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>
59 * By design including kmem.h overrides the Linux slab interfaces to provide
60 * the Illumos kmem cache interfaces. To override this and gain access to
61 * the Linux interfaces these preprocessor macros must be undefined.
63 #ifdef kmem_cache_destroy
64 #undef kmem_cache_destroy
67 #ifdef kmem_cache_create
68 #undef kmem_cache_create
71 #ifdef kmem_cache_alloc
72 #undef kmem_cache_alloc
75 #ifdef kmem_cache_free
76 #undef kmem_cache_free
79 #define ZFS_VERSION_CODE \
80 OBD_OCD_VERSION(ZFS_MAJOR, ZFS_MINOR, ZFS_PATCH, ZFS_FIX)
82 #define LUSTRE_ROOT_FID_SEQ 0
83 #define DMU_OSD_SVNAME "svname"
84 #define DMU_OSD_OI_NAME_BASE "oi"
86 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
88 /* Statfs space reservation for grant, fragmentation, and unlink space. */
89 #define OSD_STATFS_RESERVED_SIZE (16ULL << 20) /* reserve 16MB minimum */
90 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
92 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
93 #define OSD_DNODE_MIN_BLKSHIFT (DNODES_PER_BLOCK_SHIFT)
94 #define OSD_DNODE_EST_BLKSHIFT (12) /* est 4KB/dnode */
95 #define OSD_DNODE_EST_COUNT 4096
97 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
100 * Iterator's in-memory data structure for quota file.
102 struct osd_it_quota {
103 struct osd_object *oiq_obj;
104 /* DMU accounting object id */
107 zap_cursor_t *oiq_zc;
108 /** identifier for current quota record */
110 unsigned oiq_reset:1; /* 1 -- no need to advance */
114 * Iterator's in-memory data structure for ZAPs
116 * ZFS does not store . and .. on a disk, instead they are
117 * generated up on request
118 * to follow this format we do the same
121 zap_cursor_t *ozi_zc;
122 struct osd_object *ozi_obj;
123 unsigned ozi_reset:1; /* 1 -- no need to advance */
124 /* ozi_pos - position of the cursor:
125 * 0 - before any record
128 * 3 - real records */
131 char ozi_name[MAXNAMELEN]; /* file name for dir */
132 __u64 ozi_key; /* binary key for index files */
135 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
138 * regular ZFS direntry
140 struct zpl_direntry {
141 uint64_t zde_dnode:48,
144 } __attribute__((packed));
147 * lustre direntry adds a fid to regular ZFS direntry
149 struct luz_direntry {
150 struct zpl_direntry lzd_reg;
151 struct lu_fid lzd_fid;
152 } __attribute__((packed));
155 /* cached SA attributes */
160 #ifdef ZFS_PROJINHERIT
173 #define OSD_INS_CACHE_SIZE 8
176 struct osd_idmap_cache {
177 struct osd_device *oic_dev;
178 struct lu_fid oic_fid;
179 /** max 2^48 dnodes per dataset, avoid spilling into another word */
180 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
181 oic_remote:1; /* FID isn't local */
184 struct osd_inconsistent_item {
185 /* link into lustre_scrub::os_inconsistent_items,
186 * protected by lustr_scrub::os_lock. */
187 struct list_head oii_list;
189 /* The right FID <=> oid mapping. */
190 struct osd_idmap_cache oii_cache;
192 unsigned int oii_insert:1; /* insert or update mapping. */
195 struct osd_otable_it {
196 struct osd_device *ooi_dev;
197 struct lu_fid ooi_fid;
199 __u64 ooi_prefetched_dnode;
202 /* The following bits can be updated/checked w/o lock protection.
203 * If more bits will be introduced in the future and need lock to
204 * protect, please add comment. */
205 unsigned int ooi_used_outside:1, /* Some user out of OSD
206 * uses the iteration. */
207 ooi_all_cached:1, /* No more entries can be
208 * filled into cache. */
209 ooi_user_ready:1, /* The user out of OSD is
210 * ready to iterate. */
211 ooi_waiting:1; /* it::next is waiting. */
214 extern const struct dt_index_operations osd_otable_ops;
216 /* max.number of regular attributes the callers may ask for */
217 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
219 struct osd_thread_info {
220 const struct lu_env *oti_env;
222 struct lu_fid oti_fid;
224 * XXX temporary: for ->i_op calls.
