Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
4 * All Rights Reserved.
5 */
6#include "xfs.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_btree.h"
15#include "xfs_btree_staging.h"
16#include "xfs_ialloc.h"
17#include "xfs_ialloc_btree.h"
18#include "xfs_alloc.h"
19#include "xfs_error.h"
20#include "xfs_trace.h"
21#include "xfs_trans.h"
22#include "xfs_rmap.h"
23#include "xfs_ag.h"
24
25static struct kmem_cache *xfs_inobt_cur_cache;
26
27STATIC int
28xfs_inobt_get_minrecs(
29 struct xfs_btree_cur *cur,
30 int level)
31{
32 return M_IGEO(cur->bc_mp)->inobt_mnr[level != 0];
33}
34
35STATIC struct xfs_btree_cur *
36xfs_inobt_dup_cursor(
37 struct xfs_btree_cur *cur)
38{
39 return xfs_inobt_init_cursor(cur->bc_ag.pag, cur->bc_tp,
40 cur->bc_ag.agbp, cur->bc_btnum);
41}
42
43STATIC void
44xfs_inobt_set_root(
45 struct xfs_btree_cur *cur,
46 const union xfs_btree_ptr *nptr,
47 int inc) /* level change */
48{
49 struct xfs_buf *agbp = cur->bc_ag.agbp;
50 struct xfs_agi *agi = agbp->b_addr;
51
52 agi->agi_root = nptr->s;
53 be32_add_cpu(&agi->agi_level, inc);
54 xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_ROOT | XFS_AGI_LEVEL);
55}
56
57STATIC void
58xfs_finobt_set_root(
59 struct xfs_btree_cur *cur,
60 const union xfs_btree_ptr *nptr,
61 int inc) /* level change */
62{
63 struct xfs_buf *agbp = cur->bc_ag.agbp;
64 struct xfs_agi *agi = agbp->b_addr;
65
66 agi->agi_free_root = nptr->s;
67 be32_add_cpu(&agi->agi_free_level, inc);
68 xfs_ialloc_log_agi(cur->bc_tp, agbp,
69 XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL);
70}
71
72/* Update the inode btree block counter for this btree. */
73static inline void
74xfs_inobt_mod_blockcount(
75 struct xfs_btree_cur *cur,
76 int howmuch)
77{
78 struct xfs_buf *agbp = cur->bc_ag.agbp;
79 struct xfs_agi *agi = agbp->b_addr;
80
81 if (!xfs_has_inobtcounts(cur->bc_mp))
82 return;
83
84 if (cur->bc_btnum == XFS_BTNUM_FINO)
85 be32_add_cpu(&agi->agi_fblocks, howmuch);
86 else if (cur->bc_btnum == XFS_BTNUM_INO)
87 be32_add_cpu(&agi->agi_iblocks, howmuch);
88 xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_IBLOCKS);
89}
90
91STATIC int
92__xfs_inobt_alloc_block(
93 struct xfs_btree_cur *cur,
94 const union xfs_btree_ptr *start,
95 union xfs_btree_ptr *new,
96 int *stat,
97 enum xfs_ag_resv_type resv)
98{
99 xfs_alloc_arg_t args; /* block allocation args */
100 int error; /* error return value */
101 xfs_agblock_t sbno = be32_to_cpu(start->s);
102
103 memset(&args, 0, sizeof(args));
104 args.tp = cur->bc_tp;
105 args.mp = cur->bc_mp;
106 args.pag = cur->bc_ag.pag;
107 args.oinfo = XFS_RMAP_OINFO_INOBT;
108 args.minlen = 1;
109 args.maxlen = 1;
110 args.prod = 1;
111 args.resv = resv;
112
113 error = xfs_alloc_vextent_near_bno(&args,
114 XFS_AGB_TO_FSB(args.mp, args.pag->pag_agno, sbno));
115 if (error)
116 return error;
117
118 if (args.fsbno == NULLFSBLOCK) {
119 *stat = 0;
120 return 0;
121 }
122 ASSERT(args.len == 1);
123
124 new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
125 *stat = 1;
126 xfs_inobt_mod_blockcount(cur, 1);
127 return 0;
128}
129
130STATIC int
131xfs_inobt_alloc_block(
132 struct xfs_btree_cur *cur,
133 const union xfs_btree_ptr *start,
134 union xfs_btree_ptr *new,
135 int *stat)
136{
137 return __xfs_inobt_alloc_block(cur, start, new, stat, XFS_AG_RESV_NONE);
138}
139
140STATIC int
141xfs_finobt_alloc_block(
142 struct xfs_btree_cur *cur,
143 const union xfs_btree_ptr *start,
144 union xfs_btree_ptr *new,
145 int *stat)
146{
147 if (cur->bc_mp->m_finobt_nores)
148 return xfs_inobt_alloc_block(cur, start, new, stat);
149 return __xfs_inobt_alloc_block(cur, start, new, stat,
150 XFS_AG_RESV_METADATA);
151}
152
153STATIC int
154__xfs_inobt_free_block(
155 struct xfs_btree_cur *cur,
156 struct xfs_buf *bp,
157 enum xfs_ag_resv_type resv)
158{
159 xfs_fsblock_t fsbno;
160
161 xfs_inobt_mod_blockcount(cur, -1);
162 fsbno = XFS_DADDR_TO_FSB(cur->bc_mp, xfs_buf_daddr(bp));
163 return xfs_free_extent_later(cur->bc_tp, fsbno, 1,
164 &XFS_RMAP_OINFO_INOBT, resv, false);
165}
166
167STATIC int
168xfs_inobt_free_block(
169 struct xfs_btree_cur *cur,
170 struct xfs_buf *bp)
171{
172 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_NONE);
173}
174
175STATIC int
176xfs_finobt_free_block(
177 struct xfs_btree_cur *cur,
178 struct xfs_buf *bp)
179{
180 if (cur->bc_mp->m_finobt_nores)
181 return xfs_inobt_free_block(cur, bp);
182 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_METADATA);
183}
184
185STATIC int
186xfs_inobt_get_maxrecs(
187 struct xfs_btree_cur *cur,
188 int level)
189{
190 return M_IGEO(cur->bc_mp)->inobt_mxr[level != 0];
191}
192
193STATIC void
194xfs_inobt_init_key_from_rec(
195 union xfs_btree_key *key,
196 const union xfs_btree_rec *rec)
197{
198 key->inobt.ir_startino = rec->inobt.ir_startino;
199}
200
201STATIC void
202xfs_inobt_init_high_key_from_rec(
203 union xfs_btree_key *key,
204 const union xfs_btree_rec *rec)
205{
206 __u32 x;
207
208 x = be32_to_cpu(rec->inobt.ir_startino);
209 x += XFS_INODES_PER_CHUNK - 1;
210 key->inobt.ir_startino = cpu_to_be32(x);
211}
212
213STATIC void
214xfs_inobt_init_rec_from_cur(
215 struct xfs_btree_cur *cur,
216 union xfs_btree_rec *rec)
217{
218 rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
219 if (xfs_has_sparseinodes(cur->bc_mp)) {
220 rec->inobt.ir_u.sp.ir_holemask =
221 cpu_to_be16(cur->bc_rec.i.ir_holemask);
222 rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
223 rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
224 } else {
225 /* ir_holemask/ir_count not supported on-disk */
226 rec->inobt.ir_u.f.ir_freecount =
227 cpu_to_be32(cur->bc_rec.i.ir_freecount);
228 }
229 rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
230}
231
232/*
233 * initial value of ptr for lookup
234 */
235STATIC void
236xfs_inobt_init_ptr_from_cur(
237 struct xfs_btree_cur *cur,
238 union xfs_btree_ptr *ptr)
239{
240 struct xfs_agi *agi = cur->bc_ag.agbp->b_addr;
241
242 ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agi->agi_seqno));
243
244 ptr->s = agi->agi_root;
245}
246
247STATIC void
248xfs_finobt_init_ptr_from_cur(
249 struct xfs_btree_cur *cur,
250 union xfs_btree_ptr *ptr)
251{
252 struct xfs_agi *agi = cur->bc_ag.agbp->b_addr;
253
254 ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agi->agi_seqno));
255 ptr->s = agi->agi_free_root;
256}
257
258STATIC int64_t
259xfs_inobt_key_diff(
260 struct xfs_btree_cur *cur,
261 const union xfs_btree_key *key)
262{
263 return (int64_t)be32_to_cpu(key->inobt.ir_startino) -
264 cur->bc_rec.i.ir_startino;
265}
266
267STATIC int64_t
268xfs_inobt_diff_two_keys(
269 struct xfs_btree_cur *cur,
270 const union xfs_btree_key *k1,
271 const union xfs_btree_key *k2,
272 const union xfs_btree_key *mask)
273{
274 ASSERT(!mask || mask->inobt.ir_startino);
275
276 return (int64_t)be32_to_cpu(k1->inobt.ir_startino) -
277 be32_to_cpu(k2->inobt.ir_startino);
278}
279
280static xfs_failaddr_t
281xfs_inobt_verify(
282 struct xfs_buf *bp)
283{
284 struct xfs_mount *mp = bp->b_mount;
285 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
286 xfs_failaddr_t fa;
287 unsigned int level;
288
289 if (!xfs_verify_magic(bp, block->bb_magic))
290 return __this_address;
291
292 /*
293 * During growfs operations, we can't verify the exact owner as the
294 * perag is not fully initialised and hence not attached to the buffer.
