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),
85 MA_LMV_DEF = (1 << 11)
102 MDT_REG_LOCK = (1 << 0),
103 MDT_PDO_LOCK = (1 << 1)
106 /* memory structure for hsm attributes
107 * for fields description see the on disk structure hsm_attrs
108 * which is defined in lustre_idl.h
117 #define IOEPOCH_INVAL 0
119 /* memory structure for som attributes
120 * for fields description see the on disk structure som_attrs
121 * which is defined in lustre_idl.h
136 struct lu_attr ma_attr;
137 struct lu_fid ma_pfid;
138 struct md_hsm ma_hsm;
139 struct lov_mds_md *ma_lmm;
140 union lmv_mds_md *ma_lmv;
142 struct llog_cookie *ma_cookie;
143 struct lustre_capa *ma_capa;
144 struct md_som_data *ma_som;
152 /** Additional parameters for create */
155 /** symlink target */
156 const char *sp_symname;
157 /** parent FID for cross-ref mkdir */
158 const struct lu_fid *sp_pfid;
159 /** eadata for regular files */
161 /** lov objs exist already */
162 const struct lu_fid *fid;
168 /** Create flag from client: such as MDS_OPEN_CREAT, and others. */
171 /** don't create lov objects or llog cookie - this replay */
172 unsigned int no_create:1,
173 sp_cr_lookup:1, /* do lookup sanity check or not. */
174 sp_rm_entry:1; /* only remove name entry */
176 /** Current lock mode for parent dir where create is performing. */
177 mdl_mode_t sp_cr_mode;
179 /** to create directory */
180 const struct dt_index_features *sp_feat;
184 * Operations implemented for each md object (both directory and leaf).
186 struct md_object_operations {
187 int (*moo_permission)(const struct lu_env *env,
188 struct md_object *pobj, struct md_object *cobj,
189 struct md_attr *attr, int mask);
191 int (*moo_attr_get)(const struct lu_env *env, struct md_object *obj,
192 struct md_attr *attr);
194 int (*moo_attr_set)(const struct lu_env *env, struct md_object *obj,
195 const struct md_attr *attr);
197 int (*moo_xattr_get)(const struct lu_env *env, struct md_object *obj,
198 struct lu_buf *buf, const char *name);
200 int (*moo_xattr_list)(const struct lu_env *env, struct md_object *obj,
203 int (*moo_xattr_set)(const struct lu_env *env, struct md_object *obj,
204 const struct lu_buf *buf, const char *name,
207 int (*moo_xattr_del)(const struct lu_env *env, struct md_object *obj,
210 /** This method is used to swap the layouts between 2 objects */
211 int (*moo_swap_layouts)(const struct lu_env *env,
212 struct md_object *obj1, struct md_object *obj2,
215 /** \retval number of bytes actually read upon success */
216 int (*moo_readpage)(const struct lu_env *env, struct md_object *obj,
217 const struct lu_rdpg *rdpg);
219 int (*moo_readlink)(const struct lu_env *env, struct md_object *obj,
221 int (*moo_changelog)(const struct lu_env *env,
222 enum changelog_rec_type type, int flags,
223 struct md_object *obj);
224 /** part of cross-ref operation */
225 int (*moo_object_create)(const struct lu_env *env,
226 struct md_object *obj,
227 const struct md_op_spec *spec,
230 int (*moo_ref_add)(const struct lu_env *env,
231 struct md_object *obj,
232 const struct md_attr *ma);
234 int (*moo_ref_del)(const struct lu_env *env,
235 struct md_object *obj,
238 int (*moo_open)(const struct lu_env *env,
239 struct md_object *obj, int flag);
241 int (*moo_close)(const struct lu_env *env, struct md_object *obj,
242 struct md_attr *ma, int mode);
244 int (*moo_capa_get)(const struct lu_env *, struct md_object *,
245 struct lustre_capa *, int renewal);
247 int (*moo_object_sync)(const struct lu_env *, struct md_object *);
249 int (*moo_file_lock)(const struct lu_env *env, struct md_object *obj,
250 struct lov_mds_md *lmm, struct ldlm_extent *extent,
251 struct lustre_handle *lockh);
252 int (*moo_file_unlock)(const struct lu_env *env, struct md_object *obj,
253 struct lov_mds_md *lmm,
254 struct lustre_handle *lockh);
255 int (*moo_object_lock)(const struct lu_env *env, struct md_object *obj,
256 struct lustre_handle *lh,
257 struct ldlm_enqueue_info *einfo,
262 * Operations implemented for each directory object.
