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, ... */
101 * Iterator's in-memory data structure for quota file.
103 struct osd_it_quota {
104 struct osd_object *oiq_obj;
105 /* DMU accounting object id */
108 zap_cursor_t *oiq_zc;
109 /** identifier for current quota record */
111 unsigned oiq_reset:1; /* 1 -- no need to advance */
115 * Iterator's in-memory data structure for ZAPs
117 * ZFS does not store . and .. on a disk, instead they are
118 * generated up on request
119 * to follow this format we do the same
122 zap_cursor_t *ozi_zc;
123 struct osd_object *ozi_obj;
124 unsigned ozi_reset:1; /* 1 -- no need to advance */
125 /* ozi_pos - position of the cursor:
126 * 0 - before any record
129 * 3 - real records */
132 char ozi_name[MAXNAMELEN]; /* file name for dir */
133 __u64 ozi_key; /* binary key for index files */
136 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
139 * regular ZFS direntry
141 struct zpl_direntry {
142 uint64_t zde_dnode:48,
145 } __attribute__((packed));
148 * lustre direntry adds a fid to regular ZFS direntry
150 struct luz_direntry {
151 struct zpl_direntry lzd_reg;
152 struct lu_fid lzd_fid;
153 } __attribute__((packed));
156 /* cached SA attributes */
161 #ifdef ZFS_PROJINHERIT
174 #define OSD_INS_CACHE_SIZE 8
177 struct osd_idmap_cache {
178 struct osd_device *oic_dev;
179 struct lu_fid oic_fid;
180 /** max 2^48 dnodes per dataset, avoid spilling into another word */
181 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
182 oic_remote:1; /* FID isn't local */
185 struct osd_inconsistent_item {
186 /* link into lustre_scrub::os_inconsistent_items,
187 * protected by lustr_scrub::os_lock. */
188 struct list_head oii_list;
190 /* The right FID <=> oid mapping. */
191 struct osd_idmap_cache oii_cache;
193 unsigned int oii_insert:1; /* insert or update mapping. */
196 struct osd_otable_it {
197 struct osd_device *ooi_dev;
198 struct lu_fid ooi_fid;
200 __u64 ooi_prefetched_dnode;
203 /* The following bits can be updated/checked w/o lock protection.
204 * If more bits will be introduced in the future and need lock to
205 * protect, please add comment. */
206 unsigned int ooi_used_outside:1, /* Some user out of OSD
207 * uses the iteration. */
208 ooi_all_cached:1, /* No more entries can be
209 * filled into cache. */
210 ooi_user_ready:1, /* The user out of OSD is
211 * ready to iterate. */
212 ooi_waiting:1; /* it::next is waiting. */
215 extern const struct dt_index_operations osd_otable_ops;
217 /* max.number of regular attributes the callers may ask for */
218 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
220 struct osd_thread_info {
221 const struct lu_env *oti_env;
223 struct lu_fid oti_fid;
225 * XXX temporary: for ->i_op calls.
227 struct timespec oti_time;
229 struct ost_id oti_ostid;
235 char oti_key[MAXNAMELEN + 1];
236 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
237 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
239 struct lustre_mdt_attrs oti_mdt_attrs;
240 unsigned int oti_in_trans:1;
242 struct lu_attr oti_la;
243 struct osa_attr oti_osa;
244 zap_attribute_t oti_za;
245 zap_attribute_t oti_za2;
246 dmu_object_info_t oti_doi;
247 struct luz_direntry oti_zde;
249 struct lquota_id_info oti_qi;
250 struct lu_seq_range oti_seq_range;
252 /* dedicated OI cache for insert (which needs inum) */
253 struct osd_idmap_cache *oti_ins_cache;
254 int oti_ins_cache_size;
255 int oti_ins_cache_used;
256 struct lu_buf oti_xattr_lbuf;
258 zap_cursor_t oti_zc2;
261 extern struct lu_context_key osd_key;
263 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
265 return lu_context_key_get(&env->le_ctx, &osd_key);
269 struct thandle ot_super;
270 struct list_head ot_dcb_list;
271 struct list_head ot_stop_dcb_list;
272 struct list_head ot_unlinked_list;
273 struct list_head ot_sa_list;
275 struct lquota_trans ot_quota_trans;
276 __u32 ot_write_commit:1,
280 #define OSD_OI_NAME_SIZE 24
283 * Object Index (OI) instance.
