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;
172 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
173 ll_ldiskfs_find_entry(dir, dentry, de, lock)
174 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
175 ldiskfs_add_entry(handle, child, cinode, hlock)
177 #define OSD_OTABLE_IT_CACHE_SIZE 128
178 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
180 struct osd_inconsistent_item {
181 /* link into osd_scrub::os_inconsistent_items,
182 * protected by osd_scrub::os_lock. */
185 /* The right FID <=> ino#/gen mapping. */
186 struct osd_idmap_cache oii_cache;
188 unsigned int oii_insert:1; /* insert or update mapping. */
191 struct osd_otable_cache {
192 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
194 /* Index for next cache slot to be filled. */
195 int ooc_producer_idx;
197 /* Index for next cache slot to be returned by it::next(). */
198 int ooc_consumer_idx;
200 /* How many items in ooc_cache. */
201 int ooc_cached_items;
203 /* Position for up layer LFSCK iteration pre-loading. */
204 __u32 ooc_pos_preload;
207 struct osd_otable_it {
208 struct osd_device *ooi_dev;
209 struct osd_otable_cache ooi_cache;
211 /* For osd_otable_it_key. */
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. */
226 extern const int osd_dto_credits_noquota[];
233 struct dt_device od_dt_dev;
234 /* information about underlying file system */
235 struct vfsmount *od_mnt;
237 struct osd_oi **od_oi_table;
238 /* total number of OI containers */
243 unsigned int od_fl_capa:1,
244 od_is_md:1; /* set in ->ldo_prepare */
245 unsigned long od_capa_timeout;
247 struct lustre_capa_key *od_capa_keys;
248 cfs_hlist_head_t *od_capa_hash;
250 cfs_proc_dir_entry_t *od_proc_entry;
251 struct lprocfs_stats *od_stats;
253 * statfs optimization: we cache a bit.
255 cfs_time_t od_osfs_age;
256 struct obd_statfs od_statfs;
257 spinlock_t od_osfs_lock;
259 unsigned int od_noscrub:1,
262 struct fsfilt_operations *od_fsops;
264 struct lu_site od_site;
266 struct osd_obj_map *od_ost_map;
268 unsigned long long od_readcache_max_filesize;
270 int od_writethrough_cache;
272 struct brw_stats od_brw_stats;
273 cfs_atomic_t od_r_in_flight;
274 cfs_atomic_t od_w_in_flight;
276 struct mutex od_otable_mutex;
277 struct osd_otable_it *od_otable_it;
278 struct osd_scrub od_scrub;
280 /* service name associated with the osd device */
281 char od_svname[MAX_OBD_NAME];
282 char od_mntdev[MAX_OBD_NAME];
284 /* quota slave instance */
285 struct qsd_instance *od_quota_slave;
288 /* There are at most 10 uid/gids are affected in a transaction, and
289 * that's rename case:
290 * - 2 for source parent uid & gid;
291 * - 2 for source child uid & gid ('..' entry update when child is directory);
292 * - 2 for target parent uid & gid;
293 * - 2 for target child uid & gid (if the target child exists);
294 * - 2 for root uid & gid (last_rcvd, llog, etc);
296 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
297 * the id type of each id in the ot_id_array.
299 #define OSD_MAX_UGID_CNT 10
304 OSD_OT_XATTR_SET = 2,
316 #define OSD_TRACK_DECLARES
319 struct thandle ot_super;
321 struct journal_callback ot_jcb;
322 cfs_list_t ot_dcb_list;
323 /* Link to the device, for debugging. */
324 struct lu_ref_link *ot_dev_link;
325 unsigned short ot_credits;
326 unsigned short ot_id_cnt;
327 unsigned short ot_id_type;
328 uid_t ot_id_array[OSD_MAX_UGID_CNT];
329 struct lquota_trans *ot_quota_trans;
330 #if OSD_THANDLE_STATS
331 /** time when this handle was allocated */
332 cfs_time_t oth_alloced;
334 /** time when this thanle was started */
335 cfs_time_t oth_started;
340 * Basic transaction credit op
350 DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
364 LPROC_OSD_READ_BYTES = 0,
365 LPROC_OSD_WRITE_BYTES = 1,
366 LPROC_OSD_GET_PAGE = 2,
367 LPROC_OSD_NO_PAGE = 3,
368 LPROC_OSD_CACHE_ACCESS = 4,
369 LPROC_OSD_CACHE_HIT = 5,
370 LPROC_OSD_CACHE_MISS = 6,
372 #if OSD_THANDLE_STATS
373 LPROC_OSD_THANDLE_STARTING,
374 LPROC_OSD_THANDLE_OPEN,
375 LPROC_OSD_THANDLE_CLOSING,
382 * Storage representation for fids.
