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 <ldiskfs/ldiskfs.h>
53 #include <ldiskfs/ldiskfs_jbd2.h>
57 /* class_register_type(), class_unregister_type() */
58 #include <obd_class.h>
59 #include <lustre_disk.h>
60 #include <dt_object.h>
61 #include <lustre_quota.h>
65 #include "osd_scrub.h"
66 #include "osd_quota_fmt.h"
69 extern struct kmem_cache *dynlock_cachep;
71 #define OSD_COUNTERS (0)
73 /* ldiskfs special inode::i_state_flags need to be accessed with
74 * ldiskfs_{set,clear,test}_inode_state() only */
76 /* OI scrub should skip this inode. */
77 #define LDISKFS_STATE_LUSTRE_NOSCRUB 31
78 #define LDISKFS_STATE_LUSTRE_DESTROY 30
80 /** Enable thandle usage statistics */
81 #define OSD_THANDLE_STATS (0)
83 #define MAX_OBJID_GROUP (FID_SEQ_ECHO + 1)
85 #define OBJECTS "OBJECTS"
86 #define ADMIN_USR "admin_quotafile_v2.usr"
87 #define ADMIN_GRP "admin_quotafile_v2.grp"
89 /* Statfs space reservation for fragmentation and local objects */
90 #define OSD_STATFS_RESERVED (1ULL << 23) /* 8MB */
91 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
93 /* Default extent bytes when declaring write commit */
94 #define OSD_DEFAULT_EXTENT_BYTES (1U << 20)
96 /* check if ldiskfs support project quota */
97 #ifndef LDISKFS_IOC_FSSETXATTR
98 #undef HAVE_PROJECT_QUOTA
101 #define OBD_BRW_MAPPED OBD_BRW_LOCAL1
103 struct osd_directory {
104 struct iam_container od_container;
105 struct iam_descr od_descr;
109 * Object Index (oi) instance.
113 * underlying index object, where fid->id mapping in stored.
115 struct inode *oi_inode;
116 struct osd_directory oi_dir;
119 extern const int osd_dto_credits_noquota[];
122 struct dt_object oo_dt;
124 * Inode for file system object represented by this osd_object. This
125 * inode is pinned for the whole duration of lu_object life.
127 * Not modified concurrently (either setup early during object
128 * creation, or assigned by osd_create() under write lock).
130 struct inode *oo_inode;
132 * to protect index ops.
134 struct htree_lock_head *oo_hl_head;
135 struct rw_semaphore oo_ext_idx_sem;
136 struct rw_semaphore oo_sem;
137 struct osd_directory *oo_dir;
138 /** protects inode attributes. */
142 * Following two members *compat_dot* are used to indicate
143 * the presence of dot and dotdot in the given directory.
144 * This is required for interop mode (b11826).
146 __u32 oo_destroyed:1,
148 oo_compat_dot_created:1,
149 oo_compat_dotdot_created:1;
151 /* the i_flags in LMA */
154 const struct lu_env *oo_owner;
156 struct list_head oo_xattr_list;
157 struct lu_object_header *oo_header;
158 __u64 oo_dirent_count;
162 /* protects on-fly initialization */
163 int oos_subdir_count; /* subdir count for each seq */
164 struct dentry *oos_root; /* O/<seq> */
165 struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
166 u64 oos_seq; /* seq number */
167 struct list_head oos_seq_list; /* list to seq_list */
171 struct dentry *om_root; /* dentry for /O */
172 rwlock_t om_seq_list_lock; /* lock for seq_list */
173 struct list_head om_seq_list; /* list head for seq */
175 struct mutex om_dir_init_mutex;
179 struct dentry *om_root; /* AGENT/<index> */
180 u64 om_index; /* mdt index */
181 struct list_head om_list; /* list to omm_list */
184 struct osd_mdobj_map {
185 struct dentry *omm_remote_parent;
187 int osd_ldiskfs_add_entry(struct osd_thread_info *info, struct osd_device *osd,
188 handle_t *handle, struct dentry *child,
189 struct inode *inode, struct htree_lock *hlock);
191 #define OSD_OTABLE_IT_CACHE_SIZE 64
192 #define OSD_OTABLE_IT_CACHE_MASK (~(OSD_OTABLE_IT_CACHE_SIZE - 1))
194 struct osd_inconsistent_item {
195 /* link into lustre_scrub::os_inconsistent_items,
196 * protected by lustre_scrub::os_lock. */
197 struct list_head oii_list;
199 /* The right FID <=> ino#/gen mapping. */
200 struct osd_idmap_cache oii_cache;
202 unsigned int oii_insert:1; /* insert or update mapping. */
205 struct osd_otable_cache {
206 struct osd_idmap_cache ooc_cache[OSD_OTABLE_IT_CACHE_SIZE];
208 /* Index for next cache slot to be filled. */
209 int ooc_producer_idx;
211 /* Index for next cache slot to be returned by it::next(). */
212 int ooc_consumer_idx;
214 /* How many items in ooc_cache. */
215 __u64 ooc_cached_items;
217 /* Position for up layer LFSCK iteration pre-loading. */
218 __u64 ooc_pos_preload;
221 struct osd_otable_it {
222 struct osd_device *ooi_dev;
223 struct osd_otable_cache ooi_cache;
224 struct osd_iit_param ooi_iit_param;
226 /* The following bits can be updated/checked w/o lock protection.
