xfs: decouple inode and bmap btree header files
[linux-2.6-block.git] / fs / xfs / xfs_attr_list.c
CommitLineData
abec5f2b
DC
1/*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * Copyright (c) 2013 Red Hat, Inc.
4 * All Rights Reserved.
5 *
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.
9 *
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.
14 *
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
18 */
19#include "xfs.h"
20#include "xfs_fs.h"
a4fbe6ab 21#include "xfs_format.h"
239880ef
DC
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
abec5f2b 24#include "xfs_bit.h"
abec5f2b
DC
25#include "xfs_sb.h"
26#include "xfs_ag.h"
27#include "xfs_mount.h"
57062787 28#include "xfs_da_format.h"
abec5f2b 29#include "xfs_da_btree.h"
abec5f2b 30#include "xfs_inode.h"
239880ef 31#include "xfs_trans.h"
abec5f2b
DC
32#include "xfs_inode_item.h"
33#include "xfs_bmap.h"
34#include "xfs_attr.h"
a4fbe6ab
DC
35#include "xfs_attr_sf.h"
36#include "xfs_attr_remote.h"
abec5f2b
DC
37#include "xfs_attr_leaf.h"
38#include "xfs_error.h"
39#include "xfs_trace.h"
40#include "xfs_buf_item.h"
41#include "xfs_cksum.h"
a4fbe6ab 42#include "xfs_dinode.h"
abec5f2b
DC
43
44STATIC int
45xfs_attr_shortform_compare(const void *a, const void *b)
46{
47 xfs_attr_sf_sort_t *sa, *sb;
48
49 sa = (xfs_attr_sf_sort_t *)a;
50 sb = (xfs_attr_sf_sort_t *)b;
51 if (sa->hash < sb->hash) {
52 return(-1);
53 } else if (sa->hash > sb->hash) {
54 return(1);
55 } else {
56 return(sa->entno - sb->entno);
57 }
58}
59
60#define XFS_ISRESET_CURSOR(cursor) \
61 (!((cursor)->initted) && !((cursor)->hashval) && \
62 !((cursor)->blkno) && !((cursor)->offset))
63/*
64 * Copy out entries of shortform attribute lists for attr_list().
65 * Shortform attribute lists are not stored in hashval sorted order.
66 * If the output buffer is not large enough to hold them all, then we
67 * we have to calculate each entries' hashvalue and sort them before
68 * we can begin returning them to the user.
69 */
70int
71xfs_attr_shortform_list(xfs_attr_list_context_t *context)
72{
73 attrlist_cursor_kern_t *cursor;
74 xfs_attr_sf_sort_t *sbuf, *sbp;
75 xfs_attr_shortform_t *sf;
76 xfs_attr_sf_entry_t *sfe;
77 xfs_inode_t *dp;
78 int sbsize, nsbuf, count, i;
79 int error;
80
81 ASSERT(context != NULL);
82 dp = context->dp;
83 ASSERT(dp != NULL);
84 ASSERT(dp->i_afp != NULL);
85 sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data;
86 ASSERT(sf != NULL);
87 if (!sf->hdr.count)
88 return(0);
89 cursor = context->cursor;
90 ASSERT(cursor != NULL);
91
92 trace_xfs_attr_list_sf(context);
93
94 /*
95 * If the buffer is large enough and the cursor is at the start,
96 * do not bother with sorting since we will return everything in
97 * one buffer and another call using the cursor won't need to be
98 * made.
99 * Note the generous fudge factor of 16 overhead bytes per entry.
100 * If bufsize is zero then put_listent must be a search function
101 * and can just scan through what we have.
102 */
103 if (context->bufsize == 0 ||
104 (XFS_ISRESET_CURSOR(cursor) &&
105 (dp->i_afp->if_bytes + sf->hdr.count * 16) < context->bufsize)) {
106 for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) {
107 error = context->put_listent(context,
108 sfe->flags,
109 sfe->nameval,
110 (int)sfe->namelen,
111 (int)sfe->valuelen,
112 &sfe->nameval[sfe->namelen]);
113
114 /*
115 * Either search callback finished early or
116 * didn't fit it all in the buffer after all.
117 */
118 if (context->seen_enough)
119 break;
120
121 if (error)
122 return error;
123 sfe = XFS_ATTR_SF_NEXTENTRY(sfe);
124 }
125 trace_xfs_attr_list_sf_all(context);
126 return(0);
127 }
128
129 /* do no more for a search callback */
130 if (context->bufsize == 0)
131 return 0;
132
133 /*
134 * It didn't all fit, so we have to sort everything on hashval.
