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 #ifndef HAVE_ZFS_REFCOUNT_ADD
103 #define zfs_refcount_add refcount_add
106 extern struct dt_body_operations osd_body_scrub_ops;
109 * Iterator's in-memory data structure for quota file.
111 struct osd_it_quota {
112 struct osd_object *oiq_obj;
113 /* DMU accounting object id */
116 zap_cursor_t *oiq_zc;
117 /** identifier for current quota record */
119 unsigned oiq_reset:1; /* 1 -- no need to advance */
124 OZI_POS_DOT = 1, /* cursor at . */
125 OZI_POS_DOTDOT = 2, /* cursor at .. */
126 OZI_POS_REAL = 3, /* cursor at real entries */
130 * Iterator's in-memory data structure for ZAPs
132 * ZFS does not store . and .. on a disk, instead they are
133 * generated up on request
134 * to follow this format we do the same
137 zap_cursor_t *ozi_zc;
138 struct osd_object *ozi_obj;
139 unsigned ozi_reset:1; /* 1 -- no need to advance */
140 /* ozi_pos - position of the cursor */
141 enum osd_zap_pos ozi_pos;
143 char ozi_name[MAXNAMELEN]; /* file name for dir */
144 __u64 ozi_key; /* binary key for index files */
147 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
150 * regular ZFS direntry
152 struct zpl_direntry {
153 uint64_t zde_dnode:48,
156 } __attribute__((packed));
159 * lustre direntry adds a fid to regular ZFS direntry
161 struct luz_direntry {
162 struct zpl_direntry lzd_reg;
163 struct lu_fid lzd_fid;
164 } __attribute__((packed));
167 /* cached SA attributes */
172 #ifdef ZFS_PROJINHERIT
185 #define OSD_INS_CACHE_SIZE 8
188 struct osd_idmap_cache {
189 struct osd_device *oic_dev;
190 struct lu_fid oic_fid;
191 /** max 2^48 dnodes per dataset, avoid spilling into another word */
192 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
193 oic_remote:1; /* FID isn't local */
196 struct osd_inconsistent_item {
197 /* link into lustre_scrub::os_inconsistent_items,
198 * protected by lustr_scrub::os_lock. */
199 struct list_head oii_list;
201 /* The right FID <=> oid mapping. */
202 struct osd_idmap_cache oii_cache;
204 unsigned int oii_insert:1; /* insert or update mapping. */
207 struct osd_otable_it {
208 struct osd_device *ooi_dev;
209 struct lu_fid ooi_fid;
211 __u64 ooi_prefetched_dnode;
214 /* The following bits can be updated/checked w/o lock protection.
215 * If more bits will be introduced in the future and need lock to
216 * protect, please add comment. */
217 unsigned int ooi_used_outside:1, /* Some user out of OSD
218 * uses the iteration. */
219 ooi_all_cached:1, /* No more entries can be
220 * filled into cache. */
221 ooi_user_ready:1, /* The user out of OSD is
222 * ready to iterate. */
223 ooi_waiting:1; /* it::next is waiting. */
226 extern const struct dt_index_operations osd_otable_ops;
228 /* max.number of regular attributes the callers may ask for */
229 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
231 struct osd_thread_info {
232 const struct lu_env *oti_env;
234 struct lu_fid oti_fid;
236 struct ost_id oti_ostid;
242 char oti_key[MAXNAMELEN + 1];
243 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
244 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
246 struct lustre_mdt_attrs oti_mdt_attrs;
247 unsigned int oti_in_trans:1;
249 struct lu_attr oti_la;
250 struct osa_attr oti_osa;
251 zap_attribute_t oti_za;
252 zap_attribute_t oti_za2;
253 dmu_object_info_t oti_doi;
254 struct luz_direntry oti_zde;
256 struct lquota_id_info oti_qi;
257 struct lu_seq_range oti_seq_range;
259 /* dedicated OI cache for insert (which needs inum) */
260 struct osd_idmap_cache *oti_ins_cache;
261 int oti_ins_cache_size;
262 int oti_ins_cache_used;
263 /* inc by osd_trans_create and dec by osd_trans_stop */
264 int oti_ins_cache_depth;
265 struct lu_buf oti_xattr_lbuf;
267 zap_cursor_t oti_zc2;
270 extern struct lu_context_key osd_key;
272 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
274 return lu_context_key_get(&env->le_ctx, &osd_key);
278 struct thandle ot_super;
279 struct list_head ot_dcb_list;
280 struct list_head ot_stop_dcb_list;
281 struct list_head ot_unlinked_list;
282 struct list_head ot_sa_list;
284 struct lquota_trans ot_quota_trans;
288 #define OSD_OI_NAME_SIZE 24
291 * Object Index (OI) instance.
