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) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2012, 2015, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/osd-zfs/osd_index.c
38 * Author: Alex Zhuravlev <bzzz@whamcloud.com>
39 * Author: Mike Pershin <tappro@whamcloud.com>
42 #define DEBUG_SUBSYSTEM S_OSD
44 #include <lustre_ver.h>
45 #include <libcfs/libcfs.h>
46 #include <obd_support.h>
47 #include <lustre_net.h>
49 #include <obd_class.h>
50 #include <lustre_disk.h>
51 #include <lustre_fid.h>
53 #include "osd_internal.h"
55 #include <sys/dnode.h>
60 #include <sys/spa_impl.h>
61 #include <sys/zfs_znode.h>
62 #include <sys/dmu_tx.h>
63 #include <sys/dmu_objset.h>
64 #include <sys/dsl_prop.h>
65 #include <sys/sa_impl.h>
68 static inline int osd_object_is_zap(dmu_buf_t *db)
70 dmu_buf_impl_t *dbi = (dmu_buf_impl_t *) db;
76 rc = (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
77 dn->dn_type == DMU_OT_USERGROUP_USED);
83 /* We don't actually have direct access to the zap_hashbits() function
84 * so just pretend like we do for now. If this ever breaks we can look at
86 #define zap_hashbits(zc) 48
89 * | cd (16 bits) | hash (48 bits) |
90 * we need it in other form:
91 * |0| hash (48 bit) | cd (15 bit) |
92 * to be a full 64-bit ordered hash so that Lustre readdir can use it to merge
93 * the readdir hashes from multiple directory stripes uniformly on the client.
94 * Another point is sign bit, the hash range should be in [0, 2^63-1] because
95 * loff_t (for llseek) needs to be a positive value. This means the "cd" field
96 * should only be the low 15 bits.
98 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc)
100 uint64_t zfs_hash = zap_cursor_serialize(zc) & (~0ULL >> 1);
102 return (zfs_hash >> zap_hashbits(zc)) |
103 (zfs_hash << (63 - zap_hashbits(zc)));
106 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
107 uint64_t id, uint64_t dirhash)
109 uint64_t zfs_hash = ((dirhash << zap_hashbits(zc)) & (~0ULL >> 1)) |
110 (dirhash >> (63 - zap_hashbits(zc)));
112 zap_cursor_init_serialized(zc, os, id, zfs_hash);
115 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
116 uint64_t id, uint64_t dirhash)
121 if (unlikely(t == NULL))
124 osd_zap_cursor_init_serialized(t, os, id, dirhash);
130 void osd_zap_cursor_fini(zap_cursor_t *zc)
136 static inline void osd_obj_cursor_init_serialized(zap_cursor_t *zc,
137 struct osd_object *o,
140 struct osd_device *d = osd_obj2dev(o);
141 osd_zap_cursor_init_serialized(zc, d->od_os,
142 o->oo_db->db_object, dirhash);
145 static inline int osd_obj_cursor_init(zap_cursor_t **zc, struct osd_object *o,
148 struct osd_device *d = osd_obj2dev(o);
149 return osd_zap_cursor_init(zc, d->od_os, o->oo_db->db_object, dirhash);
152 static struct dt_it *osd_index_it_init(const struct lu_env *env,
153 struct dt_object *dt,
156 struct osd_thread_info *info = osd_oti_get(env);
157 struct osd_zap_it *it;
158 struct osd_object *obj = osd_dt_obj(dt);
159 struct lu_object *lo = &dt->do_lu;
163 if (obj->oo_destroyed)
164 RETURN(ERR_PTR(-ENOENT));
166 LASSERT(lu_object_exists(lo));
168 LASSERT(osd_object_is_zap(obj->oo_db));
171 OBD_SLAB_ALLOC_PTR_GFP(it, osd_zapit_cachep, GFP_NOFS);
173 RETURN(ERR_PTR(-ENOMEM));
175 rc = osd_obj_cursor_init(&it->ozi_zc, obj, 0);
177 OBD_SLAB_FREE_PTR(it, osd_zapit_cachep);
185 RETURN((struct dt_it *)it);
188 static void osd_index_it_fini(const struct lu_env *env, struct dt_it *di)
190 struct osd_zap_it *it = (struct osd_zap_it *)di;
191 struct osd_object *obj;
195 LASSERT(it->ozi_obj);
199 osd_zap_cursor_fini(it->ozi_zc);
200 lu_object_put(env, &obj->oo_dt.do_lu);
201 OBD_SLAB_FREE_PTR(it, osd_zapit_cachep);
207 static void osd_index_it_put(const struct lu_env *env, struct dt_it *di)
209 /* PBS: do nothing : ref are incremented at retrive and decreamented
213 static inline void osd_it_append_attrs(struct lu_dirent *ent, __u32 attr,
216 const unsigned align = sizeof(struct luda_type) - 1;
217 struct luda_type *lt;
219 /* check if file type is required */
220 if (attr & LUDA_TYPE) {
221 len = (len + align) & ~align;
223 lt = (void *)ent->lde_name + len;
224 lt->lt_type = cpu_to_le16(DTTOIF(type));
225 ent->lde_attrs |= LUDA_TYPE;
228 ent->lde_attrs = cpu_to_le32(ent->lde_attrs);
232 * as we don't know FID, we can't use LU object, so this function
233 * partially duplicate __osd_xattr_get() which is built around
234 * LU-object and uses it to cache data like regular EA dnode, etc
236 static int osd_find_parent_by_dnode(const struct lu_env *env,
