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_object.c
33 * Lustre Metadata Server (mdd) routines
35 * Author: Wang Di <wangdi@clusterfs.com>
38 #define DEBUG_SUBSYSTEM S_MDS
40 #include <linux/module.h>
42 #include <obd_class.h>
43 #include <obd_support.h>
44 #include <lprocfs_status.h>
45 /* fid_be_cpu(), fid_cpu_to_be(). */
46 #include <lustre_fid.h>
47 #include <lustre_idmap.h>
48 #include <uapi/linux/lustre/lustre_param.h>
49 #include <lustre_mds.h>
51 #include "mdd_internal.h"
53 static const struct lu_object_operations mdd_lu_obj_ops;
55 struct mdd_object_user {
56 struct list_head mou_list; /**< linked off mod_users */
57 u64 mou_open_flags; /**< open mode by client */
58 __u64 mou_uidgid; /**< uid_gid on client */
59 int mou_opencount; /**< # opened */
60 ktime_t mou_deniednext; /**< time of next access denied
65 static int mdd_xattr_get(const struct lu_env *env,
66 struct md_object *obj, struct lu_buf *buf,
69 static int mdd_changelog_data_store_by_fid(const struct lu_env *env,
70 struct mdd_device *mdd,
71 enum changelog_rec_type type,
72 enum changelog_rec_flags clf_flags,
73 const struct lu_fid *fid,
74 const struct lu_fid *pfid,
75 const char *xattr_name,
76 struct thandle *handle);
78 static inline bool has_prefix(const char *str, const char *prefix);
81 static u32 flags_helper(u64 open_flags)
85 if (open_flags & MDS_FMODE_EXEC) {
86 open_mode = MDS_FMODE_EXEC;
88 if (open_flags & MDS_FMODE_READ)
89 open_mode = MDS_FMODE_READ;
91 (MDS_FMODE_WRITE | MDS_OPEN_TRUNC | MDS_OPEN_APPEND))
92 open_mode |= MDS_FMODE_WRITE;
98 /** Allocate/init a user and its sub-structures.
103 * \retval mou [OUT] success valid structure
106 static struct mdd_object_user *mdd_obj_user_alloc(u64 open_flags,
107 uid_t uid, gid_t gid)
109 struct mdd_object_user *mou;
113 OBD_SLAB_ALLOC_PTR(mou, mdd_object_kmem);
115 RETURN(ERR_PTR(-ENOMEM));
117 mou->mou_open_flags = open_flags;
118 mou->mou_uidgid = ((__u64)uid << 32) | gid;
119 mou->mou_opencount = 0;
120 mou->mou_deniednext = ktime_set(0, 0);
126 * Free a user and its sub-structures.
128 * \param mou [IN] user to be freed.
130 static void mdd_obj_user_free(struct mdd_object_user *mou)
132 OBD_SLAB_FREE_PTR(mou, mdd_object_kmem);
136 * Find if UID/GID already has this file open
138 * Caller should have write-locked \param mdd_obj.
139 * \param mdd_obj [IN] mdd_obj
140 * \param uid [IN] client uid
141 * \param gid [IN] client gid
142 * \retval user pointer or NULL if not found
145 struct mdd_object_user *mdd_obj_user_find(struct mdd_object *mdd_obj,
146 uid_t uid, gid_t gid,
149 struct mdd_object_user *mou;
154 uidgid = ((__u64)uid << 32) | gid;
155 list_for_each_entry(mou, &mdd_obj->mod_users, mou_list) {
156 if (mou->mou_uidgid == uidgid &&
157 flags_helper(mou->mou_open_flags) ==
158 flags_helper(open_flags))
165 * Add a user to the list of openers for this file
167 * Caller should have write-locked \param mdd_obj.
168 * \param mdd_obj [IN] mdd_obj
169 * \param mou [IN] user
171 * \retval -ve failure
173 static int mdd_obj_user_add(struct mdd_object *mdd_obj,
174 struct mdd_object_user *mou,
177 struct mdd_device *mdd = mdd_obj2mdd_dev(mdd_obj);
178 struct mdd_object_user *tmp;
179 __u32 uid = mou->mou_uidgid >> 32;
180 __u32 gid = mou->mou_uidgid & ((1UL << 32) - 1);
183 tmp = mdd_obj_user_find(mdd_obj, uid, gid, mou->mou_open_flags);
187 list_add_tail(&mou->mou_list, &mdd_obj->mod_users);
190 /* next 'access denied' notification cannot happen before
193 mou->mou_deniednext =
194 ktime_add(ktime_get(),
195 ktime_set(mdd->mdd_cl.mc_deniednext, 0));
197 mou->mou_opencount++;
202 * Remove UID from the list
204 * Caller should have write-locked \param mdd_obj.
205 * \param mdd_obj [IN] mdd_obj
206 * \param uid [IN] user
207 * \retval -ve failure
209 static int mdd_obj_user_remove(struct mdd_object *mdd_obj,
210 struct mdd_object_user *mou)
217 list_del_init(&mou->mou_list);
219 mdd_obj_user_free(mou);
223 int mdd_la_get(const struct lu_env *env, struct mdd_object *obj,
228 if (mdd_object_exists(obj) == 0)
231 rc = mdo_attr_get(env, obj, la);
232 if (unlikely(rc != 0)) {
234 obj->mod_flags |= DEAD_OBJ;
238 if (la->la_valid & LA_FLAGS && la->la_flags & LUSTRE_ORPHAN_FL)
239 obj->mod_flags |= ORPHAN_OBJ | DEAD_OBJ;
244 struct mdd_thread_info *mdd_env_info(const struct lu_env *env)
246 return lu_env_info(env, &mdd_thread_key);
249 struct lu_buf *mdd_buf_get(const struct lu_env *env, void *area, ssize_t len)
253 buf = &mdd_env_info(env)->mdi_buf[0];
259 const struct lu_buf *mdd_buf_get_const(const struct lu_env *env,
260 const void *area, ssize_t len)
264 buf = &mdd_env_info(env)->mdi_buf[0];
265 buf->lb_buf = (void *)area;
270 struct lu_object *mdd_object_alloc(const struct lu_env *env,
271 const struct lu_object_header *hdr,
274 struct mdd_device *mdd = lu2mdd_dev(d);
275 struct mdd_object *mdd_obj;
278 OBD_SLAB_ALLOC_PTR_GFP(mdd_obj, mdd_object_kmem, GFP_NOFS);
282 o = mdd2lu_obj(mdd_obj);
283 lu_object_init(o, NULL, d);
284 mdd_obj->mod_obj.mo_ops = &mdd_obj_ops;
285 mdd_obj->mod_obj.mo_dir_ops = &mdd_dir_ops;
286 mdd_obj->mod_count = 0;
287 obd_obt_init(mdd2obd_dev(mdd));
288 o->lo_ops = &mdd_lu_obj_ops;
289 INIT_LIST_HEAD(&mdd_obj->mod_users);
294 static int mdd_object_init(const struct lu_env *env, struct lu_object *o,
295 const struct lu_object_conf *unused)
297 struct mdd_device *d = lu2mdd_dev(o->lo_dev);
298 struct mdd_object *mdd_obj = lu2mdd_obj(o);
299 struct lu_object *below;
300 struct lu_device *under;
303 mdd_obj->mod_cltime = ktime_set(0, 0);
304 under = &d->mdd_child->dd_lu_dev;
305 below = under->ld_ops->ldo_object_alloc(env, o->lo_header, under);
307 RETURN(PTR_ERR(below));
309 lu_object_add(o, below);
314 static int mdd_object_start(const struct lu_env *env, struct lu_object *o)
318 if (lu_object_exists(o)) {
319 struct mdd_object *mdd_obj = lu2mdd_obj(o);
320 struct lu_attr *attr = MDD_ENV_VAR(env, la_for_start);
322 rc = mdd_la_get(env, mdd_obj, attr);
328 static void mdd_object_free(const struct lu_env *env, struct lu_object *o)
330 struct mdd_object *mdd = lu2mdd_obj(o);
331 struct mdd_object_user *mou, *tmp2;
334 list_for_each_entry_safe(mou, tmp2, &mdd->mod_users, mou_list) {
335 list_del(&mou->mou_list);
336 mdd_obj_user_free(mou);
340 /* mdd doesn't contain an lu_object_header, so don't need call_rcu */
341 OBD_SLAB_FREE_PTR(mdd, mdd_object_kmem);
344 static int mdd_object_print(const struct lu_env *env, void *cookie,
345 lu_printer_t p, const struct lu_object *o)
347 struct mdd_object *mdd = lu2mdd_obj((struct lu_object *)o);
349 return (*p)(env, cookie,
350 LUSTRE_MDD_NAME"-object@%p(open_count=%d, valid=%x, cltime=%lldns, flags=%lx)",
351 mdd, mdd->mod_count, mdd->mod_valid,
352 ktime_to_ns(mdd->mod_cltime), mdd->mod_flags);
355 static const struct lu_object_operations mdd_lu_obj_ops = {
356 .loo_object_init = mdd_object_init,
357 .loo_object_start = mdd_object_start,
358 .loo_object_free = mdd_object_free,
359 .loo_object_print = mdd_object_print,
362 struct mdd_object *mdd_object_find(const struct lu_env *env,
363 struct mdd_device *d,
364 const struct lu_fid *f)
366 return md2mdd_obj(md_object_find_slice(env, &d->mdd_md_dev, f));
370 * No permission check is needed.
372 int mdd_attr_get(const struct lu_env *env, struct md_object *obj,
375 struct mdd_object *mdd_obj = md2mdd_obj(obj);
380 rc = mdd_la_get(env, mdd_obj, &ma->ma_attr);
381 if ((ma->ma_need & MA_INODE) != 0 && mdd_is_dead_obj(mdd_obj))
382 ma->ma_attr.la_nlink = 0;
388 * No permission check is needed.
390 static int mdd_xattr_get(const struct lu_env *env,
391 struct md_object *obj, struct lu_buf *buf,
394 struct mdd_object *mdd_obj = md2mdd_obj(obj);
395 struct mdd_device *mdd;
400 if (mdd_object_exists(mdd_obj) == 0) {
401 CERROR("%s: object "DFID" not found: rc = -2\n",
402 mdd_obj_dev_name(mdd_obj),
403 PFID(mdd_object_fid(mdd_obj)));
407 /* If the object has been destroyed, then do not get LMVEA, because
408 * it needs to load stripes from the iteration of the master object,
409 * and it will cause problem if master object has been destroyed, see
411 if (unlikely((mdd_obj->mod_flags & DEAD_OBJ) &&
412 !(mdd_obj->mod_flags & ORPHAN_OBJ) &&
413 strcmp(name, XATTR_NAME_LMV) == 0))
416 /* If the object has been delete from the namespace, then
417 * get linkEA should return -ENOENT as well */
418 if (unlikely((mdd_obj->mod_flags & (DEAD_OBJ | ORPHAN_OBJ)) &&
419 strcmp(name, XATTR_NAME_LINK) == 0))
422 mdd_read_lock(env, mdd_obj, DT_TGT_CHILD);
423 rc = mdo_xattr_get(env, mdd_obj, buf, name);
424 mdd_read_unlock(env, mdd_obj);
428 /* record only getting user xattrs and acls */
429 if (rc >= 0 && buf->lb_buf &&
430 mdd_changelog_enabled(env, mdd, CL_GETXATTR) &&
431 (has_prefix(name, XATTR_USER_PREFIX) ||
432 has_prefix(name, XATTR_NAME_POSIX_ACL_ACCESS) ||
433 has_prefix(name, XATTR_NAME_POSIX_ACL_DEFAULT))) {
434 struct thandle *handle;
437 LASSERT(mdd_object_fid(mdd_obj) != NULL);
439 handle = mdd_trans_create(env, mdd);
441 RETURN(PTR_ERR(handle));
443 rc2 = mdd_declare_changelog_store(env, mdd, CL_GETXATTR, NULL,
448 rc2 = mdd_trans_start(env, mdd, handle);
452 rc2 = mdd_changelog_data_store_by_fid(env, mdd, CL_GETXATTR, 0,
453 mdd_object_fid(mdd_obj),
457 rc2 = mdd_trans_stop(env, mdd, rc2, handle);
466 * Permission check is done when open,
467 * no need check again.
469 int mdd_readlink(const struct lu_env *env, struct md_object *obj,
472 struct mdd_object *mdd_obj = md2mdd_obj(obj);
473 struct dt_object *next;
478 if (mdd_object_exists(mdd_obj) == 0) {
479 CERROR("%s: object "DFID" not found: rc = -2\n",
480 mdd_obj_dev_name(mdd_obj),PFID(mdd_object_fid(mdd_obj)));
484 next = mdd_object_child(mdd_obj);
485 LASSERT(next != NULL);
486 LASSERT(next->do_body_ops != NULL);
487 LASSERT(next->do_body_ops->dbo_read != NULL);
488 mdd_read_lock(env, mdd_obj, DT_TGT_CHILD);
489 rc = dt_read(env, next, buf, &pos);
490 mdd_read_unlock(env, mdd_obj);
495 * No permission check is needed.
