1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
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
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
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * Implementation of cl_lock for OSC layer.
38 * Author: Nikita Danilov <nikita.danilov@sun.com>
41 /** \addtogroup osc osc @{ */
43 #define DEBUG_SUBSYSTEM S_OSC
46 # include <libcfs/libcfs.h>
48 # include <liblustre.h>
50 /* fid_build_reg_res_name() */
51 #include <lustre_fid.h>
53 #include "osc_cl_internal.h"
55 /*****************************************************************************
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);
66 int osc_lock_is_lockless(const struct osc_lock *olck)
68 return (olck->ols_cl.cls_ops == &osc_lock_lockless_ops);
72 * Returns a weak pointer to the ldlm lock identified by a handle. Returned
73 * pointer cannot be dereferenced, as lock is not protected from concurrent
74 * reclaim. This function is a helper for osc_lock_invariant().
76 static struct ldlm_lock *osc_handle_ptr(struct lustre_handle *handle)
78 struct ldlm_lock *lock;
80 lock = ldlm_handle2lock(handle);
87 * Invariant that has to be true all of the time.
89 static int osc_lock_invariant(struct osc_lock *ols)
91 struct ldlm_lock *lock = osc_handle_ptr(&ols->ols_handle);
92 struct ldlm_lock *olock = ols->ols_lock;
93 int handle_used = lustre_handle_is_used(&ols->ols_handle);
96 ergo(osc_lock_is_lockless(ols),
97 ols->ols_locklessable && ols->ols_lock == NULL) ||
98 (ergo(olock != NULL, handle_used) &&
100 olock->l_handle.h_cookie == ols->ols_handle.cookie) &&
102 * Check that ->ols_handle and ->ols_lock are consistent, but
103 * take into account that they are set at the different time.
106 ergo(lock != NULL && olock != NULL, lock == olock) &&
107 ergo(lock == NULL, olock == NULL)) &&
108 ergo(ols->ols_state == OLS_CANCELLED,
109 olock == NULL && !handle_used) &&
111 * DLM lock is destroyed only after we have seen cancellation
114 ergo(olock != NULL && ols->ols_state < OLS_CANCELLED,
115 !olock->l_destroyed) &&
116 ergo(ols->ols_state == OLS_GRANTED,
118 olock->l_req_mode == olock->l_granted_mode &&
122 /*****************************************************************************
129 * Breaks a link between osc_lock and dlm_lock.
131 static void osc_lock_detach(const struct lu_env *env, struct osc_lock *olck)
133 struct ldlm_lock *dlmlock;
135 spin_lock(&osc_ast_guard);
136 dlmlock = olck->ols_lock;
137 if (dlmlock == NULL) {
138 spin_unlock(&osc_ast_guard);
142 olck->ols_lock = NULL;
143 /* wb(); --- for all who checks (ols->ols_lock != NULL) before
144 * call to osc_lock_detach() */
145 dlmlock->l_ast_data = NULL;
146 olck->ols_handle.cookie = 0ULL;
147 spin_unlock(&osc_ast_guard);
149 lock_res_and_lock(dlmlock);
150 if (dlmlock->l_granted_mode == dlmlock->l_req_mode) {
151 struct cl_object *obj = olck->ols_cl.cls_obj;
152 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
153 __u64 old_kms = cl2osc(obj)->oo_oinfo->loi_kms;
155 /* Update the kms. Need to loop all granted locks.
156 * Not a problem for the client */
157 attr->cat_kms = ldlm_extent_shift_kms(dlmlock, old_kms);
158 unlock_res_and_lock(dlmlock);
160 cl_object_attr_lock(obj);
161 cl_object_attr_set(env, obj, attr, CAT_KMS);
162 cl_object_attr_unlock(obj);
164 unlock_res_and_lock(dlmlock);
166 /* release a reference taken in osc_lock_upcall0(). */
167 lu_ref_del(&dlmlock->l_reference, "osc_lock", olck);
168 LDLM_LOCK_RELEASE(dlmlock);
171 static int osc_lock_unuse(const struct lu_env *env,
172 const struct cl_lock_slice *slice)
174 struct osc_lock *ols = cl2osc_lock(slice);
177 LASSERT(ols->ols_state == OLS_GRANTED ||
178 ols->ols_state == OLS_UPCALL_RECEIVED);
179 LINVRNT(osc_lock_invariant(ols));
181 if (ols->ols_glimpse) {
182 LASSERT(ols->ols_hold == 0);
185 LASSERT(ols->ols_hold);
188 * Move lock into OLS_RELEASED state before calling osc_cancel_base()
189 * so that possible synchronous cancellation (that always happens
190 * e.g., for liblustre) sees that lock is released.
192 ols->ols_state = OLS_RELEASED;
194 result = osc_cancel_base(&ols->ols_handle, ols->ols_einfo.ei_mode);
195 ols->ols_has_ref = 0;
199 static void osc_lock_fini(const struct lu_env *env,
200 struct cl_lock_slice *slice)
202 struct osc_lock *ols = cl2osc_lock(slice);
204 LINVRNT(osc_lock_invariant(ols));
206 * ->ols_hold can still be true at this point if, for example, a
207 * thread that requested a lock was killed (and released a reference
208 * to the lock), before reply from a server was received. In this case
209 * lock is destroyed immediately after upcall.
212 osc_lock_unuse(env, slice);
213 LASSERT(ols->ols_lock == NULL);
215 OBD_SLAB_FREE_PTR(ols, osc_lock_kmem);
218 void osc_lock_build_res(const struct lu_env *env, const struct osc_object *obj,
219 struct ldlm_res_id *resname)
221 const struct lu_fid *fid = lu_object_fid(&obj->oo_cl.co_lu);
224 * In the perfect world of the future, where ost servers talk
227 fid_build_reg_res_name(fid, resname);
230 * In reality, where ost server expects ->lsm_object_id and
231 * ->lsm_object_gr in rename.
233 osc_build_res_name(obj->oo_oinfo->loi_id, obj->oo_oinfo->loi_gr,
238 static void osc_lock_build_policy(const struct lu_env *env,
239 const struct cl_lock *lock,
240 ldlm_policy_data_t *policy)
242 const struct cl_lock_descr *d = &lock->cll_descr;
244 osc_index2policy(policy, d->cld_obj, d->cld_start, d->cld_end);
245 policy->l_extent.gid = d->cld_gid;
248 static int osc_enq2ldlm_flags(__u32 enqflags)
252 LASSERT((enqflags & ~CEF_MASK) == 0);
254 if (enqflags & CEF_NONBLOCK)
255 result |= LDLM_FL_BLOCK_NOWAIT;
256 if (enqflags & CEF_ASYNC)
257 result |= LDLM_FL_HAS_INTENT;
258 if (enqflags & CEF_DISCARD_DATA)
259 result |= LDLM_AST_DISCARD_DATA;
264 * Global spin-lock protecting consistency of ldlm_lock::l_ast_data
265 * pointers. Initialized in osc_init().