264 struct md_dir_operations {
265 int (*mdo_is_subdir) (const struct lu_env *env, struct md_object *obj,
266 const struct lu_fid *fid, struct lu_fid *sfid);
268 int (*mdo_lookup)(const struct lu_env *env, struct md_object *obj,
269 const struct lu_name *lname, struct lu_fid *fid,
270 struct md_op_spec *spec);
272 mdl_mode_t (*mdo_lock_mode)(const struct lu_env *env,
273 struct md_object *obj,
276 int (*mdo_create)(const struct lu_env *env, struct md_object *pobj,
277 const struct lu_name *lname, struct md_object *child,
278 struct md_op_spec *spec,
281 /** This method is used for creating data object for this meta object*/
282 int (*mdo_create_data)(const struct lu_env *env, struct md_object *p,
284 const struct md_op_spec *spec,
287 int (*mdo_rename)(const struct lu_env *env, struct md_object *spobj,
288 struct md_object *tpobj, const struct lu_fid *lf,
289 const struct lu_name *lsname, struct md_object *tobj,
290 const struct lu_name *ltname, struct md_attr *ma);
292 int (*mdo_link)(const struct lu_env *env, struct md_object *tgt_obj,
293 struct md_object *src_obj, const struct lu_name *lname,
296 int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj,
297 struct md_object *cobj, const struct lu_name *lname,
298 struct md_attr *ma, int no_name);
300 /** This method is used to compare a requested layout to an existing
301 * layout (struct lov_mds_md_v1/3 vs struct lov_mds_md_v1/3) */
302 int (*mdo_lum_lmm_cmp)(const struct lu_env *env,
303 struct md_object *cobj,
304 const struct md_op_spec *spec,
307 /** partial ops for cross-ref case */
308 int (*mdo_name_insert)(const struct lu_env *env,
309 struct md_object *obj,
310 const struct lu_name *lname,
311 const struct lu_fid *fid,
312 const struct md_attr *ma);
314 int (*mdo_name_remove)(const struct lu_env *env,
315 struct md_object *obj,
316 const struct lu_name *lname,
317 const struct md_attr *ma);
319 int (*mdo_rename_tgt)(const struct lu_env *env, struct md_object *pobj,
320 struct md_object *tobj, const struct lu_fid *fid,
321 const struct lu_name *lname, struct md_attr *ma);
324 struct md_device_operations {
325 /** meta-data device related handlers. */
326 int (*mdo_root_get)(const struct lu_env *env, struct md_device *m,
329 int (*mdo_maxeasize_get)(const struct lu_env *env, struct md_device *m,
332 int (*mdo_statfs)(const struct lu_env *env, struct md_device *m,
333 struct obd_statfs *sfs);
335 int (*mdo_init_capa_ctxt)(const struct lu_env *env, struct md_device *m,
336 int mode, unsigned long timeout, __u32 alg,
337 struct lustre_capa_key *keys);
339 int (*mdo_update_capa_key)(const struct lu_env *env,
341 struct lustre_capa_key *key);
343 int (*mdo_llog_ctxt_get)(const struct lu_env *env,
344 struct md_device *m, int idx, void **h);
346 int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m,
347 unsigned int cmd, int len, void *data);
350 enum md_upcall_event {
351 /** Sync the md layer*/
352 MD_LOV_SYNC = (1 << 0),
353 /** Just for split, no need trans, for replay */
354 MD_NO_TRANS = (1 << 1),
355 MD_LOV_CONFIG = (1 << 2),
356 /** Trigger quota recovery */
357 MD_LOV_QUOTA = (1 << 3)
361 /** this lock protects upcall using against its removal
362 * read lock is for usage the upcall, write - for init/fini */
363 struct rw_semaphore mu_upcall_sem;
364 /** device to call, upper layer normally */
365 struct md_device *mu_upcall_dev;
366 /** upcall function */
367 int (*mu_upcall)(const struct lu_env *env, struct md_device *md,
368 enum md_upcall_event ev, void *data);
372 struct lu_device md_lu_dev;
373 const struct md_device_operations *md_ops;
374 struct md_upcall md_upcall;
