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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/obdclass/dt_object.c
35 * Generic functions from dt_object.h
37 * Author: Nikita Danilov <nikita@clusterfs.com>
40 #define DEBUG_SUBSYSTEM S_CLASS
42 #include <linux/list.h>
44 #include <dt_object.h>
46 #include <lustre_fid.h>
47 #include <lustre_nodemap.h>
48 #include <lustre_quota.h>
49 #include <lustre_lfsck.h>
51 /* context key constructor/destructor: dt_global_key_init, dt_global_key_fini */
52 LU_KEY_INIT(dt_global, struct dt_thread_info);
53 LU_KEY_FINI(dt_global, struct dt_thread_info);
55 struct lu_context_key dt_key = {
56 .lct_tags = LCT_MD_THREAD | LCT_DT_THREAD | LCT_MG_THREAD | LCT_LOCAL,
57 .lct_init = dt_global_key_init,
58 .lct_fini = dt_global_key_fini
61 /* no lock is necessary to protect the list, because call-backs
62 * are added during system startup. Please refer to "struct dt_device".
64 void dt_txn_callback_add(struct dt_device *dev, struct dt_txn_callback *cb)
66 list_add(&cb->dtc_linkage, &dev->dd_txn_callbacks);
68 EXPORT_SYMBOL(dt_txn_callback_add);
70 void dt_txn_callback_del(struct dt_device *dev, struct dt_txn_callback *cb)
72 list_del_init(&cb->dtc_linkage);
74 EXPORT_SYMBOL(dt_txn_callback_del);
76 int dt_txn_hook_start(const struct lu_env *env,
77 struct dt_device *dev, struct thandle *th)
80 struct dt_txn_callback *cb;
85 list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
86 struct thandle *dtc_th = th;
88 if (cb->dtc_txn_start == NULL ||
89 !(cb->dtc_tag & env->le_ctx.lc_tags))
92 /* Usually dt_txn_hook_start is called from bottom device,
93 * and if the thandle has th_top, then we need use top
94 * thandle for the callback in the top thandle layer */
95 if (th->th_top != NULL)
98 rc = cb->dtc_txn_start(env, dtc_th, cb->dtc_cookie);
104 EXPORT_SYMBOL(dt_txn_hook_start);
106 int dt_txn_hook_stop(const struct lu_env *env, struct thandle *th)
108 struct dt_device *dev = th->th_dev;
109 struct dt_txn_callback *cb;
115 if (OBD_FAIL_CHECK(OBD_FAIL_DT_TXN_STOP))
118 list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
119 struct thandle *dtc_th = th;
121 if (cb->dtc_txn_stop == NULL ||
122 !(cb->dtc_tag & env->le_ctx.lc_tags))
125 /* Usually dt_txn_hook_stop is called from bottom device,
126 * and if the thandle has th_top, then we need use top
127 * thandle for the callback in the top thandle layer */
128 if (th->th_top != NULL)
131 rc = cb->dtc_txn_stop(env, dtc_th, cb->dtc_cookie);
137 EXPORT_SYMBOL(dt_txn_hook_stop);
139 void dt_txn_hook_commit(struct thandle *th)
141 struct dt_txn_callback *cb;
146 list_for_each_entry(cb, &th->th_dev->dd_txn_callbacks,
148 /* Right now, the bottom device (OSD) will use this hook
149 * commit to notify OSP, so we do not check and replace
150 * the thandle to top thandle now */
151 if (cb->dtc_txn_commit)
152 cb->dtc_txn_commit(th, cb->dtc_cookie);
155 EXPORT_SYMBOL(dt_txn_hook_commit);
157 int dt_device_init(struct dt_device *dev, struct lu_device_type *t)
159 INIT_LIST_HEAD(&dev->dd_txn_callbacks);
160 return lu_device_init(&dev->dd_lu_dev, t);
162 EXPORT_SYMBOL(dt_device_init);
164 void dt_device_fini(struct dt_device *dev)
166 lu_device_fini(&dev->dd_lu_dev);
168 EXPORT_SYMBOL(dt_device_fini);
170 int dt_object_init(struct dt_object *obj,
171 struct lu_object_header *h, struct lu_device *d)
174 return lu_object_init(&obj->do_lu, h, d);
176 EXPORT_SYMBOL(dt_object_init);
178 void dt_object_fini(struct dt_object *obj)
180 lu_object_fini(&obj->do_lu);
182 EXPORT_SYMBOL(dt_object_fini);
184 int dt_try_as_dir(const struct lu_env *env, struct dt_object *obj)
186 if (obj->do_index_ops == NULL)
187 obj->do_ops->do_index_try(env, obj, &dt_directory_features);
188 return obj->do_index_ops != NULL;
190 EXPORT_SYMBOL(dt_try_as_dir);
192 enum dt_format_type dt_mode_to_dft(__u32 mode)
194 enum dt_format_type result;
196 switch (mode & S_IFMT) {
201 result = DFT_REGULAR;
218 EXPORT_SYMBOL(dt_mode_to_dft);
221 * lookup fid for object named \a name in directory \a dir.
