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, 2013, Intel Corporation.
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 struct md_device_operations;
67 struct obd_export *mq_exp;
70 struct md_quota *md_quota(const struct lu_env *env);
72 /** metadata attributes */
79 MA_ACL_DEF = (1 << 5),
80 MA_LOV_DEF = (1 << 6),
81 MA_LAY_GEN = (1 << 7),
101 MDT_REG_LOCK = (1 << 0),
102 MDT_PDO_LOCK = (1 << 1)
105 /* memory structure for hsm attributes
106 * for fields description see the on disk structure hsm_attrs
107 * which is defined in lustre_idl.h
116 #define IOEPOCH_INVAL 0
118 /* memory structure for som attributes
119 * for fields description see the on disk structure som_attrs
120 * which is defined in lustre_idl.h
135 struct lu_attr ma_attr;
136 struct lu_fid ma_pfid;
137 struct md_hsm ma_hsm;
138 struct lov_mds_md *ma_lmm;
139 struct lmv_stripe_md *ma_lmv;
141 struct lustre_capa *ma_capa;
142 struct md_som_data *ma_som;
149 /** Additional parameters for create */
152 /** symlink target */
153 const char *sp_symname;
154 /** parent FID for cross-ref mkdir */
155 const struct lu_fid *sp_pfid;
156 /** eadata for regular files */
158 /** lov objs exist already */
159 const struct lu_fid *fid;
165 /** Create flag from client: such as MDS_OPEN_CREAT, and others. */
168 /** don't create lov objects or llog cookie - this replay */
169 unsigned int no_create:1,
170 sp_cr_lookup:1, /* do lookup sanity check or not. */
171 sp_rm_entry:1; /* only remove name entry */
173 /** Current lock mode for parent dir where create is performing. */
174 mdl_mode_t sp_cr_mode;
176 /** to create directory */
177 const struct dt_index_features *sp_feat;
181 * Operations implemented for each md object (both directory and leaf).
183 struct md_object_operations {
184 int (*moo_permission)(const struct lu_env *env,
185 struct md_object *pobj, struct md_object *cobj,
186 struct md_attr *attr, int mask);
188 int (*moo_attr_get)(const struct lu_env *env, struct md_object *obj,
189 struct md_attr *attr);
191 int (*moo_attr_set)(const struct lu_env *env, struct md_object *obj,
192 const struct md_attr *attr);
194 int (*moo_xattr_get)(const struct lu_env *env, struct md_object *obj,
195 struct lu_buf *buf, const char *name);
197 int (*moo_xattr_list)(const struct lu_env *env, struct md_object *obj,
200 int (*moo_xattr_set)(const struct lu_env *env, struct md_object *obj,
201 const struct lu_buf *buf, const char *name,
204 int (*moo_xattr_del)(const struct lu_env *env, struct md_object *obj,
207 /** This method is used to swap the layouts between 2 objects */
208 int (*moo_swap_layouts)(const struct lu_env *env,
209 struct md_object *obj1, struct md_object *obj2,
212 /** \retval number of bytes actually read upon success */
213 int (*moo_readpage)(const struct lu_env *env, struct md_object *obj,
214 const struct lu_rdpg *rdpg);
216 int (*moo_readlink)(const struct lu_env *env, struct md_object *obj,
218 int (*moo_changelog)(const struct lu_env *env,
219 enum changelog_rec_type type, int flags,
220 struct md_object *obj);
221 /** part of cross-ref operation */
222 int (*moo_object_create)(const struct lu_env *env,
223 struct md_object *obj,
224 const struct md_op_spec *spec,
227 int (*moo_ref_add)(const struct lu_env *env,
228 struct md_object *obj,
229 const struct md_attr *ma);
231 int (*moo_ref_del)(const struct lu_env *env,
232 struct md_object *obj,
235 int (*moo_open)(const struct lu_env *env,
236 struct md_object *obj, int flag);
238 int (*moo_close)(const struct lu_env *env, struct md_object *obj,
239 struct md_attr *ma, int mode);
241 int (*moo_capa_get)(const struct lu_env *, struct md_object *,
242 struct lustre_capa *, int renewal);
244 int (*moo_object_sync)(const struct lu_env *, struct md_object *);
246 int (*moo_file_lock)(const struct lu_env *env, struct md_object *obj,
247 struct lov_mds_md *lmm, struct ldlm_extent *extent,
248 struct lustre_handle *lockh);
249 int (*moo_file_unlock)(const struct lu_env *env, struct md_object *obj,
250 struct lov_mds_md *lmm,
251 struct lustre_handle *lockh);
252 int (*moo_object_lock)(const struct lu_env *env, struct md_object *obj,
253 struct lustre_handle *lh,
254 struct ldlm_enqueue_info *einfo,
259 * Operations implemented for each directory object.
