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.gnu.org/licenses/gpl-2.0.html
  19  *
  20  * GPL HEADER END
  21  */
  22 /*
  23  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Use is subject to license terms.
  25  *
  26  * Copyright (c) 2012, 2015, Intel Corporation.
  27  */
  28 /*
  29  * This file is part of Lustre, http://www.lustre.org/
  30  * Lustre is a trademark of Sun Microsystems, Inc.
  31  *
  32  * lnet/klnds/o2iblnd/o2iblnd_cb.c
  33  *
  34  * Author: Eric Barton <eric@bartonsoftware.com>
  35  */
  36
  37 #include "o2iblnd.h"
  38
  39 #define MAX_CONN_RACES_BEFORE_ABORT 20
  40
  41 static void kiblnd_peer_alive(struct kib_peer *peer);
  42 static void kiblnd_peer_connect_failed(struct kib_peer *peer, int active, int error);
  43 static void kiblnd_init_tx_msg(struct lnet_ni *ni, struct kib_tx *tx,
  44                                int type, int body_nob);
  45 static int kiblnd_init_rdma(struct kib_conn *conn, struct kib_tx *tx, int type,
  46                             int resid, struct kib_rdma_desc *dstrd,
  47                             __u64 dstcookie);
  48 static void kiblnd_queue_tx_locked(struct kib_tx *tx, struct kib_conn *conn);
  49 static void kiblnd_queue_tx(struct kib_tx *tx, struct kib_conn *conn);
  50 static void kiblnd_unmap_tx(struct lnet_ni *ni, struct kib_tx *tx);
  51 static void kiblnd_check_sends_locked(struct kib_conn *conn);
  52
  53 static void
  54 kiblnd_tx_done(struct lnet_ni *ni, struct kib_tx *tx)
  55 {
  56         struct lnet_msg *lntmsg[2];
  57         struct kib_net *net = ni->ni_data;
  58         int rc;
  59         int i;
  60
  61         LASSERT(net);
  62         LASSERT(!in_interrupt());
  63         LASSERT(!tx->tx_queued);               /* mustn't be queued for sending */
  64         LASSERT(!tx->tx_sending);         /* mustn't be awaiting sent callback */
  65         LASSERT(!tx->tx_waiting);             /* mustn't be awaiting peer response */
  66         LASSERT(tx->tx_pool);
  67
  68         kiblnd_unmap_tx(ni, tx);
  69
  70         /* tx may have up to 2 lnet msgs to finalise */
  71         lntmsg[0] = tx->tx_lntmsg[0]; tx->tx_lntmsg[0] = NULL;
  72         lntmsg[1] = tx->tx_lntmsg[1]; tx->tx_lntmsg[1] = NULL;
  73         rc = tx->tx_status;
  74
  75         if (tx->tx_conn) {
  76                 LASSERT(ni == tx->tx_conn->ibc_peer->ibp_ni);
  77
  78                 kiblnd_conn_decref(tx->tx_conn);
  79                 tx->tx_conn = NULL;
  80         }
  81
  82         tx->tx_nwrq = 0;
  83         tx->tx_status = 0;
  84
  85         kiblnd_pool_free_node(&tx->tx_pool->tpo_pool, &tx->tx_list);
  86
  87         /* delay finalize until my descs have been freed */
  88         for (i = 0; i < 2; i++) {
  89                 if (!lntmsg[i])
  90                         continue;
  91
  92                 lnet_finalize(ni, lntmsg[i], rc);
  93         }
  94 }
  95
  96 void
  97 kiblnd_txlist_done(struct lnet_ni *ni, struct list_head *txlist, int status)
  98 {
  99         struct kib_tx *tx;
 100
 101         while (!list_empty(txlist)) {
 102                 tx = list_entry(txlist->next, struct kib_tx, tx_list);
 103
 104                 list_del(&tx->tx_list);
 105                 /* complete now */
 106                 tx->tx_waiting = 0;
 107                 tx->tx_status = status;
 108                 kiblnd_tx_done(ni, tx);
 109         }
 110 }
 111
 112 static struct kib_tx *
 113 kiblnd_get_idle_tx(struct lnet_ni *ni, lnet_nid_t target)
 114 {
 115         struct kib_net *net = (struct kib_net *)ni->ni_data;
 116         struct list_head *node;
 117         struct kib_tx *tx;
 118         struct kib_tx_poolset *tps;
 119
 120         tps = net->ibn_tx_ps[lnet_cpt_of_nid(target)];
 121         node = kiblnd_pool_alloc_node(&tps->tps_poolset);
 122         if (!node)
 123                 return NULL;
 124         tx = list_entry(node, struct kib_tx, tx_list);
 125
 126         LASSERT(!tx->tx_nwrq);
 127         LASSERT(!tx->tx_queued);
 128         LASSERT(!tx->tx_sending);
 129         LASSERT(!tx->tx_waiting);
 130         LASSERT(!tx->tx_status);
 131         LASSERT(!tx->tx_conn);
 132         LASSERT(!tx->tx_lntmsg[0]);
 133         LASSERT(!tx->tx_lntmsg[1]);
 134         LASSERT(!tx->tx_nfrags);
 135
 136         return tx;
 137 }
 138
 139 static void
 140 kiblnd_drop_rx(struct kib_rx *rx)
 141 {
 142         struct kib_conn *conn = rx->rx_conn;
 143         struct kib_sched_info *sched = conn->ibc_sched;
 144         unsigned long flags;
 145
 146         spin_lock_irqsave(&sched->ibs_lock, flags);
 147         LASSERT(conn->ibc_nrx > 0);
 148         conn->ibc_nrx--;
 149         spin_unlock_irqrestore(&sched->ibs_lock, flags);
 150
 151         kiblnd_conn_decref(conn);
 152 }
 153
 154 int
 155 kiblnd_post_rx(struct kib_rx *rx, int credit)
 156 {
 157         struct kib_conn *conn = rx->rx_conn;
 158         struct kib_net *net = conn->ibc_peer->ibp_ni->ni_data;
 159         struct ib_recv_wr *bad_wrq = NULL;
 160         int rc;
 161
 162         LASSERT(net);
 163         LASSERT(!in_interrupt());
 164         LASSERT(credit == IBLND_POSTRX_NO_CREDIT ||
 165                 credit == IBLND_POSTRX_PEER_CREDIT ||
 166                 credit == IBLND_POSTRX_RSRVD_CREDIT);
 167
 168         rx->rx_sge.lkey   = conn->ibc_hdev->ibh_pd->local_dma_lkey;
 169         rx->rx_sge.addr   = rx->rx_msgaddr;
 170         rx->rx_sge.length = IBLND_MSG_SIZE;
 171
 172         rx->rx_wrq.next    = NULL;
 173         rx->rx_wrq.sg_list = &rx->rx_sge;
 174         rx->rx_wrq.num_sge = 1;
 175         rx->rx_wrq.wr_id   = kiblnd_ptr2wreqid(rx, IBLND_WID_RX);
 176
 177         LASSERT(conn->ibc_state >= IBLND_CONN_INIT);
 178         LASSERT(rx->rx_nob >= 0);             /* not posted */
 179
 180         if (conn->ibc_state > IBLND_CONN_ESTABLISHED) {
 181                 kiblnd_drop_rx(rx);          /* No more posts for this rx */
 182                 return 0;
 183         }
 184
 185         rx->rx_nob = -1;                        /* flag posted */
 186
 187         /* NB: need an extra reference after ib_post_recv because we don't
 188          * own this rx (and rx::rx_conn) anymore, LU-5678.
 189          */
 190         kiblnd_conn_addref(conn);
 191         rc = ib_post_recv(conn->ibc_cmid->qp, &rx->rx_wrq, &bad_wrq);
 192         if (unlikely(rc)) {
 193                 CERROR("Can't post rx for %s: %d, bad_wrq: %p\n",
 194                        libcfs_nid2str(conn->ibc_peer->ibp_nid), rc, bad_wrq);
 195                 rx->rx_nob = 0;
 196         }
 197
 198         if (conn->ibc_state < IBLND_CONN_ESTABLISHED) /* Initial post */
 199                 goto out;
 200
 201         if (unlikely(rc)) {
 202                 kiblnd_close_conn(conn, rc);
 203                 kiblnd_drop_rx(rx);          /* No more posts for this rx */
 204                 goto out;
 205         }
 206
 207         if (credit == IBLND_POSTRX_NO_CREDIT)
 208                 goto out;
 209
 210         spin_lock(&conn->ibc_lock);
 211         if (credit == IBLND_POSTRX_PEER_CREDIT)
 212                 conn->ibc_outstanding_credits++;
 213         else
 214                 conn->ibc_reserved_credits++;
 215         kiblnd_check_sends_locked(conn);
 216         spin_unlock(&conn->ibc_lock);
 217
 218 out:
 219         kiblnd_conn_decref(conn);
 220         return rc;
 221 }
 222
 223 static struct kib_tx *
 224 kiblnd_find_waiting_tx_locked(struct kib_conn *conn, int txtype, __u64 cookie)
 225 {
 226         struct list_head *tmp;
 227
 228         list_for_each(tmp, &conn->ibc_active_txs) {
 229                 struct kib_tx *tx = list_entry(tmp, struct kib_tx, tx_list);
 230
 231                 LASSERT(!tx->tx_queued);
 232                 LASSERT(tx->tx_sending || tx->tx_waiting);
 233
 234                 if (tx->tx_cookie != cookie)
 235                         continue;
 236
 237                 if (tx->tx_waiting &&
 238                     tx->tx_msg->ibm_type == txtype)
 239                         return tx;
 240
 241                 CWARN("Bad completion: %swaiting, type %x (wanted %x)\n",
 242                       tx->tx_waiting ? "" : "NOT ",
 243                       tx->tx_msg->ibm_type, txtype);
 244         }
 245         return NULL;
 246 }
 247
 248 static void
 249 kiblnd_handle_completion(struct kib_conn *conn, int txtype, int status, __u64 cookie)
 250 {
 251         struct kib_tx *tx;
 252         struct lnet_ni *ni = conn->ibc_peer->ibp_ni;
 253         int idle;
 254
 255         spin_lock(&conn->ibc_lock);
 256
 257         tx = kiblnd_find_waiting_tx_locked(conn, txtype, cookie);
 258         if (!tx) {
 259                 spin_unlock(&conn->ibc_lock);
 260
 261                 CWARN("Unmatched completion type %x cookie %#llx from %s\n",
 262                       txtype, cookie, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 263                 kiblnd_close_conn(conn, -EPROTO);
 264                 return;
 265         }
 266
 267         if (!tx->tx_status) {          /* success so far */
 268                 if (status < 0) /* failed? */
 269                         tx->tx_status = status;
 270                 else if (txtype == IBLND_MSG_GET_REQ)
 271                         lnet_set_reply_msg_len(ni, tx->tx_lntmsg[1], status);
 272         }
 273
 274         tx->tx_waiting = 0;
 275
 276         idle = !tx->tx_queued && !tx->tx_sending;
 277         if (idle)
 278                 list_del(&tx->tx_list);
 279
 280         spin_unlock(&conn->ibc_lock);
 281
 282         if (idle)
 283                 kiblnd_tx_done(ni, tx);
 284 }
 285
 286 static void
 287 kiblnd_send_completion(struct kib_conn *conn, int type, int status, __u64 cookie)
 288 {
 289         struct lnet_ni *ni = conn->ibc_peer->ibp_ni;
 290         struct kib_tx *tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
 291
 292         if (!tx) {
 293                 CERROR("Can't get tx for completion %x for %s\n",
 294                        type, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 295                 return;
 296         }
 297
 298         tx->tx_msg->ibm_u.completion.ibcm_status = status;
 299         tx->tx_msg->ibm_u.completion.ibcm_cookie = cookie;
 300         kiblnd_init_tx_msg(ni, tx, type, sizeof(struct kib_completion_msg));
 301
 302         kiblnd_queue_tx(tx, conn);
 303 }
 304
 305 static void
 306 kiblnd_handle_rx(struct kib_rx *rx)
 307 {
 308         struct kib_msg *msg = rx->rx_msg;
 309         struct kib_conn *conn = rx->rx_conn;
 310         struct lnet_ni *ni = conn->ibc_peer->ibp_ni;
 311         int credits = msg->ibm_credits;
 312         struct kib_tx *tx;
 313         int rc = 0;
 314         int rc2;
 315         int post_credit;
 316
 317         LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
 318
 319         CDEBUG(D_NET, "Received %x[%d] from %s\n",
 320                msg->ibm_type, credits,
 321                libcfs_nid2str(conn->ibc_peer->ibp_nid));
 322
 323         if (credits) {
 324                 /* Have I received credits that will let me send? */
 325                 spin_lock(&conn->ibc_lock);
 326
 327                 if (conn->ibc_credits + credits >
 328                     conn->ibc_queue_depth) {
 329                         rc2 = conn->ibc_credits;
 330                         spin_unlock(&conn->ibc_lock);
 331
 332                         CERROR("Bad credits from %s: %d + %d > %d\n",
 333                                libcfs_nid2str(conn->ibc_peer->ibp_nid),
 334                                rc2, credits, conn->ibc_queue_depth);
 335
 336                         kiblnd_close_conn(conn, -EPROTO);
 337                         kiblnd_post_rx(rx, IBLND_POSTRX_NO_CREDIT);
 338                         return;
 339                 }
 340
 341                 conn->ibc_credits += credits;
 342
 343                 /* This ensures the credit taken by NOOP can be returned */
 344                 if (msg->ibm_type == IBLND_MSG_NOOP &&
 345                     !IBLND_OOB_CAPABLE(conn->ibc_version)) /* v1 only */
 346                         conn->ibc_outstanding_credits++;
 347
 348                 kiblnd_check_sends_locked(conn);
 349                 spin_unlock(&conn->ibc_lock);
 350         }
 351
 352         switch (msg->ibm_type) {
 353         default:
 354                 CERROR("Bad IBLND message type %x from %s\n",
 355                        msg->ibm_type, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 356                 post_credit = IBLND_POSTRX_NO_CREDIT;
 357                 rc = -EPROTO;
 358                 break;
 359
 360         case IBLND_MSG_NOOP:
 361                 if (IBLND_OOB_CAPABLE(conn->ibc_version)) {
 362                         post_credit = IBLND_POSTRX_NO_CREDIT;
 363                         break;
 364                 }
 365
 366                 if (credits) /* credit already posted */
 367                         post_credit = IBLND_POSTRX_NO_CREDIT;
 368                 else          /* a keepalive NOOP */
 369                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 370                 break;
 371
 372         case IBLND_MSG_IMMEDIATE:
 373                 post_credit = IBLND_POSTRX_DONT_POST;
 374                 rc = lnet_parse(ni, &msg->ibm_u.immediate.ibim_hdr,
 375                                 msg->ibm_srcnid, rx, 0);
 376                 if (rc < 0)                  /* repost on error */
 377                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 378                 break;
 379
 380         case IBLND_MSG_PUT_REQ:
 381                 post_credit = IBLND_POSTRX_DONT_POST;
 382                 rc = lnet_parse(ni, &msg->ibm_u.putreq.ibprm_hdr,
 383                                 msg->ibm_srcnid, rx, 1);
 384                 if (rc < 0)                  /* repost on error */
 385                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 386                 break;
 387
 388         case IBLND_MSG_PUT_NAK:
 389                 CWARN("PUT_NACK from %s\n",
 390                       libcfs_nid2str(conn->ibc_peer->ibp_nid));
 391                 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
 392                 kiblnd_handle_completion(conn, IBLND_MSG_PUT_REQ,
 393                                          msg->ibm_u.completion.ibcm_status,
 394                                          msg->ibm_u.completion.ibcm_cookie);
 395                 break;
 396
 397         case IBLND_MSG_PUT_ACK:
 398                 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
 399
 400                 spin_lock(&conn->ibc_lock);
 401                 tx = kiblnd_find_waiting_tx_locked(conn, IBLND_MSG_PUT_REQ,
 402                                                    msg->ibm_u.putack.ibpam_src_cookie);
 403                 if (tx)
 404                         list_del(&tx->tx_list);
 405                 spin_unlock(&conn->ibc_lock);
 406
 407                 if (!tx) {
 408                         CERROR("Unmatched PUT_ACK from %s\n",
 409                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
 410                         rc = -EPROTO;
 411                         break;
 412                 }
 413
 414                 LASSERT(tx->tx_waiting);
 415                 /*
 416                  * CAVEAT EMPTOR: I could be racing with tx_complete, but...
 417                  * (a) I can overwrite tx_msg since my peer has received it!
 418                  * (b) tx_waiting set tells tx_complete() it's not done.