294 char oi_name[OSD_OI_NAME_SIZE];
301 uint64_t *os_compat_dirs;
302 int os_subdir_count; /* subdir count for each seq */
303 u64 os_seq; /* seq number */
304 struct list_head os_seq_list; /* list to seq_list */
307 struct osd_seq_list {
308 rwlock_t osl_seq_list_lock; /* lock for seq_list */
309 struct list_head osl_seq_list; /* list head for seq */
310 struct semaphore osl_seq_init_sem;
313 #define OSD_OST_MAP_SIZE 32
320 struct dt_device od_dt_dev;
321 /* information about underlying file system */
322 struct objset *od_os;
323 uint64_t od_rootid; /* id of root znode */
324 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
325 /* SA attr mapping->id,
326 * name is the same as in ZFS to use defines SA_ZPL_...*/
327 sa_attr_type_t *z_attr_table;
329 struct proc_dir_entry *od_proc_entry;
330 struct lprocfs_stats *od_stats;
332 uint64_t od_remote_parent_dir;
333 uint64_t od_index_backup_id;
334 uint64_t od_max_blksz;
337 struct osd_oi **od_oi_table;
338 unsigned int od_oi_count;
339 struct osd_seq_list od_seq_list;
341 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
342 od_prop_rdonly:1, /**< ZFS property readonly */
348 unsigned int od_dnsize;
349 int od_index_backup_stop;
351 enum lustre_index_backup_policy od_index_backup_policy;
358 __s64 od_auto_scrub_interval;
359 struct lu_site od_site;
361 dnode_t *od_groupused_dn;
362 dnode_t *od_userused_dn;
363 #ifdef ZFS_PROJINHERIT
364 dnode_t *od_projectused_dn;
367 /* quota slave instance for inode */
368 struct qsd_instance *od_quota_slave_md;
370 /* quota slave instance for block */
371 struct qsd_instance *od_quota_slave_dt;
373 struct brw_stats od_brw_stats;
374 atomic_t od_r_in_flight;
375 atomic_t od_w_in_flight;
377 /* used to debug zerocopy logic: the fields track all
378 * allocated, loaned and referenced buffers in use.
379 * to be removed once the change is tested well. */
380 atomic_t od_zerocopy_alloc;
381 atomic_t od_zerocopy_loan;
382 atomic_t od_zerocopy_pin;
384 arc_prune_t *arc_prune_cb;
386 /* osd seq instance */
387 struct lu_client_seq *od_cl_seq;
389 struct semaphore od_otable_sem;
390 struct osd_otable_it *od_otable_it;
391 struct lustre_scrub od_scrub;
392 struct list_head od_ios_list;
393 struct list_head od_index_backup_list;
394 struct list_head od_index_restore_list;
396 unsigned long long od_readcache_max_filesize;
399 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
402 return osd->od_quota_slave_dt;
404 return osd->od_quota_slave_md;
407 enum osd_destroy_type {
408 OSD_DESTROY_NONE = 0,
409 OSD_DESTROY_SYNC = 1,
410 OSD_DESTROY_ASYNC = 2,
414 struct dt_object oo_dt;
416 * Inode for file system object represented by this osd_object. This
417 * inode is pinned for the whole duration of lu_object life.
419 * Not modified concurrently (either setup early during object
420 * creation, or assigned by osd_create() under write lock).
