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, 2014, 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 LASSERT(lu_object_exists(lo));
165 LASSERT(osd_object_is_zap(obj->oo_db));
168 OBD_SLAB_ALLOC_PTR_GFP(it, osd_zapit_cachep, GFP_NOFS);
170 RETURN(ERR_PTR(-ENOMEM));
172 rc = osd_obj_cursor_init(&it->ozi_zc, obj, 0);
174 OBD_SLAB_FREE_PTR(it, osd_zapit_cachep);
182 RETURN((struct dt_it *)it);
185 static void osd_index_it_fini(const struct lu_env *env, struct dt_it *di)
187 struct osd_zap_it *it = (struct osd_zap_it *)di;
188 struct osd_object *obj;
192 LASSERT(it->ozi_obj);
196 osd_zap_cursor_fini(it->ozi_zc);
197 lu_object_put(env, &obj->oo_dt.do_lu);
198 OBD_SLAB_FREE_PTR(it, osd_zapit_cachep);
204 static void osd_index_it_put(const struct lu_env *env, struct dt_it *di)
206 /* PBS: do nothing : ref are incremented at retrive and decreamented
210 static inline void osd_it_append_attrs(struct lu_dirent *ent, __u32 attr,
213 const unsigned align = sizeof(struct luda_type) - 1;
214 struct luda_type *lt;
216 /* check if file type is required */
217 if (attr & LUDA_TYPE) {
218 len = (len + align) & ~align;
220 lt = (void *)ent->lde_name + len;
221 lt->lt_type = cpu_to_le16(DTTOIF(type));
222 ent->lde_attrs |= LUDA_TYPE;
225 ent->lde_attrs = cpu_to_le32(ent->lde_attrs);
229 * as we don't know FID, we can't use LU object, so this function
230 * partially duplicate __osd_xattr_get() which is built around
231 * LU-object and uses it to cache data like regular EA dnode, etc
233 static int osd_find_parent_by_dnode(const struct lu_env *env,
237 struct osd_device *osd = osd_obj2dev(osd_dt_obj(o));
238 struct lustre_mdt_attrs *lma;
241 nvlist_t *nvbuf = NULL;
247 /* first of all, get parent dnode from own attributes */
248 LASSERT(osd_dt_obj(o)->oo_db);
249 rc = -sa_handle_get(osd->od_os, osd_dt_obj(o)->oo_db->db_object,
250 NULL, SA_HDL_PRIVATE, &sa_hdl);
254 dnode = ZFS_NO_OBJECT;
255 rc = -sa_lookup(sa_hdl, SA_ZPL_PARENT(osd), &dnode, 8);
256 sa_handle_destroy(sa_hdl);
260 /* now get EA buffer */
261 rc = __osd_xattr_load(osd, dnode, &nvbuf);
265 /* XXX: if we get that far.. should we cache the result? */
267 /* try to find LMA attribute */
268 LASSERT(nvbuf != NULL);
269 rc = -nvlist_lookup_byte_array(nvbuf, XATTR_NAME_LMA, &value, &size);
270 if (rc == 0 && size >= sizeof(*lma)) {
271 lma = (struct lustre_mdt_attrs *)value;
272 lustre_lma_swab(lma);
273 *fid = lma->lma_self_fid;
278 /* no LMA attribute in SA, let's try regular EA */
280 /* first of all, get parent dnode storing regular EA */
281 rc = -sa_handle_get(osd->od_os, dnode, NULL, SA_HDL_PRIVATE, &sa_hdl);
285 dnode = ZFS_NO_OBJECT;
286 rc = -sa_lookup(sa_hdl, SA_ZPL_XATTR(osd), &dnode, 8);
287 sa_handle_destroy(sa_hdl);
291 CLASSERT(sizeof(*lma) <= sizeof(osd_oti_get(env)->oti_buf));
292 buf.lb_buf = osd_oti_get(env)->oti_buf;
293 buf.lb_len = sizeof(osd_oti_get(env)->oti_buf);
295 /* now try to find LMA */
296 rc = __osd_xattr_get_large(env, osd, dnode, &buf,
297 XATTR_NAME_LMA, &size);
298 if (rc == 0 && size >= sizeof(*lma)) {
300 lustre_lma_swab(lma);
301 *fid = lma->lma_self_fid;
306 GOTO(out, rc = -EIO);
315 static int osd_find_parent_fid(const struct lu_env *env, struct dt_object *o,
318 struct link_ea_header *leh;
319 struct link_ea_entry *lee;
324 buf.lb_buf = osd_oti_get(env)->oti_buf;
325 buf.lb_len = sizeof(osd_oti_get(env)->oti_buf);
327 rc = osd_xattr_get(env, o, &buf, XATTR_NAME_LINK);
329 rc = osd_xattr_get(env, o, &LU_BUF_NULL, XATTR_NAME_LINK);
333 OBD_ALLOC(buf.lb_buf, rc);
334 if (buf.lb_buf == NULL)
337 rc = osd_xattr_get(env, o, &buf, XATTR_NAME_LINK);
341 if (rc < sizeof(*leh) + sizeof(*lee))
342 GOTO(out, rc = -EINVAL);
345 if (leh->leh_magic == __swab32(LINK_EA_MAGIC)) {
346 leh->leh_magic = LINK_EA_MAGIC;
347 leh->leh_reccount = __swab32(leh->leh_reccount);
348 leh->leh_len = __swab64(leh->leh_len);
350 if (leh->leh_magic != LINK_EA_MAGIC)
351 GOTO(out, rc = -EINVAL);
352 if (leh->leh_reccount == 0)
353 GOTO(out, rc = -ENODATA);
355 lee = (struct link_ea_entry *)(leh + 1);
356 fid_be_to_cpu(fid, (const struct lu_fid *)&lee->lee_parent_fid);
360 if (buf.lb_buf != osd_oti_get(env)->oti_buf)
361 OBD_FREE(buf.lb_buf, buf.lb_len);
364 /* this block can be enabled for additional verification
365 * it's trying to match FID from LinkEA vs. FID from LMA */
