xref: /linux/fs/xfs/libxfs/xfs_ialloc_btree.c (revision a06c3fad49a50d5d5eb078f93e70f4d3eca5d5a5)
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 
26 static struct kmem_cache	*xfs_inobt_cur_cache;
27 
28 STATIC int
29 xfs_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 
36 STATIC struct xfs_btree_cur *
37 xfs_inobt_dup_cursor(
38 	struct xfs_btree_cur	*cur)
39 {
40 	return xfs_inobt_init_cursor(cur->bc_ag.pag, cur->bc_tp,
41 			cur->bc_ag.agbp);
42 }
43 
44 STATIC struct xfs_btree_cur *
45 xfs_finobt_dup_cursor(
46 	struct xfs_btree_cur	*cur)
47 {
48 	return xfs_finobt_init_cursor(cur->bc_ag.pag, cur->bc_tp,
49 			cur->bc_ag.agbp);
50 }
51 
52 STATIC void
53 xfs_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 
66 STATIC void
67 xfs_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. */
82 static inline void
83 xfs_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 
100 STATIC 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 = cur->bc_ag.pag;
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_AGB_TO_FSB(args.mp, args.pag->pag_agno, 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 
139 STATIC int
140 xfs_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 
149 STATIC int
150 xfs_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 
162 STATIC 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 
176 STATIC int
177 xfs_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 
184 STATIC int
185 xfs_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 
194 STATIC int
195 xfs_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 
202 STATIC void
203 xfs_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 
210 STATIC void
211 xfs_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 
222 STATIC void
223 xfs_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  */
244 STATIC void
245 xfs_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_ag.pag->pag_agno == be32_to_cpu(agi->agi_seqno));
252 
253 	ptr->s = agi->agi_root;
254 }
255 
256 STATIC void
257 xfs_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_ag.pag->pag_agno == be32_to_cpu(agi->agi_seqno));
264 	ptr->s = agi->agi_free_root;
265 }
266 
267 STATIC int64_t
268 xfs_inobt_key_diff(
269 	struct xfs_btree_cur		*cur,
270 	const union xfs_btree_key	*key)
271 {
272 	return (int64_t)be32_to_cpu(key->inobt.ir_startino) -
273 			  cur->bc_rec.i.ir_startino;
274 }
275 
276 STATIC int64_t
277 xfs_inobt_diff_two_keys(
278 	struct xfs_btree_cur		*cur,
279 	const union xfs_btree_key	*k1,
280 	const union xfs_btree_key	*k2,
281 	const union xfs_btree_key	*mask)
282 {
283 	ASSERT(!mask || mask->inobt.ir_startino);
284 
285 	return (int64_t)be32_to_cpu(k1->inobt.ir_startino) -
286 			be32_to_cpu(k2->inobt.ir_startino);
287 }
288 
289 static xfs_failaddr_t
290 xfs_inobt_verify(
291 	struct xfs_buf		*bp)
292 {
293 	struct xfs_mount	*mp = bp->b_mount;
294 	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
295 	xfs_failaddr_t		fa;
296 	unsigned int		level;
297 
298 	if (!xfs_verify_magic(bp, block->bb_magic))
299 		return __this_address;
300 
301 	/*
302 	 * During growfs operations, we can't verify the exact owner as the
303 	 * perag is not fully initialised and hence not attached to the buffer.
304 	 *
305 	 * Similarly, during log recovery we will have a perag structure
306 	 * attached, but the agi information will not yet have been initialised
307 	 * from the on disk AGI. We don't currently use any of this information,
308 	 * but beware of the landmine (i.e. need to check
309 	 * xfs_perag_initialised_agi(pag)) if we ever do.