497 static int mdd_xattr_list(const struct lu_env *env, struct md_object *obj,
500 struct mdd_object *mdd_obj = md2mdd_obj(obj);
505 mdd_read_lock(env, mdd_obj, DT_TGT_CHILD);
506 rc = mdo_xattr_list(env, mdd_obj, buf);
507 mdd_read_unlock(env, mdd_obj);
509 /* If the buffer is NULL then we are only here to get the
510 * length of the xattr name list. */
511 if (rc < 0 || buf->lb_buf == NULL)
515 * Filter out XATTR_NAME_LINK if this is an orphan object. See
518 if (unlikely(mdd_obj->mod_flags & (DEAD_OBJ | ORPHAN_OBJ))) {
519 char *end = (char *)buf->lb_buf + rc;
520 char *p = buf->lb_buf;
523 char *next = p + strlen(p) + 1;
525 if (strcmp(p, XATTR_NAME_LINK) == 0) {
527 memmove(p, next, end - next);
529 CDEBUG(D_INFO, "Filtered out "XATTR_NAME_LINK
530 " of orphan "DFID"\n",
531 PFID(mdd_object_fid(mdd_obj)));
542 int mdd_invalidate(const struct lu_env *env, struct md_object *obj)
544 return mdo_invalidate(env, md2mdd_obj(obj));
547 int mdd_declare_create_object_internal(const struct lu_env *env,
548 struct mdd_object *p,
549 struct mdd_object *c,
550 struct lu_attr *attr,
551 struct thandle *handle,
552 const struct md_op_spec *spec,
553 struct dt_allocation_hint *hint)
555 struct dt_object_format *dof = &mdd_env_info(env)->mdi_dof;
556 const struct dt_index_features *feat = spec->sp_feat;
560 if (feat != &dt_directory_features && feat != NULL) {
561 dof->dof_type = DFT_INDEX;
562 dof->u.dof_idx.di_feat = feat;
564 dof->dof_type = dt_mode_to_dft(attr->la_mode);
565 if (dof->dof_type == DFT_REGULAR) {
566 dof->u.dof_reg.striped =
567 md_should_create(spec->sp_cr_flags);
568 if (spec->sp_cr_flags & MDS_OPEN_HAS_EA)
569 dof->u.dof_reg.striped = 0;
570 /* is this replay? */
572 dof->u.dof_reg.striped = 0;
576 rc = mdo_declare_create_object(env, c, attr, hint, dof, handle);
581 int mdd_create_object_internal(const struct lu_env *env, struct mdd_object *p,
582 struct mdd_object *c, struct lu_attr *attr,
583 struct thandle *handle,
584 const struct md_op_spec *spec,
585 struct dt_allocation_hint *hint)
587 struct dt_object_format *dof = &mdd_env_info(env)->mdi_dof;
591 LASSERT(!mdd_object_exists(c));
593 rc = mdo_create_object(env, c, attr, hint, dof, handle);
598 int mdd_attr_set_internal(const struct lu_env *env, struct mdd_object *obj,
599 const struct lu_attr *attr, struct thandle *handle,
605 rc = mdo_attr_set(env, obj, attr, handle);
606 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
607 if (!rc && (attr->la_valid & LA_MODE) && needacl)
608 rc = mdd_acl_chmod(env, obj, attr->la_mode, handle);
613 int mdd_update_time(const struct lu_env *env, struct mdd_object *obj,
614 const struct lu_attr *oattr, struct lu_attr *attr,
615 struct thandle *handle)
620 LASSERT(attr->la_valid & LA_CTIME);
621 LASSERT(oattr != NULL);
623 /* Make sure the ctime is increased only, however, it's not strictly
624 * reliable at here because there is not guarantee to hold lock on
625 * object, so we just bypass some unnecessary cmtime setting first
626 * and OSD has to check it again. */
627 if (attr->la_ctime < oattr->la_ctime)
628 attr->la_valid &= ~(LA_MTIME | LA_CTIME);
629 else if (attr->la_valid == LA_CTIME &&
630 attr->la_ctime == oattr->la_ctime)
631 attr->la_valid &= ~LA_CTIME;
633 if (attr->la_valid != 0)
634 rc = mdd_attr_set_internal(env, obj, attr, handle, 0);
639 static bool is_project_state_change(const struct lu_attr *oattr,
642 if (la->la_valid & LA_PROJID &&
643 oattr->la_projid != la->la_projid)
646 if ((la->la_valid & LA_FLAGS) &&
647 (la->la_flags & LUSTRE_PROJINHERIT_FL) !=
648 (oattr->la_flags & LUSTRE_PROJINHERIT_FL))
655 * This gives the same functionality as the code between
656 * sys_chmod and inode_setattr
657 * chown_common and inode_setattr
658 * utimes and inode_setattr
659 * This API is ported from mds_fix_attr but remove some unnecesssary stuff.
661 static int mdd_fix_attr(const struct lu_env *env, struct mdd_object *obj,
662 const struct lu_attr *oattr, struct lu_attr *la,
663 const struct md_attr *ma)
667 const unsigned long flags = ma->ma_attr_flags;
674 /* Do not permit change file type */
675 if (la->la_valid & LA_TYPE)
678 /* They should not be processed by setattr */
679 if (la->la_valid & (LA_NLINK | LA_RDEV | LA_BLKSIZE))
682 LASSERT(oattr != NULL);
684 uc = lu_ucred_check(env);
688 if (is_project_state_change(oattr, la)) {
689 if (!cap_raised(uc->uc_cap, CAP_SYS_RESOURCE) &&
690 !lustre_in_group_p(uc, ma->ma_enable_chprojid_gid) &&
691 !(ma->ma_enable_chprojid_gid == -1 &&
692 mdd_permission_internal(env, obj, oattr, MAY_WRITE)))
696 if (la->la_valid == LA_CTIME) {
697 if (!(flags & MDS_PERM_BYPASS))
698 /* This is only for set ctime when rename's source is
700 rc = mdd_may_delete(env, NULL, NULL, obj, oattr, NULL,
702 if (rc == 0 && la->la_ctime <= oattr->la_ctime)
703 la->la_valid &= ~LA_CTIME;
707 if (flags & MDS_CLOSE_UPDATE_TIMES &&
708 la->la_valid & (LA_ATIME | LA_MTIME | LA_CTIME)) {
709 /* This is an atime/mtime/ctime attribute update for
712 if (la->la_valid & LA_ATIME &&
713 la->la_atime <= (oattr->la_atime +
714 mdd_obj2mdd_dev(obj)->mdd_atime_diff))
715 la->la_valid &= ~LA_ATIME;
716 if (la->la_valid & LA_CTIME && la->la_ctime <= oattr->la_ctime)
717 la->la_valid &= ~LA_CTIME;
718 if (la->la_valid & LA_MTIME && la->la_mtime <= oattr->la_mtime)
719 la->la_valid &= ~LA_MTIME;
723 /* Check if flags change. */
724 if (la->la_valid & LA_FLAGS) {
725 unsigned int oldflags = oattr->la_flags &
726 (LUSTRE_IMMUTABLE_FL | LUSTRE_APPEND_FL);
727 unsigned int newflags = la->la_flags &
728 (LUSTRE_IMMUTABLE_FL | LUSTRE_APPEND_FL);
730 if ((uc->uc_fsuid != oattr->la_uid) &&
731 !cap_raised(uc->uc_cap, CAP_FOWNER))
734 /* The IMMUTABLE and APPEND_ONLY flags can
735 * only be changed by the relevant capability. */
736 if ((oldflags ^ newflags) &&
737 !cap_raised(uc->uc_cap, CAP_LINUX_IMMUTABLE))
740 if (!S_ISDIR(oattr->la_mode)) {
741 la->la_flags &= ~(LUSTRE_DIRSYNC_FL | LUSTRE_TOPDIR_FL);
742 } else if (la->la_flags & LUSTRE_ENCRYPT_FL) {
743 /* when trying to add encryption flag on dir,
744 * make sure it is empty
746 rc = mdd_dir_is_empty(env, obj);
753 if (oattr->la_flags & (LUSTRE_IMMUTABLE_FL | LUSTRE_APPEND_FL) &&
754 (la->la_valid & ~LA_FLAGS) &&
755 !(flags & MDS_PERM_BYPASS))
758 /* Check for setting the obj time. */
759 if ((la->la_valid & (LA_MTIME | LA_ATIME | LA_CTIME)) &&
760 !(la->la_valid & ~(LA_MTIME | LA_ATIME | LA_CTIME))) {
761 if ((uc->uc_fsuid != oattr->la_uid) &&
762 !cap_raised(uc->uc_cap, CAP_FOWNER)) {
763 rc = mdd_permission_internal(env, obj, oattr,
770 if (la->la_valid & LA_KILL_SUID) {
771 la->la_valid &= ~LA_KILL_SUID;
772 if ((oattr->la_mode & S_ISUID) &&
773 !(la->la_valid & LA_MODE)) {
774 la->la_mode = oattr->la_mode;
775 la->la_valid |= LA_MODE;
777 la->la_mode &= ~S_ISUID;
780 if (la->la_valid & LA_KILL_SGID) {
781 la->la_valid &= ~LA_KILL_SGID;
782 if (((oattr->la_mode & (S_ISGID | S_IXGRP)) ==
783 (S_ISGID | S_IXGRP)) &&
784 !(la->la_valid & LA_MODE)) {
785 la->la_mode = oattr->la_mode;
786 la->la_valid |= LA_MODE;
788 la->la_mode &= ~S_ISGID;
791 /* Make sure a caller can chmod. */
792 if (la->la_valid & LA_MODE) {
793 if (!(flags & MDS_PERM_BYPASS) &&
794 (uc->uc_fsuid != oattr->la_uid) &&
795 !cap_raised(uc->uc_cap, CAP_FOWNER))
798 if (la->la_mode == (umode_t) -1)
799 la->la_mode = oattr->la_mode;
801 la->la_mode = (la->la_mode & S_IALLUGO) |
802 (oattr->la_mode & ~S_IALLUGO);
804 /* Also check the setgid bit! */
805 if (!lustre_in_group_p(uc, (la->la_valid & LA_GID) ?
806 la->la_gid : oattr->la_gid) &&
807 !cap_raised(uc->uc_cap, CAP_FSETID))
808 la->la_mode &= ~S_ISGID;
810 la->la_mode = oattr->la_mode;
813 /* Make sure a caller can chown. */
814 if (la->la_valid & LA_UID) {
815 if (la->la_uid == (uid_t) -1)
816 la->la_uid = oattr->la_uid;
817 if (((uc->uc_fsuid != oattr->la_uid) ||
818 (la->la_uid != oattr->la_uid)) &&
819 !cap_raised(uc->uc_cap, CAP_CHOWN))
822 /* If the user or group of a non-directory has been
823 * changed by a non-root user, remove the setuid bit.
824 * 19981026 David C Niemi <niemi@tux.org>
826 * Changed this to apply to all users, including root,
827 * to avoid some races. This is the behavior we had in
828 * 2.0. The check for non-root was definitely wrong
829 * for 2.2 anyway, as it should have been using
830 * CAP_FSETID rather than fsuid -- 19990830 SD. */
831 if (((oattr->la_mode & S_ISUID) == S_ISUID) &&
832 !S_ISDIR(oattr->la_mode)) {
833 la->la_mode &= ~S_ISUID;
834 la->la_valid |= LA_MODE;
838 /* Make sure caller can chgrp. */
839 if (la->la_valid & LA_GID) {
840 if (la->la_gid == (gid_t) -1)
841 la->la_gid = oattr->la_gid;
842 if (((uc->uc_fsuid != oattr->la_uid) ||
843 ((la->la_gid != oattr->la_gid) &&
844 !lustre_in_group_p(uc, la->la_gid))) &&
845 !cap_raised(uc->uc_cap, CAP_CHOWN))
848 /* Likewise, if the user or group of a non-directory
849 * has been changed by a non-root user, remove the
850 * setgid bit UNLESS there is no group execute bit
851 * (this would be a file marked for mandatory
852 * locking). 19981026 David C Niemi <niemi@tux.org>
854 * Removed the fsuid check (see the comment above) --
856 if (((oattr->la_mode & (S_ISGID | S_IXGRP)) ==
857 (S_ISGID | S_IXGRP)) && !S_ISDIR(oattr->la_mode)) {
858 la->la_mode &= ~S_ISGID;
859 la->la_valid |= LA_MODE;
863 if (la->la_valid & (LA_SIZE | LA_BLOCKS)) {
864 if (!((flags & MDS_OWNEROVERRIDE) &&
865 (uc->uc_fsuid == oattr->la_uid)) &&
866 !(flags & MDS_PERM_BYPASS)) {
867 rc = mdd_permission_internal(env, obj, oattr,
874 if (la->la_valid & LA_CTIME) {
876 * The pure setattr, it has the priority over what is
877 * already set, do not drop it if ctime is equal.
879 if (la->la_ctime < oattr->la_ctime)
880 la->la_valid &= ~(LA_ATIME | LA_MTIME | LA_CTIME);
886 static int mdd_changelog_data_store_by_fid(const struct lu_env *env,
887 struct mdd_device *mdd,
888 enum changelog_rec_type type,
889 enum changelog_rec_flags clf_flags,
890 const struct lu_fid *fid,
891 const struct lu_fid *pfid,
892 const char *xattr_name,
893 struct thandle *handle)
895 const struct lu_ucred *uc = lu_ucred(env);
896 enum changelog_rec_extra_flags xflags = CLFE_INVALID;
897 struct llog_changelog_rec *rec;
902 clf_flags = (clf_flags & CLF_FLAGMASK) | CLF_VERSION | CLF_EXTRA_FLAGS;
905 if (uc->uc_jobid[0] != '\0')
906 clf_flags |= CLF_JOBID;
907 xflags |= CLFE_UIDGID;
910 if (type == CL_OPEN || type == CL_DN_OPEN)
912 if (type == CL_SETXATTR || type == CL_GETXATTR)
913 xflags |= CLFE_XATTR;
915 reclen = llog_data_len(LLOG_CHANGELOG_HDR_SZ +
916 changelog_rec_offset(clf_flags & CLF_SUPPORTED,
917 xflags & CLFE_SUPPORTED));
918 buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mdi_chlg_buf, reclen);
919 if (buf->lb_buf == NULL)
923 rec->cr_hdr.lrh_len = reclen;
924 rec->cr.cr_flags = clf_flags;
925 rec->cr.cr_type = (__u32)type;
926 rec->cr.cr_tfid = *fid;
928 rec->cr.cr_pfid = *pfid;
929 rec->cr.cr_namelen = 0;
931 if (clf_flags & CLF_JOBID)
932 mdd_changelog_rec_ext_jobid(&rec->cr, uc->uc_jobid);
934 if (clf_flags & CLF_EXTRA_FLAGS) {
935 mdd_changelog_rec_ext_extra_flags(&rec->cr, xflags);
936 if (xflags & CLFE_UIDGID)
937 mdd_changelog_rec_extra_uidgid(&rec->cr,
938 uc->uc_uid, uc->uc_gid);
939 if (xflags & CLFE_NID)
940 mdd_changelog_rec_extra_nid(&rec->cr, uc->uc_nid);
941 if (xflags & CLFE_OPEN)
942 mdd_changelog_rec_extra_omode(&rec->cr, clf_flags);
943 if (xflags & CLFE_XATTR) {
944 if (xattr_name == NULL)
946 mdd_changelog_rec_extra_xattr(&rec->cr, xattr_name);
950 rc = mdd_changelog_store(env, mdd, rec, handle);
955 /** Store a data change changelog record
956 * If this fails, we must fail the whole transaction; we don't
957 * want the change to commit without the log entry.
