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, 2012, Intel Corporation.
  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 /**
  42  * This file implements POSIX lock type for Lustre.
  43  * Its policy properties are start and end of extent and PID.
  44  *
  45  * These locks are only done through MDS due to POSIX semantics requiring
  46  * e.g. that locks could be only partially released and as such split into
  47  * two parts, and also that two adjacent locks from the same process may be
  48  * merged into a single wider lock.
  49  *
  50  * Lock modes are mapped like this:
  51  * PR and PW for READ and WRITE locks
  52  * NL to request a releasing of a portion of the lock
  53  *
  54  * These flock locks never timeout.
  55  */
  56
  57 #define DEBUG_SUBSYSTEM S_LDLM
  58
  59 #ifdef __KERNEL__
  60 #include <lustre_dlm.h>
  61 #include <obd_support.h>
  62 #include <obd_class.h>
  63 #include <lustre_lib.h>
  64 #include <libcfs/list.h>
  65 #else
  66 #include <liblustre.h>
  67 #include <obd_class.h>
  68 #endif
  69
  70 #include "ldlm_internal.h"
  71
  72 int ldlm_flock_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
  73                             void *data, int flag);
  74
  75 /**
  76  * list_for_remaining_safe - iterate over the remaining entries in a list
  77  *              and safeguard against removal of a list entry.
  78  * \param pos   the &struct list_head to use as a loop counter. pos MUST
  79  *              have been initialized prior to using it in this macro.
  80  * \param n     another &struct list_head to use as temporary storage
  81  * \param head  the head for your list.
  82  */
  83 #define list_for_remaining_safe(pos, n, head) \
  84         for (n = pos->next; pos != (head); pos = n, n = pos->next)
  85
  86 static inline int
  87 ldlm_same_flock_owner(struct ldlm_lock *lock, struct ldlm_lock *new)
  88 {
  89         return((new->l_policy_data.l_flock.owner ==
  90                 lock->l_policy_data.l_flock.owner) &&
  91                (new->l_export == lock->l_export));
  92 }
  93
  94 static inline int
  95 ldlm_flocks_overlap(struct ldlm_lock *lock, struct ldlm_lock *new)
  96 {
  97         return((new->l_policy_data.l_flock.start <=
  98                 lock->l_policy_data.l_flock.end) &&
  99                (new->l_policy_data.l_flock.end >=
 100                 lock->l_policy_data.l_flock.start));
 101 }
 102
 103 static inline void ldlm_flock_blocking_link(struct ldlm_lock *req,
 104                                             struct ldlm_lock *lock)
 105 {
 106         /* For server only */
 107         if (req->l_export == NULL)
 108                 return;
 109
 110         LASSERT(cfs_hlist_unhashed(&req->l_exp_flock_hash));
 111
 112         req->l_policy_data.l_flock.blocking_owner =
 113                 lock->l_policy_data.l_flock.owner;
 114         req->l_policy_data.l_flock.blocking_export =
 115                 lock->l_export;
 116         req->l_policy_data.l_flock.blocking_refs = 0;
 117
 118         cfs_hash_add(req->l_export->exp_flock_hash,
 119                      &req->l_policy_data.l_flock.owner,
 120                      &req->l_exp_flock_hash);
 121 }
 122
 123 static inline void ldlm_flock_blocking_unlink(struct ldlm_lock *req)
 124 {
 125         /* For server only */
 126         if (req->l_export == NULL)
 127                 return;
 128
 129         check_res_locked(req->l_resource);
 130         if (req->l_export->exp_flock_hash != NULL &&
 131             !cfs_hlist_unhashed(&req->l_exp_flock_hash))
 132                 cfs_hash_del(req->l_export->exp_flock_hash,
 133                              &req->l_policy_data.l_flock.owner,
 134                              &req->l_exp_flock_hash);
 135 }
 136
 137 static inline void
 138 ldlm_flock_destroy(struct ldlm_lock *lock, ldlm_mode_t mode, __u64 flags)
 139 {
 140         ENTRY;
 141
 142         LDLM_DEBUG(lock, "ldlm_flock_destroy(mode: %d, flags: 0x%llx)",
 143                    mode, flags);
 144
 145         /* Safe to not lock here, since it should be empty anyway */
 146         LASSERT(cfs_hlist_unhashed(&lock->l_exp_flock_hash));
 147
 148         cfs_list_del_init(&lock->l_res_link);
 149         if (flags == LDLM_FL_WAIT_NOREPROC &&
 150             !(lock->l_flags & LDLM_FL_FAILED)) {
 151                 /* client side - set a flag to prevent sending a CANCEL */
 152                 lock->l_flags |= LDLM_FL_LOCAL_ONLY | LDLM_FL_CBPENDING;
 153
 154                 /* when reaching here, it is under lock_res_and_lock(). Thus,
 155                    need call the nolock version of ldlm_lock_decref_internal*/
 156                 ldlm_lock_decref_internal_nolock(lock, mode);
 157         }
 158
 159         ldlm_lock_destroy_nolock(lock);
 160         EXIT;
 161 }
 162
 163 /**
 164  * POSIX locks deadlock detection code.
