lustre/ldlm/ldlm_lock.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  * GPL HEADER START
   5  *
   6  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 only,
  10  * as published by the Free Software Foundation.
  11  *
  12  * This program is distributed in the hope that it will be useful, but
  13  * WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * General Public License version 2 for more details (a copy is included
  16  * in the LICENSE file that accompanied this code).
  17  *
  18  * You should have received a copy of the GNU General Public License
  19  * version 2 along with this program; If not, see
  20  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  21  *
  22  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  23  * CA 95054 USA or visit www.sun.com if you need additional information or
  24  * have any questions.
  25  *
  26  * GPL HEADER END
  27  */
  28 /*
  29  * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
  30  * Use is subject to license terms.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  *
  36  * 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 #ifdef __KERNEL__
  45 # include <libcfs/libcfs.h>
  46 # ifndef HAVE_VFS_INTENT_PATCHES
  47 # include <linux/lustre_intent.h>
  48 # endif
  49 #else
  50 # include <liblustre.h>
  51 #endif
  52
  53 #include <obd_class.h>
  54 #include "ldlm_internal.h"
  55
  56 /* lock types */
  57 char *ldlm_lockname[] = {
  58         [0] "--",
  59         [LCK_EX] "EX",
  60         [LCK_PW] "PW",
  61         [LCK_PR] "PR",
  62         [LCK_CW] "CW",
  63         [LCK_CR] "CR",
  64         [LCK_NL] "NL",
  65         [LCK_GROUP] "GROUP",
  66         [LCK_COS] "COS"
  67 };
  68
  69 char *ldlm_typename[] = {
  70         [LDLM_PLAIN] "PLN",
  71         [LDLM_EXTENT] "EXT",
  72         [LDLM_FLOCK] "FLK",
  73         [LDLM_IBITS] "IBT",
  74 };
  75
  76 char *ldlm_it2str(int it)
  77 {
  78         switch (it) {
  79         case IT_OPEN:
  80                 return "open";
  81         case IT_CREAT:
  82                 return "creat";
  83         case (IT_OPEN | IT_CREAT):
  84                 return "open|creat";
  85         case IT_READDIR:
  86                 return "readdir";
  87         case IT_GETATTR:
  88                 return "getattr";
  89         case IT_LOOKUP:
  90                 return "lookup";
  91         case IT_UNLINK:
  92                 return "unlink";
  93         case IT_GETXATTR:
  94                 return "getxattr";
  95         default:
  96                 CERROR("Unknown intent %d\n", it);
  97                 return "UNKNOWN";
  98         }
  99 }
 100
 101 extern cfs_mem_cache_t *ldlm_lock_slab;
 102
 103 static ldlm_processing_policy ldlm_processing_policy_table[] = {
 104         [LDLM_PLAIN] ldlm_process_plain_lock,
 105         [LDLM_EXTENT] ldlm_process_extent_lock,
 106 #ifdef __KERNEL__
 107         [LDLM_FLOCK] ldlm_process_flock_lock,
 108 #endif
 109         [LDLM_IBITS] ldlm_process_inodebits_lock,
 110 };
 111
 112 ldlm_processing_policy ldlm_get_processing_policy(struct ldlm_resource *res)
 113 {
 114         return ldlm_processing_policy_table[res->lr_type];
 115 }
 116
 117 void ldlm_register_intent(struct ldlm_namespace *ns, ldlm_res_policy arg)
 118 {
 119         ns->ns_policy = arg;
 120 }
 121
 122 /*
 123  * REFCOUNTED LOCK OBJECTS
 124  */
 125
 126
 127 /*
 128  * Lock refcounts, during creation:
 129  *   - one special one for allocation, dec'd only once in destroy
 130  *   - one for being a lock that's in-use
 131  *   - one for the addref associated with a new lock
 132  */
 133 struct ldlm_lock *ldlm_lock_get(struct ldlm_lock *lock)
 134 {
 135         cfs_atomic_inc(&lock->l_refc);
 136         return lock;
 137 }
 138
 139 static void ldlm_lock_free(struct ldlm_lock *lock, size_t size)
 140 {
 141         LASSERT(size == sizeof(*lock));
 142         OBD_SLAB_FREE(lock, ldlm_lock_slab, sizeof(*lock));
 143 }
 144
 145 void ldlm_lock_put(struct ldlm_lock *lock)
 146 {
 147         ENTRY;
 148
 149         LASSERT(lock->l_resource != LP_POISON);
 150         LASSERT(cfs_atomic_read(&lock->l_refc) > 0);
 151         if (cfs_atomic_dec_and_test(&lock->l_refc)) {
 152                 struct ldlm_resource *res;
 153
 154                 LDLM_DEBUG(lock,
 155                            "final lock_put on destroyed lock, freeing it.");
 156
 157                 res = lock->l_resource;
 158                 LASSERT(lock->l_destroyed);
 159                 LASSERT(cfs_list_empty(&lock->l_res_link));
 160                 LASSERT(cfs_list_empty(&lock->l_pending_chain));
 161
 162                 cfs_atomic_dec(&res->lr_namespace->ns_locks);
 163                 lu_ref_del(&res->lr_reference, "lock", lock);
 164                 ldlm_resource_putref(res);
 165                 lock->l_resource = NULL;
 166                 if (lock->l_export) {
 167                         class_export_lock_put(lock->l_export, lock);
 168                         lock->l_export = NULL;
 169                 }
 170
 171                 if (lock->l_lvb_data != NULL)
 172                         OBD_FREE(lock->l_lvb_data, lock->l_lvb_len);
 173
 174                 ldlm_interval_free(ldlm_interval_detach(lock));
 175                 lu_ref_fini(&lock->l_reference);
 176                 OBD_FREE_RCU_CB(lock, sizeof(*lock), &lock->l_handle,
 177                                 ldlm_lock_free);
 178         }
 179
 180         EXIT;
 181 }
 182
 183 int ldlm_lock_remove_from_lru_nolock(struct ldlm_lock *lock)
 184 {
 185         int rc = 0;
 186         if (!cfs_list_empty(&lock->l_lru)) {
 187                 struct ldlm_namespace *ns = lock->l_resource->lr_namespace;
 188                 LASSERT(lock->l_resource->lr_type != LDLM_FLOCK);
 189                 cfs_list_del_init(&lock->l_lru);
 190                 if (lock->l_flags & LDLM_FL_SKIPPED)
 191                         lock->l_flags &= ~LDLM_FL_SKIPPED;
 192                 LASSERT(ns->ns_nr_unused > 0);
 193                 ns->ns_nr_unused--;
 194                 rc = 1;
 195         }
 196         return rc;
 197 }
 198
 199 int ldlm_lock_remove_from_lru(struct ldlm_lock *lock)
 200 {
 201         struct ldlm_namespace *ns = lock->l_resource->lr_namespace;
 202         int rc;
 203         ENTRY;
 204         cfs_spin_lock(&ns->ns_unused_lock);
 205         rc = ldlm_lock_remove_from_lru_nolock(lock);
 206         cfs_spin_unlock(&ns->ns_unused_lock);
 207         EXIT;
 208         return rc;
 209 }
 210
 211 void ldlm_lock_add_to_lru_nolock(struct ldlm_lock *lock)
 212 {
 213         struct ldlm_namespace *ns = lock->l_resource->lr_namespace;
 214         lock->l_last_used = cfs_time_current();
 215         LASSERT(cfs_list_empty(&lock->l_lru));
 216         LASSERT(lock->l_resource->lr_type != LDLM_FLOCK);
 217         cfs_list_add_tail(&lock->l_lru, &ns->ns_unused_list);
 218         LASSERT(ns->ns_nr_unused >= 0);
 219         ns->ns_nr_unused++;
 220 }
 221
 222 void ldlm_lock_add_to_lru(struct ldlm_lock *lock)
 223 {
 224         struct ldlm_namespace *ns = lock->l_resource->lr_namespace;
 225         ENTRY;
 226         cfs_spin_lock(&ns->ns_unused_lock);
 227         ldlm_lock_add_to_lru_nolock(lock);
 228         cfs_spin_unlock(&ns->ns_unused_lock);
 229         EXIT;
 230 }
 231
 232 void ldlm_lock_touch_in_lru(struct ldlm_lock *lock)
 233 {
 234         struct ldlm_namespace *ns = lock->l_resource->lr_namespace;
 235         ENTRY;
 236         cfs_spin_lock(&ns->ns_unused_lock);
 237         if (!cfs_list_empty(&lock->l_lru)) {
 238                 ldlm_lock_remove_from_lru_nolock(lock);
 239                 ldlm_lock_add_to_lru_nolock(lock);
 240         }
 241         cfs_spin_unlock(&ns->ns_unused_lock);
 242         EXIT;
 243 }
 244
 245 /* This used to have a 'strict' flag, which recovery would use to mark an
 246  * in-use lock as needing-to-die.  Lest I am ever tempted to put it back, I
 247  * shall explain why it's gone: with the new hash table scheme, once you call
 248  * ldlm_lock_destroy, you can never drop your final references on this lock.
