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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
31 * lustre/osd/osd_internal.h
33 * Shared definitions and declarations for osd module
35 * Author: Nikita Danilov <nikita@clusterfs.com>
38 #ifndef _OSD_INTERNAL_H
39 #define _OSD_INTERNAL_H
43 #include <linux/mutex.h>
44 /* struct rw_semaphore */
45 #include <linux/rwsem.h>
47 #include <linux/dcache.h>
49 #include <linux/dirent.h>
50 #include <linux/statfs.h>
51 #include <linux/bio.h>
52 #include <linux/file.h>
53 #include <ldiskfs/ldiskfs.h>
54 #include <ldiskfs/ldiskfs_jbd2.h>
58 /* class_register_type(), class_unregister_type() */
59 #include <obd_class.h>
60 #include <lustre_disk.h>
61 #include <dt_object.h>
62 #include <lustre_quota.h>
66 #include "osd_scrub.h"
67 #include "osd_quota_fmt.h"
70 extern struct kmem_cache *dynlock_cachep;
72 #define OSD_COUNTERS (0)
74 /* ldiskfs special inode::i_state_flags need to be accessed with
75 * ldiskfs_{set,clear,test}_inode_state() only */
77 /* OI scrub should skip this inode. */
78 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
79 #define LDISKFS_STATE_LUSTRE_DESTROY 30
81 /** Enable thandle usage statistics */
82 #define OSD_THANDLE_STATS (0)
84 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
86 #define OBJECTS "OBJECTS"
87 #define ADMIN_USR "admin_quotafile_v2.usr"
88 #define ADMIN_GRP "admin_quotafile_v2.grp"
90 /* Statfs space reservation for fragmentation and local objects */
91 #define OSD_STATFS_RESERVED (1ULL << 23) /* 8MB */
92 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
94 /* Default extent bytes when declaring write commit */
95 #define OSD_DEFAULT_EXTENT_BYTES (1U << 20)
97 /* check if ldiskfs support project quota */
98 #ifndef LDISKFS_IOC_FSSETXATTR
99 #undef HAVE_PROJECT_QUOTA
102 #define OBD_BRW_MAPPED OBD_BRW_LOCAL1
104 struct osd_directory {
105 struct iam_container od_container;
106 struct iam_descr od_descr;
110 * Object Index (oi) instance.
114 * underlying index object, where fid->id mapping in stored.
116 struct inode *oi_inode;
117 struct osd_directory oi_dir;
120 extern const int osd_dto_credits_noquota[];
123 struct dt_object oo_dt;
125 * Inode for file system object represented by this osd_object. This
126 * inode is pinned for the whole duration of lu_object life.
128 * Not modified concurrently (either setup early during object
129 * creation, or assigned by osd_create() under write lock).
131 struct inode *oo_inode;
133 * to protect index ops.
135 struct htree_lock_head *oo_hl_head;
136 struct rw_semaphore oo_ext_idx_sem;
137 struct rw_semaphore oo_sem;
138 struct osd_directory *oo_dir;
139 /** protects inode attributes. */
143 * Following two members *compat_dot* are used to indicate
144 * the presence of dot and dotdot in the given directory.
145 * This is required for interop mode (b11826).
147 __u32 oo_destroyed:1,
149 oo_compat_dot_created:1,
150 oo_compat_dotdot_created:1;
152 /* the i_flags in LMA */
155 const struct lu_env *oo_owner;
157 struct list_head oo_xattr_list;
158 struct lu_object_header *oo_header;
159 __u64 oo_dirent_count;
163 /* protects on-fly initialization */
164 int oos_subdir_count; /* subdir count for each seq */
165 struct dentry *oos_root; /* O/<seq> */
166 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
167 u64 oos_seq; /* seq number */
168 struct list_head oos_seq_list; /* list to seq_list */
172 struct dentry *om_root; /* dentry for /O */
173 rwlock_t om_seq_list_lock; /* lock for seq_list */
174 struct list_head om_seq_list; /* list head for seq */
176 struct mutex om_dir_init_mutex;
180 struct dentry *om_root; /* AGENT/<index> */
181 u64 om_index; /* mdt index */
182 struct list_head om_list; /* list to omm_list */
185 struct osd_mdobj_map {
186 struct dentry *omm_remote_parent;
188 int osd_ldiskfs_add_entry(struct osd_thread_info *info, struct osd_device *osd,
189 handle_t *handle, struct dentry *child,
190 struct inode *inode, struct htree_lock *hlock);
192 #define OSD_OTABLE_IT_CACHE_SIZE 64
193 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
195 struct osd_inconsistent_item {
196 /* link into lustre_scrub::os_inconsistent_items,
197 * protected by lustre_scrub::os_lock. */
198 struct list_head oii_list;
200 /* The right FID <=> ino#/gen mapping. */
201 struct osd_idmap_cache oii_cache;
203 unsigned int oii_insert:1; /* insert or update mapping. */
206 struct osd_otable_cache {
207 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
209 /* Index for next cache slot to be filled. */
210 int ooc_producer_idx;
212 /* Index for next cache slot to be returned by it::next(). */
213 int ooc_consumer_idx;
215 /* How many items in ooc_cache. */
216 __u64 ooc_cached_items;
218 /* Position for up layer LFSCK iteration pre-loading. */
219 __u64 ooc_pos_preload;
222 struct osd_otable_it {
223 struct osd_device *ooi_dev;
224 struct osd_otable_cache ooi_cache;
225 struct osd_iit_param ooi_iit_param;
227 /* The following bits can be updated/checked w/o lock protection.
