kernel/drivers/staging/lustre/lustre/ldlm/ldlm_lock.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) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2010, 2012, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  *
  36  * lustre/ldlm/ldlm_lock.c
  37  *
  38  * Author: Peter Braam <braam@clusterfs.com>
  39  * Author: Phil Schwan <phil@clusterfs.com>
  40  */
  41
  42 #define DEBUG_SUBSYSTEM S_LDLM
  43
  44 #include "../../include/linux/libcfs/libcfs.h"
  45 #include "../include/lustre_intent.h"
  46 #include "../include/obd_class.h"
  47 #include "ldlm_internal.h"
  48
  49 /* lock types */
  50 char *ldlm_lockname[] = {
  51         [0]             = "--",
  52         [LCK_EX]        = "EX",
  53         [LCK_PW]        = "PW",
  54         [LCK_PR]        = "PR",
  55         [LCK_CW]        = "CW",
  56         [LCK_CR]        = "CR",
  57         [LCK_NL]        = "NL",
  58         [LCK_GROUP]     = "GROUP",
  59         [LCK_COS]       = "COS",
  60 };
  61 EXPORT_SYMBOL(ldlm_lockname);
  62
  63 char *ldlm_typename[] = {
  64         [LDLM_PLAIN]    = "PLN",
  65         [LDLM_EXTENT]   = "EXT",
  66         [LDLM_FLOCK]    = "FLK",
  67         [LDLM_IBITS]    = "IBT",
  68 };
  69 EXPORT_SYMBOL(ldlm_typename);
  70
  71 static ldlm_policy_wire_to_local_t ldlm_policy_wire18_to_local[] = {
  72         [LDLM_PLAIN - LDLM_MIN_TYPE]    = ldlm_plain_policy_wire_to_local,
  73         [LDLM_EXTENT - LDLM_MIN_TYPE]   = ldlm_extent_policy_wire_to_local,
  74         [LDLM_FLOCK - LDLM_MIN_TYPE]    = ldlm_flock_policy_wire18_to_local,
  75         [LDLM_IBITS - LDLM_MIN_TYPE]    = ldlm_ibits_policy_wire_to_local,
  76 };
  77
  78 static ldlm_policy_wire_to_local_t ldlm_policy_wire21_to_local[] = {
  79         [LDLM_PLAIN - LDLM_MIN_TYPE]    = ldlm_plain_policy_wire_to_local,
  80         [LDLM_EXTENT - LDLM_MIN_TYPE]   = ldlm_extent_policy_wire_to_local,
  81         [LDLM_FLOCK - LDLM_MIN_TYPE]    = ldlm_flock_policy_wire21_to_local,
  82         [LDLM_IBITS - LDLM_MIN_TYPE]    = ldlm_ibits_policy_wire_to_local,
  83 };
  84
  85 static ldlm_policy_local_to_wire_t ldlm_policy_local_to_wire[] = {
  86         [LDLM_PLAIN - LDLM_MIN_TYPE]    = ldlm_plain_policy_local_to_wire,
  87         [LDLM_EXTENT - LDLM_MIN_TYPE]   = ldlm_extent_policy_local_to_wire,
  88         [LDLM_FLOCK - LDLM_MIN_TYPE]    = ldlm_flock_policy_local_to_wire,
  89         [LDLM_IBITS - LDLM_MIN_TYPE]    = ldlm_ibits_policy_local_to_wire,
  90 };
  91
  92 /**
  93  * Converts lock policy from local format to on the wire lock_desc format
  94  */
  95 void ldlm_convert_policy_to_wire(ldlm_type_t type,
  96                                  const ldlm_policy_data_t *lpolicy,
  97                                  ldlm_wire_policy_data_t *wpolicy)
  98 {
  99         ldlm_policy_local_to_wire_t convert;
 100
 101         convert = ldlm_policy_local_to_wire[type - LDLM_MIN_TYPE];
 102
 103         convert(lpolicy, wpolicy);
 104 }
 105
 106 /**
 107  * Converts lock policy from on the wire lock_desc format to local format
 108  */
 109 void ldlm_convert_policy_to_local(struct obd_export *exp, ldlm_type_t type,
 110                                   const ldlm_wire_policy_data_t *wpolicy,
 111                                   ldlm_policy_data_t *lpolicy)
 112 {
 113         ldlm_policy_wire_to_local_t convert;
 114         int new_client;
 115
 116         /** some badness for 2.0.0 clients, but 2.0.0 isn't supported */
 117         new_client = (exp_connect_flags(exp) & OBD_CONNECT_FULL20) != 0;
 118         if (new_client)
 119                 convert = ldlm_policy_wire21_to_local[type - LDLM_MIN_TYPE];
 120         else
 121                 convert = ldlm_policy_wire18_to_local[type - LDLM_MIN_TYPE];
 122
 123         convert(wpolicy, lpolicy);
 124 }
 125
 126 char *ldlm_it2str(int it)
 127 {
 128         switch (it) {
 129         case IT_OPEN:
 130                 return "open";
 131         case IT_CREAT:
 132                 return "creat";
 133         case (IT_OPEN | IT_CREAT):
 134                 return "open|creat";
 135         case IT_READDIR:
 136                 return "readdir";
 137         case IT_GETATTR:
 138                 return "getattr";
 139         case IT_LOOKUP:
 140                 return "lookup";
 141         case IT_UNLINK:
 142                 return "unlink";
 143         case IT_GETXATTR:
 144                 return "getxattr";
 145         case IT_LAYOUT:
 146                 return "layout";
 147         default:
 148                 CERROR("Unknown intent %d\n", it);
 149                 return "UNKNOWN";
 150         }
 151 }
 152 EXPORT_SYMBOL(ldlm_it2str);
 153
 154
 155 void ldlm_register_intent(struct ldlm_namespace *ns, ldlm_res_policy arg)
 156 {
 157         ns->ns_policy = arg;
 158 }
 159 EXPORT_SYMBOL(ldlm_register_intent);
 160
 161 /*
 162  * REFCOUNTED LOCK OBJECTS
 163  */
 164
 165
 166 /**
 167  * Get a reference on a lock.
 168  *
 169  * Lock refcounts, during creation:
 170  *   - one special one for allocation, dec'd only once in destroy
 171  *   - one for being a lock that's in-use
 172  *   - one for the addref associated with a new lock
 173  */
 174 struct ldlm_lock *ldlm_lock_get(struct ldlm_lock *lock)
 175 {
 176         atomic_inc(&lock->l_refc);
 177         return lock;
 178 }
 179 EXPORT_SYMBOL(ldlm_lock_get);
 180
 181 /**
 182  * Release lock reference.
 183  *
 184  * Also frees the lock if it was last reference.
 185  */
 186 void ldlm_lock_put(struct ldlm_lock *lock)
 187 {
 188         LASSERT(lock->l_resource != LP_POISON);
 189         LASSERT(atomic_read(&lock->l_refc) > 0);
 190         if (atomic_dec_and_test(&lock->l_refc)) {
 191                 struct ldlm_resource *res;
 192
 193                 LDLM_DEBUG(lock,
 194                            "final lock_put on destroyed lock, freeing it.");
 195
 196                 res = lock->l_resource;
 197                 LASSERT(lock->l_flags & LDLM_FL_DESTROYED);
 198                 LASSERT(list_empty(&lock->l_res_link));
 199                 LASSERT(list_empty(&lock->l_pending_chain));
 200
 201                 lprocfs_counter_decr(ldlm_res_to_ns(res)->ns_stats,
 202                                      LDLM_NSS_LOCKS);
 203                 lu_ref_del(&res->lr_reference, "lock", lock);
 204                 ldlm_resource_putref(res);
 205                 lock->l_resource = NULL;
 206                 if (lock->l_export) {
 207                         class_export_lock_put(lock->l_export, lock);
 208                         lock->l_export = NULL;
 209                 }
 210
 211                 if (lock->l_lvb_data != NULL)
 212                         OBD_FREE(lock->l_lvb_data, lock->l_lvb_len);
 213
 214                 ldlm_interval_free(ldlm_interval_detach(lock));
 215                 lu_ref_fini(&lock->l_reference);
 216                 OBD_FREE_RCU(lock, sizeof(*lock), &lock->l_handle);
 217         }
 218 }
 219 EXPORT_SYMBOL(ldlm_lock_put);
 220
 221 /**
 222  * Removes LDLM lock \a lock from LRU. Assumes LRU is already locked.
 223  */
 224 int ldlm_lock_remove_from_lru_nolock(struct ldlm_lock *lock)
 225 {
 226         int rc = 0;
 227
 228         if (!list_empty(&lock->l_lru)) {
 229                 struct ldlm_namespace *ns = ldlm_lock_to_ns(lock);
 230
 231                 LASSERT(lock->l_resource->lr_type != LDLM_FLOCK);
 232                 list_del_init(&lock->l_lru);
 233                 LASSERT(ns->ns_nr_unused > 0);
 234                 ns->ns_nr_unused--;
 235                 rc = 1;
 236         }
 237         return rc;
 238 }
 239
 240 /**
 241  * Removes LDLM lock \a lock from LRU. Obtains the LRU lock first.
 242  */
 243 int ldlm_lock_remove_from_lru(struct ldlm_lock *lock)
 244 {
 245         struct ldlm_namespace *ns = ldlm_lock_to_ns(lock);
 246         int rc;
 247
 248         if (lock->l_flags & LDLM_FL_NS_SRV) {
 249                 LASSERT(list_empty(&lock->l_lru));
 250                 return 0;
 251         }
 252
 253         spin_lock(&ns->ns_lock);
 254         rc = ldlm_lock_remove_from_lru_nolock(lock);
 255         spin_unlock(&ns->ns_lock);
 256         return rc;
 257 }
 258
 259 /**
 260  * Adds LDLM lock \a lock to namespace LRU. Assumes LRU is already locked.
 261  */
 262 void ldlm_lock_add_to_lru_nolock(struct ldlm_lock *lock)
 263 {
 264         struct ldlm_namespace *ns = ldlm_lock_to_ns(lock);
 265
 266         lock->l_last_used = cfs_time_current();
 267         LASSERT(list_empty(&lock->l_lru));
 268         LASSERT(lock->l_resource->lr_type != LDLM_FLOCK);
 269         list_add_tail(&lock->l_lru, &ns->ns_unused_list);
 270         if (lock->l_flags & LDLM_FL_SKIPPED)
 271                 lock->l_flags &= ~LDLM_FL_SKIPPED;
 272         LASSERT(ns->ns_nr_unused >= 0);
 273         ns->ns_nr_unused++;
 274 }
 275
 276 /**
 277  * Adds LDLM lock \a lock to namespace LRU. Obtains necessary LRU locks
 278  * first.
 279  */
 280 void ldlm_lock_add_to_lru(struct ldlm_lock *lock)
 281 {
 282         struct ldlm_namespace *ns = ldlm_lock_to_ns(lock);
 283
 284         spin_lock(&ns->ns_lock);
 285         ldlm_lock_add_to_lru_nolock(lock);
 286         spin_unlock(&ns->ns_lock);
 287 }
 288
 289 /**
 290  * Moves LDLM lock \a lock that is already in namespace LRU to the tail of
 291  * the LRU. Performs necessary LRU locking
 292  */
 293 void ldlm_lock_touch_in_lru(struct ldlm_lock *lock)
 294 {
 295         struct ldlm_namespace *ns = ldlm_lock_to_ns(lock);
 296
 297         if (lock->l_flags & LDLM_FL_NS_SRV) {
 298                 LASSERT(list_empty(&lock->l_lru));
 299                 return;
 300         }
 301
 302         spin_lock(&ns->ns_lock);
 303         if (!list_empty(&lock->l_lru)) {
 304                 ldlm_lock_remove_from_lru_nolock(lock);
 305                 ldlm_lock_add_to_lru_nolock(lock);
 306         }
 307         spin_unlock(&ns->ns_lock);
 308 }
 309
 310 /**
 311  * Helper to destroy a locked lock.
 312  *
 313  * Used by ldlm_lock_destroy and ldlm_lock_destroy_nolock
 314  * Must be called with l_lock and lr_lock held.