295 *
296 * Similarly, during log recovery we will have a perag structure
297 * attached, but the agi information will not yet have been initialised
298 * from the on disk AGI. We don't currently use any of this information,
299 * but beware of the landmine (i.e. need to check
300 * xfs_perag_initialised_agi(pag)) if we ever do.
301 */
302 if (xfs_has_crc(mp)) {
303 fa = xfs_btree_sblock_v5hdr_verify(bp);
304 if (fa)
305 return fa;
306 }
307
308 /* level verification */
309 level = be16_to_cpu(block->bb_level);
310 if (level >= M_IGEO(mp)->inobt_maxlevels)
311 return __this_address;
312
313 return xfs_btree_sblock_verify(bp,
314 M_IGEO(mp)->inobt_mxr[level != 0]);
315}
316
317static void
318xfs_inobt_read_verify(
319 struct xfs_buf *bp)
320{
321 xfs_failaddr_t fa;
322
323 if (!xfs_btree_sblock_verify_crc(bp))
324 xfs_verifier_error(bp, -EFSBADCRC, __this_address);
325 else {
326 fa = xfs_inobt_verify(bp);
327 if (fa)
328 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
329 }
330
331 if (bp->b_error)
332 trace_xfs_btree_corrupt(bp, _RET_IP_);
333}
334
335static void
336xfs_inobt_write_verify(
337 struct xfs_buf *bp)
338{
339 xfs_failaddr_t fa;
340
341 fa = xfs_inobt_verify(bp);
342 if (fa) {
343 trace_xfs_btree_corrupt(bp, _RET_IP_);
344 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
345 return;
346 }
347 xfs_btree_sblock_calc_crc(bp);
348
349}
350
351const struct xfs_buf_ops xfs_inobt_buf_ops = {
352 .name = "xfs_inobt",
353 .magic = { cpu_to_be32(XFS_IBT_MAGIC), cpu_to_be32(XFS_IBT_CRC_MAGIC) },
354 .verify_read = xfs_inobt_read_verify,
355 .verify_write = xfs_inobt_write_verify,
356 .verify_struct = xfs_inobt_verify,
357};
358
359const struct xfs_buf_ops xfs_finobt_buf_ops = {
360 .name = "xfs_finobt",
361 .magic = { cpu_to_be32(XFS_FIBT_MAGIC),
362 cpu_to_be32(XFS_FIBT_CRC_MAGIC) },
363 .verify_read = xfs_inobt_read_verify,
364 .verify_write = xfs_inobt_write_verify,
365 .verify_struct = xfs_inobt_verify,
366};
367
368STATIC int
369xfs_inobt_keys_inorder(
370 struct xfs_btree_cur *cur,
371 const union xfs_btree_key *k1,
372 const union xfs_btree_key *k2)
373{
374 return be32_to_cpu(k1->inobt.ir_startino) <
375 be32_to_cpu(k2->inobt.ir_startino);
376}
377
378STATIC int
379xfs_inobt_recs_inorder(
380 struct xfs_btree_cur *cur,
381 const union xfs_btree_rec *r1,
382 const union xfs_btree_rec *r2)
383{
384 return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
385 be32_to_cpu(r2->inobt.ir_startino);
386}
387
388STATIC enum xbtree_key_contig
389xfs_inobt_keys_contiguous(
390 struct xfs_btree_cur *cur,
391 const union xfs_btree_key *key1,
392 const union xfs_btree_key *key2,
393 const union xfs_btree_key *mask)
394{
395 ASSERT(!mask || mask->inobt.ir_startino);
396
397 return xbtree_key_contig(be32_to_cpu(key1->inobt.ir_startino),
398 be32_to_cpu(key2->inobt.ir_startino));
399}
400
401static const struct xfs_btree_ops xfs_inobt_ops = {
402 .rec_len = sizeof(xfs_inobt_rec_t),
403 .key_len = sizeof(xfs_inobt_key_t),
404
405 .dup_cursor = xfs_inobt_dup_cursor,
406 .set_root = xfs_inobt_set_root,
407 .alloc_block = xfs_inobt_alloc_block,
408 .free_block = xfs_inobt_free_block,
409 .get_minrecs = xfs_inobt_get_minrecs,
410 .get_maxrecs = xfs_inobt_get_maxrecs,
411 .init_key_from_rec = xfs_inobt_init_key_from_rec,
412 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
413 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
414 .init_ptr_from_cur = xfs_inobt_init_ptr_from_cur,
415 .key_diff = xfs_inobt_key_diff,
416 .buf_ops = &xfs_inobt_buf_ops,
417 .diff_two_keys = xfs_inobt_diff_two_keys,
418 .keys_inorder = xfs_inobt_keys_inorder,
419 .recs_inorder = xfs_inobt_recs_inorder,
420 .keys_contiguous = xfs_inobt_keys_contiguous,
421};
422
423static const struct xfs_btree_ops xfs_finobt_ops = {
424 .rec_len = sizeof(xfs_inobt_rec_t),
425 .key_len = sizeof(xfs_inobt_key_t),
426
427 .dup_cursor = xfs_inobt_dup_cursor,
428 .set_root = xfs_finobt_set_root,
429 .alloc_block = xfs_finobt_alloc_block,
430 .free_block = xfs_finobt_free_block,
431 .get_minrecs = xfs_inobt_get_minrecs,
432 .get_maxrecs = xfs_inobt_get_maxrecs,
433 .init_key_from_rec = xfs_inobt_init_key_from_rec,
434 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
435 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
436 .init_ptr_from_cur = xfs_finobt_init_ptr_from_cur,
437 .key_diff = xfs_inobt_key_diff,
438 .buf_ops = &xfs_finobt_buf_ops,
439 .diff_two_keys = xfs_inobt_diff_two_keys,
440 .keys_inorder = xfs_inobt_keys_inorder,
441 .recs_inorder = xfs_inobt_recs_inorder,
442 .keys_contiguous = xfs_inobt_keys_contiguous,
443};
444
445/*
446 * Initialize a new inode btree cursor.