286 char oi_name[OSD_OI_NAME_SIZE];
293 uint64_t *os_compat_dirs;
294 int os_subdir_count; /* subdir count for each seq */
295 u64 os_seq; /* seq number */
296 struct list_head os_seq_list; /* list to seq_list */
299 struct osd_seq_list {
300 rwlock_t osl_seq_list_lock; /* lock for seq_list */
301 struct list_head osl_seq_list; /* list head for seq */
302 struct semaphore osl_seq_init_sem;
305 #define OSD_OST_MAP_SIZE 32
312 struct dt_device od_dt_dev;
313 /* information about underlying file system */
314 struct objset *od_os;
315 uint64_t od_rootid; /* id of root znode */
316 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
317 /* SA attr mapping->id,
318 * name is the same as in ZFS to use defines SA_ZPL_...*/
319 sa_attr_type_t *z_attr_table;
321 struct proc_dir_entry *od_proc_entry;
322 struct lprocfs_stats *od_stats;
324 uint64_t od_remote_parent_dir;
325 uint64_t od_index_backup_id;
326 uint64_t od_max_blksz;
329 struct osd_oi **od_oi_table;
330 unsigned int od_oi_count;
331 struct osd_seq_list od_seq_list;
333 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
334 od_prop_rdonly:1, /**< ZFS property readonly */
339 unsigned int od_dnsize;
340 int od_index_backup_stop;
342 enum lustre_index_backup_policy od_index_backup_policy;
349 __s64 od_auto_scrub_interval;
350 struct lu_site od_site;
352 dnode_t *od_groupused_dn;
353 dnode_t *od_userused_dn;
354 #ifdef ZFS_PROJINHERIT
355 dnode_t *od_projectused_dn;
358 /* quota slave instance */
359 struct qsd_instance *od_quota_slave;
361 struct brw_stats od_brw_stats;
362 atomic_t od_r_in_flight;
363 atomic_t od_w_in_flight;
365 /* used to debug zerocopy logic: the fields track all
366 * allocated, loaned and referenced buffers in use.
367 * to be removed once the change is tested well. */
368 atomic_t od_zerocopy_alloc;
369 atomic_t od_zerocopy_loan;
370 atomic_t od_zerocopy_pin;
372 arc_prune_t *arc_prune_cb;
374 /* osd seq instance */
375 struct lu_client_seq *od_cl_seq;
377 struct semaphore od_otable_sem;
378 struct osd_otable_it *od_otable_it;
379 struct lustre_scrub od_scrub;
380 struct list_head od_ios_list;
381 struct list_head od_index_backup_list;
382 struct list_head od_index_restore_list;
386 enum osd_destroy_type {
387 OSD_DESTROY_NONE = 0,
388 OSD_DESTROY_SYNC = 1,
389 OSD_DESTROY_ASYNC = 2,
393 struct dt_object oo_dt;
395 * Inode for file system object represented by this osd_object. This
396 * inode is pinned for the whole duration of lu_object life.
398 * Not modified concurrently (either setup early during object
399 * creation, or assigned by osd_create() under write lock).