267 spinlock_t osc_ast_guard;
269 static struct osc_lock *osc_ast_data_get(struct ldlm_lock *dlm_lock)
271 struct osc_lock *olck;
273 lock_res_and_lock(dlm_lock);
274 spin_lock(&osc_ast_guard);
275 olck = dlm_lock->l_ast_data;
277 struct cl_lock *lock = olck->ols_cl.cls_lock;
279 * If osc_lock holds a reference on ldlm lock, return it even
280 * when cl_lock is in CLS_FREEING state. This way
282 * osc_ast_data_get(dlmlock) == NULL
284 * guarantees that all osc references on dlmlock were
285 * released. osc_dlm_blocking_ast0() relies on that.
287 if (lock->cll_state < CLS_FREEING || olck->ols_has_ref) {
288 cl_lock_get_trust(lock);
289 lu_ref_add_atomic(&lock->cll_reference,
290 "ast", cfs_current());
294 spin_unlock(&osc_ast_guard);
295 unlock_res_and_lock(dlm_lock);
299 static void osc_ast_data_put(const struct lu_env *env, struct osc_lock *olck)
301 struct cl_lock *lock;
303 lock = olck->ols_cl.cls_lock;
304 lu_ref_del(&lock->cll_reference, "ast", cfs_current());
305 cl_lock_put(env, lock);
309 * Updates object attributes from a lock value block (lvb) received together
310 * with the DLM lock reply from the server. Copy of osc_update_enqueue()
313 * This can be optimized to not update attributes when lock is a result of a
316 * Called under lock and resource spin-locks.
318 static void osc_lock_lvb_update(const struct lu_env *env, struct osc_lock *olck,
322 struct cl_object *obj;
323 struct lov_oinfo *oinfo;
324 struct cl_attr *attr;
329 if (!(olck->ols_flags & LDLM_FL_LVB_READY)) {
334 lvb = &olck->ols_lvb;
335 obj = olck->ols_cl.cls_obj;
336 oinfo = cl2osc(obj)->oo_oinfo;
337 attr = &osc_env_info(env)->oti_attr;
338 valid = CAT_BLOCKS | CAT_ATIME | CAT_CTIME | CAT_MTIME | CAT_SIZE;
339 cl_lvb2attr(attr, lvb);
341 cl_object_attr_lock(obj);
343 struct ldlm_lock *dlmlock;
346 dlmlock = olck->ols_lock;
347 LASSERT(dlmlock != NULL);
349 /* re-grab LVB from a dlm lock under DLM spin-locks. */
350 *lvb = *(struct ost_lvb *)dlmlock->l_lvb_data;
351 size = lvb->lvb_size;
352 /* Extend KMS up to the end of this lock and no further
353 * A lock on [x,y] means a KMS of up to y + 1 bytes! */
354 if (size > dlmlock->l_policy_data.l_extent.end)
355 size = dlmlock->l_policy_data.l_extent.end + 1;
356 if (size >= oinfo->loi_kms) {
357 LDLM_DEBUG(dlmlock, "lock acquired, setting rss="LPU64
358 ", kms="LPU64, lvb->lvb_size, size);
360 attr->cat_kms = size;
362 LDLM_DEBUG(dlmlock, "lock acquired, setting rss="
363 LPU64"; leaving kms="LPU64", end="LPU64,
364 lvb->lvb_size, oinfo->loi_kms,
365 dlmlock->l_policy_data.l_extent.end);
367 ldlm_lock_allow_match_locked(dlmlock);
368 } else if (rc == -ENAVAIL && olck->ols_glimpse) {
369 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
370 " kms="LPU64"\n", lvb->lvb_size, oinfo->loi_kms);
375 cl_object_attr_set(env, obj, attr, valid);
377 cl_object_attr_unlock(obj);
383 * Called when a lock is granted, from an upcall (when server returned a
384 * granted lock), or from completion AST, when server returned a blocked lock.
386 * Called under lock and resource spin-locks, that are released temporarily
389 static void osc_lock_granted(const struct lu_env *env, struct osc_lock *olck,
390 struct ldlm_lock *dlmlock, int rc)
392 struct ldlm_extent *ext;
393 struct cl_lock *lock;
394 struct cl_lock_descr *descr;
396 LASSERT(dlmlock->l_granted_mode == dlmlock->l_req_mode);
399 if (olck->ols_state != OLS_GRANTED) {
400 lock = olck->ols_cl.cls_lock;
401 ext = &dlmlock->l_policy_data.l_extent;
402 descr = &osc_env_info(env)->oti_descr;
403 descr->cld_obj = lock->cll_descr.cld_obj;
405 /* XXX check that ->l_granted_mode is valid. */
406 descr->cld_mode = osc_ldlm2cl_lock(dlmlock->l_granted_mode);
407 descr->cld_start = cl_index(descr->cld_obj, ext->start);
408 descr->cld_end = cl_index(descr->cld_obj, ext->end);
409 descr->cld_gid = ext->gid;
411 * tell upper layers the extent of the lock that was actually
414 olck->ols_state = OLS_GRANTED;
415 osc_lock_lvb_update(env, olck, rc);
417 /* release DLM spin-locks to allow cl_lock_{modify,signal}()
418 * to take a semaphore on a parent lock. This is safe, because
419 * spin-locks are needed to protect consistency of
420 * dlmlock->l_*_mode and LVB, and we have finished processing
422 unlock_res_and_lock(dlmlock);
423 cl_lock_modify(env, lock, descr);
424 cl_lock_signal(env, lock);
425 LINVRNT(osc_lock_invariant(olck));
426 lock_res_and_lock(dlmlock);
431 static void osc_lock_upcall0(const struct lu_env *env, struct osc_lock *olck)
434 struct ldlm_lock *dlmlock;
438 dlmlock = ldlm_handle2lock_long(&olck->ols_handle, 0);
439 LASSERT(dlmlock != NULL);
441 lock_res_and_lock(dlmlock);
442 spin_lock(&osc_ast_guard);
443 LASSERT(dlmlock->l_ast_data == olck);
444 LASSERT(olck->ols_lock == NULL);
445 olck->ols_lock = dlmlock;
446 spin_unlock(&osc_ast_guard);
449 * Lock might be not yet granted. In this case, completion ast
450 * (osc_ldlm_completion_ast()) comes later and finishes lock
453 if (dlmlock->l_granted_mode == dlmlock->l_req_mode)
454 osc_lock_granted(env, olck, dlmlock, 0);
455 unlock_res_and_lock(dlmlock);
458 * osc_enqueue_interpret() decrefs asynchronous locks, counter
461 ldlm_lock_addref(&olck->ols_handle, olck->ols_einfo.ei_mode);
462 olck->ols_hold = olck->ols_has_ref = 1;
464 /* lock reference taken by ldlm_handle2lock_long() is owned by
465 * osc_lock and released in osc_lock_detach() */
