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

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  19  *
  20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21  * CA 95054 USA or visit www.sun.com if you need additional information or
  22  * have any questions.
  23  *
  24  * GPL HEADER END
  25  */
  26 /*
  27  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2011, 2012, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  *
  36  * lnet/klnds/o2iblnd/o2iblnd.c
  37  *
  38  * Author: Eric Barton <eric@bartonsoftware.com>
  39  */
  40
  41 #include "o2iblnd.h"
  42 #include <asm/div64.h>
  43
  44 static lnd_t the_o2iblnd = {
  45         .lnd_type       = O2IBLND,
  46         .lnd_startup    = kiblnd_startup,
  47         .lnd_shutdown   = kiblnd_shutdown,
  48         .lnd_ctl        = kiblnd_ctl,
  49         .lnd_query      = kiblnd_query,
  50         .lnd_send       = kiblnd_send,
  51         .lnd_recv       = kiblnd_recv,
  52 };
  53
  54 kib_data_t            kiblnd_data;
  55
  56 static __u32
  57 kiblnd_cksum(void *ptr, int nob)
  58 {
  59         char  *c  = ptr;
  60         __u32  sum = 0;
  61
  62         while (nob-- > 0)
  63                 sum = ((sum << 1) | (sum >> 31)) + *c++;
  64
  65         /* ensure I don't return 0 (== no checksum) */
  66         return (sum == 0) ? 1 : sum;
  67 }
  68
  69 static char *
  70 kiblnd_msgtype2str(int type)
  71 {
  72         switch (type) {
  73         case IBLND_MSG_CONNREQ:
  74                 return "CONNREQ";
  75
  76         case IBLND_MSG_CONNACK:
  77                 return "CONNACK";
  78
  79         case IBLND_MSG_NOOP:
  80                 return "NOOP";
  81
  82         case IBLND_MSG_IMMEDIATE:
  83                 return "IMMEDIATE";
  84
  85         case IBLND_MSG_PUT_REQ:
  86                 return "PUT_REQ";
  87
  88         case IBLND_MSG_PUT_NAK:
  89                 return "PUT_NAK";
  90
  91         case IBLND_MSG_PUT_ACK:
  92                 return "PUT_ACK";
  93
  94         case IBLND_MSG_PUT_DONE:
  95                 return "PUT_DONE";
  96
  97         case IBLND_MSG_GET_REQ:
  98                 return "GET_REQ";
  99
 100         case IBLND_MSG_GET_DONE:
 101                 return "GET_DONE";
 102
 103         default:
 104                 return "???";
 105         }
 106 }
 107
 108 static int
 109 kiblnd_msgtype2size(int type)
 110 {
 111         const int hdr_size = offsetof(kib_msg_t, ibm_u);
 112
 113         switch (type) {
 114         case IBLND_MSG_CONNREQ:
 115         case IBLND_MSG_CONNACK:
 116                 return hdr_size + sizeof(kib_connparams_t);
 117
 118         case IBLND_MSG_NOOP:
 119                 return hdr_size;
 120
 121         case IBLND_MSG_IMMEDIATE:
 122                 return offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[0]);
 123
 124         case IBLND_MSG_PUT_REQ:
 125                 return hdr_size + sizeof(kib_putreq_msg_t);
 126
 127         case IBLND_MSG_PUT_ACK:
 128                 return hdr_size + sizeof(kib_putack_msg_t);
 129
 130         case IBLND_MSG_GET_REQ:
 131                 return hdr_size + sizeof(kib_get_msg_t);
 132
 133         case IBLND_MSG_PUT_NAK:
 134         case IBLND_MSG_PUT_DONE:
 135         case IBLND_MSG_GET_DONE:
 136                 return hdr_size + sizeof(kib_completion_msg_t);
 137         default:
 138                 return -1;
 139         }
 140 }
 141
 142 static int
 143 kiblnd_unpack_rd(kib_msg_t *msg, int flip)
 144 {
 145         kib_rdma_desc_t   *rd;
 146         int             nob;
 147         int             n;
 148         int             i;
 149
 150         LASSERT (msg->ibm_type == IBLND_MSG_GET_REQ ||
 151                  msg->ibm_type == IBLND_MSG_PUT_ACK);
 152
 153         rd = msg->ibm_type == IBLND_MSG_GET_REQ ?
 154                               &msg->ibm_u.get.ibgm_rd :
 155                               &msg->ibm_u.putack.ibpam_rd;
 156
 157         if (flip) {
 158                 __swab32s(&rd->rd_key);
 159                 __swab32s(&rd->rd_nfrags);
 160         }
 161
 162         n = rd->rd_nfrags;
 163
 164         if (n <= 0 || n > IBLND_MAX_RDMA_FRAGS) {
 165                 CERROR("Bad nfrags: %d, should be 0 < n <= %d\n",
 166                        n, IBLND_MAX_RDMA_FRAGS);
 167                 return 1;
 168         }
 169
 170         nob = offsetof (kib_msg_t, ibm_u) +
 171               kiblnd_rd_msg_size(rd, msg->ibm_type, n);
 172
 173         if (msg->ibm_nob < nob) {
 174                 CERROR("Short %s: %d(%d)\n",
 175                        kiblnd_msgtype2str(msg->ibm_type), msg->ibm_nob, nob);
 176                 return 1;
 177         }
 178
 179         if (!flip)
 180                 return 0;
 181
 182         for (i = 0; i < n; i++) {
 183                 __swab32s(&rd->rd_frags[i].rf_nob);
 184                 __swab64s(&rd->rd_frags[i].rf_addr);
 185         }
 186
 187         return 0;
 188 }
 189
 190 void
 191 kiblnd_pack_msg (lnet_ni_t *ni, kib_msg_t *msg, int version,
 192                  int credits, lnet_nid_t dstnid, __u64 dststamp)
 193 {
 194         kib_net_t *net = ni->ni_data;
 195
 196         /* CAVEAT EMPTOR! all message fields not set here should have been
 197          * initialised previously. */
 198         msg->ibm_magic    = IBLND_MSG_MAGIC;
 199         msg->ibm_version  = version;
 200         /*   ibm_type */
 201         msg->ibm_credits  = credits;
 202         /*   ibm_nob */
 203         msg->ibm_cksum    = 0;
 204         msg->ibm_srcnid   = ni->ni_nid;
 205         msg->ibm_srcstamp = net->ibn_incarnation;
 206         msg->ibm_dstnid   = dstnid;
 207         msg->ibm_dststamp = dststamp;
 208
 209         if (*kiblnd_tunables.kib_cksum) {
 210                 /* NB ibm_cksum zero while computing cksum */
 211                 msg->ibm_cksum = kiblnd_cksum(msg, msg->ibm_nob);
 212         }
 213 }
 214
 215 int
 216 kiblnd_unpack_msg(kib_msg_t *msg, int nob)
 217 {
 218         const int hdr_size = offsetof(kib_msg_t, ibm_u);
 219         __u32     msg_cksum;
 220         __u16     version;
 221         int       msg_nob;
 222         int       flip;
 223
 224         /* 6 bytes are enough to have received magic + version */
 225         if (nob < 6) {
 226                 CERROR("Short message: %d\n", nob);
 227                 return -EPROTO;
 228         }
 229
 230         if (msg->ibm_magic == IBLND_MSG_MAGIC) {
 231                 flip = 0;
 232         } else if (msg->ibm_magic == __swab32(IBLND_MSG_MAGIC)) {
 233                 flip = 1;
 234         } else {
 235                 CERROR("Bad magic: %08x\n", msg->ibm_magic);
 236                 return -EPROTO;
 237         }
 238
 239         version = flip ? __swab16(msg->ibm_version) : msg->ibm_version;
 240         if (version != IBLND_MSG_VERSION &&
 241             version != IBLND_MSG_VERSION_1) {
 242                 CERROR("Bad version: %x\n", version);
 243                 return -EPROTO;
 244         }
 245
 246         if (nob < hdr_size) {
 247                 CERROR("Short message: %d\n", nob);
 248                 return -EPROTO;
 249         }
 250
 251         msg_nob = flip ? __swab32(msg->ibm_nob) : msg->ibm_nob;
 252         if (msg_nob > nob) {
 253                 CERROR("Short message: got %d, wanted %d\n", nob, msg_nob);
 254                 return -EPROTO;
 255         }
 256
 257         /* checksum must be computed with ibm_cksum zero and BEFORE anything
 258          * gets flipped */
 259         msg_cksum = flip ? __swab32(msg->ibm_cksum) : msg->ibm_cksum;
 260         msg->ibm_cksum = 0;
 261         if (msg_cksum != 0 &&
 262             msg_cksum != kiblnd_cksum(msg, msg_nob)) {
 263                 CERROR("Bad checksum\n");
 264                 return -EPROTO;
 265         }
 266
 267         msg->ibm_cksum = msg_cksum;
 268
 269         if (flip) {
 270                 /* leave magic unflipped as a clue to peer endianness */
 271                 msg->ibm_version = version;
 272                 CLASSERT (sizeof(msg->ibm_type) == 1);
 273                 CLASSERT (sizeof(msg->ibm_credits) == 1);
 274                 msg->ibm_nob     = msg_nob;
 275                 __swab64s(&msg->ibm_srcnid);
 276                 __swab64s(&msg->ibm_srcstamp);
 277                 __swab64s(&msg->ibm_dstnid);
 278                 __swab64s(&msg->ibm_dststamp);
 279         }
 280
 281         if (msg->ibm_srcnid == LNET_NID_ANY) {
 282                 CERROR("Bad src nid: %s\n", libcfs_nid2str(msg->ibm_srcnid));
 283                 return -EPROTO;
 284         }
 285
 286         if (msg_nob < kiblnd_msgtype2size(msg->ibm_type)) {
 287                 CERROR("Short %s: %d(%d)\n", kiblnd_msgtype2str(msg->ibm_type),
 288                        msg_nob, kiblnd_msgtype2size(msg->ibm_type));
 289                 return -EPROTO;
 290         }
 291
 292         switch (msg->ibm_type) {
 293         default:
 294                 CERROR("Unknown message type %x\n", msg->ibm_type);
 295                 return -EPROTO;
 296
 297         case IBLND_MSG_NOOP:
 298         case IBLND_MSG_IMMEDIATE:
 299         case IBLND_MSG_PUT_REQ:
 300                 break;
 301
 302         case IBLND_MSG_PUT_ACK:
 303         case IBLND_MSG_GET_REQ:
 304                 if (kiblnd_unpack_rd(msg, flip))
 305                         return -EPROTO;
 306                 break;
 307
 308         case IBLND_MSG_PUT_NAK:
 309         case IBLND_MSG_PUT_DONE:
 310         case IBLND_MSG_GET_DONE:
 311                 if (flip)
 312                         __swab32s(&msg->ibm_u.completion.ibcm_status);
 313                 break;
 314
 315         case IBLND_MSG_CONNREQ:
 316         case IBLND_MSG_CONNACK:
 317                 if (flip) {
 318                         __swab16s(&msg->ibm_u.connparams.ibcp_queue_depth);
 319                         __swab16s(&msg->ibm_u.connparams.ibcp_max_frags);
 320                         __swab32s(&msg->ibm_u.connparams.ibcp_max_msg_size);
 321                 }
 322                 break;
 323         }
 324         return 0;
 325 }
 326
 327 int
 328 kiblnd_create_peer(lnet_ni_t *ni, kib_peer_t **peerp, lnet_nid_t nid)
 329 {
 330         kib_peer_t      *peer;
 331         kib_net_t       *net = ni->ni_data;
 332         int             cpt = lnet_cpt_of_nid(nid);
 333         unsigned long   flags;
 334
 335         LASSERT(net != NULL);
 336         LASSERT(nid != LNET_NID_ANY);
 337
 338         LIBCFS_CPT_ALLOC(peer, lnet_cpt_table(), cpt, sizeof(*peer));
 339         if (peer == NULL) {
 340                 CERROR("Cannot allocate peer\n");
 341                 return -ENOMEM;
 342         }
 343
 344         memset(peer, 0, sizeof(*peer));  /* zero flags etc */
 345
 346         peer->ibp_ni = ni;
 347         peer->ibp_nid = nid;
 348         peer->ibp_error = 0;
 349         peer->ibp_last_alive = 0;
 350         atomic_set(&peer->ibp_refcount, 1);  /* 1 ref for caller */
 351
 352         INIT_LIST_HEAD(&peer->ibp_list);     /* not in the peer table yet */
 353         INIT_LIST_HEAD(&peer->ibp_conns);
 354         INIT_LIST_HEAD(&peer->ibp_tx_queue);
 355
 356         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
 357
 358         /* always called with a ref on ni, which prevents ni being shutdown */
 359         LASSERT (net->ibn_shutdown == 0);
 360
 361         /* npeers only grows with the global lock held */
 362         atomic_inc(&net->ibn_npeers);
 363
 364         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
 365
 366         *peerp = peer;
 367         return 0;
 368 }
 369
 370 void
 371 kiblnd_destroy_peer (kib_peer_t *peer)
 372 {
 373         kib_net_t *net = peer->ibp_ni->ni_data;
 374
 375         LASSERT (net != NULL);
 376         LASSERT (atomic_read(&peer->ibp_refcount) == 0);
 377         LASSERT (!kiblnd_peer_active(peer));
 378         LASSERT (peer->ibp_connecting == 0);
 379         LASSERT (peer->ibp_accepting == 0);
 380         LASSERT (list_empty(&peer->ibp_conns));
 381         LASSERT (list_empty(&peer->ibp_tx_queue));
 382
 383         LIBCFS_FREE(peer, sizeof(*peer));
 384
 385         /* NB a peer's connections keep a reference on their peer until
 386          * they are destroyed, so we can be assured that _all_ state to do
 387          * with this peer has been cleaned up when its refcount drops to
 388          * zero. */
 389         atomic_dec(&net->ibn_npeers);
 390 }
 391
 392 kib_peer_t *
 393 kiblnd_find_peer_locked (lnet_nid_t nid)
 394 {
 395         /* the caller is responsible for accounting the additional reference
 396          * that this creates */
 397         struct list_head       *peer_list = kiblnd_nid2peerlist(nid);
 398         struct list_head       *tmp;
 399         kib_peer_t       *peer;
 400
 401         list_for_each (tmp, peer_list) {
 402
 403                 peer = list_entry(tmp, kib_peer_t, ibp_list);
 404
 405                 LASSERT (peer->ibp_connecting > 0 || /* creating conns */
 406                          peer->ibp_accepting > 0 ||
 407                          !list_empty(&peer->ibp_conns));  /* active conn */
 408
 409                 if (peer->ibp_nid != nid)
 410                         continue;
 411
 412                 CDEBUG(D_NET, "got peer [%p] -> %s (%d) version: %x\n",
 413                        peer, libcfs_nid2str(nid),
 414                        atomic_read(&peer->ibp_refcount),
 415                        peer->ibp_version);
 416                 return peer;
 417         }
 418         return NULL;
 419 }
 420
 421 void
 422 kiblnd_unlink_peer_locked (kib_peer_t *peer)
 423 {
 424         LASSERT (list_empty(&peer->ibp_conns));
 425
 426         LASSERT (kiblnd_peer_active(peer));
 427         list_del_init(&peer->ibp_list);
 428         /* lose peerlist's ref */
 429         kiblnd_peer_decref(peer);
 430 }
 431
 432 static int
 433 kiblnd_get_peer_info(lnet_ni_t *ni, int index,
 434                       lnet_nid_t *nidp, int *count)
 435 {
 436         kib_peer_t          *peer;
 437         struct list_head            *ptmp;
 438         int                 i;
 439         unsigned long     flags;
 440
 441         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
 442
 443         for (i = 0; i < kiblnd_data.kib_peer_hash_size; i++) {
 444
 445                 list_for_each (ptmp, &kiblnd_data.kib_peers[i]) {
 446
 447                         peer = list_entry(ptmp, kib_peer_t, ibp_list);
 448                         LASSERT (peer->ibp_connecting > 0 ||
 449                                  peer->ibp_accepting > 0 ||
 450                                  !list_empty(&peer->ibp_conns));
 451
 452                         if (peer->ibp_ni != ni)
 453                                 continue;
 454
 455                         if (index-- > 0)
 456                                 continue;
 457
 458                         *nidp = peer->ibp_nid;
 459                         *count = atomic_read(&peer->ibp_refcount);
 460
