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, 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/obdclass/dt_object.c
39 * Generic functions from dt_object.h
41 * Author: Nikita Danilov <nikita@clusterfs.com>
44 #define DEBUG_SUBSYSTEM S_CLASS
46 #include <linux/list.h>
48 #include <dt_object.h>
50 #include <lustre_fid.h>
51 #include <lustre_nodemap.h>
52 #include <lustre_quota.h>
54 /* context key constructor/destructor: dt_global_key_init, dt_global_key_fini */
55 LU_KEY_INIT(dt_global, struct dt_thread_info);
56 LU_KEY_FINI(dt_global, struct dt_thread_info);
58 struct lu_context_key dt_key = {
59 .lct_tags = LCT_MD_THREAD | LCT_DT_THREAD | LCT_MG_THREAD | LCT_LOCAL,
60 .lct_init = dt_global_key_init,
61 .lct_fini = dt_global_key_fini
64 /* no lock is necessary to protect the list, because call-backs
65 * are added during system startup. Please refer to "struct dt_device".
67 void dt_txn_callback_add(struct dt_device *dev, struct dt_txn_callback *cb)
69 list_add(&cb->dtc_linkage, &dev->dd_txn_callbacks);
71 EXPORT_SYMBOL(dt_txn_callback_add);
73 void dt_txn_callback_del(struct dt_device *dev, struct dt_txn_callback *cb)
75 list_del_init(&cb->dtc_linkage);
77 EXPORT_SYMBOL(dt_txn_callback_del);
79 int dt_txn_hook_start(const struct lu_env *env,
80 struct dt_device *dev, struct thandle *th)
83 struct dt_txn_callback *cb;
88 list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
89 struct thandle *dtc_th = th;
91 if (cb->dtc_txn_start == NULL ||
92 !(cb->dtc_tag & env->le_ctx.lc_tags))
95 /* Usually dt_txn_hook_start is called from bottom device,
96 * and if the thandle has th_top, then we need use top
97 * thandle for the callback in the top thandle layer */
98 if (th->th_top != NULL)
101 rc = cb->dtc_txn_start(env, dtc_th, cb->dtc_cookie);
107 EXPORT_SYMBOL(dt_txn_hook_start);
109 int dt_txn_hook_stop(const struct lu_env *env, struct thandle *th)
111 struct dt_device *dev = th->th_dev;
112 struct dt_txn_callback *cb;
118 if (OBD_FAIL_CHECK(OBD_FAIL_DT_TXN_STOP))
121 list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
122 struct thandle *dtc_th = th;
124 if (cb->dtc_txn_stop == NULL ||
125 !(cb->dtc_tag & env->le_ctx.lc_tags))
128 /* Usually dt_txn_hook_stop is called from bottom device,
129 * and if the thandle has th_top, then we need use top
130 * thandle for the callback in the top thandle layer */
131 if (th->th_top != NULL)
134 rc = cb->dtc_txn_stop(env, dtc_th, cb->dtc_cookie);
140 EXPORT_SYMBOL(dt_txn_hook_stop);
142 void dt_txn_hook_commit(struct thandle *th)
144 struct dt_txn_callback *cb;
149 list_for_each_entry(cb, &th->th_dev->dd_txn_callbacks,
151 /* Right now, the bottom device (OSD) will use this hook
152 * commit to notify OSP, so we do not check and replace
153 * the thandle to top thandle now */
154 if (cb->dtc_txn_commit)
155 cb->dtc_txn_commit(th, cb->dtc_cookie);
158 EXPORT_SYMBOL(dt_txn_hook_commit);
160 int dt_device_init(struct dt_device *dev, struct lu_device_type *t)
162 INIT_LIST_HEAD(&dev->dd_txn_callbacks);
163 return lu_device_init(&dev->dd_lu_dev, t);
165 EXPORT_SYMBOL(dt_device_init);
167 void dt_device_fini(struct dt_device *dev)
169 lu_device_fini(&dev->dd_lu_dev);
171 EXPORT_SYMBOL(dt_device_fini);
173 int dt_object_init(struct dt_object *obj,
174 struct lu_object_header *h, struct lu_device *d)
177 return lu_object_init(&obj->do_lu, h, d);
179 EXPORT_SYMBOL(dt_object_init);
181 void dt_object_fini(struct dt_object *obj)
183 lu_object_fini(&obj->do_lu);
185 EXPORT_SYMBOL(dt_object_fini);
187 int dt_try_as_dir(const struct lu_env *env, struct dt_object *obj)
189 if (obj->do_index_ops == NULL)
190 obj->do_ops->do_index_try(env, obj, &dt_directory_features);
191 return obj->do_index_ops != NULL;
193 EXPORT_SYMBOL(dt_try_as_dir);
195 enum dt_format_type dt_mode_to_dft(__u32 mode)
197 enum dt_format_type result;
199 switch (mode & S_IFMT) {
204 result = DFT_REGULAR;
221 EXPORT_SYMBOL(dt_mode_to_dft);
224 * lookup fid for object named \a name in directory \a dir.
