lustre/ldlm/ldlm_flock.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) 2003 Hewlett-Packard Development Company LP.
  28  * Developed under the sponsorship of the US Government under
  29  * Subcontract No. B514193
  30  *
  31  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
  32  * Use is subject to license terms.
  33  *
  34  * Copyright (c) 2010, 2011, Whamcloud, Inc.
  35  */
  36 /*
  37  * This file is part of Lustre, http://www.lustre.org/
  38  * Lustre is a trademark of Sun Microsystems, Inc.
  39  */
  40
  41 #define DEBUG_SUBSYSTEM S_LDLM
  42
  43 #ifdef __KERNEL__
  44 #include <lustre_dlm.h>
  45 #include <obd_support.h>
  46 #include <obd_class.h>
  47 #include <lustre_lib.h>
  48 #include <libcfs/list.h>
  49 #else
  50 #include <liblustre.h>
  51 #include <obd_class.h>
  52 #endif
  53
  54 #include "ldlm_internal.h"
  55
  56 int ldlm_flock_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
  57                             void *data, int flag);
  58
  59 /**
  60  * list_for_remaining_safe - iterate over the remaining entries in a list
  61  *              and safeguard against removal of a list entry.
  62  * \param pos   the &struct list_head to use as a loop counter. pos MUST
  63  *              have been initialized prior to using it in this macro.
  64  * \param n     another &struct list_head to use as temporary storage
  65  * \param head  the head for your list.
  66  */
  67 #define list_for_remaining_safe(pos, n, head) \
  68         for (n = pos->next; pos != (head); pos = n, n = pos->next)
  69
  70 static inline int
  71 ldlm_same_flock_owner(struct ldlm_lock *lock, struct ldlm_lock *new)
  72 {
  73         return((new->l_policy_data.l_flock.owner ==
  74                 lock->l_policy_data.l_flock.owner) &&
  75                (new->l_export == lock->l_export));
  76 }
  77
  78 static inline int
  79 ldlm_flocks_overlap(struct ldlm_lock *lock, struct ldlm_lock *new)
  80 {
  81         return((new->l_policy_data.l_flock.start <=
  82                 lock->l_policy_data.l_flock.end) &&
  83                (new->l_policy_data.l_flock.end >=
  84                 lock->l_policy_data.l_flock.start));
  85 }
  86
  87 static inline int ldlm_flock_blocking_link(struct ldlm_lock *req,
  88                                            struct ldlm_lock *lock)
  89 {
  90         int rc = 0;
  91
  92         /* For server only */
  93         if (req->l_export == NULL)
  94                 return 0;
  95
  96         if (unlikely(req->l_export->exp_flock_hash == NULL)) {
  97                 rc = ldlm_init_flock_export(req->l_export);
  98                 if (rc)
  99                         goto error;
 100         }
 101
 102         LASSERT(cfs_hlist_unhashed(&req->l_exp_flock_hash));
 103
 104         req->l_policy_data.l_flock.blocking_owner =
 105                 lock->l_policy_data.l_flock.owner;
 106         req->l_policy_data.l_flock.blocking_export =
 107                 lock->l_export;
 108         req->l_policy_data.l_flock.blocking_refs = 0;
 109
 110         cfs_hash_add(req->l_export->exp_flock_hash,
 111                      &req->l_policy_data.l_flock.owner,
 112                      &req->l_exp_flock_hash);
 113 error:
 114         return rc;
 115 }
 116
 117 static inline void ldlm_flock_blocking_unlink(struct ldlm_lock *req)
 118 {
 119         /* For server only */
 120         if (req->l_export == NULL)
 121                 return;
 122
 123         check_res_locked(req->l_resource);
 124         if (req->l_export->exp_flock_hash != NULL &&
 125             !cfs_hlist_unhashed(&req->l_exp_flock_hash))
 126                 cfs_hash_del(req->l_export->exp_flock_hash,
 127                              &req->l_policy_data.l_flock.owner,
 128                              &req->l_exp_flock_hash);
 129 }
 130
 131 static inline void
 132 ldlm_flock_destroy(struct ldlm_lock *lock, ldlm_mode_t mode, __u64 flags)
 133 {
 134         ENTRY;
 135
 136         LDLM_DEBUG(lock, "ldlm_flock_destroy(mode: %d, flags: 0x%llx)",
 137                    mode, flags);
 138
 139         /* Safe to not lock here, since it should be empty anyway */
