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__)
49 #include <linux/mutex.h>
50 /* struct rw_semaphore */
51 #include <linux/rwsem.h>
53 #include <linux/dcache.h>
55 #include <linux/dirent.h>
57 #include <ldiskfs/ldiskfs.h>
58 #include <ldiskfs/ldiskfs_jbd2.h>
62 /* class_register_type(), class_unregister_type(), class_get_type() */
63 #include <obd_class.h>
64 #include <lustre_disk.h>
65 #include <dt_object.h>
66 #include <lustre_quota.h>
70 #include "osd_scrub.h"
71 #include "osd_quota_fmt.h"
75 #define OSD_COUNTERS (0)
77 /* ldiskfs special inode::i_state_flags need to be accessed with
78 * ldiskfs_{set,clear,test}_inode_state() only */
80 /* OI scrub should skip this inode. */
81 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
83 /** Enable thandle usage statistics */
84 #define OSD_THANDLE_STATS (0)
86 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
88 #define OBJECTS "OBJECTS"
89 #define ADMIN_USR "admin_quotafile_v2.usr"
90 #define ADMIN_GRP "admin_quotafile_v2.grp"
92 struct osd_directory {
93 struct iam_container od_container;
94 struct iam_descr od_descr;
98 * Object Index (oi) instance.
102 * underlying index object, where fid->id mapping in stored.
104 struct inode *oi_inode;
105 struct osd_directory oi_dir;
108 extern const int osd_dto_credits_noquota[];
111 struct dt_object oo_dt;
113 * Inode for file system object represented by this osd_object. This
114 * inode is pinned for the whole duration of lu_object life.
116 * Not modified concurrently (either setup early during object
117 * creation, or assigned by osd_object_create() under write lock).
119 struct inode *oo_inode;
121 * to protect index ops.
123 struct htree_lock_head *oo_hl_head;
124 struct rw_semaphore oo_ext_idx_sem;
125 struct rw_semaphore oo_sem;
126 struct osd_directory *oo_dir;
127 /** protects inode attributes. */
130 * Following two members are used to indicate the presence of dot and
131 * dotdot in the given directory. This is required for interop mode
134 int oo_compat_dot_created;
135 int oo_compat_dotdot_created;
137 const struct lu_env *oo_owner;
138 #ifdef CONFIG_LOCKDEP
139 struct lockdep_map oo_dep_map;
144 /* protects on-fly initialization */
145 int oos_subdir_count; /* subdir count for each seq */
146 struct dentry *oos_root; /* O/<seq> */
147 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
148 obd_seq oos_seq; /* seq number */
149 struct list_head oos_seq_list; /* list to seq_list */
153 struct dentry *om_root; /* dentry for /O */
154 rwlock_t om_seq_list_lock; /* lock for seq_list */
155 struct list_head om_seq_list; /* list head for seq */
157 struct mutex om_dir_init_mutex;
161 struct dentry *om_root; /* AGENT/<index> */
162 obd_seq om_index; /* mdt index */
163 struct list_head om_list; /* list to omm_list */
166 struct osd_mdobj_map {
167 struct dentry *omm_remote_parent;
170 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
171 ldiskfs_add_entry(handle, child, cinode, hlock)
173 #define OSD_OTABLE_IT_CACHE_SIZE 64
174 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
176 struct osd_inconsistent_item {
177 /* link into osd_scrub::os_inconsistent_items,
178 * protected by osd_scrub::os_lock. */
179 struct list_head oii_list;
181 /* The right FID <=> ino#/gen mapping. */
182 struct osd_idmap_cache oii_cache;
184 unsigned int oii_insert:1; /* insert or update mapping. */
187 struct osd_otable_cache {
188 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
190 /* Index for next cache slot to be filled. */
191 int ooc_producer_idx;
193 /* Index for next cache slot to be returned by it::next(). */
194 int ooc_consumer_idx;
196 /* How many items in ooc_cache. */
197 int ooc_cached_items;
199 /* Position for up layer LFSCK iteration pre-loading. */
200 __u32 ooc_pos_preload;
203 struct osd_otable_it {
204 struct osd_device *ooi_dev;
205 struct osd_otable_cache ooi_cache;
207 /* The following bits can be updated/checked w/o lock protection.
