4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2016, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
33 * lustre/include/lustre_osc.h
35 * OSC layer structures and methods common for both OSC and MDC.
37 * This file contains OSC interfaces used by OSC and MDC. Most of them
38 * were just moved from lustre/osc/osc_cl_internal.h for Data-on-MDT
41 * Author: Nikita Danilov <nikita.danilov@sun.com>
42 * Author: Jinshan Xiong <jinshan.xiong@whamcloud.com>
43 * Author: Mikhail Pershin <mike.pershin@intel.com>
49 #include <libcfs/libcfs.h>
51 #include <cl_object.h>
57 struct osc_quota_info {
58 /** linkage for quota hash table */
59 struct hlist_node oqi_hash;
64 ASYNC_READY = 0x1, /* ap_make_ready will not be called before this
65 page is added to an rpc */
66 ASYNC_URGENT = 0x2, /* page must be put into an RPC before return */
67 ASYNC_COUNT_STABLE = 0x4, /* ap_refresh_count will not be called
68 to give the caller a chance to update
69 or cancel the size of the io */
73 struct osc_async_page {
75 unsigned short oap_cmd;
76 unsigned short oap_interrupted:1;
78 struct list_head oap_pending_item;
79 struct list_head oap_rpc_item;
82 unsigned oap_page_off;
83 enum async_flags oap_async_flags;
85 struct brw_page oap_brw_page;
87 struct ptlrpc_request *oap_request;
88 struct client_obd *oap_cli;
89 struct osc_object *oap_obj;
94 #define oap_page oap_brw_page.pg
95 #define oap_count oap_brw_page.count
96 #define oap_brw_flags oap_brw_page.flag
98 static inline struct osc_async_page *brw_page2oap(struct brw_page *pga)
100 return container_of(pga, struct osc_async_page, oap_brw_page);
103 struct osc_cache_waiter {
104 struct list_head ocw_entry;
105 wait_queue_head_t ocw_waitq;
106 struct osc_async_page *ocw_oap;
112 struct cl_device od_cl;
113 struct obd_export *od_exp;
115 /* Write stats is actually protected by client_obd's lock. */
117 uint64_t os_lockless_writes; /* by bytes */
118 uint64_t os_lockless_reads; /* by bytes */
119 uint64_t os_lockless_truncates; /* by times */
122 /* configuration item(s) */
123 int od_contention_time;
124 int od_lockless_truncate;
130 * State maintained by osc layer for each IO context.
134 struct cl_io_slice oi_cl;
135 /** true if this io is lockless. */
136 unsigned int oi_lockless:1,
137 /** true if this io is counted as active IO */
139 /** how many LRU pages are reserved for this IO */
140 unsigned long oi_lru_reserved;
142 /** active extents, we know how many bytes is going to be written,
143 * so having an active extent will prevent it from being fragmented */
144 struct osc_extent *oi_active;
145 /** partially truncated extent, we need to hold this extent to prevent
146 * page writeback from happening. */
147 struct osc_extent *oi_trunc;
148 /** write osc_lock for this IO, used by osc_extent_find(). */
149 struct osc_lock *oi_write_osclock;
151 struct osc_async_cbargs {
154 struct completion opc_sync;
159 * State maintained by osc layer for the duration of a system call.
165 #define OTI_PVEC_SIZE 256
166 struct osc_thread_info {
167 struct ldlm_res_id oti_resname;
168 union ldlm_policy_data oti_policy;
169 struct cl_lock_descr oti_descr;
170 struct cl_attr oti_attr;
171 struct lustre_handle oti_handle;
172 struct cl_page_list oti_plist;
174 void *oti_pvec[OTI_PVEC_SIZE];
176 * Fields used by cl_lock_discard_pages().
178 pgoff_t oti_next_index;
179 pgoff_t oti_fn_index; /* first non-overlapped index */
180 struct cl_sync_io oti_anchor;
181 struct cl_req_attr oti_req_attr;
182 struct lu_buf oti_ladvise_buf;
186 struct cl_object oo_cl;
187 struct lov_oinfo *oo_oinfo;
189 * True if locking against this stripe got -EUSERS.
192 cfs_time_t oo_contention_time;
193 #ifdef CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK
195 * IO context used for invariant checks in osc_lock_has_pages().
197 struct cl_io oo_debug_io;
198 /** Serialization object for osc_object::oo_debug_io. */
199 struct mutex oo_debug_mutex;
202 * used by the osc to keep track of what objects to build into rpcs.
