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;
70 /** there are at most 5 fids in one operation, see rename, NOTE the last one
71 * is a temporary one used for is_subdir() */
75 struct lu_fid mc_fid[MD_CAPAINFO_MAX];
76 struct lustre_capa *mc_capa[MD_CAPAINFO_MAX];
80 struct obd_export *mq_exp;
84 * Implemented in mdd/mdd_handler.c.
86 * XXX should be moved into separate .h/.c together with all md security
87 * related definitions.
89 struct md_capainfo *md_capainfo(const struct lu_env *env);
90 struct md_quota *md_quota(const struct lu_env *env);
92 /** metadata attributes */
99 MA_ACL_DEF = (1 << 5),
100 MA_LOV_DEF = (1 << 6),
101 MA_LAY_GEN = (1 << 7),
121 MDT_REG_LOCK = (1 << 0),
122 MDT_PDO_LOCK = (1 << 1)
125 /* memory structure for hsm attributes
126 * for fields description see the on disk structure hsm_attrs
127 * which is defined in lustre_idl.h
136 #define IOEPOCH_INVAL 0
138 /* memory structure for som attributes
139 * for fields description see the on disk structure som_attrs
140 * which is defined in lustre_idl.h
155 struct lu_attr ma_attr;
156 struct lu_fid ma_pfid;
157 struct md_hsm ma_hsm;
158 struct lov_mds_md *ma_lmm;
159 struct lmv_stripe_md *ma_lmv;
161 struct llog_cookie *ma_cookie;
162 struct lustre_capa *ma_capa;
163 struct md_som_data *ma_som;
171 /** Additional parameters for create */
174 /** symlink target */
175 const char *sp_symname;
176 /** parent FID for cross-ref mkdir */
177 const struct lu_fid *sp_pfid;
178 /** eadata for regular files */
180 /** lov objs exist already */
181 const struct lu_fid *fid;
187 /** Create flag from client: such as MDS_OPEN_CREAT, and others. */
190 /** don't create lov objects or llog cookie - this replay */
191 unsigned int no_create:1,
192 sp_cr_lookup:1, /* do lookup sanity check or not. */
193 sp_rm_entry:1; /* only remove name entry */
195 /** Current lock mode for parent dir where create is performing. */
196 mdl_mode_t sp_cr_mode;
198 /** to create directory */
199 const struct dt_index_features *sp_feat;
203 * Operations implemented for each md object (both directory and leaf).
205 struct md_object_operations {
206 int (*moo_permission)(const struct lu_env *env,
207 struct md_object *pobj, struct md_object *cobj,
208 struct md_attr *attr, int mask);
210 int (*moo_attr_get)(const struct lu_env *env, struct md_object *obj,
211 struct md_attr *attr);
213 int (*moo_attr_set)(const struct lu_env *env, struct md_object *obj,
214 const struct md_attr *attr);
216 int (*moo_xattr_get)(const struct lu_env *env, struct md_object *obj,
217 struct lu_buf *buf, const char *name);
219 int (*moo_xattr_list)(const struct lu_env *env, struct md_object *obj,
222 int (*moo_xattr_set)(const struct lu_env *env, struct md_object *obj,
223 const struct lu_buf *buf, const char *name,
226 int (*moo_xattr_del)(const struct lu_env *env, struct md_object *obj,
229 /** This method is used to swap the layouts between 2 objects */
230 int (*moo_swap_layouts)(const struct lu_env *env,
231 struct md_object *obj1, struct md_object *obj2,
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 *);
268 int (*moo_file_lock)(const struct lu_env *env, struct md_object *obj,
269 struct lov_mds_md *lmm, struct ldlm_extent *extent,
270 struct lustre_handle *lockh);
271 int (*moo_file_unlock)(const struct lu_env *env, struct md_object *obj,
272 struct lov_mds_md *lmm,
273 struct lustre_handle *lockh);
274 int (*moo_object_lock)(const struct lu_env *env, struct md_object *obj,
275 struct lustre_handle *lh,
276 struct ldlm_enqueue_info *einfo,
281 * Operations implemented for each directory object.