228 * If more bits will be introduced in the future and need lock to
229 * protect, please add comment. */
230 unsigned long ooi_used_outside:1, /* Some user out of OSD
231 * uses the iteration. */
232 ooi_all_cached:1, /* No more entries can be
233 * filled into cache. */
234 ooi_user_ready:1, /* The user out of OSD is
235 * ready to iterate. */
236 ooi_waiting:1; /* it::next is waiting. */
239 struct osd_obj_orphan {
240 struct list_head oor_list;
241 struct lu_env *oor_env; /* to identify "own" records */
246 OSD_T10_TYPE_UNKNOWN = 0,
249 OSD_T10_TYPE_CRC = 0x4,
250 OSD_T10_TYPE1_CRC = OSD_T10_TYPE1 | OSD_T10_TYPE_CRC,
251 OSD_T10_TYPE3_CRC = OSD_T10_TYPE3 | OSD_T10_TYPE_CRC,
252 OSD_T10_TYPE1_IP = OSD_T10_TYPE1,
253 OSD_T10_TYPE3_IP = OSD_T10_TYPE3,
261 struct dt_device od_dt_dev;
262 /* information about underlying file system */
263 struct vfsmount *od_mnt;
265 struct osd_oi **od_oi_table;
266 /* total number of OI containers */
271 unsigned int od_fl_capa:1,
280 od_writethrough_cache:1,
282 od_enable_projid_xattr:1;
285 __u32 od_dirent_journal;
287 struct proc_dir_entry *od_proc_entry;
288 struct lprocfs_stats *od_stats;
290 spinlock_t od_osfs_lock;
292 int od_fallocate_zero_blocks;
294 struct lu_site od_site;
296 struct osd_obj_map *od_ost_map;
297 struct osd_mdobj_map *od_mdt_map;
299 /* objects with size > od_readcache_max_filesize will be
300 * served bypassing pagecache unless already cached */
301 unsigned long long od_readcache_max_filesize;
303 /* reads > od_readcache_max_iosize will be
304 * served bypassing pagecache unless already cached */
305 unsigned long od_readcache_max_iosize;
307 /* writes > od_writethough_max_iosize will be
308 * served bypassing pagecache unless already cached */
309 unsigned long od_writethrough_max_iosize;
311 struct brw_stats od_brw_stats;
312 atomic_t od_r_in_flight;
313 atomic_t od_w_in_flight;
315 struct mutex od_otable_mutex;
316 struct osd_otable_it *od_otable_it;
317 struct osd_scrub od_scrub;
318 struct list_head od_ios_list;
320 /* service name associated with the osd device */
321 char od_svname[MAX_OBD_NAME];
322 char od_mntdev[MAX_OBD_NAME];
325 /* quota slave instance for inode */
326 struct qsd_instance *od_quota_slave_md;
328 /* quota slave instance for block */
329 struct qsd_instance *od_quota_slave_dt;
331 /* osd seq instance */
332 struct lu_client_seq *od_cl_seq;
333 /* If the ratio of "the total OI mappings count" vs
334 * "the bad OI mappings count" is lower than the
335 * osd_device::od_full_scrub_ratio, then trigger
336 * OI scrub to scan the whole the device. */
337 __u64 od_full_scrub_ratio;
338 /* If the speed of found bad OI mappings (per minute)
339 * exceeds the osd_device::od_full_scrub_threshold_rate,
340 * then trigger OI scrub to scan the whole device. */
341 __u64 od_full_scrub_threshold_rate;
343 /* a list of orphaned agent inodes, protected with od_osfs_lock */
344 struct list_head od_orphan_list;
345 struct list_head od_index_backup_list;
346 struct list_head od_index_restore_list;
348 struct inode *od_index_backup_inode;
349 enum lustre_index_backup_policy od_index_backup_policy;
350 int od_index_backup_stop;
351 /* T10PI type, zero if not supported */
352 enum osd_t10_type od_t10_type;
353 atomic_t od_commit_cb_in_flight;
354 wait_queue_head_t od_commit_cb_done;
355 unsigned int __percpu *od_extent_bytes_percpu;
358 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
361 return osd->od_quota_slave_dt;
363 return osd->od_quota_slave_md;
366 enum osd_full_scrub_ratio {
367 /* Trigger OI scrub to scan the whole device directly. */
370 /* Because the bad OI mappings count cannot be larger than
371 * the total OI mappints count, then setting OFSR_NEVER means
372 * that the whole device scanning cannot be triggered by auto
373 * detected bad OI mappings during the RPC services. */
375 OFSR_DEFAULT = 10000,
378 #define FULL_SCRUB_THRESHOLD_RATE_DEFAULT 60
380 /* There are at most 15 uid/gid/projids are affected in a transaction, and
381 * that's rename case:
382 * - 3 for source parent uid & gid & projid;
383 * - 3 for source child uid & gid & projid ('..' entry update when
384 * child is directory);
385 * - 3 for target parent uid & gid & projid;
386 * - 3 for target child uid & gid & projid(if the target child exists);
387 * - 3 for root uid & gid(last_rcvd, llog, etc);
390 #define OSD_MAX_UGID_CNT 15
408 struct osd_access_lock {
409 struct list_head tl_list;
410 struct osd_object *tl_obj;
420 struct thandle ot_super;
422 struct ldiskfs_journal_cb_entry ot_jcb;
423 struct list_head ot_commit_dcb_list;
424 struct list_head ot_stop_dcb_list;
425 /* Link to the device, for debugging. */
426 struct lu_ref_link ot_dev_link;
427 unsigned int ot_credits;
428 unsigned int oh_declared_ext;
430 /* quota IDs related to the transaction */
431 unsigned short ot_id_cnt;
432 __u8 ot_id_res[OSD_MAX_UGID_CNT];
433 __u8 ot_id_types[OSD_MAX_UGID_CNT];
434 uid_t ot_id_array[OSD_MAX_UGID_CNT];
435 struct lquota_trans *ot_quota_trans;
437 unsigned int ot_remove_agents:1;
438 #if OSD_THANDLE_STATS
439 /** time when this handle was allocated */
442 /** time when this thanle was started */
445 struct list_head ot_trunc_locks;
449 * Basic transaction credit op
470 #ifdef CONFIG_PROC_FS
472 LPROC_OSD_READ_BYTES = 0,
473 LPROC_OSD_WRITE_BYTES = 1,
474 LPROC_OSD_GET_PAGE = 2,
475 LPROC_OSD_NO_PAGE = 3,
476 LPROC_OSD_CACHE_ACCESS = 4,
477 LPROC_OSD_CACHE_HIT = 5,
478 LPROC_OSD_CACHE_MISS = 6,
480 #if OSD_THANDLE_STATS
481 LPROC_OSD_THANDLE_STARTING,
482 LPROC_OSD_THANDLE_OPEN,
483 LPROC_OSD_THANDLE_CLOSING,
490 * Storage representation for fids.
