xfs_alloc.c 86.2 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2 3
 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
L
Linus Torvalds 已提交
4
 *
5 6
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
L
Linus Torvalds 已提交
7 8
 * published by the Free Software Foundation.
 *
9 10 11 12
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
L
Linus Torvalds 已提交
13
 *
14 15 16
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
L
Linus Torvalds 已提交
17 18
 */
#include "xfs.h"
19
#include "xfs_fs.h"
20
#include "xfs_format.h"
21
#include "xfs_log_format.h"
22
#include "xfs_shared.h"
23
#include "xfs_trans_resv.h"
24
#include "xfs_bit.h"
L
Linus Torvalds 已提交
25 26
#include "xfs_sb.h"
#include "xfs_mount.h"
27
#include "xfs_defer.h"
28
#include "xfs_inode.h"
L
Linus Torvalds 已提交
29
#include "xfs_btree.h"
30
#include "xfs_rmap.h"
31
#include "xfs_alloc_btree.h"
L
Linus Torvalds 已提交
32
#include "xfs_alloc.h"
33
#include "xfs_extent_busy.h"
34
#include "xfs_errortag.h"
L
Linus Torvalds 已提交
35
#include "xfs_error.h"
D
Dave Chinner 已提交
36
#include "xfs_cksum.h"
C
Christoph Hellwig 已提交
37
#include "xfs_trace.h"
38
#include "xfs_trans.h"
D
Dave Chinner 已提交
39
#include "xfs_buf_item.h"
40
#include "xfs_log.h"
41
#include "xfs_ag_resv.h"
L
Linus Torvalds 已提交
42

43
struct workqueue_struct *xfs_alloc_wq;
L
Linus Torvalds 已提交
44 45 46 47 48 49 50 51 52 53

#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 *,
54
		xfs_btree_cur_t *, xfs_agblock_t *, xfs_extlen_t *, int *);
L
Linus Torvalds 已提交
55

56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72
/*
 * 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);
}

73 74 75 76 77 78 79 80 81 82 83
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;
}

84 85 86 87
xfs_extlen_t
xfs_prealloc_blocks(
	struct xfs_mount	*mp)
{
88 89
	if (xfs_sb_version_hasreflink(&mp->m_sb))
		return xfs_refc_block(mp) + 1;
90 91 92 93 94 95 96
	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;
}

97 98 99 100 101 102 103 104 105 106 107 108
/*
 * 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.
 *
109 110
 * 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.
111 112 113 114 115
 */
unsigned int
xfs_alloc_set_aside(
	struct xfs_mount	*mp)
{
116
	return mp->m_sb.sb_agcount * (XFS_ALLOC_AGFL_RESERVE + 4);
117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
}

/*
 * 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 */
146 147
	if (xfs_sb_version_hasreflink(&mp->m_sb))
		blocks++;		/* refcount root block */
148 149 150 151

	return mp->m_sb.sb_agblocks - blocks;
}

152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
/*
 * 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.
 */
171
int				/* error */
172 173 174 175 176 177 178 179 180 181 182 183 184 185 186
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.
 */
187
int					/* error */
188 189 190 191 192 193 194 195 196 197 198
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);
}

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
/*
 * 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);
}
216

217 218 219
/*
 * Get the data from the pointed-to record.
 */
C
Christoph Hellwig 已提交
220
int					/* error */
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
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 */
{
	union xfs_btree_rec	*rec;
	int			error;

	error = xfs_btree_get_rec(cur, &rec, stat);
	if (!error && *stat == 1) {
		*bno = be32_to_cpu(rec->alloc.ar_startblock);
		*len = be32_to_cpu(rec->alloc.ar_blockcount);
	}
	return error;
}

L
Linus Torvalds 已提交
238 239 240 241
/*
 * Compute aligned version of the found extent.
 * Takes alignment and min length into account.
 */
242
STATIC bool
L
Linus Torvalds 已提交
243
xfs_alloc_compute_aligned(
244
	xfs_alloc_arg_t	*args,		/* allocation argument structure */
L
Linus Torvalds 已提交
245 246 247
	xfs_agblock_t	foundbno,	/* starting block in found extent */
	xfs_extlen_t	foundlen,	/* length in found extent */
	xfs_agblock_t	*resbno,	/* result block number */
248 249
	xfs_extlen_t	*reslen,	/* result length */
	unsigned	*busy_gen)
L
Linus Torvalds 已提交
250
{
251 252
	xfs_agblock_t	bno = foundbno;
	xfs_extlen_t	len = foundlen;
253
	xfs_extlen_t	diff;
254
	bool		busy;
L
Linus Torvalds 已提交
255

256
	/* Trim busy sections out of found extent */
257
	busy = xfs_extent_busy_trim(args, &bno, &len, busy_gen);
258

259 260 261 262 263 264 265 266 267 268 269 270
	/*
	 * 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;
		}
	}

271 272
	if (args->alignment > 1 && len >= args->minlen) {
		xfs_agblock_t	aligned_bno = roundup(bno, args->alignment);
273 274

		diff = aligned_bno - bno;
275 276 277

		*resbno = aligned_bno;
		*reslen = diff >= len ? 0 : len - diff;
L
Linus Torvalds 已提交
278
	} else {
279 280
		*resbno = bno;
		*reslen = len;
L
Linus Torvalds 已提交
281
	}
282 283

	return busy;
L
Linus Torvalds 已提交
284 285 286 287 288 289 290 291 292 293 294
}

/*
 * 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 */
295
	int		datatype,	/* are we allocating data? */
L
Linus Torvalds 已提交
296 297 298 299 300 301 302 303 304 305
	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 */
306
	bool		userdata = xfs_alloc_is_userdata(datatype);
L
Linus Torvalds 已提交
307 308 309 310

	ASSERT(freelen >= wantlen);
	freeend = freebno + freelen;
	wantend = wantbno + wantlen;
311 312 313 314 315 316 317 318
	/*
	 * 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
319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
		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;
378 379 380 381
	if (k > args->mod)
		rlen = rlen - (k - args->mod);
	else
		rlen = rlen - args->prod + (args->mod - k);
382
	/* casts to (int) catch length underflows */
383 384 385 386
	if ((int)rlen < (int)args->minlen)
		return;
	ASSERT(rlen >= args->minlen && rlen <= args->maxlen);
	ASSERT(rlen % args->prod == args->mod);
387 388
	ASSERT(args->pag->pagf_freeblks + args->pag->pagf_flcount >=
		rlen + args->minleft);
L
Linus Torvalds 已提交
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
	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 */
415 416 417
	struct xfs_mount *mp;

	mp = cnt_cur->bc_mp;
L
Linus Torvalds 已提交
418 419 420 421 422 423 424 425

	/*
	 * 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;
426
		XFS_WANT_CORRUPTED_RETURN(mp,
L
Linus Torvalds 已提交
427 428 429 430 431
			i == 1 && nfbno1 == fbno && nflen1 == flen);
#endif
	} else {
		if ((error = xfs_alloc_lookup_eq(cnt_cur, fbno, flen, &i)))
			return error;
432
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
433 434 435 436 437 438 439 440
	}
	/*
	 * 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;
441
		XFS_WANT_CORRUPTED_RETURN(mp,
L
Linus Torvalds 已提交
442 443 444 445 446
			i == 1 && nfbno1 == fbno && nflen1 == flen);
#endif
	} else {
		if ((error = xfs_alloc_lookup_eq(bno_cur, fbno, flen, &i)))
			return error;
447
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
448
	}
449

L
Linus Torvalds 已提交
450
#ifdef DEBUG
451 452 453 454 455 456
	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
457

458
		XFS_WANT_CORRUPTED_RETURN(mp,
459
			bnoblock->bb_numrecs == cntblock->bb_numrecs);
L
Linus Torvalds 已提交
460 461
	}
#endif
462

L
Linus Torvalds 已提交
463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
	/*
	 * 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.
	 */
487
	if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
488
		return error;
489
	XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
490 491 492 493 494 495
	/*
	 * Add new by-size btree entry(s).
	 */
	if (nfbno1 != NULLAGBLOCK) {
		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno1, nflen1, &i)))
			return error;
496
		XFS_WANT_CORRUPTED_RETURN(mp, i == 0);
497
		if ((error = xfs_btree_insert(cnt_cur, &i)))
L
Linus Torvalds 已提交
498
			return error;
499
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
500 501 502 503
	}
	if (nfbno2 != NULLAGBLOCK) {
		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno2, nflen2, &i)))
			return error;
504
		XFS_WANT_CORRUPTED_RETURN(mp, i == 0);
505
		if ((error = xfs_btree_insert(cnt_cur, &i)))
L
Linus Torvalds 已提交
506
			return error;
507
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
508 509 510 511 512 513 514 515
	}
	/*
	 * Fix up the by-block btree entry(s).
	 */
	if (nfbno1 == NULLAGBLOCK) {
		/*
		 * No remaining freespace, just delete the by-block tree entry.
		 */
516
		if ((error = xfs_btree_delete(bno_cur, &i)))
L
Linus Torvalds 已提交
517
			return error;
518
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
519 520 521 522 523 524 525 526 527 528 529 530 531
	} 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;
532
		XFS_WANT_CORRUPTED_RETURN(mp, i == 0);
533
		if ((error = xfs_btree_insert(bno_cur, &i)))
L
Linus Torvalds 已提交
534
			return error;
535
		XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
L
Linus Torvalds 已提交
536 537 538 539
	}
	return 0;
}

