xref: /openbmc/linux/fs/xfs/libxfs/xfs_ialloc_btree.c (revision 4da722ca19f30f7db250db808d1ab1703607a932) 
1 /*
2  * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public License as
7  * published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it would be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  * GNU General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write the Free Software Foundation,
16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17  */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_bit.h"
25 #include "xfs_mount.h"
26 #include "xfs_inode.h"
27 #include "xfs_btree.h"
28 #include "xfs_ialloc.h"
29 #include "xfs_ialloc_btree.h"
30 #include "xfs_alloc.h"
31 #include "xfs_error.h"
32 #include "xfs_trace.h"
33 #include "xfs_cksum.h"
34 #include "xfs_trans.h"
35 #include "xfs_rmap.h"
36 
37 
38 STATIC int
39 xfs_inobt_get_minrecs(
40 	struct xfs_btree_cur	*cur,
41 	int			level)
42 {
43 	return cur->bc_mp->m_inobt_mnr[level != 0];
44 }
45 
46 STATIC struct xfs_btree_cur *
47 xfs_inobt_dup_cursor(
48 	struct xfs_btree_cur	*cur)
49 {
50 	return xfs_inobt_init_cursor(cur->bc_mp, cur->bc_tp,
51 			cur->bc_private.a.agbp, cur->bc_private.a.agno,
52 			cur->bc_btnum);
53 }
54 
55 STATIC void
56 xfs_inobt_set_root(
57 	struct xfs_btree_cur	*cur,
58 	union xfs_btree_ptr	*nptr,
59 	int			inc)	/* level change */
60 {
61 	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
62 	struct xfs_agi		*agi = XFS_BUF_TO_AGI(agbp);
63 
64 	agi->agi_root = nptr->s;
65 	be32_add_cpu(&agi->agi_level, inc);
66 	xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_ROOT | XFS_AGI_LEVEL);
67 }
68 
69 STATIC void
70 xfs_finobt_set_root(
71 	struct xfs_btree_cur	*cur,
72 	union xfs_btree_ptr	*nptr,
73 	int			inc)	/* level change */
74 {
75 	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
76 	struct xfs_agi		*agi = XFS_BUF_TO_AGI(agbp);
77 
78 	agi->agi_free_root = nptr->s;
79 	be32_add_cpu(&agi->agi_free_level, inc);
80 	xfs_ialloc_log_agi(cur->bc_tp, agbp,
81 			   XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL);
82 }
83 
84 STATIC int
85 __xfs_inobt_alloc_block(
86 	struct xfs_btree_cur	*cur,
87 	union xfs_btree_ptr	*start,
88 	union xfs_btree_ptr	*new,
89 	int			*stat,
90 	enum xfs_ag_resv_type	resv)
91 {
92 	xfs_alloc_arg_t		args;		/* block allocation args */
93 	int			error;		/* error return value */
94 	xfs_agblock_t		sbno = be32_to_cpu(start->s);
95 
96 	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
97 
98 	memset(&args, 0, sizeof(args));
99 	args.tp = cur->bc_tp;
100 	args.mp = cur->bc_mp;
101 	xfs_rmap_ag_owner(&args.oinfo, XFS_RMAP_OWN_INOBT);
102 	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno, sbno);
103 	args.minlen = 1;
104 	args.maxlen = 1;
105 	args.prod = 1;
106 	args.type = XFS_ALLOCTYPE_NEAR_BNO;
107 	args.resv = resv;
108 
109 	error = xfs_alloc_vextent(&args);
110 	if (error) {
111 		XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
112 		return error;
113 	}
114 	if (args.fsbno == NULLFSBLOCK) {
115 		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
116 		*stat = 0;
117 		return 0;
118 	}
119 	ASSERT(args.len == 1);
120 	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
121 
122 	new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
123 	*stat = 1;
124 	return 0;
125 }
126 
127 STATIC int
128 xfs_inobt_alloc_block(
129 	struct xfs_btree_cur	*cur,
130 	union xfs_btree_ptr	*start,
131 	union xfs_btree_ptr	*new,
132 	int			*stat)
133 {
134 	return __xfs_inobt_alloc_block(cur, start, new, stat, XFS_AG_RESV_NONE);
135 }
136 
137 STATIC int
138 xfs_finobt_alloc_block(
139 	struct xfs_btree_cur	*cur,
140 	union xfs_btree_ptr	*start,
141 	union xfs_btree_ptr	*new,
142 	int			*stat)
143 {
144 	return __xfs_inobt_alloc_block(cur, start, new, stat,
145 			XFS_AG_RESV_METADATA);
146 }
147 
148 STATIC int
149 xfs_inobt_free_block(
150 	struct xfs_btree_cur	*cur,
151 	struct xfs_buf		*bp)
152 {
153 	struct xfs_owner_info	oinfo;
154 
155 	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INOBT);
156 	return xfs_free_extent(cur->bc_tp,
157 			XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp)), 1,
158 			&oinfo, XFS_AG_RESV_NONE);
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 int
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 	unsigned int		level;
260 
261 	/*
262 	 * During growfs operations, we can't verify the exact owner as the
263 	 * perag is not fully initialised and hence not attached to the buffer.
