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, Intel Corporation.
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;
144 /* protects on-fly initialization */
145 int oos_subdir_count; /* subdir count for each seq */
146 struct dentry *oos_root; /* O/<seq> */
147 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
148 obd_seq oos_seq; /* seq number */
149 cfs_list_t oos_seq_list; /* list to seq_list */
153 struct dentry *om_root; /* dentry for /O */
154 rwlock_t om_seq_list_lock; /* lock for seq_list */
155 cfs_list_t om_seq_list; /* list head for seq */
157 struct semaphore om_dir_init_sem;
160 #ifdef HAVE_LDISKFS_PDO
162 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
163 ll_ldiskfs_find_entry(dir, dentry, de, lock)
164 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
165 ldiskfs_add_entry(handle, child, cinode, hlock)
167 #else /* HAVE_LDISKFS_PDO */
173 struct htree_lock_head {
177 #define ldiskfs_htree_lock(lock, head, inode, op) do { LBUG(); } while (0)
178 #define ldiskfs_htree_unlock(lock) do { LBUG(); } while (0)
180 static inline struct htree_lock_head *ldiskfs_htree_lock_head_alloc(int dep)
186 #define ldiskfs_htree_lock_head_free(lh) do { LBUG(); } while (0)
188 #define LDISKFS_DUMMY_HTREE_LOCK 0xbabecafe
190 static inline struct htree_lock *ldiskfs_htree_lock_alloc(void)
192 return (struct htree_lock *)LDISKFS_DUMMY_HTREE_LOCK;
195 static inline void ldiskfs_htree_lock_free(struct htree_lock *lk)
197 LASSERT((unsigned long)lk == LDISKFS_DUMMY_HTREE_LOCK);
200 #define HTREE_HBITS_DEF 0
202 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
203 ll_ldiskfs_find_entry(dir, dentry, de)
204 #define osd_ldiskfs_add_entry(handle, child, cinode, lock) \
205 ldiskfs_add_entry(handle, child, cinode)
207 #endif /* HAVE_LDISKFS_PDO */
209 #define OSD_OTABLE_IT_CACHE_SIZE 128
210 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
212 struct osd_inconsistent_item {
213 /* link into osd_scrub::os_inconsistent_items,
214 * protected by osd_scrub::os_lock. */
217 /* The right FID <=> ino#/gen mapping. */
218 struct osd_idmap_cache oii_cache;
220 unsigned int oii_insert:1; /* insert or update mapping. */
223 struct osd_otable_cache {
224 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
226 /* Index for next cache slot to be filled. */
227 int ooc_producer_idx;
229 /* Index for next cache slot to be returned by it::next(). */
230 int ooc_consumer_idx;
232 /* How many items in ooc_cache. */
233 int ooc_cached_items;
235 /* Position for up layer LFSCK iteration pre-loading. */
236 __u32 ooc_pos_preload;
239 struct osd_otable_it {
240 struct osd_device *ooi_dev;
241 struct osd_otable_cache ooi_cache;
243 /* For osd_otable_it_key. */
246 /* The following bits can be updated/checked w/o lock protection.
247 * If more bits will be introduced in the future and need lock to
248 * protect, please add comment. */
249 unsigned long ooi_used_outside:1, /* Some user out of OSD
250 * uses the iteration. */
251 ooi_all_cached:1, /* No more entries can be
252 * filled into cache. */
253 ooi_user_ready:1, /* The user out of OSD is
254 * ready to iterate. */
255 ooi_waiting:1; /* it::next is waiting. */
258 extern const int osd_dto_credits_noquota[];
265 struct dt_device od_dt_dev;
266 /* information about underlying file system */
267 struct vfsmount *od_mnt;
269 struct osd_oi **od_oi_table;
270 /* total number of OI containers */
275 unsigned int od_fl_capa:1,
276 od_is_md:1; /* set in ->ldo_prepare */
277 unsigned long od_capa_timeout;
279 struct lustre_capa_key *od_capa_keys;
280 cfs_hlist_head_t *od_capa_hash;
282 cfs_proc_dir_entry_t *od_proc_entry;
283 struct lprocfs_stats *od_stats;
285 * statfs optimization: we cache a bit.
