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_device.c
38 * Lustre Metadata Server (mdd) routines
40 * Author: Wang Di <wangdi@clusterfs.com>
43 #define DEBUG_SUBSYSTEM S_MDS
45 #include <obd_class.h>
46 #include <lprocfs_status.h>
47 #include <lustre_fid.h>
48 #include <lustre_mds.h>
49 #include <lustre_disk.h> /* for changelogs */
50 #include <lustre_param.h>
51 #include <lustre_fid.h>
53 #include "mdd_internal.h"
55 const struct md_device_operations mdd_ops;
56 static struct lu_device_type mdd_device_type;
58 static const char mdd_root_dir_name[] = "ROOT";
59 static const char mdd_obf_dir_name[] = "fid";
61 /* Slab for MDD object allocation */
62 cfs_mem_cache_t *mdd_object_kmem;
64 static struct lu_kmem_descr mdd_caches[] = {
66 .ckd_cache = &mdd_object_kmem,
67 .ckd_name = "mdd_obj",
68 .ckd_size = sizeof(struct mdd_object)
75 static int mdd_connect_to_next(const struct lu_env *env, struct mdd_device *m,
78 struct obd_connect_data *data = NULL;
79 struct lu_device *lud = mdd2lu_dev(m);
80 struct obd_device *obd;
84 LASSERT(m->mdd_child_exp == NULL);
86 OBD_ALLOC(data, sizeof(*data));
88 GOTO(out, rc = -ENOMEM);
90 obd = class_name2obd(nextdev);
92 CERROR("can't locate next device: %s\n", nextdev);
93 GOTO(out, rc = -ENOTCONN);
96 data->ocd_connect_flags = OBD_CONNECT_VERSION;
97 data->ocd_version = LUSTRE_VERSION_CODE;
99 rc = obd_connect(NULL, &m->mdd_child_exp, obd, &obd->obd_uuid, data, NULL);
101 CERROR("cannot connect to next dev %s (%d)\n", nextdev, rc);
105 lud->ld_site = m->mdd_child_exp->exp_obd->obd_lu_dev->ld_site;
106 LASSERT(lud->ld_site);
107 m->mdd_child = lu2dt_dev(m->mdd_child_exp->exp_obd->obd_lu_dev);
108 lu_dev_add_linkage(lud->ld_site, lud);
112 OBD_FREE(data, sizeof(*data));
116 static int mdd_init0(const struct lu_env *env, struct mdd_device *mdd,
117 struct lu_device_type *t, struct lustre_cfg *lcfg)
122 mdd->mdd_md_dev.md_lu_dev.ld_ops = &mdd_lu_ops;
123 mdd->mdd_md_dev.md_ops = &mdd_ops;
125 rc = mdd_connect_to_next(env, mdd, lustre_cfg_string(lcfg, 3));
129 mdd->mdd_atime_diff = MAX_ATIME_DIFF;
130 /* sync permission changes */
131 mdd->mdd_sync_permission = 1;
133 dt_conf_get(env, mdd->mdd_child, &mdd->mdd_dt_conf);
135 /* we are using service name but not mdd obd name
136 * for compatibility reasons.
137 * It is passed from MDT in lustre_cfg[2] buffer */
138 rc = mdd_procfs_init(mdd, lustre_cfg_string(lcfg, 2));
140 obd_disconnect(mdd->mdd_child_exp);
145 static struct lu_device *mdd_device_fini(const struct lu_env *env,
148 struct mdd_device *mdd = lu2mdd_dev(d);
152 lu_dev_del_linkage(d->ld_site, d);
154 rc = mdd_procfs_fini(mdd);
156 CERROR("proc fini error %d \n", rc);
162 static void mdd_changelog_fini(const struct lu_env *env,
163 struct mdd_device *mdd);
165 static void mdd_device_shutdown(const struct lu_env *env,
166 struct mdd_device *m, struct lustre_cfg *cfg)
169 mdd_lfsck_cleanup(env, m);
170 mdd_changelog_fini(env, m);
171 if (m->mdd_dot_lustre_objs.mdd_obf)
172 mdd_object_put(env, m->mdd_dot_lustre_objs.mdd_obf);
173 if (m->mdd_dot_lustre)
174 mdd_object_put(env, m->mdd_dot_lustre);
175 orph_index_fini(env, m);
176 if (m->mdd_capa != NULL) {
177 lu_object_put(env, &m->mdd_capa->do_lu);
180 lu_site_purge(env, m->mdd_md_dev.md_lu_dev.ld_site, -1);
181 /* remove upcall device*/
182 md_upcall_fini(&m->mdd_md_dev);
184 if (m->mdd_child_exp)
185 obd_disconnect(m->mdd_child_exp);
190 static int changelog_init_cb(const struct lu_env *env, struct llog_handle *llh,
191 struct llog_rec_hdr *hdr, void *data)
193 struct mdd_device *mdd = (struct mdd_device *)data;
194 struct llog_changelog_rec *rec = (struct llog_changelog_rec *)hdr;
197 LASSERT(llh->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN);
198 LASSERT(rec->cr_hdr.lrh_type == CHANGELOG_REC);
201 "seeing record at index %d/%d/"LPU64" t=%x %.*s in log "LPX64"\n",
202 hdr->lrh_index, rec->cr_hdr.lrh_index, rec->cr.cr_index,
203 rec->cr.cr_type, rec->cr.cr_namelen, rec->cr.cr_name,
204 llh->lgh_id.lgl_oid);
206 mdd->mdd_cl.mc_index = rec->cr.cr_index;
207 RETURN(LLOG_PROC_BREAK);
210 static int changelog_user_init_cb(const struct lu_env *env,
211 struct llog_handle *llh,
212 struct llog_rec_hdr *hdr, void *data)
214 struct mdd_device *mdd = (struct mdd_device *)data;
215 struct llog_changelog_user_rec *rec =
216 (struct llog_changelog_user_rec *)hdr;
219 LASSERT(llh->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN);
220 LASSERT(rec->cur_hdr.lrh_type == CHANGELOG_USER_REC);
222 CDEBUG(D_INFO, "seeing user at index %d/%d id=%d endrec="LPU64
223 " in log "LPX64"\n", hdr->lrh_index, rec->cur_hdr.lrh_index,
224 rec->cur_id, rec->cur_endrec, llh->lgh_id.lgl_oid);
