4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
31 * lustre/osd/osd_internal.h
33 * Shared definitions and declarations for osd module
35 * Author: Nikita Danilov <nikita@clusterfs.com>
38 #ifndef _OSD_INTERNAL_H
39 #define _OSD_INTERNAL_H
43 #include <linux/mutex.h>
44 /* struct rw_semaphore */
45 #include <linux/rwsem.h>
47 #include <linux/dcache.h>
49 #include <linux/dirent.h>
50 #include <linux/statfs.h>
51 #include <linux/bio.h>
52 #include <linux/file.h>
53 #include <ldiskfs/ldiskfs.h>
54 #include <ldiskfs/ldiskfs_jbd2.h>
58 /* class_register_type(), class_unregister_type() */
59 #include <obd_class.h>
60 #include <lustre_disk.h>
61 #include <dt_object.h>
62 #include <lustre_quota.h>
66 #include "osd_scrub.h"
67 #include "osd_quota_fmt.h"
69 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)
70 #ifdef HAVE_LINUX_BLK_INTEGRITY_HEADER
71 #include <linux/blk-integrity.h>
76 extern struct kmem_cache *dynlock_cachep;
78 #define OSD_COUNTERS (0)
80 /* ldiskfs special inode::i_state_flags need to be accessed with
81 * ldiskfs_{set,clear,test}_inode_state() only */
83 /* OI scrub should skip this inode. */
84 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
85 #define LDISKFS_STATE_LUSTRE_DESTROY 30
87 /** Enable thandle usage statistics */
88 #define OSD_THANDLE_STATS (0)
90 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
92 #define OBJECTS "OBJECTS"
93 #define ADMIN_USR "admin_quotafile_v2.usr"
94 #define ADMIN_GRP "admin_quotafile_v2.grp"
96 /* Statfs space reservation for fragmentation and local objects */
97 #define OSD_STATFS_RESERVED (1ULL << 23) /* 8MB */
98 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
100 /* Default extent bytes when declaring write commit */
101 #define OSD_DEFAULT_EXTENT_BYTES (1U << 20)
103 /* check if ldiskfs support project quota */
104 #if LDISKFS_MAXQUOTAS < 3
105 #undef HAVE_PROJECT_QUOTA
108 #define OBD_BRW_MAPPED OBD_BRW_LOCAL1
110 struct osd_directory {
111 struct iam_container od_container;
112 struct iam_descr od_descr;
116 * Object Index (oi) instance.
120 * underlying index object, where fid->id mapping in stored.
122 struct inode *oi_inode;
123 struct osd_directory oi_dir;
126 extern const int osd_dto_credits_noquota[];
129 struct dt_object oo_dt;
131 * Inode for file system object represented by this osd_object. This
132 * inode is pinned for the whole duration of lu_object life.
134 * Not modified concurrently (either setup early during object
135 * creation, or assigned by osd_create() under write lock).
137 struct inode *oo_inode;
139 * to protect index ops.
141 struct htree_lock_head *oo_hl_head;
142 struct rw_semaphore oo_ext_idx_sem;
143 struct rw_semaphore oo_sem;
144 struct osd_directory *oo_dir;
145 /** protects inode attributes. */
149 * Following two members *compat_dot* are used to indicate
150 * the presence of dot and dotdot in the given directory.
151 * This is required for interop mode (b11826).
153 __u32 oo_destroyed:1,
155 oo_compat_dot_created:1,
156 oo_compat_dotdot_created:1;
158 /* the i_flags in LMA */
161 const struct lu_env *oo_owner;
163 struct list_head oo_xattr_list;
164 struct lu_object_header *oo_header;
165 __u64 oo_dirent_count;
169 /* protects on-fly initialization */
170 int oos_subdir_count; /* subdir count for each seq */
171 struct dentry *oos_root; /* O/<seq> */
172 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
173 u64 oos_seq; /* seq number */
174 struct list_head oos_seq_list; /* list to seq_list */
178 struct dentry *om_root; /* dentry for /O */
179 rwlock_t om_seq_list_lock; /* lock for seq_list */
180 struct list_head om_seq_list; /* list head for seq */
182 struct mutex om_dir_init_mutex;
186 struct dentry *om_root; /* AGENT/<index> */
187 u64 om_index; /* mdt index */
188 struct list_head om_list; /* list to omm_list */
191 struct osd_mdobj_map {
192 struct dentry *omm_remote_parent;
194 int osd_ldiskfs_add_entry(struct osd_thread_info *info, struct osd_device *osd,
195 handle_t *handle, struct dentry *child,
196 struct inode *inode, struct htree_lock *hlock);
198 #define OSD_OTABLE_IT_CACHE_SIZE 64
199 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
201 struct osd_inconsistent_item {
202 /* link into lustre_scrub::os_inconsistent_items,
203 * protected by lustre_scrub::os_lock. */
204 struct list_head oii_list;
206 /* The right FID <=> ino#/gen mapping. */
207 struct osd_idmap_cache oii_cache;
209 unsigned int oii_insert:1; /* insert or update mapping. */
212 struct osd_otable_cache {
213 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
215 /* Index for next cache slot to be filled. */
216 int ooc_producer_idx;
218 /* Index for next cache slot to be returned by it::next(). */
219 int ooc_consumer_idx;
221 /* How many items in ooc_cache. */
222 __u64 ooc_cached_items;
224 /* Position for up layer LFSCK iteration pre-loading. */
225 __u64 ooc_pos_preload;
228 struct osd_otable_it {
229 struct osd_device *ooi_dev;
230 struct osd_otable_cache ooi_cache;
231 struct osd_iit_param ooi_iit_param;
233 /* The following bits can be updated/checked w/o lock protection.