226 struct timespec oti_time;
228 struct ost_id oti_ostid;
234 char oti_key[MAXNAMELEN + 1];
235 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
236 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
238 struct lustre_mdt_attrs oti_mdt_attrs;
239 unsigned int oti_in_trans:1;
241 struct lu_attr oti_la;
242 struct osa_attr oti_osa;
243 zap_attribute_t oti_za;
244 dmu_object_info_t oti_doi;
245 struct luz_direntry oti_zde;
247 struct lquota_id_info oti_qi;
248 struct lu_seq_range oti_seq_range;
250 /* dedicated OI cache for insert (which needs inum) */
251 struct osd_idmap_cache *oti_ins_cache;
252 int oti_ins_cache_size;
253 int oti_ins_cache_used;
254 struct lu_buf oti_xattr_lbuf;
258 extern struct lu_context_key osd_key;
260 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
262 return lu_context_key_get(&env->le_ctx, &osd_key);
266 struct thandle ot_super;
267 struct list_head ot_dcb_list;
268 struct list_head ot_stop_dcb_list;
269 struct list_head ot_unlinked_list;
270 struct list_head ot_sa_list;
272 struct lquota_trans ot_quota_trans;
273 __u32 ot_write_commit:1,
277 #define OSD_OI_NAME_SIZE 24
280 * Object Index (OI) instance.
283 char oi_name[OSD_OI_NAME_SIZE];
290 uint64_t *os_compat_dirs;
291 int os_subdir_count; /* subdir count for each seq */
292 u64 os_seq; /* seq number */
293 struct list_head os_seq_list; /* list to seq_list */
296 struct osd_seq_list {
297 rwlock_t osl_seq_list_lock; /* lock for seq_list */
298 struct list_head osl_seq_list; /* list head for seq */
299 struct semaphore osl_seq_init_sem;
302 #define OSD_OST_MAP_SIZE 32
309 struct dt_device od_dt_dev;
310 /* information about underlying file system */
311 struct objset *od_os;
312 uint64_t od_rootid; /* id of root znode */
313 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
314 /* SA attr mapping->id,
315 * name is the same as in ZFS to use defines SA_ZPL_...*/
316 sa_attr_type_t *z_attr_table;
318 struct proc_dir_entry *od_proc_entry;
319 struct lprocfs_stats *od_stats;
321 uint64_t od_remote_parent_dir;
322 uint64_t od_max_blksz;
325 struct osd_oi **od_oi_table;
326 unsigned int od_oi_count;
327 struct osd_seq_list od_seq_list;
329 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
330 od_prop_rdonly:1, /**< ZFS property readonly */
335 unsigned int od_dnsize;
343 __s64 od_auto_scrub_interval;
344 struct lu_site od_site;
346 dnode_t *od_groupused_dn;
347 dnode_t *od_userused_dn;
348 #ifdef ZFS_PROJINHERIT
349 dnode_t *od_projectused_dn;
352 /* quota slave instance */
353 struct qsd_instance *od_quota_slave;
355 struct brw_stats od_brw_stats;
356 atomic_t od_r_in_flight;
357 atomic_t od_w_in_flight;
359 /* used to debug zerocopy logic: the fields track all
360 * allocated, loaned and referenced buffers in use.
361 * to be removed once the change is tested well. */
362 atomic_t od_zerocopy_alloc;
363 atomic_t od_zerocopy_loan;
364 atomic_t od_zerocopy_pin;
366 arc_prune_t *arc_prune_cb;
368 /* osd seq instance */
369 struct lu_client_seq *od_cl_seq;
371 struct semaphore od_otable_sem;
372 struct osd_otable_it *od_otable_it;
373 struct lustre_scrub od_scrub;
374 struct list_head od_ios_list;
377 enum osd_destroy_type {
378 OSD_DESTROY_NONE = 0,
379 OSD_DESTROY_SYNC = 1,
380 OSD_DESTROY_ASYNC = 2,
384 struct dt_object oo_dt;
386 * Inode for file system object represented by this osd_object. This
387 * inode is pinned for the whole duration of lu_object life.
389 * Not modified concurrently (either setup early during object
390 * creation, or assigned by osd_create() under write lock).
