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 # include <libcfs/libcfs.h>
  43 # ifdef __mips64__
  44 #  include <linux/kernel.h>
  45 # endif
  46 #endif
  47
  48 #include <obd_class.h>
  49 #include <lustre_net.h>
  50 #include <lustre_sec.h>
  51 #include "ptlrpc_internal.h"
  52
  53 lnet_handle_eq_t   ptlrpc_eq_h;
  54
  55 /*
  56  *  Client's outgoing request callback
  57  */
  58 void request_out_callback(lnet_event_t *ev)
  59 {
  60         struct ptlrpc_cb_id   *cbid = ev->md.user_ptr;
  61         struct ptlrpc_request *req = cbid->cbid_arg;
  62         ENTRY;
  63
  64         LASSERT (ev->type == LNET_EVENT_SEND ||
  65                  ev->type == LNET_EVENT_UNLINK);
  66         LASSERT (ev->unlinked);
  67
  68         DEBUG_REQ((ev->status == 0) ? D_NET : D_ERROR, req,
  69                   "type %d, status %d", ev->type, ev->status);
  70
  71         sptlrpc_request_out_callback(req);
  72
  73         if (ev->type == LNET_EVENT_UNLINK || ev->status != 0) {
  74
  75                 /* Failed send: make it seem like the reply timed out, just
  76                  * like failing sends in client.c does currently...  */
  77
  78                 cfs_spin_lock(&req->rq_lock);
  79                 req->rq_net_err = 1;
  80                 cfs_spin_unlock(&req->rq_lock);
  81
  82                 ptlrpc_client_wake_req(req);
  83         }
  84
  85         ptlrpc_req_finished(req);
  86
  87         EXIT;
  88 }
  89
  90 /*
  91  * Client's incoming reply callback
  92  */
  93 void reply_in_callback(lnet_event_t *ev)
  94 {
  95         struct ptlrpc_cb_id   *cbid = ev->md.user_ptr;
  96         struct ptlrpc_request *req = cbid->cbid_arg;
  97         ENTRY;
  98
  99         DEBUG_REQ((ev->status == 0) ? D_NET : D_ERROR, req,
 100                   "type %d, status %d", ev->type, ev->status);
 101
 102         LASSERT (ev->type == LNET_EVENT_PUT || ev->type == LNET_EVENT_UNLINK);
 103         LASSERT (ev->md.start == req->rq_repbuf);
 104         LASSERT (ev->mlength <= req->rq_repbuf_len);
 105         /* We've set LNET_MD_MANAGE_REMOTE for all outgoing requests
 106            for adaptive timeouts' early reply. */
 107         LASSERT((ev->md.options & LNET_MD_MANAGE_REMOTE) != 0);
 108
 109         cfs_spin_lock(&req->rq_lock);
 110
 111         req->rq_receiving_reply = 0;
 112         req->rq_early = 0;
 113         if (ev->unlinked)
 114                 req->rq_must_unlink = 0;
 115
 116         if (ev->status)
 117                 goto out_wake;
 118
 119         if (ev->type == LNET_EVENT_UNLINK) {
 120                 LASSERT(ev->unlinked);
 121                 DEBUG_REQ(D_RPCTRACE, req, "unlink");
 122                 goto out_wake;
 123         }
 124
 125         if (ev->mlength < ev->rlength ) {
 126                 CDEBUG(D_RPCTRACE, "truncate req %p rpc %d - %d+%d\n", req,
 127                        req->rq_replen, ev->rlength, ev->offset);
 128                 req->rq_reply_truncate = 1;
 129                 req->rq_replied = 1;
 130                 req->rq_status = -EOVERFLOW;
 131                 req->rq_nob_received = ev->rlength + ev->offset;
 132                 goto out_wake;
 133         }
 134
 135         if ((ev->offset == 0) &&
 136             ((lustre_msghdr_get_flags(req->rq_reqmsg) & MSGHDR_AT_SUPPORT))) {
 137                 /* Early reply */
 138                 DEBUG_REQ(D_ADAPTTO, req,
 139                           "Early reply received: mlen=%u offset=%d replen=%d "
 140                           "replied=%d unlinked=%d", ev->mlength, ev->offset,
 141                           req->rq_replen, req->rq_replied, ev->unlinked);
 142
 143                 req->rq_early_count++; /* number received, client side */
 144
 145                 if (req->rq_replied)   /* already got the real reply */
