]>
Commit | Line | Data |
---|---|---|
2bd0ea18 | 1 | /* |
da23017d NS |
2 | * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc. |
3 | * All Rights Reserved. | |
dfc130f3 | 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. |
dfc130f3 | 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. | |
dfc130f3 | 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 NS |
17 | */ |
18 | ||
6b803e5a | 19 | #include "libxfs.h" |
2bd0ea18 NS |
20 | #include "avl.h" |
21 | #include "globals.h" | |
22 | #include "agheader.h" | |
23 | #include "incore.h" | |
24 | #include "protos.h" | |
25 | #include "err_protos.h" | |
26 | #include "dinode.h" | |
27 | #include "scan.h" | |
28 | #include "versions.h" | |
29 | #include "bmap.h" | |
06fbdda9 | 30 | #include "progress.h" |
364a126c | 31 | #include "threads.h" |
11b9e510 | 32 | #include "slab.h" |
00efc33a | 33 | #include "rmap.h" |
2bd0ea18 | 34 | |
2bd0ea18 | 35 | static xfs_mount_t *mp = NULL; |
2bd0ea18 | 36 | |
48d49157 | 37 | /* |
364a126c | 38 | * Variables to validate AG header values against the manual count |
48d49157 | 39 | * from the btree traversal. |
48d49157 | 40 | */ |
364a126c DC |
41 | struct aghdr_cnts { |
42 | xfs_agnumber_t agno; | |
43 | xfs_extlen_t agffreeblks; | |
44 | xfs_extlen_t agflongest; | |
45 | __uint64_t agfbtreeblks; | |
46 | __uint32_t agicount; | |
47 | __uint32_t agifreecount; | |
48 | __uint64_t fdblocks; | |
0f94fa4b | 49 | __uint64_t usedblocks; |
364a126c | 50 | __uint64_t ifreecount; |
1578050f | 51 | __uint32_t fibtfreecount; |
364a126c | 52 | }; |
48d49157 | 53 | |
2bd0ea18 NS |
54 | void |
55 | set_mp(xfs_mount_t *mpp) | |
56 | { | |
f1b058f9 | 57 | libxfs_bcache_purge(); |
2bd0ea18 NS |
58 | mp = mpp; |
59 | } | |
60 | ||
8b8a6b02 | 61 | static void |
2bd0ea18 NS |
62 | scan_sbtree( |
63 | xfs_agblock_t root, | |
64 | int nlevels, | |
65 | xfs_agnumber_t agno, | |
66 | int suspect, | |
b3563c19 | 67 | void (*func)(struct xfs_btree_block *block, |
2bd0ea18 NS |
68 | int level, |
69 | xfs_agblock_t bno, | |
70 | xfs_agnumber_t agno, | |
71 | int suspect, | |
364a126c | 72 | int isroot, |
e0607266 | 73 | __uint32_t magic, |
364a126c DC |
74 | void *priv), |
75 | int isroot, | |
e0607266 DC |
76 | __uint32_t magic, |
77 | void *priv, | |
78 | const struct xfs_buf_ops *ops) | |
2bd0ea18 NS |
79 | { |
80 | xfs_buf_t *bp; | |
81 | ||
82 | bp = libxfs_readbuf(mp->m_dev, XFS_AGB_TO_DADDR(mp, agno, root), | |
e0607266 | 83 | XFS_FSB_TO_BB(mp, 1), 0, ops); |
2bd0ea18 | 84 | if (!bp) { |
507f4e33 | 85 | do_error(_("can't read btree block %d/%d\n"), agno, root); |
2bd0ea18 NS |
86 | return; |
87 | } | |
66fc04e0 | 88 | if (bp->b_error == -EFSBADCRC || bp->b_error == -EFSCORRUPTED) { |
a76a5a71 DC |
89 | do_warn(_("btree block %d/%d is suspect, error %d\n"), |
90 | agno, root, bp->b_error); | |
91 | suspect = 1; | |
92 | } | |
93 | ||
364a126c | 94 | (*func)(XFS_BUF_TO_BLOCK(bp), nlevels - 1, root, agno, suspect, |
e0607266 | 95 | isroot, magic, priv); |
2bd0ea18 NS |
96 | libxfs_putbuf(bp); |
97 | } | |
98 | ||
99 | /* | |
100 | * returns 1 on bad news (inode needs to be cleared), 0 on good | |
101 | */ | |
102 | int | |
103 | scan_lbtree( | |
5a35bf2c | 104 | xfs_fsblock_t root, |
2bd0ea18 | 105 | int nlevels, |
b3563c19 | 106 | int (*func)(struct xfs_btree_block *block, |
2bd0ea18 NS |
107 | int level, |
108 | int type, | |
109 | int whichfork, | |
5a35bf2c | 110 | xfs_fsblock_t bno, |
2bd0ea18 | 111 | xfs_ino_t ino, |
5a35bf2c | 112 | xfs_rfsblock_t *tot, |
2bd0ea18 NS |
113 | __uint64_t *nex, |
114 | blkmap_t **blkmapp, | |
115 | bmap_cursor_t *bm_cursor, | |
116 | int isroot, | |
117 | int check_dups, | |
e0607266 DC |
118 | int *dirty, |
119 | __uint64_t magic), | |
2bd0ea18 NS |
120 | int type, |
121 | int whichfork, | |
122 | xfs_ino_t ino, | |
5a35bf2c | 123 | xfs_rfsblock_t *tot, |
2bd0ea18 NS |
124 | __uint64_t *nex, |
125 | blkmap_t **blkmapp, | |