240 struct osd_device *osd = osd_obj2dev(osd_dt_obj(o));
241 struct lustre_mdt_attrs *lma;
244 nvlist_t *nvbuf = NULL;
250 /* first of all, get parent dnode from own attributes */
251 LASSERT(osd_dt_obj(o)->oo_db);
252 rc = -sa_handle_get(osd->od_os, osd_dt_obj(o)->oo_db->db_object,
253 NULL, SA_HDL_PRIVATE, &sa_hdl);
257 dnode = ZFS_NO_OBJECT;
258 rc = -sa_lookup(sa_hdl, SA_ZPL_PARENT(osd), &dnode, 8);
259 sa_handle_destroy(sa_hdl);
263 /* now get EA buffer */
264 rc = __osd_xattr_load(osd, dnode, &nvbuf);
268 /* XXX: if we get that far.. should we cache the result? */
270 /* try to find LMA attribute */
271 LASSERT(nvbuf != NULL);
272 rc = -nvlist_lookup_byte_array(nvbuf, XATTR_NAME_LMA, &value, &size);
273 if (rc == 0 && size >= sizeof(*lma)) {
274 lma = (struct lustre_mdt_attrs *)value;
275 lustre_lma_swab(lma);
276 *fid = lma->lma_self_fid;
281 /* no LMA attribute in SA, let's try regular EA */
283 /* first of all, get parent dnode storing regular EA */
284 rc = -sa_handle_get(osd->od_os, dnode, NULL, SA_HDL_PRIVATE, &sa_hdl);
288 dnode = ZFS_NO_OBJECT;
289 rc = -sa_lookup(sa_hdl, SA_ZPL_XATTR(osd), &dnode, 8);
290 sa_handle_destroy(sa_hdl);
294 CLASSERT(sizeof(*lma) <= sizeof(osd_oti_get(env)->oti_buf));
295 buf.lb_buf = osd_oti_get(env)->oti_buf;
296 buf.lb_len = sizeof(osd_oti_get(env)->oti_buf);
298 /* now try to find LMA */
299 rc = __osd_xattr_get_large(env, osd, dnode, &buf,
300 XATTR_NAME_LMA, &size);
301 if (rc == 0 && size >= sizeof(*lma)) {
303 lustre_lma_swab(lma);
304 *fid = lma->lma_self_fid;
309 GOTO(out, rc = -EIO);
318 static int osd_find_parent_fid(const struct lu_env *env, struct dt_object *o,
321 struct link_ea_header *leh;
322 struct link_ea_entry *lee;
327 buf.lb_buf = osd_oti_get(env)->oti_buf;
328 buf.lb_len = sizeof(osd_oti_get(env)->oti_buf);
330 rc = osd_xattr_get(env, o, &buf, XATTR_NAME_LINK);
332 rc = osd_xattr_get(env, o, &LU_BUF_NULL, XATTR_NAME_LINK);
336 OBD_ALLOC(buf.lb_buf, rc);
337 if (buf.lb_buf == NULL)
340 rc = osd_xattr_get(env, o, &buf, XATTR_NAME_LINK);
344 if (rc < sizeof(*leh) + sizeof(*lee))
345 GOTO(out, rc = -EINVAL);
348 if (leh->leh_magic == __swab32(LINK_EA_MAGIC)) {
349 leh->leh_magic = LINK_EA_MAGIC;
350 leh->leh_reccount = __swab32(leh->leh_reccount);
351 leh->leh_len = __swab64(leh->leh_len);
353 if (leh->leh_magic != LINK_EA_MAGIC)
354 GOTO(out, rc = -EINVAL);
355 if (leh->leh_reccount == 0)
356 GOTO(out, rc = -ENODATA);
358 lee = (struct link_ea_entry *)(leh + 1);
359 fid_be_to_cpu(fid, (const struct lu_fid *)&lee->lee_parent_fid);
363 if (buf.lb_buf != osd_oti_get(env)->oti_buf)
364 OBD_FREE(buf.lb_buf, buf.lb_len);
367 /* this block can be enabled for additional verification
368 * it's trying to match FID from LinkEA vs. FID from LMA */
372 rc2 = osd_find_parent_by_dnode(env, o, &fid2);
374 if (lu_fid_eq(fid, &fid2) == 0)
375 CERROR("wrong parent: "DFID" != "DFID"\n",
376 PFID(fid), PFID(&fid2));
380 /* no LinkEA is found, let's try to find the fid in parent's LMA */
381 if (unlikely(rc != 0))
382 rc = osd_find_parent_by_dnode(env, o, fid);
387 static int osd_dir_lookup(const struct lu_env *env, struct dt_object *dt,
388 struct dt_rec *rec, const struct dt_key *key)
390 struct osd_thread_info *oti = osd_oti_get(env);
391 struct osd_object *obj = osd_dt_obj(dt);
392 struct osd_device *osd = osd_obj2dev(obj);
393 char *name = (char *)key;
397 LASSERT(osd_object_is_zap(obj->oo_db));
399 if (name[0] == '.') {
401 const struct lu_fid *f = lu_object_fid(&dt->do_lu);
402 memcpy(rec, f, sizeof(*f));
404 } else if (name[1] == '.' && name[2] == 0) {
405 rc = osd_find_parent_fid(env, dt, (struct lu_fid *)rec);
406 RETURN(rc == 0 ? 1 : rc);
410 rc = -zap_lookup(osd->od_os, obj->oo_db->db_object,
411 (char *)key, 8, sizeof(oti->oti_zde) / 8,
412 (void *)&oti->oti_zde);
413 memcpy(rec, &oti->oti_zde.lzd_fid, sizeof(struct lu_fid));
415 RETURN(rc == 0 ? 1 : rc);
418 static int osd_declare_dir_insert(const struct lu_env *env,
419 struct dt_object *dt,
420 const struct dt_rec *rec,
421 const struct dt_key *key,
424 struct osd_object *obj = osd_dt_obj(dt);
425 struct osd_thandle *oh;
430 oh = container_of0(th, struct osd_thandle, ot_super);
432 /* This is for inserting dot/dotdot for new created dir. */
433 if (obj->oo_db == NULL)
434 object = DMU_NEW_OBJECT;
436 object = obj->oo_db->db_object;
438 dmu_tx_hold_bonus(oh->ot_tx, object);
439 dmu_tx_hold_zap(oh->ot_tx, object, TRUE, (char *)key);
445 * Find the osd object for given fid.