234 * If more bits will be introduced in the future and need lock to
235 * protect, please add comment. */
236 unsigned long ooi_used_outside:1, /* Some user out of OSD
237 * uses the iteration. */
238 ooi_all_cached:1, /* No more entries can be
239 * filled into cache. */
240 ooi_user_ready:1, /* The user out of OSD is
241 * ready to iterate. */
242 ooi_waiting:1; /* it::next is waiting. */
245 struct osd_obj_orphan {
246 struct list_head oor_list;
247 struct lu_env *oor_env; /* to identify "own" records */
252 OSD_T10_TYPE_UNKNOWN = 0x00,
253 OSD_T10_TYPE1 = 0x01,
254 OSD_T10_TYPE2 = 0x02,
255 OSD_T10_TYPE3 = 0x04,
256 OSD_T10_TYPE_CRC = 0x08,
257 OSD_T10_TYPE_IP = 0x10,
258 OSD_T10_TYPE1_CRC = OSD_T10_TYPE1 | OSD_T10_TYPE_CRC,
259 OSD_T10_TYPE2_CRC = OSD_T10_TYPE2 | OSD_T10_TYPE_CRC,
260 OSD_T10_TYPE3_CRC = OSD_T10_TYPE3 | OSD_T10_TYPE_CRC,
261 OSD_T10_TYPE1_IP = OSD_T10_TYPE1 | OSD_T10_TYPE_IP,
262 OSD_T10_TYPE2_IP = OSD_T10_TYPE2 | OSD_T10_TYPE_IP,
263 OSD_T10_TYPE3_IP = OSD_T10_TYPE3 | OSD_T10_TYPE_IP,
271 struct dt_device od_dt_dev;
272 /* information about underlying file system */
273 struct vfsmount *od_mnt;
275 struct osd_oi **od_oi_table;
276 /* total number of OI containers */
281 unsigned int od_fl_capa:1,
290 od_writethrough_cache:1,
292 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;
459 * Basic transaction credit op
480 #ifdef CONFIG_PROC_FS
482 LPROC_OSD_READ_BYTES = 0,
483 LPROC_OSD_WRITE_BYTES = 1,
484 LPROC_OSD_GET_PAGE = 2,
485 LPROC_OSD_NO_PAGE = 3,
486 LPROC_OSD_CACHE_ACCESS = 4,
487 LPROC_OSD_CACHE_HIT = 5,
488 LPROC_OSD_CACHE_MISS = 6,
490 #if OSD_THANDLE_STATS
491 LPROC_OSD_THANDLE_STARTING,
492 LPROC_OSD_THANDLE_OPEN,
493 LPROC_OSD_THANDLE_CLOSING,
500 * Storage representation for fids.
502 * Variable size, first byte contains the length of the whole record.
504 struct osd_fid_pack {
505 unsigned char fp_len;
506 char fp_area[sizeof(struct lu_fid)];
509 struct osd_it_ea_dirent {
510 struct lu_fid oied_fid;
513 unsigned short oied_namelen;
514 unsigned int oied_type;
516 } __attribute__((packed));
519 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
520 * than lu_dirent struct. osd readdir reads less number of dirent than
521 * required for mdd dir page. so buffer size need to be increased so that
522 * there would be one ext3 readdir for every mdd readdir page.
525 #define OSD_IT_EA_BUFSIZE (PAGE_SIZE + PAGE_SIZE/4)
528 * This is iterator's in-memory data structure in interoperability
529 * mode (i.e. iterator over ldiskfs style directory)
532 struct osd_object *oie_obj;
533 /** used in ldiskfs iterator, to stored file pointer */
534 struct file *oie_file;
535 /** how many entries have been read-cached from storage */
537 /** current entry is being iterated by caller */
539 /** current processing entry */
540 struct osd_it_ea_dirent *oie_dirent;
541 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
546 * Iterator's in-memory data structure for IAM mode.
549 struct osd_object *oi_obj;
550 struct iam_path_descr *oi_ipd;
551 struct iam_iterator oi_it;
554 struct osd_quota_leaf {
555 struct list_head oql_link;
560 * Iterator's in-memory data structure for quota file.
562 struct osd_it_quota {
563 struct osd_object *oiq_obj;
564 /** tree blocks path to where the entry is stored */
565 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
566 /** on-disk offset for current key where quota record can be found */
568 /** identifier for current quota record */
570 /** the record index in the leaf/index block */
571 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
572 /** list of already processed leaf blocks */
573 struct list_head oiq_list;
576 #define MAX_BLOCKS_PER_PAGE (PAGE_SIZE / 512)
579 wait_queue_head_t dr_wait;
580 atomic_t dr_numreqs; /* number of reqs being processed */
585 unsigned int dr_elapsed_valid:1; /* we really did count time */
586 unsigned int dr_rw:1;
587 struct lu_buf dr_pg_buf;
588 struct page **dr_pages;
589 struct niobuf_local **dr_lnbs;
590 struct lu_buf dr_bl_buf;
591 struct lu_buf dr_lnb_buf;
593 ktime_t dr_start_time;
594 ktime_t dr_elapsed; /* how long io took */
595 struct osd_device *dr_dev;
596 unsigned int dr_init_at; /* the line iobuf was initialized */
599 #define osd_dirty_inode(inode, flag) (inode)->i_sb->s_op->dirty_inode((inode), flag)
601 #ifndef HAVE_ALLOC_FILE_PSEUDO
603 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
604 (flag & __FMODE_NONOTIFY)))
606 struct file *alloc_file_pseudo(struct inode *inode, struct vfsmount *mnt,
607 const char *name, int flags,
608 const struct file_operations *fops)
610 struct qstr this = QSTR_INIT(name, strlen(name));
614 path.dentry = d_alloc_pseudo(mnt->mnt_sb, &this);
616 return ERR_PTR(-ENOMEM);
617 path.mnt = mntget(mnt);
618 d_instantiate(path.dentry, inode);
619 file = alloc_file(&path, OPEN_FMODE(flags), fops);
624 file->f_flags = flags;
628 #endif /* !HAVE_ALLOC_FILE_PSEUDO */
630 #ifdef HAVE_INODE_TIMESPEC64
631 # define osd_timespec timespec64
633 # define osd_timespec timespec
636 static inline struct osd_timespec osd_inode_time(struct inode *inode,
639 struct osd_timespec ts = { .tv_sec = seconds };
644 #define OSD_INS_CACHE_SIZE 8
646 struct osd_thread_info {
647 const struct lu_env *oti_env;
649 * used for index operations.