540
static xfs_failaddr_t
541
xfs_agfl_verify(
542 543 544 545 546 547
	struct xfs_buf	*bp)
{
	struct xfs_mount *mp = bp->b_target->bt_mount;
	struct xfs_agfl	*agfl = XFS_BUF_TO_AGFL(bp);
	int		i;

548 549 550 551 552 553 554 555 556
	/*
	 * 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;

557
	if (!uuid_equal(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid))
558
		return __this_address;
559
	if (be32_to_cpu(agfl->agfl_magicnum) != XFS_AGFL_MAGIC)
560
		return __this_address;
561 562 563 564 565 566 567
	/*
	 * 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)
568
		return __this_address;
569

570
	for (i = 0; i < xfs_agfl_size(mp); i++) {
571
		if (be32_to_cpu(agfl->agfl_bno[i]) != NULLAGBLOCK &&
572
		    be32_to_cpu(agfl->agfl_bno[i]) >= mp->m_sb.sb_agblocks)
573
			return __this_address;
574
	}
575

576 577 578
	if (!xfs_log_check_lsn(mp, be64_to_cpu(XFS_BUF_TO_AGFL(bp)->agfl_lsn)))
		return __this_address;
	return NULL;
579 580 581 582 583 584 585
}

static void
xfs_agfl_read_verify(
	struct xfs_buf	*bp)
{
	struct xfs_mount *mp = bp->b_target->bt_mount;
586
	xfs_failaddr_t	fa;
587 588 589 590 591 592 593 594 595 596

	/*
	 * 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;

597
	if (!xfs_buf_verify_cksum(bp, XFS_AGFL_CRC_OFF))
598 599 600 601 602 603
		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
	else {
		fa = xfs_agfl_verify(bp);
		if (fa)
			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
	}
604 605
}

606
static void
607 608 609
xfs_agfl_write_verify(
	struct xfs_buf	*bp)
{
C
Carlos Maiolino 已提交
610 611
	struct xfs_mount	*mp = bp->b_target->bt_mount;
	struct xfs_buf_log_item	*bip = bp->b_log_item;
612
	xfs_failaddr_t		fa;
613

614 615 616 617
	/* no verification of non-crc AGFLs */
	if (!xfs_sb_version_hascrc(&mp->m_sb))
		return;

618 619 620
	fa = xfs_agfl_verify(bp);
	if (fa) {
		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
621 622 623 624 625 626
		return;
	}

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

627
	xfs_buf_update_cksum(bp, XFS_AGFL_CRC_OFF);
628 629
}

630
const struct xfs_buf_ops xfs_agfl_buf_ops = {
631
	.name = "xfs_agfl",
632 633
	.verify_read = xfs_agfl_read_verify,
	.verify_write = xfs_agfl_write_verify,
634
	.verify_struct = xfs_agfl_verify,
635 636
};

L
Linus Torvalds 已提交
637 638 639
/*
 * Read in the allocation group free block array.
 */
640
int					/* error */
L
Linus Torvalds 已提交
641 642 643 644 645 646 647 648 649 650 651 652 653
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)),
654
			XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_agfl_buf_ops);
L
Linus Torvalds 已提交
655 656
	if (error)
		return error;
657
	xfs_buf_set_ref(bp, XFS_AGFL_REF);
L
Linus Torvalds 已提交
658 659 660 661
	*bpp = bp;
	return 0;
}

662 663 664 665 666 667 668 669 670 671 672 673 674 675 676
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
677
		return -EFSCORRUPTED;
678 679 680 681 682

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

L
Linus Torvalds 已提交
683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705
/*
 * 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);
706

L
Linus Torvalds 已提交
707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
	/*
	 * 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 */
	}
725 726

	if (error || args->agbno == NULLAGBLOCK)
L
Linus Torvalds 已提交
727
		return error;
728 729 730

	ASSERT(args->len >= args->minlen);
	ASSERT(args->len <= args->maxlen);
731
	ASSERT(!args->wasfromfl || args->resv != XFS_AG_RESV_AGFL);
732 733
	ASSERT(args->agbno % args->alignment == 0);

734
	/* if not file data, insert new block into the reverse map btree */
735
	if (!xfs_rmap_should_skip_owner_update(&args->oinfo)) {
736 737 738 739 740 741
		error = xfs_rmap_alloc(args->tp, args->agbp, args->agno,
				       args->agbno, args->len, &args->oinfo);
		if (error)
			return error;
	}

742 743 744 745 746 747 748
	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
749
		ASSERT(!xfs_extent_busy_search(args->mp, args->agno,
750
					      args->agbno, args->len));
L
Linus Torvalds 已提交
751
	}
752

753
	xfs_ag_resv_alloc_extent(args->pag, args->resv, args);
754

755 756
	XFS_STATS_INC(args->mp, xs_allocx);
	XFS_STATS_ADD(args->mp, xs_allocb, args->len);
757
	return error;
L
Linus Torvalds 已提交
758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
}

/*
 * 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 */
775 776 777
	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
778
	int		i;	/* success/failure of operation */
779
	unsigned	busy_gen;
L
Linus Torvalds 已提交
780 781

	ASSERT(args->alignment == 1);
782

L
Linus Torvalds 已提交
783 784 785
	/*
	 * Allocate/initialize a cursor for the by-number freespace btree.
	 */
786
	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
787 788
					  args->agno, XFS_BTNUM_BNO);

L
Linus Torvalds 已提交
789 790 791 792 793
	/*
	 * 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.
	 */
794 795
	error = xfs_alloc_lookup_le(bno_cur, args->agbno, args->minlen, &i);
	if (error)
L
Linus Torvalds 已提交
796
		goto error0;
797 798 799
	if (!i)
		goto not_found;

L
Linus Torvalds 已提交
800 801 802
	/*
	 * Grab the freespace record.
	 */
803 804
	error = xfs_alloc_get_rec(bno_cur, &fbno, &flen, &i);
	if (error)
L
Linus Torvalds 已提交
805
		goto error0;
806
	XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
807
	ASSERT(fbno <= args->agbno);
808

L
Linus Torvalds 已提交
809
	/*
810
	 * Check for overlapping busy extents.
L
Linus Torvalds 已提交
811
	 */
812 813 814
	tbno = fbno;
	tlen = flen;
	xfs_extent_busy_trim(args, &tbno, &tlen, &busy_gen);
815 816 817 818 819 820 821 822 823 824 825

	/*
	 * 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)
826 827
		goto not_found;

L
Linus Torvalds 已提交
828 829 830
	/*
	 * End of extent will be smaller of the freespace end and the
	 * maximal requested end.
831
	 *
L
Linus Torvalds 已提交
832 833
	 * Fix the length according to mod and prod if given.
	 */
834 835
	args->len = XFS_AGBLOCK_MIN(tend, args->agbno + args->maxlen)
						- args->agbno;
L
Linus Torvalds 已提交
836
	xfs_alloc_fix_len(args);
837
	ASSERT(args->agbno + args->len <= tend);
838

L
Linus Torvalds 已提交
839
	/*
840
	 * We are allocating agbno for args->len
L
Linus Torvalds 已提交
841 842
	 * Allocate/initialize a cursor for the by-size btree.
	 */
843 844
	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_CNT);
L
Linus Torvalds 已提交
845
	ASSERT(args->agbno + args->len <=
846
		be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
847 848 849
	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
850 851 852
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
		goto error0;
	}
853

L
Linus Torvalds 已提交
854 855
	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
C
Christoph Hellwig 已提交
856

L
Linus Torvalds 已提交
857
	args->wasfromfl = 0;
858 859 860 861 862 863 864 865
	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
866 867 868 869
	return 0;

error0:
	xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
C
Christoph Hellwig 已提交
870
	trace_xfs_alloc_exact_error(args);
L
Linus Torvalds 已提交
871 872 873
	return error;
}

874 875 876 877 878 879 880 881 882 883 884
/*
 * 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 */
885 886 887
	xfs_extlen_t		*slen,	/* extent length */
	xfs_agblock_t		*sbnoa,	/* aligned extent found by search */
	xfs_extlen_t		*slena,	/* aligned extent length */
888 889 890 891 892 893
	int			dir)	/* 0 = search right, 1 = search left */
{
	xfs_agblock_t		new;
	xfs_agblock_t		sdiff;
	int			error;
	int			i;
894
	unsigned		busy_gen;
895 896 897 898 899 900 901 902 903 904 905 906

	/* 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;
907
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
908 909
		xfs_alloc_compute_aligned(args, *sbno, *slen,
				sbnoa, slena, &busy_gen);
910 911 912 913 914

