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.sun.com/software/products/lustre/docs/GPLv2.pdf
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
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2013, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef __LUSTRE_DT_OBJECT_H
38 #define __LUSTRE_DT_OBJECT_H
41 * Sub-class of lu_object with methods common for "data" objects in OST stack.
43 * Data objects behave like regular files: you can read/write them, get and
44 * set their attributes. Implementation of dt interface is supposed to
45 * implement some form of garbage collection, normally reference counting
48 * Examples: osd (lustre/osd) is an implementation of dt interface.
52 #include <obd_support.h>
54 * super-class definitions.
56 #include <lu_object.h>
58 #include <libcfs/libcfs.h>
61 struct proc_dir_entry;
67 struct dt_index_features;
70 struct ldlm_enqueue_info;
73 MNTOPT_USERXATTR = 0x00000001,
74 MNTOPT_ACL = 0x00000002,
77 struct dt_device_param {
78 unsigned ddp_max_name_len;
79 unsigned ddp_max_nlink;
80 unsigned ddp_block_shift;
82 unsigned ddp_max_ea_size;
84 unsigned long long ddp_maxbytes;
85 /* percentage of available space to reserve for grant error margin */
86 int ddp_grant_reserved;
87 /* per-inode space consumption */
89 /* per-fragment grant overhead to be used by client for grant
95 * Per-transaction commit callback function
97 struct dt_txn_commit_cb;
98 typedef void (*dt_cb_t)(struct lu_env *env, struct thandle *th,
99 struct dt_txn_commit_cb *cb, int err);
101 * Special per-transaction callback for cases when just commit callback
102 * is needed and per-device callback are not convenient to use
104 #define TRANS_COMMIT_CB_MAGIC 0xa0a00a0a
105 #define MAX_COMMIT_CB_STR_LEN 32
107 struct dt_txn_commit_cb {
108 struct list_head dcb_linkage;
111 char dcb_name[MAX_COMMIT_CB_STR_LEN];
115 * Operations on dt device.
117 struct dt_device_operations {
119 * Return device-wide statistics.
121 int (*dt_statfs)(const struct lu_env *env,
122 struct dt_device *dev, struct obd_statfs *osfs);
124 * Create transaction, described by \a param.
126 struct thandle *(*dt_trans_create)(const struct lu_env *env,
127 struct dt_device *dev);
129 * Start transaction, described by \a param.
131 int (*dt_trans_start)(const struct lu_env *env,
132 struct dt_device *dev, struct thandle *th);
134 * Finish previously started transaction.
136 int (*dt_trans_stop)(const struct lu_env *env, struct dt_device *dev,
139 * Add commit callback to the transaction.
141 int (*dt_trans_cb_add)(struct thandle *th,
142 struct dt_txn_commit_cb *dcb);
144 * Return fid of root index object.
146 int (*dt_root_get)(const struct lu_env *env,
147 struct dt_device *dev, struct lu_fid *f);
149 * Return device configuration data.
151 void (*dt_conf_get)(const struct lu_env *env,
152 const struct dt_device *dev,
153 struct dt_device_param *param);
155 * handling device state, mostly for tests
157 int (*dt_sync)(const struct lu_env *env, struct dt_device *dev);
158 int (*dt_ro)(const struct lu_env *env, struct dt_device *dev);
160 * Start a transaction commit asynchronously
162 * \param env environment
163 * \param dev dt_device to start commit on
165 * \return 0 success, negative value if error
167 int (*dt_commit_async)(const struct lu_env *env,
168 struct dt_device *dev);
170 * Initialize capability context.
172 int (*dt_init_capa_ctxt)(const struct lu_env *env,
173 struct dt_device *dev,
174 int mode, unsigned long timeout,
175 __u32 alg, struct lustre_capa_key *keys);
178 struct dt_index_features {
179 /** required feature flags from enum dt_index_flags */
181 /** minimal required key size */
182 size_t dif_keysize_min;
183 /** maximal required key size, 0 if no limit */
184 size_t dif_keysize_max;
185 /** minimal required record size */
186 size_t dif_recsize_min;
187 /** maximal required record size, 0 if no limit */
188 size_t dif_recsize_max;
189 /** pointer size for record */
193 enum dt_index_flags {
194 /** index supports variable sized keys */
195 DT_IND_VARKEY = 1 << 0,
196 /** index supports variable sized records */
197 DT_IND_VARREC = 1 << 1,
198 /** index can be modified */
199 DT_IND_UPDATE = 1 << 2,
200 /** index supports records with non-unique (duplicate) keys */
201 DT_IND_NONUNQ = 1 << 3,
203 * index support fixed-size keys sorted with natural numerical way
204 * and is able to return left-side value if no exact value found
206 DT_IND_RANGE = 1 << 4,
210 * Features, required from index to support file system directories (mapping
213 extern const struct dt_index_features dt_directory_features;
214 extern const struct dt_index_features dt_otable_features;
215 extern const struct dt_index_features dt_lfsck_orphan_features;
216 extern const struct dt_index_features dt_lfsck_features;
218 /* index features supported by the accounting objects */
219 extern const struct dt_index_features dt_acct_features;
221 /* index features supported by the quota global indexes */
222 extern const struct dt_index_features dt_quota_glb_features;
224 /* index features supported by the quota slave indexes */
225 extern const struct dt_index_features dt_quota_slv_features;
228 * This is a general purpose dt allocation hint.
229 * It now contains the parent object.
230 * It can contain any allocation hint in the future.
232 struct dt_allocation_hint {
233 struct dt_object *dah_parent;
234 const void *dah_eadata;
240 * object type specifier.
243 enum dt_format_type {
248 /** for special index */
250 /** for symbolic link */
255 * object format specifier.
257 struct dt_object_format {
258 /** type for dt object */
259 enum dt_format_type dof_type;
269 * special index need feature as parameter to create
273 const struct dt_index_features *di_feat;
278 enum dt_format_type dt_mode_to_dft(__u32 mode);
280 typedef __u64 dt_obj_version_t;
282 union ldlm_policy_data;
285 * Per-dt-object operations.