227 * If more bits will be introduced in the future and need lock to
228 * protect, please add comment. */
229 unsigned long ooi_used_outside:1, /* Some user out of OSD
230 * uses the iteration. */
231 ooi_all_cached:1, /* No more entries can be
232 * filled into cache. */
233 ooi_user_ready:1, /* The user out of OSD is
234 * ready to iterate. */
235 ooi_waiting:1; /* it::next is waiting. */
238 struct osd_obj_orphan {
239 struct list_head oor_list;
240 struct lu_env *oor_env; /* to identify "own" records */
245 OSD_T10_TYPE_UNKNOWN = 0,
248 OSD_T10_TYPE_CRC = 0x4,
249 OSD_T10_TYPE1_CRC = OSD_T10_TYPE1 | OSD_T10_TYPE_CRC,
250 OSD_T10_TYPE3_CRC = OSD_T10_TYPE3 | OSD_T10_TYPE_CRC,
251 OSD_T10_TYPE1_IP = OSD_T10_TYPE1,
252 OSD_T10_TYPE3_IP = OSD_T10_TYPE3,
260 struct dt_device od_dt_dev;
261 /* information about underlying file system */
262 struct vfsmount *od_mnt;
264 struct osd_oi **od_oi_table;
265 /* total number of OI containers */
270 unsigned int od_fl_capa:1,
279 od_writethrough_cache:1,
283 __s64 od_auto_scrub_interval;
284 __u32 od_dirent_journal;
286 struct proc_dir_entry *od_proc_entry;
287 struct lprocfs_stats *od_stats;
289 spinlock_t od_osfs_lock;
291 int od_fallocate_zero_blocks;
293 struct lu_site od_site;
295 struct osd_obj_map *od_ost_map;
296 struct osd_mdobj_map *od_mdt_map;
298 /* objects with size > od_readcache_max_filesize will be
299 * served bypassing pagecache unless already cached */
300 unsigned long long od_readcache_max_filesize;
302 /* reads > od_readcache_max_iosize will be
303 * served bypassing pagecache unless already cached */
304 unsigned long od_readcache_max_iosize;
306 /* writes > od_writethough_max_iosize will be
307 * served bypassing pagecache unless already cached */
308 unsigned long od_writethrough_max_iosize;
310 struct brw_stats od_brw_stats;
311 atomic_t od_r_in_flight;
312 atomic_t od_w_in_flight;
314 struct mutex od_otable_mutex;
315 struct osd_otable_it *od_otable_it;
316 struct osd_scrub od_scrub;
317 struct list_head od_ios_list;
319 /* service name associated with the osd device */
320 char od_svname[MAX_OBD_NAME];
321 char od_mntdev[MAX_OBD_NAME];
324 /* quota slave instance for inode */
325 struct qsd_instance *od_quota_slave_md;
327 /* quota slave instance for block */
328 struct qsd_instance *od_quota_slave_dt;
330 /* osd seq instance */
331 struct lu_client_seq *od_cl_seq;
332 /* If the ratio of "the total OI mappings count" vs
333 * "the bad OI mappings count" is lower than the
334 * osd_device::od_full_scrub_ratio, then trigger
335 * OI scrub to scan the whole the device. */
336 __u64 od_full_scrub_ratio;
337 /* If the speed of found bad OI mappings (per minute)
338 * exceeds the osd_device::od_full_scrub_threshold_rate,
339 * then trigger OI scrub to scan the whole device. */
340 __u64 od_full_scrub_threshold_rate;
342 /* a list of orphaned agent inodes, protected with od_osfs_lock */
343 struct list_head od_orphan_list;
344 struct list_head od_index_backup_list;
345 struct list_head od_index_restore_list;
347 struct inode *od_index_backup_inode;
348 enum lustre_index_backup_policy od_index_backup_policy;
349 int od_index_backup_stop;
350 /* T10PI type, zero if not supported */
351 enum osd_t10_type od_t10_type;
352 atomic_t od_commit_cb_in_flight;
353 wait_queue_head_t od_commit_cb_done;
354 unsigned int __percpu *od_extent_bytes_percpu;
357 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
360 return osd->od_quota_slave_dt;
362 return osd->od_quota_slave_md;
365 enum osd_full_scrub_ratio {
366 /* Trigger OI scrub to scan the whole device directly. */
369 /* Because the bad OI mappings count cannot be larger than
370 * the total OI mappints count, then setting OFSR_NEVER means
371 * that the whole device scanning cannot be triggered by auto
372 * detected bad OI mappings during the RPC services. */
374 OFSR_DEFAULT = 10000,
377 #define FULL_SCRUB_THRESHOLD_RATE_DEFAULT 60
379 /* There are at most 15 uid/gid/projids are affected in a transaction, and
380 * that's rename case:
381 * - 3 for source parent uid & gid & projid;
382 * - 3 for source child uid & gid & projid ('..' entry update when
383 * child is directory);
384 * - 3 for target parent uid & gid & projid;
385 * - 3 for target child uid & gid & projid(if the target child exists);
386 * - 3 for root uid & gid(last_rcvd, llog, etc);
389 #define OSD_MAX_UGID_CNT 15
407 struct osd_access_lock {
408 struct list_head tl_list;
409 struct osd_object *tl_obj;
419 struct thandle ot_super;
421 struct ldiskfs_journal_cb_entry ot_jcb;
422 struct list_head ot_commit_dcb_list;
423 struct list_head ot_stop_dcb_list;
424 /* Link to the device, for debugging. */
425 struct lu_ref_link ot_dev_link;
426 unsigned int ot_credits;
427 unsigned int oh_declared_ext;
429 /* quota IDs related to the transaction */
430 unsigned short ot_id_cnt;
431 __u8 ot_id_res[OSD_MAX_UGID_CNT];
432 __u8 ot_id_types[OSD_MAX_UGID_CNT];
433 uid_t ot_id_array[OSD_MAX_UGID_CNT];
434 struct lquota_trans *ot_quota_trans;
436 unsigned int ot_remove_agents:1;
437 #if OSD_THANDLE_STATS
438 /** time when this handle was allocated */
441 /** time when this thanle was started */
444 struct list_head ot_trunc_locks;
448 * Basic transaction credit op
469 #ifdef CONFIG_PROC_FS
471 LPROC_OSD_READ_BYTES = 0,
472 LPROC_OSD_WRITE_BYTES = 1,
473 LPROC_OSD_GET_PAGE = 2,
474 LPROC_OSD_NO_PAGE = 3,
475 LPROC_OSD_CACHE_ACCESS = 4,
476 LPROC_OSD_CACHE_HIT = 5,
477 LPROC_OSD_CACHE_MISS = 6,
479 #if OSD_THANDLE_STATS
480 LPROC_OSD_THANDLE_STARTING,
481 LPROC_OSD_THANDLE_OPEN,
482 LPROC_OSD_THANDLE_CLOSING,
489 * Storage representation for fids.
491 * Variable size, first byte contains the length of the whole record.