		/*
		 * The good extent is closer than this one.
		 */
		if (!dir) {
915 916
			if (*sbnoa > args->max_agbno)
				goto out_use_good;
917
			if (*sbnoa >= args->agbno + gdiff)
918 919
				goto out_use_good;
		} else {
920 921
			if (*sbnoa < args->min_agbno)
				goto out_use_good;
922
			if (*sbnoa <= args->agbno - gdiff)
923 924 925 926 927 928 929 930 931 932 933
				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,
934
						       args->alignment,
935
						       args->datatype, *sbnoa,
936
						       *slena, &new);
937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968

			/*
			 * 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
969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985
/*
 * 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 */
986
	xfs_extlen_t	gtlena;		/* aligned ... */
L
Linus Torvalds 已提交
987 988 989 990 991 992 993 994
	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 */
995
	xfs_extlen_t	ltlena;		/* aligned ... */
L
Linus Torvalds 已提交
996 997
	xfs_agblock_t	ltnew;		/* useful start bno of left side */
	xfs_extlen_t	rlen;		/* length of returned extent */
998 999
	bool		busy;
	unsigned	busy_gen;
1000
#ifdef DEBUG
L
Linus Torvalds 已提交
1001 1002 1003 1004 1005
	/*
	 * Randomly don't execute the first algorithm.
	 */
	int		dofirst;	/* set to do first algorithm */

1006
	dofirst = prandom_u32() & 1;
L
Linus Torvalds 已提交
1007
#endif
1008

1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
	/* 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;

1020 1021 1022 1023 1024 1025
restart:
	bno_cur_lt = NULL;
	bno_cur_gt = NULL;
	ltlen = 0;
	gtlena = 0;
	ltlena = 0;
1026
	busy = false;
1027

L
Linus Torvalds 已提交
1028 1029 1030
	/*
	 * Get a cursor for the by-size btree.
	 */
1031 1032
	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_CNT);
1033

L
Linus Torvalds 已提交
1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
	/*
	 * 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);
1049
			trace_xfs_alloc_near_noentry(args);
L
Linus Torvalds 已提交
1050 1051 1052 1053 1054
			return 0;
		}
		ASSERT(i == 1);
	}
	args->wasfromfl = 0;
1055

L
Linus Torvalds 已提交
1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
	/*
	 * 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;

1072 1073
#ifdef DEBUG
		if (dofirst)
L
Linus Torvalds 已提交
1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
			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;
1088
				XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1089 1090
				if (ltlen >= args->minlen)
					break;
1091
				if ((error = xfs_btree_increment(cnt_cur, 0, &i)))
L
Linus Torvalds 已提交
1092 1093 1094 1095 1096 1097 1098 1099 1100
					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);
1101
		     error = xfs_btree_increment(cnt_cur, 0, &j)) {
L
Linus Torvalds 已提交
1102 1103 1104 1105 1106 1107
			/*
			 * 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;
1108
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1109 1110
			busy = xfs_alloc_compute_aligned(args, ltbno, ltlen,
					&ltbnoa, &ltlena, &busy_gen);
1111
			if (ltlena < args->minlen)
L
Linus Torvalds 已提交
1112
				continue;
1113 1114
			if (ltbnoa < args->min_agbno || ltbnoa > args->max_agbno)
				continue;
L
Linus Torvalds 已提交
1115 1116 1117 1118 1119 1120
			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,
1121
				args->alignment, args->datatype, ltbnoa,
1122
				ltlena, &ltnew);
L
Linus Torvalds 已提交
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
			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;
1143
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1144
		ASSERT(ltbno + ltlen <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
L
Linus Torvalds 已提交
1145
		args->len = blen;
1146

L
Linus Torvalds 已提交
1147 1148 1149 1150 1151
		/*
		 * We are allocating starting at bnew for blen blocks.
		 */
		args->agbno = bnew;
		ASSERT(bnew >= ltbno);
1152
		ASSERT(bnew + blen <= ltbno + ltlen);
L
Linus Torvalds 已提交
1153 1154 1155
		/*
		 * Set up a cursor for the by-bno tree.
		 */
1156 1157
		bno_cur_lt = xfs_allocbt_init_cursor(args->mp, args->tp,
			args->agbp, args->agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1158 1159 1160 1161 1162 1163 1164 1165
		/*
		 * 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
1166 1167

		trace_xfs_alloc_near_first(args);
L
Linus Torvalds 已提交
1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
		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.
	 */
1185 1186
	bno_cur_lt = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
	/*
	 * 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.
	 */
1209
	if ((error = xfs_btree_increment(bno_cur_gt, 0, &i)))
L
Linus Torvalds 已提交
1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
		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;
1227
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1228 1229
			busy |= xfs_alloc_compute_aligned(args, ltbno, ltlen,
					&ltbnoa, &ltlena, &busy_gen);
1230
			if (ltlena >= args->minlen && ltbnoa >= args->min_agbno)
L
Linus Torvalds 已提交
1231
				break;
1232
			if ((error = xfs_btree_decrement(bno_cur_lt, 0, &i)))
L
Linus Torvalds 已提交
1233
				goto error0;
1234
			if (!i || ltbnoa < args->min_agbno) {
L
Linus Torvalds 已提交
1235 1236 1237 1238 1239 1240 1241 1242
				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;
1243
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1244 1245
			busy |= xfs_alloc_compute_aligned(args, gtbno, gtlen,
					&gtbnoa, &gtlena, &busy_gen);
1246
			if (gtlena >= args->minlen && gtbnoa <= args->max_agbno)
L
Linus Torvalds 已提交
1247
				break;
1248
			if ((error = xfs_btree_increment(bno_cur_gt, 0, &i)))
L
Linus Torvalds 已提交
1249
				goto error0;
1250
			if (!i || gtbnoa > args->max_agbno) {
L
Linus Torvalds 已提交
1251 1252 1253 1254 1255 1256
				xfs_btree_del_cursor(bno_cur_gt,
						     XFS_BTREE_NOERROR);
				bno_cur_gt = NULL;
			}
		}
	} while (bno_cur_lt || bno_cur_gt);
1257

L
Linus Torvalds 已提交
1258 1259 1260 1261 1262 1263
	/*
	 * Got both cursors still active, need to find better entry.
	 */
	if (bno_cur_lt && bno_cur_gt) {
		if (ltlena >= args->minlen) {
			/*
1264
			 * Left side is good, look for a right side entry.
L
Linus Torvalds 已提交
1265 1266 1267
			 */
			args->len = XFS_EXTLEN_MIN(ltlena, args->maxlen);
			xfs_alloc_fix_len(args);
1268
			ltdiff = xfs_alloc_compute_diff(args->agbno, args->len,
1269
				args->alignment, args->datatype, ltbnoa,
1270
				ltlena, &ltnew);
1271 1272 1273

			error = xfs_alloc_find_best_extent(args,
						&bno_cur_lt, &bno_cur_gt,
1274 1275
						ltdiff, &gtbno, &gtlen,
						&gtbnoa, &gtlena,
1276 1277 1278 1279
						0 /* search right */);
		} else {
			ASSERT(gtlena >= args->minlen);

L
Linus Torvalds 已提交
1280
			/*
1281
			 * Right side is good, look for a left side entry.
L
Linus Torvalds 已提交
1282 1283 1284
			 */
			args->len = XFS_EXTLEN_MIN(gtlena, args->maxlen);
			xfs_alloc_fix_len(args);
1285
			gtdiff = xfs_alloc_compute_diff(args->agbno, args->len,
1286
				args->alignment, args->datatype, gtbnoa,
1287
				gtlena, &gtnew);
1288 1289 1290

			error = xfs_alloc_find_best_extent(args,
						&bno_cur_gt, &bno_cur_lt,
1291 1292
						gtdiff, &ltbno, &ltlen,
						&ltbnoa, &ltlena,
1293
						1 /* search left */);
L
Linus Torvalds 已提交
1294
		}
1295 1296 1297

		if (error)
			goto error0;
L
Linus Torvalds 已提交
1298
	}
1299

L
Linus Torvalds 已提交
1300 1301 1302 1303
	/*
	 * If we couldn't get anything, give up.
	 */
	if (bno_cur_lt == NULL && bno_cur_gt == NULL) {
1304 1305
		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);

1306
		if (busy) {
1307
			trace_xfs_alloc_near_busy(args);
1308
			xfs_extent_busy_flush(args->mp, args->pag, busy_gen);
1309 1310
			goto restart;
		}
C
Christoph Hellwig 已提交
1311
		trace_xfs_alloc_size_neither(args);
L
Linus Torvalds 已提交
1312 1313 1314
		args->agbno = NULLAGBLOCK;
		return 0;
	}
1315

L
Linus Torvalds 已提交
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331
	/*
	 * 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;
1332