 165  *
 166  * Given a new lock \a req and an existing lock \a bl_lock it conflicts
 167  * with, we need to iterate through all blocked POSIX locks for this
 168  * export and see if there is a deadlock condition arising. (i.e. when
 169  * one client holds a lock on something and want a lock on something
 170  * else and at the same time another client has the opposite situation).
 171  */
 172 static int
 173 ldlm_flock_deadlock(struct ldlm_lock *req, struct ldlm_lock *bl_lock)
 174 {
 175         struct obd_export *req_exp = req->l_export;
 176         struct obd_export *bl_exp = bl_lock->l_export;
 177         __u64 req_owner = req->l_policy_data.l_flock.owner;
 178         __u64 bl_owner = bl_lock->l_policy_data.l_flock.owner;
 179
 180         /* For server only */
 181         if (req_exp == NULL)
 182                 return 0;
 183
 184         class_export_get(bl_exp);
 185         while (1) {
 186                 struct obd_export *bl_exp_new;
 187                 struct ldlm_lock *lock = NULL;
 188                 struct ldlm_flock *flock;
 189
 190                 if (bl_exp->exp_flock_hash != NULL)
 191                         lock = cfs_hash_lookup(bl_exp->exp_flock_hash,
 192                                                &bl_owner);
 193                 if (lock == NULL)
 194                         break;
 195
 196                 LASSERT(req != lock);
 197                 flock = &lock->l_policy_data.l_flock;
 198                 LASSERT(flock->owner == bl_owner);
 199                 bl_owner = flock->blocking_owner;
 200                 bl_exp_new = class_export_get(flock->blocking_export);
 201                 class_export_put(bl_exp);
 202
 203                 cfs_hash_put(bl_exp->exp_flock_hash, &lock->l_exp_flock_hash);
 204                 bl_exp = bl_exp_new;
 205
 206                 if (bl_owner == req_owner && bl_exp == req_exp) {
 207                         class_export_put(bl_exp);
 208                         return 1;
 209                 }
 210         }
 211         class_export_put(bl_exp);
 212
 213         return 0;
 214 }
 215
 216 /**
 217  * Process a granting attempt for flock lock.
 218  * Must be called under ns lock held.
 219  *
 220  * This function looks for any conflicts for \a lock in the granted or
 221  * waiting queues. The lock is granted if no conflicts are found in
 222  * either queue.
 223  *
 224  * It is also responsible for splitting a lock if a portion of the lock
 225  * is released.
 226  *
 227  * If \a first_enq is 0 (ie, called from ldlm_reprocess_queue):
 228  *   - blocking ASTs have already been sent
 229  *
 230  * If \a first_enq is 1 (ie, called from ldlm_lock_enqueue):
 231  *   - blocking ASTs have not been sent yet, so list of conflicting locks
 232  *     would be collected and ASTs sent.
