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_inode.h"
27#include "xfs_trans.h"
28#include "xfs_inode_item.h"
29#include "xfs_alloc.h"
30#include "xfs_btree.h"
31#include "xfs_bmap_btree.h"
32#include "xfs_bmap.h"
33#include "xfs_error.h"
34#include "xfs_quota.h"
35#include "xfs_trace.h"
36#include "xfs_cksum.h"
37
38/*
39 * Determine the extent state.
40 */
41/* ARGSUSED */
42STATIC xfs_exntst_t
43xfs_extent_state(
44	xfs_filblks_t		blks,
45	int			extent_flag)
46{
47	if (extent_flag) {
48		ASSERT(blks != 0);	/* saved for DMIG */
49		return XFS_EXT_UNWRITTEN;
50	}
51	return XFS_EXT_NORM;
52}
53
54/*
55 * Convert on-disk form of btree root to in-memory form.
56 */
57void
58xfs_bmdr_to_bmbt(
59	struct xfs_inode	*ip,
60	xfs_bmdr_block_t	*dblock,
61	int			dblocklen,
62	struct xfs_btree_block	*rblock,
63	int			rblocklen)
64{
65	struct xfs_mount	*mp = ip->i_mount;
66	int			dmxr;
67	xfs_bmbt_key_t		*fkp;
68	__be64			*fpp;
69	xfs_bmbt_key_t		*tkp;
70	__be64			*tpp;
71
72	if (xfs_sb_version_hascrc(&mp->m_sb))
73		xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
74				 XFS_BMAP_CRC_MAGIC, 0, 0, ip->i_ino,
75				 XFS_BTREE_LONG_PTRS | XFS_BTREE_CRC_BLOCKS);
76	else
77		xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
78				 XFS_BMAP_MAGIC, 0, 0, ip->i_ino,
79				 XFS_BTREE_LONG_PTRS);
80
81	rblock->bb_level = dblock->bb_level;
82	ASSERT(be16_to_cpu(rblock->bb_level) > 0);
83	rblock->bb_numrecs = dblock->bb_numrecs;
84	dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
85	fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
86	tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
87	fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
88	tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
89	dmxr = be16_to_cpu(dblock->bb_numrecs);
90	memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
91	memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
92}
93
94/*
95 * Convert a compressed bmap extent record to an uncompressed form.
96 * This code must be in sync with the routines xfs_bmbt_get_startoff,
97 * xfs_bmbt_get_startblock, xfs_bmbt_get_blockcount and xfs_bmbt_get_state.
98 */
99STATIC void
100__xfs_bmbt_get_all(
101		__uint64_t l0,
102		__uint64_t l1,
103		xfs_bmbt_irec_t *s)
104{
105	int	ext_flag;
106	xfs_exntst_t st;
107
108	ext_flag = (int)(l0 >> (64 - BMBT_EXNTFLAG_BITLEN));
109	s->br_startoff = ((xfs_fileoff_t)l0 &
110			   xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
111	s->br_startblock = (((xfs_fsblock_t)l0 & xfs_mask64lo(9)) << 43) |
112			   (((xfs_fsblock_t)l1) >> 21);
113	s->br_blockcount = (xfs_filblks_t)(l1 & xfs_mask64lo(21));
114	/* This is xfs_extent_state() in-line */
115	if (ext_flag) {
116		ASSERT(s->br_blockcount != 0);	/* saved for DMIG */
117		st = XFS_EXT_UNWRITTEN;
118	} else
119		st = XFS_EXT_NORM;
120	s->br_state = st;
121}
122
123void
124xfs_bmbt_get_all(
125	xfs_bmbt_rec_host_t *r,
126	xfs_bmbt_irec_t *s)
127{
128	__xfs_bmbt_get_all(r->l0, r->l1, s);
129}
130
131/*
132 * Extract the blockcount field from an in memory bmap extent record.
133 */
134xfs_filblks_t
135xfs_bmbt_get_blockcount(
136	xfs_bmbt_rec_host_t	*r)
137{
138	return (xfs_filblks_t)(r->l1 & xfs_mask64lo(21));
139}
140
141/*
142 * Extract the startblock field from an in memory bmap extent record.
