xref: /openbmc/linux/fs/xfs/libxfs/xfs_alloc.c (revision be471fe3)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #include "xfs.h"
7 #include "xfs_fs.h"
8 #include "xfs_format.h"
9 #include "xfs_log_format.h"
10 #include "xfs_shared.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_bit.h"
13 #include "xfs_sb.h"
14 #include "xfs_mount.h"
15 #include "xfs_defer.h"
16 #include "xfs_btree.h"
17 #include "xfs_rmap.h"
18 #include "xfs_alloc_btree.h"
19 #include "xfs_alloc.h"
20 #include "xfs_extent_busy.h"
21 #include "xfs_errortag.h"
22 #include "xfs_error.h"
23 #include "xfs_trace.h"
24 #include "xfs_trans.h"
25 #include "xfs_buf_item.h"
26 #include "xfs_log.h"
27 #include "xfs_ag_resv.h"
28 #include "xfs_bmap.h"
29 
30 extern kmem_zone_t	*xfs_bmap_free_item_zone;
31 
32 struct workqueue_struct *xfs_alloc_wq;
33 
34 #define XFS_ABSDIFF(a,b)	(((a) <= (b)) ? ((b) - (a)) : ((a) - (b)))
35 
36 #define	XFSA_FIXUP_BNO_OK	1
37 #define	XFSA_FIXUP_CNT_OK	2
38 
39 STATIC int xfs_alloc_ag_vextent_exact(xfs_alloc_arg_t *);
40 STATIC int xfs_alloc_ag_vextent_near(xfs_alloc_arg_t *);
41 STATIC int xfs_alloc_ag_vextent_size(xfs_alloc_arg_t *);
42 
43 /*
44  * Size of the AGFL.  For CRC-enabled filesystes we steal a couple of slots in
45  * the beginning of the block for a proper header with the location information
46  * and CRC.
47  */
48 unsigned int
49 xfs_agfl_size(
50 	struct xfs_mount	*mp)
51 {
52 	unsigned int		size = mp->m_sb.sb_sectsize;
53 
54 	if (xfs_sb_version_hascrc(&mp->m_sb))
55 		size -= sizeof(struct xfs_agfl);
56 
57 	return size / sizeof(xfs_agblock_t);
58 }
59 
60 unsigned int
61 xfs_refc_block(
62 	struct xfs_mount	*mp)
63 {
64 	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
65 		return XFS_RMAP_BLOCK(mp) + 1;
66 	if (xfs_sb_version_hasfinobt(&mp->m_sb))
67 		return XFS_FIBT_BLOCK(mp) + 1;
68 	return XFS_IBT_BLOCK(mp) + 1;
69 }
70 
71 xfs_extlen_t
72 xfs_prealloc_blocks(
73 	struct xfs_mount	*mp)
74 {
75 	if (xfs_sb_version_hasreflink(&mp->m_sb))
76 		return xfs_refc_block(mp) + 1;
77 	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
78 		return XFS_RMAP_BLOCK(mp) + 1;
79 	if (xfs_sb_version_hasfinobt(&mp->m_sb))
80 		return XFS_FIBT_BLOCK(mp) + 1;
81 	return XFS_IBT_BLOCK(mp) + 1;
82 }
83 
84 /*
85  * In order to avoid ENOSPC-related deadlock caused by out-of-order locking of
86  * AGF buffer (PV 947395), we place constraints on the relationship among
87  * actual allocations for data blocks, freelist blocks, and potential file data
88  * bmap btree blocks. However, these restrictions may result in no actual space
89  * allocated for a delayed extent, for example, a data block in a certain AG is
90  * allocated but there is no additional block for the additional bmap btree
91  * block due to a split of the bmap btree of the file. The result of this may
92  * lead to an infinite loop when the file gets flushed to disk and all delayed
93  * extents need to be actually allocated. To get around this, we explicitly set
94  * aside a few blocks which will not be reserved in delayed allocation.
95  *
96  * We need to reserve 4 fsbs _per AG_ for the freelist and 4 more to handle a
97  * potential split of the file's bmap btree.
98  */
99 unsigned int
100 xfs_alloc_set_aside(
101 	struct xfs_mount	*mp)
102 {
103 	return mp->m_sb.sb_agcount * (XFS_ALLOC_AGFL_RESERVE + 4);
104 }
105 
106 /*
107  * When deciding how much space to allocate out of an AG, we limit the
108  * allocation maximum size to the size the AG. However, we cannot use all the
109  * blocks in the AG - some are permanently used by metadata. These
110  * blocks are generally:
111  *	- the AG superblock, AGF, AGI and AGFL
112  *	- the AGF (bno and cnt) and AGI btree root blocks, and optionally
113  *	  the AGI free inode and rmap btree root blocks.
114  *	- blocks on the AGFL according to xfs_alloc_set_aside() limits
115  *	- the rmapbt root block
116  *
117  * The AG headers are sector sized, so the amount of space they take up is
118  * dependent on filesystem geometry. The others are all single blocks.
119  */
120 unsigned int
121 xfs_alloc_ag_max_usable(
122 	struct xfs_mount	*mp)
123 {
124 	unsigned int		blocks;
125 
126 	blocks = XFS_BB_TO_FSB(mp, XFS_FSS_TO_BB(mp, 4)); /* ag headers */
127 	blocks += XFS_ALLOC_AGFL_RESERVE;
128 	blocks += 3;			/* AGF, AGI btree root blocks */
129 	if (xfs_sb_version_hasfinobt(&mp->m_sb))
130 		blocks++;		/* finobt root block */
131 	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
132 		blocks++; 		/* rmap root block */
133 	if (xfs_sb_version_hasreflink(&mp->m_sb))
134 		blocks++;		/* refcount root block */
135 
136 	return mp->m_sb.sb_agblocks - blocks;
137 }
138 
139 /*
140  * Lookup the record equal to [bno, len] in the btree given by cur.
141  */
142 STATIC int				/* error */
143 xfs_alloc_lookup_eq(
144 	struct xfs_btree_cur	*cur,	/* btree cursor */
145 	xfs_agblock_t		bno,	/* starting block of extent */
146 	xfs_extlen_t		len,	/* length of extent */
147 	int			*stat)	/* success/failure */
148 {
149 	int			error;
150 
151 	cur->bc_rec.a.ar_startblock = bno;
152 	cur->bc_rec.a.ar_blockcount = len;
153 	error = xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
154 	cur->bc_ag.abt.active = (*stat == 1);
155 	return error;
156 }
157 
158 /*
159  * Lookup the first record greater than or equal to [bno, len]
160  * in the btree given by cur.
161  */
162 int				/* error */
163 xfs_alloc_lookup_ge(
164 	struct xfs_btree_cur	*cur,	/* btree cursor */
165 	xfs_agblock_t		bno,	/* starting block of extent */
166 	xfs_extlen_t		len,	/* length of extent */
167 	int			*stat)	/* success/failure */
168 {
169 	int			error;
170 
171 	cur->bc_rec.a.ar_startblock = bno;
172 	cur->bc_rec.a.ar_blockcount = len;
173 	error = xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
174 	cur->bc_ag.abt.active = (*stat == 1);
175 	return error;
176 }
177 
178 /*
179  * Lookup the first record less than or equal to [bno, len]
180  * in the btree given by cur.
181  */
182 int					/* error */
183 xfs_alloc_lookup_le(
184 	struct xfs_btree_cur	*cur,	/* btree cursor */
185 	xfs_agblock_t		bno,	/* starting block of extent */
186 	xfs_extlen_t		len,	/* length of extent */
187 	int			*stat)	/* success/failure */
188 {
189 	int			error;
190 	cur->bc_rec.a.ar_startblock = bno;
191 	cur->bc_rec.a.ar_blockcount = len;
192 	error = xfs_btree_lookup(cur, XFS_LOOKUP_LE, stat);
193 	cur->bc_ag.abt.active = (*stat == 1);
194 	return error;
195 }
196 
197 static inline bool
198 xfs_alloc_cur_active(
199 	struct xfs_btree_cur	*cur)
200 {
201 	return cur && cur->bc_ag.abt.active;
202 }
203 
204 /*
205  * Update the record referred to by cur to the value given
206  * by [bno, len].
207  * This either works (return 0) or gets an EFSCORRUPTED error.
208  */
209 STATIC int				/* error */
210 xfs_alloc_update(
211 	struct xfs_btree_cur	*cur,	/* btree cursor */
212 	xfs_agblock_t		bno,	/* starting block of extent */
213 	xfs_extlen_t		len)	/* length of extent */
214 {
215 	union xfs_btree_rec	rec;
216 
217 	rec.alloc.ar_startblock = cpu_to_be32(bno);
218 	rec.alloc.ar_blockcount = cpu_to_be32(len);
219 	return xfs_btree_update(cur, &rec);
220 }
221 
222 /*
223  * Get the data from the pointed-to record.
224  */
225 int					/* error */
226 xfs_alloc_get_rec(
227 	struct xfs_btree_cur	*cur,	/* btree cursor */
228 	xfs_agblock_t		*bno,	/* output: starting block of extent */
229 	xfs_extlen_t		*len,	/* output: length of extent */
230 	int			*stat)	/* output: success/failure */
231 {
232 	struct xfs_mount	*mp = cur->bc_mp;
233 	xfs_agnumber_t		agno = cur->bc_ag.agno;
234 	union xfs_btree_rec	*rec;
235 	int			error;
236 
237 	error = xfs_btree_get_rec(cur, &rec, stat);
238 	if (error || !(*stat))
239 		return error;
240 
241 	*bno = be32_to_cpu(rec->alloc.ar_startblock);
242 	*len = be32_to_cpu(rec->alloc.ar_blockcount);
243 
244 	if (*len == 0)
245 		goto out_bad_rec;
246 
247 	/* check for valid extent range, including overflow */
248 	if (!xfs_verify_agbno(mp, agno, *bno))
249 		goto out_bad_rec;
250 	if (*bno > *bno + *len)
251 		goto out_bad_rec;
252 	if (!xfs_verify_agbno(mp, agno, *bno + *len - 1))
253 		goto out_bad_rec;
254 
255 	return 0;
256 
257 out_bad_rec:
258 	xfs_warn(mp,
259 		"%s Freespace BTree record corruption in AG %d detected!",
260 		cur->bc_btnum == XFS_BTNUM_BNO ? "Block" : "Size", agno);
261 	xfs_warn(mp,
262 		"start block 0x%x block count 0x%x", *bno, *len);
263 	return -EFSCORRUPTED;
264 }
265 
266 /*
267  * Compute aligned version of the found extent.
268  * Takes alignment and min length into account.
269  */
270 STATIC bool
271 xfs_alloc_compute_aligned(
272 	xfs_alloc_arg_t	*args,		/* allocation argument structure */
273 	xfs_agblock_t	foundbno,	/* starting block in found extent */
274 	xfs_extlen_t	foundlen,	/* length in found extent */
275 	xfs_agblock_t	*resbno,	/* result block number */
276 	xfs_extlen_t	*reslen,	/* result length */
277 	unsigned	*busy_gen)
278 {
279 	xfs_agblock_t	bno = foundbno;
280 	xfs_extlen_t	len = foundlen;
281 	xfs_extlen_t	diff;
282 	bool		busy;
283 
284 	/* Trim busy sections out of found extent */
285 	busy = xfs_extent_busy_trim(args, &bno, &len, busy_gen);
286 
287 	/*
288 	 * If we have a largish extent that happens to start before min_agbno,
289 	 * see if we can shift it into range...
290 	 */
291 	if (bno < args->min_agbno && bno + len > args->min_agbno) {
292 		diff = args->min_agbno - bno;
293 		if (len > diff) {
294 			bno += diff;
295 			len -= diff;
296 		}
297 	}
298 
299 	if (args->alignment > 1 && len >= args->minlen) {
300 		xfs_agblock_t	aligned_bno = roundup(bno, args->alignment);
301 
302 		diff = aligned_bno - bno;
303 
304 		*resbno = aligned_bno;
305 		*reslen = diff >= len ? 0 : len - diff;
306 	} else {
307 		*resbno = bno;
308 		*reslen = len;
309 	}
310 
311 	return busy;
312 }
313 
314 /*
315  * Compute best start block and diff for "near" allocations.
316  * freelen >= wantlen already checked by caller.
317  */
318 STATIC xfs_extlen_t			/* difference value (absolute) */
319 xfs_alloc_compute_diff(
320 	xfs_agblock_t	wantbno,	/* target starting block */
321 	xfs_extlen_t	wantlen,	/* target length */
322 	xfs_extlen_t	alignment,	/* target alignment */
323 	int		datatype,	/* are we allocating data? */
324 	xfs_agblock_t	freebno,	/* freespace's starting block */
325 	xfs_extlen_t	freelen,	/* freespace's length */
326 	xfs_agblock_t	*newbnop)	/* result: best start block from free */
327 {
328 	xfs_agblock_t	freeend;	/* end of freespace extent */
329 	xfs_agblock_t	newbno1;	/* return block number */
330 	xfs_agblock_t	newbno2;	/* other new block number */
331 	xfs_extlen_t	newlen1=0;	/* length with newbno1 */
332 	xfs_extlen_t	newlen2=0;	/* length with newbno2 */
333 	xfs_agblock_t	wantend;	/* end of target extent */
334 	bool		userdata = datatype & XFS_ALLOC_USERDATA;
335 
336 	ASSERT(freelen >= wantlen);
337 	freeend = freebno + freelen;
338 	wantend = wantbno + wantlen;
339 	/*
340 	 * We want to allocate from the start of a free extent if it is past
341 	 * the desired block or if we are allocating user data and the free
342 	 * extent is before desired block. The second case is there to allow
343 	 * for contiguous allocation from the remaining free space if the file
344 	 * grows in the short term.
345 	 */
346 	if (freebno >= wantbno || (userdata && freeend < wantend)) {
347 		if ((newbno1 = roundup(freebno, alignment)) >= freeend)
348 			newbno1 = NULLAGBLOCK;
349 	} else if (freeend >= wantend && alignment > 1) {
350 		newbno1 = roundup(wantbno, alignment);
351 		newbno2 = newbno1 - alignment;
352 		if (newbno1 >= freeend)
353 			newbno1 = NULLAGBLOCK;
354 		else
355 			newlen1 = XFS_EXTLEN_MIN(wantlen, freeend - newbno1);
356 		if (newbno2 < freebno)
357 			newbno2 = NULLAGBLOCK;
358 		else
359 			newlen2 = XFS_EXTLEN_MIN(wantlen, freeend - newbno2);
360 		if (newbno1 != NULLAGBLOCK && newbno2 != NULLAGBLOCK) {
361 			if (newlen1 < newlen2 ||
362 			    (newlen1 == newlen2 &&
363 			     XFS_ABSDIFF(newbno1, wantbno) >
364 			     XFS_ABSDIFF(newbno2, wantbno)))
365 				newbno1 = newbno2;
366 		} else if (newbno2 != NULLAGBLOCK)
367 			newbno1 = newbno2;
368 	} else if (freeend >= wantend) {
369 		newbno1 = wantbno;
370 	} else if (alignment > 1) {
371 		newbno1 = roundup(freeend - wantlen, alignment);
372 		if (newbno1 > freeend - wantlen &&
373 		    newbno1 - alignment >= freebno)
374 			newbno1 -= alignment;
375 		else if (newbno1 >= freeend)
376 			newbno1 = NULLAGBLOCK;
377 	} else
378 		newbno1 = freeend - wantlen;
379 	*newbnop = newbno1;
380 	return newbno1 == NULLAGBLOCK ? 0 : XFS_ABSDIFF(newbno1, wantbno);
381 }
382 
383 /*
384  * Fix up the length, based on mod and prod.
385  * len should be k * prod + mod for some k.
386  * If len is too small it is returned unchanged.
387  * If len hits maxlen it is left alone.
388  */
389 STATIC void
390 xfs_alloc_fix_len(
391 	xfs_alloc_arg_t	*args)		/* allocation argument structure */
392 {
393 	xfs_extlen_t	k;
394 	xfs_extlen_t	rlen;
395 
396 	ASSERT(args->mod < args->prod);
397 	rlen = args->len;
398 	ASSERT(rlen >= args->minlen);
399 	ASSERT(rlen <= args->maxlen);
400 	if (args->prod <= 1 || rlen < args->mod || rlen == args->maxlen ||
401 	    (args->mod == 0 && rlen < args->prod))
402 		return;
403 	k = rlen % args->prod;
404 	if (k == args->mod)
405 		return;
406 	if (k > args->mod)
407 		rlen = rlen - (k - args->mod);
408 	else
409 		rlen = rlen - args->prod + (args->mod - k);
410 	/* casts to (int) catch length underflows */
411 	if ((int)rlen < (int)args->minlen)
412 		return;
413 	ASSERT(rlen >= args->minlen && rlen <= args->maxlen);
414 	ASSERT(rlen % args->prod == args->mod);
415 	ASSERT(args->pag->pagf_freeblks + args->pag->pagf_flcount >=
416 		rlen + args->minleft);
417 	args->len = rlen;
418 }
419 
420 /*
421  * Update the two btrees, logically removing from freespace the extent
422  * starting at rbno, rlen blocks.  The extent is contained within the
423  * actual (current) free extent fbno for flen blocks.
424  * Flags are passed in indicating whether the cursors are set to the
425  * relevant records.