447 * \param fid need to find the osd object having this fid
449 * \retval osd_object on success
450 * \retval -ve on error
452 struct osd_object *osd_object_find(const struct lu_env *env,
453 struct dt_object *dt,
454 const struct lu_fid *fid)
456 struct lu_device *ludev = dt->do_lu.lo_dev;
457 struct osd_object *child = NULL;
458 struct lu_object *luch;
459 struct lu_object *lo;
462 * at this point topdev might not exist yet
463 * (i.e. MGS is preparing profiles). so we can
464 * not rely on topdev and instead lookup with
465 * our device passed as topdev. this can't work
466 * if the object isn't cached yet (as osd doesn't
467 * allocate lu_header). IOW, the object must be
468 * in the cache, otherwise lu_object_alloc() crashes
471 luch = lu_object_find_at(env, ludev, fid, NULL);
475 if (lu_object_exists(luch)) {
476 lo = lu_object_locate(luch->lo_header, ludev->ld_type);
480 LU_OBJECT_DEBUG(D_ERROR, env, luch,
481 "%s: object can't be located "DFID"\n",
482 osd_dev(ludev)->od_svname, PFID(fid));
485 lu_object_put(env, luch);
486 CERROR("%s: Unable to get osd_object "DFID"\n",
487 osd_dev(ludev)->od_svname, PFID(fid));
488 child = ERR_PTR(-ENOENT);
491 LU_OBJECT_DEBUG(D_ERROR, env, luch,
492 "%s: lu_object does not exists "DFID"\n",
493 osd_dev(ludev)->od_svname, PFID(fid));
494 lu_object_put(env, luch);
495 child = ERR_PTR(-ENOENT);
502 * Put the osd object once done with it.
504 * \param obj osd object that needs to be put
506 static inline void osd_object_put(const struct lu_env *env,
507 struct osd_object *obj)
509 lu_object_put(env, &obj->oo_dt.do_lu);
512 static int osd_seq_exists(const struct lu_env *env, struct osd_device *osd,
515 struct lu_seq_range *range = &osd_oti_get(env)->oti_seq_range;
516 struct seq_server_site *ss = osd_seq_site(osd);
521 LASSERT(ss->ss_server_fld != NULL);
523 rc = osd_fld_lookup(env, osd, seq, range);
526 CERROR("%s: Can not lookup fld for "LPX64"\n",
531 RETURN(ss->ss_node_id == range->lsr_index);
534 static int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
535 const struct lu_fid *fid)
537 struct seq_server_site *ss = osd_seq_site(osd);
540 /* FID seqs not in FLDB, must be local seq */
541 if (unlikely(!fid_seq_in_fldb(fid_seq(fid))))
544 /* If FLD is not being initialized yet, it only happens during the
545 * initialization, likely during mgs initialization, and we assume
546 * this is local FID. */
547 if (ss == NULL || ss->ss_server_fld == NULL)
550 /* Only check the local FLDB here */
551 if (osd_seq_exists(env, osd, fid_seq(fid)))
558 * Inserts (key, value) pair in \a directory object.
560 * \param dt osd index object
561 * \param key key for index
562 * \param rec record reference
563 * \param th transaction handler
564 * \param ignore_quota update should not affect quota
567 * \retval -ve failure
569 static int osd_dir_insert(const struct lu_env *env, struct dt_object *dt,
570 const struct dt_rec *rec, const struct dt_key *key,
571 struct thandle *th, int ignore_quota)
573 struct osd_thread_info *oti = osd_oti_get(env);
574 struct osd_object *parent = osd_dt_obj(dt);
575 struct osd_device *osd = osd_obj2dev(parent);
576 struct dt_insert_rec *rec1 = (struct dt_insert_rec *)rec;
577 const struct lu_fid *fid = rec1->rec_fid;
578 struct osd_thandle *oh;
579 struct osd_object *child = NULL;
581 char *name = (char *)key;
585 LASSERT(parent->oo_db);
586 LASSERT(osd_object_is_zap(parent->oo_db));
588 LASSERT(dt_object_exists(dt));
589 LASSERT(osd_invariant(parent));
592 oh = container_of0(th, struct osd_thandle, ot_super);
594 rc = osd_remote_fid(env, osd, fid);
596 CERROR("%s: Can not find object "DFID": rc = %d\n",
597 osd->od_svname, PFID(fid), rc);
601 if (unlikely(rc == 1)) {
602 /* Insert remote entry */
603 memset(&oti->oti_zde.lzd_reg, 0, sizeof(oti->oti_zde.lzd_reg));
604 oti->oti_zde.lzd_reg.zde_type = IFTODT(rec1->rec_type & S_IFMT);
607 * To simulate old Orion setups with ./.. stored in the
610 /* Insert local entry */
611 child = osd_object_find(env, dt, fid);
613 RETURN(PTR_ERR(child));
615 LASSERT(child->oo_db);
616 if (name[0] == '.') {
618 /* do not store ".", instead generate it
619 * during iteration */
621 } else if (name[1] == '.' && name[2] == 0) {
622 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_BAD_PARENT)) {
623 struct lu_fid tfid = *fid;
625 osd_object_put(env, child);
627 child = osd_object_find(env, dt, &tfid);
629 RETURN(PTR_ERR(child));
631 LASSERT(child->oo_db);
634 /* update parent dnode in the child.