226 cfs_spin_lock(&mdd->mdd_cl.mc_user_lock);
227 mdd->mdd_cl.mc_lastuser = rec->cur_id;
228 cfs_spin_unlock(&mdd->mdd_cl.mc_user_lock);
230 RETURN(LLOG_PROC_BREAK);
234 static int mdd_changelog_llog_init(struct mdd_device *mdd)
236 struct obd_device *obd = mdd2obd_dev(mdd);
237 struct llog_ctxt *ctxt;
240 /* Find last changelog entry number */
241 ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
243 CERROR("no changelog context\n");
246 if (!ctxt->loc_handle) {
251 rc = llog_cat_reverse_process(NULL, ctxt->loc_handle,
252 changelog_init_cb, mdd);
256 CERROR("changelog init failed: %d\n", rc);
259 CDEBUG(D_IOCTL, "changelog starting index="LPU64"\n",
260 mdd->mdd_cl.mc_index);
262 /* Find last changelog user id */
263 ctxt = llog_get_context(obd, LLOG_CHANGELOG_USER_ORIG_CTXT);
265 CERROR("no changelog user context\n");
268 if (!ctxt->loc_handle) {
273 rc = llog_cat_reverse_process(NULL, ctxt->loc_handle,
274 changelog_user_init_cb, mdd);
278 CERROR("changelog user init failed: %d\n", rc);
282 /* If we have registered users, assume we want changelogs on */
283 if (mdd->mdd_cl.mc_lastuser > 0)
284 rc = mdd_changelog_on(mdd, 1);
289 static int mdd_changelog_init(const struct lu_env *env, struct mdd_device *mdd)
293 mdd->mdd_cl.mc_index = 0;
294 cfs_spin_lock_init(&mdd->mdd_cl.mc_lock);
295 mdd->mdd_cl.mc_starttime = cfs_time_current_64();
296 mdd->mdd_cl.mc_flags = 0; /* off by default */
297 mdd->mdd_cl.mc_mask = CHANGELOG_DEFMASK;
298 cfs_spin_lock_init(&mdd->mdd_cl.mc_user_lock);
299 mdd->mdd_cl.mc_lastuser = 0;
301 rc = mdd_changelog_llog_init(mdd);
303 CERROR("Changelog setup during init failed %d\n", rc);
304 mdd->mdd_cl.mc_flags |= CLM_ERR;
310 static void mdd_changelog_fini(const struct lu_env *env, struct mdd_device *mdd)
312 mdd->mdd_cl.mc_flags = 0;
315 /* Start / stop recording */
316 int mdd_changelog_on(struct mdd_device *mdd, int on)
320 if ((on == 1) && ((mdd->mdd_cl.mc_flags & CLM_ON) == 0)) {
321 LCONSOLE_INFO("%s: changelog on\n", mdd2obd_dev(mdd)->obd_name);
322 if (mdd->mdd_cl.mc_flags & CLM_ERR) {
323 CERROR("Changelogs cannot be enabled due to error "
324 "condition (see %s log).\n",
325 mdd2obd_dev(mdd)->obd_name);
328 cfs_spin_lock(&mdd->mdd_cl.mc_lock);
329 mdd->mdd_cl.mc_flags |= CLM_ON;
330 cfs_spin_unlock(&mdd->mdd_cl.mc_lock);
331 rc = mdd_changelog_write_header(mdd, CLM_START);
333 } else if ((on == 0) && ((mdd->mdd_cl.mc_flags & CLM_ON) == CLM_ON)) {
334 LCONSOLE_INFO("%s: changelog off\n",mdd2obd_dev(mdd)->obd_name);
335 rc = mdd_changelog_write_header(mdd, CLM_FINI);
336 cfs_spin_lock(&mdd->mdd_cl.mc_lock);
337 mdd->mdd_cl.mc_flags &= ~CLM_ON;
338 cfs_spin_unlock(&mdd->mdd_cl.mc_lock);
343 static __u64 cl_time(void) {
346 cfs_fs_time_current(&time);
347 return (((__u64)time.tv_sec) << 30) + time.tv_nsec;
350 /** Add a changelog entry \a rec to the changelog llog
353 * \param handle - currently ignored since llogs start their own transaction;
354 * this will hopefully be fixed in llog rewrite
357 int mdd_changelog_llog_write(struct mdd_device *mdd,
358 struct llog_changelog_rec *rec,
359 struct thandle *handle)
361 struct obd_device *obd = mdd2obd_dev(mdd);
362 struct llog_ctxt *ctxt;
365 rec->cr_hdr.lrh_len = llog_data_len(sizeof(*rec) + rec->cr.cr_namelen);
366 /* llog_lvfs_write_rec sets the llog tail len */
367 rec->cr_hdr.lrh_type = CHANGELOG_REC;
368 rec->cr.cr_time = cl_time();
369 cfs_spin_lock(&mdd->mdd_cl.mc_lock);
370 /* NB: I suppose it's possible llog_add adds out of order wrt cr_index,
371 but as long as the MDD transactions are ordered correctly for e.g.
372 rename conflicts, I don't think this should matter. */
373 rec->cr.cr_index = ++mdd->mdd_cl.mc_index;
374 cfs_spin_unlock(&mdd->mdd_cl.mc_lock);
375 ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
379 /* nested journal transaction */
380 rc = llog_obd_add(NULL, ctxt, &rec->cr_hdr, NULL, NULL, 0);
386 /** Add a changelog_ext entry \a rec to the changelog llog
389 * \param handle - currently ignored since llogs start their own transaction;
390 * this will hopefully be fixed in llog rewrite
393 int mdd_changelog_ext_llog_write(struct mdd_device *mdd,
394 struct llog_changelog_ext_rec *rec,
395 struct thandle *handle)
397 struct obd_device *obd = mdd2obd_dev(mdd);
398 struct llog_ctxt *ctxt;
401 rec->cr_hdr.lrh_len = llog_data_len(sizeof(*rec) + rec->cr.cr_namelen);
402 /* llog_lvfs_write_rec sets the llog tail len */
403 rec->cr_hdr.lrh_type = CHANGELOG_REC;
404 rec->cr.cr_time = cl_time();
405 cfs_spin_lock(&mdd->mdd_cl.mc_lock);
406 /* NB: I suppose it's possible llog_add adds out of order wrt cr_index,
407 * but as long as the MDD transactions are ordered correctly for e.g.