426  */
427 STATIC int				/* error code */
428 xfs_alloc_fixup_trees(
429 	xfs_btree_cur_t	*cnt_cur,	/* cursor for by-size btree */
430 	xfs_btree_cur_t	*bno_cur,	/* cursor for by-block btree */
431 	xfs_agblock_t	fbno,		/* starting block of free extent */
432 	xfs_extlen_t	flen,		/* length of free extent */
433 	xfs_agblock_t	rbno,		/* starting block of returned extent */
434 	xfs_extlen_t	rlen,		/* length of returned extent */
435 	int		flags)		/* flags, XFSA_FIXUP_... */
436 {
437 	int		error;		/* error code */
438 	int		i;		/* operation results */
439 	xfs_agblock_t	nfbno1;		/* first new free startblock */
440 	xfs_agblock_t	nfbno2;		/* second new free startblock */
441 	xfs_extlen_t	nflen1=0;	/* first new free length */
442 	xfs_extlen_t	nflen2=0;	/* second new free length */
443 	struct xfs_mount *mp;
444 
445 	mp = cnt_cur->bc_mp;
446 
447 	/*
448 	 * Look up the record in the by-size tree if necessary.
449 	 */
450 	if (flags & XFSA_FIXUP_CNT_OK) {
451 #ifdef DEBUG
452 		if ((error = xfs_alloc_get_rec(cnt_cur, &nfbno1, &nflen1, &i)))
453 			return error;
454 		if (XFS_IS_CORRUPT(mp,
455 				   i != 1 ||
456 				   nfbno1 != fbno ||
457 				   nflen1 != flen))
458 			return -EFSCORRUPTED;
459 #endif
460 	} else {
461 		if ((error = xfs_alloc_lookup_eq(cnt_cur, fbno, flen, &i)))
462 			return error;
463 		if (XFS_IS_CORRUPT(mp, i != 1))
464 			return -EFSCORRUPTED;
465 	}
466 	/*
467 	 * Look up the record in the by-block tree if necessary.
468 	 */
469 	if (flags & XFSA_FIXUP_BNO_OK) {
470 #ifdef DEBUG
471 		if ((error = xfs_alloc_get_rec(bno_cur, &nfbno1, &nflen1, &i)))
472 			return error;
473 		if (XFS_IS_CORRUPT(mp,
474 				   i != 1 ||
475 				   nfbno1 != fbno ||
476 				   nflen1 != flen))
477 			return -EFSCORRUPTED;
478 #endif
479 	} else {
480 		if ((error = xfs_alloc_lookup_eq(bno_cur, fbno, flen, &i)))
481 			return error;
482 		if (XFS_IS_CORRUPT(mp, i != 1))
483 			return -EFSCORRUPTED;
484 	}
485 
486 #ifdef DEBUG
487 	if (bno_cur->bc_nlevels == 1 && cnt_cur->bc_nlevels == 1) {
488 		struct xfs_btree_block	*bnoblock;
489 		struct xfs_btree_block	*cntblock;
490 
491 		bnoblock = XFS_BUF_TO_BLOCK(bno_cur->bc_bufs[0]);
492 		cntblock = XFS_BUF_TO_BLOCK(cnt_cur->bc_bufs[0]);
493 
494 		if (XFS_IS_CORRUPT(mp,
495 				   bnoblock->bb_numrecs !=
496 				   cntblock->bb_numrecs))
497 			return -EFSCORRUPTED;
498 	}
499 #endif
500 
501 	/*
502 	 * Deal with all four cases: the allocated record is contained
503 	 * within the freespace record, so we can have new freespace
504 	 * at either (or both) end, or no freespace remaining.
505 	 */
506 	if (rbno == fbno && rlen == flen)
507 		nfbno1 = nfbno2 = NULLAGBLOCK;
508 	else if (rbno == fbno) {
509 		nfbno1 = rbno + rlen;
510 		nflen1 = flen - rlen;
511 		nfbno2 = NULLAGBLOCK;
512 	} else if (rbno + rlen == fbno + flen) {
513 		nfbno1 = fbno;
514 		nflen1 = flen - rlen;
515 		nfbno2 = NULLAGBLOCK;
516 	} else {
517 		nfbno1 = fbno;
518 		nflen1 = rbno - fbno;
519 		nfbno2 = rbno + rlen;
520 		nflen2 = (fbno + flen) - nfbno2;
521 	}
522 	/*
523 	 * Delete the entry from the by-size btree.
524 	 */
525 	if ((error = xfs_btree_delete(cnt_cur, &i)))
526 		return error;
527 	if (XFS_IS_CORRUPT(mp, i != 1))
528 		return -EFSCORRUPTED;
529 	/*
530 	 * Add new by-size btree entry(s).
531 	 */
532 	if (nfbno1 != NULLAGBLOCK) {
533 		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno1, nflen1, &i)))
534 			return error;
535 		if (XFS_IS_CORRUPT(mp, i != 0))
536 			return -EFSCORRUPTED;
537 		if ((error = xfs_btree_insert(cnt_cur, &i)))
538 			return error;
539 		if (XFS_IS_CORRUPT(mp, i != 1))
540 			return -EFSCORRUPTED;
541 	}
542 	if (nfbno2 != NULLAGBLOCK) {
543 		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno2, nflen2, &i)))
544 			return error;
545 		if (XFS_IS_CORRUPT(mp, i != 0))
546 			return -EFSCORRUPTED;
547 		if ((error = xfs_btree_insert(cnt_cur, &i)))
548 			return error;
549 		if (XFS_IS_CORRUPT(mp, i != 1))
550 			return -EFSCORRUPTED;
551 	}
552 	/*
553 	 * Fix up the by-block btree entry(s).
554 	 */
555 	if (nfbno1 == NULLAGBLOCK) {
556 		/*
557 		 * No remaining freespace, just delete the by-block tree entry.
558 		 */
559 		if ((error = xfs_btree_delete(bno_cur, &i)))
560 			return error;
561 		if (XFS_IS_CORRUPT(mp, i != 1))
562 			return -EFSCORRUPTED;
563 	} else {
564 		/*
565 		 * Update the by-block entry to start later|be shorter.
566 		 */
567 		if ((error = xfs_alloc_update(bno_cur, nfbno1, nflen1)))
568 			return error;
569 	}
570 	if (nfbno2 != NULLAGBLOCK) {
571 		/*
572 		 * 2 resulting free entries, need to add one.
573 		 */
574 		if ((error = xfs_alloc_lookup_eq(bno_cur, nfbno2, nflen2, &i)))
575 			return error;
576 		if (XFS_IS_CORRUPT(mp, i != 0))
577 			return -EFSCORRUPTED;
578 		if ((error = xfs_btree_insert(bno_cur, &i)))
579 			return error;
580 		if (XFS_IS_CORRUPT(mp, i != 1))
581 			return -EFSCORRUPTED;
582 	}
583 	return 0;
584 }
585 
586 static xfs_failaddr_t
587 xfs_agfl_verify(
588 	struct xfs_buf	*bp)
589 {
590 	struct xfs_mount *mp = bp->b_mount;
591 	struct xfs_agfl	*agfl = XFS_BUF_TO_AGFL(bp);
592 	__be32		*agfl_bno = xfs_buf_to_agfl_bno(bp);
593 	int		i;
594 
595 	/*
596 	 * There is no verification of non-crc AGFLs because mkfs does not
597 	 * initialise the AGFL to zero or NULL. Hence the only valid part of the
598 	 * AGFL is what the AGF says is active. We can't get to the AGF, so we
599 	 * can't verify just those entries are valid.
600 	 */
601 	if (!xfs_sb_version_hascrc(&mp->m_sb))
602 		return NULL;
603 
604 	if (!xfs_verify_magic(bp, agfl->agfl_magicnum))
605 		return __this_address;
606 	if (!uuid_equal(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid))
607 		return __this_address;
608 	/*
609 	 * during growfs operations, the perag is not fully initialised,
610 	 * so we can't use it for any useful checking. growfs ensures we can't
611 	 * use it by using uncached buffers that don't have the perag attached
612 	 * so we can detect and avoid this problem.
613 	 */
614 	if (bp->b_pag && be32_to_cpu(agfl->agfl_seqno) != bp->b_pag->pag_agno)
615 		return __this_address;
616 
617 	for (i = 0; i < xfs_agfl_size(mp); i++) {
618 		if (be32_to_cpu(agfl_bno[i]) != NULLAGBLOCK &&
619 		    be32_to_cpu(agfl_bno[i]) >= mp->m_sb.sb_agblocks)
620 			return __this_address;
621 	}
622 
623 	if (!xfs_log_check_lsn(mp, be64_to_cpu(XFS_BUF_TO_AGFL(bp)->agfl_lsn)))
624 		return __this_address;
625 	return NULL;
626 }
627 
628 static void
629 xfs_agfl_read_verify(
630 	struct xfs_buf	*bp)
631 {
632 	struct xfs_mount *mp = bp->b_mount;
633 	xfs_failaddr_t	fa;
634 
635 	/*
636 	 * There is no verification of non-crc AGFLs because mkfs does not
637 	 * initialise the AGFL to zero or NULL. Hence the only valid part of the
638 	 * AGFL is what the AGF says is active. We can't get to the AGF, so we
639 	 * can't verify just those entries are valid.
640 	 */
641 	if (!xfs_sb_version_hascrc(&mp->m_sb))
642 		return;
643 
644 	if (!xfs_buf_verify_cksum(bp, XFS_AGFL_CRC_OFF))
645 		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
646 	else {
647 		fa = xfs_agfl_verify(bp);
648 		if (fa)
649 			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
650 	}
651 }
652 
653 static void
654 xfs_agfl_write_verify(
655 	struct xfs_buf	*bp)
656 {
657 	struct xfs_mount	*mp = bp->b_mount;
658 	struct xfs_buf_log_item	*bip = bp->b_log_item;
659 	xfs_failaddr_t		fa;
660 
661 	/* no verification of non-crc AGFLs */
662 	if (!xfs_sb_version_hascrc(&mp->m_sb))
663 		return;
664 
665 	fa = xfs_agfl_verify(bp);
666 	if (fa) {
667 		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
668 		return;
669 	}
670 
671 	if (bip)
672 		XFS_BUF_TO_AGFL(bp)->agfl_lsn = cpu_to_be64(bip->bli_item.li_lsn);
673 
674 	xfs_buf_update_cksum(bp, XFS_AGFL_CRC_OFF);
675 }
676 
677 const struct xfs_buf_ops xfs_agfl_buf_ops = {
678 	.name = "xfs_agfl",
679 	.magic = { cpu_to_be32(XFS_AGFL_MAGIC), cpu_to_be32(XFS_AGFL_MAGIC) },
680 	.verify_read = xfs_agfl_read_verify,
681 	.verify_write = xfs_agfl_write_verify,
682 	.verify_struct = xfs_agfl_verify,
683 };
684 
685 /*
686  * Read in the allocation group free block array.
687  */
688 int					/* error */
689 xfs_alloc_read_agfl(
690 	xfs_mount_t	*mp,		/* mount point structure */
691 	xfs_trans_t	*tp,		/* transaction pointer */
692 	xfs_agnumber_t	agno,		/* allocation group number */
693 	struct xfs_buf	**bpp)		/* buffer for the ag free block array */
694 {
695 	struct xfs_buf	*bp;		/* return value */
696 	int		error;
697 
698 	ASSERT(agno != NULLAGNUMBER);
699 	error = xfs_trans_read_buf(
700 			mp, tp, mp->m_ddev_targp,
701 			XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
702 			XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_agfl_buf_ops);
703 	if (error)
704 		return error;
705 	xfs_buf_set_ref(bp, XFS_AGFL_REF);
706 	*bpp = bp;
707 	return 0;
708 }
709 
710 STATIC int
711 xfs_alloc_update_counters(
712 	struct xfs_trans	*tp,
713 	struct xfs_buf		*agbp,
714 	long			len)
715 {
716 	struct xfs_agf		*agf = agbp->b_addr;
717 
718 	agbp->b_pag->pagf_freeblks += len;
719 	be32_add_cpu(&agf->agf_freeblks, len);
720 
721 	if (unlikely(be32_to_cpu(agf->agf_freeblks) >
722 		     be32_to_cpu(agf->agf_length))) {
723 		xfs_buf_mark_corrupt(agbp);
724 		return -EFSCORRUPTED;
725 	}
726 
727 	xfs_alloc_log_agf(tp, agbp, XFS_AGF_FREEBLKS);
728 	return 0;
729 }
730 
731 /*
732  * Block allocation algorithm and data structures.
733  */
734 struct xfs_alloc_cur {
735 	struct xfs_btree_cur		*cnt;	/* btree cursors */
736 	struct xfs_btree_cur		*bnolt;
737 	struct xfs_btree_cur		*bnogt;
738 	xfs_extlen_t			cur_len;/* current search length */
739 	xfs_agblock_t			rec_bno;/* extent startblock */
740 	xfs_extlen_t			rec_len;/* extent length */
741 	xfs_agblock_t			bno;	/* alloc bno */
742 	xfs_extlen_t			len;	/* alloc len */
743 	xfs_extlen_t			diff;	/* diff from search bno */
744 	unsigned int			busy_gen;/* busy state */
745 	bool				busy;
746 };
747 
748 /*
749  * Set up cursors, etc. in the extent allocation cursor. This function can be
750  * called multiple times to reset an initialized structure without having to
751  * reallocate cursors.
752  */
753 static int
754 xfs_alloc_cur_setup(
755 	struct xfs_alloc_arg	*args,
756 	struct xfs_alloc_cur	*acur)
757 {
758 	int			error;
759 	int			i;
760 
761 	ASSERT(args->alignment == 1 || args->type != XFS_ALLOCTYPE_THIS_BNO);
762 
763 	acur->cur_len = args->maxlen;
764 	acur->rec_bno = 0;
765 	acur->rec_len = 0;
766 	acur->bno = 0;
767 	acur->len = 0;
768 	acur->diff = -1;
769 	acur->busy = false;
770 	acur->busy_gen = 0;
771 
772 	/*
773 	 * Perform an initial cntbt lookup to check for availability of maxlen
774 	 * extents. If this fails, we'll return -ENOSPC to signal the caller to
775 	 * attempt a small allocation.
776 	 */
777 	if (!acur->cnt)
778 		acur->cnt = xfs_allocbt_init_cursor(args->mp, args->tp,
779 					args->agbp, args->agno, XFS_BTNUM_CNT);
780 	error = xfs_alloc_lookup_ge(acur->cnt, 0, args->maxlen, &i);
781 	if (error)
782 		return error;
783 
784 	/*
785 	 * Allocate the bnobt left and right search cursors.
786 	 */
787 	if (!acur->bnolt)
788 		acur->bnolt = xfs_allocbt_init_cursor(args->mp, args->tp,
789 					args->agbp, args->agno, XFS_BTNUM_BNO);
790 	if (!acur->bnogt)
791 		acur->bnogt = xfs_allocbt_init_cursor(args->mp, args->tp,
792 					args->agbp, args->agno, XFS_BTNUM_BNO);
793 	return i == 1 ? 0 : -ENOSPC;
794 }
795 
796 static void
797 xfs_alloc_cur_close(
798 	struct xfs_alloc_cur	*acur,
799 	bool			error)
800 {
801 	int			cur_error = XFS_BTREE_NOERROR;
802 
803 	if (error)
804 		cur_error = XFS_BTREE_ERROR;
805 
806 	if (acur->cnt)
807 		xfs_btree_del_cursor(acur->cnt, cur_error);
808 	if (acur->bnolt)
809 		xfs_btree_del_cursor(acur->bnolt, cur_error);
810 	if (acur->bnogt)
811 		xfs_btree_del_cursor(acur->bnogt, cur_error);
812 	acur->cnt = acur->bnolt = acur->bnogt = NULL;
813 }
814 
815 /*
816  * Check an extent for allocation and track the best available candidate in the
817  * allocation structure. The cursor is deactivated if it has entered an out of
818  * range state based on allocation arguments. Optionally return the extent
819  * extent geometry and allocation status if requested by the caller.
820  */
821 static int
822 xfs_alloc_cur_check(
823 	struct xfs_alloc_arg	*args,
824 	struct xfs_alloc_cur	*acur,
825 	struct xfs_btree_cur	*cur,
826 	int			*new)
827 {
828 	int			error, i;
829 	xfs_agblock_t		bno, bnoa, bnew;
830 	xfs_extlen_t		len, lena, diff = -1;
831 	bool			busy;
832 	unsigned		busy_gen = 0;
833 	bool			deactivate = false;
834 	bool			isbnobt = cur->bc_btnum == XFS_BTNUM_BNO;
835 
836 	*new = 0;
837 
838 	error = xfs_alloc_get_rec(cur, &bno, &len, &i);
839 	if (error)
840 		return error;
841 	if (XFS_IS_CORRUPT(args->mp, i != 1))
842 		return -EFSCORRUPTED;
843 
844 	/*
845 	 * Check minlen and deactivate a cntbt cursor if out of acceptable size
846 	 * range (i.e., walking backwards looking for a minlen extent).
847 	 */
848 	if (len < args->minlen) {
849 		deactivate = !isbnobt;
850 		goto out;
851 	}
852 
853 	busy = xfs_alloc_compute_aligned(args, bno, len, &bnoa, &lena,
854 					 &busy_gen);
855 	acur->busy |= busy;
856 	if (busy)
857 		acur->busy_gen = busy_gen;
858 	/* deactivate a bnobt cursor outside of locality range */
859 	if (bnoa < args->min_agbno || bnoa > args->max_agbno) {
860 		deactivate = isbnobt;
861 		goto out;
862 	}
863 	if (lena < args->minlen)
864 		goto out;
865 
866 	args->len = XFS_EXTLEN_MIN(lena, args->maxlen);
867 	xfs_alloc_fix_len(args);
868 	ASSERT(args->len >= args->minlen);
869 	if (args->len < acur->len)
870 		goto out;
871 
872 	/*
873 	 * We have an aligned record that satisfies minlen and beats or matches
874 	 * the candidate extent size. Compare locality for near allocation mode.