208 * If more bits will be introduced in the future and need lock to
209 * protect, please add comment. */
210 unsigned long ooi_used_outside:1, /* Some user out of OSD
211 * uses the iteration. */
212 ooi_all_cached:1, /* No more entries can be
213 * filled into cache. */
214 ooi_user_ready:1, /* The user out of OSD is
215 * ready to iterate. */
216 ooi_waiting:1; /* it::next is waiting. */
224 struct dt_device od_dt_dev;
225 /* information about underlying file system */
226 struct vfsmount *od_mnt;
228 struct osd_oi **od_oi_table;
229 /* total number of OI containers */
234 unsigned int od_fl_capa:1,
241 od_lma_self_repair:1;
243 unsigned long od_capa_timeout;
245 struct lustre_capa_key *od_capa_keys;
246 struct hlist_head *od_capa_hash;
248 struct proc_dir_entry *od_proc_entry;
249 struct lprocfs_stats *od_stats;
251 spinlock_t od_osfs_lock;
254 struct lu_site od_site;
256 struct osd_obj_map *od_ost_map;
257 struct osd_mdobj_map *od_mdt_map;
259 unsigned long long od_readcache_max_filesize;
261 int od_writethrough_cache;
263 struct brw_stats od_brw_stats;
264 atomic_t od_r_in_flight;
265 atomic_t od_w_in_flight;
267 struct mutex od_otable_mutex;
268 struct osd_otable_it *od_otable_it;
269 struct osd_scrub od_scrub;
270 struct list_head od_ios_list;
272 /* service name associated with the osd device */
273 char od_svname[MAX_OBD_NAME];
274 char od_mntdev[MAX_OBD_NAME];
276 /* quota slave instance */
277 struct qsd_instance *od_quota_slave;
279 /* osd seq instance */
280 struct lu_client_seq *od_cl_seq;
283 /* There are at most 10 uid/gids are affected in a transaction, and
284 * that's rename case:
285 * - 2 for source parent uid & gid;
286 * - 2 for source child uid & gid ('..' entry update when child is directory);
287 * - 2 for target parent uid & gid;
288 * - 2 for target child uid & gid (if the target child exists);
289 * - 2 for root uid & gid (last_rcvd, llog, etc);
291 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
292 * the id type of each id in the ot_id_array.
294 #define OSD_MAX_UGID_CNT 10
299 OSD_OT_XATTR_SET = 2,
313 struct thandle ot_super;
315 struct ldiskfs_journal_cb_entry ot_jcb;
316 struct list_head ot_dcb_list;
317 /* Link to the device, for debugging. */
318 struct lu_ref_link ot_dev_link;
319 unsigned short ot_credits;
320 unsigned short ot_id_cnt;
321 unsigned short ot_id_type;
322 uid_t ot_id_array[OSD_MAX_UGID_CNT];
323 struct lquota_trans *ot_quota_trans;
324 #if OSD_THANDLE_STATS
325 /** time when this handle was allocated */
326 cfs_time_t oth_alloced;
328 /** time when this thanle was started */
329 cfs_time_t oth_started;
334 * Basic transaction credit op
357 LPROC_OSD_READ_BYTES = 0,
358 LPROC_OSD_WRITE_BYTES = 1,
359 LPROC_OSD_GET_PAGE = 2,
360 LPROC_OSD_NO_PAGE = 3,
361 LPROC_OSD_CACHE_ACCESS = 4,
362 LPROC_OSD_CACHE_HIT = 5,
363 LPROC_OSD_CACHE_MISS = 6,
365 #if OSD_THANDLE_STATS
366 LPROC_OSD_THANDLE_STARTING,
367 LPROC_OSD_THANDLE_OPEN,
368 LPROC_OSD_THANDLE_CLOSING,
375 * Storage representation for fids.