408 * rename conflicts, I don't think this should matter. */
409 rec->cr.cr_index = ++mdd->mdd_cl.mc_index;
410 cfs_spin_unlock(&mdd->mdd_cl.mc_lock);
411 ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
415 /* nested journal transaction */
416 rc = llog_obd_add(NULL, ctxt, &rec->cr_hdr, NULL, NULL, 0);
422 /** Remove entries with indicies up to and including \a endrec from the
428 int mdd_changelog_llog_cancel(const struct lu_env *env,
429 struct mdd_device *mdd, long long endrec)
431 struct obd_device *obd = mdd2obd_dev(mdd);
432 struct llog_ctxt *ctxt;
433 long long unsigned cur;
436 ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
440 cfs_spin_lock(&mdd->mdd_cl.mc_lock);
441 cur = (long long)mdd->mdd_cl.mc_index;
442 cfs_spin_unlock(&mdd->mdd_cl.mc_lock);
446 /* purge to "0" is shorthand for everything */
450 /* If purging all records, write a header entry so we don't have an
451 empty catalog and we're sure to have a valid starting index next
452 time. In case of crash, we just restart with old log so we're
455 /* XXX: transaction is started by llog itself */
456 rc = mdd_changelog_write_header(mdd, CLM_PURGE);
461 /* Some records were purged, so reset repeat-access time (so we
462 record new mtime update records, so users can see a file has been
463 changed since the last purge) */
464 mdd->mdd_cl.mc_starttime = cfs_time_current_64();
466 /* XXX: transaction is started by llog itself */
467 rc = llog_cancel(env, ctxt, NULL, 1, (struct llog_cookie *)&endrec, 0);
473 /** Add a CL_MARK record to the changelog
475 * \param markerflags - CLM_*
478 int mdd_changelog_write_header(struct mdd_device *mdd, int markerflags)
480 struct obd_device *obd = mdd2obd_dev(mdd);
481 struct llog_changelog_rec *rec;
483 int len = strlen(obd->obd_name);
487 reclen = llog_data_len(sizeof(*rec) + len);
488 OBD_ALLOC(rec, reclen);
492 rec->cr.cr_flags = CLF_VERSION;
493 rec->cr.cr_type = CL_MARK;
494 rec->cr.cr_namelen = len;
495 memcpy(rec->cr.cr_name, obd->obd_name, rec->cr.cr_namelen);
496 /* Status and action flags */
497 rec->cr.cr_markerflags = mdd->mdd_cl.mc_flags | markerflags;
499 /* XXX: transaction is started by llog itself */
500 rc = (mdd->mdd_cl.mc_mask & (1 << CL_MARK)) ?
501 mdd_changelog_llog_write(mdd, rec, NULL) : 0;
503 /* assume on or off event; reset repeat-access time */
504 mdd->mdd_cl.mc_starttime = cfs_time_current_64();
506 OBD_FREE(rec, reclen);
511 * Create ".lustre" directory.
513 static int create_dot_lustre_dir(const struct lu_env *env, struct mdd_device *m)
515 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
516 struct md_object *mdo;
519 memcpy(fid, &LU_DOT_LUSTRE_FID, sizeof(struct lu_fid));
520 mdo = llo_store_create_index(env, &m->mdd_md_dev, m->mdd_child,
521 mdd_root_dir_name, dot_lustre_name,
522 fid, &dt_directory_features);
523 /* .lustre dir may be already present */
524 if (IS_ERR(mdo) && PTR_ERR(mdo) != -EEXIST) {
526 CERROR("creating obj [%s] fid = "DFID" rc = %d\n",
527 dot_lustre_name, PFID(fid), rc);
532 lu_object_put(env, &mdo->mo_lu);
538 static int dot_lustre_mdd_permission(const struct lu_env *env,
539 struct md_object *pobj,
540 struct md_object *cobj,
541 struct md_attr *attr, int mask)
543 if (mask & ~(MAY_READ | MAY_EXEC))
549 static int dot_lustre_mdd_xattr_get(const struct lu_env *env,
550 struct md_object *obj, struct lu_buf *buf,
556 static int dot_lustre_mdd_xattr_list(const struct lu_env *env,
557 struct md_object *obj, struct lu_buf *buf)
562 static int dot_lustre_mdd_xattr_set(const struct lu_env *env,
563 struct md_object *obj,
564 const struct lu_buf *buf, const char *name,
570 static int dot_lustre_mdd_xattr_del(const struct lu_env *env,
571 struct md_object *obj,
577 static int dot_lustre_mdd_readlink(const struct lu_env *env,
578 struct md_object *obj, struct lu_buf *buf)
583 static int dot_lustre_mdd_object_create(const struct lu_env *env,
584 struct md_object *obj,
585 const struct md_op_spec *spec,
591 static int dot_lustre_mdd_ref_add(const struct lu_env *env,
592 struct md_object *obj,
593 const struct md_attr *ma)
598 static int dot_lustre_mdd_ref_del(const struct lu_env *env,
599 struct md_object *obj,
605 static int dot_lustre_mdd_open(const struct lu_env *env, struct md_object *obj,
608 struct mdd_object *mdd_obj = md2mdd_obj(obj);
610 mdd_write_lock(env, mdd_obj, MOR_TGT_CHILD);
611 mdd_obj->mod_count++;
612 mdd_write_unlock(env, mdd_obj);
617 static int dot_lustre_mdd_close(const struct lu_env *env, struct md_object *obj,
618 struct md_attr *ma, int mode)
620 struct mdd_object *mdd_obj = md2mdd_obj(obj);
622 mdd_write_lock(env, mdd_obj, MOR_TGT_CHILD);
623 mdd_obj->mod_count--;
624 mdd_write_unlock(env, mdd_obj);
629 static int dot_lustre_mdd_object_sync(const struct lu_env *env,
630 struct md_object *obj)
635 static int dot_lustre_mdd_path(const struct lu_env *env, struct md_object *obj,
636 char *path, int pathlen, __u64 *recno, int *linkno)
641 static int dot_file_lock(const struct lu_env *env, struct md_object *obj,
642 struct lov_mds_md *lmm, struct ldlm_extent *extent,
643 struct lustre_handle *lockh)
648 static int dot_file_unlock(const struct lu_env *env, struct md_object *obj,
649 struct lov_mds_md *lmm, struct lustre_handle *lockh)
654 static struct md_object_operations mdd_dot_lustre_obj_ops = {
655 .moo_permission = dot_lustre_mdd_permission,
656 .moo_attr_get = mdd_attr_get,
657 .moo_attr_set = mdd_attr_set,
658 .moo_xattr_get = dot_lustre_mdd_xattr_get,
659 .moo_xattr_list = dot_lustre_mdd_xattr_list,
