xref: /openbmc/linux/fs/xfs/libxfs/xfs_bmap_btree.c (revision b5572597)
1 /*
2  * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public License as
7  * published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it would be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  * GNU General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write the Free Software Foundation,
16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17  */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_bit.h"
25 #include "xfs_mount.h"
26 #include "xfs_defer.h"
27 #include "xfs_inode.h"
28 #include "xfs_trans.h"
29 #include "xfs_inode_item.h"
30 #include "xfs_alloc.h"
31 #include "xfs_btree.h"
32 #include "xfs_bmap_btree.h"
33 #include "xfs_bmap.h"
34 #include "xfs_error.h"
35 #include "xfs_quota.h"
36 #include "xfs_trace.h"
37 #include "xfs_cksum.h"
38 #include "xfs_rmap.h"
39 
40 /*
41  * Convert on-disk form of btree root to in-memory form.
42  */
43 void
44 xfs_bmdr_to_bmbt(
45 	struct xfs_inode	*ip,
46 	xfs_bmdr_block_t	*dblock,
47 	int			dblocklen,
48 	struct xfs_btree_block	*rblock,
49 	int			rblocklen)
50 {
51 	struct xfs_mount	*mp = ip->i_mount;
52 	int			dmxr;
53 	xfs_bmbt_key_t		*fkp;
54 	__be64			*fpp;
55 	xfs_bmbt_key_t		*tkp;
56 	__be64			*tpp;
57 
58 	xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
59 				 XFS_BTNUM_BMAP, 0, 0, ip->i_ino,
60 				 XFS_BTREE_LONG_PTRS);
61 	rblock->bb_level = dblock->bb_level;
62 	ASSERT(be16_to_cpu(rblock->bb_level) > 0);
63 	rblock->bb_numrecs = dblock->bb_numrecs;
64 	dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
65 	fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
66 	tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
67 	fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
68 	tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
69 	dmxr = be16_to_cpu(dblock->bb_numrecs);
70 	memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
71 	memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
72 }
73 
74 void
75 xfs_bmbt_disk_get_all(
76 	struct xfs_bmbt_rec	*rec,
77 	struct xfs_bmbt_irec	*irec)
78 {
79 	uint64_t		l0 = get_unaligned_be64(&rec->l0);
80 	uint64_t		l1 = get_unaligned_be64(&rec->l1);
81 
82 	irec->br_startoff = (l0 & xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
83 	irec->br_startblock = ((l0 & xfs_mask64lo(9)) << 43) | (l1 >> 21);
84 	irec->br_blockcount = l1 & xfs_mask64lo(21);
85 	if (l0 >> (64 - BMBT_EXNTFLAG_BITLEN))
86 		irec->br_state = XFS_EXT_UNWRITTEN;
87 	else
88 		irec->br_state = XFS_EXT_NORM;
89 }
90 
91 /*
92  * Extract the blockcount field from an on disk bmap extent record.
93  */
94 xfs_filblks_t
95 xfs_bmbt_disk_get_blockcount(
96 	xfs_bmbt_rec_t	*r)
97 {
98 	return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
99 }
100 
101 /*
102  * Extract the startoff field from a disk format bmap extent record.
103  */
104 xfs_fileoff_t
105 xfs_bmbt_disk_get_startoff(
106 	xfs_bmbt_rec_t	*r)
107 {
108 	return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
109 		 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
110 }
111 
112 /*
113  * Set all the fields in a bmap extent record from the uncompressed form.