369 rc2 = osd_find_parent_by_dnode(env, o, &fid2);
371 if (lu_fid_eq(fid, &fid2) == 0)
372 CERROR("wrong parent: "DFID" != "DFID"\n",
373 PFID(fid), PFID(&fid2));
377 /* no LinkEA is found, let's try to find the fid in parent's LMA */
378 if (unlikely(rc != 0))
379 rc = osd_find_parent_by_dnode(env, o, fid);
384 static int osd_dir_lookup(const struct lu_env *env, struct dt_object *dt,
385 struct dt_rec *rec, const struct dt_key *key)
387 struct osd_thread_info *oti = osd_oti_get(env);
388 struct osd_object *obj = osd_dt_obj(dt);
389 struct osd_device *osd = osd_obj2dev(obj);
390 char *name = (char *)key;
394 LASSERT(osd_object_is_zap(obj->oo_db));
396 if (name[0] == '.') {
398 const struct lu_fid *f = lu_object_fid(&dt->do_lu);
399 memcpy(rec, f, sizeof(*f));
401 } else if (name[1] == '.' && name[2] == 0) {
402 rc = osd_find_parent_fid(env, dt, (struct lu_fid *)rec);
403 RETURN(rc == 0 ? 1 : rc);
407 rc = -zap_lookup(osd->od_os, obj->oo_db->db_object,
408 (char *)key, 8, sizeof(oti->oti_zde) / 8,
409 (void *)&oti->oti_zde);
410 memcpy(rec, &oti->oti_zde.lzd_fid, sizeof(struct lu_fid));
412 RETURN(rc == 0 ? 1 : rc);
415 static int osd_declare_dir_insert(const struct lu_env *env,
416 struct dt_object *dt,
417 const struct dt_rec *rec,
418 const struct dt_key *key,
421 struct osd_object *obj = osd_dt_obj(dt);
422 struct osd_thandle *oh;
427 oh = container_of0(th, struct osd_thandle, ot_super);
429 /* This is for inserting dot/dotdot for new created dir. */
430 if (obj->oo_db == NULL)
431 object = DMU_NEW_OBJECT;
433 object = obj->oo_db->db_object;
435 dmu_tx_hold_bonus(oh->ot_tx, object);
436 dmu_tx_hold_zap(oh->ot_tx, object, TRUE, (char *)key);
442 * Find the osd object for given fid.
444 * \param fid need to find the osd object having this fid
446 * \retval osd_object on success
447 * \retval -ve on error
449 struct osd_object *osd_object_find(const struct lu_env *env,
450 struct dt_object *dt,
451 const struct lu_fid *fid)
453 struct lu_device *ludev = dt->do_lu.lo_dev;
454 struct osd_object *child = NULL;
455 struct lu_object *luch;
456 struct lu_object *lo;
459 * at this point topdev might not exist yet
460 * (i.e. MGS is preparing profiles). so we can
461 * not rely on topdev and instead lookup with
462 * our device passed as topdev. this can't work
463 * if the object isn't cached yet (as osd doesn't
464 * allocate lu_header). IOW, the object must be
465 * in the cache, otherwise lu_object_alloc() crashes
468 luch = lu_object_find_at(env, ludev, fid, NULL);
472 if (lu_object_exists(luch)) {
473 lo = lu_object_locate(luch->lo_header, ludev->ld_type);
477 LU_OBJECT_DEBUG(D_ERROR, env, luch,
478 "%s: object can't be located "DFID"\n",
479 osd_dev(ludev)->od_svname, PFID(fid));
482 lu_object_put(env, luch);
483 CERROR("%s: Unable to get osd_object "DFID"\n",
484 osd_dev(ludev)->od_svname, PFID(fid));
485 child = ERR_PTR(-ENOENT);
488 LU_OBJECT_DEBUG(D_ERROR, env, luch,
489 "%s: lu_object does not exists "DFID"\n",
490 osd_dev(ludev)->od_svname, PFID(fid));
491 lu_object_put(env, luch);
492 child = ERR_PTR(-ENOENT);
499 * Put the osd object once done with it.
501 * \param obj osd object that needs to be put
503 static inline void osd_object_put(const struct lu_env *env,
504 struct osd_object *obj)
506 lu_object_put(env, &obj->oo_dt.do_lu);
509 static int osd_seq_exists(const struct lu_env *env, struct osd_device *osd,
512 struct lu_seq_range *range = &osd_oti_get(env)->oti_seq_range;
513 struct seq_server_site *ss = osd_seq_site(osd);
518 LASSERT(ss->ss_server_fld != NULL);
520 rc = osd_fld_lookup(env, osd, seq, range);
523 CERROR("%s: Can not lookup fld for "LPX64"\n",
528 RETURN(ss->ss_node_id == range->lsr_index);
531 static int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
532 const struct lu_fid *fid)
534 struct seq_server_site *ss = osd_seq_site(osd);
537 /* FID seqs not in FLDB, must be local seq */
538 if (unlikely(!fid_seq_in_fldb(fid_seq(fid))))
541 /* If FLD is not being initialized yet, it only happens during the
542 * initialization, likely during mgs initialization, and we assume
543 * this is local FID. */
544 if (ss == NULL || ss->ss_server_fld == NULL)
547 /* Only check the local FLDB here */
548 if (osd_seq_exists(env, osd, fid_seq(fid)))
555 * Inserts (key, value) pair in \a directory object.