 315  *
 316  * Does not actually free the lock data, but rather marks the lock as
 317  * destroyed by setting l_destroyed field in the lock to 1.  Destroys a
 318  * handle->lock association too, so that the lock can no longer be found
 319  * and removes the lock from LRU list.  Actual lock freeing occurs when
 320  * last lock reference goes away.
 321  *
 322  * Original comment (of some historical value):
 323  * This used to have a 'strict' flag, which recovery would use to mark an
 324  * in-use lock as needing-to-die.  Lest I am ever tempted to put it back, I
 325  * shall explain why it's gone: with the new hash table scheme, once you call
 326  * ldlm_lock_destroy, you can never drop your final references on this lock.
 327  * Because it's not in the hash table anymore.  -phil
 328  */
 329 int ldlm_lock_destroy_internal(struct ldlm_lock *lock)
 330 {
 331         if (lock->l_readers || lock->l_writers) {
 332                 LDLM_ERROR(lock, "lock still has references");
 333                 LBUG();
 334         }
 335
 336         if (!list_empty(&lock->l_res_link)) {
 337                 LDLM_ERROR(lock, "lock still on resource");
 338                 LBUG();
 339         }
 340
 341         if (lock->l_flags & LDLM_FL_DESTROYED) {
 342                 LASSERT(list_empty(&lock->l_lru));
 343                 return 0;
 344         }
 345         lock->l_flags |= LDLM_FL_DESTROYED;
 346
 347         if (lock->l_export && lock->l_export->exp_lock_hash) {
 348                 /* NB: it's safe to call cfs_hash_del() even lock isn't
 349                  * in exp_lock_hash. */
 350                 /* In the function below, .hs_keycmp resolves to
 351                  * ldlm_export_lock_keycmp() */
 352                 /* coverity[overrun-buffer-val] */
 353                 cfs_hash_del(lock->l_export->exp_lock_hash,
 354                              &lock->l_remote_handle, &lock->l_exp_hash);
 355         }
 356
 357         ldlm_lock_remove_from_lru(lock);
 358         class_handle_unhash(&lock->l_handle);
 359
 360 #if 0
 361         /* Wake anyone waiting for this lock */
 362         /* FIXME: I should probably add yet another flag, instead of using
 363          * l_export to only call this on clients */
 364         if (lock->l_export)
 365                 class_export_put(lock->l_export);
 366         lock->l_export = NULL;
 367         if (lock->l_export && lock->l_completion_ast)
 368                 lock->l_completion_ast(lock, 0);
 369 #endif
 370         return 1;
 371 }
 372
 373 /**
 374  * Destroys a LDLM lock \a lock. Performs necessary locking first.
 375  */
 376 void ldlm_lock_destroy(struct ldlm_lock *lock)
 377 {
 378         int first;
 379
 380         lock_res_and_lock(lock);
 381         first = ldlm_lock_destroy_internal(lock);
 382         unlock_res_and_lock(lock);
 383
 384         /* drop reference from hashtable only for first destroy */
 385         if (first) {
 386                 lu_ref_del(&lock->l_reference, "hash", lock);
 387                 LDLM_LOCK_RELEASE(lock);
 388         }
 389 }
 390
 391 /**
 392  * Destroys a LDLM lock \a lock that is already locked.
 393  */
 394 void ldlm_lock_destroy_nolock(struct ldlm_lock *lock)
 395 {
 396         int first;
 397
 398         first = ldlm_lock_destroy_internal(lock);
 399         /* drop reference from hashtable only for first destroy */
 400         if (first) {
 401                 lu_ref_del(&lock->l_reference, "hash", lock);
 402                 LDLM_LOCK_RELEASE(lock);
 403         }
 404 }
 405
 406 /* this is called by portals_handle2object with the handle lock taken */
 407 static void lock_handle_addref(void *lock)
 408 {
 409         LDLM_LOCK_GET((struct ldlm_lock *)lock);
 410 }
 411
 412 static void lock_handle_free(void *lock, int size)
 413 {
 414         LASSERT(size == sizeof(struct ldlm_lock));
 415         OBD_SLAB_FREE(lock, ldlm_lock_slab, size);
 416 }
 417
 418 struct portals_handle_ops lock_handle_ops = {
 419         .hop_addref = lock_handle_addref,
 420         .hop_free   = lock_handle_free,
 421 };
 422
 423 /**
 424  *
 425  * Allocate and initialize new lock structure.
 426  *
 427  * usage: pass in a resource on which you have done ldlm_resource_get
 428  *      new lock will take over the refcount.
 429  * returns: lock with refcount 2 - one for current caller and one for remote
 430  */
 431 static struct ldlm_lock *ldlm_lock_new(struct ldlm_resource *resource)
 432 {
 433         struct ldlm_lock *lock;
 434
 435         if (resource == NULL)
 436                 LBUG();
 437
 438         OBD_SLAB_ALLOC_PTR_GFP(lock, ldlm_lock_slab, GFP_NOFS);
 439         if (lock == NULL)
 440                 return NULL;
 441
 442         spin_lock_init(&lock->l_lock);
 443         lock->l_resource = resource;
 444         lu_ref_add(&resource->lr_reference, "lock", lock);
 445
 446         atomic_set(&lock->l_refc, 2);
 447         INIT_LIST_HEAD(&lock->l_res_link);
 448         INIT_LIST_HEAD(&lock->l_lru);
 449         INIT_LIST_HEAD(&lock->l_pending_chain);
 450         INIT_LIST_HEAD(&lock->l_bl_ast);
 451         INIT_LIST_HEAD(&lock->l_cp_ast);
 452         INIT_LIST_HEAD(&lock->l_rk_ast);
 453         init_waitqueue_head(&lock->l_waitq);
 454         lock->l_blocking_lock = NULL;
 455         INIT_LIST_HEAD(&lock->l_sl_mode);
 456         INIT_LIST_HEAD(&lock->l_sl_policy);
 457         INIT_HLIST_NODE(&lock->l_exp_hash);
 458         INIT_HLIST_NODE(&lock->l_exp_flock_hash);
 459
 460         lprocfs_counter_incr(ldlm_res_to_ns(resource)->ns_stats,
 461                              LDLM_NSS_LOCKS);
 462         INIT_LIST_HEAD(&lock->l_handle.h_link);
 463         class_handle_hash(&lock->l_handle, &lock_handle_ops);
 464
 465         lu_ref_init(&lock->l_reference);
 466         lu_ref_add(&lock->l_reference, "hash", lock);
 467         lock->l_callback_timeout = 0;
 468
 469 #if LUSTRE_TRACKS_LOCK_EXP_REFS
 470         INIT_LIST_HEAD(&lock->l_exp_refs_link);
 471         lock->l_exp_refs_nr = 0;
 472         lock->l_exp_refs_target = NULL;
 473 #endif
 474         INIT_LIST_HEAD(&lock->l_exp_list);
 475
 476         return lock;
 477 }
 478
 479 /**
 480  * Moves LDLM lock \a lock to another resource.
 481  * This is used on client when server returns some other lock than requested
 482  * (typically as a result of intent operation)
 483  */
 484 int ldlm_lock_change_resource(struct ldlm_namespace *ns, struct ldlm_lock *lock,
 485                               const struct ldlm_res_id *new_resid)
 486 {
 487         struct ldlm_resource *oldres = lock->l_resource;
 488         struct ldlm_resource *newres;
 489         int type;
 490
 491         LASSERT(ns_is_client(ns));
 492
 493         lock_res_and_lock(lock);
 494         if (memcmp(new_resid, &lock->l_resource->lr_name,
 495                    sizeof(lock->l_resource->lr_name)) == 0) {
 496                 /* Nothing to do */
 497                 unlock_res_and_lock(lock);
 498                 return 0;
 499         }
 500
 501         LASSERT(new_resid->name[0] != 0);
 502
 503         /* This function assumes that the lock isn't on any lists */
 504         LASSERT(list_empty(&lock->l_res_link));
 505
 506         type = oldres->lr_type;
 507         unlock_res_and_lock(lock);
 508
 509         newres = ldlm_resource_get(ns, NULL, new_resid, type, 1);
 510         if (newres == NULL)
 511                 return -ENOMEM;
 512
 513         lu_ref_add(&newres->lr_reference, "lock", lock);
 514         /*
 515          * To flip the lock from the old to the new resource, lock, oldres and
 516          * newres have to be locked. Resource spin-locks are nested within
 517          * lock->l_lock, and are taken in the memory address order to avoid
 518          * dead-locks.
 519          */
 520         spin_lock(&lock->l_lock);
 521         oldres = lock->l_resource;
 522         if (oldres < newres) {
 523                 lock_res(oldres);
 524                 lock_res_nested(newres, LRT_NEW);
 525         } else {
 526                 lock_res(newres);
 527                 lock_res_nested(oldres, LRT_NEW);
 528         }
 529         LASSERT(memcmp(new_resid, &oldres->lr_name,
 530                        sizeof(oldres->lr_name)) != 0);
 531         lock->l_resource = newres;
 532         unlock_res(oldres);
 533         unlock_res_and_lock(lock);
 534
 535         /* ...and the flowers are still standing! */
 536         lu_ref_del(&oldres->lr_reference, "lock", lock);
 537         ldlm_resource_putref(oldres);
 538
 539         return 0;
 540 }
 541 EXPORT_SYMBOL(ldlm_lock_change_resource);
 542
 543 /** \defgroup ldlm_handles LDLM HANDLES
 544  * Ways to get hold of locks without any addresses.
 545  * @{
 546  */
 547
 548 /**
 549  * Fills in handle for LDLM lock \a lock into supplied \a lockh
 550  * Does not take any references.
 551  */
 552 void ldlm_lock2handle(const struct ldlm_lock *lock, struct lustre_handle *lockh)
 553 {
 554         lockh->cookie = lock->l_handle.h_cookie;
 555 }
 556 EXPORT_SYMBOL(ldlm_lock2handle);
 557
 558 /**
 559  * Obtain a lock reference by handle.
 560  *
 561  * if \a flags: atomically get the lock and set the flags.
 562  *            Return NULL if flag already set
 563  */
 564 struct ldlm_lock *__ldlm_handle2lock(const struct lustre_handle *handle,
 565                                      __u64 flags)
 566 {
 567         struct ldlm_lock *lock;
 568
 569         LASSERT(handle);
 570
 571         lock = class_handle2object(handle->cookie);
 572         if (lock == NULL)
 573                 return NULL;
 574
 575         /* It's unlikely but possible that someone marked the lock as
 576          * destroyed after we did handle2object on it */
 577         if (flags == 0 && ((lock->l_flags & LDLM_FL_DESTROYED) == 0)) {
 578                 lu_ref_add(&lock->l_reference, "handle", current);
 579                 return lock;
 580         }
 581
 582         lock_res_and_lock(lock);
 583
 584         LASSERT(lock->l_resource != NULL);
 585
 586         lu_ref_add_atomic(&lock->l_reference, "handle", current);
 587         if (unlikely(lock->l_flags & LDLM_FL_DESTROYED)) {
 588                 unlock_res_and_lock(lock);
 589                 CDEBUG(D_INFO, "lock already destroyed: lock %p\n", lock);
 590                 LDLM_LOCK_PUT(lock);
 591                 return NULL;
 592         }
 593
 594         if (flags && (lock->l_flags & flags)) {
 595                 unlock_res_and_lock(lock);
 596                 LDLM_LOCK_PUT(lock);
 597                 return NULL;
 598         }
 599
 600         if (flags)
 601                 lock->l_flags |= flags;
 602
 603         unlock_res_and_lock(lock);
 604         return lock;
 605 }
 606 EXPORT_SYMBOL(__ldlm_handle2lock);
 607 /** @} ldlm_handles */
 608
 609 /**
 610  * Fill in "on the wire" representation for given LDLM lock into supplied
 611  * lock descriptor \a desc structure.
