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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/md_object.h
34 * Extention of lu_object.h for metadata objects
37 #ifndef _LUSTRE_MD_OBJECT_H
38 #define _LUSTRE_MD_OBJECT_H
40 #ifndef HAVE_SERVER_SUPPORT
41 # error "client code should not depend on md_object.h"
42 #endif /* !HAVE_SERVER_SUPPORT */
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;
66 /** metadata attributes */
78 MA_FORCE_LOG = 1 << 10, /* forced close logged in mdt_mfd_close */
95 MDT_REG_LOCK = (1 << 0),
96 MDT_PDO_LOCK = (1 << 1)
99 /* lfs rgetfacl permission check */
100 #define MAY_RGETFACL (1 << 14)
102 /* memory structure for hsm attributes
103 * for fields description see the on disk structure hsm_attrs
104 * which is defined in lustre_idl.h
114 /* memory structure for SOM attributes
115 * for fields description see the on disk structure som_attrs
116 * which is defined in lustre_idl.h
128 struct lu_attr ma_attr;
129 struct lu_fid ma_pfid;
130 struct md_hsm ma_hsm;
131 struct md_som ma_som;
132 struct lov_mds_md *ma_lmm;
133 union lmv_mds_md *ma_lmv;
134 struct lmv_user_md *ma_default_lmv;
138 int ma_default_lmv_size;
140 int ma_enable_chprojid_gid;
143 /** Additional parameters for create */
146 /** symlink target */
147 const char *sp_symname;
148 /** eadata for regular files */
155 /** Open flags from client: such as MDS_OPEN_CREAT, and others. */
158 /* File security context for creates. */
159 const char *sp_cr_file_secctx_name; /* (security) xattr name */
160 void *sp_cr_file_secctx; /* xattr value */
161 size_t sp_cr_file_secctx_size; /* xattr value size */
163 /* Archive ID used for auto PCC attach when create newly files. */
166 /** don't create lov objects or llog cookie - this replay */
167 unsigned int no_create:1,
168 sp_cr_lookup:1, /* do lookup sanity check or not. */
169 sp_rm_entry:1, /* only remove name entry */
170 sp_permitted:1, /* do not check permission */
171 sp_migrate_close:1; /* close the file during migrate */
172 /** Current lock mode for parent dir where create is performing. */
173 mdl_mode_t sp_cr_mode;
175 /** to create directory */
176 const struct dt_index_features *sp_feat;
181 MD_LAYOUT_WRITE, /* FLR: write the file */
182 MD_LAYOUT_RESYNC, /* FLR: resync starts */
183 MD_LAYOUT_RESYNC_DONE, /* FLR: resync done */
184 MD_LAYOUT_ATTACH, /* attach stripes to target dir */
185 MD_LAYOUT_DETACH, /* detach stripes from dir */
186 MD_LAYOUT_SHRINK, /* shrink stripes (check empty and destroy) */
191 * Parameters for layout change API.
193 struct md_layout_change {
194 enum md_layout_opc mlc_opc;
196 struct layout_intent *mlc_intent;
197 struct lu_buf mlc_buf;
198 struct lustre_som_attrs mlc_som;
199 size_t mlc_resync_count;
200 __u32 *mlc_resync_ids;
203 union ldlm_policy_data;
205 * Operations implemented for each md object (both directory and leaf).
207 struct md_object_operations {
208 int (*moo_permission)(const struct lu_env *env,
209 struct md_object *pobj, struct md_object *cobj,
210 struct md_attr *attr, int mask);
212 int (*moo_attr_get)(const struct lu_env *env, struct md_object *obj,
213 struct md_attr *attr);
215 int (*moo_attr_set)(const struct lu_env *env, struct md_object *obj,
216 const struct md_attr *attr);
218 int (*moo_xattr_get)(const struct lu_env *env, struct md_object *obj,
219 struct lu_buf *buf, const char *name);
221 int (*moo_xattr_list)(const struct lu_env *env, struct md_object *obj,
224 int (*moo_xattr_set)(const struct lu_env *env, struct md_object *obj,
225 const struct lu_buf *buf, const char *name,
228 int (*moo_xattr_del)(const struct lu_env *env, struct md_object *obj,
231 /** This method is used to swap the layouts between 2 objects */
232 int (*moo_swap_layouts)(const struct lu_env *env,
233 struct md_object *obj1, struct md_object *obj2,
236 /** \retval number of bytes actually read upon success */
237 int (*moo_readpage)(const struct lu_env *env, struct md_object *obj,
238 const struct lu_rdpg *rdpg);
240 int (*moo_readlink)(const struct lu_env *env, struct md_object *obj,
243 int (*moo_changelog)(const struct lu_env *env,
244 enum changelog_rec_type type,
245 enum changelog_rec_flags clf_flags,
246 struct md_device *m, const struct lu_fid *fid);
248 int (*moo_open)(const struct lu_env *env, struct md_object *obj,
249 u64 open_flags, struct md_op_spec*);
251 int (*moo_close)(const struct lu_env *env, struct md_object *obj,
252 struct md_attr *ma, u64 open_flags);
254 int (*moo_object_sync)(const struct lu_env *, struct md_object *);
256 int (*moo_object_lock)(const struct lu_env *env, struct md_object *obj,
257 struct lustre_handle *lh,
258 struct ldlm_enqueue_info *einfo,
259 union ldlm_policy_data *policy);
260 int (*moo_object_unlock)(const struct lu_env *env,
261 struct md_object *obj,
262 struct ldlm_enqueue_info *einfo,
263 union ldlm_policy_data *policy);
265 int (*moo_invalidate)(const struct lu_env *env, struct md_object *obj);
267 * Trying to write to un-instantiated layout component.