224 int dt_lookup_dir(const struct lu_env *env, struct dt_object *dir,
225 const char *name, struct lu_fid *fid)
227 if (dt_try_as_dir(env, dir))
228 return dt_lookup(env, dir, (struct dt_rec *)fid,
229 (const struct dt_key *)name);
232 EXPORT_SYMBOL(dt_lookup_dir);
234 /* this differs from dt_locate by top_dev as parameter
235 * but not one from lu_site */
236 struct dt_object *dt_locate_at(const struct lu_env *env,
237 struct dt_device *dev,
238 const struct lu_fid *fid,
239 struct lu_device *top_dev,
240 const struct lu_object_conf *conf)
242 struct lu_object *lo;
245 lo = lu_object_find_at(env, top_dev, fid, conf);
247 return ERR_PTR(PTR_ERR(lo));
251 list_for_each_entry(n, &lo->lo_header->loh_layers, lo_linkage) {
252 if (n->lo_dev == &dev->dd_lu_dev)
253 return container_of0(n, struct dt_object, do_lu);
256 lu_object_put(env, lo);
257 return ERR_PTR(-ENOENT);
259 EXPORT_SYMBOL(dt_locate_at);
262 * find an object named \a entry in given \a dfh->dfh_o directory.
264 static int dt_find_entry(const struct lu_env *env, const char *entry,
267 struct dt_find_hint *dfh = data;
268 struct dt_device *dt = dfh->dfh_dt;
269 struct lu_fid *fid = dfh->dfh_fid;
270 struct dt_object *obj = dfh->dfh_o;
273 rc = dt_lookup_dir(env, obj, entry, fid);
274 dt_object_put(env, obj);
276 obj = dt_locate(env, dt, fid);
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, struct dt_device *dt,
372 const char *dirname, const char *filename,
375 struct dt_object *file;
376 struct dt_object *dir;
378 dir = dt_store_resolve(env, dt, dirname, fid);
380 file = dt_reg_open(env, dt, dir, filename, fid);
381 dt_object_put(env, dir);
389 struct dt_object *dt_find_or_create(const struct lu_env *env,
390 struct dt_device *dt,
391 const struct lu_fid *fid,
392 struct dt_object_format *dof,
395 struct dt_object *dto;
401 dto = dt_locate(env, dt, fid);
405 LASSERT(dto != NULL);
406 if (dt_object_exists(dto))
409 th = dt_trans_create(env, dt);
411 GOTO(out, rc = PTR_ERR(th));
413 rc = dt_declare_create(env, dto, at, NULL, dof, th);
415 GOTO(trans_stop, rc);
417 rc = dt_trans_start_local(env, dt, th);
419 GOTO(trans_stop, rc);
421 dt_write_lock(env, dto, 0);
422 if (dt_object_exists(dto))
423 GOTO(unlock, rc = 0);
425 CDEBUG(D_OTHER, "create new object "DFID"\n", PFID(fid));
427 rc = dt_create(env, dto, at, NULL, dof, th);
430 LASSERT(dt_object_exists(dto));
432 dt_write_unlock(env, dto);
434 dt_trans_stop(env, dt, th);
437 dt_object_put(env, dto);
443 EXPORT_SYMBOL(dt_find_or_create);
445 /* dt class init function. */
446 int dt_global_init(void)
450 LU_CONTEXT_KEY_INIT(&dt_key);
451 result = lu_context_key_register(&dt_key);
455 void dt_global_fini(void)
457 lu_context_key_degister(&dt_key);
461 * Generic read helper. May return an error for partial reads.