227 int dt_lookup_dir(const struct lu_env *env, struct dt_object *dir,
228 const char *name, struct lu_fid *fid)
230 if (dt_try_as_dir(env, dir))
231 return dt_lookup(env, dir, (struct dt_rec *)fid,
232 (const struct dt_key *)name);
235 EXPORT_SYMBOL(dt_lookup_dir);
237 /* this differs from dt_locate by top_dev as parameter
238 * but not one from lu_site */
239 struct dt_object *dt_locate_at(const struct lu_env *env,
240 struct dt_device *dev,
241 const struct lu_fid *fid,
242 struct lu_device *top_dev,
243 const struct lu_object_conf *conf)
245 struct lu_object *lo;
248 lo = lu_object_find_at(env, top_dev, fid, conf);
250 return ERR_PTR(PTR_ERR(lo));
254 list_for_each_entry(n, &lo->lo_header->loh_layers, lo_linkage) {
255 if (n->lo_dev == &dev->dd_lu_dev)
256 return container_of0(n, struct dt_object, do_lu);
259 return ERR_PTR(-ENOENT);
261 EXPORT_SYMBOL(dt_locate_at);
264 * find an object named \a entry in given \a dfh->dfh_o directory.
266 static int dt_find_entry(const struct lu_env *env, const char *entry, void *data)
268 struct dt_find_hint *dfh = data;
269 struct dt_device *dt = dfh->dfh_dt;
270 struct lu_fid *fid = dfh->dfh_fid;
271 struct dt_object *obj = dfh->dfh_o;
274 result = dt_lookup_dir(env, obj, entry, fid);
275 lu_object_put(env, &obj->do_lu);
277 obj = dt_locate(env, dt, fid);
279 result = PTR_ERR(obj);
286 * Abstract function which parses path name. This function feeds
287 * path component to \a entry_func.
289 int dt_path_parser(const struct lu_env *env,
290 char *path, dt_entry_func_t entry_func,
297 e = strsep(&path, "/");
302 if (!path || path[0] == '\0')
306 rc = entry_func(env, e, data);
315 dt_store_resolve(const struct lu_env *env, struct dt_device *dt,
316 const char *path, struct lu_fid *fid)
318 struct dt_thread_info *info = dt_info(env);
319 struct dt_find_hint *dfh = &info->dti_dfh;
320 struct dt_object *obj;
327 strlcpy(info->dti_buf, path, sizeof(info->dti_buf));
329 result = dt->dd_ops->dt_root_get(env, dt, fid);
331 obj = dt_locate(env, dt, fid);
334 result = dt_path_parser(env, info->dti_buf,
337 obj = ERR_PTR(result);
342 obj = ERR_PTR(result);
347 static struct dt_object *dt_reg_open(const struct lu_env *env,
348 struct dt_device *dt,
356 result = dt_lookup_dir(env, p, name, fid);
358 o = dt_locate(env, dt, fid);
367 * Open dt object named \a filename from \a dirname directory.