 612  */
 613 void ldlm_lock2desc(struct ldlm_lock *lock, struct ldlm_lock_desc *desc)
 614 {
 615         ldlm_res2desc(lock->l_resource, &desc->l_resource);
 616         desc->l_req_mode = lock->l_req_mode;
 617         desc->l_granted_mode = lock->l_granted_mode;
 618         ldlm_convert_policy_to_wire(lock->l_resource->lr_type,
 619                                     &lock->l_policy_data,
 620                                     &desc->l_policy_data);
 621 }
 622 EXPORT_SYMBOL(ldlm_lock2desc);
 623
 624 /**
 625  * Add a lock to list of conflicting locks to send AST to.
 626  *
 627  * Only add if we have not sent a blocking AST to the lock yet.
 628  */
 629 void ldlm_add_bl_work_item(struct ldlm_lock *lock, struct ldlm_lock *new,
 630                            struct list_head *work_list)
 631 {
 632         if ((lock->l_flags & LDLM_FL_AST_SENT) == 0) {
 633                 LDLM_DEBUG(lock, "lock incompatible; sending blocking AST.");
 634                 lock->l_flags |= LDLM_FL_AST_SENT;
 635                 /* If the enqueuing client said so, tell the AST recipient to
 636                  * discard dirty data, rather than writing back. */
 637                 if (new->l_flags & LDLM_FL_AST_DISCARD_DATA)
 638                         lock->l_flags |= LDLM_FL_DISCARD_DATA;
 639                 LASSERT(list_empty(&lock->l_bl_ast));
 640                 list_add(&lock->l_bl_ast, work_list);
 641                 LDLM_LOCK_GET(lock);
 642                 LASSERT(lock->l_blocking_lock == NULL);
 643                 lock->l_blocking_lock = LDLM_LOCK_GET(new);
 644         }
 645 }
 646
 647 /**
 648  * Add a lock to list of just granted locks to send completion AST to.
 649  */
 650 void ldlm_add_cp_work_item(struct ldlm_lock *lock, struct list_head *work_list)
 651 {
 652         if ((lock->l_flags & LDLM_FL_CP_REQD) == 0) {
 653                 lock->l_flags |= LDLM_FL_CP_REQD;
 654                 LDLM_DEBUG(lock, "lock granted; sending completion AST.");
 655                 LASSERT(list_empty(&lock->l_cp_ast));
 656                 list_add(&lock->l_cp_ast, work_list);
 657                 LDLM_LOCK_GET(lock);
 658         }
 659 }
 660
 661 /**
 662  * Aggregator function to add AST work items into a list. Determines
 663  * what sort of an AST work needs to be done and calls the proper
 664  * adding function.
 665  * Must be called with lr_lock held.
 666  */
 667 void ldlm_add_ast_work_item(struct ldlm_lock *lock, struct ldlm_lock *new,
 668                             struct list_head *work_list)
 669 {
 670         check_res_locked(lock->l_resource);
 671         if (new)
 672                 ldlm_add_bl_work_item(lock, new, work_list);
 673         else
 674                 ldlm_add_cp_work_item(lock, work_list);
 675 }
 676
 677 /**
 678  * Add specified reader/writer reference to LDLM lock with handle \a lockh.
 679  * r/w reference type is determined by \a mode
 680  * Calls ldlm_lock_addref_internal.
 681  */
 682 void ldlm_lock_addref(struct lustre_handle *lockh, __u32 mode)
 683 {
 684         struct ldlm_lock *lock;
 685
 686         lock = ldlm_handle2lock(lockh);
 687         LASSERT(lock != NULL);
 688         ldlm_lock_addref_internal(lock, mode);
 689         LDLM_LOCK_PUT(lock);
 690 }
 691 EXPORT_SYMBOL(ldlm_lock_addref);
 692
 693 /**
 694  * Helper function.
 695  * Add specified reader/writer reference to LDLM lock \a lock.
 696  * r/w reference type is determined by \a mode
 697  * Removes lock from LRU if it is there.
 698  * Assumes the LDLM lock is already locked.
 699  */
 700 void ldlm_lock_addref_internal_nolock(struct ldlm_lock *lock, __u32 mode)
 701 {
 702         ldlm_lock_remove_from_lru(lock);
 703         if (mode & (LCK_NL | LCK_CR | LCK_PR)) {
 704                 lock->l_readers++;
 705                 lu_ref_add_atomic(&lock->l_reference, "reader", lock);
 706         }
 707         if (mode & (LCK_EX | LCK_CW | LCK_PW | LCK_GROUP | LCK_COS)) {
 708                 lock->l_writers++;
 709                 lu_ref_add_atomic(&lock->l_reference, "writer", lock);
 710         }
 711         LDLM_LOCK_GET(lock);
 712         lu_ref_add_atomic(&lock->l_reference, "user", lock);
 713         LDLM_DEBUG(lock, "ldlm_lock_addref(%s)", ldlm_lockname[mode]);
 714 }
 715
 716 /**
 717  * Attempts to add reader/writer reference to a lock with handle \a lockh, and
 718  * fails if lock is already LDLM_FL_CBPENDING or destroyed.
 719  *
 720  * \retval 0 success, lock was addref-ed
 721  *
 722  * \retval -EAGAIN lock is being canceled.
 723  */
 724 int ldlm_lock_addref_try(struct lustre_handle *lockh, __u32 mode)
 725 {
 726         struct ldlm_lock *lock;
 727         int            result;
 728
 729         result = -EAGAIN;
 730         lock = ldlm_handle2lock(lockh);
 731         if (lock != NULL) {
 732                 lock_res_and_lock(lock);
 733                 if (lock->l_readers != 0 || lock->l_writers != 0 ||
 734                     !(lock->l_flags & LDLM_FL_CBPENDING)) {
 735                         ldlm_lock_addref_internal_nolock(lock, mode);
 736                         result = 0;
 737                 }
 738                 unlock_res_and_lock(lock);
 739                 LDLM_LOCK_PUT(lock);
 740         }
 741         return result;
 742 }
 743 EXPORT_SYMBOL(ldlm_lock_addref_try);
 744
 745 /**
 746  * Add specified reader/writer reference to LDLM lock \a lock.
 747  * Locks LDLM lock and calls ldlm_lock_addref_internal_nolock to do the work.
 748  * Only called for local locks.
 749  */
 750 void ldlm_lock_addref_internal(struct ldlm_lock *lock, __u32 mode)
 751 {
 752         lock_res_and_lock(lock);
 753         ldlm_lock_addref_internal_nolock(lock, mode);
 754         unlock_res_and_lock(lock);
 755 }
 756
 757 /**
 758  * Removes reader/writer reference for LDLM lock \a lock.
 759  * Assumes LDLM lock is already locked.
 760  * only called in ldlm_flock_destroy and for local locks.
 761  * Does NOT add lock to LRU if no r/w references left to accommodate flock locks
 762  * that cannot be placed in LRU.
 763  */
 764 void ldlm_lock_decref_internal_nolock(struct ldlm_lock *lock, __u32 mode)
 765 {
 766         LDLM_DEBUG(lock, "ldlm_lock_decref(%s)", ldlm_lockname[mode]);
 767         if (mode & (LCK_NL | LCK_CR | LCK_PR)) {
 768                 LASSERT(lock->l_readers > 0);
 769                 lu_ref_del(&lock->l_reference, "reader", lock);
 770                 lock->l_readers--;
 771         }
 772         if (mode & (LCK_EX | LCK_CW | LCK_PW | LCK_GROUP | LCK_COS)) {
 773                 LASSERT(lock->l_writers > 0);
 774                 lu_ref_del(&lock->l_reference, "writer", lock);
 775                 lock->l_writers--;
 776         }
 777
 778         lu_ref_del(&lock->l_reference, "user", lock);
 779         LDLM_LOCK_RELEASE(lock);    /* matches the LDLM_LOCK_GET() in addref */
 780 }
 781
 782 /**
 783  * Removes reader/writer reference for LDLM lock \a lock.
 784  * Locks LDLM lock first.
 785  * If the lock is determined to be client lock on a client and r/w refcount
 786  * drops to zero and the lock is not blocked, the lock is added to LRU lock
 787  * on the namespace.
 788  * For blocked LDLM locks if r/w count drops to zero, blocking_ast is called.
 789  */
 790 void ldlm_lock_decref_internal(struct ldlm_lock *lock, __u32 mode)
 791 {
 792         struct ldlm_namespace *ns;
 793
 794         lock_res_and_lock(lock);
 795
 796         ns = ldlm_lock_to_ns(lock);
 797
 798         ldlm_lock_decref_internal_nolock(lock, mode);
 799
 800         if (lock->l_flags & LDLM_FL_LOCAL &&
 801             !lock->l_readers && !lock->l_writers) {
 802                 /* If this is a local lock on a server namespace and this was
 803                  * the last reference, cancel the lock. */
 804                 CDEBUG(D_INFO, "forcing cancel of local lock\n");
 805                 lock->l_flags |= LDLM_FL_CBPENDING;
 806         }
 807
 808         if (!lock->l_readers && !lock->l_writers &&
 809             (lock->l_flags & LDLM_FL_CBPENDING)) {
 810                 /* If we received a blocked AST and this was the last reference,
 811                  * run the callback. */
 812                 if ((lock->l_flags & LDLM_FL_NS_SRV) && lock->l_export)
 813                         CERROR("FL_CBPENDING set on non-local lock--just a warning\n");
 814
 815                 LDLM_DEBUG(lock, "final decref done on cbpending lock");
 816
 817                 LDLM_LOCK_GET(lock); /* dropped by bl thread */
 818                 ldlm_lock_remove_from_lru(lock);
 819                 unlock_res_and_lock(lock);
 820
 821                 if (lock->l_flags & LDLM_FL_FAIL_LOC)
 822                         OBD_RACE(OBD_FAIL_LDLM_CP_BL_RACE);
 823
 824                 if ((lock->l_flags & LDLM_FL_ATOMIC_CB) ||
 825                     ldlm_bl_to_thread_lock(ns, NULL, lock) != 0)
 826                         ldlm_handle_bl_callback(ns, NULL, lock);
 827         } else if (ns_is_client(ns) &&
 828                    !lock->l_readers && !lock->l_writers &&
 829                    !(lock->l_flags & LDLM_FL_NO_LRU) &&
 830                    !(lock->l_flags & LDLM_FL_BL_AST)) {
 831
 832                 LDLM_DEBUG(lock, "add lock into lru list");
 833
 834                 /* If this is a client-side namespace and this was the last
 835                  * reference, put it on the LRU. */
 836                 ldlm_lock_add_to_lru(lock);
 837                 unlock_res_and_lock(lock);
 838
 839                 if (lock->l_flags & LDLM_FL_FAIL_LOC)
 840                         OBD_RACE(OBD_FAIL_LDLM_CP_BL_RACE);
 841
 842                 /* Call ldlm_cancel_lru() only if EARLY_CANCEL and LRU RESIZE
 843                  * are not supported by the server, otherwise, it is done on
 844                  * enqueue. */
 845                 if (!exp_connect_cancelset(lock->l_conn_export) &&
 846                     !ns_connect_lru_resize(ns))
 847                         ldlm_cancel_lru(ns, 0, LCF_ASYNC, 0);
 848         } else {
 849                 LDLM_DEBUG(lock, "do not add lock into lru list");
 850                 unlock_res_and_lock(lock);
 851         }
 852 }
 853
 854 /**
 855  * Decrease reader/writer refcount for LDLM lock with handle \a lockh
 856  */
 857 void ldlm_lock_decref(struct lustre_handle *lockh, __u32 mode)
 858 {
 859         struct ldlm_lock *lock = __ldlm_handle2lock(lockh, 0);
 860
 861         LASSERTF(lock != NULL, "Non-existing lock: %#llx\n", lockh->cookie);
 862         ldlm_lock_decref_internal(lock, mode);
 863         LDLM_LOCK_PUT(lock);
 864 }
 865 EXPORT_SYMBOL(ldlm_lock_decref);
 866
 867 /**
 868  * Decrease reader/writer refcount for LDLM lock with handle
 869  * \a lockh and mark it for subsequent cancellation once r/w refcount
 870  * drops to zero instead of putting into LRU.
 871  *
 872  * Typical usage is for GROUP locks which we cannot allow to be cached.