651 struct dentry oti_obj_dentry;
652 struct dentry oti_child_dentry;
654 /** dentry for Iterator context. */
655 struct dentry oti_it_dentry;
658 struct kstatfs oti_ksfs;
660 struct htree_lock *oti_hlock;
662 struct lu_fid oti_fid;
663 struct lu_fid oti_fid2;
664 struct lu_fid oti_fid3;
665 struct osd_inode_id oti_id;
666 struct osd_inode_id oti_id2;
667 struct osd_inode_id oti_id3;
668 struct ost_id oti_ostid;
671 * following ipd and it structures are used for osd_index_iam_lookup()
672 * these are defined separately as we might do index operation
673 * in open iterator session.
676 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
678 unsigned int oti_it_ea_buf_used:1;
680 /* IAM iterator for index operation. */
681 struct iam_iterator oti_idx_it;
683 /** union to guarantee that ->oti_ipd[] has proper alignment. */
686 char oti_it_ipd[DX_IPD_MAX_SIZE];
687 long long oti_alignment_lieutenant;
691 char oti_idx_ipd[DX_IPD_MAX_SIZE];
692 long long oti_alignment_lieutenant_colonel;
695 struct osd_idmap_cache oti_cache;
697 /* dedicated OI cache for insert (which needs inum) */
698 struct osd_idmap_cache *oti_ins_cache;
699 int oti_ins_cache_size;
700 int oti_ins_cache_used;
701 /* inc by osd_trans_create and dec by osd_trans_stop */
702 int oti_ins_cache_depth;
707 /** used in osd_fid_set() to put xattr */
708 struct lu_buf oti_buf;
709 struct lu_buf oti_big_buf;
710 /** used in osd_ea_fid_set() to set fid into common ea */
712 struct lustre_ost_attrs oti_ost_attrs;
713 struct filter_fid_18_23 oti_ff_old;
714 struct filter_fid oti_ff;
717 struct osd_iobuf oti_iobuf;
718 /* used to access objects in /O */
719 struct inode *oti_inode;
720 #define OSD_FID_REC_SZ 32
721 char oti_ldp[OSD_FID_REC_SZ];
722 char oti_ldp2[OSD_FID_REC_SZ];
724 /* used by quota code */
726 #if defined(HAVE_DQUOT_QC_DQBLK)
727 struct qc_dqblk oti_qdq;
729 struct fs_disk_quota oti_fdq;
731 struct if_dqinfo oti_dqinfo;
733 struct lquota_id_info oti_qi;
734 struct lquota_trans oti_quota_trans;
735 union lquota_rec oti_quota_rec;
737 struct lu_seq_range oti_seq_range;
739 /* Tracking for transaction credits, to allow debugging and optimizing
740 * cases where a large number of credits are being allocated for
741 * single transaction. */
742 unsigned int oti_credits_before;
743 unsigned int oti_declare_ops[OSD_OT_MAX];
744 unsigned int oti_declare_ops_cred[OSD_OT_MAX];
745 unsigned int oti_declare_ops_used[OSD_OT_MAX];
746 struct osd_directory oti_iam;
748 struct page **oti_dio_pages;
749 int oti_dio_pages_used;
751 struct osd_it_ea_dirent *oti_seq_dirent;
752 struct osd_it_ea_dirent *oti_dir_dirent;
755 extern int ldiskfs_pdo;
757 /* autoconf test is in lustre-build-ldiskfs.m4 */
758 #ifdef HAVE_BVEC_ITER_ALL
759 #define DECLARE_BVEC_ITER_ALL(iter) struct bvec_iter_all iter
761 #define DECLARE_BVEC_ITER_ALL(iter) int iter
764 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
765 const char *name, void *buf, int len)
770 dentry->d_inode = inode;
771 dentry->d_sb = inode->i_sb;
772 return ll_vfs_getxattr(dentry, inode, name, buf, len);
775 static inline int __osd_xattr_set(struct osd_thread_info *info,
776 struct inode *inode, const char *name,
777 const void *buf, int buflen, int fl)
779 struct dentry *dentry = &info->oti_child_dentry;
781 dquot_initialize(inode);
782 dentry->d_inode = inode;
783 dentry->d_sb = inode->i_sb;
784 return ll_vfs_setxattr(dentry, inode, name, buf, buflen, fl);
787 static inline char *osd_oid_name(char *name, size_t name_size,
788 const struct lu_fid *fid, u64 id)
790 snprintf(name, name_size,
791 (fid_seq_is_rsvd(fid_seq(fid)) ||
792 fid_seq_is_mdt0(fid_seq(fid)) ||
793 fid_seq_is_idif(fid_seq(fid))) ? "%llu" : "%llx", id);
798 #ifdef CONFIG_PROC_FS
800 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
801 int osd_procfs_init(struct osd_device *osd, const char *name);
802 int osd_procfs_fini(struct osd_device *osd);
803 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
804 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 0, 52, 0)
805 int osd_register_proc_index_in_idif(struct osd_device *osd);
809 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
810 struct obd_statfs *sfs, struct obd_statfs_info *info);
811 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
812 struct osd_inode_id *id);
814 osd_iget_fid(struct osd_thread_info *info, struct osd_device *dev,
815 struct osd_inode_id *id, struct lu_fid *fid);
816 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
817 const struct lu_fid *fid, __u32 compat, __u32 incompat);
818 struct osd_obj_seq *osd_seq_load(struct osd_thread_info *info,
819 struct osd_device *osd, u64 seq);
820 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
821 struct dentry *dentry, struct lustre_ost_attrs *loa);
822 void osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
823 struct osd_inode_id *id, const struct lu_fid *fid);
824 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
825 struct dentry *dentry, struct lu_fid *fid);
826 struct osd_it_ea *osd_it_dir_init(const struct lu_env *env,
827 struct osd_device *dev, struct inode *inode,
829 void osd_it_dir_fini(const struct lu_env *env, struct osd_it_ea *oie,
830 struct inode *inode);
831 int osd_ldiskfs_it_fill(const struct lu_env *env, const struct dt_it *di);
833 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
834 void osd_obj_map_fini(struct osd_device *dev);
835 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
836 const struct lu_fid *fid, struct osd_inode_id *id);
837 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
838 const struct lu_fid *fid, const struct osd_inode_id *id,
840 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
841 const struct lu_fid *fid, handle_t *th);
842 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
843 const struct lu_fid *fid, const struct osd_inode_id *id,
845 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
846 struct inode *src_parent, struct dentry *src_child,
847 const struct lu_fid *fid);
