]> git.ipfire.org Git - thirdparty/xfsprogs-dev.git/blame - libxfs/xfs_bmap_btree.c
xfs: validate btree records on retrieval
[thirdparty/xfsprogs-dev.git] / libxfs / xfs_bmap_btree.c
CommitLineData
2bd0ea18 1/*
da23017d
NS
2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
5000d01d 4 *
da23017d
NS
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
2bd0ea18 7 * published by the Free Software Foundation.
5000d01d 8 *
da23017d
NS
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.
5000d01d 13 *
da23017d
NS
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
2bd0ea18 17 */
9c799827 18#include "libxfs_priv.h"
b626fb59
DC
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"
f944d3d0 26#include "xfs_defer.h"
b626fb59
DC
27#include "xfs_inode.h"
28#include "xfs_trans.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_trace.h"
34#include "xfs_cksum.h"
85aec44f 35#include "xfs_rmap.h"
2bd0ea18 36
2bd0ea18
NS
37/*
38 * Convert on-disk form of btree root to in-memory form.
39 */
40void
41xfs_bmdr_to_bmbt(
5dfa5cd2 42 struct xfs_inode *ip,
2bd0ea18
NS
43 xfs_bmdr_block_t *dblock,
44 int dblocklen,
b3563c19 45 struct xfs_btree_block *rblock,
2bd0ea18
NS
46 int rblocklen)
47{
5dfa5cd2 48 struct xfs_mount *mp = ip->i_mount;
2bd0ea18
NS
49 int dmxr;
50 xfs_bmbt_key_t *fkp;
5e656dbb 51 __be64 *fpp;
2bd0ea18 52 xfs_bmbt_key_t *tkp;
5e656dbb 53 __be64 *tpp;
2bd0ea18 54
e394a4b1
ES
55 xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
56 XFS_BTNUM_BMAP, 0, 0, ip->i_ino,
f4241a08 57 XFS_BTREE_LONG_PTRS);
6e3140c7
NS
58 rblock->bb_level = dblock->bb_level;
59 ASSERT(be16_to_cpu(rblock->bb_level) > 0);
60 rblock->bb_numrecs = dblock->bb_numrecs;
ff105f75 61 dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
b3563c19
BN
62 fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
63 tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
64 fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
65 tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
6e3140c7 66 dmxr = be16_to_cpu(dblock->bb_numrecs);
32181a02 67 memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
5e656dbb 68 memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
2bd0ea18
NS
69}
70
f9e56f43 71void
b37d753d
CH
72xfs_bmbt_disk_get_all(
73 struct xfs_bmbt_rec *rec,
74 struct xfs_bmbt_irec *irec)
2bd0ea18 75{
b37d753d
CH
76 uint64_t l0 = get_unaligned_be64(&rec->l0);
77 uint64_t l1 = get_unaligned_be64(&rec->l1);
f9e56f43 78
b37d753d
CH
79 irec->br_startoff = (l0 & xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
80 irec->br_startblock = ((l0 & xfs_mask64lo(9)) << 43) | (l1 >> 21);
81 irec->br_blockcount = l1 & xfs_mask64lo(21);
82 if (l0 >> (64 - BMBT_EXNTFLAG_BITLEN))
83 irec->br_state = XFS_EXT_UNWRITTEN;
84 else
85 irec->br_state = XFS_EXT_NORM;
f9e56f43
NS
86}
87
2bd0ea18 88/*
b194c7d8 89 * Extract the blockcount field from an on disk bmap extent record.
2bd0ea18 90 */
b194c7d8
BN
91xfs_filblks_t
92xfs_bmbt_disk_get_blockcount(
93 xfs_bmbt_rec_t *r)
2bd0ea18 94{
56b2de80 95 return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
2bd0ea18
NS
96}
97
2bd0ea18 98/*
b194c7d8 99 * Extract the startoff field from a disk format bmap extent record.
2bd0ea18 100 */
b194c7d8
BN
101xfs_fileoff_t
102xfs_bmbt_disk_get_startoff(
103 xfs_bmbt_rec_t *r)
2bd0ea18 104{
b194c7d8 105 return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
56b2de80 106 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
2bd0ea18
NS
107}
108
2bd0ea18 109/*
d0e5f1ff 110 * Set all the fields in a bmap extent record from the uncompressed form.
