mm: fix exec activate_mm vs TLB shootdown and lazy tlb switching race
[linux/fpc-iii.git] / fs / xfs / libxfs / xfs_bmap_btree.c
bloba6331ffa51e3e69f04a789de7f335da99ab856dd
1 /*
2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
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.
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.
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
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"
41 * Determine the extent state.
43 /* ARGSUSED */
44 STATIC xfs_exntst_t
45 xfs_extent_state(
46 xfs_filblks_t blks,
47 int extent_flag)
49 if (extent_flag) {
50 ASSERT(blks != 0); /* saved for DMIG */
51 return XFS_EXT_UNWRITTEN;
53 return XFS_EXT_NORM;
57 * Convert on-disk form of btree root to in-memory form.
59 void
60 xfs_bmdr_to_bmbt(
61 struct xfs_inode *ip,
62 xfs_bmdr_block_t *dblock,
63 int dblocklen,
64 struct xfs_btree_block *rblock,
65 int rblocklen)
67 struct xfs_mount *mp = ip->i_mount;
68 int dmxr;
69 xfs_bmbt_key_t *fkp;
70 __be64 *fpp;
71 xfs_bmbt_key_t *tkp;
72 __be64 *tpp;
74 xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
75 XFS_BTNUM_BMAP, 0, 0, ip->i_ino,
76 XFS_BTREE_LONG_PTRS);
77 rblock->bb_level = dblock->bb_level;
78 ASSERT(be16_to_cpu(rblock->bb_level) > 0);
79 rblock->bb_numrecs = dblock->bb_numrecs;
80 dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
81 fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
82 tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
83 fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
84 tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
85 dmxr = be16_to_cpu(dblock->bb_numrecs);
86 memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
87 memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
91 * Convert a compressed bmap extent record to an uncompressed form.
92 * This code must be in sync with the routines xfs_bmbt_get_startoff,
93 * xfs_bmbt_get_startblock, xfs_bmbt_get_blockcount and xfs_bmbt_get_state.
95 STATIC void
96 __xfs_bmbt_get_all(
97 uint64_t l0,
98 uint64_t l1,
99 xfs_bmbt_irec_t *s)
101 int ext_flag;
102 xfs_exntst_t st;
104 ext_flag = (int)(l0 >> (64 - BMBT_EXNTFLAG_BITLEN));
105 s->br_startoff = ((xfs_fileoff_t)l0 &
106 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
107 s->br_startblock = (((xfs_fsblock_t)l0 & xfs_mask64lo(9)) << 43) |
108 (((xfs_fsblock_t)l1) >> 21);
109 s->br_blockcount = (xfs_filblks_t)(l1 & xfs_mask64lo(21));
110 /* This is xfs_extent_state() in-line */
111 if (ext_flag) {
112 ASSERT(s->br_blockcount != 0); /* saved for DMIG */
113 st = XFS_EXT_UNWRITTEN;
114 } else
115 st = XFS_EXT_NORM;
116 s->br_state = st;
119 void
120 xfs_bmbt_get_all(
121 xfs_bmbt_rec_host_t *r,
122 xfs_bmbt_irec_t *s)
124 __xfs_bmbt_get_all(r->l0, r->l1, s);
128 * Extract the blockcount field from an in memory bmap extent record.
130 xfs_filblks_t
131 xfs_bmbt_get_blockcount(
132 xfs_bmbt_rec_host_t *r)
134 return (xfs_filblks_t)(r->l1 & xfs_mask64lo(21));
138 * Extract the startblock field from an in memory bmap extent record.
140 xfs_fsblock_t
141 xfs_bmbt_get_startblock(
142 xfs_bmbt_rec_host_t *r)
144 return (((xfs_fsblock_t)r->l0 & xfs_mask64lo(9)) << 43) |
145 (((xfs_fsblock_t)r->l1) >> 21);
149 * Extract the startoff field from an in memory bmap extent record.
151 xfs_fileoff_t
152 xfs_bmbt_get_startoff(
153 xfs_bmbt_rec_host_t *r)
155 return ((xfs_fileoff_t)r->l0 &
156 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
159 xfs_exntst_t
160 xfs_bmbt_get_state(
161 xfs_bmbt_rec_host_t *r)
163 int ext_flag;
165 ext_flag = (int)((r->l0) >> (64 - BMBT_EXNTFLAG_BITLEN));
166 return xfs_extent_state(xfs_bmbt_get_blockcount(r),
167 ext_flag);
171 * Extract the blockcount field from an on disk bmap extent record.