 873  */
 874 void ldlm_lock_decref_and_cancel(struct lustre_handle *lockh, __u32 mode)
 875 {
 876         struct ldlm_lock *lock = __ldlm_handle2lock(lockh, 0);
 877
 878         LASSERT(lock != NULL);
 879
 880         LDLM_DEBUG(lock, "ldlm_lock_decref(%s)", ldlm_lockname[mode]);
 881         lock_res_and_lock(lock);
 882         lock->l_flags |= LDLM_FL_CBPENDING;
 883         unlock_res_and_lock(lock);
 884         ldlm_lock_decref_internal(lock, mode);
 885         LDLM_LOCK_PUT(lock);
 886 }
 887 EXPORT_SYMBOL(ldlm_lock_decref_and_cancel);
 888
 889 struct sl_insert_point {
 890         struct list_head *res_link;
 891         struct list_head *mode_link;
 892         struct list_head *policy_link;
 893 };
 894
 895 /**
 896  * Finds a position to insert the new lock into granted lock list.
 897  *
 898  * Used for locks eligible for skiplist optimization.
 899  *
 900  * Parameters:
 901  *      queue [input]:  the granted list where search acts on;
 902  *      req [input]:    the lock whose position to be located;
 903  *      prev [output]:  positions within 3 lists to insert @req to
 904  * Return Value:
 905  *      filled @prev
 906  * NOTE: called by
 907  *  - ldlm_grant_lock_with_skiplist
 908  */
 909 static void search_granted_lock(struct list_head *queue,
 910                                 struct ldlm_lock *req,
 911                                 struct sl_insert_point *prev)
 912 {
 913         struct list_head *tmp;
 914         struct ldlm_lock *lock, *mode_end, *policy_end;
 915
 916         list_for_each(tmp, queue) {
 917                 lock = list_entry(tmp, struct ldlm_lock, l_res_link);
 918
 919                 mode_end = list_entry(lock->l_sl_mode.prev,
 920                                           struct ldlm_lock, l_sl_mode);
 921
 922                 if (lock->l_req_mode != req->l_req_mode) {
 923                         /* jump to last lock of mode group */
 924                         tmp = &mode_end->l_res_link;
 925                         continue;
 926                 }
 927
 928                 /* suitable mode group is found */
 929                 if (lock->l_resource->lr_type == LDLM_PLAIN) {
 930                         /* insert point is last lock of the mode group */
 931                         prev->res_link = &mode_end->l_res_link;
 932                         prev->mode_link = &mode_end->l_sl_mode;
 933                         prev->policy_link = &req->l_sl_policy;
 934                         return;
 935                 } else if (lock->l_resource->lr_type == LDLM_IBITS) {
 936                         for (;;) {
 937                                 policy_end =
 938                                         list_entry(lock->l_sl_policy.prev,
 939                                                        struct ldlm_lock,
 940                                                        l_sl_policy);
 941
 942                                 if (lock->l_policy_data.l_inodebits.bits ==
 943                                     req->l_policy_data.l_inodebits.bits) {
 944                                         /* insert point is last lock of
 945                                          * the policy group */
 946                                         prev->res_link =
 947                                                 &policy_end->l_res_link;
 948                                         prev->mode_link =
 949                                                 &policy_end->l_sl_mode;
 950                                         prev->policy_link =
 951                                                 &policy_end->l_sl_policy;
 952                                         return;
 953                                 }
 954
 955                                 if (policy_end == mode_end)
 956                                         /* done with mode group */
 957                                         break;
 958
 959                                 /* go to next policy group within mode group */
 960                                 tmp = policy_end->l_res_link.next;
 961                                 lock = list_entry(tmp, struct ldlm_lock,
 962                                                       l_res_link);
 963                         }  /* loop over policy groups within the mode group */
 964
 965                         /* insert point is last lock of the mode group,
 966                          * new policy group is started */
 967                         prev->res_link = &mode_end->l_res_link;
 968                         prev->mode_link = &mode_end->l_sl_mode;
 969                         prev->policy_link = &req->l_sl_policy;
 970                         return;
 971                 } else {
 972                         LDLM_ERROR(lock,
 973                                    "is not LDLM_PLAIN or LDLM_IBITS lock");
 974                         LBUG();
 975                 }
 976         }
 977
 978         /* insert point is last lock on the queue,
 979          * new mode group and new policy group are started */
 980         prev->res_link = queue->prev;
 981         prev->mode_link = &req->l_sl_mode;
 982         prev->policy_link = &req->l_sl_policy;
 983 }
 984
 985 /**
 986  * Add a lock into resource granted list after a position described by
 987  * \a prev.
 988  */
 989 static void ldlm_granted_list_add_lock(struct ldlm_lock *lock,
 990                                        struct sl_insert_point *prev)
 991 {
 992         struct ldlm_resource *res = lock->l_resource;
 993
 994         check_res_locked(res);
 995
 996         ldlm_resource_dump(D_INFO, res);
 997         LDLM_DEBUG(lock, "About to add lock:");
 998
 999         if (lock->l_flags & LDLM_FL_DESTROYED) {
1000                 CDEBUG(D_OTHER, "Lock destroyed, not adding to resource\n");
1001                 return;
1002         }
1003
1004         LASSERT(list_empty(&lock->l_res_link));
1005         LASSERT(list_empty(&lock->l_sl_mode));
1006         LASSERT(list_empty(&lock->l_sl_policy));
1007
1008         /*
1009          * lock->link == prev->link means lock is first starting the group.
1010          * Don't re-add to itself to suppress kernel warnings.
1011          */
1012         if (&lock->l_res_link != prev->res_link)
1013                 list_add(&lock->l_res_link, prev->res_link);
1014         if (&lock->l_sl_mode != prev->mode_link)
1015                 list_add(&lock->l_sl_mode, prev->mode_link);
1016         if (&lock->l_sl_policy != prev->policy_link)
1017                 list_add(&lock->l_sl_policy, prev->policy_link);
1018 }
1019
1020 /**
1021  * Add a lock to granted list on a resource maintaining skiplist
1022  * correctness.
1023  */
1024 static void ldlm_grant_lock_with_skiplist(struct ldlm_lock *lock)
1025 {
1026         struct sl_insert_point prev;
1027
1028         LASSERT(lock->l_req_mode == lock->l_granted_mode);
1029
1030         search_granted_lock(&lock->l_resource->lr_granted, lock, &prev);
1031         ldlm_granted_list_add_lock(lock, &prev);
1032 }
1033
1034 /**
1035  * Perform lock granting bookkeeping.
1036  *
1037  * Includes putting the lock into granted list and updating lock mode.
1038  * NOTE: called by
1039  *  - ldlm_lock_enqueue
1040  *  - ldlm_reprocess_queue
1041  *  - ldlm_lock_convert
1042  *
1043  * must be called with lr_lock held
1044  */
1045 void ldlm_grant_lock(struct ldlm_lock *lock, struct list_head *work_list)
1046 {
1047         struct ldlm_resource *res = lock->l_resource;
1048
1049         check_res_locked(res);
1050
1051         lock->l_granted_mode = lock->l_req_mode;
1052         if (res->lr_type == LDLM_PLAIN || res->lr_type == LDLM_IBITS)
1053                 ldlm_grant_lock_with_skiplist(lock);
1054         else if (res->lr_type == LDLM_EXTENT)
1055                 ldlm_extent_add_lock(res, lock);
1056         else
1057                 ldlm_resource_add_lock(res, &res->lr_granted, lock);
1058
1059         if (lock->l_granted_mode < res->lr_most_restr)
1060                 res->lr_most_restr = lock->l_granted_mode;
1061
1062         if (work_list && lock->l_completion_ast != NULL)
1063                 ldlm_add_ast_work_item(lock, NULL, work_list);
1064
1065         ldlm_pool_add(&ldlm_res_to_ns(res)->ns_pool, lock);
1066 }
1067
1068 /**
1069  * Search for a lock with given properties in a queue.
1070  *
1071  * \retval a referenced lock or NULL.  See the flag descriptions below, in the
1072  * comment above ldlm_lock_match
1073  */
1074 static struct ldlm_lock *search_queue(struct list_head *queue,
1075                                       ldlm_mode_t *mode,
1076                                       ldlm_policy_data_t *policy,
1077                                       struct ldlm_lock *old_lock,
1078                                       __u64 flags, int unref)
1079 {
1080         struct ldlm_lock *lock;
1081         struct list_head       *tmp;
1082
1083         list_for_each(tmp, queue) {
1084                 ldlm_mode_t match;
1085
1086                 lock = list_entry(tmp, struct ldlm_lock, l_res_link);
1087
1088                 if (lock == old_lock)
1089                         break;
1090
1091                 /* Check if this lock can be matched.
1092                  * Used by LU-2919(exclusive open) for open lease lock */
1093                 if (ldlm_is_excl(lock))
1094                         continue;
1095
1096                 /* llite sometimes wants to match locks that will be
1097                  * canceled when their users drop, but we allow it to match
1098                  * if it passes in CBPENDING and the lock still has users.
1099                  * this is generally only going to be used by children
1100                  * whose parents already hold a lock so forward progress
1101                  * can still happen. */
1102                 if (lock->l_flags & LDLM_FL_CBPENDING &&
1103                     !(flags & LDLM_FL_CBPENDING))
1104                         continue;
1105                 if (!unref && lock->l_flags & LDLM_FL_CBPENDING &&
1106                     lock->l_readers == 0 && lock->l_writers == 0)
1107                         continue;
1108
1109                 if (!(lock->l_req_mode & *mode))
1110                         continue;
1111                 match = lock->l_req_mode;
1112
1113                 if (lock->l_resource->lr_type == LDLM_EXTENT &&
1114                     (lock->l_policy_data.l_extent.start >
1115                      policy->l_extent.start ||
1116                      lock->l_policy_data.l_extent.end < policy->l_extent.end))
1117                         continue;
1118
1119                 if (unlikely(match == LCK_GROUP) &&
1120                     lock->l_resource->lr_type == LDLM_EXTENT &&
1121                     lock->l_policy_data.l_extent.gid != policy->l_extent.gid)
1122                         continue;
1123
1124                 /* We match if we have existing lock with same or wider set
1125                    of bits. */
1126                 if (lock->l_resource->lr_type == LDLM_IBITS &&
1127                      ((lock->l_policy_data.l_inodebits.bits &
1128                       policy->l_inodebits.bits) !=
1129                       policy->l_inodebits.bits))
1130                         continue;
1131
1132                 if (!unref && (lock->l_flags & LDLM_FL_GONE_MASK))
1133                         continue;
1134
1135                 if ((flags & LDLM_FL_LOCAL_ONLY) &&
1136                     !(lock->l_flags & LDLM_FL_LOCAL))
1137                         continue;
1138
1139                 if (flags & LDLM_FL_TEST_LOCK) {
1140                         LDLM_LOCK_GET(lock);
1141                         ldlm_lock_touch_in_lru(lock);
1142                 } else {
1143                         ldlm_lock_addref_internal_nolock(lock, match);
1144                 }
1145                 *mode = match;
1146                 return lock;
1147         }
1148
1149         return NULL;
1150 }
1151
1152 void ldlm_lock_fail_match_locked(struct ldlm_lock *lock)
1153 {
1154         if ((lock->l_flags & LDLM_FL_FAIL_NOTIFIED) == 0) {
1155                 lock->l_flags |= LDLM_FL_FAIL_NOTIFIED;
1156                 wake_up_all(&lock->l_waitq);
1157         }
1158 }
1159 EXPORT_SYMBOL(ldlm_lock_fail_match_locked);
1160
1161 void ldlm_lock_fail_match(struct ldlm_lock *lock)
1162 {
1163         lock_res_and_lock(lock);
1164         ldlm_lock_fail_match_locked(lock);
1165         unlock_res_and_lock(lock);
1166 }
1167 EXPORT_SYMBOL(ldlm_lock_fail_match);
1168
1169 /**
1170  * Mark lock as "matchable" by OST.
1171  *
1172  * Used to prevent certain races in LOV/OSC where the lock is granted, but LVB
1173  * is not yet valid.
1174  * Assumes LDLM lock is already locked.