126 | bmap_cursor_t *bm_cursor, | |
127 | int isroot, | |
e0607266 DC |
128 | int check_dups, |
129 | __uint64_t magic, | |
130 | const struct xfs_buf_ops *ops) | |
2bd0ea18 NS |
131 | { |
132 | xfs_buf_t *bp; | |
133 | int err; | |
134 | int dirty = 0; | |
a76a5a71 | 135 | bool badcrc = false; |
2bd0ea18 NS |
136 | |
137 | bp = libxfs_readbuf(mp->m_dev, XFS_FSB_TO_DADDR(mp, root), | |
e0607266 | 138 | XFS_FSB_TO_BB(mp, 1), 0, ops); |
2bd0ea18 | 139 | if (!bp) { |
507f4e33 | 140 | do_error(_("can't read btree block %d/%d\n"), |
2bd0ea18 NS |
141 | XFS_FSB_TO_AGNO(mp, root), |
142 | XFS_FSB_TO_AGBNO(mp, root)); | |
143 | return(1); | |
144 | } | |
a76a5a71 DC |
145 | |
146 | /* | |
147 | * only check for bad CRC here - caller will determine if there | |
148 | * is a corruption or not and whether it got corrected and so needs | |
149 | * writing back. CRC errors always imply we need to write the block. | |
150 | */ | |
66fc04e0 | 151 | if (bp->b_error == -EFSBADCRC) { |
a76a5a71 DC |
152 | do_warn(_("btree block %d/%d is suspect, error %d\n"), |
153 | XFS_FSB_TO_AGNO(mp, root), | |
154 | XFS_FSB_TO_AGBNO(mp, root), bp->b_error); | |
155 | badcrc = true; | |
156 | } | |
157 | ||
b3563c19 | 158 | err = (*func)(XFS_BUF_TO_BLOCK(bp), nlevels - 1, |
2bd0ea18 | 159 | type, whichfork, root, ino, tot, nex, blkmapp, |
e0607266 DC |
160 | bm_cursor, isroot, check_dups, &dirty, |
161 | magic); | |
2bd0ea18 | 162 | |
27527004 | 163 | ASSERT(dirty == 0 || (dirty && !no_modify)); |
2bd0ea18 | 164 | |
a76a5a71 | 165 | if ((dirty || badcrc) && !no_modify) |
2bd0ea18 NS |
166 | libxfs_writebuf(bp, 0); |
167 | else | |
168 | libxfs_putbuf(bp); | |
169 | ||
170 | return(err); | |
171 | } | |
172 | ||
173 | int | |
e0607266 | 174 | scan_bmapbt( |
b3563c19 | 175 | struct xfs_btree_block *block, |
2bd0ea18 NS |
176 | int level, |
177 | int type, | |
178 | int whichfork, | |
5a35bf2c | 179 | xfs_fsblock_t bno, |
2bd0ea18 | 180 | xfs_ino_t ino, |
5a35bf2c | 181 | xfs_rfsblock_t *tot, |
2bd0ea18 NS |
182 | __uint64_t *nex, |
183 | blkmap_t **blkmapp, | |
184 | bmap_cursor_t *bm_cursor, | |
185 | int isroot, | |
186 | int check_dups, | |
e0607266 DC |
187 | int *dirty, |
188 | __uint64_t magic) | |
2bd0ea18 | 189 | { |
2bd0ea18 NS |
190 | int i; |
191 | int err; | |
192 | xfs_bmbt_ptr_t *pp; | |
193 | xfs_bmbt_key_t *pkey; | |
5e656dbb | 194 | xfs_bmbt_rec_t *rp; |
5a35bf2c DC |
195 | xfs_fileoff_t first_key; |
196 | xfs_fileoff_t last_key; | |
beed0dc8 | 197 | char *forkname = get_forkname(whichfork); |
5e656dbb | 198 | int numrecs; |
95650c4d BN |
199 | xfs_agnumber_t agno; |
200 | xfs_agblock_t agbno; | |
201 | int state; | |
00efc33a | 202 | int error; |
2bd0ea18 | 203 | |
2bd0ea18 | 204 | /* |
dfc130f3 | 205 | * unlike the ag freeblock btrees, if anything looks wrong |
2bd0ea18 NS |
206 | * in an inode bmap tree, just bail. it's possible that |
207 | * we'll miss a case where the to-be-toasted inode and | |
208 | * another inode are claiming the same block but that's | |
209 | * highly unlikely. | |
210 | */ | |
e0607266 | 211 | if (be32_to_cpu(block->bb_magic) != magic) { |
5d1b7f0f CH |
212 | do_warn( |
213 | _("bad magic # %#x in inode %" PRIu64 " (%s fork) bmbt block %" PRIu64 "\n"), | |
214 | be32_to_cpu(block->bb_magic), ino, forkname, bno); | |
2bd0ea18 NS |
215 | return(1); |
216 | } | |
5e656dbb | 217 | if (be16_to_cpu(block->bb_level) != level) { |
5d1b7f0f CH |
218 | do_warn( |
219 | _("expected level %d got %d in inode %" PRIu64 ", (%s fork) bmbt block %" PRIu64 "\n"), | |
220 | level, be16_to_cpu(block->bb_level), | |
221 | ino, forkname, bno); | |
2bd0ea18 NS |
222 | return(1); |
223 | } | |
224 | ||
e0607266 DC |
225 | if (magic == XFS_BMAP_CRC_MAGIC) { |
226 | /* verify owner */ | |
227 | if (be64_to_cpu(block->bb_u.l.bb_owner) != ino) { | |
228 | do_warn( | |
229 | _("expected owner inode %" PRIu64 ", got %llu, bmbt block %" PRIu64 "\n"), | |