447 */
448static struct xfs_btree_cur *
449xfs_inobt_init_common(
450 struct xfs_perag *pag,
451 struct xfs_trans *tp, /* transaction pointer */
452 xfs_btnum_t btnum) /* ialloc or free ino btree */
453{
454 struct xfs_mount *mp = pag->pag_mount;
455 struct xfs_btree_cur *cur;
456
457 cur = xfs_btree_alloc_cursor(mp, tp, btnum,
458 M_IGEO(mp)->inobt_maxlevels, xfs_inobt_cur_cache);
459 if (btnum == XFS_BTNUM_INO) {
460 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_ibt_2);
461 cur->bc_ops = &xfs_inobt_ops;
462 } else {
463 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_fibt_2);
464 cur->bc_ops = &xfs_finobt_ops;
465 }
466
467 if (xfs_has_crc(mp))
468 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
469
470 cur->bc_ag.pag = xfs_perag_hold(pag);
471 return cur;
472}
473
474/* Create an inode btree cursor. */
475struct xfs_btree_cur *
476xfs_inobt_init_cursor(
477 struct xfs_perag *pag,
478 struct xfs_trans *tp,
479 struct xfs_buf *agbp,
480 xfs_btnum_t btnum)
481{
482 struct xfs_btree_cur *cur;
483 struct xfs_agi *agi = agbp->b_addr;
484
485 cur = xfs_inobt_init_common(pag, tp, btnum);
486 if (btnum == XFS_BTNUM_INO)
487 cur->bc_nlevels = be32_to_cpu(agi->agi_level);
488 else
489 cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
490 cur->bc_ag.agbp = agbp;
491 return cur;
492}
493
494/* Create an inode btree cursor with a fake root for staging. */
495struct xfs_btree_cur *
496xfs_inobt_stage_cursor(
497 struct xfs_perag *pag,
498 struct xbtree_afakeroot *afake,
499 xfs_btnum_t btnum)
500{
501 struct xfs_btree_cur *cur;
502
503 cur = xfs_inobt_init_common(pag, NULL, btnum);
504 xfs_btree_stage_afakeroot(cur, afake);
505 return cur;
506}
507
508/*
509 * Install a new inobt btree root. Caller is responsible for invalidating
510 * and freeing the old btree blocks.
511 */
512void
513xfs_inobt_commit_staged_btree(
514 struct xfs_btree_cur *cur,
515 struct xfs_trans *tp,
516 struct xfs_buf *agbp)
517{
518 struct xfs_agi *agi = agbp->b_addr;
519 struct xbtree_afakeroot *afake = cur->bc_ag.afake;
520 int fields;
521
522 ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
523
524 if (cur->bc_btnum == XFS_BTNUM_INO) {
525 fields = XFS_AGI_ROOT | XFS_AGI_LEVEL;
526 agi->agi_root = cpu_to_be32(afake->af_root);
527 agi->agi_level = cpu_to_be32(afake->af_levels);
528 if (xfs_has_inobtcounts(cur->bc_mp)) {
529 agi->agi_iblocks = cpu_to_be32(afake->af_blocks);
530 fields |= XFS_AGI_IBLOCKS;
531 }
532 xfs_ialloc_log_agi(tp, agbp, fields);
533 xfs_btree_commit_afakeroot(cur, tp, agbp, &xfs_inobt_ops);
534 } else {
535 fields = XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL;
536 agi->agi_free_root = cpu_to_be32(afake->af_root);
537 agi->agi_free_level = cpu_to_be32(afake->af_levels);
538 if (xfs_has_inobtcounts(cur->bc_mp)) {
539 agi->agi_fblocks = cpu_to_be32(afake->af_blocks);
540 fields |= XFS_AGI_IBLOCKS;
541 }
542 xfs_ialloc_log_agi(tp, agbp, fields);
543 xfs_btree_commit_afakeroot(cur, tp, agbp, &xfs_finobt_ops);
544 }
545}
546
547/* Calculate number of records in an inode btree block. */
548static inline unsigned int
549xfs_inobt_block_maxrecs(
550 unsigned int blocklen,
551 bool leaf)
552{
553 if (leaf)
554 return blocklen / sizeof(xfs_inobt_rec_t);
555 return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
556}
557
558/*
559 * Calculate number of records in an inobt btree block.
560 */
561int
562xfs_inobt_maxrecs(
563 struct xfs_mount *mp,
564 int blocklen,
565 int leaf)
566{
567 blocklen -= XFS_INOBT_BLOCK_LEN(mp);
568 return xfs_inobt_block_maxrecs(blocklen, leaf);
569}
570
571/*
572 * Maximum number of inode btree records per AG. Pretend that we can fill an
573 * entire AG completely full of inodes except for the AG headers.
574 */
575#define XFS_MAX_INODE_RECORDS \
576 ((XFS_MAX_AG_BYTES - (4 * BBSIZE)) / XFS_DINODE_MIN_SIZE) / \
577 XFS_INODES_PER_CHUNK
578
579/* Compute the max possible height for the inode btree. */
580static inline unsigned int
581xfs_inobt_maxlevels_ondisk(void)
582{
583 unsigned int minrecs[2];
584 unsigned int blocklen;
585
586 blocklen = min(XFS_MIN_BLOCKSIZE - XFS_BTREE_SBLOCK_LEN,
587 XFS_MIN_CRC_BLOCKSIZE - XFS_BTREE_SBLOCK_CRC_LEN);
588
589 minrecs[0] = xfs_inobt_block_maxrecs(blocklen, true) / 2;
590 minrecs[1] = xfs_inobt_block_maxrecs(blocklen, false) / 2;
591
592 return xfs_btree_compute_maxlevels(minrecs, XFS_MAX_INODE_RECORDS);
593}
594
595/* Compute the max possible height for the free inode btree. */
596static inline unsigned int
597xfs_finobt_maxlevels_ondisk(void)
598{
599 unsigned int minrecs[2];
600 unsigned int blocklen;
601
602 blocklen = XFS_MIN_CRC_BLOCKSIZE - XFS_BTREE_SBLOCK_CRC_LEN;
603
604 minrecs[0] = xfs_inobt_block_maxrecs(blocklen, true) / 2;
605 minrecs[1] = xfs_inobt_block_maxrecs(blocklen, false) / 2;
606
607 return xfs_btree_compute_maxlevels(minrecs, XFS_MAX_INODE_RECORDS);
608}
609
610/* Compute the max possible height for either inode btree. */
611unsigned int
612xfs_iallocbt_maxlevels_ondisk(void)
613{
614 return max(xfs_inobt_maxlevels_ondisk(),
615 xfs_finobt_maxlevels_ondisk());
616}
617
618/*
619 * Convert the inode record holemask to an inode allocation bitmap. The inode
620 * allocation bitmap is inode granularity and specifies whether an inode is
621 * physically allocated on disk (not whether the inode is considered allocated
622 * or free by the fs).
