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/
31 * lustre/osd-zfs/osd_internal.h
32 * Shared definitions and declarations for zfs/dmu osd
34 * Author: Alex Zhuravlev <bzzz@whamcloud.com>
35 * Author: Mike Pershin <tappro@whamcloud.com>
36 * Author: Johann Lombardi <johann@whamcloud.com>
39 #ifndef _OSD_INTERNAL_H
40 #define _OSD_INTERNAL_H
42 #include <linux/refcount.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 #ifdef HAVE_ZFS_REFCOUNT_HEADER
54 #include <sys/zfs_refcount.h>
56 #include <sys/zfs_znode.h>
59 #include <sys/dmu_objset.h>
60 #include <lustre_scrub.h>
63 * By design including kmem.h overrides the Linux slab interfaces to provide
64 * the Illumos kmem cache interfaces. To override this and gain access to
65 * the Linux interfaces these preprocessor macros must be undefined.
67 #ifdef kmem_cache_destroy
68 #undef kmem_cache_destroy
71 #ifdef kmem_cache_create
72 #undef kmem_cache_create
75 #ifdef kmem_cache_alloc
76 #undef kmem_cache_alloc
79 #ifdef kmem_cache_free
80 #undef kmem_cache_free
83 #define ZFS_VERSION_CODE \
84 OBD_OCD_VERSION(ZFS_MAJOR, ZFS_MINOR, ZFS_PATCH, ZFS_FIX)
86 #define LUSTRE_ROOT_FID_SEQ 0
87 #define DMU_OSD_SVNAME "svname"
88 #define DMU_OSD_OI_NAME_BASE "oi"
90 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
92 /* Statfs space reservation for grant, fragmentation, and unlink space. */
93 #define OSD_STATFS_RESERVED_SIZE (16ULL << 20) /* reserve 16MB minimum */
94 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
96 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
97 #define OSD_DNODE_MIN_BLKSHIFT (DNODES_PER_BLOCK_SHIFT)
98 #define OSD_DNODE_EST_BLKSHIFT (12) /* est 4KB/dnode */
99 #define OSD_DNODE_EST_COUNT 4096
101 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
103 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
105 #ifndef HAVE_ZFS_REFCOUNT_HEADER
106 #ifndef HAVE_ZFS_REFCOUNT_ADD
107 #define zfs_refcount_add refcount_add
111 extern const struct dt_body_operations osd_body_scrub_ops;
112 extern const struct dt_body_operations osd_body_ops;
113 extern struct kmem_cache *osd_object_kmem;
116 * Iterator's in-memory data structure for quota file.
118 struct osd_it_quota {
119 struct osd_object *oiq_obj;
120 /* DMU accounting object id */
123 zap_cursor_t *oiq_zc;
124 /** identifier for current quota record */
126 unsigned oiq_reset:1; /* 1 -- no need to advance */
131 OZI_POS_DOT = 1, /* cursor at . */
132 OZI_POS_DOTDOT = 2, /* cursor at .. */
133 OZI_POS_REAL = 3, /* cursor at real entries */
137 * regular ZFS direntry
139 struct zpl_direntry {
140 uint64_t zde_dnode:48,
143 } __attribute__((packed));
146 * lustre direntry adds a fid to regular ZFS direntry
148 struct luz_direntry {
149 struct zpl_direntry lzd_reg;
150 struct lu_fid lzd_fid;
151 } __attribute__((packed));
154 * Iterator's in-memory data structure for ZAPs
156 * ZFS does not store . and .. on a disk, instead they are
157 * generated up on request
158 * to follow this format we do the same
161 zap_cursor_t *ozi_zc;
162 struct osd_object *ozi_obj;
163 unsigned ozi_reset:1; /* 1 -- no need to advance */
164 /* ozi_pos - position of the cursor */
165 enum osd_zap_pos ozi_pos;
166 struct luz_direntry ozi_zde;
167 zap_attribute_t ozi_za;
169 char ozi_name[MAXNAMELEN]; /* file name for dir */
170 __u64 ozi_key; /* binary key for index files */
173 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
175 /* cached SA attributes */
180 #ifdef ZFS_PROJINHERIT
194 #define OSD_INS_CACHE_SIZE 8
197 struct osd_idmap_cache {
198 struct osd_device *oic_dev;
199 struct lu_fid oic_fid;
200 /** max 2^48 dnodes per dataset, avoid spilling into another word */
201 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
202 oic_remote:1; /* FID isn't local */
205 struct osd_inconsistent_item {
206 /* link into lustre_scrub::os_inconsistent_items,
207 * protected by lustr_scrub::os_lock. */
208 struct list_head oii_list;
210 /* The right FID <=> oid mapping. */
211 struct osd_idmap_cache oii_cache;
213 unsigned int oii_insert:1; /* insert or update mapping. */
216 struct osd_otable_it {
217 struct osd_device *ooi_dev;
218 struct lu_fid ooi_fid;
220 __u64 ooi_prefetched_dnode;
223 /* The following bits can be updated/checked w/o lock protection.
