4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Whamcloud, Inc.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/mdd/mdd_orphans.c
38 * Orphan handling code
40 * Author: Mike Pershin <tappro@clusterfs.com>
41 * Pravin B Shelar <pravin.shelar@sun.com>
45 # define EXPORT_SYMTAB
47 #define DEBUG_SUBSYSTEM S_MDS
50 #include <obd_class.h>
51 #include <lustre_ver.h>
52 #include <obd_support.h>
53 #include <lustre_fid.h>
54 #include "mdd_internal.h"
56 const char orph_index_name[] = "PENDING";
57 const char *dotdot = "..";
64 #define ORPHAN_FILE_NAME_FORMAT "%016llx:%08x:%08x:%2x"
65 #define ORPHAN_FILE_NAME_FORMAT_18 "%llx:%08x"
67 static struct dt_key* orph_key_fill(const struct lu_env *env,
68 const struct lu_fid *lf, __u32 op)
70 char *key = mdd_env_info(env)->mti_orph_key;
74 rc = snprintf(key, NAME_MAX + 1, ORPHAN_FILE_NAME_FORMAT,
75 (long long unsigned int)fid_seq(lf),
76 fid_oid(lf), fid_ver(lf), op);
78 return (struct dt_key*) key;
83 static struct dt_key* orph_key_fill_18(const struct lu_env *env,
84 const struct lu_fid *lf)
86 char *key = mdd_env_info(env)->mti_orph_key;
90 rc = snprintf(key, NAME_MAX + 1, ORPHAN_FILE_NAME_FORMAT_18,
91 (unsigned long long)fid_seq(lf), fid_oid(lf));
93 return (struct dt_key*) key;
98 static int orphan_key_to_fid(char *key, struct lu_fid *lf)
103 rc = sscanf(key, ORPHAN_FILE_NAME_FORMAT,
104 (long long unsigned int *)&lf->f_seq, &lf->f_oid,
110 rc = sscanf(key, ORPHAN_FILE_NAME_FORMAT_18,
111 (long long unsigned int *)&lf->f_seq, &lf->f_oid);
117 CERROR("can not parse orphan file name %s\n", key);
121 static inline void mdd_orphan_write_lock(const struct lu_env *env,
122 struct mdd_device *mdd)
125 struct dt_object *dor = mdd->mdd_orphans;
126 dor->do_ops->do_write_lock(env, dor, MOR_TGT_ORPHAN);
129 static inline void mdd_orphan_write_unlock(const struct lu_env *env,
130 struct mdd_device *mdd)
133 struct dt_object *dor = mdd->mdd_orphans;
134 dor->do_ops->do_write_unlock(env, dor);
137 static inline int mdd_orphan_insert_obj(const struct lu_env *env,
138 struct mdd_device *mdd,
139 struct mdd_object *obj,
143 struct dt_object *dor = mdd->mdd_orphans;
144 const struct lu_fid *lf = mdo2fid(obj);
145 struct dt_key *key = orph_key_fill(env, lf, op);
148 return dor->do_index_ops->dio_insert(env, dor,
154 static inline int mdd_orphan_delete_obj(const struct lu_env *env,
155 struct mdd_device *mdd ,
159 struct dt_object *dor = mdd->mdd_orphans;
161 return dor->do_index_ops->dio_delete(env, dor,
166 static inline void mdd_orphan_ref_add(const struct lu_env *env,
167 struct mdd_device *mdd,
170 struct dt_object *dor = mdd->mdd_orphans;
171 dor->do_ops->do_ref_add(env, dor, th);
174 static inline void mdd_orphan_ref_del(const struct lu_env *env,
175 struct mdd_device *mdd,
178 struct dt_object *dor = mdd->mdd_orphans;
179 dor->do_ops->do_ref_del(env, dor, th);
183 int orph_declare_index_insert(const struct lu_env *env,