173 xfs_filblks_t
174 xfs_bmbt_disk_get_blockcount(
175 xfs_bmbt_rec_t *r)
177 return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
181 * Extract the startoff field from a disk format bmap extent record.
183 xfs_fileoff_t
184 xfs_bmbt_disk_get_startoff(
185 xfs_bmbt_rec_t *r)
187 return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
188 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
193 * Set all the fields in a bmap extent record from the arguments.
195 void
196 xfs_bmbt_set_allf(
197 xfs_bmbt_rec_host_t *r,
198 xfs_fileoff_t startoff,
199 xfs_fsblock_t startblock,
200 xfs_filblks_t blockcount,
201 xfs_exntst_t state)
203 int extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
205 ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
206 ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
207 ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
209 ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
211 r->l0 = ((xfs_bmbt_rec_base_t)extent_flag << 63) |
212 ((xfs_bmbt_rec_base_t)startoff << 9) |
213 ((xfs_bmbt_rec_base_t)startblock >> 43);
214 r->l1 = ((xfs_bmbt_rec_base_t)startblock << 21) |
215 ((xfs_bmbt_rec_base_t)blockcount &
216 (xfs_bmbt_rec_base_t)xfs_mask64lo(21));
220 * Set all the fields in a bmap extent record from the uncompressed form.
222 void
223 xfs_bmbt_set_all(
224 xfs_bmbt_rec_host_t *r,
225 xfs_bmbt_irec_t *s)
227 xfs_bmbt_set_allf(r, s->br_startoff, s->br_startblock,
228 s->br_blockcount, s->br_state);
233 * Set all the fields in a disk format bmap extent record from the arguments.
235 void
236 xfs_bmbt_disk_set_allf(
237 xfs_bmbt_rec_t *r,
238 xfs_fileoff_t startoff,
239 xfs_fsblock_t startblock,
240 xfs_filblks_t blockcount,
241 xfs_exntst_t state)
243 int extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
245 ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
246 ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
247 ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
248 ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
250 r->l0 = cpu_to_be64(
251 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
252 ((xfs_bmbt_rec_base_t)startoff << 9) |
253 ((xfs_bmbt_rec_base_t)startblock >> 43));
254 r->l1 = cpu_to_be64(
255 ((xfs_bmbt_rec_base_t)startblock << 21) |
256 ((xfs_bmbt_rec_base_t)blockcount &
257 (xfs_bmbt_rec_base_t)xfs_mask64lo(21)));
261 * Set all the fields in a bmap extent record from the uncompressed form.
263 STATIC void
264 xfs_bmbt_disk_set_all(
265 xfs_bmbt_rec_t *r,
266 xfs_bmbt_irec_t *s)
268 xfs_bmbt_disk_set_allf(r, s->br_startoff, s->br_startblock,
269 s->br_blockcount, s->br_state);
273 * Set the blockcount field in a bmap extent record.
275 void
276 xfs_bmbt_set_blockcount(
277 xfs_bmbt_rec_host_t *r,
278 xfs_filblks_t v)
280 ASSERT((v & xfs_mask64hi(43)) == 0);
281 r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64hi(43)) |
282 (xfs_bmbt_rec_base_t)(v & xfs_mask64lo(21));
286 * Set the startblock field in a bmap extent record.
288 void
289 xfs_bmbt_set_startblock(
290 xfs_bmbt_rec_host_t *r,
291 xfs_fsblock_t v)
293 ASSERT((v & xfs_mask64hi(12)) == 0);
294 r->l0 = (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64hi(55)) |
295 (xfs_bmbt_rec_base_t)(v >> 43);
296 r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64lo(21)) |
297 (xfs_bmbt_rec_base_t)(v << 21);
301 * Set the startoff field in a bmap extent record.
303 void
304 xfs_bmbt_set_startoff(
305 xfs_bmbt_rec_host_t *r,
306 xfs_fileoff_t v)
308 ASSERT((v & xfs_mask64hi(9)) == 0);
309 r->l0 = (r->l0 & (xfs_bmbt_rec_base_t) xfs_mask64hi(1)) |
310 ((xfs_bmbt_rec_base_t)v << 9) |
311 (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64lo(9));
315 * Set the extent state field in a bmap extent record.