 140         LASSERT(cfs_hlist_unhashed(&lock->l_exp_flock_hash));
 141
 142         cfs_list_del_init(&lock->l_res_link);
 143         if (flags == LDLM_FL_WAIT_NOREPROC &&
 144             !(lock->l_flags & LDLM_FL_FAILED)) {
 145                 /* client side - set a flag to prevent sending a CANCEL */
 146                 lock->l_flags |= LDLM_FL_LOCAL_ONLY | LDLM_FL_CBPENDING;
 147
 148                 /* when reaching here, it is under lock_res_and_lock(). Thus,
 149                    need call the nolock version of ldlm_lock_decref_internal*/
 150                 ldlm_lock_decref_internal_nolock(lock, mode);
 151         }
 152
 153         ldlm_lock_destroy_nolock(lock);
 154         EXIT;
 155 }
 156
 157 static int
 158 ldlm_flock_deadlock(struct ldlm_lock *req, struct ldlm_lock *bl_lock)
 159 {
 160         struct obd_export *req_exp = req->l_export;
 161         struct obd_export *bl_exp = bl_lock->l_export;
 162         __u64 req_owner = req->l_policy_data.l_flock.owner;
 163         __u64 bl_owner = bl_lock->l_policy_data.l_flock.owner;
 164
 165         /* For server only */
 166         if (req_exp == NULL)
 167                 return 0;
 168
 169         class_export_get(bl_exp);
 170         while (1) {
 171                 struct obd_export *bl_exp_new;
 172                 struct ldlm_lock *lock = NULL;
 173                 struct ldlm_flock *flock;
 174
 175                 if (bl_exp->exp_flock_hash != NULL)
 176                         lock = cfs_hash_lookup(bl_exp->exp_flock_hash,
 177                                                &bl_owner);
 178                 if (lock == NULL)
 179                         break;
 180
 181                 flock = &lock->l_policy_data.l_flock;
 182                 LASSERT(flock->owner == bl_owner);
 183                 bl_owner = flock->blocking_owner;
 184                 bl_exp_new = class_export_get(flock->blocking_export);
 185                 class_export_put(bl_exp);
 186
 187                 cfs_hash_put(bl_exp->exp_flock_hash, &lock->l_exp_flock_hash);
 188                 bl_exp = bl_exp_new;
 189
 190                 if (bl_owner == req_owner && bl_exp == req_exp) {
 191                         class_export_put(bl_exp);
 192                         return 1;
 193                 }
 194         }
 195         class_export_put(bl_exp);
 196
 197         return 0;
 198 }
 199
 200 int
 201 ldlm_process_flock_lock(struct ldlm_lock *req, __u64 *flags, int first_enq,
 202                         ldlm_error_t *err, cfs_list_t *work_list)
 203 {
 204         struct ldlm_resource *res = req->l_resource;
 205         struct ldlm_namespace *ns = ldlm_res_to_ns(res);
 206         cfs_list_t *tmp;
 207         cfs_list_t *ownlocks = NULL;
 208         struct ldlm_lock *lock = NULL;
 209         struct ldlm_lock *new = req;
 210         struct ldlm_lock *new2 = NULL;
 211         ldlm_mode_t mode = req->l_req_mode;
 212         int local = ns_is_client(ns);
 213         int added = (mode == LCK_NL);
 214         int overlaps = 0;
 215         int splitted = 0;
 216         const struct ldlm_callback_suite null_cbs = { NULL };
 217         int rc;
 218         ENTRY;
 219
 220         CDEBUG(D_DLMTRACE, "flags %#llx owner "LPU64" pid %u mode %u start "
 221                LPU64" end "LPU64"\n", *flags,
 222                new->l_policy_data.l_flock.owner,
 223                new->l_policy_data.l_flock.pid, mode,
 224                req->l_policy_data.l_flock.start,
 225                req->l_policy_data.l_flock.end);
 226
 227         *err = ELDLM_OK;
 228
 229         if (local) {
 230                 /* No blocking ASTs are sent to the clients for
 231                  * Posix file & record locks */
 232                 req->l_blocking_ast = NULL;
 233         } else {
 234                 /* Called on the server for lock cancels. */
 235                 req->l_blocking_ast = ldlm_flock_blocking_ast;
 236         }
 237
 238 reprocess:
 239         if ((*flags == LDLM_FL_WAIT_NOREPROC) || (mode == LCK_NL)) {
 240                 /* This loop determines where this processes locks start