958 * \param mdd_obj - mdd_object of change
959 * \param handle - transaction handle
960 * \param pfid - parent FID for CL_MTIME changelogs
962 int mdd_changelog_data_store(const struct lu_env *env, struct mdd_device *mdd,
963 enum changelog_rec_type type,
964 enum changelog_rec_flags clf_flags,
965 struct mdd_object *mdd_obj, struct thandle *handle,
966 const struct lu_fid *pfid)
970 LASSERT(mdd_obj != NULL);
971 LASSERT(handle != NULL);
973 if (!mdd_changelog_enabled(env, mdd, type))
976 if (mdd_is_volatile_obj(mdd_obj))
979 if ((type >= CL_MTIME) && (type <= CL_ATIME) &&
980 ktime_before(mdd->mdd_cl.mc_starttime, mdd_obj->mod_cltime)) {
981 /* Don't need multiple updates in this log */
982 /* Don't check under lock - no big deal if we get an extra
987 rc = mdd_changelog_data_store_by_fid(env, mdd, type, clf_flags,
988 mdd_object_fid(mdd_obj), pfid,
991 mdd_obj->mod_cltime = ktime_get();
996 int mdd_changelog_data_store_xattr(const struct lu_env *env,
997 struct mdd_device *mdd,
998 enum changelog_rec_type type,
999 enum changelog_rec_flags clf_flags,
1000 struct mdd_object *mdd_obj,
1001 const char *xattr_name,
1002 struct thandle *handle)
1006 LASSERT(mdd_obj != NULL);
1007 LASSERT(handle != NULL);
1009 if (!mdd_changelog_enabled(env, mdd, type))
1012 if (mdd_is_volatile_obj(mdd_obj))
1015 if ((type >= CL_MTIME) && (type <= CL_ATIME) &&
1016 ktime_before(mdd->mdd_cl.mc_starttime, mdd_obj->mod_cltime)) {
1017 /* Don't need multiple updates in this log */
1018 /* Don't check under lock - no big deal if we get an extra
1024 rc = mdd_changelog_data_store_by_fid(env, mdd, type, clf_flags,
1025 mdd_object_fid(mdd_obj), NULL,
1026 xattr_name, handle);
1028 mdd_obj->mod_cltime = ktime_get();
1033 /* only the bottom CLF_FLAGSHIFT bits of @flags are stored in the record,
1034 * except for open flags have a dedicated record to store 32 bits of flags */
1035 static int mdd_changelog(const struct lu_env *env, enum changelog_rec_type type,
1036 enum changelog_rec_flags clf_flags,
1037 struct md_device *m, const struct lu_fid *fid)
1039 struct thandle *handle;
1040 struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
1044 LASSERT(fid != NULL);
1046 /* We'll check this again below, but we check now before we
1047 * start a transaction. */
1048 if (!mdd_changelog_enabled(env, mdd, type))
1051 handle = mdd_trans_create(env, mdd);
1053 RETURN(PTR_ERR(handle));
1055 rc = mdd_declare_changelog_store(env, mdd, type, NULL, NULL, handle);
1059 rc = mdd_trans_start(env, mdd, handle);
1063 rc = mdd_changelog_data_store_by_fid(env, mdd, type, clf_flags,
1064 fid, NULL, NULL, handle);
1067 rc = mdd_trans_stop(env, mdd, rc, handle);
1073 * Save LMA extended attributes with data from \a ma.
1075 * HSM and Size-On-MDS data will be extracted from \ma if they are valid, if
1076 * not, LMA EA will be first read from disk, modified and write back.
1079 /* Precedence for choosing record type when multiple
1080 * attributes change: setattr > mtime > ctime > atime
1081 * (ctime changes when mtime does, plus chmod/chown.
1082 * atime and ctime are independent.) */
1083 static int mdd_attr_set_changelog(const struct lu_env *env,
1084 struct md_object *obj, struct thandle *handle,
1085 const struct lu_fid *pfid, __u64 valid)
1087 struct mdd_device *mdd = mdo2mdd(obj);
1090 bits = (valid & LA_SIZE) ? BIT(CL_TRUNC) : 0;
1091 bits |= (valid & ~(LA_CTIME|LA_MTIME|LA_ATIME)) ? BIT(CL_SETATTR) : 0;
1092 bits |= (valid & LA_MTIME) ? BIT(CL_MTIME) : 0;
1093 bits |= (valid & LA_CTIME) ? BIT(CL_CTIME) : 0;
1094 bits |= (valid & LA_ATIME) ? BIT(CL_ATIME) : 0;
1095 bits = bits & mdd->mdd_cl.mc_current_mask;
1096 /* This is an implementation limit rather than a protocol limit */
1097 BUILD_BUG_ON(CL_LAST > sizeof(int) * 8);
1101 /* The record type is the lowest non-masked set bit */
1104 /* XXX: we only store the low CLF_FLAGMASK bits of la_valid */
1105 return mdd_changelog_data_store(env, mdd, type, valid, md2mdd_obj(obj),
1109 static int mdd_declare_attr_set(const struct lu_env *env,
1110 struct mdd_device *mdd,
1111 struct mdd_object *obj,
1112 const struct lu_attr *attr,
1113 struct thandle *handle)
1117 rc = mdo_declare_attr_set(env, obj, attr, handle);
1121 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1122 if (attr->la_valid & LA_MODE) {
1123 mdd_read_lock(env, obj, DT_TGT_CHILD);
1124 rc = mdo_xattr_get(env, obj, &LU_BUF_NULL,
1125 XATTR_NAME_ACL_ACCESS);
1126 mdd_read_unlock(env, obj);
1127 if (rc == -EOPNOTSUPP || rc == -ENODATA)
1133 struct lu_buf *buf = mdd_buf_get(env, NULL, rc);
1134 rc = mdo_declare_xattr_set(env, obj, buf,
1135 XATTR_NAME_ACL_ACCESS, 0,
1143 rc = mdd_declare_changelog_store(env, mdd, CL_SETXATTR, NULL, NULL,
1152 * permission changes may require sync operation, to mitigate performance
1153 * impact, only do this for dir and when permission is reduced.
1155 * For regular files, version is updated with permission change (see VBR), async
1156 * permission won't cause any issue, while missing permission change on
1157 * directory may affect accessibility of other objects after recovery.
1159 static inline bool permission_needs_sync(const struct lu_attr *old,
1160 const struct lu_attr *new)
1162 if (!S_ISDIR(old->la_mode))
1165 if (new->la_valid & LA_UID && old->la_uid != new->la_uid)
1168 if (new->la_valid & LA_GID && old->la_gid != new->la_gid)
1171 if (new->la_valid & LA_MODE) {
1172 /* turned on sticky bit */
1173 if (!(old->la_mode & S_ISVTX) && (new->la_mode & S_ISVTX))
1176 /* set-GID has no impact on what is allowed, not checked */
1178 /* turned off setuid bit, or one of rwx for someone */
1179 if (((new->la_mode & old->la_mode) & (0777 | S_ISUID)) !=
1180 (old->la_mode & (0777 | S_ISUID)))
1187 static inline __u64 mdd_lmm_dom_size(void *buf)
1189 struct lov_mds_md *lmm = buf;
1190 struct lov_comp_md_v1 *comp_v1;
1191 struct lov_mds_md *v1;
1197 if (le32_to_cpu(lmm->lmm_magic) != LOV_MAGIC_COMP_V1)
1200 comp_v1 = (struct lov_comp_md_v1 *)lmm;
1201 off = le32_to_cpu(comp_v1->lcm_entries[0].lcme_offset);
1202 v1 = (struct lov_mds_md *)((char *)comp_v1 + off);
1204 /* DoM entry is the first entry always */
1205 if (lov_pattern(le32_to_cpu(v1->lmm_pattern)) == LOV_PATTERN_MDT)
1206 return le64_to_cpu(comp_v1->lcm_entries[0].lcme_extent.e_end);
1211 /* set attr and LOV EA at once, return updated attr */
1212 int mdd_attr_set(const struct lu_env *env, struct md_object *obj,
1213 const struct md_attr *ma)
1215 struct mdd_object *mdd_obj = md2mdd_obj(obj);
1216 struct mdd_device *mdd = mdo2mdd(obj);
1217 struct thandle *handle = NULL;
1218 struct lu_attr *la_copy = &mdd_env_info(env)->mdi_la_for_fix;
1219 struct lu_attr *attr = MDD_ENV_VAR(env, cattr);
1220 const struct lu_attr *la = &ma->ma_attr;
1221 struct lu_ucred *uc;
1222 struct lquota_id_info qi;
1223 bool quota_reserved = false;
1224 bool chrgrp_by_unprivileged_user = false;
1228 /* we do not use ->attr_set() for LOV/HSM EA any more */
1229 LASSERT((ma->ma_valid & MA_LOV) == 0);
1230 LASSERT((ma->ma_valid & MA_HSM) == 0);
1232 rc = mdd_la_get(env, mdd_obj, attr);
1236 *la_copy = ma->ma_attr;
1237 rc = mdd_fix_attr(env, mdd_obj, attr, la_copy, ma);
1241 /* no need to setattr anymore */
1242 if (la_copy->la_valid == 0) {
1244 "%s: no valid attribute on "DFID", previous was %#llx\n",
1245 mdd_obj_dev_name(mdd_obj),
1246 PFID(mdd_object_fid(mdd_obj)), la->la_valid);
1251 /* If an unprivileged user changes group of some file,
1252 * the setattr operation will be processed synchronously to
1253 * honor the quota limit of the corresponding group. see LU-5152 */
1254 uc = lu_ucred_check(env);
1255 memset(&qi, 0, sizeof(qi));
1256 if (S_ISREG(attr->la_mode) && la->la_valid & LA_GID &&
1257 la->la_gid != attr->la_gid && uc != NULL && uc->uc_fsuid != 0) {
1258 CDEBUG(D_QUOTA, "%s: reserve quota for changing group: gid=%u size=%llu\n",
1259 mdd2obd_dev(mdd)->obd_name, la->la_gid, la->la_size);
1261 if (la->la_valid & LA_BLOCKS)
1262 qi.lqi_space = la->la_blocks << 9;
1263 else if (la->la_valid & LA_SIZE)
1264 qi.lqi_space = la->la_size;
1265 /* use local attr gotten above */
1266 else if (attr->la_valid & LA_BLOCKS)
1267 qi.lqi_space = attr->la_blocks << 9;
1268 else if (attr->la_valid & LA_SIZE)
1269 qi.lqi_space = attr->la_size;
1271 if (qi.lqi_space > 0) {
1272 qi.lqi_id.qid_gid = la->la_gid;
1273 qi.lqi_type = GRPQUOTA;
1274 qi.lqi_space = toqb(qi.lqi_space);
1275 qi.lqi_is_blk = true;
1276 rc = dt_reserve_or_free_quota(env, mdd->mdd_bottom,
1280 CDEBUG(D_QUOTA, "%s: failed to reserve quota for gid %d size %llu\n",
1281 mdd2obd_dev(mdd)->obd_name,
1282 la->la_gid, qi.lqi_space);
1287 quota_reserved = true;
1288 la_copy->la_valid |= LA_FLAGS;
1291 chrgrp_by_unprivileged_user = true;
1293 /* Flush the possible existing client setattr requests to OSTs
1294 * to keep the order with the current setattr operation that
1295 * will be sent directly to OSTs. see LU-5152 */
1296 /* LU-11303 disable sync as this is too heavyweight.
1297 * This should be replaced with a sync only for the object
1298 * being modified here, not the whole filesystem.
1299 rc = dt_sync(env, mdd->mdd_child);
1305 handle = mdd_trans_create(env, mdd);
1306 if (IS_ERR(handle)) {
1307 rc = PTR_ERR(handle);
1313 rc = mdd_declare_attr_set(env, mdd, mdd_obj, la_copy, handle);
1317 rc = mdd_trans_start(env, mdd, handle);
1321 if (!chrgrp_by_unprivileged_user && mdd->mdd_sync_permission &&
1322 permission_needs_sync(attr, la))
1323 handle->th_sync = 1;
1325 if (la->la_valid & (LA_MTIME | LA_CTIME))
1326 CDEBUG(D_INODE, "setting mtime %llu, ctime %llu\n",
1327 la->la_mtime, la->la_ctime);
1329 mdd_write_lock(env, mdd_obj, DT_TGT_CHILD);
1331 /* LU-10509: setattr of LA_SIZE should be skipped case of DOM,
1332 * otherwise following truncate will do nothing and truncated
1333 * data may be read again. This is a quick fix until LU-11033
1336 if (la_copy->la_valid & LA_SIZE) {
1337 struct lu_buf *lov_buf = mdd_buf_get(env, NULL, 0);
1339 rc = mdd_stripe_get(env, mdd_obj, lov_buf, XATTR_NAME_LOV);
1343 if (mdd_lmm_dom_size(lov_buf->lb_buf) > 0)
1344 la_copy->la_valid &= ~LA_SIZE;
1345 lu_buf_free(lov_buf);
1349 if (la_copy->la_valid) {
1350 rc = mdd_attr_set_internal(env, mdd_obj, la_copy, handle, 1);
1352 if (rc == -EDQUOT && la_copy->la_flags & LUSTRE_SET_SYNC_FL) {
1353 /* rollback to the original gid */
1354 la_copy->la_flags &= ~LUSTRE_SET_SYNC_FL;
1355 la_copy->la_gid = attr->la_gid;
1356 mdd_attr_set_internal(env, mdd_obj, la_copy, handle, 1);
1359 mdd_write_unlock(env, mdd_obj);
1363 rc = mdd_attr_set_changelog(env, obj, handle, &ma->ma_pfid,
1366 if (rc == 0 && quota_reserved) {
1367 struct thandle *sub_th;
1369 sub_th = thandle_get_sub_by_dt(env, handle, mdd->mdd_bottom);
1370 if (unlikely(IS_ERR(sub_th))) {
1372 dt_reserve_or_free_quota(env, mdd->mdd_bottom, &qi);
1374 sub_th->th_reserved_quota = qi;
1376 } else if (quota_reserved) {
1378 dt_reserve_or_free_quota(env, mdd->mdd_bottom, &qi);
1382 rc = mdd_trans_stop(env, mdd, rc, handle);
1387 static int mdd_xattr_sanity_check(const struct lu_env *env,
1388 struct mdd_object *obj,
1389 const struct lu_attr *attr,
1392 struct lu_ucred *uc = lu_ucred_assert(env);
1395 if (attr->la_flags & (LUSTRE_IMMUTABLE_FL | LUSTRE_APPEND_FL))
1398 if (strncmp(XATTR_USER_PREFIX, name,
1399 sizeof(XATTR_USER_PREFIX) - 1) == 0) {
1400 /* For sticky directories, only the owner and privileged user
1401 * can write attributes. */
1402 if (S_ISDIR(attr->la_mode) && (attr->la_mode & S_ISVTX) &&
1403 (uc->uc_fsuid != attr->la_uid) &&
1404 !cap_raised(uc->uc_cap, CAP_FOWNER))
1406 } else if (strcmp(name, XATTR_NAME_SOM) != 0 &&
1407 (uc->uc_fsuid != attr->la_uid) &&
1408 !cap_raised(uc->uc_cap, CAP_FOWNER)) {
1416 * Check if a string begins with a given prefix.
1418 * \param str String to check
1419 * \param prefix Substring to check at the beginning of \a str
1420 * \return true/false whether the condition is verified.
1422 static inline bool has_prefix(const char *str, const char *prefix)
1424 return strncmp(prefix, str, strlen(prefix)) == 0;
1428 * Indicate the kind of changelog to store (if any) for a xattr set/del.
1430 * \param[in] xattr_name Full extended attribute name.
