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/
31 * lustre/mdd/mdd_dir.c
33 * Lustre Metadata Server (mdd) routines
35 * Author: Wang Di <wangdi@intel.com>
38 #define DEBUG_SUBSYSTEM S_MDS
40 #include <obd_class.h>
41 #include <obd_support.h>
42 #include <lustre_mds.h>
43 #include <lustre_fid.h>
44 #include <lustre_lmv.h>
45 #include <lustre_idmap.h>
47 #include "mdd_internal.h"
49 static const char dot[] = ".";
50 static const char dotdot[] = "..";
52 static struct lu_name lname_dotdot = {
53 .ln_name = (char *) dotdot,
54 .ln_namelen = sizeof(dotdot) - 1,
58 mdd_name_check(struct mdd_device *m, const struct lu_name *ln)
60 if (!lu_name_is_valid(ln))
62 else if (ln->ln_namelen > m->mdd_dt_conf.ddp_max_name_len)
68 /* Get FID from name and parent */
70 __mdd_lookup(const struct lu_env *env, struct md_object *pobj,
71 const struct lu_attr *pattr, const struct lu_name *lname,
72 struct lu_fid *fid, unsigned int may_mask)
74 const char *name = lname->ln_name;
75 const struct dt_key *key = (const struct dt_key *)name;
76 struct mdd_object *mdd_obj = md2mdd_obj(pobj);
77 struct dt_object *dir = mdd_object_child(mdd_obj);
82 if (unlikely(mdd_is_dead_obj(mdd_obj)))
85 if (!mdd_object_exists(mdd_obj))
88 if (mdd_object_remote(mdd_obj)) {
89 CDEBUG(D_INFO, "%s: Object "DFID" located on remote server\n",
90 mdd_obj_dev_name(mdd_obj),
91 PFID(mdd_object_fid(mdd_obj)));
94 rc = mdd_permission_internal_locked(env, mdd_obj, pattr, may_mask,
99 if (likely(S_ISDIR(mdd_object_type(mdd_obj)) &&
100 dt_try_as_dir(env, dir)))
101 rc = dt_lookup(env, dir, (struct dt_rec *)fid, key);
108 int mdd_lookup(const struct lu_env *env,
109 struct md_object *pobj, const struct lu_name *lname,
110 struct lu_fid *fid, struct md_op_spec *spec)
112 struct lu_attr *pattr = MDD_ENV_VAR(env, pattr);
116 rc = mdd_la_get(env, md2mdd_obj(pobj), pattr);
120 rc = __mdd_lookup(env, pobj, pattr, lname, fid,
121 (spec != NULL && spec->sp_permitted) ? 0 : MAY_EXEC);
125 /** Read the link EA into a temp buffer.
126 * Uses the mdd_thread_info::mdi_link_buf since it is generally large.
127 * A pointer to the buffer is stored in \a ldata::ld_buf.
131 static int __mdd_links_read(const struct lu_env *env,
132 struct mdd_object *mdd_obj,
133 struct linkea_data *ldata)
137 if (!mdd_object_exists(mdd_obj))
140 /* First try a small buf */
141 LASSERT(env != NULL);
142 ldata->ld_buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mdi_link_buf,
144 if (ldata->ld_buf->lb_buf == NULL)
147 rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf, XATTR_NAME_LINK);
149 /* Buf was too small, figure out what we need. */
150 lu_buf_free(ldata->ld_buf);
151 rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf,
155 ldata->ld_buf = lu_buf_check_and_alloc(ldata->ld_buf, rc);
156 if (ldata->ld_buf->lb_buf == NULL)
158 rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf,
162 lu_buf_free(ldata->ld_buf);
163 ldata->ld_buf = NULL;
167 return linkea_init(ldata);
170 int mdd_links_read(const struct lu_env *env,
171 struct mdd_object *mdd_obj,
172 struct linkea_data *ldata)
176 rc = __mdd_links_read(env, mdd_obj, ldata);
178 rc = linkea_init(ldata);
183 static int mdd_links_read_with_rec(const struct lu_env *env,
184 struct mdd_object *mdd_obj,
185 struct linkea_data *ldata)
189 rc = __mdd_links_read(env, mdd_obj, ldata);
191 rc = linkea_init_with_rec(ldata);
197 * Get parent FID of the directory
199 * Read parent FID from linkEA, if that fails, then do lookup
200 * dotdot to get the parent FID.
202 * \param[in] env execution environment
203 * \param[in] obj object from which to find the parent FID
204 * \param[in] attr attribute of the object
205 * \param[out] fid fid to get the parent FID
207 * \retval 0 if getting the parent FID succeeds.
208 * \retval negative errno if getting the parent FID fails.
210 static inline int mdd_parent_fid(const struct lu_env *env,
211 struct mdd_object *obj,
212 const struct lu_attr *attr,
215 struct mdd_thread_info *info = mdd_env_info(env);
216 struct linkea_data ldata = { NULL };
217 struct lu_buf *buf = &info->mdi_link_buf;
218 struct lu_name lname;
223 LASSERTF(S_ISDIR(mdd_object_type(obj)),
224 "%s: FID "DFID" is not a directory type = %o\n",
225 mdd_obj_dev_name(obj), PFID(mdd_object_fid(obj)),
226 mdd_object_type(obj));
228 buf = lu_buf_check_and_alloc(buf, PATH_MAX);
229 if (buf->lb_buf == NULL)
230 GOTO(lookup, rc = 0);
233 rc = mdd_links_read_with_rec(env, obj, &ldata);
237 /* the obj is not locked, don't cache attributes */
238 mdd_invalidate(env, &obj->mod_obj);
240 LASSERT(ldata.ld_leh != NULL);
241 /* Directory should only have 1 parent */
242 if (ldata.ld_leh->leh_reccount > 1)
245 ldata.ld_lee = (struct link_ea_entry *)(ldata.ld_leh + 1);
247 linkea_entry_unpack(ldata.ld_lee, &ldata.ld_reclen, &lname, fid);
248 if (likely(fid_is_sane(fid)))
251 rc = __mdd_lookup(env, &obj->mod_obj, attr, &lname_dotdot, fid, 0);
256 * For root fid use special function, which does not compare version component
257 * of fid. Version component is different for root fids on all MDTs.
259 int mdd_is_root(struct mdd_device *mdd, const struct lu_fid *fid)
261 return fid_seq(&mdd->mdd_root_fid) == fid_seq(fid) &&
262 fid_oid(&mdd->mdd_root_fid) == fid_oid(fid);
266 * return 1: if \a tfid is the fid of the ancestor of \a mo;
268 * otherwise: values < 0, errors.
270 static int mdd_is_parent(const struct lu_env *env,
271 struct mdd_device *mdd,
272 struct mdd_object *mo,
273 const struct lu_attr *attr,
274 const struct lu_fid *tfid)
276 struct mdd_object *mp;
280 LASSERT(!lu_fid_eq(mdd_object_fid(mo), tfid));
281 pfid = &mdd_env_info(env)->mdi_fid;
283 if (mdd_is_root(mdd, mdd_object_fid(mo)))
286 if (mdd_is_root(mdd, tfid))
289 rc = mdd_parent_fid(env, mo, attr, pfid);
294 if (lu_fid_eq(pfid, tfid))
297 if (mdd_is_root(mdd, pfid))
300 mp = mdd_object_find(env, mdd, pfid);
304 if (!mdd_object_exists(mp)) {
305 mdd_object_put(env, mp);
309 rc = mdd_parent_fid(env, mp, attr, pfid);
310 mdd_object_put(env, mp);
319 * No permission check is needed.
321 * returns 1: if fid is ancestor of @mo;
322 * returns 0: if fid is not an ancestor of @mo;
323 * returns < 0: if error
325 int mdd_is_subdir(const struct lu_env *env, struct md_object *mo,
326 const struct lu_fid *fid)
328 struct mdd_device *mdd = mdo2mdd(mo);
329 struct lu_attr *attr = MDD_ENV_VAR(env, cattr);
333 if (!mdd_object_exists(md2mdd_obj(mo)))
336 if (!S_ISDIR(mdd_object_type(md2mdd_obj(mo))))
339 rc = mdd_la_get(env, md2mdd_obj(mo), attr);
343 rc = mdd_is_parent(env, mdd, md2mdd_obj(mo), attr, fid);
348 * Check that @dir contains no entries except (possibly) dot and dotdot.
353 * -ENOTDIR not a directory object
354 * -ENOTEMPTY not empty
358 int mdd_dir_is_empty(const struct lu_env *env, struct mdd_object *dir)
361 struct dt_object *obj;
362 const struct dt_it_ops *iops;
366 obj = mdd_object_child(dir);
367 if (!dt_try_as_dir(env, obj))
370 iops = &obj->do_index_ops->dio_it;
371 it = iops->init(env, obj, LUDA_64BITHASH);
373 result = iops->get(env, it, (const struct dt_key *)"");
376 for (result = 0, i = 0; result == 0 && i < 3; ++i)
377 result = iops->next(env, it);
380 else if (result == 1)
382 } else if (result == 0)
384 * Huh? Index contains no zero key?
391 result = PTR_ERR(it);
392 /* -ENODEV means no valid stripe */
393 if (result == -ENODEV)
400 * Determine if the target object can be hard linked, and right now it only
401 * checks if the link count reach the maximum limit. Note: for ldiskfs, the
402 * directory nlink count might exceed the maximum link count(see
403 * osd_object_ref_add), so it only check nlink for non-directories.
405 * \param[in] env thread environment
406 * \param[in] obj object being linked to
407 * \param[in] la attributes of \a obj
409 * \retval 0 if \a obj can be hard linked
410 * \retval negative error if \a obj is a directory or has too
413 static int __mdd_may_link(const struct lu_env *env, struct mdd_object *obj,
414 const struct lu_attr *la)
416 struct mdd_device *m = mdd_obj2mdd_dev(obj);
421 /* Subdir count limitation can be broken through
422 * (see osd_object_ref_add), so only check non-directory here. */
423 if (!S_ISDIR(la->la_mode) &&
424 la->la_nlink >= m->mdd_dt_conf.ddp_max_nlink)
431 * Check whether it may create the cobj under the pobj.
433 * \param[in] env execution environment
434 * \param[in] pobj the parent directory
435 * \param[in] pattr the attribute of the parent directory
436 * \param[in] cobj the child to be created
437 * \param[in] check_perm if check WRITE|EXEC permission for parent
439 * \retval = 0 create the child under this dir is allowed
440 * \retval negative errno create the child under this dir is
443 int mdd_may_create(const struct lu_env *env, struct mdd_object *pobj,
444 const struct lu_attr *pattr, struct mdd_object *cobj,
450 if (cobj && mdd_object_exists(cobj))
453 if (mdd_is_dead_obj(pobj))
457 rc = mdd_permission_internal_locked(env, pobj, pattr,
458 MAY_WRITE | MAY_EXEC,
464 * Check whether can unlink from the pobj in the case of "cobj == NULL".
466 int mdd_may_unlink(const struct lu_env *env, struct mdd_object *pobj,
467 const struct lu_attr *pattr, const struct lu_attr *attr)
472 if (mdd_is_dead_obj(pobj))
475 if (attr->la_flags & (LUSTRE_APPEND_FL | LUSTRE_IMMUTABLE_FL))
478 rc = mdd_permission_internal_locked(env, pobj, pattr,
479 MAY_WRITE | MAY_EXEC,
484 if (pattr->la_flags & LUSTRE_APPEND_FL)
491 * pobj == NULL is remote ops case, under such case, pobj's
492 * VTX feature has been checked already, no need check again.
494 static inline int mdd_is_sticky(const struct lu_env *env,
495 struct mdd_object *pobj,
496 const struct lu_attr *pattr,
497 struct mdd_object *cobj,
498 const struct lu_attr *cattr)
500 struct lu_ucred *uc = lu_ucred_assert(env);
503 LASSERT(pattr != NULL);
504 if (!(pattr->la_mode & S_ISVTX) ||
505 (pattr->la_uid == uc->uc_fsuid))
509 LASSERT(cattr != NULL);
510 if (cattr->la_uid == uc->uc_fsuid)
513 return !md_capable(uc, CAP_FOWNER);
516 static int mdd_may_delete_entry(const struct lu_env *env,
517 struct mdd_object *pobj,
518 const struct lu_attr *pattr,
523 LASSERT(pobj != NULL);
524 if (!mdd_object_exists(pobj))
527 if (mdd_is_dead_obj(pobj))
532 rc = mdd_permission_internal_locked(env, pobj, pattr,
533 MAY_WRITE | MAY_EXEC,
539 if (pattr->la_flags & LUSTRE_APPEND_FL)
546 * Check whether it may delete the cobj from the pobj.
549 int mdd_may_delete(const struct lu_env *env, struct mdd_object *tpobj,
550 const struct lu_attr *tpattr, struct mdd_object *tobj,
551 const struct lu_attr *tattr, const struct lu_attr *cattr,
552 int check_perm, int check_empty)
558 LASSERT(tpattr != NULL);
559 rc = mdd_may_delete_entry(env, tpobj, tpattr, check_perm);
567 if (!mdd_object_exists(tobj))
570 if (mdd_is_dead_obj(tobj))
573 if (mdd_is_sticky(env, tpobj, tpattr, tobj, tattr))
576 if (tattr->la_flags & (LUSTRE_APPEND_FL | LUSTRE_IMMUTABLE_FL))
579 /* additional check the rename case */
581 if (S_ISDIR(cattr->la_mode)) {
582 if (!S_ISDIR(tattr->la_mode))
585 if (mdd_is_root(mdo2mdd(&tobj->mod_obj),
586 mdd_object_fid(tobj)))
588 } else if (S_ISDIR(tattr->la_mode))
592 if (S_ISDIR(tattr->la_mode) && check_empty)
593 rc = mdd_dir_is_empty(env, tobj);
599 * Check whether it can create the link file(linked to @src_obj) under
600 * the target directory(@tgt_obj), and src_obj has been locked by
603 * \param[in] env execution environment
604 * \param[in] tgt_obj the target directory
605 * \param[in] tattr attributes of target directory
606 * \param[in] lname the link name
607 * \param[in] src_obj source object for link
608 * \param[in] cattr attributes for source object
610 * \retval = 0 it is allowed to create the link file under tgt_obj
611 * \retval negative error not allowed to create the link file
613 static int mdd_link_sanity_check(const struct lu_env *env,
614 struct mdd_object *tgt_obj,
615 const struct lu_attr *tattr,
616 const struct lu_name *lname,
617 struct mdd_object *src_obj,
618 const struct lu_attr *cattr)
620 struct mdd_device *m = mdd_obj2mdd_dev(src_obj);
624 if (!mdd_object_exists(src_obj))
627 if (mdd_is_dead_obj(src_obj))
630 /* Local ops, no lookup before link, check filename length here. */
631 rc = mdd_name_check(m, lname);
635 if (cattr->la_flags & (LUSTRE_IMMUTABLE_FL | LUSTRE_APPEND_FL))
638 if (S_ISDIR(mdd_object_type(src_obj)))
641 LASSERT(src_obj != tgt_obj);
642 rc = mdd_may_create(env, tgt_obj, tattr, NULL, true);
646 rc = __mdd_may_link(env, src_obj, cattr);
651 static int __mdd_index_delete_only(const struct lu_env *env, struct mdd_object *pobj,
652 const char *name, struct thandle *handle)
654 struct dt_object *next = mdd_object_child(pobj);
658 if (dt_try_as_dir(env, next))
659 rc = dt_delete(env, next, (struct dt_key *)name, handle);
666 static int __mdd_index_insert_only(const struct lu_env *env,
667 struct mdd_object *pobj,
668 const struct lu_fid *lf, __u32 type,
669 const char *name, struct thandle *handle)
671 struct dt_object *next = mdd_object_child(pobj);
675 if (dt_try_as_dir(env, next)) {
676 struct dt_insert_rec *rec = &mdd_env_info(env)->mdi_dt_rec;
679 rec->rec_type = type;
680 rc = dt_insert(env, next, (const struct dt_rec *)rec,
681 (const struct dt_key *)name, handle);
688 /* insert named index, add reference if isdir */
689 static int __mdd_index_insert(const struct lu_env *env, struct mdd_object *pobj,
690 const struct lu_fid *lf, __u32 type,
691 const char *name, struct thandle *handle)
696 rc = __mdd_index_insert_only(env, pobj, lf, type, name, handle);
697 if (rc == 0 && S_ISDIR(type)) {
698 mdd_write_lock(env, pobj, DT_TGT_PARENT);
699 mdo_ref_add(env, pobj, handle);
700 mdd_write_unlock(env, pobj);
706 /* delete named index, drop reference if isdir */
707 static int __mdd_index_delete(const struct lu_env *env, struct mdd_object *pobj,
708 const char *name, int is_dir,
709 struct thandle *handle)
714 rc = __mdd_index_delete_only(env, pobj, name, handle);
715 if (rc == 0 && is_dir) {
716 mdd_write_lock(env, pobj, DT_TGT_PARENT);
717 mdo_ref_del(env, pobj, handle);
718 mdd_write_unlock(env, pobj);
724 static int mdd_llog_record_calc_size(const struct lu_env *env,
725 const struct lu_name *tname,
726 const struct lu_name *sname)
728 const struct lu_ucred *uc = lu_ucred(env);
729 enum changelog_rec_flags clf_flags = CLF_EXTRA_FLAGS;
730 enum changelog_rec_extra_flags crfe = CLFE_UIDGID | CLFE_NID;
733 clf_flags |= CLF_RENAME;
735 if (uc != NULL && uc->uc_jobid[0] != '\0')
736 clf_flags |= CLF_JOBID;
738 return llog_data_len(LLOG_CHANGELOG_HDR_SZ +
739 changelog_rec_offset(clf_flags, crfe) +
740 (tname != NULL ? tname->ln_namelen : 0) +
741 (sname != NULL ? 1 + sname->ln_namelen : 0));
744 int mdd_declare_changelog_store(const struct lu_env *env,
745 struct mdd_device *mdd,
746 enum changelog_rec_type type,
747 const struct lu_name *tname,
748 const struct lu_name *sname,
749 struct thandle *handle)
751 struct obd_device *obd = mdd2obd_dev(mdd);
752 struct llog_ctxt *ctxt;
753 struct llog_rec_hdr rec_hdr;
754 struct thandle *llog_th;
757 if (!mdd_changelog_enabled(env, mdd, type))
760 rec_hdr.lrh_len = mdd_llog_record_calc_size(env, tname, sname);
761 rec_hdr.lrh_type = CHANGELOG_REC;
763 ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
767 llog_th = thandle_get_sub(env, handle, ctxt->loc_handle->lgh_obj);
769 GOTO(out_put, rc = PTR_ERR(llog_th));
771 rc = llog_declare_add(env, ctxt->loc_handle, &rec_hdr, llog_th);
779 int mdd_changelog_write_rec(const struct lu_env *env,
780 struct llog_handle *loghandle,
781 struct llog_rec_hdr *r,
782 struct llog_cookie *cookie,
783 int idx, struct thandle *th)
787 if (r->lrh_type == CHANGELOG_REC) {
788 struct mdd_device *mdd;
789 struct llog_changelog_rec *rec;
791 mdd = lu2mdd_dev(loghandle->lgh_ctxt->loc_obd->obd_lu_dev);
792 rec = container_of(r, struct llog_changelog_rec, cr_hdr);
794 spin_lock(&mdd->mdd_cl.mc_lock);
795 rec->cr.cr_index = mdd->mdd_cl.mc_index + 1;
796 spin_unlock(&mdd->mdd_cl.mc_lock);
798 rc = llog_osd_ops.lop_write_rec(env, loghandle, r,
802 * if current llog is full, we will generate a new
803 * llog, and since it's actually not an error, let's
804 * avoid increasing index so that userspace apps
805 * should not see a gap in the changelog sequence
807 if (!(rc == -ENOSPC && llog_is_full(loghandle))) {
808 spin_lock(&mdd->mdd_cl.mc_lock);
809 ++mdd->mdd_cl.mc_index;
810 spin_unlock(&mdd->mdd_cl.mc_lock);
813 rc = llog_osd_ops.lop_write_rec(env, loghandle, r,
820 /** Add a changelog entry \a rec to the changelog llog
823 * \param handle - currently ignored since llogs start their own transaction;
824 * this will hopefully be fixed in llog rewrite
827 int mdd_changelog_store(const struct lu_env *env, struct mdd_device *mdd,
828 struct llog_changelog_rec *rec, struct thandle *th)
830 struct obd_device *obd = mdd2obd_dev(mdd);
831 struct llog_ctxt *ctxt;
832 struct thandle *llog_th;
835 rec->cr_hdr.lrh_len = llog_data_len(sizeof(*rec) +
836 changelog_rec_varsize(&rec->cr));
838 /* llog_lvfs_write_rec sets the llog tail len */
839 rec->cr_hdr.lrh_type = CHANGELOG_REC;
840 rec->cr.cr_time = cl_time();
842 ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
846 llog_th = thandle_get_sub(env, th, ctxt->loc_handle->lgh_obj);
848 GOTO(out_put, rc = PTR_ERR(llog_th));
850 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_CHANGELOG_REORDER, cfs_fail_val);
851 /* nested journal transaction */
852 rc = llog_add(env, ctxt->loc_handle, &rec->cr_hdr, NULL, llog_th);
854 /* time to recover some space ?? */
855 if (likely(!mdd->mdd_changelog_gc ||
856 mdd->mdd_cl.mc_gc_task != MDD_CHLG_GC_NONE ||
857 mdd->mdd_changelog_min_gc_interval >=
858 ktime_get_real_seconds() - mdd->mdd_cl.mc_gc_time))
859 /* save a spin_lock trip */
861 spin_lock(&mdd->mdd_cl.mc_lock);