623 *
624 * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
625 */
626uint64_t
627xfs_inobt_irec_to_allocmask(
628 const struct xfs_inobt_rec_incore *rec)
629{
630 uint64_t bitmap = 0;
631 uint64_t inodespbit;
632 int nextbit;
633 uint allocbitmap;
634
635 /*
636 * The holemask has 16-bits for a 64 inode record. Therefore each
637 * holemask bit represents multiple inodes. Create a mask of bits to set
638 * in the allocmask for each holemask bit.
639 */
640 inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
641
642 /*
643 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
644 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
645 * anything beyond the 16 holemask bits since this casts to a larger
646 * type.
647 */
648 allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
649
650 /*
651 * allocbitmap is the inverted holemask so every set bit represents
652 * allocated inodes. To expand from 16-bit holemask granularity to
653 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
654 * bitmap for every holemask bit.
655 */
656 nextbit = xfs_next_bit(&allocbitmap, 1, 0);
657 while (nextbit != -1) {
658 ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
659
660 bitmap |= (inodespbit <<
661 (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
662
663 nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
664 }
665
666 return bitmap;
667}
668
669#if defined(DEBUG) || defined(XFS_WARN)
670/*
671 * Verify that an in-core inode record has a valid inode count.
672 */
673int
674xfs_inobt_rec_check_count(
675 struct xfs_mount *mp,
676 struct xfs_inobt_rec_incore *rec)
677{
678 int inocount = 0;
679 int nextbit = 0;
680 uint64_t allocbmap;
681 int wordsz;
682
683 wordsz = sizeof(allocbmap) / sizeof(unsigned int);
684 allocbmap = xfs_inobt_irec_to_allocmask(rec);
685
686 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
687 while (nextbit != -1) {
688 inocount++;
689 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
690 nextbit + 1);
691 }
692
693 if (inocount != rec->ir_count)
694 return -EFSCORRUPTED;
695
696 return 0;
697}
698#endif /* DEBUG */
699
700static xfs_extlen_t
701xfs_inobt_max_size(
702 struct xfs_perag *pag)
703{
704 struct xfs_mount *mp = pag->pag_mount;
705 xfs_agblock_t agblocks = pag->block_count;
706
707 /* Bail out if we're uninitialized, which can happen in mkfs. */
708 if (M_IGEO(mp)->inobt_mxr[0] == 0)
709 return 0;
710
711 /*
712 * The log is permanently allocated, so the space it occupies will
713 * never be available for the kinds of things that would require btree
714 * expansion. We therefore can pretend the space isn't there.
715 */
716 if (xfs_ag_contains_log(mp, pag->pag_agno))
717 agblocks -= mp->m_sb.sb_logblocks;
718
719 return xfs_btree_calc_size(M_IGEO(mp)->inobt_mnr,
720 (uint64_t)agblocks * mp->m_sb.sb_inopblock /
721 XFS_INODES_PER_CHUNK);
722}
723
724/* Read AGI and create inobt cursor. */
725int
726xfs_inobt_cur(
727 struct xfs_perag *pag,
728 struct xfs_trans *tp,
729 xfs_btnum_t which,
730 struct xfs_btree_cur **curpp,
731 struct xfs_buf **agi_bpp)
732{
733 struct xfs_btree_cur *cur;
734 int error;
735
736 ASSERT(*agi_bpp == NULL);
737 ASSERT(*curpp == NULL);
738
739 error = xfs_ialloc_read_agi(pag, tp, agi_bpp);
740 if (error)
741 return error;
742
743 cur = xfs_inobt_init_cursor(pag, tp, *agi_bpp, which);
744 *curpp = cur;
745 return 0;
746}
747
748static int
749xfs_inobt_count_blocks(
750 struct xfs_perag *pag,
751 struct xfs_trans *tp,
752 xfs_btnum_t btnum,
753 xfs_extlen_t *tree_blocks)
754{
755 struct xfs_buf *agbp = NULL;
756 struct xfs_btree_cur *cur = NULL;
757 int error;
758
759 error = xfs_inobt_cur(pag, tp, btnum, &cur, &agbp);
760 if (error)
761 return error;
762
763 error = xfs_btree_count_blocks(cur, tree_blocks);
764 xfs_btree_del_cursor(cur, error);
765 xfs_trans_brelse(tp, agbp);
766
767 return error;
768}
769
770/* Read finobt block count from AGI header. */
771static int
772xfs_finobt_read_blocks(
773 struct xfs_perag *pag,
774 struct xfs_trans *tp,
775 xfs_extlen_t *tree_blocks)
776{
777 struct xfs_buf *agbp;
778 struct xfs_agi *agi;
779 int error;
780
781 error = xfs_ialloc_read_agi(pag, tp, &agbp);
782 if (error)
783 return error;
784
785 agi = agbp->b_addr;
786 *tree_blocks = be32_to_cpu(agi->agi_fblocks);
787 xfs_trans_brelse(tp, agbp);
788 return 0;
789}
790
791/*
792 * Figure out how many blocks to reserve and how many are used by this btree.
793 */
794int
795xfs_finobt_calc_reserves(
796 struct xfs_perag *pag,
797 struct xfs_trans *tp,
798 xfs_extlen_t *ask,
799 xfs_extlen_t *used)
800{
801 xfs_extlen_t tree_len = 0;
802 int error;
803
804 if (!xfs_has_finobt(pag->pag_mount))
805 return 0;
806
807 if (xfs_has_inobtcounts(pag->pag_mount))
808 error = xfs_finobt_read_blocks(pag, tp, &tree_len);
809 else
810 error = xfs_inobt_count_blocks(pag, tp, XFS_BTNUM_FINO,
811 &tree_len);
812 if (error)
813 return error;
814
815 *ask += xfs_inobt_max_size(pag);
816 *used += tree_len;
817 return 0;
818}
819
820/* Calculate the inobt btree size for some records. */
821xfs_extlen_t
822xfs_iallocbt_calc_size(
823 struct xfs_mount *mp,
824 unsigned long long len)
825{
826 return xfs_btree_calc_size(M_IGEO(mp)->inobt_mnr, len);
827}
828
829int __init
830xfs_inobt_init_cur_cache(void)
831{
832 xfs_inobt_cur_cache = kmem_cache_create("xfs_inobt_cur",
833 xfs_btree_cur_sizeof(xfs_inobt_maxlevels_ondisk()),
834 0, 0, NULL);
835
836 if (!xfs_inobt_cur_cache)
837 return -ENOMEM;
838 return 0;
839}
840
841void
842xfs_inobt_destroy_cur_cache(void)
843{
844 kmem_cache_destroy(xfs_inobt_cur_cache);
845 xfs_inobt_cur_cache = NULL;
846}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
4 * All Rights Reserved.
