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;
82 /** there are at most 5 fids in one operation, see rename, NOTE the last one
83 * is a temporary one used for is_subdir() */
87 struct lu_fid mc_fid[MD_CAPAINFO_MAX];
88 struct lustre_capa *mc_capa[MD_CAPAINFO_MAX];
92 struct obd_export *mq_exp;
96 * Implemented in mdd/mdd_handler.c.
98 * XXX should be moved into separate .h/.c together with all md security
99 * related definitions.
101 struct md_capainfo *md_capainfo(const struct lu_env *env);
102 struct md_quota *md_quota(const struct lu_env *env);
104 /** metadata attributes */
108 MA_COOKIE = (1 << 2),
111 MA_ACL_DEF = (1 << 5),
112 MA_LOV_DEF = (1 << 6),
113 MA_LAY_GEN = (1 << 7),
133 MDT_REG_LOCK = (1 << 0),
134 MDT_PDO_LOCK = (1 << 1)
144 #define IOEPOCH_INVAL 0
159 struct lu_attr ma_attr;
160 struct lu_fid ma_pfid;
161 struct md_hsm ma_hsm;
162 struct lov_mds_md *ma_lmm;
163 struct lmv_stripe_md *ma_lmv;
165 struct llog_cookie *ma_cookie;
166 struct lustre_capa *ma_capa;
167 struct md_som_data *ma_som;
169 int ma_big_lmm_used:1;
176 /** Additional parameters for create */
179 /** symlink target */
180 const char *sp_symname;
181 /** parent FID for cross-ref mkdir */
182 const struct lu_fid *sp_pfid;
183 /** eadata for regular files */
185 /** lov objs exist already */
186 const struct lu_fid *fid;
191 /** don't create lov objects or llog cookie - this replay */
194 /** Create flag from client: such as MDS_OPEN_CREAT, and others. */
197 /** Should mdd do lookup sanity check or not. */
200 /** Current lock mode for parent dir where create is performing. */
201 mdl_mode_t sp_cr_mode;
203 /** to create directory */
204 const struct dt_index_features *sp_feat;
208 * Operations implemented for each md object (both directory and leaf).
210 struct md_object_operations {
211 int (*moo_permission)(const struct lu_env *env,
212 struct md_object *pobj, struct md_object *cobj,
213 struct md_attr *attr, int mask);
215 int (*moo_attr_get)(const struct lu_env *env, struct md_object *obj,
216 struct md_attr *attr);
218 int (*moo_attr_set)(const struct lu_env *env, struct md_object *obj,
219 const struct md_attr *attr);
221 int (*moo_xattr_get)(const struct lu_env *env, struct md_object *obj,
222 struct lu_buf *buf, const char *name);
224 int (*moo_xattr_list)(const struct lu_env *env, struct md_object *obj,
227 int (*moo_xattr_set)(const struct lu_env *env, struct md_object *obj,
228 const struct lu_buf *buf, const char *name,
231 int (*moo_xattr_del)(const struct lu_env *env, struct md_object *obj,
234 /** \retval number of bytes actually read upon success */
235 int (*moo_readpage)(const struct lu_env *env, struct md_object *obj,
236 const struct lu_rdpg *rdpg);
238 int (*moo_readlink)(const struct lu_env *env, struct md_object *obj,
240 int (*moo_changelog)(const struct lu_env *env,
241 enum changelog_rec_type type, int flags,
242 struct md_object *obj);
243 /** part of cross-ref operation */
244 int (*moo_object_create)(const struct lu_env *env,
245 struct md_object *obj,
246 const struct md_op_spec *spec,
249 int (*moo_ref_add)(const struct lu_env *env,
250 struct md_object *obj,
251 const struct md_attr *ma);
253 int (*moo_ref_del)(const struct lu_env *env,
254 struct md_object *obj,
257 int (*moo_open)(const struct lu_env *env,
258 struct md_object *obj, int flag);
260 int (*moo_close)(const struct lu_env *env, struct md_object *obj,
261 struct md_attr *ma, int mode);
263 int (*moo_capa_get)(const struct lu_env *, struct md_object *,
264 struct lustre_capa *, int renewal);
266 int (*moo_object_sync)(const struct lu_env *, struct md_object *);
267 int (*moo_path)(const struct lu_env *env, struct md_object *obj,
268 char *path, int pathlen, __u64 *recno, int *linkno);
269 int (*moo_file_lock)(const struct lu_env *env, struct md_object *obj,
270 struct lov_mds_md *lmm, struct ldlm_extent *extent,
271 struct lustre_handle *lockh);
272 int (*moo_file_unlock)(const struct lu_env *env, struct md_object *obj,
273 struct lov_mds_md *lmm,
274 struct lustre_handle *lockh);
278 * Operations implemented for each directory object.