377 * Variable size, first byte contains the length of the whole record.
379 struct osd_fid_pack {
380 unsigned char fp_len;
381 char fp_area[sizeof(struct lu_fid)];
384 struct osd_it_ea_dirent {
385 struct lu_fid oied_fid;
388 unsigned short oied_namelen;
389 unsigned int oied_type;
391 } __attribute__((packed));
394 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
395 * than lu_dirent struct. osd readdir reads less number of dirent than
396 * required for mdd dir page. so buffer size need to be increased so that
397 * there would be one ext3 readdir for every mdd readdir page.
400 #define OSD_IT_EA_BUFSIZE (PAGE_CACHE_SIZE + PAGE_CACHE_SIZE/4)
403 * This is iterator's in-memory data structure in interoperability
404 * mode (i.e. iterator over ldiskfs style directory)
407 struct osd_object *oie_obj;
408 /** used in ldiskfs iterator, to stored file pointer */
409 struct file oie_file;
410 /** how many entries have been read-cached from storage */
412 /** current entry is being iterated by caller */
414 /** current processing entry */
415 struct osd_it_ea_dirent *oie_dirent;
416 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
421 * Iterator's in-memory data structure for IAM mode.
424 struct osd_object *oi_obj;
425 struct iam_path_descr *oi_ipd;
426 struct iam_iterator oi_it;
429 struct osd_quota_leaf {
430 struct list_head oql_link;
435 * Iterator's in-memory data structure for quota file.
437 struct osd_it_quota {
438 struct osd_object *oiq_obj;
439 /** tree blocks path to where the entry is stored */
440 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
441 /** on-disk offset for current key where quota record can be found */
443 /** identifier for current quota record */
445 /** the record index in the leaf/index block */
446 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
447 /** list of already processed leaf blocks */
448 struct list_head oiq_list;
451 #define MAX_BLOCKS_PER_PAGE (PAGE_CACHE_SIZE / 512)
454 wait_queue_head_t dr_wait;
455 atomic_t dr_numreqs; /* number of reqs being processed */
460 unsigned int dr_ignore_quota:1;
461 unsigned int dr_elapsed_valid:1; /* we really did count time */
462 unsigned int dr_rw:1;
463 struct lu_buf dr_pg_buf;
464 struct page **dr_pages;
465 struct lu_buf dr_bl_buf;
466 unsigned long *dr_blocks;
467 unsigned long dr_start_time;
468 unsigned long dr_elapsed; /* how long io took */
469 struct osd_device *dr_dev;
470 unsigned int dr_init_at; /* the line iobuf was initialized */
473 struct osd_thread_info {
474 const struct lu_env *oti_env;
476 * used for index operations.
478 struct dentry oti_obj_dentry;
479 struct dentry oti_child_dentry;
481 /** dentry for Iterator context. */
482 struct dentry oti_it_dentry;
483 struct htree_lock *oti_hlock;
485 struct lu_fid oti_fid;
486 struct lu_fid oti_fid2;
487 struct lu_fid oti_fid3;
488 struct osd_inode_id oti_id;
489 struct osd_inode_id oti_id2;
490 struct osd_inode_id oti_id3;
491 struct ost_id oti_ostid;
494 * XXX temporary: for ->i_op calls.
496 struct timespec oti_time;
498 * XXX temporary: fake struct file for osd_object_sync
500 struct file oti_file;
502 * XXX temporary: for capa operations.
504 struct lustre_capa_key oti_capa_key;
505 struct lustre_capa oti_capa;
507 /** osd_device reference, initialized in osd_trans_start() and
508 used in osd_trans_stop() */
509 struct osd_device *oti_dev;
512 * following ipd and it structures are used for osd_index_iam_lookup()
513 * these are defined separately as we might do index operation
514 * in open iterator session.