384 * Variable size, first byte contains the length of the whole record.
386 struct osd_fid_pack {
387 unsigned char fp_len;
388 char fp_area[sizeof(struct lu_fid)];
391 struct osd_it_ea_dirent {
392 struct lu_fid oied_fid;
395 unsigned short oied_namelen;
396 unsigned int oied_type;
398 } __attribute__((packed));
401 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
402 * than lu_dirent struct. osd readdir reads less number of dirent than
403 * required for mdd dir page. so buffer size need to be increased so that
404 * there would be one ext3 readdir for every mdd readdir page.
407 #define OSD_IT_EA_BUFSIZE (CFS_PAGE_SIZE + CFS_PAGE_SIZE/4)
410 * This is iterator's in-memory data structure in interoperability
411 * mode (i.e. iterator over ldiskfs style directory)
414 struct osd_object *oie_obj;
415 /** used in ldiskfs iterator, to stored file pointer */
416 struct file oie_file;
417 /** how many entries have been read-cached from storage */
419 /** current entry is being iterated by caller */
421 /** current processing entry */
422 struct osd_it_ea_dirent *oie_dirent;
423 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
428 * Iterator's in-memory data structure for IAM mode.
431 struct osd_object *oi_obj;
432 struct iam_path_descr *oi_ipd;
433 struct iam_iterator oi_it;
436 struct osd_quota_leaf {
442 * Iterator's in-memory data structure for quota file.
444 struct osd_it_quota {
445 struct osd_object *oiq_obj;
446 /** tree blocks path to where the entry is stored */
447 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
448 /** on-disk offset for current key where quota record can be found */
450 /** identifier for current quota record */
452 /** the record index in the leaf/index block */
453 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
454 /** list of already processed leaf blocks */
458 #define MAX_BLOCKS_PER_PAGE (CFS_PAGE_SIZE / 512)
462 cfs_atomic_t dr_numreqs; /* number of reqs being processed */
467 unsigned int dr_ignore_quota:1;
468 unsigned int dr_elapsed_valid:1; /* we really did count time */
469 unsigned int dr_rw:1;
470 struct page *dr_pages[PTLRPC_MAX_BRW_PAGES];
471 unsigned long dr_blocks[PTLRPC_MAX_BRW_PAGES*MAX_BLOCKS_PER_PAGE];
472 unsigned long dr_start_time;
473 unsigned long dr_elapsed; /* how long io took */
474 struct osd_device *dr_dev;
475 unsigned int dr_init_at; /* the line iobuf was initialized */
478 struct osd_thread_info {
479 const struct lu_env *oti_env;
481 * used for index operations.
483 struct dentry oti_obj_dentry;
484 struct dentry oti_child_dentry;
486 /** dentry for Iterator context. */
487 struct dentry oti_it_dentry;
488 struct htree_lock *oti_hlock;
490 struct lu_fid oti_fid;
491 struct lu_fid oti_fid2;
492 struct osd_inode_id oti_id;
493 struct osd_inode_id oti_id2;
494 struct ost_id oti_ostid;
497 * XXX temporary: for ->i_op calls.
499 struct timespec oti_time;
501 * XXX temporary: fake struct file for osd_object_sync
503 struct file oti_file;
505 * XXX temporary: for capa operations.
