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.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Whamcloud, Inc.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/md_object.h
38 * Extention of lu_object.h for metadata objects
41 #ifndef _LUSTRE_MD_OBJECT_H
42 #define _LUSTRE_MD_OBJECT_H
45 * Sub-class of lu_object with methods common for "meta-data" objects in MDT
48 * Meta-data objects implement namespace operations: you can link, unlink
49 * them, and treat them as directories.
51 * Examples: mdt, cmm, and mdt are implementations of md interface.
57 * super-class definitions.
59 #include <dt_object.h>
62 /* LU-1051, temperary solution to reduce llog credits */
63 #define DECLARE_LLOG_REWRITE 0
64 #define DECLARE_LLOG_WRITE INT_MAX
67 struct md_device_operations;
91 cfs_group_info_t *mu_ginfo;
92 struct md_identity *mu_identity;
99 /** there are at most 5 fids in one operation, see rename, NOTE the last one
100 * is a temporary one used for is_subdir() */
104 struct lu_fid mc_fid[MD_CAPAINFO_MAX];
105 struct lustre_capa *mc_capa[MD_CAPAINFO_MAX];
109 struct obd_export *mq_exp;
113 * Implemented in mdd/mdd_handler.c.
115 * XXX should be moved into separate .h/.c together with all md security
116 * related definitions.
118 struct md_ucred *md_ucred(const struct lu_env *env);
119 struct md_capainfo *md_capainfo(const struct lu_env *env);
120 struct md_quota *md_quota(const struct lu_env *env);
122 /** metadata attributes */
126 MA_COOKIE = (1 << 2),
129 MA_ACL_DEF = (1 << 5),
130 MA_LOV_DEF = (1 << 6),
131 MA_LAY_GEN = (1 << 7),
151 MDT_REG_LOCK = (1 << 0),
152 MDT_PDO_LOCK = (1 << 1)
157 __u32 mh_archive_number;
160 #define IOEPOCH_INVAL 0
173 struct lu_attr ma_attr;
174 struct lu_fid ma_pfid;
175 struct md_hsm ma_hsm;
176 struct lov_mds_md *ma_lmm;
177 struct lmv_stripe_md *ma_lmv;
179 struct llog_cookie *ma_cookie;
180 struct lustre_capa *ma_capa;
181 struct md_som_data *ma_som;
183 int ma_big_lmm_used:1;
190 /** Additional parameters for create */
193 /** symlink target */
194 const char *sp_symname;
195 /** parent FID for cross-ref mkdir */
196 const struct lu_fid *sp_pfid;
197 /** eadata for regular files */
199 /** lov objs exist already */
200 const struct lu_fid *fid;
205 /** don't create lov objects or llog cookie - this replay */
208 /** Create flag from client: such as MDS_OPEN_CREAT, and others. */
211 /** Should mdd do lookup sanity check or not. */
214 /** Current lock mode for parent dir where create is performing. */
215 mdl_mode_t sp_cr_mode;
217 /** to create directory */
218 const struct dt_index_features *sp_feat;
222 * Operations implemented for each md object (both directory and leaf).
224 struct md_object_operations {
225 int (*moo_permission)(const struct lu_env *env,
226 struct md_object *pobj, struct md_object *cobj,
227 struct md_attr *attr, int mask);
229 int (*moo_attr_get)(const struct lu_env *env, struct md_object *obj,
230 struct md_attr *attr);
232 int (*moo_attr_set)(const struct lu_env *env, struct md_object *obj,
233 const struct md_attr *attr);
235 int (*moo_xattr_get)(const struct lu_env *env, struct md_object *obj,
236 struct lu_buf *buf, const char *name);
238 int (*moo_xattr_list)(const struct lu_env *env, struct md_object *obj,
241 int (*moo_xattr_set)(const struct lu_env *env, struct md_object *obj,
242 const struct lu_buf *buf, const char *name,
245 int (*moo_xattr_del)(const struct lu_env *env, struct md_object *obj,
248 /** \retval number of bytes actually read upon success */
249 int (*moo_readpage)(const struct lu_env *env, struct md_object *obj,
250 const struct lu_rdpg *rdpg);
252 int (*moo_readlink)(const struct lu_env *env, struct md_object *obj,
254 int (*moo_changelog)(const struct lu_env *env,
255 enum changelog_rec_type type, int flags,
256 struct md_object *obj);
257 /** part of cross-ref operation */
258 int (*moo_object_create)(const struct lu_env *env,
259 struct md_object *obj,
260 const struct md_op_spec *spec,
263 int (*moo_ref_add)(const struct lu_env *env,
264 struct md_object *obj,
265 const struct md_attr *ma);
267 int (*moo_ref_del)(const struct lu_env *env,
268 struct md_object *obj,
271 int (*moo_open)(const struct lu_env *env,
272 struct md_object *obj, int flag);
274 int (*moo_close)(const struct lu_env *env, struct md_object *obj,
275 struct md_attr *ma, int mode);
277 int (*moo_capa_get)(const struct lu_env *, struct md_object *,
278 struct lustre_capa *, int renewal);
280 int (*moo_object_sync)(const struct lu_env *, struct md_object *);
281 int (*moo_path)(const struct lu_env *env, struct md_object *obj,
282 char *path, int pathlen, __u64 *recno, int *linkno);
283 int (*moo_file_lock)(const struct lu_env *env, struct md_object *obj,
284 struct lov_mds_md *lmm, struct ldlm_extent *extent,
285 struct lustre_handle *lockh);
286 int (*moo_file_unlock)(const struct lu_env *env, struct md_object *obj,
287 struct lov_mds_md *lmm,
288 struct lustre_handle *lockh);
292 * Operations implemented for each directory object.