143 */
144xfs_fsblock_t
145xfs_bmbt_get_startblock(
146	xfs_bmbt_rec_host_t	*r)
147{
148	return (((xfs_fsblock_t)r->l0 & xfs_mask64lo(9)) << 43) |
149	       (((xfs_fsblock_t)r->l1) >> 21);
150}
151
152/*
153 * Extract the startoff field from an in memory bmap extent record.
154 */
155xfs_fileoff_t
156xfs_bmbt_get_startoff(
157	xfs_bmbt_rec_host_t	*r)
158{
159	return ((xfs_fileoff_t)r->l0 &
160		 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
161}
162
163xfs_exntst_t
164xfs_bmbt_get_state(
165	xfs_bmbt_rec_host_t	*r)
166{
167	int	ext_flag;
168
169	ext_flag = (int)((r->l0) >> (64 - BMBT_EXNTFLAG_BITLEN));
170	return xfs_extent_state(xfs_bmbt_get_blockcount(r),
171				ext_flag);
172}
173
174/*
175 * Extract the blockcount field from an on disk bmap extent record.
176 */
177xfs_filblks_t
178xfs_bmbt_disk_get_blockcount(
179	xfs_bmbt_rec_t	*r)
180{
181	return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
182}
183
184/*
185 * Extract the startoff field from a disk format bmap extent record.
186 */
187xfs_fileoff_t
188xfs_bmbt_disk_get_startoff(
189	xfs_bmbt_rec_t	*r)
190{
191	return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
192		 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
193}
194
195
196/*
197 * Set all the fields in a bmap extent record from the arguments.
198 */
199void
200xfs_bmbt_set_allf(
201	xfs_bmbt_rec_host_t	*r,
202	xfs_fileoff_t		startoff,
203	xfs_fsblock_t		startblock,
204	xfs_filblks_t		blockcount,
205	xfs_exntst_t		state)
206{
207	int		extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
208
209	ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
210	ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
211	ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
212
213	ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
214
215	r->l0 = ((xfs_bmbt_rec_base_t)extent_flag << 63) |
216		((xfs_bmbt_rec_base_t)startoff << 9) |
217		((xfs_bmbt_rec_base_t)startblock >> 43);
218	r->l1 = ((xfs_bmbt_rec_base_t)startblock << 21) |
219		((xfs_bmbt_rec_base_t)blockcount &
220		(xfs_bmbt_rec_base_t)xfs_mask64lo(21));
221}
222
223/*
224 * Set all the fields in a bmap extent record from the uncompressed form.
225 */
226void
227xfs_bmbt_set_all(
228	xfs_bmbt_rec_host_t *r,
229	xfs_bmbt_irec_t	*s)
230{
231	xfs_bmbt_set_allf(r, s->br_startoff, s->br_startblock,
232			     s->br_blockcount, s->br_state);
233}
234
235
236/*
237 * Set all the fields in a disk format bmap extent record from the arguments.
238 */
239void
240xfs_bmbt_disk_set_allf(
241	xfs_bmbt_rec_t		*r,
242	xfs_fileoff_t		startoff,
243	xfs_fsblock_t		startblock,
244	xfs_filblks_t		blockcount,
245	xfs_exntst_t		state)
246{
247	int			extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
248
249	ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
250	ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
251	ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
252	ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
253
254	r->l0 = cpu_to_be64(
255		((xfs_bmbt_rec_base_t)extent_flag << 63) |
256		 ((xfs_bmbt_rec_base_t)startoff << 9) |
257		 ((xfs_bmbt_rec_base_t)startblock >> 43));
258	r->l1 = cpu_to_be64(
259		((xfs_bmbt_rec_base_t)startblock << 21) |
260		 ((xfs_bmbt_rec_base_t)blockcount &
261		  (xfs_bmbt_rec_base_t)xfs_mask64lo(21)));
262}
263
264/*
265 * Set all the fields in a bmap extent record from the uncompressed form.