517 /** osd iterator context used for iterator session */
520 struct osd_it_iam oti_it;
521 /* ldiskfs iterator data structure,
522 * see osd_it_ea_{init, fini} */
523 struct osd_it_ea oti_it_ea;
524 struct osd_it_quota oti_it_quota;
527 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
530 struct kstatfs oti_ksfs;
532 /** IAM iterator for index operation. */
533 struct iam_iterator oti_idx_it;
535 /** union to guarantee that ->oti_ipd[] has proper alignment. */
537 char oti_it_ipd[DX_IPD_MAX_SIZE];
538 long long oti_alignment_lieutenant;
542 char oti_idx_ipd[DX_IPD_MAX_SIZE];
543 long long oti_alignment_lieutenant_colonel;
546 struct osd_idmap_cache oti_cache;
551 /** used in osd_fid_set() to put xattr */
552 struct lu_buf oti_buf;
553 /** used in osd_ea_fid_set() to set fid into common ea */
555 struct lustre_mdt_attrs oti_mdt_attrs;
556 /* old LMA for compatibility */
557 char oti_mdt_attrs_old[LMA_OLD_SIZE];
560 struct osd_iobuf oti_iobuf;
561 struct inode oti_inode;
562 #define OSD_FID_REC_SZ 32
563 char oti_ldp[OSD_FID_REC_SZ];
564 char oti_ldp2[OSD_FID_REC_SZ];
566 /* used by quota code */
568 #ifdef HAVE_DQUOT_FS_DISK_QUOTA
569 struct fs_disk_quota oti_fdq;
571 struct if_dqblk oti_dqblk;
573 struct if_dqinfo oti_dqinfo;
575 struct lquota_id_info oti_qi;
576 struct lquota_trans oti_quota_trans;
577 union lquota_rec oti_quota_rec;
579 struct lu_seq_range oti_seq_range;
581 /* Tracking for transaction credits, to allow debugging and optimizing
582 * cases where a large number of credits are being allocated for
583 * single transaction. */
584 unsigned short oti_declare_ops[OSD_OT_MAX];
585 unsigned short oti_declare_ops_rb[OSD_OT_MAX];
586 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
591 struct filter_fid_old oti_ff;
592 struct filter_fid oti_ff_new;
596 extern int ldiskfs_pdo;
598 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
599 const char *name, void *buf, int len)
604 dentry->d_inode = inode;
605 dentry->d_sb = inode->i_sb;
606 return inode->i_op->getxattr(dentry, name, buf, len);
609 static inline int __osd_xattr_set(struct osd_thread_info *info,
610 struct inode *inode, const char *name,
611 const void *buf, int buflen, int fl)
613 struct dentry *dentry = &info->oti_child_dentry;
615 ll_vfs_dq_init(inode);
616 dentry->d_inode = inode;
617 dentry->d_sb = inode->i_sb;
618 return inode->i_op->setxattr(dentry, name, buf, buflen, fl);
623 extern struct lprocfs_seq_vars lprocfs_osd_obd_vars[];
624 extern struct lprocfs_seq_vars lprocfs_osd_module_vars[];
625 int osd_procfs_init(struct osd_device *osd, const char *name);
626 int osd_procfs_fini(struct osd_device *osd);
627 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
630 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
631 struct obd_statfs *sfs);
632 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
633 struct lustre_capa *capa, __u64 opc);
634 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
635 struct osd_inode_id *id);
636 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
637 const struct lu_fid *fid, __u32 compat, __u32 incompat);
638 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
639 struct dentry *dentry, struct lustre_mdt_attrs *lma);
640 int osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
641 struct osd_inode_id *id, const struct lu_fid *fid);
642 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
643 struct dentry *dentry, struct lu_fid *fid);
645 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
646 void osd_obj_map_fini(struct osd_device *dev);
647 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
648 const struct lu_fid *fid, struct osd_inode_id *id);
649 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
650 const struct lu_fid *fid, const struct osd_inode_id *id,
652 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
653 const struct lu_fid *fid, handle_t *th);
654 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
655 const struct lu_fid *fid, const struct osd_inode_id *id,
657 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
658 struct inode *src_parent, struct dentry *src_child,