 233  */
 234 int
 235 ldlm_process_flock_lock(struct ldlm_lock *req, __u64 *flags, int first_enq,
 236                         ldlm_error_t *err, cfs_list_t *work_list)
 237 {
 238         struct ldlm_resource *res = req->l_resource;
 239         struct ldlm_namespace *ns = ldlm_res_to_ns(res);
 240         cfs_list_t *tmp;
 241         cfs_list_t *ownlocks = NULL;
 242         struct ldlm_lock *lock = NULL;
 243         struct ldlm_lock *new = req;
 244         struct ldlm_lock *new2 = NULL;
 245         ldlm_mode_t mode = req->l_req_mode;
 246         int local = ns_is_client(ns);
 247         int added = (mode == LCK_NL);
 248         int overlaps = 0;
 249         int splitted = 0;
 250         const struct ldlm_callback_suite null_cbs = { NULL };
 251         ENTRY;
 252
 253         CDEBUG(D_DLMTRACE, "flags %#llx owner "LPU64" pid %u mode %u start "
 254                LPU64" end "LPU64"\n", *flags,
 255                new->l_policy_data.l_flock.owner,
 256                new->l_policy_data.l_flock.pid, mode,
 257                req->l_policy_data.l_flock.start,
 258                req->l_policy_data.l_flock.end);
 259
 260         *err = ELDLM_OK;
 261
 262         if (local) {
 263                 /* No blocking ASTs are sent to the clients for
 264                  * Posix file & record locks */
 265                 req->l_blocking_ast = NULL;
 266         } else {
 267                 /* Called on the server for lock cancels. */
 268                 req->l_blocking_ast = ldlm_flock_blocking_ast;
 269         }
 270
 271 reprocess:
 272         if ((*flags == LDLM_FL_WAIT_NOREPROC) || (mode == LCK_NL)) {
 273                 /* This loop determines where this processes locks start
 274                  * in the resource lr_granted list. */
 275                 cfs_list_for_each(tmp, &res->lr_granted) {
 276                         lock = cfs_list_entry(tmp, struct ldlm_lock,
 277                                               l_res_link);
 278                         if (ldlm_same_flock_owner(lock, req)) {
 279                                 ownlocks = tmp;
 280                                 break;
 281                         }
 282                 }
 283         } else {
 284                 lockmode_verify(mode);
 285
 286                 /* This loop determines if there are existing locks
 287                  * that conflict with the new lock request. */
 288                 cfs_list_for_each(tmp, &res->lr_granted) {
 289                         lock = cfs_list_entry(tmp, struct ldlm_lock,
 290                                               l_res_link);
 291
 292                         if (ldlm_same_flock_owner(lock, req)) {
 293                                 if (!ownlocks)
 294                                         ownlocks = tmp;
 295                                 continue;
 296                         }
 297
 298                         /* locks are compatible, overlap doesn't matter */
 299                         if (lockmode_compat(lock->l_granted_mode, mode))
 300                                 continue;
 301
 302                         if (!ldlm_flocks_overlap(lock, req))
 303                                 continue;
 304
 305                         if (!first_enq)
 306                                 RETURN(LDLM_ITER_CONTINUE);
 307
 308                         if (*flags & LDLM_FL_BLOCK_NOWAIT) {
 309                                 ldlm_flock_destroy(req, mode, *flags);
 310                                 *err = -EAGAIN;
 311                                 RETURN(LDLM_ITER_STOP);
 312                         }
 313
 314                         if (*flags & LDLM_FL_TEST_LOCK) {
 315                                 ldlm_flock_destroy(req, mode, *flags);
 316                                 req->l_req_mode = lock->l_granted_mode;
 317                                 req->l_policy_data.l_flock.pid =
 318                                         lock->l_policy_data.l_flock.pid;
 319                                 req->l_policy_data.l_flock.start =
 320                                         lock->l_policy_data.l_flock.start;
 321                                 req->l_policy_data.l_flock.end =
 322                                         lock->l_policy_data.l_flock.end;
 323                                 *flags |= LDLM_FL_LOCK_CHANGED;
 324                                 RETURN(LDLM_ITER_STOP);
 325                         }
 326
 327                         /* add lock to blocking list before deadlock
 328                          * check to prevent race */
 329                         ldlm_flock_blocking_link(req, lock);
 330
 331                         if (ldlm_flock_deadlock(req, lock)) {
 332                                 ldlm_flock_blocking_unlink(req);
 333                                 ldlm_flock_destroy(req, mode, *flags);
 334                                 *err = -EDEADLK;
 335                                 RETURN(LDLM_ITER_STOP);
 336                         }
 337
 338                         ldlm_resource_add_lock(res, &res->lr_waiting, req);
 339                         *flags |= LDLM_FL_BLOCK_GRANTED;
 340                         RETURN(LDLM_ITER_STOP);
 341                 }
 342         }
 343
 344         if (*flags & LDLM_FL_TEST_LOCK) {
 345                 ldlm_flock_destroy(req, mode, *flags);
 346                 req->l_req_mode = LCK_NL;
 347                 *flags |= LDLM_FL_LOCK_CHANGED;
 348                 RETURN(LDLM_ITER_STOP);
 349         }
 350
 351         /* In case we had slept on this lock request take it off of the
 352          * deadlock detection hash list. */
 353         ldlm_flock_blocking_unlink(req);
 354
 355         /* Scan the locks owned by this process that overlap this request.