635 * later it will be used to generate ".." */
636 rc = osd_object_sa_update(parent,
638 &child->oo_db->db_object,
644 CLASSERT(sizeof(oti->oti_zde.lzd_reg) == 8);
645 CLASSERT(sizeof(oti->oti_zde) % 8 == 0);
646 attr = child->oo_dt.do_lu.lo_header ->loh_attr;
647 oti->oti_zde.lzd_reg.zde_type = IFTODT(attr & S_IFMT);
648 oti->oti_zde.lzd_reg.zde_dnode = child->oo_db->db_object;
651 oti->oti_zde.lzd_fid = *fid;
652 /* Insert (key,oid) into ZAP */
653 rc = -zap_add(osd->od_os, parent->oo_db->db_object,
654 (char *)key, 8, sizeof(oti->oti_zde) / 8,
655 (void *)&oti->oti_zde, oh->ot_tx);
656 if (unlikely(rc == -EEXIST &&
657 name[0] == '.' && name[1] == '.' && name[2] == 0))
658 /* Update (key,oid) in ZAP */
659 rc = -zap_update(osd->od_os, parent->oo_db->db_object,
660 (char *)key, 8, sizeof(oti->oti_zde) / 8,
661 (void *)&oti->oti_zde, oh->ot_tx);
665 osd_object_put(env, child);
670 static int osd_declare_dir_delete(const struct lu_env *env,
671 struct dt_object *dt,
672 const struct dt_key *key,
675 struct osd_object *obj = osd_dt_obj(dt);
676 struct osd_thandle *oh;
680 LASSERT(dt_object_exists(dt));
681 LASSERT(osd_invariant(obj));
684 oh = container_of0(th, struct osd_thandle, ot_super);
686 if (dt_object_exists(dt)) {
688 LASSERT(osd_object_is_zap(obj->oo_db));
689 dnode = obj->oo_db->db_object;
691 dnode = DMU_NEW_OBJECT;
693 dmu_tx_hold_zap(oh->ot_tx, dnode, TRUE, (char *)key);
698 static int osd_dir_delete(const struct lu_env *env, struct dt_object *dt,
699 const struct dt_key *key, struct thandle *th)
701 struct osd_object *obj = osd_dt_obj(dt);
702 struct osd_device *osd = osd_obj2dev(obj);
703 struct osd_thandle *oh;
704 dmu_buf_t *zap_db = obj->oo_db;
705 char *name = (char *)key;
710 LASSERT(osd_object_is_zap(zap_db));
713 oh = container_of0(th, struct osd_thandle, ot_super);
716 * In Orion . and .. were stored in the directory (not generated upon
717 * request as now). we preserve them for backward compatibility
719 if (name[0] == '.') {
722 } else if (name[1] == '.' && name[2] == 0) {
727 /* Remove key from the ZAP */
728 rc = -zap_remove(osd->od_os, zap_db->db_object,
729 (char *) key, oh->ot_tx);
731 if (unlikely(rc && rc != -ENOENT))
732 CERROR("%s: zap_remove failed: rc = %d\n", osd->od_svname, rc);
737 static struct dt_it *osd_dir_it_init(const struct lu_env *env,
738 struct dt_object *dt,
741 struct osd_zap_it *it;
743 it = (struct osd_zap_it *)osd_index_it_init(env, dt, unused);
747 RETURN((struct dt_it *)it);
751 * Move Iterator to record specified by \a key
753 * \param di osd iterator
754 * \param key key for index
756 * \retval +ve di points to record with least key not larger than key
757 * \retval 0 di points to exact matched key
758 * \retval -ve failure
760 static int osd_dir_it_get(const struct lu_env *env,
761 struct dt_it *di, const struct dt_key *key)
763 struct osd_zap_it *it = (struct osd_zap_it *)di;
764 struct osd_object *obj = it->ozi_obj;
765 char *name = (char *)key;
772 /* reset the cursor */
773 zap_cursor_fini(it->ozi_zc);
774 osd_obj_cursor_init_serialized(it->ozi_zc, obj, 0);
776 /* XXX: implementation of the API is broken at the moment */
777 LASSERT(((const char *)key)[0] == 0);
784 if (name[0] == '.') {
788 } else if (name[1] == '.' && name[2] == 0) {
794 /* neither . nor .. - some real record */
802 static void osd_dir_it_put(const struct lu_env *env, struct dt_it *di)
804 /* PBS: do nothing : ref are incremented at retrive and decreamented
809 * in Orion . and .. were stored in the directory, while ZPL
810 * and current osd-zfs generate them up on request. so, we
811 * need to ignore previously stored . and ..
813 static int osd_index_retrieve_skip_dots(struct osd_zap_it *it,
819 rc = -zap_cursor_retrieve(it->ozi_zc, za);
822 if (unlikely(rc == 0 && za->za_name[0] == '.')) {
823 if (za->za_name[1] == 0) {
825 } else if (za->za_name[1] == '.' &&
826 za->za_name[2] == 0) {
830 zap_cursor_advance(it->ozi_zc);
832 } while (unlikely(rc == 0 && isdot));
838 * to load a directory entry at a time and stored it in
839 * iterator's in-memory data structure.
841 * \param di, struct osd_it_ea, iterator's in memory structure
843 * \retval +ve, iterator reached to end
844 * \retval 0, iterator not reached to end
845 * \retval -ve, on error
847 static int osd_dir_it_next(const struct lu_env *env, struct dt_it *di)
849 struct osd_zap_it *it = (struct osd_zap_it *)di;
850 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
855 /* temp. storage should be enough for any key supported by ZFS */
856 CLASSERT(sizeof(za->za_name) <= sizeof(it->ozi_name));
859 * the first ->next() moves the cursor to .
860 * the second ->next() moves the cursor to ..
861 * then we get to the real records and have to verify any exist
863 if (it->ozi_pos <= 2) {
869 zap_cursor_advance(it->ozi_zc);
873 * According to current API we need to return error if its last entry.