377 static inline void md_upcall_init(struct md_device *m, void *upcl)
379 init_rwsem(&m->md_upcall.mu_upcall_sem);
380 m->md_upcall.mu_upcall_dev = NULL;
381 m->md_upcall.mu_upcall = upcl;
384 static inline void md_upcall_dev_set(struct md_device *m, struct md_device *up)
386 down_write(&m->md_upcall.mu_upcall_sem);
387 m->md_upcall.mu_upcall_dev = up;
388 up_write(&m->md_upcall.mu_upcall_sem);
391 static inline void md_upcall_fini(struct md_device *m)
393 down_write(&m->md_upcall.mu_upcall_sem);
394 m->md_upcall.mu_upcall_dev = NULL;
395 m->md_upcall.mu_upcall = NULL;
396 up_write(&m->md_upcall.mu_upcall_sem);
399 static inline int md_do_upcall(const struct lu_env *env, struct md_device *m,
400 enum md_upcall_event ev, void *data)
403 down_read(&m->md_upcall.mu_upcall_sem);
404 if (m->md_upcall.mu_upcall_dev != NULL &&
405 m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall != NULL) {
406 rc = m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall(env,
407 m->md_upcall.mu_upcall_dev,
410 up_read(&m->md_upcall.mu_upcall_sem);
415 struct lu_object mo_lu;
416 const struct md_object_operations *mo_ops;
417 const struct md_dir_operations *mo_dir_ops;
423 struct seq_server_site {
424 struct lu_site *ss_lu;
426 * mds number of this site.
430 * Fid location database
432 struct lu_server_fld *ss_server_fld;
433 struct lu_client_fld *ss_client_fld;
438 struct lu_server_seq *ss_server_seq;
441 * Controller Seq Manager
443 struct lu_server_seq *ss_control_seq;
444 struct obd_export *ss_control_exp;
449 struct lu_client_seq *ss_client_seq;
452 static inline struct md_device *lu2md_dev(const struct lu_device *d)
454 LASSERT(IS_ERR(d) || lu_device_is_md(d));
455 return container_of0(d, struct md_device, md_lu_dev);
458 static inline struct lu_device *md2lu_dev(struct md_device *d)
460 return &d->md_lu_dev;
463 static inline struct md_object *lu2md(const struct lu_object *o)
465 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev));
466 return container_of0(o, struct md_object, mo_lu);
469 static inline struct md_object *md_object_next(const struct md_object *obj)
471 return (obj ? lu2md(lu_object_next(&obj->mo_lu)) : NULL);
474 static inline struct md_device *md_obj2dev(const struct md_object *o)
476 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->mo_lu.lo_dev));
477 return container_of0(o->mo_lu.lo_dev, struct md_device, md_lu_dev);
480 static inline int md_device_init(struct md_device *md, struct lu_device_type *t)
482 return lu_device_init(&md->md_lu_dev, t);
485 static inline void md_device_fini(struct md_device *md)
487 lu_device_fini(&md->md_lu_dev);
490 static inline struct md_object *md_object_find_slice(const struct lu_env *env,
491 struct md_device *md,
492 const struct lu_fid *f)
494 return lu2md(lu_object_find_slice(env, md2lu_dev(md), f, NULL));
499 static inline int mo_permission(const struct lu_env *env,
505 LASSERT(c->mo_ops->moo_permission);
506 return c->mo_ops->moo_permission(env, p, c, at, mask);
509 static inline int mo_attr_get(const struct lu_env *env,
513 LASSERT(m->mo_ops->moo_attr_get);
514 return m->mo_ops->moo_attr_get(env, m, at);
517 static inline int mo_readlink(const struct lu_env *env,
521 LASSERT(m->mo_ops->moo_readlink);
522 return m->mo_ops->moo_readlink(env, m, buf);
525 static inline int mo_changelog(const struct lu_env *env,
526 enum changelog_rec_type type,
527 int flags, struct md_object *m)
529 LASSERT(m->mo_ops->moo_changelog);
530 return m->mo_ops->moo_changelog(env, type, flags, m);
533 static inline int mo_attr_set(const struct lu_env *env,
535 const struct md_attr *at)
537 LASSERT(m->mo_ops->moo_attr_set);
