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
48 #include <linux/mutex.h>
49 /* struct rw_semaphore */
50 #include <linux/rwsem.h>
52 #include <linux/dcache.h>
54 #include <linux/dirent.h>
55 #include <linux/statfs.h>
56 #include <ldiskfs/ldiskfs.h>
57 #include <ldiskfs/ldiskfs_jbd2.h>
61 /* class_register_type(), class_unregister_type(), class_get_type() */
62 #include <obd_class.h>
63 #include <lustre_disk.h>
64 #include <dt_object.h>
65 #include <lustre_quota.h>
69 #include "osd_scrub.h"
70 #include "osd_quota_fmt.h"
74 #define OSD_COUNTERS (0)
76 /* ldiskfs special inode::i_state_flags need to be accessed with
77 * ldiskfs_{set,clear,test}_inode_state() only */
79 /* OI scrub should skip this inode. */
80 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
82 /** Enable thandle usage statistics */
83 #define OSD_THANDLE_STATS (0)
85 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
87 #define OBJECTS "OBJECTS"
88 #define ADMIN_USR "admin_quotafile_v2.usr"
89 #define ADMIN_GRP "admin_quotafile_v2.grp"
91 struct osd_directory {
92 struct iam_container od_container;
93 struct iam_descr od_descr;
97 * Object Index (oi) instance.
101 * underlying index object, where fid->id mapping in stored.
103 struct inode *oi_inode;
104 struct osd_directory oi_dir;
107 extern const int osd_dto_credits_noquota[];
110 struct dt_object oo_dt;
112 * Inode for file system object represented by this osd_object. This
113 * inode is pinned for the whole duration of lu_object life.
115 * Not modified concurrently (either setup early during object
116 * creation, or assigned by osd_object_create() under write lock).
118 struct inode *oo_inode;
120 * to protect index ops.
122 struct htree_lock_head *oo_hl_head;
123 struct rw_semaphore oo_ext_idx_sem;
124 struct rw_semaphore oo_sem;
125 struct osd_directory *oo_dir;
126 /** protects inode attributes. */
129 * Following two members are used to indicate the presence of dot and
130 * dotdot in the given directory. This is required for interop mode
133 int oo_compat_dot_created;
134 int oo_compat_dotdot_created;
136 const struct lu_env *oo_owner;
137 #ifdef CONFIG_LOCKDEP
138 struct lockdep_map oo_dep_map;
143 /* protects on-fly initialization */
144 int oos_subdir_count; /* subdir count for each seq */
145 struct dentry *oos_root; /* O/<seq> */
146 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
147 obd_seq oos_seq; /* seq number */
148 struct list_head oos_seq_list; /* list to seq_list */
152 struct dentry *om_root; /* dentry for /O */
153 rwlock_t om_seq_list_lock; /* lock for seq_list */
154 struct list_head om_seq_list; /* list head for seq */
156 struct mutex om_dir_init_mutex;
160 struct dentry *om_root; /* AGENT/<index> */
161 obd_seq om_index; /* mdt index */
162 struct list_head om_list; /* list to omm_list */
165 struct osd_mdobj_map {
166 struct dentry *omm_remote_parent;
169 #define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
170 ldiskfs_add_entry(handle, child, cinode, hlock)
172 #define OSD_OTABLE_IT_CACHE_SIZE 64
173 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
175 struct osd_inconsistent_item {
176 /* link into osd_scrub::os_inconsistent_items,
177 * protected by osd_scrub::os_lock. */
178 struct list_head oii_list;
180 /* The right FID <=> ino#/gen mapping. */
181 struct osd_idmap_cache oii_cache;
183 unsigned int oii_insert:1; /* insert or update mapping. */
186 struct osd_otable_cache {
187 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
189 /* Index for next cache slot to be filled. */
190 int ooc_producer_idx;
192 /* Index for next cache slot to be returned by it::next(). */
193 int ooc_consumer_idx;
195 /* How many items in ooc_cache. */
196 int ooc_cached_items;
198 /* Position for up layer LFSCK iteration pre-loading. */
199 __u32 ooc_pos_preload;
202 struct osd_otable_it {
203 struct osd_device *ooi_dev;
204 struct osd_otable_cache ooi_cache;
206 /* The following bits can be updated/checked w/o lock protection.
