4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/mdt/mdt_reint.c
34 * Lustre Metadata Target (mdt) reintegration routines
36 * Author: Peter Braam <braam@clusterfs.com>
37 * Author: Andreas Dilger <adilger@clusterfs.com>
38 * Author: Phil Schwan <phil@clusterfs.com>
39 * Author: Huang Hua <huanghua@clusterfs.com>
40 * Author: Yury Umanets <umka@clusterfs.com>
43 #define DEBUG_SUBSYSTEM S_MDS
45 #include <lprocfs_status.h>
46 #include "mdt_internal.h"
47 #include <lustre_lmv.h>
49 static inline void mdt_reint_init_ma(struct mdt_thread_info *info,
52 ma->ma_need = MA_INODE;
57 * Get version of object by fid.
59 * Return real version or ENOENT_VERSION if object doesn't exist
61 static void mdt_obj_version_get(struct mdt_thread_info *info,
62 struct mdt_object *o, __u64 *version)
66 if (mdt_object_exists(o) && !mdt_object_remote(o) &&
67 !fid_is_obf(mdt_object_fid(o)))
68 *version = dt_version_get(info->mti_env, mdt_obj2dt(o));
70 *version = ENOENT_VERSION;
71 CDEBUG(D_INODE, "FID "DFID" version is %#llx\n",
72 PFID(mdt_object_fid(o)), *version);
76 * Check version is correct.
78 * Should be called only during replay.
80 static int mdt_version_check(struct ptlrpc_request *req,
81 __u64 version, int idx)
83 __u64 *pre_ver = lustre_msg_get_versions(req->rq_reqmsg);
86 if (!exp_connect_vbr(req->rq_export))
89 LASSERT(req_is_replay(req));
90 /** VBR: version is checked always because costs nothing */
91 LASSERT(idx < PTLRPC_NUM_VERSIONS);
92 /** Sanity check for malformed buffers */
93 if (pre_ver == NULL) {
94 CERROR("No versions in request buffer\n");
95 spin_lock(&req->rq_export->exp_lock);
96 req->rq_export->exp_vbr_failed = 1;
97 spin_unlock(&req->rq_export->exp_lock);
99 } else if (pre_ver[idx] != version) {
100 CDEBUG(D_INODE, "Version mismatch %#llx != %#llx\n",
101 pre_ver[idx], version);
102 spin_lock(&req->rq_export->exp_lock);
103 req->rq_export->exp_vbr_failed = 1;
104 spin_unlock(&req->rq_export->exp_lock);
111 * Save pre-versions in reply.
113 static void mdt_version_save(struct ptlrpc_request *req, __u64 version,
118 if (!exp_connect_vbr(req->rq_export))
121 LASSERT(!req_is_replay(req));
122 LASSERT(req->rq_repmsg != NULL);
123 reply_ver = lustre_msg_get_versions(req->rq_repmsg);
125 reply_ver[idx] = version;
129 * Save enoent version, it is needed when it is obvious that object doesn't
130 * exist, e.g. child during create.
132 static void mdt_enoent_version_save(struct mdt_thread_info *info, int idx)
134 /* save version of file name for replay, it must be ENOENT here */
135 if (!req_is_replay(mdt_info_req(info))) {
136 info->mti_ver[idx] = ENOENT_VERSION;
137 mdt_version_save(mdt_info_req(info), info->mti_ver[idx], idx);
142 * Get version from disk and save in reply buffer.
144 * Versions are saved in reply only during normal operations not replays.
146 void mdt_version_get_save(struct mdt_thread_info *info,
147 struct mdt_object *mto, int idx)
149 /* don't save versions during replay */
150 if (!req_is_replay(mdt_info_req(info))) {
151 mdt_obj_version_get(info, mto, &info->mti_ver[idx]);
152 mdt_version_save(mdt_info_req(info), info->mti_ver[idx], idx);
157 * Get version from disk and check it, no save in reply.
159 int mdt_version_get_check(struct mdt_thread_info *info,
160 struct mdt_object *mto, int idx)
162 /* only check versions during replay */
163 if (!req_is_replay(mdt_info_req(info)))
166 mdt_obj_version_get(info, mto, &info->mti_ver[idx]);
167 return mdt_version_check(mdt_info_req(info), info->mti_ver[idx], idx);
171 * Get version from disk and check if recovery or just save.
173 int mdt_version_get_check_save(struct mdt_thread_info *info,
174 struct mdt_object *mto, int idx)
178 mdt_obj_version_get(info, mto, &info->mti_ver[idx]);
179 if (req_is_replay(mdt_info_req(info)))
180 rc = mdt_version_check(mdt_info_req(info), info->mti_ver[idx],
183 mdt_version_save(mdt_info_req(info), info->mti_ver[idx], idx);
188 * Lookup with version checking.
190 * This checks version of 'name'. Many reint functions uses 'name' for child not
191 * FID, therefore we need to get object by name and check its version.
193 int mdt_lookup_version_check(struct mdt_thread_info *info,
194 struct mdt_object *p,
195 const struct lu_name *lname,
196 struct lu_fid *fid, int idx)
200 rc = mdo_lookup(info->mti_env, mdt_object_child(p), lname, fid,
202 /* Check version only during replay */
203 if (!req_is_replay(mdt_info_req(info)))
206 info->mti_ver[idx] = ENOENT_VERSION;
208 struct mdt_object *child;
209 child = mdt_object_find(info->mti_env, info->mti_mdt, fid);
210 if (likely(!IS_ERR(child))) {
211 mdt_obj_version_get(info, child, &info->mti_ver[idx]);
212 mdt_object_put(info->mti_env, child);
215 vbrc = mdt_version_check(mdt_info_req(info), info->mti_ver[idx], idx);
216 return vbrc ? vbrc : rc;
220 static int mdt_unlock_slaves(struct mdt_thread_info *mti,
221 struct mdt_object *obj,
222 struct ldlm_enqueue_info *einfo,
225 union ldlm_policy_data *policy = &mti->mti_policy;
226 struct mdt_lock_handle *lh = &mti->mti_lh[MDT_LH_LOCAL];
227 struct lustre_handle_array *slave_locks = einfo->ei_cbdata;
230 LASSERT(S_ISDIR(obj->mot_header.loh_attr));
231 LASSERT(slave_locks);
233 memset(policy, 0, sizeof(*policy));
234 policy->l_inodebits.bits = einfo->ei_inodebits;
235 mdt_lock_handle_init(lh);
236 mdt_lock_reg_init(lh, einfo->ei_mode);
237 for (i = 0; i < slave_locks->ha_count; i++) {
238 if (test_bit(i, (void *)slave_locks->ha_map))
239 lh->mlh_rreg_lh = slave_locks->ha_handles[i];
241 lh->mlh_reg_lh = slave_locks->ha_handles[i];
242 mdt_object_unlock(mti, NULL, lh, decref);
243 slave_locks->ha_handles[i].cookie = 0ull;
246 return mo_object_unlock(mti->mti_env, mdt_object_child(obj), einfo,
250 static inline int mdt_object_striped(struct mdt_thread_info *mti,
251 struct mdt_object *obj)
253 struct lu_device *bottom_dev;
254 struct lu_object *bottom_obj;
257 if (!S_ISDIR(obj->mot_header.loh_attr))
260 /* getxattr from bottom obj to avoid reading in shard FIDs */
261 bottom_dev = dt2lu_dev(mti->mti_mdt->mdt_bottom);
262 bottom_obj = lu_object_find_slice(mti->mti_env, bottom_dev,
263 mdt_object_fid(obj), NULL);
264 if (IS_ERR(bottom_obj))
265 return PTR_ERR(bottom_obj);
267 rc = dt_xattr_get(mti->mti_env, lu2dt(bottom_obj), &LU_BUF_NULL,
269 lu_object_put(mti->mti_env, bottom_obj);
271 return (rc > 0) ? 1 : (rc == -ENODATA) ? 0 : rc;
275 * Lock slave stripes if necessary, the lock handles of slave stripes
276 * will be stored in einfo->ei_cbdata.
278 static int mdt_lock_slaves(struct mdt_thread_info *mti, struct mdt_object *obj,
279 enum ldlm_mode mode, __u64 ibits,
280 struct ldlm_enqueue_info *einfo)
282 union ldlm_policy_data *policy = &mti->mti_policy;
284 LASSERT(S_ISDIR(obj->mot_header.loh_attr));
286 einfo->ei_type = LDLM_IBITS;
287 einfo->ei_mode = mode;
288 einfo->ei_cb_bl = mdt_remote_blocking_ast;
289 einfo->ei_cb_local_bl = mdt_blocking_ast;
290 einfo->ei_cb_cp = ldlm_completion_ast;
291 einfo->ei_enq_slave = 1;
292 einfo->ei_namespace = mti->mti_mdt->mdt_namespace;
293 einfo->ei_inodebits = ibits;
294 memset(policy, 0, sizeof(*policy));
295 policy->l_inodebits.bits = ibits;
297 return mo_object_lock(mti->mti_env, mdt_object_child(obj), NULL, einfo,
301 int mdt_reint_striped_lock(struct mdt_thread_info *info,
302 struct mdt_object *o,
303 struct mdt_lock_handle *lh,
305 struct ldlm_enqueue_info *einfo,
310 LASSERT(!mdt_object_remote(o));
312 memset(einfo, 0, sizeof(*einfo));
314 rc = mdt_reint_object_lock(info, o, lh, ibits, cos_incompat);
318 rc = mdt_object_striped(info, o);
321 mdt_object_unlock(info, o, lh, rc);
325 rc = mdt_lock_slaves(info, o, lh->mlh_reg_mode, ibits, einfo);
327 mdt_object_unlock(info, o, lh, rc);
328 if (rc == -EIO && OBD_FAIL_CHECK(OBD_FAIL_LFSCK_BAD_SLAVE_NAME))
335 void mdt_reint_striped_unlock(struct mdt_thread_info *info,
336 struct mdt_object *o,
337 struct mdt_lock_handle *lh,
338 struct ldlm_enqueue_info *einfo, int decref)
340 if (einfo->ei_cbdata)
341 mdt_unlock_slaves(info, o, einfo, decref);
342 mdt_object_unlock(info, o, lh, decref);
345 static int mdt_restripe(struct mdt_thread_info *info,
346 struct mdt_object *parent,
347 const struct lu_name *lname,
348 const struct lu_fid *tfid,
349 struct md_op_spec *spec,
352 struct mdt_device *mdt = info->mti_mdt;
353 struct lu_fid *fid = &info->mti_tmp_fid2;
354 struct ldlm_enqueue_info *einfo = &info->mti_einfo[0];
355 struct lmv_mds_md_v1 *lmv;
356 struct mdt_object *child;
357 struct mdt_lock_handle *lhp;
358 struct mdt_lock_handle *lhc;
359 struct mdt_body *repbody;
364 if (!mdt->mdt_enable_dir_restripe)
367 rc = mdt_version_get_check_save(info, parent, 0);
371 lhp = &info->mti_lh[MDT_LH_PARENT];
372 mdt_lock_pdo_init(lhp, LCK_PW, lname);
373 rc = mdt_reint_object_lock(info, parent, lhp, MDS_INODELOCK_UPDATE,
378 rc = mdt_stripe_get(info, parent, ma, XATTR_NAME_LMV);
380 GOTO(unlock_parent, rc);
382 if (ma->ma_valid & MA_LMV) {
383 /* don't allow restripe if parent dir layout is changing */
384 lmv = &ma->ma_lmv->lmv_md_v1;
385 if (!lmv_is_sane(lmv))
386 GOTO(unlock_parent, rc = -EBADF);
388 if (lmv_is_layout_changing(lmv))
389 GOTO(unlock_parent, rc = -EBUSY);
393 rc = mdt_lookup_version_check(info, parent, lname, fid, 1);
395 GOTO(unlock_parent, rc);
397 child = mdt_object_find(info->mti_env, mdt, fid);
399 GOTO(unlock_parent, rc = PTR_ERR(child));
401 if (!mdt_object_exists(child))
402 GOTO(out_child, rc = -ENOENT);
404 if (mdt_object_remote(child)) {
405 struct mdt_body *repbody;
407 repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY);
409 GOTO(out_child, rc = -EPROTO);
411 repbody->mbo_fid1 = *fid;
412 repbody->mbo_valid |= (OBD_MD_FLID | OBD_MD_MDS);
413 GOTO(out_child, rc = -EREMOTE);
417 lhc = &info->mti_lh[MDT_LH_CHILD];
418 mdt_lock_reg_init(lhc, LCK_EX);
420 /* enqueue object remote LOOKUP lock */
421 if (mdt_object_remote(parent)) {
422 rc = mdt_remote_object_lock(info, parent, fid,
425 MDS_INODELOCK_LOOKUP, false);
430 rc = mdt_reint_striped_lock(info, child, lhc, MDS_INODELOCK_FULL, einfo,
433 GOTO(unlock_child, rc);
435 tgt_vbr_obj_set(info->mti_env, mdt_obj2dt(child));
436 rc = mdt_version_get_check_save(info, child, 1);
438 GOTO(unlock_child, rc);
440 spin_lock(&mdt->mdt_restriper.mdr_lock);
441 if (child->mot_restriping) {
443 spin_unlock(&mdt->mdt_restriper.mdr_lock);
444 GOTO(unlock_child, rc = -EBUSY);
446 child->mot_restriping = 1;
447 spin_unlock(&mdt->mdt_restriper.mdr_lock);
450 rc = mdt_restripe_internal(info, parent, child, lname, fid, spec, ma);
452 GOTO(restriping_clear, rc);
454 repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY);
456 GOTO(restriping_clear, rc = -EPROTO);
458 mdt_pack_attr2body(info, repbody, &ma->ma_attr, fid);
462 child->mot_restriping = 0;
464 mdt_reint_striped_unlock(info, child, lhc, einfo, rc);
466 mdt_object_put(info->mti_env, child);
468 mdt_object_unlock(info, parent, lhp, rc);
474 * VBR: we save three versions in reply:
475 * 0 - parent. Check that parent version is the same during replay.
