4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, 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
47 #include <dt_object.h>
48 #include <linux/list.h>
50 #include <lustre_fid.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
63 EXPORT_SYMBOL(dt_key
);
65 /* no lock is necessary to protect the list, because call-backs
66 * are added during system startup. Please refer to "struct dt_device".
68 void dt_txn_callback_add(struct dt_device
*dev
, struct dt_txn_callback
*cb
)
70 list_add(&cb
->dtc_linkage
, &dev
->dd_txn_callbacks
);
72 EXPORT_SYMBOL(dt_txn_callback_add
);
74 void dt_txn_callback_del(struct dt_device
*dev
, struct dt_txn_callback
*cb
)
76 list_del_init(&cb
->dtc_linkage
);
78 EXPORT_SYMBOL(dt_txn_callback_del
);
80 int dt_txn_hook_start(const struct lu_env
*env
,
81 struct dt_device
*dev
, struct thandle
*th
)
84 struct dt_txn_callback
*cb
;
89 list_for_each_entry(cb
, &dev
->dd_txn_callbacks
, dtc_linkage
) {
90 if (cb
->dtc_txn_start
== NULL
||
91 !(cb
->dtc_tag
& env
->le_ctx
.lc_tags
))
93 rc
= cb
->dtc_txn_start(env
, th
, cb
->dtc_cookie
);
99 EXPORT_SYMBOL(dt_txn_hook_start
);
101 int dt_txn_hook_stop(const struct lu_env
*env
, struct thandle
*txn
)
103 struct dt_device
*dev
= txn
->th_dev
;
104 struct dt_txn_callback
*cb
;
110 list_for_each_entry(cb
, &dev
->dd_txn_callbacks
, dtc_linkage
) {
111 if (cb
->dtc_txn_stop
== NULL
||
112 !(cb
->dtc_tag
& env
->le_ctx
.lc_tags
))
114 rc
= cb
->dtc_txn_stop(env
, txn
, cb
->dtc_cookie
);
120 EXPORT_SYMBOL(dt_txn_hook_stop
);
122 void dt_txn_hook_commit(struct thandle
*txn
)
124 struct dt_txn_callback
*cb
;
129 list_for_each_entry(cb
, &txn
->th_dev
->dd_txn_callbacks
,
131 if (cb
->dtc_txn_commit
)
132 cb
->dtc_txn_commit(txn
, cb
->dtc_cookie
);
135 EXPORT_SYMBOL(dt_txn_hook_commit
);
137 int dt_device_init(struct dt_device
*dev
, struct lu_device_type
*t
)
140 INIT_LIST_HEAD(&dev
->dd_txn_callbacks
);
141 return lu_device_init(&dev
->dd_lu_dev
, t
);
143 EXPORT_SYMBOL(dt_device_init
);
145 void dt_device_fini(struct dt_device
*dev
)
147 lu_device_fini(&dev
->dd_lu_dev
);
149 EXPORT_SYMBOL(dt_device_fini
);
151 int dt_object_init(struct dt_object
*obj
,
152 struct lu_object_header
*h
, struct lu_device
*d
)
155 return lu_object_init(&obj
->do_lu
, h
, d
);
157 EXPORT_SYMBOL(dt_object_init
);
159 void dt_object_fini(struct dt_object
*obj
)
161 lu_object_fini(&obj
->do_lu
);
163 EXPORT_SYMBOL(dt_object_fini
);
165 int dt_try_as_dir(const struct lu_env
*env
, struct dt_object
*obj
)
167 if (obj
->do_index_ops
== NULL
)
168 obj
->do_ops
->do_index_try(env
, obj
, &dt_directory_features
);
169 return obj
->do_index_ops
!= NULL
;
171 EXPORT_SYMBOL(dt_try_as_dir
);
173 enum dt_format_type
dt_mode_to_dft(__u32 mode
)
175 enum dt_format_type result
;
177 switch (mode
& S_IFMT
) {
182 result
= DFT_REGULAR
;
199 EXPORT_SYMBOL(dt_mode_to_dft
);
202 * lookup fid for object named \a name in directory \a dir.