287 struct dt_object_operations {
288 void (*do_read_lock)(const struct lu_env *env,
289 struct dt_object *dt, unsigned role);
290 void (*do_write_lock)(const struct lu_env *env,
291 struct dt_object *dt, unsigned role);
292 void (*do_read_unlock)(const struct lu_env *env,
293 struct dt_object *dt);
294 void (*do_write_unlock)(const struct lu_env *env,
295 struct dt_object *dt);
296 int (*do_write_locked)(const struct lu_env *env,
297 struct dt_object *dt);
299 * Note: following ->do_{x,}attr_{set,get}() operations are very
300 * similar to ->moo_{x,}attr_{set,get}() operations in struct
301 * md_object_operations (see md_object.h). These operations are not in
302 * lu_object_operations, because ->do_{x,}attr_set() versions take
303 * transaction handle as an argument (this transaction is started by
304 * caller). We might factor ->do_{x,}attr_get() into
305 * lu_object_operations, but that would break existing symmetry.
308 int (*do_declare_attr_get)(const struct lu_env *env,
309 struct dt_object *dt,
310 struct lustre_capa *capa);
312 * Return standard attributes.
314 * precondition: lu_object_exists(&dt->do_lu);
316 int (*do_attr_get)(const struct lu_env *env,
317 struct dt_object *dt, struct lu_attr *attr,
318 struct lustre_capa *capa);
320 * Set standard attributes.
322 * precondition: dt_object_exists(dt);
324 int (*do_declare_attr_set)(const struct lu_env *env,
325 struct dt_object *dt,
326 const struct lu_attr *attr,
327 struct thandle *handle);
328 int (*do_attr_set)(const struct lu_env *env,
329 struct dt_object *dt,
330 const struct lu_attr *attr,
331 struct thandle *handle,
332 struct lustre_capa *capa);
334 int (*do_declare_xattr_get)(const struct lu_env *env,
335 struct dt_object *dt,
338 struct lustre_capa *capa);
341 * Return a value of an extended attribute.
343 * precondition: dt_object_exists(dt);
345 int (*do_xattr_get)(const struct lu_env *env, struct dt_object *dt,
346 struct lu_buf *buf, const char *name,
347 struct lustre_capa *capa);
349 * Set value of an extended attribute.
351 * \a fl - flags from enum lu_xattr_flags
353 * precondition: dt_object_exists(dt);
355 int (*do_declare_xattr_set)(const struct lu_env *env,
356 struct dt_object *dt,
357 const struct lu_buf *buf,
358 const char *name, int fl,
359 struct thandle *handle);
360 int (*do_xattr_set)(const struct lu_env *env,
361 struct dt_object *dt, const struct lu_buf *buf,
362 const char *name, int fl, struct thandle *handle,
363 struct lustre_capa *capa);
365 * Delete existing extended attribute.
367 * precondition: dt_object_exists(dt);
369 int (*do_declare_xattr_del)(const struct lu_env *env,
370 struct dt_object *dt,
371 const char *name, struct thandle *handle);
372 int (*do_xattr_del)(const struct lu_env *env,
373 struct dt_object *dt,
374 const char *name, struct thandle *handle,
375 struct lustre_capa *capa);
377 * Place list of existing extended attributes into \a buf (which has
380 * precondition: dt_object_exists(dt);
382 int (*do_xattr_list)(const struct lu_env *env,
383 struct dt_object *dt, struct lu_buf *buf,
384 struct lustre_capa *capa);
386 * Init allocation hint using parent object and child mode.
387 * (1) The \a parent might be NULL if this is a partial creation for
389 * (2) The type of child is in \a child_mode.
390 * (3) The result hint is stored in \a ah;
392 void (*do_ah_init)(const struct lu_env *env,
393 struct dt_allocation_hint *ah,
394 struct dt_object *parent,
395 struct dt_object *child,
398 * Create new object on this device.
400 * precondition: !dt_object_exists(dt);
401 * postcondition: ergo(result == 0, dt_object_exists(dt));
403 int (*do_declare_create)(const struct lu_env *env,
404 struct dt_object *dt,
405 struct lu_attr *attr,
406 struct dt_allocation_hint *hint,
407 struct dt_object_format *dof,
409 int (*do_create)(const struct lu_env *env, struct dt_object *dt,
410 struct lu_attr *attr,
411 struct dt_allocation_hint *hint,
412 struct dt_object_format *dof,
416 Destroy object on this device
417 * precondition: !dt_object_exists(dt);
418 * postcondition: ergo(result == 0, dt_object_exists(dt));
420 int (*do_declare_destroy)(const struct lu_env *env,
421 struct dt_object *dt,
423 int (*do_destroy)(const struct lu_env *env, struct dt_object *dt,
427 * Announce that this object is going to be used as an index. This
428 * operation check that object supports indexing operations and
429 * installs appropriate dt_index_operations vector on success.
431 * Also probes for features. Operation is successful if all required
432 * features are supported.
434 int (*do_index_try)(const struct lu_env *env,
435 struct dt_object *dt,
436 const struct dt_index_features *feat);
438 * Add nlink of the object
439 * precondition: dt_object_exists(dt);
441 int (*do_declare_ref_add)(const struct lu_env *env,
442 struct dt_object *dt, struct thandle *th);
443 int (*do_ref_add)(const struct lu_env *env,
444 struct dt_object *dt, struct thandle *th);
446 * Del nlink of the object
447 * precondition: dt_object_exists(dt);
449 int (*do_declare_ref_del)(const struct lu_env *env,
450 struct dt_object *dt, struct thandle *th);
451 int (*do_ref_del)(const struct lu_env *env,
452 struct dt_object *dt, struct thandle *th);
454 struct obd_capa *(*do_capa_get)(const struct lu_env *env,
455 struct dt_object *dt,
456 struct lustre_capa *old,
458 int (*do_object_sync)(const struct lu_env *env, struct dt_object *obj,
459 __u64 start, __u64 end);
461 * Get object info of next level. Currently, only get inode from osd.
462 * This is only used by quota b=16542
463 * precondition: dt_object_exists(dt);
465 int (*do_data_get)(const struct lu_env *env, struct dt_object *dt,
471 int (*do_object_lock)(const struct lu_env *env, struct dt_object *dt,
472 struct lustre_handle *lh,
473 struct ldlm_enqueue_info *einfo,
474 union ldlm_policy_data *policy);
476 int (*do_object_unlock)(const struct lu_env *env, struct dt_object *dt,
477 struct ldlm_enqueue_info *einfo,
478 union ldlm_policy_data *policy);
482 * Per-dt-object operations on "file body".
