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_config.h>
56 #include <ldiskfs/ldiskfs.h>
57 #include <ldiskfs/ldiskfs_jbd2.h>
59 /* fsfilt_{get|put}_ops */
60 #include <lustre_fsfilt.h>
64 /* class_register_type(), class_unregister_type(), class_get_type() */
65 #include <obd_class.h>
66 #include <lustre_disk.h>
67 #include <dt_object.h>
68 #include <lustre_quota.h>
72 #include "osd_scrub.h"
73 #include "osd_quota_fmt.h"
77 #define OSD_COUNTERS (0)
79 /* ldiskfs special inode::i_state_flags need to be accessed with
80 * ldiskfs_{set,clear,test}_inode_state() only */
82 /* OI scrub should skip this inode. */
83 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
85 /* Do not add OI mapping for this inode. */
86 #define LDISKFS_STATE_LUSTRE_NO_OI 30
88 /** Enable thandle usage statistics */
89 #define OSD_THANDLE_STATS (0)
91 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
93 #define OBJECTS "OBJECTS"
94 #define ADMIN_USR "admin_quotafile_v2.usr"
95 #define ADMIN_GRP "admin_quotafile_v2.grp"
97 struct osd_directory {
98 struct iam_container od_container;
99 struct iam_descr od_descr;
103 * Object Index (oi) instance.
107 * underlying index object, where fid->id mapping in stored.
109 struct inode *oi_inode;
110 struct osd_directory oi_dir;
113 extern const int osd_dto_credits_noquota[];
116 struct dt_object oo_dt;
118 * Inode for file system object represented by this osd_object. This
119 * inode is pinned for the whole duration of lu_object life.
121 * Not modified concurrently (either setup early during object
122 * creation, or assigned by osd_object_create() under write lock).
124 struct inode *oo_inode;
126 * to protect index ops.
128 struct htree_lock_head *oo_hl_head;
129 struct rw_semaphore oo_ext_idx_sem;
130 struct rw_semaphore oo_sem;
131 struct osd_directory *oo_dir;
132 /** protects inode attributes. */
135 * Following two members are used to indicate the presence of dot and
136 * dotdot in the given directory. This is required for interop mode
139 int oo_compat_dot_created;
140 int oo_compat_dotdot_created;
142 const struct lu_env *oo_owner;
143 #ifdef CONFIG_LOCKDEP
144 struct lockdep_map oo_dep_map;
149 /* protects on-fly initialization */
150 int oos_subdir_count; /* subdir count for each seq */
151 struct dentry *oos_root; /* O/<seq> */
152 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
153 obd_seq oos_seq; /* seq number */
154 cfs_list_t oos_seq_list; /* list to seq_list */
158 struct dentry *om_root; /* dentry for /O */
159 rwlock_t om_seq_list_lock; /* lock for seq_list */
160 cfs_list_t om_seq_list; /* list head for seq */
162 struct semaphore om_dir_init_sem;
166 struct dentry *om_root; /* AGENT/<index> */
167 obd_seq om_index; /* mdt index */
168 cfs_list_t om_list; /* list to omm_list */
171 struct osd_mdobj_map {
172 struct dentry *omm_remote_parent;
175 #define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
176 ll_ldiskfs_find_entry(dir, dentry, de, lock)
177 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
178 ldiskfs_add_entry(handle, child, cinode, hlock)
180 #define OSD_OTABLE_IT_CACHE_SIZE 64
181 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
183 struct osd_inconsistent_item {
184 /* link into osd_scrub::os_inconsistent_items,
185 * protected by osd_scrub::os_lock. */
188 /* The right FID <=> ino#/gen mapping. */
189 struct osd_idmap_cache oii_cache;
191 unsigned int oii_insert:1; /* insert or update mapping. */
194 struct osd_otable_cache {
195 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
197 /* Index for next cache slot to be filled. */
198 int ooc_producer_idx;
200 /* Index for next cache slot to be returned by it::next(). */
201 int ooc_consumer_idx;
203 /* How many items in ooc_cache. */
204 int ooc_cached_items;
206 /* Position for up layer LFSCK iteration pre-loading. */
207 __u32 ooc_pos_preload;
210 struct osd_otable_it {
211 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. */
231 struct dt_device od_dt_dev;
232 /* information about underlying file system */
233 struct vfsmount *od_mnt;
235 struct osd_oi **od_oi_table;
236 /* total number of OI containers */
241 unsigned int od_fl_capa:1,
249 unsigned long od_capa_timeout;
251 struct lustre_capa_key *od_capa_keys;
252 cfs_hlist_head_t *od_capa_hash;
254 cfs_proc_dir_entry_t *od_proc_entry;
255 struct lprocfs_stats *od_stats;
257 * statfs optimization: we cache a bit.