203 * Protected by client_obd->cli_loi_list_lock.
205 struct list_head oo_ready_item;
206 struct list_head oo_hp_ready_item;
207 struct list_head oo_write_item;
208 struct list_head oo_read_item;
211 * extent is a red black tree to manage (async) dirty pages.
213 struct rb_root oo_root;
215 * Manage write(dirty) extents.
217 struct list_head oo_hp_exts; /* list of hp extents */
218 struct list_head oo_urgent_exts; /* list of writeback extents */
219 struct list_head oo_full_exts;
221 struct list_head oo_reading_exts;
223 atomic_t oo_nr_reads;
224 atomic_t oo_nr_writes;
226 /** Protect extent tree. Will be used to protect
227 * oo_{read|write}_pages soon. */
231 * Radix tree for caching pages
233 spinlock_t oo_tree_lock;
234 struct radix_tree_root oo_tree;
235 unsigned long oo_npages;
237 /* Protect osc_lock this osc_object has */
238 struct list_head oo_ol_list;
239 spinlock_t oo_ol_spin;
241 /** number of active IOs of this object */
243 wait_queue_head_t oo_io_waitq;
248 static inline void osc_object_lock(struct osc_object *obj)
250 spin_lock(&obj->oo_lock);
253 static inline int osc_object_trylock(struct osc_object *obj)
255 return spin_trylock(&obj->oo_lock);
258 static inline void osc_object_unlock(struct osc_object *obj)
260 spin_unlock(&obj->oo_lock);
263 static inline int osc_object_is_locked(struct osc_object *obj)
265 #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
266 return spin_is_locked(&obj->oo_lock);
269 * It is not perfect to return true all the time.
270 * But since this function is only used for assertion
271 * and checking, it seems OK.
278 * Lock "micro-states" for osc layer.
280 enum osc_lock_state {
289 * osc-private state of cl_lock.
291 * Interaction with DLM.
293 * Once receive upcall is invoked, osc_lock remembers a handle of DLM lock in
294 * osc_lock::ols_handle and a pointer to that lock in osc_lock::ols_dlmlock.
296 * This pointer is protected through a reference, acquired by
297 * osc_lock_upcall0(). Also, an additional reference is acquired by
298 * ldlm_lock_addref() call protecting the lock from cancellation, until
299 * osc_lock_unuse() releases it.
301 * Below is a description of how lock references are acquired and released
304 * - When new lock is created and enqueued to the server (ldlm_cli_enqueue())
305 * - ldlm_lock_create()
306 * - ldlm_lock_new(): initializes a lock with 2 references. One for
307 * the caller (released when reply from the server is received, or on
308 * error), and another for the hash table.
309 * - ldlm_lock_addref_internal(): protects the lock from cancellation.
311 * - When reply is received from the server (osc_enqueue_interpret())
312 * - ldlm_cli_enqueue_fini()
313 * - LDLM_LOCK_PUT(): releases caller reference acquired by
316 * ldlm_lock_decref(): error case: matches ldlm_cli_enqueue().
317 * - ldlm_lock_decref(): for async locks, matches ldlm_cli_enqueue().
319 * - When lock is being cancelled (ldlm_lock_cancel())
320 * - ldlm_lock_destroy()
321 * - LDLM_LOCK_PUT(): releases hash-table reference acquired by
324 * osc_lock is detached from ldlm_lock by osc_lock_detach() that is called
325 * either when lock is cancelled (osc_lock_blocking()), or when locks is
326 * deleted without cancellation (e.g., from cl_locks_prune()). In the latter
327 * case ldlm lock remains in memory, and can be re-attached to osc_lock in the
331 struct cl_lock_slice ols_cl;
332 /** Internal lock to protect states, etc. */
334 /** Owner sleeps on this channel for state change */
335 struct cl_sync_io *ols_owner;
336 /** waiting list for this lock to be cancelled */
337 struct list_head ols_waiting_list;
338 /** wait entry of ols_waiting_list */
339 struct list_head ols_wait_entry;
340 /** list entry for osc_object::oo_ol_list */
341 struct list_head ols_nextlock_oscobj;
343 /** underlying DLM lock */
344 struct ldlm_lock *ols_dlmlock;
345 /** DLM flags with which osc_lock::ols_lock was enqueued */
347 /** osc_lock::ols_lock handle */
348 struct lustre_handle ols_handle;
349 struct ldlm_enqueue_info ols_einfo;
350 enum osc_lock_state ols_state;
351 /** lock value block */
352 struct ost_lvb ols_lvb;
355 * true, if ldlm_lock_addref() was called against
356 * osc_lock::ols_lock. This is used for sanity checking.