484 struct dt_body_operations {
486 * precondition: dt_object_exists(dt);
488 ssize_t (*dbo_read)(const struct lu_env *env, struct dt_object *dt,
489 struct lu_buf *buf, loff_t *pos,
490 struct lustre_capa *capa);
492 * precondition: dt_object_exists(dt);
494 ssize_t (*dbo_declare_write)(const struct lu_env *env,
495 struct dt_object *dt,
496 const struct lu_buf *buf, loff_t pos,
497 struct thandle *handle);
498 ssize_t (*dbo_write)(const struct lu_env *env, struct dt_object *dt,
499 const struct lu_buf *buf, loff_t *pos,
500 struct thandle *handle, struct lustre_capa *capa,
503 * methods for zero-copy IO
507 * precondition: dt_object_exists(dt);
511 * > 0 - number of local buffers prepared
513 int (*dbo_bufs_get)(const struct lu_env *env, struct dt_object *dt,
514 loff_t pos, ssize_t len, struct niobuf_local *lb,
515 int rw, struct lustre_capa *capa);
517 * precondition: dt_object_exists(dt);
519 int (*dbo_bufs_put)(const struct lu_env *env, struct dt_object *dt,
520 struct niobuf_local *lb, int nr);
522 * precondition: dt_object_exists(dt);
524 int (*dbo_write_prep)(const struct lu_env *env, struct dt_object *dt,
525 struct niobuf_local *lb, int nr);
527 * precondition: dt_object_exists(dt);
529 int (*dbo_declare_write_commit)(const struct lu_env *env,
530 struct dt_object *dt,
531 struct niobuf_local *,
532 int, struct thandle *);
534 * precondition: dt_object_exists(dt);
536 int (*dbo_write_commit)(const struct lu_env *env, struct dt_object *dt,
537 struct niobuf_local *, int, struct thandle *);
539 * precondition: dt_object_exists(dt);
541 int (*dbo_read_prep)(const struct lu_env *env, struct dt_object *dt,
542 struct niobuf_local *lnb, int nr);
543 int (*dbo_fiemap_get)(const struct lu_env *env, struct dt_object *dt,
544 struct ll_user_fiemap *fm);
546 * Punch object's content
547 * precondition: regular object, not index
549 int (*dbo_declare_punch)(const struct lu_env *, struct dt_object *,
550 __u64, __u64, struct thandle *th);
551 int (*dbo_punch)(const struct lu_env *env, struct dt_object *dt,
552 __u64 start, __u64 end, struct thandle *th,
553 struct lustre_capa *capa);
557 * Incomplete type of index record.
562 * Incomplete type of index key.
567 * Incomplete type of dt iterator.
572 * Per-dt-object operations on object as index.
574 struct dt_index_operations {
576 * precondition: dt_object_exists(dt);
578 int (*dio_lookup)(const struct lu_env *env, struct dt_object *dt,
579 struct dt_rec *rec, const struct dt_key *key,
580 struct lustre_capa *capa);
582 * precondition: dt_object_exists(dt);
584 int (*dio_declare_insert)(const struct lu_env *env,
585 struct dt_object *dt,
586 const struct dt_rec *rec,
587 const struct dt_key *key,
588 struct thandle *handle);
589 int (*dio_insert)(const struct lu_env *env, struct dt_object *dt,
590 const struct dt_rec *rec, const struct dt_key *key,
591 struct thandle *handle, struct lustre_capa *capa,
594 * precondition: dt_object_exists(dt);
596 int (*dio_declare_delete)(const struct lu_env *env,
597 struct dt_object *dt,
598 const struct dt_key *key,
599 struct thandle *handle);
600 int (*dio_delete)(const struct lu_env *env, struct dt_object *dt,
601 const struct dt_key *key, struct thandle *handle,
602 struct lustre_capa *capa);
608 * Allocate and initialize new iterator.
610 * precondition: dt_object_exists(dt);
612 struct dt_it *(*init)(const struct lu_env *env,
613 struct dt_object *dt,
615 struct lustre_capa *capa);
616 void (*fini)(const struct lu_env *env,
618 int (*get)(const struct lu_env *env,
620 const struct dt_key *key);
621 void (*put)(const struct lu_env *env,
623 int (*next)(const struct lu_env *env,
625 struct dt_key *(*key)(const struct lu_env *env,
626 const struct dt_it *di);
627 int (*key_size)(const struct lu_env *env,
628 const struct dt_it *di);
629 int (*rec)(const struct lu_env *env,
630 const struct dt_it *di,
633 int (*rec_size)(const struct lu_env *env,
634 const struct dt_it *di,
636 __u64 (*store)(const struct lu_env *env,
637 const struct dt_it *di);
638 int (*load)(const struct lu_env *env,
639 const struct dt_it *di, __u64 hash);
640 int (*key_rec)(const struct lu_env *env,
641 const struct dt_it *di, void* key_rec);
645 enum dt_otable_it_valid {
646 DOIV_ERROR_HANDLE = 0x0001,
647 DOIV_DRYRUN = 0x0002,
650 enum dt_otable_it_flags {
651 /* Exit when fail. */
652 DOIF_FAILOUT = 0x0001,
654 /* Reset iteration position to the device beginning. */
657 /* There is up layer component uses the iteration. */
658 DOIF_OUTUSED = 0x0004,
660 /* Check only without repairing. */
661 DOIF_DRYRUN = 0x0008,
664 /* otable based iteration needs to use the common DT interation APIs.
665 * To initialize the iteration, it needs call dio_it::init() firstly.
666 * Here is how the otable based iteration should prepare arguments to
667 * call dt_it_ops::init().
669 * For otable based iteration, the 32-bits 'attr' for dt_it_ops::init()
670 * is composed of two parts:
671 * low 16-bits is for valid bits, high 16-bits is for flags bits. */
672 #define DT_OTABLE_IT_FLAGS_SHIFT 16
673 #define DT_OTABLE_IT_FLAGS_MASK 0xffff0000
676 struct lu_device dd_lu_dev;
677 const struct dt_device_operations *dd_ops;
680 * List of dt_txn_callback (see below). This is not protected in any
681 * way, because callbacks are supposed to be added/deleted only during
682 * single-threaded start-up shut-down procedures.
