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;
282 __u32 ot_write_commit:1,
286 #define OSD_OI_NAME_SIZE 24
289 * Object Index (OI) instance.
292 char oi_name[OSD_OI_NAME_SIZE];
299 uint64_t *os_compat_dirs;
300 int os_subdir_count; /* subdir count for each seq */
301 u64 os_seq; /* seq number */
302 struct list_head os_seq_list; /* list to seq_list */
305 struct osd_seq_list {
306 rwlock_t osl_seq_list_lock; /* lock for seq_list */
307 struct list_head osl_seq_list; /* list head for seq */
308 struct semaphore osl_seq_init_sem;
311 #define OSD_OST_MAP_SIZE 32
318 struct dt_device od_dt_dev;
319 /* information about underlying file system */
320 struct objset *od_os;
321 uint64_t od_rootid; /* id of root znode */
322 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
323 /* SA attr mapping->id,
324 * name is the same as in ZFS to use defines SA_ZPL_...*/
325 sa_attr_type_t *z_attr_table;
327 struct proc_dir_entry *od_proc_entry;
328 struct lprocfs_stats *od_stats;
330 uint64_t od_remote_parent_dir;
331 uint64_t od_index_backup_id;
332 uint64_t od_max_blksz;
335 struct osd_oi **od_oi_table;
336 unsigned int od_oi_count;
337 struct osd_seq_list od_seq_list;
339 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
340 od_prop_rdonly:1, /**< ZFS property readonly */
345 unsigned int od_dnsize;
346 int od_index_backup_stop;
348 enum lustre_index_backup_policy od_index_backup_policy;
355 __s64 od_auto_scrub_interval;
356 struct lu_site od_site;
358 dnode_t *od_groupused_dn;
359 dnode_t *od_userused_dn;
360 #ifdef ZFS_PROJINHERIT
361 dnode_t *od_projectused_dn;
364 /* quota slave instance */
365 struct qsd_instance *od_quota_slave;
367 struct brw_stats od_brw_stats;
368 atomic_t od_r_in_flight;
369 atomic_t od_w_in_flight;
371 /* used to debug zerocopy logic: the fields track all
372 * allocated, loaned and referenced buffers in use.
373 * to be removed once the change is tested well. */
374 atomic_t od_zerocopy_alloc;
375 atomic_t od_zerocopy_loan;
376 atomic_t od_zerocopy_pin;
378 arc_prune_t *arc_prune_cb;
380 /* osd seq instance */
381 struct lu_client_seq *od_cl_seq;
383 struct semaphore od_otable_sem;
384 struct osd_otable_it *od_otable_it;
385 struct lustre_scrub od_scrub;
386 struct list_head od_ios_list;
387 struct list_head od_index_backup_list;
388 struct list_head od_index_restore_list;
390 unsigned long long od_readcache_max_filesize;
393 enum osd_destroy_type {
394 OSD_DESTROY_NONE = 0,
395 OSD_DESTROY_SYNC = 1,
396 OSD_DESTROY_ASYNC = 2,
400 struct dt_object oo_dt;
402 * Inode for file system object represented by this osd_object. This
403 * inode is pinned for the whole duration of lu_object life.
405 * Not modified concurrently (either setup early during object
406 * creation, or assigned by osd_create() under write lock).