 356          * We may have to merge or split existing locks. */
 357
 358         if (!ownlocks)
 359                 ownlocks = &res->lr_granted;
 360
 361         list_for_remaining_safe(ownlocks, tmp, &res->lr_granted) {
 362                 lock = cfs_list_entry(ownlocks, struct ldlm_lock, l_res_link);
 363
 364                 if (!ldlm_same_flock_owner(lock, new))
 365                         break;
 366
 367                 if (lock->l_granted_mode == mode) {
 368                         /* If the modes are the same then we need to process
 369                          * locks that overlap OR adjoin the new lock. The extra
 370                          * logic condition is necessary to deal with arithmetic
 371                          * overflow and underflow. */
 372                         if ((new->l_policy_data.l_flock.start >
 373                              (lock->l_policy_data.l_flock.end + 1))
 374                             && (lock->l_policy_data.l_flock.end !=
 375                                 OBD_OBJECT_EOF))
 376                                 continue;
 377
 378                         if ((new->l_policy_data.l_flock.end <
 379                              (lock->l_policy_data.l_flock.start - 1))
 380                             && (lock->l_policy_data.l_flock.start != 0))
 381                                 break;
 382
 383                         if (new->l_policy_data.l_flock.start <
 384                             lock->l_policy_data.l_flock.start) {
 385                                 lock->l_policy_data.l_flock.start =
 386                                         new->l_policy_data.l_flock.start;
 387                         } else {
 388                                 new->l_policy_data.l_flock.start =
 389                                         lock->l_policy_data.l_flock.start;
 390                         }
 391
 392                         if (new->l_policy_data.l_flock.end >
 393                             lock->l_policy_data.l_flock.end) {
 394                                 lock->l_policy_data.l_flock.end =
 395                                         new->l_policy_data.l_flock.end;
 396                         } else {
 397                                 new->l_policy_data.l_flock.end =
 398                                         lock->l_policy_data.l_flock.end;
 399                         }
 400
 401                         if (added) {
 402                                 ldlm_flock_destroy(lock, mode, *flags);
 403                         } else {
 404                                 new = lock;
 405                                 added = 1;
 406                         }
 407                         continue;
 408                 }
 409
 410                 if (new->l_policy_data.l_flock.start >
 411                     lock->l_policy_data.l_flock.end)
 412                         continue;
 413
 414                 if (new->l_policy_data.l_flock.end <
 415                     lock->l_policy_data.l_flock.start)
 416                         break;
 417
 418                 ++overlaps;
 419
 420                 if (new->l_policy_data.l_flock.start <=
 421                     lock->l_policy_data.l_flock.start) {
 422                         if (new->l_policy_data.l_flock.end <
 423                             lock->l_policy_data.l_flock.end) {
 424                                 lock->l_policy_data.l_flock.start =
 425                                         new->l_policy_data.l_flock.end + 1;
 426                                 break;
 427                         }
 428                         ldlm_flock_destroy(lock, lock->l_req_mode, *flags);
 429                         continue;
 430                 }
 431                 if (new->l_policy_data.l_flock.end >=
 432                     lock->l_policy_data.l_flock.end) {
 433                         lock->l_policy_data.l_flock.end =
 434                                 new->l_policy_data.l_flock.start - 1;
 435                         continue;
 436                 }
 437
 438                 /* split the existing lock into two locks */
 439
 440                 /* if this is an F_UNLCK operation then we could avoid
 441                  * allocating a new lock and use the req lock passed in
 442                  * with the request but this would complicate the reply
 443                  * processing since updates to req get reflected in the
 444                  * reply. The client side replays the lock request so
 445                  * it must see the original lock data in the reply. */
 446
 447                 /* XXX - if ldlm_lock_new() can sleep we should
 448                  * release the lr_lock, allocate the new lock,
 449                  * and restart processing this lock. */
 450                 if (!new2) {
 451                         unlock_res_and_lock(req);
 452                         new2 = ldlm_lock_create(ns, &res->lr_name, LDLM_FLOCK,
 453                                                 lock->l_granted_mode, &null_cbs,
 454                                                 NULL, 0, LVB_T_NONE);
 455                         lock_res_and_lock(req);
 456                         if (!new2) {
 457                                 ldlm_flock_destroy(req, lock->l_granted_mode,
 458                                                    *flags);
 459                                 *err = -ENOLCK;
 460                                 RETURN(LDLM_ITER_STOP);
 461                         }
 462                         goto reprocess;
 463                 }
 464
 465                 splitted = 1;
 466
 467                 new2->l_granted_mode = lock->l_granted_mode;
