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;
294 __u32 od_dirent_journal;
296 struct proc_dir_entry *od_proc_entry;
297 struct lprocfs_stats *od_stats;
299 spinlock_t od_osfs_lock;
301 int od_fallocate_zero_blocks;
303 struct lu_site od_site;
305 struct osd_obj_map *od_ost_map;
306 struct osd_mdobj_map *od_mdt_map;
308 /* objects with size > od_readcache_max_filesize will be
309 * served bypassing pagecache unless already cached */
310 unsigned long long od_readcache_max_filesize;
312 /* reads >= od_readcache_max_iosize will be
313 * served bypassing pagecache unless already cached */
314 unsigned long od_readcache_max_iosize;
316 /* writes >= od_writethough_max_iosize will be
317 * served bypassing pagecache unless already cached */
318 unsigned long od_writethrough_max_iosize;
320 struct brw_stats od_brw_stats;
321 atomic_t od_r_in_flight;
322 atomic_t od_w_in_flight;
324 struct mutex od_otable_mutex;
325 struct osd_otable_it *od_otable_it;
326 struct osd_scrub od_scrub;
327 struct list_head od_ios_list;
329 /* service name associated with the osd device */
330 char od_svname[MAX_OBD_NAME];
331 char od_mntdev[MAX_OBD_NAME];
334 /* quota slave instance for inode */
335 struct qsd_instance *od_quota_slave_md;
337 /* quota slave instance for block */
338 struct qsd_instance *od_quota_slave_dt;
340 /* osd seq instance */
341 struct lu_client_seq *od_cl_seq;
342 /* If the ratio of "the total OI mappings count" vs
343 * "the bad OI mappings count" is lower than the
344 * osd_device::od_full_scrub_ratio, then trigger
345 * OI scrub to scan the whole the device. */
346 __u64 od_full_scrub_ratio;
347 /* If the speed of found bad OI mappings (per minute)
348 * exceeds the osd_device::od_full_scrub_threshold_rate,
349 * then trigger OI scrub to scan the whole device. */
350 __u64 od_full_scrub_threshold_rate;
352 /* a list of orphaned agent inodes, protected with od_osfs_lock */
353 struct list_head od_orphan_list;
354 struct list_head od_index_backup_list;
355 struct list_head od_index_restore_list;
357 struct inode *od_index_backup_inode;
358 enum lustre_index_backup_policy od_index_backup_policy;
359 int od_index_backup_stop;
360 /* T10PI type, zero if not supported */
361 enum osd_t10_type od_t10_type;
362 atomic_t od_commit_cb_in_flight;
363 wait_queue_head_t od_commit_cb_done;
364 unsigned int __percpu *od_extent_bytes_percpu;
367 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
370 return osd->od_quota_slave_dt;
372 return osd->od_quota_slave_md;
375 enum osd_full_scrub_ratio {
376 /* Trigger OI scrub to scan the whole device directly. */
379 /* Because the bad OI mappings count cannot be larger than
380 * the total OI mappints count, then setting OFSR_NEVER means
381 * that the whole device scanning cannot be triggered by auto
382 * detected bad OI mappings during the RPC services. */
384 OFSR_DEFAULT = 10000,
387 #define FULL_SCRUB_THRESHOLD_RATE_DEFAULT 60
389 /* There are at most 15 uid/gid/projids are affected in a transaction, and
390 * that's rename case:
391 * - 3 for source parent uid & gid & projid;
392 * - 3 for source child uid & gid & projid ('..' entry update when
393 * child is directory);
394 * - 3 for target parent uid & gid & projid;
395 * - 3 for target child uid & gid & projid(if the target child exists);
396 * - 3 for root uid & gid(last_rcvd, llog, etc);
399 #define OSD_MAX_UGID_CNT 15
417 struct osd_access_lock {
418 struct list_head tl_list;
419 struct osd_object *tl_obj;
429 struct thandle ot_super;
431 struct ldiskfs_journal_cb_entry ot_jcb;
432 struct list_head ot_commit_dcb_list;
433 struct list_head ot_stop_dcb_list;
434 /* Link to the device, for debugging. */
435 struct lu_ref_link ot_dev_link;
436 unsigned int ot_credits;
437 unsigned int oh_declared_ext;
439 /* quota IDs related to the transaction */
440 unsigned short ot_id_cnt;
441 __u8 ot_id_res[OSD_MAX_UGID_CNT];
442 __u8 ot_id_types[OSD_MAX_UGID_CNT];
443 uid_t ot_id_array[OSD_MAX_UGID_CNT];
444 struct lquota_trans *ot_quota_trans;
446 unsigned int ot_remove_agents:1;
447 #if OSD_THANDLE_STATS
448 /** time when this handle was allocated */
451 /** time when this thanle was started */
454 struct list_head ot_trunc_locks;
456 * list of declarations, used for large transactions to check
457 * for duplicates, like llogs
459 struct list_head ot_declare_list;
463 * Basic transaction credit op
480 struct osd_obj_declare {
481 struct list_head old_list;
482 struct lu_fid old_fid;
483 enum dt_txn_op old_op;
491 #ifdef CONFIG_PROC_FS
493 LPROC_OSD_READ_BYTES = 0,
494 LPROC_OSD_WRITE_BYTES = 1,
495 LPROC_OSD_GET_PAGE = 2,
496 LPROC_OSD_NO_PAGE = 3,
497 LPROC_OSD_CACHE_ACCESS = 4,
498 LPROC_OSD_CACHE_HIT = 5,
499 LPROC_OSD_CACHE_MISS = 6,
501 #if OSD_THANDLE_STATS
502 LPROC_OSD_THANDLE_STARTING,
503 LPROC_OSD_THANDLE_OPEN,
504 LPROC_OSD_THANDLE_CLOSING,
506 LPROC_OSD_TOO_MANY_CREDITS,
512 * Storage representation for fids.