L
Linus Torvalds 已提交
1333 1334 1335 1336 1337 1338
	/*
	 * 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;
1339
	(void)xfs_alloc_compute_diff(args->agbno, rlen, args->alignment,
1340
				     args->datatype, ltbnoa, ltlena, &ltnew);
L
Linus Torvalds 已提交
1341
	ASSERT(ltnew >= ltbno);
1342
	ASSERT(ltnew + rlen <= ltbnoa + ltlena);
1343
	ASSERT(ltnew + rlen <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
1344
	ASSERT(ltnew >= args->min_agbno && ltnew <= args->max_agbno);
L
Linus Torvalds 已提交
1345
	args->agbno = ltnew;
1346

L
Linus Torvalds 已提交
1347 1348 1349
	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
1350 1351 1352 1353 1354 1355

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

L
Linus Torvalds 已提交
1356 1357 1358 1359 1360
	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
1361
	trace_xfs_alloc_near_error(args);
L
Linus Torvalds 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
	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 */
1389 1390
	bool		busy;
	unsigned	busy_gen;
L
Linus Torvalds 已提交
1391

1392
restart:
L
Linus Torvalds 已提交
1393 1394 1395
	/*
	 * Allocate and initialize a cursor for the by-size btree.
	 */
1396 1397
	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_CNT);
L
Linus Torvalds 已提交
1398
	bno_cur = NULL;
1399
	busy = false;
1400

L
Linus Torvalds 已提交
1401 1402 1403 1404 1405 1406
	/*
	 * 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;
1407

L
Linus Torvalds 已提交
1408
	/*
1409 1410 1411 1412
	 * 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
1413
	 * are no smaller extents available.
L
Linus Torvalds 已提交
1414
	 */
1415
	if (!i) {
1416 1417 1418
		error = xfs_alloc_ag_vextent_small(args, cnt_cur,
						   &fbno, &flen, &i);
		if (error)
L
Linus Torvalds 已提交
1419 1420 1421
			goto error0;
		if (i == 0 || flen == 0) {
			xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
C
Christoph Hellwig 已提交
1422
			trace_xfs_alloc_size_noentry(args);
L
Linus Torvalds 已提交
1423 1424 1425
			return 0;
		}
		ASSERT(i == 1);
1426 1427
		busy = xfs_alloc_compute_aligned(args, fbno, flen, &rbno,
				&rlen, &busy_gen);
1428 1429 1430 1431 1432 1433 1434 1435
	} 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;
1436
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
1437

1438 1439
			busy = xfs_alloc_compute_aligned(args, fbno, flen,
					&rbno, &rlen, &busy_gen);
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450

			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
1451
				 * retrying.
1452 1453 1454 1455
				 */
				xfs_btree_del_cursor(cnt_cur,
						     XFS_BTREE_NOERROR);
				trace_xfs_alloc_size_busy(args);
1456 1457
				xfs_extent_busy_flush(args->mp,
							args->pag, busy_gen);
1458 1459 1460
				goto restart;
			}
		}
L
Linus Torvalds 已提交
1461
	}
1462

L
Linus Torvalds 已提交
1463 1464 1465 1466 1467 1468 1469
	/*
	 * 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);
1470
	XFS_WANT_CORRUPTED_GOTO(args->mp, rlen == 0 ||
L
Linus Torvalds 已提交
1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
			(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 (;;) {
1483
			if ((error = xfs_btree_decrement(cnt_cur, 0, &i)))
L
Linus Torvalds 已提交
1484 1485 1486 1487 1488 1489
				goto error0;
			if (i == 0)
				break;
			if ((error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen,
					&i)))
				goto error0;
1490
			XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1491 1492
			if (flen < bestrlen)
				break;
1493 1494
			busy = xfs_alloc_compute_aligned(args, fbno, flen,
					&rbno, &rlen, &busy_gen);
L
Linus Torvalds 已提交
1495
			rlen = XFS_EXTLEN_MIN(args->maxlen, rlen);
1496
			XFS_WANT_CORRUPTED_GOTO(args->mp, rlen == 0 ||
L
Linus Torvalds 已提交
1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
				(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;
1511
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521
		rlen = bestrlen;
		rbno = bestrbno;
		flen = bestflen;
		fbno = bestfbno;
	}
	args->wasfromfl = 0;
	/*
	 * Fix up the length.
	 */
	args->len = rlen;
1522
	if (rlen < args->minlen) {
1523
		if (busy) {
1524 1525
			xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
			trace_xfs_alloc_size_busy(args);
1526
			xfs_extent_busy_flush(args->mp, args->pag, busy_gen);
1527 1528 1529
			goto restart;
		}
		goto out_nominleft;
L
Linus Torvalds 已提交
1530
	}
1531 1532
	xfs_alloc_fix_len(args);

L
Linus Torvalds 已提交
1533
	rlen = args->len;
1534
	XFS_WANT_CORRUPTED_GOTO(args->mp, rlen <= flen, error0);
L
Linus Torvalds 已提交
1535 1536 1537
	/*
	 * Allocate and initialize a cursor for the by-block tree.
	 */
1538 1539
	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
		args->agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1540 1541 1542 1543 1544 1545 1546 1547
	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;
1548
	XFS_WANT_CORRUPTED_GOTO(args->mp,
L
Linus Torvalds 已提交
1549
		args->agbno + args->len <=
1550
			be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length),
L
Linus Torvalds 已提交
1551
		error0);
C
Christoph Hellwig 已提交
1552
	trace_xfs_alloc_size_done(args);
L
Linus Torvalds 已提交
1553 1554 1555
	return 0;

error0:
C
Christoph Hellwig 已提交
1556
	trace_xfs_alloc_size_error(args);
L
Linus Torvalds 已提交
1557 1558 1559 1560 1561
	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;
1562 1563 1564 1565 1566 1567

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
1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582
}

/*
 * 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 */
{
1583
	struct xfs_owner_info	oinfo;
L
Linus Torvalds 已提交
1584 1585 1586 1587 1588
	int		error;
	xfs_agblock_t	fbno;
	xfs_extlen_t	flen;
	int		i;

1589
	if ((error = xfs_btree_decrement(ccur, 0, &i)))
L
Linus Torvalds 已提交
1590 1591 1592 1593
		goto error0;
	if (i) {
		if ((error = xfs_alloc_get_rec(ccur, &fbno, &flen, &i)))
			goto error0;
1594
		XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0);
L
Linus Torvalds 已提交
1595 1596 1597 1598 1599 1600
	}
	/*
	 * Nothing in the btree, try the freelist.  Make sure
	 * to respect minleft even when pulling from the
	 * freelist.
	 */
1601
	else if (args->minlen == 1 && args->alignment == 1 &&
1602
		 args->resv != XFS_AG_RESV_AGFL &&
1603 1604
		 (be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_flcount)
		  > args->minleft)) {
D
David Chinner 已提交
1605 1606
		error = xfs_alloc_get_freelist(args->tp, args->agbp, &fbno, 0);
		if (error)
L
Linus Torvalds 已提交
1607 1608
			goto error0;
		if (fbno != NULLAGBLOCK) {
D
Dave Chinner 已提交
1609
			xfs_extent_busy_reuse(args->mp, args->agno, fbno, 1,
1610
			      xfs_alloc_allow_busy_reuse(args->datatype));
1611

1612
			if (xfs_alloc_is_userdata(args->datatype)) {
L
Linus Torvalds 已提交
1613 1614 1615 1616
				xfs_buf_t	*bp;

				bp = xfs_btree_get_bufs(args->mp, args->tp,
					args->agno, fbno, 0);
1617 1618 1619 1620
				if (!bp) {
					error = -EFSCORRUPTED;
					goto error0;
				}
L
Linus Torvalds 已提交
1621 1622 1623 1624
				xfs_trans_binval(args->tp, bp);
			}
			args->len = 1;
			args->agbno = fbno;
1625
			XFS_WANT_CORRUPTED_GOTO(args->mp,
L
Linus Torvalds 已提交
1626
				args->agbno + args->len <=
1627
				be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length),
L
Linus Torvalds 已提交
1628 1629
				error0);
			args->wasfromfl = 1;
C
Christoph Hellwig 已提交
1630
			trace_xfs_alloc_small_freelist(args);
1631 1632 1633 1634

			/*
			 * If we're feeding an AGFL block to something that
			 * doesn't live in the free space, we need to clear
1635
			 * out the OWN_AG rmap.
1636 1637 1638 1639 1640 1641 1642
			 */
			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
1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654
			*stat = 0;
			return 0;
		}
		/*
		 * Nothing in the freelist.
		 */
		else
			flen = 0;
	}
	/*
	 * Can't allocate from the freelist for some reason.
	 */
1655 1656
	else {
		fbno = NULLAGBLOCK;
L
Linus Torvalds 已提交
1657
		flen = 0;
1658
	}
L
Linus Torvalds 已提交
1659 1660 1661 1662 1663
	/*
	 * Can't do the allocation, give up.
	 */
	if (flen < args->minlen) {
		args->agbno = NULLAGBLOCK;
C
Christoph Hellwig 已提交
1664
		trace_xfs_alloc_small_notenough(args);
L
Linus Torvalds 已提交
1665 1666 1667 1668 1669
		flen = 0;
	}
	*fbnop = fbno;
	*flenp = flen;
	*stat = 1;
C
Christoph Hellwig 已提交
1670
	trace_xfs_alloc_small_done(args);
L
Linus Torvalds 已提交
1671 1672 1673
	return 0;

error0:
C
Christoph Hellwig 已提交
1674
	trace_xfs_alloc_small_error(args);
L
Linus Torvalds 已提交
1675 1676 1677 1678 1679 1680
	return error;
}

