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) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/osd-zfs/osd_internal.h
33 * Shared definitions and declarations for zfs/dmu osd
35 * Author: Alex Zhuravlev <bzzz@whamcloud.com>
36 * Author: Mike Pershin <tappro@whamcloud.com>
37 * Author: Johann Lombardi <johann@whamcloud.com>
40 #ifndef _OSD_INTERNAL_H
41 #define _OSD_INTERNAL_H
43 #include <dt_object.h>
44 #include <md_object.h>
45 #include <lustre_quota.h>
46 #include <lustre_scrub.h>
52 #include <sys/nvpair.h>
53 #include <sys/zfs_znode.h>
56 #include <sys/dmu_objset.h>
57 #include <lustre_scrub.h>
60 * By design including kmem.h overrides the Linux slab interfaces to provide
61 * the Illumos kmem cache interfaces. To override this and gain access to
62 * the Linux interfaces these preprocessor macros must be undefined.
64 #ifdef kmem_cache_destroy
65 #undef kmem_cache_destroy
68 #ifdef kmem_cache_create
69 #undef kmem_cache_create
72 #ifdef kmem_cache_alloc
73 #undef kmem_cache_alloc
76 #ifdef kmem_cache_free
77 #undef kmem_cache_free
80 #define ZFS_VERSION_CODE \
81 OBD_OCD_VERSION(ZFS_MAJOR, ZFS_MINOR, ZFS_PATCH, ZFS_FIX)
83 #define LUSTRE_ROOT_FID_SEQ 0
84 #define DMU_OSD_SVNAME "svname"
85 #define DMU_OSD_OI_NAME_BASE "oi"
87 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
89 /* Statfs space reservation for grant, fragmentation, and unlink space. */
90 #define OSD_STATFS_RESERVED_SIZE (16ULL << 20) /* reserve 16MB minimum */
91 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
93 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
94 #define OSD_DNODE_MIN_BLKSHIFT (DNODES_PER_BLOCK_SHIFT)
95 #define OSD_DNODE_EST_BLKSHIFT (12) /* est 4KB/dnode */
96 #define OSD_DNODE_EST_COUNT 4096
98 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
100 extern struct dt_body_operations osd_body_scrub_ops;
103 * Iterator's in-memory data structure for quota file.
105 struct osd_it_quota {
106 struct osd_object *oiq_obj;
107 /* DMU accounting object id */
110 zap_cursor_t *oiq_zc;
111 /** identifier for current quota record */
113 unsigned oiq_reset:1; /* 1 -- no need to advance */
117 * Iterator's in-memory data structure for ZAPs
119 * ZFS does not store . and .. on a disk, instead they are
120 * generated up on request
121 * to follow this format we do the same
124 zap_cursor_t *ozi_zc;
125 struct osd_object *ozi_obj;
126 unsigned ozi_reset:1; /* 1 -- no need to advance */
127 /* ozi_pos - position of the cursor:
128 * 0 - before any record
131 * 3 - real records */
134 char ozi_name[MAXNAMELEN]; /* file name for dir */
135 __u64 ozi_key; /* binary key for index files */
138 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
141 * regular ZFS direntry
143 struct zpl_direntry {
144 uint64_t zde_dnode:48,
147 } __attribute__((packed));
150 * lustre direntry adds a fid to regular ZFS direntry
152 struct luz_direntry {
153 struct zpl_direntry lzd_reg;
154 struct lu_fid lzd_fid;
155 } __attribute__((packed));
158 /* cached SA attributes */
163 #ifdef ZFS_PROJINHERIT
176 #define OSD_INS_CACHE_SIZE 8
179 struct osd_idmap_cache {
180 struct osd_device *oic_dev;
181 struct lu_fid oic_fid;
182 /** max 2^48 dnodes per dataset, avoid spilling into another word */
183 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
184 oic_remote:1; /* FID isn't local */
187 struct osd_inconsistent_item {
188 /* link into lustre_scrub::os_inconsistent_items,
189 * protected by lustr_scrub::os_lock. */
190 struct list_head oii_list;
192 /* The right FID <=> oid mapping. */
193 struct osd_idmap_cache oii_cache;
195 unsigned int oii_insert:1; /* insert or update mapping. */
198 struct osd_otable_it {
199 struct osd_device *ooi_dev;
200 struct lu_fid ooi_fid;
202 __u64 ooi_prefetched_dnode;
205 /* The following bits can be updated/checked w/o lock protection.