848 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
849 const struct lu_fid *fid, struct osd_inode_id *id,
850 enum oi_check_flags flags);
851 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
852 const struct lu_fid *fid, const struct osd_inode_id *id,
854 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
855 const struct lu_fid *fid, const struct osd_inode_id *id,
857 int osd_obj_del_entry(struct osd_thread_info *info, struct osd_device *osd,
858 struct dentry *dird, char *name, int namelen,
861 char *osd_lf_fid2name(const struct lu_fid *fid);
862 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
864 void osd_scrub_stop(struct osd_device *dev);
865 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev,
867 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
868 int osd_scrub_oi_insert(struct osd_device *dev, const struct lu_fid *fid,
869 struct osd_inode_id *id, int insert);
870 void osd_scrub_oi_resurrect(struct lustre_scrub *scrub,
871 const struct lu_fid *fid);
872 void osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
874 struct dentry *osd_lookup_one_len_unlocked(struct osd_device *dev,
876 struct dentry *base, int len);
877 struct dentry *osd_lookup_one_len(struct osd_device *dev, const char *name,
878 struct dentry *base, int len);
880 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
881 u64 seq, struct lu_seq_range *range);
883 int osd_delete_from_remote_parent(const struct lu_env *env,
884 struct osd_device *osd,
885 struct osd_object *obj,
886 struct osd_thandle *oh, bool destroy);
887 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
888 struct osd_object *obj, struct osd_thandle *oh);
889 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
890 struct osd_device *osd,
891 const struct lu_fid *fid,
892 struct osd_inode_id *id);
894 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
896 int osd_scrub_refresh_mapping(struct osd_thread_info *info,
897 struct osd_device *dev,
898 const struct lu_fid *fid,
899 const struct osd_inode_id *id,
901 enum oi_check_flags flags, bool *exist);
903 /* osd_quota_fmt.c */
904 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
905 int type, uint blk, int depth, uint index,
906 struct osd_it_quota *it);
907 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
908 int type, uint blk, uint index,
909 struct osd_it_quota *it);
910 loff_t find_tree_dqentry(const struct lu_env *env,
911 struct osd_object *obj, int type,
912 qid_t dqid, uint blk, int depth,
913 struct osd_it_quota *it);
915 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
916 struct lquota_id_info *qi, struct osd_object *obj,
917 bool enforce, enum osd_quota_local_flags *local_flags);
918 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
919 __u32 projid, long long space, struct osd_thandle *oh,
920 struct osd_object *obj,
921 enum osd_quota_local_flags *local_flags,
922 enum osd_qid_declare_flags);
923 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
924 const struct dt_rec *rec,
925 union lquota_rec *quota_rec);
926 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
928 #ifdef HAVE_PROJECT_QUOTA
929 static inline __u32 i_projid_read(struct inode *inode)
931 return (__u32)from_kprojid(&init_user_ns, LDISKFS_I(inode)->i_projid);
934 static inline void i_projid_write(struct inode *inode, __u32 projid)
937 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
938 LDISKFS_I(inode)->i_projid = kprojid;
941 static inline uid_t i_projid_read(struct inode *inode)
945 static inline void i_projid_write(struct inode *inode, __u32 projid)
951 #ifdef HAVE_LDISKFS_IGET_WITH_FLAGS
952 # define osd_ldiskfs_iget(sb, ino) \
953 ldiskfs_iget((sb), (ino), \
954 LDISKFS_IGET_HANDLE | LDISKFS_IGET_SPECIAL)
956 # define osd_ldiskfs_iget(sb, ino) ldiskfs_iget((sb), (ino))
959 #ifdef HAVE_LDISKFS_INFO_JINODE
960 # define osd_attach_jinode(inode) ldiskfs_inode_attach_jinode(inode)
961 #else /* HAVE_LDISKFS_INFO_JINODE */
962 # define osd_attach_jinode(inode) 0
963 #endif /* HAVE_LDISKFS_INFO_JINODE */
965 #ifndef HAVE_JBD2_JOURNAL_GET_MAX_TXN_BUFS
966 #define jbd2_journal_get_max_txn_bufs(jrnl) \
967 (jrnl->j_max_transaction_buffers)
970 #ifdef LDISKFS_HT_MISC
971 # define osd_journal_start_sb(sb, type, nblock) \
972 ldiskfs_journal_start_sb(sb, type, nblock)
973 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
975 ldiskfs_lblk_t *nblock)
979 rc = osd_attach_jinode(inode);
982 return ldiskfs_append(handle, inode, nblock);
985 # ifdef HAVE___LDISKFS_FIND_ENTRY
986 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
987 (__ldiskfs_find_entry(dir, name, de, inlined, lock) ?: \
990 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
991 (ldiskfs_find_entry_locked(dir, name, de, inlined, lock) ?: \
995 # define osd_journal_start(inode, type, nblocks) \
996 ldiskfs_journal_start(inode, type, nblocks)
997 # define osd_transaction_size(dev) \
998 (jbd2_journal_get_max_txn_bufs(osd_journal(dev)) / 2)
999 #else /* ! defined LDISKFS_HT_MISC */
1000 # define LDISKFS_HT_MISC 0
1001 # define osd_journal_start_sb(sb, type, nblock) \
1002 ldiskfs_journal_start_sb(sb, nblock)
1004 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
1005 struct inode *inode,
1006 ldiskfs_lblk_t *nblock)
1008 struct buffer_head *bh;
1011 bh = ldiskfs_append(handle, inode, nblock, &err);
1018 # ifdef HAVE___LDISKFS_FIND_ENTRY
1019 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1020 (__ldiskfs_find_entry(dir, name, de, lock) ?: \
1023 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
1024 (ldiskfs_find_entry_locked(dir, name, de, lock) ?: \
1027 # define osd_journal_start(inode, type, nblocks) \
1028 ldiskfs_journal_start(inode, nblocks)
1029 # define osd_transaction_size(dev) \
1030 jbd2_journal_get_max_txn_bufs(osd_journal(dev))
1031 #endif /* LDISKFS_HT_MISC */
1033 #ifndef HAVE___LDISKFS_FIND_ENTRY
1034 # define __ldiskfs_add_entry(handle, child, inode, hlock) \
1035 ldiskfs_add_entry_locked(handle, child, inode, hlock)
1039 * Invariants, assertions.