287 cfs_time_t od_osfs_age;
288 struct obd_statfs od_statfs;
289 spinlock_t od_osfs_lock;
291 unsigned int od_noscrub:1;
293 struct fsfilt_operations *od_fsops;
295 struct lu_site od_site;
297 struct osd_obj_map *od_ost_map;
299 unsigned long long od_readcache_max_filesize;
301 int od_writethrough_cache;
303 struct brw_stats od_brw_stats;
304 cfs_atomic_t od_r_in_flight;
305 cfs_atomic_t od_w_in_flight;
307 struct mutex od_otable_mutex;
308 struct osd_otable_it *od_otable_it;
309 struct osd_scrub od_scrub;
311 /* service name associated with the osd device */
312 char od_svname[MAX_OBD_NAME];
313 char od_mntdev[MAX_OBD_NAME];
315 /* quota slave instance */
316 struct qsd_instance *od_quota_slave;
319 #define OSD_TRACK_DECLARES
320 #ifdef OSD_TRACK_DECLARES
321 #define OSD_DECLARE_OP(oh, op, credits) \
323 LASSERT((oh)->ot_handle == NULL); \
324 ((oh)->ot_declare_ ##op)++; \
325 ((oh)->ot_declare_ ##op ##_cred) += (credits); \
326 (oh)->ot_credits += (credits); \
328 #define OSD_EXEC_OP(handle, op) \
330 struct osd_thandle *oh = container_of(handle, typeof(*oh), ot_super); \
331 LASSERT((oh)->ot_declare_ ##op > 0); \
332 ((oh)->ot_declare_ ##op)--; \
335 #define OSD_DECLARE_OP(oh, op, credits) (oh)->ot_credits += (credits)
336 #define OSD_EXEC_OP(oh, op)
339 /* There are at most 10 uid/gids are affected in a transaction, and
340 * that's rename case:
341 * - 2 for source parent uid & gid;
342 * - 2 for source child uid & gid ('..' entry update when child is directory);
343 * - 2 for target parent uid & gid;
344 * - 2 for target child uid & gid (if the target child exists);
345 * - 2 for root uid & gid (last_rcvd, llog, etc);
347 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
348 * the id type of each id in the ot_id_array.
350 #define OSD_MAX_UGID_CNT 10
353 struct thandle ot_super;
355 struct journal_callback ot_jcb;
356 cfs_list_t ot_dcb_list;
357 /* Link to the device, for debugging. */
358 struct lu_ref_link *ot_dev_link;
359 unsigned short ot_credits;
360 unsigned short ot_id_cnt;
361 unsigned short ot_id_type;
362 uid_t ot_id_array[OSD_MAX_UGID_CNT];
363 struct lquota_trans *ot_quota_trans;
365 #ifdef OSD_TRACK_DECLARES
366 /* Tracking for transaction credits, to allow debugging and optimizing
367 * cases where a large number of credits are being allocated for
368 * single transaction. */
369 unsigned char ot_declare_attr_set;
370 unsigned char ot_declare_punch;
371 unsigned char ot_declare_xattr_set;
372 unsigned char ot_declare_create;
373 unsigned char ot_declare_destroy;
374 unsigned char ot_declare_ref_add;
375 unsigned char ot_declare_ref_del;
376 unsigned char ot_declare_write;
377 unsigned char ot_declare_insert;
378 unsigned char ot_declare_delete;
379 unsigned char ot_declare_quota;
381 unsigned short ot_declare_attr_set_cred;
382 unsigned short ot_declare_punch_cred;
383 unsigned short ot_declare_xattr_set_cred;
384 unsigned short ot_declare_create_cred;
385 unsigned short ot_declare_destroy_cred;
386 unsigned short ot_declare_ref_add_cred;
387 unsigned short ot_declare_ref_del_cred;
388 unsigned short ot_declare_write_cred;
389 unsigned short ot_declare_insert_cred;
390 unsigned short ot_declare_delete_cred;
391 unsigned short ot_declare_quota_cred;
394 #if OSD_THANDLE_STATS
395 /** time when this handle was allocated */
396 cfs_time_t oth_alloced;
398 /** time when this thanle was started */
399 cfs_time_t oth_started;
404 * Basic transaction credit op
414 DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
428 LPROC_OSD_READ_BYTES = 0,
429 LPROC_OSD_WRITE_BYTES = 1,
430 LPROC_OSD_GET_PAGE = 2,
431 LPROC_OSD_NO_PAGE = 3,
432 LPROC_OSD_CACHE_ACCESS = 4,
433 LPROC_OSD_CACHE_HIT = 5,
434 LPROC_OSD_CACHE_MISS = 6,
436 #if OSD_THANDLE_STATS
437 LPROC_OSD_THANDLE_STARTING,
438 LPROC_OSD_THANDLE_OPEN,
439 LPROC_OSD_THANDLE_CLOSING,
446 * Storage representation for fids.