684 struct list_head dd_txn_callbacks;
685 unsigned int dd_record_fid_accessed:1;
688 int dt_device_init(struct dt_device *dev, struct lu_device_type *t);
689 void dt_device_fini(struct dt_device *dev);
691 static inline int lu_device_is_dt(const struct lu_device *d)
693 return ergo(d != NULL, d->ld_type->ldt_tags & LU_DEVICE_DT);
696 static inline struct dt_device * lu2dt_dev(struct lu_device *l)
698 LASSERT(lu_device_is_dt(l));
699 return container_of0(l, struct dt_device, dd_lu_dev);
703 struct lu_object do_lu;
704 const struct dt_object_operations *do_ops;
705 const struct dt_body_operations *do_body_ops;
706 const struct dt_index_operations *do_index_ops;
710 * In-core representation of per-device local object OID storage
712 struct local_oid_storage {
713 /* all initialized llog systems on this node linked by this */
714 struct list_head los_list;
716 /* how many handle's reference this los has */
717 atomic_t los_refcount;
718 struct dt_device *los_dev;
719 struct dt_object *los_obj;
721 /* data used to generate new fids */
722 struct mutex los_id_lock;
727 static inline struct lu_device *dt2lu_dev(struct dt_device *d)
729 return &d->dd_lu_dev;
732 static inline struct dt_object *lu2dt(struct lu_object *l)
734 LASSERT(l == NULL || IS_ERR(l) || lu_device_is_dt(l->lo_dev));
735 return container_of0(l, struct dt_object, do_lu);
738 int dt_object_init(struct dt_object *obj,
739 struct lu_object_header *h, struct lu_device *d);
741 void dt_object_fini(struct dt_object *obj);
743 static inline int dt_object_exists(const struct dt_object *dt)
745 return lu_object_exists(&dt->do_lu);
748 static inline int dt_object_remote(const struct dt_object *dt)
750 return lu_object_remote(&dt->do_lu);
753 static inline struct dt_object *lu2dt_obj(struct lu_object *o)
755 LASSERT(ergo(o != NULL, lu_device_is_dt(o->lo_dev)));
756 return container_of0(o, struct dt_object, do_lu);
759 struct thandle_update {
760 /* In DNE, one transaction can be disassembled into
761 * updates on several different MDTs, and these updates
762 * will be attached to tu_remote_update_list per target.
763 * Only single thread will access the list, no need lock
765 struct list_head tu_remote_update_list;
767 /* sent after or before local transaction */
768 unsigned int tu_sent_after_local_trans:1,
769 tu_only_remote_trans:1;
773 * This is the general purpose transaction handle.
774 * 1. Transaction Life Cycle
775 * This transaction handle is allocated upon starting a new transaction,
776 * and deallocated after this transaction is committed.
777 * 2. Transaction Nesting
778 * We do _NOT_ support nested transaction. So, every thread should only
779 * have one active transaction, and a transaction only belongs to one
780 * thread. Due to this, transaction handle need no reference count.
781 * 3. Transaction & dt_object locking
782 * dt_object locks should be taken inside transaction.
783 * 4. Transaction & RPC
784 * No RPC request should be issued inside transaction.
787 /** the dt device on which the transactions are executed */
788 struct dt_device *th_dev;
791 /* the size of transaction */
794 /** context for this transaction, tag is LCT_TX_HANDLE */
795 struct lu_context th_ctx;
797 /** additional tags (layers can add in declare) */
800 /** the last operation result in this transaction.
801 * this value is used in recovery */
804 /** whether we need sync commit */
805 unsigned int th_sync:1;
807 /* local transation, no need to inform other layers */
808 unsigned int th_local:1;
810 struct thandle_update *th_update;
813 static inline void thandle_get(struct thandle *thandle)
815 atomic_inc(&thandle->th_refc);
818 static inline void thandle_put(struct thandle *thandle)
820 if (atomic_dec_and_test(&thandle->th_refc)) {
821 if (thandle->th_update != NULL)
822 OBD_FREE_PTR(thandle->th_update);
823 OBD_FREE(thandle, thandle->th_alloc_size);
827 * Transaction call-backs.
829 * These are invoked by osd (or underlying transaction engine) when
830 * transaction changes state.
832 * Call-backs are used by upper layers to modify transaction parameters and to
833 * perform some actions on for each transaction state transition. Typical
834 * example is mdt registering call-back to write into last-received file
835 * before each transaction commit.
837 struct dt_txn_callback {
838 int (*dtc_txn_start)(const struct lu_env *env,
839 struct thandle *txn, void *cookie);
840 int (*dtc_txn_stop)(const struct lu_env *env,
841 struct thandle *txn, void *cookie);
842 void (*dtc_txn_commit)(struct thandle *txn, void *cookie);
845 struct list_head dtc_linkage;
848 void dt_txn_callback_add(struct dt_device *dev, struct dt_txn_callback *cb);
849 void dt_txn_callback_del(struct dt_device *dev, struct dt_txn_callback *cb);
851 int dt_txn_hook_start(const struct lu_env *env,
852 struct dt_device *dev, struct thandle *txn);
853 int dt_txn_hook_stop(const struct lu_env *env, struct thandle *txn);
854 void dt_txn_hook_commit(struct thandle *txn);
856 int dt_try_as_dir(const struct lu_env *env, struct dt_object *obj);
859 * Callback function used for parsing path.