230 | ino, be64_to_cpu(block->bb_u.l.bb_owner), bno); | |
0519f662 DW |
231 | return 1; |
232 | } | |
233 | /* verify block number */ | |
234 | if (be64_to_cpu(block->bb_u.l.bb_blkno) != | |
235 | XFS_FSB_TO_DADDR(mp, bno)) { | |
236 | do_warn( | |
237 | _("expected block %" PRIu64 ", got %llu, bmbt block %" PRIu64 "\n"), | |
238 | XFS_FSB_TO_DADDR(mp, bno), | |
239 | be64_to_cpu(block->bb_u.l.bb_blkno), bno); | |
240 | return 1; | |
241 | } | |
242 | /* verify uuid */ | |
243 | if (platform_uuid_compare(&block->bb_u.l.bb_uuid, | |
f31982ae | 244 | &mp->m_sb.sb_meta_uuid) != 0) { |
0519f662 DW |
245 | do_warn( |
246 | _("wrong FS UUID, bmbt block %" PRIu64 "\n"), | |
247 | bno); | |
248 | return 1; | |
e0607266 DC |
249 | } |
250 | } | |
251 | ||
2bd0ea18 NS |
252 | if (check_dups == 0) { |
253 | /* | |
254 | * check sibling pointers. if bad we have a conflict | |
255 | * between the sibling pointers and the child pointers | |
256 | * in the parent block. blow out the inode if that happens | |
257 | */ | |
5a35bf2c | 258 | if (bm_cursor->level[level].fsbno != NULLFSBLOCK) { |
2bd0ea18 NS |
259 | /* |
260 | * this is not the first block on this level | |
261 | * so the cursor for this level has recorded the | |
262 | * values for this's block left-sibling. | |
263 | */ | |
264 | if (bno != bm_cursor->level[level].right_fsbno) { | |
265 | do_warn( | |
5d1b7f0f CH |
266 | _("bad fwd (right) sibling pointer (saw %" PRIu64 " parent block says %" PRIu64 ")\n" |
267 | "\tin inode %" PRIu64 " (%s fork) bmap btree block %" PRIu64 "\n"), | |
2bd0ea18 | 268 | bm_cursor->level[level].right_fsbno, |
507f4e33 | 269 | bno, ino, forkname, |
2bd0ea18 NS |
270 | bm_cursor->level[level].fsbno); |
271 | return(1); | |
272 | } | |
b3563c19 | 273 | if (be64_to_cpu(block->bb_u.l.bb_leftsib) != |
2bd0ea18 NS |
274 | bm_cursor->level[level].fsbno) { |
275 | do_warn( | |
5d1b7f0f CH |
276 | _("bad back (left) sibling pointer (saw %llu parent block says %" PRIu64 ")\n" |
277 | "\tin inode %" PRIu64 " (%s fork) bmap btree block %" PRIu64 "\n"), | |
278 | (unsigned long long) | |
279 | be64_to_cpu(block->bb_u.l.bb_leftsib), | |
507f4e33 | 280 | bm_cursor->level[level].fsbno, |
2bd0ea18 NS |
281 | ino, forkname, bno); |
282 | return(1); | |
283 | } | |
284 | } else { | |
285 | /* | |
286 | * This is the first or only block on this level. | |
287 | * Check that the left sibling pointer is NULL | |
288 | */ | |
5a35bf2c | 289 | if (be64_to_cpu(block->bb_u.l.bb_leftsib) != NULLFSBLOCK) { |
2bd0ea18 | 290 | do_warn( |
507f4e33 | 291 | _("bad back (left) sibling pointer (saw %llu should be NULL (0))\n" |
5d1b7f0f CH |
292 | "\tin inode %" PRIu64 " (%s fork) bmap btree block %" PRIu64 "\n"), |
293 | (unsigned long long) | |
294 | be64_to_cpu(block->bb_u.l.bb_leftsib), | |
2bd0ea18 NS |
295 | ino, forkname, bno); |
296 | return(1); | |
297 | } | |
298 | } | |
299 | ||
300 | /* | |
301 | * update cursor block pointers to reflect this block | |
302 | */ | |
303 | bm_cursor->level[level].fsbno = bno; | |
507f4e33 | 304 | bm_cursor->level[level].left_fsbno = |
b3563c19 | 305 | be64_to_cpu(block->bb_u.l.bb_leftsib); |
507f4e33 | 306 | bm_cursor->level[level].right_fsbno = |
b3563c19 | 307 | be64_to_cpu(block->bb_u.l.bb_rightsib); |
2bd0ea18 | 308 | |
95650c4d BN |
309 | agno = XFS_FSB_TO_AGNO(mp, bno); |
310 | agbno = XFS_FSB_TO_AGBNO(mp, bno); | |
311 | ||
586f8abf | 312 | pthread_mutex_lock(&ag_locks[agno].lock); |
95650c4d BN |
313 | state = get_bmap(agno, agbno); |
314 | switch (state) { | |
0f94fa4b DW |
315 | case XR_E_INUSE1: |
316 | /* | |
317 | * block was claimed as in use data by the rmap | |
318 | * btree, but has not been found in the data extent | |
319 | * map for the inode. That means this bmbt block hasn't | |
320 | * yet been claimed as in use, which means -it's ours- | |
321 | */ | |
2bd0ea18 NS |
322 | case XR_E_UNKNOWN: |