184 struct mdd_object *obj,
187 struct mdd_device *mdd = mdo2mdd(&obj->mod_obj);
190 rc = dt_declare_insert(env, mdd->mdd_orphans, NULL, NULL, th);
194 rc = mdo_declare_ref_add(env, obj, th);
198 if (!S_ISDIR(mdd_object_type(obj)))
201 rc = mdo_declare_ref_add(env, obj, th);
205 rc = dt_declare_ref_add(env, mdd->mdd_orphans, th);
209 rc = mdo_declare_index_delete(env, obj, dotdot, th);
213 rc = mdo_declare_index_insert(env, obj, NULL, dotdot, th);
218 static int orph_index_insert(const struct lu_env *env,
219 struct mdd_object *obj,
223 struct mdd_device *mdd = mdo2mdd(&obj->mod_obj);
224 struct dt_object *dor = mdd->mdd_orphans;
225 const struct lu_fid *lf_dor = lu_object_fid(&dor->do_lu);
226 struct dt_object *next = mdd_object_child(obj);
230 LASSERT(mdd_write_locked(env, obj) != 0);
231 LASSERT(!(obj->mod_flags & ORPHAN_OBJ));
232 LASSERT(obj->mod_count > 0);
234 mdd_orphan_write_lock(env, mdd);
236 rc = mdd_orphan_insert_obj(env, mdd, obj, op, th);
240 mdo_ref_add(env, obj, th);
241 if (!S_ISDIR(mdd_object_type(obj)))
244 mdo_ref_add(env, obj, th);
245 mdd_orphan_ref_add(env, mdd, th);
247 /* try best to fixup directory, dont return errors
249 if (!dt_try_as_dir(env, next))
251 next->do_index_ops->dio_delete(env, next,
252 (const struct dt_key *)dotdot,
255 next->do_index_ops->dio_insert(env, next,
256 (struct dt_rec *)lf_dor,
257 (const struct dt_key *)dotdot,
262 obj->mod_flags |= ORPHAN_OBJ;
264 mdd_orphan_write_unlock(env, mdd);
270 * Destroy OSD object on MDD and associated OST objects.
272 * \param obj orphan object
273 * \param mdd used for sending llog msg to osts
278 static int orphan_object_kill(const struct lu_env *env,
279 struct mdd_object *obj,
280 struct mdd_device *mdd,
283 struct lu_attr *la = &mdd_env_info(env)->mti_la;
287 /* No need to lock this object as its recovery phase, and
288 * no other thread can access it. But we need to lock it
289 * as its precondition for osd api we using. */
291 mdo_ref_del(env, obj, th);
292 if (S_ISDIR(mdd_object_type(obj))) {
293 mdo_ref_del(env, obj, th);
294 mdd_orphan_ref_del(env, mdd, th);
296 /* regular file , cleanup linked ost objects */
297 rc = mdd_la_get(env, obj, la, BYPASS_CAPA);
299 rc = mdd_lov_destroy(env, mdd, obj, la);
301 mdo_destroy(env, obj, th);
305 int orph_declare_index_delete(const struct lu_env *env,
306 struct mdd_object *obj,
309 struct mdd_device *mdd = mdo2mdd(&obj->mod_obj);
312 rc = dt_declare_delete(env, mdd->mdd_orphans, NULL, th);
316 rc = mdo_declare_ref_del(env, obj, th);
320 if (S_ISDIR(mdd_object_type(obj))) {
321 rc = mdo_declare_ref_del(env, obj, th);
325 rc = dt_declare_ref_del(env, mdd->mdd_orphans, th);
331 static int orph_index_delete(const struct lu_env *env,
332 struct mdd_object *obj,
336 struct mdd_device *mdd = mdo2mdd(&obj->mod_obj);
337 struct dt_object *dor = mdd->mdd_orphans;
343 LASSERT(mdd_write_locked(env, obj) != 0);
344 LASSERT(obj->mod_flags & ORPHAN_OBJ);
