kernel/drivers/staging/lustre/lustre/osc/osc_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) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2011, 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  * Implementation of cl_lock for OSC layer.
  37  *
  38  *   Author: Nikita Danilov <nikita.danilov@sun.com>
  39  */
  40
  41 #define DEBUG_SUBSYSTEM S_OSC
  42
  43 #include "../../include/linux/libcfs/libcfs.h"
  44 /* fid_build_reg_res_name() */
  45 #include "../include/lustre_fid.h"
  46
  47 #include "osc_cl_internal.h"
  48
  49 /** \addtogroup osc
  50  *  @{
  51  */
  52
  53 #define _PAGEREF_MAGIC  (-10000000)
  54
  55 /*****************************************************************************
  56  *
  57  * Type conversions.
  58  *
  59  */
  60
  61 static const struct cl_lock_operations osc_lock_ops;
  62 static const struct cl_lock_operations osc_lock_lockless_ops;
  63 static void osc_lock_to_lockless(const struct lu_env *env,
  64                                  struct osc_lock *ols, int force);
  65 static int osc_lock_has_pages(struct osc_lock *olck);
  66
  67 int osc_lock_is_lockless(const struct osc_lock *olck)
  68 {
  69         return (olck->ols_cl.cls_ops == &osc_lock_lockless_ops);
  70 }
  71
  72 /**
  73  * Returns a weak pointer to the ldlm lock identified by a handle. Returned
  74  * pointer cannot be dereferenced, as lock is not protected from concurrent
  75  * reclaim. This function is a helper for osc_lock_invariant().
  76  */
  77 static struct ldlm_lock *osc_handle_ptr(struct lustre_handle *handle)
  78 {
  79         struct ldlm_lock *lock;
  80
  81         lock = ldlm_handle2lock(handle);
  82         if (lock != NULL)
  83                 LDLM_LOCK_PUT(lock);
  84         return lock;
  85 }
  86
  87 /**
  88  * Invariant that has to be true all of the time.
  89  */
  90 static int osc_lock_invariant(struct osc_lock *ols)
  91 {
  92         struct ldlm_lock *lock        = osc_handle_ptr(&ols->ols_handle);
  93         struct ldlm_lock *olock       = ols->ols_lock;
  94         int               handle_used = lustre_handle_is_used(&ols->ols_handle);
  95
  96         if (ergo(osc_lock_is_lockless(ols),
  97                  ols->ols_locklessable && ols->ols_lock == NULL))
  98                 return 1;
  99
 100         /*
 101          * If all the following "ergo"s are true, return 1, otherwise 0
 102          */
 103         if (!ergo(olock != NULL, handle_used))
 104                 return 0;
 105
 106         if (!ergo(olock != NULL,
 107                    olock->l_handle.h_cookie == ols->ols_handle.cookie))
 108                 return 0;
 109
 110         if (!ergo(handle_used,
 111                    ergo(lock != NULL && olock != NULL, lock == olock) &&
 112                    ergo(lock == NULL, olock == NULL)))
 113                 return 0;
 114         /*
 115          * Check that ->ols_handle and ->ols_lock are consistent, but
 116          * take into account that they are set at the different time.
 117          */
 118         if (!ergo(ols->ols_state == OLS_CANCELLED,
 119                    olock == NULL && !handle_used))
 120                 return 0;
 121         /*
 122          * DLM lock is destroyed only after we have seen cancellation
 123          * ast.
 124          */
 125         if (!ergo(olock != NULL && ols->ols_state < OLS_CANCELLED,
 126                    ((olock->l_flags & LDLM_FL_DESTROYED) == 0)))
 127                 return 0;
 128
 129         if (!ergo(ols->ols_state == OLS_GRANTED,
 130                    olock != NULL &&
 131                    olock->l_req_mode == olock->l_granted_mode &&
 132                    ols->ols_hold))
 133                 return 0;
 134         return 1;
 135 }
 136
 137 /*****************************************************************************
 138  *
 139  * Lock operations.
 140  *
 141  */
 142
 143 /**
 144  * Breaks a link between osc_lock and dlm_lock.
 145  */
 146 static void osc_lock_detach(const struct lu_env *env, struct osc_lock *olck)
 147 {
 148         struct ldlm_lock *dlmlock;
 149
 150         spin_lock(&osc_ast_guard);
 151         dlmlock = olck->ols_lock;
 152         if (dlmlock == NULL) {
 153                 spin_unlock(&osc_ast_guard);
 154                 return;
 155         }
 156
 157         olck->ols_lock = NULL;
 158         /* wb(); --- for all who checks (ols->ols_lock != NULL) before
 159          * call to osc_lock_detach() */
 160         dlmlock->l_ast_data = NULL;
 161         olck->ols_handle.cookie = 0ULL;
 162         spin_unlock(&osc_ast_guard);
 163
 164         lock_res_and_lock(dlmlock);
 165         if (dlmlock->l_granted_mode == dlmlock->l_req_mode) {
 166                 struct cl_object *obj = olck->ols_cl.cls_obj;
 167                 struct cl_attr *attr  = &osc_env_info(env)->oti_attr;
 168                 __u64 old_kms;
 169
 170                 cl_object_attr_lock(obj);
 171                 /* Must get the value under the lock to avoid possible races. */
 172                 old_kms = cl2osc(obj)->oo_oinfo->loi_kms;
 173                 /* Update the kms. Need to loop all granted locks.
 174                  * Not a problem for the client */
 175                 attr->cat_kms = ldlm_extent_shift_kms(dlmlock, old_kms);
 176
 177                 cl_object_attr_set(env, obj, attr, CAT_KMS);
 178                 cl_object_attr_unlock(obj);
 179         }
 180         unlock_res_and_lock(dlmlock);
 181
 182         /* release a reference taken in osc_lock_upcall0(). */
 183         LASSERT(olck->ols_has_ref);
 184         lu_ref_del(&dlmlock->l_reference, "osc_lock", olck);
 185         LDLM_LOCK_RELEASE(dlmlock);
 186         olck->ols_has_ref = 0;
 187 }
 188
 189 static int osc_lock_unhold(struct osc_lock *ols)
 190 {
 191         int result = 0;
 192
 193         if (ols->ols_hold) {
 194                 ols->ols_hold = 0;
 195                 result = osc_cancel_base(&ols->ols_handle,
 196                                          ols->ols_einfo.ei_mode);
 197         }
 198         return result;
 199 }
 200
 201 static int osc_lock_unuse(const struct lu_env *env,
 202                           const struct cl_lock_slice *slice)
 203 {
 204         struct osc_lock *ols = cl2osc_lock(slice);
 205
 206         LINVRNT(osc_lock_invariant(ols));
 207
 208         switch (ols->ols_state) {
 209         case OLS_NEW:
 210                 LASSERT(!ols->ols_hold);
 211                 LASSERT(ols->ols_agl);
 212                 return 0;
 213         case OLS_UPCALL_RECEIVED:
 214                 osc_lock_unhold(ols);
 215         case OLS_ENQUEUED:
 216                 LASSERT(!ols->ols_hold);
 217                 osc_lock_detach(env, ols);
 218                 ols->ols_state = OLS_NEW;
 219                 return 0;
 220         case OLS_GRANTED:
 221                 LASSERT(!ols->ols_glimpse);
 222                 LASSERT(ols->ols_hold);
 223                 /*
 224                  * Move lock into OLS_RELEASED state before calling
 225                  * osc_cancel_base() so that possible synchronous cancellation
 226                  * (that always happens e.g., for liblustre) sees that lock is
 227                  * released.
 228                  */
 229                 ols->ols_state = OLS_RELEASED;
 230                 return osc_lock_unhold(ols);
 231         default:
 232                 CERROR("Impossible state: %d\n", ols->ols_state);
 233                 LBUG();
 234         }
 235 }
 236
 237 static void osc_lock_fini(const struct lu_env *env,
 238                           struct cl_lock_slice *slice)
 239 {
 240         struct osc_lock  *ols = cl2osc_lock(slice);
 241
 242         LINVRNT(osc_lock_invariant(ols));
 243         /*
 244          * ->ols_hold can still be true at this point if, for example, a
 245          * thread that requested a lock was killed (and released a reference
 246          * to the lock), before reply from a server was received. In this case
 247          * lock is destroyed immediately after upcall.
