2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * Copyright (c) 2013 Red Hat, Inc.
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as
8 * published by the Free Software Foundation.
10 * This program is distributed in the hope that it would be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
25 #include "xfs_mount.h"
26 #include "xfs_da_format.h"
27 #include "xfs_da_btree.h"
28 #include "xfs_inode.h"
29 #include "xfs_trans.h"
30 #include "xfs_inode_item.h"
33 #include "xfs_attr_sf.h"
34 #include "xfs_attr_remote.h"
35 #include "xfs_attr_leaf.h"
36 #include "xfs_error.h"
37 #include "xfs_trace.h"
38 #include "xfs_buf_item.h"
39 #include "xfs_cksum.h"
43 xfs_attr_shortform_compare(const void *a
, const void *b
)
45 xfs_attr_sf_sort_t
*sa
, *sb
;
47 sa
= (xfs_attr_sf_sort_t
*)a
;
48 sb
= (xfs_attr_sf_sort_t
*)b
;
49 if (sa
->hash
< sb
->hash
) {
51 } else if (sa
->hash
> sb
->hash
) {
54 return sa
->entno
- sb
->entno
;
58 #define XFS_ISRESET_CURSOR(cursor) \
59 (!((cursor)->initted) && !((cursor)->hashval) && \
60 !((cursor)->blkno) && !((cursor)->offset))
62 * Copy out entries of shortform attribute lists for attr_list().
63 * Shortform attribute lists are not stored in hashval sorted order.
64 * If the output buffer is not large enough to hold them all, then we
65 * we have to calculate each entries' hashvalue and sort them before
66 * we can begin returning them to the user.
69 xfs_attr_shortform_list(xfs_attr_list_context_t
*context
)
71 attrlist_cursor_kern_t
*cursor
;
72 xfs_attr_sf_sort_t
*sbuf
, *sbp
;
73 xfs_attr_shortform_t
*sf
;
74 xfs_attr_sf_entry_t
*sfe
;
76 int sbsize
, nsbuf
, count
, i
;
78 ASSERT(context
!= NULL
);
81 ASSERT(dp
->i_afp
!= NULL
);
82 sf
= (xfs_attr_shortform_t
*)dp
->i_afp
->if_u1
.if_data
;
86 cursor
= context
->cursor
;
87 ASSERT(cursor
!= NULL
);
89 trace_xfs_attr_list_sf(context
);
92 * If the buffer is large enough and the cursor is at the start,
93 * do not bother with sorting since we will return everything in
94 * one buffer and another call using the cursor won't need to be
96 * Note the generous fudge factor of 16 overhead bytes per entry.
97 * If bufsize is zero then put_listent must be a search function
98 * and can just scan through what we have.
100 if (context
->bufsize
== 0 ||
101 (XFS_ISRESET_CURSOR(cursor
) &&
102 (dp
->i_afp
->if_bytes
+ sf
->hdr
.count
* 16) < context
->bufsize
)) {
103 for (i
= 0, sfe
= &sf
->list
[0]; i
< sf
->hdr
.count
; i
++) {
104 context
->put_listent(context
,
110 * Either search callback finished early or
111 * didn't fit it all in the buffer after all.
113 if (context
->seen_enough
)
115 sfe
= XFS_ATTR_SF_NEXTENTRY(sfe
);
117 trace_xfs_attr_list_sf_all(context
);
121 /* do no more for a search callback */
122 if (context
->bufsize
== 0)
126 * It didn't all fit, so we have to sort everything on hashval.
128 sbsize
= sf
->hdr
.count
* sizeof(*sbuf
);
129 sbp
= sbuf
= kmem_alloc(sbsize
, KM_SLEEP
| KM_NOFS
);
132 * Scan the attribute list for the rest of the entries, storing
133 * the relevant info from only those that match into a buffer.