207 * If more bits will be introduced in the future and need lock to
208 * protect, please add comment. */
209 unsigned long ooi_used_outside:1, /* Some user out of OSD
210 * uses the iteration. */
211 ooi_all_cached:1, /* No more entries can be
212 * filled into cache. */
213 ooi_user_ready:1, /* The user out of OSD is
214 * ready to iterate. */
215 ooi_waiting:1; /* it::next is waiting. */
223 struct dt_device od_dt_dev;
224 /* information about underlying file system */
225 struct vfsmount *od_mnt;
227 struct osd_oi **od_oi_table;
228 /* total number of OI containers */
233 unsigned int od_fl_capa:1,
240 od_lma_self_repair:1;
242 unsigned long od_capa_timeout;
244 struct lustre_capa_key *od_capa_keys;
245 struct hlist_head *od_capa_hash;
247 struct proc_dir_entry *od_proc_entry;
248 struct lprocfs_stats *od_stats;
250 spinlock_t od_osfs_lock;
253 struct lu_site od_site;
255 struct osd_obj_map *od_ost_map;
256 struct osd_mdobj_map *od_mdt_map;
258 unsigned long long od_readcache_max_filesize;
260 int od_writethrough_cache;
262 struct brw_stats od_brw_stats;
263 atomic_t od_r_in_flight;
264 atomic_t od_w_in_flight;
266 struct mutex od_otable_mutex;
267 struct osd_otable_it *od_otable_it;
268 struct osd_scrub od_scrub;
269 struct list_head od_ios_list;
271 /* service name associated with the osd device */
272 char od_svname[MAX_OBD_NAME];
273 char od_mntdev[MAX_OBD_NAME];
275 /* quota slave instance */
276 struct qsd_instance *od_quota_slave;
278 /* osd seq instance */
279 struct lu_client_seq *od_cl_seq;
282 /* There are at most 10 uid/gids are affected in a transaction, and
283 * that's rename case:
284 * - 2 for source parent uid & gid;
285 * - 2 for source child uid & gid ('..' entry update when child is directory);
286 * - 2 for target parent uid & gid;
287 * - 2 for target child uid & gid (if the target child exists);
288 * - 2 for root uid & gid (last_rcvd, llog, etc);
290 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
291 * the id type of each id in the ot_id_array.
293 #define OSD_MAX_UGID_CNT 10
298 OSD_OT_XATTR_SET = 2,
312 struct thandle ot_super;
314 struct ldiskfs_journal_cb_entry ot_jcb;
315 struct list_head ot_dcb_list;
316 /* Link to the device, for debugging. */
317 struct lu_ref_link ot_dev_link;
318 unsigned short ot_credits;
319 unsigned short ot_id_cnt;
320 unsigned short ot_id_type;
321 uid_t ot_id_array[OSD_MAX_UGID_CNT];
322 struct lquota_trans *ot_quota_trans;
323 #if OSD_THANDLE_STATS
324 /** time when this handle was allocated */
325 cfs_time_t oth_alloced;
327 /** time when this thanle was started */
328 cfs_time_t oth_started;
333 * Basic transaction credit op
356 LPROC_OSD_READ_BYTES = 0,
357 LPROC_OSD_WRITE_BYTES = 1,
358 LPROC_OSD_GET_PAGE = 2,
359 LPROC_OSD_NO_PAGE = 3,
360 LPROC_OSD_CACHE_ACCESS = 4,
361 LPROC_OSD_CACHE_HIT = 5,
362 LPROC_OSD_CACHE_MISS = 6,
364 #if OSD_THANDLE_STATS
365 LPROC_OSD_THANDLE_STARTING,
366 LPROC_OSD_THANDLE_OPEN,
367 LPROC_OSD_THANDLE_CLOSING,
374 * Storage representation for fids.