538 return m->mo_ops->moo_attr_set(env, m, at);
541 static inline int mo_xattr_get(const struct lu_env *env,
546 LASSERT(m->mo_ops->moo_xattr_get);
547 return m->mo_ops->moo_xattr_get(env, m, buf, name);
550 static inline int mo_xattr_del(const struct lu_env *env,
554 LASSERT(m->mo_ops->moo_xattr_del);
555 return m->mo_ops->moo_xattr_del(env, m, name);
558 static inline int mo_xattr_set(const struct lu_env *env,
560 const struct lu_buf *buf,
564 LASSERT(m->mo_ops->moo_xattr_set);
565 return m->mo_ops->moo_xattr_set(env, m, buf, name, flags);
568 static inline int mo_xattr_list(const struct lu_env *env,
572 LASSERT(m->mo_ops->moo_xattr_list);
573 return m->mo_ops->moo_xattr_list(env, m, buf);
576 static inline int mo_swap_layouts(const struct lu_env *env,
577 struct md_object *o1,
578 struct md_object *o2, __u64 flags)
580 LASSERT(o1->mo_ops->moo_swap_layouts);
581 LASSERT(o2->mo_ops->moo_swap_layouts);
582 if (o1->mo_ops->moo_swap_layouts != o2->mo_ops->moo_swap_layouts)
584 return o1->mo_ops->moo_swap_layouts(env, o1, o2, flags);
587 static inline int mo_open(const struct lu_env *env,
591 LASSERT(m->mo_ops->moo_open);
592 return m->mo_ops->moo_open(env, m, flags);
595 static inline int mo_close(const struct lu_env *env,
600 LASSERT(m->mo_ops->moo_close);
601 return m->mo_ops->moo_close(env, m, ma, mode);
604 static inline int mo_readpage(const struct lu_env *env,
606 const struct lu_rdpg *rdpg)
608 LASSERT(m->mo_ops->moo_readpage);
609 return m->mo_ops->moo_readpage(env, m, rdpg);
612 static inline int mo_object_create(const struct lu_env *env,
614 const struct md_op_spec *spc,
617 LASSERT(m->mo_ops->moo_object_create);
618 return m->mo_ops->moo_object_create(env, m, spc, at);
621 static inline int mo_ref_add(const struct lu_env *env,
623 const struct md_attr *ma)
625 LASSERT(m->mo_ops->moo_ref_add);
626 return m->mo_ops->moo_ref_add(env, m, ma);
629 static inline int mo_ref_del(const struct lu_env *env,
633 LASSERT(m->mo_ops->moo_ref_del);
634 return m->mo_ops->moo_ref_del(env, m, ma);
637 static inline int mo_capa_get(const struct lu_env *env,
639 struct lustre_capa *c,
642 LASSERT(m->mo_ops->moo_capa_get);
643 return m->mo_ops->moo_capa_get(env, m, c, renewal);
646 static inline int mo_object_sync(const struct lu_env *env, struct md_object *m)
648 LASSERT(m->mo_ops->moo_object_sync);
649 return m->mo_ops->moo_object_sync(env, m);
652 static inline int mo_file_lock(const struct lu_env *env, struct md_object *m,
653 struct lov_mds_md *lmm,
654 struct ldlm_extent *extent,
655 struct lustre_handle *lockh)
657 LASSERT(m->mo_ops->moo_file_lock);
658 return m->mo_ops->moo_file_lock(env, m, lmm, extent, lockh);
661 static inline int mo_file_unlock(const struct lu_env *env, struct md_object *m,
662 struct lov_mds_md *lmm,
663 struct lustre_handle *lockh)
665 LASSERT(m->mo_ops->moo_file_unlock);
666 return m->mo_ops->moo_file_unlock(env, m, lmm, lockh);
669 static inline int mo_object_lock(const struct lu_env *env,
671 struct lustre_handle *lh,
672 struct ldlm_enqueue_info *einfo,
675 LASSERT(m->mo_ops->moo_object_lock);
676 return m->mo_ops->moo_object_lock(env, m, lh, einfo, policy);
679 static inline int mdo_lookup(const struct lu_env *env,
681 const struct lu_name *lname,
683 struct md_op_spec *spec)
685 LASSERT(p->mo_dir_ops->mdo_lookup);
686 return p->mo_dir_ops->mdo_lookup(env, p, lname, f, spec);
689 static inline mdl_mode_t mdo_lock_mode(const struct lu_env *env,
690 struct md_object *mo,
693 if (mo->mo_dir_ops->mdo_lock_mode == NULL)
695 return mo->mo_dir_ops->mdo_lock_mode(env, mo, lm);
698 static inline int mdo_create(const struct lu_env *env,