 248          */
 249         osc_lock_unhold(ols);
 250         LASSERT(ols->ols_lock == NULL);
 251         LASSERT(atomic_read(&ols->ols_pageref) == 0 ||
 252                 atomic_read(&ols->ols_pageref) == _PAGEREF_MAGIC);
 253
 254         OBD_SLAB_FREE_PTR(ols, osc_lock_kmem);
 255 }
 256
 257 static void osc_lock_build_policy(const struct lu_env *env,
 258                                   const struct cl_lock *lock,
 259                                   ldlm_policy_data_t *policy)
 260 {
 261         const struct cl_lock_descr *d = &lock->cll_descr;
 262
 263         osc_index2policy(policy, d->cld_obj, d->cld_start, d->cld_end);
 264         policy->l_extent.gid = d->cld_gid;
 265 }
 266
 267 static __u64 osc_enq2ldlm_flags(__u32 enqflags)
 268 {
 269         __u64 result = 0;
 270
 271         LASSERT((enqflags & ~CEF_MASK) == 0);
 272
 273         if (enqflags & CEF_NONBLOCK)
 274                 result |= LDLM_FL_BLOCK_NOWAIT;
 275         if (enqflags & CEF_ASYNC)
 276                 result |= LDLM_FL_HAS_INTENT;
 277         if (enqflags & CEF_DISCARD_DATA)
 278                 result |= LDLM_FL_AST_DISCARD_DATA;
 279         return result;
 280 }
 281
 282 /**
 283  * Global spin-lock protecting consistency of ldlm_lock::l_ast_data
 284  * pointers. Initialized in osc_init().
 285  */
 286 spinlock_t osc_ast_guard;
 287
 288 static struct osc_lock *osc_ast_data_get(struct ldlm_lock *dlm_lock)
 289 {
 290         struct osc_lock *olck;
 291
 292         lock_res_and_lock(dlm_lock);
 293         spin_lock(&osc_ast_guard);
 294         olck = dlm_lock->l_ast_data;
 295         if (olck != NULL) {
 296                 struct cl_lock *lock = olck->ols_cl.cls_lock;
 297                 /*
 298                  * If osc_lock holds a reference on ldlm lock, return it even
 299                  * when cl_lock is in CLS_FREEING state. This way
 300                  *
 301                  *       osc_ast_data_get(dlmlock) == NULL
 302                  *
 303                  * guarantees that all osc references on dlmlock were
 304                  * released. osc_dlm_blocking_ast0() relies on that.
 305                  */
 306                 if (lock->cll_state < CLS_FREEING || olck->ols_has_ref) {
 307                         cl_lock_get_trust(lock);
 308                         lu_ref_add_atomic(&lock->cll_reference,
 309                                           "ast", current);
 310                 } else
 311                         olck = NULL;
 312         }
 313         spin_unlock(&osc_ast_guard);
 314         unlock_res_and_lock(dlm_lock);
 315         return olck;
 316 }
 317
 318 static void osc_ast_data_put(const struct lu_env *env, struct osc_lock *olck)
 319 {
 320         struct cl_lock *lock;
 321
 322         lock = olck->ols_cl.cls_lock;
 323         lu_ref_del(&lock->cll_reference, "ast", current);
 324         cl_lock_put(env, lock);
 325 }
 326
 327 /**
 328  * Updates object attributes from a lock value block (lvb) received together
 329  * with the DLM lock reply from the server. Copy of osc_update_enqueue()
 330  * logic.
 331  *
 332  * This can be optimized to not update attributes when lock is a result of a
 333  * local match.
 334  *
 335  * Called under lock and resource spin-locks.
 336  */
 337 static void osc_lock_lvb_update(const struct lu_env *env, struct osc_lock *olck,
 338                                 int rc)
 339 {
 340         struct ost_lvb    *lvb;
 341         struct cl_object  *obj;
 342         struct lov_oinfo  *oinfo;
 343         struct cl_attr    *attr;
 344         unsigned           valid;
 345
 346         if (!(olck->ols_flags & LDLM_FL_LVB_READY))
 347                 return;
 348
 349         lvb   = &olck->ols_lvb;
 350         obj   = olck->ols_cl.cls_obj;
 351         oinfo = cl2osc(obj)->oo_oinfo;
 352         attr  = &osc_env_info(env)->oti_attr;
 353         valid = CAT_BLOCKS | CAT_ATIME | CAT_CTIME | CAT_MTIME | CAT_SIZE;
 354         cl_lvb2attr(attr, lvb);
 355
 356         cl_object_attr_lock(obj);
 357         if (rc == 0) {
 358                 struct ldlm_lock  *dlmlock;
 359                 __u64 size;
 360
 361                 dlmlock = olck->ols_lock;
 362                 LASSERT(dlmlock != NULL);
 363
 364                 /* re-grab LVB from a dlm lock under DLM spin-locks. */
 365                 *lvb = *(struct ost_lvb *)dlmlock->l_lvb_data;
 366                 size = lvb->lvb_size;
 367                 /* Extend KMS up to the end of this lock and no further
 368                  * A lock on [x,y] means a KMS of up to y + 1 bytes! */
 369                 if (size > dlmlock->l_policy_data.l_extent.end)
 370                         size = dlmlock->l_policy_data.l_extent.end + 1;
 371                 if (size >= oinfo->loi_kms) {
 372                         LDLM_DEBUG(dlmlock, "lock acquired, setting rss=%llu, kms=%llu",
 373                                    lvb->lvb_size, size);
 374                         valid |= CAT_KMS;
 375                         attr->cat_kms = size;
 376                 } else {
 377                         LDLM_DEBUG(dlmlock, "lock acquired, setting rss=%llu; leaving kms=%llu, end=%llu",
 378                                    lvb->lvb_size, oinfo->loi_kms,
 379                                    dlmlock->l_policy_data.l_extent.end);
 380                 }
 381                 ldlm_lock_allow_match_locked(dlmlock);
 382         } else if (rc == -ENAVAIL && olck->ols_glimpse) {
 383                 CDEBUG(D_INODE, "glimpsed, setting rss=%llu; leaving kms=%llu\n",
 384                        lvb->lvb_size, oinfo->loi_kms);
 385         } else
 386                 valid = 0;
 387
 388         if (valid != 0)
 389                 cl_object_attr_set(env, obj, attr, valid);
 390
 391         cl_object_attr_unlock(obj);
 392 }
 393
 394 /**
 395  * Called when a lock is granted, from an upcall (when server returned a
 396  * granted lock), or from completion AST, when server returned a blocked lock.
 397  *
 398  * Called under lock and resource spin-locks, that are released temporarily
 399  * here.
 400  */
 401 static void osc_lock_granted(const struct lu_env *env, struct osc_lock *olck,
 402                              struct ldlm_lock *dlmlock, int rc)
 403 {
 404         struct ldlm_extent   *ext;
 405         struct cl_lock       *lock;
 406         struct cl_lock_descr *descr;
 407
 408         LASSERT(dlmlock->l_granted_mode == dlmlock->l_req_mode);
 409
 410         if (olck->ols_state < OLS_GRANTED) {
 411                 lock  = olck->ols_cl.cls_lock;
 412                 ext   = &dlmlock->l_policy_data.l_extent;
 413                 descr = &osc_env_info(env)->oti_descr;
 414                 descr->cld_obj = lock->cll_descr.cld_obj;
 415
 416                 /* XXX check that ->l_granted_mode is valid. */
 417                 descr->cld_mode  = osc_ldlm2cl_lock(dlmlock->l_granted_mode);
 418                 descr->cld_start = cl_index(descr->cld_obj, ext->start);
 419                 descr->cld_end   = cl_index(descr->cld_obj, ext->end);
 420                 descr->cld_gid   = ext->gid;
 421                 /*
 422                  * tell upper layers the extent of the lock that was actually
 423                  * granted
 424                  */
 425                 olck->ols_state = OLS_GRANTED;
 426                 osc_lock_lvb_update(env, olck, rc);
 427
 428                 /* release DLM spin-locks to allow cl_lock_{modify,signal}()
 429                  * to take a semaphore on a parent lock. This is safe, because
 430                  * spin-locks are needed to protect consistency of
 431                  * dlmlock->l_*_mode and LVB, and we have finished processing
 432                  * them. */
 433                 unlock_res_and_lock(dlmlock);
 434                 cl_lock_modify(env, lock, descr);
 435                 cl_lock_signal(env, lock);
 436                 LINVRNT(osc_lock_invariant(olck));
 437                 lock_res_and_lock(dlmlock);
 438         }
 439 }
 440
 441 static void osc_lock_upcall0(const struct lu_env *env, struct osc_lock *olck)
 442
 443 {
 444         struct ldlm_lock *dlmlock;
 445
 446         dlmlock = ldlm_handle2lock_long(&olck->ols_handle, 0);
 447         LASSERT(dlmlock != NULL);
 448
 449         lock_res_and_lock(dlmlock);
 450         spin_lock(&osc_ast_guard);
 451         LASSERT(dlmlock->l_ast_data == olck);
 452         LASSERT(olck->ols_lock == NULL);
 453         olck->ols_lock = dlmlock;
 454         spin_unlock(&osc_ast_guard);
 455
 456         /*
 457          * Lock might be not yet granted. In this case, completion ast
 458          * (osc_ldlm_completion_ast()) comes later and finishes lock
 459          * granting.
