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>
55 #include <lustre_compat.h>
59 /* class_register_type(), class_unregister_type() */
60 #include <obd_class.h>
61 #include <lustre_disk.h>
62 #include <dt_object.h>
63 #include <lustre_quota.h>
67 #include "osd_scrub.h"
68 #include "osd_quota_fmt.h"
70 #ifdef HAVE_LINUX_BLK_INTEGRITY_HEADER
71 #include <linux/blk-integrity.h>
75 extern struct kmem_cache *dynlock_cachep;
77 #define OSD_COUNTERS (0)
79 /* ldiskfs special inode::i_state_flags need to be accessed with
80 * ldiskfs_{set,clear,test}_inode_state() only */
82 /* OI scrub should skip this inode. */
83 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
84 #define LDISKFS_STATE_LUSTRE_DESTROY 30
86 /** Enable thandle usage statistics */
87 #define OSD_THANDLE_STATS (0)
89 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
91 #define OBJECTS "OBJECTS"
92 #define ADMIN_USR "admin_quotafile_v2.usr"
93 #define ADMIN_GRP "admin_quotafile_v2.grp"
95 /* Statfs space reservation for fragmentation and local objects */
96 #define OSD_STATFS_RESERVED (1ULL << 23) /* 8MB */
97 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
99 /* Default extent bytes when declaring write commit */
100 #define OSD_DEFAULT_EXTENT_BYTES (1U << 20)
102 /* check if ldiskfs support project quota */
103 #if LDISKFS_MAXQUOTAS < 3
104 #undef HAVE_PROJECT_QUOTA
107 #define OBD_BRW_MAPPED OBD_BRW_LOCAL1
109 struct osd_directory {
110 struct iam_container od_container;
111 struct iam_descr od_descr;
115 * Object Index (oi) instance.
119 * underlying index object, where fid->id mapping in stored.
121 struct inode *oi_inode;
122 struct osd_directory oi_dir;
125 extern const int osd_dto_credits_noquota[];
128 struct dt_object oo_dt;
130 * Inode for file system object represented by this osd_object. This
131 * inode is pinned for the whole duration of lu_object life.
133 * Not modified concurrently (either setup early during object
134 * creation, or assigned by osd_create() under write lock).
136 struct inode *oo_inode;
138 * to protect index ops.
140 struct htree_lock_head *oo_hl_head;
141 struct rw_semaphore oo_ext_idx_sem;
142 struct rw_semaphore oo_sem;
143 struct osd_directory *oo_dir;
144 /** protects inode attributes. */
148 * Following two members *compat_dot* are used to indicate
149 * the presence of dot and dotdot in the given directory.
150 * This is required for interop mode (b11826).
152 __u32 oo_destroyed:1,
154 oo_compat_dot_created:1,
155 oo_compat_dotdot_created:1;
157 /* the i_flags in LMA */
159 atomic_t oo_dirent_count;
161 const struct lu_env *oo_owner;
163 struct list_head oo_xattr_list;
164 struct lu_object_header *oo_header;
168 /* protects on-fly initialization */
169 int oos_subdir_count; /* subdir count for each seq */
170 struct dentry *oos_root; /* O/<seq> */
171 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
172 u64 oos_seq; /* seq number */
173 struct list_head oos_seq_list; /* list to seq_list */
177 struct dentry *om_root; /* dentry for /O */
178 rwlock_t om_seq_list_lock; /* lock for seq_list */
179 struct list_head om_seq_list; /* list head for seq */
181 struct mutex om_dir_init_mutex;
185 struct dentry *om_root; /* AGENT/<index> */
186 u64 om_index; /* mdt index */
187 struct list_head om_list; /* list to omm_list */
190 struct osd_mdobj_map {
191 struct dentry *omm_remote_parent;
193 int osd_ldiskfs_add_entry(struct osd_thread_info *info, struct osd_device *osd,
194 handle_t *handle, struct dentry *child,
195 struct inode *inode, struct htree_lock *hlock);
197 #define OSD_OTABLE_IT_CACHE_SIZE 64
198 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
200 struct osd_inconsistent_item {
201 /* link into lustre_scrub::os_inconsistent_items,
202 * protected by lustre_scrub::os_lock. */
203 struct list_head oii_list;
205 /* The right FID <=> ino#/gen mapping. */
206 struct osd_idmap_cache oii_cache;
208 unsigned int oii_insert:1; /* insert or update mapping. */
211 struct osd_otable_cache {
212 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
214 /* Index for next cache slot to be filled. */
215 int ooc_producer_idx;
217 /* Index for next cache slot to be returned by it::next(). */
218 int ooc_consumer_idx;
220 /* How many items in ooc_cache. */
221 __u64 ooc_cached_items;
223 /* Position for up layer LFSCK iteration pre-loading. */
224 __u64 ooc_pos_preload;
227 struct osd_otable_it {
228 struct osd_device *ooi_dev;
229 struct osd_otable_cache ooi_cache;
230 struct osd_iit_param ooi_iit_param;
232 /* The following bits can be updated/checked w/o lock protection.
