aboutsummaryrefslogtreecommitdiffstats
path: root/fs/xfs/libxfs/xfs_bmap_btree.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/xfs/libxfs/xfs_bmap_btree.c')
-rw-r--r--fs/xfs/libxfs/xfs_bmap_btree.c967
1 files changed, 967 insertions, 0 deletions
diff --git a/fs/xfs/libxfs/xfs_bmap_btree.c b/fs/xfs/libxfs/xfs_bmap_btree.c
new file mode 100644
index 000000000000..a388de4ceaa1
--- /dev/null
+++ b/fs/xfs/libxfs/xfs_bmap_btree.c
@@ -0,0 +1,967 @@
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_sb.h"
26#include "xfs_ag.h"
27#include "xfs_mount.h"
28#include "xfs_inode.h"
29#include "xfs_trans.h"
30#include "xfs_inode_item.h"
31#include "xfs_alloc.h"
32#include "xfs_btree.h"
33#include "xfs_bmap_btree.h"
34#include "xfs_bmap.h"
35#include "xfs_error.h"
36#include "xfs_quota.h"
37#include "xfs_trace.h"
38#include "xfs_cksum.h"
39#include "xfs_dinode.h"
40
41/*
42 * Determine the extent state.
43 */
44/* ARGSUSED */
45STATIC xfs_exntst_t
46xfs_extent_state(
47 xfs_filblks_t blks,
48 int extent_flag)
49{
50 if (extent_flag) {
51 ASSERT(blks != 0); /* saved for DMIG */
52 return XFS_EXT_UNWRITTEN;
53 }
54 return XFS_EXT_NORM;
55}
56
57/*
58 * Convert on-disk form of btree root to in-memory form.
59 */
60void
61xfs_bmdr_to_bmbt(
62 struct xfs_inode *ip,
63 xfs_bmdr_block_t *dblock,
64 int dblocklen,
65 struct xfs_btree_block *rblock,
66 int rblocklen)
67{
68 struct xfs_mount *mp = ip->i_mount;
69 int dmxr;
70 xfs_bmbt_key_t *fkp;
71 __be64 *fpp;
72 xfs_bmbt_key_t *tkp;
73 __be64 *tpp;
74
75 if (xfs_sb_version_hascrc(&mp->m_sb))
76 xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
77 XFS_BMAP_CRC_MAGIC, 0, 0, ip->i_ino,
78 XFS_BTREE_LONG_PTRS | XFS_BTREE_CRC_BLOCKS);
79 else
80 xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
81 XFS_BMAP_MAGIC, 0, 0, ip->i_ino,
82 XFS_BTREE_LONG_PTRS);
83
84 rblock->bb_level = dblock->bb_level;
85 ASSERT(be16_to_cpu(rblock->bb_level) > 0);
86 rblock->bb_numrecs = dblock->bb_numrecs;
87 dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
88 fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
89 tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
90 fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
91 tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
92 dmxr = be16_to_cpu(dblock->bb_numrecs);
93 memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
94 memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
95}
96
97/*
98 * Convert a compressed bmap extent record to an uncompressed form.
99 * This code must be in sync with the routines xfs_bmbt_get_startoff,
100 * xfs_bmbt_get_startblock, xfs_bmbt_get_blockcount and xfs_bmbt_get_state.