448 * Variable size, first byte contains the length of the whole record.
450 struct osd_fid_pack {
451 unsigned char fp_len;
452 char fp_area[sizeof(struct lu_fid)];
455 struct osd_it_ea_dirent {
456 struct lu_fid oied_fid;
459 unsigned short oied_namelen;
460 unsigned int oied_type;
462 } __attribute__((packed));
465 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
466 * than lu_dirent struct. osd readdir reads less number of dirent than
467 * required for mdd dir page. so buffer size need to be increased so that
468 * there would be one ext3 readdir for every mdd readdir page.
471 #define OSD_IT_EA_BUFSIZE (CFS_PAGE_SIZE + CFS_PAGE_SIZE/4)
474 * This is iterator's in-memory data structure in interoperability
475 * mode (i.e. iterator over ldiskfs style directory)
478 struct osd_object *oie_obj;
479 /** used in ldiskfs iterator, to stored file pointer */
480 struct file oie_file;
481 /** how many entries have been read-cached from storage */
483 /** current entry is being iterated by caller */
485 /** current processing entry */
486 struct osd_it_ea_dirent *oie_dirent;
487 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
492 * Iterator's in-memory data structure for IAM mode.
495 struct osd_object *oi_obj;
496 struct iam_path_descr *oi_ipd;
497 struct iam_iterator oi_it;
500 struct osd_quota_leaf {
506 * Iterator's in-memory data structure for quota file.
508 struct osd_it_quota {
509 struct osd_object *oiq_obj;
510 /** tree blocks path to where the entry is stored */
511 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
512 /** on-disk offset for current key where quota record can be found */
514 /** identifier for current quota record */
516 /** the record index in the leaf/index block */
517 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
518 /** list of already processed leaf blocks */
522 #define MAX_BLOCKS_PER_PAGE (CFS_PAGE_SIZE / 512)
526 cfs_atomic_t dr_numreqs; /* number of reqs being processed */
531 unsigned int dr_ignore_quota:1;
532 unsigned int dr_elapsed_valid:1; /* we really did count time */
533 unsigned int dr_rw:1;
534 struct page *dr_pages[PTLRPC_MAX_BRW_PAGES];
535 unsigned long dr_blocks[PTLRPC_MAX_BRW_PAGES*MAX_BLOCKS_PER_PAGE];
536 unsigned long dr_start_time;
537 unsigned long dr_elapsed; /* how long io took */
538 struct osd_device *dr_dev;
539 unsigned int dr_init_at; /* the line iobuf was initialized */
542 struct osd_thread_info {
543 const struct lu_env *oti_env;
545 * used for index operations.
547 struct dentry oti_obj_dentry;
548 struct dentry oti_child_dentry;
550 /** dentry for Iterator context. */
551 struct dentry oti_it_dentry;
552 struct htree_lock *oti_hlock;
554 struct lu_fid oti_fid;
555 struct lu_fid oti_fid2;
556 struct osd_inode_id oti_id;
557 struct osd_inode_id oti_id2;
558 struct ost_id oti_ostid;
561 * XXX temporary: for ->i_op calls.
563 struct timespec oti_time;
565 * XXX temporary: fake struct file for osd_object_sync
567 struct file oti_file;
569 * XXX temporary: for capa operations.
571 struct lustre_capa_key oti_capa_key;
572 struct lustre_capa oti_capa;
574 /** osd_device reference, initialized in osd_trans_start() and
575 used in osd_trans_stop() */
576 struct osd_device *oti_dev;
579 * following ipd and it structures are used for osd_index_iam_lookup()
580 * these are defined separately as we might do index operation
581 * in open iterator session.