259 cfs_time_t od_osfs_age;
260 struct obd_statfs od_statfs;
261 spinlock_t od_osfs_lock;
263 struct fsfilt_operations *od_fsops;
265 struct lu_site od_site;
267 struct osd_obj_map *od_ost_map;
268 struct osd_mdobj_map *od_mdt_map;
270 unsigned long long od_readcache_max_filesize;
272 int od_writethrough_cache;
274 struct brw_stats od_brw_stats;
275 cfs_atomic_t od_r_in_flight;
276 cfs_atomic_t od_w_in_flight;
278 struct mutex od_otable_mutex;
279 struct osd_otable_it *od_otable_it;
280 struct osd_scrub od_scrub;
281 cfs_list_t od_ios_list;
283 /* service name associated with the osd device */
284 char od_svname[MAX_OBD_NAME];
285 char od_mntdev[MAX_OBD_NAME];
287 /* quota slave instance */
288 struct qsd_instance *od_quota_slave;
291 /* There are at most 10 uid/gids are affected in a transaction, and
292 * that's rename case:
293 * - 2 for source parent uid & gid;
294 * - 2 for source child uid & gid ('..' entry update when child is directory);
295 * - 2 for target parent uid & gid;
296 * - 2 for target child uid & gid (if the target child exists);
297 * - 2 for root uid & gid (last_rcvd, llog, etc);
299 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
300 * the id type of each id in the ot_id_array.
302 #define OSD_MAX_UGID_CNT 10
307 OSD_OT_XATTR_SET = 2,
321 struct thandle ot_super;
323 struct ldiskfs_journal_cb_entry ot_jcb;
324 cfs_list_t ot_dcb_list;
325 /* Link to the device, for debugging. */
326 struct lu_ref_link ot_dev_link;
327 unsigned short ot_credits;
328 unsigned short ot_id_cnt;
329 unsigned short ot_id_type;
330 uid_t ot_id_array[OSD_MAX_UGID_CNT];
331 struct lquota_trans *ot_quota_trans;
332 #if OSD_THANDLE_STATS
333 /** time when this handle was allocated */
334 cfs_time_t oth_alloced;
336 /** time when this thanle was started */
337 cfs_time_t oth_started;
342 * Basic transaction credit op
352 DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
366 LPROC_OSD_READ_BYTES = 0,
367 LPROC_OSD_WRITE_BYTES = 1,
368 LPROC_OSD_GET_PAGE = 2,
369 LPROC_OSD_NO_PAGE = 3,
370 LPROC_OSD_CACHE_ACCESS = 4,
371 LPROC_OSD_CACHE_HIT = 5,
372 LPROC_OSD_CACHE_MISS = 6,
374 #if OSD_THANDLE_STATS
375 LPROC_OSD_THANDLE_STARTING,
376 LPROC_OSD_THANDLE_OPEN,
377 LPROC_OSD_THANDLE_CLOSING,
384 * Storage representation for fids.
386 * Variable size, first byte contains the length of the whole record.
388 struct osd_fid_pack {
389 unsigned char fp_len;
390 char fp_area[sizeof(struct lu_fid)];
393 struct osd_it_ea_dirent {
394 struct lu_fid oied_fid;
397 unsigned short oied_namelen;
398 unsigned int oied_type;
400 } __attribute__((packed));
403 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
404 * than lu_dirent struct. osd readdir reads less number of dirent than
405 * required for mdd dir page. so buffer size need to be increased so that
406 * there would be one ext3 readdir for every mdd readdir page.
