drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  19  *
  20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21  * CA 95054 USA or visit www.sun.com if you need additional information or
  22  * have any questions.
  23  *
  24  * GPL HEADER END
  25  */
  26 /*
  27  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2012, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  *
  36  * lnet/klnds/o2iblnd/o2iblnd_cb.c
  37  *
  38  * Author: Eric Barton <eric@bartonsoftware.com>
  39  */
  40
  41 #include "o2iblnd.h"
  42
  43 static void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx);
  44
  45 static void
  46 kiblnd_tx_done(lnet_ni_t *ni, kib_tx_t *tx)
  47 {
  48         lnet_msg_t *lntmsg[2];
  49         kib_net_t *net = ni->ni_data;
  50         int rc;
  51         int i;
  52
  53         LASSERT(net != NULL);
  54         LASSERT(!in_interrupt());
  55         LASSERT(!tx->tx_queued);               /* mustn't be queued for sending */
  56         LASSERT(tx->tx_sending == 0);     /* mustn't be awaiting sent callback */
  57         LASSERT(!tx->tx_waiting);             /* mustn't be awaiting peer response */
  58         LASSERT(tx->tx_pool != NULL);
  59
  60         kiblnd_unmap_tx(ni, tx);
  61
  62         /* tx may have up to 2 lnet msgs to finalise */
  63         lntmsg[0] = tx->tx_lntmsg[0]; tx->tx_lntmsg[0] = NULL;
  64         lntmsg[1] = tx->tx_lntmsg[1]; tx->tx_lntmsg[1] = NULL;
  65         rc = tx->tx_status;
  66
  67         if (tx->tx_conn != NULL) {
  68                 LASSERT(ni == tx->tx_conn->ibc_peer->ibp_ni);
  69
  70                 kiblnd_conn_decref(tx->tx_conn);
  71                 tx->tx_conn = NULL;
  72         }
  73
  74         tx->tx_nwrq = 0;
  75         tx->tx_status = 0;
  76
  77         kiblnd_pool_free_node(&tx->tx_pool->tpo_pool, &tx->tx_list);
  78
  79         /* delay finalize until my descs have been freed */
  80         for (i = 0; i < 2; i++) {
  81                 if (lntmsg[i] == NULL)
  82                         continue;
  83
  84                 lnet_finalize(ni, lntmsg[i], rc);
  85         }
  86 }
  87
  88 void
  89 kiblnd_txlist_done(lnet_ni_t *ni, struct list_head *txlist, int status)
  90 {
  91         kib_tx_t *tx;
  92
  93         while (!list_empty(txlist)) {
  94                 tx = list_entry(txlist->next, kib_tx_t, tx_list);
  95
  96                 list_del(&tx->tx_list);
  97                 /* complete now */
  98                 tx->tx_waiting = 0;
  99                 tx->tx_status = status;
 100                 kiblnd_tx_done(ni, tx);
 101         }
 102 }
 103
 104 static kib_tx_t *
 105 kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target)
 106 {
 107         kib_net_t *net = (kib_net_t *)ni->ni_data;
 108         struct list_head *node;
 109         kib_tx_t *tx;
 110         kib_tx_poolset_t *tps;
 111
 112         tps = net->ibn_tx_ps[lnet_cpt_of_nid(target)];
 113         node = kiblnd_pool_alloc_node(&tps->tps_poolset);
 114         if (node == NULL)
 115                 return NULL;
 116         tx = container_of(node, kib_tx_t, tx_list);
 117
 118         LASSERT(tx->tx_nwrq == 0);
 119         LASSERT(!tx->tx_queued);
 120         LASSERT(tx->tx_sending == 0);
 121         LASSERT(!tx->tx_waiting);
 122         LASSERT(tx->tx_status == 0);
 123         LASSERT(tx->tx_conn == NULL);
 124         LASSERT(tx->tx_lntmsg[0] == NULL);
 125         LASSERT(tx->tx_lntmsg[1] == NULL);
 126         LASSERT(tx->tx_nfrags == 0);
 127
 128         return tx;
 129 }
 130
 131 static void
 132 kiblnd_drop_rx(kib_rx_t *rx)
 133 {
 134         kib_conn_t *conn = rx->rx_conn;
 135         struct kib_sched_info *sched = conn->ibc_sched;
 136         unsigned long flags;
 137
 138         spin_lock_irqsave(&sched->ibs_lock, flags);
 139         LASSERT(conn->ibc_nrx > 0);
 140         conn->ibc_nrx--;
 141         spin_unlock_irqrestore(&sched->ibs_lock, flags);
 142
 143         kiblnd_conn_decref(conn);
 144 }
 145
 146 int
 147 kiblnd_post_rx(kib_rx_t *rx, int credit)
 148 {
 149         kib_conn_t *conn = rx->rx_conn;
 150         kib_net_t *net = conn->ibc_peer->ibp_ni->ni_data;
 151         struct ib_recv_wr *bad_wrq = NULL;
 152         struct ib_mr *mr;
 153         int rc;
 154
 155         LASSERT(net != NULL);
 156         LASSERT(!in_interrupt());
 157         LASSERT(credit == IBLND_POSTRX_NO_CREDIT ||
 158                 credit == IBLND_POSTRX_PEER_CREDIT ||
 159                 credit == IBLND_POSTRX_RSRVD_CREDIT);
 160
 161         mr = kiblnd_find_dma_mr(conn->ibc_hdev, rx->rx_msgaddr, IBLND_MSG_SIZE);
 162         LASSERT(mr != NULL);
 163
 164         rx->rx_sge.lkey   = mr->lkey;
 165         rx->rx_sge.addr   = rx->rx_msgaddr;
 166         rx->rx_sge.length = IBLND_MSG_SIZE;
 167
 168         rx->rx_wrq.next    = NULL;
 169         rx->rx_wrq.sg_list = &rx->rx_sge;
 170         rx->rx_wrq.num_sge = 1;
 171         rx->rx_wrq.wr_id   = kiblnd_ptr2wreqid(rx, IBLND_WID_RX);
 172
 173         LASSERT(conn->ibc_state >= IBLND_CONN_INIT);
 174         LASSERT(rx->rx_nob >= 0);             /* not posted */
 175
 176         if (conn->ibc_state > IBLND_CONN_ESTABLISHED) {
 177                 kiblnd_drop_rx(rx);          /* No more posts for this rx */
 178                 return 0;
 179         }
 180
 181         rx->rx_nob = -1;                        /* flag posted */
 182
 183         /* NB: need an extra reference after ib_post_recv because we don't
 184          * own this rx (and rx::rx_conn) anymore, LU-5678.
 185          */
 186         kiblnd_conn_addref(conn);
 187         rc = ib_post_recv(conn->ibc_cmid->qp, &rx->rx_wrq, &bad_wrq);
 188         if (unlikely(rc != 0)) {
 189                 CERROR("Can't post rx for %s: %d, bad_wrq: %p\n",
 190                        libcfs_nid2str(conn->ibc_peer->ibp_nid), rc, bad_wrq);
 191                 rx->rx_nob = 0;
 192         }
 193
 194         if (conn->ibc_state < IBLND_CONN_ESTABLISHED) /* Initial post */
 195                 goto out;
 196
 197         if (unlikely(rc != 0)) {
 198                 kiblnd_close_conn(conn, rc);
 199                 kiblnd_drop_rx(rx);          /* No more posts for this rx */
 200                 goto out;
 201         }
 202
 203         if (credit == IBLND_POSTRX_NO_CREDIT)
 204                 goto out;
 205
 206         spin_lock(&conn->ibc_lock);
 207         if (credit == IBLND_POSTRX_PEER_CREDIT)
 208                 conn->ibc_outstanding_credits++;
 209         else
 210                 conn->ibc_reserved_credits++;
 211         spin_unlock(&conn->ibc_lock);
 212
 213         kiblnd_check_sends(conn);
 214 out:
 215         kiblnd_conn_decref(conn);
 216         return rc;
 217 }
 218
 219 static kib_tx_t *
 220 kiblnd_find_waiting_tx_locked(kib_conn_t *conn, int txtype, __u64 cookie)
 221 {
 222         struct list_head *tmp;
 223
 224         list_for_each(tmp, &conn->ibc_active_txs) {
 225                 kib_tx_t *tx = list_entry(tmp, kib_tx_t, tx_list);
 226
 227                 LASSERT(!tx->tx_queued);
 228                 LASSERT(tx->tx_sending != 0 || tx->tx_waiting);
 229
 230                 if (tx->tx_cookie != cookie)
 231                         continue;
 232
 233                 if (tx->tx_waiting &&
 234                     tx->tx_msg->ibm_type == txtype)
 235                         return tx;
 236
 237                 CWARN("Bad completion: %swaiting, type %x (wanted %x)\n",
 238                       tx->tx_waiting ? "" : "NOT ",
 239                       tx->tx_msg->ibm_type, txtype);
 240         }
 241         return NULL;
 242 }
 243
 244 static void
 245 kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie)
 246 {
 247         kib_tx_t *tx;
 248         lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
 249         int idle;
 250
 251         spin_lock(&conn->ibc_lock);
 252
 253         tx = kiblnd_find_waiting_tx_locked(conn, txtype, cookie);
 254         if (tx == NULL) {
 255                 spin_unlock(&conn->ibc_lock);
 256
 257                 CWARN("Unmatched completion type %x cookie %#llx from %s\n",
 258                       txtype, cookie, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 259                 kiblnd_close_conn(conn, -EPROTO);
 260                 return;
 261         }
 262
 263         if (tx->tx_status == 0) {              /* success so far */
 264                 if (status < 0) /* failed? */
 265                         tx->tx_status = status;
 266                 else if (txtype == IBLND_MSG_GET_REQ)
 267                         lnet_set_reply_msg_len(ni, tx->tx_lntmsg[1], status);
 268         }
 269
 270         tx->tx_waiting = 0;
 271
 272         idle = !tx->tx_queued && (tx->tx_sending == 0);
 273         if (idle)
 274                 list_del(&tx->tx_list);
 275
 276         spin_unlock(&conn->ibc_lock);
 277
 278         if (idle)
 279                 kiblnd_tx_done(ni, tx);
 280 }
 281
 282 static void
 283 kiblnd_send_completion(kib_conn_t *conn, int type, int status, __u64 cookie)
 284 {
 285         lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
 286         kib_tx_t *tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
 287
 288         if (tx == NULL) {
 289                 CERROR("Can't get tx for completion %x for %s\n",
 290                        type, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 291                 return;
 292         }
 293
 294         tx->tx_msg->ibm_u.completion.ibcm_status = status;
 295         tx->tx_msg->ibm_u.completion.ibcm_cookie = cookie;
 296         kiblnd_init_tx_msg(ni, tx, type, sizeof(kib_completion_msg_t));
 297
 298         kiblnd_queue_tx(tx, conn);
 299 }
 300
 301 static void
 302 kiblnd_handle_rx(kib_rx_t *rx)
 303 {
 304         kib_msg_t *msg = rx->rx_msg;
 305         kib_conn_t *conn = rx->rx_conn;
 306         lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
 307         int credits = msg->ibm_credits;
 308         kib_tx_t *tx;
 309         int rc = 0;
 310         int rc2;
 311         int post_credit;
 312
 313         LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
 314
 315         CDEBUG(D_NET, "Received %x[%d] from %s\n",
 316                msg->ibm_type, credits,
 317                libcfs_nid2str(conn->ibc_peer->ibp_nid));
 318
 319         if (credits != 0) {
 320                 /* Have I received credits that will let me send? */
 321                 spin_lock(&conn->ibc_lock);
 322
 323                 if (conn->ibc_credits + credits >
 324                     IBLND_MSG_QUEUE_SIZE(conn->ibc_version)) {
 325                         rc2 = conn->ibc_credits;
 326                         spin_unlock(&conn->ibc_lock);
 327
 328                         CERROR("Bad credits from %s: %d + %d > %d\n",
 329                                libcfs_nid2str(conn->ibc_peer->ibp_nid),
 330                                rc2, credits,
 331                                IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
 332
 333                         kiblnd_close_conn(conn, -EPROTO);
 334                         kiblnd_post_rx(rx, IBLND_POSTRX_NO_CREDIT);
 335                         return;
 336                 }
 337
 338                 conn->ibc_credits += credits;
 339
 340                 /* This ensures the credit taken by NOOP can be returned */
 341                 if (msg->ibm_type == IBLND_MSG_NOOP &&
 342                     !IBLND_OOB_CAPABLE(conn->ibc_version)) /* v1 only */
 343                         conn->ibc_outstanding_credits++;
 344
 345                 spin_unlock(&conn->ibc_lock);
 346                 kiblnd_check_sends(conn);
 347         }
 348
 349         switch (msg->ibm_type) {
 350         default:
 351                 CERROR("Bad IBLND message type %x from %s\n",
 352                        msg->ibm_type, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 353                 post_credit = IBLND_POSTRX_NO_CREDIT;
 354                 rc = -EPROTO;
 355                 break;
 356
 357         case IBLND_MSG_NOOP:
 358                 if (IBLND_OOB_CAPABLE(conn->ibc_version)) {
 359                         post_credit = IBLND_POSTRX_NO_CREDIT;
 360                         break;
 361                 }
 362
 363                 if (credits != 0) /* credit already posted */
 364                         post_credit = IBLND_POSTRX_NO_CREDIT;
 365                 else          /* a keepalive NOOP */
 366                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 367                 break;
 368
 369         case IBLND_MSG_IMMEDIATE:
 370                 post_credit = IBLND_POSTRX_DONT_POST;
 371                 rc = lnet_parse(ni, &msg->ibm_u.immediate.ibim_hdr,
 372                                 msg->ibm_srcnid, rx, 0);
 373                 if (rc < 0)                  /* repost on error */
 374                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 375                 break;
 376
 377         case IBLND_MSG_PUT_REQ:
 378                 post_credit = IBLND_POSTRX_DONT_POST;
 379                 rc = lnet_parse(ni, &msg->ibm_u.putreq.ibprm_hdr,
 380                                 msg->ibm_srcnid, rx, 1);
 381                 if (rc < 0)                  /* repost on error */
 382                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 383                 break;
 384
 385         case IBLND_MSG_PUT_NAK:
 386                 CWARN("PUT_NACK from %s\n",
 387                       libcfs_nid2str(conn->ibc_peer->ibp_nid));
 388                 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
 389                 kiblnd_handle_completion(conn, IBLND_MSG_PUT_REQ,
 390                                          msg->ibm_u.completion.ibcm_status,
 391                                          msg->ibm_u.completion.ibcm_cookie);
 392                 break;
 393
 394         case IBLND_MSG_PUT_ACK:
 395                 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
 396
 397                 spin_lock(&conn->ibc_lock);
 398                 tx = kiblnd_find_waiting_tx_locked(conn, IBLND_MSG_PUT_REQ,
 399                                         msg->ibm_u.putack.ibpam_src_cookie);
 400                 if (tx != NULL)
 401                         list_del(&tx->tx_list);
 402                 spin_unlock(&conn->ibc_lock);
 403
 404                 if (tx == NULL) {
 405                         CERROR("Unmatched PUT_ACK from %s\n",
 406                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
 407                         rc = -EPROTO;
 408                         break;
 409                 }
 410
 411                 LASSERT(tx->tx_waiting);
 412                 /* CAVEAT EMPTOR: I could be racing with tx_complete, but...
 413                  * (a) I can overwrite tx_msg since my peer has received it!