206 * If more bits will be introduced in the future and need lock to
207 * protect, please add comment. */
208 unsigned int ooi_used_outside:1, /* Some user out of OSD
209 * uses the iteration. */
210 ooi_all_cached:1, /* No more entries can be
211 * filled into cache. */
212 ooi_user_ready:1, /* The user out of OSD is
213 * ready to iterate. */
214 ooi_waiting:1; /* it::next is waiting. */
217 extern const struct dt_index_operations osd_otable_ops;
219 /* max.number of regular attributes the callers may ask for */
220 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
222 struct osd_thread_info {
223 const struct lu_env *oti_env;
225 struct lu_fid oti_fid;
227 * XXX temporary: for ->i_op calls.
229 struct timespec oti_time;
231 struct ost_id oti_ostid;
237 char oti_key[MAXNAMELEN + 1];
238 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
239 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
241 struct lustre_mdt_attrs oti_mdt_attrs;
242 unsigned int oti_in_trans:1;
244 struct lu_attr oti_la;
245 struct osa_attr oti_osa;
246 zap_attribute_t oti_za;
247 zap_attribute_t oti_za2;
248 dmu_object_info_t oti_doi;
249 struct luz_direntry oti_zde;
251 struct lquota_id_info oti_qi;
252 struct lu_seq_range oti_seq_range;
254 /* dedicated OI cache for insert (which needs inum) */
255 struct osd_idmap_cache *oti_ins_cache;
256 int oti_ins_cache_size;
257 int oti_ins_cache_used;
258 /* inc by osd_trans_create and dec by osd_trans_stop */
259 int oti_ins_cache_depth;
260 struct lu_buf oti_xattr_lbuf;
262 zap_cursor_t oti_zc2;
265 extern struct lu_context_key osd_key;
267 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
269 return lu_context_key_get(&env->le_ctx, &osd_key);
273 struct thandle ot_super;
274 struct list_head ot_dcb_list;
275 struct list_head ot_stop_dcb_list;
276 struct list_head ot_unlinked_list;
277 struct list_head ot_sa_list;
279 struct lquota_trans ot_quota_trans;
280 __u32 ot_write_commit:1,
284 #define OSD_OI_NAME_SIZE 24
287 * Object Index (OI) instance.
290 char oi_name[OSD_OI_NAME_SIZE];
297 uint64_t *os_compat_dirs;
298 int os_subdir_count; /* subdir count for each seq */
299 u64 os_seq; /* seq number */
300 struct list_head os_seq_list; /* list to seq_list */
303 struct osd_seq_list {
304 rwlock_t osl_seq_list_lock; /* lock for seq_list */
305 struct list_head osl_seq_list; /* list head for seq */
306 struct semaphore osl_seq_init_sem;
309 #define OSD_OST_MAP_SIZE 32
316 struct dt_device od_dt_dev;
317 /* information about underlying file system */
318 struct objset *od_os;
319 uint64_t od_rootid; /* id of root znode */
320 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
321 /* SA attr mapping->id,
322 * name is the same as in ZFS to use defines SA_ZPL_...*/
323 sa_attr_type_t *z_attr_table;
325 struct proc_dir_entry *od_proc_entry;
326 struct lprocfs_stats *od_stats;
328 uint64_t od_remote_parent_dir;
329 uint64_t od_index_backup_id;
330 uint64_t od_max_blksz;
333 struct osd_oi **od_oi_table;
334 unsigned int od_oi_count;
335 struct osd_seq_list od_seq_list;
337 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
338 od_prop_rdonly:1, /**< ZFS property readonly */
343 unsigned int od_dnsize;
344 int od_index_backup_stop;
346 enum lustre_index_backup_policy od_index_backup_policy;
353 __s64 od_auto_scrub_interval;
354 struct lu_site od_site;
356 dnode_t *od_groupused_dn;
357 dnode_t *od_userused_dn;
358 #ifdef ZFS_PROJINHERIT
359 dnode_t *od_projectused_dn;
362 /* quota slave instance */
363 struct qsd_instance *od_quota_slave;
365 struct brw_stats od_brw_stats;
366 atomic_t od_r_in_flight;
367 atomic_t od_w_in_flight;
369 /* used to debug zerocopy logic: the fields track all
370 * allocated, loaned and referenced buffers in use.