 249  * Because it's not in the hash table anymore.  -phil */
 250 int ldlm_lock_destroy_internal(struct ldlm_lock *lock)
 251 {
 252         ENTRY;
 253
 254         if (lock->l_readers || lock->l_writers) {
 255                 LDLM_ERROR(lock, "lock still has references");
 256                 ldlm_lock_dump(D_ERROR, lock, 0);
 257                 LBUG();
 258         }
 259
 260         if (!cfs_list_empty(&lock->l_res_link)) {
 261                 LDLM_ERROR(lock, "lock still on resource");
 262                 ldlm_lock_dump(D_ERROR, lock, 0);
 263                 LBUG();
 264         }
 265
 266         if (lock->l_destroyed) {
 267                 LASSERT(cfs_list_empty(&lock->l_lru));
 268                 EXIT;
 269                 return 0;
 270         }
 271         lock->l_destroyed = 1;
 272
 273         if (lock->l_export && lock->l_export->exp_lock_hash &&
 274             !cfs_hlist_unhashed(&lock->l_exp_hash))
 275                 cfs_hash_del(lock->l_export->exp_lock_hash,
 276                              &lock->l_remote_handle, &lock->l_exp_hash);
 277
 278         ldlm_lock_remove_from_lru(lock);
 279         class_handle_unhash(&lock->l_handle);
 280
 281 #if 0
 282         /* Wake anyone waiting for this lock */
 283         /* FIXME: I should probably add yet another flag, instead of using
 284          * l_export to only call this on clients */
 285         if (lock->l_export)
 286                 class_export_put(lock->l_export);
 287         lock->l_export = NULL;
 288         if (lock->l_export && lock->l_completion_ast)
 289                 lock->l_completion_ast(lock, 0);
 290 #endif
 291         EXIT;
 292         return 1;
 293 }
 294
 295 void ldlm_lock_destroy(struct ldlm_lock *lock)
 296 {
 297         int first;
 298         ENTRY;
 299         lock_res_and_lock(lock);
 300         first = ldlm_lock_destroy_internal(lock);
 301         unlock_res_and_lock(lock);
 302
 303         /* drop reference from hashtable only for first destroy */
 304         if (first) {
 305                 lu_ref_del(&lock->l_reference, "hash", lock);
 306                 LDLM_LOCK_RELEASE(lock);
 307         }
 308         EXIT;
 309 }
 310
 311 void ldlm_lock_destroy_nolock(struct ldlm_lock *lock)
 312 {
 313         int first;
 314         ENTRY;
 315         first = ldlm_lock_destroy_internal(lock);
 316         /* drop reference from hashtable only for first destroy */
 317         if (first) {
 318                 lu_ref_del(&lock->l_reference, "hash", lock);
 319                 LDLM_LOCK_RELEASE(lock);
 320         }
 321         EXIT;
 322 }
 323
 324 /* this is called by portals_handle2object with the handle lock taken */
 325 static void lock_handle_addref(void *lock)
 326 {
 327         LDLM_LOCK_GET((struct ldlm_lock *)lock);
 328 }
 329
 330 /*
 331  * usage: pass in a resource on which you have done ldlm_resource_get
 332  *        pass in a parent lock on which you have done a ldlm_lock_get
 333  *        after return, ldlm_*_put the resource and parent
 334  * returns: lock with refcount 2 - one for current caller and one for remote
 335  */
 336 static struct ldlm_lock *ldlm_lock_new(struct ldlm_resource *resource)
 337 {
 338         struct ldlm_lock *lock;
 339         ENTRY;
 340
 341         if (resource == NULL)
 342                 LBUG();
 343
 344         OBD_SLAB_ALLOC_PTR_GFP(lock, ldlm_lock_slab, CFS_ALLOC_IO);
 345         if (lock == NULL)
 346                 RETURN(NULL);
 347
 348         cfs_spin_lock_init(&lock->l_lock);
 349         lock->l_resource = ldlm_resource_getref(resource);
 350         lu_ref_add(&resource->lr_reference, "lock", lock);
 351
 352         cfs_atomic_set(&lock->l_refc, 2);
 353         CFS_INIT_LIST_HEAD(&lock->l_res_link);
 354         CFS_INIT_LIST_HEAD(&lock->l_lru);
 355         CFS_INIT_LIST_HEAD(&lock->l_pending_chain);
 356         CFS_INIT_LIST_HEAD(&lock->l_bl_ast);
 357         CFS_INIT_LIST_HEAD(&lock->l_cp_ast);
 358         CFS_INIT_LIST_HEAD(&lock->l_rk_ast);
 359         cfs_waitq_init(&lock->l_waitq);
 360         lock->l_blocking_lock = NULL;
 361         CFS_INIT_LIST_HEAD(&lock->l_sl_mode);
 362         CFS_INIT_LIST_HEAD(&lock->l_sl_policy);
 363         CFS_INIT_HLIST_NODE(&lock->l_exp_hash);
 364
 365         cfs_atomic_inc(&resource->lr_namespace->ns_locks);
 366         CFS_INIT_LIST_HEAD(&lock->l_handle.h_link);
 367         class_handle_hash(&lock->l_handle, lock_handle_addref);
 368
 369         CFS_INIT_LIST_HEAD(&lock->l_extents_list);
 370         cfs_spin_lock_init(&lock->l_extents_list_lock);
 371         CFS_INIT_LIST_HEAD(&lock->l_cache_locks_list);
 372         lu_ref_init(&lock->l_reference);
 373         lu_ref_add(&lock->l_reference, "hash", lock);
 374         lock->l_callback_timeout = 0;
 375
 376 #if LUSTRE_TRACKS_LOCK_EXP_REFS
 377         CFS_INIT_LIST_HEAD(&lock->l_exp_refs_link);
 378         lock->l_exp_refs_nr = 0;
 379         lock->l_exp_refs_target = NULL;
 380 #endif
 381
 382         RETURN(lock);
 383 }
 384
 385 int ldlm_lock_change_resource(struct ldlm_namespace *ns, struct ldlm_lock *lock,
 386                               const struct ldlm_res_id *new_resid)
 387 {
 388         struct ldlm_resource *oldres = lock->l_resource;
 389         struct ldlm_resource *newres;
 390         int type;
 391         ENTRY;
 392
 393         LASSERT(ns_is_client(ns));
 394
 395         lock_res_and_lock(lock);
 396         if (memcmp(new_resid, &lock->l_resource->lr_name,
 397                    sizeof(lock->l_resource->lr_name)) == 0) {
 398                 /* Nothing to do */
 399                 unlock_res_and_lock(lock);
 400                 RETURN(0);
 401         }
 402
 403         LASSERT(new_resid->name[0] != 0);
 404
 405         /* This function assumes that the lock isn't on any lists */
 406         LASSERT(cfs_list_empty(&lock->l_res_link));
 407
 408         type = oldres->lr_type;
 409         unlock_res_and_lock(lock);
 410
 411         newres = ldlm_resource_get(ns, NULL, new_resid, type, 1);
 412         lu_ref_add(&newres->lr_reference, "lock", lock);
 413         if (newres == NULL)
 414                 RETURN(-ENOMEM);
 415         /*
 416          * To flip the lock from the old to the new resource, lock, oldres and
 417          * newres have to be locked. Resource spin-locks are nested within
 418          * lock->l_lock, and are taken in the memory address order to avoid
 419          * dead-locks.
 420          */
 421         cfs_spin_lock(&lock->l_lock);
 422         oldres = lock->l_resource;
 423         if (oldres < newres) {
 424                 lock_res(oldres);
 425                 lock_res_nested(newres, LRT_NEW);
 426         } else {
 427                 lock_res(newres);
 428                 lock_res_nested(oldres, LRT_NEW);
 429         }
 430         LASSERT(memcmp(new_resid, &oldres->lr_name,
 431                        sizeof oldres->lr_name) != 0);
 432         lock->l_resource = newres;
 433         unlock_res(oldres);
 434         unlock_res_and_lock(lock);
 435
 436         /* ...and the flowers are still standing! */
 437         lu_ref_del(&oldres->lr_reference, "lock", lock);
 438         ldlm_resource_putref(oldres);
 439
 440         RETURN(0);
 441 }
 442
 443 /*
 444  *  HANDLES
 445  */
 446
 447 void ldlm_lock2handle(const struct ldlm_lock *lock, struct lustre_handle *lockh)
 448 {
 449         lockh->cookie = lock->l_handle.h_cookie;
 450 }
 451
 452 /* if flags: atomically get the lock and set the flags.
 453  *           Return NULL if flag already set
 454  */
 455
 456 struct ldlm_lock *__ldlm_handle2lock(const struct lustre_handle *handle,
 457                                      int flags)
 458 {
 459         struct ldlm_namespace *ns;
 460         struct ldlm_lock *lock, *retval = NULL;
 461         ENTRY;
 462
 463         LASSERT(handle);
 464
 465         lock = class_handle2object(handle->cookie);
 466         if (lock == NULL)
 467                 RETURN(NULL);
 468
 469         LASSERT(lock->l_resource != NULL);
 470         ns = lock->l_resource->lr_namespace;
 471         LASSERT(ns != NULL);
 472
 473         lu_ref_add_atomic(&lock->l_reference, "handle", cfs_current());
 474         lock_res_and_lock(lock);
 475
 476         /* It's unlikely but possible that someone marked the lock as
 477          * destroyed after we did handle2object on it */
 478         if (lock->l_destroyed) {
 479                 unlock_res_and_lock(lock);
 480                 CDEBUG(D_INFO, "lock already destroyed: lock %p\n", lock);
 481                 LDLM_LOCK_PUT(lock);
 482                 GOTO(out, retval);
 483         }
 484
 485         if (flags && (lock->l_flags & flags)) {
 486                 unlock_res_and_lock(lock);
 487                 LDLM_LOCK_PUT(lock);
 488                 GOTO(out, retval);
 489         }
 490
 491         if (flags)
 492                 lock->l_flags |= flags;
 493
 494         unlock_res_and_lock(lock);
 495         retval = lock;
 496         EXIT;
 497  out:
 498         return retval;
 499 }
 500
 501 void ldlm_lock2desc(struct ldlm_lock *lock, struct ldlm_lock_desc *desc)
 502 {
 503         struct obd_export *exp = lock->l_export?:lock->l_conn_export;
 504         /* INODEBITS_INTEROP: If the other side does not support
 505          * inodebits, reply with a plain lock descriptor.
