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>
58 /* fsfilt_{get|put}_ops */
59 #include <lustre_fsfilt.h>
63 /* class_register_type(), class_unregister_type(), class_get_type() */
64 #include <obd_class.h>
65 #include <lustre_disk.h>
66 #include <dt_object.h>
67 #include <lustre_quota.h>
71 #include "osd_scrub.h"
72 #include "osd_quota_fmt.h"
76 #define OSD_COUNTERS (0)
78 /* ldiskfs special inode::i_state_flags need to be accessed with
79 * ldiskfs_{set,clear,test}_inode_state() only */
81 /* OI scrub should skip this inode. */
82 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
84 /* Do not add OI mapping for this inode. */
85 #define LDISKFS_STATE_LUSTRE_NO_OI 30
87 /** Enable thandle usage statistics */
88 #define OSD_THANDLE_STATS (0)
90 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
92 #define OBJECTS "OBJECTS"
93 #define ADMIN_USR "admin_quotafile_v2.usr"
94 #define ADMIN_GRP "admin_quotafile_v2.grp"
96 struct osd_directory {
97 struct iam_container od_container;
98 struct iam_descr od_descr;
102 * Object Index (oi) instance.
106 * underlying index object, where fid->id mapping in stored.
108 struct inode *oi_inode;
109 struct osd_directory oi_dir;
112 extern const int osd_dto_credits_noquota[];
115 struct dt_object oo_dt;
117 * Inode for file system object represented by this osd_object. This
118 * inode is pinned for the whole duration of lu_object life.
120 * Not modified concurrently (either setup early during object
121 * creation, or assigned by osd_object_create() under write lock).
123 struct inode *oo_inode;
125 * to protect index ops.
127 struct htree_lock_head *oo_hl_head;
128 struct rw_semaphore oo_ext_idx_sem;
129 struct rw_semaphore oo_sem;
130 struct osd_directory *oo_dir;
131 /** protects inode attributes. */
134 * Following two members are used to indicate the presence of dot and
135 * dotdot in the given directory. This is required for interop mode
138 int oo_compat_dot_created;
139 int oo_compat_dotdot_created;
141 const struct lu_env *oo_owner;
142 #ifdef CONFIG_LOCKDEP
143 struct lockdep_map oo_dep_map;
148 /* protects on-fly initialization */
149 int oos_subdir_count; /* subdir count for each seq */
150 struct dentry *oos_root; /* O/<seq> */
151 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
152 obd_seq oos_seq; /* seq number */
153 cfs_list_t oos_seq_list; /* list to seq_list */
157 struct dentry *om_root; /* dentry for /O */
158 rwlock_t om_seq_list_lock; /* lock for seq_list */
159 cfs_list_t om_seq_list; /* list head for seq */
161 struct semaphore om_dir_init_sem;
165 struct dentry *om_root; /* AGENT/<index> */
166 obd_seq om_index; /* mdt index */
167 cfs_list_t om_list; /* list to omm_list */
170 struct osd_mdobj_map {
171 struct dentry *omm_remote_parent;
174 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
175 ll_ldiskfs_find_entry(dir, dentry, de, lock)
176 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
177 ldiskfs_add_entry(handle, child, cinode, hlock)
179 #define OSD_OTABLE_IT_CACHE_SIZE 128
180 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
182 struct osd_inconsistent_item {
183 /* link into osd_scrub::os_inconsistent_items,
184 * protected by osd_scrub::os_lock. */
187 /* The right FID <=> ino#/gen mapping. */
188 struct osd_idmap_cache oii_cache;
190 unsigned int oii_insert:1; /* insert or update mapping. */
193 struct osd_otable_cache {
194 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
196 /* Index for next cache slot to be filled. */
197 int ooc_producer_idx;
199 /* Index for next cache slot to be returned by it::next(). */
200 int ooc_consumer_idx;
202 /* How many items in ooc_cache. */
203 int ooc_cached_items;
205 /* Position for up layer LFSCK iteration pre-loading. */
206 __u32 ooc_pos_preload;
209 struct osd_otable_it {
210 struct osd_device *ooi_dev;
212 struct osd_otable_cache ooi_cache;
214 /* The following bits can be updated/checked w/o lock protection.