 241                  * in the resource lr_granted list. */
 242                 cfs_list_for_each(tmp, &res->lr_granted) {
 243                         lock = cfs_list_entry(tmp, struct ldlm_lock,
 244                                               l_res_link);
 245                         if (ldlm_same_flock_owner(lock, req)) {
 246                                 ownlocks = tmp;
 247                                 break;
 248                         }
 249                 }
 250         } else {
 251                 lockmode_verify(mode);
 252
 253                 /* This loop determines if there are existing locks
 254                  * that conflict with the new lock request. */
 255                 cfs_list_for_each(tmp, &res->lr_granted) {
 256                         lock = cfs_list_entry(tmp, struct ldlm_lock,
 257                                               l_res_link);
 258
 259                         if (ldlm_same_flock_owner(lock, req)) {
 260                                 if (!ownlocks)
 261                                         ownlocks = tmp;
 262                                 continue;
 263                         }
 264
 265                         /* locks are compatible, overlap doesn't matter */
 266                         if (lockmode_compat(lock->l_granted_mode, mode))
 267                                 continue;
 268
 269                         if (!ldlm_flocks_overlap(lock, req))
 270                                 continue;
 271
 272                         if (!first_enq)
 273                                 RETURN(LDLM_ITER_CONTINUE);
 274
 275                         if (*flags & LDLM_FL_BLOCK_NOWAIT) {
 276                                 ldlm_flock_destroy(req, mode, *flags);
 277                                 *err = -EAGAIN;
 278                                 RETURN(LDLM_ITER_STOP);
 279                         }
 280
 281                         if (*flags & LDLM_FL_TEST_LOCK) {
 282                                 ldlm_flock_destroy(req, mode, *flags);
 283                                 req->l_req_mode = lock->l_granted_mode;
 284                                 req->l_policy_data.l_flock.pid =
 285                                         lock->l_policy_data.l_flock.pid;
 286                                 req->l_policy_data.l_flock.start =
 287                                         lock->l_policy_data.l_flock.start;
 288                                 req->l_policy_data.l_flock.end =
 289                                         lock->l_policy_data.l_flock.end;
 290                                 *flags |= LDLM_FL_LOCK_CHANGED;
 291                                 RETURN(LDLM_ITER_STOP);
 292                         }
 293
 294                         if (ldlm_flock_deadlock(req, lock)) {
 295                                 ldlm_flock_destroy(req, mode, *flags);
 296                                 *err = -EDEADLK;
 297                                 RETURN(LDLM_ITER_STOP);
 298                         }
 299
 300                         rc = ldlm_flock_blocking_link(req, lock);
 301                         if (rc) {
 302                                 ldlm_flock_destroy(req, mode, *flags);
 303                                 *err = rc;
 304                                 RETURN(LDLM_ITER_STOP);
 305                         }
 306                         ldlm_resource_add_lock(res, &res->lr_waiting, req);
 307                         *flags |= LDLM_FL_BLOCK_GRANTED;
 308                         RETURN(LDLM_ITER_STOP);
 309                 }
 310         }
 311
 312         if (*flags & LDLM_FL_TEST_LOCK) {
 313                 ldlm_flock_destroy(req, mode, *flags);
 314                 req->l_req_mode = LCK_NL;
 315                 *flags |= LDLM_FL_LOCK_CHANGED;
 316                 RETURN(LDLM_ITER_STOP);
 317         }
 318
 319         /* In case we had slept on this lock request take it off of the
 320          * deadlock detection hash list. */
 321         ldlm_flock_blocking_unlink(req);
 322
 323         /* Scan the locks owned by this process that overlap this request.