101 */
102STATIC void
103__xfs_bmbt_get_all(
104 __uint64_t l0,
105 __uint64_t l1,
106 xfs_bmbt_irec_t *s)
107{
108 int ext_flag;
109 xfs_exntst_t st;
110
111 ext_flag = (int)(l0 >> (64 - BMBT_EXNTFLAG_BITLEN));
112 s->br_startoff = ((xfs_fileoff_t)l0 &
113 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
114#if XFS_BIG_BLKNOS
115 s->br_startblock = (((xfs_fsblock_t)l0 & xfs_mask64lo(9)) << 43) |
116 (((xfs_fsblock_t)l1) >> 21);
117#else
118#ifdef DEBUG
119 {
120 xfs_dfsbno_t b;
121
122 b = (((xfs_dfsbno_t)l0 & xfs_mask64lo(9)) << 43) |
123 (((xfs_dfsbno_t)l1) >> 21);
124 ASSERT((b >> 32) == 0 || isnulldstartblock(b));
125 s->br_startblock = (xfs_fsblock_t)b;
126 }
127#else /* !DEBUG */
128 s->br_startblock = (xfs_fsblock_t)(((xfs_dfsbno_t)l1) >> 21);
129#endif /* DEBUG */
130#endif /* XFS_BIG_BLKNOS */
131 s->br_blockcount = (xfs_filblks_t)(l1 & xfs_mask64lo(21));
132 /* This is xfs_extent_state() in-line */
133 if (ext_flag) {
134 ASSERT(s->br_blockcount != 0); /* saved for DMIG */
135 st = XFS_EXT_UNWRITTEN;
136 } else
137 st = XFS_EXT_NORM;
138 s->br_state = st;
139}
140
141void
142xfs_bmbt_get_all(
143 xfs_bmbt_rec_host_t *r,
144 xfs_bmbt_irec_t *s)
145{
146 __xfs_bmbt_get_all(r->l0, r->l1, s);
147}
148
149/*
150 * Extract the blockcount field from an in memory bmap extent record.
151 */
152xfs_filblks_t
153xfs_bmbt_get_blockcount(
154 xfs_bmbt_rec_host_t *r)
155{
156 return (xfs_filblks_t)(r->l1 & xfs_mask64lo(21));
157}
158
159/*
160 * Extract the startblock field from an in memory bmap extent record.
161 */
162xfs_fsblock_t
163xfs_bmbt_get_startblock(
164 xfs_bmbt_rec_host_t *r)
165{
166#if XFS_BIG_BLKNOS
167 return (((xfs_fsblock_t)r->l0 & xfs_mask64lo(9)) << 43) |
168 (((xfs_fsblock_t)r->l1) >> 21);
169#else
170#ifdef DEBUG
171 xfs_dfsbno_t b;
172
173 b = (((xfs_dfsbno_t)r->l0 & xfs_mask64lo(9)) << 43) |
174 (((xfs_dfsbno_t)r->l1) >> 21);
175 ASSERT((b >> 32) == 0 || isnulldstartblock(b));
176 return (xfs_fsblock_t)b;
177#else /* !DEBUG */
178 return (xfs_fsblock_t)(((xfs_dfsbno_t)r->l1) >> 21);
179#endif /* DEBUG */
180#endif /* XFS_BIG_BLKNOS */
181}
182
183/*
184 * Extract the startoff field from an in memory bmap extent record.
185 */
186xfs_fileoff_t
187xfs_bmbt_get_startoff(
188 xfs_bmbt_rec_host_t *r)
189{
190 return ((xfs_fileoff_t)r->l0 &
191 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
192}
193
194xfs_exntst_t
195xfs_bmbt_get_state(
196 xfs_bmbt_rec_host_t *r)
197{
198 int ext_flag;
199
200 ext_flag = (int)((r->l0) >> (64 - BMBT_EXNTFLAG_BITLEN));
201 return xfs_extent_state(xfs_bmbt_get_blockcount(r),
202 ext_flag);
203}
204
205/*
206 * Extract the blockcount field from an on disk bmap extent record.
207 */
208xfs_filblks_t
209xfs_bmbt_disk_get_blockcount(
210 xfs_bmbt_rec_t *r)
211{
212 return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
213}
214
215/*
216 * Extract the startoff field from a disk format bmap extent record.
217 */
218xfs_fileoff_t
219xfs_bmbt_disk_get_startoff(
220 xfs_bmbt_rec_t *r)
221{
222 return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
223 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
224}
225
226
227/*
228 * Set all the fields in a bmap extent record from the arguments.