409 sa_handle_t *oo_sa_hdl;
410 nvlist_t *oo_sa_xattr;
411 struct list_head oo_sa_linkage;
413 /* used to implement osd_object_*_{lock|unlock} */
414 struct rw_semaphore oo_sem;
416 /* to serialize some updates: destroy vs. others,
417 * xattr_set, object block size change etc */
418 struct rw_semaphore oo_guard;
420 /* protected by oo_guard */
421 struct list_head oo_unlinked_linkage;
423 /* cached attributes */
424 rwlock_t oo_attr_lock;
425 struct lu_attr oo_attr;
427 /* external dnode holding large EAs, protected by oo_guard */
429 enum osd_destroy_type oo_destroy;
431 __u32 oo_destroyed:1,
433 #ifdef ZFS_PROJINHERIT
439 /* the i_flags in LMA */
442 int oo_ea_in_bonus; /* EA bytes we expect */
444 /* record size for index file */
445 unsigned char oo_keysize;
446 unsigned char oo_recsize;
447 unsigned char oo_recusize; /* unit size */
449 uint64_t oo_parent; /* used only at object creation */
451 struct lu_object_header *oo_header;
454 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
455 extern const struct dt_index_operations osd_acct_index_ops;
456 extern struct lu_device_operations osd_lu_ops;
457 extern struct dt_index_operations osd_dir_ops;
458 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
459 qid_t uid, qid_t gid, qid_t projid, long long space,
460 struct osd_thandle *oh, int *flags,
461 enum osd_qid_declare_flags osd_qid_declare_flags);
462 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
463 uint64_t nrblocks, uint64_t est_maxblockshift);
464 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
470 static inline int lu_device_is_osd(const struct lu_device *d)
472 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
475 static inline struct osd_object *osd_obj(const struct lu_object *o)
477 LASSERT(lu_device_is_osd(o->lo_dev));
478 return container_of0(o, struct osd_object, oo_dt.do_lu);
481 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
483 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
484 return container_of0(d, struct osd_device, od_dt_dev);
487 static inline struct osd_device *osd_dev(const struct lu_device *d)
489 LASSERT(lu_device_is_osd(d));
490 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
493 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
495 return osd_obj(&d->do_lu);
498 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
500 return osd_dev(o->oo_dt.do_lu.lo_dev);
503 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
505 return &osd->od_dt_dev.dd_lu_dev;
508 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
510 return osd_dev(o->do_lu.lo_dev)->od_os;
513 static inline int osd_invariant(const struct osd_object *obj)
519 * Put the osd object once done with it.
521 * \param obj osd object that needs to be put
523 static inline void osd_object_put(const struct lu_env *env,
524 struct osd_object *obj)
526 dt_object_put(env, &obj->oo_dt);
529 static inline int osd_object_invariant(const struct lu_object *l)
531 return osd_invariant(osd_obj(l));
534 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
536 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
539 static inline char *osd_name(struct osd_device *osd)
541 return osd->od_svname;
544 static inline void zfs_set_bit(int nr, __u8 *addr)
546 set_bit(nr, (unsigned long *)addr);
549 static inline int zfs_test_bit(int nr, __u8 *addr)
551 return test_bit(nr, (const unsigned long *)addr);
554 static inline int osd_oi_fid2idx(struct osd_device *dev,
555 const struct lu_fid *fid)
557 return fid->f_seq & (dev->od_oi_count - 1);
560 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
561 const struct lu_fid *fid)
563 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
564 "%s: "DFID", oi_count %d\n",
565 osd_name(osd), PFID(fid), osd->od_oi_count);
567 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
570 #ifdef CONFIG_PROC_FS
572 LPROC_OSD_READ_BYTES = 0,
573 LPROC_OSD_WRITE_BYTES = 1,
574 LPROC_OSD_GET_PAGE = 2,
575 LPROC_OSD_NO_PAGE = 3,
576 LPROC_OSD_CACHE_ACCESS = 4,
577 LPROC_OSD_CACHE_HIT = 5,
578 LPROC_OSD_CACHE_MISS = 6,
579 LPROC_OSD_COPY_IO = 7,
580 LPROC_OSD_ZEROCOPY_IO = 8,
581 LPROC_OSD_TAIL_IO = 9,
585 extern struct kmem_cache *osd_zapit_cachep;
587 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
589 int osd_procfs_init(struct osd_device *osd, const char *name);
590 int osd_procfs_fini(struct osd_device *osd);
593 extern char *osd_obj_tag;
594 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
595 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
596 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
597 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
598 uint64_t oid, dmu_buf_t **dbp);
599 struct lu_object *osd_object_alloc(const struct lu_env *env,
600 const struct lu_object_header *hdr,
601 struct lu_device *d);
602 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
603 void *buf, uint32_t buflen, struct osd_thandle *oh);
604 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
605 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
606 unsigned dnsize, zap_flags_t flags);
607 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
608 struct osd_object *obj, const struct lu_fid *fid,
609 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
610 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
611 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
612 struct lu_attr *la, uint64_t parent, nvlist_t *);
613 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
614 uint64_t oid, dnode_t **dnp);
617 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
618 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
619 int osd_fid_lookup(const struct lu_env *env,
620 struct osd_device *, const struct lu_fid *, uint64_t *);
621 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
622 const struct lu_fid *fid, char *buf, int bufsize,
624 int osd_options_init(void);
625 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
627 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
628 struct osd_object *obj);
629 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
630 struct osd_device *osd,
631 const struct lu_fid *fid);
632 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
633 struct osd_device *osd,