875 	 */
876 	ASSERT(args->type == XFS_ALLOCTYPE_NEAR_BNO);
877 	diff = xfs_alloc_compute_diff(args->agbno, args->len,
878 				      args->alignment, args->datatype,
879 				      bnoa, lena, &bnew);
880 	if (bnew == NULLAGBLOCK)
881 		goto out;
882 
883 	/*
884 	 * Deactivate a bnobt cursor with worse locality than the current best.
885 	 */
886 	if (diff > acur->diff) {
887 		deactivate = isbnobt;
888 		goto out;
889 	}
890 
891 	ASSERT(args->len > acur->len ||
892 	       (args->len == acur->len && diff <= acur->diff));
893 	acur->rec_bno = bno;
894 	acur->rec_len = len;
895 	acur->bno = bnew;
896 	acur->len = args->len;
897 	acur->diff = diff;
898 	*new = 1;
899 
900 	/*
901 	 * We're done if we found a perfect allocation. This only deactivates
902 	 * the current cursor, but this is just an optimization to terminate a
903 	 * cntbt search that otherwise runs to the edge of the tree.
904 	 */
905 	if (acur->diff == 0 && acur->len == args->maxlen)
906 		deactivate = true;
907 out:
908 	if (deactivate)
909 		cur->bc_ag.abt.active = false;
910 	trace_xfs_alloc_cur_check(args->mp, cur->bc_btnum, bno, len, diff,
911 				  *new);
912 	return 0;
913 }
914 
915 /*
916  * Complete an allocation of a candidate extent. Remove the extent from both
917  * trees and update the args structure.
918  */
919 STATIC int
920 xfs_alloc_cur_finish(
921 	struct xfs_alloc_arg	*args,
922 	struct xfs_alloc_cur	*acur)
923 {
924 	struct xfs_agf __maybe_unused *agf = args->agbp->b_addr;
925 	int			error;
926 
927 	ASSERT(acur->cnt && acur->bnolt);
928 	ASSERT(acur->bno >= acur->rec_bno);
929 	ASSERT(acur->bno + acur->len <= acur->rec_bno + acur->rec_len);
930 	ASSERT(acur->rec_bno + acur->rec_len <= be32_to_cpu(agf->agf_length));
931 
932 	error = xfs_alloc_fixup_trees(acur->cnt, acur->bnolt, acur->rec_bno,
933 				      acur->rec_len, acur->bno, acur->len, 0);
934 	if (error)
935 		return error;
936 
937 	args->agbno = acur->bno;
938 	args->len = acur->len;
939 	args->wasfromfl = 0;
940 
941 	trace_xfs_alloc_cur(args);
942 	return 0;
943 }
944 
945 /*
946  * Locality allocation lookup algorithm. This expects a cntbt cursor and uses
947  * bno optimized lookup to search for extents with ideal size and locality.
948  */
949 STATIC int
950 xfs_alloc_cntbt_iter(
951 	struct xfs_alloc_arg		*args,
952 	struct xfs_alloc_cur		*acur)
953 {
954 	struct xfs_btree_cur	*cur = acur->cnt;
955 	xfs_agblock_t		bno;
956 	xfs_extlen_t		len, cur_len;
957 	int			error;
958 	int			i;
959 
960 	if (!xfs_alloc_cur_active(cur))
961 		return 0;
962 
963 	/* locality optimized lookup */
964 	cur_len = acur->cur_len;
965 	error = xfs_alloc_lookup_ge(cur, args->agbno, cur_len, &i);
966 	if (error)
967 		return error;
968 	if (i == 0)
969 		return 0;
970 	error = xfs_alloc_get_rec(cur, &bno, &len, &i);
971 	if (error)
972 		return error;
973 
974 	/* check the current record and update search length from it */
975 	error = xfs_alloc_cur_check(args, acur, cur, &i);
976 	if (error)
977 		return error;
978 	ASSERT(len >= acur->cur_len);
979 	acur->cur_len = len;
980 
981 	/*
982 	 * We looked up the first record >= [agbno, len] above. The agbno is a
983 	 * secondary key and so the current record may lie just before or after
984 	 * agbno. If it is past agbno, check the previous record too so long as
985 	 * the length matches as it may be closer. Don't check a smaller record
986 	 * because that could deactivate our cursor.
987 	 */
988 	if (bno > args->agbno) {
989 		error = xfs_btree_decrement(cur, 0, &i);
990 		if (!error && i) {
991 			error = xfs_alloc_get_rec(cur, &bno, &len, &i);
992 			if (!error && i && len == acur->cur_len)
993 				error = xfs_alloc_cur_check(args, acur, cur,
994 							    &i);
995 		}
996 		if (error)
997 			return error;
998 	}
999 
1000 	/*
1001 	 * Increment the search key until we find at least one allocation
1002 	 * candidate or if the extent we found was larger. Otherwise, double the
1003 	 * search key to optimize the search. Efficiency is more important here
1004 	 * than absolute best locality.
1005 	 */
1006 	cur_len <<= 1;
1007 	if (!acur->len || acur->cur_len >= cur_len)
1008 		acur->cur_len++;
1009 	else
1010 		acur->cur_len = cur_len;
1011 
1012 	return error;
1013 }
1014 
1015 /*
1016  * Deal with the case where only small freespaces remain. Either return the
1017  * contents of the last freespace record, or allocate space from the freelist if
1018  * there is nothing in the tree.
1019  */
1020 STATIC int			/* error */
1021 xfs_alloc_ag_vextent_small(
1022 	struct xfs_alloc_arg	*args,	/* allocation argument structure */
1023 	struct xfs_btree_cur	*ccur,	/* optional by-size cursor */
1024 	xfs_agblock_t		*fbnop,	/* result block number */
1025 	xfs_extlen_t		*flenp,	/* result length */
1026 	int			*stat)	/* status: 0-freelist, 1-normal/none */
1027 {
1028 	struct xfs_agf		*agf = args->agbp->b_addr;
1029 	int			error = 0;
1030 	xfs_agblock_t		fbno = NULLAGBLOCK;
1031 	xfs_extlen_t		flen = 0;
1032 	int			i = 0;
1033 
1034 	/*
1035 	 * If a cntbt cursor is provided, try to allocate the largest record in
1036 	 * the tree. Try the AGFL if the cntbt is empty, otherwise fail the
1037 	 * allocation. Make sure to respect minleft even when pulling from the
1038 	 * freelist.
1039 	 */
1040 	if (ccur)
1041 		error = xfs_btree_decrement(ccur, 0, &i);
1042 	if (error)
1043 		goto error;
1044 	if (i) {
1045 		error = xfs_alloc_get_rec(ccur, &fbno, &flen, &i);
1046 		if (error)
1047 			goto error;
1048 		if (XFS_IS_CORRUPT(args->mp, i != 1)) {
1049 			error = -EFSCORRUPTED;
1050 			goto error;
1051 		}
1052 		goto out;
1053 	}
1054 
1055 	if (args->minlen != 1 || args->alignment != 1 ||
1056 	    args->resv == XFS_AG_RESV_AGFL ||
1057 	    be32_to_cpu(agf->agf_flcount) <= args->minleft)
1058 		goto out;
1059 
1060 	error = xfs_alloc_get_freelist(args->tp, args->agbp, &fbno, 0);
1061 	if (error)
1062 		goto error;
1063 	if (fbno == NULLAGBLOCK)
1064 		goto out;
1065 
1066 	xfs_extent_busy_reuse(args->mp, args->agno, fbno, 1,
1067 			      (args->datatype & XFS_ALLOC_NOBUSY));
1068 
1069 	if (args->datatype & XFS_ALLOC_USERDATA) {
1070 		struct xfs_buf	*bp;
1071 
1072 		error = xfs_trans_get_buf(args->tp, args->mp->m_ddev_targp,
1073 				XFS_AGB_TO_DADDR(args->mp, args->agno, fbno),
1074 				args->mp->m_bsize, 0, &bp);
1075 		if (error)
1076 			goto error;
1077 		xfs_trans_binval(args->tp, bp);
1078 	}
1079 	*fbnop = args->agbno = fbno;
1080 	*flenp = args->len = 1;
1081 	if (XFS_IS_CORRUPT(args->mp, fbno >= be32_to_cpu(agf->agf_length))) {
1082 		error = -EFSCORRUPTED;
1083 		goto error;
1084 	}
1085 	args->wasfromfl = 1;
1086 	trace_xfs_alloc_small_freelist(args);
1087 
1088 	/*
1089 	 * If we're feeding an AGFL block to something that doesn't live in the
1090 	 * free space, we need to clear out the OWN_AG rmap.
1091 	 */
1092 	error = xfs_rmap_free(args->tp, args->agbp, args->agno, fbno, 1,
1093 			      &XFS_RMAP_OINFO_AG);
1094 	if (error)
1095 		goto error;
1096 
1097 	*stat = 0;
1098 	return 0;
1099 
1100 out:
1101 	/*
1102 	 * Can't do the allocation, give up.
1103 	 */
1104 	if (flen < args->minlen) {
1105 		args->agbno = NULLAGBLOCK;
1106 		trace_xfs_alloc_small_notenough(args);
1107 		flen = 0;
1108 	}
1109 	*fbnop = fbno;
1110 	*flenp = flen;
1111 	*stat = 1;
1112 	trace_xfs_alloc_small_done(args);
1113 	return 0;
1114 
1115 error:
1116 	trace_xfs_alloc_small_error(args);
1117 	return error;
1118 }
1119 
1120 /*
1121  * Allocate a variable extent in the allocation group agno.
1122  * Type and bno are used to determine where in the allocation group the
1123  * extent will start.
1124  * Extent's length (returned in *len) will be between minlen and maxlen,
1125  * and of the form k * prod + mod unless there's nothing that large.
1126  * Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
1127  */
1128 STATIC int			/* error */
1129 xfs_alloc_ag_vextent(
1130 	xfs_alloc_arg_t	*args)	/* argument structure for allocation */
1131 {
1132 	int		error=0;
1133 
1134 	ASSERT(args->minlen > 0);
1135 	ASSERT(args->maxlen > 0);
1136 	ASSERT(args->minlen <= args->maxlen);
1137 	ASSERT(args->mod < args->prod);
1138 	ASSERT(args->alignment > 0);
1139 
1140 	/*
1141 	 * Branch to correct routine based on the type.
1142 	 */
1143 	args->wasfromfl = 0;
1144 	switch (args->type) {
1145 	case XFS_ALLOCTYPE_THIS_AG:
1146 		error = xfs_alloc_ag_vextent_size(args);
1147 		break;
1148 	case XFS_ALLOCTYPE_NEAR_BNO:
1149 		error = xfs_alloc_ag_vextent_near(args);
1150 		break;
1151 	case XFS_ALLOCTYPE_THIS_BNO:
1152 		error = xfs_alloc_ag_vextent_exact(args);
1153 		break;
1154 	default:
1155 		ASSERT(0);
1156 		/* NOTREACHED */
1157 	}
1158 
1159 	if (error || args->agbno == NULLAGBLOCK)
1160 		return error;
1161 
1162 	ASSERT(args->len >= args->minlen);
1163 	ASSERT(args->len <= args->maxlen);
1164 	ASSERT(!args->wasfromfl || args->resv != XFS_AG_RESV_AGFL);
1165 	ASSERT(args->agbno % args->alignment == 0);
1166 
1167 	/* if not file data, insert new block into the reverse map btree */
1168 	if (!xfs_rmap_should_skip_owner_update(&args->oinfo)) {
1169 		error = xfs_rmap_alloc(args->tp, args->agbp, args->agno,
1170 				       args->agbno, args->len, &args->oinfo);
1171 		if (error)
1172 			return error;
1173 	}
1174 
1175 	if (!args->wasfromfl) {
1176 		error = xfs_alloc_update_counters(args->tp, args->agbp,
1177 						  -((long)(args->len)));
1178 		if (error)
1179 			return error;
1180 
1181 		ASSERT(!xfs_extent_busy_search(args->mp, args->agno,
1182 					      args->agbno, args->len));
1183 	}
1184 
1185 	xfs_ag_resv_alloc_extent(args->pag, args->resv, args);
1186 
1187 	XFS_STATS_INC(args->mp, xs_allocx);
1188 	XFS_STATS_ADD(args->mp, xs_allocb, args->len);
1189 	return error;
1190 }
1191 
1192 /*
1193  * Allocate a variable extent at exactly agno/bno.
1194  * Extent's length (returned in *len) will be between minlen and maxlen,
1195  * and of the form k * prod + mod unless there's nothing that large.
1196  * Return the starting a.g. block (bno), or NULLAGBLOCK if we can't do it.
1197  */
1198 STATIC int			/* error */
1199 xfs_alloc_ag_vextent_exact(
1200 	xfs_alloc_arg_t	*args)	/* allocation argument structure */
1201 {
1202 	struct xfs_agf __maybe_unused *agf = args->agbp->b_addr;
1203 	xfs_btree_cur_t	*bno_cur;/* by block-number btree cursor */
1204 	xfs_btree_cur_t	*cnt_cur;/* by count btree cursor */
1205 	int		error;
1206 	xfs_agblock_t	fbno;	/* start block of found extent */
1207 	xfs_extlen_t	flen;	/* length of found extent */
1208 	xfs_agblock_t	tbno;	/* start block of busy extent */
1209 	xfs_extlen_t	tlen;	/* length of busy extent */
1210 	xfs_agblock_t	tend;	/* end block of busy extent */
1211 	int		i;	/* success/failure of operation */
1212 	unsigned	busy_gen;
1213 
1214 	ASSERT(args->alignment == 1);
1215 
1216 	/*
1217 	 * Allocate/initialize a cursor for the by-number freespace btree.
1218 	 */
1219 	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
1220 					  args->agno, XFS_BTNUM_BNO);
1221 
1222 	/*
1223 	 * Lookup bno and minlen in the btree (minlen is irrelevant, really).
1224 	 * Look for the closest free block <= bno, it must contain bno
1225 	 * if any free block does.
1226 	 */
1227 	error = xfs_alloc_lookup_le(bno_cur, args->agbno, args->minlen, &i);
1228 	if (error)
1229 		goto error0;
1230 	if (!i)
1231 		goto not_found;
1232 
1233 	/*
1234 	 * Grab the freespace record.
1235 	 */
1236 	error = xfs_alloc_get_rec(bno_cur, &fbno, &flen, &i);
1237 	if (error)
1238 		goto error0;
1239 	if (XFS_IS_CORRUPT(args->mp, i != 1)) {
1240 		error = -EFSCORRUPTED;
1241 		goto error0;
1242 	}
1243 	ASSERT(fbno <= args->agbno);
1244 
1245 	/*
1246 	 * Check for overlapping busy extents.
1247 	 */
1248 	tbno = fbno;
1249 	tlen = flen;
1250 	xfs_extent_busy_trim(args, &tbno, &tlen, &busy_gen);
1251 
1252 	/*
1253 	 * Give up if the start of the extent is busy, or the freespace isn't
1254 	 * long enough for the minimum request.
1255 	 */
1256 	if (tbno > args->agbno)
1257 		goto not_found;
1258 	if (tlen < args->minlen)
1259 		goto not_found;
1260 	tend = tbno + tlen;
1261 	if (tend < args->agbno + args->minlen)
1262 		goto not_found;
1263 
1264 	/*
1265 	 * End of extent will be smaller of the freespace end and the
1266 	 * maximal requested end.
1267 	 *
1268 	 * Fix the length according to mod and prod if given.
1269 	 */
1270 	args->len = XFS_AGBLOCK_MIN(tend, args->agbno + args->maxlen)
1271 						- args->agbno;
1272 	xfs_alloc_fix_len(args);
1273 	ASSERT(args->agbno + args->len <= tend);
1274 
1275 	/*
1276 	 * We are allocating agbno for args->len
1277 	 * Allocate/initialize a cursor for the by-size btree.
1278 	 */
1279 	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
1280 		args->agno, XFS_BTNUM_CNT);
1281 	ASSERT(args->agbno + args->len <= be32_to_cpu(agf->agf_length));
1282 	error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen, args->agbno,
1283 				      args->len, XFSA_FIXUP_BNO_OK);
1284 	if (error) {
1285 		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
1286 		goto error0;
1287 	}
1288 
1289 	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
1290 	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
1291 
1292 	args->wasfromfl = 0;
1293 	trace_xfs_alloc_exact_done(args);
1294 	return 0;
1295 
1296 not_found:
1297 	/* Didn't find it, return null. */
1298 	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
1299 	args->agbno = NULLAGBLOCK;
1300 	trace_xfs_alloc_exact_notfound(args);
1301 	return 0;
1302 
1303 error0:
1304 	xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
1305 	trace_xfs_alloc_exact_error(args);
1306 	return error;
1307 }
1308 
1309 /*
1310  * Search a given number of btree records in a given direction. Check each
1311  * record against the good extent we've already found.
1312  */
1313 STATIC int
1314 xfs_alloc_walk_iter(
1315 	struct xfs_alloc_arg	*args,
1316 	struct xfs_alloc_cur	*acur,
1317 	struct xfs_btree_cur	*cur,
1318 	bool			increment,
1319 	bool			find_one, /* quit on first candidate */
1320 	int			count,    /* rec count (-1 for infinite) */
1321 	int			*stat)
1322 {
1323 	int			error;
1324 	int			i;
1325 
1326 	*stat = 0;
1327 
1328 	/*
1329 	 * Search so long as the cursor is active or we find a better extent.