376 * Variable size, first byte contains the length of the whole record.
378 struct osd_fid_pack {
379 unsigned char fp_len;
380 char fp_area[sizeof(struct lu_fid)];
383 struct osd_it_ea_dirent {
384 struct lu_fid oied_fid;
387 unsigned short oied_namelen;
388 unsigned int oied_type;
390 } __attribute__((packed));
393 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
394 * than lu_dirent struct. osd readdir reads less number of dirent than
395 * required for mdd dir page. so buffer size need to be increased so that
396 * there would be one ext3 readdir for every mdd readdir page.
399 #define OSD_IT_EA_BUFSIZE (PAGE_CACHE_SIZE + PAGE_CACHE_SIZE/4)
402 * This is iterator's in-memory data structure in interoperability
403 * mode (i.e. iterator over ldiskfs style directory)
406 struct osd_object *oie_obj;
407 /** used in ldiskfs iterator, to stored file pointer */
408 struct file oie_file;
409 /** how many entries have been read-cached from storage */
411 /** current entry is being iterated by caller */
413 /** current processing entry */
414 struct osd_it_ea_dirent *oie_dirent;
415 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
420 * Iterator's in-memory data structure for IAM mode.
423 struct osd_object *oi_obj;
424 struct iam_path_descr *oi_ipd;
425 struct iam_iterator oi_it;
428 struct osd_quota_leaf {
429 struct list_head oql_link;
434 * Iterator's in-memory data structure for quota file.
436 struct osd_it_quota {
437 struct osd_object *oiq_obj;
438 /** tree blocks path to where the entry is stored */
439 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
440 /** on-disk offset for current key where quota record can be found */
442 /** identifier for current quota record */
444 /** the record index in the leaf/index block */
445 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
446 /** list of already processed leaf blocks */
447 struct list_head oiq_list;
450 #define MAX_BLOCKS_PER_PAGE (PAGE_CACHE_SIZE / 512)
453 wait_queue_head_t dr_wait;
454 atomic_t dr_numreqs; /* number of reqs being processed */
459 unsigned int dr_ignore_quota:1;
460 unsigned int dr_elapsed_valid:1; /* we really did count time */
461 unsigned int dr_rw:1;
462 struct lu_buf dr_pg_buf;
463 struct page **dr_pages;
464 struct lu_buf dr_bl_buf;
465 unsigned long *dr_blocks;
466 unsigned long dr_start_time;
467 unsigned long dr_elapsed; /* how long io took */
468 struct osd_device *dr_dev;
469 unsigned int dr_init_at; /* the line iobuf was initialized */
472 struct osd_thread_info {
473 const struct lu_env *oti_env;
475 * used for index operations.
477 struct dentry oti_obj_dentry;
478 struct dentry oti_child_dentry;
480 /** dentry for Iterator context. */
481 struct dentry oti_it_dentry;
482 struct htree_lock *oti_hlock;
484 struct lu_fid oti_fid;
485 struct lu_fid oti_fid2;
486 struct lu_fid oti_fid3;
487 struct osd_inode_id oti_id;
488 struct osd_inode_id oti_id2;
489 struct osd_inode_id oti_id3;
490 struct ost_id oti_ostid;
493 * XXX temporary: for ->i_op calls.
495 struct timespec oti_time;
497 * XXX temporary: fake struct file for osd_object_sync
499 struct file oti_file;
501 * XXX temporary: for capa operations.
503 struct lustre_capa_key oti_capa_key;
504 struct lustre_capa oti_capa;
506 /** osd_device reference, initialized in osd_trans_start() and
507 used in osd_trans_stop() */
508 struct osd_device *oti_dev;
511 * following ipd and it structures are used for osd_index_iam_lookup()
512 * these are defined separately as we might do index operation
513 * in open iterator session.