229 */
230void
231xfs_bmbt_set_allf(
232 xfs_bmbt_rec_host_t *r,
233 xfs_fileoff_t startoff,
234 xfs_fsblock_t startblock,
235 xfs_filblks_t blockcount,
236 xfs_exntst_t state)
237{
238 int extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
239
240 ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
241 ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
242 ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
243
244#if XFS_BIG_BLKNOS
245 ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
246
247 r->l0 = ((xfs_bmbt_rec_base_t)extent_flag << 63) |
248 ((xfs_bmbt_rec_base_t)startoff << 9) |
249 ((xfs_bmbt_rec_base_t)startblock >> 43);
250 r->l1 = ((xfs_bmbt_rec_base_t)startblock << 21) |
251 ((xfs_bmbt_rec_base_t)blockcount &
252 (xfs_bmbt_rec_base_t)xfs_mask64lo(21));
253#else /* !XFS_BIG_BLKNOS */
254 if (isnullstartblock(startblock)) {
255 r->l0 = ((xfs_bmbt_rec_base_t)extent_flag << 63) |
256 ((xfs_bmbt_rec_base_t)startoff << 9) |
257 (xfs_bmbt_rec_base_t)xfs_mask64lo(9);
258 r->l1 = xfs_mask64hi(11) |
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 } else {
263 r->l0 = ((xfs_bmbt_rec_base_t)extent_flag << 63) |
264 ((xfs_bmbt_rec_base_t)startoff << 9);
265 r->l1 = ((xfs_bmbt_rec_base_t)startblock << 21) |
266 ((xfs_bmbt_rec_base_t)blockcount &
267 (xfs_bmbt_rec_base_t)xfs_mask64lo(21));
268 }
269#endif /* XFS_BIG_BLKNOS */
270}
271
272/*
273 * Set all the fields in a bmap extent record from the uncompressed form.
274 */
275void
276xfs_bmbt_set_all(
277 xfs_bmbt_rec_host_t *r,
278 xfs_bmbt_irec_t *s)
279{
280 xfs_bmbt_set_allf(r, s->br_startoff, s->br_startblock,
281 s->br_blockcount, s->br_state);
282}
283
284
285/*
286 * Set all the fields in a disk format bmap extent record from the arguments.
287 */
288void
289xfs_bmbt_disk_set_allf(
290 xfs_bmbt_rec_t *r,
291 xfs_fileoff_t startoff,
292 xfs_fsblock_t startblock,
293 xfs_filblks_t blockcount,
294 xfs_exntst_t state)
295{
296 int extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
297
298 ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
299 ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
300 ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
301
302#if XFS_BIG_BLKNOS
303 ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
304
305 r->l0 = cpu_to_be64(
306 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
307 ((xfs_bmbt_rec_base_t)startoff << 9) |
308 ((xfs_bmbt_rec_base_t)startblock >> 43));
309 r->l1 = cpu_to_be64(
310 ((xfs_bmbt_rec_base_t)startblock << 21) |
311 ((xfs_bmbt_rec_base_t)blockcount &
312 (xfs_bmbt_rec_base_t)xfs_mask64lo(21)));
313#else /* !XFS_BIG_BLKNOS */
314 if (isnullstartblock(startblock)) {
315 r->l0 = cpu_to_be64(
316 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
317 ((xfs_bmbt_rec_base_t)startoff << 9) |
318 (xfs_bmbt_rec_base_t)xfs_mask64lo(9));
319 r->l1 = cpu_to_be64(xfs_mask64hi(11) |
320 ((xfs_bmbt_rec_base_t)startblock << 21) |
321 ((xfs_bmbt_rec_base_t)blockcount &
322 (xfs_bmbt_rec_base_t)xfs_mask64lo(21)));
323 } else {
324 r->l0 = cpu_to_be64(
325 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
326 ((xfs_bmbt_rec_base_t)startoff << 9));
327 r->l1 = cpu_to_be64(
328 ((xfs_bmbt_rec_base_t)startblock << 21) |
329 ((xfs_bmbt_rec_base_t)blockcount &
330 (xfs_bmbt_rec_base_t)xfs_mask64lo(21)));
331 }
332#endif /* XFS_BIG_BLKNOS */
333}
334
335/*
336 * Set all the fields in a bmap extent record from the uncompressed form.
337 */
338STATIC void
339xfs_bmbt_disk_set_all(
340 xfs_bmbt_rec_t *r,
341 xfs_bmbt_irec_t *s)
342{
343 xfs_bmbt_disk_set_allf(r, s->br_startoff, s->br_startblock,
344 s->br_blockcount, s->br_state);
345}
346
347/*
348 * Set the blockcount field in a bmap extent record.