294 struct md_dir_operations {
295 int (*mdo_is_subdir) (const struct lu_env *env, struct md_object *obj,
296 const struct lu_fid *fid, struct lu_fid *sfid);
298 int (*mdo_lookup)(const struct lu_env *env, struct md_object *obj,
299 const struct lu_name *lname, struct lu_fid *fid,
300 struct md_op_spec *spec);
302 mdl_mode_t (*mdo_lock_mode)(const struct lu_env *env,
303 struct md_object *obj,
306 int (*mdo_create)(const struct lu_env *env, struct md_object *pobj,
307 const struct lu_name *lname, struct md_object *child,
308 struct md_op_spec *spec,
311 /** This method is used for creating data object for this meta object*/
312 int (*mdo_create_data)(const struct lu_env *env, struct md_object *p,
314 const struct md_op_spec *spec,
317 int (*mdo_rename)(const struct lu_env *env, struct md_object *spobj,
318 struct md_object *tpobj, const struct lu_fid *lf,
319 const struct lu_name *lsname, struct md_object *tobj,
320 const struct lu_name *ltname, struct md_attr *ma);
322 int (*mdo_link)(const struct lu_env *env, struct md_object *tgt_obj,
323 struct md_object *src_obj, const struct lu_name *lname,
326 int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj,
327 struct md_object *cobj, const struct lu_name *lname,
330 /** This method is used to compare a requested layout to an existing
331 * layout (struct lov_mds_md_v1/3 vs struct lov_mds_md_v1/3) */
332 int (*mdo_lum_lmm_cmp)(const struct lu_env *env,
333 struct md_object *cobj,
334 const struct md_op_spec *spec,
337 /** partial ops for cross-ref case */
338 int (*mdo_name_insert)(const struct lu_env *env,
339 struct md_object *obj,
340 const struct lu_name *lname,
341 const struct lu_fid *fid,
342 const struct md_attr *ma);
344 int (*mdo_name_remove)(const struct lu_env *env,
345 struct md_object *obj,
346 const struct lu_name *lname,
347 const struct md_attr *ma);
349 int (*mdo_rename_tgt)(const struct lu_env *env, struct md_object *pobj,
350 struct md_object *tobj, const struct lu_fid *fid,
351 const struct lu_name *lname, struct md_attr *ma);
354 struct md_device_operations {
355 /** meta-data device related handlers. */
356 int (*mdo_root_get)(const struct lu_env *env, struct md_device *m,
359 int (*mdo_maxsize_get)(const struct lu_env *env, struct md_device *m,
360 int *md_size, int *cookie_size);
362 int (*mdo_statfs)(const struct lu_env *env, struct md_device *m,
363 struct obd_statfs *sfs);
365 int (*mdo_init_capa_ctxt)(const struct lu_env *env, struct md_device *m,
366 int mode, unsigned long timeout, __u32 alg,
367 struct lustre_capa_key *keys);
369 int (*mdo_update_capa_key)(const struct lu_env *env,
371 struct lustre_capa_key *key);
373 int (*mdo_llog_ctxt_get)(const struct lu_env *env,
374 struct md_device *m, int idx, void **h);
376 int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m,
377 unsigned int cmd, int len, void *data);
380 enum md_upcall_event {
381 /** Sync the md layer*/
382 MD_LOV_SYNC = (1 << 0),
383 /** Just for split, no need trans, for replay */
384 MD_NO_TRANS = (1 << 1),
385 MD_LOV_CONFIG = (1 << 2),
386 /** Trigger quota recovery */
387 MD_LOV_QUOTA = (1 << 3)
391 /** this lock protects upcall using against its removal
392 * read lock is for usage the upcall, write - for init/fini */
393 struct rw_semaphore mu_upcall_sem;
394 /** device to call, upper layer normally */
395 struct md_device *mu_upcall_dev;
396 /** upcall function */
397 int (*mu_upcall)(const struct lu_env *env, struct md_device *md,
398 enum md_upcall_event ev, void *data);
402 struct lu_device md_lu_dev;