860 * \see llo_store_resolve
862 typedef int (*dt_entry_func_t)(const struct lu_env *env,
866 #define DT_MAX_PATH 1024
868 int dt_path_parser(const struct lu_env *env,
869 char *local, dt_entry_func_t entry_func,
873 dt_store_resolve(const struct lu_env *env, struct dt_device *dt,
874 const char *path, struct lu_fid *fid);
876 struct dt_object *dt_store_open(const struct lu_env *env,
877 struct dt_device *dt,
879 const char *filename,
882 struct dt_object *dt_find_or_create(const struct lu_env *env,
883 struct dt_device *dt,
884 const struct lu_fid *fid,
885 struct dt_object_format *dof,
886 struct lu_attr *attr);
888 struct dt_object *dt_locate_at(const struct lu_env *env,
889 struct dt_device *dev,
890 const struct lu_fid *fid,
891 struct lu_device *top_dev,
892 const struct lu_object_conf *conf);
894 static inline struct dt_object *
895 dt_locate(const struct lu_env *env, struct dt_device *dev,
896 const struct lu_fid *fid)
898 return dt_locate_at(env, dev, fid,
899 dev->dd_lu_dev.ld_site->ls_top_dev, NULL);
902 int local_oid_storage_init(const struct lu_env *env, struct dt_device *dev,
903 const struct lu_fid *first_fid,
904 struct local_oid_storage **los);
905 void local_oid_storage_fini(const struct lu_env *env,
906 struct local_oid_storage *los);
907 int local_object_fid_generate(const struct lu_env *env,
908 struct local_oid_storage *los,
910 int local_object_declare_create(const struct lu_env *env,
911 struct local_oid_storage *los,
913 struct lu_attr *attr,
914 struct dt_object_format *dof,
916 int local_object_create(const struct lu_env *env,
917 struct local_oid_storage *los,
919 struct lu_attr *attr, struct dt_object_format *dof,
921 struct dt_object *local_file_find_or_create(const struct lu_env *env,
922 struct local_oid_storage *los,
923 struct dt_object *parent,
924 const char *name, __u32 mode);
925 struct dt_object *local_file_find_or_create_with_fid(const struct lu_env *env,
926 struct dt_device *dt,
927 const struct lu_fid *fid,
928 struct dt_object *parent,
932 local_index_find_or_create(const struct lu_env *env,
933 struct local_oid_storage *los,
934 struct dt_object *parent,
935 const char *name, __u32 mode,
936 const struct dt_index_features *ft);
938 local_index_find_or_create_with_fid(const struct lu_env *env,
939 struct dt_device *dt,
940 const struct lu_fid *fid,
941 struct dt_object *parent,
942 const char *name, __u32 mode,
943 const struct dt_index_features *ft);
944 int local_object_unlink(const struct lu_env *env, struct dt_device *dt,
945 struct dt_object *parent, const char *name);
947 static inline int dt_object_lock(const struct lu_env *env,
948 struct dt_object *o, struct lustre_handle *lh,
949 struct ldlm_enqueue_info *einfo,
950 union ldlm_policy_data *policy)
953 LASSERT(o->do_ops != NULL);
954 LASSERT(o->do_ops->do_object_lock != NULL);
955 return o->do_ops->do_object_lock(env, o, lh, einfo, policy);
958 static inline int dt_object_unlock(const struct lu_env *env,
960 struct ldlm_enqueue_info *einfo,
961 union ldlm_policy_data *policy)
964 LASSERT(o->do_ops != NULL);
965 LASSERT(o->do_ops->do_object_unlock != NULL);
966 return o->do_ops->do_object_unlock(env, o, einfo, policy);
969 int dt_lookup_dir(const struct lu_env *env, struct dt_object *dir,
970 const char *name, struct lu_fid *fid);
972 static inline int dt_object_sync(const struct lu_env *env, struct dt_object *o,
973 __u64 start, __u64 end)
977 LASSERT(o->do_ops->do_object_sync);
978 return o->do_ops->do_object_sync(env, o, start, end);
981 int dt_declare_version_set(const struct lu_env *env, struct dt_object *o,
983 void dt_version_set(const struct lu_env *env, struct dt_object *o,
984 dt_obj_version_t version, struct thandle *th);
985 dt_obj_version_t dt_version_get(const struct lu_env *env, struct dt_object *o);
988 int dt_read(const struct lu_env *env, struct dt_object *dt,
989 struct lu_buf *buf, loff_t *pos);
990 int dt_record_read(const struct lu_env *env, struct dt_object *dt,
991 struct lu_buf *buf, loff_t *pos);
992 int dt_record_write(const struct lu_env *env, struct dt_object *dt,
993 const struct lu_buf *buf, loff_t *pos, struct thandle *th);
994 typedef int (*dt_index_page_build_t)(const struct lu_env *env,
995 union lu_page *lp, int nob,
996 const struct dt_it_ops *iops,
997 struct dt_it *it, __u32 attr, void *arg);
998 int dt_index_walk(const struct lu_env *env, struct dt_object *obj,
999 const struct lu_rdpg *rdpg, dt_index_page_build_t filler,
1001 int dt_index_read(const struct lu_env *env, struct dt_device *dev,
1002 struct idx_info *ii, const struct lu_rdpg *rdpg);
1004 static inline struct thandle *dt_trans_create(const struct lu_env *env,
1005 struct dt_device *d)
1007 LASSERT(d->dd_ops->dt_trans_create);
1008 return d->dd_ops->dt_trans_create(env, d);
1011 static inline int dt_trans_start(const struct lu_env *env,
1012 struct dt_device *d, struct thandle *th)
1014 LASSERT(d->dd_ops->dt_trans_start);
1015 return d->dd_ops->dt_trans_start(env, d, th);
1018 /* for this transaction hooks shouldn't be called */
