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;
280 /* If the ratio of "the total OI mappings count" vs
281 * "the bad OI mappings count" is lower than the
282 * osd_device::od_full_scrub_ratio, then trigger
283 * OI scrub to scan the whole the device. */
284 __u64 od_full_scrub_ratio;
285 /* If the speed of found bad OI mappings (per minute)
286 * exceeds the osd_device::od_full_scrub_speed, then
287 * trigger OI scrub to scan the whole the device. */
288 __u64 od_full_scrub_speed;
291 enum osd_full_scrub_ratio {
292 /* Trigger OI scrub to scan the whole device directly. */
295 /* Because the bad OI mappings count cannot be larger than
296 * the total OI mappints count, then setting OFSR_NEVER means
297 * that the whole device scanning cannot be triggered by auto
298 * detected bad OI mappings during the RPC services. */
300 OFSR_DEFAULT = 10000,
303 #define FULL_SCRUB_SPEED_DEFULT 60
305 /* There are at most 10 uid/gids are affected in a transaction, and
306 * that's rename case:
307 * - 2 for source parent uid & gid;
308 * - 2 for source child uid & gid ('..' entry update when child is directory);
309 * - 2 for target parent uid & gid;
310 * - 2 for target child uid & gid (if the target child exists);
311 * - 2 for root uid & gid (last_rcvd, llog, etc);
313 * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
314 * the id type of each id in the ot_id_array.
316 #define OSD_MAX_UGID_CNT 10
321 OSD_OT_XATTR_SET = 2,
335 struct thandle ot_super;
337 struct ldiskfs_journal_cb_entry ot_jcb;
338 struct list_head ot_dcb_list;
339 /* Link to the device, for debugging. */
340 struct lu_ref_link ot_dev_link;
341 unsigned short ot_credits;
342 unsigned short ot_id_cnt;
343 unsigned short ot_id_type;
344 uid_t ot_id_array[OSD_MAX_UGID_CNT];
345 struct lquota_trans *ot_quota_trans;
346 #if OSD_THANDLE_STATS
347 /** time when this handle was allocated */
348 cfs_time_t oth_alloced;
350 /** time when this thanle was started */
351 cfs_time_t oth_started;
356 * Basic transaction credit op
379 LPROC_OSD_READ_BYTES = 0,
380 LPROC_OSD_WRITE_BYTES = 1,
381 LPROC_OSD_GET_PAGE = 2,
382 LPROC_OSD_NO_PAGE = 3,
383 LPROC_OSD_CACHE_ACCESS = 4,
384 LPROC_OSD_CACHE_HIT = 5,
385 LPROC_OSD_CACHE_MISS = 6,
387 #if OSD_THANDLE_STATS
388 LPROC_OSD_THANDLE_STARTING,
389 LPROC_OSD_THANDLE_OPEN,
390 LPROC_OSD_THANDLE_CLOSING,
397 * Storage representation for fids.
399 * Variable size, first byte contains the length of the whole record.
401 struct osd_fid_pack {
402 unsigned char fp_len;
403 char fp_area[sizeof(struct lu_fid)];
406 struct osd_it_ea_dirent {
407 struct lu_fid oied_fid;
410 unsigned short oied_namelen;
411 unsigned int oied_type;
413 } __attribute__((packed));
416 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
417 * than lu_dirent struct. osd readdir reads less number of dirent than
418 * required for mdd dir page. so buffer size need to be increased so that
419 * there would be one ext3 readdir for every mdd readdir page.
422 #define OSD_IT_EA_BUFSIZE (PAGE_CACHE_SIZE + PAGE_CACHE_SIZE/4)
425 * This is iterator's in-memory data structure in interoperability
426 * mode (i.e. iterator over ldiskfs style directory)
429 struct osd_object *oie_obj;
430 /** used in ldiskfs iterator, to stored file pointer */
431 struct file oie_file;
432 /** how many entries have been read-cached from storage */
434 /** current entry is being iterated by caller */
436 /** current processing entry */
437 struct osd_it_ea_dirent *oie_dirent;
438 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
443 * Iterator's in-memory data structure for IAM mode.
446 struct osd_object *oi_obj;
447 struct iam_path_descr *oi_ipd;
448 struct iam_iterator oi_it;
451 struct osd_quota_leaf {
452 struct list_head oql_link;
457 * Iterator's in-memory data structure for quota file.