2bd0ea18
NS
111 */
112void
d0e5f1ff
CH
113xfs_bmbt_disk_set_all(
114 struct xfs_bmbt_rec *r,
115 struct xfs_bmbt_irec *s)
2bd0ea18 116{
d0e5f1ff 117 int extent_flag = (s->br_state != XFS_EXT_NORM);
5e656dbb 118
d0e5f1ff
CH
119 ASSERT(s->br_state == XFS_EXT_NORM || s->br_state == XFS_EXT_UNWRITTEN);
120 ASSERT(!(s->br_startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)));
121 ASSERT(!(s->br_blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)));
122 ASSERT(!(s->br_startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)));
5e656dbb 123
c2e6c704 124 put_unaligned_be64(
5e656dbb 125 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
d0e5f1ff 126 ((xfs_bmbt_rec_base_t)s->br_startoff << 9) |
c2e6c704
CH
127 ((xfs_bmbt_rec_base_t)s->br_startblock >> 43), &r->l0);
128 put_unaligned_be64(
d0e5f1ff
CH
129 ((xfs_bmbt_rec_base_t)s->br_startblock << 21) |
130 ((xfs_bmbt_rec_base_t)s->br_blockcount &
c2e6c704 131 (xfs_bmbt_rec_base_t)xfs_mask64lo(21)), &r->l1);
2bd0ea18 132}
5e656dbb 133
2bd0ea18
NS
134/*
135 * Convert in-memory form of btree root to on-disk form.
136 */
137void
138xfs_bmbt_to_bmdr(
b3563c19
BN
139 struct xfs_mount *mp,
140 struct xfs_btree_block *rblock,
2bd0ea18
NS
141 int rblocklen,
142 xfs_bmdr_block_t *dblock,
143 int dblocklen)
144{
145 int dmxr;
146 xfs_bmbt_key_t *fkp;
5e656dbb 147 __be64 *fpp;
2bd0ea18 148 xfs_bmbt_key_t *tkp;
5e656dbb 149 __be64 *tpp;
2bd0ea18 150
5dfa5cd2
DC
151 if (xfs_sb_version_hascrc(&mp->m_sb)) {
152 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
9c4e12fb
ES
153 ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid,
154 &mp->m_sb.sb_meta_uuid));
5dfa5cd2
DC
155 ASSERT(rblock->bb_u.l.bb_blkno ==
156 cpu_to_be64(XFS_BUF_DADDR_NULL));
157 } else
158 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
5a35bf2c
DC
159 ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
160 ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
a2ceac1f 161 ASSERT(rblock->bb_level != 0);
6e3140c7
NS
162 dblock->bb_level = rblock->bb_level;
163 dblock->bb_numrecs = rblock->bb_numrecs;
ff105f75 164 dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
b3563c19
BN
165 fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
166 tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
167 fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
168 tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
6e3140c7 169 dmxr = be16_to_cpu(dblock->bb_numrecs);
32181a02 170 memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
5e656dbb 171 memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
2bd0ea18
NS
172}
173
b194c7d8
BN
174STATIC struct xfs_btree_cur *
175xfs_bmbt_dup_cursor(
176 struct xfs_btree_cur *cur)
177{
178 struct xfs_btree_cur *new;
179
180 new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
181 cur->bc_private.b.ip, cur->bc_private.b.whichfork);
182
183 /*
f33cea1a 184 * Copy the firstblock, dfops, and flags values,
b194c7d8
BN
185 * since init cursor doesn't get them.
186 */
187 new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
f33cea1a 188 new->bc_private.b.dfops = cur->bc_private.b.dfops;
b194c7d8
BN
189 new->bc_private.b.flags = cur->bc_private.b.flags;
190
191 return new;
192}
193
194STATIC void
195xfs_bmbt_update_cursor(
196 struct xfs_btree_cur *src,
197 struct xfs_btree_cur *dst)
198{
199 ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
200 (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
f33cea1a 201 ASSERT(dst->bc_private.b.dfops == src->bc_private.b.dfops);
b194c7d8
BN
202
203 dst->bc_private.b.allocated += src->bc_private.b.allocated;
204 dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
205
206 src->bc_private.b.allocated = 0;
207}
208
209STATIC int
210xfs_bmbt_alloc_block(
211 struct xfs_btree_cur *cur,
212 union xfs_btree_ptr *start,
213 union xfs_btree_ptr *new,
b194c7d8
BN
214 int *stat)
215{
216 xfs_alloc_arg_t args; /* block allocation args */
217 int error; /* error return value */
218
219 memset(&args, 0, sizeof(args));
220 args.tp = cur->bc_tp;
221 args.mp = cur->bc_mp;
222 args.fsbno = cur->bc_private.b.firstblock;
223 args.firstblock = args.fsbno;
85aec44f
DW
224 xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_private.b.ip->i_ino,
225 cur->bc_private.b.whichfork);
b194c7d8
BN
226
227 if (args.fsbno == NULLFSBLOCK) {
228 args.fsbno = be64_to_cpu(start->l);
229 args.type = XFS_ALLOCTYPE_START_BNO;
230 /*
231 * Make sure there is sufficient room left in the AG to
232 * complete a full tree split for an extent insert. If
233 * we are converting the middle part of an extent then
234 * we may need space for two tree splits.