659 const struct lu_fid *fid);
660 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
661 const struct lu_fid *fid, struct osd_inode_id *id);
662 int osd_obj_spec_insert(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_spec_update(struct osd_thread_info *info, struct osd_device *osd,
666 const struct lu_fid *fid, const struct osd_inode_id *id,
669 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
670 int osd_scrub_file_store(struct osd_scrub *scrub);
671 char *osd_lf_fid2name(const struct lu_fid *fid);
672 int osd_scrub_start(struct osd_device *dev);
673 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
674 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
675 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
677 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
678 struct osd_inode_id *id);
679 int osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
681 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
682 obd_seq seq, struct lu_seq_range *range);
684 int osd_delete_from_remote_parent(const struct lu_env *env,
685 struct osd_device *osd,
686 struct osd_object *obj,
687 struct osd_thandle *oh);
688 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
689 struct osd_object *obj, struct osd_thandle *oh);
690 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
691 struct osd_device *osd,
692 const struct lu_fid *fid,
693 struct osd_inode_id *id);
695 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
697 /* osd_quota_fmt.c */
698 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
699 int type, uint blk, int depth, uint index,
700 struct osd_it_quota *it);
701 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
702 int type, uint blk, uint index,
703 struct osd_it_quota *it);
704 loff_t find_tree_dqentry(const struct lu_env *env,
705 struct osd_object *obj, int type,
706 qid_t dqid, uint blk, int depth,
707 struct osd_it_quota *it);
709 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
710 struct lquota_id_info *qi, bool allocated, int *flags);
711 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
712 long long space, struct osd_thandle *oh,
713 bool is_blk, bool allocated, int *flags, bool force);
714 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
715 const struct dt_rec *rec,
716 union lquota_rec *quota_rec);
717 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
718 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
719 const struct dt_index_features *feat);
721 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
723 return (feat == &dt_quota_iusr_features ||
724 feat == &dt_quota_busr_features ||
725 feat == &dt_quota_igrp_features ||
726 feat == &dt_quota_bgrp_features) ? true : false;
729 #ifndef HAVE_I_UID_READ
730 static inline uid_t i_uid_read(const struct inode *inode)
735 static inline gid_t i_gid_read(const struct inode *inode)
740 static inline void i_uid_write(struct inode *inode, uid_t uid)
745 static inline void i_gid_write(struct inode *inode, gid_t gid)
752 * Invariants, assertions.
756 * XXX: do not enable this, until invariant checking code is made thread safe
757 * in the face of pdirops locking.
759 #define OSD_INVARIANT_CHECKS (0)
761 #if OSD_INVARIANT_CHECKS
762 static inline int osd_invariant(const struct osd_object *obj)
766 ergo(obj->oo_inode != NULL,
767 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
768 atomic_read(&obj->oo_inode->i_count) > 0) &&
769 ergo(obj->oo_dir != NULL &&
770 obj->oo_dir->od_conationer.ic_object != NULL,
771 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
774 #define osd_invariant(obj) (1)
777 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
779 extern const struct dt_index_operations osd_otable_ops;
781 static inline int osd_oi_fid2idx(struct osd_device *dev,
782 const struct lu_fid *fid)
784 return fid->f_seq & (dev->od_oi_count - 1);
787 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
788 const struct lu_fid *fid)
790 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
791 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
792 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