476 * 1 - name. Version of 'name' if file exists with the same name or
477 * ENOENT_VERSION, it is needed because file may appear due to missed replays.
478 * 2 - child. Version of child by FID. Must be ENOENT. It is mostly sanity
481 static int mdt_create(struct mdt_thread_info *info)
483 struct mdt_device *mdt = info->mti_mdt;
484 struct mdt_object *parent;
485 struct mdt_object *child;
486 struct mdt_lock_handle *lh;
487 struct mdt_body *repbody;
488 struct md_attr *ma = &info->mti_attr;
489 struct mdt_reint_record *rr = &info->mti_rr;
490 struct md_op_spec *spec = &info->mti_spec;
491 bool restripe = false;
495 DEBUG_REQ(D_INODE, mdt_info_req(info),
496 "Create ("DNAME"->"DFID") in "DFID,
497 PNAME(&rr->rr_name), PFID(rr->rr_fid2), PFID(rr->rr_fid1));
499 if (!fid_is_md_operative(rr->rr_fid1))
502 if (S_ISDIR(ma->ma_attr.la_mode) &&
503 spec->u.sp_ea.eadata != NULL && spec->u.sp_ea.eadatalen != 0) {
504 const struct lmv_user_md *lum = spec->u.sp_ea.eadata;
505 struct lu_ucred *uc = mdt_ucred(info);
506 struct obd_export *exp = mdt_info_req(info)->rq_export;
508 /* Only new clients can create remote dir( >= 2.4) and
509 * striped dir(>= 2.6), old client will return -ENOTSUPP */
510 if (!mdt_is_dne_client(exp))
513 if (le32_to_cpu(lum->lum_stripe_count) > 1) {
514 if (!mdt_is_striped_client(exp))
517 if (!mdt->mdt_enable_striped_dir)
519 } else if (!mdt->mdt_enable_remote_dir) {
523 if ((!(exp_connect_flags2(exp) & OBD_CONNECT2_CRUSH)) &&
524 (le32_to_cpu(lum->lum_hash_type) & LMV_HASH_TYPE_MASK) ==
528 if (!md_capable(uc, CFS_CAP_SYS_ADMIN) &&
529 uc->uc_gid != mdt->mdt_enable_remote_dir_gid &&
530 mdt->mdt_enable_remote_dir_gid != -1)
533 /* restripe if later found dir exists */
534 if (le32_to_cpu(lum->lum_stripe_offset) == LMV_OFFSET_DEFAULT)
538 repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY);
540 parent = mdt_object_find(info->mti_env, info->mti_mdt, rr->rr_fid1);
542 RETURN(PTR_ERR(parent));
544 if (!mdt_object_exists(parent))
545 GOTO(put_parent, rc = -ENOENT);
548 * LU-10235: check if name exists locklessly first to avoid massive
549 * lock recalls on existing directories.
551 rc = mdt_lookup_version_check(info, parent, &rr->rr_name,
552 &info->mti_tmp_fid1, 1);
555 GOTO(put_parent, rc = -EEXIST);
557 rc = mdt_restripe(info, parent, &rr->rr_name, rr->rr_fid2, spec,
561 /* -ENOENT is expected here */
563 GOTO(put_parent, rc);
565 /* save version of file name for replay, it must be ENOENT here */
566 mdt_enoent_version_save(info, 1);
568 OBD_RACE(OBD_FAIL_MDS_CREATE_RACE);
570 lh = &info->mti_lh[MDT_LH_PARENT];
571 mdt_lock_pdo_init(lh, LCK_PW, &rr->rr_name);
572 rc = mdt_object_lock(info, parent, lh, MDS_INODELOCK_UPDATE);
574 GOTO(put_parent, rc);
576 if (!mdt_object_remote(parent)) {
577 rc = mdt_version_get_check_save(info, parent, 0);
579 GOTO(unlock_parent, rc);
582 child = mdt_object_new(info->mti_env, mdt, rr->rr_fid2);
583 if (unlikely(IS_ERR(child)))
584 GOTO(unlock_parent, rc = PTR_ERR(child));
586 ma->ma_need = MA_INODE;
589 mdt_fail_write(info->mti_env, info->mti_mdt->mdt_bottom,
590 OBD_FAIL_MDS_REINT_CREATE_WRITE);
592 /* Version of child will be updated on disk. */
593 tgt_vbr_obj_set(info->mti_env, mdt_obj2dt(child));
594 rc = mdt_version_get_check_save(info, child, 2);
598 /* Let lower layer know current lock mode. */
599 info->mti_spec.sp_cr_mode = mdt_dlm_mode2mdl_mode(lh->mlh_pdo_mode);
602 * Do not perform lookup sanity check. We know that name does
605 info->mti_spec.sp_cr_lookup = 0;
606 info->mti_spec.sp_feat = &dt_directory_features;
608 rc = mdo_create(info->mti_env, mdt_object_child(parent), &rr->rr_name,
609 mdt_object_child(child), &info->mti_spec, ma);
611 rc = mdt_attr_get_complex(info, child, ma);
617 * On DNE, we need to eliminate dependey between 'mkdir a' and
618 * 'mkdir a/b' if b is a striped directory, to achieve this, two
619 * things are done below:
620 * 1. save child and slaves lock.
621 * 2. if the child is a striped directory, relock parent so to
622 * compare against with COS locks to ensure parent was
625 if (mdt_slc_is_enabled(mdt) && S_ISDIR(ma->ma_attr.la_mode)) {
626 struct mdt_lock_handle *lhc;
627 struct ldlm_enqueue_info *einfo = &info->mti_einfo[0];
630 rc = mdt_object_striped(info, child);
636 if (!mdt_object_remote(parent)) {
637 mdt_object_unlock(info, parent, lh, 1);
638 mdt_lock_pdo_init(lh, LCK_PW, &rr->rr_name);
639 rc = mdt_reint_object_lock(info, parent, lh,
640 MDS_INODELOCK_UPDATE,
647 lhc = &info->mti_lh[MDT_LH_CHILD];
648 mdt_lock_handle_init(lhc);
649 mdt_lock_reg_init(lhc, LCK_PW);
650 rc = mdt_reint_striped_lock(info, child, lhc,
651 MDS_INODELOCK_UPDATE, einfo,
656 mdt_reint_striped_unlock(info, child, lhc, einfo, rc);
659 /* Return fid & attr to client. */
660 if (ma->ma_valid & MA_INODE)
661 mdt_pack_attr2body(info, repbody, &ma->ma_attr,
662 mdt_object_fid(child));
665 mdt_object_put(info->mti_env, child);
667 mdt_object_unlock(info, parent, lh, rc);
669 mdt_object_put(info->mti_env, parent);
673 static int mdt_attr_set(struct mdt_thread_info *info, struct mdt_object *mo,
676 struct mdt_lock_handle *lh;
677 int do_vbr = ma->ma_attr.la_valid &
678 (LA_MODE | LA_UID | LA_GID | LA_PROJID | LA_FLAGS);
679 __u64 lockpart = MDS_INODELOCK_UPDATE;
680 struct ldlm_enqueue_info *einfo = &info->mti_einfo[0];
685 rc = mdt_object_striped(info, mo);
691 lh = &info->mti_lh[MDT_LH_PARENT];
692 mdt_lock_reg_init(lh, LCK_PW);
694 /* Even though the new MDT will grant PERM lock to the old
695 * client, but the old client will almost ignore that during
696 * So it needs to revoke both LOOKUP and PERM lock here, so
697 * both new and old client can cancel the dcache */
698 if (ma->ma_attr.la_valid & (LA_MODE|LA_UID|LA_GID))
699 lockpart |= MDS_INODELOCK_LOOKUP | MDS_INODELOCK_PERM;
701 rc = mdt_reint_striped_lock(info, mo, lh, lockpart, einfo,
706 /* all attrs are packed into mti_attr in unpack_setattr */
707 mdt_fail_write(info->mti_env, info->mti_mdt->mdt_bottom,
708 OBD_FAIL_MDS_REINT_SETATTR_WRITE);
710 /* VBR: update version if attr changed are important for recovery */
712 /* update on-disk version of changed object */
713 tgt_vbr_obj_set(info->mti_env, mdt_obj2dt(mo));
714 rc = mdt_version_get_check_save(info, mo, 0);
716 GOTO(out_unlock, rc);
719 /* Ensure constant striping during chown(). See LU-2789. */
720 if (ma->ma_attr.la_valid & (LA_UID|LA_GID|LA_PROJID))
721 mutex_lock(&mo->mot_lov_mutex);
723 /* all attrs are packed into mti_attr in unpack_setattr */
724 rc = mo_attr_set(info->mti_env, mdt_object_child(mo), ma);
726 if (ma->ma_attr.la_valid & (LA_UID|LA_GID|LA_PROJID))
727 mutex_unlock(&mo->mot_lov_mutex);
730 GOTO(out_unlock, rc);
731 mdt_dom_obj_lvb_update(info->mti_env, mo, false);
734 mdt_reint_striped_unlock(info, mo, lh, einfo, rc);
739 * Check HSM flags and add HS_DIRTY flag if relevant.
741 * A file could be set dirty only if it has a copy in the backend (HS_EXISTS)
742 * and is not RELEASED.