233 * If more bits will be introduced in the future and need lock to
234 * protect, please add comment. */
235 unsigned long ooi_used_outside:1, /* Some user out of OSD
236 * uses the iteration. */
237 ooi_all_cached:1, /* No more entries can be
238 * filled into cache. */
239 ooi_user_ready:1, /* The user out of OSD is
240 * ready to iterate. */
241 ooi_waiting:1; /* it::next is waiting. */
244 struct osd_obj_orphan {
245 struct list_head oor_list;
246 struct lu_env *oor_env; /* to identify "own" records */
251 OSD_T10_TYPE_UNKNOWN = 0x00,
252 OSD_T10_TYPE1 = 0x01,
253 OSD_T10_TYPE2 = 0x02,
254 OSD_T10_TYPE3 = 0x04,
255 OSD_T10_TYPE_CRC = 0x08,
256 OSD_T10_TYPE_IP = 0x10,
257 OSD_T10_TYPE1_CRC = OSD_T10_TYPE1 | OSD_T10_TYPE_CRC,
258 OSD_T10_TYPE2_CRC = OSD_T10_TYPE2 | OSD_T10_TYPE_CRC,
259 OSD_T10_TYPE3_CRC = OSD_T10_TYPE3 | OSD_T10_TYPE_CRC,
260 OSD_T10_TYPE1_IP = OSD_T10_TYPE1 | OSD_T10_TYPE_IP,
261 OSD_T10_TYPE2_IP = OSD_T10_TYPE2 | OSD_T10_TYPE_IP,
262 OSD_T10_TYPE3_IP = OSD_T10_TYPE3 | OSD_T10_TYPE_IP,
270 struct dt_device od_dt_dev;
271 /* information about underlying file system */
272 struct vfsmount *od_mnt;
274 struct osd_oi **od_oi_table;
275 /* total number of OI containers */
280 unsigned int od_fl_capa:1,
289 od_writethrough_cache:1,
291 od_enable_projid_xattr:1,
295 __u32 od_dirent_journal;
297 struct proc_dir_entry *od_proc_entry;
298 struct lprocfs_stats *od_stats;
300 spinlock_t od_osfs_lock;
302 int od_fallocate_zero_blocks;
304 struct lu_site od_site;
306 struct osd_obj_map *od_ost_map;
307 struct osd_mdobj_map *od_mdt_map;
309 /* objects with size > od_readcache_max_filesize will be
310 * served bypassing pagecache unless already cached */
311 unsigned long long od_readcache_max_filesize;
313 /* reads >= od_readcache_max_iosize will be
314 * served bypassing pagecache unless already cached */
315 unsigned long od_readcache_max_iosize;
317 /* writes >= od_writethough_max_iosize will be
318 * served bypassing pagecache unless already cached */
319 unsigned long od_writethrough_max_iosize;
321 struct brw_stats od_brw_stats;
322 atomic_t od_r_in_flight;
323 atomic_t od_w_in_flight;
325 struct mutex od_otable_mutex;
326 struct osd_otable_it *od_otable_it;
327 struct osd_scrub od_scrub;
328 struct list_head od_ios_list;
330 /* service name associated with the osd device */
331 char od_svname[MAX_OBD_NAME];
332 char od_mntdev[MAX_OBD_NAME];
335 /* quota slave instance for inode */
336 struct qsd_instance *od_quota_slave_md;
338 /* quota slave instance for block */
339 struct qsd_instance *od_quota_slave_dt;
341 /* osd seq instance */
342 struct lu_client_seq *od_cl_seq;
343 /* If the ratio of "the total OI mappings count" vs
344 * "the bad OI mappings count" is lower than the
345 * osd_device::od_full_scrub_ratio, then trigger
346 * OI scrub to scan the whole the device. */
347 __u64 od_full_scrub_ratio;
348 /* If the speed of found bad OI mappings (per minute)
349 * exceeds the osd_device::od_full_scrub_threshold_rate,
350 * then trigger OI scrub to scan the whole device. */
351 __u64 od_full_scrub_threshold_rate;
353 /* a list of orphaned agent inodes, protected with od_osfs_lock */
354 struct list_head od_orphan_list;
355 struct list_head od_index_backup_list;
356 struct list_head od_index_restore_list;
358 struct inode *od_index_backup_inode;
359 enum lustre_index_backup_policy od_index_backup_policy;
360 int od_index_backup_stop;
361 /* T10PI type, zero if not supported */
362 enum osd_t10_type od_t10_type;
363 atomic_t od_commit_cb_in_flight;
364 wait_queue_head_t od_commit_cb_done;
365 unsigned int __percpu *od_extent_bytes_percpu;
368 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
371 return osd->od_quota_slave_dt;
373 return osd->od_quota_slave_md;
376 enum osd_full_scrub_ratio {
377 /* Trigger OI scrub to scan the whole device directly. */
380 /* Because the bad OI mappings count cannot be larger than
381 * the total OI mappints count, then setting OFSR_NEVER means
382 * that the whole device scanning cannot be triggered by auto
383 * detected bad OI mappings during the RPC services. */
385 OFSR_DEFAULT = 10000,
388 #define FULL_SCRUB_THRESHOLD_RATE_DEFAULT 60
390 /* There are at most 15 uid/gid/projids are affected in a transaction, and
391 * that's rename case:
392 * - 3 for source parent uid & gid & projid;
393 * - 3 for source child uid & gid & projid ('..' entry update when
394 * child is directory);
395 * - 3 for target parent uid & gid & projid;
396 * - 3 for target child uid & gid & projid(if the target child exists);
397 * - 3 for root uid & gid(last_rcvd, llog, etc);
400 #define OSD_MAX_UGID_CNT 15
418 struct osd_access_lock {
419 struct list_head tl_list;
420 struct osd_object *tl_obj;
430 struct thandle ot_super;
432 struct ldiskfs_journal_cb_entry ot_jcb;
433 struct list_head ot_commit_dcb_list;
434 struct list_head ot_stop_dcb_list;
435 /* Link to the device, for debugging. */
436 struct lu_ref_link ot_dev_link;
437 unsigned int ot_credits;
438 unsigned int oh_declared_ext;
440 /* quota IDs related to the transaction */
441 unsigned short ot_id_cnt;
442 __u8 ot_id_res[OSD_MAX_UGID_CNT];
443 __u8 ot_id_types[OSD_MAX_UGID_CNT];
444 uid_t ot_id_array[OSD_MAX_UGID_CNT];
445 struct lquota_trans *ot_quota_trans;
447 unsigned int ot_remove_agents:1;
448 #if OSD_THANDLE_STATS
449 /** time when this handle was allocated */
452 /** time when this thanle was started */
455 struct list_head ot_trunc_locks;
457 * list of declarations, used for large transactions to check
458 * for duplicates, like llogs
460 struct list_head ot_declare_list;
464 * Basic transaction credit op
481 struct osd_obj_declare {
482 struct list_head old_list;
483 struct lu_fid old_fid;
484 enum dt_txn_op old_op;
492 #ifdef CONFIG_PROC_FS
494 LPROC_OSD_READ_BYTES = 0,
495 LPROC_OSD_WRITE_BYTES = 1,
496 LPROC_OSD_GET_PAGE = 2,
497 LPROC_OSD_NO_PAGE = 3,
498 LPROC_OSD_CACHE_ACCESS = 4,
499 LPROC_OSD_CACHE_HIT = 5,
500 LPROC_OSD_CACHE_MISS = 6,
502 #if OSD_THANDLE_STATS
503 LPROC_OSD_THANDLE_STARTING,
504 LPROC_OSD_THANDLE_OPEN,
505 LPROC_OSD_THANDLE_CLOSING,
507 LPROC_OSD_TOO_MANY_CREDITS,
513 * Storage representation for fids.
515 * Variable size, first byte contains the length of the whole record.
517 struct osd_fid_pack {
518 unsigned char fp_len;
519 char fp_area[sizeof(struct lu_fid)];
522 struct osd_it_ea_dirent {
523 struct lu_fid oied_fid;
526 unsigned short oied_namelen;
527 unsigned int oied_type;
529 } __attribute__((packed));
532 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
533 * than lu_dirent struct. osd readdir reads less number of dirent than
534 * required for mdd dir page. so buffer size need to be increased so that
535 * there would be one ext3 readdir for every mdd readdir page.
