lustre/ptlrpc/events.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  * GPL HEADER START
   5  *
   6  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 only,
  10  * as published by the Free Software Foundation.
  11  *
  12  * This program is distributed in the hope that it will be useful, but
  13  * WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * General Public License version 2 for more details (a copy is included
  16  * in the LICENSE file that accompanied this code).
  17  *
  18  * You should have received a copy of the GNU General Public License
  19  * version 2 along with this program; If not, see
  20  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  21  *
  22  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  23  * CA 95054 USA or visit www.sun.com if you need additional information or
  24  * have any questions.
  25  *
  26  * GPL HEADER END
  27  */
  28 /*
  29  * Copyright  2008 Sun Microsystems, Inc. All rights reserved
  30  * Use is subject to license terms.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  */
  36
  37 #define DEBUG_SUBSYSTEM S_RPC
  38
  39 #ifndef __KERNEL__
  40 # include <liblustre.h>
  41 #else
  42 # ifdef __mips64__
  43 #  include <linux/kernel.h>
  44 # endif
  45 #endif
  46 #include <obd_class.h>
  47 #include <lustre_net.h>
  48 #include "ptlrpc_internal.h"
  49
  50 lnet_handle_eq_t   ptlrpc_eq_h;
  51
  52 /*
  53  *  Client's outgoing request callback
  54  */
  55 void request_out_callback(lnet_event_t *ev)
  56 {
  57         struct ptlrpc_cb_id   *cbid = ev->md.user_ptr;
  58         struct ptlrpc_request *req = cbid->cbid_arg;
  59         ENTRY;
  60
  61         LASSERT (ev->type == LNET_EVENT_SEND ||
  62                  ev->type == LNET_EVENT_UNLINK);
  63         LASSERT (ev->unlinked);
  64
  65         DEBUG_REQ((ev->status == 0) ? D_NET : D_ERROR, req,
  66                   "type %d, status %d", ev->type, ev->status);
  67
  68         if (ev->type == LNET_EVENT_UNLINK || ev->status != 0) {
  69
  70                 /* Failed send: make it seem like the reply timed out, just
  71                  * like failing sends in client.c does currently...  */
  72
  73                 spin_lock(&req->rq_lock);
  74                 req->rq_net_err = 1;
  75                 spin_unlock(&req->rq_lock);
  76
  77                 ptlrpc_client_wake_req(req);
  78         }
  79
  80         ptlrpc_req_finished(req);
  81
  82         EXIT;
  83 }
  84
  85 /*
  86  * Client's incoming reply callback
  87  */
  88 void reply_in_callback(lnet_event_t *ev)
  89 {
  90         struct ptlrpc_cb_id   *cbid = ev->md.user_ptr;
  91         struct ptlrpc_request *req = cbid->cbid_arg;
  92         ENTRY;
  93
  94         DEBUG_REQ((ev->status == 0) ? D_NET : D_ERROR, req,
  95                   "type %d, status %d", ev->type, ev->status);
  96
  97         LASSERT(ev->type == LNET_EVENT_PUT || ev->type == LNET_EVENT_UNLINK);
  98         LASSERT(ev->md.start == req->rq_repbuf);
  99         LASSERT(ev->mlength <= req->rq_replen);
 100         /* We've set LNET_MD_MANAGE_REMOTE for all outgoing requests
 101            for adaptive timeouts' early reply. */
 102         LASSERT((ev->md.options & LNET_MD_MANAGE_REMOTE) != 0);
 103
 104         spin_lock(&req->rq_lock);
 105
 106         req->rq_receiving_reply = 0;
 107         req->rq_early = 0;
 108         if (ev->unlinked)
 109                 req->rq_must_unlink = 0;
 110
 111         if (ev->status)
 112                 goto out_wake;
 113
 114         if (ev->type == LNET_EVENT_UNLINK) {
 115                 LASSERT(ev->unlinked);
 116                 DEBUG_REQ(D_RPCTRACE, req, "unlink");
 117                 goto out_wake;
 118         }
 119
 120         if (ev->mlength < ev->rlength ) {
 121                 CDEBUG(D_RPCTRACE, "truncate req %p rpc %d - %d+%d\n", req,
 122                        req->rq_replen, ev->rlength, ev->offset);
 123                 req->rq_reply_truncate = 1;