 468                 new2->l_policy_data.l_flock.pid =
 469                         new->l_policy_data.l_flock.pid;
 470                 new2->l_policy_data.l_flock.owner =
 471                         new->l_policy_data.l_flock.owner;
 472                 new2->l_policy_data.l_flock.start =
 473                         lock->l_policy_data.l_flock.start;
 474                 new2->l_policy_data.l_flock.end =
 475                         new->l_policy_data.l_flock.start - 1;
 476                 lock->l_policy_data.l_flock.start =
 477                         new->l_policy_data.l_flock.end + 1;
 478                 new2->l_conn_export = lock->l_conn_export;
 479                 if (lock->l_export != NULL) {
 480                         new2->l_export = class_export_lock_get(lock->l_export, new2);
 481                         if (new2->l_export->exp_lock_hash &&
 482                             cfs_hlist_unhashed(&new2->l_exp_hash))
 483                                 cfs_hash_add(new2->l_export->exp_lock_hash,
 484                                              &new2->l_remote_handle,
 485                                              &new2->l_exp_hash);
 486                 }
 487                 if (*flags == LDLM_FL_WAIT_NOREPROC)
 488                         ldlm_lock_addref_internal_nolock(new2,
 489                                                          lock->l_granted_mode);
 490
 491                 /* insert new2 at lock */
 492                 ldlm_resource_add_lock(res, ownlocks, new2);
 493                 LDLM_LOCK_RELEASE(new2);
 494                 break;
 495         }
 496
 497         /* if new2 is created but never used, destroy it*/
 498         if (splitted == 0 && new2 != NULL)
 499                 ldlm_lock_destroy_nolock(new2);
 500
 501         /* At this point we're granting the lock request. */
 502         req->l_granted_mode = req->l_req_mode;
 503
 504         /* Add req to the granted queue before calling ldlm_reprocess_all(). */
 505         if (!added) {
 506                 cfs_list_del_init(&req->l_res_link);
 507                 /* insert new lock before ownlocks in list. */
 508                 ldlm_resource_add_lock(res, ownlocks, req);
 509         }
 510
 511         if (*flags != LDLM_FL_WAIT_NOREPROC) {
 512 #ifdef HAVE_SERVER_SUPPORT
 513                 if (first_enq) {
 514                         /* If this is an unlock, reprocess the waitq and
 515                          * send completions ASTs for locks that can now be
 516                          * granted. The only problem with doing this
 517                          * reprocessing here is that the completion ASTs for
 518                          * newly granted locks will be sent before the unlock
 519                          * completion is sent. It shouldn't be an issue. Also
 520                          * note that ldlm_process_flock_lock() will recurse,
 521                          * but only once because first_enq will be false from
 522                          * ldlm_reprocess_queue. */
 523                         if ((mode == LCK_NL) && overlaps) {
 524                                 CFS_LIST_HEAD(rpc_list);
 525                                 int rc;
 526 restart:
 527                                 ldlm_reprocess_queue(res, &res->lr_waiting,
 528                                                      &rpc_list);
 529
 530                                 unlock_res_and_lock(req);
 531                                 rc = ldlm_run_ast_work(ns, &rpc_list,
 532                                                        LDLM_WORK_CP_AST);
 533                                 lock_res_and_lock(req);
 534                                 if (rc == -ERESTART)
 535                                         GOTO(restart, -ERESTART);
 536                        }
 537                 } else {
 538                         LASSERT(req->l_completion_ast);
 539                         ldlm_add_ast_work_item(req, NULL, work_list);
 540                 }
 541 #else /* !HAVE_SERVER_SUPPORT */
 542                 /* The only one possible case for client-side calls flock
 543                  * policy function is ldlm_flock_completion_ast inside which
 544                  * carries LDLM_FL_WAIT_NOREPROC flag. */
 545                 CERROR("Illegal parameter for client-side-only module.\n");
 546                 LBUG();
 547 #endif /* HAVE_SERVER_SUPPORT */
 548         }
 549
 550         /* In case we're reprocessing the requested lock we can't destroy
 551          * it until after calling ldlm_add_ast_work_item() above so that laawi()
 552          * can bump the reference count on \a req. Otherwise \a req
 553          * could be freed before the completion AST can be sent.  */
 554         if (added)
 555                 ldlm_flock_destroy(req, mode, *flags);
 556
 557         ldlm_resource_dump(D_INFO, res);
 558         RETURN(LDLM_ITER_CONTINUE);
 559 }
 560
 561 struct ldlm_flock_wait_data {
 562         struct ldlm_lock *fwd_lock;
 563         int               fwd_generation;
 564 };
 565
 566 static void
 567 ldlm_flock_interrupted_wait(void *data)
 568 {
 569         struct ldlm_lock *lock;
 570         ENTRY;
 571
 572         lock = ((struct ldlm_flock_wait_data *)data)->fwd_lock;
 573
 574         /* take lock off the deadlock detection hash list. */
 575         lock_res_and_lock(lock);
 576         ldlm_flock_blocking_unlink(lock);
 577
 578         /* client side - set flag to prevent lock from being put on LRU list */
 579         lock->l_flags |= LDLM_FL_CBPENDING;
 580         unlock_res_and_lock(lock);
 581
 582         EXIT;
 583 }
 584
 585 /**
 586  * Flock completion callback function.