215 * If more bits will be introduced in the future and need lock to
216 * protect, please add comment. */
217 unsigned long ooi_used_outside:1, /* Some user out of OSD
218 * uses the iteration. */
219 ooi_all_cached:1, /* No more entries can be
220 * filled into cache. */
221 ooi_user_ready:1, /* The user out of OSD is
222 * ready to iterate. */
223 ooi_waiting:1, /* it::next is waiting. */
224 ooi_stopping:1; /* Someone is stopping
228 extern const int osd_dto_credits_noquota[];
235 struct dt_device od_dt_dev;
236 /* information about underlying file system */
237 struct vfsmount *od_mnt;
239 struct osd_oi **od_oi_table;
240 /* total number of OI containers */
245 unsigned int od_fl_capa:1,
246 od_is_md:1; /* set in ->ldo_prepare */
247 unsigned long od_capa_timeout;
249 struct lustre_capa_key *od_capa_keys;
250 cfs_hlist_head_t *od_capa_hash;
252 cfs_proc_dir_entry_t *od_proc_entry;
253 struct lprocfs_stats *od_stats;
255 * statfs optimization: we cache a bit.
257 cfs_time_t od_osfs_age;
258 struct obd_statfs od_statfs;
259 spinlock_t od_osfs_lock;
261 unsigned int od_noscrub:1,
265 struct fsfilt_operations *od_fsops;
267 struct lu_site od_site;
269 struct osd_obj_map *od_ost_map;
270 struct osd_mdobj_map *od_mdt_map;
272 unsigned long long od_readcache_max_filesize;
274 int od_writethrough_cache;
276 struct brw_stats od_brw_stats;
277 cfs_atomic_t od_r_in_flight;
278 cfs_atomic_t od_w_in_flight;
280 struct mutex od_otable_mutex;
281 struct osd_otable_it *od_otable_it;
282 struct osd_scrub od_scrub;
283 cfs_list_t od_ios_list;
285 /* service name associated with the osd device */
286 char od_svname[MAX_OBD_NAME];
287 char od_mntdev[MAX_OBD_NAME];
289 /* quota slave instance */
290 struct qsd_instance *od_quota_slave;
293 /* There are at most 10 uid/gids are affected in a transaction, and
294 * that's rename case:
295 * - 2 for source parent uid & gid;
296 * - 2 for source child uid & gid ('..' entry update when child is directory);
297 * - 2 for target parent uid & gid;
298 * - 2 for target child uid & gid (if the target child exists);
299 * - 2 for root uid & gid (last_rcvd, llog, etc);
301 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
302 * the id type of each id in the ot_id_array.
304 #define OSD_MAX_UGID_CNT 10
309 OSD_OT_XATTR_SET = 2,
321 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 3, 90, 0)
322 # define OSD_TRACK_DECLARES
326 struct thandle ot_super;
328 struct ldiskfs_journal_cb_entry ot_jcb;
329 cfs_list_t ot_dcb_list;
330 /* Link to the device, for debugging. */
331 struct lu_ref_link *ot_dev_link;
332 unsigned short ot_credits;
333 unsigned short ot_id_cnt;
334 unsigned short ot_id_type;
335 uid_t ot_id_array[OSD_MAX_UGID_CNT];
336 struct lquota_trans *ot_quota_trans;
337 #if OSD_THANDLE_STATS
338 /** time when this handle was allocated */
339 cfs_time_t oth_alloced;
341 /** time when this thanle was started */
342 cfs_time_t oth_started;
347 * Basic transaction credit op
357 DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
371 LPROC_OSD_READ_BYTES = 0,
372 LPROC_OSD_WRITE_BYTES = 1,
373 LPROC_OSD_GET_PAGE = 2,
374 LPROC_OSD_NO_PAGE = 3,
375 LPROC_OSD_CACHE_ACCESS = 4,
376 LPROC_OSD_CACHE_HIT = 5,
377 LPROC_OSD_CACHE_MISS = 6,
379 #if OSD_THANDLE_STATS
380 LPROC_OSD_THANDLE_STARTING,
381 LPROC_OSD_THANDLE_OPEN,
382 LPROC_OSD_THANDLE_CLOSING,
389 * Storage representation for fids.