493 struct osd_fid_pack {
494 unsigned char fp_len;
495 char fp_area[sizeof(struct lu_fid)];
498 struct osd_it_ea_dirent {
499 struct lu_fid oied_fid;
502 unsigned short oied_namelen;
503 unsigned int oied_type;
505 } __attribute__((packed));
508 * as osd_it_ea_dirent (in memory dirent struct for osd) is greater
509 * than lu_dirent struct. osd readdir reads less number of dirent than
510 * required for mdd dir page. so buffer size need to be increased so that
511 * there would be one ext3 readdir for every mdd readdir page.
514 #define OSD_IT_EA_BUFSIZE (PAGE_SIZE + PAGE_SIZE/4)
517 * This is iterator's in-memory data structure in interoperability
518 * mode (i.e. iterator over ldiskfs style directory)
521 struct osd_object *oie_obj;
522 /** used in ldiskfs iterator, to stored file pointer */
523 struct file oie_file;
524 /** how many entries have been read-cached from storage */
526 /** current entry is being iterated by caller */
528 /** current processing entry */
529 struct osd_it_ea_dirent *oie_dirent;
530 /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
532 struct dentry oie_dentry;
536 * Iterator's in-memory data structure for IAM mode.
539 struct osd_object *oi_obj;
540 struct iam_path_descr *oi_ipd;
541 struct iam_iterator oi_it;
544 struct osd_quota_leaf {
545 struct list_head oql_link;
550 * Iterator's in-memory data structure for quota file.
552 struct osd_it_quota {
553 struct osd_object *oiq_obj;
554 /** tree blocks path to where the entry is stored */
555 uint oiq_blk[LUSTRE_DQTREEDEPTH + 1];
556 /** on-disk offset for current key where quota record can be found */
558 /** identifier for current quota record */
560 /** the record index in the leaf/index block */
561 uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
562 /** list of already processed leaf blocks */
563 struct list_head oiq_list;
566 #define MAX_BLOCKS_PER_PAGE (PAGE_SIZE / 512)
569 wait_queue_head_t dr_wait;
570 atomic_t dr_numreqs; /* number of reqs being processed */
575 unsigned int dr_elapsed_valid:1; /* we really did count time */
576 unsigned int dr_rw:1;
577 struct lu_buf dr_pg_buf;
578 struct page **dr_pages;
579 struct niobuf_local **dr_lnbs;
580 struct lu_buf dr_bl_buf;
581 struct lu_buf dr_lnb_buf;
583 ktime_t dr_start_time;
584 ktime_t dr_elapsed; /* how long io took */
585 struct osd_device *dr_dev;
586 unsigned int dr_init_at; /* the line iobuf was initialized */
589 int osd_security_file_alloc(struct file *file);
591 #define osd_dirty_inode(inode, flag) (inode)->i_sb->s_op->dirty_inode((inode), flag)
593 #ifdef HAVE_INODE_TIMESPEC64
594 # define osd_timespec timespec64
596 # define osd_timespec timespec
599 static inline struct osd_timespec osd_inode_time(struct inode *inode,
602 struct osd_timespec ts = { .tv_sec = seconds };
607 #define OSD_INS_CACHE_SIZE 8
609 struct osd_thread_info {
610 const struct lu_env *oti_env;
612 * used for index operations.
614 struct dentry oti_obj_dentry;
615 struct dentry oti_child_dentry;
617 /** dentry for Iterator context. */
618 struct dentry oti_it_dentry;
621 /* fake struct file for osd_object_sync */
622 struct file oti_file;
624 struct kstatfs oti_ksfs;
627 struct htree_lock *oti_hlock;
629 struct lu_fid oti_fid;
630 struct lu_fid oti_fid2;
631 struct lu_fid oti_fid3;
632 struct osd_inode_id oti_id;
633 struct osd_inode_id oti_id2;
634 struct osd_inode_id oti_id3;
635 struct ost_id oti_ostid;
638 * following ipd and it structures are used for osd_index_iam_lookup()
639 * these are defined separately as we might do index operation
640 * in open iterator session.
643 /** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
645 unsigned int oti_it_ea_buf_used:1;
647 /* IAM iterator for index operation. */
648 struct iam_iterator oti_idx_it;
650 /** union to guarantee that ->oti_ipd[] has proper alignment. */
653 char oti_it_ipd[DX_IPD_MAX_SIZE];
654 long long oti_alignment_lieutenant;
658 char oti_idx_ipd[DX_IPD_MAX_SIZE];
659 long long oti_alignment_lieutenant_colonel;
662 struct osd_idmap_cache oti_cache;
664 /* dedicated OI cache for insert (which needs inum) */
665 struct osd_idmap_cache *oti_ins_cache;
666 int oti_ins_cache_size;
667 int oti_ins_cache_used;
668 /* inc by osd_trans_create and dec by osd_trans_stop */
669 int oti_ins_cache_depth;
674 /** used in osd_fid_set() to put xattr */
675 struct lu_buf oti_buf;
676 struct lu_buf oti_big_buf;
677 /** used in osd_ea_fid_set() to set fid into common ea */
679 struct lustre_ost_attrs oti_ost_attrs;
680 struct filter_fid_18_23 oti_ff_old;
681 struct filter_fid oti_ff;
684 struct osd_iobuf oti_iobuf;
685 /* used to access objects in /O */
686 struct inode *oti_inode;
687 #define OSD_FID_REC_SZ 32
688 char oti_ldp[OSD_FID_REC_SZ];
689 char oti_ldp2[OSD_FID_REC_SZ];
691 /* used by quota code */
693 #if defined(HAVE_DQUOT_QC_DQBLK)
694 struct qc_dqblk oti_qdq;
696 struct fs_disk_quota oti_fdq;
698 struct if_dqinfo oti_dqinfo;
700 struct lquota_id_info oti_qi;
701 struct lquota_trans oti_quota_trans;
702 union lquota_rec oti_quota_rec;
704 struct lu_seq_range oti_seq_range;
706 /* Tracking for transaction credits, to allow debugging and optimizing
707 * cases where a large number of credits are being allocated for
708 * single transaction. */
709 unsigned int oti_credits_before;
710 unsigned int oti_declare_ops[OSD_OT_MAX];
711 unsigned int oti_declare_ops_cred[OSD_OT_MAX];
712 unsigned int oti_declare_ops_used[OSD_OT_MAX];
713 struct osd_directory oti_iam;
715 struct page **oti_dio_pages;
716 int oti_dio_pages_used;
718 struct osd_it_ea_dirent *oti_seq_dirent;
719 struct osd_it_ea_dirent *oti_dir_dirent;
722 extern int ldiskfs_pdo;
724 /* autoconf test is in lustre-build-ldiskfs.m4 */
725 #ifdef HAVE_BVEC_ITER_ALL
726 #define DECLARE_BVEC_ITER_ALL(iter) struct bvec_iter_all iter
728 #define DECLARE_BVEC_ITER_ALL(iter) int iter
731 static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