744 int mdt_add_dirty_flag(struct mdt_thread_info *info, struct mdt_object *mo,
747 struct lu_ucred *uc = mdt_ucred(info);
752 /* If the file was modified, add the dirty flag */
753 ma->ma_need = MA_HSM;
754 rc = mdt_attr_get_complex(info, mo, ma);
756 CERROR("file attribute read error for "DFID": %d.\n",
757 PFID(mdt_object_fid(mo)), rc);
761 /* If an up2date copy exists in the backend, add dirty flag */
762 if ((ma->ma_valid & MA_HSM) && (ma->ma_hsm.mh_flags & HS_EXISTS)
763 && !(ma->ma_hsm.mh_flags & (HS_DIRTY|HS_RELEASED))) {
764 ma->ma_hsm.mh_flags |= HS_DIRTY;
766 /* Bump cap so that closes from non-owner writers can
767 * set the HSM state to dirty. */
768 cap_saved = uc->uc_cap;
769 uc->uc_cap |= MD_CAP_TO_MASK(CFS_CAP_FOWNER);
770 rc = mdt_hsm_attr_set(info, mo, &ma->ma_hsm);
771 uc->uc_cap = cap_saved;
773 CERROR("file attribute change error for "DFID": %d\n",
774 PFID(mdt_object_fid(mo)), rc);
780 static int mdt_reint_setattr(struct mdt_thread_info *info,
781 struct mdt_lock_handle *lhc)
783 struct mdt_device *mdt = info->mti_mdt;
784 struct md_attr *ma = &info->mti_attr;
785 struct mdt_reint_record *rr = &info->mti_rr;
786 struct ptlrpc_request *req = mdt_info_req(info);
787 struct mdt_object *mo;
788 struct mdt_body *repbody;
792 DEBUG_REQ(D_INODE, req, "setattr "DFID" %x", PFID(rr->rr_fid1),
793 (unsigned int)ma->ma_attr.la_valid);
795 if (info->mti_dlm_req)
796 ldlm_request_cancel(req, info->mti_dlm_req, 0, LATF_SKIP);
798 OBD_RACE(OBD_FAIL_PTLRPC_RESEND_RACE);
800 repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY);
801 mo = mdt_object_find(info->mti_env, mdt, rr->rr_fid1);
803 GOTO(out, rc = PTR_ERR(mo));
805 if (!mdt_object_exists(mo))
806 GOTO(out_put, rc = -ENOENT);
808 if (mdt_object_remote(mo))
809 GOTO(out_put, rc = -EREMOTE);
811 ma->ma_enable_chprojid_gid = mdt->mdt_enable_chprojid_gid;
812 /* revoke lease lock if size is going to be changed */
813 if (unlikely(ma->ma_attr.la_valid & LA_SIZE &&
814 !(ma->ma_attr_flags & MDS_TRUNC_KEEP_LEASE) &&
815 atomic_read(&mo->mot_lease_count) > 0)) {
816 down_read(&mo->mot_open_sem);
818 if (atomic_read(&mo->mot_lease_count) > 0) { /* lease exists */
819 lhc = &info->mti_lh[MDT_LH_LOCAL];
820 mdt_lock_reg_init(lhc, LCK_CW);
822 rc = mdt_object_lock(info, mo, lhc, MDS_INODELOCK_OPEN);
824 up_read(&mo->mot_open_sem);
828 /* revoke lease lock */
829 mdt_object_unlock(info, mo, lhc, 1);
831 up_read(&mo->mot_open_sem);
834 if (ma->ma_attr.la_valid & LA_SIZE || rr->rr_flags & MRF_OPEN_TRUNC) {
835 /* Check write access for the O_TRUNC case */
836 if (mdt_write_read(mo) < 0)
837 GOTO(out_put, rc = -ETXTBSY);
839 /* LU-10286: compatibility check for FLR.
840 * Please check the comment in mdt_finish_open() for details */
841 if (!exp_connect_flr(info->mti_exp) ||
842 !exp_connect_overstriping(info->mti_exp)) {
843 rc = mdt_big_xattr_get(info, mo, XATTR_NAME_LOV);
844 if (rc < 0 && rc != -ENODATA)
847 if (!exp_connect_flr(info->mti_exp)) {
849 mdt_lmm_is_flr(info->mti_big_lmm))
850 GOTO(out_put, rc = -EOPNOTSUPP);
853 if (!exp_connect_overstriping(info->mti_exp)) {
855 mdt_lmm_is_overstriping(info->mti_big_lmm))
856 GOTO(out_put, rc = -EOPNOTSUPP);
860 /* For truncate, the file size sent from client
861 * is believable, but the blocks are incorrect,
862 * which makes the block size in LSOM attribute
863 * inconsisent with the real block size.
865 rc = mdt_lsom_update(info, mo, true);
870 if ((ma->ma_valid & MA_INODE) && ma->ma_attr.la_valid) {
871 if (ma->ma_valid & MA_LOV)
872 GOTO(out_put, rc = -EPROTO);
874 /* MDT supports FMD for regular files due to Data-on-MDT */
875 if (S_ISREG(lu_object_attr(&mo->mot_obj)) &&
876 ma->ma_attr.la_valid & (LA_ATIME | LA_MTIME | LA_CTIME))
877 tgt_fmd_update(info->mti_exp, mdt_object_fid(mo),
880 rc = mdt_attr_set(info, mo, ma);
883 } else if ((ma->ma_valid & (MA_LOV | MA_LMV)) &&
884 (ma->ma_valid & MA_INODE)) {
885 struct lu_buf *buf = &info->mti_buf;
886 struct lu_ucred *uc = mdt_ucred(info);
887 struct mdt_lock_handle *lh;
889 __u64 lockpart = MDS_INODELOCK_XATTR;
891 /* reject if either remote or striped dir is disabled */
892 if (ma->ma_valid & MA_LMV) {
893 if (!mdt->mdt_enable_remote_dir ||
894 !mdt->mdt_enable_striped_dir)
895 GOTO(out_put, rc = -EPERM);
897 if (!md_capable(uc, CFS_CAP_SYS_ADMIN) &&
898 uc->uc_gid != mdt->mdt_enable_remote_dir_gid &&
899 mdt->mdt_enable_remote_dir_gid != -1)
900 GOTO(out_put, rc = -EPERM);
903 if (!S_ISDIR(lu_object_attr(&mo->mot_obj)))
904 GOTO(out_put, rc = -ENOTDIR);
906 if (ma->ma_attr.la_valid != 0)
907 GOTO(out_put, rc = -EPROTO);
909 if (ma->ma_valid & MA_LOV) {
910 buf->lb_buf = ma->ma_lmm;
911 buf->lb_len = ma->ma_lmm_size;
912 name = XATTR_NAME_LOV;
914 struct lmv_user_md *lmu = &ma->ma_lmv->lmv_user_md;
917 buf->lb_len = ma->ma_lmv_size;
918 name = XATTR_NAME_DEFAULT_LMV;
919 /* force client to update dir default layout */
920 lockpart |= MDS_INODELOCK_LOOKUP;
923 lh = &info->mti_lh[MDT_LH_PARENT];
924 mdt_lock_reg_init(lh, LCK_PW);
926 rc = mdt_object_lock(info, mo, lh, lockpart);
930 rc = mo_xattr_set(info->mti_env, mdt_object_child(mo), buf,
933 mdt_object_unlock(info, mo, lh, rc);
937 GOTO(out_put, rc = -EPROTO);
940 /* If file data is modified, add the dirty flag */
941 if (ma->ma_attr_flags & MDS_DATA_MODIFIED)
942 rc = mdt_add_dirty_flag(info, mo, ma);
944 ma->ma_need = MA_INODE;
946 rc = mdt_attr_get_complex(info, mo, ma);
950 mdt_pack_attr2body(info, repbody, &ma->ma_attr, mdt_object_fid(mo));
954 mdt_object_put(info->mti_env, mo);
957 mdt_counter_incr(req, LPROC_MDT_SETATTR);
959 mdt_client_compatibility(info);
960 rc2 = mdt_fix_reply(info);
966 static int mdt_reint_create(struct mdt_thread_info *info,
967 struct mdt_lock_handle *lhc)
969 struct ptlrpc_request *req = mdt_info_req(info);
973 if (OBD_FAIL_CHECK(OBD_FAIL_MDS_REINT_CREATE))
974 RETURN(err_serious(-ESTALE));
976 if (info->mti_dlm_req)
977 ldlm_request_cancel(mdt_info_req(info),
978 info->mti_dlm_req, 0, LATF_SKIP);
980 if (!lu_name_is_valid(&info->mti_rr.rr_name))
983 switch (info->mti_attr.ma_attr.la_mode & S_IFMT) {
985 mdt_counter_incr(req, LPROC_MDT_MKDIR);
993 /* Special file should stay on the same node as parent. */
994 mdt_counter_incr(req, LPROC_MDT_MKNOD);
997 CERROR("%s: Unsupported mode %o\n",
998 mdt_obd_name(info->mti_mdt),
999 info->mti_attr.ma_attr.la_mode);
1000 RETURN(err_serious(-EOPNOTSUPP));
1003 rc = mdt_create(info);
1008 * VBR: save parent version in reply and child version getting by its name.
1009 * Version of child is getting and checking during its lookup. If
1011 static int mdt_reint_unlink(struct mdt_thread_info *info,
1012 struct mdt_lock_handle *lhc)
1014 struct mdt_reint_record *rr = &info->mti_rr;
1015 struct ptlrpc_request *req = mdt_info_req(info);
1016 struct md_attr *ma = &info->mti_attr;
1017 struct lu_fid *child_fid = &info->mti_tmp_fid1;
1018 struct mdt_object *mp;
1019 struct mdt_object *mc;
1020 struct mdt_lock_handle *parent_lh;
1021 struct mdt_lock_handle *child_lh;
1022 struct ldlm_enqueue_info *einfo = &info->mti_einfo[0];
1024 bool cos_incompat = false;
1030 DEBUG_REQ(D_INODE, req, "unlink "DFID"/"DNAME"", PFID(rr->rr_fid1),
1031 PNAME(&rr->rr_name));
1033 if (info->mti_dlm_req)
1034 ldlm_request_cancel(req, info->mti_dlm_req, 0, LATF_SKIP);
1036 if (OBD_FAIL_CHECK(OBD_FAIL_MDS_REINT_UNLINK))
1037 RETURN(err_serious(-ENOENT));
1039 if (!fid_is_md_operative(rr->rr_fid1))
1042 mp = mdt_object_find(info->mti_env, info->mti_mdt, rr->rr_fid1);
1044 RETURN(PTR_ERR(mp));
1046 if (mdt_object_remote(mp)) {
1047 cos_incompat = true;
1049 rc = mdt_version_get_check_save(info, mp, 0);
1051 GOTO(put_parent, rc);
1055 parent_lh = &info->mti_lh[MDT_LH_PARENT];
1056 mdt_lock_pdo_init(parent_lh, LCK_PW, &rr->rr_name);
1057 rc = mdt_reint_object_lock(info, mp, parent_lh, MDS_INODELOCK_UPDATE,
1060 GOTO(put_parent, rc);
1062 /* lookup child object along with version checking */
1063 fid_zero(child_fid);
1064 rc = mdt_lookup_version_check(info, mp, &rr->rr_name, child_fid, 1);
1066 /* Name might not be able to find during resend of
1067 * remote unlink, considering following case.
1068 * dir_A is a remote directory, the name entry of
1069 * dir_A is on MDT0, the directory is on MDT1,
1071 * 1. client sends unlink req to MDT1.
1072 * 2. MDT1 sends name delete update to MDT0.
1073 * 3. name entry is being deleted in MDT0 synchronously.
1074 * 4. MDT1 is restarted.
1075 * 5. client resends unlink req to MDT1. So it can not
1076 * find the name entry on MDT0 anymore.
1077 * In this case, MDT1 only needs to destory the local
1080 if (mdt_object_remote(mp) && rc == -ENOENT &&
1081 !fid_is_zero(rr->rr_fid2) &&
1082 lustre_msg_get_flags(req->rq_reqmsg) & MSG_RESENT) {
1084 *child_fid = *rr->rr_fid2;
1086 GOTO(unlock_parent, rc);
1090 if (!fid_is_md_operative(child_fid))
1091 GOTO(unlock_parent, rc = -EPERM);
1093 /* We will lock the child regardless it is local or remote. No harm. */
1094 mc = mdt_object_find(info->mti_env, info->mti_mdt, child_fid);
1096 GOTO(unlock_parent, rc = PTR_ERR(mc));
1098 if (!cos_incompat) {
1099 rc = mdt_object_striped(info, mc);
1101 GOTO(put_child, rc);
1105 mdt_object_put(info->mti_env, mc);
1106 mdt_object_unlock(info, mp, parent_lh, -EAGAIN);
1111 child_lh = &info->mti_lh[MDT_LH_CHILD];
1112 mdt_lock_reg_init(child_lh, LCK_EX);
1113 if (info->mti_spec.sp_rm_entry) {
1114 struct lu_ucred *uc = mdt_ucred(info);
1116 if (!mdt_is_dne_client(req->rq_export))
1117 /* Return -ENOTSUPP for old client */
1118 GOTO(put_child, rc = -ENOTSUPP);
1120 if (!md_capable(uc, CFS_CAP_SYS_ADMIN))
1121 GOTO(put_child, rc = -EPERM);
1123 ma->ma_need = MA_INODE;
1125 rc = mdo_unlink(info->mti_env, mdt_object_child(mp),
1126 NULL, &rr->rr_name, ma, no_name);
1127 GOTO(put_child, rc);
1130 if (mdt_object_remote(mc)) {
1131 struct mdt_body *repbody;
1133 if (!fid_is_zero(rr->rr_fid2)) {
1134 CDEBUG(D_INFO, "%s: name "DNAME" cannot find "DFID"\n",
1135 mdt_obd_name(info->mti_mdt),
1136 PNAME(&rr->rr_name), PFID(mdt_object_fid(mc)));
1137 GOTO(put_child, rc = -ENOENT);
1139 CDEBUG(D_INFO, "%s: name "DNAME": "DFID" is on another MDT\n",
1140 mdt_obd_name(info->mti_mdt),
1141 PNAME(&rr->rr_name), PFID(mdt_object_fid(mc)));
1143 if (!mdt_is_dne_client(req->rq_export))
1144 /* Return -ENOTSUPP for old client */
1145 GOTO(put_child, rc = -ENOTSUPP);
1147 /* Revoke the LOOKUP lock of the remote object granted by
1148 * this MDT. Since the unlink will happen on another MDT,
1149 * it will release the LOOKUP lock right away. Then What
1150 * would happen if another client try to grab the LOOKUP
1151 * lock at the same time with unlink XXX */
1152 mdt_object_lock(info, mc, child_lh, MDS_INODELOCK_LOOKUP);
1153 repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY);
1154 LASSERT(repbody != NULL);
1155 repbody->mbo_fid1 = *mdt_object_fid(mc);
1156 repbody->mbo_valid |= (OBD_MD_FLID | OBD_MD_MDS);
1157 GOTO(unlock_child, rc = -EREMOTE);
1159 /* We used to acquire MDS_INODELOCK_FULL here but we can't do
1160 * this now because a running HSM restore on the child (unlink
1161 * victim) will hold the layout lock. See LU-4002. */
1162 lock_ibits = MDS_INODELOCK_LOOKUP | MDS_INODELOCK_UPDATE;
1163 if (mdt_object_remote(mp)) {
1164 /* Enqueue lookup lock from parent MDT */
1165 rc = mdt_remote_object_lock(info, mp, mdt_object_fid(mc),
1166 &child_lh->mlh_rreg_lh,
1167 child_lh->mlh_rreg_mode,
1168 MDS_INODELOCK_LOOKUP, false);
1170 GOTO(put_child, rc);
1172 lock_ibits &= ~MDS_INODELOCK_LOOKUP;
1175 rc = mdt_reint_striped_lock(info, mc, child_lh, lock_ibits, einfo,
1178 GOTO(put_child, rc);
1181 * Now we can only make sure we need MA_INODE, in mdd layer, will check
1182 * whether need MA_LOV and MA_COOKIE.