557 * \param dt osd index object
558 * \param key key for index
559 * \param rec record reference
560 * \param th transaction handler
561 * \param ignore_quota update should not affect quota
564 * \retval -ve failure
566 static int osd_dir_insert(const struct lu_env *env, struct dt_object *dt,
567 const struct dt_rec *rec, const struct dt_key *key,
568 struct thandle *th, int ignore_quota)
570 struct osd_thread_info *oti = osd_oti_get(env);
571 struct osd_object *parent = osd_dt_obj(dt);
572 struct osd_device *osd = osd_obj2dev(parent);
573 struct dt_insert_rec *rec1 = (struct dt_insert_rec *)rec;
574 const struct lu_fid *fid = rec1->rec_fid;
575 struct osd_thandle *oh;
576 struct osd_object *child = NULL;
578 char *name = (char *)key;
582 LASSERT(parent->oo_db);
583 LASSERT(osd_object_is_zap(parent->oo_db));
585 LASSERT(dt_object_exists(dt));
586 LASSERT(osd_invariant(parent));
589 oh = container_of0(th, struct osd_thandle, ot_super);
591 rc = osd_remote_fid(env, osd, fid);
593 CERROR("%s: Can not find object "DFID": rc = %d\n",
594 osd->od_svname, PFID(fid), rc);
598 if (unlikely(rc == 1)) {
599 /* Insert remote entry */
600 memset(&oti->oti_zde.lzd_reg, 0, sizeof(oti->oti_zde.lzd_reg));
601 oti->oti_zde.lzd_reg.zde_type = IFTODT(rec1->rec_type & S_IFMT);
604 * To simulate old Orion setups with ./.. stored in the
607 /* Insert local entry */
608 child = osd_object_find(env, dt, fid);
610 RETURN(PTR_ERR(child));
612 LASSERT(child->oo_db);
613 if (name[0] == '.') {
615 /* do not store ".", instead generate it
616 * during iteration */
618 } else if (name[1] == '.' && name[2] == 0) {
619 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_BAD_PARENT)) {
620 struct lu_fid tfid = *fid;
622 osd_object_put(env, child);
624 child = osd_object_find(env, dt, &tfid);
626 RETURN(PTR_ERR(child));
628 LASSERT(child->oo_db);
631 /* update parent dnode in the child.
632 * later it will be used to generate ".." */
633 rc = osd_object_sa_update(parent,
635 &child->oo_db->db_object,
641 CLASSERT(sizeof(oti->oti_zde.lzd_reg) == 8);
642 CLASSERT(sizeof(oti->oti_zde) % 8 == 0);
643 attr = child->oo_dt.do_lu.lo_header ->loh_attr;
644 oti->oti_zde.lzd_reg.zde_type = IFTODT(attr & S_IFMT);
645 oti->oti_zde.lzd_reg.zde_dnode = child->oo_db->db_object;
648 oti->oti_zde.lzd_fid = *fid;
649 /* Insert (key,oid) into ZAP */
650 rc = -zap_add(osd->od_os, parent->oo_db->db_object,
651 (char *)key, 8, sizeof(oti->oti_zde) / 8,
652 (void *)&oti->oti_zde, oh->ot_tx);
653 if (unlikely(rc == -EEXIST &&
654 name[0] == '.' && name[1] == '.' && name[2] == 0))
655 /* Update (key,oid) in ZAP */
656 rc = -zap_update(osd->od_os, parent->oo_db->db_object,
657 (char *)key, 8, sizeof(oti->oti_zde) / 8,
658 (void *)&oti->oti_zde, oh->ot_tx);
662 osd_object_put(env, child);
667 static int osd_declare_dir_delete(const struct lu_env *env,
668 struct dt_object *dt,
669 const struct dt_key *key,
672 struct osd_object *obj = osd_dt_obj(dt);
673 struct osd_thandle *oh;
677 LASSERT(dt_object_exists(dt));
678 LASSERT(osd_invariant(obj));
681 oh = container_of0(th, struct osd_thandle, ot_super);
683 if (dt_object_exists(dt)) {
685 LASSERT(osd_object_is_zap(obj->oo_db));
686 dnode = obj->oo_db->db_object;
688 dnode = DMU_NEW_OBJECT;
690 dmu_tx_hold_zap(oh->ot_tx, dnode, TRUE, (char *)key);
695 static int osd_dir_delete(const struct lu_env *env, struct dt_object *dt,
696 const struct dt_key *key, struct thandle *th)
698 struct osd_object *obj = osd_dt_obj(dt);
699 struct osd_device *osd = osd_obj2dev(obj);
700 struct osd_thandle *oh;
701 dmu_buf_t *zap_db = obj->oo_db;
702 char *name = (char *)key;
707 LASSERT(osd_object_is_zap(zap_db));
710 oh = container_of0(th, struct osd_thandle, ot_super);
713 * In Orion . and .. were stored in the directory (not generated upon
714 * request as now). we preserve them for backward compatibility
716 if (name[0] == '.') {
719 } else if (name[1] == '.' && name[2] == 0) {
724 /* Remove key from the ZAP */
725 rc = -zap_remove(osd->od_os, zap_db->db_object,
726 (char *) key, oh->ot_tx);
728 if (unlikely(rc && rc != -ENOENT))