 419                  */
 420                 tx->tx_nwrq = 0;                /* overwrite PUT_REQ */
 421
 422                 rc2 = kiblnd_init_rdma(conn, tx, IBLND_MSG_PUT_DONE,
 423                                        kiblnd_rd_size(&msg->ibm_u.putack.ibpam_rd),
 424                                        &msg->ibm_u.putack.ibpam_rd,
 425                                        msg->ibm_u.putack.ibpam_dst_cookie);
 426                 if (rc2 < 0)
 427                         CERROR("Can't setup rdma for PUT to %s: %d\n",
 428                                libcfs_nid2str(conn->ibc_peer->ibp_nid), rc2);
 429
 430                 spin_lock(&conn->ibc_lock);
 431                 tx->tx_waiting = 0;     /* clear waiting and queue atomically */
 432                 kiblnd_queue_tx_locked(tx, conn);
 433                 spin_unlock(&conn->ibc_lock);
 434                 break;
 435
 436         case IBLND_MSG_PUT_DONE:
 437                 post_credit = IBLND_POSTRX_PEER_CREDIT;
 438                 kiblnd_handle_completion(conn, IBLND_MSG_PUT_ACK,
 439                                          msg->ibm_u.completion.ibcm_status,
 440                                          msg->ibm_u.completion.ibcm_cookie);
 441                 break;
 442
 443         case IBLND_MSG_GET_REQ:
 444                 post_credit = IBLND_POSTRX_DONT_POST;
 445                 rc = lnet_parse(ni, &msg->ibm_u.get.ibgm_hdr,
 446                                 msg->ibm_srcnid, rx, 1);
 447                 if (rc < 0)                  /* repost on error */
 448                         post_credit = IBLND_POSTRX_PEER_CREDIT;
 449                 break;
 450
 451         case IBLND_MSG_GET_DONE:
 452                 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
 453                 kiblnd_handle_completion(conn, IBLND_MSG_GET_REQ,
 454                                          msg->ibm_u.completion.ibcm_status,
 455                                          msg->ibm_u.completion.ibcm_cookie);
 456                 break;
 457         }
 458
 459         if (rc < 0)                          /* protocol error */
 460                 kiblnd_close_conn(conn, rc);
 461
 462         if (post_credit != IBLND_POSTRX_DONT_POST)
 463                 kiblnd_post_rx(rx, post_credit);
 464 }
 465
 466 static void
 467 kiblnd_rx_complete(struct kib_rx *rx, int status, int nob)
 468 {
 469         struct kib_msg *msg = rx->rx_msg;
 470         struct kib_conn *conn = rx->rx_conn;
 471         struct lnet_ni *ni = conn->ibc_peer->ibp_ni;
 472         struct kib_net *net = ni->ni_data;
 473         int rc;
 474         int err = -EIO;
 475
 476         LASSERT(net);
 477         LASSERT(rx->rx_nob < 0);               /* was posted */
 478         rx->rx_nob = 0;                  /* isn't now */
 479
 480         if (conn->ibc_state > IBLND_CONN_ESTABLISHED)
 481                 goto ignore;
 482
 483         if (status != IB_WC_SUCCESS) {
 484                 CNETERR("Rx from %s failed: %d\n",
 485                         libcfs_nid2str(conn->ibc_peer->ibp_nid), status);
 486                 goto failed;
 487         }
 488
 489         LASSERT(nob >= 0);
 490         rx->rx_nob = nob;
 491
 492         rc = kiblnd_unpack_msg(msg, rx->rx_nob);
 493         if (rc) {
 494                 CERROR("Error %d unpacking rx from %s\n",
 495                        rc, libcfs_nid2str(conn->ibc_peer->ibp_nid));
 496                 goto failed;
 497         }
 498
 499         if (msg->ibm_srcnid != conn->ibc_peer->ibp_nid ||
 500             msg->ibm_dstnid != ni->ni_nid ||
 501             msg->ibm_srcstamp != conn->ibc_incarnation ||
 502             msg->ibm_dststamp != net->ibn_incarnation) {
 503                 CERROR("Stale rx from %s\n",
 504                        libcfs_nid2str(conn->ibc_peer->ibp_nid));
 505                 err = -ESTALE;
 506                 goto failed;
 507         }
 508
 509         /* set time last known alive */
 510         kiblnd_peer_alive(conn->ibc_peer);
 511
 512         /* racing with connection establishment/teardown! */
 513
 514         if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
 515                 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
 516                 unsigned long flags;
 517
 518                 write_lock_irqsave(g_lock, flags);
 519                 /* must check holding global lock to eliminate race */
 520                 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
 521                         list_add_tail(&rx->rx_list, &conn->ibc_early_rxs);
 522                         write_unlock_irqrestore(g_lock, flags);
 523                         return;
 524                 }
 525                 write_unlock_irqrestore(g_lock, flags);
 526         }
 527         kiblnd_handle_rx(rx);
 528         return;
 529
 530  failed:
 531         CDEBUG(D_NET, "rx %p conn %p\n", rx, conn);
 532         kiblnd_close_conn(conn, err);
 533  ignore:
 534         kiblnd_drop_rx(rx);                  /* Don't re-post rx. */
 535 }
 536
 537 static struct page *
 538 kiblnd_kvaddr_to_page(unsigned long vaddr)
 539 {
 540         struct page *page;
 541
 542         if (is_vmalloc_addr((void *)vaddr)) {
 543                 page = vmalloc_to_page((void *)vaddr);
 544                 LASSERT(page);
 545                 return page;
 546         }
 547 #ifdef CONFIG_HIGHMEM
 548         if (vaddr >= PKMAP_BASE &&
 549             vaddr < (PKMAP_BASE + LAST_PKMAP * PAGE_SIZE)) {
 550                 /* No highmem pages only used for bulk (kiov) I/O */
 551                 CERROR("find page for address in highmem\n");
 552                 LBUG();
 553         }
 554 #endif
 555         page = virt_to_page(vaddr);
 556         LASSERT(page);
 557         return page;
 558 }
 559
 560 static int
 561 kiblnd_fmr_map_tx(struct kib_net *net, struct kib_tx *tx, struct kib_rdma_desc *rd, __u32 nob)
 562 {
 563         struct kib_hca_dev *hdev;
 564         struct kib_fmr_poolset *fps;
 565         int cpt;
 566         int rc;
 567
 568         LASSERT(tx->tx_pool);
 569         LASSERT(tx->tx_pool->tpo_pool.po_owner);
 570
 571         hdev = tx->tx_pool->tpo_hdev;
 572         cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
 573
 574         fps = net->ibn_fmr_ps[cpt];
 575         rc = kiblnd_fmr_pool_map(fps, tx, rd, nob, 0, &tx->fmr);
 576         if (rc) {
 577                 CERROR("Can't map %u bytes: %d\n", nob, rc);
 578                 return rc;
 579         }
 580
 581         /*
 582          * If rd is not tx_rd, it's going to get sent to a peer, who will need
 583          * the rkey
 584          */
 585         rd->rd_key = tx->fmr.fmr_key;
 586         rd->rd_frags[0].rf_addr &= ~hdev->ibh_page_mask;
 587         rd->rd_frags[0].rf_nob = nob;
 588         rd->rd_nfrags = 1;
 589
 590         return 0;
 591 }
 592
 593 static void kiblnd_unmap_tx(struct lnet_ni *ni, struct kib_tx *tx)
 594 {
 595         struct kib_net *net = ni->ni_data;
 596
 597         LASSERT(net);
 598
 599         if (net->ibn_fmr_ps)
 600                 kiblnd_fmr_pool_unmap(&tx->fmr, tx->tx_status);
 601
 602         if (tx->tx_nfrags) {
 603                 kiblnd_dma_unmap_sg(tx->tx_pool->tpo_hdev->ibh_ibdev,
 604                                     tx->tx_frags, tx->tx_nfrags, tx->tx_dmadir);
 605                 tx->tx_nfrags = 0;
 606         }
 607 }
 608
 609 static int kiblnd_map_tx(struct lnet_ni *ni, struct kib_tx *tx,
 610                          struct kib_rdma_desc *rd, int nfrags)
 611 {
 612         struct kib_net *net = ni->ni_data;
 613         struct kib_hca_dev *hdev = net->ibn_dev->ibd_hdev;
 614         __u32 nob;
 615         int i;
 616
 617         /*
 618          * If rd is not tx_rd, it's going to get sent to a peer and I'm the
 619          * RDMA sink
 620          */
 621         tx->tx_dmadir = (rd != tx->tx_rd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
 622         tx->tx_nfrags = nfrags;
 623
 624         rd->rd_nfrags = kiblnd_dma_map_sg(hdev->ibh_ibdev, tx->tx_frags,
 625                                           tx->tx_nfrags, tx->tx_dmadir);
 626
 627         for (i = 0, nob = 0; i < rd->rd_nfrags; i++) {
 628                 rd->rd_frags[i].rf_nob  = kiblnd_sg_dma_len(
 629                         hdev->ibh_ibdev, &tx->tx_frags[i]);
 630                 rd->rd_frags[i].rf_addr = kiblnd_sg_dma_address(
 631                         hdev->ibh_ibdev, &tx->tx_frags[i]);
 632                 nob += rd->rd_frags[i].rf_nob;
 633         }
 634
 635         if (net->ibn_fmr_ps)
 636                 return kiblnd_fmr_map_tx(net, tx, rd, nob);
 637
 638         return -EINVAL;
 639 }
 640
 641 static int
 642 kiblnd_setup_rd_iov(struct lnet_ni *ni, struct kib_tx *tx,
 643                     struct kib_rdma_desc *rd, unsigned int niov,
 644                     const struct kvec *iov, int offset, int nob)
 645 {
 646         struct kib_net *net = ni->ni_data;
 647         struct page *page;
 648         struct scatterlist *sg;
 649         unsigned long vaddr;
 650         int fragnob;
 651         int page_offset;
 652
 653         LASSERT(nob > 0);
 654         LASSERT(niov > 0);
 655         LASSERT(net);
 656
 657         while (offset >= iov->iov_len) {
 658                 offset -= iov->iov_len;
 659                 niov--;
 660                 iov++;
 661                 LASSERT(niov > 0);
 662         }
 663
 664         sg = tx->tx_frags;
 665         do {
 666                 LASSERT(niov > 0);
 667
 668                 vaddr = ((unsigned long)iov->iov_base) + offset;
 669                 page_offset = vaddr & (PAGE_SIZE - 1);
 670                 page = kiblnd_kvaddr_to_page(vaddr);
 671                 if (!page) {
 672                         CERROR("Can't find page\n");
 673                         return -EFAULT;
 674                 }
 675
 676                 fragnob = min((int)(iov->iov_len - offset), nob);
 677                 fragnob = min(fragnob, (int)PAGE_SIZE - page_offset);
 678
 679                 sg_set_page(sg, page, fragnob, page_offset);
 680                 sg = sg_next(sg);
 681                 if (!sg) {
 682                         CERROR("lacking enough sg entries to map tx\n");
 683                         return -EFAULT;
 684                 }
 685
 686                 if (offset + fragnob < iov->iov_len) {
 687                         offset += fragnob;
 688                 } else {
 689                         offset = 0;
 690                         iov++;
 691                         niov--;
 692                 }
 693                 nob -= fragnob;
 694         } while (nob > 0);
 695
 696         return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
 697 }
 698
 699 static int
 700 kiblnd_setup_rd_kiov(struct lnet_ni *ni, struct kib_tx *tx,
 701                      struct kib_rdma_desc *rd, int nkiov,
 702                      const struct bio_vec *kiov, int offset, int nob)
 703 {
 704         struct kib_net *net = ni->ni_data;
 705         struct scatterlist *sg;
 706         int fragnob;
 707
 708         CDEBUG(D_NET, "niov %d offset %d nob %d\n", nkiov, offset, nob);
 709
 710         LASSERT(nob > 0);
 711         LASSERT(nkiov > 0);
 712         LASSERT(net);
 713
 714         while (offset >= kiov->bv_len) {
 715                 offset -= kiov->bv_len;
 716                 nkiov--;
 717                 kiov++;
 718                 LASSERT(nkiov > 0);
 719         }
 720
 721         sg = tx->tx_frags;
 722         do {
 723                 LASSERT(nkiov > 0);
 724
 725                 fragnob = min((int)(kiov->bv_len - offset), nob);
 726
 727                 sg_set_page(sg, kiov->bv_page, fragnob,
 728                             kiov->bv_offset + offset);
 729                 sg = sg_next(sg);
 730                 if (!sg) {
 731                         CERROR("lacking enough sg entries to map tx\n");
 732                         return -EFAULT;
 733                 }
 734
 735                 offset = 0;
 736                 kiov++;
 737                 nkiov--;
 738                 nob -= fragnob;
 739         } while (nob > 0);
 740
 741         return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
 742 }
 743
 744 static int
 745 kiblnd_post_tx_locked(struct kib_conn *conn, struct kib_tx *tx, int credit)
 746         __must_hold(&conn->ibc_lock)
 747 {
 748         struct kib_msg *msg = tx->tx_msg;
 749         struct kib_peer *peer = conn->ibc_peer;
 750         struct lnet_ni *ni = peer->ibp_ni;
 751         int ver = conn->ibc_version;
 752         int rc;
 753         int done;
 754
 755         LASSERT(tx->tx_queued);
 756         /* We rely on this for QP sizing */
 757         LASSERT(tx->tx_nwrq > 0);
 758
 759         LASSERT(!credit || credit == 1);
 760         LASSERT(conn->ibc_outstanding_credits >= 0);
 761         LASSERT(conn->ibc_outstanding_credits <= conn->ibc_queue_depth);
 762         LASSERT(conn->ibc_credits >= 0);
 763         LASSERT(conn->ibc_credits <= conn->ibc_queue_depth);
 764
 765         if (conn->ibc_nsends_posted == kiblnd_concurrent_sends(ver, ni)) {
 766                 /* tx completions outstanding... */
 767                 CDEBUG(D_NET, "%s: posted enough\n",
 768                        libcfs_nid2str(peer->ibp_nid));
 769                 return -EAGAIN;
 770         }
 771
 772         if (credit && !conn->ibc_credits) {   /* no credits */
 773                 CDEBUG(D_NET, "%s: no credits\n",
 774                        libcfs_nid2str(peer->ibp_nid));
 775                 return -EAGAIN;
 776         }
 777
 778         if (credit && !IBLND_OOB_CAPABLE(ver) &&
 779             conn->ibc_credits == 1 &&   /* last credit reserved */
 780             msg->ibm_type != IBLND_MSG_NOOP) {      /* for NOOP */
 781                 CDEBUG(D_NET, "%s: not using last credit\n",
 782                        libcfs_nid2str(peer->ibp_nid));
 783                 return -EAGAIN;
 784         }
 785
 786         /* NB don't drop ibc_lock before bumping tx_sending */
 787         list_del(&tx->tx_list);
 788         tx->tx_queued = 0;
 789
 790         if (msg->ibm_type == IBLND_MSG_NOOP &&
 791             (!kiblnd_need_noop(conn) ||     /* redundant NOOP */
 792              (IBLND_OOB_CAPABLE(ver) && /* posted enough NOOP */
 793               conn->ibc_noops_posted == IBLND_OOB_MSGS(ver)))) {
 794                 /*
 795                  * OK to drop when posted enough NOOPs, since
 796                  * kiblnd_check_sends_locked will queue NOOP again when
 797                  * posted NOOPs complete
 798                  */
 799                 spin_unlock(&conn->ibc_lock);
 800                 kiblnd_tx_done(peer->ibp_ni, tx);
 801                 spin_lock(&conn->ibc_lock);
 802                 CDEBUG(D_NET, "%s(%d): redundant or enough NOOP\n",
 803                        libcfs_nid2str(peer->ibp_nid),
 804                        conn->ibc_noops_posted);
 805                 return 0;
 806         }
 807
 808         kiblnd_pack_msg(peer->ibp_ni, msg, ver, conn->ibc_outstanding_credits,
 809                         peer->ibp_nid, conn->ibc_incarnation);
 810
 811         conn->ibc_credits -= credit;
 812         conn->ibc_outstanding_credits = 0;
 813         conn->ibc_nsends_posted++;
 814         if (msg->ibm_type == IBLND_MSG_NOOP)
 815                 conn->ibc_noops_posted++;
 816
 817         /*
 818          * CAVEAT EMPTOR!  This tx could be the PUT_DONE of an RDMA
 819          * PUT.  If so, it was first queued here as a PUT_REQ, sent and
 820          * stashed on ibc_active_txs, matched by an incoming PUT_ACK,
 821          * and then re-queued here.  It's (just) possible that
 822          * tx_sending is non-zero if we've not done the tx_complete()
 823          * from the first send; hence the ++ rather than = below.
 824          */
 825         tx->tx_sending++;
 826         list_add(&tx->tx_list, &conn->ibc_active_txs);
 827
 828         /* I'm still holding ibc_lock! */
 829         if (conn->ibc_state != IBLND_CONN_ESTABLISHED) {
 830                 rc = -ECONNABORTED;
 831         } else if (tx->tx_pool->tpo_pool.po_failed ||
 832                  conn->ibc_hdev != tx->tx_pool->tpo_hdev) {
 833                 /* close_conn will launch failover */
 834                 rc = -ENETDOWN;
 835         } else {
 836                 struct kib_fast_reg_descriptor *frd = tx->fmr.fmr_frd;
 837                 struct ib_send_wr *bad = &tx->tx_wrq[tx->tx_nwrq - 1].wr;
 838                 struct ib_send_wr *wrq = &tx->tx_wrq[0].wr;
 839
 840                 if (frd) {
 841                         if (!frd->frd_valid) {
 842                                 wrq = &frd->frd_inv_wr;
 843                                 wrq->next = &frd->frd_fastreg_wr.wr;
 844                         } else {
 845                                 wrq = &frd->frd_fastreg_wr.wr;
 846                         }
 847                         frd->frd_fastreg_wr.wr.next = &tx->tx_wrq[0].wr;
 848                 }
 849
 850                 LASSERTF(bad->wr_id == kiblnd_ptr2wreqid(tx, IBLND_WID_TX),
 851                          "bad wr_id %llx, opc %d, flags %d, peer: %s\n",
 852                          bad->wr_id, bad->opcode, bad->send_flags,
 853                          libcfs_nid2str(conn->ibc_peer->ibp_nid));
 854                 bad = NULL;
 855                 rc = ib_post_send(conn->ibc_cmid->qp, wrq, &bad);
 856         }
 857
 858         conn->ibc_last_send = jiffies;
 859
 860         if (!rc)
 861                 return 0;
 862
 863         /*
 864          * NB credits are transferred in the actual
 865          * message, which can only be the last work item
 866          */
 867         conn->ibc_credits += credit;
 868         conn->ibc_outstanding_credits += msg->ibm_credits;
 869         conn->ibc_nsends_posted--;
 870         if (msg->ibm_type == IBLND_MSG_NOOP)
 871                 conn->ibc_noops_posted--;
 872
 873         tx->tx_status = rc;
 874         tx->tx_waiting = 0;
 875         tx->tx_sending--;
 876
 877         done = !tx->tx_sending;
 878         if (done)
 879                 list_del(&tx->tx_list);
 880
 881         spin_unlock(&conn->ibc_lock);
 882
 883         if (conn->ibc_state == IBLND_CONN_ESTABLISHED)
 884                 CERROR("Error %d posting transmit to %s\n",
 885                        rc, libcfs_nid2str(peer->ibp_nid));
 886         else
 887                 CDEBUG(D_NET, "Error %d posting transmit to %s\n",
 888                        rc, libcfs_nid2str(peer->ibp_nid));
 889
 890         kiblnd_close_conn(conn, rc);
 891
 892         if (done)
 893                 kiblnd_tx_done(peer->ibp_ni, tx);
 894
 895         spin_lock(&conn->ibc_lock);
 896
 897         return -EIO;
 898 }
 899
 900 static void
 901 kiblnd_check_sends_locked(struct kib_conn *conn)
 902 {
 903         int ver = conn->ibc_version;
 904         struct lnet_ni *ni = conn->ibc_peer->ibp_ni;
 905         struct kib_tx *tx;
 906
 907         /* Don't send anything until after the connection is established */
 908         if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
 909                 CDEBUG(D_NET, "%s too soon\n",
 910                        libcfs_nid2str(conn->ibc_peer->ibp_nid));
 911                 return;
 912         }
 913
 914         LASSERT(conn->ibc_nsends_posted <= kiblnd_concurrent_sends(ver, ni));
 915         LASSERT(!IBLND_OOB_CAPABLE(ver) ||
 916                 conn->ibc_noops_posted <= IBLND_OOB_MSGS(ver));
 917         LASSERT(conn->ibc_reserved_credits >= 0);
 918
 919         while (conn->ibc_reserved_credits > 0 &&
 920                !list_empty(&conn->ibc_tx_queue_rsrvd)) {
 921                 tx = list_entry(conn->ibc_tx_queue_rsrvd.next,
 922                                 struct kib_tx, tx_list);
 923                 list_del(&tx->tx_list);
 924                 list_add_tail(&tx->tx_list, &conn->ibc_tx_queue);
 925                 conn->ibc_reserved_credits--;
 926         }
 927
 928         if (kiblnd_need_noop(conn)) {
 929                 spin_unlock(&conn->ibc_lock);
 930
 931                 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
 932                 if (tx)
 933                         kiblnd_init_tx_msg(ni, tx, IBLND_MSG_NOOP, 0);
 934
 935                 spin_lock(&conn->ibc_lock);
 936                 if (tx)
 937                         kiblnd_queue_tx_locked(tx, conn);
 938         }
 939
 940         for (;;) {
 941                 int credit;
 942
 943                 if (!list_empty(&conn->ibc_tx_queue_nocred)) {
 944                         credit = 0;
 945                         tx = list_entry(conn->ibc_tx_queue_nocred.next,
 946                                         struct kib_tx, tx_list);
 947                 } else if (!list_empty(&conn->ibc_tx_noops)) {
 948                         LASSERT(!IBLND_OOB_CAPABLE(ver));
 949                         credit = 1;
 950                         tx = list_entry(conn->ibc_tx_noops.next,
 951                                         struct kib_tx, tx_list);
 952                 } else if (!list_empty(&conn->ibc_tx_queue)) {
 953                         credit = 1;
 954                         tx = list_entry(conn->ibc_tx_queue.next,
 955                                         struct kib_tx, tx_list);
 956                 } else {
 957                         break;
 958                 }
 959
 960                 if (kiblnd_post_tx_locked(conn, tx, credit))
 961                         break;
 962         }
 963 }
 964
 965 static void
 966 kiblnd_tx_complete(struct kib_tx *tx, int status)
 967 {
 968         int failed = (status != IB_WC_SUCCESS);
 969         struct kib_conn *conn = tx->tx_conn;
 970         int idle;
 971
 972         LASSERT(tx->tx_sending > 0);
 973
 974         if (failed) {
 975                 if (conn->ibc_state == IBLND_CONN_ESTABLISHED)
 976                         CNETERR("Tx -> %s cookie %#llx sending %d waiting %d: failed %d\n",
 977                                 libcfs_nid2str(conn->ibc_peer->ibp_nid),
 978                                 tx->tx_cookie, tx->tx_sending, tx->tx_waiting,
 979                                 status);
 980
 981                 kiblnd_close_conn(conn, -EIO);
 982         } else {
 983                 kiblnd_peer_alive(conn->ibc_peer);
 984         }
 985
 986         spin_lock(&conn->ibc_lock);
 987
 988         /*
 989          * I could be racing with rdma completion.  Whoever makes 'tx' idle
 990          * gets to free it, which also drops its ref on 'conn'.