224 * If more bits will be introduced in the future and need lock to
225 * protect, please add comment. */
226 unsigned int ooi_used_outside:1, /* Some user out of OSD
227 * uses the iteration. */
228 ooi_all_cached:1, /* No more entries can be
229 * filled into cache. */
230 ooi_user_ready:1, /* The user out of OSD is
231 * ready to iterate. */
232 ooi_waiting:1; /* it::next is waiting. */
235 extern const struct dt_index_operations osd_otable_ops;
237 /* max.number of regular attributes the callers may ask for */
238 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
240 struct osd_thread_info {
241 const struct lu_env *oti_env;
243 struct lu_fid oti_fid;
245 struct ost_id oti_ostid;
251 char oti_key[MAXNAMELEN + 1];
252 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
253 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
255 struct lustre_mdt_attrs oti_mdt_attrs;
256 unsigned int oti_in_trans:1;
258 struct lu_attr oti_la;
259 struct osa_attr oti_osa;
260 zap_attribute_t oti_za;
261 zap_attribute_t oti_za2;
262 dmu_object_info_t oti_doi;
263 struct luz_direntry oti_zde;
265 struct lquota_id_info oti_qi;
266 struct lu_seq_range oti_seq_range;
268 /* dedicated OI cache for insert (which needs inum) */
269 struct osd_idmap_cache *oti_ins_cache;
270 int oti_ins_cache_size;
271 int oti_ins_cache_used;
272 /* inc by osd_trans_create and dec by osd_trans_stop */
273 int oti_ins_cache_depth;
274 struct lu_buf oti_xattr_lbuf;
276 zap_cursor_t oti_zc2;
282 extern struct lu_context_key osd_key;
284 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
286 return lu_context_key_get(&env->le_ctx, &osd_key);
290 struct thandle ot_super;
291 struct list_head ot_dcb_list;
292 struct list_head ot_stop_dcb_list;
293 struct list_head ot_unlinked_list;
294 struct list_head ot_sa_list;
296 struct lquota_trans ot_quota_trans;
300 #define OSD_OI_NAME_SIZE 24
303 * Object Index (OI) instance.