310 	 */
311 	if (xfs_has_crc(mp)) {
312 		fa = xfs_btree_agblock_v5hdr_verify(bp);
313 		if (fa)
314 			return fa;
315 	}
316 
317 	/* level verification */
318 	level = be16_to_cpu(block->bb_level);
319 	if (level >= M_IGEO(mp)->inobt_maxlevels)
320 		return __this_address;
321 
322 	return xfs_btree_agblock_verify(bp,
323 			M_IGEO(mp)->inobt_mxr[level != 0]);
324 }
325 
326 static void
327 xfs_inobt_read_verify(
328 	struct xfs_buf	*bp)
329 {
330 	xfs_failaddr_t	fa;
331 
332 	if (!xfs_btree_agblock_verify_crc(bp))
333 		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
334 	else {
335 		fa = xfs_inobt_verify(bp);
336 		if (fa)
337 			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
338 	}
339 
340 	if (bp->b_error)
341 		trace_xfs_btree_corrupt(bp, _RET_IP_);
342 }
343 
344 static void
345 xfs_inobt_write_verify(
346 	struct xfs_buf	*bp)
347 {
348 	xfs_failaddr_t	fa;
349 
350 	fa = xfs_inobt_verify(bp);
351 	if (fa) {
352 		trace_xfs_btree_corrupt(bp, _RET_IP_);
353 		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
354 		return;
355 	}
356 	xfs_btree_agblock_calc_crc(bp);
357 
358 }
359 
360 const struct xfs_buf_ops xfs_inobt_buf_ops = {
361 	.name = "xfs_inobt",
362 	.magic = { cpu_to_be32(XFS_IBT_MAGIC), cpu_to_be32(XFS_IBT_CRC_MAGIC) },
363 	.verify_read = xfs_inobt_read_verify,
364 	.verify_write = xfs_inobt_write_verify,
365 	.verify_struct = xfs_inobt_verify,
366 };
367 
368 const struct xfs_buf_ops xfs_finobt_buf_ops = {
369 	.name = "xfs_finobt",
370 	.magic = { cpu_to_be32(XFS_FIBT_MAGIC),
371 		   cpu_to_be32(XFS_FIBT_CRC_MAGIC) },
372 	.verify_read = xfs_inobt_read_verify,
373 	.verify_write = xfs_inobt_write_verify,
374 	.verify_struct = xfs_inobt_verify,
375 };
376 
377 STATIC int
378 xfs_inobt_keys_inorder(
379 	struct xfs_btree_cur		*cur,
380 	const union xfs_btree_key	*k1,
381 	const union xfs_btree_key	*k2)
382 {
383 	return be32_to_cpu(k1->inobt.ir_startino) <
384 		be32_to_cpu(k2->inobt.ir_startino);
385 }
386 
387 STATIC int
388 xfs_inobt_recs_inorder(
389 	struct xfs_btree_cur		*cur,
390 	const union xfs_btree_rec	*r1,
391 	const union xfs_btree_rec	*r2)
392 {
393 	return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
394 		be32_to_cpu(r2->inobt.ir_startino);
395 }
396 
397 STATIC enum xbtree_key_contig
398 xfs_inobt_keys_contiguous(
399 	struct xfs_btree_cur		*cur,
400 	const union xfs_btree_key	*key1,
401 	const union xfs_btree_key	*key2,
402 	const union xfs_btree_key	*mask)
403 {
404 	ASSERT(!mask || mask->inobt.ir_startino);
405 
406 	return xbtree_key_contig(be32_to_cpu(key1->inobt.ir_startino),
407 				 be32_to_cpu(key2->inobt.ir_startino));
408 }
409 
410 const struct xfs_btree_ops xfs_inobt_ops = {
411 	.name			= "ino",
412 	.type			= XFS_BTREE_TYPE_AG,
413 
414 	.rec_len		= sizeof(xfs_inobt_rec_t),
415 	.key_len		= sizeof(xfs_inobt_key_t),