368 * \param dt dt device
369 * \param fid on success, object fid is stored in *fid
371 struct dt_object *dt_store_open(const struct lu_env *env,
372 struct dt_device *dt,
374 const char *filename,
377 struct dt_object *file;
378 struct dt_object *dir;
380 dir = dt_store_resolve(env, dt, dirname, fid);
382 file = dt_reg_open(env, dt, dir,
384 lu_object_put(env, &dir->do_lu);
391 struct dt_object *dt_find_or_create(const struct lu_env *env,
392 struct dt_device *dt,
393 const struct lu_fid *fid,
394 struct dt_object_format *dof,
397 struct dt_object *dto;
403 dto = dt_locate(env, dt, fid);
407 LASSERT(dto != NULL);
408 if (dt_object_exists(dto))
411 th = dt_trans_create(env, dt);
413 GOTO(out, rc = PTR_ERR(th));
415 rc = dt_declare_create(env, dto, at, NULL, dof, th);
417 GOTO(trans_stop, rc);
419 rc = dt_trans_start_local(env, dt, th);
421 GOTO(trans_stop, rc);
423 dt_write_lock(env, dto, 0);
424 if (dt_object_exists(dto))
425 GOTO(unlock, rc = 0);
427 CDEBUG(D_OTHER, "create new object "DFID"\n", PFID(fid));
429 rc = dt_create(env, dto, at, NULL, dof, th);
432 LASSERT(dt_object_exists(dto));
434 dt_write_unlock(env, dto);
436 dt_trans_stop(env, dt, th);
439 lu_object_put(env, &dto->do_lu);
444 EXPORT_SYMBOL(dt_find_or_create);
446 /* dt class init function. */
447 int dt_global_init(void)
451 LU_CONTEXT_KEY_INIT(&dt_key);
452 result = lu_context_key_register(&dt_key);
456 void dt_global_fini(void)
458 lu_context_key_degister(&dt_key);
462 * Generic read helper. May return an error for partial reads.
464 * \param env lustre environment
465 * \param dt object to be read
466 * \param buf lu_buf to be filled, with buffer pointer and length
467 * \param pos position to start reading, updated as data is read
469 * \retval real size of data read
470 * \retval -ve errno on failure
472 int dt_read(const struct lu_env *env, struct dt_object *dt,
473 struct lu_buf *buf, loff_t *pos)
475 LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
476 return dt->do_body_ops->dbo_read(env, dt, buf, pos);
478 EXPORT_SYMBOL(dt_read);
481 * Read structures of fixed size from storage. Unlike dt_read(), using
482 * dt_record_read() will return an error for partial reads.
484 * \param env lustre environment
485 * \param dt object to be read
486 * \param buf lu_buf to be filled, with buffer pointer and length
487 * \param pos position to start reading, updated as data is read
489 * \retval 0 on successfully reading full buffer
490 * \retval -EFAULT on short read
491 * \retval -ve errno on failure
493 int dt_record_read(const struct lu_env *env, struct dt_object *dt,
494 struct lu_buf *buf, loff_t *pos)
498 LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
500 size = dt->do_body_ops->dbo_read(env, dt, buf, pos);
503 return (size == (ssize_t)buf->lb_len) ? 0 : -EFAULT;
505 EXPORT_SYMBOL(dt_record_read);
507 int dt_record_write(const struct lu_env *env, struct dt_object *dt,
508 const struct lu_buf *buf, loff_t *pos, struct thandle *th)
512 LASSERTF(dt != NULL, "dt is NULL when we want to write record\n");
514 LASSERT(dt->do_body_ops);
515 LASSERT(dt->do_body_ops->dbo_write);
517 size = dt->do_body_ops->dbo_write(env, dt, buf, pos, th, 1);
520 return (size == (ssize_t)buf->lb_len) ? 0 : -EFAULT;
522 EXPORT_SYMBOL(dt_record_write);
524 int dt_declare_version_set(const struct lu_env *env, struct dt_object *o,
528 char *xname = XATTR_NAME_VERSION;
532 vbuf.lb_len = sizeof(dt_obj_version_t);
533 return dt_declare_xattr_set(env, o, &vbuf, xname, 0, th);
536 EXPORT_SYMBOL(dt_declare_version_set);
538 void dt_version_set(const struct lu_env *env, struct dt_object *o,