 461                         read_unlock_irqrestore(&kiblnd_data.kib_global_lock,
 462                                                flags);
 463                         return 0;
 464                 }
 465         }
 466
 467         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
 468         return -ENOENT;
 469 }
 470
 471 static void
 472 kiblnd_del_peer_locked(kib_peer_t *peer)
 473 {
 474         struct list_head           *ctmp;
 475         struct list_head           *cnxt;
 476         kib_conn_t         *conn;
 477
 478         if (list_empty(&peer->ibp_conns)) {
 479                 kiblnd_unlink_peer_locked(peer);
 480         } else {
 481                 list_for_each_safe (ctmp, cnxt, &peer->ibp_conns) {
 482                         conn = list_entry(ctmp, kib_conn_t, ibc_list);
 483
 484                         kiblnd_close_conn_locked(conn, 0);
 485                 }
 486                 /* NB closing peer's last conn unlinked it. */
 487         }
 488         /* NB peer now unlinked; might even be freed if the peer table had the
 489          * last ref on it. */
 490 }
 491
 492 static int
 493 kiblnd_del_peer(lnet_ni_t *ni, lnet_nid_t nid)
 494 {
 495         LIST_HEAD        (zombies);
 496         struct list_head            *ptmp;
 497         struct list_head            *pnxt;
 498         kib_peer_t          *peer;
 499         int                 lo;
 500         int                 hi;
 501         int                 i;
 502         unsigned long     flags;
 503         int                 rc = -ENOENT;
 504
 505         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
 506
 507         if (nid != LNET_NID_ANY) {
 508                 lo = hi = kiblnd_nid2peerlist(nid) - kiblnd_data.kib_peers;
 509         } else {
 510                 lo = 0;
 511                 hi = kiblnd_data.kib_peer_hash_size - 1;
 512         }
 513
 514         for (i = lo; i <= hi; i++) {
 515                 list_for_each_safe (ptmp, pnxt, &kiblnd_data.kib_peers[i]) {
 516                         peer = list_entry(ptmp, kib_peer_t, ibp_list);
 517                         LASSERT (peer->ibp_connecting > 0 ||
 518                                  peer->ibp_accepting > 0 ||
 519                                  !list_empty(&peer->ibp_conns));
 520
 521                         if (peer->ibp_ni != ni)
 522                                 continue;
 523
 524                         if (!(nid == LNET_NID_ANY || peer->ibp_nid == nid))
 525                                 continue;
 526
 527                         if (!list_empty(&peer->ibp_tx_queue)) {
 528                                 LASSERT (list_empty(&peer->ibp_conns));
 529
 530                                 list_splice_init(&peer->ibp_tx_queue,
 531                                                      &zombies);
 532                         }
 533
 534                         kiblnd_del_peer_locked(peer);
 535                         rc = 0;  /* matched something */
 536                 }
 537         }
 538
 539         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
 540
 541         kiblnd_txlist_done(ni, &zombies, -EIO);
 542
 543         return rc;
 544 }
 545
 546 static kib_conn_t *
 547 kiblnd_get_conn_by_idx(lnet_ni_t *ni, int index)
 548 {
 549         kib_peer_t          *peer;
 550         struct list_head            *ptmp;
 551         kib_conn_t          *conn;
 552         struct list_head            *ctmp;
 553         int                 i;
 554         unsigned long     flags;
 555
 556         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
 557
 558         for (i = 0; i < kiblnd_data.kib_peer_hash_size; i++) {
 559                 list_for_each (ptmp, &kiblnd_data.kib_peers[i]) {
 560
 561                         peer = list_entry(ptmp, kib_peer_t, ibp_list);
 562                         LASSERT (peer->ibp_connecting > 0 ||
 563                                  peer->ibp_accepting > 0 ||
 564                                  !list_empty(&peer->ibp_conns));
 565
 566                         if (peer->ibp_ni != ni)
 567                                 continue;
 568
 569                         list_for_each (ctmp, &peer->ibp_conns) {
 570                                 if (index-- > 0)
 571                                         continue;
 572
 573                                 conn = list_entry(ctmp, kib_conn_t,
 574                                                       ibc_list);
 575                                 kiblnd_conn_addref(conn);
 576                                 read_unlock_irqrestore(&kiblnd_data.kib_global_lock,
 577                                                        flags);
 578                                 return conn;
 579                         }
 580                 }
 581         }
 582
 583         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
 584         return NULL;
 585 }
 586
 587 int
 588 kiblnd_translate_mtu(int value)
 589 {
 590         switch (value) {
 591         default:
 592                 return -1;
 593         case 0:
 594                 return 0;
 595         case 256:
 596                 return IB_MTU_256;
 597         case 512:
 598                 return IB_MTU_512;
 599         case 1024:
 600                 return IB_MTU_1024;
 601         case 2048:
 602                 return IB_MTU_2048;
 603         case 4096:
 604                 return IB_MTU_4096;
 605         }
 606 }
 607
 608 static void
 609 kiblnd_setup_mtu_locked(struct rdma_cm_id *cmid)
 610 {
 611         int        mtu;
 612
 613         /* XXX There is no path record for iWARP, set by netdev->change_mtu? */
 614         if (cmid->route.path_rec == NULL)
 615                 return;
 616
 617         mtu = kiblnd_translate_mtu(*kiblnd_tunables.kib_ib_mtu);
 618         LASSERT (mtu >= 0);
 619         if (mtu != 0)
 620                 cmid->route.path_rec->mtu = mtu;
 621 }
 622
 623 static int
 624 kiblnd_get_completion_vector(kib_conn_t *conn, int cpt)
 625 {
 626         cpumask_t       *mask;
 627         int             vectors;
 628         int             off;
 629         int             i;
 630         lnet_nid_t      nid = conn->ibc_peer->ibp_nid;
 631
 632         vectors = conn->ibc_cmid->device->num_comp_vectors;
 633         if (vectors <= 1)
 634                 return 0;
 635
 636         mask = cfs_cpt_cpumask(lnet_cpt_table(), cpt);
 637         if (mask == NULL)
 638                 return 0;
 639
 640         /* hash NID to CPU id in this partition... */
 641         off = do_div(nid, cpus_weight(*mask));
 642         for_each_cpu_mask(i, *mask) {
 643                 if (off-- == 0)
 644                         return i % vectors;
 645         }
 646
 647         LBUG();
 648         return 1;
 649 }
 650
 651 kib_conn_t *
 652 kiblnd_create_conn(kib_peer_t *peer, struct rdma_cm_id *cmid,
 653                    int state, int version)
 654 {
 655         /* CAVEAT EMPTOR:
 656          * If the new conn is created successfully it takes over the caller's
 657          * ref on 'peer'.  It also "owns" 'cmid' and destroys it when it itself
 658          * is destroyed.  On failure, the caller's ref on 'peer' remains and
 659          * she must dispose of 'cmid'.  (Actually I'd block forever if I tried
 660          * to destroy 'cmid' here since I'm called from the CM which still has
 661          * its ref on 'cmid'). */
 662         rwlock_t                *glock = &kiblnd_data.kib_global_lock;
 663         kib_net_t             *net = peer->ibp_ni->ni_data;
 664         kib_dev_t             *dev;
 665         struct ib_qp_init_attr *init_qp_attr;
 666         struct kib_sched_info   *sched;
 667         kib_conn_t              *conn;
 668         struct ib_cq            *cq;
 669         unsigned long           flags;
 670         int                     cpt;
 671         int                     rc;
 672         int                     i;
 673
 674         LASSERT(net != NULL);
 675         LASSERT(!in_interrupt());
 676
 677         dev = net->ibn_dev;
 678
 679         cpt = lnet_cpt_of_nid(peer->ibp_nid);
 680         sched = kiblnd_data.kib_scheds[cpt];
 681
 682         LASSERT(sched->ibs_nthreads > 0);
 683
 684         LIBCFS_CPT_ALLOC(init_qp_attr, lnet_cpt_table(), cpt,
 685                          sizeof(*init_qp_attr));
 686         if (init_qp_attr == NULL) {
 687                 CERROR("Can't allocate qp_attr for %s\n",
 688                        libcfs_nid2str(peer->ibp_nid));
 689                 goto failed_0;
 690         }
 691
 692         LIBCFS_CPT_ALLOC(conn, lnet_cpt_table(), cpt, sizeof(*conn));
 693         if (conn == NULL) {
 694                 CERROR("Can't allocate connection for %s\n",
 695                        libcfs_nid2str(peer->ibp_nid));
 696                 goto failed_1;
 697         }
 698
 699         conn->ibc_state = IBLND_CONN_INIT;
 700         conn->ibc_version = version;
 701         conn->ibc_peer = peer;            /* I take the caller's ref */
 702         cmid->context = conn;              /* for future CM callbacks */
 703         conn->ibc_cmid = cmid;
 704
 705         INIT_LIST_HEAD(&conn->ibc_early_rxs);
 706         INIT_LIST_HEAD(&conn->ibc_tx_noops);
 707         INIT_LIST_HEAD(&conn->ibc_tx_queue);
 708         INIT_LIST_HEAD(&conn->ibc_tx_queue_rsrvd);
 709         INIT_LIST_HEAD(&conn->ibc_tx_queue_nocred);
 710         INIT_LIST_HEAD(&conn->ibc_active_txs);
 711         spin_lock_init(&conn->ibc_lock);
 712
 713         LIBCFS_CPT_ALLOC(conn->ibc_connvars, lnet_cpt_table(), cpt,
 714                          sizeof(*conn->ibc_connvars));
 715         if (conn->ibc_connvars == NULL) {
 716                 CERROR("Can't allocate in-progress connection state\n");
 717                 goto failed_2;
 718         }
 719
 720         write_lock_irqsave(glock, flags);
 721         if (dev->ibd_failover) {
 722                 write_unlock_irqrestore(glock, flags);
 723                 CERROR("%s: failover in progress\n", dev->ibd_ifname);
 724                 goto failed_2;
 725         }
 726
 727         if (dev->ibd_hdev->ibh_ibdev != cmid->device) {
 728                 /* wakeup failover thread and teardown connection */
 729                 if (kiblnd_dev_can_failover(dev)) {
 730                         list_add_tail(&dev->ibd_fail_list,
 731                                       &kiblnd_data.kib_failed_devs);
 732                         wake_up(&kiblnd_data.kib_failover_waitq);
 733                 }
 734
 735                 write_unlock_irqrestore(glock, flags);
 736                 CERROR("cmid HCA(%s), kib_dev(%s) need failover\n",
 737                        cmid->device->name, dev->ibd_ifname);
 738                 goto failed_2;
 739         }
 740
 741         kiblnd_hdev_addref_locked(dev->ibd_hdev);
 742         conn->ibc_hdev = dev->ibd_hdev;
 743
 744         kiblnd_setup_mtu_locked(cmid);
 745
 746         write_unlock_irqrestore(glock, flags);
 747
 748         LIBCFS_CPT_ALLOC(conn->ibc_rxs, lnet_cpt_table(), cpt,
 749                          IBLND_RX_MSGS(version) * sizeof(kib_rx_t));
 750         if (conn->ibc_rxs == NULL) {
 751                 CERROR("Cannot allocate RX buffers\n");
 752                 goto failed_2;
 753         }
 754
 755         rc = kiblnd_alloc_pages(&conn->ibc_rx_pages, cpt,
 756                                 IBLND_RX_MSG_PAGES(version));
 757         if (rc != 0)
 758                 goto failed_2;
 759
 760         kiblnd_map_rx_descs(conn);
 761
 762         cq = ib_create_cq(cmid->device,
 763                           kiblnd_cq_completion, kiblnd_cq_event, conn,
 764                           IBLND_CQ_ENTRIES(version),
 765                           kiblnd_get_completion_vector(conn, cpt));
 766         if (IS_ERR(cq)) {
 767                 CERROR("Can't create CQ: %ld, cqe: %d\n",
 768                        PTR_ERR(cq), IBLND_CQ_ENTRIES(version));
 769                 goto failed_2;
 770         }
 771
 772         conn->ibc_cq = cq;
 773
 774         rc = ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
 775         if (rc != 0) {
 776                 CERROR("Can't request completion notificiation: %d\n", rc);
 777                 goto failed_2;
 778         }
 779
 780         init_qp_attr->event_handler = kiblnd_qp_event;
 781         init_qp_attr->qp_context = conn;
 782         init_qp_attr->cap.max_send_wr = IBLND_SEND_WRS(version);
 783         init_qp_attr->cap.max_recv_wr = IBLND_RECV_WRS(version);
 784         init_qp_attr->cap.max_send_sge = 1;
 785         init_qp_attr->cap.max_recv_sge = 1;
 786         init_qp_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
 787         init_qp_attr->qp_type = IB_QPT_RC;
 788         init_qp_attr->send_cq = cq;
 789         init_qp_attr->recv_cq = cq;
 790
 791         conn->ibc_sched = sched;
 792
 793         rc = rdma_create_qp(cmid, conn->ibc_hdev->ibh_pd, init_qp_attr);
 794         if (rc != 0) {
 795                 CERROR("Can't create QP: %d, send_wr: %d, recv_wr: %d\n",
 796                        rc, init_qp_attr->cap.max_send_wr,
 797                        init_qp_attr->cap.max_recv_wr);
 798                 goto failed_2;
 799         }
 800
 801         LIBCFS_FREE(init_qp_attr, sizeof(*init_qp_attr));
 802
 803         /* 1 ref for caller and each rxmsg */
 804         atomic_set(&conn->ibc_refcount, 1 + IBLND_RX_MSGS(version));
 805         conn->ibc_nrx = IBLND_RX_MSGS(version);
 806
 807         /* post receives */
 808         for (i = 0; i < IBLND_RX_MSGS(version); i++) {
 809                 rc = kiblnd_post_rx(&conn->ibc_rxs[i],
 810                                     IBLND_POSTRX_NO_CREDIT);
 811                 if (rc != 0) {
 812                         CERROR("Can't post rxmsg: %d\n", rc);
 813
 814                         /* Make posted receives complete */
 815                         kiblnd_abort_receives(conn);
 816
 817                         /* correct # of posted buffers
 818                          * NB locking needed now I'm racing with completion */
 819                         spin_lock_irqsave(&sched->ibs_lock, flags);
 820                         conn->ibc_nrx -= IBLND_RX_MSGS(version) - i;
 821                         spin_unlock_irqrestore(&sched->ibs_lock, flags);
 822
 823                         /* cmid will be destroyed by CM(ofed) after cm_callback
 824                          * returned, so we can't refer it anymore
 825                          * (by kiblnd_connd()->kiblnd_destroy_conn) */
 826                         rdma_destroy_qp(conn->ibc_cmid);
 827                         conn->ibc_cmid = NULL;
 828
 829                         /* Drop my own and unused rxbuffer refcounts */
 830                         while (i++ <= IBLND_RX_MSGS(version))
 831                                 kiblnd_conn_decref(conn);
 832
 833                         return NULL;
 834                 }
 835         }
 836
 837         /* Init successful! */
 838         LASSERT (state == IBLND_CONN_ACTIVE_CONNECT ||