235 *
236 * We are relying on the caller to make the correct block
237 * reservation for this operation to succeed. If the
238 * reservation amount is insufficient then we may fail a
239 * block allocation here and corrupt the filesystem.
240 */
0268fdc3 241 args.minleft = args.tp->t_blk_res;
f33cea1a 242 } else if (cur->bc_private.b.dfops->dop_low) {
b194c7d8
BN
243 args.type = XFS_ALLOCTYPE_START_BNO;
244 } else {
245 args.type = XFS_ALLOCTYPE_NEAR_BNO;
246 }
247
248 args.minlen = args.maxlen = args.prod = 1;
249 args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
0268fdc3 250 if (!args.wasdel && args.tp->t_blk_res == 0) {
12b53197 251 error = -ENOSPC;
b194c7d8
BN
252 goto error0;
253 }
254 error = xfs_alloc_vextent(&args);
255 if (error)
256 goto error0;
257
258 if (args.fsbno == NULLFSBLOCK && args.minleft) {
259 /*
260 * Could not find an AG with enough free space to satisfy
a3b4a951 261 * a full btree split. Try again and if
b194c7d8
BN
262 * successful activate the lowspace algorithm.
263 */
264 args.fsbno = 0;
265 args.type = XFS_ALLOCTYPE_FIRST_AG;
b194c7d8
BN
266 error = xfs_alloc_vextent(&args);
267 if (error)
268 goto error0;
f33cea1a 269 cur->bc_private.b.dfops->dop_low = true;
b194c7d8 270 }
c1587ecf 271 if (WARN_ON_ONCE(args.fsbno == NULLFSBLOCK)) {
b194c7d8
BN
272 *stat = 0;
273 return 0;
274 }
97b3ffd0 275
b194c7d8
BN
276 ASSERT(args.len == 1);
277 cur->bc_private.b.firstblock = args.fsbno;
278 cur->bc_private.b.allocated++;
279 cur->bc_private.b.ip->i_d.di_nblocks++;
280 xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
56b2de80 281 xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
b194c7d8
BN
282 XFS_TRANS_DQ_BCOUNT, 1L);
283
284 new->l = cpu_to_be64(args.fsbno);
285
b194c7d8
BN
286 *stat = 1;
287 return 0;
288
289 error0:
b194c7d8
BN
290 return error;
291}
292
293STATIC int
294xfs_bmbt_free_block(
295 struct xfs_btree_cur *cur,
296 struct xfs_buf *bp)
297{
298 struct xfs_mount *mp = cur->bc_mp;
299 struct xfs_inode *ip = cur->bc_private.b.ip;
300 struct xfs_trans *tp = cur->bc_tp;
301 xfs_fsblock_t fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
85aec44f 302 struct xfs_owner_info oinfo;
b194c7d8 303
85aec44f
DW
304 xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, cur->bc_private.b.whichfork);
305 xfs_bmap_add_free(mp, cur->bc_private.b.dfops, fsbno, 1, &oinfo);
b194c7d8
BN
306 ip->i_d.di_nblocks--;
307
308 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
56b2de80 309 xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
b194c7d8
BN
310 return 0;
311}
312
313STATIC int
314xfs_bmbt_get_minrecs(
315 struct xfs_btree_cur *cur,
316 int level)
317{
b3563c19
BN
318 if (level == cur->bc_nlevels - 1) {
319 struct xfs_ifork *ifp;
320
321 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
322 cur->bc_private.b.whichfork);
323
324 return xfs_bmbt_maxrecs(cur->bc_mp,
325 ifp->if_broot_bytes, level == 0) / 2;
326 }
327
328 return cur->bc_mp->m_bmap_dmnr[level != 0];
b194c7d8
BN
329}
330
b3563c19 331int
b194c7d8
BN
332xfs_bmbt_get_maxrecs(
333 struct xfs_btree_cur *cur,
334 int level)
335{
b3563c19
BN
336 if (level == cur->bc_nlevels - 1) {
337 struct xfs_ifork *ifp;
338
339 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
340 cur->bc_private.b.whichfork);
341
342 return xfs_bmbt_maxrecs(cur->bc_mp,
343 ifp->if_broot_bytes, level == 0);
344 }
345
346 return cur->bc_mp->m_bmap_dmxr[level != 0];
347
b194c7d8
BN
348}
349
350/*
351 * Get the maximum records we could store in the on-disk format.
