]> git.ipfire.org Git - thirdparty/xfsprogs-dev.git/blob - libxfs/xfs_ialloc_btree.c
xfsprogs: Release v6.7.0
[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 args.oinfo = XFS_RMAP_OINFO_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_finobt_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 return xfs_free_extent(cur->bc_tp,
139 XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp)), 1,
140 &XFS_RMAP_OINFO_INOBT, resv);
141 }
142
143 STATIC int
144 xfs_inobt_free_block(
145 struct xfs_btree_cur *cur,
146 struct xfs_buf *bp)
147 {
148 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_NONE);
149 }
150
151 STATIC int
152 xfs_finobt_free_block(
153 struct xfs_btree_cur *cur,
154 struct xfs_buf *bp)
155 {
156 if (cur->bc_mp->m_finobt_nores)
157 return xfs_inobt_free_block(cur, bp);
158 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_METADATA);
159 }
160
161 STATIC int
162 xfs_inobt_get_maxrecs(
163 struct xfs_btree_cur *cur,
164 int level)
165 {
166 return cur->bc_mp->m_inobt_mxr[level != 0];
167 }
168
169 STATIC void
170 xfs_inobt_init_key_from_rec(
171 union xfs_btree_key *key,
172 union xfs_btree_rec *rec)
173 {
174 key->inobt.ir_startino = rec->inobt.ir_startino;
175 }
176
177 STATIC void
178 xfs_inobt_init_high_key_from_rec(
179 union xfs_btree_key *key,
180 union xfs_btree_rec *rec)
181 {
182 __u32 x;
183
184 x = be32_to_cpu(rec->inobt.ir_startino);
185 x += XFS_INODES_PER_CHUNK - 1;
186 key->inobt.ir_startino = cpu_to_be32(x);
187 }
188
189 STATIC void
190 xfs_inobt_init_rec_from_cur(
191 struct xfs_btree_cur *cur,
192 union xfs_btree_rec *rec)
193 {
194 rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
195 if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
196 rec->inobt.ir_u.sp.ir_holemask =
197 cpu_to_be16(cur->bc_rec.i.ir_holemask);
198 rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
199 rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
200 } else {
201 /* ir_holemask/ir_count not supported on-disk */
202 rec->inobt.ir_u.f.ir_freecount =
203 cpu_to_be32(cur->bc_rec.i.ir_freecount);
204 }
205 rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
206 }
207
208 /*
209 * initial value of ptr for lookup
210 */
211 STATIC void
212 xfs_inobt_init_ptr_from_cur(
213 struct xfs_btree_cur *cur,
214 union xfs_btree_ptr *ptr)
215 {
216 struct xfs_agi *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
217
218 ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
219
220 ptr->s = agi->agi_root;
221 }
222
223 STATIC void
224 xfs_finobt_init_ptr_from_cur(
225 struct xfs_btree_cur *cur,
226 union xfs_btree_ptr *ptr)
227 {
228 struct xfs_agi *agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
229
230 ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
231 ptr->s = agi->agi_free_root;
232 }
233
234 STATIC int64_t
235 xfs_inobt_key_diff(
236 struct xfs_btree_cur *cur,
237 union xfs_btree_key *key)
238 {
239 return (int64_t)be32_to_cpu(key->inobt.ir_startino) -
240 cur->bc_rec.i.ir_startino;
241 }
242
243 STATIC int64_t
244 xfs_inobt_diff_two_keys(
245 struct xfs_btree_cur *cur,
246 union xfs_btree_key *k1,
247 union xfs_btree_key *k2)
248 {
249 return (int64_t)be32_to_cpu(k1->inobt.ir_startino) -
250 be32_to_cpu(k2->inobt.ir_startino);
251 }
252
253 static xfs_failaddr_t
254 xfs_inobt_verify(
255 struct xfs_buf *bp)
256 {
257 struct xfs_mount *mp = bp->b_target->bt_mount;
258 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