514 * Variable size, first byte contains the length of the whole record.
516 struct osd_fid_pack {
517 unsigned char fp_len;
518 char fp_area[sizeof(struct lu_fid)];
521 struct osd_it_ea_dirent {
522 struct lu_fid oied_fid;
525 unsigned short oied_namelen;
526 unsigned int oied_type;
528 } __attribute__((packed));
531 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
532 * than lu_dirent struct. osd readdir reads less number of dirent than
533 * required for mdd dir page. so buffer size need to be increased so that
534 * there would be one ext3 readdir for every mdd readdir page.
537 #define OSD_IT_EA_BUFSIZE (PAGE_SIZE + PAGE_SIZE/4)
540 * This is iterator's in-memory data structure in interoperability
541 * mode (i.e. iterator over ldiskfs style directory)
544 struct osd_object *oie_obj;
545 /** used in ldiskfs iterator, to stored file pointer */
546 struct file *oie_file;
547 /** how many entries have been read-cached from storage */
549 /** current entry is being iterated by caller */
551 /** current processing entry */
552 struct osd_it_ea_dirent *oie_dirent;
553 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
558 * Iterator's in-memory data structure for IAM mode.
561 struct osd_object *oi_obj;
562 struct iam_path_descr *oi_ipd;
563 struct iam_iterator oi_it;
566 struct osd_quota_leaf {
567 struct list_head oql_link;
572 * Iterator's in-memory data structure for quota file.
574 struct osd_it_quota {
575 struct osd_object *oiq_obj;
576 /** tree blocks path to where the entry is stored */
577 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
578 /** on-disk offset for current key where quota record can be found */
580 /** identifier for current quota record */
582 /** the record index in the leaf/index block */
583 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
584 /** list of already processed leaf blocks */
585 struct list_head oiq_list;
588 #define MAX_BLOCKS_PER_PAGE (PAGE_SIZE / 512)
591 wait_queue_head_t dr_wait;
592 atomic_t dr_numreqs; /* number of reqs being processed */
595 unsigned int dr_pextents; /* number block extents */
596 unsigned int dr_lextents; /* number a logical extents */
599 unsigned int dr_init_at:16, /* the line iobuf was initialized */
600 dr_elapsed_valid:1, /* we really did count time */
603 struct niobuf_local **dr_lnbs;
604 struct lu_buf dr_bl_buf;
605 struct lu_buf dr_lnb_buf;
607 ktime_t dr_start_time;
608 ktime_t dr_elapsed; /* how long io took */
609 struct osd_device *dr_dev;
610 /* Already written blocks of the start page */
611 unsigned int dr_start_pg_wblks;
612 struct inode *dr_inode;
615 #define osd_dirty_inode(inode, flag) (inode)->i_sb->s_op->dirty_inode((inode), flag)
616 #define osd_i_blocks(inode, size) ((size) >> (inode)->i_blkbits)
619 #ifdef HAVE_INODE_TIMESPEC64
620 # define osd_timespec timespec64
622 # define osd_timespec timespec
625 static inline struct osd_timespec osd_inode_time(struct inode *inode,
628 struct osd_timespec ts = { .tv_sec = seconds };
634 * any object can be identified unique way:
636 * - inode + generation (unique object within a given filesystem)
638 * version increments each insert/delete in a directory
640 struct osd_lookup_cache_object {
641 struct osd_device *lco_osd;
647 struct osd_lookup_cache_entry {
648 struct osd_lookup_cache_object lce_obj;
649 struct lu_fid lce_fid;
652 char lce_name[LDISKFS_NAME_LEN + 1];
655 #define OSD_LOOKUP_CACHE_MAX 3
657 struct osd_lookup_cache {
658 struct osd_lookup_cache_entry olc_entry[OSD_LOOKUP_CACHE_MAX];
663 #define OSD_INS_CACHE_SIZE 8
665 struct osd_thread_info {
666 const struct lu_env *oti_env;
668 * used for index operations.
670 struct dentry oti_obj_dentry;
671 struct dentry oti_child_dentry;
673 /** dentry for Iterator context. */
674 struct dentry oti_it_dentry;
677 struct kstatfs oti_ksfs;
679 struct htree_lock *oti_hlock;
681 struct lu_fid oti_fid;
682 struct lu_fid oti_fid2;
683 struct lu_fid oti_fid3;
684 struct osd_inode_id oti_id;
685 struct osd_inode_id oti_id2;
686 struct osd_inode_id oti_id3;
687 struct ost_id oti_ostid;
690 * following ipd and it structures are used for osd_index_iam_lookup()
691 * these are defined separately as we might do index operation
692 * in open iterator session.