1043 * XXX: do not enable this, until invariant checking code is made thread safe
1044 * in the face of pdirops locking.
1046 #define OSD_INVARIANT_CHECKS (0)
1048 #if OSD_INVARIANT_CHECKS
1049 static inline int osd_invariant(const struct osd_object *obj)
1053 ergo(obj->oo_inode != NULL,
1054 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
1055 atomic_read(&obj->oo_inode->i_count) > 0) &&
1056 ergo(obj->oo_dir != NULL &&
1057 obj->oo_dir->od_conationer.ic_object != NULL,
1058 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
1061 #define osd_invariant(obj) (1)
1064 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
1065 #define OSD_READCACHE_MAX_IO_MB 8
1066 #define OSD_WRITECACHE_MAX_IO_MB 8
1068 extern const struct dt_index_operations osd_otable_ops;
1070 static inline int osd_oi_fid2idx(struct osd_device *dev,
1071 const struct lu_fid *fid)
1073 return fid->f_seq & (dev->od_oi_count - 1);
1076 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
1077 const struct lu_fid *fid)
1079 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
1080 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
1081 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
1082 DFID"\n", PFID(fid));
1083 /* It can work even od_oi_count equals to 1 although it's unexpected,
1084 * the only reason we set it to 1 is for performance measurement */
1085 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
1088 extern const struct lu_device_operations osd_lu_ops;
1090 static inline int lu_device_is_osd(const struct lu_device *d)
1092 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
1095 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
1097 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
1098 return container_of(d, struct osd_device, od_dt_dev);
1101 static inline struct osd_device *osd_dev(const struct lu_device *d)
1103 LASSERT(lu_device_is_osd(d));
1104 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
1107 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
1109 return osd_dev(o->oo_dt.do_lu.lo_dev);
1112 static inline struct super_block *osd_sb(const struct osd_device *dev)
1114 return dev->od_mnt->mnt_sb;
1117 static inline const char *osd_sb2name(const struct super_block *sb)
1119 /* this is LDISKFS_SB(sb), but preserves "const" */
1120 const struct ldiskfs_sb_info *sbi = sb->s_fs_info;
1122 return sbi->s_es->s_volume_name;
1125 static inline const char *osd_dev2name(const struct osd_device *dev)
1127 return osd_sb2name(osd_sb(dev));
1130 static inline const char *osd_ino2name(const struct inode *inode)
1132 return osd_sb2name(inode->i_sb);
1136 * Put the osd object once done with it.
1138 * \param obj osd object that needs to be put
1140 static inline void osd_object_put(const struct lu_env *env,
1141 struct osd_object *obj)
1143 dt_object_put(env, &obj->oo_dt);
1146 static inline int osd_object_is_root(const struct osd_object *obj)
1148 return obj->oo_inode && is_root_inode(obj->oo_inode);
1151 static inline struct osd_object *osd_obj(const struct lu_object *o)
1153 LASSERT(lu_device_is_osd(o->lo_dev));
1154 return container_of(o, struct osd_object, oo_dt.do_lu);
1157 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
1159 return osd_obj(&d->do_lu);
1162 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
1164 return &osd->od_dt_dev.dd_lu_dev;
1167 static inline journal_t *osd_journal(const struct osd_device *dev)
1169 return LDISKFS_SB(osd_sb(dev))->s_journal;
1172 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
1174 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
1177 static inline char *osd_name(struct osd_device *osd)
1179 return osd->od_svname;
1182 static inline bool osd_is_ea_inode(struct inode *inode)
1184 return !!(LDISKFS_I(inode)->i_flags & LDISKFS_EA_INODE_FL);
1187 extern const struct dt_body_operations osd_body_ops;
1188 extern struct lu_context_key osd_key;
1190 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
1192 return lu_context_key_get(&env->le_ctx, &osd_key);
1195 extern const struct dt_body_operations osd_body_ops_new;
1201 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
1202 const struct iam_container *bag)
1204 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1205 osd_oti_get(env)->oti_it_ipd);
1209 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
1210 const struct iam_container *bag)
1212 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1213 osd_oti_get(env)->oti_idx_ipd);
1216 static inline void osd_ipd_put(const struct lu_env *env,
1217 const struct iam_container *bag,
1218 struct iam_path_descr *ipd)
1220 bag->ic_descr->id_ops->id_ipd_free(ipd);
1223 int osd_calc_bkmap_credits(struct super_block *sb, struct inode *inode,
1224 const loff_t size, const loff_t pos,
1227 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
1230 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
1231 struct inode *inode,
1232 const char *name, const int namelen)
1234 struct osd_thread_info *info = osd_oti_get(env);
1235 struct dentry *child_dentry = &info->oti_child_dentry;
1236 struct dentry *obj_dentry = &info->oti_obj_dentry;
1238 obj_dentry->d_inode = inode;
1239 obj_dentry->d_sb = inode->i_sb;
1240 obj_dentry->d_name.hash = 0;
1242 child_dentry->d_name.hash = 0;
1243 child_dentry->d_parent = obj_dentry;
1244 child_dentry->d_name.name = name;
1245 child_dentry->d_name.len = namelen;
1246 return child_dentry;
1249 extern int osd_trans_declare_op2rb[];
1250 extern int ldiskfs_track_declares_assert;
1251 void osd_trans_dump_creds(const struct lu_env *env, struct thandle *th);
1253 static inline void osd_trans_declare_op(const struct lu_env *env,