259 xfs_failaddr_t fa;
260 unsigned int level;
261
262 if (!xfs_verify_magic(bp, block->bb_magic))
263 return __this_address;
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 if (xfs_sb_version_hascrc(&mp->m_sb)) {
276 fa = xfs_btree_sblock_v5hdr_verify(bp);
277 if (fa)
278 return fa;
279 }
280
281 /* level verification */
282 level = be16_to_cpu(block->bb_level);
283 if (level >= mp->m_in_maxlevels)
284 return __this_address;
285
286 return xfs_btree_sblock_verify(bp, mp->m_inobt_mxr[level != 0]);
287 }
288
289 static void
290 xfs_inobt_read_verify(
291 struct xfs_buf *bp)
292 {
293 xfs_failaddr_t fa;
294
295 if (!xfs_btree_sblock_verify_crc(bp))
296 xfs_verifier_error(bp, -EFSBADCRC, __this_address);
297 else {
298 fa = xfs_inobt_verify(bp);
299 if (fa)
300 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
301 }
302
303 if (bp->b_error)
304 trace_xfs_btree_corrupt(bp, _RET_IP_);
305 }
306
307 static void
308 xfs_inobt_write_verify(
309 struct xfs_buf *bp)
310 {
311 xfs_failaddr_t fa;
312
313 fa = xfs_inobt_verify(bp);
314 if (fa) {
315 trace_xfs_btree_corrupt(bp, _RET_IP_);
316 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
317 return;
318 }
319 xfs_btree_sblock_calc_crc(bp);
320
321 }
322
323 const struct xfs_buf_ops xfs_inobt_buf_ops = {
324 .name = "xfs_inobt",
325 .magic = { cpu_to_be32(XFS_IBT_MAGIC), cpu_to_be32(XFS_IBT_CRC_MAGIC) },
326 .verify_read = xfs_inobt_read_verify,
327 .verify_write = xfs_inobt_write_verify,
328 .verify_struct = xfs_inobt_verify,
329 };
330
331 const struct xfs_buf_ops xfs_finobt_buf_ops = {
332 .name = "xfs_finobt",
333 .magic = { cpu_to_be32(XFS_FIBT_MAGIC),
334 cpu_to_be32(XFS_FIBT_CRC_MAGIC) },
335 .verify_read = xfs_inobt_read_verify,
336 .verify_write = xfs_inobt_write_verify,
337 .verify_struct = xfs_inobt_verify,
338 };
339
340 STATIC int
341 xfs_inobt_keys_inorder(
342 struct xfs_btree_cur *cur,
343 union xfs_btree_key *k1,
344 union xfs_btree_key *k2)
345 {
346 return be32_to_cpu(k1->inobt.ir_startino) <
347 be32_to_cpu(k2->inobt.ir_startino);
348 }
349
350 STATIC int
351 xfs_inobt_recs_inorder(
352 struct xfs_btree_cur *cur,
353 union xfs_btree_rec *r1,
354 union xfs_btree_rec *r2)
355 {
356 return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
357 be32_to_cpu(r2->inobt.ir_startino);
358 }
359
360 static const struct xfs_btree_ops xfs_inobt_ops = {
361 .rec_len = sizeof(xfs_inobt_rec_t),
362 .key_len = sizeof(xfs_inobt_key_t),
363
364 .dup_cursor = xfs_inobt_dup_cursor,
365 .set_root = xfs_inobt_set_root,
366 .alloc_block = xfs_inobt_alloc_block,
367 .free_block = xfs_inobt_free_block,
368 .get_minrecs = xfs_inobt_get_minrecs,
369 .get_maxrecs = xfs_inobt_get_maxrecs,
370 .init_key_from_rec = xfs_inobt_init_key_from_rec,
371 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
372 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
373 .init_ptr_from_cur = xfs_inobt_init_ptr_from_cur,
374 .key_diff = xfs_inobt_key_diff,
375 .buf_ops = &xfs_inobt_buf_ops,
376 .diff_two_keys = xfs_inobt_diff_two_keys,
377 .keys_inorder = xfs_inobt_keys_inorder,
378 .recs_inorder = xfs_inobt_recs_inorder,
379 };
380
381 static const struct xfs_btree_ops xfs_finobt_ops = {
382 .rec_len = sizeof(xfs_inobt_rec_t),
383 .key_len = sizeof(xfs_inobt_key_t),
384
385 .dup_cursor = xfs_inobt_dup_cursor,
386 .set_root = xfs_finobt_set_root,
387 .alloc_block = xfs_finobt_alloc_block,
388 .free_block = xfs_finobt_free_block,
389 .get_minrecs = xfs_inobt_get_minrecs,