700 const struct lu_name *lchild_name,
702 struct md_op_spec *spc,
705 LASSERT(p->mo_dir_ops->mdo_create);
706 return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at);
709 static inline int mdo_create_data(const struct lu_env *env,
712 const struct md_op_spec *spec,
715 LASSERT(c->mo_dir_ops->mdo_create_data);
716 return c->mo_dir_ops->mdo_create_data(env, p, c, spec, ma);
719 static inline int mdo_rename(const struct lu_env *env,
720 struct md_object *sp,
721 struct md_object *tp,
722 const struct lu_fid *lf,
723 const struct lu_name *lsname,
725 const struct lu_name *ltname,
728 LASSERT(tp->mo_dir_ops->mdo_rename);
729 return tp->mo_dir_ops->mdo_rename(env, sp, tp, lf, lsname, t, ltname,
733 static inline int mdo_is_subdir(const struct lu_env *env,
734 struct md_object *mo,
735 const struct lu_fid *fid,
738 LASSERT(mo->mo_dir_ops->mdo_is_subdir);
739 return mo->mo_dir_ops->mdo_is_subdir(env, mo, fid, sfid);
742 static inline int mdo_link(const struct lu_env *env,
745 const struct lu_name *lname,
748 LASSERT(s->mo_dir_ops->mdo_link);
749 return s->mo_dir_ops->mdo_link(env, p, s, lname, ma);
752 static inline int mdo_unlink(const struct lu_env *env,
755 const struct lu_name *lname,
756 struct md_attr *ma, int no_name)
758 LASSERT(p->mo_dir_ops->mdo_unlink);
759 return p->mo_dir_ops->mdo_unlink(env, p, c, lname, ma, no_name);
762 static inline int mdo_lum_lmm_cmp(const struct lu_env *env,
764 const struct md_op_spec *spec,
767 LASSERT(c->mo_dir_ops->mdo_lum_lmm_cmp);
768 return c->mo_dir_ops->mdo_lum_lmm_cmp(env, c, spec, ma);
771 static inline int mdo_name_insert(const struct lu_env *env,
773 const struct lu_name *lname,
774 const struct lu_fid *f,
775 const struct md_attr *ma)
777 LASSERT(p->mo_dir_ops->mdo_name_insert);
778 return p->mo_dir_ops->mdo_name_insert(env, p, lname, f, ma);
781 static inline int mdo_name_remove(const struct lu_env *env,
783 const struct lu_name *lname,
784 const struct md_attr *ma)
786 LASSERT(p->mo_dir_ops->mdo_name_remove);
787 return p->mo_dir_ops->mdo_name_remove(env, p, lname, ma);
790 static inline int mdo_rename_tgt(const struct lu_env *env,
793 const struct lu_fid *lf,
794 const struct lu_name *lname,
798 LASSERT(t->mo_dir_ops->mdo_rename_tgt);
799 return t->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
801 LASSERT(p->mo_dir_ops->mdo_rename_tgt);
802 return p->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
807 * Used in MDD/OUT layer for object lock rule
809 enum mdd_object_role {
819 * Structure to hold object information. This is used to create object
820 * \pre llod_dir exist
822 struct lu_local_obj_desc {
823 const char *llod_dir;
824 const char *llod_name;
827 const struct dt_index_features *llod_feat;
828 cfs_list_t llod_linkage;
831 int lustre_buf2som(void *buf, int rc, struct md_som_data *msd);
832 int lustre_buf2hsm(void *buf, int rc, struct md_hsm *mh);
833 void lustre_hsm2buf(void *buf, const struct md_hsm *mh);
852 __u32 uc_suppgids[2];
855 struct group_info *uc_ginfo;
856 struct md_identity *uc_identity;
859 struct lu_ucred *lu_ucred(const struct lu_env *env);
861 struct lu_ucred *lu_ucred_check(const struct lu_env *env);
863 struct lu_ucred *lu_ucred_assert(const struct lu_env *env);
865 int lu_ucred_global_init(void);
867 void lu_ucred_global_fini(void);
869 #define md_cap_t(x) (x)
871 #define MD_CAP_TO_MASK(x) (1 << (x))
873 #define md_cap_raised(c, flag) (md_cap_t(c) & MD_CAP_TO_MASK(flag))
875 /* capable() is copied from linux kernel! */
876 static inline int md_capable(struct lu_ucred *uc, cfs_cap_t cap)
878 if (md_cap_raised(uc->uc_cap, cap))
884 #endif /* _LINUX_MD_OBJECT_H */