729 CERROR("%s: zap_remove failed: rc = %d\n", osd->od_svname, rc);
734 static struct dt_it *osd_dir_it_init(const struct lu_env *env,
735 struct dt_object *dt,
738 struct osd_zap_it *it;
740 it = (struct osd_zap_it *)osd_index_it_init(env, dt, unused);
744 RETURN((struct dt_it *)it);
748 * Move Iterator to record specified by \a key
750 * \param di osd iterator
751 * \param key key for index
753 * \retval +ve di points to record with least key not larger than key
754 * \retval 0 di points to exact matched key
755 * \retval -ve failure
757 static int osd_dir_it_get(const struct lu_env *env,
758 struct dt_it *di, const struct dt_key *key)
760 struct osd_zap_it *it = (struct osd_zap_it *)di;
761 struct osd_object *obj = it->ozi_obj;
762 char *name = (char *)key;
769 /* reset the cursor */
770 zap_cursor_fini(it->ozi_zc);
771 osd_obj_cursor_init_serialized(it->ozi_zc, obj, 0);
773 /* XXX: implementation of the API is broken at the moment */
774 LASSERT(((const char *)key)[0] == 0);
781 if (name[0] == '.') {
785 } else if (name[1] == '.' && name[2] == 0) {
791 /* neither . nor .. - some real record */
799 static void osd_dir_it_put(const struct lu_env *env, struct dt_it *di)
801 /* PBS: do nothing : ref are incremented at retrive and decreamented
806 * in Orion . and .. were stored in the directory, while ZPL
807 * and current osd-zfs generate them up on request. so, we
808 * need to ignore previously stored . and ..
810 static int osd_index_retrieve_skip_dots(struct osd_zap_it *it,
816 rc = -zap_cursor_retrieve(it->ozi_zc, za);
819 if (unlikely(rc == 0 && za->za_name[0] == '.')) {
820 if (za->za_name[1] == 0) {
822 } else if (za->za_name[1] == '.' &&
823 za->za_name[2] == 0) {
827 zap_cursor_advance(it->ozi_zc);
829 } while (unlikely(rc == 0 && isdot));
835 * to load a directory entry at a time and stored it in
836 * iterator's in-memory data structure.
838 * \param di, struct osd_it_ea, iterator's in memory structure
840 * \retval +ve, iterator reached to end
841 * \retval 0, iterator not reached to end
842 * \retval -ve, on error
844 static int osd_dir_it_next(const struct lu_env *env, struct dt_it *di)
846 struct osd_zap_it *it = (struct osd_zap_it *)di;
847 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
852 /* temp. storage should be enough for any key supported by ZFS */
853 CLASSERT(sizeof(za->za_name) <= sizeof(it->ozi_name));
856 * the first ->next() moves the cursor to .
857 * the second ->next() moves the cursor to ..
858 * then we get to the real records and have to verify any exist
860 if (it->ozi_pos <= 2) {
866 zap_cursor_advance(it->ozi_zc);
870 * According to current API we need to return error if its last entry.
871 * zap_cursor_advance() does not return any value. So we need to call
872 * retrieve to check if there is any record. We should make
873 * changes to Iterator API to not return status for this API
875 rc = osd_index_retrieve_skip_dots(it, za);
877 if (rc == -ENOENT) /* end of dir */
883 static struct dt_key *osd_dir_it_key(const struct lu_env *env,
884 const struct dt_it *di)
886 struct osd_zap_it *it = (struct osd_zap_it *)di;
887 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
891 if (it->ozi_pos <= 1) {
893 RETURN((struct dt_key *)".");
894 } else if (it->ozi_pos == 2) {
895 RETURN((struct dt_key *)"..");
898 if ((rc = -zap_cursor_retrieve(it->ozi_zc, za)))
901 strcpy(it->ozi_name, za->za_name);
903 RETURN((struct dt_key *)it->ozi_name);
906 static int osd_dir_it_key_size(const struct lu_env *env, const struct dt_it *di)
908 struct osd_zap_it *it = (struct osd_zap_it *)di;
909 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
913 if (it->ozi_pos <= 1) {
916 } else if (it->ozi_pos == 2) {
920 if ((rc = -zap_cursor_retrieve(it->ozi_zc, za)) == 0)
921 rc = strlen(za->za_name);
926 static int osd_dir_it_rec(const struct lu_env *env, const struct dt_it *di,
927 struct dt_rec *dtrec, __u32 attr)
929 struct osd_zap_it *it = (struct osd_zap_it *)di;
930 struct lu_dirent *lde = (struct lu_dirent *)dtrec;
931 struct luz_direntry *zde = &osd_oti_get(env)->oti_zde;