317 void
318 xfs_bmbt_set_state(
319 xfs_bmbt_rec_host_t *r,
320 xfs_exntst_t v)
322 ASSERT(v == XFS_EXT_NORM || v == XFS_EXT_UNWRITTEN);
323 if (v == XFS_EXT_NORM)
324 r->l0 &= xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN);
325 else
326 r->l0 |= xfs_mask64hi(BMBT_EXNTFLAG_BITLEN);
330 * Convert in-memory form of btree root to on-disk form.
332 void
333 xfs_bmbt_to_bmdr(
334 struct xfs_mount *mp,
335 struct xfs_btree_block *rblock,
336 int rblocklen,
337 xfs_bmdr_block_t *dblock,
338 int dblocklen)
340 int dmxr;
341 xfs_bmbt_key_t *fkp;
342 __be64 *fpp;
343 xfs_bmbt_key_t *tkp;
344 __be64 *tpp;
346 if (xfs_sb_version_hascrc(&mp->m_sb)) {
347 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
348 ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid,
349 &mp->m_sb.sb_meta_uuid));
350 ASSERT(rblock->bb_u.l.bb_blkno ==
351 cpu_to_be64(XFS_BUF_DADDR_NULL));
352 } else
353 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
354 ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
355 ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
356 ASSERT(rblock->bb_level != 0);
357 dblock->bb_level = rblock->bb_level;
358 dblock->bb_numrecs = rblock->bb_numrecs;
359 dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
360 fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
361 tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
362 fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
363 tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
364 dmxr = be16_to_cpu(dblock->bb_numrecs);
365 memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
366 memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
369 STATIC struct xfs_btree_cur *
370 xfs_bmbt_dup_cursor(
371 struct xfs_btree_cur *cur)
373 struct xfs_btree_cur *new;
375 new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
376 cur->bc_private.b.ip, cur->bc_private.b.whichfork);
379 * Copy the firstblock, dfops, and flags values,
380 * since init cursor doesn't get them.
382 new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
383 new->bc_private.b.dfops = cur->bc_private.b.dfops;
384 new->bc_private.b.flags = cur->bc_private.b.flags;
386 return new;
389 STATIC void
390 xfs_bmbt_update_cursor(
391 struct xfs_btree_cur *src,
392 struct xfs_btree_cur *dst)
394 ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
395 (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
396 ASSERT(dst->bc_private.b.dfops == src->bc_private.b.dfops);
398 dst->bc_private.b.allocated += src->bc_private.b.allocated;
399 dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
401 src->bc_private.b.allocated = 0;
404 STATIC int
405 xfs_bmbt_alloc_block(
406 struct xfs_btree_cur *cur,
407 union xfs_btree_ptr *start,
408 union xfs_btree_ptr *new,
409 int *stat)
411 xfs_alloc_arg_t args; /* block allocation args */
412 int error; /* error return value */
414 memset(&args, 0, sizeof(args));
415 args.tp = cur->bc_tp;
416 args.mp = cur->bc_mp;
417 args.fsbno = cur->bc_private.b.firstblock;
418 args.firstblock = args.fsbno;
419 xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_private.b.ip->i_ino,
420 cur->bc_private.b.whichfork);
422 if (args.fsbno == NULLFSBLOCK) {
423 args.fsbno = be64_to_cpu(start->l);
424 args.type = XFS_ALLOCTYPE_START_BNO;
426 * Make sure there is sufficient room left in the AG to
427 * complete a full tree split for an extent insert. If
428 * we are converting the middle part of an extent then
429 * we may need space for two tree splits.
431 * We are relying on the caller to make the correct block
432 * reservation for this operation to succeed. If the
433 * reservation amount is insufficient then we may fail a
434 * block allocation here and corrupt the filesystem.
436 args.minleft = args.tp->t_blk_res;
437 } else if (cur->bc_private.b.dfops->dop_low) {
438 args.type = XFS_ALLOCTYPE_START_BNO;
439 } else {
440 args.type = XFS_ALLOCTYPE_NEAR_BNO;
443 args.minlen = args.maxlen = args.prod = 1;
444 args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
445 if (!args.wasdel && args.tp->t_blk_res == 0) {
446 error = -ENOSPC;
447 goto error0;
449 error = xfs_alloc_vextent(&args);
450 if (error)
451 goto error0;
453 if (args.fsbno == NULLFSBLOCK && args.minleft) {
455 * Could not find an AG with enough free space to satisfy
456 * a full btree split. Try again and if
457 * successful activate the lowspace algorithm.