 839                  state == IBLND_CONN_PASSIVE_WAIT);
 840         conn->ibc_state = state;
 841
 842         /* 1 more conn */
 843         atomic_inc(&net->ibn_nconns);
 844         return conn;
 845
 846  failed_2:
 847         kiblnd_destroy_conn(conn);
 848  failed_1:
 849         LIBCFS_FREE(init_qp_attr, sizeof(*init_qp_attr));
 850  failed_0:
 851         return NULL;
 852 }
 853
 854 void
 855 kiblnd_destroy_conn (kib_conn_t *conn)
 856 {
 857         struct rdma_cm_id *cmid = conn->ibc_cmid;
 858         kib_peer_t      *peer = conn->ibc_peer;
 859         int             rc;
 860
 861         LASSERT (!in_interrupt());
 862         LASSERT (atomic_read(&conn->ibc_refcount) == 0);
 863         LASSERT (list_empty(&conn->ibc_early_rxs));
 864         LASSERT (list_empty(&conn->ibc_tx_noops));
 865         LASSERT (list_empty(&conn->ibc_tx_queue));
 866         LASSERT (list_empty(&conn->ibc_tx_queue_rsrvd));
 867         LASSERT (list_empty(&conn->ibc_tx_queue_nocred));
 868         LASSERT (list_empty(&conn->ibc_active_txs));
 869         LASSERT (conn->ibc_noops_posted == 0);
 870         LASSERT (conn->ibc_nsends_posted == 0);
 871
 872         switch (conn->ibc_state) {
 873         default:
 874                 /* conn must be completely disengaged from the network */
 875                 LBUG();
 876
 877         case IBLND_CONN_DISCONNECTED:
 878                 /* connvars should have been freed already */
 879                 LASSERT (conn->ibc_connvars == NULL);
 880                 break;
 881
 882         case IBLND_CONN_INIT:
 883                 break;
 884         }
 885
 886         /* conn->ibc_cmid might be destroyed by CM already */
 887         if (cmid != NULL && cmid->qp != NULL)
 888                 rdma_destroy_qp(cmid);
 889
 890         if (conn->ibc_cq != NULL) {
 891                 rc = ib_destroy_cq(conn->ibc_cq);
 892                 if (rc != 0)
 893                         CWARN("Error destroying CQ: %d\n", rc);
 894         }
 895
 896         if (conn->ibc_rx_pages != NULL)
 897                 kiblnd_unmap_rx_descs(conn);
 898
 899         if (conn->ibc_rxs != NULL) {
 900                 LIBCFS_FREE(conn->ibc_rxs,
 901                             IBLND_RX_MSGS(conn->ibc_version) * sizeof(kib_rx_t));
 902         }
 903
 904         if (conn->ibc_connvars != NULL)
 905                 LIBCFS_FREE(conn->ibc_connvars, sizeof(*conn->ibc_connvars));
 906
 907         if (conn->ibc_hdev != NULL)
 908                 kiblnd_hdev_decref(conn->ibc_hdev);
 909
 910         /* See CAVEAT EMPTOR above in kiblnd_create_conn */
 911         if (conn->ibc_state != IBLND_CONN_INIT) {
 912                 kib_net_t *net = peer->ibp_ni->ni_data;
 913
 914                 kiblnd_peer_decref(peer);
 915                 rdma_destroy_id(cmid);
 916                 atomic_dec(&net->ibn_nconns);
 917         }
 918
 919         LIBCFS_FREE(conn, sizeof(*conn));
 920 }
 921
 922 int
 923 kiblnd_close_peer_conns_locked (kib_peer_t *peer, int why)
 924 {
 925         kib_conn_t           *conn;
 926         struct list_head             *ctmp;
 927         struct list_head             *cnxt;
 928         int                  count = 0;
 929
 930         list_for_each_safe (ctmp, cnxt, &peer->ibp_conns) {
 931                 conn = list_entry(ctmp, kib_conn_t, ibc_list);
 932
 933                 CDEBUG(D_NET, "Closing conn -> %s, "
 934                               "version: %x, reason: %d\n",
 935                        libcfs_nid2str(peer->ibp_nid),
 936                        conn->ibc_version, why);
 937
 938                 kiblnd_close_conn_locked(conn, why);
 939                 count++;
 940         }
 941
 942         return count;
 943 }
 944
 945 int
 946 kiblnd_close_stale_conns_locked (kib_peer_t *peer,
 947                                  int version, __u64 incarnation)
 948 {
 949         kib_conn_t           *conn;
 950         struct list_head             *ctmp;
 951         struct list_head             *cnxt;
 952         int                  count = 0;
 953
 954         list_for_each_safe (ctmp, cnxt, &peer->ibp_conns) {
 955                 conn = list_entry(ctmp, kib_conn_t, ibc_list);
 956
 957                 if (conn->ibc_version     == version &&
 958                     conn->ibc_incarnation == incarnation)
 959                         continue;
 960
 961                 CDEBUG(D_NET, "Closing stale conn -> %s version: %x, "
 962                               "incarnation:%#llx(%x, %#llx)\n",
 963                        libcfs_nid2str(peer->ibp_nid),
 964                        conn->ibc_version, conn->ibc_incarnation,
 965                        version, incarnation);
 966
 967                 kiblnd_close_conn_locked(conn, -ESTALE);
 968                 count++;
 969         }
 970
 971         return count;
 972 }
 973
 974 static int
 975 kiblnd_close_matching_conns(lnet_ni_t *ni, lnet_nid_t nid)
 976 {
 977         kib_peer_t           *peer;
 978         struct list_head             *ptmp;
 979         struct list_head             *pnxt;
 980         int                  lo;
 981         int                  hi;
 982         int                  i;
 983         unsigned long      flags;
 984         int                  count = 0;
 985
 986         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
 987
 988         if (nid != LNET_NID_ANY)
 989                 lo = hi = kiblnd_nid2peerlist(nid) - kiblnd_data.kib_peers;
 990         else {
 991                 lo = 0;
 992                 hi = kiblnd_data.kib_peer_hash_size - 1;
 993         }
 994
 995         for (i = lo; i <= hi; i++) {
 996                 list_for_each_safe (ptmp, pnxt, &kiblnd_data.kib_peers[i]) {
 997
 998                         peer = list_entry(ptmp, kib_peer_t, ibp_list);
 999                         LASSERT (peer->ibp_connecting > 0 ||
1000                                  peer->ibp_accepting > 0 ||
1001                                  !list_empty(&peer->ibp_conns));
1002
1003                         if (peer->ibp_ni != ni)
1004                                 continue;
1005
1006                         if (!(nid == LNET_NID_ANY || nid == peer->ibp_nid))
1007                                 continue;
1008
1009                         count += kiblnd_close_peer_conns_locked(peer, 0);
1010                 }
1011         }
1012
1013         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1014
1015         /* wildcards always succeed */
1016         if (nid == LNET_NID_ANY)
1017                 return 0;
1018
1019         return (count == 0) ? -ENOENT : 0;
1020 }
1021
1022 int
1023 kiblnd_ctl(lnet_ni_t *ni, unsigned int cmd, void *arg)
1024 {
1025         struct libcfs_ioctl_data *data = arg;
1026         int                    rc = -EINVAL;
1027
1028         switch (cmd) {
1029         case IOC_LIBCFS_GET_PEER: {
1030                 lnet_nid_t   nid = 0;
1031                 int       count = 0;
1032
1033                 rc = kiblnd_get_peer_info(ni, data->ioc_count,
1034                                           &nid, &count);
1035                 data->ioc_nid    = nid;
1036                 data->ioc_count  = count;
1037                 break;
1038         }
1039
1040         case IOC_LIBCFS_DEL_PEER: {
1041                 rc = kiblnd_del_peer(ni, data->ioc_nid);
1042                 break;
1043         }
1044         case IOC_LIBCFS_GET_CONN: {
1045                 kib_conn_t *conn;
1046
1047                 rc = 0;
1048                 conn = kiblnd_get_conn_by_idx(ni, data->ioc_count);
1049                 if (conn == NULL) {
1050                         rc = -ENOENT;
1051                         break;
1052                 }
1053
1054                 LASSERT (conn->ibc_cmid != NULL);
1055                 data->ioc_nid = conn->ibc_peer->ibp_nid;
1056                 if (conn->ibc_cmid->route.path_rec == NULL)
1057                         data->ioc_u32[0] = 0; /* iWarp has no path MTU */
1058                 else
1059                         data->ioc_u32[0] =
1060                         ib_mtu_enum_to_int(conn->ibc_cmid->route.path_rec->mtu);
1061                 kiblnd_conn_decref(conn);
1062                 break;
1063         }
1064         case IOC_LIBCFS_CLOSE_CONNECTION: {
1065                 rc = kiblnd_close_matching_conns(ni, data->ioc_nid);
1066                 break;
1067         }
1068
1069         default:
1070                 break;
1071         }
1072
1073         return rc;
1074 }
1075
1076 void
1077 kiblnd_query (lnet_ni_t *ni, lnet_nid_t nid, unsigned long *when)
1078 {
1079         unsigned long   last_alive = 0;
1080         unsigned long   now = cfs_time_current();
1081         rwlock_t        *glock = &kiblnd_data.kib_global_lock;
1082         kib_peer_t      *peer;
1083         unsigned long   flags;
1084
1085         read_lock_irqsave(glock, flags);
1086
1087         peer = kiblnd_find_peer_locked(nid);
1088         if (peer != NULL) {
1089                 LASSERT (peer->ibp_connecting > 0 || /* creating conns */
1090                          peer->ibp_accepting > 0 ||
1091                          !list_empty(&peer->ibp_conns));  /* active conn */
1092                 last_alive = peer->ibp_last_alive;
1093         }
1094
1095         read_unlock_irqrestore(glock, flags);
1096
1097         if (last_alive != 0)
1098                 *when = last_alive;
1099
1100         /* peer is not persistent in hash, trigger peer creation
1101          * and connection establishment with a NULL tx */
1102         if (peer == NULL)
1103                 kiblnd_launch_tx(ni, NULL, nid);
1104
1105         CDEBUG(D_NET, "Peer %s %p, alive %ld secs ago\n",
1106                libcfs_nid2str(nid), peer,
1107                last_alive ? cfs_duration_sec(now - last_alive) : -1);
1108         return;
1109 }
1110
1111 void
1112 kiblnd_free_pages(kib_pages_t *p)
1113 {
1114         int     npages = p->ibp_npages;
1115         int     i;
1116
1117         for (i = 0; i < npages; i++) {
1118                 if (p->ibp_pages[i] != NULL)
1119                         __free_page(p->ibp_pages[i]);
1120         }
1121
1122         LIBCFS_FREE(p, offsetof(kib_pages_t, ibp_pages[npages]));
1123 }
1124
1125 int
1126 kiblnd_alloc_pages(kib_pages_t **pp, int cpt, int npages)
1127 {
1128         kib_pages_t     *p;
1129         int             i;
1130
1131         LIBCFS_CPT_ALLOC(p, lnet_cpt_table(), cpt,
1132                          offsetof(kib_pages_t, ibp_pages[npages]));
1133         if (p == NULL) {
1134                 CERROR("Can't allocate descriptor for %d pages\n", npages);
1135                 return -ENOMEM;
1136         }
1137
1138         memset(p, 0, offsetof(kib_pages_t, ibp_pages[npages]));
1139         p->ibp_npages = npages;
1140
1141         for (i = 0; i < npages; i++) {
1142                 p->ibp_pages[i] = alloc_pages_node(
1143                                     cfs_cpt_spread_node(lnet_cpt_table(), cpt),
1144                                     GFP_NOFS, 0);
1145                 if (p->ibp_pages[i] == NULL) {
1146                         CERROR("Can't allocate page %d of %d\n", i, npages);
1147                         kiblnd_free_pages(p);
1148                         return -ENOMEM;
1149                 }
1150         }
1151
1152         *pp = p;
1153         return 0;
1154 }
1155
1156 void
1157 kiblnd_unmap_rx_descs(kib_conn_t *conn)
1158 {
1159         kib_rx_t *rx;
1160         int       i;
1161
1162         LASSERT (conn->ibc_rxs != NULL);
1163         LASSERT (conn->ibc_hdev != NULL);
1164
1165         for (i = 0; i < IBLND_RX_MSGS(conn->ibc_version); i++) {
1166                 rx = &conn->ibc_rxs[i];
1167
1168                 LASSERT (rx->rx_nob >= 0); /* not posted */
1169
1170                 kiblnd_dma_unmap_single(conn->ibc_hdev->ibh_ibdev,
1171                                         KIBLND_UNMAP_ADDR(rx, rx_msgunmap,
1172                                                           rx->rx_msgaddr),
1173                                         IBLND_MSG_SIZE, DMA_FROM_DEVICE);
1174         }
1175
1176         kiblnd_free_pages(conn->ibc_rx_pages);
1177
1178         conn->ibc_rx_pages = NULL;
1179 }
1180
1181 void
1182 kiblnd_map_rx_descs(kib_conn_t *conn)
1183 {
1184         kib_rx_t       *rx;
1185         struct page    *pg;
1186         int          pg_off;
1187         int          ipg;
1188         int          i;
1189
1190         for (pg_off = ipg = i = 0;
1191              i < IBLND_RX_MSGS(conn->ibc_version); i++) {
1192                 pg = conn->ibc_rx_pages->ibp_pages[ipg];
1193                 rx = &conn->ibc_rxs[i];
1194
1195                 rx->rx_conn = conn;
1196                 rx->rx_msg = (kib_msg_t *)(((char *)page_address(pg)) + pg_off);
1197
1198                 rx->rx_msgaddr = kiblnd_dma_map_single(conn->ibc_hdev->ibh_ibdev,
1199                                                        rx->rx_msg, IBLND_MSG_SIZE,
1200                                                        DMA_FROM_DEVICE);
1201                 LASSERT (!kiblnd_dma_mapping_error(conn->ibc_hdev->ibh_ibdev,
1202                                                    rx->rx_msgaddr));
1203                 KIBLND_UNMAP_ADDR_SET(rx, rx_msgunmap, rx->rx_msgaddr);
1204
1205                 CDEBUG(D_NET, "rx %d: %p %#llx(%#llx)\n",
1206                        i, rx->rx_msg, rx->rx_msgaddr,
1207                        lnet_page2phys(pg) + pg_off);
1208
1209                 pg_off += IBLND_MSG_SIZE;
1210                 LASSERT (pg_off <= PAGE_SIZE);
1211
1212                 if (pg_off == PAGE_SIZE) {
1213                         pg_off = 0;
1214                         ipg++;
1215                         LASSERT (ipg <= IBLND_RX_MSG_PAGES(conn->ibc_version));
1216                 }
1217         }
1218 }
1219
1220 static void
1221 kiblnd_unmap_tx_pool(kib_tx_pool_t *tpo)
1222 {
1223         kib_hca_dev_t  *hdev = tpo->tpo_hdev;
1224         kib_tx_t       *tx;
1225         int          i;
1226
1227         LASSERT (tpo->tpo_pool.po_allocated == 0);
1228
1229         if (hdev == NULL)
1230                 return;
1231
1232         for (i = 0; i < tpo->tpo_pool.po_size; i++) {
1233                 tx = &tpo->tpo_tx_descs[i];
1234                 kiblnd_dma_unmap_single(hdev->ibh_ibdev,
1235                                         KIBLND_UNMAP_ADDR(tx, tx_msgunmap,
1236                                                           tx->tx_msgaddr),
1237                                         IBLND_MSG_SIZE, DMA_TO_DEVICE);
1238         }
1239
1240         kiblnd_hdev_decref(hdev);
1241         tpo->tpo_hdev = NULL;
1242 }
1243
1244 static kib_hca_dev_t *
1245 kiblnd_current_hdev(kib_dev_t *dev)
1246 {
1247         kib_hca_dev_t *hdev;
1248         unsigned long  flags;
1249         int         i = 0;
1250
1251         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1252         while (dev->ibd_failover) {
1253                 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1254                 if (i++ % 50 == 0)