1019 static inline int dt_trans_start_local(const struct lu_env *env,
1020 struct dt_device *d, struct thandle *th)
1022 LASSERT(d->dd_ops->dt_trans_start);
1024 return d->dd_ops->dt_trans_start(env, d, th);
1027 static inline int dt_trans_stop(const struct lu_env *env,
1028 struct dt_device *d, struct thandle *th)
1030 LASSERT(d->dd_ops->dt_trans_stop);
1031 return d->dd_ops->dt_trans_stop(env, d, th);
1034 static inline int dt_trans_cb_add(struct thandle *th,
1035 struct dt_txn_commit_cb *dcb)
1037 LASSERT(th->th_dev->dd_ops->dt_trans_cb_add);
1038 dcb->dcb_magic = TRANS_COMMIT_CB_MAGIC;
1039 return th->th_dev->dd_ops->dt_trans_cb_add(th, dcb);
1044 static inline int dt_declare_record_write(const struct lu_env *env,
1045 struct dt_object *dt,
1046 const struct lu_buf *buf,
1052 LASSERTF(dt != NULL, "dt is NULL when we want to write record\n");
1053 LASSERT(th != NULL);
1054 LASSERT(dt->do_body_ops);
1055 LASSERT(dt->do_body_ops->dbo_declare_write);
1056 rc = dt->do_body_ops->dbo_declare_write(env, dt, buf, pos, th);
1060 static inline int dt_declare_create(const struct lu_env *env,
1061 struct dt_object *dt,
1062 struct lu_attr *attr,
1063 struct dt_allocation_hint *hint,
1064 struct dt_object_format *dof,
1068 LASSERT(dt->do_ops);
1069 LASSERT(dt->do_ops->do_declare_create);
1070 return dt->do_ops->do_declare_create(env, dt, attr, hint, dof, th);
1073 static inline int dt_create(const struct lu_env *env,
1074 struct dt_object *dt,
1075 struct lu_attr *attr,
1076 struct dt_allocation_hint *hint,
1077 struct dt_object_format *dof,
1081 LASSERT(dt->do_ops);
1082 LASSERT(dt->do_ops->do_create);
1083 return dt->do_ops->do_create(env, dt, attr, hint, dof, th);
1086 static inline int dt_declare_destroy(const struct lu_env *env,
1087 struct dt_object *dt,
1091 LASSERT(dt->do_ops);
1092 LASSERT(dt->do_ops->do_declare_destroy);
1093 return dt->do_ops->do_declare_destroy(env, dt, th);
1096 static inline int dt_destroy(const struct lu_env *env,
1097 struct dt_object *dt,
1101 LASSERT(dt->do_ops);
1102 LASSERT(dt->do_ops->do_destroy);
1103 return dt->do_ops->do_destroy(env, dt, th);
1106 static inline void dt_read_lock(const struct lu_env *env,
1107 struct dt_object *dt,
1111 LASSERT(dt->do_ops);
1112 LASSERT(dt->do_ops->do_read_lock);
1113 dt->do_ops->do_read_lock(env, dt, role);
1116 static inline void dt_write_lock(const struct lu_env *env,
1117 struct dt_object *dt,
1121 LASSERT(dt->do_ops);
1122 LASSERT(dt->do_ops->do_write_lock);
1123 dt->do_ops->do_write_lock(env, dt, role);
1126 static inline void dt_read_unlock(const struct lu_env *env,
1127 struct dt_object *dt)
1130 LASSERT(dt->do_ops);
1131 LASSERT(dt->do_ops->do_read_unlock);
1132 dt->do_ops->do_read_unlock(env, dt);
1135 static inline void dt_write_unlock(const struct lu_env *env,
1136 struct dt_object *dt)
1139 LASSERT(dt->do_ops);
1140 LASSERT(dt->do_ops->do_write_unlock);
1141 dt->do_ops->do_write_unlock(env, dt);
1144 static inline int dt_write_locked(const struct lu_env *env,
1145 struct dt_object *dt)
1148 LASSERT(dt->do_ops);
1149 LASSERT(dt->do_ops->do_write_locked);
1150 return dt->do_ops->do_write_locked(env, dt);
1153 static inline int dt_declare_attr_get(const struct lu_env *env,
1154 struct dt_object *dt,
1155 struct lustre_capa *capa)
1158 LASSERT(dt->do_ops);
1159 LASSERT(dt->do_ops->do_declare_attr_get);
1160 return dt->do_ops->do_declare_attr_get(env, dt, capa);
1163 static inline int dt_attr_get(const struct lu_env *env, struct dt_object *dt,
1164 struct lu_attr *la, void *arg)
1167 LASSERT(dt->do_ops);
1168 LASSERT(dt->do_ops->do_attr_get);
1169 return dt->do_ops->do_attr_get(env, dt, la, arg);
1172 static inline int dt_declare_attr_set(const struct lu_env *env,
1173 struct dt_object *dt,
1174 const struct lu_attr *la,
1178 LASSERT(dt->do_ops);
1179 LASSERT(dt->do_ops->do_declare_attr_set);
1180 return dt->do_ops->do_declare_attr_set(env, dt, la, th);
1183 static inline int dt_attr_set(const struct lu_env *env, struct dt_object *dt,
1184 const struct lu_attr *la, struct thandle *th,
1185 struct lustre_capa *capa)
1188 LASSERT(dt->do_ops);
1189 LASSERT(dt->do_ops->do_attr_set);
1190 return dt->do_ops->do_attr_set(env, dt, la, th, capa);
1193 static inline int dt_declare_ref_add(const struct lu_env *env,
1194 struct dt_object *dt, struct thandle *th)
1197 LASSERT(dt->do_ops);
1198 LASSERT(dt->do_ops->do_declare_ref_add);
1199 return dt->do_ops->do_declare_ref_add(env, dt, th);
1202 static inline int dt_ref_add(const struct lu_env *env,
1203 struct dt_object *dt, struct thandle *th)
1206 LASSERT(dt->do_ops);
1207 LASSERT(dt->do_ops->do_ref_add);
1208 return dt->do_ops->do_ref_add(env, dt, th);
1211 static inline int dt_declare_ref_del(const struct lu_env *env,
1212 struct dt_object *dt, struct thandle *th)
1215 LASSERT(dt->do_ops);
1216 LASSERT(dt->do_ops->do_declare_ref_del);
1217 return dt->do_ops->do_declare_ref_del(env, dt, th);
1220 static inline int dt_ref_del(const struct lu_env *env,
1221 struct dt_object *dt, struct thandle *th)
1224 LASSERT(dt->do_ops);