 460          */
 461         if (dlmlock->l_granted_mode == dlmlock->l_req_mode)
 462                 osc_lock_granted(env, olck, dlmlock, 0);
 463         unlock_res_and_lock(dlmlock);
 464
 465         /*
 466          * osc_enqueue_interpret() decrefs asynchronous locks, counter
 467          * this.
 468          */
 469         ldlm_lock_addref(&olck->ols_handle, olck->ols_einfo.ei_mode);
 470         olck->ols_hold = 1;
 471
 472         /* lock reference taken by ldlm_handle2lock_long() is owned by
 473          * osc_lock and released in osc_lock_detach() */
 474         lu_ref_add(&dlmlock->l_reference, "osc_lock", olck);
 475         olck->ols_has_ref = 1;
 476 }
 477
 478 /**
 479  * Lock upcall function that is executed either when a reply to ENQUEUE rpc is
 480  * received from a server, or after osc_enqueue_base() matched a local DLM
 481  * lock.
 482  */
 483 static int osc_lock_upcall(void *cookie, int errcode)
 484 {
 485         struct osc_lock  *olck  = cookie;
 486         struct cl_lock_slice    *slice = &olck->ols_cl;
 487         struct cl_lock    *lock  = slice->cls_lock;
 488         struct lu_env      *env;
 489         struct cl_env_nest       nest;
 490
 491         env = cl_env_nested_get(&nest);
 492         if (!IS_ERR(env)) {
 493                 int rc;
 494
 495                 cl_lock_mutex_get(env, lock);
 496
 497                 LASSERT(lock->cll_state >= CLS_QUEUING);
 498                 if (olck->ols_state == OLS_ENQUEUED) {
 499                         olck->ols_state = OLS_UPCALL_RECEIVED;
 500                         rc = ldlm_error2errno(errcode);
 501                 } else if (olck->ols_state == OLS_CANCELLED) {
 502                         rc = -EIO;
 503                 } else {
 504                         CERROR("Impossible state: %d\n", olck->ols_state);
 505                         LBUG();
 506                 }
 507                 if (rc) {
 508                         struct ldlm_lock *dlmlock;
 509
 510                         dlmlock = ldlm_handle2lock(&olck->ols_handle);
 511                         if (dlmlock != NULL) {
 512                                 lock_res_and_lock(dlmlock);
 513                                 spin_lock(&osc_ast_guard);
 514                                 LASSERT(olck->ols_lock == NULL);
 515                                 dlmlock->l_ast_data = NULL;
 516                                 olck->ols_handle.cookie = 0ULL;
 517                                 spin_unlock(&osc_ast_guard);
 518                                 ldlm_lock_fail_match_locked(dlmlock);
 519                                 unlock_res_and_lock(dlmlock);
 520                                 LDLM_LOCK_PUT(dlmlock);
 521                         }
 522                 } else {
 523                         if (olck->ols_glimpse)
 524                                 olck->ols_glimpse = 0;
 525                         osc_lock_upcall0(env, olck);
 526                 }
 527
 528                 /* Error handling, some errors are tolerable. */
 529                 if (olck->ols_locklessable && rc == -EUSERS) {
 530                         /* This is a tolerable error, turn this lock into
 531                          * lockless lock.
 532                          */
 533                         osc_object_set_contended(cl2osc(slice->cls_obj));
 534                         LASSERT(slice->cls_ops == &osc_lock_ops);
 535
 536                         /* Change this lock to ldlmlock-less lock. */
 537                         osc_lock_to_lockless(env, olck, 1);
 538                         olck->ols_state = OLS_GRANTED;
 539                         rc = 0;
 540                 } else if (olck->ols_glimpse && rc == -ENAVAIL) {
 541                         osc_lock_lvb_update(env, olck, rc);
 542                         cl_lock_delete(env, lock);
 543                         /* Hide the error. */
 544                         rc = 0;
 545                 }
 546
 547                 if (rc == 0) {
 548                         /* For AGL case, the RPC sponsor may exits the cl_lock
 549                         *  processing without wait() called before related OSC
 550                         *  lock upcall(). So update the lock status according
 551                         *  to the enqueue result inside AGL upcall(). */
 552                         if (olck->ols_agl) {
 553                                 lock->cll_flags |= CLF_FROM_UPCALL;
 554                                 cl_wait_try(env, lock);
 555                                 lock->cll_flags &= ~CLF_FROM_UPCALL;
 556                                 if (!olck->ols_glimpse)
 557                                         olck->ols_agl = 0;
 558                         }
 559                         cl_lock_signal(env, lock);
 560                         /* del user for lock upcall cookie */
 561                         cl_unuse_try(env, lock);
 562                 } else {
 563                         /* del user for lock upcall cookie */
 564                         cl_lock_user_del(env, lock);
 565                         cl_lock_error(env, lock, rc);
 566                 }
 567
 568                 /* release cookie reference, acquired by osc_lock_enqueue() */
 569                 cl_lock_hold_release(env, lock, "upcall", lock);
 570                 cl_lock_mutex_put(env, lock);
 571
 572                 lu_ref_del(&lock->cll_reference, "upcall", lock);
 573                 /* This maybe the last reference, so must be called after
 574                  * cl_lock_mutex_put(). */
 575                 cl_lock_put(env, lock);
 576
 577                 cl_env_nested_put(&nest, env);
 578         } else {
 579                 /* should never happen, similar to osc_ldlm_blocking_ast(). */
 580                 LBUG();
 581         }
 582         return errcode;
 583 }
 584
 585 /**
 586  * Core of osc_dlm_blocking_ast() logic.
 587  */
 588 static void osc_lock_blocking(const struct lu_env *env,
 589                               struct ldlm_lock *dlmlock,
 590                               struct osc_lock *olck, int blocking)
 591 {
 592         struct cl_lock *lock = olck->ols_cl.cls_lock;
 593
 594         LASSERT(olck->ols_lock == dlmlock);
 595         CLASSERT(OLS_BLOCKED < OLS_CANCELLED);
 596         LASSERT(!osc_lock_is_lockless(olck));
 597
 598         /*
 599          * Lock might be still addref-ed here, if e.g., blocking ast
 600          * is sent for a failed lock.
 601          */
 602         osc_lock_unhold(olck);
 603
 604         if (blocking && olck->ols_state < OLS_BLOCKED)
 605                 /*
 606                  * Move osc_lock into OLS_BLOCKED before canceling the lock,
 607                  * because it recursively re-enters osc_lock_blocking(), with
 608                  * the state set to OLS_CANCELLED.
 609                  */
 610                 olck->ols_state = OLS_BLOCKED;
 611         /*
 612          * cancel and destroy lock at least once no matter how blocking ast is
 613          * entered (see comment above osc_ldlm_blocking_ast() for use
 614          * cases). cl_lock_cancel() and cl_lock_delete() are idempotent.
 615          */
 616         cl_lock_cancel(env, lock);
 617         cl_lock_delete(env, lock);
 618 }
 619
 620 /**
 621  * Helper for osc_dlm_blocking_ast() handling discrepancies between cl_lock
 622  * and ldlm_lock caches.