 506          */
 507         if ((lock->l_resource->lr_type == LDLM_IBITS) &&
 508             (exp && !(exp->exp_connect_flags & OBD_CONNECT_IBITS))) {
 509                 /* Make sure all the right bits are set in this lock we
 510                    are going to pass to client */
 511                 LASSERTF(lock->l_policy_data.l_inodebits.bits ==
 512                          (MDS_INODELOCK_LOOKUP|MDS_INODELOCK_UPDATE),
 513                          "Inappropriate inode lock bits during "
 514                          "conversion " LPU64 "\n",
 515                          lock->l_policy_data.l_inodebits.bits);
 516
 517                 ldlm_res2desc(lock->l_resource, &desc->l_resource);
 518                 desc->l_resource.lr_type = LDLM_PLAIN;
 519
 520                 /* Convert "new" lock mode to something old client can
 521                    understand */
 522                 if ((lock->l_req_mode == LCK_CR) ||
 523                     (lock->l_req_mode == LCK_CW))
 524                         desc->l_req_mode = LCK_PR;
 525                 else
 526                         desc->l_req_mode = lock->l_req_mode;
 527                 if ((lock->l_granted_mode == LCK_CR) ||
 528                     (lock->l_granted_mode == LCK_CW)) {
 529                         desc->l_granted_mode = LCK_PR;
 530                 } else {
 531                         /* We never grant PW/EX locks to clients */
 532                         LASSERT((lock->l_granted_mode != LCK_PW) &&
 533                                 (lock->l_granted_mode != LCK_EX));
 534                         desc->l_granted_mode = lock->l_granted_mode;
 535                 }
 536
 537                 /* We do not copy policy here, because there is no
 538                    policy for plain locks */
 539         } else {
 540                 ldlm_res2desc(lock->l_resource, &desc->l_resource);
 541                 desc->l_req_mode = lock->l_req_mode;
 542                 desc->l_granted_mode = lock->l_granted_mode;
 543                 desc->l_policy_data = lock->l_policy_data;
 544         }
 545 }
 546
 547 void ldlm_add_bl_work_item(struct ldlm_lock *lock, struct ldlm_lock *new,
 548                            cfs_list_t *work_list)
 549 {
 550         if ((lock->l_flags & LDLM_FL_AST_SENT) == 0) {
 551                 LDLM_DEBUG(lock, "lock incompatible; sending blocking AST.");
 552                 lock->l_flags |= LDLM_FL_AST_SENT;
 553                 /* If the enqueuing client said so, tell the AST recipient to
 554                  * discard dirty data, rather than writing back. */
 555                 if (new->l_flags & LDLM_AST_DISCARD_DATA)
 556                         lock->l_flags |= LDLM_FL_DISCARD_DATA;
 557                 LASSERT(cfs_list_empty(&lock->l_bl_ast));
 558                 cfs_list_add(&lock->l_bl_ast, work_list);
 559                 LDLM_LOCK_GET(lock);
 560                 LASSERT(lock->l_blocking_lock == NULL);
 561                 lock->l_blocking_lock = LDLM_LOCK_GET(new);
 562         }
 563 }
 564
 565 void ldlm_add_cp_work_item(struct ldlm_lock *lock, cfs_list_t *work_list)
 566 {
 567         if ((lock->l_flags & LDLM_FL_CP_REQD) == 0) {
 568                 lock->l_flags |= LDLM_FL_CP_REQD;
 569                 LDLM_DEBUG(lock, "lock granted; sending completion AST.");
 570                 LASSERT(cfs_list_empty(&lock->l_cp_ast));
 571                 cfs_list_add(&lock->l_cp_ast, work_list);
 572                 LDLM_LOCK_GET(lock);
 573         }
 574 }
 575
 576 /* must be called with lr_lock held */
 577 void ldlm_add_ast_work_item(struct ldlm_lock *lock, struct ldlm_lock *new,
 578                             cfs_list_t *work_list)
 579 {
 580         ENTRY;
 581         check_res_locked(lock->l_resource);
 582         if (new)
 583                 ldlm_add_bl_work_item(lock, new, work_list);
 584         else
 585                 ldlm_add_cp_work_item(lock, work_list);
 586         EXIT;
 587 }
 588
 589 void ldlm_lock_addref(struct lustre_handle *lockh, __u32 mode)
 590 {
 591         struct ldlm_lock *lock;
 592
 593         lock = ldlm_handle2lock(lockh);
 594         LASSERT(lock != NULL);
 595         ldlm_lock_addref_internal(lock, mode);
 596         LDLM_LOCK_PUT(lock);
 597 }
 598
 599 void ldlm_lock_addref_internal_nolock(struct ldlm_lock *lock, __u32 mode)
 600 {
 601         ldlm_lock_remove_from_lru(lock);
 602         if (mode & (LCK_NL | LCK_CR | LCK_PR)) {
 603                 lock->l_readers++;
 604                 lu_ref_add_atomic(&lock->l_reference, "reader", lock);
 605         }
 606         if (mode & (LCK_EX | LCK_CW | LCK_PW | LCK_GROUP | LCK_COS)) {
 607                 lock->l_writers++;
 608                 lu_ref_add_atomic(&lock->l_reference, "writer", lock);
 609         }
 610         LDLM_LOCK_GET(lock);
 611         lu_ref_add_atomic(&lock->l_reference, "user", lock);
 612         LDLM_DEBUG(lock, "ldlm_lock_addref(%s)", ldlm_lockname[mode]);
 613 }
 614
 615 /**
 616  * Attempts to addref a lock, and fails if lock is already LDLM_FL_CBPENDING
 617  * or destroyed.
 618  *
 619  * \retval 0 success, lock was addref-ed
 620  *
 621  * \retval -EAGAIN lock is being canceled.
 622  */
 623 int ldlm_lock_addref_try(struct lustre_handle *lockh, __u32 mode)
 624 {
 625         struct ldlm_lock *lock;
 626         int               result;
 627
 628         result = -EAGAIN;
 629         lock = ldlm_handle2lock(lockh);
 630         if (lock != NULL) {
 631                 lock_res_and_lock(lock);
 632                 if (lock->l_readers != 0 || lock->l_writers != 0 ||
 633                     !(lock->l_flags & LDLM_FL_CBPENDING)) {
 634                         ldlm_lock_addref_internal_nolock(lock, mode);
 635                         result = 0;
 636                 }
 637                 unlock_res_and_lock(lock);
 638                 LDLM_LOCK_PUT(lock);
 639         }
 640         return result;
 641 }
 642
 643 /* only called for local locks */
 644 void ldlm_lock_addref_internal(struct ldlm_lock *lock, __u32 mode)
 645 {
 646         lock_res_and_lock(lock);
 647         ldlm_lock_addref_internal_nolock(lock, mode);
 648         unlock_res_and_lock(lock);
 649 }
 650
 651 /* only called in ldlm_flock_destroy and for local locks.
 652  *  * for LDLM_FLOCK type locks, l_blocking_ast is null, and
 653  *   * ldlm_lock_remove_from_lru() does nothing, it is safe
 654  *    * for ldlm_flock_destroy usage by dropping some code */
 655 void ldlm_lock_decref_internal_nolock(struct ldlm_lock *lock, __u32 mode)
 656 {
 657         LDLM_DEBUG(lock, "ldlm_lock_decref(%s)", ldlm_lockname[mode]);
 658         if (mode & (LCK_NL | LCK_CR | LCK_PR)) {
 659                 LASSERT(lock->l_readers > 0);
 660                 lu_ref_del(&lock->l_reference, "reader", lock);
 661                 lock->l_readers--;
 662         }
 663         if (mode & (LCK_EX | LCK_CW | LCK_PW | LCK_GROUP | LCK_COS)) {
 664                 LASSERT(lock->l_writers > 0);
 665                 lu_ref_del(&lock->l_reference, "writer", lock);
 666                 lock->l_writers--;
 667         }
 668
 669         lu_ref_del(&lock->l_reference, "user", lock);
 670         LDLM_LOCK_RELEASE(lock);    /* matches the LDLM_LOCK_GET() in addref */
 671 }
 672
 673 void ldlm_lock_decref_internal(struct ldlm_lock *lock, __u32 mode)
 674 {
 675         struct ldlm_namespace *ns;
 676         ENTRY;
 677
 678         lock_res_and_lock(lock);
 679
 680         ns = lock->l_resource->lr_namespace;
 681
 682         ldlm_lock_decref_internal_nolock(lock, mode);
 683
 684         if (lock->l_flags & LDLM_FL_LOCAL &&
 685             !lock->l_readers && !lock->l_writers) {
 686                 /* If this is a local lock on a server namespace and this was
 687                  * the last reference, cancel the lock. */
 688                 CDEBUG(D_INFO, "forcing cancel of local lock\n");
 689                 lock->l_flags |= LDLM_FL_CBPENDING;
 690         }
 691
 692         if (!lock->l_readers && !lock->l_writers &&
 693             (lock->l_flags & LDLM_FL_CBPENDING)) {
 694                 /* If we received a blocked AST and this was the last reference,
 695                  * run the callback. */
 696                 if (ns_is_server(ns) && lock->l_export)
 697                         CERROR("FL_CBPENDING set on non-local lock--just a "
 698                                "warning\n");
 699
 700                 LDLM_DEBUG(lock, "final decref done on cbpending lock");
 701
 702                 LDLM_LOCK_GET(lock); /* dropped by bl thread */
 703                 ldlm_lock_remove_from_lru(lock);
 704                 unlock_res_and_lock(lock);
 705
 706                 if (lock->l_flags & LDLM_FL_FAIL_LOC)
 707                         OBD_RACE(OBD_FAIL_LDLM_CP_BL_RACE);
 708
 709                 if ((lock->l_flags & LDLM_FL_ATOMIC_CB) ||
 710                     ldlm_bl_to_thread_lock(ns, NULL, lock) != 0)
 711                         ldlm_handle_bl_callback(ns, NULL, lock);
 712         } else if (ns_is_client(ns) &&
 713                    !lock->l_readers && !lock->l_writers &&
 714                    !(lock->l_flags & LDLM_FL_BL_AST)) {
 715                 /* If this is a client-side namespace and this was the last
 716                  * reference, put it on the LRU. */
 717                 ldlm_lock_add_to_lru(lock);
 718                 unlock_res_and_lock(lock);
 719
 720                 if (lock->l_flags & LDLM_FL_FAIL_LOC)
 721                         OBD_RACE(OBD_FAIL_LDLM_CP_BL_RACE);
 722
 723                 /* Call ldlm_cancel_lru() only if EARLY_CANCEL and LRU RESIZE
 724                  * are not supported by the server, otherwise, it is done on
 725                  * enqueue. */
 726                 if (!exp_connect_cancelset(lock->l_conn_export) &&
 727                     !ns_connect_lru_resize(ns))
 728                         ldlm_cancel_lru(ns, 0, LDLM_ASYNC, 0);
 729         } else {
 730                 unlock_res_and_lock(lock);
 731         }
 732
 733         EXIT;
 734 }
 735
 736 void ldlm_lock_decref(struct lustre_handle *lockh, __u32 mode)
 737 {
 738         struct ldlm_lock *lock = __ldlm_handle2lock(lockh, 0);
 739         LASSERTF(lock != NULL, "Non-existing lock: "LPX64"\n", lockh->cookie);
 740         ldlm_lock_decref_internal(lock, mode);
 741         LDLM_LOCK_PUT(lock);
 742 }
 743
 744 /* This will drop a lock reference and mark it for destruction, but will not
 745  * necessarily cancel the lock before returning. */
 746 void ldlm_lock_decref_and_cancel(struct lustre_handle *lockh, __u32 mode)
 747 {
 748         struct ldlm_lock *lock = __ldlm_handle2lock(lockh, 0);
 749         ENTRY;
 750
 751         LASSERT(lock != NULL);
 752
 753         LDLM_DEBUG(lock, "ldlm_lock_decref(%s)", ldlm_lockname[mode]);
 754         lock_res_and_lock(lock);
 755         lock->l_flags |= LDLM_FL_CBPENDING;
 756         unlock_res_and_lock(lock);
 757         ldlm_lock_decref_internal(lock, mode);
 758         LDLM_LOCK_PUT(lock);
 759 }
 760
 761 struct sl_insert_point {
 762         cfs_list_t *res_link;
 763         cfs_list_t *mode_link;
 764         cfs_list_t *policy_link;
 765 };
 766
 767 /*
 768  * search_granted_lock
 769  *
 770  * Description:
 771  *      Finds a position to insert the new lock.