634 const struct lu_fid *fid);
635 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
636 struct osd_device *osd,
637 const struct lu_fid *fid,
639 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
640 const struct lu_fid *fid);
641 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
642 uint64_t parent, const char *name, uint64_t *child,
643 const struct lu_fid *fid, bool isdir);
645 extern unsigned int osd_oi_count;
648 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
649 uint64_t oid, struct lu_fid *fid);
650 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
651 const struct dt_index_features *feat);
652 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
653 u64 seq, struct lu_seq_range *range);
654 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
655 uint64_t id, uint64_t dirhash);
656 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
657 uint64_t id, uint64_t dirhash);
658 void osd_zap_cursor_fini(zap_cursor_t *zc);
659 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
660 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
661 const struct lu_fid *fid);
662 int osd_add_to_remote_parent(const struct lu_env *env,
663 struct osd_device *osd,
664 struct osd_object *obj,
665 struct osd_thandle *oh);
666 int osd_delete_from_remote_parent(const struct lu_env *env,
667 struct osd_device *osd,
668 struct osd_object *obj,
669 struct osd_thandle *oh, bool destroy);
670 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
674 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
675 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
676 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
678 void osd_scrub_stop(struct osd_device *dev);
679 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
680 const struct lu_fid *fid, uint64_t oid, bool insert);
681 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
685 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
686 struct osd_object *obj,
687 struct osd_thandle *oh);
688 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
689 struct osd_thandle *oh);
690 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
691 struct osd_thandle *oh);
692 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
694 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
695 uint64_t xattr, struct lu_buf *buf,
696 const char *name, int *sizep);
697 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
698 struct lu_buf *buf, const char *name, int *sizep);
699 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
700 struct lu_buf *buf, const char *name);
701 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
702 const struct lu_buf *buf, const char *name,
703 int fl, struct thandle *handle);
704 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
705 const struct lu_buf *buf, const char *name, int fl,
706 struct thandle *handle);
707 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
708 const char *name, struct thandle *handle);
709 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
710 const char *name, struct thandle *handle);
711 void osd_declare_xattrs_destroy(const struct lu_env *env,
712 struct osd_object *obj,
713 struct osd_thandle *oh);
714 int osd_xattrs_destroy(const struct lu_env *env,
715 struct osd_object *obj, struct osd_thandle *oh);
716 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
717 const struct lu_buf *lb);
718 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
719 int vallen, const char *name, struct osd_thandle *oh);
720 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
721 const struct lu_buf *buf, const char *name, int fl,
722 struct osd_thandle *oh);;
723 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
724 const struct lu_buf *buf, const char *name, int fl,
725 struct osd_thandle *oh);
726 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
727 struct osd_thandle *oh);
729 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
730 sizeof(__u64) /* VBR VERSION */ + \
731 sizeof(struct lustre_mdt_attrs) /* LMA */)
733 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
734 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
737 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
739 return DN_MAX_BONUSLEN;
743 static inline int osd_object_is_zap(dnode_t *dn)
745 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
746 dn->dn_type == DMU_OT_USERGROUP_USED);
749 /* XXX: f_ver is not counted, but may differ too */
750 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
752 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
756 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
757 const struct lu_buf *buf, const char *name, int fl,
758 struct osd_thandle *oh)
762 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
766 if (osd_obj2dev(obj)->od_xattr_in_sa) {
767 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
769 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
771 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
777 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
779 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
780 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
781 #ifdef ZFS_PROJINHERIT
782 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
784 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
787 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
789 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
790 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
791 #ifdef ZFS_PROJINHERIT
792 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
794 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
799 #ifndef HAVE_DSL_POOL_CONFIG
800 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
804 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
809 #ifdef HAVE_SPA_MAXBLOCKSIZE
810 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
811 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
813 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
814 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
815 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
818 #ifdef HAVE_SA_SPILL_ALLOC
820 osd_zio_buf_alloc(size_t size)
822 return sa_spill_alloc(KM_SLEEP);
826 osd_zio_buf_free(void *buf, size_t size)