 146                         goto out_wake;
 147
 148                 req->rq_early = 1;
 149                 req->rq_reply_off = ev->offset;
 150                 req->rq_nob_received = ev->mlength;
 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_reply_off = ev->offset;
 158                 req->rq_nob_received = ev->mlength;
 159                 /* LNetMDUnlink can't be called under the LNET_LOCK,
 160                    so we must unlink in ptlrpc_unregister_reply */
 161                 DEBUG_REQ(D_INFO, req,
 162                           "reply in flags=%x mlen=%u offset=%d replen=%d",
 163                           lustre_msg_get_flags(req->rq_reqmsg),
 164                           ev->mlength, ev->offset, req->rq_replen);
 165         }
 166
 167         req->rq_import->imp_last_reply_time = cfs_time_current_sec();
 168
 169 out_wake:
 170         /* NB don't unlock till after wakeup; req can disappear under us
 171          * since we don't have our own ref */
 172         ptlrpc_client_wake_req(req);
 173         cfs_spin_unlock(&req->rq_lock);
 174         EXIT;
 175 }
 176
 177 /*
 178  * Client's bulk has been written/read
 179  */
 180 void client_bulk_callback (lnet_event_t *ev)
 181 {
 182         struct ptlrpc_cb_id     *cbid = ev->md.user_ptr;
 183         struct ptlrpc_bulk_desc *desc = cbid->cbid_arg;
 184         ENTRY;
 185
 186         LASSERT ((desc->bd_type == BULK_PUT_SINK &&
 187                   ev->type == LNET_EVENT_PUT) ||
 188                  (desc->bd_type == BULK_GET_SOURCE &&
 189                   ev->type == LNET_EVENT_GET) ||
 190                  ev->type == LNET_EVENT_UNLINK);
 191         LASSERT (ev->unlinked);
 192
 193         CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
 194                "event type %d, status %d, desc %p\n",
 195                ev->type, ev->status, desc);
 196
 197         cfs_spin_lock(&desc->bd_lock);
 198
 199         LASSERT(desc->bd_network_rw);
 200         desc->bd_network_rw = 0;
 201
 202         if (ev->type != LNET_EVENT_UNLINK && ev->status == 0) {
 203                 desc->bd_success = 1;
 204                 desc->bd_nob_transferred = ev->mlength;
 205                 desc->bd_sender = ev->sender;
 206         }
 207
 208         /* release the encrypted pages for write */
 209         if (desc->bd_req->rq_bulk_write)
 210                 sptlrpc_enc_pool_put_pages(desc);
 211
 212         /* NB don't unlock till after wakeup; desc can disappear under us
 213          * otherwise */
 214         ptlrpc_client_wake_req(desc->bd_req);
 215
 216         cfs_spin_unlock(&desc->bd_lock);
 217         EXIT;
 218 }
 219
 220 /*
 221  * Server's incoming request callback
 222  */
 223 void request_in_callback(lnet_event_t *ev)
 224 {
 225         struct ptlrpc_cb_id               *cbid = ev->md.user_ptr;
 226         struct ptlrpc_request_buffer_desc *rqbd = cbid->cbid_arg;
 227         struct ptlrpc_service             *service = rqbd->rqbd_service;
 228         struct ptlrpc_request             *req;
 229         ENTRY;
 230
 231         LASSERT (ev->type == LNET_EVENT_PUT ||
 232                  ev->type == LNET_EVENT_UNLINK);
 233         LASSERT ((char *)ev->md.start >= rqbd->rqbd_buffer);
 234         LASSERT ((char *)ev->md.start + ev->offset + ev->mlength <=
 235                  rqbd->rqbd_buffer + service->srv_buf_size);
 236
 237         CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
 238                "event type %d, status %d, service %s\n",
 239                ev->type, ev->status, service->srv_name);
 240
 241         if (ev->unlinked) {
 242                 /* If this is the last request message to fit in the
 243                  * request buffer we can use the request object embedded in
 244                  * rqbd.  Note that if we failed to allocate a request,