135 */
136 sbsize = sf->hdr.count * sizeof(*sbuf);
137 sbp = sbuf = kmem_alloc(sbsize, KM_SLEEP | KM_NOFS);
138
139 /*
140 * Scan the attribute list for the rest of the entries, storing
141 * the relevant info from only those that match into a buffer.
142 */
143 nsbuf = 0;
144 for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) {
145 if (unlikely(
146 ((char *)sfe < (char *)sf) ||
147 ((char *)sfe >= ((char *)sf + dp->i_afp->if_bytes)))) {
148 XFS_CORRUPTION_ERROR("xfs_attr_shortform_list",
149 XFS_ERRLEVEL_LOW,
150 context->dp->i_mount, sfe);
151 kmem_free(sbuf);
152 return XFS_ERROR(EFSCORRUPTED);
153 }
154
155 sbp->entno = i;
156 sbp->hash = xfs_da_hashname(sfe->nameval, sfe->namelen);
157 sbp->name = sfe->nameval;
158 sbp->namelen = sfe->namelen;
159 /* These are bytes, and both on-disk, don't endian-flip */
160 sbp->valuelen = sfe->valuelen;
161 sbp->flags = sfe->flags;
162 sfe = XFS_ATTR_SF_NEXTENTRY(sfe);
163 sbp++;
164 nsbuf++;
165 }
166
167 /*
168 * Sort the entries on hash then entno.
169 */
170 xfs_sort(sbuf, nsbuf, sizeof(*sbuf), xfs_attr_shortform_compare);
171
172 /*
173 * Re-find our place IN THE SORTED LIST.
174 */
175 count = 0;
176 cursor->initted = 1;
177 cursor->blkno = 0;
178 for (sbp = sbuf, i = 0; i < nsbuf; i++, sbp++) {
179 if (sbp->hash == cursor->hashval) {
180 if (cursor->offset == count) {
181 break;
182 }
183 count++;
184 } else if (sbp->hash > cursor->hashval) {
185 break;
186 }
187 }
188 if (i == nsbuf) {
189 kmem_free(sbuf);
190 return(0);
191 }
192
193 /*
194 * Loop putting entries into the user buffer.
195 */
196 for ( ; i < nsbuf; i++, sbp++) {
197 if (cursor->hashval != sbp->hash) {
198 cursor->hashval = sbp->hash;
199 cursor->offset = 0;
200 }
201 error = context->put_listent(context,
202 sbp->flags,
203 sbp->name,
204 sbp->namelen,
205 sbp->valuelen,
206 &sbp->name[sbp->namelen]);
207 if (error)
208 return error;
209 if (context->seen_enough)
210 break;
211 cursor->offset++;
212 }
213
214 kmem_free(sbuf);
215 return(0);
216}
217
218STATIC int
219xfs_attr_node_list(xfs_attr_list_context_t *context)
220{
221 attrlist_cursor_kern_t *cursor;
222 xfs_attr_leafblock_t *leaf;
223 xfs_da_intnode_t *node;
224 struct xfs_attr3_icleaf_hdr leafhdr;
225 struct xfs_da3_icnode_hdr nodehdr;
226 struct xfs_da_node_entry *btree;
227 int error, i;
228 struct xfs_buf *bp;
229
230 trace_xfs_attr_node_list(context);
231
232 cursor = context->cursor;
233 cursor->initted = 1;
234
235 /*
236 * Do all sorts of validation on the passed-in cursor structure.
237 * If anything is amiss, ignore the cursor and look up the hashval
238 * starting from the btree root.