261 struct md_dir_operations {
262 int (*mdo_is_subdir) (const struct lu_env *env, struct md_object *obj,
263 const struct lu_fid *fid, struct lu_fid *sfid);
265 int (*mdo_lookup)(const struct lu_env *env, struct md_object *obj,
266 const struct lu_name *lname, struct lu_fid *fid,
267 struct md_op_spec *spec);
269 mdl_mode_t (*mdo_lock_mode)(const struct lu_env *env,
270 struct md_object *obj,
273 int (*mdo_create)(const struct lu_env *env, struct md_object *pobj,
274 const struct lu_name *lname, struct md_object *child,
275 struct md_op_spec *spec,
278 /** This method is used for creating data object for this meta object*/
279 int (*mdo_create_data)(const struct lu_env *env, struct md_object *p,
281 const struct md_op_spec *spec,
284 int (*mdo_rename)(const struct lu_env *env, struct md_object *spobj,
285 struct md_object *tpobj, const struct lu_fid *lf,
286 const struct lu_name *lsname, struct md_object *tobj,
287 const struct lu_name *ltname, struct md_attr *ma);
289 int (*mdo_link)(const struct lu_env *env, struct md_object *tgt_obj,
290 struct md_object *src_obj, const struct lu_name *lname,
293 int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj,
294 struct md_object *cobj, const struct lu_name *lname,
295 struct md_attr *ma, int no_name);
297 /** This method is used to compare a requested layout to an existing
298 * layout (struct lov_mds_md_v1/3 vs struct lov_mds_md_v1/3) */
299 int (*mdo_lum_lmm_cmp)(const struct lu_env *env,
300 struct md_object *cobj,
301 const struct md_op_spec *spec,
304 /** partial ops for cross-ref case */
305 int (*mdo_name_insert)(const struct lu_env *env,
306 struct md_object *obj,
307 const struct lu_name *lname,
308 const struct lu_fid *fid,
309 const struct md_attr *ma);
311 int (*mdo_name_remove)(const struct lu_env *env,
312 struct md_object *obj,
313 const struct lu_name *lname,
314 const struct md_attr *ma);
316 int (*mdo_rename_tgt)(const struct lu_env *env, struct md_object *pobj,
317 struct md_object *tobj, const struct lu_fid *fid,
318 const struct lu_name *lname, struct md_attr *ma);
321 struct md_device_operations {
322 /** meta-data device related handlers. */
323 int (*mdo_root_get)(const struct lu_env *env, struct md_device *m,
326 int (*mdo_maxeasize_get)(const struct lu_env *env, struct md_device *m,
329 int (*mdo_statfs)(const struct lu_env *env, struct md_device *m,
330 struct obd_statfs *sfs);
332 int (*mdo_init_capa_ctxt)(const struct lu_env *env, struct md_device *m,
333 int mode, unsigned long timeout, __u32 alg,
334 struct lustre_capa_key *keys);
336 int (*mdo_update_capa_key)(const struct lu_env *env,
338 struct lustre_capa_key *key);
340 int (*mdo_llog_ctxt_get)(const struct lu_env *env,
341 struct md_device *m, int idx, void **h);
343 int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m,
344 unsigned int cmd, int len, void *data);
347 enum md_upcall_event {
348 /** Sync the md layer*/
349 MD_LOV_SYNC = (1 << 0),
350 /** Just for split, no need trans, for replay */
351 MD_NO_TRANS = (1 << 1),
352 MD_LOV_CONFIG = (1 << 2),
353 /** Trigger quota recovery */
354 MD_LOV_QUOTA = (1 << 3)
358 /** this lock protects upcall using against its removal
359 * read lock is for usage the upcall, write - for init/fini */
360 struct rw_semaphore mu_upcall_sem;
361 /** device to call, upper layer normally */
362 struct md_device *mu_upcall_dev;
363 /** upcall function */
364 int (*mu_upcall)(const struct lu_env *env, struct md_device *md,
365 enum md_upcall_event ev, void *data);
369 struct lu_device md_lu_dev;
370 const struct md_device_operations *md_ops;