507 struct lustre_capa_key oti_capa_key;
508 struct lustre_capa oti_capa;
510 /** osd_device reference, initialized in osd_trans_start() and
511 used in osd_trans_stop() */
512 struct osd_device *oti_dev;
515 * following ipd and it structures are used for osd_index_iam_lookup()
516 * these are defined separately as we might do index operation
517 * in open iterator session.
520 /** osd iterator context used for iterator session */
523 struct osd_it_iam oti_it;
524 /* ldiskfs iterator data structure,
525 * see osd_it_ea_{init, fini} */
526 struct osd_it_ea oti_it_ea;
527 struct osd_it_quota oti_it_quota;
530 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
533 cfs_kstatfs_t oti_ksfs;
535 /** IAM iterator for index operation. */
536 struct iam_iterator oti_idx_it;
538 /** union to guarantee that ->oti_ipd[] has proper alignment. */
540 char oti_it_ipd[DX_IPD_MAX_SIZE];
541 long long oti_alignment_lieutenant;
545 char oti_idx_ipd[DX_IPD_MAX_SIZE];
546 long long oti_alignment_lieutenant_colonel;
549 struct osd_idmap_cache oti_cache;
554 /** used in osd_fid_set() to put xattr */
555 struct lu_buf oti_buf;
556 /** used in osd_ea_fid_set() to set fid into common ea */
558 struct lustre_mdt_attrs oti_mdt_attrs;
559 /* old LMA for compatibility */
560 char oti_mdt_attrs_old[LMA_OLD_SIZE];
563 struct osd_iobuf oti_iobuf;
564 struct inode oti_inode;
565 int oti_created[PTLRPC_MAX_BRW_PAGES];
566 struct lu_env oti_obj_delete_tx_env;
567 #define OSD_FID_REC_SZ 32
568 char oti_ldp[OSD_FID_REC_SZ];
569 char oti_ldp2[OSD_FID_REC_SZ];
571 /* used by quota code */
573 struct if_dqblk oti_dqblk;
574 struct if_dqinfo oti_dqinfo;
576 struct lquota_id_info oti_qi;
577 struct lquota_trans oti_quota_trans;
578 union lquota_rec oti_quota_rec;
580 struct lu_seq_range oti_seq_range;
582 #ifdef OSD_TRACK_DECLARES
583 /* Tracking for transaction credits, to allow debugging and optimizing
584 * cases where a large number of credits are being allocated for
585 * single transaction. */
586 unsigned char oti_declare_ops[OSD_OT_MAX];
587 unsigned char oti_declare_ops_rb[OSD_OT_MAX];
588 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
594 extern int ldiskfs_pdo;
598 void lprocfs_osd_init_vars(struct lprocfs_static_vars *lvars);
599 int osd_procfs_init(struct osd_device *osd, const char *name);
600 int osd_procfs_fini(struct osd_device *osd);
601 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
604 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
605 struct obd_statfs *sfs);
606 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
607 struct lustre_capa *capa, __u64 opc);
608 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
609 struct osd_inode_id *id);
610 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
611 struct dentry *dentry, struct lustre_mdt_attrs *lma);
613 int osd_obj_map_init(struct osd_device *osd);
614 void osd_obj_map_fini(struct osd_device *dev);
615 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
616 const struct lu_fid *fid, struct osd_inode_id *id);
617 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
618 const struct lu_fid *fid, const struct osd_inode_id *id,
620 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
621 const struct lu_fid *fid, struct thandle *th);
622 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
623 const struct lu_fid *fid, struct osd_inode_id *id);
624 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
625 const struct lu_fid *fid, const struct osd_inode_id *id,
628 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
629 int osd_scrub_file_store(struct osd_scrub *scrub);
630 int osd_scrub_start(struct osd_device *dev);
631 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
632 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
633 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
635 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
636 struct osd_inode_id *id);
637 int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
639 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
640 const struct lu_fid *fid, struct lu_seq_range *range);
641 /* osd_quota_fmt.c */
642 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
643 int type, uint blk, int depth, uint index,
644 struct osd_it_quota *it);
645 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
646 int type, uint blk, uint index,
647 struct osd_it_quota *it);
648 loff_t find_tree_dqentry(const struct lu_env *env,
649 struct osd_object *obj, int type,
650 qid_t dqid, uint blk, int depth,
651 struct osd_it_quota *it);
653 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
654 struct lquota_id_info *qi, bool allocated, int *flags);
655 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
656 long long space, struct osd_thandle *oh,
657 bool is_blk, bool allocated, int *flags, bool force);
658 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
659 const struct dt_rec *rec,
660 union lquota_rec *quota_rec);
661 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
662 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
663 const struct dt_index_features *feat);
665 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
667 return (feat == &dt_quota_iusr_features ||
668 feat == &dt_quota_busr_features ||
669 feat == &dt_quota_igrp_features ||
670 feat == &dt_quota_bgrp_features) ? true : false;
674 * Invariants, assertions.