349 */
350void
351xfs_bmbt_set_blockcount(
352 xfs_bmbt_rec_host_t *r,
353 xfs_filblks_t v)
354{
355 ASSERT((v & xfs_mask64hi(43)) == 0);
356 r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64hi(43)) |
357 (xfs_bmbt_rec_base_t)(v & xfs_mask64lo(21));
358}
359
360/*
361 * Set the startblock field in a bmap extent record.
362 */
363void
364xfs_bmbt_set_startblock(
365 xfs_bmbt_rec_host_t *r,
366 xfs_fsblock_t v)
367{
368#if XFS_BIG_BLKNOS
369 ASSERT((v & xfs_mask64hi(12)) == 0);
370 r->l0 = (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64hi(55)) |
371 (xfs_bmbt_rec_base_t)(v >> 43);
372 r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64lo(21)) |
373 (xfs_bmbt_rec_base_t)(v << 21);
374#else /* !XFS_BIG_BLKNOS */
375 if (isnullstartblock(v)) {
376 r->l0 |= (xfs_bmbt_rec_base_t)xfs_mask64lo(9);
377 r->l1 = (xfs_bmbt_rec_base_t)xfs_mask64hi(11) |
378 ((xfs_bmbt_rec_base_t)v << 21) |
379 (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64lo(21));
380 } else {
381 r->l0 &= ~(xfs_bmbt_rec_base_t)xfs_mask64lo(9);
382 r->l1 = ((xfs_bmbt_rec_base_t)v << 21) |
383 (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64lo(21));
384 }
385#endif /* XFS_BIG_BLKNOS */
386}
387
388/*
389 * Set the startoff field in a bmap extent record.
390 */
391void
392xfs_bmbt_set_startoff(
393 xfs_bmbt_rec_host_t *r,
394 xfs_fileoff_t v)
395{
396 ASSERT((v & xfs_mask64hi(9)) == 0);
397 r->l0 = (r->l0 & (xfs_bmbt_rec_base_t) xfs_mask64hi(1)) |
398 ((xfs_bmbt_rec_base_t)v << 9) |
399 (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64lo(9));
400}
401
402/*
403 * Set the extent state field in a bmap extent record.
404 */
405void
406xfs_bmbt_set_state(
407 xfs_bmbt_rec_host_t *r,
408 xfs_exntst_t v)
409{
410 ASSERT(v == XFS_EXT_NORM || v == XFS_EXT_UNWRITTEN);
411 if (v == XFS_EXT_NORM)
412 r->l0 &= xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN);
413 else
414 r->l0 |= xfs_mask64hi(BMBT_EXNTFLAG_BITLEN);
415}
416
417/*
418 * Convert in-memory form of btree root to on-disk form.
419 */
420void
421xfs_bmbt_to_bmdr(
422 struct xfs_mount *mp,
423 struct xfs_btree_block *rblock,
424 int rblocklen,
425 xfs_bmdr_block_t *dblock,
426 int dblocklen)
427{
428 int dmxr;
429 xfs_bmbt_key_t *fkp;
430 __be64 *fpp;
431 xfs_bmbt_key_t *tkp;
432 __be64 *tpp;
433
434 if (xfs_sb_version_hascrc(&mp->m_sb)) {
435 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
436 ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid, &mp->m_sb.sb_uuid));
437 ASSERT(rblock->bb_u.l.bb_blkno ==
438 cpu_to_be64(XFS_BUF_DADDR_NULL));
439 } else
440 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
441 ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLDFSBNO));
442 ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLDFSBNO));
443 ASSERT(rblock->bb_level != 0);
444 dblock->bb_level = rblock->bb_level;
445 dblock->bb_numrecs = rblock->bb_numrecs;
446 dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
447 fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
448 tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
449 fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
450 tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
451 dmxr = be16_to_cpu(dblock->bb_numrecs);
452 memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
453 memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
454}
455
456/*
457 * Check extent records, which have just been read, for
458 * any bit in the extent flag field. ASSERT on debug
459 * kernels, as this condition should not occur.
460 * Return an error condition (1) if any flags found,
461 * otherwise return 0.