423 sa_handle_t *oo_sa_hdl;
424 nvlist_t *oo_sa_xattr;
425 struct list_head oo_sa_linkage;
427 /* used to implement osd_object_*_{lock|unlock} */
428 struct rw_semaphore oo_sem;
430 /* to serialize some updates: destroy vs. others,
431 * xattr_set, object block size change etc */
432 struct rw_semaphore oo_guard;
434 /* protected by oo_guard */
435 struct list_head oo_unlinked_linkage;
437 /* cached attributes */
438 rwlock_t oo_attr_lock;
439 struct lu_attr oo_attr;
441 /* external dnode holding large EAs, protected by oo_guard */
443 enum osd_destroy_type oo_destroy;
445 __u32 oo_destroyed:1,
447 #ifdef ZFS_PROJINHERIT
453 /* the i_flags in LMA */
456 int oo_ea_in_bonus; /* EA bytes we expect */
458 /* record size for index file */
459 unsigned char oo_keysize;
460 unsigned char oo_recsize;
461 unsigned char oo_recusize; /* unit size */
463 uint64_t oo_parent; /* used only at object creation */
465 struct lu_object_header *oo_header;
468 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
469 extern const struct dt_index_operations osd_acct_index_ops;
470 extern struct lu_device_operations osd_lu_ops;
471 extern struct dt_index_operations osd_dir_ops;
472 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
473 qid_t uid, qid_t gid, qid_t projid, long long space,
474 struct osd_thandle *oh, int *flags,
475 enum osd_qid_declare_flags osd_qid_declare_flags);
476 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
477 uint64_t nrblocks, uint64_t est_maxblockshift);
478 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
484 static inline int lu_device_is_osd(const struct lu_device *d)
486 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
489 static inline struct osd_object *osd_obj(const struct lu_object *o)
491 LASSERT(lu_device_is_osd(o->lo_dev));
492 return container_of0(o, struct osd_object, oo_dt.do_lu);
495 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
497 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
498 return container_of0(d, struct osd_device, od_dt_dev);
501 static inline struct osd_device *osd_dev(const struct lu_device *d)
503 LASSERT(lu_device_is_osd(d));
504 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
507 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
509 return osd_obj(&d->do_lu);
512 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
514 return osd_dev(o->oo_dt.do_lu.lo_dev);
517 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
519 return &osd->od_dt_dev.dd_lu_dev;
522 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
524 return osd_dev(o->do_lu.lo_dev)->od_os;
527 static inline int osd_invariant(const struct osd_object *obj)
533 * Put the osd object once done with it.
535 * \param obj osd object that needs to be put
537 static inline void osd_object_put(const struct lu_env *env,
538 struct osd_object *obj)
540 dt_object_put(env, &obj->oo_dt);
543 static inline int osd_object_invariant(const struct lu_object *l)
545 return osd_invariant(osd_obj(l));
548 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
550 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
553 static inline char *osd_name(struct osd_device *osd)
555 return osd->od_svname;
558 static inline void zfs_set_bit(int nr, __u8 *addr)
560 set_bit(nr, (unsigned long *)addr);
563 static inline int zfs_test_bit(int nr, __u8 *addr)
565 return test_bit(nr, (const unsigned long *)addr);
568 static inline int osd_oi_fid2idx(struct osd_device *dev,
569 const struct lu_fid *fid)
571 return fid->f_seq & (dev->od_oi_count - 1);
574 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
575 const struct lu_fid *fid)
577 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
578 "%s: "DFID", oi_count %d\n",
579 osd_name(osd), PFID(fid), osd->od_oi_count);
581 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
584 #ifdef CONFIG_PROC_FS
586 LPROC_OSD_READ_BYTES = 0,
587 LPROC_OSD_WRITE_BYTES = 1,
588 LPROC_OSD_GET_PAGE = 2,
589 LPROC_OSD_NO_PAGE = 3,
590 LPROC_OSD_CACHE_ACCESS = 4,
591 LPROC_OSD_CACHE_HIT = 5,
592 LPROC_OSD_CACHE_MISS = 6,
593 LPROC_OSD_COPY_IO = 7,
594 LPROC_OSD_ZEROCOPY_IO = 8,
595 LPROC_OSD_TAIL_IO = 9,
599 extern struct kmem_cache *osd_zapit_cachep;
601 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
603 int osd_procfs_init(struct osd_device *osd, const char *name);
604 int osd_procfs_fini(struct osd_device *osd);
607 extern char *osd_obj_tag;
608 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
609 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
610 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
611 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
612 uint64_t oid, dmu_buf_t **dbp);
613 struct lu_object *osd_object_alloc(const struct lu_env *env,
614 const struct lu_object_header *hdr,
615 struct lu_device *d);
616 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
617 void *buf, uint32_t buflen, struct osd_thandle *oh);
618 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
619 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
620 unsigned dnsize, zap_flags_t flags);
621 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
622 struct osd_object *obj, const struct lu_fid *fid,
623 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
624 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
625 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
626 struct lu_attr *la, uint64_t parent, nvlist_t *);
627 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
628 uint64_t oid, dnode_t **dnp);
631 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
632 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
633 int osd_fid_lookup(const struct lu_env *env,
634 struct osd_device *, const struct lu_fid *, uint64_t *);
635 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
636 const struct lu_fid *fid, char *buf, int bufsize,
638 int osd_options_init(void);
639 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
641 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
642 struct osd_object *obj);
643 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
644 struct osd_device *osd,
645 const struct lu_fid *fid);
646 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
647 struct osd_device *osd,
648 const struct lu_fid *fid);
649 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