409 #define OSD_IT_EA_BUFSIZE (PAGE_CACHE_SIZE + PAGE_CACHE_SIZE/4)
412 * This is iterator's in-memory data structure in interoperability
413 * mode (i.e. iterator over ldiskfs style directory)
416 struct osd_object *oie_obj;
417 /** used in ldiskfs iterator, to stored file pointer */
418 struct file oie_file;
419 /** how many entries have been read-cached from storage */
421 /** current entry is being iterated by caller */
423 /** current processing entry */
424 struct osd_it_ea_dirent *oie_dirent;
425 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
430 * Iterator's in-memory data structure for IAM mode.
433 struct osd_object *oi_obj;
434 struct iam_path_descr *oi_ipd;
435 struct iam_iterator oi_it;
438 struct osd_quota_leaf {
444 * Iterator's in-memory data structure for quota file.
446 struct osd_it_quota {
447 struct osd_object *oiq_obj;
448 /** tree blocks path to where the entry is stored */
449 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
450 /** on-disk offset for current key where quota record can be found */
452 /** identifier for current quota record */
454 /** the record index in the leaf/index block */
455 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
456 /** list of already processed leaf blocks */
460 #define MAX_BLOCKS_PER_PAGE (PAGE_CACHE_SIZE / 512)
464 cfs_atomic_t dr_numreqs; /* number of reqs being processed */
469 unsigned int dr_ignore_quota:1;
470 unsigned int dr_elapsed_valid:1; /* we really did count time */
471 unsigned int dr_rw:1;
472 struct lu_buf dr_pg_buf;
473 struct page **dr_pages;
474 struct lu_buf dr_bl_buf;
475 unsigned long *dr_blocks;
476 unsigned long dr_start_time;
477 unsigned long dr_elapsed; /* how long io took */
478 struct osd_device *dr_dev;
479 unsigned int dr_init_at; /* the line iobuf was initialized */
482 struct osd_thread_info {
483 const struct lu_env *oti_env;
485 * used for index operations.
487 struct dentry oti_obj_dentry;
488 struct dentry oti_child_dentry;
490 /** dentry for Iterator context. */
491 struct dentry oti_it_dentry;
492 struct htree_lock *oti_hlock;
494 struct lu_fid oti_fid;
495 struct lu_fid oti_fid2;
496 struct lu_fid oti_fid3;
497 struct osd_inode_id oti_id;
498 struct osd_inode_id oti_id2;
499 struct osd_inode_id oti_id3;
500 struct ost_id oti_ostid;
503 * XXX temporary: for ->i_op calls.
505 struct timespec oti_time;
507 * XXX temporary: fake struct file for osd_object_sync
509 struct file oti_file;
511 * XXX temporary: for capa operations.
513 struct lustre_capa_key oti_capa_key;
514 struct lustre_capa oti_capa;
516 /** osd_device reference, initialized in osd_trans_start() and
517 used in osd_trans_stop() */
518 struct osd_device *oti_dev;
521 * following ipd and it structures are used for osd_index_iam_lookup()
522 * these are defined separately as we might do index operation
523 * in open iterator session.
526 /** osd iterator context used for iterator session */
529 struct osd_it_iam oti_it;
530 /* ldiskfs iterator data structure,
531 * see osd_it_ea_{init, fini} */
532 struct osd_it_ea oti_it_ea;
533 struct osd_it_quota oti_it_quota;
536 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
539 struct kstatfs oti_ksfs;
541 /** IAM iterator for index operation. */
542 struct iam_iterator oti_idx_it;
544 /** union to guarantee that ->oti_ipd[] has proper alignment. */
546 char oti_it_ipd[DX_IPD_MAX_SIZE];
547 long long oti_alignment_lieutenant;
551 char oti_idx_ipd[DX_IPD_MAX_SIZE];
552 long long oti_alignment_lieutenant_colonel;
555 struct osd_idmap_cache oti_cache;
560 /** used in osd_fid_set() to put xattr */
561 struct lu_buf oti_buf;
562 /** used in osd_ea_fid_set() to set fid into common ea */
564 struct lustre_mdt_attrs oti_mdt_attrs;
565 /* old LMA for compatibility */
566 char oti_mdt_attrs_old[LMA_OLD_SIZE];
569 struct osd_iobuf oti_iobuf;
570 struct inode oti_inode;
571 #define OSD_FID_REC_SZ 32
572 char oti_ldp[OSD_FID_REC_SZ];
573 char oti_ldp2[OSD_FID_REC_SZ];
575 /* used by quota code */
577 #ifdef HAVE_DQUOT_FS_DISK_QUOTA
578 struct fs_disk_quota oti_fdq;
580 struct if_dqblk oti_dqblk;
582 struct if_dqinfo oti_dqinfo;
584 struct lquota_id_info oti_qi;
585 struct lquota_trans oti_quota_trans;
586 union lquota_rec oti_quota_rec;
588 struct lu_seq_range oti_seq_range;
590 /* Tracking for transaction credits, to allow debugging and optimizing
591 * cases where a large number of credits are being allocated for
592 * single transaction. */
593 unsigned short oti_declare_ops[OSD_OT_MAX];
594 unsigned short oti_declare_ops_rb[OSD_OT_MAX];
595 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
600 struct filter_fid_old oti_ff;
601 struct filter_fid oti_ff_new;
605 extern int ldiskfs_pdo;
607 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
608 const char *name, void *buf, int len)
613 dentry->d_inode = inode;
614 dentry->d_sb = inode->i_sb;