1254 struct osd_thandle *oh,
1255 enum osd_op_type op, int credits)
1257 struct osd_thread_info *oti = osd_oti_get(env);
1259 LASSERT(oh->ot_handle == NULL);
1260 if (unlikely(op >= OSD_OT_MAX)) {
1261 if (unlikely(ldiskfs_track_declares_assert)) {
1262 LASSERT(op < OSD_OT_MAX);
1264 CWARN("%s: Invalid operation index %d\n",
1265 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op);
1266 libcfs_debug_dumpstack(NULL);
1269 oti->oti_declare_ops[op]++;
1270 oti->oti_declare_ops_cred[op] += credits;
1272 oh->ot_credits += credits;
1275 /* linux: v5.4-rc3-21-g933f1c1e0b75 renamed h_buffer_credits */
1276 #ifdef HAVE_JOURNAL_TOTAL_CREDITS
1277 #define h_buffer_credits h_total_credits
1280 static inline void osd_trans_exec_op(const struct lu_env *env,
1282 enum osd_op_type op)
1284 struct osd_thread_info *oti = osd_oti_get(env);
1285 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1287 unsigned int rb, left;
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: opcode %u: invalid value >= %u\n",
1295 osd_name(osd_dt_dev(oh->ot_super.th_dev)),
1297 libcfs_debug_dumpstack(NULL);
1302 /* find rollback (or reverse) operation for the given one
1303 * such an operation doesn't require additional credits
1304 * as the same set of blocks are modified */
1305 rb = osd_trans_declare_op2rb[op];
1307 /* check whether credits for this operation were reserved at all */
1308 if (unlikely(oti->oti_declare_ops_cred[op] == 0 &&
1309 oti->oti_declare_ops_cred[rb] == 0)) {
1310 /* the API is not perfect yet: CREATE does REF_ADD internally
1311 * while DESTROY does not. To rollback CREATE the callers
1312 * needs to call REF_DEL+DESTROY which is hard to detect using
1313 * a simple table of rollback operations */
1314 if (op == OSD_OT_REF_DEL &&
1315 oti->oti_declare_ops_cred[OSD_OT_CREATE] > 0)
1317 if (op == OSD_OT_REF_ADD &&
1318 oti->oti_declare_ops_cred[OSD_OT_DESTROY] > 0)
1320 CWARN("%s: opcode %u: credits = 0, rollback = %u\n",
1321 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
1322 osd_trans_dump_creds(env, th);
1323 LASSERT(!ldiskfs_track_declares_assert);
1327 /* remember how many credits we have unused before the operation */
1328 oti->oti_credits_before = oh->ot_handle->h_buffer_credits;
1329 left = oti->oti_declare_ops_cred[op] - oti->oti_declare_ops_used[op];
1330 if (unlikely(oti->oti_credits_before < left)) {
1331 CWARN("%s: opcode %u: before %u < left %u, rollback = %u\n",
1332 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1333 oti->oti_credits_before, left, rb);
1334 osd_trans_dump_creds(env, th);
1335 /* on a very small fs (testing?) it's possible that
1336 * the transaction can't fit 1/4 of journal, so we
1337 * just request less credits (see osd_trans_start()).
1338 * ignore the same case here */
1339 rb = osd_transaction_size(osd_dt_dev(th->th_dev));
1340 if (unlikely(oh->ot_credits < rb))
1341 LASSERT(!ldiskfs_track_declares_assert);
1345 static inline void osd_trans_exec_check(const struct lu_env *env,
1347 enum osd_op_type op)
1349 struct osd_thread_info *oti = osd_oti_get(env);
1350 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1352 int used, over, quota;
1354 /* how many credits have been used by the operation */
1355 used = oti->oti_credits_before - oh->ot_handle->h_buffer_credits;
1357 if (unlikely(used < 0)) {
1358 /* if some block was allocated and released in the same
1359 * transaction, then it won't be a part of the transaction
1360 * and delta can be negative */
1365 /* rollback operations (e.g. when we destroy just created
1366 * object) should not consume any credits. there is no point
1367 * to confuse the checks below */
1371 oti->oti_declare_ops_used[op] += used;
1372 if (oti->oti_declare_ops_used[op] <= oti->oti_declare_ops_cred[op])
1375 /* we account quota for a whole transaction and any operation can
1376 * consume corresponding credits */
1377 over = oti->oti_declare_ops_used[op] -
1378 oti->oti_declare_ops_cred[op];
1379 quota = oti->oti_declare_ops_cred[OSD_OT_QUOTA] -
1380 oti->oti_declare_ops_used[OSD_OT_QUOTA];
1381 if (over <= quota) {
1382 /* probably that credits were consumed by
1383 * quota indirectly (in the depths of ldiskfs) */
1384 oti->oti_declare_ops_used[OSD_OT_QUOTA] += over;
1385 oti->oti_declare_ops_used[op] -= over;
1387 CWARN("%s: opcode %d: used %u, used now %u, reserved %u\n",
1388 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1389 oti->oti_declare_ops_used[op], used,
1390 oti->oti_declare_ops_cred[op]);
1391 osd_trans_dump_creds(env, th);
1392 if (unlikely(ldiskfs_track_declares_assert))
1398 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1401 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1402 struct lu_fid *befider)
1404 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1405 memcpy(pack->fp_area, befider, sizeof(*befider));
1406 pack->fp_len = sizeof(*befider) + 1;
1410 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1415 switch (pack->fp_len) {
1416 case sizeof *fid + 1:
1417 memcpy(fid, pack->fp_area, sizeof *fid);
1418 fid_be_to_cpu(fid, fid);
1421 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1428 * Quota/Accounting handling
1430 extern const struct dt_index_operations osd_acct_index_ops;
1431 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1432 const struct lu_fid *fid, struct osd_inode_id *id);
1434 /* copy from fs/ext4/dir.c */
1435 static inline int is_32bit_api(void)
1437 #ifdef CONFIG_COMPAT
1438 return in_compat_syscall();
1440 return (BITS_PER_LONG == 32);
1444 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1446 if ((filp->f_mode & FMODE_32BITHASH) ||