 772  * Parameters:
 773  *      queue [input]:  the granted list where search acts on;
 774  *      req [input]:    the lock whose position to be located;
 775  *      prev [output]:  positions within 3 lists to insert @req to
 776  * Return Value:
 777  *      filled @prev
 778  * NOTE: called by
 779  *  - ldlm_grant_lock_with_skiplist
 780  */
 781 static void search_granted_lock(cfs_list_t *queue,
 782                                 struct ldlm_lock *req,
 783                                 struct sl_insert_point *prev)
 784 {
 785         cfs_list_t *tmp;
 786         struct ldlm_lock *lock, *mode_end, *policy_end;
 787         ENTRY;
 788
 789         cfs_list_for_each(tmp, queue) {
 790                 lock = cfs_list_entry(tmp, struct ldlm_lock, l_res_link);
 791
 792                 mode_end = cfs_list_entry(lock->l_sl_mode.prev,
 793                                           struct ldlm_lock, l_sl_mode);
 794
 795                 if (lock->l_req_mode != req->l_req_mode) {
 796                         /* jump to last lock of mode group */
 797                         tmp = &mode_end->l_res_link;
 798                         continue;
 799                 }
 800
 801                 /* suitable mode group is found */
 802                 if (lock->l_resource->lr_type == LDLM_PLAIN) {
 803                         /* insert point is last lock of the mode group */
 804                         prev->res_link = &mode_end->l_res_link;
 805                         prev->mode_link = &mode_end->l_sl_mode;
 806                         prev->policy_link = &req->l_sl_policy;
 807                         EXIT;
 808                         return;
 809                 } else if (lock->l_resource->lr_type == LDLM_IBITS) {
 810                         for (;;) {
 811                                 policy_end =
 812                                         cfs_list_entry(lock->l_sl_policy.prev,
 813                                                        struct ldlm_lock,
 814                                                        l_sl_policy);
 815
 816                                 if (lock->l_policy_data.l_inodebits.bits ==
 817                                     req->l_policy_data.l_inodebits.bits) {
 818                                         /* insert point is last lock of
 819                                          * the policy group */
 820                                         prev->res_link =
 821                                                 &policy_end->l_res_link;
 822                                         prev->mode_link =
 823                                                 &policy_end->l_sl_mode;
 824                                         prev->policy_link =
 825                                                 &policy_end->l_sl_policy;
 826                                         EXIT;
 827                                         return;
 828                                 }
 829
 830                                 if (policy_end == mode_end)
 831                                         /* done with mode group */
 832                                         break;
 833
 834                                 /* go to next policy group within mode group */
 835                                 tmp = policy_end->l_res_link.next;
 836                                 lock = cfs_list_entry(tmp, struct ldlm_lock,
 837                                                       l_res_link);
 838                         }  /* loop over policy groups within the mode group */
 839
 840                         /* insert point is last lock of the mode group,
 841                          * new policy group is started */
 842                         prev->res_link = &mode_end->l_res_link;
 843                         prev->mode_link = &mode_end->l_sl_mode;
 844                         prev->policy_link = &req->l_sl_policy;
 845                         EXIT;
 846                         return;
 847                 } else {
 848                         LDLM_ERROR(lock,"is not LDLM_PLAIN or LDLM_IBITS lock");
 849                         LBUG();
 850                 }
 851         }
 852
 853         /* insert point is last lock on the queue,
 854          * new mode group and new policy group are started */
 855         prev->res_link = queue->prev;
 856         prev->mode_link = &req->l_sl_mode;
 857         prev->policy_link = &req->l_sl_policy;
 858         EXIT;
 859         return;
 860 }
 861
 862 static void ldlm_granted_list_add_lock(struct ldlm_lock *lock,
 863                                        struct sl_insert_point *prev)
 864 {
 865         struct ldlm_resource *res = lock->l_resource;
 866         ENTRY;
 867
 868         check_res_locked(res);
 869
 870         ldlm_resource_dump(D_INFO, res);
 871         CDEBUG(D_OTHER, "About to add this lock:\n");
 872         ldlm_lock_dump(D_OTHER, lock, 0);
 873
 874         if (lock->l_destroyed) {
 875                 CDEBUG(D_OTHER, "Lock destroyed, not adding to resource\n");
 876                 return;
 877         }
 878
 879         LASSERT(cfs_list_empty(&lock->l_res_link));
 880         LASSERT(cfs_list_empty(&lock->l_sl_mode));
 881         LASSERT(cfs_list_empty(&lock->l_sl_policy));
 882
 883         cfs_list_add(&lock->l_res_link, prev->res_link);
 884         cfs_list_add(&lock->l_sl_mode, prev->mode_link);
 885         cfs_list_add(&lock->l_sl_policy, prev->policy_link);
 886
 887         EXIT;
 888 }
 889
 890 static void ldlm_grant_lock_with_skiplist(struct ldlm_lock *lock)
 891 {
 892         struct sl_insert_point prev;
 893         ENTRY;
 894
 895         LASSERT(lock->l_req_mode == lock->l_granted_mode);
 896
 897         search_granted_lock(&lock->l_resource->lr_granted, lock, &prev);
 898         ldlm_granted_list_add_lock(lock, &prev);
 899         EXIT;
 900 }
 901
 902 /* NOTE: called by
 903  *  - ldlm_lock_enqueue
 904  *  - ldlm_reprocess_queue
 905  *  - ldlm_lock_convert
 906  *
 907  * must be called with lr_lock held
 908  */
 909 void ldlm_grant_lock(struct ldlm_lock *lock, cfs_list_t *work_list)
 910 {
 911         struct ldlm_resource *res = lock->l_resource;
 912         ENTRY;
 913
 914         check_res_locked(res);
 915
 916         lock->l_granted_mode = lock->l_req_mode;
 917         if (res->lr_type == LDLM_PLAIN || res->lr_type == LDLM_IBITS)
 918                 ldlm_grant_lock_with_skiplist(lock);
 919         else if (res->lr_type == LDLM_EXTENT)
 920                 ldlm_extent_add_lock(res, lock);
 921         else
 922                 ldlm_resource_add_lock(res, &res->lr_granted, lock);
 923
 924         if (lock->l_granted_mode < res->lr_most_restr)
 925                 res->lr_most_restr = lock->l_granted_mode;
 926
 927         if (work_list && lock->l_completion_ast != NULL)
 928                 ldlm_add_ast_work_item(lock, NULL, work_list);
 929
 930         ldlm_pool_add(&res->lr_namespace->ns_pool, lock);
 931         EXIT;
 932 }
 933
 934 /* returns a referenced lock or NULL.  See the flag descriptions below, in the
 935  * comment above ldlm_lock_match */
 936 static struct ldlm_lock *search_queue(cfs_list_t *queue,
 937                                       ldlm_mode_t *mode,
 938                                       ldlm_policy_data_t *policy,
 939                                       struct ldlm_lock *old_lock,
 940                                       int flags, int unref)
 941 {
 942         struct ldlm_lock *lock;
 943         cfs_list_t       *tmp;
 944
 945         cfs_list_for_each(tmp, queue) {
 946                 ldlm_mode_t match;
 947
 948                 lock = cfs_list_entry(tmp, struct ldlm_lock, l_res_link);
 949
 950                 if (lock == old_lock)
 951                         break;
 952
 953                 /* llite sometimes wants to match locks that will be
 954                  * canceled when their users drop, but we allow it to match
 955                  * if it passes in CBPENDING and the lock still has users.
 956                  * this is generally only going to be used by children
 957                  * whose parents already hold a lock so forward progress
 958                  * can still happen. */
 959                 if (lock->l_flags & LDLM_FL_CBPENDING &&
 960                     !(flags & LDLM_FL_CBPENDING))
 961                         continue;
 962                 if (!unref && lock->l_flags & LDLM_FL_CBPENDING &&
 963                     lock->l_readers == 0 && lock->l_writers == 0)
 964                         continue;
 965
 966                 if (!(lock->l_req_mode & *mode))
 967                         continue;
 968                 match = lock->l_req_mode;
 969
 970                 if (lock->l_resource->lr_type == LDLM_EXTENT &&
 971                     (lock->l_policy_data.l_extent.start >
 972                      policy->l_extent.start ||
 973                      lock->l_policy_data.l_extent.end < policy->l_extent.end))
 974                         continue;
 975
 976                 if (unlikely(match == LCK_GROUP) &&
 977                     lock->l_resource->lr_type == LDLM_EXTENT &&
 978                     lock->l_policy_data.l_extent.gid != policy->l_extent.gid)
 979                         continue;
 980
 981                 /* We match if we have existing lock with same or wider set
 982                    of bits. */
 983                 if (lock->l_resource->lr_type == LDLM_IBITS &&
 984                      ((lock->l_policy_data.l_inodebits.bits &
 985                       policy->l_inodebits.bits) !=
 986                       policy->l_inodebits.bits))
 987                         continue;
 988
 989                 if (!unref &&
 990                     (lock->l_destroyed || (lock->l_flags & LDLM_FL_FAILED)))
 991                         continue;
 992
 993                 if ((flags & LDLM_FL_LOCAL_ONLY) &&
 994                     !(lock->l_flags & LDLM_FL_LOCAL))
 995                         continue;
 996
 997                 if (flags & LDLM_FL_TEST_LOCK) {
 998                         LDLM_LOCK_GET(lock);
 999                         ldlm_lock_touch_in_lru(lock);
1000                 } else {
1001                         ldlm_lock_addref_internal_nolock(lock, match);
1002                 }
1003                 *mode = match;
1004                 return lock;
1005         }
1006
1007         return NULL;
1008 }
1009
1010 void ldlm_lock_allow_match_locked(struct ldlm_lock *lock)
1011 {
1012         lock->l_flags |= LDLM_FL_LVB_READY;
1013         cfs_waitq_signal(&lock->l_waitq);
1014 }
1015
1016 void ldlm_lock_allow_match(struct ldlm_lock *lock)
1017 {
1018         lock_res_and_lock(lock);
1019         ldlm_lock_allow_match_locked(lock);
1020         unlock_res_and_lock(lock);
1021 }
1022
1023 /* Can be called in two ways:
1024  *
1025  * If 'ns' is NULL, then lockh describes an existing lock that we want to look
1026  * for a duplicate of.
1027  *
1028  * Otherwise, all of the fields must be filled in, to match against.
1029  *
1030  * If 'flags' contains LDLM_FL_LOCAL_ONLY, then only match local locks on the
1031  *     server (ie, connh is NULL)
1032  * If 'flags' contains LDLM_FL_BLOCK_GRANTED, then only locks on the granted
1033  *     list will be considered
1034  * If 'flags' contains LDLM_FL_CBPENDING, then locks that have been marked
1035  *     to be canceled can still be matched as long as they still have reader
1036  *     or writer refernces
1037  * If 'flags' contains LDLM_FL_TEST_LOCK, then don't actually reference a lock,
1038  *     just tell us if we would have matched.
1039  *
1040  * Returns 1 if it finds an already-existing lock that is compatible; in this
1041  * case, lockh is filled in with a addref()ed lock
1042  *
1043  * we also check security context, if that failed we simply return 0 (to keep
1044  * caller code unchanged), the context failure will be discovered by caller
1045  * sometime later.
