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 <dt_object.h>
43 #include <md_object.h>
44 #include <lustre_quota.h>
45 #include <lustre_scrub.h>
51 #include <sys/nvpair.h>
52 #include <sys/zfs_znode.h>
55 #include <sys/dmu_objset.h>
56 #include <lustre_scrub.h>
59 * By design including kmem.h overrides the Linux slab interfaces to provide
60 * the Illumos kmem cache interfaces. To override this and gain access to
61 * the Linux interfaces these preprocessor macros must be undefined.
63 #ifdef kmem_cache_destroy
64 #undef kmem_cache_destroy
67 #ifdef kmem_cache_create
68 #undef kmem_cache_create
71 #ifdef kmem_cache_alloc
72 #undef kmem_cache_alloc
75 #ifdef kmem_cache_free
76 #undef kmem_cache_free
79 #define ZFS_VERSION_CODE \
80 OBD_OCD_VERSION(ZFS_MAJOR, ZFS_MINOR, ZFS_PATCH, ZFS_FIX)
82 #define LUSTRE_ROOT_FID_SEQ 0
83 #define DMU_OSD_SVNAME "svname"
84 #define DMU_OSD_OI_NAME_BASE "oi"
86 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
88 /* Statfs space reservation for grant, fragmentation, and unlink space. */
89 #define OSD_STATFS_RESERVED_SIZE (16ULL << 20) /* reserve 16MB minimum */
90 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
92 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
93 #define OSD_DNODE_MIN_BLKSHIFT (DNODES_PER_BLOCK_SHIFT)
94 #define OSD_DNODE_EST_BLKSHIFT (12) /* est 4KB/dnode */
95 #define OSD_DNODE_EST_COUNT 4096
97 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
99 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
101 #ifndef HAVE_ZFS_REFCOUNT_ADD
102 #define zfs_refcount_add refcount_add
105 extern const struct dt_body_operations osd_body_scrub_ops;
106 extern const struct dt_body_operations osd_body_ops;
107 extern struct kmem_cache *osd_object_kmem;
110 * Iterator's in-memory data structure for quota file.
112 struct osd_it_quota {
113 struct osd_object *oiq_obj;
114 /* DMU accounting object id */
117 zap_cursor_t *oiq_zc;
118 /** identifier for current quota record */
120 unsigned oiq_reset:1; /* 1 -- no need to advance */
125 OZI_POS_DOT = 1, /* cursor at . */
126 OZI_POS_DOTDOT = 2, /* cursor at .. */
127 OZI_POS_REAL = 3, /* cursor at real entries */
131 * regular ZFS direntry
133 struct zpl_direntry {
134 uint64_t zde_dnode:48,
137 } __attribute__((packed));
140 * lustre direntry adds a fid to regular ZFS direntry
142 struct luz_direntry {
143 struct zpl_direntry lzd_reg;
144 struct lu_fid lzd_fid;
145 } __attribute__((packed));
148 * Iterator's in-memory data structure for ZAPs
150 * ZFS does not store . and .. on a disk, instead they are
151 * generated up on request
152 * to follow this format we do the same
155 zap_cursor_t *ozi_zc;
156 struct osd_object *ozi_obj;
157 unsigned ozi_reset:1; /* 1 -- no need to advance */
158 /* ozi_pos - position of the cursor */
159 enum osd_zap_pos ozi_pos;
160 struct luz_direntry ozi_zde;
161 zap_attribute_t ozi_za;
163 char ozi_name[MAXNAMELEN]; /* file name for dir */
164 __u64 ozi_key; /* binary key for index files */
167 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
169 /* cached SA attributes */
174 #ifdef ZFS_PROJINHERIT
188 #define OSD_INS_CACHE_SIZE 8
191 struct osd_idmap_cache {
192 struct osd_device *oic_dev;
193 struct lu_fid oic_fid;
194 /** max 2^48 dnodes per dataset, avoid spilling into another word */
195 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
196 oic_remote:1; /* FID isn't local */
199 struct osd_inconsistent_item {
200 /* link into lustre_scrub::os_inconsistent_items,
201 * protected by lustr_scrub::os_lock. */
202 struct list_head oii_list;
204 /* The right FID <=> oid mapping. */
205 struct osd_idmap_cache oii_cache;
207 unsigned int oii_insert:1; /* insert or update mapping. */
210 struct osd_otable_it {
211 struct osd_device *ooi_dev;
212 struct lu_fid ooi_fid;
214 __u64 ooi_prefetched_dnode;
217 /* The following bits can be updated/checked w/o lock protection.