402 sa_handle_t *oo_sa_hdl;
403 nvlist_t *oo_sa_xattr;
404 struct list_head oo_sa_linkage;
406 /* used to implement osd_object_*_{lock|unlock} */
407 struct rw_semaphore oo_sem;
409 /* to serialize some updates: destroy vs. others,
410 * xattr_set, object block size change etc */
411 struct rw_semaphore oo_guard;
413 /* protected by oo_guard */
414 struct list_head oo_unlinked_linkage;
416 /* cached attributes */
417 rwlock_t oo_attr_lock;
418 struct lu_attr oo_attr;
420 /* external dnode holding large EAs, protected by oo_guard */
422 enum osd_destroy_type oo_destroy;
424 __u32 oo_destroyed:1,
426 #ifdef ZFS_PROJINHERIT
432 /* the i_flags in LMA */
435 int oo_ea_in_bonus; /* EA bytes we expect */
437 /* record size for index file */
438 unsigned char oo_keysize;
439 unsigned char oo_recsize;
440 unsigned char oo_recusize; /* unit size */
442 uint64_t oo_parent; /* used only at object creation */
444 struct lu_object_header *oo_header;
447 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
448 extern const struct dt_index_operations osd_acct_index_ops;
449 extern struct lu_device_operations osd_lu_ops;
450 extern struct dt_index_operations osd_dir_ops;
451 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
452 qid_t uid, qid_t gid, qid_t projid, long long space,
453 struct osd_thandle *oh, int *flags,
454 enum osd_qid_declare_flags osd_qid_declare_flags);
455 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
456 uint64_t nrblocks, uint64_t est_maxblockshift);
457 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
463 static inline int lu_device_is_osd(const struct lu_device *d)
465 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
468 static inline struct osd_object *osd_obj(const struct lu_object *o)
470 LASSERT(lu_device_is_osd(o->lo_dev));
471 return container_of0(o, struct osd_object, oo_dt.do_lu);
474 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
476 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
477 return container_of0(d, struct osd_device, od_dt_dev);
480 static inline struct osd_device *osd_dev(const struct lu_device *d)
482 LASSERT(lu_device_is_osd(d));
483 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
486 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
488 return osd_obj(&d->do_lu);
491 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
493 return osd_dev(o->oo_dt.do_lu.lo_dev);
496 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
498 return &osd->od_dt_dev.dd_lu_dev;
501 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
503 return osd_dev(o->do_lu.lo_dev)->od_os;
506 static inline int osd_invariant(const struct osd_object *obj)
512 * Put the osd object once done with it.
514 * \param obj osd object that needs to be put
516 static inline void osd_object_put(const struct lu_env *env,
517 struct osd_object *obj)
519 dt_object_put(env, &obj->oo_dt);
522 static inline int osd_object_invariant(const struct lu_object *l)
524 return osd_invariant(osd_obj(l));
527 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
529 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
532 static inline char *osd_name(struct osd_device *osd)
534 return osd->od_svname;
537 static inline void zfs_set_bit(int nr, __u8 *addr)
539 set_bit(nr, (unsigned long *)addr);
542 static inline int zfs_test_bit(int nr, __u8 *addr)
544 return test_bit(nr, (const unsigned long *)addr);
547 static inline int osd_oi_fid2idx(struct osd_device *dev,
548 const struct lu_fid *fid)
550 return fid->f_seq & (dev->od_oi_count - 1);
553 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
554 const struct lu_fid *fid)
556 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
557 "%s: "DFID", oi_count %d\n",
558 osd_name(osd), PFID(fid), osd->od_oi_count);
560 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
563 #ifdef CONFIG_PROC_FS
565 LPROC_OSD_READ_BYTES = 0,
566 LPROC_OSD_WRITE_BYTES = 1,
567 LPROC_OSD_GET_PAGE = 2,
568 LPROC_OSD_NO_PAGE = 3,
569 LPROC_OSD_CACHE_ACCESS = 4,
570 LPROC_OSD_CACHE_HIT = 5,
571 LPROC_OSD_CACHE_MISS = 6,
572 LPROC_OSD_COPY_IO = 7,
573 LPROC_OSD_ZEROCOPY_IO = 8,
574 LPROC_OSD_TAIL_IO = 9,
578 extern struct kmem_cache *osd_zapit_cachep;
580 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
582 int osd_procfs_init(struct osd_device *osd, const char *name);
583 int osd_procfs_fini(struct osd_device *osd);
586 extern char *osd_obj_tag;
587 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
588 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
589 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
590 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
591 uint64_t oid, dmu_buf_t **dbp);
592 struct lu_object *osd_object_alloc(const struct lu_env *env,
593 const struct lu_object_header *hdr,
594 struct lu_device *d);
595 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
596 void *buf, uint32_t buflen, struct osd_thandle *oh);
597 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
598 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
599 unsigned dnsize, zap_flags_t flags);
600 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
601 struct osd_object *obj, const struct lu_fid *fid,
602 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
603 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
604 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
605 struct lu_attr *la, uint64_t parent, nvlist_t *);
606 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
607 uint64_t oid, dnode_t **dnp);
608 int osd_object_init0(const struct lu_env *env, struct osd_object *obj);
611 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
612 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
613 int osd_fid_lookup(const struct lu_env *env,
614 struct osd_device *, const struct lu_fid *, uint64_t *);
615 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
616 const struct lu_fid *fid, char *buf, int bufsize,
618 int osd_options_init(void);
619 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
621 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
622 struct osd_object *obj);
623 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
624 struct osd_device *osd,
625 const struct lu_fid *fid);
626 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