466 lu_ref_add(&dlmlock->l_reference, "osc_lock", olck);
470 * Lock upcall function that is executed either when a reply to ENQUEUE rpc is
471 * received from a server, or after osc_enqueue_base() matched a local DLM
474 static int osc_lock_upcall(void *cookie, int errcode)
476 struct osc_lock *olck = cookie;
477 struct cl_lock_slice *slice = &olck->ols_cl;
478 struct cl_lock *lock = slice->cls_lock;
480 struct cl_env_nest nest;
483 env = cl_env_nested_get(&nest);
487 cl_lock_mutex_get(env, lock);
489 LASSERT(lock->cll_state >= CLS_QUEUING);
490 if (olck->ols_state == OLS_ENQUEUED) {
491 olck->ols_state = OLS_UPCALL_RECEIVED;
492 rc = ldlm_error2errno(errcode);
493 } else if (olck->ols_state == OLS_CANCELLED) {
496 CERROR("Impossible state: %i\n", olck->ols_state);
500 struct ldlm_lock *dlmlock;
502 dlmlock = ldlm_handle2lock(&olck->ols_handle);
503 if (dlmlock != NULL) {
504 lock_res_and_lock(dlmlock);
505 spin_lock(&osc_ast_guard);
506 LASSERT(olck->ols_lock == NULL);
507 dlmlock->l_ast_data = NULL;
508 olck->ols_handle.cookie = 0ULL;
509 spin_unlock(&osc_ast_guard);
510 unlock_res_and_lock(dlmlock);
511 LDLM_LOCK_PUT(dlmlock);
514 if (olck->ols_glimpse)
515 olck->ols_glimpse = 0;
516 osc_lock_upcall0(env, olck);
519 /* Error handling, some errors are tolerable. */
520 if (olck->ols_locklessable && rc == -EUSERS) {
521 /* This is a tolerable error, turn this lock into
524 osc_object_set_contended(cl2osc(slice->cls_obj));
525 LASSERT(slice->cls_ops == &osc_lock_ops);
527 /* Change this lock to ldlmlock-less lock. */
528 osc_lock_to_lockless(env, olck, 1);
529 olck->ols_state = OLS_GRANTED;
531 } else if (olck->ols_glimpse && rc == -ENAVAIL) {
532 osc_lock_lvb_update(env, olck, rc);
533 cl_lock_delete(env, lock);
534 /* Hide the error. */
539 /* on error, lock was signaled by cl_lock_error() */
540 cl_lock_signal(env, lock);
542 cl_lock_error(env, lock, rc);
544 cl_lock_mutex_put(env, lock);
546 /* release cookie reference, acquired by osc_lock_enqueue() */
547 lu_ref_del(&lock->cll_reference, "upcall", lock);
548 cl_lock_put(env, lock);
549 cl_env_nested_put(&nest, env);
551 /* should never happen, similar to osc_ldlm_blocking_ast(). */
557 * Core of osc_dlm_blocking_ast() logic.
559 static void osc_lock_blocking(const struct lu_env *env,
560 struct ldlm_lock *dlmlock,
561 struct osc_lock *olck, int blocking)
563 struct cl_lock *lock = olck->ols_cl.cls_lock;
565 LASSERT(olck->ols_lock == dlmlock);
566 CLASSERT(OLS_BLOCKED < OLS_CANCELLED);
567 LASSERT(!osc_lock_is_lockless(olck));
571 * Lock might be still addref-ed here, if e.g., blocking ast
572 * is sent for a failed lock.
574 osc_lock_unuse(env, &olck->ols_cl);
576 if (blocking && olck->ols_state < OLS_BLOCKED)
578 * Move osc_lock into OLS_BLOCKED before canceling the lock,
579 * because it recursively re-enters osc_lock_blocking(), with
580 * the state set to OLS_CANCELLED.
582 olck->ols_state = OLS_BLOCKED;
584 * cancel and destroy lock at least once no matter how blocking ast is
585 * entered (see comment above osc_ldlm_blocking_ast() for use
586 * cases). cl_lock_cancel() and cl_lock_delete() are idempotent.
588 cl_lock_cancel(env, lock);
589 cl_lock_delete(env, lock);
593 * Helper for osc_dlm_blocking_ast() handling discrepancies between cl_lock
594 * and ldlm_lock caches.
596 static int osc_dlm_blocking_ast0(const struct lu_env *env,
597 struct ldlm_lock *dlmlock,
598 void *data, int flag)
600 struct osc_lock *olck;
601 struct cl_lock *lock;
605 LASSERT(flag == LDLM_CB_BLOCKING || flag == LDLM_CB_CANCELING);
608 olck = osc_ast_data_get(dlmlock);
610 lock = olck->ols_cl.cls_lock;
611 cl_lock_mutex_get(env, lock);
612 LINVRNT(osc_lock_invariant(olck));
613 if (olck->ols_ast_wait) {
614 /* wake up osc_lock_use() */
615 cl_lock_signal(env, lock);
616 olck->ols_ast_wait = 0;
619 * Lock might have been canceled while this thread was
620 * sleeping for lock mutex, but olck is pinned in memory.
622 if (olck == dlmlock->l_ast_data) {
624 * NOTE: DLM sends blocking AST's for failed locks
625 * (that are still in pre-OLS_GRANTED state)
626 * too, and they have to be canceled otherwise
627 * DLM lock is never destroyed and stuck in
630 * Alternatively, ldlm_cli_cancel() can be
631 * called here directly for osc_locks with
632 * ols_state < OLS_GRANTED to maintain an
633 * invariant that ->clo_cancel() is only called
634 * for locks that were granted.
636 LASSERT(data == olck);
637 osc_lock_blocking(env, dlmlock,
638 olck, flag == LDLM_CB_BLOCKING);
641 cl_lock_mutex_put(env, lock);
642 osc_ast_data_put(env, olck);
645 * DLM lock exists, but there is no cl_lock attached to it.
646 * This is a `normal' race. cl_object and its cl_lock's can be
647 * removed by memory pressure, together with all pages.
649 cancel = (flag == LDLM_CB_BLOCKING);
652 struct lustre_handle *lockh;
654 lockh = &osc_env_info(env)->oti_handle;
655 ldlm_lock2handle(dlmlock, lockh);
656 result = ldlm_cli_cancel(lockh);
663 * Blocking ast invoked by ldlm when dlm lock is either blocking progress of
664 * some other lock, or is canceled. This function is installed as a
665 * ldlm_lock::l_blocking_ast() for client extent locks.
667 * Control flow is tricky, because ldlm uses the same call-back
668 * (ldlm_lock::l_blocking_ast()) for both blocking and cancellation ast's.
670 * \param dlmlock lock for which ast occurred.
672 * \param new description of a conflicting lock in case of blocking ast.