/*
 * Free the extent starting at agno/bno for length.
 */
1681
STATIC int
L
Linus Torvalds 已提交
1682
xfs_free_ag_extent(
1683 1684 1685 1686 1687 1688
	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,
1689
	enum xfs_ag_resv_type	type)
L
Linus Torvalds 已提交
1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703
{
	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 */
1704
	xfs_perag_t	*pag;		/* per allocation group data */
L
Linus Torvalds 已提交
1705

1706
	bno_cur = cnt_cur = NULL;
L
Linus Torvalds 已提交
1707
	mp = tp->t_mountp;
1708

1709
	if (!xfs_rmap_should_skip_owner_update(oinfo)) {
1710 1711 1712 1713 1714
		error = xfs_rmap_free(tp, agbp, agno, bno, len, oinfo);
		if (error)
			goto error0;
	}

L
Linus Torvalds 已提交
1715 1716 1717
	/*
	 * Allocate and initialize a cursor for the by-block btree.
	 */
1718
	bno_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_BNO);
L
Linus Torvalds 已提交
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
	/*
	 * 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;
1731
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742
		/*
		 * 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.
			 */
1743 1744
			XFS_WANT_CORRUPTED_GOTO(mp,
						ltbno + ltlen <= bno, error0);
L
Linus Torvalds 已提交
1745 1746 1747 1748 1749 1750
		}
	}
	/*
	 * Look for a neighboring block on the right (higher block numbers)
	 * that is contiguous with this space.
	 */
1751
	if ((error = xfs_btree_increment(bno_cur, 0, &haveright)))
L
Linus Torvalds 已提交
1752 1753 1754 1755 1756 1757 1758
		goto error0;
	if (haveright) {
		/*
		 * There is a block to our right.
		 */
		if ((error = xfs_alloc_get_rec(bno_cur, &gtbno, &gtlen, &i)))
			goto error0;
1759
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770
		/*
		 * 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.
			 */
1771
			XFS_WANT_CORRUPTED_GOTO(mp, gtbno >= bno + len, error0);
L
Linus Torvalds 已提交
1772 1773 1774 1775 1776
		}
	}
	/*
	 * Now allocate and initialize a cursor for the by-size tree.
	 */
1777
	cnt_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_CNT);
L
Linus Torvalds 已提交
1778 1779 1780 1781 1782 1783 1784 1785 1786 1787
	/*
	 * 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;
1788
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1789
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1790
			goto error0;
1791
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1792 1793 1794 1795 1796
		/*
		 * Delete the old by-size entry on the right.
		 */
		if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
			goto error0;
1797
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1798
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1799
			goto error0;
1800
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1801 1802 1803
		/*
		 * Delete the old by-block entry for the right block.
		 */
1804
		if ((error = xfs_btree_delete(bno_cur, &i)))
L
Linus Torvalds 已提交
1805
			goto error0;
1806
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1807 1808 1809
		/*
		 * Move the by-block cursor back to the left neighbor.
		 */
1810
		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
L
Linus Torvalds 已提交
1811
			goto error0;
1812
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824
#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;
1825
			XFS_WANT_CORRUPTED_GOTO(mp,
L
Linus Torvalds 已提交
1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847
				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;
1848
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1849
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1850
			goto error0;
1851
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1852 1853 1854 1855
		/*
		 * Back up the by-block cursor to the left neighbor, and
		 * update its length.
		 */
1856
		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
L
Linus Torvalds 已提交
1857
			goto error0;
1858
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
		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;
1874
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
1875
		if ((error = xfs_btree_delete(cnt_cur, &i)))
L
Linus Torvalds 已提交
1876
			goto error0;
1877
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893
		/*
		 * 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;
1894
		if ((error = xfs_btree_insert(bno_cur, &i)))
L
Linus Torvalds 已提交
1895
			goto error0;
1896
		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1897 1898 1899 1900 1901 1902 1903 1904
	}
	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;
1905
	XFS_WANT_CORRUPTED_GOTO(mp, i == 0, error0);
1906
	if ((error = xfs_btree_insert(cnt_cur, &i)))
L
Linus Torvalds 已提交
1907
		goto error0;
1908
	XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0);
L
Linus Torvalds 已提交
1909 1910
	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
	cnt_cur = NULL;
1911

L
Linus Torvalds 已提交
1912 1913 1914
	/*
	 * Update the freespace totals in the ag and superblock.
	 */
1915 1916
	pag = xfs_perag_get(mp, agno);
	error = xfs_alloc_update_counters(tp, pag, agbp, len);
1917
	xfs_ag_resv_free_extent(pag, type, tp, len);
1918 1919 1920 1921
	xfs_perag_put(pag);
	if (error)
		goto error0;

1922 1923
	XFS_STATS_INC(mp, xs_freex);
	XFS_STATS_ADD(mp, xs_freeb, len);
C
Christoph Hellwig 已提交
1924

1925
	trace_xfs_free_extent(mp, agno, bno, len, type, haveleft, haveright);
L
Linus Torvalds 已提交
1926 1927 1928 1929

	return 0;

 error0:
1930
	trace_xfs_free_extent(mp, agno, bno, len, type, -1, -1);
L
Linus Torvalds 已提交
1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
	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 */
{
1950
	mp->m_ag_maxlevels = xfs_btree_compute_maxlevels(mp->m_alloc_mnr,
1951
			(mp->m_sb.sb_agblocks + 1) / 2);
L
Linus Torvalds 已提交
1952 1953
}

1954
/*
1955 1956 1957 1958
 * 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.
1959 1960 1961
 */
xfs_extlen_t
xfs_alloc_longest_free_extent(
1962
	struct xfs_perag	*pag,
1963 1964
	xfs_extlen_t		need,
	xfs_extlen_t		reserved)
1965
{
1966
	xfs_extlen_t		delta = 0;
1967

1968 1969 1970 1971
	/*
	 * If the AGFL needs a recharge, we'll have to subtract that from the
	 * longest extent.
	 */
1972 1973 1974
	if (need > pag->pagf_flcount)
		delta = need - pag->pagf_flcount;

1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
	/*
	 * 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.
	 */
1987 1988
	if (pag->pagf_longest > delta)
		return pag->pagf_longest - delta;
1989 1990

	/* Otherwise, let the caller try for 1 block if there's space. */
1991 1992 1993
	return pag->pagf_flcount > 0 || pag->pagf_longest > 0;
}

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
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);
2007 2008 2009 2010 2011
	/* 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);
2012 2013 2014 2015

	return min_free;
}

2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028
/*
 * 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;
2029
	xfs_extlen_t		alloc_len, longest;
2030
	xfs_extlen_t		reservation; /* blocks that are still reserved */
2031 2032 2033 2034 2035
	int			available;

	if (flags & XFS_ALLOC_FLAG_FREEING)
		return true;

2036 2037
	reservation = xfs_ag_resv_needed(pag, args->resv);

2038
	/* do we have enough contiguous free space for the allocation? */
2039
	alloc_len = args->minlen + (args->alignment - 1) + args->minalignslop;
2040
	longest = xfs_alloc_longest_free_extent(pag, min_free, reservation);
2041
	if (longest < alloc_len)
2042 2043
		return false;

2044
	/* do we have enough free space remaining for the allocation? */
2045
	available = (int)(pag->pagf_freeblks + pag->pagf_flcount -
2046
			  reservation - min_free - args->minleft);
2047
	if (available < (int)max(args->total, alloc_len))
2048 2049
		return false;

2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
	/*
	 * 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);
	}

2060 2061 2062
	return true;
}

2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086
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;
}

2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 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
/*
 * 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;
}

L
Linus Torvalds 已提交
2174 2175 2176 2177
/*
 * Decide whether to use this allocation group for this allocation.
 * If so, fix up the btree freelist's size.
 */
2178
int			/* error */
L
Linus Torvalds 已提交
2179
xfs_alloc_fix_freelist(
2180 2181
	struct xfs_alloc_arg	*args,	/* allocation argument structure */
	int			flags)	/* XFS_ALLOC_FLAG_... */
L
Linus Torvalds 已提交
2182
{
2183 2184 2185 2186 2187 2188 2189 2190
	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 */
2191
	int			error = 0;
2192