345 LASSERT(obj->mod_count == 0);
349 key = orph_key_fill(env, mdo2fid(obj), op);
350 mdd_orphan_write_lock(env, mdd);
352 rc = mdd_orphan_delete_obj(env, mdd, key, th);
355 key = orph_key_fill_18(env, mdo2fid(obj));
356 rc = mdd_orphan_delete_obj(env, mdd, key, th);
360 /* lov objects will be destroyed by caller */
361 mdo_ref_del(env, obj, th);
362 if (S_ISDIR(mdd_object_type(obj))) {
363 mdo_ref_del(env, obj, th);
364 mdd_orphan_ref_del(env, mdd, th);
366 obj->mod_flags &= ~ORPHAN_OBJ;
368 CERROR("could not delete object: rc = %d\n",rc);
371 mdd_orphan_write_unlock(env, mdd);
376 static int orphan_object_destroy(const struct lu_env *env,
377 struct mdd_object *obj,
380 struct thandle *th = NULL;
381 struct mdd_device *mdd = mdo2mdd(&obj->mod_obj);
382 struct md_attr *ma = &mdd_env_info(env)->mti_ma;
387 ma->ma_lmm_size = mdd_lov_mdsize(env, mdd);
388 ma->ma_lmm = mdd_max_lmm_get(env, mdd);
389 ma->ma_cookie_size = mdd_lov_cookiesize(env, mdd);
390 ma->ma_cookie = mdd_max_cookie_get(env, mdd);
391 ma->ma_need = MA_INODE | MA_LOV | MA_COOKIE;
394 th = mdd_trans_create(env, mdd);
396 CERROR("Cannot get thandle\n");
399 rc = orph_declare_index_delete(env, obj, th);
403 rc = mdd_declare_object_kill(env, obj, ma, th);
407 rc = mdd_trans_start(env, mdd, th);
411 mdd_write_lock(env, obj, MOR_TGT_CHILD);
412 if (likely(obj->mod_count == 0)) {
413 mdd_orphan_write_lock(env, mdd);
414 rc = mdd_orphan_delete_obj(env, mdd, key, th);
416 orphan_object_kill(env, obj, mdd, th);
418 CERROR("could not delete object: rc = %d\n",rc);
419 mdd_orphan_write_unlock(env, mdd);
421 mdd_write_unlock(env, obj);
424 mdd_trans_stop(env, mdd, 0, th);
430 * Delete unused orphan with FID \a lf from PENDING directory
432 * \param mdd MDD device finishing recovery
433 * \param lf FID of file or directory to delete
434 * \param key cookie for this entry in index iterator
439 static int orph_key_test_and_del(const struct lu_env *env,
440 struct mdd_device *mdd,
444 struct mdd_object *mdo;
447 mdo = mdd_object_find(env, mdd, lf);
453 if (mdo->mod_count == 0) {
454 CDEBUG(D_HA, "Found orphan "DFID", delete it\n", PFID(lf));
455 rc = orphan_object_destroy(env, mdo, key);
456 if (rc) /* so replay-single.sh test_37 works */
457 CERROR("%s: error unlinking orphan "DFID" from "
458 "PENDING: rc = %d\n",
459 mdd->mdd_obd_dev->obd_name, PFID(lf), rc);
461 mdd_write_lock(env, mdo, MOR_TGT_CHILD);
462 if (likely(mdo->mod_count > 0)) {
463 CDEBUG(D_HA, "Found orphan "DFID" count %d, skip it\n",
464 PFID(lf), mdo->mod_count);
465 mdo->mod_flags |= ORPHAN_OBJ;
467 mdd_write_unlock(env, mdo);
470 mdd_object_put(env, mdo);
475 * delete unreferenced files and directories in the PENDING directory
477 * Files that remain in PENDING after client->MDS recovery has completed
478 * have to be referenced (opened) by some client during recovery, or they
479 * will be deleted here (for clients that did not complete recovery).