393 sa_handle_t *oo_sa_hdl;
394 nvlist_t *oo_sa_xattr;
395 struct list_head oo_sa_linkage;
397 /* used to implement osd_object_*_{lock|unlock} */
398 struct rw_semaphore oo_sem;
400 /* to serialize some updates: destroy vs. others,
401 * xattr_set, object block size change etc */
402 struct rw_semaphore oo_guard;
404 /* protected by oo_guard */
405 struct list_head oo_unlinked_linkage;
407 /* cached attributes */
408 rwlock_t oo_attr_lock;
409 struct lu_attr oo_attr;
411 /* external dnode holding large EAs, protected by oo_guard */
413 enum osd_destroy_type oo_destroy;
415 __u32 oo_destroyed:1,
417 #ifdef ZFS_PROJINHERIT
423 /* the i_flags in LMA */
426 int oo_ea_in_bonus; /* EA bytes we expect */
428 /* record size for index file */
429 unsigned char oo_keysize;
430 unsigned char oo_recsize;
431 unsigned char oo_recusize; /* unit size */
433 uint64_t oo_parent; /* used only at object creation */
435 struct lu_object_header *oo_header;
438 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
439 extern const struct dt_index_operations osd_acct_index_ops;
440 extern struct lu_device_operations osd_lu_ops;
441 extern struct dt_index_operations osd_dir_ops;
442 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
443 qid_t uid, qid_t gid, qid_t projid, long long space,
444 struct osd_thandle *oh, int *flags,
445 enum osd_qid_declare_flags osd_qid_declare_flags);
446 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
447 uint64_t nrblocks, uint64_t est_maxblockshift);
448 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
454 static inline int lu_device_is_osd(const struct lu_device *d)
456 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
459 static inline struct osd_object *osd_obj(const struct lu_object *o)
461 LASSERT(lu_device_is_osd(o->lo_dev));
462 return container_of0(o, struct osd_object, oo_dt.do_lu);
465 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
467 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
468 return container_of0(d, struct osd_device, od_dt_dev);
471 static inline struct osd_device *osd_dev(const struct lu_device *d)
473 LASSERT(lu_device_is_osd(d));
474 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
477 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
479 return osd_obj(&d->do_lu);
482 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
484 return osd_dev(o->oo_dt.do_lu.lo_dev);
487 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
489 return &osd->od_dt_dev.dd_lu_dev;
492 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
494 return osd_dev(o->do_lu.lo_dev)->od_os;
497 static inline int osd_invariant(const struct osd_object *obj)
503 * Put the osd object once done with it.
505 * \param obj osd object that needs to be put
507 static inline void osd_object_put(const struct lu_env *env,
508 struct osd_object *obj)
510 dt_object_put(env, &obj->oo_dt);
513 static inline int osd_object_invariant(const struct lu_object *l)
515 return osd_invariant(osd_obj(l));
518 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
520 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
523 static inline char *osd_name(struct osd_device *osd)
525 return osd->od_svname;
528 static inline void zfs_set_bit(int nr, __u8 *addr)
530 set_bit(nr, (unsigned long *)addr);
533 static inline int zfs_test_bit(int nr, __u8 *addr)
535 return test_bit(nr, (const unsigned long *)addr);
538 static inline int osd_oi_fid2idx(struct osd_device *dev,
539 const struct lu_fid *fid)
541 return fid->f_seq & (dev->od_oi_count - 1);
544 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
545 const struct lu_fid *fid)
547 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
548 "%s: "DFID", oi_count %d\n",
549 osd_name(osd), PFID(fid), osd->od_oi_count);
551 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
554 #ifdef CONFIG_PROC_FS
556 LPROC_OSD_READ_BYTES = 0,
557 LPROC_OSD_WRITE_BYTES = 1,
558 LPROC_OSD_GET_PAGE = 2,
559 LPROC_OSD_NO_PAGE = 3,
560 LPROC_OSD_CACHE_ACCESS = 4,
561 LPROC_OSD_CACHE_HIT = 5,
562 LPROC_OSD_CACHE_MISS = 6,
563 LPROC_OSD_COPY_IO = 7,
564 LPROC_OSD_ZEROCOPY_IO = 8,
565 LPROC_OSD_TAIL_IO = 9,
569 extern struct kmem_cache *osd_zapit_cachep;
571 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
573 int osd_procfs_init(struct osd_device *osd, const char *name);
574 int osd_procfs_fini(struct osd_device *osd);
577 extern char *osd_obj_tag;
578 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
579 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
580 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
581 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
582 uint64_t oid, dmu_buf_t **dbp);
583 struct lu_object *osd_object_alloc(const struct lu_env *env,
584 const struct lu_object_header *hdr,
585 struct lu_device *d);
586 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
587 void *buf, uint32_t buflen, struct osd_thandle *oh);
588 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
589 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
590 unsigned dnsize, zap_flags_t flags);
591 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
592 struct osd_object *obj, const struct lu_fid *fid,
593 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
594 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
595 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
596 struct lu_attr *la, uint64_t parent, nvlist_t *);
597 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
598 uint64_t oid, dnode_t **dnp);
599 int osd_object_init0(const struct lu_env *env, struct osd_object *obj);
602 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
603 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
604 int osd_fid_lookup(const struct lu_env *env,
605 struct osd_device *, const struct lu_fid *, uint64_t *);
606 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
607 const struct lu_fid *fid, char *buf, int bufsize,
609 int osd_options_init(void);