 124                 req->rq_replied = 1;
 125                 req->rq_status = -EOVERFLOW;
 126                 req->rq_nob_received = ev->rlength + ev->offset;
 127                 goto out_wake;
 128         }
 129
 130         if ((ev->offset == 0) &&
 131             (lustre_msghdr_get_flags(req->rq_reqmsg) & MSGHDR_AT_SUPPORT)) {
 132                 /* Early reply */
 133                 DEBUG_REQ(D_ADAPTTO, req,
 134                           "Early reply received: mlen=%u offset=%d replen=%d "
 135                           "replied=%d unlinked=%d", ev->mlength, ev->offset,
 136                           req->rq_replen, req->rq_replied, ev->unlinked);
 137
 138                 if (unlikely(ev->mlength != lustre_msg_early_size(req)))
 139                         CERROR("early reply sized %u, expect %u\n",
 140                                ev->mlength, lustre_msg_early_size(req));
 141
 142                 req->rq_early_count++; /* number received, client side */
 143
 144                 if (req->rq_replied)   /* already got the real reply */
 145                         goto out_wake;
 146
 147                 req->rq_early = 1;
 148                 req->rq_nob_received = ev->mlength;
 149                 /* repmsg points to early reply */
 150                 req->rq_repmsg = req->rq_repbuf;
 151                 /* And we're still receiving */
 152                 req->rq_receiving_reply = 1;
 153         } else {
 154                 /* Real reply */
 155                 req->rq_rep_swab_mask = 0;
 156                 req->rq_replied = 1;
 157                 req->rq_nob_received = ev->mlength;
 158                 /* repmsg points to real reply */
 159                 req->rq_repmsg = (struct lustre_msg *)((char *)req->rq_repbuf +
 160                                                        ev->offset);
 161                 /* LNetMDUnlink can't be called under the LNET_LOCK,
 162                    so we must unlink in ptlrpc_unregister_reply */
 163                 DEBUG_REQ(D_INFO, req,
 164                           "reply in flags=%x mlen=%u offset=%d replen=%d",
 165                           lustre_msg_get_flags(req->rq_reqmsg),
 166                           ev->mlength, ev->offset, req->rq_replen);
 167         }
 168
 169         req->rq_import->imp_last_reply_time = cfs_time_current_sec();
 170
 171 out_wake:
 172         /* NB don't unlock till after wakeup; req can disappear under us
 173          * since we don't have our own ref */
 174         ptlrpc_client_wake_req(req);
 175         spin_unlock(&req->rq_lock);
 176         EXIT;
 177 }
 178
 179 /*
 180  * Client's bulk has been written/read
 181  */
 182 void client_bulk_callback (lnet_event_t *ev)
 183 {
 184         struct ptlrpc_cb_id     *cbid = ev->md.user_ptr;
 185         struct ptlrpc_bulk_desc *desc = cbid->cbid_arg;
 186         ENTRY;
 187
 188         if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_CLIENT_BULK_CB))
 189                 ev->status = -EIO;
 190
 191         LASSERT ((desc->bd_type == BULK_PUT_SINK &&
 192                   ev->type == LNET_EVENT_PUT) ||
 193                  (desc->bd_type == BULK_GET_SOURCE &&
 194                   ev->type == LNET_EVENT_GET) ||
 195                  ev->type == LNET_EVENT_UNLINK);
 196         LASSERT (ev->unlinked);
 197
 198         CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
 199                "event type %d, status %d, desc %p\n",
 200                ev->type, ev->status, desc);
 201
 202         spin_lock(&desc->bd_lock);
 203
 204         LASSERT(desc->bd_network_rw);
 205         desc->bd_network_rw = 0;
 206
 207         if (ev->type != LNET_EVENT_UNLINK && ev->status == 0) {
 208                 desc->bd_success = 1;
 209                 desc->bd_nob_transferred = ev->mlength;
 210                 desc->bd_sender = ev->sender;
 211         }
 212
 213         /* NB don't unlock till after wakeup; desc can disappear under us