280 struct md_dir_operations {
281 int (*mdo_is_subdir) (const struct lu_env *env, struct md_object *obj,
282 const struct lu_fid *fid, struct lu_fid *sfid);
284 int (*mdo_lookup)(const struct lu_env *env, struct md_object *obj,
285 const struct lu_name *lname, struct lu_fid *fid,
286 struct md_op_spec *spec);
288 mdl_mode_t (*mdo_lock_mode)(const struct lu_env *env,
289 struct md_object *obj,
292 int (*mdo_create)(const struct lu_env *env, struct md_object *pobj,
293 const struct lu_name *lname, struct md_object *child,
294 struct md_op_spec *spec,
297 /** This method is used for creating data object for this meta object*/
298 int (*mdo_create_data)(const struct lu_env *env, struct md_object *p,
300 const struct md_op_spec *spec,
303 int (*mdo_rename)(const struct lu_env *env, struct md_object *spobj,
304 struct md_object *tpobj, const struct lu_fid *lf,
305 const struct lu_name *lsname, struct md_object *tobj,
306 const struct lu_name *ltname, struct md_attr *ma);
308 int (*mdo_link)(const struct lu_env *env, struct md_object *tgt_obj,
309 struct md_object *src_obj, const struct lu_name *lname,
312 int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj,
313 struct md_object *cobj, const struct lu_name *lname,
316 /** This method is used to compare a requested layout to an existing
317 * layout (struct lov_mds_md_v1/3 vs struct lov_mds_md_v1/3) */
318 int (*mdo_lum_lmm_cmp)(const struct lu_env *env,
319 struct md_object *cobj,
320 const struct md_op_spec *spec,
323 /** partial ops for cross-ref case */
324 int (*mdo_name_insert)(const struct lu_env *env,
325 struct md_object *obj,
326 const struct lu_name *lname,
327 const struct lu_fid *fid,
328 const struct md_attr *ma);
330 int (*mdo_name_remove)(const struct lu_env *env,
331 struct md_object *obj,
332 const struct lu_name *lname,
333 const struct md_attr *ma);
335 int (*mdo_rename_tgt)(const struct lu_env *env, struct md_object *pobj,
336 struct md_object *tobj, const struct lu_fid *fid,
337 const struct lu_name *lname, struct md_attr *ma);
340 struct md_device_operations {
341 /** meta-data device related handlers. */
342 int (*mdo_root_get)(const struct lu_env *env, struct md_device *m,
345 int (*mdo_maxsize_get)(const struct lu_env *env, struct md_device *m,
346 int *md_size, int *cookie_size);
348 int (*mdo_statfs)(const struct lu_env *env, struct md_device *m,
349 struct obd_statfs *sfs);
351 int (*mdo_init_capa_ctxt)(const struct lu_env *env, struct md_device *m,
352 int mode, unsigned long timeout, __u32 alg,
353 struct lustre_capa_key *keys);
355 int (*mdo_update_capa_key)(const struct lu_env *env,
357 struct lustre_capa_key *key);
359 int (*mdo_llog_ctxt_get)(const struct lu_env *env,
360 struct md_device *m, int idx, void **h);
362 int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m,
363 unsigned int cmd, int len, void *data);
366 enum md_upcall_event {
367 /** Sync the md layer*/
368 MD_LOV_SYNC = (1 << 0),
369 /** Just for split, no need trans, for replay */
370 MD_NO_TRANS = (1 << 1),
371 MD_LOV_CONFIG = (1 << 2),
372 /** Trigger quota recovery */
373 MD_LOV_QUOTA = (1 << 3)
377 /** this lock protects upcall using against its removal
378 * read lock is for usage the upcall, write - for init/fini */
379 struct rw_semaphore mu_upcall_sem;
380 /** device to call, upper layer normally */
381 struct md_device *mu_upcall_dev;
382 /** upcall function */
383 int (*mu_upcall)(const struct lu_env *env, struct md_device *md,
384 enum md_upcall_event ev, void *data);
388 struct lu_device md_lu_dev;
389 const struct md_device_operations *md_ops;