650 struct osd_device *osd,
651 const struct lu_fid *fid,
653 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
654 const struct lu_fid *fid);
655 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
656 uint64_t parent, const char *name, uint64_t *child,
657 const struct lu_fid *fid, bool isdir);
659 extern unsigned int osd_oi_count;
662 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
663 uint64_t oid, struct lu_fid *fid);
664 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
665 const struct dt_index_features *feat);
666 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
667 u64 seq, struct lu_seq_range *range);
668 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
669 uint64_t id, uint64_t dirhash);
670 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
671 uint64_t id, uint64_t dirhash);
672 void osd_zap_cursor_fini(zap_cursor_t *zc);
673 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
674 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
675 const struct lu_fid *fid);
676 int osd_add_to_remote_parent(const struct lu_env *env,
677 struct osd_device *osd,
678 struct osd_object *obj,
679 struct osd_thandle *oh);
680 int osd_delete_from_remote_parent(const struct lu_env *env,
681 struct osd_device *osd,
682 struct osd_object *obj,
683 struct osd_thandle *oh, bool destroy);
684 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
688 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
689 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
690 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
692 void osd_scrub_stop(struct osd_device *dev);
693 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
694 const struct lu_fid *fid, uint64_t oid, bool insert);
695 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
699 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
700 struct osd_object *obj,
701 struct osd_thandle *oh);
702 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
703 struct osd_thandle *oh);
704 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
705 struct osd_thandle *oh);
706 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
708 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
709 uint64_t xattr, struct lu_buf *buf,
710 const char *name, int *sizep);
711 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
712 struct lu_buf *buf, const char *name, int *sizep);
713 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
714 struct lu_buf *buf, const char *name);
715 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
716 const struct lu_buf *buf, const char *name,
717 int fl, struct thandle *handle);
718 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
719 const struct lu_buf *buf, const char *name, int fl,
720 struct thandle *handle);
721 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
722 const char *name, struct thandle *handle);
723 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
724 const char *name, struct thandle *handle);
725 void osd_declare_xattrs_destroy(const struct lu_env *env,
726 struct osd_object *obj,
727 struct osd_thandle *oh);
728 int osd_xattrs_destroy(const struct lu_env *env,
729 struct osd_object *obj, struct osd_thandle *oh);
730 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
731 const struct lu_buf *lb);
732 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
733 int vallen, const char *name, struct osd_thandle *oh);
734 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
735 const struct lu_buf *buf, const char *name, int fl,
736 struct osd_thandle *oh);;
737 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
738 const struct lu_buf *buf, const char *name, int fl,
739 struct osd_thandle *oh);
740 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
741 struct osd_thandle *oh);
743 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
744 sizeof(__u64) /* VBR VERSION */ + \
745 sizeof(struct lustre_mdt_attrs) /* LMA */)
747 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
748 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
751 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
753 return DN_MAX_BONUSLEN;
757 static inline int osd_object_is_zap(dnode_t *dn)
759 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
760 dn->dn_type == DMU_OT_USERGROUP_USED);
763 /* XXX: f_ver is not counted, but may differ too */
764 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
766 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
770 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
771 const struct lu_buf *buf, const char *name, int fl,
772 struct osd_thandle *oh)
776 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
780 if (osd_obj2dev(obj)->od_xattr_in_sa) {
781 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
783 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
785 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
791 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
793 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
794 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
795 #ifdef ZFS_PROJINHERIT
796 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
798 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
801 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
803 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
804 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
805 #ifdef ZFS_PROJINHERIT
806 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
808 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
813 #ifndef HAVE_DSL_POOL_CONFIG
814 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
818 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
823 #ifdef HAVE_SPA_MAXBLOCKSIZE
824 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
825 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
827 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
828 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
829 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
832 #ifdef HAVE_SA_SPILL_ALLOC
834 osd_zio_buf_alloc(size_t size)
836 return sa_spill_alloc(KM_SLEEP);
840 osd_zio_buf_free(void *buf, size_t size)
845 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
846 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