492 * Variable size, first byte contains the length of the whole record.
494 struct osd_fid_pack {
495 unsigned char fp_len;
496 char fp_area[sizeof(struct lu_fid)];
499 struct osd_it_ea_dirent {
500 struct lu_fid oied_fid;
503 unsigned short oied_namelen;
504 unsigned int oied_type;
506 } __attribute__((packed));
509 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
510 * than lu_dirent struct. osd readdir reads less number of dirent than
511 * required for mdd dir page. so buffer size need to be increased so that
512 * there would be one ext3 readdir for every mdd readdir page.
515 #define OSD_IT_EA_BUFSIZE (PAGE_SIZE + PAGE_SIZE/4)
518 * This is iterator's in-memory data structure in interoperability
519 * mode (i.e. iterator over ldiskfs style directory)
522 struct osd_object *oie_obj;
523 /** used in ldiskfs iterator, to stored file pointer */
524 struct file *oie_file;
525 /** how many entries have been read-cached from storage */
527 /** current entry is being iterated by caller */
529 /** current processing entry */
530 struct osd_it_ea_dirent *oie_dirent;
531 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
536 * Iterator's in-memory data structure for IAM mode.
539 struct osd_object *oi_obj;
540 struct iam_path_descr *oi_ipd;
541 struct iam_iterator oi_it;
544 struct osd_quota_leaf {
545 struct list_head oql_link;
550 * Iterator's in-memory data structure for quota file.
552 struct osd_it_quota {
553 struct osd_object *oiq_obj;
554 /** tree blocks path to where the entry is stored */
555 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
556 /** on-disk offset for current key where quota record can be found */
558 /** identifier for current quota record */
560 /** the record index in the leaf/index block */
561 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
562 /** list of already processed leaf blocks */
563 struct list_head oiq_list;
566 #define MAX_BLOCKS_PER_PAGE (PAGE_SIZE / 512)
569 wait_queue_head_t dr_wait;
570 atomic_t dr_numreqs; /* number of reqs being processed */
575 unsigned int dr_elapsed_valid:1; /* we really did count time */
576 unsigned int dr_rw:1;
577 struct lu_buf dr_pg_buf;
578 struct page **dr_pages;
579 struct niobuf_local **dr_lnbs;
580 struct lu_buf dr_bl_buf;
581 struct lu_buf dr_lnb_buf;
583 ktime_t dr_start_time;
584 ktime_t dr_elapsed; /* how long io took */
585 struct osd_device *dr_dev;
586 unsigned int dr_init_at; /* the line iobuf was initialized */
589 #define osd_dirty_inode(inode, flag) (inode)->i_sb->s_op->dirty_inode((inode), flag)
591 #ifndef HAVE_ALLOC_FILE_PSEUDO
593 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
594 (flag & __FMODE_NONOTIFY)))
596 struct file *alloc_file_pseudo(struct inode *inode, struct vfsmount *mnt,
597 const char *name, int flags,
598 const struct file_operations *fops)
600 struct qstr this = QSTR_INIT(name, strlen(name));
604 path.dentry = d_alloc_pseudo(mnt->mnt_sb, &this);
606 return ERR_PTR(-ENOMEM);
607 path.mnt = mntget(mnt);
608 d_instantiate(path.dentry, inode);
609 file = alloc_file(&path, OPEN_FMODE(flags), fops);
614 file->f_flags = flags;
618 #endif /* !HAVE_ALLOC_FILE_PSEUDO */
620 #ifdef HAVE_INODE_TIMESPEC64
621 # define osd_timespec timespec64
623 # define osd_timespec timespec
626 static inline struct osd_timespec osd_inode_time(struct inode *inode,
629 struct osd_timespec ts = { .tv_sec = seconds };
634 #define OSD_INS_CACHE_SIZE 8
636 struct osd_thread_info {
637 const struct lu_env *oti_env;
639 * used for index operations.
641 struct dentry oti_obj_dentry;
642 struct dentry oti_child_dentry;
644 /** dentry for Iterator context. */
645 struct dentry oti_it_dentry;
648 struct kstatfs oti_ksfs;
650 struct htree_lock *oti_hlock;
652 struct lu_fid oti_fid;
653 struct lu_fid oti_fid2;
654 struct lu_fid oti_fid3;
655 struct osd_inode_id oti_id;
656 struct osd_inode_id oti_id2;
657 struct osd_inode_id oti_id3;
658 struct ost_id oti_ostid;
661 * following ipd and it structures are used for osd_index_iam_lookup()
662 * these are defined separately as we might do index operation
663 * in open iterator session.