 414                  * (b) tx_waiting set tells tx_complete() it's not done. */
 415
 416                 tx->tx_nwrq = 0;                /* overwrite PUT_REQ */
 417
 418                 rc2 = kiblnd_init_rdma(conn, tx, IBLND_MSG_PUT_DONE,
 419                                        kiblnd_rd_size(&msg->ibm_u.putack.ibpam_rd),
 420                                        &msg->ibm_u.putack.ibpam_rd,
 421                                        msg->ibm_u.putack.ibpam_dst_cookie);
 422                 if (rc2 < 0)
 423                         CERROR("Can't setup rdma for PUT to %s: %d\n",
 424                                libcfs_nid2str(conn->ibc_peer->ibp_nid), rc2);
 425
 426                 spin_lock(&conn->ibc_lock);
 427                 tx->tx_waiting = 0;     /* clear waiting and queue atomically */
 428                 kiblnd_queue_tx_locked(tx, conn);
 429                 spin_unlock(&conn->ibc_lock);
 430                 break;
 431
 432         case IBLND_MSG_PUT_DONE:
 433                 post_credit = IBLND_POSTRX_PEER_CREDIT;
 434                 kiblnd_handle_completion(conn, IBLND_MSG_PUT_ACK,
 435                                          msg->ibm_u.completion.ibcm_status,
 436                                          msg->ibm_u.completion.ibcm_cookie);
 437                 break;
 438
 439         case IBLND_MSG_GET_REQ:
 440                 post_credit = IBLND_POSTRX_DONT_POST;
 441                 rc = lnet_parse(ni, &msg->ibm_u.get.ibgm_hdr,
 442                                 msg->ibm_srcnid, rx, 1);
 443                 if (rc < 0)                  /* repost on error */
 444                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 445                 break;
 446
 447         case IBLND_MSG_GET_DONE:
 448                 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
 449                 kiblnd_handle_completion(conn, IBLND_MSG_GET_REQ,
 450                                          msg->ibm_u.completion.ibcm_status,
 451                                          msg->ibm_u.completion.ibcm_cookie);
 452                 break;
 453         }
 454
 455         if (rc < 0)                          /* protocol error */
 456                 kiblnd_close_conn(conn, rc);
 457
 458         if (post_credit != IBLND_POSTRX_DONT_POST)
 459                 kiblnd_post_rx(rx, post_credit);
 460 }
 461
 462 static void
 463 kiblnd_rx_complete(kib_rx_t *rx, int status, int nob)
 464 {
 465         kib_msg_t *msg = rx->rx_msg;
 466         kib_conn_t *conn = rx->rx_conn;
 467         lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
 468         kib_net_t *net = ni->ni_data;
 469         int rc;
 470         int err = -EIO;
 471
 472         LASSERT(net != NULL);
 473         LASSERT(rx->rx_nob < 0);               /* was posted */
 474         rx->rx_nob = 0;                  /* isn't now */
 475
 476         if (conn->ibc_state > IBLND_CONN_ESTABLISHED)
 477                 goto ignore;
 478
 479         if (status != IB_WC_SUCCESS) {
 480                 CNETERR("Rx from %s failed: %d\n",
 481                         libcfs_nid2str(conn->ibc_peer->ibp_nid), status);
 482                 goto failed;
 483         }
 484
 485         LASSERT(nob >= 0);
 486         rx->rx_nob = nob;
 487
 488         rc = kiblnd_unpack_msg(msg, rx->rx_nob);
 489         if (rc != 0) {
 490                 CERROR("Error %d unpacking rx from %s\n",
 491                         rc, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 492                 goto failed;
 493         }
 494
 495         if (msg->ibm_srcnid != conn->ibc_peer->ibp_nid ||
 496             msg->ibm_dstnid != ni->ni_nid ||
 497             msg->ibm_srcstamp != conn->ibc_incarnation ||
 498             msg->ibm_dststamp != net->ibn_incarnation) {
 499                 CERROR("Stale rx from %s\n",
 500                         libcfs_nid2str(conn->ibc_peer->ibp_nid));
 501                 err = -ESTALE;
 502                 goto failed;
 503         }
 504
 505         /* set time last known alive */
 506         kiblnd_peer_alive(conn->ibc_peer);
 507
 508         /* racing with connection establishment/teardown! */
 509
 510         if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
 511                 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
 512                 unsigned long flags;
 513
 514                 write_lock_irqsave(g_lock, flags);
 515                 /* must check holding global lock to eliminate race */
 516                 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
 517                         list_add_tail(&rx->rx_list, &conn->ibc_early_rxs);
 518                         write_unlock_irqrestore(g_lock, flags);
 519                         return;
 520                 }
 521                 write_unlock_irqrestore(g_lock, flags);
 522         }
 523         kiblnd_handle_rx(rx);
 524         return;
 525
 526  failed:
 527         CDEBUG(D_NET, "rx %p conn %p\n", rx, conn);
 528         kiblnd_close_conn(conn, err);
 529  ignore:
 530         kiblnd_drop_rx(rx);                  /* Don't re-post rx. */
 531 }
 532
 533 static struct page *
 534 kiblnd_kvaddr_to_page(unsigned long vaddr)
 535 {
 536         struct page *page;
 537
 538         if (is_vmalloc_addr((void *)vaddr)) {
 539                 page = vmalloc_to_page((void *)vaddr);
 540                 LASSERT(page != NULL);
 541                 return page;
 542         }
 543 #ifdef CONFIG_HIGHMEM
 544         if (vaddr >= PKMAP_BASE &&
 545             vaddr < (PKMAP_BASE + LAST_PKMAP * PAGE_SIZE)) {
 546                 /* No highmem pages only used for bulk (kiov) I/O */
 547                 CERROR("find page for address in highmem\n");
 548                 LBUG();
 549         }
 550 #endif
 551         page = virt_to_page(vaddr);
 552         LASSERT(page != NULL);
 553         return page;
 554 }
 555
 556 static int
 557 kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob)
 558 {
 559         kib_hca_dev_t *hdev;
 560         __u64 *pages = tx->tx_pages;
 561         kib_fmr_poolset_t *fps;
 562         int npages;
 563         int size;
 564         int cpt;
 565         int rc;
 566         int i;
 567
 568         LASSERT(tx->tx_pool != NULL);
 569         LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL);
 570
 571         hdev = tx->tx_pool->tpo_hdev;
 572
 573         for (i = 0, npages = 0; i < rd->rd_nfrags; i++) {
 574                 for (size = 0; size <  rd->rd_frags[i].rf_nob;
 575                                size += hdev->ibh_page_size) {
 576                         pages[npages++] = (rd->rd_frags[i].rf_addr &
 577                                             hdev->ibh_page_mask) + size;
 578                 }
 579         }
 580
 581         cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
 582
 583         fps = net->ibn_fmr_ps[cpt];
 584         rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->fmr);
 585         if (rc != 0) {
 586                 CERROR("Can't map %d pages: %d\n", npages, rc);
 587                 return rc;
 588         }
 589
 590         /* If rd is not tx_rd, it's going to get sent to a peer, who will need
 591          * the rkey */
 592         rd->rd_key = (rd != tx->tx_rd) ? tx->fmr.fmr_pfmr->fmr->rkey :
 593                                          tx->fmr.fmr_pfmr->fmr->lkey;
 594         rd->rd_frags[0].rf_addr &= ~hdev->ibh_page_mask;
 595         rd->rd_frags[0].rf_nob = nob;
 596         rd->rd_nfrags = 1;
 597
 598         return 0;
 599 }
 600
 601 static void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx)
 602 {
 603         kib_net_t *net = ni->ni_data;
 604
 605         LASSERT(net != NULL);
 606
 607         if (net->ibn_fmr_ps && tx->fmr.fmr_pfmr) {
 608                 kiblnd_fmr_pool_unmap(&tx->fmr, tx->tx_status);
 609                 tx->fmr.fmr_pfmr = NULL;
 610         }
 611
 612         if (tx->tx_nfrags != 0) {
 613                 kiblnd_dma_unmap_sg(tx->tx_pool->tpo_hdev->ibh_ibdev,
 614                                     tx->tx_frags, tx->tx_nfrags, tx->tx_dmadir);
 615                 tx->tx_nfrags = 0;
 616         }
 617 }
 618
 619 static int kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
 620                          int nfrags)
 621 {
 622         kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
 623         kib_net_t *net = ni->ni_data;
 624         struct ib_mr *mr    = NULL;
 625         __u32 nob;
 626         int i;
 627
 628         /* If rd is not tx_rd, it's going to get sent to a peer and I'm the
 629          * RDMA sink */
 630         tx->tx_dmadir = (rd != tx->tx_rd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
 631         tx->tx_nfrags = nfrags;
 632
 633         rd->rd_nfrags = kiblnd_dma_map_sg(hdev->ibh_ibdev, tx->tx_frags,
 634                                           tx->tx_nfrags, tx->tx_dmadir);
 635
 636         for (i = 0, nob = 0; i < rd->rd_nfrags; i++) {
 637                 rd->rd_frags[i].rf_nob  = kiblnd_sg_dma_len(
 638                         hdev->ibh_ibdev, &tx->tx_frags[i]);
 639                 rd->rd_frags[i].rf_addr = kiblnd_sg_dma_address(
 640                         hdev->ibh_ibdev, &tx->tx_frags[i]);
 641                 nob += rd->rd_frags[i].rf_nob;
 642         }
 643
 644         /* looking for pre-mapping MR */
 645         mr = kiblnd_find_rd_dma_mr(hdev, rd);
 646         if (mr != NULL) {
 647                 /* found pre-mapping MR */
 648                 rd->rd_key = (rd != tx->tx_rd) ? mr->rkey : mr->lkey;
 649                 return 0;
 650         }
 651
 652         if (net->ibn_fmr_ps != NULL)
 653                 return kiblnd_fmr_map_tx(net, tx, rd, nob);
 654
 655         return -EINVAL;
 656 }
 657
 658 static int
 659 kiblnd_setup_rd_iov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
 660                     unsigned int niov, struct kvec *iov, int offset, int nob)
 661 {
 662         kib_net_t *net = ni->ni_data;
 663         struct page *page;
 664         struct scatterlist *sg;
 665         unsigned long vaddr;
 666         int fragnob;
 667         int page_offset;
 668
 669         LASSERT(nob > 0);
 670         LASSERT(niov > 0);
 671         LASSERT(net != NULL);
 672
 673         while (offset >= iov->iov_len) {
 674                 offset -= iov->iov_len;
 675                 niov--;
 676                 iov++;
 677                 LASSERT(niov > 0);
 678         }
 679
 680         sg = tx->tx_frags;
 681         do {
 682                 LASSERT(niov > 0);
 683
 684                 vaddr = ((unsigned long)iov->iov_base) + offset;
 685                 page_offset = vaddr & (PAGE_SIZE - 1);
 686                 page = kiblnd_kvaddr_to_page(vaddr);
 687                 if (page == NULL) {
 688                         CERROR("Can't find page\n");
 689                         return -EFAULT;
 690                 }
 691
 692                 fragnob = min((int)(iov->iov_len - offset), nob);
 693                 fragnob = min(fragnob, (int)PAGE_SIZE - page_offset);
 694
 695                 sg_set_page(sg, page, fragnob, page_offset);
 696                 sg++;
 697
 698                 if (offset + fragnob < iov->iov_len) {
 699                         offset += fragnob;
 700                 } else {
 701                         offset = 0;
 702                         iov++;
 703                         niov--;
 704                 }
 705                 nob -= fragnob;
 706         } while (nob > 0);
 707
 708         return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
 709 }
 710
 711 static int
 712 kiblnd_setup_rd_kiov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
 713                       int nkiov, lnet_kiov_t *kiov, int offset, int nob)
 714 {
 715         kib_net_t *net = ni->ni_data;
 716         struct scatterlist *sg;
 717         int fragnob;
 718
 719         CDEBUG(D_NET, "niov %d offset %d nob %d\n", nkiov, offset, nob);
 720
 721         LASSERT(nob > 0);
 722         LASSERT(nkiov > 0);
 723         LASSERT(net != NULL);
 724
 725         while (offset >= kiov->kiov_len) {
 726                 offset -= kiov->kiov_len;
 727                 nkiov--;
 728                 kiov++;
 729                 LASSERT(nkiov > 0);
 730         }
 731
 732         sg = tx->tx_frags;
 733         do {
 734                 LASSERT(nkiov > 0);
 735
 736                 fragnob = min((int)(kiov->kiov_len - offset), nob);
 737
 738                 sg_set_page(sg, kiov->kiov_page, fragnob,
 739                             kiov->kiov_offset + offset);
 740                 sg++;
 741
 742                 offset = 0;
 743                 kiov++;
 744                 nkiov--;
 745                 nob -= fragnob;
 746         } while (nob > 0);
 747
 748         return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
 749 }
 750
 751 static int
 752 kiblnd_post_tx_locked(kib_conn_t *conn, kib_tx_t *tx, int credit)
 753         __releases(conn->ibc_lock)
 754         __acquires(conn->ibc_lock)
 755 {
 756         kib_msg_t *msg = tx->tx_msg;
 757         kib_peer_t *peer = conn->ibc_peer;
 758         int ver = conn->ibc_version;
 759         int rc;
 760         int done;
 761         struct ib_send_wr *bad_wrq;
 762
 763         LASSERT(tx->tx_queued);
 764         /* We rely on this for QP sizing */
 765         LASSERT(tx->tx_nwrq > 0);
 766         LASSERT(tx->tx_nwrq <= 1 + IBLND_RDMA_FRAGS(ver));
 767
 768         LASSERT(credit == 0 || credit == 1);
 769         LASSERT(conn->ibc_outstanding_credits >= 0);
 770         LASSERT(conn->ibc_outstanding_credits <= IBLND_MSG_QUEUE_SIZE(ver));
 771         LASSERT(conn->ibc_credits >= 0);
 772         LASSERT(conn->ibc_credits <= IBLND_MSG_QUEUE_SIZE(ver));
 773
 774         if (conn->ibc_nsends_posted == IBLND_CONCURRENT_SENDS(ver)) {
 775                 /* tx completions outstanding... */
 776                 CDEBUG(D_NET, "%s: posted enough\n",
 777                        libcfs_nid2str(peer->ibp_nid));
 778                 return -EAGAIN;
 779         }
 780
 781         if (credit != 0 && conn->ibc_credits == 0) {   /* no credits */
 782                 CDEBUG(D_NET, "%s: no credits\n",
 783                        libcfs_nid2str(peer->ibp_nid));
 784                 return -EAGAIN;
 785         }
 786
 787         if (credit != 0 && !IBLND_OOB_CAPABLE(ver) &&
 788             conn->ibc_credits == 1 &&   /* last credit reserved */
 789             msg->ibm_type != IBLND_MSG_NOOP) {      /* for NOOP */
 790                 CDEBUG(D_NET, "%s: not using last credit\n",
 791                        libcfs_nid2str(peer->ibp_nid));
 792                 return -EAGAIN;
 793         }
 794
 795         /* NB don't drop ibc_lock before bumping tx_sending */
 796         list_del(&tx->tx_list);
 797         tx->tx_queued = 0;
 798
 799         if (msg->ibm_type == IBLND_MSG_NOOP &&
 800             (!kiblnd_need_noop(conn) ||     /* redundant NOOP */
 801              (IBLND_OOB_CAPABLE(ver) && /* posted enough NOOP */
 802               conn->ibc_noops_posted == IBLND_OOB_MSGS(ver)))) {
 803                 /* OK to drop when posted enough NOOPs, since
 804                  * kiblnd_check_sends will queue NOOP again when
 805                  * posted NOOPs complete */
 806                 spin_unlock(&conn->ibc_lock);
 807                 kiblnd_tx_done(peer->ibp_ni, tx);
 808                 spin_lock(&conn->ibc_lock);
 809                 CDEBUG(D_NET, "%s(%d): redundant or enough NOOP\n",
 810                        libcfs_nid2str(peer->ibp_nid),
 811                        conn->ibc_noops_posted);
 812                 return 0;
 813         }
 814
 815         kiblnd_pack_msg(peer->ibp_ni, msg, ver, conn->ibc_outstanding_credits,
 816                         peer->ibp_nid, conn->ibc_incarnation);
 817
 818         conn->ibc_credits -= credit;
 819         conn->ibc_outstanding_credits = 0;
 820         conn->ibc_nsends_posted++;
 821         if (msg->ibm_type == IBLND_MSG_NOOP)
 822                 conn->ibc_noops_posted++;
 823
 824         /* CAVEAT EMPTOR!  This tx could be the PUT_DONE of an RDMA
 825          * PUT.  If so, it was first queued here as a PUT_REQ, sent and
 826          * stashed on ibc_active_txs, matched by an incoming PUT_ACK,
 827          * and then re-queued here.  It's (just) possible that
 828          * tx_sending is non-zero if we've not done the tx_complete()
 829          * from the first send; hence the ++ rather than = below. */
 830         tx->tx_sending++;
 831         list_add(&tx->tx_list, &conn->ibc_active_txs);
 832
 833         /* I'm still holding ibc_lock! */
 834         if (conn->ibc_state != IBLND_CONN_ESTABLISHED) {
 835                 rc = -ECONNABORTED;
 836         } else if (tx->tx_pool->tpo_pool.po_failed ||
 837                  conn->ibc_hdev != tx->tx_pool->tpo_hdev) {
 838                 /* close_conn will launch failover */
 839                 rc = -ENETDOWN;
 840         } else {
 841                 rc = ib_post_send(conn->ibc_cmid->qp, &tx->tx_wrq->wr, &bad_wrq);
 842         }
 843
 844         conn->ibc_last_send = jiffies;
 845
 846         if (rc == 0)
 847                 return 0;
 848
 849         /* NB credits are transferred in the actual
 850          * message, which can only be the last work item */
 851         conn->ibc_credits += credit;
 852         conn->ibc_outstanding_credits += msg->ibm_credits;
 853         conn->ibc_nsends_posted--;
 854         if (msg->ibm_type == IBLND_MSG_NOOP)
 855                 conn->ibc_noops_posted--;
 856
 857         tx->tx_status = rc;
 858         tx->tx_waiting = 0;
 859         tx->tx_sending--;
 860
 861         done = (tx->tx_sending == 0);
 862         if (done)
 863                 list_del(&tx->tx_list);
 864
 865         spin_unlock(&conn->ibc_lock);
 866
 867         if (conn->ibc_state == IBLND_CONN_ESTABLISHED)
 868                 CERROR("Error %d posting transmit to %s\n",
 869                        rc, libcfs_nid2str(peer->ibp_nid));
 870         else
 871                 CDEBUG(D_NET, "Error %d posting transmit to %s\n",
 872                        rc, libcfs_nid2str(peer->ibp_nid));
 873
 874         kiblnd_close_conn(conn, rc);
 875
 876         if (done)
 877                 kiblnd_tx_done(peer->ibp_ni, tx);
 878
 879         spin_lock(&conn->ibc_lock);
 880
 881         return -EIO;
 882 }
 883
 884 void
 885 kiblnd_check_sends(kib_conn_t *conn)
 886 {
 887         int ver = conn->ibc_version;
 888         lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
 889         kib_tx_t *tx;
 890
 891         /* Don't send anything until after the connection is established */
 892         if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
 893                 CDEBUG(D_NET, "%s too soon\n",
 894                        libcfs_nid2str(conn->ibc_peer->ibp_nid));
 895                 return;
 896         }
 897
 898         spin_lock(&conn->ibc_lock);
 899
 900         LASSERT(conn->ibc_nsends_posted <= IBLND_CONCURRENT_SENDS(ver));