 214          * otherwise */
 215         ptlrpc_client_wake_req(desc->bd_req);
 216
 217         spin_unlock(&desc->bd_lock);
 218         EXIT;
 219 }
 220
 221 /*
 222  * Server's incoming request callback
 223  */
 224 void request_in_callback(lnet_event_t *ev)
 225 {
 226         struct ptlrpc_cb_id               *cbid = ev->md.user_ptr;
 227         struct ptlrpc_request_buffer_desc *rqbd = cbid->cbid_arg;
 228         struct ptlrpc_service             *service = rqbd->rqbd_service;
 229         struct ptlrpc_request             *req;
 230         ENTRY;
 231
 232         LASSERT (ev->type == LNET_EVENT_PUT ||
 233                  ev->type == LNET_EVENT_UNLINK);
 234         LASSERT ((char *)ev->md.start >= rqbd->rqbd_buffer);
 235         LASSERT ((char *)ev->md.start + ev->offset + ev->mlength <=
 236                  rqbd->rqbd_buffer + service->srv_buf_size);
 237
 238         CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
 239                "event type %d, status %d, service %s\n",
 240                ev->type, ev->status, service->srv_name);
 241
 242         if (ev->unlinked) {
 243                 /* If this is the last request message to fit in the
 244                  * request buffer we can use the request object embedded in
 245                  * rqbd.  Note that if we failed to allocate a request,
 246                  * we'd have to re-post the rqbd, which we can't do in this
 247                  * context. */
 248                 req = &rqbd->rqbd_req;
 249                 memset(req, 0, sizeof (*req));
 250         } else {
 251                 LASSERT (ev->type == LNET_EVENT_PUT);
 252                 if (ev->status != 0) {
 253                         /* We moaned above already... */
 254                         return;
 255                 }
 256                 OBD_ALLOC_GFP(req, sizeof(*req), CFS_ALLOC_ATOMIC_TRY);
 257                 if (req == NULL) {
 258                         CERROR("Can't allocate incoming request descriptor: "
 259                                "Dropping %s RPC from %s\n",
 260                                service->srv_name,
 261                                libcfs_id2str(ev->initiator));
 262                         return;
 263                 }
 264         }
 265
 266         /* NB we ABSOLUTELY RELY on req being zeroed, so pointers are NULL,
 267          * flags are reset and scalars are zero.  We only set the message
 268          * size to non-zero if this was a successful receive. */
 269         req->rq_xid = ev->match_bits;
 270         req->rq_reqmsg = ev->md.start + ev->offset;
 271         if (ev->type == LNET_EVENT_PUT && ev->status == 0)
 272                 req->rq_reqlen = ev->mlength;
 273         do_gettimeofday(&req->rq_arrival_time);
 274         req->rq_peer = ev->initiator;
 275         req->rq_self = ev->target.nid;
 276         req->rq_rqbd = rqbd;
 277         req->rq_phase = RQ_PHASE_NEW;
 278 #ifdef CRAY_XT3
 279         req->rq_uid = ev->uid;
 280 #endif
 281         spin_lock_init(&req->rq_lock);
 282         CFS_INIT_LIST_HEAD(&req->rq_timed_list);
 283         atomic_set(&req->rq_refcount, 1);
 284         if (ev->type == LNET_EVENT_PUT)
 285                 DEBUG_REQ(D_RPCTRACE, req, "incoming req");
 286
 287         spin_lock(&service->srv_lock);
 288
 289         req->rq_history_seq = service->srv_request_seq++;
 290         list_add_tail(&req->rq_history_list, &service->srv_request_history);
 291
 292         if (ev->unlinked) {
 293                 service->srv_nrqbd_receiving--;
 294                 CDEBUG(D_INFO, "Buffer complete: %d buffers still posted\n",
 295                        service->srv_nrqbd_receiving);
 296
 297                 /* Normally, don't complain about 0 buffers posted; LNET won't
 298                  * drop incoming reqs since we set the portal lazy */
 299                 if (test_req_buffer_pressure &&
 300                     ev->type != LNET_EVENT_UNLINK &&
