xfs_alloc_btree.c 11.6 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2 3
 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
L
Linus Torvalds 已提交
4
 *
5 6
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
L
Linus Torvalds 已提交
7 8
 * published by the Free Software Foundation.
 *
9 10 11 12
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
L
Linus Torvalds 已提交
13
 *
14 15 16
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
L
Linus Torvalds 已提交
17 18
 */
#include "xfs.h"
19
#include "xfs_fs.h"
20
#include "xfs_shared.h"
21
#include "xfs_format.h"
22 23
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
L
Linus Torvalds 已提交
24 25 26
#include "xfs_sb.h"
#include "xfs_mount.h"
#include "xfs_btree.h"
27
#include "xfs_alloc_btree.h"
L
Linus Torvalds 已提交
28
#include "xfs_alloc.h"
29
#include "xfs_extent_busy.h"
L
Linus Torvalds 已提交
30
#include "xfs_error.h"
C
Christoph Hellwig 已提交
31
#include "xfs_trace.h"
32
#include "xfs_cksum.h"
33
#include "xfs_trans.h"
L
Linus Torvalds 已提交
34 35


36 37 38 39 40 41 42 43 44
STATIC struct xfs_btree_cur *
xfs_allocbt_dup_cursor(
	struct xfs_btree_cur	*cur)
{
	return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
			cur->bc_private.a.agbp, cur->bc_private.a.agno,
			cur->bc_btnum);
}

45 46 47 48 49 50 51 52 53 54
STATIC void
xfs_allocbt_set_root(
	struct xfs_btree_cur	*cur,
	union xfs_btree_ptr	*ptr,
	int			inc)
{
	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
	int			btnum = cur->bc_btnum;
55
	struct xfs_perag	*pag = xfs_perag_get(cur->bc_mp, seqno);
56 57 58 59 60

	ASSERT(ptr->s != 0);

	agf->agf_roots[btnum] = ptr->s;
	be32_add_cpu(&agf->agf_levels[btnum], inc);
61 62
	pag->pagf_levels[btnum] += inc;
	xfs_perag_put(pag);
63 64 65 66

	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
}

67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
STATIC int
xfs_allocbt_alloc_block(
	struct xfs_btree_cur	*cur,
	union xfs_btree_ptr	*start,
	union xfs_btree_ptr	*new,
	int			*stat)
{
	int			error;
	xfs_agblock_t		bno;

	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);

	/* Allocate the new block from the freelist. If we can't, give up.  */
	error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_private.a.agbp,
				       &bno, 1);
	if (error) {
		XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
		return error;
	}

	if (bno == NULLAGBLOCK) {
		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
		*stat = 0;
		return 0;
	}
92

D
Dave Chinner 已提交
93
	xfs_extent_busy_reuse(cur->bc_mp, cur->bc_private.a.agno, bno, 1, false);
94 95 96 97 98 99 100 101 102

	xfs_trans_agbtree_delta(cur->bc_tp, 1);
	new->s = cpu_to_be32(bno);

	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
	*stat = 1;
	return 0;
}

103 104 105 106 107 108 109 110 111 112
STATIC int
xfs_allocbt_free_block(
	struct xfs_btree_cur	*cur,
	struct xfs_buf		*bp)
{
	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
	xfs_agblock_t		bno;
	int			error;

113
	bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
114 115 116 117
	error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
	if (error)
		return error;

D
Dave Chinner 已提交
118 119
	xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
			      XFS_EXTENT_BUSY_SKIP_DISCARD);
120 121 122 123
	xfs_trans_agbtree_delta(cur->bc_tp, -1);
	return 0;
}

L
Linus Torvalds 已提交
124
/*
125
 * Update the longest extent in the AGF
L
Linus Torvalds 已提交
126
 */
127 128 129 130 131 132 133
STATIC void
xfs_allocbt_update_lastrec(
	struct xfs_btree_cur	*cur,
	struct xfs_btree_block	*block,
	union xfs_btree_rec	*rec,
	int			ptr,
	int			reason)
L
Linus Torvalds 已提交
134
{
135 136
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
137
	struct xfs_perag	*pag;
138
	__be32			len;
139
	int			numrecs;
L
Linus Torvalds 已提交
140