1046  */
1047 ldlm_mode_t ldlm_lock_match(struct ldlm_namespace *ns, int flags,
1048                             const struct ldlm_res_id *res_id, ldlm_type_t type,
1049                             ldlm_policy_data_t *policy, ldlm_mode_t mode,
1050                             struct lustre_handle *lockh, int unref)
1051 {
1052         struct ldlm_resource *res;
1053         struct ldlm_lock *lock, *old_lock = NULL;
1054         int rc = 0;
1055         ENTRY;
1056
1057         if (ns == NULL) {
1058                 old_lock = ldlm_handle2lock(lockh);
1059                 LASSERT(old_lock);
1060
1061                 ns = old_lock->l_resource->lr_namespace;
1062                 res_id = &old_lock->l_resource->lr_name;
1063                 type = old_lock->l_resource->lr_type;
1064                 mode = old_lock->l_req_mode;
1065         }
1066
1067         res = ldlm_resource_get(ns, NULL, res_id, type, 0);
1068         if (res == NULL) {
1069                 LASSERT(old_lock == NULL);
1070                 RETURN(0);
1071         }
1072
1073         LDLM_RESOURCE_ADDREF(res);
1074         lock_res(res);
1075
1076         lock = search_queue(&res->lr_granted, &mode, policy, old_lock,
1077                             flags, unref);
1078         if (lock != NULL)
1079                 GOTO(out, rc = 1);
1080         if (flags & LDLM_FL_BLOCK_GRANTED)
1081                 GOTO(out, rc = 0);
1082         lock = search_queue(&res->lr_converting, &mode, policy, old_lock,
1083                             flags, unref);
1084         if (lock != NULL)
1085                 GOTO(out, rc = 1);
1086         lock = search_queue(&res->lr_waiting, &mode, policy, old_lock,
1087                             flags, unref);
1088         if (lock != NULL)
1089                 GOTO(out, rc = 1);
1090
1091         EXIT;
1092  out:
1093         unlock_res(res);
1094         LDLM_RESOURCE_DELREF(res);
1095         ldlm_resource_putref(res);
1096
1097         if (lock) {
1098                 ldlm_lock2handle(lock, lockh);
1099                 if ((flags & LDLM_FL_LVB_READY) &&
1100                     (!(lock->l_flags & LDLM_FL_LVB_READY))) {
1101                         struct l_wait_info lwi;
1102                         if (lock->l_completion_ast) {
1103                                 int err = lock->l_completion_ast(lock,
1104                                                           LDLM_FL_WAIT_NOREPROC,
1105                                                                  NULL);
1106                                 if (err) {
1107                                         if (flags & LDLM_FL_TEST_LOCK)
1108                                                 LDLM_LOCK_RELEASE(lock);
1109                                         else
1110                                                 ldlm_lock_decref_internal(lock,
1111                                                                           mode);
1112                                         rc = 0;
1113                                         goto out2;
1114                                 }
1115                         }
1116
1117                         lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(obd_timeout),
1118                                                NULL, LWI_ON_SIGNAL_NOOP, NULL);
1119
1120                         /* XXX FIXME see comment on CAN_MATCH in lustre_dlm.h */
1121                         l_wait_event(lock->l_waitq,
1122                                      (lock->l_flags & LDLM_FL_LVB_READY), &lwi);
1123                 }
1124         }
1125  out2:
1126         if (rc) {
1127                 LDLM_DEBUG(lock, "matched ("LPU64" "LPU64")",
1128                            (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1129                                 res_id->name[2] : policy->l_extent.start,
1130                            (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1131                                 res_id->name[3] : policy->l_extent.end);
1132
1133                 /* check user's security context */
1134                 if (lock->l_conn_export &&
1135                     sptlrpc_import_check_ctx(
1136                                 class_exp2cliimp(lock->l_conn_export))) {
1137                         if (!(flags & LDLM_FL_TEST_LOCK))
1138                                 ldlm_lock_decref_internal(lock, mode);
1139                         rc = 0;
1140                 }
1141
1142                 if (flags & LDLM_FL_TEST_LOCK)
1143                         LDLM_LOCK_RELEASE(lock);
1144
1145         } else if (!(flags & LDLM_FL_TEST_LOCK)) {/*less verbose for test-only*/
1146                 LDLM_DEBUG_NOLOCK("not matched ns %p type %u mode %u res "
1147                                   LPU64"/"LPU64" ("LPU64" "LPU64")", ns,
1148                                   type, mode, res_id->name[0], res_id->name[1],
1149                                   (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1150                                         res_id->name[2] :policy->l_extent.start,
1151                                   (type == LDLM_PLAIN || type == LDLM_IBITS) ?
1152                                         res_id->name[3] : policy->l_extent.end);
1153         }
1154         if (old_lock)
1155                 LDLM_LOCK_PUT(old_lock);
1156
1157         return rc ? mode : 0;
1158 }
1159
1160 /* Returns a referenced lock */
1161 struct ldlm_lock *ldlm_lock_create(struct ldlm_namespace *ns,
1162                                    const struct ldlm_res_id *res_id,
1163                                    ldlm_type_t type,
1164                                    ldlm_mode_t mode,
1165                                    const struct ldlm_callback_suite *cbs,
1166                                    void *data, __u32 lvb_len)
1167 {
1168         struct ldlm_lock *lock;
1169         struct ldlm_resource *res;
1170         ENTRY;
1171
1172         res = ldlm_resource_get(ns, NULL, res_id, type, 1);
1173         if (res == NULL)
1174                 RETURN(NULL);
1175
1176         lock = ldlm_lock_new(res);
1177         ldlm_resource_putref(res);
1178
1179         if (lock == NULL)
1180                 RETURN(NULL);
1181
1182         lock->l_req_mode = mode;
1183         lock->l_ast_data = data;
1184         lock->l_pid = cfs_curproc_pid();
1185         if (cbs) {
1186                 lock->l_blocking_ast = cbs->lcs_blocking;
1187                 lock->l_completion_ast = cbs->lcs_completion;
1188                 lock->l_glimpse_ast = cbs->lcs_glimpse;
1189                 lock->l_weigh_ast = cbs->lcs_weigh;
1190         }
1191
1192         lock->l_tree_node = NULL;
1193         /* if this is the extent lock, allocate the interval tree node */
1194         if (type == LDLM_EXTENT) {
1195                 if (ldlm_interval_alloc(lock) == NULL)
1196                         GOTO(out, 0);
1197         }
1198
1199         if (lvb_len) {
1200                 lock->l_lvb_len = lvb_len;
1201                 OBD_ALLOC(lock->l_lvb_data, lvb_len);
1202                 if (lock->l_lvb_data == NULL)
1203                         GOTO(out, 0);
1204         }
1205
1206         if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_NEW_LOCK))
1207                 GOTO(out, 0);
1208
1209         RETURN(lock);
1210
1211 out:
1212         ldlm_lock_destroy(lock);
1213         LDLM_LOCK_RELEASE(lock);
1214         return NULL;
1215 }
1216
1217 ldlm_error_t ldlm_lock_enqueue(struct ldlm_namespace *ns,
1218                                struct ldlm_lock **lockp,
1219                                void *cookie, int *flags)
1220 {
1221         struct ldlm_lock *lock = *lockp;
1222         struct ldlm_resource *res = lock->l_resource;
1223         int local = ns_is_client(res->lr_namespace);
1224         ldlm_processing_policy policy;
1225         ldlm_error_t rc = ELDLM_OK;
1226         struct ldlm_interval *node = NULL;
1227         ENTRY;
1228
1229         lock->l_last_activity = cfs_time_current_sec();
1230         /* policies are not executed on the client or during replay */
1231         if ((*flags & (LDLM_FL_HAS_INTENT|LDLM_FL_REPLAY)) == LDLM_FL_HAS_INTENT
1232             && !local && ns->ns_policy) {
1233                 rc = ns->ns_policy(ns, lockp, cookie, lock->l_req_mode, *flags,
1234                                    NULL);
1235                 if (rc == ELDLM_LOCK_REPLACED) {
1236                         /* The lock that was returned has already been granted,
1237                          * and placed into lockp.  If it's not the same as the
1238                          * one we passed in, then destroy the old one and our
1239                          * work here is done. */
1240                         if (lock != *lockp) {
1241                                 ldlm_lock_destroy(lock);
1242                                 LDLM_LOCK_RELEASE(lock);
1243                         }
1244                         *flags |= LDLM_FL_LOCK_CHANGED;
1245                         RETURN(0);
1246                 } else if (rc != ELDLM_OK ||
1247                            (rc == ELDLM_OK && (*flags & LDLM_FL_INTENT_ONLY))) {
1248                         ldlm_lock_destroy(lock);
1249                         RETURN(rc);
1250                 }
1251         }
1252
1253         /* For a replaying lock, it might be already in granted list. So
1254          * unlinking the lock will cause the interval node to be freed, we
1255          * have to allocate the interval node early otherwise we can't regrant
1256          * this lock in the future. - jay */
1257         if (!local && (*flags & LDLM_FL_REPLAY) && res->lr_type == LDLM_EXTENT)
1258                 OBD_SLAB_ALLOC_PTR_GFP(node, ldlm_interval_slab, CFS_ALLOC_IO);
1259
1260         lock_res_and_lock(lock);
1261         if (local && lock->l_req_mode == lock->l_granted_mode) {
1262                 /* The server returned a blocked lock, but it was granted
1263                  * before we got a chance to actually enqueue it.  We don't
1264                  * need to do anything else. */
1265                 *flags &= ~(LDLM_FL_BLOCK_GRANTED |
1266                             LDLM_FL_BLOCK_CONV | LDLM_FL_BLOCK_WAIT);
1267                 GOTO(out, ELDLM_OK);
1268         }
1269
1270         ldlm_resource_unlink_lock(lock);
1271         if (res->lr_type == LDLM_EXTENT && lock->l_tree_node == NULL) {
1272                 if (node == NULL) {
1273                         ldlm_lock_destroy_nolock(lock);
1274                         GOTO(out, rc = -ENOMEM);
1275                 }
1276
1277                 CFS_INIT_LIST_HEAD(&node->li_group);
1278                 ldlm_interval_attach(node, lock);
1279                 node = NULL;
1280         }
1281
1282         /* Some flags from the enqueue want to make it into the AST, via the
1283          * lock's l_flags. */
1284         lock->l_flags |= *flags & LDLM_AST_DISCARD_DATA;
1285
1286         /* This distinction between local lock trees is very important; a client
1287          * namespace only has information about locks taken by that client, and
1288          * thus doesn't have enough information to decide for itself if it can
1289          * be granted (below).  In this case, we do exactly what the server
1290          * tells us to do, as dictated by the 'flags'.
1291          *
1292          * We do exactly the same thing during recovery, when the server is
1293          * more or less trusting the clients not to lie.
