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 /** This method is used to swap the layouts between 2 objects */
235 int (*moo_swap_layouts)(const struct lu_env *env,
236 struct md_object *obj1, struct md_object *obj2,
239 /** \retval number of bytes actually read upon success */
240 int (*moo_readpage)(const struct lu_env *env, struct md_object *obj,
241 const struct lu_rdpg *rdpg);
243 int (*moo_readlink)(const struct lu_env *env, struct md_object *obj,
245 int (*moo_changelog)(const struct lu_env *env,
246 enum changelog_rec_type type, int flags,
247 struct md_object *obj);
248 /** part of cross-ref operation */
249 int (*moo_object_create)(const struct lu_env *env,
250 struct md_object *obj,
251 const struct md_op_spec *spec,
254 int (*moo_ref_add)(const struct lu_env *env,
255 struct md_object *obj,
256 const struct md_attr *ma);
258 int (*moo_ref_del)(const struct lu_env *env,
259 struct md_object *obj,
262 int (*moo_open)(const struct lu_env *env,
263 struct md_object *obj, int flag);
265 int (*moo_close)(const struct lu_env *env, struct md_object *obj,
266 struct md_attr *ma, int mode);
268 int (*moo_capa_get)(const struct lu_env *, struct md_object *,
269 struct lustre_capa *, int renewal);
271 int (*moo_object_sync)(const struct lu_env *, struct md_object *);
272 int (*moo_path)(const struct lu_env *env, struct md_object *obj,
273 char *path, int pathlen, __u64 *recno, int *linkno);
274 int (*moo_file_lock)(const struct lu_env *env, struct md_object *obj,
275 struct lov_mds_md *lmm, struct ldlm_extent *extent,
276 struct lustre_handle *lockh);
277 int (*moo_file_unlock)(const struct lu_env *env, struct md_object *obj,
278 struct lov_mds_md *lmm,
279 struct lustre_handle *lockh);
283 * Operations implemented for each directory object.
285 struct md_dir_operations {
286 int (*mdo_is_subdir) (const struct lu_env *env, struct md_object *obj,
287 const struct lu_fid *fid, struct lu_fid *sfid);
289 int (*mdo_lookup)(const struct lu_env *env, struct md_object *obj,
290 const struct lu_name *lname, struct lu_fid *fid,
291 struct md_op_spec *spec);
293 mdl_mode_t (*mdo_lock_mode)(const struct lu_env *env,
294 struct md_object *obj,
297 int (*mdo_create)(const struct lu_env *env, struct md_object *pobj,
298 const struct lu_name *lname, struct md_object *child,
299 struct md_op_spec *spec,
302 /** This method is used for creating data object for this meta object*/
303 int (*mdo_create_data)(const struct lu_env *env, struct md_object *p,
305 const struct md_op_spec *spec,
308 int (*mdo_rename)(const struct lu_env *env, struct md_object *spobj,
309 struct md_object *tpobj, const struct lu_fid *lf,
310 const struct lu_name *lsname, struct md_object *tobj,
311 const struct lu_name *ltname, struct md_attr *ma);
313 int (*mdo_link)(const struct lu_env *env, struct md_object *tgt_obj,
314 struct md_object *src_obj, const struct lu_name *lname,
317 int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj,
318 struct md_object *cobj, const struct lu_name *lname,
321 /** This method is used to compare a requested layout to an existing
322 * layout (struct lov_mds_md_v1/3 vs struct lov_mds_md_v1/3) */
323 int (*mdo_lum_lmm_cmp)(const struct lu_env *env,
324 struct md_object *cobj,
325 const struct md_op_spec *spec,
328 /** partial ops for cross-ref case */