 245                  * we'd have to re-post the rqbd, which we can't do in this
 246                  * context. */
 247                 req = &rqbd->rqbd_req;
 248                 memset(req, 0, sizeof (*req));
 249         } else {
 250                 LASSERT (ev->type == LNET_EVENT_PUT);
 251                 if (ev->status != 0) {
 252                         /* We moaned above already... */
 253                         return;
 254                 }
 255                 OBD_ALLOC_GFP(req, sizeof(*req), CFS_ALLOC_ATOMIC_TRY);
 256                 if (req == NULL) {
 257                         CERROR("Can't allocate incoming request descriptor: "
 258                                "Dropping %s RPC from %s\n",
 259                                service->srv_name,
 260                                libcfs_id2str(ev->initiator));
 261                         return;
 262                 }
 263         }
 264
 265         /* NB we ABSOLUTELY RELY on req being zeroed, so pointers are NULL,
 266          * flags are reset and scalars are zero.  We only set the message
 267          * size to non-zero if this was a successful receive. */
 268         req->rq_xid = ev->match_bits;
 269         req->rq_reqbuf = ev->md.start + ev->offset;
 270         if (ev->type == LNET_EVENT_PUT && ev->status == 0)
 271                 req->rq_reqdata_len = ev->mlength;
 272         cfs_gettimeofday(&req->rq_arrival_time);
 273         req->rq_peer = ev->initiator;
 274         req->rq_self = ev->target.nid;
 275         req->rq_rqbd = rqbd;
 276         req->rq_phase = RQ_PHASE_NEW;
 277 #ifdef CRAY_XT3
 278         req->rq_uid = ev->uid;
 279 #endif
 280         cfs_spin_lock_init(&req->rq_lock);
 281         CFS_INIT_LIST_HEAD(&req->rq_timed_list);
 282         cfs_atomic_set(&req->rq_refcount, 1);
 283         if (ev->type == LNET_EVENT_PUT)
 284                 CDEBUG(D_RPCTRACE, "incoming req@%p x"LPU64" msgsize %u\n",
 285                        req, req->rq_xid, ev->mlength);
 286
 287         CDEBUG(D_RPCTRACE, "peer: %s\n", libcfs_id2str(req->rq_peer));
 288
 289         cfs_spin_lock(&service->srv_lock);
 290
 291         req->rq_history_seq = service->srv_request_seq++;
 292         cfs_list_add_tail(&req->rq_history_list, &service->srv_request_history);
 293
 294         if (ev->unlinked) {
 295                 service->srv_nrqbd_receiving--;
 296                 CDEBUG(D_INFO, "Buffer complete: %d buffers still posted\n",
 297                        service->srv_nrqbd_receiving);
 298
 299                 /* Normally, don't complain about 0 buffers posted; LNET won't
 300                  * drop incoming reqs since we set the portal lazy */
 301                 if (test_req_buffer_pressure &&
 302                     ev->type != LNET_EVENT_UNLINK &&
 303                     service->srv_nrqbd_receiving == 0)
 304                         CWARN("All %s request buffers busy\n",
 305                               service->srv_name);
 306
 307                 /* req takes over the network's ref on rqbd */
 308         } else {
 309                 /* req takes a ref on rqbd */
 310                 rqbd->rqbd_refcount++;
 311         }
 312
 313         cfs_list_add_tail(&req->rq_list, &service->srv_req_in_queue);
 314         service->srv_n_queued_reqs++;
 315
 316         /* NB everything can disappear under us once the request
 317          * has been queued and we unlock, so do the wake now... */
 318         cfs_waitq_signal(&service->srv_waitq);
 319
 320         cfs_spin_unlock(&service->srv_lock);
 321         EXIT;
 322 }
 323
 324 /*
 325  *  Server's outgoing reply callback
 326  */
 327 void reply_out_callback(lnet_event_t *ev)
 328 {
 329         struct ptlrpc_cb_id       *cbid = ev->md.user_ptr;
 330         struct ptlrpc_reply_state *rs = cbid->cbid_arg;
 331         struct ptlrpc_service     *svc = rs->rs_service;