5 */
6#include "xfs.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_btree.h"
15#include "xfs_btree_staging.h"
16#include "xfs_ialloc.h"
17#include "xfs_ialloc_btree.h"
18#include "xfs_alloc.h"
19#include "xfs_error.h"
20#include "xfs_health.h"
21#include "xfs_trace.h"
22#include "xfs_trans.h"
23#include "xfs_rmap.h"
24#include "xfs_ag.h"
25
26static struct kmem_cache *xfs_inobt_cur_cache;
27
28STATIC int
29xfs_inobt_get_minrecs(
30 struct xfs_btree_cur *cur,
31 int level)
32{
33 return M_IGEO(cur->bc_mp)->inobt_mnr[level != 0];
34}
35
36STATIC struct xfs_btree_cur *
37xfs_inobt_dup_cursor(
38 struct xfs_btree_cur *cur)
39{
40 return xfs_inobt_init_cursor(to_perag(cur->bc_group), cur->bc_tp,
41 cur->bc_ag.agbp);
42}
43
44STATIC struct xfs_btree_cur *
45xfs_finobt_dup_cursor(
46 struct xfs_btree_cur *cur)
47{
48 return xfs_finobt_init_cursor(to_perag(cur->bc_group), cur->bc_tp,
49 cur->bc_ag.agbp);
50}
51
52STATIC void
53xfs_inobt_set_root(
54 struct xfs_btree_cur *cur,
55 const union xfs_btree_ptr *nptr,
56 int inc) /* level change */
57{
58 struct xfs_buf *agbp = cur->bc_ag.agbp;
59 struct xfs_agi *agi = agbp->b_addr;
60
61 agi->agi_root = nptr->s;
62 be32_add_cpu(&agi->agi_level, inc);
63 xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_ROOT | XFS_AGI_LEVEL);
64}
65
66STATIC void
67xfs_finobt_set_root(
68 struct xfs_btree_cur *cur,
69 const union xfs_btree_ptr *nptr,
70 int inc) /* level change */
71{
72 struct xfs_buf *agbp = cur->bc_ag.agbp;
73 struct xfs_agi *agi = agbp->b_addr;
74
75 agi->agi_free_root = nptr->s;
76 be32_add_cpu(&agi->agi_free_level, inc);
77 xfs_ialloc_log_agi(cur->bc_tp, agbp,
78 XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL);
79}
80
81/* Update the inode btree block counter for this btree. */
82static inline void
83xfs_inobt_mod_blockcount(
84 struct xfs_btree_cur *cur,
85 int howmuch)
86{
87 struct xfs_buf *agbp = cur->bc_ag.agbp;
88 struct xfs_agi *agi = agbp->b_addr;
89
90 if (!xfs_has_inobtcounts(cur->bc_mp))
91 return;
92
93 if (xfs_btree_is_fino(cur->bc_ops))
94 be32_add_cpu(&agi->agi_fblocks, howmuch);
95 else
96 be32_add_cpu(&agi->agi_iblocks, howmuch);
97 xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_IBLOCKS);
98}
99
100STATIC int
101__xfs_inobt_alloc_block(
102 struct xfs_btree_cur *cur,
103 const union xfs_btree_ptr *start,
104 union xfs_btree_ptr *new,
105 int *stat,
106 enum xfs_ag_resv_type resv)
107{
108 xfs_alloc_arg_t args; /* block allocation args */
109 int error; /* error return value */
110 xfs_agblock_t sbno = be32_to_cpu(start->s);
111
112 memset(&args, 0, sizeof(args));
113 args.tp = cur->bc_tp;
114 args.mp = cur->bc_mp;
115 args.pag = to_perag(cur->bc_group);
116 args.oinfo = XFS_RMAP_OINFO_INOBT;
117 args.minlen = 1;
118 args.maxlen = 1;
119 args.prod = 1;
120 args.resv = resv;
121
122 error = xfs_alloc_vextent_near_bno(&args,
123 xfs_agbno_to_fsb(args.pag, sbno));
124 if (error)
125 return error;
126
127 if (args.fsbno == NULLFSBLOCK) {
128 *stat = 0;
129 return 0;
130 }
131 ASSERT(args.len == 1);
132
133 new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
134 *stat = 1;
135 xfs_inobt_mod_blockcount(cur, 1);
136 return 0;
137}
138
139STATIC int
140xfs_inobt_alloc_block(
141 struct xfs_btree_cur *cur,
142 const union xfs_btree_ptr *start,
143 union xfs_btree_ptr *new,
144 int *stat)
145{
146 return __xfs_inobt_alloc_block(cur, start, new, stat, XFS_AG_RESV_NONE);
147}
148
149STATIC int
150xfs_finobt_alloc_block(
151 struct xfs_btree_cur *cur,
152 const union xfs_btree_ptr *start,
153 union xfs_btree_ptr *new,
154 int *stat)
155{
156 if (cur->bc_mp->m_finobt_nores)
157 return xfs_inobt_alloc_block(cur, start, new, stat);
158 return __xfs_inobt_alloc_block(cur, start, new, stat,
159 XFS_AG_RESV_METADATA);
160}
161
162STATIC int
163__xfs_inobt_free_block(
164 struct xfs_btree_cur *cur,
165 struct xfs_buf *bp,
166 enum xfs_ag_resv_type resv)
167{
168 xfs_fsblock_t fsbno;
169
170 xfs_inobt_mod_blockcount(cur, -1);
171 fsbno = XFS_DADDR_TO_FSB(cur->bc_mp, xfs_buf_daddr(bp));
172 return xfs_free_extent_later(cur->bc_tp, fsbno, 1,
173 &XFS_RMAP_OINFO_INOBT, resv, 0);
174}
175
176STATIC int
177xfs_inobt_free_block(
178 struct xfs_btree_cur *cur,
179 struct xfs_buf *bp)
180{
181 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_NONE);
182}
183
184STATIC int
185xfs_finobt_free_block(
186 struct xfs_btree_cur *cur,
187 struct xfs_buf *bp)
188{
189 if (cur->bc_mp->m_finobt_nores)
190 return xfs_inobt_free_block(cur, bp);
191 return __xfs_inobt_free_block(cur, bp, XFS_AG_RESV_METADATA);
192}
193
194STATIC int
195xfs_inobt_get_maxrecs(
196 struct xfs_btree_cur *cur,
197 int level)
198{
199 return M_IGEO(cur->bc_mp)->inobt_mxr[level != 0];
200}
201
202STATIC void
203xfs_inobt_init_key_from_rec(
204 union xfs_btree_key *key,
205 const union xfs_btree_rec *rec)
206{
207 key->inobt.ir_startino = rec->inobt.ir_startino;
208}
209
210STATIC void
211xfs_inobt_init_high_key_from_rec(
212 union xfs_btree_key *key,
213 const union xfs_btree_rec *rec)
214{
215 __u32 x;
216
217 x = be32_to_cpu(rec->inobt.ir_startino);
218 x += XFS_INODES_PER_CHUNK - 1;
219 key->inobt.ir_startino = cpu_to_be32(x);
220}
221
222STATIC void
223xfs_inobt_init_rec_from_cur(
224 struct xfs_btree_cur *cur,
225 union xfs_btree_rec *rec)
226{
227 rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
228 if (xfs_has_sparseinodes(cur->bc_mp)) {
229 rec->inobt.ir_u.sp.ir_holemask =
230 cpu_to_be16(cur->bc_rec.i.ir_holemask);
231 rec->inobt.ir_u.sp.ir_count = cur->bc_rec.i.ir_count;
232 rec->inobt.ir_u.sp.ir_freecount = cur->bc_rec.i.ir_freecount;
233 } else {
234 /* ir_holemask/ir_count not supported on-disk */
235 rec->inobt.ir_u.f.ir_freecount =
236 cpu_to_be32(cur->bc_rec.i.ir_freecount);
237 }
238 rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
239}
240
241/*
242 * initial value of ptr for lookup
243 */
244STATIC void
245xfs_inobt_init_ptr_from_cur(
246 struct xfs_btree_cur *cur,
247 union xfs_btree_ptr *ptr)
248{
249 struct xfs_agi *agi = cur->bc_ag.agbp->b_addr;
250
251 ASSERT(cur->bc_group->xg_gno == be32_to_cpu(agi->agi_seqno));
252
253 ptr->s = agi->agi_root;
254}
255
256STATIC void
257xfs_finobt_init_ptr_from_cur(
258 struct xfs_btree_cur *cur,
259 union xfs_btree_ptr *ptr)