463 * \param env lustre environment
464 * \param dt object to be read
465 * \param buf lu_buf to be filled, with buffer pointer and length
466 * \param pos position to start reading, updated as data is read
468 * \retval real size of data read
469 * \retval -ve errno on failure
471 int dt_read(const struct lu_env *env, struct dt_object *dt,
472 struct lu_buf *buf, loff_t *pos)
474 LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
475 return dt->do_body_ops->dbo_read(env, dt, buf, pos);
477 EXPORT_SYMBOL(dt_read);
480 * Read structures of fixed size from storage. Unlike dt_read(), using
481 * dt_record_read() will return an error for partial reads.
483 * \param env lustre environment
484 * \param dt object to be read
485 * \param buf lu_buf to be filled, with buffer pointer and length
486 * \param pos position to start reading, updated as data is read
488 * \retval 0 on successfully reading full buffer
489 * \retval -EFAULT on short read
490 * \retval -ve errno on failure
492 int dt_record_read(const struct lu_env *env, struct dt_object *dt,
493 struct lu_buf *buf, loff_t *pos)
497 LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
499 size = dt->do_body_ops->dbo_read(env, dt, buf, pos);
502 return (size == (ssize_t)buf->lb_len) ? 0 : -EFAULT;
504 EXPORT_SYMBOL(dt_record_read);
506 int dt_record_write(const struct lu_env *env, struct dt_object *dt,
507 const struct lu_buf *buf, loff_t *pos, struct thandle *th)
511 LASSERTF(dt != NULL, "dt is NULL when we want to write record\n");
513 LASSERT(dt->do_body_ops);
514 LASSERT(dt->do_body_ops->dbo_write);
516 size = dt->do_body_ops->dbo_write(env, dt, buf, pos, th, 1);
519 return (size == (ssize_t)buf->lb_len) ? 0 : -EFAULT;
521 EXPORT_SYMBOL(dt_record_write);
523 int dt_declare_version_set(const struct lu_env *env, struct dt_object *o,
527 char *xname = XATTR_NAME_VERSION;
531 vbuf.lb_len = sizeof(dt_obj_version_t);
532 return dt_declare_xattr_set(env, o, &vbuf, xname, 0, th);
535 EXPORT_SYMBOL(dt_declare_version_set);
537 void dt_version_set(const struct lu_env *env, struct dt_object *o,
538 dt_obj_version_t version, struct thandle *th)
541 char *xname = XATTR_NAME_VERSION;
545 vbuf.lb_buf = &version;
546 vbuf.lb_len = sizeof(version);
548 rc = dt_xattr_set(env, o, &vbuf, xname, 0, th);
550 CDEBUG(D_INODE, "Can't set version, rc %d\n", rc);
553 EXPORT_SYMBOL(dt_version_set);
555 dt_obj_version_t dt_version_get(const struct lu_env *env, struct dt_object *o)
558 char *xname = XATTR_NAME_VERSION;
559 dt_obj_version_t version;
563 vbuf.lb_buf = &version;
564 vbuf.lb_len = sizeof(version);
565 rc = dt_xattr_get(env, o, &vbuf, xname);
566 if (rc != sizeof(version)) {
567 CDEBUG(D_INODE, "Can't get version, rc %d\n", rc);
572 EXPORT_SYMBOL(dt_version_get);
574 /* list of all supported index types */
577 const struct dt_index_features dt_directory_features;
578 EXPORT_SYMBOL(dt_directory_features);
581 const struct dt_index_features dt_otable_features;
582 EXPORT_SYMBOL(dt_otable_features);
584 /* lfsck layout orphan */
585 const struct dt_index_features dt_lfsck_layout_orphan_features = {
587 .dif_keysize_min = sizeof(struct lu_fid),
588 .dif_keysize_max = sizeof(struct lu_fid),
589 .dif_recsize_min = sizeof(struct lu_orphan_rec_v2),