610 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
612 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
613 struct osd_object *obj);
614 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
615 struct osd_device *osd,
616 const struct lu_fid *fid);
617 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
618 struct osd_device *osd,
619 const struct lu_fid *fid);
620 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
621 struct osd_device *osd,
622 const struct lu_fid *fid,
624 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
625 const struct lu_fid *fid);
626 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
627 uint64_t parent, const char *name, uint64_t *child,
628 const struct lu_fid *fid, bool isdir);
630 extern unsigned int osd_oi_count;
633 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
634 const struct dt_index_features *feat);
635 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
636 u64 seq, struct lu_seq_range *range);
637 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
638 uint64_t id, uint64_t dirhash);
639 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
640 uint64_t id, uint64_t dirhash);
641 void osd_zap_cursor_fini(zap_cursor_t *zc);
642 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
643 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
644 const struct lu_fid *fid);
645 int osd_add_to_remote_parent(const struct lu_env *env,
646 struct osd_device *osd,
647 struct osd_object *obj,
648 struct osd_thandle *oh);
649 int osd_delete_from_remote_parent(const struct lu_env *env,
650 struct osd_device *osd,
651 struct osd_object *obj,
652 struct osd_thandle *oh, bool destroy);
653 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
657 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
658 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
659 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
661 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
662 const struct lu_fid *fid, uint64_t oid, bool insert);
663 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
667 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
668 struct osd_object *obj,
669 struct osd_thandle *oh);
670 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
671 struct osd_thandle *oh);
672 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
673 struct osd_thandle *oh);
674 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
676 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
677 uint64_t xattr, struct lu_buf *buf,
678 const char *name, int *sizep);
679 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
680 struct lu_buf *buf, const char *name, int *sizep);
681 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
682 struct lu_buf *buf, const char *name);
683 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
684 const struct lu_buf *buf, const char *name,
685 int fl, struct thandle *handle);
686 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
687 const struct lu_buf *buf, const char *name, int fl,
688 struct thandle *handle);
689 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
690 const char *name, struct thandle *handle);
691 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
692 const char *name, struct thandle *handle);
693 void osd_declare_xattrs_destroy(const struct lu_env *env,
694 struct osd_object *obj,
695 struct osd_thandle *oh);
696 int osd_xattrs_destroy(const struct lu_env *env,
697 struct osd_object *obj, struct osd_thandle *oh);
698 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
699 const struct lu_buf *lb);
700 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
701 int vallen, const char *name, struct osd_thandle *oh);
702 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
703 const struct lu_buf *buf, const char *name, int fl,
704 struct osd_thandle *oh);;
705 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
706 const struct lu_buf *buf, const char *name, int fl,
707 struct osd_thandle *oh);
708 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
709 struct osd_thandle *oh);
711 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
712 sizeof(__u64) /* VBR VERSION */ + \
713 sizeof(struct lustre_mdt_attrs) /* LMA */)
715 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
716 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
719 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
721 return DN_MAX_BONUSLEN;
725 /* XXX: f_ver is not counted, but may differ too */
726 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
728 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
732 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
733 const struct lu_buf *buf, const char *name, int fl,
734 struct osd_thandle *oh)
738 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
742 if (osd_obj2dev(obj)->od_xattr_in_sa) {
743 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
745 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
747 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
753 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
755 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
756 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
757 #ifdef ZFS_PROJINHERIT
758 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
760 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
763 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
765 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
766 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
767 #ifdef ZFS_PROJINHERIT
768 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
770 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
775 #ifndef HAVE_DSL_POOL_CONFIG
776 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
780 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
785 #ifdef HAVE_SPA_MAXBLOCKSIZE
786 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
787 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