695 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
697 unsigned int oti_it_ea_buf_used:1;
699 /* IAM iterator for index operation. */
700 struct iam_iterator oti_idx_it;
702 /** union to guarantee that ->oti_ipd[] has proper alignment. */
705 char oti_it_ipd[DX_IPD_MAX_SIZE];
706 long long oti_alignment_lieutenant;
710 char oti_idx_ipd[DX_IPD_MAX_SIZE];
711 long long oti_alignment_lieutenant_colonel;
714 struct osd_idmap_cache oti_cache;
716 /* dedicated OI cache for insert (which needs inum) */
717 struct osd_idmap_cache *oti_ins_cache;
718 int oti_ins_cache_size;
719 int oti_ins_cache_used;
720 /* inc by osd_trans_create and dec by osd_trans_stop */
721 int oti_ins_cache_depth;
726 /** used in osd_fid_set() to put xattr */
727 struct lu_buf oti_buf;
728 struct lu_buf oti_big_buf;
729 /** used in osd_ea_fid_set() to set fid into common ea */
731 struct lustre_ost_attrs oti_ost_attrs;
732 struct filter_fid_18_23 oti_ff_old;
733 struct filter_fid oti_ff;
736 struct osd_iobuf oti_iobuf;
737 /* used to access objects in /O */
738 struct inode *oti_inode;
739 #define OSD_FID_REC_SZ 32
740 char oti_ldp[OSD_FID_REC_SZ];
741 char oti_ldp2[OSD_FID_REC_SZ];
743 /* used by quota code */
745 #if defined(HAVE_DQUOT_QC_DQBLK)
746 struct qc_dqblk oti_qdq;
748 struct fs_disk_quota oti_fdq;
750 struct if_dqinfo oti_dqinfo;
752 struct lquota_id_info oti_qi;
753 struct lquota_trans oti_quota_trans;
754 union lquota_rec oti_quota_rec;
756 struct lu_seq_range oti_seq_range;
758 /* Tracking for transaction credits, to allow debugging and optimizing
759 * cases where a large number of credits are being allocated for
760 * single transaction. */
761 unsigned int oti_credits_before;
762 unsigned int oti_declare_ops[OSD_OT_MAX];
763 unsigned int oti_declare_ops_cred[OSD_OT_MAX];
764 unsigned int oti_declare_ops_used[OSD_OT_MAX];
765 struct osd_directory oti_iam;
767 struct page **oti_dio_pages;
768 int oti_dio_pages_used;
770 struct osd_it_ea_dirent *oti_seq_dirent;
771 struct osd_it_ea_dirent *oti_dir_dirent;
773 struct osd_lookup_cache_object oti_cobj; /* cache object id */
774 struct osd_lookup_cache *oti_lookup_cache;
777 extern int ldiskfs_pdo;
779 /* autoconf test is in lustre-build-ldiskfs.m4 */
780 #ifdef HAVE_BVEC_ITER_ALL
781 #define DECLARE_BVEC_ITER_ALL(iter) struct bvec_iter_all iter
783 #define DECLARE_BVEC_ITER_ALL(iter) int iter
786 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
787 const char *name, void *buf, int len)
792 dentry->d_inode = inode;
793 dentry->d_sb = inode->i_sb;
794 return ll_vfs_getxattr(dentry, inode, name, buf, len);
797 static inline int __osd_xattr_set(struct osd_thread_info *info,
798 struct inode *inode, const char *name,
799 const void *buf, int buflen, int fl)
801 struct dentry *dentry = &info->oti_child_dentry;
803 dquot_initialize(inode);
804 dentry->d_inode = inode;
805 dentry->d_sb = inode->i_sb;
806 return ll_vfs_setxattr(dentry, inode, name, buf, buflen, fl);
809 static inline char *osd_oid_name(char *name, size_t name_size,
810 const struct lu_fid *fid, u64 id)
812 snprintf(name, name_size,
813 (fid_seq_is_rsvd(fid_seq(fid)) ||
814 fid_seq_is_mdt0(fid_seq(fid)) ||
815 fid_seq_is_idif(fid_seq(fid))) ? "%llu" : "%llx", id);
820 #ifdef CONFIG_PROC_FS
822 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
823 int osd_procfs_init(struct osd_device *osd, const char *name);
824 int osd_procfs_fini(struct osd_device *osd);
825 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
826 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 0, 52, 0)
827 int osd_register_proc_index_in_idif(struct osd_device *osd);
831 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
832 struct obd_statfs *sfs, struct obd_statfs_info *info);
833 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
834 struct osd_inode_id *id);
836 osd_iget_fid(struct osd_thread_info *info, struct osd_device *dev,
837 struct osd_inode_id *id, struct lu_fid *fid);
838 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
839 const struct lu_fid *fid, __u32 compat, __u32 incompat);
840 struct osd_obj_seq *osd_seq_load(struct osd_thread_info *info,
841 struct osd_device *osd, u64 seq);
842 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
843 struct dentry *dentry, struct lustre_ost_attrs *loa);
844 void osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
845 struct osd_inode_id *id, const struct lu_fid *fid);
846 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
847 struct dentry *dentry, struct lu_fid *fid);
848 struct osd_it_ea *osd_it_dir_init(const struct lu_env *env,
849 struct osd_device *dev, struct inode *inode,
851 void osd_it_dir_fini(const struct lu_env *env, struct osd_it_ea *oie,
852 struct inode *inode);
853 int osd_ldiskfs_it_fill(const struct lu_env *env, const struct dt_it *di);
855 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
856 void osd_obj_map_fini(struct osd_device *dev);
857 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
858 const struct lu_fid *fid, struct osd_inode_id *id);
859 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
860 const struct lu_fid *fid, const struct osd_inode_id *id,
862 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
863 const struct lu_fid *fid, handle_t *th);
864 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
865 const struct lu_fid *fid, const struct osd_inode_id *id,
867 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
868 struct inode *src_parent, struct dentry *src_child,
869 const struct lu_fid *fid);
870 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
871 const struct lu_fid *fid, struct osd_inode_id *id,
872 enum oi_check_flags flags);
873 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
874 const struct lu_fid *fid, const struct osd_inode_id *id,
876 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
877 const struct lu_fid *fid, const struct osd_inode_id *id,