590 .dif_recsize_max = sizeof(struct lu_orphan_rec_v2),
593 EXPORT_SYMBOL(dt_lfsck_layout_orphan_features);
595 /* lfsck layout dangling */
596 const struct dt_index_features dt_lfsck_layout_dangling_features = {
597 .dif_flags = DT_IND_UPDATE,
598 .dif_keysize_min = sizeof(struct lfsck_layout_dangling_key),
599 .dif_keysize_max = sizeof(struct lfsck_layout_dangling_key),
600 .dif_recsize_min = sizeof(struct lu_fid),
601 .dif_recsize_max = sizeof(struct lu_fid),
604 EXPORT_SYMBOL(dt_lfsck_layout_dangling_features);
606 /* lfsck namespace */
607 const struct dt_index_features dt_lfsck_namespace_features = {
608 .dif_flags = DT_IND_UPDATE,
609 .dif_keysize_min = sizeof(struct lu_fid),
610 .dif_keysize_max = sizeof(struct lu_fid),
611 .dif_recsize_min = sizeof(__u8),
612 .dif_recsize_max = sizeof(__u8),
615 EXPORT_SYMBOL(dt_lfsck_namespace_features);
617 /* accounting indexes */
618 const struct dt_index_features dt_acct_features = {
619 .dif_flags = DT_IND_UPDATE,
620 .dif_keysize_min = sizeof(__u64), /* 64-bit uid/gid */
621 .dif_keysize_max = sizeof(__u64), /* 64-bit uid/gid */
622 .dif_recsize_min = sizeof(struct lquota_acct_rec), /* 16 bytes */
623 .dif_recsize_max = sizeof(struct lquota_acct_rec), /* 16 bytes */
626 EXPORT_SYMBOL(dt_acct_features);
628 /* global quota files */
629 const struct dt_index_features dt_quota_glb_features = {
630 .dif_flags = DT_IND_UPDATE,
631 /* a different key would have to be used for per-directory quota */
632 .dif_keysize_min = sizeof(__u64), /* 64-bit uid/gid */
633 .dif_keysize_max = sizeof(__u64), /* 64-bit uid/gid */
634 .dif_recsize_min = sizeof(struct lquota_glb_rec), /* 32 bytes */
635 .dif_recsize_max = sizeof(struct lquota_glb_rec), /* 32 bytes */
638 EXPORT_SYMBOL(dt_quota_glb_features);
640 /* slave quota files */
641 const struct dt_index_features dt_quota_slv_features = {
642 .dif_flags = DT_IND_UPDATE,
643 /* a different key would have to be used for per-directory quota */
644 .dif_keysize_min = sizeof(__u64), /* 64-bit uid/gid */
645 .dif_keysize_max = sizeof(__u64), /* 64-bit uid/gid */
646 .dif_recsize_min = sizeof(struct lquota_slv_rec), /* 8 bytes */
647 .dif_recsize_max = sizeof(struct lquota_slv_rec), /* 8 bytes */
650 EXPORT_SYMBOL(dt_quota_slv_features);
652 /* nodemap files, nodemap_rec size asserted in nodemap_storage.c */
653 const struct dt_index_features dt_nodemap_features = {
654 .dif_flags = DT_IND_UPDATE,
655 .dif_keysize_min = sizeof(__u64), /* 64-bit nodemap/record id */
656 .dif_keysize_max = sizeof(__u64), /* 64-bit nodemap/record id */
657 .dif_recsize_min = sizeof(union nodemap_rec), /* 32 bytes */
658 .dif_recsize_max = sizeof(union nodemap_rec), /* 32 bytes */
661 EXPORT_SYMBOL(dt_nodemap_features);
663 /* helper function returning what dt_index_features structure should be used
664 * based on the FID sequence. This is used by OBD_IDX_READ RPC */
665 static inline const struct dt_index_features *dt_index_feat_select(__u64 seq,
668 if (seq == FID_SEQ_QUOTA_GLB) {
669 /* global quota index */
671 /* global quota index should be a regular file */
672 return ERR_PTR(-ENOENT);
673 return &dt_quota_glb_features;
674 } else if (seq == FID_SEQ_QUOTA) {
675 /* quota slave index */
677 /* slave index should be a regular file */
678 return ERR_PTR(-ENOENT);