266 */
267STATIC void
268xfs_bmbt_disk_set_all(
269	xfs_bmbt_rec_t	*r,
270	xfs_bmbt_irec_t *s)
271{
272	xfs_bmbt_disk_set_allf(r, s->br_startoff, s->br_startblock,
273				  s->br_blockcount, s->br_state);
274}
275
276/*
277 * Set the blockcount field in a bmap extent record.
278 */
279void
280xfs_bmbt_set_blockcount(
281	xfs_bmbt_rec_host_t *r,
282	xfs_filblks_t	v)
283{
284	ASSERT((v & xfs_mask64hi(43)) == 0);
285	r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64hi(43)) |
286		  (xfs_bmbt_rec_base_t)(v & xfs_mask64lo(21));
287}
288
289/*
290 * Set the startblock field in a bmap extent record.
291 */
292void
293xfs_bmbt_set_startblock(
294	xfs_bmbt_rec_host_t *r,
295	xfs_fsblock_t	v)
296{
297	ASSERT((v & xfs_mask64hi(12)) == 0);
298	r->l0 = (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64hi(55)) |
299		  (xfs_bmbt_rec_base_t)(v >> 43);
300	r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64lo(21)) |
301		  (xfs_bmbt_rec_base_t)(v << 21);
302}
303
304/*
305 * Set the startoff field in a bmap extent record.
306 */
307void
308xfs_bmbt_set_startoff(
309	xfs_bmbt_rec_host_t *r,
310	xfs_fileoff_t	v)
311{
312	ASSERT((v & xfs_mask64hi(9)) == 0);
313	r->l0 = (r->l0 & (xfs_bmbt_rec_base_t) xfs_mask64hi(1)) |
314		((xfs_bmbt_rec_base_t)v << 9) |
315		  (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64lo(9));
316}
317
318/*
319 * Set the extent state field in a bmap extent record.
320 */
321void
322xfs_bmbt_set_state(
323	xfs_bmbt_rec_host_t *r,
324	xfs_exntst_t	v)
325{
326	ASSERT(v == XFS_EXT_NORM || v == XFS_EXT_UNWRITTEN);
327	if (v == XFS_EXT_NORM)
328		r->l0 &= xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN);
329	else
330		r->l0 |= xfs_mask64hi(BMBT_EXNTFLAG_BITLEN);
331}
332
333/*
334 * Convert in-memory form of btree root to on-disk form.
335 */
336void
337xfs_bmbt_to_bmdr(
338	struct xfs_mount	*mp,
339	struct xfs_btree_block	*rblock,
340	int			rblocklen,
341	xfs_bmdr_block_t	*dblock,
342	int			dblocklen)
343{
344	int			dmxr;
345	xfs_bmbt_key_t		*fkp;
346	__be64			*fpp;
347	xfs_bmbt_key_t		*tkp;
348	__be64			*tpp;
349
350	if (xfs_sb_version_hascrc(&mp->m_sb)) {
351		ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
352		ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid, &mp->m_sb.sb_uuid));
353		ASSERT(rblock->bb_u.l.bb_blkno ==
354		       cpu_to_be64(XFS_BUF_DADDR_NULL));
355	} else
356		ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
357	ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
358	ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
359	ASSERT(rblock->bb_level != 0);
360	dblock->bb_level = rblock->bb_level;
361	dblock->bb_numrecs = rblock->bb_numrecs;
362	dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
363	fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
364	tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
365	fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
366	tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
367	dmxr = be16_to_cpu(dblock->bb_numrecs);
368	memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
369	memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
370}
371
372/*
373 * Check extent records, which have just been read, for
374 * any bit in the extent flag field. ASSERT on debug
375 * kernels, as this condition should not occur.
376 * Return an error condition (1) if any flags found,
377 * otherwise return 0.
378 */
379
380int
381xfs_check_nostate_extents(
382	xfs_ifork_t		*ifp,
383	xfs_extnum_t		idx,
384	xfs_extnum_t		num)
385{
386	for (; num > 0; num--, idx++) {
387		xfs_bmbt_rec_host_t *ep = xfs_iext_get_ext(ifp, idx);
388		if ((ep->l0 >>
389		     (64 - BMBT_EXNTFLAG_BITLEN)) != 0) {
390			ASSERT(0);
391			return 1;
392		}
393	}
394	return 0;
395}
396
397
398STATIC struct xfs_btree_cur *
399xfs_bmbt_dup_cursor(
400	struct xfs_btree_cur	*cur)
401{
402	struct xfs_btree_cur	*new;
403
404	new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
405			cur->bc_private.b.ip, cur->bc_private.b.whichfork);
406
407	/*
408	 * Copy the firstblock, flist, and flags values,
409	 * since init cursor doesn't get them.
