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 /* ldiskfs special inode::i_state_flags need to be accessed with
87 * ldiskfs_{set,clear,test}_inode_state() only */
89 /* OI scrub should skip this inode. */
90 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
92 /* Do not add OI mapping for this inode. */
93 #define LDISKFS_STATE_LUSTRE_NO_OI 30
95 /** Enable thandle usage statistics */
96 #define OSD_THANDLE_STATS (0)
98 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
100 #define OBJECTS "OBJECTS"
101 #define ADMIN_USR "admin_quotafile_v2.usr"
102 #define ADMIN_GRP "admin_quotafile_v2.grp"
104 struct osd_directory {
105 struct iam_container od_container;
106 struct iam_descr od_descr;
110 * Object Index (oi) instance.
114 * underlying index object, where fid->id mapping in stored.
116 struct inode *oi_inode;
117 struct osd_directory oi_dir;
120 extern const int osd_dto_credits_noquota[];
123 struct dt_object oo_dt;
125 * Inode for file system object represented by this osd_object. This
126 * inode is pinned for the whole duration of lu_object life.
128 * Not modified concurrently (either setup early during object
129 * creation, or assigned by osd_object_create() under write lock).
131 struct inode *oo_inode;
133 * to protect index ops.
135 struct htree_lock_head *oo_hl_head;
136 struct rw_semaphore oo_ext_idx_sem;
137 struct rw_semaphore oo_sem;
138 struct osd_directory *oo_dir;
139 /** protects inode attributes. */
142 * Following two members are used to indicate the presence of dot and
143 * dotdot in the given directory. This is required for interop mode
146 int oo_compat_dot_created;
147 int oo_compat_dotdot_created;
149 const struct lu_env *oo_owner;
150 #ifdef CONFIG_LOCKDEP
151 struct lockdep_map oo_dep_map;
156 /* protects on-fly initialization */
157 int oos_subdir_count; /* subdir count for each seq */
158 struct dentry *oos_root; /* O/<seq> */
159 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
160 obd_seq oos_seq; /* seq number */
161 cfs_list_t oos_seq_list; /* list to seq_list */
165 struct dentry *om_root; /* dentry for /O */
166 rwlock_t om_seq_list_lock; /* lock for seq_list */
167 cfs_list_t om_seq_list; /* list head for seq */
169 struct semaphore om_dir_init_sem;
173 struct dentry *om_root; /* AGENT/<index> */
174 obd_seq om_index; /* mdt index */
175 cfs_list_t om_list; /* list to omm_list */
178 struct osd_mdobj_map {
179 struct dentry *omm_agent_dentry;
182 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
183 ll_ldiskfs_find_entry(dir, dentry, de, lock)
184 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
185 ldiskfs_add_entry(handle, child, cinode, hlock)
187 #define OSD_OTABLE_IT_CACHE_SIZE 128
188 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
190 struct osd_inconsistent_item {
191 /* link into osd_scrub::os_inconsistent_items,
192 * protected by osd_scrub::os_lock. */
195 /* The right FID <=> ino#/gen mapping. */
196 struct osd_idmap_cache oii_cache;
198 unsigned int oii_insert:1; /* insert or update mapping. */
201 struct osd_otable_cache {
202 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
204 /* Index for next cache slot to be filled. */
205 int ooc_producer_idx;
207 /* Index for next cache slot to be returned by it::next(). */
208 int ooc_consumer_idx;
210 /* How many items in ooc_cache. */
211 int ooc_cached_items;
213 /* Position for up layer LFSCK iteration pre-loading. */
214 __u32 ooc_pos_preload;
217 struct osd_otable_it {
218 struct osd_device *ooi_dev;
219 struct osd_otable_cache ooi_cache;
221 /* For osd_otable_it_key. */
224 /* The following bits can be updated/checked w/o lock protection.