666 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
668 unsigned int oti_it_ea_buf_used:1;
670 /* IAM iterator for index operation. */
671 struct iam_iterator oti_idx_it;
673 /** union to guarantee that ->oti_ipd[] has proper alignment. */
676 char oti_it_ipd[DX_IPD_MAX_SIZE];
677 long long oti_alignment_lieutenant;
681 char oti_idx_ipd[DX_IPD_MAX_SIZE];
682 long long oti_alignment_lieutenant_colonel;
685 struct osd_idmap_cache oti_cache;
687 /* dedicated OI cache for insert (which needs inum) */
688 struct osd_idmap_cache *oti_ins_cache;
689 int oti_ins_cache_size;
690 int oti_ins_cache_used;
691 /* inc by osd_trans_create and dec by osd_trans_stop */
692 int oti_ins_cache_depth;
697 /** used in osd_fid_set() to put xattr */
698 struct lu_buf oti_buf;
699 struct lu_buf oti_big_buf;
700 /** used in osd_ea_fid_set() to set fid into common ea */
702 struct lustre_ost_attrs oti_ost_attrs;
703 struct filter_fid_18_23 oti_ff_old;
704 struct filter_fid oti_ff;
707 struct osd_iobuf oti_iobuf;
708 /* used to access objects in /O */
709 struct inode *oti_inode;
710 #define OSD_FID_REC_SZ 32
711 char oti_ldp[OSD_FID_REC_SZ];
712 char oti_ldp2[OSD_FID_REC_SZ];
714 /* used by quota code */
716 #if defined(HAVE_DQUOT_QC_DQBLK)
717 struct qc_dqblk oti_qdq;
719 struct fs_disk_quota oti_fdq;
721 struct if_dqinfo oti_dqinfo;
723 struct lquota_id_info oti_qi;
724 struct lquota_trans oti_quota_trans;
725 union lquota_rec oti_quota_rec;
727 struct lu_seq_range oti_seq_range;
729 /* Tracking for transaction credits, to allow debugging and optimizing
730 * cases where a large number of credits are being allocated for
731 * single transaction. */
732 unsigned int oti_credits_before;
733 unsigned int oti_declare_ops[OSD_OT_MAX];
734 unsigned int oti_declare_ops_cred[OSD_OT_MAX];
735 unsigned int oti_declare_ops_used[OSD_OT_MAX];
736 struct osd_directory oti_iam;
738 struct page **oti_dio_pages;
739 int oti_dio_pages_used;
741 struct osd_it_ea_dirent *oti_seq_dirent;
742 struct osd_it_ea_dirent *oti_dir_dirent;
745 extern int ldiskfs_pdo;
747 /* autoconf test is in lustre-build-ldiskfs.m4 */
748 #ifdef HAVE_BVEC_ITER_ALL
749 #define DECLARE_BVEC_ITER_ALL(iter) struct bvec_iter_all iter
751 #define DECLARE_BVEC_ITER_ALL(iter) int iter
754 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
755 const char *name, void *buf, int len)
760 dentry->d_inode = inode;
761 dentry->d_sb = inode->i_sb;
762 return ll_vfs_getxattr(dentry, inode, name, buf, len);
765 static inline int __osd_xattr_set(struct osd_thread_info *info,
766 struct inode *inode, const char *name,
767 const void *buf, int buflen, int fl)
769 struct dentry *dentry = &info->oti_child_dentry;
771 dquot_initialize(inode);
772 dentry->d_inode = inode;
773 dentry->d_sb = inode->i_sb;
774 return ll_vfs_setxattr(dentry, inode, name, buf, buflen, fl);
777 static inline char *osd_oid_name(char *name, size_t name_size,
778 const struct lu_fid *fid, u64 id)
780 snprintf(name, name_size,
781 (fid_seq_is_rsvd(fid_seq(fid)) ||
782 fid_seq_is_mdt0(fid_seq(fid)) ||
783 fid_seq_is_idif(fid_seq(fid))) ? "%llu" : "%llx", id);
788 #ifdef CONFIG_PROC_FS
790 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
791 int osd_procfs_init(struct osd_device *osd, const char *name);
792 int osd_procfs_fini(struct osd_device *osd);
793 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
794 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 0, 52, 0)
795 int osd_register_proc_index_in_idif(struct osd_device *osd);
799 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
800 struct obd_statfs *sfs, struct obd_statfs_info *info);
801 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
802 struct osd_inode_id *id);
804 osd_iget_fid(struct osd_thread_info *info, struct osd_device *dev,
805 struct osd_inode_id *id, struct lu_fid *fid);
806 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
807 const struct lu_fid *fid, __u32 compat, __u32 incompat);
808 struct osd_obj_seq *osd_seq_load(struct osd_thread_info *info,
809 struct osd_device *osd, u64 seq);
810 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
811 struct dentry *dentry, struct lustre_ost_attrs *loa);
812 void osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
813 struct osd_inode_id *id, const struct lu_fid *fid);
814 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
815 struct dentry *dentry, struct lu_fid *fid);
816 struct osd_it_ea *osd_it_dir_init(const struct lu_env *env,
817 struct osd_device *dev, struct inode *inode,
819 void osd_it_dir_fini(const struct lu_env *env, struct osd_it_ea *oie,
820 struct inode *inode);
821 int osd_ldiskfs_it_fill(const struct lu_env *env, const struct dt_it *di);
823 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
824 void osd_obj_map_fini(struct osd_device *dev);
825 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
826 const struct lu_fid *fid, struct osd_inode_id *id);
827 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
828 const struct lu_fid *fid, const struct osd_inode_id *id,
830 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
831 const struct lu_fid *fid, handle_t *th);
832 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
833 const struct lu_fid *fid, const struct osd_inode_id *id,
835 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
836 struct inode *src_parent, struct dentry *src_child,
837 const struct lu_fid *fid);
838 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
839 const struct lu_fid *fid, struct osd_inode_id *id,
840 enum oi_check_flags flags);
841 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
842 const struct lu_fid *fid, const struct osd_inode_id *id,
844 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
845 const struct lu_fid *fid, const struct osd_inode_id *id,
847 int osd_obj_del_entry(struct osd_thread_info *info, struct osd_device *osd,
848 struct dentry *dird, char *name, int namelen,
851 char *osd_lf_fid2name(const struct lu_fid *fid);
852 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
854 void osd_scrub_stop(struct osd_device *dev);
855 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev,
857 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
858 int osd_oii_insert(struct osd_device *dev, const struct lu_fid *fid,
859 struct osd_inode_id *id, int insert);
860 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
861 struct osd_inode_id *id);
862 void osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
864 struct dentry *osd_lookup_one_len_unlocked(struct osd_device *dev,
866 struct dentry *base, int len);
867 struct dentry *osd_lookup_one_len(struct osd_device *dev, const char *name,
868 struct dentry *base, int len);
870 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
871 u64 seq, struct lu_seq_range *range);
873 int osd_delete_from_remote_parent(const struct lu_env *env,
874 struct osd_device *osd,
875 struct osd_object *obj,
876 struct osd_thandle *oh, bool destroy);