260{
261 struct xfs_agi *agi = cur->bc_ag.agbp->b_addr;
262
263 ASSERT(cur->bc_group->xg_gno == be32_to_cpu(agi->agi_seqno));
264
265 ptr->s = agi->agi_free_root;
266}
267
268STATIC int64_t
269xfs_inobt_key_diff(
270 struct xfs_btree_cur *cur,
271 const union xfs_btree_key *key)
272{
273 return (int64_t)be32_to_cpu(key->inobt.ir_startino) -
274 cur->bc_rec.i.ir_startino;
275}
276
277STATIC int64_t
278xfs_inobt_diff_two_keys(
279 struct xfs_btree_cur *cur,
280 const union xfs_btree_key *k1,
281 const union xfs_btree_key *k2,
282 const union xfs_btree_key *mask)
283{
284 ASSERT(!mask || mask->inobt.ir_startino);
285
286 return (int64_t)be32_to_cpu(k1->inobt.ir_startino) -
287 be32_to_cpu(k2->inobt.ir_startino);
288}
289
290static xfs_failaddr_t
291xfs_inobt_verify(
292 struct xfs_buf *bp)
293{
294 struct xfs_mount *mp = bp->b_mount;
295 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
296 xfs_failaddr_t fa;
297 unsigned int level;
298
299 if (!xfs_verify_magic(bp, block->bb_magic))
300 return __this_address;
301
302 /*
303 * During growfs operations, we can't verify the exact owner as the
304 * perag is not fully initialised and hence not attached to the buffer.
305 *
306 * Similarly, during log recovery we will have a perag structure
307 * attached, but the agi information will not yet have been initialised
308 * from the on disk AGI. We don't currently use any of this information,
309 * but beware of the landmine (i.e. need to check
310 * xfs_perag_initialised_agi(pag)) if we ever do.
311 */
312 if (xfs_has_crc(mp)) {
313 fa = xfs_btree_agblock_v5hdr_verify(bp);
314 if (fa)
315 return fa;
316 }
317
318 /* level verification */
319 level = be16_to_cpu(block->bb_level);
320 if (level >= M_IGEO(mp)->inobt_maxlevels)
321 return __this_address;
322
323 return xfs_btree_agblock_verify(bp,
324 M_IGEO(mp)->inobt_mxr[level != 0]);
325}
326
327static void
328xfs_inobt_read_verify(
329 struct xfs_buf *bp)
330{
331 xfs_failaddr_t fa;
332
333 if (!xfs_btree_agblock_verify_crc(bp))
334 xfs_verifier_error(bp, -EFSBADCRC, __this_address);
335 else {
336 fa = xfs_inobt_verify(bp);
337 if (fa)
338 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
339 }
340
341 if (bp->b_error)
342 trace_xfs_btree_corrupt(bp, _RET_IP_);
343}
344
345static void
346xfs_inobt_write_verify(
347 struct xfs_buf *bp)
348{
349 xfs_failaddr_t fa;
350
351 fa = xfs_inobt_verify(bp);
352 if (fa) {
353 trace_xfs_btree_corrupt(bp, _RET_IP_);
354 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
355 return;
356 }
357 xfs_btree_agblock_calc_crc(bp);
358
359}
360
361const struct xfs_buf_ops xfs_inobt_buf_ops = {
362 .name = "xfs_inobt",
363 .magic = { cpu_to_be32(XFS_IBT_MAGIC), cpu_to_be32(XFS_IBT_CRC_MAGIC) },
364 .verify_read = xfs_inobt_read_verify,
365 .verify_write = xfs_inobt_write_verify,
366 .verify_struct = xfs_inobt_verify,
367};
368
369const struct xfs_buf_ops xfs_finobt_buf_ops = {
370 .name = "xfs_finobt",
371 .magic = { cpu_to_be32(XFS_FIBT_MAGIC),
372 cpu_to_be32(XFS_FIBT_CRC_MAGIC) },
373 .verify_read = xfs_inobt_read_verify,
374 .verify_write = xfs_inobt_write_verify,
375 .verify_struct = xfs_inobt_verify,
376};
377
378STATIC int
379xfs_inobt_keys_inorder(
380 struct xfs_btree_cur *cur,
381 const union xfs_btree_key *k1,
382 const union xfs_btree_key *k2)
383{
384 return be32_to_cpu(k1->inobt.ir_startino) <
385 be32_to_cpu(k2->inobt.ir_startino);
386}
387
388STATIC int
389xfs_inobt_recs_inorder(
390 struct xfs_btree_cur *cur,
391 const union xfs_btree_rec *r1,
392 const union xfs_btree_rec *r2)
393{
394 return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
395 be32_to_cpu(r2->inobt.ir_startino);
396}
397
398STATIC enum xbtree_key_contig
399xfs_inobt_keys_contiguous(
400 struct xfs_btree_cur *cur,
401 const union xfs_btree_key *key1,
402 const union xfs_btree_key *key2,
403 const union xfs_btree_key *mask)
404{
405 ASSERT(!mask || mask->inobt.ir_startino);
406
407 return xbtree_key_contig(be32_to_cpu(key1->inobt.ir_startino),
408 be32_to_cpu(key2->inobt.ir_startino));
409}
410
411const struct xfs_btree_ops xfs_inobt_ops = {
412 .name = "ino",
413 .type = XFS_BTREE_TYPE_AG,
414
415 .rec_len = sizeof(xfs_inobt_rec_t),
416 .key_len = sizeof(xfs_inobt_key_t),
417 .ptr_len = XFS_BTREE_SHORT_PTR_LEN,
418
419 .lru_refs = XFS_INO_BTREE_REF,
420 .statoff = XFS_STATS_CALC_INDEX(xs_ibt_2),
421 .sick_mask = XFS_SICK_AG_INOBT,
422
423 .dup_cursor = xfs_inobt_dup_cursor,
424 .set_root = xfs_inobt_set_root,
425 .alloc_block = xfs_inobt_alloc_block,
426 .free_block = xfs_inobt_free_block,
427 .get_minrecs = xfs_inobt_get_minrecs,
428 .get_maxrecs = xfs_inobt_get_maxrecs,
429 .init_key_from_rec = xfs_inobt_init_key_from_rec,
430 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
431 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
432 .init_ptr_from_cur = xfs_inobt_init_ptr_from_cur,
433 .key_diff = xfs_inobt_key_diff,
434 .buf_ops = &xfs_inobt_buf_ops,
435 .diff_two_keys = xfs_inobt_diff_two_keys,
436 .keys_inorder = xfs_inobt_keys_inorder,
437 .recs_inorder = xfs_inobt_recs_inorder,
438 .keys_contiguous = xfs_inobt_keys_contiguous,
439};
440
441const struct xfs_btree_ops xfs_finobt_ops = {
442 .name = "fino",
443 .type = XFS_BTREE_TYPE_AG,
444
445 .rec_len = sizeof(xfs_inobt_rec_t),
446 .key_len = sizeof(xfs_inobt_key_t),
447 .ptr_len = XFS_BTREE_SHORT_PTR_LEN,
448
449 .lru_refs = XFS_INO_BTREE_REF,
450 .statoff = XFS_STATS_CALC_INDEX(xs_fibt_2),
451 .sick_mask = XFS_SICK_AG_FINOBT,
452
453 .dup_cursor = xfs_finobt_dup_cursor,
454 .set_root = xfs_finobt_set_root,
455 .alloc_block = xfs_finobt_alloc_block,
456 .free_block = xfs_finobt_free_block,
457 .get_minrecs = xfs_inobt_get_minrecs,
458 .get_maxrecs = xfs_inobt_get_maxrecs,
459 .init_key_from_rec = xfs_inobt_init_key_from_rec,
460 .init_high_key_from_rec = xfs_inobt_init_high_key_from_rec,
461 .init_rec_from_cur = xfs_inobt_init_rec_from_cur,
462 .init_ptr_from_cur = xfs_finobt_init_ptr_from_cur,
463 .key_diff = xfs_inobt_key_diff,
464 .buf_ops = &xfs_finobt_buf_ops,
465 .diff_two_keys = xfs_inobt_diff_two_keys,
466 .keys_inorder = xfs_inobt_keys_inorder,
467 .recs_inorder = xfs_inobt_recs_inorder,
468 .keys_contiguous = xfs_inobt_keys_contiguous,
469};
470
471/*
472 * Create an inode btree cursor.