225 * If more bits will be introduced in the future and need lock to
226 * protect, please add comment. */
227 unsigned long ooi_used_outside:1, /* Some user out of OSD
228 * uses the iteration. */
229 ooi_all_cached:1, /* No more entries can be
230 * filled into cache. */
231 ooi_user_ready:1, /* The user out of OSD is
232 * ready to iterate. */
233 ooi_waiting:1; /* it::next is waiting. */
236 extern const int osd_dto_credits_noquota[];
243 struct dt_device od_dt_dev;
244 /* information about underlying file system */
245 struct vfsmount *od_mnt;
247 struct osd_oi **od_oi_table;
248 /* total number of OI containers */
253 unsigned int od_fl_capa:1,
254 od_is_md:1; /* set in ->ldo_prepare */
255 unsigned long od_capa_timeout;
257 struct lustre_capa_key *od_capa_keys;
258 cfs_hlist_head_t *od_capa_hash;
260 cfs_proc_dir_entry_t *od_proc_entry;
261 struct lprocfs_stats *od_stats;
263 * statfs optimization: we cache a bit.
265 cfs_time_t od_osfs_age;
266 struct obd_statfs od_statfs;
267 spinlock_t od_osfs_lock;
269 unsigned int od_noscrub:1,
273 struct fsfilt_operations *od_fsops;
275 struct lu_site od_site;
277 struct osd_obj_map *od_ost_map;
278 struct osd_mdobj_map *od_mdt_map;
280 unsigned long long od_readcache_max_filesize;
282 int od_writethrough_cache;
284 struct brw_stats od_brw_stats;
285 cfs_atomic_t od_r_in_flight;
286 cfs_atomic_t od_w_in_flight;
288 struct mutex od_otable_mutex;
289 struct osd_otable_it *od_otable_it;
290 struct osd_scrub od_scrub;
291 cfs_list_t od_ios_list;
293 /* service name associated with the osd device */
294 char od_svname[MAX_OBD_NAME];
295 char od_mntdev[MAX_OBD_NAME];
297 /* quota slave instance */
298 struct qsd_instance *od_quota_slave;
301 /* There are at most 10 uid/gids are affected in a transaction, and
302 * that's rename case:
303 * - 2 for source parent uid & gid;
304 * - 2 for source child uid & gid ('..' entry update when child is directory);
305 * - 2 for target parent uid & gid;
306 * - 2 for target child uid & gid (if the target child exists);
307 * - 2 for root uid & gid (last_rcvd, llog, etc);
309 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
310 * the id type of each id in the ot_id_array.
312 #define OSD_MAX_UGID_CNT 10
317 OSD_OT_XATTR_SET = 2,
329 #define OSD_TRACK_DECLARES
332 struct thandle ot_super;
334 struct journal_callback ot_jcb;
335 cfs_list_t ot_dcb_list;
336 /* Link to the device, for debugging. */
337 struct lu_ref_link *ot_dev_link;
338 unsigned short ot_credits;
339 unsigned short ot_id_cnt;
340 unsigned short ot_id_type;
341 uid_t ot_id_array[OSD_MAX_UGID_CNT];
342 struct lquota_trans *ot_quota_trans;
343 #if OSD_THANDLE_STATS
344 /** time when this handle was allocated */
345 cfs_time_t oth_alloced;
347 /** time when this thanle was started */
348 cfs_time_t oth_started;
353 * Basic transaction credit op
363 DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
377 LPROC_OSD_READ_BYTES = 0,
378 LPROC_OSD_WRITE_BYTES = 1,
379 LPROC_OSD_GET_PAGE = 2,
380 LPROC_OSD_NO_PAGE = 3,
381 LPROC_OSD_CACHE_ACCESS = 4,
382 LPROC_OSD_CACHE_HIT = 5,
383 LPROC_OSD_CACHE_MISS = 6,
385 #if OSD_THANDLE_STATS
386 LPROC_OSD_THANDLE_STARTING,
387 LPROC_OSD_THANDLE_OPEN,
388 LPROC_OSD_THANDLE_CLOSING,
395 * Storage representation for fids.
397 * Variable size, first byte contains the length of the whole record.