459 args.fsbno = 0;
460 args.type = XFS_ALLOCTYPE_FIRST_AG;
461 error = xfs_alloc_vextent(&args);
462 if (error)
463 goto error0;
464 cur->bc_private.b.dfops->dop_low = true;
466 if (WARN_ON_ONCE(args.fsbno == NULLFSBLOCK)) {
467 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
468 *stat = 0;
469 return 0;
471 ASSERT(args.len == 1);
472 cur->bc_private.b.firstblock = args.fsbno;
473 cur->bc_private.b.allocated++;
474 cur->bc_private.b.ip->i_d.di_nblocks++;
475 xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
476 xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
477 XFS_TRANS_DQ_BCOUNT, 1L);
479 new->l = cpu_to_be64(args.fsbno);
481 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
482 *stat = 1;
483 return 0;
485 error0:
486 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
487 return error;
490 STATIC int
491 xfs_bmbt_free_block(
492 struct xfs_btree_cur *cur,
493 struct xfs_buf *bp)
495 struct xfs_mount *mp = cur->bc_mp;
496 struct xfs_inode *ip = cur->bc_private.b.ip;
497 struct xfs_trans *tp = cur->bc_tp;
498 xfs_fsblock_t fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
499 struct xfs_owner_info oinfo;
501 xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, cur->bc_private.b.whichfork);
502 xfs_bmap_add_free(mp, cur->bc_private.b.dfops, fsbno, 1, &oinfo);
503 ip->i_d.di_nblocks--;
505 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
506 xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
507 return 0;
510 STATIC int
511 xfs_bmbt_get_minrecs(
512 struct xfs_btree_cur *cur,
513 int level)
515 if (level == cur->bc_nlevels - 1) {
516 struct xfs_ifork *ifp;
518 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
519 cur->bc_private.b.whichfork);
521 return xfs_bmbt_maxrecs(cur->bc_mp,
522 ifp->if_broot_bytes, level == 0) / 2;
525 return cur->bc_mp->m_bmap_dmnr[level != 0];
529 xfs_bmbt_get_maxrecs(
530 struct xfs_btree_cur *cur,
531 int level)
533 if (level == cur->bc_nlevels - 1) {
534 struct xfs_ifork *ifp;
536 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
537 cur->bc_private.b.whichfork);
539 return xfs_bmbt_maxrecs(cur->bc_mp,
540 ifp->if_broot_bytes, level == 0);
543 return cur->bc_mp->m_bmap_dmxr[level != 0];
548 * Get the maximum records we could store in the on-disk format.
550 * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
551 * for the root node this checks the available space in the dinode fork
552 * so that we can resize the in-memory buffer to match it. After a
553 * resize to the maximum size this function returns the same value
554 * as xfs_bmbt_get_maxrecs for the root node, too.
556 STATIC int
557 xfs_bmbt_get_dmaxrecs(
558 struct xfs_btree_cur *cur,
559 int level)
561 if (level != cur->bc_nlevels - 1)
562 return cur->bc_mp->m_bmap_dmxr[level != 0];
563 return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
566 STATIC void
567 xfs_bmbt_init_key_from_rec(
568 union xfs_btree_key *key,
569 union xfs_btree_rec *rec)
571 key->bmbt.br_startoff =
572 cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
575 STATIC void
576 xfs_bmbt_init_high_key_from_rec(
577 union xfs_btree_key *key,
578 union xfs_btree_rec *rec)
580 key->bmbt.br_startoff = cpu_to_be64(
581 xfs_bmbt_disk_get_startoff(&rec->bmbt) +
582 xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
585 STATIC void
586 xfs_bmbt_init_rec_from_cur(
587 struct xfs_btree_cur *cur,
588 union xfs_btree_rec *rec)
590 xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
593 STATIC void
594 xfs_bmbt_init_ptr_from_cur(
595 struct xfs_btree_cur *cur,
596 union xfs_btree_ptr *ptr)
598 ptr->l = 0;
601 STATIC int64_t
602 xfs_bmbt_key_diff(
603 struct xfs_btree_cur *cur,
604 union xfs_btree_key *key)
606 return (int64_t)be64_to_cpu(key->bmbt.br_startoff) -
607 cur->bc_rec.b.br_startoff;
610 STATIC int64_t
611 xfs_bmbt_diff_two_keys(
612 struct xfs_btree_cur *cur,
613 union xfs_btree_key *k1,
614 union xfs_btree_key *k2)
616 return (int64_t)be64_to_cpu(k1->bmbt.br_startoff) -
617 be64_to_cpu(k2->bmbt.br_startoff);
620 static bool
621 xfs_bmbt_verify(
622 struct xfs_buf *bp)
624 struct xfs_mount *mp = bp->b_target->bt_mount;
625 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
626 unsigned int level;
628 switch (block->bb_magic) {
629 case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
630 if (!xfs_sb_version_hascrc(&mp->m_sb))
631 return false;
632 if (!uuid_equal(&block->bb_u.l.bb_uuid, &mp->m_sb.sb_meta_uuid))
633 return false;
634 if (be64_to_cpu(block->bb_u.l.bb_blkno) != bp->b_bn)
635 return false;
637 * XXX: need a better way of verifying the owner here. Right now
638 * just make sure there has been one set.