358 * \see osc_lock::ols_has_ref
360 unsigned ols_hold :1,
362 * this is much like osc_lock::ols_hold, except that this bit is
363 * cleared _after_ reference in released in osc_lock_unuse(). This
364 * fine distinction is needed because:
366 * - if ldlm lock still has a reference, osc_ast_data_get() needs
367 * to return associated cl_lock (so that a flag is needed that is
368 * cleared after ldlm_lock_decref() returned), and
370 * - ldlm_lock_decref() can invoke blocking ast (for a
371 * LDLM_FL_CBPENDING lock), and osc_lock functions like
372 * osc_lock_cancel() called from there need to know whether to
373 * release lock reference (so that a flag is needed that is
374 * cleared before ldlm_lock_decref() is called).
378 * inherit the lockless attribute from top level cl_io.
379 * If true, osc_lock_enqueue is able to tolerate the -EUSERS error.
383 * if set, the osc_lock is a glimpse lock. For glimpse locks, we treat
384 * the EVAVAIL error as torerable, this will make upper logic happy
385 * to wait all glimpse locks to each OSTs to be completed.
386 * Glimpse lock converts to normal lock if the server lock is granted.
387 * Glimpse lock should be destroyed immediately after use.
391 * For async glimpse lock.
395 * for speculative locks - asynchronous glimpse locks and ladvise
396 * lockahead manual lock requests
398 * Used to tell osc layer to not wait for the ldlm reply from the
399 * server, so the osc lock will be short lived - It only exists to
400 * create the ldlm request and is not updated on request completion.
407 * Page state private for osc layer.
410 struct cl_page_slice ops_cl;
412 * Page queues used by osc to detect when RPC can be formed.
414 struct osc_async_page ops_oap;
416 * An offset within page from which next transfer starts. This is used
417 * by cl_page_clip() to submit partial page transfers.
421 * An offset within page at which next transfer ends.
423 * \see osc_page::ops_from.
427 * Boolean, true iff page is under transfer. Used for sanity checking.
429 unsigned ops_transfer_pinned:1,
435 * Set if the page must be transferred with OBD_BRW_SRVLOCK.
439 * lru page list. See osc_lru_{del|use}() in osc_page.c for usage.
441 struct list_head ops_lru;
443 * Submit time - the time when the page is starting RPC. For debugging.
445 cfs_time_t ops_submit_time;
448 extern struct kmem_cache *osc_lock_kmem;
449 extern struct kmem_cache *osc_object_kmem;
450 extern struct kmem_cache *osc_thread_kmem;
451 extern struct kmem_cache *osc_session_kmem;
452 extern struct kmem_cache *osc_extent_kmem;
453 extern struct kmem_cache *osc_quota_kmem;
455 extern struct lu_context_key osc_key;
456 extern struct lu_context_key osc_session_key;
458 #define OSC_FLAGS (ASYNC_URGENT|ASYNC_READY)
461 int osc_page_init(const struct lu_env *env, struct cl_object *obj,
462 struct cl_page *page, pgoff_t ind);
463 void osc_index2policy(union ldlm_policy_data *policy, const struct cl_object *obj,
464 pgoff_t start, pgoff_t end);
465 void osc_lru_add_batch(struct client_obd *cli, struct list_head *list);
466 void osc_page_submit(const struct lu_env *env, struct osc_page *opg,
467 enum cl_req_type crt, int brw_flags);
468 int lru_queue_work(const struct lu_env *env, void *data);
471 int osc_cancel_async_page(const struct lu_env *env, struct osc_page *ops);
472 int osc_set_async_flags(struct osc_object *obj, struct osc_page *opg,
474 int osc_prep_async_page(struct osc_object *osc, struct osc_page *ops,
475 struct page *page, loff_t offset);
476 int osc_queue_async_io(const struct lu_env *env, struct cl_io *io,
477 struct osc_page *ops);
478 int osc_page_cache_add(const struct lu_env *env,
479 const struct cl_page_slice *slice, struct cl_io *io);
480 int osc_teardown_async_page(const struct lu_env *env, struct osc_object *obj,
481 struct osc_page *ops);
482 int osc_flush_async_page(const struct lu_env *env, struct cl_io *io,
483 struct osc_page *ops);
484 int osc_queue_sync_pages(const struct lu_env *env, struct osc_object *obj,
485 struct list_head *list, int cmd, int brw_flags);
486 int osc_cache_truncate_start(const struct lu_env *env, struct osc_object *obj,
487 __u64 size, struct osc_extent **extp);