539 dt_obj_version_t version, struct thandle *th)
542 char *xname = XATTR_NAME_VERSION;
546 vbuf.lb_buf = &version;
547 vbuf.lb_len = sizeof(version);
549 rc = dt_xattr_set(env, o, &vbuf, xname, 0, th);
551 CDEBUG(D_INODE, "Can't set version, rc %d\n", rc);
554 EXPORT_SYMBOL(dt_version_set);
556 dt_obj_version_t dt_version_get(const struct lu_env *env, struct dt_object *o)
559 char *xname = XATTR_NAME_VERSION;
560 dt_obj_version_t version;
564 vbuf.lb_buf = &version;
565 vbuf.lb_len = sizeof(version);
566 rc = dt_xattr_get(env, o, &vbuf, xname);
567 if (rc != sizeof(version)) {
568 CDEBUG(D_INODE, "Can't get version, rc %d\n", rc);
573 EXPORT_SYMBOL(dt_version_get);
575 /* list of all supported index types */
578 const struct dt_index_features dt_directory_features;
579 EXPORT_SYMBOL(dt_directory_features);
582 const struct dt_index_features dt_otable_features;
583 EXPORT_SYMBOL(dt_otable_features);
586 const struct dt_index_features dt_lfsck_orphan_features = {
588 .dif_keysize_min = sizeof(struct lu_fid),
589 .dif_keysize_max = sizeof(struct lu_fid),
590 .dif_recsize_min = sizeof(struct lu_orphan_rec),
591 .dif_recsize_max = sizeof(struct lu_orphan_rec),
594 EXPORT_SYMBOL(dt_lfsck_orphan_features);
597 const struct dt_index_features dt_lfsck_features = {
598 .dif_flags = DT_IND_UPDATE,
599 .dif_keysize_min = sizeof(struct lu_fid),
600 .dif_keysize_max = sizeof(struct lu_fid),
601 .dif_recsize_min = sizeof(__u8),
602 .dif_recsize_max = sizeof(__u8),
605 EXPORT_SYMBOL(dt_lfsck_features);
607 /* accounting indexes */
608 const struct dt_index_features dt_acct_features = {
609 .dif_flags = DT_IND_UPDATE,
610 .dif_keysize_min = sizeof(__u64), /* 64-bit uid/gid */
611 .dif_keysize_max = sizeof(__u64), /* 64-bit uid/gid */
612 .dif_recsize_min = sizeof(struct lquota_acct_rec), /* 16 bytes */
613 .dif_recsize_max = sizeof(struct lquota_acct_rec), /* 16 bytes */
616 EXPORT_SYMBOL(dt_acct_features);
618 /* global quota files */
619 const struct dt_index_features dt_quota_glb_features = {
620 .dif_flags = DT_IND_UPDATE,
621 /* a different key would have to be used for per-directory quota */
622 .dif_keysize_min = sizeof(__u64), /* 64-bit uid/gid */
623 .dif_keysize_max = sizeof(__u64), /* 64-bit uid/gid */
624 .dif_recsize_min = sizeof(struct lquota_glb_rec), /* 32 bytes */
625 .dif_recsize_max = sizeof(struct lquota_glb_rec), /* 32 bytes */
628 EXPORT_SYMBOL(dt_quota_glb_features);
630 /* slave quota files */
631 const struct dt_index_features dt_quota_slv_features = {
632 .dif_flags = DT_IND_UPDATE,
633 /* a different key would have to be used for per-directory quota */
634 .dif_keysize_min = sizeof(__u64), /* 64-bit uid/gid */
635 .dif_keysize_max = sizeof(__u64), /* 64-bit uid/gid */
636 .dif_recsize_min = sizeof(struct lquota_slv_rec), /* 8 bytes */
637 .dif_recsize_max = sizeof(struct lquota_slv_rec), /* 8 bytes */
640 EXPORT_SYMBOL(dt_quota_slv_features);
642 /* nodemap files, nodemap_rec size asserted in nodemap_storage.c */
643 const struct dt_index_features dt_nodemap_features = {
644 .dif_flags = DT_IND_UPDATE,
645 .dif_keysize_min = sizeof(__u64), /* 64-bit nodemap/record id */
646 .dif_keysize_max = sizeof(__u64), /* 64-bit nodemap/record id */
647 .dif_recsize_min = sizeof(union nodemap_rec), /* 32 bytes */
648 .dif_recsize_max = sizeof(union nodemap_rec), /* 32 bytes */
651 EXPORT_SYMBOL(dt_nodemap_features);
653 /* helper function returning what dt_index_features structure should be used
654 * based on the FID sequence. This is used by OBD_IDX_READ RPC */
655 static inline const struct dt_index_features *dt_index_feat_select(__u64 seq,