390 struct md_upcall md_upcall;
393 static inline void md_upcall_init(struct md_device *m, void *upcl)
395 init_rwsem(&m->md_upcall.mu_upcall_sem);
396 m->md_upcall.mu_upcall_dev = NULL;
397 m->md_upcall.mu_upcall = upcl;
400 static inline void md_upcall_dev_set(struct md_device *m, struct md_device *up)
402 down_write(&m->md_upcall.mu_upcall_sem);
403 m->md_upcall.mu_upcall_dev = up;
404 up_write(&m->md_upcall.mu_upcall_sem);
407 static inline void md_upcall_fini(struct md_device *m)
409 down_write(&m->md_upcall.mu_upcall_sem);
410 m->md_upcall.mu_upcall_dev = NULL;
411 m->md_upcall.mu_upcall = NULL;
412 up_write(&m->md_upcall.mu_upcall_sem);
415 static inline int md_do_upcall(const struct lu_env *env, struct md_device *m,
416 enum md_upcall_event ev, void *data)
419 down_read(&m->md_upcall.mu_upcall_sem);
420 if (m->md_upcall.mu_upcall_dev != NULL &&
421 m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall != NULL) {
422 rc = m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall(env,
423 m->md_upcall.mu_upcall_dev,
426 up_read(&m->md_upcall.mu_upcall_sem);
431 struct lu_object mo_lu;
432 const struct md_object_operations *mo_ops;
433 const struct md_dir_operations *mo_dir_ops;
440 struct lu_site *ms_lu;
442 * mds number of this site.
446 * Fid location database
448 struct lu_server_fld *ms_server_fld;
449 struct lu_client_fld *ms_client_fld;
454 struct lu_server_seq *ms_server_seq;
457 * Controller Seq Manager
459 struct lu_server_seq *ms_control_seq;
460 struct obd_export *ms_control_exp;
465 struct lu_client_seq *ms_client_seq;
468 static inline int lu_device_is_md(const struct lu_device *d)
470 return ergo(d != NULL, d->ld_type->ldt_tags & LU_DEVICE_MD);
473 static inline struct md_device *lu2md_dev(const struct lu_device *d)
475 LASSERT(IS_ERR(d) || lu_device_is_md(d));
476 return container_of0(d, struct md_device, md_lu_dev);
479 static inline struct lu_device *md2lu_dev(struct md_device *d)
481 return &d->md_lu_dev;
484 static inline struct md_object *lu2md(const struct lu_object *o)
486 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev));
487 return container_of0(o, struct md_object, mo_lu);
490 static inline struct md_object *md_object_next(const struct md_object *obj)
492 return (obj ? lu2md(lu_object_next(&obj->mo_lu)) : NULL);
495 static inline struct md_device *md_obj2dev(const struct md_object *o)
497 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->mo_lu.lo_dev));
498 return container_of0(o->mo_lu.lo_dev, struct md_device, md_lu_dev);
501 static inline struct md_site *lu_site2md(const struct lu_site *s)
503 return s->ld_md_site;
506 static inline int md_device_init(struct md_device *md, struct lu_device_type *t)
508 return lu_device_init(&md->md_lu_dev, t);
511 static inline void md_device_fini(struct md_device *md)
513 lu_device_fini(&md->md_lu_dev);
516 static inline struct md_object *md_object_find_slice(const struct lu_env *env,
517 struct md_device *md,
518 const struct lu_fid *f)
520 return lu2md(lu_object_find_slice(env, md2lu_dev(md), f, NULL));
525 static inline int mo_permission(const struct lu_env *env,
531 LASSERT(c->mo_ops->moo_permission);
532 return c->mo_ops->moo_permission(env, p, c, at, mask);
535 static inline int mo_attr_get(const struct lu_env *env,
539 LASSERT(m->mo_ops->moo_attr_get);
540 return m->mo_ops->moo_attr_get(env, m, at);
543 static inline int mo_readlink(const struct lu_env *env,
547 LASSERT(m->mo_ops->moo_readlink);
548 return m->mo_ops->moo_readlink(env, m, buf);
551 static inline int mo_changelog(const struct lu_env *env,
552 enum changelog_rec_type type,