538 #define OSD_IT_EA_BUFSIZE (PAGE_SIZE + PAGE_SIZE/4)
541 * This is iterator's in-memory data structure in interoperability
542 * mode (i.e. iterator over ldiskfs style directory)
545 struct osd_object *oie_obj;
546 /** used in ldiskfs iterator, to stored file pointer */
547 struct file *oie_file;
548 /** how many entries have been read-cached from storage */
550 /** current entry is being iterated by caller */
552 /** current processing entry */
553 struct osd_it_ea_dirent *oie_dirent;
554 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
559 * Iterator's in-memory data structure for IAM mode.
562 struct osd_object *oi_obj;
563 struct iam_path_descr *oi_ipd;
564 struct iam_iterator oi_it;
567 struct osd_quota_leaf {
568 struct list_head oql_link;
573 * Iterator's in-memory data structure for quota file.
575 struct osd_it_quota {
576 struct osd_object *oiq_obj;
577 /** tree blocks path to where the entry is stored */
578 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
579 /** on-disk offset for current key where quota record can be found */
581 /** identifier for current quota record */
583 /** the record index in the leaf/index block */
584 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
585 /** list of already processed leaf blocks */
586 struct list_head oiq_list;
589 #define MAX_BLOCKS_PER_PAGE (PAGE_SIZE / 512)
592 wait_queue_head_t dr_wait;
593 atomic_t dr_numreqs; /* number of reqs being processed */
596 unsigned int dr_pextents; /* number block extents */
597 unsigned int dr_lextents; /* number a logical extents */
600 unsigned int dr_init_at:16, /* the line iobuf was initialized */
601 dr_elapsed_valid:1, /* we really did count time */
604 struct niobuf_local **dr_lnbs;
605 struct lu_buf dr_bl_buf;
606 struct lu_buf dr_lnb_buf;
608 ktime_t dr_start_time;
609 ktime_t dr_elapsed; /* how long io took */
610 struct osd_device *dr_dev;
611 /* Already written blocks of the start page */
612 unsigned int dr_start_pg_wblks;
613 struct inode *dr_inode;
616 #define osd_dirty_inode(inode, flag) (inode)->i_sb->s_op->dirty_inode((inode), flag)
617 #define osd_i_blocks(inode, size) ((size) >> (inode)->i_blkbits)
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 };
635 * any object can be identified unique way:
637 * - inode + generation (unique object within a given filesystem)
639 * version increments each insert/delete in a directory
641 struct osd_lookup_cache_object {
642 struct osd_device *lco_osd;
648 struct osd_lookup_cache_entry {
649 struct osd_lookup_cache_object lce_obj;
650 struct lu_fid lce_fid;
653 char lce_name[LDISKFS_NAME_LEN + 1];
656 #define OSD_LOOKUP_CACHE_MAX 3
658 struct osd_lookup_cache {
659 struct osd_lookup_cache_entry olc_entry[OSD_LOOKUP_CACHE_MAX];
664 #define OSD_INS_CACHE_SIZE 8
666 struct osd_thread_info {
667 const struct lu_env *oti_env;
669 * used for index operations.
671 struct dentry oti_obj_dentry;
672 struct dentry oti_child_dentry;
674 /** dentry for Iterator context. */
675 struct dentry oti_it_dentry;
678 struct kstatfs oti_ksfs;
680 struct htree_lock *oti_hlock;
682 struct lu_fid oti_fid;
683 struct lu_fid oti_fid2;
684 struct lu_fid oti_fid3;
685 struct osd_inode_id oti_id;
686 struct osd_inode_id oti_id2;
687 struct osd_inode_id oti_id3;
688 struct ost_id oti_ostid;
691 * following ipd and it structures are used for osd_index_iam_lookup()
692 * these are defined separately as we might do index operation
693 * in open iterator session.
696 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
698 unsigned int oti_it_ea_buf_used:1;
700 /* IAM iterator for index operation. */
701 struct iam_iterator oti_idx_it;
703 /** union to guarantee that ->oti_ipd[] has proper alignment. */
706 char oti_it_ipd[DX_IPD_MAX_SIZE];
707 long long oti_alignment_lieutenant;
711 char oti_idx_ipd[DX_IPD_MAX_SIZE];
712 long long oti_alignment_lieutenant_colonel;
715 struct osd_idmap_cache oti_cache;
717 /* dedicated OI cache for insert (which needs inum) */
718 struct osd_idmap_cache *oti_ins_cache;
719 int oti_ins_cache_size;
720 int oti_ins_cache_used;
721 /* inc by osd_trans_create and dec by osd_trans_stop */
722 int oti_ins_cache_depth;
727 /** used in osd_fid_set() to put xattr */
728 struct lu_buf oti_buf;
729 struct lu_buf oti_big_buf;
730 /** used in osd_ea_fid_set() to set fid into common ea */
732 struct lustre_ost_attrs oti_ost_attrs;
733 struct filter_fid_18_23 oti_ff_old;
734 struct filter_fid oti_ff;
737 struct osd_iobuf oti_iobuf;
738 /* used to access objects in /O */
739 struct inode *oti_inode;
740 #define OSD_FID_REC_SZ 32
741 char oti_ldp[OSD_FID_REC_SZ];
742 char oti_ldp2[OSD_FID_REC_SZ];
744 /* used by quota code */
746 #if defined(HAVE_DQUOT_QC_DQBLK)
747 struct qc_dqblk oti_qdq;
749 struct fs_disk_quota oti_fdq;
751 struct if_dqinfo oti_dqinfo;
753 struct lquota_id_info oti_qi;
754 struct lquota_trans oti_quota_trans;
755 union lquota_rec oti_quota_rec;
757 struct lu_seq_range oti_seq_range;
759 /* Tracking for transaction credits, to allow debugging and optimizing
760 * cases where a large number of credits are being allocated for
761 * single transaction. */
762 unsigned int oti_credits_before;
763 unsigned int oti_declare_ops[OSD_OT_MAX];
764 unsigned int oti_declare_ops_cred[OSD_OT_MAX];
765 unsigned int oti_declare_ops_used[OSD_OT_MAX];
766 struct osd_directory oti_iam;
768 struct page **oti_dio_pages;
769 int oti_dio_pages_used;
771 struct osd_it_ea_dirent *oti_seq_dirent;
772 struct osd_it_ea_dirent *oti_dir_dirent;
774 struct osd_lookup_cache_object oti_cobj; /* cache object id */
775 struct osd_lookup_cache *oti_lookup_cache;
778 extern int ldiskfs_pdo;
780 /* autoconf test is in lustre-build-ldiskfs.m4 */
781 #ifdef HAVE_BVEC_ITER_ALL
782 #define DECLARE_BVEC_ITER_ALL(iter) struct bvec_iter_all iter
784 #define DECLARE_BVEC_ITER_ALL(iter) int iter
787 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
788 const char *name, void *buf, int len)
793 dentry->d_inode = inode;
794 dentry->d_sb = inode->i_sb;