1255                         CDEBUG(D_NET, "%s: Wait for failover\n",
1256                                dev->ibd_ifname);
1257                 schedule_timeout(cfs_time_seconds(1) / 100);
1258
1259                 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1260         }
1261
1262         kiblnd_hdev_addref_locked(dev->ibd_hdev);
1263         hdev = dev->ibd_hdev;
1264
1265         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1266
1267         return hdev;
1268 }
1269
1270 static void
1271 kiblnd_map_tx_pool(kib_tx_pool_t *tpo)
1272 {
1273         kib_pages_t    *txpgs = tpo->tpo_tx_pages;
1274         kib_pool_t     *pool  = &tpo->tpo_pool;
1275         kib_net_t      *net   = pool->po_owner->ps_net;
1276         kib_dev_t      *dev;
1277         struct page    *page;
1278         kib_tx_t       *tx;
1279         int          page_offset;
1280         int          ipage;
1281         int          i;
1282
1283         LASSERT (net != NULL);
1284
1285         dev = net->ibn_dev;
1286
1287         /* pre-mapped messages are not bigger than 1 page */
1288         CLASSERT (IBLND_MSG_SIZE <= PAGE_SIZE);
1289
1290         /* No fancy arithmetic when we do the buffer calculations */
1291         CLASSERT (PAGE_SIZE % IBLND_MSG_SIZE == 0);
1292
1293         tpo->tpo_hdev = kiblnd_current_hdev(dev);
1294
1295         for (ipage = page_offset = i = 0; i < pool->po_size; i++) {
1296                 page = txpgs->ibp_pages[ipage];
1297                 tx = &tpo->tpo_tx_descs[i];
1298
1299                 tx->tx_msg = (kib_msg_t *)(((char *)page_address(page)) +
1300                                            page_offset);
1301
1302                 tx->tx_msgaddr = kiblnd_dma_map_single(
1303                         tpo->tpo_hdev->ibh_ibdev, tx->tx_msg,
1304                         IBLND_MSG_SIZE, DMA_TO_DEVICE);
1305                 LASSERT (!kiblnd_dma_mapping_error(tpo->tpo_hdev->ibh_ibdev,
1306                                                    tx->tx_msgaddr));
1307                 KIBLND_UNMAP_ADDR_SET(tx, tx_msgunmap, tx->tx_msgaddr);
1308
1309                 list_add(&tx->tx_list, &pool->po_free_list);
1310
1311                 page_offset += IBLND_MSG_SIZE;
1312                 LASSERT (page_offset <= PAGE_SIZE);
1313
1314                 if (page_offset == PAGE_SIZE) {
1315                         page_offset = 0;
1316                         ipage++;
1317                         LASSERT (ipage <= txpgs->ibp_npages);
1318                 }
1319         }
1320 }
1321
1322 struct ib_mr *
1323 kiblnd_find_dma_mr(kib_hca_dev_t *hdev, __u64 addr, __u64 size)
1324 {
1325         __u64   index;
1326
1327         LASSERT (hdev->ibh_mrs[0] != NULL);
1328
1329         if (hdev->ibh_nmrs == 1)
1330                 return hdev->ibh_mrs[0];
1331
1332         index = addr >> hdev->ibh_mr_shift;
1333
1334         if (index <  hdev->ibh_nmrs &&
1335             index == ((addr + size - 1) >> hdev->ibh_mr_shift))
1336                 return hdev->ibh_mrs[index];
1337
1338         return NULL;
1339 }
1340
1341 struct ib_mr *
1342 kiblnd_find_rd_dma_mr(kib_hca_dev_t *hdev, kib_rdma_desc_t *rd)
1343 {
1344         struct ib_mr *prev_mr;
1345         struct ib_mr *mr;
1346         int        i;
1347
1348         LASSERT (hdev->ibh_mrs[0] != NULL);
1349
1350         if (*kiblnd_tunables.kib_map_on_demand > 0 &&
1351             *kiblnd_tunables.kib_map_on_demand <= rd->rd_nfrags)
1352                 return NULL;
1353
1354         if (hdev->ibh_nmrs == 1)
1355                 return hdev->ibh_mrs[0];
1356
1357         for (i = 0, mr = prev_mr = NULL;
1358              i < rd->rd_nfrags; i++) {
1359                 mr = kiblnd_find_dma_mr(hdev,
1360                                         rd->rd_frags[i].rf_addr,
1361                                         rd->rd_frags[i].rf_nob);
1362                 if (prev_mr == NULL)
1363                         prev_mr = mr;
1364
1365                 if (mr == NULL || prev_mr != mr) {
1366                         /* Can't covered by one single MR */
1367                         mr = NULL;
1368                         break;
1369                 }
1370         }
1371
1372         return mr;
1373 }
1374
1375 static void
1376 kiblnd_destroy_fmr_pool(kib_fmr_pool_t *pool)
1377 {
1378         LASSERT (pool->fpo_map_count == 0);
1379
1380         if (pool->fpo_fmr_pool != NULL)
1381                 ib_destroy_fmr_pool(pool->fpo_fmr_pool);
1382
1383         if (pool->fpo_hdev != NULL)
1384                 kiblnd_hdev_decref(pool->fpo_hdev);
1385
1386         LIBCFS_FREE(pool, sizeof(kib_fmr_pool_t));
1387 }
1388
1389 static void
1390 kiblnd_destroy_fmr_pool_list(struct list_head *head)
1391 {
1392         kib_fmr_pool_t *pool;
1393
1394         while (!list_empty(head)) {
1395                 pool = list_entry(head->next, kib_fmr_pool_t, fpo_list);
1396                 list_del(&pool->fpo_list);
1397                 kiblnd_destroy_fmr_pool(pool);
1398         }
1399 }
1400
1401 static int kiblnd_fmr_pool_size(int ncpts)
1402 {
1403         int size = *kiblnd_tunables.kib_fmr_pool_size / ncpts;
1404
1405         return max(IBLND_FMR_POOL, size);
1406 }
1407
1408 static int kiblnd_fmr_flush_trigger(int ncpts)
1409 {
1410         int size = *kiblnd_tunables.kib_fmr_flush_trigger / ncpts;
1411
1412         return max(IBLND_FMR_POOL_FLUSH, size);
1413 }
1414
1415 static int
1416 kiblnd_create_fmr_pool(kib_fmr_poolset_t *fps, kib_fmr_pool_t **pp_fpo)
1417 {
1418         /* FMR pool for RDMA */
1419         kib_dev_t              *dev = fps->fps_net->ibn_dev;
1420         kib_fmr_pool_t    *fpo;
1421         struct ib_fmr_pool_param param = {
1422                 .max_pages_per_fmr = LNET_MAX_PAYLOAD/PAGE_SIZE,
1423                 .page_shift     = PAGE_SHIFT,
1424                 .access     = (IB_ACCESS_LOCAL_WRITE |
1425                                       IB_ACCESS_REMOTE_WRITE),
1426                 .pool_size         = fps->fps_pool_size,
1427                 .dirty_watermark   = fps->fps_flush_trigger,
1428                 .flush_function    = NULL,
1429                 .flush_arg       = NULL,
1430                 .cache       = !!*kiblnd_tunables.kib_fmr_cache};
1431         int rc;
1432
1433         LIBCFS_CPT_ALLOC(fpo, lnet_cpt_table(), fps->fps_cpt, sizeof(*fpo));
1434         if (fpo == NULL)
1435                 return -ENOMEM;
1436
1437         fpo->fpo_hdev = kiblnd_current_hdev(dev);
1438
1439         fpo->fpo_fmr_pool = ib_create_fmr_pool(fpo->fpo_hdev->ibh_pd, &param);
1440         if (IS_ERR(fpo->fpo_fmr_pool)) {
1441                 rc = PTR_ERR(fpo->fpo_fmr_pool);
1442                 CERROR("Failed to create FMR pool: %d\n", rc);
1443
1444                 kiblnd_hdev_decref(fpo->fpo_hdev);
1445                 LIBCFS_FREE(fpo, sizeof(kib_fmr_pool_t));
1446                 return rc;
1447         }
1448
1449         fpo->fpo_deadline = cfs_time_shift(IBLND_POOL_DEADLINE);
1450         fpo->fpo_owner    = fps;
1451         *pp_fpo = fpo;
1452
1453         return 0;
1454 }
1455
1456 static void
1457 kiblnd_fail_fmr_poolset(kib_fmr_poolset_t *fps, struct list_head *zombies)
1458 {
1459         if (fps->fps_net == NULL) /* intialized? */
1460                 return;
1461
1462         spin_lock(&fps->fps_lock);
1463
1464         while (!list_empty(&fps->fps_pool_list)) {
1465                 kib_fmr_pool_t *fpo = list_entry(fps->fps_pool_list.next,
1466                                                  kib_fmr_pool_t, fpo_list);
1467                 fpo->fpo_failed = 1;
1468                 list_del(&fpo->fpo_list);
1469                 if (fpo->fpo_map_count == 0)
1470                         list_add(&fpo->fpo_list, zombies);
1471                 else
1472                         list_add(&fpo->fpo_list, &fps->fps_failed_pool_list);
1473         }
1474
1475         spin_unlock(&fps->fps_lock);
1476 }
1477
1478 static void
1479 kiblnd_fini_fmr_poolset(kib_fmr_poolset_t *fps)
1480 {
1481         if (fps->fps_net != NULL) { /* initialized? */
1482                 kiblnd_destroy_fmr_pool_list(&fps->fps_failed_pool_list);
1483                 kiblnd_destroy_fmr_pool_list(&fps->fps_pool_list);
1484         }
1485 }
1486
1487 static int
1488 kiblnd_init_fmr_poolset(kib_fmr_poolset_t *fps, int cpt, kib_net_t *net,
1489                         int pool_size, int flush_trigger)
1490 {
1491         kib_fmr_pool_t *fpo;
1492         int          rc;
1493
1494         memset(fps, 0, sizeof(kib_fmr_poolset_t));
1495
1496         fps->fps_net = net;
1497         fps->fps_cpt = cpt;
1498         fps->fps_pool_size = pool_size;
1499         fps->fps_flush_trigger = flush_trigger;
1500         spin_lock_init(&fps->fps_lock);
1501         INIT_LIST_HEAD(&fps->fps_pool_list);
1502         INIT_LIST_HEAD(&fps->fps_failed_pool_list);
1503
1504         rc = kiblnd_create_fmr_pool(fps, &fpo);
1505         if (rc == 0)
1506                 list_add_tail(&fpo->fpo_list, &fps->fps_pool_list);
1507
1508         return rc;
1509 }
1510
1511 static int
1512 kiblnd_fmr_pool_is_idle(kib_fmr_pool_t *fpo, unsigned long now)
1513 {
1514         if (fpo->fpo_map_count != 0) /* still in use */
1515                 return 0;
1516         if (fpo->fpo_failed)
1517                 return 1;
1518         return cfs_time_aftereq(now, fpo->fpo_deadline);
1519 }
1520
1521 void
1522 kiblnd_fmr_pool_unmap(kib_fmr_t *fmr, int status)
1523 {
1524         LIST_HEAD     (zombies);
1525         kib_fmr_pool_t    *fpo = fmr->fmr_pool;
1526         kib_fmr_poolset_t *fps = fpo->fpo_owner;
1527         unsigned long    now = cfs_time_current();
1528         kib_fmr_pool_t    *tmp;
1529         int             rc;
1530
1531         rc = ib_fmr_pool_unmap(fmr->fmr_pfmr);
1532         LASSERT (rc == 0);
1533
1534         if (status != 0) {
1535                 rc = ib_flush_fmr_pool(fpo->fpo_fmr_pool);
1536                 LASSERT (rc == 0);
1537         }
1538
1539         fmr->fmr_pool = NULL;
1540         fmr->fmr_pfmr = NULL;
1541
1542         spin_lock(&fps->fps_lock);
1543         fpo->fpo_map_count --;  /* decref the pool */
1544
1545         list_for_each_entry_safe(fpo, tmp, &fps->fps_pool_list, fpo_list) {
1546                 /* the first pool is persistent */
1547                 if (fps->fps_pool_list.next == &fpo->fpo_list)
1548                         continue;
1549
1550                 if (kiblnd_fmr_pool_is_idle(fpo, now)) {
1551                         list_move(&fpo->fpo_list, &zombies);
1552                         fps->fps_version ++;
1553                 }
1554         }
1555         spin_unlock(&fps->fps_lock);
1556
1557         if (!list_empty(&zombies))
1558                 kiblnd_destroy_fmr_pool_list(&zombies);
1559 }
1560
1561 int
1562 kiblnd_fmr_pool_map(kib_fmr_poolset_t *fps, __u64 *pages, int npages,
1563                     __u64 iov, kib_fmr_t *fmr)
1564 {
1565         struct ib_pool_fmr *pfmr;
1566         kib_fmr_pool_t     *fpo;
1567         __u64          version;
1568         int              rc;
1569
1570  again:
1571         spin_lock(&fps->fps_lock);
1572         version = fps->fps_version;
1573         list_for_each_entry(fpo, &fps->fps_pool_list, fpo_list) {
1574                 fpo->fpo_deadline = cfs_time_shift(IBLND_POOL_DEADLINE);
1575                 fpo->fpo_map_count++;
1576                 spin_unlock(&fps->fps_lock);
1577
1578                 pfmr = ib_fmr_pool_map_phys(fpo->fpo_fmr_pool,
1579                                             pages, npages, iov);
1580                 if (likely(!IS_ERR(pfmr))) {
1581                         fmr->fmr_pool = fpo;
1582                         fmr->fmr_pfmr = pfmr;
1583                         return 0;
1584                 }
1585
1586                 spin_lock(&fps->fps_lock);
1587                 fpo->fpo_map_count--;
1588                 if (PTR_ERR(pfmr) != -EAGAIN) {
1589                         spin_unlock(&fps->fps_lock);
1590                         return PTR_ERR(pfmr);
1591                 }
1592
1593                 /* EAGAIN and ... */
1594                 if (version != fps->fps_version) {
1595                         spin_unlock(&fps->fps_lock);
1596                         goto again;
1597                 }
1598         }
1599
1600         if (fps->fps_increasing) {
1601                 spin_unlock(&fps->fps_lock);
1602                 CDEBUG(D_NET, "Another thread is allocating new "
1603                        "FMR pool, waiting for her to complete\n");
1604                 schedule();
1605                 goto again;
1606
1607         }
1608
1609         if (time_before(cfs_time_current(), fps->fps_next_retry)) {
1610                 /* someone failed recently */
1611                 spin_unlock(&fps->fps_lock);
1612                 return -EAGAIN;
1613         }
1614
1615         fps->fps_increasing = 1;
1616         spin_unlock(&fps->fps_lock);
1617
1618         CDEBUG(D_NET, "Allocate new FMR pool\n");
1619         rc = kiblnd_create_fmr_pool(fps, &fpo);
1620         spin_lock(&fps->fps_lock);
1621         fps->fps_increasing = 0;
1622         if (rc == 0) {
1623                 fps->fps_version++;
1624                 list_add_tail(&fpo->fpo_list, &fps->fps_pool_list);
1625         } else {
1626                 fps->fps_next_retry = cfs_time_shift(IBLND_POOL_RETRY);
1627         }
1628         spin_unlock(&fps->fps_lock);
1629
1630         goto again;
1631 }
1632
1633 static void
1634 kiblnd_fini_pool(kib_pool_t *pool)
1635 {
1636         LASSERT (list_empty(&pool->po_free_list));
1637         LASSERT (pool->po_allocated == 0);
1638
1639         CDEBUG(D_NET, "Finalize %s pool\n", pool->po_owner->ps_name);
1640 }
1641
1642 static void
1643 kiblnd_init_pool(kib_poolset_t *ps, kib_pool_t *pool, int size)
1644 {
1645         CDEBUG(D_NET, "Initialize %s pool\n", ps->ps_name);
1646
1647         memset(pool, 0, sizeof(kib_pool_t));
1648         INIT_LIST_HEAD(&pool->po_free_list);
1649         pool->po_deadline = cfs_time_shift(IBLND_POOL_DEADLINE);
1650         pool->po_owner    = ps;
1651         pool->po_size     = size;
1652 }
1653
1654 static void
1655 kiblnd_destroy_pool_list(struct list_head *head)
1656 {
1657         kib_pool_t *pool;
1658
1659         while (!list_empty(head)) {
1660                 pool = list_entry(head->next, kib_pool_t, po_list);
1661                 list_del(&pool->po_list);
1662
1663                 LASSERT (pool->po_owner != NULL);
1664                 pool->po_owner->ps_pool_destroy(pool);
1665         }
1666 }
1667
1668 static void
1669 kiblnd_fail_poolset(kib_poolset_t *ps, struct list_head *zombies)
1670 {
1671         if (ps->ps_net == NULL) /* intialized? */
1672                 return;
1673
1674         spin_lock(&ps->ps_lock);
1675         while (!list_empty(&ps->ps_pool_list)) {
1676                 kib_pool_t *po = list_entry(ps->ps_pool_list.next,
1677                                             kib_pool_t, po_list);
1678                 po->po_failed = 1;
1679                 list_del(&po->po_list);