462 */
463
464int
465xfs_check_nostate_extents(
466 xfs_ifork_t *ifp,
467 xfs_extnum_t idx,
468 xfs_extnum_t num)
469{
470 for (; num > 0; num--, idx++) {
471 xfs_bmbt_rec_host_t *ep = xfs_iext_get_ext(ifp, idx);
472 if ((ep->l0 >>
473 (64 - BMBT_EXNTFLAG_BITLEN)) != 0) {
474 ASSERT(0);
475 return 1;
476 }
477 }
478 return 0;
479}
480
481
482STATIC struct xfs_btree_cur *
483xfs_bmbt_dup_cursor(
484 struct xfs_btree_cur *cur)
485{
486 struct xfs_btree_cur *new;
487
488 new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
489 cur->bc_private.b.ip, cur->bc_private.b.whichfork);
490
491 /*
492 * Copy the firstblock, flist, and flags values,
493 * since init cursor doesn't get them.
494 */
495 new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
496 new->bc_private.b.flist = cur->bc_private.b.flist;
497 new->bc_private.b.flags = cur->bc_private.b.flags;
498
499 return new;
500}
501
502STATIC void
503xfs_bmbt_update_cursor(
504 struct xfs_btree_cur *src,
505 struct xfs_btree_cur *dst)
506{
507 ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
508 (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
509 ASSERT(dst->bc_private.b.flist == src->bc_private.b.flist);
510
511 dst->bc_private.b.allocated += src->bc_private.b.allocated;
512 dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
513
514 src->bc_private.b.allocated = 0;
515}
516
517STATIC int
518xfs_bmbt_alloc_block(
519 struct xfs_btree_cur *cur,
520 union xfs_btree_ptr *start,
521 union xfs_btree_ptr *new,
522 int *stat)
523{
524 xfs_alloc_arg_t args; /* block allocation args */
525 int error; /* error return value */
526
527 memset(&args, 0, sizeof(args));
528 args.tp = cur->bc_tp;
529 args.mp = cur->bc_mp;
530 args.fsbno = cur->bc_private.b.firstblock;
531 args.firstblock = args.fsbno;
532
533 if (args.fsbno == NULLFSBLOCK) {
534 args.fsbno = be64_to_cpu(start->l);
535 args.type = XFS_ALLOCTYPE_START_BNO;
536 /*
537 * Make sure there is sufficient room left in the AG to
538 * complete a full tree split for an extent insert. If
539 * we are converting the middle part of an extent then
540 * we may need space for two tree splits.
541 *
542 * We are relying on the caller to make the correct block
543 * reservation for this operation to succeed. If the
544 * reservation amount is insufficient then we may fail a
545 * block allocation here and corrupt the filesystem.
546 */
547 args.minleft = xfs_trans_get_block_res(args.tp);
548 } else if (cur->bc_private.b.flist->xbf_low) {
549 args.type = XFS_ALLOCTYPE_START_BNO;
550 } else {
551 args.type = XFS_ALLOCTYPE_NEAR_BNO;
552 }
553
554 args.minlen = args.maxlen = args.prod = 1;
555 args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
556 if (!args.wasdel && xfs_trans_get_block_res(args.tp) == 0) {
557 error = -ENOSPC;
558 goto error0;
559 }
560 error = xfs_alloc_vextent(&args);
561 if (error)
562 goto error0;
563
564 if (args.fsbno == NULLFSBLOCK && args.minleft) {
565 /*
566 * Could not find an AG with enough free space to satisfy
567 * a full btree split. Try again without minleft and if
568 * successful activate the lowspace algorithm.