674 * \param data value of dlmlock->l_ast_data
676 * \param flag LDLM_CB_BLOCKING or LDLM_CB_CANCELING. Used to distinguish
677 * cancellation and blocking ast's.
679 * Possible use cases:
681 * - ldlm calls dlmlock->l_blocking_ast(..., LDLM_CB_CANCELING) to cancel
682 * lock due to lock lru pressure, or explicit user request to purge
685 * - ldlm calls dlmlock->l_blocking_ast(..., LDLM_CB_BLOCKING) to notify
686 * us that dlmlock conflicts with another lock that some client is
687 * enqueing. Lock is canceled.
689 * - cl_lock_cancel() is called. osc_lock_cancel() calls
690 * ldlm_cli_cancel() that calls
692 * dlmlock->l_blocking_ast(..., LDLM_CB_CANCELING)
694 * recursively entering osc_ldlm_blocking_ast().
696 * - client cancels lock voluntary (e.g., as a part of early cancellation):
699 * osc_lock_cancel()->
700 * ldlm_cli_cancel()->
701 * dlmlock->l_blocking_ast(..., LDLM_CB_CANCELING)
704 static int osc_ldlm_blocking_ast(struct ldlm_lock *dlmlock,
705 struct ldlm_lock_desc *new, void *data,
709 struct cl_env_nest nest;
713 * This can be called in the context of outer IO, e.g.,
716 * ->osc_enqueue_base()->...
717 * ->ldlm_prep_elc_req()->...
718 * ->ldlm_cancel_callback()->...
719 * ->osc_ldlm_blocking_ast()
721 * new environment has to be created to not corrupt outer context.
723 env = cl_env_nested_get(&nest);
725 result = osc_dlm_blocking_ast0(env, dlmlock, data, flag);
726 cl_env_nested_put(&nest, env);
728 result = PTR_ERR(env);
730 * XXX This should never happen, as cl_lock is
731 * stuck. Pre-allocated environment a la vvp_inode_fini_env
737 if (result == -ENODATA)
740 CERROR("BAST failed: %d\n", result);
745 static int osc_ldlm_completion_ast(struct ldlm_lock *dlmlock,
746 int flags, void *data)
748 struct cl_env_nest nest;
750 struct osc_lock *olck;
751 struct cl_lock *lock;
755 /* first, do dlm part of the work */
756 dlmrc = ldlm_completion_ast_async(dlmlock, flags, data);
757 /* then, notify cl_lock */
758 env = cl_env_nested_get(&nest);
760 olck = osc_ast_data_get(dlmlock);
762 lock = olck->ols_cl.cls_lock;
763 cl_lock_mutex_get(env, lock);
765 * ldlm_handle_cp_callback() copied LVB from request
766 * to lock->l_lvb_data, store it in osc_lock.
768 LASSERT(dlmlock->l_lvb_data != NULL);
769 lock_res_and_lock(dlmlock);
770 olck->ols_lvb = *(struct ost_lvb *)dlmlock->l_lvb_data;
771 if (olck->ols_lock == NULL)
773 * upcall (osc_lock_upcall()) hasn't yet been
774 * called. Do nothing now, upcall will bind
775 * olck to dlmlock and signal the waiters.
777 * This maintains an invariant that osc_lock
778 * and ldlm_lock are always bound when
779 * osc_lock is in OLS_GRANTED state.
782 else if (dlmlock->l_granted_mode != LCK_MINMODE)
783 osc_lock_granted(env, olck, dlmlock, dlmrc);
784 unlock_res_and_lock(dlmlock);
786 cl_lock_error(env, lock, dlmrc);
787 cl_lock_mutex_put(env, lock);
788 osc_ast_data_put(env, olck);
791 result = -ELDLM_NO_LOCK_DATA;
792 cl_env_nested_put(&nest, env);
794 result = PTR_ERR(env);
795 return dlmrc ?: result;
798 static int osc_ldlm_glimpse_ast(struct ldlm_lock *dlmlock, void *data)
800 struct ptlrpc_request *req = data;
801 struct osc_lock *olck;
802 struct cl_lock *lock;
803 struct cl_object *obj;
804 struct cl_env_nest nest;
807 struct req_capsule *cap;
810 LASSERT(lustre_msg_get_opc(req->rq_reqmsg) == LDLM_GL_CALLBACK);
812 env = cl_env_nested_get(&nest);
815 * osc_ast_data_get() has to go after environment is
816 * allocated, because osc_ast_data() acquires a
817 * reference to a lock, and it can only be released in
820 olck = osc_ast_data_get(dlmlock);
822 lock = olck->ols_cl.cls_lock;
823 cl_lock_mutex_get(env, lock);
825 req_capsule_extend(cap, &RQF_LDLM_GL_CALLBACK);
826 req_capsule_set_size(cap, &RMF_DLM_LVB, RCL_SERVER,
828 result = req_capsule_server_pack(cap);
830 lvb = req_capsule_server_get(cap, &RMF_DLM_LVB);
831 obj = lock->cll_descr.cld_obj;
832 result = cl_object_glimpse(env, obj, lvb);
834 cl_lock_mutex_put(env, lock);
835 osc_ast_data_put(env, olck);
838 * These errors are normal races, so we don't want to
839 * fill the console with messages by calling
842 lustre_pack_reply(req, 1, NULL, NULL);
843 result = -ELDLM_NO_LOCK_DATA;
845 cl_env_nested_put(&nest, env);
847 result = PTR_ERR(env);
848 req->rq_status = result;
852 static unsigned long osc_lock_weigh(const struct lu_env *env,
853 const struct cl_lock_slice *slice)
856 * don't need to grab coh_page_guard since we don't care the exact #
859 return cl_object_header(slice->cls_obj)->coh_pages;
863 * Get the weight of dlm lock for early cancellation.
865 * XXX: it should return the pages covered by this \a dlmlock.
867 static unsigned long osc_ldlm_weigh_ast(struct ldlm_lock *dlmlock)
869 struct cl_env_nest nest;
871 struct osc_lock *lock;
873 unsigned long weight;
878 * osc_ldlm_weigh_ast has a complex context since it might be called
879 * because of lock canceling, or from user's input. We have to make
880 * a new environment for it. Probably it is implementation safe to use
881 * the upper context because cl_lock_put don't modify environment
882 * variables. But in case of ..
884 env = cl_env_nested_get(&nest);
886 /* Mostly because lack of memory, tend to eliminate this lock*/
889 LASSERT(dlmlock->l_resource->lr_type == LDLM_EXTENT);
890 lock = osc_ast_data_get(dlmlock);
892 /* cl_lock was destroyed because of memory pressure.