283 struct md_dir_operations {
284 int (*mdo_is_subdir) (const struct lu_env *env, struct md_object *obj,
285 const struct lu_fid *fid, struct lu_fid *sfid);
287 int (*mdo_lookup)(const struct lu_env *env, struct md_object *obj,
288 const struct lu_name *lname, struct lu_fid *fid,
289 struct md_op_spec *spec);
291 mdl_mode_t (*mdo_lock_mode)(const struct lu_env *env,
292 struct md_object *obj,
295 int (*mdo_create)(const struct lu_env *env, struct md_object *pobj,
296 const struct lu_name *lname, struct md_object *child,
297 struct md_op_spec *spec,
300 /** This method is used for creating data object for this meta object*/
301 int (*mdo_create_data)(const struct lu_env *env, struct md_object *p,
303 const struct md_op_spec *spec,
306 int (*mdo_rename)(const struct lu_env *env, struct md_object *spobj,
307 struct md_object *tpobj, const struct lu_fid *lf,
308 const struct lu_name *lsname, struct md_object *tobj,
309 const struct lu_name *ltname, struct md_attr *ma);
311 int (*mdo_link)(const struct lu_env *env, struct md_object *tgt_obj,
312 struct md_object *src_obj, const struct lu_name *lname,
315 int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj,
316 struct md_object *cobj, const struct lu_name *lname,
317 struct md_attr *ma, int no_name);
319 /** This method is used to compare a requested layout to an existing
320 * layout (struct lov_mds_md_v1/3 vs struct lov_mds_md_v1/3) */
321 int (*mdo_lum_lmm_cmp)(const struct lu_env *env,
322 struct md_object *cobj,
323 const struct md_op_spec *spec,
326 /** partial ops for cross-ref case */
327 int (*mdo_name_insert)(const struct lu_env *env,
328 struct md_object *obj,
329 const struct lu_name *lname,
330 const struct lu_fid *fid,
331 const struct md_attr *ma);
333 int (*mdo_name_remove)(const struct lu_env *env,
334 struct md_object *obj,
335 const struct lu_name *lname,
336 const struct md_attr *ma);
338 int (*mdo_rename_tgt)(const struct lu_env *env, struct md_object *pobj,
339 struct md_object *tobj, const struct lu_fid *fid,
340 const struct lu_name *lname, struct md_attr *ma);
343 struct md_device_operations {
344 /** meta-data device related handlers. */
345 int (*mdo_root_get)(const struct lu_env *env, struct md_device *m,
348 int (*mdo_maxeasize_get)(const struct lu_env *env, struct md_device *m,
351 int (*mdo_statfs)(const struct lu_env *env, struct md_device *m,
352 struct obd_statfs *sfs);
354 int (*mdo_init_capa_ctxt)(const struct lu_env *env, struct md_device *m,
355 int mode, unsigned long timeout, __u32 alg,
356 struct lustre_capa_key *keys);
358 int (*mdo_update_capa_key)(const struct lu_env *env,
360 struct lustre_capa_key *key);
362 int (*mdo_llog_ctxt_get)(const struct lu_env *env,
363 struct md_device *m, int idx, void **h);
365 int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m,
366 unsigned int cmd, int len, void *data);
369 enum md_upcall_event {
370 /** Sync the md layer*/
371 MD_LOV_SYNC = (1 << 0),
372 /** Just for split, no need trans, for replay */
373 MD_NO_TRANS = (1 << 1),
374 MD_LOV_CONFIG = (1 << 2),
375 /** Trigger quota recovery */
376 MD_LOV_QUOTA = (1 << 3)
380 /** this lock protects upcall using against its removal
381 * read lock is for usage the upcall, write - for init/fini */
382 struct rw_semaphore mu_upcall_sem;
383 /** device to call, upper layer normally */
384 struct md_device *mu_upcall_dev;
385 /** upcall function */
386 int (*mu_upcall)(const struct lu_env *env, struct md_device *md,
387 enum md_upcall_event ev, void *data);
391 struct lu_device md_lu_dev;
392 const struct md_device_operations *md_ops;
393 struct md_upcall md_upcall;
396 static inline void md_upcall_init(struct md_device *m, void *upcl)
398 init_rwsem(&m->md_upcall.mu_upcall_sem);
399 m->md_upcall.mu_upcall_dev = NULL;
400 m->md_upcall.mu_upcall = upcl;
403 static inline void md_upcall_dev_set(struct md_device *m, struct md_device *up)
405 down_write(&m->md_upcall.mu_upcall_sem);
406 m->md_upcall.mu_upcall_dev = up;
407 up_write(&m->md_upcall.mu_upcall_sem);
410 static inline void md_upcall_fini(struct md_device *m)
412 down_write(&m->md_upcall.mu_upcall_sem);
413 m->md_upcall.mu_upcall_dev = NULL;
414 m->md_upcall.mu_upcall = NULL;
415 up_write(&m->md_upcall.mu_upcall_sem);
418 static inline int md_do_upcall(const struct lu_env *env, struct md_device *m,
419 enum md_upcall_event ev, void *data)
422 down_read(&m->md_upcall.mu_upcall_sem);
423 if (m->md_upcall.mu_upcall_dev != NULL &&
424 m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall != NULL) {
425 rc = m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall(env,
426 m->md_upcall.mu_upcall_dev,
429 up_read(&m->md_upcall.mu_upcall_sem);
434 struct lu_object mo_lu;
435 const struct md_object_operations *mo_ops;
436 const struct md_dir_operations *mo_dir_ops;
442 struct seq_server_site {
443 struct lu_site *ss_lu;
445 * mds number of this site.