1432 * \return type of changelog to use, or CL_NONE if no changelog is to be emitted
1434 static enum changelog_rec_type
1435 mdd_xattr_changelog_type(const struct lu_env *env, struct mdd_device *mdd,
1436 const char *xattr_name)
1438 /* Layout changes systematically recorded */
1439 if (strcmp(XATTR_NAME_LOV, xattr_name) == 0 ||
1440 strcmp(XATTR_LUSTRE_LOV, xattr_name) == 0 ||
1441 allowed_lustre_lov(xattr_name))
1444 /* HSM information changes systematically recorded */
1445 if (strcmp(XATTR_NAME_HSM, xattr_name) == 0)
1448 /* Avoid logging SOM xattr for every file */
1449 if (strcmp(XATTR_NAME_SOM, xattr_name) == 0)
1452 if (has_prefix(xattr_name, XATTR_USER_PREFIX) ||
1453 has_prefix(xattr_name, XATTR_NAME_POSIX_ACL_ACCESS) ||
1454 has_prefix(xattr_name, XATTR_NAME_POSIX_ACL_DEFAULT) ||
1455 has_prefix(xattr_name, XATTR_TRUSTED_PREFIX) ||
1456 has_prefix(xattr_name, XATTR_SECURITY_PREFIX))
1462 static int mdd_declare_xattr_set(const struct lu_env *env,
1463 struct mdd_device *mdd,
1464 struct mdd_object *obj,
1465 const struct lu_buf *buf,
1467 int fl, struct thandle *handle)
1469 enum changelog_rec_type type;
1472 rc = mdo_declare_xattr_set(env, obj, buf, name, fl, handle);
1476 type = mdd_xattr_changelog_type(env, mdd, name);
1478 return 0; /* no changelog to store */
1480 return mdd_declare_changelog_store(env, mdd, type, NULL, NULL, handle);
1484 * Compare current and future data of HSM EA and add a changelog if needed.
1486 * Caller should have write-locked \param obj.
1488 * \param buf - Future HSM EA content.
1489 * \retval 0 if no changelog is needed or changelog was added properly.
1490 * \retval -ve errno if there was a problem
1492 static int mdd_hsm_update_locked(const struct lu_env *env,
1493 struct md_object *obj,
1494 const struct lu_buf *buf,
1495 struct thandle *handle,
1496 enum changelog_rec_flags *clf_flags)
1498 struct mdd_thread_info *info = mdd_env_info(env);
1499 struct mdd_object *mdd_obj = md2mdd_obj(obj);
1500 struct lu_buf *current_buf;
1501 struct md_hsm *current_mh;
1502 struct md_hsm *new_mh;
1506 OBD_ALLOC_PTR(current_mh);
1507 if (current_mh == NULL)
1510 /* Read HSM attrs from disk */
1511 current_buf = lu_buf_check_and_alloc(&info->mdi_xattr_buf,
1513 mdd_obj2mdd_dev(mdd_obj)->mdd_dt_conf.ddp_max_ea_size,
1515 rc = mdo_xattr_get(env, mdd_obj, current_buf, XATTR_NAME_HSM);
1516 rc = lustre_buf2hsm(current_buf->lb_buf, rc, current_mh);
1517 if (rc < 0 && rc != -ENODATA)
1519 else if (rc == -ENODATA)
1520 current_mh->mh_flags = 0;
1522 /* Map future HSM xattr */
1523 OBD_ALLOC_PTR(new_mh);
1525 GOTO(free, rc = -ENOMEM);
1526 lustre_buf2hsm(buf->lb_buf, buf->lb_len, new_mh);
1530 /* Flags differ, set flags for the changelog that will be added */
1531 if (current_mh->mh_flags != new_mh->mh_flags) {
1532 hsm_set_cl_event(clf_flags, HE_STATE);
1533 if (new_mh->mh_flags & HS_DIRTY)
1534 hsm_set_cl_flags(clf_flags, CLF_HSM_DIRTY);
1537 OBD_FREE_PTR(new_mh);
1540 OBD_FREE_PTR(current_mh);
1544 static int mdd_object_pfid_replace(const struct lu_env *env,
1545 struct mdd_object *o)
1547 struct mdd_device *mdd = mdo2mdd(&o->mod_obj);
1548 struct thandle *handle;
1551 handle = mdd_trans_create(env, mdd);
1553 RETURN(PTR_ERR(handle));
1555 handle->th_complex = 1;
1557 /* it doesn't need to track the PFID update via llog, because LFSCK
1558 * will repair it even it goes wrong */
1559 rc = mdd_declare_xattr_set(env, mdd, o, NULL, XATTR_NAME_FID,
1564 rc = mdd_trans_start(env, mdd, handle);
1568 rc = mdo_xattr_set(env, o, NULL, XATTR_NAME_FID, 0, handle);
1573 mdd_trans_stop(env, mdd, rc, handle);
1578 static int mdd_declare_xattr_del(const struct lu_env *env,
1579 struct mdd_device *mdd,
1580 struct mdd_object *obj,
1582 struct thandle *handle);
1584 static int mdd_xattr_del(const struct lu_env *env, struct md_object *obj,
1587 static int mdd_xattr_merge(const struct lu_env *env, struct md_object *md_obj,
1588 struct md_object *md_vic)
1590 struct mdd_device *mdd = mdo2mdd(md_obj);
1591 struct mdd_object *obj = md2mdd_obj(md_obj);
1592 struct mdd_object *vic = md2mdd_obj(md_vic);
1593 struct lu_buf *buf = &mdd_env_info(env)->mdi_buf[0];
1594 struct lu_buf *buf_vic = &mdd_env_info(env)->mdi_buf[1];
1595 struct lov_mds_md *lmm;
1596 struct thandle *handle;
1600 lock_order = lu_fid_cmp(mdd_object_fid(obj), mdd_object_fid(vic));
1601 if (lock_order == 0) /* same fid */
1604 handle = mdd_trans_create(env, mdd);
1606 RETURN(PTR_ERR(handle));
1608 /* get EA of victim file */
1609 memset(buf_vic, 0, sizeof(*buf_vic));
1610 rc = mdd_stripe_get(env, vic, buf_vic, XATTR_NAME_LOV);
1617 /* parse the layout of victim file */
1618 lmm = buf_vic->lb_buf;
1619 if (le32_to_cpu(lmm->lmm_magic) != LOV_MAGIC_COMP_V1)
1620 GOTO(stop, rc = -EINVAL);
1622 /* save EA of target file for restore */
1623 memset(buf, 0, sizeof(*buf));
1624 rc = mdd_stripe_get(env, obj, buf, XATTR_NAME_LOV);
1628 /* Get rid of the layout from victim object */
1629 rc = mdd_declare_xattr_del(env, mdd, vic, XATTR_NAME_LOV, handle);
1633 rc = mdd_declare_xattr_set(env, mdd, obj, buf_vic, XATTR_NAME_LOV,
1634 LU_XATTR_MERGE, handle);
1638 rc = mdd_trans_start(env, mdd, handle);
1642 if (lock_order > 0) {
1643 mdd_write_lock(env, obj, DT_TGT_CHILD);
1644 mdd_write_lock(env, vic, DT_TGT_CHILD);
1646 mdd_write_lock(env, vic, DT_TGT_CHILD);
1647 mdd_write_lock(env, obj, DT_TGT_CHILD);
1650 rc = mdo_xattr_set(env, obj, buf_vic, XATTR_NAME_LOV, LU_XATTR_MERGE,
1655 rc = mdo_xattr_del(env, vic, XATTR_NAME_LOV, handle);
1657 GOTO(out_restore, rc);
1659 (void)mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, obj, handle,
1661 (void)mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, vic, handle,
1667 int rc2 = mdo_xattr_set(env, obj, buf, XATTR_NAME_LOV,
1668 LU_XATTR_REPLACE, handle);
1670 CERROR("%s: failed rollback of "DFID" layout: file state unknown: rc = %d\n",
1671 mdd_obj_dev_name(obj),
1672 PFID(mdd_object_fid(obj)), rc2);
1676 mdd_write_unlock(env, obj);
1677 mdd_write_unlock(env, vic);
1679 mdd_trans_stop(env, mdd, rc, handle);
1681 lu_buf_free(buf_vic);
1684 (void) mdd_object_pfid_replace(env, obj);
1690 * Extract the mirror with specified mirror id, and store the splitted
1691 * mirror layout to @buf.
1693 * \param[in] comp_v1 mirrored layout
1694 * \param[in] mirror_id the mirror with mirror_id to be extracted
1695 * \param[out] buf store the layout excluding the extracted mirror,
1696 * caller free the buffer we allocated in this function
1697 * \param[out] buf_vic store the extracted layout, caller free the buffer
1698 * we allocated in this function
1700 * \retval 0 on success; < 0 if error happens
1702 static int mdd_split_ea(struct lov_comp_md_v1 *comp_v1, __u16 mirror_id,
1703 struct lu_buf *buf, struct lu_buf *buf_vic)
1705 struct lov_comp_md_v1 *comp_rem;
1706 struct lov_comp_md_v1 *comp_vic;
1707 struct lov_comp_md_entry_v1 *entry;
1708 struct lov_comp_md_entry_v1 *entry_rem;
1709 struct lov_comp_md_entry_v1 *entry_vic;
1711 __u16 comp_cnt, count = 0;
1712 int lmm_size, lmm_size_vic = 0;
1714 int offset, offset_rem, offset_vic;
1716 mirror_cnt = le16_to_cpu(comp_v1->lcm_mirror_count) + 1;
1717 /* comp_v1 should contains more than 1 mirror */
1718 if (mirror_cnt <= 1)
1720 comp_cnt = le16_to_cpu(comp_v1->lcm_entry_count);
1721 lmm_size = le32_to_cpu(comp_v1->lcm_size);
1723 for (i = 0; i < comp_cnt; i++) {
1724 entry = &comp_v1->lcm_entries[i];
1725 if (mirror_id_of(le32_to_cpu(entry->lcme_id)) == mirror_id) {
1727 lmm_size_vic += sizeof(*entry);
1728 lmm_size_vic += le32_to_cpu(entry->lcme_size);
1729 } else if (count > 0) {
1730 /* find the specified mirror */
1738 lu_buf_alloc(buf, lmm_size - lmm_size_vic);
1742 lu_buf_alloc(buf_vic, sizeof(*comp_vic) + lmm_size_vic);
1743 if (!buf_vic->lb_buf) {
1748 comp_rem = (struct lov_comp_md_v1 *)buf->lb_buf;
1749 comp_vic = (struct lov_comp_md_v1 *)buf_vic->lb_buf;
1751 memcpy(comp_rem, comp_v1, sizeof(*comp_v1));
1752 comp_rem->lcm_mirror_count = cpu_to_le16(mirror_cnt - 2);
1753 comp_rem->lcm_entry_count = cpu_to_le32(comp_cnt - count);
1754 comp_rem->lcm_size = cpu_to_le32(lmm_size - lmm_size_vic);
1755 if (!comp_rem->lcm_mirror_count)
1756 comp_rem->lcm_flags = cpu_to_le16(LCM_FL_NONE);
1758 memset(comp_vic, 0, sizeof(*comp_v1));
1759 comp_vic->lcm_magic = cpu_to_le32(LOV_MAGIC_COMP_V1);
1760 comp_vic->lcm_mirror_count = 0;
1761 comp_vic->lcm_entry_count = cpu_to_le32(count);
1762 comp_vic->lcm_size = cpu_to_le32(lmm_size_vic + sizeof(*comp_vic));
1763 comp_vic->lcm_flags = cpu_to_le16(LCM_FL_NONE);
1764 comp_vic->lcm_layout_gen = 0;
1766 offset = sizeof(*comp_v1) + sizeof(*entry) * comp_cnt;
1767 offset_rem = sizeof(*comp_rem) +
1768 sizeof(*entry_rem) * (comp_cnt - count);
1769 offset_vic = sizeof(*comp_vic) + sizeof(*entry_vic) * count;
1770 for (i = j = k = 0; i < comp_cnt; i++) {
1771 struct lov_mds_md *lmm, *lmm_dst;
1774 entry = &comp_v1->lcm_entries[i];
1775 entry_vic = &comp_vic->lcm_entries[j];
1776 entry_rem = &comp_rem->lcm_entries[k];
1778 if (mirror_id_of(le32_to_cpu(entry->lcme_id)) == mirror_id)
1781 /* copy component entry */
1783 memcpy(entry_vic, entry, sizeof(*entry));
1784 entry_vic->lcme_flags &= cpu_to_le32(LCME_FL_INIT);
1785 entry_vic->lcme_offset = cpu_to_le32(offset_vic);
1788 memcpy(entry_rem, entry, sizeof(*entry));
1789 entry_rem->lcme_offset = cpu_to_le32(offset_rem);
1793 lmm = (struct lov_mds_md *)((char *)comp_v1 + offset);
1795 lmm_dst = (struct lov_mds_md *)
1796 ((char *)comp_vic + offset_vic);
1798 lmm_dst = (struct lov_mds_md *)
1799 ((char *)comp_rem + offset_rem);
1801 /* copy component entry blob */
1802 memcpy(lmm_dst, lmm, le32_to_cpu(entry->lcme_size));
1804 /* blob offset advance */
1805 offset += le32_to_cpu(entry->lcme_size);
1807 offset_vic += le32_to_cpu(entry->lcme_size);
1809 offset_rem += le32_to_cpu(entry->lcme_size);
1815 static int mdd_dom_data_truncate(const struct lu_env *env,
1816 struct mdd_device *mdd, struct mdd_object *mo);
1818 static int mdd_xattr_split(const struct lu_env *env, struct md_object *md_obj,
1819 struct md_rejig_data *mrd)
1821 struct mdd_device *mdd = mdo2mdd(md_obj);
1822 struct mdd_object *obj = md2mdd_obj(md_obj);
1823 struct mdd_object *vic = NULL;
1824 struct lu_buf *buf = &mdd_env_info(env)->mdi_buf[0];
1825 struct lu_buf *buf_save = &mdd_env_info(env)->mdi_buf[1];
1826 struct lu_buf *buf_vic = &mdd_env_info(env)->mdi_buf[2];
1827 struct lov_comp_md_v1 *lcm;
1828 struct thandle *handle;
1830 bool dom_stripe = false;
1835 * NULL @mrd_obj means mirror deleting, and use NULL vic to indicate
1839 vic = md2mdd_obj(mrd->mrd_obj);
1841 handle = mdd_trans_create(env, mdd);
1843 RETURN(PTR_ERR(handle));
1845 /* get EA of mirrored file */
1846 memset(buf_save, 0, sizeof(*buf));
1847 rc = mdd_stripe_get(env, obj, buf_save, XATTR_NAME_LOV);
1851 lcm = buf_save->lb_buf;
1852 if (le32_to_cpu(lcm->lcm_magic) != LOV_MAGIC_COMP_V1)
1853 GOTO(stop, rc = -EINVAL);
1856 * Extract the mirror with specified mirror id, and store the splitted
1857 * mirror layout to the victim buffer.