L
Linus Torvalds 已提交
2193
	if (!pag->pagf_init) {
2194 2195 2196
		error = xfs_alloc_read_agf(mp, tp, args->agno, flags, &agbp);
		if (error)
			goto out_no_agbp;
L
Linus Torvalds 已提交
2197
		if (!pag->pagf_init) {
2198 2199
			ASSERT(flags & XFS_ALLOC_FLAG_TRYLOCK);
			ASSERT(!(flags & XFS_ALLOC_FLAG_FREEING));
2200
			goto out_agbp_relse;
L
Linus Torvalds 已提交
2201
		}
2202
	}
L
Linus Torvalds 已提交
2203

2204
	/*
2205 2206 2207
	 * 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
2208
	 */
2209
	if (pag->pagf_metadata && xfs_alloc_is_userdata(args->datatype) &&
2210 2211
	    (flags & XFS_ALLOC_FLAG_TRYLOCK)) {
		ASSERT(!(flags & XFS_ALLOC_FLAG_FREEING));
2212
		goto out_agbp_relse;
L
Linus Torvalds 已提交
2213 2214
	}

2215
	need = xfs_alloc_min_freelist(mp, pag);
2216 2217
	if (!xfs_alloc_space_available(args, need, flags |
			XFS_ALLOC_FLAG_CHECK))
2218
		goto out_agbp_relse;
2219

L
Linus Torvalds 已提交
2220 2221 2222 2223
	/*
	 * Get the a.g. freespace buffer.
	 * Can fail if we're not blocking on locks, and it's held.
	 */
2224 2225 2226 2227 2228
	if (!agbp) {
		error = xfs_alloc_read_agf(mp, tp, args->agno, flags, &agbp);
		if (error)
			goto out_no_agbp;
		if (!agbp) {
2229 2230
			ASSERT(flags & XFS_ALLOC_FLAG_TRYLOCK);
			ASSERT(!(flags & XFS_ALLOC_FLAG_FREEING));
2231
			goto out_no_agbp;
2232
		}
L
Linus Torvalds 已提交
2233
	}
2234

2235 2236 2237 2238
	/* reset a padding mismatched agfl before final free space check */
	if (pag->pagf_agflreset)
		xfs_agfl_reset(tp, agbp, pag);

2239
	/* If there isn't enough total space or single-extent, reject it. */
2240
	need = xfs_alloc_min_freelist(mp, pag);
2241 2242
	if (!xfs_alloc_space_available(args, need, flags))
		goto out_agbp_relse;
2243

L
Linus Torvalds 已提交
2244 2245
	/*
	 * Make the freelist shorter if it's too long.
2246
	 *
2247 2248 2249 2250 2251 2252 2253 2254
	 * 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.
	 *
2255 2256 2257 2258
	 * 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?
2259 2260 2261 2262 2263 2264 2265 2266
	 *
	 * 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
2267
	 */
2268 2269 2270 2271 2272 2273
	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
2274 2275
		error = xfs_alloc_get_freelist(tp, agbp, &bno, 0);
		if (error)
2276
			goto out_agbp_relse;
2277 2278 2279

		error = xfs_free_agfl_block(tp, args->agno, bno, agbp,
					    &targs.oinfo);
2280
		if (error)
2281
			goto out_agbp_relse;
L
Linus Torvalds 已提交
2282
	}
2283

L
Linus Torvalds 已提交
2284 2285 2286 2287
	targs.tp = tp;
	targs.mp = mp;
	targs.agbp = agbp;
	targs.agno = args->agno;
2288
	targs.alignment = targs.minlen = targs.prod = 1;
L
Linus Torvalds 已提交
2289 2290
	targs.type = XFS_ALLOCTYPE_THIS_AG;
	targs.pag = pag;
2291 2292
	error = xfs_alloc_read_agfl(mp, tp, targs.agno, &agflbp);
	if (error)
2293
		goto out_agbp_relse;
2294 2295 2296

	/* Make the freelist longer if it's too short. */
	while (pag->pagf_flcount < need) {
L
Linus Torvalds 已提交
2297
		targs.agbno = 0;
2298
		targs.maxlen = need - pag->pagf_flcount;
2299
		targs.resv = XFS_AG_RESV_AGFL;
2300 2301 2302

		/* Allocate as many blocks as possible at once. */
		error = xfs_alloc_ag_vextent(&targs);
2303 2304 2305
		if (error)
			goto out_agflbp_relse;

L
Linus Torvalds 已提交
2306 2307 2308 2309 2310
		/*
		 * 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.
		 */
2311
		if (targs.agbno == NULLAGBLOCK) {
2312 2313
			if (flags & XFS_ALLOC_FLAG_FREEING)
				break;
2314
			goto out_agflbp_relse;
2315
		}
L
Linus Torvalds 已提交
2316 2317 2318 2319
		/*
		 * Put each allocated block on the list.
		 */
		for (bno = targs.agbno; bno < targs.agbno + targs.len; bno++) {
D
David Chinner 已提交
2320 2321 2322
			error = xfs_alloc_put_freelist(tp, agbp,
							agflbp, bno, 0);
			if (error)
2323
				goto out_agflbp_relse;
L
Linus Torvalds 已提交
2324 2325
		}
	}
2326
	xfs_trans_brelse(tp, agflbp);
L
Linus Torvalds 已提交
2327 2328
	args->agbp = agbp;
	return 0;
2329 2330 2331 2332 2333 2334 2335 2336 2337

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
2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
}

/*
 * 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
2348 2349
	xfs_agblock_t	*bnop,	/* block address retrieved from freelist */
	int		btreeblk) /* destination is a AGF btree */
L
Linus Torvalds 已提交
2350 2351 2352 2353
{
	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 */
2354
	__be32		*agfl_bno;
L
Linus Torvalds 已提交
2355
	int		error;
D
David Chinner 已提交
2356
	int		logflags;
2357
	xfs_mount_t	*mp = tp->t_mountp;
L
Linus Torvalds 已提交
2358 2359 2360 2361 2362
	xfs_perag_t	*pag;	/* per allocation group data */

	/*
	 * Freelist is empty, give up.
	 */
2363
	agf = XFS_BUF_TO_AGF(agbp);
L
Linus Torvalds 已提交
2364 2365 2366 2367 2368 2369 2370
	if (!agf->agf_flcount) {
		*bnop = NULLAGBLOCK;
		return 0;
	}
	/*
	 * Read the array of free blocks.
	 */
2371 2372 2373
	error = xfs_alloc_read_agfl(mp, tp, be32_to_cpu(agf->agf_seqno),
				    &agflbp);
	if (error)
L
Linus Torvalds 已提交
2374
		return error;
2375 2376


L
Linus Torvalds 已提交
2377 2378 2379
	/*
	 * Get the block number and update the data structures.
	 */
2380 2381
	agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, agflbp);
	bno = be32_to_cpu(agfl_bno[be32_to_cpu(agf->agf_flfirst)]);
2382
	be32_add_cpu(&agf->agf_flfirst, 1);
L
Linus Torvalds 已提交
2383
	xfs_trans_brelse(tp, agflbp);
2384
	if (be32_to_cpu(agf->agf_flfirst) == xfs_agfl_size(mp))
L
Linus Torvalds 已提交
2385
		agf->agf_flfirst = 0;
2386 2387

	pag = xfs_perag_get(mp, be32_to_cpu(agf->agf_seqno));
2388
	ASSERT(!pag->pagf_agflreset);
2389
	be32_add_cpu(&agf->agf_flcount, -1);
L
Linus Torvalds 已提交
2390 2391
	xfs_trans_agflist_delta(tp, -1);
	pag->pagf_flcount--;
2392
	xfs_perag_put(pag);
D
David Chinner 已提交
2393 2394 2395

	logflags = XFS_AGF_FLFIRST | XFS_AGF_FLCOUNT;
	if (btreeblk) {
2396
		be32_add_cpu(&agf->agf_btreeblks, 1);
D
David Chinner 已提交
2397 2398 2399 2400 2401
		pag->pagf_btreeblks++;
		logflags |= XFS_AGF_BTREEBLKS;
	}

	xfs_alloc_log_agf(tp, agbp, logflags);
L
Linus Torvalds 已提交
2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429
	*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
2430
		offsetof(xfs_agf_t, agf_btreeblks),
D
Dave Chinner 已提交
2431
		offsetof(xfs_agf_t, agf_uuid),
2432
		offsetof(xfs_agf_t, agf_rmap_blocks),
2433 2434 2435
		offsetof(xfs_agf_t, agf_refcount_blocks),
		offsetof(xfs_agf_t, agf_refcount_root),
		offsetof(xfs_agf_t, agf_refcount_level),
2436 2437
		/* 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
2438 2439 2440
		sizeof(xfs_agf_t)
	};

C
Christoph Hellwig 已提交
2441 2442
	trace_xfs_agf(tp->t_mountp, XFS_BUF_TO_AGF(bp), fields, _RET_IP_);

2443
	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_AGF_BUF);
D
Dave Chinner 已提交
2444