1680                 if (po->po_allocated == 0)
1681                         list_add(&po->po_list, zombies);
1682                 else
1683                         list_add(&po->po_list, &ps->ps_failed_pool_list);
1684         }
1685         spin_unlock(&ps->ps_lock);
1686 }
1687
1688 static void
1689 kiblnd_fini_poolset(kib_poolset_t *ps)
1690 {
1691         if (ps->ps_net != NULL) { /* initialized? */
1692                 kiblnd_destroy_pool_list(&ps->ps_failed_pool_list);
1693                 kiblnd_destroy_pool_list(&ps->ps_pool_list);
1694         }
1695 }
1696
1697 static int
1698 kiblnd_init_poolset(kib_poolset_t *ps, int cpt,
1699                     kib_net_t *net, char *name, int size,
1700                     kib_ps_pool_create_t po_create,
1701                     kib_ps_pool_destroy_t po_destroy,
1702                     kib_ps_node_init_t nd_init,
1703                     kib_ps_node_fini_t nd_fini)
1704 {
1705         kib_pool_t      *pool;
1706         int             rc;
1707
1708         memset(ps, 0, sizeof(kib_poolset_t));
1709
1710         ps->ps_cpt          = cpt;
1711         ps->ps_net        = net;
1712         ps->ps_pool_create  = po_create;
1713         ps->ps_pool_destroy = po_destroy;
1714         ps->ps_node_init    = nd_init;
1715         ps->ps_node_fini    = nd_fini;
1716         ps->ps_pool_size    = size;
1717         if (strlcpy(ps->ps_name, name, sizeof(ps->ps_name))
1718             >= sizeof(ps->ps_name))
1719                 return -E2BIG;
1720         spin_lock_init(&ps->ps_lock);
1721         INIT_LIST_HEAD(&ps->ps_pool_list);
1722         INIT_LIST_HEAD(&ps->ps_failed_pool_list);
1723
1724         rc = ps->ps_pool_create(ps, size, &pool);
1725         if (rc == 0)
1726                 list_add(&pool->po_list, &ps->ps_pool_list);
1727         else
1728                 CERROR("Failed to create the first pool for %s\n", ps->ps_name);
1729
1730         return rc;
1731 }
1732
1733 static int
1734 kiblnd_pool_is_idle(kib_pool_t *pool, unsigned long now)
1735 {
1736         if (pool->po_allocated != 0) /* still in use */
1737                 return 0;
1738         if (pool->po_failed)
1739                 return 1;
1740         return cfs_time_aftereq(now, pool->po_deadline);
1741 }
1742
1743 void
1744 kiblnd_pool_free_node(kib_pool_t *pool, struct list_head *node)
1745 {
1746         LIST_HEAD  (zombies);
1747         kib_poolset_t  *ps = pool->po_owner;
1748         kib_pool_t     *tmp;
1749         unsigned long      now = cfs_time_current();
1750
1751         spin_lock(&ps->ps_lock);
1752
1753         if (ps->ps_node_fini != NULL)
1754                 ps->ps_node_fini(pool, node);
1755
1756         LASSERT (pool->po_allocated > 0);
1757         list_add(node, &pool->po_free_list);
1758         pool->po_allocated --;
1759
1760         list_for_each_entry_safe(pool, tmp, &ps->ps_pool_list, po_list) {
1761                 /* the first pool is persistent */
1762                 if (ps->ps_pool_list.next == &pool->po_list)
1763                         continue;
1764
1765                 if (kiblnd_pool_is_idle(pool, now))
1766                         list_move(&pool->po_list, &zombies);
1767         }
1768         spin_unlock(&ps->ps_lock);
1769
1770         if (!list_empty(&zombies))
1771                 kiblnd_destroy_pool_list(&zombies);
1772 }
1773
1774 struct list_head *
1775 kiblnd_pool_alloc_node(kib_poolset_t *ps)
1776 {
1777         struct list_head            *node;
1778         kib_pool_t          *pool;
1779         int                 rc;
1780
1781  again:
1782         spin_lock(&ps->ps_lock);
1783         list_for_each_entry(pool, &ps->ps_pool_list, po_list) {
1784                 if (list_empty(&pool->po_free_list))
1785                         continue;
1786
1787                 pool->po_allocated ++;
1788                 pool->po_deadline = cfs_time_shift(IBLND_POOL_DEADLINE);
1789                 node = pool->po_free_list.next;
1790                 list_del(node);
1791
1792                 if (ps->ps_node_init != NULL) {
1793                         /* still hold the lock */
1794                         ps->ps_node_init(pool, node);
1795                 }
1796                 spin_unlock(&ps->ps_lock);
1797                 return node;
1798         }
1799
1800         /* no available tx pool and ... */
1801         if (ps->ps_increasing) {
1802                 /* another thread is allocating a new pool */
1803                 spin_unlock(&ps->ps_lock);
1804                 CDEBUG(D_NET, "Another thread is allocating new "
1805                        "%s pool, waiting for her to complete\n",
1806                        ps->ps_name);
1807                 schedule();
1808                 goto again;
1809         }
1810
1811         if (time_before(cfs_time_current(), ps->ps_next_retry)) {
1812                 /* someone failed recently */
1813                 spin_unlock(&ps->ps_lock);
1814                 return NULL;
1815         }
1816
1817         ps->ps_increasing = 1;
1818         spin_unlock(&ps->ps_lock);
1819
1820         CDEBUG(D_NET, "%s pool exhausted, allocate new pool\n", ps->ps_name);
1821
1822         rc = ps->ps_pool_create(ps, ps->ps_pool_size, &pool);
1823
1824         spin_lock(&ps->ps_lock);
1825         ps->ps_increasing = 0;
1826         if (rc == 0) {
1827                 list_add_tail(&pool->po_list, &ps->ps_pool_list);
1828         } else {
1829                 ps->ps_next_retry = cfs_time_shift(IBLND_POOL_RETRY);
1830                 CERROR("Can't allocate new %s pool because out of memory\n",
1831                        ps->ps_name);
1832         }
1833         spin_unlock(&ps->ps_lock);
1834
1835         goto again;
1836 }
1837
1838 void
1839 kiblnd_pmr_pool_unmap(kib_phys_mr_t *pmr)
1840 {
1841         kib_pmr_pool_t      *ppo = pmr->pmr_pool;
1842         struct ib_mr    *mr  = pmr->pmr_mr;
1843
1844         pmr->pmr_mr = NULL;
1845         kiblnd_pool_free_node(&ppo->ppo_pool, &pmr->pmr_list);
1846         if (mr != NULL)
1847                 ib_dereg_mr(mr);
1848 }
1849
1850 int
1851 kiblnd_pmr_pool_map(kib_pmr_poolset_t *pps, kib_hca_dev_t *hdev,
1852                     kib_rdma_desc_t *rd, __u64 *iova, kib_phys_mr_t **pp_pmr)
1853 {
1854         kib_phys_mr_t *pmr;
1855         struct list_head    *node;
1856         int         rc;
1857         int         i;
1858
1859         node = kiblnd_pool_alloc_node(&pps->pps_poolset);
1860         if (node == NULL) {
1861                 CERROR("Failed to allocate PMR descriptor\n");
1862                 return -ENOMEM;
1863         }
1864
1865         pmr = container_of(node, kib_phys_mr_t, pmr_list);
1866         if (pmr->pmr_pool->ppo_hdev != hdev) {
1867                 kiblnd_pool_free_node(&pmr->pmr_pool->ppo_pool, node);
1868                 return -EAGAIN;
1869         }
1870
1871         for (i = 0; i < rd->rd_nfrags; i ++) {
1872                 pmr->pmr_ipb[i].addr = rd->rd_frags[i].rf_addr;
1873                 pmr->pmr_ipb[i].size = rd->rd_frags[i].rf_nob;
1874         }
1875
1876         pmr->pmr_mr = ib_reg_phys_mr(hdev->ibh_pd,
1877                                      pmr->pmr_ipb, rd->rd_nfrags,
1878                                      IB_ACCESS_LOCAL_WRITE |
1879                                      IB_ACCESS_REMOTE_WRITE,
1880                                      iova);
1881         if (!IS_ERR(pmr->pmr_mr)) {
1882                 pmr->pmr_iova = *iova;
1883                 *pp_pmr = pmr;
1884                 return 0;
1885         }
1886
1887         rc = PTR_ERR(pmr->pmr_mr);
1888         CERROR("Failed ib_reg_phys_mr: %d\n", rc);
1889
1890         pmr->pmr_mr = NULL;
1891         kiblnd_pool_free_node(&pmr->pmr_pool->ppo_pool, node);
1892
1893         return rc;
1894 }
1895
1896 static void
1897 kiblnd_destroy_pmr_pool(kib_pool_t *pool)
1898 {
1899         kib_pmr_pool_t *ppo = container_of(pool, kib_pmr_pool_t, ppo_pool);
1900         kib_phys_mr_t  *pmr;
1901
1902         LASSERT (pool->po_allocated == 0);
1903
1904         while (!list_empty(&pool->po_free_list)) {
1905                 pmr = list_entry(pool->po_free_list.next,
1906                                      kib_phys_mr_t, pmr_list);
1907
1908                 LASSERT (pmr->pmr_mr == NULL);
1909                 list_del(&pmr->pmr_list);
1910
1911                 if (pmr->pmr_ipb != NULL) {
1912                         LIBCFS_FREE(pmr->pmr_ipb,
1913                                     IBLND_MAX_RDMA_FRAGS *
1914                                     sizeof(struct ib_phys_buf));
1915                 }
1916
1917                 LIBCFS_FREE(pmr, sizeof(kib_phys_mr_t));
1918         }
1919
1920         kiblnd_fini_pool(pool);
1921         if (ppo->ppo_hdev != NULL)
1922                 kiblnd_hdev_decref(ppo->ppo_hdev);
1923
1924         LIBCFS_FREE(ppo, sizeof(kib_pmr_pool_t));
1925 }
1926
1927 static inline int kiblnd_pmr_pool_size(int ncpts)
1928 {
1929         int size = *kiblnd_tunables.kib_pmr_pool_size / ncpts;
1930
1931         return max(IBLND_PMR_POOL, size);
1932 }
1933
1934 static int
1935 kiblnd_create_pmr_pool(kib_poolset_t *ps, int size, kib_pool_t **pp_po)
1936 {
1937         struct kib_pmr_pool     *ppo;
1938         struct kib_pool         *pool;
1939         kib_phys_mr_t           *pmr;
1940         int                     i;
1941
1942         LIBCFS_CPT_ALLOC(ppo, lnet_cpt_table(),
1943                          ps->ps_cpt, sizeof(kib_pmr_pool_t));
1944         if (ppo == NULL) {
1945                 CERROR("Failed to allocate PMR pool\n");
1946                 return -ENOMEM;
1947         }
1948
1949         pool = &ppo->ppo_pool;
1950         kiblnd_init_pool(ps, pool, size);
1951
1952         for (i = 0; i < size; i++) {
1953                 LIBCFS_CPT_ALLOC(pmr, lnet_cpt_table(),
1954                                  ps->ps_cpt, sizeof(kib_phys_mr_t));
1955                 if (pmr == NULL)
1956                         break;
1957
1958                 pmr->pmr_pool = ppo;
1959                 LIBCFS_CPT_ALLOC(pmr->pmr_ipb, lnet_cpt_table(), ps->ps_cpt,
1960                                  IBLND_MAX_RDMA_FRAGS * sizeof(*pmr->pmr_ipb));
1961                 if (pmr->pmr_ipb == NULL)
1962                         break;
1963
1964                 list_add(&pmr->pmr_list, &pool->po_free_list);
1965         }
1966
1967         if (i < size) {
1968                 ps->ps_pool_destroy(pool);
1969                 return -ENOMEM;
1970         }
1971
1972         ppo->ppo_hdev = kiblnd_current_hdev(ps->ps_net->ibn_dev);
1973         *pp_po = pool;
1974         return 0;
1975 }
1976
1977 static void
1978 kiblnd_destroy_tx_pool(kib_pool_t *pool)
1979 {
1980         kib_tx_pool_t  *tpo = container_of(pool, kib_tx_pool_t, tpo_pool);
1981         int          i;
1982
1983         LASSERT (pool->po_allocated == 0);
1984
1985         if (tpo->tpo_tx_pages != NULL) {
1986                 kiblnd_unmap_tx_pool(tpo);
1987                 kiblnd_free_pages(tpo->tpo_tx_pages);
1988         }
1989
1990         if (tpo->tpo_tx_descs == NULL)
1991                 goto out;
1992
1993         for (i = 0; i < pool->po_size; i++) {
1994                 kib_tx_t *tx = &tpo->tpo_tx_descs[i];
1995
1996                 list_del(&tx->tx_list);
1997                 if (tx->tx_pages != NULL)
1998                         LIBCFS_FREE(tx->tx_pages,
1999                                     LNET_MAX_IOV *
2000                                     sizeof(*tx->tx_pages));
2001                 if (tx->tx_frags != NULL)
2002                         LIBCFS_FREE(tx->tx_frags,
2003                                     IBLND_MAX_RDMA_FRAGS *
2004                                             sizeof(*tx->tx_frags));
2005                 if (tx->tx_wrq != NULL)
2006                         LIBCFS_FREE(tx->tx_wrq,
2007                                     (1 + IBLND_MAX_RDMA_FRAGS) *
2008                                     sizeof(*tx->tx_wrq));
2009                 if (tx->tx_sge != NULL)
2010                         LIBCFS_FREE(tx->tx_sge,
2011                                     (1 + IBLND_MAX_RDMA_FRAGS) *
2012                                     sizeof(*tx->tx_sge));
2013                 if (tx->tx_rd != NULL)
2014                         LIBCFS_FREE(tx->tx_rd,
2015                                     offsetof(kib_rdma_desc_t,
2016                                              rd_frags[IBLND_MAX_RDMA_FRAGS]));
2017         }
2018
2019         LIBCFS_FREE(tpo->tpo_tx_descs,
2020                     pool->po_size * sizeof(kib_tx_t));
2021 out:
2022         kiblnd_fini_pool(pool);
2023         LIBCFS_FREE(tpo, sizeof(kib_tx_pool_t));
2024 }
2025
2026 static int kiblnd_tx_pool_size(int ncpts)
2027 {
2028         int ntx = *kiblnd_tunables.kib_ntx / ncpts;
2029
2030         return max(IBLND_TX_POOL, ntx);
2031 }
2032
2033 static int
2034 kiblnd_create_tx_pool(kib_poolset_t *ps, int size, kib_pool_t **pp_po)
2035 {
2036         int         i;
2037         int         npg;
2038         kib_pool_t    *pool;
2039         kib_tx_pool_t *tpo;
2040
2041         LIBCFS_CPT_ALLOC(tpo, lnet_cpt_table(), ps->ps_cpt, sizeof(*tpo));
2042         if (tpo == NULL) {
2043                 CERROR("Failed to allocate TX pool\n");
2044                 return -ENOMEM;
2045         }
2046
2047         pool = &tpo->tpo_pool;
2048         kiblnd_init_pool(ps, pool, size);
2049         tpo->tpo_tx_descs = NULL;
2050         tpo->tpo_tx_pages = NULL;
2051
2052         npg = (size * IBLND_MSG_SIZE + PAGE_SIZE - 1) / PAGE_SIZE;
2053         if (kiblnd_alloc_pages(&tpo->tpo_tx_pages, ps->ps_cpt, npg) != 0) {
2054                 CERROR("Can't allocate tx pages: %d\n", npg);
2055                 LIBCFS_FREE(tpo, sizeof(kib_tx_pool_t));
2056                 return -ENOMEM;
2057         }
2058
2059         LIBCFS_CPT_ALLOC(tpo->tpo_tx_descs, lnet_cpt_table(), ps->ps_cpt,
2060                          size * sizeof(kib_tx_t));
2061         if (tpo->tpo_tx_descs == NULL) {
2062                 CERROR("Can't allocate %d tx descriptors\n", size);
2063                 ps->ps_pool_destroy(pool);
2064                 return -ENOMEM;
2065         }
2066
2067         memset(tpo->tpo_tx_descs, 0, size * sizeof(kib_tx_t));
2068
2069         for (i = 0; i < size; i++) {
2070                 kib_tx_t *tx = &tpo->tpo_tx_descs[i];
2071
2072                 tx->tx_pool = tpo;
2073                 if (ps->ps_net->ibn_fmr_ps != NULL) {
2074                         LIBCFS_CPT_ALLOC(tx->tx_pages,
2075                                          lnet_cpt_table(), ps->ps_cpt,
2076                                          LNET_MAX_IOV * sizeof(*tx->tx_pages));
2077                         if (tx->tx_pages == NULL)
2078                                 break;
2079                 }
2080
2081                 LIBCFS_CPT_ALLOC(tx->tx_frags, lnet_cpt_table(), ps->ps_cpt,
2082                                  IBLND_MAX_RDMA_FRAGS * sizeof(*tx->tx_frags));
2083                 if (tx->tx_frags == NULL)
2084                         break;
2085
2086                 sg_init_table(tx->tx_frags, IBLND_MAX_RDMA_FRAGS);