136 for (i
= 0, sfe
= &sf
->list
[0]; i
< sf
->hdr
.count
; i
++) {
138 ((char *)sfe
< (char *)sf
) ||
139 ((char *)sfe
>= ((char *)sf
+ dp
->i_afp
->if_bytes
)))) {
140 XFS_CORRUPTION_ERROR("xfs_attr_shortform_list",
142 context
->dp
->i_mount
, sfe
);
144 return -EFSCORRUPTED
;
148 sbp
->hash
= xfs_da_hashname(sfe
->nameval
, sfe
->namelen
);
149 sbp
->name
= sfe
->nameval
;
150 sbp
->namelen
= sfe
->namelen
;
151 /* These are bytes, and both on-disk, don't endian-flip */
152 sbp
->valuelen
= sfe
->valuelen
;
153 sbp
->flags
= sfe
->flags
;
154 sfe
= XFS_ATTR_SF_NEXTENTRY(sfe
);
160 * Sort the entries on hash then entno.
162 xfs_sort(sbuf
, nsbuf
, sizeof(*sbuf
), xfs_attr_shortform_compare
);
165 * Re-find our place IN THE SORTED LIST.
170 for (sbp
= sbuf
, i
= 0; i
< nsbuf
; i
++, sbp
++) {
171 if (sbp
->hash
== cursor
->hashval
) {
172 if (cursor
->offset
== count
) {
176 } else if (sbp
->hash
> cursor
->hashval
) {
186 * Loop putting entries into the user buffer.
188 for ( ; i
< nsbuf
; i
++, sbp
++) {
189 if (cursor
->hashval
!= sbp
->hash
) {
190 cursor
->hashval
= sbp
->hash
;
193 context
->put_listent(context
,
198 if (context
->seen_enough
)
208 * We didn't find the block & hash mentioned in the cursor state, so
209 * walk down the attr btree looking for the hash.
212 xfs_attr_node_list_lookup(
213 struct xfs_attr_list_context
*context
,
214 struct attrlist_cursor_kern
*cursor
,
215 struct xfs_buf
**pbp
)
217 struct xfs_da3_icnode_hdr nodehdr
;
218 struct xfs_da_intnode
*node
;
219 struct xfs_da_node_entry
*btree
;
220 struct xfs_inode
*dp
= context
->dp
;
221 struct xfs_mount
*mp
= dp
->i_mount
;
222 struct xfs_trans
*tp
= context
->tp
;
226 unsigned int expected_level
= 0;
229 ASSERT(*pbp
== NULL
);
232 error
= xfs_da3_node_read(tp
, dp
, cursor
->blkno
, -1, &bp
,
237 magic
= be16_to_cpu(node
->hdr
.info
.magic
);
238 if (magic
== XFS_ATTR_LEAF_MAGIC
||
239 magic
== XFS_ATTR3_LEAF_MAGIC
)
241 if (magic
!= XFS_DA_NODE_MAGIC
&&
242 magic
!= XFS_DA3_NODE_MAGIC
) {
243 XFS_CORRUPTION_ERROR(__func__
, XFS_ERRLEVEL_LOW
, mp
,
248 dp
->d_ops
->node_hdr_from_disk(&nodehdr
, node
);
250 /* Tree taller than we can handle; bail out! */
251 if (nodehdr
.level
>= XFS_DA_NODE_MAXDEPTH
)
254 /* Check the level from the root node. */
255 if (cursor
->blkno
== 0)
256 expected_level
= nodehdr
.level
- 1;
257 else if (expected_level
!= nodehdr
.level
)
262 btree
= dp
->d_ops
->node_tree_p(node
);
263 for (i
= 0; i
< nodehdr
.count
; btree
++, i
++) {
264 if (cursor
->hashval
<= be32_to_cpu(btree
->hashval
)) {
265 cursor
->blkno
= be32_to_cpu(btree
->before
);
266 trace_xfs_attr_list_node_descend(context
,
271 xfs_trans_brelse(tp
, bp
);
273 if (i
== nodehdr
.count
)
276 /* We can't point back to the root. */
277 if (cursor
->blkno
== 0)
278 return -EFSCORRUPTED
;
281 if (expected_level
!= 0)
288 xfs_trans_brelse(tp
, bp
);
289 return -EFSCORRUPTED
;
294 struct xfs_attr_list_context
*context
)
296 struct xfs_attr3_icleaf_hdr leafhdr
;
297 struct attrlist_cursor_kern
*cursor
;
298 struct xfs_attr_leafblock
*leaf
;
299 struct xfs_da_intnode
*node
;
301 struct xfs_inode
*dp
= context
->dp
;
302 struct xfs_mount
*mp
= dp
->i_mount
;
305 trace_xfs_attr_node_list(context
);
307 cursor
= context
->cursor
;
311 * Do all sorts of validation on the passed-in cursor structure.