 901         LASSERT(!IBLND_OOB_CAPABLE(ver) ||
 902                  conn->ibc_noops_posted <= IBLND_OOB_MSGS(ver));
 903         LASSERT(conn->ibc_reserved_credits >= 0);
 904
 905         while (conn->ibc_reserved_credits > 0 &&
 906                !list_empty(&conn->ibc_tx_queue_rsrvd)) {
 907                 tx = list_entry(conn->ibc_tx_queue_rsrvd.next,
 908                                     kib_tx_t, tx_list);
 909                 list_del(&tx->tx_list);
 910                 list_add_tail(&tx->tx_list, &conn->ibc_tx_queue);
 911                 conn->ibc_reserved_credits--;
 912         }
 913
 914         if (kiblnd_need_noop(conn)) {
 915                 spin_unlock(&conn->ibc_lock);
 916
 917                 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
 918                 if (tx != NULL)
 919                         kiblnd_init_tx_msg(ni, tx, IBLND_MSG_NOOP, 0);
 920
 921                 spin_lock(&conn->ibc_lock);
 922                 if (tx != NULL)
 923                         kiblnd_queue_tx_locked(tx, conn);
 924         }
 925
 926         kiblnd_conn_addref(conn); /* 1 ref for me.... (see b21911) */
 927
 928         for (;;) {
 929                 int credit;
 930
 931                 if (!list_empty(&conn->ibc_tx_queue_nocred)) {
 932                         credit = 0;
 933                         tx = list_entry(conn->ibc_tx_queue_nocred.next,
 934                                             kib_tx_t, tx_list);
 935                 } else if (!list_empty(&conn->ibc_tx_noops)) {
 936                         LASSERT(!IBLND_OOB_CAPABLE(ver));
 937                         credit = 1;
 938                         tx = list_entry(conn->ibc_tx_noops.next,
 939                                         kib_tx_t, tx_list);
 940                 } else if (!list_empty(&conn->ibc_tx_queue)) {
 941                         credit = 1;
 942                         tx = list_entry(conn->ibc_tx_queue.next,
 943                                             kib_tx_t, tx_list);
 944                 } else
 945                         break;
 946
 947                 if (kiblnd_post_tx_locked(conn, tx, credit) != 0)
 948                         break;
 949         }
 950
 951         spin_unlock(&conn->ibc_lock);
 952
 953         kiblnd_conn_decref(conn); /* ...until here */
 954 }
 955
 956 static void
 957 kiblnd_tx_complete(kib_tx_t *tx, int status)
 958 {
 959         int failed = (status != IB_WC_SUCCESS);
 960         kib_conn_t *conn = tx->tx_conn;
 961         int idle;
 962
 963         LASSERT(tx->tx_sending > 0);
 964
 965         if (failed) {
 966                 if (conn->ibc_state == IBLND_CONN_ESTABLISHED)
 967                         CNETERR("Tx -> %s cookie %#llx sending %d waiting %d: failed %d\n",
 968                                 libcfs_nid2str(conn->ibc_peer->ibp_nid),
 969                                 tx->tx_cookie, tx->tx_sending, tx->tx_waiting,
 970                                 status);
 971
 972                 kiblnd_close_conn(conn, -EIO);
 973         } else {
 974                 kiblnd_peer_alive(conn->ibc_peer);
 975         }
 976
 977         spin_lock(&conn->ibc_lock);
 978
 979         /* I could be racing with rdma completion.  Whoever makes 'tx' idle
 980          * gets to free it, which also drops its ref on 'conn'. */
 981
 982         tx->tx_sending--;
 983         conn->ibc_nsends_posted--;
 984         if (tx->tx_msg->ibm_type == IBLND_MSG_NOOP)
 985                 conn->ibc_noops_posted--;
 986
 987         if (failed) {
 988                 tx->tx_waiting = 0;          /* don't wait for peer */
 989                 tx->tx_status = -EIO;
 990         }
 991
 992         idle = (tx->tx_sending == 0) &&  /* This is the final callback */
 993                !tx->tx_waiting &&              /* Not waiting for peer */
 994                !tx->tx_queued;            /* Not re-queued (PUT_DONE) */
 995         if (idle)
 996                 list_del(&tx->tx_list);
 997
 998         kiblnd_conn_addref(conn);              /* 1 ref for me.... */
 999
1000         spin_unlock(&conn->ibc_lock);
1001
1002         if (idle)
1003                 kiblnd_tx_done(conn->ibc_peer->ibp_ni, tx);
1004
1005         kiblnd_check_sends(conn);
1006
1007         kiblnd_conn_decref(conn);              /* ...until here */
1008 }
1009
1010 void
1011 kiblnd_init_tx_msg(lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob)
1012 {
1013         kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
1014         struct ib_sge *sge = &tx->tx_sge[tx->tx_nwrq];
1015         struct ib_rdma_wr *wrq = &tx->tx_wrq[tx->tx_nwrq];
1016         int nob = offsetof(kib_msg_t, ibm_u) + body_nob;
1017         struct ib_mr *mr;
1018
1019         LASSERT(tx->tx_nwrq >= 0);
1020         LASSERT(tx->tx_nwrq < IBLND_MAX_RDMA_FRAGS + 1);
1021         LASSERT(nob <= IBLND_MSG_SIZE);
1022
1023         kiblnd_init_msg(tx->tx_msg, type, body_nob);
1024
1025         mr = kiblnd_find_dma_mr(hdev, tx->tx_msgaddr, nob);
1026         LASSERT(mr != NULL);
1027
1028         sge->lkey   = mr->lkey;
1029         sge->addr   = tx->tx_msgaddr;
1030         sge->length = nob;
1031
1032         memset(wrq, 0, sizeof(*wrq));
1033
1034         wrq->wr.next       = NULL;
1035         wrq->wr.wr_id      = kiblnd_ptr2wreqid(tx, IBLND_WID_TX);
1036         wrq->wr.sg_list    = sge;
1037         wrq->wr.num_sge    = 1;
1038         wrq->wr.opcode     = IB_WR_SEND;
1039         wrq->wr.send_flags = IB_SEND_SIGNALED;
1040
1041         tx->tx_nwrq++;
1042 }
1043
1044 int
1045 kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type,
1046                   int resid, kib_rdma_desc_t *dstrd, __u64 dstcookie)
1047 {
1048         kib_msg_t *ibmsg = tx->tx_msg;
1049         kib_rdma_desc_t *srcrd = tx->tx_rd;
1050         struct ib_sge *sge = &tx->tx_sge[0];
1051         struct ib_rdma_wr *wrq = &tx->tx_wrq[0], *next;
1052         int rc  = resid;
1053         int srcidx;
1054         int dstidx;
1055         int wrknob;
1056
1057         LASSERT(!in_interrupt());
1058         LASSERT(tx->tx_nwrq == 0);
1059         LASSERT(type == IBLND_MSG_GET_DONE ||
1060                  type == IBLND_MSG_PUT_DONE);
1061
1062         srcidx = dstidx = 0;
1063
1064         while (resid > 0) {
1065                 if (srcidx >= srcrd->rd_nfrags) {
1066                         CERROR("Src buffer exhausted: %d frags\n", srcidx);
1067                         rc = -EPROTO;
1068                         break;
1069                 }
1070
1071                 if (dstidx == dstrd->rd_nfrags) {
1072                         CERROR("Dst buffer exhausted: %d frags\n", dstidx);
1073                         rc = -EPROTO;
1074                         break;
1075                 }
1076
1077                 if (tx->tx_nwrq == IBLND_RDMA_FRAGS(conn->ibc_version)) {
1078                         CERROR("RDMA too fragmented for %s (%d): %d/%d src %d/%d dst frags\n",
1079                                libcfs_nid2str(conn->ibc_peer->ibp_nid),
1080                                IBLND_RDMA_FRAGS(conn->ibc_version),
1081                                srcidx, srcrd->rd_nfrags,
1082                                dstidx, dstrd->rd_nfrags);
1083                         rc = -EMSGSIZE;
1084                         break;
1085                 }
1086
1087                 wrknob = min(min(kiblnd_rd_frag_size(srcrd, srcidx),
1088                                  kiblnd_rd_frag_size(dstrd, dstidx)),
1089                              (__u32) resid);
1090
1091                 sge = &tx->tx_sge[tx->tx_nwrq];
1092                 sge->addr   = kiblnd_rd_frag_addr(srcrd, srcidx);
1093                 sge->lkey   = kiblnd_rd_frag_key(srcrd, srcidx);
1094                 sge->length = wrknob;
1095
1096                 wrq = &tx->tx_wrq[tx->tx_nwrq];
1097                 next = wrq + 1;
1098
1099                 wrq->wr.next       = &next->wr;
1100                 wrq->wr.wr_id      = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA);
1101                 wrq->wr.sg_list    = sge;
1102                 wrq->wr.num_sge    = 1;
1103                 wrq->wr.opcode     = IB_WR_RDMA_WRITE;
1104                 wrq->wr.send_flags = 0;
1105
1106                 wrq->remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx);
1107                 wrq->rkey        = kiblnd_rd_frag_key(dstrd, dstidx);
1108
1109                 srcidx = kiblnd_rd_consume_frag(srcrd, srcidx, wrknob);
1110                 dstidx = kiblnd_rd_consume_frag(dstrd, dstidx, wrknob);
1111
1112                 resid -= wrknob;
1113
1114                 tx->tx_nwrq++;
1115                 wrq++;
1116                 sge++;
1117         }
1118
1119         if (rc < 0)                          /* no RDMA if completing with failure */
1120                 tx->tx_nwrq = 0;
1121
1122         ibmsg->ibm_u.completion.ibcm_status = rc;
1123         ibmsg->ibm_u.completion.ibcm_cookie = dstcookie;
1124         kiblnd_init_tx_msg(conn->ibc_peer->ibp_ni, tx,
1125                            type, sizeof(kib_completion_msg_t));
1126
1127         return rc;
1128 }
1129
1130 void
1131 kiblnd_queue_tx_locked(kib_tx_t *tx, kib_conn_t *conn)
1132 {
1133         struct list_head *q;
1134
1135         LASSERT(tx->tx_nwrq > 0);             /* work items set up */
1136         LASSERT(!tx->tx_queued);               /* not queued for sending already */
1137         LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1138
1139         tx->tx_queued = 1;
1140         tx->tx_deadline = jiffies + (*kiblnd_tunables.kib_timeout * HZ);
1141
1142         if (tx->tx_conn == NULL) {
1143                 kiblnd_conn_addref(conn);
1144                 tx->tx_conn = conn;
1145                 LASSERT(tx->tx_msg->ibm_type != IBLND_MSG_PUT_DONE);
1146         } else {
1147                 /* PUT_DONE first attached to conn as a PUT_REQ */
1148                 LASSERT(tx->tx_conn == conn);
1149                 LASSERT(tx->tx_msg->ibm_type == IBLND_MSG_PUT_DONE);
1150         }
1151
1152         switch (tx->tx_msg->ibm_type) {
1153         default:
1154                 LBUG();
1155
1156         case IBLND_MSG_PUT_REQ:
1157         case IBLND_MSG_GET_REQ:
1158                 q = &conn->ibc_tx_queue_rsrvd;
1159                 break;
1160
1161         case IBLND_MSG_PUT_NAK:
1162         case IBLND_MSG_PUT_ACK:
1163         case IBLND_MSG_PUT_DONE:
1164         case IBLND_MSG_GET_DONE:
1165                 q = &conn->ibc_tx_queue_nocred;
1166                 break;
1167
1168         case IBLND_MSG_NOOP:
1169                 if (IBLND_OOB_CAPABLE(conn->ibc_version))
1170                         q = &conn->ibc_tx_queue_nocred;
1171                 else
1172                         q = &conn->ibc_tx_noops;
1173                 break;
1174
1175         case IBLND_MSG_IMMEDIATE:
1176                 q = &conn->ibc_tx_queue;
1177                 break;
1178         }
1179
1180         list_add_tail(&tx->tx_list, q);
1181 }
1182
1183 void
1184 kiblnd_queue_tx(kib_tx_t *tx, kib_conn_t *conn)
1185 {
1186         spin_lock(&conn->ibc_lock);
1187         kiblnd_queue_tx_locked(tx, conn);
1188         spin_unlock(&conn->ibc_lock);
1189
1190         kiblnd_check_sends(conn);
1191 }
1192
1193 static int kiblnd_resolve_addr(struct rdma_cm_id *cmid,
1194                                struct sockaddr_in *srcaddr,
1195                                struct sockaddr_in *dstaddr,
1196                                int timeout_ms)
1197 {
1198         unsigned short port;
1199         int rc;
1200
1201         /* allow the port to be reused */
1202         rc = rdma_set_reuseaddr(cmid, 1);
1203         if (rc != 0) {
1204                 CERROR("Unable to set reuse on cmid: %d\n", rc);
1205                 return rc;
1206         }
1207
1208         /* look for a free privileged port */
1209         for (port = PROT_SOCK-1; port > 0; port--) {
1210                 srcaddr->sin_port = htons(port);
1211                 rc = rdma_resolve_addr(cmid,
1212                                        (struct sockaddr *)srcaddr,
1213                                        (struct sockaddr *)dstaddr,
1214                                        timeout_ms);
1215                 if (rc == 0) {
1216                         CDEBUG(D_NET, "bound to port %hu\n", port);
1217                         return 0;
1218                 } else if (rc == -EADDRINUSE || rc == -EADDRNOTAVAIL) {
1219                         CDEBUG(D_NET, "bind to port %hu failed: %d\n",
1220                                port, rc);
1221                 } else {
1222                         return rc;
1223                 }
1224         }
1225
1226         CERROR("Failed to bind to a free privileged port\n");
1227         return rc;
1228 }
1229
1230 static void
1231 kiblnd_connect_peer(kib_peer_t *peer)
1232 {
1233         struct rdma_cm_id *cmid;
1234         kib_dev_t *dev;
1235         kib_net_t *net = peer->ibp_ni->ni_data;
1236         struct sockaddr_in srcaddr;
1237         struct sockaddr_in dstaddr;
1238         int rc;
1239
1240         LASSERT(net != NULL);
1241         LASSERT(peer->ibp_connecting > 0);
1242
1243         cmid = kiblnd_rdma_create_id(kiblnd_cm_callback, peer, RDMA_PS_TCP,
1244                                      IB_QPT_RC);
1245
1246         if (IS_ERR(cmid)) {
1247                 CERROR("Can't create CMID for %s: %ld\n",
1248                        libcfs_nid2str(peer->ibp_nid), PTR_ERR(cmid));
1249                 rc = PTR_ERR(cmid);
1250                 goto failed;
1251         }
1252
1253         dev = net->ibn_dev;
1254         memset(&srcaddr, 0, sizeof(srcaddr));
1255         srcaddr.sin_family = AF_INET;
1256         srcaddr.sin_addr.s_addr = htonl(dev->ibd_ifip);
1257
1258         memset(&dstaddr, 0, sizeof(dstaddr));
1259         dstaddr.sin_family = AF_INET;
1260         dstaddr.sin_port = htons(*kiblnd_tunables.kib_service);
1261         dstaddr.sin_addr.s_addr = htonl(LNET_NIDADDR(peer->ibp_nid));
1262
1263         kiblnd_peer_addref(peer);              /* cmid's ref */
1264
1265         if (*kiblnd_tunables.kib_use_priv_port) {
1266                 rc = kiblnd_resolve_addr(cmid, &srcaddr, &dstaddr,
1267                                          *kiblnd_tunables.kib_timeout * 1000);
1268         } else {
1269                 rc = rdma_resolve_addr(cmid,
1270                                        (struct sockaddr *)&srcaddr,
1271                                        (struct sockaddr *)&dstaddr,
1272                                        *kiblnd_tunables.kib_timeout * 1000);
1273         }
1274         if (rc != 0) {
1275                 /* Can't initiate address resolution:  */
1276                 CERROR("Can't resolve addr for %s: %d\n",
1277                        libcfs_nid2str(peer->ibp_nid), rc);
1278                 goto failed2;
1279         }
1280
1281         LASSERT(cmid->device != NULL);
1282         CDEBUG(D_NET, "%s: connection bound to %s:%pI4h:%s\n",
1283                libcfs_nid2str(peer->ibp_nid), dev->ibd_ifname,
1284                &dev->ibd_ifip, cmid->device->name);
1285
1286         return;
1287
1288  failed2:
1289         kiblnd_peer_decref(peer);              /* cmid's ref */
1290         rdma_destroy_id(cmid);
1291  failed:
1292         kiblnd_peer_connect_failed(peer, 1, rc);
1293 }
1294
1295 void
1296 kiblnd_launch_tx(lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid)
1297 {
1298         kib_peer_t *peer;
1299         kib_peer_t *peer2;
1300         kib_conn_t *conn;
1301         rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
1302         unsigned long flags;
1303         int rc;
1304
1305         /* If I get here, I've committed to send, so I complete the tx with
1306          * failure on any problems */
1307
1308         LASSERT(tx == NULL || tx->tx_conn == NULL); /* only set when assigned a conn */
1309         LASSERT(tx == NULL || tx->tx_nwrq > 0);     /* work items have been set up */
1310
1311         /* First time, just use a read lock since I expect to find my peer
1312          * connected */
1313         read_lock_irqsave(g_lock, flags);
1314
1315         peer = kiblnd_find_peer_locked(nid);
1316         if (peer != NULL && !list_empty(&peer->ibp_conns)) {
1317                 /* Found a peer with an established connection */
1318                 conn = kiblnd_get_conn_locked(peer);
1319                 kiblnd_conn_addref(conn); /* 1 ref for me... */
1320
1321                 read_unlock_irqrestore(g_lock, flags);
1322
1323                 if (tx != NULL)
1324                         kiblnd_queue_tx(tx, conn);
1325                 kiblnd_conn_decref(conn); /* ...to here */
1326                 return;
1327         }
1328
1329         read_unlock(g_lock);
1330         /* Re-try with a write lock */
1331         write_lock(g_lock);
1332
1333         peer = kiblnd_find_peer_locked(nid);
1334         if (peer != NULL) {
1335                 if (list_empty(&peer->ibp_conns)) {
1336                         /* found a peer, but it's still connecting... */
1337                         LASSERT(peer->ibp_connecting != 0 ||
1338                                  peer->ibp_accepting != 0);
1339                         if (tx != NULL)
1340                                 list_add_tail(&tx->tx_list,
1341                                                   &peer->ibp_tx_queue);
1342                         write_unlock_irqrestore(g_lock, flags);
1343                 } else {
1344                         conn = kiblnd_get_conn_locked(peer);
1345                         kiblnd_conn_addref(conn); /* 1 ref for me... */
1346
1347                         write_unlock_irqrestore(g_lock, flags);
1348
1349                         if (tx != NULL)
1350                                 kiblnd_queue_tx(tx, conn);
1351                         kiblnd_conn_decref(conn); /* ...to here */
1352                 }
1353                 return;
1354         }
1355
1356         write_unlock_irqrestore(g_lock, flags);
1357
1358         /* Allocate a peer ready to add to the peer table and retry */
1359         rc = kiblnd_create_peer(ni, &peer, nid);
1360         if (rc != 0) {
1361                 CERROR("Can't create peer %s\n", libcfs_nid2str(nid));
1362                 if (tx != NULL) {
1363                         tx->tx_status = -EHOSTUNREACH;
1364                         tx->tx_waiting = 0;
1365                         kiblnd_tx_done(ni, tx);
1366                 }
1367                 return;
1368         }
1369
1370         write_lock_irqsave(g_lock, flags);
1371
1372         peer2 = kiblnd_find_peer_locked(nid);
1373         if (peer2 != NULL) {
1374                 if (list_empty(&peer2->ibp_conns)) {
1375                         /* found a peer, but it's still connecting... */
1376                         LASSERT(peer2->ibp_connecting != 0 ||
1377                                  peer2->ibp_accepting != 0);
1378                         if (tx != NULL)
1379                                 list_add_tail(&tx->tx_list,
1380                                                   &peer2->ibp_tx_queue);