874 * zap_cursor_advance() does not return any value. So we need to call
875 * retrieve to check if there is any record. We should make
876 * changes to Iterator API to not return status for this API
878 rc = osd_index_retrieve_skip_dots(it, za);
880 if (rc == -ENOENT) /* end of dir */
886 static struct dt_key *osd_dir_it_key(const struct lu_env *env,
887 const struct dt_it *di)
889 struct osd_zap_it *it = (struct osd_zap_it *)di;
890 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
894 if (it->ozi_pos <= 1) {
896 RETURN((struct dt_key *)".");
897 } else if (it->ozi_pos == 2) {
898 RETURN((struct dt_key *)"..");
901 if ((rc = -zap_cursor_retrieve(it->ozi_zc, za)))
904 strcpy(it->ozi_name, za->za_name);
906 RETURN((struct dt_key *)it->ozi_name);
909 static int osd_dir_it_key_size(const struct lu_env *env, const struct dt_it *di)
911 struct osd_zap_it *it = (struct osd_zap_it *)di;
912 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
916 if (it->ozi_pos <= 1) {
919 } else if (it->ozi_pos == 2) {
923 if ((rc = -zap_cursor_retrieve(it->ozi_zc, za)) == 0)
924 rc = strlen(za->za_name);
929 static int osd_dir_it_rec(const struct lu_env *env, const struct dt_it *di,
930 struct dt_rec *dtrec, __u32 attr)
932 struct osd_zap_it *it = (struct osd_zap_it *)di;
933 struct lu_dirent *lde = (struct lu_dirent *)dtrec;
934 struct luz_direntry *zde = &osd_oti_get(env)->oti_zde;
935 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
939 if (it->ozi_pos <= 1) {
940 lde->lde_hash = cpu_to_le64(1);
941 strcpy(lde->lde_name, ".");
942 lde->lde_namelen = cpu_to_le16(1);
943 lde->lde_fid = *lu_object_fid(&it->ozi_obj->oo_dt.do_lu);
944 lde->lde_attrs = LUDA_FID;
945 /* append lustre attributes */
946 osd_it_append_attrs(lde, attr, 1, IFTODT(S_IFDIR));
947 lde->lde_reclen = cpu_to_le16(lu_dirent_calc_size(1, attr));
951 } else if (it->ozi_pos == 2) {
952 lde->lde_hash = cpu_to_le64(2);
953 strcpy(lde->lde_name, "..");
954 lde->lde_namelen = cpu_to_le16(2);
955 lde->lde_attrs = LUDA_FID;
956 /* append lustre attributes */
957 osd_it_append_attrs(lde, attr, 2, IFTODT(S_IFDIR));
958 lde->lde_reclen = cpu_to_le16(lu_dirent_calc_size(2, attr));
959 rc = osd_find_parent_fid(env, &it->ozi_obj->oo_dt, &lde->lde_fid);
961 /* ENOENT happens at the root of filesystem so ignore it */
969 rc = -zap_cursor_retrieve(it->ozi_zc, za);
970 if (unlikely(rc != 0))
973 lde->lde_hash = cpu_to_le64(osd_zap_cursor_serialize(it->ozi_zc));
974 namelen = strlen(za->za_name);
975 if (namelen > NAME_MAX)
976 GOTO(out, rc = -EOVERFLOW);
977 strcpy(lde->lde_name, za->za_name);
978 lde->lde_namelen = cpu_to_le16(namelen);
980 if (za->za_integer_length != 8 || za->za_num_integers < 3) {
981 CERROR("%s: unsupported direntry format: %d %d\n",
982 osd_obj2dev(it->ozi_obj)->od_svname,
983 za->za_integer_length, (int)za->za_num_integers);
985 GOTO(out, rc = -EIO);
988 rc = -zap_lookup(it->ozi_zc->zc_objset, it->ozi_zc->zc_zapobj,
989 za->za_name, za->za_integer_length, 3, zde);
993 lde->lde_fid = zde->lzd_fid;
994 lde->lde_attrs = LUDA_FID;
996 /* append lustre attributes */
997 osd_it_append_attrs(lde, attr, namelen, zde->lzd_reg.zde_type);
999 lde->lde_reclen = cpu_to_le16(lu_dirent_calc_size(namelen, attr));
1005 static int osd_dir_it_rec_size(const struct lu_env *env, const struct dt_it *di,
1008 struct osd_zap_it *it = (struct osd_zap_it *)di;
1009 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1014 if (it->ozi_pos <= 1)
1016 else if (it->ozi_pos == 2)
1020 rc = lu_dirent_calc_size(namelen, attr);
1024 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1025 if (unlikely(rc != 0))
1028 if (za->za_integer_length != 8 || za->za_num_integers < 3) {
1029 CERROR("%s: unsupported direntry format: %d %d\n",
1030 osd_obj2dev(it->ozi_obj)->od_svname,
1031 za->za_integer_length, (int)za->za_num_integers);
1035 namelen = strlen(za->za_name);
1036 if (namelen > NAME_MAX)
1039 rc = lu_dirent_calc_size(namelen, attr);
1044 static __u64 osd_dir_it_store(const struct lu_env *env, const struct dt_it *di)
1046 struct osd_zap_it *it = (struct osd_zap_it *)di;
1050 if (it->ozi_pos <= 2)
1053 pos = osd_zap_cursor_serialize(it->ozi_zc);
1060 * rc == 0 -> end of directory.
1061 * rc > 0 -> ok, proceed.
1062 * rc < 0 -> error. ( EOVERFLOW can be masked.)
1064 static int osd_dir_it_load(const struct lu_env *env,
1065 const struct dt_it *di, __u64 hash)
1067 struct osd_zap_it *it = (struct osd_zap_it *)di;
1068 struct osd_object *obj = it->ozi_obj;
1069 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1073 /* reset the cursor */
1074 zap_cursor_fini(it->ozi_zc);
1075 osd_obj_cursor_init_serialized(it->ozi_zc, obj, hash);
1082 /* to return whether the end has been reached */
1083 rc = osd_index_retrieve_skip_dots(it, za);
1086 else if (rc == -ENOENT)
1093 struct dt_index_operations osd_dir_ops = {
1094 .dio_lookup = osd_dir_lookup,
1095 .dio_declare_insert = osd_declare_dir_insert,
1096 .dio_insert = osd_dir_insert,
1097 .dio_declare_delete = osd_declare_dir_delete,
1098 .dio_delete = osd_dir_delete,
1100 .init = osd_dir_it_init,
1101 .fini = osd_index_it_fini,
1102 .get = osd_dir_it_get,
1103 .put = osd_dir_it_put,
1104 .next = osd_dir_it_next,
1105 .key = osd_dir_it_key,
1106 .key_size = osd_dir_it_key_size,
1107 .rec = osd_dir_it_rec,
1108 .rec_size = osd_dir_it_rec_size,
1109 .store = osd_dir_it_store,
1110 .load = osd_dir_it_load
1115 * Primitives for index files using binary keys.