1859 memset(buf, 0, sizeof(*buf));
1860 memset(buf_vic, 0, sizeof(*buf_vic));
1861 rc = mdd_split_ea(lcm, mrd->mrd_mirror_id, buf, buf_vic);
1865 * @buf stores layout w/o the specified mirror, @buf_vic stores the
1869 dom_stripe = mdd_lmm_dom_size(buf_vic->lb_buf) > 0;
1873 * non delete mirror split
1875 * declare obj set remaining layout in @buf, will set obj's
1878 rc = mdd_declare_xattr_set(env, mdd, obj, buf, XATTR_NAME_LOV,
1879 LU_XATTR_SPLIT, handle);
1883 /* declare vic set splitted layout in @buf_vic */
1884 rc = mdd_declare_xattr_set(env, mdd, vic, buf_vic,
1885 XATTR_NAME_LOV, LU_XATTR_SPLIT,
1891 * declare delete mirror objects in @buf_vic, will change obj's
1894 rc = mdd_declare_xattr_set(env, mdd, obj, buf_vic,
1895 XATTR_NAME_LOV, LU_XATTR_PURGE,
1900 /* declare obj set remaining layout in @buf */
1901 rc = mdd_declare_xattr_set(env, mdd, obj, buf,
1902 XATTR_NAME_LOV, LU_XATTR_SPLIT,
1908 rc = mdd_trans_start(env, mdd, handle);
1913 /* don't use the same file to save the splitted mirror */
1914 rc = lu_fid_cmp(mdd_object_fid(obj), mdd_object_fid(vic));
1916 GOTO(stop, rc = -EPERM);
1919 mdd_write_lock(env, obj, DT_TGT_CHILD);
1920 mdd_write_lock(env, vic, DT_TGT_CHILD);
1922 mdd_write_lock(env, vic, DT_TGT_CHILD);
1923 mdd_write_lock(env, obj, DT_TGT_CHILD);
1926 mdd_write_lock(env, obj, DT_TGT_CHILD);
1929 /* set obj's layout in @buf */
1930 rc = mdo_xattr_set(env, obj, buf, XATTR_NAME_LOV, LU_XATTR_SPLIT,
1936 /* set vic's layout in @buf_vic */
1937 rc = mdo_xattr_set(env, vic, buf_vic, XATTR_NAME_LOV,
1938 LU_XATTR_CREATE, handle);
1940 GOTO(out_restore, rc);
1942 /* delete mirror objects */
1943 rc = mdo_xattr_set(env, obj, buf_vic, XATTR_NAME_LOV,
1944 LU_XATTR_PURGE, handle);
1946 GOTO(out_restore, rc);
1949 rc = mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, obj, handle,
1952 GOTO(out_restore, rc);
1955 rc = mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, vic,
1958 GOTO(out_restore, rc);
1963 /* restore obj's in-memory and on-disk layout */
1964 int rc2 = mdo_xattr_set(env, obj, buf_save, XATTR_NAME_LOV,
1965 LU_XATTR_REPLACE, handle);
1967 CERROR("%s: failed rollback "DFID
1968 " layout: file state unknown: rc = %d\n",
1969 mdd_obj_dev_name(obj),
1970 PFID(mdd_object_fid(obj)), rc);
1974 mdd_write_unlock(env, obj);
1976 mdd_write_unlock(env, vic);
1978 rc = mdd_trans_stop(env, mdd, rc, handle);
1980 /* Truncate local DOM data if all went well */
1981 if (!rc && dom_stripe)
1982 mdd_dom_data_truncate(env, mdd, obj);
1984 lu_buf_free(buf_save);
1986 lu_buf_free(buf_vic);
1989 (void) mdd_object_pfid_replace(env, obj);
1994 static int mdd_layout_merge_allowed(const struct lu_env *env,
1995 struct md_object *target,
1996 struct md_object *victim)
1998 struct mdd_object *o1 = md2mdd_obj(target);
2000 /* cannot extend directory's LOVEA */
2001 if (S_ISDIR(mdd_object_type(o1))) {
2002 CERROR("%s: Don't extend directory's LOVEA, just set it.\n",
2003 mdd_obj_dev_name(o1));
2011 * The caller should guarantee to update the object ctime
2012 * after xattr_set if needed.
2014 static int mdd_xattr_set(const struct lu_env *env, struct md_object *obj,
2015 const struct lu_buf *buf, const char *name,
2018 struct mdd_object *mdd_obj = md2mdd_obj(obj);
2019 struct lu_attr *attr = MDD_ENV_VAR(env, cattr);
2020 struct mdd_device *mdd = mdo2mdd(obj);
2021 struct thandle *handle;
2022 enum changelog_rec_type cl_type;
2023 enum changelog_rec_flags clf_flags = 0;
2027 rc = mdd_la_get(env, mdd_obj, attr);
2031 rc = mdd_xattr_sanity_check(env, mdd_obj, attr, name);
2035 if (strcmp(name, XATTR_LUSTRE_LOV) == 0 &&
2036 (fl == LU_XATTR_MERGE || fl == LU_XATTR_SPLIT)) {
2037 struct md_rejig_data *mrd = buf->lb_buf;
2038 struct md_object *victim = mrd->mrd_obj;
2040 if (buf->lb_len != sizeof(*mrd))
2044 if (fl == LU_XATTR_MERGE) {
2045 rc = mdd_layout_merge_allowed(env, obj, victim);
2048 /* merge layout of victim as a mirror of obj's. */
2049 rc = mdd_xattr_merge(env, obj, victim);
2051 rc = mdd_xattr_split(env, obj, mrd);
2056 if (strcmp(name, XATTR_NAME_ACL_ACCESS) == 0 ||
2057 strcmp(name, XATTR_NAME_ACL_DEFAULT) == 0) {
2058 struct posix_acl *acl;
2060 /* user may set empty ACL, which should be treated as removing
2062 acl = posix_acl_from_xattr(&init_user_ns, buf->lb_buf,
2065 RETURN(PTR_ERR(acl));
2067 rc = mdd_xattr_del(env, obj, name);
2070 posix_acl_release(acl);
2073 if (!strcmp(name, XATTR_NAME_ACL_ACCESS)) {
2074 rc = mdd_acl_set(env, mdd_obj, attr, buf, fl);
2078 handle = mdd_trans_create(env, mdd);
2080 RETURN(PTR_ERR(handle));
2082 rc = mdd_declare_xattr_set(env, mdd, mdd_obj, buf, name, fl, handle);
2086 rc = mdd_trans_start(env, mdd, handle);
2090 mdd_write_lock(env, mdd_obj, DT_TGT_CHILD);
2092 if (strcmp(XATTR_NAME_HSM, name) == 0) {
2093 rc = mdd_hsm_update_locked(env, obj, buf, handle, &clf_flags);
2095 mdd_write_unlock(env, mdd_obj);
2100 rc = mdo_xattr_set(env, mdd_obj, buf, name, fl, handle);
2101 mdd_write_unlock(env, mdd_obj);
2105 cl_type = mdd_xattr_changelog_type(env, mdd, name);
2109 rc = mdd_changelog_data_store_xattr(env, mdd, cl_type, clf_flags,
2110 mdd_obj, name, handle);
2114 return mdd_trans_stop(env, mdd, rc, handle);
2117 static int mdd_declare_xattr_del(const struct lu_env *env,
2118 struct mdd_device *mdd,
2119 struct mdd_object *obj,
2121 struct thandle *handle)
2123 enum changelog_rec_type type;
2126 rc = mdo_declare_xattr_del(env, obj, name, handle);
2130 type = mdd_xattr_changelog_type(env, mdd, name);
2132 return 0; /* no changelog to store */
2134 return mdd_declare_changelog_store(env, mdd, type, NULL, NULL, handle);
2138 * The caller should guarantee to update the object ctime
2139 * after xattr_set if needed.
2141 static int mdd_xattr_del(const struct lu_env *env, struct md_object *obj,
2144 struct mdd_object *mdd_obj = md2mdd_obj(obj);
2145 struct lu_attr *attr = MDD_ENV_VAR(env, cattr);
2146 struct mdd_device *mdd = mdo2mdd(obj);
2147 struct thandle *handle;
2151 rc = mdd_la_get(env, mdd_obj, attr);
2155 rc = mdd_xattr_sanity_check(env, mdd_obj, attr, name);
2159 handle = mdd_trans_create(env, mdd);
2161 RETURN(PTR_ERR(handle));
2163 rc = mdd_declare_xattr_del(env, mdd, mdd_obj, name, handle);
2167 rc = mdd_trans_start(env, mdd, handle);
2171 mdd_write_lock(env, mdd_obj, DT_TGT_CHILD);
2172 rc = mdo_xattr_del(env, mdd_obj, name, handle);
2173 mdd_write_unlock(env, mdd_obj);
2177 if (mdd_xattr_changelog_type(env, mdd, name) < 0)
2180 rc = mdd_changelog_data_store_xattr(env, mdd, CL_SETXATTR, 0, mdd_obj,
2185 return mdd_trans_stop(env, mdd, rc, handle);
2189 * read lov/lmv EA of an object
2190 * return the lov/lmv EA in an allocated lu_buf
2192 int mdd_stripe_get(const struct lu_env *env, struct mdd_object *obj,
2193 struct lu_buf *lmm_buf, const char *name)
2195 struct lu_buf *buf = &mdd_env_info(env)->mdi_big_buf;
2200 if (buf->lb_buf == NULL) {
2201 buf = lu_buf_check_and_alloc(buf, 4096);
2202 if (buf->lb_buf == NULL)
2207 rc = mdo_xattr_get(env, obj, buf, name);
2208 if (rc == -ERANGE) {
2209 /* mdi_big_buf is allocated but is too small
2210 * we need to increase it */
2211 buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mdi_big_buf,
2213 if (buf->lb_buf == NULL)
2216 } else if (rc < 0) {
2218 } else if (rc == 0) {
2222 lu_buf_alloc(lmm_buf, rc);
2223 if (lmm_buf->lb_buf == NULL)
2227 * we don't use lmm_buf directly, because we don't know xattr size, so
2228 * by using mdi_big_buf we can avoid calling mdo_xattr_get() twice.
2230 memcpy(lmm_buf->lb_buf, buf->lb_buf, rc);
2235 static int mdd_xattr_hsm_replace(const struct lu_env *env,
2236 struct mdd_object *o, struct lu_buf *buf,
2237 struct thandle *handle)
2239 struct hsm_attrs *attrs;
2240 enum hsm_states hsm_flags;
2241 enum changelog_rec_flags clf_flags = 0;
2245 rc = mdo_xattr_set(env, o, buf, XATTR_NAME_HSM, LU_XATTR_REPLACE,
2250 attrs = buf->lb_buf;
2251 hsm_flags = le32_to_cpu(attrs->hsm_flags);
2252 if (!(hsm_flags & HS_RELEASED) || mdd_is_dead_obj(o))
2255 /* Add a changelog record for release. */
2256 hsm_set_cl_event(&clf_flags, HE_RELEASE);
2257 rc = mdd_changelog_data_store(env, mdo2mdd(&o->mod_obj), CL_HSM,
2258 clf_flags, o, handle, NULL);
2263 * check if layout swapping between 2 objects is allowed
2265 * - only normal FIDs or non-system IGIFs
2266 * - same type of objects
2267 * - same owner/group (so quotas are still valid) unless this is from HSM
2270 static int mdd_layout_swap_allowed(const struct lu_env *env,
2271 struct mdd_object *o1,
2272 const struct lu_attr *attr1,
2273 struct mdd_object *o2,
2274 const struct lu_attr *attr2,
2277 const struct lu_fid *fid1, *fid2;
2280 fid1 = mdd_object_fid(o1);
2281 fid2 = mdd_object_fid(o2);
2283 if (!fid_is_norm(fid1) &&
2284 (!fid_is_igif(fid1) || IS_ERR(mdd_links_get(env, o1))))
2287 if (!fid_is_norm(fid2) &&
2288 (!fid_is_igif(fid2) || IS_ERR(mdd_links_get(env, o2))))
2291 if (mdd_object_type(o1) != mdd_object_type(o2)) {
2292 if (S_ISDIR(mdd_object_type(o1)))
2294 if (S_ISREG(mdd_object_type(o1)))
2299 if (flags & SWAP_LAYOUTS_MDS_HSM)
2302 if ((attr1->la_uid != attr2->la_uid) ||
2303 (attr1->la_gid != attr2->la_gid))
2309 /* XXX To set the proper lmm_oi & lmm_layout_gen when swap layouts, we have to
2310 * look into the layout in MDD layer. */
2311 static int mdd_lmm_oi(struct lov_mds_md *lmm, struct ost_id *oi, bool get)
2313 struct lov_comp_md_v1 *comp_v1;
2314 struct lov_mds_md *v1;
2318 if (le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_COMP_V1) {
2319 comp_v1 = (struct lov_comp_md_v1 *)lmm;
2320 ent_count = le16_to_cpu(comp_v1->lcm_entry_count);
2326 off = le32_to_cpu(comp_v1->lcm_entries[0].lcme_offset);
2327 v1 = (struct lov_mds_md *)((char *)comp_v1 + off);
2330 for (i = 0; i < le32_to_cpu(ent_count); i++) {
2331 off = le32_to_cpu(comp_v1->lcm_entries[i].