1330 	 * The cursor is deactivated if it extends beyond the range of the
1331 	 * current allocation candidate.
1332 	 */
1333 	while (xfs_alloc_cur_active(cur) && count) {
1334 		error = xfs_alloc_cur_check(args, acur, cur, &i);
1335 		if (error)
1336 			return error;
1337 		if (i == 1) {
1338 			*stat = 1;
1339 			if (find_one)
1340 				break;
1341 		}
1342 		if (!xfs_alloc_cur_active(cur))
1343 			break;
1344 
1345 		if (increment)
1346 			error = xfs_btree_increment(cur, 0, &i);
1347 		else
1348 			error = xfs_btree_decrement(cur, 0, &i);
1349 		if (error)
1350 			return error;
1351 		if (i == 0)
1352 			cur->bc_ag.abt.active = false;
1353 
1354 		if (count > 0)
1355 			count--;
1356 	}
1357 
1358 	return 0;
1359 }
1360 
1361 /*
1362  * Search the by-bno and by-size btrees in parallel in search of an extent with
1363  * ideal locality based on the NEAR mode ->agbno locality hint.
1364  */
1365 STATIC int
1366 xfs_alloc_ag_vextent_locality(
1367 	struct xfs_alloc_arg	*args,
1368 	struct xfs_alloc_cur	*acur,
1369 	int			*stat)
1370 {
1371 	struct xfs_btree_cur	*fbcur = NULL;
1372 	int			error;
1373 	int			i;
1374 	bool			fbinc;
1375 
1376 	ASSERT(acur->len == 0);
1377 	ASSERT(args->type == XFS_ALLOCTYPE_NEAR_BNO);
1378 
1379 	*stat = 0;
1380 
1381 	error = xfs_alloc_lookup_ge(acur->cnt, args->agbno, acur->cur_len, &i);
1382 	if (error)
1383 		return error;
1384 	error = xfs_alloc_lookup_le(acur->bnolt, args->agbno, 0, &i);
1385 	if (error)
1386 		return error;
1387 	error = xfs_alloc_lookup_ge(acur->bnogt, args->agbno, 0, &i);
1388 	if (error)
1389 		return error;
1390 
1391 	/*
1392 	 * Search the bnobt and cntbt in parallel. Search the bnobt left and
1393 	 * right and lookup the closest extent to the locality hint for each
1394 	 * extent size key in the cntbt. The entire search terminates
1395 	 * immediately on a bnobt hit because that means we've found best case
1396 	 * locality. Otherwise the search continues until the cntbt cursor runs
1397 	 * off the end of the tree. If no allocation candidate is found at this
1398 	 * point, give up on locality, walk backwards from the end of the cntbt
1399 	 * and take the first available extent.
1400 	 *
1401 	 * The parallel tree searches balance each other out to provide fairly
1402 	 * consistent performance for various situations. The bnobt search can
1403 	 * have pathological behavior in the worst case scenario of larger
1404 	 * allocation requests and fragmented free space. On the other hand, the
1405 	 * bnobt is able to satisfy most smaller allocation requests much more
1406 	 * quickly than the cntbt. The cntbt search can sift through fragmented
1407 	 * free space and sets of free extents for larger allocation requests
1408 	 * more quickly than the bnobt. Since the locality hint is just a hint
1409 	 * and we don't want to scan the entire bnobt for perfect locality, the
1410 	 * cntbt search essentially bounds the bnobt search such that we can
1411 	 * find good enough locality at reasonable performance in most cases.
1412 	 */
1413 	while (xfs_alloc_cur_active(acur->bnolt) ||
1414 	       xfs_alloc_cur_active(acur->bnogt) ||
1415 	       xfs_alloc_cur_active(acur->cnt)) {
1416 
1417 		trace_xfs_alloc_cur_lookup(args);
1418 
1419 		/*
1420 		 * Search the bnobt left and right. In the case of a hit, finish
1421 		 * the search in the opposite direction and we're done.
1422 		 */
1423 		error = xfs_alloc_walk_iter(args, acur, acur->bnolt, false,
1424 					    true, 1, &i);
1425 		if (error)
1426 			return error;
1427 		if (i == 1) {
1428 			trace_xfs_alloc_cur_left(args);
1429 			fbcur = acur->bnogt;
1430 			fbinc = true;
1431 			break;
1432 		}
1433 		error = xfs_alloc_walk_iter(args, acur, acur->bnogt, true, true,
1434 					    1, &i);
1435 		if (error)
1436 			return error;
1437 		if (i == 1) {
1438 			trace_xfs_alloc_cur_right(args);
1439 			fbcur = acur->bnolt;
1440 			fbinc = false;
1441 			break;
1442 		}
1443 
1444 		/*
1445 		 * Check the extent with best locality based on the current
1446 		 * extent size search key and keep track of the best candidate.
1447 		 */
1448 		error = xfs_alloc_cntbt_iter(args, acur);
1449 		if (error)
1450 			return error;
1451 		if (!xfs_alloc_cur_active(acur->cnt)) {
1452 			trace_xfs_alloc_cur_lookup_done(args);
1453 			break;
1454 		}
1455 	}
1456 
1457 	/*
1458 	 * If we failed to find anything due to busy extents, return empty
1459 	 * handed so the caller can flush and retry. If no busy extents were
1460 	 * found, walk backwards from the end of the cntbt as a last resort.
1461 	 */
1462 	if (!xfs_alloc_cur_active(acur->cnt) && !acur->len && !acur->busy) {
1463 		error = xfs_btree_decrement(acur->cnt, 0, &i);
1464 		if (error)
1465 			return error;
1466 		if (i) {
1467 			acur->cnt->bc_ag.abt.active = true;
1468 			fbcur = acur->cnt;
1469 			fbinc = false;
1470 		}
1471 	}
1472 
1473 	/*
1474 	 * Search in the opposite direction for a better entry in the case of
1475 	 * a bnobt hit or walk backwards from the end of the cntbt.
1476 	 */
1477 	if (fbcur) {
1478 		error = xfs_alloc_walk_iter(args, acur, fbcur, fbinc, true, -1,
1479 					    &i);
1480 		if (error)
1481 			return error;
1482 	}
1483 
1484 	if (acur->len)
1485 		*stat = 1;
1486 
1487 	return 0;
1488 }
1489 
1490 /* Check the last block of the cnt btree for allocations. */
1491 static int
1492 xfs_alloc_ag_vextent_lastblock(
1493 	struct xfs_alloc_arg	*args,
1494 	struct xfs_alloc_cur	*acur,
1495 	xfs_agblock_t		*bno,
1496 	xfs_extlen_t		*len,
1497 	bool			*allocated)
1498 {
1499 	int			error;
1500 	int			i;
1501 
1502 #ifdef DEBUG
1503 	/* Randomly don't execute the first algorithm. */
1504 	if (prandom_u32() & 1)
1505 		return 0;
1506 #endif
1507 
1508 	/*
1509 	 * Start from the entry that lookup found, sequence through all larger
1510 	 * free blocks.  If we're actually pointing at a record smaller than
1511 	 * maxlen, go to the start of this block, and skip all those smaller
1512 	 * than minlen.
1513 	 */
1514 	if (*len || args->alignment > 1) {
1515 		acur->cnt->bc_ptrs[0] = 1;
1516 		do {
1517 			error = xfs_alloc_get_rec(acur->cnt, bno, len, &i);
1518 			if (error)
1519 				return error;
1520 			if (XFS_IS_CORRUPT(args->mp, i != 1))
1521 				return -EFSCORRUPTED;
1522 			if (*len >= args->minlen)
1523 				break;
1524 			error = xfs_btree_increment(acur->cnt, 0, &i);
1525 			if (error)
1526 				return error;
1527 		} while (i);
1528 		ASSERT(*len >= args->minlen);
1529 		if (!i)
1530 			return 0;
1531 	}
1532 
1533 	error = xfs_alloc_walk_iter(args, acur, acur->cnt, true, false, -1, &i);
1534 	if (error)
1535 		return error;
1536 
1537 	/*
1538 	 * It didn't work.  We COULD be in a case where there's a good record
1539 	 * somewhere, so try again.
1540 	 */
1541 	if (acur->len == 0)
1542 		return 0;
1543 
1544 	trace_xfs_alloc_near_first(args);
1545 	*allocated = true;
1546 	return 0;
1547 }
1548 
1549 /*
1550  * Allocate a variable extent near bno in the allocation group agno.
1551  * Extent's length (returned in len) will be between minlen and maxlen,
1552  * and of the form k * prod + mod unless there's nothing that large.
1553  * Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
1554  */
1555 STATIC int
1556 xfs_alloc_ag_vextent_near(
1557 	struct xfs_alloc_arg	*args)
1558 {
1559 	struct xfs_alloc_cur	acur = {};
1560 	int			error;		/* error code */
1561 	int			i;		/* result code, temporary */
1562 	xfs_agblock_t		bno;
1563 	xfs_extlen_t		len;
1564 
1565 	/* handle uninitialized agbno range so caller doesn't have to */
1566 	if (!args->min_agbno && !args->max_agbno)
1567 		args->max_agbno = args->mp->m_sb.sb_agblocks - 1;
1568 	ASSERT(args->min_agbno <= args->max_agbno);
1569 
1570 	/* clamp agbno to the range if it's outside */
1571 	if (args->agbno < args->min_agbno)
1572 		args->agbno = args->min_agbno;
1573 	if (args->agbno > args->max_agbno)
1574 		args->agbno = args->max_agbno;
1575 
1576 restart:
1577 	len = 0;
1578 
1579 	/*
1580 	 * Set up cursors and see if there are any free extents as big as
1581 	 * maxlen. If not, pick the last entry in the tree unless the tree is
1582 	 * empty.
1583 	 */
1584 	error = xfs_alloc_cur_setup(args, &acur);
1585 	if (error == -ENOSPC) {
1586 		error = xfs_alloc_ag_vextent_small(args, acur.cnt, &bno,
1587 				&len, &i);
1588 		if (error)
1589 			goto out;
1590 		if (i == 0 || len == 0) {
1591 			trace_xfs_alloc_near_noentry(args);
1592 			goto out;
1593 		}
1594 		ASSERT(i == 1);
1595 	} else if (error) {
1596 		goto out;
1597 	}
1598 
1599 	/*
1600 	 * First algorithm.
1601 	 * If the requested extent is large wrt the freespaces available
1602 	 * in this a.g., then the cursor will be pointing to a btree entry
1603 	 * near the right edge of the tree.  If it's in the last btree leaf
1604 	 * block, then we just examine all the entries in that block
1605 	 * that are big enough, and pick the best one.
1606 	 */
1607 	if (xfs_btree_islastblock(acur.cnt, 0)) {
1608 		bool		allocated = false;
1609 
1610 		error = xfs_alloc_ag_vextent_lastblock(args, &acur, &bno, &len,
1611 				&allocated);
1612 		if (error)
1613 			goto out;
1614 		if (allocated)
1615 			goto alloc_finish;
1616 	}
1617 
1618 	/*
1619 	 * Second algorithm. Combined cntbt and bnobt search to find ideal
1620 	 * locality.
1621 	 */
1622 	error = xfs_alloc_ag_vextent_locality(args, &acur, &i);
1623 	if (error)
1624 		goto out;
1625 
1626 	/*
1627 	 * If we couldn't get anything, give up.
1628 	 */
1629 	if (!acur.len) {
1630 		if (acur.busy) {
1631 			trace_xfs_alloc_near_busy(args);
1632 			xfs_extent_busy_flush(args->mp, args->pag,
1633 					      acur.busy_gen);
1634 			goto restart;
1635 		}
1636 		trace_xfs_alloc_size_neither(args);
1637 		args->agbno = NULLAGBLOCK;
1638 		goto out;
1639 	}
1640 
1641 alloc_finish:
1642 	/* fix up btrees on a successful allocation */
1643 	error = xfs_alloc_cur_finish(args, &acur);
1644 
1645 out:
1646 	xfs_alloc_cur_close(&acur, error);
1647 	return error;
1648 }
1649 
1650 /*
1651  * Allocate a variable extent anywhere in the allocation group agno.
1652  * Extent's length (returned in len) will be between minlen and maxlen,
1653  * and of the form k * prod + mod unless there's nothing that large.
1654  * Return the starting a.g. block, or NULLAGBLOCK if we can't do it.
1655  */
1656 STATIC int				/* error */
1657 xfs_alloc_ag_vextent_size(
1658 	xfs_alloc_arg_t	*args)		/* allocation argument structure */
1659 {
1660 	struct xfs_agf	*agf = args->agbp->b_addr;
1661 	xfs_btree_cur_t	*bno_cur;	/* cursor for bno btree */
1662 	xfs_btree_cur_t	*cnt_cur;	/* cursor for cnt btree */
1663 	int		error;		/* error result */
1664 	xfs_agblock_t	fbno;		/* start of found freespace */
1665 	xfs_extlen_t	flen;		/* length of found freespace */
1666 	int		i;		/* temp status variable */
1667 	xfs_agblock_t	rbno;		/* returned block number */
1668 	xfs_extlen_t	rlen;		/* length of returned extent */
1669 	bool		busy;
1670 	unsigned	busy_gen;
1671 
1672 restart:
1673 	/*
1674 	 * Allocate and initialize a cursor for the by-size btree.
1675 	 */
1676 	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
1677 		args->agno, XFS_BTNUM_CNT);
1678 	bno_cur = NULL;
1679 	busy = false;
1680 
1681 	/*
1682 	 * Look for an entry >= maxlen+alignment-1 blocks.
1683 	 */
1684 	if ((error = xfs_alloc_lookup_ge(cnt_cur, 0,
1685 			args->maxlen + args->alignment - 1, &i)))
1686 		goto error0;
1687 
1688 	/*
1689 	 * If none then we have to settle for a smaller extent. In the case that
1690 	 * there are no large extents, this will return the last entry in the
1691 	 * tree unless the tree is empty. In the case that there are only busy
1692 	 * large extents, this will return the largest small extent unless there
1693 	 * are no smaller extents available.
1694 	 */
1695 	if (!i) {
1696 		error = xfs_alloc_ag_vextent_small(args, cnt_cur,
1697 						   &fbno, &flen, &i);
1698 		if (error)
1699 			goto error0;
1700 		if (i == 0 || flen == 0) {
1701 			xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
1702 			trace_xfs_alloc_size_noentry(args);
1703 			return 0;
1704 		}
1705 		ASSERT(i == 1);
1706 		busy = xfs_alloc_compute_aligned(args, fbno, flen, &rbno,
1707 				&rlen, &busy_gen);
1708 	} else {
1709 		/*
1710 		 * Search for a non-busy extent that is large enough.
1711 		 */
1712 		for (;;) {
1713 			error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen, &i);
1714 			if (error)
1715 				goto error0;
1716 			if (XFS_IS_CORRUPT(args->mp, i != 1)) {
1717 				error = -EFSCORRUPTED;
1718 				goto error0;
1719 			}
1720 
1721 			busy = xfs_alloc_compute_aligned(args, fbno, flen,
1722 					&rbno, &rlen, &busy_gen);
1723 
1724 			if (rlen >= args->maxlen)
1725 				break;
1726 
1727 			error = xfs_btree_increment(cnt_cur, 0, &i);
1728 			if (error)
1729 				goto error0;
1730 			if (i == 0) {
1731 				/*
1732 				 * Our only valid extents must have been busy.
1733 				 * Make it unbusy by forcing the log out and
1734 				 * retrying.
1735 				 */
1736 				xfs_btree_del_cursor(cnt_cur,
1737 						     XFS_BTREE_NOERROR);
1738 				trace_xfs_alloc_size_busy(args);
1739 				xfs_extent_busy_flush(args->mp,
1740 							args->pag, busy_gen);
1741 				goto restart;
1742 			}
1743 		}
1744 	}
1745 
1746 	/*
1747 	 * In the first case above, we got the last entry in the
1748 	 * by-size btree.  Now we check to see if the space hits maxlen
1749 	 * once aligned; if not, we search left for something better.
1750 	 * This can't happen in the second case above.
1751 	 */
1752 	rlen = XFS_EXTLEN_MIN(args->maxlen, rlen);
1753 	if (XFS_IS_CORRUPT(args->mp,
1754 			   rlen != 0 &&
1755 			   (rlen > flen ||
1756 			    rbno + rlen > fbno + flen))) {
1757 		error = -EFSCORRUPTED;
1758 		goto error0;
1759 	}
1760 	if (rlen < args->maxlen) {
1761 		xfs_agblock_t	bestfbno;
1762 		xfs_extlen_t	bestflen;
1763 		xfs_agblock_t	bestrbno;
1764 		xfs_extlen_t	bestrlen;
1765 
1766 		bestrlen = rlen;
1767 		bestrbno = rbno;
1768 		bestflen = flen;
1769 		bestfbno = fbno;
1770 		for (;;) {
1771 			if ((error = xfs_btree_decrement(cnt_cur, 0, &i)))
1772 				goto error0;
1773 			if (i == 0)
1774 				break;
1775 			if ((error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen,
1776 					&i)))
1777 				goto error0;
1778 			if (XFS_IS_CORRUPT(args->mp, i != 1)) {
1779 				error = -EFSCORRUPTED;
1780 				goto error0;
1781 			}
1782 			if (flen < bestrlen)
1783 				break;
1784 			busy = xfs_alloc_compute_aligned(args, fbno, flen,
1785 					&rbno, &rlen, &busy_gen);
1786 			rlen = XFS_EXTLEN_MIN(args->maxlen, rlen);
1787 			if (XFS_IS_CORRUPT(args->mp,
1788 					   rlen != 0 &&
1789 					   (rlen > flen ||
1790 					    rbno + rlen > fbno + flen))) {
1791 				error = -EFSCORRUPTED;
1792 				goto error0;
1793 			}
1794 			if (rlen > bestrlen) {
1795 				bestrlen = rlen;
1796 				bestrbno = rbno;
1797 				bestflen = flen;
1798 				bestfbno = fbno;
1799 				if (rlen == args->maxlen)
1800 					break;
1801 			}
1802 		}
1803 		if ((error = xfs_alloc_lookup_eq(cnt_cur, bestfbno, bestflen,
1804 				&i)))
1805 			goto error0;
1806 		if (XFS_IS_CORRUPT(args->mp, i != 1)) {
1807 			error = -EFSCORRUPTED;
1808 			goto error0;
1809 		}
1810 		rlen = bestrlen;
1811 		rbno = bestrbno;
1812 		flen = bestflen;
1813 		fbno = bestfbno;
1814 	}
1815 	args->wasfromfl = 0;
1816 	/*
1817 	 * Fix up the length.