264 	 *
265 	 * Similarly, during log recovery we will have a perag structure
266 	 * attached, but the agi information will not yet have been initialised
267 	 * from the on disk AGI. We don't currently use any of this information,
268 	 * but beware of the landmine (i.e. need to check pag->pagi_init) if we
269 	 * ever do.
270 	 */
271 	switch (block->bb_magic) {
272 	case cpu_to_be32(XFS_IBT_CRC_MAGIC):
273 	case cpu_to_be32(XFS_FIBT_CRC_MAGIC):
274 		if (!xfs_btree_sblock_v5hdr_verify(bp))
275 			return false;
276 		/* fall through */
277 	case cpu_to_be32(XFS_IBT_MAGIC):
278 	case cpu_to_be32(XFS_FIBT_MAGIC):
279 		break;
280 	default:
281 		return 0;
282 	}
283 
284 	/* level verification */
285 	level = be16_to_cpu(block->bb_level);
286 	if (level >= mp->m_in_maxlevels)
287 		return false;
288 
289 	return xfs_btree_sblock_verify(bp, mp->m_inobt_mxr[level != 0]);
290 }
291 
292 static void
293 xfs_inobt_read_verify(
294 	struct xfs_buf	*bp)
295 {
296 	if (!xfs_btree_sblock_verify_crc(bp))
297 		xfs_buf_ioerror(bp, -EFSBADCRC);
298 	else if (!xfs_inobt_verify(bp))
299 		xfs_buf_ioerror(bp, -EFSCORRUPTED);
300 
301 	if (bp->b_error) {
302 		trace_xfs_btree_corrupt(bp, _RET_IP_);
303 		xfs_verifier_error(bp);
304 	}
305 }
306 
307 static void
308 xfs_inobt_write_verify(
309 	struct xfs_buf	*bp)
310 {
311 	if (!xfs_inobt_verify(bp)) {
312 		trace_xfs_btree_corrupt(bp, _RET_IP_);
313 		xfs_buf_ioerror(bp, -EFSCORRUPTED);
314 		xfs_verifier_error(bp);
315 		return;
316 	}
317 	xfs_btree_sblock_calc_crc(bp);
318 
319 }
320 
321 const struct xfs_buf_ops xfs_inobt_buf_ops = {
322 	.name = "xfs_inobt",
323 	.verify_read = xfs_inobt_read_verify,
324 	.verify_write = xfs_inobt_write_verify,
325 };
326 
327 STATIC int
328 xfs_inobt_keys_inorder(
329 	struct xfs_btree_cur	*cur,
330 	union xfs_btree_key	*k1,
331 	union xfs_btree_key	*k2)
332 {
333 	return be32_to_cpu(k1->inobt.ir_startino) <
334 		be32_to_cpu(k2->inobt.ir_startino);
335 }
336 
337 STATIC int
338 xfs_inobt_recs_inorder(
339 	struct xfs_btree_cur	*cur,
340 	union xfs_btree_rec	*r1,
341 	union xfs_btree_rec	*r2)
342 {
343 	return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
344 		be32_to_cpu(r2->inobt.ir_startino);
345 }
346 
347 static const struct xfs_btree_ops xfs_inobt_ops = {
348 	.rec_len		= sizeof(xfs_inobt_rec_t),
349 	.key_len		= sizeof(xfs_inobt_key_t),
350 
351 	.dup_cursor		= xfs_inobt_dup_cursor,
352 	.set_root		= xfs_inobt_set_root,
353 	.alloc_block		= xfs_inobt_alloc_block,
354 	.free_block		= xfs_inobt_free_block,
355 	.get_minrecs		= xfs_inobt_get_minrecs,
356 	.get_maxrecs		= xfs_inobt_get_maxrecs,
357 	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
358 	.init_high_key_from_rec	= xfs_inobt_init_high_key_from_rec,
359 	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
360 	.init_ptr_from_cur	= xfs_inobt_init_ptr_from_cur,
361 	.key_diff		= xfs_inobt_key_diff,
362 	.buf_ops		= &xfs_inobt_buf_ops,
363 	.diff_two_keys		= xfs_inobt_diff_two_keys,
364 	.keys_inorder		= xfs_inobt_keys_inorder,
365 	.recs_inorder		= xfs_inobt_recs_inorder,
366 };
367 
368 static const struct xfs_btree_ops xfs_finobt_ops = {
369 	.rec_len		= sizeof(xfs_inobt_rec_t),