473 *
474 * For staging cursors tp and agbp are NULL.
475 */
476struct xfs_btree_cur *
477xfs_inobt_init_cursor(
478 struct xfs_perag *pag,
479 struct xfs_trans *tp,
480 struct xfs_buf *agbp)
481{
482 struct xfs_mount *mp = pag_mount(pag);
483 struct xfs_btree_cur *cur;
484
485 cur = xfs_btree_alloc_cursor(mp, tp, &xfs_inobt_ops,
486 M_IGEO(mp)->inobt_maxlevels, xfs_inobt_cur_cache);
487 cur->bc_group = xfs_group_hold(pag_group(pag));
488 cur->bc_ag.agbp = agbp;
489 if (agbp) {
490 struct xfs_agi *agi = agbp->b_addr;
491
492 cur->bc_nlevels = be32_to_cpu(agi->agi_level);
493 }
494 return cur;
495}
496
497/*
498 * Create a free inode btree cursor.
499 *
500 * For staging cursors tp and agbp are NULL.
501 */
502struct xfs_btree_cur *
503xfs_finobt_init_cursor(
504 struct xfs_perag *pag,
505 struct xfs_trans *tp,
506 struct xfs_buf *agbp)
507{
508 struct xfs_mount *mp = pag_mount(pag);
509 struct xfs_btree_cur *cur;
510
511 cur = xfs_btree_alloc_cursor(mp, tp, &xfs_finobt_ops,
512 M_IGEO(mp)->inobt_maxlevels, xfs_inobt_cur_cache);
513 cur->bc_group = xfs_group_hold(pag_group(pag));
514 cur->bc_ag.agbp = agbp;
515 if (agbp) {
516 struct xfs_agi *agi = agbp->b_addr;
517
518 cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
519 }
520 return cur;
521}
522
523/*
524 * Install a new inobt btree root. Caller is responsible for invalidating
525 * and freeing the old btree blocks.
526 */
527void
528xfs_inobt_commit_staged_btree(
529 struct xfs_btree_cur *cur,
530 struct xfs_trans *tp,
531 struct xfs_buf *agbp)
532{
533 struct xfs_agi *agi = agbp->b_addr;
534 struct xbtree_afakeroot *afake = cur->bc_ag.afake;
535 int fields;
536
537 ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
538
539 if (xfs_btree_is_ino(cur->bc_ops)) {
540 fields = XFS_AGI_ROOT | XFS_AGI_LEVEL;
541 agi->agi_root = cpu_to_be32(afake->af_root);
542 agi->agi_level = cpu_to_be32(afake->af_levels);
543 if (xfs_has_inobtcounts(cur->bc_mp)) {
544 agi->agi_iblocks = cpu_to_be32(afake->af_blocks);
545 fields |= XFS_AGI_IBLOCKS;
546 }
547 xfs_ialloc_log_agi(tp, agbp, fields);
548 xfs_btree_commit_afakeroot(cur, tp, agbp);
549 } else {
550 fields = XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL;
551 agi->agi_free_root = cpu_to_be32(afake->af_root);
552 agi->agi_free_level = cpu_to_be32(afake->af_levels);
553 if (xfs_has_inobtcounts(cur->bc_mp)) {
554 agi->agi_fblocks = cpu_to_be32(afake->af_blocks);
555 fields |= XFS_AGI_IBLOCKS;
556 }
557 xfs_ialloc_log_agi(tp, agbp, fields);
558 xfs_btree_commit_afakeroot(cur, tp, agbp);
559 }
560}
561
562/* Calculate number of records in an inode btree block. */
563static inline unsigned int
564xfs_inobt_block_maxrecs(
565 unsigned int blocklen,
566 bool leaf)
567{
568 if (leaf)
569 return blocklen / sizeof(xfs_inobt_rec_t);
570 return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
571}
572
573/*
574 * Calculate number of records in an inobt btree block.
575 */
576unsigned int
577xfs_inobt_maxrecs(
578 struct xfs_mount *mp,
579 unsigned int blocklen,
580 bool leaf)
581{
582 blocklen -= XFS_INOBT_BLOCK_LEN(mp);
583 return xfs_inobt_block_maxrecs(blocklen, leaf);
584}
585
586/*
587 * Maximum number of inode btree records per AG. Pretend that we can fill an
588 * entire AG completely full of inodes except for the AG headers.