660 .moo_xattr_set = dot_lustre_mdd_xattr_set,
661 .moo_xattr_del = dot_lustre_mdd_xattr_del,
662 .moo_readpage = mdd_readpage,
663 .moo_readlink = dot_lustre_mdd_readlink,
664 .moo_object_create = dot_lustre_mdd_object_create,
665 .moo_ref_add = dot_lustre_mdd_ref_add,
666 .moo_ref_del = dot_lustre_mdd_ref_del,
667 .moo_open = dot_lustre_mdd_open,
668 .moo_close = dot_lustre_mdd_close,
669 .moo_capa_get = mdd_capa_get,
670 .moo_object_sync = dot_lustre_mdd_object_sync,
671 .moo_path = dot_lustre_mdd_path,
672 .moo_file_lock = dot_file_lock,
673 .moo_file_unlock = dot_file_unlock,
677 static int dot_lustre_mdd_lookup(const struct lu_env *env, struct md_object *p,
678 const struct lu_name *lname, struct lu_fid *f,
679 struct md_op_spec *spec)
681 if (strcmp(lname->ln_name, mdd_obf_dir_name) == 0)
689 static mdl_mode_t dot_lustre_mdd_lock_mode(const struct lu_env *env,
690 struct md_object *obj,
696 static int dot_lustre_mdd_create(const struct lu_env *env,
697 struct md_object *pobj,
698 const struct lu_name *lname,
699 struct md_object *child,
700 struct md_op_spec *spec,
706 static int dot_lustre_mdd_create_data(const struct lu_env *env,
709 const struct md_op_spec *spec,
715 static int dot_lustre_mdd_rename(const struct lu_env *env,
716 struct md_object *src_pobj,
717 struct md_object *tgt_pobj,
718 const struct lu_fid *lf,
719 const struct lu_name *lsname,
720 struct md_object *tobj,
721 const struct lu_name *ltname,
727 static int dot_lustre_mdd_link(const struct lu_env *env,
728 struct md_object *tgt_obj,
729 struct md_object *src_obj,
730 const struct lu_name *lname,
736 static int dot_lustre_mdd_unlink(const struct lu_env *env,
737 struct md_object *pobj,
738 struct md_object *cobj,
739 const struct lu_name *lname,
745 static int dot_lustre_mdd_name_insert(const struct lu_env *env,
746 struct md_object *obj,
747 const struct lu_name *lname,
748 const struct lu_fid *fid,
749 const struct md_attr *ma)
754 static int dot_lustre_mdd_name_remove(const struct lu_env *env,
755 struct md_object *obj,
756 const struct lu_name *lname,
757 const struct md_attr *ma)
762 static int dot_lustre_mdd_rename_tgt(const struct lu_env *env,
763 struct md_object *pobj,
764 struct md_object *tobj,
765 const struct lu_fid *fid,
766 const struct lu_name *lname,
773 static struct md_dir_operations mdd_dot_lustre_dir_ops = {
774 .mdo_is_subdir = mdd_is_subdir,
775 .mdo_lookup = dot_lustre_mdd_lookup,
776 .mdo_lock_mode = dot_lustre_mdd_lock_mode,
777 .mdo_create = dot_lustre_mdd_create,
778 .mdo_create_data = dot_lustre_mdd_create_data,
779 .mdo_rename = dot_lustre_mdd_rename,
780 .mdo_link = dot_lustre_mdd_link,
781 .mdo_unlink = dot_lustre_mdd_unlink,
782 .mdo_name_insert = dot_lustre_mdd_name_insert,
783 .mdo_name_remove = dot_lustre_mdd_name_remove,
784 .mdo_rename_tgt = dot_lustre_mdd_rename_tgt,
787 static int obf_attr_get(const struct lu_env *env, struct md_object *obj,
790 struct mdd_device *mdd = mdo2mdd(obj);
792 /* "fid" is a virtual object and hence does not have any "real"
793 * attributes. So we reuse attributes of .lustre for "fid" dir */
794 return mdd_attr_get(env, &mdd->mdd_dot_lustre->mod_obj, ma);
797 static int obf_attr_set(const struct lu_env *env, struct md_object *obj,
798 const struct md_attr *ma)
803 static int obf_xattr_list(const struct lu_env *env,
804 struct md_object *obj, struct lu_buf *buf)
809 static int obf_xattr_get(const struct lu_env *env,
810 struct md_object *obj, struct lu_buf *buf,
813 struct mdd_device *mdd = mdo2mdd(obj);
814 struct mdd_object *root;
815 struct lu_fid rootfid;
819 * .lustre returns default striping which is 'stored'
822 if (strcmp(name, XATTR_NAME_LOV) == 0) {
823 rc = dt_root_get(env, mdd->mdd_child, &rootfid);
826 root = mdd_object_find(env, mdd, &rootfid);
828 return PTR_ERR(root);
829 rc = mdo_xattr_get(env, root, buf, name,
830 mdd_object_capa(env, md2mdd_obj(obj)));
831 mdd_object_put(env, root);
837 static int obf_xattr_set(const struct lu_env *env,
838 struct md_object *obj,
839 const struct lu_buf *buf, const char *name,
845 static int obf_xattr_del(const struct lu_env *env,
846 struct md_object *obj,
852 static int obf_mdd_open(const struct lu_env *env, struct md_object *obj,
855 struct mdd_object *mdd_obj = md2mdd_obj(obj);
857 mdd_write_lock(env, mdd_obj, MOR_TGT_CHILD);
858 mdd_obj->mod_count++;
859 mdd_write_unlock(env, mdd_obj);
864 static int obf_mdd_close(const struct lu_env *env, struct md_object *obj,
865 struct md_attr *ma, int mode)
867 struct mdd_object *mdd_obj = md2mdd_obj(obj);
869 mdd_write_lock(env, mdd_obj, MOR_TGT_CHILD);
870 mdd_obj->mod_count--;
871 mdd_write_unlock(env, mdd_obj);
876 /** Nothing to list in "fid" directory */
877 static int obf_mdd_readpage(const struct lu_env *env, struct md_object *obj,
878 const struct lu_rdpg *rdpg)
883 static int obf_path(const struct lu_env *env, struct md_object *obj,
884 char *path, int pathlen, __u64 *recno, int *linkno)
889 static struct md_object_operations mdd_obf_obj_ops = {
890 .moo_attr_get = obf_attr_get,
891 .moo_attr_set = obf_attr_set,
892 .moo_xattr_list = obf_xattr_list,
893 .moo_xattr_get = obf_xattr_get,
894 .moo_xattr_set = obf_xattr_set,
895 .moo_xattr_del = obf_xattr_del,
896 .moo_open = obf_mdd_open,
897 .moo_close = obf_mdd_close,
898 .moo_readpage = obf_mdd_readpage,
903 * Lookup method for "fid" object. Only filenames with correct SEQ:OID format
904 * are valid. We also check if object with passed fid exists or not.