862 if (likely(mdd->mdd_changelog_gc &&
863 mdd->mdd_cl.mc_gc_task == MDD_CHLG_GC_NONE &&
864 ktime_get_real_seconds() - mdd->mdd_cl.mc_gc_time >
865 mdd->mdd_changelog_min_gc_interval)) {
866 if (unlikely(llog_cat_free_space(ctxt->loc_handle) <=
867 mdd->mdd_changelog_min_free_cat_entries ||
868 OBD_FAIL_CHECK(OBD_FAIL_FORCE_GC_THREAD))) {
869 CWARN("%s:%s low on changelog_catalog free entries, "
870 "starting ChangeLog garbage collection thread\n",
872 OBD_FAIL_CHECK(OBD_FAIL_FORCE_GC_THREAD) ?
875 /* indicate further kthread run will occur outside
876 * right after current journal transaction filling has
879 mdd->mdd_cl.mc_gc_task = MDD_CHLG_GC_NEED;
881 /* next check in mdd_changelog_min_gc_interval anyway
883 mdd->mdd_cl.mc_gc_time = ktime_get_real_seconds();
885 spin_unlock(&mdd->mdd_cl.mc_lock);
893 static void mdd_changelog_rec_ext_rename(struct changelog_rec *rec,
894 const struct lu_fid *sfid,
895 const struct lu_fid *spfid,
896 const struct lu_name *sname)
898 struct changelog_ext_rename *rnm = changelog_rec_rename(rec);
901 LASSERT(sfid != NULL);
902 LASSERT(spfid != NULL);
903 LASSERT(sname != NULL);
905 extsize = sname->ln_namelen + 1;
907 rnm->cr_sfid = *sfid;
908 rnm->cr_spfid = *spfid;
910 changelog_rec_name(rec)[rec->cr_namelen] = '\0';
911 strlcpy(changelog_rec_sname(rec), sname->ln_name, extsize);
912 rec->cr_namelen += extsize;
915 void mdd_changelog_rec_ext_jobid(struct changelog_rec *rec, const char *jobid)
917 struct changelog_ext_jobid *jid = changelog_rec_jobid(rec);
919 if (jobid == NULL || jobid[0] == '\0')
922 strlcpy(jid->cr_jobid, jobid, sizeof(jid->cr_jobid));
925 void mdd_changelog_rec_ext_extra_flags(struct changelog_rec *rec, __u64 eflags)
927 struct changelog_ext_extra_flags *ef = changelog_rec_extra_flags(rec);
929 ef->cr_extra_flags = eflags;
932 void mdd_changelog_rec_extra_uidgid(struct changelog_rec *rec,
933 __u64 uid, __u64 gid)
935 struct changelog_ext_uidgid *uidgid = changelog_rec_uidgid(rec);
937 uidgid->cr_uid = uid;
938 uidgid->cr_gid = gid;
941 void mdd_changelog_rec_extra_nid(struct changelog_rec *rec,
944 struct changelog_ext_nid *clnid = changelog_rec_nid(rec);
949 void mdd_changelog_rec_extra_omode(struct changelog_rec *rec, u32 flags)
951 struct changelog_ext_openmode *omd = changelog_rec_openmode(rec);
953 omd->cr_openflags = flags;
956 void mdd_changelog_rec_extra_xattr(struct changelog_rec *rec,
957 const char *xattr_name)
959 struct changelog_ext_xattr *xattr = changelog_rec_xattr(rec);
961 strlcpy(xattr->cr_xattr, xattr_name, sizeof(xattr->cr_xattr));
964 /** Store a namespace change changelog record
965 * If this fails, we must fail the whole transaction; we don't
966 * want the change to commit without the log entry.
967 * \param target - mdd_object of change
968 * \param tpfid - target parent dir/object fid
969 * \param sfid - source object fid
970 * \param spfid - source parent fid
971 * \param tname - target name string
972 * \param sname - source name string
973 * \param handle - transaction handle
975 int mdd_changelog_ns_store(const struct lu_env *env,
976 struct mdd_device *mdd,
977 enum changelog_rec_type type,
978 enum changelog_rec_flags clf_flags,
979 struct mdd_object *target,
980 const struct lu_fid *tpfid,
981 const struct lu_fid *sfid,
982 const struct lu_fid *spfid,
983 const struct lu_name *tname,
984 const struct lu_name *sname,
985 struct thandle *handle)
987 const struct lu_ucred *uc = lu_ucred(env);
988 struct llog_changelog_rec *rec;
991 __u64 xflags = CLFE_INVALID;
995 if (!mdd_changelog_enabled(env, mdd, type))
998 LASSERT(tpfid != NULL);
999 LASSERT(tname != NULL);
1000 LASSERT(handle != NULL);
1002 reclen = mdd_llog_record_calc_size(env, tname, sname);
1003 buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mdi_chlg_buf, reclen);
1004 if (buf->lb_buf == NULL)
1008 clf_flags &= CLF_FLAGMASK;
1009 clf_flags |= CLF_EXTRA_FLAGS;
1012 if (uc->uc_jobid[0] != '\0')
1013 clf_flags |= CLF_JOBID;
1014 xflags |= CLFE_UIDGID;
1019 clf_flags |= CLF_RENAME;
1021 clf_flags |= CLF_VERSION;
1023 rec->cr.cr_flags = clf_flags;
1025 if (clf_flags & CLF_EXTRA_FLAGS) {
1026 mdd_changelog_rec_ext_extra_flags(&rec->cr, xflags);
1027 if (xflags & CLFE_UIDGID)
1028 mdd_changelog_rec_extra_uidgid(&rec->cr,
1029 uc->uc_uid, uc->uc_gid);
1030 if (xflags & CLFE_NID)
1031 mdd_changelog_rec_extra_nid(&rec->cr, uc->uc_nid);
1034 rec->cr.cr_type = (__u32)type;
1035 rec->cr.cr_pfid = *tpfid;
1036 rec->cr.cr_namelen = tname->ln_namelen;
1037 memcpy(changelog_rec_name(&rec->cr), tname->ln_name, tname->ln_namelen);
1039 if (clf_flags & CLF_RENAME)
1040 mdd_changelog_rec_ext_rename(&rec->cr, sfid, spfid, sname);
1042 if (clf_flags & CLF_JOBID)
1043 mdd_changelog_rec_ext_jobid(&rec->cr, uc->uc_jobid);
1045 if (likely(target != NULL)) {
1046 rec->cr.cr_tfid = *mdd_object_fid(target);
1047 target->mod_cltime = ktime_get();
1049 fid_zero(&rec->cr.cr_tfid);
1052 rc = mdd_changelog_store(env, mdd, rec, handle);
1054 CERROR("%s: cannot store changelog record: type = %d, "
1055 "name = '%s', t = "DFID", p = "DFID": rc = %d\n",
1056 mdd2obd_dev(mdd)->obd_name, type, tname->ln_name,
1057 PFID(&rec->cr.cr_tfid), PFID(&rec->cr.cr_pfid), rc);
1064 static int __mdd_links_add(const struct lu_env *env,
1065 struct mdd_object *mdd_obj,
1066 struct linkea_data *ldata,
1067 const struct lu_name *lname,
1068 const struct lu_fid *pfid,
1069 int first, int check)
1071 /* cattr is set in mdd_link */
1072 struct lu_attr *cattr = MDD_ENV_VAR(env, cattr);
1075 if (ldata->ld_leh == NULL) {
1076 rc = first ? -ENODATA : mdd_links_read(env, mdd_obj, ldata);
1080 rc = linkea_data_new(ldata,
1081 &mdd_env_info(env)->mdi_link_buf);
1088 rc = linkea_links_find(ldata, lname, pfid);
1089 if (rc && rc != -ENOENT)
1095 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_LINKEA_MORE)) {
1096 struct lu_fid *tfid = &mdd_env_info(env)->mdi_fid2;
1100 linkea_add_buf(ldata, lname, tfid, false);
1103 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_LINKEA_MORE2))
1104 linkea_add_buf(ldata, lname, pfid, false);
1106 /* For encrypted file, we want to limit number of hard links to what
1107 * linkEA can contain. So ask to return error in case of overflow.
1108 * Currently linkEA stores 4KiB of links, that is 14 NAME_MAX links,
1109 * or 119 16-byte names.
1111 return linkea_add_buf(ldata, lname, pfid,
1112 cattr->la_valid & LA_FLAGS &&
1113 cattr->la_flags & LUSTRE_ENCRYPT_FL);
1116 static int __mdd_links_del(const struct lu_env *env,
1117 struct mdd_object *mdd_obj,
1118 struct linkea_data *ldata,
1119 const struct lu_name *lname,
1120 const struct lu_fid *pfid)
1122 /* cattr is set in mdd_link */
1123 struct lu_attr *cattr = MDD_ENV_VAR(env, cattr);
1126 if (ldata->ld_leh == NULL) {
1127 rc = mdd_links_read(env, mdd_obj, ldata);
1132 rc = linkea_links_find(ldata, lname, pfid);
1136 linkea_del_buf(ldata, lname,
1137 cattr->la_valid & LA_FLAGS &&
1138 cattr->la_flags & LUSTRE_ENCRYPT_FL);
1142 static int mdd_linkea_prepare(const struct lu_env *env,
1143 struct mdd_object *mdd_obj,
1144 const struct lu_fid *oldpfid,
1145 const struct lu_name *oldlname,
1146 const struct lu_fid *newpfid,
1147 const struct lu_name *newlname,
1148 int first, int check,
1149 struct linkea_data *ldata)
1154 if (OBD_FAIL_CHECK(OBD_FAIL_FID_IGIF))
1157 LASSERT(oldpfid != NULL || newpfid != NULL);
1159 if (mdd_obj->mod_flags & DEAD_OBJ)
1160 /* Unnecessary to update linkEA for dead object. */
1163 if (oldpfid != NULL) {
1164 rc = __mdd_links_del(env, mdd_obj, ldata, oldlname, oldpfid);
1166 if ((check == 1) || (rc != -ENODATA && rc != -ENOENT))
1169 /* No changes done. */
1174 /* If renaming, add the new record */
1175 if (newpfid != NULL)
1176 rc = __mdd_links_add(env, mdd_obj, ldata, newlname, newpfid,
1182 int mdd_links_rename(const struct lu_env *env,
1183 struct mdd_object *mdd_obj,
1184 const struct lu_fid *oldpfid,
1185 const struct lu_name *oldlname,
1186 const struct lu_fid *newpfid,
1187 const struct lu_name *newlname,
1188 struct thandle *handle,
1189 struct linkea_data *ldata,
1190 int first, int check)
1195 if (ldata == NULL) {
1196 ldata = &mdd_env_info(env)->mdi_link_data;
1197 memset(ldata, 0, sizeof(*ldata));
1198 rc = mdd_linkea_prepare(env, mdd_obj, oldpfid, oldlname,
1199 newpfid, newlname, first, check, ldata);
1204 if (!(mdd_obj->mod_flags & DEAD_OBJ))
1205 rc = mdd_links_write(env, mdd_obj, ldata, handle);
1211 if (newlname == NULL)
1212 CERROR("link_ea add failed %d "DFID"\n",
1213 rc, PFID(mdd_object_fid(mdd_obj)));
1214 else if (oldpfid == NULL)
1215 CERROR("link_ea add '%.*s' failed %d "DFID"\n",
1216 newlname->ln_namelen, newlname->ln_name, rc,
1217 PFID(mdd_object_fid(mdd_obj)));
1218 else if (newpfid == NULL)
1219 CERROR("link_ea del '%.*s' failed %d "DFID"\n",
1220 oldlname->ln_namelen, oldlname->ln_name, rc,
1221 PFID(mdd_object_fid(mdd_obj)));
1223 CERROR("link_ea rename '%.*s'->'%.*s' failed %d "DFID
1224 "\n", oldlname->ln_namelen, oldlname->ln_name,
1225 newlname->ln_namelen, newlname->ln_name, rc,
1226 PFID(mdd_object_fid(mdd_obj)));
1229 if (is_vmalloc_addr(ldata->ld_buf))
1230 /* if we vmalloced a large buffer drop it */
1231 lu_buf_free(ldata->ld_buf);
1236 static inline int mdd_links_add(const struct lu_env *env,
1237 struct mdd_object *mdd_obj,
1238 const struct lu_fid *pfid,
1239 const struct lu_name *lname,
1240 struct thandle *handle,
1241 struct linkea_data *ldata, int first)
1243 return mdd_links_rename(env, mdd_obj, NULL, NULL,
1244 pfid, lname, handle, ldata, first, 0);
1247 static inline int mdd_links_del(const struct lu_env *env,
1248 struct mdd_object *mdd_obj,
1249 const struct lu_fid *pfid,
1250 const struct lu_name *lname,
1251 struct thandle *handle)
1253 return mdd_links_rename(env, mdd_obj, pfid, lname,
1254 NULL, NULL, handle, NULL, 0, 0);
1257 /** Read the link EA into a temp buffer.
1258 * Uses the name_buf since it is generally large.
1259 * \retval IS_ERR err
1260 * \retval ptr to \a lu_buf (always \a mdi_link_buf)
1262 struct lu_buf *mdd_links_get(const struct lu_env *env,
1263 struct mdd_object *mdd_obj)
1265 struct linkea_data ldata = { NULL };
1268 rc = mdd_links_read(env, mdd_obj, &ldata);
1269 return rc ? ERR_PTR(rc) : ldata.ld_buf;
1272 int mdd_links_write(const struct lu_env *env, struct mdd_object *mdd_obj,
1273 struct linkea_data *ldata, struct thandle *handle)
1275 const struct lu_buf *buf;
1278 if (ldata == NULL || ldata->ld_buf == NULL ||
1279 ldata->ld_leh == NULL)
1282 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_LINKEA))
1286 buf = mdd_buf_get_const(env, ldata->ld_buf->lb_buf,
1287 ldata->ld_leh->leh_len);
1288 rc = mdo_xattr_set(env, mdd_obj, buf, XATTR_NAME_LINK, 0, handle);
1289 if (unlikely(rc == -ENOSPC)) {
1290 rc = linkea_overflow_shrink(ldata);
1298 static int mdd_declare_links_add(const struct lu_env *env,
1299 struct mdd_object *mdd_obj,
1300 struct thandle *handle,
1301 struct linkea_data *ldata)
1307 if (ldata != NULL && ldata->ld_leh != NULL) {
1308 ea_len = ldata->ld_leh->leh_len;
1309 linkea = ldata->ld_buf->lb_buf;
1311 ea_len = MAX_LINKEA_SIZE;
1315 rc = mdo_declare_xattr_set(env, mdd_obj,
1316 mdd_buf_get_const(env, linkea, ea_len),
1317 XATTR_NAME_LINK, 0, handle);
1322 static inline int mdd_declare_links_del(const struct lu_env *env,
1323 struct mdd_object *c,
1324 struct thandle *handle)
1328 /* For directory, the linkEA will be removed together
1329 * with the object. */
1330 if (!S_ISDIR(mdd_object_type(c)))
1331 rc = mdd_declare_links_add(env, c, handle, NULL);
1336 static int mdd_declare_link(const struct lu_env *env,
1337 struct mdd_device *mdd,
1338 struct mdd_object *p,
1339 struct mdd_object *c,
1340 const struct lu_name *name,
1341 struct thandle *handle,
1343 struct linkea_data *data)
1345 struct lu_fid tfid = *mdd_object_fid(c);
1348 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING3))
1349 tfid.f_oid = cfs_fail_val;
1351 rc = mdo_declare_index_insert(env, p, &tfid, mdd_object_type(c),
1352 name->ln_name, handle);
1356 rc = mdo_declare_ref_add(env, c, handle);
1360 la->la_valid = LA_CTIME | LA_MTIME;
1361 rc = mdo_declare_attr_set(env, p, la, handle);
1365 la->la_valid = LA_CTIME;
1366 rc = mdo_declare_attr_set(env, c, la, handle);
1370 rc = mdd_declare_links_add(env, c, handle, data);
1374 rc = mdd_declare_changelog_store(env, mdd, CL_HARDLINK, name, NULL,
1380 static int mdd_link(const struct lu_env *env, struct md_object *tgt_obj,
1381 struct md_object *src_obj, const struct lu_name *lname,
1384 const char *name = lname->ln_name;
1385 struct lu_attr *la = &mdd_env_info(env)->mdi_la_for_fix;
1386 struct mdd_object *mdd_tobj = md2mdd_obj(tgt_obj);
1387 struct mdd_object *mdd_sobj = md2mdd_obj(src_obj);
1388 struct lu_attr *cattr = MDD_ENV_VAR(env, cattr);
1389 struct lu_attr *tattr = MDD_ENV_VAR(env, tattr);
1390 struct mdd_device *mdd = mdo2mdd(src_obj);
1391 struct thandle *handle;
1392 struct lu_fid *tfid = &mdd_env_info(env)->mdi_fid2;
1393 struct linkea_data *ldata = &mdd_env_info(env)->mdi_link_data;
1397 rc = mdd_la_get(env, mdd_sobj, cattr);
1401 rc = mdd_la_get(env, mdd_tobj, tattr);
1406 * If we are using project inheritance, we only allow hard link
1407 * creation in our tree when the project IDs are the same;
1408 * otherwise the tree quota mechanism could be circumvented.
