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, 2013, 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>
60 /* class_register_type(), class_unregister_type(), class_get_type() */
61 #include <obd_class.h>
62 #include <lustre_disk.h>
63 #include <dt_object.h>
64 #include <lustre_quota.h>
68 #include "osd_scrub.h"
69 #include "osd_quota_fmt.h"
73 #define OSD_COUNTERS (0)
75 /* ldiskfs special inode::i_state_flags need to be accessed with
76 * ldiskfs_{set,clear,test}_inode_state() only */
78 /* OI scrub should skip this inode. */
79 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
81 /** Enable thandle usage statistics */
82 #define OSD_THANDLE_STATS (0)
84 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
86 #define OBJECTS "OBJECTS"
87 #define ADMIN_USR "admin_quotafile_v2.usr"
88 #define ADMIN_GRP "admin_quotafile_v2.grp"
90 struct osd_directory {
91 struct iam_container od_container;
92 struct iam_descr od_descr;
96 * Object Index (oi) instance.
100 * underlying index object, where fid->id mapping in stored.
102 struct inode *oi_inode;
103 struct osd_directory oi_dir;
106 extern const int osd_dto_credits_noquota[];
109 struct dt_object oo_dt;
111 * Inode for file system object represented by this osd_object. This
112 * inode is pinned for the whole duration of lu_object life.
114 * Not modified concurrently (either setup early during object
115 * creation, or assigned by osd_object_create() under write lock).
117 struct inode *oo_inode;
119 * to protect index ops.
121 struct htree_lock_head *oo_hl_head;
122 struct rw_semaphore oo_ext_idx_sem;
123 struct rw_semaphore oo_sem;
124 struct osd_directory *oo_dir;
125 /** protects inode attributes. */
128 * Following two members are used to indicate the presence of dot and
129 * dotdot in the given directory. This is required for interop mode
132 int oo_compat_dot_created;
133 int oo_compat_dotdot_created;
135 const struct lu_env *oo_owner;
136 #ifdef CONFIG_LOCKDEP
137 struct lockdep_map oo_dep_map;
142 /* protects on-fly initialization */
143 int oos_subdir_count; /* subdir count for each seq */
144 struct dentry *oos_root; /* O/<seq> */
145 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
146 obd_seq oos_seq; /* seq number */
147 cfs_list_t oos_seq_list; /* list to seq_list */
151 struct dentry *om_root; /* dentry for /O */
152 rwlock_t om_seq_list_lock; /* lock for seq_list */
153 cfs_list_t om_seq_list; /* list head for seq */
155 struct semaphore om_dir_init_sem;
159 struct dentry *om_root; /* AGENT/<index> */
160 obd_seq om_index; /* mdt index */
161 cfs_list_t om_list; /* list to omm_list */
164 struct osd_mdobj_map {
165 struct dentry *omm_remote_parent;
168 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
169 ldiskfs_add_entry(handle, child, cinode, hlock)
171 #define OSD_OTABLE_IT_CACHE_SIZE 64
172 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
174 struct osd_inconsistent_item {
175 /* link into osd_scrub::os_inconsistent_items,
176 * protected by osd_scrub::os_lock. */
179 /* The right FID <=> ino#/gen mapping. */
180 struct osd_idmap_cache oii_cache;
182 unsigned int oii_insert:1; /* insert or update mapping. */
185 struct osd_otable_cache {
186 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
188 /* Index for next cache slot to be filled. */
189 int ooc_producer_idx;
191 /* Index for next cache slot to be returned by it::next(). */
192 int ooc_consumer_idx;
194 /* How many items in ooc_cache. */
195 int ooc_cached_items;
197 /* Position for up layer LFSCK iteration pre-loading. */
198 __u32 ooc_pos_preload;
201 struct osd_otable_it {
202 struct osd_device *ooi_dev;
203 struct osd_otable_cache ooi_cache;
205 /* The following bits can be updated/checked w/o lock protection.
