xref: /openbmc/linux/fs/xfs/libxfs/xfs_alloc_btree.c (revision 29c37341)
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_sb.h"
13 #include "xfs_mount.h"
14 #include "xfs_btree.h"
15 #include "xfs_btree_staging.h"
16 #include "xfs_alloc_btree.h"
17 #include "xfs_alloc.h"
18 #include "xfs_extent_busy.h"
19 #include "xfs_error.h"
20 #include "xfs_trace.h"
21 #include "xfs_trans.h"
22 
23 
24 STATIC struct xfs_btree_cur *
25 xfs_allocbt_dup_cursor(
26 	struct xfs_btree_cur	*cur)
27 {
28 	return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
29 			cur->bc_ag.agbp, cur->bc_ag.agno,
30 			cur->bc_btnum);
31 }
32 
33 STATIC void
34 xfs_allocbt_set_root(
35 	struct xfs_btree_cur	*cur,
36 	union xfs_btree_ptr	*ptr,
37 	int			inc)
38 {
39 	struct xfs_buf		*agbp = cur->bc_ag.agbp;
40 	struct xfs_agf		*agf = agbp->b_addr;
41 	int			btnum = cur->bc_btnum;
42 	struct xfs_perag	*pag = agbp->b_pag;
43 
44 	ASSERT(ptr->s != 0);
45 
46 	agf->agf_roots[btnum] = ptr->s;
47 	be32_add_cpu(&agf->agf_levels[btnum], inc);
48 	pag->pagf_levels[btnum] += inc;
49 
50 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
51 }
52 
53 STATIC int
54 xfs_allocbt_alloc_block(
55 	struct xfs_btree_cur	*cur,
56 	union xfs_btree_ptr	*start,
57 	union xfs_btree_ptr	*new,
58 	int			*stat)
59 {
60 	int			error;
61 	xfs_agblock_t		bno;
62 
63 	/* Allocate the new block from the freelist. If we can't, give up.  */
64 	error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_ag.agbp,
65 				       &bno, 1);
66 	if (error)
67 		return error;
68 
69 	if (bno == NULLAGBLOCK) {
70 		*stat = 0;
71 		return 0;
72 	}
73 
74 	xfs_extent_busy_reuse(cur->bc_mp, cur->bc_ag.agno, bno, 1, false);
75 
76 	xfs_trans_agbtree_delta(cur->bc_tp, 1);
77 	new->s = cpu_to_be32(bno);
78 
79 	*stat = 1;
80 	return 0;
81 }
82 
83 STATIC int
84 xfs_allocbt_free_block(
85 	struct xfs_btree_cur	*cur,
86 	struct xfs_buf		*bp)
87 {
88 	struct xfs_buf		*agbp = cur->bc_ag.agbp;
89 	struct xfs_agf		*agf = agbp->b_addr;
90 	xfs_agblock_t		bno;
91 	int			error;
92 
93 	bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
94 	error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
95 	if (error)
96 		return error;
97 
98 	xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
99 			      XFS_EXTENT_BUSY_SKIP_DISCARD);
100 	xfs_trans_agbtree_delta(cur->bc_tp, -1);
101 	return 0;
102 }
103 
104 /*
105  * Update the longest extent in the AGF
106  */
107 STATIC void
108 xfs_allocbt_update_lastrec(
109 	struct xfs_btree_cur	*cur,
110 	struct xfs_btree_block	*block,
111 	union xfs_btree_rec	*rec,
112 	int			ptr,
113 	int			reason)
114 {
115 	struct xfs_agf		*agf = cur->bc_ag.agbp->b_addr;
116 	struct xfs_perag	*pag;
117 	__be32			len;
118 	int			numrecs;
119 
120 	ASSERT(cur->bc_btnum == XFS_BTNUM_CNT);
121 
122 	switch (reason) {
123 	case LASTREC_UPDATE:
124 		/*
125 		 * If this is the last leaf block and it's the last record,
126 		 * then update the size of the longest extent in the AG.