370 	.key_len		= sizeof(xfs_inobt_key_t),
371 
372 	.dup_cursor		= xfs_inobt_dup_cursor,
373 	.set_root		= xfs_finobt_set_root,
374 	.alloc_block		= xfs_finobt_alloc_block,
375 	.free_block		= xfs_inobt_free_block,
376 	.get_minrecs		= xfs_inobt_get_minrecs,
377 	.get_maxrecs		= xfs_inobt_get_maxrecs,
378 	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
379 	.init_high_key_from_rec	= xfs_inobt_init_high_key_from_rec,
380 	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
381 	.init_ptr_from_cur	= xfs_finobt_init_ptr_from_cur,
382 	.key_diff		= xfs_inobt_key_diff,
383 	.buf_ops		= &xfs_inobt_buf_ops,
384 	.diff_two_keys		= xfs_inobt_diff_two_keys,
385 	.keys_inorder		= xfs_inobt_keys_inorder,
386 	.recs_inorder		= xfs_inobt_recs_inorder,
387 };
388 
389 /*
390  * Allocate a new inode btree cursor.
391  */
392 struct xfs_btree_cur *				/* new inode btree cursor */
393 xfs_inobt_init_cursor(
394 	struct xfs_mount	*mp,		/* file system mount point */
395 	struct xfs_trans	*tp,		/* transaction pointer */
396 	struct xfs_buf		*agbp,		/* buffer for agi structure */
397 	xfs_agnumber_t		agno,		/* allocation group number */
398 	xfs_btnum_t		btnum)		/* ialloc or free ino btree */
399 {
400 	struct xfs_agi		*agi = XFS_BUF_TO_AGI(agbp);
401 	struct xfs_btree_cur	*cur;
402 
403 	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
404 
405 	cur->bc_tp = tp;
406 	cur->bc_mp = mp;
407 	cur->bc_btnum = btnum;
408 	if (btnum == XFS_BTNUM_INO) {
409 		cur->bc_nlevels = be32_to_cpu(agi->agi_level);
410 		cur->bc_ops = &xfs_inobt_ops;
411 		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_ibt_2);
412 	} else {
413 		cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
414 		cur->bc_ops = &xfs_finobt_ops;
415 		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_fibt_2);
416 	}
417 
418 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
419 
420 	if (xfs_sb_version_hascrc(&mp->m_sb))
421 		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
422 
423 	cur->bc_private.a.agbp = agbp;
424 	cur->bc_private.a.agno = agno;
425 
426 	return cur;
427 }
428 
429 /*
430  * Calculate number of records in an inobt btree block.
431  */
432 int
433 xfs_inobt_maxrecs(
434 	struct xfs_mount	*mp,
435 	int			blocklen,
436 	int			leaf)
437 {
438 	blocklen -= XFS_INOBT_BLOCK_LEN(mp);
439 
440 	if (leaf)
441 		return blocklen / sizeof(xfs_inobt_rec_t);
442 	return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
443 }
444 
445 /*
446  * Convert the inode record holemask to an inode allocation bitmap. The inode
447  * allocation bitmap is inode granularity and specifies whether an inode is
448  * physically allocated on disk (not whether the inode is considered allocated
449  * or free by the fs).
450  *
451  * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
452  */
453 uint64_t
454 xfs_inobt_irec_to_allocmask(
455 	struct xfs_inobt_rec_incore	*rec)
456 {
457 	uint64_t			bitmap = 0;
458 	uint64_t			inodespbit;
459 	int				nextbit;
460 	uint				allocbitmap;
461 
462 	/*
463 	 * The holemask has 16-bits for a 64 inode record. Therefore each
464 	 * holemask bit represents multiple inodes. Create a mask of bits to set
465 	 * in the allocmask for each holemask bit.