218 * If more bits will be introduced in the future and need lock to
219 * protect, please add comment. */
220 unsigned int ooi_used_outside:1, /* Some user out of OSD
221 * uses the iteration. */
222 ooi_all_cached:1, /* No more entries can be
223 * filled into cache. */
224 ooi_user_ready:1, /* The user out of OSD is
225 * ready to iterate. */
226 ooi_waiting:1; /* it::next is waiting. */
229 extern const struct dt_index_operations osd_otable_ops;
231 /* max.number of regular attributes the callers may ask for */
232 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
234 struct osd_thread_info {
235 const struct lu_env *oti_env;
237 struct lu_fid oti_fid;
239 struct ost_id oti_ostid;
245 char oti_key[MAXNAMELEN + 1];
246 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
247 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
249 struct lustre_mdt_attrs oti_mdt_attrs;
250 unsigned int oti_in_trans:1;
252 struct lu_attr oti_la;
253 struct osa_attr oti_osa;
254 zap_attribute_t oti_za;
255 zap_attribute_t oti_za2;
256 dmu_object_info_t oti_doi;
257 struct luz_direntry oti_zde;
259 struct lquota_id_info oti_qi;
260 struct lu_seq_range oti_seq_range;
262 /* dedicated OI cache for insert (which needs inum) */
263 struct osd_idmap_cache *oti_ins_cache;
264 int oti_ins_cache_size;
265 int oti_ins_cache_used;
266 /* inc by osd_trans_create and dec by osd_trans_stop */
267 int oti_ins_cache_depth;
268 struct lu_buf oti_xattr_lbuf;
270 zap_cursor_t oti_zc2;
276 extern struct lu_context_key osd_key;
278 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
280 return lu_context_key_get(&env->le_ctx, &osd_key);
284 struct thandle ot_super;
285 struct list_head ot_dcb_list;
286 struct list_head ot_stop_dcb_list;
287 struct list_head ot_unlinked_list;
288 struct list_head ot_sa_list;
290 struct lquota_trans ot_quota_trans;
294 #define OSD_OI_NAME_SIZE 24
297 * Object Index (OI) instance.
300 char oi_name[OSD_OI_NAME_SIZE];
307 uint64_t *os_compat_dirs;
308 int os_subdir_count; /* subdir count for each seq */
309 u64 os_seq; /* seq number */
310 struct list_head os_seq_list; /* list to seq_list */
313 struct osd_seq_list {
314 rwlock_t osl_seq_list_lock; /* lock for seq_list */
315 struct list_head osl_seq_list; /* list head for seq */
316 struct semaphore osl_seq_init_sem;
319 #define OSD_OST_MAP_SIZE 32
326 struct dt_device od_dt_dev;
327 /* information about underlying file system */
328 struct objset *od_os;
329 uint64_t od_rootid; /* id of root znode */
330 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
331 /* SA attr mapping->id,
332 * name is the same as in ZFS to use defines SA_ZPL_...*/
333 sa_attr_type_t *z_attr_table;
335 struct proc_dir_entry *od_proc_entry;
336 struct lprocfs_stats *od_stats;
338 uint64_t od_remote_parent_dir;
339 uint64_t od_index_backup_id;
340 uint64_t od_max_blksz;
343 struct osd_oi **od_oi_table;
344 unsigned int od_oi_count;
345 struct osd_seq_list od_seq_list;
347 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
348 od_prop_rdonly:1, /**< ZFS property readonly */
355 unsigned int od_dnsize;
356 int od_index_backup_stop;
358 enum lustre_index_backup_policy od_index_backup_policy;
365 struct lu_site od_site;
367 dnode_t *od_groupused_dn;
368 dnode_t *od_userused_dn;
369 #ifdef ZFS_PROJINHERIT
370 dnode_t *od_projectused_dn;
373 /* quota slave instance for inode */
374 struct qsd_instance *od_quota_slave_md;
376 /* quota slave instance for block */
377 struct qsd_instance *od_quota_slave_dt;
379 struct brw_stats od_brw_stats;
380 atomic_t od_r_in_flight;
381 atomic_t od_w_in_flight;
383 /* used to debug zerocopy logic: the fields track all
384 * allocated, loaned and referenced buffers in use.