795 return ll_vfs_getxattr(dentry, inode, name, buf, len);
798 static inline int __osd_xattr_set(struct osd_thread_info *info,
799 struct inode *inode, const char *name,
800 const void *buf, int buflen, int fl)
802 struct dentry *dentry = &info->oti_child_dentry;
804 dquot_initialize(inode);
805 dentry->d_inode = inode;
806 dentry->d_sb = inode->i_sb;
807 return ll_vfs_setxattr(dentry, inode, name, buf, buflen, fl);
810 static inline char *osd_oid_name(char *name, size_t name_size,
811 const struct lu_fid *fid, u64 id)
813 snprintf(name, name_size,
814 (fid_seq_is_rsvd(fid_seq(fid)) ||
815 fid_seq_is_mdt0(fid_seq(fid)) ||
816 fid_seq_is_idif(fid_seq(fid))) ? "%llu" : "%llx", id);
821 #ifdef CONFIG_PROC_FS
823 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
824 int osd_procfs_init(struct osd_device *osd, const char *name);
825 int osd_procfs_fini(struct osd_device *osd);
826 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
827 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 0, 52, 0)
828 int osd_register_proc_index_in_idif(struct osd_device *osd);
832 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
833 struct obd_statfs *sfs, struct obd_statfs_info *info);
834 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
835 struct osd_inode_id *id);
837 osd_iget_fid(struct osd_thread_info *info, struct osd_device *dev,
838 struct osd_inode_id *id, struct lu_fid *fid);
839 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
840 const struct lu_fid *fid, __u32 compat, __u32 incompat);
841 struct osd_obj_seq *osd_seq_load(struct osd_thread_info *info,
842 struct osd_device *osd, u64 seq);
843 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
844 struct dentry *dentry, struct lustre_ost_attrs *loa);
845 void osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
846 struct osd_inode_id *id, const struct lu_fid *fid);
847 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
848 struct dentry *dentry, struct lu_fid *fid);
849 struct osd_it_ea *osd_it_dir_init(const struct lu_env *env,
850 struct osd_device *dev, struct inode *inode,
852 void osd_it_dir_fini(const struct lu_env *env, struct osd_it_ea *oie,
853 struct inode *inode);
854 int osd_ldiskfs_it_fill(const struct lu_env *env, const struct dt_it *di);
856 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
857 void osd_obj_map_fini(struct osd_device *dev);
858 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
859 const struct lu_fid *fid, struct osd_inode_id *id);
860 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
861 const struct lu_fid *fid, const struct osd_inode_id *id,
863 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
864 const struct lu_fid *fid, handle_t *th);
865 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
866 const struct lu_fid *fid, const struct osd_inode_id *id,
868 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
869 struct inode *src_parent, struct dentry *src_child,
870 const struct lu_fid *fid);
871 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
872 const struct lu_fid *fid, struct osd_inode_id *id,
873 enum oi_check_flags flags);
874 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
875 const struct lu_fid *fid, const struct osd_inode_id *id,
877 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
878 const struct lu_fid *fid, const struct osd_inode_id *id,
880 int osd_obj_del_entry(struct osd_thread_info *info, struct osd_device *osd,
881 struct dentry *dird, char *name, int namelen,
884 char *osd_lf_fid2name(const struct lu_fid *fid);
885 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
887 void osd_scrub_stop(struct osd_device *dev);
888 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev,
890 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
891 int osd_scrub_oi_insert(struct osd_device *dev, const struct lu_fid *fid,
892 struct osd_inode_id *id, int insert);
893 bool osd_scrub_oi_resurrect(struct lustre_scrub *scrub,
894 const struct lu_fid *fid);
895 void osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
897 struct dentry *osd_lookup_one_len_unlocked(struct osd_device *dev,
899 struct dentry *base, int len);
900 struct dentry *osd_lookup_one_len(struct osd_device *dev, const char *name,
901 struct dentry *base, int len);
903 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
904 u64 seq, struct lu_seq_range *range);
906 int osd_delete_from_remote_parent(const struct lu_env *env,
907 struct osd_device *osd,
908 struct osd_object *obj,
909 struct osd_thandle *oh, bool destroy);
910 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
911 struct osd_object *obj, struct osd_thandle *oh);
912 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
913 struct osd_device *osd,
914 const struct lu_fid *fid,
915 struct osd_inode_id *id);
917 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
919 int osd_scrub_refresh_mapping(struct osd_thread_info *info,
920 struct osd_device *dev,
921 const struct lu_fid *fid,
922 const struct osd_inode_id *id,
924 enum oi_check_flags flags, bool *exist);
926 /* osd_quota_fmt.c */
927 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
928 int type, uint blk, int depth, uint index,
929 struct osd_it_quota *it);
930 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
931 int type, uint blk, uint index,
932 struct osd_it_quota *it);
933 loff_t find_tree_dqentry(const struct lu_env *env,
934 struct osd_object *obj, int type,
935 qid_t dqid, uint blk, int depth,
936 struct osd_it_quota *it);
938 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
939 struct lquota_id_info *qi, struct osd_object *obj,
940 bool enforce, enum osd_quota_local_flags *local_flags);
941 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
942 __u32 projid, long long space, struct osd_thandle *oh,
943 struct osd_object *obj,
944 enum osd_quota_local_flags *local_flags,
945 enum osd_qid_declare_flags);
946 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
947 const struct dt_rec *rec,
