]> git.ipfire.org Git - thirdparty/xfsprogs-dev.git/blob - libxfs/xfs_ialloc_btree.c
xfs: fix transaction leak on remote attr set/remove failure
[thirdparty/xfsprogs-dev.git] / libxfs / xfs_ialloc_btree.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
4 * All Rights Reserved.
5 */
6 #include "libxfs_priv.h"
7 #include "xfs_fs.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_bit.h"
13 #include "xfs_mount.h"
14 #include "xfs_inode.h"
15 #include "xfs_btree.h"
16 #include "xfs_ialloc.h"
17 #include "xfs_ialloc_btree.h"
18 #include "xfs_alloc.h"
19 #include "xfs_trace.h"
20 #include "xfs_cksum.h"
21 #include "xfs_trans.h"
22 #include "xfs_rmap.h"
23
24
25 STATIC int
26 xfs_inobt_get_minrecs(
27 struct xfs_btree_cur *cur,
28 int level)
29 {
30 return cur->bc_mp->m_inobt_mnr[level != 0];
31 }
32
33 STATIC struct xfs_btree_cur *
34 xfs_inobt_dup_cursor(
35 struct xfs_btree_cur *cur)
36 {
37 return xfs_inobt_init_cursor(cur->bc_mp, cur->bc_tp,
38 cur->bc_private.a.agbp, cur->bc_private.a.agno,
39 cur->bc_btnum);
40 }
41
42 STATIC void
43 xfs_inobt_set_root(
44 struct xfs_btree_cur *cur,
45 union xfs_btree_ptr *nptr,
46 int inc) /* level change */
47 {
48 struct xfs_buf *agbp = cur->bc_private.a.agbp;
49 struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp);
50
51 agi->agi_root = nptr->s;
52 be32_add_cpu(&agi->agi_level, inc);
53 xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_ROOT | XFS_AGI_LEVEL);
54 }
55
56 STATIC void
57 xfs_finobt_set_root(
58 struct xfs_btree_cur *cur,
59 union xfs_btree_ptr *nptr,
60 int inc) /* level change */
61 {
62 struct xfs_buf *agbp = cur->bc_private.a.agbp;
63 struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp);
64
65 agi->agi_free_root = nptr->s;
66 be32_add_cpu(&agi->agi_free_level, inc);
67 xfs_ialloc_log_agi(cur->bc_tp, agbp,
68 XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL);
69 }
70
71 STATIC int
72 __xfs_inobt_alloc_block(
73 struct xfs_btree_cur *cur,
74 union xfs_btree_ptr *start,
75 union xfs_btree_ptr *new,
76 int *stat,
77 enum xfs_ag_resv_type resv)
78 {
79 xfs_alloc_arg_t args; /* block allocation args */
80 int error; /* error return value */
81 xfs_agblock_t sbno = be32_to_cpu(start->s);
82
83 memset(&args, 0, sizeof(args));
84 args.tp = cur->bc_tp;
85 args.mp = cur->bc_mp;
86 xfs_rmap_ag_owner(&args.oinfo, XFS_RMAP_OWN_INOBT);
87 args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno, sbno);
88 args.minlen = 1;
89 args.maxlen = 1;
90 args.prod = 1;
91 args.type = XFS_ALLOCTYPE_NEAR_BNO;
92 args.resv = resv;
93
94 error = xfs_alloc_vextent(&args);
95 if (error)
96 return error;
97
98 if (args.fsbno == NULLFSBLOCK) {
99 *stat = 0;
100 return 0;
101 }
102 ASSERT(args.len == 1);
103
104 new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
105 *stat = 1;
106 return 0;
107 }
108
109 STATIC int
110 xfs_inobt_alloc_block(
111 struct xfs_btree_cur *cur,
112 union xfs_btree_ptr *start,
113 union xfs_btree_ptr *new,
114 int *stat)
115 {
116 return __xfs_inobt_alloc_block(cur, start, new, stat, XFS_AG_RESV_NONE);
117 }
118
119 STATIC int
120 xfs_finobt_alloc_block(
121 struct xfs_btree_cur *cur,
122 union xfs_btree_ptr *start,
123 union xfs_btree_ptr *new,
124 int *stat)
125 {
126 if (cur->bc_mp->m_inotbt_nores)
127 return xfs_inobt_alloc_block(cur, start, new, stat);
128 return __xfs_inobt_alloc_block(cur, start, new, stat,
129 XFS_AG_RESV_METADATA);
130 }
131
132 STATIC int
133 __xfs_inobt_free_block(
134 struct xfs_btree_cur *cur,