932 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
936 if (it->ozi_pos <= 1) {
937 lde->lde_hash = cpu_to_le64(1);
938 strcpy(lde->lde_name, ".");
939 lde->lde_namelen = cpu_to_le16(1);
940 lde->lde_fid = *lu_object_fid(&it->ozi_obj->oo_dt.do_lu);
941 lde->lde_attrs = LUDA_FID;
942 /* append lustre attributes */
943 osd_it_append_attrs(lde, attr, 1, IFTODT(S_IFDIR));
944 lde->lde_reclen = cpu_to_le16(lu_dirent_calc_size(1, attr));
948 } else if (it->ozi_pos == 2) {
949 lde->lde_hash = cpu_to_le64(2);
950 strcpy(lde->lde_name, "..");
951 lde->lde_namelen = cpu_to_le16(2);
952 lde->lde_attrs = LUDA_FID;
953 /* append lustre attributes */
954 osd_it_append_attrs(lde, attr, 2, IFTODT(S_IFDIR));
955 lde->lde_reclen = cpu_to_le16(lu_dirent_calc_size(2, attr));
956 rc = osd_find_parent_fid(env, &it->ozi_obj->oo_dt, &lde->lde_fid);
958 /* ENOENT happens at the root of filesystem so ignore it */
966 rc = -zap_cursor_retrieve(it->ozi_zc, za);
967 if (unlikely(rc != 0))
970 lde->lde_hash = cpu_to_le64(osd_zap_cursor_serialize(it->ozi_zc));
971 namelen = strlen(za->za_name);
972 if (namelen > NAME_MAX)
973 GOTO(out, rc = -EOVERFLOW);
974 strcpy(lde->lde_name, za->za_name);
975 lde->lde_namelen = cpu_to_le16(namelen);
977 if (za->za_integer_length != 8 || za->za_num_integers < 3) {
978 CERROR("%s: unsupported direntry format: %d %d\n",
979 osd_obj2dev(it->ozi_obj)->od_svname,
980 za->za_integer_length, (int)za->za_num_integers);
982 GOTO(out, rc = -EIO);
985 rc = -zap_lookup(it->ozi_zc->zc_objset, it->ozi_zc->zc_zapobj,
986 za->za_name, za->za_integer_length, 3, zde);
990 lde->lde_fid = zde->lzd_fid;
991 lde->lde_attrs = LUDA_FID;
993 /* append lustre attributes */
994 osd_it_append_attrs(lde, attr, namelen, zde->lzd_reg.zde_type);
996 lde->lde_reclen = cpu_to_le16(lu_dirent_calc_size(namelen, attr));
1002 static int osd_dir_it_rec_size(const struct lu_env *env, const struct dt_it *di,
1005 struct osd_zap_it *it = (struct osd_zap_it *)di;
1006 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1011 if (it->ozi_pos <= 1)
1013 else if (it->ozi_pos == 2)
1017 rc = lu_dirent_calc_size(namelen, attr);
1021 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1022 if (unlikely(rc != 0))
1025 if (za->za_integer_length != 8 || za->za_num_integers < 3) {
1026 CERROR("%s: unsupported direntry format: %d %d\n",
1027 osd_obj2dev(it->ozi_obj)->od_svname,
1028 za->za_integer_length, (int)za->za_num_integers);
1032 namelen = strlen(za->za_name);
1033 if (namelen > NAME_MAX)
1036 rc = lu_dirent_calc_size(namelen, attr);
1041 static __u64 osd_dir_it_store(const struct lu_env *env, const struct dt_it *di)
1043 struct osd_zap_it *it = (struct osd_zap_it *)di;
1047 if (it->ozi_pos <= 2)
1050 pos = osd_zap_cursor_serialize(it->ozi_zc);
1057 * rc == 0 -> end of directory.
1058 * rc > 0 -> ok, proceed.
1059 * rc < 0 -> error. ( EOVERFLOW can be masked.)
1061 static int osd_dir_it_load(const struct lu_env *env,
1062 const struct dt_it *di, __u64 hash)
1064 struct osd_zap_it *it = (struct osd_zap_it *)di;
1065 struct osd_object *obj = it->ozi_obj;
1066 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1070 /* reset the cursor */
1071 zap_cursor_fini(it->ozi_zc);
1072 osd_obj_cursor_init_serialized(it->ozi_zc, obj, hash);
1079 /* to return whether the end has been reached */
1080 rc = osd_index_retrieve_skip_dots(it, za);
1083 else if (rc == -ENOENT)
1090 struct dt_index_operations osd_dir_ops = {
1091 .dio_lookup = osd_dir_lookup,
1092 .dio_declare_insert = osd_declare_dir_insert,
1093 .dio_insert = osd_dir_insert,
1094 .dio_declare_delete = osd_declare_dir_delete,
1095 .dio_delete = osd_dir_delete,
1097 .init = osd_dir_it_init,
1098 .fini = osd_index_it_fini,
1099 .get = osd_dir_it_get,
1100 .put = osd_dir_it_put,
1101 .next = osd_dir_it_next,
1102 .key = osd_dir_it_key,
1103 .key_size = osd_dir_it_key_size,
1104 .rec = osd_dir_it_rec,
1105 .rec_size = osd_dir_it_rec_size,
1106 .store = osd_dir_it_store,
1107 .load = osd_dir_it_load
1112 * Primitives for index files using binary keys.