584 /** osd iterator context used for iterator session */
587 struct osd_it_iam oti_it;
588 /* ldiskfs iterator data structure,
589 * see osd_it_ea_{init, fini} */
590 struct osd_it_ea oti_it_ea;
591 struct osd_it_quota oti_it_quota;
594 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
597 cfs_kstatfs_t oti_ksfs;
599 /** IAM iterator for index operation. */
600 struct iam_iterator oti_idx_it;
602 /** union to guarantee that ->oti_ipd[] has proper alignment. */
604 char oti_it_ipd[DX_IPD_MAX_SIZE];
605 long long oti_alignment_lieutenant;
609 char oti_idx_ipd[DX_IPD_MAX_SIZE];
610 long long oti_alignment_lieutenant_colonel;
613 struct osd_idmap_cache oti_cache;
618 /** used in osd_fid_set() to put xattr */
619 struct lu_buf oti_buf;
620 /** used in osd_ea_fid_set() to set fid into common ea */
622 struct lustre_mdt_attrs oti_mdt_attrs;
623 /* old LMA for compatibility */
624 char oti_mdt_attrs_old[LMA_OLD_SIZE];
627 struct osd_iobuf oti_iobuf;
628 struct inode oti_inode;
629 int oti_created[PTLRPC_MAX_BRW_PAGES];
630 struct lu_env oti_obj_delete_tx_env;
631 #define OSD_FID_REC_SZ 32
632 char oti_ldp[OSD_FID_REC_SZ];
633 char oti_ldp2[OSD_FID_REC_SZ];
635 /* used by quota code */
637 struct if_dqblk oti_dqblk;
638 struct if_dqinfo oti_dqinfo;
640 struct lquota_id_info oti_qi;
641 struct lquota_trans oti_quota_trans;
642 union lquota_rec oti_quota_rec;
646 extern int ldiskfs_pdo;
650 void lprocfs_osd_init_vars(struct lprocfs_static_vars *lvars);
651 int osd_procfs_init(struct osd_device *osd, const char *name);
652 int osd_procfs_fini(struct osd_device *osd);
653 void osd_lprocfs_time_start(const struct lu_env *env);
654 void osd_lprocfs_time_end(const struct lu_env *env,
655 struct osd_device *osd, int op);
656 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
659 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
660 struct obd_statfs *sfs);
661 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
662 struct lustre_capa *capa, __u64 opc);
663 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
664 struct osd_inode_id *id);
665 struct inode *osd_iget_fid(struct osd_thread_info *info, struct osd_device *dev,
666 struct osd_inode_id *id, struct lu_fid *fid);
668 int osd_obj_map_init(struct osd_device *osd);
669 void osd_obj_map_fini(struct osd_device *dev);
670 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
671 const struct lu_fid *fid, struct osd_inode_id *id);
672 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
673 const struct lu_fid *fid, const struct osd_inode_id *id,
675 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
676 const struct lu_fid *fid, struct thandle *th);
677 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
678 const struct lu_fid *fid, struct osd_inode_id *id);
679 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
680 const struct lu_fid *fid, const struct osd_inode_id *id,
683 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
684 int osd_scrub_file_store(struct osd_scrub *scrub);
685 int osd_scrub_start(struct osd_device *dev);
686 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
687 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
688 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
690 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
691 struct osd_inode_id *id);
692 int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
694 /* osd_quota_fmt.c */
695 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
696 int type, uint blk, int depth, uint index,
697 struct osd_it_quota *it);
698 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
699 int type, uint blk, uint index,
700 struct osd_it_quota *it);
701 loff_t find_tree_dqentry(const struct lu_env *env,
702 struct osd_object *obj, int type,
703 qid_t dqid, uint blk, int depth,
704 struct osd_it_quota *it);
706 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
707 struct lquota_id_info *qi, bool allocated, int *flags);
708 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
709 long long space, struct osd_thandle *oh,
710 bool is_blk, bool allocated, int *flags, bool force);
711 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
712 const struct dt_rec *rec,
713 union lquota_rec *quota_rec);
714 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
715 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
716 const struct dt_index_features *feat);
718 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
720 return (feat == &dt_quota_iusr_features ||
721 feat == &dt_quota_busr_features ||
722 feat == &dt_quota_igrp_features ||
723 feat == &dt_quota_bgrp_features) ? true : false;
727 * Invariants, assertions.
731 * XXX: do not enable this, until invariant checking code is made thread safe
732 * in the face of pdirops locking.