L
Linus Torvalds 已提交
2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476
	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
2477 2478
	xfs_agblock_t		bno,	/* block being freed */
	int			btreeblk) /* block came from a AGF btree */
L
Linus Torvalds 已提交
2479 2480
{
	xfs_agf_t		*agf;	/* a.g. freespace structure */
2481
	__be32			*blockp;/* pointer to array entry */
L
Linus Torvalds 已提交
2482
	int			error;
D
David Chinner 已提交
2483
	int			logflags;
L
Linus Torvalds 已提交
2484 2485
	xfs_mount_t		*mp;	/* mount structure */
	xfs_perag_t		*pag;	/* per allocation group data */
2486 2487
	__be32			*agfl_bno;
	int			startoff;
L
Linus Torvalds 已提交
2488 2489 2490 2491 2492

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

	if (!agflbp && (error = xfs_alloc_read_agfl(mp, tp,
2493
			be32_to_cpu(agf->agf_seqno), &agflbp)))
L
Linus Torvalds 已提交
2494
		return error;
2495
	be32_add_cpu(&agf->agf_fllast, 1);
2496
	if (be32_to_cpu(agf->agf_fllast) == xfs_agfl_size(mp))
L
Linus Torvalds 已提交
2497
		agf->agf_fllast = 0;
2498 2499

	pag = xfs_perag_get(mp, be32_to_cpu(agf->agf_seqno));
2500
	ASSERT(!pag->pagf_agflreset);
2501
	be32_add_cpu(&agf->agf_flcount, 1);
L
Linus Torvalds 已提交
2502 2503
	xfs_trans_agflist_delta(tp, 1);
	pag->pagf_flcount++;
D
David Chinner 已提交
2504 2505 2506

	logflags = XFS_AGF_FLLAST | XFS_AGF_FLCOUNT;
	if (btreeblk) {
2507
		be32_add_cpu(&agf->agf_btreeblks, -1);
D
David Chinner 已提交
2508 2509 2510
		pag->pagf_btreeblks--;
		logflags |= XFS_AGF_BTREEBLKS;
	}
2511
	xfs_perag_put(pag);
D
David Chinner 已提交
2512 2513 2514

	xfs_alloc_log_agf(tp, agbp, logflags);

2515
	ASSERT(be32_to_cpu(agf->agf_flcount) <= xfs_agfl_size(mp));
2516 2517 2518

	agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, agflbp);
	blockp = &agfl_bno[be32_to_cpu(agf->agf_fllast)];
2519
	*blockp = cpu_to_be32(bno);
2520 2521
	startoff = (char *)blockp - (char *)agflbp->b_addr;

D
David Chinner 已提交
2522
	xfs_alloc_log_agf(tp, agbp, logflags);
2523

2524
	xfs_trans_buf_set_type(tp, agflbp, XFS_BLFT_AGFL_BUF);
2525 2526
	xfs_trans_log_buf(tp, agflbp, startoff,
			  startoff + sizeof(xfs_agblock_t) - 1);
L
Linus Torvalds 已提交
2527 2528 2529
	return 0;
}

2530
static xfs_failaddr_t
2531
xfs_agf_verify(
2532 2533 2534 2535
	struct xfs_buf		*bp)
{
	struct xfs_mount	*mp = bp->b_target->bt_mount;
	struct xfs_agf		*agf = XFS_BUF_TO_AGF(bp);
2536

2537 2538
	if (xfs_sb_version_hascrc(&mp->m_sb)) {
		if (!uuid_equal(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid))
2539
			return __this_address;
2540 2541
		if (!xfs_log_check_lsn(mp,
				be64_to_cpu(XFS_BUF_TO_AGF(bp)->agf_lsn)))
2542
			return __this_address;
2543
	}
2544

D
Dave Chinner 已提交
2545 2546 2547
	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) &&
2548 2549 2550
	      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)))
2551
		return __this_address;
2552

2553 2554 2555
	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 ||
2556
	    be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]) > XFS_BTREE_MAXLEVELS)
2557
		return __this_address;
2558

2559
	if (xfs_sb_version_hasrmapbt(&mp->m_sb) &&
2560 2561
	    (be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) < 1 ||
	     be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) > XFS_BTREE_MAXLEVELS))
2562
		return __this_address;
2563

2564 2565 2566 2567 2568 2569
	/*
	 * 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
2570
	if (bp->b_pag && be32_to_cpu(agf->agf_seqno) != bp->b_pag->pag_agno)
2571
		return __this_address;
2572

D
Dave Chinner 已提交
2573 2574
	if (xfs_sb_version_haslazysbcount(&mp->m_sb) &&
	    be32_to_cpu(agf->agf_btreeblks) > be32_to_cpu(agf->agf_length))
2575
		return __this_address;
D
Dave Chinner 已提交
2576

2577
	if (xfs_sb_version_hasreflink(&mp->m_sb) &&
2578 2579
	    (be32_to_cpu(agf->agf_refcount_level) < 1 ||
	     be32_to_cpu(agf->agf_refcount_level) > XFS_BTREE_MAXLEVELS))
2580
		return __this_address;
2581

2582
	return NULL;
2583

2584 2585
}

2586 2587
static void
xfs_agf_read_verify(
2588 2589
	struct xfs_buf	*bp)
{
D
Dave Chinner 已提交
2590
	struct xfs_mount *mp = bp->b_target->bt_mount;
2591
	xfs_failaddr_t	fa;
D
Dave Chinner 已提交
2592

2593 2594
	if (xfs_sb_version_hascrc(&mp->m_sb) &&
	    !xfs_buf_verify_cksum(bp, XFS_AGF_CRC_OFF))
2595 2596
		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
	else {
2597
		fa = xfs_agf_verify(bp);
2598 2599 2600
		if (XFS_TEST_ERROR(fa, mp, XFS_ERRTAG_ALLOC_READ_AGF))
			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
	}
2601
}
2602

2603
static void
2604
xfs_agf_write_verify(
2605 2606
	struct xfs_buf	*bp)
{
C
Carlos Maiolino 已提交
2607 2608
	struct xfs_mount	*mp = bp->b_target->bt_mount;
	struct xfs_buf_log_item	*bip = bp->b_log_item;
2609
	xfs_failaddr_t		fa;
D
Dave Chinner 已提交
2610

2611
	fa = xfs_agf_verify(bp);
2612 2613
	if (fa) {
		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
D
Dave Chinner 已提交
2614 2615 2616 2617 2618 2619 2620 2621 2622
		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);

2623
	xfs_buf_update_cksum(bp, XFS_AGF_CRC_OFF);
2624 2625
}

2626
const struct xfs_buf_ops xfs_agf_buf_ops = {
2627
	.name = "xfs_agf",
2628 2629
	.verify_read = xfs_agf_read_verify,
	.verify_write = xfs_agf_write_verify,
2630
	.verify_struct = xfs_agf_verify,
2631 2632
};

L
Linus Torvalds 已提交
2633 2634 2635 2636
/*
 * Read in the allocation group header (free/alloc section).
 */
int					/* error */
2637 2638 2639 2640 2641 2642
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
2643 2644 2645
{
	int		error;

2646 2647
	trace_xfs_read_agf(mp, agno);

L
Linus Torvalds 已提交
2648 2649 2650 2651
	ASSERT(agno != NULLAGNUMBER);
	error = xfs_trans_read_buf(
			mp, tp, mp->m_ddev_targp,
			XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
2652
			XFS_FSS_TO_BB(mp, 1), flags, bpp, &xfs_agf_buf_ops);
L
Linus Torvalds 已提交
2653 2654
	if (error)
		return error;
2655
	if (!*bpp)
L
Linus Torvalds 已提交
2656
		return 0;
2657

2658
	ASSERT(!(*bpp)->b_error);
2659
	xfs_buf_set_ref(*bpp, XFS_AGF_REF);
2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677
	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;

2678
	trace_xfs_alloc_read_agf(mp, agno);
2679

2680
	ASSERT(agno != NULLAGNUMBER);
2681
	error = xfs_read_agf(mp, tp, agno,
2682
			(flags & XFS_ALLOC_FLAG_TRYLOCK) ? XBF_TRYLOCK : 0,
2683 2684 2685 2686 2687
			bpp);
	if (error)
		return error;
	if (!*bpp)
		return 0;
2688
	ASSERT(!(*bpp)->b_error);
2689 2690