1118 /* key integer_size is 8 */
1119 static int osd_prepare_key_uint64(struct osd_object *o, __u64 *dst,
1120 const struct dt_key *src)
1127 /* align keysize to 64bit */
1128 size = (o->oo_keysize + sizeof(__u64) - 1) / sizeof(__u64);
1129 size *= sizeof(__u64);
1131 LASSERT(size <= MAXNAMELEN);
1133 if (unlikely(size > o->oo_keysize))
1134 memset(dst + o->oo_keysize, 0, size - o->oo_keysize);
1135 memcpy(dst, (const char *)src, o->oo_keysize);
1137 return (size/sizeof(__u64));
1140 static int osd_index_lookup(const struct lu_env *env, struct dt_object *dt,
1141 struct dt_rec *rec, const struct dt_key *key)
1143 struct osd_object *obj = osd_dt_obj(dt);
1144 struct osd_device *osd = osd_obj2dev(obj);
1145 __u64 *k = osd_oti_get(env)->oti_key64;
1149 rc = osd_prepare_key_uint64(obj, k, key);
1151 rc = -zap_lookup_uint64(osd->od_os, obj->oo_db->db_object,
1152 k, rc, obj->oo_recusize, obj->oo_recsize,
1154 RETURN(rc == 0 ? 1 : rc);
1157 static int osd_declare_index_insert(const struct lu_env *env,
1158 struct dt_object *dt,
1159 const struct dt_rec *rec,
1160 const struct dt_key *key,
1163 struct osd_object *obj = osd_dt_obj(dt);
1164 struct osd_thandle *oh;
1167 LASSERT(th != NULL);
1168 oh = container_of0(th, struct osd_thandle, ot_super);
1170 LASSERT(obj->oo_db);
1172 dmu_tx_hold_bonus(oh->ot_tx, obj->oo_db->db_object);
1174 /* It is not clear what API should be used for binary keys, so we pass
1175 * a null name which has the side effect of over-reserving space,
1176 * accounting for the worst case. See zap_count_write() */
1177 dmu_tx_hold_zap(oh->ot_tx, obj->oo_db->db_object, TRUE, NULL);
1182 static int osd_index_insert(const struct lu_env *env, struct dt_object *dt,
1183 const struct dt_rec *rec, const struct dt_key *key,
1184 struct thandle *th, int ignore_quota)
1186 struct osd_object *obj = osd_dt_obj(dt);
1187 struct osd_device *osd = osd_obj2dev(obj);
1188 struct osd_thandle *oh;
1189 __u64 *k = osd_oti_get(env)->oti_key64;
1193 LASSERT(obj->oo_db);
1194 LASSERT(dt_object_exists(dt));
1195 LASSERT(osd_invariant(obj));
1196 LASSERT(th != NULL);
1198 oh = container_of0(th, struct osd_thandle, ot_super);
1200 rc = osd_prepare_key_uint64(obj, k, key);
1202 /* Insert (key,oid) into ZAP */
1203 rc = -zap_add_uint64(osd->od_os, obj->oo_db->db_object,
1204 k, rc, obj->oo_recusize, obj->oo_recsize,
1205 (void *)rec, oh->ot_tx);
1209 static int osd_declare_index_delete(const struct lu_env *env,
1210 struct dt_object *dt,
1211 const struct dt_key *key,
1214 struct osd_object *obj = osd_dt_obj(dt);
1215 struct osd_thandle *oh;
1218 LASSERT(dt_object_exists(dt));
1219 LASSERT(osd_invariant(obj));
1220 LASSERT(th != NULL);
1221 LASSERT(obj->oo_db);
1223 oh = container_of0(th, struct osd_thandle, ot_super);
1224 dmu_tx_hold_zap(oh->ot_tx, obj->oo_db->db_object, TRUE, NULL);
1229 static int osd_index_delete(const struct lu_env *env, struct dt_object *dt,
1230 const struct dt_key *key, struct thandle *th)
1232 struct osd_object *obj = osd_dt_obj(dt);
1233 struct osd_device *osd = osd_obj2dev(obj);
1234 struct osd_thandle *oh;
1235 __u64 *k = osd_oti_get(env)->oti_key64;
1239 LASSERT(obj->oo_db);
1240 LASSERT(th != NULL);
1241 oh = container_of0(th, struct osd_thandle, ot_super);
1243 rc = osd_prepare_key_uint64(obj, k, key);
1245 /* Remove binary key from the ZAP */
1246 rc = -zap_remove_uint64(osd->od_os, obj->oo_db->db_object,
1251 static int osd_index_it_get(const struct lu_env *env, struct dt_it *di,
1252 const struct dt_key *key)
1254 struct osd_zap_it *it = (struct osd_zap_it *)di;
1255 struct osd_object *obj = it->ozi_obj;
1256 struct osd_device *osd = osd_obj2dev(obj);
1260 LASSERT(it->ozi_zc);
1263 * XXX: we need a binary version of zap_cursor_move_to_key()
1264 * to implement this API */
1265 if (*((const __u64 *)key) != 0)
1266 CERROR("NOT IMPLEMETED YET (move to "LPX64")\n",
1269 zap_cursor_fini(it->ozi_zc);
1270 zap_cursor_init(it->ozi_zc, osd->od_os, obj->oo_db->db_object);
1276 static int osd_index_it_next(const struct lu_env *env, struct dt_it *di)
1278 struct osd_zap_it *it = (struct osd_zap_it *)di;
1279 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1283 if (it->ozi_reset == 0)
1284 zap_cursor_advance(it->ozi_zc);
1288 * According to current API we need to return error if it's last entry.