416 	.ptr_len		= XFS_BTREE_SHORT_PTR_LEN,
417 
418 	.lru_refs		= XFS_INO_BTREE_REF,
419 	.statoff		= XFS_STATS_CALC_INDEX(xs_ibt_2),
420 	.sick_mask		= XFS_SICK_AG_INOBT,
421 
422 	.dup_cursor		= xfs_inobt_dup_cursor,
423 	.set_root		= xfs_inobt_set_root,
424 	.alloc_block		= xfs_inobt_alloc_block,
425 	.free_block		= xfs_inobt_free_block,
426 	.get_minrecs		= xfs_inobt_get_minrecs,
427 	.get_maxrecs		= xfs_inobt_get_maxrecs,
428 	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
429 	.init_high_key_from_rec	= xfs_inobt_init_high_key_from_rec,
430 	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
431 	.init_ptr_from_cur	= xfs_inobt_init_ptr_from_cur,
432 	.key_diff		= xfs_inobt_key_diff,
433 	.buf_ops		= &xfs_inobt_buf_ops,
434 	.diff_two_keys		= xfs_inobt_diff_two_keys,
435 	.keys_inorder		= xfs_inobt_keys_inorder,
436 	.recs_inorder		= xfs_inobt_recs_inorder,
437 	.keys_contiguous	= xfs_inobt_keys_contiguous,
438 };
439 
440 const struct xfs_btree_ops xfs_finobt_ops = {
441 	.name			= "fino",
442 	.type			= XFS_BTREE_TYPE_AG,
443 
444 	.rec_len		= sizeof(xfs_inobt_rec_t),
445 	.key_len		= sizeof(xfs_inobt_key_t),
446 	.ptr_len		= XFS_BTREE_SHORT_PTR_LEN,
447 
448 	.lru_refs		= XFS_INO_BTREE_REF,
449 	.statoff		= XFS_STATS_CALC_INDEX(xs_fibt_2),
450 	.sick_mask		= XFS_SICK_AG_FINOBT,
451 
452 	.dup_cursor		= xfs_finobt_dup_cursor,
453 	.set_root		= xfs_finobt_set_root,
454 	.alloc_block		= xfs_finobt_alloc_block,
455 	.free_block		= xfs_finobt_free_block,
456 	.get_minrecs		= xfs_inobt_get_minrecs,
457 	.get_maxrecs		= xfs_inobt_get_maxrecs,
458 	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
459 	.init_high_key_from_rec	= xfs_inobt_init_high_key_from_rec,
460 	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
461 	.init_ptr_from_cur	= xfs_finobt_init_ptr_from_cur,
462 	.key_diff		= xfs_inobt_key_diff,
463 	.buf_ops		= &xfs_finobt_buf_ops,
464 	.diff_two_keys		= xfs_inobt_diff_two_keys,
465 	.keys_inorder		= xfs_inobt_keys_inorder,
466 	.recs_inorder		= xfs_inobt_recs_inorder,
467 	.keys_contiguous	= xfs_inobt_keys_contiguous,
468 };
469 
470 /*
471  * Create an inode btree cursor.
472  *
473  * For staging cursors tp and agbp are NULL.
474  */
475 struct xfs_btree_cur *
476 xfs_inobt_init_cursor(
477 	struct xfs_perag	*pag,
478 	struct xfs_trans	*tp,
479 	struct xfs_buf		*agbp)
480 {
481 	struct xfs_mount	*mp = pag->pag_mount;
482 	struct xfs_btree_cur	*cur;
483 
484 	cur = xfs_btree_alloc_cursor(mp, tp, &xfs_inobt_ops,
485 			M_IGEO(mp)->inobt_maxlevels, xfs_inobt_cur_cache);
486 	cur->bc_ag.pag = xfs_perag_hold(pag);
487 	cur->bc_ag.agbp = agbp;
488 	if (agbp) {
489 		struct xfs_agi		*agi = agbp->b_addr;
490 
491 		cur->bc_nlevels = be32_to_cpu(agi->agi_level);
492 	}
493 	return cur;
494 }
495 
496 /*
497  * Create a free inode btree cursor.
498  *
499  * For staging cursors tp and agbp are NULL.