459 struct osd_it_quota {
460 struct osd_object *oiq_obj;
461 /** tree blocks path to where the entry is stored */
462 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
463 /** on-disk offset for current key where quota record can be found */
465 /** identifier for current quota record */
467 /** the record index in the leaf/index block */
468 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
469 /** list of already processed leaf blocks */
470 struct list_head oiq_list;
473 #define MAX_BLOCKS_PER_PAGE (PAGE_CACHE_SIZE / 512)
476 wait_queue_head_t dr_wait;
477 atomic_t dr_numreqs; /* number of reqs being processed */
482 unsigned int dr_ignore_quota:1;
483 unsigned int dr_elapsed_valid:1; /* we really did count time */
484 unsigned int dr_rw:1;
485 struct lu_buf dr_pg_buf;
486 struct page **dr_pages;
487 struct lu_buf dr_bl_buf;
488 unsigned long *dr_blocks;
489 unsigned long dr_start_time;
490 unsigned long dr_elapsed; /* how long io took */
491 struct osd_device *dr_dev;
492 unsigned int dr_init_at; /* the line iobuf was initialized */
495 struct osd_thread_info {
496 const struct lu_env *oti_env;
498 * used for index operations.
500 struct dentry oti_obj_dentry;
501 struct dentry oti_child_dentry;
503 /** dentry for Iterator context. */
504 struct dentry oti_it_dentry;
505 struct htree_lock *oti_hlock;
507 struct lu_fid oti_fid;
508 struct lu_fid oti_fid2;
509 struct lu_fid oti_fid3;
510 struct osd_inode_id oti_id;
511 struct osd_inode_id oti_id2;
512 struct osd_inode_id oti_id3;
513 struct ost_id oti_ostid;
516 * XXX temporary: for ->i_op calls.
518 struct timespec oti_time;
520 * XXX temporary: fake struct file for osd_object_sync
522 struct file oti_file;
524 * XXX temporary: for capa operations.
526 struct lustre_capa_key oti_capa_key;
527 struct lustre_capa oti_capa;
529 /** osd_device reference, initialized in osd_trans_start() and
530 used in osd_trans_stop() */
531 struct osd_device *oti_dev;
534 * following ipd and it structures are used for osd_index_iam_lookup()
535 * these are defined separately as we might do index operation
536 * in open iterator session.
539 /** osd iterator context used for iterator session */
542 struct osd_it_iam oti_it;
543 /* ldiskfs iterator data structure,
544 * see osd_it_ea_{init, fini} */
545 struct osd_it_ea oti_it_ea;
546 struct osd_it_quota oti_it_quota;
549 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
552 struct kstatfs oti_ksfs;
554 /** IAM iterator for index operation. */
555 struct iam_iterator oti_idx_it;
557 /** union to guarantee that ->oti_ipd[] has proper alignment. */
559 char oti_it_ipd[DX_IPD_MAX_SIZE];
560 long long oti_alignment_lieutenant;
564 char oti_idx_ipd[DX_IPD_MAX_SIZE];
565 long long oti_alignment_lieutenant_colonel;
568 struct osd_idmap_cache oti_cache;
570 unsigned int oti_it_inline:1;
574 /** used in osd_fid_set() to put xattr */
575 struct lu_buf oti_buf;
576 struct lu_buf oti_big_buf;
577 /** used in osd_ea_fid_set() to set fid into common ea */
579 struct lustre_mdt_attrs oti_mdt_attrs;
580 /* old LMA for compatibility */
581 char oti_mdt_attrs_old[LMA_OLD_SIZE];
584 struct osd_iobuf oti_iobuf;
585 struct inode oti_inode;
586 #define OSD_FID_REC_SZ 32
587 char oti_ldp[OSD_FID_REC_SZ];
588 char oti_ldp2[OSD_FID_REC_SZ];
590 /* used by quota code */
592 #ifdef HAVE_DQUOT_FS_DISK_QUOTA
593 struct fs_disk_quota oti_fdq;
595 struct if_dqblk oti_dqblk;
597 struct if_dqinfo oti_dqinfo;
599 struct lquota_id_info oti_qi;
600 struct lquota_trans oti_quota_trans;
601 union lquota_rec oti_quota_rec;
603 struct lu_seq_range oti_seq_range;
605 /* Tracking for transaction credits, to allow debugging and optimizing
606 * cases where a large number of credits are being allocated for
607 * single transaction. */
608 unsigned short oti_declare_ops[OSD_OT_MAX];
609 unsigned short oti_declare_ops_rb[OSD_OT_MAX];
610 unsigned short oti_declare_ops_cred[OSD_OT_MAX];
615 struct filter_fid_old oti_ff;
616 struct filter_fid oti_ff_new;
620 extern int ldiskfs_pdo;
622 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
623 const char *name, void *buf, int len)
628 dentry->d_inode = inode;
629 dentry->d_sb = inode->i_sb;
630 return inode->i_op->getxattr(dentry, name, buf, len);
633 static inline int __osd_xattr_set(struct osd_thread_info *info,
634 struct inode *inode, const char *name,
635 const void *buf, int buflen, int fl)
637 struct dentry *dentry = &info->oti_child_dentry;
639 ll_vfs_dq_init(inode);
640 dentry->d_inode = inode;
641 dentry->d_sb = inode->i_sb;
642 return inode->i_op->setxattr(dentry, name, buf, buflen, fl);
647 extern struct lprocfs_seq_vars lprocfs_osd_obd_vars[];
648 extern struct lprocfs_seq_vars lprocfs_osd_module_vars[];
649 int osd_procfs_init(struct osd_device *osd, const char *name);