205 int dt_lookup_dir(const struct lu_env
*env
, struct dt_object
*dir
,
206 const char *name
, struct lu_fid
*fid
)
208 if (dt_try_as_dir(env
, dir
))
209 return dt_lookup(env
, dir
, (struct dt_rec
*)fid
,
210 (const struct dt_key
*)name
, BYPASS_CAPA
);
213 EXPORT_SYMBOL(dt_lookup_dir
);
215 /* this differs from dt_locate by top_dev as parameter
216 * but not one from lu_site */
217 struct dt_object
*dt_locate_at(const struct lu_env
*env
,
218 struct dt_device
*dev
, const struct lu_fid
*fid
,
219 struct lu_device
*top_dev
)
221 struct lu_object
*lo
, *n
;
224 lo
= lu_object_find_at(env
, top_dev
, fid
, NULL
);
230 list_for_each_entry(n
, &lo
->lo_header
->loh_layers
, lo_linkage
) {
231 if (n
->lo_dev
== &dev
->dd_lu_dev
)
232 return container_of0(n
, struct dt_object
, do_lu
);
234 return ERR_PTR(-ENOENT
);
236 EXPORT_SYMBOL(dt_locate_at
);
239 * find a object named \a entry in given \a dfh->dfh_o directory.
241 static int dt_find_entry(const struct lu_env
*env
, const char *entry
, void *data
)
243 struct dt_find_hint
*dfh
= data
;
244 struct dt_device
*dt
= dfh
->dfh_dt
;
245 struct lu_fid
*fid
= dfh
->dfh_fid
;
246 struct dt_object
*obj
= dfh
->dfh_o
;
249 result
= dt_lookup_dir(env
, obj
, entry
, fid
);
250 lu_object_put(env
, &obj
->do_lu
);
252 obj
= dt_locate(env
, dt
, fid
);
254 result
= PTR_ERR(obj
);
261 * Abstract function which parses path name. This function feeds
262 * path component to \a entry_func.
264 int dt_path_parser(const struct lu_env
*env
,
265 char *path
, dt_entry_func_t entry_func
,
272 e
= strsep(&path
, "/");
277 if (!path
|| path
[0] == '\0')
281 rc
= entry_func(env
, e
, data
);
290 dt_store_resolve(const struct lu_env
*env
, struct dt_device
*dt
,
291 const char *path
, struct lu_fid
*fid
)
293 struct dt_thread_info
*info
= dt_info(env
);
294 struct dt_find_hint
*dfh
= &info
->dti_dfh
;
295 struct dt_object
*obj
;
296 char *local
= info
->dti_buf
;
303 strncpy(local
, path
, DT_MAX_PATH
);
304 local
[DT_MAX_PATH
- 1] = '\0';
306 result
= dt
->dd_ops
->dt_root_get(env
, dt
, fid
);
308 obj
= dt_locate(env
, dt
, fid
);
311 result
= dt_path_parser(env
, local
, dt_find_entry
, dfh
);
313 obj
= ERR_PTR(result
);
318 obj
= ERR_PTR(result
);
322 EXPORT_SYMBOL(dt_store_resolve
);
324 static struct dt_object
*dt_reg_open(const struct lu_env
*env
,
325 struct dt_device
*dt
,
333 result
= dt_lookup_dir(env
, p
, name
, fid
);
335 o
= dt_locate(env
, dt
, fid
);
344 * Open dt object named \a filename from \a dirname directory.
345 * \param dt dt device
346 * \param fid on success, object fid is stored in *fid
348 struct dt_object
*dt_store_open(const struct lu_env
*env
,
349 struct dt_device
*dt
,
351 const char *filename
,
354 struct dt_object
*file
;
355 struct dt_object
*dir
;
357 dir
= dt_store_resolve(env
, dt
, dirname
, fid
);
359 file
= dt_reg_open(env
, dt
, dir
,
361 lu_object_put(env
, &dir
->do_lu
);
367 EXPORT_SYMBOL(dt_store_open
);
369 struct dt_object
*dt_find_or_create(const struct lu_env
*env
,
370 struct dt_device
*dt
,
371 const struct lu_fid
*fid
,
372 struct dt_object_format
*dof
,
375 struct dt_object
*dto
;
381 dto
= dt_locate(env
, dt
, fid
);
385 LASSERT(dto
!= NULL
);
386 if (dt_object_exists(dto
))
389 th
= dt_trans_create(env
, dt
);
391 GOTO(out
, rc
= PTR_ERR(th
));
393 rc
= dt_declare_create(env
, dto
, at
, NULL
, dof
, th
);
395 GOTO(trans_stop
, rc
);
397 rc
= dt_trans_start_local(env
, dt
, th
);
399 GOTO(trans_stop
, rc
);
401 dt_write_lock(env
, dto
, 0);
402 if (dt_object_exists(dto
))
403 GOTO(unlock
, rc
= 0);
405 CDEBUG(D_OTHER
, "create new object "DFID
"\n", PFID(fid
));
407 rc
= dt_create(env
, dto
, at
, NULL
, dof
, th
);
410 LASSERT(dt_object_exists(dto
));
412 dt_write_unlock(env
, dto
);
414 dt_trans_stop(env
, dt
, th
);
417 lu_object_put(env
, &dto
->do_lu
);
422 EXPORT_SYMBOL(dt_find_or_create
);
424 /* dt class init function. */
425 int dt_global_init(void)
429 LU_CONTEXT_KEY_INIT(&dt_key
);
430 result
= lu_context_key_register(&dt_key
);
434 void dt_global_fini(void)
436 lu_context_key_degister(&dt_key
);
440 * Generic read helper. May return an error for partial reads.