306 char oi_name[OSD_OI_NAME_SIZE];
313 uint64_t *os_compat_dirs;
314 int os_subdir_count; /* subdir count for each seq */
315 u64 os_seq; /* seq number */
316 struct list_head os_seq_list; /* list to seq_list */
319 struct osd_seq_list {
320 rwlock_t osl_seq_list_lock; /* lock for seq_list */
321 struct list_head osl_seq_list; /* list head for seq */
322 struct semaphore osl_seq_init_sem;
325 #define OSD_OST_MAP_SIZE 32
332 struct dt_device od_dt_dev;
333 /* information about underlying file system */
334 struct objset *od_os;
335 uint64_t od_rootid; /* id of root znode */
336 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
337 /* SA attr mapping->id,
338 * name is the same as in ZFS to use defines SA_ZPL_...*/
339 sa_attr_type_t *z_attr_table;
341 struct proc_dir_entry *od_proc_entry;
342 struct lprocfs_stats *od_stats;
344 uint64_t od_remote_parent_dir;
345 uint64_t od_index_backup_id;
346 uint64_t od_max_blksz;
349 struct osd_oi **od_oi_table;
350 unsigned int od_oi_count;
351 struct osd_seq_list od_seq_list;
353 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
354 od_prop_rdonly:1, /**< ZFS property readonly */
361 unsigned int od_dnsize;
362 int od_index_backup_stop;
364 enum lustre_index_backup_policy od_index_backup_policy;
371 struct lu_site od_site;
373 dnode_t *od_groupused_dn;
374 dnode_t *od_userused_dn;
375 #ifdef ZFS_PROJINHERIT
376 dnode_t *od_projectused_dn;
379 /* quota slave instance for inode */
380 struct qsd_instance *od_quota_slave_md;
382 /* quota slave instance for block */
383 struct qsd_instance *od_quota_slave_dt;
385 struct brw_stats od_brw_stats;
386 atomic_t od_r_in_flight;
387 atomic_t od_w_in_flight;
389 /* used to debug zerocopy logic: the fields track all
390 * allocated, loaned and referenced buffers in use.
391 * to be removed once the change is tested well. */
392 atomic_t od_zerocopy_alloc;
393 atomic_t od_zerocopy_loan;
394 atomic_t od_zerocopy_pin;
396 arc_prune_t *arc_prune_cb;
398 /* osd seq instance */
399 struct lu_client_seq *od_cl_seq;
401 struct semaphore od_otable_sem;
402 struct osd_otable_it *od_otable_it;
403 struct lustre_scrub od_scrub;
404 struct list_head od_ios_list;
405 struct list_head od_index_backup_list;
406 struct list_head od_index_restore_list;
408 unsigned long long od_readcache_max_filesize;
411 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
414 return osd->od_quota_slave_dt;
416 return osd->od_quota_slave_md;
419 enum osd_destroy_type {
420 OSD_DESTROY_NONE = 0,
421 OSD_DESTROY_SYNC = 1,
422 OSD_DESTROY_ASYNC = 2,
425 #define OSD_MAX_DBUFS 2 /* how many dbufs to cache in object */
428 struct dt_object oo_dt;
430 * Inode for file system object represented by this osd_object. This
431 * inode is pinned for the whole duration of lu_object life.
433 * Not modified concurrently (either setup early during object
434 * creation, or assigned by osd_create() under write lock).
437 sa_handle_t *oo_sa_hdl;
438 nvlist_t *oo_sa_xattr;
439 struct list_head oo_sa_linkage;
441 /* used to implement osd_object_*_{lock|unlock} */
442 struct rw_semaphore oo_sem;
444 /* to serialize some updates: destroy vs. others,
445 * xattr_set, object block size change etc */
446 struct rw_semaphore oo_guard;
448 /* protected by oo_guard */
449 struct list_head oo_unlinked_linkage;
451 /* cached attributes */
452 rwlock_t oo_attr_lock;
453 struct lu_attr oo_attr;
455 /* external dnode holding large EAs, protected by oo_guard */
457 enum osd_destroy_type oo_destroy;
459 __u32 oo_destroyed:1,
461 #ifdef ZFS_PROJINHERIT
467 /* the i_flags in LMA */
470 int oo_ea_in_bonus; /* EA bytes we expect */
472 /* record size for index file */
473 unsigned char oo_keysize;
474 unsigned char oo_recsize;
475 unsigned char oo_recusize; /* unit size */
477 uint64_t oo_parent; /* used only at object creation */
479 struct lu_object_header *oo_header;
480 dmu_buf_t *oo_dbs[OSD_MAX_DBUFS];
483 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *,
484 struct obd_statfs_info *);
485 extern const struct dt_index_operations osd_acct_index_ops;
486 extern const struct lu_device_operations osd_lu_ops;
487 extern const struct dt_index_operations osd_dir_ops;
488 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
489 qid_t uid, qid_t gid, qid_t projid, long long space,
490 struct osd_thandle *oh,
491 enum osd_quota_local_flags *local_flags,
492 enum osd_qid_declare_flags osd_qid_declare_flags);
493 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
494 uint64_t nrblocks, uint64_t est_maxblockshift);
495 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
501 static inline int lu_device_is_osd(const struct lu_device *d)
503 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
506 static inline struct osd_object *osd_obj(const struct lu_object *o)
508 LASSERT(lu_device_is_osd(o->lo_dev));
509 return container_of(o, struct osd_object, oo_dt.do_lu);
512 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
514 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
515 return container_of(d, struct osd_device, od_dt_dev);
518 static inline struct osd_device *osd_dev(const struct lu_device *d)
520 LASSERT(lu_device_is_osd(d));
521 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
524 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
526 return osd_obj(&d->do_lu);
529 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
531 return osd_dev(o->oo_dt.do_lu.lo_dev);
534 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
536 return &osd->od_dt_dev.dd_lu_dev;
539 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
541 return osd_dev(o->do_lu.lo_dev)->od_os;
544 static inline int osd_invariant(const struct osd_object *obj)
550 * Put the osd object once done with it.