114  */
115 void
116 xfs_bmbt_disk_set_all(
117 	struct xfs_bmbt_rec	*r,
118 	struct xfs_bmbt_irec	*s)
119 {
120 	int			extent_flag = (s->br_state != XFS_EXT_NORM);
121 
122 	ASSERT(s->br_state == XFS_EXT_NORM || s->br_state == XFS_EXT_UNWRITTEN);
123 	ASSERT(!(s->br_startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)));
124 	ASSERT(!(s->br_blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)));
125 	ASSERT(!(s->br_startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)));
126 
127 	put_unaligned_be64(
128 		((xfs_bmbt_rec_base_t)extent_flag << 63) |
129 		 ((xfs_bmbt_rec_base_t)s->br_startoff << 9) |
130 		 ((xfs_bmbt_rec_base_t)s->br_startblock >> 43), &r->l0);
131 	put_unaligned_be64(
132 		((xfs_bmbt_rec_base_t)s->br_startblock << 21) |
133 		 ((xfs_bmbt_rec_base_t)s->br_blockcount &
134 		  (xfs_bmbt_rec_base_t)xfs_mask64lo(21)), &r->l1);
135 }
136 
137 /*
138  * Convert in-memory form of btree root to on-disk form.
139  */
140 void
141 xfs_bmbt_to_bmdr(
142 	struct xfs_mount	*mp,
143 	struct xfs_btree_block	*rblock,
144 	int			rblocklen,
145 	xfs_bmdr_block_t	*dblock,
146 	int			dblocklen)
147 {
148 	int			dmxr;
149 	xfs_bmbt_key_t		*fkp;
150 	__be64			*fpp;
151 	xfs_bmbt_key_t		*tkp;
152 	__be64			*tpp;
153 
154 	if (xfs_sb_version_hascrc(&mp->m_sb)) {
155 		ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
156 		ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid,
157 		       &mp->m_sb.sb_meta_uuid));
158 		ASSERT(rblock->bb_u.l.bb_blkno ==
159 		       cpu_to_be64(XFS_BUF_DADDR_NULL));
160 	} else
161 		ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
162 	ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
163 	ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
164 	ASSERT(rblock->bb_level != 0);
165 	dblock->bb_level = rblock->bb_level;
166 	dblock->bb_numrecs = rblock->bb_numrecs;
167 	dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
168 	fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
169 	tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
170 	fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
171 	tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
172 	dmxr = be16_to_cpu(dblock->bb_numrecs);
173 	memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
174 	memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
175 }
176 
177 STATIC struct xfs_btree_cur *
178 xfs_bmbt_dup_cursor(
179 	struct xfs_btree_cur	*cur)
180 {
181 	struct xfs_btree_cur	*new;
182 
183 	new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
184 			cur->bc_private.b.ip, cur->bc_private.b.whichfork);
185 
186 	/*
187 	 * Copy the firstblock, dfops, and flags values,
188 	 * since init cursor doesn't get them.
189 	 */
190 	new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
191 	new->bc_private.b.dfops = cur->bc_private.b.dfops;
192 	new->bc_private.b.flags = cur->bc_private.b.flags;
193 
194 	return new;
195 }
196 
197 STATIC void
198 xfs_bmbt_update_cursor(
199 	struct xfs_btree_cur	*src,
200 	struct xfs_btree_cur	*dst)
201 {
202 	ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
203 	       (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
204 	ASSERT(dst->bc_private.b.dfops == src->bc_private.b.dfops);
205 
206 	dst->bc_private.b.allocated += src->bc_private.b.allocated;
207 	dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
208 
209 	src->bc_private.b.allocated = 0;
210 }
211 
212 STATIC int
213 xfs_bmbt_alloc_block(
214 	struct xfs_btree_cur	*cur,
215 	union xfs_btree_ptr	*start,
216 	union xfs_btree_ptr	*new,
217 	int			*stat)
218 {
219 	xfs_alloc_arg_t		args;		/* block allocation args */
220 	int			error;		/* error return value */
221 
222 	memset(&args, 0, sizeof(args));
223 	args.tp = cur->bc_tp;
224 	args.mp = cur->bc_mp;
225 	args.fsbno = cur->bc_private.b.firstblock;
226 	args.firstblock = args.fsbno;
227 	xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_private.b.ip->i_ino,
228 			cur->bc_private.b.whichfork);
229 
230 	if (args.fsbno == NULLFSBLOCK) {
231 		args.fsbno = be64_to_cpu(start->l);
232 		args.type = XFS_ALLOCTYPE_START_BNO;
233 		/*
234 		 * Make sure there is sufficient room left in the AG to
235 		 * complete a full tree split for an extent insert.  If
236 		 * we are converting the middle part of an extent then
237 		 * we may need space for two tree splits.
