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