1115 /* key integer_size is 8 */
1116 static int osd_prepare_key_uint64(struct osd_object *o, __u64 *dst,
1117 const struct dt_key *src)
1124 /* align keysize to 64bit */
1125 size = (o->oo_keysize + sizeof(__u64) - 1) / sizeof(__u64);
1126 size *= sizeof(__u64);
1128 LASSERT(size <= MAXNAMELEN);
1130 if (unlikely(size > o->oo_keysize))
1131 memset(dst + o->oo_keysize, 0, size - o->oo_keysize);
1132 memcpy(dst, (const char *)src, o->oo_keysize);
1134 return (size/sizeof(__u64));
1137 static int osd_index_lookup(const struct lu_env *env, struct dt_object *dt,
1138 struct dt_rec *rec, const struct dt_key *key)
1140 struct osd_object *obj = osd_dt_obj(dt);
1141 struct osd_device *osd = osd_obj2dev(obj);
1142 __u64 *k = osd_oti_get(env)->oti_key64;
1146 rc = osd_prepare_key_uint64(obj, k, key);
1148 rc = -zap_lookup_uint64(osd->od_os, obj->oo_db->db_object,
1149 k, rc, obj->oo_recusize, obj->oo_recsize,
1151 RETURN(rc == 0 ? 1 : rc);
1154 static int osd_declare_index_insert(const struct lu_env *env,
1155 struct dt_object *dt,
1156 const struct dt_rec *rec,
1157 const struct dt_key *key,
1160 struct osd_object *obj = osd_dt_obj(dt);
1161 struct osd_thandle *oh;
1164 LASSERT(th != NULL);
1165 oh = container_of0(th, struct osd_thandle, ot_super);
1167 LASSERT(obj->oo_db);
1169 dmu_tx_hold_bonus(oh->ot_tx, obj->oo_db->db_object);
1171 /* It is not clear what API should be used for binary keys, so we pass
1172 * a null name which has the side effect of over-reserving space,
1173 * accounting for the worst case. See zap_count_write() */
1174 dmu_tx_hold_zap(oh->ot_tx, obj->oo_db->db_object, TRUE, NULL);
1179 static int osd_index_insert(const struct lu_env *env, struct dt_object *dt,
1180 const struct dt_rec *rec, const struct dt_key *key,
1181 struct thandle *th, int ignore_quota)
1183 struct osd_object *obj = osd_dt_obj(dt);
1184 struct osd_device *osd = osd_obj2dev(obj);
1185 struct osd_thandle *oh;
1186 __u64 *k = osd_oti_get(env)->oti_key64;
1190 LASSERT(obj->oo_db);
1191 LASSERT(dt_object_exists(dt));
1192 LASSERT(osd_invariant(obj));
1193 LASSERT(th != NULL);
1195 oh = container_of0(th, struct osd_thandle, ot_super);
1197 rc = osd_prepare_key_uint64(obj, k, key);
1199 /* Insert (key,oid) into ZAP */
1200 rc = -zap_add_uint64(osd->od_os, obj->oo_db->db_object,
1201 k, rc, obj->oo_recusize, obj->oo_recsize,
1202 (void *)rec, oh->ot_tx);
1206 static int osd_declare_index_delete(const struct lu_env *env,
1207 struct dt_object *dt,
1208 const struct dt_key *key,
1211 struct osd_object *obj = osd_dt_obj(dt);
1212 struct osd_thandle *oh;
1215 LASSERT(dt_object_exists(dt));
1216 LASSERT(osd_invariant(obj));
1217 LASSERT(th != NULL);
1218 LASSERT(obj->oo_db);
1220 oh = container_of0(th, struct osd_thandle, ot_super);
1221 dmu_tx_hold_zap(oh->ot_tx, obj->oo_db->db_object, TRUE, NULL);
1226 static int osd_index_delete(const struct lu_env *env, struct dt_object *dt,
1227 const struct dt_key *key, struct thandle *th)
1229 struct osd_object *obj = osd_dt_obj(dt);
1230 struct osd_device *osd = osd_obj2dev(obj);
1231 struct osd_thandle *oh;
1232 __u64 *k = osd_oti_get(env)->oti_key64;
1236 LASSERT(obj->oo_db);
1237 LASSERT(th != NULL);
1238 oh = container_of0(th, struct osd_thandle, ot_super);
1240 rc = osd_prepare_key_uint64(obj, k, key);
1242 /* Remove binary key from the ZAP */
1243 rc = -zap_remove_uint64(osd->od_os, obj->oo_db->db_object,
1248 static int osd_index_it_get(const struct lu_env *env, struct dt_it *di,
1249 const struct dt_key *key)
1251 struct osd_zap_it *it = (struct osd_zap_it *)di;
1252 struct osd_object *obj = it->ozi_obj;
1253 struct osd_device *osd = osd_obj2dev(obj);
1257 LASSERT(it->ozi_zc);
1260 * XXX: we need a binary version of zap_cursor_move_to_key()
1261 * to implement this API */
1262 if (*((const __u64 *)key) != 0)
1263 CERROR("NOT IMPLEMETED YET (move to "LPX64")\n",
1266 zap_cursor_fini(it->ozi_zc);
1267 zap_cursor_init(it->ozi_zc, osd->od_os, obj->oo_db->db_object);
1273 static int osd_index_it_next(const struct lu_env *env, struct dt_it *di)
1275 struct osd_zap_it *it = (struct osd_zap_it *)di;
1276 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1280 if (it->ozi_reset == 0)
1281 zap_cursor_advance(it->ozi_zc);
1285 * According to current API we need to return error if it's last entry.