1447 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1448 return LDISKFS_HTREE_EOF_32BIT;
1450 return LDISKFS_HTREE_EOF_64BIT;
1453 static inline int fid_is_internal(const struct lu_fid *fid)
1455 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1458 static inline bool is_remote_parent_ino(struct osd_device *o, unsigned long ino)
1463 LASSERT(o->od_mdt_map != NULL);
1465 return ino == o->od_mdt_map->omm_remote_parent->d_inode->i_ino;
1469 * ext4_bread/ldiskfs_bread has either 5 or 4 parameters. The error
1470 * return code has been removed and integrated into the pointer in the
1473 static inline struct buffer_head *__ldiskfs_bread(handle_t *handle,
1474 struct inode *inode,
1475 ldiskfs_lblk_t block,
1479 struct buffer_head *bh;
1482 rc = osd_attach_jinode(inode);
1486 #ifdef HAVE_EXT4_BREAD_4ARGS
1487 bh = ldiskfs_bread(handle, inode, block, create);
1490 bh = ldiskfs_bread(handle, inode, block, create, &rc);
1491 if (bh == NULL && rc != 0)
1497 #ifndef HAVE_BIO_INTEGRITY_ENABLED
1498 bool bio_integrity_enabled(struct bio *bio);
1502 # define bio_get_dev(bio) ((bio)->bi_bdev)
1503 # define bio_get_disk(bio) (bio_get_dev(bio)->bd_disk)
1504 # define bio_get_queue(bio) bdev_get_queue(bio_get_dev(bio))
1506 # ifndef HAVE_BIO_SET_DEV
1507 # define bio_set_dev(bio, bdev) (bio_get_dev(bio) = (bdev))
1510 # define bio_get_disk(bio) ((bio)->bi_disk)
1511 # define bio_get_queue(bio) (bio_get_disk(bio)->queue)
1514 #ifdef HAVE_EXT4_JOURNAL_GET_WRITE_ACCESS_4ARGS
1515 # define osd_ldiskfs_journal_get_write_access(handle, sb, bh, flags) \
1516 ldiskfs_journal_get_write_access((handle), (sb), (bh), (flags))
1518 # define LDISKFS_JTR_NONE 0
1519 # define osd_ldiskfs_journal_get_write_access(handle, sb, bh, flags) \
1520 ldiskfs_journal_get_write_access((handle), (bh))
1521 #endif /* HAVE_EXT4_JOURNAL_GET_WRITE_ACCESS_4ARGS */
1523 #ifdef HAVE_EXT4_INC_DEC_COUNT_2ARGS
1524 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((h), (inode))
1525 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((h), (inode))
1527 #define osd_ldiskfs_inc_count(h, inode) ldiskfs_inc_count((inode))
1528 #define osd_ldiskfs_dec_count(h, inode) ldiskfs_dec_count((inode))
1529 #endif /* HAVE_EXT4_INC_DEC_COUNT_2ARGS */
1531 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1534 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1535 __u32 keysize, __u32 recsize)
1537 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1538 &osd->od_index_backup_list, &osd->od_lock,
1539 &osd->od_index_backup_stop,
1540 fid, keysize, recsize);
1544 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1546 struct osd_thread_info *info = osd_oti_get(env);
1547 struct lu_fid *fid = &info->oti_fid3;
1548 struct osd_inode_id *id = &info->oti_id3;
1550 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1551 osd_id_gen(id, osd->od_index_backup_inode->i_ino,
1552 osd->od_index_backup_inode->i_generation);
1553 osd_add_oi_cache(info, osd, id, fid);
1555 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1556 &osd->od_index_backup_list, &osd->od_lock,
1557 &osd->od_index_backup_stop, backup);
1560 #ifdef LDISKFS_HAS_INCOMPAT_FEATURE
1562 # ifdef LDISKFS_FEATURE_INCOMPAT_EXTENTS
1563 # define ldiskfs_has_feature_extents(sb) \
1564 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EXTENTS)
1566 # ifdef LDISKFS_FEATURE_INCOMPAT_EA_INODE
1567 # define ldiskfs_has_feature_ea_inode(sb) \
1568 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EA_INODE)
1570 # ifdef LDISKFS_FEATURE_INCOMPAT_DIRDATA
1571 # define ldiskfs_has_feature_dirdata(sb) \
1572 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_DIRDATA)
1574 # ifdef LDISKFS_FEATURE_INCOMPAT_LARGEDIR
1575 # define ldiskfs_has_feature_largedir(sb) \
1576 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1577 # define ldiskfs_set_feature_largedir(sb) \
1578 LDISKFS_SET_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1580 # ifdef LDISKFS_FEATURE_COMPAT_HAS_JOURNAL
1581 # define ldiskfs_has_feature_journal(sb) \
1582 LDISKFS_HAS_COMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_HAS_JOURNAL)
1584 # ifdef LDISKFS_FEATURE_RO_COMPAT_QUOTA
1585 # define ldiskfs_has_feature_quota(sb) \
1586 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_QUOTA)
1588 # ifdef LDISKFS_FEATURE_RO_COMPAT_PROJECT
1589 # define ldiskfs_has_feature_project(sb) \
1590 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_PROJECT)
1592 # ifdef LDISKFS_FEATURE_COMPAT_FAST_COMMIT
1593 # define ldiskfs_has_feature_fast_commit(sb) \
1594 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_FAST_COMMIT)
1598 #ifndef ldiskfs_has_feature_fast_commit
1599 #define ldiskfs_has_feature_fast_commit(sb) false
1602 int osd_trunc_lock(struct osd_object *obj, struct osd_thandle *oh,
1604 void osd_trunc_unlock_all(const struct lu_env *env, struct list_head *list);
1605 int osd_process_truncates(const struct lu_env *env, struct list_head *list);
1606 void osd_execute_truncate(struct osd_object *obj);
1608 #ifdef HAVE_BIO_ENDIO_USES_ONE_ARG
1609 #define osd_dio_complete_routine(bio, error) dio_complete_routine(bio)
1611 #define osd_dio_complete_routine(bio, error) dio_complete_routine(bio, error)
1614 #ifndef HAVE___BI_CNT
1615 #define __bi_cnt bi_cnt
1619 #define bi_opf bi_rw
1622 #ifndef HAVE_CLEAN_BDEV_ALIASES
1623 #define clean_bdev_aliases(bdev, block, len) \
1624 unmap_underlying_metadata((bdev), (block))
1627 #ifndef HAVE_BI_STATUS
1628 #define bi_status bi_error
1632 * Maximum size of xattr attributes for FEATURE_INCOMPAT_EA_INODE 1Mb
1633 * This limit is arbitrary, but is reasonable for the xattr API.
