xfs_alloc.c 87.5 KB
Newer Older
D
Dave Chinner 已提交
1
// SPDX-License-Identifier: GPL-2.0
L
Linus Torvalds 已提交
2
/*
3 4
 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
L
Linus Torvalds 已提交
5 6
 */
#include "xfs.h"
7
#include "xfs_fs.h"
8
#include "xfs_format.h"
9
#include "xfs_log_format.h"
10
#include "xfs_shared.h"
11
#include "xfs_trans_resv.h"
12
#include "xfs_bit.h"
L
Linus Torvalds 已提交
13 14
#include "xfs_sb.h"
#include "xfs_mount.h"
15
#include "xfs_defer.h"
16
#include "xfs_inode.h"
L
Linus Torvalds 已提交
17
#include "xfs_btree.h"
18
#include "xfs_rmap.h"
19
#include "xfs_alloc_btree.h"
L
Linus Torvalds 已提交
20
#include "xfs_alloc.h"
21
#include "xfs_extent_busy.h"
22
#include "xfs_errortag.h"
L
Linus Torvalds 已提交
23
#include "xfs_error.h"
D
Dave Chinner 已提交
24
#include "xfs_cksum.h"
C
Christoph Hellwig 已提交
25
#include "xfs_trace.h"
26
#include "xfs_trans.h"
D
Dave Chinner 已提交
27
#include "xfs_buf_item.h"
28
#include "xfs_log.h"
29
#include "xfs_ag_resv.h"
30 31 32
#include "xfs_bmap.h"

extern kmem_zone_t	*xfs_bmap_free_item_zone;
L
Linus Torvalds 已提交
33

34
struct workqueue_struct *xfs_alloc_wq;
L
Linus Torvalds 已提交
35 36 37 38 39 40 41 42 43 44

#define XFS_ABSDIFF(a,b)	(((a) <= (b)) ? ((b) - (a)) : ((a) - (b)))

#define	XFSA_FIXUP_BNO_OK	1
#define	XFSA_FIXUP_CNT_OK	2

STATIC int xfs_alloc_ag_vextent_exact(xfs_alloc_arg_t *);
STATIC int xfs_alloc_ag_vextent_near(xfs_alloc_arg_t *);
STATIC int xfs_alloc_ag_vextent_size(xfs_alloc_arg_t *);
STATIC int xfs_alloc_ag_vextent_small(xfs_alloc_arg_t *,
45
		xfs_btree_cur_t *, xfs_agblock_t *, xfs_extlen_t *, int *);
L
Linus Torvalds 已提交
46

47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63
/*
 * Size of the AGFL.  For CRC-enabled filesystes we steal a couple of slots in
 * the beginning of the block for a proper header with the location information
 * and CRC.
 */
unsigned int
xfs_agfl_size(
	struct xfs_mount	*mp)
{
	unsigned int		size = mp->m_sb.sb_sectsize;

	if (xfs_sb_version_hascrc(&mp->m_sb))
		size -= sizeof(struct xfs_agfl);

	return size / sizeof(xfs_agblock_t);
}

64 65 66 67 68 69 70 71 72 73 74
unsigned int
xfs_refc_block(
	struct xfs_mount	*mp)
{
	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
		return XFS_RMAP_BLOCK(mp) + 1;
	if (xfs_sb_version_hasfinobt(&mp->m_sb))
		return XFS_FIBT_BLOCK(mp) + 1;
	return XFS_IBT_BLOCK(mp) + 1;
}

75 76 77 78
xfs_extlen_t
xfs_prealloc_blocks(
	struct xfs_mount	*mp)
{
79 80
	if (xfs_sb_version_hasreflink(&mp->m_sb))
		return xfs_refc_block(mp) + 1;
81 82 83 84 85 86 87
	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
		return XFS_RMAP_BLOCK(mp) + 1;
	if (xfs_sb_version_hasfinobt(&mp->m_sb))
		return XFS_FIBT_BLOCK(mp) + 1;
	return XFS_IBT_BLOCK(mp) + 1;
}

88 89 90 91 92 93 94 95 96 97 98 99
/*
 * In order to avoid ENOSPC-related deadlock caused by out-of-order locking of
 * AGF buffer (PV 947395), we place constraints on the relationship among
 * actual allocations for data blocks, freelist blocks, and potential file data
 * bmap btree blocks. However, these restrictions may result in no actual space
 * allocated for a delayed extent, for example, a data block in a certain AG is
 * allocated but there is no additional block for the additional bmap btree
 * block due to a split of the bmap btree of the file. The result of this may
 * lead to an infinite loop when the file gets flushed to disk and all delayed
 * extents need to be actually allocated. To get around this, we explicitly set
 * aside a few blocks which will not be reserved in delayed allocation.
 *
100 101
 * We need to reserve 4 fsbs _per AG_ for the freelist and 4 more to handle a
 * potential split of the file's bmap btree.
102 103 104 105 106
 */
unsigned int
xfs_alloc_set_aside(
	struct xfs_mount	*mp)
{
107
	return mp->m_sb.sb_agcount * (XFS_ALLOC_AGFL_RESERVE + 4);
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
}

/*
 * When deciding how much space to allocate out of an AG, we limit the
 * allocation maximum size to the size the AG. However, we cannot use all the
 * blocks in the AG - some are permanently used by metadata. These
 * blocks are generally:
 *	- the AG superblock, AGF, AGI and AGFL
 *	- the AGF (bno and cnt) and AGI btree root blocks, and optionally
 *	  the AGI free inode and rmap btree root blocks.
 *	- blocks on the AGFL according to xfs_alloc_set_aside() limits
 *	- the rmapbt root block
 *
 * The AG headers are sector sized, so the amount of space they take up is
 * dependent on filesystem geometry. The others are all single blocks.
 */
unsigned int
xfs_alloc_ag_max_usable(
	struct xfs_mount	*mp)
{
	unsigned int		blocks;

	blocks = XFS_BB_TO_FSB(mp, XFS_FSS_TO_BB(mp, 4)); /* ag headers */
	blocks += XFS_ALLOC_AGFL_RESERVE;
	blocks += 3;			/* AGF, AGI btree root blocks */
	if (xfs_sb_version_hasfinobt(&mp->m_sb))
		blocks++;		/* finobt root block */
	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
		blocks++; 		/* rmap root block */
137 138
	if (xfs_sb_version_hasreflink(&mp->m_sb))
		blocks++;		/* refcount root block */
139 140 141 142

	return mp->m_sb.sb_agblocks - blocks;
}

143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
/*
 * Lookup the record equal to [bno, len] in the btree given by cur.
 */
STATIC int				/* error */
xfs_alloc_lookup_eq(
	struct xfs_btree_cur	*cur,	/* btree cursor */
	xfs_agblock_t		bno,	/* starting block of extent */
	xfs_extlen_t		len,	/* length of extent */
	int			*stat)	/* success/failure */
{
	cur->bc_rec.a.ar_startblock = bno;
	cur->bc_rec.a.ar_blockcount = len;
	return xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
}

/*
 * Lookup the first record greater than or equal to [bno, len]
 * in the btree given by cur.
 */
162
int				/* error */
163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
xfs_alloc_lookup_ge(
	struct xfs_btree_cur	*cur,	/* btree cursor */
	xfs_agblock_t		bno,	/* starting block of extent */
	xfs_extlen_t		len,	/* length of extent */
	int			*stat)	/* success/failure */
{
	cur->bc_rec.a.ar_startblock = bno;
	cur->bc_rec.a.ar_blockcount = len;
	return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
}

/*
 * Lookup the first record less than or equal to [bno, len]
 * in the btree given by cur.
 */
178
int					/* error */
179 180 181 182 183 184 185 186 187 188 189
xfs_alloc_lookup_le(
	struct xfs_btree_cur	*cur,	/* btree cursor */
	xfs_agblock_t		bno,	/* starting block of extent */
	xfs_extlen_t		len,	/* length of extent */
	int			*stat)	/* success/failure */
{
	cur->bc_rec.a.ar_startblock = bno;
	cur->bc_rec.a.ar_blockcount = len;
	return xfs_btree_lookup(cur, XFS_LOOKUP_LE, stat);
}

190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206
/*
 * Update the record referred to by cur to the value given
 * by [bno, len].
 * This either works (return 0) or gets an EFSCORRUPTED error.
 */
STATIC int				/* error */
xfs_alloc_update(
	struct xfs_btree_cur	*cur,	/* btree cursor */
	xfs_agblock_t		bno,	/* starting block of extent */
	xfs_extlen_t		len)	/* length of extent */
{
	union xfs_btree_rec	rec;

	rec.alloc.ar_startblock = cpu_to_be32(bno);
	rec.alloc.ar_blockcount = cpu_to_be32(len);
	return xfs_btree_update(cur, &rec);
}
207

208 209 210
/*
 * Get the data from the pointed-to record.
 */
C
Christoph Hellwig 已提交
211
int					/* error */
212 213 214 215 216 217
xfs_alloc_get_rec(
	struct xfs_btree_cur	*cur,	/* btree cursor */
	xfs_agblock_t		*bno,	/* output: starting block of extent */
	xfs_extlen_t		*len,	/* output: length of extent */
	int			*stat)	/* output: success/failure */
{
218 219
	struct xfs_mount	*mp = cur->bc_mp;
	xfs_agnumber_t		agno = cur->bc_private.a.agno;
220 221 222 223
	union xfs_btree_rec	*rec;
	int			error;

	error = xfs_btree_get_rec(cur, &rec, stat);
224 225 226
	if (error || !(*stat))
		return error;
	if (rec->alloc.ar_blockcount == 0)
227
		goto out_bad_rec;
228 229 230 231

	*bno = be32_to_cpu(rec->alloc.ar_startblock);
	*len = be32_to_cpu(rec->alloc.ar_blockcount);

232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
	/* check for valid extent range, including overflow */
	if (!xfs_verify_agbno(mp, agno, *bno))
		goto out_bad_rec;
	if (*bno > *bno + *len)
		goto out_bad_rec;
	if (!xfs_verify_agbno(mp, agno, *bno + *len - 1))
		goto out_bad_rec;

	return 0;

out_bad_rec:
	xfs_warn(mp,
		"%s Freespace BTree record corruption in AG %d detected!",
		cur->bc_btnum == XFS_BTNUM_BNO ? "Block" : "Size", agno);
	xfs_warn(mp,
		"start block 0x%x block count 0x%x", *bno, *len);
	return -EFSCORRUPTED;
249 250
}

L
Linus Torvalds 已提交
251 252 253 254
/*
 * Compute aligned version of the found extent.
 * Takes alignment and min length into account.
 */
255
STATIC bool
L
Linus Torvalds 已提交
256
xfs_alloc_compute_aligned(
257
	xfs_alloc_arg_t	*args,		/* allocation argument structure */
L
Linus Torvalds 已提交
258 259 260
	xfs_agblock_t	foundbno,	/* starting block in found extent */
	xfs_extlen_t	foundlen,	/* length in found extent */
	xfs_agblock_t	*resbno,	/* result block number */
261 262
	xfs_extlen_t	*reslen,	/* result length */
	unsigned	*busy_gen)
L
Linus Torvalds 已提交
263
{
264 265
	xfs_agblock_t	bno = foundbno;
	xfs_extlen_t	len = foundlen;
266
	xfs_extlen_t	diff;
267
	bool		busy;
L
Linus Torvalds 已提交
268

269
	/* Trim busy sections out of found extent */
270
	busy = xfs_extent_busy_trim(args, &bno, &len, busy_gen);
271

272 273 274 275 276 277 278 279 280 281 282 283
	/*
	 * If we have a largish extent that happens to start before min_agbno,
	 * see if we can shift it into range...
	 */
	if (bno < args->min_agbno && bno + len > args->min_agbno) {
		diff = args->min_agbno - bno;
		if (len > diff) {
			bno += diff;
			len -= diff;
		}
	}

284 285
	if (args->alignment > 1 && len >= args->minlen) {
		xfs_agblock_t	aligned_bno = roundup(bno, args->alignment);
286 287

		diff = aligned_bno - bno;
288 289 290

		*resbno = aligned_bno;
		*reslen = diff >= len ? 0 : len - diff;
L
Linus Torvalds 已提交
291
	} else {
292 293
		*resbno = bno;
		*reslen = len;
L
Linus Torvalds 已提交
294
	}
295 296

	return busy;
L
Linus Torvalds 已提交
297 298 299 300 301 302 303 304 305 306 307
}

/*
 * Compute best start block and diff for "near" allocations.
 * freelen >= wantlen already checked by caller.
 */
STATIC xfs_extlen_t			/* difference value (absolute) */
xfs_alloc_compute_diff(
	xfs_agblock_t	wantbno,	/* target starting block */
	xfs_extlen_t	wantlen,	/* target length */
	xfs_extlen_t	alignment,	/* target alignment */
308
	int		datatype,	/* are we allocating data? */
L
Linus Torvalds 已提交
309 310 311 312 313 314 315 316 317 318
	xfs_agblock_t	freebno,	/* freespace's starting block */
	xfs_extlen_t	freelen,	/* freespace's length */
	xfs_agblock_t	*newbnop)	/* result: best start block from free */
{
	xfs_agblock_t	freeend;	/* end of freespace extent */
	xfs_agblock_t	newbno1;	/* return block number */
	xfs_agblock_t	newbno2;	/* other new block number */
	xfs_extlen_t	newlen1=0;	/* length with newbno1 */
	xfs_extlen_t	newlen2=0;	/* length with newbno2 */
	xfs_agblock_t	wantend;	/* end of target extent */
319
	bool		userdata = xfs_alloc_is_userdata(datatype);
L
Linus Torvalds 已提交
320 321 322 323

	ASSERT(freelen >= wantlen);
	freeend = freebno + freelen;
	wantend = wantbno + wantlen;
324 325 326 327 328 329 330 331
	/*
	 * We want to allocate from the start of a free extent if it is past
	 * the desired block or if we are allocating user data and the free
	 * extent is before desired block. The second case is there to allow
	 * for contiguous allocation from the remaining free space if the file
	 * grows in the short term.
	 */
	if (freebno >= wantbno || (userdata && freeend < wantend)) {
L
Linus Torvalds 已提交
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390
		if ((newbno1 = roundup(freebno, alignment)) >= freeend)
			newbno1 = NULLAGBLOCK;
	} else if (freeend >= wantend && alignment > 1) {
		newbno1 = roundup(wantbno, alignment);
		newbno2 = newbno1 - alignment;
		if (newbno1 >= freeend)
			newbno1 = NULLAGBLOCK;
		else
			newlen1 = XFS_EXTLEN_MIN(wantlen, freeend - newbno1);
		if (newbno2 < freebno)
			newbno2 = NULLAGBLOCK;
		else
			newlen2 = XFS_EXTLEN_MIN(wantlen, freeend - newbno2);
		if (newbno1 != NULLAGBLOCK && newbno2 != NULLAGBLOCK) {
			if (newlen1 < newlen2 ||
			    (newlen1 == newlen2 &&
			     XFS_ABSDIFF(newbno1, wantbno) >
			     XFS_ABSDIFF(newbno2, wantbno)))
				newbno1 = newbno2;
		} else if (newbno2 != NULLAGBLOCK)
			newbno1 = newbno2;
	} else if (freeend >= wantend) {
		newbno1 = wantbno;
	} else if (alignment > 1) {
		newbno1 = roundup(freeend - wantlen, alignment);
		if (newbno1 > freeend - wantlen &&
		    newbno1 - alignment >= freebno)
			newbno1 -= alignment;
		else if (newbno1 >= freeend)
			newbno1 = NULLAGBLOCK;
	} else
		newbno1 = freeend - wantlen;
	*newbnop = newbno1;
	return newbno1 == NULLAGBLOCK ? 0 : XFS_ABSDIFF(newbno1, wantbno);
}

/*
 * Fix up the length, based on mod and prod.
 * len should be k * prod + mod for some k.
 * If len is too small it is returned unchanged.
 * If len hits maxlen it is left alone.
 */
STATIC void
xfs_alloc_fix_len(
	xfs_alloc_arg_t	*args)		/* allocation argument structure */
{
	xfs_extlen_t	k;
	xfs_extlen_t	rlen;