678 * XXX: do not enable this, until invariant checking code is made thread safe
679 * in the face of pdirops locking.
681 #define OSD_INVARIANT_CHECKS (0)
683 #if OSD_INVARIANT_CHECKS
684 static inline int osd_invariant(const struct osd_object *obj)
688 ergo(obj->oo_inode != NULL,
689 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
690 atomic_read(&obj->oo_inode->i_count) > 0) &&
691 ergo(obj->oo_dir != NULL &&
692 obj->oo_dir->od_conationer.ic_object != NULL,
693 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
696 #define osd_invariant(obj) (1)
699 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
701 extern const struct dt_index_operations osd_otable_ops;
703 static inline int osd_oi_fid2idx(struct osd_device *dev,
704 const struct lu_fid *fid)
706 return fid->f_seq & (dev->od_oi_count - 1);
709 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
710 const struct lu_fid *fid)
712 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
713 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
714 LASSERT(osd->od_oi_table != NULL && osd->od_oi_count >= 1);
715 /* It can work even od_oi_count equals to 1 although it's unexpected,
716 * the only reason we set it to 1 is for performance measurement */
717 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
720 extern const struct lu_device_operations osd_lu_ops;
722 static inline int lu_device_is_osd(const struct lu_device *d)
724 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
727 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
729 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
730 return container_of0(d, struct osd_device, od_dt_dev);
733 static inline struct osd_device *osd_dev(const struct lu_device *d)
735 LASSERT(lu_device_is_osd(d));
736 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
739 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
741 return osd_dev(o->oo_dt.do_lu.lo_dev);
744 static inline struct super_block *osd_sb(const struct osd_device *dev)
746 return dev->od_mnt->mnt_sb;
749 static inline int osd_object_is_root(const struct osd_object *obj)
751 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
754 static inline struct osd_object *osd_obj(const struct lu_object *o)
756 LASSERT(lu_device_is_osd(o->lo_dev));
757 return container_of0(o, struct osd_object, oo_dt.do_lu);
760 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
762 return osd_obj(&d->do_lu);
765 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
767 return &osd->od_dt_dev.dd_lu_dev;
770 static inline journal_t *osd_journal(const struct osd_device *dev)
772 return LDISKFS_SB(osd_sb(dev))->s_journal;
775 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
777 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
780 static inline char *osd_name(struct osd_device *osd)
782 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
786 extern const struct dt_body_operations osd_body_ops;
787 extern struct lu_context_key osd_key;
789 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
791 return lu_context_key_get(&env->le_ctx, &osd_key);
794 extern const struct dt_body_operations osd_body_ops_new;
800 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
801 const struct iam_container *bag)
803 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
804 osd_oti_get(env)->oti_it_ipd);
808 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
809 const struct iam_container *bag)
811 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
812 osd_oti_get(env)->oti_idx_ipd);
815 static inline void osd_ipd_put(const struct lu_env *env,
816 const struct iam_container *bag,
817 struct iam_path_descr *ipd)
819 bag->ic_descr->id_ops->id_ipd_free(ipd);