734 #define OSD_INVARIANT_CHECKS (0)
736 #if OSD_INVARIANT_CHECKS
737 static inline int osd_invariant(const struct osd_object *obj)
741 ergo(obj->oo_inode != NULL,
742 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
743 atomic_read(&obj->oo_inode->i_count) > 0) &&
744 ergo(obj->oo_dir != NULL &&
745 obj->oo_dir->od_conationer.ic_object != NULL,
746 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
749 #define osd_invariant(obj) (1)
752 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
754 extern const struct dt_index_operations osd_otable_ops;
756 static inline int osd_oi_fid2idx(struct osd_device *dev,
757 const struct lu_fid *fid)
759 return fid->f_seq & (dev->od_oi_count - 1);
762 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
763 const struct lu_fid *fid)
765 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
766 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
767 LASSERTF(!fid_is_igif(fid), DFID"\n", PFID(fid));
768 LASSERT(osd->od_oi_table != NULL && osd->od_oi_count >= 1);
769 /* It can work even od_oi_count equals to 1 although it's unexpected,
770 * the only reason we set it to 1 is for performance measurement */
771 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
774 extern const struct lu_device_operations osd_lu_ops;
776 static inline int lu_device_is_osd(const struct lu_device *d)
778 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
781 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
783 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
784 return container_of0(d, struct osd_device, od_dt_dev);
787 static inline struct osd_device *osd_dev(const struct lu_device *d)
789 LASSERT(lu_device_is_osd(d));
790 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
793 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
795 return osd_dev(o->oo_dt.do_lu.lo_dev);
798 static inline struct super_block *osd_sb(const struct osd_device *dev)
800 return dev->od_mnt->mnt_sb;
803 static inline int osd_object_is_root(const struct osd_object *obj)
805 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
808 static inline struct osd_object *osd_obj(const struct lu_object *o)
810 LASSERT(lu_device_is_osd(o->lo_dev));
811 return container_of0(o, struct osd_object, oo_dt.do_lu);
814 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
816 return osd_obj(&d->do_lu);
819 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
821 return &osd->od_dt_dev.dd_lu_dev;
824 static inline journal_t *osd_journal(const struct osd_device *dev)
826 return LDISKFS_SB(osd_sb(dev))->s_journal;
829 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
831 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
834 static inline char *osd_name(struct osd_device *osd)
836 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
840 extern const struct dt_body_operations osd_body_ops;
841 extern struct lu_context_key osd_key;
843 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
845 return lu_context_key_get(&env->le_ctx, &osd_key);
848 extern const struct dt_body_operations osd_body_ops_new;
854 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
855 const struct iam_container *bag)
857 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
858 osd_oti_get(env)->oti_it_ipd);
862 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
863 const struct iam_container *bag)
865 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
866 osd_oti_get(env)->oti_idx_ipd);
869 static inline void osd_ipd_put(const struct lu_env *env,
870 const struct iam_container *bag,
871 struct iam_path_descr *ipd)
873 bag->ic_descr->id_ops->id_ipd_free(ipd);
876 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
877 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
878 int write_NUL, loff_t *offs, handle_t *handle);
881 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
883 const char *name, const int namelen)
885 struct osd_thread_info *info = osd_oti_get(env);
886 struct dentry *child_dentry = &info->oti_child_dentry;
887 struct dentry *obj_dentry = &info->oti_obj_dentry;
889 obj_dentry->d_inode = inode;
890 obj_dentry->d_sb = inode->i_sb;
891 obj_dentry->d_name.hash = 0;
893 child_dentry->d_name.hash = 0;
894 child_dentry->d_parent = obj_dentry;
895 child_dentry->d_name.name = name;
896 child_dentry->d_name.len = namelen;
901 * Helper function to pack the fid, ldiskfs stores fid in packed format.
904 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
905 struct lu_fid *befider)
907 fid_cpu_to_be(befider, (struct lu_fid *)fid);
908 memcpy(pack->fp_area, befider, sizeof(*befider));
909 pack->fp_len = sizeof(*befider) + 1;
913 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
918 switch (pack->fp_len) {
919 case sizeof *fid + 1:
920 memcpy(fid, pack->fp_area, sizeof *fid);
921 fid_be_to_cpu(fid, fid);
924 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
931 * Quota/Accounting handling
933 extern const struct dt_index_operations osd_acct_index_ops;
934 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
935 const struct lu_fid *fid, struct osd_inode_id *id);
937 /* copy from fs/ext4/dir.c */
938 static inline int is_32bit_api(void)
941 return is_compat_task();
943 return (BITS_PER_LONG == 32);
947 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
949 if ((filp->f_mode & FMODE_32BITHASH) ||
950 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
951 return LDISKFS_HTREE_EOF_32BIT;
953 return LDISKFS_HTREE_EOF_64BIT;
956 #endif /* __KERNEL__ */
957 #endif /* _OSD_INTERNAL_H */