679 return &dt_quota_slv_features;
680 } else if (seq == FID_SEQ_LAYOUT_RBTREE){
681 return &dt_lfsck_layout_orphan_features;
682 } else if (seq >= FID_SEQ_NORMAL) {
683 /* object is part of the namespace, verify that it is a
686 /* sorry, we can only deal with directory */
687 return ERR_PTR(-ENOTDIR);
688 return &dt_directory_features;
691 return ERR_PTR(-EOPNOTSUPP);
695 * Fill a lu_idxpage with key/record pairs read for transfer via OBD_IDX_READ
698 * \param env - is the environment passed by the caller
699 * \param lp - is a pointer to the lu_page to fill
700 * \param nob - is the maximum number of bytes that should be copied
701 * \param iops - is the index operation vector associated with the index object
702 * \param it - is a pointer to the current iterator
703 * \param attr - is the index attribute to pass to iops->rec()
704 * \param arg - is a pointer to the idx_info structure
706 static int dt_index_page_build(const struct lu_env *env, union lu_page *lp,
707 size_t nob, const struct dt_it_ops *iops,
708 struct dt_it *it, __u32 attr, void *arg)
710 struct idx_info *ii = (struct idx_info *)arg;
711 struct lu_idxpage *lip = &lp->lp_idx;
717 if (nob < LIP_HDR_SIZE)
720 /* initialize the header of the new container */
721 memset(lip, 0, LIP_HDR_SIZE);
722 lip->lip_magic = LIP_MAGIC;
725 /* compute size needed to store a key/record pair */
726 size = ii->ii_recsize + ii->ii_keysize;
727 if ((ii->ii_flags & II_FL_NOHASH) == 0)
728 /* add hash if the client wants it */
729 size += sizeof(__u64);
731 entry = lip->lip_entries;
733 char *tmp_entry = entry;
739 /* fetch 64-bit hash value */
740 hash = iops->store(env, it);
741 ii->ii_hash_end = hash;
743 if (OBD_FAIL_CHECK(OBD_FAIL_OBD_IDX_READ_BREAK)) {
744 if (lip->lip_nr != 0)
749 if (lip->lip_nr == 0)
750 GOTO(out, rc = -EINVAL);
754 if (!(ii->ii_flags & II_FL_NOHASH)) {
755 /* client wants to the 64-bit hash value associated with
757 memcpy(tmp_entry, &hash, sizeof(hash));
758 tmp_entry += sizeof(hash);
761 if (ii->ii_flags & II_FL_VARKEY)
762 keysize = iops->key_size(env, it);
764 keysize = ii->ii_keysize;
766 if (!(ii->ii_flags & II_FL_NOKEY)) {
767 /* then the key value */
768 key = iops->key(env, it);
769 memcpy(tmp_entry, key, keysize);
770 tmp_entry += keysize;
773 /* and finally the record */
774 rc = iops->rec(env, it, (struct dt_rec *)tmp_entry, attr);
779 /* hash/key/record successfully copied! */
781 if (unlikely(lip->lip_nr == 1 && ii->ii_count == 0))
782 ii->ii_hash_start = hash;
784 if (ii->ii_flags & II_FL_VARREC)
785 recsize = iops->rec_size(env, it, attr);
787 recsize = ii->ii_recsize;
789 entry = tmp_entry + recsize;
793 /* move on to the next record */
795 rc = iops->next(env, it);
796 } while (rc == -ESTALE);
802 if (rc >= 0 && lip->lip_nr > 0)
803 /* one more container */
806 /* no more entries */
807 ii->ii_hash_end = II_END_OFF;
813 * Walk index and fill lu_page containers with key/record pairs
815 * \param env - is the environment passed by the caller
816 * \param obj - is the index object to parse
817 * \param rdpg - is the lu_rdpg descriptor associated with the transfer
818 * \param filler - is the callback function responsible for filling a lu_page
819 * with key/record pairs in the format wanted by the caller.