488 void osc_cache_truncate_end(const struct lu_env *env, struct osc_extent *ext);
489 int osc_cache_writeback_range(const struct lu_env *env, struct osc_object *obj,
490 pgoff_t start, pgoff_t end, int hp, int discard);
491 int osc_cache_wait_range(const struct lu_env *env, struct osc_object *obj,
492 pgoff_t start, pgoff_t end);
493 int osc_io_unplug0(const struct lu_env *env, struct client_obd *cli,
494 struct osc_object *osc, int async);
496 static inline int osc_io_unplug_async(const struct lu_env *env,
497 struct client_obd *cli,
498 struct osc_object *osc)
500 return osc_io_unplug0(env, cli, osc, 1);
503 static inline void osc_io_unplug(const struct lu_env *env,
504 struct client_obd *cli,
505 struct osc_object *osc)
507 (void)osc_io_unplug0(env, cli, osc, 0);
510 void osc_object_set_contended(struct osc_object *obj);
511 void osc_object_clear_contended(struct osc_object *obj);
512 int osc_object_is_contended(struct osc_object *obj);
513 int osc_lock_is_lockless(const struct osc_lock *olck);
516 int osc_device_init(const struct lu_env *env, struct lu_device *d,
517 const char *name, struct lu_device *next);
518 struct lu_device *osc_device_fini(const struct lu_env *env,
519 struct lu_device *d);
520 struct lu_device *osc_device_free(const struct lu_env *env,
521 struct lu_device *d);
524 int osc_object_init(const struct lu_env *env, struct lu_object *obj,
525 const struct lu_object_conf *conf);
526 void osc_object_free(const struct lu_env *env, struct lu_object *obj);
527 int osc_lvb_print(const struct lu_env *env, void *cookie,
528 lu_printer_t p, const struct ost_lvb *lvb);
529 int osc_object_print(const struct lu_env *env, void *cookie,
530 lu_printer_t p, const struct lu_object *obj);
531 int osc_attr_get(const struct lu_env *env, struct cl_object *obj,
532 struct cl_attr *attr);
533 int osc_attr_update(const struct lu_env *env, struct cl_object *obj,
534 const struct cl_attr *attr, unsigned valid);
535 int osc_object_glimpse(const struct lu_env *env, const struct cl_object *obj,
536 struct ost_lvb *lvb);
537 int osc_object_invalidate(const struct lu_env *env, struct osc_object *osc);
540 void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd);
541 int osc_setup_common(struct obd_device *obd, struct lustre_cfg *lcfg);
542 int osc_precleanup_common(struct obd_device *obd);
543 int osc_cleanup_common(struct obd_device *obd);
544 int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
545 u32 keylen, void *key, u32 vallen, void *val,
546 struct ptlrpc_request_set *set);
547 int osc_ldlm_resource_invalidate(struct cfs_hash *hs, struct cfs_hash_bd *bd,
548 struct hlist_node *hnode, void *arg);
550 int osc_punch_send(struct obd_export *exp, struct obdo *oa,
551 obd_enqueue_update_f upcall, void *cookie);
554 int osc_io_submit(const struct lu_env *env, const struct cl_io_slice *ios,
555 enum cl_req_type crt, struct cl_2queue *queue);
556 int osc_io_commit_async(const struct lu_env *env,
557 const struct cl_io_slice *ios,
558 struct cl_page_list *qin, int from, int to,
560 int osc_io_iter_init(const struct lu_env *env, const struct cl_io_slice *ios);
561 void osc_io_iter_fini(const struct lu_env *env,
562 const struct cl_io_slice *ios);
563 int osc_io_write_iter_init(const struct lu_env *env,
564 const struct cl_io_slice *ios);
565 void osc_io_write_iter_fini(const struct lu_env *env,
566 const struct cl_io_slice *ios);
567 int osc_io_fault_start(const struct lu_env *env, const struct cl_io_slice *ios);
568 void osc_io_setattr_end(const struct lu_env *env,
569 const struct cl_io_slice *slice);
570 int osc_io_read_start(const struct lu_env *env,
571 const struct cl_io_slice *slice);
572 int osc_io_write_start(const struct lu_env *env,
573 const struct cl_io_slice *slice);
574 void osc_io_end(const struct lu_env *env, const struct cl_io_slice *slice);
576 int osc_io_fsync_start(const struct lu_env *env,
577 const struct cl_io_slice *slice);
578 void osc_io_fsync_end(const struct lu_env *env,
579 const struct cl_io_slice *slice);
581 /*****************************************************************************
583 * Accessors and type conversions.