732 const char *name, void *buf, int len)
737 dentry->d_inode = inode;
738 dentry->d_sb = inode->i_sb;
739 return ll_vfs_getxattr(dentry, inode, name, buf, len);
742 static inline int __osd_xattr_set(struct osd_thread_info *info,
743 struct inode *inode, const char *name,
744 const void *buf, int buflen, int fl)
746 struct dentry *dentry = &info->oti_child_dentry;
748 dquot_initialize(inode);
749 dentry->d_inode = inode;
750 dentry->d_sb = inode->i_sb;
751 return ll_vfs_setxattr(dentry, inode, name, buf, buflen, fl);
754 static inline char *osd_oid_name(char *name, size_t name_size,
755 const struct lu_fid *fid, u64 id)
757 snprintf(name, name_size,
758 (fid_seq_is_rsvd(fid_seq(fid)) ||
759 fid_seq_is_mdt0(fid_seq(fid)) ||
760 fid_seq_is_idif(fid_seq(fid))) ? "%llu" : "%llx", id);
765 #ifdef CONFIG_PROC_FS
767 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
768 int osd_procfs_init(struct osd_device *osd, const char *name);
769 int osd_procfs_fini(struct osd_device *osd);
770 void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
771 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 0, 52, 0)
772 int osd_register_proc_index_in_idif(struct osd_device *osd);
776 int osd_statfs(const struct lu_env *env, struct dt_device *dev,
777 struct obd_statfs *sfs, struct obd_statfs_info *info);
778 struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
779 struct osd_inode_id *id);
781 osd_iget_fid(struct osd_thread_info *info, struct osd_device *dev,
782 struct osd_inode_id *id, struct lu_fid *fid);
783 int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
784 const struct lu_fid *fid, __u32 compat, __u32 incompat);
785 struct osd_obj_seq *osd_seq_load(struct osd_thread_info *info,
786 struct osd_device *osd, u64 seq);
787 int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
788 struct dentry *dentry, struct lustre_ost_attrs *loa);
789 void osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
790 struct osd_inode_id *id, const struct lu_fid *fid);
791 int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
792 struct dentry *dentry, struct lu_fid *fid);
793 struct osd_it_ea *osd_it_dir_init(const struct lu_env *env, struct inode *inode,
795 void osd_it_dir_fini(const struct lu_env *env, struct osd_it_ea *oie,
796 struct inode *inode);
797 int osd_ldiskfs_it_fill(const struct lu_env *env, const struct dt_it *di);
799 int osd_obj_map_init(const struct lu_env *env, struct osd_device *osd);
800 void osd_obj_map_fini(struct osd_device *dev);
801 int osd_obj_map_lookup(struct osd_thread_info *info, struct osd_device *osd,
802 const struct lu_fid *fid, struct osd_inode_id *id);
803 int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
804 const struct lu_fid *fid, const struct osd_inode_id *id,
806 int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
807 const struct lu_fid *fid, handle_t *th);
808 int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
809 const struct lu_fid *fid, const struct osd_inode_id *id,
811 int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
812 struct inode *src_parent, struct dentry *src_child,
813 const struct lu_fid *fid);
814 int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
815 const struct lu_fid *fid, struct osd_inode_id *id,
816 enum oi_check_flags flags);
817 int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
818 const struct lu_fid *fid, const struct osd_inode_id *id,
820 int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
821 const struct lu_fid *fid, const struct osd_inode_id *id,
823 int osd_obj_del_entry(struct osd_thread_info *info, struct osd_device *osd,
824 struct dentry *dird, char *name, int namelen,
827 char *osd_lf_fid2name(const struct lu_fid *fid);
828 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
830 void osd_scrub_stop(struct osd_device *dev);
831 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev,
833 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
834 int osd_oii_insert(struct osd_device *dev, const struct lu_fid *fid,
835 struct osd_inode_id *id, int insert);
836 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
837 struct osd_inode_id *id);
838 void osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
840 struct dentry *osd_lookup_one_len_unlocked(struct osd_device *dev,
842 struct dentry *base, int len);
843 struct dentry *osd_lookup_one_len(struct osd_device *dev, const char *name,
844 struct dentry *base, int len);
846 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
847 u64 seq, struct lu_seq_range *range);
849 int osd_delete_from_remote_parent(const struct lu_env *env,
850 struct osd_device *osd,
851 struct osd_object *obj,
852 struct osd_thandle *oh, bool destroy);
853 int osd_add_to_remote_parent(const struct lu_env *env, struct osd_device *osd,
854 struct osd_object *obj, struct osd_thandle *oh);
855 int osd_lookup_in_remote_parent(struct osd_thread_info *oti,
856 struct osd_device *osd,
857 const struct lu_fid *fid,
858 struct osd_inode_id *id);
860 int osd_ost_seq_exists(struct osd_thread_info *info, struct osd_device *osd,
862 int osd_scrub_refresh_mapping(struct osd_thread_info *info,
863 struct osd_device *dev,
864 const struct lu_fid *fid,
865 const struct osd_inode_id *id,
867 enum oi_check_flags flags, bool *exist);
869 /* osd_quota_fmt.c */
870 int walk_tree_dqentry(const struct lu_env *env, struct osd_object *obj,
871 int type, uint blk, int depth, uint index,
872 struct osd_it_quota *it);
873 int walk_block_dqentry(const struct lu_env *env, struct osd_object *obj,
874 int type, uint blk, uint index,
875 struct osd_it_quota *it);
876 loff_t find_tree_dqentry(const struct lu_env *env,
877 struct osd_object *obj, int type,
878 qid_t dqid, uint blk, int depth,
879 struct osd_it_quota *it);
881 int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
882 struct lquota_id_info *qi, struct osd_object *obj,