820 * If NULL, uses dt_index_page_build
821 * \param arg - is an opaq argument passed to the filler function
823 * \retval sum (in bytes) of all filled lu_pages
824 * \retval -ve errno on failure
826 int dt_index_walk(const struct lu_env *env, struct dt_object *obj,
827 const struct lu_rdpg *rdpg, dt_index_page_build_t filler,
831 const struct dt_it_ops *iops;
832 size_t pageidx, nob, nlupgs = 0;
836 LASSERT(rdpg->rp_pages != NULL);
837 LASSERT(obj->do_index_ops != NULL);
840 filler = dt_index_page_build;
842 nob = rdpg->rp_count;
846 /* Iterate through index and fill containers from @rdpg */
847 iops = &obj->do_index_ops->dio_it;
848 LASSERT(iops != NULL);
849 it = iops->init(env, obj, rdpg->rp_attrs);
853 rc = iops->load(env, it, rdpg->rp_hash);
856 * Iterator didn't find record with exactly the key requested.
858 * It is currently either
860 * - positioned above record with key less than
861 * requested---skip it.
862 * - or not positioned at all (is in IAM_IT_SKEWED
863 * state)---position it on the next item.
865 rc = iops->next(env, it);
874 /* Fill containers one after the other. There might be multiple
875 * containers per physical page.
877 * At this point and across for-loop:
878 * rc == 0 -> ok, proceed.
879 * rc > 0 -> end of index.
880 * rc < 0 -> error. */
881 for (pageidx = 0; rc == 0 && nob > 0; pageidx++) {
885 LASSERT(pageidx < rdpg->rp_npages);
886 lp = kmap(rdpg->rp_pages[pageidx]);
889 for (i = 0; i < LU_PAGE_COUNT; i++, lp++, nob -= LU_PAGE_SIZE) {
890 rc = filler(env, lp, min_t(size_t, nob, LU_PAGE_SIZE),
891 iops, it, rdpg->rp_attrs, arg);
894 /* one more lu_page */
900 kunmap(rdpg->rp_pages[i]);
908 rc = min_t(size_t, nlupgs * LU_PAGE_SIZE, rdpg->rp_count);
912 EXPORT_SYMBOL(dt_index_walk);
915 * Walk key/record pairs of an index and copy them into 4KB containers to be
916 * transferred over the network. This is the common handler for OBD_IDX_READ
919 * \param env - is the environment passed by the caller
920 * \param dev - is the dt_device storing the index
921 * \param ii - is the idx_info structure packed by the client in the
922 * OBD_IDX_READ request
923 * \param rdpg - is the lu_rdpg descriptor
925 * \retval on success, return sum (in bytes) of all filled containers
926 * \retval appropriate error otherwise.
928 int dt_index_read(const struct lu_env *env, struct dt_device *dev,
929 struct idx_info *ii, const struct lu_rdpg *rdpg)
931 const struct dt_index_features *feat;
932 struct dt_object *obj;
936 /* rp_count shouldn't be null and should be a multiple of the container
938 if (rdpg->rp_count == 0 || (rdpg->rp_count & (LU_PAGE_SIZE - 1)) != 0)
941 if (!fid_is_quota(&ii->ii_fid) && !fid_is_layout_rbtree(&ii->ii_fid) &&
942 !fid_is_norm(&ii->ii_fid))
945 /* lookup index object subject to the transfer */
946 obj = dt_locate(env, dev, &ii->ii_fid);
948 RETURN(PTR_ERR(obj));
949 if (dt_object_exists(obj) == 0)
950 GOTO(out, rc = -ENOENT);
952 /* fetch index features associated with index object */
953 feat = dt_index_feat_select(fid_seq(&ii->ii_fid),
954 lu_object_attr(&obj->do_lu));
956 GOTO(out, rc = PTR_ERR(feat));