879 int osd_obj_del_entry(struct osd_thread_info *info, struct osd_device *osd,
880 struct dentry *dird, char *name, int namelen,
883 char *osd_lf_fid2name(const struct lu_fid *fid);
884 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
886 void osd_scrub_stop(struct osd_device *dev);
887 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev,
889 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
890 int osd_scrub_oi_insert(struct osd_device *dev, const struct lu_fid *fid,
891 struct osd_inode_id *id, int insert);
892 bool osd_scrub_oi_resurrect(struct lustre_scrub *scrub,
893 const struct lu_fid *fid);
894 void osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
896 struct dentry *osd_lookup_one_len_unlocked(struct osd_device *dev,
898 struct dentry *base, int len);
899 struct dentry *osd_lookup_one_len(struct osd_device *dev, const char *name,
900 struct dentry *base, int len);
902 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
903 u64 seq, struct lu_seq_range *range);
905 int osd_delete_from_remote_parent(const struct lu_env *env,
906 struct osd_device *osd,
907 struct osd_object *obj,
908 struct osd_thandle *oh, bool destroy);
909 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
910 struct osd_object *obj, struct osd_thandle *oh);
911 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
912 struct osd_device *osd,
913 const struct lu_fid *fid,
914 struct osd_inode_id *id);
916 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
918 int osd_scrub_refresh_mapping(struct osd_thread_info *info,
919 struct osd_device *dev,
920 const struct lu_fid *fid,
921 const struct osd_inode_id *id,
923 enum oi_check_flags flags, bool *exist);
925 /* osd_quota_fmt.c */
926 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
927 int type, uint blk, int depth, uint index,
928 struct osd_it_quota *it);
929 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
930 int type, uint blk, uint index,
931 struct osd_it_quota *it);
932 loff_t find_tree_dqentry(const struct lu_env *env,
933 struct osd_object *obj, int type,
934 qid_t dqid, uint blk, int depth,
935 struct osd_it_quota *it);
937 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
938 struct lquota_id_info *qi, struct osd_object *obj,
939 bool enforce, enum osd_quota_local_flags *local_flags);
940 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
941 __u32 projid, long long space, struct osd_thandle *oh,
942 struct osd_object *obj,
943 enum osd_quota_local_flags *local_flags,
944 enum osd_qid_declare_flags);
945 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
946 const struct dt_rec *rec,
947 union lquota_rec *quota_rec);
948 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
950 #ifdef HAVE_PROJECT_QUOTA
951 static inline __u32 i_projid_read(struct inode *inode)
953 return (__u32)from_kprojid(&init_user_ns, LDISKFS_I(inode)->i_projid);
956 static inline void i_projid_write(struct inode *inode, __u32 projid)
959 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
960 LDISKFS_I(inode)->i_projid = kprojid;
963 static inline uid_t i_projid_read(struct inode *inode)
967 static inline void i_projid_write(struct inode *inode, __u32 projid)
973 #ifdef HAVE_LDISKFS_IGET_WITH_FLAGS
974 # define osd_ldiskfs_iget(sb, ino) \
975 ldiskfs_iget((sb), (ino), \
976 LDISKFS_IGET_HANDLE | LDISKFS_IGET_SPECIAL)
978 # define osd_ldiskfs_iget(sb, ino) ldiskfs_iget((sb), (ino))
981 #ifdef HAVE_LDISKFS_INFO_JINODE
982 # define osd_attach_jinode(inode) ldiskfs_inode_attach_jinode(inode)
983 #else /* HAVE_LDISKFS_INFO_JINODE */
984 # define osd_attach_jinode(inode) 0
985 #endif /* HAVE_LDISKFS_INFO_JINODE */
987 #ifndef HAVE_JBD2_JOURNAL_GET_MAX_TXN_BUFS
988 #define jbd2_journal_get_max_txn_bufs(jrnl) \
989 (jrnl->j_max_transaction_buffers)
992 #ifdef LDISKFS_HT_MISC
993 # define osd_journal_start_sb(sb, type, nblock) \
994 ldiskfs_journal_start_sb(sb, type, nblock)
995 #ifdef HAVE_LDISKFS_JOURNAL_ENSURE_CREDITS
996 # define osd_journal_start_with_revoke(sb, type, nblock, revoke) \
997 ldiskfs_journal_start_with_revoke((sb)->s_root->d_inode, type, \
1000 # define osd_journal_start_with_revoke(sb, type, nblock, revoke) \
1001 ldiskfs_journal_start_sb(sb, type, nblock)
1003 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
1004 struct inode *inode,
1005 ldiskfs_lblk_t *nblock)
1009 rc = osd_attach_jinode(inode);
1012 return ldiskfs_append(handle, inode, nblock);
1015 # ifdef HAVE___LDISKFS_FIND_ENTRY
1016 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1017 (__ldiskfs_find_entry(dir, name, de, inlined, lock) ?: \
1020 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1021 (ldiskfs_find_entry_locked(dir, name, de, inlined, lock) ?: \
1025 # define osd_journal_start(inode, type, nblocks) \
1026 ldiskfs_journal_start(inode, type, nblocks)
1027 # define osd_transaction_size(dev) \
1028 (jbd2_journal_get_max_txn_bufs(osd_journal(dev)) / 2)
1029 #else /* ! defined LDISKFS_HT_MISC */
1030 # define LDISKFS_HT_MISC 0
1031 # define osd_journal_start_sb(sb, type, nblock) \
1032 ldiskfs_journal_start_sb(sb, nblock)
1034 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
1035 struct inode *inode,
1036 ldiskfs_lblk_t *nblock)
1038 struct buffer_head *bh;
1041 bh = ldiskfs_append(handle, inode, nblock, &err);
1048 # ifdef HAVE___LDISKFS_FIND_ENTRY
1049 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1050 (__ldiskfs_find_entry(dir, name, de, lock) ?: \
1053 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1054 (ldiskfs_find_entry_locked(dir, name, de, lock) ?: \
1057 # define osd_journal_start(inode, type, nblocks) \
1058 ldiskfs_journal_start(inode, nblocks)
1059 # define osd_transaction_size(dev) \
1060 jbd2_journal_get_max_txn_bufs(osd_journal(dev))
1061 #endif /* LDISKFS_HT_MISC */
1063 #ifndef HAVE___LDISKFS_FIND_ENTRY
1064 # define __ldiskfs_add_entry(handle, child, inode, hlock) \
1065 ldiskfs_add_entry_locked(handle, child, inode, hlock)
1069 * Invariants, assertions.