1410 if ((tattr->la_flags & LUSTRE_PROJINHERIT_FL) &&
1411 (tattr->la_projid != cattr->la_projid))
1414 handle = mdd_trans_create(env, mdd);
1416 GOTO(out_pending, rc = PTR_ERR(handle));
1418 memset(ldata, 0, sizeof(*ldata));
1420 LASSERT(ma->ma_attr.la_valid & LA_CTIME);
1421 la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime;
1423 /* Note: even this function will change ldata, but it comes from
1424 * thread_info, which is completely temporary and only seen in
1425 * this function, so we do not need reset ldata once it fails.*/
1426 rc = mdd_linkea_prepare(env, mdd_sobj, NULL, NULL,
1427 mdd_object_fid(mdd_tobj), lname, 0, 0, ldata);
1431 rc = mdd_declare_link(env, mdd, mdd_tobj, mdd_sobj, lname, handle,
1436 rc = mdd_trans_start(env, mdd, handle);
1440 mdd_write_lock(env, mdd_sobj, DT_TGT_CHILD);
1441 rc = mdd_link_sanity_check(env, mdd_tobj, tattr, lname, mdd_sobj,
1444 GOTO(out_unlock, rc);
1446 if (!OBD_FAIL_CHECK(OBD_FAIL_LFSCK_LESS_NLINK)) {
1447 rc = mdo_ref_add(env, mdd_sobj, handle);
1449 GOTO(out_unlock, rc);
1452 *tfid = *mdd_object_fid(mdd_sobj);
1453 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING3))
1454 tfid->f_oid = cfs_fail_val;
1456 rc = __mdd_index_insert_only(env, mdd_tobj, tfid,
1457 mdd_object_type(mdd_sobj), name, handle);
1459 mdo_ref_del(env, mdd_sobj, handle);
1460 GOTO(out_unlock, rc);
1463 la->la_valid = LA_CTIME | LA_MTIME;
1464 rc = mdd_update_time(env, mdd_tobj, tattr, la, handle);
1466 GOTO(out_unlock, rc);
1468 la->la_valid = LA_CTIME;
1469 rc = mdd_update_time(env, mdd_sobj, cattr, la, handle);
1471 /* Note: The failure of links_add should not cause the
1472 * link failure, so do not check return value. */
1473 mdd_links_add(env, mdd_sobj, mdd_object_fid(mdd_tobj),
1474 lname, handle, ldata, 0);
1478 mdd_write_unlock(env, mdd_sobj);
1480 rc = mdd_changelog_ns_store(env, mdd, CL_HARDLINK, 0, mdd_sobj,
1481 mdd_object_fid(mdd_tobj), NULL,
1482 NULL, lname, NULL, handle);
1484 rc = mdd_trans_stop(env, mdd, rc, handle);
1485 if (is_vmalloc_addr(ldata->ld_buf))
1486 /* if we vmalloced a large buffer drop it */
1487 lu_buf_free(ldata->ld_buf);
1492 static int mdd_mark_orphan_object(const struct lu_env *env,
1493 struct mdd_object *obj, struct thandle *handle,
1496 struct lu_attr *attr = MDD_ENV_VAR(env, la_for_start);
1499 attr->la_valid = LA_FLAGS;
1500 attr->la_flags = LUSTRE_ORPHAN_FL;
1503 rc = mdo_declare_attr_set(env, obj, attr, handle);
1505 rc = mdo_attr_set(env, obj, attr, handle);
1510 static int mdd_declare_finish_unlink(const struct lu_env *env,
1511 struct mdd_object *obj,
1512 struct thandle *handle)
1516 /* Sigh, we do not know if the unlink object will become orphan in
1517 * declare phase, but fortunately the flags here does not matter
1518 * in current declare implementation */
1519 rc = mdd_mark_orphan_object(env, obj, handle, true);
1523 rc = mdo_declare_destroy(env, obj, handle);
1527 rc = mdd_orphan_declare_insert(env, obj, mdd_object_type(obj), handle);
1531 return mdd_declare_links_del(env, obj, handle);
1534 /* caller should take a lock before calling */
1535 int mdd_finish_unlink(const struct lu_env *env,
1536 struct mdd_object *obj, struct md_attr *ma,
1537 struct mdd_object *pobj,
1538 const struct lu_name *lname,
1542 int is_dir = S_ISDIR(ma->ma_attr.la_mode);
1545 LASSERT(mdd_write_locked(env, obj) != 0);
1547 if (ma->ma_attr.la_nlink == 0 || is_dir) {
1548 /* add new orphan and the object
1549 * will be deleted during mdd_close() */
1550 obj->mod_flags |= DEAD_OBJ;
1551 if (obj->mod_count) {
1552 rc = mdd_orphan_insert(env, obj, th);
1554 CDEBUG(D_HA, "Object "DFID" is inserted into "
1555 "orphan list, open count = %d\n",
1556 PFID(mdd_object_fid(obj)),
1559 CERROR("Object "DFID" fail to be an orphan, "
1560 "open count = %d, maybe cause failed "
1562 PFID(mdd_object_fid(obj)),
1565 /* mark object as an orphan here, not
1566 * before mdd_orphan_insert() as racing
1567 * mdd_la_get() may propagate ORPHAN_OBJ
1568 * causing the asserition */
1569 rc = mdd_mark_orphan_object(env, obj, th, false);
1571 rc = mdo_destroy(env, obj, th);
1573 } else if (!is_dir) {
1574 /* old files may not have link ea; ignore errors */
1575 mdd_links_del(env, obj, mdd_object_fid(pobj), lname, th);
1583 * has mdd_write_lock on cobj already, but not on pobj yet
1585 int mdd_unlink_sanity_check(const struct lu_env *env, struct mdd_object *pobj,
1586 const struct lu_attr *pattr,
1587 struct mdd_object *cobj,
1588 const struct lu_attr *cattr)
1593 rc = mdd_may_delete(env, pobj, pattr, cobj, cattr, NULL, 1, 1);
1598 static int mdd_declare_unlink(const struct lu_env *env, struct mdd_device *mdd,
1599 struct mdd_object *p, struct mdd_object *c,
1600 const struct lu_name *name, struct md_attr *ma,
1601 struct thandle *handle, int no_name, int is_dir)
1603 struct lu_attr *la = &mdd_env_info(env)->mdi_la_for_fix;
1606 if (!OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING2)) {
1607 if (likely(no_name == 0)) {
1608 rc = mdo_declare_index_delete(env, p, name->ln_name,
1615 rc = mdo_declare_ref_del(env, p, handle);
1621 LASSERT(ma->ma_attr.la_valid & LA_CTIME);
1622 la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime;
1623 la->la_valid = LA_CTIME | LA_MTIME;
1624 rc = mdo_declare_attr_set(env, p, la, handle);
1629 rc = mdo_declare_ref_del(env, c, handle);
1633 rc = mdo_declare_ref_del(env, c, handle);
1637 la->la_valid = LA_CTIME;
1638 rc = mdo_declare_attr_set(env, c, la, handle);
1642 rc = mdd_declare_finish_unlink(env, c, handle);
1646 /* FIXME: need changelog for remove entry */
1647 rc = mdd_declare_changelog_store(env, mdd, CL_UNLINK, name,
1655 * test if a file has an HSM archive
1656 * if HSM attributes are not found in ma update them from
1659 static bool mdd_hsm_archive_exists(const struct lu_env *env,
1660 struct mdd_object *obj,
1665 if (!(ma->ma_valid & MA_HSM)) {
1666 /* no HSM MD provided, read xattr */
1667 struct lu_buf *hsm_buf;
1668 const size_t buflen = sizeof(struct hsm_attrs);
1671 hsm_buf = mdd_buf_get(env, NULL, 0);
1672 lu_buf_alloc(hsm_buf, buflen);
1673 rc = mdo_xattr_get(env, obj, hsm_buf, XATTR_NAME_HSM);
1674 rc = lustre_buf2hsm(hsm_buf->lb_buf, rc, &ma->ma_hsm);
1675 lu_buf_free(hsm_buf);
1679 ma->ma_valid |= MA_HSM;
1681 if (ma->ma_hsm.mh_flags & HS_EXISTS)
1687 * Delete name entry and the object.
1688 * Note: no_name == 1 means it only destory the object, i.e. name_entry
1689 * does not exist for this object, and it could only happen during resending
1690 * of remote unlink. see the comments in mdt_reint_unlink. Unfortunately, lname
1691 * is also needed in this case(needed by changelog), so we have to add another
1692 * parameter(no_name)here. XXX: this is only needed in DNE phase I, on Phase II,
1693 * the ENOENT failure should be able to be fixed by redo mechanism.
1695 static int mdd_unlink(const struct lu_env *env, struct md_object *pobj,
1696 struct md_object *cobj, const struct lu_name *lname,
1697 struct md_attr *ma, int no_name)
1699 const char *name = lname->ln_name;
1700 struct lu_attr *pattr = MDD_ENV_VAR(env, pattr);
1701 struct lu_attr *cattr = MDD_ENV_VAR(env, cattr);
1702 struct lu_attr *la = &mdd_env_info(env)->mdi_la_for_fix;
1703 struct mdd_object *mdd_pobj = md2mdd_obj(pobj);
1704 struct mdd_object *mdd_cobj = NULL;
1705 struct mdd_device *mdd = mdo2mdd(pobj);
1706 struct thandle *handle;
1707 int rc, is_dir = 0, cl_flags = 0;
1710 /* let shutdown to start */
1711 CFS_FAIL_TIMEOUT(OBD_FAIL_TGT_REPLY_DATA_RACE, 1);
1713 /* cobj == NULL means only delete name entry */
1714 if (likely(cobj != NULL)) {
1715 mdd_cobj = md2mdd_obj(cobj);
1716 if (mdd_object_exists(mdd_cobj) == 0)
1720 rc = mdd_la_get(env, mdd_pobj, pattr);
1724 if (likely(mdd_cobj != NULL)) {
1726 rc = mdd_la_get(env, mdd_cobj, cattr);
1730 is_dir = S_ISDIR(cattr->la_mode);
1731 /* search for an existing archive.
1732 * we should check ahead as the object
1733 * can be destroyed in this transaction */
1734 if (mdd_hsm_archive_exists(env, mdd_cobj, ma))
1735 cl_flags |= CLF_UNLINK_HSM_EXISTS;
1738 rc = mdd_unlink_sanity_check(env, mdd_pobj, pattr, mdd_cobj, cattr);
1742 handle = mdd_trans_create(env, mdd);
1744 RETURN(PTR_ERR(handle));
1746 rc = mdd_declare_unlink(env, mdd, mdd_pobj, mdd_cobj,
1747 lname, ma, handle, no_name, is_dir);
1751 rc = mdd_trans_start(env, mdd, handle);
1755 if (likely(mdd_cobj != NULL))
1756 mdd_write_lock(env, mdd_cobj, DT_TGT_CHILD);
1758 if (likely(no_name == 0) && !OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING2)) {
1759 rc = __mdd_index_delete(env, mdd_pobj, name, is_dir, handle);
1764 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_MUL_REF) ||
1765 OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_NAMEENTRY))
1766 GOTO(cleanup, rc = 0);
1768 if (likely(mdd_cobj != NULL)) {
1769 rc = mdo_ref_del(env, mdd_cobj, handle);
1771 __mdd_index_insert_only(env, mdd_pobj,
1772 mdd_object_fid(mdd_cobj),
1773 mdd_object_type(mdd_cobj),
1780 mdo_ref_del(env, mdd_cobj, handle);
1782 /* fetch updated nlink */
1783 rc = mdd_la_get(env, mdd_cobj, cattr);
1788 LASSERT(ma->ma_attr.la_valid & LA_CTIME);
1789 la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime;
1791 la->la_valid = LA_CTIME | LA_MTIME;
1792 rc = mdd_update_time(env, mdd_pobj, pattr, la, handle);
1796 /* Enough for only unlink the entry */
1797 if (unlikely(mdd_cobj == NULL))
1800 if (cattr->la_nlink > 0 || mdd_cobj->mod_count > 0) {
1801 /* update ctime of an unlinked file only if it is still
1802 * opened or a link still exists */
1803 la->la_valid = LA_CTIME;
1804 rc = mdd_update_time(env, mdd_cobj, cattr, la, handle);
1809 /* XXX: this transfer to ma will be removed with LOD/OSP */
1810 ma->ma_attr = *cattr;
1811 ma->ma_valid |= MA_INODE;
1812 rc = mdd_finish_unlink(env, mdd_cobj, ma, mdd_pobj, lname, handle);
1816 /* fetch updated nlink */
1817 rc = mdd_la_get(env, mdd_cobj, cattr);
1818 /* if object is removed then we can't get its attrs,
1820 if (rc == -ENOENT) {
1821 cattr->la_nlink = 0;
1825 if (cattr->la_nlink == 0) {
1826 ma->ma_attr = *cattr;
1827 ma->ma_valid |= MA_INODE;
1832 if (likely(mdd_cobj != NULL))
1833 mdd_write_unlock(env, mdd_cobj);
1836 if (cattr->la_nlink == 0)
1837 cl_flags |= CLF_UNLINK_LAST;
1839 cl_flags &= ~CLF_UNLINK_HSM_EXISTS;
1841 rc = mdd_changelog_ns_store(env, mdd,
1842 is_dir ? CL_RMDIR : CL_UNLINK, cl_flags,
1843 mdd_cobj, mdd_object_fid(mdd_pobj), NULL, NULL,
1844 lname, NULL, handle);
1848 rc = mdd_trans_stop(env, mdd, rc, handle);
1854 * The permission has been checked when obj created, no need check again.
1856 static int mdd_cd_sanity_check(const struct lu_env *env,
1857 struct mdd_object *obj)
1862 if (!obj || mdd_is_dead_obj(obj))
1868 static int mdd_create_data(const struct lu_env *env, struct md_object *pobj,
1869 struct md_object *cobj,
1870 const struct md_op_spec *spec, struct md_attr *ma)
1872 struct mdd_device *mdd = mdo2mdd(cobj);
1873 struct mdd_object *mdd_pobj = md2mdd_obj(pobj);
1874 struct mdd_object *son = md2mdd_obj(cobj);
1875 struct thandle *handle;
1876 const struct lu_buf *buf;
1877 struct lu_attr *attr = MDD_ENV_VAR(env, cattr);
1878 struct dt_allocation_hint *hint = &mdd_env_info(env)->mdi_hint;
1882 rc = mdd_cd_sanity_check(env, son);
1886 if (!md_should_create(spec->sp_cr_flags))
1890 * there are following use cases for this function:
1891 * 1) late striping - file was created with MDS_OPEN_DELAY_CREATE
1892 * striping can be specified or not
1895 rc = mdd_la_get(env, son, attr);
1899 /* calling ->ah_make_hint() is used to transfer information from parent */
1900 mdd_object_make_hint(env, mdd_pobj, son, attr, spec, hint);
1902 handle = mdd_trans_create(env, mdd);
1904 GOTO(out_free, rc = PTR_ERR(handle));
1907 * XXX: Setting the lov ea is not locked but setting the attr is locked?
1908 * Should this be fixed?
1910 CDEBUG(D_OTHER, "ea %p/%u, cr_flags %#llo, no_create %u\n",
1911 spec->u.sp_ea.eadata, spec->u.sp_ea.eadatalen,
1912 spec->sp_cr_flags, spec->no_create);
1914 if (spec->no_create || (spec->sp_cr_flags & MDS_OPEN_HAS_EA)) {
1915 buf = mdd_buf_get_const(env, spec->u.sp_ea.eadata,
1916 spec->u.sp_ea.eadatalen);
1921 rc = dt_declare_xattr_set(env, mdd_object_child(son), buf,
1922 XATTR_NAME_LOV, 0, handle);
1926 rc = mdd_declare_changelog_store(env, mdd, CL_LAYOUT, NULL, NULL,
1931 rc = mdd_trans_start(env, mdd, handle);
1935 rc = dt_xattr_set(env, mdd_object_child(son), buf, XATTR_NAME_LOV,
1941 rc = mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, son, handle,
1945 rc = mdd_trans_stop(env, mdd, rc, handle);
1951 static int mdd_declare_object_initialize(const struct lu_env *env,
1952 struct mdd_object *parent,
1953 struct mdd_object *child,
1954 const struct lu_attr *attr,
1955 struct thandle *handle)
1960 LASSERT(attr->la_valid & (LA_MODE | LA_TYPE));
1961 if (!S_ISDIR(attr->la_mode))
1964 rc = mdo_declare_index_insert(env, child, mdd_object_fid(child),
1965 S_IFDIR, dot, handle);
1969 rc = mdo_declare_ref_add(env, child, handle);
1973 rc = mdo_declare_index_insert(env, child, mdd_object_fid(parent),
1974 S_IFDIR, dotdot, handle);
1979 static int mdd_object_initialize(const struct lu_env *env,
1980 const struct lu_fid *pfid,
1981 struct mdd_object *child,
1982 struct lu_attr *attr,
1983 struct thandle *handle)
1988 if (S_ISDIR(attr->la_mode)) {
1989 /* Add "." and ".." for newly created dir */
1990 mdo_ref_add(env, child, handle);
1991 rc = __mdd_index_insert_only(env, child, mdd_object_fid(child),
1992 S_IFDIR, dot, handle);
1994 rc = __mdd_index_insert_only(env, child, pfid, S_IFDIR,
1997 mdo_ref_del(env, child, handle);
2004 * This function checks whether it can create a file/dir under the
2005 * directory(@pobj). The directory(@pobj) is not being locked by
2008 * \param[in] env execution environment
2009 * \param[in] pobj the directory to create files
2010 * \param[in] pattr the attributes of the directory
2011 * \param[in] lname the name of the created file/dir
2012 * \param[in] cattr the attributes of the file/dir
2013 * \param[in] spec create specification
2015 * \retval = 0 it is allowed to create file/dir under
2017 * \retval negative error not allowed to create file/dir
2018 * under the directory
2020 static int mdd_create_sanity_check(const struct lu_env *env,
2021 struct md_object *pobj,
2022 const struct lu_attr *pattr,
2023 const struct lu_name *lname,
2024 struct lu_attr *cattr,
2025 struct md_op_spec *spec)
2027 struct mdd_thread_info *info = mdd_env_info(env);
2028 struct lu_fid *fid = &info->mdi_fid;
2029 struct mdd_object *obj = md2mdd_obj(pobj);
2030 struct mdd_device *m = mdo2mdd(pobj);
2031 bool check_perm = true;
2036 if (mdd_is_dead_obj(obj))
2040 * In some cases this lookup is not needed - we know before if name
2041 * exists or not because MDT performs lookup for it.