206 * If more bits will be introduced in the future and need lock to
207 * protect, please add comment. */
208 unsigned long ooi_used_outside:1, /* Some user out of OSD
209 * uses the iteration. */
210 ooi_all_cached:1, /* No more entries can be
211 * filled into cache. */
212 ooi_user_ready:1, /* The user out of OSD is
213 * ready to iterate. */
214 ooi_waiting:1; /* it::next is waiting. */
222 struct dt_device od_dt_dev;
223 /* information about underlying file system */
224 struct vfsmount *od_mnt;
226 struct osd_oi **od_oi_table;
227 /* total number of OI containers */
232 unsigned int od_fl_capa:1,
239 od_lma_self_repair:1;
241 unsigned long od_capa_timeout;
243 struct lustre_capa_key *od_capa_keys;
244 cfs_hlist_head_t *od_capa_hash;
246 cfs_proc_dir_entry_t *od_proc_entry;
247 struct lprocfs_stats *od_stats;
249 spinlock_t od_osfs_lock;
252 struct lu_site od_site;
254 struct osd_obj_map *od_ost_map;
255 struct osd_mdobj_map *od_mdt_map;
257 unsigned long long od_readcache_max_filesize;
259 int od_writethrough_cache;
261 struct brw_stats od_brw_stats;
262 atomic_t od_r_in_flight;
263 atomic_t od_w_in_flight;
265 struct mutex od_otable_mutex;
266 struct osd_otable_it *od_otable_it;
267 struct osd_scrub od_scrub;
268 cfs_list_t od_ios_list;
270 /* service name associated with the osd device */
271 char od_svname[MAX_OBD_NAME];
272 char od_mntdev[MAX_OBD_NAME];
274 /* quota slave instance */
275 struct qsd_instance *od_quota_slave;
278 /* There are at most 10 uid/gids are affected in a transaction, and
279 * that's rename case:
280 * - 2 for source parent uid & gid;
281 * - 2 for source child uid & gid ('..' entry update when child is directory);
282 * - 2 for target parent uid & gid;
283 * - 2 for target child uid & gid (if the target child exists);
284 * - 2 for root uid & gid (last_rcvd, llog, etc);
286 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
287 * the id type of each id in the ot_id_array.
289 #define OSD_MAX_UGID_CNT 10
294 OSD_OT_XATTR_SET = 2,
308 struct thandle ot_super;
310 struct ldiskfs_journal_cb_entry ot_jcb;
311 cfs_list_t ot_dcb_list;
312 /* Link to the device, for debugging. */
313 struct lu_ref_link ot_dev_link;
314 unsigned short ot_credits;
315 unsigned short ot_id_cnt;
316 unsigned short ot_id_type;
317 uid_t ot_id_array[OSD_MAX_UGID_CNT];
318 struct lquota_trans *ot_quota_trans;
319 #if OSD_THANDLE_STATS
320 /** time when this handle was allocated */
321 cfs_time_t oth_alloced;
323 /** time when this thanle was started */
324 cfs_time_t oth_started;
329 * Basic transaction credit op
339 DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
353 LPROC_OSD_READ_BYTES = 0,
354 LPROC_OSD_WRITE_BYTES = 1,
355 LPROC_OSD_GET_PAGE = 2,
356 LPROC_OSD_NO_PAGE = 3,
357 LPROC_OSD_CACHE_ACCESS = 4,
358 LPROC_OSD_CACHE_HIT = 5,
359 LPROC_OSD_CACHE_MISS = 6,
361 #if OSD_THANDLE_STATS
362 LPROC_OSD_THANDLE_STARTING,
363 LPROC_OSD_THANDLE_OPEN,
364 LPROC_OSD_THANDLE_CLOSING,
371 * Storage representation for fids.
373 * Variable size, first byte contains the length of the whole record.
375 struct osd_fid_pack {
376 unsigned char fp_len;
377 char fp_area[sizeof(struct lu_fid)];
380 struct osd_it_ea_dirent {
381 struct lu_fid oied_fid;
384 unsigned short oied_namelen;
385 unsigned int oied_type;
387 } __attribute__((packed));
390 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
391 * than lu_dirent struct. osd readdir reads less number of dirent than
392 * required for mdd dir page. so buffer size need to be increased so that
393 * there would be one ext3 readdir for every mdd readdir page.
396 #define OSD_IT_EA_BUFSIZE (PAGE_CACHE_SIZE + PAGE_CACHE_SIZE/4)
399 * This is iterator's in-memory data structure in interoperability
400 * mode (i.e. iterator over ldiskfs style directory)
403 struct osd_object *oie_obj;
404 /** used in ldiskfs iterator, to stored file pointer */
405 struct file oie_file;
406 /** how many entries have been read-cached from storage */
408 /** current entry is being iterated by caller */
410 /** current processing entry */
411 struct osd_it_ea_dirent *oie_dirent;
412 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
417 * Iterator's in-memory data structure for IAM mode.