 324          * We may have to merge or split existing locks. */
 325
 326         if (!ownlocks)
 327                 ownlocks = &res->lr_granted;
 328
 329         list_for_remaining_safe(ownlocks, tmp, &res->lr_granted) {
 330                 lock = cfs_list_entry(ownlocks, struct ldlm_lock, l_res_link);
 331
 332                 if (!ldlm_same_flock_owner(lock, new))
 333                         break;
 334
 335                 if (lock->l_granted_mode == mode) {
 336                         /* If the modes are the same then we need to process
 337                          * locks that overlap OR adjoin the new lock. The extra
 338                          * logic condition is necessary to deal with arithmetic
 339                          * overflow and underflow. */
 340                         if ((new->l_policy_data.l_flock.start >
 341                              (lock->l_policy_data.l_flock.end + 1))
 342                             && (lock->l_policy_data.l_flock.end !=
 343                                 OBD_OBJECT_EOF))
 344                                 continue;
 345
 346                         if ((new->l_policy_data.l_flock.end <
 347                              (lock->l_policy_data.l_flock.start - 1))
 348                             && (lock->l_policy_data.l_flock.start != 0))
 349                                 break;
 350
 351                         if (new->l_policy_data.l_flock.start <
 352                             lock->l_policy_data.l_flock.start) {
 353                                 lock->l_policy_data.l_flock.start =
 354                                         new->l_policy_data.l_flock.start;
 355                         } else {
 356                                 new->l_policy_data.l_flock.start =
 357                                         lock->l_policy_data.l_flock.start;
 358                         }
 359
 360                         if (new->l_policy_data.l_flock.end >
 361                             lock->l_policy_data.l_flock.end) {
 362                                 lock->l_policy_data.l_flock.end =
 363                                         new->l_policy_data.l_flock.end;
 364                         } else {
 365                                 new->l_policy_data.l_flock.end =
 366                                         lock->l_policy_data.l_flock.end;
 367                         }
 368
 369                         if (added) {
 370                                 ldlm_flock_destroy(lock, mode, *flags);
 371                         } else {
 372                                 new = lock;
 373                                 added = 1;
 374                         }
 375                         continue;
 376                 }
 377
 378                 if (new->l_policy_data.l_flock.start >
 379                     lock->l_policy_data.l_flock.end)
 380                         continue;
 381
 382                 if (new->l_policy_data.l_flock.end <
 383                     lock->l_policy_data.l_flock.start)
 384                         break;
 385
 386                 ++overlaps;
 387
 388                 if (new->l_policy_data.l_flock.start <=
 389                     lock->l_policy_data.l_flock.start) {
 390                         if (new->l_policy_data.l_flock.end <
 391                             lock->l_policy_data.l_flock.end) {
 392                                 lock->l_policy_data.l_flock.start =
 393                                         new->l_policy_data.l_flock.end + 1;
 394                                 break;
 395                         }
 396                         ldlm_flock_destroy(lock, lock->l_req_mode, *flags);
 397                         continue;
 398                 }
 399                 if (new->l_policy_data.l_flock.end >=
 400                     lock->l_policy_data.l_flock.end) {
 401                         lock->l_policy_data.l_flock.end =
 402                                 new->l_policy_data.l_flock.start - 1;
 403                         continue;
 404                 }
 405
 406                 /* split the existing lock into two locks */
 407
 408                 /* if this is an F_UNLCK operation then we could avoid
 409                  * allocating a new lock and use the req lock passed in
 410                  * with the request but this would complicate the reply
 411                  * processing since updates to req get reflected in the
 412                  * reply. The client side replays the lock request so
 413                  * it must see the original lock data in the reply. */
 414
 415                 /* XXX - if ldlm_lock_new() can sleep we should
 416                  * release the lr_lock, allocate the new lock,
 417                  * and restart processing this lock. */
 418                 if (!new2) {
 419                         unlock_res_and_lock(req);
 420                          new2 = ldlm_lock_create(ns, &res->lr_name, LDLM_FLOCK,
 421                                         lock->l_granted_mode, &null_cbs,
 422                                         NULL, 0);
 423                         lock_res_and_lock(req);
 424                         if (!new2) {
 425                                 ldlm_flock_destroy(req, lock->l_granted_mode,
 426                                                    *flags);
 427                                 *err = -ENOLCK;
 428                                 RETURN(LDLM_ITER_STOP);
 429                         }
 430                         goto reprocess;
 431                 }
 432
 433                 splitted = 1;
 434
 435                 new2->l_granted_mode = lock->l_granted_mode;