948 union lquota_rec *quota_rec);
949 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
951 #ifdef HAVE_PROJECT_QUOTA
952 static inline __u32 i_projid_read(struct inode *inode)
954 return (__u32)from_kprojid(&init_user_ns, LDISKFS_I(inode)->i_projid);
957 static inline void i_projid_write(struct inode *inode, __u32 projid)
960 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
961 LDISKFS_I(inode)->i_projid = kprojid;
964 static inline uid_t i_projid_read(struct inode *inode)
968 static inline void i_projid_write(struct inode *inode, __u32 projid)
974 #ifdef HAVE_LDISKFS_IGET_WITH_FLAGS
975 # define osd_ldiskfs_iget(sb, ino) \
976 ldiskfs_iget((sb), (ino), \
977 LDISKFS_IGET_HANDLE | LDISKFS_IGET_SPECIAL)
979 # define osd_ldiskfs_iget(sb, ino) ldiskfs_iget((sb), (ino))
982 #ifdef HAVE_LDISKFS_INFO_JINODE
983 # define osd_attach_jinode(inode) ldiskfs_inode_attach_jinode(inode)
984 #else /* HAVE_LDISKFS_INFO_JINODE */
985 # define osd_attach_jinode(inode) 0
986 #endif /* HAVE_LDISKFS_INFO_JINODE */
988 #ifndef HAVE_JBD2_JOURNAL_GET_MAX_TXN_BUFS
989 #define jbd2_journal_get_max_txn_bufs(jrnl) \
990 (jrnl->j_max_transaction_buffers)
993 #ifdef LDISKFS_HT_MISC
994 # define osd_journal_start_sb(sb, type, nblock) \
995 ldiskfs_journal_start_sb(sb, type, nblock)
996 #ifdef HAVE_LDISKFS_JOURNAL_ENSURE_CREDITS
997 # define osd_journal_start_with_revoke(sb, type, nblock, revoke) \
998 ldiskfs_journal_start_with_revoke((sb)->s_root->d_inode, type, \
1001 # define osd_journal_start_with_revoke(sb, type, nblock, revoke) \
1002 ldiskfs_journal_start_sb(sb, type, nblock)
1004 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
1005 struct inode *inode,
1006 ldiskfs_lblk_t *nblock)
1010 rc = osd_attach_jinode(inode);
1013 return ldiskfs_append(handle, inode, nblock);
1016 # ifdef HAVE___LDISKFS_FIND_ENTRY
1017 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1018 (__ldiskfs_find_entry(dir, name, de, inlined, lock) ?: \
1021 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1022 (ldiskfs_find_entry_locked(dir, name, de, inlined, lock) ?: \
1026 # define osd_journal_start(inode, type, nblocks) \
1027 ldiskfs_journal_start(inode, type, nblocks)
1028 # define osd_transaction_size(dev) \
1029 (jbd2_journal_get_max_txn_bufs(osd_journal(dev)) / 2)
1030 #else /* ! defined LDISKFS_HT_MISC */
1031 # define LDISKFS_HT_MISC 0
1032 # define osd_journal_start_sb(sb, type, nblock) \
1033 ldiskfs_journal_start_sb(sb, nblock)
1035 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
1036 struct inode *inode,
1037 ldiskfs_lblk_t *nblock)
1039 struct buffer_head *bh;
1042 bh = ldiskfs_append(handle, inode, nblock, &err);
1049 # ifdef HAVE___LDISKFS_FIND_ENTRY
1050 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1051 (__ldiskfs_find_entry(dir, name, de, lock) ?: \
1054 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1055 (ldiskfs_find_entry_locked(dir, name, de, lock) ?: \
1058 # define osd_journal_start(inode, type, nblocks) \
1059 ldiskfs_journal_start(inode, nblocks)
1060 # define osd_transaction_size(dev) \
1061 jbd2_journal_get_max_txn_bufs(osd_journal(dev))
1062 #endif /* LDISKFS_HT_MISC */
1064 #ifndef HAVE___LDISKFS_FIND_ENTRY
1065 # define __ldiskfs_add_entry(handle, child, inode, hlock) \
1066 ldiskfs_add_entry_locked(handle, child, inode, hlock)
1070 * Invariants, assertions.
1074 * XXX: do not enable this, until invariant checking code is made thread safe
1075 * in the face of pdirops locking.
1077 #define OSD_INVARIANT_CHECKS (0)
1079 #if OSD_INVARIANT_CHECKS
1080 static inline int osd_invariant(const struct osd_object *obj)
1084 ergo(obj->oo_inode != NULL,
1085 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
1086 atomic_read(&obj->oo_inode->i_count) > 0) &&
1087 ergo(obj->oo_dir != NULL &&
1088 obj->oo_dir->od_conationer.ic_object != NULL,
1089 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
1092 #define osd_invariant(obj) (1)
1095 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
1096 #define OSD_READCACHE_MAX_IO_MB 8
1097 #define OSD_WRITECACHE_MAX_IO_MB 8
1099 extern const struct dt_index_operations osd_otable_ops;
1101 static inline int osd_oi_fid2idx(struct osd_device *dev,
1102 const struct lu_fid *fid)
1104 return fid->f_seq & (dev->od_oi_count - 1);
1107 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
1108 const struct lu_fid *fid)
1110 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
1111 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
1112 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
1113 DFID"\n", PFID(fid));
1114 /* It can work even od_oi_count equals to 1 although it's unexpected,
1115 * the only reason we set it to 1 is for performance measurement */
1116 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
1119 extern const struct lu_device_operations osd_lu_ops;
1121 static inline int lu_device_is_osd(const struct lu_device *d)
1123 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
1126 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
1128 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
1129 return container_of(d, struct osd_device, od_dt_dev);
1132 static inline struct osd_device *osd_dev(const struct lu_device *d)
1134 LASSERT(lu_device_is_osd(d));
1135 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
1138 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
1140 return osd_dev(o->oo_dt.do_lu.lo_dev);
1143 static inline struct super_block *osd_sb(const struct osd_device *dev)
1145 return dev->od_mnt->mnt_sb;
1148 static inline const char *osd_sb2name(const struct super_block *sb)
1150 /* this is LDISKFS_SB(sb), but preserves "const" */
1151 const struct ldiskfs_sb_info *sbi = sb->s_fs_info;
1153 return sbi->s_es->s_volume_name;
1156 static inline const char *osd_dev2name(const struct osd_device *dev)
1158 return osd_sb2name(osd_sb(dev));
1161 static inline const char *osd_ino2name(const struct inode *inode)
1163 return osd_sb2name(inode->i_sb);
1167 * Put the osd object once done with it.