1818 	 */
1819 	args->len = rlen;
1820 	if (rlen < args->minlen) {
1821 		if (busy) {
1822 			xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
1823 			trace_xfs_alloc_size_busy(args);
1824 			xfs_extent_busy_flush(args->mp, args->pag, busy_gen);
1825 			goto restart;
1826 		}
1827 		goto out_nominleft;
1828 	}
1829 	xfs_alloc_fix_len(args);
1830 
1831 	rlen = args->len;
1832 	if (XFS_IS_CORRUPT(args->mp, rlen > flen)) {
1833 		error = -EFSCORRUPTED;
1834 		goto error0;
1835 	}
1836 	/*
1837 	 * Allocate and initialize a cursor for the by-block tree.
1838 	 */
1839 	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
1840 		args->agno, XFS_BTNUM_BNO);
1841 	if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen,
1842 			rbno, rlen, XFSA_FIXUP_CNT_OK)))
1843 		goto error0;
1844 	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
1845 	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
1846 	cnt_cur = bno_cur = NULL;
1847 	args->len = rlen;
1848 	args->agbno = rbno;
1849 	if (XFS_IS_CORRUPT(args->mp,
1850 			   args->agbno + args->len >
1851 			   be32_to_cpu(agf->agf_length))) {
1852 		error = -EFSCORRUPTED;
1853 		goto error0;
1854 	}
1855 	trace_xfs_alloc_size_done(args);
1856 	return 0;
1857 
1858 error0:
1859 	trace_xfs_alloc_size_error(args);
1860 	if (cnt_cur)
1861 		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
1862 	if (bno_cur)
1863 		xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
1864 	return error;
1865 
1866 out_nominleft:
1867 	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
1868 	trace_xfs_alloc_size_nominleft(args);
1869 	args->agbno = NULLAGBLOCK;
1870 	return 0;
1871 }
1872 
1873 /*
1874  * Free the extent starting at agno/bno for length.
1875  */
1876 STATIC int
1877 xfs_free_ag_extent(
1878 	struct xfs_trans		*tp,
1879 	struct xfs_buf			*agbp,
1880 	xfs_agnumber_t			agno,
1881 	xfs_agblock_t			bno,
1882 	xfs_extlen_t			len,
1883 	const struct xfs_owner_info	*oinfo,
1884 	enum xfs_ag_resv_type		type)
1885 {
1886 	struct xfs_mount		*mp;
1887 	struct xfs_btree_cur		*bno_cur;
1888 	struct xfs_btree_cur		*cnt_cur;
1889 	xfs_agblock_t			gtbno; /* start of right neighbor */
1890 	xfs_extlen_t			gtlen; /* length of right neighbor */
1891 	xfs_agblock_t			ltbno; /* start of left neighbor */
1892 	xfs_extlen_t			ltlen; /* length of left neighbor */
1893 	xfs_agblock_t			nbno; /* new starting block of freesp */
1894 	xfs_extlen_t			nlen; /* new length of freespace */
1895 	int				haveleft; /* have a left neighbor */
1896 	int				haveright; /* have a right neighbor */
1897 	int				i;
1898 	int				error;
1899 
1900 	bno_cur = cnt_cur = NULL;
1901 	mp = tp->t_mountp;
1902 
1903 	if (!xfs_rmap_should_skip_owner_update(oinfo)) {
1904 		error = xfs_rmap_free(tp, agbp, agno, bno, len, oinfo);
1905 		if (error)
1906 			goto error0;
1907 	}
1908 
1909 	/*
1910 	 * Allocate and initialize a cursor for the by-block btree.
1911 	 */
1912 	bno_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_BNO);
1913 	/*
1914 	 * Look for a neighboring block on the left (lower block numbers)
1915 	 * that is contiguous with this space.
1916 	 */
1917 	if ((error = xfs_alloc_lookup_le(bno_cur, bno, len, &haveleft)))
1918 		goto error0;
1919 	if (haveleft) {
1920 		/*
1921 		 * There is a block to our left.
1922 		 */
1923 		if ((error = xfs_alloc_get_rec(bno_cur, &ltbno, &ltlen, &i)))
1924 			goto error0;
1925 		if (XFS_IS_CORRUPT(mp, i != 1)) {
1926 			error = -EFSCORRUPTED;
1927 			goto error0;
1928 		}
1929 		/*
1930 		 * It's not contiguous, though.
1931 		 */
1932 		if (ltbno + ltlen < bno)
1933 			haveleft = 0;
1934 		else {
1935 			/*
1936 			 * If this failure happens the request to free this
1937 			 * space was invalid, it's (partly) already free.
1938 			 * Very bad.
1939 			 */
1940 			if (XFS_IS_CORRUPT(mp, ltbno + ltlen > bno)) {
1941 				error = -EFSCORRUPTED;
1942 				goto error0;
1943 			}
1944 		}
1945 	}
1946 	/*
1947 	 * Look for a neighboring block on the right (higher block numbers)
1948 	 * that is contiguous with this space.
1949 	 */
1950 	if ((error = xfs_btree_increment(bno_cur, 0, &haveright)))
1951 		goto error0;
1952 	if (haveright) {
1953 		/*
1954 		 * There is a block to our right.
1955 		 */
1956 		if ((error = xfs_alloc_get_rec(bno_cur, &gtbno, &gtlen, &i)))
1957 			goto error0;
1958 		if (XFS_IS_CORRUPT(mp, i != 1)) {
1959 			error = -EFSCORRUPTED;
1960 			goto error0;
1961 		}
1962 		/*
1963 		 * It's not contiguous, though.
1964 		 */
1965 		if (bno + len < gtbno)
1966 			haveright = 0;
1967 		else {
1968 			/*
1969 			 * If this failure happens the request to free this
1970 			 * space was invalid, it's (partly) already free.
1971 			 * Very bad.
1972 			 */
1973 			if (XFS_IS_CORRUPT(mp, bno + len > gtbno)) {
1974 				error = -EFSCORRUPTED;
1975 				goto error0;
1976 			}
1977 		}
1978 	}
1979 	/*
1980 	 * Now allocate and initialize a cursor for the by-size tree.
1981 	 */
1982 	cnt_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_CNT);
1983 	/*
1984 	 * Have both left and right contiguous neighbors.
1985 	 * Merge all three into a single free block.
1986 	 */
1987 	if (haveleft && haveright) {
1988 		/*
1989 		 * Delete the old by-size entry on the left.
1990 		 */
1991 		if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
1992 			goto error0;
1993 		if (XFS_IS_CORRUPT(mp, i != 1)) {
1994 			error = -EFSCORRUPTED;
1995 			goto error0;
1996 		}
1997 		if ((error = xfs_btree_delete(cnt_cur, &i)))
1998 			goto error0;
1999 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2000 			error = -EFSCORRUPTED;
2001 			goto error0;
2002 		}
2003 		/*
2004 		 * Delete the old by-size entry on the right.
2005 		 */
2006 		if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
2007 			goto error0;
2008 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2009 			error = -EFSCORRUPTED;
2010 			goto error0;
2011 		}
2012 		if ((error = xfs_btree_delete(cnt_cur, &i)))
2013 			goto error0;
2014 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2015 			error = -EFSCORRUPTED;
2016 			goto error0;
2017 		}
2018 		/*
2019 		 * Delete the old by-block entry for the right block.
2020 		 */
2021 		if ((error = xfs_btree_delete(bno_cur, &i)))
2022 			goto error0;
2023 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2024 			error = -EFSCORRUPTED;
2025 			goto error0;
2026 		}
2027 		/*
2028 		 * Move the by-block cursor back to the left neighbor.
2029 		 */
2030 		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
2031 			goto error0;
2032 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2033 			error = -EFSCORRUPTED;
2034 			goto error0;
2035 		}
2036 #ifdef DEBUG
2037 		/*
2038 		 * Check that this is the right record: delete didn't
2039 		 * mangle the cursor.
2040 		 */
2041 		{
2042 			xfs_agblock_t	xxbno;
2043 			xfs_extlen_t	xxlen;
2044 
2045 			if ((error = xfs_alloc_get_rec(bno_cur, &xxbno, &xxlen,
2046 					&i)))
2047 				goto error0;
2048 			if (XFS_IS_CORRUPT(mp,
2049 					   i != 1 ||
2050 					   xxbno != ltbno ||
2051 					   xxlen != ltlen)) {
2052 				error = -EFSCORRUPTED;
2053 				goto error0;
2054 			}
2055 		}
2056 #endif
2057 		/*
2058 		 * Update remaining by-block entry to the new, joined block.
2059 		 */
2060 		nbno = ltbno;
2061 		nlen = len + ltlen + gtlen;
2062 		if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
2063 			goto error0;
2064 	}
2065 	/*
2066 	 * Have only a left contiguous neighbor.
2067 	 * Merge it together with the new freespace.
2068 	 */
2069 	else if (haveleft) {
2070 		/*
2071 		 * Delete the old by-size entry on the left.
2072 		 */
2073 		if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
2074 			goto error0;
2075 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2076 			error = -EFSCORRUPTED;
2077 			goto error0;
2078 		}
2079 		if ((error = xfs_btree_delete(cnt_cur, &i)))
2080 			goto error0;
2081 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2082 			error = -EFSCORRUPTED;
2083 			goto error0;
2084 		}
2085 		/*
2086 		 * Back up the by-block cursor to the left neighbor, and
2087 		 * update its length.
2088 		 */
2089 		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
2090 			goto error0;
2091 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2092 			error = -EFSCORRUPTED;
2093 			goto error0;
2094 		}
2095 		nbno = ltbno;
2096 		nlen = len + ltlen;
2097 		if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
2098 			goto error0;
2099 	}
2100 	/*
2101 	 * Have only a right contiguous neighbor.
2102 	 * Merge it together with the new freespace.
2103 	 */
2104 	else if (haveright) {
2105 		/*
2106 		 * Delete the old by-size entry on the right.
2107 		 */
2108 		if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
2109 			goto error0;
2110 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2111 			error = -EFSCORRUPTED;
2112 			goto error0;
2113 		}
2114 		if ((error = xfs_btree_delete(cnt_cur, &i)))
2115 			goto error0;
2116 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2117 			error = -EFSCORRUPTED;
2118 			goto error0;
2119 		}
2120 		/*
2121 		 * Update the starting block and length of the right
2122 		 * neighbor in the by-block tree.
2123 		 */
2124 		nbno = bno;
2125 		nlen = len + gtlen;
2126 		if ((error = xfs_alloc_update(bno_cur, nbno, nlen)))
2127 			goto error0;
2128 	}
2129 	/*
2130 	 * No contiguous neighbors.
2131 	 * Insert the new freespace into the by-block tree.
2132 	 */
2133 	else {
2134 		nbno = bno;
2135 		nlen = len;
2136 		if ((error = xfs_btree_insert(bno_cur, &i)))
2137 			goto error0;
2138 		if (XFS_IS_CORRUPT(mp, i != 1)) {
2139 			error = -EFSCORRUPTED;
2140 			goto error0;
2141 		}
2142 	}
2143 	xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR);
2144 	bno_cur = NULL;
2145 	/*
2146 	 * In all cases we need to insert the new freespace in the by-size tree.
2147 	 */
2148 	if ((error = xfs_alloc_lookup_eq(cnt_cur, nbno, nlen, &i)))
2149 		goto error0;
2150 	if (XFS_IS_CORRUPT(mp, i != 0)) {
2151 		error = -EFSCORRUPTED;
2152 		goto error0;
2153 	}
2154 	if ((error = xfs_btree_insert(cnt_cur, &i)))
2155 		goto error0;
2156 	if (XFS_IS_CORRUPT(mp, i != 1)) {
2157 		error = -EFSCORRUPTED;
2158 		goto error0;
2159 	}
2160 	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
2161 	cnt_cur = NULL;
2162 
2163 	/*
2164 	 * Update the freespace totals in the ag and superblock.
2165 	 */
2166 	error = xfs_alloc_update_counters(tp, agbp, len);
2167 	xfs_ag_resv_free_extent(agbp->b_pag, type, tp, len);
2168 	if (error)
2169 		goto error0;
2170 
2171 	XFS_STATS_INC(mp, xs_freex);
2172 	XFS_STATS_ADD(mp, xs_freeb, len);
2173 
2174 	trace_xfs_free_extent(mp, agno, bno, len, type, haveleft, haveright);
2175 
2176 	return 0;
2177 
2178  error0:
2179 	trace_xfs_free_extent(mp, agno, bno, len, type, -1, -1);
2180 	if (bno_cur)
2181 		xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
2182 	if (cnt_cur)
2183 		xfs_btree_del_cursor(cnt_cur, XFS_BTREE_ERROR);
2184 	return error;
2185 }
2186 
2187 /*
2188  * Visible (exported) allocation/free functions.
2189  * Some of these are used just by xfs_alloc_btree.c and this file.
2190  */
2191 
2192 /*
2193  * Compute and fill in value of m_ag_maxlevels.
2194  */
2195 void
2196 xfs_alloc_compute_maxlevels(
2197 	xfs_mount_t	*mp)	/* file system mount structure */
2198 {
2199 	mp->m_ag_maxlevels = xfs_btree_compute_maxlevels(mp->m_alloc_mnr,
2200 			(mp->m_sb.sb_agblocks + 1) / 2);
2201 }
2202 
2203 /*
2204  * Find the length of the longest extent in an AG.  The 'need' parameter
2205  * specifies how much space we're going to need for the AGFL and the
2206  * 'reserved' parameter tells us how many blocks in this AG are reserved for
2207  * other callers.
2208  */
2209 xfs_extlen_t
2210 xfs_alloc_longest_free_extent(
2211 	struct xfs_perag	*pag,
2212 	xfs_extlen_t		need,
2213 	xfs_extlen_t		reserved)
2214 {
2215 	xfs_extlen_t		delta = 0;
2216 
2217 	/*
2218 	 * If the AGFL needs a recharge, we'll have to subtract that from the
2219 	 * longest extent.
2220 	 */
2221 	if (need > pag->pagf_flcount)
2222 		delta = need - pag->pagf_flcount;
2223 
2224 	/*
2225 	 * If we cannot maintain others' reservations with space from the
2226 	 * not-longest freesp extents, we'll have to subtract /that/ from
2227 	 * the longest extent too.
2228 	 */
2229 	if (pag->pagf_freeblks - pag->pagf_longest < reserved)
2230 		delta += reserved - (pag->pagf_freeblks - pag->pagf_longest);
2231 
2232 	/*
2233 	 * If the longest extent is long enough to satisfy all the
2234 	 * reservations and AGFL rules in place, we can return this extent.
2235 	 */
2236 	if (pag->pagf_longest > delta)
2237 		return min_t(xfs_extlen_t, pag->pag_mount->m_ag_max_usable,
2238 				pag->pagf_longest - delta);
2239 
2240 	/* Otherwise, let the caller try for 1 block if there's space. */
2241 	return pag->pagf_flcount > 0 || pag->pagf_longest > 0;
2242 }
2243 
2244 /*
2245  * Compute the minimum length of the AGFL in the given AG.  If @pag is NULL,
2246  * return the largest possible minimum length.
2247  */
2248 unsigned int
2249 xfs_alloc_min_freelist(
2250 	struct xfs_mount	*mp,
2251 	struct xfs_perag	*pag)
2252 {
2253 	/* AG btrees have at least 1 level. */
2254 	static const uint8_t	fake_levels[XFS_BTNUM_AGF] = {1, 1, 1};
2255 	const uint8_t		*levels = pag ? pag->pagf_levels : fake_levels;
2256 	unsigned int		min_free;
2257 
2258 	ASSERT(mp->m_ag_maxlevels > 0);
2259 
2260 	/* space needed by-bno freespace btree */
2261 	min_free = min_t(unsigned int, levels[XFS_BTNUM_BNOi] + 1,
2262 				       mp->m_ag_maxlevels);
2263 	/* space needed by-size freespace btree */
2264 	min_free += min_t(unsigned int, levels[XFS_BTNUM_CNTi] + 1,
2265 				       mp->m_ag_maxlevels);
2266 	/* space needed reverse mapping used space btree */
2267 	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
2268 		min_free += min_t(unsigned int, levels[XFS_BTNUM_RMAPi] + 1,
2269 						mp->m_rmap_maxlevels);
2270 
2271 	return min_free;
2272 }
2273 
2274 /*
2275  * Check if the operation we are fixing up the freelist for should go ahead or
2276  * not. If we are freeing blocks, we always allow it, otherwise the allocation
2277  * is dependent on whether the size and shape of free space available will
2278  * permit the requested allocation to take place.