650 int osd_procfs_fini(struct osd_device *osd);
651 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
654 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
655 struct obd_statfs *sfs);
656 int osd_object_auth(const struct lu_env *env, struct dt_object *dt,
657 struct lustre_capa *capa, __u64 opc);
658 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
659 struct osd_inode_id *id);
660 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
661 const struct lu_fid *fid, __u32 compat, __u32 incompat);
662 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
663 struct dentry *dentry, struct lustre_mdt_attrs *lma);
664 int osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
665 struct osd_inode_id *id, const struct lu_fid *fid);
666 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
667 struct dentry *dentry, struct lu_fid *fid);
669 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
670 void osd_obj_map_fini(struct osd_device *dev);
671 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
672 const struct lu_fid *fid, struct osd_inode_id *id);
673 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
674 const struct lu_fid *fid, const struct osd_inode_id *id,
676 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
677 const struct lu_fid *fid, handle_t *th);
678 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
679 const struct lu_fid *fid, const struct osd_inode_id *id,
681 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
682 struct inode *src_parent, struct dentry *src_child,
683 const struct lu_fid *fid);
684 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
685 const struct lu_fid *fid, struct osd_inode_id *id);
686 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
687 const struct lu_fid *fid, const struct osd_inode_id *id,
689 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
690 const struct lu_fid *fid, const struct osd_inode_id *id,
693 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
694 int osd_scrub_file_store(struct osd_scrub *scrub);
695 char *osd_lf_fid2name(const struct lu_fid *fid);
696 int osd_scrub_start(struct osd_device *dev, __u32 flags);
697 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
698 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
699 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
701 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
702 struct osd_inode_id *id);
703 int osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
705 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
706 obd_seq seq, struct lu_seq_range *range);
708 int osd_delete_from_remote_parent(const struct lu_env *env,
709 struct osd_device *osd,
710 struct osd_object *obj,
711 struct osd_thandle *oh);
712 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
713 struct osd_object *obj, struct osd_thandle *oh);
714 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
715 struct osd_device *osd,
716 const struct lu_fid *fid,
717 struct osd_inode_id *id);
719 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
721 /* osd_quota_fmt.c */
722 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
723 int type, uint blk, int depth, uint index,
724 struct osd_it_quota *it);
725 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
726 int type, uint blk, uint index,
727 struct osd_it_quota *it);
728 loff_t find_tree_dqentry(const struct lu_env *env,
729 struct osd_object *obj, int type,
730 qid_t dqid, uint blk, int depth,
731 struct osd_it_quota *it);
733 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
734 struct lquota_id_info *qi, struct osd_object *obj,
735 bool enforce, int *flags);
736 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
737 long long space, struct osd_thandle *oh,
738 struct osd_object *obj, bool is_blk, int *flags,
740 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
741 const struct dt_rec *rec,
742 union lquota_rec *quota_rec);
743 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
744 int osd_quota_migration(const struct lu_env *env, struct dt_object *dt,
745 const struct dt_index_features *feat);
747 static inline bool is_quota_glb_feat(const struct dt_index_features *feat)
749 return (feat == &dt_quota_iusr_features ||
750 feat == &dt_quota_busr_features ||
751 feat == &dt_quota_igrp_features ||
752 feat == &dt_quota_bgrp_features) ? true : false;
755 #ifndef HAVE_I_UID_READ
756 static inline uid_t i_uid_read(const struct inode *inode)
761 static inline gid_t i_gid_read(const struct inode *inode)
766 static inline void i_uid_write(struct inode *inode, uid_t uid)
771 static inline void i_gid_write(struct inode *inode, gid_t gid)
778 * Invariants, assertions.
782 * XXX: do not enable this, until invariant checking code is made thread safe
783 * in the face of pdirops locking.