1169 * \param obj osd object that needs to be put
1171 static inline void osd_object_put(const struct lu_env *env,
1172 struct osd_object *obj)
1174 dt_object_put(env, &obj->oo_dt);
1177 static inline int osd_object_is_root(const struct osd_object *obj)
1179 return obj->oo_inode && is_root_inode(obj->oo_inode);
1182 static inline struct osd_object *osd_obj(const struct lu_object *o)
1184 LASSERT(lu_device_is_osd(o->lo_dev));
1185 return container_of(o, struct osd_object, oo_dt.do_lu);
1188 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
1190 return osd_obj(&d->do_lu);
1193 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
1195 return &osd->od_dt_dev.dd_lu_dev;
1198 static inline journal_t *osd_journal(const struct osd_device *dev)
1200 return LDISKFS_SB(osd_sb(dev))->s_journal;
1203 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
1205 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
1208 static inline char *osd_name(struct osd_device *osd)
1210 return osd->od_svname;
1213 static inline bool osd_is_ea_inode(struct inode *inode)
1215 return !!(LDISKFS_I(inode)->i_flags & LDISKFS_EA_INODE_FL);
1218 extern const struct dt_body_operations osd_body_ops;
1219 extern struct lu_context_key osd_key;
1221 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
1223 return lu_context_key_get(&env->le_ctx, &osd_key);
1226 extern const struct dt_body_operations osd_body_ops_new;
1232 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
1233 const struct iam_container *bag)
1235 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1236 osd_oti_get(env)->oti_it_ipd);
1240 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
1241 const struct iam_container *bag)
1243 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1244 osd_oti_get(env)->oti_idx_ipd);
1247 static inline void osd_ipd_put(const struct lu_env *env,
1248 const struct iam_container *bag,
1249 struct iam_path_descr *ipd)
1251 bag->ic_descr->id_ops->id_ipd_free(ipd);
1254 int osd_calc_bkmap_credits(struct super_block *sb, struct inode *inode,
1255 const loff_t size, const loff_t pos,
1258 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
1261 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
1262 struct inode *inode,
1263 const char *name, const int namelen)
1265 struct osd_thread_info *info = osd_oti_get(env);
1266 struct dentry *child_dentry = &info->oti_child_dentry;
1267 struct dentry *obj_dentry = &info->oti_obj_dentry;
1269 obj_dentry->d_inode = inode;
1270 obj_dentry->d_sb = inode->i_sb;
1271 obj_dentry->d_name.hash = 0;
1273 child_dentry->d_name.hash = 0;
1274 child_dentry->d_parent = obj_dentry;
1275 child_dentry->d_name.name = name;
1276 child_dentry->d_name.len = namelen;
1277 return child_dentry;
1280 extern int osd_trans_declare_op2rb[];
1281 extern int ldiskfs_track_declares_assert;
1282 void osd_trans_dump_creds(const struct lu_env *env, struct thandle *th);
1284 static inline void osd_trans_declare_op(const struct lu_env *env,
1285 struct osd_thandle *oh,
1286 enum osd_op_type op, int credits)
1288 struct osd_thread_info *oti = osd_oti_get(env);
1290 LASSERT(oh->ot_handle == NULL);
1291 if (unlikely(op >= OSD_OT_MAX)) {
1292 if (unlikely(ldiskfs_track_declares_assert)) {
1293 LASSERT(op < OSD_OT_MAX);
1295 CWARN("%s: Invalid operation index %d\n",
1296 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op);
1300 oti->oti_declare_ops[op]++;
1301 oti->oti_declare_ops_cred[op] += credits;
1303 oh->ot_credits += credits;
1306 /* linux: v5.4-rc3-21-g933f1c1e0b75 renamed h_buffer_credits */
1307 #ifdef HAVE_JOURNAL_TOTAL_CREDITS
1308 #define h_buffer_credits h_total_credits
1311 static inline void osd_trans_exec_op(const struct lu_env *env,
1313 enum osd_op_type op)
1315 struct osd_thread_info *oti = osd_oti_get(env);
1316 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1318 unsigned int rb, left;
1320 LASSERT(oh->ot_handle != NULL);
1321 if (unlikely(op >= OSD_OT_MAX)) {
1322 if (unlikely(ldiskfs_track_declares_assert))
1323 LASSERT(op < OSD_OT_MAX);
1325 CWARN("%s: opcode %u: invalid value >= %u\n",
1326 osd_name(osd_dt_dev(oh->ot_super.th_dev)),
1333 /* find rollback (or reverse) operation for the given one
1334 * such an operation doesn't require additional credits
1335 * as the same set of blocks are modified */
1336 rb = osd_trans_declare_op2rb[op];
1338 /* check whether credits for this operation were reserved at all */
1339 if (unlikely(oti->oti_declare_ops_cred[op] == 0 &&
1340 oti->oti_declare_ops_cred[rb] == 0)) {
1341 /* the API is not perfect yet: CREATE does REF_ADD internally
1342 * while DESTROY does not. To rollback CREATE the callers
1343 * needs to call REF_DEL+DESTROY which is hard to detect using
1344 * a simple table of rollback operations */
1345 if (op == OSD_OT_REF_DEL &&
1346 oti->oti_declare_ops_cred[OSD_OT_CREATE] > 0)
1348 if (op == OSD_OT_REF_ADD &&
1349 oti->oti_declare_ops_cred[OSD_OT_DESTROY] > 0)
1351 CWARN("%s: opcode %u: credits = 0, rollback = %u\n",
1352 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
1353 osd_trans_dump_creds(env, th);
1354 LASSERT(!ldiskfs_track_declares_assert);
1358 /* remember how many credits we have unused before the operation */
1359 oti->oti_credits_before = oh->ot_handle->h_buffer_credits;
1360 left = oti->oti_declare_ops_cred[op] - oti->oti_declare_ops_used[op];
1361 if (unlikely(oti->oti_credits_before < left)) {
1362 CWARN("%s: opcode %u: before %u < left %u, rollback = %u\n",
1363 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1364 oti->oti_credits_before, left, rb);
1365 osd_trans_dump_creds(env, th);
1366 /* on a very small fs (testing?) it's possible that
1367 * the transaction can't fit 1/4 of journal, so we
1368 * just request less credits (see osd_trans_start()).