658 if (seq == FID_SEQ_QUOTA_GLB) {
659 /* global quota index */
661 /* global quota index should be a regular file */
662 return ERR_PTR(-ENOENT);
663 return &dt_quota_glb_features;
664 } else if (seq == FID_SEQ_QUOTA) {
665 /* quota slave index */
667 /* slave index should be a regular file */
668 return ERR_PTR(-ENOENT);
669 return &dt_quota_slv_features;
670 } else if (seq == FID_SEQ_LAYOUT_RBTREE){
671 return &dt_lfsck_orphan_features;
672 } else if (seq >= FID_SEQ_NORMAL) {
673 /* object is part of the namespace, verify that it is a
676 /* sorry, we can only deal with directory */
677 return ERR_PTR(-ENOTDIR);
678 return &dt_directory_features;
681 return ERR_PTR(-EOPNOTSUPP);
685 * Fill a lu_idxpage with key/record pairs read for transfer via OBD_IDX_READ
688 * \param env - is the environment passed by the caller
689 * \param lp - is a pointer to the lu_page to fill
690 * \param nob - is the maximum number of bytes that should be copied
691 * \param iops - is the index operation vector associated with the index object
692 * \param it - is a pointer to the current iterator
693 * \param attr - is the index attribute to pass to iops->rec()
694 * \param arg - is a pointer to the idx_info structure
696 static int dt_index_page_build(const struct lu_env *env, union lu_page *lp,
697 size_t nob, const struct dt_it_ops *iops,
698 struct dt_it *it, __u32 attr, void *arg)
700 struct idx_info *ii = (struct idx_info *)arg;
701 struct lu_idxpage *lip = &lp->lp_idx;
707 if (nob < LIP_HDR_SIZE)
710 /* initialize the header of the new container */
711 memset(lip, 0, LIP_HDR_SIZE);
712 lip->lip_magic = LIP_MAGIC;
715 /* compute size needed to store a key/record pair */
716 size = ii->ii_recsize + ii->ii_keysize;
717 if ((ii->ii_flags & II_FL_NOHASH) == 0)
718 /* add hash if the client wants it */
719 size += sizeof(__u64);
721 entry = lip->lip_entries;
723 char *tmp_entry = entry;
729 /* fetch 64-bit hash value */
730 hash = iops->store(env, it);
731 ii->ii_hash_end = hash;
733 if (OBD_FAIL_CHECK(OBD_FAIL_OBD_IDX_READ_BREAK)) {
734 if (lip->lip_nr != 0)
739 if (lip->lip_nr == 0)
740 GOTO(out, rc = -EINVAL);
744 if (!(ii->ii_flags & II_FL_NOHASH)) {
745 /* client wants to the 64-bit hash value associated with
747 memcpy(tmp_entry, &hash, sizeof(hash));
748 tmp_entry += sizeof(hash);
751 if (ii->ii_flags & II_FL_VARKEY)
752 keysize = iops->key_size(env, it);
754 keysize = ii->ii_keysize;
756 if (!(ii->ii_flags & II_FL_NOKEY)) {
757 /* then the key value */
758 key = iops->key(env, it);
759 memcpy(tmp_entry, key, keysize);
760 tmp_entry += keysize;
763 /* and finally the record */
764 rc = iops->rec(env, it, (struct dt_rec *)tmp_entry, attr);
769 /* hash/key/record successfully copied! */
771 if (unlikely(lip->lip_nr == 1 && ii->ii_count == 0))
772 ii->ii_hash_start = hash;
774 if (ii->ii_flags & II_FL_VARREC)
775 recsize = iops->rec_size(env, it, attr);
777 recsize = ii->ii_recsize;
779 entry = tmp_entry + recsize;
783 /* move on to the next record */
785 rc = iops->next(env, it);
786 } while (rc == -ESTALE);
792 if (rc >= 0 && lip->lip_nr > 0)
793 /* one more container */
796 /* no more entries */
797 ii->ii_hash_end = II_END_OFF;
803 * Walk index and fill lu_page containers with key/record pairs
805 * \param env - is the environment passed by the caller
806 * \param obj - is the index object to parse
807 * \param rdpg - is the lu_rdpg descriptor associated with the transfer
808 * \param filler - is the callback function responsible for filling a lu_page
809 * with key/record pairs in the format wanted by the caller.