329 int (*mdo_name_insert)(const struct lu_env *env,
330 struct md_object *obj,
331 const struct lu_name *lname,
332 const struct lu_fid *fid,
333 const struct md_attr *ma);
335 int (*mdo_name_remove)(const struct lu_env *env,
336 struct md_object *obj,
337 const struct lu_name *lname,
338 const struct md_attr *ma);
340 int (*mdo_rename_tgt)(const struct lu_env *env, struct md_object *pobj,
341 struct md_object *tobj, const struct lu_fid *fid,
342 const struct lu_name *lname, struct md_attr *ma);
345 struct md_device_operations {
346 /** meta-data device related handlers. */
347 int (*mdo_root_get)(const struct lu_env *env, struct md_device *m,
350 int (*mdo_maxsize_get)(const struct lu_env *env, struct md_device *m,
351 int *md_size, int *cookie_size);
353 int (*mdo_statfs)(const struct lu_env *env, struct md_device *m,
354 struct obd_statfs *sfs);
356 int (*mdo_init_capa_ctxt)(const struct lu_env *env, struct md_device *m,
357 int mode, unsigned long timeout, __u32 alg,
358 struct lustre_capa_key *keys);
360 int (*mdo_update_capa_key)(const struct lu_env *env,
362 struct lustre_capa_key *key);
364 int (*mdo_llog_ctxt_get)(const struct lu_env *env,
365 struct md_device *m, int idx, void **h);
367 int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m,
368 unsigned int cmd, int len, void *data);
371 enum md_upcall_event {
372 /** Sync the md layer*/
373 MD_LOV_SYNC = (1 << 0),
374 /** Just for split, no need trans, for replay */
375 MD_NO_TRANS = (1 << 1),
376 MD_LOV_CONFIG = (1 << 2),
377 /** Trigger quota recovery */
378 MD_LOV_QUOTA = (1 << 3)
382 /** this lock protects upcall using against its removal
383 * read lock is for usage the upcall, write - for init/fini */
384 struct rw_semaphore mu_upcall_sem;
385 /** device to call, upper layer normally */
386 struct md_device *mu_upcall_dev;
387 /** upcall function */
388 int (*mu_upcall)(const struct lu_env *env, struct md_device *md,
389 enum md_upcall_event ev, void *data);
393 struct lu_device md_lu_dev;
394 const struct md_device_operations *md_ops;
395 struct md_upcall md_upcall;
398 static inline void md_upcall_init(struct md_device *m, void *upcl)
400 init_rwsem(&m->md_upcall.mu_upcall_sem);
401 m->md_upcall.mu_upcall_dev = NULL;
402 m->md_upcall.mu_upcall = upcl;
405 static inline void md_upcall_dev_set(struct md_device *m, struct md_device *up)
407 down_write(&m->md_upcall.mu_upcall_sem);
408 m->md_upcall.mu_upcall_dev = up;
409 up_write(&m->md_upcall.mu_upcall_sem);
412 static inline void md_upcall_fini(struct md_device *m)
414 down_write(&m->md_upcall.mu_upcall_sem);
415 m->md_upcall.mu_upcall_dev = NULL;
416 m->md_upcall.mu_upcall = NULL;
417 up_write(&m->md_upcall.mu_upcall_sem);
420 static inline int md_do_upcall(const struct lu_env *env, struct md_device *m,
421 enum md_upcall_event ev, void *data)
424 down_read(&m->md_upcall.mu_upcall_sem);
425 if (m->md_upcall.mu_upcall_dev != NULL &&
426 m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall != NULL) {
427 rc = m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall(env,
428 m->md_upcall.mu_upcall_dev,
431 up_read(&m->md_upcall.mu_upcall_sem);
436 struct lu_object mo_lu;
437 const struct md_object_operations *mo_ops;
438 const struct md_dir_operations *mo_dir_ops;
445 struct lu_site *ms_lu;
447 * mds number of this site.