1369 * ignore the same case here */
1370 rb = osd_transaction_size(osd_dt_dev(th->th_dev));
1371 if (unlikely(oh->ot_credits < rb))
1372 LASSERT(!ldiskfs_track_declares_assert);
1376 static inline void osd_trans_exec_check(const struct lu_env *env,
1378 enum osd_op_type op)
1380 struct osd_thread_info *oti = osd_oti_get(env);
1381 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1383 int used, over, quota;
1385 /* how many credits have been used by the operation */
1386 used = oti->oti_credits_before - oh->ot_handle->h_buffer_credits;
1388 if (unlikely(used < 0)) {
1389 /* if some block was allocated and released in the same
1390 * transaction, then it won't be a part of the transaction
1391 * and delta can be negative */
1396 /* rollback operations (e.g. when we destroy just created
1397 * object) should not consume any credits. there is no point
1398 * to confuse the checks below */
1402 oti->oti_declare_ops_used[op] += used;
1403 if (oti->oti_declare_ops_used[op] <= oti->oti_declare_ops_cred[op])
1406 /* we account quota for a whole transaction and any operation can
1407 * consume corresponding credits */
1408 over = oti->oti_declare_ops_used[op] -
1409 oti->oti_declare_ops_cred[op];
1410 quota = oti->oti_declare_ops_cred[OSD_OT_QUOTA] -
1411 oti->oti_declare_ops_used[OSD_OT_QUOTA];
1412 if (over <= quota) {
1413 /* probably that credits were consumed by
1414 * quota indirectly (in the depths of ldiskfs) */
1415 oti->oti_declare_ops_used[OSD_OT_QUOTA] += over;
1416 oti->oti_declare_ops_used[op] -= over;
1418 CWARN("%s: opcode %d: used %u, used now %u, reserved %u\n",
1419 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1420 oti->oti_declare_ops_used[op], used,
1421 oti->oti_declare_ops_cred[op]);
1422 osd_trans_dump_creds(env, th);
1423 if (unlikely(ldiskfs_track_declares_assert))
1429 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1432 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1433 struct lu_fid *befider)
1435 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1436 memcpy(pack->fp_area, befider, sizeof(*befider));
1437 pack->fp_len = sizeof(*befider) + 1;
1441 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1446 switch (pack->fp_len) {
1447 case sizeof *fid + 1:
1448 memcpy(fid, pack->fp_area, sizeof *fid);
1449 fid_be_to_cpu(fid, fid);
1452 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1459 * Quota/Accounting handling
1461 extern const struct dt_index_operations osd_acct_index_ops;
1462 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1463 const struct lu_fid *fid, struct osd_inode_id *id);
1465 /* copy from fs/ext4/dir.c */
1466 static inline int is_32bit_api(void)
1468 #ifdef CONFIG_COMPAT
1469 return in_compat_syscall();
1471 return (BITS_PER_LONG == 32);
1475 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1477 if ((filp->f_mode & FMODE_32BITHASH) ||
1478 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1479 return LDISKFS_HTREE_EOF_32BIT;
1481 return LDISKFS_HTREE_EOF_64BIT;
1484 static inline int fid_is_internal(const struct lu_fid *fid)
1486 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1489 static inline bool is_remote_parent_ino(struct osd_device *o, unsigned long ino)
1494 LASSERT(o->od_mdt_map != NULL);
1496 return ino == o->od_mdt_map->omm_remote_parent->d_inode->i_ino;
1500 * ext4_bread/ldiskfs_bread has either 5 or 4 parameters. The error
1501 * return code has been removed and integrated into the pointer in the
1504 static inline struct buffer_head *__ldiskfs_bread(handle_t *handle,
1505 struct inode *inode,
1506 ldiskfs_lblk_t block,
1510 struct buffer_head *bh;
1513 rc = osd_attach_jinode(inode);
1517 #ifdef HAVE_EXT4_BREAD_4ARGS
1518 bh = ldiskfs_bread(handle, inode, block, create);
1521 bh = ldiskfs_bread(handle, inode, block, create, &rc);
1522 if (bh == NULL && rc != 0)
1528 #ifdef HAVE_EXT4_JOURNAL_GET_WRITE_ACCESS_4ARGS
1529 # define osd_ldiskfs_journal_get_write_access(handle, sb, bh, flags) \
1530 ldiskfs_journal_get_write_access((handle), (sb), (bh), (flags))
1532 # define LDISKFS_JTR_NONE 0
1533 # define osd_ldiskfs_journal_get_write_access(handle, sb, bh, flags) \
1534 ldiskfs_journal_get_write_access((handle), (bh))
1535 #endif /* HAVE_EXT4_JOURNAL_GET_WRITE_ACCESS_4ARGS */
1537 #ifdef HAVE_EXT4_INC_DEC_COUNT_2ARGS
1538 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((h), (inode))
1539 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((h), (inode))
1541 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((inode))
1542 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((inode))
1543 #endif /* HAVE_EXT4_INC_DEC_COUNT_2ARGS */
1545 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1548 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1549 __u32 keysize, __u32 recsize)
1551 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1552 &osd->od_index_backup_list, &osd->od_lock,
1553 &osd->od_index_backup_stop,
1554 fid, keysize, recsize);
1558 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1560 struct osd_thread_info *info = osd_oti_get(env);
1561 struct lu_fid *fid = &info->oti_fid3;
1562 struct osd_inode_id *id = &info->oti_id3;
1564 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1565 osd_id_gen(id, osd->od_index_backup_inode->i_ino,
1566 osd->od_index_backup_inode->i_generation);
1567 osd_add_oi_cache(info, osd, id, fid);
1569 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1570 &osd->od_index_backup_list, &osd->od_lock,
1571 &osd->od_index_backup_stop, backup);
1574 #ifdef LDISKFS_HAS_INCOMPAT_FEATURE
1576 # ifdef LDISKFS_FEATURE_INCOMPAT_EXTENTS
1577 # define ldiskfs_has_feature_extents(sb) \
1578 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EXTENTS)
1580 # ifdef LDISKFS_FEATURE_INCOMPAT_EA_INODE
1581 # define ldiskfs_has_feature_ea_inode(sb) \
1582 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EA_INODE)
1584 # ifdef LDISKFS_FEATURE_INCOMPAT_DIRDATA
1585 # define ldiskfs_has_feature_dirdata(sb) \
1586 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_DIRDATA)
1588 # ifdef LDISKFS_FEATURE_INCOMPAT_LARGEDIR
1589 # define ldiskfs_has_feature_largedir(sb) \
1590 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1591 # define ldiskfs_set_feature_largedir(sb) \
1592 LDISKFS_SET_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1594 # ifdef LDISKFS_FEATURE_COMPAT_HAS_JOURNAL
1595 # define ldiskfs_has_feature_journal(sb) \
1596 LDISKFS_HAS_COMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_HAS_JOURNAL)
1598 # ifdef LDISKFS_FEATURE_RO_COMPAT_QUOTA
1599 # define ldiskfs_has_feature_quota(sb) \
1600 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_QUOTA)
1602 # ifdef LDISKFS_FEATURE_RO_COMPAT_PROJECT
1603 # define ldiskfs_has_feature_project(sb) \
1604 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_PROJECT)
1606 # ifdef LDISKFS_FEATURE_COMPAT_FAST_COMMIT
1607 # define ldiskfs_has_feature_fast_commit(sb) \
1608 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_FAST_COMMIT)
1612 #ifndef ldiskfs_has_feature_fast_commit
1613 #define ldiskfs_has_feature_fast_commit(sb) false
1616 int osd_trunc_lock(struct osd_object *obj, struct osd_thandle *oh,
1618 void osd_trunc_unlock_all(const struct lu_env *env, struct list_head *list);
1619 int osd_process_truncates(const struct lu_env *env, struct list_head *list);
1620 void osd_execute_truncate(struct osd_object *obj);
1622 #ifndef HAVE___BI_CNT
1623 #define __bi_cnt bi_cnt
1626 #ifndef HAVE_CLEAN_BDEV_ALIASES
1627 #define clean_bdev_aliases(bdev, block, len) \
1628 unmap_underlying_metadata((bdev), (block))
1631 #ifndef HAVE_BI_STATUS
1632 #define bi_status bi_error
1636 * Maximum size of xattr attributes for FEATURE_INCOMPAT_EA_INODE 1Mb
1637 * This limit is arbitrary, but is reasonable for the xattr API.