420 struct osd_object *oi_obj;
421 struct iam_path_descr *oi_ipd;
422 struct iam_iterator oi_it;
425 struct osd_quota_leaf {
431 * Iterator's in-memory data structure for quota file.
433 struct osd_it_quota {
434 struct osd_object *oiq_obj;
435 /** tree blocks path to where the entry is stored */
436 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
437 /** on-disk offset for current key where quota record can be found */
439 /** identifier for current quota record */
441 /** the record index in the leaf/index block */
442 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
443 /** list of already processed leaf blocks */
447 #define MAX_BLOCKS_PER_PAGE (PAGE_CACHE_SIZE / 512)
450 wait_queue_head_t dr_wait;
451 atomic_t dr_numreqs; /* number of reqs being processed */
456 unsigned int dr_ignore_quota:1;
457 unsigned int dr_elapsed_valid:1; /* we really did count time */
458 unsigned int dr_rw:1;
459 struct lu_buf dr_pg_buf;
460 struct page **dr_pages;
461 struct lu_buf dr_bl_buf;
462 unsigned long *dr_blocks;
463 unsigned long dr_start_time;
464 unsigned long dr_elapsed; /* how long io took */
465 struct osd_device *dr_dev;
466 unsigned int dr_init_at; /* the line iobuf was initialized */
469 struct osd_thread_info {
470 const struct lu_env *oti_env;
472 * used for index operations.
474 struct dentry oti_obj_dentry;
475 struct dentry oti_child_dentry;
477 /** dentry for Iterator context. */
478 struct dentry oti_it_dentry;
479 struct htree_lock *oti_hlock;
481 struct lu_fid oti_fid;
482 struct lu_fid oti_fid2;
483 struct lu_fid oti_fid3;
484 struct osd_inode_id oti_id;
485 struct osd_inode_id oti_id2;
486 struct osd_inode_id oti_id3;
487 struct ost_id oti_ostid;
490 * XXX temporary: for ->i_op calls.
492 struct timespec oti_time;
494 * XXX temporary: fake struct file for osd_object_sync
496 struct file oti_file;
498 * XXX temporary: for capa operations.
500 struct lustre_capa_key oti_capa_key;
501 struct lustre_capa oti_capa;
503 /** osd_device reference, initialized in osd_trans_start() and
504 used in osd_trans_stop() */
505 struct osd_device *oti_dev;
508 * following ipd and it structures are used for osd_index_iam_lookup()
509 * these are defined separately as we might do index operation
510 * in open iterator session.
513 /** osd iterator context used for iterator session */
516 struct osd_it_iam oti_it;
517 /* ldiskfs iterator data structure,
518 * see osd_it_ea_{init, fini} */
519 struct osd_it_ea oti_it_ea;
520 struct osd_it_quota oti_it_quota;
523 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
526 struct kstatfs oti_ksfs;
528 /** IAM iterator for index operation. */
529 struct iam_iterator oti_idx_it;
531 /** union to guarantee that ->oti_ipd[] has proper alignment. */
533 char oti_it_ipd[DX_IPD_MAX_SIZE];
534 long long oti_alignment_lieutenant;
538 char oti_idx_ipd[DX_IPD_MAX_SIZE];
539 long long oti_alignment_lieutenant_colonel;
542 struct osd_idmap_cache oti_cache;
547 /** used in osd_fid_set() to put xattr */
548 struct lu_buf oti_buf;
549 /** used in osd_ea_fid_set() to set fid into common ea */
551 struct lustre_mdt_attrs oti_mdt_attrs;
552 /* old LMA for compatibility */
553 char oti_mdt_attrs_old[LMA_OLD_SIZE];
556 struct osd_iobuf oti_iobuf;
557 struct inode oti_inode;
558 #define OSD_FID_REC_SZ 32
559 char oti_ldp[OSD_FID_REC_SZ];
560 char oti_ldp2[OSD_FID_REC_SZ];
562 /* used by quota code */
564 #ifdef HAVE_DQUOT_FS_DISK_QUOTA
565 struct fs_disk_quota oti_fdq;
567 struct if_dqblk oti_dqblk;
569 struct if_dqinfo oti_dqinfo;
571 struct lquota_id_info oti_qi;
572 struct lquota_trans oti_quota_trans;
573 union lquota_rec oti_quota_rec;
575 struct lu_seq_range oti_seq_range;
577 /* Tracking for transaction credits, to allow debugging and optimizing
578 * cases where a large number of credits are being allocated for
579 * single transaction. */
580 unsigned short oti_declare_ops[OSD_OT_MAX];
581 unsigned short oti_declare_ops_rb[OSD_OT_MAX];
582 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
587 struct filter_fid_old oti_ff;
588 struct filter_fid oti_ff_new;
592 extern int ldiskfs_pdo;
594 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
595 const char *name, void *buf, int len)
600 dentry->d_inode = inode;
601 dentry->d_sb = inode->i_sb;
602 return inode->i_op->getxattr(dentry, name, buf, len);
605 static inline int __osd_xattr_set(struct osd_thread_info *info,
606 struct inode *inode, const char *name,
607 const void *buf, int buflen, int fl)
609 struct dentry *dentry = &info->oti_child_dentry;
611 ll_vfs_dq_init(inode);
612 dentry->d_inode = inode;
613 dentry->d_sb = inode->i_sb;