553 int flags, struct md_object *m)
555 LASSERT(m->mo_ops->moo_changelog);
556 return m->mo_ops->moo_changelog(env, type, flags, m);
559 static inline int mo_attr_set(const struct lu_env *env,
561 const struct md_attr *at)
563 LASSERT(m->mo_ops->moo_attr_set);
564 return m->mo_ops->moo_attr_set(env, m, at);
567 static inline int mo_xattr_get(const struct lu_env *env,
572 LASSERT(m->mo_ops->moo_xattr_get);
573 return m->mo_ops->moo_xattr_get(env, m, buf, name);
576 static inline int mo_xattr_del(const struct lu_env *env,
580 LASSERT(m->mo_ops->moo_xattr_del);
581 return m->mo_ops->moo_xattr_del(env, m, name);
584 static inline int mo_xattr_set(const struct lu_env *env,
586 const struct lu_buf *buf,
590 LASSERT(m->mo_ops->moo_xattr_set);
591 return m->mo_ops->moo_xattr_set(env, m, buf, name, flags);
594 static inline int mo_xattr_list(const struct lu_env *env,
598 LASSERT(m->mo_ops->moo_xattr_list);
599 return m->mo_ops->moo_xattr_list(env, m, buf);
602 static inline int mo_open(const struct lu_env *env,
606 LASSERT(m->mo_ops->moo_open);
607 return m->mo_ops->moo_open(env, m, flags);
610 static inline int mo_close(const struct lu_env *env,
615 LASSERT(m->mo_ops->moo_close);
616 return m->mo_ops->moo_close(env, m, ma, mode);
619 static inline int mo_readpage(const struct lu_env *env,
621 const struct lu_rdpg *rdpg)
623 LASSERT(m->mo_ops->moo_readpage);
624 return m->mo_ops->moo_readpage(env, m, rdpg);
627 static inline int mo_object_create(const struct lu_env *env,
629 const struct md_op_spec *spc,
632 LASSERT(m->mo_ops->moo_object_create);
633 return m->mo_ops->moo_object_create(env, m, spc, at);
636 static inline int mo_ref_add(const struct lu_env *env,
638 const struct md_attr *ma)
640 LASSERT(m->mo_ops->moo_ref_add);
641 return m->mo_ops->moo_ref_add(env, m, ma);
644 static inline int mo_ref_del(const struct lu_env *env,
648 LASSERT(m->mo_ops->moo_ref_del);
649 return m->mo_ops->moo_ref_del(env, m, ma);
652 static inline int mo_capa_get(const struct lu_env *env,
654 struct lustre_capa *c,
657 LASSERT(m->mo_ops->moo_capa_get);
658 return m->mo_ops->moo_capa_get(env, m, c, renewal);
661 static inline int mo_path(const struct lu_env *env, struct md_object *m,
662 char *path, int pathlen, __u64 *recno, int *linkno)
664 if (m->mo_ops->moo_path == NULL)
666 return m->mo_ops->moo_path(env, m, path, pathlen, recno, linkno);
669 static inline int mo_object_sync(const struct lu_env *env, struct md_object *m)
671 LASSERT(m->mo_ops->moo_object_sync);
672 return m->mo_ops->moo_object_sync(env, m);
675 static inline int mo_file_lock(const struct lu_env *env, struct md_object *m,
676 struct lov_mds_md *lmm,
677 struct ldlm_extent *extent,
678 struct lustre_handle *lockh)
680 LASSERT(m->mo_ops->moo_file_lock);
681 return m->mo_ops->moo_file_lock(env, m, lmm, extent, lockh);
684 static inline int mo_file_unlock(const struct lu_env *env, struct md_object *m,
685 struct lov_mds_md *lmm,
686 struct lustre_handle *lockh)
688 LASSERT(m->mo_ops->moo_file_unlock);
689 return m->mo_ops->moo_file_unlock(env, m, lmm, lockh);
692 static inline int mdo_lookup(const struct lu_env *env,
694 const struct lu_name *lname,
696 struct md_op_spec *spec)
698 LASSERT(p->mo_dir_ops->mdo_lookup);
699 return p->mo_dir_ops->mdo_lookup(env, p, lname, f, spec);
702 static inline mdl_mode_t mdo_lock_mode(const struct lu_env *env,
703 struct md_object *mo,
706 if (mo->mo_dir_ops->mdo_lock_mode == NULL)
708 return mo->mo_dir_ops->mdo_lock_mode(env, mo, lm);
711 static inline int mdo_create(const struct lu_env *env,
713 const struct lu_name *lchild_name,
715 struct md_op_spec *spc,