883 bool enforce, enum osd_quota_local_flags *local_flags);
884 int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
885 __u32 projid, long long space, struct osd_thandle *oh,
886 struct osd_object *obj,
887 enum osd_quota_local_flags *local_flags,
888 enum osd_qid_declare_flags);
889 const struct dt_rec *osd_quota_pack(struct osd_object *obj,
890 const struct dt_rec *rec,
891 union lquota_rec *quota_rec);
892 void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
894 #ifdef HAVE_PROJECT_QUOTA
895 static inline __u32 i_projid_read(struct inode *inode)
897 return (__u32)from_kprojid(&init_user_ns, LDISKFS_I(inode)->i_projid);
900 static inline void i_projid_write(struct inode *inode, __u32 projid)
903 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
904 LDISKFS_I(inode)->i_projid = kprojid;
907 static inline uid_t i_projid_read(struct inode *inode)
911 static inline void i_projid_write(struct inode *inode, __u32 projid)
917 #ifdef HAVE_LDISKFS_IGET_WITH_FLAGS
918 # define osd_ldiskfs_iget(sb, ino) \
919 ldiskfs_iget((sb), (ino), \
920 LDISKFS_IGET_HANDLE | LDISKFS_IGET_SPECIAL)
922 # define osd_ldiskfs_iget(sb, ino) ldiskfs_iget((sb), (ino))
925 #ifdef HAVE_LDISKFS_INFO_JINODE
926 # define osd_attach_jinode(inode) ldiskfs_inode_attach_jinode(inode)
927 #else /* HAVE_LDISKFS_INFO_JINODE */
928 # define osd_attach_jinode(inode) 0
929 #endif /* HAVE_LDISKFS_INFO_JINODE */
931 #ifdef LDISKFS_HT_MISC
932 # define osd_journal_start_sb(sb, type, nblock) \
933 ldiskfs_journal_start_sb(sb, type, nblock)
934 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
936 ldiskfs_lblk_t *nblock)
940 rc = osd_attach_jinode(inode);
943 return ldiskfs_append(handle, inode, nblock);
946 # ifdef HAVE___LDISKFS_FIND_ENTRY
947 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
948 (__ldiskfs_find_entry(dir, name, de, inlined, lock) ?: \
951 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
952 (ldiskfs_find_entry_locked(dir, name, de, inlined, lock) ?: \
956 # define osd_journal_start(inode, type, nblocks) \
957 ldiskfs_journal_start(inode, type, nblocks)
958 # define osd_transaction_size(dev) \
959 (osd_journal(dev)->j_max_transaction_buffers / 2)
960 #else /* ! defined LDISKFS_HT_MISC */
961 # define LDISKFS_HT_MISC 0
962 # define osd_journal_start_sb(sb, type, nblock) \
963 ldiskfs_journal_start_sb(sb, nblock)
965 static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
967 ldiskfs_lblk_t *nblock)
969 struct buffer_head *bh;
972 bh = ldiskfs_append(handle, inode, nblock, &err);
979 # ifdef HAVE___LDISKFS_FIND_ENTRY
980 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
981 (__ldiskfs_find_entry(dir, name, de, lock) ?: \
984 # define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
985 (ldiskfs_find_entry_locked(dir, name, de, lock) ?: \
988 # define osd_journal_start(inode, type, nblocks) \
989 ldiskfs_journal_start(inode, nblocks)
990 # define osd_transaction_size(dev) \
991 (osd_journal(dev)->j_max_transaction_buffers)
992 #endif /* LDISKFS_HT_MISC */
994 #ifndef HAVE___LDISKFS_FIND_ENTRY
995 # define __ldiskfs_add_entry(handle, child, inode, hlock) \
996 ldiskfs_add_entry_locked(handle, child, inode, hlock)
1000 * Invariants, assertions.
1004 * XXX: do not enable this, until invariant checking code is made thread safe
1005 * in the face of pdirops locking.
1007 #define OSD_INVARIANT_CHECKS (0)
1009 #if OSD_INVARIANT_CHECKS
1010 static inline int osd_invariant(const struct osd_object *obj)
1014 ergo(obj->oo_inode != NULL,
1015 obj->oo_inode->i_sb == osd_sb(osd_obj2dev(obj)) &&
1016 atomic_read(&obj->oo_inode->i_count) > 0) &&
1017 ergo(obj->oo_dir != NULL &&
1018 obj->oo_dir->od_conationer.ic_object != NULL,
1019 obj->oo_dir->od_conationer.ic_object == obj->oo_inode);
1022 #define osd_invariant(obj) (1)
1025 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
1026 #define OSD_READCACHE_MAX_IO_MB 8
1027 #define OSD_WRITECACHE_MAX_IO_MB 8
1029 extern const struct dt_index_operations osd_otable_ops;
1031 static inline int osd_oi_fid2idx(struct osd_device *dev,
1032 const struct lu_fid *fid)
1034 return fid->f_seq & (dev->od_oi_count - 1);
1037 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
1038 const struct lu_fid *fid)
1040 LASSERTF(!fid_is_idif(fid), DFID"\n", PFID(fid));
1041 LASSERTF(!fid_is_last_id(fid), DFID"\n", PFID(fid));
1042 LASSERTF(osd->od_oi_table != NULL && osd->od_oi_count >= 1,
1043 DFID"\n", PFID(fid));
1044 /* It can work even od_oi_count equals to 1 although it's unexpected,
1045 * the only reason we set it to 1 is for performance measurement */
1046 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
1049 extern const struct lu_device_operations osd_lu_ops;
1051 static inline int lu_device_is_osd(const struct lu_device *d)
1053 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
1056 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
1058 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
1059 return container_of(d, struct osd_device, od_dt_dev);
1062 static inline struct osd_device *osd_dev(const struct lu_device *d)
1064 LASSERT(lu_device_is_osd(d));
1065 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
1068 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
1070 return osd_dev(o->oo_dt.do_lu.lo_dev);
1073 static inline struct super_block *osd_sb(const struct osd_device *dev)
1075 return dev->od_mnt->mnt_sb;
1078 static inline const char *osd_sb2name(const struct super_block *sb)
1080 /* this is LDISKFS_SB(sb), but preserves "const" */
1081 const struct ldiskfs_sb_info *sbi = sb->s_fs_info;
1083 return sbi->s_es->s_volume_name;
1086 static inline const char *osd_dev2name(const struct osd_device *dev)
1088 return osd_sb2name(osd_sb(dev));
1091 static inline const char *osd_ino2name(const struct inode *inode)
1093 return osd_sb2name(inode->i_sb);
1097 * Put the osd object once done with it.