877 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
878 struct osd_object *obj, struct osd_thandle *oh);
879 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
880 struct osd_device *osd,
881 const struct lu_fid *fid,
882 struct osd_inode_id *id);
884 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
886 int osd_scrub_refresh_mapping(struct osd_thread_info *info,
887 struct osd_device *dev,
888 const struct lu_fid *fid,
889 const struct osd_inode_id *id,
891 enum oi_check_flags flags, bool *exist);
893 /* osd_quota_fmt.c */
894 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
895 int type, uint blk, int depth, uint index,
896 struct osd_it_quota *it);
897 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
898 int type, uint blk, uint index,
899 struct osd_it_quota *it);
900 loff_t find_tree_dqentry(const struct lu_env *env,
901 struct osd_object *obj, int type,
902 qid_t dqid, uint blk, int depth,
903 struct osd_it_quota *it);
905 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
906 struct lquota_id_info *qi, struct osd_object *obj,
907 bool enforce, enum osd_quota_local_flags *local_flags);
908 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
909 __u32 projid, long long space, struct osd_thandle *oh,
910 struct osd_object *obj,
911 enum osd_quota_local_flags *local_flags,
912 enum osd_qid_declare_flags);
913 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
914 const struct dt_rec *rec,
915 union lquota_rec *quota_rec);
916 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
918 #ifdef HAVE_PROJECT_QUOTA
919 static inline __u32 i_projid_read(struct inode *inode)
921 return (__u32)from_kprojid(&init_user_ns, LDISKFS_I(inode)->i_projid);
924 static inline void i_projid_write(struct inode *inode, __u32 projid)
927 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
928 LDISKFS_I(inode)->i_projid = kprojid;
931 static inline uid_t i_projid_read(struct inode *inode)
935 static inline void i_projid_write(struct inode *inode, __u32 projid)
941 #ifdef HAVE_LDISKFS_IGET_WITH_FLAGS
942 # define osd_ldiskfs_iget(sb, ino) \
943 ldiskfs_iget((sb), (ino), \
944 LDISKFS_IGET_HANDLE | LDISKFS_IGET_SPECIAL)
946 # define osd_ldiskfs_iget(sb, ino) ldiskfs_iget((sb), (ino))
949 #ifdef HAVE_LDISKFS_INFO_JINODE
950 # define osd_attach_jinode(inode) ldiskfs_inode_attach_jinode(inode)
951 #else /* HAVE_LDISKFS_INFO_JINODE */
952 # define osd_attach_jinode(inode) 0
953 #endif /* HAVE_LDISKFS_INFO_JINODE */
955 #ifdef LDISKFS_HT_MISC
956 # define osd_journal_start_sb(sb, type, nblock) \
957 ldiskfs_journal_start_sb(sb, type, nblock)
958 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
960 ldiskfs_lblk_t *nblock)
964 rc = osd_attach_jinode(inode);
967 return ldiskfs_append(handle, inode, nblock);
970 # ifdef HAVE___LDISKFS_FIND_ENTRY
971 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
972 (__ldiskfs_find_entry(dir, name, de, inlined, lock) ?: \
975 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
976 (ldiskfs_find_entry_locked(dir, name, de, inlined, lock) ?: \
980 # define osd_journal_start(inode, type, nblocks) \
981 ldiskfs_journal_start(inode, type, nblocks)
982 # define osd_transaction_size(dev) \
983 (osd_journal(dev)->j_max_transaction_buffers / 2)
984 #else /* ! defined LDISKFS_HT_MISC */
985 # define LDISKFS_HT_MISC 0
986 # define osd_journal_start_sb(sb, type, nblock) \
987 ldiskfs_journal_start_sb(sb, nblock)
989 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
991 ldiskfs_lblk_t *nblock)
993 struct buffer_head *bh;
996 bh = ldiskfs_append(handle, inode, nblock, &err);
1003 # ifdef HAVE___LDISKFS_FIND_ENTRY
1004 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1005 (__ldiskfs_find_entry(dir, name, de, lock) ?: \
1008 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1009 (ldiskfs_find_entry_locked(dir, name, de, lock) ?: \
1012 # define osd_journal_start(inode, type, nblocks) \
1013 ldiskfs_journal_start(inode, nblocks)
1014 # define osd_transaction_size(dev) \
1015 (osd_journal(dev)->j_max_transaction_buffers)
1016 #endif /* LDISKFS_HT_MISC */
1018 #ifndef HAVE___LDISKFS_FIND_ENTRY
1019 # define __ldiskfs_add_entry(handle, child, inode, hlock) \
1020 ldiskfs_add_entry_locked(handle, child, inode, hlock)
1024 * Invariants, assertions.
1028 * XXX: do not enable this, until invariant checking code is made thread safe
1029 * in the face of pdirops locking.
1031 #define OSD_INVARIANT_CHECKS (0)
1033 #if OSD_INVARIANT_CHECKS
1034 static inline int osd_invariant(const struct osd_object *obj)
1038 ergo(obj->oo_inode != NULL,
1039 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
1040 atomic_read(&obj->oo_inode->i_count) > 0) &&
1041 ergo(obj->oo_dir != NULL &&
1042 obj->oo_dir->od_conationer.ic_object != NULL,
1043 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
1046 #define osd_invariant(obj) (1)
1049 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
1050 #define OSD_READCACHE_MAX_IO_MB 8
1051 #define OSD_WRITECACHE_MAX_IO_MB 8
1053 extern const struct dt_index_operations osd_otable_ops;
1055 static inline int osd_oi_fid2idx(struct osd_device *dev,
1056 const struct lu_fid *fid)
1058 return fid->f_seq & (dev->od_oi_count - 1);
1061 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
1062 const struct lu_fid *fid)
1064 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
1065 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
1066 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
1067 DFID"\n", PFID(fid));
1068 /* It can work even od_oi_count equals to 1 although it's unexpected,
1069 * the only reason we set it to 1 is for performance measurement */
1070 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
1073 extern const struct lu_device_operations osd_lu_ops;
1075 static inline int lu_device_is_osd(const struct lu_device *d)
1077 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
1080 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
1082 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
1083 return container_of(d, struct osd_device, od_dt_dev);
1086 static inline struct osd_device *osd_dev(const struct lu_device *d)
1088 LASSERT(lu_device_is_osd(d));
1089 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
1092 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
1094 return osd_dev(o->oo_dt.do_lu.lo_dev);
1097 static inline struct super_block *osd_sb(const struct osd_device *dev)
1099 return dev->od_mnt->mnt_sb;
1102 static inline const char *osd_sb2name(const struct super_block *sb)
1104 /* this is LDISKFS_SB(sb), but preserves "const" */
1105 const struct ldiskfs_sb_info *sbi = sb->s_fs_info;
1107 return sbi->s_es->s_volume_name;
1110 static inline const char *osd_dev2name(const struct osd_device *dev)
1112 return osd_sb2name(osd_sb(dev));
1115 static inline const char *osd_ino2name(const struct inode *inode)
1117 return osd_sb2name(inode->i_sb);
1121 * Put the osd object once done with it.
