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, 2012, Whamcloud, Inc.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/osd/osd_internal.h
38 * Shared definitions and declarations for osd module
40 * Author: Nikita Danilov <nikita@clusterfs.com>
43 #ifndef _OSD_INTERNAL_H
44 #define _OSD_INTERNAL_H
46 #if defined(__KERNEL__)
48 /* struct rw_semaphore */
49 #include <linux/rwsem.h>
51 #include <linux/dcache.h>
53 #include <linux/dirent.h>
55 #include <ldiskfs/ldiskfs.h>
56 #include <ldiskfs/ldiskfs_jbd2.h>
57 #ifdef HAVE_LDISKFS_JOURNAL_CALLBACK_ADD
58 # define journal_callback ldiskfs_journal_cb_entry
59 # define osd_journal_callback_set(handle, func, jcb) \
60 ldiskfs_journal_callback_add(handle, func, jcb)
62 # define osd_journal_callback_set(handle, func, jcb) \
63 jbd2_journal_callback_set(handle, func, jcb)
66 /* fsfilt_{get|put}_ops */
67 #include <lustre_fsfilt.h>
71 /* class_register_type(), class_unregister_type(), class_get_type() */
72 #include <obd_class.h>
73 #include <lustre_disk.h>
74 #include <dt_object.h>
75 #include <lustre_quota.h>
79 #include "osd_scrub.h"
80 #include "osd_quota_fmt.h"
84 #define OSD_COUNTERS (0)
86 /* Lustre special inode::i_state to indicate OI scrub skip this inode. */
87 #define I_LUSTRE_NOSCRUB (1 << 31)
89 /** Enable thandle usage statistics */
90 #define OSD_THANDLE_STATS (0)
92 struct osd_directory {
93 struct iam_container od_container;
94 struct iam_descr od_descr;
98 * Object Index (oi) instance.
102 * underlying index object, where fid->id mapping in stored.
104 struct inode *oi_inode;
105 struct osd_directory oi_dir;
108 extern const int osd_dto_credits_noquota[];
111 struct dt_object oo_dt;
113 * Inode for file system object represented by this osd_object. This
114 * inode is pinned for the whole duration of lu_object life.
116 * Not modified concurrently (either setup early during object
117 * creation, or assigned by osd_object_create() under write lock).
119 struct inode *oo_inode;
121 * to protect index ops.
123 struct htree_lock_head *oo_hl_head;
124 struct rw_semaphore oo_ext_idx_sem;
125 struct rw_semaphore oo_sem;
126 struct osd_directory *oo_dir;
127 /** protects inode attributes. */
130 * Following two members are used to indicate the presence of dot and
131 * dotdot in the given directory. This is required for interop mode
134 int oo_compat_dot_created;
135 int oo_compat_dotdot_created;
137 const struct lu_env *oo_owner;
138 #ifdef CONFIG_LOCKDEP
139 struct lockdep_map oo_dep_map;
143 #ifdef HAVE_LDISKFS_PDO
145 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
146 ll_ldiskfs_find_entry(dir, dentry, de, lock)
147 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
148 ldiskfs_add_entry(handle, child, cinode, hlock)
150 #else /* HAVE_LDISKFS_PDO */
156 struct htree_lock_head {
160 #define ldiskfs_htree_lock(lock, head, inode, op) do { LBUG(); } while (0)
161 #define ldiskfs_htree_unlock(lock) do { LBUG(); } while (0)
163 static inline struct htree_lock_head *ldiskfs_htree_lock_head_alloc(int dep)
169 #define ldiskfs_htree_lock_head_free(lh) do { LBUG(); } while (0)
171 #define LDISKFS_DUMMY_HTREE_LOCK 0xbabecafe
173 static inline struct htree_lock *ldiskfs_htree_lock_alloc(void)
175 return (struct htree_lock *)LDISKFS_DUMMY_HTREE_LOCK;
178 static inline void ldiskfs_htree_lock_free(struct htree_lock *lk)
180 LASSERT((unsigned long)lk == LDISKFS_DUMMY_HTREE_LOCK);
183 #define HTREE_HBITS_DEF 0
185 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
186 ll_ldiskfs_find_entry(dir, dentry, de)
187 #define osd_ldiskfs_add_entry(handle, child, cinode, lock) \
188 ldiskfs_add_entry(handle, child, cinode)
190 #endif /* HAVE_LDISKFS_PDO */
192 #define OSD_OTABLE_IT_CACHE_SIZE 128