958 /* load index feature if not done already */
959 if (obj->do_index_ops == NULL) {
960 rc = obj->do_ops->do_index_try(env, obj, feat);
965 /* fill ii_flags with supported index features */
966 ii->ii_flags &= (II_FL_NOHASH | II_FL_NOKEY | II_FL_VARKEY |
969 if (!(feat->dif_flags & DT_IND_VARKEY))
970 ii->ii_keysize = feat->dif_keysize_max;
972 if (!(feat->dif_flags & DT_IND_VARREC))
973 ii->ii_recsize = feat->dif_recsize_max;
975 if (feat->dif_flags & DT_IND_NONUNQ)
976 /* key isn't necessarily unique */
977 ii->ii_flags |= II_FL_NONUNQ;
979 if (!fid_is_layout_rbtree(&ii->ii_fid)) {
980 dt_read_lock(env, obj, 0);
981 /* fetch object version before walking the index */
982 ii->ii_version = dt_version_get(env, obj);
985 /* walk the index and fill lu_idxpages with key/record pairs */
986 rc = dt_index_walk(env, obj, rdpg, dt_index_page_build, ii);
987 if (!fid_is_layout_rbtree(&ii->ii_fid))
988 dt_read_unlock(env, obj);
992 LASSERT(ii->ii_count == 0);
993 ii->ii_hash_end = II_END_OFF;
998 dt_object_put(env, obj);
1001 EXPORT_SYMBOL(dt_index_read);
1003 #ifdef CONFIG_PROC_FS
1004 int lprocfs_dt_blksize_seq_show(struct seq_file *m, void *v)
1006 struct dt_device *dt = m->private;
1007 struct obd_statfs osfs;
1009 int rc = dt_statfs(NULL, dt, &osfs);
1011 seq_printf(m, "%u\n", (unsigned) osfs.os_bsize);
1014 EXPORT_SYMBOL(lprocfs_dt_blksize_seq_show);
1016 int lprocfs_dt_kbytestotal_seq_show(struct seq_file *m, void *v)
1018 struct dt_device *dt = m->private;
1019 struct obd_statfs osfs;
1021 int rc = dt_statfs(NULL, dt, &osfs);
1023 __u32 blk_size = osfs.os_bsize >> 10;
1024 __u64 result = osfs.os_blocks;
1026 while (blk_size >>= 1)
1029 seq_printf(m, "%llu\n", result);
1033 EXPORT_SYMBOL(lprocfs_dt_kbytestotal_seq_show);
1035 int lprocfs_dt_kbytesfree_seq_show(struct seq_file *m, void *v)
1037 struct dt_device *dt = m->private;
1038 struct obd_statfs osfs;
1040 int rc = dt_statfs(NULL, dt, &osfs);
1042 __u32 blk_size = osfs.os_bsize >> 10;
1043 __u64 result = osfs.os_bfree;
1045 while (blk_size >>= 1)
1048 seq_printf(m, "%llu\n", result);
1052 EXPORT_SYMBOL(lprocfs_dt_kbytesfree_seq_show);
1054 int lprocfs_dt_kbytesavail_seq_show(struct seq_file *m, void *v)
1056 struct dt_device *dt = m->private;
1057 struct obd_statfs osfs;
1059 int rc = dt_statfs(NULL, dt, &osfs);
1061 __u32 blk_size = osfs.os_bsize >> 10;
1062 __u64 result = osfs.os_bavail;
1064 while (blk_size >>= 1)
1067 seq_printf(m, "%llu\n", result);
1071 EXPORT_SYMBOL(lprocfs_dt_kbytesavail_seq_show);
1073 int lprocfs_dt_filestotal_seq_show(struct seq_file *m, void *v)
1075 struct dt_device *dt = m->private;
1076 struct obd_statfs osfs;
1078 int rc = dt_statfs(NULL, dt, &osfs);
1080 seq_printf(m, "%llu\n", osfs.os_files);
1083 EXPORT_SYMBOL(lprocfs_dt_filestotal_seq_show);
1085 int lprocfs_dt_filesfree_seq_show(struct seq_file *m, void *v)
1087 struct dt_device *dt = m->private;
1088 struct obd_statfs osfs;
1090 int rc = dt_statfs(NULL, dt, &osfs);
1092 seq_printf(m, "%llu\n", osfs.os_ffree);
1095 EXPORT_SYMBOL(lprocfs_dt_filesfree_seq_show);
1097 #endif /* CONFIG_PROC_FS */