	ASSERT(args->mod < args->prod);
	rlen = args->len;
	ASSERT(rlen >= args->minlen);
	ASSERT(rlen <= args->maxlen);
	if (args->prod <= 1 || rlen < args->mod || rlen == args->maxlen ||
	    (args->mod == 0 && rlen < args->prod))
		return;
	k = rlen % args->prod;
	if (k == args->mod)
		return;
391 392 393 394
	if (k > args->mod)
		rlen = rlen - (k - args->mod);
	else
		rlen = rlen - args->prod + (args->mod - k);
395
	/* casts to (int) catch length underflows */
396 397 398 399
	if ((int)rlen < (int)args->minlen)
		return;
	ASSERT(rlen >= args->minlen && rlen <= args->maxlen);
	ASSERT(rlen % args->prod == args->mod);
400 401
	ASSERT(args->pag->pagf_freeblks + args->pag->pagf_flcount >=
		rlen + args->minleft);
L
Linus Torvalds 已提交
402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
	args->len = rlen;
}

/*
 * Update the two btrees, logically removing from freespace the extent
 * starting at rbno, rlen blocks.  The extent is contained within the
 * actual (current) free extent fbno for flen blocks.
 * Flags are passed in indicating whether the cursors are set to the
 * relevant records.
 */
STATIC int				/* error code */
xfs_alloc_fixup_trees(
	xfs_btree_cur_t	*cnt_cur,	/* cursor for by-size btree */
	xfs_btree_cur_t	*bno_cur,	/* cursor for by-block btree */
	xfs_agblock_t	fbno,		/* starting block of free extent */
	xfs_extlen_t	flen,		/* length of free extent */
	xfs_agblock_t	rbno,		/* starting block of returned extent */
	xfs_extlen_t	rlen,		/* length of returned extent */
	int		flags)		/* flags, XFSA_FIXUP_... */
{
	int		error;		/* error code */
	int		i;		/* operation results */
	xfs_agblock_t	nfbno1;		/* first new free startblock */
	xfs_agblock_t	nfbno2;		/* second new free startblock */
	xfs_extlen_t	nflen1=0;	/* first new free length */
	xfs_extlen_t	nflen2=0;	/* second new free length */
428 429 430
	struct xfs_mount *mp;

	mp = cnt_cur->bc_mp;
L
Linus Torvalds 已提交
431 432 433 434 435 436 437 438

	/*
	 * Look up the record in the by-size tree if necessary.
	 */
	if (flags & XFSA_FIXUP_CNT_OK) {
#ifdef DEBUG
		if ((error = xfs_alloc_get_rec(cnt_cur, &nfbno1, &nflen1, &i)))
			return error;
439
		XFS_WANT_CORRUPTED_RETURN(mp,
L
Linus Torvalds 已提交
440 441 442 443 444
			i == 1 && nfbno1 == fbno && nflen1 == flen);
#endif
	} else {
		if ((error = xfs_alloc_lookup_eq(cnt_cur, fbno, flen, &i)))
			return error;
445
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
446 447 448 449 450 451 452 453
	}
	/*
	 * Look up the record in the by-block tree if necessary.
	 */
	if (flags & XFSA_FIXUP_BNO_OK) {
#ifdef DEBUG
		if ((error = xfs_alloc_get_rec(bno_cur, &nfbno1, &nflen1, &i)))
			return error;
454
		XFS_WANT_CORRUPTED_RETURN(mp,
L
Linus Torvalds 已提交
455 456 457 458 459
			i == 1 && nfbno1 == fbno && nflen1 == flen);
#endif
	} else {
		if ((error = xfs_alloc_lookup_eq(bno_cur, fbno, flen, &i)))
			return error;
460
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
461
	}
462

L
Linus Torvalds 已提交
463
#ifdef DEBUG
464 465 466 467 468 469
	if (bno_cur->bc_nlevels == 1 && cnt_cur->bc_nlevels == 1) {
		struct xfs_btree_block	*bnoblock;
		struct xfs_btree_block	*cntblock;

		bnoblock = XFS_BUF_TO_BLOCK(bno_cur->bc_bufs[0]);
		cntblock = XFS_BUF_TO_BLOCK(cnt_cur->bc_bufs[0]);
L
Linus Torvalds 已提交
470

471
		XFS_WANT_CORRUPTED_RETURN(mp,
472
			bnoblock->bb_numrecs == cntblock->bb_numrecs);
L
Linus Torvalds 已提交
473 474
	}
#endif
475

L
Linus Torvalds 已提交
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499
	/*
	 * Deal with all four cases: the allocated record is contained
	 * within the freespace record, so we can have new freespace
	 * at either (or both) end, or no freespace remaining.
	 */
	if (rbno == fbno && rlen == flen)
		nfbno1 = nfbno2 = NULLAGBLOCK;
	else if (rbno == fbno) {
		nfbno1 = rbno + rlen;
		nflen1 = flen - rlen;
		nfbno2 = NULLAGBLOCK;
	} else if (rbno + rlen == fbno + flen) {
		nfbno1 = fbno;
		nflen1 = flen - rlen;
		nfbno2 = NULLAGBLOCK;
	} else {
		nfbno1 = fbno;
		nflen1 = rbno - fbno;
		nfbno2 = rbno + rlen;
		nflen2 = (fbno + flen) - nfbno2;
	}
	/*
	 * Delete the entry from the by-size btree.
	 */
500
	if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
501
		return error;
502
	XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
503 504 505 506 507 508
	/*
	 * Add new by-size btree entry(s).
	 */
	if (nfbno1 != NULLAGBLOCK) {
		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno1, nflen1, &i)))
			return error;
509
		XFS_WANT_CORRUPTED_RETURN(mp, i == 0);
510
		if ((error = xfs_btree_insert(cnt_cur, &i)))
L
Linus Torvalds 已提交
511
			return error;
512
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
513 514 515 516
	}
	if (nfbno2 != NULLAGBLOCK) {
		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno2, nflen2, &i)))
			return error;
517
		XFS_WANT_CORRUPTED_RETURN(mp, i == 0);
518
		if ((error = xfs_btree_insert(cnt_cur, &i)))
L
Linus Torvalds 已提交
519
			return error;
520
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
521 522 523 524 525 526 527 528
	}
	/*
	 * Fix up the by-block btree entry(s).
	 */
	if (nfbno1 == NULLAGBLOCK) {
		/*
		 * No remaining freespace, just delete the by-block tree entry.
		 */
529
		if ((error = xfs_btree_delete(bno_cur, &i)))
L
Linus Torvalds 已提交
530
			return error;
531
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
532 533 534 535 536 537 538 539 540 541 542 543 544
	} else {
		/*
		 * Update the by-block entry to start later|be shorter.
		 */
		if ((error = xfs_alloc_update(bno_cur, nfbno1, nflen1)))
			return error;
	}
	if (nfbno2 != NULLAGBLOCK) {
		/*
		 * 2 resulting free entries, need to add one.
		 */
		if ((error = xfs_alloc_lookup_eq(bno_cur, nfbno2, nflen2, &i)))
			return error;
545
		XFS_WANT_CORRUPTED_RETURN(mp, i == 0);
546
		if ((error = xfs_btree_insert(bno_cur, &i)))
L
Linus Torvalds 已提交
547
			return error;
548
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
549 550 551 552
	}
	return 0;
}

553
static xfs_failaddr_t
554
xfs_agfl_verify(
555 556 557 558 559 560
	struct xfs_buf	*bp)
{
	struct xfs_mount *mp = bp->b_target->bt_mount;
	struct xfs_agfl	*agfl = XFS_BUF_TO_AGFL(bp);
	int		i;

561 562 563 564 565 566 567 568 569
	/*
	 * There is no verification of non-crc AGFLs because mkfs does not
	 * initialise the AGFL to zero or NULL. Hence the only valid part of the
	 * AGFL is what the AGF says is active. We can't get to the AGF, so we
	 * can't verify just those entries are valid.
	 */
	if (!xfs_sb_version_hascrc(&mp->m_sb))
		return NULL;

570
	if (!uuid_equal(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid))
571
		return __this_address;
572
	if (be32_to_cpu(agfl->agfl_magicnum) != XFS_AGFL_MAGIC)
573
		return __this_address;
574 575 576 577 578 579 580
	/*
	 * during growfs operations, the perag is not fully initialised,
	 * so we can't use it for any useful checking. growfs ensures we can't
	 * use it by using uncached buffers that don't have the perag attached
	 * so we can detect and avoid this problem.
	 */
	if (bp->b_pag && be32_to_cpu(agfl->agfl_seqno) != bp->b_pag->pag_agno)
581
		return __this_address;
582

583
	for (i = 0; i < xfs_agfl_size(mp); i++) {
584
		if (be32_to_cpu(agfl->agfl_bno[i]) != NULLAGBLOCK &&
585
		    be32_to_cpu(agfl->agfl_bno[i]) >= mp->m_sb.sb_agblocks)
586
			return __this_address;
587
	}
588

589 590 591
	if (!xfs_log_check_lsn(mp, be64_to_cpu(XFS_BUF_TO_AGFL(bp)->agfl_lsn)))
		return __this_address;
	return NULL;
592 593 594 595 596 597 598
}

static void
xfs_agfl_read_verify(
	struct xfs_buf	*bp)
{
	struct xfs_mount *mp = bp->b_target->bt_mount;
599
	xfs_failaddr_t	fa;
600 601 602 603 604 605 606 607 608 609

	/*
	 * There is no verification of non-crc AGFLs because mkfs does not
	 * initialise the AGFL to zero or NULL. Hence the only valid part of the
	 * AGFL is what the AGF says is active. We can't get to the AGF, so we
	 * can't verify just those entries are valid.
	 */
	if (!xfs_sb_version_hascrc(&mp->m_sb))
		return;

610
	if (!xfs_buf_verify_cksum(bp, XFS_AGFL_CRC_OFF))
611 612 613 614 615 616
		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
	else {
		fa = xfs_agfl_verify(bp);
		if (fa)
			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
	}
617 618
}

619
static void
620 621 622
xfs_agfl_write_verify(
	struct xfs_buf	*bp)
{
C
Carlos Maiolino 已提交
623 624
	struct xfs_mount	*mp = bp->b_target->bt_mount;
	struct xfs_buf_log_item	*bip = bp->b_log_item;
625
	xfs_failaddr_t		fa;
626

627 628 629 630
	/* no verification of non-crc AGFLs */
	if (!xfs_sb_version_hascrc(&mp->m_sb))
		return;

631 632 633
	fa = xfs_agfl_verify(bp);
	if (fa) {
		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
634 635 636 637 638 639
		return;
	}

	if (bip)
		XFS_BUF_TO_AGFL(bp)->agfl_lsn = cpu_to_be64(bip->bli_item.li_lsn);

640
	xfs_buf_update_cksum(bp, XFS_AGFL_CRC_OFF);
641 642
}

643
const struct xfs_buf_ops xfs_agfl_buf_ops = {
644
	.name = "xfs_agfl",
645 646
	.verify_read = xfs_agfl_read_verify,
	.verify_write = xfs_agfl_write_verify,
647
	.verify_struct = xfs_agfl_verify,
648 649
};

L
Linus Torvalds 已提交
650 651 652
/*
 * Read in the allocation group free block array.
 */
653
int					/* error */
L
Linus Torvalds 已提交
654 655 656 657 658 659 660 661 662 663 664 665 666
xfs_alloc_read_agfl(
	xfs_mount_t	*mp,		/* mount point structure */
	xfs_trans_t	*tp,		/* transaction pointer */
	xfs_agnumber_t	agno,		/* allocation group number */
	xfs_buf_t	**bpp)		/* buffer for the ag free block array */
{
	xfs_buf_t	*bp;		/* return value */
	int		error;

	ASSERT(agno != NULLAGNUMBER);
	error = xfs_trans_read_buf(
			mp, tp, mp->m_ddev_targp,
			XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
667
			XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_agfl_buf_ops);
L
Linus Torvalds 已提交
668 669
	if (error)
		return error;
670
	xfs_buf_set_ref(bp, XFS_AGFL_REF);
L
Linus Torvalds 已提交
671 672 673 674
	*bpp = bp;
	return 0;
}

675 676 677 678 679 680 681 682 683 684 685 686 687 688 689
STATIC int
xfs_alloc_update_counters(
	struct xfs_trans	*tp,
	struct xfs_perag	*pag,
	struct xfs_buf		*agbp,
	long			len)
{
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);

	pag->pagf_freeblks += len;
	be32_add_cpu(&agf->agf_freeblks, len);

	xfs_trans_agblocks_delta(tp, len);
	if (unlikely(be32_to_cpu(agf->agf_freeblks) >
		     be32_to_cpu(agf->agf_length)))
D
Dave Chinner 已提交
690
		return -EFSCORRUPTED;
691 692 693 694 695

	xfs_alloc_log_agf(tp, agbp, XFS_AGF_FREEBLKS);
	return 0;
}

L
Linus Torvalds 已提交
696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
/*
 * Allocation group level functions.
 */

/*
 * Allocate a variable extent in the allocation group agno.
 * Type and bno are used to determine where in the allocation group the
 * extent will start.
 * Extent's length (returned in *len) will be between minlen and maxlen,
 * and of the form k * prod + mod unless there's nothing that large.
 * Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
 */
STATIC int			/* error */
xfs_alloc_ag_vextent(
	xfs_alloc_arg_t	*args)	/* argument structure for allocation */
{
	int		error=0;

	ASSERT(args->minlen > 0);
	ASSERT(args->maxlen > 0);
	ASSERT(args->minlen <= args->maxlen);
	ASSERT(args->mod < args->prod);
	ASSERT(args->alignment > 0);
719

L
Linus Torvalds 已提交
720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737
	/*
	 * Branch to correct routine based on the type.
	 */
	args->wasfromfl = 0;
	switch (args->type) {
	case XFS_ALLOCTYPE_THIS_AG:
		error = xfs_alloc_ag_vextent_size(args);
		break;
	case XFS_ALLOCTYPE_NEAR_BNO:
		error = xfs_alloc_ag_vextent_near(args);
		break;
	case XFS_ALLOCTYPE_THIS_BNO:
		error = xfs_alloc_ag_vextent_exact(args);
		break;
	default:
		ASSERT(0);
		/* NOTREACHED */
	}
738 739

	if (error || args->agbno == NULLAGBLOCK)
L
Linus Torvalds 已提交
740
		return error;
741 742 743

	ASSERT(args->len >= args->minlen);
	ASSERT(args->len <= args->maxlen);
744
	ASSERT(!args->wasfromfl || args->resv != XFS_AG_RESV_AGFL);
745 746
	ASSERT(args->agbno % args->alignment == 0);

747
	/* if not file data, insert new block into the reverse map btree */
748
	if (!xfs_rmap_should_skip_owner_update(&args->oinfo)) {
749 750 751 752 753 754
		error = xfs_rmap_alloc(args->tp, args->agbp, args->agno,
				       args->agbno, args->len, &args->oinfo);
		if (error)
			return error;
	}

755 756 757 758 759 760 761
	if (!args->wasfromfl) {
		error = xfs_alloc_update_counters(args->tp, args->pag,
						  args->agbp,
						  -((long)(args->len)));
		if (error)
			return error;

D
Dave Chinner 已提交
762
		ASSERT(!xfs_extent_busy_search(args->mp, args->agno,
763
					      args->agbno, args->len));
L
Linus Torvalds 已提交
764
	}
765

766
	xfs_ag_resv_alloc_extent(args->pag, args->resv, args);
767

768 769
	XFS_STATS_INC(args->mp, xs_allocx);
	XFS_STATS_ADD(args->mp, xs_allocb, args->len);
770
	return error;
L
Linus Torvalds 已提交
771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787
}

/*
 * Allocate a variable extent at exactly agno/bno.
 * Extent's length (returned in *len) will be between minlen and maxlen,
 * and of the form k * prod + mod unless there's nothing that large.
 * Return the starting a.g. block (bno), or NULLAGBLOCK if we can't do it.
 */
STATIC int			/* error */
xfs_alloc_ag_vextent_exact(
	xfs_alloc_arg_t	*args)	/* allocation argument structure */
{
	xfs_btree_cur_t	*bno_cur;/* by block-number btree cursor */
	xfs_btree_cur_t	*cnt_cur;/* by count btree cursor */
	int		error;
	xfs_agblock_t	fbno;	/* start block of found extent */
	xfs_extlen_t	flen;	/* length of found extent */
788 789 790
	xfs_agblock_t	tbno;	/* start block of busy extent */
	xfs_extlen_t	tlen;	/* length of busy extent */
	xfs_agblock_t	tend;	/* end block of busy extent */
L
Linus Torvalds 已提交
791
	int		i;	/* success/failure of operation */
792
	unsigned	busy_gen;
L
Linus Torvalds 已提交
793 794