352 *
353 * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
354 * for the root node this checks the available space in the dinode fork
355 * so that we can resize the in-memory buffer to match it. After a
356 * resize to the maximum size this function returns the same value
357 * as xfs_bmbt_get_maxrecs for the root node, too.
358 */
359STATIC int
360xfs_bmbt_get_dmaxrecs(
361 struct xfs_btree_cur *cur,
362 int level)
363{
b3563c19
BN
364 if (level != cur->bc_nlevels - 1)
365 return cur->bc_mp->m_bmap_dmxr[level != 0];
ff105f75 366 return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
b194c7d8
BN
367}
368
369STATIC void
370xfs_bmbt_init_key_from_rec(
371 union xfs_btree_key *key,
372 union xfs_btree_rec *rec)
373{
374 key->bmbt.br_startoff =
375 cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
376}
377
4e388bc9
DW
378STATIC void
379xfs_bmbt_init_high_key_from_rec(
380 union xfs_btree_key *key,
381 union xfs_btree_rec *rec)
382{
383 key->bmbt.br_startoff = cpu_to_be64(
384 xfs_bmbt_disk_get_startoff(&rec->bmbt) +
385 xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
386}
387
b194c7d8
BN
388STATIC void
389xfs_bmbt_init_rec_from_cur(
390 struct xfs_btree_cur *cur,
391 union xfs_btree_rec *rec)
392{
393 xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
394}
395
396STATIC void
397xfs_bmbt_init_ptr_from_cur(
398 struct xfs_btree_cur *cur,
399 union xfs_btree_ptr *ptr)
400{
401 ptr->l = 0;
402}
403
4a492e72 404STATIC int64_t
b194c7d8
BN
405xfs_bmbt_key_diff(
406 struct xfs_btree_cur *cur,
407 union xfs_btree_key *key)
408{
4a492e72 409 return (int64_t)be64_to_cpu(key->bmbt.br_startoff) -
b194c7d8
BN
410 cur->bc_rec.b.br_startoff;
411}
412
4e388bc9
DW
413STATIC int64_t
414xfs_bmbt_diff_two_keys(
415 struct xfs_btree_cur *cur,
416 union xfs_btree_key *k1,
417 union xfs_btree_key *k2)
418{
419 return (int64_t)be64_to_cpu(k1->bmbt.br_startoff) -
420 be64_to_cpu(k2->bmbt.br_startoff);
421}
422
bc01119d 423static xfs_failaddr_t
a2ceac1f
DC
424xfs_bmbt_verify(
425 struct xfs_buf *bp)
426{
427 struct xfs_mount *mp = bp->b_target->bt_mount;
428 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
bc01119d 429 xfs_failaddr_t fa;
a2ceac1f 430 unsigned int level;
a2ceac1f 431
c0adfb03
DC
432 switch (block->bb_magic) {
433 case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
5dfa5cd2
DC
434 /*
435 * XXX: need a better way of verifying the owner here. Right now
436 * just make sure there has been one set.
437 */
bc01119d
DW
438 fa = xfs_btree_lblock_v5hdr_verify(bp, XFS_RMAP_OWN_UNKNOWN);
439 if (fa)
440 return fa;
5dfa5cd2 441 /* fall through */
c0adfb03 442 case cpu_to_be32(XFS_BMAP_MAGIC):
5dfa5cd2
DC
443 break;
444 default:
bc01119d 445 return __this_address;
5dfa5cd2
DC
446 }
447
448 /*
449 * numrecs and level verification.
a2ceac1f 450 *
5dfa5cd2 451 * We don't know what fork we belong to, so just verify that the level
a2ceac1f
DC
452 * is less than the maximum of the two. Later checks will be more
453 * precise.