193 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
195 struct osd_inconsistent_item {
196 /* link into osd_scrub::os_inconsistent_items,
197 * protected by osd_scrub::os_lock. */
200 /* The right FID <=> ino#/gen mapping. */
201 struct osd_idmap_cache oii_cache;
203 unsigned int oii_insert:1; /* insert or update mapping. */
206 struct osd_otable_cache {
207 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
209 /* Index for next cache slot to be filled. */
210 int ooc_producer_idx;
212 /* Index for next cache slot to be returned by it::next(). */
213 int ooc_consumer_idx;
215 /* How many items in ooc_cache. */
216 int ooc_cached_items;
218 /* Position for up layer LFSCK iteration pre-loading. */
219 __u32 ooc_pos_preload;
222 struct osd_otable_it {
223 struct osd_device *ooi_dev;
224 struct osd_otable_cache ooi_cache;
226 /* For osd_otable_it_key. */
229 /* The following bits can be updated/checked w/o lock protection.
230 * If more bits will be introduced in the future and need lock to
231 * protect, please add comment. */
232 unsigned long ooi_used_outside:1, /* Some user out of OSD
233 * uses the iteration. */
234 ooi_all_cached:1, /* No more entries can be
235 * filled into cache. */
236 ooi_user_ready:1, /* The user out of OSD is
237 * ready to iterate. */
238 ooi_waiting:1; /* it::next is waiting. */
241 extern const int osd_dto_credits_noquota[];
248 struct dt_device od_dt_dev;
249 /* information about underlying file system */
250 struct vfsmount *od_mnt;
252 struct osd_oi **od_oi_table;
253 /* total number of OI containers */
258 unsigned int od_fl_capa:1,
259 od_is_md:1; /* set in ->ldo_prepare */
260 unsigned long od_capa_timeout;
262 struct lustre_capa_key *od_capa_keys;
263 cfs_hlist_head_t *od_capa_hash;
265 cfs_proc_dir_entry_t *od_proc_entry;
266 struct lprocfs_stats *od_stats;
268 * statfs optimization: we cache a bit.
270 cfs_time_t od_osfs_age;
271 struct obd_statfs od_statfs;
272 spinlock_t od_osfs_lock;
274 unsigned int od_noscrub:1;
276 struct fsfilt_operations *od_fsops;
278 struct lu_site od_site;
281 * mapping for legacy OST objids
283 struct osd_compat_objid *od_ost_map;
285 unsigned long long od_readcache_max_filesize;
287 int od_writethrough_cache;
289 struct brw_stats od_brw_stats;
290 cfs_atomic_t od_r_in_flight;
291 cfs_atomic_t od_w_in_flight;
293 struct mutex od_otable_mutex;
294 struct osd_otable_it *od_otable_it;
295 struct osd_scrub od_scrub;
297 /* service name associated with the osd device */
298 char od_svname[MAX_OBD_NAME];
299 char od_mntdev[MAX_OBD_NAME];
301 /* quota slave instance */
302 struct qsd_instance *od_quota_slave;
305 #define OSD_TRACK_DECLARES
306 #ifdef OSD_TRACK_DECLARES
307 #define OSD_DECLARE_OP(oh, op, credits) \
309 LASSERT((oh)->ot_handle == NULL); \
310 ((oh)->ot_declare_ ##op)++; \
311 ((oh)->ot_declare_ ##op ##_cred) += (credits); \
312 (oh)->ot_credits += (credits); \
314 #define OSD_EXEC_OP(handle, op) \
316 struct osd_thandle *oh = container_of(handle, typeof(*oh), ot_super); \
317 LASSERT((oh)->ot_declare_ ##op > 0); \
318 ((oh)->ot_declare_ ##op)--; \
321 #define OSD_DECLARE_OP(oh, op, credits) (oh)->ot_credits += (credits)
322 #define OSD_EXEC_OP(oh, op)
325 /* There are at most 10 uid/gids are affected in a transaction, and
326 * that's rename case:
327 * - 2 for source parent uid & gid;
328 * - 2 for source child uid & gid ('..' entry update when child is directory);
329 * - 2 for target parent uid & gid;
330 * - 2 for target child uid & gid (if the target child exists);
331 * - 2 for root uid & gid (last_rcvd, llog, etc);
333 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
334 * the id type of each id in the ot_id_array.