391 * Variable size, first byte contains the length of the whole record.
393 struct osd_fid_pack {
394 unsigned char fp_len;
395 char fp_area[sizeof(struct lu_fid)];
398 struct osd_it_ea_dirent {
399 struct lu_fid oied_fid;
402 unsigned short oied_namelen;
403 unsigned int oied_type;
405 } __attribute__((packed));
408 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
409 * than lu_dirent struct. osd readdir reads less number of dirent than
410 * required for mdd dir page. so buffer size need to be increased so that
411 * there would be one ext3 readdir for every mdd readdir page.
414 #define OSD_IT_EA_BUFSIZE (CFS_PAGE_SIZE + CFS_PAGE_SIZE/4)
417 * This is iterator's in-memory data structure in interoperability
418 * mode (i.e. iterator over ldiskfs style directory)
421 struct osd_object *oie_obj;
422 /** used in ldiskfs iterator, to stored file pointer */
423 struct file oie_file;
424 /** how many entries have been read-cached from storage */
426 /** current entry is being iterated by caller */
428 /** current processing entry */
429 struct osd_it_ea_dirent *oie_dirent;
430 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
435 * Iterator's in-memory data structure for IAM mode.
438 struct osd_object *oi_obj;
439 struct iam_path_descr *oi_ipd;
440 struct iam_iterator oi_it;
443 struct osd_quota_leaf {
449 * Iterator's in-memory data structure for quota file.
451 struct osd_it_quota {
452 struct osd_object *oiq_obj;
453 /** tree blocks path to where the entry is stored */
454 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
455 /** on-disk offset for current key where quota record can be found */
457 /** identifier for current quota record */
459 /** the record index in the leaf/index block */
460 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
461 /** list of already processed leaf blocks */
465 #define MAX_BLOCKS_PER_PAGE (CFS_PAGE_SIZE / 512)
469 cfs_atomic_t dr_numreqs; /* number of reqs being processed */
474 unsigned int dr_ignore_quota:1;
475 unsigned int dr_elapsed_valid:1; /* we really did count time */
476 unsigned int dr_rw:1;
477 struct lu_buf dr_pg_buf;
478 struct page **dr_pages;
479 struct lu_buf dr_bl_buf;
480 unsigned long *dr_blocks;
481 unsigned long dr_start_time;
482 unsigned long dr_elapsed; /* how long io took */
483 struct osd_device *dr_dev;
484 unsigned int dr_init_at; /* the line iobuf was initialized */
487 struct osd_thread_info {
488 const struct lu_env *oti_env;
490 * used for index operations.
492 struct dentry oti_obj_dentry;
493 struct dentry oti_child_dentry;
495 /** dentry for Iterator context. */
496 struct dentry oti_it_dentry;
497 struct htree_lock *oti_hlock;
499 struct lu_fid oti_fid;
500 struct lu_fid oti_fid2;
501 struct osd_inode_id oti_id;
502 struct osd_inode_id oti_id2;
503 struct ost_id oti_ostid;
506 * XXX temporary: for ->i_op calls.
508 struct timespec oti_time;
510 * XXX temporary: fake struct file for osd_object_sync
512 struct file oti_file;
514 * XXX temporary: for capa operations.
516 struct lustre_capa_key oti_capa_key;
517 struct lustre_capa oti_capa;
519 /** osd_device reference, initialized in osd_trans_start() and
520 used in osd_trans_stop() */
521 struct osd_device *oti_dev;
524 * following ipd and it structures are used for osd_index_iam_lookup()
525 * these are defined separately as we might do index operation
526 * in open iterator session.