454 */
455 level = be16_to_cpu(block->bb_level);
e6d77a21 456 if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
bc01119d 457 return __this_address;
5dfa5cd2 458
fb75464f 459 return xfs_btree_lblock_verify(bp, mp->m_bmap_dmxr[level != 0]);
a2ceac1f
DC
460}
461
462static void
463xfs_bmbt_read_verify(
464 struct xfs_buf *bp)
465{
1e697959
DW
466 xfs_failaddr_t fa;
467
45922933 468 if (!xfs_btree_lblock_verify_crc(bp))
1e697959
DW
469 xfs_verifier_error(bp, -EFSBADCRC, __this_address);
470 else {
471 fa = xfs_bmbt_verify(bp);
472 if (fa)
473 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
474 }
45922933 475
7e6c95f1 476 if (bp->b_error)
45922933 477 trace_xfs_btree_corrupt(bp, _RET_IP_);
a2ceac1f
DC
478}
479
480static void
481xfs_bmbt_write_verify(
482 struct xfs_buf *bp)
483{
1e697959
DW
484 xfs_failaddr_t fa;
485
486 fa = xfs_bmbt_verify(bp);
487 if (fa) {
5dfa5cd2 488 trace_xfs_btree_corrupt(bp, _RET_IP_);
1e697959 489 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
5dfa5cd2
DC
490 return;
491 }
492 xfs_btree_lblock_calc_crc(bp);
a2ceac1f
DC
493}
494
495const struct xfs_buf_ops xfs_bmbt_buf_ops = {
a3fac935 496 .name = "xfs_bmbt",
a2ceac1f
DC
497 .verify_read = xfs_bmbt_read_verify,
498 .verify_write = xfs_bmbt_write_verify,
95d9582b 499 .verify_struct = xfs_bmbt_verify,
a2ceac1f
DC
500};
501
502
b194c7d8
BN
503STATIC int
504xfs_bmbt_keys_inorder(
505 struct xfs_btree_cur *cur,
506 union xfs_btree_key *k1,
507 union xfs_btree_key *k2)
508{
509 return be64_to_cpu(k1->bmbt.br_startoff) <
510 be64_to_cpu(k2->bmbt.br_startoff);
511}
512
513STATIC int
514xfs_bmbt_recs_inorder(
515 struct xfs_btree_cur *cur,
516 union xfs_btree_rec *r1,
517 union xfs_btree_rec *r2)
518{
519 return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
520 xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
521 xfs_bmbt_disk_get_startoff(&r2->bmbt);
522}
b194c7d8 523
b194c7d8
BN
524static const struct xfs_btree_ops xfs_bmbt_ops = {
525 .rec_len = sizeof(xfs_bmbt_rec_t),
526 .key_len = sizeof(xfs_bmbt_key_t),
527
528 .dup_cursor = xfs_bmbt_dup_cursor,
529 .update_cursor = xfs_bmbt_update_cursor,
530 .alloc_block = xfs_bmbt_alloc_block,
531 .free_block = xfs_bmbt_free_block,
532 .get_maxrecs = xfs_bmbt_get_maxrecs,
533 .get_minrecs = xfs_bmbt_get_minrecs,
534 .get_dmaxrecs = xfs_bmbt_get_dmaxrecs,
535 .init_key_from_rec = xfs_bmbt_init_key_from_rec,
4e388bc9 536 .init_high_key_from_rec = xfs_bmbt_init_high_key_from_rec,
b194c7d8
BN
537 .init_rec_from_cur = xfs_bmbt_init_rec_from_cur,
538 .init_ptr_from_cur = xfs_bmbt_init_ptr_from_cur,
539 .key_diff = xfs_bmbt_key_diff,
4e388bc9 540 .diff_two_keys = xfs_bmbt_diff_two_keys,
a2ceac1f 541 .buf_ops = &xfs_bmbt_buf_ops,
b194c7d8
BN
542 .keys_inorder = xfs_bmbt_keys_inorder,
543 .recs_inorder = xfs_bmbt_recs_inorder,
b194c7d8
BN
544};
545
546/*
547 * Allocate a new bmap btree cursor.