810 * If NULL, uses dt_index_page_build
811 * \param arg - is an opaq argument passed to the filler function
813 * \retval sum (in bytes) of all filled lu_pages
814 * \retval -ve errno on failure
816 int dt_index_walk(const struct lu_env *env, struct dt_object *obj,
817 const struct lu_rdpg *rdpg, dt_index_page_build_t filler,
821 const struct dt_it_ops *iops;
822 size_t pageidx, nob, nlupgs = 0;
826 LASSERT(rdpg->rp_pages != NULL);
827 LASSERT(obj->do_index_ops != NULL);
830 filler = dt_index_page_build;
832 nob = rdpg->rp_count;
836 /* Iterate through index and fill containers from @rdpg */
837 iops = &obj->do_index_ops->dio_it;
838 LASSERT(iops != NULL);
839 it = iops->init(env, obj, rdpg->rp_attrs);
843 rc = iops->load(env, it, rdpg->rp_hash);
846 * Iterator didn't find record with exactly the key requested.
848 * It is currently either
850 * - positioned above record with key less than
851 * requested---skip it.
852 * - or not positioned at all (is in IAM_IT_SKEWED
853 * state)---position it on the next item.
855 rc = iops->next(env, it);
864 /* Fill containers one after the other. There might be multiple
865 * containers per physical page.
867 * At this point and across for-loop:
868 * rc == 0 -> ok, proceed.
869 * rc > 0 -> end of index.
870 * rc < 0 -> error. */
871 for (pageidx = 0; rc == 0 && nob > 0; pageidx++) {
875 LASSERT(pageidx < rdpg->rp_npages);
876 lp = kmap(rdpg->rp_pages[pageidx]);
879 for (i = 0; i < LU_PAGE_COUNT; i++, lp++, nob -= LU_PAGE_SIZE) {
880 rc = filler(env, lp, min_t(size_t, nob, LU_PAGE_SIZE),
881 iops, it, rdpg->rp_attrs, arg);
884 /* one more lu_page */
890 kunmap(rdpg->rp_pages[i]);
898 rc = min_t(size_t, nlupgs * LU_PAGE_SIZE, rdpg->rp_count);
902 EXPORT_SYMBOL(dt_index_walk);
905 * Walk key/record pairs of an index and copy them into 4KB containers to be
906 * transferred over the network. This is the common handler for OBD_IDX_READ
909 * \param env - is the environment passed by the caller
910 * \param dev - is the dt_device storing the index
911 * \param ii - is the idx_info structure packed by the client in the
912 * OBD_IDX_READ request
913 * \param rdpg - is the lu_rdpg descriptor
915 * \retval on success, return sum (in bytes) of all filled containers
916 * \retval appropriate error otherwise.
918 int dt_index_read(const struct lu_env *env, struct dt_device *dev,
919 struct idx_info *ii, const struct lu_rdpg *rdpg)
921 const struct dt_index_features *feat;
922 struct dt_object *obj;
926 /* rp_count shouldn't be null and should be a multiple of the container
928 if (rdpg->rp_count == 0 || (rdpg->rp_count & (LU_PAGE_SIZE - 1)) != 0)
931 if (!fid_is_quota(&ii->ii_fid) && !fid_is_layout_rbtree(&ii->ii_fid) &&
932 !fid_is_norm(&ii->ii_fid))
935 /* lookup index object subject to the transfer */
936 obj = dt_locate(env, dev, &ii->ii_fid);
938 RETURN(PTR_ERR(obj));
939 if (dt_object_exists(obj) == 0)
940 GOTO(out, rc = -ENOENT);
942 /* fetch index features associated with index object */
943 feat = dt_index_feat_select(fid_seq(&ii->ii_fid),
944 lu_object_attr(&obj->do_lu));
946 GOTO(out, rc = PTR_ERR(feat));