1073 * XXX: do not enable this, until invariant checking code is made thread safe
1074 * in the face of pdirops locking.
1076 #define OSD_INVARIANT_CHECKS (0)
1078 #if OSD_INVARIANT_CHECKS
1079 static inline int osd_invariant(const struct osd_object *obj)
1083 ergo(obj->oo_inode != NULL,
1084 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
1085 atomic_read(&obj->oo_inode->i_count) > 0) &&
1086 ergo(obj->oo_dir != NULL &&
1087 obj->oo_dir->od_conationer.ic_object != NULL,
1088 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
1091 #define osd_invariant(obj) (1)
1094 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
1095 #define OSD_READCACHE_MAX_IO_MB 8
1096 #define OSD_WRITECACHE_MAX_IO_MB 8
1098 extern const struct dt_index_operations osd_otable_ops;
1100 static inline int osd_oi_fid2idx(struct osd_device *dev,
1101 const struct lu_fid *fid)
1103 return fid->f_seq & (dev->od_oi_count - 1);
1106 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
1107 const struct lu_fid *fid)
1109 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
1110 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
1111 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
1112 DFID"\n", PFID(fid));
1113 /* It can work even od_oi_count equals to 1 although it's unexpected,
1114 * the only reason we set it to 1 is for performance measurement */
1115 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
1118 extern const struct lu_device_operations osd_lu_ops;
1120 static inline int lu_device_is_osd(const struct lu_device *d)
1122 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
1125 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
1127 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
1128 return container_of(d, struct osd_device, od_dt_dev);
1131 static inline struct osd_device *osd_dev(const struct lu_device *d)
1133 LASSERT(lu_device_is_osd(d));
1134 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
1137 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
1139 return osd_dev(o->oo_dt.do_lu.lo_dev);
1142 static inline struct super_block *osd_sb(const struct osd_device *dev)
1144 return dev->od_mnt->mnt_sb;
1147 static inline const char *osd_sb2name(const struct super_block *sb)
1149 /* this is LDISKFS_SB(sb), but preserves "const" */
1150 const struct ldiskfs_sb_info *sbi = sb->s_fs_info;
1152 return sbi->s_es->s_volume_name;
1155 static inline const char *osd_dev2name(const struct osd_device *dev)
1157 return osd_sb2name(osd_sb(dev));
1160 static inline const char *osd_ino2name(const struct inode *inode)
1162 return osd_sb2name(inode->i_sb);
1166 * Put the osd object once done with it.
1168 * \param obj osd object that needs to be put
1170 static inline void osd_object_put(const struct lu_env *env,
1171 struct osd_object *obj)
1173 dt_object_put(env, &obj->oo_dt);
1176 static inline int osd_object_is_root(const struct osd_object *obj)
1178 return obj->oo_inode && is_root_inode(obj->oo_inode);
1181 static inline struct osd_object *osd_obj(const struct lu_object *o)
1183 LASSERT(lu_device_is_osd(o->lo_dev));
1184 return container_of(o, struct osd_object, oo_dt.do_lu);
1187 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
1189 return osd_obj(&d->do_lu);
1192 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
1194 return &osd->od_dt_dev.dd_lu_dev;
1197 static inline journal_t *osd_journal(const struct osd_device *dev)
1199 return LDISKFS_SB(osd_sb(dev))->s_journal;
1202 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
1204 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
1207 static inline char *osd_name(struct osd_device *osd)
1209 return osd->od_svname;
1212 static inline bool osd_is_ea_inode(struct inode *inode)
1214 return !!(LDISKFS_I(inode)->i_flags & LDISKFS_EA_INODE_FL);
1217 extern const struct dt_body_operations osd_body_ops;
1218 extern struct lu_context_key osd_key;
1220 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
1222 return lu_context_key_get(&env->le_ctx, &osd_key);
1225 extern const struct dt_body_operations osd_body_ops_new;
1231 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
1232 const struct iam_container *bag)
1234 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1235 osd_oti_get(env)->oti_it_ipd);
1239 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
1240 const struct iam_container *bag)
1242 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1243 osd_oti_get(env)->oti_idx_ipd);
1246 static inline void osd_ipd_put(const struct lu_env *env,
1247 const struct iam_container *bag,
1248 struct iam_path_descr *ipd)
1250 bag->ic_descr->id_ops->id_ipd_free(ipd);
1253 int osd_calc_bkmap_credits(struct super_block *sb, struct inode *inode,
1254 const loff_t size, const loff_t pos,
1257 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
1260 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
1261 struct inode *inode,
1262 const char *name, const int namelen)
1264 struct osd_thread_info *info = osd_oti_get(env);
1265 struct dentry *child_dentry = &info->oti_child_dentry;
1266 struct dentry *obj_dentry = &info->oti_obj_dentry;
1268 obj_dentry->d_inode = inode;
1269 obj_dentry->d_sb = inode->i_sb;
1270 obj_dentry->d_name.hash = 0;
1272 child_dentry->d_name.hash = 0;
1273 child_dentry->d_parent = obj_dentry;
1274 child_dentry->d_name.name = name;
1275 child_dentry->d_name.len = namelen;
1276 return child_dentry;