793 DFID"\n", PFID(fid));
794 /* It can work even od_oi_count equals to 1 although it's unexpected,
795 * the only reason we set it to 1 is for performance measurement */
796 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
799 extern const struct lu_device_operations osd_lu_ops;
801 static inline int lu_device_is_osd(const struct lu_device *d)
803 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
806 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
808 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
809 return container_of0(d, struct osd_device, od_dt_dev);
812 static inline struct osd_device *osd_dev(const struct lu_device *d)
814 LASSERT(lu_device_is_osd(d));
815 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
818 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
820 return osd_dev(o->oo_dt.do_lu.lo_dev);
823 static inline struct super_block *osd_sb(const struct osd_device *dev)
825 return dev->od_mnt->mnt_sb;
828 static inline int osd_object_is_root(const struct osd_object *obj)
830 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
833 static inline struct osd_object *osd_obj(const struct lu_object *o)
835 LASSERT(lu_device_is_osd(o->lo_dev));
836 return container_of0(o, struct osd_object, oo_dt.do_lu);
839 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
841 return osd_obj(&d->do_lu);
844 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
846 return &osd->od_dt_dev.dd_lu_dev;
849 static inline journal_t *osd_journal(const struct osd_device *dev)
851 return LDISKFS_SB(osd_sb(dev))->s_journal;
854 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
856 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
859 static inline char *osd_name(struct osd_device *osd)
861 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
864 extern const struct dt_body_operations osd_body_ops;
865 extern struct lu_context_key osd_key;
867 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
869 return lu_context_key_get(&env->le_ctx, &osd_key);
872 extern const struct dt_body_operations osd_body_ops_new;
878 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
879 const struct iam_container *bag)
881 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
882 osd_oti_get(env)->oti_it_ipd);
886 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
887 const struct iam_container *bag)
889 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
890 osd_oti_get(env)->oti_idx_ipd);
893 static inline void osd_ipd_put(const struct lu_env *env,
894 const struct iam_container *bag,
895 struct iam_path_descr *ipd)
897 bag->ic_descr->id_ops->id_ipd_free(ipd);
900 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
901 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
902 int write_NUL, loff_t *offs, handle_t *handle);
905 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
907 const char *name, const int namelen)
909 struct osd_thread_info *info = osd_oti_get(env);
910 struct dentry *child_dentry = &info->oti_child_dentry;
911 struct dentry *obj_dentry = &info->oti_obj_dentry;
913 obj_dentry->d_inode = inode;
914 obj_dentry->d_sb = inode->i_sb;
915 obj_dentry->d_name.hash = 0;
917 child_dentry->d_name.hash = 0;
918 child_dentry->d_parent = obj_dentry;
919 child_dentry->d_name.name = name;
920 child_dentry->d_name.len = namelen;
924 extern int osd_trans_declare_op2rb[];
925 extern int ldiskfs_track_declares_assert;
927 static inline void osd_trans_declare_op(const struct lu_env *env,
928 struct osd_thandle *oh,
929 unsigned int op, int credits)
931 struct osd_thread_info *oti = osd_oti_get(env);
933 LASSERT(oh->ot_handle == NULL);
934 if (unlikely(op >= OSD_OT_MAX)) {
935 if (unlikely(ldiskfs_track_declares_assert)) {
936 LASSERT(op < OSD_OT_MAX);
938 CWARN("%s: Invalid operation index %d\n",
939 osd_name(oti->oti_dev), op);
940 libcfs_debug_dumpstack(NULL);
943 oti->oti_declare_ops[op]++;
944 oti->oti_declare_ops_cred[op] += credits;
946 oh->ot_credits += credits;
949 static inline void osd_trans_exec_op(const struct lu_env *env,
950 struct thandle *th, unsigned int op)
952 struct osd_thread_info *oti = osd_oti_get(env);