323 | case XR_E_FREE1: | |
324 | case XR_E_FREE: | |
95650c4d | 325 | set_bmap(agno, agbno, XR_E_INUSE); |
2bd0ea18 NS |
326 | break; |
327 | case XR_E_FS_MAP: | |
328 | case XR_E_INUSE: | |
329 | /* | |
330 | * we'll try and continue searching here since | |
331 | * the block looks like it's been claimed by file | |
332 | * to store user data, a directory to store directory | |
333 | * data, or the space allocation btrees but since | |
334 | * we made it here, the block probably | |
335 | * contains btree data. | |
336 | */ | |
95650c4d | 337 | set_bmap(agno, agbno, XR_E_MULT); |
2bd0ea18 | 338 | do_warn( |
d9db5c4e | 339 | _("inode 0x%" PRIx64 "bmap block 0x%" PRIx64 " claimed, state is %d\n"), |
5d1b7f0f | 340 | ino, bno, state); |
2bd0ea18 NS |
341 | break; |
342 | case XR_E_MULT: | |
343 | case XR_E_INUSE_FS: | |
95650c4d | 344 | set_bmap(agno, agbno, XR_E_MULT); |
2bd0ea18 | 345 | do_warn( |
d9db5c4e | 346 | _("inode 0x%" PRIx64 " bmap block 0x%" PRIx64 " claimed, state is %d\n"), |
5d1b7f0f | 347 | ino, bno, state); |
2bd0ea18 NS |
348 | /* |
349 | * if we made it to here, this is probably a bmap block | |
350 | * that is being used by *another* file as a bmap block | |
351 | * so the block will be valid. Both files should be | |
352 | * trashed along with any other file that impinges on | |
353 | * any blocks referenced by either file. So we | |
354 | * continue searching down this btree to mark all | |
355 | * blocks duplicate | |
356 | */ | |
357 | break; | |
358 | case XR_E_BAD_STATE: | |
359 | default: | |
360 | do_warn( | |
2e686aa8 | 361 | _("bad state %d, inode %" PRIu64 " bmap block 0x%" PRIx64 "\n"), |
5d1b7f0f | 362 | state, ino, bno); |
2bd0ea18 NS |
363 | break; |
364 | } | |
586f8abf | 365 | pthread_mutex_unlock(&ag_locks[agno].lock); |
2bd0ea18 NS |
366 | } else { |
367 | /* | |
368 | * attribute fork for realtime files is in the regular | |
369 | * filesystem | |
370 | */ | |
371 | if (type != XR_INO_RTDATA || whichfork != XFS_DATA_FORK) { | |
79872d6e BN |
372 | if (search_dup_extent(XFS_FSB_TO_AGNO(mp, bno), |
373 | XFS_FSB_TO_AGBNO(mp, bno), | |
374 | XFS_FSB_TO_AGBNO(mp, bno) + 1)) | |
2bd0ea18 NS |
375 | return(1); |
376 | } else { | |
377 | if (search_rt_dup_extent(mp, bno)) | |
378 | return(1); | |
379 | } | |
380 | } | |
381 | (*tot)++; | |
5e656dbb BN |
382 | numrecs = be16_to_cpu(block->bb_numrecs); |
383 | ||
00efc33a DW |
384 | /* Record BMBT blocks in the reverse-mapping data. */ |
385 | if (check_dups && collect_rmaps) { | |
2d273771 | 386 | error = rmap_add_bmbt_rec(mp, ino, whichfork, bno); |
00efc33a DW |
387 | if (error) |
388 | do_error( | |
389 | _("couldn't add inode %"PRIu64" bmbt block %"PRIu64" reverse-mapping data."), | |
390 | ino, bno); | |
391 | } | |
392 | ||
2bd0ea18 | 393 | if (level == 0) { |
5e656dbb BN |
394 | if (numrecs > mp->m_bmap_dmxr[0] || (isroot == 0 && numrecs < |
395 | mp->m_bmap_dmnr[0])) { | |
507f4e33 | 396 | do_warn( |
b52923f3 | 397 | _("inode %" PRIu64 " bad # of bmap records (%u, min - %u, max - %u)\n"), |
5e656dbb BN |
398 | ino, numrecs, mp->m_bmap_dmnr[0], |
399 | mp->m_bmap_dmxr[0]); | |
2bd0ea18 NS |
400 | return(1); |
401 | } | |
b3563c19 | 402 | rp = XFS_BMBT_REC_ADDR(mp, block, 1); |
5e656dbb | 403 | *nex += numrecs; |
2bd0ea18 NS |
404 | /* |
405 | * XXX - if we were going to fix up the btree record, | |
406 | * we'd do it right here. For now, if there's a problem, | |
407 | * we'll bail out and presumably clear the inode. | |
408 | */ | |
409 | if (check_dups == 0) { | |
e1f43b4c CH |
410 | err = process_bmbt_reclist(mp, rp, &numrecs, type, ino, |
411 | tot, blkmapp, &first_key, | |
412 | &last_key, whichfork); | |
2bd0ea18 | 413 | if (err) |
e1f43b4c CH |
414 | return 1; |
415 | ||
2bd0ea18 NS |
416 | /* |
417 | * check that key ordering is monotonically increasing. | |
418 | * if the last_key value in the cursor is set to | |