906 static int obf_lookup(const struct lu_env *env, struct md_object *p,
907 const struct lu_name *lname, struct lu_fid *f,
908 struct md_op_spec *spec)
910 char *name = (char *)lname->ln_name;
911 struct mdd_device *mdd = mdo2mdd(p);
912 struct mdd_object *child;
918 sscanf(name, SFID, RFID(f));
919 if (!fid_is_sane(f)) {
920 CWARN("%s: bad FID format [%s], should be "DFID"\n",
921 mdd2obd_dev(mdd)->obd_name, lname->ln_name,
922 (__u64)FID_SEQ_NORMAL, 1, 0);
923 GOTO(out, rc = -EINVAL);
926 if (!fid_is_norm(f)) {
927 CWARN("%s: "DFID" is invalid, sequence should be "
928 ">= "LPX64"\n", mdd2obd_dev(mdd)->obd_name, PFID(f),
929 (__u64)FID_SEQ_NORMAL);
930 GOTO(out, rc = -EINVAL);
933 /* Check if object with this fid exists */
934 child = mdd_object_find(env, mdd, f);
938 GOTO(out, rc = PTR_ERR(child));
940 if (mdd_object_exists(child) == 0)
943 mdd_object_put(env, child);
949 static int obf_create(const struct lu_env *env, struct md_object *pobj,
950 const struct lu_name *lname, struct md_object *child,
951 struct md_op_spec *spec, struct md_attr* ma)
956 static int obf_rename(const struct lu_env *env,
957 struct md_object *src_pobj, struct md_object *tgt_pobj,
958 const struct lu_fid *lf, const struct lu_name *lsname,
959 struct md_object *tobj, const struct lu_name *ltname,
965 static int obf_link(const struct lu_env *env, struct md_object *tgt_obj,
966 struct md_object *src_obj, const struct lu_name *lname,
972 static int obf_unlink(const struct lu_env *env, struct md_object *pobj,
973 struct md_object *cobj, const struct lu_name *lname,
979 static struct md_dir_operations mdd_obf_dir_ops = {
980 .mdo_lookup = obf_lookup,
981 .mdo_create = obf_create,
982 .mdo_rename = obf_rename,
983 .mdo_link = obf_link,
984 .mdo_unlink = obf_unlink
988 * Create special in-memory "fid" object for open-by-fid.
990 static int mdd_obf_setup(const struct lu_env *env, struct mdd_device *m)
992 struct mdd_object *mdd_obf;
993 struct lu_object *obf_lu_obj;
996 m->mdd_dot_lustre_objs.mdd_obf = mdd_object_find(env, m,
998 if (m->mdd_dot_lustre_objs.mdd_obf == NULL ||
999 IS_ERR(m->mdd_dot_lustre_objs.mdd_obf))
1000 GOTO(out, rc = -ENOENT);
1002 mdd_obf = m->mdd_dot_lustre_objs.mdd_obf;
1003 mdd_obf->mod_obj.mo_dir_ops = &mdd_obf_dir_ops;
1004 mdd_obf->mod_obj.mo_ops = &mdd_obf_obj_ops;
1005 /* Don't allow objects to be created in "fid" dir */
1006 mdd_obf->mod_flags |= IMMUTE_OBJ;
1008 obf_lu_obj = mdd2lu_obj(mdd_obf);
1009 obf_lu_obj->lo_header->loh_attr |= (LOHA_EXISTS | S_IFDIR);
1015 /** Setup ".lustre" directory object */
1016 static int mdd_dot_lustre_setup(const struct lu_env *env, struct mdd_device *m)
1018 struct dt_object *dt_dot_lustre;
1019 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
1022 rc = create_dot_lustre_dir(env, m);
1026 dt_dot_lustre = dt_store_open(env, m->mdd_child, mdd_root_dir_name,
1027 dot_lustre_name, fid);
1028 if (IS_ERR(dt_dot_lustre)) {
1029 rc = PTR_ERR(dt_dot_lustre);
1033 /* references are released in mdd_device_shutdown() */
1034 m->mdd_dot_lustre = lu2mdd_obj(lu_object_locate(dt_dot_lustre->do_lu.lo_header,
1037 m->mdd_dot_lustre->mod_obj.mo_dir_ops = &mdd_dot_lustre_dir_ops;
1038 m->mdd_dot_lustre->mod_obj.mo_ops = &mdd_dot_lustre_obj_ops;
1040 rc = mdd_obf_setup(env, m);
1042 CERROR("Error initializing \"fid\" object - %d.\n", rc);
1048 static int mdd_process_config(const struct lu_env *env,
1049 struct lu_device *d, struct lustre_cfg *cfg)
1051 struct mdd_device *m = lu2mdd_dev(d);
1052 struct dt_device *dt = m->mdd_child;
1053 struct lu_device *next = &dt->dd_lu_dev;
1057 switch (cfg->lcfg_command) {
1059 struct lprocfs_static_vars lvars;
1061 lprocfs_mdd_init_vars(&lvars);
1062 rc = class_process_proc_param(PARAM_MDD, lvars.obd_vars, cfg,m);
1063 if (rc > 0 || rc == -ENOSYS)
1064 /* we don't understand; pass it on */
1065 rc = next->ld_ops->ldo_process_config(env, next, cfg);
1069 rc = next->ld_ops->ldo_process_config(env, next, cfg);
1072 dt->dd_ops->dt_conf_get(env, dt, &m->mdd_dt_conf);
1074 mdd_changelog_init(env, m);
1077 rc = next->ld_ops->ldo_process_config(env, next, cfg);
1078 lu_dev_del_linkage(d->ld_site, d);
1079 mdd_device_shutdown(env, m, cfg);
1082 rc = next->ld_ops->ldo_process_config(env, next, cfg);
1089 static int mdd_recovery_complete(const struct lu_env *env,
1090 struct lu_device *d)
1092 struct mdd_device *mdd = lu2mdd_dev(d);
1093 struct lu_device *next = &mdd->mdd_child->dd_lu_dev;
1097 LASSERT(mdd != NULL);
1099 /* XXX: orphans handling. */
1100 __mdd_orphan_cleanup(env, mdd);
1101 rc = next->ld_ops->ldo_recovery_complete(env, next);
1106 static int mdd_prepare(const struct lu_env *env,
1107 struct lu_device *pdev,
1108 struct lu_device *cdev)
1110 struct mdd_device *mdd = lu2mdd_dev(cdev);
1111 struct lu_device *next = &mdd->mdd_child->dd_lu_dev;
1112 struct dt_object *root;
1117 rc = next->ld_ops->ldo_prepare(env, cdev, next);
1121 root = dt_store_open(env, mdd->mdd_child, "", mdd_root_dir_name,
1122 &mdd->mdd_root_fid);
1123 if (!IS_ERR(root)) {
1124 LASSERT(root != NULL);
1125 lu_object_put(env, &root->do_lu);
1126 rc = orph_index_init(env, mdd);
1133 rc = mdd_dot_lustre_setup(env, mdd);
1135 CERROR("Error(%d) initializing .lustre objects\n", rc);
1139 /* we use capa file to declare llog changes,
1140 * will be fixed with new llog in 2.3 */
1141 root = dt_store_open(env, mdd->mdd_child, "", CAPA_KEYS, &fid);
1143 GOTO(out, rc = PTR_ERR(root));
1145 mdd->mdd_capa = root;
1146 rc = mdd_lfsck_setup(env, mdd);
1154 const struct lu_device_operations mdd_lu_ops = {
1155 .ldo_object_alloc = mdd_object_alloc,
1156 .ldo_process_config = mdd_process_config,
1157 .ldo_recovery_complete = mdd_recovery_complete,
1158 .ldo_prepare = mdd_prepare,
1162 * No permission check is needed.