718 LASSERT(p->mo_dir_ops->mdo_create);
719 return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at);
722 static inline int mdo_create_data(const struct lu_env *env,
725 const struct md_op_spec *spec,
728 LASSERT(c->mo_dir_ops->mdo_create_data);
729 return c->mo_dir_ops->mdo_create_data(env, p, c, spec, ma);
732 static inline int mdo_rename(const struct lu_env *env,
733 struct md_object *sp,
734 struct md_object *tp,
735 const struct lu_fid *lf,
736 const struct lu_name *lsname,
738 const struct lu_name *ltname,
741 LASSERT(tp->mo_dir_ops->mdo_rename);
742 return tp->mo_dir_ops->mdo_rename(env, sp, tp, lf, lsname, t, ltname,
746 static inline int mdo_is_subdir(const struct lu_env *env,
747 struct md_object *mo,
748 const struct lu_fid *fid,
751 LASSERT(mo->mo_dir_ops->mdo_is_subdir);
752 return mo->mo_dir_ops->mdo_is_subdir(env, mo, fid, sfid);
755 static inline int mdo_link(const struct lu_env *env,
758 const struct lu_name *lname,
761 LASSERT(s->mo_dir_ops->mdo_link);
762 return s->mo_dir_ops->mdo_link(env, p, s, lname, ma);
765 static inline int mdo_unlink(const struct lu_env *env,
768 const struct lu_name *lname,
771 LASSERT(c->mo_dir_ops->mdo_unlink);
772 return c->mo_dir_ops->mdo_unlink(env, p, c, lname, ma);
775 static inline int mdo_lum_lmm_cmp(const struct lu_env *env,
777 const struct md_op_spec *spec,
780 LASSERT(c->mo_dir_ops->mdo_lum_lmm_cmp);
781 return c->mo_dir_ops->mdo_lum_lmm_cmp(env, c, spec, ma);
784 static inline int mdo_name_insert(const struct lu_env *env,
786 const struct lu_name *lname,
787 const struct lu_fid *f,
788 const struct md_attr *ma)
790 LASSERT(p->mo_dir_ops->mdo_name_insert);
791 return p->mo_dir_ops->mdo_name_insert(env, p, lname, f, ma);
794 static inline int mdo_name_remove(const struct lu_env *env,
796 const struct lu_name *lname,
797 const struct md_attr *ma)
799 LASSERT(p->mo_dir_ops->mdo_name_remove);
800 return p->mo_dir_ops->mdo_name_remove(env, p, lname, ma);
803 static inline int mdo_rename_tgt(const struct lu_env *env,
806 const struct lu_fid *lf,
807 const struct lu_name *lname,
811 LASSERT(t->mo_dir_ops->mdo_rename_tgt);
812 return t->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
814 LASSERT(p->mo_dir_ops->mdo_rename_tgt);
815 return p->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
821 * Structure to hold object information. This is used to create object
822 * \pre llod_dir exist
824 struct lu_local_obj_desc {
825 const char *llod_dir;
826 const char *llod_name;
829 const struct dt_index_features *llod_feat;
830 cfs_list_t llod_linkage;
833 struct md_object *llo_store_resolve(const struct lu_env *env,
834 struct md_device *md,
835 struct dt_device *dt,
839 struct md_object *llo_store_open(const struct lu_env *env,
840 struct md_device *md,
841 struct dt_device *dt,
846 struct md_object *llo_store_create_index(const struct lu_env *env,
847 struct md_device *md,
848 struct dt_device *dt,
851 const struct lu_fid *fid,
852 const struct dt_index_features *feat);
854 struct md_object *llo_store_create(const struct lu_env *env,
855 struct md_device *md,
856 struct dt_device *dt,
859 const struct lu_fid *fid);
861 void llo_local_obj_register(struct lu_local_obj_desc *);
862 void llo_local_obj_unregister(struct lu_local_obj_desc *);
864 int llo_local_objects_setup(const struct lu_env *env,
865 struct md_device * md,
866 struct dt_device * dt);
868 int lustre_buf2som(void *buf, int rc, struct md_som_data *msd);
869 int lustre_buf2hsm(void *buf, int rc, struct md_hsm *mh);
870 void lustre_hsm2buf(void *buf, struct md_hsm *mh);
872 #endif /* _LINUX_MD_OBJECT_H */