569 */
570 args.fsbno = 0;
571 args.type = XFS_ALLOCTYPE_FIRST_AG;
572 args.minleft = 0;
573 error = xfs_alloc_vextent(&args);
574 if (error)
575 goto error0;
576 cur->bc_private.b.flist->xbf_low = 1;
577 }
578 if (args.fsbno == NULLFSBLOCK) {
579 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
580 *stat = 0;
581 return 0;
582 }
583 ASSERT(args.len == 1);
584 cur->bc_private.b.firstblock = args.fsbno;
585 cur->bc_private.b.allocated++;
586 cur->bc_private.b.ip->i_d.di_nblocks++;
587 xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
588 xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
589 XFS_TRANS_DQ_BCOUNT, 1L);
590
591 new->l = cpu_to_be64(args.fsbno);
592
593 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
594 *stat = 1;
595 return 0;
596
597 error0:
598 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
599 return error;
600}
601
602STATIC int
603xfs_bmbt_free_block(
604 struct xfs_btree_cur *cur,
605 struct xfs_buf *bp)
606{
607 struct xfs_mount *mp = cur->bc_mp;
608 struct xfs_inode *ip = cur->bc_private.b.ip;
609 struct xfs_trans *tp = cur->bc_tp;
610 xfs_fsblock_t fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
611
612 xfs_bmap_add_free(fsbno, 1, cur->bc_private.b.flist, mp);
613 ip->i_d.di_nblocks--;
614
615 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
616 xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
617 xfs_trans_binval(tp, bp);
618 return 0;
619}
620
621STATIC int
622xfs_bmbt_get_minrecs(
623 struct xfs_btree_cur *cur,
624 int level)
625{
626 if (level == cur->bc_nlevels - 1) {
627 struct xfs_ifork *ifp;
628
629 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
630 cur->bc_private.b.whichfork);
631
632 return xfs_bmbt_maxrecs(cur->bc_mp,
633 ifp->if_broot_bytes, level == 0) / 2;
634 }
635
636 return cur->bc_mp->m_bmap_dmnr[level != 0];
637}
638
639int
640xfs_bmbt_get_maxrecs(
641 struct xfs_btree_cur *cur,
642 int level)
643{
644 if (level == cur->bc_nlevels - 1) {
645 struct xfs_ifork *ifp;
646
647 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
648 cur->bc_private.b.whichfork);
649
650 return xfs_bmbt_maxrecs(cur->bc_mp,
651 ifp->if_broot_bytes, level == 0);
652 }
653
654 return cur->bc_mp->m_bmap_dmxr[level != 0];
655
656}
657
658/*
659 * Get the maximum records we could store in the on-disk format.
660 *
661 * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
662 * for the root node this checks the available space in the dinode fork
663 * so that we can resize the in-memory buffer to match it. After a
664 * resize to the maximum size this function returns the same value
665 * as xfs_bmbt_get_maxrecs for the root node, too.
666 */
667STATIC int
668xfs_bmbt_get_dmaxrecs(
669 struct xfs_btree_cur *cur,
670 int level)
671{
672 if (level != cur->bc_nlevels - 1)
673 return cur->bc_mp->m_bmap_dmxr[level != 0];
674 return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
675}
676
677STATIC void
678xfs_bmbt_init_key_from_rec(
679 union xfs_btree_key *key,
680 union xfs_btree_rec *rec)
681{
682 key->bmbt.br_startoff =
683 cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
684}
685
686STATIC void
687xfs_bmbt_init_rec_from_key(
688 union xfs_btree_key *key,
689 union xfs_btree_rec *rec)
690{
691 ASSERT(key->bmbt.br_startoff != 0);
692
693 xfs_bmbt_disk_set_allf(&rec->bmbt, be64_to_cpu(key->bmbt.br_startoff),
694 0, 0, XFS_EXT_NORM);
695}
696
697STATIC void
698xfs_bmbt_init_rec_from_cur(
699 struct xfs_btree_cur *cur,
700 union xfs_btree_rec *rec)
701{
702 xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
703}
704
705STATIC void
706xfs_bmbt_init_ptr_from_cur(
707 struct xfs_btree_cur *cur,
708 union xfs_btree_ptr *ptr)
709{
710 ptr->l = 0;
711}
712
713STATIC __int64_t
714xfs_bmbt_key_diff(
715 struct xfs_btree_cur *cur,
716 union xfs_btree_key *key)
717{
718 return (__int64_t)be64_to_cpu(key->bmbt.br_startoff) -
719 cur->bc_rec.b.br_startoff;
720}
721
722static bool
723xfs_bmbt_verify(
724 struct xfs_buf *bp)
725{
726 struct xfs_mount *mp = bp->b_target->bt_mount;
727 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
728 unsigned int level;
729
730 switch (block->bb_magic) {
731 case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
732 if (!xfs_sb_version_hascrc(&mp->m_sb))
733 return false;
734 if (!uuid_equal(&block->bb_u.l.bb_uuid, &mp->m_sb.sb_uuid))
735 return false;
736 if (be64_to_cpu(block->bb_u.l.bb_blkno) != bp->b_bn)
737 return false;
738 /*
739 * XXX: need a better way of verifying the owner here. Right now
740 * just make sure there has been one set.