1184 ma->ma_need = MA_INODE;
1187 mdt_fail_write(info->mti_env, info->mti_mdt->mdt_bottom,
1188 OBD_FAIL_MDS_REINT_UNLINK_WRITE);
1189 /* save version when object is locked */
1190 mdt_version_get_save(info, mc, 1);
1192 mutex_lock(&mc->mot_lov_mutex);
1194 rc = mdo_unlink(info->mti_env, mdt_object_child(mp),
1195 mdt_object_child(mc), &rr->rr_name, ma, no_name);
1197 mutex_unlock(&mc->mot_lov_mutex);
1199 GOTO(unlock_child, rc);
1201 if (!lu_object_is_dying(&mc->mot_header)) {
1202 rc = mdt_attr_get_complex(info, mc, ma);
1205 } else if (mdt_dom_check_for_discard(info, mc)) {
1206 mdt_dom_discard_data(info, mc);
1208 mdt_handle_last_unlink(info, mc, ma);
1211 if (ma->ma_valid & MA_INODE) {
1212 switch (ma->ma_attr.la_mode & S_IFMT) {
1214 mdt_counter_incr(req, LPROC_MDT_RMDIR);
1222 mdt_counter_incr(req, LPROC_MDT_UNLINK);
1225 LASSERTF(0, "bad file type %o unlinking\n",
1226 ma->ma_attr.la_mode);
1233 mdt_reint_striped_unlock(info, mc, child_lh, einfo, rc);
1235 mdt_object_put(info->mti_env, mc);
1237 mdt_object_unlock(info, mp, parent_lh, rc);
1239 mdt_object_put(info->mti_env, mp);
1240 CFS_RACE_WAKEUP(OBD_FAIL_OBD_ZERO_NLINK_RACE);
1245 * VBR: save versions in reply: 0 - parent; 1 - child by fid; 2 - target by
1248 static int mdt_reint_link(struct mdt_thread_info *info,
1249 struct mdt_lock_handle *lhc)
1251 struct mdt_reint_record *rr = &info->mti_rr;
1252 struct ptlrpc_request *req = mdt_info_req(info);
1253 struct md_attr *ma = &info->mti_attr;
1254 struct mdt_object *ms;
1255 struct mdt_object *mp;
1256 struct mdt_lock_handle *lhs;
1257 struct mdt_lock_handle *lhp;
1262 DEBUG_REQ(D_INODE, req, "link "DFID" to "DFID"/"DNAME,
1263 PFID(rr->rr_fid1), PFID(rr->rr_fid2), PNAME(&rr->rr_name));
1265 if (OBD_FAIL_CHECK(OBD_FAIL_MDS_REINT_LINK))
1266 RETURN(err_serious(-ENOENT));
1268 if (OBD_FAIL_PRECHECK(OBD_FAIL_PTLRPC_RESEND_RACE)) {
1269 req->rq_no_reply = 1;
1270 RETURN(err_serious(-ENOENT));
1273 if (info->mti_dlm_req)
1274 ldlm_request_cancel(req, info->mti_dlm_req, 0, LATF_SKIP);
1276 /* Invalid case so return error immediately instead of
1278 if (lu_fid_eq(rr->rr_fid1, rr->rr_fid2))
1281 if (!fid_is_md_operative(rr->rr_fid1) ||
1282 !fid_is_md_operative(rr->rr_fid2))
1285 /* step 1: find target parent dir */
1286 mp = mdt_object_find(info->mti_env, info->mti_mdt, rr->rr_fid2);
1288 RETURN(PTR_ERR(mp));
1290 rc = mdt_version_get_check_save(info, mp, 0);
1292 GOTO(put_parent, rc);
1294 /* step 2: find source */
1295 ms = mdt_object_find(info->mti_env, info->mti_mdt, rr->rr_fid1);
1297 GOTO(put_parent, rc = PTR_ERR(ms));
1299 if (!mdt_object_exists(ms)) {
1300 CDEBUG(D_INFO, "%s: "DFID" does not exist.\n",
1301 mdt_obd_name(info->mti_mdt), PFID(rr->rr_fid1));
1302 GOTO(put_source, rc = -ENOENT);
1305 cos_incompat = (mdt_object_remote(mp) || mdt_object_remote(ms));
1307 lhp = &info->mti_lh[MDT_LH_PARENT];
1308 mdt_lock_pdo_init(lhp, LCK_PW, &rr->rr_name);
1309 rc = mdt_reint_object_lock(info, mp, lhp, MDS_INODELOCK_UPDATE,
1312 GOTO(put_source, rc);
1314 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_RENAME3, 5);
1316 lhs = &info->mti_lh[MDT_LH_CHILD];
1317 mdt_lock_reg_init(lhs, LCK_EX);
1318 rc = mdt_reint_object_lock(info, ms, lhs,
1319 MDS_INODELOCK_UPDATE | MDS_INODELOCK_XATTR,
1322 GOTO(unlock_parent, rc);
1324 /* step 3: link it */
1325 mdt_fail_write(info->mti_env, info->mti_mdt->mdt_bottom,
1326 OBD_FAIL_MDS_REINT_LINK_WRITE);
1328 tgt_vbr_obj_set(info->mti_env, mdt_obj2dt(ms));
1329 rc = mdt_version_get_check_save(info, ms, 1);
1331 GOTO(unlock_source, rc);
1333 /** check target version by name during replay */
1334 rc = mdt_lookup_version_check(info, mp, &rr->rr_name,
1335 &info->mti_tmp_fid1, 2);
1336 if (rc != 0 && rc != -ENOENT)
1337 GOTO(unlock_source, rc);
1338 /* save version of file name for replay, it must be ENOENT here */
1339 if (!req_is_replay(mdt_info_req(info))) {
1340 if (rc != -ENOENT) {
1341 CDEBUG(D_INFO, "link target "DNAME" existed!\n",
1342 PNAME(&rr->rr_name));
1343 GOTO(unlock_source, rc = -EEXIST);
1345 info->mti_ver[2] = ENOENT_VERSION;
1346 mdt_version_save(mdt_info_req(info), info->mti_ver[2], 2);
1349 rc = mdo_link(info->mti_env, mdt_object_child(mp),
1350 mdt_object_child(ms), &rr->rr_name, ma);
1353 mdt_counter_incr(req, LPROC_MDT_LINK);
1357 mdt_object_unlock(info, ms, lhs, rc);
1359 mdt_object_unlock(info, mp, lhp, rc);
1361 mdt_object_put(info->mti_env, ms);
1363 mdt_object_put(info->mti_env, mp);
1367 * lock the part of the directory according to the hash of the name
1368 * (lh->mlh_pdo_hash) in parallel directory lock.
1370 static int mdt_pdir_hash_lock(struct mdt_thread_info *info,
1371 struct mdt_lock_handle *lh,
1372 struct mdt_object *obj, __u64 ibits,
1375 struct ldlm_res_id *res = &info->mti_res_id;
1376 struct ldlm_namespace *ns = info->mti_mdt->mdt_namespace;
1377 union ldlm_policy_data *policy = &info->mti_policy;
1378 __u64 dlmflags = LDLM_FL_LOCAL_ONLY | LDLM_FL_ATOMIC_CB;
1382 * Finish res_id initializing by name hash marking part of
1383 * directory which is taking modification.
1385 LASSERT(lh->mlh_pdo_hash != 0);
1386 fid_build_pdo_res_name(mdt_object_fid(obj), lh->mlh_pdo_hash, res);
1387 memset(policy, 0, sizeof(*policy));
1388 policy->l_inodebits.bits = ibits;
1390 (lh->mlh_reg_mode == LCK_PW || lh->mlh_reg_mode == LCK_EX))
1391 dlmflags |= LDLM_FL_COS_INCOMPAT;
1393 * Use LDLM_FL_LOCAL_ONLY for this lock. We do not know yet if it is
1394 * going to be sent to client. If it is - mdt_intent_policy() path will
1395 * fix it up and turn FL_LOCAL flag off.
1397 rc = mdt_fid_lock(info->mti_env, ns, &lh->mlh_reg_lh, lh->mlh_reg_mode,
1398 policy, res, dlmflags,
1399 &info->mti_exp->exp_handle.h_cookie);
1404 * Get BFL lock for rename or migrate process.
1406 static int mdt_rename_lock(struct mdt_thread_info *info,
1407 struct lustre_handle *lh)
1412 if (mdt_seq_site(info->mti_mdt)->ss_node_id != 0) {
1413 struct lu_fid *fid = &info->mti_tmp_fid1;
1414 struct mdt_object *obj;
1416 /* XXX, right now, it has to use object API to
1417 * enqueue lock cross MDT, so it will enqueue
1418 * rename lock(with LUSTRE_BFL_FID) by root object */
1420 obj = mdt_object_find(info->mti_env, info->mti_mdt, fid);
1422 RETURN(PTR_ERR(obj));
1424 rc = mdt_remote_object_lock(info, obj,
1425 &LUSTRE_BFL_FID, lh,
1427 MDS_INODELOCK_UPDATE, false);
1428 mdt_object_put(info->mti_env, obj);
1430 struct ldlm_namespace *ns = info->mti_mdt->mdt_namespace;
1431 union ldlm_policy_data *policy = &info->mti_policy;
1432 struct ldlm_res_id *res_id = &info->mti_res_id;
1435 fid_build_reg_res_name(&LUSTRE_BFL_FID, res_id);
1436 memset(policy, 0, sizeof *policy);
1437 policy->l_inodebits.bits = MDS_INODELOCK_UPDATE;
1438 flags = LDLM_FL_LOCAL_ONLY | LDLM_FL_ATOMIC_CB;
1439 rc = ldlm_cli_enqueue_local(info->mti_env, ns, res_id,
1440 LDLM_IBITS, policy, LCK_EX, &flags,
1442 ldlm_completion_ast, NULL, NULL, 0,
1444 &info->mti_exp->exp_handle.h_cookie,
1451 static void mdt_rename_unlock(struct lustre_handle *lh)
1454 LASSERT(lustre_handle_is_used(lh));
1455 /* Cancel the single rename lock right away */
1456 ldlm_lock_decref_and_cancel(lh, LCK_EX);
1460 static struct mdt_object *mdt_parent_find_check(struct mdt_thread_info *info,
1461 const struct lu_fid *fid,
1464 struct mdt_object *dir;
1469 dir = mdt_object_find(info->mti_env, info->mti_mdt, fid);
1473 /* check early, the real version will be saved after locking */
1474 rc = mdt_version_get_check(info, dir, idx);
1478 if (!mdt_object_exists(dir))
1479 GOTO(out_put, rc = -ENOENT);
1481 if (!S_ISDIR(lu_object_attr(&dir->mot_obj)))
1482 GOTO(out_put, rc = -ENOTDIR);
1486 mdt_object_put(info->mti_env, dir);
1491 * in case obj is remote obj on its parent, revoke LOOKUP lock,
1492 * herein we don't really check it, just do revoke.