2279  */
2280 static bool
2281 xfs_alloc_space_available(
2282 	struct xfs_alloc_arg	*args,
2283 	xfs_extlen_t		min_free,
2284 	int			flags)
2285 {
2286 	struct xfs_perag	*pag = args->pag;
2287 	xfs_extlen_t		alloc_len, longest;
2288 	xfs_extlen_t		reservation; /* blocks that are still reserved */
2289 	int			available;
2290 	xfs_extlen_t		agflcount;
2291 
2292 	if (flags & XFS_ALLOC_FLAG_FREEING)
2293 		return true;
2294 
2295 	reservation = xfs_ag_resv_needed(pag, args->resv);
2296 
2297 	/* do we have enough contiguous free space for the allocation? */
2298 	alloc_len = args->minlen + (args->alignment - 1) + args->minalignslop;
2299 	longest = xfs_alloc_longest_free_extent(pag, min_free, reservation);
2300 	if (longest < alloc_len)
2301 		return false;
2302 
2303 	/*
2304 	 * Do we have enough free space remaining for the allocation? Don't
2305 	 * account extra agfl blocks because we are about to defer free them,
2306 	 * making them unavailable until the current transaction commits.
2307 	 */
2308 	agflcount = min_t(xfs_extlen_t, pag->pagf_flcount, min_free);
2309 	available = (int)(pag->pagf_freeblks + agflcount -
2310 			  reservation - min_free - args->minleft);
2311 	if (available < (int)max(args->total, alloc_len))
2312 		return false;
2313 
2314 	/*
2315 	 * Clamp maxlen to the amount of free space available for the actual
2316 	 * extent allocation.
2317 	 */
2318 	if (available < (int)args->maxlen && !(flags & XFS_ALLOC_FLAG_CHECK)) {
2319 		args->maxlen = available;
2320 		ASSERT(args->maxlen > 0);
2321 		ASSERT(args->maxlen >= args->minlen);
2322 	}
2323 
2324 	return true;
2325 }
2326 
2327 int
2328 xfs_free_agfl_block(
2329 	struct xfs_trans	*tp,
2330 	xfs_agnumber_t		agno,
2331 	xfs_agblock_t		agbno,
2332 	struct xfs_buf		*agbp,
2333 	struct xfs_owner_info	*oinfo)
2334 {
2335 	int			error;
2336 	struct xfs_buf		*bp;
2337 
2338 	error = xfs_free_ag_extent(tp, agbp, agno, agbno, 1, oinfo,
2339 				   XFS_AG_RESV_AGFL);
2340 	if (error)
2341 		return error;
2342 
2343 	error = xfs_trans_get_buf(tp, tp->t_mountp->m_ddev_targp,
2344 			XFS_AGB_TO_DADDR(tp->t_mountp, agno, agbno),
2345 			tp->t_mountp->m_bsize, 0, &bp);
2346 	if (error)
2347 		return error;
2348 	xfs_trans_binval(tp, bp);
2349 
2350 	return 0;
2351 }
2352 
2353 /*
2354  * Check the agfl fields of the agf for inconsistency or corruption. The purpose
2355  * is to detect an agfl header padding mismatch between current and early v5
2356  * kernels. This problem manifests as a 1-slot size difference between the
2357  * on-disk flcount and the active [first, last] range of a wrapped agfl. This
2358  * may also catch variants of agfl count corruption unrelated to padding. Either
2359  * way, we'll reset the agfl and warn the user.
2360  *
2361  * Return true if a reset is required before the agfl can be used, false
2362  * otherwise.
2363  */
2364 static bool
2365 xfs_agfl_needs_reset(
2366 	struct xfs_mount	*mp,
2367 	struct xfs_agf		*agf)
2368 {
2369 	uint32_t		f = be32_to_cpu(agf->agf_flfirst);
2370 	uint32_t		l = be32_to_cpu(agf->agf_fllast);
2371 	uint32_t		c = be32_to_cpu(agf->agf_flcount);
2372 	int			agfl_size = xfs_agfl_size(mp);
2373 	int			active;
2374 
2375 	/* no agfl header on v4 supers */
2376 	if (!xfs_sb_version_hascrc(&mp->m_sb))
2377 		return false;
2378 
2379 	/*
2380 	 * The agf read verifier catches severe corruption of these fields.
2381 	 * Repeat some sanity checks to cover a packed -> unpacked mismatch if
2382 	 * the verifier allows it.
2383 	 */
2384 	if (f >= agfl_size || l >= agfl_size)
2385 		return true;
2386 	if (c > agfl_size)
2387 		return true;
2388 
2389 	/*
2390 	 * Check consistency between the on-disk count and the active range. An
2391 	 * agfl padding mismatch manifests as an inconsistent flcount.
2392 	 */
2393 	if (c && l >= f)
2394 		active = l - f + 1;
2395 	else if (c)
2396 		active = agfl_size - f + l + 1;
2397 	else
2398 		active = 0;
2399 
2400 	return active != c;
2401 }
2402 
2403 /*
2404  * Reset the agfl to an empty state. Ignore/drop any existing blocks since the
2405  * agfl content cannot be trusted. Warn the user that a repair is required to
2406  * recover leaked blocks.
2407  *
2408  * The purpose of this mechanism is to handle filesystems affected by the agfl
2409  * header padding mismatch problem. A reset keeps the filesystem online with a
2410  * relatively minor free space accounting inconsistency rather than suffer the
2411  * inevitable crash from use of an invalid agfl block.
2412  */
2413 static void
2414 xfs_agfl_reset(
2415 	struct xfs_trans	*tp,
2416 	struct xfs_buf		*agbp,
2417 	struct xfs_perag	*pag)
2418 {
2419 	struct xfs_mount	*mp = tp->t_mountp;
2420 	struct xfs_agf		*agf = agbp->b_addr;
2421 
2422 	ASSERT(pag->pagf_agflreset);
2423 	trace_xfs_agfl_reset(mp, agf, 0, _RET_IP_);
2424 
2425 	xfs_warn(mp,
2426 	       "WARNING: Reset corrupted AGFL on AG %u. %d blocks leaked. "
2427 	       "Please unmount and run xfs_repair.",
2428 	         pag->pag_agno, pag->pagf_flcount);
2429 
2430 	agf->agf_flfirst = 0;
2431 	agf->agf_fllast = cpu_to_be32(xfs_agfl_size(mp) - 1);
2432 	agf->agf_flcount = 0;
2433 	xfs_alloc_log_agf(tp, agbp, XFS_AGF_FLFIRST | XFS_AGF_FLLAST |
2434 				    XFS_AGF_FLCOUNT);
2435 
2436 	pag->pagf_flcount = 0;
2437 	pag->pagf_agflreset = false;
2438 }
2439 
2440 /*
2441  * Defer an AGFL block free. This is effectively equivalent to
2442  * xfs_bmap_add_free() with some special handling particular to AGFL blocks.
2443  *
2444  * Deferring AGFL frees helps prevent log reservation overruns due to too many
2445  * allocation operations in a transaction. AGFL frees are prone to this problem
2446  * because for one they are always freed one at a time. Further, an immediate
2447  * AGFL block free can cause a btree join and require another block free before
2448  * the real allocation can proceed. Deferring the free disconnects freeing up
2449  * the AGFL slot from freeing the block.
2450  */
2451 STATIC void
2452 xfs_defer_agfl_block(
2453 	struct xfs_trans		*tp,
2454 	xfs_agnumber_t			agno,
2455 	xfs_fsblock_t			agbno,
2456 	struct xfs_owner_info		*oinfo)
2457 {
2458 	struct xfs_mount		*mp = tp->t_mountp;
2459 	struct xfs_extent_free_item	*new;		/* new element */
2460 
2461 	ASSERT(xfs_bmap_free_item_zone != NULL);
2462 	ASSERT(oinfo != NULL);
2463 
2464 	new = kmem_cache_alloc(xfs_bmap_free_item_zone,
2465 			       GFP_KERNEL | __GFP_NOFAIL);
2466 	new->xefi_startblock = XFS_AGB_TO_FSB(mp, agno, agbno);
2467 	new->xefi_blockcount = 1;
2468 	new->xefi_oinfo = *oinfo;
2469 	new->xefi_skip_discard = false;
2470 
2471 	trace_xfs_agfl_free_defer(mp, agno, 0, agbno, 1);
2472 
2473 	xfs_defer_add(tp, XFS_DEFER_OPS_TYPE_AGFL_FREE, &new->xefi_list);
2474 }
2475 
2476 #ifdef DEBUG
2477 /*
2478  * Check if an AGF has a free extent record whose length is equal to
2479  * args->minlen.
2480  */
2481 STATIC int
2482 xfs_exact_minlen_extent_available(
2483 	struct xfs_alloc_arg	*args,
2484 	struct xfs_buf		*agbp,
2485 	int			*stat)
2486 {
2487 	struct xfs_btree_cur	*cnt_cur;
2488 	xfs_agblock_t		fbno;
2489 	xfs_extlen_t		flen;
2490 	int			error = 0;
2491 
2492 	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, agbp,
2493 			args->agno, XFS_BTNUM_CNT);
2494 	error = xfs_alloc_lookup_ge(cnt_cur, 0, args->minlen, stat);
2495 	if (error)
2496 		goto out;
2497 
2498 	if (*stat == 0) {
2499 		error = -EFSCORRUPTED;
2500 		goto out;
2501 	}
2502 
2503 	error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen, stat);
2504 	if (error)
2505 		goto out;
2506 
2507 	if (*stat == 1 && flen != args->minlen)
2508 		*stat = 0;
2509 
2510 out:
2511 	xfs_btree_del_cursor(cnt_cur, error);
2512 
2513 	return error;
2514 }
2515 #endif
2516 
2517 /*
2518  * Decide whether to use this allocation group for this allocation.
2519  * If so, fix up the btree freelist's size.
2520  */
2521 int			/* error */
2522 xfs_alloc_fix_freelist(
2523 	struct xfs_alloc_arg	*args,	/* allocation argument structure */
2524 	int			flags)	/* XFS_ALLOC_FLAG_... */
2525 {
2526 	struct xfs_mount	*mp = args->mp;
2527 	struct xfs_perag	*pag = args->pag;
2528 	struct xfs_trans	*tp = args->tp;
2529 	struct xfs_buf		*agbp = NULL;
2530 	struct xfs_buf		*agflbp = NULL;
2531 	struct xfs_alloc_arg	targs;	/* local allocation arguments */
2532 	xfs_agblock_t		bno;	/* freelist block */
2533 	xfs_extlen_t		need;	/* total blocks needed in freelist */
2534 	int			error = 0;
2535 
2536 	/* deferred ops (AGFL block frees) require permanent transactions */
2537 	ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
2538 
2539 	if (!pag->pagf_init) {
2540 		error = xfs_alloc_read_agf(mp, tp, args->agno, flags, &agbp);
2541 		if (error) {
2542 			/* Couldn't lock the AGF so skip this AG. */
2543 			if (error == -EAGAIN)
2544 				error = 0;
2545 			goto out_no_agbp;
2546 		}
2547 	}
2548 
2549 	/*
2550 	 * If this is a metadata preferred pag and we are user data then try
2551 	 * somewhere else if we are not being asked to try harder at this
2552 	 * point
2553 	 */
2554 	if (pag->pagf_metadata && (args->datatype & XFS_ALLOC_USERDATA) &&
2555 	    (flags & XFS_ALLOC_FLAG_TRYLOCK)) {
2556 		ASSERT(!(flags & XFS_ALLOC_FLAG_FREEING));
2557 		goto out_agbp_relse;
2558 	}
2559 
2560 	need = xfs_alloc_min_freelist(mp, pag);
2561 	if (!xfs_alloc_space_available(args, need, flags |
2562 			XFS_ALLOC_FLAG_CHECK))
2563 		goto out_agbp_relse;
2564 
2565 	/*
2566 	 * Get the a.g. freespace buffer.
2567 	 * Can fail if we're not blocking on locks, and it's held.
2568 	 */
2569 	if (!agbp) {
2570 		error = xfs_alloc_read_agf(mp, tp, args->agno, flags, &agbp);
2571 		if (error) {
2572 			/* Couldn't lock the AGF so skip this AG. */
2573 			if (error == -EAGAIN)
2574 				error = 0;
2575 			goto out_no_agbp;
2576 		}
2577 	}
2578 
2579 	/* reset a padding mismatched agfl before final free space check */
2580 	if (pag->pagf_agflreset)
2581 		xfs_agfl_reset(tp, agbp, pag);
2582 
2583 	/* If there isn't enough total space or single-extent, reject it. */
2584 	need = xfs_alloc_min_freelist(mp, pag);
2585 	if (!xfs_alloc_space_available(args, need, flags))
2586 		goto out_agbp_relse;
2587 
2588 #ifdef DEBUG
2589 	if (args->alloc_minlen_only) {
2590 		int stat;
2591 
2592 		error = xfs_exact_minlen_extent_available(args, agbp, &stat);
2593 		if (error || !stat)
2594 			goto out_agbp_relse;
2595 	}
2596 #endif
2597 	/*
2598 	 * Make the freelist shorter if it's too long.
2599 	 *
2600 	 * Note that from this point onwards, we will always release the agf and
2601 	 * agfl buffers on error. This handles the case where we error out and
2602 	 * the buffers are clean or may not have been joined to the transaction
2603 	 * and hence need to be released manually. If they have been joined to
2604 	 * the transaction, then xfs_trans_brelse() will handle them
2605 	 * appropriately based on the recursion count and dirty state of the
2606 	 * buffer.
2607 	 *
2608 	 * XXX (dgc): When we have lots of free space, does this buy us
2609 	 * anything other than extra overhead when we need to put more blocks
2610 	 * back on the free list? Maybe we should only do this when space is
2611 	 * getting low or the AGFL is more than half full?
2612 	 *
2613 	 * The NOSHRINK flag prevents the AGFL from being shrunk if it's too
2614 	 * big; the NORMAP flag prevents AGFL expand/shrink operations from
2615 	 * updating the rmapbt.  Both flags are used in xfs_repair while we're
2616 	 * rebuilding the rmapbt, and neither are used by the kernel.  They're
2617 	 * both required to ensure that rmaps are correctly recorded for the
2618 	 * regenerated AGFL, bnobt, and cntbt.  See repair/phase5.c and
2619 	 * repair/rmap.c in xfsprogs for details.
2620 	 */
2621 	memset(&targs, 0, sizeof(targs));
2622 	/* struct copy below */
2623 	if (flags & XFS_ALLOC_FLAG_NORMAP)
2624 		targs.oinfo = XFS_RMAP_OINFO_SKIP_UPDATE;
2625 	else
2626 		targs.oinfo = XFS_RMAP_OINFO_AG;
2627 	while (!(flags & XFS_ALLOC_FLAG_NOSHRINK) && pag->pagf_flcount > need) {
2628 		error = xfs_alloc_get_freelist(tp, agbp, &bno, 0);
2629 		if (error)
2630 			goto out_agbp_relse;
2631 
2632 		/* defer agfl frees */
2633 		xfs_defer_agfl_block(tp, args->agno, bno, &targs.oinfo);
2634 	}
2635 
2636 	targs.tp = tp;
2637 	targs.mp = mp;
2638 	targs.agbp = agbp;
2639 	targs.agno = args->agno;
2640 	targs.alignment = targs.minlen = targs.prod = 1;
2641 	targs.type = XFS_ALLOCTYPE_THIS_AG;
2642 	targs.pag = pag;
2643 	error = xfs_alloc_read_agfl(mp, tp, targs.agno, &agflbp);
2644 	if (error)
2645 		goto out_agbp_relse;
2646 
2647 	/* Make the freelist longer if it's too short. */
2648 	while (pag->pagf_flcount < need) {
2649 		targs.agbno = 0;
2650 		targs.maxlen = need - pag->pagf_flcount;
2651 		targs.resv = XFS_AG_RESV_AGFL;
2652 
2653 		/* Allocate as many blocks as possible at once. */
2654 		error = xfs_alloc_ag_vextent(&targs);
2655 		if (error)
2656 			goto out_agflbp_relse;
2657 
2658 		/*
2659 		 * Stop if we run out.  Won't happen if callers are obeying
2660 		 * the restrictions correctly.  Can happen for free calls
2661 		 * on a completely full ag.
2662 		 */
2663 		if (targs.agbno == NULLAGBLOCK) {
2664 			if (flags & XFS_ALLOC_FLAG_FREEING)
2665 				break;
2666 			goto out_agflbp_relse;
2667 		}
2668 		/*
2669 		 * Put each allocated block on the list.
2670 		 */
2671 		for (bno = targs.agbno; bno < targs.agbno + targs.len; bno++) {
2672 			error = xfs_alloc_put_freelist(tp, agbp,
2673 							agflbp, bno, 0);
2674 			if (error)
2675 				goto out_agflbp_relse;
2676 		}
2677 	}
2678 	xfs_trans_brelse(tp, agflbp);
2679 	args->agbp = agbp;
2680 	return 0;
2681 
2682 out_agflbp_relse:
2683 	xfs_trans_brelse(tp, agflbp);
2684 out_agbp_relse:
2685 	if (agbp)
2686 		xfs_trans_brelse(tp, agbp);
2687 out_no_agbp:
2688 	args->agbp = NULL;
2689 	return error;
2690 }
2691 
2692 /*
2693  * Get a block from the freelist.
2694  * Returns with the buffer for the block gotten.