269 * The caller should have held layout lock.
271 * This API can be extended to support every other layout changing
272 * operations, such as component {add,del,change}, layout swap,
273 * layout merge, etc. One of the benefits by doing this is that the MDT
274 * no longer needs to understand layout.
276 * However, layout creation, removal, and fetch should still use
277 * xattr_{get,set}() because they don't interpret layout on the
280 * \param[in] env execution environment
281 * \param[in] obj MD object
282 * \param[in] layout data structure to describe the changes to
283 * the MD object's layout
286 * \retval -ne error code
288 int (*moo_layout_change)(const struct lu_env *env,
289 struct md_object *obj,
290 struct md_layout_change *layout);
294 * Operations implemented for each directory object.
296 struct md_dir_operations {
297 int (*mdo_is_subdir)(const struct lu_env *env, struct md_object *obj,
298 const struct lu_fid *fid);
300 int (*mdo_lookup)(const struct lu_env *env, struct md_object *obj,
301 const struct lu_name *lname, struct lu_fid *fid,
302 struct md_op_spec *spec);
304 mdl_mode_t (*mdo_lock_mode)(const struct lu_env *env,
305 struct md_object *obj,
308 int (*mdo_create)(const struct lu_env *env, struct md_object *pobj,
309 const struct lu_name *lname, struct md_object *child,
310 struct md_op_spec *spec,
313 /** This method is used for creating data object for this meta object*/
314 int (*mdo_create_data)(const struct lu_env *env, struct md_object *p,
316 const struct md_op_spec *spec,
319 int (*mdo_rename)(const struct lu_env *env, struct md_object *spobj,
320 struct md_object *tpobj, const struct lu_fid *lf,
321 const struct lu_name *lsname, struct md_object *tobj,
322 const struct lu_name *ltname, struct md_attr *ma);
324 int (*mdo_link)(const struct lu_env *env, struct md_object *tgt_obj,
325 struct md_object *src_obj, const struct lu_name *lname,
328 int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj,
329 struct md_object *cobj, const struct lu_name *lname,
330 struct md_attr *ma, int no_name);
332 int (*mdo_migrate)(const struct lu_env *env, struct md_object *pobj,
333 struct md_object *sobj, const struct lu_name *lname,
334 struct md_object *tobj, struct md_op_spec *spec,
338 struct md_device_operations {
339 /** meta-data device related handlers. */
340 int (*mdo_root_get)(const struct lu_env *env, struct md_device *m,
343 const struct dt_device_param *(*mdo_dtconf_get)(const struct lu_env *e,
344 struct md_device *m);
346 int (*mdo_statfs)(const struct lu_env *env, struct md_device *m,
347 struct obd_statfs *sfs);
349 int (*mdo_llog_ctxt_get)(const struct lu_env *env,
350 struct md_device *m, int idx, void **h);
352 int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m,
353 unsigned int cmd, int len, void *data);
357 struct lu_device md_lu_dev;
358 const struct md_device_operations *md_ops;
362 struct lu_object mo_lu;
363 const struct md_object_operations *mo_ops;
364 const struct md_dir_operations *mo_dir_ops;
367 static inline struct md_device *lu2md_dev(const struct lu_device *d)
369 LASSERT(IS_ERR(d) || lu_device_is_md(d));
370 return container_of0(d, struct md_device, md_lu_dev);
373 static inline struct lu_device *md2lu_dev(struct md_device *d)
375 return &d->md_lu_dev;
378 static inline struct md_object *lu2md(const struct lu_object *o)
380 LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev));
381 return container_of0(o, struct md_object, mo_lu);
384 static inline int md_device_init(struct md_device *md, struct lu_device_type *t)
386 return lu_device_init(&md->md_lu_dev, t);
389 static inline void md_device_fini(struct md_device *md)
391 lu_device_fini(&md->md_lu_dev);