893 * It is much reasonable to assign this type of lock
896 GOTO(out, weight = 0);
899 cll = lock->ols_cl.cls_lock;
900 cl_lock_mutex_get(env, cll);
901 weight = cl_lock_weigh(env, cll);
902 cl_lock_mutex_put(env, cll);
903 osc_ast_data_put(env, lock);
907 cl_env_nested_put(&nest, env);
911 static void osc_lock_build_einfo(const struct lu_env *env,
912 const struct cl_lock *clock,
913 struct osc_lock *lock,
914 struct ldlm_enqueue_info *einfo)
916 enum cl_lock_mode mode;
918 mode = clock->cll_descr.cld_mode;
919 if (mode == CLM_PHANTOM)
921 * For now, enqueue all glimpse locks in read mode. In the
922 * future, client might choose to enqueue LCK_PW lock for
923 * glimpse on a file opened for write.
927 einfo->ei_type = LDLM_EXTENT;
928 einfo->ei_mode = osc_cl_lock2ldlm(mode);
929 einfo->ei_cb_bl = osc_ldlm_blocking_ast;
930 einfo->ei_cb_cp = osc_ldlm_completion_ast;
931 einfo->ei_cb_gl = osc_ldlm_glimpse_ast;
932 einfo->ei_cb_wg = osc_ldlm_weigh_ast;
933 einfo->ei_cbdata = lock; /* value to be put into ->l_ast_data */
936 static int osc_lock_delete0(struct cl_lock *conflict)
938 struct cl_env_nest nest;
942 env = cl_env_nested_get(&nest);
944 cl_lock_delete(env, conflict);
945 cl_env_nested_put(&nest, env);
951 * Cancels \a conflict lock and waits until it reached CLS_FREEING state. This
952 * is called as a part of enqueuing to cancel conflicting locks early.
954 * \retval 0: success, \a conflict was cancelled and destroyed.
956 * \retval CLO_REPEAT: \a conflict was cancelled, but \a lock mutex was
957 * released in the process. Repeat enqueing.
959 * \retval -EWOULDBLOCK: \a conflict cannot be cancelled immediately, and
960 * either \a lock is non-blocking, or current thread
961 * holds other locks, that prevent it from waiting
962 * for cancel to complete.
964 * \retval -ve: other error, including -EINTR.
967 static int osc_lock_cancel_wait(const struct lu_env *env, struct cl_lock *lock,
968 struct cl_lock *conflict, int canwait)
972 LASSERT(cl_lock_is_mutexed(lock));
973 LASSERT(cl_lock_is_mutexed(conflict));
976 if (conflict->cll_state != CLS_FREEING) {
977 cl_lock_cancel(env, conflict);
978 rc = osc_lock_delete0(conflict);
981 if (conflict->cll_flags & (CLF_CANCELPEND|CLF_DOOMED)) {
983 if (cl_lock_nr_mutexed(env) > 2)
985 * If mutices of locks other than @lock and
986 * @scan are held by the current thread, it
987 * cannot wait on @scan state change in a
988 * dead-lock safe matter, so simply skip early
989 * cancellation in this case.
991 * This means that early cancellation doesn't
992 * work when there is even slight mutex
993 * contention, as top-lock's mutex is usually
998 /* Waiting for @scan to be destroyed */
999 cl_lock_mutex_put(env, lock);
1001 rc = cl_lock_state_wait(env, conflict);
1003 conflict->cll_state < CLS_FREEING);
1004 /* mutex was released, repeat enqueue. */
1005 rc = rc ?: CLO_REPEAT;
1006 cl_lock_mutex_get(env, lock);
1009 LASSERT(ergo(!rc, conflict->cll_state == CLS_FREEING));
1010 CDEBUG(D_INFO, "lock %p was %s freed now, rc (%d)\n",
1011 conflict, rc ? "not":"", rc);
1017 * Determine if the lock should be converted into a lockless lock.
1020 * - if the lock has an explicite requirment for a non-lockless lock;
1021 * - if the io lock request type ci_lockreq;
1022 * - send the enqueue rpc to ost to make the further decision;
1023 * - special treat to truncate lockless lock
1025 * Additional policy can be implemented here, e.g., never do lockless-io
1026 * for large extents.
1028 static void osc_lock_to_lockless(const struct lu_env *env,
1029 struct osc_lock *ols, int force)
1031 struct cl_lock_slice *slice = &ols->ols_cl;
1032 struct cl_lock *lock = slice->cls_lock;
1034 LASSERT(ols->ols_state == OLS_NEW ||
1035 ols->ols_state == OLS_UPCALL_RECEIVED);
1038 ols->ols_locklessable = 1;
1039 LASSERT(cl_lock_is_mutexed(lock));
1040 slice->cls_ops = &osc_lock_lockless_ops;
1042 struct osc_io *oio = osc_env_io(env);
1043 struct cl_io *io = oio->oi_cl.cis_io;
1044 struct cl_object *obj = slice->cls_obj;
1045 struct osc_object *oob = cl2osc(obj);
1046 const struct osc_device *osd = lu2osc_dev(obj->co_lu.lo_dev);
1047 struct obd_connect_data *ocd;
1049 LASSERT(io->ci_lockreq == CILR_MANDATORY ||
1050 io->ci_lockreq == CILR_MAYBE ||
1051 io->ci_lockreq == CILR_NEVER);
1053 ocd = &class_exp2cliimp(osc_export(oob))->imp_connect_data;
1054 ols->ols_locklessable = (io->ci_type != CIT_TRUNC) &&
1055 (io->ci_lockreq == CILR_MAYBE) &&
1056 (ocd->ocd_connect_flags & OBD_CONNECT_SRVLOCK);
1057 if (io->ci_lockreq == CILR_NEVER ||
1059 (ols->ols_locklessable && osc_object_is_contended(oob)) ||
1060 /* lockless truncate */
1061 (io->ci_type == CIT_TRUNC &&
1062 (ocd->ocd_connect_flags & OBD_CONNECT_TRUNCLOCK) &&
1063 osd->od_lockless_truncate)) {
1064 ols->ols_locklessable = 1;
1065 slice->cls_ops = &osc_lock_lockless_ops;
1068 LASSERT(ergo(ols->ols_glimpse, !osc_lock_is_lockless(ols)));
1072 * Cancel all conflicting locks and wait for them to be destroyed.
1074 * This function is used for two purposes:
1076 * - early cancel all conflicting locks before starting IO, and
1078 * - guarantee that pages added to the page cache by lockless IO are never
1079 * covered by locks other than lockless IO lock, and, hence, are not
1080 * visible to other threads.