410	 */
411	new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
412	new->bc_private.b.flist = cur->bc_private.b.flist;
413	new->bc_private.b.flags = cur->bc_private.b.flags;
414
415	return new;
416}
417
418STATIC void
419xfs_bmbt_update_cursor(
420	struct xfs_btree_cur	*src,
421	struct xfs_btree_cur	*dst)
422{
423	ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
424	       (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
425	ASSERT(dst->bc_private.b.flist == src->bc_private.b.flist);
426
427	dst->bc_private.b.allocated += src->bc_private.b.allocated;
428	dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
429
430	src->bc_private.b.allocated = 0;
431}
432
433STATIC int
434xfs_bmbt_alloc_block(
435	struct xfs_btree_cur	*cur,
436	union xfs_btree_ptr	*start,
437	union xfs_btree_ptr	*new,
438	int			*stat)
439{
440	xfs_alloc_arg_t		args;		/* block allocation args */
441	int			error;		/* error return value */
442
443	memset(&args, 0, sizeof(args));
444	args.tp = cur->bc_tp;
445	args.mp = cur->bc_mp;
446	args.fsbno = cur->bc_private.b.firstblock;
447	args.firstblock = args.fsbno;
448
449	if (args.fsbno == NULLFSBLOCK) {
450		args.fsbno = be64_to_cpu(start->l);
451		args.type = XFS_ALLOCTYPE_START_BNO;
452		/*
453		 * Make sure there is sufficient room left in the AG to
454		 * complete a full tree split for an extent insert.  If
455		 * we are converting the middle part of an extent then
456		 * we may need space for two tree splits.
457		 *
458		 * We are relying on the caller to make the correct block
459		 * reservation for this operation to succeed.  If the
460		 * reservation amount is insufficient then we may fail a
461		 * block allocation here and corrupt the filesystem.
462		 */
463		args.minleft = xfs_trans_get_block_res(args.tp);
464	} else if (cur->bc_private.b.flist->xbf_low) {
465		args.type = XFS_ALLOCTYPE_START_BNO;
466	} else {
467		args.type = XFS_ALLOCTYPE_NEAR_BNO;
468	}
469
470	args.minlen = args.maxlen = args.prod = 1;
471	args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
472	if (!args.wasdel && xfs_trans_get_block_res(args.tp) == 0) {
473		error = -ENOSPC;
474		goto error0;
475	}
476	error = xfs_alloc_vextent(&args);
477	if (error)
478		goto error0;
479
480	if (args.fsbno == NULLFSBLOCK && args.minleft) {
481		/*
482		 * Could not find an AG with enough free space to satisfy
483		 * a full btree split.  Try again without minleft and if
484		 * successful activate the lowspace algorithm.