516 /** osd iterator context used for iterator session */
519 struct osd_it_iam oti_it;
520 /* ldiskfs iterator data structure,
521 * see osd_it_ea_{init, fini} */
522 struct osd_it_ea oti_it_ea;
523 struct osd_it_quota oti_it_quota;
526 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
529 struct kstatfs oti_ksfs;
531 /** IAM iterator for index operation. */
532 struct iam_iterator oti_idx_it;
534 /** union to guarantee that ->oti_ipd[] has proper alignment. */
536 char oti_it_ipd[DX_IPD_MAX_SIZE];
537 long long oti_alignment_lieutenant;
541 char oti_idx_ipd[DX_IPD_MAX_SIZE];
542 long long oti_alignment_lieutenant_colonel;
545 struct osd_idmap_cache oti_cache;
550 /** used in osd_fid_set() to put xattr */
551 struct lu_buf oti_buf;
552 /** used in osd_ea_fid_set() to set fid into common ea */
554 struct lustre_mdt_attrs oti_mdt_attrs;
555 /* old LMA for compatibility */
556 char oti_mdt_attrs_old[LMA_OLD_SIZE];
559 struct osd_iobuf oti_iobuf;
560 struct inode oti_inode;
561 #define OSD_FID_REC_SZ 32
562 char oti_ldp[OSD_FID_REC_SZ];
563 char oti_ldp2[OSD_FID_REC_SZ];
565 /* used by quota code */
567 #ifdef HAVE_DQUOT_FS_DISK_QUOTA
568 struct fs_disk_quota oti_fdq;
570 struct if_dqblk oti_dqblk;
572 struct if_dqinfo oti_dqinfo;
574 struct lquota_id_info oti_qi;
575 struct lquota_trans oti_quota_trans;
576 union lquota_rec oti_quota_rec;
578 struct lu_seq_range oti_seq_range;
580 /* Tracking for transaction credits, to allow debugging and optimizing
581 * cases where a large number of credits are being allocated for
582 * single transaction. */
583 unsigned short oti_declare_ops[OSD_OT_MAX];
584 unsigned short oti_declare_ops_rb[OSD_OT_MAX];
585 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
590 struct filter_fid_old oti_ff;
591 struct filter_fid oti_ff_new;
595 extern int ldiskfs_pdo;
597 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
598 const char *name, void *buf, int len)
603 dentry->d_inode = inode;
604 dentry->d_sb = inode->i_sb;
605 return inode->i_op->getxattr(dentry, name, buf, len);
608 static inline int __osd_xattr_set(struct osd_thread_info *info,
609 struct inode *inode, const char *name,
610 const void *buf, int buflen, int fl)
612 struct dentry *dentry = &info->oti_child_dentry;
614 ll_vfs_dq_init(inode);
615 dentry->d_inode = inode;
616 dentry->d_sb = inode->i_sb;
617 return inode->i_op->setxattr(dentry, name, buf, buflen, fl);
622 extern struct lprocfs_seq_vars lprocfs_osd_obd_vars[];
623 extern struct lprocfs_seq_vars lprocfs_osd_module_vars[];
624 int osd_procfs_init(struct osd_device *osd, const char *name);
625 int osd_procfs_fini(struct osd_device *osd);
626 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
629 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
630 struct obd_statfs *sfs);
631 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
632 struct lustre_capa *capa, __u64 opc);
633 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
634 struct osd_inode_id *id);
635 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
636 const struct lu_fid *fid, __u32 compat, __u32 incompat);
637 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
638 struct dentry *dentry, struct lustre_mdt_attrs *lma);
639 int osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
640 struct osd_inode_id *id, const struct lu_fid *fid);
641 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
642 struct dentry *dentry, struct lu_fid *fid);
644 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
645 void osd_obj_map_fini(struct osd_device *dev);
646 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
647 const struct lu_fid *fid, struct osd_inode_id *id);
648 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
649 const struct lu_fid *fid, const struct osd_inode_id *id,
651 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
652 const struct lu_fid *fid, handle_t *th);
653 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
654 const struct lu_fid *fid, const struct osd_inode_id *id,
656 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
657 struct inode *src_parent, struct dentry *src_child,