127 		 */
128 		if (ptr != xfs_btree_get_numrecs(block))
129 			return;
130 		len = rec->alloc.ar_blockcount;
131 		break;
132 	case LASTREC_INSREC:
133 		if (be32_to_cpu(rec->alloc.ar_blockcount) <=
134 		    be32_to_cpu(agf->agf_longest))
135 			return;
136 		len = rec->alloc.ar_blockcount;
137 		break;
138 	case LASTREC_DELREC:
139 		numrecs = xfs_btree_get_numrecs(block);
140 		if (ptr <= numrecs)
141 			return;
142 		ASSERT(ptr == numrecs + 1);
143 
144 		if (numrecs) {
145 			xfs_alloc_rec_t *rrp;
146 
147 			rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs);
148 			len = rrp->ar_blockcount;
149 		} else {
150 			len = 0;
151 		}
152 
153 		break;
154 	default:
155 		ASSERT(0);
156 		return;
157 	}
158 
159 	agf->agf_longest = len;
160 	pag = cur->bc_ag.agbp->b_pag;
161 	pag->pagf_longest = be32_to_cpu(len);
162 	xfs_alloc_log_agf(cur->bc_tp, cur->bc_ag.agbp, XFS_AGF_LONGEST);
163 }
164 
165 STATIC int
166 xfs_allocbt_get_minrecs(
167 	struct xfs_btree_cur	*cur,
168 	int			level)
169 {
170 	return cur->bc_mp->m_alloc_mnr[level != 0];
171 }
172 
173 STATIC int
174 xfs_allocbt_get_maxrecs(
175 	struct xfs_btree_cur	*cur,
176 	int			level)
177 {
178 	return cur->bc_mp->m_alloc_mxr[level != 0];
179 }
180 
181 STATIC void
182 xfs_allocbt_init_key_from_rec(
183 	union xfs_btree_key	*key,
184 	union xfs_btree_rec	*rec)
185 {
186 	key->alloc.ar_startblock = rec->alloc.ar_startblock;
187 	key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
188 }
189 
190 STATIC void
191 xfs_bnobt_init_high_key_from_rec(
192 	union xfs_btree_key	*key,
193 	union xfs_btree_rec	*rec)
194 {
195 	__u32			x;
196 
197 	x = be32_to_cpu(rec->alloc.ar_startblock);
198 	x += be32_to_cpu(rec->alloc.ar_blockcount) - 1;
199 	key->alloc.ar_startblock = cpu_to_be32(x);
200 	key->alloc.ar_blockcount = 0;
201 }
202 
203 STATIC void
204 xfs_cntbt_init_high_key_from_rec(
205 	union xfs_btree_key	*key,
206 	union xfs_btree_rec	*rec)
207 {
208 	key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
209 	key->alloc.ar_startblock = 0;
210 }
211 
212 STATIC void
213 xfs_allocbt_init_rec_from_cur(
214 	struct xfs_btree_cur	*cur,
215 	union xfs_btree_rec	*rec)
216 {
217 	rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
218 	rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
219 }
220 
221 STATIC void
222 xfs_allocbt_init_ptr_from_cur(
223 	struct xfs_btree_cur	*cur,
224 	union xfs_btree_ptr	*ptr)
225 {
226 	struct xfs_agf		*agf = cur->bc_ag.agbp->b_addr;
227 
228 	ASSERT(cur->bc_ag.agno == be32_to_cpu(agf->agf_seqno));
229 
230 	ptr->s = agf->agf_roots[cur->bc_btnum];
231 }
232 
233 STATIC int64_t
234 xfs_bnobt_key_diff(
235 	struct xfs_btree_cur	*cur,
236 	union xfs_btree_key	*key)
237 {
238 	xfs_alloc_rec_incore_t	*rec = &cur->bc_rec.a;
239 	xfs_alloc_key_t		*kp = &key->alloc;
240 
241 	return (int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
242 }
243 
244 STATIC int64_t
245 xfs_cntbt_key_diff(
246 	struct xfs_btree_cur	*cur,
247 	union xfs_btree_key	*key)
248 {
249 	xfs_alloc_rec_incore_t	*rec = &cur->bc_rec.a;
250 	xfs_alloc_key_t		*kp = &key->alloc;
251 	int64_t			diff;
252 
253 	diff = (int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
254 	if (diff)
255 		return diff;
256 
257 	return (int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
258 }
259 
260 STATIC int64_t
261 xfs_bnobt_diff_two_keys(
262 	struct xfs_btree_cur	*cur,
263 	union xfs_btree_key	*k1,
264 	union xfs_btree_key	*k2)
265 {
266 	return (int64_t)be32_to_cpu(k1->alloc.ar_startblock) -
267 			  be32_to_cpu(k2->alloc.ar_startblock);
268 }
269 
270 STATIC int64_t
271 xfs_cntbt_diff_two_keys(
272 	struct xfs_btree_cur	*cur,
273 	union xfs_btree_key	*k1,
274 	union xfs_btree_key	*k2)
275 {
276 	int64_t			diff;
277 
278 	diff =  be32_to_cpu(k1->alloc.ar_blockcount) -
279 		be32_to_cpu(k2->alloc.ar_blockcount);
280 	if (diff)
281 		return diff;
282 
283 	return  be32_to_cpu(k1->alloc.ar_startblock) -
284 		be32_to_cpu(k2->alloc.ar_startblock);
285 }
286 
287 static xfs_failaddr_t
288 xfs_allocbt_verify(
289 	struct xfs_buf		*bp)
290 {
291 	struct xfs_mount	*mp = bp->b_mount;
292 	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
293 	struct xfs_perag	*pag = bp->b_pag;
294 	xfs_failaddr_t		fa;
295 	unsigned int		level;
296 	xfs_btnum_t		btnum = XFS_BTNUM_BNOi;
297 
298 	if (!xfs_verify_magic(bp, block->bb_magic))
299 		return __this_address;
300 
301 	if (xfs_sb_version_hascrc(&mp->m_sb)) {
302 		fa = xfs_btree_sblock_v5hdr_verify(bp);
303 		if (fa)
304 			return fa;
305 	}
306 
307 	/*
308 	 * The perag may not be attached during grow operations or fully
309 	 * initialized from the AGF during log recovery. Therefore we can only
310 	 * check against maximum tree depth from those contexts.