500  */
501 struct xfs_btree_cur *
502 xfs_finobt_init_cursor(
503 	struct xfs_perag	*pag,
504 	struct xfs_trans	*tp,
505 	struct xfs_buf		*agbp)
506 {
507 	struct xfs_mount	*mp = pag->pag_mount;
508 	struct xfs_btree_cur	*cur;
509 
510 	cur = xfs_btree_alloc_cursor(mp, tp, &xfs_finobt_ops,
511 			M_IGEO(mp)->inobt_maxlevels, xfs_inobt_cur_cache);
512 	cur->bc_ag.pag = xfs_perag_hold(pag);
513 	cur->bc_ag.agbp = agbp;
514 	if (agbp) {
515 		struct xfs_agi		*agi = agbp->b_addr;
516 
517 		cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
518 	}
519 	return cur;
520 }
521 
522 /*
523  * Install a new inobt btree root.  Caller is responsible for invalidating
524  * and freeing the old btree blocks.
525  */
526 void
527 xfs_inobt_commit_staged_btree(
528 	struct xfs_btree_cur	*cur,
529 	struct xfs_trans	*tp,
530 	struct xfs_buf		*agbp)
531 {
532 	struct xfs_agi		*agi = agbp->b_addr;
533 	struct xbtree_afakeroot	*afake = cur->bc_ag.afake;
534 	int			fields;
535 
536 	ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
537 
538 	if (xfs_btree_is_ino(cur->bc_ops)) {
539 		fields = XFS_AGI_ROOT | XFS_AGI_LEVEL;
540 		agi->agi_root = cpu_to_be32(afake->af_root);
541 		agi->agi_level = cpu_to_be32(afake->af_levels);
542 		if (xfs_has_inobtcounts(cur->bc_mp)) {
543 			agi->agi_iblocks = cpu_to_be32(afake->af_blocks);
544 			fields |= XFS_AGI_IBLOCKS;
545 		}
546 		xfs_ialloc_log_agi(tp, agbp, fields);
547 		xfs_btree_commit_afakeroot(cur, tp, agbp);
548 	} else {
549 		fields = XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL;
550 		agi->agi_free_root = cpu_to_be32(afake->af_root);
551 		agi->agi_free_level = cpu_to_be32(afake->af_levels);
552 		if (xfs_has_inobtcounts(cur->bc_mp)) {
553 			agi->agi_fblocks = cpu_to_be32(afake->af_blocks);
554 			fields |= XFS_AGI_IBLOCKS;
555 		}
556 		xfs_ialloc_log_agi(tp, agbp, fields);
557 		xfs_btree_commit_afakeroot(cur, tp, agbp);
558 	}
559 }
560 
561 /* Calculate number of records in an inode btree block. */
562 static inline unsigned int
563 xfs_inobt_block_maxrecs(
564 	unsigned int		blocklen,
565 	bool			leaf)
566 {
567 	if (leaf)
568 		return blocklen / sizeof(xfs_inobt_rec_t);
569 	return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
570 }
571 
572 /*
573  * Calculate number of records in an inobt btree block.
574  */
575 int
576 xfs_inobt_maxrecs(
577 	struct xfs_mount	*mp,
578 	int			blocklen,
579 	int			leaf)
580 {
581 	blocklen -= XFS_INOBT_BLOCK_LEN(mp);
582 	return xfs_inobt_block_maxrecs(blocklen, leaf);
583 }
584 
585 /*
586  * Maximum number of inode btree records per AG.  Pretend that we can fill an
587  * entire AG completely full of inodes except for the AG headers.