 301                     service->srv_nrqbd_receiving == 0)
 302                         CWARN("All %s request buffers busy\n",
 303                               service->srv_name);
 304
 305                 /* req takes over the network's ref on rqbd */
 306         } else {
 307                 /* req takes a ref on rqbd */
 308                 rqbd->rqbd_refcount++;
 309         }
 310
 311         list_add_tail(&req->rq_list, &service->srv_req_in_queue);
 312         service->srv_n_queued_reqs++;
 313
 314         /* NB everything can disappear under us once the request
 315          * has been queued and we unlock, so do the wake now... */
 316         cfs_waitq_signal(&service->srv_waitq);
 317
 318         spin_unlock(&service->srv_lock);
 319         EXIT;
 320 }
 321
 322 /*
 323  *  Server's outgoing reply callback
 324  */
 325 void reply_out_callback(lnet_event_t *ev)
 326 {
 327         struct ptlrpc_cb_id       *cbid = ev->md.user_ptr;
 328         struct ptlrpc_reply_state *rs = cbid->cbid_arg;
 329         struct ptlrpc_service     *svc = rs->rs_service;
 330         ENTRY;
 331
 332         LASSERT (ev->type == LNET_EVENT_SEND ||
 333                  ev->type == LNET_EVENT_ACK ||
 334                  ev->type == LNET_EVENT_UNLINK);
 335
 336         if (!rs->rs_difficult) {
 337                 /* 'Easy' replies have no further processing so I drop the
 338                  * net's ref on 'rs' */
 339                 LASSERT (ev->unlinked);
 340                 ptlrpc_rs_decref(rs);
 341                 atomic_dec (&svc->srv_outstanding_replies);
 342                 EXIT;
 343                 return;
 344         }
 345
 346         LASSERT (rs->rs_on_net);
 347
 348         if (ev->unlinked) {
 349                 /* Last network callback.  The net's ref on 'rs' stays put
 350                  * until ptlrpc_server_handle_reply() is done with it */
 351                 spin_lock(&svc->srv_lock);
 352                 rs->rs_on_net = 0;
 353                 ptlrpc_schedule_difficult_reply (rs);
 354                 spin_unlock(&svc->srv_lock);
 355         }
 356
 357         EXIT;
 358 }
 359
 360 /*
 361  * Server's bulk completion callback
 362  */
 363 void server_bulk_callback (lnet_event_t *ev)
 364 {
 365         struct ptlrpc_cb_id     *cbid = ev->md.user_ptr;
 366         struct ptlrpc_bulk_desc *desc = cbid->cbid_arg;
 367         ENTRY;
 368
 369         LASSERT (ev->type == LNET_EVENT_SEND ||
 370                  ev->type == LNET_EVENT_UNLINK ||
 371                  (desc->bd_type == BULK_PUT_SOURCE &&
 372                   ev->type == LNET_EVENT_ACK) ||
 373                  (desc->bd_type == BULK_GET_SINK &&
 374                   ev->type == LNET_EVENT_REPLY));
 375
 376         CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
 377                "event type %d, status %d, desc %p\n",
 378                ev->type, ev->status, desc);
 379
 380         spin_lock(&desc->bd_lock);
 381
 382         if ((ev->type == LNET_EVENT_ACK ||
 383              ev->type == LNET_EVENT_REPLY) &&
 384             ev->status == 0) {
 385                 /* We heard back from the peer, so even if we get this
 386                  * before the SENT event (oh yes we can), we know we
 387                  * read/wrote the peer buffer and how much... */
 388                 desc->bd_success = 1;
 389                 desc->bd_nob_transferred = ev->mlength;
 390                 desc->bd_sender = ev->sender;
 391         }
 392
 393         if (ev->unlinked) {
 394                 /* This is the last callback no matter what... */
 395                 desc->bd_network_rw = 0;
 396                 cfs_waitq_signal(&desc->bd_waitq);
 397         }
 398
 399         spin_unlock(&desc->bd_lock);
 400         EXIT;
 401 }
 402
 403 static void ptlrpc_master_callback(lnet_event_t *ev)
 404 {
 405         struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
 406         void (*callback)(lnet_event_t *ev) = cbid->cbid_fn;