614 return inode->i_op->setxattr(dentry, name, buf, buflen, fl);
619 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
620 extern struct lprocfs_vars lprocfs_osd_module_vars[];
621 int osd_procfs_init(struct osd_device *osd, const char *name);
622 int osd_procfs_fini(struct osd_device *osd);
623 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
626 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
627 struct obd_statfs *sfs);
628 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
629 struct lustre_capa *capa, __u64 opc);
630 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
631 struct osd_inode_id *id);
632 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
633 const struct lu_fid *fid, __u32 compat, __u32 incompat);
634 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
635 struct dentry *dentry, struct lustre_mdt_attrs *lma);
636 int osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
637 struct osd_inode_id *id, const struct lu_fid *fid);
638 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
639 struct dentry *dentry, struct lu_fid *fid);
641 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
642 void osd_obj_map_fini(struct osd_device *dev);
643 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
644 const struct lu_fid *fid, struct osd_inode_id *id);
645 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
646 const struct lu_fid *fid, const struct osd_inode_id *id,
648 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
649 const struct lu_fid *fid, handle_t *th);
650 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
651 const struct lu_fid *fid, const struct osd_inode_id *id,
653 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
654 struct inode *src_parent, struct dentry *src_child,
655 const struct lu_fid *fid);
656 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
657 const struct lu_fid *fid, struct osd_inode_id *id);
658 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
659 const struct lu_fid *fid, const struct osd_inode_id *id,
661 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
662 const struct lu_fid *fid, const struct osd_inode_id *id,
665 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
666 int osd_scrub_file_store(struct osd_scrub *scrub);
667 char *osd_lf_fid2name(const struct lu_fid *fid);
668 int osd_scrub_start(struct osd_device *dev);
669 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
670 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
671 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
673 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
674 struct osd_inode_id *id);
675 int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
677 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
678 obd_seq seq, struct lu_seq_range *range);
680 int osd_delete_from_remote_parent(const struct lu_env *env,
681 struct osd_device *osd,
682 struct osd_object *obj,
683 struct osd_thandle *oh);
684 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
685 struct osd_object *obj, struct osd_thandle *oh);
686 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
687 struct osd_device *osd,
688 const struct lu_fid *fid,
689 struct osd_inode_id *id);
691 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
693 /* osd_quota_fmt.c */
694 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
695 int type, uint blk, int depth, uint index,
696 struct osd_it_quota *it);
697 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
698 int type, uint blk, uint index,
699 struct osd_it_quota *it);
700 loff_t find_tree_dqentry(const struct lu_env *env,
701 struct osd_object *obj, int type,
702 qid_t dqid, uint blk, int depth,
703 struct osd_it_quota *it);
705 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
706 struct lquota_id_info *qi, bool allocated, int *flags);
707 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
708 long long space, struct osd_thandle *oh,
709 bool is_blk, bool allocated, int *flags, bool force);
710 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
711 const struct dt_rec *rec,
712 union lquota_rec *quota_rec);
713 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
714 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
715 const struct dt_index_features *feat);
717 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
719 return (feat == &dt_quota_iusr_features ||
720 feat == &dt_quota_busr_features ||
721 feat == &dt_quota_igrp_features ||
722 feat == &dt_quota_bgrp_features) ? true : false;
726 * Invariants, assertions.
730 * XXX: do not enable this, until invariant checking code is made thread safe
731 * in the face of pdirops locking.