135 struct xfs_buf *bp,
136 enum xfs_ag_resv_type resv)
137 {
138 struct xfs_owner_info oinfo;
139
140 xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INOBT);
141 return xfs_free_extent(cur->bc_tp,
142 XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp)), 1,
143 &oinfo, resv);
144 }
145
146 STATIC int
147 xfs_inobt_free_block(
148 struct xfs_btree_cur *cur,
149 struct xfs_buf *bp)
150 {
151 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_NONE);
152 }
153
154 STATIC int
155 xfs_finobt_free_block(
156 struct xfs_btree_cur *cur,
157 struct xfs_buf *bp)
158 {
159 if (cur->bc_mp->m_inotbt_nores)
160 return xfs_inobt_free_block(cur, bp);
161 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_METADATA);
162 }
163
164 STATIC int
165 xfs_inobt_get_maxrecs(
166 struct xfs_btree_cur *cur,
167 int level)
168 {
169 return cur->bc_mp->m_inobt_mxr[level != 0];
170 }
171
172 STATIC void
173 xfs_inobt_init_key_from_rec(
174 union xfs_btree_key *key,
175 union xfs_btree_rec *rec)
176 {
177 key->inobt.ir_startino = rec->inobt.ir_startino;
178 }
179
180 STATIC void
181 xfs_inobt_init_high_key_from_rec(
182 union xfs_btree_key *key,
183 union xfs_btree_rec *rec)
184 {
185 __u32 x;
186
187 x = be32_to_cpu(rec->inobt.ir_startino);
188 x += XFS_INODES_PER_CHUNK - 1;
189 key->inobt.ir_startino = cpu_to_be32(x);
190 }
191
192 STATIC void
193 xfs_inobt_init_rec_from_cur(
194 struct xfs_btree_cur *cur,
195 union xfs_btree_rec *rec)
196 {
197 rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
198 if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
199 rec->inobt.ir_u.sp.ir_holemask =
200 cpu_to_be16(cur->bc_rec.i.ir_holemask);
201 rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
202 rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
203 } else {
204 /* ir_holemask/ir_count not supported on-disk */
205 rec->inobt.ir_u.f.ir_freecount =
206 cpu_to_be32(cur->bc_rec.i.ir_freecount);
207 }
208 rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
209 }
210
211 /*
212 * initial value of ptr for lookup
213 */
214 STATIC void
215 xfs_inobt_init_ptr_from_cur(
216 struct xfs_btree_cur *cur,
217 union xfs_btree_ptr *ptr)
218 {
219 struct xfs_agi *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
220
221 ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
222
223 ptr->s = agi->agi_root;
224 }
225
226 STATIC void
227 xfs_finobt_init_ptr_from_cur(
228 struct xfs_btree_cur *cur,
229 union xfs_btree_ptr *ptr)
230 {
231 struct xfs_agi *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
232
233 ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
234 ptr->s = agi->agi_free_root;
235 }
236
237 STATIC int64_t
238 xfs_inobt_key_diff(
239 struct xfs_btree_cur *cur,
240 union xfs_btree_key *key)
241 {
242 return (int64_t)be32_to_cpu(key->inobt.ir_startino) -
243 cur->bc_rec.i.ir_startino;
244 }
245
246 STATIC int64_t
247 xfs_inobt_diff_two_keys(
248 struct xfs_btree_cur *cur,
249 union xfs_btree_key *k1,
250 union xfs_btree_key *k2)
251 {
252 return (int64_t)be32_to_cpu(k1->inobt.ir_startino) -
253 be32_to_cpu(k2->inobt.ir_startino);
254 }
255
256 static xfs_failaddr_t
257 xfs_inobt_verify(
258 struct xfs_buf *bp)
259 {
260 struct xfs_mount *mp = bp->b_target->bt_mount;
261 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
262 xfs_failaddr_t fa;
263 unsigned int level;
264
265 /*
266 * During growfs operations, we can't verify the exact owner as the
267 * perag is not fully initialised and hence not attached to the buffer.