2333 v1 = (struct lov_mds_md *)((char *)comp_v1 +
2338 } else if (le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_V1 ||
2339 le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_V3) {
2350 static inline int mdd_get_lmm_oi(struct lov_mds_md *lmm, struct ost_id *oi)
2352 return mdd_lmm_oi(lmm, oi, true);
2355 static inline int mdd_set_lmm_oi(struct lov_mds_md *lmm, struct ost_id *oi)
2357 return mdd_lmm_oi(lmm, oi, false);
2360 static int mdd_lmm_gen(struct lov_mds_md *lmm, __u32 *gen, bool get)
2362 struct lov_comp_md_v1 *comp_v1;
2364 if (le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_COMP_V1) {
2365 comp_v1 = (struct lov_comp_md_v1 *)lmm;
2367 *gen = le32_to_cpu(comp_v1->lcm_layout_gen);
2369 comp_v1->lcm_layout_gen = cpu_to_le32(*gen);
2370 } else if (le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_V1 ||
2371 le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_V3) {
2372 __u16 tmp_gen = *gen;
2374 *gen = le16_to_cpu(lmm->lmm_layout_gen);
2376 lmm->lmm_layout_gen = cpu_to_le16(tmp_gen);
2383 static inline int mdd_get_lmm_gen(struct lov_mds_md *lmm, __u32 *gen)
2385 return mdd_lmm_gen(lmm, gen, true);
2388 static inline int mdd_set_lmm_gen(struct lov_mds_md *lmm, __u32 *gen)
2390 return mdd_lmm_gen(lmm, gen, false);
2393 static int mdd_dom_data_truncate(const struct lu_env *env,
2394 struct mdd_device *mdd, struct mdd_object *mo)
2397 struct dt_object *dom;
2400 dom = dt_object_locate(mdd_object_child(mo), mdd->mdd_bottom);
2402 GOTO(out, rc = -ENODATA);
2404 th = dt_trans_create(env, mdd->mdd_bottom);
2406 GOTO(out, rc = PTR_ERR(th));
2408 rc = dt_declare_punch(env, dom, 0, OBD_OBJECT_EOF, th);
2412 rc = dt_trans_start_local(env, mdd->mdd_bottom, th);
2416 rc = dt_punch(env, dom, 0, OBD_OBJECT_EOF, th);
2418 dt_trans_stop(env, mdd->mdd_bottom, th);
2420 /* Ignore failure but report the error */
2422 CERROR("%s: can't truncate DOM inode "DFID" data: rc = %d\n",
2423 mdd_obj_dev_name(mo), PFID(mdd_object_fid(mo)), rc);
2428 * swap layouts between 2 lustre objects
2430 static int mdd_swap_layouts(const struct lu_env *env, struct md_object *obj1,
2431 struct md_object *obj2, __u64 flags)
2433 struct mdd_thread_info *info = mdd_env_info(env);
2434 struct mdd_object *fst_o = md2mdd_obj(obj1);
2435 struct mdd_object *snd_o = md2mdd_obj(obj2);
2436 struct lu_attr *fst_la = MDD_ENV_VAR(env, cattr);
2437 struct lu_attr *snd_la = MDD_ENV_VAR(env, tattr);
2438 struct mdd_device *mdd = mdo2mdd(obj1);
2439 struct lov_mds_md *fst_lmm, *snd_lmm;
2440 struct lu_buf *fst_buf = &info->mdi_buf[0];
2441 struct lu_buf *snd_buf = &info->mdi_buf[1];
2442 struct lu_buf *fst_hsm_buf = &info->mdi_buf[2];
2443 struct lu_buf *snd_hsm_buf = &info->mdi_buf[3];
2444 struct ost_id *saved_oi = NULL;
2445 struct thandle *handle;
2446 struct mdd_object *dom_o = NULL;
2447 __u64 domsize_dom, domsize_vlt;
2448 __u32 fst_gen, snd_gen, saved_gen;
2454 BUILD_BUG_ON(ARRAY_SIZE(info->mdi_buf) < 4);
2455 memset(info->mdi_buf, 0, sizeof(info->mdi_buf));
2457 /* we have to sort the 2 obj, so locking will always
2458 * be in the same order, even in case of 2 concurrent swaps */
2459 rc = lu_fid_cmp(mdd_object_fid(fst_o), mdd_object_fid(snd_o));
2460 if (rc == 0) /* same fid ? */
2466 rc = mdd_la_get(env, fst_o, fst_la);
2470 rc = mdd_la_get(env, snd_o, snd_la);
2474 /* check if layout swapping is allowed */
2475 rc = mdd_layout_swap_allowed(env, fst_o, fst_la, snd_o, snd_la, flags);
2479 handle = mdd_trans_create(env, mdd);
2481 RETURN(PTR_ERR(handle));
2483 /* objects are already sorted */
2484 mdd_write_lock(env, fst_o, DT_TGT_CHILD);
2485 mdd_write_lock(env, snd_o, DT_TGT_CHILD);
2487 rc = mdd_stripe_get(env, fst_o, fst_buf, XATTR_NAME_LOV);
2488 if (rc < 0 && rc != -ENODATA)
2491 rc = mdd_stripe_get(env, snd_o, snd_buf, XATTR_NAME_LOV);
2492 if (rc < 0 && rc != -ENODATA)
2495 /* check if file has DoM. DoM file can be migrated only to another
2496 * DoM layout with the same DoM component size or to an non-DOM
2497 * layout. After migration to OSTs layout, local MDT inode data
2498 * should be truncated.
2499 * Objects are sorted by FIDs, considering that original file's FID
2500 * is always smaller the snd_o is always original file we are migrating
2503 domsize_dom = mdd_lmm_dom_size(snd_buf->lb_buf);
2504 domsize_vlt = mdd_lmm_dom_size(fst_buf->lb_buf);
2506 /* Only migration is supported for DoM files, not 'swap_layouts' so
2507 * target file must be volatile and orphan.
2509 if (fst_o->mod_flags & (ORPHAN_OBJ | VOLATILE_OBJ)) {
2510 dom_o = domsize_dom ? snd_o : NULL;
2511 } else if (snd_o->mod_flags & (ORPHAN_OBJ | VOLATILE_OBJ)) {
2512 swap(domsize_dom, domsize_vlt);
2513 dom_o = domsize_dom ? fst_o : NULL;
2514 } else if (domsize_dom > 0 || domsize_vlt > 0) {
2515 /* 'lfs swap_layouts' case, neither file should have DoM */
2517 CDEBUG(D_LAYOUT, "cannot swap layouts with DOM component, "
2518 "use migration instead: rc = %d\n", rc);
2522 if (domsize_vlt > 0 && domsize_dom == 0) {
2525 "%s: cannot swap "DFID" layout: OST to DOM migration not supported: rc = %d\n",
2526 mdd_obj_dev_name(snd_o),
2527 PFID(mdd_object_fid(snd_o)), rc);
2529 } else if (domsize_vlt > 0 && domsize_dom != domsize_vlt) {
2532 "%s: cannot swap "DFID" layout: new layout must have same DoM component size: rc = %d\n",
2533 mdd_obj_dev_name(fst_o),
2534 PFID(mdd_object_fid(fst_o)), rc);
2536 } else if (domsize_vlt > 0) {
2537 /* Migration with the same DOM component size, no need to
2538 * truncate local data, it is still being used */
2542 /* swapping 2 non existant layouts is a success */
2543 if (fst_buf->lb_buf == NULL && snd_buf->lb_buf == NULL)
2546 /* to help inode migration between MDT, it is better to
2547 * start by the no layout file (if one), so we order the swap */
2548 if (snd_buf->lb_buf == NULL) {
2550 swap(fst_buf, snd_buf);
2553 fst_gen = snd_gen = 0;
2554 /* lmm and generation layout initialization */
2555 if (fst_buf->lb_buf != NULL) {
2556 fst_lmm = fst_buf->lb_buf;
2557 mdd_get_lmm_gen(fst_lmm, &fst_gen);
2558 fst_fl = LU_XATTR_REPLACE;
2562 fst_fl = LU_XATTR_CREATE;
2565 snd_lmm = snd_buf->lb_buf;
2566 mdd_get_lmm_gen(snd_lmm, &snd_gen);
2568 saved_gen = fst_gen;
2569 /* increase the generation layout numbers */
2574 * XXX The layout generation is used to generate component IDs for
2575 * the composite file, we have to do some special tweaks to make
2576 * sure the layout generation is always adequate for that job.
2579 /* Skip invalid generation number for composite layout */
2580 if ((snd_gen & LCME_ID_MASK) == 0)
2582 if ((fst_gen & LCME_ID_MASK) == 0)
2584 /* Make sure the generation is greater than all the component IDs */
2585 if (fst_gen < snd_gen)
2587 else if (fst_gen > snd_gen)
2590 /* set the file specific informations in lmm */
2591 if (fst_lmm != NULL) {
2592 struct ost_id temp_oi;
2594 saved_oi = &info->mdi_oa.o_oi;
2595 mdd_get_lmm_oi(fst_lmm, saved_oi);
2596 mdd_get_lmm_oi(snd_lmm, &temp_oi);
2597 mdd_set_lmm_gen(fst_lmm, &snd_gen);
2598 mdd_set_lmm_oi(fst_lmm, &temp_oi);
2599 mdd_set_lmm_oi(snd_lmm, saved_oi);
2601 if ((snd_lmm->lmm_magic & cpu_to_le32(LOV_MAGIC_MASK)) ==
2602 cpu_to_le32(LOV_MAGIC_MAGIC))
2603 snd_lmm->lmm_magic |= cpu_to_le32(LOV_MAGIC_DEFINED);
2605 GOTO(stop, rc = -EPROTO);
2607 mdd_set_lmm_gen(snd_lmm, &fst_gen);
2609 /* Prepare HSM attribute if it's required */
2610 if (flags & SWAP_LAYOUTS_MDS_HSM) {
2611 const int buflen = sizeof(struct hsm_attrs);
2613 lu_buf_alloc(fst_hsm_buf, buflen);
2614 lu_buf_alloc(snd_hsm_buf, buflen);
2615 if (fst_hsm_buf->lb_buf == NULL || snd_hsm_buf->lb_buf == NULL)
2616 GOTO(stop, rc = -ENOMEM);
2618 /* Read HSM attribute */
2619 rc = mdo_xattr_get(env, fst_o, fst_hsm_buf, XATTR_NAME_HSM);
2623 rc = mdo_xattr_get(env, snd_o, snd_hsm_buf, XATTR_NAME_HSM);
2627 rc = mdd_declare_xattr_set(env, mdd, fst_o, snd_hsm_buf,
2628 XATTR_NAME_HSM, LU_XATTR_REPLACE,
2633 rc = mdd_declare_xattr_set(env, mdd, snd_o, fst_hsm_buf,
2634 XATTR_NAME_HSM, LU_XATTR_REPLACE,
2640 /* prepare transaction */
2641 rc = mdd_declare_xattr_set(env, mdd, fst_o, snd_buf, XATTR_NAME_LOV,
2646 if (fst_buf->lb_buf != NULL)
2647 rc = mdd_declare_xattr_set(env, mdd, snd_o, fst_buf,
2648 XATTR_NAME_LOV, LU_XATTR_REPLACE,
2651 rc = mdd_declare_xattr_del(env, mdd, snd_o, XATTR_NAME_LOV,
2656 rc = mdd_trans_start(env, mdd, handle);
2660 if (flags & SWAP_LAYOUTS_MDS_HSM) {
2661 rc = mdd_xattr_hsm_replace(env, fst_o, snd_hsm_buf, handle);
2665 rc = mdd_xattr_hsm_replace(env, snd_o, fst_hsm_buf, handle);
2667 rc2 = mdd_xattr_hsm_replace(env, fst_o, fst_hsm_buf,
2670 CERROR("%s: HSM error restoring "DFID": rc = %d/%d\n",
2671 mdd_obj_dev_name(fst_o),
2672 PFID(mdd_object_fid(fst_o)), rc, rc2);
2677 rc = mdo_xattr_set(env, fst_o, snd_buf, XATTR_NAME_LOV, fst_fl, handle);
2681 if (unlikely(OBD_FAIL_CHECK(OBD_FAIL_MDS_HSM_SWAP_LAYOUTS))) {
2684 if (fst_buf->lb_buf != NULL)
2685 rc = mdo_xattr_set(env, snd_o, fst_buf, XATTR_NAME_LOV,
2686 LU_XATTR_REPLACE, handle);
2688 rc = mdo_xattr_del(env, snd_o, XATTR_NAME_LOV, handle);
2691 GOTO(out_restore, rc);
2693 /* Issue one changelog record per file */
2694 rc = mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, fst_o, handle,
2699 rc = mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, snd_o, handle,
2709 /* failure on second file, but first was done, so we have
2710 * to roll back first. */
2711 if (fst_buf->lb_buf != NULL) {
2712 mdd_set_lmm_oi(fst_lmm, saved_oi);
2713 mdd_set_lmm_gen(fst_lmm, &saved_gen);
2714 rc2 = mdo_xattr_set(env, fst_o, fst_buf, XATTR_NAME_LOV,
2715 LU_XATTR_REPLACE, handle);
2717 rc2 = mdo_xattr_del(env, fst_o, XATTR_NAME_LOV, handle);
2722 if (flags & SWAP_LAYOUTS_MDS_HSM) {
2724 rc2 = mdd_xattr_hsm_replace(env, fst_o, fst_hsm_buf,
2730 rc2 = mdd_xattr_hsm_replace(env, snd_o, snd_hsm_buf,
2737 CERROR("%s: unable to roll back layout swap of "DFID" and "DFID", steps: %d: rc = %d/%d\n",
2738 mdd_obj_dev_name(fst_o),
2739 PFID(mdd_object_fid(snd_o)),
2740 PFID(mdd_object_fid(fst_o)),
2742 /* a solution to avoid journal commit is to panic,
2743 * but it has strong consequences so we use LBUG to
2744 * allow sysdamin to choose to panic or not
2751 rc = mdd_trans_stop(env, mdd, rc, handle);
2753 /* Truncate local DOM data if all went well */
2755 mdd_dom_data_truncate(env, mdd, dom_o);
2757 mdd_write_unlock(env, snd_o);
2758 mdd_write_unlock(env, fst_o);
2760 lu_buf_free(fst_buf);
2761 lu_buf_free(snd_buf);
2762 lu_buf_free(fst_hsm_buf);
2763 lu_buf_free(snd_hsm_buf);
2766 (void) mdd_object_pfid_replace(env, fst_o);
2767 (void) mdd_object_pfid_replace(env, snd_o);
2772 static int mdd_declare_layout_change(const struct lu_env *env,
2773 struct mdd_device *mdd,
2774 struct mdd_object *obj,
2775 struct md_layout_change *mlc,
2776 struct thandle *handle)
2780 rc = mdo_declare_layout_change(env, obj, mlc, handle);
2784 return mdd_declare_changelog_store(env, mdd, CL_LAYOUT, NULL, NULL,
2788 /* For PFL, this is used to instantiate necessary component objects. */
2790 mdd_layout_instantiate_component(const struct lu_env *env,
2791 struct mdd_object *obj, struct md_layout_change *mlc,
2792 struct thandle *handle)
2794 struct mdd_device *mdd = mdd_obj2mdd_dev(obj);
2798 if (mlc->mlc_opc != MD_LAYOUT_WRITE)
2801 rc = mdd_declare_layout_change(env, mdd, obj, mlc, handle);
2803 * It's possible that another layout write intent has already
2804 * instantiated our objects, so a -EALREADY returned, and we need to
2808 RETURN(rc == -EALREADY ? 0 : rc);
2810 rc = mdd_trans_start(env, mdd, handle);
2814 mdd_write_lock(env, obj, DT_TGT_CHILD);
2815 rc = mdo_layout_change(env, obj, mlc, handle);
2816 mdd_write_unlock(env, obj);
2820 rc = mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, obj, handle,
2826 * Change the FLR layout from RDONLY to WRITE_PENDING.
2828 * It picks the primary mirror, and bumps the layout version, and set
2829 * layout version xattr to OST objects in a sync tx. In order to facilitate
2830 * the handling of phantom writers from evicted clients, the clients carry
2831 * layout version of the file with write RPC, so that the OSTs can verify
2832 * if the write RPCs are legitimate, meaning not from evicted clients.