1164 static int mdd_root_get(const struct lu_env *env,
1165 struct md_device *m, struct lu_fid *f)
1167 struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
1170 *f = mdd->mdd_root_fid;
1175 * No permission check is needed.
1177 static int mdd_statfs(const struct lu_env *env, struct md_device *m,
1178 struct obd_statfs *sfs)
1180 struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
1185 rc = mdd_child_ops(mdd)->dt_statfs(env, mdd->mdd_child, sfs);
1191 * No permission check is needed.
1193 static int mdd_maxsize_get(const struct lu_env *env, struct md_device *m,
1194 int *md_size, int *cookie_size)
1196 struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
1199 *md_size = mdd_lov_mdsize(env, mdd);
1200 *cookie_size = mdd_lov_cookiesize(env, mdd);
1205 static int mdd_init_capa_ctxt(const struct lu_env *env, struct md_device *m,
1206 int mode, unsigned long timeout, __u32 alg,
1207 struct lustre_capa_key *keys)
1209 struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
1210 struct mds_obd *mds = &mdd2obd_dev(mdd)->u.mds;
1214 /* need barrier for mds_capa_keys access. */
1215 cfs_down_write(&mds->mds_notify_lock);
1216 mds->mds_capa_keys = keys;
1217 cfs_up_write(&mds->mds_notify_lock);
1219 rc = mdd_child_ops(mdd)->dt_init_capa_ctxt(env, mdd->mdd_child, mode,
1220 timeout, alg, keys);
1224 static int mdd_update_capa_key(const struct lu_env *env,
1225 struct md_device *m,
1226 struct lustre_capa_key *key)
1228 /* we do not support capabilities ... */
1232 static int mdd_llog_ctxt_get(const struct lu_env *env, struct md_device *m,
1235 struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
1237 *h = llog_group_get_ctxt(&mdd2obd_dev(mdd)->obd_olg, idx);
1238 return (*h == NULL ? -ENOENT : 0);
1241 static struct lu_device *mdd_device_free(const struct lu_env *env,
1242 struct lu_device *lu)
1244 struct mdd_device *m = lu2mdd_dev(lu);
1247 LASSERT(cfs_atomic_read(&lu->ld_ref) == 0);
1248 md_device_fini(&m->mdd_md_dev);
1253 static struct lu_device *mdd_device_alloc(const struct lu_env *env,
1254 struct lu_device_type *t,
1255 struct lustre_cfg *lcfg)
1257 struct lu_device *l;
1258 struct mdd_device *m;
1262 l = ERR_PTR(-ENOMEM);
1267 md_device_init(&m->mdd_md_dev, t);
1268 rc = mdd_init0(env, m, t, lcfg);
1270 mdd_device_free(env, l);
1279 * we use exports to track all mdd users
1281 static int mdd_obd_connect(const struct lu_env *env, struct obd_export **exp,
1282 struct obd_device *obd, struct obd_uuid *cluuid,
1283 struct obd_connect_data *data, void *localdata)
1285 struct mdd_device *mdd = lu2mdd_dev(obd->obd_lu_dev);
1286 struct lustre_handle conn;
1290 CDEBUG(D_CONFIG, "connect #%d\n", mdd->mdd_connects);
1292 rc = class_connect(&conn, obd, cluuid);
1296 *exp = class_conn2export(&conn);
1298 /* Why should there ever be more than 1 connect? */
1299 LASSERT(mdd->mdd_connects == 0);
1300 mdd->mdd_connects++;
1306 * once last export (we don't count self-export) disappeared
1307 * mdd can be released
1309 static int mdd_obd_disconnect(struct obd_export *exp)
1311 struct obd_device *obd = exp->exp_obd;
1312 struct mdd_device *mdd = lu2mdd_dev(obd->obd_lu_dev);
1313 int rc, release = 0;
1316 mdd->mdd_connects--;
1317 if (mdd->mdd_connects == 0)
1320 rc = class_disconnect(exp);
1322 if (rc == 0 && release)
1323 class_manual_cleanup(obd);
1327 static int mdd_obd_health_check(const struct lu_env *env,
1328 struct obd_device *obd)
1330 struct mdd_device *mdd = lu2mdd_dev(obd->obd_lu_dev);
1335 rc = obd_health_check(env, mdd->mdd_child_exp->exp_obd);
1339 static struct obd_ops mdd_obd_device_ops = {
1340 .o_owner = THIS_MODULE,
1341 .o_connect = mdd_obd_connect,
1342 .o_disconnect = mdd_obd_disconnect,
1343 .o_health_check = mdd_obd_health_check
1346 /* context key constructor/destructor: mdd_ucred_key_init, mdd_ucred_key_fini */
1347 LU_KEY_INIT_FINI(mdd_ucred, struct md_ucred);
1349 static struct lu_context_key mdd_ucred_key = {
1350 .lct_tags = LCT_SESSION,
1351 .lct_init = mdd_ucred_key_init,
1352 .lct_fini = mdd_ucred_key_fini
1355 struct md_ucred *md_ucred(const struct lu_env *env)
1357 LASSERT(env->le_ses != NULL);
1358 return lu_context_key_get(env->le_ses, &mdd_ucred_key);
1360 EXPORT_SYMBOL(md_ucred);
1363 * context key constructor/destructor:
1364 * mdd_capainfo_key_init, mdd_capainfo_key_fini
1366 LU_KEY_INIT_FINI(mdd_capainfo, struct md_capainfo);
1368 struct lu_context_key mdd_capainfo_key = {
1369 .lct_tags = LCT_SESSION,
1370 .lct_init = mdd_capainfo_key_init,
1371 .lct_fini = mdd_capainfo_key_fini
1374 struct md_capainfo *md_capainfo(const struct lu_env *env)
1376 /* NB, in mdt_init0 */
1377 if (env->le_ses == NULL)
1379 return lu_context_key_get(env->le_ses, &mdd_capainfo_key);
1381 EXPORT_SYMBOL(md_capainfo);
1383 static int mdd_changelog_user_register(struct mdd_device *mdd, int *id)
1385 struct llog_ctxt *ctxt;
1386 struct llog_changelog_user_rec *rec;
1390 ctxt = llog_get_context(mdd2obd_dev(mdd),LLOG_CHANGELOG_USER_ORIG_CTXT);
1396 llog_ctxt_put(ctxt);
1400 /* Assume we want it on since somebody registered */
1401 rc = mdd_changelog_on(mdd, 1);
1405 rec->cur_hdr.lrh_len = sizeof(*rec);
1406 rec->cur_hdr.lrh_type = CHANGELOG_USER_REC;
1407 cfs_spin_lock(&mdd->mdd_cl.mc_user_lock);
1408 if (mdd->mdd_cl.mc_lastuser == (unsigned int)(-1)) {
1409 cfs_spin_unlock(&mdd->mdd_cl.mc_user_lock);
1410 CERROR("Maximum number of changelog users exceeded!\n");