403 const struct md_device_operations *md_ops;
404 struct md_upcall md_upcall;
407 static inline void md_upcall_init(struct md_device *m, void *upcl)
409 init_rwsem(&m->md_upcall.mu_upcall_sem);
410 m->md_upcall.mu_upcall_dev = NULL;
411 m->md_upcall.mu_upcall = upcl;
414 static inline void md_upcall_dev_set(struct md_device *m, struct md_device *up)
416 down_write(&m->md_upcall.mu_upcall_sem);
417 m->md_upcall.mu_upcall_dev = up;
418 up_write(&m->md_upcall.mu_upcall_sem);
421 static inline void md_upcall_fini(struct md_device *m)
423 down_write(&m->md_upcall.mu_upcall_sem);
424 m->md_upcall.mu_upcall_dev = NULL;
425 m->md_upcall.mu_upcall = NULL;
426 up_write(&m->md_upcall.mu_upcall_sem);
429 static inline int md_do_upcall(const struct lu_env *env, struct md_device *m,
430 enum md_upcall_event ev, void *data)
433 down_read(&m->md_upcall.mu_upcall_sem);
434 if (m->md_upcall.mu_upcall_dev != NULL &&
435 m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall != NULL) {
436 rc = m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall(env,
437 m->md_upcall.mu_upcall_dev,
440 up_read(&m->md_upcall.mu_upcall_sem);
445 struct lu_object mo_lu;
446 const struct md_object_operations *mo_ops;
447 const struct md_dir_operations *mo_dir_ops;
454 struct lu_site *ms_lu;
456 * mds number of this site.
460 * Fid location database
462 struct lu_server_fld *ms_server_fld;
463 struct lu_client_fld *ms_client_fld;
468 struct lu_server_seq *ms_server_seq;
471 * Controller Seq Manager
473 struct lu_server_seq *ms_control_seq;
474 struct obd_export *ms_control_exp;
479 struct lu_client_seq *ms_client_seq;
482 static inline int lu_device_is_md(const struct lu_device *d)
484 return ergo(d != NULL, d->ld_type->ldt_tags & LU_DEVICE_MD);
487 static inline struct md_device *lu2md_dev(const struct lu_device *d)
489 LASSERT(IS_ERR(d) || lu_device_is_md(d));
490 return container_of0(d, struct md_device, md_lu_dev);
493 static inline struct lu_device *md2lu_dev(struct md_device *d)
495 return &d->md_lu_dev;
498 static inline struct md_object *lu2md(const struct lu_object *o)
500 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev));
501 return container_of0(o, struct md_object, mo_lu);
504 static inline struct md_object *md_object_next(const struct md_object *obj)
506 return (obj ? lu2md(lu_object_next(&obj->mo_lu)) : NULL);
509 static inline struct md_device *md_obj2dev(const struct md_object *o)
511 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->mo_lu.lo_dev));
512 return container_of0(o->mo_lu.lo_dev, struct md_device, md_lu_dev);
515 static inline struct md_site *lu_site2md(const struct lu_site *s)
517 return s->ld_md_site;
520 static inline int md_device_init(struct md_device *md, struct lu_device_type *t)
522 return lu_device_init(&md->md_lu_dev, t);
525 static inline void md_device_fini(struct md_device *md)
527 lu_device_fini(&md->md_lu_dev);
530 static inline struct md_object *md_object_find_slice(const struct lu_env *env,
531 struct md_device *md,
532 const struct lu_fid *f)
534 return lu2md(lu_object_find_slice(env, md2lu_dev(md), f, NULL));
539 static inline int mo_permission(const struct lu_env *env,
545 LASSERT(c->mo_ops->moo_permission);
546 return c->mo_ops->moo_permission(env, p, c, at, mask);
549 static inline int mo_attr_get(const struct lu_env *env,
553 LASSERT(m->mo_ops->moo_attr_get);
554 return m->mo_ops->moo_attr_get(env, m, at);
557 static inline int mo_readlink(const struct lu_env *env,
561 LASSERT(m->mo_ops->moo_readlink);
562 return m->mo_ops->moo_readlink(env, m, buf);