451 * Fid location database
453 struct lu_server_fld *ms_server_fld;
454 struct lu_client_fld *ms_client_fld;
459 struct lu_server_seq *ms_server_seq;
462 * Controller Seq Manager
464 struct lu_server_seq *ms_control_seq;
465 struct obd_export *ms_control_exp;
470 struct lu_client_seq *ms_client_seq;
473 static inline int lu_device_is_md(const struct lu_device *d)
475 return ergo(d != NULL, d->ld_type->ldt_tags & LU_DEVICE_MD);
478 static inline struct md_device *lu2md_dev(const struct lu_device *d)
480 LASSERT(IS_ERR(d) || lu_device_is_md(d));
481 return container_of0(d, struct md_device, md_lu_dev);
484 static inline struct lu_device *md2lu_dev(struct md_device *d)
486 return &d->md_lu_dev;
489 static inline struct md_object *lu2md(const struct lu_object *o)
491 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev));
492 return container_of0(o, struct md_object, mo_lu);
495 static inline struct md_object *md_object_next(const struct md_object *obj)
497 return (obj ? lu2md(lu_object_next(&obj->mo_lu)) : NULL);
500 static inline struct md_device *md_obj2dev(const struct md_object *o)
502 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->mo_lu.lo_dev));
503 return container_of0(o->mo_lu.lo_dev, struct md_device, md_lu_dev);
506 static inline struct md_site *lu_site2md(const struct lu_site *s)
508 return s->ld_md_site;
511 static inline int md_device_init(struct md_device *md, struct lu_device_type *t)
513 return lu_device_init(&md->md_lu_dev, t);
516 static inline void md_device_fini(struct md_device *md)
518 lu_device_fini(&md->md_lu_dev);
521 static inline struct md_object *md_object_find_slice(const struct lu_env *env,
522 struct md_device *md,
523 const struct lu_fid *f)
525 return lu2md(lu_object_find_slice(env, md2lu_dev(md), f, NULL));
530 static inline int mo_permission(const struct lu_env *env,
536 LASSERT(c->mo_ops->moo_permission);
537 return c->mo_ops->moo_permission(env, p, c, at, mask);
540 static inline int mo_attr_get(const struct lu_env *env,
544 LASSERT(m->mo_ops->moo_attr_get);
545 return m->mo_ops->moo_attr_get(env, m, at);
548 static inline int mo_readlink(const struct lu_env *env,
552 LASSERT(m->mo_ops->moo_readlink);
553 return m->mo_ops->moo_readlink(env, m, buf);
556 static inline int mo_changelog(const struct lu_env *env,
557 enum changelog_rec_type type,
558 int flags, struct md_object *m)
560 LASSERT(m->mo_ops->moo_changelog);
561 return m->mo_ops->moo_changelog(env, type, flags, m);
564 static inline int mo_attr_set(const struct lu_env *env,
566 const struct md_attr *at)
568 LASSERT(m->mo_ops->moo_attr_set);
569 return m->mo_ops->moo_attr_set(env, m, at);
572 static inline int mo_xattr_get(const struct lu_env *env,
577 LASSERT(m->mo_ops->moo_xattr_get);
578 return m->mo_ops->moo_xattr_get(env, m, buf, name);
581 static inline int mo_xattr_del(const struct lu_env *env,
585 LASSERT(m->mo_ops->moo_xattr_del);
586 return m->mo_ops->moo_xattr_del(env, m, name);
589 static inline int mo_xattr_set(const struct lu_env *env,
591 const struct lu_buf *buf,
595 LASSERT(m->mo_ops->moo_xattr_set);
596 return m->mo_ops->moo_xattr_set(env, m, buf, name, flags);
599 static inline int mo_xattr_list(const struct lu_env *env,
603 LASSERT(m->mo_ops->moo_xattr_list);
604 return m->mo_ops->moo_xattr_list(env, m, buf);