849 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
850 static inline uint64_t
851 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
852 int dnodesize, dmu_tx_t *tx)
855 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
857 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
858 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
861 static inline uint64_t
862 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
863 dmu_object_type_t ot, int leaf_blockshift,
864 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
867 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
869 return zap_create_flags_dnsize(os, normflags, flags, ot,
870 leaf_blockshift, indirect_blockshift,
871 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
876 osd_obj_bonuslen(struct osd_object *obj)
878 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
880 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
881 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
883 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
887 dnodesize = dmu_objset_dnodesize(os);
889 bonuslen = DN_BONUS_SIZE(dnodesize);
896 static inline uint64_t
897 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
898 int dnodesize, dmu_tx_t *tx)
900 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
901 DN_MAX_BONUSLEN, tx);
904 static inline uint64_t
905 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
906 dmu_object_type_t ot, int leaf_blockshift,
907 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
909 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
910 indirect_blockshift, DMU_OT_SA,
911 DN_MAX_BONUSLEN, tx);
915 osd_obj_bonuslen(struct osd_object *obj)
917 return DN_MAX_BONUSLEN;
919 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
921 #ifdef HAVE_DMU_PREFETCH_6ARG
922 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
923 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
925 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
926 dmu_prefetch((os), (obj), (lvl), (off))
929 static inline int osd_sa_handle_get(struct osd_object *obj)
931 struct osd_device *osd = osd_obj2dev(obj);
932 dnode_t *dn = obj->oo_dn;
938 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
939 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
940 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
943 zfs_refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
947 static inline void osd_dnode_rele(dnode_t *dn)
951 LASSERT(dn->dn_bonus);
955 dmu_buf_rele(&db->db, osd_obj_tag);
958 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
960 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
961 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
963 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
965 return dmu_objset_userobjspace_present(osd->od_os);
969 #define OSD_DMU_USEROBJ_PREFIX "obj-"
970 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
972 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
976 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
978 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
979 dnode_t *dn, const char *key,
980 int int_size, int int_num,
981 const void *val, dmu_tx_t *tx)
985 #ifdef HAVE_ZAP_ADD_BY_DNODE
987 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
989 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
992 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
993 dnode_t *dn, const char *key,
998 #ifdef HAVE_ZAP_ADD_BY_DNODE
1000 return -zap_remove_by_dnode(dn, key, tx);
1002 return -zap_remove(osd->od_os, zap, key, tx);
1006 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
1007 dnode_t *dn, const char *key,
1008 int int_size, int int_num, void *v)
1012 #ifdef HAVE_ZAP_ADD_BY_DNODE
1014 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
1016 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
1019 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
1020 dnode_t *dn, int add, const char *name)
1022 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1024 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
1028 dmu_tx_hold_zap(tx, zap, add, name);
1031 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
1032 dnode_t *dn, uint64_t off, int len)
1034 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1036 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1040 dmu_tx_hold_write(tx, oid, off, len);
1043 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1044 uint64_t offset, uint64_t size,
1045 const char *buf, dmu_tx_t *tx)
1048 #ifdef HAVE_DMU_WRITE_BY_DNODE
1049 dmu_write_by_dnode(dn, offset, size, buf, tx);
1051 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1055 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1056 uint64_t offset, uint64_t size,
1057 char *buf, int flags)
1060 #ifdef HAVE_DMU_READ_BY_DNODE
1061 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1063 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1067 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1068 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1069 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1071 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1072 dmu_objset_own((name), (type), (ronly), (tag), (os))
1075 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1076 #define osd_dmu_objset_disown(os, decrypt, tag) \
1077 dmu_objset_disown((os), (decrypt), (tag))
1079 #define osd_dmu_objset_disown(os, decrypt, tag) \
1080 dmu_objset_disown((os), (tag))
1084 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1085 __u32 keysize, __u32 recsize)
1087 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1088 &osd->od_index_backup_list, &osd->od_lock,
1089 &osd->od_index_backup_stop,
1090 fid, keysize, recsize);
1094 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1096 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1099 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1100 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1101 osd->od_index_backup_id);
1105 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1106 &osd->od_index_backup_list, &osd->od_lock,
1107 &osd->od_index_backup_stop, backup);
1110 #ifndef HAVE_DMU_TX_MARK_NETFREE
1111 #define dmu_tx_mark_netfree(tx)
1114 #ifndef HAVE_ZFS_INODE_TIMESPEC
1115 #define inode_timespec_t timestruc_t
1118 #endif /* _OSD_INTERNAL_H */