1123 * \param obj osd object that needs to be put
1125 static inline void osd_object_put(const struct lu_env *env,
1126 struct osd_object *obj)
1128 dt_object_put(env, &obj->oo_dt);
1131 static inline int osd_object_is_root(const struct osd_object *obj)
1133 return obj->oo_inode && is_root_inode(obj->oo_inode);
1136 static inline struct osd_object *osd_obj(const struct lu_object *o)
1138 LASSERT(lu_device_is_osd(o->lo_dev));
1139 return container_of(o, struct osd_object, oo_dt.do_lu);
1142 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
1144 return osd_obj(&d->do_lu);
1147 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
1149 return &osd->od_dt_dev.dd_lu_dev;
1152 static inline journal_t *osd_journal(const struct osd_device *dev)
1154 return LDISKFS_SB(osd_sb(dev))->s_journal;
1157 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
1159 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
1162 static inline char *osd_name(struct osd_device *osd)
1164 return osd->od_svname;
1167 static inline bool osd_is_ea_inode(struct inode *inode)
1169 return !!(LDISKFS_I(inode)->i_flags & LDISKFS_EA_INODE_FL);
1172 extern const struct dt_body_operations osd_body_ops;
1173 extern struct lu_context_key osd_key;
1175 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
1177 return lu_context_key_get(&env->le_ctx, &osd_key);
1180 extern const struct dt_body_operations osd_body_ops_new;
1186 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
1187 const struct iam_container *bag)
1189 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1190 osd_oti_get(env)->oti_it_ipd);
1194 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
1195 const struct iam_container *bag)
1197 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1198 osd_oti_get(env)->oti_idx_ipd);
1201 static inline void osd_ipd_put(const struct lu_env *env,
1202 const struct iam_container *bag,
1203 struct iam_path_descr *ipd)
1205 bag->ic_descr->id_ops->id_ipd_free(ipd);
1208 int osd_calc_bkmap_credits(struct super_block *sb, struct inode *inode,
1209 const loff_t size, const loff_t pos,
1212 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
1215 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
1216 struct inode *inode,
1217 const char *name, const int namelen)
1219 struct osd_thread_info *info = osd_oti_get(env);
1220 struct dentry *child_dentry = &info->oti_child_dentry;
1221 struct dentry *obj_dentry = &info->oti_obj_dentry;
1223 obj_dentry->d_inode = inode;
1224 obj_dentry->d_sb = inode->i_sb;
1225 obj_dentry->d_name.hash = 0;
1227 child_dentry->d_name.hash = 0;
1228 child_dentry->d_parent = obj_dentry;
1229 child_dentry->d_name.name = name;
1230 child_dentry->d_name.len = namelen;
1231 return child_dentry;
1234 extern int osd_trans_declare_op2rb[];
1235 extern int ldiskfs_track_declares_assert;
1236 void osd_trans_dump_creds(const struct lu_env *env, struct thandle *th);
1238 static inline void osd_trans_declare_op(const struct lu_env *env,
1239 struct osd_thandle *oh,
1240 enum osd_op_type op, int credits)
1242 struct osd_thread_info *oti = osd_oti_get(env);
1244 LASSERT(oh->ot_handle == NULL);
1245 if (unlikely(op >= OSD_OT_MAX)) {
1246 if (unlikely(ldiskfs_track_declares_assert)) {
1247 LASSERT(op < OSD_OT_MAX);
1249 CWARN("%s: Invalid operation index %d\n",
1250 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op);
1251 libcfs_debug_dumpstack(NULL);
1254 oti->oti_declare_ops[op]++;
1255 oti->oti_declare_ops_cred[op] += credits;
1257 oh->ot_credits += credits;
1260 /* linux: v5.4-rc3-21-g933f1c1e0b75 renamed h_buffer_credits */
1261 #ifdef HAVE_JOURNAL_TOTAL_CREDITS
1262 #define h_buffer_credits h_total_credits
1265 static inline void osd_trans_exec_op(const struct lu_env *env,
1267 enum osd_op_type op)
1269 struct osd_thread_info *oti = osd_oti_get(env);
1270 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1272 unsigned int rb, left;
1274 LASSERT(oh->ot_handle != NULL);
1275 if (unlikely(op >= OSD_OT_MAX)) {
1276 if (unlikely(ldiskfs_track_declares_assert))
1277 LASSERT(op < OSD_OT_MAX);
1279 CWARN("%s: opcode %u: invalid value >= %u\n",
1280 osd_name(osd_dt_dev(oh->ot_super.th_dev)),
1282 libcfs_debug_dumpstack(NULL);
1287 /* find rollback (or reverse) operation for the given one
1288 * such an operation doesn't require additional credits
1289 * as the same set of blocks are modified */
1290 rb = osd_trans_declare_op2rb[op];
1292 /* check whether credits for this operation were reserved at all */
1293 if (unlikely(oti->oti_declare_ops_cred[op] == 0 &&
1294 oti->oti_declare_ops_cred[rb] == 0)) {
1295 /* the API is not perfect yet: CREATE does REF_ADD internally
1296 * while DESTROY does not. To rollback CREATE the callers
1297 * needs to call REF_DEL+DESTROY which is hard to detect using
1298 * a simple table of rollback operations */
1299 if (op == OSD_OT_REF_DEL &&
1300 oti->oti_declare_ops_cred[OSD_OT_CREATE] > 0)
1302 if (op == OSD_OT_REF_ADD &&
1303 oti->oti_declare_ops_cred[OSD_OT_DESTROY] > 0)
1305 CWARN("%s: opcode %u: credits = 0, rollback = %u\n",
1306 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
1307 osd_trans_dump_creds(env, th);
1308 LASSERT(!ldiskfs_track_declares_assert);
1312 /* remember how many credits we have unused before the operation */
1313 oti->oti_credits_before = oh->ot_handle->h_buffer_credits;
1314 left = oti->oti_declare_ops_cred[op] - oti->oti_declare_ops_used[op];
1315 if (unlikely(oti->oti_credits_before < left)) {
1316 CWARN("%s: opcode %u: before %u < left %u, rollback = %u\n",
1317 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1318 oti->oti_credits_before, left, rb);
1319 osd_trans_dump_creds(env, th);
1320 /* on a very small fs (testing?) it's possible that
1321 * the transaction can't fit 1/4 of journal, so we
1322 * just request less credits (see osd_trans_start()).