399 struct osd_fid_pack {
400 unsigned char fp_len;
401 char fp_area[sizeof(struct lu_fid)];
404 struct osd_it_ea_dirent {
405 struct lu_fid oied_fid;
408 unsigned short oied_namelen;
409 unsigned int oied_type;
411 } __attribute__((packed));
414 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
415 * than lu_dirent struct. osd readdir reads less number of dirent than
416 * required for mdd dir page. so buffer size need to be increased so that
417 * there would be one ext3 readdir for every mdd readdir page.
420 #define OSD_IT_EA_BUFSIZE (CFS_PAGE_SIZE + CFS_PAGE_SIZE/4)
423 * This is iterator's in-memory data structure in interoperability
424 * mode (i.e. iterator over ldiskfs style directory)
427 struct osd_object *oie_obj;
428 /** used in ldiskfs iterator, to stored file pointer */
429 struct file oie_file;
430 /** how many entries have been read-cached from storage */
432 /** current entry is being iterated by caller */
434 /** current processing entry */
435 struct osd_it_ea_dirent *oie_dirent;
436 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
441 * Iterator's in-memory data structure for IAM mode.
444 struct osd_object *oi_obj;
445 struct iam_path_descr *oi_ipd;
446 struct iam_iterator oi_it;
449 struct osd_quota_leaf {
455 * Iterator's in-memory data structure for quota file.
457 struct osd_it_quota {
458 struct osd_object *oiq_obj;
459 /** tree blocks path to where the entry is stored */
460 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
461 /** on-disk offset for current key where quota record can be found */
463 /** identifier for current quota record */
465 /** the record index in the leaf/index block */
466 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
467 /** list of already processed leaf blocks */
471 #define MAX_BLOCKS_PER_PAGE (CFS_PAGE_SIZE / 512)
475 cfs_atomic_t dr_numreqs; /* number of reqs being processed */
480 unsigned int dr_ignore_quota:1;
481 unsigned int dr_elapsed_valid:1; /* we really did count time */
482 unsigned int dr_rw:1;
483 struct page *dr_pages[PTLRPC_MAX_BRW_PAGES];
484 unsigned long dr_blocks[PTLRPC_MAX_BRW_PAGES*MAX_BLOCKS_PER_PAGE];
485 unsigned long dr_start_time;
486 unsigned long dr_elapsed; /* how long io took */
487 struct osd_device *dr_dev;
488 unsigned int dr_init_at; /* the line iobuf was initialized */
491 struct osd_thread_info {
492 const struct lu_env *oti_env;
494 * used for index operations.
496 struct dentry oti_obj_dentry;
497 struct dentry oti_child_dentry;
499 /** dentry for Iterator context. */
500 struct dentry oti_it_dentry;
501 struct htree_lock *oti_hlock;
503 struct lu_fid oti_fid;
504 struct lu_fid oti_fid2;
505 struct osd_inode_id oti_id;
506 struct osd_inode_id oti_id2;
507 struct ost_id oti_ostid;
510 * XXX temporary: for ->i_op calls.
512 struct timespec oti_time;
514 * XXX temporary: fake struct file for osd_object_sync
516 struct file oti_file;
518 * XXX temporary: for capa operations.
520 struct lustre_capa_key oti_capa_key;
521 struct lustre_capa oti_capa;
523 /** osd_device reference, initialized in osd_trans_start() and
524 used in osd_trans_stop() */
525 struct osd_device *oti_dev;
528 * following ipd and it structures are used for osd_index_iam_lookup()
529 * these are defined separately as we might do index operation
530 * in open iterator session.