449 * Fid location database
451 struct lu_server_fld *ss_server_fld;
452 struct lu_client_fld *ss_client_fld;
457 struct lu_server_seq *ss_server_seq;
460 * Controller Seq Manager
462 struct lu_server_seq *ss_control_seq;
463 struct obd_export *ss_control_exp;
468 struct lu_client_seq *ss_client_seq;
471 static inline struct md_device *lu2md_dev(const struct lu_device *d)
473 LASSERT(IS_ERR(d) || lu_device_is_md(d));
474 return container_of0(d, struct md_device, md_lu_dev);
477 static inline struct lu_device *md2lu_dev(struct md_device *d)
479 return &d->md_lu_dev;
482 static inline struct md_object *lu2md(const struct lu_object *o)
484 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev));
485 return container_of0(o, struct md_object, mo_lu);
488 static inline struct md_object *md_object_next(const struct md_object *obj)
490 return (obj ? lu2md(lu_object_next(&obj->mo_lu)) : NULL);
493 static inline struct md_device *md_obj2dev(const struct md_object *o)
495 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->mo_lu.lo_dev));
496 return container_of0(o->mo_lu.lo_dev, struct md_device, md_lu_dev);
499 static inline int md_device_init(struct md_device *md, struct lu_device_type *t)
501 return lu_device_init(&md->md_lu_dev, t);
504 static inline void md_device_fini(struct md_device *md)
506 lu_device_fini(&md->md_lu_dev);
509 static inline struct md_object *md_object_find_slice(const struct lu_env *env,
510 struct md_device *md,
511 const struct lu_fid *f)
513 return lu2md(lu_object_find_slice(env, md2lu_dev(md), f, NULL));
518 static inline int mo_permission(const struct lu_env *env,
524 LASSERT(c->mo_ops->moo_permission);
525 return c->mo_ops->moo_permission(env, p, c, at, mask);
528 static inline int mo_attr_get(const struct lu_env *env,
532 LASSERT(m->mo_ops->moo_attr_get);
533 return m->mo_ops->moo_attr_get(env, m, at);
536 static inline int mo_readlink(const struct lu_env *env,
540 LASSERT(m->mo_ops->moo_readlink);
541 return m->mo_ops->moo_readlink(env, m, buf);
544 static inline int mo_changelog(const struct lu_env *env,
545 enum changelog_rec_type type,
546 int flags, struct md_object *m)
548 LASSERT(m->mo_ops->moo_changelog);
549 return m->mo_ops->moo_changelog(env, type, flags, m);
552 static inline int mo_attr_set(const struct lu_env *env,
554 const struct md_attr *at)
556 LASSERT(m->mo_ops->moo_attr_set);
557 return m->mo_ops->moo_attr_set(env, m, at);
560 static inline int mo_xattr_get(const struct lu_env *env,
565 LASSERT(m->mo_ops->moo_xattr_get);
566 return m->mo_ops->moo_xattr_get(env, m, buf, name);
569 static inline int mo_xattr_del(const struct lu_env *env,
573 LASSERT(m->mo_ops->moo_xattr_del);
574 return m->mo_ops->moo_xattr_del(env, m, name);
577 static inline int mo_xattr_set(const struct lu_env *env,
579 const struct lu_buf *buf,
583 LASSERT(m->mo_ops->moo_xattr_set);
584 return m->mo_ops->moo_xattr_set(env, m, buf, name, flags);
587 static inline int mo_xattr_list(const struct lu_env *env,
591 LASSERT(m->mo_ops->moo_xattr_list);
592 return m->mo_ops->moo_xattr_list(env, m, buf);
595 static inline int mo_swap_layouts(const struct lu_env *env,
596 struct md_object *o1,
597 struct md_object *o2, __u64 flags)