2835 mdd_layout_update_rdonly(const struct lu_env *env, struct mdd_object *obj,
2836 struct md_layout_change *mlc, struct thandle *handle)
2838 struct mdd_device *mdd = mdd_obj2mdd_dev(obj);
2839 struct lu_buf *som_buf = &mdd_env_info(env)->mdi_buf[1];
2840 struct lustre_som_attrs *som = &mlc->mlc_som;
2845 /* Verify acceptable operations */
2846 switch (mlc->mlc_opc) {
2847 case MD_LAYOUT_WRITE:
2848 case MD_LAYOUT_RESYNC:
2849 /* these are legal operations - this represents the case that
2850 * a few mirrors were missed in the last resync. */
2852 case MD_LAYOUT_RESYNC_DONE:
2857 som_buf->lb_buf = som;
2858 som_buf->lb_len = sizeof(*som);
2859 rc = mdo_xattr_get(env, obj, som_buf, XATTR_NAME_SOM);
2860 if (rc < 0 && rc != -ENODATA)
2864 lustre_som_swab(som);
2865 if (som->lsa_valid & SOM_FL_STRICT)
2866 fl = LU_XATTR_REPLACE;
2868 if (mlc->mlc_opc == MD_LAYOUT_WRITE &&
2869 mlc->mlc_intent->li_extent.e_end > som->lsa_size) {
2870 som->lsa_size = mlc->mlc_intent->li_extent.e_end + 1;
2871 fl = LU_XATTR_REPLACE;
2875 rc = mdd_declare_layout_change(env, mdd, obj, mlc, handle);
2880 rc = mdd_declare_xattr_set(env, mdd, obj, som_buf,
2881 XATTR_NAME_SOM, fl, handle);
2886 /* record a changelog for data mover to consume */
2887 rc = mdd_declare_changelog_store(env, mdd, CL_FLRW, NULL, NULL, handle);
2891 rc = mdd_trans_start(env, mdd, handle);
2895 /* it needs a sync tx to make FLR to work properly */
2896 handle->th_sync = 1;
2898 mdd_write_lock(env, obj, DT_TGT_CHILD);
2899 rc = mdo_layout_change(env, obj, mlc, handle);
2901 /* SOM state transition from STRICT to STALE */
2902 som->lsa_valid = SOM_FL_STALE;
2903 lustre_som_swab(som);
2904 rc = mdo_xattr_set(env, obj, som_buf, XATTR_NAME_SOM,
2907 mdd_write_unlock(env, obj);
2911 rc = mdd_changelog_data_store(env, mdd, CL_FLRW, 0, obj, handle, NULL);
2922 * Handle mirrored file state transition when it's in WRITE_PENDING.
2924 * Only MD_LAYOUT_RESYNC, which represents start of resync, is allowed when
2925 * the file is in WRITE_PENDING state. If everything goes fine, the file's
2926 * layout version will be increased, and the file's state will be changed to
2930 mdd_layout_update_write_pending(const struct lu_env *env,
2931 struct mdd_object *obj, struct md_layout_change *mlc,
2932 struct thandle *handle)
2934 struct mdd_device *mdd = mdd_obj2mdd_dev(obj);
2935 struct lu_buf *som_buf = &mdd_env_info(env)->mdi_buf[1];
2936 struct lustre_som_attrs *som = &mlc->mlc_som;
2941 switch (mlc->mlc_opc) {
2942 case MD_LAYOUT_RESYNC:
2943 /* Upon receiving the resync request, it should
2944 * instantiate all stale components right away to get ready
2945 * for mirror copy. In order to avoid layout version change,
2946 * client should avoid sending LAYOUT_WRITE request at the
2949 case MD_LAYOUT_WRITE:
2951 * legal race for concurrent write, the file state has been
2952 * changed by another client. Or a jump over file size and
2955 som_buf->lb_buf = som;
2956 som_buf->lb_len = sizeof(*som);
2957 rc = mdo_xattr_get(env, obj, som_buf, XATTR_NAME_SOM);
2958 if (rc < 0 && rc != -ENODATA)
2962 lustre_som_swab(som);
2963 if (mlc->mlc_intent->li_extent.e_end > som->lsa_size) {
2965 mlc->mlc_intent->li_extent.e_end + 1;
2966 fl = LU_XATTR_REPLACE;
2974 rc = mdd_declare_layout_change(env, mdd, obj, mlc, handle);
2979 rc = mdd_declare_xattr_set(env, mdd, obj, som_buf,
2980 XATTR_NAME_SOM, fl, handle);
2985 rc = mdd_trans_start(env, mdd, handle);
2989 /* it needs a sync tx to make FLR to work properly */
2990 handle->th_sync = 1;
2992 mdd_write_lock(env, obj, DT_TGT_CHILD);
2993 rc = mdo_layout_change(env, obj, mlc, handle);
2995 som->lsa_valid = SOM_FL_STALE;
2996 lustre_som_swab(som);
2997 rc = mdo_xattr_set(env, obj, som_buf, XATTR_NAME_SOM,
3000 mdd_write_unlock(env, obj);
3011 * Handle the requests when a FLR file's state is in SYNC_PENDING.
3013 * Only concurrent write and sync complete requests are possible when the
3014 * file is in SYNC_PENDING. For the latter request, it will pass in the
3015 * mirrors that have been synchronized, then the stale bit will be cleared
3016 * to make the file's state turn into RDONLY.
3017 * For concurrent write reqeust, it just needs to change the file's state
3018 * to WRITE_PENDING in a sync tx. It doesn't have to change the layout
3019 * version because the version will be increased in the transition to
3020 * SYNC_PENDING later so that it can deny the write request from potential
3021 * evicted SYNC clients. */
3023 mdd_object_update_sync_pending(const struct lu_env *env, struct mdd_object *obj,
3024 struct md_layout_change *mlc, struct thandle *handle)
3026 struct mdd_device *mdd = mdd_obj2mdd_dev(obj);
3027 struct lu_buf *som_buf = &mdd_env_info(env)->mdi_buf[1];
3032 /* operation validation */
3033 switch (mlc->mlc_opc) {
3034 case MD_LAYOUT_RESYNC_DONE:
3035 /* resync complete. */
3036 case MD_LAYOUT_WRITE:
3037 /* concurrent write. */
3039 case MD_LAYOUT_RESYNC:
3040 /* resync again, most likely the previous run failed.
3041 * no-op if it's already in SYNC_PENDING state */
3047 if (mlc->mlc_som.lsa_valid & SOM_FL_STRICT) {
3048 rc = mdo_xattr_get(env, obj, &LU_BUF_NULL, XATTR_NAME_SOM);
3049 if (rc < 0 && rc != -ENODATA)
3052 fl = rc == -ENODATA ? LU_XATTR_CREATE : LU_XATTR_REPLACE;
3053 lustre_som_swab(&mlc->mlc_som);
3054 som_buf->lb_buf = &mlc->mlc_som;
3055 som_buf->lb_len = sizeof(mlc->mlc_som);
3058 rc = mdd_declare_layout_change(env, mdd, obj, mlc, handle);
3062 /* record a changelog for the completion of resync */
3063 rc = mdd_declare_changelog_store(env, mdd, CL_RESYNC, NULL, NULL,
3068 /* RESYNC_DONE has piggybacked size and blocks */
3070 rc = mdd_declare_xattr_set(env, mdd, obj, som_buf,
3071 XATTR_NAME_SOM, fl, handle);
3076 rc = mdd_trans_start(env, mdd, handle);
3080 /* it needs a sync tx to make FLR to work properly */
3081 handle->th_sync = 1;
3083 rc = mdo_layout_change(env, obj, mlc, handle);
3088 rc = mdo_xattr_set(env, obj, som_buf, XATTR_NAME_SOM,
3094 rc = mdd_changelog_data_store(env, mdd, CL_RESYNC, 0, obj, handle,
3104 * Layout change callback for object.
3106 * This is only used by FLR for now. In the future, it can be exteneded to
3107 * handle all layout change.
3110 mdd_layout_change(const struct lu_env *env, struct md_object *o,
3111 struct md_layout_change *mlc)
3113 struct mdd_object *obj = md2mdd_obj(o);
3114 struct mdd_device *mdd = mdd_obj2mdd_dev(obj);
3115 struct lu_buf *buf = mdd_buf_get(env, NULL, 0);
3116 struct lov_comp_md_v1 *lcm;
3117 struct thandle *handle;
3123 if (S_ISDIR(mdd_object_type(obj))) {
3124 switch (mlc->mlc_opc) {
3125 case MD_LAYOUT_SHRINK:
3126 rc = mdd_dir_layout_shrink(env, o, mlc);
3128 case MD_LAYOUT_SPLIT:
3129 rc = mdd_dir_layout_split(env, o, mlc);
3138 /* Verify acceptable operations */
3139 switch (mlc->mlc_opc) {
3140 case MD_LAYOUT_WRITE:
3141 case MD_LAYOUT_RESYNC:
3142 case MD_LAYOUT_RESYNC_DONE:
3148 handle = mdd_trans_create(env, mdd);
3150 RETURN(PTR_ERR(handle));
3152 rc = mdd_stripe_get(env, obj, buf, XATTR_NAME_LOV);
3159 /* analyze the layout to make sure it's a FLR file */
3161 if (le32_to_cpu(lcm->lcm_magic) != LOV_MAGIC_COMP_V1)
3162 GOTO(out, rc = -EINVAL);
3164 flr_state = le16_to_cpu(lcm->lcm_flags) & LCM_FL_FLR_MASK;
3166 /* please refer to HLD of FLR for state transition */
3167 switch (flr_state) {
3169 rc = mdd_layout_instantiate_component(env, obj, mlc, handle);
3171 case LCM_FL_WRITE_PENDING:
3172 rc = mdd_layout_update_write_pending(env, obj, mlc, handle);
3175 rc = mdd_layout_update_rdonly(env, obj, mlc, handle);
3177 case LCM_FL_SYNC_PENDING:
3178 rc = mdd_object_update_sync_pending(env, obj, mlc, handle);
3187 mdd_trans_stop(env, mdd, rc, handle);
3192 void mdd_object_make_hint(const struct lu_env *env, struct mdd_object *parent,
3193 struct mdd_object *child, const struct lu_attr *attr,
3194 const struct md_op_spec *spec,
3195 struct dt_allocation_hint *hint)
3197 struct dt_object *np = parent ? mdd_object_child(parent) : NULL;
3198 struct mdd_device *mdd = mdd_obj2mdd_dev(child);
3199 struct dt_object *nc = mdd_object_child(child);
3201 memset(hint, 0, sizeof(*hint));
3203 /* For striped directory, give striping EA to lod_ah_init, which will
3204 * decide the stripe_offset and stripe count by it. */
3205 if (S_ISDIR(attr->la_mode) &&
3206 unlikely(spec != NULL && spec->sp_cr_flags & MDS_OPEN_HAS_EA)) {
3207 hint->dah_eadata = spec->u.sp_ea.eadata;
3208 hint->dah_eadata_len = spec->u.sp_ea.eadatalen;
3209 } else if (S_ISREG(attr->la_mode) &&
3210 spec->sp_cr_flags & MDS_OPEN_APPEND) {
3211 hint->dah_append_stripe_count = mdd->mdd_append_stripe_count;
3212 hint->dah_append_pool = mdd->mdd_append_pool;
3215 CDEBUG(D_INFO, DFID" eadata %p len %d\n", PFID(mdd_object_fid(child)),
3216 hint->dah_eadata, hint->dah_eadata_len);
3217 /* @hint will be initialized by underlying device. */
3218 nc->do_ops->do_ah_init(env, hint, np, nc, attr->la_mode & S_IFMT);
3221 static int mdd_accmode(const struct lu_env *env, const struct lu_attr *la,
3224 /* Sadly, NFSD reopens a file repeatedly during operation, so the
3225 * "acc_mode = 0" allowance for newly-created files isn't honoured.
3226 * NFSD uses the MDS_OPEN_OWNEROVERRIDE flag to say that a file
3227 * owner can write to a file even if it is marked readonly to hide
3228 * its brokenness. (bug 5781) */
3229 if (open_flags & MDS_OPEN_OWNEROVERRIDE) {
3230 struct lu_ucred *uc = lu_ucred_check(env);
3232 if ((uc == NULL) || (la->la_uid == uc->uc_fsuid))
3236 return mds_accmode(open_flags);
3239 static int mdd_open_sanity_check(const struct lu_env *env,
3240 struct mdd_object *obj,
3241 const struct lu_attr *attr, u64 open_flags,
3244 unsigned int may_mask;
3248 /* EEXIST check, also opening of *open* orphans is allowed so we can
3249 * open-by-handle unlinked files
3251 if (mdd_is_dead_obj(obj) && !is_replay &&
3252 likely(!(mdd_is_orphan_obj(obj) && obj->mod_count > 0)))
3255 if (S_ISLNK(attr->la_mode))
3258 may_mask = mdd_accmode(env, attr, open_flags);
3260 if (S_ISDIR(attr->la_mode) && (may_mask & MAY_WRITE))
3263 if (!(open_flags & MDS_OPEN_CREATED)) {
3264 rc = mdd_permission_internal(env, obj, attr, may_mask);
3269 if (S_ISFIFO(attr->la_mode) || S_ISSOCK(attr->la_mode) ||
3270 S_ISBLK(attr->la_mode) || S_ISCHR(attr->la_mode))
3271 open_flags &= ~MDS_OPEN_TRUNC;
3273 /* For writing append-only file must open it with append mode. */
3274 if (attr->la_flags & LUSTRE_APPEND_FL) {
3275 if ((open_flags & MDS_FMODE_WRITE) &&
3276 !(open_flags & MDS_OPEN_APPEND))
3278 if (open_flags & MDS_OPEN_TRUNC)
3285 static int mdd_open(const struct lu_env *env, struct md_object *obj,
3286 u64 open_flags, struct md_op_spec *spec)
3288 struct mdd_object *mdd_obj = md2mdd_obj(obj);
3289 struct md_device *md_dev = lu2md_dev(mdd2lu_dev(mdo2mdd(obj)));
3290 struct lu_attr *attr = MDD_ENV_VAR(env, cattr);
3291 struct mdd_object_user *mou = NULL;
3292 const struct lu_ucred *uc = lu_ucred(env);
3293 struct mdd_device *mdd = mdo2mdd(obj);
3294 enum changelog_rec_type type = CL_OPEN;
3298 mdd_write_lock(env, mdd_obj, DT_TGT_CHILD);
3300 rc = mdd_la_get(env, mdd_obj, attr);
3304 rc = mdd_open_sanity_check(env, mdd_obj, attr, open_flags,
3306 if ((rc == -EACCES) && (mdd->mdd_cl.mc_current_mask & BIT(CL_DN_OPEN)))
3311 mdd_obj->mod_count++;
3313 if (!mdd_changelog_enabled(env, mdd, type))
3317 /* look for existing opener in list under mdd_write_lock */
3318 mou = mdd_obj_user_find(mdd_obj, uc->uc_uid, uc->uc_gid, open_flags);
3323 /* add user to list */
3324 mou = mdd_obj_user_alloc(open_flags, uc->uc_uid, uc->uc_gid);
3330 rc2 = mdd_obj_user_add(mdd_obj, mou, type == CL_DN_OPEN);
3332 mdd_obj_user_free(mou);
3337 if (type == CL_DN_OPEN) {
3338 if (ktime_before(ktime_get(), mou->mou_deniednext))
3339 /* same user denied again same access within
3340 * time interval: do not record
3344 /* this user already denied, but some time ago:
3345 * update denied time
3347 mou->mou_deniednext =
3348 ktime_add(ktime_get(),
3349 ktime_set(mdd->mdd_cl.mc_deniednext,
3352 mou->mou_opencount++;
3353 /* same user opening file again with same flags:
3360 /* FYI, only the bottom 32 bits of open_flags are recorded */
3361 mdd_changelog(env, type, open_flags, md_dev, mdd_object_fid(mdd_obj));
3365 mdd_write_unlock(env, mdd_obj);
3369 static int mdd_declare_close(const struct lu_env *env, struct mdd_object *obj,
3370 struct md_attr *ma, struct thandle *handle)
3374 rc = mdd_orphan_declare_delete(env, obj, handle);
3378 return mdo_declare_destroy(env, obj, handle);
3382 * No permission check is needed.