 587  *
 588  * \param lock [in,out]: A lock to be handled
 589  * \param flags    [in]: flags
 590  * \param *data    [in]: ldlm_work_cp_ast_lock() will use ldlm_cb_set_arg
 591  *
 592  * \retval 0    : success
 593  * \retval <0   : failure
 594  */
 595 int
 596 ldlm_flock_completion_ast(struct ldlm_lock *lock, __u64 flags, void *data)
 597 {
 598         struct file_lock                *getlk = lock->l_ast_data;
 599         struct obd_device              *obd;
 600         struct obd_import              *imp = NULL;
 601         struct ldlm_flock_wait_data     fwd;
 602         struct l_wait_info              lwi;
 603         ldlm_error_t                    err;
 604         int                             rc = 0;
 605         ENTRY;
 606
 607         CDEBUG(D_DLMTRACE, "flags: 0x%llx data: %p getlk: %p\n",
 608                flags, data, getlk);
 609
 610         /* Import invalidation. We need to actually release the lock
 611          * references being held, so that it can go away. No point in
 612          * holding the lock even if app still believes it has it, since
 613          * server already dropped it anyway. Only for granted locks too. */
 614         if ((lock->l_flags & (LDLM_FL_FAILED|LDLM_FL_LOCAL_ONLY)) ==
 615             (LDLM_FL_FAILED|LDLM_FL_LOCAL_ONLY)) {
 616                 if (lock->l_req_mode == lock->l_granted_mode &&
 617                     lock->l_granted_mode != LCK_NL &&
 618                     NULL == data)
 619                         ldlm_lock_decref_internal(lock, lock->l_req_mode);
 620
 621                 /* Need to wake up the waiter if we were evicted */
 622                 cfs_waitq_signal(&lock->l_waitq);
 623                 RETURN(0);
 624         }
 625
 626         LASSERT(flags != LDLM_FL_WAIT_NOREPROC);
 627
 628         if (!(flags & (LDLM_FL_BLOCK_WAIT | LDLM_FL_BLOCK_GRANTED |
 629                        LDLM_FL_BLOCK_CONV))) {
 630                 if (NULL == data)
 631                         /* mds granted the lock in the reply */
 632                         goto granted;
 633                 /* CP AST RPC: lock get granted, wake it up */
 634                 cfs_waitq_signal(&lock->l_waitq);
 635                 RETURN(0);
 636         }
 637
 638         LDLM_DEBUG(lock, "client-side enqueue returned a blocked lock, "
 639                    "sleeping");
 640         fwd.fwd_lock = lock;
 641         obd = class_exp2obd(lock->l_conn_export);
 642
 643         /* if this is a local lock, there is no import */
 644         if (NULL != obd)
 645                 imp = obd->u.cli.cl_import;
 646
 647         if (NULL != imp) {
 648                 spin_lock(&imp->imp_lock);
 649                 fwd.fwd_generation = imp->imp_generation;
 650                 spin_unlock(&imp->imp_lock);
 651         }
 652
 653         lwi = LWI_TIMEOUT_INTR(0, NULL, ldlm_flock_interrupted_wait, &fwd);
 654
 655         /* Go to sleep until the lock is granted. */
 656         rc = l_wait_event(lock->l_waitq, is_granted_or_cancelled(lock), &lwi);
 657
 658         if (rc) {
 659                 LDLM_DEBUG(lock, "client-side enqueue waking up: failed (%d)",
 660                            rc);
 661                 RETURN(rc);
 662         }
 663
 664 granted:
 665         OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_CP_CB_WAIT, 10);
 666
 667         if (lock->l_flags & LDLM_FL_DESTROYED) {
 668                 LDLM_DEBUG(lock, "client-side enqueue waking up: destroyed");
 669                 RETURN(0);
 670         }
 671
 672         if (lock->l_flags & LDLM_FL_FAILED) {
 673                 LDLM_DEBUG(lock, "client-side enqueue waking up: failed");
 674                 RETURN(-EIO);
 675         }
 676
 677         if (rc) {
 678                 LDLM_DEBUG(lock, "client-side enqueue waking up: failed (%d)",
 679                            rc);
 680                 RETURN(rc);
 681         }
 682
 683         LDLM_DEBUG(lock, "client-side enqueue granted");
 684
 685         lock_res_and_lock(lock);
