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 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
102 extern struct dt_body_operations osd_body_scrub_ops;
105 * Iterator's in-memory data structure for quota file.
107 struct osd_it_quota {
108 struct osd_object *oiq_obj;
109 /* DMU accounting object id */
112 zap_cursor_t *oiq_zc;
113 /** identifier for current quota record */
115 unsigned oiq_reset:1; /* 1 -- no need to advance */
119 * Iterator's in-memory data structure for ZAPs
121 * ZFS does not store . and .. on a disk, instead they are
122 * generated up on request
123 * to follow this format we do the same
126 zap_cursor_t *ozi_zc;
127 struct osd_object *ozi_obj;
128 unsigned ozi_reset:1; /* 1 -- no need to advance */
129 /* ozi_pos - position of the cursor:
130 * 0 - before any record
133 * 3 - real records */
136 char ozi_name[MAXNAMELEN]; /* file name for dir */
137 __u64 ozi_key; /* binary key for index files */
140 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
143 * regular ZFS direntry
145 struct zpl_direntry {
146 uint64_t zde_dnode:48,
149 } __attribute__((packed));
152 * lustre direntry adds a fid to regular ZFS direntry
154 struct luz_direntry {
155 struct zpl_direntry lzd_reg;
156 struct lu_fid lzd_fid;
157 } __attribute__((packed));
160 /* cached SA attributes */
165 #ifdef ZFS_PROJINHERIT
178 #define OSD_INS_CACHE_SIZE 8
181 struct osd_idmap_cache {
182 struct osd_device *oic_dev;
183 struct lu_fid oic_fid;
184 /** max 2^48 dnodes per dataset, avoid spilling into another word */
185 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
186 oic_remote:1; /* FID isn't local */
189 struct osd_inconsistent_item {
190 /* link into lustre_scrub::os_inconsistent_items,
191 * protected by lustr_scrub::os_lock. */
192 struct list_head oii_list;
194 /* The right FID <=> oid mapping. */
195 struct osd_idmap_cache oii_cache;
197 unsigned int oii_insert:1; /* insert or update mapping. */
200 struct osd_otable_it {
201 struct osd_device *ooi_dev;
202 struct lu_fid ooi_fid;
204 __u64 ooi_prefetched_dnode;
207 /* The following bits can be updated/checked w/o lock protection.
208 * If more bits will be introduced in the future and need lock to
209 * protect, please add comment. */
210 unsigned int ooi_used_outside:1, /* Some user out of OSD
211 * uses the iteration. */
212 ooi_all_cached:1, /* No more entries can be
213 * filled into cache. */
214 ooi_user_ready:1, /* The user out of OSD is
215 * ready to iterate. */
216 ooi_waiting:1; /* it::next is waiting. */
219 extern const struct dt_index_operations osd_otable_ops;
221 /* max.number of regular attributes the callers may ask for */
222 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
224 struct osd_thread_info {
225 const struct lu_env *oti_env;
227 struct lu_fid oti_fid;
229 * XXX temporary: for ->i_op calls.
231 struct timespec oti_time;
233 struct ost_id oti_ostid;
239 char oti_key[MAXNAMELEN + 1];
240 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
241 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
243 struct lustre_mdt_attrs oti_mdt_attrs;
244 unsigned int oti_in_trans:1;
246 struct lu_attr oti_la;
247 struct osa_attr oti_osa;
248 zap_attribute_t oti_za;
249 zap_attribute_t oti_za2;
250 dmu_object_info_t oti_doi;
251 struct luz_direntry oti_zde;
253 struct lquota_id_info oti_qi;
254 struct lu_seq_range oti_seq_range;
256 /* dedicated OI cache for insert (which needs inum) */
257 struct osd_idmap_cache *oti_ins_cache;
258 int oti_ins_cache_size;
259 int oti_ins_cache_used;
260 /* inc by osd_trans_create and dec by osd_trans_stop */
261 int oti_ins_cache_depth;
262 struct lu_buf oti_xattr_lbuf;
264 zap_cursor_t oti_zc2;
267 extern struct lu_context_key osd_key;
269 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
271 return lu_context_key_get(&env->le_ctx, &osd_key);
275 struct thandle ot_super;
276 struct list_head ot_dcb_list;
277 struct list_head ot_stop_dcb_list;
278 struct list_head ot_unlinked_list;
279 struct list_head ot_sa_list;
281 struct lquota_trans ot_quota_trans;
285 #define OSD_OI_NAME_SIZE 24
288 * Object Index (OI) instance.