442 * \param env lustre environment
443 * \param dt object to be read
444 * \param buf lu_buf to be filled, with buffer pointer and length
445 * \param pos position to start reading, updated as data is read
447 * \retval real size of data read
448 * \retval -ve errno on failure
450 int dt_read(const struct lu_env
*env
, struct dt_object
*dt
,
451 struct lu_buf
*buf
, loff_t
*pos
)
453 LASSERTF(dt
!= NULL
, "dt is NULL when we want to read record\n");
454 return dt
->do_body_ops
->dbo_read(env
, dt
, buf
, pos
, BYPASS_CAPA
);
456 EXPORT_SYMBOL(dt_read
);
459 * Read structures of fixed size from storage. Unlike dt_read(), using
460 * dt_record_read() will return an error for partial reads.
462 * \param env lustre environment
463 * \param dt object to be read
464 * \param buf lu_buf to be filled, with buffer pointer and length
465 * \param pos position to start reading, updated as data is read
467 * \retval 0 on successfully reading full buffer
468 * \retval -EFAULT on short read
469 * \retval -ve errno on failure
471 int dt_record_read(const struct lu_env
*env
, struct dt_object
*dt
,
472 struct lu_buf
*buf
, loff_t
*pos
)
476 LASSERTF(dt
!= NULL
, "dt is NULL when we want to read record\n");
478 rc
= dt
->do_body_ops
->dbo_read(env
, dt
, buf
, pos
, BYPASS_CAPA
);
480 if (rc
== buf
->lb_len
)
486 EXPORT_SYMBOL(dt_record_read
);
488 int dt_record_write(const struct lu_env
*env
, struct dt_object
*dt
,
489 const struct lu_buf
*buf
, loff_t
*pos
, struct thandle
*th
)
493 LASSERTF(dt
!= NULL
, "dt is NULL when we want to write record\n");
495 LASSERT(dt
->do_body_ops
);
496 LASSERT(dt
->do_body_ops
->dbo_write
);
497 rc
= dt
->do_body_ops
->dbo_write(env
, dt
, buf
, pos
, th
, BYPASS_CAPA
, 1);
498 if (rc
== buf
->lb_len
)
504 EXPORT_SYMBOL(dt_record_write
);
506 int dt_declare_version_set(const struct lu_env
*env
, struct dt_object
*o
,
510 char *xname
= XATTR_NAME_VERSION
;
514 vbuf
.lb_len
= sizeof(dt_obj_version_t
);
515 return dt_declare_xattr_set(env
, o
, &vbuf
, xname
, 0, th
);
518 EXPORT_SYMBOL(dt_declare_version_set
);
520 void dt_version_set(const struct lu_env
*env
, struct dt_object
*o
,
521 dt_obj_version_t version
, struct thandle
*th
)
524 char *xname
= XATTR_NAME_VERSION
;
528 vbuf
.lb_buf
= &version
;
529 vbuf
.lb_len
= sizeof(version
);
531 rc
= dt_xattr_set(env
, o
, &vbuf
, xname
, 0, th
, BYPASS_CAPA
);
533 CDEBUG(D_INODE
, "Can't set version, rc %d\n", rc
);
536 EXPORT_SYMBOL(dt_version_set
);
538 dt_obj_version_t
dt_version_get(const struct lu_env
*env
, struct dt_object
*o
)
541 char *xname
= XATTR_NAME_VERSION
;
542 dt_obj_version_t version
;
546 vbuf
.lb_buf
= &version
;
547 vbuf
.lb_len
= sizeof(version
);
548 rc
= dt_xattr_get(env
, o
, &vbuf
, xname
, BYPASS_CAPA
);
549 if (rc
!= sizeof(version
)) {
550 CDEBUG(D_INODE
, "Can't get version, rc %d\n", rc
);
555 EXPORT_SYMBOL(dt_version_get
);
557 /* list of all supported index types */
560 const struct dt_index_features dt_directory_features
;
561 EXPORT_SYMBOL(dt_directory_features
);
564 const struct dt_index_features dt_otable_features
;
565 EXPORT_SYMBOL(dt_otable_features
);