238 		 *
239 		 * We are relying on the caller to make the correct block
240 		 * reservation for this operation to succeed.  If the
241 		 * reservation amount is insufficient then we may fail a
242 		 * block allocation here and corrupt the filesystem.
243 		 */
244 		args.minleft = args.tp->t_blk_res;
245 	} else if (cur->bc_private.b.dfops->dop_low) {
246 		args.type = XFS_ALLOCTYPE_START_BNO;
247 	} else {
248 		args.type = XFS_ALLOCTYPE_NEAR_BNO;
249 	}
250 
251 	args.minlen = args.maxlen = args.prod = 1;
252 	args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
253 	if (!args.wasdel && args.tp->t_blk_res == 0) {
254 		error = -ENOSPC;
255 		goto error0;
256 	}
257 	error = xfs_alloc_vextent(&args);
258 	if (error)
259 		goto error0;
260 
261 	if (args.fsbno == NULLFSBLOCK && args.minleft) {
262 		/*
263 		 * Could not find an AG with enough free space to satisfy
264 		 * a full btree split.  Try again and if
265 		 * successful activate the lowspace algorithm.
266 		 */
267 		args.fsbno = 0;
268 		args.type = XFS_ALLOCTYPE_FIRST_AG;
269 		error = xfs_alloc_vextent(&args);
270 		if (error)
271 			goto error0;
272 		cur->bc_private.b.dfops->dop_low = true;
273 	}
274 	if (WARN_ON_ONCE(args.fsbno == NULLFSBLOCK)) {
275 		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
276 		*stat = 0;
277 		return 0;
278 	}
279 	ASSERT(args.len == 1);
280 	cur->bc_private.b.firstblock = args.fsbno;
281 	cur->bc_private.b.allocated++;
282 	cur->bc_private.b.ip->i_d.di_nblocks++;
283 	xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
284 	xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
285 			XFS_TRANS_DQ_BCOUNT, 1L);
286 
287 	new->l = cpu_to_be64(args.fsbno);
288 
289 	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
290 	*stat = 1;
291 	return 0;
292 
293  error0:
294 	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
295 	return error;
296 }
297 
298 STATIC int
299 xfs_bmbt_free_block(
300 	struct xfs_btree_cur	*cur,
301 	struct xfs_buf		*bp)
302 {
303 	struct xfs_mount	*mp = cur->bc_mp;
304 	struct xfs_inode	*ip = cur->bc_private.b.ip;
305 	struct xfs_trans	*tp = cur->bc_tp;
306 	xfs_fsblock_t		fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
307 	struct xfs_owner_info	oinfo;
308 
309 	xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, cur->bc_private.b.whichfork);
310 	xfs_bmap_add_free(mp, cur->bc_private.b.dfops, fsbno, 1, &oinfo);
311 	ip->i_d.di_nblocks--;
312 
313 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
314 	xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
315 	return 0;
316 }
317 
318 STATIC int
319 xfs_bmbt_get_minrecs(
320 	struct xfs_btree_cur	*cur,
321 	int			level)
322 {
323 	if (level == cur->bc_nlevels - 1) {
324 		struct xfs_ifork	*ifp;
325 
326 		ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
327 				    cur->bc_private.b.whichfork);
328 
329 		return xfs_bmbt_maxrecs(cur->bc_mp,
330 					ifp->if_broot_bytes, level == 0) / 2;
331 	}
332 
333 	return cur->bc_mp->m_bmap_dmnr[level != 0];
334 }
335 
336 int
337 xfs_bmbt_get_maxrecs(
338 	struct xfs_btree_cur	*cur,
339 	int			level)
340 {
341 	if (level == cur->bc_nlevels - 1) {
342 		struct xfs_ifork	*ifp;
343 
344 		ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
345 				    cur->bc_private.b.whichfork);
346 
347 		return xfs_bmbt_maxrecs(cur->bc_mp,
348 					ifp->if_broot_bytes, level == 0);
349 	}
350 
351 	return cur->bc_mp->m_bmap_dmxr[level != 0];
352 
353 }
354 
355 /*
356  * Get the maximum records we could store in the on-disk format.