 436                 new2->l_policy_data.l_flock.pid =
 437                         new->l_policy_data.l_flock.pid;
 438                 new2->l_policy_data.l_flock.owner =
 439                         new->l_policy_data.l_flock.owner;
 440                 new2->l_policy_data.l_flock.start =
 441                         lock->l_policy_data.l_flock.start;
 442                 new2->l_policy_data.l_flock.end =
 443                         new->l_policy_data.l_flock.start - 1;
 444                 lock->l_policy_data.l_flock.start =
 445                         new->l_policy_data.l_flock.end + 1;
 446                 new2->l_conn_export = lock->l_conn_export;
 447                 if (lock->l_export != NULL) {
 448                         new2->l_export = class_export_lock_get(lock->l_export, new2);
 449                         if (new2->l_export->exp_lock_hash &&
 450                             cfs_hlist_unhashed(&new2->l_exp_hash))
 451                                 cfs_hash_add(new2->l_export->exp_lock_hash,
 452                                              &new2->l_remote_handle,
 453                                              &new2->l_exp_hash);
 454                 }
 455                 if (*flags == LDLM_FL_WAIT_NOREPROC)
 456                         ldlm_lock_addref_internal_nolock(new2,
 457                                                          lock->l_granted_mode);
 458
 459                 /* insert new2 at lock */
 460                 ldlm_resource_add_lock(res, ownlocks, new2);
 461                 LDLM_LOCK_RELEASE(new2);
 462                 break;
 463         }
 464
 465         /* if new2 is created but never used, destroy it*/
 466         if (splitted == 0 && new2 != NULL)
 467                 ldlm_lock_destroy_nolock(new2);
 468
 469         /* At this point we're granting the lock request. */
 470         req->l_granted_mode = req->l_req_mode;
 471
 472         /* Add req to the granted queue before calling ldlm_reprocess_all(). */
 473         if (!added) {
 474                 cfs_list_del_init(&req->l_res_link);
 475                 /* insert new lock before ownlocks in list. */
 476                 ldlm_resource_add_lock(res, ownlocks, req);
 477         }
 478
 479         if (*flags != LDLM_FL_WAIT_NOREPROC) {
 480 #ifdef HAVE_SERVER_SUPPORT
 481                 if (first_enq) {
 482                         /* If this is an unlock, reprocess the waitq and
 483                          * send completions ASTs for locks that can now be
 484                          * granted. The only problem with doing this
 485                          * reprocessing here is that the completion ASTs for
 486                          * newly granted locks will be sent before the unlock
 487                          * completion is sent. It shouldn't be an issue. Also
 488                          * note that ldlm_process_flock_lock() will recurse,
 489                          * but only once because first_enq will be false from
 490                          * ldlm_reprocess_queue. */
 491                         if ((mode == LCK_NL) && overlaps) {
 492                                 CFS_LIST_HEAD(rpc_list);
 493                                 int rc;
 494 restart:
 495                                 ldlm_reprocess_queue(res, &res->lr_waiting,
 496                                                      &rpc_list);
 497
 498                                 unlock_res_and_lock(req);
 499                                 rc = ldlm_run_ast_work(ns, &rpc_list,
 500                                                        LDLM_WORK_CP_AST);
 501                                 lock_res_and_lock(req);
 502                                 if (rc == -ERESTART)
 503                                         GOTO(restart, -ERESTART);
 504                        }
 505                 } else {
 506                         LASSERT(req->l_completion_ast);
 507                         ldlm_add_ast_work_item(req, NULL, work_list);
 508                 }
 509 #else /* !HAVE_SERVER_SUPPORT */
 510                 /* The only one possible case for client-side calls flock
 511                  * policy function is ldlm_flock_completion_ast inside which
 512                  * carries LDLM_FL_WAIT_NOREPROC flag. */
 513                 CERROR("Illegal parameter for client-side-only module.\n");
 514                 LBUG();
 515 #endif /* HAVE_SERVER_SUPPORT */
 516         }
 517
 518         /* In case we're reprocessing the requested lock we can't destroy
 519          * it until after calling ldlm_ast_work_item() above so that lawi()
 520          * can bump the reference count on req. Otherwise req could be freed
 521          * before the completion AST can be sent.  */
 522         if (added)
 523                 ldlm_flock_destroy(req, mode, *flags);
 524
 525         ldlm_resource_dump(D_INFO, res);
 526         RETURN(LDLM_ITER_CONTINUE);
 527 }
 528
 529 struct ldlm_flock_wait_data {
 530         struct ldlm_lock *fwd_lock;
 531         int               fwd_generation;
 532 };
 533
 534 static void
 535 ldlm_flock_interrupted_wait(void *data)
 536 {
 537         struct ldlm_lock *lock;
 538         ENTRY;
 539
 540         lock = ((struct ldlm_flock_wait_data *)data)->fwd_lock;
 541
 542         /* take lock off the deadlock detection hash list. */
 543         lock_res_and_lock(lock);
 544         ldlm_flock_blocking_unlink(lock);
 545
 546         /* client side - set flag to prevent lock from being put on lru list */
 547         lock->l_flags |= LDLM_FL_CBPENDING;
 548         unlock_res_and_lock(lock);
 549
 550         EXIT;
 551 }
 552
 553 /**
 554  * Flock completion calback function.