 991          */
 992         tx->tx_sending--;
 993         conn->ibc_nsends_posted--;
 994         if (tx->tx_msg->ibm_type == IBLND_MSG_NOOP)
 995                 conn->ibc_noops_posted--;
 996
 997         if (failed) {
 998                 tx->tx_waiting = 0;          /* don't wait for peer */
 999                 tx->tx_status = -EIO;
1000         }
1001
1002         idle = !tx->tx_sending &&        /* This is the final callback */
1003                !tx->tx_waiting &&              /* Not waiting for peer */
1004                !tx->tx_queued;            /* Not re-queued (PUT_DONE) */
1005         if (idle)
1006                 list_del(&tx->tx_list);
1007
1008         kiblnd_check_sends_locked(conn);
1009         spin_unlock(&conn->ibc_lock);
1010
1011         if (idle)
1012                 kiblnd_tx_done(conn->ibc_peer->ibp_ni, tx);
1013 }
1014
1015 static void
1016 kiblnd_init_tx_msg(struct lnet_ni *ni, struct kib_tx *tx, int type,
1017                    int body_nob)
1018 {
1019         struct kib_hca_dev *hdev = tx->tx_pool->tpo_hdev;
1020         struct ib_sge *sge = &tx->tx_sge[tx->tx_nwrq];
1021         struct ib_rdma_wr *wrq = &tx->tx_wrq[tx->tx_nwrq];
1022         int nob = offsetof(struct kib_msg, ibm_u) + body_nob;
1023
1024         LASSERT(tx->tx_nwrq >= 0);
1025         LASSERT(tx->tx_nwrq < IBLND_MAX_RDMA_FRAGS + 1);
1026         LASSERT(nob <= IBLND_MSG_SIZE);
1027
1028         kiblnd_init_msg(tx->tx_msg, type, body_nob);
1029
1030         sge->lkey   = hdev->ibh_pd->local_dma_lkey;
1031         sge->addr   = tx->tx_msgaddr;
1032         sge->length = nob;
1033
1034         memset(wrq, 0, sizeof(*wrq));
1035
1036         wrq->wr.next       = NULL;
1037         wrq->wr.wr_id      = kiblnd_ptr2wreqid(tx, IBLND_WID_TX);
1038         wrq->wr.sg_list    = sge;
1039         wrq->wr.num_sge    = 1;
1040         wrq->wr.opcode     = IB_WR_SEND;
1041         wrq->wr.send_flags = IB_SEND_SIGNALED;
1042
1043         tx->tx_nwrq++;
1044 }
1045
1046 static int
1047 kiblnd_init_rdma(struct kib_conn *conn, struct kib_tx *tx, int type,
1048                  int resid, struct kib_rdma_desc *dstrd, __u64 dstcookie)
1049 {
1050         struct kib_msg *ibmsg = tx->tx_msg;
1051         struct kib_rdma_desc *srcrd = tx->tx_rd;
1052         struct ib_sge *sge = &tx->tx_sge[0];
1053         struct ib_rdma_wr *wrq, *next;
1054         int rc  = resid;
1055         int srcidx = 0;
1056         int dstidx = 0;
1057         int wrknob;
1058
1059         LASSERT(!in_interrupt());
1060         LASSERT(!tx->tx_nwrq);
1061         LASSERT(type == IBLND_MSG_GET_DONE ||
1062                 type == IBLND_MSG_PUT_DONE);
1063
1064         if (kiblnd_rd_size(srcrd) > conn->ibc_max_frags << PAGE_SHIFT) {
1065                 CERROR("RDMA is too large for peer %s (%d), src size: %d dst size: %d\n",
1066                        libcfs_nid2str(conn->ibc_peer->ibp_nid),
1067                        conn->ibc_max_frags << PAGE_SHIFT,
1068                        kiblnd_rd_size(srcrd), kiblnd_rd_size(dstrd));
1069                 rc = -EMSGSIZE;
1070                 goto too_big;
1071         }
1072
1073         while (resid > 0) {
1074                 if (srcidx >= srcrd->rd_nfrags) {
1075                         CERROR("Src buffer exhausted: %d frags\n", srcidx);
1076                         rc = -EPROTO;
1077                         break;
1078                 }
1079
1080                 if (dstidx == dstrd->rd_nfrags) {
1081                         CERROR("Dst buffer exhausted: %d frags\n", dstidx);
1082                         rc = -EPROTO;
1083                         break;
1084                 }
1085
1086                 if (tx->tx_nwrq >= IBLND_MAX_RDMA_FRAGS) {
1087                         CERROR("RDMA has too many fragments for peer %s (%d), src idx/frags: %d/%d dst idx/frags: %d/%d\n",
1088                                libcfs_nid2str(conn->ibc_peer->ibp_nid),
1089                                IBLND_MAX_RDMA_FRAGS,
1090                                srcidx, srcrd->rd_nfrags,
1091                                dstidx, dstrd->rd_nfrags);
1092                         rc = -EMSGSIZE;
1093                         break;
1094                 }
1095
1096                 wrknob = min3(kiblnd_rd_frag_size(srcrd, srcidx),
1097                               kiblnd_rd_frag_size(dstrd, dstidx),
1098                               (__u32)resid);
1099
1100                 sge = &tx->tx_sge[tx->tx_nwrq];
1101                 sge->addr   = kiblnd_rd_frag_addr(srcrd, srcidx);
1102                 sge->lkey   = kiblnd_rd_frag_key(srcrd, srcidx);
1103                 sge->length = wrknob;
1104
1105                 wrq = &tx->tx_wrq[tx->tx_nwrq];
1106                 next = wrq + 1;
1107
1108                 wrq->wr.next       = &next->wr;
1109                 wrq->wr.wr_id      = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA);
1110                 wrq->wr.sg_list    = sge;
1111                 wrq->wr.num_sge    = 1;
1112                 wrq->wr.opcode     = IB_WR_RDMA_WRITE;
1113                 wrq->wr.send_flags = 0;
1114
1115                 wrq->remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx);
1116                 wrq->rkey        = kiblnd_rd_frag_key(dstrd, dstidx);
1117
1118                 srcidx = kiblnd_rd_consume_frag(srcrd, srcidx, wrknob);
1119                 dstidx = kiblnd_rd_consume_frag(dstrd, dstidx, wrknob);
1120
1121                 resid -= wrknob;
1122
1123                 tx->tx_nwrq++;
1124                 wrq++;
1125                 sge++;
1126         }
1127 too_big:
1128         if (rc < 0)                          /* no RDMA if completing with failure */
1129                 tx->tx_nwrq = 0;
1130
1131         ibmsg->ibm_u.completion.ibcm_status = rc;
1132         ibmsg->ibm_u.completion.ibcm_cookie = dstcookie;
1133         kiblnd_init_tx_msg(conn->ibc_peer->ibp_ni, tx,
1134                            type, sizeof(struct kib_completion_msg));
1135
1136         return rc;
1137 }
1138
1139 static void
1140 kiblnd_queue_tx_locked(struct kib_tx *tx, struct kib_conn *conn)
1141 {
1142         struct list_head *q;
1143
1144         LASSERT(tx->tx_nwrq > 0);             /* work items set up */
1145         LASSERT(!tx->tx_queued);               /* not queued for sending already */
1146         LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1147
1148         tx->tx_queued = 1;
1149         tx->tx_deadline = jiffies +
1150                           msecs_to_jiffies(*kiblnd_tunables.kib_timeout *
1151                                            MSEC_PER_SEC);
1152
1153         if (!tx->tx_conn) {
1154                 kiblnd_conn_addref(conn);
1155                 tx->tx_conn = conn;
1156                 LASSERT(tx->tx_msg->ibm_type != IBLND_MSG_PUT_DONE);
1157         } else {
1158                 /* PUT_DONE first attached to conn as a PUT_REQ */
1159                 LASSERT(tx->tx_conn == conn);
1160                 LASSERT(tx->tx_msg->ibm_type == IBLND_MSG_PUT_DONE);
1161         }
1162
1163         switch (tx->tx_msg->ibm_type) {
1164         default:
1165                 LBUG();
1166
1167         case IBLND_MSG_PUT_REQ:
1168         case IBLND_MSG_GET_REQ:
1169                 q = &conn->ibc_tx_queue_rsrvd;
1170                 break;
1171
1172         case IBLND_MSG_PUT_NAK:
1173         case IBLND_MSG_PUT_ACK:
1174         case IBLND_MSG_PUT_DONE:
1175         case IBLND_MSG_GET_DONE:
1176                 q = &conn->ibc_tx_queue_nocred;
1177                 break;
1178
1179         case IBLND_MSG_NOOP:
1180                 if (IBLND_OOB_CAPABLE(conn->ibc_version))
1181                         q = &conn->ibc_tx_queue_nocred;
1182                 else
1183                         q = &conn->ibc_tx_noops;
1184                 break;
1185
1186         case IBLND_MSG_IMMEDIATE:
1187                 q = &conn->ibc_tx_queue;
1188                 break;
1189         }
1190
1191         list_add_tail(&tx->tx_list, q);
1192 }
1193
1194 static void
1195 kiblnd_queue_tx(struct kib_tx *tx, struct kib_conn *conn)
1196 {
1197         spin_lock(&conn->ibc_lock);
1198         kiblnd_queue_tx_locked(tx, conn);
1199         kiblnd_check_sends_locked(conn);
1200         spin_unlock(&conn->ibc_lock);
1201 }
1202
1203 static int kiblnd_resolve_addr(struct rdma_cm_id *cmid,
1204                                struct sockaddr_in *srcaddr,
1205                                struct sockaddr_in *dstaddr,
1206                                int timeout_ms)
1207 {
1208         unsigned short port;
1209         int rc;
1210
1211         /* allow the port to be reused */
1212         rc = rdma_set_reuseaddr(cmid, 1);
1213         if (rc) {
1214                 CERROR("Unable to set reuse on cmid: %d\n", rc);
1215                 return rc;
1216         }
1217
1218         /* look for a free privileged port */
1219         for (port = PROT_SOCK - 1; port > 0; port--) {
1220                 srcaddr->sin_port = htons(port);
1221                 rc = rdma_resolve_addr(cmid,
1222                                        (struct sockaddr *)srcaddr,
1223                                        (struct sockaddr *)dstaddr,
1224                                        timeout_ms);
1225                 if (!rc) {
1226                         CDEBUG(D_NET, "bound to port %hu\n", port);
1227                         return 0;
1228                 } else if (rc == -EADDRINUSE || rc == -EADDRNOTAVAIL) {
1229                         CDEBUG(D_NET, "bind to port %hu failed: %d\n",
1230                                port, rc);
1231                 } else {
1232                         return rc;
1233                 }
1234         }
1235
1236         CERROR("Failed to bind to a free privileged port\n");
1237         return rc;
1238 }
1239
1240 static void
1241 kiblnd_connect_peer(struct kib_peer *peer)
1242 {
1243         struct rdma_cm_id *cmid;
1244         struct kib_dev *dev;
1245         struct kib_net *net = peer->ibp_ni->ni_data;
1246         struct sockaddr_in srcaddr;
1247         struct sockaddr_in dstaddr;
1248         int rc;
1249
1250         LASSERT(net);
1251         LASSERT(peer->ibp_connecting > 0);
1252         LASSERT(!peer->ibp_reconnecting);
1253
1254         cmid = kiblnd_rdma_create_id(kiblnd_cm_callback, peer, RDMA_PS_TCP,
1255                                      IB_QPT_RC);
1256
1257         if (IS_ERR(cmid)) {
1258                 CERROR("Can't create CMID for %s: %ld\n",
1259                        libcfs_nid2str(peer->ibp_nid), PTR_ERR(cmid));
1260                 rc = PTR_ERR(cmid);
1261                 goto failed;
1262         }
1263
1264         dev = net->ibn_dev;
1265         memset(&srcaddr, 0, sizeof(srcaddr));
1266         srcaddr.sin_family = AF_INET;
1267         srcaddr.sin_addr.s_addr = htonl(dev->ibd_ifip);
1268
1269         memset(&dstaddr, 0, sizeof(dstaddr));
1270         dstaddr.sin_family = AF_INET;
1271         dstaddr.sin_port = htons(*kiblnd_tunables.kib_service);
1272         dstaddr.sin_addr.s_addr = htonl(LNET_NIDADDR(peer->ibp_nid));
1273
1274         kiblnd_peer_addref(peer);              /* cmid's ref */
1275
1276         if (*kiblnd_tunables.kib_use_priv_port) {
1277                 rc = kiblnd_resolve_addr(cmid, &srcaddr, &dstaddr,
1278                                          *kiblnd_tunables.kib_timeout * 1000);
1279         } else {
1280                 rc = rdma_resolve_addr(cmid,
1281                                        (struct sockaddr *)&srcaddr,
1282                                        (struct sockaddr *)&dstaddr,
1283                                        *kiblnd_tunables.kib_timeout * 1000);
1284         }
1285         if (rc) {
1286                 /* Can't initiate address resolution:  */
1287                 CERROR("Can't resolve addr for %s: %d\n",
1288                        libcfs_nid2str(peer->ibp_nid), rc);
1289                 goto failed2;
1290         }
1291
1292         LASSERT(cmid->device);
1293         CDEBUG(D_NET, "%s: connection bound to %s:%pI4h:%s\n",
1294                libcfs_nid2str(peer->ibp_nid), dev->ibd_ifname,
1295                &dev->ibd_ifip, cmid->device->name);
1296
1297         return;
1298
1299  failed2:
1300         kiblnd_peer_connect_failed(peer, 1, rc);
1301         kiblnd_peer_decref(peer);              /* cmid's ref */
1302         rdma_destroy_id(cmid);
1303         return;
1304  failed:
1305         kiblnd_peer_connect_failed(peer, 1, rc);
1306 }
1307
1308 bool
1309 kiblnd_reconnect_peer(struct kib_peer *peer)
1310 {
1311         rwlock_t *glock = &kiblnd_data.kib_global_lock;
1312         char *reason = NULL;
1313         struct list_head txs;
1314         unsigned long flags;
1315
1316         INIT_LIST_HEAD(&txs);
1317
1318         write_lock_irqsave(glock, flags);
1319         if (!peer->ibp_reconnecting) {
1320                 if (peer->ibp_accepting)
1321                         reason = "accepting";
1322                 else if (peer->ibp_connecting)
1323                         reason = "connecting";
1324                 else if (!list_empty(&peer->ibp_conns))
1325                         reason = "connected";
1326                 else /* connected then closed */
1327                         reason = "closed";
1328
1329                 goto no_reconnect;
1330         }
1331
1332         LASSERT(!peer->ibp_accepting && !peer->ibp_connecting &&
1333                 list_empty(&peer->ibp_conns));
1334         peer->ibp_reconnecting = 0;
1335
1336         if (!kiblnd_peer_active(peer)) {
1337                 list_splice_init(&peer->ibp_tx_queue, &txs);
1338                 reason = "unlinked";
1339                 goto no_reconnect;
1340         }
1341
1342         peer->ibp_connecting++;
1343         peer->ibp_reconnected++;
1344         write_unlock_irqrestore(glock, flags);
1345
1346         kiblnd_connect_peer(peer);
1347         return true;
1348
1349 no_reconnect:
1350         write_unlock_irqrestore(glock, flags);
1351
1352         CWARN("Abort reconnection of %s: %s\n",
1353               libcfs_nid2str(peer->ibp_nid), reason);
1354         kiblnd_txlist_done(peer->ibp_ni, &txs, -ECONNABORTED);
1355         return false;
1356 }
1357
1358 void
1359 kiblnd_launch_tx(struct lnet_ni *ni, struct kib_tx *tx, lnet_nid_t nid)
1360 {
1361         struct kib_peer *peer;
1362         struct kib_peer *peer2;
1363         struct kib_conn *conn;
1364         rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
1365         unsigned long flags;
1366         int rc;
1367
1368         /*
1369          * If I get here, I've committed to send, so I complete the tx with
1370          * failure on any problems
1371          */
1372         LASSERT(!tx || !tx->tx_conn); /* only set when assigned a conn */
1373         LASSERT(!tx || tx->tx_nwrq > 0);     /* work items have been set up */
1374
1375         /*
1376          * First time, just use a read lock since I expect to find my peer
1377          * connected
1378          */
1379         read_lock_irqsave(g_lock, flags);
1380
1381         peer = kiblnd_find_peer_locked(nid);
1382         if (peer && !list_empty(&peer->ibp_conns)) {
1383                 /* Found a peer with an established connection */
1384                 conn = kiblnd_get_conn_locked(peer);
1385                 kiblnd_conn_addref(conn); /* 1 ref for me... */
1386
1387                 read_unlock_irqrestore(g_lock, flags);
1388
1389                 if (tx)
1390                         kiblnd_queue_tx(tx, conn);
1391                 kiblnd_conn_decref(conn); /* ...to here */
1392                 return;
1393         }
1394
1395         read_unlock(g_lock);
1396         /* Re-try with a write lock */
1397         write_lock(g_lock);
1398
1399         peer = kiblnd_find_peer_locked(nid);
1400         if (peer) {
1401                 if (list_empty(&peer->ibp_conns)) {
1402                         /* found a peer, but it's still connecting... */
1403                         LASSERT(kiblnd_peer_connecting(peer));
1404                         if (tx)
1405                                 list_add_tail(&tx->tx_list,
1406                                               &peer->ibp_tx_queue);
1407                         write_unlock_irqrestore(g_lock, flags);
1408                 } else {
1409                         conn = kiblnd_get_conn_locked(peer);
1410                         kiblnd_conn_addref(conn); /* 1 ref for me... */
1411
1412                         write_unlock_irqrestore(g_lock, flags);
1413
1414                         if (tx)
1415                                 kiblnd_queue_tx(tx, conn);
1416                         kiblnd_conn_decref(conn); /* ...to here */
1417                 }
1418                 return;
1419         }
1420
1421         write_unlock_irqrestore(g_lock, flags);
1422
1423         /* Allocate a peer ready to add to the peer table and retry */
1424         rc = kiblnd_create_peer(ni, &peer, nid);
1425         if (rc) {
1426                 CERROR("Can't create peer %s\n", libcfs_nid2str(nid));
1427                 if (tx) {
1428                         tx->tx_status = -EHOSTUNREACH;
1429                         tx->tx_waiting = 0;
1430                         kiblnd_tx_done(ni, tx);
1431                 }
1432                 return;
1433         }
1434
1435         write_lock_irqsave(g_lock, flags);
1436
1437         peer2 = kiblnd_find_peer_locked(nid);
1438         if (peer2) {
1439                 if (list_empty(&peer2->ibp_conns)) {
1440                         /* found a peer, but it's still connecting... */
1441                         LASSERT(kiblnd_peer_connecting(peer2));
1442                         if (tx)
1443                                 list_add_tail(&tx->tx_list,
1444                                               &peer2->ibp_tx_queue);
1445                         write_unlock_irqrestore(g_lock, flags);
1446                 } else {
1447                         conn = kiblnd_get_conn_locked(peer2);
1448                         kiblnd_conn_addref(conn); /* 1 ref for me... */
1449
1450                         write_unlock_irqrestore(g_lock, flags);
1451
1452                         if (tx)
1453                                 kiblnd_queue_tx(tx, conn);
1454                         kiblnd_conn_decref(conn); /* ...to here */
1455                 }
1456
1457                 kiblnd_peer_decref(peer);