5a35bf2c | 419 | * NULLFILEOFF, then we know this is the first block |
2bd0ea18 NS |
420 | * on the leaf level and we shouldn't check the |
421 | * last_key value. | |
422 | */ | |
423 | if (first_key <= bm_cursor->level[level].last_key && | |
424 | bm_cursor->level[level].last_key != | |
5a35bf2c | 425 | NULLFILEOFF) { |
2bd0ea18 | 426 | do_warn( |
5d1b7f0f | 427 | _("out-of-order bmap key (file offset) in inode %" PRIu64 ", %s fork, fsbno %" PRIu64 "\n"), |
2bd0ea18 NS |
428 | ino, forkname, bno); |
429 | return(1); | |
430 | } | |
431 | /* | |
432 | * update cursor keys to reflect this block. | |
433 | * don't have to check if last_key is > first_key | |
434 | * since that gets checked by process_bmbt_reclist. | |
435 | */ | |
436 | bm_cursor->level[level].first_key = first_key; | |
437 | bm_cursor->level[level].last_key = last_key; | |
438 | ||
e1f43b4c CH |
439 | return 0; |
440 | } else { | |
441 | return scan_bmbt_reclist(mp, rp, &numrecs, type, ino, | |
442 | tot, whichfork); | |
443 | } | |
2bd0ea18 | 444 | } |
5e656dbb BN |
445 | if (numrecs > mp->m_bmap_dmxr[1] || (isroot == 0 && numrecs < |
446 | mp->m_bmap_dmnr[1])) { | |
507f4e33 | 447 | do_warn( |
2e686aa8 | 448 | _("inode %" PRIu64 " bad # of bmap records (%u, min - %u, max - %u)\n"), |
5e656dbb | 449 | ino, numrecs, mp->m_bmap_dmnr[1], mp->m_bmap_dmxr[1]); |
2bd0ea18 NS |
450 | return(1); |
451 | } | |
b3563c19 BN |
452 | pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]); |
453 | pkey = XFS_BMBT_KEY_ADDR(mp, block, 1); | |
2bd0ea18 | 454 | |
5a35bf2c | 455 | last_key = NULLFILEOFF; |
2bd0ea18 | 456 | |
5e656dbb | 457 | for (i = 0, err = 0; i < numrecs; i++) { |
2bd0ea18 NS |
458 | /* |
459 | * XXX - if we were going to fix up the interior btree nodes, | |
460 | * we'd do it right here. For now, if there's a problem, | |
461 | * we'll bail out and presumably clear the inode. | |
462 | */ | |
5e656dbb | 463 | if (!verify_dfsbno(mp, be64_to_cpu(pp[i]))) { |
5d1b7f0f CH |
464 | do_warn( |
465 | _("bad bmap btree ptr 0x%llx in ino %" PRIu64 "\n"), | |
466 | (unsigned long long) be64_to_cpu(pp[i]), ino); | |
2bd0ea18 NS |
467 | return(1); |
468 | } | |
469 | ||
e0607266 | 470 | err = scan_lbtree(be64_to_cpu(pp[i]), level, scan_bmapbt, |
5e656dbb | 471 | type, whichfork, ino, tot, nex, blkmapp, |
e0607266 DC |
472 | bm_cursor, 0, check_dups, magic, |
473 | &xfs_bmbt_buf_ops); | |
2bd0ea18 NS |
474 | if (err) |
475 | return(1); | |
476 | ||
477 | /* | |
478 | * fix key (offset) mismatches between the first key | |
479 | * in the child block (as recorded in the cursor) and the | |
480 | * key in the interior node referencing the child block. | |
481 | * | |
482 | * fixes cases where entries have been shifted between | |
483 | * child blocks but the parent hasn't been updated. We | |
484 | * don't have to worry about the key values in the cursor | |
485 | * not being set since we only look at the key values of | |
486 | * our child and those are guaranteed to be set by the | |
487 | * call to scan_lbtree() above. | |
488 | */ | |
5e656dbb | 489 | if (check_dups == 0 && be64_to_cpu(pkey[i].br_startoff) != |
2bd0ea18 NS |
490 | bm_cursor->level[level-1].first_key) { |
491 | if (!no_modify) { | |
492 | do_warn( | |
5d1b7f0f CH |
493 | _("correcting bt key (was %llu, now %" PRIu64 ") in inode %" PRIu64 "\n" |
494 | "\t\t%s fork, btree block %" PRIu64 "\n"), | |
495 | (unsigned long long) | |
496 | be64_to_cpu(pkey[i].br_startoff), | |
2bd0ea18 | 497 | bm_cursor->level[level-1].first_key, |
507f4e33 | 498 | ino, |
2bd0ea18 NS |
499 | forkname, bno); |
500 | *dirty = 1; | |
5e656dbb | 501 | pkey[i].br_startoff = cpu_to_be64( |
507f4e33 | 502 | bm_cursor->level[level-1].first_key); |
2bd0ea18 NS |
503 | } else { |
504 | do_warn( | |
5d1b7f0f CH |
505 | _("bad btree key (is %llu, should be %" PRIu64 ") in inode %" PRIu64 "\n" |
506 | "\t\t%s fork, btree block %" PRIu64 "\n"), | |