1635 #define LDISKFS_XATTR_MAX_LARGE_EA_SIZE (1024 * 1024)
1637 struct osd_bio_private {
1638 struct osd_iobuf *obp_iobuf;
1639 /* Start page index in the obp_iobuf for the bio */
1640 int obp_start_page_idx;
1643 #ifdef HAVE_BIO_INTEGRITY_PREP_FN
1644 # ifdef HAVE_BLK_INTEGRITY_ITER
1645 # define integrity_gen_fn integrity_processing_fn
1646 # define integrity_vrfy_fn integrity_processing_fn
1648 int osd_get_integrity_profile(struct osd_device *osd,
1649 integrity_gen_fn **generate_fn,
1650 integrity_vrfy_fn **verify_fn);
1651 #endif /* HAVE_BIO_INTEGRITY_PREP_FN */
1653 #ifdef HAVE_BIO_BI_PHYS_SEGMENTS
1654 #define osd_bio_nr_segs(bio) ((bio)->bi_phys_segments)
1656 #define osd_bio_nr_segs(bio) bio_segments((bio))
1657 #endif /* HAVE_BIO_BI_PHYS_SEGMENTS */
1659 #ifdef HAVE_GET_INODE_USAGE
1660 #define lock_dquot_transfer(inode) down_read(&LDISKFS_I(inode)->xattr_sem)
1661 #define unlock_dquot_transfer(inode) up_read(&LDISKFS_I(inode)->xattr_sem)
1663 #define lock_dquot_transfer(inode) do {} while (0)
1664 #define unlock_dquot_transfer(inode) do {} while (0)
1667 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)
1668 static inline unsigned short blk_integrity_interval(struct blk_integrity *bi)
1670 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1671 return bi->interval_exp ? 1 << bi->interval_exp : 0;
1672 #elif defined(HAVE_INTERVAL_BLK_INTEGRITY)
1673 return bi->interval;
1675 return bi->sector_size;
1676 #endif /* !HAVE_INTERVAL_EXP_BLK_INTEGRITY */
1679 static inline const char *blk_integrity_name(struct blk_integrity *bi)
1681 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1682 return bi->profile->name;
1688 static inline unsigned int bip_size(struct bio_integrity_payload *bip)
1690 #ifdef HAVE_BIP_ITER_BIO_INTEGRITY_PAYLOAD
1691 return bip->bip_iter.bi_size;
1693 return bip->bip_size;
1696 #else /* !CONFIG_BLK_DEV_INTEGRITY */
1697 static inline unsigned short blk_integrity_interval(struct blk_integrity *bi)
1701 static inline const char *blk_integrity_name(struct blk_integrity *bi)
1703 /* gcc8 dislikes when strcmp() is called against NULL */
1706 #endif /* !CONFIG_BLK_DEV_INTEGRITY */
1708 #ifndef INTEGRITY_FLAG_READ
1709 #define INTEGRITY_FLAG_READ BLK_INTEGRITY_VERIFY
1712 #ifndef INTEGRITY_FLAG_WRITE
1713 #define INTEGRITY_FLAG_WRITE BLK_INTEGRITY_GENERATE
1716 static inline bool bdev_integrity_enabled(struct block_device *bdev, int rw)
1718 #if IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)
1719 struct blk_integrity *bi = bdev_get_integrity(bdev);
1724 #ifdef HAVE_INTERVAL_EXP_BLK_INTEGRITY
1725 if (rw == 0 && bi->profile->verify_fn != NULL &&
1726 (bi->flags & INTEGRITY_FLAG_READ))
1729 if (rw == 1 && bi->profile->generate_fn != NULL &&
1730 (bi->flags & INTEGRITY_FLAG_WRITE))
1733 if (rw == 0 && bi->verify_fn != NULL &&
1734 (bi->flags & INTEGRITY_FLAG_READ))
1737 if (rw == 1 && bi->generate_fn != NULL &&
1738 (bi->flags & INTEGRITY_FLAG_WRITE))
1740 #endif /* !HAVE_INTERVAL_EXP_BLK_INTEGRITY */
1741 #endif /* !CONFIG_BLK_DEV_INTEGRITY */
1746 #endif /* _OSD_INTERNAL_H */