2087
2088                 LIBCFS_CPT_ALLOC(tx->tx_wrq, lnet_cpt_table(), ps->ps_cpt,
2089                                  (1 + IBLND_MAX_RDMA_FRAGS) *
2090                                  sizeof(*tx->tx_wrq));
2091                 if (tx->tx_wrq == NULL)
2092                         break;
2093
2094                 LIBCFS_CPT_ALLOC(tx->tx_sge, lnet_cpt_table(), ps->ps_cpt,
2095                                  (1 + IBLND_MAX_RDMA_FRAGS) *
2096                                  sizeof(*tx->tx_sge));
2097                 if (tx->tx_sge == NULL)
2098                         break;
2099
2100                 LIBCFS_CPT_ALLOC(tx->tx_rd, lnet_cpt_table(), ps->ps_cpt,
2101                                  offsetof(kib_rdma_desc_t,
2102                                           rd_frags[IBLND_MAX_RDMA_FRAGS]));
2103                 if (tx->tx_rd == NULL)
2104                         break;
2105         }
2106
2107         if (i == size) {
2108                 kiblnd_map_tx_pool(tpo);
2109                 *pp_po = pool;
2110                 return 0;
2111         }
2112
2113         ps->ps_pool_destroy(pool);
2114         return -ENOMEM;
2115 }
2116
2117 static void
2118 kiblnd_tx_init(kib_pool_t *pool, struct list_head *node)
2119 {
2120         kib_tx_poolset_t *tps = container_of(pool->po_owner, kib_tx_poolset_t,
2121                                              tps_poolset);
2122         kib_tx_t         *tx  = list_entry(node, kib_tx_t, tx_list);
2123
2124         tx->tx_cookie = tps->tps_next_tx_cookie ++;
2125 }
2126
2127 static void
2128 kiblnd_net_fini_pools(kib_net_t *net)
2129 {
2130         int     i;
2131
2132         cfs_cpt_for_each(i, lnet_cpt_table()) {
2133                 kib_tx_poolset_t        *tps;
2134                 kib_fmr_poolset_t       *fps;
2135                 kib_pmr_poolset_t       *pps;
2136
2137                 if (net->ibn_tx_ps != NULL) {
2138                         tps = net->ibn_tx_ps[i];
2139                         kiblnd_fini_poolset(&tps->tps_poolset);
2140                 }
2141
2142                 if (net->ibn_fmr_ps != NULL) {
2143                         fps = net->ibn_fmr_ps[i];
2144                         kiblnd_fini_fmr_poolset(fps);
2145                 }
2146
2147                 if (net->ibn_pmr_ps != NULL) {
2148                         pps = net->ibn_pmr_ps[i];
2149                         kiblnd_fini_poolset(&pps->pps_poolset);
2150                 }
2151         }
2152
2153         if (net->ibn_tx_ps != NULL) {
2154                 cfs_percpt_free(net->ibn_tx_ps);
2155                 net->ibn_tx_ps = NULL;
2156         }
2157
2158         if (net->ibn_fmr_ps != NULL) {
2159                 cfs_percpt_free(net->ibn_fmr_ps);
2160                 net->ibn_fmr_ps = NULL;
2161         }
2162
2163         if (net->ibn_pmr_ps != NULL) {
2164                 cfs_percpt_free(net->ibn_pmr_ps);
2165                 net->ibn_pmr_ps = NULL;
2166         }
2167 }
2168
2169 static int
2170 kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts)
2171 {
2172         unsigned long   flags;
2173         int             cpt;
2174         int             rc;
2175         int             i;
2176
2177         read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2178         if (*kiblnd_tunables.kib_map_on_demand == 0 &&
2179             net->ibn_dev->ibd_hdev->ibh_nmrs == 1) {
2180                 read_unlock_irqrestore(&kiblnd_data.kib_global_lock,
2181                                            flags);
2182                 goto create_tx_pool;
2183         }
2184
2185         read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2186
2187         if (*kiblnd_tunables.kib_fmr_pool_size <
2188             *kiblnd_tunables.kib_ntx / 4) {
2189                 CERROR("Can't set fmr pool size (%d) < ntx / 4(%d)\n",
2190                        *kiblnd_tunables.kib_fmr_pool_size,
2191                        *kiblnd_tunables.kib_ntx / 4);
2192                 rc = -EINVAL;
2193                 goto failed;
2194         }
2195
2196         /* TX pool must be created later than FMR/PMR, see LU-2268
2197          * for details */
2198         LASSERT(net->ibn_tx_ps == NULL);
2199
2200         /* premapping can fail if ibd_nmr > 1, so we always create
2201          * FMR/PMR pool and map-on-demand if premapping failed */
2202
2203         net->ibn_fmr_ps = cfs_percpt_alloc(lnet_cpt_table(),
2204                                            sizeof(kib_fmr_poolset_t));
2205         if (net->ibn_fmr_ps == NULL) {
2206                 CERROR("Failed to allocate FMR pool array\n");
2207                 rc = -ENOMEM;
2208                 goto failed;
2209         }
2210
2211         for (i = 0; i < ncpts; i++) {
2212                 cpt = (cpts == NULL) ? i : cpts[i];
2213                 rc = kiblnd_init_fmr_poolset(net->ibn_fmr_ps[cpt], cpt, net,
2214                                              kiblnd_fmr_pool_size(ncpts),
2215                                              kiblnd_fmr_flush_trigger(ncpts));
2216                 if (rc == -ENOSYS && i == 0) /* no FMR */
2217                         break; /* create PMR pool */
2218
2219                 if (rc != 0) { /* a real error */
2220                         CERROR("Can't initialize FMR pool for CPT %d: %d\n",
2221                                cpt, rc);
2222                         goto failed;
2223                 }
2224         }
2225
2226         if (i > 0) {
2227                 LASSERT(i == ncpts);
2228                 goto create_tx_pool;
2229         }
2230
2231         cfs_percpt_free(net->ibn_fmr_ps);
2232         net->ibn_fmr_ps = NULL;
2233
2234         CWARN("Device does not support FMR, failing back to PMR\n");
2235
2236         if (*kiblnd_tunables.kib_pmr_pool_size <
2237             *kiblnd_tunables.kib_ntx / 4) {
2238                 CERROR("Can't set pmr pool size (%d) < ntx / 4(%d)\n",
2239                        *kiblnd_tunables.kib_pmr_pool_size,
2240                        *kiblnd_tunables.kib_ntx / 4);
2241                 rc = -EINVAL;
2242                 goto failed;
2243         }
2244
2245         net->ibn_pmr_ps = cfs_percpt_alloc(lnet_cpt_table(),
2246                                            sizeof(kib_pmr_poolset_t));
2247         if (net->ibn_pmr_ps == NULL) {
2248                 CERROR("Failed to allocate PMR pool array\n");
2249                 rc = -ENOMEM;
2250                 goto failed;
2251         }
2252
2253         for (i = 0; i < ncpts; i++) {
2254                 cpt = (cpts == NULL) ? i : cpts[i];
2255                 rc = kiblnd_init_poolset(&net->ibn_pmr_ps[cpt]->pps_poolset,
2256                                          cpt, net, "PMR",
2257                                          kiblnd_pmr_pool_size(ncpts),
2258                                          kiblnd_create_pmr_pool,
2259                                          kiblnd_destroy_pmr_pool, NULL, NULL);
2260                 if (rc != 0) {
2261                         CERROR("Can't initialize PMR pool for CPT %d: %d\n",
2262                                cpt, rc);
2263                         goto failed;
2264                 }
2265         }
2266
2267  create_tx_pool:
2268         net->ibn_tx_ps = cfs_percpt_alloc(lnet_cpt_table(),
2269                                           sizeof(kib_tx_poolset_t));
2270         if (net->ibn_tx_ps == NULL) {
2271                 CERROR("Failed to allocate tx pool array\n");
2272                 rc = -ENOMEM;
2273                 goto failed;
2274         }
2275
2276         for (i = 0; i < ncpts; i++) {
2277                 cpt = (cpts == NULL) ? i : cpts[i];
2278                 rc = kiblnd_init_poolset(&net->ibn_tx_ps[cpt]->tps_poolset,
2279                                          cpt, net, "TX",
2280                                          kiblnd_tx_pool_size(ncpts),
2281                                          kiblnd_create_tx_pool,
2282                                          kiblnd_destroy_tx_pool,
2283                                          kiblnd_tx_init, NULL);
2284                 if (rc != 0) {
2285                         CERROR("Can't initialize TX pool for CPT %d: %d\n",
2286                                cpt, rc);
2287                         goto failed;
2288                 }
2289         }
2290
2291         return 0;
2292  failed:
2293         kiblnd_net_fini_pools(net);
2294         LASSERT(rc != 0);
2295         return rc;
2296 }
2297
2298 static int
2299 kiblnd_hdev_get_attr(kib_hca_dev_t *hdev)
2300 {
2301         struct ib_device_attr *attr;
2302         int                 rc;
2303
2304         /* It's safe to assume a HCA can handle a page size
2305          * matching that of the native system */
2306         hdev->ibh_page_shift = PAGE_SHIFT;
2307         hdev->ibh_page_size  = 1 << PAGE_SHIFT;
2308         hdev->ibh_page_mask  = ~((__u64)hdev->ibh_page_size - 1);
2309
2310         LIBCFS_ALLOC(attr, sizeof(*attr));
2311         if (attr == NULL) {
2312                 CERROR("Out of memory\n");
2313                 return -ENOMEM;
2314         }
2315
2316         rc = ib_query_device(hdev->ibh_ibdev, attr);
2317         if (rc == 0)
2318                 hdev->ibh_mr_size = attr->max_mr_size;
2319
2320         LIBCFS_FREE(attr, sizeof(*attr));
2321
2322         if (rc != 0) {
2323                 CERROR("Failed to query IB device: %d\n", rc);
2324                 return rc;
2325         }
2326
2327         if (hdev->ibh_mr_size == ~0ULL) {
2328                 hdev->ibh_mr_shift = 64;
2329                 return 0;
2330         }
2331
2332         for (hdev->ibh_mr_shift = 0;
2333              hdev->ibh_mr_shift < 64; hdev->ibh_mr_shift ++) {
2334                 if (hdev->ibh_mr_size == (1ULL << hdev->ibh_mr_shift) ||
2335                     hdev->ibh_mr_size == (1ULL << hdev->ibh_mr_shift) - 1)
2336                         return 0;
2337         }
2338
2339         CERROR("Invalid mr size: %#llx\n", hdev->ibh_mr_size);
2340         return -EINVAL;
2341 }
2342
2343 static void
2344 kiblnd_hdev_cleanup_mrs(kib_hca_dev_t *hdev)
2345 {
2346         int     i;
2347
2348         if (hdev->ibh_nmrs == 0 || hdev->ibh_mrs == NULL)
2349                 return;
2350
2351         for (i = 0; i < hdev->ibh_nmrs; i++) {
2352                 if (hdev->ibh_mrs[i] == NULL)
2353                         break;
2354
2355                 ib_dereg_mr(hdev->ibh_mrs[i]);
2356         }
2357
2358         LIBCFS_FREE(hdev->ibh_mrs, sizeof(*hdev->ibh_mrs) * hdev->ibh_nmrs);
2359         hdev->ibh_mrs  = NULL;
2360         hdev->ibh_nmrs = 0;
2361 }
2362
2363 void
2364 kiblnd_hdev_destroy(kib_hca_dev_t *hdev)
2365 {
2366         kiblnd_hdev_cleanup_mrs(hdev);
2367
2368         if (hdev->ibh_pd != NULL)
2369                 ib_dealloc_pd(hdev->ibh_pd);
2370
2371         if (hdev->ibh_cmid != NULL)
2372                 rdma_destroy_id(hdev->ibh_cmid);
2373
2374         LIBCFS_FREE(hdev, sizeof(*hdev));
2375 }
2376
2377 static int
2378 kiblnd_hdev_setup_mrs(kib_hca_dev_t *hdev)
2379 {
2380         struct ib_mr *mr;
2381         int        i;
2382         int        rc;
2383         __u64    mm_size;
2384         __u64    mr_size;
2385         int        acflags = IB_ACCESS_LOCAL_WRITE |
2386                                 IB_ACCESS_REMOTE_WRITE;
2387
2388         rc = kiblnd_hdev_get_attr(hdev);
2389         if (rc != 0)
2390                 return rc;
2391
2392         if (hdev->ibh_mr_shift == 64) {
2393                 LIBCFS_ALLOC(hdev->ibh_mrs, 1 * sizeof(*hdev->ibh_mrs));
2394                 if (hdev->ibh_mrs == NULL) {
2395                         CERROR("Failed to allocate MRs table\n");
2396                         return -ENOMEM;
2397                 }
2398
2399                 hdev->ibh_mrs[0] = NULL;
2400                 hdev->ibh_nmrs   = 1;
2401
2402                 mr = ib_get_dma_mr(hdev->ibh_pd, acflags);
2403                 if (IS_ERR(mr)) {
2404                         CERROR("Failed ib_get_dma_mr : %ld\n", PTR_ERR(mr));
2405                         kiblnd_hdev_cleanup_mrs(hdev);
2406                         return PTR_ERR(mr);
2407                 }
2408
2409                 hdev->ibh_mrs[0] = mr;
2410
2411                 goto out;
2412         }
2413
2414         mr_size = 1ULL << hdev->ibh_mr_shift;
2415         mm_size = (unsigned long)high_memory - PAGE_OFFSET;
2416
2417         hdev->ibh_nmrs = (int)((mm_size + mr_size - 1) >> hdev->ibh_mr_shift);
2418
2419         if (hdev->ibh_mr_shift < 32 || hdev->ibh_nmrs > 1024) {
2420                 /* it's 4T..., assume we will re-code at that time */
2421                 CERROR("Can't support memory size: x%#llx with MR size: x%#llx\n",
2422                        mm_size, mr_size);
2423                 return -EINVAL;
2424         }
2425
2426         /* create an array of MRs to cover all memory */
2427         LIBCFS_ALLOC(hdev->ibh_mrs, sizeof(*hdev->ibh_mrs) * hdev->ibh_nmrs);
2428         if (hdev->ibh_mrs == NULL) {
2429                 CERROR("Failed to allocate MRs' table\n");
2430                 return -ENOMEM;
2431         }
2432
2433         for (i = 0; i < hdev->ibh_nmrs; i++) {
2434                 struct ib_phys_buf ipb;
2435                 __u64         iova;
2436
2437                 ipb.size = hdev->ibh_mr_size;
2438                 ipb.addr = i * mr_size;
2439                 iova     = ipb.addr;
2440
2441                 mr = ib_reg_phys_mr(hdev->ibh_pd, &ipb, 1, acflags, &iova);
2442                 if (IS_ERR(mr)) {
2443                         CERROR("Failed ib_reg_phys_mr addr %#llx size %#llx : %ld\n",
2444                                ipb.addr, ipb.size, PTR_ERR(mr));
2445                         kiblnd_hdev_cleanup_mrs(hdev);
2446                         return PTR_ERR(mr);
2447                 }
2448
2449                 LASSERT (iova == ipb.addr);
2450
2451                 hdev->ibh_mrs[i] = mr;
2452         }
2453
2454 out:
2455         if (hdev->ibh_mr_size != ~0ULL || hdev->ibh_nmrs != 1)
2456                 LCONSOLE_INFO("Register global MR array, MR size: %#llx, array size: %d\n",
2457                               hdev->ibh_mr_size, hdev->ibh_nmrs);
2458         return 0;
2459 }
2460
2461 static int
2462 kiblnd_dummy_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event)
2463 {       /* DUMMY */
2464         return 0;
2465 }
2466
2467 static int
2468 kiblnd_dev_need_failover(kib_dev_t *dev)
2469 {
2470         struct rdma_cm_id  *cmid;
2471         struct sockaddr_in  srcaddr;
2472         struct sockaddr_in  dstaddr;
2473         int              rc;
2474
2475         if (dev->ibd_hdev == NULL || /* initializing */
2476             dev->ibd_hdev->ibh_cmid == NULL || /* listener is dead */
2477             *kiblnd_tunables.kib_dev_failover > 1) /* debugging */
2478                 return 1;
2479
2480         /* XXX: it's UGLY, but I don't have better way to find
2481          * ib-bonding HCA failover because:
2482          *
2483          * a. no reliable CM event for HCA failover...