 686
 687         /* take lock off the deadlock detection hash list. */
 688         ldlm_flock_blocking_unlink(lock);
 689
 690         /* ldlm_lock_enqueue() has already placed lock on the granted list. */
 691         cfs_list_del_init(&lock->l_res_link);
 692
 693         if (flags & LDLM_FL_TEST_LOCK) {
 694                 /* fcntl(F_GETLK) request */
 695                 /* The old mode was saved in getlk->fl_type so that if the mode
 696                  * in the lock changes we can decref the appropriate refcount.*/
 697                 ldlm_flock_destroy(lock, flock_type(getlk),
 698                                    LDLM_FL_WAIT_NOREPROC);
 699                 switch (lock->l_granted_mode) {
 700                 case LCK_PR:
 701                         flock_set_type(getlk, F_RDLCK);
 702                         break;
 703                 case LCK_PW:
 704                         flock_set_type(getlk, F_WRLCK);
 705                         break;
 706                 default:
 707                         flock_set_type(getlk, F_UNLCK);
 708                 }
 709                 flock_set_pid(getlk, (pid_t)lock->l_policy_data.l_flock.pid);
 710                 flock_set_start(getlk,
 711                                 (loff_t)lock->l_policy_data.l_flock.start);
 712                 flock_set_end(getlk,
 713                               (loff_t)lock->l_policy_data.l_flock.end);
 714         } else {
 715                 __u64 noreproc = LDLM_FL_WAIT_NOREPROC;
 716
 717                 /* We need to reprocess the lock to do merges or splits
 718                  * with existing locks owned by this process. */
 719                 ldlm_process_flock_lock(lock, &noreproc, 1, &err, NULL);
 720         }
 721         unlock_res_and_lock(lock);
 722         RETURN(0);
 723 }
 724 EXPORT_SYMBOL(ldlm_flock_completion_ast);
 725
 726 int ldlm_flock_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
 727                             void *data, int flag)
 728 {
 729         ENTRY;
 730
 731         LASSERT(lock);
 732         LASSERT(flag == LDLM_CB_CANCELING);
 733
 734         /* take lock off the deadlock detection hash list. */
 735         lock_res_and_lock(lock);
 736         ldlm_flock_blocking_unlink(lock);
 737         unlock_res_and_lock(lock);
 738         RETURN(0);
 739 }
 740
 741 void ldlm_flock_policy_wire18_to_local(const ldlm_wire_policy_data_t *wpolicy,
 742                                        ldlm_policy_data_t *lpolicy)
 743 {
 744         memset(lpolicy, 0, sizeof(*lpolicy));
 745         lpolicy->l_flock.start = wpolicy->l_flock.lfw_start;
 746         lpolicy->l_flock.end = wpolicy->l_flock.lfw_end;
 747         lpolicy->l_flock.pid = wpolicy->l_flock.lfw_pid;
 748         /* Compat code, old clients had no idea about owner field and
 749          * relied solely on pid for ownership. Introduced in LU-104, 2.1,
 750          * April 2011 */
 751         lpolicy->l_flock.owner = wpolicy->l_flock.lfw_pid;
 752 }
 753
 754
 755 void ldlm_flock_policy_wire21_to_local(const ldlm_wire_policy_data_t *wpolicy,
 756                                        ldlm_policy_data_t *lpolicy)
 757 {
 758         memset(lpolicy, 0, sizeof(*lpolicy));
 759         lpolicy->l_flock.start = wpolicy->l_flock.lfw_start;
 760         lpolicy->l_flock.end = wpolicy->l_flock.lfw_end;
 761         lpolicy->l_flock.pid = wpolicy->l_flock.lfw_pid;
 762         lpolicy->l_flock.owner = wpolicy->l_flock.lfw_owner;
 763 }
 764
 765 void ldlm_flock_policy_local_to_wire(const ldlm_policy_data_t *lpolicy,
 766                                      ldlm_wire_policy_data_t *wpolicy)
 767 {
 768         memset(wpolicy, 0, sizeof(*wpolicy));
 769         wpolicy->l_flock.lfw_start = lpolicy->l_flock.start;
 770         wpolicy->l_flock.lfw_end = lpolicy->l_flock.end;
 771         wpolicy->l_flock.lfw_pid = lpolicy->l_flock.pid;
 772         wpolicy->l_flock.lfw_owner = lpolicy->l_flock.owner;
 773 }
 774
 775 /*
 776  * Export handle<->flock hash operations.