311 	 *
312 	 * Otherwise check against the per-tree limit. Peek at one of the
313 	 * verifier magic values to determine the type of tree we're verifying
314 	 * against.
315 	 */
316 	level = be16_to_cpu(block->bb_level);
317 	if (bp->b_ops->magic[0] == cpu_to_be32(XFS_ABTC_MAGIC))
318 		btnum = XFS_BTNUM_CNTi;
319 	if (pag && pag->pagf_init) {
320 		if (level >= pag->pagf_levels[btnum])
321 			return __this_address;
322 	} else if (level >= mp->m_ag_maxlevels)
323 		return __this_address;
324 
325 	return xfs_btree_sblock_verify(bp, mp->m_alloc_mxr[level != 0]);
326 }
327 
328 static void
329 xfs_allocbt_read_verify(
330 	struct xfs_buf	*bp)
331 {
332 	xfs_failaddr_t	fa;
333 
334 	if (!xfs_btree_sblock_verify_crc(bp))
335 		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
336 	else {
337 		fa = xfs_allocbt_verify(bp);
338 		if (fa)
339 			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
340 	}
341 
342 	if (bp->b_error)
343 		trace_xfs_btree_corrupt(bp, _RET_IP_);
344 }
345 
346 static void
347 xfs_allocbt_write_verify(
348 	struct xfs_buf	*bp)
349 {
350 	xfs_failaddr_t	fa;
351 
352 	fa = xfs_allocbt_verify(bp);
353 	if (fa) {
354 		trace_xfs_btree_corrupt(bp, _RET_IP_);
355 		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
356 		return;
357 	}
358 	xfs_btree_sblock_calc_crc(bp);
359 
360 }
361 
362 const struct xfs_buf_ops xfs_bnobt_buf_ops = {
363 	.name = "xfs_bnobt",
364 	.magic = { cpu_to_be32(XFS_ABTB_MAGIC),
365 		   cpu_to_be32(XFS_ABTB_CRC_MAGIC) },
366 	.verify_read = xfs_allocbt_read_verify,
367 	.verify_write = xfs_allocbt_write_verify,
368 	.verify_struct = xfs_allocbt_verify,
369 };
370 
371 const struct xfs_buf_ops xfs_cntbt_buf_ops = {
372 	.name = "xfs_cntbt",
373 	.magic = { cpu_to_be32(XFS_ABTC_MAGIC),
374 		   cpu_to_be32(XFS_ABTC_CRC_MAGIC) },
375 	.verify_read = xfs_allocbt_read_verify,
376 	.verify_write = xfs_allocbt_write_verify,
377 	.verify_struct = xfs_allocbt_verify,
378 };
379 
380 STATIC int
381 xfs_bnobt_keys_inorder(
382 	struct xfs_btree_cur	*cur,
383 	union xfs_btree_key	*k1,
384 	union xfs_btree_key	*k2)
385 {
386 	return be32_to_cpu(k1->alloc.ar_startblock) <
387 	       be32_to_cpu(k2->alloc.ar_startblock);
388 }
389 
390 STATIC int
391 xfs_bnobt_recs_inorder(
392 	struct xfs_btree_cur	*cur,
393 	union xfs_btree_rec	*r1,
394 	union xfs_btree_rec	*r2)
395 {
396 	return be32_to_cpu(r1->alloc.ar_startblock) +
397 		be32_to_cpu(r1->alloc.ar_blockcount) <=
398 		be32_to_cpu(r2->alloc.ar_startblock);
399 }
400 
401 STATIC int
402 xfs_cntbt_keys_inorder(
403 	struct xfs_btree_cur	*cur,
404 	union xfs_btree_key	*k1,
405 	union xfs_btree_key	*k2)
406 {
407 	return be32_to_cpu(k1->alloc.ar_blockcount) <
408 		be32_to_cpu(k2->alloc.ar_blockcount) ||
409 		(k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
410 		 be32_to_cpu(k1->alloc.ar_startblock) <
411 		 be32_to_cpu(k2->alloc.ar_startblock));
412 }
413 
414 STATIC int
415 xfs_cntbt_recs_inorder(
416 	struct xfs_btree_cur	*cur,
417 	union xfs_btree_rec	*r1,
418 	union xfs_btree_rec	*r2)
419 {
420 	return be32_to_cpu(r1->alloc.ar_blockcount) <
421 		be32_to_cpu(r2->alloc.ar_blockcount) ||