1279 extern int osd_trans_declare_op2rb[];
1280 extern int ldiskfs_track_declares_assert;
1281 void osd_trans_dump_creds(const struct lu_env *env, struct thandle *th);
1283 static inline void osd_trans_declare_op(const struct lu_env *env,
1284 struct osd_thandle *oh,
1285 enum osd_op_type op, int credits)
1287 struct osd_thread_info *oti = osd_oti_get(env);
1289 LASSERT(oh->ot_handle == NULL);
1290 if (unlikely(op >= OSD_OT_MAX)) {
1291 if (unlikely(ldiskfs_track_declares_assert)) {
1292 LASSERT(op < OSD_OT_MAX);
1294 CWARN("%s: Invalid operation index %d\n",
1295 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op);
1299 oti->oti_declare_ops[op]++;
1300 oti->oti_declare_ops_cred[op] += credits;
1302 oh->ot_credits += credits;
1305 /* linux: v5.4-rc3-21-g933f1c1e0b75 renamed h_buffer_credits */
1306 #ifdef HAVE_JOURNAL_TOTAL_CREDITS
1307 #define h_buffer_credits h_total_credits
1310 static inline void osd_trans_exec_op(const struct lu_env *env,
1312 enum osd_op_type op)
1314 struct osd_thread_info *oti = osd_oti_get(env);
1315 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1317 unsigned int rb, left;
1319 LASSERT(oh->ot_handle != NULL);
1320 if (unlikely(op >= OSD_OT_MAX)) {
1321 if (unlikely(ldiskfs_track_declares_assert))
1322 LASSERT(op < OSD_OT_MAX);
1324 CWARN("%s: opcode %u: invalid value >= %u\n",
1325 osd_name(osd_dt_dev(oh->ot_super.th_dev)),
1332 /* find rollback (or reverse) operation for the given one
1333 * such an operation doesn't require additional credits
1334 * as the same set of blocks are modified */
1335 rb = osd_trans_declare_op2rb[op];
1337 /* check whether credits for this operation were reserved at all */
1338 if (unlikely(oti->oti_declare_ops_cred[op] == 0 &&
1339 oti->oti_declare_ops_cred[rb] == 0)) {
1340 /* the API is not perfect yet: CREATE does REF_ADD internally
1341 * while DESTROY does not. To rollback CREATE the callers
1342 * needs to call REF_DEL+DESTROY which is hard to detect using
1343 * a simple table of rollback operations */
1344 if (op == OSD_OT_REF_DEL &&
1345 oti->oti_declare_ops_cred[OSD_OT_CREATE] > 0)
1347 if (op == OSD_OT_REF_ADD &&
1348 oti->oti_declare_ops_cred[OSD_OT_DESTROY] > 0)
1350 CWARN("%s: opcode %u: credits = 0, rollback = %u\n",
1351 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
1352 osd_trans_dump_creds(env, th);
1353 LASSERT(!ldiskfs_track_declares_assert);
1357 /* remember how many credits we have unused before the operation */
1358 oti->oti_credits_before = oh->ot_handle->h_buffer_credits;
1359 left = oti->oti_declare_ops_cred[op] - oti->oti_declare_ops_used[op];
1360 if (unlikely(oti->oti_credits_before < left)) {
1361 CWARN("%s: opcode %u: before %u < left %u, rollback = %u\n",
1362 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1363 oti->oti_credits_before, left, rb);
1364 osd_trans_dump_creds(env, th);
1365 /* on a very small fs (testing?) it's possible that
1366 * the transaction can't fit 1/4 of journal, so we
1367 * just request less credits (see osd_trans_start()).
1368 * ignore the same case here */
1369 rb = osd_transaction_size(osd_dt_dev(th->th_dev));
1370 if (unlikely(oh->ot_credits < rb))
1371 LASSERT(!ldiskfs_track_declares_assert);
1375 static inline void osd_trans_exec_check(const struct lu_env *env,
1377 enum osd_op_type op)
1379 struct osd_thread_info *oti = osd_oti_get(env);
1380 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1382 int used, over, quota;
1384 /* how many credits have been used by the operation */
1385 used = oti->oti_credits_before - oh->ot_handle->h_buffer_credits;
1387 if (unlikely(used < 0)) {
1388 /* if some block was allocated and released in the same
1389 * transaction, then it won't be a part of the transaction
1390 * and delta can be negative */
1395 /* rollback operations (e.g. when we destroy just created
1396 * object) should not consume any credits. there is no point
1397 * to confuse the checks below */
1401 oti->oti_declare_ops_used[op] += used;
1402 if (oti->oti_declare_ops_used[op] <= oti->oti_declare_ops_cred[op])
1405 /* we account quota for a whole transaction and any operation can
1406 * consume corresponding credits */
1407 over = oti->oti_declare_ops_used[op] -
1408 oti->oti_declare_ops_cred[op];
1409 quota = oti->oti_declare_ops_cred[OSD_OT_QUOTA] -
1410 oti->oti_declare_ops_used[OSD_OT_QUOTA];
1411 if (over <= quota) {
1412 /* probably that credits were consumed by
1413 * quota indirectly (in the depths of ldiskfs) */
1414 oti->oti_declare_ops_used[OSD_OT_QUOTA] += over;
1415 oti->oti_declare_ops_used[op] -= over;
1417 CWARN("%s: opcode %d: used %u, used now %u, reserved %u\n",
1418 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1419 oti->oti_declare_ops_used[op], used,
1420 oti->oti_declare_ops_cred[op]);
1421 osd_trans_dump_creds(env, th);
1422 if (unlikely(ldiskfs_track_declares_assert))
1428 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1431 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1432 struct lu_fid *befider)
1434 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1435 memcpy(pack->fp_area, befider, sizeof(*befider));
1436 pack->fp_len = sizeof(*befider) + 1;
1440 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1445 switch (pack->fp_len) {
1446 case sizeof *fid + 1:
1447 memcpy(fid, pack->fp_area, sizeof *fid);
1448 fid_be_to_cpu(fid, fid);
1451 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1458 * Quota/Accounting handling
