xref: /openbmc/linux/fs/xfs/libxfs/xfs_alloc_btree.c (revision 06ed5c2b)
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_sb.h"
25 #include "xfs_ag.h"
26 #include "xfs_mount.h"
27 #include "xfs_btree.h"
28 #include "xfs_alloc_btree.h"
29 #include "xfs_alloc.h"
30 #include "xfs_extent_busy.h"
31 #include "xfs_error.h"
32 #include "xfs_trace.h"
33 #include "xfs_cksum.h"
34 #include "xfs_trans.h"
35 
36 
37 STATIC struct xfs_btree_cur *
38 xfs_allocbt_dup_cursor(
39 	struct xfs_btree_cur	*cur)
40 {
41 	return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
42 			cur->bc_private.a.agbp, cur->bc_private.a.agno,
43 			cur->bc_btnum);
44 }
45 
46 STATIC void
47 xfs_allocbt_set_root(
48 	struct xfs_btree_cur	*cur,
49 	union xfs_btree_ptr	*ptr,
50 	int			inc)
51 {
52 	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
53 	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
54 	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
55 	int			btnum = cur->bc_btnum;
56 	struct xfs_perag	*pag = xfs_perag_get(cur->bc_mp, seqno);
57 
58 	ASSERT(ptr->s != 0);
59 
60 	agf->agf_roots[btnum] = ptr->s;
61 	be32_add_cpu(&agf->agf_levels[btnum], inc);
62 	pag->pagf_levels[btnum] += inc;
63 	xfs_perag_put(pag);
64 
65 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
66 }
67 
68 STATIC int
69 xfs_allocbt_alloc_block(
70 	struct xfs_btree_cur	*cur,
71 	union xfs_btree_ptr	*start,
72 	union xfs_btree_ptr	*new,
73 	int			*stat)
74 {
75 	int			error;
76 	xfs_agblock_t		bno;
77 
78 	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
79 
80 	/* Allocate the new block from the freelist. If we can't, give up.  */
81 	error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_private.a.agbp,
82 				       &bno, 1);
83 	if (error) {
84 		XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
85 		return error;
86 	}
87 
88 	if (bno == NULLAGBLOCK) {
89 		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
90 		*stat = 0;
91 		return 0;
92 	}
93 
94 	xfs_extent_busy_reuse(cur->bc_mp, cur->bc_private.a.agno, bno, 1, false);
95 
96 	xfs_trans_agbtree_delta(cur->bc_tp, 1);
97 	new->s = cpu_to_be32(bno);
98 
99 	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
100 	*stat = 1;
101 	return 0;
102 }
103 
104 STATIC int
105 xfs_allocbt_free_block(
106 	struct xfs_btree_cur	*cur,
107 	struct xfs_buf		*bp)
108 {
109 	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
110 	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
111 	xfs_agblock_t		bno;
112 	int			error;
113 
114 	bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
115 	error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
116 	if (error)
117 		return error;
118 
119 	xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
120 			      XFS_EXTENT_BUSY_SKIP_DISCARD);
121 	xfs_trans_agbtree_delta(cur->bc_tp, -1);
122 
123 	xfs_trans_binval(cur->bc_tp, bp);
124 	return 0;
125 }
126 
127 /*
128  * Update the longest extent in the AGF
129  */
130 STATIC void
131 xfs_allocbt_update_lastrec(
132 	struct xfs_btree_cur	*cur,
133 	struct xfs_btree_block	*block,
134 	union xfs_btree_rec	*rec,
135 	int			ptr,
136 	int			reason)
137 {
138 	struct xfs_agf		*agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
139 	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
140 	struct xfs_perag	*pag;
141 	__be32			len;
142 	int			numrecs;
143 
144 	ASSERT(cur->bc_btnum == XFS_BTNUM_CNT);
145 
146 	switch (reason) {
147 	case LASTREC_UPDATE:
148 		/*
149 		 * If this is the last leaf block and it's the last record,
150 		 * then update the size of the longest extent in the AG.