385 * to be removed once the change is tested well. */
386 atomic_t od_zerocopy_alloc;
387 atomic_t od_zerocopy_loan;
388 atomic_t od_zerocopy_pin;
390 arc_prune_t *arc_prune_cb;
392 /* osd seq instance */
393 struct lu_client_seq *od_cl_seq;
395 struct semaphore od_otable_sem;
396 struct osd_otable_it *od_otable_it;
397 struct lustre_scrub od_scrub;
398 struct list_head od_ios_list;
399 struct list_head od_index_backup_list;
400 struct list_head od_index_restore_list;
402 unsigned long long od_readcache_max_filesize;
405 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
408 return osd->od_quota_slave_dt;
410 return osd->od_quota_slave_md;
413 enum osd_destroy_type {
414 OSD_DESTROY_NONE = 0,
415 OSD_DESTROY_SYNC = 1,
416 OSD_DESTROY_ASYNC = 2,
420 struct dt_object oo_dt;
422 * Inode for file system object represented by this osd_object. This
423 * inode is pinned for the whole duration of lu_object life.
425 * Not modified concurrently (either setup early during object
426 * creation, or assigned by osd_create() under write lock).
429 sa_handle_t *oo_sa_hdl;
430 nvlist_t *oo_sa_xattr;
431 struct list_head oo_sa_linkage;
433 /* used to implement osd_object_*_{lock|unlock} */
434 struct rw_semaphore oo_sem;
436 /* to serialize some updates: destroy vs. others,
437 * xattr_set, object block size change etc */
438 struct rw_semaphore oo_guard;
440 /* protected by oo_guard */
441 struct list_head oo_unlinked_linkage;
443 /* cached attributes */
444 rwlock_t oo_attr_lock;
445 struct lu_attr oo_attr;
447 /* external dnode holding large EAs, protected by oo_guard */
449 enum osd_destroy_type oo_destroy;
451 __u32 oo_destroyed:1,
453 #ifdef ZFS_PROJINHERIT
459 /* the i_flags in LMA */
462 int oo_ea_in_bonus; /* EA bytes we expect */
464 /* record size for index file */
465 unsigned char oo_keysize;
466 unsigned char oo_recsize;
467 unsigned char oo_recusize; /* unit size */
469 uint64_t oo_parent; /* used only at object creation */
471 struct lu_object_header *oo_header;
474 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *,
475 struct obd_statfs_info *);
476 extern const struct dt_index_operations osd_acct_index_ops;
477 extern const struct lu_device_operations osd_lu_ops;
478 extern const struct dt_index_operations osd_dir_ops;
479 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
480 qid_t uid, qid_t gid, qid_t projid, long long space,
481 struct osd_thandle *oh,
482 enum osd_quota_local_flags *local_flags,
483 enum osd_qid_declare_flags osd_qid_declare_flags);
484 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
485 uint64_t nrblocks, uint64_t est_maxblockshift);
486 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
492 static inline int lu_device_is_osd(const struct lu_device *d)
494 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
497 static inline struct osd_object *osd_obj(const struct lu_object *o)
499 LASSERT(lu_device_is_osd(o->lo_dev));
500 return container_of(o, struct osd_object, oo_dt.do_lu);
503 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
505 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
506 return container_of(d, struct osd_device, od_dt_dev);
509 static inline struct osd_device *osd_dev(const struct lu_device *d)
511 LASSERT(lu_device_is_osd(d));
512 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
515 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
517 return osd_obj(&d->do_lu);
520 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
522 return osd_dev(o->oo_dt.do_lu.lo_dev);
525 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
527 return &osd->od_dt_dev.dd_lu_dev;
530 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
532 return osd_dev(o->do_lu.lo_dev)->od_os;
535 static inline int osd_invariant(const struct osd_object *obj)
541 * Put the osd object once done with it.