141
	ASSERT(cur->bc_btnum == XFS_BTNUM_CNT);
L
Linus Torvalds 已提交
142

143 144
	switch (reason) {
	case LASTREC_UPDATE:
L
Linus Torvalds 已提交
145
		/*
146 147
		 * If this is the last leaf block and it's the last record,
		 * then update the size of the longest extent in the AG.
L
Linus Torvalds 已提交
148
		 */
149 150 151 152
		if (ptr != xfs_btree_get_numrecs(block))
			return;
		len = rec->alloc.ar_blockcount;
		break;
153 154 155 156 157
	case LASTREC_INSREC:
		if (be32_to_cpu(rec->alloc.ar_blockcount) <=
		    be32_to_cpu(agf->agf_longest))
			return;
		len = rec->alloc.ar_blockcount;
158 159 160 161 162 163 164 165 166 167
		break;
	case LASTREC_DELREC:
		numrecs = xfs_btree_get_numrecs(block);
		if (ptr <= numrecs)
			return;
		ASSERT(ptr == numrecs + 1);

		if (numrecs) {
			xfs_alloc_rec_t *rrp;

168
			rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs);
169 170 171 172 173
			len = rrp->ar_blockcount;
		} else {
			len = 0;
		}

174
		break;
175 176 177
	default:
		ASSERT(0);
		return;
L
Linus Torvalds 已提交
178
	}
179

180
	agf->agf_longest = len;
181 182 183
	pag = xfs_perag_get(cur->bc_mp, seqno);
	pag->pagf_longest = be32_to_cpu(len);
	xfs_perag_put(pag);
184
	xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp, XFS_AGF_LONGEST);
185 186
}

187 188 189 190 191 192 193 194
STATIC int
xfs_allocbt_get_minrecs(
	struct xfs_btree_cur	*cur,
	int			level)
{
	return cur->bc_mp->m_alloc_mnr[level != 0];
}

195 196 197 198 199 200 201 202
STATIC int
xfs_allocbt_get_maxrecs(
	struct xfs_btree_cur	*cur,
	int			level)
{
	return cur->bc_mp->m_alloc_mxr[level != 0];
}

203 204 205 206 207 208 209 210 211 212 213
STATIC void
xfs_allocbt_init_key_from_rec(
	union xfs_btree_key	*key,
	union xfs_btree_rec	*rec)
{
	ASSERT(rec->alloc.ar_startblock != 0);

	key->alloc.ar_startblock = rec->alloc.ar_startblock;
	key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
}

214 215 216 217 218 219 220 221 222 223 224
STATIC void
xfs_allocbt_init_rec_from_cur(
	struct xfs_btree_cur	*cur,
	union xfs_btree_rec	*rec)
{
	ASSERT(cur->bc_rec.a.ar_startblock != 0);

	rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
	rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
}

225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
STATIC void
xfs_allocbt_init_ptr_from_cur(
	struct xfs_btree_cur	*cur,
	union xfs_btree_ptr	*ptr)
{
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);

	ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno));
	ASSERT(agf->agf_roots[cur->bc_btnum] != 0);

	ptr->s = agf->agf_roots[cur->bc_btnum];
}

STATIC __int64_t
xfs_allocbt_key_diff(
	struct xfs_btree_cur	*cur,
	union xfs_btree_key	*key)
{
	xfs_alloc_rec_incore_t	*rec = &cur->bc_rec.a;
	xfs_alloc_key_t		*kp = &key->alloc;
	__int64_t		diff;

	if (cur->bc_btnum == XFS_BTNUM_BNO) {
		return (__int64_t)be32_to_cpu(kp->ar_startblock) -
				rec->ar_startblock;
	}

	diff = (__int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
	if (diff)
		return diff;

	return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
}

259
static bool
260
xfs_allocbt_verify(
261 262 263 264 265 266 267 268 269 270
	struct xfs_buf		*bp)
{
	struct xfs_mount	*mp = bp->b_target->bt_mount;
	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
	struct xfs_perag	*pag = bp->b_pag;
	unsigned int		level;