2042 * name length check is done in lookup.
2044 if (spec->sp_cr_lookup) {
2046 * Check if the name already exist, though it will be checked in
2047 * _index_insert also, for avoiding rolling back if exists
2050 rc = __mdd_lookup(env, pobj, pattr, lname, fid,
2051 MAY_WRITE | MAY_EXEC);
2053 RETURN(rc ? : -EEXIST);
2055 /* Permission is already being checked in mdd_lookup */
2059 if (S_ISDIR(cattr->la_mode) &&
2060 unlikely(spec != NULL && spec->sp_cr_flags & MDS_OPEN_HAS_EA) &&
2061 spec->u.sp_ea.eadata != NULL && spec->u.sp_ea.eadatalen > 0) {
2062 const struct lmv_user_md *lum = spec->u.sp_ea.eadata;
2064 if (!lmv_user_magic_supported(le32_to_cpu(lum->lum_magic)) &&
2065 le32_to_cpu(lum->lum_magic) != LMV_USER_MAGIC_V0) {
2067 CERROR("%s: invalid lmv_user_md: magic = %x, "
2068 "stripe_offset = %d, stripe_count = %u: "
2069 "rc = %d\n", mdd2obd_dev(m)->obd_name,
2070 le32_to_cpu(lum->lum_magic),
2071 (int)le32_to_cpu(lum->lum_stripe_offset),
2072 le32_to_cpu(lum->lum_stripe_count), rc);
2077 rc = mdd_may_create(env, obj, pattr, NULL, check_perm);
2082 if (pattr->la_mode & S_ISGID) {
2083 struct lu_ucred *uc = lu_ucred(env);
2085 cattr->la_gid = pattr->la_gid;
2087 /* Directories are special, and always inherit S_ISGID */
2088 if (S_ISDIR(cattr->la_mode)) {
2089 cattr->la_mode |= S_ISGID;
2090 cattr->la_valid |= LA_MODE;
2091 } else if ((cattr->la_mode & (S_ISGID | S_IXGRP))
2092 == (S_ISGID | S_IXGRP) &&
2093 !lustre_in_group_p(uc,
2094 (cattr->la_valid & LA_GID) ?
2095 cattr->la_gid : pattr->la_gid) &&
2096 !md_capable(uc, CAP_FSETID)) {
2097 cattr->la_mode &= ~S_ISGID;
2098 cattr->la_valid |= LA_MODE;
2102 /* Inherit project ID from parent directory */
2103 if (pattr->la_flags & LUSTRE_PROJINHERIT_FL) {
2104 cattr->la_projid = pattr->la_projid;
2105 if (S_ISDIR(cattr->la_mode)) {
2106 cattr->la_flags |= LUSTRE_PROJINHERIT_FL;
2107 cattr->la_valid |= LA_FLAGS;
2109 cattr->la_valid |= LA_PROJID;
2112 rc = mdd_name_check(m, lname);
2116 switch (cattr->la_mode & S_IFMT) {
2118 unsigned int symlen = strlen(spec->u.sp_symname) + 1;
2120 if (symlen > m->mdd_dt_conf.ddp_symlink_max)
2121 RETURN(-ENAMETOOLONG);
2140 static int mdd_declare_create_object(const struct lu_env *env,
2141 struct mdd_device *mdd,
2142 struct mdd_object *p, struct mdd_object *c,
2143 struct lu_attr *attr,
2144 struct thandle *handle,
2145 const struct md_op_spec *spec,
2146 struct lu_buf *def_acl_buf,
2147 struct lu_buf *acl_buf,
2148 struct lu_buf *hsm_buf,
2149 struct dt_allocation_hint *hint)
2151 const struct lu_buf *buf;
2154 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
2155 /* ldiskfs OSD needs this information for credit allocation */
2157 hint->dah_acl_len = def_acl_buf->lb_len;
2159 rc = mdd_declare_create_object_internal(env, p, c, attr, handle, spec,
2164 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
2165 if (def_acl_buf && def_acl_buf->lb_len > 0 && S_ISDIR(attr->la_mode)) {
2166 /* if dir, then can inherit default ACl */
2167 rc = mdo_declare_xattr_set(env, c, def_acl_buf,
2168 XATTR_NAME_ACL_DEFAULT,
2174 if (acl_buf && acl_buf->lb_len > 0) {
2175 rc = mdo_declare_attr_set(env, c, attr, handle);
2179 rc = mdo_declare_xattr_set(env, c, acl_buf,
2180 XATTR_NAME_ACL_ACCESS, 0, handle);
2185 rc = mdd_declare_object_initialize(env, p, c, attr, handle);
2189 /* replay case, create LOV EA from client data */
2190 if ((!(spec->sp_cr_flags & MDS_OPEN_DELAY_CREATE) && spec->no_create) ||
2191 (spec->sp_cr_flags & MDS_OPEN_HAS_EA && S_ISREG(attr->la_mode))) {
2192 buf = mdd_buf_get_const(env, spec->u.sp_ea.eadata,
2193 spec->u.sp_ea.eadatalen);
2194 rc = mdo_declare_xattr_set(env, c, buf,
2195 S_ISDIR(attr->la_mode) ?
2196 XATTR_NAME_LMV : XATTR_NAME_LOV,
2197 LU_XATTR_CREATE, handle);
2201 if (spec->sp_cr_flags & MDS_OPEN_PCC) {
2202 rc = mdo_declare_xattr_set(env, c, hsm_buf,
2210 if (S_ISLNK(attr->la_mode)) {
2211 const char *target_name = spec->u.sp_symname;
2212 int sym_len = strlen(target_name);
2213 const struct lu_buf *buf;
2215 buf = mdd_buf_get_const(env, target_name, sym_len);
2216 rc = dt_declare_record_write(env, mdd_object_child(c),
2222 if (spec->sp_cr_file_secctx_name != NULL) {
2223 buf = mdd_buf_get_const(env, spec->sp_cr_file_secctx,
2224 spec->sp_cr_file_secctx_size);
2225 rc = mdo_declare_xattr_set(env, c, buf,
2226 spec->sp_cr_file_secctx_name, 0,
2232 if (spec->sp_cr_file_encctx != NULL) {
2233 buf = mdd_buf_get_const(env, spec->sp_cr_file_encctx,
2234 spec->sp_cr_file_encctx_size);
2235 rc = mdo_declare_xattr_set(env, c, buf,
2236 LL_XATTR_NAME_ENCRYPTION_CONTEXT, 0,
2245 static int mdd_declare_create(const struct lu_env *env, struct mdd_device *mdd,
2246 struct mdd_object *p, struct mdd_object *c,
2247 const struct lu_name *name,
2248 struct lu_attr *attr,
2249 struct thandle *handle,
2250 const struct md_op_spec *spec,
2251 struct linkea_data *ldata,
2252 struct lu_buf *def_acl_buf,
2253 struct lu_buf *acl_buf,
2254 struct lu_buf *hsm_buf,
2255 struct dt_allocation_hint *hint)
2259 rc = mdd_declare_create_object(env, mdd, p, c, attr, handle, spec,
2260 def_acl_buf, acl_buf, hsm_buf, hint);
2264 if (S_ISDIR(attr->la_mode)) {
2265 rc = mdo_declare_ref_add(env, p, handle);
2270 if (unlikely(spec->sp_cr_flags & MDS_OPEN_VOLATILE)) {
2271 rc = mdd_orphan_declare_insert(env, c, attr->la_mode, handle);
2275 struct lu_attr *la = &mdd_env_info(env)->mdi_la_for_fix;
2276 enum changelog_rec_type type;
2278 rc = mdo_declare_index_insert(env, p, mdd_object_fid(c),
2279 attr->la_mode, name->ln_name,
2284 rc = mdd_declare_links_add(env, c, handle, ldata);
2289 la->la_valid = LA_CTIME | LA_MTIME;
2290 rc = mdo_declare_attr_set(env, p, la, handle);
2294 type = S_ISDIR(attr->la_mode) ? CL_MKDIR :
2295 S_ISREG(attr->la_mode) ? CL_CREATE :
2296 S_ISLNK(attr->la_mode) ? CL_SOFTLINK : CL_MKNOD;
2298 rc = mdd_declare_changelog_store(env, mdd, type, name, NULL,
2307 static int mdd_acl_init(const struct lu_env *env, struct mdd_object *pobj,
2308 struct lu_attr *la, struct lu_buf *def_acl_buf,
2309 struct lu_buf *acl_buf)
2315 if (S_ISLNK(la->la_mode)) {
2316 acl_buf->lb_len = 0;
2317 def_acl_buf->lb_len = 0;
2321 mdd_read_lock(env, pobj, DT_TGT_PARENT);
2322 rc = mdo_xattr_get(env, pobj, def_acl_buf,
2323 XATTR_NAME_ACL_DEFAULT);
2324 mdd_read_unlock(env, pobj);
2326 /* ACL buffer size is not enough, need realloc */
2327 if (rc > acl_buf->lb_len)
2330 /* If there are default ACL, fix mode/ACL by default ACL */
2331 def_acl_buf->lb_len = rc;
2332 memcpy(acl_buf->lb_buf, def_acl_buf->lb_buf, rc);
2333 acl_buf->lb_len = rc;
2334 rc = __mdd_fix_mode_acl(env, acl_buf, &la->la_mode);
2337 } else if (rc == -ENODATA || rc == -EOPNOTSUPP) {
2338 /* If there are no default ACL, fix mode by mask */
2339 struct lu_ucred *uc = lu_ucred(env);
2341 /* The create triggered by MDT internal events, such as
2342 * LFSCK reset, will not contain valid "uc". */
2343 if (unlikely(uc != NULL))
2344 la->la_mode &= ~uc->uc_umask;
2346 acl_buf->lb_len = 0;
2347 def_acl_buf->lb_len = 0;
2354 * Create a metadata object and initialize it, set acl, xattr.
2356 static int mdd_create_object(const struct lu_env *env, struct mdd_object *pobj,
2357 struct mdd_object *son, struct lu_attr *attr,
2358 struct md_op_spec *spec, struct lu_buf *acl_buf,
2359 struct lu_buf *def_acl_buf,
2360 struct lu_buf *hsm_buf,
2361 struct dt_allocation_hint *hint,
2362 struct thandle *handle, bool initsecctx)
2364 const struct lu_buf *buf;
2367 mdd_write_lock(env, son, DT_TGT_CHILD);
2368 rc = mdd_create_object_internal(env, NULL, son, attr, handle, spec,
2373 /* Note: In DNE phase I, for striped dir, though sub-stripes will be
2374 * created in declare phase, they also needs to be added to master
2375 * object as sub-directory entry. So it has to initialize the master
2376 * object, then set dir striped EA.(in mdo_xattr_set) */
2377 rc = mdd_object_initialize(env, mdd_object_fid(pobj), son, attr,
2380 GOTO(err_destroy, rc);
2383 * in case of replay we just set LOVEA provided by the client
2384 * XXX: I think it would be interesting to try "old" way where
2385 * MDT calls this xattr_set(LOV) in a different transaction.
2386 * probably this way we code can be made better.
2389 /* During creation, there are only a few cases we need do xattr_set to
2391 * 1. regular file: see comments above.
2392 * 2. dir: inherit default striping or pool settings from parent.
2393 * 3. create striped directory with provided stripeEA.
2394 * 4. create striped directory because inherit default layout from the
2397 if (spec->no_create ||
2398 (S_ISREG(attr->la_mode) && spec->sp_cr_flags & MDS_OPEN_HAS_EA) ||
2399 S_ISDIR(attr->la_mode)) {
2400 buf = mdd_buf_get_const(env, spec->u.sp_ea.eadata,
2401 spec->u.sp_ea.eadatalen);
2402 rc = mdo_xattr_set(env, son, buf,
2403 S_ISDIR(attr->la_mode) ? XATTR_NAME_LMV :
2405 LU_XATTR_CREATE, handle);
2407 GOTO(err_destroy, rc);
2410 if (S_ISREG(attr->la_mode) && spec->sp_cr_flags & MDS_OPEN_PCC) {
2413 memset(&mh, 0, sizeof(mh));
2414 mh.mh_flags = HS_EXISTS | HS_ARCHIVED | HS_RELEASED;
2415 mh.mh_arch_id = spec->sp_archive_id;
2416 lustre_hsm2buf(hsm_buf->lb_buf, &mh);
2417 rc = mdo_xattr_set(env, son, hsm_buf, XATTR_NAME_HSM,
2420 GOTO(err_destroy, rc);
2423 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
2424 if (def_acl_buf != NULL && def_acl_buf->lb_len > 0 &&
2425 S_ISDIR(attr->la_mode)) {
2426 /* set default acl */
2427 rc = mdo_xattr_set(env, son, def_acl_buf,
2428 XATTR_NAME_ACL_DEFAULT, 0,
2431 GOTO(err_destroy, rc);
2433 /* set its own acl */
2434 if (acl_buf != NULL && acl_buf->lb_len > 0) {
2435 rc = mdo_xattr_set(env, son, acl_buf,
2436 XATTR_NAME_ACL_ACCESS,
2439 GOTO(err_destroy, rc);
2443 if (S_ISLNK(attr->la_mode)) {
2444 struct dt_object *dt = mdd_object_child(son);
2445 const char *target_name = spec->u.sp_symname;
2446 int sym_len = strlen(target_name);
2449 buf = mdd_buf_get_const(env, target_name, sym_len);
2450 rc = dt->do_body_ops->dbo_write(env, dt, buf, &pos, handle);
2454 GOTO(err_initlized, rc = -EFAULT);
2457 if (initsecctx && spec->sp_cr_file_secctx_name != NULL) {
2458 buf = mdd_buf_get_const(env, spec->sp_cr_file_secctx,
2459 spec->sp_cr_file_secctx_size);
2460 rc = mdo_xattr_set(env, son, buf, spec->sp_cr_file_secctx_name,
2463 GOTO(err_initlized, rc);
2466 if (spec->sp_cr_file_encctx != NULL) {
2467 buf = mdd_buf_get_const(env, spec->sp_cr_file_encctx,
2468 spec->sp_cr_file_encctx_size);
2469 rc = mdo_xattr_set(env, son, buf,
2470 LL_XATTR_NAME_ENCRYPTION_CONTEXT, 0,
2473 GOTO(err_initlized, rc);
2477 if (unlikely(rc != 0)) {
2479 if (S_ISDIR(attr->la_mode)) {
2480 /* Drop the reference, no need to delete "."/"..",
2481 * because the object to be destroied directly. */
2482 rc2 = mdo_ref_del(env, son, handle);
2486 rc2 = mdo_ref_del(env, son, handle);
2490 mdo_destroy(env, son, handle);
2493 mdd_write_unlock(env, son);
2497 static int mdd_index_delete(const struct lu_env *env,
2498 struct mdd_object *mdd_pobj,
2499 struct lu_attr *cattr,
2500 const struct lu_name *lname)
2502 struct mdd_device *mdd = mdo2mdd(&mdd_pobj->mod_obj);
2503 struct thandle *handle;
2507 handle = mdd_trans_create(env, mdd);
2509 RETURN(PTR_ERR(handle));
2511 rc = mdo_declare_index_delete(env, mdd_pobj, lname->ln_name,
2516 if (S_ISDIR(cattr->la_mode)) {
2517 rc = mdo_declare_ref_del(env, mdd_pobj, handle);
2522 /* Since this will only be used in the error handler path,
2523 * Let's set the thandle to be local and not mess the transno */
2524 handle->th_local = 1;
2525 rc = mdd_trans_start(env, mdd, handle);
2529 rc = __mdd_index_delete(env, mdd_pobj, lname->ln_name,
2530 S_ISDIR(cattr->la_mode), handle);
2534 rc = mdd_trans_stop(env, mdd, rc, handle);
2540 * Create object and insert it into namespace.
2542 * Two operations have to be performed:
2544 * - an allocation of a new object (->do_create()), and
2545 * - an insertion into a parent index (->dio_insert()).
2547 * Due to locking, operation order is not important, when both are
2548 * successful, *but* error handling cases are quite different:
2550 * - if insertion is done first, and following object creation fails,
2551 * insertion has to be rolled back, but this operation might fail
2552 * also leaving us with dangling index entry.
2554 * - if creation is done first, is has to be undone if insertion fails,
2555 * leaving us with leaked space, which is not good but not fatal.
2557 * It seems that creation-first is simplest solution, but it is sub-optimal
2563 * case, because second mkdir is bound to create object, only to
2564 * destroy it immediately.
2566 * To avoid this follow local file systems that do double lookup:
2568 * 0. lookup -> -EEXIST (mdd_create_sanity_check())
2569 * 1. create (mdd_create_object_internal())
2570 * 2. insert (__mdd_index_insert(), lookup again)
2572 * \param[in] pobj parent object
2573 * \param[in] lname name of child being created
2574 * \param[in,out] child child object being created
2575 * \param[in] spec additional create parameters
2576 * \param[in] ma attributes for new child object
2578 * \retval 0 on success
2579 * \retval negative errno on failure
2581 int mdd_create(const struct lu_env *env, struct md_object *pobj,
2582 const struct lu_name *lname, struct md_object *child,
2583 struct md_op_spec *spec, struct md_attr *ma)
2585 struct mdd_thread_info *info = mdd_env_info(env);
2586 struct lu_attr *la = &info->mdi_la_for_fix;
2587 struct mdd_object *mdd_pobj = md2mdd_obj(pobj);
2588 struct mdd_object *son = md2mdd_obj(child);
2589 struct mdd_device *mdd = mdo2mdd(pobj);
2590 struct lu_attr *attr = &ma->ma_attr;
2591 struct thandle *handle;
2592 struct lu_attr *pattr = &info->mdi_pattr;
2593 struct lu_buf acl_buf;
2594 struct lu_buf def_acl_buf;
2595 struct lu_buf hsm_buf;
2596 struct linkea_data *ldata = &info->mdi_link_data;
2597 const char *name = lname->ln_name;
2598 struct dt_allocation_hint *hint = &mdd_env_info(env)->mdi_hint;
2599 int acl_size = LUSTRE_POSIX_ACL_MAX_SIZE_OLD;
2604 rc = mdd_la_get(env, mdd_pobj, pattr);
2608 /* Sanity checks before big job. */
2609 rc = mdd_create_sanity_check(env, pobj, pattr, lname, attr, spec);
2613 if (OBD_FAIL_CHECK(OBD_FAIL_MDS_DQACQ_NET))
2614 GOTO(out_free, rc = -EINPROGRESS);
2616 handle = mdd_trans_create(env, mdd);
2618 GOTO(out_free, rc = PTR_ERR(handle));
2621 acl_buf = *lu_buf_check_and_alloc(&info->mdi_xattr_buf, acl_size);
2622 if (!acl_buf.lb_buf)
2623 GOTO(out_stop, rc = -ENOMEM);
2625 def_acl_buf = *lu_buf_check_and_alloc(&info->mdi_big_buf, acl_size);
2626 if (!def_acl_buf.lb_buf)
2627 GOTO(out_stop, rc = -ENOMEM);
2629 rc = mdd_acl_init(env, mdd_pobj, attr, &def_acl_buf, &acl_buf);
2630 if (unlikely(rc == -ERANGE &&
2631 acl_size == LUSTRE_POSIX_ACL_MAX_SIZE_OLD)) {
2632 /* use maximum-sized xattr buffer for too-big default ACL */
2633 acl_size = min_t(unsigned int, mdd->mdd_dt_conf.ddp_max_ea_size,
2635 goto use_bigger_buffer;
2640 if (S_ISDIR(attr->la_mode)) {
2641 struct lmv_user_md *lmu = spec->u.sp_ea.eadata;
2644 * migrate may create 1-stripe directory, so lod_ah_init()
2645 * doesn't adjust stripe count from lmu.