 332         ENTRY;
 333
 334         LASSERT (ev->type == LNET_EVENT_SEND ||
 335                  ev->type == LNET_EVENT_ACK ||
 336                  ev->type == LNET_EVENT_UNLINK);
 337
 338         if (!rs->rs_difficult) {
 339                 /* 'Easy' replies have no further processing so I drop the
 340                  * net's ref on 'rs' */
 341                 LASSERT (ev->unlinked);
 342                 ptlrpc_rs_decref(rs);
 343                 cfs_atomic_dec (&svc->srv_outstanding_replies);
 344                 EXIT;
 345                 return;
 346         }
 347
 348         LASSERT (rs->rs_on_net);
 349
 350         if (ev->unlinked) {
 351                 /* Last network callback. The net's ref on 'rs' stays put
 352                  * until ptlrpc_handle_rs() is done with it */
 353                 cfs_spin_lock(&svc->srv_lock);
 354                 cfs_spin_lock(&rs->rs_lock);
 355                 rs->rs_on_net = 0;
 356                 if (!rs->rs_no_ack ||
 357                     rs->rs_transno <= rs->rs_export->exp_obd->obd_last_committed)
 358                         ptlrpc_schedule_difficult_reply (rs);
 359                 cfs_spin_unlock(&rs->rs_lock);
 360                 cfs_spin_unlock(&svc->srv_lock);
 361         }
 362
 363         EXIT;
 364 }
 365
 366 /*
 367  * Server's bulk completion callback
 368  */
 369 void server_bulk_callback (lnet_event_t *ev)
 370 {
 371         struct ptlrpc_cb_id     *cbid = ev->md.user_ptr;
 372         struct ptlrpc_bulk_desc *desc = cbid->cbid_arg;
 373         ENTRY;
 374
 375         LASSERT (ev->type == LNET_EVENT_SEND ||
 376                  ev->type == LNET_EVENT_UNLINK ||
 377                  (desc->bd_type == BULK_PUT_SOURCE &&
 378                   ev->type == LNET_EVENT_ACK) ||
 379                  (desc->bd_type == BULK_GET_SINK &&
 380                   ev->type == LNET_EVENT_REPLY));
 381
 382         CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
 383                "event type %d, status %d, desc %p\n",
 384                ev->type, ev->status, desc);
 385
 386         cfs_spin_lock(&desc->bd_lock);
 387
 388         if ((ev->type == LNET_EVENT_ACK ||
 389              ev->type == LNET_EVENT_REPLY) &&
 390             ev->status == 0) {
 391                 /* We heard back from the peer, so even if we get this
 392                  * before the SENT event (oh yes we can), we know we
 393                  * read/wrote the peer buffer and how much... */
 394                 desc->bd_success = 1;
 395                 desc->bd_nob_transferred = ev->mlength;
 396                 desc->bd_sender = ev->sender;
 397         }
 398
 399         if (ev->unlinked) {
 400                 /* This is the last callback no matter what... */
 401                 desc->bd_network_rw = 0;
 402                 cfs_waitq_signal(&desc->bd_waitq);
 403         }
 404
 405         cfs_spin_unlock(&desc->bd_lock);
 406         EXIT;
 407 }
 408
 409 static void ptlrpc_master_callback(lnet_event_t *ev)
 410 {
 411         struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
 412         void (*callback)(lnet_event_t *ev) = cbid->cbid_fn;
 413
 414         /* Honestly, it's best to find out early. */
 415         LASSERT (cbid->cbid_arg != LP_POISON);
 416         LASSERT (callback == request_out_callback ||
 417                  callback == reply_in_callback ||
 418                  callback == client_bulk_callback ||
 419                  callback == request_in_callback ||
 420                  callback == reply_out_callback ||
 421                  callback == server_bulk_callback);
 422
 423         callback (ev);
 424 }
 425
 426 int ptlrpc_uuid_to_peer (struct obd_uuid *uuid,
 427                          lnet_process_id_t *peer, lnet_nid_t *self)
 428 {