239 */
240 bp = NULL;
241 if (cursor->blkno > 0) {
242 error = xfs_da3_node_read(NULL, context->dp, cursor->blkno, -1,
243 &bp, XFS_ATTR_FORK);
244 if ((error != 0) && (error != EFSCORRUPTED))
245 return(error);
246 if (bp) {
247 struct xfs_attr_leaf_entry *entries;
248
249 node = bp->b_addr;
250 switch (be16_to_cpu(node->hdr.info.magic)) {
251 case XFS_DA_NODE_MAGIC:
252 case XFS_DA3_NODE_MAGIC:
253 trace_xfs_attr_list_wrong_blk(context);
254 xfs_trans_brelse(NULL, bp);
255 bp = NULL;
256 break;
257 case XFS_ATTR_LEAF_MAGIC:
258 case XFS_ATTR3_LEAF_MAGIC:
259 leaf = bp->b_addr;
260 xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf);
261 entries = xfs_attr3_leaf_entryp(leaf);
262 if (cursor->hashval > be32_to_cpu(
263 entries[leafhdr.count - 1].hashval)) {
264 trace_xfs_attr_list_wrong_blk(context);
265 xfs_trans_brelse(NULL, bp);
266 bp = NULL;
267 } else if (cursor->hashval <= be32_to_cpu(
268 entries[0].hashval)) {
269 trace_xfs_attr_list_wrong_blk(context);
270 xfs_trans_brelse(NULL, bp);
271 bp = NULL;
272 }
273 break;
274 default:
275 trace_xfs_attr_list_wrong_blk(context);
276 xfs_trans_brelse(NULL, bp);
277 bp = NULL;
278 }
279 }
280 }
281
282 /*
283 * We did not find what we expected given the cursor's contents,
284 * so we start from the top and work down based on the hash value.
285 * Note that start of node block is same as start of leaf block.
286 */
287 if (bp == NULL) {
288 cursor->blkno = 0;
289 for (;;) {
290 __uint16_t magic;
291
292 error = xfs_da3_node_read(NULL, context->dp,
293 cursor->blkno, -1, &bp,
294 XFS_ATTR_FORK);
295 if (error)
296 return(error);
297 node = bp->b_addr;
298 magic = be16_to_cpu(node->hdr.info.magic);
299 if (magic == XFS_ATTR_LEAF_MAGIC ||
300 magic == XFS_ATTR3_LEAF_MAGIC)
301 break;
302 if (magic != XFS_DA_NODE_MAGIC &&
303 magic != XFS_DA3_NODE_MAGIC) {
304 XFS_CORRUPTION_ERROR("xfs_attr_node_list(3)",
305 XFS_ERRLEVEL_LOW,
306 context->dp->i_mount,
307 node);
308 xfs_trans_brelse(NULL, bp);
309 return XFS_ERROR(EFSCORRUPTED);
310 }
311
312 xfs_da3_node_hdr_from_disk(&nodehdr, node);
313 btree = xfs_da3_node_tree_p(node);
314 for (i = 0; i < nodehdr.count; btree++, i++) {
315 if (cursor->hashval
316 <= be32_to_cpu(btree->hashval)) {
317 cursor->blkno = be32_to_cpu(btree->before);
318 trace_xfs_attr_list_node_descend(context,
319 btree);
320 break;
321 }
322 }
323 if (i == nodehdr.count) {
324 xfs_trans_brelse(NULL, bp);
325 return 0;
326 }
327 xfs_trans_brelse(NULL, bp);
328 }
329 }
330 ASSERT(bp != NULL);
331
332 /*
333 * Roll upward through the blocks, processing each leaf block in
334 * order. As long as there is space in the result buffer, keep
335 * adding the information.
336 */
337 for (;;) {
338 leaf = bp->b_addr;
339 error = xfs_attr3_leaf_list_int(bp, context);
340 if (error) {
341 xfs_trans_brelse(NULL, bp);
342 return error;
343 }
344 xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf);
345 if (context->seen_enough || leafhdr.forw == 0)
346 break;
347 cursor->blkno = leafhdr.forw;
348 xfs_trans_brelse(NULL, bp);
349 error = xfs_attr3_leaf_read(NULL, context->dp, cursor->blkno, -1,
350 &bp);
351 if (error)
352 return error;
353 }
354 xfs_trans_brelse(NULL, bp);
355 return 0;
356}
357
358/*
359 * Copy out attribute list entries for attr_list(), for leaf attribute lists.
360 */
361int
362xfs_attr3_leaf_list_int(
363 struct xfs_buf *bp,
364 struct xfs_attr_list_context *context)
365{
366 struct attrlist_cursor_kern *cursor;
367 struct xfs_attr_leafblock *leaf;
368 struct xfs_attr3_icleaf_hdr ichdr;
369 struct xfs_attr_leaf_entry *entries;
370 struct xfs_attr_leaf_entry *entry;
371 int retval;
372 int i;
373
374 trace_xfs_attr_list_leaf(context);
375
376 leaf = bp->b_addr;
377 xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf);
378 entries = xfs_attr3_leaf_entryp(leaf);
379
380 cursor = context->cursor;
381 cursor->initted = 1;
382
383 /*
384 * Re-find our place in the leaf block if this is a new syscall.