390 .get_maxrecs = xfs_inobt_get_maxrecs,
391 .init_key_from_rec = xfs_inobt_init_key_from_rec,
392 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
393 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
394 .init_ptr_from_cur = xfs_finobt_init_ptr_from_cur,
395 .key_diff = xfs_inobt_key_diff,
396 .buf_ops = &xfs_finobt_buf_ops,
397 .diff_two_keys = xfs_inobt_diff_two_keys,
398 .keys_inorder = xfs_inobt_keys_inorder,
399 .recs_inorder = xfs_inobt_recs_inorder,
400 };
401
402 /*
403 * Allocate a new inode btree cursor.
404 */
405 struct xfs_btree_cur * /* new inode btree cursor */
406 xfs_inobt_init_cursor(
407 struct xfs_mount *mp, /* file system mount point */
408 struct xfs_trans *tp, /* transaction pointer */
409 struct xfs_buf *agbp, /* buffer for agi structure */
410 xfs_agnumber_t agno, /* allocation group number */
411 xfs_btnum_t btnum) /* ialloc or free ino btree */
412 {
413 struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp);
414 struct xfs_btree_cur *cur;
415
416 cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
417
418 cur->bc_tp = tp;
419 cur->bc_mp = mp;
420 cur->bc_btnum = btnum;
421 if (btnum == XFS_BTNUM_INO) {
422 cur->bc_nlevels = be32_to_cpu(agi->agi_level);
423 cur->bc_ops = &xfs_inobt_ops;
424 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_ibt_2);
425 } else {
426 cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
427 cur->bc_ops = &xfs_finobt_ops;
428 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_fibt_2);
429 }
430
431 cur->bc_blocklog = mp->m_sb.sb_blocklog;
432
433 if (xfs_sb_version_hascrc(&mp->m_sb))
434 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
435
436 cur->bc_private.a.agbp = agbp;
437 cur->bc_private.a.agno = agno;
438
439 return cur;
440 }
441
442 /*
443 * Calculate number of records in an inobt btree block.
444 */
445 int
446 xfs_inobt_maxrecs(
447 struct xfs_mount *mp,
448 int blocklen,
449 int leaf)
450 {
451 blocklen -= XFS_INOBT_BLOCK_LEN(mp);
452
453 if (leaf)
454 return blocklen / sizeof(xfs_inobt_rec_t);
455 return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
456 }
457
458 /*
459 * Convert the inode record holemask to an inode allocation bitmap. The inode
460 * allocation bitmap is inode granularity and specifies whether an inode is
461 * physically allocated on disk (not whether the inode is considered allocated
462 * or free by the fs).
463 *
464 * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
465 */
466 uint64_t
467 xfs_inobt_irec_to_allocmask(
468 struct xfs_inobt_rec_incore *rec)
469 {
470 uint64_t bitmap = 0;
471 uint64_t inodespbit;
472 int nextbit;
473 uint allocbitmap;
474
475 /*
476 * The holemask has 16-bits for a 64 inode record. Therefore each
477 * holemask bit represents multiple inodes. Create a mask of bits to set
478 * in the allocmask for each holemask bit.
479 */
480 inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
481
482 /*
483 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
484 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
485 * anything beyond the 16 holemask bits since this casts to a larger
486 * type.
487 */
488 allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
489
490 /*
491 * allocbitmap is the inverted holemask so every set bit represents
492 * allocated inodes. To expand from 16-bit holemask granularity to
493 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
494 * bitmap for every holemask bit.