543 * \param obj osd object that needs to be put
545 static inline void osd_object_put(const struct lu_env *env,
546 struct osd_object *obj)
548 dt_object_put(env, &obj->oo_dt);
551 static inline int osd_object_invariant(const struct lu_object *l)
553 return osd_invariant(osd_obj(l));
556 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
558 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
561 static inline char *osd_name(struct osd_device *osd)
563 return osd->od_svname;
566 static inline void zfs_set_bit(int nr, __u8 *addr)
568 set_bit(nr, (unsigned long *)addr);
571 static inline int zfs_test_bit(int nr, __u8 *addr)
573 return test_bit(nr, (const unsigned long *)addr);
576 static inline int osd_oi_fid2idx(struct osd_device *dev,
577 const struct lu_fid *fid)
579 return fid->f_seq & (dev->od_oi_count - 1);
582 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
583 const struct lu_fid *fid)
585 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
586 "%s: "DFID", oi_count %d\n",
587 osd_name(osd), PFID(fid), osd->od_oi_count);
589 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
592 #ifdef CONFIG_PROC_FS
594 LPROC_OSD_READ_BYTES = 0,
595 LPROC_OSD_WRITE_BYTES = 1,
596 LPROC_OSD_GET_PAGE = 2,
597 LPROC_OSD_NO_PAGE = 3,
598 LPROC_OSD_CACHE_ACCESS = 4,
599 LPROC_OSD_CACHE_HIT = 5,
600 LPROC_OSD_CACHE_MISS = 6,
601 LPROC_OSD_COPY_IO = 7,
602 LPROC_OSD_ZEROCOPY_IO = 8,
603 LPROC_OSD_TAIL_IO = 9,
607 extern struct kmem_cache *osd_zapit_cachep;
609 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
611 int osd_procfs_init(struct osd_device *osd, const char *name);
612 int osd_procfs_fini(struct osd_device *osd);
615 extern char *osd_obj_tag;
616 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
617 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
618 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
619 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
620 uint64_t oid, dmu_buf_t **dbp);
621 struct lu_object *osd_object_alloc(const struct lu_env *env,
622 const struct lu_object_header *hdr,
623 struct lu_device *d);
624 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
625 void *buf, uint32_t buflen, struct osd_thandle *oh);
626 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
627 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
628 unsigned dnsize, zap_flags_t flags);
629 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
630 struct osd_object *obj, const struct lu_fid *fid,
631 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
632 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
633 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
634 struct lu_attr *la, uint64_t parent, nvlist_t *);
635 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
636 uint64_t oid, dnode_t **dnp);
639 int osd_oi_init(const struct lu_env *env, struct osd_device *o, bool reset);
640 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
641 int osd_fid_lookup(const struct lu_env *env,
642 struct osd_device *, const struct lu_fid *, uint64_t *);
643 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
644 const struct lu_fid *fid, char *buf, int bufsize,
646 int osd_options_init(void);
647 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
649 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
650 struct osd_object *obj);
651 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
652 struct osd_device *osd,
653 const struct lu_fid *fid);
654 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
655 struct osd_device *osd,
656 const struct lu_fid *fid);
657 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
658 struct osd_device *osd,
659 const struct lu_fid *fid,
661 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
662 const struct lu_fid *fid);
663 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
664 uint64_t parent, const char *name, uint64_t *child,
665 const struct lu_fid *fid, bool isdir);
667 extern unsigned int osd_oi_count;
670 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
671 uint64_t oid, struct lu_fid *fid);
672 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
673 const struct dt_index_features *feat);
674 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
675 u64 seq, struct lu_seq_range *range);
676 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
677 uint64_t id, uint64_t dirhash);
678 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
679 uint64_t id, uint64_t dirhash);
680 void osd_zap_cursor_fini(zap_cursor_t *zc);
681 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
682 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
683 const struct lu_fid *fid);
684 int osd_add_to_remote_parent(const struct lu_env *env,
685 struct osd_device *osd,
686 struct osd_object *obj,
687 struct osd_thandle *oh);
688 int osd_delete_from_remote_parent(const struct lu_env *env,
689 struct osd_device *osd,
690 struct osd_object *obj,
691 struct osd_thandle *oh, bool destroy);
692 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
696 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev,
697 time64_t interval, bool resetoi);
698 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
699 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
701 void osd_scrub_stop(struct osd_device *dev);
702 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
703 const struct lu_fid *fid, uint64_t oid, bool insert);
704 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
708 * Basic transaction credit op
717 int osd_scrub_refresh_mapping(const struct lu_env *env,
718 struct osd_device *dev,
719 const struct lu_fid *fid,
720 uint64_t oid, enum dt_txn_op ops,
721 bool force, const char *name);