1323 * ignore the same case here */
1324 rb = osd_transaction_size(osd_dt_dev(th->th_dev));
1325 if (unlikely(oh->ot_credits < rb))
1326 LASSERT(!ldiskfs_track_declares_assert);
1330 static inline void osd_trans_exec_check(const struct lu_env *env,
1332 enum osd_op_type op)
1334 struct osd_thread_info *oti = osd_oti_get(env);
1335 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1337 int used, over, quota;
1339 /* how many credits have been used by the operation */
1340 used = oti->oti_credits_before - oh->ot_handle->h_buffer_credits;
1342 if (unlikely(used < 0)) {
1343 /* if some block was allocated and released in the same
1344 * transaction, then it won't be a part of the transaction
1345 * and delta can be negative */
1350 /* rollback operations (e.g. when we destroy just created
1351 * object) should not consume any credits. there is no point
1352 * to confuse the checks below */
1356 oti->oti_declare_ops_used[op] += used;
1357 if (oti->oti_declare_ops_used[op] <= oti->oti_declare_ops_cred[op])
1360 /* we account quota for a whole transaction and any operation can
1361 * consume corresponding credits */
1362 over = oti->oti_declare_ops_used[op] -
1363 oti->oti_declare_ops_cred[op];
1364 quota = oti->oti_declare_ops_cred[OSD_OT_QUOTA] -
1365 oti->oti_declare_ops_used[OSD_OT_QUOTA];
1366 if (over <= quota) {
1367 /* probably that credits were consumed by
1368 * quota indirectly (in the depths of ldiskfs) */
1369 oti->oti_declare_ops_used[OSD_OT_QUOTA] += over;
1370 oti->oti_declare_ops_used[op] -= over;
1372 CWARN("%s: opcode %d: used %u, used now %u, reserved %u\n",
1373 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1374 oti->oti_declare_ops_used[op], used,
1375 oti->oti_declare_ops_cred[op]);
1376 osd_trans_dump_creds(env, th);
1377 if (unlikely(ldiskfs_track_declares_assert))
1383 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1386 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1387 struct lu_fid *befider)
1389 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1390 memcpy(pack->fp_area, befider, sizeof(*befider));
1391 pack->fp_len = sizeof(*befider) + 1;
1395 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1400 switch (pack->fp_len) {
1401 case sizeof *fid + 1:
1402 memcpy(fid, pack->fp_area, sizeof *fid);
1403 fid_be_to_cpu(fid, fid);
1406 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1413 * Quota/Accounting handling
1415 extern const struct dt_index_operations osd_acct_index_ops;
1416 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1417 const struct lu_fid *fid, struct osd_inode_id *id);
1419 /* copy from fs/ext4/dir.c */
1420 static inline int is_32bit_api(void)
1422 #ifdef CONFIG_COMPAT
1423 return in_compat_syscall();
1425 return (BITS_PER_LONG == 32);
1429 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1431 if ((filp->f_mode & FMODE_32BITHASH) ||
1432 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1433 return LDISKFS_HTREE_EOF_32BIT;
1435 return LDISKFS_HTREE_EOF_64BIT;
1438 static inline int fid_is_internal(const struct lu_fid *fid)
1440 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1443 static inline bool is_remote_parent_ino(struct osd_device *o, unsigned long ino)
1448 LASSERT(o->od_mdt_map != NULL);
1450 return ino == o->od_mdt_map->omm_remote_parent->d_inode->i_ino;
1454 * ext4_bread/ldiskfs_bread has either 5 or 4 parameters. The error
1455 * return code has been removed and integrated into the pointer in the
1458 static inline struct buffer_head *__ldiskfs_bread(handle_t *handle,
1459 struct inode *inode,
1460 ldiskfs_lblk_t block,
1464 struct buffer_head *bh;
1467 rc = osd_attach_jinode(inode);
1471 #ifdef HAVE_EXT4_BREAD_4ARGS
1472 bh = ldiskfs_bread(handle, inode, block, create);
1475 bh = ldiskfs_bread(handle, inode, block, create, &rc);
1476 if (bh == NULL && rc != 0)
1482 #ifndef HAVE_BIO_INTEGRITY_ENABLED
1483 bool bio_integrity_enabled(struct bio *bio);
1487 # define bio_get_dev(bio) ((bio)->bi_bdev)
1488 # define bio_get_disk(bio) (bio_get_dev(bio)->bd_disk)
1489 # define bio_get_queue(bio) bdev_get_queue(bio_get_dev(bio))
1490 # define bio_set_dev(bio, bdev) (bio_get_dev(bio) = (bdev))
1492 # define bio_get_disk(bio) ((bio)->bi_disk)
1493 # define bio_get_queue(bio) (bio_get_disk(bio)->queue)
1496 #ifdef HAVE_EXT4_INC_DEC_COUNT_2ARGS
1497 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((h), (inode))
1498 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((h), (inode))
1500 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((inode))
1501 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((inode))
1502 #endif /* HAVE_EXT4_INC_DEC_COUNT_2ARGS */