1494 int mdt_revoke_remote_lookup_lock(struct mdt_thread_info *info,
1495 struct mdt_object *pobj,
1496 struct mdt_object *obj)
1498 struct mdt_lock_handle *lh = &info->mti_lh[MDT_LH_LOCAL];
1501 mdt_lock_handle_init(lh);
1502 mdt_lock_reg_init(lh, LCK_EX);
1504 if (mdt_object_remote(pobj)) {
1505 /* don't bother to check if pobj and obj are on the same MDT. */
1506 rc = mdt_remote_object_lock(info, pobj, mdt_object_fid(obj),
1507 &lh->mlh_rreg_lh, LCK_EX,
1508 MDS_INODELOCK_LOOKUP, false);
1509 } else if (mdt_object_remote(obj)) {
1510 struct ldlm_res_id *res = &info->mti_res_id;
1511 union ldlm_policy_data *policy = &info->mti_policy;
1512 __u64 dlmflags = LDLM_FL_LOCAL_ONLY | LDLM_FL_ATOMIC_CB |
1513 LDLM_FL_COS_INCOMPAT;
1515 fid_build_reg_res_name(mdt_object_fid(obj), res);
1516 memset(policy, 0, sizeof(*policy));
1517 policy->l_inodebits.bits = MDS_INODELOCK_LOOKUP;
1518 rc = mdt_fid_lock(info->mti_env, info->mti_mdt->mdt_namespace,
1519 &lh->mlh_reg_lh, LCK_EX, policy, res,
1522 /* do nothing if both are local */
1530 * TODO, currently we don't save this lock because there is no place to
1531 * hold this lock handle, but to avoid race we need to save this lock.
1533 mdt_object_unlock(info, NULL, lh, 1);
1539 * operation may takes locks of linkea, or directory stripes, group them in
1542 struct mdt_sub_lock {
1543 struct mdt_object *msl_obj;
1544 struct mdt_lock_handle msl_lh;
1545 struct list_head msl_linkage;
1548 static void mdt_unlock_list(struct mdt_thread_info *info,
1549 struct list_head *list, int decref)
1551 struct mdt_sub_lock *msl;
1552 struct mdt_sub_lock *tmp;
1554 list_for_each_entry_safe(msl, tmp, list, msl_linkage) {
1555 mdt_object_unlock_put(info, msl->msl_obj, &msl->msl_lh, decref);
1556 list_del(&msl->msl_linkage);
1561 static inline void mdt_migrate_object_unlock(struct mdt_thread_info *info,
1562 struct mdt_object *obj,
1563 struct mdt_lock_handle *lh,
1564 struct ldlm_enqueue_info *einfo,
1565 struct list_head *slave_locks,
1568 if (mdt_object_remote(obj)) {
1569 mdt_unlock_list(info, slave_locks, decref);
1570 mdt_object_unlock(info, obj, lh, decref);
1572 mdt_reint_striped_unlock(info, obj, lh, einfo, decref);
1577 * lock parents of links, and also check whether total locks don't exceed
1580 * \retval 0 on success, and locks can be saved in ptlrpc_reply_stat
1581 * \retval 1 on success, but total lock count may exceed RS_MAX_LOCKS
1582 * \retval -ev negative errno upon error
1584 static int mdt_link_parents_lock(struct mdt_thread_info *info,
1585 struct mdt_object *pobj,
1586 const struct md_attr *ma,
1587 struct mdt_object *obj,
1588 struct mdt_lock_handle *lhp,
1589 struct ldlm_enqueue_info *peinfo,
1590 struct list_head *parent_slave_locks,
1591 struct list_head *link_locks)
1593 struct mdt_device *mdt = info->mti_mdt;
1594 struct lu_buf *buf = &info->mti_big_buf;
1595 struct lu_name *lname = &info->mti_name;
1596 struct linkea_data ldata = { NULL };
1597 bool blocked = false;
1598 int local_lnkp_cnt = 0;
1603 if (S_ISDIR(lu_object_attr(&obj->mot_obj)))
1606 buf = lu_buf_check_and_alloc(buf, MAX_LINKEA_SIZE);
1607 if (buf->lb_buf == NULL)
1611 rc = mdt_links_read(info, obj, &ldata);
1613 if (rc == -ENOENT || rc == -ENODATA)
1618 for (linkea_first_entry(&ldata); ldata.ld_lee && !rc;
1619 linkea_next_entry(&ldata)) {
1620 struct mdt_object *lnkp;
1621 struct mdt_sub_lock *msl;
1625 linkea_entry_unpack(ldata.ld_lee, &ldata.ld_reclen, lname,
1628 /* check if it's also linked to parent */
1629 if (lu_fid_eq(mdt_object_fid(pobj), &fid)) {
1630 CDEBUG(D_INFO, "skip parent "DFID", reovke "DNAME"\n",
1631 PFID(&fid), PNAME(lname));
1632 /* in case link is remote object, revoke LOOKUP lock */
1633 rc = mdt_revoke_remote_lookup_lock(info, pobj, obj);
1639 /* check if it's linked to a stripe of parent */
1640 if (ma->ma_valid & MA_LMV) {
1641 struct lmv_mds_md_v1 *lmv = &ma->ma_lmv->lmv_md_v1;
1642 struct lu_fid *stripe_fid = &info->mti_tmp_fid1;
1645 for (; j < le32_to_cpu(lmv->lmv_stripe_count); j++) {
1646 fid_le_to_cpu(stripe_fid,
1647 &lmv->lmv_stripe_fids[j]);
1648 if (lu_fid_eq(stripe_fid, &fid)) {
1649 CDEBUG(D_INFO, "skip stripe "DFID
1650 ", reovke "DNAME"\n",
1651 PFID(&fid), PNAME(lname));
1652 lnkp = mdt_object_find(info->mti_env,
1655 GOTO(out, rc = PTR_ERR(lnkp));
1661 rc = mdt_revoke_remote_lookup_lock(info, lnkp,
1663 mdt_object_put(info->mti_env, lnkp);
1668 /* Check if it's already locked */
1669 list_for_each_entry(msl, link_locks, msl_linkage) {
1670 if (lu_fid_eq(mdt_object_fid(msl->msl_obj), &fid)) {
1672 DFID" was locked, revoke "DNAME"\n",
1673 PFID(&fid), PNAME(lname));
1674 lnkp = msl->msl_obj;
1680 rc = mdt_revoke_remote_lookup_lock(info, lnkp, obj);
1684 CDEBUG(D_INFO, "lock "DFID":"DNAME"\n",
1685 PFID(&fid), PNAME(lname));
1687 lnkp = mdt_object_find(info->mti_env, mdt, &fid);
1689 CWARN("%s: cannot find obj "DFID": %ld\n",
1690 mdt_obd_name(mdt), PFID(&fid), PTR_ERR(lnkp));
1694 if (!mdt_object_exists(lnkp)) {
1695 CDEBUG(D_INFO, DFID" doesn't exist, skip "DNAME"\n",
1696 PFID(&fid), PNAME(lname));
1697 mdt_object_put(info->mti_env, lnkp);
1701 if (!mdt_object_remote(lnkp))
1706 GOTO(out, rc = -ENOMEM);
1709 * we can't follow parent-child lock order like other MD
1710 * operations, use lock_try here to avoid deadlock, if the lock
1711 * cannot be taken, drop all locks taken, revoke the blocked
1712 * one, and continue processing the remaining entries, and in
1713 * the end of the loop restart from beginning.
1715 mdt_lock_pdo_init(&msl->msl_lh, LCK_PW, lname);
1717 rc = mdt_object_lock_try(info, lnkp, &msl->msl_lh, &ibits,
1718 MDS_INODELOCK_UPDATE, true);
1719 if (!(ibits & MDS_INODELOCK_UPDATE)) {
1721 CDEBUG(D_INFO, "busy lock on "DFID" "DNAME"\n",
1722 PFID(&fid), PNAME(lname));
1724 mdt_unlock_list(info, link_locks, 1);
1725 /* also unlock parent locks to avoid deadlock */
1727 mdt_migrate_object_unlock(info, pobj, lhp,
1734 mdt_lock_pdo_init(&msl->msl_lh, LCK_PW, lname);
1735 rc = mdt_object_lock(info, lnkp, &msl->msl_lh,
1736 MDS_INODELOCK_UPDATE);
1738 mdt_object_put(info->mti_env, lnkp);
1743 if (mdt_object_remote(lnkp)) {
1744 struct ldlm_lock *lock;
1747 * for remote object, set lock cb_atomic,
1748 * so lock can be released in blocking_ast()
1749 * immediately, then the next lock_try will
1750 * have better chance of success.
1752 lock = ldlm_handle2lock(
1753 &msl->msl_lh.mlh_rreg_lh);
1754 LASSERT(lock != NULL);
1755 lock_res_and_lock(lock);
1756 ldlm_set_atomic_cb(lock);
1757 unlock_res_and_lock(lock);
1758 LDLM_LOCK_PUT(lock);
1761 mdt_object_unlock_put(info, lnkp, &msl->msl_lh, 1);
1766 INIT_LIST_HEAD(&msl->msl_linkage);
1767 msl->msl_obj = lnkp;
1768 list_add_tail(&msl->msl_linkage, link_locks);
1770 rc = mdt_revoke_remote_lookup_lock(info, lnkp, obj);
1774 GOTO(out, rc = -EBUSY);
1779 mdt_unlock_list(info, link_locks, rc);
1780 else if (local_lnkp_cnt > RS_MAX_LOCKS - 6)
1782 * parent may have 3 local objects: master object and 2 stripes
1783 * (if it's being migrated too); source may have 2 local
1784 * objects: master and 1 stripe; target has 1 local object.
1790 static int mdt_lock_remote_slaves(struct mdt_thread_info *info,
1791 struct mdt_object *obj,
1792 const struct md_attr *ma,
1793 struct list_head *slave_locks)
1795 struct mdt_device *mdt = info->mti_mdt;
1796 const struct lmv_mds_md_v1 *lmv = &ma->ma_lmv->lmv_md_v1;
1797 struct lu_fid *fid = &info->mti_tmp_fid1;
1798 struct mdt_object *slave;
1799 struct mdt_sub_lock *msl;
1805 LASSERT(mdt_object_remote(obj));
1806 LASSERT(ma->ma_valid & MA_LMV);
1809 if (!lmv_is_sane(lmv))
1812 for (i = 0; i < le32_to_cpu(lmv->lmv_stripe_count); i++) {
1813 fid_le_to_cpu(fid, &lmv->lmv_stripe_fids[i]);
1815 if (!fid_is_sane(fid))
1818 slave = mdt_object_find(info->mti_env, mdt, fid);
1820 GOTO(out, rc = PTR_ERR(slave));
1824 mdt_object_put(info->mti_env, slave);
1825 GOTO(out, rc = -ENOMEM);
1828 mdt_lock_reg_init(&msl->msl_lh, LCK_EX);
1829 rc = mdt_reint_object_lock(info, slave, &msl->msl_lh,
1830 MDS_INODELOCK_UPDATE, true);
1833 mdt_object_put(info->mti_env, slave);
1837 INIT_LIST_HEAD(&msl->msl_linkage);
1838 msl->msl_obj = slave;
1839 list_add_tail(&msl->msl_linkage, slave_locks);
1845 mdt_unlock_list(info, slave_locks, rc);
1849 /* lock parent and its stripes */
1850 static int mdt_migrate_parent_lock(struct mdt_thread_info *info,
1851 struct mdt_object *obj,
1852 const struct md_attr *ma,
1853 struct mdt_lock_handle *lh,
1854 struct ldlm_enqueue_info *einfo,
1855 struct list_head *slave_locks)
1859 if (mdt_object_remote(obj)) {
1860 rc = mdt_remote_object_lock(info, obj, mdt_object_fid(obj),
1861 &lh->mlh_rreg_lh, LCK_PW,
1862 MDS_INODELOCK_UPDATE, false);
1867 * if obj is remote and striped, lock its stripes explicitly
1868 * because it's not striped in LOD layer on this MDT.