268 *
269 * Similarly, during log recovery we will have a perag structure
270 * attached, but the agi information will not yet have been initialised
271 * from the on disk AGI. We don't currently use any of this information,
272 * but beware of the landmine (i.e. need to check pag->pagi_init) if we
273 * ever do.
274 */
275 switch (block->bb_magic) {
276 case cpu_to_be32(XFS_IBT_CRC_MAGIC):
277 case cpu_to_be32(XFS_FIBT_CRC_MAGIC):
278 fa = xfs_btree_sblock_v5hdr_verify(bp);
279 if (fa)
280 return fa;
281 /* fall through */
282 case cpu_to_be32(XFS_IBT_MAGIC):
283 case cpu_to_be32(XFS_FIBT_MAGIC):
284 break;
285 default:
286 return __this_address;
287 }
288
289 /* level verification */
290 level = be16_to_cpu(block->bb_level);
291 if (level >= mp->m_in_maxlevels)
292 return __this_address;
293
294 return xfs_btree_sblock_verify(bp, mp->m_inobt_mxr[level != 0]);
295 }
296
297 static void
298 xfs_inobt_read_verify(
299 struct xfs_buf *bp)
300 {
301 xfs_failaddr_t fa;
302
303 if (!xfs_btree_sblock_verify_crc(bp))
304 xfs_verifier_error(bp, -EFSBADCRC, __this_address);
305 else {
306 fa = xfs_inobt_verify(bp);
307 if (fa)
308 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
309 }
310
311 if (bp->b_error)
312 trace_xfs_btree_corrupt(bp, _RET_IP_);
313 }
314
315 static void
316 xfs_inobt_write_verify(
317 struct xfs_buf *bp)
318 {
319 xfs_failaddr_t fa;
320
321 fa = xfs_inobt_verify(bp);
322 if (fa) {
323 trace_xfs_btree_corrupt(bp, _RET_IP_);
324 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
325 return;
326 }
327 xfs_btree_sblock_calc_crc(bp);
328
329 }
330
331 const struct xfs_buf_ops xfs_inobt_buf_ops = {
332 .name = "xfs_inobt",
333 .verify_read = xfs_inobt_read_verify,
334 .verify_write = xfs_inobt_write_verify,
335 .verify_struct = xfs_inobt_verify,
336 };
337
338 STATIC int
339 xfs_inobt_keys_inorder(
340 struct xfs_btree_cur *cur,
341 union xfs_btree_key *k1,
342 union xfs_btree_key *k2)
343 {
344 return be32_to_cpu(k1->inobt.ir_startino) <
345 be32_to_cpu(k2->inobt.ir_startino);
346 }
347
348 STATIC int
349 xfs_inobt_recs_inorder(
350 struct xfs_btree_cur *cur,
351 union xfs_btree_rec *r1,
352 union xfs_btree_rec *r2)
353 {
354 return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
355 be32_to_cpu(r2->inobt.ir_startino);
356 }
357
358 static const struct xfs_btree_ops xfs_inobt_ops = {
359 .rec_len = sizeof(xfs_inobt_rec_t),
360 .key_len = sizeof(xfs_inobt_key_t),
361
362 .dup_cursor = xfs_inobt_dup_cursor,
363 .set_root = xfs_inobt_set_root,
364 .alloc_block = xfs_inobt_alloc_block,
365 .free_block = xfs_inobt_free_block,
366 .get_minrecs = xfs_inobt_get_minrecs,
367 .get_maxrecs = xfs_inobt_get_maxrecs,
368 .init_key_from_rec = xfs_inobt_init_key_from_rec,
369 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
370 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
371 .init_ptr_from_cur = xfs_inobt_init_ptr_from_cur,
372 .key_diff = xfs_inobt_key_diff,