1225 LASSERT(dt->do_ops->do_ref_del);
1226 return dt->do_ops->do_ref_del(env, dt, th);
1229 static inline struct obd_capa *dt_capa_get(const struct lu_env *env,
1230 struct dt_object *dt,
1231 struct lustre_capa *old, __u64 opc)
1234 LASSERT(dt->do_ops);
1235 LASSERT(dt->do_ops->do_capa_get);
1236 return dt->do_ops->do_capa_get(env, dt, old, opc);
1239 static inline int dt_bufs_get(const struct lu_env *env, struct dt_object *d,
1240 struct niobuf_remote *rnb,
1241 struct niobuf_local *lnb, int rw,
1242 struct lustre_capa *capa)
1245 LASSERT(d->do_body_ops);
1246 LASSERT(d->do_body_ops->dbo_bufs_get);
1247 return d->do_body_ops->dbo_bufs_get(env, d, rnb->rnb_offset,
1248 rnb->rnb_len, lnb, rw, capa);
1251 static inline int dt_bufs_put(const struct lu_env *env, struct dt_object *d,
1252 struct niobuf_local *lnb, int n)
1255 LASSERT(d->do_body_ops);
1256 LASSERT(d->do_body_ops->dbo_bufs_put);
1257 return d->do_body_ops->dbo_bufs_put(env, d, lnb, n);
1260 static inline int dt_write_prep(const struct lu_env *env, struct dt_object *d,
1261 struct niobuf_local *lnb, int n)
1264 LASSERT(d->do_body_ops);
1265 LASSERT(d->do_body_ops->dbo_write_prep);
1266 return d->do_body_ops->dbo_write_prep(env, d, lnb, n);
1269 static inline int dt_declare_write_commit(const struct lu_env *env,
1270 struct dt_object *d,
1271 struct niobuf_local *lnb,
1272 int n, struct thandle *th)
1274 LASSERTF(d != NULL, "dt is NULL when we want to declare write\n");
1275 LASSERT(th != NULL);
1276 return d->do_body_ops->dbo_declare_write_commit(env, d, lnb, n, th);
1280 static inline int dt_write_commit(const struct lu_env *env,
1281 struct dt_object *d, struct niobuf_local *lnb,
1282 int n, struct thandle *th)
1285 LASSERT(d->do_body_ops);
1286 LASSERT(d->do_body_ops->dbo_write_commit);
1287 return d->do_body_ops->dbo_write_commit(env, d, lnb, n, th);
1290 static inline int dt_read_prep(const struct lu_env *env, struct dt_object *d,
1291 struct niobuf_local *lnb, int n)
1294 LASSERT(d->do_body_ops);
1295 LASSERT(d->do_body_ops->dbo_read_prep);
1296 return d->do_body_ops->dbo_read_prep(env, d, lnb, n);
1299 static inline int dt_declare_punch(const struct lu_env *env,
1300 struct dt_object *dt, __u64 start,
1301 __u64 end, struct thandle *th)
1304 LASSERT(dt->do_body_ops);
1305 LASSERT(dt->do_body_ops->dbo_declare_punch);
1306 return dt->do_body_ops->dbo_declare_punch(env, dt, start, end, th);
1309 static inline int dt_punch(const struct lu_env *env, struct dt_object *dt,
1310 __u64 start, __u64 end, struct thandle *th,
1311 struct lustre_capa *capa)
1314 LASSERT(dt->do_body_ops);
1315 LASSERT(dt->do_body_ops->dbo_punch);
1316 return dt->do_body_ops->dbo_punch(env, dt, start, end, th, capa);
1319 static inline int dt_fiemap_get(const struct lu_env *env, struct dt_object *d,
1320 struct ll_user_fiemap *fm)
1323 if (d->do_body_ops == NULL)
1325 if (d->do_body_ops->dbo_fiemap_get == NULL)
1327 return d->do_body_ops->dbo_fiemap_get(env, d, fm);
1330 static inline int dt_statfs(const struct lu_env *env, struct dt_device *dev,
1331 struct obd_statfs *osfs)
1334 LASSERT(dev->dd_ops);
1335 LASSERT(dev->dd_ops->dt_statfs);
1336 return dev->dd_ops->dt_statfs(env, dev, osfs);
1339 static inline int dt_root_get(const struct lu_env *env, struct dt_device *dev,
1343 LASSERT(dev->dd_ops);
1344 LASSERT(dev->dd_ops->dt_root_get);
1345 return dev->dd_ops->dt_root_get(env, dev, f);
1348 static inline void dt_conf_get(const struct lu_env *env,
1349 const struct dt_device *dev,
1350 struct dt_device_param *param)
1353 LASSERT(dev->dd_ops);
1354 LASSERT(dev->dd_ops->dt_conf_get);
1355 return dev->dd_ops->dt_conf_get(env, dev, param);
1358 static inline int dt_sync(const struct lu_env *env, struct dt_device *dev)
1361 LASSERT(dev->dd_ops);
1362 LASSERT(dev->dd_ops->dt_sync);
1363 return dev->dd_ops->dt_sync(env, dev);
1366 static inline int dt_ro(const struct lu_env *env, struct dt_device *dev)
1369 LASSERT(dev->dd_ops);
1370 LASSERT(dev->dd_ops->dt_ro);
1371 return dev->dd_ops->dt_ro(env, dev);
1374 static inline int dt_declare_insert(const struct lu_env *env,
1375 struct dt_object *dt,
1376 const struct dt_rec *rec,
1377 const struct dt_key *key,
1381 LASSERT(dt->do_index_ops);
1382 LASSERT(dt->do_index_ops->dio_declare_insert);
1383 return dt->do_index_ops->dio_declare_insert(env, dt, rec, key, th);
1386 static inline int dt_insert(const struct lu_env *env,
1387 struct dt_object *dt,
1388 const struct dt_rec *rec,
1389 const struct dt_key *key,
1391 struct lustre_capa *capa,
1395 LASSERT(dt->do_index_ops);
1396 LASSERT(dt->do_index_ops->dio_insert);
1397 return dt->do_index_ops->dio_insert(env, dt, rec, key, th,
1401 static inline int dt_declare_xattr_del(const struct lu_env *env,
1402 struct dt_object *dt,
1407 LASSERT(dt->do_ops);
1408 LASSERT(dt->do_ops->do_declare_xattr_del);
1409 return dt->do_ops->do_declare_xattr_del(env, dt, name, th);
1412 static inline int dt_xattr_del(const struct lu_env *env,
1413 struct dt_object *dt, const char *name,
1415 struct lustre_capa *capa)