785 #define OSD_INVARIANT_CHECKS (0)
787 #if OSD_INVARIANT_CHECKS
788 static inline int osd_invariant(const struct osd_object *obj)
792 ergo(obj->oo_inode != NULL,
793 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
794 atomic_read(&obj->oo_inode->i_count) > 0) &&
795 ergo(obj->oo_dir != NULL &&
796 obj->oo_dir->od_conationer.ic_object != NULL,
797 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
800 #define osd_invariant(obj) (1)
803 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
805 extern const struct dt_index_operations osd_otable_ops;
807 static inline int osd_oi_fid2idx(struct osd_device *dev,
808 const struct lu_fid *fid)
810 return fid->f_seq & (dev->od_oi_count - 1);
813 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
814 const struct lu_fid *fid)
816 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
817 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
818 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
819 DFID"\n", PFID(fid));
820 /* It can work even od_oi_count equals to 1 although it's unexpected,
821 * the only reason we set it to 1 is for performance measurement */
822 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
825 extern const struct lu_device_operations osd_lu_ops;
827 static inline int lu_device_is_osd(const struct lu_device *d)
829 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
832 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
834 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
835 return container_of0(d, struct osd_device, od_dt_dev);
838 static inline struct osd_device *osd_dev(const struct lu_device *d)
840 LASSERT(lu_device_is_osd(d));
841 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
844 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
846 return osd_dev(o->oo_dt.do_lu.lo_dev);
849 static inline struct super_block *osd_sb(const struct osd_device *dev)
851 return dev->od_mnt->mnt_sb;
854 static inline int osd_object_is_root(const struct osd_object *obj)
856 return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
859 static inline struct osd_object *osd_obj(const struct lu_object *o)
861 LASSERT(lu_device_is_osd(o->lo_dev));
862 return container_of0(o, struct osd_object, oo_dt.do_lu);
865 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
867 return osd_obj(&d->do_lu);
870 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
872 return &osd->od_dt_dev.dd_lu_dev;
875 static inline journal_t *osd_journal(const struct osd_device *dev)
877 return LDISKFS_SB(osd_sb(dev))->s_journal;
880 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
882 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
885 static inline char *osd_name(struct osd_device *osd)
887 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
890 extern const struct dt_body_operations osd_body_ops;
891 extern struct lu_context_key osd_key;
893 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
895 return lu_context_key_get(&env->le_ctx, &osd_key);
898 extern const struct dt_body_operations osd_body_ops_new;
904 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
905 const struct iam_container *bag)
907 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
908 osd_oti_get(env)->oti_it_ipd);
912 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
913 const struct iam_container *bag)
915 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
916 osd_oti_get(env)->oti_idx_ipd);
919 static inline void osd_ipd_put(const struct lu_env *env,
920 const struct iam_container *bag,
921 struct iam_path_descr *ipd)
923 bag->ic_descr->id_ops->id_ipd_free(ipd);
926 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
927 int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
928 int write_NUL, loff_t *offs, handle_t *handle);
931 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
933 const char *name, const int namelen)
935 struct osd_thread_info *info = osd_oti_get(env);
936 struct dentry *child_dentry = &info->oti_child_dentry;
937 struct dentry *obj_dentry = &info->oti_obj_dentry;
939 obj_dentry->d_inode = inode;
940 obj_dentry->d_sb = inode->i_sb;
941 obj_dentry->d_name.hash = 0;
943 child_dentry->d_name.hash = 0;
944 child_dentry->d_parent = obj_dentry;
945 child_dentry->d_name.name = name;
946 child_dentry->d_name.len = namelen;
950 extern int osd_trans_declare_op2rb[];
951 extern int ldiskfs_track_declares_assert;
953 static inline void osd_trans_declare_op(const struct lu_env *env,
954 struct osd_thandle *oh,
955 unsigned int op, int credits)
957 struct osd_thread_info *oti = osd_oti_get(env);
959 LASSERT(oh->ot_handle == NULL);
960 if (unlikely(op >= OSD_OT_MAX)) {