	/*
	 * magic number and level verification
	 *
271 272 273 274 275 276 277 278
	 * During growfs operations, we can't verify the exact level or owner as
	 * the perag is not fully initialised and hence not attached to the
	 * buffer.  In this case, check against the maximum tree depth.
	 *
	 * Similarly, during log recovery we will have a perag structure
	 * attached, but the agf information will not yet have been initialised
	 * from the on disk AGF. Again, we can only check against maximum limits
	 * in this case.
279 280 281
	 */
	level = be16_to_cpu(block->bb_level);
	switch (block->bb_magic) {
282
	case cpu_to_be32(XFS_ABTB_CRC_MAGIC):
283
		if (!xfs_btree_sblock_v5hdr_verify(bp))
284 285
			return false;
		/* fall through */
286
	case cpu_to_be32(XFS_ABTB_MAGIC):
287 288 289 290 291
		if (pag && pag->pagf_init) {
			if (level >= pag->pagf_levels[XFS_BTNUM_BNOi])
				return false;
		} else if (level >= mp->m_ag_maxlevels)
			return false;
292
		break;
293
	case cpu_to_be32(XFS_ABTC_CRC_MAGIC):
294
		if (!xfs_btree_sblock_v5hdr_verify(bp))
295 296
			return false;
		/* fall through */
297
	case cpu_to_be32(XFS_ABTC_MAGIC):
298 299 300 301 302
		if (pag && pag->pagf_init) {
			if (level >= pag->pagf_levels[XFS_BTNUM_CNTi])
				return false;
		} else if (level >= mp->m_ag_maxlevels)
			return false;
303 304
		break;
	default:
305
		return false;
306 307
	}

308
	return xfs_btree_sblock_verify(bp, mp->m_alloc_mxr[level != 0]);
309
}
310

311
static void
312
xfs_allocbt_read_verify(
313 314
	struct xfs_buf	*bp)
{
315
	if (!xfs_btree_sblock_verify_crc(bp))
D
Dave Chinner 已提交
316
		xfs_buf_ioerror(bp, -EFSBADCRC);
317
	else if (!xfs_allocbt_verify(bp))
D
Dave Chinner 已提交
318
		xfs_buf_ioerror(bp, -EFSCORRUPTED);
319 320 321 322

	if (bp->b_error) {
		trace_xfs_btree_corrupt(bp, _RET_IP_);
		xfs_verifier_error(bp);
323
	}
324 325
}

326 327
static void
xfs_allocbt_write_verify(
328 329
	struct xfs_buf	*bp)
{
330 331
	if (!xfs_allocbt_verify(bp)) {
		trace_xfs_btree_corrupt(bp, _RET_IP_);
D
Dave Chinner 已提交
332
		xfs_buf_ioerror(bp, -EFSCORRUPTED);
333
		xfs_verifier_error(bp);
334
		return;
335 336 337
	}
	xfs_btree_sblock_calc_crc(bp);

338 339
}

340
const struct xfs_buf_ops xfs_allocbt_buf_ops = {
341
	.name = "xfs_allocbt",
342 343 344 345 346
	.verify_read = xfs_allocbt_read_verify,
	.verify_write = xfs_allocbt_write_verify,
};


D
Dave Chinner 已提交
347
#if defined(DEBUG) || defined(XFS_WARN)
348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
STATIC int
xfs_allocbt_keys_inorder(
	struct xfs_btree_cur	*cur,
	union xfs_btree_key	*k1,
	union xfs_btree_key	*k2)
{
	if (cur->bc_btnum == XFS_BTNUM_BNO) {
		return be32_to_cpu(k1->alloc.ar_startblock) <
		       be32_to_cpu(k2->alloc.ar_startblock);
	} else {
		return be32_to_cpu(k1->alloc.ar_blockcount) <
			be32_to_cpu(k2->alloc.ar_blockcount) ||
			(k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
			 be32_to_cpu(k1->alloc.ar_startblock) <
			 be32_to_cpu(k2->alloc.ar_startblock));
	}
}