533 /** osd iterator context used for iterator session */
536 struct osd_it_iam oti_it;
537 /* ldiskfs iterator data structure,
538 * see osd_it_ea_{init, fini} */
539 struct osd_it_ea oti_it_ea;
540 struct osd_it_quota oti_it_quota;
543 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
546 cfs_kstatfs_t oti_ksfs;
548 /** IAM iterator for index operation. */
549 struct iam_iterator oti_idx_it;
551 /** union to guarantee that ->oti_ipd[] has proper alignment. */
553 char oti_it_ipd[DX_IPD_MAX_SIZE];
554 long long oti_alignment_lieutenant;
558 char oti_idx_ipd[DX_IPD_MAX_SIZE];
559 long long oti_alignment_lieutenant_colonel;
562 struct osd_idmap_cache oti_cache;
567 /** used in osd_fid_set() to put xattr */
568 struct lu_buf oti_buf;
569 /** used in osd_ea_fid_set() to set fid into common ea */
571 struct lustre_mdt_attrs oti_mdt_attrs;
572 /* old LMA for compatibility */
573 char oti_mdt_attrs_old[LMA_OLD_SIZE];
576 struct osd_iobuf oti_iobuf;
577 struct inode oti_inode;
578 int oti_created[PTLRPC_MAX_BRW_PAGES];
579 struct lu_env oti_obj_delete_tx_env;
580 #define OSD_FID_REC_SZ 32
581 char oti_ldp[OSD_FID_REC_SZ];
582 char oti_ldp2[OSD_FID_REC_SZ];
584 /* used by quota code */
586 struct if_dqblk oti_dqblk;
587 struct if_dqinfo oti_dqinfo;
589 struct lquota_id_info oti_qi;
590 struct lquota_trans oti_quota_trans;
591 union lquota_rec oti_quota_rec;
593 struct lu_seq_range oti_seq_range;
595 #ifdef OSD_TRACK_DECLARES
596 /* Tracking for transaction credits, to allow debugging and optimizing
597 * cases where a large number of credits are being allocated for
598 * single transaction. */
599 unsigned char oti_declare_ops[OSD_OT_MAX];
600 unsigned char oti_declare_ops_rb[OSD_OT_MAX];
601 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
608 extern int ldiskfs_pdo;
612 void lprocfs_osd_init_vars(struct lprocfs_static_vars *lvars);
613 int osd_procfs_init(struct osd_device *osd, const char *name);
614 int osd_procfs_fini(struct osd_device *osd);
615 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
618 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
619 struct obd_statfs *sfs);
620 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
621 struct lustre_capa *capa, __u64 opc);
622 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
623 struct osd_inode_id *id);
624 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
625 const struct lu_fid *fid);
626 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
627 struct dentry *dentry, struct lustre_mdt_attrs *lma);
629 int osd_obj_map_init(struct osd_device *osd);
630 void osd_obj_map_fini(struct osd_device *dev);
631 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
632 const struct lu_fid *fid, struct osd_inode_id *id);
633 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
634 const struct lu_fid *fid, const struct osd_inode_id *id,
636 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
637 const struct lu_fid *fid, struct thandle *th);
638 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
639 const struct lu_fid *fid, struct osd_inode_id *id);
640 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
641 const struct lu_fid *fid, const struct osd_inode_id *id,
644 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
645 int osd_scrub_file_store(struct osd_scrub *scrub);
646 char *osd_lf_fid2name(const struct lu_fid *fid);
647 int osd_scrub_start(struct osd_device *dev);
648 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
649 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
650 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
652 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
653 struct osd_inode_id *id);
654 int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
656 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
657 const struct lu_fid *fid, struct lu_seq_range *range);
658 struct dentry *osd_agent_lookup(struct osd_mdobj_map *omm, int index);
659 struct dentry *osd_agent_load(const struct osd_device *osd, int mdt_index,
662 int osd_delete_agent_inode(const struct lu_env *env, struct osd_device *osd,
663 struct osd_object *obj, struct osd_thandle *oh);
664 int osd_create_agent_inode(const struct lu_env *env, struct osd_device *osd,
665 struct osd_object *obj, struct osd_thandle *oh);
667 /* osd_quota_fmt.c */
668 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
669 int type, uint blk, int depth, uint index,
670 struct osd_it_quota *it);
671 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
672 int type, uint blk, uint index,
673 struct osd_it_quota *it);
674 loff_t find_tree_dqentry(const struct lu_env *env,
675 struct osd_object *obj, int type,
676 qid_t dqid, uint blk, int depth,
677 struct osd_it_quota *it);
679 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
680 struct lquota_id_info *qi, bool allocated, int *flags);
681 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
682 long long space, struct osd_thandle *oh,
683 bool is_blk, bool allocated, int *flags, bool force);
684 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
685 const struct dt_rec *rec,
686 union lquota_rec *quota_rec);
687 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
688 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
689 const struct dt_index_features *feat);
692 struct osd_obj_seq *osd_seq_load(struct osd_device *osd, obd_seq seq);
694 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
696 return (feat == &dt_quota_iusr_features ||
697 feat == &dt_quota_busr_features ||
698 feat == &dt_quota_igrp_features ||
699 feat == &dt_quota_bgrp_features) ? true : false;
703 * Invariants, assertions.