394 static inline struct md_object *md_object_find_slice(const struct lu_env *env,
395 struct md_device *md,
396 const struct lu_fid *f)
398 return lu2md(lu_object_find_slice(env, md2lu_dev(md), f, NULL));
403 static inline int mo_permission(const struct lu_env *env, struct md_object *p,
404 struct md_object *c, struct md_attr *at,
407 LASSERT(c->mo_ops->moo_permission);
408 return c->mo_ops->moo_permission(env, p, c, at, mask);
411 static inline int mo_attr_get(const struct lu_env *env, struct md_object *m,
414 LASSERT(m->mo_ops->moo_attr_get);
415 return m->mo_ops->moo_attr_get(env, m, at);
418 static inline int mo_readlink(const struct lu_env *env,
422 LASSERT(m->mo_ops->moo_readlink);
423 return m->mo_ops->moo_readlink(env, m, buf);
426 static inline int mo_changelog(const struct lu_env *env,
427 enum changelog_rec_type type,
428 enum changelog_rec_flags clf_flags,
429 struct md_device *m, const struct lu_fid *fid)
431 struct lu_fid rootfid;
432 struct md_object *root;
435 rc = m->md_ops->mdo_root_get(env, m, &rootfid);
439 root = md_object_find_slice(env, m, &rootfid);
441 RETURN(PTR_ERR(root));
443 LASSERT(root->mo_ops->moo_changelog);
444 rc = root->mo_ops->moo_changelog(env, type, clf_flags, m, fid);
446 lu_object_put(env, &root->mo_lu);
451 static inline int mo_attr_set(const struct lu_env *env,
453 const struct md_attr *at)
455 LASSERT(m->mo_ops->moo_attr_set);
456 return m->mo_ops->moo_attr_set(env, m, at);
459 static inline int mo_xattr_get(const struct lu_env *env,
464 LASSERT(m->mo_ops->moo_xattr_get);
465 return m->mo_ops->moo_xattr_get(env, m, buf, name);
468 static inline int mo_xattr_del(const struct lu_env *env,
472 LASSERT(m->mo_ops->moo_xattr_del);
473 return m->mo_ops->moo_xattr_del(env, m, name);
476 static inline int mo_xattr_set(const struct lu_env *env,
478 const struct lu_buf *buf,
482 LASSERT(m->mo_ops->moo_xattr_set);
483 return m->mo_ops->moo_xattr_set(env, m, buf, name, flags);
486 static inline int mo_xattr_list(const struct lu_env *env,
490 LASSERT(m->mo_ops->moo_xattr_list);
491 return m->mo_ops->moo_xattr_list(env, m, buf);
494 static inline int mo_invalidate(const struct lu_env *env, struct md_object *m)
496 LASSERT(m->mo_ops->moo_invalidate);
497 return m->mo_ops->moo_invalidate(env, m);
500 static inline int mo_layout_change(const struct lu_env *env,
502 struct md_layout_change *layout)
504 /* need instantiate objects which in the access range */
505 LASSERT(m->mo_ops->moo_layout_change);
506 return m->mo_ops->moo_layout_change(env, m, layout);
509 static inline int mo_swap_layouts(const struct lu_env *env,
510 struct md_object *o1,
511 struct md_object *o2, __u64 flags)
513 LASSERT(o1->mo_ops->moo_swap_layouts);
514 LASSERT(o2->mo_ops->moo_swap_layouts);
515 if (o1->mo_ops->moo_swap_layouts != o2->mo_ops->moo_swap_layouts)
517 return o1->mo_ops->moo_swap_layouts(env, o1, o2, flags);
520 static inline int mo_open(const struct lu_env *env, struct md_object *m,
521 u64 open_flags, struct md_op_spec *spec)
523 LASSERT(m->mo_ops->moo_open);
524 return m->mo_ops->moo_open(env, m, open_flags, spec);
527 static inline int mo_close(const struct lu_env *env, struct md_object *m,
528 struct md_attr *ma, u64 open_flags)
530 LASSERT(m->mo_ops->moo_close);
531 return m->mo_ops->moo_close(env, m, ma, open_flags);
534 static inline int mo_readpage(const struct lu_env *env,
536 const struct lu_rdpg *rdpg)
538 LASSERT(m->mo_ops->moo_readpage);
539 return m->mo_ops->moo_readpage(env, m, rdpg);
542 static inline int mo_object_sync(const struct lu_env *env, struct md_object *m)
544 LASSERT(m->mo_ops->moo_object_sync);
545 return m->mo_ops->moo_object_sync(env, m);
548 static inline int mo_object_lock(const struct lu_env *env,