371 * to be removed once the change is tested well. */
372 atomic_t od_zerocopy_alloc;
373 atomic_t od_zerocopy_loan;
374 atomic_t od_zerocopy_pin;
376 arc_prune_t *arc_prune_cb;
378 /* osd seq instance */
379 struct lu_client_seq *od_cl_seq;
381 struct semaphore od_otable_sem;
382 struct osd_otable_it *od_otable_it;
383 struct lustre_scrub od_scrub;
384 struct list_head od_ios_list;
385 struct list_head od_index_backup_list;
386 struct list_head od_index_restore_list;
390 enum osd_destroy_type {
391 OSD_DESTROY_NONE = 0,
392 OSD_DESTROY_SYNC = 1,
393 OSD_DESTROY_ASYNC = 2,
397 struct dt_object oo_dt;
399 * Inode for file system object represented by this osd_object. This
400 * inode is pinned for the whole duration of lu_object life.
402 * Not modified concurrently (either setup early during object
403 * creation, or assigned by osd_create() under write lock).
406 sa_handle_t *oo_sa_hdl;
407 nvlist_t *oo_sa_xattr;
408 struct list_head oo_sa_linkage;
410 /* used to implement osd_object_*_{lock|unlock} */
411 struct rw_semaphore oo_sem;
413 /* to serialize some updates: destroy vs. others,
414 * xattr_set, object block size change etc */
415 struct rw_semaphore oo_guard;
417 /* protected by oo_guard */
418 struct list_head oo_unlinked_linkage;
420 /* cached attributes */
421 rwlock_t oo_attr_lock;
422 struct lu_attr oo_attr;
424 /* external dnode holding large EAs, protected by oo_guard */
426 enum osd_destroy_type oo_destroy;
428 __u32 oo_destroyed:1,
430 #ifdef ZFS_PROJINHERIT
436 /* the i_flags in LMA */
439 int oo_ea_in_bonus; /* EA bytes we expect */
441 /* record size for index file */
442 unsigned char oo_keysize;
443 unsigned char oo_recsize;
444 unsigned char oo_recusize; /* unit size */
446 uint64_t oo_parent; /* used only at object creation */
448 struct lu_object_header *oo_header;
451 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
452 extern const struct dt_index_operations osd_acct_index_ops;
453 extern struct lu_device_operations osd_lu_ops;
454 extern struct dt_index_operations osd_dir_ops;
455 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
456 qid_t uid, qid_t gid, qid_t projid, long long space,
457 struct osd_thandle *oh, int *flags,
458 enum osd_qid_declare_flags osd_qid_declare_flags);
459 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
460 uint64_t nrblocks, uint64_t est_maxblockshift);
461 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
467 static inline int lu_device_is_osd(const struct lu_device *d)
469 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
472 static inline struct osd_object *osd_obj(const struct lu_object *o)
474 LASSERT(lu_device_is_osd(o->lo_dev));
475 return container_of0(o, struct osd_object, oo_dt.do_lu);
478 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
480 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
481 return container_of0(d, struct osd_device, od_dt_dev);
484 static inline struct osd_device *osd_dev(const struct lu_device *d)
486 LASSERT(lu_device_is_osd(d));
487 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
490 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
492 return osd_obj(&d->do_lu);
495 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
497 return osd_dev(o->oo_dt.do_lu.lo_dev);
500 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
502 return &osd->od_dt_dev.dd_lu_dev;
505 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
507 return osd_dev(o->do_lu.lo_dev)->od_os;
510 static inline int osd_invariant(const struct osd_object *obj)
516 * Put the osd object once done with it.