312 * If anything is amiss, ignore the cursor and look up the hashval
313 * starting from the btree root.
316 if (cursor
->blkno
> 0) {
317 error
= xfs_da3_node_read(context
->tp
, dp
, cursor
->blkno
, -1,
319 if ((error
!= 0) && (error
!= -EFSCORRUPTED
))
322 struct xfs_attr_leaf_entry
*entries
;
325 switch (be16_to_cpu(node
->hdr
.info
.magic
)) {
326 case XFS_DA_NODE_MAGIC
:
327 case XFS_DA3_NODE_MAGIC
:
328 trace_xfs_attr_list_wrong_blk(context
);
329 xfs_trans_brelse(context
->tp
, bp
);
332 case XFS_ATTR_LEAF_MAGIC
:
333 case XFS_ATTR3_LEAF_MAGIC
:
335 xfs_attr3_leaf_hdr_from_disk(mp
->m_attr_geo
,
337 entries
= xfs_attr3_leaf_entryp(leaf
);
338 if (cursor
->hashval
> be32_to_cpu(
339 entries
[leafhdr
.count
- 1].hashval
)) {
340 trace_xfs_attr_list_wrong_blk(context
);
341 xfs_trans_brelse(context
->tp
, bp
);
343 } else if (cursor
->hashval
<= be32_to_cpu(
344 entries
[0].hashval
)) {
345 trace_xfs_attr_list_wrong_blk(context
);
346 xfs_trans_brelse(context
->tp
, bp
);
351 trace_xfs_attr_list_wrong_blk(context
);
352 xfs_trans_brelse(context
->tp
, bp
);
359 * We did not find what we expected given the cursor's contents,
360 * so we start from the top and work down based on the hash value.
361 * Note that start of node block is same as start of leaf block.
364 error
= xfs_attr_node_list_lookup(context
, cursor
, &bp
);
371 * Roll upward through the blocks, processing each leaf block in
372 * order. As long as there is space in the result buffer, keep
373 * adding the information.
377 xfs_attr3_leaf_list_int(bp
, context
);
378 xfs_attr3_leaf_hdr_from_disk(mp
->m_attr_geo
, &leafhdr
, leaf
);
379 if (context
->seen_enough
|| leafhdr
.forw
== 0)
381 cursor
->blkno
= leafhdr
.forw
;
382 xfs_trans_brelse(context
->tp
, bp
);
383 error
= xfs_attr3_leaf_read(context
->tp
, dp
, cursor
->blkno
, -1, &bp
);
387 xfs_trans_brelse(context
->tp
, bp
);
392 * Copy out attribute list entries for attr_list(), for leaf attribute lists.