1082 static int osc_lock_enqueue_wait(const struct lu_env *env,
1083 const struct osc_lock *olck)
1085 struct cl_lock *lock = olck->ols_cl.cls_lock;
1086 struct cl_lock_descr *descr = &lock->cll_descr;
1087 struct cl_object_header *hdr = cl_object_header(descr->cld_obj);
1088 struct cl_lock_closure *closure = &osc_env_info(env)->oti_closure;
1089 struct cl_lock *scan;
1090 struct cl_lock *temp;
1091 int lockless = osc_lock_is_lockless(olck);
1097 LASSERT(cl_lock_is_mutexed(lock));
1098 LASSERT(lock->cll_state == CLS_QUEUING);
1101 * XXX This function could be sped up if we had asynchronous
1106 !(olck->ols_flags & LDLM_FL_BLOCK_NOWAIT) &&
1107 cl_lock_nr_mutexed(env) == 1;
1108 cl_lock_closure_init(env, closure, lock, canwait);
1109 spin_lock(&hdr->coh_lock_guard);
1110 list_for_each_entry_safe(scan, temp, &hdr->coh_locks, cll_linkage) {
1114 if (scan->cll_state < CLS_QUEUING ||
1115 scan->cll_state == CLS_FREEING ||
1116 scan->cll_descr.cld_start > descr->cld_end ||
1117 scan->cll_descr.cld_end < descr->cld_start)
1120 /* overlapped and living locks. */
1122 /* We're not supposed to give up group lock. */
1123 if (scan->cll_descr.cld_mode == CLM_GROUP) {
1124 LASSERT(descr->cld_mode != CLM_GROUP ||
1125 descr->cld_gid != scan->cll_descr.cld_gid);
1129 /* A tricky case for lockless pages:
1130 * We need to cancel the compatible locks if we're enqueuing
1131 * a lockless lock, for example:
1132 * imagine that client has PR lock on [0, 1000], and thread T0
1133 * is doing lockless IO in [500, 1500] region. Concurrent
1134 * thread T1 can see lockless data in [500, 1000], which is
1135 * wrong, because these data are possibly stale.
1137 if (!lockless && cl_lock_compatible(scan, lock))
1140 /* Now @scan is conflicting with @lock, this means current
1141 * thread have to sleep for @scan being destroyed. */
1142 cl_lock_get_trust(scan);
1143 if (&temp->cll_linkage != &hdr->coh_locks)
1144 cl_lock_get_trust(temp);
1145 spin_unlock(&hdr->coh_lock_guard);
1146 lu_ref_add(&scan->cll_reference, "cancel-wait", lock);
1148 LASSERT(list_empty(&closure->clc_list));
1149 rc = cl_lock_closure_build(env, scan, closure);
1151 rc = osc_lock_cancel_wait(env, lock, scan, canwait);
1152 cl_lock_disclosure(env, closure);
1153 if (rc == -EWOULDBLOCK)
1156 if (rc == CLO_REPEAT && !canwait)
1157 /* cannot wait... no early cancellation. */
1160 lu_ref_del(&scan->cll_reference, "cancel-wait", lock);
1161 cl_lock_put(env, scan);
1162 spin_lock(&hdr->coh_lock_guard);
1164 * Lock list could have been modified, while spin-lock was
1165 * released. Check that it is safe to continue.
1167 stop = list_empty(&temp->cll_linkage);
1168 if (&temp->cll_linkage != &hdr->coh_locks)
1169 cl_lock_put(env, temp);
1170 if (stop || rc != 0)
1173 spin_unlock(&hdr->coh_lock_guard);
1174 cl_lock_closure_fini(closure);
1179 * Deadlock avoidance for osc_lock_enqueue(). Consider following scenario:
1181 * - Thread0: obtains PR:[0, 10]. Lock is busy.
1183 * - Thread1: enqueues PW:[5, 50]. Blocking ast is sent to
1184 * PR:[0, 10], but cancellation of busy lock is postponed.
1186 * - Thread0: enqueue PR:[30, 40]. Lock is locally matched to
1187 * PW:[5, 50], and thread0 waits for the lock completion never
1188 * releasing PR:[0, 10]---deadlock.
1190 * The second PR lock can be glimpse (it is to deal with that situation that
1191 * ll_glimpse_size() has second argument, preventing local match of
1192 * not-yet-granted locks, see bug 10295). Similar situation is possible in the
1193 * case of memory mapped user level buffer.
1195 * To prevent this we can detect a situation when current "thread" or "io"
1196 * already holds a lock on this object and either add LDLM_FL_BLOCK_GRANTED to
1197 * the ols->ols_flags, or prevent local match with PW locks.
1199 static int osc_deadlock_is_possible(const struct lu_env *env,
1200 struct cl_lock *lock)
1202 struct cl_object *obj;
1203 struct cl_object_header *head;
1204 struct cl_lock *scan;
1211 LASSERT(cl_lock_is_mutexed(lock));
1213 oio = osc_env_io(env);
1214 obj = lock->cll_descr.cld_obj;
1215 head = cl_object_header(obj);
1218 spin_lock(&head->coh_lock_guard);
1219 list_for_each_entry(scan, &head->coh_locks, cll_linkage) {
1221 struct osc_lock *oscan;
1223 oscan = osc_lock_at(scan);
1224 LASSERT(oscan != NULL);
1225 if (oscan->ols_owner == oio) {
1231 spin_unlock(&head->coh_lock_guard);
1236 * Implementation of cl_lock_operations::clo_enqueue() method for osc
1237 * layer. This initiates ldlm enqueue:
1239 * - checks for possible dead-lock conditions (osc_deadlock_is_possible());
1241 * - cancels conflicting locks early (osc_lock_enqueue_wait());
1243 * - calls osc_enqueue_base() to do actual enqueue.
1245 * osc_enqueue_base() is supplied with an upcall function that is executed
1246 * when lock is received either after a local cached ldlm lock is matched, or
1247 * when a reply from the server is received.
1249 * This function does not wait for the network communication to complete.