615 return inode->i_op->getxattr(dentry, name, buf, len);
618 static inline int __osd_xattr_set(struct osd_thread_info *info,
619 struct inode *inode, const char *name,
620 const void *buf, int buflen, int fl)
622 struct dentry *dentry = &info->oti_child_dentry;
624 ll_vfs_dq_init(inode);
625 dentry->d_inode = inode;
626 dentry->d_sb = inode->i_sb;
627 return inode->i_op->setxattr(dentry, name, buf, buflen, fl);
632 void lprocfs_osd_init_vars(struct lprocfs_static_vars *lvars);
633 int osd_procfs_init(struct osd_device *osd, const char *name);
634 int osd_procfs_fini(struct osd_device *osd);
635 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
638 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
639 struct obd_statfs *sfs);
640 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
641 struct lustre_capa *capa, __u64 opc);
642 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
643 struct osd_inode_id *id);
644 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
645 const struct lu_fid *fid, __u32 compat, __u32 incompat);
646 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
647 struct dentry *dentry, struct lustre_mdt_attrs *lma);
648 int osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
649 struct osd_inode_id *id, const struct lu_fid *fid);
650 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
651 struct dentry *dentry, struct lu_fid *fid);
653 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
654 void osd_obj_map_fini(struct osd_device *dev);
655 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
656 const struct lu_fid *fid, struct osd_inode_id *id);
657 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
658 const struct lu_fid *fid, const struct osd_inode_id *id,
660 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
661 const struct lu_fid *fid, struct thandle *th);
662 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
663 const struct lu_fid *fid, const struct osd_inode_id *id,
665 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
666 struct inode *src_parent, struct dentry *src_child,
667 const struct lu_fid *fid);
668 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
669 const struct lu_fid *fid, struct osd_inode_id *id);
670 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
671 const struct lu_fid *fid, const struct osd_inode_id *id,
673 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
674 const struct lu_fid *fid, const struct osd_inode_id *id,
677 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
678 int osd_scrub_file_store(struct osd_scrub *scrub);
679 char *osd_lf_fid2name(const struct lu_fid *fid);
680 int osd_scrub_start(struct osd_device *dev);
681 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
682 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
683 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
685 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
686 struct osd_inode_id *id);
687 int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
689 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
690 obd_seq seq, struct lu_seq_range *range);
692 int osd_delete_from_remote_parent(const struct lu_env *env,
693 struct osd_device *osd,
694 struct osd_object *obj,
695 struct osd_thandle *oh);
696 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
697 struct osd_object *obj, struct osd_thandle *oh);
698 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
699 struct osd_device *osd,
700 const struct lu_fid *fid,
701 struct osd_inode_id *id);
703 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
705 /* osd_quota_fmt.c */
706 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
707 int type, uint blk, int depth, uint index,
708 struct osd_it_quota *it);
709 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
710 int type, uint blk, uint index,
711 struct osd_it_quota *it);
712 loff_t find_tree_dqentry(const struct lu_env *env,
713 struct osd_object *obj, int type,
714 qid_t dqid, uint blk, int depth,
715 struct osd_it_quota *it);
717 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
718 struct lquota_id_info *qi, bool allocated, int *flags);
719 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
720 long long space, struct osd_thandle *oh,
721 bool is_blk, bool allocated, int *flags, bool force);
722 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
723 const struct dt_rec *rec,
724 union lquota_rec *quota_rec);
725 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
726 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
727 const struct dt_index_features *feat);
729 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
731 return (feat == &dt_quota_iusr_features ||
732 feat == &dt_quota_busr_features ||
733 feat == &dt_quota_igrp_features ||
734 feat == &dt_quota_bgrp_features) ? true : false;
738 * Invariants, assertions.