599 LASSERT(o1->mo_ops->moo_swap_layouts);
600 LASSERT(o2->mo_ops->moo_swap_layouts);
601 if (o1->mo_ops->moo_swap_layouts != o2->mo_ops->moo_swap_layouts)
603 return o1->mo_ops->moo_swap_layouts(env, o1, o2, flags);
606 static inline int mo_open(const struct lu_env *env,
610 LASSERT(m->mo_ops->moo_open);
611 return m->mo_ops->moo_open(env, m, flags);
614 static inline int mo_close(const struct lu_env *env,
619 LASSERT(m->mo_ops->moo_close);
620 return m->mo_ops->moo_close(env, m, ma, mode);
623 static inline int mo_readpage(const struct lu_env *env,
625 const struct lu_rdpg *rdpg)
627 LASSERT(m->mo_ops->moo_readpage);
628 return m->mo_ops->moo_readpage(env, m, rdpg);
631 static inline int mo_object_create(const struct lu_env *env,
633 const struct md_op_spec *spc,
636 LASSERT(m->mo_ops->moo_object_create);
637 return m->mo_ops->moo_object_create(env, m, spc, at);
640 static inline int mo_ref_add(const struct lu_env *env,
642 const struct md_attr *ma)
644 LASSERT(m->mo_ops->moo_ref_add);
645 return m->mo_ops->moo_ref_add(env, m, ma);
648 static inline int mo_ref_del(const struct lu_env *env,
652 LASSERT(m->mo_ops->moo_ref_del);
653 return m->mo_ops->moo_ref_del(env, m, ma);
656 static inline int mo_capa_get(const struct lu_env *env,
658 struct lustre_capa *c,
661 LASSERT(m->mo_ops->moo_capa_get);
662 return m->mo_ops->moo_capa_get(env, m, c, renewal);
665 static inline int mo_object_sync(const struct lu_env *env, struct md_object *m)
667 LASSERT(m->mo_ops->moo_object_sync);
668 return m->mo_ops->moo_object_sync(env, m);
671 static inline int mo_file_lock(const struct lu_env *env, struct md_object *m,
672 struct lov_mds_md *lmm,
673 struct ldlm_extent *extent,
674 struct lustre_handle *lockh)
676 LASSERT(m->mo_ops->moo_file_lock);
677 return m->mo_ops->moo_file_lock(env, m, lmm, extent, lockh);
680 static inline int mo_file_unlock(const struct lu_env *env, struct md_object *m,
681 struct lov_mds_md *lmm,
682 struct lustre_handle *lockh)
684 LASSERT(m->mo_ops->moo_file_unlock);
685 return m->mo_ops->moo_file_unlock(env, m, lmm, lockh);
688 static inline int mo_object_lock(const struct lu_env *env,
690 struct lustre_handle *lh,
691 struct ldlm_enqueue_info *einfo,
694 LASSERT(m->mo_ops->moo_object_lock);
695 return m->mo_ops->moo_object_lock(env, m, lh, einfo, policy);
698 static inline int mdo_lookup(const struct lu_env *env,
700 const struct lu_name *lname,
702 struct md_op_spec *spec)
704 LASSERT(p->mo_dir_ops->mdo_lookup);
705 return p->mo_dir_ops->mdo_lookup(env, p, lname, f, spec);
708 static inline mdl_mode_t mdo_lock_mode(const struct lu_env *env,
709 struct md_object *mo,
712 if (mo->mo_dir_ops->mdo_lock_mode == NULL)
714 return mo->mo_dir_ops->mdo_lock_mode(env, mo, lm);
717 static inline int mdo_create(const struct lu_env *env,
719 const struct lu_name *lchild_name,
721 struct md_op_spec *spc,
724 LASSERT(p->mo_dir_ops->mdo_create);
725 return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at);
728 static inline int mdo_create_data(const struct lu_env *env,
731 const struct md_op_spec *spec,
734 LASSERT(c->mo_dir_ops->mdo_create_data);
735 return c->mo_dir_ops->mdo_create_data(env, p, c, spec, ma);
738 static inline int mdo_rename(const struct lu_env *env,