529 /** osd iterator context used for iterator session */
532 struct osd_it_iam oti_it;
533 /* ldiskfs iterator data structure,
534 * see osd_it_ea_{init, fini} */
535 struct osd_it_ea oti_it_ea;
536 struct osd_it_quota oti_it_quota;
539 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
542 struct kstatfs oti_ksfs;
544 /** IAM iterator for index operation. */
545 struct iam_iterator oti_idx_it;
547 /** union to guarantee that ->oti_ipd[] has proper alignment. */
549 char oti_it_ipd[DX_IPD_MAX_SIZE];
550 long long oti_alignment_lieutenant;
554 char oti_idx_ipd[DX_IPD_MAX_SIZE];
555 long long oti_alignment_lieutenant_colonel;
558 struct osd_idmap_cache oti_cache;
563 /** used in osd_fid_set() to put xattr */
564 struct lu_buf oti_buf;
565 /** used in osd_ea_fid_set() to set fid into common ea */
567 struct lustre_mdt_attrs oti_mdt_attrs;
568 /* old LMA for compatibility */
569 char oti_mdt_attrs_old[LMA_OLD_SIZE];
572 struct osd_iobuf oti_iobuf;
573 struct inode oti_inode;
574 #define OSD_FID_REC_SZ 32
575 char oti_ldp[OSD_FID_REC_SZ];
576 char oti_ldp2[OSD_FID_REC_SZ];
578 /* used by quota code */
580 #ifdef HAVE_DQUOT_FS_DISK_QUOTA
581 struct fs_disk_quota oti_fdq;
583 struct if_dqblk oti_dqblk;
585 struct if_dqinfo oti_dqinfo;
587 struct lquota_id_info oti_qi;
588 struct lquota_trans oti_quota_trans;
589 union lquota_rec oti_quota_rec;
591 struct lu_seq_range oti_seq_range;
593 #ifdef OSD_TRACK_DECLARES
594 /* Tracking for transaction credits, to allow debugging and optimizing
595 * cases where a large number of credits are being allocated for
596 * single transaction. */
597 unsigned char oti_declare_ops[OSD_OT_MAX];
598 unsigned char oti_declare_ops_rb[OSD_OT_MAX];
599 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
606 extern int ldiskfs_pdo;
608 static inline int __osd_xattr_set(struct osd_thread_info *info,
609 struct inode *inode, const char *name,
610 const void *buf, int buflen, int fl)
612 struct dentry *dentry = &info->oti_child_dentry;
614 ll_vfs_dq_init(inode);
615 dentry->d_inode = inode;
616 return inode->i_op->setxattr(dentry, name, buf, buflen, fl);
621 void lprocfs_osd_init_vars(struct lprocfs_static_vars *lvars);
622 int osd_procfs_init(struct osd_device *osd, const char *name);
623 int osd_procfs_fini(struct osd_device *osd);
624 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
627 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
628 struct obd_statfs *sfs);
629 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
630 struct lustre_capa *capa, __u64 opc);
631 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
632 struct osd_inode_id *id);
633 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
634 const struct lu_fid *fid, __u64 flags);
635 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
636 struct dentry *dentry, struct lustre_mdt_attrs *lma);
638 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
639 void osd_obj_map_fini(struct osd_device *dev);
640 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
641 const struct lu_fid *fid, struct osd_inode_id *id);
642 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
643 const struct lu_fid *fid, const struct osd_inode_id *id,
645 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
646 const struct lu_fid *fid, struct thandle *th);
647 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
648 const struct lu_fid *fid, struct osd_inode_id *id);
649 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
650 const struct lu_fid *fid, const struct osd_inode_id *id,
653 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
654 int osd_scrub_file_store(struct osd_scrub *scrub);
655 char *osd_lf_fid2name(const struct lu_fid *fid);
656 int osd_scrub_start(struct osd_device *dev);
657 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
658 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
659 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
661 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
662 struct osd_inode_id *id);
663 int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
665 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
666 const struct lu_fid *fid, struct lu_seq_range *range);
668 int osd_delete_from_remote_parent(const struct lu_env *env,
669 struct osd_device *osd,
670 struct osd_object *obj,
671 struct osd_thandle *oh);
672 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
673 struct osd_object *obj, struct osd_thandle *oh);
675 /* osd_quota_fmt.c */
676 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
677 int type, uint blk, int depth, uint index,
678 struct osd_it_quota *it);
679 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
680 int type, uint blk, uint index,
681 struct osd_it_quota *it);
682 loff_t find_tree_dqentry(const struct lu_env *env,
683 struct osd_object *obj, int type,
684 qid_t dqid, uint blk, int depth,
685 struct osd_it_quota *it);
687 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
688 struct lquota_id_info *qi, bool allocated, int *flags);
689 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
690 long long space, struct osd_thandle *oh,
691 bool is_blk, bool allocated, int *flags, bool force);
692 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
693 const struct dt_rec *rec,
694 union lquota_rec *quota_rec);
695 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
696 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
697 const struct dt_index_features *feat);
700 struct osd_obj_seq *osd_seq_load(struct osd_device *osd, obd_seq seq);
702 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
704 return (feat == &dt_quota_iusr_features ||
705 feat == &dt_quota_busr_features ||
706 feat == &dt_quota_igrp_features ||
707 feat == &dt_quota_bgrp_features) ? true : false;
711 * Invariants, assertions.