658 const struct lu_fid *fid);
659 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
660 const struct lu_fid *fid, struct osd_inode_id *id);
661 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
662 const struct lu_fid *fid, const struct osd_inode_id *id,
664 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
665 const struct lu_fid *fid, const struct osd_inode_id *id,
668 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
669 int osd_scrub_file_store(struct osd_scrub *scrub);
670 char *osd_lf_fid2name(const struct lu_fid *fid);
671 int osd_scrub_start(struct osd_device *dev);
672 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
673 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
674 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
676 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
677 struct osd_inode_id *id);
678 int osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
680 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
681 obd_seq seq, struct lu_seq_range *range);
683 int osd_delete_from_remote_parent(const struct lu_env *env,
684 struct osd_device *osd,
685 struct osd_object *obj,
686 struct osd_thandle *oh);
687 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
688 struct osd_object *obj, struct osd_thandle *oh);
689 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
690 struct osd_device *osd,
691 const struct lu_fid *fid,
692 struct osd_inode_id *id);
694 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
696 /* osd_quota_fmt.c */
697 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
698 int type, uint blk, int depth, uint index,
699 struct osd_it_quota *it);
700 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
701 int type, uint blk, uint index,
702 struct osd_it_quota *it);
703 loff_t find_tree_dqentry(const struct lu_env *env,
704 struct osd_object *obj, int type,
705 qid_t dqid, uint blk, int depth,
706 struct osd_it_quota *it);
708 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
709 struct lquota_id_info *qi, bool allocated, int *flags);
710 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
711 long long space, struct osd_thandle *oh,
712 bool is_blk, bool allocated, int *flags, bool force);
713 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
714 const struct dt_rec *rec,
715 union lquota_rec *quota_rec);
716 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
717 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
718 const struct dt_index_features *feat);
720 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
722 return (feat == &dt_quota_iusr_features ||
723 feat == &dt_quota_busr_features ||
724 feat == &dt_quota_igrp_features ||
725 feat == &dt_quota_bgrp_features) ? true : false;
728 #ifndef HAVE_I_UID_READ
729 static inline uid_t i_uid_read(const struct inode *inode)
734 static inline gid_t i_gid_read(const struct inode *inode)
739 static inline void i_uid_write(struct inode *inode, uid_t uid)
744 static inline void i_gid_write(struct inode *inode, gid_t gid)
751 * Invariants, assertions.
755 * XXX: do not enable this, until invariant checking code is made thread safe
756 * in the face of pdirops locking.
758 #define OSD_INVARIANT_CHECKS (0)
760 #if OSD_INVARIANT_CHECKS
761 static inline int osd_invariant(const struct osd_object *obj)
765 ergo(obj->oo_inode != NULL,
766 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
767 atomic_read(&obj->oo_inode->i_count) > 0) &&
768 ergo(obj->oo_dir != NULL &&
769 obj->oo_dir->od_conationer.ic_object != NULL,
770 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
773 #define osd_invariant(obj) (1)
776 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
778 extern const struct dt_index_operations osd_otable_ops;
780 static inline int osd_oi_fid2idx(struct osd_device *dev,
781 const struct lu_fid *fid)
783 return fid->f_seq & (dev->od_oi_count - 1);
786 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
787 const struct lu_fid *fid)
789 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
790 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