1251 static int osc_lock_enqueue(const struct lu_env *env,
1252 const struct cl_lock_slice *slice,
1253 struct cl_io *_, __u32 enqflags)
1255 struct osc_lock *ols = cl2osc_lock(slice);
1256 struct cl_lock *lock = ols->ols_cl.cls_lock;
1257 struct osc_object *obj = cl2osc(slice->cls_obj);
1258 struct osc_thread_info *info = osc_env_info(env);
1259 struct ldlm_res_id *resname = &info->oti_resname;
1260 ldlm_policy_data_t *policy = &info->oti_policy;
1261 struct ldlm_enqueue_info *einfo = &ols->ols_einfo;
1265 LASSERT(cl_lock_is_mutexed(lock));
1266 LASSERT(lock->cll_state == CLS_QUEUING);
1267 LASSERT(ols->ols_state == OLS_NEW);
1269 osc_lock_build_res(env, obj, resname);
1270 osc_lock_build_policy(env, lock, policy);
1271 ols->ols_flags = osc_enq2ldlm_flags(enqflags);
1272 if (osc_deadlock_is_possible(env, lock))
1273 ols->ols_flags |= LDLM_FL_BLOCK_GRANTED;
1274 if (ols->ols_flags & LDLM_FL_HAS_INTENT)
1275 ols->ols_glimpse = 1;
1277 result = osc_lock_enqueue_wait(env, ols);
1279 if (!(enqflags & CEF_MUST))
1280 /* try to convert this lock to a lockless lock */
1281 osc_lock_to_lockless(env, ols, (enqflags & CEF_NEVER));
1282 if (!osc_lock_is_lockless(ols)) {
1283 if (ols->ols_locklessable)
1284 ols->ols_flags |= LDLM_FL_DENY_ON_CONTENTION;
1286 /* a reference for lock, passed as an upcall cookie */
1288 lu_ref_add(&lock->cll_reference, "upcall", lock);
1289 ols->ols_state = OLS_ENQUEUED;
1292 * XXX: this is possible blocking point as
1293 * ldlm_lock_match(LDLM_FL_LVB_READY) waits for
1296 result = osc_enqueue_base(osc_export(obj), resname,
1297 &ols->ols_flags, policy,
1299 obj->oo_oinfo->loi_kms_valid,
1301 ols, einfo, &ols->ols_handle,
1304 lu_ref_del(&lock->cll_reference,
1306 cl_lock_put(env, lock);
1309 ols->ols_state = OLS_GRANTED;
1312 LASSERT(ergo(ols->ols_glimpse, !osc_lock_is_lockless(ols)));
1316 static int osc_lock_wait(const struct lu_env *env,
1317 const struct cl_lock_slice *slice)
1319 struct osc_lock *olck = cl2osc_lock(slice);
1320 struct cl_lock *lock = olck->ols_cl.cls_lock;
1322 LINVRNT(osc_lock_invariant(olck));
1323 if (olck->ols_glimpse && olck->ols_state >= OLS_UPCALL_RECEIVED)
1326 LASSERT(equi(olck->ols_state >= OLS_UPCALL_RECEIVED &&
1327 lock->cll_error == 0, olck->ols_lock != NULL));
1329 return lock->cll_error ?: olck->ols_state >= OLS_GRANTED ? 0 : CLO_WAIT;
1333 * An implementation of cl_lock_operations::clo_use() method that pins cached
1336 static int osc_lock_use(const struct lu_env *env,
1337 const struct cl_lock_slice *slice)
1339 struct osc_lock *olck = cl2osc_lock(slice);
1342 LASSERT(!olck->ols_hold);
1344 * Atomically check for LDLM_FL_CBPENDING and addref a lock if this
1345 * flag is not set. This protects us from a concurrent blocking ast.
1347 rc = ldlm_lock_addref_try(&olck->ols_handle, olck->ols_einfo.ei_mode);
1349 olck->ols_hold = olck->ols_has_ref = 1;
1350 olck->ols_state = OLS_GRANTED;
1352 struct cl_lock *lock;
1355 * Lock is being cancelled somewhere within
1356 * ldlm_handle_bl_callback(): LDLM_FL_CBPENDING is already
1357 * set, but osc_ldlm_blocking_ast() hasn't yet acquired
1360 lock = slice->cls_lock;
1361 LASSERT(lock->cll_state == CLS_CACHED);
1362 LASSERT(lock->cll_users > 0);
1363 LASSERT(olck->ols_lock->l_flags & LDLM_FL_CBPENDING);
1364 /* set a flag for osc_dlm_blocking_ast0() to signal the
1366 olck->ols_ast_wait = 1;
1372 static int osc_lock_flush(struct osc_lock *ols, int discard)
1374 struct cl_lock *lock = ols->ols_cl.cls_lock;
1375 struct cl_env_nest nest;
1379 env = cl_env_nested_get(&nest);
1381 result = cl_lock_page_out(env, lock, discard);
1382 cl_env_nested_put(&nest, env);
1384 result = PTR_ERR(env);
1391 * Implements cl_lock_operations::clo_cancel() method for osc layer. This is
1392 * called (as part of cl_lock_cancel()) when lock is canceled either voluntary
1393 * (LRU pressure, early cancellation, umount, etc.) or due to the conflict
1394 * with some other lock some where in the cluster. This function does the
1397 * - invalidates all pages protected by this lock (after sending dirty
1398 * ones to the server, as necessary);
1400 * - decref's underlying ldlm lock;
1402 * - cancels ldlm lock (ldlm_cli_cancel()).
1404 static void osc_lock_cancel(const struct lu_env *env,
1405 const struct cl_lock_slice *slice)
1407 struct cl_lock *lock = slice->cls_lock;
1408 struct osc_lock *olck = cl2osc_lock(slice);
1409 struct ldlm_lock *dlmlock = olck->ols_lock;
1413 LASSERT(cl_lock_is_mutexed(lock));
1414 LINVRNT(osc_lock_invariant(olck));
1416 if (dlmlock != NULL) {
1419 discard = dlmlock->l_flags & LDLM_FL_DISCARD_DATA;
1420 result = osc_lock_flush(olck, discard);
1422 osc_lock_unuse(env, slice);
1424 lock_res_and_lock(dlmlock);
1425 /* Now that we're the only user of dlm read/write reference,
1426 * mostly the ->l_readers + ->l_writers should be zero.
1427 * However, there is a corner case.
1428 * See bug 18829 for details.*/
1429 do_cancel = (dlmlock->l_readers == 0 &&
1430 dlmlock->l_writers == 0);
1431 dlmlock->l_flags |= LDLM_FL_CBPENDING;
1432 unlock_res_and_lock(dlmlock);
1434 result = ldlm_cli_cancel(&olck->ols_handle);
1436 CL_LOCK_DEBUG(D_ERROR, env, lock,
1437 "lock %p cancel failure with error(%d)\n",
1440 olck->ols_state = OLS_CANCELLED;
1441 osc_lock_detach(env, olck);
1444 void cl_lock_page_list_fixup(const struct lu_env *env,
1445 struct cl_io *io, struct cl_lock *lock,
1446 struct cl_page_list *queue);
1448 #ifdef INVARIANT_CHECK
1450 * Returns true iff there are pages under \a olck not protected by other
1453 static int osc_lock_has_pages(struct osc_lock *olck)
1455 struct cl_lock *lock;
1456 struct cl_lock_descr *descr;
1457 struct cl_object *obj;
1458 struct osc_object *oob;
1459 struct cl_page_list *plist;
1460 struct cl_page *page;
1461 struct cl_env_nest nest;
1466 env = cl_env_nested_get(&nest);
1468 obj = olck->ols_cl.cls_obj;
1470 io = &oob->oo_debug_io;
1471 lock = olck->ols_cl.cls_lock;
1472 descr = &lock->cll_descr;
1473 plist = &osc_env_info(env)->oti_plist;
1474 cl_page_list_init(plist);
1476 mutex_lock(&oob->oo_debug_mutex);
1478 io->ci_obj = cl_object_top(obj);
1479 cl_io_init(env, io, CIT_MISC, io->ci_obj);
1480 cl_page_gang_lookup(env, obj, io,
1481 descr->cld_start, descr->cld_end, plist);
1482 cl_lock_page_list_fixup(env, io, lock, plist);
1483 if (plist->pl_nr > 0) {
1484 CL_LOCK_DEBUG(D_ERROR, env, lock, "still has pages\n");
1485 cl_page_list_for_each(page, plist)
1486 CL_PAGE_DEBUG(D_ERROR, env, page, "\n");
1488 result = plist->pl_nr > 0;
1489 cl_page_list_disown(env, io, plist);
1490 cl_page_list_fini(env, plist);
1491 cl_io_fini(env, io);
1492 mutex_unlock(&oob->oo_debug_mutex);
1493 cl_env_nested_put(&nest, env);
1499 # define osc_lock_has_pages(olck) (0)
1500 #endif /* INVARIANT_CHECK */
1502 static void osc_lock_delete(const struct lu_env *env,
1503 const struct cl_lock_slice *slice)
1505 struct osc_lock *olck;
1507 olck = cl2osc_lock(slice);
1508 LINVRNT(osc_lock_invariant(olck));
1509 LINVRNT(!osc_lock_has_pages(olck));
1512 osc_lock_unuse(env, slice);
1513 osc_lock_detach(env, olck);
1517 * Implements cl_lock_operations::clo_state() method for osc layer.