395 xfs_attr3_leaf_list_int(
397 struct xfs_attr_list_context
*context
)
399 struct attrlist_cursor_kern
*cursor
;
400 struct xfs_attr_leafblock
*leaf
;
401 struct xfs_attr3_icleaf_hdr ichdr
;
402 struct xfs_attr_leaf_entry
*entries
;
403 struct xfs_attr_leaf_entry
*entry
;
405 struct xfs_mount
*mp
= context
->dp
->i_mount
;
407 trace_xfs_attr_list_leaf(context
);
410 xfs_attr3_leaf_hdr_from_disk(mp
->m_attr_geo
, &ichdr
, leaf
);
411 entries
= xfs_attr3_leaf_entryp(leaf
);
413 cursor
= context
->cursor
;
417 * Re-find our place in the leaf block if this is a new syscall.
419 if (context
->resynch
) {
421 for (i
= 0; i
< ichdr
.count
; entry
++, i
++) {
422 if (be32_to_cpu(entry
->hashval
) == cursor
->hashval
) {
423 if (cursor
->offset
== context
->dupcnt
) {
428 } else if (be32_to_cpu(entry
->hashval
) >
434 if (i
== ichdr
.count
) {
435 trace_xfs_attr_list_notfound(context
);
442 context
->resynch
= 0;
445 * We have found our place, start copying out the new attributes.
447 for (; i
< ichdr
.count
; entry
++, i
++) {
449 int namelen
, valuelen
;
451 if (be32_to_cpu(entry
->hashval
) != cursor
->hashval
) {
452 cursor
->hashval
= be32_to_cpu(entry
->hashval
);
456 if ((entry
->flags
& XFS_ATTR_INCOMPLETE
) &&
457 !(context
->flags
& ATTR_INCOMPLETE
))
458 continue; /* skip incomplete entries */
460 if (entry
->flags
& XFS_ATTR_LOCAL
) {
461 xfs_attr_leaf_name_local_t
*name_loc
;
463 name_loc
= xfs_attr3_leaf_name_local(leaf
, i
);
464 name
= name_loc
->nameval
;
465 namelen
= name_loc
->namelen
;
466 valuelen
= be16_to_cpu(name_loc
->valuelen
);
468 xfs_attr_leaf_name_remote_t
*name_rmt
;
470 name_rmt
= xfs_attr3_leaf_name_remote(leaf
, i
);
471 name
= name_rmt
->name
;
472 namelen
= name_rmt
->namelen
;
473 valuelen
= be32_to_cpu(name_rmt
->valuelen
);
476 context
->put_listent(context
, entry
->flags
,
477 name
, namelen
, valuelen
);
478 if (context
->seen_enough
)
482 trace_xfs_attr_list_leaf_end(context
);
487 * Copy out attribute entries for attr_list(), for leaf attribute lists.
490 xfs_attr_leaf_list(xfs_attr_list_context_t
*context
)
495 trace_xfs_attr_leaf_list(context
);
497 context
->cursor
->blkno
= 0;
498 error
= xfs_attr3_leaf_read(context
->tp
, context
->dp
, 0, -1, &bp
);
502 xfs_attr3_leaf_list_int(bp
, context
);
503 xfs_trans_brelse(context
->tp
, bp
);
508 xfs_attr_list_int_ilocked(
509 struct xfs_attr_list_context
*context
)
511 struct xfs_inode
*dp
= context
->dp
;
513 ASSERT(xfs_isilocked(dp
, XFS_ILOCK_SHARED
| XFS_ILOCK_EXCL
));
516 * Decide on what work routines to call based on the inode size.