291 char oi_name[OSD_OI_NAME_SIZE];
298 uint64_t *os_compat_dirs;
299 int os_subdir_count; /* subdir count for each seq */
300 u64 os_seq; /* seq number */
301 struct list_head os_seq_list; /* list to seq_list */
304 struct osd_seq_list {
305 rwlock_t osl_seq_list_lock; /* lock for seq_list */
306 struct list_head osl_seq_list; /* list head for seq */
307 struct semaphore osl_seq_init_sem;
310 #define OSD_OST_MAP_SIZE 32
317 struct dt_device od_dt_dev;
318 /* information about underlying file system */
319 struct objset *od_os;
320 uint64_t od_rootid; /* id of root znode */
321 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
322 /* SA attr mapping->id,
323 * name is the same as in ZFS to use defines SA_ZPL_...*/
324 sa_attr_type_t *z_attr_table;
326 struct proc_dir_entry *od_proc_entry;
327 struct lprocfs_stats *od_stats;
329 uint64_t od_remote_parent_dir;
330 uint64_t od_index_backup_id;
331 uint64_t od_max_blksz;
334 struct osd_oi **od_oi_table;
335 unsigned int od_oi_count;
336 struct osd_seq_list od_seq_list;
338 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
339 od_prop_rdonly:1, /**< ZFS property readonly */
344 unsigned int od_dnsize;
345 int od_index_backup_stop;
347 enum lustre_index_backup_policy od_index_backup_policy;
354 __s64 od_auto_scrub_interval;
355 struct lu_site od_site;
357 dnode_t *od_groupused_dn;
358 dnode_t *od_userused_dn;
359 #ifdef ZFS_PROJINHERIT
360 dnode_t *od_projectused_dn;
363 /* quota slave instance */
364 struct qsd_instance *od_quota_slave;
366 struct brw_stats od_brw_stats;
367 atomic_t od_r_in_flight;
368 atomic_t od_w_in_flight;
370 /* used to debug zerocopy logic: the fields track all
371 * allocated, loaned and referenced buffers in use.
372 * to be removed once the change is tested well. */
373 atomic_t od_zerocopy_alloc;
374 atomic_t od_zerocopy_loan;
375 atomic_t od_zerocopy_pin;
377 arc_prune_t *arc_prune_cb;
379 /* osd seq instance */
380 struct lu_client_seq *od_cl_seq;
382 struct semaphore od_otable_sem;
383 struct osd_otable_it *od_otable_it;
384 struct lustre_scrub od_scrub;
385 struct list_head od_ios_list;
386 struct list_head od_index_backup_list;
387 struct list_head od_index_restore_list;
389 unsigned long long od_readcache_max_filesize;
392 enum osd_destroy_type {
393 OSD_DESTROY_NONE = 0,
394 OSD_DESTROY_SYNC = 1,
395 OSD_DESTROY_ASYNC = 2,
399 struct dt_object oo_dt;
401 * Inode for file system object represented by this osd_object. This
402 * inode is pinned for the whole duration of lu_object life.
404 * Not modified concurrently (either setup early during object
405 * creation, or assigned by osd_create() under write lock).