586 static inline struct osc_thread_info *osc_env_info(const struct lu_env *env)
588 struct osc_thread_info *info;
590 info = lu_context_key_get(&env->le_ctx, &osc_key);
591 LASSERT(info != NULL);
595 static inline struct osc_session *osc_env_session(const struct lu_env *env)
597 struct osc_session *ses;
599 ses = lu_context_key_get(env->le_ses, &osc_session_key);
600 LASSERT(ses != NULL);
604 static inline struct osc_io *osc_env_io(const struct lu_env *env)
606 return &osc_env_session(env)->os_io;
609 static inline struct osc_device *lu2osc_dev(const struct lu_device *d)
611 return container_of0(d, struct osc_device, od_cl.cd_lu_dev);
614 static inline struct obd_export *osc_export(const struct osc_object *obj)
616 return lu2osc_dev(obj->oo_cl.co_lu.lo_dev)->od_exp;
619 static inline struct client_obd *osc_cli(const struct osc_object *obj)
621 return &osc_export(obj)->exp_obd->u.cli;
624 static inline struct osc_object *cl2osc(const struct cl_object *obj)
626 return container_of0(obj, struct osc_object, oo_cl);
629 static inline struct cl_object *osc2cl(const struct osc_object *obj)
631 return (struct cl_object *)&obj->oo_cl;
634 static inline struct osc_device *obd2osc_dev(const struct obd_device *d)
636 return container_of0(d->obd_lu_dev, struct osc_device,
640 static inline struct lu_device *osc2lu_dev(struct osc_device *osc)
642 return &osc->od_cl.cd_lu_dev;
645 static inline struct lu_object *osc2lu(struct osc_object *osc)
647 return &osc->oo_cl.co_lu;
650 static inline struct osc_object *lu2osc(const struct lu_object *obj)
652 return container_of0(obj, struct osc_object, oo_cl.co_lu);
655 static inline struct osc_io *cl2osc_io(const struct lu_env *env,
656 const struct cl_io_slice *slice)
658 struct osc_io *oio = container_of0(slice, struct osc_io, oi_cl);
660 LINVRNT(oio == osc_env_io(env));
664 static inline enum ldlm_mode osc_cl_lock2ldlm(enum cl_lock_mode mode)
666 LASSERT(mode == CLM_READ || mode == CLM_WRITE || mode == CLM_GROUP);
667 if (mode == CLM_READ)
669 if (mode == CLM_WRITE)
674 static inline enum cl_lock_mode osc_ldlm2cl_lock(enum ldlm_mode mode)
676 LASSERT(mode == LCK_PR || mode == LCK_PW || mode == LCK_GROUP);
684 static inline struct osc_page *cl2osc_page(const struct cl_page_slice *slice)
686 return container_of0(slice, struct osc_page, ops_cl);
689 static inline struct osc_page *oap2osc(struct osc_async_page *oap)
691 return container_of0(oap, struct osc_page, ops_oap);
694 static inline pgoff_t osc_index(struct osc_page *opg)
696 return opg->ops_cl.cpl_index;
699 static inline struct cl_page *oap2cl_page(struct osc_async_page *oap)
701 return oap2osc(oap)->ops_cl.cpl_page;
704 static inline struct osc_page *oap2osc_page(struct osc_async_page *oap)
706 return (struct osc_page *)container_of(oap, struct osc_page, ops_oap);
709 static inline struct osc_page *
710 osc_cl_page_osc(struct cl_page *page, struct osc_object *osc)
712 const struct cl_page_slice *slice;
714 LASSERT(osc != NULL);
715 slice = cl_object_page_slice(&osc->oo_cl, page);
716 return cl2osc_page(slice);
719 static inline struct osc_lock *cl2osc_lock(const struct cl_lock_slice *slice)
721 return container_of0(slice, struct osc_lock, ols_cl);
724 static inline int osc_io_srvlock(struct osc_io *oio)
726 return (oio->oi_lockless && !oio->oi_cl.cis_io->ci_no_srvlock);
729 enum osc_extent_state {
730 OES_INV = 0, /** extent is just initialized or destroyed */
731 OES_ACTIVE = 1, /** process is using this extent */
732 OES_CACHE = 2, /** extent is ready for IO */
733 OES_LOCKING = 3, /** locking page to prepare IO */
734 OES_LOCK_DONE = 4, /** locking finished, ready to send */
735 OES_RPC = 5, /** in RPC */
736 OES_TRUNC = 6, /** being truncated */
741 * osc_extent data to manage dirty pages.