 407
 408         /* Honestly, it's best to find out early. */
 409         LASSERT (cbid->cbid_arg != LP_POISON);
 410         LASSERT (callback == request_out_callback ||
 411                  callback == reply_in_callback ||
 412                  callback == client_bulk_callback ||
 413                  callback == request_in_callback ||
 414                  callback == reply_out_callback ||
 415                  callback == server_bulk_callback);
 416
 417         callback (ev);
 418 }
 419
 420 int ptlrpc_uuid_to_peer (struct obd_uuid *uuid,
 421                          lnet_process_id_t *peer, lnet_nid_t *self)
 422 {
 423         int               best_dist = 0;
 424         __u32             best_order = 0;
 425         int               count = 0;
 426         int               rc = -ENOENT;
 427         int               portals_compatibility;
 428         int               dist;
 429         __u32             order;
 430         lnet_nid_t        dst_nid;
 431         lnet_nid_t        src_nid;
 432
 433         portals_compatibility = LNetCtl(IOC_LIBCFS_PORTALS_COMPATIBILITY, NULL);
 434
 435         peer->pid = LUSTRE_SRV_LNET_PID;
 436
 437         /* Choose the matching UUID that's closest */
 438         while (lustre_uuid_to_peer(uuid->uuid, &dst_nid, count++) == 0) {
 439                 dist = LNetDist(dst_nid, &src_nid, &order);
 440                 if (dist < 0)
 441                         continue;
 442
 443                 if (dist == 0) {                /* local! use loopback LND */
 444                         peer->nid = *self = LNET_MKNID(LNET_MKNET(LOLND, 0), 0);
 445                         rc = 0;
 446                         break;
 447                 }
 448
 449                 if (rc < 0 ||
 450                     dist < best_dist ||
 451                     (dist == best_dist && order < best_order)) {
 452                         best_dist = dist;
 453                         best_order = order;
 454
 455                         if (portals_compatibility > 1) {
 456                                 /* Strong portals compatibility: Zero the nid's
 457                                  * NET, so if I'm reading new config logs, or
 458                                  * getting configured by (new) lconf I can
 459                                  * still talk to old servers. */
 460                                 dst_nid = LNET_MKNID(0, LNET_NIDADDR(dst_nid));
 461                                 src_nid = LNET_MKNID(0, LNET_NIDADDR(src_nid));
 462                         }
 463                         peer->nid = dst_nid;
 464                         *self = src_nid;
 465                         rc = 0;
 466                 }
 467         }
 468
 469         CDEBUG(D_NET,"%s->%s\n", uuid->uuid, libcfs_id2str(*peer));
 470         if (rc != 0)
 471                 CERROR("No NID found for %s\n", uuid->uuid);
 472         return rc;
 473 }
 474
 475 void ptlrpc_ni_fini(void)
 476 {
 477         cfs_waitq_t         waitq;
 478         struct l_wait_info  lwi;
 479         int                 rc;
 480         int                 retries;
 481
 482         /* Wait for the event queue to become idle since there may still be
 483          * messages in flight with pending events (i.e. the fire-and-forget
 484          * messages == client requests and "non-difficult" server
 485          * replies */
 486
 487         for (retries = 0;; retries++) {
 488                 rc = LNetEQFree(ptlrpc_eq_h);
 489                 switch (rc) {
 490                 default:
 491                         LBUG();
 492
 493                 case 0:
 494                         LNetNIFini();
 495                         return;
 496
 497                 case -EBUSY:
 498                         if (retries != 0)
 499                                 CWARN("Event queue still busy\n");
 500
 501                         /* Wait for a bit */
 502                         cfs_waitq_init(&waitq);