385 */
386 if (context->resynch) {
387 entry = &entries[0];
388 for (i = 0; i < ichdr.count; entry++, i++) {
389 if (be32_to_cpu(entry->hashval) == cursor->hashval) {
390 if (cursor->offset == context->dupcnt) {
391 context->dupcnt = 0;
392 break;
393 }
394 context->dupcnt++;
395 } else if (be32_to_cpu(entry->hashval) >
396 cursor->hashval) {
397 context->dupcnt = 0;
398 break;
399 }
400 }
401 if (i == ichdr.count) {
402 trace_xfs_attr_list_notfound(context);
403 return 0;
404 }
405 } else {
406 entry = &entries[0];
407 i = 0;
408 }
409 context->resynch = 0;
410
411 /*
412 * We have found our place, start copying out the new attributes.
413 */
414 retval = 0;
415 for (; i < ichdr.count; entry++, i++) {
416 if (be32_to_cpu(entry->hashval) != cursor->hashval) {
417 cursor->hashval = be32_to_cpu(entry->hashval);
418 cursor->offset = 0;
419 }
420
421 if (entry->flags & XFS_ATTR_INCOMPLETE)
422 continue; /* skip incomplete entries */
423
424 if (entry->flags & XFS_ATTR_LOCAL) {
425 xfs_attr_leaf_name_local_t *name_loc =
426 xfs_attr3_leaf_name_local(leaf, i);
427
428 retval = context->put_listent(context,
429 entry->flags,
430 name_loc->nameval,
431 (int)name_loc->namelen,
432 be16_to_cpu(name_loc->valuelen),
433 &name_loc->nameval[name_loc->namelen]);
434 if (retval)
435 return retval;
436 } else {
437 xfs_attr_leaf_name_remote_t *name_rmt =
438 xfs_attr3_leaf_name_remote(leaf, i);
439
440 int valuelen = be32_to_cpu(name_rmt->valuelen);
441
442 if (context->put_value) {
443 xfs_da_args_t args;
444
445 memset((char *)&args, 0, sizeof(args));
446 args.dp = context->dp;
447 args.whichfork = XFS_ATTR_FORK;
448 args.valuelen = valuelen;
449 args.value = kmem_alloc(valuelen, KM_SLEEP | KM_NOFS);
450 args.rmtblkno = be32_to_cpu(name_rmt->valueblk);
451 args.rmtblkcnt = xfs_attr3_rmt_blocks(
452 args.dp->i_mount, valuelen);
453 retval = xfs_attr_rmtval_get(&args);
454 if (retval)
455 return retval;
456 retval = context->put_listent(context,
457 entry->flags,
458 name_rmt->name,
459 (int)name_rmt->namelen,
460 valuelen,
461 args.value);
462 kmem_free(args.value);
463 } else {
464 retval = context->put_listent(context,
465 entry->flags,
466 name_rmt->name,
467 (int)name_rmt->namelen,
468 valuelen,
469 NULL);
470 }
471 if (retval)
472 return retval;
473 }
474 if (context->seen_enough)
475 break;
476 cursor->offset++;
477 }
478 trace_xfs_attr_list_leaf_end(context);
479 return retval;
480}
481
482/*
483 * Copy out attribute entries for attr_list(), for leaf attribute lists.
484 */
485STATIC int
486xfs_attr_leaf_list(xfs_attr_list_context_t *context)
487{
488 int error;
489 struct xfs_buf *bp;
490
491 trace_xfs_attr_leaf_list(context);
492
493 context->cursor->blkno = 0;
494 error = xfs_attr3_leaf_read(NULL, context->dp, 0, -1, &bp);
495 if (error)
496 return XFS_ERROR(error);
497
498 error = xfs_attr3_leaf_list_int(bp, context);
499 xfs_trans_brelse(NULL, bp);
500 return XFS_ERROR(error);
501}
502
503int
504xfs_attr_list_int(
505 xfs_attr_list_context_t *context)
506{
507 int error;
508 xfs_inode_t *dp = context->dp;
509
510 XFS_STATS_INC(xs_attr_list);
511
512 if (XFS_FORCED_SHUTDOWN(dp->i_mount))
513 return EIO;
514
515 xfs_ilock(dp, XFS_ILOCK_SHARED);
516
517 /*
518 * Decide on what work routines to call based on the inode size.