336 #define OSD_MAX_UGID_CNT 10
339 struct thandle ot_super;
341 struct journal_callback ot_jcb;
342 cfs_list_t ot_dcb_list;
343 /* Link to the device, for debugging. */
344 struct lu_ref_link *ot_dev_link;
345 unsigned short ot_credits;
346 unsigned short ot_id_cnt;
347 unsigned short ot_id_type;
348 uid_t ot_id_array[OSD_MAX_UGID_CNT];
349 struct lquota_trans *ot_quota_trans;
351 #ifdef OSD_TRACK_DECLARES
352 /* Tracking for transaction credits, to allow debugging and optimizing
353 * cases where a large number of credits are being allocated for
354 * single transaction. */
355 unsigned char ot_declare_attr_set;
356 unsigned char ot_declare_punch;
357 unsigned char ot_declare_xattr_set;
358 unsigned char ot_declare_create;
359 unsigned char ot_declare_destroy;
360 unsigned char ot_declare_ref_add;
361 unsigned char ot_declare_ref_del;
362 unsigned char ot_declare_write;
363 unsigned char ot_declare_insert;
364 unsigned char ot_declare_delete;
365 unsigned char ot_declare_quota;
367 unsigned short ot_declare_attr_set_cred;
368 unsigned short ot_declare_punch_cred;
369 unsigned short ot_declare_xattr_set_cred;
370 unsigned short ot_declare_create_cred;
371 unsigned short ot_declare_destroy_cred;
372 unsigned short ot_declare_ref_add_cred;
373 unsigned short ot_declare_ref_del_cred;
374 unsigned short ot_declare_write_cred;
375 unsigned short ot_declare_insert_cred;
376 unsigned short ot_declare_delete_cred;
377 unsigned short ot_declare_quota_cred;
380 #if OSD_THANDLE_STATS
381 /** time when this handle was allocated */
382 cfs_time_t oth_alloced;
384 /** time when this thanle was started */
385 cfs_time_t oth_started;
390 * Basic transaction credit op
400 DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
414 LPROC_OSD_READ_BYTES = 0,
415 LPROC_OSD_WRITE_BYTES = 1,
416 LPROC_OSD_GET_PAGE = 2,
417 LPROC_OSD_NO_PAGE = 3,
418 LPROC_OSD_CACHE_ACCESS = 4,
419 LPROC_OSD_CACHE_HIT = 5,
420 LPROC_OSD_CACHE_MISS = 6,
422 #if OSD_THANDLE_STATS
423 LPROC_OSD_THANDLE_STARTING,
424 LPROC_OSD_THANDLE_OPEN,
425 LPROC_OSD_THANDLE_CLOSING,
432 * Storage representation for fids.
434 * Variable size, first byte contains the length of the whole record.
436 struct osd_fid_pack {
437 unsigned char fp_len;
438 char fp_area[sizeof(struct lu_fid)];
441 struct osd_it_ea_dirent {
442 struct lu_fid oied_fid;
445 unsigned short oied_namelen;
446 unsigned int oied_type;
448 } __attribute__((packed));
451 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
452 * than lu_dirent struct. osd readdir reads less number of dirent than
453 * required for mdd dir page. so buffer size need to be increased so that
454 * there would be one ext3 readdir for every mdd readdir page.
457 #define OSD_IT_EA_BUFSIZE (CFS_PAGE_SIZE + CFS_PAGE_SIZE/4)
460 * This is iterator's in-memory data structure in interoperability
461 * mode (i.e. iterator over ldiskfs style directory)
464 struct osd_object *oie_obj;
465 /** used in ldiskfs iterator, to stored file pointer */
466 struct file oie_file;
467 /** how many entries have been read-cached from storage */
469 /** current entry is being iterated by caller */
471 /** current processing entry */
472 struct osd_it_ea_dirent *oie_dirent;
473 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
478 * Iterator's in-memory data structure for IAM mode.