2695  */
2696 int				/* error */
2697 xfs_alloc_get_freelist(
2698 	xfs_trans_t	*tp,	/* transaction pointer */
2699 	struct xfs_buf	*agbp,	/* buffer containing the agf structure */
2700 	xfs_agblock_t	*bnop,	/* block address retrieved from freelist */
2701 	int		btreeblk) /* destination is a AGF btree */
2702 {
2703 	struct xfs_agf	*agf = agbp->b_addr;
2704 	struct xfs_buf	*agflbp;/* buffer for a.g. freelist structure */
2705 	xfs_agblock_t	bno;	/* block number returned */
2706 	__be32		*agfl_bno;
2707 	int		error;
2708 	int		logflags;
2709 	xfs_mount_t	*mp = tp->t_mountp;
2710 	xfs_perag_t	*pag;	/* per allocation group data */
2711 
2712 	/*
2713 	 * Freelist is empty, give up.
2714 	 */
2715 	if (!agf->agf_flcount) {
2716 		*bnop = NULLAGBLOCK;
2717 		return 0;
2718 	}
2719 	/*
2720 	 * Read the array of free blocks.
2721 	 */
2722 	error = xfs_alloc_read_agfl(mp, tp, be32_to_cpu(agf->agf_seqno),
2723 				    &agflbp);
2724 	if (error)
2725 		return error;
2726 
2727 
2728 	/*
2729 	 * Get the block number and update the data structures.
2730 	 */
2731 	agfl_bno = xfs_buf_to_agfl_bno(agflbp);
2732 	bno = be32_to_cpu(agfl_bno[be32_to_cpu(agf->agf_flfirst)]);
2733 	be32_add_cpu(&agf->agf_flfirst, 1);
2734 	xfs_trans_brelse(tp, agflbp);
2735 	if (be32_to_cpu(agf->agf_flfirst) == xfs_agfl_size(mp))
2736 		agf->agf_flfirst = 0;
2737 
2738 	pag = agbp->b_pag;
2739 	ASSERT(!pag->pagf_agflreset);
2740 	be32_add_cpu(&agf->agf_flcount, -1);
2741 	pag->pagf_flcount--;
2742 
2743 	logflags = XFS_AGF_FLFIRST | XFS_AGF_FLCOUNT;
2744 	if (btreeblk) {
2745 		be32_add_cpu(&agf->agf_btreeblks, 1);
2746 		pag->pagf_btreeblks++;
2747 		logflags |= XFS_AGF_BTREEBLKS;
2748 	}
2749 
2750 	xfs_alloc_log_agf(tp, agbp, logflags);
2751 	*bnop = bno;
2752 
2753 	return 0;
2754 }
2755 
2756 /*
2757  * Log the given fields from the agf structure.
2758  */
2759 void
2760 xfs_alloc_log_agf(
2761 	xfs_trans_t	*tp,	/* transaction pointer */
2762 	struct xfs_buf	*bp,	/* buffer for a.g. freelist header */
2763 	int		fields)	/* mask of fields to be logged (XFS_AGF_...) */
2764 {
2765 	int	first;		/* first byte offset */
2766 	int	last;		/* last byte offset */
2767 	static const short	offsets[] = {
2768 		offsetof(xfs_agf_t, agf_magicnum),
2769 		offsetof(xfs_agf_t, agf_versionnum),
2770 		offsetof(xfs_agf_t, agf_seqno),
2771 		offsetof(xfs_agf_t, agf_length),
2772 		offsetof(xfs_agf_t, agf_roots[0]),
2773 		offsetof(xfs_agf_t, agf_levels[0]),
2774 		offsetof(xfs_agf_t, agf_flfirst),
2775 		offsetof(xfs_agf_t, agf_fllast),
2776 		offsetof(xfs_agf_t, agf_flcount),
2777 		offsetof(xfs_agf_t, agf_freeblks),
2778 		offsetof(xfs_agf_t, agf_longest),
2779 		offsetof(xfs_agf_t, agf_btreeblks),
2780 		offsetof(xfs_agf_t, agf_uuid),
2781 		offsetof(xfs_agf_t, agf_rmap_blocks),
2782 		offsetof(xfs_agf_t, agf_refcount_blocks),
2783 		offsetof(xfs_agf_t, agf_refcount_root),
2784 		offsetof(xfs_agf_t, agf_refcount_level),
2785 		/* needed so that we don't log the whole rest of the structure: */
2786 		offsetof(xfs_agf_t, agf_spare64),
2787 		sizeof(xfs_agf_t)
2788 	};
2789 
2790 	trace_xfs_agf(tp->t_mountp, bp->b_addr, fields, _RET_IP_);
2791 
2792 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_AGF_BUF);
2793 
2794 	xfs_btree_offsets(fields, offsets, XFS_AGF_NUM_BITS, &first, &last);
2795 	xfs_trans_log_buf(tp, bp, (uint)first, (uint)last);
2796 }
2797 
2798 /*
2799  * Interface for inode allocation to force the pag data to be initialized.
2800  */
2801 int					/* error */
2802 xfs_alloc_pagf_init(
2803 	xfs_mount_t		*mp,	/* file system mount structure */
2804 	xfs_trans_t		*tp,	/* transaction pointer */
2805 	xfs_agnumber_t		agno,	/* allocation group number */
2806 	int			flags)	/* XFS_ALLOC_FLAGS_... */
2807 {
2808 	struct xfs_buf		*bp;
2809 	int			error;
2810 
2811 	error = xfs_alloc_read_agf(mp, tp, agno, flags, &bp);
2812 	if (!error)
2813 		xfs_trans_brelse(tp, bp);
2814 	return error;
2815 }
2816 
2817 /*
2818  * Put the block on the freelist for the allocation group.
2819  */
2820 int					/* error */
2821 xfs_alloc_put_freelist(
2822 	xfs_trans_t		*tp,	/* transaction pointer */
2823 	struct xfs_buf		*agbp,	/* buffer for a.g. freelist header */
2824 	struct xfs_buf		*agflbp,/* buffer for a.g. free block array */
2825 	xfs_agblock_t		bno,	/* block being freed */
2826 	int			btreeblk) /* block came from a AGF btree */
2827 {
2828 	struct xfs_mount	*mp = tp->t_mountp;
2829 	struct xfs_agf		*agf = agbp->b_addr;
2830 	__be32			*blockp;/* pointer to array entry */
2831 	int			error;
2832 	int			logflags;
2833 	xfs_perag_t		*pag;	/* per allocation group data */
2834 	__be32			*agfl_bno;
2835 	int			startoff;
2836 
2837 	if (!agflbp && (error = xfs_alloc_read_agfl(mp, tp,
2838 			be32_to_cpu(agf->agf_seqno), &agflbp)))
2839 		return error;
2840 	be32_add_cpu(&agf->agf_fllast, 1);
2841 	if (be32_to_cpu(agf->agf_fllast) == xfs_agfl_size(mp))
2842 		agf->agf_fllast = 0;
2843 
2844 	pag = agbp->b_pag;
2845 	ASSERT(!pag->pagf_agflreset);
2846 	be32_add_cpu(&agf->agf_flcount, 1);
2847 	pag->pagf_flcount++;
2848 
2849 	logflags = XFS_AGF_FLLAST | XFS_AGF_FLCOUNT;
2850 	if (btreeblk) {
2851 		be32_add_cpu(&agf->agf_btreeblks, -1);
2852 		pag->pagf_btreeblks--;
2853 		logflags |= XFS_AGF_BTREEBLKS;
2854 	}
2855 
2856 	xfs_alloc_log_agf(tp, agbp, logflags);
2857 
2858 	ASSERT(be32_to_cpu(agf->agf_flcount) <= xfs_agfl_size(mp));
2859 
2860 	agfl_bno = xfs_buf_to_agfl_bno(agflbp);
2861 	blockp = &agfl_bno[be32_to_cpu(agf->agf_fllast)];
2862 	*blockp = cpu_to_be32(bno);
2863 	startoff = (char *)blockp - (char *)agflbp->b_addr;
2864 
2865 	xfs_alloc_log_agf(tp, agbp, logflags);
2866 
2867 	xfs_trans_buf_set_type(tp, agflbp, XFS_BLFT_AGFL_BUF);
2868 	xfs_trans_log_buf(tp, agflbp, startoff,
2869 			  startoff + sizeof(xfs_agblock_t) - 1);
2870 	return 0;
2871 }
2872 
2873 static xfs_failaddr_t
2874 xfs_agf_verify(
2875 	struct xfs_buf		*bp)
2876 {
2877 	struct xfs_mount	*mp = bp->b_mount;
2878 	struct xfs_agf		*agf = bp->b_addr;
2879 
2880 	if (xfs_sb_version_hascrc(&mp->m_sb)) {
2881 		if (!uuid_equal(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid))
2882 			return __this_address;
2883 		if (!xfs_log_check_lsn(mp, be64_to_cpu(agf->agf_lsn)))
2884 			return __this_address;
2885 	}
2886 
2887 	if (!xfs_verify_magic(bp, agf->agf_magicnum))
2888 		return __this_address;
2889 
2890 	if (!(XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
2891 	      be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
2892 	      be32_to_cpu(agf->agf_flfirst) < xfs_agfl_size(mp) &&
2893 	      be32_to_cpu(agf->agf_fllast) < xfs_agfl_size(mp) &&
2894 	      be32_to_cpu(agf->agf_flcount) <= xfs_agfl_size(mp)))
2895 		return __this_address;
2896 
2897 	if (be32_to_cpu(agf->agf_length) > mp->m_sb.sb_dblocks)
2898 		return __this_address;
2899 
2900 	if (be32_to_cpu(agf->agf_freeblks) < be32_to_cpu(agf->agf_longest) ||
2901 	    be32_to_cpu(agf->agf_freeblks) > be32_to_cpu(agf->agf_length))
2902 		return __this_address;
2903 
2904 	if (be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]) < 1 ||
2905 	    be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]) < 1 ||
2906 	    be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]) > mp->m_ag_maxlevels ||
2907 	    be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]) > mp->m_ag_maxlevels)
2908 		return __this_address;
2909 
2910 	if (xfs_sb_version_hasrmapbt(&mp->m_sb) &&
2911 	    (be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) < 1 ||
2912 	     be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]) > mp->m_rmap_maxlevels))
2913 		return __this_address;
2914 
2915 	if (xfs_sb_version_hasrmapbt(&mp->m_sb) &&
2916 	    be32_to_cpu(agf->agf_rmap_blocks) > be32_to_cpu(agf->agf_length))
2917 		return __this_address;
2918 
2919 	/*
2920 	 * during growfs operations, the perag is not fully initialised,
2921 	 * so we can't use it for any useful checking. growfs ensures we can't
2922 	 * use it by using uncached buffers that don't have the perag attached
2923 	 * so we can detect and avoid this problem.
2924 	 */
2925 	if (bp->b_pag && be32_to_cpu(agf->agf_seqno) != bp->b_pag->pag_agno)
2926 		return __this_address;
2927 
2928 	if (xfs_sb_version_haslazysbcount(&mp->m_sb) &&
2929 	    be32_to_cpu(agf->agf_btreeblks) > be32_to_cpu(agf->agf_length))
2930 		return __this_address;
2931 
2932 	if (xfs_sb_version_hasreflink(&mp->m_sb) &&
2933 	    be32_to_cpu(agf->agf_refcount_blocks) >
2934 	    be32_to_cpu(agf->agf_length))
2935 		return __this_address;
2936 
2937 	if (xfs_sb_version_hasreflink(&mp->m_sb) &&
2938 	    (be32_to_cpu(agf->agf_refcount_level) < 1 ||
2939 	     be32_to_cpu(agf->agf_refcount_level) > mp->m_refc_maxlevels))
2940 		return __this_address;
2941 
2942 	return NULL;
2943 
2944 }
2945 
2946 static void
2947 xfs_agf_read_verify(
2948 	struct xfs_buf	*bp)
2949 {
2950 	struct xfs_mount *mp = bp->b_mount;
2951 	xfs_failaddr_t	fa;
2952 
2953 	if (xfs_sb_version_hascrc(&mp->m_sb) &&
2954 	    !xfs_buf_verify_cksum(bp, XFS_AGF_CRC_OFF))
2955 		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
2956 	else {
2957 		fa = xfs_agf_verify(bp);
2958 		if (XFS_TEST_ERROR(fa, mp, XFS_ERRTAG_ALLOC_READ_AGF))
2959 			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
2960 	}
2961 }
2962 
2963 static void
2964 xfs_agf_write_verify(
2965 	struct xfs_buf	*bp)
2966 {
2967 	struct xfs_mount	*mp = bp->b_mount;
2968 	struct xfs_buf_log_item	*bip = bp->b_log_item;
2969 	struct xfs_agf		*agf = bp->b_addr;
2970 	xfs_failaddr_t		fa;
2971 
2972 	fa = xfs_agf_verify(bp);
2973 	if (fa) {
2974 		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
2975 		return;
2976 	}
2977 
2978 	if (!xfs_sb_version_hascrc(&mp->m_sb))
2979 		return;
2980 
2981 	if (bip)
2982 		agf->agf_lsn = cpu_to_be64(bip->bli_item.li_lsn);
2983 
2984 	xfs_buf_update_cksum(bp, XFS_AGF_CRC_OFF);
2985 }
2986 
2987 const struct xfs_buf_ops xfs_agf_buf_ops = {
2988 	.name = "xfs_agf",
2989 	.magic = { cpu_to_be32(XFS_AGF_MAGIC), cpu_to_be32(XFS_AGF_MAGIC) },
2990 	.verify_read = xfs_agf_read_verify,
2991 	.verify_write = xfs_agf_write_verify,
2992 	.verify_struct = xfs_agf_verify,
2993 };
2994 
2995 /*
2996  * Read in the allocation group header (free/alloc section).
2997  */
2998 int					/* error */
2999 xfs_read_agf(
3000 	struct xfs_mount	*mp,	/* mount point structure */
3001 	struct xfs_trans	*tp,	/* transaction pointer */
3002 	xfs_agnumber_t		agno,	/* allocation group number */
3003 	int			flags,	/* XFS_BUF_ */
3004 	struct xfs_buf		**bpp)	/* buffer for the ag freelist header */
3005 {
3006 	int		error;
3007 
3008 	trace_xfs_read_agf(mp, agno);
3009 
3010 	ASSERT(agno != NULLAGNUMBER);
3011 	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
3012 			XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
3013 			XFS_FSS_TO_BB(mp, 1), flags, bpp, &xfs_agf_buf_ops);
3014 	if (error)
3015 		return error;
3016 
3017 	ASSERT(!(*bpp)->b_error);
3018 	xfs_buf_set_ref(*bpp, XFS_AGF_REF);
3019 	return 0;
3020 }
3021 
3022 /*
3023  * Read in the allocation group header (free/alloc section).
3024  */
3025 int					/* error */
3026 xfs_alloc_read_agf(
3027 	struct xfs_mount	*mp,	/* mount point structure */
3028 	struct xfs_trans	*tp,	/* transaction pointer */
3029 	xfs_agnumber_t		agno,	/* allocation group number */
3030 	int			flags,	/* XFS_ALLOC_FLAG_... */
3031 	struct xfs_buf		**bpp)	/* buffer for the ag freelist header */
3032 {
3033 	struct xfs_agf		*agf;		/* ag freelist header */
3034 	struct xfs_perag	*pag;		/* per allocation group data */
3035 	int			error;
3036 	int			allocbt_blks;
3037 
3038 	trace_xfs_alloc_read_agf(mp, agno);
3039 
3040 	/* We don't support trylock when freeing. */
3041 	ASSERT((flags & (XFS_ALLOC_FLAG_FREEING | XFS_ALLOC_FLAG_TRYLOCK)) !=
3042 			(XFS_ALLOC_FLAG_FREEING | XFS_ALLOC_FLAG_TRYLOCK));
3043 	ASSERT(agno != NULLAGNUMBER);
3044 	error = xfs_read_agf(mp, tp, agno,
3045 			(flags & XFS_ALLOC_FLAG_TRYLOCK) ? XBF_TRYLOCK : 0,
3046 			bpp);
3047 	if (error)
3048 		return error;
3049 	ASSERT(!(*bpp)->b_error);
3050 
3051 	agf = (*bpp)->b_addr;
3052 	pag = (*bpp)->b_pag;
3053 	if (!pag->pagf_init) {
3054 		pag->pagf_freeblks = be32_to_cpu(agf->agf_freeblks);
3055 		pag->pagf_btreeblks = be32_to_cpu(agf->agf_btreeblks);
3056 		pag->pagf_flcount = be32_to_cpu(agf->agf_flcount);
3057 		pag->pagf_longest = be32_to_cpu(agf->agf_longest);
3058 		pag->pagf_levels[XFS_BTNUM_BNOi] =
3059 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]);
3060 		pag->pagf_levels[XFS_BTNUM_CNTi] =
3061 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]);
3062 		pag->pagf_levels[XFS_BTNUM_RMAPi] =
3063 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
3064 		pag->pagf_refcount_level = be32_to_cpu(agf->agf_refcount_level);
3065 		pag->pagf_init = 1;
3066 		pag->pagf_agflreset = xfs_agfl_needs_reset(mp, agf);
3067 
3068 		/*
3069 		 * Update the in-core allocbt counter. Filter out the rmapbt
3070 		 * subset of the btreeblks counter because the rmapbt is managed
3071 		 * by perag reservation. Subtract one for the rmapbt root block
3072 		 * because the rmap counter includes it while the btreeblks
3073 		 * counter only tracks non-root blocks.