1099 * \param obj osd object that needs to be put
1101 static inline void osd_object_put(const struct lu_env *env,
1102 struct osd_object *obj)
1104 dt_object_put(env, &obj->oo_dt);
1107 static inline int osd_object_is_root(const struct osd_object *obj)
1109 return obj->oo_inode && is_root_inode(obj->oo_inode);
1112 static inline struct osd_object *osd_obj(const struct lu_object *o)
1114 LASSERT(lu_device_is_osd(o->lo_dev));
1115 return container_of(o, struct osd_object, oo_dt.do_lu);
1118 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
1120 return osd_obj(&d->do_lu);
1123 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
1125 return &osd->od_dt_dev.dd_lu_dev;
1128 static inline journal_t *osd_journal(const struct osd_device *dev)
1130 return LDISKFS_SB(osd_sb(dev))->s_journal;
1133 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
1135 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
1138 static inline char *osd_name(struct osd_device *osd)
1140 return osd->od_svname;
1143 static inline bool osd_is_ea_inode(struct inode *inode)
1145 return !!(LDISKFS_I(inode)->i_flags & LDISKFS_EA_INODE_FL);
1148 extern const struct dt_body_operations osd_body_ops;
1149 extern struct lu_context_key osd_key;
1151 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
1153 return lu_context_key_get(&env->le_ctx, &osd_key);
1156 extern const struct dt_body_operations osd_body_ops_new;
1162 struct iam_path_descr *osd_it_ipd_get(const struct lu_env *env,
1163 const struct iam_container *bag)
1165 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1166 osd_oti_get(env)->oti_it_ipd);
1170 struct iam_path_descr *osd_idx_ipd_get(const struct lu_env *env,
1171 const struct iam_container *bag)
1173 return bag->ic_descr->id_ops->id_ipd_alloc(bag,
1174 osd_oti_get(env)->oti_idx_ipd);
1177 static inline void osd_ipd_put(const struct lu_env *env,
1178 const struct iam_container *bag,
1179 struct iam_path_descr *ipd)
1181 bag->ic_descr->id_ops->id_ipd_free(ipd);
1184 int osd_calc_bkmap_credits(struct super_block *sb, struct inode *inode,
1185 const loff_t size, const loff_t pos,
1188 int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
1191 struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
1192 struct inode *inode,
1193 const char *name, const int namelen)
1195 struct osd_thread_info *info = osd_oti_get(env);
1196 struct dentry *child_dentry = &info->oti_child_dentry;
1197 struct dentry *obj_dentry = &info->oti_obj_dentry;
1199 obj_dentry->d_inode = inode;
1200 obj_dentry->d_sb = inode->i_sb;
1201 obj_dentry->d_name.hash = 0;
1203 child_dentry->d_name.hash = 0;
1204 child_dentry->d_parent = obj_dentry;
1205 child_dentry->d_name.name = name;
1206 child_dentry->d_name.len = namelen;
1207 return child_dentry;
1210 /* build quasi file structure when it is needed to call an inode i_fop */
1211 static inline struct file *osd_quasi_file_init(const struct lu_env *env,
1212 struct dentry *dentry,
1213 struct inode *inode)
1215 struct osd_thread_info *info = osd_oti_get(env);
1217 info->oti_file.f_path.dentry = dentry;
1218 info->oti_file.f_mapping = inode->i_mapping;
1219 info->oti_file.f_op = inode->i_fop;
1220 info->oti_file.f_inode = inode;
1221 info->oti_file.f_pos = 0;
1222 info->oti_file.private_data = NULL;
1223 info->oti_file.f_cred = current_cred();
1224 info->oti_file.f_flags = O_NOATIME;
1225 info->oti_file.f_mode = FMODE_64BITHASH | FMODE_NONOTIFY;
1227 return &info->oti_file;
1230 static inline struct file *osd_quasi_file(const struct lu_env *env,
1231 struct inode *inode)
1233 struct osd_thread_info *info = osd_oti_get(env);
1235 info->oti_obj_dentry.d_inode = inode;
1236 info->oti_obj_dentry.d_sb = inode->i_sb;
1238 return osd_quasi_file_init(env, &info->oti_obj_dentry, inode);
1241 static inline struct file *osd_quasi_file_by_dentry(const struct lu_env *env,
1242 struct dentry *dentry)
1244 return osd_quasi_file_init(env, dentry, dentry->d_inode);
1247 extern int osd_trans_declare_op2rb[];
1248 extern int ldiskfs_track_declares_assert;
1249 void osd_trans_dump_creds(const struct lu_env *env, struct thandle *th);
1251 static inline void osd_trans_declare_op(const struct lu_env *env,
1252 struct osd_thandle *oh,
1253 enum osd_op_type op, int credits)
1255 struct osd_thread_info *oti = osd_oti_get(env);
1257 LASSERT(oh->ot_handle == NULL);
1258 if (unlikely(op >= OSD_OT_MAX)) {
1259 if (unlikely(ldiskfs_track_declares_assert)) {
1260 LASSERT(op < OSD_OT_MAX);
1262 CWARN("%s: Invalid operation index %d\n",
1263 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op);
1264 libcfs_debug_dumpstack(NULL);
1267 oti->oti_declare_ops[op]++;
1268 oti->oti_declare_ops_cred[op] += credits;
1270 oh->ot_credits += credits;
1273 /* linux: v5.4-rc3-21-g933f1c1e0b75 renamed h_buffer_credits */
1274 #ifdef HAVE_JOURNAL_TOTAL_CREDITS
1275 #define h_buffer_credits h_total_credits
1278 static inline void osd_trans_exec_op(const struct lu_env *env,
1280 enum osd_op_type op)
1282 struct osd_thread_info *oti = osd_oti_get(env);
1283 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1285 unsigned int rb, left;
1287 LASSERT(oh->ot_handle != NULL);
1288 if (unlikely(op >= OSD_OT_MAX)) {
1289 if (unlikely(ldiskfs_track_declares_assert))