565 static inline int mo_changelog(const struct lu_env *env,
566 enum changelog_rec_type type,
567 int flags, struct md_object *m)
569 LASSERT(m->mo_ops->moo_changelog);
570 return m->mo_ops->moo_changelog(env, type, flags, m);
573 static inline int mo_attr_set(const struct lu_env *env,
575 const struct md_attr *at)
577 LASSERT(m->mo_ops->moo_attr_set);
578 return m->mo_ops->moo_attr_set(env, m, at);
581 static inline int mo_xattr_get(const struct lu_env *env,
586 LASSERT(m->mo_ops->moo_xattr_get);
587 return m->mo_ops->moo_xattr_get(env, m, buf, name);
590 static inline int mo_xattr_del(const struct lu_env *env,
594 LASSERT(m->mo_ops->moo_xattr_del);
595 return m->mo_ops->moo_xattr_del(env, m, name);
598 static inline int mo_xattr_set(const struct lu_env *env,
600 const struct lu_buf *buf,
604 LASSERT(m->mo_ops->moo_xattr_set);
605 return m->mo_ops->moo_xattr_set(env, m, buf, name, flags);
608 static inline int mo_xattr_list(const struct lu_env *env,
612 LASSERT(m->mo_ops->moo_xattr_list);
613 return m->mo_ops->moo_xattr_list(env, m, buf);
616 static inline int mo_open(const struct lu_env *env,
620 LASSERT(m->mo_ops->moo_open);
621 return m->mo_ops->moo_open(env, m, flags);
624 static inline int mo_close(const struct lu_env *env,
629 LASSERT(m->mo_ops->moo_close);
630 return m->mo_ops->moo_close(env, m, ma, mode);
633 static inline int mo_readpage(const struct lu_env *env,
635 const struct lu_rdpg *rdpg)
637 LASSERT(m->mo_ops->moo_readpage);
638 return m->mo_ops->moo_readpage(env, m, rdpg);
641 static inline int mo_object_create(const struct lu_env *env,
643 const struct md_op_spec *spc,
646 LASSERT(m->mo_ops->moo_object_create);
647 return m->mo_ops->moo_object_create(env, m, spc, at);
650 static inline int mo_ref_add(const struct lu_env *env,
652 const struct md_attr *ma)
654 LASSERT(m->mo_ops->moo_ref_add);
655 return m->mo_ops->moo_ref_add(env, m, ma);
658 static inline int mo_ref_del(const struct lu_env *env,
662 LASSERT(m->mo_ops->moo_ref_del);
663 return m->mo_ops->moo_ref_del(env, m, ma);
666 static inline int mo_capa_get(const struct lu_env *env,
668 struct lustre_capa *c,
671 LASSERT(m->mo_ops->moo_capa_get);
672 return m->mo_ops->moo_capa_get(env, m, c, renewal);
675 static inline int mo_path(const struct lu_env *env, struct md_object *m,
676 char *path, int pathlen, __u64 *recno, int *linkno)
678 if (m->mo_ops->moo_path == NULL)
680 return m->mo_ops->moo_path(env, m, path, pathlen, recno, linkno);
683 static inline int mo_object_sync(const struct lu_env *env, struct md_object *m)
685 LASSERT(m->mo_ops->moo_object_sync);
686 return m->mo_ops->moo_object_sync(env, m);
689 static inline int mo_file_lock(const struct lu_env *env, struct md_object *m,
690 struct lov_mds_md *lmm,
691 struct ldlm_extent *extent,
692 struct lustre_handle *lockh)
694 LASSERT(m->mo_ops->moo_file_lock);
695 return m->mo_ops->moo_file_lock(env, m, lmm, extent, lockh);
698 static inline int mo_file_unlock(const struct lu_env *env, struct md_object *m,
699 struct lov_mds_md *lmm,
700 struct lustre_handle *lockh)
702 LASSERT(m->mo_ops->moo_file_unlock);
703 return m->mo_ops->moo_file_unlock(env, m, lmm, lockh);
706 static inline int mdo_lookup(const struct lu_env *env,
708 const struct lu_name *lname,
710 struct md_op_spec *spec)
712 LASSERT(p->mo_dir_ops->mdo_lookup);
713 return p->mo_dir_ops->mdo_lookup(env, p, lname, f, spec);
716 static inline mdl_mode_t mdo_lock_mode(const struct lu_env *env,
717 struct md_object *mo,
720 if (mo->mo_dir_ops->mdo_lock_mode == NULL)
722 return mo->mo_dir_ops->mdo_lock_mode(env, mo, lm);