371 struct md_upcall md_upcall;
374 static inline void md_upcall_init(struct md_device *m, void *upcl)
376 init_rwsem(&m->md_upcall.mu_upcall_sem);
377 m->md_upcall.mu_upcall_dev = NULL;
378 m->md_upcall.mu_upcall = upcl;
381 static inline void md_upcall_dev_set(struct md_device *m, struct md_device *up)
383 down_write(&m->md_upcall.mu_upcall_sem);
384 m->md_upcall.mu_upcall_dev = up;
385 up_write(&m->md_upcall.mu_upcall_sem);
388 static inline void md_upcall_fini(struct md_device *m)
390 down_write(&m->md_upcall.mu_upcall_sem);
391 m->md_upcall.mu_upcall_dev = NULL;
392 m->md_upcall.mu_upcall = NULL;
393 up_write(&m->md_upcall.mu_upcall_sem);
396 static inline int md_do_upcall(const struct lu_env *env, struct md_device *m,
397 enum md_upcall_event ev, void *data)
400 down_read(&m->md_upcall.mu_upcall_sem);
401 if (m->md_upcall.mu_upcall_dev != NULL &&
402 m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall != NULL) {
403 rc = m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall(env,
404 m->md_upcall.mu_upcall_dev,
407 up_read(&m->md_upcall.mu_upcall_sem);
412 struct lu_object mo_lu;
413 const struct md_object_operations *mo_ops;
414 const struct md_dir_operations *mo_dir_ops;
420 struct seq_server_site {
421 struct lu_site *ss_lu;
423 * mds number of this site.
427 * Fid location database
429 struct lu_server_fld *ss_server_fld;
430 struct lu_client_fld *ss_client_fld;
435 struct lu_server_seq *ss_server_seq;
438 * Controller Seq Manager
440 struct lu_server_seq *ss_control_seq;
441 struct obd_export *ss_control_exp;
446 struct lu_client_seq *ss_client_seq;
449 static inline struct md_device *lu2md_dev(const struct lu_device *d)
451 LASSERT(IS_ERR(d) || lu_device_is_md(d));
452 return container_of0(d, struct md_device, md_lu_dev);
455 static inline struct lu_device *md2lu_dev(struct md_device *d)
457 return &d->md_lu_dev;
460 static inline struct md_object *lu2md(const struct lu_object *o)
462 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev));
463 return container_of0(o, struct md_object, mo_lu);
466 static inline struct md_object *md_object_next(const struct md_object *obj)
468 return (obj ? lu2md(lu_object_next(&obj->mo_lu)) : NULL);
471 static inline int md_device_init(struct md_device *md, struct lu_device_type *t)
473 return lu_device_init(&md->md_lu_dev, t);
476 static inline void md_device_fini(struct md_device *md)
478 lu_device_fini(&md->md_lu_dev);
481 static inline struct md_object *md_object_find_slice(const struct lu_env *env,
482 struct md_device *md,
483 const struct lu_fid *f)
485 return lu2md(lu_object_find_slice(env, md2lu_dev(md), f, NULL));
490 static inline int mo_permission(const struct lu_env *env,
496 LASSERT(c->mo_ops->moo_permission);
497 return c->mo_ops->moo_permission(env, p, c, at, mask);
500 static inline int mo_attr_get(const struct lu_env *env,
504 LASSERT(m->mo_ops->moo_attr_get);
505 return m->mo_ops->moo_attr_get(env, m, at);
508 static inline int mo_readlink(const struct lu_env *env,
512 LASSERT(m->mo_ops->moo_readlink);
513 return m->mo_ops->moo_readlink(env, m, buf);
516 static inline int mo_changelog(const struct lu_env *env,
517 enum changelog_rec_type type,
518 int flags, struct md_object *m)
520 LASSERT(m->mo_ops->moo_changelog);
521 return m->mo_ops->moo_changelog(env, type, flags, m);
524 static inline int mo_attr_set(const struct lu_env *env,
526 const struct md_attr *at)
528 LASSERT(m->mo_ops->moo_attr_set);
529 return m->mo_ops->moo_attr_set(env, m, at);
532 static inline int mo_xattr_get(const struct lu_env *env,