741 */
742 if (be64_to_cpu(block->bb_u.l.bb_owner) == 0)
743 return false;
744 /* fall through */
745 case cpu_to_be32(XFS_BMAP_MAGIC):
746 break;
747 default:
748 return false;
749 }
750
751 /*
752 * numrecs and level verification.
753 *
754 * We don't know what fork we belong to, so just verify that the level
755 * is less than the maximum of the two. Later checks will be more
756 * precise.
757 */
758 level = be16_to_cpu(block->bb_level);
759 if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
760 return false;
761 if (be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
762 return false;
763
764 /* sibling pointer verification */
765 if (!block->bb_u.l.bb_leftsib ||
766 (block->bb_u.l.bb_leftsib != cpu_to_be64(NULLDFSBNO) &&
767 !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_leftsib))))
768 return false;
769 if (!block->bb_u.l.bb_rightsib ||
770 (block->bb_u.l.bb_rightsib != cpu_to_be64(NULLDFSBNO) &&
771 !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_rightsib))))
772 return false;
773
774 return true;
775}
776
777static void
778xfs_bmbt_read_verify(
779 struct xfs_buf *bp)
780{
781 if (!xfs_btree_lblock_verify_crc(bp))
782 xfs_buf_ioerror(bp, -EFSBADCRC);
783 else if (!xfs_bmbt_verify(bp))
784 xfs_buf_ioerror(bp, -EFSCORRUPTED);
785
786 if (bp->b_error) {
787 trace_xfs_btree_corrupt(bp, _RET_IP_);
788 xfs_verifier_error(bp);
789 }
790}
791
792static void
793xfs_bmbt_write_verify(
794 struct xfs_buf *bp)
795{
796 if (!xfs_bmbt_verify(bp)) {
797 trace_xfs_btree_corrupt(bp, _RET_IP_);
798 xfs_buf_ioerror(bp, -EFSCORRUPTED);
799 xfs_verifier_error(bp);
800 return;
801 }
802 xfs_btree_lblock_calc_crc(bp);
803}
804
805const struct xfs_buf_ops xfs_bmbt_buf_ops = {
806 .verify_read = xfs_bmbt_read_verify,
807 .verify_write = xfs_bmbt_write_verify,
808};
809
810
811#if defined(DEBUG) || defined(XFS_WARN)
812STATIC int
813xfs_bmbt_keys_inorder(
814 struct xfs_btree_cur *cur,
815 union xfs_btree_key *k1,
816 union xfs_btree_key *k2)
817{
818 return be64_to_cpu(k1->bmbt.br_startoff) <
819 be64_to_cpu(k2->bmbt.br_startoff);
820}
821
822STATIC int
823xfs_bmbt_recs_inorder(
824 struct xfs_btree_cur *cur,
825 union xfs_btree_rec *r1,
826 union xfs_btree_rec *r2)
827{
828 return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
829 xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
830 xfs_bmbt_disk_get_startoff(&r2->bmbt);
831}
832#endif /* DEBUG */
833
834static const struct xfs_btree_ops xfs_bmbt_ops = {
835 .rec_len = sizeof(xfs_bmbt_rec_t),
836 .key_len = sizeof(xfs_bmbt_key_t),
837
838 .dup_cursor = xfs_bmbt_dup_cursor,
839 .update_cursor = xfs_bmbt_update_cursor,
840 .alloc_block = xfs_bmbt_alloc_block,
841 .free_block = xfs_bmbt_free_block,
842 .get_maxrecs = xfs_bmbt_get_maxrecs,
843 .get_minrecs = xfs_bmbt_get_minrecs,
844 .get_dmaxrecs = xfs_bmbt_get_dmaxrecs,
845 .init_key_from_rec = xfs_bmbt_init_key_from_rec,
846 .init_rec_from_key = xfs_bmbt_init_rec_from_key,
847 .init_rec_from_cur = xfs_bmbt_init_rec_from_cur,
848 .init_ptr_from_cur = xfs_bmbt_init_ptr_from_cur,
849 .key_diff = xfs_bmbt_key_diff,
850 .buf_ops = &xfs_bmbt_buf_ops,
851#if defined(DEBUG) || defined(XFS_WARN)
852 .keys_inorder = xfs_bmbt_keys_inorder,
853 .recs_inorder = xfs_bmbt_recs_inorder,
854#endif
855};
856
857/*
858 * Allocate a new bmap btree cursor.