791 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
792 DFID"\n", PFID(fid));
793 /* It can work even od_oi_count equals to 1 although it's unexpected,
794 * the only reason we set it to 1 is for performance measurement */
795 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
798 extern const struct lu_device_operations osd_lu_ops;
800 static inline int lu_device_is_osd(const struct lu_device *d)
802 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
805 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
807 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
808 return container_of0(d, struct osd_device, od_dt_dev);
811 static inline struct osd_device *osd_dev(const struct lu_device *d)
813 LASSERT(lu_device_is_osd(d));
814 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
817 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
819 return osd_dev(o->oo_dt.do_lu.lo_dev);
822 static inline struct super_block *osd_sb(const struct osd_device *dev)
824 return dev->od_mnt->mnt_sb;
827 static inline int osd_object_is_root(const struct osd_object *obj)
829 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
832 static inline struct osd_object *osd_obj(const struct lu_object *o)
834 LASSERT(lu_device_is_osd(o->lo_dev));
835 return container_of0(o, struct osd_object, oo_dt.do_lu);
838 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
840 return osd_obj(&d->do_lu);
843 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
845 return &osd->od_dt_dev.dd_lu_dev;
848 static inline journal_t *osd_journal(const struct osd_device *dev)
850 return LDISKFS_SB(osd_sb(dev))->s_journal;
853 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
855 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
858 static inline char *osd_name(struct osd_device *osd)
860 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
863 extern const struct dt_body_operations osd_body_ops;
864 extern struct lu_context_key osd_key;
866 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
868 return lu_context_key_get(&env->le_ctx, &osd_key);
871 extern const struct dt_body_operations osd_body_ops_new;
877 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
878 const struct iam_container *bag)
880 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
881 osd_oti_get(env)->oti_it_ipd);
885 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
886 const struct iam_container *bag)
888 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
889 osd_oti_get(env)->oti_idx_ipd);
892 static inline void osd_ipd_put(const struct lu_env *env,
893 const struct iam_container *bag,
894 struct iam_path_descr *ipd)
896 bag->ic_descr->id_ops->id_ipd_free(ipd);
899 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
900 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
901 int write_NUL, loff_t *offs, handle_t *handle);
904 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
906 const char *name, const int namelen)
908 struct osd_thread_info *info = osd_oti_get(env);
909 struct dentry *child_dentry = &info->oti_child_dentry;
910 struct dentry *obj_dentry = &info->oti_obj_dentry;
912 obj_dentry->d_inode = inode;
913 obj_dentry->d_sb = inode->i_sb;
914 obj_dentry->d_name.hash = 0;
916 child_dentry->d_name.hash = 0;
917 child_dentry->d_parent = obj_dentry;
918 child_dentry->d_name.name = name;
919 child_dentry->d_name.len = namelen;
923 extern int osd_trans_declare_op2rb[];
924 extern int ldiskfs_track_declares_assert;
926 static inline void osd_trans_declare_op(const struct lu_env *env,
927 struct osd_thandle *oh,
928 unsigned int op, int credits)
930 struct osd_thread_info *oti = osd_oti_get(env);
932 LASSERT(oh->ot_handle == NULL);
933 if (unlikely(op >= OSD_OT_MAX)) {
934 if (unlikely(ldiskfs_track_declares_assert)) {
935 LASSERT(op < OSD_OT_MAX);
937 CWARN("%s: Invalid operation index %d\n",
938 osd_name(oti->oti_dev), op);
939 libcfs_debug_dumpstack(NULL);
942 oti->oti_declare_ops[op]++;
943 oti->oti_declare_ops_cred[op] += credits;
945 oh->ot_credits += credits;
948 static inline void osd_trans_exec_op(const struct lu_env *env,
949 struct thandle *th, unsigned int op)