588  */
589 #define XFS_MAX_INODE_RECORDS \
590 	((XFS_MAX_AG_BYTES - (4 * BBSIZE)) / XFS_DINODE_MIN_SIZE) / \
591 			XFS_INODES_PER_CHUNK
592 
593 /* Compute the max possible height for the inode btree. */
594 static inline unsigned int
595 xfs_inobt_maxlevels_ondisk(void)
596 {
597 	unsigned int		minrecs[2];
598 	unsigned int		blocklen;
599 
600 	blocklen = min(XFS_MIN_BLOCKSIZE - XFS_BTREE_SBLOCK_LEN,
601 		       XFS_MIN_CRC_BLOCKSIZE - XFS_BTREE_SBLOCK_CRC_LEN);
602 
603 	minrecs[0] = xfs_inobt_block_maxrecs(blocklen, true) / 2;
604 	minrecs[1] = xfs_inobt_block_maxrecs(blocklen, false) / 2;
605 
606 	return xfs_btree_compute_maxlevels(minrecs, XFS_MAX_INODE_RECORDS);
607 }
608 
609 /* Compute the max possible height for the free inode btree. */
610 static inline unsigned int
611 xfs_finobt_maxlevels_ondisk(void)
612 {
613 	unsigned int		minrecs[2];
614 	unsigned int		blocklen;
615 
616 	blocklen = XFS_MIN_CRC_BLOCKSIZE - XFS_BTREE_SBLOCK_CRC_LEN;
617 
618 	minrecs[0] = xfs_inobt_block_maxrecs(blocklen, true) / 2;
619 	minrecs[1] = xfs_inobt_block_maxrecs(blocklen, false) / 2;
620 
621 	return xfs_btree_compute_maxlevels(minrecs, XFS_MAX_INODE_RECORDS);
622 }
623 
624 /* Compute the max possible height for either inode btree. */
625 unsigned int
626 xfs_iallocbt_maxlevels_ondisk(void)
627 {
628 	return max(xfs_inobt_maxlevels_ondisk(),
629 		   xfs_finobt_maxlevels_ondisk());
630 }
631 
632 /*
633  * Convert the inode record holemask to an inode allocation bitmap. The inode
634  * allocation bitmap is inode granularity and specifies whether an inode is
635  * physically allocated on disk (not whether the inode is considered allocated
636  * or free by the fs).
637  *
638  * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
639  */
640 uint64_t
641 xfs_inobt_irec_to_allocmask(
642 	const struct xfs_inobt_rec_incore	*rec)
643 {
644 	uint64_t			bitmap = 0;
645 	uint64_t			inodespbit;
646 	int				nextbit;
647 	uint				allocbitmap;
648 
649 	/*
650 	 * The holemask has 16-bits for a 64 inode record. Therefore each
651 	 * holemask bit represents multiple inodes. Create a mask of bits to set
652 	 * in the allocmask for each holemask bit.
653 	 */
654 	inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
655 
656 	/*
657 	 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
658 	 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
659 	 * anything beyond the 16 holemask bits since this casts to a larger
660 	 * type.
661 	 */
662 	allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
663 
664 	/*
665 	 * allocbitmap is the inverted holemask so every set bit represents
666 	 * allocated inodes. To expand from 16-bit holemask granularity to
667 	 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
668 	 * bitmap for every holemask bit.
669 	 */
670 	nextbit = xfs_next_bit(&allocbitmap, 1, 0);
671 	while (nextbit != -1) {
672 		ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
673 
674 		bitmap |= (inodespbit <<
675 			   (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
676 
677 		nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
678 	}
679 
680 	return bitmap;
681 }
682 
683 #if defined(DEBUG) || defined(XFS_WARN)
684 /*
685  * Verify that an in-core inode record has a valid inode count.
686  */
687 int
688 xfs_inobt_rec_check_count(
689 	struct xfs_mount		*mp,
690 	struct xfs_inobt_rec_incore	*rec)
691 {
692 	int				inocount = 0;
693 	int				nextbit = 0;
694 	uint64_t			allocbmap;
695 	int				wordsz;
696 
697 	wordsz = sizeof(allocbmap) / sizeof(unsigned int);
698 	allocbmap = xfs_inobt_irec_to_allocmask(rec);
699 
700 	nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
701 	while (nextbit != -1) {
702 		inocount++;
703 		nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
704 				       nextbit + 1);
705 	}
706 
707 	if (inocount != rec->ir_count)
708 		return -EFSCORRUPTED;
709 
710 	return 0;
711 }
712 #endif	/* DEBUG */
713 
714 static xfs_extlen_t
715 xfs_inobt_max_size(
716 	struct xfs_perag	*pag)
717 {
718 	struct xfs_mount	*mp = pag->pag_mount;
719 	xfs_agblock_t		agblocks = pag->block_count;
720 
721 	/* Bail out if we're uninitialized, which can happen in mkfs. */
722 	if (M_IGEO(mp)->inobt_mxr[0] == 0)
723 		return 0;
724 
725 	/*
726 	 * The log is permanently allocated, so the space it occupies will
727 	 * never be available for the kinds of things that would require btree
728 	 * expansion.  We therefore can pretend the space isn't there.