 555  *
 556  * \param lock [in,out]: A lock to be handled
 557  * \param flags    [in]: flags
 558  * \param *data    [in]: ldlm_work_cp_ast_lock() will use ldlm_cb_set_arg
 559  *
 560  * \retval 0    : success
 561  * \retval <0   : failure
 562  */
 563 int
 564 ldlm_flock_completion_ast(struct ldlm_lock *lock, __u64 flags, void *data)
 565 {
 566         cfs_flock_t                    *getlk = lock->l_ast_data;
 567         struct obd_device              *obd;
 568         struct obd_import              *imp = NULL;
 569         struct ldlm_flock_wait_data     fwd;
 570         struct l_wait_info              lwi;
 571         ldlm_error_t                    err;
 572         int                             rc = 0;
 573         ENTRY;
 574
 575         CDEBUG(D_DLMTRACE, "flags: 0x%llx data: %p getlk: %p\n",
 576                flags, data, getlk);
 577
 578         /* Import invalidation. We need to actually release the lock
 579          * references being held, so that it can go away. No point in
 580          * holding the lock even if app still believes it has it, since
 581          * server already dropped it anyway. Only for granted locks too. */
 582         if ((lock->l_flags & (LDLM_FL_FAILED|LDLM_FL_LOCAL_ONLY)) ==
 583             (LDLM_FL_FAILED|LDLM_FL_LOCAL_ONLY)) {
 584                 if (lock->l_req_mode == lock->l_granted_mode &&
 585                     lock->l_granted_mode != LCK_NL &&
 586                     NULL == data)
 587                         ldlm_lock_decref_internal(lock, lock->l_req_mode);
 588
 589                 /* Need to wake up the waiter if we were evicted */
 590                 cfs_waitq_signal(&lock->l_waitq);
 591                 RETURN(0);
 592         }
 593
 594         LASSERT(flags != LDLM_FL_WAIT_NOREPROC);
 595
 596         if (!(flags & (LDLM_FL_BLOCK_WAIT | LDLM_FL_BLOCK_GRANTED |
 597                        LDLM_FL_BLOCK_CONV))) {
 598                 if (NULL == data)
 599                         /* mds granted the lock in the reply */
 600                         goto granted;
 601                 /* CP AST RPC: lock get granted, wake it up */
 602                 cfs_waitq_signal(&lock->l_waitq);
 603                 RETURN(0);
 604         }
 605
 606         LDLM_DEBUG(lock, "client-side enqueue returned a blocked lock, "
 607                    "sleeping");
 608         fwd.fwd_lock = lock;
 609         obd = class_exp2obd(lock->l_conn_export);
 610
 611         /* if this is a local lock, there is no import */
 612         if (NULL != obd)
 613                 imp = obd->u.cli.cl_import;
 614
 615         if (NULL != imp) {
 616                 cfs_spin_lock(&imp->imp_lock);
 617                 fwd.fwd_generation = imp->imp_generation;
 618                 cfs_spin_unlock(&imp->imp_lock);
 619         }
 620
 621         lwi = LWI_TIMEOUT_INTR(0, NULL, ldlm_flock_interrupted_wait, &fwd);
 622
 623         /* Go to sleep until the lock is granted. */
 624         rc = l_wait_event(lock->l_waitq, is_granted_or_cancelled(lock), &lwi);
 625
 626         if (rc) {
 627                 LDLM_DEBUG(lock, "client-side enqueue waking up: failed (%d)",
 628                            rc);
 629                 RETURN(rc);
 630         }
 631
 632 granted:
 633         OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_CP_CB_WAIT, 10);
 634
 635         if (lock->l_destroyed) {
 636                 LDLM_DEBUG(lock, "client-side enqueue waking up: destroyed");
 637                 RETURN(0);
 638         }
 639
 640         if (lock->l_flags & LDLM_FL_FAILED) {
 641                 LDLM_DEBUG(lock, "client-side enqueue waking up: failed");
 642                 RETURN(-EIO);
 643         }
 644
 645         if (rc) {
 646                 LDLM_DEBUG(lock, "client-side enqueue waking up: failed (%d)",
 647                            rc);
 648                 RETURN(rc);
 649         }
 650
 651         LDLM_DEBUG(lock, "client-side enqueue granted");
 652
 653         lock_res_and_lock(lock);
 654
 655         /* take lock off the deadlock detection hash list. */