	ASSERT(args->alignment == 1);
795

L
Linus Torvalds 已提交
796 797 798
	/*
	 * Allocate/initialize a cursor for the by-number freespace btree.
	 */
799
	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
800 801
					  args->agno, XFS_BTNUM_BNO);

L
Linus Torvalds 已提交
802 803 804 805 806
	/*
	 * Lookup bno and minlen in the btree (minlen is irrelevant, really).
	 * Look for the closest free block <= bno, it must contain bno
	 * if any free block does.
	 */
807 808
	error = xfs_alloc_lookup_le(bno_cur, args->agbno, args->minlen, &i);
	if (error)
L
Linus Torvalds 已提交
809
		goto error0;
810 811 812
	if (!i)
		goto not_found;

L
Linus Torvalds 已提交
813 814 815
	/*
	 * Grab the freespace record.
	 */
816 817
	error = xfs_alloc_get_rec(bno_cur, &fbno, &flen, &i);
	if (error)
L
Linus Torvalds 已提交
818
		goto error0;
819
	XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
820
	ASSERT(fbno <= args->agbno);
821

L
Linus Torvalds 已提交
822
	/*
823
	 * Check for overlapping busy extents.
L
Linus Torvalds 已提交
824
	 */
825 826 827
	tbno = fbno;
	tlen = flen;
	xfs_extent_busy_trim(args, &tbno, &tlen, &busy_gen);
828 829 830 831 832 833 834 835 836 837 838

	/*
	 * Give up if the start of the extent is busy, or the freespace isn't
	 * long enough for the minimum request.
	 */
	if (tbno > args->agbno)
		goto not_found;
	if (tlen < args->minlen)
		goto not_found;
	tend = tbno + tlen;
	if (tend < args->agbno + args->minlen)
839 840
		goto not_found;

L
Linus Torvalds 已提交
841 842 843
	/*
	 * End of extent will be smaller of the freespace end and the
	 * maximal requested end.
844
	 *
L
Linus Torvalds 已提交
845 846
	 * Fix the length according to mod and prod if given.
	 */
847 848
	args->len = XFS_AGBLOCK_MIN(tend, args->agbno + args->maxlen)
						- args->agbno;
L
Linus Torvalds 已提交
849
	xfs_alloc_fix_len(args);
850
	ASSERT(args->agbno + args->len <= tend);
851

L
Linus Torvalds 已提交
852
	/*
853
	 * We are allocating agbno for args->len
L
Linus Torvalds 已提交
854 855
	 * Allocate/initialize a cursor for the by-size btree.
	 */
856 857
	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_CNT);
L
Linus Torvalds 已提交
858
	ASSERT(args->agbno + args->len <=
859
		be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
860 861 862
	error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen, args->agbno,
				      args->len, XFSA_FIXUP_BNO_OK);
	if (error) {
L
Linus Torvalds 已提交
863 864 865
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
		goto error0;
	}
866

L
Linus Torvalds 已提交
867 868
	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
C
Christoph Hellwig 已提交
869

L
Linus Torvalds 已提交
870
	args->wasfromfl = 0;
871 872 873 874 875 876 877 878
	trace_xfs_alloc_exact_done(args);
	return 0;

not_found:
	/* Didn't find it, return null. */
	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
	args->agbno = NULLAGBLOCK;
	trace_xfs_alloc_exact_notfound(args);
L
Linus Torvalds 已提交
879 880 881 882
	return 0;

error0:
	xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
C
Christoph Hellwig 已提交
883
	trace_xfs_alloc_exact_error(args);
L
Linus Torvalds 已提交
884 885 886
	return error;
}

887 888 889 890 891 892 893 894 895 896 897
/*
 * Search the btree in a given direction via the search cursor and compare
 * the records found against the good extent we've already found.
 */
STATIC int
xfs_alloc_find_best_extent(
	struct xfs_alloc_arg	*args,	/* allocation argument structure */
	struct xfs_btree_cur	**gcur,	/* good cursor */
	struct xfs_btree_cur	**scur,	/* searching cursor */
	xfs_agblock_t		gdiff,	/* difference for search comparison */
	xfs_agblock_t		*sbno,	/* extent found by search */
898 899 900
	xfs_extlen_t		*slen,	/* extent length */
	xfs_agblock_t		*sbnoa,	/* aligned extent found by search */
	xfs_extlen_t		*slena,	/* aligned extent length */
901 902 903 904 905 906
	int			dir)	/* 0 = search right, 1 = search left */
{
	xfs_agblock_t		new;
	xfs_agblock_t		sdiff;
	int			error;
	int			i;
907
	unsigned		busy_gen;
908 909 910 911 912 913 914 915 916 917 918 919

	/* The good extent is perfect, no need to  search. */
	if (!gdiff)
		goto out_use_good;

	/*
	 * Look until we find a better one, run out of space or run off the end.
	 */
	do {
		error = xfs_alloc_get_rec(*scur, sbno, slen, &i);
		if (error)
			goto error0;
920
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
921 922
		xfs_alloc_compute_aligned(args, *sbno, *slen,
				sbnoa, slena, &busy_gen);
923 924 925 926 927

		/*
		 * The good extent is closer than this one.
		 */
		if (!dir) {
928 929
			if (*sbnoa > args->max_agbno)
				goto out_use_good;
930
			if (*sbnoa >= args->agbno + gdiff)
931 932
				goto out_use_good;
		} else {
933 934
			if (*sbnoa < args->min_agbno)
				goto out_use_good;
935
			if (*sbnoa <= args->agbno - gdiff)
936 937 938 939 940 941 942 943 944 945 946
				goto out_use_good;
		}

		/*
		 * Same distance, compare length and pick the best.
		 */
		if (*slena >= args->minlen) {
			args->len = XFS_EXTLEN_MIN(*slena, args->maxlen);
			xfs_alloc_fix_len(args);

			sdiff = xfs_alloc_compute_diff(args->agbno, args->len,
947
						       args->alignment,
948
						       args->datatype, *sbnoa,
949
						       *slena, &new);
950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981

			/*
			 * Choose closer size and invalidate other cursor.
			 */
			if (sdiff < gdiff)
				goto out_use_search;
			goto out_use_good;
		}

		if (!dir)
			error = xfs_btree_increment(*scur, 0, &i);
		else
			error = xfs_btree_decrement(*scur, 0, &i);
		if (error)
			goto error0;
	} while (i);

out_use_good:
	xfs_btree_del_cursor(*scur, XFS_BTREE_NOERROR);
	*scur = NULL;
	return 0;

out_use_search:
	xfs_btree_del_cursor(*gcur, XFS_BTREE_NOERROR);
	*gcur = NULL;
	return 0;

error0:
	/* caller invalidates cursors */
	return error;
}

L
Linus Torvalds 已提交
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998
/*
 * Allocate a variable extent near bno in the allocation group agno.
 * Extent's length (returned in len) will be between minlen and maxlen,
 * and of the form k * prod + mod unless there's nothing that large.
 * Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
 */
STATIC int				/* error */
xfs_alloc_ag_vextent_near(
	xfs_alloc_arg_t	*args)		/* allocation argument structure */
{
	xfs_btree_cur_t	*bno_cur_gt;	/* cursor for bno btree, right side */
	xfs_btree_cur_t	*bno_cur_lt;	/* cursor for bno btree, left side */
	xfs_btree_cur_t	*cnt_cur;	/* cursor for count btree */
	xfs_agblock_t	gtbno;		/* start bno of right side entry */
	xfs_agblock_t	gtbnoa;		/* aligned ... */
	xfs_extlen_t	gtdiff;		/* difference to right side entry */
	xfs_extlen_t	gtlen;		/* length of right side entry */
999
	xfs_extlen_t	gtlena;		/* aligned ... */
L
Linus Torvalds 已提交
1000 1001 1002 1003 1004 1005 1006 1007
	xfs_agblock_t	gtnew;		/* useful start bno of right side */
	int		error;		/* error code */
	int		i;		/* result code, temporary */
	int		j;		/* result code, temporary */
	xfs_agblock_t	ltbno;		/* start bno of left side entry */
	xfs_agblock_t	ltbnoa;		/* aligned ... */
	xfs_extlen_t	ltdiff;		/* difference to left side entry */
	xfs_extlen_t	ltlen;		/* length of left side entry */
1008
	xfs_extlen_t	ltlena;		/* aligned ... */
L
Linus Torvalds 已提交
1009 1010
	xfs_agblock_t	ltnew;		/* useful start bno of left side */
	xfs_extlen_t	rlen;		/* length of returned extent */
1011 1012
	bool		busy;
	unsigned	busy_gen;
1013
#ifdef DEBUG
L
Linus Torvalds 已提交
1014 1015 1016 1017 1018
	/*
	 * Randomly don't execute the first algorithm.
	 */
	int		dofirst;	/* set to do first algorithm */

1019
	dofirst = prandom_u32() & 1;
L
Linus Torvalds 已提交
1020
#endif
1021

1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
	/* handle unitialized agbno range so caller doesn't have to */
	if (!args->min_agbno && !args->max_agbno)
		args->max_agbno = args->mp->m_sb.sb_agblocks - 1;
	ASSERT(args->min_agbno <= args->max_agbno);

	/* clamp agbno to the range if it's outside */
	if (args->agbno < args->min_agbno)
		args->agbno = args->min_agbno;
	if (args->agbno > args->max_agbno)
		args->agbno = args->max_agbno;

1033 1034 1035 1036 1037 1038
restart:
	bno_cur_lt = NULL;
	bno_cur_gt = NULL;
	ltlen = 0;
	gtlena = 0;
	ltlena = 0;
1039
	busy = false;
1040

L
Linus Torvalds 已提交
1041 1042 1043
	/*
	 * Get a cursor for the by-size btree.
	 */
1044 1045
	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_CNT);
1046

L
Linus Torvalds 已提交
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
	/*
	 * See if there are any free extents as big as maxlen.
	 */
	if ((error = xfs_alloc_lookup_ge(cnt_cur, 0, args->maxlen, &i)))
		goto error0;
	/*
	 * If none, then pick up the last entry in the tree unless the
	 * tree is empty.
	 */
	if (!i) {
		if ((error = xfs_alloc_ag_vextent_small(args, cnt_cur, &ltbno,
				&ltlen, &i)))
			goto error0;
		if (i == 0 || ltlen == 0) {
			xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
1062
			trace_xfs_alloc_near_noentry(args);
L
Linus Torvalds 已提交
1063 1064 1065 1066 1067
			return 0;
		}
		ASSERT(i == 1);
	}
	args->wasfromfl = 0;
1068

L
Linus Torvalds 已提交
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
	/*
	 * First algorithm.
	 * If the requested extent is large wrt the freespaces available
	 * in this a.g., then the cursor will be pointing to a btree entry
	 * near the right edge of the tree.  If it's in the last btree leaf
	 * block, then we just examine all the entries in that block
	 * that are big enough, and pick the best one.
	 * This is written as a while loop so we can break out of it,
	 * but we never loop back to the top.
	 */
	while (xfs_btree_islastblock(cnt_cur, 0)) {
		xfs_extlen_t	bdiff;
		int		besti=0;
		xfs_extlen_t	blen=0;
		xfs_agblock_t	bnew=0;

1085 1086
#ifdef DEBUG
		if (dofirst)
L
Linus Torvalds 已提交
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
			break;
#endif
		/*
		 * Start from the entry that lookup found, sequence through
		 * all larger free blocks.  If we're actually pointing at a
		 * record smaller than maxlen, go to the start of this block,
		 * and skip all those smaller than minlen.
		 */
		if (ltlen || args->alignment > 1) {
			cnt_cur->bc_ptrs[0] = 1;
			do {
				if ((error = xfs_alloc_get_rec(cnt_cur, &ltbno,
						&ltlen, &i)))
					goto error0;
1101
				XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1102 1103
				if (ltlen >= args->minlen)
					break;
1104
				if ((error = xfs_btree_increment(cnt_cur, 0, &i)))
L
Linus Torvalds 已提交
1105 1106 1107 1108 1109 1110 1111 1112 1113
					goto error0;
			} while (i);
			ASSERT(ltlen >= args->minlen);
			if (!i)
				break;
		}
		i = cnt_cur->bc_ptrs[0];
		for (j = 1, blen = 0, bdiff = 0;
		     !error && j && (blen < args->maxlen || bdiff > 0);
1114
		     error = xfs_btree_increment(cnt_cur, 0, &j)) {
L
Linus Torvalds 已提交
1115 1116 1117 1118 1119 1120
			/*
			 * For each entry, decide if it's better than
			 * the previous best entry.
			 */
			if ((error = xfs_alloc_get_rec(cnt_cur, &ltbno, &ltlen, &i)))
				goto error0;
1121
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1122 1123
			busy = xfs_alloc_compute_aligned(args, ltbno, ltlen,
					&ltbnoa, &ltlena, &busy_gen);
1124
			if (ltlena < args->minlen)
L
Linus Torvalds 已提交
1125
				continue;
1126 1127
			if (ltbnoa < args->min_agbno || ltbnoa > args->max_agbno)
				continue;
L
Linus Torvalds 已提交
1128 1129 1130 1131 1132 1133
			args->len = XFS_EXTLEN_MIN(ltlena, args->maxlen);
			xfs_alloc_fix_len(args);
			ASSERT(args->len >= args->minlen);
			if (args->len < blen)
				continue;
			ltdiff = xfs_alloc_compute_diff(args->agbno, args->len,
1134
				args->alignment, args->datatype, ltbnoa,
1135
				ltlena, &ltnew);
L
Linus Torvalds 已提交
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
			if (ltnew != NULLAGBLOCK &&
			    (args->len > blen || ltdiff < bdiff)) {
				bdiff = ltdiff;
				bnew = ltnew;
				blen = args->len;
				besti = cnt_cur->bc_ptrs[0];
			}
		}
		/*
		 * It didn't work.  We COULD be in a case where
		 * there's a good record somewhere, so try again.
		 */
		if (blen == 0)
			break;
		/*
		 * Point at the best entry, and retrieve it again.
		 */
		cnt_cur->bc_ptrs[0] = besti;
		if ((error = xfs_alloc_get_rec(cnt_cur, &ltbno, &ltlen, &i)))
			goto error0;
1156
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1157
		ASSERT(ltbno + ltlen <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
L
Linus Torvalds 已提交
1158
		args->len = blen;
1159