1289 * zap_cursor_advance() does not return any value. So we need to call
1290 * retrieve to check if there is any record. We should make
1291 * changes to Iterator API to not return status for this API
1293 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1300 static struct dt_key *osd_index_it_key(const struct lu_env *env,
1301 const struct dt_it *di)
1303 struct osd_zap_it *it = (struct osd_zap_it *)di;
1304 struct osd_object *obj = it->ozi_obj;
1305 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1310 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1312 RETURN(ERR_PTR(rc));
1314 /* the binary key is stored in the name */
1315 memcpy(&it->ozi_key, za->za_name, obj->oo_keysize);
1317 RETURN((struct dt_key *)&it->ozi_key);
1320 static int osd_index_it_key_size(const struct lu_env *env,
1321 const struct dt_it *di)
1323 struct osd_zap_it *it = (struct osd_zap_it *)di;
1324 struct osd_object *obj = it->ozi_obj;
1325 RETURN(obj->oo_keysize);
1328 static int osd_index_it_rec(const struct lu_env *env, const struct dt_it *di,
1329 struct dt_rec *rec, __u32 attr)
1331 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1332 struct osd_zap_it *it = (struct osd_zap_it *)di;
1333 struct osd_object *obj = it->ozi_obj;
1334 struct osd_device *osd = osd_obj2dev(obj);
1335 __u64 *k = osd_oti_get(env)->oti_key64;
1340 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1344 rc = osd_prepare_key_uint64(obj, k, (const struct dt_key *)za->za_name);
1346 rc = -zap_lookup_uint64(osd->od_os, obj->oo_db->db_object,
1347 k, rc, obj->oo_recusize, obj->oo_recsize,
1352 static __u64 osd_index_it_store(const struct lu_env *env,
1353 const struct dt_it *di)
1355 struct osd_zap_it *it = (struct osd_zap_it *)di;
1358 RETURN((__u64)zap_cursor_serialize(it->ozi_zc));
1361 static int osd_index_it_load(const struct lu_env *env, const struct dt_it *di,
1364 struct osd_zap_it *it = (struct osd_zap_it *)di;
1365 struct osd_object *obj = it->ozi_obj;
1366 struct osd_device *osd = osd_obj2dev(obj);
1367 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1371 /* reset the cursor */
1372 zap_cursor_fini(it->ozi_zc);
1373 zap_cursor_init_serialized(it->ozi_zc, osd->od_os,
1374 obj->oo_db->db_object, hash);
1377 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1380 else if (rc == -ENOENT)
1386 static struct dt_index_operations osd_index_ops = {
1387 .dio_lookup = osd_index_lookup,
1388 .dio_declare_insert = osd_declare_index_insert,
1389 .dio_insert = osd_index_insert,
1390 .dio_declare_delete = osd_declare_index_delete,
1391 .dio_delete = osd_index_delete,
1393 .init = osd_index_it_init,
1394 .fini = osd_index_it_fini,
1395 .get = osd_index_it_get,
1396 .put = osd_index_it_put,
1397 .next = osd_index_it_next,
1398 .key = osd_index_it_key,
1399 .key_size = osd_index_it_key_size,
1400 .rec = osd_index_it_rec,
1401 .store = osd_index_it_store,
1402 .load = osd_index_it_load
1406 struct osd_metadnode_it {
1407 struct osd_device *mit_dev;
1409 struct lu_fid mit_fid;
1411 __u64 mit_prefetched_dnode;
1414 static struct dt_it *osd_zfs_otable_it_init(const struct lu_env *env,
1415 struct dt_object *dt, __u32 attr)
1417 struct osd_device *dev = osd_dev(dt->do_lu.lo_dev);
1418 struct osd_metadnode_it *it;
1422 if (unlikely(it == NULL))
1423 RETURN(ERR_PTR(-ENOMEM));
1427 /* XXX: dmu_object_next() does NOT find dnodes allocated
1428 * in the current non-committed txg, so we force txg
1429 * commit to find all existing dnodes ... */
1430 txg_wait_synced(dmu_objset_pool(dev->od_os), 0ULL);
1432 RETURN((struct dt_it *)it);
1435 static void osd_zfs_otable_it_fini(const struct lu_env *env, struct dt_it *di)
1437 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1442 static int osd_zfs_otable_it_get(const struct lu_env *env,
1443 struct dt_it *di, const struct dt_key *key)
1448 static void osd_zfs_otable_it_put(const struct lu_env *env, struct dt_it *di)
1452 #define OTABLE_PREFETCH 256
1454 static void osd_zfs_otable_prefetch(const struct lu_env *env,
1455 struct osd_metadnode_it *it)
1457 struct osd_device *dev = it->mit_dev;
1460 /* can go negative on the very first access to the iterator
1461 * or if some non-Lustre objects were found */
1462 if (unlikely(it->mit_prefetched < 0))
1463 it->mit_prefetched = 0;
1465 if (it->mit_prefetched >= (OTABLE_PREFETCH >> 1))
1468 if (it->mit_prefetched_dnode == 0)
1469 it->mit_prefetched_dnode = it->mit_pos;
1471 while (it->mit_prefetched < OTABLE_PREFETCH) {
1472 rc = -dmu_object_next(dev->od_os, &it->mit_prefetched_dnode,
1474 if (unlikely(rc != 0))
1477 /* dmu_prefetch() was exported in 0.6.2, if you use with
1478 * an older release, just comment it out - this is an
1480 dmu_prefetch(dev->od_os, it->mit_prefetched_dnode, 0, 0);
1482 it->mit_prefetched++;
1486 static int osd_zfs_otable_it_next(const struct lu_env *env, struct dt_it *di)
1488 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1489 struct lustre_mdt_attrs *lma;
1490 struct osd_device *dev = it->mit_dev;
1491 nvlist_t *nvbuf = NULL;
1496 memset(&it->mit_fid, 0, sizeof(it->mit_fid));
1498 dnode = it->mit_pos;
1500 rc = -dmu_object_next(dev->od_os, &it->mit_pos, B_FALSE, 0);
1501 if (unlikely(rc != 0))
1503 it->mit_prefetched--;
1505 /* LMA is required for this to be a Lustre object.