568 const struct dt_index_features dt_lfsck_features
= {
569 .dif_flags
= DT_IND_UPDATE
,
570 .dif_keysize_min
= sizeof(struct lu_fid
),
571 .dif_keysize_max
= sizeof(struct lu_fid
),
572 .dif_recsize_min
= sizeof(__u8
),
573 .dif_recsize_max
= sizeof(__u8
),
576 EXPORT_SYMBOL(dt_lfsck_features
);
578 /* accounting indexes */
579 const struct dt_index_features dt_acct_features
= {
580 .dif_flags
= DT_IND_UPDATE
,
581 .dif_keysize_min
= sizeof(__u64
), /* 64-bit uid/gid */
582 .dif_keysize_max
= sizeof(__u64
), /* 64-bit uid/gid */
583 .dif_recsize_min
= sizeof(struct lquota_acct_rec
), /* 16 bytes */
584 .dif_recsize_max
= sizeof(struct lquota_acct_rec
), /* 16 bytes */
587 EXPORT_SYMBOL(dt_acct_features
);
589 /* global quota files */
590 const struct dt_index_features dt_quota_glb_features
= {
591 .dif_flags
= DT_IND_UPDATE
,
592 /* a different key would have to be used for per-directory quota */
593 .dif_keysize_min
= sizeof(__u64
), /* 64-bit uid/gid */
594 .dif_keysize_max
= sizeof(__u64
), /* 64-bit uid/gid */
595 .dif_recsize_min
= sizeof(struct lquota_glb_rec
), /* 32 bytes */
596 .dif_recsize_max
= sizeof(struct lquota_glb_rec
), /* 32 bytes */
599 EXPORT_SYMBOL(dt_quota_glb_features
);
601 /* slave quota files */
602 const struct dt_index_features dt_quota_slv_features
= {
603 .dif_flags
= DT_IND_UPDATE
,
604 /* a different key would have to be used for per-directory quota */
605 .dif_keysize_min
= sizeof(__u64
), /* 64-bit uid/gid */
606 .dif_keysize_max
= sizeof(__u64
), /* 64-bit uid/gid */
607 .dif_recsize_min
= sizeof(struct lquota_slv_rec
), /* 8 bytes */
608 .dif_recsize_max
= sizeof(struct lquota_slv_rec
), /* 8 bytes */
611 EXPORT_SYMBOL(dt_quota_slv_features
);
613 /* helper function returning what dt_index_features structure should be used
614 * based on the FID sequence. This is used by OBD_IDX_READ RPC */
615 static inline const struct dt_index_features
*dt_index_feat_select(__u64 seq
,
618 if (seq
== FID_SEQ_QUOTA_GLB
) {
619 /* global quota index */
621 /* global quota index should be a regular file */
622 return ERR_PTR(-ENOENT
);
623 return &dt_quota_glb_features
;
624 } else if (seq
== FID_SEQ_QUOTA
) {
625 /* quota slave index */
627 /* slave index should be a regular file */
628 return ERR_PTR(-ENOENT
);
629 return &dt_quota_slv_features
;
630 } else if (seq
>= FID_SEQ_NORMAL
) {
631 /* object is part of the namespace, verify that it is a
634 /* sorry, we can only deal with directory */
635 return ERR_PTR(-ENOTDIR
);
636 return &dt_directory_features
;
639 return ERR_PTR(-EOPNOTSUPP
);
643 * Fill a lu_idxpage with key/record pairs read for transfer via OBD_IDX_READ
646 * \param env - is the environment passed by the caller
647 * \param lp - is a pointer to the lu_page to fill
648 * \param nob - is the maximum number of bytes that should be copied
649 * \param iops - is the index operation vector associated with the index object
650 * \param it - is a pointer to the current iterator
651 * \param attr - is the index attribute to pass to iops->rec()
652 * \param arg - is a pointer to the idx_info structure
654 static int dt_index_page_build(const struct lu_env