408 sa_handle_t *oo_sa_hdl;
409 nvlist_t *oo_sa_xattr;
410 struct list_head oo_sa_linkage;
412 /* used to implement osd_object_*_{lock|unlock} */
413 struct rw_semaphore oo_sem;
415 /* to serialize some updates: destroy vs. others,
416 * xattr_set, object block size change etc */
417 struct rw_semaphore oo_guard;
419 /* protected by oo_guard */
420 struct list_head oo_unlinked_linkage;
422 /* cached attributes */
423 rwlock_t oo_attr_lock;
424 struct lu_attr oo_attr;
426 /* external dnode holding large EAs, protected by oo_guard */
428 enum osd_destroy_type oo_destroy;
430 __u32 oo_destroyed:1,
432 #ifdef ZFS_PROJINHERIT
438 /* the i_flags in LMA */
441 int oo_ea_in_bonus; /* EA bytes we expect */
443 /* record size for index file */
444 unsigned char oo_keysize;
445 unsigned char oo_recsize;
446 unsigned char oo_recusize; /* unit size */
448 uint64_t oo_parent; /* used only at object creation */
450 struct lu_object_header *oo_header;
453 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
454 extern const struct dt_index_operations osd_acct_index_ops;
455 extern struct lu_device_operations osd_lu_ops;
456 extern struct dt_index_operations osd_dir_ops;
457 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
458 qid_t uid, qid_t gid, qid_t projid, long long space,
459 struct osd_thandle *oh, int *flags,
460 enum osd_qid_declare_flags osd_qid_declare_flags);
461 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
462 uint64_t nrblocks, uint64_t est_maxblockshift);
463 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
469 static inline int lu_device_is_osd(const struct lu_device *d)
471 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
474 static inline struct osd_object *osd_obj(const struct lu_object *o)
476 LASSERT(lu_device_is_osd(o->lo_dev));
477 return container_of0(o, struct osd_object, oo_dt.do_lu);
480 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
482 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
483 return container_of0(d, struct osd_device, od_dt_dev);
486 static inline struct osd_device *osd_dev(const struct lu_device *d)
488 LASSERT(lu_device_is_osd(d));
489 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
492 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
494 return osd_obj(&d->do_lu);
497 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
499 return osd_dev(o->oo_dt.do_lu.lo_dev);
502 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
504 return &osd->od_dt_dev.dd_lu_dev;
507 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
509 return osd_dev(o->do_lu.lo_dev)->od_os;
512 static inline int osd_invariant(const struct osd_object *obj)
518 * Put the osd object once done with it.
520 * \param obj osd object that needs to be put
522 static inline void osd_object_put(const struct lu_env *env,
523 struct osd_object *obj)
525 dt_object_put(env, &obj->oo_dt);
528 static inline int osd_object_invariant(const struct lu_object *l)
530 return osd_invariant(osd_obj(l));
533 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
535 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
538 static inline char *osd_name(struct osd_device *osd)
540 return osd->od_svname;
543 static inline void zfs_set_bit(int nr, __u8 *addr)
545 set_bit(nr, (unsigned long *)addr);
548 static inline int zfs_test_bit(int nr, __u8 *addr)
550 return test_bit(nr, (const unsigned long *)addr);
553 static inline int osd_oi_fid2idx(struct osd_device *dev,
554 const struct lu_fid *fid)
556 return fid->f_seq & (dev->od_oi_count - 1);
559 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
560 const struct lu_fid *fid)
562 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
563 "%s: "DFID", oi_count %d\n",
564 osd_name(osd), PFID(fid), osd->od_oi_count);
566 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
569 #ifdef CONFIG_PROC_FS
571 LPROC_OSD_READ_BYTES = 0,
572 LPROC_OSD_WRITE_BYTES = 1,
573 LPROC_OSD_GET_PAGE = 2,
574 LPROC_OSD_NO_PAGE = 3,
575 LPROC_OSD_CACHE_ACCESS = 4,
576 LPROC_OSD_CACHE_HIT = 5,
577 LPROC_OSD_CACHE_MISS = 6,
578 LPROC_OSD_COPY_IO = 7,
579 LPROC_OSD_ZEROCOPY_IO = 8,
580 LPROC_OSD_TAIL_IO = 9,
584 extern struct kmem_cache *osd_zapit_cachep;
586 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
588 int osd_procfs_init(struct osd_device *osd, const char *name);
589 int osd_procfs_fini(struct osd_device *osd);
592 extern char *osd_obj_tag;
593 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
594 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
595 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
596 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
597 uint64_t oid, dmu_buf_t **dbp);
598 struct lu_object *osd_object_alloc(const struct lu_env *env,
599 const struct lu_object_header *hdr,
600 struct lu_device *d);
601 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
602 void *buf, uint32_t buflen, struct osd_thandle *oh);
603 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
604 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
605 unsigned dnsize, zap_flags_t flags);
606 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
607 struct osd_object *obj, const struct lu_fid *fid,
608 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
609 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
610 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
611 struct lu_attr *la, uint64_t parent, nvlist_t *);
612 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
613 uint64_t oid, dnode_t **dnp);
616 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
617 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
618 int osd_fid_lookup(const struct lu_env *env,
619 struct osd_device *, const struct lu_fid *, uint64_t *);
620 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
621 const struct lu_fid *fid, char *buf, int bufsize,
623 int osd_options_init(void);
624 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
626 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
627 struct osd_object *obj);
628 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
629 struct osd_device *osd,
630 const struct lu_fid *fid);
631 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
632 struct osd_device *osd,
633 const struct lu_fid *fid);
634 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