357  *
358  * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
359  * for the root node this checks the available space in the dinode fork
360  * so that we can resize the in-memory buffer to match it.  After a
361  * resize to the maximum size this function returns the same value
362  * as xfs_bmbt_get_maxrecs for the root node, too.
363  */
364 STATIC int
365 xfs_bmbt_get_dmaxrecs(
366 	struct xfs_btree_cur	*cur,
367 	int			level)
368 {
369 	if (level != cur->bc_nlevels - 1)
370 		return cur->bc_mp->m_bmap_dmxr[level != 0];
371 	return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
372 }
373 
374 STATIC void
375 xfs_bmbt_init_key_from_rec(
376 	union xfs_btree_key	*key,
377 	union xfs_btree_rec	*rec)
378 {
379 	key->bmbt.br_startoff =
380 		cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
381 }
382 
383 STATIC void
384 xfs_bmbt_init_high_key_from_rec(
385 	union xfs_btree_key	*key,
386 	union xfs_btree_rec	*rec)
387 {
388 	key->bmbt.br_startoff = cpu_to_be64(
389 			xfs_bmbt_disk_get_startoff(&rec->bmbt) +
390 			xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
391 }
392 
393 STATIC void
394 xfs_bmbt_init_rec_from_cur(
395 	struct xfs_btree_cur	*cur,
396 	union xfs_btree_rec	*rec)
397 {
398 	xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
399 }
400 
401 STATIC void
402 xfs_bmbt_init_ptr_from_cur(
403 	struct xfs_btree_cur	*cur,
404 	union xfs_btree_ptr	*ptr)
405 {
406 	ptr->l = 0;
407 }
408 
409 STATIC int64_t
410 xfs_bmbt_key_diff(
411 	struct xfs_btree_cur	*cur,
412 	union xfs_btree_key	*key)
413 {
414 	return (int64_t)be64_to_cpu(key->bmbt.br_startoff) -
415 				      cur->bc_rec.b.br_startoff;
416 }
417 
418 STATIC int64_t
419 xfs_bmbt_diff_two_keys(
420 	struct xfs_btree_cur	*cur,
421 	union xfs_btree_key	*k1,
422 	union xfs_btree_key	*k2)
423 {
424 	return (int64_t)be64_to_cpu(k1->bmbt.br_startoff) -
425 			  be64_to_cpu(k2->bmbt.br_startoff);
426 }
427 
428 static xfs_failaddr_t
429 xfs_bmbt_verify(
430 	struct xfs_buf		*bp)
431 {
432 	struct xfs_mount	*mp = bp->b_target->bt_mount;
433 	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
434 	xfs_failaddr_t		fa;
435 	unsigned int		level;
436 
437 	switch (block->bb_magic) {
438 	case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
439 		/*
440 		 * XXX: need a better way of verifying the owner here. Right now
441 		 * just make sure there has been one set.
442 		 */
443 		fa = xfs_btree_lblock_v5hdr_verify(bp, XFS_RMAP_OWN_UNKNOWN);
444 		if (fa)
445 			return fa;
446 		/* fall through */
447 	case cpu_to_be32(XFS_BMAP_MAGIC):
448 		break;
449 	default:
450 		return __this_address;
451 	}
452 
453 	/*
454 	 * numrecs and level verification.