L
Linus Torvalds 已提交
1160 1161 1162 1163 1164
		/*
		 * We are allocating starting at bnew for blen blocks.
		 */
		args->agbno = bnew;
		ASSERT(bnew >= ltbno);
1165
		ASSERT(bnew + blen <= ltbno + ltlen);
L
Linus Torvalds 已提交
1166 1167 1168
		/*
		 * Set up a cursor for the by-bno tree.
		 */
1169 1170
		bno_cur_lt = xfs_allocbt_init_cursor(args->mp, args->tp,
			args->agbp, args->agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1171 1172 1173 1174 1175 1176 1177 1178
		/*
		 * Fix up the btree entries.
		 */
		if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur_lt, ltbno,
				ltlen, bnew, blen, XFSA_FIXUP_CNT_OK)))
			goto error0;
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
		xfs_btree_del_cursor(bno_cur_lt, XFS_BTREE_NOERROR);
C
Christoph Hellwig 已提交
1179 1180

		trace_xfs_alloc_near_first(args);
L
Linus Torvalds 已提交
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
		return 0;
	}
	/*
	 * Second algorithm.
	 * Search in the by-bno tree to the left and to the right
	 * simultaneously, until in each case we find a space big enough,
	 * or run into the edge of the tree.  When we run into the edge,
	 * we deallocate that cursor.
	 * If both searches succeed, we compare the two spaces and pick
	 * the better one.
	 * With alignment, it's possible for both to fail; the upper
	 * level algorithm that picks allocation groups for allocations
	 * is not supposed to do this.
	 */
	/*
	 * Allocate and initialize the cursor for the leftward search.
	 */
1198 1199
	bno_cur_lt = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
	/*
	 * Lookup <= bno to find the leftward search's starting point.
	 */
	if ((error = xfs_alloc_lookup_le(bno_cur_lt, args->agbno, args->maxlen, &i)))
		goto error0;
	if (!i) {
		/*
		 * Didn't find anything; use this cursor for the rightward
		 * search.
		 */
		bno_cur_gt = bno_cur_lt;
		bno_cur_lt = NULL;
	}
	/*
	 * Found something.  Duplicate the cursor for the rightward search.
	 */
	else if ((error = xfs_btree_dup_cursor(bno_cur_lt, &bno_cur_gt)))
		goto error0;
	/*
	 * Increment the cursor, so we will point at the entry just right
	 * of the leftward entry if any, or to the leftmost entry.
	 */
1222
	if ((error = xfs_btree_increment(bno_cur_gt, 0, &i)))
L
Linus Torvalds 已提交
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
		goto error0;
	if (!i) {
		/*
		 * It failed, there are no rightward entries.
		 */
		xfs_btree_del_cursor(bno_cur_gt, XFS_BTREE_NOERROR);
		bno_cur_gt = NULL;
	}
	/*
	 * Loop going left with the leftward cursor, right with the
	 * rightward cursor, until either both directions give up or
	 * we find an entry at least as big as minlen.
	 */
	do {
		if (bno_cur_lt) {
			if ((error = xfs_alloc_get_rec(bno_cur_lt, &ltbno, &ltlen, &i)))
				goto error0;
1240
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1241 1242
			busy |= xfs_alloc_compute_aligned(args, ltbno, ltlen,
					&ltbnoa, &ltlena, &busy_gen);
1243
			if (ltlena >= args->minlen && ltbnoa >= args->min_agbno)
L
Linus Torvalds 已提交
1244
				break;
1245
			if ((error = xfs_btree_decrement(bno_cur_lt, 0, &i)))
L
Linus Torvalds 已提交
1246
				goto error0;
1247
			if (!i || ltbnoa < args->min_agbno) {
L
Linus Torvalds 已提交
1248 1249 1250 1251 1252 1253 1254 1255
				xfs_btree_del_cursor(bno_cur_lt,
						     XFS_BTREE_NOERROR);
				bno_cur_lt = NULL;
			}
		}
		if (bno_cur_gt) {
			if ((error = xfs_alloc_get_rec(bno_cur_gt, &gtbno, &gtlen, &i)))
				goto error0;
1256
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1257 1258
			busy |= xfs_alloc_compute_aligned(args, gtbno, gtlen,
					&gtbnoa, &gtlena, &busy_gen);
1259
			if (gtlena >= args->minlen && gtbnoa <= args->max_agbno)
L
Linus Torvalds 已提交
1260
				break;
1261
			if ((error = xfs_btree_increment(bno_cur_gt, 0, &i)))
L
Linus Torvalds 已提交
1262
				goto error0;
1263
			if (!i || gtbnoa > args->max_agbno) {
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268 1269
				xfs_btree_del_cursor(bno_cur_gt,
						     XFS_BTREE_NOERROR);
				bno_cur_gt = NULL;
			}
		}
	} while (bno_cur_lt || bno_cur_gt);
1270

L
Linus Torvalds 已提交
1271 1272 1273 1274 1275 1276
	/*
	 * Got both cursors still active, need to find better entry.
	 */
	if (bno_cur_lt && bno_cur_gt) {
		if (ltlena >= args->minlen) {
			/*
1277
			 * Left side is good, look for a right side entry.
L
Linus Torvalds 已提交
1278 1279 1280
			 */
			args->len = XFS_EXTLEN_MIN(ltlena, args->maxlen);
			xfs_alloc_fix_len(args);
1281
			ltdiff = xfs_alloc_compute_diff(args->agbno, args->len,
1282
				args->alignment, args->datatype, ltbnoa,
1283
				ltlena, &ltnew);
1284 1285 1286

			error = xfs_alloc_find_best_extent(args,
						&bno_cur_lt, &bno_cur_gt,
1287 1288
						ltdiff, &gtbno, &gtlen,
						&gtbnoa, &gtlena,
1289 1290 1291 1292
						0 /* search right */);
		} else {
			ASSERT(gtlena >= args->minlen);

L
Linus Torvalds 已提交
1293
			/*
1294
			 * Right side is good, look for a left side entry.
L
Linus Torvalds 已提交
1295 1296 1297
			 */
			args->len = XFS_EXTLEN_MIN(gtlena, args->maxlen);
			xfs_alloc_fix_len(args);
1298
			gtdiff = xfs_alloc_compute_diff(args->agbno, args->len,
1299
				args->alignment, args->datatype, gtbnoa,
1300
				gtlena, &gtnew);
1301 1302 1303

			error = xfs_alloc_find_best_extent(args,
						&bno_cur_gt, &bno_cur_lt,
1304 1305
						gtdiff, &ltbno, &ltlen,
						&ltbnoa, &ltlena,
1306
						1 /* search left */);
L
Linus Torvalds 已提交
1307
		}
1308 1309 1310

		if (error)
			goto error0;
L
Linus Torvalds 已提交
1311
	}
1312

L
Linus Torvalds 已提交
1313 1314 1315 1316
	/*
	 * If we couldn't get anything, give up.
	 */
	if (bno_cur_lt == NULL && bno_cur_gt == NULL) {
1317 1318
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);

1319
		if (busy) {
1320
			trace_xfs_alloc_near_busy(args);
1321
			xfs_extent_busy_flush(args->mp, args->pag, busy_gen);
1322 1323
			goto restart;
		}
C
Christoph Hellwig 已提交
1324
		trace_xfs_alloc_size_neither(args);
L
Linus Torvalds 已提交
1325 1326 1327
		args->agbno = NULLAGBLOCK;
		return 0;
	}
1328

L
Linus Torvalds 已提交
1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344
	/*
	 * At this point we have selected a freespace entry, either to the
	 * left or to the right.  If it's on the right, copy all the
	 * useful variables to the "left" set so we only have one
	 * copy of this code.
	 */
	if (bno_cur_gt) {
		bno_cur_lt = bno_cur_gt;
		bno_cur_gt = NULL;
		ltbno = gtbno;
		ltbnoa = gtbnoa;
		ltlen = gtlen;
		ltlena = gtlena;
		j = 1;
	} else
		j = 0;
1345

L
Linus Torvalds 已提交
1346 1347 1348 1349 1350 1351
	/*
	 * Fix up the length and compute the useful address.
	 */
	args->len = XFS_EXTLEN_MIN(ltlena, args->maxlen);
	xfs_alloc_fix_len(args);
	rlen = args->len;
1352
	(void)xfs_alloc_compute_diff(args->agbno, rlen, args->alignment,
1353
				     args->datatype, ltbnoa, ltlena, &ltnew);
L
Linus Torvalds 已提交
1354
	ASSERT(ltnew >= ltbno);
1355
	ASSERT(ltnew + rlen <= ltbnoa + ltlena);
1356
	ASSERT(ltnew + rlen <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
1357
	ASSERT(ltnew >= args->min_agbno && ltnew <= args->max_agbno);
L
Linus Torvalds 已提交
1358
	args->agbno = ltnew;
1359

L
Linus Torvalds 已提交
1360 1361 1362
	if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur_lt, ltbno, ltlen,
			ltnew, rlen, XFSA_FIXUP_BNO_OK)))
		goto error0;
C
Christoph Hellwig 已提交
1363 1364 1365 1366 1367 1368

	if (j)
		trace_xfs_alloc_near_greater(args);
	else
		trace_xfs_alloc_near_lesser(args);

L
Linus Torvalds 已提交
1369 1370 1371 1372 1373
	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
	xfs_btree_del_cursor(bno_cur_lt, XFS_BTREE_NOERROR);
	return 0;

 error0:
C
Christoph Hellwig 已提交
1374
	trace_xfs_alloc_near_error(args);
L
Linus Torvalds 已提交
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
	if (cnt_cur != NULL)
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
	if (bno_cur_lt != NULL)
		xfs_btree_del_cursor(bno_cur_lt, XFS_BTREE_ERROR);
	if (bno_cur_gt != NULL)
		xfs_btree_del_cursor(bno_cur_gt, XFS_BTREE_ERROR);
	return error;
}

/*
 * Allocate a variable extent anywhere in the allocation group agno.
 * Extent's length (returned in len) will be between minlen and maxlen,
 * and of the form k * prod + mod unless there's nothing that large.
 * Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
 */
STATIC int				/* error */
xfs_alloc_ag_vextent_size(
	xfs_alloc_arg_t	*args)		/* allocation argument structure */
{
	xfs_btree_cur_t	*bno_cur;	/* cursor for bno btree */
	xfs_btree_cur_t	*cnt_cur;	/* cursor for cnt btree */
	int		error;		/* error result */
	xfs_agblock_t	fbno;		/* start of found freespace */
	xfs_extlen_t	flen;		/* length of found freespace */
	int		i;		/* temp status variable */
	xfs_agblock_t	rbno;		/* returned block number */
	xfs_extlen_t	rlen;		/* length of returned extent */
1402 1403
	bool		busy;
	unsigned	busy_gen;
L
Linus Torvalds 已提交
1404

1405
restart:
L
Linus Torvalds 已提交
1406 1407 1408
	/*
	 * Allocate and initialize a cursor for the by-size btree.
	 */
1409 1410
	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_CNT);
L
Linus Torvalds 已提交
1411
	bno_cur = NULL;
1412
	busy = false;
1413

L
Linus Torvalds 已提交
1414 1415 1416 1417 1418 1419
	/*
	 * Look for an entry >= maxlen+alignment-1 blocks.
	 */
	if ((error = xfs_alloc_lookup_ge(cnt_cur, 0,
			args->maxlen + args->alignment - 1, &i)))
		goto error0;
1420

L
Linus Torvalds 已提交
1421
	/*
1422 1423 1424 1425
	 * If none then we have to settle for a smaller extent. In the case that
	 * there are no large extents, this will return the last entry in the
	 * tree unless the tree is empty. In the case that there are only busy
	 * large extents, this will return the largest small extent unless there
1426
	 * are no smaller extents available.
L
Linus Torvalds 已提交
1427
	 */
1428
	if (!i) {
1429 1430 1431
		error = xfs_alloc_ag_vextent_small(args, cnt_cur,
						   &fbno, &flen, &i);
		if (error)
L
Linus Torvalds 已提交
1432 1433 1434
			goto error0;
		if (i == 0 || flen == 0) {
			xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
C
Christoph Hellwig 已提交
1435
			trace_xfs_alloc_size_noentry(args);
L
Linus Torvalds 已提交
1436 1437 1438
			return 0;
		}
		ASSERT(i == 1);
1439 1440
		busy = xfs_alloc_compute_aligned(args, fbno, flen, &rbno,
				&rlen, &busy_gen);
1441 1442 1443 1444 1445 1446 1447 1448
	} else {
		/*
		 * Search for a non-busy extent that is large enough.
		 */
		for (;;) {
			error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen, &i);
			if (error)
				goto error0;
1449
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1450

1451 1452
			busy = xfs_alloc_compute_aligned(args, fbno, flen,
					&rbno, &rlen, &busy_gen);
1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463

			if (rlen >= args->maxlen)
				break;

			error = xfs_btree_increment(cnt_cur, 0, &i);
			if (error)
				goto error0;
			if (i == 0) {
				/*
				 * Our only valid extents must have been busy.
				 * Make it unbusy by forcing the log out and
1464
				 * retrying.
1465 1466 1467 1468
				 */
				xfs_btree_del_cursor(cnt_cur,
						     XFS_BTREE_NOERROR);
				trace_xfs_alloc_size_busy(args);
1469 1470
				xfs_extent_busy_flush(args->mp,
							args->pag, busy_gen);
1471 1472 1473
				goto restart;
			}
		}
L
Linus Torvalds 已提交
1474
	}
1475

L
Linus Torvalds 已提交
1476 1477 1478 1479 1480 1481 1482
	/*
	 * In the first case above, we got the last entry in the
	 * by-size btree.  Now we check to see if the space hits maxlen
	 * once aligned; if not, we search left for something better.
	 * This can't happen in the second case above.
	 */
	rlen = XFS_EXTLEN_MIN(args->maxlen, rlen);
1483
	XFS_WANT_CORRUPTED_GOTO(args->mp, rlen == 0 ||
L
Linus Torvalds 已提交
1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495
			(rlen <= flen && rbno + rlen <= fbno + flen), error0);
	if (rlen < args->maxlen) {
		xfs_agblock_t	bestfbno;
		xfs_extlen_t	bestflen;
		xfs_agblock_t	bestrbno;
		xfs_extlen_t	bestrlen;

		bestrlen = rlen;
		bestrbno = rbno;
		bestflen = flen;
		bestfbno = fbno;
		for (;;) {
1496
			if ((error = xfs_btree_decrement(cnt_cur, 0, &i)))
L
Linus Torvalds 已提交
1497 1498 1499 1500 1501 1502
				goto error0;
			if (i == 0)
				break;
			if ((error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen,
					&i)))
				goto error0;
1503
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1504 1505
			if (flen < bestrlen)
				break;
1506 1507
			busy = xfs_alloc_compute_aligned(args, fbno, flen,
					&rbno, &rlen, &busy_gen);
L
Linus Torvalds 已提交
1508
			rlen = XFS_EXTLEN_MIN(args->maxlen, rlen);
1509
			XFS_WANT_CORRUPTED_GOTO(args->mp, rlen == 0 ||
L
Linus Torvalds 已提交
1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
				(rlen <= flen && rbno + rlen <= fbno + flen),
				error0);
			if (rlen > bestrlen) {
				bestrlen = rlen;
				bestrbno = rbno;
				bestflen = flen;
				bestfbno = fbno;
				if (rlen == args->maxlen)
					break;
			}
		}
		if ((error = xfs_alloc_lookup_eq(cnt_cur, bestfbno, bestflen,
				&i)))
			goto error0;
1524
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
		rlen = bestrlen;
		rbno = bestrbno;
		flen = bestflen;
		fbno = bestfbno;
	}
	args->wasfromfl = 0;
	/*
	 * Fix up the length.
	 */
	args->len = rlen;
1535
	if (rlen < args->minlen) {
1536
		if (busy) {
1537 1538
			xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
			trace_xfs_alloc_size_busy(args);
1539
			xfs_extent_busy_flush(args->mp, args->pag, busy_gen);
1540 1541 1542
			goto restart;
		}
		goto out_nominleft;
L
Linus Torvalds 已提交
1543
	}
1544 1545
	xfs_alloc_fix_len(args);

L
Linus Torvalds 已提交
1546
	rlen = args->len;
1547
	XFS_WANT_CORRUPTED_GOTO(args->mp, rlen <= flen, error0);
L
Linus Torvalds 已提交
1548 1549 1550
	/*
	 * Allocate and initialize a cursor for the by-block tree.
	 */
1551 1552
	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1553 1554 1555 1556 1557 1558 1559 1560
	if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen,
			rbno, rlen, XFSA_FIXUP_CNT_OK)))
		goto error0;
	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
	cnt_cur = bno_cur = NULL;
	args->len = rlen;
	args->agbno = rbno;
1561
	XFS_WANT_CORRUPTED_GOTO(args->mp,
L
Linus Torvalds 已提交
1562
		args->agbno + args->len <=
1563
			be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length),
L
Linus Torvalds 已提交
1564
		error0);
C
Christoph Hellwig 已提交
1565
	trace_xfs_alloc_size_done(args);
L
Linus Torvalds 已提交
1566 1567 1568
	return 0;

error0:
C
Christoph Hellwig 已提交
1569
	trace_xfs_alloc_size_error(args);
L
Linus Torvalds 已提交
1570 1571 1572 1573 1574
	if (cnt_cur)
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
	if (bno_cur)
		xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
	return error;
1575 1576 1577 1578 1579 1580

out_nominleft:
	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
	trace_xfs_alloc_size_nominleft(args);
	args->agbno = NULLAGBLOCK;
	return 0;
L
Linus Torvalds 已提交
1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
}

/*
 * Deal with the case where only small freespaces remain.
 * Either return the contents of the last freespace record,
 * or allocate space from the freelist if there is nothing in the tree.
 */
STATIC int			/* error */
xfs_alloc_ag_vextent_small(
	xfs_alloc_arg_t	*args,	/* allocation argument structure */
	xfs_btree_cur_t	*ccur,	/* by-size cursor */
	xfs_agblock_t	*fbnop,	/* result block number */
	xfs_extlen_t	*flenp,	/* result length */
	int		*stat)	/* status: 0-freelist, 1-normal/none */
{
1596
	struct xfs_owner_info	oinfo;
L
Linus Torvalds 已提交
1597 1598 1599 1600 1601
	int		error;
	xfs_agblock_t	fbno;
	xfs_extlen_t	flen;
	int		i;