*env
, union lu_page
*lp
,
655 int nob
, const struct dt_it_ops
*iops
,
656 struct dt_it
*it
, __u32 attr
, void *arg
)
658 struct idx_info
*ii
= (struct idx_info
*)arg
;
659 struct lu_idxpage
*lip
= &lp
->lp_idx
;
664 /* no support for variable key & record size for now */
665 LASSERT((ii
->ii_flags
& II_FL_VARKEY
) == 0);
666 LASSERT((ii
->ii_flags
& II_FL_VARREC
) == 0);
668 /* initialize the header of the new container */
669 memset(lip
, 0, LIP_HDR_SIZE
);
670 lip
->lip_magic
= LIP_MAGIC
;
673 /* compute size needed to store a key/record pair */
674 size
= ii
->ii_recsize
+ ii
->ii_keysize
;
675 if ((ii
->ii_flags
& II_FL_NOHASH
) == 0)
676 /* add hash if the client wants it */
677 size
+= sizeof(__u64
);
679 entry
= lip
->lip_entries
;
681 char *tmp_entry
= entry
;
685 /* fetch 64-bit hash value */
686 hash
= iops
->store(env
, it
);
687 ii
->ii_hash_end
= hash
;
689 if (OBD_FAIL_CHECK(OBD_FAIL_OBD_IDX_READ_BREAK
)) {
690 if (lip
->lip_nr
!= 0)
695 if (lip
->lip_nr
== 0)
696 GOTO(out
, rc
= -EINVAL
);
700 if ((ii
->ii_flags
& II_FL_NOHASH
) == 0) {
701 /* client wants to the 64-bit hash value associated with
703 memcpy(tmp_entry
, &hash
, sizeof(hash
));
704 tmp_entry
+= sizeof(hash
);
707 /* then the key value */
708 LASSERT(iops
->key_size(env
, it
) == ii
->ii_keysize
);
709 key
= iops
->key(env
, it
);
710 memcpy(tmp_entry
, key
, ii
->ii_keysize
);
711 tmp_entry
+= ii
->ii_keysize
;
713 /* and finally the record */
714 rc
= iops
->rec(env
, it
, (struct dt_rec
*)tmp_entry
, attr
);
719 /* hash/key/record successfully copied! */
721 if (unlikely(lip
->lip_nr
== 1 && ii
->ii_count
== 0))
722 ii
->ii_hash_start
= hash
;
723 entry
= tmp_entry
+ ii
->ii_recsize
;
727 /* move on to the next record */
729 rc
= iops
->next(env
, it
);
730 } while (rc
== -ESTALE
);
736 if (rc
>= 0 && lip
->lip_nr
> 0)
737 /* one more container */
740 /* no more entries */
741 ii
->ii_hash_end
= II_END_OFF
;
746 * Walk index and fill lu_page containers with key/record pairs
748 * \param env - is the environment passed by the caller
749 * \param obj - is the index object to parse
750 * \param rdpg - is the lu_rdpg descriptor associated with the transfer
751 * \param filler - is the callback function responsible for filling a lu_page
752 * with key/record pairs in the format wanted by the caller
753 * \param arg - is an opaq argument passed to the filler function
755 * \retval sum (in bytes) of all filled lu_pages
756 * \retval -ve errno on failure
758 int dt_index_walk(const struct lu_env
*env
, struct dt_object
*obj
,
759 const struct lu_rdpg
*rdpg
, dt_index_page_build_t filler
,
763 const struct dt_it_ops
*iops
;
764 unsigned int pageidx
, nob
, nlupgs
= 0;
768 LASSERT(rdpg
->rp_pages
!= NULL
);
769 LASSERT(obj
->do_index_ops
!= NULL
);
771 nob
= rdpg
->rp_count
;
775 /* Iterate through index and fill containers from @rdpg */
776 iops
= &obj
->do_index_ops
->dio_it
;
777 LASSERT(iops
!= NULL
);
778 it
= iops
->init(env
, obj
, rdpg
->rp_attrs
, BYPASS_CAPA
);
782 rc
= iops
->load(env
, it
, rdpg
->rp_hash
);
785 * Iterator didn't find record with exactly the key requested.