455 	 *
456 	 * We don't know what fork we belong to, so just verify that the level
457 	 * is less than the maximum of the two. Later checks will be more
458 	 * precise.
459 	 */
460 	level = be16_to_cpu(block->bb_level);
461 	if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
462 		return __this_address;
463 
464 	return xfs_btree_lblock_verify(bp, mp->m_bmap_dmxr[level != 0]);
465 }
466 
467 static void
468 xfs_bmbt_read_verify(
469 	struct xfs_buf	*bp)
470 {
471 	xfs_failaddr_t	fa;
472 
473 	if (!xfs_btree_lblock_verify_crc(bp))
474 		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
475 	else {
476 		fa = xfs_bmbt_verify(bp);
477 		if (fa)
478 			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
479 	}
480 
481 	if (bp->b_error)
482 		trace_xfs_btree_corrupt(bp, _RET_IP_);
483 }
484 
485 static void
486 xfs_bmbt_write_verify(
487 	struct xfs_buf	*bp)
488 {
489 	xfs_failaddr_t	fa;
490 
491 	fa = xfs_bmbt_verify(bp);
492 	if (fa) {
493 		trace_xfs_btree_corrupt(bp, _RET_IP_);
494 		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
495 		return;
496 	}
497 	xfs_btree_lblock_calc_crc(bp);
498 }
499 
500 const struct xfs_buf_ops xfs_bmbt_buf_ops = {
501 	.name = "xfs_bmbt",
502 	.verify_read = xfs_bmbt_read_verify,
503 	.verify_write = xfs_bmbt_write_verify,
504 	.verify_struct = xfs_bmbt_verify,
505 };
506 
507 
508 STATIC int
509 xfs_bmbt_keys_inorder(
510 	struct xfs_btree_cur	*cur,
511 	union xfs_btree_key	*k1,
512 	union xfs_btree_key	*k2)
513 {
514 	return be64_to_cpu(k1->bmbt.br_startoff) <
515 		be64_to_cpu(k2->bmbt.br_startoff);
516 }
517 
518 STATIC int
519 xfs_bmbt_recs_inorder(
520 	struct xfs_btree_cur	*cur,
521 	union xfs_btree_rec	*r1,
522 	union xfs_btree_rec	*r2)
523 {
524 	return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
525 		xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
526 		xfs_bmbt_disk_get_startoff(&r2->bmbt);
527 }
528 
529 static const struct xfs_btree_ops xfs_bmbt_ops = {
530 	.rec_len		= sizeof(xfs_bmbt_rec_t),
531 	.key_len		= sizeof(xfs_bmbt_key_t),
532 
533 	.dup_cursor		= xfs_bmbt_dup_cursor,
534 	.update_cursor		= xfs_bmbt_update_cursor,
535 	.alloc_block		= xfs_bmbt_alloc_block,
536 	.free_block		= xfs_bmbt_free_block,
537 	.get_maxrecs		= xfs_bmbt_get_maxrecs,
538 	.get_minrecs		= xfs_bmbt_get_minrecs,
539 	.get_dmaxrecs		= xfs_bmbt_get_dmaxrecs,
540 	.init_key_from_rec	= xfs_bmbt_init_key_from_rec,
541 	.init_high_key_from_rec	= xfs_bmbt_init_high_key_from_rec,
542 	.init_rec_from_cur	= xfs_bmbt_init_rec_from_cur,
543 	.init_ptr_from_cur	= xfs_bmbt_init_ptr_from_cur,
544 	.key_diff		= xfs_bmbt_key_diff,
545 	.diff_two_keys		= xfs_bmbt_diff_two_keys,
546 	.buf_ops		= &xfs_bmbt_buf_ops,
547 	.keys_inorder		= xfs_bmbt_keys_inorder,
548 	.recs_inorder		= xfs_bmbt_recs_inorder,
549 };
550 
551 /*
552  * Allocate a new bmap btree cursor.