466 	 */
467 	inodespbit = (1 << XFS_INODES_PER_HOLEMASK_BIT) - 1;
468 
469 	/*
470 	 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
471 	 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
472 	 * anything beyond the 16 holemask bits since this casts to a larger
473 	 * type.
474 	 */
475 	allocbitmap = ~rec->ir_holemask & ((1 << XFS_INOBT_HOLEMASK_BITS) - 1);
476 
477 	/*
478 	 * allocbitmap is the inverted holemask so every set bit represents
479 	 * allocated inodes. To expand from 16-bit holemask granularity to
480 	 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
481 	 * bitmap for every holemask bit.
482 	 */
483 	nextbit = xfs_next_bit(&allocbitmap, 1, 0);
484 	while (nextbit != -1) {
485 		ASSERT(nextbit < (sizeof(rec->ir_holemask) * NBBY));
486 
487 		bitmap |= (inodespbit <<
488 			   (nextbit * XFS_INODES_PER_HOLEMASK_BIT));
489 
490 		nextbit = xfs_next_bit(&allocbitmap, 1, nextbit + 1);
491 	}
492 
493 	return bitmap;
494 }
495 
496 #if defined(DEBUG) || defined(XFS_WARN)
497 /*
498  * Verify that an in-core inode record has a valid inode count.
499  */
500 int
501 xfs_inobt_rec_check_count(
502 	struct xfs_mount		*mp,
503 	struct xfs_inobt_rec_incore	*rec)
504 {
505 	int				inocount = 0;
506 	int				nextbit = 0;
507 	uint64_t			allocbmap;
508 	int				wordsz;
509 
510 	wordsz = sizeof(allocbmap) / sizeof(unsigned int);
511 	allocbmap = xfs_inobt_irec_to_allocmask(rec);
512 
513 	nextbit = xfs_next_bit((uint *) &allocbmap, wordsz, nextbit);
514 	while (nextbit != -1) {
515 		inocount++;
516 		nextbit = xfs_next_bit((uint *) &allocbmap, wordsz,
517 				       nextbit + 1);
518 	}
519 
520 	if (inocount != rec->ir_count)
521 		return -EFSCORRUPTED;
522 
523 	return 0;
524 }
525 #endif	/* DEBUG */
526 
527 static xfs_extlen_t
528 xfs_inobt_max_size(
529 	struct xfs_mount	*mp)
530 {
531 	/* Bail out if we're uninitialized, which can happen in mkfs. */
532 	if (mp->m_inobt_mxr[0] == 0)
533 		return 0;
534 
535 	return xfs_btree_calc_size(mp, mp->m_inobt_mnr,
536 		(uint64_t)mp->m_sb.sb_agblocks * mp->m_sb.sb_inopblock /
537 				XFS_INODES_PER_CHUNK);
538 }
539 
540 static int
541 xfs_inobt_count_blocks(
542 	struct xfs_mount	*mp,
543 	xfs_agnumber_t		agno,
544 	xfs_btnum_t		btnum,
545 	xfs_extlen_t		*tree_blocks)
546 {
547 	struct xfs_buf		*agbp;
548 	struct xfs_btree_cur	*cur;
549 	int			error;
550 
551 	error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
552 	if (error)
553 		return error;
554 
555 	cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, btnum);
556 	error = xfs_btree_count_blocks(cur, tree_blocks);
557 	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
558 	xfs_buf_relse(agbp);
559 
560 	return error;
561 }
562 
563 /*
564  * Figure out how many blocks to reserve and how many are used by this btree.
565  */
566 int
567 xfs_finobt_calc_reserves(
568 	struct xfs_mount	*mp,
569 	xfs_agnumber_t		agno,
570 	xfs_extlen_t		*ask,
571 	xfs_extlen_t		*used)
572 {
573 	xfs_extlen_t		tree_len = 0;
574 	int			error;
575 
576 	if (!xfs_sb_version_hasfinobt(&mp->m_sb))
577 		return 0;
578 
579 	error = xfs_inobt_count_blocks(mp, agno, XFS_BTNUM_FINO, &tree_len);
580 	if (error)
581 		return error;
582 
583 	*ask += xfs_inobt_max_size(mp);
584 	*used += tree_len;
585 	return 0;
586 }
587