951 struct osd_thread_info *oti = osd_oti_get(env);
952 struct osd_thandle *oh = container_of(th, struct osd_thandle,
956 LASSERT(oh->ot_handle != NULL);
957 if (unlikely(op >= OSD_OT_MAX)) {
958 if (unlikely(ldiskfs_track_declares_assert))
959 LASSERT(op < OSD_OT_MAX);
961 CWARN("%s: Invalid operation index %d\n",
962 osd_name(oti->oti_dev), op);
963 libcfs_debug_dumpstack(NULL);
968 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
969 oti->oti_declare_ops[op]--;
970 oti->oti_declare_ops_rb[op]++;
972 /* all future updates are considered rollback */
973 oti->oti_rollback = true;
974 rb = osd_trans_declare_op2rb[op];
975 if (unlikely(rb >= OSD_OT_MAX)) {
976 if (unlikely(ldiskfs_track_declares_assert))
977 LASSERTF(rb < OSD_OT_MAX, "rb = %u\n", rb);
979 CWARN("%s: Invalid rollback index %d\n",
980 osd_name(oti->oti_dev), rb);
981 libcfs_debug_dumpstack(NULL);
985 if (unlikely(oti->oti_declare_ops_rb[rb] == 0)) {
986 if (unlikely(ldiskfs_track_declares_assert))
987 LASSERTF(oti->oti_declare_ops_rb[rb] > 0,
990 CWARN("%s: Overflow in tracking declares for "
992 osd_name(oti->oti_dev), rb);
993 libcfs_debug_dumpstack(NULL);
997 oti->oti_declare_ops_rb[rb]--;
1001 static inline void osd_trans_declare_rb(const struct lu_env *env,
1002 struct thandle *th, unsigned int op)
1004 struct osd_thread_info *oti = osd_oti_get(env);
1005 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1008 LASSERT(oh->ot_handle != NULL);
1009 if (unlikely(op >= OSD_OT_MAX)) {
1010 if (unlikely(ldiskfs_track_declares_assert))
1011 LASSERT(op < OSD_OT_MAX);
1013 CWARN("%s: Invalid operation index %d\n",
1014 osd_name(oti->oti_dev), op);
1015 libcfs_debug_dumpstack(NULL);
1019 oti->oti_declare_ops_rb[op]++;
1024 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1027 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1028 struct lu_fid *befider)
1030 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1031 memcpy(pack->fp_area, befider, sizeof(*befider));
1032 pack->fp_len = sizeof(*befider) + 1;
1036 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1041 switch (pack->fp_len) {
1042 case sizeof *fid + 1:
1043 memcpy(fid, pack->fp_area, sizeof *fid);
1044 fid_be_to_cpu(fid, fid);
1047 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1054 * Quota/Accounting handling
1056 extern const struct dt_index_operations osd_acct_index_ops;
1057 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1058 const struct lu_fid *fid, struct osd_inode_id *id);
1060 /* copy from fs/ext4/dir.c */
1061 static inline int is_32bit_api(void)
1063 #ifdef CONFIG_COMPAT
1064 return is_compat_task();
1066 return (BITS_PER_LONG == 32);
1070 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1072 if ((filp->f_mode & FMODE_32BITHASH) ||
1073 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1074 return LDISKFS_HTREE_EOF_32BIT;
1076 return LDISKFS_HTREE_EOF_64BIT;
1079 static inline int fid_is_internal(const struct lu_fid *fid)
1081 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1084 #ifdef JOURNAL_START_HAS_3ARGS
1085 # define osd_journal_start_sb(sb, type, nblock) \
1086 ldiskfs_journal_start_sb(sb, type, nblock)
1087 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1088 ldiskfs_append(handle, inode, nblock)
1089 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1090 ldiskfs_find_entry(dir, name, de, inlined, lock)
1091 # define osd_journal_start(inode, type, nblocks) \
1092 ldiskfs_journal_start(inode, type, nblocks);
1094 # define LDISKFS_HT_MISC 0
1095 # define osd_journal_start_sb(sb, type, nblock) \
1096 ldiskfs_journal_start_sb(sb, nblock)
1097 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1098 ldiskfs_append(handle, inode, nblock, err)
1099 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1100 ldiskfs_find_entry(dir, name, de, lock)
1101 # define osd_journal_start(inode, type, nblocks) \
1102 ldiskfs_journal_start(inode, nblocks);
1105 void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
1106 void ldiskfs_dec_count(handle_t *handle, struct inode *inode);
1108 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1110 #endif /* _OSD_INTERNAL_H */