1458                 return;
1459         }
1460
1461         /* Brand new peer */
1462         LASSERT(!peer->ibp_connecting);
1463         peer->ibp_connecting = 1;
1464
1465         /* always called with a ref on ni, which prevents ni being shutdown */
1466         LASSERT(!((struct kib_net *)ni->ni_data)->ibn_shutdown);
1467
1468         if (tx)
1469                 list_add_tail(&tx->tx_list, &peer->ibp_tx_queue);
1470
1471         kiblnd_peer_addref(peer);
1472         list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
1473
1474         write_unlock_irqrestore(g_lock, flags);
1475
1476         kiblnd_connect_peer(peer);
1477         kiblnd_peer_decref(peer);
1478 }
1479
1480 int
1481 kiblnd_send(struct lnet_ni *ni, void *private, struct lnet_msg *lntmsg)
1482 {
1483         struct lnet_hdr *hdr = &lntmsg->msg_hdr;
1484         int type = lntmsg->msg_type;
1485         struct lnet_process_id target = lntmsg->msg_target;
1486         int target_is_router = lntmsg->msg_target_is_router;
1487         int routing = lntmsg->msg_routing;
1488         unsigned int payload_niov = lntmsg->msg_niov;
1489         struct kvec *payload_iov = lntmsg->msg_iov;
1490         struct bio_vec *payload_kiov = lntmsg->msg_kiov;
1491         unsigned int payload_offset = lntmsg->msg_offset;
1492         unsigned int payload_nob = lntmsg->msg_len;
1493         struct iov_iter from;
1494         struct kib_msg *ibmsg;
1495         struct kib_rdma_desc  *rd;
1496         struct kib_tx *tx;
1497         int nob;
1498         int rc;
1499
1500         /* NB 'private' is different depending on what we're sending.... */
1501
1502         CDEBUG(D_NET, "sending %d bytes in %d frags to %s\n",
1503                payload_nob, payload_niov, libcfs_id2str(target));
1504
1505         LASSERT(!payload_nob || payload_niov > 0);
1506         LASSERT(payload_niov <= LNET_MAX_IOV);
1507
1508         /* Thread context */
1509         LASSERT(!in_interrupt());
1510         /* payload is either all vaddrs or all pages */
1511         LASSERT(!(payload_kiov && payload_iov));
1512
1513         if (payload_kiov)
1514                 iov_iter_bvec(&from, ITER_BVEC | WRITE,
1515                               payload_kiov, payload_niov,
1516                               payload_nob + payload_offset);
1517         else
1518                 iov_iter_kvec(&from, ITER_KVEC | WRITE,
1519                               payload_iov, payload_niov,
1520                               payload_nob + payload_offset);
1521
1522         iov_iter_advance(&from, payload_offset);
1523
1524         switch (type) {
1525         default:
1526                 LBUG();
1527                 return -EIO;
1528
1529         case LNET_MSG_ACK:
1530                 LASSERT(!payload_nob);
1531                 break;
1532
1533         case LNET_MSG_GET:
1534                 if (routing || target_is_router)
1535                         break;            /* send IMMEDIATE */
1536
1537                 /* is the REPLY message too small for RDMA? */
1538                 nob = offsetof(struct kib_msg, ibm_u.immediate.ibim_payload[lntmsg->msg_md->md_length]);
1539                 if (nob <= IBLND_MSG_SIZE)
1540                         break;            /* send IMMEDIATE */
1541
1542                 tx = kiblnd_get_idle_tx(ni, target.nid);
1543                 if (!tx) {
1544                         CERROR("Can't allocate txd for GET to %s\n",
1545                                libcfs_nid2str(target.nid));
1546                         return -ENOMEM;
1547                 }
1548
1549                 ibmsg = tx->tx_msg;
1550                 rd = &ibmsg->ibm_u.get.ibgm_rd;
1551                 if (!(lntmsg->msg_md->md_options & LNET_MD_KIOV))
1552                         rc = kiblnd_setup_rd_iov(ni, tx, rd,
1553                                                  lntmsg->msg_md->md_niov,
1554                                                  lntmsg->msg_md->md_iov.iov,
1555                                                  0, lntmsg->msg_md->md_length);
1556                 else
1557                         rc = kiblnd_setup_rd_kiov(ni, tx, rd,
1558                                                   lntmsg->msg_md->md_niov,
1559                                                   lntmsg->msg_md->md_iov.kiov,
1560                                                   0, lntmsg->msg_md->md_length);
1561                 if (rc) {
1562                         CERROR("Can't setup GET sink for %s: %d\n",
1563                                libcfs_nid2str(target.nid), rc);
1564                         kiblnd_tx_done(ni, tx);
1565                         return -EIO;
1566                 }
1567
1568                 nob = offsetof(struct kib_get_msg, ibgm_rd.rd_frags[rd->rd_nfrags]);
1569                 ibmsg->ibm_u.get.ibgm_cookie = tx->tx_cookie;
1570                 ibmsg->ibm_u.get.ibgm_hdr = *hdr;
1571
1572                 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_GET_REQ, nob);
1573
1574                 tx->tx_lntmsg[1] = lnet_create_reply_msg(ni, lntmsg);
1575                 if (!tx->tx_lntmsg[1]) {
1576                         CERROR("Can't create reply for GET -> %s\n",
1577                                libcfs_nid2str(target.nid));
1578                         kiblnd_tx_done(ni, tx);
1579                         return -EIO;
1580                 }
1581
1582                 tx->tx_lntmsg[0] = lntmsg;      /* finalise lntmsg[0,1] on completion */
1583                 tx->tx_waiting = 1;          /* waiting for GET_DONE */
1584                 kiblnd_launch_tx(ni, tx, target.nid);
1585                 return 0;
1586
1587         case LNET_MSG_REPLY:
1588         case LNET_MSG_PUT:
1589                 /* Is the payload small enough not to need RDMA? */
1590                 nob = offsetof(struct kib_msg, ibm_u.immediate.ibim_payload[payload_nob]);
1591                 if (nob <= IBLND_MSG_SIZE)
1592                         break;            /* send IMMEDIATE */
1593
1594                 tx = kiblnd_get_idle_tx(ni, target.nid);
1595                 if (!tx) {
1596                         CERROR("Can't allocate %s txd for %s\n",
1597                                type == LNET_MSG_PUT ? "PUT" : "REPLY",
1598                                libcfs_nid2str(target.nid));
1599                         return -ENOMEM;
1600                 }
1601
1602                 if (!payload_kiov)
1603                         rc = kiblnd_setup_rd_iov(ni, tx, tx->tx_rd,
1604                                                  payload_niov, payload_iov,
1605                                                  payload_offset, payload_nob);
1606                 else
1607                         rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1608                                                   payload_niov, payload_kiov,
1609                                                   payload_offset, payload_nob);
1610                 if (rc) {
1611                         CERROR("Can't setup PUT src for %s: %d\n",
1612                                libcfs_nid2str(target.nid), rc);
1613                         kiblnd_tx_done(ni, tx);
1614                         return -EIO;
1615                 }
1616
1617                 ibmsg = tx->tx_msg;
1618                 ibmsg->ibm_u.putreq.ibprm_hdr = *hdr;
1619                 ibmsg->ibm_u.putreq.ibprm_cookie = tx->tx_cookie;
1620                 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_PUT_REQ, sizeof(struct kib_putreq_msg));
1621
1622                 tx->tx_lntmsg[0] = lntmsg;      /* finalise lntmsg on completion */
1623                 tx->tx_waiting = 1;          /* waiting for PUT_{ACK,NAK} */
1624                 kiblnd_launch_tx(ni, tx, target.nid);
1625                 return 0;
1626         }
1627
1628         /* send IMMEDIATE */
1629
1630         LASSERT(offsetof(struct kib_msg, ibm_u.immediate.ibim_payload[payload_nob])
1631                  <= IBLND_MSG_SIZE);
1632
1633         tx = kiblnd_get_idle_tx(ni, target.nid);
1634         if (!tx) {
1635                 CERROR("Can't send %d to %s: tx descs exhausted\n",
1636                        type, libcfs_nid2str(target.nid));
1637                 return -ENOMEM;
1638         }
1639
1640         ibmsg = tx->tx_msg;
1641         ibmsg->ibm_u.immediate.ibim_hdr = *hdr;
1642
1643         rc = copy_from_iter(&ibmsg->ibm_u.immediate.ibim_payload, payload_nob,
1644                             &from);
1645         if (rc != payload_nob) {
1646                 kiblnd_pool_free_node(&tx->tx_pool->tpo_pool, &tx->tx_list);
1647                 return -EFAULT;
1648         }
1649
1650         nob = offsetof(struct kib_immediate_msg, ibim_payload[payload_nob]);
1651         kiblnd_init_tx_msg(ni, tx, IBLND_MSG_IMMEDIATE, nob);
1652
1653         tx->tx_lntmsg[0] = lntmsg;            /* finalise lntmsg on completion */
1654         kiblnd_launch_tx(ni, tx, target.nid);
1655         return 0;
1656 }
1657
1658 static void
1659 kiblnd_reply(struct lnet_ni *ni, struct kib_rx *rx, struct lnet_msg *lntmsg)
1660 {
1661         struct lnet_process_id target = lntmsg->msg_target;
1662         unsigned int niov = lntmsg->msg_niov;
1663         struct kvec *iov = lntmsg->msg_iov;
1664         struct bio_vec *kiov = lntmsg->msg_kiov;
1665         unsigned int offset = lntmsg->msg_offset;
1666         unsigned int nob = lntmsg->msg_len;
1667         struct kib_tx *tx;
1668         int rc;
1669
1670         tx = kiblnd_get_idle_tx(ni, rx->rx_conn->ibc_peer->ibp_nid);
1671         if (!tx) {
1672                 CERROR("Can't get tx for REPLY to %s\n",
1673                        libcfs_nid2str(target.nid));
1674                 goto failed_0;
1675         }
1676
1677         if (!nob)
1678                 rc = 0;
1679         else if (!kiov)
1680                 rc = kiblnd_setup_rd_iov(ni, tx, tx->tx_rd,
1681                                          niov, iov, offset, nob);
1682         else
1683                 rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1684                                           niov, kiov, offset, nob);
1685
1686         if (rc) {
1687                 CERROR("Can't setup GET src for %s: %d\n",
1688                        libcfs_nid2str(target.nid), rc);
1689                 goto failed_1;
1690         }
1691
1692         rc = kiblnd_init_rdma(rx->rx_conn, tx,
1693                               IBLND_MSG_GET_DONE, nob,
1694                               &rx->rx_msg->ibm_u.get.ibgm_rd,
1695                               rx->rx_msg->ibm_u.get.ibgm_cookie);
1696         if (rc < 0) {
1697                 CERROR("Can't setup rdma for GET from %s: %d\n",
1698                        libcfs_nid2str(target.nid), rc);
1699                 goto failed_1;
1700         }
1701
1702         if (!nob) {
1703                 /* No RDMA: local completion may happen now! */
1704                 lnet_finalize(ni, lntmsg, 0);
1705         } else {
1706                 /* RDMA: lnet_finalize(lntmsg) when it completes */
1707                 tx->tx_lntmsg[0] = lntmsg;
1708         }
1709
1710         kiblnd_queue_tx(tx, rx->rx_conn);
1711         return;
1712
1713  failed_1:
1714         kiblnd_tx_done(ni, tx);
1715  failed_0:
1716         lnet_finalize(ni, lntmsg, -EIO);
1717 }
1718
1719 int
1720 kiblnd_recv(struct lnet_ni *ni, void *private, struct lnet_msg *lntmsg,
1721             int delayed, struct iov_iter *to, unsigned int rlen)
1722 {
1723         struct kib_rx *rx = private;
1724         struct kib_msg *rxmsg = rx->rx_msg;
1725         struct kib_conn *conn = rx->rx_conn;
1726         struct kib_tx *tx;
1727         int nob;
1728         int post_credit = IBLND_POSTRX_PEER_CREDIT;
1729         int rc = 0;
1730
1731         LASSERT(iov_iter_count(to) <= rlen);
1732         LASSERT(!in_interrupt());
1733         /* Either all pages or all vaddrs */
1734
1735         switch (rxmsg->ibm_type) {
1736         default:
1737                 LBUG();
1738
1739         case IBLND_MSG_IMMEDIATE:
1740                 nob = offsetof(struct kib_msg, ibm_u.immediate.ibim_payload[rlen]);
1741                 if (nob > rx->rx_nob) {
1742                         CERROR("Immediate message from %s too big: %d(%d)\n",
1743                                libcfs_nid2str(rxmsg->ibm_u.immediate.ibim_hdr.src_nid),
1744                                nob, rx->rx_nob);
1745                         rc = -EPROTO;
1746                         break;
1747                 }
1748
1749                 rc = copy_to_iter(&rxmsg->ibm_u.immediate.ibim_payload, rlen,
1750                                   to);
1751                 if (rc != rlen) {
1752                         rc = -EFAULT;
1753                         break;
1754                 }
1755
1756                 rc = 0;
1757                 lnet_finalize(ni, lntmsg, 0);
1758                 break;
1759
1760         case IBLND_MSG_PUT_REQ: {
1761                 struct kib_msg  *txmsg;
1762                 struct kib_rdma_desc *rd;
1763
1764                 if (!iov_iter_count(to)) {
1765                         lnet_finalize(ni, lntmsg, 0);
1766                         kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, 0,
1767                                                rxmsg->ibm_u.putreq.ibprm_cookie);
1768                         break;
1769                 }
1770
1771                 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
1772                 if (!tx) {
1773                         CERROR("Can't allocate tx for %s\n",
1774                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
1775                         /* Not replying will break the connection */
1776                         rc = -ENOMEM;
1777                         break;
1778                 }
1779
1780                 txmsg = tx->tx_msg;
1781                 rd = &txmsg->ibm_u.putack.ibpam_rd;
1782                 if (!(to->type & ITER_BVEC))
1783                         rc = kiblnd_setup_rd_iov(ni, tx, rd,
1784                                                  to->nr_segs, to->kvec,
1785                                                  to->iov_offset,
1786                                                  iov_iter_count(to));
1787                 else
1788                         rc = kiblnd_setup_rd_kiov(ni, tx, rd,
1789                                                   to->nr_segs, to->bvec,
1790                                                   to->iov_offset,
1791                                                   iov_iter_count(to));
1792                 if (rc) {
1793                         CERROR("Can't setup PUT sink for %s: %d\n",
1794                                libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
1795                         kiblnd_tx_done(ni, tx);
1796                         /* tell peer it's over */
1797                         kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, rc,
1798                                                rxmsg->ibm_u.putreq.ibprm_cookie);
1799                         break;
1800                 }
1801
1802                 nob = offsetof(struct kib_putack_msg, ibpam_rd.rd_frags[rd->rd_nfrags]);
1803                 txmsg->ibm_u.putack.ibpam_src_cookie = rxmsg->ibm_u.putreq.ibprm_cookie;
1804                 txmsg->ibm_u.putack.ibpam_dst_cookie = tx->tx_cookie;
1805
1806                 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_PUT_ACK, nob);
1807
1808                 tx->tx_lntmsg[0] = lntmsg;      /* finalise lntmsg on completion */
1809                 tx->tx_waiting = 1;          /* waiting for PUT_DONE */
1810                 kiblnd_queue_tx(tx, conn);
1811
1812                 /* reposted buffer reserved for PUT_DONE */
1813                 post_credit = IBLND_POSTRX_NO_CREDIT;
1814                 break;
1815                 }
1816
1817         case IBLND_MSG_GET_REQ:
1818                 if (lntmsg) {
1819                         /* Optimized GET; RDMA lntmsg's payload */
1820                         kiblnd_reply(ni, rx, lntmsg);
1821                 } else {
1822                         /* GET didn't match anything */
1823                         kiblnd_send_completion(rx->rx_conn, IBLND_MSG_GET_DONE,
1824                                                -ENODATA,
1825                                                rxmsg->ibm_u.get.ibgm_cookie);
1826                 }
1827                 break;
1828         }
1829
1830         kiblnd_post_rx(rx, post_credit);
1831         return rc;
1832 }
1833
1834 int
1835 kiblnd_thread_start(int (*fn)(void *arg), void *arg, char *name)
1836 {
1837         struct task_struct *task = kthread_run(fn, arg, "%s", name);
1838
1839         if (IS_ERR(task))
1840                 return PTR_ERR(task);
1841
1842         atomic_inc(&kiblnd_data.kib_nthreads);
1843         return 0;
1844 }
1845
1846 static void
1847 kiblnd_thread_fini(void)
1848 {
1849         atomic_dec(&kiblnd_data.kib_nthreads);
1850 }
1851
1852 static void
1853 kiblnd_peer_alive(struct kib_peer *peer)
1854 {
1855         /* This is racy, but everyone's only writing cfs_time_current() */
1856         peer->ibp_last_alive = cfs_time_current();
1857         mb();
1858 }
1859
1860 static void
1861 kiblnd_peer_notify(struct kib_peer *peer)
1862 {
1863         int error = 0;
1864         unsigned long last_alive = 0;
1865         unsigned long flags;
1866
1867         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1868
1869         if (kiblnd_peer_idle(peer) && peer->ibp_error) {
1870                 error = peer->ibp_error;
1871                 peer->ibp_error = 0;
1872
1873                 last_alive = peer->ibp_last_alive;
1874         }
1875
1876         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1877
1878         if (error)
1879                 lnet_notify(peer->ibp_ni,
1880                             peer->ibp_nid, 0, last_alive);
1881 }
1882
1883 void
1884 kiblnd_close_conn_locked(struct kib_conn *conn, int error)
1885 {
1886         /*
1887          * This just does the immediate housekeeping. 'error' is zero for a
1888          * normal shutdown which can happen only after the connection has been
1889          * established.  If the connection is established, schedule the
1890          * connection to be finished off by the connd. Otherwise the connd is
1891          * already dealing with it (either to set it up or tear it down).