 429         int               best_dist = 0;
 430         __u32             best_order = 0;
 431         int               count = 0;
 432         int               rc = -ENOENT;
 433         int               portals_compatibility;
 434         int               dist;
 435         __u32             order;
 436         lnet_nid_t        dst_nid;
 437         lnet_nid_t        src_nid;
 438
 439         portals_compatibility = LNetCtl(IOC_LIBCFS_PORTALS_COMPATIBILITY, NULL);
 440
 441         peer->pid = LUSTRE_SRV_LNET_PID;
 442
 443         /* Choose the matching UUID that's closest */
 444         while (lustre_uuid_to_peer(uuid->uuid, &dst_nid, count++) == 0) {
 445                 dist = LNetDist(dst_nid, &src_nid, &order);
 446                 if (dist < 0)
 447                         continue;
 448
 449                 if (dist == 0) {                /* local! use loopback LND */
 450                         peer->nid = *self = LNET_MKNID(LNET_MKNET(LOLND, 0), 0);
 451                         rc = 0;
 452                         break;
 453                 }
 454
 455                 if (rc < 0 ||
 456                     dist < best_dist ||
 457                     (dist == best_dist && order < best_order)) {
 458                         best_dist = dist;
 459                         best_order = order;
 460
 461                         if (portals_compatibility > 1) {
 462                                 /* Strong portals compatibility: Zero the nid's
 463                                  * NET, so if I'm reading new config logs, or
 464                                  * getting configured by (new) lconf I can
 465                                  * still talk to old servers. */
 466                                 dst_nid = LNET_MKNID(0, LNET_NIDADDR(dst_nid));
 467                                 src_nid = LNET_MKNID(0, LNET_NIDADDR(src_nid));
 468                         }
 469                         peer->nid = dst_nid;
 470                         *self = src_nid;
 471                         rc = 0;
 472                 }
 473         }
 474
 475         CDEBUG(D_NET,"%s->%s\n", uuid->uuid, libcfs_id2str(*peer));
 476         if (rc != 0)
 477                 CERROR("No NID found for %s\n", uuid->uuid);
 478         return rc;
 479 }
 480
 481 void ptlrpc_ni_fini(void)
 482 {
 483         cfs_waitq_t         waitq;
 484         struct l_wait_info  lwi;
 485         int                 rc;
 486         int                 retries;
 487
 488         /* Wait for the event queue to become idle since there may still be
 489          * messages in flight with pending events (i.e. the fire-and-forget
 490          * messages == client requests and "non-difficult" server
 491          * replies */
 492
 493         for (retries = 0;; retries++) {
 494                 rc = LNetEQFree(ptlrpc_eq_h);
 495                 switch (rc) {
 496                 default:
 497                         LBUG();
 498
 499                 case 0:
 500                         LNetNIFini();
 501                         return;
 502
 503                 case -EBUSY:
 504                         if (retries != 0)
 505                                 CWARN("Event queue still busy\n");
 506
 507                         /* Wait for a bit */
 508                         cfs_waitq_init(&waitq);
 509                         lwi = LWI_TIMEOUT(cfs_time_seconds(2), NULL, NULL);
 510                         l_wait_event(waitq, 0, &lwi);
 511                         break;
 512                 }
 513         }
 514         /* notreached */
 515 }
 516
 517 lnet_pid_t ptl_get_pid(void)
 518 {
 519         lnet_pid_t        pid;
 520
 521 #ifndef  __KERNEL__
 522         pid = getpid();
 523 #else
 524         pid = LUSTRE_SRV_LNET_PID;
 525 #endif
 526         return pid;
 527 }
 528
 529 int ptlrpc_ni_init(void)