2484          * b. no OFED API to get ib_device for current net_device...
2485          *
2486          * We have only two choices at this point:
2487          *
2488          * a. rdma_bind_addr(), it will conflict with listener cmid
2489          * b. rdma_resolve_addr() to zero addr */
2490         cmid = kiblnd_rdma_create_id(kiblnd_dummy_callback, dev, RDMA_PS_TCP,
2491                                      IB_QPT_RC);
2492         if (IS_ERR(cmid)) {
2493                 rc = PTR_ERR(cmid);
2494                 CERROR("Failed to create cmid for failover: %d\n", rc);
2495                 return rc;
2496         }
2497
2498         memset(&srcaddr, 0, sizeof(srcaddr));
2499         srcaddr.sin_family      = AF_INET;
2500         srcaddr.sin_addr.s_addr = (__force u32)htonl(dev->ibd_ifip);
2501
2502         memset(&dstaddr, 0, sizeof(dstaddr));
2503         dstaddr.sin_family = AF_INET;
2504         rc = rdma_resolve_addr(cmid, (struct sockaddr *)&srcaddr,
2505                                (struct sockaddr *)&dstaddr, 1);
2506         if (rc != 0 || cmid->device == NULL) {
2507                 CERROR("Failed to bind %s:%pI4h to device(%p): %d\n",
2508                        dev->ibd_ifname, &dev->ibd_ifip,
2509                        cmid->device, rc);
2510                 rdma_destroy_id(cmid);
2511                 return rc;
2512         }
2513
2514         if (dev->ibd_hdev->ibh_ibdev == cmid->device) {
2515                 /* don't need device failover */
2516                 rdma_destroy_id(cmid);
2517                 return 0;
2518         }
2519
2520         return 1;
2521 }
2522
2523 int
2524 kiblnd_dev_failover(kib_dev_t *dev)
2525 {
2526         LIST_HEAD      (zombie_tpo);
2527         LIST_HEAD      (zombie_ppo);
2528         LIST_HEAD      (zombie_fpo);
2529         struct rdma_cm_id  *cmid  = NULL;
2530         kib_hca_dev_t      *hdev  = NULL;
2531         kib_hca_dev_t      *old;
2532         struct ib_pd       *pd;
2533         kib_net_t         *net;
2534         struct sockaddr_in  addr;
2535         unsigned long       flags;
2536         int              rc = 0;
2537         int                 i;
2538
2539         LASSERT (*kiblnd_tunables.kib_dev_failover > 1 ||
2540                  dev->ibd_can_failover ||
2541                  dev->ibd_hdev == NULL);
2542
2543         rc = kiblnd_dev_need_failover(dev);
2544         if (rc <= 0)
2545                 goto out;
2546
2547         if (dev->ibd_hdev != NULL &&
2548             dev->ibd_hdev->ibh_cmid != NULL) {
2549                 /* XXX it's not good to close old listener at here,
2550                  * because we can fail to create new listener.
2551                  * But we have to close it now, otherwise rdma_bind_addr
2552                  * will return EADDRINUSE... How crap! */
2553                 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2554
2555                 cmid = dev->ibd_hdev->ibh_cmid;
2556                 /* make next schedule of kiblnd_dev_need_failover()
2557                  * return 1 for me */
2558                 dev->ibd_hdev->ibh_cmid  = NULL;
2559                 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2560
2561                 rdma_destroy_id(cmid);
2562         }
2563
2564         cmid = kiblnd_rdma_create_id(kiblnd_cm_callback, dev, RDMA_PS_TCP,
2565                                      IB_QPT_RC);
2566         if (IS_ERR(cmid)) {
2567                 rc = PTR_ERR(cmid);
2568                 CERROR("Failed to create cmid for failover: %d\n", rc);
2569                 goto out;
2570         }
2571
2572         memset(&addr, 0, sizeof(addr));
2573         addr.sin_family      = AF_INET;
2574         addr.sin_addr.s_addr = (__force u32)htonl(dev->ibd_ifip);
2575         addr.sin_port   = htons(*kiblnd_tunables.kib_service);
2576
2577         /* Bind to failover device or port */
2578         rc = rdma_bind_addr(cmid, (struct sockaddr *)&addr);
2579         if (rc != 0 || cmid->device == NULL) {
2580                 CERROR("Failed to bind %s:%pI4h to device(%p): %d\n",
2581                        dev->ibd_ifname, &dev->ibd_ifip,
2582                        cmid->device, rc);
2583                 rdma_destroy_id(cmid);
2584                 goto out;
2585         }
2586
2587         LIBCFS_ALLOC(hdev, sizeof(*hdev));
2588         if (hdev == NULL) {
2589                 CERROR("Failed to allocate kib_hca_dev\n");
2590                 rdma_destroy_id(cmid);
2591                 rc = -ENOMEM;
2592                 goto out;
2593         }
2594
2595         atomic_set(&hdev->ibh_ref, 1);
2596         hdev->ibh_dev   = dev;
2597         hdev->ibh_cmid  = cmid;
2598         hdev->ibh_ibdev = cmid->device;
2599
2600         pd = ib_alloc_pd(cmid->device);
2601         if (IS_ERR(pd)) {
2602                 rc = PTR_ERR(pd);
2603                 CERROR("Can't allocate PD: %d\n", rc);
2604                 goto out;
2605         }
2606
2607         hdev->ibh_pd = pd;
2608
2609         rc = rdma_listen(cmid, 0);
2610         if (rc != 0) {
2611                 CERROR("Can't start new listener: %d\n", rc);
2612                 goto out;
2613         }
2614
2615         rc = kiblnd_hdev_setup_mrs(hdev);
2616         if (rc != 0) {
2617                 CERROR("Can't setup device: %d\n", rc);
2618                 goto out;
2619         }
2620
2621         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2622
2623         old = dev->ibd_hdev;
2624         dev->ibd_hdev = hdev; /* take over the refcount */
2625         hdev = old;
2626
2627         list_for_each_entry(net, &dev->ibd_nets, ibn_list) {
2628                 cfs_cpt_for_each(i, lnet_cpt_table()) {
2629                         kiblnd_fail_poolset(&net->ibn_tx_ps[i]->tps_poolset,
2630                                             &zombie_tpo);
2631
2632                         if (net->ibn_fmr_ps != NULL) {
2633                                 kiblnd_fail_fmr_poolset(net->ibn_fmr_ps[i],
2634                                                         &zombie_fpo);
2635
2636                         } else if (net->ibn_pmr_ps != NULL) {
2637                                 kiblnd_fail_poolset(&net->ibn_pmr_ps[i]->
2638                                                     pps_poolset, &zombie_ppo);
2639                         }
2640                 }
2641         }
2642
2643         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2644  out:
2645         if (!list_empty(&zombie_tpo))
2646                 kiblnd_destroy_pool_list(&zombie_tpo);
2647         if (!list_empty(&zombie_ppo))
2648                 kiblnd_destroy_pool_list(&zombie_ppo);
2649         if (!list_empty(&zombie_fpo))
2650                 kiblnd_destroy_fmr_pool_list(&zombie_fpo);
2651         if (hdev != NULL)
2652                 kiblnd_hdev_decref(hdev);
2653
2654         if (rc != 0)
2655                 dev->ibd_failed_failover++;
2656         else
2657                 dev->ibd_failed_failover = 0;
2658
2659         return rc;
2660 }
2661
2662 void
2663 kiblnd_destroy_dev (kib_dev_t *dev)
2664 {
2665         LASSERT (dev->ibd_nnets == 0);
2666         LASSERT (list_empty(&dev->ibd_nets));
2667
2668         list_del(&dev->ibd_fail_list);
2669         list_del(&dev->ibd_list);
2670
2671         if (dev->ibd_hdev != NULL)
2672                 kiblnd_hdev_decref(dev->ibd_hdev);
2673
2674         LIBCFS_FREE(dev, sizeof(*dev));
2675 }
2676
2677 static kib_dev_t *
2678 kiblnd_create_dev(char *ifname)
2679 {
2680         struct net_device *netdev;
2681         kib_dev_t        *dev;
2682         __u32         netmask;
2683         __u32         ip;
2684         int             up;
2685         int             rc;
2686
2687         rc = libcfs_ipif_query(ifname, &up, &ip, &netmask);
2688         if (rc != 0) {
2689                 CERROR("Can't query IPoIB interface %s: %d\n",
2690                        ifname, rc);
2691                 return NULL;
2692         }
2693
2694         if (!up) {
2695                 CERROR("Can't query IPoIB interface %s: it's down\n", ifname);
2696                 return NULL;
2697         }
2698
2699         LIBCFS_ALLOC(dev, sizeof(*dev));
2700         if (dev == NULL)
2701                 return NULL;
2702
2703         netdev = dev_get_by_name(&init_net, ifname);
2704         if (netdev == NULL) {
2705                 dev->ibd_can_failover = 0;
2706         } else {
2707                 dev->ibd_can_failover = !!(netdev->flags & IFF_MASTER);
2708                 dev_put(netdev);
2709         }
2710
2711         INIT_LIST_HEAD(&dev->ibd_nets);
2712         INIT_LIST_HEAD(&dev->ibd_list); /* not yet in kib_devs */
2713         INIT_LIST_HEAD(&dev->ibd_fail_list);
2714         dev->ibd_ifip = ip;
2715         strcpy(&dev->ibd_ifname[0], ifname);
2716
2717         /* initialize the device */
2718         rc = kiblnd_dev_failover(dev);
2719         if (rc != 0) {
2720                 CERROR("Can't initialize device: %d\n", rc);
2721                 LIBCFS_FREE(dev, sizeof(*dev));
2722                 return NULL;
2723         }
2724
2725         list_add_tail(&dev->ibd_list,
2726                           &kiblnd_data.kib_devs);
2727         return dev;
2728 }
2729
2730 static void
2731 kiblnd_base_shutdown(void)
2732 {
2733         struct kib_sched_info   *sched;
2734         int                     i;
2735
2736         LASSERT (list_empty(&kiblnd_data.kib_devs));
2737
2738         CDEBUG(D_MALLOC, "before LND base cleanup: kmem %d\n",
2739                atomic_read(&libcfs_kmemory));
2740
2741         switch (kiblnd_data.kib_init) {
2742         default:
2743                 LBUG();
2744
2745         case IBLND_INIT_ALL:
2746         case IBLND_INIT_DATA:
2747                 LASSERT (kiblnd_data.kib_peers != NULL);
2748                 for (i = 0; i < kiblnd_data.kib_peer_hash_size; i++) {
2749                         LASSERT (list_empty(&kiblnd_data.kib_peers[i]));
2750                 }
2751                 LASSERT (list_empty(&kiblnd_data.kib_connd_zombies));
2752                 LASSERT (list_empty(&kiblnd_data.kib_connd_conns));
2753
2754                 /* flag threads to terminate; wake and wait for them to die */
2755                 kiblnd_data.kib_shutdown = 1;
2756
2757                 /* NB: we really want to stop scheduler threads net by net
2758                  * instead of the whole module, this should be improved
2759                  * with dynamic configuration LNet */
2760                 cfs_percpt_for_each(sched, i, kiblnd_data.kib_scheds)
2761                         wake_up_all(&sched->ibs_waitq);
2762
2763                 wake_up_all(&kiblnd_data.kib_connd_waitq);
2764                 wake_up_all(&kiblnd_data.kib_failover_waitq);
2765
2766                 i = 2;
2767                 while (atomic_read(&kiblnd_data.kib_nthreads) != 0) {
2768                         i++;
2769                         CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET, /* power of 2? */
2770                                "Waiting for %d threads to terminate\n",
2771                                atomic_read(&kiblnd_data.kib_nthreads));
2772                         set_current_state(TASK_UNINTERRUPTIBLE);
2773                         schedule_timeout(cfs_time_seconds(1));
2774                 }
2775
2776                 /* fall through */
2777
2778         case IBLND_INIT_NOTHING:
2779                 break;
2780         }
2781
2782         if (kiblnd_data.kib_peers != NULL) {
2783                 LIBCFS_FREE(kiblnd_data.kib_peers,
2784                             sizeof(struct list_head) *
2785                             kiblnd_data.kib_peer_hash_size);
2786         }
2787
2788         if (kiblnd_data.kib_scheds != NULL)
2789                 cfs_percpt_free(kiblnd_data.kib_scheds);
2790
2791         CDEBUG(D_MALLOC, "after LND base cleanup: kmem %d\n",
2792                atomic_read(&libcfs_kmemory));
2793
2794         kiblnd_data.kib_init = IBLND_INIT_NOTHING;
2795         module_put(THIS_MODULE);
2796 }
2797
2798 void
2799 kiblnd_shutdown (lnet_ni_t *ni)
2800 {
2801         kib_net_t       *net = ni->ni_data;
2802         rwlock_t     *g_lock = &kiblnd_data.kib_global_lock;
2803         int            i;
2804         unsigned long     flags;
2805
2806         LASSERT(kiblnd_data.kib_init == IBLND_INIT_ALL);
2807
2808         if (net == NULL)
2809                 goto out;
2810
2811         CDEBUG(D_MALLOC, "before LND net cleanup: kmem %d\n",
2812                atomic_read(&libcfs_kmemory));
2813
2814         write_lock_irqsave(g_lock, flags);
2815         net->ibn_shutdown = 1;
2816         write_unlock_irqrestore(g_lock, flags);
2817
2818         switch (net->ibn_init) {
2819         default:
2820                 LBUG();
2821
2822         case IBLND_INIT_ALL:
2823                 /* nuke all existing peers within this net */
2824                 kiblnd_del_peer(ni, LNET_NID_ANY);
2825
2826                 /* Wait for all peer state to clean up */
2827                 i = 2;
2828                 while (atomic_read(&net->ibn_npeers) != 0) {
2829                         i++;
2830                         CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET, /* 2**n? */