 503                         lwi = LWI_TIMEOUT(cfs_time_seconds(2), NULL, NULL);
 504                         l_wait_event(waitq, 0, &lwi);
 505                         break;
 506                 }
 507         }
 508         /* notreached */
 509 }
 510
 511 lnet_pid_t ptl_get_pid(void)
 512 {
 513         lnet_pid_t        pid;
 514
 515 #ifndef  __KERNEL__
 516         pid = getpid();
 517 #else
 518         pid = LUSTRE_SRV_LNET_PID;
 519 #endif
 520         return pid;
 521 }
 522
 523 int ptlrpc_ni_init(void)
 524 {
 525         int              rc;
 526         lnet_pid_t       pid;
 527
 528         pid = ptl_get_pid();
 529         CDEBUG(D_NET, "My pid is: %x\n", pid);
 530
 531         /* We're not passing any limits yet... */
 532         rc = LNetNIInit(pid);
 533         if (rc < 0) {
 534                 CDEBUG (D_NET, "Can't init network interface: %d\n", rc);
 535                 return (-ENOENT);
 536         }
 537
 538         /* CAVEAT EMPTOR: how we process portals events is _radically_
 539          * different depending on... */
 540 #ifdef __KERNEL__
 541         /* kernel portals calls our master callback when events are added to
 542          * the event queue.  In fact lustre never pulls events off this queue,
 543          * so it's only sized for some debug history. */
 544         rc = LNetEQAlloc(1024, ptlrpc_master_callback, &ptlrpc_eq_h);
 545 #else
 546         /* liblustre calls the master callback when it removes events from the
 547          * event queue.  The event queue has to be big enough not to drop
 548          * anything */
 549         rc = LNetEQAlloc(10240, LNET_EQ_HANDLER_NONE, &ptlrpc_eq_h);
 550 #endif
 551         if (rc == 0)
 552                 return 0;
 553
 554         CERROR ("Failed to allocate event queue: %d\n", rc);
 555         LNetNIFini();
 556
 557         return (-ENOMEM);
 558 }
 559
 560 #ifndef __KERNEL__
 561 CFS_LIST_HEAD(liblustre_wait_callbacks);
 562 CFS_LIST_HEAD(liblustre_idle_callbacks);
 563 void *liblustre_services_callback;
 564
 565 void *
 566 liblustre_register_waitidle_callback (struct list_head *callback_list,
 567                                       const char *name,
 568                                       int (*fn)(void *arg), void *arg)
 569 {
 570         struct liblustre_wait_callback *llwc;
 571
 572         OBD_ALLOC(llwc, sizeof(*llwc));
 573         LASSERT (llwc != NULL);
 574
 575         llwc->llwc_name = name;
 576         llwc->llwc_fn = fn;
 577         llwc->llwc_arg = arg;
 578         list_add_tail(&llwc->llwc_list, callback_list);
 579
 580         return (llwc);
 581 }
 582
 583 void
 584 liblustre_deregister_waitidle_callback (void *opaque)
 585 {
 586         struct liblustre_wait_callback *llwc = opaque;
 587
 588         list_del(&llwc->llwc_list);
 589         OBD_FREE(llwc, sizeof(*llwc));
 590 }
 591
 592 void *
 593 liblustre_register_wait_callback (const char *name,
 594                                   int (*fn)(void *arg), void *arg)
 595 {
 596         return liblustre_register_waitidle_callback(&liblustre_wait_callbacks,
 597                                                     name, fn, arg);
 598 }
 599
 600 void
 601 liblustre_deregister_wait_callback (void *opaque)
 602 {
 603         liblustre_deregister_waitidle_callback(opaque);
 604 }
 605
 606 void *
 607 liblustre_register_idle_callback (const char *name,
 608                                   int (*fn)(void *arg), void *arg)
 609 {
 610         return liblustre_register_waitidle_callback(&liblustre_idle_callbacks,
 611                                                     name, fn, arg);
 612 }
 613
 614 void
 615 liblustre_deregister_idle_callback (void *opaque)
 616 {
 617         liblustre_deregister_waitidle_callback(opaque);
 618 }
 619
 620 int
 621 liblustre_check_events (int timeout)
 622 {
 623         lnet_event_t ev;
 624         int         rc;
 625         int         i;
 626         ENTRY;
 627