1639 #define LDISKFS_XATTR_MAX_LARGE_EA_SIZE (1024 * 1024)
1641 struct osd_bio_private {
1642 struct osd_iobuf *obp_iobuf;
1643 /* Start page index in the obp_iobuf for the bio */
1644 int obp_start_page_idx;
1646 extern struct kmem_cache *biop_cachep;
1648 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && defined(HAVE_BIO_INTEGRITY_PREP_FN)
1649 int osd_bio_integrity_handle(struct osd_device *osd, struct bio *bio,
1650 struct osd_iobuf *iobuf);
1651 #else /* !CONFIG_BLK_DEV_INTEGRITY */
1652 #define osd_bio_integrity_handle(osd, bio, iobuf) (0)
1655 #ifdef HAVE_BIO_INTEGRITY_PREP_FN
1656 # ifdef HAVE_BLK_INTEGRITY_ITER
1657 # define integrity_gen_fn integrity_processing_fn
1658 # define integrity_vrfy_fn integrity_processing_fn
1660 int osd_get_integrity_profile(struct osd_device *osd,
1661 integrity_gen_fn **generate_fn,
1662 integrity_vrfy_fn **verify_fn);
1663 #endif /* HAVE_BIO_INTEGRITY_PREP_FN */
1665 #ifdef HAVE_BIO_BI_PHYS_SEGMENTS
1666 #define osd_bio_nr_segs(bio) ((bio)->bi_phys_segments)
1668 #define osd_bio_nr_segs(bio) bio_segments((bio))
1669 #endif /* HAVE_BIO_BI_PHYS_SEGMENTS */
1671 #ifdef HAVE_GET_INODE_USAGE
1672 #define lock_dquot_transfer(inode) down_read(&LDISKFS_I(inode)->xattr_sem)
1673 #define unlock_dquot_transfer(inode) up_read(&LDISKFS_I(inode)->xattr_sem)
1675 #define lock_dquot_transfer(inode) do {} while (0)
1676 #define unlock_dquot_transfer(inode) do {} while (0)
1679 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)
1680 static inline unsigned short blk_integrity_interval(struct blk_integrity *bi)
1682 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1683 return bi->interval_exp ? 1 << bi->interval_exp : 0;
1684 #elif defined(HAVE_INTERVAL_BLK_INTEGRITY)
1685 return bi->interval;
1687 return bi->sector_size;
1688 #endif /* !HAVE_INTERVAL_EXP_BLK_INTEGRITY */
1691 static inline const char *blk_integrity_name(struct blk_integrity *bi)
1693 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1694 return bi->profile->name;
1700 static inline unsigned int bip_size(struct bio_integrity_payload *bip)
1702 #ifdef HAVE_BIP_ITER_BIO_INTEGRITY_PAYLOAD
1703 return bip->bip_iter.bi_size;
1705 return bip->bip_size;
1708 #else /* !CONFIG_BLK_DEV_INTEGRITY */
1709 static inline unsigned short blk_integrity_interval(struct blk_integrity *bi)
1713 static inline const char *blk_integrity_name(struct blk_integrity *bi)
1715 /* gcc8 dislikes when strcmp() is called against NULL */
1718 #endif /* !CONFIG_BLK_DEV_INTEGRITY */
1720 #ifndef INTEGRITY_FLAG_READ
1721 #define INTEGRITY_FLAG_READ BLK_INTEGRITY_VERIFY
1724 #ifndef INTEGRITY_FLAG_WRITE
1725 #define INTEGRITY_FLAG_WRITE BLK_INTEGRITY_GENERATE
1728 static inline bool bdev_integrity_enabled(struct block_device *bdev, int rw)
1730 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)
1731 struct blk_integrity *bi = bdev_get_integrity(bdev);
1736 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1737 if (rw == 0 && bi->profile->verify_fn != NULL &&
1738 (bi->flags & INTEGRITY_FLAG_READ))
1741 if (rw == 1 && bi->profile->generate_fn != NULL &&
1742 (bi->flags & INTEGRITY_FLAG_WRITE))
1745 if (rw == 0 && bi->verify_fn != NULL &&
1746 (bi->flags & INTEGRITY_FLAG_READ))
1749 if (rw == 1 && bi->generate_fn != NULL &&
1750 (bi->flags & INTEGRITY_FLAG_WRITE))
1752 #endif /* !HAVE_INTERVAL_EXP_BLK_INTEGRITY */
1753 #endif /* !CONFIG_BLK_DEV_INTEGRITY */
1758 bool osd_tx_was_declared(const struct lu_env *env, struct osd_thandle *oth,
1759 struct dt_object *dt, enum dt_txn_op op, loff_t p);
1761 #ifdef HAVE_DQUOT_TRANSFER_WITH_USER_NS
1762 #define osd_dquot_transfer(ns, i, a) dquot_transfer((ns), (i), (a))
1764 #define osd_dquot_transfer(ns, i, a) dquot_transfer((i), (a))
1767 #ifdef HAVE_FILLDIR_USE_CTX_RETURN_BOOL
1768 #define WRAP_FILLDIR_FN(prefix, fill_fn) \
1769 static bool fill_fn(struct dir_context *buf, const char *name, int namelen, \
1770 loff_t offset, __u64 ino, unsigned int d_type) \
1772 return !prefix##fill_fn(buf, name, namelen, offset, ino, d_type); \
1774 #elif defined(HAVE_FILLDIR_USE_CTX)
1775 #define WRAP_FILLDIR_FN(prefix, fill_fn) \
1776 static int fill_fn(struct dir_context *buf, const char *name, int namelen, \
1777 loff_t offset, __u64 ino, unsigned int d_type) \
1779 return prefix##fill_fn(buf, name, namelen, offset, ino, d_type); \
1782 #define WRAP_FILLDIR_FN(prefix, fill_fn)
1785 #endif /* _OSD_INTERNAL_H */