2831                                "%s: waiting for %d peers to disconnect\n",
2832                                libcfs_nid2str(ni->ni_nid),
2833                                atomic_read(&net->ibn_npeers));
2834                         set_current_state(TASK_UNINTERRUPTIBLE);
2835                         schedule_timeout(cfs_time_seconds(1));
2836                 }
2837
2838                 kiblnd_net_fini_pools(net);
2839
2840                 write_lock_irqsave(g_lock, flags);
2841                 LASSERT(net->ibn_dev->ibd_nnets > 0);
2842                 net->ibn_dev->ibd_nnets--;
2843                 list_del(&net->ibn_list);
2844                 write_unlock_irqrestore(g_lock, flags);
2845
2846                 /* fall through */
2847
2848         case IBLND_INIT_NOTHING:
2849                 LASSERT (atomic_read(&net->ibn_nconns) == 0);
2850
2851                 if (net->ibn_dev != NULL &&
2852                     net->ibn_dev->ibd_nnets == 0)
2853                         kiblnd_destroy_dev(net->ibn_dev);
2854
2855                 break;
2856         }
2857
2858         CDEBUG(D_MALLOC, "after LND net cleanup: kmem %d\n",
2859                atomic_read(&libcfs_kmemory));
2860
2861         net->ibn_init = IBLND_INIT_NOTHING;
2862         ni->ni_data = NULL;
2863
2864         LIBCFS_FREE(net, sizeof(*net));
2865
2866 out:
2867         if (list_empty(&kiblnd_data.kib_devs))
2868                 kiblnd_base_shutdown();
2869         return;
2870 }
2871
2872 static int
2873 kiblnd_base_startup(void)
2874 {
2875         struct kib_sched_info   *sched;
2876         int                     rc;
2877         int                     i;
2878
2879         LASSERT (kiblnd_data.kib_init == IBLND_INIT_NOTHING);
2880
2881         try_module_get(THIS_MODULE);
2882         memset(&kiblnd_data, 0, sizeof(kiblnd_data)); /* zero pointers, flags etc */
2883
2884         rwlock_init(&kiblnd_data.kib_global_lock);
2885
2886         INIT_LIST_HEAD(&kiblnd_data.kib_devs);
2887         INIT_LIST_HEAD(&kiblnd_data.kib_failed_devs);
2888
2889         kiblnd_data.kib_peer_hash_size = IBLND_PEER_HASH_SIZE;
2890         LIBCFS_ALLOC(kiblnd_data.kib_peers,
2891                      sizeof(struct list_head) *
2892                             kiblnd_data.kib_peer_hash_size);
2893         if (kiblnd_data.kib_peers == NULL) {
2894                 goto failed;
2895         }
2896         for (i = 0; i < kiblnd_data.kib_peer_hash_size; i++)
2897                 INIT_LIST_HEAD(&kiblnd_data.kib_peers[i]);
2898
2899         spin_lock_init(&kiblnd_data.kib_connd_lock);
2900         INIT_LIST_HEAD(&kiblnd_data.kib_connd_conns);
2901         INIT_LIST_HEAD(&kiblnd_data.kib_connd_zombies);
2902         init_waitqueue_head(&kiblnd_data.kib_connd_waitq);
2903         init_waitqueue_head(&kiblnd_data.kib_failover_waitq);
2904
2905         kiblnd_data.kib_scheds = cfs_percpt_alloc(lnet_cpt_table(),
2906                                                   sizeof(*sched));
2907         if (kiblnd_data.kib_scheds == NULL)
2908                 goto failed;
2909
2910         cfs_percpt_for_each(sched, i, kiblnd_data.kib_scheds) {
2911                 int     nthrs;
2912
2913                 spin_lock_init(&sched->ibs_lock);
2914                 INIT_LIST_HEAD(&sched->ibs_conns);
2915                 init_waitqueue_head(&sched->ibs_waitq);
2916
2917                 nthrs = cfs_cpt_weight(lnet_cpt_table(), i);
2918                 if (*kiblnd_tunables.kib_nscheds > 0) {
2919                         nthrs = min(nthrs, *kiblnd_tunables.kib_nscheds);
2920                 } else {
2921                         /* max to half of CPUs, another half is reserved for
2922                          * upper layer modules */
2923                         nthrs = min(max(IBLND_N_SCHED, nthrs >> 1), nthrs);
2924                 }
2925
2926                 sched->ibs_nthreads_max = nthrs;
2927                 sched->ibs_cpt = i;
2928         }
2929
2930         kiblnd_data.kib_error_qpa.qp_state = IB_QPS_ERR;
2931
2932         /* lists/ptrs/locks initialised */
2933         kiblnd_data.kib_init = IBLND_INIT_DATA;
2934         /*****************************************************/
2935
2936         rc = kiblnd_thread_start(kiblnd_connd, NULL, "kiblnd_connd");
2937         if (rc != 0) {
2938                 CERROR("Can't spawn o2iblnd connd: %d\n", rc);
2939                 goto failed;
2940         }
2941
2942         if (*kiblnd_tunables.kib_dev_failover != 0)
2943                 rc = kiblnd_thread_start(kiblnd_failover_thread, NULL,
2944                                          "kiblnd_failover");
2945
2946         if (rc != 0) {
2947                 CERROR("Can't spawn o2iblnd failover thread: %d\n", rc);
2948                 goto failed;
2949         }
2950
2951         /* flag everything initialised */
2952         kiblnd_data.kib_init = IBLND_INIT_ALL;
2953         /*****************************************************/
2954
2955         return 0;
2956
2957  failed:
2958         kiblnd_base_shutdown();
2959         return -ENETDOWN;
2960 }
2961
2962 static int
2963 kiblnd_start_schedulers(struct kib_sched_info *sched)
2964 {
2965         int     rc = 0;
2966         int     nthrs;
2967         int     i;
2968
2969         if (sched->ibs_nthreads == 0) {
2970                 if (*kiblnd_tunables.kib_nscheds > 0) {
2971                         nthrs = sched->ibs_nthreads_max;
2972                 } else {
2973                         nthrs = cfs_cpt_weight(lnet_cpt_table(),
2974                                                sched->ibs_cpt);
2975                         nthrs = min(max(IBLND_N_SCHED, nthrs >> 1), nthrs);
2976                         nthrs = min(IBLND_N_SCHED_HIGH, nthrs);
2977                 }
2978         } else {
2979                 LASSERT(sched->ibs_nthreads <= sched->ibs_nthreads_max);
2980                 /* increase one thread if there is new interface */
2981                 nthrs = (sched->ibs_nthreads < sched->ibs_nthreads_max);
2982         }
2983
2984         for (i = 0; i < nthrs; i++) {
2985                 long    id;
2986                 char    name[20];
2987                 id = KIB_THREAD_ID(sched->ibs_cpt, sched->ibs_nthreads + i);
2988                 snprintf(name, sizeof(name), "kiblnd_sd_%02ld_%02ld",
2989                          KIB_THREAD_CPT(id), KIB_THREAD_TID(id));
2990                 rc = kiblnd_thread_start(kiblnd_scheduler, (void *)id, name);
2991                 if (rc == 0)
2992                         continue;
2993
2994                 CERROR("Can't spawn thread %d for scheduler[%d]: %d\n",
2995                        sched->ibs_cpt, sched->ibs_nthreads + i, rc);
2996                 break;
2997         }
2998
2999         sched->ibs_nthreads += i;
3000         return rc;
3001 }
3002
3003 static int
3004 kiblnd_dev_start_threads(kib_dev_t *dev, int newdev, __u32 *cpts, int ncpts)
3005 {
3006         int     cpt;
3007         int     rc;
3008         int     i;
3009
3010         for (i = 0; i < ncpts; i++) {
3011                 struct kib_sched_info *sched;
3012
3013                 cpt = (cpts == NULL) ? i : cpts[i];
3014                 sched = kiblnd_data.kib_scheds[cpt];
3015
3016                 if (!newdev && sched->ibs_nthreads > 0)
3017                         continue;
3018
3019                 rc = kiblnd_start_schedulers(kiblnd_data.kib_scheds[cpt]);
3020                 if (rc != 0) {
3021                         CERROR("Failed to start scheduler threads for %s\n",
3022                                dev->ibd_ifname);
3023                         return rc;
3024                 }
3025         }
3026         return 0;
3027 }
3028
3029 static kib_dev_t *
3030 kiblnd_dev_search(char *ifname)
3031 {
3032         kib_dev_t       *alias = NULL;
3033         kib_dev_t       *dev;
3034         char            *colon;
3035         char            *colon2;
3036
3037         colon = strchr(ifname, ':');
3038         list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
3039                 if (strcmp(&dev->ibd_ifname[0], ifname) == 0)
3040                         return dev;
3041
3042                 if (alias != NULL)
3043                         continue;
3044
3045                 colon2 = strchr(dev->ibd_ifname, ':');
3046                 if (colon != NULL)
3047                         *colon = 0;
3048                 if (colon2 != NULL)
3049                         *colon2 = 0;
3050
3051                 if (strcmp(&dev->ibd_ifname[0], ifname) == 0)
3052                         alias = dev;
3053
3054                 if (colon != NULL)
3055                         *colon = ':';
3056                 if (colon2 != NULL)
3057                         *colon2 = ':';
3058         }
3059         return alias;
3060 }
3061
3062 int
3063 kiblnd_startup (lnet_ni_t *ni)
3064 {
3065         char                 *ifname;
3066         kib_dev_t               *ibdev = NULL;
3067         kib_net_t               *net;
3068         struct timeval      tv;
3069         unsigned long        flags;
3070         int                    rc;
3071         int                       newdev;
3072
3073         LASSERT (ni->ni_lnd == &the_o2iblnd);
3074
3075         if (kiblnd_data.kib_init == IBLND_INIT_NOTHING) {
3076                 rc = kiblnd_base_startup();
3077                 if (rc != 0)
3078                         return rc;
3079         }
3080
3081         LIBCFS_ALLOC(net, sizeof(*net));
3082         ni->ni_data = net;
3083         if (net == NULL)
3084                 goto net_failed;
3085
3086         do_gettimeofday(&tv);
3087         net->ibn_incarnation = (((__u64)tv.tv_sec) * 1000000) + tv.tv_usec;
3088
3089         ni->ni_peertimeout    = *kiblnd_tunables.kib_peertimeout;
3090         ni->ni_maxtxcredits   = *kiblnd_tunables.kib_credits;
3091         ni->ni_peertxcredits  = *kiblnd_tunables.kib_peertxcredits;
3092         ni->ni_peerrtrcredits = *kiblnd_tunables.kib_peerrtrcredits;
3093
3094         if (ni->ni_interfaces[0] != NULL) {
3095                 /* Use the IPoIB interface specified in 'networks=' */
3096
3097                 CLASSERT (LNET_MAX_INTERFACES > 1);
3098                 if (ni->ni_interfaces[1] != NULL) {
3099                         CERROR("Multiple interfaces not supported\n");
3100                         goto failed;
3101                 }
3102
3103                 ifname = ni->ni_interfaces[0];
3104         } else {
3105                 ifname = *kiblnd_tunables.kib_default_ipif;
3106         }
3107
3108         if (strlen(ifname) >= sizeof(ibdev->ibd_ifname)) {
3109                 CERROR("IPoIB interface name too long: %s\n", ifname);
3110                 goto failed;
3111         }
3112
3113         ibdev = kiblnd_dev_search(ifname);
3114
3115         newdev = ibdev == NULL;
3116         /* hmm...create kib_dev even for alias */
3117         if (ibdev == NULL || strcmp(&ibdev->ibd_ifname[0], ifname) != 0)
3118                 ibdev = kiblnd_create_dev(ifname);
3119
3120         if (ibdev == NULL)
3121                 goto failed;
3122
3123         net->ibn_dev = ibdev;
3124         ni->ni_nid = LNET_MKNID(LNET_NIDNET(ni->ni_nid), ibdev->ibd_ifip);
3125
3126         rc = kiblnd_dev_start_threads(ibdev, newdev,
3127                                       ni->ni_cpts, ni->ni_ncpts);
3128         if (rc != 0)
3129                 goto failed;
3130
3131         rc = kiblnd_net_init_pools(net, ni->ni_cpts, ni->ni_ncpts);
3132         if (rc != 0) {
3133                 CERROR("Failed to initialize NI pools: %d\n", rc);
3134                 goto failed;
3135         }
3136
3137         write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
3138         ibdev->ibd_nnets++;
3139         list_add_tail(&net->ibn_list, &ibdev->ibd_nets);
3140         write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
3141
3142         net->ibn_init = IBLND_INIT_ALL;
3143
3144         return 0;
3145
3146 failed:
3147         if (net->ibn_dev == NULL && ibdev != NULL)
3148                 kiblnd_destroy_dev(ibdev);
3149
3150 net_failed:
3151         kiblnd_shutdown(ni);
3152
3153         CDEBUG(D_NET, "kiblnd_startup failed\n");
3154         return -ENETDOWN;
3155 }
3156
3157 static void __exit
3158 kiblnd_module_fini (void)
3159 {
3160         lnet_unregister_lnd(&the_o2iblnd);
3161 }
3162
3163 static int __init
3164 kiblnd_module_init (void)
3165 {
3166         int    rc;
3167
3168         CLASSERT (sizeof(kib_msg_t) <= IBLND_MSG_SIZE);
3169         CLASSERT (offsetof(kib_msg_t, ibm_u.get.ibgm_rd.rd_frags[IBLND_MAX_RDMA_FRAGS])
3170                   <= IBLND_MSG_SIZE);
3171         CLASSERT (offsetof(kib_msg_t, ibm_u.putack.ibpam_rd.rd_frags[IBLND_MAX_RDMA_FRAGS])
3172                   <= IBLND_MSG_SIZE);
3173
3174         rc = kiblnd_tunables_init();
3175         if (rc != 0)
3176                 return rc;
3177
3178         lnet_register_lnd(&the_o2iblnd);
3179
3180         return 0;
3181 }
3182
3183 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
3184 MODULE_DESCRIPTION("Kernel OpenIB gen2 LND v2.00");
3185 MODULE_LICENSE("GPL");
3186
3187 module_init(kiblnd_module_init);
3188 module_exit(kiblnd_module_fini);