548 */
549struct xfs_btree_cur * /* new bmap btree cursor */
550xfs_bmbt_init_cursor(
551 struct xfs_mount *mp, /* file system mount point */
552 struct xfs_trans *tp, /* transaction pointer */
553 struct xfs_inode *ip, /* inode owning the btree */
554 int whichfork) /* data or attr fork */
555{
556 struct xfs_ifork *ifp = XFS_IFORK_PTR(ip, whichfork);
557 struct xfs_btree_cur *cur;
cb8a004a 558 ASSERT(whichfork != XFS_COW_FORK);
b194c7d8 559
762989ef 560 cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
b194c7d8
BN
561
562 cur->bc_tp = tp;
563 cur->bc_mp = mp;
564 cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
565 cur->bc_btnum = XFS_BTNUM_BMAP;
566 cur->bc_blocklog = mp->m_sb.sb_blocklog;
5d8acc46 567 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_bmbt_2);
b194c7d8
BN
568
569 cur->bc_ops = &xfs_bmbt_ops;
570 cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
5dfa5cd2
DC
571 if (xfs_sb_version_hascrc(&mp->m_sb))
572 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
b194c7d8
BN
573
574 cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
575 cur->bc_private.b.ip = ip;
576 cur->bc_private.b.firstblock = NULLFSBLOCK;
f33cea1a 577 cur->bc_private.b.dfops = NULL;
b194c7d8
BN
578 cur->bc_private.b.allocated = 0;
579 cur->bc_private.b.flags = 0;
580 cur->bc_private.b.whichfork = whichfork;
581
582 return cur;
583}
b3563c19
BN
584
585/*
586 * Calculate number of records in a bmap btree block.
587 */
588int
589xfs_bmbt_maxrecs(
590 struct xfs_mount *mp,
591 int blocklen,
592 int leaf)
593{
594 blocklen -= XFS_BMBT_BLOCK_LEN(mp);
595
596 if (leaf)
597 return blocklen / sizeof(xfs_bmbt_rec_t);
598 return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
599}
600
601/*
602 * Calculate number of records in a bmap btree inode root.
603 */
604int
605xfs_bmdr_maxrecs(
b3563c19
BN
606 int blocklen,
607 int leaf)
608{
609 blocklen -= sizeof(xfs_bmdr_block_t);
610
611 if (leaf)
612 return blocklen / sizeof(xfs_bmdr_rec_t);
613 return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
614}
9c6ebc42
DC
615
616/*
ff105f75 617 * Change the owner of a btree format fork fo the inode passed in. Change it to
9c6ebc42
DC
618 * the owner of that is passed in so that we can change owners before or after
619 * we switch forks between inodes. The operation that the caller is doing will
620 * determine whether is needs to change owner before or after the switch.
621 *
622 * For demand paged transactional modification, the fork switch should be done
623 * after reading in all the blocks, modifying them and pinning them in the
624 * transaction. For modification when the buffers are already pinned in memory,
625 * the fork switch can be done before changing the owner as we won't need to
626 * validate the owner until the btree buffers are unpinned and writes can occur
627 * again.
628 *
629 * For recovery based ownership change, there is no transactional context and
630 * so a buffer list must be supplied so that we can record the buffers that we
631 * modified for the caller to issue IO on.
632 */
633int
634xfs_bmbt_change_owner(
635 struct xfs_trans *tp,
636 struct xfs_inode *ip,
637 int whichfork,
638 xfs_ino_t new_owner,
639 struct list_head *buffer_list)
640{
641 struct xfs_btree_cur *cur;
642 int error;
643
644 ASSERT(tp || buffer_list);
645 ASSERT(!(tp && buffer_list));
646 if (whichfork == XFS_DATA_FORK)
647 ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
648 else
649 ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
650
651 cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
652 if (!cur)
12b53197 653 return -ENOMEM;
c7ba1731 654 cur->bc_private.b.flags |= XFS_BTCUR_BPRV_INVALID_OWNER;
9c6ebc42
DC
655
656 error = xfs_btree_change_owner(cur, new_owner, buffer_list);
657 xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
658 return error;
659}
5f728383
DW
660
661/* Calculate the bmap btree size for some records. */
662unsigned long long
663xfs_bmbt_calc_size(
664 struct xfs_mount *mp,
665 unsigned long long len)
666{
667 return xfs_btree_calc_size(mp->m_bmap_dmnr, len);
668}