1460 extern const struct dt_index_operations osd_acct_index_ops;
1461 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1462 const struct lu_fid *fid, struct osd_inode_id *id);
1464 /* copy from fs/ext4/dir.c */
1465 static inline int is_32bit_api(void)
1467 #ifdef CONFIG_COMPAT
1468 return in_compat_syscall();
1470 return (BITS_PER_LONG == 32);
1474 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1476 if ((filp->f_mode & FMODE_32BITHASH) ||
1477 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1478 return LDISKFS_HTREE_EOF_32BIT;
1480 return LDISKFS_HTREE_EOF_64BIT;
1483 static inline int fid_is_internal(const struct lu_fid *fid)
1485 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1488 static inline bool is_remote_parent_ino(struct osd_device *o, unsigned long ino)
1493 LASSERT(o->od_mdt_map != NULL);
1495 return ino == o->od_mdt_map->omm_remote_parent->d_inode->i_ino;
1499 * ext4_bread/ldiskfs_bread has either 5 or 4 parameters. The error
1500 * return code has been removed and integrated into the pointer in the
1503 static inline struct buffer_head *__ldiskfs_bread(handle_t *handle,
1504 struct inode *inode,
1505 ldiskfs_lblk_t block,
1509 struct buffer_head *bh;
1512 rc = osd_attach_jinode(inode);
1516 #ifdef HAVE_EXT4_BREAD_4ARGS
1517 bh = ldiskfs_bread(handle, inode, block, create);
1520 bh = ldiskfs_bread(handle, inode, block, create, &rc);
1521 if (bh == NULL && rc != 0)
1527 #ifdef HAVE_EXT4_JOURNAL_GET_WRITE_ACCESS_4ARGS
1528 # define osd_ldiskfs_journal_get_write_access(handle, sb, bh, flags) \
1529 ldiskfs_journal_get_write_access((handle), (sb), (bh), (flags))
1531 # define LDISKFS_JTR_NONE 0
1532 # define osd_ldiskfs_journal_get_write_access(handle, sb, bh, flags) \
1533 ldiskfs_journal_get_write_access((handle), (bh))
1534 #endif /* HAVE_EXT4_JOURNAL_GET_WRITE_ACCESS_4ARGS */
1536 #ifdef HAVE_EXT4_INC_DEC_COUNT_2ARGS
1537 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((h), (inode))
1538 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((h), (inode))
1540 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((inode))
1541 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((inode))
1542 #endif /* HAVE_EXT4_INC_DEC_COUNT_2ARGS */
1544 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1547 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1548 __u32 keysize, __u32 recsize)
1550 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1551 &osd->od_index_backup_list, &osd->od_lock,
1552 &osd->od_index_backup_stop,
1553 fid, keysize, recsize);
1557 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1559 struct osd_thread_info *info = osd_oti_get(env);
1560 struct lu_fid *fid = &info->oti_fid3;
1561 struct osd_inode_id *id = &info->oti_id3;
1563 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1564 osd_id_gen(id, osd->od_index_backup_inode->i_ino,
1565 osd->od_index_backup_inode->i_generation);
1566 osd_add_oi_cache(info, osd, id, fid);
1568 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1569 &osd->od_index_backup_list, &osd->od_lock,
1570 &osd->od_index_backup_stop, backup);
1573 #ifdef LDISKFS_HAS_INCOMPAT_FEATURE
1575 # ifdef LDISKFS_FEATURE_INCOMPAT_EXTENTS
1576 # define ldiskfs_has_feature_extents(sb) \
1577 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EXTENTS)
1579 # ifdef LDISKFS_FEATURE_INCOMPAT_EA_INODE
1580 # define ldiskfs_has_feature_ea_inode(sb) \
1581 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EA_INODE)
1583 # ifdef LDISKFS_FEATURE_INCOMPAT_DIRDATA
1584 # define ldiskfs_has_feature_dirdata(sb) \
1585 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_DIRDATA)
1587 # ifdef LDISKFS_FEATURE_INCOMPAT_LARGEDIR
1588 # define ldiskfs_has_feature_largedir(sb) \
1589 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1590 # define ldiskfs_set_feature_largedir(sb) \
1591 LDISKFS_SET_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1593 # ifdef LDISKFS_FEATURE_COMPAT_HAS_JOURNAL
1594 # define ldiskfs_has_feature_journal(sb) \
1595 LDISKFS_HAS_COMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_HAS_JOURNAL)
1597 # ifdef LDISKFS_FEATURE_RO_COMPAT_QUOTA
1598 # define ldiskfs_has_feature_quota(sb) \
1599 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_QUOTA)
1601 # ifdef LDISKFS_FEATURE_RO_COMPAT_PROJECT
1602 # define ldiskfs_has_feature_project(sb) \
1603 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_PROJECT)
1605 # ifdef LDISKFS_FEATURE_COMPAT_FAST_COMMIT
1606 # define ldiskfs_has_feature_fast_commit(sb) \
1607 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_FAST_COMMIT)
1611 #ifndef ldiskfs_has_feature_fast_commit
1612 #define ldiskfs_has_feature_fast_commit(sb) false
1615 int osd_trunc_lock(struct osd_object *obj, struct osd_thandle *oh,
1617 void osd_trunc_unlock_all(const struct lu_env *env, struct list_head *list);
1618 int osd_process_truncates(const struct lu_env *env, struct list_head *list);
1619 void osd_execute_truncate(struct osd_object *obj);
1621 #ifndef HAVE___BI_CNT
1622 #define __bi_cnt bi_cnt
1625 #ifndef HAVE_CLEAN_BDEV_ALIASES
1626 #define clean_bdev_aliases(bdev, block, len) \
1627 unmap_underlying_metadata((bdev), (block))
1630 #ifndef HAVE_BI_STATUS
1631 #define bi_status bi_error
1635 * Maximum size of xattr attributes for FEATURE_INCOMPAT_EA_INODE 1Mb
1636 * This limit is arbitrary, but is reasonable for the xattr API.