1602
	if ((error = xfs_btree_decrement(ccur, 0, &i)))
L
Linus Torvalds 已提交
1603 1604 1605 1606
		goto error0;
	if (i) {
		if ((error = xfs_alloc_get_rec(ccur, &fbno, &flen, &i)))
			goto error0;
1607
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1608 1609 1610 1611 1612 1613
	}
	/*
	 * Nothing in the btree, try the freelist.  Make sure
	 * to respect minleft even when pulling from the
	 * freelist.
	 */
1614
	else if (args->minlen == 1 && args->alignment == 1 &&
1615
		 args->resv != XFS_AG_RESV_AGFL &&
1616 1617
		 (be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_flcount)
		  > args->minleft)) {
D
David Chinner 已提交
1618 1619
		error = xfs_alloc_get_freelist(args->tp, args->agbp, &fbno, 0);
		if (error)
L
Linus Torvalds 已提交
1620 1621
			goto error0;
		if (fbno != NULLAGBLOCK) {
D
Dave Chinner 已提交
1622
			xfs_extent_busy_reuse(args->mp, args->agno, fbno, 1,
1623
			      xfs_alloc_allow_busy_reuse(args->datatype));
1624

1625
			if (xfs_alloc_is_userdata(args->datatype)) {
L
Linus Torvalds 已提交
1626 1627 1628 1629
				xfs_buf_t	*bp;

				bp = xfs_btree_get_bufs(args->mp, args->tp,
					args->agno, fbno, 0);
1630 1631 1632 1633
				if (!bp) {
					error = -EFSCORRUPTED;
					goto error0;
				}
L
Linus Torvalds 已提交
1634 1635 1636 1637
				xfs_trans_binval(args->tp, bp);
			}
			args->len = 1;
			args->agbno = fbno;
1638
			XFS_WANT_CORRUPTED_GOTO(args->mp,
L
Linus Torvalds 已提交
1639
				args->agbno + args->len <=
1640
				be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length),
L
Linus Torvalds 已提交
1641 1642
				error0);
			args->wasfromfl = 1;
C
Christoph Hellwig 已提交
1643
			trace_xfs_alloc_small_freelist(args);
1644 1645 1646 1647

			/*
			 * If we're feeding an AGFL block to something that
			 * doesn't live in the free space, we need to clear
1648
			 * out the OWN_AG rmap.
1649 1650 1651 1652 1653 1654 1655
			 */
			xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
			error = xfs_rmap_free(args->tp, args->agbp, args->agno,
					fbno, 1, &oinfo);
			if (error)
				goto error0;

L
Linus Torvalds 已提交
1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
			*stat = 0;
			return 0;
		}
		/*
		 * Nothing in the freelist.
		 */
		else
			flen = 0;
	}
	/*
	 * Can't allocate from the freelist for some reason.
	 */
1668 1669
	else {
		fbno = NULLAGBLOCK;
L
Linus Torvalds 已提交
1670
		flen = 0;
1671
	}
L
Linus Torvalds 已提交
1672 1673 1674 1675 1676
	/*
	 * Can't do the allocation, give up.
	 */
	if (flen < args->minlen) {
		args->agbno = NULLAGBLOCK;
C
Christoph Hellwig 已提交
1677
		trace_xfs_alloc_small_notenough(args);
L
Linus Torvalds 已提交
1678 1679 1680 1681 1682
		flen = 0;
	}
	*fbnop = fbno;
	*flenp = flen;
	*stat = 1;
C
Christoph Hellwig 已提交
1683
	trace_xfs_alloc_small_done(args);
L
Linus Torvalds 已提交
1684 1685 1686
	return 0;

error0:
C
Christoph Hellwig 已提交
1687
	trace_xfs_alloc_small_error(args);
L
Linus Torvalds 已提交
1688 1689 1690 1691 1692 1693
	return error;
}

/*
 * Free the extent starting at agno/bno for length.
 */
1694
STATIC int
L
Linus Torvalds 已提交
1695
xfs_free_ag_extent(
1696 1697 1698 1699 1700 1701
	xfs_trans_t		*tp,
	xfs_buf_t		*agbp,
	xfs_agnumber_t		agno,
	xfs_agblock_t		bno,
	xfs_extlen_t		len,
	struct xfs_owner_info	*oinfo,
1702
	enum xfs_ag_resv_type	type)
L
Linus Torvalds 已提交
1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
{
	xfs_btree_cur_t	*bno_cur;	/* cursor for by-block btree */
	xfs_btree_cur_t	*cnt_cur;	/* cursor for by-size btree */
	int		error;		/* error return value */
	xfs_agblock_t	gtbno;		/* start of right neighbor block */
	xfs_extlen_t	gtlen;		/* length of right neighbor block */
	int		haveleft;	/* have a left neighbor block */
	int		haveright;	/* have a right neighbor block */
	int		i;		/* temp, result code */
	xfs_agblock_t	ltbno;		/* start of left neighbor block */
	xfs_extlen_t	ltlen;		/* length of left neighbor block */
	xfs_mount_t	*mp;		/* mount point struct for filesystem */
	xfs_agblock_t	nbno;		/* new starting block of freespace */
	xfs_extlen_t	nlen;		/* new length of freespace */
1717
	xfs_perag_t	*pag;		/* per allocation group data */
L
Linus Torvalds 已提交
1718

1719
	bno_cur = cnt_cur = NULL;
L
Linus Torvalds 已提交
1720
	mp = tp->t_mountp;
1721

1722
	if (!xfs_rmap_should_skip_owner_update(oinfo)) {
1723 1724 1725 1726 1727
		error = xfs_rmap_free(tp, agbp, agno, bno, len, oinfo);
		if (error)
			goto error0;
	}

L
Linus Torvalds 已提交
1728 1729 1730
	/*
	 * Allocate and initialize a cursor for the by-block btree.
	 */
1731
	bno_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
	/*
	 * Look for a neighboring block on the left (lower block numbers)
	 * that is contiguous with this space.
	 */
	if ((error = xfs_alloc_lookup_le(bno_cur, bno, len, &haveleft)))
		goto error0;
	if (haveleft) {
		/*
		 * There is a block to our left.
		 */
		if ((error = xfs_alloc_get_rec(bno_cur, &ltbno, &ltlen, &i)))
			goto error0;
1744
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755
		/*
		 * It's not contiguous, though.
		 */
		if (ltbno + ltlen < bno)
			haveleft = 0;
		else {
			/*
			 * If this failure happens the request to free this
			 * space was invalid, it's (partly) already free.
			 * Very bad.
			 */
1756 1757
			XFS_WANT_CORRUPTED_GOTO(mp,
						ltbno + ltlen <= bno, error0);
L
Linus Torvalds 已提交
1758 1759 1760 1761 1762 1763
		}
	}
	/*
	 * Look for a neighboring block on the right (higher block numbers)
	 * that is contiguous with this space.
	 */
1764
	if ((error = xfs_btree_increment(bno_cur, 0, &haveright)))
L
Linus Torvalds 已提交
1765 1766 1767 1768 1769 1770 1771
		goto error0;
	if (haveright) {
		/*
		 * There is a block to our right.
		 */
		if ((error = xfs_alloc_get_rec(bno_cur, &gtbno, &gtlen, &i)))
			goto error0;
1772
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783
		/*
		 * It's not contiguous, though.
		 */
		if (bno + len < gtbno)
			haveright = 0;
		else {
			/*
			 * If this failure happens the request to free this
			 * space was invalid, it's (partly) already free.
			 * Very bad.
			 */
1784
			XFS_WANT_CORRUPTED_GOTO(mp, gtbno >= bno + len, error0);
L
Linus Torvalds 已提交
1785 1786 1787 1788 1789
		}
	}
	/*
	 * Now allocate and initialize a cursor for the by-size tree.
	 */
1790
	cnt_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_CNT);
L
Linus Torvalds 已提交
1791 1792 1793 1794 1795 1796 1797 1798 1799 1800
	/*
	 * Have both left and right contiguous neighbors.
	 * Merge all three into a single free block.
	 */
	if (haveleft && haveright) {
		/*
		 * Delete the old by-size entry on the left.
		 */
		if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
			goto error0;
1801
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1802
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1803
			goto error0;
1804
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1805 1806 1807 1808 1809
		/*
		 * Delete the old by-size entry on the right.
		 */
		if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
			goto error0;
1810
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1811
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1812
			goto error0;
1813
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1814 1815 1816
		/*
		 * Delete the old by-block entry for the right block.
		 */
1817
		if ((error = xfs_btree_delete(bno_cur, &i)))
L
Linus Torvalds 已提交
1818
			goto error0;
1819
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1820 1821 1822
		/*
		 * Move the by-block cursor back to the left neighbor.
		 */
1823
		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
L
Linus Torvalds 已提交
1824
			goto error0;
1825
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837
#ifdef DEBUG
		/*
		 * Check that this is the right record: delete didn't
		 * mangle the cursor.
		 */
		{
			xfs_agblock_t	xxbno;
			xfs_extlen_t	xxlen;

			if ((error = xfs_alloc_get_rec(bno_cur, &xxbno, &xxlen,
					&i)))
				goto error0;
1838
			XFS_WANT_CORRUPTED_GOTO(mp,
L
Linus Torvalds 已提交
1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
				i == 1 && xxbno == ltbno && xxlen == ltlen,
				error0);
		}
#endif
		/*
		 * Update remaining by-block entry to the new, joined block.
		 */
		nbno = ltbno;
		nlen = len + ltlen + gtlen;
		if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
			goto error0;
	}
	/*
	 * Have only a left contiguous neighbor.
	 * Merge it together with the new freespace.
	 */
	else if (haveleft) {
		/*
		 * Delete the old by-size entry on the left.
		 */
		if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
			goto error0;
1861
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1862
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1863
			goto error0;
1864
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1865 1866 1867 1868
		/*
		 * Back up the by-block cursor to the left neighbor, and
		 * update its length.
		 */
1869
		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
L
Linus Torvalds 已提交
1870
			goto error0;
1871
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886
		nbno = ltbno;
		nlen = len + ltlen;
		if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
			goto error0;
	}
	/*
	 * Have only a right contiguous neighbor.
	 * Merge it together with the new freespace.
	 */
	else if (haveright) {
		/*
		 * Delete the old by-size entry on the right.
		 */
		if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
			goto error0;
1887
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1888
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1889
			goto error0;
1890
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
		/*
		 * Update the starting block and length of the right
		 * neighbor in the by-block tree.
		 */
		nbno = bno;
		nlen = len + gtlen;
		if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
			goto error0;
	}
	/*
	 * No contiguous neighbors.
	 * Insert the new freespace into the by-block tree.
	 */
	else {
		nbno = bno;
		nlen = len;
1907
		if ((error = xfs_btree_insert(bno_cur, &i)))
L
Linus Torvalds 已提交
1908
			goto error0;
1909
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1910 1911 1912 1913 1914 1915 1916 1917
	}
	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
	bno_cur = NULL;
	/*
	 * In all cases we need to insert the new freespace in the by-size tree.
	 */
	if ((error = xfs_alloc_lookup_eq(cnt_cur, nbno, nlen, &i)))
		goto error0;
1918
	XFS_WANT_CORRUPTED_GOTO(mp, i == 0, error0);
1919
	if ((error = xfs_btree_insert(cnt_cur, &i)))
L
Linus Torvalds 已提交
1920
		goto error0;
1921
	XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1922 1923
	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
	cnt_cur = NULL;
1924

L
Linus Torvalds 已提交
1925 1926 1927
	/*
	 * Update the freespace totals in the ag and superblock.
	 */
1928 1929
	pag = xfs_perag_get(mp, agno);
	error = xfs_alloc_update_counters(tp, pag, agbp, len);
1930
	xfs_ag_resv_free_extent(pag, type, tp, len);
1931 1932 1933 1934
	xfs_perag_put(pag);
	if (error)
		goto error0;

1935 1936
	XFS_STATS_INC(mp, xs_freex);
	XFS_STATS_ADD(mp, xs_freeb, len);
C
Christoph Hellwig 已提交
1937

1938
	trace_xfs_free_extent(mp, agno, bno, len, type, haveleft, haveright);
L
Linus Torvalds 已提交
1939 1940 1941 1942

	return 0;

 error0:
1943
	trace_xfs_free_extent(mp, agno, bno, len, type, -1, -1);
L
Linus Torvalds 已提交
1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962
	if (bno_cur)
		xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
	if (cnt_cur)
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
	return error;
}

/*
 * Visible (exported) allocation/free functions.
 * Some of these are used just by xfs_alloc_btree.c and this file.
 */

/*
 * Compute and fill in value of m_ag_maxlevels.
 */
void
xfs_alloc_compute_maxlevels(
	xfs_mount_t	*mp)	/* file system mount structure */
{
1963
	mp->m_ag_maxlevels = xfs_btree_compute_maxlevels(mp->m_alloc_mnr,
1964
			(mp->m_sb.sb_agblocks + 1) / 2);
L
Linus Torvalds 已提交
1965 1966
}

1967
/*
1968 1969 1970 1971
 * Find the length of the longest extent in an AG.  The 'need' parameter
 * specifies how much space we're going to need for the AGFL and the
 * 'reserved' parameter tells us how many blocks in this AG are reserved for
 * other callers.
1972 1973 1974
 */
xfs_extlen_t
xfs_alloc_longest_free_extent(
1975
	struct xfs_perag	*pag,
1976 1977
	xfs_extlen_t		need,
	xfs_extlen_t		reserved)
1978
{
1979
	xfs_extlen_t		delta = 0;
1980

1981 1982 1983 1984
	/*
	 * If the AGFL needs a recharge, we'll have to subtract that from the
	 * longest extent.
	 */
1985 1986 1987
	if (need > pag->pagf_flcount)
		delta = need - pag->pagf_flcount;

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
	/*
	 * If we cannot maintain others' reservations with space from the
	 * not-longest freesp extents, we'll have to subtract /that/ from
	 * the longest extent too.
	 */
	if (pag->pagf_freeblks - pag->pagf_longest < reserved)
		delta += reserved - (pag->pagf_freeblks - pag->pagf_longest);

	/*
	 * If the longest extent is long enough to satisfy all the
	 * reservations and AGFL rules in place, we can return this extent.
	 */
2000 2001
	if (pag->pagf_longest > delta)
		return pag->pagf_longest - delta;
2002 2003

	/* Otherwise, let the caller try for 1 block if there's space. */
2004 2005 2006
	return pag->pagf_flcount > 0 || pag->pagf_longest > 0;
}

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
unsigned int
xfs_alloc_min_freelist(
	struct xfs_mount	*mp,
	struct xfs_perag	*pag)
{
	unsigned int		min_free;

	/* space needed by-bno freespace btree */
	min_free = min_t(unsigned int, pag->pagf_levels[XFS_BTNUM_BNOi] + 1,
				       mp->m_ag_maxlevels);
	/* space needed by-size freespace btree */
	min_free += min_t(unsigned int, pag->pagf_levels[XFS_BTNUM_CNTi] + 1,
				       mp->m_ag_maxlevels);
2020 2021 2022 2023 2024
	/* space needed reverse mapping used space btree */
	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
		min_free += min_t(unsigned int,
				  pag->pagf_levels[XFS_BTNUM_RMAPi] + 1,
				  mp->m_rmap_maxlevels);
2025 2026 2027 2028

	return min_free;
}

2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
/*
 * Check if the operation we are fixing up the freelist for should go ahead or
 * not. If we are freeing blocks, we always allow it, otherwise the allocation
 * is dependent on whether the size and shape of free space available will
 * permit the requested allocation to take place.
 */
static bool
xfs_alloc_space_available(
	struct xfs_alloc_arg	*args,
	xfs_extlen_t		min_free,
	int			flags)
{
	struct xfs_perag	*pag = args->pag;
2042
	xfs_extlen_t		alloc_len, longest;
2043
	xfs_extlen_t		reservation; /* blocks that are still reserved */
2044 2045 2046 2047 2048
	int			available;

	if (flags & XFS_ALLOC_FLAG_FREEING)
		return true;

2049 2050
	reservation = xfs_ag_resv_needed(pag, args->resv);

2051
	/* do we have enough contiguous free space for the allocation? */
2052
	alloc_len = args->minlen + (args->alignment - 1) + args->minalignslop;
2053
	longest = xfs_alloc_longest_free_extent(pag, min_free, reservation);
2054
	if (longest < alloc_len)
2055 2056
		return false;

2057
	/* do we have enough free space remaining for the allocation? */
2058
	available = (int)(pag->pagf_freeblks + pag->pagf_flcount -
2059
			  reservation - min_free - args->minleft);
2060
	if (available < (int)max(args->total, alloc_len))
2061 2062
		return false;