1381                         write_unlock_irqrestore(g_lock, flags);
1382                 } else {
1383                         conn = kiblnd_get_conn_locked(peer2);
1384                         kiblnd_conn_addref(conn); /* 1 ref for me... */
1385
1386                         write_unlock_irqrestore(g_lock, flags);
1387
1388                         if (tx != NULL)
1389                                 kiblnd_queue_tx(tx, conn);
1390                         kiblnd_conn_decref(conn); /* ...to here */
1391                 }
1392
1393                 kiblnd_peer_decref(peer);
1394                 return;
1395         }
1396
1397         /* Brand new peer */
1398         LASSERT(peer->ibp_connecting == 0);
1399         peer->ibp_connecting = 1;
1400
1401         /* always called with a ref on ni, which prevents ni being shutdown */
1402         LASSERT(((kib_net_t *)ni->ni_data)->ibn_shutdown == 0);
1403
1404         if (tx != NULL)
1405                 list_add_tail(&tx->tx_list, &peer->ibp_tx_queue);
1406
1407         kiblnd_peer_addref(peer);
1408         list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
1409
1410         write_unlock_irqrestore(g_lock, flags);
1411
1412         kiblnd_connect_peer(peer);
1413         kiblnd_peer_decref(peer);
1414 }
1415
1416 int
1417 kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg)
1418 {
1419         lnet_hdr_t *hdr = &lntmsg->msg_hdr;
1420         int type = lntmsg->msg_type;
1421         lnet_process_id_t target = lntmsg->msg_target;
1422         int target_is_router = lntmsg->msg_target_is_router;
1423         int routing = lntmsg->msg_routing;
1424         unsigned int payload_niov = lntmsg->msg_niov;
1425         struct kvec *payload_iov = lntmsg->msg_iov;
1426         lnet_kiov_t *payload_kiov = lntmsg->msg_kiov;
1427         unsigned int payload_offset = lntmsg->msg_offset;
1428         unsigned int payload_nob = lntmsg->msg_len;
1429         kib_msg_t *ibmsg;
1430         kib_rdma_desc_t  *rd;
1431         kib_tx_t *tx;
1432         int nob;
1433         int rc;
1434
1435         /* NB 'private' is different depending on what we're sending.... */
1436
1437         CDEBUG(D_NET, "sending %d bytes in %d frags to %s\n",
1438                payload_nob, payload_niov, libcfs_id2str(target));
1439
1440         LASSERT(payload_nob == 0 || payload_niov > 0);
1441         LASSERT(payload_niov <= LNET_MAX_IOV);
1442
1443         /* Thread context */
1444         LASSERT(!in_interrupt());
1445         /* payload is either all vaddrs or all pages */
1446         LASSERT(!(payload_kiov != NULL && payload_iov != NULL));
1447
1448         switch (type) {
1449         default:
1450                 LBUG();
1451                 return -EIO;
1452
1453         case LNET_MSG_ACK:
1454                 LASSERT(payload_nob == 0);
1455                 break;
1456
1457         case LNET_MSG_GET:
1458                 if (routing || target_is_router)
1459                         break;            /* send IMMEDIATE */
1460
1461                 /* is the REPLY message too small for RDMA? */
1462                 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[lntmsg->msg_md->md_length]);
1463                 if (nob <= IBLND_MSG_SIZE)
1464                         break;            /* send IMMEDIATE */
1465
1466                 tx = kiblnd_get_idle_tx(ni, target.nid);
1467                 if (tx == NULL) {
1468                         CERROR("Can't allocate txd for GET to %s\n",
1469                                libcfs_nid2str(target.nid));
1470                         return -ENOMEM;
1471                 }
1472
1473                 ibmsg = tx->tx_msg;
1474                 rd = &ibmsg->ibm_u.get.ibgm_rd;
1475                 if ((lntmsg->msg_md->md_options & LNET_MD_KIOV) == 0)
1476                         rc = kiblnd_setup_rd_iov(ni, tx, rd,
1477                                                  lntmsg->msg_md->md_niov,
1478                                                  lntmsg->msg_md->md_iov.iov,
1479                                                  0, lntmsg->msg_md->md_length);
1480                 else
1481                         rc = kiblnd_setup_rd_kiov(ni, tx, rd,
1482                                                   lntmsg->msg_md->md_niov,
1483                                                   lntmsg->msg_md->md_iov.kiov,
1484                                                   0, lntmsg->msg_md->md_length);
1485                 if (rc != 0) {
1486                         CERROR("Can't setup GET sink for %s: %d\n",
1487                                libcfs_nid2str(target.nid), rc);
1488                         kiblnd_tx_done(ni, tx);
1489                         return -EIO;
1490                 }
1491
1492                 nob = offsetof(kib_get_msg_t, ibgm_rd.rd_frags[rd->rd_nfrags]);
1493                 ibmsg->ibm_u.get.ibgm_cookie = tx->tx_cookie;
1494                 ibmsg->ibm_u.get.ibgm_hdr = *hdr;
1495
1496                 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_GET_REQ, nob);
1497
1498                 tx->tx_lntmsg[1] = lnet_create_reply_msg(ni, lntmsg);
1499                 if (tx->tx_lntmsg[1] == NULL) {
1500                         CERROR("Can't create reply for GET -> %s\n",
1501                                libcfs_nid2str(target.nid));
1502                         kiblnd_tx_done(ni, tx);
1503                         return -EIO;
1504                 }
1505
1506                 tx->tx_lntmsg[0] = lntmsg;      /* finalise lntmsg[0,1] on completion */
1507                 tx->tx_waiting = 1;          /* waiting for GET_DONE */
1508                 kiblnd_launch_tx(ni, tx, target.nid);
1509                 return 0;
1510
1511         case LNET_MSG_REPLY:
1512         case LNET_MSG_PUT:
1513                 /* Is the payload small enough not to need RDMA? */
1514                 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[payload_nob]);
1515                 if (nob <= IBLND_MSG_SIZE)
1516                         break;            /* send IMMEDIATE */
1517
1518                 tx = kiblnd_get_idle_tx(ni, target.nid);
1519                 if (tx == NULL) {
1520                         CERROR("Can't allocate %s txd for %s\n",
1521                                type == LNET_MSG_PUT ? "PUT" : "REPLY",
1522                                libcfs_nid2str(target.nid));
1523                         return -ENOMEM;
1524                 }
1525
1526                 if (payload_kiov == NULL)
1527                         rc = kiblnd_setup_rd_iov(ni, tx, tx->tx_rd,
1528                                                  payload_niov, payload_iov,
1529                                                  payload_offset, payload_nob);
1530                 else
1531                         rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1532                                                   payload_niov, payload_kiov,
1533                                                   payload_offset, payload_nob);
1534                 if (rc != 0) {
1535                         CERROR("Can't setup PUT src for %s: %d\n",
1536                                libcfs_nid2str(target.nid), rc);
1537                         kiblnd_tx_done(ni, tx);
1538                         return -EIO;
1539                 }
1540
1541                 ibmsg = tx->tx_msg;
1542                 ibmsg->ibm_u.putreq.ibprm_hdr = *hdr;
1543                 ibmsg->ibm_u.putreq.ibprm_cookie = tx->tx_cookie;
1544                 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_PUT_REQ, sizeof(kib_putreq_msg_t));
1545
1546                 tx->tx_lntmsg[0] = lntmsg;      /* finalise lntmsg on completion */
1547                 tx->tx_waiting = 1;          /* waiting for PUT_{ACK,NAK} */
1548                 kiblnd_launch_tx(ni, tx, target.nid);
1549                 return 0;
1550         }
1551
1552         /* send IMMEDIATE */
1553
1554         LASSERT(offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[payload_nob])
1555                  <= IBLND_MSG_SIZE);
1556
1557         tx = kiblnd_get_idle_tx(ni, target.nid);
1558         if (tx == NULL) {
1559                 CERROR("Can't send %d to %s: tx descs exhausted\n",
1560                         type, libcfs_nid2str(target.nid));
1561                 return -ENOMEM;
1562         }
1563
1564         ibmsg = tx->tx_msg;
1565         ibmsg->ibm_u.immediate.ibim_hdr = *hdr;
1566
1567         if (payload_kiov != NULL)
1568                 lnet_copy_kiov2flat(IBLND_MSG_SIZE, ibmsg,
1569                                     offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1570                                     payload_niov, payload_kiov,
1571                                     payload_offset, payload_nob);
1572         else
1573                 lnet_copy_iov2flat(IBLND_MSG_SIZE, ibmsg,
1574                                    offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1575                                    payload_niov, payload_iov,
1576                                    payload_offset, payload_nob);
1577
1578         nob = offsetof(kib_immediate_msg_t, ibim_payload[payload_nob]);
1579         kiblnd_init_tx_msg(ni, tx, IBLND_MSG_IMMEDIATE, nob);
1580
1581         tx->tx_lntmsg[0] = lntmsg;            /* finalise lntmsg on completion */
1582         kiblnd_launch_tx(ni, tx, target.nid);
1583         return 0;
1584 }
1585
1586 static void
1587 kiblnd_reply(lnet_ni_t *ni, kib_rx_t *rx, lnet_msg_t *lntmsg)
1588 {
1589         lnet_process_id_t target = lntmsg->msg_target;
1590         unsigned int niov = lntmsg->msg_niov;
1591         struct kvec *iov = lntmsg->msg_iov;
1592         lnet_kiov_t *kiov = lntmsg->msg_kiov;
1593         unsigned int offset = lntmsg->msg_offset;
1594         unsigned int nob = lntmsg->msg_len;
1595         kib_tx_t *tx;
1596         int rc;
1597
1598         tx = kiblnd_get_idle_tx(ni, rx->rx_conn->ibc_peer->ibp_nid);
1599         if (tx == NULL) {
1600                 CERROR("Can't get tx for REPLY to %s\n",
1601                        libcfs_nid2str(target.nid));
1602                 goto failed_0;
1603         }
1604
1605         if (nob == 0)
1606                 rc = 0;
1607         else if (kiov == NULL)
1608                 rc = kiblnd_setup_rd_iov(ni, tx, tx->tx_rd,
1609                                          niov, iov, offset, nob);
1610         else
1611                 rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1612                                           niov, kiov, offset, nob);
1613
1614         if (rc != 0) {
1615                 CERROR("Can't setup GET src for %s: %d\n",
1616                        libcfs_nid2str(target.nid), rc);
1617                 goto failed_1;
1618         }
1619
1620         rc = kiblnd_init_rdma(rx->rx_conn, tx,
1621                               IBLND_MSG_GET_DONE, nob,
1622                               &rx->rx_msg->ibm_u.get.ibgm_rd,
1623                               rx->rx_msg->ibm_u.get.ibgm_cookie);
1624         if (rc < 0) {
1625                 CERROR("Can't setup rdma for GET from %s: %d\n",
1626                        libcfs_nid2str(target.nid), rc);
1627                 goto failed_1;
1628         }
1629
1630         if (nob == 0) {
1631                 /* No RDMA: local completion may happen now! */
1632                 lnet_finalize(ni, lntmsg, 0);
1633         } else {
1634                 /* RDMA: lnet_finalize(lntmsg) when it
1635                  * completes */
1636                 tx->tx_lntmsg[0] = lntmsg;
1637         }
1638
1639         kiblnd_queue_tx(tx, rx->rx_conn);
1640         return;
1641
1642  failed_1:
1643         kiblnd_tx_done(ni, tx);
1644  failed_0:
1645         lnet_finalize(ni, lntmsg, -EIO);
1646 }
1647
1648 int
1649 kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed,
1650              unsigned int niov, struct kvec *iov, lnet_kiov_t *kiov,
1651              unsigned int offset, unsigned int mlen, unsigned int rlen)
1652 {
1653         kib_rx_t *rx = private;
1654         kib_msg_t *rxmsg = rx->rx_msg;
1655         kib_conn_t *conn = rx->rx_conn;
1656         kib_tx_t *tx;
1657         int nob;
1658         int post_credit = IBLND_POSTRX_PEER_CREDIT;
1659         int rc = 0;
1660
1661         LASSERT(mlen <= rlen);
1662         LASSERT(!in_interrupt());
1663         /* Either all pages or all vaddrs */
1664         LASSERT(!(kiov != NULL && iov != NULL));
1665
1666         switch (rxmsg->ibm_type) {
1667         default:
1668                 LBUG();
1669
1670         case IBLND_MSG_IMMEDIATE:
1671                 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[rlen]);
1672                 if (nob > rx->rx_nob) {
1673                         CERROR("Immediate message from %s too big: %d(%d)\n",
1674                                 libcfs_nid2str(rxmsg->ibm_u.immediate.ibim_hdr.src_nid),
1675                                 nob, rx->rx_nob);
1676                         rc = -EPROTO;
1677                         break;
1678                 }
1679
1680                 if (kiov != NULL)
1681                         lnet_copy_flat2kiov(niov, kiov, offset,
1682                                             IBLND_MSG_SIZE, rxmsg,
1683                                             offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1684                                             mlen);
1685                 else
1686                         lnet_copy_flat2iov(niov, iov, offset,
1687                                            IBLND_MSG_SIZE, rxmsg,
1688                                            offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1689                                            mlen);
1690                 lnet_finalize(ni, lntmsg, 0);
1691                 break;
1692
1693         case IBLND_MSG_PUT_REQ: {
1694                 kib_msg_t       *txmsg;
1695                 kib_rdma_desc_t *rd;
1696
1697                 if (mlen == 0) {
1698                         lnet_finalize(ni, lntmsg, 0);
1699                         kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, 0,
1700                                                rxmsg->ibm_u.putreq.ibprm_cookie);
1701                         break;
1702                 }
1703
1704                 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
1705                 if (tx == NULL) {
1706                         CERROR("Can't allocate tx for %s\n",
1707                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
1708                         /* Not replying will break the connection */
1709                         rc = -ENOMEM;
1710                         break;
1711                 }
1712
1713                 txmsg = tx->tx_msg;
1714                 rd = &txmsg->ibm_u.putack.ibpam_rd;
1715                 if (kiov == NULL)
1716                         rc = kiblnd_setup_rd_iov(ni, tx, rd,
1717                                                  niov, iov, offset, mlen);
1718                 else
1719                         rc = kiblnd_setup_rd_kiov(ni, tx, rd,
1720                                                   niov, kiov, offset, mlen);
1721                 if (rc != 0) {
1722                         CERROR("Can't setup PUT sink for %s: %d\n",
1723                                libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
1724                         kiblnd_tx_done(ni, tx);
1725                         /* tell peer it's over */
1726                         kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, rc,
1727                                                rxmsg->ibm_u.putreq.ibprm_cookie);
1728                         break;
1729                 }
1730
1731                 nob = offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[rd->rd_nfrags]);
1732                 txmsg->ibm_u.putack.ibpam_src_cookie = rxmsg->ibm_u.putreq.ibprm_cookie;
1733                 txmsg->ibm_u.putack.ibpam_dst_cookie = tx->tx_cookie;
1734
1735                 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_PUT_ACK, nob);
1736
1737                 tx->tx_lntmsg[0] = lntmsg;      /* finalise lntmsg on completion */
1738                 tx->tx_waiting = 1;          /* waiting for PUT_DONE */
1739                 kiblnd_queue_tx(tx, conn);
1740
1741                 /* reposted buffer reserved for PUT_DONE */
1742                 post_credit = IBLND_POSTRX_NO_CREDIT;
1743                 break;
1744                 }
1745
1746         case IBLND_MSG_GET_REQ:
1747                 if (lntmsg != NULL) {
1748                         /* Optimized GET; RDMA lntmsg's payload */
1749                         kiblnd_reply(ni, rx, lntmsg);
1750                 } else {
1751                         /* GET didn't match anything */
1752                         kiblnd_send_completion(rx->rx_conn, IBLND_MSG_GET_DONE,
1753                                                -ENODATA,
1754                                                rxmsg->ibm_u.get.ibgm_cookie);
1755                 }
1756                 break;
1757         }
1758
1759         kiblnd_post_rx(rx, post_credit);
1760         return rc;
1761 }
1762
1763 int
1764 kiblnd_thread_start(int (*fn)(void *arg), void *arg, char *name)
1765 {
1766         struct task_struct *task = kthread_run(fn, arg, "%s", name);
1767
1768         if (IS_ERR(task))
1769                 return PTR_ERR(task);
1770
1771         atomic_inc(&kiblnd_data.kib_nthreads);
1772         return 0;
1773 }
1774
1775 static void
1776 kiblnd_thread_fini(void)
1777 {
1778         atomic_dec(&kiblnd_data.kib_nthreads);
1779 }
1780
1781 void
1782 kiblnd_peer_alive(kib_peer_t *peer)
1783 {
1784         /* This is racy, but everyone's only writing cfs_time_current() */
1785         peer->ibp_last_alive = cfs_time_current();
1786         mb();
1787 }
1788
1789 static void
1790 kiblnd_peer_notify(kib_peer_t *peer)
1791 {
1792         int error = 0;
1793         unsigned long last_alive = 0;
1794         unsigned long flags;
1795
1796         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1797
1798         if (list_empty(&peer->ibp_conns) &&
1799             peer->ibp_accepting == 0 &&
1800             peer->ibp_connecting == 0 &&
1801             peer->ibp_error != 0) {
1802                 error = peer->ibp_error;
1803                 peer->ibp_error = 0;
1804
1805                 last_alive = peer->ibp_last_alive;
1806         }
1807
1808         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1809
1810         if (error != 0)
1811                 lnet_notify(peer->ibp_ni,
1812                             peer->ibp_nid, 0, last_alive);
1813 }
1814
1815 void
1816 kiblnd_close_conn_locked(kib_conn_t *conn, int error)
1817 {
1818         /* This just does the immediate housekeeping.  'error' is zero for a
1819          * normal shutdown which can happen only after the connection has been
1820          * established.  If the connection is established, schedule the
1821          * connection to be finished off by the connd.  Otherwise the connd is
1822          * already dealing with it (either to set it up or tear it down).