422 		(r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
423 		 be32_to_cpu(r1->alloc.ar_startblock) <
424 		 be32_to_cpu(r2->alloc.ar_startblock));
425 }
426 
427 static const struct xfs_btree_ops xfs_bnobt_ops = {
428 	.rec_len		= sizeof(xfs_alloc_rec_t),
429 	.key_len		= sizeof(xfs_alloc_key_t),
430 
431 	.dup_cursor		= xfs_allocbt_dup_cursor,
432 	.set_root		= xfs_allocbt_set_root,
433 	.alloc_block		= xfs_allocbt_alloc_block,
434 	.free_block		= xfs_allocbt_free_block,
435 	.update_lastrec		= xfs_allocbt_update_lastrec,
436 	.get_minrecs		= xfs_allocbt_get_minrecs,
437 	.get_maxrecs		= xfs_allocbt_get_maxrecs,
438 	.init_key_from_rec	= xfs_allocbt_init_key_from_rec,
439 	.init_high_key_from_rec	= xfs_bnobt_init_high_key_from_rec,
440 	.init_rec_from_cur	= xfs_allocbt_init_rec_from_cur,
441 	.init_ptr_from_cur	= xfs_allocbt_init_ptr_from_cur,
442 	.key_diff		= xfs_bnobt_key_diff,
443 	.buf_ops		= &xfs_bnobt_buf_ops,
444 	.diff_two_keys		= xfs_bnobt_diff_two_keys,
445 	.keys_inorder		= xfs_bnobt_keys_inorder,
446 	.recs_inorder		= xfs_bnobt_recs_inorder,
447 };
448 
449 static const struct xfs_btree_ops xfs_cntbt_ops = {
450 	.rec_len		= sizeof(xfs_alloc_rec_t),
451 	.key_len		= sizeof(xfs_alloc_key_t),
452 
453 	.dup_cursor		= xfs_allocbt_dup_cursor,
454 	.set_root		= xfs_allocbt_set_root,
455 	.alloc_block		= xfs_allocbt_alloc_block,
456 	.free_block		= xfs_allocbt_free_block,
457 	.update_lastrec		= xfs_allocbt_update_lastrec,
458 	.get_minrecs		= xfs_allocbt_get_minrecs,
459 	.get_maxrecs		= xfs_allocbt_get_maxrecs,
460 	.init_key_from_rec	= xfs_allocbt_init_key_from_rec,
461 	.init_high_key_from_rec	= xfs_cntbt_init_high_key_from_rec,
462 	.init_rec_from_cur	= xfs_allocbt_init_rec_from_cur,
463 	.init_ptr_from_cur	= xfs_allocbt_init_ptr_from_cur,
464 	.key_diff		= xfs_cntbt_key_diff,
465 	.buf_ops		= &xfs_cntbt_buf_ops,
466 	.diff_two_keys		= xfs_cntbt_diff_two_keys,
467 	.keys_inorder		= xfs_cntbt_keys_inorder,
468 	.recs_inorder		= xfs_cntbt_recs_inorder,
469 };
470 
471 /* Allocate most of a new allocation btree cursor. */
472 STATIC struct xfs_btree_cur *
473 xfs_allocbt_init_common(
474 	struct xfs_mount	*mp,
475 	struct xfs_trans	*tp,
476 	xfs_agnumber_t		agno,
477 	xfs_btnum_t		btnum)
478 {
479 	struct xfs_btree_cur	*cur;
480 
481 	ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT);
482 
483 	cur = kmem_cache_zalloc(xfs_btree_cur_zone, GFP_NOFS | __GFP_NOFAIL);
484 
485 	cur->bc_tp = tp;
486 	cur->bc_mp = mp;
487 	cur->bc_btnum = btnum;
488 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
489 
490 	if (btnum == XFS_BTNUM_CNT) {
491 		cur->bc_ops = &xfs_cntbt_ops;
492 		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtc_2);
493 		cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
494 	} else {
495 		cur->bc_ops = &xfs_bnobt_ops;
496 		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtb_2);
497 	}
498 
499 	cur->bc_ag.agno = agno;
500 	cur->bc_ag.abt.active = false;
501 
502 	if (xfs_sb_version_hascrc(&mp->m_sb))
503 		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
504 
505 	return cur;
506 }
507 
508 /*
509  * Allocate a new allocation btree cursor.