1892          * Caller holds kib_global_lock exclusively in irq context
1893          */
1894         struct kib_peer *peer = conn->ibc_peer;
1895         struct kib_dev *dev;
1896         unsigned long flags;
1897
1898         LASSERT(error || conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1899
1900         if (error && !conn->ibc_comms_error)
1901                 conn->ibc_comms_error = error;
1902
1903         if (conn->ibc_state != IBLND_CONN_ESTABLISHED)
1904                 return; /* already being handled  */
1905
1906         if (!error &&
1907             list_empty(&conn->ibc_tx_noops) &&
1908             list_empty(&conn->ibc_tx_queue) &&
1909             list_empty(&conn->ibc_tx_queue_rsrvd) &&
1910             list_empty(&conn->ibc_tx_queue_nocred) &&
1911             list_empty(&conn->ibc_active_txs)) {
1912                 CDEBUG(D_NET, "closing conn to %s\n",
1913                        libcfs_nid2str(peer->ibp_nid));
1914         } else {
1915                 CNETERR("Closing conn to %s: error %d%s%s%s%s%s\n",
1916                         libcfs_nid2str(peer->ibp_nid), error,
1917                         list_empty(&conn->ibc_tx_queue) ? "" : "(sending)",
1918                         list_empty(&conn->ibc_tx_noops) ? "" : "(sending_noops)",
1919                         list_empty(&conn->ibc_tx_queue_rsrvd) ? "" : "(sending_rsrvd)",
1920                         list_empty(&conn->ibc_tx_queue_nocred) ? "" : "(sending_nocred)",
1921                         list_empty(&conn->ibc_active_txs) ? "" : "(waiting)");
1922         }
1923
1924         dev = ((struct kib_net *)peer->ibp_ni->ni_data)->ibn_dev;
1925         list_del(&conn->ibc_list);
1926         /* connd (see below) takes over ibc_list's ref */
1927
1928         if (list_empty(&peer->ibp_conns) &&    /* no more conns */
1929             kiblnd_peer_active(peer)) {  /* still in peer table */
1930                 kiblnd_unlink_peer_locked(peer);
1931
1932                 /* set/clear error on last conn */
1933                 peer->ibp_error = conn->ibc_comms_error;
1934         }
1935
1936         kiblnd_set_conn_state(conn, IBLND_CONN_CLOSING);
1937
1938         if (error &&
1939             kiblnd_dev_can_failover(dev)) {
1940                 list_add_tail(&dev->ibd_fail_list,
1941                               &kiblnd_data.kib_failed_devs);
1942                 wake_up(&kiblnd_data.kib_failover_waitq);
1943         }
1944
1945         spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
1946
1947         list_add_tail(&conn->ibc_list, &kiblnd_data.kib_connd_conns);
1948         wake_up(&kiblnd_data.kib_connd_waitq);
1949
1950         spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
1951 }
1952
1953 void
1954 kiblnd_close_conn(struct kib_conn *conn, int error)
1955 {
1956         unsigned long flags;
1957
1958         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1959
1960         kiblnd_close_conn_locked(conn, error);
1961
1962         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1963 }
1964
1965 static void
1966 kiblnd_handle_early_rxs(struct kib_conn *conn)
1967 {
1968         unsigned long flags;
1969         struct kib_rx *rx;
1970         struct kib_rx *tmp;
1971
1972         LASSERT(!in_interrupt());
1973         LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1974
1975         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1976         list_for_each_entry_safe(rx, tmp, &conn->ibc_early_rxs, rx_list) {
1977                 list_del(&rx->rx_list);
1978                 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1979
1980                 kiblnd_handle_rx(rx);
1981
1982                 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1983         }
1984         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1985 }
1986
1987 static void
1988 kiblnd_abort_txs(struct kib_conn *conn, struct list_head *txs)
1989 {
1990         LIST_HEAD(zombies);
1991         struct list_head *tmp;
1992         struct list_head *nxt;
1993         struct kib_tx *tx;
1994
1995         spin_lock(&conn->ibc_lock);
1996
1997         list_for_each_safe(tmp, nxt, txs) {
1998                 tx = list_entry(tmp, struct kib_tx, tx_list);
1999
2000                 if (txs == &conn->ibc_active_txs) {
2001                         LASSERT(!tx->tx_queued);
2002                         LASSERT(tx->tx_waiting || tx->tx_sending);
2003                 } else {
2004                         LASSERT(tx->tx_queued);
2005                 }
2006
2007                 tx->tx_status = -ECONNABORTED;
2008                 tx->tx_waiting = 0;
2009
2010                 if (!tx->tx_sending) {
2011                         tx->tx_queued = 0;
2012                         list_del(&tx->tx_list);
2013                         list_add(&tx->tx_list, &zombies);
2014                 }
2015         }
2016
2017         spin_unlock(&conn->ibc_lock);
2018
2019         kiblnd_txlist_done(conn->ibc_peer->ibp_ni, &zombies, -ECONNABORTED);
2020 }
2021
2022 static void
2023 kiblnd_finalise_conn(struct kib_conn *conn)
2024 {
2025         LASSERT(!in_interrupt());
2026         LASSERT(conn->ibc_state > IBLND_CONN_INIT);
2027
2028         kiblnd_set_conn_state(conn, IBLND_CONN_DISCONNECTED);
2029
2030         /*
2031          * abort_receives moves QP state to IB_QPS_ERR.  This is only required
2032          * for connections that didn't get as far as being connected, because
2033          * rdma_disconnect() does this for free.
2034          */
2035         kiblnd_abort_receives(conn);
2036
2037         /*
2038          * Complete all tx descs not waiting for sends to complete.
2039          * NB we should be safe from RDMA now that the QP has changed state
2040          */
2041         kiblnd_abort_txs(conn, &conn->ibc_tx_noops);
2042         kiblnd_abort_txs(conn, &conn->ibc_tx_queue);
2043         kiblnd_abort_txs(conn, &conn->ibc_tx_queue_rsrvd);
2044         kiblnd_abort_txs(conn, &conn->ibc_tx_queue_nocred);
2045         kiblnd_abort_txs(conn, &conn->ibc_active_txs);
2046
2047         kiblnd_handle_early_rxs(conn);
2048 }
2049
2050 static void
2051 kiblnd_peer_connect_failed(struct kib_peer *peer, int active, int error)
2052 {
2053         LIST_HEAD(zombies);
2054         unsigned long flags;
2055
2056         LASSERT(error);
2057         LASSERT(!in_interrupt());
2058
2059         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2060
2061         if (active) {
2062                 LASSERT(peer->ibp_connecting > 0);
2063                 peer->ibp_connecting--;
2064         } else {
2065                 LASSERT(peer->ibp_accepting > 0);
2066                 peer->ibp_accepting--;
2067         }
2068
2069         if (kiblnd_peer_connecting(peer)) {
2070                 /* another connection attempt under way... */
2071                 write_unlock_irqrestore(&kiblnd_data.kib_global_lock,
2072                                         flags);
2073                 return;
2074         }
2075
2076         peer->ibp_reconnected = 0;
2077         if (list_empty(&peer->ibp_conns)) {
2078                 /* Take peer's blocked transmits to complete with error */
2079                 list_add(&zombies, &peer->ibp_tx_queue);
2080                 list_del_init(&peer->ibp_tx_queue);
2081
2082                 if (kiblnd_peer_active(peer))
2083                         kiblnd_unlink_peer_locked(peer);
2084
2085                 peer->ibp_error = error;
2086         } else {
2087                 /* Can't have blocked transmits if there are connections */
2088                 LASSERT(list_empty(&peer->ibp_tx_queue));
2089         }
2090
2091         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2092
2093         kiblnd_peer_notify(peer);
2094
2095         if (list_empty(&zombies))
2096                 return;
2097
2098         CNETERR("Deleting messages for %s: connection failed\n",
2099                 libcfs_nid2str(peer->ibp_nid));
2100
2101         kiblnd_txlist_done(peer->ibp_ni, &zombies, -EHOSTUNREACH);
2102 }
2103
2104 static void
2105 kiblnd_connreq_done(struct kib_conn *conn, int status)
2106 {
2107         struct kib_peer *peer = conn->ibc_peer;
2108         struct kib_tx *tx;
2109         struct kib_tx *tmp;
2110         struct list_head txs;
2111         unsigned long flags;
2112         int active;
2113
2114         active = (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2115
2116         CDEBUG(D_NET, "%s: active(%d), version(%x), status(%d)\n",
2117                libcfs_nid2str(peer->ibp_nid), active,
2118                conn->ibc_version, status);
2119
2120         LASSERT(!in_interrupt());
2121         LASSERT((conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT &&
2122                  peer->ibp_connecting > 0) ||
2123                  (conn->ibc_state == IBLND_CONN_PASSIVE_WAIT &&
2124                  peer->ibp_accepting > 0));
2125
2126         LIBCFS_FREE(conn->ibc_connvars, sizeof(*conn->ibc_connvars));
2127         conn->ibc_connvars = NULL;
2128
2129         if (status) {
2130                 /* failed to establish connection */
2131                 kiblnd_peer_connect_failed(peer, active, status);
2132                 kiblnd_finalise_conn(conn);
2133                 return;
2134         }
2135
2136         /* connection established */
2137         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2138
2139         conn->ibc_last_send = jiffies;
2140         kiblnd_set_conn_state(conn, IBLND_CONN_ESTABLISHED);
2141         kiblnd_peer_alive(peer);
2142
2143         /*
2144          * Add conn to peer's list and nuke any dangling conns from a different
2145          * peer instance...
2146          */
2147         kiblnd_conn_addref(conn);              /* +1 ref for ibc_list */
2148         list_add(&conn->ibc_list, &peer->ibp_conns);
2149         peer->ibp_reconnected = 0;
2150         if (active)
2151                 peer->ibp_connecting--;
2152         else
2153                 peer->ibp_accepting--;
2154
2155         if (!peer->ibp_version) {
2156                 peer->ibp_version     = conn->ibc_version;
2157                 peer->ibp_incarnation = conn->ibc_incarnation;
2158         }
2159
2160         if (peer->ibp_version     != conn->ibc_version ||
2161             peer->ibp_incarnation != conn->ibc_incarnation) {
2162                 kiblnd_close_stale_conns_locked(peer, conn->ibc_version,
2163                                                 conn->ibc_incarnation);
2164                 peer->ibp_version     = conn->ibc_version;
2165                 peer->ibp_incarnation = conn->ibc_incarnation;
2166         }
2167
2168         /* grab pending txs while I have the lock */
2169         list_add(&txs, &peer->ibp_tx_queue);
2170         list_del_init(&peer->ibp_tx_queue);
2171
2172         if (!kiblnd_peer_active(peer) ||        /* peer has been deleted */
2173             conn->ibc_comms_error) {       /* error has happened already */
2174                 struct lnet_ni *ni = peer->ibp_ni;
2175
2176                 /* start to shut down connection */
2177                 kiblnd_close_conn_locked(conn, -ECONNABORTED);
2178                 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2179
2180                 kiblnd_txlist_done(ni, &txs, -ECONNABORTED);
2181
2182                 return;
2183         }
2184
2185         /*
2186          * +1 ref for myself, this connection is visible to other threads
2187          * now, refcount of peer:ibp_conns can be released by connection
2188          * close from either a different thread, or the calling of
2189          * kiblnd_check_sends_locked() below. See bz21911 for details.
2190          */
2191         kiblnd_conn_addref(conn);
2192         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2193
2194         /* Schedule blocked txs */
2195         spin_lock(&conn->ibc_lock);
2196         list_for_each_entry_safe(tx, tmp, &txs, tx_list) {
2197                 list_del(&tx->tx_list);
2198
2199                 kiblnd_queue_tx_locked(tx, conn);
2200         }
2201         kiblnd_check_sends_locked(conn);
2202         spin_unlock(&conn->ibc_lock);
2203
2204         /* schedule blocked rxs */
2205         kiblnd_handle_early_rxs(conn);
2206
2207         kiblnd_conn_decref(conn);
2208 }
2209
2210 static void
2211 kiblnd_reject(struct rdma_cm_id *cmid, struct kib_rej *rej)
2212 {
2213         int rc;
2214
2215         rc = rdma_reject(cmid, rej, sizeof(*rej));
2216
2217         if (rc)
2218                 CWARN("Error %d sending reject\n", rc);
2219 }
2220
2221 static int
2222 kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob)
2223 {
2224         rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
2225         struct kib_msg *reqmsg = priv;
2226         struct kib_msg *ackmsg;
2227         struct kib_dev *ibdev;
2228         struct kib_peer *peer;
2229         struct kib_peer *peer2;
2230         struct kib_conn *conn;
2231         struct lnet_ni *ni  = NULL;
2232         struct kib_net *net = NULL;
2233         lnet_nid_t nid;
2234         struct rdma_conn_param cp;
2235         struct kib_rej rej;
2236         int version = IBLND_MSG_VERSION;
2237         unsigned long flags;
2238         int max_frags;
2239         int rc;
2240         struct sockaddr_in *peer_addr;
2241
2242         LASSERT(!in_interrupt());
2243
2244         /* cmid inherits 'context' from the corresponding listener id */
2245         ibdev = (struct kib_dev *)cmid->context;
2246         LASSERT(ibdev);
2247
2248         memset(&rej, 0, sizeof(rej));
2249         rej.ibr_magic = IBLND_MSG_MAGIC;
2250         rej.ibr_why = IBLND_REJECT_FATAL;
2251         rej.ibr_cp.ibcp_max_msg_size = IBLND_MSG_SIZE;
2252
2253         peer_addr = (struct sockaddr_in *)&cmid->route.addr.dst_addr;
2254         if (*kiblnd_tunables.kib_require_priv_port &&
2255             ntohs(peer_addr->sin_port) >= PROT_SOCK) {
2256                 __u32 ip = ntohl(peer_addr->sin_addr.s_addr);
2257
2258                 CERROR("Peer's port (%pI4h:%hu) is not privileged\n",
2259                        &ip, ntohs(peer_addr->sin_port));
2260                 goto failed;
2261         }
2262
2263         if (priv_nob < offsetof(struct kib_msg, ibm_type)) {
2264                 CERROR("Short connection request\n");
2265                 goto failed;
2266         }
2267
2268         /*
2269          * Future protocol version compatibility support!  If the
2270          * o2iblnd-specific protocol changes, or when LNET unifies
2271          * protocols over all LNDs, the initial connection will
2272          * negotiate a protocol version.  I trap this here to avoid
2273          * console errors; the reject tells the peer which protocol I
2274          * speak.