1823          * Caller holds kib_global_lock exclusively in irq context */
1824         kib_peer_t *peer = conn->ibc_peer;
1825         kib_dev_t *dev;
1826         unsigned long flags;
1827
1828         LASSERT(error != 0 || conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1829
1830         if (error != 0 && conn->ibc_comms_error == 0)
1831                 conn->ibc_comms_error = error;
1832
1833         if (conn->ibc_state != IBLND_CONN_ESTABLISHED)
1834                 return; /* already being handled  */
1835
1836         if (error == 0 &&
1837             list_empty(&conn->ibc_tx_noops) &&
1838             list_empty(&conn->ibc_tx_queue) &&
1839             list_empty(&conn->ibc_tx_queue_rsrvd) &&
1840             list_empty(&conn->ibc_tx_queue_nocred) &&
1841             list_empty(&conn->ibc_active_txs)) {
1842                 CDEBUG(D_NET, "closing conn to %s\n",
1843                        libcfs_nid2str(peer->ibp_nid));
1844         } else {
1845                 CNETERR("Closing conn to %s: error %d%s%s%s%s%s\n",
1846                        libcfs_nid2str(peer->ibp_nid), error,
1847                        list_empty(&conn->ibc_tx_queue) ? "" : "(sending)",
1848                        list_empty(&conn->ibc_tx_noops) ? "" : "(sending_noops)",
1849                        list_empty(&conn->ibc_tx_queue_rsrvd) ? "" : "(sending_rsrvd)",
1850                        list_empty(&conn->ibc_tx_queue_nocred) ? "" : "(sending_nocred)",
1851                        list_empty(&conn->ibc_active_txs) ? "" : "(waiting)");
1852         }
1853
1854         dev = ((kib_net_t *)peer->ibp_ni->ni_data)->ibn_dev;
1855         list_del(&conn->ibc_list);
1856         /* connd (see below) takes over ibc_list's ref */
1857
1858         if (list_empty(&peer->ibp_conns) &&    /* no more conns */
1859             kiblnd_peer_active(peer)) {  /* still in peer table */
1860                 kiblnd_unlink_peer_locked(peer);
1861
1862                 /* set/clear error on last conn */
1863                 peer->ibp_error = conn->ibc_comms_error;
1864         }
1865
1866         kiblnd_set_conn_state(conn, IBLND_CONN_CLOSING);
1867
1868         if (error != 0 &&
1869             kiblnd_dev_can_failover(dev)) {
1870                 list_add_tail(&dev->ibd_fail_list,
1871                               &kiblnd_data.kib_failed_devs);
1872                 wake_up(&kiblnd_data.kib_failover_waitq);
1873         }
1874
1875         spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
1876
1877         list_add_tail(&conn->ibc_list, &kiblnd_data.kib_connd_conns);
1878         wake_up(&kiblnd_data.kib_connd_waitq);
1879
1880         spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
1881 }
1882
1883 void
1884 kiblnd_close_conn(kib_conn_t *conn, int error)
1885 {
1886         unsigned long flags;
1887
1888         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1889
1890         kiblnd_close_conn_locked(conn, error);
1891
1892         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1893 }
1894
1895 static void
1896 kiblnd_handle_early_rxs(kib_conn_t *conn)
1897 {
1898         unsigned long flags;
1899         kib_rx_t *rx;
1900         kib_rx_t *tmp;
1901
1902         LASSERT(!in_interrupt());
1903         LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1904
1905         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1906         list_for_each_entry_safe(rx, tmp, &conn->ibc_early_rxs, rx_list) {
1907                 list_del(&rx->rx_list);
1908                 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1909
1910                 kiblnd_handle_rx(rx);
1911
1912                 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1913         }
1914         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1915 }
1916
1917 static void
1918 kiblnd_abort_txs(kib_conn_t *conn, struct list_head *txs)
1919 {
1920         LIST_HEAD(zombies);
1921         struct list_head *tmp;
1922         struct list_head *nxt;
1923         kib_tx_t *tx;
1924
1925         spin_lock(&conn->ibc_lock);
1926
1927         list_for_each_safe(tmp, nxt, txs) {
1928                 tx = list_entry(tmp, kib_tx_t, tx_list);
1929
1930                 if (txs == &conn->ibc_active_txs) {
1931                         LASSERT(!tx->tx_queued);
1932                         LASSERT(tx->tx_waiting ||
1933                                  tx->tx_sending != 0);
1934                 } else {
1935                         LASSERT(tx->tx_queued);
1936                 }
1937
1938                 tx->tx_status = -ECONNABORTED;
1939                 tx->tx_waiting = 0;
1940
1941                 if (tx->tx_sending == 0) {
1942                         tx->tx_queued = 0;
1943                         list_del(&tx->tx_list);
1944                         list_add(&tx->tx_list, &zombies);
1945                 }
1946         }
1947
1948         spin_unlock(&conn->ibc_lock);
1949
1950         kiblnd_txlist_done(conn->ibc_peer->ibp_ni, &zombies, -ECONNABORTED);
1951 }
1952
1953 static void
1954 kiblnd_finalise_conn(kib_conn_t *conn)
1955 {
1956         LASSERT(!in_interrupt());
1957         LASSERT(conn->ibc_state > IBLND_CONN_INIT);
1958
1959         kiblnd_set_conn_state(conn, IBLND_CONN_DISCONNECTED);
1960
1961         /* abort_receives moves QP state to IB_QPS_ERR.  This is only required
1962          * for connections that didn't get as far as being connected, because
1963          * rdma_disconnect() does this for free. */
1964         kiblnd_abort_receives(conn);
1965
1966         /* Complete all tx descs not waiting for sends to complete.
1967          * NB we should be safe from RDMA now that the QP has changed state */
1968
1969         kiblnd_abort_txs(conn, &conn->ibc_tx_noops);
1970         kiblnd_abort_txs(conn, &conn->ibc_tx_queue);
1971         kiblnd_abort_txs(conn, &conn->ibc_tx_queue_rsrvd);
1972         kiblnd_abort_txs(conn, &conn->ibc_tx_queue_nocred);
1973         kiblnd_abort_txs(conn, &conn->ibc_active_txs);
1974
1975         kiblnd_handle_early_rxs(conn);
1976 }
1977
1978 void
1979 kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error)
1980 {
1981         LIST_HEAD(zombies);
1982         unsigned long flags;
1983
1984         LASSERT(error != 0);
1985         LASSERT(!in_interrupt());
1986
1987         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1988
1989         if (active) {
1990                 LASSERT(peer->ibp_connecting > 0);
1991                 peer->ibp_connecting--;
1992         } else {
1993                 LASSERT(peer->ibp_accepting > 0);
1994                 peer->ibp_accepting--;
1995         }
1996
1997         if (peer->ibp_connecting != 0 ||
1998             peer->ibp_accepting != 0) {
1999                 /* another connection attempt under way... */
2000                 write_unlock_irqrestore(&kiblnd_data.kib_global_lock,
2001                                             flags);
2002                 return;
2003         }
2004
2005         if (list_empty(&peer->ibp_conns)) {
2006                 /* Take peer's blocked transmits to complete with error */
2007                 list_add(&zombies, &peer->ibp_tx_queue);
2008                 list_del_init(&peer->ibp_tx_queue);
2009
2010                 if (kiblnd_peer_active(peer))
2011                         kiblnd_unlink_peer_locked(peer);
2012
2013                 peer->ibp_error = error;
2014         } else {
2015                 /* Can't have blocked transmits if there are connections */
2016                 LASSERT(list_empty(&peer->ibp_tx_queue));
2017         }
2018
2019         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2020
2021         kiblnd_peer_notify(peer);
2022
2023         if (list_empty(&zombies))
2024                 return;
2025
2026         CNETERR("Deleting messages for %s: connection failed\n",
2027                 libcfs_nid2str(peer->ibp_nid));
2028
2029         kiblnd_txlist_done(peer->ibp_ni, &zombies, -EHOSTUNREACH);
2030 }
2031
2032 void
2033 kiblnd_connreq_done(kib_conn_t *conn, int status)
2034 {
2035         kib_peer_t *peer = conn->ibc_peer;
2036         kib_tx_t *tx;
2037         kib_tx_t *tmp;
2038         struct list_head txs;
2039         unsigned long flags;
2040         int active;
2041
2042         active = (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2043
2044         CDEBUG(D_NET, "%s: active(%d), version(%x), status(%d)\n",
2045                libcfs_nid2str(peer->ibp_nid), active,
2046                conn->ibc_version, status);
2047
2048         LASSERT(!in_interrupt());
2049         LASSERT((conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT &&
2050                   peer->ibp_connecting > 0) ||
2051                  (conn->ibc_state == IBLND_CONN_PASSIVE_WAIT &&
2052                   peer->ibp_accepting > 0));
2053
2054         LIBCFS_FREE(conn->ibc_connvars, sizeof(*conn->ibc_connvars));
2055         conn->ibc_connvars = NULL;
2056
2057         if (status != 0) {
2058                 /* failed to establish connection */
2059                 kiblnd_peer_connect_failed(peer, active, status);
2060                 kiblnd_finalise_conn(conn);
2061                 return;
2062         }
2063
2064         /* connection established */
2065         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2066
2067         conn->ibc_last_send = jiffies;
2068         kiblnd_set_conn_state(conn, IBLND_CONN_ESTABLISHED);
2069         kiblnd_peer_alive(peer);
2070
2071         /* Add conn to peer's list and nuke any dangling conns from a different
2072          * peer instance... */
2073         kiblnd_conn_addref(conn);              /* +1 ref for ibc_list */
2074         list_add(&conn->ibc_list, &peer->ibp_conns);
2075         if (active)
2076                 peer->ibp_connecting--;
2077         else
2078                 peer->ibp_accepting--;
2079
2080         if (peer->ibp_version == 0) {
2081                 peer->ibp_version     = conn->ibc_version;
2082                 peer->ibp_incarnation = conn->ibc_incarnation;
2083         }
2084
2085         if (peer->ibp_version     != conn->ibc_version ||
2086             peer->ibp_incarnation != conn->ibc_incarnation) {
2087                 kiblnd_close_stale_conns_locked(peer, conn->ibc_version,
2088                                                 conn->ibc_incarnation);
2089                 peer->ibp_version     = conn->ibc_version;
2090                 peer->ibp_incarnation = conn->ibc_incarnation;
2091         }
2092
2093         /* grab pending txs while I have the lock */
2094         list_add(&txs, &peer->ibp_tx_queue);
2095         list_del_init(&peer->ibp_tx_queue);
2096
2097         if (!kiblnd_peer_active(peer) ||        /* peer has been deleted */
2098             conn->ibc_comms_error != 0) {       /* error has happened already */
2099                 lnet_ni_t *ni = peer->ibp_ni;
2100
2101                 /* start to shut down connection */
2102                 kiblnd_close_conn_locked(conn, -ECONNABORTED);
2103                 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2104
2105                 kiblnd_txlist_done(ni, &txs, -ECONNABORTED);
2106
2107                 return;
2108         }
2109
2110         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2111
2112         /* Schedule blocked txs */
2113         spin_lock(&conn->ibc_lock);
2114         list_for_each_entry_safe(tx, tmp, &txs, tx_list) {
2115                 list_del(&tx->tx_list);
2116
2117                 kiblnd_queue_tx_locked(tx, conn);
2118         }
2119         spin_unlock(&conn->ibc_lock);
2120
2121         kiblnd_check_sends(conn);
2122
2123         /* schedule blocked rxs */
2124         kiblnd_handle_early_rxs(conn);
2125 }
2126
2127 static void
2128 kiblnd_reject(struct rdma_cm_id *cmid, kib_rej_t *rej)
2129 {
2130         int rc;
2131
2132         rc = rdma_reject(cmid, rej, sizeof(*rej));
2133
2134         if (rc != 0)
2135                 CWARN("Error %d sending reject\n", rc);
2136 }
2137
2138 static int
2139 kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob)
2140 {
2141         rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
2142         kib_msg_t *reqmsg = priv;
2143         kib_msg_t *ackmsg;
2144         kib_dev_t *ibdev;
2145         kib_peer_t *peer;
2146         kib_peer_t *peer2;
2147         kib_conn_t *conn;
2148         lnet_ni_t *ni  = NULL;
2149         kib_net_t *net = NULL;
2150         lnet_nid_t nid;
2151         struct rdma_conn_param cp;
2152         kib_rej_t rej;
2153         int version = IBLND_MSG_VERSION;
2154         unsigned long flags;
2155         int rc;
2156         struct sockaddr_in *peer_addr;
2157
2158         LASSERT(!in_interrupt());
2159
2160         /* cmid inherits 'context' from the corresponding listener id */
2161         ibdev = (kib_dev_t *)cmid->context;
2162         LASSERT(ibdev != NULL);
2163
2164         memset(&rej, 0, sizeof(rej));
2165         rej.ibr_magic = IBLND_MSG_MAGIC;
2166         rej.ibr_why = IBLND_REJECT_FATAL;
2167         rej.ibr_cp.ibcp_max_msg_size = IBLND_MSG_SIZE;
2168
2169         peer_addr = (struct sockaddr_in *)&(cmid->route.addr.dst_addr);
2170         if (*kiblnd_tunables.kib_require_priv_port &&
2171             ntohs(peer_addr->sin_port) >= PROT_SOCK) {
2172                 __u32 ip = ntohl(peer_addr->sin_addr.s_addr);
2173
2174                 CERROR("Peer's port (%pI4h:%hu) is not privileged\n",
2175                        &ip, ntohs(peer_addr->sin_port));
2176                 goto failed;
2177         }
2178
2179         if (priv_nob < offsetof(kib_msg_t, ibm_type)) {
2180                 CERROR("Short connection request\n");
2181                 goto failed;
2182         }
2183
2184         /* Future protocol version compatibility support!  If the
2185          * o2iblnd-specific protocol changes, or when LNET unifies
2186          * protocols over all LNDs, the initial connection will
2187          * negotiate a protocol version.  I trap this here to avoid
2188          * console errors; the reject tells the peer which protocol I
2189          * speak. */
2190         if (reqmsg->ibm_magic == LNET_PROTO_MAGIC ||
2191             reqmsg->ibm_magic == __swab32(LNET_PROTO_MAGIC))
2192                 goto failed;
2193         if (reqmsg->ibm_magic == IBLND_MSG_MAGIC &&
2194             reqmsg->ibm_version != IBLND_MSG_VERSION &&