1870 if (ma->ma_valid & MA_LMV) {
1871 rc = mdt_lock_remote_slaves(info, obj, ma, slave_locks);
1873 mdt_object_unlock(info, obj, lh, rc);
1876 rc = mdt_reint_striped_lock(info, obj, lh, MDS_INODELOCK_UPDATE,
1884 * in migration, object may be remote, and we need take full lock of it and its
1885 * stripes if it's directory, besides, object may be a remote object on its
1886 * parent, revoke its LOOKUP lock on where its parent is located.
1888 static int mdt_migrate_object_lock(struct mdt_thread_info *info,
1889 struct mdt_object *pobj,
1890 struct mdt_object *obj,
1891 struct mdt_lock_handle *lh,
1892 struct ldlm_enqueue_info *einfo,
1893 struct list_head *slave_locks)
1897 if (mdt_object_remote(obj)) {
1898 rc = mdt_revoke_remote_lookup_lock(info, pobj, obj);
1902 rc = mdt_remote_object_lock(info, obj, mdt_object_fid(obj),
1903 &lh->mlh_rreg_lh, LCK_EX,
1904 MDS_INODELOCK_FULL, false);
1909 * if obj is remote and striped, lock its stripes explicitly
1910 * because it's not striped in LOD layer on this MDT.
1912 if (S_ISDIR(lu_object_attr(&obj->mot_obj))) {
1913 struct md_attr *ma = &info->mti_attr;
1915 rc = mdt_stripe_get(info, obj, ma, XATTR_NAME_LMV);
1917 mdt_object_unlock(info, obj, lh, rc);
1921 if (ma->ma_valid & MA_LMV) {
1922 rc = mdt_lock_remote_slaves(info, obj, ma,
1925 mdt_object_unlock(info, obj, lh, rc);
1929 if (mdt_object_remote(pobj)) {
1930 rc = mdt_revoke_remote_lookup_lock(info, pobj, obj);
1935 rc = mdt_reint_striped_lock(info, obj, lh, MDS_INODELOCK_FULL,
1943 * lookup source by name, if parent is striped directory, we need to find the
1944 * corresponding stripe where source is located, and then lookup there.
1946 * besides, if parent is migrating too, and file is already in target stripe,
1947 * this should be a redo of 'lfs migrate' on client side.
1949 static int mdt_migrate_lookup(struct mdt_thread_info *info,
1950 struct mdt_object *pobj,
1951 const struct md_attr *ma,
1952 const struct lu_name *lname,
1953 struct mdt_object **spobj,
1954 struct mdt_object **sobj)
1956 const struct lu_env *env = info->mti_env;
1957 struct lu_fid *fid = &info->mti_tmp_fid1;
1958 struct mdt_object *stripe;
1961 if (ma->ma_valid & MA_LMV) {
1962 /* if parent is striped, lookup on corresponding stripe */
1963 struct lmv_mds_md_v1 *lmv = &ma->ma_lmv->lmv_md_v1;
1965 if (!lmv_is_sane(lmv))
1968 rc = lmv_name_to_stripe_index_old(lmv, lname->ln_name,
1973 fid_le_to_cpu(fid, &lmv->lmv_stripe_fids[rc]);
1975 stripe = mdt_object_find(env, info->mti_mdt, fid);
1977 return PTR_ERR(stripe);
1980 rc = mdo_lookup(env, mdt_object_child(stripe), lname, fid,
1982 if (rc == -ENOENT && lmv_is_layout_changing(lmv)) {
1984 * if parent layout is changeing, and lookup child
1985 * failed on source stripe, lookup again on target
1986 * stripe, if it exists, it means previous migration
1987 * was interrupted, and current file was migrated
1990 mdt_object_put(env, stripe);
1992 rc = lmv_name_to_stripe_index(lmv, lname->ln_name,
1997 fid_le_to_cpu(fid, &lmv->lmv_stripe_fids[rc]);
1999 stripe = mdt_object_find(env, info->mti_mdt, fid);
2001 return PTR_ERR(stripe);
2004 rc = mdo_lookup(env, mdt_object_child(stripe), lname,
2005 fid, &info->mti_spec);
2006 mdt_object_put(env, stripe);
2007 return rc ?: -EALREADY;
2009 mdt_object_put(env, stripe);
2014 rc = mdo_lookup(env, mdt_object_child(pobj), lname, fid,
2020 mdt_object_get(env, stripe);
2025 *sobj = mdt_object_find(env, info->mti_mdt, fid);
2026 if (IS_ERR(*sobj)) {
2027 mdt_object_put(env, stripe);
2028 rc = PTR_ERR(*sobj);
2036 /* end lease and close file for regular file */
2037 static int mdd_migrate_close(struct mdt_thread_info *info,
2038 struct mdt_object *obj)
2040 struct close_data *data;
2041 struct mdt_body *repbody;
2042 struct ldlm_lock *lease;
2047 if (!req_capsule_field_present(info->mti_pill, &RMF_MDT_EPOCH,
2049 !req_capsule_field_present(info->mti_pill, &RMF_CLOSE_DATA,
2053 data = req_capsule_client_get(info->mti_pill, &RMF_CLOSE_DATA);
2058 lease = ldlm_handle2lock(&data->cd_handle);
2062 /* check if the lease was already canceled */
2063 lock_res_and_lock(lease);
2064 rc = ldlm_is_cancel(lease);
2065 unlock_res_and_lock(lease);
2069 LDLM_DEBUG(lease, DFID" lease broken",
2070 PFID(mdt_object_fid(obj)));
2074 * cancel server side lease, client side counterpart should have been
2075 * cancelled, it's okay to cancel it now as we've held mot_open_sem.
2077 ldlm_lock_cancel(lease);
2078 ldlm_reprocess_all(lease->l_resource, lease);
2079 LDLM_LOCK_PUT(lease);
2082 rc2 = mdt_close_internal(info, mdt_info_req(info), NULL);
2083 repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY);
2084 repbody->mbo_valid |= OBD_MD_CLOSE_INTENT_EXECED;
2090 * migrate file in below steps:
2091 * 1. lock parent and its stripes
2092 * 2. lookup source by name
2093 * 3. lock parents of source links if source is not directory
2094 * 4. reject if source is in HSM
2095 * 5. take source open_sem and close file if source is regular file
2096 * 6. lock source and its stripes if it's directory
2097 * 7. lock target so subsequent change to it can trigger COS
2099 * 9. unlock above locks
2100 * 10. sync device if source has links
2102 int mdt_reint_migrate(struct mdt_thread_info *info,
2103 struct mdt_lock_handle *unused)
2105 const struct lu_env *env = info->mti_env;
2106 struct mdt_device *mdt = info->mti_mdt;
2107 struct ptlrpc_request *req = mdt_info_req(info);
2108 struct mdt_reint_record *rr = &info->mti_rr;
2109 struct lu_ucred *uc = mdt_ucred(info);
2110 struct md_attr *ma = &info->mti_attr;
2111 struct ldlm_enqueue_info *peinfo = &info->mti_einfo[0];
2112 struct ldlm_enqueue_info *seinfo = &info->mti_einfo[1];
2113 struct mdt_object *pobj;
2114 struct mdt_object *spobj = NULL;
2115 struct mdt_object *sobj = NULL;
2116 struct mdt_object *tobj;
2117 struct lustre_handle rename_lh = { 0 };
2118 struct mdt_lock_handle *lhp;
2119 struct mdt_lock_handle *lhs;
2120 struct mdt_lock_handle *lht;
2121 LIST_HEAD(parent_slave_locks);
2122 LIST_HEAD(child_slave_locks);
2123 LIST_HEAD(link_locks);
2124 int lock_retries = 5;
2125 bool open_sem_locked = false;
2126 bool do_sync = false;
2130 CDEBUG(D_INODE, "migrate "DFID"/"DNAME" to "DFID"\n", PFID(rr->rr_fid1),
2131 PNAME(&rr->rr_name), PFID(rr->rr_fid2));
2133 if (info->mti_dlm_req)
2134 ldlm_request_cancel(req, info->mti_dlm_req, 0, LATF_SKIP);
2136 if (!fid_is_md_operative(rr->rr_fid1) ||
2137 !fid_is_md_operative(rr->rr_fid2))
2140 /* don't allow migrate . or .. */
2141 if (lu_name_is_dot_or_dotdot(&rr->rr_name))
2144 if (!mdt->mdt_enable_remote_dir || !mdt->mdt_enable_dir_migration)
2147 if (uc && !md_capable(uc, CFS_CAP_SYS_ADMIN) &&
2148 uc->uc_gid != mdt->mdt_enable_remote_dir_gid &&
2149 mdt->mdt_enable_remote_dir_gid != -1)
2153 * Note: do not enqueue rename lock for replay request, because
2154 * if other MDT holds rename lock, but being blocked to wait for
2155 * this MDT to finish its recovery, and the failover MDT can not
2156 * get rename lock, which will cause deadlock.
2158 * req is NULL if this is called by directory auto-split.
2160 if (req && !req_is_replay(req)) {
2161 rc = mdt_rename_lock(info, &rename_lh);
2163 CERROR("%s: can't lock FS for rename: rc = %d\n",
2164 mdt_obd_name(info->mti_mdt), rc);
2169 /* pobj is master object of parent */
2170 pobj = mdt_object_find(env, mdt, rr->rr_fid1);
2172 GOTO(unlock_rename, rc = PTR_ERR(pobj));
2175 rc = mdt_version_get_check(info, pobj, 0);
2177 GOTO(put_parent, rc);
2180 if (!mdt_object_exists(pobj))
2181 GOTO(put_parent, rc = -ENOENT);
2183 if (!S_ISDIR(lu_object_attr(&pobj->mot_obj)))
2184 GOTO(put_parent, rc = -ENOTDIR);
2186 rc = mdt_stripe_get(info, pobj, ma, XATTR_NAME_LMV);
2188 GOTO(put_parent, rc);
2191 /* lock parent object */
2192 lhp = &info->mti_lh[MDT_LH_PARENT];
2193 mdt_lock_reg_init(lhp, LCK_PW);
2194 rc = mdt_migrate_parent_lock(info, pobj, ma, lhp, peinfo,
2195 &parent_slave_locks);
2197 GOTO(put_parent, rc);
2200 * spobj is the corresponding stripe against name if pobj is striped
2201 * directory, which is the real parent, and no need to lock, because
2202 * we've taken full lock of pobj.
2204 rc = mdt_migrate_lookup(info, pobj, ma, &rr->rr_name, &spobj, &sobj);
2206 GOTO(unlock_parent, rc);
2208 /* lock parents of source links, and revoke LOOKUP lock of links */
2209 rc = mdt_link_parents_lock(info, pobj, ma, sobj, lhp, peinfo,
2210 &parent_slave_locks, &link_locks);
2211 if (rc == -EBUSY && lock_retries-- > 0) {
2212 mdt_object_put(env, sobj);
2213 mdt_object_put(env, spobj);
2218 GOTO(put_source, rc);
2221 * RS_MAX_LOCKS is the limit of number of locks that can be saved along
2222 * with one request, if total lock count exceeds this limit, we will
2223 * drop all locks after migration, and synchronous device in the end.