485		 */
486		args.fsbno = 0;
487		args.type = XFS_ALLOCTYPE_FIRST_AG;
488		args.minleft = 0;
489		error = xfs_alloc_vextent(&args);
490		if (error)
491			goto error0;
492		cur->bc_private.b.flist->xbf_low = 1;
493	}
494	if (args.fsbno == NULLFSBLOCK) {
495		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
496		*stat = 0;
497		return 0;
498	}
499	ASSERT(args.len == 1);
500	cur->bc_private.b.firstblock = args.fsbno;
501	cur->bc_private.b.allocated++;
502	cur->bc_private.b.ip->i_d.di_nblocks++;
503	xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
504	xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
505			XFS_TRANS_DQ_BCOUNT, 1L);
506
507	new->l = cpu_to_be64(args.fsbno);
508
509	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
510	*stat = 1;
511	return 0;
512
513 error0:
514	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
515	return error;
516}
517
518STATIC int
519xfs_bmbt_free_block(
520	struct xfs_btree_cur	*cur,
521	struct xfs_buf		*bp)
522{
523	struct xfs_mount	*mp = cur->bc_mp;
524	struct xfs_inode	*ip = cur->bc_private.b.ip;
525	struct xfs_trans	*tp = cur->bc_tp;
526	xfs_fsblock_t		fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
527
528	xfs_bmap_add_free(fsbno, 1, cur->bc_private.b.flist, mp);
529	ip->i_d.di_nblocks--;
530
531	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
532	xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
533	xfs_trans_binval(tp, bp);
534	return 0;
535}
536
537STATIC int
538xfs_bmbt_get_minrecs(
539	struct xfs_btree_cur	*cur,
540	int			level)
541{
542	if (level == cur->bc_nlevels - 1) {
543		struct xfs_ifork	*ifp;
544
545		ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
546				    cur->bc_private.b.whichfork);
547
548		return xfs_bmbt_maxrecs(cur->bc_mp,
549					ifp->if_broot_bytes, level == 0) / 2;
550	}
551
552	return cur->bc_mp->m_bmap_dmnr[level != 0];
553}
554
555int
556xfs_bmbt_get_maxrecs(
557	struct xfs_btree_cur	*cur,
558	int			level)
559{
560	if (level == cur->bc_nlevels - 1) {
561		struct xfs_ifork	*ifp;
562
563		ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
564				    cur->bc_private.b.whichfork);
565
566		return xfs_bmbt_maxrecs(cur->bc_mp,
567					ifp->if_broot_bytes, level == 0);
568	}
569
570	return cur->bc_mp->m_bmap_dmxr[level != 0];
571
572}
573
574/*
575 * Get the maximum records we could store in the on-disk format.
576 *
577 * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
578 * for the root node this checks the available space in the dinode fork
579 * so that we can resize the in-memory buffer to match it.  After a
580 * resize to the maximum size this function returns the same value
581 * as xfs_bmbt_get_maxrecs for the root node, too.
582 */
583STATIC int
584xfs_bmbt_get_dmaxrecs(
585	struct xfs_btree_cur	*cur,
586	int			level)
587{
588	if (level != cur->bc_nlevels - 1)
589		return cur->bc_mp->m_bmap_dmxr[level != 0];
590	return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
591}
592
593STATIC void
594xfs_bmbt_init_key_from_rec(
595	union xfs_btree_key	*key,
596	union xfs_btree_rec	*rec)
597{
598	key->bmbt.br_startoff =
599		cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
600}
601
602STATIC void
603xfs_bmbt_init_rec_from_key(
604	union xfs_btree_key	*key,
605	union xfs_btree_rec	*rec)
606{
607	ASSERT(key->bmbt.br_startoff != 0);
608
609	xfs_bmbt_disk_set_allf(&rec->bmbt, be64_to_cpu(key->bmbt.br_startoff),
610			       0, 0, XFS_EXT_NORM);
611}
612
613STATIC void
614xfs_bmbt_init_rec_from_cur(
615	struct xfs_btree_cur	*cur,
616	union xfs_btree_rec	*rec)
617{
618	xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
619}
620
621STATIC void
622xfs_bmbt_init_ptr_from_cur(
623	struct xfs_btree_cur	*cur,
624	union xfs_btree_ptr	*ptr)
625{
626	ptr->l = 0;
627}
628
629STATIC __int64_t
630xfs_bmbt_key_diff(
631	struct xfs_btree_cur	*cur,
632	union xfs_btree_key	*key)
633{
634	return (__int64_t)be64_to_cpu(key->bmbt.br_startoff) -
635				      cur->bc_rec.b.br_startoff;
636}
637
638static bool
639xfs_bmbt_verify(
640	struct xfs_buf		*bp)
641{
642	struct xfs_mount	*mp = bp->b_target->bt_mount;
643	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
644	unsigned int		level;
645
646	switch (block->bb_magic) {
647	case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
648		if (!xfs_sb_version_hascrc(&mp->m_sb))
649			return false;
650		if (!uuid_equal(&block->bb_u.l.bb_uuid, &mp->m_sb.sb_uuid))
651			return false;
652		if (be64_to_cpu(block->bb_u.l.bb_blkno) != bp->b_bn)
653			return false;
654		/*
655		 * XXX: need a better way of verifying the owner here. Right now
656		 * just make sure there has been one set.