151 		 */
152 		if (ptr != xfs_btree_get_numrecs(block))
153 			return;
154 		len = rec->alloc.ar_blockcount;
155 		break;
156 	case LASTREC_INSREC:
157 		if (be32_to_cpu(rec->alloc.ar_blockcount) <=
158 		    be32_to_cpu(agf->agf_longest))
159 			return;
160 		len = rec->alloc.ar_blockcount;
161 		break;
162 	case LASTREC_DELREC:
163 		numrecs = xfs_btree_get_numrecs(block);
164 		if (ptr <= numrecs)
165 			return;
166 		ASSERT(ptr == numrecs + 1);
167 
168 		if (numrecs) {
169 			xfs_alloc_rec_t *rrp;
170 
171 			rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs);
172 			len = rrp->ar_blockcount;
173 		} else {
174 			len = 0;
175 		}
176 
177 		break;
178 	default:
179 		ASSERT(0);
180 		return;
181 	}
182 
183 	agf->agf_longest = len;
184 	pag = xfs_perag_get(cur->bc_mp, seqno);
185 	pag->pagf_longest = be32_to_cpu(len);
186 	xfs_perag_put(pag);
187 	xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp, XFS_AGF_LONGEST);
188 }
189 
190 STATIC int
191 xfs_allocbt_get_minrecs(
192 	struct xfs_btree_cur	*cur,
193 	int			level)
194 {
195 	return cur->bc_mp->m_alloc_mnr[level != 0];
196 }
197 
198 STATIC int
199 xfs_allocbt_get_maxrecs(
200 	struct xfs_btree_cur	*cur,
201 	int			level)
202 {
203 	return cur->bc_mp->m_alloc_mxr[level != 0];
204 }
205 
206 STATIC void
207 xfs_allocbt_init_key_from_rec(
208 	union xfs_btree_key	*key,
209 	union xfs_btree_rec	*rec)
210 {
211 	ASSERT(rec->alloc.ar_startblock != 0);
212 
213 	key->alloc.ar_startblock = rec->alloc.ar_startblock;
214 	key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
215 }
216 
217 STATIC void
218 xfs_allocbt_init_rec_from_key(
219 	union xfs_btree_key	*key,
220 	union xfs_btree_rec	*rec)
221 {
222 	ASSERT(key->alloc.ar_startblock != 0);
223 
224 	rec->alloc.ar_startblock = key->alloc.ar_startblock;
225 	rec->alloc.ar_blockcount = key->alloc.ar_blockcount;
226 }
227 
228 STATIC void
229 xfs_allocbt_init_rec_from_cur(
230 	struct xfs_btree_cur	*cur,
231 	union xfs_btree_rec	*rec)
232 {
233 	ASSERT(cur->bc_rec.a.ar_startblock != 0);
234 
235 	rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
236 	rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
237 }
238 
239 STATIC void
240 xfs_allocbt_init_ptr_from_cur(
241 	struct xfs_btree_cur	*cur,
242 	union xfs_btree_ptr	*ptr)
243 {
244 	struct xfs_agf		*agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
245 
246 	ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno));
247 	ASSERT(agf->agf_roots[cur->bc_btnum] != 0);
248 
249 	ptr->s = agf->agf_roots[cur->bc_btnum];
250 }
251 
252 STATIC __int64_t
253 xfs_allocbt_key_diff(
254 	struct xfs_btree_cur	*cur,
255 	union xfs_btree_key	*key)
256 {
257 	xfs_alloc_rec_incore_t	*rec = &cur->bc_rec.a;
258 	xfs_alloc_key_t		*kp = &key->alloc;
259 	__int64_t		diff;
260 
261 	if (cur->bc_btnum == XFS_BTNUM_BNO) {
262 		return (__int64_t)be32_to_cpu(kp->ar_startblock) -
263 				rec->ar_startblock;
264 	}
265 
266 	diff = (__int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
267 	if (diff)
268 		return diff;
269 
270 	return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
271 }
272 
273 static bool
274 xfs_allocbt_verify(
275 	struct xfs_buf		*bp)
276 {
277 	struct xfs_mount	*mp = bp->b_target->bt_mount;
278 	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
279 	struct xfs_perag	*pag = bp->b_pag;
280 	unsigned int		level;
281 
282 	/*
283 	 * magic number and level verification
284 	 *
285 	 * During growfs operations, we can't verify the exact level or owner as
286 	 * the perag is not fully initialised and hence not attached to the
287 	 * buffer.  In this case, check against the maximum tree depth.