 623  */
 624 static int osc_dlm_blocking_ast0(const struct lu_env *env,
 625                                  struct ldlm_lock *dlmlock,
 626                                  void *data, int flag)
 627 {
 628         struct osc_lock *olck;
 629         struct cl_lock  *lock;
 630         int result;
 631         int cancel;
 632
 633         LASSERT(flag == LDLM_CB_BLOCKING || flag == LDLM_CB_CANCELING);
 634
 635         cancel = 0;
 636         olck = osc_ast_data_get(dlmlock);
 637         if (olck != NULL) {
 638                 lock = olck->ols_cl.cls_lock;
 639                 cl_lock_mutex_get(env, lock);
 640                 LINVRNT(osc_lock_invariant(olck));
 641                 if (olck->ols_ast_wait) {
 642                         /* wake up osc_lock_use() */
 643                         cl_lock_signal(env, lock);
 644                         olck->ols_ast_wait = 0;
 645                 }
 646                 /*
 647                  * Lock might have been canceled while this thread was
 648                  * sleeping for lock mutex, but olck is pinned in memory.
 649                  */
 650                 if (olck == dlmlock->l_ast_data) {
 651                         /*
 652                          * NOTE: DLM sends blocking AST's for failed locks
 653                          *       (that are still in pre-OLS_GRANTED state)
 654                          *       too, and they have to be canceled otherwise
 655                          *       DLM lock is never destroyed and stuck in
 656                          *       the memory.
 657                          *
 658                          *       Alternatively, ldlm_cli_cancel() can be
 659                          *       called here directly for osc_locks with
 660                          *       ols_state < OLS_GRANTED to maintain an
 661                          *       invariant that ->clo_cancel() is only called
 662                          *       for locks that were granted.
 663                          */
 664                         LASSERT(data == olck);
 665                         osc_lock_blocking(env, dlmlock,
 666                                           olck, flag == LDLM_CB_BLOCKING);
 667                 } else
 668                         cancel = 1;
 669                 cl_lock_mutex_put(env, lock);
 670                 osc_ast_data_put(env, olck);
 671         } else
 672                 /*
 673                  * DLM lock exists, but there is no cl_lock attached to it.
 674                  * This is a `normal' race. cl_object and its cl_lock's can be
 675                  * removed by memory pressure, together with all pages.
 676                  */
 677                 cancel = (flag == LDLM_CB_BLOCKING);
 678
 679         if (cancel) {
 680                 struct lustre_handle *lockh;
 681
 682                 lockh = &osc_env_info(env)->oti_handle;
 683                 ldlm_lock2handle(dlmlock, lockh);
 684                 result = ldlm_cli_cancel(lockh, LCF_ASYNC);
 685         } else
 686                 result = 0;
 687         return result;
 688 }
 689
 690 /**
 691  * Blocking ast invoked by ldlm when dlm lock is either blocking progress of
 692  * some other lock, or is canceled. This function is installed as a
 693  * ldlm_lock::l_blocking_ast() for client extent locks.
 694  *
 695  * Control flow is tricky, because ldlm uses the same call-back
 696  * (ldlm_lock::l_blocking_ast()) for both blocking and cancellation ast's.
 697  *
 698  * \param dlmlock lock for which ast occurred.
 699  *
 700  * \param new description of a conflicting lock in case of blocking ast.
 701  *
 702  * \param data value of dlmlock->l_ast_data
 703  *
 704  * \param flag LDLM_CB_BLOCKING or LDLM_CB_CANCELING. Used to distinguish
 705  *           cancellation and blocking ast's.
 706  *
 707  * Possible use cases:
 708  *
 709  *     - ldlm calls dlmlock->l_blocking_ast(..., LDLM_CB_CANCELING) to cancel
 710  *       lock due to lock lru pressure, or explicit user request to purge
 711  *       locks.
 712  *
 713  *     - ldlm calls dlmlock->l_blocking_ast(..., LDLM_CB_BLOCKING) to notify
 714  *       us that dlmlock conflicts with another lock that some client is
 715  *       enqueing. Lock is canceled.
 716  *
 717  *         - cl_lock_cancel() is called. osc_lock_cancel() calls
 718  *           ldlm_cli_cancel() that calls
 719  *
 720  *                dlmlock->l_blocking_ast(..., LDLM_CB_CANCELING)
 721  *
 722  *           recursively entering osc_ldlm_blocking_ast().
 723  *
 724  *     - client cancels lock voluntary (e.g., as a part of early cancellation):
 725  *
 726  *         cl_lock_cancel()->
 727  *           osc_lock_cancel()->
 728  *             ldlm_cli_cancel()->
 729  *               dlmlock->l_blocking_ast(..., LDLM_CB_CANCELING)
 730  *
 731  */
 732 static int osc_ldlm_blocking_ast(struct ldlm_lock *dlmlock,
 733                                  struct ldlm_lock_desc *new, void *data,
 734                                  int flag)
 735 {
 736         struct lu_env     *env;
 737         struct cl_env_nest nest;
 738         int             result;
 739
 740         /*
 741          * This can be called in the context of outer IO, e.g.,
 742          *
 743          *     cl_enqueue()->...
 744          *       ->osc_enqueue_base()->...
 745          *       ->ldlm_prep_elc_req()->...
 746          *         ->ldlm_cancel_callback()->...
 747          *           ->osc_ldlm_blocking_ast()
 748          *
 749          * new environment has to be created to not corrupt outer context.
 750          */
 751         env = cl_env_nested_get(&nest);
 752         if (!IS_ERR(env)) {
 753                 result = osc_dlm_blocking_ast0(env, dlmlock, data, flag);
 754                 cl_env_nested_put(&nest, env);
 755         } else {
 756                 result = PTR_ERR(env);
 757                 /*
 758                  * XXX This should never happen, as cl_lock is
 759                  * stuck. Pre-allocated environment a la vvp_inode_fini_env
 760                  * should be used.
 761                  */
 762                 LBUG();
 763         }
 764         if (result != 0) {
 765                 if (result == -ENODATA)
 766                         result = 0;
 767                 else
 768                         CERROR("BAST failed: %d\n", result);
 769         }
 770         return result;
 771 }
 772
 773 static int osc_ldlm_completion_ast(struct ldlm_lock *dlmlock,
 774                                    __u64 flags, void *data)
 775 {
 776         struct cl_env_nest nest;
 777         struct lu_env     *env;
 778         struct osc_lock   *olck;
 779         struct cl_lock    *lock;
 780         int result;
 781         int dlmrc;
 782
 783         /* first, do dlm part of the work */
 784         dlmrc = ldlm_completion_ast_async(dlmlock, flags, data);
 785         /* then, notify cl_lock */
 786         env = cl_env_nested_get(&nest);
 787         if (!IS_ERR(env)) {
 788                 olck = osc_ast_data_get(dlmlock);
 789                 if (olck != NULL) {
 790                         lock = olck->ols_cl.cls_lock;
 791                         cl_lock_mutex_get(env, lock);
 792                         /*
 793                          * ldlm_handle_cp_callback() copied LVB from request
 794                          * to lock->l_lvb_data, store it in osc_lock.
 795                          */
 796                         LASSERT(dlmlock->l_lvb_data != NULL);
 797                         lock_res_and_lock(dlmlock);
 798                         olck->ols_lvb = *(struct ost_lvb *)dlmlock->l_lvb_data;
 799                         if (olck->ols_lock == NULL) {
 800                                 /*
 801                                  * upcall (osc_lock_upcall()) hasn't yet been
 802                                  * called. Do nothing now, upcall will bind
 803                                  * olck to dlmlock and signal the waiters.
 804                                  *
 805                                  * This maintains an invariant that osc_lock
 806                                  * and ldlm_lock are always bound when
 807                                  * osc_lock is in OLS_GRANTED state.
 808                                  */
 809                         } else if (dlmlock->l_granted_mode ==
 810                                    dlmlock->l_req_mode) {
 811                                 osc_lock_granted(env, olck, dlmlock, dlmrc);
 812                         }
 813                         unlock_res_and_lock(dlmlock);
 814
 815                         if (dlmrc != 0) {
 816                                 CL_LOCK_DEBUG(D_ERROR, env, lock,
 817                                               "dlmlock returned %d\n", dlmrc);
 818                                 cl_lock_error(env, lock, dlmrc);
 819                         }
 820                         cl_lock_mutex_put(env, lock);
 821                         osc_ast_data_put(env, olck);
 822                         result = 0;
 823                 } else
 824                         result = -ELDLM_NO_LOCK_DATA;
 825                 cl_env_nested_put(&nest, env);
 826         } else
 827                 result = PTR_ERR(env);
 828         return dlmrc ?: result;
 829 }
 830
 831 static int osc_ldlm_glimpse_ast(struct ldlm_lock *dlmlock, void *data)
 832 {
 833         struct ptlrpc_request  *req  = data;
 834         struct osc_lock *olck;
 835         struct cl_lock   *lock;
 836         struct cl_object       *obj;
 837         struct cl_env_nest      nest;
 838         struct lu_env     *env;
 839         struct ost_lvb   *lvb;
 840         struct req_capsule     *cap;
 841         int                  result;
 842
 843         LASSERT(lustre_msg_get_opc(req->rq_reqmsg) == LDLM_GL_CALLBACK);
 844
 845         env = cl_env_nested_get(&nest);
 846         if (!IS_ERR(env)) {
 847                 /* osc_ast_data_get() has to go after environment is
 848                  * allocated, because osc_ast_data() acquires a
 849                  * reference to a lock, and it can only be released in
 850                  * environment.