742 * XXX: do not enable this, until invariant checking code is made thread safe
743 * in the face of pdirops locking.
745 #define OSD_INVARIANT_CHECKS (0)
747 #if OSD_INVARIANT_CHECKS
748 static inline int osd_invariant(const struct osd_object *obj)
752 ergo(obj->oo_inode != NULL,
753 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
754 atomic_read(&obj->oo_inode->i_count) > 0) &&
755 ergo(obj->oo_dir != NULL &&
756 obj->oo_dir->od_conationer.ic_object != NULL,
757 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
760 #define osd_invariant(obj) (1)
763 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
765 extern const struct dt_index_operations osd_otable_ops;
767 static inline int osd_oi_fid2idx(struct osd_device *dev,
768 const struct lu_fid *fid)
770 return fid->f_seq & (dev->od_oi_count - 1);
773 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
774 const struct lu_fid *fid)
776 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
777 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
778 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
779 DFID"\n", PFID(fid));
780 /* It can work even od_oi_count equals to 1 although it's unexpected,
781 * the only reason we set it to 1 is for performance measurement */
782 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
785 extern const struct lu_device_operations osd_lu_ops;
787 static inline int lu_device_is_osd(const struct lu_device *d)
789 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
792 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
794 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
795 return container_of0(d, struct osd_device, od_dt_dev);
798 static inline struct osd_device *osd_dev(const struct lu_device *d)
800 LASSERT(lu_device_is_osd(d));
801 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
804 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
806 return osd_dev(o->oo_dt.do_lu.lo_dev);
809 static inline struct super_block *osd_sb(const struct osd_device *dev)
811 return dev->od_mnt->mnt_sb;
814 static inline int osd_object_is_root(const struct osd_object *obj)
816 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
819 static inline struct osd_object *osd_obj(const struct lu_object *o)
821 LASSERT(lu_device_is_osd(o->lo_dev));
822 return container_of0(o, struct osd_object, oo_dt.do_lu);
825 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
827 return osd_obj(&d->do_lu);
830 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
832 return &osd->od_dt_dev.dd_lu_dev;
835 static inline journal_t *osd_journal(const struct osd_device *dev)
837 return LDISKFS_SB(osd_sb(dev))->s_journal;
840 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
842 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
845 static inline char *osd_name(struct osd_device *osd)
847 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
850 extern const struct dt_body_operations osd_body_ops;
851 extern struct lu_context_key osd_key;
853 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
855 return lu_context_key_get(&env->le_ctx, &osd_key);
858 extern const struct dt_body_operations osd_body_ops_new;
864 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
865 const struct iam_container *bag)
867 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
868 osd_oti_get(env)->oti_it_ipd);
872 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
873 const struct iam_container *bag)
875 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
876 osd_oti_get(env)->oti_idx_ipd);
879 static inline void osd_ipd_put(const struct lu_env *env,
880 const struct iam_container *bag,
881 struct iam_path_descr *ipd)
883 bag->ic_descr->id_ops->id_ipd_free(ipd);
886 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
887 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
888 int write_NUL, loff_t *offs, handle_t *handle);
891 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
893 const char *name, const int namelen)
895 struct osd_thread_info *info = osd_oti_get(env);
896 struct dentry *child_dentry = &info->oti_child_dentry;