2275          */
2276         if (reqmsg->ibm_magic == LNET_PROTO_MAGIC ||
2277             reqmsg->ibm_magic == __swab32(LNET_PROTO_MAGIC))
2278                 goto failed;
2279         if (reqmsg->ibm_magic == IBLND_MSG_MAGIC &&
2280             reqmsg->ibm_version != IBLND_MSG_VERSION &&
2281             reqmsg->ibm_version != IBLND_MSG_VERSION_1)
2282                 goto failed;
2283         if (reqmsg->ibm_magic == __swab32(IBLND_MSG_MAGIC) &&
2284             reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION) &&
2285             reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION_1))
2286                 goto failed;
2287
2288         rc = kiblnd_unpack_msg(reqmsg, priv_nob);
2289         if (rc) {
2290                 CERROR("Can't parse connection request: %d\n", rc);
2291                 goto failed;
2292         }
2293
2294         nid = reqmsg->ibm_srcnid;
2295         ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid));
2296
2297         if (ni) {
2298                 net = (struct kib_net *)ni->ni_data;
2299                 rej.ibr_incarnation = net->ibn_incarnation;
2300         }
2301
2302         if (!ni ||                       /* no matching net */
2303             ni->ni_nid != reqmsg->ibm_dstnid ||   /* right NET, wrong NID! */
2304             net->ibn_dev != ibdev) {          /* wrong device */
2305                 CERROR("Can't accept conn from %s on %s (%s:%d:%pI4h): bad dst nid %s\n",
2306                        libcfs_nid2str(nid),
2307                        !ni ? "NA" : libcfs_nid2str(ni->ni_nid),
2308                        ibdev->ibd_ifname, ibdev->ibd_nnets,
2309                        &ibdev->ibd_ifip,
2310                        libcfs_nid2str(reqmsg->ibm_dstnid));
2311
2312                 goto failed;
2313         }
2314
2315        /* check time stamp as soon as possible */
2316         if (reqmsg->ibm_dststamp &&
2317             reqmsg->ibm_dststamp != net->ibn_incarnation) {
2318                 CWARN("Stale connection request\n");
2319                 rej.ibr_why = IBLND_REJECT_CONN_STALE;
2320                 goto failed;
2321         }
2322
2323         /* I can accept peer's version */
2324         version = reqmsg->ibm_version;
2325
2326         if (reqmsg->ibm_type != IBLND_MSG_CONNREQ) {
2327                 CERROR("Unexpected connreq msg type: %x from %s\n",
2328                        reqmsg->ibm_type, libcfs_nid2str(nid));
2329                 goto failed;
2330         }
2331
2332         if (reqmsg->ibm_u.connparams.ibcp_queue_depth >
2333             kiblnd_msg_queue_size(version, ni)) {
2334                 CERROR("Can't accept conn from %s, queue depth too large: %d (<=%d wanted)\n",
2335                        libcfs_nid2str(nid),
2336                        reqmsg->ibm_u.connparams.ibcp_queue_depth,
2337                        kiblnd_msg_queue_size(version, ni));
2338
2339                 if (version == IBLND_MSG_VERSION)
2340                         rej.ibr_why = IBLND_REJECT_MSG_QUEUE_SIZE;
2341
2342                 goto failed;
2343         }
2344
2345         max_frags = reqmsg->ibm_u.connparams.ibcp_max_frags >> IBLND_FRAG_SHIFT;
2346         if (max_frags > kiblnd_rdma_frags(version, ni)) {
2347                 CWARN("Can't accept conn from %s (version %x): max message size %d is too large (%d wanted)\n",
2348                       libcfs_nid2str(nid), version, max_frags,
2349                       kiblnd_rdma_frags(version, ni));
2350
2351                 if (version >= IBLND_MSG_VERSION)
2352                         rej.ibr_why = IBLND_REJECT_RDMA_FRAGS;
2353
2354                 goto failed;
2355         } else if (max_frags < kiblnd_rdma_frags(version, ni) &&
2356                    !net->ibn_fmr_ps) {
2357                 CWARN("Can't accept conn from %s (version %x): max message size %d incompatible without FMR pool (%d wanted)\n",
2358                       libcfs_nid2str(nid), version, max_frags,
2359                       kiblnd_rdma_frags(version, ni));
2360
2361                 if (version == IBLND_MSG_VERSION)
2362                         rej.ibr_why = IBLND_REJECT_RDMA_FRAGS;
2363
2364                 goto failed;
2365         }
2366
2367         if (reqmsg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2368                 CERROR("Can't accept %s: message size %d too big (%d max)\n",
2369                        libcfs_nid2str(nid),
2370                        reqmsg->ibm_u.connparams.ibcp_max_msg_size,
2371                        IBLND_MSG_SIZE);
2372                 goto failed;
2373         }
2374
2375         /* assume 'nid' is a new peer; create  */
2376         rc = kiblnd_create_peer(ni, &peer, nid);
2377         if (rc) {
2378                 CERROR("Can't create peer for %s\n", libcfs_nid2str(nid));
2379                 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2380                 goto failed;
2381         }
2382
2383         /* We have validated the peer's parameters so use those */
2384         peer->ibp_max_frags = max_frags;
2385         peer->ibp_queue_depth = reqmsg->ibm_u.connparams.ibcp_queue_depth;
2386
2387         write_lock_irqsave(g_lock, flags);
2388
2389         peer2 = kiblnd_find_peer_locked(nid);
2390         if (peer2) {
2391                 if (!peer2->ibp_version) {
2392                         peer2->ibp_version     = version;
2393                         peer2->ibp_incarnation = reqmsg->ibm_srcstamp;
2394                 }
2395
2396                 /* not the guy I've talked with */
2397                 if (peer2->ibp_incarnation != reqmsg->ibm_srcstamp ||
2398                     peer2->ibp_version     != version) {
2399                         kiblnd_close_peer_conns_locked(peer2, -ESTALE);
2400
2401                         if (kiblnd_peer_active(peer2)) {
2402                                 peer2->ibp_incarnation = reqmsg->ibm_srcstamp;
2403                                 peer2->ibp_version = version;
2404                         }
2405                         write_unlock_irqrestore(g_lock, flags);
2406
2407                         CWARN("Conn stale %s version %x/%x incarnation %llu/%llu\n",
2408                               libcfs_nid2str(nid), peer2->ibp_version, version,
2409                               peer2->ibp_incarnation, reqmsg->ibm_srcstamp);
2410
2411                         kiblnd_peer_decref(peer);
2412                         rej.ibr_why = IBLND_REJECT_CONN_STALE;
2413                         goto failed;
2414                 }
2415
2416                 /*
2417                  * Tie-break connection race in favour of the higher NID.
2418                  * If we keep running into a race condition multiple times,
2419                  * we have to assume that the connection attempt with the
2420                  * higher NID is stuck in a connecting state and will never
2421                  * recover.  As such, we pass through this if-block and let
2422                  * the lower NID connection win so we can move forward.
2423                  */
2424                 if (peer2->ibp_connecting &&
2425                     nid < ni->ni_nid && peer2->ibp_races <
2426                     MAX_CONN_RACES_BEFORE_ABORT) {
2427                         peer2->ibp_races++;
2428                         write_unlock_irqrestore(g_lock, flags);
2429
2430                         CDEBUG(D_NET, "Conn race %s\n",
2431                                libcfs_nid2str(peer2->ibp_nid));
2432
2433                         kiblnd_peer_decref(peer);
2434                         rej.ibr_why = IBLND_REJECT_CONN_RACE;
2435                         goto failed;
2436                 }
2437                 if (peer2->ibp_races >= MAX_CONN_RACES_BEFORE_ABORT)
2438                         CNETERR("Conn race %s: unresolved after %d attempts, letting lower NID win\n",
2439                                 libcfs_nid2str(peer2->ibp_nid),
2440                                 MAX_CONN_RACES_BEFORE_ABORT);
2441                 /**
2442                  * passive connection is allowed even this peer is waiting for
2443                  * reconnection.
2444                  */
2445                 peer2->ibp_reconnecting = 0;
2446                 peer2->ibp_races = 0;
2447                 peer2->ibp_accepting++;
2448                 kiblnd_peer_addref(peer2);
2449
2450                 /**
2451                  * Race with kiblnd_launch_tx (active connect) to create peer
2452                  * so copy validated parameters since we now know what the
2453                  * peer's limits are
2454                  */
2455                 peer2->ibp_max_frags = peer->ibp_max_frags;
2456                 peer2->ibp_queue_depth = peer->ibp_queue_depth;
2457
2458                 write_unlock_irqrestore(g_lock, flags);
2459                 kiblnd_peer_decref(peer);
2460                 peer = peer2;
2461         } else {
2462                 /* Brand new peer */
2463                 LASSERT(!peer->ibp_accepting);
2464                 LASSERT(!peer->ibp_version &&
2465                         !peer->ibp_incarnation);
2466
2467                 peer->ibp_accepting   = 1;
2468                 peer->ibp_version     = version;
2469                 peer->ibp_incarnation = reqmsg->ibm_srcstamp;
2470
2471                 /* I have a ref on ni that prevents it being shutdown */
2472                 LASSERT(!net->ibn_shutdown);
2473
2474                 kiblnd_peer_addref(peer);
2475                 list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
2476
2477                 write_unlock_irqrestore(g_lock, flags);
2478         }
2479
2480         conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_PASSIVE_WAIT,
2481                                   version);
2482         if (!conn) {
2483                 kiblnd_peer_connect_failed(peer, 0, -ENOMEM);
2484                 kiblnd_peer_decref(peer);
2485                 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2486                 goto failed;
2487         }
2488
2489         /*
2490          * conn now "owns" cmid, so I return success from here on to ensure the
2491          * CM callback doesn't destroy cmid.
2492          */
2493         conn->ibc_incarnation      = reqmsg->ibm_srcstamp;
2494         conn->ibc_credits          = conn->ibc_queue_depth;
2495         conn->ibc_reserved_credits = conn->ibc_queue_depth;
2496         LASSERT(conn->ibc_credits + conn->ibc_reserved_credits +
2497                 IBLND_OOB_MSGS(version) <= IBLND_RX_MSGS(conn));
2498
2499         ackmsg = &conn->ibc_connvars->cv_msg;
2500         memset(ackmsg, 0, sizeof(*ackmsg));
2501
2502         kiblnd_init_msg(ackmsg, IBLND_MSG_CONNACK,
2503                         sizeof(ackmsg->ibm_u.connparams));
2504         ackmsg->ibm_u.connparams.ibcp_queue_depth = conn->ibc_queue_depth;
2505         ackmsg->ibm_u.connparams.ibcp_max_frags = conn->ibc_max_frags << IBLND_FRAG_SHIFT;
2506         ackmsg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2507
2508         kiblnd_pack_msg(ni, ackmsg, version, 0, nid, reqmsg->ibm_srcstamp);
2509
2510         memset(&cp, 0, sizeof(cp));
2511         cp.private_data = ackmsg;
2512         cp.private_data_len = ackmsg->ibm_nob;
2513         cp.responder_resources = 0;          /* No atomic ops or RDMA reads */
2514         cp.initiator_depth = 0;
2515         cp.flow_control = 1;
2516         cp.retry_count = *kiblnd_tunables.kib_retry_count;
2517         cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
2518
2519         CDEBUG(D_NET, "Accept %s\n", libcfs_nid2str(nid));
2520
2521         rc = rdma_accept(cmid, &cp);
2522         if (rc) {
2523                 CERROR("Can't accept %s: %d\n", libcfs_nid2str(nid), rc);
2524                 rej.ibr_version = version;
2525                 rej.ibr_why     = IBLND_REJECT_FATAL;
2526
2527                 kiblnd_reject(cmid, &rej);
2528                 kiblnd_connreq_done(conn, rc);
2529                 kiblnd_conn_decref(conn);
2530         }
2531
2532         lnet_ni_decref(ni);
2533         return 0;
2534
2535  failed:
2536         if (ni) {
2537                 rej.ibr_cp.ibcp_queue_depth = kiblnd_msg_queue_size(version, ni);
2538                 rej.ibr_cp.ibcp_max_frags = kiblnd_rdma_frags(version, ni);
2539                 lnet_ni_decref(ni);
2540         }
2541
2542         rej.ibr_version             = version;
2543         kiblnd_reject(cmid, &rej);
2544
2545         return -ECONNREFUSED;
2546 }
2547
2548 static void
2549 kiblnd_check_reconnect(struct kib_conn *conn, int version,
2550                        __u64 incarnation, int why, struct kib_connparams *cp)
2551 {
2552         rwlock_t *glock = &kiblnd_data.kib_global_lock;
2553         struct kib_peer *peer = conn->ibc_peer;
2554         char *reason;
2555         int msg_size = IBLND_MSG_SIZE;
2556         int frag_num = -1;
2557         int queue_dep = -1;
2558         bool reconnect;
2559         unsigned long flags;
2560
2561         LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2562         LASSERT(peer->ibp_connecting > 0);     /* 'conn' at least */
2563         LASSERT(!peer->ibp_reconnecting);
2564
2565         if (cp) {
2566                 msg_size = cp->ibcp_max_msg_size;
2567                 frag_num        = cp->ibcp_max_frags << IBLND_FRAG_SHIFT;
2568                 queue_dep = cp->ibcp_queue_depth;
2569         }
2570
2571         write_lock_irqsave(glock, flags);
2572         /**
2573          * retry connection if it's still needed and no other connection
2574          * attempts (active or passive) are in progress
2575          * NB: reconnect is still needed even when ibp_tx_queue is
2576          * empty if ibp_version != version because reconnect may be
2577          * initiated by kiblnd_query()
2578          */
2579         reconnect = (!list_empty(&peer->ibp_tx_queue) ||
2580                      peer->ibp_version != version) &&
2581                     peer->ibp_connecting == 1 &&
2582                     !peer->ibp_accepting;
2583         if (!reconnect) {
2584                 reason = "no need";
2585                 goto out;
2586         }
2587
2588         switch (why) {
2589         default:
2590                 reason = "Unknown";
2591                 break;
2592
2593         case IBLND_REJECT_RDMA_FRAGS: {
2594                 struct lnet_ioctl_config_lnd_tunables *tunables;
2595
2596                 if (!cp) {
2597                         reason = "can't negotiate max frags";
2598                         goto out;
2599                 }
2600                 tunables = peer->ibp_ni->ni_lnd_tunables;
2601                 if (!tunables->lt_tun_u.lt_o2ib.lnd_map_on_demand) {
2602                         reason = "map_on_demand must be enabled";
2603                         goto out;
2604                 }
2605                 if (conn->ibc_max_frags <= frag_num) {
2606                         reason = "unsupported max frags";
2607                         goto out;
2608                 }
2609
2610                 peer->ibp_max_frags = frag_num;
2611                 reason = "rdma fragments";
2612                 break;
2613         }
2614         case IBLND_REJECT_MSG_QUEUE_SIZE:
2615                 if (!cp) {
2616                         reason = "can't negotiate queue depth";
2617                         goto out;
2618                 }
2619                 if (conn->ibc_queue_depth <= queue_dep) {
2620                         reason = "unsupported queue depth";
2621                         goto out;
2622                 }
2623
2624                 peer->ibp_queue_depth = queue_dep;
2625                 reason = "queue depth";
2626                 break;
2627
2628         case IBLND_REJECT_CONN_STALE:
2629                 reason = "stale";
2630                 break;
2631
2632         case IBLND_REJECT_CONN_RACE:
2633                 reason = "conn race";
2634                 break;
2635
2636         case IBLND_REJECT_CONN_UNCOMPAT:
2637                 reason = "version negotiation";
2638                 break;
2639         }
2640
2641         conn->ibc_reconnect = 1;
2642         peer->ibp_reconnecting = 1;
2643         peer->ibp_version = version;
2644         if (incarnation)
2645                 peer->ibp_incarnation = incarnation;
2646 out:
2647         write_unlock_irqrestore(glock, flags);
2648
2649         CNETERR("%s: %s (%s), %x, %x, msg_size: %d, queue_depth: %d/%d, max_frags: %d/%d\n",
2650                 libcfs_nid2str(peer->ibp_nid),
2651                 reconnect ? "reconnect" : "don't reconnect",
2652                 reason, IBLND_MSG_VERSION, version, msg_size,
2653                 conn->ibc_queue_depth, queue_dep,
2654                 conn->ibc_max_frags, frag_num);
2655         /**
2656          * if conn::ibc_reconnect is TRUE, connd will reconnect to the peer
2657          * while destroying the zombie
2658          */
2659 }
2660
2661 static void
2662 kiblnd_rejected(struct kib_conn *conn, int reason, void *priv, int priv_nob)
2663 {
2664         struct kib_peer *peer = conn->ibc_peer;
2665
2666         LASSERT(!in_interrupt());
2667         LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2668
2669         switch (reason) {
2670         case IB_CM_REJ_STALE_CONN:
2671                 kiblnd_check_reconnect(conn, IBLND_MSG_VERSION, 0,
2672                                        IBLND_REJECT_CONN_STALE, NULL);
2673                 break;
2674
2675         case IB_CM_REJ_INVALID_SERVICE_ID:
2676                 CNETERR("%s rejected: no listener at %d\n",
2677                         libcfs_nid2str(peer->ibp_nid),
2678                         *kiblnd_tunables.kib_service);
2679                 break;
2680
2681         case IB_CM_REJ_CONSUMER_DEFINED:
2682                 if (priv_nob >= offsetof(struct kib_rej, ibr_padding)) {
2683                         struct kib_rej *rej = priv;
2684                         struct kib_connparams *cp = NULL;
2685                         int flip = 0;
2686                         __u64 incarnation = -1;
2687
2688                         /* NB. default incarnation is -1 because:
2689                          * a) V1 will ignore dst incarnation in connreq.
2690                          * b) V2 will provide incarnation while rejecting me,
2691                          *    -1 will be overwrote.
2692                          *
2693                          * if I try to connect to a V1 peer with V2 protocol,
2694                          * it rejected me then upgrade to V2, I have no idea
2695                          * about the upgrading and try to reconnect with V1,
2696                          * in this case upgraded V2 can find out I'm trying to
2697                          * talk to the old guy and reject me(incarnation is -1).
2698                          */
2699
2700                         if (rej->ibr_magic == __swab32(IBLND_MSG_MAGIC) ||
2701                             rej->ibr_magic == __swab32(LNET_PROTO_MAGIC)) {
2702                                 __swab32s(&rej->ibr_magic);
2703                                 __swab16s(&rej->ibr_version);
2704                                 flip = 1;
2705                         }
2706
2707                         if (priv_nob >= sizeof(struct kib_rej) &&
2708                             rej->ibr_version > IBLND_MSG_VERSION_1) {
2709                                 /*
2710                                  * priv_nob is always 148 in current version
2711                                  * of OFED, so we still need to check version.