2227 /* TODO: DoM migration is not supported, migrate dirent only */
2228 if (S_ISREG(lu_object_attr(&sobj->mot_obj))) {
2229 rc = mdt_stripe_get(info, sobj, ma, XATTR_NAME_LOV);
2231 GOTO(unlock_links, rc);
2233 if (ma->ma_valid & MA_LOV && mdt_lmm_dom_stripesize(ma->ma_lmm))
2234 info->mti_spec.sp_migrate_nsonly = 1;
2237 /* if migration HSM is allowed */
2238 if (!mdt->mdt_opts.mo_migrate_hsm_allowed) {
2239 ma->ma_need = MA_HSM;
2241 rc = mdt_attr_get_complex(info, sobj, ma);
2243 GOTO(unlock_links, rc);
2245 if ((ma->ma_valid & MA_HSM) && ma->ma_hsm.mh_flags != 0)
2246 GOTO(unlock_links, rc = -EOPNOTSUPP);
2249 /* end lease and close file for regular file */
2250 if (info->mti_spec.sp_migrate_close) {
2251 /* try to hold open_sem so that nobody else can open the file */
2252 if (!down_write_trylock(&sobj->mot_open_sem)) {
2254 mdd_migrate_close(info, sobj);
2255 GOTO(unlock_links, rc = -EBUSY);
2257 open_sem_locked = true;
2258 rc = mdd_migrate_close(info, sobj);
2260 GOTO(unlock_open_sem, rc);
2265 lhs = &info->mti_lh[MDT_LH_OLD];
2266 mdt_lock_reg_init(lhs, LCK_EX);
2267 rc = mdt_migrate_object_lock(info, spobj, sobj, lhs, seinfo,
2268 &child_slave_locks);
2270 GOTO(unlock_open_sem, rc);
2273 tobj = mdt_object_find(env, mdt, rr->rr_fid2);
2275 GOTO(unlock_source, rc = PTR_ERR(tobj));
2277 lht = &info->mti_lh[MDT_LH_NEW];
2278 mdt_lock_reg_init(lht, LCK_EX);
2279 rc = mdt_reint_object_lock(info, tobj, lht, MDS_INODELOCK_FULL, true);
2281 GOTO(put_target, rc);
2283 /* Don't do lookup sanity check. We know name doesn't exist. */
2284 info->mti_spec.sp_cr_lookup = 0;
2285 info->mti_spec.sp_feat = &dt_directory_features;
2287 rc = mdo_migrate(env, mdt_object_child(pobj),
2288 mdt_object_child(sobj), &rr->rr_name,
2289 mdt_object_child(tobj),
2290 &info->mti_spec, ma);
2292 lprocfs_counter_incr(mdt->mdt_lu_dev.ld_obd->obd_md_stats,
2293 LPROC_MDT_MIGRATE + LPROC_MD_LAST_OPC);
2296 mdt_object_unlock(info, tobj, lht, rc);
2298 mdt_object_put(env, tobj);
2300 mdt_migrate_object_unlock(info, sobj, lhs, seinfo,
2301 &child_slave_locks, rc);
2303 if (open_sem_locked)
2304 up_write(&sobj->mot_open_sem);
2306 /* if we've got too many locks to save into RPC,
2307 * then just commit before the locks are released */
2309 mdt_device_sync(env, mdt);
2310 mdt_unlock_list(info, &link_locks, do_sync ? 1 : rc);
2312 mdt_object_put(env, sobj);
2313 mdt_object_put(env, spobj);
2315 mdt_migrate_object_unlock(info, pobj, lhp, peinfo,
2316 &parent_slave_locks, rc);
2318 mdt_object_put(env, pobj);
2320 if (lustre_handle_is_used(&rename_lh))
2321 mdt_rename_unlock(&rename_lh);
2326 static int mdt_object_lock_save(struct mdt_thread_info *info,
2327 struct mdt_object *dir,
2328 struct mdt_lock_handle *lh,
2329 int idx, bool cos_incompat)
2333 /* we lock the target dir if it is local */
2334 rc = mdt_reint_object_lock(info, dir, lh, MDS_INODELOCK_UPDATE,
2339 /* get and save correct version after locking */
2340 mdt_version_get_save(info, dir, idx);
2345 * determine lock order of sobj and tobj
2347 * there are two situations we need to lock tobj before sobj:
2348 * 1. sobj is child of tobj
2349 * 2. sobj and tobj are stripes of a directory, and stripe index of sobj is
2350 * larger than that of tobj
2352 * \retval 1 lock tobj before sobj
2353 * \retval 0 lock sobj before tobj
2354 * \retval -ev negative errno upon error
2356 static int mdt_rename_determine_lock_order(struct mdt_thread_info *info,
2357 struct mdt_object *sobj,
2358 struct mdt_object *tobj)
2360 struct md_attr *ma = &info->mti_attr;
2361 struct lu_fid *spfid = &info->mti_tmp_fid1;
2362 struct lu_fid *tpfid = &info->mti_tmp_fid2;
2363 struct lmv_mds_md_v1 *lmv;
2368 /* sobj and tobj are the same */
2372 if (fid_is_root(mdt_object_fid(sobj)))
2375 if (fid_is_root(mdt_object_fid(tobj)))
2378 /* check whether sobj is child of tobj */
2379 rc = mdo_is_subdir(info->mti_env, mdt_object_child(sobj),
2380 mdt_object_fid(tobj));
2387 /* check whether sobj and tobj are children of the same parent */
2388 rc = mdt_attr_get_pfid(info, sobj, spfid);
2392 rc = mdt_attr_get_pfid(info, tobj, tpfid);
2396 if (!lu_fid_eq(spfid, tpfid))
2399 /* check whether sobj and tobj are sibling stripes */
2400 rc = mdt_stripe_get(info, sobj, ma, XATTR_NAME_LMV);
2404 if (!(ma->ma_valid & MA_LMV))
2407 lmv = &ma->ma_lmv->lmv_md_v1;
2408 if (!(le32_to_cpu(lmv->lmv_magic) & LMV_MAGIC_STRIPE))
2410 sindex = le32_to_cpu(lmv->lmv_master_mdt_index);
2413 rc = mdt_stripe_get(info, tobj, ma, XATTR_NAME_LMV);
2417 if (!(ma->ma_valid & MA_LMV))
2420 lmv = &ma->ma_lmv->lmv_md_v1;
2421 if (!(le32_to_cpu(lmv->lmv_magic) & LMV_MAGIC_STRIPE))
2423 tindex = le32_to_cpu(lmv->lmv_master_mdt_index);
2425 /* check stripe index of sobj and tobj */
2426 if (sindex == tindex)
2429 return sindex < tindex ? 0 : 1;
2433 * VBR: rename versions in reply: 0 - srcdir parent; 1 - tgtdir parent;
2434 * 2 - srcdir child; 3 - tgtdir child.
2435 * Update on disk version of srcdir child.
2437 static int mdt_reint_rename(struct mdt_thread_info *info,
2438 struct mdt_lock_handle *unused)
2440 struct mdt_device *mdt = info->mti_mdt;
2441 struct mdt_reint_record *rr = &info->mti_rr;
2442 struct md_attr *ma = &info->mti_attr;
2443 struct ptlrpc_request *req = mdt_info_req(info);
2444 struct mdt_object *msrcdir = NULL;
2445 struct mdt_object *mtgtdir = NULL;
2446 struct mdt_object *mold;
2447 struct mdt_object *mnew = NULL;
2448 struct lustre_handle rename_lh = { 0 };
2449 struct mdt_lock_handle *lh_srcdirp;
2450 struct mdt_lock_handle *lh_tgtdirp;
2451 struct mdt_lock_handle *lh_oldp = NULL;
2452 struct mdt_lock_handle *lh_newp = NULL;
2453 struct lu_fid *old_fid = &info->mti_tmp_fid1;
2454 struct lu_fid *new_fid = &info->mti_tmp_fid2;
2456 bool reverse = false, discard = false;
2461 DEBUG_REQ(D_INODE, req, "rename "DFID"/"DNAME" to "DFID"/"DNAME,
2462 PFID(rr->rr_fid1), PNAME(&rr->rr_name),
2463 PFID(rr->rr_fid2), PNAME(&rr->rr_tgt_name));
2465 if (info->mti_dlm_req)
2466 ldlm_request_cancel(req, info->mti_dlm_req, 0, LATF_SKIP);
2468 if (!fid_is_md_operative(rr->rr_fid1) ||
2469 !fid_is_md_operative(rr->rr_fid2))
2472 /* find both parents. */
2473 msrcdir = mdt_parent_find_check(info, rr->rr_fid1, 0);
2474 if (IS_ERR(msrcdir))
2475 RETURN(PTR_ERR(msrcdir));
2477 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_RENAME3, 5);
2479 if (lu_fid_eq(rr->rr_fid1, rr->rr_fid2)) {
2481 mdt_object_get(info->mti_env, mtgtdir);
2483 mtgtdir = mdt_parent_find_check(info, rr->rr_fid2, 1);
2484 if (IS_ERR(mtgtdir))
2485 GOTO(out_put_srcdir, rc = PTR_ERR(mtgtdir));
2489 * Note: do not enqueue rename lock for replay request, because
2490 * if other MDT holds rename lock, but being blocked to wait for
2491 * this MDT to finish its recovery, and the failover MDT can not
2492 * get rename lock, which will cause deadlock.
2494 if (!req_is_replay(req)) {
2496 * Normally rename RPC is handled on the MDT with the target
2497 * directory (if target exists, it's on the MDT with the
2498 * target), if the source directory is remote, it's a hint that
2499 * source is remote too (this may not be true, but it won't
2500 * cause any issue), return -EXDEV early to avoid taking
2503 if (!mdt->mdt_enable_remote_rename &&
2504 mdt_object_remote(msrcdir))
2505 GOTO(out_put_tgtdir, rc = -EXDEV);
2507 rc = mdt_rename_lock(info, &rename_lh);
2509 CERROR("%s: can't lock FS for rename: rc = %d\n",
2510 mdt_obd_name(mdt), rc);
2511 GOTO(out_put_tgtdir, rc);
2515 rc = mdt_rename_determine_lock_order(info, msrcdir, mtgtdir);
2517 GOTO(out_unlock_rename, rc);
2521 /* source needs to be looked up after locking source parent, otherwise
2522 * this rename may race with unlink source, and cause rename hang, see
2523 * sanityn.sh 55b, so check parents first, if later we found source is
2524 * remote, relock parents. */
2525 cos_incompat = (mdt_object_remote(msrcdir) ||
2526 mdt_object_remote(mtgtdir));
2528 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_RENAME4, 5);
2530 /* lock parents in the proper order. */
2531 lh_srcdirp = &info->mti_lh[MDT_LH_PARENT];
2532 lh_tgtdirp = &info->mti_lh[MDT_LH_CHILD];
2535 mdt_lock_pdo_init(lh_srcdirp, LCK_PW, &rr->rr_name);
2536 mdt_lock_pdo_init(lh_tgtdirp, LCK_PW, &rr->rr_tgt_name);
2539 rc = mdt_object_lock_save(info, mtgtdir, lh_tgtdirp, 1,
2542 GOTO(out_unlock_rename, rc);
2544 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_RENAME, 5);
2546 rc = mdt_object_lock_save(info, msrcdir, lh_srcdirp, 0,
2549 mdt_object_unlock(info, mtgtdir, lh_tgtdirp, rc);
2550 GOTO(out_unlock_rename, rc);
2553 rc = mdt_object_lock_save(info, msrcdir, lh_srcdirp, 0,
2556 GOTO(out_unlock_rename, rc);
2558 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_RENAME, 5);
2560 if (mtgtdir != msrcdir) {
2561 rc = mdt_object_lock_save(info, mtgtdir, lh_tgtdirp, 1,
2563 } else if (!mdt_object_remote(mtgtdir) &&
2564 lh_srcdirp->mlh_pdo_hash !=
2565 lh_tgtdirp->mlh_pdo_hash) {
2566 rc = mdt_pdir_hash_lock(info, lh_tgtdirp, mtgtdir,
2567 MDS_INODELOCK_UPDATE,
2569 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_PDO_LOCK2, 10);
2572 mdt_object_unlock(info, msrcdir, lh_srcdirp, rc);
2573 GOTO(out_unlock_rename, rc);
2577 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_RENAME4, 5);
2578 OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_RENAME2, 5);
2580 /* find mold object. */
2582 rc = mdt_lookup_version_check(info, msrcdir, &rr->rr_name, old_fid, 2);
2584 GOTO(out_unlock_parents, rc);
2586 if (lu_fid_eq(old_fid, rr->rr_fid1) || lu_fid_eq(old_fid, rr->rr_fid2))
2587 GOTO(out_unlock_parents, rc = -EINVAL);
2589 if (!fid_is_md_operative(old_fid))