1175  */
1176 void ldlm_lock_allow_match_locked(struct ldlm_lock *lock)
1177 {
1178         lock->l_flags |= LDLM_FL_LVB_READY;
1179         wake_up_all(&lock->l_waitq);
1180 }
1181 EXPORT_SYMBOL(ldlm_lock_allow_match_locked);
1182
1183 /**
1184  * Mark lock as "matchable" by OST.
1185  * Locks the lock and then \see ldlm_lock_allow_match_locked
1186  */
1187 void ldlm_lock_allow_match(struct ldlm_lock *lock)
1188 {
1189         lock_res_and_lock(lock);
1190         ldlm_lock_allow_match_locked(lock);
1191         unlock_res_and_lock(lock);
1192 }
1193 EXPORT_SYMBOL(ldlm_lock_allow_match);
1194
1195 /**
1196  * Attempt to find a lock with specified properties.
1197  *
1198  * Typically returns a reference to matched lock unless LDLM_FL_TEST_LOCK is
1199  * set in \a flags
1200  *
1201  * Can be called in two ways:
1202  *
1203  * If 'ns' is NULL, then lockh describes an existing lock that we want to look
1204  * for a duplicate of.
1205  *
1206  * Otherwise, all of the fields must be filled in, to match against.
1207  *
1208  * If 'flags' contains LDLM_FL_LOCAL_ONLY, then only match local locks on the
1209  *     server (ie, connh is NULL)
1210  * If 'flags' contains LDLM_FL_BLOCK_GRANTED, then only locks on the granted
1211  *     list will be considered
1212  * If 'flags' contains LDLM_FL_CBPENDING, then locks that have been marked
1213  *     to be canceled can still be matched as long as they still have reader
1214  *     or writer referneces
1215  * If 'flags' contains LDLM_FL_TEST_LOCK, then don't actually reference a lock,
1216  *     just tell us if we would have matched.
1217  *
1218  * \retval 1 if it finds an already-existing lock that is compatible; in this
1219  * case, lockh is filled in with a addref()ed lock
1220  *
1221  * We also check security context, and if that fails we simply return 0 (to
1222  * keep caller code unchanged), the context failure will be discovered by
1223  * caller sometime later.
1224  */
1225 ldlm_mode_t ldlm_lock_match(struct ldlm_namespace *ns, __u64 flags,
1226                             const struct ldlm_res_id *res_id, ldlm_type_t type,
1227                             ldlm_policy_data_t *policy, ldlm_mode_t mode,
1228                             struct lustre_handle *lockh, int unref)
1229 {
1230         struct ldlm_resource *res;
1231         struct ldlm_lock *lock, *old_lock = NULL;
1232         int rc = 0;
1233
1234         if (ns == NULL) {
1235                 old_lock = ldlm_handle2lock(lockh);
1236                 LASSERT(old_lock);
1237
1238                 ns = ldlm_lock_to_ns(old_lock);
1239                 res_id = &old_lock->l_resource->lr_name;
1240                 type = old_lock->l_resource->lr_type;
1241                 mode = old_lock->l_req_mode;
1242         }
1243
1244         res = ldlm_resource_get(ns, NULL, res_id, type, 0);
1245         if (res == NULL) {
1246                 LASSERT(old_lock == NULL);
1247                 return 0;
1248         }
1249
1250         LDLM_RESOURCE_ADDREF(res);
1251         lock_res(res);
1252
1253         lock = search_queue(&res->lr_granted, &mode, policy, old_lock,
1254                             flags, unref);
1255         if (lock != NULL) {
1256                 rc = 1;
1257                 goto out;
1258         }
1259         if (flags & LDLM_FL_BLOCK_GRANTED) {
1260                 rc = 0;
1261                 goto out;
1262         }
1263         lock = search_queue(&res->lr_converting, &mode, policy, old_lock,
1264                             flags, unref);
1265         if (lock != NULL) {
1266                 rc = 1;
1267                 goto out;
1268         }
1269         lock = search_queue(&res->lr_waiting, &mode, policy, old_lock,
1270                             flags, unref);
1271         if (lock != NULL) {
1272                 rc = 1;
1273                 goto out;
1274         }
1275
1276  out:
1277         unlock_res(res);
1278         LDLM_RESOURCE_DELREF(res);
1279         ldlm_resource_putref(res);
1280
1281         if (lock) {
1282                 ldlm_lock2handle(lock, lockh);
1283                 if ((flags & LDLM_FL_LVB_READY) &&
1284                     (!(lock->l_flags & LDLM_FL_LVB_READY))) {
1285                         __u64 wait_flags = LDLM_FL_LVB_READY |
1286                                 LDLM_FL_DESTROYED | LDLM_FL_FAIL_NOTIFIED;
1287                         struct l_wait_info lwi;
1288
1289                         if (lock->l_completion_ast) {
1290                                 int err = lock->l_completion_ast(lock,
1291                                                           LDLM_FL_WAIT_NOREPROC,
1292                                                                  NULL);
1293                                 if (err) {
1294                                         if (flags & LDLM_FL_TEST_LOCK)
1295                                                 LDLM_LOCK_RELEASE(lock);
1296                                         else
1297                                                 ldlm_lock_decref_internal(lock,
1298                                                                           mode);
1299                                         rc = 0;
1300                                         goto out2;
1301                                 }
1302                         }
1303
1304                         lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(obd_timeout),
1305                                                NULL, LWI_ON_SIGNAL_NOOP, NULL);
1306
1307                         /* XXX FIXME see comment on CAN_MATCH in lustre_dlm.h */
1308                         l_wait_event(lock->l_waitq,
1309                                      lock->l_flags & wait_flags,
1310                                      &lwi);
1311                         if (!(lock->l_flags & LDLM_FL_LVB_READY)) {
1312                                 if (flags & LDLM_FL_TEST_LOCK)
1313                                         LDLM_LOCK_RELEASE(lock);
1314                                 else
1315                                         ldlm_lock_decref_internal(lock, mode);
1316                                 rc = 0;
1317                         }
1318                 }
1319         }
1320  out2:
1321         if (rc) {
1322                 LDLM_DEBUG(lock, "matched (%llu %llu)",
1323                            (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1324                                 res_id->name[2] : policy->l_extent.start,
1325                            (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1326                                 res_id->name[3] : policy->l_extent.end);
1327
1328                 /* check user's security context */
1329                 if (lock->l_conn_export &&
1330                     sptlrpc_import_check_ctx(
1331                                 class_exp2cliimp(lock->l_conn_export))) {
1332                         if (!(flags & LDLM_FL_TEST_LOCK))
1333                                 ldlm_lock_decref_internal(lock, mode);
1334                         rc = 0;
1335                 }
1336
1337                 if (flags & LDLM_FL_TEST_LOCK)
1338                         LDLM_LOCK_RELEASE(lock);
1339
1340         } else if (!(flags & LDLM_FL_TEST_LOCK)) {/*less verbose for test-only*/
1341                 LDLM_DEBUG_NOLOCK("not matched ns %p type %u mode %u res %llu/%llu (%llu %llu)",
1342                                   ns, type, mode, res_id->name[0],
1343                                   res_id->name[1],
1344                                   (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1345                                         res_id->name[2] : policy->l_extent.start,
1346                                   (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1347                                         res_id->name[3] : policy->l_extent.end);
1348         }
1349         if (old_lock)
1350                 LDLM_LOCK_PUT(old_lock);
1351
1352         return rc ? mode : 0;
1353 }
1354 EXPORT_SYMBOL(ldlm_lock_match);
1355
1356 ldlm_mode_t ldlm_revalidate_lock_handle(struct lustre_handle *lockh,
1357                                         __u64 *bits)
1358 {
1359         struct ldlm_lock *lock;
1360         ldlm_mode_t mode = 0;
1361
1362         lock = ldlm_handle2lock(lockh);
1363         if (lock != NULL) {
1364                 lock_res_and_lock(lock);
1365                 if (lock->l_flags & LDLM_FL_GONE_MASK)
1366                         goto out;
1367
1368                 if (lock->l_flags & LDLM_FL_CBPENDING &&
1369                     lock->l_readers == 0 && lock->l_writers == 0)
1370                         goto out;
1371
1372                 if (bits)
1373                         *bits = lock->l_policy_data.l_inodebits.bits;
1374                 mode = lock->l_granted_mode;
1375                 ldlm_lock_addref_internal_nolock(lock, mode);
1376         }
1377
1378 out:
1379         if (lock != NULL) {
1380                 unlock_res_and_lock(lock);
1381                 LDLM_LOCK_PUT(lock);
1382         }
1383         return mode;
1384 }
1385 EXPORT_SYMBOL(ldlm_revalidate_lock_handle);
1386
1387 /** The caller must guarantee that the buffer is large enough. */
1388 int ldlm_fill_lvb(struct ldlm_lock *lock, struct req_capsule *pill,
1389                   enum req_location loc, void *data, int size)
1390 {
1391         void *lvb;
1392
1393         LASSERT(data != NULL);
1394         LASSERT(size >= 0);
1395
1396         switch (lock->l_lvb_type) {
1397         case LVB_T_OST:
1398                 if (size == sizeof(struct ost_lvb)) {
1399                         if (loc == RCL_CLIENT)
1400                                 lvb = req_capsule_client_swab_get(pill,
1401                                                 &RMF_DLM_LVB,
1402                                                 lustre_swab_ost_lvb);
1403                         else
1404                                 lvb = req_capsule_server_swab_get(pill,
1405                                                 &RMF_DLM_LVB,
1406                                                 lustre_swab_ost_lvb);
1407                         if (unlikely(lvb == NULL)) {
1408                                 LDLM_ERROR(lock, "no LVB");
1409                                 return -EPROTO;
1410                         }
1411
1412                         memcpy(data, lvb, size);
1413                 } else if (size == sizeof(struct ost_lvb_v1)) {
1414                         struct ost_lvb *olvb = data;
1415
1416                         if (loc == RCL_CLIENT)
1417                                 lvb = req_capsule_client_swab_get(pill,
1418                                                 &RMF_DLM_LVB,
1419                                                 lustre_swab_ost_lvb_v1);
1420                         else
1421                                 lvb = req_capsule_server_sized_swab_get(pill,
1422                                                 &RMF_DLM_LVB, size,
1423                                                 lustre_swab_ost_lvb_v1);
1424                         if (unlikely(lvb == NULL)) {
1425                                 LDLM_ERROR(lock, "no LVB");
1426                                 return -EPROTO;
1427                         }
1428
1429                         memcpy(data, lvb, size);
1430                         olvb->lvb_mtime_ns = 0;
1431                         olvb->lvb_atime_ns = 0;
1432                         olvb->lvb_ctime_ns = 0;
1433                 } else {
1434                         LDLM_ERROR(lock, "Replied unexpected ost LVB size %d",
1435                                    size);
1436                         return -EINVAL;
1437                 }
1438                 break;
1439         case LVB_T_LQUOTA:
1440                 if (size == sizeof(struct lquota_lvb)) {
1441                         if (loc == RCL_CLIENT)
1442                                 lvb = req_capsule_client_swab_get(pill,
1443                                                 &RMF_DLM_LVB,
1444                                                 lustre_swab_lquota_lvb);
1445                         else
1446                                 lvb = req_capsule_server_swab_get(pill,
1447                                                 &RMF_DLM_LVB,
1448                                                 lustre_swab_lquota_lvb);
1449                         if (unlikely(lvb == NULL)) {
1450                                 LDLM_ERROR(lock, "no LVB");
1451                                 return -EPROTO;
1452                         }
1453
1454                         memcpy(data, lvb, size);
1455                 } else {
1456                         LDLM_ERROR(lock,
1457                                    "Replied unexpected lquota LVB size %d",
1458                                    size);
1459                         return -EINVAL;
1460                 }
1461                 break;
1462         case LVB_T_LAYOUT:
1463                 if (size == 0)
1464                         break;
1465
1466                 if (loc == RCL_CLIENT)
1467                         lvb = req_capsule_client_get(pill, &RMF_DLM_LVB);
1468                 else
1469                         lvb = req_capsule_server_get(pill, &RMF_DLM_LVB);
1470                 if (unlikely(lvb == NULL)) {
1471                         LDLM_ERROR(lock, "no LVB");
1472                         return -EPROTO;
1473                 }
1474
1475                 memcpy(data, lvb, size);
1476                 break;
1477         default:
1478                 LDLM_ERROR(lock, "Unknown LVB type: %d\n", lock->l_lvb_type);
1479                 dump_stack();
1480                 return -EINVAL;
1481         }
1482
1483         return 0;
1484 }
1485
1486 /**
1487  * Create and fill in new LDLM lock with specified properties.