507 | (unsigned long long) | |
508 | be64_to_cpu(pkey[i].br_startoff), | |
2bd0ea18 | 509 | bm_cursor->level[level-1].first_key, |
5e656dbb | 510 | ino, forkname, bno); |
2bd0ea18 NS |
511 | } |
512 | } | |
513 | } | |
514 | ||
515 | /* | |
e8cb94ee | 516 | * If we're the last node at our level, check that the last child |
4dcea5f4 | 517 | * block's forward sibling pointer is NULL. |
2bd0ea18 | 518 | */ |
dfc130f3 | 519 | if (check_dups == 0 && |
5a35bf2c DC |
520 | bm_cursor->level[level].right_fsbno == NULLFSBLOCK && |
521 | bm_cursor->level[level - 1].right_fsbno != NULLFSBLOCK) { | |
2bd0ea18 | 522 | do_warn( |
5a35bf2c | 523 | _("bad fwd (right) sibling pointer (saw %" PRIu64 " should be NULLFSBLOCK)\n" |
5d1b7f0f | 524 | "\tin inode %" PRIu64 " (%s fork) bmap btree block %" PRIu64 "\n"), |
507f4e33 | 525 | bm_cursor->level[level - 1].right_fsbno, |
5e656dbb | 526 | ino, forkname, bm_cursor->level[level - 1].fsbno); |
2bd0ea18 NS |
527 | return(1); |
528 | } | |
529 | ||
530 | /* | |
531 | * update cursor keys to reflect this block | |
532 | */ | |
533 | if (check_dups == 0) { | |
534 | bm_cursor->level[level].first_key = | |
5e656dbb | 535 | be64_to_cpu(pkey[0].br_startoff); |
2bd0ea18 | 536 | bm_cursor->level[level].last_key = |
5e656dbb | 537 | be64_to_cpu(pkey[numrecs - 1].br_startoff); |
2bd0ea18 NS |
538 | } |
539 | ||
540 | return(0); | |
541 | } | |
542 | ||
364a126c | 543 | static void |
e0607266 | 544 | scan_allocbt( |
b3563c19 | 545 | struct xfs_btree_block *block, |
2bd0ea18 NS |
546 | int level, |
547 | xfs_agblock_t bno, | |
548 | xfs_agnumber_t agno, | |
549 | int suspect, | |
64a0b0ab | 550 | int isroot, |
364a126c | 551 | __uint32_t magic, |
e0607266 | 552 | void *priv) |
2bd0ea18 | 553 | { |
e0607266 | 554 | struct aghdr_cnts *agcnts = priv; |
64a0b0ab | 555 | const char *name; |
2bd0ea18 NS |
556 | int i; |
557 | xfs_alloc_ptr_t *pp; | |
558 | xfs_alloc_rec_t *rp; | |
559 | int hdr_errors = 0; | |
560 | int numrecs; | |
561 | int state; | |
c9638a82 CH |
562 | xfs_extlen_t lastcount = 0; |
563 | xfs_agblock_t lastblock = 0; | |
2bd0ea18 | 564 | |
e0607266 DC |
565 | switch (magic) { |
566 | case XFS_ABTB_CRC_MAGIC: | |
567 | case XFS_ABTB_MAGIC: | |
568 | name = "bno"; | |
569 | break; | |
570 | case XFS_ABTC_CRC_MAGIC: | |
571 | case XFS_ABTC_MAGIC: | |
572 | name = "cnt"; | |
573 | break; | |
574 | default: | |
6bddecbc | 575 | name = "(unknown)"; |
e0607266 DC |
576 | assert(0); |
577 | break; | |
578 | } | |
64a0b0ab CH |
579 | |
580 | if (be32_to_cpu(block->bb_magic) != magic) { | |
581 | do_warn(_("bad magic # %#x in bt%s block %d/%d\n"), | |
582 | be32_to_cpu(block->bb_magic), name, agno, bno); | |
2bd0ea18 NS |
583 | hdr_errors++; |
584 | if (suspect) | |
585 | return; | |
586 | } | |
48d49157 CH |
587 | |
588 | /* | |
589 | * All freespace btree blocks except the roots are freed for a | |
590 | * fully used filesystem, thus they are counted towards the | |
591 | * free data block counter. | |
592 | */ | |
593 | if (!isroot) { | |
364a126c DC |
594 | agcnts->agfbtreeblks++; |
595 | agcnts->fdblocks++; | |
48d49157 CH |
596 | } |
597 | ||
5e656dbb | 598 | if (be16_to_cpu(block->bb_level) != level) { |
64a0b0ab CH |
599 | do_warn(_("expected level %d got %d in bt%s block %d/%d\n"), |
600 | level, be16_to_cpu(block->bb_level), name, agno, bno); | |
2bd0ea18 NS |
601 | hdr_errors++; |
602 | if (suspect) | |
603 | return; | |
604 | } | |
605 | ||
606 | /* | |
607 | * check for btree blocks multiply claimed | |
608 | */ | |
95650c4d BN |
609 | state = get_bmap(agno, bno); |
610 | if (state != XR_E_UNKNOWN) { | |
611 | set_bmap(agno, bno, XR_E_MULT); | |
2bd0ea18 | 612 | do_warn( |
64a0b0ab CH |
613 | _("%s freespace btree block claimed (state %d), agno %d, bno %d, suspect %d\n"), |
614 | name, state, agno, bno, suspect); | |
2bd0ea18 NS |
615 | return; |
616 | } | |