589 */
590#define XFS_MAX_INODE_RECORDS \
591 ((XFS_MAX_AG_BYTES - (4 * BBSIZE)) / XFS_DINODE_MIN_SIZE) / \
592 XFS_INODES_PER_CHUNK
593
594/* Compute the max possible height for the inode btree. */
595static inline unsigned int
596xfs_inobt_maxlevels_ondisk(void)
597{
598 unsigned int minrecs[2];
599 unsigned int blocklen;
600
601 blocklen = min(XFS_MIN_BLOCKSIZE - XFS_BTREE_SBLOCK_LEN,
602 XFS_MIN_CRC_BLOCKSIZE - XFS_BTREE_SBLOCK_CRC_LEN);
603
604 minrecs[0] = xfs_inobt_block_maxrecs(blocklen, true) / 2;
605 minrecs[1] = xfs_inobt_block_maxrecs(blocklen, false) / 2;
606
607 return xfs_btree_compute_maxlevels(minrecs, XFS_MAX_INODE_RECORDS);
608}
609
610/* Compute the max possible height for the free inode btree. */
611static inline unsigned int
612xfs_finobt_maxlevels_ondisk(void)
613{
614 unsigned int minrecs[2];
615 unsigned int blocklen;
616
617 blocklen = XFS_MIN_CRC_BLOCKSIZE - XFS_BTREE_SBLOCK_CRC_LEN;
618
619 minrecs[0] = xfs_inobt_block_maxrecs(blocklen, true) / 2;
620 minrecs[1] = xfs_inobt_block_maxrecs(blocklen, false) / 2;
621
622 return xfs_btree_compute_maxlevels(minrecs, XFS_MAX_INODE_RECORDS);
623}
624
625/* Compute the max possible height for either inode btree. */
626unsigned int
627xfs_iallocbt_maxlevels_ondisk(void)
628{
629 return max(xfs_inobt_maxlevels_ondisk(),
630 xfs_finobt_maxlevels_ondisk());
631}
632
633/*
634 * Convert the inode record holemask to an inode allocation bitmap. The inode
635 * allocation bitmap is inode granularity and specifies whether an inode is
636 * physically allocated on disk (not whether the inode is considered allocated
637 * or free by the fs).
638 *
639 * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
640 */
641uint64_t
642xfs_inobt_irec_to_allocmask(
643 const struct xfs_inobt_rec_incore *rec)
644{
645 uint64_t bitmap = 0;
646 uint64_t inodespbit;
647 int nextbit;
648 uint allocbitmap;
649
650 /*
651 * The holemask has 16-bits for a 64 inode record. Therefore each
652 * holemask bit represents multiple inodes. Create a mask of bits to set
653 * in the allocmask for each holemask bit.
654 */
655 inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
656
657 /*
658 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
659 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
660 * anything beyond the 16 holemask bits since this casts to a larger
661 * type.
662 */
663 allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
664
665 /*
666 * allocbitmap is the inverted holemask so every set bit represents
667 * allocated inodes. To expand from 16-bit holemask granularity to
668 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
669 * bitmap for every holemask bit.
670 */
671 nextbit = xfs_next_bit(&allocbitmap, 1, 0);
672 while (nextbit != -1) {
673 ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
674
675 bitmap |= (inodespbit <<
676 (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
677
678 nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
679 }
680
681 return bitmap;
682}
683
684#if defined(DEBUG) || defined(XFS_WARN)
685/*
686 * Verify that an in-core inode record has a valid inode count.
687 */
688int
689xfs_inobt_rec_check_count(
690 struct xfs_mount *mp,
691 struct xfs_inobt_rec_incore *rec)
692{
693 int inocount = 0;
694 int nextbit = 0;
695 uint64_t allocbmap;
696 int wordsz;
697
698 wordsz = sizeof(allocbmap) / sizeof(unsigned int);
699 allocbmap = xfs_inobt_irec_to_allocmask(rec);
700
701 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
702 while (nextbit != -1) {
703 inocount++;
704 nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
705 nextbit + 1);
706 }
707
708 if (inocount != rec->ir_count)
709 return -EFSCORRUPTED;
710
711 return 0;
712}
713#endif /* DEBUG */
714
715static xfs_extlen_t
716xfs_inobt_max_size(
717 struct xfs_perag *pag)
718{
719 struct xfs_mount *mp = pag_mount(pag);
720 xfs_agblock_t agblocks = pag_group(pag)->xg_block_count;
721
722 /* Bail out if we're uninitialized, which can happen in mkfs. */
723 if (M_IGEO(mp)->inobt_mxr[0] == 0)
724 return 0;
725
726 /*
727 * The log is permanently allocated, so the space it occupies will
728 * never be available for the kinds of things that would require btree
729 * expansion. We therefore can pretend the space isn't there.
730 */
731 if (xfs_ag_contains_log(mp, pag_agno(pag)))
732 agblocks -= mp->m_sb.sb_logblocks;
733
734 return xfs_btree_calc_size(M_IGEO(mp)->inobt_mnr,
735 (uint64_t)agblocks * mp->m_sb.sb_inopblock /
736 XFS_INODES_PER_CHUNK);
737}
738
739static int
740xfs_finobt_count_blocks(
741 struct xfs_perag *pag,
742 struct xfs_trans *tp,
743 xfs_extlen_t *tree_blocks)
744{
745 struct xfs_buf *agbp = NULL;
746 struct xfs_btree_cur *cur;
747 xfs_filblks_t blocks;
748 int error;
749
750 error = xfs_ialloc_read_agi(pag, tp, 0, &agbp);
751 if (error)
752 return error;
753
754 cur = xfs_finobt_init_cursor(pag, tp, agbp);
755 error = xfs_btree_count_blocks(cur, &blocks);
756 xfs_btree_del_cursor(cur, error);
757 xfs_trans_brelse(tp, agbp);
758 *tree_blocks = blocks;
759
760 return error;
761}
762
763/* Read finobt block count from AGI header. */
764static int
765xfs_finobt_read_blocks(
766 struct xfs_perag *pag,
767 struct xfs_trans *tp,
768 xfs_extlen_t *tree_blocks)
769{
770 struct xfs_buf *agbp;
771 struct xfs_agi *agi;
772 int error;
773
774 error = xfs_ialloc_read_agi(pag, tp, 0, &agbp);
775 if (error)
776 return error;
777
778 agi = agbp->b_addr;
779 *tree_blocks = be32_to_cpu(agi->agi_fblocks);
780 xfs_trans_brelse(tp, agbp);
781 return 0;
782}
783
784/*
785 * Figure out how many blocks to reserve and how many are used by this btree.
786 */
787int
788xfs_finobt_calc_reserves(
789 struct xfs_perag *pag,
790 struct xfs_trans *tp,
791 xfs_extlen_t *ask,
792 xfs_extlen_t *used)
793{
794 xfs_extlen_t tree_len = 0;
795 int error;
796
797 if (!xfs_has_finobt(pag_mount(pag)))
798 return 0;
799
800 if (xfs_has_inobtcounts(pag_mount(pag)))
801 error = xfs_finobt_read_blocks(pag, tp, &tree_len);
802 else
803 error = xfs_finobt_count_blocks(pag, tp, &tree_len);
804 if (error)
805 return error;
806
807 *ask += xfs_inobt_max_size(pag);
808 *used += tree_len;
809 return 0;
810}
811
812/* Calculate the inobt btree size for some records. */
813xfs_extlen_t
814xfs_iallocbt_calc_size(
815 struct xfs_mount *mp,
816 unsigned long long len)
817{
818 return xfs_btree_calc_size(M_IGEO(mp)->inobt_mnr, len);
819}
820
821int __init
822xfs_inobt_init_cur_cache(void)
823{
824 xfs_inobt_cur_cache = kmem_cache_create("xfs_inobt_cur",
825 xfs_btree_cur_sizeof(xfs_inobt_maxlevels_ondisk()),
826 0, 0, NULL);
827
828 if (!xfs_inobt_cur_cache)
829 return -ENOMEM;
830 return 0;
831}
832
833void
834xfs_inobt_destroy_cur_cache(void)
835{
836 kmem_cache_destroy(xfs_inobt_cur_cache);
837 xfs_inobt_cur_cache = NULL;
838}