 628         rc = LNetEQPoll(&ptlrpc_eq_h, 1, timeout * 1000, &ev, &i);
 629         if (rc == 0)
 630                 RETURN(0);
 631
 632         LASSERT (rc == -EOVERFLOW || rc == 1);
 633
 634         /* liblustre: no asynch callback so we can't affort to miss any
 635          * events... */
 636         if (rc == -EOVERFLOW) {
 637                 CERROR ("Dropped an event!!!\n");
 638                 abort();
 639         }
 640
 641         ptlrpc_master_callback (&ev);
 642         RETURN(1);
 643 }
 644
 645 int liblustre_waiting = 0;
 646
 647 int
 648 liblustre_wait_event (int timeout)
 649 {
 650         struct list_head               *tmp;
 651         struct liblustre_wait_callback *llwc;
 652         int                             found_something = 0;
 653
 654         /* single threaded recursion check... */
 655         liblustre_waiting = 1;
 656
 657         for (;;) {
 658                 /* Deal with all pending events */
 659                 while (liblustre_check_events(0))
 660                         found_something = 1;
 661
 662                 /* Give all registered callbacks a bite at the cherry */
 663                 list_for_each(tmp, &liblustre_wait_callbacks) {
 664                         llwc = list_entry(tmp, struct liblustre_wait_callback,
 665                                           llwc_list);
 666
 667                         if (llwc->llwc_fn(llwc->llwc_arg))
 668                                 found_something = 1;
 669                 }
 670
 671                 if (found_something || timeout == 0)
 672                         break;
 673
 674                 /* Nothing so far, but I'm allowed to block... */
 675                 found_something = liblustre_check_events(timeout);
 676                 if (!found_something)           /* still nothing */
 677                         break;                  /* I timed out */
 678         }
 679
 680         liblustre_waiting = 0;
 681
 682         return found_something;
 683 }
 684
 685 void
 686 liblustre_wait_idle(void)
 687 {
 688         static int recursed = 0;
 689
 690         struct list_head               *tmp;
 691         struct liblustre_wait_callback *llwc;
 692         int                             idle = 0;
 693
 694         LASSERT(!recursed);
 695         recursed = 1;
 696
 697         do {
 698                 liblustre_wait_event(0);
 699
 700                 idle = 1;
 701
 702                 list_for_each(tmp, &liblustre_idle_callbacks) {
 703                         llwc = list_entry(tmp, struct liblustre_wait_callback,
 704                                           llwc_list);
 705
 706                         if (!llwc->llwc_fn(llwc->llwc_arg)) {
 707                                 idle = 0;
 708                                 break;
 709                         }
 710                 }
 711
 712         } while (!idle);
 713
 714         recursed = 0;
 715 }
 716
 717 #endif /* __KERNEL__ */
 718
 719 int ptlrpc_init_portals(void)
 720 {
 721         int   rc = ptlrpc_ni_init();
 722
 723         if (rc != 0) {
 724                 CERROR("network initialisation failed\n");
 725                 return -EIO;
 726         }
 727 #ifndef __KERNEL__
 728         liblustre_services_callback =
 729                 liblustre_register_wait_callback("liblustre_check_services",
 730                                                  &liblustre_check_services, NULL);
 731 #endif
 732         rc = ptlrpcd_addref();
 733         if (rc == 0)
 734                 return 0;
 735
 736         CERROR("rpcd initialisation failed\n");
 737 #ifndef __KERNEL__
 738         liblustre_deregister_wait_callback(liblustre_services_callback);
 739 #endif
 740         ptlrpc_ni_fini();
 741         return rc;
 742 }
 743
 744 void ptlrpc_exit_portals(void)
 745 {
 746 #ifndef __KERNEL__
 747         liblustre_deregister_wait_callback(liblustre_services_callback);
 748 #endif
 749         ptlrpcd_decref();
 750         ptlrpc_ni_fini();
 751 }