725 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
726 struct osd_object *obj,
727 struct osd_thandle *oh);
728 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
729 struct osd_thandle *oh);
730 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
731 struct osd_thandle *oh);
732 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
734 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
735 uint64_t xattr, struct lu_buf *buf,
736 const char *name, int *sizep);
737 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
738 struct lu_buf *buf, const char *name, int *sizep);
739 int osd_xattr_get_lma(const struct lu_env *env, struct osd_object *obj,
741 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
742 struct lu_buf *buf, const char *name);
743 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
744 const struct lu_buf *buf, const char *name,
745 int fl, struct thandle *handle);
746 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
747 const struct lu_buf *buf, const char *name, int fl,
748 struct thandle *handle);
749 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
750 const char *name, struct thandle *handle);
751 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
752 const char *name, struct thandle *handle);
753 void osd_declare_xattrs_destroy(const struct lu_env *env,
754 struct osd_object *obj,
755 struct osd_thandle *oh);
756 int osd_xattrs_destroy(const struct lu_env *env,
757 struct osd_object *obj, struct osd_thandle *oh);
758 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
759 const struct lu_buf *lb);
760 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
761 int vallen, const char *name, struct osd_thandle *oh);
762 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
763 const struct lu_buf *buf, const char *name, int fl,
764 struct osd_thandle *oh);;
765 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
766 const struct lu_buf *buf, const char *name, int fl,
767 struct osd_thandle *oh);
768 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
769 struct osd_thandle *oh);
771 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
772 sizeof(__u64) /* VBR VERSION */ + \
773 sizeof(struct lustre_mdt_attrs) /* LMA */)
775 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
776 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
779 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
781 return DN_MAX_BONUSLEN;
785 static inline int osd_object_is_zap(dnode_t *dn)
787 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
788 dn->dn_type == DMU_OT_USERGROUP_USED);
791 /* XXX: f_ver is not counted, but may differ too */
792 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
794 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
798 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
799 const struct lu_buf *buf, const char *name, int fl,
800 struct osd_thandle *oh)
804 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
808 if (osd_obj2dev(obj)->od_xattr_in_sa) {
809 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
811 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
813 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
819 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
821 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
822 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
823 #ifdef ZFS_PROJINHERIT
824 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
826 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
829 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
831 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
832 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
833 #ifdef ZFS_PROJINHERIT
834 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
836 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
841 #ifndef HAVE_DSL_POOL_CONFIG
842 static inline void dsl_pool_config_enter(dsl_pool_t *dp, void *name)
846 static inline void dsl_pool_config_exit(dsl_pool_t *dp, void *name)
851 #ifdef HAVE_SPA_MAXBLOCKSIZE
852 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
853 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
855 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
856 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
857 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
860 #ifdef HAVE_SA_SPILL_ALLOC
862 osd_zio_buf_alloc(size_t size)
864 return sa_spill_alloc(KM_SLEEP);
868 osd_zio_buf_free(void *buf, size_t size)
873 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
874 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
877 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
878 static inline uint64_t
879 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
880 int dnodesize, dmu_tx_t *tx)
883 dnodesize = max_t(int, dmu_objset_dnodesize(os),
886 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
887 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
890 static inline uint64_t
891 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
892 dmu_object_type_t ot, int leaf_blockshift,
893 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
896 dnodesize = max_t(int, dmu_objset_dnodesize(os),
899 return zap_create_flags_dnsize(os, normflags, flags, ot,
900 leaf_blockshift, indirect_blockshift,
901 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
906 osd_obj_bonuslen(struct osd_object *obj)
908 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
910 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
911 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
913 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
917 dnodesize = dmu_objset_dnodesize(os);
919 bonuslen = DN_BONUS_SIZE(dnodesize);