2195             reqmsg->ibm_version != IBLND_MSG_VERSION_1)
2196                 goto failed;
2197         if (reqmsg->ibm_magic == __swab32(IBLND_MSG_MAGIC) &&
2198             reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION) &&
2199             reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION_1))
2200                 goto failed;
2201
2202         rc = kiblnd_unpack_msg(reqmsg, priv_nob);
2203         if (rc != 0) {
2204                 CERROR("Can't parse connection request: %d\n", rc);
2205                 goto failed;
2206         }
2207
2208         nid = reqmsg->ibm_srcnid;
2209         ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid));
2210
2211         if (ni != NULL) {
2212                 net = (kib_net_t *)ni->ni_data;
2213                 rej.ibr_incarnation = net->ibn_incarnation;
2214         }
2215
2216         if (ni == NULL ||                        /* no matching net */
2217             ni->ni_nid != reqmsg->ibm_dstnid ||   /* right NET, wrong NID! */
2218             net->ibn_dev != ibdev) {          /* wrong device */
2219                 CERROR("Can't accept %s on %s (%s:%d:%pI4h): bad dst nid %s\n",
2220                        libcfs_nid2str(nid),
2221                        ni == NULL ? "NA" : libcfs_nid2str(ni->ni_nid),
2222                        ibdev->ibd_ifname, ibdev->ibd_nnets,
2223                        &ibdev->ibd_ifip,
2224                        libcfs_nid2str(reqmsg->ibm_dstnid));
2225
2226                 goto failed;
2227         }
2228
2229        /* check time stamp as soon as possible */
2230         if (reqmsg->ibm_dststamp != 0 &&
2231             reqmsg->ibm_dststamp != net->ibn_incarnation) {
2232                 CWARN("Stale connection request\n");
2233                 rej.ibr_why = IBLND_REJECT_CONN_STALE;
2234                 goto failed;
2235         }
2236
2237         /* I can accept peer's version */
2238         version = reqmsg->ibm_version;
2239
2240         if (reqmsg->ibm_type != IBLND_MSG_CONNREQ) {
2241                 CERROR("Unexpected connreq msg type: %x from %s\n",
2242                        reqmsg->ibm_type, libcfs_nid2str(nid));
2243                 goto failed;
2244         }
2245
2246         if (reqmsg->ibm_u.connparams.ibcp_queue_depth !=
2247             IBLND_MSG_QUEUE_SIZE(version)) {
2248                 CERROR("Can't accept %s: incompatible queue depth %d (%d wanted)\n",
2249                        libcfs_nid2str(nid), reqmsg->ibm_u.connparams.ibcp_queue_depth,
2250                        IBLND_MSG_QUEUE_SIZE(version));
2251
2252                 if (version == IBLND_MSG_VERSION)
2253                         rej.ibr_why = IBLND_REJECT_MSG_QUEUE_SIZE;
2254
2255                 goto failed;
2256         }
2257
2258         if (reqmsg->ibm_u.connparams.ibcp_max_frags !=
2259             IBLND_RDMA_FRAGS(version)) {
2260                 CERROR("Can't accept %s(version %x): incompatible max_frags %d (%d wanted)\n",
2261                        libcfs_nid2str(nid), version,
2262                        reqmsg->ibm_u.connparams.ibcp_max_frags,
2263                        IBLND_RDMA_FRAGS(version));
2264
2265                 if (version == IBLND_MSG_VERSION)
2266                         rej.ibr_why = IBLND_REJECT_RDMA_FRAGS;
2267
2268                 goto failed;
2269
2270         }
2271
2272         if (reqmsg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2273                 CERROR("Can't accept %s: message size %d too big (%d max)\n",
2274                        libcfs_nid2str(nid),
2275                        reqmsg->ibm_u.connparams.ibcp_max_msg_size,
2276                        IBLND_MSG_SIZE);
2277                 goto failed;
2278         }
2279
2280         /* assume 'nid' is a new peer; create  */
2281         rc = kiblnd_create_peer(ni, &peer, nid);
2282         if (rc != 0) {
2283                 CERROR("Can't create peer for %s\n", libcfs_nid2str(nid));
2284                 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2285                 goto failed;
2286         }
2287
2288         write_lock_irqsave(g_lock, flags);
2289
2290         peer2 = kiblnd_find_peer_locked(nid);
2291         if (peer2 != NULL) {
2292                 if (peer2->ibp_version == 0) {
2293                         peer2->ibp_version     = version;
2294                         peer2->ibp_incarnation = reqmsg->ibm_srcstamp;
2295                 }
2296
2297                 /* not the guy I've talked with */
2298                 if (peer2->ibp_incarnation != reqmsg->ibm_srcstamp ||
2299                     peer2->ibp_version     != version) {
2300                         kiblnd_close_peer_conns_locked(peer2, -ESTALE);
2301                         write_unlock_irqrestore(g_lock, flags);
2302
2303                         CWARN("Conn stale %s [old ver: %x, new ver: %x]\n",
2304                               libcfs_nid2str(nid), peer2->ibp_version, version);
2305
2306                         kiblnd_peer_decref(peer);
2307                         rej.ibr_why = IBLND_REJECT_CONN_STALE;
2308                         goto failed;
2309                 }
2310
2311                 /* tie-break connection race in favour of the higher NID */
2312                 if (peer2->ibp_connecting != 0 &&
2313                     nid < ni->ni_nid) {
2314                         write_unlock_irqrestore(g_lock, flags);
2315
2316                         CWARN("Conn race %s\n", libcfs_nid2str(peer2->ibp_nid));
2317
2318                         kiblnd_peer_decref(peer);
2319                         rej.ibr_why = IBLND_REJECT_CONN_RACE;
2320                         goto failed;
2321                 }
2322
2323                 peer2->ibp_accepting++;
2324                 kiblnd_peer_addref(peer2);
2325
2326                 write_unlock_irqrestore(g_lock, flags);
2327                 kiblnd_peer_decref(peer);
2328                 peer = peer2;
2329         } else {
2330                 /* Brand new peer */
2331                 LASSERT(peer->ibp_accepting == 0);
2332                 LASSERT(peer->ibp_version == 0 &&
2333                          peer->ibp_incarnation == 0);
2334
2335                 peer->ibp_accepting   = 1;
2336                 peer->ibp_version     = version;
2337                 peer->ibp_incarnation = reqmsg->ibm_srcstamp;
2338
2339                 /* I have a ref on ni that prevents it being shutdown */
2340                 LASSERT(net->ibn_shutdown == 0);
2341
2342                 kiblnd_peer_addref(peer);
2343                 list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
2344
2345                 write_unlock_irqrestore(g_lock, flags);
2346         }
2347
2348         conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_PASSIVE_WAIT, version);
2349         if (conn == NULL) {
2350                 kiblnd_peer_connect_failed(peer, 0, -ENOMEM);
2351                 kiblnd_peer_decref(peer);
2352                 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2353                 goto failed;
2354         }
2355
2356         /* conn now "owns" cmid, so I return success from here on to ensure the
2357          * CM callback doesn't destroy cmid. */
2358
2359         conn->ibc_incarnation      = reqmsg->ibm_srcstamp;
2360         conn->ibc_credits          = IBLND_MSG_QUEUE_SIZE(version);
2361         conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(version);
2362         LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(version)
2363                  <= IBLND_RX_MSGS(version));
2364
2365         ackmsg = &conn->ibc_connvars->cv_msg;
2366         memset(ackmsg, 0, sizeof(*ackmsg));
2367
2368         kiblnd_init_msg(ackmsg, IBLND_MSG_CONNACK,
2369                         sizeof(ackmsg->ibm_u.connparams));
2370         ackmsg->ibm_u.connparams.ibcp_queue_depth  = IBLND_MSG_QUEUE_SIZE(version);
2371         ackmsg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2372         ackmsg->ibm_u.connparams.ibcp_max_frags    = IBLND_RDMA_FRAGS(version);
2373
2374         kiblnd_pack_msg(ni, ackmsg, version, 0, nid, reqmsg->ibm_srcstamp);
2375
2376         memset(&cp, 0, sizeof(cp));
2377         cp.private_data = ackmsg;
2378         cp.private_data_len = ackmsg->ibm_nob;
2379         cp.responder_resources = 0;          /* No atomic ops or RDMA reads */
2380         cp.initiator_depth = 0;
2381         cp.flow_control = 1;
2382         cp.retry_count = *kiblnd_tunables.kib_retry_count;
2383         cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
2384
2385         CDEBUG(D_NET, "Accept %s\n", libcfs_nid2str(nid));
2386
2387         rc = rdma_accept(cmid, &cp);
2388         if (rc != 0) {
2389                 CERROR("Can't accept %s: %d\n", libcfs_nid2str(nid), rc);
2390                 rej.ibr_version = version;
2391                 rej.ibr_why     = IBLND_REJECT_FATAL;
2392
2393                 kiblnd_reject(cmid, &rej);
2394                 kiblnd_connreq_done(conn, rc);
2395                 kiblnd_conn_decref(conn);
2396         }
2397
2398         lnet_ni_decref(ni);
2399         return 0;
2400
2401  failed:
2402         if (ni != NULL)
2403                 lnet_ni_decref(ni);
2404
2405         rej.ibr_version             = version;
2406         rej.ibr_cp.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
2407         rej.ibr_cp.ibcp_max_frags   = IBLND_RDMA_FRAGS(version);
2408         kiblnd_reject(cmid, &rej);
2409
2410         return -ECONNREFUSED;
2411 }
2412
2413 static void
2414 kiblnd_reconnect(kib_conn_t *conn, int version,
2415                   __u64 incarnation, int why, kib_connparams_t *cp)
2416 {
2417         kib_peer_t *peer = conn->ibc_peer;
2418         char *reason;
2419         int retry = 0;
2420         unsigned long flags;
2421
2422         LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2423         LASSERT(peer->ibp_connecting > 0);     /* 'conn' at least */
2424
2425         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2426
2427         /* retry connection if it's still needed and no other connection
2428          * attempts (active or passive) are in progress
2429          * NB: reconnect is still needed even when ibp_tx_queue is
2430          * empty if ibp_version != version because reconnect may be
2431          * initiated by kiblnd_query() */
2432         if ((!list_empty(&peer->ibp_tx_queue) ||
2433              peer->ibp_version != version) &&
2434             peer->ibp_connecting == 1 &&
2435             peer->ibp_accepting == 0) {
2436                 retry = 1;
2437                 peer->ibp_connecting++;
2438
2439                 peer->ibp_version     = version;
2440                 peer->ibp_incarnation = incarnation;
2441         }
2442
2443         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2444
2445         if (!retry)
2446                 return;
2447
2448         switch (why) {
2449         default:
2450                 reason = "Unknown";
2451                 break;
2452
2453         case IBLND_REJECT_CONN_STALE:
2454                 reason = "stale";
2455                 break;
2456
2457         case IBLND_REJECT_CONN_RACE:
2458                 reason = "conn race";
2459                 break;
2460
2461         case IBLND_REJECT_CONN_UNCOMPAT:
2462                 reason = "version negotiation";
2463                 break;
2464         }
2465
2466         CNETERR("%s: retrying (%s), %x, %x, queue_dep: %d, max_frag: %d, msg_size: %d\n",
2467                 libcfs_nid2str(peer->ibp_nid),
2468                 reason, IBLND_MSG_VERSION, version,
2469                 cp != NULL ? cp->ibcp_queue_depth  : IBLND_MSG_QUEUE_SIZE(version),
2470                 cp != NULL ? cp->ibcp_max_frags    : IBLND_RDMA_FRAGS(version),
2471                 cp != NULL ? cp->ibcp_max_msg_size : IBLND_MSG_SIZE);
2472
2473         kiblnd_connect_peer(peer);
2474 }
2475
2476 static void
2477 kiblnd_rejected(kib_conn_t *conn, int reason, void *priv, int priv_nob)
2478 {
2479         kib_peer_t *peer = conn->ibc_peer;
2480
2481         LASSERT(!in_interrupt());
2482         LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2483
2484         switch (reason) {
2485         case IB_CM_REJ_STALE_CONN:
2486                 kiblnd_reconnect(conn, IBLND_MSG_VERSION, 0,
2487                                  IBLND_REJECT_CONN_STALE, NULL);
2488                 break;
2489
2490         case IB_CM_REJ_INVALID_SERVICE_ID:
2491                 CNETERR("%s rejected: no listener at %d\n",
2492                         libcfs_nid2str(peer->ibp_nid),
2493                         *kiblnd_tunables.kib_service);
2494                 break;
2495
2496         case IB_CM_REJ_CONSUMER_DEFINED:
2497                 if (priv_nob >= offsetof(kib_rej_t, ibr_padding)) {
2498                         kib_rej_t *rej = priv;
2499                         kib_connparams_t *cp = NULL;
2500                         int flip = 0;
2501                         __u64 incarnation = -1;
2502
2503                         /* NB. default incarnation is -1 because:
2504                          * a) V1 will ignore dst incarnation in connreq.
2505                          * b) V2 will provide incarnation while rejecting me,
2506                          *    -1 will be overwrote.
2507                          *
2508                          * if I try to connect to a V1 peer with V2 protocol,
2509                          * it rejected me then upgrade to V2, I have no idea
2510                          * about the upgrading and try to reconnect with V1,
2511                          * in this case upgraded V2 can find out I'm trying to
2512                          * talk to the old guy and reject me(incarnation is -1).
2513                          */
2514
2515                         if (rej->ibr_magic == __swab32(IBLND_MSG_MAGIC) ||
2516                             rej->ibr_magic == __swab32(LNET_PROTO_MAGIC)) {
2517                                 __swab32s(&rej->ibr_magic);
2518                                 __swab16s(&rej->ibr_version);
2519                                 flip = 1;
2520                         }
2521
2522                         if (priv_nob >= sizeof(kib_rej_t) &&
2523                             rej->ibr_version > IBLND_MSG_VERSION_1) {
2524                                 /* priv_nob is always 148 in current version
2525                                  * of OFED, so we still need to check version.