481 struct osd_object *oi_obj;
482 struct iam_path_descr *oi_ipd;
483 struct iam_iterator oi_it;
486 struct osd_quota_leaf {
492 * Iterator's in-memory data structure for quota file.
494 struct osd_it_quota {
495 struct osd_object *oiq_obj;
496 /** tree blocks path to where the entry is stored */
497 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
498 /** on-disk offset for current key where quota record can be found */
500 /** identifier for current quota record */
502 /** the record index in the leaf/index block */
503 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
504 /** list of already processed leaf blocks */
508 #define MAX_BLOCKS_PER_PAGE (CFS_PAGE_SIZE / 512)
512 cfs_atomic_t dr_numreqs; /* number of reqs being processed */
517 unsigned int dr_ignore_quota:1;
518 unsigned int dr_elapsed_valid:1; /* we really did count time */
519 unsigned int dr_rw:1;
520 struct page *dr_pages[PTLRPC_MAX_BRW_PAGES];
521 unsigned long dr_blocks[PTLRPC_MAX_BRW_PAGES*MAX_BLOCKS_PER_PAGE];
522 unsigned long dr_start_time;
523 unsigned long dr_elapsed; /* how long io took */
524 struct osd_device *dr_dev;
525 unsigned int dr_init_at; /* the line iobuf was initialized */
528 struct osd_thread_info {
529 const struct lu_env *oti_env;
531 * used for index operations.
533 struct dentry oti_obj_dentry;
534 struct dentry oti_child_dentry;
536 /** dentry for Iterator context. */
537 struct dentry oti_it_dentry;
538 struct htree_lock *oti_hlock;
540 struct lu_fid oti_fid;
541 struct lu_fid oti_fid2;
542 struct osd_inode_id oti_id;
543 struct osd_inode_id oti_id2;
544 struct ost_id oti_ostid;
547 * XXX temporary: for ->i_op calls.
549 struct timespec oti_time;
551 * XXX temporary: fake struct file for osd_object_sync
553 struct file oti_file;
555 * XXX temporary: for capa operations.
557 struct lustre_capa_key oti_capa_key;
558 struct lustre_capa oti_capa;
560 /** osd_device reference, initialized in osd_trans_start() and
561 used in osd_trans_stop() */
562 struct osd_device *oti_dev;
565 * following ipd and it structures are used for osd_index_iam_lookup()
566 * these are defined separately as we might do index operation
567 * in open iterator session.
570 /** osd iterator context used for iterator session */
573 struct osd_it_iam oti_it;
574 /* ldiskfs iterator data structure,
575 * see osd_it_ea_{init, fini} */
576 struct osd_it_ea oti_it_ea;
577 struct osd_it_quota oti_it_quota;
580 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
583 cfs_kstatfs_t oti_ksfs;
585 /** IAM iterator for index operation. */
586 struct iam_iterator oti_idx_it;
588 /** union to guarantee that ->oti_ipd[] has proper alignment. */
590 char oti_it_ipd[DX_IPD_MAX_SIZE];
591 long long oti_alignment_lieutenant;
595 char oti_idx_ipd[DX_IPD_MAX_SIZE];
596 long long oti_alignment_lieutenant_colonel;
599 struct osd_idmap_cache oti_cache;
604 /** used in osd_fid_set() to put xattr */
605 struct lu_buf oti_buf;
606 /** used in osd_ea_fid_set() to set fid into common ea */
608 struct lustre_mdt_attrs oti_mdt_attrs;
609 /* old LMA for compatibility */
610 char oti_mdt_attrs_old[LMA_OLD_SIZE];
613 struct osd_iobuf oti_iobuf;
614 struct inode oti_inode;
615 int oti_created[PTLRPC_MAX_BRW_PAGES];
616 struct lu_env oti_obj_delete_tx_env;
617 #define OSD_FID_REC_SZ 32
618 char oti_ldp[OSD_FID_REC_SZ];
619 char oti_ldp2[OSD_FID_REC_SZ];
621 /* used by quota code */
623 struct if_dqblk oti_dqblk;
624 struct if_dqinfo oti_dqinfo;
626 struct lquota_id_info oti_qi;
627 struct lquota_trans oti_quota_trans;
628 union lquota_rec oti_quota_rec;
632 extern int ldiskfs_pdo;