 656         ldlm_flock_blocking_unlink(lock);
 657
 658         /* ldlm_lock_enqueue() has already placed lock on the granted list. */
 659         cfs_list_del_init(&lock->l_res_link);
 660
 661         if (flags & LDLM_FL_TEST_LOCK) {
 662                 /* fcntl(F_GETLK) request */
 663                 /* The old mode was saved in getlk->fl_type so that if the mode
 664                  * in the lock changes we can decref the appropriate refcount.*/
 665                 ldlm_flock_destroy(lock, cfs_flock_type(getlk),
 666                                    LDLM_FL_WAIT_NOREPROC);
 667                 switch (lock->l_granted_mode) {
 668                 case LCK_PR:
 669                         cfs_flock_set_type(getlk, F_RDLCK);
 670                         break;
 671                 case LCK_PW:
 672                         cfs_flock_set_type(getlk, F_WRLCK);
 673                         break;
 674                 default:
 675                         cfs_flock_set_type(getlk, F_UNLCK);
 676                 }
 677                 cfs_flock_set_pid(getlk,
 678                                   (pid_t)lock->l_policy_data.l_flock.pid);
 679                 cfs_flock_set_start(getlk,
 680                                     (loff_t)lock->l_policy_data.l_flock.start);
 681                 cfs_flock_set_end(getlk,
 682                                   (loff_t)lock->l_policy_data.l_flock.end);
 683         } else {
 684                 __u64 noreproc = LDLM_FL_WAIT_NOREPROC;
 685
 686                 /* We need to reprocess the lock to do merges or splits
 687                  * with existing locks owned by this process. */
 688                 ldlm_process_flock_lock(lock, &noreproc, 1, &err, NULL);
 689         }
 690         unlock_res_and_lock(lock);
 691         RETURN(0);
 692 }
 693 EXPORT_SYMBOL(ldlm_flock_completion_ast);
 694
 695 int ldlm_flock_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
 696                             void *data, int flag)
 697 {
 698         ENTRY;
 699
 700         LASSERT(lock);
 701         LASSERT(flag == LDLM_CB_CANCELING);
 702
 703         /* take lock off the deadlock detection hash list. */
 704         lock_res_and_lock(lock);
 705         ldlm_flock_blocking_unlink(lock);
 706         unlock_res_and_lock(lock);
 707         RETURN(0);
 708 }
 709
 710 void ldlm_flock_policy_wire18_to_local(const ldlm_wire_policy_data_t *wpolicy,
 711                                        ldlm_policy_data_t *lpolicy)
 712 {
 713         memset(lpolicy, 0, sizeof(*lpolicy));
 714         lpolicy->l_flock.start = wpolicy->l_flock.lfw_start;
 715         lpolicy->l_flock.end = wpolicy->l_flock.lfw_end;
 716         lpolicy->l_flock.pid = wpolicy->l_flock.lfw_pid;
 717         /* Compat code, old clients had no idea about owner field and
 718          * relied solely on pid for ownership. Introduced in LU-104, 2.1,
 719          * April 2011 */
 720         lpolicy->l_flock.owner = wpolicy->l_flock.lfw_pid;
 721 }
 722
 723
 724 void ldlm_flock_policy_wire21_to_local(const ldlm_wire_policy_data_t *wpolicy,
 725                                        ldlm_policy_data_t *lpolicy)
 726 {
 727         memset(lpolicy, 0, sizeof(*lpolicy));
 728         lpolicy->l_flock.start = wpolicy->l_flock.lfw_start;
 729         lpolicy->l_flock.end = wpolicy->l_flock.lfw_end;
 730         lpolicy->l_flock.pid = wpolicy->l_flock.lfw_pid;
 731         lpolicy->l_flock.owner = wpolicy->l_flock.lfw_owner;
 732 }
 733
 734 void ldlm_flock_policy_local_to_wire(const ldlm_policy_data_t *lpolicy,
 735                                      ldlm_wire_policy_data_t *wpolicy)
 736 {
 737         memset(wpolicy, 0, sizeof(*wpolicy));
 738         wpolicy->l_flock.lfw_start = lpolicy->l_flock.start;
 739         wpolicy->l_flock.lfw_end = lpolicy->l_flock.end;
 740         wpolicy->l_flock.lfw_pid = lpolicy->l_flock.pid;
 741         wpolicy->l_flock.lfw_owner = lpolicy->l_flock.owner;
 742 }
 743
 744 /*
 745  * Export handle<->flock hash operations.