2526                                  * (define of IB_CM_REJ_PRIVATE_DATA_SIZE) */
2527                                 cp = &rej->ibr_cp;
2528
2529                                 if (flip) {
2530                                         __swab64s(&rej->ibr_incarnation);
2531                                         __swab16s(&cp->ibcp_queue_depth);
2532                                         __swab16s(&cp->ibcp_max_frags);
2533                                         __swab32s(&cp->ibcp_max_msg_size);
2534                                 }
2535
2536                                 incarnation = rej->ibr_incarnation;
2537                         }
2538
2539                         if (rej->ibr_magic != IBLND_MSG_MAGIC &&
2540                             rej->ibr_magic != LNET_PROTO_MAGIC) {
2541                                 CERROR("%s rejected: consumer defined fatal error\n",
2542                                        libcfs_nid2str(peer->ibp_nid));
2543                                 break;
2544                         }
2545
2546                         if (rej->ibr_version != IBLND_MSG_VERSION &&
2547                             rej->ibr_version != IBLND_MSG_VERSION_1) {
2548                                 CERROR("%s rejected: o2iblnd version %x error\n",
2549                                        libcfs_nid2str(peer->ibp_nid),
2550                                        rej->ibr_version);
2551                                 break;
2552                         }
2553
2554                         if (rej->ibr_why     == IBLND_REJECT_FATAL &&
2555                             rej->ibr_version == IBLND_MSG_VERSION_1) {
2556                                 CDEBUG(D_NET, "rejected by old version peer %s: %x\n",
2557                                        libcfs_nid2str(peer->ibp_nid), rej->ibr_version);
2558
2559                                 if (conn->ibc_version != IBLND_MSG_VERSION_1)
2560                                         rej->ibr_why = IBLND_REJECT_CONN_UNCOMPAT;
2561                         }
2562
2563                         switch (rej->ibr_why) {
2564                         case IBLND_REJECT_CONN_RACE:
2565                         case IBLND_REJECT_CONN_STALE:
2566                         case IBLND_REJECT_CONN_UNCOMPAT:
2567                                 kiblnd_reconnect(conn, rej->ibr_version,
2568                                                  incarnation, rej->ibr_why, cp);
2569                                 break;
2570
2571                         case IBLND_REJECT_MSG_QUEUE_SIZE:
2572                                 CERROR("%s rejected: incompatible message queue depth %d, %d\n",
2573                                        libcfs_nid2str(peer->ibp_nid),
2574                                        cp != NULL ? cp->ibcp_queue_depth :
2575                                        IBLND_MSG_QUEUE_SIZE(rej->ibr_version),
2576                                        IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
2577                                 break;
2578
2579                         case IBLND_REJECT_RDMA_FRAGS:
2580                                 CERROR("%s rejected: incompatible # of RDMA fragments %d, %d\n",
2581                                        libcfs_nid2str(peer->ibp_nid),
2582                                        cp != NULL ? cp->ibcp_max_frags :
2583                                        IBLND_RDMA_FRAGS(rej->ibr_version),
2584                                        IBLND_RDMA_FRAGS(conn->ibc_version));
2585                                 break;
2586
2587                         case IBLND_REJECT_NO_RESOURCES:
2588                                 CERROR("%s rejected: o2iblnd no resources\n",
2589                                        libcfs_nid2str(peer->ibp_nid));
2590                                 break;
2591
2592                         case IBLND_REJECT_FATAL:
2593                                 CERROR("%s rejected: o2iblnd fatal error\n",
2594                                        libcfs_nid2str(peer->ibp_nid));
2595                                 break;
2596
2597                         default:
2598                                 CERROR("%s rejected: o2iblnd reason %d\n",
2599                                        libcfs_nid2str(peer->ibp_nid),
2600                                        rej->ibr_why);
2601                                 break;
2602                         }
2603                         break;
2604                 }
2605                 /* fall through */
2606         default:
2607                 CNETERR("%s rejected: reason %d, size %d\n",
2608                         libcfs_nid2str(peer->ibp_nid), reason, priv_nob);
2609                 break;
2610         }
2611
2612         kiblnd_connreq_done(conn, -ECONNREFUSED);
2613 }
2614
2615 static void
2616 kiblnd_check_connreply(kib_conn_t *conn, void *priv, int priv_nob)
2617 {
2618         kib_peer_t *peer = conn->ibc_peer;
2619         lnet_ni_t *ni = peer->ibp_ni;
2620         kib_net_t *net = ni->ni_data;
2621         kib_msg_t *msg = priv;
2622         int ver = conn->ibc_version;
2623         int rc = kiblnd_unpack_msg(msg, priv_nob);
2624         unsigned long flags;
2625
2626         LASSERT(net != NULL);
2627
2628         if (rc != 0) {
2629                 CERROR("Can't unpack connack from %s: %d\n",
2630                        libcfs_nid2str(peer->ibp_nid), rc);
2631                 goto failed;
2632         }
2633
2634         if (msg->ibm_type != IBLND_MSG_CONNACK) {
2635                 CERROR("Unexpected message %d from %s\n",
2636                        msg->ibm_type, libcfs_nid2str(peer->ibp_nid));
2637                 rc = -EPROTO;
2638                 goto failed;
2639         }
2640
2641         if (ver != msg->ibm_version) {
2642                 CERROR("%s replied version %x is different with requested version %x\n",
2643                        libcfs_nid2str(peer->ibp_nid), msg->ibm_version, ver);
2644                 rc = -EPROTO;
2645                 goto failed;
2646         }
2647
2648         if (msg->ibm_u.connparams.ibcp_queue_depth !=
2649             IBLND_MSG_QUEUE_SIZE(ver)) {
2650                 CERROR("%s has incompatible queue depth %d(%d wanted)\n",
2651                        libcfs_nid2str(peer->ibp_nid),
2652                        msg->ibm_u.connparams.ibcp_queue_depth,
2653                        IBLND_MSG_QUEUE_SIZE(ver));
2654                 rc = -EPROTO;
2655                 goto failed;
2656         }
2657
2658         if (msg->ibm_u.connparams.ibcp_max_frags !=
2659             IBLND_RDMA_FRAGS(ver)) {
2660                 CERROR("%s has incompatible max_frags %d (%d wanted)\n",
2661                        libcfs_nid2str(peer->ibp_nid),
2662                        msg->ibm_u.connparams.ibcp_max_frags,
2663                        IBLND_RDMA_FRAGS(ver));
2664                 rc = -EPROTO;
2665                 goto failed;
2666         }
2667
2668         if (msg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2669                 CERROR("%s max message size %d too big (%d max)\n",
2670                        libcfs_nid2str(peer->ibp_nid),
2671                        msg->ibm_u.connparams.ibcp_max_msg_size,
2672                        IBLND_MSG_SIZE);
2673                 rc = -EPROTO;
2674                 goto failed;
2675         }
2676
2677         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2678         if (msg->ibm_dstnid == ni->ni_nid &&
2679             msg->ibm_dststamp == net->ibn_incarnation)
2680                 rc = 0;
2681         else
2682                 rc = -ESTALE;
2683         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2684
2685         if (rc != 0) {
2686                 CERROR("Bad connection reply from %s, rc = %d, version: %x max_frags: %d\n",
2687                        libcfs_nid2str(peer->ibp_nid), rc,
2688                        msg->ibm_version, msg->ibm_u.connparams.ibcp_max_frags);
2689                 goto failed;
2690         }
2691
2692         conn->ibc_incarnation = msg->ibm_srcstamp;
2693         conn->ibc_credits =
2694         conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(ver);
2695         LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(ver)
2696                  <= IBLND_RX_MSGS(ver));
2697
2698         kiblnd_connreq_done(conn, 0);
2699         return;
2700
2701  failed:
2702         /* NB My QP has already established itself, so I handle anything going
2703          * wrong here by setting ibc_comms_error.
2704          * kiblnd_connreq_done(0) moves the conn state to ESTABLISHED, but then
2705          * immediately tears it down. */
2706
2707         LASSERT(rc != 0);
2708         conn->ibc_comms_error = rc;
2709         kiblnd_connreq_done(conn, 0);
2710 }
2711
2712 static int
2713 kiblnd_active_connect(struct rdma_cm_id *cmid)
2714 {
2715         kib_peer_t *peer = (kib_peer_t *)cmid->context;
2716         kib_conn_t *conn;
2717         kib_msg_t *msg;
2718         struct rdma_conn_param cp;
2719         int version;
2720         __u64 incarnation;
2721         unsigned long flags;
2722         int rc;
2723
2724         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2725
2726         incarnation = peer->ibp_incarnation;
2727         version = (peer->ibp_version == 0) ? IBLND_MSG_VERSION :
2728                                              peer->ibp_version;
2729
2730         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2731
2732         conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_ACTIVE_CONNECT, version);
2733         if (conn == NULL) {
2734                 kiblnd_peer_connect_failed(peer, 1, -ENOMEM);
2735                 kiblnd_peer_decref(peer); /* lose cmid's ref */
2736                 return -ENOMEM;
2737         }
2738
2739         /* conn "owns" cmid now, so I return success from here on to ensure the
2740          * CM callback doesn't destroy cmid. conn also takes over cmid's ref
2741          * on peer */
2742
2743         msg = &conn->ibc_connvars->cv_msg;
2744
2745         memset(msg, 0, sizeof(*msg));
2746         kiblnd_init_msg(msg, IBLND_MSG_CONNREQ, sizeof(msg->ibm_u.connparams));
2747         msg->ibm_u.connparams.ibcp_queue_depth  = IBLND_MSG_QUEUE_SIZE(version);
2748         msg->ibm_u.connparams.ibcp_max_frags    = IBLND_RDMA_FRAGS(version);
2749         msg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2750
2751         kiblnd_pack_msg(peer->ibp_ni, msg, version,
2752                         0, peer->ibp_nid, incarnation);
2753
2754         memset(&cp, 0, sizeof(cp));
2755         cp.private_data = msg;
2756         cp.private_data_len    = msg->ibm_nob;
2757         cp.responder_resources = 0;          /* No atomic ops or RDMA reads */
2758         cp.initiator_depth     = 0;
2759         cp.flow_control        = 1;
2760         cp.retry_count         = *kiblnd_tunables.kib_retry_count;
2761         cp.rnr_retry_count     = *kiblnd_tunables.kib_rnr_retry_count;
2762
2763         LASSERT(cmid->context == (void *)conn);
2764         LASSERT(conn->ibc_cmid == cmid);
2765
2766         rc = rdma_connect(cmid, &cp);
2767         if (rc != 0) {
2768                 CERROR("Can't connect to %s: %d\n",
2769                        libcfs_nid2str(peer->ibp_nid), rc);
2770                 kiblnd_connreq_done(conn, rc);
2771                 kiblnd_conn_decref(conn);
2772         }
2773
2774         return 0;
2775 }
2776
2777 int
2778 kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event)
2779 {
2780         kib_peer_t *peer;
2781         kib_conn_t *conn;
2782         int rc;
2783
2784         switch (event->event) {
2785         default:
2786                 CERROR("Unexpected event: %d, status: %d\n",
2787                        event->event, event->status);
2788                 LBUG();
2789
2790         case RDMA_CM_EVENT_CONNECT_REQUEST:
2791                 /* destroy cmid on failure */
2792                 rc = kiblnd_passive_connect(cmid,
2793                                             (void *)KIBLND_CONN_PARAM(event),
2794                                             KIBLND_CONN_PARAM_LEN(event));
2795                 CDEBUG(D_NET, "connreq: %d\n", rc);
2796                 return rc;
2797
2798         case RDMA_CM_EVENT_ADDR_ERROR:
2799                 peer = (kib_peer_t *)cmid->context;
2800                 CNETERR("%s: ADDR ERROR %d\n",
2801                        libcfs_nid2str(peer->ibp_nid), event->status);
2802                 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
2803                 kiblnd_peer_decref(peer);
2804                 return -EHOSTUNREACH;      /* rc != 0 destroys cmid */
2805
2806         case RDMA_CM_EVENT_ADDR_RESOLVED:
2807                 peer = (kib_peer_t *)cmid->context;
2808
2809                 CDEBUG(D_NET, "%s Addr resolved: %d\n",
2810                        libcfs_nid2str(peer->ibp_nid), event->status);
2811
2812                 if (event->status != 0) {
2813                         CNETERR("Can't resolve address for %s: %d\n",
2814                                 libcfs_nid2str(peer->ibp_nid), event->status);
2815                         rc = event->status;
2816                 } else {
2817                         rc = rdma_resolve_route(
2818                                 cmid, *kiblnd_tunables.kib_timeout * 1000);
2819                         if (rc == 0)
2820                                 return 0;
2821                         /* Can't initiate route resolution */
2822                         CERROR("Can't resolve route for %s: %d\n",
2823                                libcfs_nid2str(peer->ibp_nid), rc);
2824                 }
2825                 kiblnd_peer_connect_failed(peer, 1, rc);
2826                 kiblnd_peer_decref(peer);
2827                 return rc;                    /* rc != 0 destroys cmid */
2828
2829         case RDMA_CM_EVENT_ROUTE_ERROR:
2830                 peer = (kib_peer_t *)cmid->context;
2831                 CNETERR("%s: ROUTE ERROR %d\n",
2832                         libcfs_nid2str(peer->ibp_nid), event->status);
2833                 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
2834                 kiblnd_peer_decref(peer);
2835                 return -EHOSTUNREACH;      /* rc != 0 destroys cmid */
2836
2837         case RDMA_CM_EVENT_ROUTE_RESOLVED:
2838                 peer = (kib_peer_t *)cmid->context;
2839                 CDEBUG(D_NET, "%s Route resolved: %d\n",
2840                        libcfs_nid2str(peer->ibp_nid), event->status);
2841
2842                 if (event->status == 0)
2843                         return kiblnd_active_connect(cmid);
2844
2845                 CNETERR("Can't resolve route for %s: %d\n",
2846                        libcfs_nid2str(peer->ibp_nid), event->status);
2847                 kiblnd_peer_connect_failed(peer, 1, event->status);
2848                 kiblnd_peer_decref(peer);
2849                 return event->status;      /* rc != 0 destroys cmid */
2850
2851         case RDMA_CM_EVENT_UNREACHABLE:
2852                 conn = (kib_conn_t *)cmid->context;
2853                 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
2854                         conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
2855                 CNETERR("%s: UNREACHABLE %d\n",
2856                        libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
2857                 kiblnd_connreq_done(conn, -ENETDOWN);
2858                 kiblnd_conn_decref(conn);
2859                 return 0;
2860
2861         case RDMA_CM_EVENT_CONNECT_ERROR:
2862                 conn = (kib_conn_t *)cmid->context;
2863                 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
2864                         conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
2865                 CNETERR("%s: CONNECT ERROR %d\n",
2866                         libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
2867                 kiblnd_connreq_done(conn, -ENOTCONN);
2868                 kiblnd_conn_decref(conn);
2869                 return 0;
2870
2871         case RDMA_CM_EVENT_REJECTED:
2872                 conn = (kib_conn_t *)cmid->context;
2873                 switch (conn->ibc_state) {
2874                 default:
2875                         LBUG();
2876
2877                 case IBLND_CONN_PASSIVE_WAIT:
2878                         CERROR("%s: REJECTED %d\n",
2879                                 libcfs_nid2str(conn->ibc_peer->ibp_nid),
2880                                 event->status);
2881                         kiblnd_connreq_done(conn, -ECONNRESET);
2882                         break;
2883
2884                 case IBLND_CONN_ACTIVE_CONNECT:
2885                         kiblnd_rejected(conn, event->status,
2886                                         (void *)KIBLND_CONN_PARAM(event),
2887                                         KIBLND_CONN_PARAM_LEN(event));
2888                         break;
2889                 }
2890                 kiblnd_conn_decref(conn);
2891                 return 0;
2892
2893         case RDMA_CM_EVENT_ESTABLISHED:
2894                 conn = (kib_conn_t *)cmid->context;
2895                 switch (conn->ibc_state) {
2896                 default:
2897                         LBUG();
2898
2899                 case IBLND_CONN_PASSIVE_WAIT:
2900                         CDEBUG(D_NET, "ESTABLISHED (passive): %s\n",
2901                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
2902                         kiblnd_connreq_done(conn, 0);
2903                         break;
2904
2905                 case IBLND_CONN_ACTIVE_CONNECT:
2906                         CDEBUG(D_NET, "ESTABLISHED(active): %s\n",
2907                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
2908                         kiblnd_check_connreply(conn,
2909                                                (void *)KIBLND_CONN_PARAM(event),
2910                                                KIBLND_CONN_PARAM_LEN(event));
2911                         break;
2912                 }
2913                 /* net keeps its ref on conn! */
2914                 return 0;
2915
2916         case RDMA_CM_EVENT_TIMEWAIT_EXIT:
2917                 CDEBUG(D_NET, "Ignore TIMEWAIT_EXIT event\n");
2918                 return 0;
2919         case RDMA_CM_EVENT_DISCONNECTED:
2920                 conn = (kib_conn_t *)cmid->context;
2921                 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
2922                         CERROR("%s DISCONNECTED\n",
2923                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
2924                         kiblnd_connreq_done(conn, -ECONNRESET);
2925                 } else {
2926                         kiblnd_close_conn(conn, 0);
2927                 }
2928                 kiblnd_conn_decref(conn);
2929                 cmid->context = NULL;
2930                 return 0;
2931
2932         case RDMA_CM_EVENT_DEVICE_REMOVAL:
2933                 LCONSOLE_ERROR_MSG(0x131,
2934                                    "Received notification of device removal\n"
2935                                    "Please shutdown LNET to allow this to proceed\n");
2936                 /* Can't remove network from underneath LNET for now, so I have
2937                  * to ignore this */
2938                 return 0;
2939
2940         case RDMA_CM_EVENT_ADDR_CHANGE:
2941                 LCONSOLE_INFO("Physical link changed (eg hca/port)\n");
2942                 return 0;
2943         }
2944 }
2945
2946 static int
2947 kiblnd_check_txs_locked(kib_conn_t *conn, struct list_head *txs)
2948 {
2949         kib_tx_t *tx;
2950         struct list_head *ttmp;
2951
2952         list_for_each(ttmp, txs) {
2953                 tx = list_entry(ttmp, kib_tx_t, tx_list);
2954
2955                 if (txs != &conn->ibc_active_txs) {
2956                         LASSERT(tx->tx_queued);
2957                 } else {
2958                         LASSERT(!tx->tx_queued);
2959                         LASSERT(tx->tx_waiting || tx->tx_sending != 0);
2960                 }
2961
2962                 if (cfs_time_aftereq(jiffies, tx->tx_deadline)) {
2963                         CERROR("Timed out tx: %s, %lu seconds\n",
2964                                kiblnd_queue2str(conn, txs),
2965                                cfs_duration_sec(jiffies - tx->tx_deadline));
2966                         return 1;
2967                 }
2968         }
2969
2970         return 0;
2971 }
2972
2973 static int
2974 kiblnd_conn_timed_out_locked(kib_conn_t *conn)
2975 {
2976         return  kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue) ||
2977                 kiblnd_check_txs_locked(conn, &conn->ibc_tx_noops) ||
2978                 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_rsrvd) ||
2979                 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_nocred) ||
2980                 kiblnd_check_txs_locked(conn, &conn->ibc_active_txs);
2981 }
2982
2983 static void
2984 kiblnd_check_conns(int idx)
2985 {
2986         LIST_HEAD(closes);
2987         LIST_HEAD(checksends);