640 if (be64_to_cpu(block->bb_u.l.bb_owner) == 0)
641 return false;
642 /* fall through */
643 case cpu_to_be32(XFS_BMAP_MAGIC):
644 break;
645 default:
646 return false;
650 * numrecs and level verification.
652 * We don't know what fork we belong to, so just verify that the level
653 * is less than the maximum of the two. Later checks will be more
654 * precise.
656 level = be16_to_cpu(block->bb_level);
657 if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
658 return false;
659 if (be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
660 return false;
662 /* sibling pointer verification */
663 if (!block->bb_u.l.bb_leftsib ||
664 (block->bb_u.l.bb_leftsib != cpu_to_be64(NULLFSBLOCK) &&
665 !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_leftsib))))
666 return false;
667 if (!block->bb_u.l.bb_rightsib ||
668 (block->bb_u.l.bb_rightsib != cpu_to_be64(NULLFSBLOCK) &&
669 !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_rightsib))))
670 return false;
672 return true;
675 static void
676 xfs_bmbt_read_verify(
677 struct xfs_buf *bp)
679 if (!xfs_btree_lblock_verify_crc(bp))
680 xfs_buf_ioerror(bp, -EFSBADCRC);
681 else if (!xfs_bmbt_verify(bp))
682 xfs_buf_ioerror(bp, -EFSCORRUPTED);
684 if (bp->b_error) {
685 trace_xfs_btree_corrupt(bp, _RET_IP_);
686 xfs_verifier_error(bp);
690 static void
691 xfs_bmbt_write_verify(
692 struct xfs_buf *bp)
694 if (!xfs_bmbt_verify(bp)) {
695 trace_xfs_btree_corrupt(bp, _RET_IP_);
696 xfs_buf_ioerror(bp, -EFSCORRUPTED);
697 xfs_verifier_error(bp);
698 return;
700 xfs_btree_lblock_calc_crc(bp);
703 const struct xfs_buf_ops xfs_bmbt_buf_ops = {
704 .name = "xfs_bmbt",
705 .verify_read = xfs_bmbt_read_verify,
706 .verify_write = xfs_bmbt_write_verify,
710 STATIC int
711 xfs_bmbt_keys_inorder(
712 struct xfs_btree_cur *cur,
713 union xfs_btree_key *k1,
714 union xfs_btree_key *k2)
716 return be64_to_cpu(k1->bmbt.br_startoff) <
717 be64_to_cpu(k2->bmbt.br_startoff);
720 STATIC int
721 xfs_bmbt_recs_inorder(
722 struct xfs_btree_cur *cur,
723 union xfs_btree_rec *r1,
724 union xfs_btree_rec *r2)
726 return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
727 xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
728 xfs_bmbt_disk_get_startoff(&r2->bmbt);
731 static const struct xfs_btree_ops xfs_bmbt_ops = {
732 .rec_len = sizeof(xfs_bmbt_rec_t),
733 .key_len = sizeof(xfs_bmbt_key_t),
735 .dup_cursor = xfs_bmbt_dup_cursor,
736 .update_cursor = xfs_bmbt_update_cursor,
737 .alloc_block = xfs_bmbt_alloc_block,
738 .free_block = xfs_bmbt_free_block,
739 .get_maxrecs = xfs_bmbt_get_maxrecs,
740 .get_minrecs = xfs_bmbt_get_minrecs,
741 .get_dmaxrecs = xfs_bmbt_get_dmaxrecs,
742 .init_key_from_rec = xfs_bmbt_init_key_from_rec,
743 .init_high_key_from_rec = xfs_bmbt_init_high_key_from_rec,
744 .init_rec_from_cur = xfs_bmbt_init_rec_from_cur,
745 .init_ptr_from_cur = xfs_bmbt_init_ptr_from_cur,
746 .key_diff = xfs_bmbt_key_diff,
747 .diff_two_keys = xfs_bmbt_diff_two_keys,
748 .buf_ops = &xfs_bmbt_buf_ops,
749 .keys_inorder = xfs_bmbt_keys_inorder,
750 .recs_inorder = xfs_bmbt_recs_inorder,
754 * Allocate a new bmap btree cursor.