707 * XXX: do not enable this, until invariant checking code is made thread safe
708 * in the face of pdirops locking.
710 #define OSD_INVARIANT_CHECKS (0)
712 #if OSD_INVARIANT_CHECKS
713 static inline int osd_invariant(const struct osd_object *obj)
717 ergo(obj->oo_inode != NULL,
718 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
719 atomic_read(&obj->oo_inode->i_count) > 0) &&
720 ergo(obj->oo_dir != NULL &&
721 obj->oo_dir->od_conationer.ic_object != NULL,
722 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
725 #define osd_invariant(obj) (1)
728 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
730 extern const struct dt_index_operations osd_otable_ops;
732 static inline int osd_oi_fid2idx(struct osd_device *dev,
733 const struct lu_fid *fid)
735 return fid->f_seq & (dev->od_oi_count - 1);
738 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
739 const struct lu_fid *fid)
741 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
742 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
743 LASSERT(osd->od_oi_table != NULL && osd->od_oi_count >= 1);
744 /* It can work even od_oi_count equals to 1 although it's unexpected,
745 * the only reason we set it to 1 is for performance measurement */
746 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
749 extern const struct lu_device_operations osd_lu_ops;
751 static inline int lu_device_is_osd(const struct lu_device *d)
753 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
756 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
758 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
759 return container_of0(d, struct osd_device, od_dt_dev);
762 static inline struct osd_device *osd_dev(const struct lu_device *d)
764 LASSERT(lu_device_is_osd(d));
765 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
768 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
770 return osd_dev(o->oo_dt.do_lu.lo_dev);
773 static inline struct super_block *osd_sb(const struct osd_device *dev)
775 return dev->od_mnt->mnt_sb;
778 static inline int osd_object_is_root(const struct osd_object *obj)
780 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
783 static inline struct osd_object *osd_obj(const struct lu_object *o)
785 LASSERT(lu_device_is_osd(o->lo_dev));
786 return container_of0(o, struct osd_object, oo_dt.do_lu);
789 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
791 return osd_obj(&d->do_lu);
794 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
796 return &osd->od_dt_dev.dd_lu_dev;
799 static inline journal_t *osd_journal(const struct osd_device *dev)
801 return LDISKFS_SB(osd_sb(dev))->s_journal;
804 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
806 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
809 static inline char *osd_name(struct osd_device *osd)
811 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
815 extern const struct dt_body_operations osd_body_ops;
816 extern struct lu_context_key osd_key;
818 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
820 return lu_context_key_get(&env->le_ctx, &osd_key);
823 extern const struct dt_body_operations osd_body_ops_new;
829 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
830 const struct iam_container *bag)
832 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
833 osd_oti_get(env)->oti_it_ipd);
837 struct iam_path_descr *osd_idx_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_idx_ipd);
844 static inline void osd_ipd_put(const struct lu_env *env,
845 const struct iam_container *bag,
846 struct iam_path_descr *ipd)
848 bag->ic_descr->id_ops->id_ipd_free(ipd);
851 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