95650c4d | 617 | set_bmap(agno, bno, XR_E_FS_MAP); |
2bd0ea18 | 618 | |
5e656dbb BN |
619 | numrecs = be16_to_cpu(block->bb_numrecs); |
620 | ||
2bd0ea18 | 621 | if (level == 0) { |
5e656dbb | 622 | if (numrecs > mp->m_alloc_mxr[0]) { |
2bd0ea18 NS |
623 | numrecs = mp->m_alloc_mxr[0]; |
624 | hdr_errors++; | |
625 | } | |
5e656dbb | 626 | if (isroot == 0 && numrecs < mp->m_alloc_mnr[0]) { |
2bd0ea18 NS |
627 | numrecs = mp->m_alloc_mnr[0]; |
628 | hdr_errors++; | |
629 | } | |
630 | ||
c9638a82 CH |
631 | if (hdr_errors) { |
632 | do_warn( | |
633 | _("bad btree nrecs (%u, min=%u, max=%u) in bt%s block %u/%u\n"), | |
634 | be16_to_cpu(block->bb_numrecs), | |
635 | mp->m_alloc_mnr[0], mp->m_alloc_mxr[0], | |
636 | name, agno, bno); | |
2bd0ea18 | 637 | suspect++; |
c9638a82 | 638 | } |
2bd0ea18 | 639 | |
b3563c19 | 640 | rp = XFS_ALLOC_REC_ADDR(mp, block, 1); |
2bd0ea18 | 641 | for (i = 0; i < numrecs; i++) { |
08382cf4 | 642 | xfs_agblock_t b, end; |
8961bfde | 643 | xfs_extlen_t len, blen; |
08382cf4 BN |
644 | |
645 | b = be32_to_cpu(rp[i].ar_startblock); | |
646 | len = be32_to_cpu(rp[i].ar_blockcount); | |
647 | end = b + len; | |
2bd0ea18 | 648 | |
c83b756d CH |
649 | if (b == 0 || !verify_agbno(mp, agno, b)) { |
650 | do_warn( | |
5d1b7f0f | 651 | _("invalid start block %u in record %u of %s btree block %u/%u\n"), |
c83b756d | 652 | b, i, name, agno, bno); |
08382cf4 | 653 | continue; |
c83b756d CH |
654 | } |
655 | if (len == 0 || !verify_agbno(mp, agno, end - 1)) { | |
656 | do_warn( | |
5d1b7f0f | 657 | _("invalid length %u in record %u of %s btree block %u/%u\n"), |
c83b756d | 658 | len, i, name, agno, bno); |
909daa84 | 659 | continue; |
c83b756d | 660 | } |
08382cf4 | 661 | |
e0607266 DC |
662 | if (magic == XFS_ABTB_MAGIC || |
663 | magic == XFS_ABTB_CRC_MAGIC) { | |
c9638a82 CH |
664 | if (b <= lastblock) { |
665 | do_warn(_( | |
666 | "out-of-order bno btree record %d (%u %u) block %u/%u\n"), | |
667 | i, b, len, agno, bno); | |
668 | } else { | |
669 | lastblock = b; | |
670 | } | |
671 | } else { | |
364a126c DC |
672 | agcnts->fdblocks += len; |
673 | agcnts->agffreeblks += len; | |
674 | if (len > agcnts->agflongest) | |
675 | agcnts->agflongest = len; | |
c9638a82 CH |
676 | if (len < lastcount) { |
677 | do_warn(_( | |
678 | "out-of-order cnt btree record %d (%u %u) block %u/%u\n"), | |
679 | i, b, len, agno, bno); | |
680 | } else { | |
681 | lastcount = len; | |
682 | } | |
683 | } | |
684 | ||
8961bfde BN |
685 | for ( ; b < end; b += blen) { |
686 | state = get_bmap_ext(agno, b, end, &blen); | |
64a0b0ab CH |
687 | switch (state) { |
688 | case XR_E_UNKNOWN: | |
95650c4d | 689 | set_bmap(agno, b, XR_E_FREE1); |
64a0b0ab CH |
690 | break; |
691 | case XR_E_FREE1: | |
692 | /* | |
693 | * no warning messages -- we'll catch | |
694 | * FREE1 blocks later | |
695 | */ | |
e0607266 DC |
696 | if (magic == XFS_ABTC_MAGIC || |
697 | magic == XFS_ABTC_CRC_MAGIC) { | |
8961bfde BN |
698 | set_bmap_ext(agno, b, blen, |
699 | XR_E_FREE); | |
64a0b0ab CH |
700 | break; |
701 | } | |
702 | default: | |
507f4e33 | 703 | do_warn( |
8961bfde BN |
704 | _("block (%d,%d-%d) multiply claimed by %s space tree, state - %d\n"), |
705 | agno, b, b + blen - 1, | |
706 | name, state); | |
64a0b0ab | 707 | break; |
2bd0ea18 NS |
708 | } |
709 | } | |
710 | } | |
711 | return; | |
712 | } | |
713 | ||
714 | /* | |
715 | * interior record | |
716 | */ | |
b3563c19 | 717 | pp = XFS_ALLOC_PTR_ADDR(mp, block, 1, mp->m_alloc_mxr[1]); |
2bd0ea18 | 718 | |
5e656dbb | 719 | if (numrecs > mp->m_alloc_mxr[1]) { |
2bd0ea18 NS |
720 | numrecs = mp->m_alloc_mxr[1]; |
721 | hdr_errors++; | |
722 | } | |
5e656dbb | 723 | if (isroot == 0 && numrecs < mp->m_alloc_mnr[1]) { |
2bd0ea18 NS |
724 | numrecs = mp->m_alloc_mnr[1]; |
725 | hdr_errors++; | |
726 | } | |
727 | ||
728 | /* | |
729 | * don't pass bogus tree flag down further if this block | |
