n", queued, timeo);
1130                 if (timeo > 0 || timeo == MAX_SCHEDULE_TIMEOUT)
1131                         continue;
1132
1133                 ret = timeo;
1134                 if (ret == 0)
1135                         ret = -ETIMEDOUT;
1136                 goto out;
1137         }
1138
1139         /*
1140          * By now we've committed to the send.  We reuse rds_send_worker()
1141          * to retry sends in the rds thread if the transport asks us to.
1142          */
1143         rds_stats_inc(s_send_queued);
1144
1145         ret = rds_send_xmit(cpath);
1146         if (ret == -ENOMEM || ret == -EAGAIN)
1147                 queue_delayed_work(rds_wq, &cpath->cp_send_w, 1);
1148
1149         rds_message_put(rm);
1150         return payload_len;
1151
1152 out:
1153         /* If the user included a RDMA_MAP cmsg, we allocated a MR on the fly.
1154          * If the sendmsg goes through, we keep the MR. If it fails with EAGAIN
1155          * or in any other way, we need to destroy the MR again */
1156         if (allocated_mr)
1157                 rds_rdma_unuse(rs, rds_rdma_cookie_key(rm->m_rdma_cookie), 1);
1158
1159         if (rm)
1160                 rds_message_put(rm);
1161         return ret;
1162 }
1163
1164 /*
1165  * send out a probe. Can be shared by rds_send_ping,
1166  * rds_send_pong, rds_send_hb.
1167  * rds_send_hb should use h_flags
1168  *   RDS_FLAG_HB_PING|RDS_FLAG_ACK_REQUIRED
1169  * or
1170  *   RDS_FLAG_HB_PONG|RDS_FLAG_ACK_REQUIRED
1171  */
1172 int
1173 rds_send_probe(struct rds_conn_path *cp, __be16 sport,
1174                __be16 dport, u8 h_flags)
1175 {
1176         struct rds_message *rm;
1177         unsigned long flags;
1178         int ret = 0;
1179
1180         rm = rds_message_alloc(0, GFP_ATOMIC);
1181         if (!rm) {
1182                 ret = -ENOMEM;
1183                 goto out;
1184         }
1185
1186         rm->m_daddr = cp->cp_conn->c_faddr;
1187         rm->data.op_active = 1;
1188
1189         rds_conn_path_connect_if_down(cp);
1190
1191         ret = rds_cong_wait(cp->cp_conn->c_fcong, dport, 1, NULL);
1192         if (ret)
1193                 goto out;
1194
1195         spin_lock_irqsave(&cp->cp_lock, flags);
1196         list_add_tail(&rm->m_conn_item, &cp->cp_send_queue);
1197         set_bit(RDS_MSG_ON_CONN, &rm->m_flags);
1198         rds_message_addref(rm);
1199         rm->m_inc.i_conn = cp->cp_conn;
1200         rm->m_inc.i_conn_path = cp;
1201
1202         rds_message_populate_header(&rm->m_inc.i_hdr, sport, dport,
1203                                     cp->cp_next_tx_seq);
1204         rm->m_inc.i_hdr.h_flags |= h_flags;
1205         cp->cp_next_tx_seq++;
1206
1207         if (RDS_HS_PROBE(sport, dport) && cp->cp_conn->c_trans->t_mp_capable) {
1208                 u16 npaths = RDS_MPATH_WORKERS;
1209
1210                 rds_message_add_extension(&rm->m_inc.i_hdr,
1211                                           RDS_EXTHDR_NPATHS, &npaths,
1212                                           sizeof(npaths));
1213                 rds_message_add_extension(&rm->m_inc.i_hdr,
1214                                           RDS_EXTHDR_GEN_NUM,
1215                                           &cp->cp_conn->c_my_gen_num,
1216                                           sizeof(u32));
1217         }
1218         spin_unlock_irqrestore(&cp->cp_lock, flags);
1219
1220         rds_stats_inc(s_send_queued);
1221         rds_stats_inc(s_send_pong);
1222
1223         /* schedule the send work on rds_wq */
1224         queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
1225
1226         rds_message_put(rm);
1227         return 0;
1228
1229 out:
1230         if (rm)
1231                 rds_message_put(rm);
1232         return ret;
1233 }
1234
1235 int
1236 rds_send_pong(struct rds_conn_path *cp, __be16 dport)
1237 {
1238         return rds_send_probe(cp, 0, dport, 0);
1239 }
1240
1241 void
1242 rds_send_ping(struct rds_connection *conn)
1243 {
1244         unsigned long flags;
1245         struct rds_conn_path *cp = &conn->c_path[0];
1246
1247         spin_lock_irqsave(&cp->cp_lock, flags);
1248         if (conn->c_ping_triggered) {
1249                 spin_unlock_irqrestore(&cp->cp_lock, flags);
1250                 return;
1251         }
1252         conn->c_ping_triggered = 1;
1253         spin_unlock_irqrestore(&cp->cp_lock, flags);
1254         rds_send_probe(&conn->c_path[0], RDS_FLAG_PROBE_PORT, 0, 0);
1255 }