2712                                  * (define of IB_CM_REJ_PRIVATE_DATA_SIZE)
2713                                  */
2714                                 cp = &rej->ibr_cp;
2715
2716                                 if (flip) {
2717                                         __swab64s(&rej->ibr_incarnation);
2718                                         __swab16s(&cp->ibcp_queue_depth);
2719                                         __swab16s(&cp->ibcp_max_frags);
2720                                         __swab32s(&cp->ibcp_max_msg_size);
2721                                 }
2722
2723                                 incarnation = rej->ibr_incarnation;
2724                         }
2725
2726                         if (rej->ibr_magic != IBLND_MSG_MAGIC &&
2727                             rej->ibr_magic != LNET_PROTO_MAGIC) {
2728                                 CERROR("%s rejected: consumer defined fatal error\n",
2729                                        libcfs_nid2str(peer->ibp_nid));
2730                                 break;
2731                         }
2732
2733                         if (rej->ibr_version != IBLND_MSG_VERSION &&
2734                             rej->ibr_version != IBLND_MSG_VERSION_1) {
2735                                 CERROR("%s rejected: o2iblnd version %x error\n",
2736                                        libcfs_nid2str(peer->ibp_nid),
2737                                        rej->ibr_version);
2738                                 break;
2739                         }
2740
2741                         if (rej->ibr_why     == IBLND_REJECT_FATAL &&
2742                             rej->ibr_version == IBLND_MSG_VERSION_1) {
2743                                 CDEBUG(D_NET, "rejected by old version peer %s: %x\n",
2744                                        libcfs_nid2str(peer->ibp_nid), rej->ibr_version);
2745
2746                                 if (conn->ibc_version != IBLND_MSG_VERSION_1)
2747                                         rej->ibr_why = IBLND_REJECT_CONN_UNCOMPAT;
2748                         }
2749
2750                         switch (rej->ibr_why) {
2751                         case IBLND_REJECT_CONN_RACE:
2752                         case IBLND_REJECT_CONN_STALE:
2753                         case IBLND_REJECT_CONN_UNCOMPAT:
2754                         case IBLND_REJECT_MSG_QUEUE_SIZE:
2755                         case IBLND_REJECT_RDMA_FRAGS:
2756                                 kiblnd_check_reconnect(conn, rej->ibr_version,
2757                                                        incarnation,
2758                                                        rej->ibr_why, cp);
2759                                 break;
2760
2761                         case IBLND_REJECT_NO_RESOURCES:
2762                                 CERROR("%s rejected: o2iblnd no resources\n",
2763                                        libcfs_nid2str(peer->ibp_nid));
2764                                 break;
2765
2766                         case IBLND_REJECT_FATAL:
2767                                 CERROR("%s rejected: o2iblnd fatal error\n",
2768                                        libcfs_nid2str(peer->ibp_nid));
2769                                 break;
2770
2771                         default:
2772                                 CERROR("%s rejected: o2iblnd reason %d\n",
2773                                        libcfs_nid2str(peer->ibp_nid),
2774                                        rej->ibr_why);
2775                                 break;
2776                         }
2777                         break;
2778                 }
2779                 /* fall through */
2780         default:
2781                 CNETERR("%s rejected: reason %d, size %d\n",
2782                         libcfs_nid2str(peer->ibp_nid), reason, priv_nob);
2783                 break;
2784         }
2785
2786         kiblnd_connreq_done(conn, -ECONNREFUSED);
2787 }
2788
2789 static void
2790 kiblnd_check_connreply(struct kib_conn *conn, void *priv, int priv_nob)
2791 {
2792         struct kib_peer *peer = conn->ibc_peer;
2793         struct lnet_ni *ni = peer->ibp_ni;
2794         struct kib_net *net = ni->ni_data;
2795         struct kib_msg *msg = priv;
2796         int ver = conn->ibc_version;
2797         int rc = kiblnd_unpack_msg(msg, priv_nob);
2798         unsigned long flags;
2799
2800         LASSERT(net);
2801
2802         if (rc) {
2803                 CERROR("Can't unpack connack from %s: %d\n",
2804                        libcfs_nid2str(peer->ibp_nid), rc);
2805                 goto failed;
2806         }
2807
2808         if (msg->ibm_type != IBLND_MSG_CONNACK) {
2809                 CERROR("Unexpected message %d from %s\n",
2810                        msg->ibm_type, libcfs_nid2str(peer->ibp_nid));
2811                 rc = -EPROTO;
2812                 goto failed;
2813         }
2814
2815         if (ver != msg->ibm_version) {
2816                 CERROR("%s replied version %x is different with requested version %x\n",
2817                        libcfs_nid2str(peer->ibp_nid), msg->ibm_version, ver);
2818                 rc = -EPROTO;
2819                 goto failed;
2820         }
2821
2822         if (msg->ibm_u.connparams.ibcp_queue_depth >
2823             conn->ibc_queue_depth) {
2824                 CERROR("%s has incompatible queue depth %d (<=%d wanted)\n",
2825                        libcfs_nid2str(peer->ibp_nid),
2826                        msg->ibm_u.connparams.ibcp_queue_depth,
2827                        conn->ibc_queue_depth);
2828                 rc = -EPROTO;
2829                 goto failed;
2830         }
2831
2832         if ((msg->ibm_u.connparams.ibcp_max_frags >> IBLND_FRAG_SHIFT) >
2833             conn->ibc_max_frags) {
2834                 CERROR("%s has incompatible max_frags %d (<=%d wanted)\n",
2835                        libcfs_nid2str(peer->ibp_nid),
2836                        msg->ibm_u.connparams.ibcp_max_frags >> IBLND_FRAG_SHIFT,
2837                        conn->ibc_max_frags);
2838                 rc = -EPROTO;
2839                 goto failed;
2840         }
2841
2842         if (msg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2843                 CERROR("%s max message size %d too big (%d max)\n",
2844                        libcfs_nid2str(peer->ibp_nid),
2845                        msg->ibm_u.connparams.ibcp_max_msg_size,
2846                        IBLND_MSG_SIZE);
2847                 rc = -EPROTO;
2848                 goto failed;
2849         }
2850
2851         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2852         if (msg->ibm_dstnid == ni->ni_nid &&
2853             msg->ibm_dststamp == net->ibn_incarnation)
2854                 rc = 0;
2855         else
2856                 rc = -ESTALE;
2857         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2858
2859         if (rc) {
2860                 CERROR("Bad connection reply from %s, rc = %d, version: %x max_frags: %d\n",
2861                        libcfs_nid2str(peer->ibp_nid), rc,
2862                        msg->ibm_version, msg->ibm_u.connparams.ibcp_max_frags);
2863                 goto failed;
2864         }
2865
2866         conn->ibc_incarnation = msg->ibm_srcstamp;
2867         conn->ibc_credits = msg->ibm_u.connparams.ibcp_queue_depth;
2868         conn->ibc_reserved_credits = msg->ibm_u.connparams.ibcp_queue_depth;
2869         conn->ibc_queue_depth = msg->ibm_u.connparams.ibcp_queue_depth;
2870         conn->ibc_max_frags = msg->ibm_u.connparams.ibcp_max_frags >> IBLND_FRAG_SHIFT;
2871         LASSERT(conn->ibc_credits + conn->ibc_reserved_credits +
2872                 IBLND_OOB_MSGS(ver) <= IBLND_RX_MSGS(conn));
2873
2874         kiblnd_connreq_done(conn, 0);
2875         return;
2876
2877  failed:
2878         /*
2879          * NB My QP has already established itself, so I handle anything going
2880          * wrong here by setting ibc_comms_error.
2881          * kiblnd_connreq_done(0) moves the conn state to ESTABLISHED, but then
2882          * immediately tears it down.
2883          */
2884         LASSERT(rc);
2885         conn->ibc_comms_error = rc;
2886         kiblnd_connreq_done(conn, 0);
2887 }
2888
2889 static int
2890 kiblnd_active_connect(struct rdma_cm_id *cmid)
2891 {
2892         struct kib_peer *peer = (struct kib_peer *)cmid->context;
2893         struct kib_conn *conn;
2894         struct kib_msg *msg;
2895         struct rdma_conn_param cp;
2896         int version;
2897         __u64 incarnation;
2898         unsigned long flags;
2899         int rc;
2900
2901         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2902
2903         incarnation = peer->ibp_incarnation;
2904         version = !peer->ibp_version ? IBLND_MSG_VERSION :
2905                                        peer->ibp_version;
2906
2907         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2908
2909         conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_ACTIVE_CONNECT,
2910                                   version);
2911         if (!conn) {
2912                 kiblnd_peer_connect_failed(peer, 1, -ENOMEM);
2913                 kiblnd_peer_decref(peer); /* lose cmid's ref */
2914                 return -ENOMEM;
2915         }
2916
2917         /*
2918          * conn "owns" cmid now, so I return success from here on to ensure the
2919          * CM callback doesn't destroy cmid. conn also takes over cmid's ref
2920          * on peer
2921          */
2922         msg = &conn->ibc_connvars->cv_msg;
2923
2924         memset(msg, 0, sizeof(*msg));
2925         kiblnd_init_msg(msg, IBLND_MSG_CONNREQ, sizeof(msg->ibm_u.connparams));
2926         msg->ibm_u.connparams.ibcp_queue_depth = conn->ibc_queue_depth;
2927         msg->ibm_u.connparams.ibcp_max_frags = conn->ibc_max_frags << IBLND_FRAG_SHIFT;
2928         msg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2929
2930         kiblnd_pack_msg(peer->ibp_ni, msg, version,
2931                         0, peer->ibp_nid, incarnation);
2932
2933         memset(&cp, 0, sizeof(cp));
2934         cp.private_data = msg;
2935         cp.private_data_len    = msg->ibm_nob;
2936         cp.responder_resources = 0;          /* No atomic ops or RDMA reads */
2937         cp.initiator_depth     = 0;
2938         cp.flow_control        = 1;
2939         cp.retry_count         = *kiblnd_tunables.kib_retry_count;
2940         cp.rnr_retry_count     = *kiblnd_tunables.kib_rnr_retry_count;
2941
2942         LASSERT(cmid->context == (void *)conn);
2943         LASSERT(conn->ibc_cmid == cmid);
2944
2945         rc = rdma_connect(cmid, &cp);
2946         if (rc) {
2947                 CERROR("Can't connect to %s: %d\n",
2948                        libcfs_nid2str(peer->ibp_nid), rc);
2949                 kiblnd_connreq_done(conn, rc);
2950                 kiblnd_conn_decref(conn);
2951         }
2952
2953         return 0;
2954 }
2955
2956 int
2957 kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event)
2958 {
2959         struct kib_peer *peer;
2960         struct kib_conn *conn;
2961         int rc;
2962
2963         switch (event->event) {
2964         default:
2965                 CERROR("Unexpected event: %d, status: %d\n",
2966                        event->event, event->status);
2967                 LBUG();
2968
2969         case RDMA_CM_EVENT_CONNECT_REQUEST:
2970                 /* destroy cmid on failure */
2971                 rc = kiblnd_passive_connect(cmid,
2972                                             (void *)KIBLND_CONN_PARAM(event),
2973                                             KIBLND_CONN_PARAM_LEN(event));
2974                 CDEBUG(D_NET, "connreq: %d\n", rc);
2975                 return rc;
2976
2977         case RDMA_CM_EVENT_ADDR_ERROR:
2978                 peer = (struct kib_peer *)cmid->context;
2979                 CNETERR("%s: ADDR ERROR %d\n",
2980                         libcfs_nid2str(peer->ibp_nid), event->status);
2981                 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
2982                 kiblnd_peer_decref(peer);
2983                 return -EHOSTUNREACH;      /* rc destroys cmid */
2984
2985         case RDMA_CM_EVENT_ADDR_RESOLVED:
2986                 peer = (struct kib_peer *)cmid->context;
2987
2988                 CDEBUG(D_NET, "%s Addr resolved: %d\n",
2989                        libcfs_nid2str(peer->ibp_nid), event->status);
2990
2991                 if (event->status) {
2992                         CNETERR("Can't resolve address for %s: %d\n",
2993                                 libcfs_nid2str(peer->ibp_nid), event->status);
2994                         rc = event->status;
2995                 } else {
2996                         rc = rdma_resolve_route(
2997                                 cmid, *kiblnd_tunables.kib_timeout * 1000);
2998                         if (!rc)
2999                                 return 0;
3000                         /* Can't initiate route resolution */
3001                         CERROR("Can't resolve route for %s: %d\n",
3002                                libcfs_nid2str(peer->ibp_nid), rc);
3003                 }
3004                 kiblnd_peer_connect_failed(peer, 1, rc);
3005                 kiblnd_peer_decref(peer);
3006                 return rc;                    /* rc destroys cmid */
3007
3008         case RDMA_CM_EVENT_ROUTE_ERROR:
3009                 peer = (struct kib_peer *)cmid->context;
3010                 CNETERR("%s: ROUTE ERROR %d\n",
3011                         libcfs_nid2str(peer->ibp_nid), event->status);
3012                 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
3013                 kiblnd_peer_decref(peer);
3014                 return -EHOSTUNREACH;      /* rc destroys cmid */
3015
3016         case RDMA_CM_EVENT_ROUTE_RESOLVED:
3017                 peer = (struct kib_peer *)cmid->context;
3018                 CDEBUG(D_NET, "%s Route resolved: %d\n",
3019                        libcfs_nid2str(peer->ibp_nid), event->status);
3020
3021                 if (!event->status)
3022                         return kiblnd_active_connect(cmid);
3023
3024                 CNETERR("Can't resolve route for %s: %d\n",
3025                         libcfs_nid2str(peer->ibp_nid), event->status);
3026                 kiblnd_peer_connect_failed(peer, 1, event->status);
3027                 kiblnd_peer_decref(peer);
3028                 return event->status;      /* rc destroys cmid */
3029
3030         case RDMA_CM_EVENT_UNREACHABLE:
3031                 conn = (struct kib_conn *)cmid->context;
3032                 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
3033                         conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
3034                 CNETERR("%s: UNREACHABLE %d\n",
3035                         libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
3036                 kiblnd_connreq_done(conn, -ENETDOWN);
3037                 kiblnd_conn_decref(conn);
3038                 return 0;
3039
3040         case RDMA_CM_EVENT_CONNECT_ERROR:
3041                 conn = (struct kib_conn *)cmid->context;
3042                 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
3043                         conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
3044                 CNETERR("%s: CONNECT ERROR %d\n",
3045                         libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
3046                 kiblnd_connreq_done(conn, -ENOTCONN);
3047                 kiblnd_conn_decref(conn);
3048                 return 0;
3049
3050         case RDMA_CM_EVENT_REJECTED:
3051                 conn = (struct kib_conn *)cmid->context;
3052                 switch (conn->ibc_state) {
3053                 default:
3054                         LBUG();
3055
3056                 case IBLND_CONN_PASSIVE_WAIT:
3057                         CERROR("%s: REJECTED %d\n",
3058                                libcfs_nid2str(conn->ibc_peer->ibp_nid),
3059                                event->status);
3060                         kiblnd_connreq_done(conn, -ECONNRESET);
3061                         break;
3062
3063                 case IBLND_CONN_ACTIVE_CONNECT:
3064                         kiblnd_rejected(conn, event->status,
3065                                         (void *)KIBLND_CONN_PARAM(event),
3066                                         KIBLND_CONN_PARAM_LEN(event));
3067                         break;
3068                 }
3069                 kiblnd_conn_decref(conn);
3070                 return 0;
3071
3072         case RDMA_CM_EVENT_ESTABLISHED:
3073                 conn = (struct kib_conn *)cmid->context;
3074                 switch (conn->ibc_state) {
3075                 default:
3076                         LBUG();
3077
3078                 case IBLND_CONN_PASSIVE_WAIT:
3079                         CDEBUG(D_NET, "ESTABLISHED (passive): %s\n",
3080                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
3081                         kiblnd_connreq_done(conn, 0);
3082                         break;
3083
3084                 case IBLND_CONN_ACTIVE_CONNECT:
3085                         CDEBUG(D_NET, "ESTABLISHED(active): %s\n",
3086                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
3087                         kiblnd_check_connreply(conn,
3088                                                (void *)KIBLND_CONN_PARAM(event),
3089                                                KIBLND_CONN_PARAM_LEN(event));
3090                         break;
3091                 }
3092                 /* net keeps its ref on conn! */
3093                 return 0;
3094
3095         case RDMA_CM_EVENT_TIMEWAIT_EXIT:
3096                 CDEBUG(D_NET, "Ignore TIMEWAIT_EXIT event\n");
3097                 return 0;
3098         case RDMA_CM_EVENT_DISCONNECTED:
3099                 conn = (struct kib_conn *)cmid->context;
3100                 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
3101                         CERROR("%s DISCONNECTED\n",
3102                                libcfs_nid2str(conn->ibc_peer->ibp_nid));
3103                         kiblnd_connreq_done(conn, -ECONNRESET);
3104                 } else {
3105                         kiblnd_close_conn(conn, 0);
3106                 }
3107                 kiblnd_conn_decref(conn);
3108                 cmid->context = NULL;
3109                 return 0;
3110
3111         case RDMA_CM_EVENT_DEVICE_REMOVAL:
3112                 LCONSOLE_ERROR_MSG(0x131,
3113                                    "Received notification of device removal\n"
3114                                    "Please shutdown LNET to allow this to proceed\n");
3115                 /*
3116                  * Can't remove network from underneath LNET for now, so I have
3117                  * to ignore this
3118                  */
3119                 return 0;
3120
3121         case RDMA_CM_EVENT_ADDR_CHANGE:
3122                 LCONSOLE_INFO("Physical link changed (eg hca/port)\n");
3123                 return 0;
3124         }
3125 }
3126
3127 static int
3128 kiblnd_check_txs_locked(struct kib_conn *conn, struct list_head *txs)
3129 {
3130         struct kib_tx *tx;
3131         struct list_head *ttmp;
3132
3133         list_for_each(ttmp, txs) {
3134                 tx = list_entry(ttmp, struct kib_tx, tx_list);
3135
3136                 if (txs != &conn->ibc_active_txs) {
3137                         LASSERT(tx->tx_queued);
3138                 } else {
3139                         LASSERT(!tx->tx_queued);
3140                         LASSERT(tx->tx_waiting || tx->tx_sending);
3141                 }
3142
3143                 if (cfs_time_aftereq(jiffies, tx->tx_deadline)) {
3144                         CERROR("Timed out tx: %s, %lu seconds\n",
3145                                kiblnd_queue2str(conn, txs),
3146                                cfs_duration_sec(jiffies - tx->tx_deadline));
3147                         return 1;
3148                 }
3149         }
3150
3151         return 0;
3152 }
3153
3154 static int
3155 kiblnd_conn_timed_out_locked(struct kib_conn *conn)
3156 {
3157         return  kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue) ||
3158                 kiblnd_check_txs_locked(conn, &conn->ibc_tx_noops) ||
3159                 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_rsrvd) ||
3160                 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_nocred) ||
3161                 kiblnd_check_txs_locked(conn, &conn->ibc_active_txs);
3162 }
3163
3164 static void
3165 kiblnd_check_conns(int idx)
3166 {
3167         LIST_HEAD(closes);
3168         LIST_HEAD(checksends);
3169         struct list_head *peers = &kiblnd_data.kib_peers[idx];
3170         struct list_head *ptmp;
3171         struct kib_peer *peer;
3172         struct kib_conn *conn;
3173         struct kib_conn *temp;
3174         struct kib_conn *tmp;
3175         struct list_head *ctmp;
3176         unsigned long flags;
3177
3178         /*
3179          * NB. We expect to have a look at all the peers and not find any
3180          * RDMAs to time out, so we just use a shared lock while we
3181          * take a look...
3182          */
3183         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
3184
3185         list_for_each(ptmp, peers) {
3186                 peer = list_entry(ptmp, struct kib_peer, ibp_list);
3187
3188                 list_for_each(ctmp, &peer->ibp_conns) {
3189                         int timedout;
3190                         int sendnoop;
3191
3192                         conn = list_entry(ctmp, struct kib_conn, ibc_list);
3193
3194                         LASSERT(conn->ibc_state == IBLND_CONN_ESTABLISHED);
3195
3196                         spin_lock(&conn->ibc_lock);
3197
3198                         sendnoop = kiblnd_need_noop(conn);
3199                         timedout = kiblnd_conn_timed_out_locked(conn);
3200                         if (!sendnoop && !timedout) {
3201                                 spin_unlock(&conn->ibc_lock);
3202                                 continue;
3203                         }
3204
3205                         if (timedout) {
3206                                 CERROR("Timed out RDMA with %s (%lu): c: %u, oc: %u, rc: %u\n",
3207                                        libcfs_nid2str(peer->ibp_nid),
3208                                        cfs_duration_sec(cfs_time_current() -
3209                                                         peer->ibp_last_alive),
3210                                        conn->ibc_credits,
3211                                        conn->ibc_outstanding_credits,
3212                                        conn->ibc_reserved_credits);
3213                                 list_add(&conn->ibc_connd_list, &closes);
3214                         } else {
3215                                 list_add(&conn->ibc_connd_list, &checksends);
3216                         }
3217                         /* +ref for 'closes' or 'checksends' */
3218                         kiblnd_conn_addref(conn);
3219
3220                         spin_unlock(&conn->ibc_lock);
3221                 }
3222         }
3223
3224         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
3225
3226         /*
3227          * Handle timeout by closing the whole
3228          * connection. We can only be sure RDMA activity
3229          * has ceased once the QP has been modified.