 851                  */
 852                 olck = osc_ast_data_get(dlmlock);
 853                 if (olck != NULL) {
 854                         lock = olck->ols_cl.cls_lock;
 855                         /* Do not grab the mutex of cl_lock for glimpse.
 856                          * See LU-1274 for details.
 857                          * BTW, it's okay for cl_lock to be cancelled during
 858                          * this period because server can handle this race.
 859                          * See ldlm_server_glimpse_ast() for details.
 860                          * cl_lock_mutex_get(env, lock); */
 861                         cap = &req->rq_pill;
 862                         req_capsule_extend(cap, &RQF_LDLM_GL_CALLBACK);
 863                         req_capsule_set_size(cap, &RMF_DLM_LVB, RCL_SERVER,
 864                                              sizeof(*lvb));
 865                         result = req_capsule_server_pack(cap);
 866                         if (result == 0) {
 867                                 lvb = req_capsule_server_get(cap, &RMF_DLM_LVB);
 868                                 obj = lock->cll_descr.cld_obj;
 869                                 result = cl_object_glimpse(env, obj, lvb);
 870                         }
 871                         if (!exp_connect_lvb_type(req->rq_export))
 872                                 req_capsule_shrink(&req->rq_pill,
 873                                                    &RMF_DLM_LVB,
 874                                                    sizeof(struct ost_lvb_v1),
 875                                                    RCL_SERVER);
 876                         osc_ast_data_put(env, olck);
 877                 } else {
 878                         /*
 879                          * These errors are normal races, so we don't want to
 880                          * fill the console with messages by calling
 881                          * ptlrpc_error()
 882                          */
 883                         lustre_pack_reply(req, 1, NULL, NULL);
 884                         result = -ELDLM_NO_LOCK_DATA;
 885                 }
 886                 cl_env_nested_put(&nest, env);
 887         } else
 888                 result = PTR_ERR(env);
 889         req->rq_status = result;
 890         return result;
 891 }
 892
 893 static unsigned long osc_lock_weigh(const struct lu_env *env,
 894                                     const struct cl_lock_slice *slice)
 895 {
 896         /*
 897          * don't need to grab coh_page_guard since we don't care the exact #
 898          * of pages..
 899          */
 900         return cl_object_header(slice->cls_obj)->coh_pages;
 901 }
 902
 903 static void osc_lock_build_einfo(const struct lu_env *env,
 904                                  const struct cl_lock *clock,
 905                                  struct osc_lock *lock,
 906                                  struct ldlm_enqueue_info *einfo)
 907 {
 908         enum cl_lock_mode mode;
 909
 910         mode = clock->cll_descr.cld_mode;
 911         if (mode == CLM_PHANTOM)
 912                 /*
 913                  * For now, enqueue all glimpse locks in read mode. In the
 914                  * future, client might choose to enqueue LCK_PW lock for
 915                  * glimpse on a file opened for write.
 916                  */
 917                 mode = CLM_READ;
 918
 919         einfo->ei_type   = LDLM_EXTENT;
 920         einfo->ei_mode   = osc_cl_lock2ldlm(mode);
 921         einfo->ei_cb_bl  = osc_ldlm_blocking_ast;
 922         einfo->ei_cb_cp  = osc_ldlm_completion_ast;
 923         einfo->ei_cb_gl  = osc_ldlm_glimpse_ast;
 924         einfo->ei_cbdata = lock; /* value to be put into ->l_ast_data */
 925 }
 926
 927 /**
 928  * Determine if the lock should be converted into a lockless lock.
 929  *
 930  * Steps to check:
 931  * - if the lock has an explicit requirement for a non-lockless lock;
 932  * - if the io lock request type ci_lockreq;
 933  * - send the enqueue rpc to ost to make the further decision;
 934  * - special treat to truncate lockless lock
 935  *
 936  *  Additional policy can be implemented here, e.g., never do lockless-io
 937  *  for large extents.
 938  */
 939 static void osc_lock_to_lockless(const struct lu_env *env,
 940                                  struct osc_lock *ols, int force)
 941 {
 942         struct cl_lock_slice *slice = &ols->ols_cl;
 943
 944         LASSERT(ols->ols_state == OLS_NEW ||
 945                 ols->ols_state == OLS_UPCALL_RECEIVED);
 946
 947         if (force) {
 948                 ols->ols_locklessable = 1;
 949                 slice->cls_ops = &osc_lock_lockless_ops;
 950         } else {
 951                 struct osc_io *oio     = osc_env_io(env);
 952                 struct cl_io  *io      = oio->oi_cl.cis_io;
 953                 struct cl_object *obj  = slice->cls_obj;
 954                 struct osc_object *oob = cl2osc(obj);
 955                 const struct osc_device *osd = lu2osc_dev(obj->co_lu.lo_dev);
 956                 struct obd_connect_data *ocd;
 957
 958                 LASSERT(io->ci_lockreq == CILR_MANDATORY ||
 959                         io->ci_lockreq == CILR_MAYBE ||
 960                         io->ci_lockreq == CILR_NEVER);
 961
 962                 ocd = &class_exp2cliimp(osc_export(oob))->imp_connect_data;
 963                 ols->ols_locklessable = (io->ci_type != CIT_SETATTR) &&
 964                                 (io->ci_lockreq == CILR_MAYBE) &&
 965                                 (ocd->ocd_connect_flags & OBD_CONNECT_SRVLOCK);
 966                 if (io->ci_lockreq == CILR_NEVER ||
 967                         /* lockless IO */
 968                     (ols->ols_locklessable && osc_object_is_contended(oob)) ||
 969                         /* lockless truncate */
 970                     (cl_io_is_trunc(io) &&
 971                      (ocd->ocd_connect_flags & OBD_CONNECT_TRUNCLOCK) &&
 972                       osd->od_lockless_truncate)) {
 973                         ols->ols_locklessable = 1;
 974                         slice->cls_ops = &osc_lock_lockless_ops;
 975                 }
 976         }
 977         LASSERT(ergo(ols->ols_glimpse, !osc_lock_is_lockless(ols)));
 978 }
 979
 980 static int osc_lock_compatible(const struct osc_lock *qing,
 981                                const struct osc_lock *qed)
 982 {
 983         enum cl_lock_mode qing_mode;
 984         enum cl_lock_mode qed_mode;
 985
 986         qing_mode = qing->ols_cl.cls_lock->cll_descr.cld_mode;
 987         if (qed->ols_glimpse &&
 988             (qed->ols_state >= OLS_UPCALL_RECEIVED || qing_mode == CLM_READ))
 989                 return 1;
 990
 991         qed_mode = qed->ols_cl.cls_lock->cll_descr.cld_mode;
 992         return ((qing_mode == CLM_READ) && (qed_mode == CLM_READ));
 993 }
 994
 995 /**
 996  * Cancel all conflicting locks and wait for them to be destroyed.
 997  *
 998  * This function is used for two purposes:
 999  *
1000  *     - early cancel all conflicting locks before starting IO, and
1001  *
1002  *     - guarantee that pages added to the page cache by lockless IO are never
1003  *       covered by locks other than lockless IO lock, and, hence, are not
1004  *       visible to other threads.