2647 if (lmu && lmu->lum_stripe_count == cpu_to_le32(1))
2648 lmu->lum_stripe_count = 0;
2651 mdd_object_make_hint(env, mdd_pobj, son, attr, spec, hint);
2653 memset(ldata, 0, sizeof(*ldata));
2654 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_BAD_PARENT)) {
2655 struct lu_fid tfid = *mdd_object_fid(mdd_pobj);
2658 rc = mdd_linkea_prepare(env, son, NULL, NULL,
2659 &tfid, lname, 1, 0, ldata);
2661 rc = mdd_linkea_prepare(env, son, NULL, NULL,
2662 mdd_object_fid(mdd_pobj),
2663 lname, 1, 0, ldata);
2666 if (spec->sp_cr_flags & MDS_OPEN_PCC) {
2667 LASSERT(spec->sp_cr_flags & MDS_OPEN_HAS_EA);
2669 memset(&hsm_buf, 0, sizeof(hsm_buf));
2670 lu_buf_alloc(&hsm_buf, sizeof(struct hsm_attrs));
2671 if (hsm_buf.lb_buf == NULL)
2672 GOTO(out_stop, rc = -ENOMEM);
2675 rc = mdd_declare_create(env, mdd, mdd_pobj, son, lname, attr,
2676 handle, spec, ldata, &def_acl_buf, &acl_buf,
2681 rc = mdd_trans_start(env, mdd, handle);
2685 rc = mdd_create_object(env, mdd_pobj, son, attr, spec, &acl_buf,
2686 &def_acl_buf, &hsm_buf, hint, handle, true);
2690 if (unlikely(spec->sp_cr_flags & MDS_OPEN_VOLATILE)) {
2691 mdd_write_lock(env, son, DT_TGT_CHILD);
2692 son->mod_flags |= VOLATILE_OBJ;
2693 rc = mdd_orphan_insert(env, son, handle);
2694 GOTO(out_volatile, rc);
2696 rc = __mdd_index_insert(env, mdd_pobj, mdd_object_fid(son),
2697 attr->la_mode, name, handle);
2699 GOTO(err_created, rc);
2701 mdd_links_add(env, son, mdd_object_fid(mdd_pobj), lname,
2704 /* update parent directory mtime/ctime */
2706 la->la_valid = LA_CTIME | LA_MTIME;
2707 rc = mdd_update_time(env, mdd_pobj, pattr, la, handle);
2709 GOTO(err_insert, rc);
2715 if (spec->sp_cr_flags & MDS_OPEN_VOLATILE)
2716 rc2 = mdd_orphan_delete(env, son, handle);
2718 rc2 = __mdd_index_delete(env, mdd_pobj, name,
2719 S_ISDIR(attr->la_mode),
2725 mdd_write_lock(env, son, DT_TGT_CHILD);
2726 if (S_ISDIR(attr->la_mode)) {
2727 /* Drop the reference, no need to delete "."/"..",
2728 * because the object is to be destroyed directly. */
2729 rc2 = mdo_ref_del(env, son, handle);
2731 mdd_write_unlock(env, son);
2736 /* For volatile files drop one link immediately, since there is
2737 * no filename in the namespace, and save any error returned. */
2738 rc2 = mdo_ref_del(env, son, handle);
2740 mdd_write_unlock(env, son);
2741 if (unlikely(rc == 0))
2746 /* Don't destroy the volatile object on success */
2747 if (likely(rc != 0))
2748 mdo_destroy(env, son, handle);
2749 mdd_write_unlock(env, son);
2752 if (rc == 0 && fid_is_namespace_visible(mdd_object_fid(son)) &&
2753 likely((spec->sp_cr_flags & MDS_OPEN_VOLATILE) == 0))
2754 rc = mdd_changelog_ns_store(env, mdd,
2755 S_ISDIR(attr->la_mode) ? CL_MKDIR :
2756 S_ISREG(attr->la_mode) ? CL_CREATE :
2757 S_ISLNK(attr->la_mode) ? CL_SOFTLINK : CL_MKNOD,
2758 0, son, mdd_object_fid(mdd_pobj), NULL, NULL,
2759 lname, NULL, handle);
2761 rc2 = mdd_trans_stop(env, mdd, rc, handle);
2763 /* If creation fails, it is most likely due to the remote update
2764 * failure, because local transaction will mostly succeed at
2765 * this stage. There is no easy way to rollback all of previous
2766 * updates, so let's remove the object from namespace, and
2767 * LFSCK should handle the orphan object. */
2768 if (rc2 < 0 && !mdd_object_remote(mdd_pobj))
2769 mdd_index_delete(env, mdd_pobj, attr, lname);
2773 if (is_vmalloc_addr(ldata->ld_buf))
2774 /* if we vmalloced a large buffer drop it */
2775 lu_buf_free(ldata->ld_buf);
2777 if (spec->sp_cr_flags & MDS_OPEN_PCC)
2778 lu_buf_free(&hsm_buf);
2780 /* The child object shouldn't be cached anymore */
2782 set_bit(LU_OBJECT_HEARD_BANSHEE,
2783 &child->mo_lu.lo_header->loh_flags);
2787 /* has not mdd_write{read}_lock on any obj yet. */
2788 static int mdd_rename_sanity_check(const struct lu_env *env,
2789 struct mdd_object *src_pobj,
2790 const struct lu_attr *pattr,
2791 struct mdd_object *tgt_pobj,
2792 const struct lu_attr *tpattr,
2793 struct mdd_object *sobj,
2794 const struct lu_attr *cattr,
2795 struct mdd_object *tobj,
2796 const struct lu_attr *tattr)
2801 /* XXX: when get here, sobj must NOT be NULL,
2802 * the other case has been processed in cld_rename
2803 * before mdd_rename and enable MDS_PERM_BYPASS. */
2807 * If we are using project inheritance, we only allow renames
2808 * into our tree when the project IDs are the same; otherwise
2809 * tree quota mechanism would be circumvented.
2811 if (((tpattr->la_flags & LUSTRE_PROJINHERIT_FL) &&
2812 tpattr->la_projid != cattr->la_projid) ||
2813 ((pattr->la_flags & LUSTRE_PROJINHERIT_FL) &&
2814 (pattr->la_projid != tpattr->la_projid)))
2817 /* we prevent an encrypted file from being renamed
2818 * into an unencrypted dir
2820 if ((cattr->la_valid & LA_FLAGS &&
2821 cattr->la_flags & LUSTRE_ENCRYPT_FL) &&
2822 !(tpattr->la_valid & LA_FLAGS &&
2823 tpattr->la_flags & LUSTRE_ENCRYPT_FL))
2826 rc = mdd_may_delete(env, src_pobj, pattr, sobj, cattr, NULL, 1, 0);
2830 /* XXX: when get here, "tobj == NULL" means tobj must
2831 * NOT exist (neither on remote MDS, such case has been
2832 * processed in cld_rename before mdd_rename and enable
2834 * So check may_create, but not check may_unlink. */
2836 rc = mdd_may_create(env, tgt_pobj, tpattr, NULL,
2837 (src_pobj != tgt_pobj));
2839 rc = mdd_may_delete(env, tgt_pobj, tpattr, tobj, tattr, cattr,
2840 (src_pobj != tgt_pobj), 1);
2842 if (!rc && !tobj && (src_pobj != tgt_pobj) && S_ISDIR(cattr->la_mode))
2843 rc = __mdd_may_link(env, tgt_pobj, tpattr);
2848 static int mdd_declare_rename(const struct lu_env *env,
2849 struct mdd_device *mdd,
2850 struct mdd_object *mdd_spobj,
2851 struct mdd_object *mdd_tpobj,
2852 struct mdd_object *mdd_sobj,
2853 struct mdd_object *mdd_tobj,
2854 const struct lu_name *sname,
2855 const struct lu_name *tname,
2857 struct linkea_data *ldata,
2858 struct thandle *handle)
2860 struct lu_attr *la = &mdd_env_info(env)->mdi_la_for_fix;
2863 LASSERT(ma->ma_attr.la_valid & LA_CTIME);
2864 la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime;
2870 /* name from source dir */
2871 rc = mdo_declare_index_delete(env, mdd_spobj, sname->ln_name, handle);
2875 /* .. from source child */
2876 if (S_ISDIR(mdd_object_type(mdd_sobj))) {
2877 /* source child can be directory, count by source dir's nlink */
2878 rc = mdo_declare_ref_del(env, mdd_spobj, handle);
2881 if (mdd_spobj != mdd_tpobj) {
2882 rc = mdo_declare_index_delete(env, mdd_sobj, dotdot,
2887 rc = mdo_declare_index_insert(env, mdd_sobj,
2888 mdd_object_fid(mdd_tpobj),
2889 S_IFDIR, dotdot, handle);
2894 /* new target child can be directory,
2895 * counted by target dir's nlink */
2896 rc = mdo_declare_ref_add(env, mdd_tpobj, handle);
2901 la->la_valid = LA_CTIME | LA_MTIME;
2902 rc = mdo_declare_attr_set(env, mdd_spobj, la, handle);
2906 rc = mdo_declare_attr_set(env, mdd_tpobj, la, handle);
2910 la->la_valid = LA_CTIME;
2911 rc = mdo_declare_attr_set(env, mdd_sobj, la, handle);
2915 rc = mdd_declare_links_add(env, mdd_sobj, handle, ldata);
2920 rc = mdo_declare_index_insert(env, mdd_tpobj, mdd_object_fid(mdd_sobj),
2921 mdd_object_type(mdd_sobj),
2922 tname->ln_name, handle);
2926 if (mdd_tobj && mdd_object_exists(mdd_tobj)) {
2927 /* delete target child in target parent directory */
2928 rc = mdo_declare_index_delete(env, mdd_tpobj, tname->ln_name,
2933 rc = mdo_declare_ref_del(env, mdd_tobj, handle);
2937 if (S_ISDIR(mdd_object_type(mdd_tobj))) {
2938 /* target child can be directory,
2939 * delete "." reference in target child directory */
2940 rc = mdo_declare_ref_del(env, mdd_tobj, handle);
2944 /* delete ".." reference in target parent directory */
2945 rc = mdo_declare_ref_del(env, mdd_tpobj, handle);
2950 la->la_valid = LA_CTIME;
2951 rc = mdo_declare_attr_set(env, mdd_tobj, la, handle);
2955 rc = mdd_declare_finish_unlink(env, mdd_tobj, handle);
2960 rc = mdd_declare_changelog_store(env, mdd, CL_RENAME, tname, sname,
2968 static int mdd_migrate_object(const struct lu_env *env,
2969 struct mdd_object *spobj,
2970 struct mdd_object *tpobj,
2971 struct mdd_object *sobj,
2972 struct mdd_object *tobj,
2973 const struct lu_name *sname,
2974 const struct lu_name *tname,
2975 struct md_op_spec *spec,
2976 struct md_attr *ma);
2978 /* src object can be remote that is why we use only fid and type of object */
2979 static int mdd_rename(const struct lu_env *env,
2980 struct md_object *src_pobj, struct md_object *tgt_pobj,
2981 const struct lu_fid *lf, const struct lu_name *lsname,
2982 struct md_object *tobj, const struct lu_name *ltname,
2985 const char *sname = lsname->ln_name;
2986 const char *tname = ltname->ln_name;
2987 struct lu_attr *la = &mdd_env_info(env)->mdi_la_for_fix;
2988 struct mdd_object *mdd_spobj = md2mdd_obj(src_pobj); /* source parent */
2989 struct mdd_object *mdd_tpobj = md2mdd_obj(tgt_pobj);
2990 struct mdd_device *mdd = mdo2mdd(src_pobj);
2991 struct mdd_object *mdd_sobj = NULL; /* source object */
2992 struct mdd_object *mdd_tobj = NULL;
2993 struct lu_attr *cattr = MDD_ENV_VAR(env, cattr);
2994 struct lu_attr *pattr = MDD_ENV_VAR(env, pattr);
2995 struct lu_attr *tattr = MDD_ENV_VAR(env, tattr);
2996 struct lu_attr *tpattr = MDD_ENV_VAR(env, tpattr);
2997 struct thandle *handle;
2998 struct linkea_data *ldata = &mdd_env_info(env)->mdi_link_data;
2999 const struct lu_fid *tpobj_fid = mdd_object_fid(mdd_tpobj);
3000 const struct lu_fid *spobj_fid = mdd_object_fid(mdd_spobj);
3003 bool tobj_locked = 0;
3004 unsigned cl_flags = 0;
3008 /* let unlink to complete and commit */
3009 CFS_FAIL_TIMEOUT(OBD_FAIL_TGT_REPLY_DATA_RACE, 2 + cfs_fail_val);
3012 mdd_tobj = md2mdd_obj(tobj);
3014 mdd_sobj = mdd_object_find(env, mdd, lf);
3015 if (IS_ERR(mdd_sobj))
3016 RETURN(PTR_ERR(mdd_sobj));
3018 rc = mdd_la_get(env, mdd_sobj, cattr);
3020 GOTO(out_pending, rc);
3022 /* if rename is cross MDTs, migrate symlink if it doesn't have other
3023 * hard links, and target doesn't exist.
3025 if (mdd_object_remote(mdd_sobj) && S_ISLNK(cattr->la_mode) &&
3026 cattr->la_nlink == 1 && !tobj) {
3027 struct md_op_spec *spec = &mdd_env_info(env)->mdi_spec;
3028 struct lu_device *ld = &mdd->mdd_md_dev.md_lu_dev;
3031 rc = ld->ld_ops->ldo_fid_alloc(env, ld, &tfid, &tgt_pobj->mo_lu,
3034 GOTO(out_pending, rc);
3036 mdd_tobj = mdd_object_find(env, mdd, &tfid);
3037 if (IS_ERR(mdd_tobj))
3038 GOTO(out_pending, rc = PTR_ERR(mdd_tobj));
3040 memset(spec, 0, sizeof(*spec));
3041 rc = mdd_migrate_object(env, mdd_spobj, mdd_tpobj, mdd_sobj,
3042 mdd_tobj, lsname, ltname, spec, ma);
3043 mdd_object_put(env, mdd_tobj);
3044 GOTO(out_pending, rc);
3047 rc = mdd_la_get(env, mdd_spobj, pattr);
3049 GOTO(out_pending, rc);
3052 rc = mdd_la_get(env, mdd_tobj, tattr);
3054 GOTO(out_pending, rc);
3055 /* search for an existing archive.
3056 * we should check ahead as the object
3057 * can be destroyed in this transaction */
3058 if (mdd_hsm_archive_exists(env, mdd_tobj, ma))
3059 cl_flags |= CLF_RENAME_LAST_EXISTS;
3062 rc = mdd_la_get(env, mdd_tpobj, tpattr);
3064 GOTO(out_pending, rc);
3066 rc = mdd_rename_sanity_check(env, mdd_spobj, pattr, mdd_tpobj, tpattr,
3067 mdd_sobj, cattr, mdd_tobj, tattr);
3069 GOTO(out_pending, rc);
3071 rc = mdd_name_check(mdd, ltname);
3073 GOTO(out_pending, rc);
3075 handle = mdd_trans_create(env, mdd);
3077 GOTO(out_pending, rc = PTR_ERR(handle));
3079 memset(ldata, 0, sizeof(*ldata));
3080 rc = mdd_linkea_prepare(env, mdd_sobj, mdd_object_fid(mdd_spobj),
3081 lsname, mdd_object_fid(mdd_tpobj), ltname,
3086 rc = mdd_declare_rename(env, mdd, mdd_spobj, mdd_tpobj, mdd_sobj,
3087 mdd_tobj, lsname, ltname, ma, ldata, handle);
3091 rc = mdd_trans_start(env, mdd, handle);
3095 is_dir = S_ISDIR(cattr->la_mode);
3097 /* Remove source name from source directory */
3098 rc = __mdd_index_delete(env, mdd_spobj, sname, is_dir, handle);
3102 /* "mv dir1 dir2" needs "dir1/.." link update */
3103 if (is_dir && !lu_fid_eq(spobj_fid, tpobj_fid)) {
3104 rc = __mdd_index_delete_only(env, mdd_sobj, dotdot, handle);
3106 GOTO(fixup_spobj2, rc);
3108 rc = __mdd_index_insert_only(env, mdd_sobj, tpobj_fid, S_IFDIR,
3111 GOTO(fixup_spobj, rc);
3114 if (mdd_tobj != NULL && mdd_object_exists(mdd_tobj)) {
3115 rc = __mdd_index_delete(env, mdd_tpobj, tname, is_dir, handle);
3117 /* tname might been renamed to something else */
3118 GOTO(fixup_spobj, rc);
3121 /* Insert new fid with target name into target dir */
3122 rc = __mdd_index_insert(env, mdd_tpobj, lf, cattr->la_mode,
3125 GOTO(fixup_tpobj, rc);
3127 LASSERT(ma->ma_attr.la_valid & LA_CTIME);
3128 la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime;
3130 /* XXX: mdd_sobj must be local one if it is NOT NULL. */
3131 la->la_valid = LA_CTIME;
3132 rc = mdd_update_time(env, mdd_sobj, cattr, la, handle);
3134 GOTO(fixup_tpobj, rc);
3136 /* Update the linkEA for the source object */
3137 mdd_write_lock(env, mdd_sobj, DT_SRC_CHILD);
3138 rc = mdd_links_rename(env, mdd_sobj, mdd_object_fid(mdd_spobj),
3139 lsname, mdd_object_fid(mdd_tpobj), ltname,
3140 handle, ldata, 0, 0);
3142 /* Old files might not have EA entry */
3143 mdd_links_add(env, mdd_sobj, mdd_object_fid(mdd_spobj),
3144 lsname, handle, NULL, 0);
3145 mdd_write_unlock(env, mdd_sobj);
3146 /* We don't fail the transaction if the link ea can't be
3147 updated -- fid2path will use alternate lookup method. */
3150 /* Remove old target object
3151 * For tobj is remote case cmm layer has processed
3152 * and set tobj to NULL then. So when tobj is NOT NULL,
3153 * it must be local one.