715 * XXX: do not enable this, until invariant checking code is made thread safe
716 * in the face of pdirops locking.
718 #define OSD_INVARIANT_CHECKS (0)
720 #if OSD_INVARIANT_CHECKS
721 static inline int osd_invariant(const struct osd_object *obj)
725 ergo(obj->oo_inode != NULL,
726 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
727 atomic_read(&obj->oo_inode->i_count) > 0) &&
728 ergo(obj->oo_dir != NULL &&
729 obj->oo_dir->od_conationer.ic_object != NULL,
730 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
733 #define osd_invariant(obj) (1)
736 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
738 extern const struct dt_index_operations osd_otable_ops;
740 static inline int osd_oi_fid2idx(struct osd_device *dev,
741 const struct lu_fid *fid)
743 return fid->f_seq & (dev->od_oi_count - 1);
746 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
747 const struct lu_fid *fid)
749 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
750 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
751 LASSERT(osd->od_oi_table != NULL && osd->od_oi_count >= 1);
752 /* It can work even od_oi_count equals to 1 although it's unexpected,
753 * the only reason we set it to 1 is for performance measurement */
754 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
757 extern const struct lu_device_operations osd_lu_ops;
759 static inline int lu_device_is_osd(const struct lu_device *d)
761 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
764 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
766 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
767 return container_of0(d, struct osd_device, od_dt_dev);
770 static inline struct osd_device *osd_dev(const struct lu_device *d)
772 LASSERT(lu_device_is_osd(d));
773 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
776 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
778 return osd_dev(o->oo_dt.do_lu.lo_dev);
781 static inline struct super_block *osd_sb(const struct osd_device *dev)
783 return dev->od_mnt->mnt_sb;
786 static inline int osd_object_is_root(const struct osd_object *obj)
788 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
791 static inline struct osd_object *osd_obj(const struct lu_object *o)
793 LASSERT(lu_device_is_osd(o->lo_dev));
794 return container_of0(o, struct osd_object, oo_dt.do_lu);
797 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
799 return osd_obj(&d->do_lu);
802 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
804 return &osd->od_dt_dev.dd_lu_dev;
807 static inline journal_t *osd_journal(const struct osd_device *dev)
809 return LDISKFS_SB(osd_sb(dev))->s_journal;
812 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
814 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
817 static inline char *osd_name(struct osd_device *osd)
819 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
823 extern const struct dt_body_operations osd_body_ops;
824 extern struct lu_context_key osd_key;
826 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
828 return lu_context_key_get(&env->le_ctx, &osd_key);
831 extern const struct dt_body_operations osd_body_ops_new;
837 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
838 const struct iam_container *bag)
840 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
841 osd_oti_get(env)->oti_it_ipd);
845 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
846 const struct iam_container *bag)
848 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
849 osd_oti_get(env)->oti_idx_ipd);
852 static inline void osd_ipd_put(const struct lu_env *env,
853 const struct iam_container *bag,
854 struct iam_path_descr *ipd)
856 bag->ic_descr->id_ops->id_ipd_free(ipd);