1411 GOTO(out, rc = -EOVERFLOW);
1413 *id = rec->cur_id = ++mdd->mdd_cl.mc_lastuser;
1414 rec->cur_endrec = mdd->mdd_cl.mc_index;
1415 cfs_spin_unlock(&mdd->mdd_cl.mc_user_lock);
1417 rc = llog_obd_add(NULL, ctxt, &rec->cur_hdr, NULL, NULL, 0);
1419 CDEBUG(D_IOCTL, "Registered changelog user %d\n", *id);
1422 llog_ctxt_put(ctxt);
1426 int mdd_declare_llog_cancel(const struct lu_env *env, struct mdd_device *mdd,
1427 struct thandle *handle)
1432 /* XXX: this is a temporary solution to declare llog changes
1433 * will be fixed in 2.3 with new llog implementation */
1435 LASSERT(mdd->mdd_capa);
1437 /* the llog record could be canceled either by modifying
1438 * the plain llog's header or by destroying the llog itself
1439 * when this record is the last one in it, it can't be known
1440 * here, but the catlog's header will also be modified for
1441 * the second case, then the first case can be covered and
1442 * is no need to declare it */
1444 /* destroy empty plain log */
1445 rc = dt_declare_destroy(env, mdd->mdd_capa, handle);
1449 /* record the catlog's header if an empty plain log was destroyed */
1450 rc = dt_declare_record_write(env, mdd->mdd_capa,
1451 sizeof(struct llog_logid_rec), 0, handle);
1455 struct mdd_changelog_user_data {
1456 __u64 mcud_endrec; /**< purge record for this user */
1457 __u64 mcud_minrec; /**< lowest changelog recno still referenced */
1459 __u32 mcud_minid; /**< user id with lowest rec reference */
1460 __u32 mcud_usercount;
1462 struct mdd_device *mcud_mdd;
1464 #define MCUD_UNREGISTER -1LL
1467 * 1. Find the smallest record everyone is willing to purge
1468 * 2. Update the last purgeable record for this user
1470 static int mdd_changelog_user_purge_cb(const struct lu_env *env,
1471 struct llog_handle *llh,
1472 struct llog_rec_hdr *hdr, void *data)
1474 struct llog_changelog_user_rec *rec;
1475 struct mdd_changelog_user_data *mcud = data;
1480 LASSERT(llh->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN);
1482 rec = (struct llog_changelog_user_rec *)hdr;
1484 mcud->mcud_usercount++;
1486 /* If we have a new endrec for this id, use it for the following
1487 min check instead of its old value */
1488 if (rec->cur_id == mcud->mcud_id)
1489 rec->cur_endrec = max(rec->cur_endrec, mcud->mcud_endrec);
1491 /* Track the minimum referenced record */
1492 if (mcud->mcud_minid == 0 || mcud->mcud_minrec > rec->cur_endrec) {
1493 mcud->mcud_minid = rec->cur_id;
1494 mcud->mcud_minrec = rec->cur_endrec;
1497 if (rec->cur_id != mcud->mcud_id)
1500 /* Update this user's record */
1501 mcud->mcud_found = 1;
1503 /* Special case: unregister this user */
1504 if (mcud->mcud_endrec == MCUD_UNREGISTER) {
1505 struct llog_cookie cookie;
1507 struct mdd_device *mdd = mcud->mcud_mdd;
1509 cookie.lgc_lgl = llh->lgh_id;
1510 cookie.lgc_index = hdr->lrh_index;
1512 /* XXX This is a workaround for the deadlock of changelog
1513 * adding vs. changelog cancelling. LU-81. */
1514 th = mdd_trans_create(env, mdd);
1516 CERROR("Cannot get thandle\n");
1520 rc = mdd_declare_llog_cancel(env, mdd, th);
1524 rc = mdd_trans_start(env, mdd, th);
1528 rc = llog_cat_cancel_records(env, llh->u.phd.phd_cat_handle,
1531 mcud->mcud_usercount--;
1534 mdd_trans_stop(env, mdd, 0, th);
1538 /* Update the endrec */
1539 CDEBUG(D_IOCTL, "Rewriting changelog user %d endrec to "LPU64"\n",
1540 mcud->mcud_id, rec->cur_endrec);
1542 /* hdr+1 is loc of data */
1543 hdr->lrh_len -= sizeof(*hdr) + sizeof(struct llog_rec_tail);
1544 rc = llog_write_rec(env, llh, hdr, NULL, 0, (void *)(hdr + 1),
1545 hdr->lrh_index, NULL);
1550 static int mdd_changelog_user_purge(const struct lu_env *env,
1551 struct mdd_device *mdd, int id,
1554 struct mdd_changelog_user_data data;
1555 struct llog_ctxt *ctxt;
1559 CDEBUG(D_IOCTL, "Purge request: id=%d, endrec=%lld\n", id, endrec);
1562 data.mcud_minid = 0;
1563 data.mcud_minrec = 0;
1564 data.mcud_usercount = 0;
1565 data.mcud_endrec = endrec;
1566 data.mcud_mdd = mdd;
1567 cfs_spin_lock(&mdd->mdd_cl.mc_lock);
1568 endrec = mdd->mdd_cl.mc_index;
1569 cfs_spin_unlock(&mdd->mdd_cl.mc_lock);
1570 if ((data.mcud_endrec == 0) ||
1571 ((data.mcud_endrec > endrec) &&
1572 (data.mcud_endrec != MCUD_UNREGISTER)))
1573 data.mcud_endrec = endrec;
1575 ctxt = llog_get_context(mdd2obd_dev(mdd),LLOG_CHANGELOG_USER_ORIG_CTXT);
1578 LASSERT(ctxt->loc_handle->lgh_hdr->llh_flags & LLOG_F_IS_CAT);
1580 rc = llog_cat_process(env, ctxt->loc_handle,
1581 mdd_changelog_user_purge_cb, (void *)&data,
1583 if ((rc >= 0) && (data.mcud_minrec > 0)) {
1584 CDEBUG(D_IOCTL, "Purging changelog entries up to "LPD64
1585 ", referenced by "CHANGELOG_USER_PREFIX"%d\n",
1586 data.mcud_minrec, data.mcud_minid);
1587 rc = mdd_changelog_llog_cancel(env, mdd, data.mcud_minrec);
1589 CWARN("Could not determine changelog records to purge; rc=%d\n",
1593 llog_ctxt_put(ctxt);
1595 if (!data.mcud_found) {
1596 CWARN("No entry for user %d. Last changelog reference is "
1597 LPD64" by changelog user %d\n", data.mcud_id,
1598 data.mcud_minrec, data.mcud_minid);
1602 if (!rc && data.mcud_usercount == 0)
1603 /* No more users; turn changelogs off */
1604 rc = mdd_changelog_on(mdd, 0);
1610 * May be called remotely from mdt_iocontrol_handle or locally from
1611 * mdt_iocontrol. Data may be freeform - remote handling doesn't enforce
1612 * an obd_ioctl_data format (but local ioctl handler does).