3155 if (tobj && mdd_object_exists(mdd_tobj)) {
3156 mdd_write_lock(env, mdd_tobj, DT_TGT_CHILD);
3158 if (mdd_is_dead_obj(mdd_tobj)) {
3159 /* shld not be dead, something is wrong */
3160 CERROR("tobj is dead, something is wrong\n");
3164 mdo_ref_del(env, mdd_tobj, handle);
3166 /* Remove dot reference. */
3167 if (S_ISDIR(tattr->la_mode))
3168 mdo_ref_del(env, mdd_tobj, handle);
3171 /* fetch updated nlink */
3172 rc = mdd_la_get(env, mdd_tobj, tattr);
3174 CERROR("%s: Failed to get nlink for tobj "
3176 mdd2obd_dev(mdd)->obd_name,
3177 PFID(tpobj_fid), rc);
3178 GOTO(fixup_tpobj, rc);
3181 la->la_valid = LA_CTIME;
3182 rc = mdd_update_time(env, mdd_tobj, tattr, la, handle);
3184 CERROR("%s: Failed to set ctime for tobj "
3186 mdd2obd_dev(mdd)->obd_name,
3187 PFID(tpobj_fid), rc);
3188 GOTO(fixup_tpobj, rc);
3191 /* XXX: this transfer to ma will be removed with LOD/OSP */
3192 ma->ma_attr = *tattr;
3193 ma->ma_valid |= MA_INODE;
3194 rc = mdd_finish_unlink(env, mdd_tobj, ma, mdd_tpobj, ltname,
3197 CERROR("%s: Failed to unlink tobj "
3199 mdd2obd_dev(mdd)->obd_name,
3200 PFID(tpobj_fid), rc);
3201 GOTO(fixup_tpobj, rc);
3204 /* fetch updated nlink */
3205 rc = mdd_la_get(env, mdd_tobj, tattr);
3206 if (rc == -ENOENT) {
3207 /* the object got removed, let's
3208 * return the latest known attributes */
3209 tattr->la_nlink = 0;
3211 } else if (rc != 0) {
3212 CERROR("%s: Failed to get nlink for tobj "
3214 mdd2obd_dev(mdd)->obd_name,
3215 PFID(tpobj_fid), rc);
3216 GOTO(fixup_tpobj, rc);
3218 /* XXX: this transfer to ma will be removed with LOD/OSP */
3219 ma->ma_attr = *tattr;
3220 ma->ma_valid |= MA_INODE;
3222 if (tattr->la_nlink == 0)
3223 cl_flags |= CLF_RENAME_LAST;
3225 cl_flags &= ~CLF_RENAME_LAST_EXISTS;
3228 la->la_valid = LA_CTIME | LA_MTIME;
3229 rc = mdd_update_time(env, mdd_spobj, pattr, la, handle);
3231 GOTO(fixup_tpobj, rc);
3233 if (mdd_spobj != mdd_tpobj) {
3234 la->la_valid = LA_CTIME | LA_MTIME;
3235 rc = mdd_update_time(env, mdd_tpobj, tpattr, la, handle);
3237 GOTO(fixup_tpobj, rc);
3244 rc2 = __mdd_index_delete(env, mdd_tpobj, tname, is_dir, handle);
3246 CWARN("tp obj fix error %d\n",rc2);
3248 if (mdd_tobj && mdd_object_exists(mdd_tobj) &&
3249 !mdd_is_dead_obj(mdd_tobj)) {
3251 mdo_ref_add(env, mdd_tobj, handle);
3253 mdo_ref_add(env, mdd_tobj, handle);
3256 rc2 = __mdd_index_insert(env, mdd_tpobj,
3257 mdd_object_fid(mdd_tobj),
3258 mdd_object_type(mdd_tobj),
3261 CWARN("tp obj fix error: rc = %d\n", rc2);
3266 if (rc && is_dir && mdd_sobj && mdd_spobj != mdd_tpobj) {
3267 rc2 = __mdd_index_delete_only(env, mdd_sobj, dotdot, handle);
3269 CWARN("%s: sp obj dotdot delete error: rc = %d\n",
3270 mdd2obd_dev(mdd)->obd_name, rc2);
3273 rc2 = __mdd_index_insert_only(env, mdd_sobj, spobj_fid, S_IFDIR,
3276 CWARN("%s: sp obj dotdot insert error: rc = %d\n",
3277 mdd2obd_dev(mdd)->obd_name, rc2);
3282 rc2 = __mdd_index_insert(env, mdd_spobj, lf,
3283 mdd_object_type(mdd_sobj), sname,
3286 CWARN("sp obj fix error: rc = %d\n", rc2);
3291 mdd_write_unlock(env, mdd_tobj);
3294 rc = mdd_changelog_ns_store(env, mdd, CL_RENAME, cl_flags,
3295 mdd_tobj, tpobj_fid, lf, spobj_fid,
3296 ltname, lsname, handle);
3299 rc = mdd_trans_stop(env, mdd, rc, handle);
3302 mdd_object_put(env, mdd_sobj);
3307 * Check whether we should migrate the file/dir
3309 * < 0 permission check failed or other error.
3310 * = 0 the file can be migrated.
3312 static int mdd_migrate_sanity_check(const struct lu_env *env,
3313 struct mdd_device *mdd,
3314 struct mdd_object *spobj,
3315 struct mdd_object *tpobj,
3316 struct mdd_object *sobj,
3317 struct mdd_object *tobj,
3318 const struct lu_attr *spattr,
3319 const struct lu_attr *tpattr,
3320 const struct lu_attr *attr)
3326 if (!mdd_object_remote(sobj)) {
3327 mdd_read_lock(env, sobj, DT_SRC_CHILD);
3328 if (sobj->mod_count > 0) {
3329 CDEBUG(D_INFO, "%s: "DFID" is opened, count %d\n",
3330 mdd_obj_dev_name(sobj),
3331 PFID(mdd_object_fid(sobj)),
3333 mdd_read_unlock(env, sobj);
3336 mdd_read_unlock(env, sobj);
3339 if (mdd_object_exists(tobj))
3342 rc = mdd_may_delete(env, spobj, spattr, sobj, attr, NULL, 1, 0);
3346 rc = mdd_may_create(env, tpobj, tpattr, NULL, true);
3351 typedef int (*mdd_xattr_cb)(const struct lu_env *env,
3352 struct mdd_object *obj,
3353 const struct lu_buf *buf,
3355 int fl, struct thandle *handle);
3357 /* iterate xattrs, but ignore LMA, LMV, and LINKEA if 'skip_linkea' is set. */
3358 static int mdd_iterate_xattrs(const struct lu_env *env,
3359 struct mdd_object *sobj,
3360 struct mdd_object *tobj,
3362 struct thandle *handle,
3365 struct mdd_thread_info *info = mdd_env_info(env);
3367 struct lu_buf list_xbuf;
3368 struct lu_buf cbxbuf;
3369 struct lu_buf xbuf = { NULL };
3378 /* retrieve xattr list from the old object */
3379 list_xsize = mdo_xattr_list(env, sobj, &LU_BUF_NULL);
3380 if (list_xsize == -ENODATA)
3386 lu_buf_check_and_alloc(&info->mdi_big_buf, list_xsize);
3387 if (info->mdi_big_buf.lb_buf == NULL)
3390 list_xbuf.lb_buf = info->mdi_big_buf.lb_buf;
3391 list_xbuf.lb_len = list_xsize;
3392 rc = mdo_xattr_list(env, sobj, &list_xbuf);
3398 xname = list_xbuf.lb_buf;
3400 xlen = strnlen(xname, rem - 1) + 1;
3401 if (strcmp(XATTR_NAME_LMA, xname) == 0 ||
3402 strcmp(XATTR_NAME_LMV, xname) == 0)
3406 strcmp(XATTR_NAME_LINK, xname) == 0)
3409 xsize = mdo_xattr_get(env, sobj, &LU_BUF_NULL, xname);
3410 if (xsize == -ENODATA)
3413 GOTO(out, rc = xsize);
3415 lu_buf_check_and_alloc(&xbuf, xsize);
3416 if (xbuf.lb_buf == NULL)
3417 GOTO(out, rc = -ENOMEM);
3419 rc = mdo_xattr_get(env, sobj, &xbuf, xname);
3426 cbxbuf.lb_len = xsize;
3428 rc = cb(env, tobj, &cbxbuf, xname, 0, handle);
3429 if (unlikely(rc == -ENOSPC &&
3430 strcmp(xname, XATTR_NAME_LINK) == 0)) {
3431 rc = linkea_overflow_shrink(
3432 (struct linkea_data *)(cbxbuf.lb_buf));
3433 if (likely(rc > 0)) {
3451 typedef int (*mdd_linkea_cb)(const struct lu_env *env,
3452 struct mdd_object *sobj,
3453 struct mdd_object *tobj,
3454 const struct lu_name *sname,
3455 const struct lu_fid *sfid,
3456 const struct lu_name *lname,
3457 const struct lu_fid *fid,
3459 struct thandle *handle);
3461 static int mdd_declare_update_link(const struct lu_env *env,
3462 struct mdd_object *sobj,
3463 struct mdd_object *tobj,
3464 const struct lu_name *tname,
3465 const struct lu_fid *tpfid,
3466 const struct lu_name *lname,
3467 const struct lu_fid *fid,
3469 struct thandle *handle)
3471 struct mdd_device *mdd = mdo2mdd(&sobj->mod_obj);
3472 struct mdd_object *pobj;
3476 if (lu_fid_eq(tpfid, fid) && tname->ln_namelen == lname->ln_namelen &&
3477 !strcmp(tname->ln_name, lname->ln_name))
3480 pobj = mdd_object_find(env, mdd, fid);
3482 return PTR_ERR(pobj);
3485 rc = mdo_declare_index_delete(env, pobj, lname->ln_name, handle);
3487 rc = mdo_declare_index_insert(env, pobj, mdd_object_fid(tobj),
3488 mdd_object_type(sobj),
3489 lname->ln_name, handle);
3490 mdd_object_put(env, pobj);
3494 rc = mdo_declare_ref_add(env, tobj, handle);
3498 rc = mdo_declare_ref_del(env, sobj, handle);
3502 static int mdd_update_link(const struct lu_env *env,
3503 struct mdd_object *sobj,
3504 struct mdd_object *tobj,
3505 const struct lu_name *tname,
3506 const struct lu_fid *tpfid,
3507 const struct lu_name *lname,
3508 const struct lu_fid *fid,
3510 struct thandle *handle)
3512 struct mdd_device *mdd = mdo2mdd(&sobj->mod_obj);
3513 struct mdd_object *pobj;
3518 LASSERT(lu_name_is_valid(lname));
3521 if (lu_fid_eq(tpfid, fid) && tname->ln_namelen == lname->ln_namelen &&
3522 !strncmp(tname->ln_name, lname->ln_name, lname->ln_namelen))
3525 CDEBUG(D_INFO, "update "DFID"/"DNAME":"DFID"\n",
3526 PFID(fid), PNAME(lname), PFID(mdd_object_fid(tobj)));
3528 pobj = mdd_object_find(env, mdd, fid);
3530 CWARN("%s: cannot find obj "DFID": %ld\n",
3531 mdd2obd_dev(mdd)->obd_name, PFID(fid), PTR_ERR(pobj));
3532 RETURN(PTR_ERR(pobj));
3535 if (!mdd_object_exists(pobj)) {
3536 CDEBUG(D_INFO, DFID" doesn't exist\n", PFID(fid));
3537 mdd_object_put(env, pobj);
3541 mdd_write_lock(env, pobj, DT_TGT_PARENT);
3542 rc = __mdd_index_delete_only(env, pobj, lname->ln_name, handle);
3544 rc = __mdd_index_insert_only(env, pobj, mdd_object_fid(tobj),
3545 mdd_object_type(sobj),
3546 lname->ln_name, handle);
3547 mdd_write_unlock(env, pobj);
3548 mdd_object_put(env, pobj);
3552 mdd_write_lock(env, tobj, DT_TGT_CHILD);
3553 rc = mdo_ref_add(env, tobj, handle);
3554 mdd_write_unlock(env, tobj);
3558 mdd_write_lock(env, sobj, DT_SRC_CHILD);
3559 rc = mdo_ref_del(env, sobj, handle);
3560 mdd_write_unlock(env, sobj);
3565 static inline int mdd_fld_lookup(const struct lu_env *env,
3566 struct mdd_device *mdd,
3567 const struct lu_fid *fid,
3570 struct lu_seq_range *range = &mdd_env_info(env)->mdi_range;
3571 struct seq_server_site *ss;
3574 ss = mdd->mdd_md_dev.md_lu_dev.ld_site->ld_seq_site;
3576 range->lsr_flags = LU_SEQ_RANGE_MDT;
3577 rc = fld_server_lookup(env, ss->ss_server_fld, fid->f_seq, range);
3581 *mdt_index = range->lsr_index;
3586 static int mdd_is_link_on_source_mdt(const struct lu_env *env,
3587 struct mdd_object *sobj,
3588 struct mdd_object *tobj,
3589 const struct lu_name *tname,
3590 const struct lu_fid *tpfid,
3591 const struct lu_name *lname,
3592 const struct lu_fid *fid,
3594 struct thandle *handle)
3596 struct mdd_device *mdd = mdo2mdd(&sobj->mod_obj);
3597 __u32 source_mdt_index = *(__u32 *)opaque;
3598 __u32 link_mdt_index;
3604 if (lu_fid_eq(tpfid, fid) && tname->ln_namelen == lname->ln_namelen &&
3605 !strcmp(tname->ln_name, lname->ln_name))
3608 rc = mdd_fld_lookup(env, mdd, fid, &link_mdt_index);
3612 RETURN(link_mdt_index == source_mdt_index);
3615 static int mdd_iterate_linkea(const struct lu_env *env,
3616 struct mdd_object *sobj,
3617 struct mdd_object *tobj,
3618 const struct lu_name *tname,
3619 const struct lu_fid *tpfid,
3620 struct linkea_data *ldata,
3622 struct thandle *handle,
3625 struct mdd_thread_info *info = mdd_env_info(env);
3626 char *filename = info->mdi_name;
3627 struct lu_name lname;
3634 for (linkea_first_entry(ldata); ldata->ld_lee && !rc;
3635 linkea_next_entry(ldata)) {
3636 linkea_entry_unpack(ldata->ld_lee, &ldata->ld_reclen, &lname,
3639 /* Note: lname might miss \0 at the end */
3640 snprintf(filename, sizeof(info->mdi_name), "%.*s",
3641 lname.ln_namelen, lname.ln_name);
3642 lname.ln_name = filename;
3644 CDEBUG(D_INFO, DFID"/"DNAME"\n", PFID(&fid), PNAME(&lname));
3646 rc = cb(env, sobj, tobj, tname, tpfid, &lname, &fid, opaque,
3654 * Prepare linkea, and check whether file needs migrate: if source still has
3655 * link on source MDT, no need to migrate, just update namespace on source and
3658 * \retval 0 do migrate
3659 * \retval 1 don't migrate
3660 * \retval -errno on failure
3662 static int mdd_migrate_linkea_prepare(const struct lu_env *env,
3663 struct mdd_device *mdd,
3664 struct mdd_object *spobj,
3665 struct mdd_object *tpobj,
3666 struct mdd_object *sobj,
3667 const struct lu_name *sname,
3668 const struct lu_name *tname,
3669 const struct lu_attr *attr,
3670 struct linkea_data *ldata)
3672 __u32 source_mdt_index;
3677 memset(ldata, 0, sizeof(*ldata));
3678 rc = mdd_linkea_prepare(env, sobj, mdd_object_fid(spobj), sname,
3679 mdd_object_fid(tpobj), tname, 1, 0, ldata);
3684 * Then it will check if the file should be migrated. If the file has
3685 * mulitple links, we only need migrate the file if all of its entries
3686 * has been migrated to the remote MDT.
3688 if (S_ISDIR(attr->la_mode) || attr->la_nlink < 2)
3691 /* If there are still links locally, don't migrate this file */
3692 LASSERT(ldata->ld_leh != NULL);
3695 * If linkEA is overflow, it means there are some unknown name entries
3696 * under unknown parents, which will prevent the migration.
3698 if (unlikely(ldata->ld_leh->leh_overflow_time))
3701 rc = mdd_fld_lookup(env, mdd, mdd_object_fid(sobj), &source_mdt_index);
3705 rc = mdd_iterate_linkea(env, sobj, NULL, tname, mdd_object_fid(tpobj),
3706 ldata, &source_mdt_index, NULL,
3707 mdd_is_link_on_source_mdt);
3711 static int mdd_declare_migrate_update(const struct lu_env *env,
3712 struct mdd_object *spobj,
3713 struct mdd_object *tpobj,
3714 struct mdd_object *obj,
3715 const struct lu_name *sname,
3716 const struct lu_name *tname,
3717 struct lu_attr *attr,
3718 struct lu_attr *spattr,
3719 struct lu_attr *tpattr,
3720 struct linkea_data *ldata,
3722 struct thandle *handle)
3724 struct mdd_thread_info *info = mdd_env_info(env);
3725 struct lu_attr *la = &info->mdi_la_for_fix;
3728 rc = mdo_declare_index_delete(env, spobj, sname->ln_name, handle);
3732 if (S_ISDIR(attr->la_mode)) {
3733 rc = mdo_declare_ref_del(env, spobj, handle);
3738 rc = mdo_declare_index_insert(env, tpobj, mdd_object_fid(obj),
3739 attr->la_mode & S_IFMT,
3740 tname->ln_name, handle);
3744 rc = mdd_declare_links_add(env, obj, handle, ldata);
3748 if (S_ISDIR(attr->la_mode)) {
3749 rc = mdo_declare_ref_add(env, tpobj, handle);
3754 la->la_valid = LA_CTIME | LA_MTIME;
3755 rc = mdo_declare_attr_set(env, spobj, la, handle);
3759 if (tpobj != spobj) {
3760 rc = mdo_declare_attr_set(env, tpobj, la, handle);
3768 static int mdd_declare_migrate_create(const struct lu_env *env,
3769 struct mdd_object *spobj,
3770 struct mdd_object *tpobj,
3771 struct mdd_object *sobj,
3772 struct mdd_object *tobj,
3773 const struct lu_name *sname,
3774 const struct lu_name *tname,
3775 struct lu_attr *spattr,
3776 struct lu_attr *tpattr,
3777 struct lu_attr *attr,
3778 struct lu_buf *sbuf,
3779 struct linkea_data *ldata,
3781 struct md_op_spec *spec,
3782 struct dt_allocation_hint *hint,
3783 struct thandle *handle)
3785 struct mdd_thread_info *info = mdd_env_info(env);
3786 struct md_layout_change *mlc = &info->mdi_mlc;
3787 struct lmv_mds_md_v1 *lmv = sbuf->lb_buf;
3792 if (S_ISDIR(attr->la_mode)) {
3793 struct lmv_user_md *lum = spec->u.sp_ea.eadata;
3795 mlc->mlc_opc = MD_LAYOUT_DETACH;
3796 rc = mdo_declare_layout_change(env, sobj, mlc, handle);
3800 lum->lum_hash_type |= cpu_to_le32(LMV_HASH_FLAG_MIGRATION);
3801 } else if (S_ISLNK(attr->la_mode)) {
3802 spec->u.sp_symname = sbuf->lb_buf;
3803 } else if (S_ISREG(attr->la_mode)) {
3804 spec->sp_cr_flags |= MDS_OPEN_DELAY_CREATE;
3805 spec->sp_cr_flags &= ~MDS_OPEN_HAS_EA;
3808 mdd_object_make_hint(env, tpobj, tobj, attr, spec, hint);
3810 rc = mdd_declare_create(env, mdo2mdd(&tpobj->mod_obj), tpobj, tobj,
3811 tname, attr, handle, spec, ldata, NULL, NULL,
3817 * tobj mode will be used in mdo_declare_layout_change(), but it's not
3818 * createb yet, copy from sobj.