518 * \param obj osd object that needs to be put
520 static inline void osd_object_put(const struct lu_env *env,
521 struct osd_object *obj)
523 dt_object_put(env, &obj->oo_dt);
526 static inline int osd_object_invariant(const struct lu_object *l)
528 return osd_invariant(osd_obj(l));
531 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
533 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
536 static inline char *osd_name(struct osd_device *osd)
538 return osd->od_svname;
541 static inline void zfs_set_bit(int nr, __u8 *addr)
543 set_bit(nr, (unsigned long *)addr);
546 static inline int zfs_test_bit(int nr, __u8 *addr)
548 return test_bit(nr, (const unsigned long *)addr);
551 static inline int osd_oi_fid2idx(struct osd_device *dev,
552 const struct lu_fid *fid)
554 return fid->f_seq & (dev->od_oi_count - 1);
557 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
558 const struct lu_fid *fid)
560 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
561 "%s: "DFID", oi_count %d\n",
562 osd_name(osd), PFID(fid), osd->od_oi_count);
564 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
567 #ifdef CONFIG_PROC_FS
569 LPROC_OSD_READ_BYTES = 0,
570 LPROC_OSD_WRITE_BYTES = 1,
571 LPROC_OSD_GET_PAGE = 2,
572 LPROC_OSD_NO_PAGE = 3,
573 LPROC_OSD_CACHE_ACCESS = 4,
574 LPROC_OSD_CACHE_HIT = 5,
575 LPROC_OSD_CACHE_MISS = 6,
576 LPROC_OSD_COPY_IO = 7,
577 LPROC_OSD_ZEROCOPY_IO = 8,
578 LPROC_OSD_TAIL_IO = 9,
582 extern struct kmem_cache *osd_zapit_cachep;
584 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
586 int osd_procfs_init(struct osd_device *osd, const char *name);
587 int osd_procfs_fini(struct osd_device *osd);
590 extern char *osd_obj_tag;
591 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
592 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
593 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
594 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
595 uint64_t oid, dmu_buf_t **dbp);
596 struct lu_object *osd_object_alloc(const struct lu_env *env,
597 const struct lu_object_header *hdr,
598 struct lu_device *d);
599 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
600 void *buf, uint32_t buflen, struct osd_thandle *oh);
601 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
602 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
603 unsigned dnsize, zap_flags_t flags);
604 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
605 struct osd_object *obj, const struct lu_fid *fid,
606 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
607 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
608 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
609 struct lu_attr *la, uint64_t parent, nvlist_t *);
610 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
611 uint64_t oid, dnode_t **dnp);
612 int osd_object_init0(const struct lu_env *env, struct osd_object *obj);
615 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
616 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
617 int osd_fid_lookup(const struct lu_env *env,
618 struct osd_device *, const struct lu_fid *, uint64_t *);
619 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
620 const struct lu_fid *fid, char *buf, int bufsize,
622 int osd_options_init(void);
623 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
625 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
626 struct osd_object *obj);
627 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
628 struct osd_device *osd,
629 const struct lu_fid *fid);
630 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
631 struct osd_device *osd,
632 const struct lu_fid *fid);
633 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
634 struct osd_device *osd,
635 const struct lu_fid *fid,
637 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
638 const struct lu_fid *fid);
639 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
640 uint64_t parent, const char *name, uint64_t *child,
641 const struct lu_fid *fid, bool isdir);
643 extern unsigned int osd_oi_count;
646 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
647 uint64_t oid, struct lu_fid *fid);
648 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
649 const struct dt_index_features *feat);
650 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
651 u64 seq, struct lu_seq_range *range);
652 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
653 uint64_t id, uint64_t dirhash);
654 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
655 uint64_t id, uint64_t dirhash);
656 void osd_zap_cursor_fini(zap_cursor_t *zc);
657 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
658 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
659 const struct lu_fid *fid);
660 int osd_add_to_remote_parent(const struct lu_env *env,
661 struct osd_device *osd,
662 struct osd_object *obj,
663 struct osd_thandle *oh);
664 int osd_delete_from_remote_parent(const struct lu_env *env,
665 struct osd_device *osd,
666 struct osd_object *obj,
667 struct osd_thandle *oh, bool destroy);
668 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
672 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
673 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
674 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
676 void osd_scrub_stop(struct osd_device *dev);
677 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