1286 * zap_cursor_advance() does not return any value. So we need to call
1287 * retrieve to check if there is any record. We should make
1288 * changes to Iterator API to not return status for this API
1290 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1297 static struct dt_key *osd_index_it_key(const struct lu_env *env,
1298 const struct dt_it *di)
1300 struct osd_zap_it *it = (struct osd_zap_it *)di;
1301 struct osd_object *obj = it->ozi_obj;
1302 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1307 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1309 RETURN(ERR_PTR(rc));
1311 /* the binary key is stored in the name */
1312 memcpy(&it->ozi_key, za->za_name, obj->oo_keysize);
1314 RETURN((struct dt_key *)&it->ozi_key);
1317 static int osd_index_it_key_size(const struct lu_env *env,
1318 const struct dt_it *di)
1320 struct osd_zap_it *it = (struct osd_zap_it *)di;
1321 struct osd_object *obj = it->ozi_obj;
1322 RETURN(obj->oo_keysize);
1325 static int osd_index_it_rec(const struct lu_env *env, const struct dt_it *di,
1326 struct dt_rec *rec, __u32 attr)
1328 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1329 struct osd_zap_it *it = (struct osd_zap_it *)di;
1330 struct osd_object *obj = it->ozi_obj;
1331 struct osd_device *osd = osd_obj2dev(obj);
1332 __u64 *k = osd_oti_get(env)->oti_key64;
1337 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1341 rc = osd_prepare_key_uint64(obj, k, (const struct dt_key *)za->za_name);
1343 rc = -zap_lookup_uint64(osd->od_os, obj->oo_db->db_object,
1344 k, rc, obj->oo_recusize, obj->oo_recsize,
1349 static __u64 osd_index_it_store(const struct lu_env *env,
1350 const struct dt_it *di)
1352 struct osd_zap_it *it = (struct osd_zap_it *)di;
1355 RETURN((__u64)zap_cursor_serialize(it->ozi_zc));
1358 static int osd_index_it_load(const struct lu_env *env, const struct dt_it *di,
1361 struct osd_zap_it *it = (struct osd_zap_it *)di;
1362 struct osd_object *obj = it->ozi_obj;
1363 struct osd_device *osd = osd_obj2dev(obj);
1364 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
1368 /* reset the cursor */
1369 zap_cursor_fini(it->ozi_zc);
1370 zap_cursor_init_serialized(it->ozi_zc, osd->od_os,
1371 obj->oo_db->db_object, hash);
1374 rc = -zap_cursor_retrieve(it->ozi_zc, za);
1377 else if (rc == -ENOENT)
1383 static struct dt_index_operations osd_index_ops = {
1384 .dio_lookup = osd_index_lookup,
1385 .dio_declare_insert = osd_declare_index_insert,
1386 .dio_insert = osd_index_insert,
1387 .dio_declare_delete = osd_declare_index_delete,
1388 .dio_delete = osd_index_delete,
1390 .init = osd_index_it_init,
1391 .fini = osd_index_it_fini,
1392 .get = osd_index_it_get,
1393 .put = osd_index_it_put,
1394 .next = osd_index_it_next,
1395 .key = osd_index_it_key,
1396 .key_size = osd_index_it_key_size,
1397 .rec = osd_index_it_rec,
1398 .store = osd_index_it_store,
1399 .load = osd_index_it_load
1403 struct osd_metadnode_it {
1404 struct osd_device *mit_dev;
1406 struct lu_fid mit_fid;
1408 __u64 mit_prefetched_dnode;
1411 static struct dt_it *osd_zfs_otable_it_init(const struct lu_env *env,
1412 struct dt_object *dt, __u32 attr)
1414 struct osd_device *dev = osd_dev(dt->do_lu.lo_dev);
1415 struct osd_metadnode_it *it;
1419 if (unlikely(it == NULL))
1420 RETURN(ERR_PTR(-ENOMEM));
1424 /* XXX: dmu_object_next() does NOT find dnodes allocated
1425 * in the current non-committed txg, so we force txg
1426 * commit to find all existing dnodes ... */
1427 txg_wait_synced(dmu_objset_pool(dev->od_os), 0ULL);
1429 RETURN((struct dt_it *)it);
1432 static void osd_zfs_otable_it_fini(const struct lu_env *env, struct dt_it *di)
1434 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1439 static int osd_zfs_otable_it_get(const struct lu_env *env,
1440 struct dt_it *di, const struct dt_key *key)
1445 static void osd_zfs_otable_it_put(const struct lu_env *env, struct dt_it *di)
1449 #define OTABLE_PREFETCH 256
1451 static void osd_zfs_otable_prefetch(const struct lu_env *env,
1452 struct osd_metadnode_it *it)
1454 struct osd_device *dev = it->mit_dev;
1457 /* can go negative on the very first access to the iterator
1458 * or if some non-Lustre objects were found */
1459 if (unlikely(it->mit_prefetched < 0))
1460 it->mit_prefetched = 0;
1462 if (it->mit_prefetched >= (OTABLE_PREFETCH >> 1))
1465 if (it->mit_prefetched_dnode == 0)
1466 it->mit_prefetched_dnode = it->mit_pos;
1468 while (it->mit_prefetched < OTABLE_PREFETCH) {
1469 rc = -dmu_object_next(dev->od_os, &it->mit_prefetched_dnode,
1471 if (unlikely(rc != 0))
1474 /* dmu_prefetch() was exported in 0.6.2, if you use with
1475 * an older release, just comment it out - this is an
1477 dmu_prefetch(dev->od_os, it->mit_prefetched_dnode, 0, 0);
1479 it->mit_prefetched++;
1483 static int osd_zfs_otable_it_next(const struct lu_env *env, struct dt_it *di)
1485 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1486 struct lustre_mdt_attrs *lma;
1487 struct osd_device *dev = it->mit_dev;
1488 nvlist_t *nvbuf = NULL;
1493 memset(&it->mit_fid, 0, sizeof(it->mit_fid));
1495 dnode = it->mit_pos;
1497 rc = -dmu_object_next(dev->od_os, &it->mit_pos, B_FALSE, 0);
1498 if (unlikely(rc != 0))
1500 it->mit_prefetched--;
1502 /* LMA is required for this to be a Lustre object.