729 	 */
730 	if (xfs_ag_contains_log(mp, pag->pag_agno))
731 		agblocks -= mp->m_sb.sb_logblocks;
732 
733 	return xfs_btree_calc_size(M_IGEO(mp)->inobt_mnr,
734 				(uint64_t)agblocks * mp->m_sb.sb_inopblock /
735 					XFS_INODES_PER_CHUNK);
736 }
737 
738 static int
739 xfs_finobt_count_blocks(
740 	struct xfs_perag	*pag,
741 	struct xfs_trans	*tp,
742 	xfs_extlen_t		*tree_blocks)
743 {
744 	struct xfs_buf		*agbp = NULL;
745 	struct xfs_btree_cur	*cur;
746 	int			error;
747 
748 	error = xfs_ialloc_read_agi(pag, tp, 0, &agbp);
749 	if (error)
750 		return error;
751 
752 	cur = xfs_inobt_init_cursor(pag, tp, agbp);
753 	error = xfs_btree_count_blocks(cur, tree_blocks);
754 	xfs_btree_del_cursor(cur, error);
755 	xfs_trans_brelse(tp, agbp);
756 
757 	return error;
758 }
759 
760 /* Read finobt block count from AGI header. */
761 static int
762 xfs_finobt_read_blocks(
763 	struct xfs_perag	*pag,
764 	struct xfs_trans	*tp,
765 	xfs_extlen_t		*tree_blocks)
766 {
767 	struct xfs_buf		*agbp;
768 	struct xfs_agi		*agi;
769 	int			error;
770 
771 	error = xfs_ialloc_read_agi(pag, tp, 0, &agbp);
772 	if (error)
773 		return error;
774 
775 	agi = agbp->b_addr;
776 	*tree_blocks = be32_to_cpu(agi->agi_fblocks);
777 	xfs_trans_brelse(tp, agbp);
778 	return 0;
779 }
780 
781 /*
782  * Figure out how many blocks to reserve and how many are used by this btree.
783  */
784 int
785 xfs_finobt_calc_reserves(
786 	struct xfs_perag	*pag,
787 	struct xfs_trans	*tp,
788 	xfs_extlen_t		*ask,
789 	xfs_extlen_t		*used)
790 {
791 	xfs_extlen_t		tree_len = 0;
792 	int			error;
793 
794 	if (!xfs_has_finobt(pag->pag_mount))
795 		return 0;
796 
797 	if (xfs_has_inobtcounts(pag->pag_mount))
798 		error = xfs_finobt_read_blocks(pag, tp, &tree_len);
799 	else
800 		error = xfs_finobt_count_blocks(pag, tp, &tree_len);
801 	if (error)
802 		return error;
803 
804 	*ask += xfs_inobt_max_size(pag);
805 	*used += tree_len;
806 	return 0;
807 }
808 
809 /* Calculate the inobt btree size for some records. */
810 xfs_extlen_t
811 xfs_iallocbt_calc_size(
812 	struct xfs_mount	*mp,
813 	unsigned long long	len)
814 {
815 	return xfs_btree_calc_size(M_IGEO(mp)->inobt_mnr, len);
816 }
817 
818 int __init
819 xfs_inobt_init_cur_cache(void)
820 {
821 	xfs_inobt_cur_cache = kmem_cache_create("xfs_inobt_cur",
822 			xfs_btree_cur_sizeof(xfs_inobt_maxlevels_ondisk()),
823 			0, 0, NULL);
824 
825 	if (!xfs_inobt_cur_cache)
826 		return -ENOMEM;
827 	return 0;
828 }
829 
830 void
831 xfs_inobt_destroy_cur_cache(void)
832 {
833 	kmem_cache_destroy(xfs_inobt_cur_cache);
834 	xfs_inobt_cur_cache = NULL;
835 }
836