550 struct lustre_handle *lh,
551 struct ldlm_enqueue_info *einfo,
552 union ldlm_policy_data *policy)
554 LASSERT(m->mo_ops->moo_object_lock);
555 return m->mo_ops->moo_object_lock(env, m, lh, einfo, policy);
558 static inline int mo_object_unlock(const struct lu_env *env,
560 struct ldlm_enqueue_info *einfo,
561 union ldlm_policy_data *policy)
563 LASSERT(m->mo_ops->moo_object_unlock);
564 return m->mo_ops->moo_object_unlock(env, m, einfo, policy);
567 static inline int mdo_lookup(const struct lu_env *env,
569 const struct lu_name *lname,
571 struct md_op_spec *spec)
573 LASSERT(p->mo_dir_ops->mdo_lookup);
574 return p->mo_dir_ops->mdo_lookup(env, p, lname, f, spec);
577 static inline mdl_mode_t mdo_lock_mode(const struct lu_env *env,
578 struct md_object *mo,
581 if (mo->mo_dir_ops->mdo_lock_mode == NULL)
583 return mo->mo_dir_ops->mdo_lock_mode(env, mo, lm);
586 static inline int mdo_create(const struct lu_env *env,
588 const struct lu_name *lchild_name,
590 struct md_op_spec *spc,
593 LASSERT(p->mo_dir_ops->mdo_create);
594 return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at);
597 static inline int mdo_create_data(const struct lu_env *env,
600 const struct md_op_spec *spec,
603 LASSERT(c->mo_dir_ops->mdo_create_data);
604 return c->mo_dir_ops->mdo_create_data(env, p, c, spec, ma);
607 static inline int mdo_rename(const struct lu_env *env,
608 struct md_object *sp,
609 struct md_object *tp,
610 const struct lu_fid *lf,
611 const struct lu_name *lsname,
613 const struct lu_name *ltname,
616 LASSERT(tp->mo_dir_ops->mdo_rename);
617 return tp->mo_dir_ops->mdo_rename(env, sp, tp, lf, lsname, t, ltname,
621 static inline int mdo_migrate(const struct lu_env *env,
622 struct md_object *pobj,
623 struct md_object *sobj,
624 const struct lu_name *lname,
625 struct md_object *tobj,
626 struct md_op_spec *spec,
629 LASSERT(pobj->mo_dir_ops->mdo_migrate);
630 return pobj->mo_dir_ops->mdo_migrate(env, pobj, sobj, lname, tobj, spec,
634 static inline int mdo_is_subdir(const struct lu_env *env,
635 struct md_object *mo,
636 const struct lu_fid *fid)
638 LASSERT(mo->mo_dir_ops->mdo_is_subdir);
639 return mo->mo_dir_ops->mdo_is_subdir(env, mo, fid);
642 static inline int mdo_link(const struct lu_env *env,
645 const struct lu_name *lname,
648 LASSERT(s->mo_dir_ops->mdo_link);
649 return s->mo_dir_ops->mdo_link(env, p, s, lname, ma);
652 static inline int mdo_unlink(const struct lu_env *env,
655 const struct lu_name *lname,
656 struct md_attr *ma, int no_name)
658 LASSERT(p->mo_dir_ops->mdo_unlink);
659 return p->mo_dir_ops->mdo_unlink(env, p, c, lname, ma, no_name);
662 static inline int mdo_statfs(const struct lu_env *env,
664 struct obd_statfs *sfs)
666 LASSERT(m->md_ops->mdo_statfs);
667 return m->md_ops->mdo_statfs(env, m, sfs);
672 void lustre_som_swab(struct lustre_som_attrs *attrs);
673 int lustre_buf2hsm(void *buf, int rc, struct md_hsm *mh);
674 void lustre_hsm2buf(void *buf, const struct md_hsm *mh);
693 __u32 uc_suppgids[2];
696 struct group_info *uc_ginfo;
697 struct md_identity *uc_identity;
698 char uc_jobid[LUSTRE_JOBID_SIZE];
700 bool uc_enable_audit;
703 struct lu_ucred *lu_ucred(const struct lu_env *env);
705 struct lu_ucred *lu_ucred_check(const struct lu_env *env);
707 struct lu_ucred *lu_ucred_assert(const struct lu_env *env);
709 int lu_ucred_global_init(void);
711 void lu_ucred_global_fini(void);
713 #define md_cap_t(x) (x)
715 #define MD_CAP_TO_MASK(x) (1 << (x))
717 #define md_cap_raised(c, flag) (md_cap_t(c) & MD_CAP_TO_MASK(flag))
719 /* capable() is copied from linux kernel! */
720 static inline int md_capable(struct lu_ucred *uc, cfs_cap_t cap)
722 if (md_cap_raised(uc->uc_cap, cap))
728 #endif /* _LINUX_MD_OBJECT_H */