2063 2064 2065 2066 2067 2068 2069 2070 2071 2072
	/*
	 * Clamp maxlen to the amount of free space available for the actual
	 * extent allocation.
	 */
	if (available < (int)args->maxlen && !(flags & XFS_ALLOC_FLAG_CHECK)) {
		args->maxlen = available;
		ASSERT(args->maxlen > 0);
		ASSERT(args->maxlen >= args->minlen);
	}

2073 2074 2075
	return true;
}

2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099
int
xfs_free_agfl_block(
	struct xfs_trans	*tp,
	xfs_agnumber_t		agno,
	xfs_agblock_t		agbno,
	struct xfs_buf		*agbp,
	struct xfs_owner_info	*oinfo)
{
	int			error;
	struct xfs_buf		*bp;

	error = xfs_free_ag_extent(tp, agbp, agno, agbno, 1, oinfo,
				   XFS_AG_RESV_AGFL);
	if (error)
		return error;

	bp = xfs_btree_get_bufs(tp->t_mountp, tp, agno, agbno, 0);
	if (!bp)
		return -EFSCORRUPTED;
	xfs_trans_binval(tp, bp);

	return 0;
}

2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186
/*
 * Check the agfl fields of the agf for inconsistency or corruption. The purpose
 * is to detect an agfl header padding mismatch between current and early v5
 * kernels. This problem manifests as a 1-slot size difference between the
 * on-disk flcount and the active [first, last] range of a wrapped agfl. This
 * may also catch variants of agfl count corruption unrelated to padding. Either
 * way, we'll reset the agfl and warn the user.
 *
 * Return true if a reset is required before the agfl can be used, false
 * otherwise.
 */
static bool
xfs_agfl_needs_reset(
	struct xfs_mount	*mp,
	struct xfs_agf		*agf)
{
	uint32_t		f = be32_to_cpu(agf->agf_flfirst);
	uint32_t		l = be32_to_cpu(agf->agf_fllast);
	uint32_t		c = be32_to_cpu(agf->agf_flcount);
	int			agfl_size = xfs_agfl_size(mp);
	int			active;

	/* no agfl header on v4 supers */
	if (!xfs_sb_version_hascrc(&mp->m_sb))
		return false;

	/*
	 * The agf read verifier catches severe corruption of these fields.
	 * Repeat some sanity checks to cover a packed -> unpacked mismatch if
	 * the verifier allows it.
	 */
	if (f >= agfl_size || l >= agfl_size)
		return true;
	if (c > agfl_size)
		return true;

	/*
	 * Check consistency between the on-disk count and the active range. An
	 * agfl padding mismatch manifests as an inconsistent flcount.
	 */
	if (c && l >= f)
		active = l - f + 1;
	else if (c)
		active = agfl_size - f + l + 1;
	else
		active = 0;

	return active != c;
}

/*
 * Reset the agfl to an empty state. Ignore/drop any existing blocks since the
 * agfl content cannot be trusted. Warn the user that a repair is required to
 * recover leaked blocks.
 *
 * The purpose of this mechanism is to handle filesystems affected by the agfl
 * header padding mismatch problem. A reset keeps the filesystem online with a
 * relatively minor free space accounting inconsistency rather than suffer the
 * inevitable crash from use of an invalid agfl block.
 */
static void
xfs_agfl_reset(
	struct xfs_trans	*tp,
	struct xfs_buf		*agbp,
	struct xfs_perag	*pag)
{
	struct xfs_mount	*mp = tp->t_mountp;
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);

	ASSERT(pag->pagf_agflreset);
	trace_xfs_agfl_reset(mp, agf, 0, _RET_IP_);

	xfs_warn(mp,
	       "WARNING: Reset corrupted AGFL on AG %u. %d blocks leaked. "
	       "Please unmount and run xfs_repair.",
	         pag->pag_agno, pag->pagf_flcount);

	agf->agf_flfirst = 0;
	agf->agf_fllast = cpu_to_be32(xfs_agfl_size(mp) - 1);
	agf->agf_flcount = 0;
	xfs_alloc_log_agf(tp, agbp, XFS_AGF_FLFIRST | XFS_AGF_FLLAST |
				    XFS_AGF_FLCOUNT);

	pag->pagf_flcount = 0;
	pag->pagf_agflreset = false;
}

2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220
/*
 * Defer an AGFL block free. This is effectively equivalent to
 * xfs_bmap_add_free() with some special handling particular to AGFL blocks.
 *
 * Deferring AGFL frees helps prevent log reservation overruns due to too many
 * allocation operations in a transaction. AGFL frees are prone to this problem
 * because for one they are always freed one at a time. Further, an immediate
 * AGFL block free can cause a btree join and require another block free before
 * the real allocation can proceed. Deferring the free disconnects freeing up
 * the AGFL slot from freeing the block.
 */
STATIC void
xfs_defer_agfl_block(
	struct xfs_mount		*mp,
	struct xfs_defer_ops		*dfops,
	xfs_agnumber_t			agno,
	xfs_fsblock_t			agbno,
	struct xfs_owner_info		*oinfo)
{
	struct xfs_extent_free_item	*new;		/* new element */

	ASSERT(xfs_bmap_free_item_zone != NULL);
	ASSERT(oinfo != NULL);

	new = kmem_zone_alloc(xfs_bmap_free_item_zone, KM_SLEEP);
	new->xefi_startblock = XFS_AGB_TO_FSB(mp, agno, agbno);
	new->xefi_blockcount = 1;
	new->xefi_oinfo = *oinfo;

	trace_xfs_agfl_free_defer(mp, agno, 0, agbno, 1);

	xfs_defer_add(dfops, XFS_DEFER_OPS_TYPE_AGFL_FREE, &new->xefi_list);
}

L
Linus Torvalds 已提交
2221 2222 2223 2224
/*
 * Decide whether to use this allocation group for this allocation.
 * If so, fix up the btree freelist's size.
 */
2225
int			/* error */
L
Linus Torvalds 已提交
2226
xfs_alloc_fix_freelist(
2227 2228
	struct xfs_alloc_arg	*args,	/* allocation argument structure */
	int			flags)	/* XFS_ALLOC_FLAG_... */
L
Linus Torvalds 已提交
2229
{
2230 2231 2232 2233 2234 2235 2236 2237
	struct xfs_mount	*mp = args->mp;
	struct xfs_perag	*pag = args->pag;
	struct xfs_trans	*tp = args->tp;
	struct xfs_buf		*agbp = NULL;
	struct xfs_buf		*agflbp = NULL;
	struct xfs_alloc_arg	targs;	/* local allocation arguments */
	xfs_agblock_t		bno;	/* freelist block */
	xfs_extlen_t		need;	/* total blocks needed in freelist */
2238
	int			error = 0;
2239

L
Linus Torvalds 已提交
2240
	if (!pag->pagf_init) {
2241 2242 2243
		error = xfs_alloc_read_agf(mp, tp, args->agno, flags, &agbp);
		if (error)
			goto out_no_agbp;
L
Linus Torvalds 已提交
2244
		if (!pag->pagf_init) {
2245 2246
			ASSERT(flags & XFS_ALLOC_FLAG_TRYLOCK);
			ASSERT(!(flags & XFS_ALLOC_FLAG_FREEING));
2247
			goto out_agbp_relse;
L
Linus Torvalds 已提交
2248
		}
2249
	}
L
Linus Torvalds 已提交
2250

2251
	/*
2252 2253 2254
	 * If this is a metadata preferred pag and we are user data then try
	 * somewhere else if we are not being asked to try harder at this
	 * point
L
Linus Torvalds 已提交
2255
	 */
2256
	if (pag->pagf_metadata && xfs_alloc_is_userdata(args->datatype) &&
2257 2258
	    (flags & XFS_ALLOC_FLAG_TRYLOCK)) {
		ASSERT(!(flags & XFS_ALLOC_FLAG_FREEING));
2259
		goto out_agbp_relse;
L
Linus Torvalds 已提交
2260 2261
	}

2262
	need = xfs_alloc_min_freelist(mp, pag);
2263 2264
	if (!xfs_alloc_space_available(args, need, flags |
			XFS_ALLOC_FLAG_CHECK))
2265
		goto out_agbp_relse;
2266

L
Linus Torvalds 已提交
2267 2268 2269 2270
	/*
	 * Get the a.g. freespace buffer.
	 * Can fail if we're not blocking on locks, and it's held.
	 */
2271 2272 2273 2274 2275
	if (!agbp) {
		error = xfs_alloc_read_agf(mp, tp, args->agno, flags, &agbp);
		if (error)
			goto out_no_agbp;
		if (!agbp) {
2276 2277
			ASSERT(flags & XFS_ALLOC_FLAG_TRYLOCK);
			ASSERT(!(flags & XFS_ALLOC_FLAG_FREEING));
2278
			goto out_no_agbp;
2279
		}
L
Linus Torvalds 已提交
2280
	}
2281

2282 2283 2284 2285
	/* reset a padding mismatched agfl before final free space check */
	if (pag->pagf_agflreset)
		xfs_agfl_reset(tp, agbp, pag);

2286
	/* If there isn't enough total space or single-extent, reject it. */
2287
	need = xfs_alloc_min_freelist(mp, pag);
2288 2289
	if (!xfs_alloc_space_available(args, need, flags))
		goto out_agbp_relse;
2290

L
Linus Torvalds 已提交
2291 2292
	/*
	 * Make the freelist shorter if it's too long.
2293
	 *
2294 2295 2296 2297 2298 2299 2300 2301
	 * Note that from this point onwards, we will always release the agf and
	 * agfl buffers on error. This handles the case where we error out and
	 * the buffers are clean or may not have been joined to the transaction
	 * and hence need to be released manually. If they have been joined to
	 * the transaction, then xfs_trans_brelse() will handle them
	 * appropriately based on the recursion count and dirty state of the
	 * buffer.
	 *
2302 2303 2304 2305
	 * XXX (dgc): When we have lots of free space, does this buy us
	 * anything other than extra overhead when we need to put more blocks
	 * back on the free list? Maybe we should only do this when space is
	 * getting low or the AGFL is more than half full?
2306 2307 2308 2309 2310 2311 2312 2313
	 *
	 * The NOSHRINK flag prevents the AGFL from being shrunk if it's too
	 * big; the NORMAP flag prevents AGFL expand/shrink operations from
	 * updating the rmapbt.  Both flags are used in xfs_repair while we're
	 * rebuilding the rmapbt, and neither are used by the kernel.  They're
	 * both required to ensure that rmaps are correctly recorded for the
	 * regenerated AGFL, bnobt, and cntbt.  See repair/phase5.c and
	 * repair/rmap.c in xfsprogs for details.
L
Linus Torvalds 已提交
2314
	 */
2315 2316 2317 2318 2319 2320
	memset(&targs, 0, sizeof(targs));
	if (flags & XFS_ALLOC_FLAG_NORMAP)
		xfs_rmap_skip_owner_update(&targs.oinfo);
	else
		xfs_rmap_ag_owner(&targs.oinfo, XFS_RMAP_OWN_AG);
	while (!(flags & XFS_ALLOC_FLAG_NOSHRINK) && pag->pagf_flcount > need) {
D
David Chinner 已提交
2321 2322
		error = xfs_alloc_get_freelist(tp, agbp, &bno, 0);
		if (error)
2323
			goto out_agbp_relse;
2324

2325
		/* defer agfl frees if dfops is provided */
2326 2327
		if (tp->t_dfops) {
			xfs_defer_agfl_block(mp, tp->t_dfops, args->agno,
2328 2329 2330 2331 2332 2333 2334
					     bno, &targs.oinfo);
		} else {
			error = xfs_free_agfl_block(tp, args->agno, bno, agbp,
						    &targs.oinfo);
			if (error)
				goto out_agbp_relse;
		}
L
Linus Torvalds 已提交
2335
	}
2336

L
Linus Torvalds 已提交
2337 2338 2339 2340
	targs.tp = tp;
	targs.mp = mp;
	targs.agbp = agbp;
	targs.agno = args->agno;
2341
	targs.alignment = targs.minlen = targs.prod = 1;
L
Linus Torvalds 已提交
2342 2343
	targs.type = XFS_ALLOCTYPE_THIS_AG;
	targs.pag = pag;
2344 2345
	error = xfs_alloc_read_agfl(mp, tp, targs.agno, &agflbp);
	if (error)
2346
		goto out_agbp_relse;
2347 2348 2349

	/* Make the freelist longer if it's too short. */
	while (pag->pagf_flcount < need) {
L
Linus Torvalds 已提交
2350
		targs.agbno = 0;
2351
		targs.maxlen = need - pag->pagf_flcount;
2352
		targs.resv = XFS_AG_RESV_AGFL;
2353 2354 2355

		/* Allocate as many blocks as possible at once. */
		error = xfs_alloc_ag_vextent(&targs);
2356 2357 2358
		if (error)
			goto out_agflbp_relse;

L
Linus Torvalds 已提交
2359 2360 2361 2362 2363
		/*
		 * Stop if we run out.  Won't happen if callers are obeying
		 * the restrictions correctly.  Can happen for free calls
		 * on a completely full ag.
		 */
2364
		if (targs.agbno == NULLAGBLOCK) {
2365 2366
			if (flags & XFS_ALLOC_FLAG_FREEING)
				break;
2367
			goto out_agflbp_relse;
2368
		}
L
Linus Torvalds 已提交
2369 2370 2371 2372
		/*
		 * Put each allocated block on the list.
		 */
		for (bno = targs.agbno; bno < targs.agbno + targs.len; bno++) {
D
David Chinner 已提交
2373 2374 2375
			error = xfs_alloc_put_freelist(tp, agbp,
							agflbp, bno, 0);
			if (error)
2376
				goto out_agflbp_relse;
L
Linus Torvalds 已提交
2377 2378
		}
	}
2379
	xfs_trans_brelse(tp, agflbp);
L
Linus Torvalds 已提交
2380 2381
	args->agbp = agbp;
	return 0;
2382 2383 2384 2385 2386 2387 2388 2389 2390

out_agflbp_relse:
	xfs_trans_brelse(tp, agflbp);
out_agbp_relse:
	if (agbp)
		xfs_trans_brelse(tp, agbp);
out_no_agbp:
	args->agbp = NULL;
	return error;
L
Linus Torvalds 已提交
2391 2392 2393 2394 2395 2396 2397 2398 2399 2400
}

/*
 * Get a block from the freelist.
 * Returns with the buffer for the block gotten.
 */
int				/* error */
xfs_alloc_get_freelist(
	xfs_trans_t	*tp,	/* transaction pointer */
	xfs_buf_t	*agbp,	/* buffer containing the agf structure */
D
David Chinner 已提交
2401 2402
	xfs_agblock_t	*bnop,	/* block address retrieved from freelist */
	int		btreeblk) /* destination is a AGF btree */
L
Linus Torvalds 已提交
2403 2404 2405 2406
{
	xfs_agf_t	*agf;	/* a.g. freespace structure */
	xfs_buf_t	*agflbp;/* buffer for a.g. freelist structure */
	xfs_agblock_t	bno;	/* block number returned */
2407
	__be32		*agfl_bno;
L
Linus Torvalds 已提交
2408
	int		error;
D
David Chinner 已提交
2409
	int		logflags;
2410
	xfs_mount_t	*mp = tp->t_mountp;
L
Linus Torvalds 已提交
2411 2412 2413 2414 2415
	xfs_perag_t	*pag;	/* per allocation group data */

	/*
	 * Freelist is empty, give up.
	 */
2416
	agf = XFS_BUF_TO_AGF(agbp);
L
Linus Torvalds 已提交
2417 2418 2419 2420 2421 2422 2423
	if (!agf->agf_flcount) {
		*bnop = NULLAGBLOCK;
		return 0;
	}
	/*
	 * Read the array of free blocks.
	 */
2424 2425 2426
	error = xfs_alloc_read_agfl(mp, tp, be32_to_cpu(agf->agf_seqno),
				    &agflbp);
	if (error)
L
Linus Torvalds 已提交
2427
		return error;
2428 2429