552 * \param obj osd object that needs to be put
554 static inline void osd_object_put(const struct lu_env *env,
555 struct osd_object *obj)
557 dt_object_put(env, &obj->oo_dt);
560 static inline int osd_object_invariant(const struct lu_object *l)
562 return osd_invariant(osd_obj(l));
565 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
567 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
570 static inline char *osd_name(struct osd_device *osd)
572 return osd->od_svname;
575 static inline void zfs_set_bit(int nr, __u8 *addr)
577 set_bit(nr, (unsigned long *)addr);
580 static inline int zfs_test_bit(int nr, __u8 *addr)
582 return test_bit(nr, (const unsigned long *)addr);
585 static inline int osd_oi_fid2idx(struct osd_device *dev,
586 const struct lu_fid *fid)
588 return fid->f_seq & (dev->od_oi_count - 1);
591 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
592 const struct lu_fid *fid)
594 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
595 "%s: "DFID", oi_count %d\n",
596 osd_name(osd), PFID(fid), osd->od_oi_count);
598 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
601 #ifdef CONFIG_PROC_FS
603 LPROC_OSD_READ_BYTES = 0,
604 LPROC_OSD_WRITE_BYTES = 1,
605 LPROC_OSD_GET_PAGE = 2,
606 LPROC_OSD_NO_PAGE = 3,
607 LPROC_OSD_CACHE_ACCESS = 4,
608 LPROC_OSD_CACHE_HIT = 5,
609 LPROC_OSD_CACHE_MISS = 6,
610 LPROC_OSD_COPY_IO = 7,
611 LPROC_OSD_ZEROCOPY_IO = 8,
612 LPROC_OSD_TAIL_IO = 9,
616 extern struct kmem_cache *osd_zapit_cachep;
618 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
620 int osd_procfs_init(struct osd_device *osd, const char *name);
621 int osd_procfs_fini(struct osd_device *osd);
624 extern char *osd_obj_tag;
625 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
626 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
627 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
628 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
629 uint64_t oid, dmu_buf_t **dbp);
630 struct lu_object *osd_object_alloc(const struct lu_env *env,
631 const struct lu_object_header *hdr,
632 struct lu_device *d);
633 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
634 void *buf, uint32_t buflen, struct osd_thandle *oh);
635 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
636 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
637 unsigned dnsize, zap_flags_t flags);
638 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
639 struct osd_object *obj, const struct lu_fid *fid,
640 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
641 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
642 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
643 struct lu_attr *la, uint64_t parent, nvlist_t *);
644 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
645 uint64_t oid, dnode_t **dnp);
648 int osd_oi_init(const struct lu_env *env, struct osd_device *o, bool reset);
649 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
650 int osd_fid_lookup(const struct lu_env *env,
651 struct osd_device *, const struct lu_fid *, uint64_t *);
652 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
653 const struct lu_fid *fid, char *buf, int bufsize,
655 int osd_options_init(void);
656 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
658 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
659 struct osd_object *obj);
660 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
661 struct osd_device *osd,
662 const struct lu_fid *fid);
663 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
664 struct osd_device *osd,
665 const struct lu_fid *fid);
666 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
667 struct osd_device *osd,
668 const struct lu_fid *fid,
670 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
671 const struct lu_fid *fid);
672 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
673 uint64_t parent, const char *name, uint64_t *child,
674 const struct lu_fid *fid, bool isdir);
676 extern unsigned int osd_oi_count;
679 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
680 uint64_t oid, struct lu_fid *fid);
681 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
682 const struct dt_index_features *feat);
683 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
684 u64 seq, struct lu_seq_range *range);
685 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
686 uint64_t id, uint64_t dirhash);
687 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
688 uint64_t id, uint64_t dirhash);
689 void osd_zap_cursor_fini(zap_cursor_t *zc);
690 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
691 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
692 const struct lu_fid *fid);
693 int osd_add_to_remote_parent(const struct lu_env *env,
694 struct osd_device *osd,
695 struct osd_object *obj,
696 struct osd_thandle *oh);