926 static inline uint64_t
927 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
928 int dnodesize, dmu_tx_t *tx)
930 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
931 DN_MAX_BONUSLEN, tx);
934 static inline uint64_t
935 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
936 dmu_object_type_t ot, int leaf_blockshift,
937 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
939 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
940 indirect_blockshift, DMU_OT_SA,
941 DN_MAX_BONUSLEN, tx);
945 osd_obj_bonuslen(struct osd_object *obj)
947 return DN_MAX_BONUSLEN;
949 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
951 #ifdef HAVE_DMU_PREFETCH_6ARG
952 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
953 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
955 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
956 dmu_prefetch((os), (obj), (lvl), (off))
959 static inline int osd_sa_handle_get(struct osd_object *obj)
961 struct osd_device *osd = osd_obj2dev(obj);
962 dnode_t *dn = obj->oo_dn;
968 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
969 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
970 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
973 zfs_refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
977 static inline void osd_dnode_rele(dnode_t *dn)
981 LASSERT(dn->dn_bonus);
984 dmu_buf_rele(&db->db, osd_obj_tag);
987 static inline uint64_t osd_db_dirty_txg(dmu_buf_impl_t *db)
989 dbuf_dirty_record_t *dr;
992 mutex_enter(&db->db_mtx);
993 #ifdef HAVE_DB_DIRTY_RECORDS_LIST
994 dr = list_head(&db->db_dirty_records);
996 dr = db->db_last_dirty;
1000 mutex_exit(&db->db_mtx);
1005 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
1007 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
1008 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
1010 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
1012 return dmu_objset_userobjspace_present(osd->od_os);
1016 #define OSD_DMU_USEROBJ_PREFIX "obj-"
1017 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
1019 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
1023 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
1025 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
1026 dnode_t *dn, const char *key,
1027 int int_size, int int_num,
1028 const void *val, dmu_tx_t *tx)
1032 #ifdef HAVE_ZAP_ADD_BY_DNODE
1034 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
1036 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
1039 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
1040 dnode_t *dn, const char *key,
1045 #ifdef HAVE_ZAP_ADD_BY_DNODE
1047 return -zap_remove_by_dnode(dn, key, tx);
1049 return -zap_remove(osd->od_os, zap, key, tx);
1053 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
1054 dnode_t *dn, const char *key,
1055 int int_size, int int_num, void *v)
1059 #ifdef HAVE_ZAP_ADD_BY_DNODE
1061 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
1063 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
1066 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
1067 dnode_t *dn, int add, const char *name)
1069 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1071 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
1075 dmu_tx_hold_zap(tx, zap, add, name);
1078 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
1079 dnode_t *dn, uint64_t off, int len)
1081 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1083 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1087 dmu_tx_hold_write(tx, oid, off, len);
1090 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1091 uint64_t offset, uint64_t size,
1092 const char *buf, dmu_tx_t *tx)
1095 #ifdef HAVE_DMU_WRITE_BY_DNODE
1096 dmu_write_by_dnode(dn, offset, size, buf, tx);
1098 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1102 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1103 uint64_t offset, uint64_t size,
1104 char *buf, int flags)
1107 #ifdef HAVE_DMU_READ_BY_DNODE
1108 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1110 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1114 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1115 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1116 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1118 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1119 dmu_objset_own((name), (type), (ronly), (tag), (os))
1122 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1123 #define osd_dmu_objset_disown(os, decrypt, tag) \
1124 dmu_objset_disown((os), (decrypt), (tag))
1126 #define osd_dmu_objset_disown(os, decrypt, tag) \
1127 dmu_objset_disown((os), (tag))
1131 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1132 __u32 keysize, __u32 recsize)
1134 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1135 &osd->od_index_backup_list, &osd->od_lock,
1136 &osd->od_index_backup_stop,
1137 fid, keysize, recsize);
1141 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1143 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1146 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1147 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1148 osd->od_index_backup_id);
1152 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1153 &osd->od_index_backup_list, &osd->od_lock,
1154 &osd->od_index_backup_stop, backup);
1157 #ifndef HAVE_DMU_TX_MARK_NETFREE
1158 #define dmu_tx_mark_netfree(tx)
1161 #ifndef HAVE_ZFS_INODE_TIMESPEC
1162 #define inode_timespec_t timestruc_t
1165 #ifdef HAVE_DMU_OFFSET_NEXT
1166 #define osd_dmu_offset_next(os, obj, hole, res) \
1167 dmu_offset_next((os), (obj), (hole), (res))
1169 #define osd_dmu_offset_next(os, obj, hole, res) (EOPNOTSUPP)
1172 #endif /* _OSD_INTERNAL_H */