2590 GOTO(out_unlock_parents, rc = -EPERM);
2592 mold = mdt_object_find(info->mti_env, info->mti_mdt, old_fid);
2594 GOTO(out_unlock_parents, rc = PTR_ERR(mold));
2596 if (!mdt_object_exists(mold)) {
2597 LU_OBJECT_DEBUG(D_INODE, info->mti_env,
2599 "object does not exist");
2600 GOTO(out_put_old, rc = -ENOENT);
2603 if (mdt_object_remote(mold) && !mdt->mdt_enable_remote_rename)
2604 GOTO(out_put_old, rc = -EXDEV);
2606 /* Check if @mtgtdir is subdir of @mold, before locking child
2607 * to avoid reverse locking. */
2608 if (mtgtdir != msrcdir) {
2609 rc = mdo_is_subdir(info->mti_env, mdt_object_child(mtgtdir),
2614 GOTO(out_put_old, rc);
2618 tgt_vbr_obj_set(info->mti_env, mdt_obj2dt(mold));
2619 /* save version after locking */
2620 mdt_version_get_save(info, mold, 2);
2622 if (!cos_incompat && mdt_object_remote(mold)) {
2623 cos_incompat = true;
2624 mdt_object_put(info->mti_env, mold);
2625 mdt_object_unlock(info, mtgtdir, lh_tgtdirp, -EAGAIN);
2626 mdt_object_unlock(info, msrcdir, lh_srcdirp, -EAGAIN);
2630 /* find mnew object:
2631 * mnew target object may not exist now
2632 * lookup with version checking */
2634 rc = mdt_lookup_version_check(info, mtgtdir, &rr->rr_tgt_name, new_fid,
2637 /* the new_fid should have been filled at this moment */
2638 if (lu_fid_eq(old_fid, new_fid))
2639 GOTO(out_put_old, rc);
2641 if (lu_fid_eq(new_fid, rr->rr_fid1) ||
2642 lu_fid_eq(new_fid, rr->rr_fid2))
2643 GOTO(out_put_old, rc = -EINVAL);
2645 if (!fid_is_md_operative(new_fid))
2646 GOTO(out_put_old, rc = -EPERM);
2648 mnew = mdt_object_find(info->mti_env, info->mti_mdt, new_fid);
2650 GOTO(out_put_old, rc = PTR_ERR(mnew));
2652 if (!mdt_object_exists(mnew)) {
2653 LU_OBJECT_DEBUG(D_INODE, info->mti_env,
2655 "object does not exist");
2656 GOTO(out_put_new, rc = -ENOENT);
2659 if (mdt_object_remote(mnew)) {
2660 struct mdt_body *repbody;
2662 /* Always send rename req to the target child MDT */
2663 repbody = req_capsule_server_get(info->mti_pill,
2665 LASSERT(repbody != NULL);
2666 repbody->mbo_fid1 = *new_fid;
2667 repbody->mbo_valid |= (OBD_MD_FLID | OBD_MD_MDS);
2668 GOTO(out_put_new, rc = -EXDEV);
2670 /* Before locking the target dir, check we do not replace
2671 * a dir with a non-dir, otherwise it may deadlock with
2672 * link op which tries to create a link in this dir
2673 * back to this non-dir. */
2674 if (S_ISDIR(lu_object_attr(&mnew->mot_obj)) &&
2675 !S_ISDIR(lu_object_attr(&mold->mot_obj)))
2676 GOTO(out_put_new, rc = -EISDIR);
2678 lh_oldp = &info->mti_lh[MDT_LH_OLD];
2679 mdt_lock_reg_init(lh_oldp, LCK_EX);
2680 lock_ibits = MDS_INODELOCK_LOOKUP | MDS_INODELOCK_XATTR;
2681 if (mdt_object_remote(msrcdir)) {
2682 /* Enqueue lookup lock from the parent MDT */
2683 rc = mdt_remote_object_lock(info, msrcdir,
2684 mdt_object_fid(mold),
2685 &lh_oldp->mlh_rreg_lh,
2686 lh_oldp->mlh_rreg_mode,
2687 MDS_INODELOCK_LOOKUP,
2690 GOTO(out_put_new, rc);
2692 lock_ibits &= ~MDS_INODELOCK_LOOKUP;
2695 rc = mdt_reint_object_lock(info, mold, lh_oldp, lock_ibits,
2698 GOTO(out_unlock_old, rc);
2700 /* Check if @msrcdir is subdir of @mnew, before locking child
2701 * to avoid reverse locking. */
2702 if (mtgtdir != msrcdir) {
2703 rc = mdo_is_subdir(info->mti_env,
2704 mdt_object_child(msrcdir), new_fid);
2708 GOTO(out_unlock_old, rc);
2712 /* We used to acquire MDS_INODELOCK_FULL here but we
2713 * can't do this now because a running HSM restore on
2714 * the rename onto victim will hold the layout
2715 * lock. See LU-4002. */
2717 lh_newp = &info->mti_lh[MDT_LH_NEW];
2718 mdt_lock_reg_init(lh_newp, LCK_EX);
2719 rc = mdt_reint_object_lock(info, mnew, lh_newp,
2720 MDS_INODELOCK_LOOKUP |
2721 MDS_INODELOCK_UPDATE,
2724 GOTO(out_unlock_old, rc);
2726 /* get and save version after locking */
2727 mdt_version_get_save(info, mnew, 3);
2728 } else if (rc != -EREMOTE && rc != -ENOENT) {
2729 GOTO(out_put_old, rc);
2731 lh_oldp = &info->mti_lh[MDT_LH_OLD];
2732 mdt_lock_reg_init(lh_oldp, LCK_EX);
2733 lock_ibits = MDS_INODELOCK_LOOKUP | MDS_INODELOCK_XATTR;
2734 if (mdt_object_remote(msrcdir)) {
2735 /* Enqueue lookup lock from the parent MDT */
2736 rc = mdt_remote_object_lock(info, msrcdir,
2737 mdt_object_fid(mold),
2738 &lh_oldp->mlh_rreg_lh,
2739 lh_oldp->mlh_rreg_mode,
2740 MDS_INODELOCK_LOOKUP,
2743 GOTO(out_put_old, rc);
2745 lock_ibits &= ~MDS_INODELOCK_LOOKUP;
2748 rc = mdt_reint_object_lock(info, mold, lh_oldp, lock_ibits,
2751 GOTO(out_unlock_old, rc);
2753 mdt_enoent_version_save(info, 3);
2756 /* step 5: rename it */
2757 mdt_reint_init_ma(info, ma);
2759 mdt_fail_write(info->mti_env, info->mti_mdt->mdt_bottom,
2760 OBD_FAIL_MDS_REINT_RENAME_WRITE);
2763 mutex_lock(&mnew->mot_lov_mutex);
2765 rc = mdo_rename(info->mti_env, mdt_object_child(msrcdir),
2766 mdt_object_child(mtgtdir), old_fid, &rr->rr_name,
2767 mnew != NULL ? mdt_object_child(mnew) : NULL,
2768 &rr->rr_tgt_name, ma);
2771 mutex_unlock(&mnew->mot_lov_mutex);
2773 /* handle last link of tgt object */
2775 mdt_counter_incr(req, LPROC_MDT_RENAME);
2777 mdt_handle_last_unlink(info, mnew, ma);
2778 discard = mdt_dom_check_for_discard(info, mnew);
2780 mdt_rename_counter_tally(info, info->mti_mdt, req,
2786 mdt_object_unlock(info, mnew, lh_newp, rc);
2788 mdt_object_unlock(info, mold, lh_oldp, rc);
2790 if (mnew && !discard)
2791 mdt_object_put(info->mti_env, mnew);
2793 mdt_object_put(info->mti_env, mold);
2795 mdt_object_unlock(info, mtgtdir, lh_tgtdirp, rc);
2796 mdt_object_unlock(info, msrcdir, lh_srcdirp, rc);
2798 if (lustre_handle_is_used(&rename_lh))
2799 mdt_rename_unlock(&rename_lh);
2801 mdt_object_put(info->mti_env, mtgtdir);
2803 mdt_object_put(info->mti_env, msrcdir);
2805 /* The DoM discard can be done right in the place above where it is
2806 * assigned, meanwhile it is done here after rename unlock due to
2807 * compatibility with old clients, for them the discard blocks
2808 * the main thread until completion. Check LU-11359 for details.
2811 mdt_dom_discard_data(info, mnew);
2812 mdt_object_put(info->mti_env, mnew);
2817 static int mdt_reint_resync(struct mdt_thread_info *info,
2818 struct mdt_lock_handle *lhc)
2820 struct mdt_reint_record *rr = &info->mti_rr;
2821 struct ptlrpc_request *req = mdt_info_req(info);
2822 struct md_attr *ma = &info->mti_attr;
2823 struct mdt_object *mo;
2824 struct ldlm_lock *lease;
2825 struct mdt_body *repbody;
2826 struct md_layout_change layout = { .mlc_mirror_id = rr->rr_mirror_id };
2832 DEBUG_REQ(D_INODE, req, DFID", FLR file resync", PFID(rr->rr_fid1));
2834 if (info->mti_dlm_req)
2835 ldlm_request_cancel(req, info->mti_dlm_req, 0, LATF_SKIP);
2837 mo = mdt_object_find(info->mti_env, info->mti_mdt, rr->rr_fid1);
2839 GOTO(out, rc = PTR_ERR(mo));
2841 if (!mdt_object_exists(mo))
2842 GOTO(out_obj, rc = -ENOENT);
2844 if (!S_ISREG(lu_object_attr(&mo->mot_obj)))
2845 GOTO(out_obj, rc = -EINVAL);
2847 if (mdt_object_remote(mo))
2848 GOTO(out_obj, rc = -EREMOTE);
2850 lease = ldlm_handle2lock(rr->rr_lease_handle);
2852 GOTO(out_obj, rc = -ESTALE);
2854 /* It's really necessary to grab open_sem and check if the lease lock
2855 * has been lost. There would exist a concurrent writer coming in and
2856 * generating some dirty data in memory cache, the writeback would fail
2857 * after the layout version is increased by MDS_REINT_RESYNC RPC. */
2858 if (!down_write_trylock(&mo->mot_open_sem))
2859 GOTO(out_put_lease, rc = -EBUSY);
2861 lock_res_and_lock(lease);
2862 lease_broken = ldlm_is_cancel(lease);
2863 unlock_res_and_lock(lease);
2865 GOTO(out_unlock, rc = -EBUSY);
2867 /* the file has yet opened by anyone else after we took the lease. */
2868 layout.mlc_opc = MD_LAYOUT_RESYNC;
2869 lhc = &info->mti_lh[MDT_LH_LOCAL];
2870 rc = mdt_layout_change(info, mo, lhc, &layout);
2872 GOTO(out_unlock, rc);
2874 mdt_object_unlock(info, mo, lhc, 0);
2876 ma->ma_need = MA_INODE;
2878 rc = mdt_attr_get_complex(info, mo, ma);
2880 GOTO(out_unlock, rc);
2882 repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY);
2883 mdt_pack_attr2body(info, repbody, &ma->ma_attr, mdt_object_fid(mo));
2887 up_write(&mo->mot_open_sem);
2889 LDLM_LOCK_PUT(lease);
2891 mdt_object_put(info->mti_env, mo);
2893 mdt_client_compatibility(info);
2894 rc2 = mdt_fix_reply(info);
2900 struct mdt_reinter {
2901 int (*mr_handler)(struct mdt_thread_info *, struct mdt_lock_handle *);
2902 enum lprocfs_extra_opc mr_extra_opc;
2905 static const struct mdt_reinter mdt_reinters[] = {
2907 .mr_handler = &mdt_reint_setattr,
2908 .mr_extra_opc = MDS_REINT_SETATTR,
2911 .mr_handler = &mdt_reint_create,
2912 .mr_extra_opc = MDS_REINT_CREATE,
2915 .mr_handler = &mdt_reint_link,
2916 .mr_extra_opc = MDS_REINT_LINK,
2919 .mr_handler = &mdt_reint_unlink,
2920 .mr_extra_opc = MDS_REINT_UNLINK,
2923 .mr_handler = &mdt_reint_rename,
2924 .mr_extra_opc = MDS_REINT_RENAME,
2927 .mr_handler = &mdt_reint_open,
2928 .mr_extra_opc = MDS_REINT_OPEN,
2930 [REINT_SETXATTR] = {
2931 .mr_handler = &mdt_reint_setxattr,
2932 .mr_extra_opc = MDS_REINT_SETXATTR,
2935 .mr_handler = &mdt_reint_unlink,
2936 .mr_extra_opc = MDS_REINT_UNLINK,
2939 .mr_handler = &mdt_reint_migrate,
2940 .mr_extra_opc = MDS_REINT_RENAME,
2943 .mr_handler = &mdt_reint_resync,
2944 .mr_extra_opc = MDS_REINT_RESYNC,
2948 int mdt_reint_rec(struct mdt_thread_info *info,
2949 struct mdt_lock_handle *lhc)
2951 const struct mdt_reinter *mr;
2955 if (!(info->mti_rr.rr_opcode < ARRAY_SIZE(mdt_reinters)))
2958 mr = &mdt_reinters[info->mti_rr.rr_opcode];
2959 if (mr->mr_handler == NULL)
2962 rc = (*mr->mr_handler)(info, lhc);
2964 lprocfs_counter_incr(ptlrpc_req2svc(mdt_info_req(info))->srv_stats,
2965 PTLRPC_LAST_CNTR + mr->mr_extra_opc);