657		 */
658		if (be64_to_cpu(block->bb_u.l.bb_owner) == 0)
659			return false;
660		/* fall through */
661	case cpu_to_be32(XFS_BMAP_MAGIC):
662		break;
663	default:
664		return false;
665	}
666
667	/*
668	 * numrecs and level verification.
669	 *
670	 * We don't know what fork we belong to, so just verify that the level
671	 * is less than the maximum of the two. Later checks will be more
672	 * precise.
673	 */
674	level = be16_to_cpu(block->bb_level);
675	if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
676		return false;
677	if (be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
678		return false;
679
680	/* sibling pointer verification */
681	if (!block->bb_u.l.bb_leftsib ||
682	    (block->bb_u.l.bb_leftsib != cpu_to_be64(NULLFSBLOCK) &&
683	     !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_leftsib))))
684		return false;
685	if (!block->bb_u.l.bb_rightsib ||
686	    (block->bb_u.l.bb_rightsib != cpu_to_be64(NULLFSBLOCK) &&
687	     !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_rightsib))))
688		return false;
689
690	return true;
691}
692
693static void
694xfs_bmbt_read_verify(
695	struct xfs_buf	*bp)
696{
697	if (!xfs_btree_lblock_verify_crc(bp))
698		xfs_buf_ioerror(bp, -EFSBADCRC);
699	else if (!xfs_bmbt_verify(bp))
700		xfs_buf_ioerror(bp, -EFSCORRUPTED);
701
702	if (bp->b_error) {
703		trace_xfs_btree_corrupt(bp, _RET_IP_);
704		xfs_verifier_error(bp);
705	}
706}
707
708static void
709xfs_bmbt_write_verify(
710	struct xfs_buf	*bp)
711{
712	if (!xfs_bmbt_verify(bp)) {
713		trace_xfs_btree_corrupt(bp, _RET_IP_);
714		xfs_buf_ioerror(bp, -EFSCORRUPTED);
715		xfs_verifier_error(bp);
716		return;
717	}
718	xfs_btree_lblock_calc_crc(bp);
719}
720
721const struct xfs_buf_ops xfs_bmbt_buf_ops = {
722	.verify_read = xfs_bmbt_read_verify,
723	.verify_write = xfs_bmbt_write_verify,
724};
725
726
727#if defined(DEBUG) || defined(XFS_WARN)
728STATIC int
729xfs_bmbt_keys_inorder(
730	struct xfs_btree_cur	*cur,
731	union xfs_btree_key	*k1,
732	union xfs_btree_key	*k2)
733{
734	return be64_to_cpu(k1->bmbt.br_startoff) <
735		be64_to_cpu(k2->bmbt.br_startoff);
736}
737
738STATIC int
739xfs_bmbt_recs_inorder(
740	struct xfs_btree_cur	*cur,
741	union xfs_btree_rec	*r1,
742	union xfs_btree_rec	*r2)
743{
744	return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
745		xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
746		xfs_bmbt_disk_get_startoff(&r2->bmbt);
747}
748#endif	/* DEBUG */
749
750static const struct xfs_btree_ops xfs_bmbt_ops = {
751	.rec_len		= sizeof(xfs_bmbt_rec_t),
752	.key_len		= sizeof(xfs_bmbt_key_t),
753
754	.dup_cursor		= xfs_bmbt_dup_cursor,
755	.update_cursor		= xfs_bmbt_update_cursor,
756	.alloc_block		= xfs_bmbt_alloc_block,
757	.free_block		= xfs_bmbt_free_block,
758	.get_maxrecs		= xfs_bmbt_get_maxrecs,
759	.get_minrecs		= xfs_bmbt_get_minrecs,
760	.get_dmaxrecs		= xfs_bmbt_get_dmaxrecs,
761	.init_key_from_rec	= xfs_bmbt_init_key_from_rec,
762	.init_rec_from_key	= xfs_bmbt_init_rec_from_key,
763	.init_rec_from_cur	= xfs_bmbt_init_rec_from_cur,
764	.init_ptr_from_cur	= xfs_bmbt_init_ptr_from_cur,
765	.key_diff		= xfs_bmbt_key_diff,
766	.buf_ops		= &xfs_bmbt_buf_ops,
767#if defined(DEBUG) || defined(XFS_WARN)
768	.keys_inorder		= xfs_bmbt_keys_inorder,
769	.recs_inorder		= xfs_bmbt_recs_inorder,
770#endif
771};
772
773/*
774 * Allocate a new bmap btree cursor.