789 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
790 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
791 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
794 #ifdef HAVE_SA_SPILL_ALLOC
796 osd_zio_buf_alloc(size_t size)
798 return sa_spill_alloc(KM_SLEEP);
802 osd_zio_buf_free(void *buf, size_t size)
807 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
808 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
811 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
812 static inline uint64_t
813 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
814 int dnodesize, dmu_tx_t *tx)
817 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
819 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
820 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
823 static inline uint64_t
824 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
825 dmu_object_type_t ot, int leaf_blockshift,
826 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
829 dnodesize = MAX(dmu_objset_dnodesize(os), DNODE_MIN_SIZE);
831 return zap_create_flags_dnsize(os, normflags, flags, ot,
832 leaf_blockshift, indirect_blockshift,
833 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
838 osd_obj_bonuslen(struct osd_object *obj)
840 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
842 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
843 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
845 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
849 dnodesize = dmu_objset_dnodesize(os);
851 bonuslen = DN_BONUS_SIZE(dnodesize);
858 static inline uint64_t
859 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
860 int dnodesize, dmu_tx_t *tx)
862 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
863 DN_MAX_BONUSLEN, tx);
866 static inline uint64_t
867 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
868 dmu_object_type_t ot, int leaf_blockshift,
869 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
871 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
872 indirect_blockshift, DMU_OT_SA,
873 DN_MAX_BONUSLEN, tx);
877 osd_obj_bonuslen(struct osd_object *obj)
879 return DN_MAX_BONUSLEN;
881 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
883 #ifdef HAVE_DMU_PREFETCH_6ARG
884 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
885 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
887 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
888 dmu_prefetch((os), (obj), (lvl), (off))
891 static inline int osd_sa_handle_get(struct osd_object *obj)
893 struct osd_device *osd = osd_obj2dev(obj);
894 dnode_t *dn = obj->oo_dn;
900 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
901 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
902 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
905 refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
909 static inline void osd_dnode_rele(dnode_t *dn)
913 LASSERT(dn->dn_bonus);
917 dmu_buf_rele(&db->db, osd_obj_tag);
920 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
922 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
923 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
925 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
927 return dmu_objset_userobjspace_present(osd->od_os);
931 #define OSD_DMU_USEROBJ_PREFIX "obj-"
932 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
934 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
938 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
940 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
941 dnode_t *dn, const char *key,
942 int int_size, int int_num,
943 const void *val, dmu_tx_t *tx)
947 #ifdef HAVE_ZAP_ADD_BY_DNODE
949 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
951 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
954 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
955 dnode_t *dn, const char *key,
960 #ifdef HAVE_ZAP_ADD_BY_DNODE
962 return -zap_remove_by_dnode(dn, key, tx);
964 return -zap_remove(osd->od_os, zap, key, tx);
968 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
969 dnode_t *dn, const char *key,
970 int int_size, int int_num, void *v)
974 #ifdef HAVE_ZAP_ADD_BY_DNODE
976 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
978 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
981 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
982 dnode_t *dn, int add, const char *name)
984 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
986 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
990 dmu_tx_hold_zap(tx, zap, add, name);
993 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
994 dnode_t *dn, uint64_t off, int len)
996 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
998 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1002 dmu_tx_hold_write(tx, oid, off, len);
1005 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1006 uint64_t offset, uint64_t size,
1007 const char *buf, dmu_tx_t *tx)
1010 #ifdef HAVE_DMU_WRITE_BY_DNODE
1011 dmu_write_by_dnode(dn, offset, size, buf, tx);
1013 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1017 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1018 uint64_t offset, uint64_t size,
1019 char *buf, int flags)
1022 #ifdef HAVE_DMU_READ_BY_DNODE
1023 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1025 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1029 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1030 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1031 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1033 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1034 dmu_objset_own((name), (type), (ronly), (tag), (os))
1037 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1038 #define osd_dmu_objset_disown(os, decrypt, tag) \
1039 dmu_objset_disown((os), (decrypt), (tag))
1041 #define osd_dmu_objset_disown(os, decrypt, tag) \
1042 dmu_objset_disown((os), (tag))
1045 #endif /* _OSD_INTERNAL_H */