 530 {
 531         int              rc;
 532         lnet_pid_t       pid;
 533
 534         pid = ptl_get_pid();
 535         CDEBUG(D_NET, "My pid is: %x\n", pid);
 536
 537         /* We're not passing any limits yet... */
 538         rc = LNetNIInit(pid);
 539         if (rc < 0) {
 540                 CDEBUG (D_NET, "Can't init network interface: %d\n", rc);
 541                 return (-ENOENT);
 542         }
 543
 544         /* CAVEAT EMPTOR: how we process portals events is _radically_
 545          * different depending on... */
 546 #ifdef __KERNEL__
 547         /* kernel portals calls our master callback when events are added to
 548          * the event queue.  In fact lustre never pulls events off this queue,
 549          * so it's only sized for some debug history. */
 550         rc = LNetEQAlloc(1024, ptlrpc_master_callback, &ptlrpc_eq_h);
 551 #else
 552         /* liblustre calls the master callback when it removes events from the
 553          * event queue.  The event queue has to be big enough not to drop
 554          * anything */
 555         rc = LNetEQAlloc(10240, LNET_EQ_HANDLER_NONE, &ptlrpc_eq_h);
 556 #endif
 557         if (rc == 0)
 558                 return 0;
 559
 560         CERROR ("Failed to allocate event queue: %d\n", rc);
 561         LNetNIFini();
 562
 563         return (-ENOMEM);
 564 }
 565
 566 #ifndef __KERNEL__
 567 CFS_LIST_HEAD(liblustre_wait_callbacks);
 568 CFS_LIST_HEAD(liblustre_idle_callbacks);
 569 void *liblustre_services_callback;
 570
 571 void *
 572 liblustre_register_waitidle_callback (cfs_list_t *callback_list,
 573                                       const char *name,
 574                                       int (*fn)(void *arg), void *arg)
 575 {
 576         struct liblustre_wait_callback *llwc;
 577
 578         OBD_ALLOC(llwc, sizeof(*llwc));
 579         LASSERT (llwc != NULL);
 580
 581         llwc->llwc_name = name;
 582         llwc->llwc_fn = fn;
 583         llwc->llwc_arg = arg;
 584         cfs_list_add_tail(&llwc->llwc_list, callback_list);
 585
 586         return (llwc);
 587 }
 588
 589 void
 590 liblustre_deregister_waitidle_callback (void *opaque)
 591 {
 592         struct liblustre_wait_callback *llwc = opaque;
 593
 594         cfs_list_del(&llwc->llwc_list);
 595         OBD_FREE(llwc, sizeof(*llwc));
 596 }
 597
 598 void *
 599 liblustre_register_wait_callback (const char *name,
 600                                   int (*fn)(void *arg), void *arg)
 601 {
 602         return liblustre_register_waitidle_callback(&liblustre_wait_callbacks,
 603                                                     name, fn, arg);
 604 }
 605
 606 void
 607 liblustre_deregister_wait_callback (void *opaque)
 608 {
 609         liblustre_deregister_waitidle_callback(opaque);
 610 }
 611
 612 void *
 613 liblustre_register_idle_callback (const char *name,
 614                                   int (*fn)(void *arg), void *arg)
 615 {
 616         return liblustre_register_waitidle_callback(&liblustre_idle_callbacks,
 617                                                     name, fn, arg);
 618 }
 619
 620 void
 621 liblustre_deregister_idle_callback (void *opaque)
 622 {
 623         liblustre_deregister_waitidle_callback(opaque);
 624 }
 625
 626 int
 627 liblustre_check_events (int timeout)
 628 {
 629         lnet_event_t ev;
 630         int         rc;
 631         int         i;
 632         ENTRY;
 633
 634         rc = LNetEQPoll(&ptlrpc_eq_h, 1, timeout * 1000, &ev, &i);
 635         if (rc == 0)
 636                 RETURN(0);
 637
 638         LASSERT (rc == -EOVERFLOW || rc == 1);
 639
 640         /* liblustre: no asynch callback so we can't afford to miss any
 641          * events... */
 642         if (rc == -EOVERFLOW) {
 643                 CERROR ("Dropped an event!!!\n");