L
Linus Torvalds 已提交
2430 2431 2432
	/*
	 * Get the block number and update the data structures.
	 */
2433 2434
	agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, agflbp);
	bno = be32_to_cpu(agfl_bno[be32_to_cpu(agf->agf_flfirst)]);
2435
	be32_add_cpu(&agf->agf_flfirst, 1);
L
Linus Torvalds 已提交
2436
	xfs_trans_brelse(tp, agflbp);
2437
	if (be32_to_cpu(agf->agf_flfirst) == xfs_agfl_size(mp))
L
Linus Torvalds 已提交
2438
		agf->agf_flfirst = 0;
2439 2440

	pag = xfs_perag_get(mp, be32_to_cpu(agf->agf_seqno));
2441
	ASSERT(!pag->pagf_agflreset);
2442
	be32_add_cpu(&agf->agf_flcount, -1);
L
Linus Torvalds 已提交
2443 2444
	xfs_trans_agflist_delta(tp, -1);
	pag->pagf_flcount--;
2445
	xfs_perag_put(pag);
D
David Chinner 已提交
2446 2447 2448

	logflags = XFS_AGF_FLFIRST | XFS_AGF_FLCOUNT;
	if (btreeblk) {
2449
		be32_add_cpu(&agf->agf_btreeblks, 1);
D
David Chinner 已提交
2450 2451 2452 2453 2454
		pag->pagf_btreeblks++;
		logflags |= XFS_AGF_BTREEBLKS;
	}

	xfs_alloc_log_agf(tp, agbp, logflags);
L
Linus Torvalds 已提交
2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482
	*bnop = bno;

	return 0;
}

/*
 * Log the given fields from the agf structure.
 */
void
xfs_alloc_log_agf(
	xfs_trans_t	*tp,	/* transaction pointer */
	xfs_buf_t	*bp,	/* buffer for a.g. freelist header */
	int		fields)	/* mask of fields to be logged (XFS_AGF_...) */
{
	int	first;		/* first byte offset */
	int	last;		/* last byte offset */
	static const short	offsets[] = {
		offsetof(xfs_agf_t, agf_magicnum),
		offsetof(xfs_agf_t, agf_versionnum),
		offsetof(xfs_agf_t, agf_seqno),
		offsetof(xfs_agf_t, agf_length),
		offsetof(xfs_agf_t, agf_roots[0]),
		offsetof(xfs_agf_t, agf_levels[0]),
		offsetof(xfs_agf_t, agf_flfirst),
		offsetof(xfs_agf_t, agf_fllast),
		offsetof(xfs_agf_t, agf_flcount),
		offsetof(xfs_agf_t, agf_freeblks),
		offsetof(xfs_agf_t, agf_longest),
D
David Chinner 已提交
2483
		offsetof(xfs_agf_t, agf_btreeblks),
D
Dave Chinner 已提交
2484
		offsetof(xfs_agf_t, agf_uuid),
2485
		offsetof(xfs_agf_t, agf_rmap_blocks),
2486 2487 2488
		offsetof(xfs_agf_t, agf_refcount_blocks),
		offsetof(xfs_agf_t, agf_refcount_root),
		offsetof(xfs_agf_t, agf_refcount_level),
2489 2490
		/* needed so that we don't log the whole rest of the structure: */
		offsetof(xfs_agf_t, agf_spare64),
L
Linus Torvalds 已提交
2491 2492 2493
		sizeof(xfs_agf_t)
	};

C
Christoph Hellwig 已提交
2494 2495
	trace_xfs_agf(tp->t_mountp, XFS_BUF_TO_AGF(bp), fields, _RET_IP_);

2496
	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_AGF_BUF);
D
Dave Chinner 已提交
2497

L
Linus Torvalds 已提交
2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529
	xfs_btree_offsets(fields, offsets, XFS_AGF_NUM_BITS, &first, &last);
	xfs_trans_log_buf(tp, bp, (uint)first, (uint)last);
}

/*
 * Interface for inode allocation to force the pag data to be initialized.
 */
int					/* error */
xfs_alloc_pagf_init(
	xfs_mount_t		*mp,	/* file system mount structure */
	xfs_trans_t		*tp,	/* transaction pointer */
	xfs_agnumber_t		agno,	/* allocation group number */
	int			flags)	/* XFS_ALLOC_FLAGS_... */
{
	xfs_buf_t		*bp;
	int			error;

	if ((error = xfs_alloc_read_agf(mp, tp, agno, flags, &bp)))
		return error;
	if (bp)
		xfs_trans_brelse(tp, bp);
	return 0;
}

/*
 * Put the block on the freelist for the allocation group.
 */
int					/* error */
xfs_alloc_put_freelist(
	xfs_trans_t		*tp,	/* transaction pointer */
	xfs_buf_t		*agbp,	/* buffer for a.g. freelist header */
	xfs_buf_t		*agflbp,/* buffer for a.g. free block array */
D
David Chinner 已提交
2530 2531
	xfs_agblock_t		bno,	/* block being freed */
	int			btreeblk) /* block came from a AGF btree */
L
Linus Torvalds 已提交
2532 2533
{
	xfs_agf_t		*agf;	/* a.g. freespace structure */
2534
	__be32			*blockp;/* pointer to array entry */
L
Linus Torvalds 已提交
2535
	int			error;
D
David Chinner 已提交
2536
	int			logflags;
L
Linus Torvalds 已提交
2537 2538
	xfs_mount_t		*mp;	/* mount structure */
	xfs_perag_t		*pag;	/* per allocation group data */
2539 2540
	__be32			*agfl_bno;
	int			startoff;
L
Linus Torvalds 已提交
2541 2542 2543 2544 2545

	agf = XFS_BUF_TO_AGF(agbp);
	mp = tp->t_mountp;

	if (!agflbp && (error = xfs_alloc_read_agfl(mp, tp,
2546
			be32_to_cpu(agf->agf_seqno), &agflbp)))
L
Linus Torvalds 已提交
2547
		return error;
2548
	be32_add_cpu(&agf->agf_fllast, 1);
2549
	if (be32_to_cpu(agf->agf_fllast) == xfs_agfl_size(mp))
L
Linus Torvalds 已提交
2550
		agf->agf_fllast = 0;
2551 2552

	pag = xfs_perag_get(mp, be32_to_cpu(agf->agf_seqno));
2553
	ASSERT(!pag->pagf_agflreset);
2554
	be32_add_cpu(&agf->agf_flcount, 1);
L
Linus Torvalds 已提交
2555 2556
	xfs_trans_agflist_delta(tp, 1);
	pag->pagf_flcount++;
D
David Chinner 已提交
2557 2558 2559

	logflags = XFS_AGF_FLLAST | XFS_AGF_FLCOUNT;
	if (btreeblk) {
2560
		be32_add_cpu(&agf->agf_btreeblks, -1);
D
David Chinner 已提交
2561 2562 2563
		pag->pagf_btreeblks--;
		logflags |= XFS_AGF_BTREEBLKS;
	}
2564
	xfs_perag_put(pag);
D
David Chinner 已提交
2565 2566 2567

	xfs_alloc_log_agf(tp, agbp, logflags);

2568
	ASSERT(be32_to_cpu(agf->agf_flcount) <= xfs_agfl_size(mp));
2569 2570 2571

	agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, agflbp);
	blockp = &agfl_bno[be32_to_cpu(agf->agf_fllast)];
2572
	*blockp = cpu_to_be32(bno);
2573 2574
	startoff = (char *)blockp - (char *)agflbp->b_addr;

D
David Chinner 已提交
2575
	xfs_alloc_log_agf(tp, agbp, logflags);
2576

2577
	xfs_trans_buf_set_type(tp, agflbp, XFS_BLFT_AGFL_BUF);
2578 2579
	xfs_trans_log_buf(tp, agflbp, startoff,
			  startoff + sizeof(xfs_agblock_t) - 1);
L
Linus Torvalds 已提交
2580 2581 2582
	return 0;
}

2583
static xfs_failaddr_t
2584
xfs_agf_verify(
2585 2586 2587 2588
	struct xfs_buf		*bp)
{
	struct xfs_mount	*mp = bp->b_target->bt_mount;
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(bp);
2589

2590 2591
	if (xfs_sb_version_hascrc(&mp->m_sb)) {
		if (!uuid_equal(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid))
2592
			return __this_address;
2593 2594
		if (!xfs_log_check_lsn(mp,
				be64_to_cpu(XFS_BUF_TO_AGF(bp)->agf_lsn)))
2595
			return __this_address;
2596
	}
2597

D
Dave Chinner 已提交
2598 2599 2600
	if (!(agf->agf_magicnum == cpu_to_be32(XFS_AGF_MAGIC) &&
	      XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
	      be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
2601 2602 2603
	      be32_to_cpu(agf->agf_flfirst) < xfs_agfl_size(mp) &&
	      be32_to_cpu(agf->agf_fllast) < xfs_agfl_size(mp) &&
	      be32_to_cpu(agf->agf_flcount) <= xfs_agfl_size(mp)))
2604
		return __this_address;
2605

2606 2607 2608
	if (be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]) < 1 ||
	    be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]) < 1 ||
	    be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]) > XFS_BTREE_MAXLEVELS ||
2609
	    be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]) > XFS_BTREE_MAXLEVELS)
2610
		return __this_address;
2611

2612
	if (xfs_sb_version_hasrmapbt(&mp->m_sb) &&
2613 2614
	    (be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) < 1 ||
	     be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) > XFS_BTREE_MAXLEVELS))
2615
		return __this_address;
2616

2617 2618 2619 2620 2621 2622
	/*
	 * during growfs operations, the perag is not fully initialised,
	 * so we can't use it for any useful checking. growfs ensures we can't
	 * use it by using uncached buffers that don't have the perag attached
	 * so we can detect and avoid this problem.
	 */
D
Dave Chinner 已提交
2623
	if (bp->b_pag && be32_to_cpu(agf->agf_seqno) != bp->b_pag->pag_agno)
2624
		return __this_address;
2625

D
Dave Chinner 已提交
2626 2627
	if (xfs_sb_version_haslazysbcount(&mp->m_sb) &&
	    be32_to_cpu(agf->agf_btreeblks) > be32_to_cpu(agf->agf_length))
2628
		return __this_address;
D
Dave Chinner 已提交
2629

2630
	if (xfs_sb_version_hasreflink(&mp->m_sb) &&
2631 2632
	    (be32_to_cpu(agf->agf_refcount_level) < 1 ||
	     be32_to_cpu(agf->agf_refcount_level) > XFS_BTREE_MAXLEVELS))
2633
		return __this_address;
2634

2635
	return NULL;
2636

2637 2638
}

2639 2640
static void
xfs_agf_read_verify(
2641 2642
	struct xfs_buf	*bp)
{
D
Dave Chinner 已提交
2643
	struct xfs_mount *mp = bp->b_target->bt_mount;
2644
	xfs_failaddr_t	fa;
D
Dave Chinner 已提交
2645

2646 2647
	if (xfs_sb_version_hascrc(&mp->m_sb) &&
	    !xfs_buf_verify_cksum(bp, XFS_AGF_CRC_OFF))
2648 2649
		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
	else {
2650
		fa = xfs_agf_verify(bp);
2651 2652 2653
		if (XFS_TEST_ERROR(fa, mp, XFS_ERRTAG_ALLOC_READ_AGF))
			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
	}
2654
}
2655

2656
static void
2657
xfs_agf_write_verify(
2658 2659
	struct xfs_buf	*bp)
{
C
Carlos Maiolino 已提交
2660 2661
	struct xfs_mount	*mp = bp->b_target->bt_mount;
	struct xfs_buf_log_item	*bip = bp->b_log_item;
2662
	xfs_failaddr_t		fa;
D
Dave Chinner 已提交
2663

2664
	fa = xfs_agf_verify(bp);
2665 2666
	if (fa) {
		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
D
Dave Chinner 已提交
2667 2668 2669 2670 2671 2672 2673 2674 2675
		return;
	}

	if (!xfs_sb_version_hascrc(&mp->m_sb))
		return;

	if (bip)
		XFS_BUF_TO_AGF(bp)->agf_lsn = cpu_to_be64(bip->bli_item.li_lsn);

2676
	xfs_buf_update_cksum(bp, XFS_AGF_CRC_OFF);
2677 2678
}

2679
const struct xfs_buf_ops xfs_agf_buf_ops = {
2680
	.name = "xfs_agf",
2681 2682
	.verify_read = xfs_agf_read_verify,
	.verify_write = xfs_agf_write_verify,
2683
	.verify_struct = xfs_agf_verify,
2684 2685
};

L
Linus Torvalds 已提交
2686 2687 2688 2689
/*
 * Read in the allocation group header (free/alloc section).
 */
int					/* error */
2690 2691 2692 2693 2694 2695
xfs_read_agf(
	struct xfs_mount	*mp,	/* mount point structure */
	struct xfs_trans	*tp,	/* transaction pointer */
	xfs_agnumber_t		agno,	/* allocation group number */
	int			flags,	/* XFS_BUF_ */
	struct xfs_buf		**bpp)	/* buffer for the ag freelist header */
L
Linus Torvalds 已提交
2696 2697 2698
{
	int		error;

2699 2700
	trace_xfs_read_agf(mp, agno);

L
Linus Torvalds 已提交
2701 2702 2703 2704
	ASSERT(agno != NULLAGNUMBER);
	error = xfs_trans_read_buf(
			mp, tp, mp->m_ddev_targp,
			XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
2705
			XFS_FSS_TO_BB(mp, 1), flags, bpp, &xfs_agf_buf_ops);
L
Linus Torvalds 已提交
2706 2707
	if (error)
		return error;
2708
	if (!*bpp)
L
Linus Torvalds 已提交
2709
		return 0;
2710

2711
	ASSERT(!(*bpp)->b_error);
2712
	xfs_buf_set_ref(*bpp, XFS_AGF_REF);
2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730
	return 0;
}

/*
 * Read in the allocation group header (free/alloc section).
 */
int					/* error */
xfs_alloc_read_agf(
	struct xfs_mount	*mp,	/* mount point structure */
	struct xfs_trans	*tp,	/* transaction pointer */
	xfs_agnumber_t		agno,	/* allocation group number */
	int			flags,	/* XFS_ALLOC_FLAG_... */
	struct xfs_buf		**bpp)	/* buffer for the ag freelist header */
{
	struct xfs_agf		*agf;		/* ag freelist header */
	struct xfs_perag	*pag;		/* per allocation group data */
	int			error;

2731
	trace_xfs_alloc_read_agf(mp, agno);
2732

2733
	ASSERT(agno != NULLAGNUMBER);
2734
	error = xfs_read_agf(mp, tp, agno,
2735
			(flags & XFS_ALLOC_FLAG_TRYLOCK) ? XBF_TRYLOCK : 0,
2736 2737 2738 2739 2740
			bpp);
	if (error)
		return error;
	if (!*bpp)
		return 0;
2741
	ASSERT(!(*bpp)->b_error);
2742 2743

	agf = XFS_BUF_TO_AGF(*bpp);
2744
	pag = xfs_perag_get(mp, agno);
L
Linus Torvalds 已提交
2745
	if (!pag->pagf_init) {
2746
		pag->pagf_freeblks = be32_to_cpu(agf->agf_freeblks);
D
David Chinner 已提交
2747
		pag->pagf_btreeblks = be32_to_cpu(agf->agf_btreeblks);
2748 2749
		pag->pagf_flcount = be32_to_cpu(agf->agf_flcount);
		pag->pagf_longest = be32_to_cpu(agf->agf_longest);
L
Linus Torvalds 已提交
2750
		pag->pagf_levels[XFS_BTNUM_BNOi] =
2751
			be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]);
L
Linus Torvalds 已提交
2752
		pag->pagf_levels[XFS_BTNUM_CNTi] =
2753
			be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]);
2754 2755
		pag->pagf_levels[XFS_BTNUM_RMAPi] =
			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
2756
		pag->pagf_refcount_level = be32_to_cpu(agf->agf_refcount_level);
E
Eric Sandeen 已提交
2757
		spin_lock_init(&pag->pagb_lock);
2758
		pag->pagb_count = 0;
2759
		pag->pagb_tree = RB_ROOT;
L
Linus Torvalds 已提交
2760
		pag->pagf_init = 1;
2761
		pag->pagf_agflreset = xfs_agfl_needs_reset(mp, agf);
L
Linus Torvalds 已提交
2762 2763 2764
	}
#ifdef DEBUG
	else if (!XFS_FORCED_SHUTDOWN(mp)) {
2765
		ASSERT(pag->pagf_freeblks == be32_to_cpu(agf->agf_freeblks));
2766
		ASSERT(pag->pagf_btreeblks == be32_to_cpu(agf->agf_btreeblks));
2767 2768
		ASSERT(pag->pagf_flcount == be32_to_cpu(agf->agf_flcount));
		ASSERT(pag->pagf_longest == be32_to_cpu(agf->agf_longest));
L
Linus Torvalds 已提交
2769
		ASSERT(pag->pagf_levels[XFS_BTNUM_BNOi] ==
2770
		       be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]));
L
Linus Torvalds 已提交
2771
		ASSERT(pag->pagf_levels[XFS_BTNUM_CNTi] ==
2772
		       be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]));
L
Linus Torvalds 已提交
2773 2774
	}
#endif
2775
	xfs_perag_put(pag);
L
Linus Torvalds 已提交
2776 2777 2778 2779 2780 2781 2782 2783 2784
	return 0;
}