852 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
853 int write_NUL, loff_t *offs, handle_t *handle);
856 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
858 const char *name, const int namelen)
860 struct osd_thread_info *info = osd_oti_get(env);
861 struct dentry *child_dentry = &info->oti_child_dentry;
862 struct dentry *obj_dentry = &info->oti_obj_dentry;
864 obj_dentry->d_inode = inode;
865 obj_dentry->d_sb = inode->i_sb;
866 obj_dentry->d_name.hash = 0;
868 child_dentry->d_name.hash = 0;
869 child_dentry->d_parent = obj_dentry;
870 child_dentry->d_name.name = name;
871 child_dentry->d_name.len = namelen;
875 #ifdef OSD_TRACK_DECLARES
876 extern int osd_trans_declare_op2rb[];
878 static inline void osd_trans_declare_op(const struct lu_env *env,
879 struct osd_thandle *oh,
880 unsigned int op, int credits)
882 struct osd_thread_info *oti = osd_oti_get(env);
884 LASSERT(oh->ot_handle == NULL);
885 LASSERT(op < OSD_OT_MAX);
887 oti->oti_declare_ops[op]++;
888 oti->oti_declare_ops_cred[op] += credits;
889 oh->ot_credits += credits;
892 static inline void osd_trans_exec_op(const struct lu_env *env,
893 struct thandle *th, unsigned int op)
895 struct osd_thread_info *oti = osd_oti_get(env);
896 struct osd_thandle *oh = container_of(th, struct osd_thandle,
900 LASSERT(oh->ot_handle != NULL);
901 LASSERT(op < OSD_OT_MAX);
903 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
904 oti->oti_declare_ops[op]--;
905 oti->oti_declare_ops_rb[op]++;
907 /* all future updates are considered rollback */
908 oti->oti_rollback = true;
909 rb = osd_trans_declare_op2rb[op];
910 LASSERTF(rb < OSD_OT_MAX, "op = %u\n", op);
911 LASSERTF(oti->oti_declare_ops_rb[rb] > 0, "rb = %u\n", rb);
912 oti->oti_declare_ops_rb[rb]--;
916 static inline void osd_trans_declare_rb(const struct lu_env *env,
917 struct thandle *th, unsigned int op)
919 struct osd_thread_info *oti = osd_oti_get(env);
920 struct osd_thandle *oh = container_of(th, struct osd_thandle,
923 LASSERT(oh->ot_handle != NULL);
924 LASSERT(op < OSD_OT_MAX);
926 oti->oti_declare_ops_rb[op]++;
929 static inline void osd_trans_declare_op(const struct lu_env *env,
930 struct osd_thandle *oh,
931 unsigned int op, int credits)
933 oh->ot_credits += credits;
936 static inline void osd_trans_exec_op(const struct lu_env *env,
937 struct thandle *th, unsigned int op)
941 static inline void osd_trans_declare_rb(const struct lu_env *env,
942 struct thandle *th, unsigned int op)
948 * Helper function to pack the fid, ldiskfs stores fid in packed format.
951 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
952 struct lu_fid *befider)
954 fid_cpu_to_be(befider, (struct lu_fid *)fid);
955 memcpy(pack->fp_area, befider, sizeof(*befider));
956 pack->fp_len = sizeof(*befider) + 1;
960 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
965 switch (pack->fp_len) {
966 case sizeof *fid + 1:
967 memcpy(fid, pack->fp_area, sizeof *fid);
968 fid_be_to_cpu(fid, fid);
971 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
978 * Quota/Accounting handling
980 extern const struct dt_index_operations osd_acct_index_ops;
981 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
982 const struct lu_fid *fid, struct osd_inode_id *id);
984 /* copy from fs/ext4/dir.c */
985 static inline int is_32bit_api(void)
988 return is_compat_task();
990 return (BITS_PER_LONG == 32);
994 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
996 if ((filp->f_mode & FMODE_32BITHASH) ||
997 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
998 return LDISKFS_HTREE_EOF_32BIT;
1000 return LDISKFS_HTREE_EOF_64BIT;
1003 #endif /* __KERNEL__ */
1004 #endif /* _OSD_INTERNAL_H */