1614 * \param len - data len
1615 * \param karg - ioctl data, in kernel space
1617 static int mdd_iocontrol(const struct lu_env *env, struct md_device *m,
1618 unsigned int cmd, int len, void *karg)
1620 struct mdd_device *mdd;
1621 struct obd_ioctl_data *data = karg;
1625 mdd = lu2mdd_dev(&m->md_lu_dev);
1627 /* Doesn't use obd_ioctl_data */
1629 case OBD_IOC_CHANGELOG_CLEAR: {
1630 struct changelog_setinfo *cs = karg;
1631 rc = mdd_changelog_user_purge(env, mdd, cs->cs_id,
1635 case OBD_IOC_GET_MNTOPT: {
1636 mntopt_t *mntopts = (mntopt_t *)karg;
1637 *mntopts = mdd->mdd_dt_conf.ddp_mntopts;
1640 case OBD_IOC_START_LFSCK: {
1641 struct lfsck_start *start = karg;
1642 struct md_lfsck *lfsck = &mdd->mdd_lfsck;
1644 /* Return the kernel service version. */
1645 /* XXX: version can be used for compatibility in the future. */
1646 start->ls_version = lfsck->ml_version;
1647 rc = mdd_lfsck_start(env, lfsck, start);
1650 case OBD_IOC_STOP_LFSCK: {
1651 rc = mdd_lfsck_stop(env, &mdd->mdd_lfsck);
1656 /* Below ioctls use obd_ioctl_data */
1657 if (len != sizeof(*data)) {
1658 CERROR("Bad ioctl size %d\n", len);
1661 if (data->ioc_version != OBD_IOCTL_VERSION) {
1662 CERROR("Bad magic %x != %x\n", data->ioc_version,
1668 case OBD_IOC_CHANGELOG_REG:
1669 rc = mdd_changelog_user_register(mdd, &data->ioc_u32_1);
1671 case OBD_IOC_CHANGELOG_DEREG:
1672 rc = mdd_changelog_user_purge(env, mdd, data->ioc_u32_1,
1682 /* type constructor/destructor: mdd_type_init, mdd_type_fini */
1683 LU_TYPE_INIT_FINI(mdd, &mdd_thread_key, &mdd_ucred_key, &mdd_capainfo_key);
1685 const struct md_device_operations mdd_ops = {
1686 .mdo_statfs = mdd_statfs,
1687 .mdo_root_get = mdd_root_get,
1688 .mdo_maxsize_get = mdd_maxsize_get,
1689 .mdo_init_capa_ctxt = mdd_init_capa_ctxt,
1690 .mdo_update_capa_key= mdd_update_capa_key,
1691 .mdo_llog_ctxt_get = mdd_llog_ctxt_get,
1692 .mdo_iocontrol = mdd_iocontrol,
1695 static struct lu_device_type_operations mdd_device_type_ops = {
1696 .ldto_init = mdd_type_init,
1697 .ldto_fini = mdd_type_fini,
1699 .ldto_start = mdd_type_start,
1700 .ldto_stop = mdd_type_stop,
1702 .ldto_device_alloc = mdd_device_alloc,
1703 .ldto_device_free = mdd_device_free,
1705 .ldto_device_fini = mdd_device_fini
1708 static struct lu_device_type mdd_device_type = {
1709 .ldt_tags = LU_DEVICE_MD,
1710 .ldt_name = LUSTRE_MDD_NAME,
1711 .ldt_ops = &mdd_device_type_ops,
1712 .ldt_ctx_tags = LCT_MD_THREAD
1715 /* context key constructor: mdd_key_init */
1716 LU_KEY_INIT(mdd, struct mdd_thread_info);
1718 static void mdd_key_fini(const struct lu_context *ctx,
1719 struct lu_context_key *key, void *data)
1721 struct mdd_thread_info *info = data;
1722 if (info->mti_max_lmm != NULL)
1723 OBD_FREE(info->mti_max_lmm, info->mti_max_lmm_size);
1724 if (info->mti_max_cookie != NULL)
1725 OBD_FREE(info->mti_max_cookie, info->mti_max_cookie_size);
1726 mdd_buf_put(&info->mti_big_buf);
1731 /* context key: mdd_thread_key */
1732 LU_CONTEXT_KEY_DEFINE(mdd, LCT_MD_THREAD);
1734 static struct lu_local_obj_desc llod_capa_key = {
1735 .llod_name = CAPA_KEYS,
1736 .llod_oid = MDD_CAPA_KEYS_OID,
1740 static struct lu_local_obj_desc llod_mdd_orphan = {
1741 .llod_name = orph_index_name,
1742 .llod_oid = MDD_ORPHAN_OID,
1744 .llod_feat = &dt_directory_features,
1747 static struct lu_local_obj_desc llod_mdd_root = {
1748 .llod_name = mdd_root_dir_name,
1749 .llod_oid = MDD_ROOT_INDEX_OID,
1751 .llod_feat = &dt_directory_features,
1754 static struct lu_local_obj_desc llod_lfsck_bookmark = {
1755 .llod_name = lfsck_bookmark_name,
1756 .llod_oid = LFSCK_BOOKMARK_OID,
1760 static int __init mdd_mod_init(void)
1762 struct lprocfs_static_vars lvars;
1765 lprocfs_mdd_init_vars(&lvars);
1767 rc = lu_kmem_init(mdd_caches);
1771 llo_local_obj_register(&llod_capa_key);
1772 llo_local_obj_register(&llod_mdd_orphan);
1773 llo_local_obj_register(&llod_mdd_root);
1774 llo_local_obj_register(&llod_lfsck_bookmark);
1776 rc = class_register_type(&mdd_obd_device_ops, NULL, lvars.module_vars,
1777 LUSTRE_MDD_NAME, &mdd_device_type);
1779 lu_kmem_fini(mdd_caches);
1783 static void __exit mdd_mod_exit(void)
1785 llo_local_obj_unregister(&llod_capa_key);
1786 llo_local_obj_unregister(&llod_mdd_orphan);
1787 llo_local_obj_unregister(&llod_mdd_root);
1788 llo_local_obj_unregister(&llod_lfsck_bookmark);
1790 class_unregister_type(LUSTRE_MDD_NAME);
1791 lu_kmem_fini(mdd_caches);
1794 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
1795 MODULE_DESCRIPTION("Lustre Meta-data Device Prototype ("LUSTRE_MDD_NAME")");
1796 MODULE_LICENSE("GPL");
1798 cfs_module(mdd, "0.1.0", mdd_mod_init, mdd_mod_exit);