1519 * Maintains osc_lock::ols_owner field.
1521 * This assumes that lock always enters CLS_HELD (from some other state) in
1522 * the same IO context as one that requested the lock. This should not be a
1523 * problem, because context is by definition shared by all activity pertaining
1524 * to the same high-level IO.
1526 static void osc_lock_state(const struct lu_env *env,
1527 const struct cl_lock_slice *slice,
1528 enum cl_lock_state state)
1530 struct osc_lock *lock = cl2osc_lock(slice);
1531 struct osc_io *oio = osc_env_io(env);
1534 * XXX multiple io contexts can use the lock at the same time.
1536 LINVRNT(osc_lock_invariant(lock));
1537 if (state == CLS_HELD && slice->cls_lock->cll_state != CLS_HELD) {
1538 LASSERT(lock->ols_owner == NULL);
1539 lock->ols_owner = oio;
1540 } else if (state != CLS_HELD)
1541 lock->ols_owner = NULL;
1544 static int osc_lock_print(const struct lu_env *env, void *cookie,
1545 lu_printer_t p, const struct cl_lock_slice *slice)
1547 struct osc_lock *lock = cl2osc_lock(slice);
1550 * XXX print ldlm lock and einfo properly.
1552 (*p)(env, cookie, "%p %08x "LPU64" %d %p ",
1553 lock->ols_lock, lock->ols_flags, lock->ols_handle.cookie,
1554 lock->ols_state, lock->ols_owner);
1555 osc_lvb_print(env, cookie, p, &lock->ols_lvb);
1559 static const struct cl_lock_operations osc_lock_ops = {
1560 .clo_fini = osc_lock_fini,
1561 .clo_enqueue = osc_lock_enqueue,
1562 .clo_wait = osc_lock_wait,
1563 .clo_unuse = osc_lock_unuse,
1564 .clo_use = osc_lock_use,
1565 .clo_delete = osc_lock_delete,
1566 .clo_state = osc_lock_state,
1567 .clo_cancel = osc_lock_cancel,
1568 .clo_weigh = osc_lock_weigh,
1569 .clo_print = osc_lock_print
1572 static int osc_lock_lockless_enqueue(const struct lu_env *env,
1573 const struct cl_lock_slice *slice,
1574 struct cl_io *_, __u32 enqflags)
1580 static int osc_lock_lockless_unuse(const struct lu_env *env,
1581 const struct cl_lock_slice *slice)
1583 struct osc_lock *ols = cl2osc_lock(slice);
1584 struct cl_lock *lock = slice->cls_lock;
1586 LASSERT(ols->ols_state == OLS_GRANTED);
1587 LINVRNT(osc_lock_invariant(ols));
1589 cl_lock_cancel(env, lock);
1590 cl_lock_delete(env, lock);
1594 static void osc_lock_lockless_cancel(const struct lu_env *env,
1595 const struct cl_lock_slice *slice)
1597 struct osc_lock *ols = cl2osc_lock(slice);
1600 result = osc_lock_flush(ols, 0);
1602 CERROR("Pages for lockless lock %p were not purged(%d)\n",
1604 ols->ols_state = OLS_CANCELLED;
1607 static int osc_lock_lockless_wait(const struct lu_env *env,
1608 const struct cl_lock_slice *slice)
1610 struct osc_lock *olck = cl2osc_lock(slice);
1611 struct cl_lock *lock = olck->ols_cl.cls_lock;
1613 LINVRNT(osc_lock_invariant(olck));
1614 LASSERT(olck->ols_state >= OLS_UPCALL_RECEIVED);
1616 return lock->cll_error;
1619 static void osc_lock_lockless_state(const struct lu_env *env,
1620 const struct cl_lock_slice *slice,
1621 enum cl_lock_state state)
1623 struct osc_lock *lock = cl2osc_lock(slice);
1624 struct osc_io *oio = osc_env_io(env);
1626 LINVRNT(osc_lock_invariant(lock));
1627 if (state == CLS_HELD) {
1628 LASSERT(lock->ols_owner == NULL);
1629 lock->ols_owner = oio;
1631 /* set the io to be lockless if this lock is for io's
1633 if (cl_object_same(oio->oi_cl.cis_obj, slice->cls_obj))
1634 oio->oi_lockless = 1;
1636 lock->ols_owner = NULL;
1639 static int osc_lock_lockless_fits_into(const struct lu_env *env,
1640 const struct cl_lock_slice *slice,
1641 const struct cl_lock_descr *need,
1642 const struct cl_io *io)
1647 static const struct cl_lock_operations osc_lock_lockless_ops = {
1648 .clo_fini = osc_lock_fini,
1649 .clo_enqueue = osc_lock_lockless_enqueue,
1650 .clo_wait = osc_lock_lockless_wait,
1651 .clo_unuse = osc_lock_lockless_unuse,
1652 .clo_state = osc_lock_lockless_state,
1653 .clo_fits_into = osc_lock_lockless_fits_into,
1654 .clo_cancel = osc_lock_lockless_cancel,
1655 .clo_print = osc_lock_print
1658 int osc_lock_init(const struct lu_env *env,
1659 struct cl_object *obj, struct cl_lock *lock,
1660 const struct cl_io *_)
1662 struct osc_lock *clk;
1665 OBD_SLAB_ALLOC_PTR_GFP(clk, osc_lock_kmem, CFS_ALLOC_IO);
1667 osc_lock_build_einfo(env, lock, clk, &clk->ols_einfo);
1668 clk->ols_state = OLS_NEW;
1669 cl_lock_slice_add(lock, &clk->ols_cl, obj, &osc_lock_ops);