627 struct osd_device *osd,
628 const struct lu_fid *fid);
629 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
630 struct osd_device *osd,
631 const struct lu_fid *fid,
633 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
634 const struct lu_fid *fid);
635 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
636 uint64_t parent, const char *name, uint64_t *child,
637 const struct lu_fid *fid, bool isdir);
639 extern unsigned int osd_oi_count;
642 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
643 uint64_t oid, struct lu_fid *fid);
644 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
645 const struct dt_index_features *feat);
646 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
647 u64 seq, struct lu_seq_range *range);
648 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
649 uint64_t id, uint64_t dirhash);
650 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
651 uint64_t id, uint64_t dirhash);
652 void osd_zap_cursor_fini(zap_cursor_t *zc);
653 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
654 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
655 const struct lu_fid *fid);
656 int osd_add_to_remote_parent(const struct lu_env *env,
657 struct osd_device *osd,
658 struct osd_object *obj,
659 struct osd_thandle *oh);
660 int osd_delete_from_remote_parent(const struct lu_env *env,
661 struct osd_device *osd,
662 struct osd_object *obj,
663 struct osd_thandle *oh, bool destroy);
664 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
668 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
669 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
670 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
672 void osd_scrub_stop(struct osd_device *dev);
673 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
674 const struct lu_fid *fid, uint64_t oid, bool insert);
675 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
679 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
680 struct osd_object *obj,
681 struct osd_thandle *oh);
682 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
683 struct osd_thandle *oh);
684 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
685 struct osd_thandle *oh);
686 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
688 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
689 uint64_t xattr, struct lu_buf *buf,
690 const char *name, int *sizep);
691 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
692 struct lu_buf *buf, const char *name, int *sizep);
693 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
694 struct lu_buf *buf, const char *name);
695 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
696 const struct lu_buf *buf, const char *name,
697 int fl, struct thandle *handle);
698 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
699 const struct lu_buf *buf, const char *name, int fl,
700 struct thandle *handle);
701 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
702 const char *name, struct thandle *handle);
703 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
704 const char *name, struct thandle *handle);
705 void osd_declare_xattrs_destroy(const struct lu_env *env,
706 struct osd_object *obj,
707 struct osd_thandle *oh);
708 int osd_xattrs_destroy(const struct lu_env *env,
709 struct osd_object *obj, struct osd_thandle *oh);
710 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
711 const struct lu_buf *lb);
712 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
713 int vallen, const char *name, struct osd_thandle *oh);
714 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
715 const struct lu_buf *buf, const char *name, int fl,
716 struct osd_thandle *oh);;
717 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
718 const struct lu_buf *buf, const char *name, int fl,
719 struct osd_thandle *oh);
720 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
721 struct osd_thandle *oh);
723 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
724 sizeof(__u64) /* VBR VERSION */ + \
725 sizeof(struct lustre_mdt_attrs) /* LMA */)
727 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
728 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
731 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
733 return DN_MAX_BONUSLEN;
737 static inline int osd_object_is_zap(dnode_t *dn)
739 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
740 dn->dn_type == DMU_OT_USERGROUP_USED);
743 /* XXX: f_ver is not counted, but may differ too */
744 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
746 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
750 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
751 const struct lu_buf *buf, const char *name, int fl,
752 struct osd_thandle *oh)
756 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
760 if (osd_obj2dev(obj)->od_xattr_in_sa) {
761 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
763 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
765 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
771 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
773 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
774 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
775 #ifdef ZFS_PROJINHERIT
776 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
778 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
781 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
783 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
784 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
785 #ifdef ZFS_PROJINHERIT
786 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
788 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
793 #ifndef HAVE_DSL_POOL_CONFIG
794 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
798 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
803 #ifdef HAVE_SPA_MAXBLOCKSIZE
804 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
805 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
807 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
808 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
809 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
812 #ifdef HAVE_SA_SPILL_ALLOC
814 osd_zio_buf_alloc(size_t size)
816 return sa_spill_alloc(KM_SLEEP);
820 osd_zio_buf_free(void *buf, size_t size)