607 static inline int mo_swap_layouts(const struct lu_env *env,
608 struct md_object *o1,
609 struct md_object *o2, __u64 flags)
611 LASSERT(o1->mo_ops->moo_swap_layouts);
612 LASSERT(o2->mo_ops->moo_swap_layouts);
613 if (o1->mo_ops->moo_swap_layouts != o2->mo_ops->moo_swap_layouts)
615 return o1->mo_ops->moo_swap_layouts(env, o1, o2, flags);
618 static inline int mo_open(const struct lu_env *env,
622 LASSERT(m->mo_ops->moo_open);
623 return m->mo_ops->moo_open(env, m, flags);
626 static inline int mo_close(const struct lu_env *env,
631 LASSERT(m->mo_ops->moo_close);
632 return m->mo_ops->moo_close(env, m, ma, mode);
635 static inline int mo_readpage(const struct lu_env *env,
637 const struct lu_rdpg *rdpg)
639 LASSERT(m->mo_ops->moo_readpage);
640 return m->mo_ops->moo_readpage(env, m, rdpg);
643 static inline int mo_object_create(const struct lu_env *env,
645 const struct md_op_spec *spc,
648 LASSERT(m->mo_ops->moo_object_create);
649 return m->mo_ops->moo_object_create(env, m, spc, at);
652 static inline int mo_ref_add(const struct lu_env *env,
654 const struct md_attr *ma)
656 LASSERT(m->mo_ops->moo_ref_add);
657 return m->mo_ops->moo_ref_add(env, m, ma);
660 static inline int mo_ref_del(const struct lu_env *env,
664 LASSERT(m->mo_ops->moo_ref_del);
665 return m->mo_ops->moo_ref_del(env, m, ma);
668 static inline int mo_capa_get(const struct lu_env *env,
670 struct lustre_capa *c,
673 LASSERT(m->mo_ops->moo_capa_get);
674 return m->mo_ops->moo_capa_get(env, m, c, renewal);
677 static inline int mo_path(const struct lu_env *env, struct md_object *m,
678 char *path, int pathlen, __u64 *recno, int *linkno)
680 if (m->mo_ops->moo_path == NULL)
682 return m->mo_ops->moo_path(env, m, path, pathlen, recno, linkno);
685 static inline int mo_object_sync(const struct lu_env *env, struct md_object *m)
687 LASSERT(m->mo_ops->moo_object_sync);
688 return m->mo_ops->moo_object_sync(env, m);
691 static inline int mo_file_lock(const struct lu_env *env, struct md_object *m,
692 struct lov_mds_md *lmm,
693 struct ldlm_extent *extent,
694 struct lustre_handle *lockh)
696 LASSERT(m->mo_ops->moo_file_lock);
697 return m->mo_ops->moo_file_lock(env, m, lmm, extent, lockh);
700 static inline int mo_file_unlock(const struct lu_env *env, struct md_object *m,
701 struct lov_mds_md *lmm,
702 struct lustre_handle *lockh)
704 LASSERT(m->mo_ops->moo_file_unlock);
705 return m->mo_ops->moo_file_unlock(env, m, lmm, lockh);
708 static inline int mdo_lookup(const struct lu_env *env,
710 const struct lu_name *lname,
712 struct md_op_spec *spec)
714 LASSERT(p->mo_dir_ops->mdo_lookup);
715 return p->mo_dir_ops->mdo_lookup(env, p, lname, f, spec);
718 static inline mdl_mode_t mdo_lock_mode(const struct lu_env *env,
719 struct md_object *mo,
722 if (mo->mo_dir_ops->mdo_lock_mode == NULL)
724 return mo->mo_dir_ops->mdo_lock_mode(env, mo, lm);
727 static inline int mdo_create(const struct lu_env *env,
729 const struct lu_name *lchild_name,
731 struct md_op_spec *spc,
734 LASSERT(p->mo_dir_ops->mdo_create);
735 return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at);
738 static inline int mdo_create_data(const struct lu_env *env,
741 const struct md_op_spec *spec,
744 LASSERT(c->mo_dir_ops->mdo_create_data);
745 return c->mo_dir_ops->mdo_create_data(env, p, c, spec, ma);