742 * osc_extent has the following attributes:
743 * 1. all pages in the same must be in one RPC in write back;
744 * 2. # of pages must be less than max_pages_per_rpc - implied by 1;
745 * 3. must be covered by only 1 osc_lock;
746 * 4. exclusive. It's impossible to have overlapped osc_extent.
748 * The lifetime of an extent is from when the 1st page is dirtied to when
749 * all pages inside it are written out.
753 * page lock -> client_obd_list_lock -> object lock(osc_object::oo_lock)
756 /** red-black tree node */
757 struct rb_node oe_node;
758 /** osc_object of this extent */
759 struct osc_object *oe_obj;
760 /** refcount, removed from red-black tree if reaches zero. */
762 /** busy if non-zero */
764 /** link list of osc_object's oo_{hp|urgent|locking}_exts. */
765 struct list_head oe_link;
766 /** state of this extent */
767 enum osc_extent_state oe_state;
768 /** flags for this extent. */
769 unsigned int oe_intree:1,
770 /** 0 is write, 1 is read */
772 /** sync extent, queued by osc_queue_sync_pages() */
774 /** set if this extent has partial, sync pages.
775 * Extents with partial page(s) can't merge with others in RPC */
779 /** an ACTIVE extent is going to be truncated, so when this extent
780 * is released, it will turn into TRUNC state instead of CACHE. */
782 /** this extent should be written asap and someone may wait for the
783 * write to finish. This bit is usually set along with urgent if
784 * the extent was CACHE state.
785 * fsync_wait extent can't be merged because new extent region may
786 * exceed fsync range. */
788 /** covering lock is being canceled */
790 /** this extent should be written back asap. set if one of pages is
791 * called by page WB daemon, or sync write or reading requests. */
793 /** how many grants allocated for this extent.
794 * Grant allocated for this extent. There is no grant allocated
795 * for reading extents and sync write extents. */
796 unsigned int oe_grants;
797 /** # of dirty pages in this extent */
798 unsigned int oe_nr_pages;
799 /** list of pending oap pages. Pages in this list are NOT sorted. */
800 struct list_head oe_pages;
801 /** Since an extent has to be written out in atomic, this is used to
802 * remember the next page need to be locked to write this extent out.
803 * Not used right now.
805 struct osc_page *oe_next_page;
806 /** start and end index of this extent, include start and end
807 * themselves. Page offset here is the page index of osc_pages.
808 * oe_start is used as keyword for red-black tree. */
811 /** maximum ending index of this extent, this is limited by
812 * max_pages_per_rpc, lock extent and chunk size. */
814 /** waitqueue - for those who want to be notified if this extent's
815 * state has changed. */
816 wait_queue_head_t oe_waitq;
817 /** lock covering this extent */
818 struct ldlm_lock *oe_dlmlock;
819 /** terminator of this extent. Must be true if this extent is in IO. */
820 struct task_struct *oe_owner;
821 /** return value of writeback. If somebody is waiting for this extent,
822 * this value can be known by outside world. */
824 /** max pages per rpc when this extent was created */
825 unsigned int oe_mppr;
828 int osc_extent_finish(const struct lu_env *env, struct osc_extent *ext,
830 int osc_extent_release(const struct lu_env *env, struct osc_extent *ext);
832 int osc_lock_discard_pages(const struct lu_env *env, struct osc_object *osc,
833 pgoff_t start, pgoff_t end, bool discard_pages);
835 typedef int (*osc_page_gang_cbt)(const struct lu_env *, struct cl_io *,
836 struct osc_page *, void *);
837 int osc_page_gang_lookup(const struct lu_env *env, struct cl_io *io,
838 struct osc_object *osc, pgoff_t start, pgoff_t end,
839 osc_page_gang_cbt cb, void *cbdata);
842 #endif /* LUSTRE_OSC_H */