1294          *
1295          * FIXME (bug 268): Detect obvious lies by checking compatibility in
1296          * granted/converting queues. */
1297         if (local) {
1298                 if (*flags & LDLM_FL_BLOCK_CONV)
1299                         ldlm_resource_add_lock(res, &res->lr_converting, lock);
1300                 else if (*flags & (LDLM_FL_BLOCK_WAIT | LDLM_FL_BLOCK_GRANTED))
1301                         ldlm_resource_add_lock(res, &res->lr_waiting, lock);
1302                 else
1303                         ldlm_grant_lock(lock, NULL);
1304                 GOTO(out, ELDLM_OK);
1305         } else if (*flags & LDLM_FL_REPLAY) {
1306                 if (*flags & LDLM_FL_BLOCK_CONV) {
1307                         ldlm_resource_add_lock(res, &res->lr_converting, lock);
1308                         GOTO(out, ELDLM_OK);
1309                 } else if (*flags & LDLM_FL_BLOCK_WAIT) {
1310                         ldlm_resource_add_lock(res, &res->lr_waiting, lock);
1311                         GOTO(out, ELDLM_OK);
1312                 } else if (*flags & LDLM_FL_BLOCK_GRANTED) {
1313                         ldlm_grant_lock(lock, NULL);
1314                         GOTO(out, ELDLM_OK);
1315                 }
1316                 /* If no flags, fall through to normal enqueue path. */
1317         }
1318
1319         policy = ldlm_processing_policy_table[res->lr_type];
1320         policy(lock, flags, 1, &rc, NULL);
1321         GOTO(out, rc);
1322 out:
1323         unlock_res_and_lock(lock);
1324         if (node)
1325                 OBD_SLAB_FREE(node, ldlm_interval_slab, sizeof(*node));
1326         return rc;
1327 }
1328
1329 /* Must be called with namespace taken: queue is waiting or converting. */
1330 int ldlm_reprocess_queue(struct ldlm_resource *res, cfs_list_t *queue,
1331                          cfs_list_t *work_list)
1332 {
1333         cfs_list_t *tmp, *pos;
1334         ldlm_processing_policy policy;
1335         int flags;
1336         int rc = LDLM_ITER_CONTINUE;
1337         ldlm_error_t err;
1338         ENTRY;
1339
1340         check_res_locked(res);
1341
1342         policy = ldlm_processing_policy_table[res->lr_type];
1343         LASSERT(policy);
1344
1345         cfs_list_for_each_safe(tmp, pos, queue) {
1346                 struct ldlm_lock *pending;
1347                 pending = cfs_list_entry(tmp, struct ldlm_lock, l_res_link);
1348
1349                 CDEBUG(D_INFO, "Reprocessing lock %p\n", pending);
1350
1351                 flags = 0;
1352                 rc = policy(pending, &flags, 0, &err, work_list);
1353                 if (rc != LDLM_ITER_CONTINUE)
1354                         break;
1355         }
1356
1357         RETURN(rc);
1358 }
1359
1360 /* Helper function for ldlm_run_ast_work().
1361  *
1362  * Send an existing rpc set specified by @arg->set and then
1363  * destroy it. Create new one if @do_create flag is set. */
1364 static void
1365 ldlm_send_and_maybe_create_set(struct ldlm_cb_set_arg *arg, int do_create)
1366 {
1367         ENTRY;
1368
1369         ptlrpc_set_wait(arg->set);
1370         if (arg->type == LDLM_BL_CALLBACK)
1371                 OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_GLIMPSE, 2);
1372         ptlrpc_set_destroy(arg->set);
1373
1374         if (do_create)
1375                 arg->set = ptlrpc_prep_set();
1376
1377         EXIT;
1378 }
1379
1380 static int
1381 ldlm_work_bl_ast_lock(cfs_list_t *tmp, struct ldlm_cb_set_arg *arg)
1382 {
1383         struct ldlm_lock_desc d;
1384         struct ldlm_lock *lock = cfs_list_entry(tmp, struct ldlm_lock,
1385                                                 l_bl_ast);
1386         ENTRY;
1387
1388         /* nobody should touch l_bl_ast */
1389         lock_res_and_lock(lock);
1390         cfs_list_del_init(&lock->l_bl_ast);
1391
1392         LASSERT(lock->l_flags & LDLM_FL_AST_SENT);
1393         LASSERT(lock->l_bl_ast_run == 0);
1394         LASSERT(lock->l_blocking_lock);
1395         lock->l_bl_ast_run++;
1396         unlock_res_and_lock(lock);
1397
1398         ldlm_lock2desc(lock->l_blocking_lock, &d);
1399
1400         lock->l_blocking_ast(lock, &d, (void *)arg,
1401                              LDLM_CB_BLOCKING);
1402         LDLM_LOCK_RELEASE(lock->l_blocking_lock);
1403         lock->l_blocking_lock = NULL;
1404         LDLM_LOCK_RELEASE(lock);
1405
1406         RETURN(1);
1407 }
1408
1409 static int
1410 ldlm_work_cp_ast_lock(cfs_list_t *tmp, struct ldlm_cb_set_arg *arg)
1411 {
1412         struct ldlm_lock *lock = cfs_list_entry(tmp, struct ldlm_lock, l_cp_ast);
1413         ldlm_completion_callback completion_callback;
1414         int rc = 0;
1415         ENTRY;
1416
1417         /* It's possible to receive a completion AST before we've set
1418          * the l_completion_ast pointer: either because the AST arrived
1419          * before the reply, or simply because there's a small race
1420          * window between receiving the reply and finishing the local
1421          * enqueue. (bug 842)
1422          *
1423          * This can't happen with the blocking_ast, however, because we
1424          * will never call the local blocking_ast until we drop our
1425          * reader/writer reference, which we won't do until we get the
1426          * reply and finish enqueueing. */
1427
1428         /* nobody should touch l_cp_ast */
1429         lock_res_and_lock(lock);
1430         cfs_list_del_init(&lock->l_cp_ast);
1431         LASSERT(lock->l_flags & LDLM_FL_CP_REQD);
1432         /* save l_completion_ast since it can be changed by
1433          * mds_intent_policy(), see bug 14225 */
1434         completion_callback = lock->l_completion_ast;
1435         lock->l_flags &= ~LDLM_FL_CP_REQD;
1436         unlock_res_and_lock(lock);
1437
1438         if (completion_callback != NULL) {
1439                 completion_callback(lock, 0, (void *)arg);
1440                 rc = 1;
1441         }
1442         LDLM_LOCK_RELEASE(lock);
1443
1444         RETURN(rc);
1445 }
1446
1447 static int
1448 ldlm_work_revoke_ast_lock(cfs_list_t *tmp, struct ldlm_cb_set_arg *arg)
1449 {
1450         struct ldlm_lock_desc desc;
1451         struct ldlm_lock *lock = cfs_list_entry(tmp, struct ldlm_lock,
1452                                                 l_rk_ast);
1453         ENTRY;
1454
1455         cfs_list_del_init(&lock->l_rk_ast);
1456
1457         /* the desc just pretend to exclusive */
1458         ldlm_lock2desc(lock, &desc);
1459         desc.l_req_mode = LCK_EX;
1460         desc.l_granted_mode = 0;
1461
1462         lock->l_blocking_ast(lock, &desc, (void*)arg, LDLM_CB_BLOCKING);
1463         LDLM_LOCK_RELEASE(lock);
1464
1465         RETURN(1);
1466 }
1467
1468 int ldlm_run_ast_work(cfs_list_t *rpc_list, ldlm_desc_ast_t ast_type)
1469 {
1470         struct ldlm_cb_set_arg arg;
1471         cfs_list_t *tmp, *pos;
1472         int (*work_ast_lock)(cfs_list_t *tmp, struct ldlm_cb_set_arg *arg);
1473         int ast_count;
1474         ENTRY;
1475
1476         if (cfs_list_empty(rpc_list))
1477                 RETURN(0);
1478
1479         arg.set = ptlrpc_prep_set();
1480         if (NULL == arg.set)
1481                 RETURN(-ERESTART);
1482         cfs_atomic_set(&arg.restart, 0);
1483         switch (ast_type) {
1484         case LDLM_WORK_BL_AST:
1485                 arg.type = LDLM_BL_CALLBACK;
1486                 work_ast_lock = ldlm_work_bl_ast_lock;
1487                 break;
1488         case LDLM_WORK_CP_AST:
1489                 arg.type = LDLM_CP_CALLBACK;
1490                 work_ast_lock = ldlm_work_cp_ast_lock;
1491                 break;
1492         case LDLM_WORK_REVOKE_AST:
1493                 arg.type = LDLM_BL_CALLBACK;
1494                 work_ast_lock = ldlm_work_revoke_ast_lock;
1495                 break;
1496         default:
1497                 LBUG();
1498         }
1499
1500         ast_count = 0;
1501         cfs_list_for_each_safe(tmp, pos, rpc_list) {
1502                 ast_count += work_ast_lock(tmp, &arg);
1503
1504                 /* Send the request set if it exceeds the PARALLEL_AST_LIMIT,
1505                  * and create a new set for requests that remained in
1506                  * @rpc_list */
1507                 if (unlikely(ast_count == PARALLEL_AST_LIMIT)) {
1508                         ldlm_send_and_maybe_create_set(&arg, 1);
1509                         ast_count = 0;
1510                 }
1511         }
1512
1513         if (ast_count > 0)
1514                 ldlm_send_and_maybe_create_set(&arg, 0);
1515         else
1516                 /* In case when number of ASTs is multiply of
1517                  * PARALLEL_AST_LIMIT or @rpc_list was initially empty,
1518                  * @arg.set must be destroyed here, otherwise we get
1519                  * write memory leaking. */
1520                 ptlrpc_set_destroy(arg.set);
1521
1522         RETURN(cfs_atomic_read(&arg.restart) ? -ERESTART : 0);
1523 }
1524
1525 static int reprocess_one_queue(struct ldlm_resource *res, void *closure)
1526 {
1527         ldlm_reprocess_all(res);
1528         return LDLM_ITER_CONTINUE;
1529 }
1530
1531 void ldlm_reprocess_all_ns(struct ldlm_namespace *ns)
1532 {
1533         cfs_list_t *tmp;
1534         int i, rc;
1535
1536         if (ns == NULL)
1537                 return;
1538
1539         ENTRY;
1540         cfs_spin_lock(&ns->ns_hash_lock);
1541         for (i = 0; i < RES_HASH_SIZE; i++) {
1542                 tmp = ns->ns_hash[i].next;
1543                 while (tmp != &(ns->ns_hash[i])) {
1544                         struct ldlm_resource *res =
1545                                 cfs_list_entry(tmp, struct ldlm_resource,
1546                                                lr_hash);
1547
1548                         ldlm_resource_getref(res);
1549                         cfs_spin_unlock(&ns->ns_hash_lock);
1550                         LDLM_RESOURCE_ADDREF(res);
1551
1552                         rc = reprocess_one_queue(res, NULL);
1553
1554                         LDLM_RESOURCE_DELREF(res);
1555                         cfs_spin_lock(&ns->ns_hash_lock);
1556                         tmp = tmp->next;
1557                         ldlm_resource_putref_locked(res);
1558
1559                         if (rc == LDLM_ITER_STOP)
1560                                 GOTO(out, rc);
1561                 }
1562         }
1563  out:
1564         cfs_spin_unlock(&ns->ns_hash_lock);
1565         EXIT;
1566 }
1567
1568 void ldlm_reprocess_all(struct ldlm_resource *res)
1569 {
1570         CFS_LIST_HEAD(rpc_list);
1571         int rc;
1572         ENTRY;
1573