2988         struct list_head *peers = &kiblnd_data.kib_peers[idx];
2989         struct list_head *ptmp;
2990         kib_peer_t *peer;
2991         kib_conn_t *conn;
2992         kib_conn_t *tmp;
2993         struct list_head *ctmp;
2994         unsigned long flags;
2995
2996         /* NB. We expect to have a look at all the peers and not find any
2997          * RDMAs to time out, so we just use a shared lock while we
2998          * take a look... */
2999         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
3000
3001         list_for_each(ptmp, peers) {
3002                 peer = list_entry(ptmp, kib_peer_t, ibp_list);
3003
3004                 list_for_each(ctmp, &peer->ibp_conns) {
3005                         int timedout;
3006                         int sendnoop;
3007
3008                         conn = list_entry(ctmp, kib_conn_t, ibc_list);
3009
3010                         LASSERT(conn->ibc_state == IBLND_CONN_ESTABLISHED);
3011
3012                         spin_lock(&conn->ibc_lock);
3013
3014                         sendnoop = kiblnd_need_noop(conn);
3015                         timedout = kiblnd_conn_timed_out_locked(conn);
3016                         if (!sendnoop && !timedout) {
3017                                 spin_unlock(&conn->ibc_lock);
3018                                 continue;
3019                         }
3020
3021                         if (timedout) {
3022                                 CERROR("Timed out RDMA with %s (%lu): c: %u, oc: %u, rc: %u\n",
3023                                        libcfs_nid2str(peer->ibp_nid),
3024                                        cfs_duration_sec(cfs_time_current() -
3025                                                         peer->ibp_last_alive),
3026                                        conn->ibc_credits,
3027                                        conn->ibc_outstanding_credits,
3028                                        conn->ibc_reserved_credits);
3029                                 list_add(&conn->ibc_connd_list, &closes);
3030                         } else {
3031                                 list_add(&conn->ibc_connd_list,
3032                                              &checksends);
3033                         }
3034                         /* +ref for 'closes' or 'checksends' */
3035                         kiblnd_conn_addref(conn);
3036
3037                         spin_unlock(&conn->ibc_lock);
3038                 }
3039         }
3040
3041         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
3042
3043         /* Handle timeout by closing the whole
3044          * connection. We can only be sure RDMA activity
3045          * has ceased once the QP has been modified. */
3046         list_for_each_entry_safe(conn, tmp, &closes, ibc_connd_list) {
3047                 list_del(&conn->ibc_connd_list);
3048                 kiblnd_close_conn(conn, -ETIMEDOUT);
3049                 kiblnd_conn_decref(conn);
3050         }
3051
3052         /* In case we have enough credits to return via a
3053          * NOOP, but there were no non-blocking tx descs
3054          * free to do it last time... */
3055         while (!list_empty(&checksends)) {
3056                 conn = list_entry(checksends.next,
3057                                       kib_conn_t, ibc_connd_list);
3058                 list_del(&conn->ibc_connd_list);
3059                 kiblnd_check_sends(conn);
3060                 kiblnd_conn_decref(conn);
3061         }
3062 }
3063
3064 static void
3065 kiblnd_disconnect_conn(kib_conn_t *conn)
3066 {
3067         LASSERT(!in_interrupt());
3068         LASSERT(current == kiblnd_data.kib_connd);
3069         LASSERT(conn->ibc_state == IBLND_CONN_CLOSING);
3070
3071         rdma_disconnect(conn->ibc_cmid);
3072         kiblnd_finalise_conn(conn);
3073
3074         kiblnd_peer_notify(conn->ibc_peer);
3075 }
3076
3077 int
3078 kiblnd_connd(void *arg)
3079 {
3080         wait_queue_t wait;
3081         unsigned long flags;
3082         kib_conn_t *conn;
3083         int timeout;
3084         int i;
3085         int dropped_lock;
3086         int peer_index = 0;
3087         unsigned long deadline = jiffies;
3088
3089         cfs_block_allsigs();
3090
3091         init_waitqueue_entry(&wait, current);
3092         kiblnd_data.kib_connd = current;
3093
3094         spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3095
3096         while (!kiblnd_data.kib_shutdown) {
3097
3098                 dropped_lock = 0;
3099
3100                 if (!list_empty(&kiblnd_data.kib_connd_zombies)) {
3101                         conn = list_entry(kiblnd_data.kib_connd_zombies.next,
3102                                               kib_conn_t, ibc_list);
3103                         list_del(&conn->ibc_list);
3104
3105                         spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
3106                                                flags);
3107                         dropped_lock = 1;
3108
3109                         kiblnd_destroy_conn(conn);
3110
3111                         spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3112                 }
3113
3114                 if (!list_empty(&kiblnd_data.kib_connd_conns)) {
3115                         conn = list_entry(kiblnd_data.kib_connd_conns.next,
3116                                               kib_conn_t, ibc_list);
3117                         list_del(&conn->ibc_list);
3118
3119                         spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
3120                                                flags);
3121                         dropped_lock = 1;
3122
3123                         kiblnd_disconnect_conn(conn);
3124                         kiblnd_conn_decref(conn);
3125
3126                         spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3127                 }
3128
3129                 /* careful with the jiffy wrap... */
3130                 timeout = (int)(deadline - jiffies);
3131                 if (timeout <= 0) {
3132                         const int n = 4;
3133                         const int p = 1;
3134                         int chunk = kiblnd_data.kib_peer_hash_size;
3135
3136                         spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3137                         dropped_lock = 1;
3138
3139                         /* Time to check for RDMA timeouts on a few more
3140                          * peers: I do checks every 'p' seconds on a
3141                          * proportion of the peer table and I need to check
3142                          * every connection 'n' times within a timeout
3143                          * interval, to ensure I detect a timeout on any
3144                          * connection within (n+1)/n times the timeout
3145                          * interval. */
3146
3147                         if (*kiblnd_tunables.kib_timeout > n * p)
3148                                 chunk = (chunk * n * p) /
3149                                         *kiblnd_tunables.kib_timeout;
3150                         if (chunk == 0)
3151                                 chunk = 1;
3152
3153                         for (i = 0; i < chunk; i++) {
3154                                 kiblnd_check_conns(peer_index);
3155                                 peer_index = (peer_index + 1) %
3156                                              kiblnd_data.kib_peer_hash_size;
3157                         }
3158
3159                         deadline += p * HZ;
3160                         spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3161                 }
3162
3163                 if (dropped_lock)
3164                         continue;
3165
3166                 /* Nothing to do for 'timeout'  */
3167                 set_current_state(TASK_INTERRUPTIBLE);
3168                 add_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3169                 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3170
3171                 schedule_timeout(timeout);
3172
3173                 remove_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3174                 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3175         }
3176
3177         spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3178
3179         kiblnd_thread_fini();
3180         return 0;
3181 }
3182
3183 void
3184 kiblnd_qp_event(struct ib_event *event, void *arg)
3185 {
3186         kib_conn_t *conn = arg;
3187
3188         switch (event->event) {
3189         case IB_EVENT_COMM_EST:
3190                 CDEBUG(D_NET, "%s established\n",
3191                        libcfs_nid2str(conn->ibc_peer->ibp_nid));
3192                 return;
3193
3194         default:
3195                 CERROR("%s: Async QP event type %d\n",
3196                        libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3197                 return;
3198         }
3199 }
3200
3201 static void
3202 kiblnd_complete(struct ib_wc *wc)
3203 {
3204         switch (kiblnd_wreqid2type(wc->wr_id)) {
3205         default:
3206                 LBUG();
3207
3208         case IBLND_WID_RDMA:
3209                 /* We only get RDMA completion notification if it fails.  All
3210                  * subsequent work items, including the final SEND will fail
3211                  * too.  However we can't print out any more info about the
3212                  * failing RDMA because 'tx' might be back on the idle list or
3213                  * even reused already if we didn't manage to post all our work
3214                  * items */
3215                 CNETERR("RDMA (tx: %p) failed: %d\n",
3216                         kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3217                 return;
3218
3219         case IBLND_WID_TX:
3220                 kiblnd_tx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3221                 return;
3222
3223         case IBLND_WID_RX:
3224                 kiblnd_rx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status,
3225                                    wc->byte_len);
3226                 return;
3227         }
3228 }
3229
3230 void
3231 kiblnd_cq_completion(struct ib_cq *cq, void *arg)
3232 {
3233         /* NB I'm not allowed to schedule this conn once its refcount has
3234          * reached 0.  Since fundamentally I'm racing with scheduler threads
3235          * consuming my CQ I could be called after all completions have
3236          * occurred.  But in this case, ibc_nrx == 0 && ibc_nsends_posted == 0
3237          * and this CQ is about to be destroyed so I NOOP. */
3238         kib_conn_t *conn = arg;
3239         struct kib_sched_info *sched = conn->ibc_sched;
3240         unsigned long flags;
3241
3242         LASSERT(cq == conn->ibc_cq);
3243
3244         spin_lock_irqsave(&sched->ibs_lock, flags);
3245
3246         conn->ibc_ready = 1;
3247
3248         if (!conn->ibc_scheduled &&
3249             (conn->ibc_nrx > 0 ||
3250              conn->ibc_nsends_posted > 0)) {
3251                 kiblnd_conn_addref(conn); /* +1 ref for sched_conns */
3252                 conn->ibc_scheduled = 1;
3253                 list_add_tail(&conn->ibc_sched_list, &sched->ibs_conns);
3254
3255                 if (waitqueue_active(&sched->ibs_waitq))
3256                         wake_up(&sched->ibs_waitq);
3257         }
3258
3259         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3260 }
3261
3262 void
3263 kiblnd_cq_event(struct ib_event *event, void *arg)
3264 {
3265         kib_conn_t *conn = arg;
3266
3267         CERROR("%s: async CQ event type %d\n",
3268                libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3269 }
3270
3271 int
3272 kiblnd_scheduler(void *arg)
3273 {
3274         long id = (long)arg;
3275         struct kib_sched_info *sched;
3276         kib_conn_t *conn;
3277         wait_queue_t wait;
3278         unsigned long flags;
3279         struct ib_wc wc;
3280         int did_something;
3281         int busy_loops = 0;
3282         int rc;
3283
3284         cfs_block_allsigs();
3285
3286         init_waitqueue_entry(&wait, current);
3287
3288         sched = kiblnd_data.kib_scheds[KIB_THREAD_CPT(id)];
3289
3290         rc = cfs_cpt_bind(lnet_cpt_table(), sched->ibs_cpt);
3291         if (rc != 0) {
3292                 CWARN("Failed to bind on CPT %d, please verify whether all CPUs are healthy and reload modules if necessary, otherwise your system might under risk of low performance\n",
3293                       sched->ibs_cpt);
3294         }
3295
3296         spin_lock_irqsave(&sched->ibs_lock, flags);
3297
3298         while (!kiblnd_data.kib_shutdown) {
3299                 if (busy_loops++ >= IBLND_RESCHED) {
3300                         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3301
3302                         cond_resched();
3303                         busy_loops = 0;
3304
3305                         spin_lock_irqsave(&sched->ibs_lock, flags);
3306                 }
3307
3308                 did_something = 0;
3309
3310                 if (!list_empty(&sched->ibs_conns)) {
3311                         conn = list_entry(sched->ibs_conns.next,
3312                                               kib_conn_t, ibc_sched_list);
3313                         /* take over kib_sched_conns' ref on conn... */
3314                         LASSERT(conn->ibc_scheduled);
3315                         list_del(&conn->ibc_sched_list);
3316                         conn->ibc_ready = 0;
3317
3318                         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3319
3320                         rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3321                         if (rc == 0) {
3322                                 rc = ib_req_notify_cq(conn->ibc_cq,
3323                                                       IB_CQ_NEXT_COMP);
3324                                 if (rc < 0) {
3325                                         CWARN("%s: ib_req_notify_cq failed: %d, closing connection\n",
3326                                               libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
3327                                         kiblnd_close_conn(conn, -EIO);
3328                                         kiblnd_conn_decref(conn);
3329                                         spin_lock_irqsave(&sched->ibs_lock,
3330                                                               flags);
3331                                         continue;
3332                                 }
3333
3334                                 rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3335                         }
3336
3337                         if (rc < 0) {
3338                                 CWARN("%s: ib_poll_cq failed: %d, closing connection\n",
3339                                       libcfs_nid2str(conn->ibc_peer->ibp_nid),
3340                                       rc);
3341                                 kiblnd_close_conn(conn, -EIO);
3342                                 kiblnd_conn_decref(conn);
3343                                 spin_lock_irqsave(&sched->ibs_lock, flags);
3344                                 continue;
3345                         }
3346
3347                         spin_lock_irqsave(&sched->ibs_lock, flags);
3348
3349                         if (rc != 0 || conn->ibc_ready) {
3350                                 /* There may be another completion waiting; get
3351                                  * another scheduler to check while I handle
3352                                  * this one... */
3353                                 /* +1 ref for sched_conns */
3354                                 kiblnd_conn_addref(conn);
3355                                 list_add_tail(&conn->ibc_sched_list,
3356                                                   &sched->ibs_conns);
3357                                 if (waitqueue_active(&sched->ibs_waitq))
3358                                         wake_up(&sched->ibs_waitq);
3359                         } else {
3360                                 conn->ibc_scheduled = 0;
3361                         }
3362
3363                         if (rc != 0) {
3364                                 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3365                                 kiblnd_complete(&wc);
3366
3367                                 spin_lock_irqsave(&sched->ibs_lock, flags);
3368                         }
3369
3370                         kiblnd_conn_decref(conn); /* ...drop my ref from above */
3371                         did_something = 1;
3372                 }
3373
3374                 if (did_something)
3375                         continue;
3376
3377                 set_current_state(TASK_INTERRUPTIBLE);
3378                 add_wait_queue_exclusive(&sched->ibs_waitq, &wait);
3379                 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3380
3381                 schedule();
3382                 busy_loops = 0;
3383
3384                 remove_wait_queue(&sched->ibs_waitq, &wait);
3385                 spin_lock_irqsave(&sched->ibs_lock, flags);
3386         }
3387
3388         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3389
3390         kiblnd_thread_fini();
3391         return 0;
3392 }
3393
3394 int
3395 kiblnd_failover_thread(void *arg)
3396 {
3397         rwlock_t *glock = &kiblnd_data.kib_global_lock;
3398         kib_dev_t *dev;
3399         wait_queue_t wait;
3400         unsigned long flags;
3401         int rc;
3402
3403         LASSERT(*kiblnd_tunables.kib_dev_failover != 0);
3404
3405         cfs_block_allsigs();
3406
3407         init_waitqueue_entry(&wait, current);
3408         write_lock_irqsave(glock, flags);
3409
3410         while (!kiblnd_data.kib_shutdown) {
3411                 int do_failover = 0;
3412                 int long_sleep;
3413
3414                 list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
3415                                     ibd_fail_list) {
3416                         if (time_before(cfs_time_current(),
3417                                         dev->ibd_next_failover))
3418                                 continue;
3419                         do_failover = 1;
3420                         break;
3421                 }
3422
3423                 if (do_failover) {
3424                         list_del_init(&dev->ibd_fail_list);
3425                         dev->ibd_failover = 1;
3426                         write_unlock_irqrestore(glock, flags);
3427
3428                         rc = kiblnd_dev_failover(dev);
3429
3430                         write_lock_irqsave(glock, flags);
3431
3432                         LASSERT(dev->ibd_failover);
3433                         dev->ibd_failover = 0;
3434                         if (rc >= 0) { /* Device is OK or failover succeed */
3435                                 dev->ibd_next_failover = cfs_time_shift(3);
3436                                 continue;
3437                         }
3438
3439                         /* failed to failover, retry later */
3440                         dev->ibd_next_failover =
3441                                 cfs_time_shift(min(dev->ibd_failed_failover, 10));
3442                         if (kiblnd_dev_can_failover(dev)) {
3443                                 list_add_tail(&dev->ibd_fail_list,
3444                                               &kiblnd_data.kib_failed_devs);
3445                         }
3446
3447                         continue;
3448                 }
3449
3450                 /* long sleep if no more pending failover */
3451                 long_sleep = list_empty(&kiblnd_data.kib_failed_devs);
3452
3453                 set_current_state(TASK_INTERRUPTIBLE);
3454                 add_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3455                 write_unlock_irqrestore(glock, flags);
3456
3457                 rc = schedule_timeout(long_sleep ? cfs_time_seconds(10) :
3458                                                    cfs_time_seconds(1));
3459                 remove_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3460                 write_lock_irqsave(glock, flags);
3461
3462                 if (!long_sleep || rc != 0)
3463                         continue;
3464
3465                 /* have a long sleep, routine check all active devices,
3466                  * we need checking like this because if there is not active
3467                  * connection on the dev and no SEND from local, we may listen
3468                  * on wrong HCA for ever while there is a bonding failover */
3469                 list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
3470                         if (kiblnd_dev_can_failover(dev)) {
3471                                 list_add_tail(&dev->ibd_fail_list,
3472                                               &kiblnd_data.kib_failed_devs);
3473                         }
3474                 }
3475         }
3476
3477         write_unlock_irqrestore(glock, flags);
3478
3479         kiblnd_thread_fini();
3480         return 0;
3481 }