3820 tobj->mod_obj.mo_lu.lo_header->loh_attr &= ~S_IFMT;
3821 tobj->mod_obj.mo_lu.lo_header->loh_attr |=
3822 sobj->mod_obj.mo_lu.lo_header->loh_attr & S_IFMT;
3824 if (S_ISDIR(attr->la_mode)) {
3826 /* if sobj is not striped, fake a 1-stripe LMV */
3827 LASSERT(sizeof(info->mdi_key) >
3828 lmv_mds_md_size(1, LMV_MAGIC_V1));
3829 lmv = (typeof(lmv))info->mdi_key;
3830 memset(lmv, 0, sizeof(*lmv));
3831 lmv->lmv_magic = cpu_to_le32(LMV_MAGIC_V1);
3832 lmv->lmv_stripe_count = cpu_to_le32(1);
3833 lmv->lmv_hash_type = cpu_to_le32(LMV_HASH_TYPE_DEFAULT);
3834 fid_le_to_cpu(&lmv->lmv_stripe_fids[0],
3835 mdd_object_fid(sobj));
3836 mlc->mlc_buf.lb_buf = lmv;
3837 mlc->mlc_buf.lb_len = lmv_mds_md_size(1, LMV_MAGIC_V1);
3839 mlc->mlc_buf = *sbuf;
3841 mlc->mlc_opc = MD_LAYOUT_ATTACH;
3842 rc = mdo_declare_layout_change(env, tobj, mlc, handle);
3847 rc = mdd_iterate_xattrs(env, sobj, tobj, true, handle,
3848 mdo_declare_xattr_set);
3852 if (S_ISREG(attr->la_mode)) {
3853 struct lu_buf fid_buf;
3855 handle->th_complex = 1;
3857 /* target may be remote, update PFID via sobj. */
3858 fid_buf.lb_buf = (void *)mdd_object_fid(tobj);
3859 fid_buf.lb_len = sizeof(struct lu_fid);
3860 rc = mdo_declare_xattr_set(env, sobj, &fid_buf, XATTR_NAME_FID,
3865 rc = mdo_declare_xattr_del(env, sobj, XATTR_NAME_LOV, handle);
3870 if (!S_ISDIR(attr->la_mode)) {
3871 rc = mdd_iterate_linkea(env, sobj, tobj, tname,
3872 mdd_object_fid(tpobj), ldata, NULL,
3873 handle, mdd_declare_update_link);
3878 if (!S_ISDIR(attr->la_mode) || lmv) {
3879 rc = mdo_declare_ref_del(env, sobj, handle);
3883 if (S_ISDIR(attr->la_mode)) {
3884 rc = mdo_declare_ref_del(env, sobj, handle);
3889 rc = mdo_declare_destroy(env, sobj, handle);
3894 rc = mdd_declare_migrate_update(env, spobj, tpobj, tobj, sname, tname,
3895 attr, spattr, tpattr, ldata, ma,
3901 * migrate dirent from \a spobj to \a tpobj.
3903 static int mdd_migrate_update(const struct lu_env *env,
3904 struct mdd_object *spobj,
3905 struct mdd_object *tpobj,
3906 struct mdd_object *obj,
3907 const struct lu_name *sname,
3908 const struct lu_name *tname,
3909 struct lu_attr *attr,
3910 struct lu_attr *spattr,
3911 struct lu_attr *tpattr,
3912 struct linkea_data *ldata,
3914 struct thandle *handle)
3916 struct mdd_thread_info *info = mdd_env_info(env);
3917 struct lu_attr *la = &info->mdi_la_for_fix;
3922 CDEBUG(D_INFO, "update "DFID" from "DFID"/%s to "DFID"/%s\n",
3923 PFID(mdd_object_fid(obj)), PFID(mdd_object_fid(spobj)),
3924 sname->ln_name, PFID(mdd_object_fid(tpobj)), tname->ln_name);
3926 rc = __mdd_index_delete(env, spobj, sname->ln_name,
3927 S_ISDIR(attr->la_mode), handle);
3931 rc = __mdd_index_insert(env, tpobj, mdd_object_fid(obj),
3932 attr->la_mode & S_IFMT,
3933 tname->ln_name, handle);
3937 rc = mdd_links_write(env, obj, ldata, handle);
3941 la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime;
3942 la->la_valid = LA_CTIME | LA_MTIME;
3943 mdd_write_lock(env, spobj, DT_SRC_PARENT);
3944 rc = mdd_update_time(env, spobj, spattr, la, handle);
3945 mdd_write_unlock(env, spobj);
3949 if (tpobj != spobj) {
3950 la->la_valid = LA_CTIME | LA_MTIME;
3951 mdd_write_lock(env, tpobj, DT_TGT_PARENT);
3952 rc = mdd_update_time(env, tpobj, tpattr, la, handle);
3953 mdd_write_unlock(env, tpobj);
3962 * Migrate file/dir to target MDT.
3964 * Create target according to \a spec, and then migrate xattrs, if it's
3965 * directory, migrate source stripes to target.
3967 * \param[in] env execution environment
3968 * \param[in] spobj source parent object
3969 * \param[in] tpobj target parent object
3970 * \param[in] sobj source object
3971 * \param[in] tobj target object
3972 * \param[in] lname file name
3973 * \param[in] spattr source parent attributes
3974 * \param[in] tpattr target parent attributes
3975 * \param[in] attr source attributes
3976 * \param[in] sbuf source LMV buf
3977 * \param[in] spec migrate create spec
3978 * \param[in] hint target creation hint
3979 * \param[in] handle tranasction handle
3981 * \retval 0 on success
3982 * \retval -errno on failure
3984 static int mdd_migrate_create(const struct lu_env *env,
3985 struct mdd_object *spobj,
3986 struct mdd_object *tpobj,
3987 struct mdd_object *sobj,
3988 struct mdd_object *tobj,
3989 const struct lu_name *sname,
3990 const struct lu_name *tname,
3991 struct lu_attr *spattr,
3992 struct lu_attr *tpattr,
3993 struct lu_attr *attr,
3994 const struct lu_buf *sbuf,
3995 struct linkea_data *ldata,
3997 struct md_op_spec *spec,
3998 struct dt_allocation_hint *hint,
3999 struct thandle *handle)
4006 * migrate sobj stripes to tobj if it's directory:
4007 * 1. detach stripes from sobj.
4008 * 2. attach stripes to tobj, see mdd_declare_migrate_mdt().
4009 * 3. create stripes for tobj, see lod_xattr_set_lmv().
4011 if (S_ISDIR(attr->la_mode)) {
4012 struct mdd_thread_info *info = mdd_env_info(env);
4013 struct md_layout_change *mlc = &info->mdi_mlc;
4015 mlc->mlc_opc = MD_LAYOUT_DETACH;
4017 mdd_write_lock(env, sobj, DT_SRC_PARENT);
4018 rc = mdo_layout_change(env, sobj, mlc, handle);
4019 mdd_write_unlock(env, sobj);
4024 /* don't set nlink from sobj */
4025 attr->la_valid &= ~LA_NLINK;
4027 rc = mdd_create_object(env, tpobj, tobj, attr, spec, NULL, NULL, NULL,
4028 hint, handle, false);
4032 mdd_write_lock(env, tobj, DT_TGT_CHILD);
4033 rc = mdd_iterate_xattrs(env, sobj, tobj, true, handle, mdo_xattr_set);
4034 mdd_write_unlock(env, tobj);
4038 /* for regular file, update OST objects XATTR_NAME_FID */
4039 if (S_ISREG(attr->la_mode)) {
4040 struct lu_buf fid_buf;
4042 /* target may be remote, update PFID via sobj. */
4043 fid_buf.lb_buf = (void *)mdd_object_fid(tobj);
4044 fid_buf.lb_len = sizeof(struct lu_fid);
4045 rc = mdo_xattr_set(env, sobj, &fid_buf, XATTR_NAME_FID, 0,
4050 /* delete LOV to avoid deleting OST objs when destroying sobj */
4051 mdd_write_lock(env, sobj, DT_SRC_CHILD);
4052 rc = mdo_xattr_del(env, sobj, XATTR_NAME_LOV, handle);
4053 mdd_write_unlock(env, sobj);
4054 /* O_DELAY_CREATE file may not have LOV, ignore -ENODATA */
4055 if (rc && rc != -ENODATA)
4060 /* update links FID */
4061 if (!S_ISDIR(attr->la_mode)) {
4062 rc = mdd_iterate_linkea(env, sobj, tobj, tname,
4063 mdd_object_fid(tpobj), ldata,
4064 NULL, handle, mdd_update_link);
4069 /* don't destroy sobj if it's plain directory */
4070 if (!S_ISDIR(attr->la_mode) || sbuf->lb_buf) {
4071 mdd_write_lock(env, sobj, DT_SRC_CHILD);
4072 rc = mdo_ref_del(env, sobj, handle);
4074 if (S_ISDIR(attr->la_mode))
4075 rc = mdo_ref_del(env, sobj, handle);
4077 rc = mdo_destroy(env, sobj, handle);
4079 mdd_write_unlock(env, sobj);
4084 rc = mdd_migrate_update(env, spobj, tpobj, tobj, sname, tname, attr,
4085 spattr, tpattr, ldata, ma, handle);
4090 /* NB: if user issued different migrate command, we can't adjust it silently
4091 * here, because this command will decide target MDT in subdir migration in
4094 static int mdd_migrate_cmd_check(struct mdd_device *mdd,
4095 const struct lmv_mds_md_v1 *lmv,
4096 const struct lmv_user_md_v1 *lum,
4097 const struct lu_name *lname)
4099 __u32 lum_stripe_count = lum->lum_stripe_count;
4100 __u32 lum_hash_type = lum->lum_hash_type &
4101 cpu_to_le32(LMV_HASH_TYPE_MASK);
4102 __u32 lmv_hash_type = lmv->lmv_hash_type &
4103 cpu_to_le32(LMV_HASH_TYPE_MASK);
4105 if (!lmv_is_sane(lmv))
4108 /* if stripe_count unspecified, set to 1 */
4109 if (!lum_stripe_count)
4110 lum_stripe_count = cpu_to_le32(1);
4112 /* TODO: check specific MDTs */
4113 if (lum_stripe_count != lmv->lmv_migrate_offset ||
4114 lum->lum_stripe_offset != lmv->lmv_master_mdt_index ||
4115 (lum_hash_type && lum_hash_type != lmv_hash_type)) {
4116 CERROR("%s: '"DNAME"' migration was interrupted, run 'lfs migrate -m %d -c %d -H %s "DNAME"' to finish migration.\n",
4117 mdd2obd_dev(mdd)->obd_name, PNAME(lname),
4118 le32_to_cpu(lmv->lmv_master_mdt_index),
4119 le32_to_cpu(lmv->lmv_migrate_offset),
4120 mdt_hash_name[le32_to_cpu(lmv_hash_type)],
4129 * Internal function to migrate directory or file between MDTs.
4131 * migrate source to target in following steps:
4132 * 1. create target, append source stripes after target's if it's directory,
4133 * migrate xattrs and update fid of source links.
4134 * 2. update namespace: migrate dirent from source parent to target parent,
4135 * update file linkea, and destroy source if it's not needed any more.
4137 * \param[in] env execution environment
4138 * \param[in] spobj source parent object
4139 * \param[in] tpobj target parent object
4140 * \param[in] sobj source object
4141 * \param[in] tobj target object
4142 * \param[in] sname source file name
4143 * \param[in] tname target file name
4144 * \param[in] spec target creation spec
4145 * \param[in] ma used to update \a pobj mtime and ctime
4147 * \retval 0 on success
4148 * \retval -errno on failure
4150 static int mdd_migrate_object(const struct lu_env *env,
4151 struct mdd_object *spobj,
4152 struct mdd_object *tpobj,
4153 struct mdd_object *sobj,
4154 struct mdd_object *tobj,
4155 const struct lu_name *sname,
4156 const struct lu_name *tname,
4157 struct md_op_spec *spec,
4160 struct mdd_thread_info *info = mdd_env_info(env);
4161 struct mdd_device *mdd = mdo2mdd(&spobj->mod_obj);
4162 struct lu_attr *spattr = &info->mdi_pattr;
4163 struct lu_attr *tpattr = &info->mdi_tpattr;
4164 struct lu_attr *attr = &info->mdi_cattr;
4165 struct linkea_data *ldata = &info->mdi_link_data;
4166 struct dt_allocation_hint *hint = &info->mdi_hint;
4167 struct lu_buf sbuf = { NULL };
4168 struct lmv_mds_md_v1 *lmv;
4169 struct thandle *handle;
4174 rc = mdd_la_get(env, sobj, attr);
4178 rc = mdd_la_get(env, spobj, spattr);
4182 rc = mdd_la_get(env, tpobj, tpattr);
4186 if (S_ISDIR(attr->la_mode) && !spec->sp_migrate_nsonly) {
4187 struct lmv_user_md_v1 *lum = spec->u.sp_ea.eadata;
4191 /* if user use default value '0' for stripe_count, we need to
4192 * adjust it to '1' to create a 1-stripe directory.
4194 if (lum->lum_stripe_count == 0)
4195 lum->lum_stripe_count = cpu_to_le32(1);
4197 rc = mdd_stripe_get(env, sobj, &sbuf, XATTR_NAME_LMV);
4198 if (rc && rc != -ENODATA)
4203 if (!lmv_is_sane(lmv))
4204 GOTO(out, rc = -EBADF);
4205 if (lmv_is_migrating(lmv)) {
4206 rc = mdd_migrate_cmd_check(mdd, lmv, lum,
4211 } else if (!S_ISDIR(attr->la_mode)) {
4213 GOTO(out, rc = -EALREADY);
4215 /* update namespace only if @sobj is on MDT where @tpobj is. */
4216 if (!mdd_object_remote(tpobj) && !mdd_object_remote(sobj))
4217 spec->sp_migrate_nsonly = true;
4219 if (S_ISLNK(attr->la_mode)) {
4220 lu_buf_check_and_alloc(&sbuf, attr->la_size + 1);
4222 GOTO(out, rc = -ENOMEM);
4224 rc = mdd_readlink(env, &sobj->mod_obj, &sbuf);
4227 CERROR("%s: "DFID" readlink failed: rc = %d\n",
4228 mdd2obd_dev(mdd)->obd_name,
4229 PFID(mdd_object_fid(sobj)), rc);
4235 /* linkea needs update upon FID or parent stripe change */
4236 rc = mdd_migrate_linkea_prepare(env, mdd, spobj, tpobj, sobj, sname,
4237 tname, attr, ldata);
4239 /* update namespace only if @sobj has link on its MDT. */
4240 spec->sp_migrate_nsonly = true;
4244 rc = mdd_migrate_sanity_check(env, mdd, spobj, tpobj, sobj, tobj,
4245 spattr, tpattr, attr);
4249 handle = mdd_trans_create(env, mdd);
4251 GOTO(out, rc = PTR_ERR(handle));
4253 if (spec->sp_migrate_nsonly)
4254 rc = mdd_declare_migrate_update(env, spobj, tpobj, sobj, sname,
4255 tname, attr, spattr, tpattr,
4258 rc = mdd_declare_migrate_create(env, spobj, tpobj, sobj, tobj,
4259 sname, tname, spattr, tpattr,
4260 attr, &sbuf, ldata, ma, spec,
4265 rc = mdd_declare_changelog_store(env, mdd, CL_MIGRATE, tname, sname,
4270 rc = mdd_trans_start(env, mdd, handle);
4274 if (spec->sp_migrate_nsonly)
4275 rc = mdd_migrate_update(env, spobj, tpobj, sobj, sname, tname,
4276 attr, spattr, tpattr, ldata, ma,
4279 rc = mdd_migrate_create(env, spobj, tpobj, sobj, tobj, sname,
4280 tname, spattr, tpattr, attr, &sbuf,
4281 ldata, ma, spec, hint, handle);
4285 rc = mdd_changelog_ns_store(env, mdd, CL_MIGRATE, 0,
4286 spec->sp_migrate_nsonly ? sobj : tobj,
4287 mdd_object_fid(spobj), mdd_object_fid(sobj),
4288 mdd_object_fid(tpobj), tname, sname,
4295 rc = mdd_trans_stop(env, mdd, rc, handle);
4303 * Migrate directory or file between MDTs.