1574         /* Local lock trees don't get reprocessed. */
1575         if (ns_is_client(res->lr_namespace)) {
1576                 EXIT;
1577                 return;
1578         }
1579
1580  restart:
1581         lock_res(res);
1582         rc = ldlm_reprocess_queue(res, &res->lr_converting, &rpc_list);
1583         if (rc == LDLM_ITER_CONTINUE)
1584                 ldlm_reprocess_queue(res, &res->lr_waiting, &rpc_list);
1585         unlock_res(res);
1586
1587         rc = ldlm_run_ast_work(&rpc_list, LDLM_WORK_CP_AST);
1588         if (rc == -ERESTART) {
1589                 LASSERT(cfs_list_empty(&rpc_list));
1590                 goto restart;
1591         }
1592         EXIT;
1593 }
1594
1595 void ldlm_cancel_callback(struct ldlm_lock *lock)
1596 {
1597         check_res_locked(lock->l_resource);
1598         if (!(lock->l_flags & LDLM_FL_CANCEL)) {
1599                 lock->l_flags |= LDLM_FL_CANCEL;
1600                 if (lock->l_blocking_ast) {
1601                         // l_check_no_ns_lock(ns);
1602                         unlock_res_and_lock(lock);
1603                         lock->l_blocking_ast(lock, NULL, lock->l_ast_data,
1604                                              LDLM_CB_CANCELING);
1605                         lock_res_and_lock(lock);
1606                 } else {
1607                         LDLM_DEBUG(lock, "no blocking ast");
1608                 }
1609         }
1610         lock->l_flags |= LDLM_FL_BL_DONE;
1611 }
1612
1613 void ldlm_unlink_lock_skiplist(struct ldlm_lock *req)
1614 {
1615         if (req->l_resource->lr_type != LDLM_PLAIN &&
1616             req->l_resource->lr_type != LDLM_IBITS)
1617                 return;
1618
1619         cfs_list_del_init(&req->l_sl_policy);
1620         cfs_list_del_init(&req->l_sl_mode);
1621 }
1622
1623 void ldlm_lock_cancel(struct ldlm_lock *lock)
1624 {
1625         struct ldlm_resource *res;
1626         struct ldlm_namespace *ns;
1627         ENTRY;
1628
1629         lock_res_and_lock(lock);
1630
1631         res = lock->l_resource;
1632         ns = res->lr_namespace;
1633
1634         /* Please do not, no matter how tempting, remove this LBUG without
1635          * talking to me first. -phik */
1636         if (lock->l_readers || lock->l_writers) {
1637                 LDLM_ERROR(lock, "lock still has references");
1638                 LBUG();
1639         }
1640
1641         ldlm_del_waiting_lock(lock);
1642
1643         /* Releases cancel callback. */
1644         ldlm_cancel_callback(lock);
1645
1646         /* Yes, second time, just in case it was added again while we were
1647            running with no res lock in ldlm_cancel_callback */
1648         ldlm_del_waiting_lock(lock);
1649         ldlm_resource_unlink_lock(lock);
1650         ldlm_lock_destroy_nolock(lock);
1651
1652         if (lock->l_granted_mode == lock->l_req_mode)
1653                 ldlm_pool_del(&ns->ns_pool, lock);
1654
1655         /* Make sure we will not be called again for same lock what is possible
1656          * if not to zero out lock->l_granted_mode */
1657         lock->l_granted_mode = LCK_MINMODE;
1658         unlock_res_and_lock(lock);
1659
1660         EXIT;
1661 }
1662
1663 int ldlm_lock_set_data(struct lustre_handle *lockh, void *data)
1664 {
1665         struct ldlm_lock *lock = ldlm_handle2lock(lockh);
1666         ENTRY;
1667
1668         if (lock == NULL)
1669                 RETURN(-EINVAL);
1670
1671         lock->l_ast_data = data;
1672         LDLM_LOCK_PUT(lock);
1673         RETURN(0);
1674 }
1675
1676 int ldlm_cancel_locks_for_export_cb(cfs_hash_t *hs, cfs_hash_bd_t *bd,
1677                                     cfs_hlist_node_t *hnode, void *data)
1678
1679 {
1680         struct obd_export    *exp  = data;
1681         struct ldlm_lock     *lock = cfs_hash_object(hs, hnode);
1682         struct ldlm_resource *res;
1683
1684         res = ldlm_resource_getref(lock->l_resource);
1685         LDLM_LOCK_GET(lock);
1686
1687         LDLM_DEBUG(lock, "export %p", exp);
1688         ldlm_res_lvbo_update(res, NULL, 1);
1689         ldlm_lock_cancel(lock);
1690         ldlm_reprocess_all(res);
1691         ldlm_resource_putref(res);
1692         LDLM_LOCK_RELEASE(lock);
1693         return 0;
1694 }
1695
1696 void ldlm_cancel_locks_for_export(struct obd_export *exp)
1697 {
1698         cfs_hash_for_each_empty(exp->exp_lock_hash,
1699                                 ldlm_cancel_locks_for_export_cb, exp);
1700 }
1701
1702 /**
1703  * Downgrade an exclusive lock.
1704  *
1705  * A fast variant of ldlm_lock_convert for convertion of exclusive
1706  * locks. The convertion is always successful.
1707  *
1708  * \param lock A lock to convert
1709  * \param new_mode new lock mode
1710  */
1711 void ldlm_lock_downgrade(struct ldlm_lock *lock, int new_mode)
1712 {
1713         struct ldlm_namespace *ns;
1714         ENTRY;
1715
1716         LASSERT(lock->l_granted_mode & (LCK_PW | LCK_EX));
1717         LASSERT(new_mode == LCK_COS);
1718
1719         lock_res_and_lock(lock);
1720         ldlm_resource_unlink_lock(lock);
1721         /*
1722          * Remove the lock from pool as it will be added again in
1723          * ldlm_grant_lock() called below.
1724          */
1725         ns = lock->l_resource->lr_namespace;
1726         ldlm_pool_del(&ns->ns_pool, lock);
1727
1728         lock->l_req_mode = new_mode;
1729         ldlm_grant_lock(lock, NULL);
1730         unlock_res_and_lock(lock);
1731         ldlm_reprocess_all(lock->l_resource);
1732
1733         EXIT;
1734 }
1735
1736 struct ldlm_resource *ldlm_lock_convert(struct ldlm_lock *lock, int new_mode,
1737                                         __u32 *flags)
1738 {
1739         CFS_LIST_HEAD(rpc_list);
1740         struct ldlm_resource *res;
1741         struct ldlm_namespace *ns;
1742         int granted = 0;
1743         int old_mode, rc;
1744         struct sl_insert_point prev;
1745         ldlm_error_t err;
1746         struct ldlm_interval *node;
1747         ENTRY;
1748
1749         if (new_mode == lock->l_granted_mode) { // No changes? Just return.
1750                 *flags |= LDLM_FL_BLOCK_GRANTED;
1751                 RETURN(lock->l_resource);
1752         }
1753
1754         /* I can't check the type of lock here because the bitlock of lock
1755          * is not held here, so do the allocation blindly. -jay */
1756         OBD_SLAB_ALLOC_PTR_GFP(node, ldlm_interval_slab, CFS_ALLOC_IO);
1757         if (node == NULL)  /* Actually, this causes EDEADLOCK to be returned */
1758                 RETURN(NULL);
1759
1760         LASSERTF((new_mode == LCK_PW && lock->l_granted_mode == LCK_PR),
1761                  "new_mode %u, granted %u\n", new_mode, lock->l_granted_mode);
1762
1763         lock_res_and_lock(lock);
1764
1765         res = lock->l_resource;
1766         ns = res->lr_namespace;
1767
1768         old_mode = lock->l_req_mode;
1769         lock->l_req_mode = new_mode;
1770         if (res->lr_type == LDLM_PLAIN || res->lr_type == LDLM_IBITS) {
1771                 /* remember the lock position where the lock might be
1772                  * added back to the granted list later and also
1773                  * remember the join mode for skiplist fixing. */
1774                 prev.res_link = lock->l_res_link.prev;
1775                 prev.mode_link = lock->l_sl_mode.prev;
1776                 prev.policy_link = lock->l_sl_policy.prev;
1777                 ldlm_resource_unlink_lock(lock);
1778         } else {
1779                 ldlm_resource_unlink_lock(lock);
1780                 if (res->lr_type == LDLM_EXTENT) {
1781                         /* FIXME: ugly code, I have to attach the lock to a
1782                          * interval node again since perhaps it will be granted
1783                          * soon */
1784                         CFS_INIT_LIST_HEAD(&node->li_group);
1785                         ldlm_interval_attach(node, lock);
1786                         node = NULL;
1787                 }
1788         }
1789
1790         /*
1791          * Remove old lock from the pool before adding the lock with new
1792          * mode below in ->policy()
1793          */
1794         ldlm_pool_del(&ns->ns_pool, lock);
1795
1796         /* If this is a local resource, put it on the appropriate list. */
1797         if (ns_is_client(res->lr_namespace)) {
1798                 if (*flags & (LDLM_FL_BLOCK_CONV | LDLM_FL_BLOCK_GRANTED)) {
1799                         ldlm_resource_add_lock(res, &res->lr_converting, lock);
1800                 } else {
1801                         /* This should never happen, because of the way the
1802                          * server handles conversions. */
1803                         LDLM_ERROR(lock, "Erroneous flags %d on local lock\n",
1804                                    *flags);
1805                         LBUG();
1806
1807                         ldlm_grant_lock(lock, &rpc_list);
1808                         granted = 1;
1809                         /* FIXME: completion handling not with ns_lock held ! */
1810                         if (lock->l_completion_ast)
1811                                 lock->l_completion_ast(lock, 0, NULL);
1812                 }
1813         } else {
1814                 int pflags = 0;
1815                 ldlm_processing_policy policy;
1816                 policy = ldlm_processing_policy_table[res->lr_type];
1817                 rc = policy(lock, &pflags, 0, &err, &rpc_list);
1818                 if (rc == LDLM_ITER_STOP) {
1819                         lock->l_req_mode = old_mode;
1820                         if (res->lr_type == LDLM_EXTENT)
1821                                 ldlm_extent_add_lock(res, lock);
1822                         else
1823                                 ldlm_granted_list_add_lock(lock, &prev);
1824
1825                         res = NULL;
1826                 } else {
1827                         *flags |= LDLM_FL_BLOCK_GRANTED;
1828                         granted = 1;
1829                 }
1830         }
1831         unlock_res_and_lock(lock);
1832
1833         if (granted)
1834                 ldlm_run_ast_work(&rpc_list, LDLM_WORK_CP_AST);
1835         if (node)
1836                 OBD_SLAB_FREE(node, ldlm_interval_slab, sizeof(*node));
1837         RETURN(res);
1838 }
1839
1840 void ldlm_lock_dump(int level, struct ldlm_lock *lock, int pos)
1841 {
1842         struct obd_device *obd = NULL;
1843
1844         if (!((libcfs_debug | D_ERROR) & level))
1845                 return;
1846
1847         if (!lock) {
1848                 CDEBUG(level, "  NULL LDLM lock\n");
1849                 return;