948 /* load index feature if not done already */
949 if (obj->do_index_ops == NULL) {
950 rc = obj->do_ops->do_index_try(env, obj, feat);
955 /* fill ii_flags with supported index features */
956 ii->ii_flags &= (II_FL_NOHASH | II_FL_NOKEY | II_FL_VARKEY |
959 if (!(feat->dif_flags & DT_IND_VARKEY))
960 ii->ii_keysize = feat->dif_keysize_max;
962 if (!(feat->dif_flags & DT_IND_VARREC))
963 ii->ii_recsize = feat->dif_recsize_max;
965 if (feat->dif_flags & DT_IND_NONUNQ)
966 /* key isn't necessarily unique */
967 ii->ii_flags |= II_FL_NONUNQ;
969 if (!fid_is_layout_rbtree(&ii->ii_fid)) {
970 dt_read_lock(env, obj, 0);
971 /* fetch object version before walking the index */
972 ii->ii_version = dt_version_get(env, obj);
975 /* walk the index and fill lu_idxpages with key/record pairs */
976 rc = dt_index_walk(env, obj, rdpg, dt_index_page_build, ii);
977 if (!fid_is_layout_rbtree(&ii->ii_fid))
978 dt_read_unlock(env, obj);
982 LASSERT(ii->ii_count == 0);
983 ii->ii_hash_end = II_END_OFF;
988 lu_object_put(env, &obj->do_lu);
991 EXPORT_SYMBOL(dt_index_read);
993 #ifdef CONFIG_PROC_FS
994 int lprocfs_dt_blksize_seq_show(struct seq_file *m, void *v)
996 struct dt_device *dt = m->private;
997 struct obd_statfs osfs;
999 int rc = dt_statfs(NULL, dt, &osfs);
1001 seq_printf(m, "%u\n", (unsigned) osfs.os_bsize);
1004 EXPORT_SYMBOL(lprocfs_dt_blksize_seq_show);
1006 int lprocfs_dt_kbytestotal_seq_show(struct seq_file *m, void *v)
1008 struct dt_device *dt = m->private;
1009 struct obd_statfs osfs;
1011 int rc = dt_statfs(NULL, dt, &osfs);
1013 __u32 blk_size = osfs.os_bsize >> 10;
1014 __u64 result = osfs.os_blocks;
1016 while (blk_size >>= 1)
1019 seq_printf(m, "%llu\n", result);
1023 EXPORT_SYMBOL(lprocfs_dt_kbytestotal_seq_show);
1025 int lprocfs_dt_kbytesfree_seq_show(struct seq_file *m, void *v)
1027 struct dt_device *dt = m->private;
1028 struct obd_statfs osfs;
1030 int rc = dt_statfs(NULL, dt, &osfs);
1032 __u32 blk_size = osfs.os_bsize >> 10;
1033 __u64 result = osfs.os_bfree;
1035 while (blk_size >>= 1)
1038 seq_printf(m, "%llu\n", result);
1042 EXPORT_SYMBOL(lprocfs_dt_kbytesfree_seq_show);
1044 int lprocfs_dt_kbytesavail_seq_show(struct seq_file *m, void *v)
1046 struct dt_device *dt = m->private;
1047 struct obd_statfs osfs;
1049 int rc = dt_statfs(NULL, dt, &osfs);
1051 __u32 blk_size = osfs.os_bsize >> 10;
1052 __u64 result = osfs.os_bavail;
1054 while (blk_size >>= 1)
1057 seq_printf(m, "%llu\n", result);
1061 EXPORT_SYMBOL(lprocfs_dt_kbytesavail_seq_show);
1063 int lprocfs_dt_filestotal_seq_show(struct seq_file *m, void *v)
1065 struct dt_device *dt = m->private;
1066 struct obd_statfs osfs;
1068 int rc = dt_statfs(NULL, dt, &osfs);
1070 seq_printf(m, "%llu\n", osfs.os_files);
1073 EXPORT_SYMBOL(lprocfs_dt_filestotal_seq_show);
1075 int lprocfs_dt_filesfree_seq_show(struct seq_file *m, void *v)
1077 struct dt_device *dt = m->private;
1078 struct obd_statfs osfs;
1080 int rc = dt_statfs(NULL, dt, &osfs);
1082 seq_printf(m, "%llu\n", osfs.os_ffree);
1085 EXPORT_SYMBOL(lprocfs_dt_filesfree_seq_show);
1087 #endif /* CONFIG_PROC_FS */