635 struct osd_device *osd,
636 const struct lu_fid *fid,
638 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
639 const struct lu_fid *fid);
640 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
641 uint64_t parent, const char *name, uint64_t *child,
642 const struct lu_fid *fid, bool isdir);
644 extern unsigned int osd_oi_count;
647 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
648 uint64_t oid, struct lu_fid *fid);
649 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
650 const struct dt_index_features *feat);
651 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
652 u64 seq, struct lu_seq_range *range);
653 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
654 uint64_t id, uint64_t dirhash);
655 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
656 uint64_t id, uint64_t dirhash);
657 void osd_zap_cursor_fini(zap_cursor_t *zc);
658 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
659 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
660 const struct lu_fid *fid);
661 int osd_add_to_remote_parent(const struct lu_env *env,
662 struct osd_device *osd,
663 struct osd_object *obj,
664 struct osd_thandle *oh);
665 int osd_delete_from_remote_parent(const struct lu_env *env,
666 struct osd_device *osd,
667 struct osd_object *obj,
668 struct osd_thandle *oh, bool destroy);
669 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
673 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
674 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
675 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
677 void osd_scrub_stop(struct osd_device *dev);
678 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
679 const struct lu_fid *fid, uint64_t oid, bool insert);
680 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
684 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
685 struct osd_object *obj,
686 struct osd_thandle *oh);
687 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
688 struct osd_thandle *oh);
689 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
690 struct osd_thandle *oh);
691 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
693 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
694 uint64_t xattr, struct lu_buf *buf,
695 const char *name, int *sizep);
696 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
697 struct lu_buf *buf, const char *name, int *sizep);
698 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
699 struct lu_buf *buf, const char *name);
700 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
701 const struct lu_buf *buf, const char *name,
702 int fl, struct thandle *handle);
703 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
704 const struct lu_buf *buf, const char *name, int fl,
705 struct thandle *handle);
706 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
707 const char *name, struct thandle *handle);
708 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
709 const char *name, struct thandle *handle);
710 void osd_declare_xattrs_destroy(const struct lu_env *env,
711 struct osd_object *obj,
712 struct osd_thandle *oh);
713 int osd_xattrs_destroy(const struct lu_env *env,
714 struct osd_object *obj, struct osd_thandle *oh);
715 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
716 const struct lu_buf *lb);
717 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
718 int vallen, const char *name, struct osd_thandle *oh);
719 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
720 const struct lu_buf *buf, const char *name, int fl,
721 struct osd_thandle *oh);;
722 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
723 const struct lu_buf *buf, const char *name, int fl,
724 struct osd_thandle *oh);
725 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
726 struct osd_thandle *oh);
728 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
729 sizeof(__u64) /* VBR VERSION */ + \
730 sizeof(struct lustre_mdt_attrs) /* LMA */)
732 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
733 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
736 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
738 return DN_MAX_BONUSLEN;
742 static inline int osd_object_is_zap(dnode_t *dn)
744 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
745 dn->dn_type == DMU_OT_USERGROUP_USED);
748 /* XXX: f_ver is not counted, but may differ too */
749 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
751 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
755 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
756 const struct lu_buf *buf, const char *name, int fl,
757 struct osd_thandle *oh)
761 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
765 if (osd_obj2dev(obj)->od_xattr_in_sa) {
766 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
768 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
770 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
776 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
778 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
779 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
780 #ifdef ZFS_PROJINHERIT
781 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
783 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
786 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
788 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
789 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
790 #ifdef ZFS_PROJINHERIT
791 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
793 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
798 #ifndef HAVE_DSL_POOL_CONFIG
799 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
803 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
808 #ifdef HAVE_SPA_MAXBLOCKSIZE
809 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
810 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
812 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
813 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
814 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
817 #ifdef HAVE_SA_SPILL_ALLOC
819 osd_zio_buf_alloc(size_t size)
821 return sa_spill_alloc(KM_SLEEP);
825 osd_zio_buf_free(void *buf, size_t size)
830 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
831 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