537 LASSERT(m->mo_ops->moo_xattr_get);
538 return m->mo_ops->moo_xattr_get(env, m, buf, name);
541 static inline int mo_xattr_del(const struct lu_env *env,
545 LASSERT(m->mo_ops->moo_xattr_del);
546 return m->mo_ops->moo_xattr_del(env, m, name);
549 static inline int mo_xattr_set(const struct lu_env *env,
551 const struct lu_buf *buf,
555 LASSERT(m->mo_ops->moo_xattr_set);
556 return m->mo_ops->moo_xattr_set(env, m, buf, name, flags);
559 static inline int mo_xattr_list(const struct lu_env *env,
563 LASSERT(m->mo_ops->moo_xattr_list);
564 return m->mo_ops->moo_xattr_list(env, m, buf);
567 static inline int mo_swap_layouts(const struct lu_env *env,
568 struct md_object *o1,
569 struct md_object *o2, __u64 flags)
571 LASSERT(o1->mo_ops->moo_swap_layouts);
572 LASSERT(o2->mo_ops->moo_swap_layouts);
573 if (o1->mo_ops->moo_swap_layouts != o2->mo_ops->moo_swap_layouts)
575 return o1->mo_ops->moo_swap_layouts(env, o1, o2, flags);
578 static inline int mo_open(const struct lu_env *env,
582 LASSERT(m->mo_ops->moo_open);
583 return m->mo_ops->moo_open(env, m, flags);
586 static inline int mo_close(const struct lu_env *env,
591 LASSERT(m->mo_ops->moo_close);
592 return m->mo_ops->moo_close(env, m, ma, mode);
595 static inline int mo_readpage(const struct lu_env *env,
597 const struct lu_rdpg *rdpg)
599 LASSERT(m->mo_ops->moo_readpage);
600 return m->mo_ops->moo_readpage(env, m, rdpg);
603 static inline int mo_object_create(const struct lu_env *env,
605 const struct md_op_spec *spc,
608 LASSERT(m->mo_ops->moo_object_create);
609 return m->mo_ops->moo_object_create(env, m, spc, at);
612 static inline int mo_ref_add(const struct lu_env *env,
614 const struct md_attr *ma)
616 LASSERT(m->mo_ops->moo_ref_add);
617 return m->mo_ops->moo_ref_add(env, m, ma);
620 static inline int mo_ref_del(const struct lu_env *env,
624 LASSERT(m->mo_ops->moo_ref_del);
625 return m->mo_ops->moo_ref_del(env, m, ma);
628 static inline int mo_capa_get(const struct lu_env *env,
630 struct lustre_capa *c,
633 LASSERT(m->mo_ops->moo_capa_get);
634 return m->mo_ops->moo_capa_get(env, m, c, renewal);
637 static inline int mo_object_sync(const struct lu_env *env, struct md_object *m)
639 LASSERT(m->mo_ops->moo_object_sync);
640 return m->mo_ops->moo_object_sync(env, m);
643 static inline int mo_file_lock(const struct lu_env *env, struct md_object *m,
644 struct lov_mds_md *lmm,
645 struct ldlm_extent *extent,
646 struct lustre_handle *lockh)
648 LASSERT(m->mo_ops->moo_file_lock);
649 return m->mo_ops->moo_file_lock(env, m, lmm, extent, lockh);
652 static inline int mo_file_unlock(const struct lu_env *env, struct md_object *m,
653 struct lov_mds_md *lmm,
654 struct lustre_handle *lockh)
656 LASSERT(m->mo_ops->moo_file_unlock);
657 return m->mo_ops->moo_file_unlock(env, m, lmm, lockh);
660 static inline int mo_object_lock(const struct lu_env *env,
662 struct lustre_handle *lh,
663 struct ldlm_enqueue_info *einfo,
666 LASSERT(m->mo_ops->moo_object_lock);
667 return m->mo_ops->moo_object_lock(env, m, lh, einfo, policy);
670 static inline int mdo_lookup(const struct lu_env *env,
672 const struct lu_name *lname,
674 struct md_op_spec *spec)
676 LASSERT(p->mo_dir_ops->mdo_lookup);
677 return p->mo_dir_ops->mdo_lookup(env, p, lname, f, spec);
680 static inline mdl_mode_t mdo_lock_mode(const struct lu_env *env,
681 struct md_object *mo,
684 if (mo->mo_dir_ops->mdo_lock_mode == NULL)
686 return mo->mo_dir_ops->mdo_lock_mode(env, mo, lm);
689 static inline int mdo_create(const struct lu_env *env,
691 const struct lu_name *lchild_name,
693 struct md_op_spec *spc,