748 static inline int mdo_rename(const struct lu_env *env,
749 struct md_object *sp,
750 struct md_object *tp,
751 const struct lu_fid *lf,
752 const struct lu_name *lsname,
754 const struct lu_name *ltname,
757 LASSERT(tp->mo_dir_ops->mdo_rename);
758 return tp->mo_dir_ops->mdo_rename(env, sp, tp, lf, lsname, t, ltname,
762 static inline int mdo_is_subdir(const struct lu_env *env,
763 struct md_object *mo,
764 const struct lu_fid *fid,
767 LASSERT(mo->mo_dir_ops->mdo_is_subdir);
768 return mo->mo_dir_ops->mdo_is_subdir(env, mo, fid, sfid);
771 static inline int mdo_link(const struct lu_env *env,
774 const struct lu_name *lname,
777 LASSERT(s->mo_dir_ops->mdo_link);
778 return s->mo_dir_ops->mdo_link(env, p, s, lname, ma);
781 static inline int mdo_unlink(const struct lu_env *env,
784 const struct lu_name *lname,
787 LASSERT(c->mo_dir_ops->mdo_unlink);
788 return c->mo_dir_ops->mdo_unlink(env, p, c, lname, ma);
791 static inline int mdo_lum_lmm_cmp(const struct lu_env *env,
793 const struct md_op_spec *spec,
796 LASSERT(c->mo_dir_ops->mdo_lum_lmm_cmp);
797 return c->mo_dir_ops->mdo_lum_lmm_cmp(env, c, spec, ma);
800 static inline int mdo_name_insert(const struct lu_env *env,
802 const struct lu_name *lname,
803 const struct lu_fid *f,
804 const struct md_attr *ma)
806 LASSERT(p->mo_dir_ops->mdo_name_insert);
807 return p->mo_dir_ops->mdo_name_insert(env, p, lname, f, ma);
810 static inline int mdo_name_remove(const struct lu_env *env,
812 const struct lu_name *lname,
813 const struct md_attr *ma)
815 LASSERT(p->mo_dir_ops->mdo_name_remove);
816 return p->mo_dir_ops->mdo_name_remove(env, p, lname, ma);
819 static inline int mdo_rename_tgt(const struct lu_env *env,
822 const struct lu_fid *lf,
823 const struct lu_name *lname,
827 LASSERT(t->mo_dir_ops->mdo_rename_tgt);
828 return t->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
830 LASSERT(p->mo_dir_ops->mdo_rename_tgt);
831 return p->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
837 * Structure to hold object information. This is used to create object
838 * \pre llod_dir exist
840 struct lu_local_obj_desc {
841 const char *llod_dir;
842 const char *llod_name;
845 const struct dt_index_features *llod_feat;
846 cfs_list_t llod_linkage;
849 struct md_object *llo_store_resolve(const struct lu_env *env,
850 struct md_device *md,
851 struct dt_device *dt,
855 struct md_object *llo_store_open(const struct lu_env *env,
856 struct md_device *md,
857 struct dt_device *dt,
862 struct md_object *llo_store_create_index(const struct lu_env *env,
863 struct md_device *md,
864 struct dt_device *dt,
867 const struct lu_fid *fid,
868 const struct dt_index_features *feat);
870 struct md_object *llo_store_create(const struct lu_env *env,
871 struct md_device *md,
872 struct dt_device *dt,
875 const struct lu_fid *fid);
877 void llo_local_obj_register(struct lu_local_obj_desc *);
878 void llo_local_obj_unregister(struct lu_local_obj_desc *);
880 int llo_local_objects_setup(const struct lu_env *env,
881 struct md_device * md,
882 struct dt_device * dt);
884 int lustre_buf2som(void *buf, int rc, struct md_som_data *msd);
885 int lustre_buf2hsm(void *buf, int rc, struct md_hsm *mh);
886 void lustre_hsm2buf(void *buf, struct md_hsm *mh);
888 #endif /* _LINUX_MD_OBJECT_H */