859 */
860struct xfs_btree_cur * /* new bmap btree cursor */
861xfs_bmbt_init_cursor(
862 struct xfs_mount *mp, /* file system mount point */
863 struct xfs_trans *tp, /* transaction pointer */
864 struct xfs_inode *ip, /* inode owning the btree */
865 int whichfork) /* data or attr fork */
866{
867 struct xfs_ifork *ifp = XFS_IFORK_PTR(ip, whichfork);
868 struct xfs_btree_cur *cur;
869
870 cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
871
872 cur->bc_tp = tp;
873 cur->bc_mp = mp;
874 cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
875 cur->bc_btnum = XFS_BTNUM_BMAP;
876 cur->bc_blocklog = mp->m_sb.sb_blocklog;
877
878 cur->bc_ops = &xfs_bmbt_ops;
879 cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
880 if (xfs_sb_version_hascrc(&mp->m_sb))
881 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
882
883 cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
884 cur->bc_private.b.ip = ip;
885 cur->bc_private.b.firstblock = NULLFSBLOCK;
886 cur->bc_private.b.flist = NULL;
887 cur->bc_private.b.allocated = 0;
888 cur->bc_private.b.flags = 0;
889 cur->bc_private.b.whichfork = whichfork;
890
891 return cur;
892}
893
894/*
895 * Calculate number of records in a bmap btree block.
896 */
897int
898xfs_bmbt_maxrecs(
899 struct xfs_mount *mp,
900 int blocklen,
901 int leaf)
902{
903 blocklen -= XFS_BMBT_BLOCK_LEN(mp);
904
905 if (leaf)
906 return blocklen / sizeof(xfs_bmbt_rec_t);
907 return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
908}
909
910/*
911 * Calculate number of records in a bmap btree inode root.
912 */
913int
914xfs_bmdr_maxrecs(
915 int blocklen,
916 int leaf)
917{
918 blocklen -= sizeof(xfs_bmdr_block_t);
919
920 if (leaf)
921 return blocklen / sizeof(xfs_bmdr_rec_t);
922 return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
923}
924
925/*
926 * Change the owner of a btree format fork fo the inode passed in. Change it to
927 * the owner of that is passed in so that we can change owners before or after
928 * we switch forks between inodes. The operation that the caller is doing will
929 * determine whether is needs to change owner before or after the switch.
930 *
931 * For demand paged transactional modification, the fork switch should be done
932 * after reading in all the blocks, modifying them and pinning them in the
933 * transaction. For modification when the buffers are already pinned in memory,
934 * the fork switch can be done before changing the owner as we won't need to
935 * validate the owner until the btree buffers are unpinned and writes can occur
936 * again.
937 *
938 * For recovery based ownership change, there is no transactional context and
939 * so a buffer list must be supplied so that we can record the buffers that we
940 * modified for the caller to issue IO on.
941 */
942int
943xfs_bmbt_change_owner(
944 struct xfs_trans *tp,
945 struct xfs_inode *ip,
946 int whichfork,
947 xfs_ino_t new_owner,
948 struct list_head *buffer_list)
949{
950 struct xfs_btree_cur *cur;
951 int error;
952
953 ASSERT(tp || buffer_list);
954 ASSERT(!(tp && buffer_list));
955 if (whichfork == XFS_DATA_FORK)
956 ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
957 else
958 ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
959
960 cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
961 if (!cur)
962 return -ENOMEM;
963
964 error = xfs_btree_change_owner(cur, new_owner, buffer_list);
965 xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
966 return error;
967}