1488  * Returns a referenced lock
1489  */
1490 struct ldlm_lock *ldlm_lock_create(struct ldlm_namespace *ns,
1491                                    const struct ldlm_res_id *res_id,
1492                                    ldlm_type_t type,
1493                                    ldlm_mode_t mode,
1494                                    const struct ldlm_callback_suite *cbs,
1495                                    void *data, __u32 lvb_len,
1496                                    enum lvb_type lvb_type)
1497 {
1498         struct ldlm_lock *lock;
1499         struct ldlm_resource *res;
1500
1501         res = ldlm_resource_get(ns, NULL, res_id, type, 1);
1502         if (res == NULL)
1503                 return NULL;
1504
1505         lock = ldlm_lock_new(res);
1506
1507         if (lock == NULL)
1508                 return NULL;
1509
1510         lock->l_req_mode = mode;
1511         lock->l_ast_data = data;
1512         lock->l_pid = current_pid();
1513         if (ns_is_server(ns))
1514                 lock->l_flags |= LDLM_FL_NS_SRV;
1515         if (cbs) {
1516                 lock->l_blocking_ast = cbs->lcs_blocking;
1517                 lock->l_completion_ast = cbs->lcs_completion;
1518                 lock->l_glimpse_ast = cbs->lcs_glimpse;
1519         }
1520
1521         lock->l_tree_node = NULL;
1522         /* if this is the extent lock, allocate the interval tree node */
1523         if (type == LDLM_EXTENT) {
1524                 if (ldlm_interval_alloc(lock) == NULL)
1525                         goto out;
1526         }
1527
1528         if (lvb_len) {
1529                 lock->l_lvb_len = lvb_len;
1530                 OBD_ALLOC(lock->l_lvb_data, lvb_len);
1531                 if (lock->l_lvb_data == NULL)
1532                         goto out;
1533         }
1534
1535         lock->l_lvb_type = lvb_type;
1536         if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_NEW_LOCK))
1537                 goto out;
1538
1539         return lock;
1540
1541 out:
1542         ldlm_lock_destroy(lock);
1543         LDLM_LOCK_RELEASE(lock);
1544         return NULL;
1545 }
1546
1547 /**
1548  * Enqueue (request) a lock.
1549  *
1550  * Does not block. As a result of enqueue the lock would be put
1551  * into granted or waiting list.
1552  *
1553  * If namespace has intent policy sent and the lock has LDLM_FL_HAS_INTENT flag
1554  * set, skip all the enqueueing and delegate lock processing to intent policy
1555  * function.
1556  */
1557 ldlm_error_t ldlm_lock_enqueue(struct ldlm_namespace *ns,
1558                                struct ldlm_lock **lockp,
1559                                void *cookie, __u64 *flags)
1560 {
1561         struct ldlm_lock *lock = *lockp;
1562         struct ldlm_resource *res = lock->l_resource;
1563         int local = ns_is_client(ldlm_res_to_ns(res));
1564         ldlm_error_t rc = ELDLM_OK;
1565         struct ldlm_interval *node = NULL;
1566
1567         lock->l_last_activity = get_seconds();
1568         /* policies are not executed on the client or during replay */
1569         if ((*flags & (LDLM_FL_HAS_INTENT|LDLM_FL_REPLAY)) == LDLM_FL_HAS_INTENT
1570             && !local && ns->ns_policy) {
1571                 rc = ns->ns_policy(ns, lockp, cookie, lock->l_req_mode, *flags,
1572                                    NULL);
1573                 if (rc == ELDLM_LOCK_REPLACED) {
1574                         /* The lock that was returned has already been granted,
1575                          * and placed into lockp.  If it's not the same as the
1576                          * one we passed in, then destroy the old one and our
1577                          * work here is done. */
1578                         if (lock != *lockp) {
1579                                 ldlm_lock_destroy(lock);
1580                                 LDLM_LOCK_RELEASE(lock);
1581                         }
1582                         *flags |= LDLM_FL_LOCK_CHANGED;
1583                         return 0;
1584                 } else if (rc != ELDLM_OK ||
1585                            (rc == ELDLM_OK && (*flags & LDLM_FL_INTENT_ONLY))) {
1586                         ldlm_lock_destroy(lock);
1587                         return rc;
1588                 }
1589         }
1590
1591         /* For a replaying lock, it might be already in granted list. So
1592          * unlinking the lock will cause the interval node to be freed, we
1593          * have to allocate the interval node early otherwise we can't regrant
1594          * this lock in the future. - jay */
1595         if (!local && (*flags & LDLM_FL_REPLAY) && res->lr_type == LDLM_EXTENT)
1596                 OBD_SLAB_ALLOC_PTR_GFP(node, ldlm_interval_slab, GFP_NOFS);
1597
1598         lock_res_and_lock(lock);
1599         if (local && lock->l_req_mode == lock->l_granted_mode) {
1600                 /* The server returned a blocked lock, but it was granted
1601                  * before we got a chance to actually enqueue it.  We don't
1602                  * need to do anything else. */
1603                 *flags &= ~(LDLM_FL_BLOCK_GRANTED |
1604                             LDLM_FL_BLOCK_CONV | LDLM_FL_BLOCK_WAIT);
1605                 goto out;
1606         }
1607
1608         ldlm_resource_unlink_lock(lock);
1609         if (res->lr_type == LDLM_EXTENT && lock->l_tree_node == NULL) {
1610                 if (node == NULL) {
1611                         ldlm_lock_destroy_nolock(lock);
1612                         rc = -ENOMEM;
1613                         goto out;
1614                 }
1615
1616                 INIT_LIST_HEAD(&node->li_group);
1617                 ldlm_interval_attach(node, lock);
1618                 node = NULL;
1619         }
1620
1621         /* Some flags from the enqueue want to make it into the AST, via the
1622          * lock's l_flags. */
1623         lock->l_flags |= *flags & LDLM_FL_AST_DISCARD_DATA;
1624
1625         /* This distinction between local lock trees is very important; a client
1626          * namespace only has information about locks taken by that client, and
1627          * thus doesn't have enough information to decide for itself if it can
1628          * be granted (below).  In this case, we do exactly what the server
1629          * tells us to do, as dictated by the 'flags'.
1630          *
1631          * We do exactly the same thing during recovery, when the server is
1632          * more or less trusting the clients not to lie.
1633          *
1634          * FIXME (bug 268): Detect obvious lies by checking compatibility in
1635          * granted/converting queues. */
1636         if (local) {
1637                 if (*flags & LDLM_FL_BLOCK_CONV)
1638                         ldlm_resource_add_lock(res, &res->lr_converting, lock);
1639                 else if (*flags & (LDLM_FL_BLOCK_WAIT | LDLM_FL_BLOCK_GRANTED))
1640                         ldlm_resource_add_lock(res, &res->lr_waiting, lock);
1641                 else
1642                         ldlm_grant_lock(lock, NULL);
1643                 goto out;
1644         } else {
1645                 CERROR("This is client-side-only module, cannot handle LDLM_NAMESPACE_SERVER resource type lock.\n");
1646                 LBUG();
1647         }
1648
1649 out:
1650         unlock_res_and_lock(lock);
1651         if (node)
1652                 OBD_SLAB_FREE(node, ldlm_interval_slab, sizeof(*node));
1653         return rc;
1654 }
1655
1656
1657 /**
1658  * Process a call to blocking AST callback for a lock in ast_work list
1659  */
1660 static int
1661 ldlm_work_bl_ast_lock(struct ptlrpc_request_set *rqset, void *opaq)
1662 {
1663         struct ldlm_cb_set_arg *arg = opaq;
1664         struct ldlm_lock_desc   d;
1665         int                  rc;
1666         struct ldlm_lock       *lock;
1667
1668         if (list_empty(arg->list))
1669                 return -ENOENT;
1670
1671         lock = list_entry(arg->list->next, struct ldlm_lock, l_bl_ast);
1672
1673         /* nobody should touch l_bl_ast */
1674         lock_res_and_lock(lock);
1675         list_del_init(&lock->l_bl_ast);
1676
1677         LASSERT(lock->l_flags & LDLM_FL_AST_SENT);
1678         LASSERT(lock->l_bl_ast_run == 0);
1679         LASSERT(lock->l_blocking_lock);
1680         lock->l_bl_ast_run++;
1681         unlock_res_and_lock(lock);
1682
1683         ldlm_lock2desc(lock->l_blocking_lock, &d);
1684
1685         rc = lock->l_blocking_ast(lock, &d, (void *)arg, LDLM_CB_BLOCKING);
1686         LDLM_LOCK_RELEASE(lock->l_blocking_lock);
1687         lock->l_blocking_lock = NULL;
1688         LDLM_LOCK_RELEASE(lock);
1689
1690         return rc;
1691 }
1692
1693 /**
1694  * Process a call to completion AST callback for a lock in ast_work list
1695  */
1696 static int
1697 ldlm_work_cp_ast_lock(struct ptlrpc_request_set *rqset, void *opaq)
1698 {
1699         struct ldlm_cb_set_arg  *arg = opaq;
1700         int                   rc = 0;
1701         struct ldlm_lock        *lock;
1702         ldlm_completion_callback completion_callback;
1703
1704         if (list_empty(arg->list))
1705                 return -ENOENT;
1706
1707         lock = list_entry(arg->list->next, struct ldlm_lock, l_cp_ast);
1708
1709         /* It's possible to receive a completion AST before we've set
1710          * the l_completion_ast pointer: either because the AST arrived
1711          * before the reply, or simply because there's a small race
1712          * window between receiving the reply and finishing the local
1713          * enqueue. (bug 842)
1714          *
1715          * This can't happen with the blocking_ast, however, because we
1716          * will never call the local blocking_ast until we drop our
1717          * reader/writer reference, which we won't do until we get the
1718          * reply and finish enqueueing. */
1719
1720         /* nobody should touch l_cp_ast */
1721         lock_res_and_lock(lock);
1722         list_del_init(&lock->l_cp_ast);
1723         LASSERT(lock->l_flags & LDLM_FL_CP_REQD);
1724         /* save l_completion_ast since it can be changed by
1725          * mds_intent_policy(), see bug 14225 */
1726         completion_callback = lock->l_completion_ast;
1727         lock->l_flags &= ~LDLM_FL_CP_REQD;
1728         unlock_res_and_lock(lock);
1729
1730         if (completion_callback != NULL)
1731                 rc = completion_callback(lock, 0, (void *)arg);
1732         LDLM_LOCK_RELEASE(lock);
1733
1734         return rc;
1735 }
1736
1737 /**
1738  * Process a call to revocation AST callback for a lock in ast_work list
1739  */
1740 static int
1741 ldlm_work_revoke_ast_lock(struct ptlrpc_request_set *rqset, void *opaq)
1742 {
1743         struct ldlm_cb_set_arg *arg = opaq;
1744         struct ldlm_lock_desc   desc;
1745         int                  rc;
1746         struct ldlm_lock       *lock;
1747
1748         if (list_empty(arg->list))
1749                 return -ENOENT;
1750
1751         lock = list_entry(arg->list->next, struct ldlm_lock, l_rk_ast);
1752         list_del_init(&lock->l_rk_ast);
1753
1754         /* the desc just pretend to exclusive */
1755         ldlm_lock2desc(lock, &desc);
1756         desc.l_req_mode = LCK_EX;
1757         desc.l_granted_mode = 0;
1758
1759         rc = lock->l_blocking_ast(lock, &desc, (void *)arg, LDLM_CB_BLOCKING);
1760         LDLM_LOCK_RELEASE(lock);
1761
1762         return rc;
1763 }
1764
1765 /**
1766  * Process a call to glimpse AST callback for a lock in ast_work list
1767  */
1768 int ldlm_work_gl_ast_lock(struct ptlrpc_request_set *rqset, void *opaq)
1769 {
1770         struct ldlm_cb_set_arg          *arg = opaq;
1771         struct ldlm_glimpse_work        *gl_work;
1772         struct ldlm_lock                *lock;
1773         int                              rc = 0;
1774
1775         if (list_empty(arg->list))
1776                 return -ENOENT;
1777
1778         gl_work = list_entry(arg->list->next, struct ldlm_glimpse_work,
1779                                  gl_list);
1780         list_del_init(&gl_work->gl_list);
1781
1782         lock = gl_work->gl_lock;
1783
1784         /* transfer the glimpse descriptor to ldlm_cb_set_arg */
1785         arg->gl_desc = gl_work->gl_desc;
1786
1787         /* invoke the actual glimpse callback */
1788         if (lock->l_glimpse_ast(lock, (void *)arg) == 0)
1789                 rc = 1;
1790
1791         LDLM_LOCK_RELEASE(lock);
1792
1793         if ((gl_work->gl_flags & LDLM_GL_WORK_NOFREE) == 0)
1794                 OBD_FREE_PTR(gl_work);
1795
1796         return rc;
1797 }
1798
1799 /**
1800  * Process list of locks in need of ASTs being sent.