1503 * If there is no xattr skip it. */
1504 rc = __osd_xattr_load(dev, it->mit_pos, &nvbuf);
1505 if (unlikely(rc != 0))
1508 LASSERT(nvbuf != NULL);
1509 rc = -nvlist_lookup_byte_array(nvbuf, XATTR_NAME_LMA, &v, &s);
1510 if (likely(rc == 0)) {
1512 lma = (struct lustre_mdt_attrs *)v;
1513 lustre_lma_swab(lma);
1514 it->mit_fid = lma->lma_self_fid;
1518 /* not a Lustre object, try next one */
1525 /* we aren't prefetching in the above loop because the number of
1526 * non-Lustre objects is very small and we will be repeating very
1527 * rare. in case we want to use this to iterate over non-Lustre
1528 * objects (i.e. when we convert regular ZFS in Lustre) it makes
1529 * sense to initiate prefetching in the loop */
1531 /* 0 - there are more items, +1 - the end */
1532 if (likely(rc == 0))
1533 osd_zfs_otable_prefetch(env, it);
1535 CDEBUG(D_OTHER, "advance: %llu -> %llu "DFID": %d\n", dnode,
1536 it->mit_pos, PFID(&it->mit_fid), rc);
1542 static struct dt_key *osd_zfs_otable_it_key(const struct lu_env *env,
1543 const struct dt_it *di)
1548 static int osd_zfs_otable_it_key_size(const struct lu_env *env,
1549 const struct dt_it *di)
1551 return sizeof(__u64);
1554 static int osd_zfs_otable_it_rec(const struct lu_env *env,
1555 const struct dt_it *di,
1556 struct dt_rec *rec, __u32 attr)
1558 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1559 struct lu_fid *fid = (struct lu_fid *)rec;
1568 static __u64 osd_zfs_otable_it_store(const struct lu_env *env,
1569 const struct dt_it *di)
1571 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1576 static int osd_zfs_otable_it_load(const struct lu_env *env,
1577 const struct dt_it *di, __u64 hash)
1579 struct osd_metadnode_it *it = (struct osd_metadnode_it *)di;
1582 it->mit_prefetched = 0;
1583 it->mit_prefetched_dnode = 0;
1585 return osd_zfs_otable_it_next(env, (struct dt_it *)di);
1588 static int osd_zfs_otable_it_key_rec(const struct lu_env *env,
1589 const struct dt_it *di, void *key_rec)
1594 const struct dt_index_operations osd_zfs_otable_ops = {
1596 .init = osd_zfs_otable_it_init,
1597 .fini = osd_zfs_otable_it_fini,
1598 .get = osd_zfs_otable_it_get,
1599 .put = osd_zfs_otable_it_put,
1600 .next = osd_zfs_otable_it_next,
1601 .key = osd_zfs_otable_it_key,
1602 .key_size = osd_zfs_otable_it_key_size,
1603 .rec = osd_zfs_otable_it_rec,
1604 .store = osd_zfs_otable_it_store,
1605 .load = osd_zfs_otable_it_load,
1606 .key_rec = osd_zfs_otable_it_key_rec,
1610 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
1611 const struct dt_index_features *feat)
1613 struct osd_object *obj = osd_dt_obj(dt);
1617 * XXX: implement support for fixed-size keys sorted with natural
1618 * numerical way (not using internal hash value)
1620 if (feat->dif_flags & DT_IND_RANGE)
1623 if (unlikely(feat == &dt_otable_features)) {
1624 dt->do_index_ops = &osd_zfs_otable_ops;
1628 LASSERT(!dt_object_exists(dt) || obj->oo_db != NULL);
1629 if (likely(feat == &dt_directory_features)) {
1630 if (!dt_object_exists(dt) || osd_object_is_zap(obj->oo_db))
1631 dt->do_index_ops = &osd_dir_ops;
1634 } else if (unlikely(feat == &dt_acct_features)) {
1635 LASSERT(fid_is_acct(lu_object_fid(&dt->do_lu)));
1636 dt->do_index_ops = &osd_acct_index_ops;
1637 } else if (dt->do_index_ops == NULL) {
1638 /* For index file, we don't support variable key & record sizes
1639 * and the key has to be unique */
1640 if ((feat->dif_flags & ~DT_IND_UPDATE) != 0)
1643 if (feat->dif_keysize_max > ZAP_MAXNAMELEN)
1645 if (feat->dif_keysize_max != feat->dif_keysize_min)
1648 /* As for the record size, it should be a multiple of 8 bytes
1649 * and smaller than the maximum value length supported by ZAP.
1651 if (feat->dif_recsize_max > ZAP_MAXVALUELEN)
1653 if (feat->dif_recsize_max != feat->dif_recsize_min)
1656 obj->oo_keysize = feat->dif_keysize_max;
1657 obj->oo_recsize = feat->dif_recsize_max;
1658 obj->oo_recusize = 1;
1660 /* ZFS prefers to work with array of 64bits */
1661 if ((obj->oo_recsize & 7) == 0) {
1662 obj->oo_recsize >>= 3;
1663 obj->oo_recusize = 8;
1665 dt->do_index_ops = &osd_index_ops;