288 	 *
289 	 * Similarly, during log recovery we will have a perag structure
290 	 * attached, but the agf information will not yet have been initialised
291 	 * from the on disk AGF. Again, we can only check against maximum limits
292 	 * in this case.
293 	 */
294 	level = be16_to_cpu(block->bb_level);
295 	switch (block->bb_magic) {
296 	case cpu_to_be32(XFS_ABTB_CRC_MAGIC):
297 		if (!xfs_sb_version_hascrc(&mp->m_sb))
298 			return false;
299 		if (!uuid_equal(&block->bb_u.s.bb_uuid, &mp->m_sb.sb_uuid))
300 			return false;
301 		if (block->bb_u.s.bb_blkno != cpu_to_be64(bp->b_bn))
302 			return false;
303 		if (pag &&
304 		    be32_to_cpu(block->bb_u.s.bb_owner) != pag->pag_agno)
305 			return false;
306 		/* fall through */
307 	case cpu_to_be32(XFS_ABTB_MAGIC):
308 		if (pag && pag->pagf_init) {
309 			if (level >= pag->pagf_levels[XFS_BTNUM_BNOi])
310 				return false;
311 		} else if (level >= mp->m_ag_maxlevels)
312 			return false;
313 		break;
314 	case cpu_to_be32(XFS_ABTC_CRC_MAGIC):
315 		if (!xfs_sb_version_hascrc(&mp->m_sb))
316 			return false;
317 		if (!uuid_equal(&block->bb_u.s.bb_uuid, &mp->m_sb.sb_uuid))
318 			return false;
319 		if (block->bb_u.s.bb_blkno != cpu_to_be64(bp->b_bn))
320 			return false;
321 		if (pag &&
322 		    be32_to_cpu(block->bb_u.s.bb_owner) != pag->pag_agno)
323 			return false;
324 		/* fall through */
325 	case cpu_to_be32(XFS_ABTC_MAGIC):
326 		if (pag && pag->pagf_init) {
327 			if (level >= pag->pagf_levels[XFS_BTNUM_CNTi])
328 				return false;
329 		} else if (level >= mp->m_ag_maxlevels)
330 			return false;
331 		break;
332 	default:
333 		return false;
334 	}
335 
336 	/* numrecs verification */
337 	if (be16_to_cpu(block->bb_numrecs) > mp->m_alloc_mxr[level != 0])
338 		return false;
339 
340 	/* sibling pointer verification */
341 	if (!block->bb_u.s.bb_leftsib ||
342 	    (be32_to_cpu(block->bb_u.s.bb_leftsib) >= mp->m_sb.sb_agblocks &&
343 	     block->bb_u.s.bb_leftsib != cpu_to_be32(NULLAGBLOCK)))
344 		return false;
345 	if (!block->bb_u.s.bb_rightsib ||
346 	    (be32_to_cpu(block->bb_u.s.bb_rightsib) >= mp->m_sb.sb_agblocks &&
347 	     block->bb_u.s.bb_rightsib != cpu_to_be32(NULLAGBLOCK)))
348 		return false;
349 
350 	return true;
351 }
352 
353 static void
354 xfs_allocbt_read_verify(
355 	struct xfs_buf	*bp)
356 {
357 	if (!xfs_btree_sblock_verify_crc(bp))
358 		xfs_buf_ioerror(bp, -EFSBADCRC);
359 	else if (!xfs_allocbt_verify(bp))
360 		xfs_buf_ioerror(bp, -EFSCORRUPTED);
361 
362 	if (bp->b_error) {
363 		trace_xfs_btree_corrupt(bp, _RET_IP_);
364 		xfs_verifier_error(bp);
365 	}
366 }
367 
368 static void
369 xfs_allocbt_write_verify(
370 	struct xfs_buf	*bp)
371 {
372 	if (!xfs_allocbt_verify(bp)) {
373 		trace_xfs_btree_corrupt(bp, _RET_IP_);
374 		xfs_buf_ioerror(bp, -EFSCORRUPTED);
375 		xfs_verifier_error(bp);
376 		return;
377 	}
378 	xfs_btree_sblock_calc_crc(bp);
379 
380 }
381 
382 const struct xfs_buf_ops xfs_allocbt_buf_ops = {
383 	.verify_read = xfs_allocbt_read_verify,
384 	.verify_write = xfs_allocbt_write_verify,
385 };
386 
387 
388 #if defined(DEBUG) || defined(XFS_WARN)
389 STATIC int
390 xfs_allocbt_keys_inorder(
391 	struct xfs_btree_cur	*cur,