1801  *
1802  * Used on server to send multiple ASTs together instead of sending one by
1803  * one.
1804  */
1805 int ldlm_run_ast_work(struct ldlm_namespace *ns, struct list_head *rpc_list,
1806                       enum ldlm_desc_ast_t ast_type)
1807 {
1808         struct ldlm_cb_set_arg *arg;
1809         set_producer_func       work_ast_lock;
1810         int                  rc;
1811
1812         if (list_empty(rpc_list))
1813                 return 0;
1814
1815         OBD_ALLOC_PTR(arg);
1816         if (arg == NULL)
1817                 return -ENOMEM;
1818
1819         atomic_set(&arg->restart, 0);
1820         arg->list = rpc_list;
1821
1822         switch (ast_type) {
1823         case LDLM_WORK_BL_AST:
1824                 arg->type = LDLM_BL_CALLBACK;
1825                 work_ast_lock = ldlm_work_bl_ast_lock;
1826                 break;
1827         case LDLM_WORK_CP_AST:
1828                 arg->type = LDLM_CP_CALLBACK;
1829                 work_ast_lock = ldlm_work_cp_ast_lock;
1830                 break;
1831         case LDLM_WORK_REVOKE_AST:
1832                 arg->type = LDLM_BL_CALLBACK;
1833                 work_ast_lock = ldlm_work_revoke_ast_lock;
1834                 break;
1835         case LDLM_WORK_GL_AST:
1836                 arg->type = LDLM_GL_CALLBACK;
1837                 work_ast_lock = ldlm_work_gl_ast_lock;
1838                 break;
1839         default:
1840                 LBUG();
1841         }
1842
1843         /* We create a ptlrpc request set with flow control extension.
1844          * This request set will use the work_ast_lock function to produce new
1845          * requests and will send a new request each time one completes in order
1846          * to keep the number of requests in flight to ns_max_parallel_ast */
1847         arg->set = ptlrpc_prep_fcset(ns->ns_max_parallel_ast ? : UINT_MAX,
1848                                      work_ast_lock, arg);
1849         if (arg->set == NULL) {
1850                 rc = -ENOMEM;
1851                 goto out;
1852         }
1853
1854         ptlrpc_set_wait(arg->set);
1855         ptlrpc_set_destroy(arg->set);
1856
1857         rc = atomic_read(&arg->restart) ? -ERESTART : 0;
1858         goto out;
1859 out:
1860         OBD_FREE_PTR(arg);
1861         return rc;
1862 }
1863
1864 static int reprocess_one_queue(struct ldlm_resource *res, void *closure)
1865 {
1866         ldlm_reprocess_all(res);
1867         return LDLM_ITER_CONTINUE;
1868 }
1869
1870 static int ldlm_reprocess_res(struct cfs_hash *hs, struct cfs_hash_bd *bd,
1871                               struct hlist_node *hnode, void *arg)
1872 {
1873         struct ldlm_resource *res = cfs_hash_object(hs, hnode);
1874         int    rc;
1875
1876         rc = reprocess_one_queue(res, arg);
1877
1878         return rc == LDLM_ITER_STOP;
1879 }
1880
1881 /**
1882  * Iterate through all resources on a namespace attempting to grant waiting
1883  * locks.
1884  */
1885 void ldlm_reprocess_all_ns(struct ldlm_namespace *ns)
1886 {
1887         if (ns != NULL) {
1888                 cfs_hash_for_each_nolock(ns->ns_rs_hash,
1889                                          ldlm_reprocess_res, NULL);
1890         }
1891 }
1892 EXPORT_SYMBOL(ldlm_reprocess_all_ns);
1893
1894 /**
1895  * Try to grant all waiting locks on a resource.
1896  *
1897  * Calls ldlm_reprocess_queue on converting and waiting queues.
1898  *
1899  * Typically called after some resource locks are cancelled to see
1900  * if anything could be granted as a result of the cancellation.
1901  */
1902 void ldlm_reprocess_all(struct ldlm_resource *res)
1903 {
1904         LIST_HEAD(rpc_list);
1905
1906         if (!ns_is_client(ldlm_res_to_ns(res))) {
1907                 CERROR("This is client-side-only module, cannot handle LDLM_NAMESPACE_SERVER resource type lock.\n");
1908                 LBUG();
1909         }
1910 }
1911
1912 /**
1913  * Helper function to call blocking AST for LDLM lock \a lock in a
1914  * "cancelling" mode.
1915  */
1916 void ldlm_cancel_callback(struct ldlm_lock *lock)
1917 {
1918         check_res_locked(lock->l_resource);
1919         if (!(lock->l_flags & LDLM_FL_CANCEL)) {
1920                 lock->l_flags |= LDLM_FL_CANCEL;
1921                 if (lock->l_blocking_ast) {
1922                         unlock_res_and_lock(lock);
1923                         lock->l_blocking_ast(lock, NULL, lock->l_ast_data,
1924                                              LDLM_CB_CANCELING);
1925                         lock_res_and_lock(lock);
1926                 } else {
1927                         LDLM_DEBUG(lock, "no blocking ast");
1928                 }
1929         }
1930         lock->l_flags |= LDLM_FL_BL_DONE;
1931 }
1932
1933 /**
1934  * Remove skiplist-enabled LDLM lock \a req from granted list
1935  */
1936 void ldlm_unlink_lock_skiplist(struct ldlm_lock *req)
1937 {
1938         if (req->l_resource->lr_type != LDLM_PLAIN &&
1939             req->l_resource->lr_type != LDLM_IBITS)
1940                 return;
1941
1942         list_del_init(&req->l_sl_policy);
1943         list_del_init(&req->l_sl_mode);
1944 }
1945
1946 /**
1947  * Attempts to cancel LDLM lock \a lock that has no reader/writer references.
1948  */
1949 void ldlm_lock_cancel(struct ldlm_lock *lock)
1950 {
1951         struct ldlm_resource *res;
1952         struct ldlm_namespace *ns;
1953
1954         lock_res_and_lock(lock);
1955
1956         res = lock->l_resource;
1957         ns  = ldlm_res_to_ns(res);
1958
1959         /* Please do not, no matter how tempting, remove this LBUG without
1960          * talking to me first. -phik */
1961         if (lock->l_readers || lock->l_writers) {
1962                 LDLM_ERROR(lock, "lock still has references");
1963                 LBUG();
1964         }
1965
1966         if (lock->l_flags & LDLM_FL_WAITED)
1967                 ldlm_del_waiting_lock(lock);
1968
1969         /* Releases cancel callback. */
1970         ldlm_cancel_callback(lock);
1971
1972         /* Yes, second time, just in case it was added again while we were
1973          * running with no res lock in ldlm_cancel_callback */
1974         if (lock->l_flags & LDLM_FL_WAITED)
1975                 ldlm_del_waiting_lock(lock);
1976
1977         ldlm_resource_unlink_lock(lock);
1978         ldlm_lock_destroy_nolock(lock);
1979
1980         if (lock->l_granted_mode == lock->l_req_mode)
1981                 ldlm_pool_del(&ns->ns_pool, lock);
1982
1983         /* Make sure we will not be called again for same lock what is possible
1984          * if not to zero out lock->l_granted_mode */
1985         lock->l_granted_mode = LCK_MINMODE;
1986         unlock_res_and_lock(lock);
1987 }
1988 EXPORT_SYMBOL(ldlm_lock_cancel);
1989
1990 /**
1991  * Set opaque data into the lock that only makes sense to upper layer.
1992  */
1993 int ldlm_lock_set_data(struct lustre_handle *lockh, void *data)
1994 {
1995         struct ldlm_lock *lock = ldlm_handle2lock(lockh);
1996         int rc = -EINVAL;
1997
1998         if (lock) {
1999                 if (lock->l_ast_data == NULL)
2000                         lock->l_ast_data = data;
2001                 if (lock->l_ast_data == data)
2002                         rc = 0;
2003                 LDLM_LOCK_PUT(lock);
2004         }
2005         return rc;
2006 }
2007 EXPORT_SYMBOL(ldlm_lock_set_data);
2008
2009 struct export_cl_data {
2010         struct obd_export       *ecl_exp;
2011         int                     ecl_loop;
2012 };
2013
2014 /**
2015  * Iterator function for ldlm_cancel_locks_for_export.
2016  * Cancels passed locks.
2017  */
2018 int ldlm_cancel_locks_for_export_cb(struct cfs_hash *hs, struct cfs_hash_bd *bd,
2019                                     struct hlist_node *hnode, void *data)
2020
2021 {
2022         struct export_cl_data   *ecl = (struct export_cl_data *)data;
2023         struct obd_export       *exp  = ecl->ecl_exp;
2024         struct ldlm_lock     *lock = cfs_hash_object(hs, hnode);
2025         struct ldlm_resource *res;
2026
2027         res = ldlm_resource_getref(lock->l_resource);
2028         LDLM_LOCK_GET(lock);
2029
2030         LDLM_DEBUG(lock, "export %p", exp);
2031         ldlm_res_lvbo_update(res, NULL, 1);
2032         ldlm_lock_cancel(lock);
2033         ldlm_reprocess_all(res);
2034         ldlm_resource_putref(res);
2035         LDLM_LOCK_RELEASE(lock);
2036
2037         ecl->ecl_loop++;
2038         if ((ecl->ecl_loop & -ecl->ecl_loop) == ecl->ecl_loop) {
2039                 CDEBUG(D_INFO,
2040                        "Cancel lock %p for export %p (loop %d), still have %d locks left on hash table.\n",
2041                        lock, exp, ecl->ecl_loop,
2042                        atomic_read(&hs->hs_count));
2043         }
2044
2045         return 0;
2046 }
2047
2048 /**
2049  * Cancel all locks for given export.
2050  *
2051  * Typically called on client disconnection/eviction
2052  */
2053 void ldlm_cancel_locks_for_export(struct obd_export *exp)
2054 {
2055         struct export_cl_data   ecl = {
2056                 .ecl_exp        = exp,
2057                 .ecl_loop       = 0,
2058         };
2059
2060         cfs_hash_for_each_empty(exp->exp_lock_hash,
2061                                 ldlm_cancel_locks_for_export_cb, &ecl);
2062 }
2063
2064 /**
2065  * Downgrade an exclusive lock.
2066  *
2067  * A fast variant of ldlm_lock_convert for conversion of exclusive
2068  * locks. The conversion is always successful.
2069  * Used by Commit on Sharing (COS) code.
2070  *
2071  * \param lock A lock to convert
2072  * \param new_mode new lock mode
2073  */
2074 void ldlm_lock_downgrade(struct ldlm_lock *lock, int new_mode)
2075 {
2076         LASSERT(lock->l_granted_mode & (LCK_PW | LCK_EX));
2077         LASSERT(new_mode == LCK_COS);
2078
2079         lock_res_and_lock(lock);
2080         ldlm_resource_unlink_lock(lock);
2081         /*
2082          * Remove the lock from pool as it will be added again in
2083          * ldlm_grant_lock() called below.