1638 #define LDISKFS_XATTR_MAX_LARGE_EA_SIZE (1024 * 1024)
1640 struct osd_bio_private {
1641 struct osd_iobuf *obp_iobuf;
1642 /* Start page index in the obp_iobuf for the bio */
1643 int obp_start_page_idx;
1645 extern struct kmem_cache *biop_cachep;
1647 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && defined(HAVE_BIO_INTEGRITY_PREP_FN)
1648 int osd_bio_integrity_handle(struct osd_device *osd, struct bio *bio,
1649 struct osd_iobuf *iobuf);
1650 #else /* !CONFIG_BLK_DEV_INTEGRITY */
1651 #define osd_bio_integrity_handle(osd, bio, iobuf) (0)
1654 #ifdef HAVE_BIO_INTEGRITY_PREP_FN
1655 # ifdef HAVE_BLK_INTEGRITY_ITER
1656 # define integrity_gen_fn integrity_processing_fn
1657 # define integrity_vrfy_fn integrity_processing_fn
1659 int osd_get_integrity_profile(struct osd_device *osd,
1660 integrity_gen_fn **generate_fn,
1661 integrity_vrfy_fn **verify_fn);
1662 #endif /* HAVE_BIO_INTEGRITY_PREP_FN */
1664 #ifdef HAVE_BIO_BI_PHYS_SEGMENTS
1665 #define osd_bio_nr_segs(bio) ((bio)->bi_phys_segments)
1667 #define osd_bio_nr_segs(bio) bio_segments((bio))
1668 #endif /* HAVE_BIO_BI_PHYS_SEGMENTS */
1670 #ifdef HAVE_GET_INODE_USAGE
1671 #define lock_dquot_transfer(inode) down_read(&LDISKFS_I(inode)->xattr_sem)
1672 #define unlock_dquot_transfer(inode) up_read(&LDISKFS_I(inode)->xattr_sem)
1674 #define lock_dquot_transfer(inode) do {} while (0)
1675 #define unlock_dquot_transfer(inode) do {} while (0)
1678 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)
1679 static inline unsigned short blk_integrity_interval(struct blk_integrity *bi)
1681 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1682 return bi->interval_exp ? 1 << bi->interval_exp : 0;
1683 #elif defined(HAVE_INTERVAL_BLK_INTEGRITY)
1684 return bi->interval;
1686 return bi->sector_size;
1687 #endif /* !HAVE_INTERVAL_EXP_BLK_INTEGRITY */
1690 static inline const char *blk_integrity_name(struct blk_integrity *bi)
1692 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1693 return bi->profile->name;
1699 static inline unsigned int bip_size(struct bio_integrity_payload *bip)
1701 #ifdef HAVE_BIP_ITER_BIO_INTEGRITY_PAYLOAD
1702 return bip->bip_iter.bi_size;
1704 return bip->bip_size;
1707 #else /* !CONFIG_BLK_DEV_INTEGRITY */
1708 static inline unsigned short blk_integrity_interval(struct blk_integrity *bi)
1712 static inline const char *blk_integrity_name(struct blk_integrity *bi)
1714 /* gcc8 dislikes when strcmp() is called against NULL */
1717 #endif /* !CONFIG_BLK_DEV_INTEGRITY */
1719 #ifndef INTEGRITY_FLAG_READ
1720 #define INTEGRITY_FLAG_READ BLK_INTEGRITY_VERIFY
1723 #ifndef INTEGRITY_FLAG_WRITE
1724 #define INTEGRITY_FLAG_WRITE BLK_INTEGRITY_GENERATE
1727 static inline bool bdev_integrity_enabled(struct block_device *bdev, int rw)
1729 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)
1730 struct blk_integrity *bi = bdev_get_integrity(bdev);
1735 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1736 if (rw == 0 && bi->profile->verify_fn != NULL &&
1737 (bi->flags & INTEGRITY_FLAG_READ))
1740 if (rw == 1 && bi->profile->generate_fn != NULL &&
1741 (bi->flags & INTEGRITY_FLAG_WRITE))
1744 if (rw == 0 && bi->verify_fn != NULL &&
1745 (bi->flags & INTEGRITY_FLAG_READ))
1748 if (rw == 1 && bi->generate_fn != NULL &&
1749 (bi->flags & INTEGRITY_FLAG_WRITE))
1751 #endif /* !HAVE_INTERVAL_EXP_BLK_INTEGRITY */
1752 #endif /* !CONFIG_BLK_DEV_INTEGRITY */
1757 bool osd_tx_was_declared(const struct lu_env *env, struct osd_thandle *oth,
1758 struct dt_object *dt, enum dt_txn_op op, loff_t p);
1760 #ifdef HAVE_DQUOT_TRANSFER_WITH_USER_NS
1761 #define osd_dquot_transfer(ns, i, a) dquot_transfer((ns), (i), (a))
1763 #define osd_dquot_transfer(ns, i, a) dquot_transfer((i), (a))
1766 #ifdef HAVE_FILLDIR_USE_CTX_RETURN_BOOL
1767 #define WRAP_FILLDIR_FN(prefix, fill_fn) \
1768 static bool fill_fn(struct dir_context *buf, const char *name, int namelen, \
1769 loff_t offset, __u64 ino, unsigned int d_type) \
1771 return !prefix##fill_fn(buf, name, namelen, offset, ino, d_type); \
1773 #elif defined(HAVE_FILLDIR_USE_CTX)
1774 #define WRAP_FILLDIR_FN(prefix, fill_fn) \
1775 static int fill_fn(struct dir_context *buf, const char *name, int namelen, \
1776 loff_t offset, __u64 ino, unsigned int d_type) \
1778 return prefix##fill_fn(buf, name, namelen, offset, ino, d_type); \
1781 #define WRAP_FILLDIR_FN(prefix, fill_fn)
1784 #endif /* _OSD_INTERNAL_H */