825 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
826 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
829 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
830 static inline uint64_t
831 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
832 int dnodesize, dmu_tx_t *tx)
835 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
837 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
838 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
841 static inline uint64_t
842 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
843 dmu_object_type_t ot, int leaf_blockshift,
844 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
847 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
849 return zap_create_flags_dnsize(os, normflags, flags, ot,
850 leaf_blockshift, indirect_blockshift,
851 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
856 osd_obj_bonuslen(struct osd_object *obj)
858 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
860 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
861 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
863 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
867 dnodesize = dmu_objset_dnodesize(os);
869 bonuslen = DN_BONUS_SIZE(dnodesize);
876 static inline uint64_t
877 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
878 int dnodesize, dmu_tx_t *tx)
880 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
881 DN_MAX_BONUSLEN, tx);
884 static inline uint64_t
885 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
886 dmu_object_type_t ot, int leaf_blockshift,
887 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
889 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
890 indirect_blockshift, DMU_OT_SA,
891 DN_MAX_BONUSLEN, tx);
895 osd_obj_bonuslen(struct osd_object *obj)
897 return DN_MAX_BONUSLEN;
899 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
901 #ifdef HAVE_DMU_PREFETCH_6ARG
902 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
903 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
905 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
906 dmu_prefetch((os), (obj), (lvl), (off))
909 static inline int osd_sa_handle_get(struct osd_object *obj)
911 struct osd_device *osd = osd_obj2dev(obj);
912 dnode_t *dn = obj->oo_dn;
918 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
919 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
920 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
923 refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
927 static inline void osd_dnode_rele(dnode_t *dn)
931 LASSERT(dn->dn_bonus);
935 dmu_buf_rele(&db->db, osd_obj_tag);
938 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
940 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
941 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
943 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
945 return dmu_objset_userobjspace_present(osd->od_os);
949 #define OSD_DMU_USEROBJ_PREFIX "obj-"
950 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
952 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
956 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
958 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
959 dnode_t *dn, const char *key,
960 int int_size, int int_num,
961 const void *val, dmu_tx_t *tx)
965 #ifdef HAVE_ZAP_ADD_BY_DNODE
967 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
969 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
972 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
973 dnode_t *dn, const char *key,
978 #ifdef HAVE_ZAP_ADD_BY_DNODE
980 return -zap_remove_by_dnode(dn, key, tx);
982 return -zap_remove(osd->od_os, zap, key, tx);
986 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
987 dnode_t *dn, const char *key,
988 int int_size, int int_num, void *v)
992 #ifdef HAVE_ZAP_ADD_BY_DNODE
994 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
996 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
999 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
1000 dnode_t *dn, int add, const char *name)
1002 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1004 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
1008 dmu_tx_hold_zap(tx, zap, add, name);
1011 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
1012 dnode_t *dn, uint64_t off, int len)
1014 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1016 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1020 dmu_tx_hold_write(tx, oid, off, len);
1023 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1024 uint64_t offset, uint64_t size,
1025 const char *buf, dmu_tx_t *tx)
1028 #ifdef HAVE_DMU_WRITE_BY_DNODE
1029 dmu_write_by_dnode(dn, offset, size, buf, tx);
1031 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1035 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1036 uint64_t offset, uint64_t size,
1037 char *buf, int flags)
1040 #ifdef HAVE_DMU_READ_BY_DNODE
1041 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1043 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1047 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1048 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1049 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1051 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1052 dmu_objset_own((name), (type), (ronly), (tag), (os))
1055 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1056 #define osd_dmu_objset_disown(os, decrypt, tag) \
1057 dmu_objset_disown((os), (decrypt), (tag))
1059 #define osd_dmu_objset_disown(os, decrypt, tag) \
1060 dmu_objset_disown((os), (tag))
1064 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1065 __u32 keysize, __u32 recsize)
1067 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1068 &osd->od_index_backup_list, &osd->od_lock,
1069 &osd->od_index_backup_stop,
1070 fid, keysize, recsize);
1074 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1076 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1079 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1080 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1081 osd->od_index_backup_id);
1085 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1086 &osd->od_index_backup_list, &osd->od_lock,
1087 &osd->od_index_backup_stop, backup);
1090 #endif /* _OSD_INTERNAL_H */