2084          */
2085         ldlm_pool_del(&ldlm_lock_to_ns(lock)->ns_pool, lock);
2086
2087         lock->l_req_mode = new_mode;
2088         ldlm_grant_lock(lock, NULL);
2089         unlock_res_and_lock(lock);
2090         ldlm_reprocess_all(lock->l_resource);
2091 }
2092 EXPORT_SYMBOL(ldlm_lock_downgrade);
2093
2094 /**
2095  * Attempt to convert already granted lock to a different mode.
2096  *
2097  * While lock conversion is not currently used, future client-side
2098  * optimizations could take advantage of it to avoid discarding cached
2099  * pages on a file.
2100  */
2101 struct ldlm_resource *ldlm_lock_convert(struct ldlm_lock *lock, int new_mode,
2102                                         __u32 *flags)
2103 {
2104         LIST_HEAD(rpc_list);
2105         struct ldlm_resource *res;
2106         struct ldlm_namespace *ns;
2107         int granted = 0;
2108         struct ldlm_interval *node;
2109
2110         /* Just return if mode is unchanged. */
2111         if (new_mode == lock->l_granted_mode) {
2112                 *flags |= LDLM_FL_BLOCK_GRANTED;
2113                 return lock->l_resource;
2114         }
2115
2116         /* I can't check the type of lock here because the bitlock of lock
2117          * is not held here, so do the allocation blindly. -jay */
2118         OBD_SLAB_ALLOC_PTR_GFP(node, ldlm_interval_slab, GFP_NOFS);
2119         if (node == NULL)
2120                 /* Actually, this causes EDEADLOCK to be returned */
2121                 return NULL;
2122
2123         LASSERTF((new_mode == LCK_PW && lock->l_granted_mode == LCK_PR),
2124                  "new_mode %u, granted %u\n", new_mode, lock->l_granted_mode);
2125
2126         lock_res_and_lock(lock);
2127
2128         res = lock->l_resource;
2129         ns  = ldlm_res_to_ns(res);
2130
2131         lock->l_req_mode = new_mode;
2132         if (res->lr_type == LDLM_PLAIN || res->lr_type == LDLM_IBITS) {
2133                 ldlm_resource_unlink_lock(lock);
2134         } else {
2135                 ldlm_resource_unlink_lock(lock);
2136                 if (res->lr_type == LDLM_EXTENT) {
2137                         /* FIXME: ugly code, I have to attach the lock to a
2138                          * interval node again since perhaps it will be granted
2139                          * soon */
2140                         INIT_LIST_HEAD(&node->li_group);
2141                         ldlm_interval_attach(node, lock);
2142                         node = NULL;
2143                 }
2144         }
2145
2146         /*
2147          * Remove old lock from the pool before adding the lock with new
2148          * mode below in ->policy()
2149          */
2150         ldlm_pool_del(&ns->ns_pool, lock);
2151
2152         /* If this is a local resource, put it on the appropriate list. */
2153         if (ns_is_client(ldlm_res_to_ns(res))) {
2154                 if (*flags & (LDLM_FL_BLOCK_CONV | LDLM_FL_BLOCK_GRANTED)) {
2155                         ldlm_resource_add_lock(res, &res->lr_converting, lock);
2156                 } else {
2157                         /* This should never happen, because of the way the
2158                          * server handles conversions. */
2159                         LDLM_ERROR(lock, "Erroneous flags %x on local lock\n",
2160                                    *flags);
2161                         LBUG();
2162
2163                         ldlm_grant_lock(lock, &rpc_list);
2164                         granted = 1;
2165                         /* FIXME: completion handling not with lr_lock held ! */
2166                         if (lock->l_completion_ast)
2167                                 lock->l_completion_ast(lock, 0, NULL);
2168                 }
2169         } else {
2170                 CERROR("This is client-side-only module, cannot handle LDLM_NAMESPACE_SERVER resource type lock.\n");
2171                 LBUG();
2172         }
2173         unlock_res_and_lock(lock);
2174
2175         if (granted)
2176                 ldlm_run_ast_work(ns, &rpc_list, LDLM_WORK_CP_AST);
2177         if (node)
2178                 OBD_SLAB_FREE(node, ldlm_interval_slab, sizeof(*node));
2179         return res;
2180 }
2181 EXPORT_SYMBOL(ldlm_lock_convert);
2182
2183 /**
2184  * Print lock with lock handle \a lockh description into debug log.
2185  *
2186  * Used when printing all locks on a resource for debug purposes.
2187  */
2188 void ldlm_lock_dump_handle(int level, struct lustre_handle *lockh)
2189 {
2190         struct ldlm_lock *lock;
2191
2192         if (!((libcfs_debug | D_ERROR) & level))
2193                 return;
2194
2195         lock = ldlm_handle2lock(lockh);
2196         if (lock == NULL)
2197                 return;
2198
2199         LDLM_DEBUG_LIMIT(level, lock, "###");
2200
2201         LDLM_LOCK_PUT(lock);
2202 }
2203 EXPORT_SYMBOL(ldlm_lock_dump_handle);
2204
2205 /**
2206  * Print lock information with custom message into debug log.
2207  * Helper function.
2208  */
2209 void _ldlm_lock_debug(struct ldlm_lock *lock,
2210                       struct libcfs_debug_msg_data *msgdata,
2211                       const char *fmt, ...)
2212 {
2213         va_list args;
2214         struct obd_export *exp = lock->l_export;
2215         struct ldlm_resource *resource = lock->l_resource;
2216         char *nid = "local";
2217
2218         va_start(args, fmt);
2219
2220         if (exp && exp->exp_connection) {
2221                 nid = libcfs_nid2str(exp->exp_connection->c_peer.nid);
2222         } else if (exp && exp->exp_obd != NULL) {
2223                 struct obd_import *imp = exp->exp_obd->u.cli.cl_import;
2224
2225                 nid = libcfs_nid2str(imp->imp_connection->c_peer.nid);
2226         }
2227
2228         if (resource == NULL) {
2229                 libcfs_debug_vmsg2(msgdata, fmt, args,
2230                                    " ns: \?\? lock: %p/%#llx lrc: %d/%d,%d mode: %s/%s res: \?\? rrc=\?\? type: \?\?\? flags: %#llx nid: %s remote: %#llx expref: %d pid: %u timeout: %lu lvb_type: %d\n",
2231                                    lock,
2232                                    lock->l_handle.h_cookie, atomic_read(&lock->l_refc),
2233                                    lock->l_readers, lock->l_writers,
2234                                    ldlm_lockname[lock->l_granted_mode],
2235                                    ldlm_lockname[lock->l_req_mode],
2236                                    lock->l_flags, nid, lock->l_remote_handle.cookie,
2237                                    exp ? atomic_read(&exp->exp_refcount) : -99,
2238                                    lock->l_pid, lock->l_callback_timeout, lock->l_lvb_type);
2239                 va_end(args);
2240                 return;
2241         }
2242
2243         switch (resource->lr_type) {
2244         case LDLM_EXTENT:
2245                 libcfs_debug_vmsg2(msgdata, fmt, args,
2246                                    " ns: %s lock: %p/%#llx lrc: %d/%d,%d mode: %s/%s res: " DLDLMRES " rrc: %d type: %s [%llu->%llu] (req %llu->%llu) flags: %#llx nid: %s remote: %#llx expref: %d pid: %u timeout: %lu lvb_type: %d\n",
2247                                    ldlm_lock_to_ns_name(lock), lock,
2248                                    lock->l_handle.h_cookie, atomic_read(&lock->l_refc),
2249                                    lock->l_readers, lock->l_writers,
2250                                    ldlm_lockname[lock->l_granted_mode],
2251                                    ldlm_lockname[lock->l_req_mode],
2252                                    PLDLMRES(resource),
2253                                    atomic_read(&resource->lr_refcount),
2254                                    ldlm_typename[resource->lr_type],
2255                                    lock->l_policy_data.l_extent.start,
2256                                    lock->l_policy_data.l_extent.end,
2257                                    lock->l_req_extent.start, lock->l_req_extent.end,
2258                                    lock->l_flags, nid, lock->l_remote_handle.cookie,
2259                                    exp ? atomic_read(&exp->exp_refcount) : -99,
2260                                    lock->l_pid, lock->l_callback_timeout,
2261                                    lock->l_lvb_type);
2262                 break;
2263
2264         case LDLM_FLOCK:
2265                 libcfs_debug_vmsg2(msgdata, fmt, args,
2266                                    " ns: %s lock: %p/%#llx lrc: %d/%d,%d mode: %s/%s res: " DLDLMRES " rrc: %d type: %s pid: %d [%llu->%llu] flags: %#llx nid: %s remote: %#llx expref: %d pid: %u timeout: %lu\n",
2267                                    ldlm_lock_to_ns_name(lock), lock,
2268                                    lock->l_handle.h_cookie, atomic_read(&lock->l_refc),
2269                                    lock->l_readers, lock->l_writers,
2270                                    ldlm_lockname[lock->l_granted_mode],
2271                                    ldlm_lockname[lock->l_req_mode],
2272                                    PLDLMRES(resource),
2273                                    atomic_read(&resource->lr_refcount),
2274                                    ldlm_typename[resource->lr_type],
2275                                    lock->l_policy_data.l_flock.pid,
2276                                    lock->l_policy_data.l_flock.start,
2277                                    lock->l_policy_data.l_flock.end,
2278                                    lock->l_flags, nid, lock->l_remote_handle.cookie,
2279                                    exp ? atomic_read(&exp->exp_refcount) : -99,
2280                                    lock->l_pid, lock->l_callback_timeout);
2281                 break;
2282
2283         case LDLM_IBITS:
2284                 libcfs_debug_vmsg2(msgdata, fmt, args,
2285                                    " ns: %s lock: %p/%#llx lrc: %d/%d,%d mode: %s/%s res: " DLDLMRES " bits %#llx rrc: %d type: %s flags: %#llx nid: %s remote: %#llx expref: %d pid: %u timeout: %lu lvb_type: %d\n",
2286                                    ldlm_lock_to_ns_name(lock),
2287                                    lock, lock->l_handle.h_cookie,
2288                                    atomic_read(&lock->l_refc),
2289                                    lock->l_readers, lock->l_writers,
2290                                    ldlm_lockname[lock->l_granted_mode],
2291                                    ldlm_lockname[lock->l_req_mode],
2292                                    PLDLMRES(resource),
2293                                    lock->l_policy_data.l_inodebits.bits,
2294                                    atomic_read(&resource->lr_refcount),
2295                                    ldlm_typename[resource->lr_type],
2296                                    lock->l_flags, nid, lock->l_remote_handle.cookie,
2297                                    exp ? atomic_read(&exp->exp_refcount) : -99,
2298                                    lock->l_pid, lock->l_callback_timeout,
2299                                    lock->l_lvb_type);
2300                 break;
2301
2302         default:
2303                 libcfs_debug_vmsg2(msgdata, fmt, args,
2304                                    " ns: %s lock: %p/%#llx lrc: %d/%d,%d mode: %s/%s res: " DLDLMRES " rrc: %d type: %s flags: %#llx nid: %s remote: %#llx expref: %d pid: %u timeout: %lu lvb_type: %d\n",
2305                                    ldlm_lock_to_ns_name(lock),
2306                                    lock, lock->l_handle.h_cookie,
2307                                    atomic_read(&lock->l_refc),
2308                                    lock->l_readers, lock->l_writers,
2309                                    ldlm_lockname[lock->l_granted_mode],
2310                                    ldlm_lockname[lock->l_req_mode],
2311                                    PLDLMRES(resource),
2312                                    atomic_read(&resource->lr_refcount),
2313                                    ldlm_typename[resource->lr_type],
2314                                    lock->l_flags, nid, lock->l_remote_handle.cookie,
2315                                    exp ? atomic_read(&exp->exp_refcount) : -99,
2316                                    lock->l_pid, lock->l_callback_timeout,
2317                                    lock->l_lvb_type);
2318                 break;
2319         }
2320         va_end(args);
2321 }
2322 EXPORT_SYMBOL(_ldlm_lock_debug);