518 if (!xfs_inode_hasattr(dp
))
520 else if (dp
->i_d
.di_aformat
== XFS_DINODE_FMT_LOCAL
)
521 return xfs_attr_shortform_list(context
);
522 else if (xfs_bmap_one_block(dp
, XFS_ATTR_FORK
))
523 return xfs_attr_leaf_list(context
);
524 return xfs_attr_node_list(context
);
529 xfs_attr_list_context_t
*context
)
532 xfs_inode_t
*dp
= context
->dp
;
535 XFS_STATS_INC(dp
->i_mount
, xs_attr_list
);
537 if (XFS_FORCED_SHUTDOWN(dp
->i_mount
))
540 lock_mode
= xfs_ilock_attr_map_shared(dp
);
541 error
= xfs_attr_list_int_ilocked(context
);
542 xfs_iunlock(dp
, lock_mode
);
546 #define ATTR_ENTBASESIZE /* minimum bytes used by an attr */ \
547 (((struct attrlist_ent *) 0)->a_name - (char *) 0)
548 #define ATTR_ENTSIZE(namelen) /* actual bytes used by an attr */ \
549 ((ATTR_ENTBASESIZE + (namelen) + 1 + sizeof(uint32_t)-1) \
550 & ~(sizeof(uint32_t)-1))
553 * Format an attribute and copy it out to the user's buffer.
554 * Take care to check values and protect against them changing later,
555 * we may be reading them directly out of a user buffer.
558 xfs_attr_put_listent(
559 xfs_attr_list_context_t
*context
,
565 struct attrlist
*alist
= (struct attrlist
*)context
->alist
;
569 ASSERT(!(context
->flags
& ATTR_KERNOVAL
));
570 ASSERT(context
->count
>= 0);
571 ASSERT(context
->count
< (ATTR_MAX_VALUELEN
/8));
572 ASSERT(context
->firstu
>= sizeof(*alist
));
573 ASSERT(context
->firstu
<= context
->bufsize
);
576 * Only list entries in the right namespace.
578 if (((context
->flags
& ATTR_SECURE
) == 0) !=
579 ((flags
& XFS_ATTR_SECURE
) == 0))
581 if (((context
->flags
& ATTR_ROOT
) == 0) !=
582 ((flags
& XFS_ATTR_ROOT
) == 0))
585 arraytop
= sizeof(*alist
) +
586 context
->count
* sizeof(alist
->al_offset
[0]);
587 context
->firstu
-= ATTR_ENTSIZE(namelen
);
588 if (context
->firstu
< arraytop
) {
589 trace_xfs_attr_list_full(context
);
591 context
->seen_enough
= 1;
595 aep
= (attrlist_ent_t
*)&context
->alist
[context
->firstu
];
596 aep
->a_valuelen
= valuelen
;
597 memcpy(aep
->a_name
, name
, namelen
);
598 aep
->a_name
[namelen
] = 0;
599 alist
->al_offset
[context
->count
++] = context
->firstu
;
600 alist
->al_count
= context
->count
;
601 trace_xfs_attr_list_add(context
);
606 * Generate a list of extended attribute names and optionally
607 * also value lengths. Positive return value follows the XFS
608 * convention of being an error, zero or negative return code
609 * is the length of the buffer returned (negated), indicating
618 attrlist_cursor_kern_t
*cursor
)
620 xfs_attr_list_context_t context
;
621 struct attrlist
*alist
;
625 * Validate the cursor.
627 if (cursor
->pad1
|| cursor
->pad2
)
629 if ((cursor
->initted
== 0) &&
630 (cursor
->hashval
|| cursor
->blkno
|| cursor
->offset
))
633 /* Only internal consumers can retrieve incomplete attrs. */
634 if (flags
& ATTR_INCOMPLETE
)
638 * Check for a properly aligned buffer.
640 if (((long)buffer
) & (sizeof(int)-1))
642 if (flags
& ATTR_KERNOVAL
)
646 * Initialize the output buffer.
648 memset(&context
, 0, sizeof(context
));
650 context
.cursor
= cursor
;
652 context
.flags
= flags
;
653 context
.alist
= buffer
;
654 context
.bufsize
= (bufsize
& ~(sizeof(int)-1)); /* align */
655 context
.firstu
= context
.bufsize
;
656 context
.put_listent
= xfs_attr_put_listent
;
658 alist
= (struct attrlist
*)context
.alist
;
661 alist
->al_offset
[0] = context
.bufsize
;
663 error
= xfs_attr_list_int(&context
);