392 	union xfs_btree_key	*k1,
393 	union xfs_btree_key	*k2)
394 {
395 	if (cur->bc_btnum == XFS_BTNUM_BNO) {
396 		return be32_to_cpu(k1->alloc.ar_startblock) <
397 		       be32_to_cpu(k2->alloc.ar_startblock);
398 	} else {
399 		return be32_to_cpu(k1->alloc.ar_blockcount) <
400 			be32_to_cpu(k2->alloc.ar_blockcount) ||
401 			(k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
402 			 be32_to_cpu(k1->alloc.ar_startblock) <
403 			 be32_to_cpu(k2->alloc.ar_startblock));
404 	}
405 }
406 
407 STATIC int
408 xfs_allocbt_recs_inorder(
409 	struct xfs_btree_cur	*cur,
410 	union xfs_btree_rec	*r1,
411 	union xfs_btree_rec	*r2)
412 {
413 	if (cur->bc_btnum == XFS_BTNUM_BNO) {
414 		return be32_to_cpu(r1->alloc.ar_startblock) +
415 			be32_to_cpu(r1->alloc.ar_blockcount) <=
416 			be32_to_cpu(r2->alloc.ar_startblock);
417 	} else {
418 		return be32_to_cpu(r1->alloc.ar_blockcount) <
419 			be32_to_cpu(r2->alloc.ar_blockcount) ||
420 			(r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
421 			 be32_to_cpu(r1->alloc.ar_startblock) <
422 			 be32_to_cpu(r2->alloc.ar_startblock));
423 	}
424 }
425 #endif	/* DEBUG */
426 
427 static const struct xfs_btree_ops xfs_allocbt_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_rec_from_key	= xfs_allocbt_init_rec_from_key,
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_allocbt_key_diff,
443 	.buf_ops		= &xfs_allocbt_buf_ops,
444 #if defined(DEBUG) || defined(XFS_WARN)
445 	.keys_inorder		= xfs_allocbt_keys_inorder,
446 	.recs_inorder		= xfs_allocbt_recs_inorder,
447 #endif
448 };
449 
450 /*
451  * Allocate a new allocation btree cursor.
452  */
453 struct xfs_btree_cur *			/* new alloc btree cursor */
454 xfs_allocbt_init_cursor(
455 	struct xfs_mount	*mp,		/* file system mount point */
456 	struct xfs_trans	*tp,		/* transaction pointer */
457 	struct xfs_buf		*agbp,		/* buffer for agf structure */
458 	xfs_agnumber_t		agno,		/* allocation group number */
459 	xfs_btnum_t		btnum)		/* btree identifier */
460 {
461 	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
462 	struct xfs_btree_cur	*cur;
463 
464 	ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT);
465 
466 	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
467 
468 	cur->bc_tp = tp;
469 	cur->bc_mp = mp;
470 	cur->bc_btnum = btnum;
471 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
472 	cur->bc_ops = &xfs_allocbt_ops;
473 
474 	if (btnum == XFS_BTNUM_CNT) {
475 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);
476 		cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
477 	} else {
478 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]);
479 	}
480 
481 	cur->bc_private.a.agbp = agbp;
482 	cur->bc_private.a.agno = agno;
483 
484 	if (xfs_sb_version_hascrc(&mp->m_sb))
485 		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
486 
487 	return cur;
488 }
489 
490 /*
491  * Calculate number of records in an alloc btree block.
492  */
493 int
494 xfs_allocbt_maxrecs(
495 	struct xfs_mount	*mp,
496 	int			blocklen,
497 	int			leaf)
498 {
499 	blocklen -= XFS_ALLOC_BLOCK_LEN(mp);
500 
501 	if (leaf)
502 		return blocklen / sizeof(xfs_alloc_rec_t);
503 	return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
504 }
505