678 const struct lu_fid *fid, uint64_t oid, bool insert);
679 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
683 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
684 struct osd_object *obj,
685 struct osd_thandle *oh);
686 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
687 struct osd_thandle *oh);
688 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
689 struct osd_thandle *oh);
690 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
692 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
693 uint64_t xattr, struct lu_buf *buf,
694 const char *name, int *sizep);
695 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
696 struct lu_buf *buf, const char *name, int *sizep);
697 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
698 struct lu_buf *buf, const char *name);
699 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
700 const struct lu_buf *buf, const char *name,
701 int fl, struct thandle *handle);
702 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
703 const struct lu_buf *buf, const char *name, int fl,
704 struct thandle *handle);
705 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
706 const char *name, struct thandle *handle);
707 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
708 const char *name, struct thandle *handle);
709 void osd_declare_xattrs_destroy(const struct lu_env *env,
710 struct osd_object *obj,
711 struct osd_thandle *oh);
712 int osd_xattrs_destroy(const struct lu_env *env,
713 struct osd_object *obj, struct osd_thandle *oh);
714 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
715 const struct lu_buf *lb);
716 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
717 int vallen, const char *name, struct osd_thandle *oh);
718 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
719 const struct lu_buf *buf, const char *name, int fl,
720 struct osd_thandle *oh);;
721 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
722 const struct lu_buf *buf, const char *name, int fl,
723 struct osd_thandle *oh);
724 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
725 struct osd_thandle *oh);
727 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
728 sizeof(__u64) /* VBR VERSION */ + \
729 sizeof(struct lustre_mdt_attrs) /* LMA */)
731 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
732 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
735 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
737 return DN_MAX_BONUSLEN;
741 static inline int osd_object_is_zap(dnode_t *dn)
743 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
744 dn->dn_type == DMU_OT_USERGROUP_USED);
747 /* XXX: f_ver is not counted, but may differ too */
748 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
750 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
754 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
755 const struct lu_buf *buf, const char *name, int fl,
756 struct osd_thandle *oh)
760 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
764 if (osd_obj2dev(obj)->od_xattr_in_sa) {
765 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
767 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
769 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
775 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
777 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
778 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
779 #ifdef ZFS_PROJINHERIT
780 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
782 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
785 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
787 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
788 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
789 #ifdef ZFS_PROJINHERIT
790 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
792 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
797 #ifndef HAVE_DSL_POOL_CONFIG
798 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
802 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
807 #ifdef HAVE_SPA_MAXBLOCKSIZE
808 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
809 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
811 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
812 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
813 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
816 #ifdef HAVE_SA_SPILL_ALLOC
818 osd_zio_buf_alloc(size_t size)
820 return sa_spill_alloc(KM_SLEEP);
824 osd_zio_buf_free(void *buf, size_t size)
829 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
830 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
833 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
834 static inline uint64_t
835 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
836 int dnodesize, dmu_tx_t *tx)
839 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
841 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
842 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
845 static inline uint64_t
846 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
847 dmu_object_type_t ot, int leaf_blockshift,
848 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
851 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
853 return zap_create_flags_dnsize(os, normflags, flags, ot,
854 leaf_blockshift, indirect_blockshift,
855 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
860 osd_obj_bonuslen(struct osd_object *obj)