373 .buf_ops = &xfs_inobt_buf_ops,
374 .diff_two_keys = xfs_inobt_diff_two_keys,
375 .keys_inorder = xfs_inobt_keys_inorder,
376 .recs_inorder = xfs_inobt_recs_inorder,
377 };
378
379 static const struct xfs_btree_ops xfs_finobt_ops = {
380 .rec_len = sizeof(xfs_inobt_rec_t),
381 .key_len = sizeof(xfs_inobt_key_t),
382
383 .dup_cursor = xfs_inobt_dup_cursor,
384 .set_root = xfs_finobt_set_root,
385 .alloc_block = xfs_finobt_alloc_block,
386 .free_block = xfs_finobt_free_block,
387 .get_minrecs = xfs_inobt_get_minrecs,
388 .get_maxrecs = xfs_inobt_get_maxrecs,
389 .init_key_from_rec = xfs_inobt_init_key_from_rec,
390 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
391 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
392 .init_ptr_from_cur = xfs_finobt_init_ptr_from_cur,
393 .key_diff = xfs_inobt_key_diff,
394 .buf_ops = &xfs_inobt_buf_ops,
395 .diff_two_keys = xfs_inobt_diff_two_keys,
396 .keys_inorder = xfs_inobt_keys_inorder,
397 .recs_inorder = xfs_inobt_recs_inorder,
398 };
399
400 /*
401 * Allocate a new inode btree cursor.
402 */
403 struct xfs_btree_cur * /* new inode btree cursor */
404 xfs_inobt_init_cursor(
405 struct xfs_mount *mp, /* file system mount point */
406 struct xfs_trans *tp, /* transaction pointer */
407 struct xfs_buf *agbp, /* buffer for agi structure */
408 xfs_agnumber_t agno, /* allocation group number */
409 xfs_btnum_t btnum) /* ialloc or free ino btree */
410 {
411 struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp);
412 struct xfs_btree_cur *cur;
413
414 cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
415
416 cur->bc_tp = tp;
417 cur->bc_mp = mp;
418 cur->bc_btnum = btnum;
419 if (btnum == XFS_BTNUM_INO) {
420 cur->bc_nlevels = be32_to_cpu(agi->agi_level);
421 cur->bc_ops = &xfs_inobt_ops;
422 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_ibt_2);
423 } else {
424 cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
425 cur->bc_ops = &xfs_finobt_ops;
426 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_fibt_2);
427 }
428
429 cur->bc_blocklog = mp->m_sb.sb_blocklog;
430
431 if (xfs_sb_version_hascrc(&mp->m_sb))
432 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
433
434 cur->bc_private.a.agbp = agbp;
435 cur->bc_private.a.agno = agno;
436
437 return cur;
438 }
439
440 /*
441 * Calculate number of records in an inobt btree block.
442 */
443 int
444 xfs_inobt_maxrecs(
445 struct xfs_mount *mp,
446 int blocklen,
447 int leaf)
448 {
449 blocklen -= XFS_INOBT_BLOCK_LEN(mp);
450
451 if (leaf)
452 return blocklen / sizeof(xfs_inobt_rec_t);
453 return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
454 }
455
456 /*
457 * Convert the inode record holemask to an inode allocation bitmap. The inode
458 * allocation bitmap is inode granularity and specifies whether an inode is
459 * physically allocated on disk (not whether the inode is considered allocated
460 * or free by the fs).