897 struct dentry *obj_dentry = &info->oti_obj_dentry;
899 obj_dentry->d_inode = inode;
900 obj_dentry->d_sb = inode->i_sb;
901 obj_dentry->d_name.hash = 0;
903 child_dentry->d_name.hash = 0;
904 child_dentry->d_parent = obj_dentry;
905 child_dentry->d_name.name = name;
906 child_dentry->d_name.len = namelen;
910 extern int osd_trans_declare_op2rb[];
911 extern int ldiskfs_track_declares_assert;
913 static inline void osd_trans_declare_op(const struct lu_env *env,
914 struct osd_thandle *oh,
915 unsigned int op, int credits)
917 struct osd_thread_info *oti = osd_oti_get(env);
919 LASSERT(oh->ot_handle == NULL);
920 if (unlikely(op >= OSD_OT_MAX)) {
921 if (unlikely(ldiskfs_track_declares_assert))
922 LASSERT(op < OSD_OT_MAX);
924 CWARN("%s: Invalid operation index %d\n",
925 osd_name(oti->oti_dev), op);
926 libcfs_debug_dumpstack(NULL);
929 oti->oti_declare_ops[op]++;
930 oti->oti_declare_ops_cred[op] += credits;
932 oh->ot_credits += credits;
935 static inline void osd_trans_exec_op(const struct lu_env *env,
936 struct thandle *th, unsigned int op)
938 struct osd_thread_info *oti = osd_oti_get(env);
939 struct osd_thandle *oh = container_of(th, struct osd_thandle,
943 LASSERT(oh->ot_handle != NULL);
944 if (unlikely(op >= OSD_OT_MAX)) {
945 if (unlikely(ldiskfs_track_declares_assert))
946 LASSERT(op < OSD_OT_MAX);
948 CWARN("%s: Invalid operation index %d\n",
949 osd_name(oti->oti_dev), op);
950 libcfs_debug_dumpstack(NULL);
955 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
956 oti->oti_declare_ops[op]--;
957 oti->oti_declare_ops_rb[op]++;
959 /* all future updates are considered rollback */
960 oti->oti_rollback = true;
961 rb = osd_trans_declare_op2rb[op];
962 if (unlikely(rb >= OSD_OT_MAX)) {
963 if (unlikely(ldiskfs_track_declares_assert))
964 LASSERTF(rb < OSD_OT_MAX, "rb = %u\n", rb);
966 CWARN("%s: Invalid rollback index %d\n",
967 osd_name(oti->oti_dev), rb);
968 libcfs_debug_dumpstack(NULL);
972 if (unlikely(oti->oti_declare_ops_rb[rb] == 0)) {
973 if (unlikely(ldiskfs_track_declares_assert))
974 LASSERTF(oti->oti_declare_ops_rb[rb] > 0,
977 CWARN("%s: Overflow in tracking declares for "
979 osd_name(oti->oti_dev), rb);
980 libcfs_debug_dumpstack(NULL);
984 oti->oti_declare_ops_rb[rb]--;
988 static inline void osd_trans_declare_rb(const struct lu_env *env,
989 struct thandle *th, unsigned int op)
991 struct osd_thread_info *oti = osd_oti_get(env);
992 struct osd_thandle *oh = container_of(th, struct osd_thandle,
995 LASSERT(oh->ot_handle != NULL);
996 if (unlikely(op >= OSD_OT_MAX)) {
997 if (unlikely(ldiskfs_track_declares_assert))
998 LASSERT(op < OSD_OT_MAX);
1000 CWARN("%s: Invalid operation index %d\n",
1001 osd_name(oti->oti_dev), op);
1002 libcfs_debug_dumpstack(NULL);
1006 oti->oti_declare_ops_rb[op]++;
1011 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1014 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1015 struct lu_fid *befider)
1017 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1018 memcpy(pack->fp_area, befider, sizeof(*befider));
1019 pack->fp_len = sizeof(*befider) + 1;
1023 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1028 switch (pack->fp_len) {
1029 case sizeof *fid + 1:
1030 memcpy(fid, pack->fp_area, sizeof *fid);
1031 fid_be_to_cpu(fid, fid);
1034 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1041 * Quota/Accounting handling
1043 extern const struct dt_index_operations osd_acct_index_ops;
1044 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1045 const struct lu_fid *fid, struct osd_inode_id *id);
1047 /* copy from fs/ext4/dir.c */
1048 static inline int is_32bit_api(void)
1050 #ifdef CONFIG_COMPAT
1051 return is_compat_task();
1053 return (BITS_PER_LONG == 32);
1057 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1059 if ((filp->f_mode & FMODE_32BITHASH) ||
1060 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1061 return LDISKFS_HTREE_EOF_32BIT;
1063 return LDISKFS_HTREE_EOF_64BIT;
1066 static inline int fid_is_internal(const struct lu_fid *fid)
1068 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1070 #endif /* __KERNEL__ */
1071 #endif /* _OSD_INTERNAL_H */