862 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
864 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
865 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
867 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
871 dnodesize = dmu_objset_dnodesize(os);
873 bonuslen = DN_BONUS_SIZE(dnodesize);
880 static inline uint64_t
881 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
882 int dnodesize, dmu_tx_t *tx)
884 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
885 DN_MAX_BONUSLEN, tx);
888 static inline uint64_t
889 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
890 dmu_object_type_t ot, int leaf_blockshift,
891 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
893 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
894 indirect_blockshift, DMU_OT_SA,
895 DN_MAX_BONUSLEN, tx);
899 osd_obj_bonuslen(struct osd_object *obj)
901 return DN_MAX_BONUSLEN;
903 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
905 #ifdef HAVE_DMU_PREFETCH_6ARG
906 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
907 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
909 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
910 dmu_prefetch((os), (obj), (lvl), (off))
913 static inline int osd_sa_handle_get(struct osd_object *obj)
915 struct osd_device *osd = osd_obj2dev(obj);
916 dnode_t *dn = obj->oo_dn;
922 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
923 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
924 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
927 refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
931 static inline void osd_dnode_rele(dnode_t *dn)
935 LASSERT(dn->dn_bonus);
939 dmu_buf_rele(&db->db, osd_obj_tag);
942 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
944 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
945 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
947 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
949 return dmu_objset_userobjspace_present(osd->od_os);
953 #define OSD_DMU_USEROBJ_PREFIX "obj-"
954 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
956 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
960 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
962 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
963 dnode_t *dn, const char *key,
964 int int_size, int int_num,
965 const void *val, dmu_tx_t *tx)
969 #ifdef HAVE_ZAP_ADD_BY_DNODE
971 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
973 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
976 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
977 dnode_t *dn, const char *key,
982 #ifdef HAVE_ZAP_ADD_BY_DNODE
984 return -zap_remove_by_dnode(dn, key, tx);
986 return -zap_remove(osd->od_os, zap, key, tx);
990 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
991 dnode_t *dn, const char *key,
992 int int_size, int int_num, void *v)
996 #ifdef HAVE_ZAP_ADD_BY_DNODE
998 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
1000 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
1003 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
1004 dnode_t *dn, int add, const char *name)
1006 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1008 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
1012 dmu_tx_hold_zap(tx, zap, add, name);
1015 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
1016 dnode_t *dn, uint64_t off, int len)
1018 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1020 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1024 dmu_tx_hold_write(tx, oid, off, len);
1027 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1028 uint64_t offset, uint64_t size,
1029 const char *buf, dmu_tx_t *tx)
1032 #ifdef HAVE_DMU_WRITE_BY_DNODE
1033 dmu_write_by_dnode(dn, offset, size, buf, tx);
1035 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1039 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1040 uint64_t offset, uint64_t size,
1041 char *buf, int flags)
1044 #ifdef HAVE_DMU_READ_BY_DNODE
1045 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1047 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1051 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1052 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1053 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1055 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1056 dmu_objset_own((name), (type), (ronly), (tag), (os))
1059 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1060 #define osd_dmu_objset_disown(os, decrypt, tag) \
1061 dmu_objset_disown((os), (decrypt), (tag))
1063 #define osd_dmu_objset_disown(os, decrypt, tag) \
1064 dmu_objset_disown((os), (tag))
1068 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1069 __u32 keysize, __u32 recsize)
1071 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1072 &osd->od_index_backup_list, &osd->od_lock,
1073 &osd->od_index_backup_stop,
1074 fid, keysize, recsize);
1078 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1080 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1083 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1084 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1085 osd->od_index_backup_id);
1089 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1090 &osd->od_index_backup_list, &osd->od_lock,
1091 &osd->od_index_backup_stop, backup);
1094 #endif /* _OSD_INTERNAL_H */