1005  */
1006 static int osc_lock_enqueue_wait(const struct lu_env *env,
1007                                  const struct osc_lock *olck)
1008 {
1009         struct cl_lock    *lock    = olck->ols_cl.cls_lock;
1010         struct cl_lock_descr    *descr   = &lock->cll_descr;
1011         struct cl_object_header *hdr     = cl_object_header(descr->cld_obj);
1012         struct cl_lock    *scan;
1013         struct cl_lock    *conflict = NULL;
1014         int lockless                 = osc_lock_is_lockless(olck);
1015         int rc                     = 0;
1016
1017         LASSERT(cl_lock_is_mutexed(lock));
1018
1019         /* make it enqueue anyway for glimpse lock, because we actually
1020          * don't need to cancel any conflicting locks. */
1021         if (olck->ols_glimpse)
1022                 return 0;
1023
1024         spin_lock(&hdr->coh_lock_guard);
1025         list_for_each_entry(scan, &hdr->coh_locks, cll_linkage) {
1026                 struct cl_lock_descr *cld = &scan->cll_descr;
1027                 const struct osc_lock *scan_ols;
1028
1029                 if (scan == lock)
1030                         break;
1031
1032                 if (scan->cll_state < CLS_QUEUING ||
1033                     scan->cll_state == CLS_FREEING ||
1034                     cld->cld_start > descr->cld_end ||
1035                     cld->cld_end < descr->cld_start)
1036                         continue;
1037
1038                 /* overlapped and living locks. */
1039
1040                 /* We're not supposed to give up group lock. */
1041                 if (scan->cll_descr.cld_mode == CLM_GROUP) {
1042                         LASSERT(descr->cld_mode != CLM_GROUP ||
1043                                 descr->cld_gid != scan->cll_descr.cld_gid);
1044                         continue;
1045                 }
1046
1047                 scan_ols = osc_lock_at(scan);
1048
1049                 /* We need to cancel the compatible locks if we're enqueuing
1050                  * a lockless lock, for example:
1051                  * imagine that client has PR lock on [0, 1000], and thread T0
1052                  * is doing lockless IO in [500, 1500] region. Concurrent
1053                  * thread T1 can see lockless data in [500, 1000], which is
1054                  * wrong, because these data are possibly stale. */
1055                 if (!lockless && osc_lock_compatible(olck, scan_ols))
1056                         continue;
1057
1058                 cl_lock_get_trust(scan);
1059                 conflict = scan;
1060                 break;
1061         }
1062         spin_unlock(&hdr->coh_lock_guard);
1063
1064         if (conflict) {
1065                 if (lock->cll_descr.cld_mode == CLM_GROUP) {
1066                         /* we want a group lock but a previous lock request
1067                          * conflicts, we do not wait but return 0 so the
1068                          * request is send to the server
1069                          */
1070                         CDEBUG(D_DLMTRACE, "group lock %p is conflicted with %p, no wait, send to server\n",
1071                                lock, conflict);
1072                         cl_lock_put(env, conflict);
1073                         rc = 0;
1074                 } else {
1075                         CDEBUG(D_DLMTRACE, "lock %p is conflicted with %p, will wait\n",
1076                                lock, conflict);
1077                         LASSERT(lock->cll_conflict == NULL);
1078                         lu_ref_add(&conflict->cll_reference, "cancel-wait",
1079                                    lock);
1080                         lock->cll_conflict = conflict;
1081                         rc = CLO_WAIT;
1082                 }
1083         }
1084         return rc;
1085 }
1086
1087 /**
1088  * Implementation of cl_lock_operations::clo_enqueue() method for osc
1089  * layer. This initiates ldlm enqueue:
1090  *
1091  *     - cancels conflicting locks early (osc_lock_enqueue_wait());
1092  *
1093  *     - calls osc_enqueue_base() to do actual enqueue.
1094  *
1095  * osc_enqueue_base() is supplied with an upcall function that is executed
1096  * when lock is received either after a local cached ldlm lock is matched, or
1097  * when a reply from the server is received.
1098  *
1099  * This function does not wait for the network communication to complete.
1100  */
1101 static int osc_lock_enqueue(const struct lu_env *env,
1102                             const struct cl_lock_slice *slice,
1103                             struct cl_io *unused, __u32 enqflags)
1104 {
1105         struct osc_lock   *ols     = cl2osc_lock(slice);
1106         struct cl_lock     *lock    = ols->ols_cl.cls_lock;
1107         int result;
1108
1109         LASSERT(cl_lock_is_mutexed(lock));
1110         LASSERTF(ols->ols_state == OLS_NEW,
1111                  "Impossible state: %d\n", ols->ols_state);
1112
1113         LASSERTF(ergo(ols->ols_glimpse, lock->cll_descr.cld_mode <= CLM_READ),
1114                 "lock = %p, ols = %p\n", lock, ols);
1115
1116         result = osc_lock_enqueue_wait(env, ols);
1117         if (result == 0) {
1118                 if (!osc_lock_is_lockless(ols)) {
1119                         struct osc_object       *obj = cl2osc(slice->cls_obj);
1120                         struct osc_thread_info   *info = osc_env_info(env);
1121                         struct ldlm_res_id       *resname = &info->oti_resname;
1122                         ldlm_policy_data_t       *policy = &info->oti_policy;
1123                         struct ldlm_enqueue_info *einfo = &ols->ols_einfo;
1124
1125                         /* lock will be passed as upcall cookie,
1126                          * hold ref to prevent to be released. */
1127                         cl_lock_hold_add(env, lock, "upcall", lock);
1128                         /* a user for lock also */
1129                         cl_lock_user_add(env, lock);
1130                         ols->ols_state = OLS_ENQUEUED;
1131
1132                         /*
1133                          * XXX: this is possible blocking point as
1134                          * ldlm_lock_match(LDLM_FL_LVB_READY) waits for
1135                          * LDLM_CP_CALLBACK.
1136                          */
1137                         ostid_build_res_name(&obj->oo_oinfo->loi_oi, resname);
1138                         osc_lock_build_policy(env, lock, policy);
1139                         result = osc_enqueue_base(osc_export(obj), resname,
1140                                           &ols->ols_flags, policy,
1141                                           &ols->ols_lvb,
1142                                           obj->oo_oinfo->loi_kms_valid,
1143                                           osc_lock_upcall,
1144                                           ols, einfo, &ols->ols_handle,
1145                                           PTLRPCD_SET, 1, ols->ols_agl);
1146                         if (result != 0) {
1147                                 cl_lock_user_del(env, lock);
1148                                 cl_lock_unhold(env, lock, "upcall", lock);
1149                                 if (unlikely(result == -ECANCELED)) {
1150                                         ols->ols_state = OLS_NEW;
1151                                         result = 0;
1152                                 }
1153                         }
1154                 } else {
1155                         ols->ols_state = OLS_GRANTED;
1156                         ols->ols_owner = osc_env_io(env);
1157                 }
1158         }
1159         LASSERT(ergo(ols->ols_glimpse, !osc_lock_is_lockless(ols)));
1160         return result;
1161 }
1162
1163 static int osc_lock_wait(const struct lu_env *env,
1164                          const struct cl_lock_slice *slice)
1165 {
1166         struct osc_lock *olck = cl2osc_lock(slice);
1167         struct cl_lock  *lock = olck->ols_cl.cls_lock;
1168
1169         LINVRNT(osc_lock_invariant(olck));
1170
1171         if (olck->ols_glimpse && olck->ols_state >= OLS_UPCALL_RECEIVED) {
1172                 if (olck->ols_flags & LDLM_FL_LVB_READY) {
1173                         return 0;
1174                 } else if (olck->ols_agl) {
1175                         if (lock->cll_flags & CLF_FROM_UPCALL)
1176                                 /* It is from enqueue RPC reply upcall for
1177                                  * updating state. Do not re-enqueue. */
1178                                 return -ENAVAIL;
1179                         else
1180                                 olck->ols_state = OLS_NEW;
1181                 } else {
1182                         LASSERT(lock->cll_error);
1183                         return lock->cll_error;
1184                 }
1185         }
1186
1187         if (olck->ols_state == OLS_NEW) {
1188                 int rc;
1189
1190                 LASSERT(olck->ols_agl);
1191                 olck->ols_agl = 0;
1192                 olck->ols_flags &= ~LDLM_FL_BLOCK_NOWAIT;
1193                 rc = osc_lock_enqueue(env, slice, NULL, CEF_ASYNC | CEF_MUST);
1194                 if (rc != 0)
1195                         return rc;
1196                 else
1197                         return CLO_REENQUEUED;
1198         }
1199
1200         LASSERT(equi(olck->ols_state >= OLS_UPCALL_RECEIVED &&
1201                      lock->cll_error == 0, olck->ols_lock != NULL));
1202
1203         return lock->cll_error ?: olck->ols_state >= OLS_GRANTED ? 0 : CLO_WAIT;
1204 }
1205
1206 /**
1207  * An implementation of cl_lock_operations::clo_use() method that pins cached
1208  * lock.