553  */
554 struct xfs_btree_cur *				/* new bmap btree cursor */
555 xfs_bmbt_init_cursor(
556 	struct xfs_mount	*mp,		/* file system mount point */
557 	struct xfs_trans	*tp,		/* transaction pointer */
558 	struct xfs_inode	*ip,		/* inode owning the btree */
559 	int			whichfork)	/* data or attr fork */
560 {
561 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
562 	struct xfs_btree_cur	*cur;
563 	ASSERT(whichfork != XFS_COW_FORK);
564 
565 	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
566 
567 	cur->bc_tp = tp;
568 	cur->bc_mp = mp;
569 	cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
570 	cur->bc_btnum = XFS_BTNUM_BMAP;
571 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
572 	cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_bmbt_2);
573 
574 	cur->bc_ops = &xfs_bmbt_ops;
575 	cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
576 	if (xfs_sb_version_hascrc(&mp->m_sb))
577 		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
578 
579 	cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
580 	cur->bc_private.b.ip = ip;
581 	cur->bc_private.b.firstblock = NULLFSBLOCK;
582 	cur->bc_private.b.dfops = NULL;
583 	cur->bc_private.b.allocated = 0;
584 	cur->bc_private.b.flags = 0;
585 	cur->bc_private.b.whichfork = whichfork;
586 
587 	return cur;
588 }
589 
590 /*
591  * Calculate number of records in a bmap btree block.
592  */
593 int
594 xfs_bmbt_maxrecs(
595 	struct xfs_mount	*mp,
596 	int			blocklen,
597 	int			leaf)
598 {
599 	blocklen -= XFS_BMBT_BLOCK_LEN(mp);
600 
601 	if (leaf)
602 		return blocklen / sizeof(xfs_bmbt_rec_t);
603 	return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
604 }
605 
606 /*
607  * Calculate number of records in a bmap btree inode root.
608  */
609 int
610 xfs_bmdr_maxrecs(
611 	int			blocklen,
612 	int			leaf)
613 {
614 	blocklen -= sizeof(xfs_bmdr_block_t);
615 
616 	if (leaf)
617 		return blocklen / sizeof(xfs_bmdr_rec_t);
618 	return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
619 }
620 
621 /*
622  * Change the owner of a btree format fork fo the inode passed in. Change it to
623  * the owner of that is passed in so that we can change owners before or after
624  * we switch forks between inodes. The operation that the caller is doing will
625  * determine whether is needs to change owner before or after the switch.
626  *
627  * For demand paged transactional modification, the fork switch should be done
628  * after reading in all the blocks, modifying them and pinning them in the
629  * transaction. For modification when the buffers are already pinned in memory,
630  * the fork switch can be done before changing the owner as we won't need to
631  * validate the owner until the btree buffers are unpinned and writes can occur
632  * again.
633  *
634  * For recovery based ownership change, there is no transactional context and
635  * so a buffer list must be supplied so that we can record the buffers that we
636  * modified for the caller to issue IO on.
637  */
638 int
639 xfs_bmbt_change_owner(
640 	struct xfs_trans	*tp,
641 	struct xfs_inode	*ip,
642 	int			whichfork,
643 	xfs_ino_t		new_owner,
644 	struct list_head	*buffer_list)
645 {
646 	struct xfs_btree_cur	*cur;
647 	int			error;
648 
649 	ASSERT(tp || buffer_list);
650 	ASSERT(!(tp && buffer_list));
651 	if (whichfork == XFS_DATA_FORK)
652 		ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
653 	else
654 		ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
655 
656 	cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
657 	if (!cur)
658 		return -ENOMEM;
659 	cur->bc_private.b.flags |= XFS_BTCUR_BPRV_INVALID_OWNER;
660 
661 	error = xfs_btree_change_owner(cur, new_owner, buffer_list);
662 	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
663 	return error;
664 }
665