756 struct xfs_btree_cur * /* new bmap btree cursor */
757 xfs_bmbt_init_cursor(
758 struct xfs_mount *mp, /* file system mount point */
759 struct xfs_trans *tp, /* transaction pointer */
760 struct xfs_inode *ip, /* inode owning the btree */
761 int whichfork) /* data or attr fork */
763 struct xfs_ifork *ifp = XFS_IFORK_PTR(ip, whichfork);
764 struct xfs_btree_cur *cur;
765 ASSERT(whichfork != XFS_COW_FORK);
767 cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
769 cur->bc_tp = tp;
770 cur->bc_mp = mp;
771 cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
772 cur->bc_btnum = XFS_BTNUM_BMAP;
773 cur->bc_blocklog = mp->m_sb.sb_blocklog;
774 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_bmbt_2);
776 cur->bc_ops = &xfs_bmbt_ops;
777 cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
778 if (xfs_sb_version_hascrc(&mp->m_sb))
779 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
781 cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
782 cur->bc_private.b.ip = ip;
783 cur->bc_private.b.firstblock = NULLFSBLOCK;
784 cur->bc_private.b.dfops = NULL;
785 cur->bc_private.b.allocated = 0;
786 cur->bc_private.b.flags = 0;
787 cur->bc_private.b.whichfork = whichfork;
789 return cur;
793 * Calculate number of records in a bmap btree block.
796 xfs_bmbt_maxrecs(
797 struct xfs_mount *mp,
798 int blocklen,
799 int leaf)
801 blocklen -= XFS_BMBT_BLOCK_LEN(mp);
803 if (leaf)
804 return blocklen / sizeof(xfs_bmbt_rec_t);
805 return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
809 * Calculate number of records in a bmap btree inode root.
812 xfs_bmdr_maxrecs(
813 int blocklen,
814 int leaf)
816 blocklen -= sizeof(xfs_bmdr_block_t);
818 if (leaf)
819 return blocklen / sizeof(xfs_bmdr_rec_t);
820 return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
824 * Change the owner of a btree format fork fo the inode passed in. Change it to
825 * the owner of that is passed in so that we can change owners before or after
826 * we switch forks between inodes. The operation that the caller is doing will
827 * determine whether is needs to change owner before or after the switch.
829 * For demand paged transactional modification, the fork switch should be done
830 * after reading in all the blocks, modifying them and pinning them in the
831 * transaction. For modification when the buffers are already pinned in memory,
832 * the fork switch can be done before changing the owner as we won't need to
833 * validate the owner until the btree buffers are unpinned and writes can occur
834 * again.
836 * For recovery based ownership change, there is no transactional context and
837 * so a buffer list must be supplied so that we can record the buffers that we
838 * modified for the caller to issue IO on.
841 xfs_bmbt_change_owner(
842 struct xfs_trans *tp,
843 struct xfs_inode *ip,
844 int whichfork,
845 xfs_ino_t new_owner,
846 struct list_head *buffer_list)
848 struct xfs_btree_cur *cur;
849 int error;
851 ASSERT(tp || buffer_list);
852 ASSERT(!(tp && buffer_list));
853 if (whichfork == XFS_DATA_FORK)
854 ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
855 else
856 ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
858 cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
859 if (!cur)
860 return -ENOMEM;
861 cur->bc_private.b.flags |= XFS_BTCUR_BPRV_INVALID_OWNER;
863 error = xfs_btree_change_owner(cur, new_owner, buffer_list);
864 xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
865 return error;