1209  */
1210 static int osc_lock_use(const struct lu_env *env,
1211                         const struct cl_lock_slice *slice)
1212 {
1213         struct osc_lock *olck = cl2osc_lock(slice);
1214         int rc;
1215
1216         LASSERT(!olck->ols_hold);
1217
1218         /*
1219          * Atomically check for LDLM_FL_CBPENDING and addref a lock if this
1220          * flag is not set. This protects us from a concurrent blocking ast.
1221          */
1222         rc = ldlm_lock_addref_try(&olck->ols_handle, olck->ols_einfo.ei_mode);
1223         if (rc == 0) {
1224                 olck->ols_hold = 1;
1225                 olck->ols_state = OLS_GRANTED;
1226         } else {
1227                 struct cl_lock *lock;
1228
1229                 /*
1230                  * Lock is being cancelled somewhere within
1231                  * ldlm_handle_bl_callback(): LDLM_FL_CBPENDING is already
1232                  * set, but osc_ldlm_blocking_ast() hasn't yet acquired
1233                  * cl_lock mutex.
1234                  */
1235                 lock = slice->cls_lock;
1236                 LASSERT(lock->cll_state == CLS_INTRANSIT);
1237                 LASSERT(lock->cll_users > 0);
1238                 /* set a flag for osc_dlm_blocking_ast0() to signal the
1239                  * lock.*/
1240                 olck->ols_ast_wait = 1;
1241                 rc = CLO_WAIT;
1242         }
1243         return rc;
1244 }
1245
1246 static int osc_lock_flush(struct osc_lock *ols, int discard)
1247 {
1248         struct cl_lock       *lock  = ols->ols_cl.cls_lock;
1249         struct cl_env_nest    nest;
1250         struct lu_env   *env;
1251         int result = 0;
1252
1253         env = cl_env_nested_get(&nest);
1254         if (!IS_ERR(env)) {
1255                 struct osc_object    *obj   = cl2osc(ols->ols_cl.cls_obj);
1256                 struct cl_lock_descr *descr = &lock->cll_descr;
1257                 int rc = 0;
1258
1259                 if (descr->cld_mode >= CLM_WRITE) {
1260                         result = osc_cache_writeback_range(env, obj,
1261                                         descr->cld_start, descr->cld_end,
1262                                         1, discard);
1263                         LDLM_DEBUG(ols->ols_lock,
1264                                 "lock %p: %d pages were %s.\n", lock, result,
1265                                 discard ? "discarded" : "written");
1266                         if (result > 0)
1267                                 result = 0;
1268                 }
1269
1270                 rc = cl_lock_discard_pages(env, lock);
1271                 if (result == 0 && rc < 0)
1272                         result = rc;
1273
1274                 cl_env_nested_put(&nest, env);
1275         } else
1276                 result = PTR_ERR(env);
1277         if (result == 0) {
1278                 ols->ols_flush = 1;
1279                 LINVRNT(!osc_lock_has_pages(ols));
1280         }
1281         return result;
1282 }
1283
1284 /**
1285  * Implements cl_lock_operations::clo_cancel() method for osc layer. This is
1286  * called (as part of cl_lock_cancel()) when lock is canceled either voluntary
1287  * (LRU pressure, early cancellation, umount, etc.) or due to the conflict
1288  * with some other lock some where in the cluster. This function does the
1289  * following:
1290  *
1291  *     - invalidates all pages protected by this lock (after sending dirty
1292  *       ones to the server, as necessary);
1293  *
1294  *     - decref's underlying ldlm lock;
1295  *
1296  *     - cancels ldlm lock (ldlm_cli_cancel()).
1297  */
1298 static void osc_lock_cancel(const struct lu_env *env,
1299                             const struct cl_lock_slice *slice)
1300 {
1301         struct cl_lock   *lock    = slice->cls_lock;
1302         struct osc_lock  *olck    = cl2osc_lock(slice);
1303         struct ldlm_lock *dlmlock = olck->ols_lock;
1304         int            result  = 0;
1305         int            discard;
1306
1307         LASSERT(cl_lock_is_mutexed(lock));
1308         LINVRNT(osc_lock_invariant(olck));
1309
1310         if (dlmlock != NULL) {
1311                 int do_cancel;
1312
1313                 discard = !!(dlmlock->l_flags & LDLM_FL_DISCARD_DATA);
1314                 if (olck->ols_state >= OLS_GRANTED)
1315                         result = osc_lock_flush(olck, discard);
1316                 osc_lock_unhold(olck);
1317
1318                 lock_res_and_lock(dlmlock);
1319                 /* Now that we're the only user of dlm read/write reference,
1320                  * mostly the ->l_readers + ->l_writers should be zero.
1321                  * However, there is a corner case.
1322                  * See bug 18829 for details.*/
1323                 do_cancel = (dlmlock->l_readers == 0 &&
1324                              dlmlock->l_writers == 0);
1325                 dlmlock->l_flags |= LDLM_FL_CBPENDING;
1326                 unlock_res_and_lock(dlmlock);
1327                 if (do_cancel)
1328                         result = ldlm_cli_cancel(&olck->ols_handle, LCF_ASYNC);
1329                 if (result < 0)
1330                         CL_LOCK_DEBUG(D_ERROR, env, lock,
1331                                       "lock %p cancel failure with error(%d)\n",
1332                                       lock, result);
1333         }
1334         olck->ols_state = OLS_CANCELLED;
1335         olck->ols_flags &= ~LDLM_FL_LVB_READY;
1336         osc_lock_detach(env, olck);
1337 }
1338
1339 static int osc_lock_has_pages(struct osc_lock *olck)
1340 {
1341         return 0;
1342 }
1343
1344 static void osc_lock_delete(const struct lu_env *env,
1345                             const struct cl_lock_slice *slice)
1346 {
1347         struct osc_lock *olck;
1348
1349         olck = cl2osc_lock(slice);
1350         if (olck->ols_glimpse) {
1351                 LASSERT(!olck->ols_hold);
1352                 LASSERT(!olck->ols_lock);
1353                 return;
1354         }
1355
1356         LINVRNT(osc_lock_invariant(olck));
1357         LINVRNT(!osc_lock_has_pages(olck));
1358
1359         osc_lock_unhold(olck);
1360         osc_lock_detach(env, olck);
1361 }
1362
1363 /**
1364  * Implements cl_lock_operations::clo_state() method for osc layer.
1365  *
1366  * Maintains osc_lock::ols_owner field.
1367  *
1368  * This assumes that lock always enters CLS_HELD (from some other state) in
1369  * the same IO context as one that requested the lock. This should not be a
1370  * problem, because context is by definition shared by all activity pertaining
1371  * to the same high-level IO.
1372  */
1373 static void osc_lock_state(const struct lu_env *env,
1374                            const struct cl_lock_slice *slice,
1375                            enum cl_lock_state state)
1376 {
1377         struct osc_lock *lock = cl2osc_lock(slice);
1378
1379         /*
1380          * XXX multiple io contexts can use the lock at the same time.
1381          */
1382         LINVRNT(osc_lock_invariant(lock));
1383         if (state == CLS_HELD && slice->cls_lock->cll_state != CLS_HELD) {
1384                 struct osc_io *oio = osc_env_io(env);
1385
1386                 LASSERT(lock->ols_owner == NULL);
1387                 lock->ols_owner = oio;
1388         } else if (state != CLS_HELD)
1389                 lock->ols_owner = NULL;
1390 }
1391
1392 static int osc_lock_print(const struct lu_env *env, void *cookie,
1393                           lu_printer_t p, const struct cl_lock_slice *slice)
1394 {
1395         struct osc_lock *lock = cl2osc_lock(slice);
1396
1397         /*
1398          * XXX print ldlm lock and einfo properly.
1399          */
1400         (*p)(env, cookie, "%p %#16llx %#llx %d %p ",
1401              lock->ols_lock, lock->ols_flags, lock->ols_handle.cookie,
1402              lock->ols_state, lock->ols_owner);
1403         osc_lvb_print(env, cookie, p, &lock->ols_lvb);
1404         return 0;
1405 }
1406
1407 static int osc_lock_fits_into(const struct lu_env *env,
1408                               const struct cl_lock_slice *slice,
1409                               const struct cl_lock_descr *need,
1410                               const struct cl_io *io)
1411 {
1412         struct osc_lock *ols = cl2osc_lock(slice);
1413
1414         if (need->cld_enq_flags & CEF_NEVER)
1415                 return 0;
1416
1417         if (ols->ols_state >= OLS_CANCELLED)
1418                 return 0;
1419
1420         if (need->cld_mode == CLM_PHANTOM) {
1421                 if (ols->ols_agl)
1422                         return !(ols->ols_state > OLS_RELEASED);
1423
1424                 /*
1425                  * Note: the QUEUED lock can't be matched here, otherwise
1426                  * it might cause the deadlocks.