1504 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1507 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1508 __u32 keysize, __u32 recsize)
1510 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1511 &osd->od_index_backup_list, &osd->od_lock,
1512 &osd->od_index_backup_stop,
1513 fid, keysize, recsize);
1517 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1519 struct osd_thread_info *info = osd_oti_get(env);
1520 struct lu_fid *fid = &info->oti_fid3;
1521 struct osd_inode_id *id = &info->oti_id3;
1523 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1524 osd_id_gen(id, osd->od_index_backup_inode->i_ino,
1525 osd->od_index_backup_inode->i_generation);
1526 osd_add_oi_cache(info, osd, id, fid);
1528 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1529 &osd->od_index_backup_list, &osd->od_lock,
1530 &osd->od_index_backup_stop, backup);
1533 #ifdef LDISKFS_HAS_INCOMPAT_FEATURE
1535 # ifdef LDISKFS_FEATURE_INCOMPAT_EXTENTS
1536 # define ldiskfs_has_feature_extents(sb) \
1537 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EXTENTS)
1539 # ifdef LDISKFS_FEATURE_INCOMPAT_EA_INODE
1540 # define ldiskfs_has_feature_ea_inode(sb) \
1541 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EA_INODE)
1543 # ifdef LDISKFS_FEATURE_INCOMPAT_DIRDATA
1544 # define ldiskfs_has_feature_dirdata(sb) \
1545 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_DIRDATA)
1547 # ifdef LDISKFS_FEATURE_INCOMPAT_LARGEDIR
1548 # define ldiskfs_has_feature_largedir(sb) \
1549 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1550 # define ldiskfs_set_feature_largedir(sb) \
1551 LDISKFS_SET_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1553 # ifdef LDISKFS_FEATURE_COMPAT_HAS_JOURNAL
1554 # define ldiskfs_has_feature_journal(sb) \
1555 LDISKFS_HAS_COMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_HAS_JOURNAL)
1557 # ifdef LDISKFS_FEATURE_RO_COMPAT_QUOTA
1558 # define ldiskfs_has_feature_quota(sb) \
1559 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_QUOTA)
1561 # ifdef LDISKFS_FEATURE_RO_COMPAT_PROJECT
1562 # define ldiskfs_has_feature_project(sb) \
1563 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_PROJECT)
1565 # ifdef LDISKFS_FEATURE_COMPAT_FAST_COMMIT
1566 # define ldiskfs_has_feature_fast_commit(sb) \
1567 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_FAST_COMMIT)
1571 #ifndef ldiskfs_has_feature_fast_commit
1572 #define ldiskfs_has_feature_fast_commit(sb) false
1575 int osd_trunc_lock(struct osd_object *obj, struct osd_thandle *oh,
1577 void osd_trunc_unlock_all(const struct lu_env *env, struct list_head *list);
1578 int osd_process_truncates(const struct lu_env *env, struct list_head *list);
1579 void osd_execute_truncate(struct osd_object *obj);
1581 #ifdef HAVE_BIO_ENDIO_USES_ONE_ARG
1582 #define osd_dio_complete_routine(bio, error) dio_complete_routine(bio)
1584 #define osd_dio_complete_routine(bio, error) dio_complete_routine(bio, error)
1587 #ifndef HAVE___BI_CNT
1588 #define __bi_cnt bi_cnt
1592 #define bi_opf bi_rw
1595 #ifndef HAVE_CLEAN_BDEV_ALIASES
1596 #define clean_bdev_aliases(bdev, block, len) \
1597 unmap_underlying_metadata((bdev), (block))
1600 #ifndef HAVE_BI_STATUS
1601 #define bi_status bi_error
1605 * Maximum size of xattr attributes for FEATURE_INCOMPAT_EA_INODE 1Mb
1606 * This limit is arbitrary, but is reasonable for the xattr API.
1608 #define LDISKFS_XATTR_MAX_LARGE_EA_SIZE (1024 * 1024)
1610 struct osd_bio_private {
1611 struct osd_iobuf *obp_iobuf;
1612 /* Start page index in the obp_iobuf for the bio */
1613 int obp_start_page_idx;
1616 #ifdef HAVE_BIO_INTEGRITY_PREP_FN
1617 int osd_get_integrity_profile(struct osd_device *osd,
1618 integrity_gen_fn **generate_fn,
1619 integrity_vrfy_fn **verify_fn);
1621 #define integrity_gen_fn void
1622 #define integrity_vrfy_fn int
1623 static inline int osd_get_integrity_profile(struct osd_device *osd,
1624 integrity_gen_fn **generate_fn,
1625 integrity_vrfy_fn **verify_fn)
1630 static inline int bio_integrity_prep_fn(struct bio *bio,
1631 integrity_gen_fn *generate_fn,
1632 integrity_vrfy_fn *verify_fn)
1634 #ifdef HAVE_BIO_INTEGRITY_PREP_FN_RETURNS_BOOL
1635 if (bio_integrity_prep(bio))
1640 return bio_integrity_prep(bio);
1645 #ifdef HAVE_BIO_BI_PHYS_SEGMENTS
1646 #define osd_bio_nr_segs(bio) ((bio)->bi_phys_segments)
1648 #define osd_bio_nr_segs(bio) bio_segments((bio))
1649 #endif /* HAVE_BIO_BI_PHYS_SEGMENTS */
1651 #ifdef HAVE_GET_INODE_USAGE
1652 #define lock_dquot_transfer(inode) down_read(&LDISKFS_I(inode)->xattr_sem)
1653 #define unlock_dquot_transfer(inode) up_read(&LDISKFS_I(inode)->xattr_sem)
1655 #define lock_dquot_transfer(inode) do {} while (0)
1656 #define unlock_dquot_transfer(inode) do {} while (0)
1659 #endif /* _OSD_INTERNAL_H */