1290 LASSERT(op < OSD_OT_MAX);
1292 CWARN("%s: opcode %u: invalid value >= %u\n",
1293 osd_name(osd_dt_dev(oh->ot_super.th_dev)),
1295 libcfs_debug_dumpstack(NULL);
1300 /* find rollback (or reverse) operation for the given one
1301 * such an operation doesn't require additional credits
1302 * as the same set of blocks are modified */
1303 rb = osd_trans_declare_op2rb[op];
1305 /* check whether credits for this operation were reserved at all */
1306 if (unlikely(oti->oti_declare_ops_cred[op] == 0 &&
1307 oti->oti_declare_ops_cred[rb] == 0)) {
1308 /* the API is not perfect yet: CREATE does REF_ADD internally
1309 * while DESTROY does not. To rollback CREATE the callers
1310 * needs to call REF_DEL+DESTROY which is hard to detect using
1311 * a simple table of rollback operations */
1312 if (op == OSD_OT_REF_DEL &&
1313 oti->oti_declare_ops_cred[OSD_OT_CREATE] > 0)
1315 if (op == OSD_OT_REF_ADD &&
1316 oti->oti_declare_ops_cred[OSD_OT_DESTROY] > 0)
1318 CWARN("%s: opcode %u: credits = 0, rollback = %u\n",
1319 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
1320 osd_trans_dump_creds(env, th);
1321 LASSERT(!ldiskfs_track_declares_assert);
1325 /* remember how many credits we have unused before the operation */
1326 oti->oti_credits_before = oh->ot_handle->h_buffer_credits;
1327 left = oti->oti_declare_ops_cred[op] - oti->oti_declare_ops_used[op];
1328 if (unlikely(oti->oti_credits_before < left)) {
1329 CWARN("%s: opcode %u: before %u < left %u, rollback = %u\n",
1330 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1331 oti->oti_credits_before, left, rb);
1332 osd_trans_dump_creds(env, th);
1333 /* on a very small fs (testing?) it's possible that
1334 * the transaction can't fit 1/4 of journal, so we
1335 * just request less credits (see osd_trans_start()).
1336 * ignore the same case here */
1337 rb = osd_transaction_size(osd_dt_dev(th->th_dev));
1338 if (unlikely(oh->ot_credits < rb))
1339 LASSERT(!ldiskfs_track_declares_assert);
1343 static inline void osd_trans_exec_check(const struct lu_env *env,
1345 enum osd_op_type op)
1347 struct osd_thread_info *oti = osd_oti_get(env);
1348 struct osd_thandle *oh = container_of(th, struct osd_thandle,
1350 int used, over, quota;
1352 /* how many credits have been used by the operation */
1353 used = oti->oti_credits_before - oh->ot_handle->h_buffer_credits;
1355 if (unlikely(used < 0)) {
1356 /* if some block was allocated and released in the same
1357 * transaction, then it won't be a part of the transaction
1358 * and delta can be negative */
1363 /* rollback operations (e.g. when we destroy just created
1364 * object) should not consume any credits. there is no point
1365 * to confuse the checks below */
1369 oti->oti_declare_ops_used[op] += used;
1370 if (oti->oti_declare_ops_used[op] <= oti->oti_declare_ops_cred[op])
1373 /* we account quota for a whole transaction and any operation can
1374 * consume corresponding credits */
1375 over = oti->oti_declare_ops_used[op] -
1376 oti->oti_declare_ops_cred[op];
1377 quota = oti->oti_declare_ops_cred[OSD_OT_QUOTA] -
1378 oti->oti_declare_ops_used[OSD_OT_QUOTA];
1379 if (over <= quota) {
1380 /* probably that credits were consumed by
1381 * quota indirectly (in the depths of ldiskfs) */
1382 oti->oti_declare_ops_used[OSD_OT_QUOTA] += over;
1383 oti->oti_declare_ops_used[op] -= over;
1385 CWARN("%s: opcode %d: used %u, used now %u, reserved %u\n",
1386 osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
1387 oti->oti_declare_ops_used[op], used,
1388 oti->oti_declare_ops_cred[op]);
1389 osd_trans_dump_creds(env, th);
1390 if (unlikely(ldiskfs_track_declares_assert))
1396 * Helper function to pack the fid, ldiskfs stores fid in packed format.
1399 void osd_fid_pack(struct osd_fid_pack *pack, const struct dt_rec *fid,
1400 struct lu_fid *befider)
1402 fid_cpu_to_be(befider, (struct lu_fid *)fid);
1403 memcpy(pack->fp_area, befider, sizeof(*befider));
1404 pack->fp_len = sizeof(*befider) + 1;
1408 int osd_fid_unpack(struct lu_fid *fid, const struct osd_fid_pack *pack)
1413 switch (pack->fp_len) {
1414 case sizeof *fid + 1:
1415 memcpy(fid, pack->fp_area, sizeof *fid);
1416 fid_be_to_cpu(fid, fid);
1419 CERROR("Unexpected packed fid size: %d\n", pack->fp_len);
1426 * Quota/Accounting handling
1428 extern const struct dt_index_operations osd_acct_index_ops;
1429 int osd_acct_obj_lookup(struct osd_thread_info *info, struct osd_device *osd,
1430 const struct lu_fid *fid, struct osd_inode_id *id);
1432 /* copy from fs/ext4/dir.c */
1433 static inline int is_32bit_api(void)
1435 #ifdef CONFIG_COMPAT
1436 return in_compat_syscall();
1438 return (BITS_PER_LONG == 32);
1442 static inline loff_t ldiskfs_get_htree_eof(struct file *filp)
1444 if ((filp->f_mode & FMODE_32BITHASH) ||
1445 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
1446 return LDISKFS_HTREE_EOF_32BIT;
1448 return LDISKFS_HTREE_EOF_64BIT;
1451 static inline int fid_is_internal(const struct lu_fid *fid)
1453 return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
1456 static inline bool is_remote_parent_ino(struct osd_device *o, unsigned long ino)
1461 LASSERT(o->od_mdt_map != NULL);
1463 return ino == o->od_mdt_map->omm_remote_parent->d_inode->i_ino;