834 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
835 static inline uint64_t
836 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
837 int dnodesize, dmu_tx_t *tx)
840 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
842 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
843 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
846 static inline uint64_t
847 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
848 dmu_object_type_t ot, int leaf_blockshift,
849 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
852 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
854 return zap_create_flags_dnsize(os, normflags, flags, ot,
855 leaf_blockshift, indirect_blockshift,
856 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
861 osd_obj_bonuslen(struct osd_object *obj)
863 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
865 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
866 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
868 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
872 dnodesize = dmu_objset_dnodesize(os);
874 bonuslen = DN_BONUS_SIZE(dnodesize);
881 static inline uint64_t
882 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
883 int dnodesize, dmu_tx_t *tx)
885 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
886 DN_MAX_BONUSLEN, tx);
889 static inline uint64_t
890 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
891 dmu_object_type_t ot, int leaf_blockshift,
892 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
894 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
895 indirect_blockshift, DMU_OT_SA,
896 DN_MAX_BONUSLEN, tx);
900 osd_obj_bonuslen(struct osd_object *obj)
902 return DN_MAX_BONUSLEN;
904 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
906 #ifdef HAVE_DMU_PREFETCH_6ARG
907 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
908 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
910 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
911 dmu_prefetch((os), (obj), (lvl), (off))
914 static inline int osd_sa_handle_get(struct osd_object *obj)
916 struct osd_device *osd = osd_obj2dev(obj);
917 dnode_t *dn = obj->oo_dn;
923 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
924 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
925 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
928 refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
932 static inline void osd_dnode_rele(dnode_t *dn)
936 LASSERT(dn->dn_bonus);
940 dmu_buf_rele(&db->db, osd_obj_tag);
943 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
945 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
946 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
948 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
950 return dmu_objset_userobjspace_present(osd->od_os);
954 #define OSD_DMU_USEROBJ_PREFIX "obj-"
955 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
957 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
961 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
963 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
964 dnode_t *dn, const char *key,
965 int int_size, int int_num,
966 const void *val, dmu_tx_t *tx)
970 #ifdef HAVE_ZAP_ADD_BY_DNODE
972 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
974 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
977 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
978 dnode_t *dn, const char *key,
983 #ifdef HAVE_ZAP_ADD_BY_DNODE
985 return -zap_remove_by_dnode(dn, key, tx);
987 return -zap_remove(osd->od_os, zap, key, tx);
991 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
992 dnode_t *dn, const char *key,
993 int int_size, int int_num, void *v)
997 #ifdef HAVE_ZAP_ADD_BY_DNODE
999 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
1001 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
1004 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
1005 dnode_t *dn, int add, const char *name)
1007 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1009 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
1013 dmu_tx_hold_zap(tx, zap, add, name);
1016 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
1017 dnode_t *dn, uint64_t off, int len)
1019 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1021 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1025 dmu_tx_hold_write(tx, oid, off, len);
1028 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1029 uint64_t offset, uint64_t size,
1030 const char *buf, dmu_tx_t *tx)
1033 #ifdef HAVE_DMU_WRITE_BY_DNODE
1034 dmu_write_by_dnode(dn, offset, size, buf, tx);
1036 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1040 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1041 uint64_t offset, uint64_t size,
1042 char *buf, int flags)
1045 #ifdef HAVE_DMU_READ_BY_DNODE
1046 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1048 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1052 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1053 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1054 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1056 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1057 dmu_objset_own((name), (type), (ronly), (tag), (os))
1060 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1061 #define osd_dmu_objset_disown(os, decrypt, tag) \
1062 dmu_objset_disown((os), (decrypt), (tag))
1064 #define osd_dmu_objset_disown(os, decrypt, tag) \
1065 dmu_objset_disown((os), (tag))
1069 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1070 __u32 keysize, __u32 recsize)
1072 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1073 &osd->od_index_backup_list, &osd->od_lock,
1074 &osd->od_index_backup_stop,
1075 fid, keysize, recsize);
1079 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1081 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1084 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1085 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1086 osd->od_index_backup_id);
1090 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1091 &osd->od_index_backup_list, &osd->od_lock,
1092 &osd->od_index_backup_stop, backup);
1095 #ifndef HAVE_DMU_TX_MARK_NETFREE
1096 #define dmu_tx_mark_netfree(tx)
1099 #ifndef HAVE_ZFS_INODE_TIMESPEC
1100 #define inode_timespec_t timestruc_t
1103 #endif /* _OSD_INTERNAL_H */