3074 		 */
3075 		allocbt_blks = pag->pagf_btreeblks;
3076 		if (xfs_sb_version_hasrmapbt(&mp->m_sb))
3077 			allocbt_blks -= be32_to_cpu(agf->agf_rmap_blocks) - 1;
3078 		if (allocbt_blks > 0)
3079 			atomic64_add(allocbt_blks, &mp->m_allocbt_blks);
3080 	}
3081 #ifdef DEBUG
3082 	else if (!XFS_FORCED_SHUTDOWN(mp)) {
3083 		ASSERT(pag->pagf_freeblks == be32_to_cpu(agf->agf_freeblks));
3084 		ASSERT(pag->pagf_btreeblks == be32_to_cpu(agf->agf_btreeblks));
3085 		ASSERT(pag->pagf_flcount == be32_to_cpu(agf->agf_flcount));
3086 		ASSERT(pag->pagf_longest == be32_to_cpu(agf->agf_longest));
3087 		ASSERT(pag->pagf_levels[XFS_BTNUM_BNOi] ==
3088 		       be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]));
3089 		ASSERT(pag->pagf_levels[XFS_BTNUM_CNTi] ==
3090 		       be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]));
3091 	}
3092 #endif
3093 	return 0;
3094 }
3095 
3096 /*
3097  * Allocate an extent (variable-size).
3098  * Depending on the allocation type, we either look in a single allocation
3099  * group or loop over the allocation groups to find the result.
3100  */
3101 int				/* error */
3102 xfs_alloc_vextent(
3103 	struct xfs_alloc_arg	*args)	/* allocation argument structure */
3104 {
3105 	xfs_agblock_t		agsize;	/* allocation group size */
3106 	int			error;
3107 	int			flags;	/* XFS_ALLOC_FLAG_... locking flags */
3108 	struct xfs_mount	*mp;	/* mount structure pointer */
3109 	xfs_agnumber_t		sagno;	/* starting allocation group number */
3110 	xfs_alloctype_t		type;	/* input allocation type */
3111 	int			bump_rotor = 0;
3112 	xfs_agnumber_t		rotorstep = xfs_rotorstep; /* inode32 agf stepper */
3113 
3114 	mp = args->mp;
3115 	type = args->otype = args->type;
3116 	args->agbno = NULLAGBLOCK;
3117 	/*
3118 	 * Just fix this up, for the case where the last a.g. is shorter
3119 	 * (or there's only one a.g.) and the caller couldn't easily figure
3120 	 * that out (xfs_bmap_alloc).
3121 	 */
3122 	agsize = mp->m_sb.sb_agblocks;
3123 	if (args->maxlen > agsize)
3124 		args->maxlen = agsize;
3125 	if (args->alignment == 0)
3126 		args->alignment = 1;
3127 	ASSERT(XFS_FSB_TO_AGNO(mp, args->fsbno) < mp->m_sb.sb_agcount);
3128 	ASSERT(XFS_FSB_TO_AGBNO(mp, args->fsbno) < agsize);
3129 	ASSERT(args->minlen <= args->maxlen);
3130 	ASSERT(args->minlen <= agsize);
3131 	ASSERT(args->mod < args->prod);
3132 	if (XFS_FSB_TO_AGNO(mp, args->fsbno) >= mp->m_sb.sb_agcount ||
3133 	    XFS_FSB_TO_AGBNO(mp, args->fsbno) >= agsize ||
3134 	    args->minlen > args->maxlen || args->minlen > agsize ||
3135 	    args->mod >= args->prod) {
3136 		args->fsbno = NULLFSBLOCK;
3137 		trace_xfs_alloc_vextent_badargs(args);
3138 		return 0;
3139 	}
3140 
3141 	switch (type) {
3142 	case XFS_ALLOCTYPE_THIS_AG:
3143 	case XFS_ALLOCTYPE_NEAR_BNO:
3144 	case XFS_ALLOCTYPE_THIS_BNO:
3145 		/*
3146 		 * These three force us into a single a.g.
3147 		 */
3148 		args->agno = XFS_FSB_TO_AGNO(mp, args->fsbno);
3149 		args->pag = xfs_perag_get(mp, args->agno);
3150 		error = xfs_alloc_fix_freelist(args, 0);
3151 		if (error) {
3152 			trace_xfs_alloc_vextent_nofix(args);
3153 			goto error0;
3154 		}
3155 		if (!args->agbp) {
3156 			trace_xfs_alloc_vextent_noagbp(args);
3157 			break;
3158 		}
3159 		args->agbno = XFS_FSB_TO_AGBNO(mp, args->fsbno);
3160 		if ((error = xfs_alloc_ag_vextent(args)))
3161 			goto error0;
3162 		break;
3163 	case XFS_ALLOCTYPE_START_BNO:
3164 		/*
3165 		 * Try near allocation first, then anywhere-in-ag after
3166 		 * the first a.g. fails.
3167 		 */
3168 		if ((args->datatype & XFS_ALLOC_INITIAL_USER_DATA) &&
3169 		    (mp->m_flags & XFS_MOUNT_32BITINODES)) {
3170 			args->fsbno = XFS_AGB_TO_FSB(mp,
3171 					((mp->m_agfrotor / rotorstep) %
3172 					mp->m_sb.sb_agcount), 0);
3173 			bump_rotor = 1;
3174 		}
3175 		args->agbno = XFS_FSB_TO_AGBNO(mp, args->fsbno);
3176 		args->type = XFS_ALLOCTYPE_NEAR_BNO;
3177 		/* FALLTHROUGH */
3178 	case XFS_ALLOCTYPE_FIRST_AG:
3179 		/*
3180 		 * Rotate through the allocation groups looking for a winner.
3181 		 */
3182 		if (type == XFS_ALLOCTYPE_FIRST_AG) {
3183 			/*
3184 			 * Start with allocation group given by bno.
3185 			 */
3186 			args->agno = XFS_FSB_TO_AGNO(mp, args->fsbno);
3187 			args->type = XFS_ALLOCTYPE_THIS_AG;
3188 			sagno = 0;
3189 			flags = 0;
3190 		} else {
3191 			/*
3192 			 * Start with the given allocation group.
3193 			 */
3194 			args->agno = sagno = XFS_FSB_TO_AGNO(mp, args->fsbno);
3195 			flags = XFS_ALLOC_FLAG_TRYLOCK;
3196 		}
3197 		/*
3198 		 * Loop over allocation groups twice; first time with
3199 		 * trylock set, second time without.
3200 		 */
3201 		for (;;) {
3202 			args->pag = xfs_perag_get(mp, args->agno);
3203 			error = xfs_alloc_fix_freelist(args, flags);
3204 			if (error) {
3205 				trace_xfs_alloc_vextent_nofix(args);
3206 				goto error0;
3207 			}
3208 			/*
3209 			 * If we get a buffer back then the allocation will fly.
3210 			 */
3211 			if (args->agbp) {
3212 				if ((error = xfs_alloc_ag_vextent(args)))
3213 					goto error0;
3214 				break;
3215 			}
3216 
3217 			trace_xfs_alloc_vextent_loopfailed(args);
3218 
3219 			/*
3220 			 * Didn't work, figure out the next iteration.
3221 			 */
3222 			if (args->agno == sagno &&
3223 			    type == XFS_ALLOCTYPE_START_BNO)
3224 				args->type = XFS_ALLOCTYPE_THIS_AG;
3225 			/*
3226 			* For the first allocation, we can try any AG to get
3227 			* space.  However, if we already have allocated a
3228 			* block, we don't want to try AGs whose number is below
3229 			* sagno. Otherwise, we may end up with out-of-order
3230 			* locking of AGF, which might cause deadlock.
3231 			*/
3232 			if (++(args->agno) == mp->m_sb.sb_agcount) {
3233 				if (args->tp->t_firstblock != NULLFSBLOCK)
3234 					args->agno = sagno;
3235 				else
3236 					args->agno = 0;
3237 			}
3238 			/*
3239 			 * Reached the starting a.g., must either be done
3240 			 * or switch to non-trylock mode.
3241 			 */
3242 			if (args->agno == sagno) {
3243 				if (flags == 0) {
3244 					args->agbno = NULLAGBLOCK;
3245 					trace_xfs_alloc_vextent_allfailed(args);
3246 					break;
3247 				}
3248 
3249 				flags = 0;
3250 				if (type == XFS_ALLOCTYPE_START_BNO) {
3251 					args->agbno = XFS_FSB_TO_AGBNO(mp,
3252 						args->fsbno);
3253 					args->type = XFS_ALLOCTYPE_NEAR_BNO;
3254 				}
3255 			}
3256 			xfs_perag_put(args->pag);
3257 		}
3258 		if (bump_rotor) {
3259 			if (args->agno == sagno)
3260 				mp->m_agfrotor = (mp->m_agfrotor + 1) %
3261 					(mp->m_sb.sb_agcount * rotorstep);
3262 			else
3263 				mp->m_agfrotor = (args->agno * rotorstep + 1) %
3264 					(mp->m_sb.sb_agcount * rotorstep);
3265 		}
3266 		break;
3267 	default:
3268 		ASSERT(0);
3269 		/* NOTREACHED */
3270 	}
3271 	if (args->agbno == NULLAGBLOCK)
3272 		args->fsbno = NULLFSBLOCK;
3273 	else {
3274 		args->fsbno = XFS_AGB_TO_FSB(mp, args->agno, args->agbno);
3275 #ifdef DEBUG
3276 		ASSERT(args->len >= args->minlen);
3277 		ASSERT(args->len <= args->maxlen);
3278 		ASSERT(args->agbno % args->alignment == 0);
3279 		XFS_AG_CHECK_DADDR(mp, XFS_FSB_TO_DADDR(mp, args->fsbno),
3280 			args->len);
3281 #endif
3282 
3283 	}
3284 	xfs_perag_put(args->pag);
3285 	return 0;
3286 error0:
3287 	xfs_perag_put(args->pag);
3288 	return error;
3289 }
3290 
3291 /* Ensure that the freelist is at full capacity. */
3292 int
3293 xfs_free_extent_fix_freelist(
3294 	struct xfs_trans	*tp,
3295 	xfs_agnumber_t		agno,
3296 	struct xfs_buf		**agbp)
3297 {
3298 	struct xfs_alloc_arg	args;
3299 	int			error;
3300 
3301 	memset(&args, 0, sizeof(struct xfs_alloc_arg));
3302 	args.tp = tp;
3303 	args.mp = tp->t_mountp;
3304 	args.agno = agno;
3305 
3306 	/*
3307 	 * validate that the block number is legal - the enables us to detect
3308 	 * and handle a silent filesystem corruption rather than crashing.
3309 	 */
3310 	if (args.agno >= args.mp->m_sb.sb_agcount)
3311 		return -EFSCORRUPTED;
3312 
3313 	args.pag = xfs_perag_get(args.mp, args.agno);
3314 	ASSERT(args.pag);
3315 
3316 	error = xfs_alloc_fix_freelist(&args, XFS_ALLOC_FLAG_FREEING);
3317 	if (error)
3318 		goto out;
3319 
3320 	*agbp = args.agbp;
3321 out:
3322 	xfs_perag_put(args.pag);
3323 	return error;
3324 }
3325 
3326 /*
3327  * Free an extent.
3328  * Just break up the extent address and hand off to xfs_free_ag_extent
3329  * after fixing up the freelist.
3330  */
3331 int
3332 __xfs_free_extent(
3333 	struct xfs_trans		*tp,
3334 	xfs_fsblock_t			bno,
3335 	xfs_extlen_t			len,
3336 	const struct xfs_owner_info	*oinfo,
3337 	enum xfs_ag_resv_type		type,
3338 	bool				skip_discard)
3339 {
3340 	struct xfs_mount		*mp = tp->t_mountp;
3341 	struct xfs_buf			*agbp;
3342 	xfs_agnumber_t			agno = XFS_FSB_TO_AGNO(mp, bno);
3343 	xfs_agblock_t			agbno = XFS_FSB_TO_AGBNO(mp, bno);
3344 	struct xfs_agf			*agf;
3345 	int				error;
3346 	unsigned int			busy_flags = 0;
3347 
3348 	ASSERT(len != 0);
3349 	ASSERT(type != XFS_AG_RESV_AGFL);
3350 
3351 	if (XFS_TEST_ERROR(false, mp,
3352 			XFS_ERRTAG_FREE_EXTENT))
3353 		return -EIO;
3354 
3355 	error = xfs_free_extent_fix_freelist(tp, agno, &agbp);
3356 	if (error)
3357 		return error;
3358 	agf = agbp->b_addr;
3359 
3360 	if (XFS_IS_CORRUPT(mp, agbno >= mp->m_sb.sb_agblocks)) {
3361 		error = -EFSCORRUPTED;
3362 		goto err;
3363 	}
3364 
3365 	/* validate the extent size is legal now we have the agf locked */
3366 	if (XFS_IS_CORRUPT(mp, agbno + len > be32_to_cpu(agf->agf_length))) {
3367 		error = -EFSCORRUPTED;
3368 		goto err;
3369 	}
3370 
3371 	error = xfs_free_ag_extent(tp, agbp, agno, agbno, len, oinfo, type);
3372 	if (error)
3373 		goto err;
3374 
3375 	if (skip_discard)
3376 		busy_flags |= XFS_EXTENT_BUSY_SKIP_DISCARD;
3377 	xfs_extent_busy_insert(tp, agno, agbno, len, busy_flags);
3378 	return 0;
3379 
3380 err:
3381 	xfs_trans_brelse(tp, agbp);
3382 	return error;
3383 }
3384 
3385 struct xfs_alloc_query_range_info {
3386 	xfs_alloc_query_range_fn	fn;
3387 	void				*priv;
3388 };
3389 
3390 /* Format btree record and pass to our callback. */
3391 STATIC int
3392 xfs_alloc_query_range_helper(
3393 	struct xfs_btree_cur		*cur,
3394 	union xfs_btree_rec		*rec,
3395 	void				*priv)
3396 {
3397 	struct xfs_alloc_query_range_info	*query = priv;
3398 	struct xfs_alloc_rec_incore		irec;
3399 
3400 	irec.ar_startblock = be32_to_cpu(rec->alloc.ar_startblock);
3401 	irec.ar_blockcount = be32_to_cpu(rec->alloc.ar_blockcount);
3402 	return query->fn(cur, &irec, query->priv);
3403 }
3404 
3405 /* Find all free space within a given range of blocks. */
3406 int
3407 xfs_alloc_query_range(
3408 	struct xfs_btree_cur			*cur,
3409 	struct xfs_alloc_rec_incore		*low_rec,
3410 	struct xfs_alloc_rec_incore		*high_rec,
3411 	xfs_alloc_query_range_fn		fn,
3412 	void					*priv)
3413 {
3414 	union xfs_btree_irec			low_brec;
3415 	union xfs_btree_irec			high_brec;
3416 	struct xfs_alloc_query_range_info	query;
3417 
3418 	ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
3419 	low_brec.a = *low_rec;
3420 	high_brec.a = *high_rec;
3421 	query.priv = priv;
3422 	query.fn = fn;
3423 	return xfs_btree_query_range(cur, &low_brec, &high_brec,
3424 			xfs_alloc_query_range_helper, &query);
3425 }
3426 
3427 /* Find all free space records. */
3428 int
3429 xfs_alloc_query_all(
3430 	struct xfs_btree_cur			*cur,
3431 	xfs_alloc_query_range_fn		fn,
3432 	void					*priv)
3433 {
3434 	struct xfs_alloc_query_range_info	query;
3435 
3436 	ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
3437 	query.priv = priv;
3438 	query.fn = fn;
3439 	return xfs_btree_query_all(cur, xfs_alloc_query_range_helper, &query);
3440 }
3441 
3442 /* Is there a record covering a given extent? */
3443 int
3444 xfs_alloc_has_record(
3445 	struct xfs_btree_cur	*cur,
3446 	xfs_agblock_t		bno,
3447 	xfs_extlen_t		len,
3448 	bool			*exists)
3449 {
3450 	union xfs_btree_irec	low;
3451 	union xfs_btree_irec	high;
3452 
3453 	memset(&low, 0, sizeof(low));
3454 	low.a.ar_startblock = bno;
3455 	memset(&high, 0xFF, sizeof(high));
3456 	high.a.ar_startblock = bno + len - 1;
3457 
3458 	return xfs_btree_has_record(cur, &low, &high, exists);
3459 }
3460 
3461 /*
3462  * Walk all the blocks in the AGFL.  The @walk_fn can return any negative
3463  * error code or XFS_ITER_*.
3464  */
3465 int
3466 xfs_agfl_walk(
3467 	struct xfs_mount	*mp,
3468 	struct xfs_agf		*agf,
3469 	struct xfs_buf		*agflbp,
3470 	xfs_agfl_walk_fn	walk_fn,
3471 	void			*priv)
3472 {
3473 	__be32			*agfl_bno;
3474 	unsigned int		i;
3475 	int			error;
3476 
3477 	agfl_bno = xfs_buf_to_agfl_bno(agflbp);
3478 	i = be32_to_cpu(agf->agf_flfirst);
3479 
3480 	/* Nothing to walk in an empty AGFL. */
3481 	if (agf->agf_flcount == cpu_to_be32(0))
3482 		return 0;
3483 
3484 	/* Otherwise, walk from first to last, wrapping as needed. */
3485 	for (;;) {
3486 		error = walk_fn(mp, be32_to_cpu(agfl_bno[i]), priv);
3487 		if (error)
3488 			return error;
3489 		if (i == be32_to_cpu(agf->agf_fllast))
3490 			break;
3491 		if (++i == xfs_agfl_size(mp))
3492 			i = 0;
3493 	}
3494 
3495 	return 0;
3496 }
3497