1467 * ext4_bread/ldiskfs_bread has either 5 or 4 parameters. The error
1468 * return code has been removed and integrated into the pointer in the
1471 static inline struct buffer_head *__ldiskfs_bread(handle_t *handle,
1472 struct inode *inode,
1473 ldiskfs_lblk_t block,
1477 struct buffer_head *bh;
1480 rc = osd_attach_jinode(inode);
1484 #ifdef HAVE_EXT4_BREAD_4ARGS
1485 bh = ldiskfs_bread(handle, inode, block, create);
1488 bh = ldiskfs_bread(handle, inode, block, create, &rc);
1489 if (bh == NULL && rc != 0)
1495 #ifndef HAVE_BIO_INTEGRITY_ENABLED
1496 bool bio_integrity_enabled(struct bio *bio);
1500 # define bio_get_dev(bio) ((bio)->bi_bdev)
1501 # define bio_get_disk(bio) (bio_get_dev(bio)->bd_disk)
1502 # define bio_get_queue(bio) bdev_get_queue(bio_get_dev(bio))
1503 # define bio_set_dev(bio, bdev) (bio_get_dev(bio) = (bdev))
1505 # define bio_get_disk(bio) ((bio)->bi_disk)
1506 # define bio_get_queue(bio) (bio_get_disk(bio)->queue)
1509 void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
1510 void ldiskfs_dec_count(handle_t *handle, struct inode *inode);
1512 void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
1515 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1516 __u32 keysize, __u32 recsize)
1518 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1519 &osd->od_index_backup_list, &osd->od_lock,
1520 &osd->od_index_backup_stop,
1521 fid, keysize, recsize);
1525 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1527 struct osd_thread_info *info = osd_oti_get(env);
1528 struct lu_fid *fid = &info->oti_fid3;
1529 struct osd_inode_id *id = &info->oti_id3;
1531 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1532 osd_id_gen(id, osd->od_index_backup_inode->i_ino,
1533 osd->od_index_backup_inode->i_generation);
1534 osd_add_oi_cache(info, osd, id, fid);
1536 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1537 &osd->od_index_backup_list, &osd->od_lock,
1538 &osd->od_index_backup_stop, backup);
1541 #ifdef LDISKFS_HAS_INCOMPAT_FEATURE
1543 # ifdef LDISKFS_FEATURE_INCOMPAT_EXTENTS
1544 # define ldiskfs_has_feature_extents(sb) \
1545 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EXTENTS)
1547 # ifdef LDISKFS_FEATURE_INCOMPAT_EA_INODE
1548 # define ldiskfs_has_feature_ea_inode(sb) \
1549 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_EA_INODE)
1551 # ifdef LDISKFS_FEATURE_INCOMPAT_DIRDATA
1552 # define ldiskfs_has_feature_dirdata(sb) \
1553 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_DIRDATA)
1555 # ifdef LDISKFS_FEATURE_INCOMPAT_LARGEDIR
1556 # define ldiskfs_has_feature_largedir(sb) \
1557 LDISKFS_HAS_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1558 # define ldiskfs_set_feature_largedir(sb) \
1559 LDISKFS_SET_INCOMPAT_FEATURE(sb, LDISKFS_FEATURE_INCOMPAT_LARGEDIR)
1561 # ifdef LDISKFS_FEATURE_COMPAT_HAS_JOURNAL
1562 # define ldiskfs_has_feature_journal(sb) \
1563 LDISKFS_HAS_COMPAT_FEATURE(sb, LDISKFS_FEATURE_COMPAT_HAS_JOURNAL)
1565 # ifdef LDISKFS_FEATURE_RO_COMPAT_QUOTA
1566 # define ldiskfs_has_feature_quota(sb) \
1567 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_QUOTA)
1569 # ifdef LDISKFS_FEATURE_RO_COMPAT_PROJECT
1570 # define ldiskfs_has_feature_project(sb) \
1571 LDISKFS_HAS_RO_COMPAT_FEATURE(sb, LDISKFS_FEATURE_RO_COMPAT_PROJECT)
1576 int osd_trunc_lock(struct osd_object *obj, struct osd_thandle *oh,
1578 void osd_trunc_unlock_all(const struct lu_env *env, struct list_head *list);
1579 void osd_process_truncates(const struct lu_env *env, struct list_head *list);
1580 void osd_execute_truncate(struct osd_object *obj);
1582 #ifdef HAVE_BIO_ENDIO_USES_ONE_ARG
1583 #define osd_dio_complete_routine(bio, error) dio_complete_routine(bio)
1585 #define osd_dio_complete_routine(bio, error) dio_complete_routine(bio, error)
1588 #ifndef HAVE___BI_CNT
1589 #define __bi_cnt bi_cnt
1593 #define bi_opf bi_rw
1596 #ifndef HAVE_CLEAN_BDEV_ALIASES
1597 #define clean_bdev_aliases(bdev, block, len) \
1598 unmap_underlying_metadata((bdev), (block))
1601 #ifndef HAVE_BI_STATUS
1602 #define bi_status bi_error
1606 * Maximum size of xattr attributes for FEATURE_INCOMPAT_EA_INODE 1Mb
1607 * This limit is arbitrary, but is reasonable for the xattr API.
1609 #define LDISKFS_XATTR_MAX_LARGE_EA_SIZE (1024 * 1024)
1611 struct osd_bio_private {
1612 struct osd_iobuf *obp_iobuf;
1613 /* Start page index in the obp_iobuf for the bio */
1614 int obp_start_page_idx;
1617 #ifdef HAVE_BIO_INTEGRITY_PREP_FN
1618 int osd_get_integrity_profile(struct osd_device *osd,
1619 integrity_gen_fn **generate_fn,
1620 integrity_vrfy_fn **verify_fn);
1622 #define integrity_gen_fn void
1623 #define integrity_vrfy_fn int
1624 static inline int osd_get_integrity_profile(struct osd_device *osd,
1625 integrity_gen_fn **generate_fn,
1626 integrity_vrfy_fn **verify_fn)
1631 static inline bool bio_integrity_prep_fn(struct bio *bio,
1632 integrity_gen_fn *generate_fn,
1633 integrity_vrfy_fn *verify_fn)
1635 return bio_integrity_prep(bio);
1639 #ifdef HAVE_BIO_BI_PHYS_SEGMENTS
1640 #define osd_bio_nr_segs(bio) ((bio)->bi_phys_segments)
1642 #define osd_bio_nr_segs(bio) bio_segments((bio))
1643 #endif /* HAVE_BIO_BI_PHYS_SEGMENTS */
1645 #endif /* _OSD_INTERNAL_H */