787 * It is currently either
789 * - positioned above record with key less than
790 * requested---skip it.
791 * - or not positioned at all (is in IAM_IT_SKEWED
792 * state)---position it on the next item.
794 rc
= iops
->next(env
, it
);
799 /* Fill containers one after the other. There might be multiple
800 * containers per physical page.
802 * At this point and across for-loop:
803 * rc == 0 -> ok, proceed.
804 * rc > 0 -> end of index.
805 * rc < 0 -> error. */
806 for (pageidx
= 0; rc
== 0 && nob
> 0; pageidx
++) {
810 LASSERT(pageidx
< rdpg
->rp_npages
);
811 lp
= kmap(rdpg
->rp_pages
[pageidx
]);
814 for (i
= 0; i
< LU_PAGE_COUNT
; i
++, lp
++, nob
-= LU_PAGE_SIZE
) {
815 rc
= filler(env
, lp
, min_t(int, nob
, LU_PAGE_SIZE
),
816 iops
, it
, rdpg
->rp_attrs
, arg
);
819 /* one more lu_page */
825 kunmap(rdpg
->rp_pages
[i
]);
832 rc
= min_t(unsigned int, nlupgs
* LU_PAGE_SIZE
, rdpg
->rp_count
);
836 EXPORT_SYMBOL(dt_index_walk
);
839 * Walk key/record pairs of an index and copy them into 4KB containers to be
840 * transferred over the network. This is the common handler for OBD_IDX_READ
843 * \param env - is the environment passed by the caller
844 * \param dev - is the dt_device storing the index
845 * \param ii - is the idx_info structure packed by the client in the
846 * OBD_IDX_READ request
847 * \param rdpg - is the lu_rdpg descriptor
849 * \retval on success, return sum (in bytes) of all filled containers
850 * \retval appropriate error otherwise.
852 int dt_index_read(const struct lu_env
*env
, struct dt_device
*dev
,
853 struct idx_info
*ii
, const struct lu_rdpg
*rdpg
)
855 const struct dt_index_features
*feat
;
856 struct dt_object
*obj
;
860 /* rp_count shouldn't be null and should be a multiple of the container
862 if (rdpg
->rp_count
<= 0 && (rdpg
->rp_count
& (LU_PAGE_SIZE
- 1)) != 0)
865 if (fid_seq(&ii
->ii_fid
) >= FID_SEQ_NORMAL
)
866 /* we don't support directory transfer via OBD_IDX_READ for the
870 if (!fid_is_quota(&ii
->ii_fid
))
871 /* block access to all local files except quota files */
874 /* lookup index object subject to the transfer */
875 obj
= dt_locate(env
, dev
, &ii
->ii_fid
);
877 RETURN(PTR_ERR(obj
));
878 if (dt_object_exists(obj
) == 0)
879 GOTO(out
, rc
= -ENOENT
);
881 /* fetch index features associated with index object */
882 feat
= dt_index_feat_select(fid_seq(&ii
->ii_fid
),
883 lu_object_attr(&obj
->do_lu
));
885 GOTO(out
, rc
= PTR_ERR(feat
));
887 /* load index feature if not done already */
888 if (obj
->do_index_ops
== NULL
) {
889 rc
= obj
->do_ops
->do_index_try(env
, obj
, feat
);
894 /* fill ii_flags with supported index features */
895 ii
->ii_flags
&= II_FL_NOHASH
;
897 ii
->ii_keysize
= feat
->dif_keysize_max
;
898 if ((feat
->dif_flags
& DT_IND_VARKEY
) != 0) {
899 /* key size is variable */
900 ii
->ii_flags
|= II_FL_VARKEY
;
901 /* we don't support variable key size for the time being */
902 GOTO(out
, rc
= -EOPNOTSUPP
);
905 ii
->ii_recsize
= feat
->dif_recsize_max
;
906 if ((feat
->dif_flags
& DT_IND_VARREC
) != 0) {