	agf = XFS_BUF_TO_AGF(*bpp);
2691
	pag = xfs_perag_get(mp, agno);
L
Linus Torvalds 已提交
2692
	if (!pag->pagf_init) {
2693
		pag->pagf_freeblks = be32_to_cpu(agf->agf_freeblks);
D
David Chinner 已提交
2694
		pag->pagf_btreeblks = be32_to_cpu(agf->agf_btreeblks);
2695 2696
		pag->pagf_flcount = be32_to_cpu(agf->agf_flcount);
		pag->pagf_longest = be32_to_cpu(agf->agf_longest);
L
Linus Torvalds 已提交
2697
		pag->pagf_levels[XFS_BTNUM_BNOi] =
2698
			be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]);
L
Linus Torvalds 已提交
2699
		pag->pagf_levels[XFS_BTNUM_CNTi] =
2700
			be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]);
2701 2702
		pag->pagf_levels[XFS_BTNUM_RMAPi] =
			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
2703
		pag->pagf_refcount_level = be32_to_cpu(agf->agf_refcount_level);
E
Eric Sandeen 已提交
2704
		spin_lock_init(&pag->pagb_lock);
2705
		pag->pagb_count = 0;
2706
		pag->pagb_tree = RB_ROOT;
L
Linus Torvalds 已提交
2707
		pag->pagf_init = 1;
2708
		pag->pagf_agflreset = xfs_agfl_needs_reset(mp, agf);
L
Linus Torvalds 已提交
2709 2710 2711
	}
#ifdef DEBUG
	else if (!XFS_FORCED_SHUTDOWN(mp)) {
2712
		ASSERT(pag->pagf_freeblks == be32_to_cpu(agf->agf_freeblks));
2713
		ASSERT(pag->pagf_btreeblks == be32_to_cpu(agf->agf_btreeblks));
2714 2715
		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
2716
		ASSERT(pag->pagf_levels[XFS_BTNUM_BNOi] ==
2717
		       be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]));
L
Linus Torvalds 已提交
2718
		ASSERT(pag->pagf_levels[XFS_BTNUM_CNTi] ==
2719
		       be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]));
L
Linus Torvalds 已提交
2720 2721
	}
#endif
2722
	xfs_perag_put(pag);
L
Linus Torvalds 已提交
2723 2724 2725 2726 2727 2728 2729 2730 2731
	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 */
2732
xfs_alloc_vextent(
L
Linus Torvalds 已提交
2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
	xfs_alloc_arg_t	*args)	/* allocation argument structure */
{
	xfs_agblock_t	agsize;	/* allocation group size */
	int		error;
	int		flags;	/* XFS_ALLOC_FLAG_... locking flags */
	xfs_mount_t	*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 */

	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
2767
		trace_xfs_alloc_vextent_badargs(args);
L
Linus Torvalds 已提交
2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778
		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);
2779
		args->pag = xfs_perag_get(mp, args->agno);
L
Linus Torvalds 已提交
2780 2781
		error = xfs_alloc_fix_freelist(args, 0);
		if (error) {
C
Christoph Hellwig 已提交
2782
			trace_xfs_alloc_vextent_nofix(args);
L
Linus Torvalds 已提交
2783 2784 2785
			goto error0;
		}
		if (!args->agbp) {
C
Christoph Hellwig 已提交
2786
			trace_xfs_alloc_vextent_noagbp(args);
L
Linus Torvalds 已提交
2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797
			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.
		 */
2798
		if ((args->datatype & XFS_ALLOC_INITIAL_USER_DATA) &&
L
Linus Torvalds 已提交
2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811
		    (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.
		 */
2812
		if (type == XFS_ALLOCTYPE_FIRST_AG) {
L
Linus Torvalds 已提交
2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831
			/*
			 * 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 (;;) {
2832
			args->pag = xfs_perag_get(mp, args->agno);
L
Linus Torvalds 已提交
2833 2834
			error = xfs_alloc_fix_freelist(args, flags);
			if (error) {
C
Christoph Hellwig 已提交
2835
				trace_xfs_alloc_vextent_nofix(args);
L
Linus Torvalds 已提交
2836 2837 2838 2839 2840 2841 2842 2843 2844 2845
				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
2846 2847 2848

			trace_xfs_alloc_vextent_loopfailed(args);

L
Linus Torvalds 已提交
2849 2850 2851 2852 2853 2854
			/*
			 * Didn't work, figure out the next iteration.
			 */
			if (args->agno == sagno &&
			    type == XFS_ALLOCTYPE_START_BNO)
				args->type = XFS_ALLOCTYPE_THIS_AG;
2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
			/*
			* 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) {
				if (args->firstblock != NULLFSBLOCK)
					args->agno = sagno;
				else
					args->agno = 0;
			}
L
Linus Torvalds 已提交
2868 2869 2870 2871 2872
			/*
			 * Reached the starting a.g., must either be done
			 * or switch to non-trylock mode.
			 */
			if (args->agno == sagno) {
2873
				if (flags == 0) {
L
Linus Torvalds 已提交
2874
					args->agbno = NULLAGBLOCK;
C
Christoph Hellwig 已提交
2875
					trace_xfs_alloc_vextent_allfailed(args);
L
Linus Torvalds 已提交
2876 2877
					break;
				}
2878 2879 2880 2881 2882 2883

				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
2884 2885
				}
			}
2886
			xfs_perag_put(args->pag);
L
Linus Torvalds 已提交
2887
		}
2888
		if (bump_rotor) {
L
Linus Torvalds 已提交
2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911
			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
2912 2913

		/* Zero the extent if we were asked to do so */
2914
		if (args->datatype & XFS_ALLOC_USERDATA_ZERO) {
2915 2916 2917 2918 2919
			error = xfs_zero_extent(args->ip, args->fsbno, args->len);
			if (error)
				goto error0;
		}

L
Linus Torvalds 已提交
2920
	}
2921
	xfs_perag_put(args->pag);
L
Linus Torvalds 已提交
2922 2923
	return 0;
error0:
2924
	xfs_perag_put(args->pag);
L
Linus Torvalds 已提交
2925 2926 2927
	return error;
}

2928 2929 2930 2931 2932 2933
/* 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
2934
{
2935 2936
	struct xfs_alloc_arg	args;
	int			error;
L
Linus Torvalds 已提交
2937

2938
	memset(&args, 0, sizeof(struct xfs_alloc_arg));
L
Linus Torvalds 已提交
2939 2940
	args.tp = tp;
	args.mp = tp->t_mountp;
2941
	args.agno = agno;
2942 2943 2944 2945 2946 2947

	/*
	 * 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
2948
		return -EFSCORRUPTED;
2949

2950
	args.pag = xfs_perag_get(args.mp, args.agno);
2951 2952 2953 2954
	ASSERT(args.pag);

	error = xfs_alloc_fix_freelist(&args, XFS_ALLOC_FLAG_FREEING);
	if (error)
2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971
		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 */
xfs_free_extent(
	struct xfs_trans	*tp,	/* transaction pointer */
	xfs_fsblock_t		bno,	/* starting block number of extent */
2972
	xfs_extlen_t		len,	/* length of extent */
2973 2974
	struct xfs_owner_info	*oinfo,	/* extent owner */
	enum xfs_ag_resv_type	type)	/* block reservation type */
2975 2976 2977 2978 2979 2980 2981 2982
{
	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;

	ASSERT(len != 0);
2983
	ASSERT(type != XFS_AG_RESV_AGFL);
2984

2985
	if (XFS_TEST_ERROR(false, mp,
2986
			XFS_ERRTAG_FREE_EXTENT))
2987 2988
		return -EIO;

2989 2990 2991 2992 2993
	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);
2994 2995

	/* validate the extent size is legal now we have the agf locked */
2996 2997 2998
	XFS_WANT_CORRUPTED_GOTO(mp,
		agbno + len <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_length),
				err);
2999

3000
	error = xfs_free_ag_extent(tp, agbp, agno, agbno, len, oinfo, type);
3001 3002 3003 3004 3005 3006 3007 3008
	if (error)
		goto err;

	xfs_extent_busy_insert(tp, agno, agbno, len, 0);
	return 0;

err:
	xfs_trans_brelse(tp, agbp);
L
Linus Torvalds 已提交
3009 3010
	return error;
}
3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052

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);
}
3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067

/* 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);
}
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 3110 3111 3112 3113 3114 3115 3116

/* Find the size of the AG, in blocks. */
xfs_agblock_t
xfs_ag_block_count(
	struct xfs_mount	*mp,
	xfs_agnumber_t		agno)
{
	ASSERT(agno < mp->m_sb.sb_agcount);

	if (agno < mp->m_sb.sb_agcount - 1)
		return mp->m_sb.sb_agblocks;
	return mp->m_sb.sb_dblocks - (agno * mp->m_sb.sb_agblocks);
}

/*
 * Verify that an AG block number pointer neither points outside the AG
 * nor points at static metadata.
 */
bool
xfs_verify_agbno(
	struct xfs_mount	*mp,
	xfs_agnumber_t		agno,
	xfs_agblock_t		agbno)
{
	xfs_agblock_t		eoag;

	eoag = xfs_ag_block_count(mp, agno);
	if (agbno >= eoag)
		return false;
	if (agbno <= XFS_AGFL_BLOCK(mp))
		return false;
	return true;
}

/*
 * Verify that an FS block number pointer neither points outside the
 * filesystem nor points at static AG metadata.
 */
bool
xfs_verify_fsbno(
	struct xfs_mount	*mp,
	xfs_fsblock_t		fsbno)
{
	xfs_agnumber_t		agno = XFS_FSB_TO_AGNO(mp, fsbno);

	if (agno >= mp->m_sb.sb_agcount)
		return false;
	return xfs_verify_agbno(mp, agno, XFS_FSB_TO_AGBNO(mp, fsbno));
}
3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135

/* 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);
}