 644                 abort();
 645         }
 646
 647         ptlrpc_master_callback (&ev);
 648         RETURN(1);
 649 }
 650
 651 int liblustre_waiting = 0;
 652
 653 int
 654 liblustre_wait_event (int timeout)
 655 {
 656         cfs_list_t                     *tmp;
 657         struct liblustre_wait_callback *llwc;
 658         int                             found_something = 0;
 659
 660         /* single threaded recursion check... */
 661         liblustre_waiting = 1;
 662
 663         for (;;) {
 664                 /* Deal with all pending events */
 665                 while (liblustre_check_events(0))
 666                         found_something = 1;
 667
 668                 /* Give all registered callbacks a bite at the cherry */
 669                 cfs_list_for_each(tmp, &liblustre_wait_callbacks) {
 670                         llwc = cfs_list_entry(tmp,
 671                                               struct liblustre_wait_callback,
 672                                               llwc_list);
 673
 674                         if (llwc->llwc_fn(llwc->llwc_arg))
 675                                 found_something = 1;
 676                 }
 677
 678                 if (found_something || timeout == 0)
 679                         break;
 680
 681                 /* Nothing so far, but I'm allowed to block... */
 682                 found_something = liblustre_check_events(timeout);
 683                 if (!found_something)           /* still nothing */
 684                         break;                  /* I timed out */
 685         }
 686
 687         liblustre_waiting = 0;
 688
 689         return found_something;
 690 }
 691
 692 void
 693 liblustre_wait_idle(void)
 694 {
 695         static int recursed = 0;
 696
 697         cfs_list_t                     *tmp;
 698         struct liblustre_wait_callback *llwc;
 699         int                             idle = 0;
 700
 701         LASSERT(!recursed);
 702         recursed = 1;
 703
 704         do {
 705                 liblustre_wait_event(0);
 706
 707                 idle = 1;
 708
 709                 cfs_list_for_each(tmp, &liblustre_idle_callbacks) {
 710                         llwc = cfs_list_entry(tmp,
 711                                               struct liblustre_wait_callback,
 712                                               llwc_list);
 713
 714                         if (!llwc->llwc_fn(llwc->llwc_arg)) {
 715                                 idle = 0;
 716                                 break;
 717                         }
 718                 }
 719
 720         } while (!idle);
 721
 722         recursed = 0;
 723 }
 724
 725 #endif /* __KERNEL__ */
 726
 727 int ptlrpc_init_portals(void)
 728 {
 729         int   rc = ptlrpc_ni_init();
 730
 731         if (rc != 0) {
 732                 CERROR("network initialisation failed\n");
 733                 return -EIO;
 734         }
 735 #ifndef __KERNEL__
 736         liblustre_services_callback =
 737                 liblustre_register_wait_callback("liblustre_check_services",
 738                                                  &liblustre_check_services,
 739                                                  NULL);
 740         cfs_init_completion_module(liblustre_wait_event);
 741 #endif
 742         rc = ptlrpcd_addref();
 743         if (rc == 0)
 744                 return 0;
 745
 746         CERROR("rpcd initialisation failed\n");
 747 #ifndef __KERNEL__
 748         liblustre_deregister_wait_callback(liblustre_services_callback);
 749 #endif
 750         ptlrpc_ni_fini();
 751         return rc;
 752 }
 753
 754 void ptlrpc_exit_portals(void)
 755 {
 756 #ifndef __KERNEL__
 757         liblustre_deregister_wait_callback(liblustre_services_callback);
 758 #endif
 759         ptlrpcd_decref();
 760         ptlrpc_ni_fini();
 761 }