636 void lprocfs_osd_init_vars(struct lprocfs_static_vars *lvars);
637 int osd_procfs_init(struct osd_device *osd, const char *name);
638 int osd_procfs_fini(struct osd_device *osd);
639 void osd_lprocfs_time_start(const struct lu_env *env);
640 void osd_lprocfs_time_end(const struct lu_env *env,
641 struct osd_device *osd, int op);
642 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
645 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
646 struct obd_statfs *sfs);
647 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
648 struct lustre_capa *capa, __u64 opc);
649 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
650 struct osd_inode_id *id);
651 struct inode *osd_iget_fid(struct osd_thread_info *info, struct osd_device *dev,
652 struct osd_inode_id *id, struct lu_fid *fid);
654 int osd_compat_init(struct osd_device *dev);
655 void osd_compat_fini(struct osd_device *dev);
656 int osd_compat_objid_lookup(struct osd_thread_info *info,
657 struct osd_device *osd,
658 const struct lu_fid *fid, struct osd_inode_id *id);
659 int osd_compat_objid_insert(struct osd_thread_info *info,
660 struct osd_device *osd,
661 const struct lu_fid *fid,
662 const struct osd_inode_id *id, struct thandle *th);
663 int osd_compat_objid_delete(struct osd_thread_info *info,
664 struct osd_device *osd,
665 const struct lu_fid *fid, struct thandle *th);
666 int osd_compat_spec_lookup(struct osd_thread_info *info,
667 struct osd_device *osd,
668 const struct lu_fid *fid, struct osd_inode_id *id);
669 int osd_compat_spec_insert(struct osd_thread_info *info,
670 struct osd_device *osd,
671 const struct lu_fid *fid,
672 const struct osd_inode_id *id, struct thandle *th);
674 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
675 int osd_scrub_file_store(struct osd_scrub *scrub);
676 int osd_scrub_start(struct osd_device *dev);
677 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
678 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
679 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
681 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
682 struct osd_inode_id *id);
683 int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
685 /* osd_quota_fmt.c */
686 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
687 int type, uint blk, int depth, uint index,
688 struct osd_it_quota *it);
689 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
690 int type, uint blk, uint index,
691 struct osd_it_quota *it);
692 loff_t find_tree_dqentry(const struct lu_env *env,
693 struct osd_object *obj, int type,
694 qid_t dqid, uint blk, int depth,
695 struct osd_it_quota *it);
697 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
698 struct lquota_id_info *qi, bool allocated, int *flags);
699 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
700 long long space, struct osd_thandle *oh,
701 bool is_blk, bool allocated, int *flags, bool force);
702 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
703 const struct dt_rec *rec,
704 union lquota_rec *quota_rec);
705 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
706 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
707 const struct dt_index_features *feat);
709 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
711 return (feat == &dt_quota_iusr_features ||
712 feat == &dt_quota_busr_features ||
713 feat == &dt_quota_igrp_features ||
714 feat == &dt_quota_bgrp_features) ? true : false;
718 * Invariants, assertions.
722 * XXX: do not enable this, until invariant checking code is made thread safe
723 * in the face of pdirops locking.