733 #define OSD_INVARIANT_CHECKS (0)
735 #if OSD_INVARIANT_CHECKS
736 static inline int osd_invariant(const struct osd_object *obj)
740 ergo(obj->oo_inode != NULL,
741 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
742 atomic_read(&obj->oo_inode->i_count) > 0) &&
743 ergo(obj->oo_dir != NULL &&
744 obj->oo_dir->od_conationer.ic_object != NULL,
745 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
748 #define osd_invariant(obj) (1)
751 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
753 extern const struct dt_index_operations osd_otable_ops;
755 static inline int osd_oi_fid2idx(struct osd_device *dev,
756 const struct lu_fid *fid)
758 return fid->f_seq & (dev->od_oi_count - 1);
761 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
762 const struct lu_fid *fid)
764 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
765 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
766 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
767 DFID"\n", PFID(fid));
768 /* It can work even od_oi_count equals to 1 although it's unexpected,
769 * the only reason we set it to 1 is for performance measurement */
770 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
773 extern const struct lu_device_operations osd_lu_ops;
775 static inline int lu_device_is_osd(const struct lu_device *d)
777 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
780 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
782 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
783 return container_of0(d, struct osd_device, od_dt_dev);
786 static inline struct osd_device *osd_dev(const struct lu_device *d)
788 LASSERT(lu_device_is_osd(d));
789 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
792 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
794 return osd_dev(o->oo_dt.do_lu.lo_dev);
797 static inline struct super_block *osd_sb(const struct osd_device *dev)
799 return dev->od_mnt->mnt_sb;
802 static inline int osd_object_is_root(const struct osd_object *obj)
804 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
807 static inline struct osd_object *osd_obj(const struct lu_object *o)
809 LASSERT(lu_device_is_osd(o->lo_dev));
810 return container_of0(o, struct osd_object, oo_dt.do_lu);
813 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
815 return osd_obj(&d->do_lu);
818 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
820 return &osd->od_dt_dev.dd_lu_dev;
823 static inline journal_t *osd_journal(const struct osd_device *dev)
825 return LDISKFS_SB(osd_sb(dev))->s_journal;
828 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
830 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
833 static inline char *osd_name(struct osd_device *osd)
835 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
838 extern const struct dt_body_operations osd_body_ops;
839 extern struct lu_context_key osd_key;
841 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
843 return lu_context_key_get(&env->le_ctx, &osd_key);
846 extern const struct dt_body_operations osd_body_ops_new;
852 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
853 const struct iam_container *bag)
855 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
856 osd_oti_get(env)->oti_it_ipd);
860 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
861 const struct iam_container *bag)
863 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
864 osd_oti_get(env)->oti_idx_ipd);
867 static inline void osd_ipd_put(const struct lu_env *env,
868 const struct iam_container *bag,
869 struct iam_path_descr *ipd)
871 bag->ic_descr->id_ops->id_ipd_free(ipd);
874 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
875 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
876 int write_NUL, loff_t *offs, handle_t *handle);
879 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
881 const char *name, const int namelen)
883 struct osd_thread_info *info = osd_oti_get(env);
884 struct dentry *child_dentry = &info->oti_child_dentry;
885 struct dentry *obj_dentry = &info->oti_obj_dentry;
887 obj_dentry->d_inode = inode;
888 obj_dentry->d_sb = inode->i_sb;
889 obj_dentry->d_name.hash = 0;
891 child_dentry->d_name.hash = 0;
892 child_dentry->d_parent = obj_dentry;
893 child_dentry->d_name.name = name;
894 child_dentry->d_name.len = namelen;
898 extern int osd_trans_declare_op2rb[];
899 extern int ldiskfs_track_declares_assert;
901 static inline void osd_trans_declare_op(const struct lu_env *env,
902 struct osd_thandle *oh,
903 unsigned int op, int credits)
905 struct osd_thread_info *oti = osd_oti_get(env);