 746  */
 747 static unsigned
 748 ldlm_export_flock_hash(cfs_hash_t *hs, const void *key, unsigned mask)
 749 {
 750         return cfs_hash_u64_hash(*(__u64 *)key, mask);
 751 }
 752
 753 static void *
 754 ldlm_export_flock_key(cfs_hlist_node_t *hnode)
 755 {
 756         struct ldlm_lock *lock;
 757
 758         lock = cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 759         return &lock->l_policy_data.l_flock.owner;
 760 }
 761
 762 static int
 763 ldlm_export_flock_keycmp(const void *key, cfs_hlist_node_t *hnode)
 764 {
 765         return !memcmp(ldlm_export_flock_key(hnode), key, sizeof(__u64));
 766 }
 767
 768 static void *
 769 ldlm_export_flock_object(cfs_hlist_node_t *hnode)
 770 {
 771         return cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 772 }
 773
 774 static void
 775 ldlm_export_flock_get(cfs_hash_t *hs, cfs_hlist_node_t *hnode)
 776 {
 777         struct ldlm_lock *lock;
 778         struct ldlm_flock *flock;
 779
 780         lock = cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 781         LDLM_LOCK_GET(lock);
 782
 783         flock = &lock->l_policy_data.l_flock;
 784         LASSERT(flock->blocking_export != NULL);
 785         class_export_get(flock->blocking_export);
 786         flock->blocking_refs++;
 787 }
 788
 789 static void
 790 ldlm_export_flock_put(cfs_hash_t *hs, cfs_hlist_node_t *hnode)
 791 {
 792         struct ldlm_lock *lock;
 793         struct ldlm_flock *flock;
 794
 795         lock = cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 796         LDLM_LOCK_RELEASE(lock);
 797
 798         flock = &lock->l_policy_data.l_flock;
 799         LASSERT(flock->blocking_export != NULL);
 800         class_export_put(flock->blocking_export);
 801         if (--flock->blocking_refs == 0) {
 802                 flock->blocking_owner = 0;
 803                 flock->blocking_export = NULL;
 804         }
 805 }
 806
 807 static cfs_hash_ops_t ldlm_export_flock_ops = {
 808         .hs_hash        = ldlm_export_flock_hash,
 809         .hs_key         = ldlm_export_flock_key,
 810         .hs_keycmp      = ldlm_export_flock_keycmp,
 811         .hs_object      = ldlm_export_flock_object,
 812         .hs_get         = ldlm_export_flock_get,
 813         .hs_put         = ldlm_export_flock_put,
 814         .hs_put_locked  = ldlm_export_flock_put,
 815 };
 816
 817 int ldlm_init_flock_export(struct obd_export *exp)
 818 {
 819         exp->exp_flock_hash =
 820                 cfs_hash_create(obd_uuid2str(&exp->exp_client_uuid),
 821                                 HASH_EXP_LOCK_CUR_BITS,
 822                                 HASH_EXP_LOCK_MAX_BITS,
 823                                 HASH_EXP_LOCK_BKT_BITS, 0,
 824                                 CFS_HASH_MIN_THETA, CFS_HASH_MAX_THETA,
 825                                 &ldlm_export_flock_ops,
 826                                 CFS_HASH_DEFAULT | CFS_HASH_NBLK_CHANGE);
 827         if (!exp->exp_flock_hash)
 828                 RETURN(-ENOMEM);
 829
 830         RETURN(0);
 831 }
 832 EXPORT_SYMBOL(ldlm_init_flock_export);
 833
 834 void ldlm_destroy_flock_export(struct obd_export *exp)
 835 {
 836         ENTRY;
 837         if (exp->exp_flock_hash) {
 838                 cfs_hash_putref(exp->exp_flock_hash);
 839                 exp->exp_flock_hash = NULL;
 840         }
 841         EXIT;
 842 }
 843 EXPORT_SYMBOL(ldlm_destroy_flock_export);