775 */
776struct xfs_btree_cur *				/* new bmap btree cursor */
777xfs_bmbt_init_cursor(
778	struct xfs_mount	*mp,		/* file system mount point */
779	struct xfs_trans	*tp,		/* transaction pointer */
780	struct xfs_inode	*ip,		/* inode owning the btree */
781	int			whichfork)	/* data or attr fork */
782{
783	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
784	struct xfs_btree_cur	*cur;
785
786	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
787
788	cur->bc_tp = tp;
789	cur->bc_mp = mp;
790	cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
791	cur->bc_btnum = XFS_BTNUM_BMAP;
792	cur->bc_blocklog = mp->m_sb.sb_blocklog;
793
794	cur->bc_ops = &xfs_bmbt_ops;
795	cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
796	if (xfs_sb_version_hascrc(&mp->m_sb))
797		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
798
799	cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
800	cur->bc_private.b.ip = ip;
801	cur->bc_private.b.firstblock = NULLFSBLOCK;
802	cur->bc_private.b.flist = NULL;
803	cur->bc_private.b.allocated = 0;
804	cur->bc_private.b.flags = 0;
805	cur->bc_private.b.whichfork = whichfork;
806
807	return cur;
808}
809
810/*
811 * Calculate number of records in a bmap btree block.
812 */
813int
814xfs_bmbt_maxrecs(
815	struct xfs_mount	*mp,
816	int			blocklen,
817	int			leaf)
818{
819	blocklen -= XFS_BMBT_BLOCK_LEN(mp);
820
821	if (leaf)
822		return blocklen / sizeof(xfs_bmbt_rec_t);
823	return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
824}
825
826/*
827 * Calculate number of records in a bmap btree inode root.
828 */
829int
830xfs_bmdr_maxrecs(
831	int			blocklen,
832	int			leaf)
833{
834	blocklen -= sizeof(xfs_bmdr_block_t);
835
836	if (leaf)
837		return blocklen / sizeof(xfs_bmdr_rec_t);
838	return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
839}
840
841/*
842 * Change the owner of a btree format fork fo the inode passed in. Change it to
843 * the owner of that is passed in so that we can change owners before or after
844 * we switch forks between inodes. The operation that the caller is doing will
845 * determine whether is needs to change owner before or after the switch.
846 *
847 * For demand paged transactional modification, the fork switch should be done
848 * after reading in all the blocks, modifying them and pinning them in the
849 * transaction. For modification when the buffers are already pinned in memory,
850 * the fork switch can be done before changing the owner as we won't need to
851 * validate the owner until the btree buffers are unpinned and writes can occur
852 * again.
853 *
854 * For recovery based ownership change, there is no transactional context and
855 * so a buffer list must be supplied so that we can record the buffers that we
856 * modified for the caller to issue IO on.
857 */
858int
859xfs_bmbt_change_owner(
860	struct xfs_trans	*tp,
861	struct xfs_inode	*ip,
862	int			whichfork,
863	xfs_ino_t		new_owner,
864	struct list_head	*buffer_list)
865{
866	struct xfs_btree_cur	*cur;
867	int			error;
868
869	ASSERT(tp || buffer_list);
870	ASSERT(!(tp && buffer_list));
871	if (whichfork == XFS_DATA_FORK)
872		ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
873	else
874		ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
875
876	cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
877	if (!cur)
878		return -ENOMEM;
879
880	error = xfs_btree_change_owner(cur, new_owner, buffer_list);
881	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
882	return error;
883}
884