 777  */
 778 static unsigned
 779 ldlm_export_flock_hash(cfs_hash_t *hs, const void *key, unsigned mask)
 780 {
 781         return cfs_hash_u64_hash(*(__u64 *)key, mask);
 782 }
 783
 784 static void *
 785 ldlm_export_flock_key(cfs_hlist_node_t *hnode)
 786 {
 787         struct ldlm_lock *lock;
 788
 789         lock = cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 790         return &lock->l_policy_data.l_flock.owner;
 791 }
 792
 793 static int
 794 ldlm_export_flock_keycmp(const void *key, cfs_hlist_node_t *hnode)
 795 {
 796         return !memcmp(ldlm_export_flock_key(hnode), key, sizeof(__u64));
 797 }
 798
 799 static void *
 800 ldlm_export_flock_object(cfs_hlist_node_t *hnode)
 801 {
 802         return cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 803 }
 804
 805 static void
 806 ldlm_export_flock_get(cfs_hash_t *hs, cfs_hlist_node_t *hnode)
 807 {
 808         struct ldlm_lock *lock;
 809         struct ldlm_flock *flock;
 810
 811         lock = cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 812         LDLM_LOCK_GET(lock);
 813
 814         flock = &lock->l_policy_data.l_flock;
 815         LASSERT(flock->blocking_export != NULL);
 816         class_export_get(flock->blocking_export);
 817         flock->blocking_refs++;
 818 }
 819
 820 static void
 821 ldlm_export_flock_put(cfs_hash_t *hs, cfs_hlist_node_t *hnode)
 822 {
 823         struct ldlm_lock *lock;
 824         struct ldlm_flock *flock;
 825
 826         lock = cfs_hlist_entry(hnode, struct ldlm_lock, l_exp_flock_hash);
 827         LDLM_LOCK_RELEASE(lock);
 828
 829         flock = &lock->l_policy_data.l_flock;
 830         LASSERT(flock->blocking_export != NULL);
 831         class_export_put(flock->blocking_export);
 832         if (--flock->blocking_refs == 0) {
 833                 flock->blocking_owner = 0;
 834                 flock->blocking_export = NULL;
 835         }
 836 }
 837
 838 static cfs_hash_ops_t ldlm_export_flock_ops = {
 839         .hs_hash        = ldlm_export_flock_hash,
 840         .hs_key         = ldlm_export_flock_key,
 841         .hs_keycmp      = ldlm_export_flock_keycmp,
 842         .hs_object      = ldlm_export_flock_object,
 843         .hs_get         = ldlm_export_flock_get,
 844         .hs_put         = ldlm_export_flock_put,
 845         .hs_put_locked  = ldlm_export_flock_put,
 846 };
 847
 848 int ldlm_init_flock_export(struct obd_export *exp)
 849 {
 850         if( strcmp(exp->exp_obd->obd_type->typ_name, LUSTRE_MDT_NAME) != 0)
 851                 RETURN(0);
 852
 853         exp->exp_flock_hash =
 854                 cfs_hash_create(obd_uuid2str(&exp->exp_client_uuid),
 855                                 HASH_EXP_LOCK_CUR_BITS,
 856                                 HASH_EXP_LOCK_MAX_BITS,
 857                                 HASH_EXP_LOCK_BKT_BITS, 0,
 858                                 CFS_HASH_MIN_THETA, CFS_HASH_MAX_THETA,
 859                                 &ldlm_export_flock_ops,
 860                                 CFS_HASH_DEFAULT | CFS_HASH_NBLK_CHANGE);
 861         if (!exp->exp_flock_hash)
 862                 RETURN(-ENOMEM);
 863
 864         RETURN(0);
 865 }
 866 EXPORT_SYMBOL(ldlm_init_flock_export);
 867
 868 void ldlm_destroy_flock_export(struct obd_export *exp)
 869 {
 870         ENTRY;
 871         if (exp->exp_flock_hash) {
 872                 cfs_hash_putref(exp->exp_flock_hash);
 873                 exp->exp_flock_hash = NULL;
 874         }
 875         EXIT;
 876 }
 877 EXPORT_SYMBOL(ldlm_destroy_flock_export);