696 LASSERT(p->mo_dir_ops->mdo_create);
697 return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at);
700 static inline int mdo_create_data(const struct lu_env *env,
703 const struct md_op_spec *spec,
706 LASSERT(c->mo_dir_ops->mdo_create_data);
707 return c->mo_dir_ops->mdo_create_data(env, p, c, spec, ma);
710 static inline int mdo_rename(const struct lu_env *env,
711 struct md_object *sp,
712 struct md_object *tp,
713 const struct lu_fid *lf,
714 const struct lu_name *lsname,
716 const struct lu_name *ltname,
719 LASSERT(tp->mo_dir_ops->mdo_rename);
720 return tp->mo_dir_ops->mdo_rename(env, sp, tp, lf, lsname, t, ltname,
724 static inline int mdo_is_subdir(const struct lu_env *env,
725 struct md_object *mo,
726 const struct lu_fid *fid,
729 LASSERT(mo->mo_dir_ops->mdo_is_subdir);
730 return mo->mo_dir_ops->mdo_is_subdir(env, mo, fid, sfid);
733 static inline int mdo_link(const struct lu_env *env,
736 const struct lu_name *lname,
739 LASSERT(s->mo_dir_ops->mdo_link);
740 return s->mo_dir_ops->mdo_link(env, p, s, lname, ma);
743 static inline int mdo_unlink(const struct lu_env *env,
746 const struct lu_name *lname,
747 struct md_attr *ma, int no_name)
749 LASSERT(p->mo_dir_ops->mdo_unlink);
750 return p->mo_dir_ops->mdo_unlink(env, p, c, lname, ma, no_name);
753 static inline int mdo_lum_lmm_cmp(const struct lu_env *env,
755 const struct md_op_spec *spec,
758 LASSERT(c->mo_dir_ops->mdo_lum_lmm_cmp);
759 return c->mo_dir_ops->mdo_lum_lmm_cmp(env, c, spec, ma);
762 static inline int mdo_name_insert(const struct lu_env *env,
764 const struct lu_name *lname,
765 const struct lu_fid *f,
766 const struct md_attr *ma)
768 LASSERT(p->mo_dir_ops->mdo_name_insert);
769 return p->mo_dir_ops->mdo_name_insert(env, p, lname, f, ma);
772 static inline int mdo_name_remove(const struct lu_env *env,
774 const struct lu_name *lname,
775 const struct md_attr *ma)
777 LASSERT(p->mo_dir_ops->mdo_name_remove);
778 return p->mo_dir_ops->mdo_name_remove(env, p, lname, ma);
781 static inline int mdo_rename_tgt(const struct lu_env *env,
784 const struct lu_fid *lf,
785 const struct lu_name *lname,
789 LASSERT(t->mo_dir_ops->mdo_rename_tgt);
790 return t->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
792 LASSERT(p->mo_dir_ops->mdo_rename_tgt);
793 return p->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
798 * Used in MDD/OUT layer for object lock rule
800 enum mdd_object_role {
810 * Structure to hold object information. This is used to create object
811 * \pre llod_dir exist
813 struct lu_local_obj_desc {
814 const char *llod_dir;
815 const char *llod_name;
818 const struct dt_index_features *llod_feat;
819 cfs_list_t llod_linkage;
822 int lustre_buf2som(void *buf, int rc, struct md_som_data *msd);
823 int lustre_buf2hsm(void *buf, int rc, struct md_hsm *mh);
824 void lustre_hsm2buf(void *buf, const struct md_hsm *mh);
843 __u32 uc_suppgids[2];
846 struct group_info *uc_ginfo;
847 struct md_identity *uc_identity;
850 struct lu_ucred *lu_ucred(const struct lu_env *env);
852 struct lu_ucred *lu_ucred_check(const struct lu_env *env);
854 struct lu_ucred *lu_ucred_assert(const struct lu_env *env);
856 int lu_ucred_global_init(void);
858 void lu_ucred_global_fini(void);
860 #define md_cap_t(x) (x)
862 #define MD_CAP_TO_MASK(x) (1 << (x))
864 #define md_cap_raised(c, flag) (md_cap_t(c) & MD_CAP_TO_MASK(flag))
866 /* capable() is copied from linux kernel! */
867 static inline int md_capable(struct lu_ucred *uc, cfs_cap_t cap)
869 if (md_cap_raised(uc->uc_cap, cap))
875 #endif /* _LINUX_MD_OBJECT_H */