519 */
520 if (!xfs_inode_hasattr(dp)) {
521 error = 0;
522 } else if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
523 error = xfs_attr_shortform_list(context);
524 } else if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
525 error = xfs_attr_leaf_list(context);
526 } else {
527 error = xfs_attr_node_list(context);
528 }
529
530 xfs_iunlock(dp, XFS_ILOCK_SHARED);
531
532 return error;
533}
534
535#define ATTR_ENTBASESIZE /* minimum bytes used by an attr */ \
536 (((struct attrlist_ent *) 0)->a_name - (char *) 0)
537#define ATTR_ENTSIZE(namelen) /* actual bytes used by an attr */ \
538 ((ATTR_ENTBASESIZE + (namelen) + 1 + sizeof(u_int32_t)-1) \
539 & ~(sizeof(u_int32_t)-1))
540
541/*
542 * Format an attribute and copy it out to the user's buffer.
543 * Take care to check values and protect against them changing later,
544 * we may be reading them directly out of a user buffer.
545 */
546STATIC int
547xfs_attr_put_listent(
548 xfs_attr_list_context_t *context,
549 int flags,
550 unsigned char *name,
551 int namelen,
552 int valuelen,
553 unsigned char *value)
554{
555 struct attrlist *alist = (struct attrlist *)context->alist;
556 attrlist_ent_t *aep;
557 int arraytop;
558
559 ASSERT(!(context->flags & ATTR_KERNOVAL));
560 ASSERT(context->count >= 0);
561 ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
562 ASSERT(context->firstu >= sizeof(*alist));
563 ASSERT(context->firstu <= context->bufsize);
564
565 /*
566 * Only list entries in the right namespace.
567 */
568 if (((context->flags & ATTR_SECURE) == 0) !=
569 ((flags & XFS_ATTR_SECURE) == 0))
570 return 0;
571 if (((context->flags & ATTR_ROOT) == 0) !=
572 ((flags & XFS_ATTR_ROOT) == 0))
573 return 0;
574
575 arraytop = sizeof(*alist) +
576 context->count * sizeof(alist->al_offset[0]);
577 context->firstu -= ATTR_ENTSIZE(namelen);
578 if (context->firstu < arraytop) {
579 trace_xfs_attr_list_full(context);
580 alist->al_more = 1;
581 context->seen_enough = 1;
582 return 1;
583 }
584
585 aep = (attrlist_ent_t *)&context->alist[context->firstu];
586 aep->a_valuelen = valuelen;
587 memcpy(aep->a_name, name, namelen);
588 aep->a_name[namelen] = 0;
589 alist->al_offset[context->count++] = context->firstu;
590 alist->al_count = context->count;
591 trace_xfs_attr_list_add(context);
592 return 0;
593}
594
595/*
596 * Generate a list of extended attribute names and optionally
597 * also value lengths. Positive return value follows the XFS
598 * convention of being an error, zero or negative return code
599 * is the length of the buffer returned (negated), indicating
600 * success.
601 */
602int
603xfs_attr_list(
604 xfs_inode_t *dp,
605 char *buffer,
606 int bufsize,
607 int flags,
608 attrlist_cursor_kern_t *cursor)
609{
610 xfs_attr_list_context_t context;
611 struct attrlist *alist;
612 int error;
613
614 /*
615 * Validate the cursor.
616 */
617 if (cursor->pad1 || cursor->pad2)
618 return(XFS_ERROR(EINVAL));
619 if ((cursor->initted == 0) &&
620 (cursor->hashval || cursor->blkno || cursor->offset))
621 return XFS_ERROR(EINVAL);
622
623 /*
624 * Check for a properly aligned buffer.
625 */
626 if (((long)buffer) & (sizeof(int)-1))
627 return XFS_ERROR(EFAULT);
628 if (flags & ATTR_KERNOVAL)
629 bufsize = 0;
630
631 /*
632 * Initialize the output buffer.
633 */
634 memset(&context, 0, sizeof(context));
635 context.dp = dp;
636 context.cursor = cursor;
637 context.resynch = 1;
638 context.flags = flags;
639 context.alist = buffer;
640 context.bufsize = (bufsize & ~(sizeof(int)-1)); /* align */
641 context.firstu = context.bufsize;
642 context.put_listent = xfs_attr_put_listent;
643
644 alist = (struct attrlist *)context.alist;
645 alist->al_count = 0;
646 alist->al_more = 0;
647 alist->al_offset[0] = context.bufsize;
648
649 error = xfs_attr_list_int(&context);
650 ASSERT(error >= 0);
651 return error;
652}