739 struct md_object *sp,
740 struct md_object *tp,
741 const struct lu_fid *lf,
742 const struct lu_name *lsname,
744 const struct lu_name *ltname,
747 LASSERT(tp->mo_dir_ops->mdo_rename);
748 return tp->mo_dir_ops->mdo_rename(env, sp, tp, lf, lsname, t, ltname,
752 static inline int mdo_is_subdir(const struct lu_env *env,
753 struct md_object *mo,
754 const struct lu_fid *fid,
757 LASSERT(mo->mo_dir_ops->mdo_is_subdir);
758 return mo->mo_dir_ops->mdo_is_subdir(env, mo, fid, sfid);
761 static inline int mdo_link(const struct lu_env *env,
764 const struct lu_name *lname,
767 LASSERT(s->mo_dir_ops->mdo_link);
768 return s->mo_dir_ops->mdo_link(env, p, s, lname, ma);
771 static inline int mdo_unlink(const struct lu_env *env,
774 const struct lu_name *lname,
775 struct md_attr *ma, int no_name)
777 LASSERT(p->mo_dir_ops->mdo_unlink);
778 return p->mo_dir_ops->mdo_unlink(env, p, c, lname, ma, no_name);
781 static inline int mdo_lum_lmm_cmp(const struct lu_env *env,
783 const struct md_op_spec *spec,
786 LASSERT(c->mo_dir_ops->mdo_lum_lmm_cmp);
787 return c->mo_dir_ops->mdo_lum_lmm_cmp(env, c, spec, ma);
790 static inline int mdo_name_insert(const struct lu_env *env,
792 const struct lu_name *lname,
793 const struct lu_fid *f,
794 const struct md_attr *ma)
796 LASSERT(p->mo_dir_ops->mdo_name_insert);
797 return p->mo_dir_ops->mdo_name_insert(env, p, lname, f, ma);
800 static inline int mdo_name_remove(const struct lu_env *env,
802 const struct lu_name *lname,
803 const struct md_attr *ma)
805 LASSERT(p->mo_dir_ops->mdo_name_remove);
806 return p->mo_dir_ops->mdo_name_remove(env, p, lname, ma);
809 static inline int mdo_rename_tgt(const struct lu_env *env,
812 const struct lu_fid *lf,
813 const struct lu_name *lname,
817 LASSERT(t->mo_dir_ops->mdo_rename_tgt);
818 return t->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
820 LASSERT(p->mo_dir_ops->mdo_rename_tgt);
821 return p->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma);
826 * Used in MDD/OUT layer for object lock rule
828 enum mdd_object_role {
838 * Structure to hold object information. This is used to create object
839 * \pre llod_dir exist
841 struct lu_local_obj_desc {
842 const char *llod_dir;
843 const char *llod_name;
846 const struct dt_index_features *llod_feat;
847 cfs_list_t llod_linkage;
850 int lustre_buf2som(void *buf, int rc, struct md_som_data *msd);
851 int lustre_buf2hsm(void *buf, int rc, struct md_hsm *mh);
852 void lustre_hsm2buf(void *buf, struct md_hsm *mh);
871 __u32 uc_suppgids[2];
874 cfs_group_info_t *uc_ginfo;
875 struct md_identity *uc_identity;
878 struct lu_ucred *lu_ucred(const struct lu_env *env);
880 struct lu_ucred *lu_ucred_check(const struct lu_env *env);
882 struct lu_ucred *lu_ucred_assert(const struct lu_env *env);
884 int lu_ucred_global_init(void);
886 void lu_ucred_global_fini(void);
888 #define md_cap_t(x) (x)
890 #define MD_CAP_TO_MASK(x) (1 << (x))
892 #define md_cap_raised(c, flag) (md_cap_t(c) & MD_CAP_TO_MASK(flag))
894 /* capable() is copied from linux kernel! */
895 static inline int md_capable(struct lu_ucred *uc, cfs_cap_t cap)
897 if (md_cap_raised(uc->uc_cap, cap))
903 #endif /* _LINUX_MD_OBJECT_H */