3384 static int mdd_close(const struct lu_env *env, struct md_object *obj,
3385 struct md_attr *ma, u64 open_flags)
3387 struct mdd_object *mdd_obj = md2mdd_obj(obj);
3388 struct mdd_device *mdd = mdo2mdd(obj);
3389 struct thandle *handle = NULL;
3392 bool blocked = false;
3393 bool last_close_by_uid = false;
3394 const struct lu_ucred *uc = lu_ucred(env);
3397 if (ma->ma_valid & MA_FLAGS && ma->ma_attr_flags & MDS_KEEP_ORPHAN) {
3398 mdd_write_lock(env, mdd_obj, DT_TGT_CHILD);
3399 mdd_obj->mod_count--;
3400 mdd_write_unlock(env, mdd_obj);
3402 if (mdd_obj->mod_flags & ORPHAN_OBJ && !mdd_obj->mod_count)
3403 CDEBUG(D_HA, "Object "DFID" is retained in orphan "
3404 "list\n", PFID(mdd_object_fid(mdd_obj)));
3408 /* mdd_finish_unlink() will always set orphan object as DEAD_OBJ, but
3409 * it might fail to add the object to orphan list (w/o ORPHAN_OBJ). */
3410 /* check without any lock */
3411 is_orphan = mdd_obj->mod_count == 1 &&
3412 (mdd_obj->mod_flags & (ORPHAN_OBJ | DEAD_OBJ)) != 0;
3416 /* mdd_trans_create() maybe failed because of barrier_entry(),
3417 * under such case, the orphan MDT-object will be left in the
3418 * orphan list, and when the MDT remount next time, the unused
3419 * orphans will be destroyed automatically.
3421 * One exception: the former mdd_finish_unlink may failed to
3422 * add the orphan MDT-object to the orphan list, then if the
3423 * mdd_trans_create() failed because of barrier_entry(), the
3424 * MDT-object will become real orphan that is neither in the
3425 * namespace nor in the orphan list. Such bad case should be
3426 * very rare and will be handled by e2fsck/lfsck. */
3427 handle = mdd_trans_create(env, mdo2mdd(obj));
3428 if (IS_ERR(handle)) {
3429 rc = PTR_ERR(handle);
3430 if (rc != -EINPROGRESS)
3438 rc = mdd_declare_close(env, mdd_obj, ma, handle);
3442 rc = mdd_declare_changelog_store(env, mdd, CL_CLOSE, NULL, NULL,
3447 rc = mdd_trans_start(env, mdo2mdd(obj), handle);
3453 mdd_write_lock(env, mdd_obj, DT_TGT_CHILD);
3454 rc = mdd_la_get(env, mdd_obj, &ma->ma_attr);
3456 CERROR("%s: failed to get lu_attr of "DFID": rc = %d\n",
3457 lu_dev_name(mdd2lu_dev(mdd)),
3458 PFID(mdd_object_fid(mdd_obj)), rc);
3462 /* check again with lock */
3463 is_orphan = (mdd_obj->mod_count == 1) &&
3464 ((mdd_obj->mod_flags & (ORPHAN_OBJ | DEAD_OBJ)) != 0 ||
3465 ma->ma_attr.la_nlink == 0);
3467 if (is_orphan && !handle && !blocked) {
3468 mdd_write_unlock(env, mdd_obj);
3472 mdd_obj->mod_count--; /*release open count */
3474 /* under mdd write lock */
3475 /* If recording, see if we need to remove UID from list. uc is not
3476 * initialized if the client has been evicted. */
3477 if (mdd_changelog_enabled(env, mdd, CL_OPEN) && uc) {
3478 struct mdd_object_user *mou;
3480 /* look for UID in list */
3481 /* If mou is NULL, it probably means logging was enabled after
3482 * the user had the file open. So the corresponding close
3483 * will not be logged.
3485 mou = mdd_obj_user_find(mdd_obj, uc->uc_uid, uc->uc_gid,
3488 mou->mou_opencount--;
3489 if (mou->mou_opencount == 0) {
3490 mdd_obj_user_remove(mdd_obj, mou);
3491 last_close_by_uid = true;
3496 if (!is_orphan || blocked)
3500 /* NB: Object maybe not in orphan list originally, it is rare case for
3501 * mdd_finish_unlink() failure, in that case, the object doesn't have
3502 * ORPHAN_OBJ flag */
3503 if ((mdd_obj->mod_flags & ORPHAN_OBJ) != 0) {
3504 /* remove link to object from orphan index */
3505 LASSERT(handle != NULL);
3506 rc = mdd_orphan_delete(env, mdd_obj, handle);
3508 CERROR("%s: unable to delete "DFID" from orphan list: "
3509 "rc = %d\n", lu_dev_name(mdd2lu_dev(mdd)),
3510 PFID(mdd_object_fid(mdd_obj)), rc);
3511 /* If object was not deleted from orphan list, do not
3512 * destroy OSS objects, which will be done when next
3517 CDEBUG(D_HA, "Object "DFID" is deleted from orphan "
3518 "list, OSS objects to be destroyed.\n",
3519 PFID(mdd_object_fid(mdd_obj)));
3522 rc = mdo_destroy(env, mdd_obj, handle);
3525 CERROR("%s: unable to delete "DFID" from orphan list: "
3526 "rc = %d\n", lu_dev_name(mdd2lu_dev(mdd)),
3527 PFID(mdd_object_fid(mdd_obj)), rc);
3532 mdd_write_unlock(env, mdd_obj);
3534 if (rc != 0 || blocked ||
3535 !mdd_changelog_enabled(env, mdd, CL_CLOSE))
3538 /* Record CL_CLOSE in changelog only if file was opened in write mode,
3539 * or if CL_OPEN was recorded and it's last close by user.
3540 * Changelogs mask may change between open and close operations, but
3541 * this is not a big deal if we have a CL_CLOSE entry with no matching
3542 * CL_OPEN. Plus Changelogs mask may not change often.
3544 if (((!(mdd->mdd_cl.mc_current_mask & BIT(CL_OPEN)) &&
3545 (open_flags & (MDS_FMODE_WRITE | MDS_OPEN_APPEND |
3546 MDS_OPEN_TRUNC))) ||
3547 ((mdd->mdd_cl.mc_current_mask & BIT(CL_OPEN)) &&
3548 last_close_by_uid)) &&
3549 !(ma->ma_valid & MA_FLAGS && ma->ma_attr_flags & MDS_RECOV_OPEN)) {
3550 if (handle == NULL) {
3551 handle = mdd_trans_create(env, mdo2mdd(obj));
3553 GOTO(stop, rc = PTR_ERR(handle));
3555 rc = mdd_declare_changelog_store(env, mdd, CL_CLOSE,
3556 NULL, NULL, handle);
3560 rc = mdd_trans_start(env, mdo2mdd(obj), handle);
3565 /* FYI, only the bottom 32 bits of open_flags are recorded */
3566 mdd_changelog_data_store(env, mdd, CL_CLOSE, open_flags,
3567 mdd_obj, handle, NULL);
3571 if (handle != NULL && !IS_ERR(handle))
3572 rc = mdd_trans_stop(env, mdd, rc, handle);
3578 * Permission check is done when open,
3579 * no need check again.
3581 static int mdd_readpage_sanity_check(const struct lu_env *env,
3582 struct mdd_object *obj)
3584 if (!dt_try_as_dir(env, mdd_object_child(obj), true))
3590 static int mdd_dir_page_build(const struct lu_env *env, struct dt_object *obj,
3591 union lu_page *lp, size_t bytes,
3592 const struct dt_it_ops *iops,
3593 struct dt_it *it, __u32 attr, void *arg)
3595 struct lu_dirpage *dp = &lp->lp_dir;
3598 struct lu_dirent *ent;
3599 struct lu_dirent *last = NULL;
3604 if (bytes < sizeof(*dp))
3605 GOTO(out_err, result = -EOVERFLOW);
3607 memset(area, 0, sizeof(*dp));
3608 area += sizeof(*dp);
3609 bytes -= sizeof(*dp);
3616 len = iops->key_size(env, it);
3618 /* IAM iterator can return record with zero len. */
3622 hash = iops->store(env, it);
3623 if (unlikely(first)) {
3625 dp->ldp_hash_start = cpu_to_le64(hash);
3628 /* calculate max space required for lu_dirent */
3629 recsize = lu_dirent_calc_size(len, attr);
3631 if (bytes >= recsize &&
3632 !OBD_FAIL_CHECK(OBD_FAIL_MDS_DIR_PAGE_WALK)) {
3633 result = iops->rec(env, it, (struct dt_rec *)ent, attr);
3634 if (result == -ESTALE)
3639 /* OSD might not able to pack all attributes, so
3640 * recheck record length had room to store FID
3642 recsize = le16_to_cpu(ent->lde_reclen);
3644 if (le32_to_cpu(ent->lde_attrs) & LUDA_FID) {
3645 fid_le_to_cpu(&fid, &ent->lde_fid);
3646 if (fid_is_dot_lustre(&fid))
3647 GOTO(next, recsize);
3650 result = (last != NULL) ? 0 : -EBADSLT;
3654 ent = (void *)ent + recsize;
3658 result = iops->next(env, it);
3659 if (result == -ESTALE)
3661 } while (result == 0);
3664 dp->ldp_hash_end = cpu_to_le64(hash);
3666 if (last->lde_hash == dp->ldp_hash_end)
3667 dp->ldp_flags |= cpu_to_le32(LDF_COLLIDE);
3668 last->lde_reclen = 0; /* end mark */
3672 /* end of directory */
3673 dp->ldp_hash_end = cpu_to_le64(MDS_DIR_END_OFF);
3674 else if (result < 0)
3675 CWARN("%s: build page failed for "DFID": rc = %d\n",
3676 lu_dev_name(obj->do_lu.lo_dev),
3677 PFID(lu_object_fid(&obj->do_lu)), result);
3682 int mdd_readpage(const struct lu_env *env, struct md_object *obj,
3683 const struct lu_rdpg *rdpg)
3685 struct mdd_object *mdd_obj = md2mdd_obj(obj);
3689 if (mdd_object_exists(mdd_obj) == 0) {
3690 CERROR("%s: object "DFID" not found: rc = -2\n",
3691 mdd_obj_dev_name(mdd_obj),PFID(mdd_object_fid(mdd_obj)));
3695 mdd_read_lock(env, mdd_obj, DT_TGT_CHILD);
3696 rc = mdd_readpage_sanity_check(env, mdd_obj);
3698 GOTO(out_unlock, rc);
3700 if (mdd_is_dead_obj(mdd_obj)) {
3702 struct lu_dirpage *dp;
3705 * According to POSIX, please do not return any entry to client:
3706 * even dot and dotdot should not be returned.
3708 CDEBUG(D_INODE, "readdir from dead object: "DFID"\n",
3709 PFID(mdd_object_fid(mdd_obj)));
3711 if (rdpg->rp_count <= 0)
3712 GOTO(out_unlock, rc = -EFAULT);
3713 LASSERT(rdpg->rp_pages != NULL);
3715 pg = rdpg->rp_pages[0];
3716 dp = (struct lu_dirpage *)kmap(pg);
3717 memset(dp, 0 , sizeof(struct lu_dirpage));
3718 dp->ldp_hash_start = cpu_to_le64(rdpg->rp_hash);
3719 dp->ldp_hash_end = cpu_to_le64(MDS_DIR_END_OFF);
3720 dp->ldp_flags = cpu_to_le32(LDF_EMPTY);
3722 GOTO(out_unlock, rc = LU_PAGE_SIZE);
3725 rc = dt_index_walk(env, mdd_object_child(mdd_obj), rdpg,
3726 mdd_dir_page_build, NULL);
3728 struct lu_dirpage *dp;
3730 dp = kmap(rdpg->rp_pages[0]);
3731 dp->ldp_hash_start = cpu_to_le64(rdpg->rp_hash);
3734 * No pages were processed, mark this for first page
3737 dp->ldp_hash_end = cpu_to_le64(MDS_DIR_END_OFF);
3738 dp->ldp_flags = cpu_to_le32(LDF_EMPTY);
3739 rc = min_t(unsigned int, LU_PAGE_SIZE, rdpg->rp_count);
3741 kunmap(rdpg->rp_pages[0]);
3744 GOTO(out_unlock, rc);
3746 mdd_read_unlock(env, mdd_obj);
3750 static int mdd_object_sync(const struct lu_env *env, struct md_object *obj)
3752 struct mdd_object *mdd_obj = md2mdd_obj(obj);
3754 if (mdd_object_exists(mdd_obj) == 0) {
3757 CERROR("%s: object "DFID" not found: rc = %d\n",
3758 mdd_obj_dev_name(mdd_obj),
3759 PFID(mdd_object_fid(mdd_obj)), rc);
3762 return dt_object_sync(env, mdd_object_child(mdd_obj),
3766 static int mdd_object_lock(const struct lu_env *env,
3767 struct md_object *obj,
3768 struct lustre_handle *lh,
3769 struct ldlm_enqueue_info *einfo,
3770 union ldlm_policy_data *policy)
3772 struct mdd_object *mdd_obj = md2mdd_obj(obj);
3773 return dt_object_lock(env, mdd_object_child(mdd_obj), lh,
3777 static int mdd_object_unlock(const struct lu_env *env,
3778 struct md_object *obj,
3779 struct ldlm_enqueue_info *einfo,
3780 union ldlm_policy_data *policy)
3782 struct mdd_object *mdd_obj = md2mdd_obj(obj);
3783 return dt_object_unlock(env, mdd_object_child(mdd_obj), einfo, policy);
3786 const struct md_object_operations mdd_obj_ops = {
3787 .moo_permission = mdd_permission,
3788 .moo_attr_get = mdd_attr_get,
3789 .moo_attr_set = mdd_attr_set,
3790 .moo_xattr_get = mdd_xattr_get,
3791 .moo_xattr_set = mdd_xattr_set,
3792 .moo_xattr_list = mdd_xattr_list,
3793 .moo_invalidate = mdd_invalidate,
3794 .moo_xattr_del = mdd_xattr_del,
3795 .moo_swap_layouts = mdd_swap_layouts,
3796 .moo_open = mdd_open,
3797 .moo_close = mdd_close,
3798 .moo_readpage = mdd_readpage,
3799 .moo_readlink = mdd_readlink,
3800 .moo_changelog = mdd_changelog,
3801 .moo_object_sync = mdd_object_sync,
3802 .moo_object_lock = mdd_object_lock,
3803 .moo_object_unlock = mdd_object_unlock,
3804 .moo_layout_change = mdd_layout_change,