1418 LASSERT(dt->do_ops);
1419 LASSERT(dt->do_ops->do_xattr_del);
1420 return dt->do_ops->do_xattr_del(env, dt, name, th, capa);
1423 static inline int dt_declare_xattr_set(const struct lu_env *env,
1424 struct dt_object *dt,
1425 const struct lu_buf *buf,
1426 const char *name, int fl,
1430 LASSERT(dt->do_ops);
1431 LASSERT(dt->do_ops->do_declare_xattr_set);
1432 return dt->do_ops->do_declare_xattr_set(env, dt, buf, name, fl, th);
1435 static inline int dt_xattr_set(const struct lu_env *env,
1436 struct dt_object *dt, const struct lu_buf *buf,
1437 const char *name, int fl, struct thandle *th,
1438 struct lustre_capa *capa)
1441 LASSERT(dt->do_ops);
1442 LASSERT(dt->do_ops->do_xattr_set);
1443 return dt->do_ops->do_xattr_set(env, dt, buf, name, fl, th, capa);
1446 static inline int dt_declare_xattr_get(const struct lu_env *env,
1447 struct dt_object *dt,
1450 struct lustre_capa *capa)
1453 LASSERT(dt->do_ops);
1454 LASSERT(dt->do_ops->do_declare_xattr_get);
1455 return dt->do_ops->do_declare_xattr_get(env, dt, buf, name, capa);
1458 static inline int dt_xattr_get(const struct lu_env *env,
1459 struct dt_object *dt, struct lu_buf *buf,
1460 const char *name, struct lustre_capa *capa)
1463 LASSERT(dt->do_ops);
1464 LASSERT(dt->do_ops->do_xattr_get);
1465 return dt->do_ops->do_xattr_get(env, dt, buf, name, capa);
1468 static inline int dt_xattr_list(const struct lu_env *env,
1469 struct dt_object *dt, struct lu_buf *buf,
1470 struct lustre_capa *capa)
1473 LASSERT(dt->do_ops);
1474 LASSERT(dt->do_ops->do_xattr_list);
1475 return dt->do_ops->do_xattr_list(env, dt, buf, capa);
1478 static inline int dt_declare_delete(const struct lu_env *env,
1479 struct dt_object *dt,
1480 const struct dt_key *key,
1484 LASSERT(dt->do_index_ops);
1485 LASSERT(dt->do_index_ops->dio_declare_delete);
1486 return dt->do_index_ops->dio_declare_delete(env, dt, key, th);
1489 static inline int dt_delete(const struct lu_env *env,
1490 struct dt_object *dt,
1491 const struct dt_key *key,
1493 struct lustre_capa *capa)
1496 LASSERT(dt->do_index_ops);
1497 LASSERT(dt->do_index_ops->dio_delete);
1498 return dt->do_index_ops->dio_delete(env, dt, key, th, capa);
1501 static inline int dt_commit_async(const struct lu_env *env,
1502 struct dt_device *dev)
1505 LASSERT(dev->dd_ops);
1506 LASSERT(dev->dd_ops->dt_commit_async);
1507 return dev->dd_ops->dt_commit_async(env, dev);
1510 static inline int dt_init_capa_ctxt(const struct lu_env *env,
1511 struct dt_device *dev,
1512 int mode, unsigned long timeout,
1513 __u32 alg, struct lustre_capa_key *keys)
1516 LASSERT(dev->dd_ops);
1517 LASSERT(dev->dd_ops->dt_init_capa_ctxt);
1518 return dev->dd_ops->dt_init_capa_ctxt(env, dev, mode,
1519 timeout, alg, keys);
1522 static inline int dt_lookup(const struct lu_env *env,
1523 struct dt_object *dt,
1525 const struct dt_key *key,
1526 struct lustre_capa *capa)
1531 LASSERT(dt->do_index_ops);
1532 LASSERT(dt->do_index_ops->dio_lookup);
1534 ret = dt->do_index_ops->dio_lookup(env, dt, rec, key, capa);
1542 #define LU221_BAD_TIME (0x80000000U + 24 * 3600)
1544 struct dt_find_hint {
1545 struct lu_fid *dfh_fid;
1546 struct dt_device *dfh_dt;
1547 struct dt_object *dfh_o;
1550 struct dt_insert_rec {
1552 const struct lu_fid *rec_fid;
1564 struct dt_thread_info {
1565 char dti_buf[DT_MAX_PATH];
1566 struct dt_find_hint dti_dfh;
1567 struct lu_attr dti_attr;
1568 struct lu_fid dti_fid;
1569 struct dt_object_format dti_dof;
1570 struct lustre_mdt_attrs dti_lma;
1571 struct lu_buf dti_lb;
1572 struct lu_object_conf dti_conf;
1574 struct dt_insert_rec dti_dt_rec;
1577 extern struct lu_context_key dt_key;
1579 static inline struct dt_thread_info *dt_info(const struct lu_env *env)
1581 struct dt_thread_info *dti;
1583 dti = lu_context_key_get(&env->le_ctx, &dt_key);
1588 int dt_global_init(void);
1589 void dt_global_fini(void);
1592 #ifndef HAVE_ONLY_PROCFS_SEQ
1593 int lprocfs_dt_rd_blksize(char *page, char **start, off_t off,
1594 int count, int *eof, void *data);
1595 int lprocfs_dt_rd_kbytestotal(char *page, char **start, off_t off,
1596 int count, int *eof, void *data);
1597 int lprocfs_dt_rd_kbytesfree(char *page, char **start, off_t off,
1598 int count, int *eof, void *data);
1599 int lprocfs_dt_rd_kbytesavail(char *page, char **start, off_t off,
1600 int count, int *eof, void *data);
1601 int lprocfs_dt_rd_filestotal(char *page, char **start, off_t off,
1602 int count, int *eof, void *data);
1603 int lprocfs_dt_rd_filesfree(char *page, char **start, off_t off,
1604 int count, int *eof, void *data);
1606 int lprocfs_dt_blksize_seq_show(struct seq_file *m, void *v);
1607 int lprocfs_dt_kbytestotal_seq_show(struct seq_file *m, void *v);
1608 int lprocfs_dt_kbytesfree_seq_show(struct seq_file *m, void *v);
1609 int lprocfs_dt_kbytesavail_seq_show(struct seq_file *m, void *v);
1610 int lprocfs_dt_filestotal_seq_show(struct seq_file *m, void *v);
1611 int lprocfs_dt_filesfree_seq_show(struct seq_file *m, void *v);
1612 # endif /* LPROCFS */
1614 #endif /* __LUSTRE_DT_OBJECT_H */