697 int osd_delete_from_remote_parent(const struct lu_env *env,
698 struct osd_device *osd,
699 struct osd_object *obj,
700 struct osd_thandle *oh, bool destroy);
701 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
705 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev,
706 time64_t interval, bool resetoi);
707 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
708 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
710 void osd_scrub_stop(struct osd_device *dev);
711 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
712 const struct lu_fid *fid, uint64_t oid, bool insert);
713 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
717 * Basic transaction credit op
726 int osd_scrub_refresh_mapping(const struct lu_env *env,
727 struct osd_device *dev,
728 const struct lu_fid *fid,
729 uint64_t oid, enum dt_txn_op ops,
730 bool force, const char *name);
734 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
735 struct osd_object *obj,
736 struct osd_thandle *oh);
737 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
738 struct osd_thandle *oh);
739 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
740 struct osd_thandle *oh);
741 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
743 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
744 uint64_t xattr, struct lu_buf *buf,
745 const char *name, int *sizep);
746 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
747 struct lu_buf *buf, const char *name, int *sizep);
748 int osd_xattr_get_lma(const struct lu_env *env, struct osd_object *obj,
750 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
751 struct lu_buf *buf, const char *name);
752 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
753 const struct lu_buf *buf, const char *name,
754 int fl, struct thandle *handle);
755 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
756 const struct lu_buf *buf, const char *name, int fl,
757 struct thandle *handle);
758 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
759 const char *name, struct thandle *handle);
760 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
761 const char *name, struct thandle *handle);
762 void osd_declare_xattrs_destroy(const struct lu_env *env,
763 struct osd_object *obj,
764 struct osd_thandle *oh);
765 int osd_xattrs_destroy(const struct lu_env *env,
766 struct osd_object *obj, struct osd_thandle *oh);
767 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
768 const struct lu_buf *lb);
769 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
770 int vallen, const char *name, struct osd_thandle *oh);
771 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
772 const struct lu_buf *buf, const char *name, int fl,
773 struct osd_thandle *oh);;
774 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
775 const struct lu_buf *buf, const char *name, int fl,
776 struct osd_thandle *oh);
777 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
778 struct osd_thandle *oh);
780 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
781 sizeof(__u64) /* VBR VERSION */ + \
782 sizeof(struct lustre_mdt_attrs) /* LMA */)
784 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
786 static inline int osd_object_is_zap(dnode_t *dn)
788 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
789 dn->dn_type == DMU_OT_USERGROUP_USED);
792 /* XXX: f_ver is not counted, but may differ too */
793 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
795 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
799 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
800 const struct lu_buf *buf, const char *name, int fl,
801 struct osd_thandle *oh)
805 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
809 if (osd_obj2dev(obj)->od_xattr_in_sa) {
810 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
812 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
814 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
820 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
822 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
823 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
824 #ifdef ZFS_PROJINHERIT
825 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
827 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
830 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
832 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
833 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
834 #ifdef ZFS_PROJINHERIT
835 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
837 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
842 static inline uint64_t
843 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
844 int dnodesize, dmu_tx_t *tx)
847 dnodesize = max_t(int, dmu_objset_dnodesize(os),
850 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