961 if (unlikely(ldiskfs_track_declares_assert)) {
962 LASSERT(op < OSD_OT_MAX);
964 CWARN("%s: Invalid operation index %d\n",
965 osd_name(oti->oti_dev), op);
966 libcfs_debug_dumpstack(NULL);
969 oti->oti_declare_ops[op]++;
970 oti->oti_declare_ops_cred[op] += credits;
972 oh->ot_credits += credits;
975 static inline void osd_trans_exec_op(const struct lu_env *env,
976 struct thandle *th, unsigned int op)
978 struct osd_thread_info *oti = osd_oti_get(env);
979 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);
995 if (likely(!oti->oti_rollback && oti->oti_declare_ops[op] > 0)) {
996 oti->oti_declare_ops[op]--;
997 oti->oti_declare_ops_rb[op]++;
999 /* all future updates are considered rollback */
1000 oti->oti_rollback = true;
1001 rb = osd_trans_declare_op2rb[op];
1002 if (unlikely(rb >= OSD_OT_MAX)) {
1003 if (unlikely(ldiskfs_track_declares_assert))
1004 LASSERTF(rb < OSD_OT_MAX, "rb = %u\n", rb);
1006 CWARN("%s: Invalid rollback index %d\n",
1007 osd_name(oti->oti_dev), rb);
1008 libcfs_debug_dumpstack(NULL);
1012 if (unlikely(oti->oti_declare_ops_rb[rb] == 0)) {
1013 if (unlikely(ldiskfs_track_declares_assert))
1014 LASSERTF(oti->oti_declare_ops_rb[rb] > 0,
1017 CWARN("%s: Overflow in tracking declares for "
1019 osd_name(oti->oti_dev), rb);
1020 libcfs_debug_dumpstack(NULL);
1024 oti->oti_declare_ops_rb[rb]--;
1028 static inline void osd_trans_declare_rb(const struct lu_env *env,
1029 struct thandle *th, unsigned int op)
1031 struct osd_thread_info *oti = osd_oti_get(env);
1032 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1035 LASSERT(oh->ot_handle != NULL);
1036 if (unlikely(op >= OSD_OT_MAX)) {
1037 if (unlikely(ldiskfs_track_declares_assert))
1038 LASSERT(op < OSD_OT_MAX);
1040 CWARN("%s: Invalid operation index %d\n",
1041 osd_name(oti->oti_dev), op);
1042 libcfs_debug_dumpstack(NULL);
1046 oti->oti_declare_ops_rb[op]++;
1051 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1054 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1055 struct lu_fid *befider)
1057 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1058 memcpy(pack->fp_area, befider, sizeof(*befider));
1059 pack->fp_len = sizeof(*befider) + 1;
1063 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1068 switch (pack->fp_len) {
1069 case sizeof *fid + 1:
1070 memcpy(fid, pack->fp_area, sizeof *fid);
1071 fid_be_to_cpu(fid, fid);
1074 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1081 * Quota/Accounting handling
1083 extern const struct dt_index_operations osd_acct_index_ops;
1084 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1085 const struct lu_fid *fid, struct osd_inode_id *id);
1087 /* copy from fs/ext4/dir.c */
1088 static inline int is_32bit_api(void)
1090 #ifdef CONFIG_COMPAT
1091 return is_compat_task();
1093 return (BITS_PER_LONG == 32);
1097 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1099 if ((filp->f_mode & FMODE_32BITHASH) ||
1100 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1101 return LDISKFS_HTREE_EOF_32BIT;
1103 return LDISKFS_HTREE_EOF_64BIT;
1106 static inline int fid_is_internal(const struct lu_fid *fid)
1108 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1111 static inline unsigned long osd_remote_parent_ino(struct osd_device *dev)
1113 return dev->od_mdt_map->omm_remote_parent->d_inode->i_ino;
1116 #ifdef JOURNAL_START_HAS_3ARGS
1117 # define osd_journal_start_sb(sb, type, nblock) \
1118 ldiskfs_journal_start_sb(sb, type, nblock)
1119 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1120 ldiskfs_append(handle, inode, nblock)
1121 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1122 ldiskfs_find_entry(dir, name, de, inlined, lock)
1123 # define osd_journal_start(inode, type, nblocks) \
1124 ldiskfs_journal_start(inode, type, nblocks)
1125 # define osd_transaction_size(dev) \
1126 (osd_journal(dev)->j_max_transaction_buffers / 2)
1128 # define LDISKFS_HT_MISC 0
1129 # define osd_journal_start_sb(sb, type, nblock) \
1130 ldiskfs_journal_start_sb(sb, nblock)
1131 # define osd_ldiskfs_append(handle, inode, nblock, err) \
1132 ldiskfs_append(handle, inode, nblock, err)
1133 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1134 ldiskfs_find_entry(dir, name, de, lock)
1135 # define osd_journal_start(inode, type, nblocks) \
1136 ldiskfs_journal_start(inode, nblocks)
1137 # define osd_transaction_size(dev) \
1138 (osd_journal(dev)->j_max_transaction_buffers)
1141 void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
1142 void ldiskfs_dec_count(handle_t *handle, struct inode *inode);
1144 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1146 #endif /* _OSD_INTERNAL_H */