1850         }
1851
1852         CDEBUG(level," -- Lock dump: %p/"LPX64" (rc: %d) (pos: %d) (pid: %d)\n",
1853                lock, lock->l_handle.h_cookie, cfs_atomic_read(&lock->l_refc),
1854                pos, lock->l_pid);
1855         if (lock->l_conn_export != NULL)
1856                 obd = lock->l_conn_export->exp_obd;
1857         if (lock->l_export && lock->l_export->exp_connection) {
1858                 CDEBUG(level, "  Node: NID %s (rhandle: "LPX64")\n",
1859                      libcfs_nid2str(lock->l_export->exp_connection->c_peer.nid),
1860                      lock->l_remote_handle.cookie);
1861         } else if (obd == NULL) {
1862                 CDEBUG(level, "  Node: local\n");
1863         } else {
1864                 struct obd_import *imp = obd->u.cli.cl_import;
1865                 CDEBUG(level, "  Node: NID %s (rhandle: "LPX64")\n",
1866                        libcfs_nid2str(imp->imp_connection->c_peer.nid),
1867                        lock->l_remote_handle.cookie);
1868         }
1869         CDEBUG(level, "  Resource: %p ("LPU64"/"LPU64"/"LPU64")\n",
1870                   lock->l_resource,
1871                   lock->l_resource->lr_name.name[0],
1872                   lock->l_resource->lr_name.name[1],
1873                   lock->l_resource->lr_name.name[2]);
1874         CDEBUG(level, "  Req mode: %s, grant mode: %s, rc: %u, read: %d, "
1875                "write: %d flags: "LPX64"\n", ldlm_lockname[lock->l_req_mode],
1876                ldlm_lockname[lock->l_granted_mode],
1877                cfs_atomic_read(&lock->l_refc), lock->l_readers, lock->l_writers,
1878                lock->l_flags);
1879         if (lock->l_resource->lr_type == LDLM_EXTENT)
1880                 CDEBUG(level, "  Extent: "LPU64" -> "LPU64
1881                        " (req "LPU64"-"LPU64")\n",
1882                        lock->l_policy_data.l_extent.start,
1883                        lock->l_policy_data.l_extent.end,
1884                        lock->l_req_extent.start, lock->l_req_extent.end);
1885         else if (lock->l_resource->lr_type == LDLM_FLOCK)
1886                 CDEBUG(level, "  Pid: %d Extent: "LPU64" -> "LPU64"\n",
1887                        lock->l_policy_data.l_flock.pid,
1888                        lock->l_policy_data.l_flock.start,
1889                        lock->l_policy_data.l_flock.end);
1890        else if (lock->l_resource->lr_type == LDLM_IBITS)
1891                 CDEBUG(level, "  Bits: "LPX64"\n",
1892                        lock->l_policy_data.l_inodebits.bits);
1893 }
1894
1895 void ldlm_lock_dump_handle(int level, struct lustre_handle *lockh)
1896 {
1897         struct ldlm_lock *lock;
1898
1899         if (!((libcfs_debug | D_ERROR) & level))
1900                 return;
1901
1902         lock = ldlm_handle2lock(lockh);
1903         if (lock == NULL)
1904                 return;
1905
1906         ldlm_lock_dump(D_OTHER, lock, 0);
1907
1908         LDLM_LOCK_PUT(lock);
1909 }
1910
1911 void _ldlm_lock_debug(struct ldlm_lock *lock, __u32 level,
1912                       struct libcfs_debug_msg_data *data, const char *fmt,
1913                       ...)
1914 {
1915         va_list args;
1916         cfs_debug_limit_state_t *cdls = data->msg_cdls;
1917
1918         va_start(args, fmt);
1919
1920         if (lock->l_resource == NULL) {
1921                 libcfs_debug_vmsg2(cdls, data->msg_subsys, level,data->msg_file,
1922                                    data->msg_fn, data->msg_line, fmt, args,
1923                        " ns: \?\? lock: %p/"LPX64" lrc: %d/%d,%d mode: %s/%s "
1924                        "res: \?\? rrc=\?\? type: \?\?\? flags: "LPX64" remote: "
1925                        LPX64" expref: %d pid: %u timeout: %lu\n", lock,
1926                        lock->l_handle.h_cookie, cfs_atomic_read(&lock->l_refc),
1927                        lock->l_readers, lock->l_writers,
1928                        ldlm_lockname[lock->l_granted_mode],
1929                        ldlm_lockname[lock->l_req_mode],
1930                        lock->l_flags, lock->l_remote_handle.cookie,
1931                        lock->l_export ?
1932                        cfs_atomic_read(&lock->l_export->exp_refcount) : -99,
1933                        lock->l_pid, lock->l_callback_timeout);
1934                 va_end(args);
1935                 return;
1936         }
1937
1938         switch (lock->l_resource->lr_type) {
1939         case LDLM_EXTENT:
1940                 libcfs_debug_vmsg2(cdls, data->msg_subsys, level,data->msg_file,
1941                                    data->msg_fn, data->msg_line, fmt, args,
1942                        " ns: %s lock: %p/"LPX64" lrc: %d/%d,%d mode: %s/%s "
1943                        "res: "LPU64"/"LPU64" rrc: %d type: %s ["LPU64"->"LPU64
1944                        "] (req "LPU64"->"LPU64") flags: "LPX64" remote: "LPX64
1945                        " expref: %d pid: %u timeout %lu\n",
1946                        lock->l_resource->lr_namespace->ns_name, lock,
1947                        lock->l_handle.h_cookie, cfs_atomic_read(&lock->l_refc),
1948                        lock->l_readers, lock->l_writers,
1949                        ldlm_lockname[lock->l_granted_mode],
1950                        ldlm_lockname[lock->l_req_mode],
1951                        lock->l_resource->lr_name.name[0],
1952                        lock->l_resource->lr_name.name[1],
1953                        cfs_atomic_read(&lock->l_resource->lr_refcount),
1954                        ldlm_typename[lock->l_resource->lr_type],
1955                        lock->l_policy_data.l_extent.start,
1956                        lock->l_policy_data.l_extent.end,
1957                        lock->l_req_extent.start, lock->l_req_extent.end,
1958                        lock->l_flags, lock->l_remote_handle.cookie,
1959                        lock->l_export ?
1960                        cfs_atomic_read(&lock->l_export->exp_refcount) : -99,
1961                        lock->l_pid, lock->l_callback_timeout);
1962                 break;
1963
1964         case LDLM_FLOCK:
1965                 libcfs_debug_vmsg2(cdls, data->msg_subsys, level,data->msg_file,
1966                                    data->msg_fn, data->msg_line, fmt, args,
1967                        " ns: %s lock: %p/"LPX64" lrc: %d/%d,%d mode: %s/%s "
1968                        "res: "LPU64"/"LPU64" rrc: %d type: %s pid: %d "
1969                        "["LPU64"->"LPU64"] flags: "LPX64" remote: "LPX64
1970                        " expref: %d pid: %u timeout: %lu\n",
1971                        lock->l_resource->lr_namespace->ns_name, lock,
1972                        lock->l_handle.h_cookie, cfs_atomic_read(&lock->l_refc),
1973                        lock->l_readers, lock->l_writers,
1974                        ldlm_lockname[lock->l_granted_mode],
1975                        ldlm_lockname[lock->l_req_mode],
1976                        lock->l_resource->lr_name.name[0],
1977                        lock->l_resource->lr_name.name[1],
1978                        cfs_atomic_read(&lock->l_resource->lr_refcount),
1979                        ldlm_typename[lock->l_resource->lr_type],
1980                        lock->l_policy_data.l_flock.pid,
1981                        lock->l_policy_data.l_flock.start,
1982                        lock->l_policy_data.l_flock.end,
1983                        lock->l_flags, lock->l_remote_handle.cookie,
1984                        lock->l_export ?
1985                        cfs_atomic_read(&lock->l_export->exp_refcount) : -99,
1986                        lock->l_pid, lock->l_callback_timeout);
1987                 break;
1988
1989         case LDLM_IBITS:
1990                 libcfs_debug_vmsg2(cdls, data->msg_subsys, level,data->msg_file,
1991                                    data->msg_fn, data->msg_line, fmt, args,
1992                        " ns: %s lock: %p/"LPX64" lrc: %d/%d,%d mode: %s/%s "
1993                        "res: "LPU64"/"LPU64" bits "LPX64" rrc: %d type: %s "
1994                        "flags: "LPX64" remote: "LPX64" expref: %d "
1995                        "pid: %u timeout: %lu\n",
1996                        lock->l_resource->lr_namespace->ns_name,
1997                        lock, lock->l_handle.h_cookie,
1998                        cfs_atomic_read (&lock->l_refc),
1999                        lock->l_readers, lock->l_writers,
2000                        ldlm_lockname[lock->l_granted_mode],
2001                        ldlm_lockname[lock->l_req_mode],
2002                        lock->l_resource->lr_name.name[0],
2003                        lock->l_resource->lr_name.name[1],
2004                        lock->l_policy_data.l_inodebits.bits,
2005                        cfs_atomic_read(&lock->l_resource->lr_refcount),
2006                        ldlm_typename[lock->l_resource->lr_type],
2007                        lock->l_flags, lock->l_remote_handle.cookie,
2008                        lock->l_export ?
2009                        cfs_atomic_read(&lock->l_export->exp_refcount) : -99,
2010                        lock->l_pid, lock->l_callback_timeout);
2011                 break;
2012
2013         default:
2014                 libcfs_debug_vmsg2(cdls, data->msg_subsys, level,data->msg_file,
2015                                    data->msg_fn, data->msg_line, fmt, args,
2016                        " ns: %s lock: %p/"LPX64" lrc: %d/%d,%d mode: %s/%s "
2017                        "res: "LPU64"/"LPU64" rrc: %d type: %s flags: "LPX64" "
2018                        "remote: "LPX64" expref: %d pid: %u timeout %lu\n",
2019                        lock->l_resource->lr_namespace->ns_name,
2020                        lock, lock->l_handle.h_cookie,
2021                        cfs_atomic_read (&lock->l_refc),
2022                        lock->l_readers, lock->l_writers,
2023                        ldlm_lockname[lock->l_granted_mode],
2024                        ldlm_lockname[lock->l_req_mode],
2025                        lock->l_resource->lr_name.name[0],
2026                        lock->l_resource->lr_name.name[1],
2027                        cfs_atomic_read(&lock->l_resource->lr_refcount),
2028                        ldlm_typename[lock->l_resource->lr_type],
2029                        lock->l_flags, lock->l_remote_handle.cookie,
2030                        lock->l_export ?
2031                        cfs_atomic_read(&lock->l_export->exp_refcount) : -99,
2032                        lock->l_pid, lock->l_callback_timeout);
2033                 break;
2034         }
2035         va_end(args);
2036 }
2037 EXPORT_SYMBOL(_ldlm_lock_debug);