725 #define OSD_INVARIANT_CHECKS (0)
727 #if OSD_INVARIANT_CHECKS
728 static inline int osd_invariant(const struct osd_object *obj)
732 ergo(obj->oo_inode != NULL,
733 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
734 atomic_read(&obj->oo_inode->i_count) > 0) &&
735 ergo(obj->oo_dir != NULL &&
736 obj->oo_dir->od_conationer.ic_object != NULL,
737 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
740 #define osd_invariant(obj) (1)
743 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
745 extern const struct dt_index_operations osd_otable_ops;
747 static inline int osd_oi_fid2idx(struct osd_device *dev,
748 const struct lu_fid *fid)
750 return fid->f_seq & (dev->od_oi_count - 1);
753 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
754 const struct lu_fid *fid)
756 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
757 LASSERTF(!fid_is_igif(fid), DFID"\n", PFID(fid));
758 LASSERT(osd->od_oi_table != NULL && osd->od_oi_count >= 1);
759 /* It can work even od_oi_count equals to 1 although it's unexpected,
760 * the only reason we set it to 1 is for performance measurement */
761 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
764 extern const struct lu_device_operations osd_lu_ops;
766 static inline int lu_device_is_osd(const struct lu_device *d)
768 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
771 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
773 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
774 return container_of0(d, struct osd_device, od_dt_dev);
777 static inline struct osd_device *osd_dev(const struct lu_device *d)
779 LASSERT(lu_device_is_osd(d));
780 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
783 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
785 return osd_dev(o->oo_dt.do_lu.lo_dev);
788 static inline struct super_block *osd_sb(const struct osd_device *dev)
790 return dev->od_mnt->mnt_sb;
793 static inline int osd_object_is_root(const struct osd_object *obj)
795 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
798 static inline struct osd_object *osd_obj(const struct lu_object *o)
800 LASSERT(lu_device_is_osd(o->lo_dev));
801 return container_of0(o, struct osd_object, oo_dt.do_lu);
804 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
806 return osd_obj(&d->do_lu);
809 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
811 return &osd->od_dt_dev.dd_lu_dev;
814 static inline journal_t *osd_journal(const struct osd_device *dev)
816 return LDISKFS_SB(osd_sb(dev))->s_journal;
819 extern const struct dt_body_operations osd_body_ops;
820 extern struct lu_context_key osd_key;
822 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
824 return lu_context_key_get(&env->le_ctx, &osd_key);
827 extern const struct dt_body_operations osd_body_ops_new;
833 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
834 const struct iam_container *bag)
836 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
837 osd_oti_get(env)->oti_it_ipd);
841 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
842 const struct iam_container *bag)
844 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
845 osd_oti_get(env)->oti_idx_ipd);
848 static inline void osd_ipd_put(const struct lu_env *env,
849 const struct iam_container *bag,
850 struct iam_path_descr *ipd)
852 bag->ic_descr->id_ops->id_ipd_free(ipd);
855 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
856 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
857 int write_NUL, loff_t *offs, handle_t *handle);
860 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
862 const char *name, const int namelen)
864 struct osd_thread_info *info = osd_oti_get(env);
865 struct dentry *child_dentry = &info->oti_child_dentry;
866 struct dentry *obj_dentry = &info->oti_obj_dentry;
868 obj_dentry->d_inode = inode;
869 obj_dentry->d_sb = inode->i_sb;
870 obj_dentry->d_name.hash = 0;
872 child_dentry->d_name.hash = 0;
873 child_dentry->d_parent = obj_dentry;
874 child_dentry->d_name.name = name;
875 child_dentry->d_name.len = namelen;
880 * Helper function to pack the fid, ldiskfs stores fid in packed format.
883 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
884 struct lu_fid *befider)
886 fid_cpu_to_be(befider, (struct lu_fid *)fid);
887 memcpy(pack->fp_area, befider, sizeof(*befider));
888 pack->fp_len = sizeof(*befider) + 1;
892 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
897 switch (pack->fp_len) {
898 case sizeof *fid + 1:
899 memcpy(fid, pack->fp_area, sizeof *fid);
900 fid_be_to_cpu(fid, fid);
903 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
910 * Quota/Accounting handling
912 extern const struct dt_index_operations osd_acct_index_ops;
913 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
914 const struct lu_fid *fid, struct osd_inode_id *id);
916 /* copy from fs/ext4/dir.c */
917 static inline int is_32bit_api(void)
920 return is_compat_task();
922 return (BITS_PER_LONG == 32);
926 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
928 if ((filp->f_mode & FMODE_32BITHASH) ||
929 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
930 return LDISKFS_HTREE_EOF_32BIT;
932 return LDISKFS_HTREE_EOF_64BIT;
935 #endif /* __KERNEL__ */
936 #endif /* _OSD_INTERNAL_H */