481 * \param mdd MDD device finishing recovery
486 static int orph_index_iterate(const struct lu_env *env,
487 struct mdd_device *mdd)
489 struct dt_object *dor = mdd->mdd_orphans;
490 char *mti_key = mdd_env_info(env)->mti_orph_key;
491 const struct dt_it_ops *iops;
501 /* In recovery phase, do not need for any lock here */
502 iops = &dor->do_index_ops->dio_it;
503 it = iops->init(env, dor, LUDA_64BITHASH, BYPASS_CAPA);
506 CERROR("%s: cannot clean PENDING: rc = %d\n",
507 mdd->mdd_obd_dev->obd_name, rc);
511 rc = iops->load(env, it, 0);
515 CERROR("%s: error loading iterator to clean PENDING\n",
516 mdd->mdd_obd_dev->obd_name);
517 /* Index contains no zero key? */
518 GOTO(out_put, rc = -EIO);
522 key = (void *)iops->key(env, it);
524 CERROR("%s: key failed when clean PENDING: rc = %ld\n",
525 mdd->mdd_obd_dev->obd_name, PTR_ERR(key));
528 key_sz = iops->key_size(env, it);
530 /* filter out "." and ".." entries from PENDING dir. */
534 memcpy(mti_key, key, key_sz);
537 if (orphan_key_to_fid(mti_key, &fid))
539 if (!fid_is_sane(&fid)) {
540 CERROR("%s: bad FID "DFID" cleaning PENDING\n",
541 mdd->mdd_obd_dev->obd_name, PFID(&fid));
545 /* kill orphan object */
546 cookie = iops->store(env, it);
548 rc = orph_key_test_and_del(env, mdd, &fid,
549 (struct dt_key *)mti_key);
551 /* after index delete reset iterator */
553 result = iops->get(env, it, (const void *)"");
555 result = iops->load(env, it, cookie);
557 result = iops->next(env, it);
558 } while (result == 0);
560 GOTO(out_put, rc = 0);
570 * open the PENDING directory for device \a mdd
572 * The PENDING directory persistently tracks files and directories that were
573 * unlinked from the namespace (nlink == 0) but are still held open by clients.
574 * Those inodes shouldn't be deleted if the MDS crashes, because the clients
575 * would not be able to recover and reopen those files. Instead, these inodes
576 * are linked into the PENDING directory on disk, and only deleted if all
577 * clients close them, or the MDS finishes client recovery without any client
578 * reopening them (i.e. former clients didn't join recovery).
579 * \param d mdd device being started.
582 * \retval -ve index operation error.
585 int orph_index_init(const struct lu_env *env, struct mdd_device *mdd)
592 d = dt_store_open(env, mdd->mdd_child, "", orph_index_name, &fid);
594 mdd->mdd_orphans = d;
595 if (!dt_try_as_dir(env, d)) {
597 CERROR("\"%s\" is not an index! : rc = %d\n",
598 orph_index_name, rc);
601 CERROR("cannot find \"%s\" obj %d\n",
602 orph_index_name, (int)PTR_ERR(d));
609 void orph_index_fini(const struct lu_env *env, struct mdd_device *mdd)
612 if (mdd->mdd_orphans != NULL) {
613 lu_object_put(env, &mdd->mdd_orphans->do_lu);
614 mdd->mdd_orphans = NULL;
620 * Iterate orphan index to cleanup orphan objects after recovery is done.
621 * \param d mdd device in recovery.
623 int __mdd_orphan_cleanup(const struct lu_env *env, struct mdd_device *d)
625 return orph_index_iterate(env, d);
629 * add an orphan \a obj to the orphan index.
630 * \param obj file or directory.
631 * \param th transaction for index insert.
633 * \pre obj nlink == 0 && obj->mod_count != 0
636 * \retval -ve index operation error.
638 int __mdd_orphan_add(const struct lu_env *env,
639 struct mdd_object *obj, struct thandle *th)
641 return orph_index_insert(env, obj, ORPH_OP_UNLINK, th);
645 * delete an orphan \a obj from orphan index.
646 * \param obj file or directory.
647 * \param th transaction for index deletion and object destruction.
649 * \pre obj->mod_count == 0 && ORPHAN_OBJ is set for obj.
652 * \retval -ve index operation error.
654 int __mdd_orphan_del(const struct lu_env *env,
655 struct mdd_object *obj, struct thandle *th)
657 return orph_index_delete(env, obj, ORPH_OP_UNLINK, th);