822 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
823 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
824 int write_NUL, loff_t *offs, handle_t *handle);
827 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
829 const char *name, const int namelen)
831 struct osd_thread_info *info = osd_oti_get(env);
832 struct dentry *child_dentry = &info->oti_child_dentry;
833 struct dentry *obj_dentry = &info->oti_obj_dentry;
835 obj_dentry->d_inode = inode;
836 obj_dentry->d_sb = inode->i_sb;
837 obj_dentry->d_name.hash = 0;
839 child_dentry->d_name.hash = 0;
840 child_dentry->d_parent = obj_dentry;
841 child_dentry->d_name.name = name;
842 child_dentry->d_name.len = namelen;
846 #ifdef OSD_TRACK_DECLARES
847 extern int osd_trans_declare_op2rb[];
849 static inline void osd_trans_declare_op(const struct lu_env *env,
850 struct osd_thandle *oh,
851 unsigned int op, int credits)
853 struct osd_thread_info *oti = osd_oti_get(env);
855 LASSERT(oh->ot_handle == NULL);
856 LASSERT(op < OSD_OT_MAX);
858 oti->oti_declare_ops[op]++;
859 oti->oti_declare_ops_cred[op] += credits;
860 oh->ot_credits += credits;
863 static inline void osd_trans_exec_op(const struct lu_env *env,
864 struct thandle *th, unsigned int op)
866 struct osd_thread_info *oti = osd_oti_get(env);
867 struct osd_thandle *oh = container_of(th, struct osd_thandle,
871 LASSERT(oh->ot_handle != NULL);
872 LASSERT(op < OSD_OT_MAX);
874 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
875 oti->oti_declare_ops[op]--;
876 oti->oti_declare_ops_rb[op]++;
878 /* all future updates are considered rollback */
879 oti->oti_rollback = true;
880 rb = osd_trans_declare_op2rb[op];
881 LASSERTF(rb < OSD_OT_MAX, "op = %u\n", op);
882 LASSERTF(oti->oti_declare_ops_rb[rb] > 0, "rb = %u\n", rb);
883 oti->oti_declare_ops_rb[rb]--;
887 static inline void osd_trans_declare_rb(const struct lu_env *env,
888 struct thandle *th, unsigned int op)
890 struct osd_thread_info *oti = osd_oti_get(env);
891 struct osd_thandle *oh = container_of(th, struct osd_thandle,
894 LASSERT(oh->ot_handle != NULL);
895 LASSERT(op < OSD_OT_MAX);
897 oti->oti_declare_ops_rb[op]++;
900 static inline void osd_trans_declare_op(const struct lu_env *env,
901 struct osd_thandle *oh,
902 unsigned int op, int credits)
904 oh->ot_credits += credits;
907 static inline void osd_trans_exec_op(const struct lu_env *env,
908 struct thandle *th, unsigned int op)
912 static inline void osd_trans_declare_rb(const struct lu_env *env,
913 struct thandle *th, unsigned int op)
919 * Helper function to pack the fid, ldiskfs stores fid in packed format.
922 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
923 struct lu_fid *befider)
925 fid_cpu_to_be(befider, (struct lu_fid *)fid);
926 memcpy(pack->fp_area, befider, sizeof(*befider));
927 pack->fp_len = sizeof(*befider) + 1;
931 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
936 switch (pack->fp_len) {
937 case sizeof *fid + 1:
938 memcpy(fid, pack->fp_area, sizeof *fid);
939 fid_be_to_cpu(fid, fid);
942 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
949 * Quota/Accounting handling
951 extern const struct dt_index_operations osd_acct_index_ops;
952 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
953 const struct lu_fid *fid, struct osd_inode_id *id);
955 /* copy from fs/ext4/dir.c */
956 static inline int is_32bit_api(void)
959 return is_compat_task();
961 return (BITS_PER_LONG == 32);
965 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
967 if ((filp->f_mode & FMODE_32BITHASH) ||
968 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
969 return LDISKFS_HTREE_EOF_32BIT;
971 return LDISKFS_HTREE_EOF_64BIT;
974 #endif /* __KERNEL__ */
975 #endif /* _OSD_INTERNAL_H */