1506 * If there is no xattr skip it. */
1507 rc = __osd_xattr_load(dev, it->mit_pos, &nvbuf);
1508 if (unlikely(rc != 0))
1511 LASSERT(nvbuf != NULL);
1512 rc = -nvlist_lookup_byte_array(nvbuf, XATTR_NAME_LMA, &v, &s);
1513 if (likely(rc == 0)) {
1515 lma = (struct lustre_mdt_attrs *)v;
1516 lustre_lma_swab(lma);
1517 it->mit_fid = lma->lma_self_fid;
1521 /* not a Lustre object, try next one */
1528 /* we aren't prefetching in the above loop because the number of
1529 * non-Lustre objects is very small and we will be repeating very
1530 * rare. in case we want to use this to iterate over non-Lustre
1531 * objects (i.e. when we convert regular ZFS in Lustre) it makes
1532 * sense to initiate prefetching in the loop */
1534 /* 0 - there are more items, +1 - the end */
1535 if (likely(rc == 0))
1536 osd_zfs_otable_prefetch(env, it);
1538 CDEBUG(D_OTHER, "advance: %llu -> %llu "DFID": %d\n", dnode,
1539 it->mit_pos, PFID(&it->mit_fid), rc);
1545 static struct dt_key *osd_zfs_otable_it_key(const struct lu_env *env,
1546 const struct dt_it *di)
1551 static int osd_zfs_otable_it_key_size(const struct lu_env *env,
1552 const struct dt_it *di)
1554 return sizeof(__u64);
1557 static int osd_zfs_otable_it_rec(const struct lu_env *env,
1558 const struct dt_it *di,
1559 struct dt_rec *rec, __u32 attr)
1561 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1562 struct lu_fid *fid = (struct lu_fid *)rec;
1571 static __u64 osd_zfs_otable_it_store(const struct lu_env *env,
1572 const struct dt_it *di)
1574 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1579 static int osd_zfs_otable_it_load(const struct lu_env *env,
1580 const struct dt_it *di, __u64 hash)
1582 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1585 it->mit_prefetched = 0;
1586 it->mit_prefetched_dnode = 0;
1588 return osd_zfs_otable_it_next(env, (struct dt_it *)di);
1591 static int osd_zfs_otable_it_key_rec(const struct lu_env *env,
1592 const struct dt_it *di, void *key_rec)
1597 const struct dt_index_operations osd_zfs_otable_ops = {
1599 .init = osd_zfs_otable_it_init,
1600 .fini = osd_zfs_otable_it_fini,
1601 .get = osd_zfs_otable_it_get,
1602 .put = osd_zfs_otable_it_put,
1603 .next = osd_zfs_otable_it_next,
1604 .key = osd_zfs_otable_it_key,
1605 .key_size = osd_zfs_otable_it_key_size,
1606 .rec = osd_zfs_otable_it_rec,
1607 .store = osd_zfs_otable_it_store,
1608 .load = osd_zfs_otable_it_load,
1609 .key_rec = osd_zfs_otable_it_key_rec,
1613 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
1614 const struct dt_index_features *feat)
1616 struct osd_object *obj = osd_dt_obj(dt);
1620 down_read(&obj->oo_guard);
1623 * XXX: implement support for fixed-size keys sorted with natural
1624 * numerical way (not using internal hash value)
1626 if (feat->dif_flags & DT_IND_RANGE)
1627 GOTO(out, rc = -ERANGE);
1629 if (unlikely(feat == &dt_otable_features)) {
1630 dt->do_index_ops = &osd_zfs_otable_ops;
1634 LASSERT(!dt_object_exists(dt) || obj->oo_db != NULL);
1635 if (likely(feat == &dt_directory_features)) {
1636 if (!dt_object_exists(dt) || osd_object_is_zap(obj->oo_db))
1637 dt->do_index_ops = &osd_dir_ops;
1639 GOTO(out, rc = -ENOTDIR);
1640 } else if (unlikely(feat == &dt_acct_features)) {
1641 LASSERT(fid_is_acct(lu_object_fid(&dt->do_lu)));
1642 dt->do_index_ops = &osd_acct_index_ops;
1643 } else if (dt->do_index_ops == NULL) {
1644 /* For index file, we don't support variable key & record sizes
1645 * and the key has to be unique */
1646 if ((feat->dif_flags & ~DT_IND_UPDATE) != 0)
1647 GOTO(out, rc = -EINVAL);
1649 if (feat->dif_keysize_max > ZAP_MAXNAMELEN)
1650 GOTO(out, rc = -E2BIG);
1651 if (feat->dif_keysize_max != feat->dif_keysize_min)
1652 GOTO(out, rc = -EINVAL);
1654 /* As for the record size, it should be a multiple of 8 bytes
1655 * and smaller than the maximum value length supported by ZAP.
1657 if (feat->dif_recsize_max > ZAP_MAXVALUELEN)
1658 GOTO(out, rc = -E2BIG);
1659 if (feat->dif_recsize_max != feat->dif_recsize_min)
1660 GOTO(out, rc = -EINVAL);
1662 obj->oo_keysize = feat->dif_keysize_max;
1663 obj->oo_recsize = feat->dif_recsize_max;
1664 obj->oo_recusize = 1;
1666 /* ZFS prefers to work with array of 64bits */
1667 if ((obj->oo_recsize & 7) == 0) {
1668 obj->oo_recsize >>= 3;
1669 obj->oo_recusize = 8;
1671 dt->do_index_ops = &osd_index_ops;
1675 up_read(&obj->oo_guard);