495 */
496 nextbit = xfs_next_bit(&allocbitmap, 1, 0);
497 while (nextbit != -1) {
498 ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
499
500 bitmap |= (inodespbit <<
501 (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
502
503 nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
504 }
505
506 return bitmap;
507 }
508
509 #if defined(DEBUG) || defined(XFS_WARN)
510 /*
511 * Verify that an in-core inode record has a valid inode count.
512 */
513 int
514 xfs_inobt_rec_check_count(
515 struct xfs_mount *mp,
516 struct xfs_inobt_rec_incore *rec)
517 {
518 int inocount = 0;
519 int nextbit = 0;
520 uint64_t allocbmap;
521 int wordsz;
522
523 wordsz = sizeof(allocbmap) / sizeof(unsigned int);
524 allocbmap = xfs_inobt_irec_to_allocmask(rec);
525
526 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
527 while (nextbit != -1) {
528 inocount++;
529 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
530 nextbit + 1);
531 }
532
533 if (inocount != rec->ir_count)
534 return -EFSCORRUPTED;
535
536 return 0;
537 }
538 #endif /* DEBUG */
539
540 static xfs_extlen_t
541 xfs_inobt_max_size(
542 struct xfs_mount *mp,
543 xfs_agnumber_t agno)
544 {
545 xfs_agblock_t agblocks = xfs_ag_block_count(mp, agno);
546
547 /* Bail out if we're uninitialized, which can happen in mkfs. */
548 if (mp->m_inobt_mxr[0] == 0)
549 return 0;
550
551 /*
552 * The log is permanently allocated, so the space it occupies will
553 * never be available for the kinds of things that would require btree
554 * expansion. We therefore can pretend the space isn't there.
555 */
556 if (mp->m_sb.sb_logstart &&
557 XFS_FSB_TO_AGNO(mp, mp->m_sb.sb_logstart) == agno)
558 agblocks -= mp->m_sb.sb_logblocks;
559
560 return xfs_btree_calc_size(mp->m_inobt_mnr,
561 (uint64_t)agblocks * mp->m_sb.sb_inopblock /
562 XFS_INODES_PER_CHUNK);
563 }
564
565 static int
566 xfs_inobt_count_blocks(
567 struct xfs_mount *mp,
568 struct xfs_trans *tp,
569 xfs_agnumber_t agno,
570 xfs_btnum_t btnum,
571 xfs_extlen_t *tree_blocks)
572 {
573 struct xfs_buf *agbp;
574 struct xfs_btree_cur *cur;
575 int error;
576
577 error = xfs_ialloc_read_agi(mp, tp, agno, &agbp);
578 if (error)
579 return error;
580
581 cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, btnum);
582 error = xfs_btree_count_blocks(cur, tree_blocks);
583 xfs_btree_del_cursor(cur, error);
584 xfs_trans_brelse(tp, agbp);
585
586 return error;
587 }
588
589 /*
590 * Figure out how many blocks to reserve and how many are used by this btree.
591 */
592 int
593 xfs_finobt_calc_reserves(
594 struct xfs_mount *mp,
595 struct xfs_trans *tp,
596 xfs_agnumber_t agno,
597 xfs_extlen_t *ask,
598 xfs_extlen_t *used)
599 {
600 xfs_extlen_t tree_len = 0;
601 int error;
602
603 if (!xfs_sb_version_hasfinobt(&mp->m_sb))
604 return 0;
605
606 error = xfs_inobt_count_blocks(mp, tp, agno, XFS_BTNUM_FINO, &tree_len);
607 if (error)
608 return error;
609
610 *ask += xfs_inobt_max_size(mp, agno);
611 *used += tree_len;
612 return 0;
613 }
614
615 /* Calculate the inobt btree size for some records. */
616 xfs_extlen_t
617 xfs_iallocbt_calc_size(
618 struct xfs_mount *mp,
619 unsigned long long len)
620 {
621 return xfs_btree_calc_size(mp->m_inobt_mnr, len);
622 }