851 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
854 static inline uint64_t
855 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
856 dmu_object_type_t ot, int leaf_blockshift,
857 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
860 dnodesize = max_t(int, dmu_objset_dnodesize(os),
863 return zap_create_flags_dnsize(os, normflags, flags, ot,
864 leaf_blockshift, indirect_blockshift,
865 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
870 osd_obj_bonuslen(struct osd_object *obj)
872 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
874 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
875 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
877 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
881 dnodesize = dmu_objset_dnodesize(os);
883 bonuslen = DN_BONUS_SIZE(dnodesize);
890 static inline int osd_sa_handle_get(struct osd_object *obj)
892 struct osd_device *osd = osd_obj2dev(obj);
893 dnode_t *dn = obj->oo_dn;
899 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
900 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
901 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
904 zfs_refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
908 static inline void osd_dnode_rele(dnode_t *dn)
912 LASSERT(dn->dn_bonus);
915 dmu_buf_rele(&db->db, osd_obj_tag);
918 static inline uint64_t osd_db_dirty_txg(dmu_buf_impl_t *db)
920 dbuf_dirty_record_t *dr;
923 mutex_enter(&db->db_mtx);
924 #ifdef HAVE_DB_DIRTY_RECORDS_LIST
925 dr = list_head(&db->db_dirty_records);
927 dr = db->db_last_dirty;
931 mutex_exit(&db->db_mtx);
936 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
938 return dmu_objset_userobjspace_present(osd->od_os);
941 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
942 dnode_t *dn, const char *key,
943 int int_size, int int_num,
944 const void *val, dmu_tx_t *tx)
949 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
951 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
954 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
955 dnode_t *dn, const char *key,
961 return -zap_remove_by_dnode(dn, key, tx);
963 return -zap_remove(osd->od_os, zap, key, tx);
967 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
968 dnode_t *dn, const char *key,
969 int int_size, int int_num, void *v)
974 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
976 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
979 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
980 dnode_t *dn, int add, const char *name)
983 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
986 dmu_tx_hold_zap(tx, zap, add, name);
989 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
990 dnode_t *dn, uint64_t off, int len)
993 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
996 dmu_tx_hold_write(tx, oid, off, len);
999 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1000 uint64_t offset, uint64_t size,
1001 const char *buf, dmu_tx_t *tx)
1004 dmu_write_by_dnode(dn, offset, size, buf, tx);
1007 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1008 uint64_t offset, uint64_t size,
1009 char *buf, int flags)
1012 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1015 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1016 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1017 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1019 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1020 dmu_objset_own((name), (type), (ronly), (tag), (os))
1023 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1024 #define osd_dmu_objset_disown(os, decrypt, tag) \
1025 dmu_objset_disown((os), (decrypt), (tag))
1027 #define osd_dmu_objset_disown(os, decrypt, tag) \
1028 dmu_objset_disown((os), (tag))
1032 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1033 __u32 keysize, __u32 recsize)
1035 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1036 &osd->od_index_backup_list, &osd->od_lock,
1037 &osd->od_index_backup_stop,
1038 fid, keysize, recsize);
1042 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1044 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1047 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1048 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1049 osd->od_index_backup_id);
1053 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1054 &osd->od_index_backup_list, &osd->od_lock,
1055 &osd->od_index_backup_stop, backup);
1058 #ifndef HAVE_DMU_TX_MARK_NETFREE
1059 #define dmu_tx_mark_netfree(tx)
1062 #ifndef HAVE_ZFS_INODE_TIMESPEC
1063 #define inode_timespec_t timestruc_t
1066 #ifdef HAVE_DMU_OFFSET_NEXT
1067 #define osd_dmu_offset_next(os, obj, hole, res) \
1068 dmu_offset_next((os), (obj), (hole), (res))
1070 #define osd_dmu_offset_next(os, obj, hole, res) (EOPNOTSUPP)
1073 extern char osd_0copy_tag[];
1075 #endif /* _OSD_INTERNAL_H */