725 static inline int mdo_create(const struct lu_env *env,
727 const struct lu_name *lchild_name,
729 struct md_op_spec *spc,
732 LASSERT(p->mo_dir_ops->mdo_create);
733 return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at);
736 static inline int mdo_create_data(const struct lu_env *env,
739 const struct md_op_spec *spec,
742 LASSERT(c->mo_dir_ops->mdo_create_data);
743 return c->mo_dir_ops->mdo_create_data(env, p, c, spec, ma);
746 static inline int mdo_rename(const struct lu_env *env,
747 struct md_object *sp,
748 struct md_object *tp,
749 const struct lu_fid *lf,
750 const struct lu_name *lsname,
752 const struct lu_name *ltname,
755 LASSERT(tp->mo_dir_ops->mdo_rename);
756 return tp->mo_dir_ops->mdo_rename(env, sp, tp, lf, lsname, t, ltname,
760 static inline int mdo_is_subdir(const struct lu_env *env,
761 struct md_object *mo,
762 const struct lu_fid *fid,
765 LASSERT(mo->mo_dir_ops->mdo_is_subdir);
766 return mo->mo_dir_ops->mdo_is_subdir(env, mo, fid, sfid);
769 static inline int mdo_link(const struct lu_env *env,
772 const struct lu_name *lname,
775 LASSERT(s->mo_dir_ops->mdo_link);
776 return s->mo_dir_ops->mdo_link(env, p, s, lname, ma);
779 static inline int mdo_unlink(const struct lu_env *env,
782 const struct lu_name *lname,
785 LASSERT(c->mo_dir_ops->mdo_unlink);
786 return c->mo_dir_ops->mdo_unlink(env, p, c, lname, ma);
789 static inline int mdo_lum_lmm_cmp(const struct lu_env *env,
791 const struct md_op_spec *spec,
794 LASSERT(c->mo_dir_ops->mdo_lum_lmm_cmp);
795 return c->mo_dir_ops->mdo_lum_lmm_cmp(env, c, spec, ma);
798 static inline int mdo_name_insert(const struct lu_env *env,
800 const struct lu_name *lname,
801 const struct lu_fid *f,
802 const struct md_attr *ma)
804 LASSERT(p->mo_dir_ops->mdo_name_insert);
805 return p->mo_dir_ops->mdo_name_insert(env, p, lname, f, ma);
808 static inline int mdo_name_remove(const struct lu_env *env,
810 const struct lu_name *lname,
811 const struct md_attr *ma)
813 LASSERT(p->mo_dir_ops->mdo_name_remove);
814 return p->mo_dir_ops->mdo_name_remove(env, p, lname, ma);
817 static inline int mdo_rename_tgt(const struct lu_env *env,
820 const struct lu_fid *lf,
821 const struct lu_name *lname,
825 LASSERT(t->mo_dir_ops->mdo_rename_tgt);
826 return t->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
828 LASSERT(p->mo_dir_ops->mdo_rename_tgt);
829 return p->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
835 * Structure to hold object information. This is used to create object
836 * \pre llod_dir exist
838 struct lu_local_obj_desc {
839 const char *llod_dir;
840 const char *llod_name;
843 const struct dt_index_features *llod_feat;
844 cfs_list_t llod_linkage;
847 struct md_object *llo_store_resolve(const struct lu_env *env,
848 struct md_device *md,
849 struct dt_device *dt,
853 struct md_object *llo_store_open(const struct lu_env *env,
854 struct md_device *md,
855 struct dt_device *dt,
860 struct md_object *llo_store_create_index(const struct lu_env *env,
861 struct md_device *md,
862 struct dt_device *dt,
865 const struct lu_fid *fid,
866 const struct dt_index_features *feat);
868 struct md_object *llo_store_create(const struct lu_env *env,
869 struct md_device *md,
870 struct dt_device *dt,
873 const struct lu_fid *fid);
875 void llo_local_obj_register(struct lu_local_obj_desc *);
876 void llo_local_obj_unregister(struct lu_local_obj_desc *);
878 int llo_local_objects_setup(const struct lu_env *env,
879 struct md_device * md,
880 struct dt_device * dt);
883 #endif /* _LINUX_MD_OBJECT_H */