510  */
511 struct xfs_btree_cur *			/* new alloc btree cursor */
512 xfs_allocbt_init_cursor(
513 	struct xfs_mount	*mp,		/* file system mount point */
514 	struct xfs_trans	*tp,		/* transaction pointer */
515 	struct xfs_buf		*agbp,		/* buffer for agf structure */
516 	xfs_agnumber_t		agno,		/* allocation group number */
517 	xfs_btnum_t		btnum)		/* btree identifier */
518 {
519 	struct xfs_agf		*agf = agbp->b_addr;
520 	struct xfs_btree_cur	*cur;
521 
522 	cur = xfs_allocbt_init_common(mp, tp, agno, btnum);
523 	if (btnum == XFS_BTNUM_CNT)
524 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);
525 	else
526 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]);
527 
528 	cur->bc_ag.agbp = agbp;
529 
530 	return cur;
531 }
532 
533 /* Create a free space btree cursor with a fake root for staging. */
534 struct xfs_btree_cur *
535 xfs_allocbt_stage_cursor(
536 	struct xfs_mount	*mp,
537 	struct xbtree_afakeroot	*afake,
538 	xfs_agnumber_t		agno,
539 	xfs_btnum_t		btnum)
540 {
541 	struct xfs_btree_cur	*cur;
542 
543 	cur = xfs_allocbt_init_common(mp, NULL, agno, btnum);
544 	xfs_btree_stage_afakeroot(cur, afake);
545 	return cur;
546 }
547 
548 /*
549  * Install a new free space btree root.  Caller is responsible for invalidating
550  * and freeing the old btree blocks.
551  */
552 void
553 xfs_allocbt_commit_staged_btree(
554 	struct xfs_btree_cur	*cur,
555 	struct xfs_trans	*tp,
556 	struct xfs_buf		*agbp)
557 {
558 	struct xfs_agf		*agf = agbp->b_addr;
559 	struct xbtree_afakeroot	*afake = cur->bc_ag.afake;
560 
561 	ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
562 
563 	agf->agf_roots[cur->bc_btnum] = cpu_to_be32(afake->af_root);
564 	agf->agf_levels[cur->bc_btnum] = cpu_to_be32(afake->af_levels);
565 	xfs_alloc_log_agf(tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
566 
567 	if (cur->bc_btnum == XFS_BTNUM_BNO) {
568 		xfs_btree_commit_afakeroot(cur, tp, agbp, &xfs_bnobt_ops);
569 	} else {
570 		cur->bc_flags |= XFS_BTREE_LASTREC_UPDATE;
571 		xfs_btree_commit_afakeroot(cur, tp, agbp, &xfs_cntbt_ops);
572 	}
573 }
574 
575 /*
576  * Calculate number of records in an alloc btree block.
577  */
578 int
579 xfs_allocbt_maxrecs(
580 	struct xfs_mount	*mp,
581 	int			blocklen,
582 	int			leaf)
583 {
584 	blocklen -= XFS_ALLOC_BLOCK_LEN(mp);
585 
586 	if (leaf)
587 		return blocklen / sizeof(xfs_alloc_rec_t);
588 	return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
589 }
590 
591 /* Calculate the freespace btree size for some records. */
592 xfs_extlen_t
593 xfs_allocbt_calc_size(
594 	struct xfs_mount	*mp,
595 	unsigned long long	len)
596 {
597 	return xfs_btree_calc_size(mp->m_alloc_mnr, len);
598 }
599