4305 * \param[in] env execution environment
4306 * \param[in] md_pobj parent master object
4307 * \param[in] md_sobj source object
4308 * \param[in] lname file name
4309 * \param[in] md_tobj target object
4310 * \param[in] spec target creation spec
4311 * \param[in] ma used to update \a pobj mtime and ctime
4313 * \retval 0 on success
4314 * \retval -errno on failure
4316 static int mdd_migrate(const struct lu_env *env, struct md_object *md_pobj,
4317 struct md_object *md_sobj, const struct lu_name *lname,
4318 struct md_object *md_tobj, struct md_op_spec *spec,
4321 struct mdd_thread_info *info = mdd_env_info(env);
4322 struct mdd_device *mdd = mdo2mdd(md_pobj);
4323 struct mdd_object *pobj = md2mdd_obj(md_pobj);
4324 struct mdd_object *sobj = md2mdd_obj(md_sobj);
4325 struct mdd_object *tobj = md2mdd_obj(md_tobj);
4326 struct mdd_object *spobj = NULL;
4327 struct mdd_object *tpobj = NULL;
4328 struct lu_buf pbuf = { NULL };
4329 struct lu_fid *fid = &info->mdi_fid2;
4330 struct lmv_mds_md_v1 *lmv;
4335 /* locate source and target stripe on pobj, which are the real parent */
4336 rc = mdd_stripe_get(env, pobj, &pbuf, XATTR_NAME_LMV);
4337 if (rc < 0 && rc != -ENODATA)
4344 if (!lmv_is_sane(lmv))
4345 GOTO(out, rc = -EBADF);
4347 /* locate target parent stripe */
4348 /* fail check here to make sure top dir migration succeed. */
4349 if (lmv_is_migrating(lmv) &&
4350 OBD_FAIL_CHECK_RESET(OBD_FAIL_MIGRATE_ENTRIES, 0))
4351 GOTO(out, rc = -EIO);
4353 index = lmv_name_to_stripe_index(lmv, lname->ln_name,
4356 GOTO(out, rc = index);
4358 fid_le_to_cpu(fid, &lmv->lmv_stripe_fids[index]);
4359 tpobj = mdd_object_find(env, mdd, fid);
4361 GOTO(out, rc = PTR_ERR(tpobj));
4363 /* locate source parent stripe */
4364 if (lmv_is_layout_changing(lmv)) {
4365 index = lmv_name_to_stripe_index_old(lmv,
4369 GOTO(out, rc = index);
4371 fid_le_to_cpu(fid, &lmv->lmv_stripe_fids[index]);
4372 spobj = mdd_object_find(env, mdd, fid);
4374 GOTO(out, rc = PTR_ERR(spobj));
4376 /* parent stripe unchanged */
4377 if (spobj == tpobj) {
4378 if (!lmv_is_restriping(lmv))
4379 GOTO(out, rc = -EINVAL);
4380 GOTO(out, rc = -EALREADY);
4384 mdd_object_get(spobj);
4389 mdd_object_get(tpobj);
4390 mdd_object_get(spobj);
4393 rc = mdd_migrate_object(env, spobj, tpobj, sobj, tobj, lname, lname,
4398 if (!IS_ERR_OR_NULL(spobj))
4399 mdd_object_put(env, spobj);
4400 if (!IS_ERR_OR_NULL(tpobj))
4401 mdd_object_put(env, tpobj);
4407 static int mdd_declare_1sd_collapse(const struct lu_env *env,
4408 struct mdd_object *pobj,
4409 struct mdd_object *obj,
4410 struct mdd_object *stripe,
4411 struct lu_attr *attr,
4412 struct md_layout_change *mlc,
4413 struct lu_name *lname,
4414 struct thandle *handle)
4418 mlc->mlc_opc = MD_LAYOUT_DETACH;
4419 rc = mdo_declare_layout_change(env, obj, mlc, handle);
4423 rc = mdo_declare_index_insert(env, stripe, mdd_object_fid(pobj),
4424 S_IFDIR, dotdot, handle);
4428 rc = mdd_iterate_xattrs(env, obj, stripe, false, handle,
4429 mdo_declare_xattr_set);
4433 rc = mdo_declare_xattr_del(env, stripe, XATTR_NAME_LMV, handle);
4437 rc = mdo_declare_attr_set(env, stripe, attr, handle);
4441 rc = mdo_declare_index_delete(env, pobj, lname->ln_name, handle);
4445 rc = mdo_declare_index_insert(env, pobj, mdd_object_fid(stripe),
4446 attr->la_mode, lname->ln_name, handle);
4450 rc = mdo_declare_ref_del(env, obj, handle);
4454 rc = mdo_declare_ref_del(env, obj, handle);
4458 rc = mdo_declare_destroy(env, obj, handle);
4465 /* transform one-stripe directory to a plain directory */
4466 static int mdd_1sd_collapse(const struct lu_env *env,
4467 struct mdd_object *pobj,
4468 struct mdd_object *obj,
4469 struct mdd_object *stripe,
4470 struct lu_attr *attr,
4471 struct md_layout_change *mlc,
4472 struct lu_name *lname,
4473 struct thandle *handle)
4479 /* replace 1-stripe directory with its stripe */
4480 mlc->mlc_opc = MD_LAYOUT_DETACH;
4482 mdd_write_lock(env, obj, DT_SRC_PARENT);
4483 rc = mdo_layout_change(env, obj, mlc, handle);
4484 mdd_write_unlock(env, obj);
4488 mdd_write_lock(env, pobj, DT_SRC_PARENT);
4489 mdd_write_lock(env, obj, DT_SRC_CHILD);
4491 /* insert dotdot to stripe which points to parent */
4492 rc = __mdd_index_insert_only(env, stripe, mdd_object_fid(pobj),
4493 S_IFDIR, dotdot, handle);
4497 /* copy xattrs including linkea */
4498 rc = mdd_iterate_xattrs(env, obj, stripe, false, handle, mdo_xattr_set);
4503 rc = mdo_xattr_del(env, stripe, XATTR_NAME_LMV, handle);
4507 /* don't set nlink from parent */
4508 attr->la_valid &= ~LA_NLINK;
4510 rc = mdo_attr_set(env, stripe, attr, handle);
4514 /* delete dir name from parent */
4515 rc = __mdd_index_delete_only(env, pobj, lname->ln_name, handle);
4519 /* insert stripe to parent with dir name */
4520 rc = __mdd_index_insert_only(env, pobj, mdd_object_fid(stripe),
4521 attr->la_mode, lname->ln_name, handle);
4525 /* destroy dir obj */
4526 rc = mdo_ref_del(env, obj, handle);
4530 rc = mdo_ref_del(env, obj, handle);
4534 rc = mdo_destroy(env, obj, handle);
4540 mdd_write_unlock(env, obj);
4541 mdd_write_unlock(env, pobj);
4547 * shrink directory stripes after migration/merge
4549 int mdd_dir_layout_shrink(const struct lu_env *env,
4550 struct md_object *md_obj,
4551 struct md_layout_change *mlc)
4553 struct mdd_device *mdd = mdo2mdd(md_obj);
4554 struct mdd_thread_info *info = mdd_env_info(env);
4555 struct mdd_object *obj = md2mdd_obj(md_obj);
4556 struct mdd_object *pobj = NULL;
4557 struct mdd_object *stripe = NULL;
4558 struct lu_attr *attr = &info->mdi_pattr;
4559 struct lu_fid *fid = &info->mdi_fid2;
4560 struct lu_name lname = { NULL };
4561 struct lu_buf lmv_buf = { NULL };
4562 struct lmv_mds_md_v1 *lmv;
4563 struct lmv_user_md *lmu;
4564 struct thandle *handle;
4569 rc = mdd_la_get(env, obj, attr);
4573 if (!S_ISDIR(attr->la_mode))
4576 rc = mdd_stripe_get(env, obj, &lmv_buf, XATTR_NAME_LMV);
4580 lmv = lmv_buf.lb_buf;
4581 if (!lmv_is_sane(lmv))
4584 lmu = mlc->mlc_buf.lb_buf;
4586 /* adjust the default value '0' to '1' */
4587 if (lmu->lum_stripe_count == 0)
4588 lmu->lum_stripe_count = cpu_to_le32(1);
4590 /* these were checked in MDT */
4591 LASSERT(le32_to_cpu(lmu->lum_stripe_count) <
4592 le32_to_cpu(lmv->lmv_stripe_count));
4593 LASSERT(!lmv_is_splitting(lmv));
4594 LASSERT(lmv_is_migrating(lmv) || lmv_is_merging(lmv));
4596 /* if dir stripe count will be shrunk to 1, it needs to be transformed
4597 * to a plain dir, which will cause FID change and namespace update.
4599 if (le32_to_cpu(lmu->lum_stripe_count) == 1) {
4600 struct linkea_data *ldata = &info->mdi_link_data;
4601 char *filename = info->mdi_name;
4603 rc = mdd_links_read(env, obj, ldata);
4607 if (ldata->ld_leh->leh_reccount > 1)
4608 GOTO(out, rc = -EINVAL);
4610 linkea_first_entry(ldata);
4612 GOTO(out, rc = -ENODATA);
4614 linkea_entry_unpack(ldata->ld_lee, &ldata->ld_reclen, &lname,
4617 /* Note: lname might miss \0 at the end */
4618 snprintf(filename, sizeof(info->mdi_name), "%.*s",
4619 lname.ln_namelen, lname.ln_name);
4620 lname.ln_name = filename;
4622 pobj = mdd_object_find(env, mdd, fid);
4629 fid_le_to_cpu(fid, &lmv->lmv_stripe_fids[0]);
4631 stripe = mdd_object_find(env, mdd, fid);
4632 if (IS_ERR(stripe)) {
4633 mdd_object_put(env, pobj);
4635 GOTO(out, rc = PTR_ERR(stripe));
4639 handle = mdd_trans_create(env, mdd);
4641 GOTO(out, rc = PTR_ERR(handle));
4643 mlc->mlc_opc = MD_LAYOUT_SHRINK;
4644 rc = mdo_declare_layout_change(env, obj, mlc, handle);
4646 GOTO(stop_trans, rc);
4648 if (le32_to_cpu(lmu->lum_stripe_count) == 1 && !lmv_is_fixed(lmv)) {
4649 rc = mdd_declare_1sd_collapse(env, pobj, obj, stripe, attr, mlc,
4652 GOTO(stop_trans, rc);
4655 rc = mdd_declare_changelog_store(env, mdd, CL_LAYOUT, NULL, NULL,
4658 GOTO(stop_trans, rc);
4660 rc = mdd_trans_start(env, mdd, handle);
4662 GOTO(stop_trans, rc);
4664 mdd_write_lock(env, obj, DT_SRC_PARENT);
4665 mlc->mlc_opc = MD_LAYOUT_SHRINK;
4666 rc = mdo_layout_change(env, obj, mlc, handle);
4667 mdd_write_unlock(env, obj);
4669 GOTO(stop_trans, rc);
4671 if (le32_to_cpu(lmu->lum_stripe_count) == 1 && !lmv_is_fixed(lmv)) {
4672 rc = mdd_1sd_collapse(env, pobj, obj, stripe, attr, mlc, &lname,
4675 GOTO(stop_trans, rc);
4678 rc = mdd_changelog_data_store_xattr(env, mdd, CL_LAYOUT, 0, obj,
4679 XATTR_NAME_LMV, handle);
4680 GOTO(stop_trans, rc);
4683 rc = mdd_trans_stop(env, mdd, rc, handle);
4686 mdd_object_put(env, stripe);
4687 mdd_object_put(env, pobj);
4689 lu_buf_free(&lmv_buf);
4693 static int mdd_dir_declare_split_plain(const struct lu_env *env,
4694 struct mdd_device *mdd,
4695 struct mdd_object *pobj,
4696 struct mdd_object *obj,
4697 struct mdd_object *tobj,
4698 struct md_layout_change *mlc,
4699 struct dt_allocation_hint *hint,
4700 struct thandle *handle)
4702 struct mdd_thread_info *info = mdd_env_info(env);
4703 const struct lu_name *lname = mlc->mlc_name;
4704 struct lu_attr *la = &info->mdi_la_for_fix;
4705 struct lmv_user_md_v1 *lum = mlc->mlc_spec->u.sp_ea.eadata;
4706 struct linkea_data *ldata = &info->mdi_link_data;
4707 struct lmv_mds_md_v1 *lmv;
4711 mlc->mlc_opc = MD_LAYOUT_DETACH;
4712 rc = mdo_declare_layout_change(env, obj, mlc, handle);
4716 memset(ldata, 0, sizeof(*ldata));
4717 rc = mdd_linkea_prepare(env, obj, NULL, NULL, mdd_object_fid(pobj),
4718 lname, 1, 0, ldata);
4722 count = lum->lum_stripe_count;
4723 lum->lum_stripe_count = 0;
4724 mdd_object_make_hint(env, pobj, tobj, mlc->mlc_attr, mlc->mlc_spec,
4726 rc = mdd_declare_create(env, mdo2mdd(&pobj->mod_obj), pobj, tobj,
4727 lname, mlc->mlc_attr, handle, mlc->mlc_spec,
4728 ldata, NULL, NULL, NULL, hint);
4732 /* tobj mode will be used in lod_declare_xattr_set(), but it's not
4735 tobj->mod_obj.mo_lu.lo_header->loh_attr |= S_IFDIR;
4737 lmv = (typeof(lmv))info->mdi_key;
4738 memset(lmv, 0, sizeof(*lmv));
4739 lmv->lmv_magic = cpu_to_le32(LMV_MAGIC_V1);
4740 lmv->lmv_stripe_count = cpu_to_le32(1);
4741 lmv->lmv_hash_type = cpu_to_le32(LMV_HASH_TYPE_DEFAULT);
4742 fid_le_to_cpu(&lmv->lmv_stripe_fids[0], mdd_object_fid(obj));
4744 mlc->mlc_opc = MD_LAYOUT_ATTACH;
4745 mlc->mlc_buf.lb_buf = lmv;
4746 mlc->mlc_buf.lb_len = lmv_mds_md_size(1, LMV_MAGIC_V1);
4747 rc = mdo_declare_layout_change(env, tobj, mlc, handle);
4751 rc = mdd_iterate_xattrs(env, obj, tobj, true, handle,
4752 mdo_declare_xattr_set);
4756 lum->lum_stripe_count = count;
4757 mlc->mlc_opc = MD_LAYOUT_SPLIT;
4758 rc = mdo_declare_layout_change(env, tobj, mlc, handle);
4762 rc = mdo_declare_index_delete(env, pobj, lname->ln_name, handle);
4766 rc = mdo_declare_index_insert(env, pobj, mdd_object_fid(tobj),
4767 S_IFDIR, lname->ln_name, handle);
4771 la->la_valid = LA_CTIME | LA_MTIME;
4772 rc = mdo_declare_attr_set(env, obj, la, handle);
4776 rc = mdo_declare_attr_set(env, pobj, la, handle);
4780 rc = mdd_declare_changelog_store(env, mdd, CL_MIGRATE, lname, NULL,
4786 * plain directory split:
4787 * 1. create \a tobj as plain directory.
4788 * 2. append \a obj as first stripe of \a tobj.
4789 * 3. migrate xattrs from \a obj to \a tobj.
4790 * 4. split \a tobj to specific stripe count.
4792 static int mdd_dir_split_plain(const struct lu_env *env,
4793 struct mdd_device *mdd,
4794 struct mdd_object *pobj,
4795 struct mdd_object *obj,
4796 struct mdd_object *tobj,
4797 struct md_layout_change *mlc,
4798 struct dt_allocation_hint *hint,
4799 struct thandle *handle)
4801 struct mdd_thread_info *info = mdd_env_info(env);
4802 struct lu_attr *pattr = &info->mdi_pattr;
4803 struct lu_attr *la = &info->mdi_la_for_fix;
4804 const struct lu_name *lname = mlc->mlc_name;
4805 struct linkea_data *ldata = &info->mdi_link_data;
4810 /* copy linkea out and set on target later */
4811 rc = mdd_links_read(env, obj, ldata);
4815 mlc->mlc_opc = MD_LAYOUT_DETACH;
4816 rc = mdo_layout_change(env, obj, mlc, handle);
4820 /* don't set nlink from obj */
4821 mlc->mlc_attr->la_valid &= ~LA_NLINK;
4823 rc = mdd_create_object(env, pobj, tobj, mlc->mlc_attr, mlc->mlc_spec,
4824 NULL, NULL, NULL, hint, handle, false);
4828 rc = mdd_iterate_xattrs(env, obj, tobj, true, handle, mdo_xattr_set);
4832 rc = mdd_links_write(env, tobj, ldata, handle);
4836 rc = __mdd_index_delete(env, pobj, lname->ln_name, true, handle);
4840 rc = __mdd_index_insert(env, pobj, mdd_object_fid(tobj), S_IFDIR,
4841 lname->ln_name, handle);
4845 la->la_ctime = la->la_mtime = mlc->mlc_attr->la_mtime;
4846 la->la_valid = LA_CTIME | LA_MTIME;
4848 mdd_write_lock(env, obj, DT_SRC_CHILD);
4849 rc = mdd_update_time(env, tobj, mlc->mlc_attr, la, handle);
4850 mdd_write_unlock(env, obj);
4854 rc = mdd_la_get(env, pobj, pattr);
4858 la->la_valid = LA_CTIME | LA_MTIME;
4860 mdd_write_lock(env, pobj, DT_SRC_PARENT);
4861 rc = mdd_update_time(env, pobj, pattr, la, handle);
4862 mdd_write_unlock(env, pobj);
4866 /* FID changes, record it as CL_MIGRATE */
4867 rc = mdd_changelog_ns_store(env, mdd, CL_MIGRATE, 0, tobj,
4868 mdd_object_fid(pobj), mdd_object_fid(obj),
4869 mdd_object_fid(pobj), lname, lname, handle);
4873 int mdd_dir_layout_split(const struct lu_env *env, struct md_object *o,
4874 struct md_layout_change *mlc)
4876 struct mdd_thread_info *info = mdd_env_info(env);
4877 struct mdd_device *mdd = mdo2mdd(o);
4878 struct mdd_object *obj = md2mdd_obj(o);
4879 struct mdd_object *pobj = md2mdd_obj(mlc->mlc_parent);
4880 struct mdd_object *tobj = md2mdd_obj(mlc->mlc_target);
4881 struct dt_allocation_hint *hint = &info->mdi_hint;
4882 bool is_plain = false;
4883 struct thandle *handle;
4888 LASSERT(S_ISDIR(mdd_object_type(obj)));
4890 rc = mdo_xattr_get(env, obj, &LU_BUF_NULL, XATTR_NAME_LMV);
4896 handle = mdd_trans_create(env, mdd);
4898 RETURN(PTR_ERR(handle));
4901 rc = mdd_dir_declare_split_plain(env, mdd, pobj, obj, tobj, mlc,
4904 mlc->mlc_opc = MD_LAYOUT_SPLIT;
4905 rc = mdo_declare_layout_change(env, obj, mlc, handle);
4907 GOTO(stop_trans, rc);
4909 rc = mdd_declare_changelog_store(env, mdd, CL_LAYOUT, NULL,
4913 GOTO(stop_trans, rc);
4915 rc = mdd_trans_start(env, mdd, handle);
4917 GOTO(stop_trans, rc);
4920 rc = mdd_dir_split_plain(env, mdd, pobj, obj, tobj, mlc, hint,
4923 mdd_write_lock(env, obj, DT_TGT_CHILD);
4924 rc = mdo_xattr_set(env, obj, NULL, XATTR_NAME_LMV,
4925 LU_XATTR_CREATE, handle);
4926 mdd_write_unlock(env, obj);
4928 GOTO(stop_trans, rc);
4930 rc = mdd_changelog_data_store_xattr(env, mdd, CL_LAYOUT, 0, obj,
4931 XATTR_NAME_LMV, handle);
4934 GOTO(stop_trans, rc);
4939 rc = mdd_trans_stop(env, mdd, rc, handle);
4944 const struct md_dir_operations mdd_dir_ops = {
4945 .mdo_is_subdir = mdd_is_subdir,
4946 .mdo_lookup = mdd_lookup,
4947 .mdo_create = mdd_create,
4948 .mdo_rename = mdd_rename,
4949 .mdo_link = mdd_link,
4950 .mdo_unlink = mdd_unlink,
4951 .mdo_create_data = mdd_create_data,
4952 .mdo_migrate = mdd_migrate,