STATIC int
xfs_allocbt_recs_inorder(
	struct xfs_btree_cur	*cur,
	union xfs_btree_rec	*r1,
	union xfs_btree_rec	*r2)
{
	if (cur->bc_btnum == XFS_BTNUM_BNO) {
		return be32_to_cpu(r1->alloc.ar_startblock) +
			be32_to_cpu(r1->alloc.ar_blockcount) <=
			be32_to_cpu(r2->alloc.ar_startblock);
	} else {
		return be32_to_cpu(r1->alloc.ar_blockcount) <
			be32_to_cpu(r2->alloc.ar_blockcount) ||
			(r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
			 be32_to_cpu(r1->alloc.ar_startblock) <
			 be32_to_cpu(r2->alloc.ar_startblock));
	}
}
#endif	/* DEBUG */

386
static const struct xfs_btree_ops xfs_allocbt_ops = {
387 388 389
	.rec_len		= sizeof(xfs_alloc_rec_t),
	.key_len		= sizeof(xfs_alloc_key_t),

390
	.dup_cursor		= xfs_allocbt_dup_cursor,
391
	.set_root		= xfs_allocbt_set_root,
392
	.alloc_block		= xfs_allocbt_alloc_block,
393
	.free_block		= xfs_allocbt_free_block,
394
	.update_lastrec		= xfs_allocbt_update_lastrec,
395
	.get_minrecs		= xfs_allocbt_get_minrecs,
396
	.get_maxrecs		= xfs_allocbt_get_maxrecs,
397
	.init_key_from_rec	= xfs_allocbt_init_key_from_rec,
398
	.init_rec_from_cur	= xfs_allocbt_init_rec_from_cur,
399 400
	.init_ptr_from_cur	= xfs_allocbt_init_ptr_from_cur,
	.key_diff		= xfs_allocbt_key_diff,
401
	.buf_ops		= &xfs_allocbt_buf_ops,
D
Dave Chinner 已提交
402
#if defined(DEBUG) || defined(XFS_WARN)
403 404 405
	.keys_inorder		= xfs_allocbt_keys_inorder,
	.recs_inorder		= xfs_allocbt_recs_inorder,
#endif
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
};

/*
 * Allocate a new allocation btree cursor.
 */
struct xfs_btree_cur *			/* new alloc btree cursor */
xfs_allocbt_init_cursor(
	struct xfs_mount	*mp,		/* file system mount point */
	struct xfs_trans	*tp,		/* transaction pointer */
	struct xfs_buf		*agbp,		/* buffer for agf structure */
	xfs_agnumber_t		agno,		/* allocation group number */
	xfs_btnum_t		btnum)		/* btree identifier */
{
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
	struct xfs_btree_cur	*cur;

	ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT);

424
	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
425 426 427 428 429 430

	cur->bc_tp = tp;
	cur->bc_mp = mp;
	cur->bc_btnum = btnum;
	cur->bc_blocklog = mp->m_sb.sb_blocklog;
	cur->bc_ops = &xfs_allocbt_ops;
D
Dave Chinner 已提交
431 432 433 434
	if (btnum == XFS_BTNUM_BNO)
		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtb_2);
	else
		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtc_2);
435 436 437

	if (btnum == XFS_BTNUM_CNT) {
		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);
438
		cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
439 440 441
	} else {
		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]);
	}
442 443 444 445

	cur->bc_private.a.agbp = agbp;
	cur->bc_private.a.agno = agno;

446 447 448
	if (xfs_sb_version_hascrc(&mp->m_sb))
		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;

449 450
	return cur;
}
451 452 453 454 455 456 457 458 459 460

/*
 * Calculate number of records in an alloc btree block.
 */
int
xfs_allocbt_maxrecs(
	struct xfs_mount	*mp,
	int			blocklen,
	int			leaf)
{
461
	blocklen -= XFS_ALLOC_BLOCK_LEN(mp);
462 463 464 465 466

	if (leaf)
		return blocklen / sizeof(xfs_alloc_rec_t);
	return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
}