3230          */
3231         list_for_each_entry_safe(conn, tmp, &closes, ibc_connd_list) {
3232                 list_del(&conn->ibc_connd_list);
3233                 kiblnd_close_conn(conn, -ETIMEDOUT);
3234                 kiblnd_conn_decref(conn);
3235         }
3236
3237         /*
3238          * In case we have enough credits to return via a
3239          * NOOP, but there were no non-blocking tx descs
3240          * free to do it last time...
3241          */
3242         list_for_each_entry_safe(conn, temp, &checksends, ibc_connd_list) {
3243                 list_del(&conn->ibc_connd_list);
3244
3245                 spin_lock(&conn->ibc_lock);
3246                 kiblnd_check_sends_locked(conn);
3247                 spin_unlock(&conn->ibc_lock);
3248
3249                 kiblnd_conn_decref(conn);
3250         }
3251 }
3252
3253 static void
3254 kiblnd_disconnect_conn(struct kib_conn *conn)
3255 {
3256         LASSERT(!in_interrupt());
3257         LASSERT(current == kiblnd_data.kib_connd);
3258         LASSERT(conn->ibc_state == IBLND_CONN_CLOSING);
3259
3260         rdma_disconnect(conn->ibc_cmid);
3261         kiblnd_finalise_conn(conn);
3262
3263         kiblnd_peer_notify(conn->ibc_peer);
3264 }
3265
3266 /**
3267  * High-water for reconnection to the same peer, reconnection attempt should
3268  * be delayed after trying more than KIB_RECONN_HIGH_RACE.
3269  */
3270 #define KIB_RECONN_HIGH_RACE    10
3271 /**
3272  * Allow connd to take a break and handle other things after consecutive
3273  * reconnection attempts.
3274  */
3275 #define KIB_RECONN_BREAK        100
3276
3277 int
3278 kiblnd_connd(void *arg)
3279 {
3280         spinlock_t *lock = &kiblnd_data.kib_connd_lock;
3281         wait_queue_entry_t wait;
3282         unsigned long flags;
3283         struct kib_conn *conn;
3284         int timeout;
3285         int i;
3286         int dropped_lock;
3287         int peer_index = 0;
3288         unsigned long deadline = jiffies;
3289
3290         cfs_block_allsigs();
3291
3292         init_waitqueue_entry(&wait, current);
3293         kiblnd_data.kib_connd = current;
3294
3295         spin_lock_irqsave(lock, flags);
3296
3297         while (!kiblnd_data.kib_shutdown) {
3298                 int reconn = 0;
3299
3300                 dropped_lock = 0;
3301
3302                 if (!list_empty(&kiblnd_data.kib_connd_zombies)) {
3303                         struct kib_peer *peer = NULL;
3304
3305                         conn = list_entry(kiblnd_data.kib_connd_zombies.next,
3306                                           struct kib_conn, ibc_list);
3307                         list_del(&conn->ibc_list);
3308                         if (conn->ibc_reconnect) {
3309                                 peer = conn->ibc_peer;
3310                                 kiblnd_peer_addref(peer);
3311                         }
3312
3313                         spin_unlock_irqrestore(lock, flags);
3314                         dropped_lock = 1;
3315
3316                         kiblnd_destroy_conn(conn, !peer);
3317
3318                         spin_lock_irqsave(lock, flags);
3319                         if (!peer)
3320                                 continue;
3321
3322                         conn->ibc_peer = peer;
3323                         if (peer->ibp_reconnected < KIB_RECONN_HIGH_RACE)
3324                                 list_add_tail(&conn->ibc_list,
3325                                               &kiblnd_data.kib_reconn_list);
3326                         else
3327                                 list_add_tail(&conn->ibc_list,
3328                                               &kiblnd_data.kib_reconn_wait);
3329                 }
3330
3331                 if (!list_empty(&kiblnd_data.kib_connd_conns)) {
3332                         conn = list_entry(kiblnd_data.kib_connd_conns.next,
3333                                           struct kib_conn, ibc_list);
3334                         list_del(&conn->ibc_list);
3335
3336                         spin_unlock_irqrestore(lock, flags);
3337                         dropped_lock = 1;
3338
3339                         kiblnd_disconnect_conn(conn);
3340                         kiblnd_conn_decref(conn);
3341
3342                         spin_lock_irqsave(lock, flags);
3343                 }
3344
3345                 while (reconn < KIB_RECONN_BREAK) {
3346                         if (kiblnd_data.kib_reconn_sec !=
3347                             ktime_get_real_seconds()) {
3348                                 kiblnd_data.kib_reconn_sec = ktime_get_real_seconds();
3349                                 list_splice_init(&kiblnd_data.kib_reconn_wait,
3350                                                  &kiblnd_data.kib_reconn_list);
3351                         }
3352
3353                         if (list_empty(&kiblnd_data.kib_reconn_list))
3354                                 break;
3355
3356                         conn = list_entry(kiblnd_data.kib_reconn_list.next,
3357                                           struct kib_conn, ibc_list);
3358                         list_del(&conn->ibc_list);
3359
3360                         spin_unlock_irqrestore(lock, flags);
3361                         dropped_lock = 1;
3362
3363                         reconn += kiblnd_reconnect_peer(conn->ibc_peer);
3364                         kiblnd_peer_decref(conn->ibc_peer);
3365                         LIBCFS_FREE(conn, sizeof(*conn));
3366
3367                         spin_lock_irqsave(lock, flags);
3368                 }
3369
3370                 /* careful with the jiffy wrap... */
3371                 timeout = (int)(deadline - jiffies);
3372                 if (timeout <= 0) {
3373                         const int n = 4;
3374                         const int p = 1;
3375                         int chunk = kiblnd_data.kib_peer_hash_size;
3376
3377                         spin_unlock_irqrestore(lock, flags);
3378                         dropped_lock = 1;
3379
3380                         /*
3381                          * Time to check for RDMA timeouts on a few more
3382                          * peers: I do checks every 'p' seconds on a
3383                          * proportion of the peer table and I need to check
3384                          * every connection 'n' times within a timeout
3385                          * interval, to ensure I detect a timeout on any
3386                          * connection within (n+1)/n times the timeout
3387                          * interval.
3388                          */
3389                         if (*kiblnd_tunables.kib_timeout > n * p)
3390                                 chunk = (chunk * n * p) /
3391                                         *kiblnd_tunables.kib_timeout;
3392                         if (!chunk)
3393                                 chunk = 1;
3394
3395                         for (i = 0; i < chunk; i++) {
3396                                 kiblnd_check_conns(peer_index);
3397                                 peer_index = (peer_index + 1) %
3398                                              kiblnd_data.kib_peer_hash_size;
3399                         }
3400
3401                         deadline += msecs_to_jiffies(p * MSEC_PER_SEC);
3402                         spin_lock_irqsave(lock, flags);
3403                 }
3404
3405                 if (dropped_lock)
3406                         continue;
3407
3408                 /* Nothing to do for 'timeout'  */
3409                 set_current_state(TASK_INTERRUPTIBLE);
3410                 add_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3411                 spin_unlock_irqrestore(lock, flags);
3412
3413                 schedule_timeout(timeout);
3414
3415                 remove_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3416                 spin_lock_irqsave(lock, flags);
3417         }
3418
3419         spin_unlock_irqrestore(lock, flags);
3420
3421         kiblnd_thread_fini();
3422         return 0;
3423 }
3424
3425 void
3426 kiblnd_qp_event(struct ib_event *event, void *arg)
3427 {
3428         struct kib_conn *conn = arg;
3429
3430         switch (event->event) {
3431         case IB_EVENT_COMM_EST:
3432                 CDEBUG(D_NET, "%s established\n",
3433                        libcfs_nid2str(conn->ibc_peer->ibp_nid));
3434                 /*
3435                  * We received a packet but connection isn't established
3436                  * probably handshake packet was lost, so free to
3437                  * force make connection established
3438                  */
3439                 rdma_notify(conn->ibc_cmid, IB_EVENT_COMM_EST);
3440                 return;
3441
3442         default:
3443                 CERROR("%s: Async QP event type %d\n",
3444                        libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3445                 return;
3446         }
3447 }
3448
3449 static void
3450 kiblnd_complete(struct ib_wc *wc)
3451 {
3452         switch (kiblnd_wreqid2type(wc->wr_id)) {
3453         default:
3454                 LBUG();
3455
3456         case IBLND_WID_MR:
3457                 if (wc->status != IB_WC_SUCCESS &&
3458                     wc->status != IB_WC_WR_FLUSH_ERR)
3459                         CNETERR("FastReg failed: %d\n", wc->status);
3460                 break;
3461
3462         case IBLND_WID_RDMA:
3463                 /*
3464                  * We only get RDMA completion notification if it fails.  All
3465                  * subsequent work items, including the final SEND will fail
3466                  * too.  However we can't print out any more info about the
3467                  * failing RDMA because 'tx' might be back on the idle list or
3468                  * even reused already if we didn't manage to post all our work
3469                  * items
3470                  */
3471                 CNETERR("RDMA (tx: %p) failed: %d\n",
3472                         kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3473                 return;
3474
3475         case IBLND_WID_TX:
3476                 kiblnd_tx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3477                 return;
3478
3479         case IBLND_WID_RX:
3480                 kiblnd_rx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status,
3481                                    wc->byte_len);
3482                 return;
3483         }
3484 }
3485
3486 void
3487 kiblnd_cq_completion(struct ib_cq *cq, void *arg)
3488 {
3489         /*
3490          * NB I'm not allowed to schedule this conn once its refcount has
3491          * reached 0.  Since fundamentally I'm racing with scheduler threads
3492          * consuming my CQ I could be called after all completions have
3493          * occurred.  But in this case, !ibc_nrx && !ibc_nsends_posted
3494          * and this CQ is about to be destroyed so I NOOP.
3495          */
3496         struct kib_conn *conn = arg;
3497         struct kib_sched_info *sched = conn->ibc_sched;
3498         unsigned long flags;
3499
3500         LASSERT(cq == conn->ibc_cq);
3501
3502         spin_lock_irqsave(&sched->ibs_lock, flags);
3503
3504         conn->ibc_ready = 1;
3505
3506         if (!conn->ibc_scheduled &&
3507             (conn->ibc_nrx > 0 ||
3508              conn->ibc_nsends_posted > 0)) {
3509                 kiblnd_conn_addref(conn); /* +1 ref for sched_conns */
3510                 conn->ibc_scheduled = 1;
3511                 list_add_tail(&conn->ibc_sched_list, &sched->ibs_conns);
3512
3513                 if (waitqueue_active(&sched->ibs_waitq))
3514                         wake_up(&sched->ibs_waitq);
3515         }
3516
3517         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3518 }
3519
3520 void
3521 kiblnd_cq_event(struct ib_event *event, void *arg)
3522 {
3523         struct kib_conn *conn = arg;
3524
3525         CERROR("%s: async CQ event type %d\n",
3526                libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3527 }
3528
3529 int
3530 kiblnd_scheduler(void *arg)
3531 {
3532         long id = (long)arg;
3533         struct kib_sched_info *sched;
3534         struct kib_conn *conn;
3535         wait_queue_entry_t wait;
3536         unsigned long flags;
3537         struct ib_wc wc;
3538         int did_something;
3539         int busy_loops = 0;
3540         int rc;
3541
3542         cfs_block_allsigs();
3543
3544         init_waitqueue_entry(&wait, current);
3545
3546         sched = kiblnd_data.kib_scheds[KIB_THREAD_CPT(id)];
3547
3548         rc = cfs_cpt_bind(lnet_cpt_table(), sched->ibs_cpt);
3549         if (rc) {
3550                 CWARN("Unable to bind on CPU partition %d, please verify whether all CPUs are healthy and reload modules if necessary, otherwise your system might under risk of low performance\n",
3551                       sched->ibs_cpt);
3552         }
3553
3554         spin_lock_irqsave(&sched->ibs_lock, flags);
3555
3556         while (!kiblnd_data.kib_shutdown) {
3557                 if (busy_loops++ >= IBLND_RESCHED) {
3558                         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3559
3560                         cond_resched();
3561                         busy_loops = 0;
3562
3563                         spin_lock_irqsave(&sched->ibs_lock, flags);
3564                 }
3565
3566                 did_something = 0;
3567
3568                 if (!list_empty(&sched->ibs_conns)) {
3569                         conn = list_entry(sched->ibs_conns.next, struct kib_conn,
3570                                           ibc_sched_list);
3571                         /* take over kib_sched_conns' ref on conn... */
3572                         LASSERT(conn->ibc_scheduled);
3573                         list_del(&conn->ibc_sched_list);
3574                         conn->ibc_ready = 0;
3575
3576                         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3577
3578                         wc.wr_id = IBLND_WID_INVAL;
3579
3580                         rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3581                         if (!rc) {
3582                                 rc = ib_req_notify_cq(conn->ibc_cq,
3583                                                       IB_CQ_NEXT_COMP);
3584                                 if (rc < 0) {
3585                                         CWARN("%s: ib_req_notify_cq failed: %d, closing connection\n",
3586                                               libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
3587                                         kiblnd_close_conn(conn, -EIO);
3588                                         kiblnd_conn_decref(conn);
3589                                         spin_lock_irqsave(&sched->ibs_lock,
3590                                                           flags);
3591                                         continue;
3592                                 }
3593
3594                                 rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3595                         }
3596
3597                         if (unlikely(rc > 0 && wc.wr_id == IBLND_WID_INVAL)) {
3598                                 LCONSOLE_ERROR("ib_poll_cq (rc: %d) returned invalid wr_id, opcode %d, status: %d, vendor_err: %d, conn: %s status: %d\nplease upgrade firmware and OFED or contact vendor.\n",
3599                                                rc, wc.opcode, wc.status,
3600                                                wc.vendor_err,
3601                                                libcfs_nid2str(conn->ibc_peer->ibp_nid),
3602                                                conn->ibc_state);
3603                                 rc = -EINVAL;
3604                         }
3605
3606                         if (rc < 0) {
3607                                 CWARN("%s: ib_poll_cq failed: %d, closing connection\n",
3608                                       libcfs_nid2str(conn->ibc_peer->ibp_nid),
3609                                       rc);
3610                                 kiblnd_close_conn(conn, -EIO);
3611                                 kiblnd_conn_decref(conn);
3612                                 spin_lock_irqsave(&sched->ibs_lock, flags);
3613                                 continue;
3614                         }
3615
3616                         spin_lock_irqsave(&sched->ibs_lock, flags);
3617
3618                         if (rc || conn->ibc_ready) {
3619                                 /*
3620                                  * There may be another completion waiting; get
3621                                  * another scheduler to check while I handle
3622                                  * this one...
3623                                  */
3624                                 /* +1 ref for sched_conns */
3625                                 kiblnd_conn_addref(conn);
3626                                 list_add_tail(&conn->ibc_sched_list,
3627                                               &sched->ibs_conns);
3628                                 if (waitqueue_active(&sched->ibs_waitq))
3629                                         wake_up(&sched->ibs_waitq);
3630                         } else {
3631                                 conn->ibc_scheduled = 0;
3632                         }
3633
3634                         if (rc) {
3635                                 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3636                                 kiblnd_complete(&wc);
3637
3638                                 spin_lock_irqsave(&sched->ibs_lock, flags);
3639                         }
3640
3641                         kiblnd_conn_decref(conn); /* ...drop my ref from above */
3642                         did_something = 1;
3643                 }
3644
3645                 if (did_something)
3646                         continue;
3647
3648                 set_current_state(TASK_INTERRUPTIBLE);
3649                 add_wait_queue_exclusive(&sched->ibs_waitq, &wait);
3650                 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3651
3652                 schedule();
3653                 busy_loops = 0;
3654
3655                 remove_wait_queue(&sched->ibs_waitq, &wait);
3656                 spin_lock_irqsave(&sched->ibs_lock, flags);
3657         }
3658
3659         spin_unlock_irqrestore(&sched->ibs_lock, flags);
3660
3661         kiblnd_thread_fini();
3662         return 0;
3663 }
3664
3665 int
3666 kiblnd_failover_thread(void *arg)
3667 {
3668         rwlock_t *glock = &kiblnd_data.kib_global_lock;
3669         struct kib_dev *dev;
3670         wait_queue_entry_t wait;
3671         unsigned long flags;
3672         int rc;
3673
3674         LASSERT(*kiblnd_tunables.kib_dev_failover);
3675
3676         cfs_block_allsigs();
3677
3678         init_waitqueue_entry(&wait, current);
3679         write_lock_irqsave(glock, flags);
3680
3681         while (!kiblnd_data.kib_shutdown) {
3682                 int do_failover = 0;
3683                 int long_sleep;
3684
3685                 list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
3686                                     ibd_fail_list) {
3687                         if (time_before(cfs_time_current(),
3688                                         dev->ibd_next_failover))
3689                                 continue;
3690                         do_failover = 1;
3691                         break;
3692                 }
3693
3694                 if (do_failover) {
3695                         list_del_init(&dev->ibd_fail_list);
3696                         dev->ibd_failover = 1;
3697                         write_unlock_irqrestore(glock, flags);
3698
3699                         rc = kiblnd_dev_failover(dev);
3700
3701                         write_lock_irqsave(glock, flags);
3702
3703                         LASSERT(dev->ibd_failover);
3704                         dev->ibd_failover = 0;
3705                         if (rc >= 0) { /* Device is OK or failover succeed */
3706                                 dev->ibd_next_failover = cfs_time_shift(3);
3707                                 continue;
3708                         }
3709
3710                         /* failed to failover, retry later */
3711                         dev->ibd_next_failover =
3712                                 cfs_time_shift(min(dev->ibd_failed_failover, 10));
3713                         if (kiblnd_dev_can_failover(dev)) {
3714                                 list_add_tail(&dev->ibd_fail_list,
3715                                               &kiblnd_data.kib_failed_devs);
3716                         }
3717
3718                         continue;
3719                 }
3720
3721                 /* long sleep if no more pending failover */
3722                 long_sleep = list_empty(&kiblnd_data.kib_failed_devs);
3723
3724                 set_current_state(TASK_INTERRUPTIBLE);
3725                 add_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3726                 write_unlock_irqrestore(glock, flags);
3727
3728                 rc = schedule_timeout(long_sleep ? cfs_time_seconds(10) :
3729                                                    cfs_time_seconds(1));
3730                 remove_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3731                 write_lock_irqsave(glock, flags);
3732
3733                 if (!long_sleep || rc)
3734                         continue;
3735
3736                 /*
3737                  * have a long sleep, routine check all active devices,
3738                  * we need checking like this because if there is not active
3739                  * connection on the dev and no SEND from local, we may listen
3740                  * on wrong HCA for ever while there is a bonding failover
3741                  */
3742                 list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
3743                         if (kiblnd_dev_can_failover(dev)) {
3744                                 list_add_tail(&dev->ibd_fail_list,
3745                                               &kiblnd_data.kib_failed_devs);
3746                         }
3747                 }
3748         }
3749
3750         write_unlock_irqrestore(glock, flags);
3751
3752         kiblnd_thread_fini();
3753         return 0;
3754 }