859 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
860 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
861 int write_NUL, loff_t *offs, handle_t *handle);
864 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
866 const char *name, const int namelen)
868 struct osd_thread_info *info = osd_oti_get(env);
869 struct dentry *child_dentry = &info->oti_child_dentry;
870 struct dentry *obj_dentry = &info->oti_obj_dentry;
872 obj_dentry->d_inode = inode;
873 obj_dentry->d_sb = inode->i_sb;
874 obj_dentry->d_name.hash = 0;
876 child_dentry->d_name.hash = 0;
877 child_dentry->d_parent = obj_dentry;
878 child_dentry->d_name.name = name;
879 child_dentry->d_name.len = namelen;
883 #ifdef OSD_TRACK_DECLARES
884 extern int osd_trans_declare_op2rb[];
886 static inline void osd_trans_declare_op(const struct lu_env *env,
887 struct osd_thandle *oh,
888 unsigned int op, int credits)
890 struct osd_thread_info *oti = osd_oti_get(env);
892 LASSERT(oh->ot_handle == NULL);
893 LASSERT(op < OSD_OT_MAX);
895 oti->oti_declare_ops[op]++;
896 oti->oti_declare_ops_cred[op] += credits;
897 oh->ot_credits += credits;
900 static inline void osd_trans_exec_op(const struct lu_env *env,
901 struct thandle *th, unsigned int op)
903 struct osd_thread_info *oti = osd_oti_get(env);
904 struct osd_thandle *oh = container_of(th, struct osd_thandle,
908 LASSERT(oh->ot_handle != NULL);
909 LASSERT(op < OSD_OT_MAX);
911 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
912 oti->oti_declare_ops[op]--;
913 oti->oti_declare_ops_rb[op]++;
915 /* all future updates are considered rollback */
916 oti->oti_rollback = true;
917 rb = osd_trans_declare_op2rb[op];
918 LASSERTF(rb < OSD_OT_MAX, "op = %u\n", op);
919 LASSERTF(oti->oti_declare_ops_rb[rb] > 0, "rb = %u\n", rb);
920 oti->oti_declare_ops_rb[rb]--;
924 static inline void osd_trans_declare_rb(const struct lu_env *env,
925 struct thandle *th, unsigned int op)
927 struct osd_thread_info *oti = osd_oti_get(env);
928 struct osd_thandle *oh = container_of(th, struct osd_thandle,
931 LASSERT(oh->ot_handle != NULL);
932 LASSERT(op < OSD_OT_MAX);
934 oti->oti_declare_ops_rb[op]++;
937 static inline void osd_trans_declare_op(const struct lu_env *env,
938 struct osd_thandle *oh,
939 unsigned int op, int credits)
941 oh->ot_credits += credits;
944 static inline void osd_trans_exec_op(const struct lu_env *env,
945 struct thandle *th, unsigned int op)
949 static inline void osd_trans_declare_rb(const struct lu_env *env,
950 struct thandle *th, unsigned int op)
956 * Helper function to pack the fid, ldiskfs stores fid in packed format.
959 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
960 struct lu_fid *befider)
962 fid_cpu_to_be(befider, (struct lu_fid *)fid);
963 memcpy(pack->fp_area, befider, sizeof(*befider));
964 pack->fp_len = sizeof(*befider) + 1;
968 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
973 switch (pack->fp_len) {
974 case sizeof *fid + 1:
975 memcpy(fid, pack->fp_area, sizeof *fid);
976 fid_be_to_cpu(fid, fid);
979 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
986 * Quota/Accounting handling
988 extern const struct dt_index_operations osd_acct_index_ops;
989 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
990 const struct lu_fid *fid, struct osd_inode_id *id);
992 /* copy from fs/ext4/dir.c */
993 static inline int is_32bit_api(void)
996 return is_compat_task();
998 return (BITS_PER_LONG == 32);
1002 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1004 if ((filp->f_mode & FMODE_32BITHASH) ||
1005 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1006 return LDISKFS_HTREE_EOF_32BIT;
1008 return LDISKFS_HTREE_EOF_64BIT;
1011 #endif /* __KERNEL__ */
1012 #endif /* _OSD_INTERNAL_H */