953 struct osd_thandle *oh = container_of(th, struct osd_thandle,
957 LASSERT(oh->ot_handle != NULL);
958 if (unlikely(op >= OSD_OT_MAX)) {
959 if (unlikely(ldiskfs_track_declares_assert))
960 LASSERT(op < OSD_OT_MAX);
962 CWARN("%s: Invalid operation index %d\n",
963 osd_name(oti->oti_dev), op);
964 libcfs_debug_dumpstack(NULL);
969 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
970 oti->oti_declare_ops[op]--;
971 oti->oti_declare_ops_rb[op]++;
973 /* all future updates are considered rollback */
974 oti->oti_rollback = true;
975 rb = osd_trans_declare_op2rb[op];
976 if (unlikely(rb >= OSD_OT_MAX)) {
977 if (unlikely(ldiskfs_track_declares_assert))
978 LASSERTF(rb < OSD_OT_MAX, "rb = %u\n", rb);
980 CWARN("%s: Invalid rollback index %d\n",
981 osd_name(oti->oti_dev), rb);
982 libcfs_debug_dumpstack(NULL);
986 if (unlikely(oti->oti_declare_ops_rb[rb] == 0)) {
987 if (unlikely(ldiskfs_track_declares_assert))
988 LASSERTF(oti->oti_declare_ops_rb[rb] > 0,
991 CWARN("%s: Overflow in tracking declares for "
993 osd_name(oti->oti_dev), rb);
994 libcfs_debug_dumpstack(NULL);
998 oti->oti_declare_ops_rb[rb]--;
1002 static inline void osd_trans_declare_rb(const struct lu_env *env,
1003 struct thandle *th, unsigned int op)
1005 struct osd_thread_info *oti = osd_oti_get(env);
1006 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1009 LASSERT(oh->ot_handle != NULL);
1010 if (unlikely(op >= OSD_OT_MAX)) {
1011 if (unlikely(ldiskfs_track_declares_assert))
1012 LASSERT(op < OSD_OT_MAX);
1014 CWARN("%s: Invalid operation index %d\n",
1015 osd_name(oti->oti_dev), op);
1016 libcfs_debug_dumpstack(NULL);
1020 oti->oti_declare_ops_rb[op]++;
1025 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1028 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1029 struct lu_fid *befider)
1031 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1032 memcpy(pack->fp_area, befider, sizeof(*befider));
1033 pack->fp_len = sizeof(*befider) + 1;
1037 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1042 switch (pack->fp_len) {
1043 case sizeof *fid + 1:
1044 memcpy(fid, pack->fp_area, sizeof *fid);
1045 fid_be_to_cpu(fid, fid);
1048 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1055 * Quota/Accounting handling
1057 extern const struct dt_index_operations osd_acct_index_ops;
1058 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1059 const struct lu_fid *fid, struct osd_inode_id *id);
1061 /* copy from fs/ext4/dir.c */
1062 static inline int is_32bit_api(void)
1064 #ifdef CONFIG_COMPAT
1065 return is_compat_task();
1067 return (BITS_PER_LONG == 32);
1071 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1073 if ((filp->f_mode & FMODE_32BITHASH) ||
1074 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1075 return LDISKFS_HTREE_EOF_32BIT;
1077 return LDISKFS_HTREE_EOF_64BIT;
1080 static inline int fid_is_internal(const struct lu_fid *fid)
1082 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1085 #ifdef JOURNAL_START_HAS_3ARGS
1086 # define osd_journal_start_sb(sb, type, nblock) \
1087 ldiskfs_journal_start_sb(sb, type, nblock)
1088 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1089 ldiskfs_append(handle, inode, nblock)
1090 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1091 ldiskfs_find_entry(dir, name, de, inlined, lock)
1092 # define osd_journal_start(inode, type, nblocks) \
1093 ldiskfs_journal_start(inode, type, nblocks);
1095 # define LDISKFS_HT_MISC 0
1096 # define osd_journal_start_sb(sb, type, nblock) \
1097 ldiskfs_journal_start_sb(sb, nblock)
1098 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1099 ldiskfs_append(handle, inode, nblock, err)
1100 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1101 ldiskfs_find_entry(dir, name, de, lock)
1102 # define osd_journal_start(inode, type, nblocks) \
1103 ldiskfs_journal_start(inode, nblocks);
1106 void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
1107 void ldiskfs_dec_count(handle_t *handle, struct inode *inode);
1109 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1111 #endif /* __KERNEL__ */
1112 #endif /* _OSD_INTERNAL_H */