1427                  * In read_process,
1428                  * P1: enqueued read lock, create sublock1
1429                  * P2: enqueued write lock, create sublock2(conflicted
1430                  *     with sublock1).
1431                  * P1: Grant read lock.
1432                  * P1: enqueued glimpse lock(with holding sublock1_read),
1433                  *     matched with sublock2, waiting sublock2 to be granted.
1434                  *     But sublock2 can not be granted, because P1
1435                  *     will not release sublock1. Bang!
1436                  */
1437                 if (ols->ols_state < OLS_GRANTED ||
1438                     ols->ols_state > OLS_RELEASED)
1439                         return 0;
1440         } else if (need->cld_enq_flags & CEF_MUST) {
1441                 /*
1442                  * If the lock hasn't ever enqueued, it can't be matched
1443                  * because enqueue process brings in many information
1444                  * which can be used to determine things such as lockless,
1445                  * CEF_MUST, etc.
1446                  */
1447                 if (ols->ols_state < OLS_UPCALL_RECEIVED &&
1448                     ols->ols_locklessable)
1449                         return 0;
1450         }
1451         return 1;
1452 }
1453
1454 static const struct cl_lock_operations osc_lock_ops = {
1455         .clo_fini    = osc_lock_fini,
1456         .clo_enqueue = osc_lock_enqueue,
1457         .clo_wait    = osc_lock_wait,
1458         .clo_unuse   = osc_lock_unuse,
1459         .clo_use     = osc_lock_use,
1460         .clo_delete  = osc_lock_delete,
1461         .clo_state   = osc_lock_state,
1462         .clo_cancel  = osc_lock_cancel,
1463         .clo_weigh   = osc_lock_weigh,
1464         .clo_print   = osc_lock_print,
1465         .clo_fits_into = osc_lock_fits_into,
1466 };
1467
1468 static int osc_lock_lockless_unuse(const struct lu_env *env,
1469                                    const struct cl_lock_slice *slice)
1470 {
1471         struct osc_lock *ols = cl2osc_lock(slice);
1472         struct cl_lock *lock = slice->cls_lock;
1473
1474         LASSERT(ols->ols_state == OLS_GRANTED);
1475         LINVRNT(osc_lock_invariant(ols));
1476
1477         cl_lock_cancel(env, lock);
1478         cl_lock_delete(env, lock);
1479         return 0;
1480 }
1481
1482 static void osc_lock_lockless_cancel(const struct lu_env *env,
1483                                      const struct cl_lock_slice *slice)
1484 {
1485         struct osc_lock   *ols  = cl2osc_lock(slice);
1486         int result;
1487
1488         result = osc_lock_flush(ols, 0);
1489         if (result)
1490                 CERROR("Pages for lockless lock %p were not purged(%d)\n",
1491                        ols, result);
1492         ols->ols_state = OLS_CANCELLED;
1493 }
1494
1495 static int osc_lock_lockless_wait(const struct lu_env *env,
1496                                   const struct cl_lock_slice *slice)
1497 {
1498         struct osc_lock *olck = cl2osc_lock(slice);
1499         struct cl_lock  *lock = olck->ols_cl.cls_lock;
1500
1501         LINVRNT(osc_lock_invariant(olck));
1502         LASSERT(olck->ols_state >= OLS_UPCALL_RECEIVED);
1503
1504         return lock->cll_error;
1505 }
1506
1507 static void osc_lock_lockless_state(const struct lu_env *env,
1508                                     const struct cl_lock_slice *slice,
1509                                     enum cl_lock_state state)
1510 {
1511         struct osc_lock *lock = cl2osc_lock(slice);
1512
1513         LINVRNT(osc_lock_invariant(lock));
1514         if (state == CLS_HELD) {
1515                 struct osc_io *oio  = osc_env_io(env);
1516
1517                 LASSERT(ergo(lock->ols_owner, lock->ols_owner == oio));
1518                 lock->ols_owner = oio;
1519
1520                 /* set the io to be lockless if this lock is for io's
1521                  * host object */
1522                 if (cl_object_same(oio->oi_cl.cis_obj, slice->cls_obj))
1523                         oio->oi_lockless = 1;
1524         }
1525 }
1526
1527 static int osc_lock_lockless_fits_into(const struct lu_env *env,
1528                                        const struct cl_lock_slice *slice,
1529                                        const struct cl_lock_descr *need,
1530                                        const struct cl_io *io)
1531 {
1532         struct osc_lock *lock = cl2osc_lock(slice);
1533
1534         if (!(need->cld_enq_flags & CEF_NEVER))
1535                 return 0;
1536
1537         /* lockless lock should only be used by its owning io. b22147 */
1538         return (lock->ols_owner == osc_env_io(env));
1539 }
1540
1541 static const struct cl_lock_operations osc_lock_lockless_ops = {
1542         .clo_fini      = osc_lock_fini,
1543         .clo_enqueue   = osc_lock_enqueue,
1544         .clo_wait      = osc_lock_lockless_wait,
1545         .clo_unuse     = osc_lock_lockless_unuse,
1546         .clo_state     = osc_lock_lockless_state,
1547         .clo_fits_into = osc_lock_lockless_fits_into,
1548         .clo_cancel    = osc_lock_lockless_cancel,
1549         .clo_print     = osc_lock_print
1550 };
1551
1552 int osc_lock_init(const struct lu_env *env,
1553                   struct cl_object *obj, struct cl_lock *lock,
1554                   const struct cl_io *unused)
1555 {
1556         struct osc_lock *clk;
1557         int result;
1558
1559         OBD_SLAB_ALLOC_PTR_GFP(clk, osc_lock_kmem, GFP_NOFS);
1560         if (clk != NULL) {
1561                 __u32 enqflags = lock->cll_descr.cld_enq_flags;
1562
1563                 osc_lock_build_einfo(env, lock, clk, &clk->ols_einfo);
1564                 atomic_set(&clk->ols_pageref, 0);
1565                 clk->ols_state = OLS_NEW;
1566
1567                 clk->ols_flags = osc_enq2ldlm_flags(enqflags);
1568                 clk->ols_agl = !!(enqflags & CEF_AGL);
1569                 if (clk->ols_agl)
1570                         clk->ols_flags |= LDLM_FL_BLOCK_NOWAIT;
1571                 if (clk->ols_flags & LDLM_FL_HAS_INTENT)
1572                         clk->ols_glimpse = 1;
1573
1574                 cl_lock_slice_add(lock, &clk->ols_cl, obj, &osc_lock_ops);
1575
1576                 if (!(enqflags & CEF_MUST))
1577                         /* try to convert this lock to a lockless lock */
1578                         osc_lock_to_lockless(env, clk, (enqflags & CEF_NEVER));
1579                 if (clk->ols_locklessable && !(enqflags & CEF_DISCARD_DATA))
1580                         clk->ols_flags |= LDLM_FL_DENY_ON_CONTENTION;
1581
1582                 LDLM_DEBUG_NOLOCK("lock %p, osc lock %p, flags %llx\n",
1583                                 lock, clk, clk->ols_flags);
1584
1585                 result = 0;
1586         } else
1587                 result = -ENOMEM;
1588         return result;
1589 }
1590
1591 int osc_dlm_lock_pageref(struct ldlm_lock *dlm)
1592 {
1593         struct osc_lock *olock;
1594         int           rc = 0;
1595
1596         spin_lock(&osc_ast_guard);
1597         olock = dlm->l_ast_data;
1598         /*
1599          * there's a very rare race with osc_page_addref_lock(), but that
1600          * doesn't matter because in the worst case we don't cancel a lock
1601          * which we actually can, that's no harm.
1602          */
1603         if (olock != NULL &&
1604             atomic_add_return(_PAGEREF_MAGIC,
1605                                   &olock->ols_pageref) != _PAGEREF_MAGIC) {
1606                 atomic_sub(_PAGEREF_MAGIC, &olock->ols_pageref);
1607                 rc = 1;
1608         }
1609         spin_unlock(&osc_ast_guard);
1610         return rc;
1611 }
1612
1613 /** @} osc */