831 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
832 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
835 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
836 static inline uint64_t
837 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
838 int dnodesize, dmu_tx_t *tx)
841 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
843 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
844 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
847 static inline uint64_t
848 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
849 dmu_object_type_t ot, int leaf_blockshift,
850 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
853 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
855 return zap_create_flags_dnsize(os, normflags, flags, ot,
856 leaf_blockshift, indirect_blockshift,
857 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
862 osd_obj_bonuslen(struct osd_object *obj)
864 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
866 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
867 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
869 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
873 dnodesize = dmu_objset_dnodesize(os);
875 bonuslen = DN_BONUS_SIZE(dnodesize);
882 static inline uint64_t
883 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
884 int dnodesize, dmu_tx_t *tx)
886 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
887 DN_MAX_BONUSLEN, 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)
895 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
896 indirect_blockshift, DMU_OT_SA,
897 DN_MAX_BONUSLEN, tx);
901 osd_obj_bonuslen(struct osd_object *obj)
903 return DN_MAX_BONUSLEN;
905 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
907 #ifdef HAVE_DMU_PREFETCH_6ARG
908 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
909 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
911 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
912 dmu_prefetch((os), (obj), (lvl), (off))
915 static inline int osd_sa_handle_get(struct osd_object *obj)
917 struct osd_device *osd = osd_obj2dev(obj);
918 dnode_t *dn = obj->oo_dn;
924 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
925 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
926 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
929 refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
933 static inline void osd_dnode_rele(dnode_t *dn)
937 LASSERT(dn->dn_bonus);
941 dmu_buf_rele(&db->db, osd_obj_tag);
944 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
946 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
947 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
949 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
951 return dmu_objset_userobjspace_present(osd->od_os);
955 #define OSD_DMU_USEROBJ_PREFIX "obj-"
956 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
958 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
962 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
964 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
965 dnode_t *dn, const char *key,
966 int int_size, int int_num,
967 const void *val, dmu_tx_t *tx)
971 #ifdef HAVE_ZAP_ADD_BY_DNODE
973 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
975 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
978 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
979 dnode_t *dn, const char *key,
984 #ifdef HAVE_ZAP_ADD_BY_DNODE
986 return -zap_remove_by_dnode(dn, key, tx);
988 return -zap_remove(osd->od_os, zap, key, tx);
992 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
993 dnode_t *dn, const char *key,
994 int int_size, int int_num, void *v)
998 #ifdef HAVE_ZAP_ADD_BY_DNODE
1000 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
1002 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
1005 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
1006 dnode_t *dn, int add, const char *name)
1008 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1010 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
1014 dmu_tx_hold_zap(tx, zap, add, name);
1017 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
1018 dnode_t *dn, uint64_t off, int len)
1020 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1022 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1026 dmu_tx_hold_write(tx, oid, off, len);
1029 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1030 uint64_t offset, uint64_t size,
1031 const char *buf, dmu_tx_t *tx)
1034 #ifdef HAVE_DMU_WRITE_BY_DNODE
1035 dmu_write_by_dnode(dn, offset, size, buf, tx);
1037 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1041 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1042 uint64_t offset, uint64_t size,
1043 char *buf, int flags)
1046 #ifdef HAVE_DMU_READ_BY_DNODE
1047 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1049 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1053 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1054 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1055 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1057 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1058 dmu_objset_own((name), (type), (ronly), (tag), (os))
1061 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1062 #define osd_dmu_objset_disown(os, decrypt, tag) \
1063 dmu_objset_disown((os), (decrypt), (tag))
1065 #define osd_dmu_objset_disown(os, decrypt, tag) \
1066 dmu_objset_disown((os), (tag))
1070 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1071 __u32 keysize, __u32 recsize)
1073 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1074 &osd->od_index_backup_list, &osd->od_lock,
1075 &osd->od_index_backup_stop,
1076 fid, keysize, recsize);
1080 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1082 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1085 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1086 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1087 osd->od_index_backup_id);
1091 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1092 &osd->od_index_backup_list, &osd->od_lock,
1093 &osd->od_index_backup_stop, backup);
1096 #ifndef HAVE_DMU_TX_MARK_NETFREE
1097 #define dmu_tx_mark_netfree(tx)
1100 #endif /* _OSD_INTERNAL_H */