/*
 * Allocate an extent (variable-size).
 * Depending on the allocation type, we either look in a single allocation
 * group or loop over the allocation groups to find the result.
 */
int				/* error */
2785
xfs_alloc_vextent(
2786
	struct xfs_alloc_arg	*args)	/* allocation argument structure */
L
Linus Torvalds 已提交
2787
{
2788 2789 2790 2791 2792 2793 2794 2795
	xfs_agblock_t		agsize;	/* allocation group size */
	int			error;
	int			flags;	/* XFS_ALLOC_FLAG_... locking flags */
	struct xfs_mount	*mp;	/* mount structure pointer */
	xfs_agnumber_t		sagno;	/* starting allocation group number */
	xfs_alloctype_t		type;	/* input allocation type */
	int			bump_rotor = 0;
	xfs_agnumber_t		rotorstep = xfs_rotorstep; /* inode32 agf stepper */
L
Linus Torvalds 已提交
2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819

	mp = args->mp;
	type = args->otype = args->type;
	args->agbno = NULLAGBLOCK;
	/*
	 * Just fix this up, for the case where the last a.g. is shorter
	 * (or there's only one a.g.) and the caller couldn't easily figure
	 * that out (xfs_bmap_alloc).
	 */
	agsize = mp->m_sb.sb_agblocks;
	if (args->maxlen > agsize)
		args->maxlen = agsize;
	if (args->alignment == 0)
		args->alignment = 1;
	ASSERT(XFS_FSB_TO_AGNO(mp, args->fsbno) < mp->m_sb.sb_agcount);
	ASSERT(XFS_FSB_TO_AGBNO(mp, args->fsbno) < agsize);
	ASSERT(args->minlen <= args->maxlen);
	ASSERT(args->minlen <= agsize);
	ASSERT(args->mod < args->prod);
	if (XFS_FSB_TO_AGNO(mp, args->fsbno) >= mp->m_sb.sb_agcount ||
	    XFS_FSB_TO_AGBNO(mp, args->fsbno) >= agsize ||
	    args->minlen > args->maxlen || args->minlen > agsize ||
	    args->mod >= args->prod) {
		args->fsbno = NULLFSBLOCK;
C
Christoph Hellwig 已提交
2820
		trace_xfs_alloc_vextent_badargs(args);
L
Linus Torvalds 已提交
2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831
		return 0;
	}

	switch (type) {
	case XFS_ALLOCTYPE_THIS_AG:
	case XFS_ALLOCTYPE_NEAR_BNO:
	case XFS_ALLOCTYPE_THIS_BNO:
		/*
		 * These three force us into a single a.g.
		 */
		args->agno = XFS_FSB_TO_AGNO(mp, args->fsbno);
2832
		args->pag = xfs_perag_get(mp, args->agno);
L
Linus Torvalds 已提交
2833 2834
		error = xfs_alloc_fix_freelist(args, 0);
		if (error) {
C
Christoph Hellwig 已提交
2835
			trace_xfs_alloc_vextent_nofix(args);
L
Linus Torvalds 已提交
2836 2837 2838
			goto error0;
		}
		if (!args->agbp) {
C
Christoph Hellwig 已提交
2839
			trace_xfs_alloc_vextent_noagbp(args);
L
Linus Torvalds 已提交
2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850
			break;
		}
		args->agbno = XFS_FSB_TO_AGBNO(mp, args->fsbno);
		if ((error = xfs_alloc_ag_vextent(args)))
			goto error0;
		break;
	case XFS_ALLOCTYPE_START_BNO:
		/*
		 * Try near allocation first, then anywhere-in-ag after
		 * the first a.g. fails.
		 */
2851
		if ((args->datatype & XFS_ALLOC_INITIAL_USER_DATA) &&
L
Linus Torvalds 已提交
2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
		    (mp->m_flags & XFS_MOUNT_32BITINODES)) {
			args->fsbno = XFS_AGB_TO_FSB(mp,
					((mp->m_agfrotor / rotorstep) %
					mp->m_sb.sb_agcount), 0);
			bump_rotor = 1;
		}
		args->agbno = XFS_FSB_TO_AGBNO(mp, args->fsbno);
		args->type = XFS_ALLOCTYPE_NEAR_BNO;
		/* FALLTHROUGH */
	case XFS_ALLOCTYPE_FIRST_AG:
		/*
		 * Rotate through the allocation groups looking for a winner.
		 */
2865
		if (type == XFS_ALLOCTYPE_FIRST_AG) {
L
Linus Torvalds 已提交
2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884
			/*
			 * Start with allocation group given by bno.
			 */
			args->agno = XFS_FSB_TO_AGNO(mp, args->fsbno);
			args->type = XFS_ALLOCTYPE_THIS_AG;
			sagno = 0;
			flags = 0;
		} else {
			/*
			 * Start with the given allocation group.
			 */
			args->agno = sagno = XFS_FSB_TO_AGNO(mp, args->fsbno);
			flags = XFS_ALLOC_FLAG_TRYLOCK;
		}
		/*
		 * Loop over allocation groups twice; first time with
		 * trylock set, second time without.
		 */
		for (;;) {
2885
			args->pag = xfs_perag_get(mp, args->agno);
L
Linus Torvalds 已提交
2886 2887
			error = xfs_alloc_fix_freelist(args, flags);
			if (error) {
C
Christoph Hellwig 已提交
2888
				trace_xfs_alloc_vextent_nofix(args);
L
Linus Torvalds 已提交
2889 2890 2891 2892 2893 2894 2895 2896 2897 2898
				goto error0;
			}
			/*
			 * If we get a buffer back then the allocation will fly.
			 */
			if (args->agbp) {
				if ((error = xfs_alloc_ag_vextent(args)))
					goto error0;
				break;
			}
C
Christoph Hellwig 已提交
2899 2900 2901

			trace_xfs_alloc_vextent_loopfailed(args);

L
Linus Torvalds 已提交
2902 2903 2904 2905 2906 2907
			/*
			 * Didn't work, figure out the next iteration.
			 */
			if (args->agno == sagno &&
			    type == XFS_ALLOCTYPE_START_BNO)
				args->type = XFS_ALLOCTYPE_THIS_AG;
2908 2909 2910 2911 2912 2913 2914 2915
			/*
			* For the first allocation, we can try any AG to get
			* space.  However, if we already have allocated a
			* block, we don't want to try AGs whose number is below
			* sagno. Otherwise, we may end up with out-of-order
			* locking of AGF, which might cause deadlock.
			*/
			if (++(args->agno) == mp->m_sb.sb_agcount) {
2916
				if (args->tp->t_firstblock != NULLFSBLOCK)
2917 2918 2919 2920
					args->agno = sagno;
				else
					args->agno = 0;
			}
L
Linus Torvalds 已提交
2921 2922 2923 2924 2925
			/*
			 * Reached the starting a.g., must either be done
			 * or switch to non-trylock mode.
			 */
			if (args->agno == sagno) {
2926
				if (flags == 0) {
L
Linus Torvalds 已提交
2927
					args->agbno = NULLAGBLOCK;
C
Christoph Hellwig 已提交
2928
					trace_xfs_alloc_vextent_allfailed(args);
L
Linus Torvalds 已提交
2929 2930
					break;
				}
2931 2932 2933 2934 2935 2936

				flags = 0;
				if (type == XFS_ALLOCTYPE_START_BNO) {
					args->agbno = XFS_FSB_TO_AGBNO(mp,
						args->fsbno);
					args->type = XFS_ALLOCTYPE_NEAR_BNO;
L
Linus Torvalds 已提交
2937 2938
				}
			}
2939
			xfs_perag_put(args->pag);
L
Linus Torvalds 已提交
2940
		}
2941
		if (bump_rotor) {
L
Linus Torvalds 已提交
2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964
			if (args->agno == sagno)
				mp->m_agfrotor = (mp->m_agfrotor + 1) %
					(mp->m_sb.sb_agcount * rotorstep);
			else
				mp->m_agfrotor = (args->agno * rotorstep + 1) %
					(mp->m_sb.sb_agcount * rotorstep);
		}
		break;
	default:
		ASSERT(0);
		/* NOTREACHED */
	}
	if (args->agbno == NULLAGBLOCK)
		args->fsbno = NULLFSBLOCK;
	else {
		args->fsbno = XFS_AGB_TO_FSB(mp, args->agno, args->agbno);
#ifdef DEBUG
		ASSERT(args->len >= args->minlen);
		ASSERT(args->len <= args->maxlen);
		ASSERT(args->agbno % args->alignment == 0);
		XFS_AG_CHECK_DADDR(mp, XFS_FSB_TO_DADDR(mp, args->fsbno),
			args->len);
#endif
2965 2966

		/* Zero the extent if we were asked to do so */
2967
		if (args->datatype & XFS_ALLOC_USERDATA_ZERO) {
2968 2969 2970 2971 2972
			error = xfs_zero_extent(args->ip, args->fsbno, args->len);
			if (error)
				goto error0;
		}

L
Linus Torvalds 已提交
2973
	}
2974
	xfs_perag_put(args->pag);
L
Linus Torvalds 已提交
2975 2976
	return 0;
error0:
2977
	xfs_perag_put(args->pag);
L
Linus Torvalds 已提交
2978 2979 2980
	return error;
}

2981 2982 2983 2984 2985 2986
/* Ensure that the freelist is at full capacity. */
int
xfs_free_extent_fix_freelist(
	struct xfs_trans	*tp,
	xfs_agnumber_t		agno,
	struct xfs_buf		**agbp)
L
Linus Torvalds 已提交
2987
{
2988 2989
	struct xfs_alloc_arg	args;
	int			error;
L
Linus Torvalds 已提交
2990

2991
	memset(&args, 0, sizeof(struct xfs_alloc_arg));
L
Linus Torvalds 已提交
2992 2993
	args.tp = tp;
	args.mp = tp->t_mountp;
2994
	args.agno = agno;
2995 2996 2997 2998 2999 3000

	/*
	 * validate that the block number is legal - the enables us to detect
	 * and handle a silent filesystem corruption rather than crashing.
	 */
	if (args.agno >= args.mp->m_sb.sb_agcount)
D
Dave Chinner 已提交
3001
		return -EFSCORRUPTED;
3002

3003
	args.pag = xfs_perag_get(args.mp, args.agno);
3004 3005 3006 3007
	ASSERT(args.pag);

	error = xfs_alloc_fix_freelist(&args, XFS_ALLOC_FLAG_FREEING);
	if (error)
3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021
		goto out;

	*agbp = args.agbp;
out:
	xfs_perag_put(args.pag);
	return error;
}

/*
 * Free an extent.
 * Just break up the extent address and hand off to xfs_free_ag_extent
 * after fixing up the freelist.
 */
int				/* error */
B
Brian Foster 已提交
3022
__xfs_free_extent(
3023 3024
	struct xfs_trans	*tp,	/* transaction pointer */
	xfs_fsblock_t		bno,	/* starting block number of extent */
3025
	xfs_extlen_t		len,	/* length of extent */
3026
	struct xfs_owner_info	*oinfo,	/* extent owner */
B
Brian Foster 已提交
3027 3028
	enum xfs_ag_resv_type	type,	/* block reservation type */
	bool			skip_discard)
3029 3030 3031 3032 3033 3034
{
	struct xfs_mount	*mp = tp->t_mountp;
	struct xfs_buf		*agbp;
	xfs_agnumber_t		agno = XFS_FSB_TO_AGNO(mp, bno);
	xfs_agblock_t		agbno = XFS_FSB_TO_AGBNO(mp, bno);
	int			error;
B
Brian Foster 已提交
3035
	unsigned int		busy_flags = 0;
3036 3037

	ASSERT(len != 0);
3038
	ASSERT(type != XFS_AG_RESV_AGFL);
3039

3040
	if (XFS_TEST_ERROR(false, mp,
3041
			XFS_ERRTAG_FREE_EXTENT))
3042 3043
		return -EIO;

3044 3045 3046 3047 3048
	error = xfs_free_extent_fix_freelist(tp, agno, &agbp);
	if (error)
		return error;

	XFS_WANT_CORRUPTED_GOTO(mp, agbno < mp->m_sb.sb_agblocks, err);
3049 3050

	/* validate the extent size is legal now we have the agf locked */
3051 3052 3053
	XFS_WANT_CORRUPTED_GOTO(mp,
		agbno + len <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_length),
				err);
3054

3055
	error = xfs_free_ag_extent(tp, agbp, agno, agbno, len, oinfo, type);
3056 3057 3058
	if (error)
		goto err;

B
Brian Foster 已提交
3059 3060 3061
	if (skip_discard)
		busy_flags |= XFS_EXTENT_BUSY_SKIP_DISCARD;
	xfs_extent_busy_insert(tp, agno, agbno, len, busy_flags);
3062 3063 3064 3065
	return 0;

err:
	xfs_trans_brelse(tp, agbp);
L
Linus Torvalds 已提交
3066 3067
	return error;
}
3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109

struct xfs_alloc_query_range_info {
	xfs_alloc_query_range_fn	fn;
	void				*priv;
};

/* Format btree record and pass to our callback. */
STATIC int
xfs_alloc_query_range_helper(
	struct xfs_btree_cur		*cur,
	union xfs_btree_rec		*rec,
	void				*priv)
{
	struct xfs_alloc_query_range_info	*query = priv;
	struct xfs_alloc_rec_incore		irec;

	irec.ar_startblock = be32_to_cpu(rec->alloc.ar_startblock);
	irec.ar_blockcount = be32_to_cpu(rec->alloc.ar_blockcount);
	return query->fn(cur, &irec, query->priv);
}

/* Find all free space within a given range of blocks. */
int
xfs_alloc_query_range(
	struct xfs_btree_cur			*cur,
	struct xfs_alloc_rec_incore		*low_rec,
	struct xfs_alloc_rec_incore		*high_rec,
	xfs_alloc_query_range_fn		fn,
	void					*priv)
{
	union xfs_btree_irec			low_brec;
	union xfs_btree_irec			high_brec;
	struct xfs_alloc_query_range_info	query;

	ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
	low_brec.a = *low_rec;
	high_brec.a = *high_rec;
	query.priv = priv;
	query.fn = fn;
	return xfs_btree_query_range(cur, &low_brec, &high_brec,
			xfs_alloc_query_range_helper, &query);
}
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124

/* Find all free space records. */
int
xfs_alloc_query_all(
	struct xfs_btree_cur			*cur,
	xfs_alloc_query_range_fn		fn,
	void					*priv)
{
	struct xfs_alloc_query_range_info	query;

	ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
	query.priv = priv;
	query.fn = fn;
	return xfs_btree_query_all(cur, xfs_alloc_query_range_helper, &query);
}
3125

3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143
/* Is there a record covering a given extent? */
int
xfs_alloc_has_record(
	struct xfs_btree_cur	*cur,
	xfs_agblock_t		bno,
	xfs_extlen_t		len,
	bool			*exists)
{
	union xfs_btree_irec	low;
	union xfs_btree_irec	high;

	memset(&low, 0, sizeof(low));
	low.a.ar_startblock = bno;
	memset(&high, 0xFF, sizeof(high));
	high.a.ar_startblock = bno + len - 1;

	return xfs_btree_has_record(cur, &low, &high, exists);
}
3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180

/*
 * Walk all the blocks in the AGFL.  The @walk_fn can return any negative
 * error code or XFS_BTREE_QUERY_RANGE_ABORT.
 */
int
xfs_agfl_walk(
	struct xfs_mount	*mp,
	struct xfs_agf		*agf,
	struct xfs_buf		*agflbp,
	xfs_agfl_walk_fn	walk_fn,
	void			*priv)
{
	__be32			*agfl_bno;
	unsigned int		i;
	int			error;

	agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, agflbp);
	i = be32_to_cpu(agf->agf_flfirst);

	/* Nothing to walk in an empty AGFL. */
	if (agf->agf_flcount == cpu_to_be32(0))
		return 0;

	/* Otherwise, walk from first to last, wrapping as needed. */
	for (;;) {
		error = walk_fn(mp, be32_to_cpu(agfl_bno[i]), priv);
		if (error)
			return error;
		if (i == be32_to_cpu(agf->agf_fllast))
			break;
		if (++i == xfs_agfl_size(mp))
			i = 0;
	}

	return 0;
}