907 LASSERT(oh->ot_handle == NULL);
908 if (unlikely(op >= OSD_OT_MAX)) {
909 if (unlikely(ldiskfs_track_declares_assert))
910 LASSERT(op < OSD_OT_MAX);
912 CWARN("%s: Invalid operation index %d\n",
913 osd_name(oti->oti_dev), op);
914 libcfs_debug_dumpstack(NULL);
917 oti->oti_declare_ops[op]++;
918 oti->oti_declare_ops_cred[op] += credits;
920 oh->ot_credits += credits;
923 static inline void osd_trans_exec_op(const struct lu_env *env,
924 struct thandle *th, unsigned int op)
926 struct osd_thread_info *oti = osd_oti_get(env);
927 struct osd_thandle *oh = container_of(th, struct osd_thandle,
931 LASSERT(oh->ot_handle != NULL);
932 if (unlikely(op >= OSD_OT_MAX)) {
933 if (unlikely(ldiskfs_track_declares_assert))
934 LASSERT(op < OSD_OT_MAX);
936 CWARN("%s: Invalid operation index %d\n",
937 osd_name(oti->oti_dev), op);
938 libcfs_debug_dumpstack(NULL);
943 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
944 oti->oti_declare_ops[op]--;
945 oti->oti_declare_ops_rb[op]++;
947 /* all future updates are considered rollback */
948 oti->oti_rollback = true;
949 rb = osd_trans_declare_op2rb[op];
950 if (unlikely(rb >= OSD_OT_MAX)) {
951 if (unlikely(ldiskfs_track_declares_assert))
952 LASSERTF(rb < OSD_OT_MAX, "rb = %u\n", rb);
954 CWARN("%s: Invalid rollback index %d\n",
955 osd_name(oti->oti_dev), rb);
956 libcfs_debug_dumpstack(NULL);
960 if (unlikely(oti->oti_declare_ops_rb[rb] == 0)) {
961 if (unlikely(ldiskfs_track_declares_assert))
962 LASSERTF(oti->oti_declare_ops_rb[rb] > 0,
965 CWARN("%s: Overflow in tracking declares for "
967 osd_name(oti->oti_dev), rb);
968 libcfs_debug_dumpstack(NULL);
972 oti->oti_declare_ops_rb[rb]--;
976 static inline void osd_trans_declare_rb(const struct lu_env *env,
977 struct thandle *th, unsigned int op)
979 struct osd_thread_info *oti = osd_oti_get(env);
980 struct osd_thandle *oh = container_of(th, struct osd_thandle,
983 LASSERT(oh->ot_handle != NULL);
984 if (unlikely(op >= OSD_OT_MAX)) {
985 if (unlikely(ldiskfs_track_declares_assert))
986 LASSERT(op < OSD_OT_MAX);
988 CWARN("%s: Invalid operation index %d\n",
989 osd_name(oti->oti_dev), op);
990 libcfs_debug_dumpstack(NULL);
994 oti->oti_declare_ops_rb[op]++;
999 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1002 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1003 struct lu_fid *befider)
1005 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1006 memcpy(pack->fp_area, befider, sizeof(*befider));
1007 pack->fp_len = sizeof(*befider) + 1;
1011 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1016 switch (pack->fp_len) {
1017 case sizeof *fid + 1:
1018 memcpy(fid, pack->fp_area, sizeof *fid);
1019 fid_be_to_cpu(fid, fid);
1022 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1029 * Quota/Accounting handling
1031 extern const struct dt_index_operations osd_acct_index_ops;
1032 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1033 const struct lu_fid *fid, struct osd_inode_id *id);
1035 /* copy from fs/ext4/dir.c */
1036 static inline int is_32bit_api(void)
1038 #ifdef CONFIG_COMPAT
1039 return is_compat_task();
1041 return (BITS_PER_LONG == 32);
1045 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1047 if ((filp->f_mode & FMODE_32BITHASH) ||
1048 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1049 return LDISKFS_HTREE_EOF_32BIT;
1051 return LDISKFS_HTREE_EOF_64BIT;
1054 static inline int fid_is_internal(const struct lu_fid *fid)
1056 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1059 #ifdef JOURNAL_START_HAS_3ARGS
1060 # define osd_journal_start_sb(sb, type, nblock) \
1061 ldiskfs_journal_start_sb(sb, type, nblock)
1062 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1063 ldiskfs_append(handle, inode, nblock)
1064 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1065 ldiskfs_find_entry(dir, name, de, inlined, lock)
1066 # define osd_journal_start(inode, type, nblocks) \
1067 ldiskfs_journal_start(inode, type, nblocks);
1069 # define LDISKFS_HT_MISC 0
1070 # define osd_journal_start_sb(sb, type, nblock) \
1071 ldiskfs_journal_start_sb(sb, nblock)
1072 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1073 ldiskfs_append(handle, inode, nblock, err)
1074 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1075 ldiskfs_find_entry(dir, name, de, lock)
1076 # define osd_journal_start(inode, type, nblocks) \
1077 ldiskfs_journal_start(inode, nblocks);
1080 void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
1081 void ldiskfs_dec_count(handle_t *handle, struct inode *inode);
1083 #endif /* __KERNEL__ */
1084 #endif /* _OSD_INTERNAL_H */