907 /* record size is variable */
908 ii
->ii_flags
|= II_FL_VARREC
;
909 /* we don't support variable record size for the time being */
910 GOTO(out
, rc
= -EOPNOTSUPP
);
913 if ((feat
->dif_flags
& DT_IND_NONUNQ
) != 0)
914 /* key isn't necessarily unique */
915 ii
->ii_flags
|= II_FL_NONUNQ
;
917 dt_read_lock(env
, obj
, 0);
918 /* fetch object version before walking the index */
919 ii
->ii_version
= dt_version_get(env
, obj
);
921 /* walk the index and fill lu_idxpages with key/record pairs */
922 rc
= dt_index_walk(env
, obj
, rdpg
, dt_index_page_build
,ii
);
923 dt_read_unlock(env
, obj
);
927 LASSERT(ii
->ii_count
== 0);
928 ii
->ii_hash_end
= II_END_OFF
;
933 lu_object_put(env
, &obj
->do_lu
);
936 EXPORT_SYMBOL(dt_index_read
);
940 int lprocfs_dt_rd_blksize(char *page
, char **start
, off_t off
,
941 int count
, int *eof
, void *data
)
943 struct dt_device
*dt
= data
;
944 struct obd_statfs osfs
;
946 int rc
= dt_statfs(NULL
, dt
, &osfs
);
949 rc
= snprintf(page
, count
, "%u\n",
950 (unsigned) osfs
.os_bsize
);
955 EXPORT_SYMBOL(lprocfs_dt_rd_blksize
);
957 int lprocfs_dt_rd_kbytestotal(char *page
, char **start
, off_t off
,
958 int count
, int *eof
, void *data
)
960 struct dt_device
*dt
= data
;
961 struct obd_statfs osfs
;
963 int rc
= dt_statfs(NULL
, dt
, &osfs
);
965 __u32 blk_size
= osfs
.os_bsize
>> 10;
966 __u64 result
= osfs
.os_blocks
;
968 while (blk_size
>>= 1)
972 rc
= snprintf(page
, count
, LPU64
"\n", result
);
977 EXPORT_SYMBOL(lprocfs_dt_rd_kbytestotal
);
979 int lprocfs_dt_rd_kbytesfree(char *page
, char **start
, off_t off
,
980 int count
, int *eof
, void *data
)
982 struct dt_device
*dt
= data
;
983 struct obd_statfs osfs
;
985 int rc
= dt_statfs(NULL
, dt
, &osfs
);
987 __u32 blk_size
= osfs
.os_bsize
>> 10;
988 __u64 result
= osfs
.os_bfree
;
990 while (blk_size
>>= 1)
994 rc
= snprintf(page
, count
, LPU64
"\n", result
);
999 EXPORT_SYMBOL(lprocfs_dt_rd_kbytesfree
);
1001 int lprocfs_dt_rd_kbytesavail(char *page
, char **start
, off_t off
,
1002 int count
, int *eof
, void *data
)
1004 struct dt_device
*dt
= data
;
1005 struct obd_statfs osfs
;
1007 int rc
= dt_statfs(NULL
, dt
, &osfs
);
1009 __u32 blk_size
= osfs
.os_bsize
>> 10;
1010 __u64 result
= osfs
.os_bavail
;
1012 while (blk_size
>>= 1)
1016 rc
= snprintf(page
, count
, LPU64
"\n", result
);
1021 EXPORT_SYMBOL(lprocfs_dt_rd_kbytesavail
);
1023 int lprocfs_dt_rd_filestotal(char *page
, char **start
, off_t off
,
1024 int count
, int *eof
, void *data
)
1026 struct dt_device
*dt
= data
;
1027 struct obd_statfs osfs
;
1029 int rc
= dt_statfs(NULL
, dt
, &osfs
);
1032 rc
= snprintf(page
, count
, LPU64
"\n", osfs
.os_files
);
1037 EXPORT_SYMBOL(lprocfs_dt_rd_filestotal
);
1039 int lprocfs_dt_rd_filesfree(char *page
, char **start
, off_t off
,
1040 int count
, int *eof
, void *data
)
1042 struct dt_device
*dt
= data
;
1043 struct obd_statfs osfs
;
1045 int rc
= dt_statfs(NULL
, dt
, &osfs
);
1048 rc
= snprintf(page
, count
, LPU64
"\n", osfs
.os_ffree
);
1053 EXPORT_SYMBOL(lprocfs_dt_rd_filesfree
);
1055 #endif /* LPROCFS */