461 *
462 * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
463 */
464 uint64_t
465 xfs_inobt_irec_to_allocmask(
466 struct xfs_inobt_rec_incore *rec)
467 {
468 uint64_t bitmap = 0;
469 uint64_t inodespbit;
470 int nextbit;
471 uint allocbitmap;
472
473 /*
474 * The holemask has 16-bits for a 64 inode record. Therefore each
475 * holemask bit represents multiple inodes. Create a mask of bits to set
476 * in the allocmask for each holemask bit.
477 */
478 inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
479
480 /*
481 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
482 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
483 * anything beyond the 16 holemask bits since this casts to a larger
484 * type.
485 */
486 allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
487
488 /*
489 * allocbitmap is the inverted holemask so every set bit represents
490 * allocated inodes. To expand from 16-bit holemask granularity to
491 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
492 * bitmap for every holemask bit.
493 */
494 nextbit = xfs_next_bit(&allocbitmap, 1, 0);
495 while (nextbit != -1) {
496 ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
497
498 bitmap |= (inodespbit <<
499 (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
500
501 nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
502 }
503
504 return bitmap;
505 }
506
507 #if defined(DEBUG) || defined(XFS_WARN)
508 /*
509 * Verify that an in-core inode record has a valid inode count.
510 */
511 int
512 xfs_inobt_rec_check_count(
513 struct xfs_mount *mp,
514 struct xfs_inobt_rec_incore *rec)
515 {
516 int inocount = 0;
517 int nextbit = 0;
518 uint64_t allocbmap;
519 int wordsz;
520
521 wordsz = sizeof(allocbmap) / sizeof(unsigned int);
522 allocbmap = xfs_inobt_irec_to_allocmask(rec);
523
524 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
525 while (nextbit != -1) {
526 inocount++;
527 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
528 nextbit + 1);
529 }
530
531 if (inocount != rec->ir_count)
532 return -EFSCORRUPTED;
533
534 return 0;
535 }
536 #endif /* DEBUG */
537
538 static xfs_extlen_t
539 xfs_inobt_max_size(
540 struct xfs_mount *mp)
541 {
542 /* Bail out if we're uninitialized, which can happen in mkfs. */
543 if (mp->m_inobt_mxr[0] == 0)
544 return 0;
545
546 return xfs_btree_calc_size(mp->m_inobt_mnr,
547 (uint64_t)mp->m_sb.sb_agblocks * mp->m_sb.sb_inopblock /
548 XFS_INODES_PER_CHUNK);
549 }
550
551 static int
552 xfs_inobt_count_blocks(
553 struct xfs_mount *mp,
554 xfs_agnumber_t agno,
555 xfs_btnum_t btnum,
556 xfs_extlen_t *tree_blocks)
557 {
558 struct xfs_buf *agbp;
559 struct xfs_btree_cur *cur;
560 int error;
561
562 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
563 if (error)
564 return error;
565
566 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, btnum);
567 error = xfs_btree_count_blocks(cur, tree_blocks);
568 xfs_btree_del_cursor(cur, error);
569 xfs_buf_relse(agbp);
570
571 return error;
572 }
573
574 /*
575 * Figure out how many blocks to reserve and how many are used by this btree.
576 */
577 int
578 xfs_finobt_calc_reserves(
579 struct xfs_mount *mp,
580 xfs_agnumber_t agno,
581 xfs_extlen_t *ask,
582 xfs_extlen_t *used)
583 {
584 xfs_extlen_t tree_len = 0;
585 int error;
586
587 if (!xfs_sb_version_hasfinobt(&mp->m_sb))
588 return 0;
589
590 error = xfs_inobt_count_blocks(mp, agno, XFS_BTNUM_FINO, &tree_len);
591 if (error)
592 return error;
593
594 *ask += xfs_inobt_max_size(mp);
595 *used += tree_len;
596 return 0;
597 }
598
599 /* Calculate the inobt btree size for some records. */
600 xfs_extlen_t
601 xfs_iallocbt_calc_size(
602 struct xfs_mount *mp,
603 unsigned long long len)
604 {
605 return xfs_btree_calc_size(mp->m_inobt_mnr, len);
606 }