730 | * looked ok. bail out if two levels in a row look bad. | |
731 | */ | |
2bd0ea18 | 732 | if (hdr_errors) { |
c9638a82 CH |
733 | do_warn( |
734 | _("bad btree nrecs (%u, min=%u, max=%u) in bt%s block %u/%u\n"), | |
735 | be16_to_cpu(block->bb_numrecs), | |
736 | mp->m_alloc_mnr[1], mp->m_alloc_mxr[1], | |
737 | name, agno, bno); | |
2bd0ea18 NS |
738 | if (suspect) |
739 | return; | |
c9638a82 CH |
740 | suspect++; |
741 | } else if (suspect) { | |
742 | suspect = 0; | |
2bd0ea18 NS |
743 | } |
744 | ||
745 | for (i = 0; i < numrecs; i++) { | |
08382cf4 BN |
746 | xfs_agblock_t bno = be32_to_cpu(pp[i]); |
747 | ||
2bd0ea18 NS |
748 | /* |
749 | * XXX - put sibling detection right here. | |
750 | * we know our sibling chain is good. So as we go, | |
751 | * we check the entry before and after each entry. | |
752 | * If either of the entries references a different block, | |
753 | * check the sibling pointer. If there's a sibling | |
754 | * pointer mismatch, try and extract as much data | |
dfc130f3 | 755 | * as possible. |
2bd0ea18 | 756 | */ |
08382cf4 | 757 | if (bno != 0 && verify_agbno(mp, agno, bno)) { |
e0607266 DC |
758 | switch (magic) { |
759 | case XFS_ABTB_CRC_MAGIC: | |
760 | case XFS_ABTB_MAGIC: | |
761 | scan_sbtree(bno, level, agno, suspect, | |
762 | scan_allocbt, 0, magic, priv, | |
763 | &xfs_allocbt_buf_ops); | |
764 | break; | |
765 | case XFS_ABTC_CRC_MAGIC: | |
766 | case XFS_ABTC_MAGIC: | |
767 | scan_sbtree(bno, level, agno, suspect, | |
768 | scan_allocbt, 0, magic, priv, | |
769 | &xfs_allocbt_buf_ops); | |
770 | break; | |
771 | } | |
08382cf4 | 772 | } |
2bd0ea18 NS |
773 | } |
774 | } | |
775 | ||
c749bd55 BF |
776 | static bool |
777 | ino_issparse( | |
778 | struct xfs_inobt_rec *rp, | |
779 | int offset) | |
780 | { | |
781 | if (!xfs_sb_version_hassparseinodes(&mp->m_sb)) | |
782 | return false; | |
783 | ||
784 | return xfs_inobt_is_sparse_disk(rp, offset); | |
785 | } | |
786 | ||
0f94fa4b DW |
787 | /* See if the rmapbt owners agree with our observations. */ |
788 | static void | |
789 | process_rmap_rec( | |
790 | struct xfs_mount *mp, | |
791 | xfs_agnumber_t agno, | |
792 | xfs_agblock_t b, | |
793 | xfs_agblock_t end, | |
794 | xfs_extlen_t blen, | |
795 | int64_t owner, | |
796 | int state, | |
797 | const char *name) | |
798 | { | |
799 | switch (state) { | |
800 | case XR_E_UNKNOWN: | |
801 | switch (owner) { | |
802 | case XFS_RMAP_OWN_FS: | |
803 | case XFS_RMAP_OWN_LOG: | |
804 | set_bmap_ext(agno, b, blen, XR_E_INUSE_FS1); | |
805 | break; | |
806 | case XFS_RMAP_OWN_AG: | |
807 | case XFS_RMAP_OWN_INOBT: | |
808 | set_bmap_ext(agno, b, blen, XR_E_FS_MAP1); | |
809 | break; | |
810 | case XFS_RMAP_OWN_INODES: | |
811 | set_bmap_ext(agno, b, blen, XR_E_INO1); | |
812 | break; | |
813 | case XFS_RMAP_OWN_NULL: | |
814 | /* still unknown */ | |
815 | break; | |
816 | default: | |
817 | /* file data */ | |
818 | set_bmap_ext(agno, b, blen, XR_E_INUSE1); | |
819 | break; | |
820 | } | |
821 | break; | |
822 | case XR_E_INUSE_FS: | |
823 | if (owner == XFS_RMAP_OWN_FS || | |
824 | owner == XFS_RMAP_OWN_LOG) | |
825 | break; | |
826 | do_warn( | |
827 | _("Static meta block (%d,%d-%d) mismatch in %s tree, state - %d,%" PRIx64 "\n"), | |
828 | agno, b, b + blen - 1, | |
829 | name, state, owner); | |
830 | break; | |
831 | case XR_E_FS_MAP: | |
832 | if (owner == XFS_RMAP_OWN_AG || | |
833 | owner == XFS_RMAP_OWN_INOBT) | |
834 | break; | |
835 | do_warn( | |
836 | _("AG meta block (%d,%d-%d) mismatch in %s tree, state - %d,%" PRIx64 "\n"), | |
837 | agno, b, b + blen - 1, | |
838 | name, state, owner); | |
839 | break; | |
840 | case XR_E_INO: | |
841 | if (owner == XFS_RMAP_OWN_INODES) | |
842 | break; | |
843 | do_warn( | |
844 | _("inode block (%d,%d-%d) mismatch in %s tree, state - %d,%" PRIx64 "\n"), | |
845 | agno, b, b + blen - 1, | |
846 | name, state, owner); | |
847 | break; | |
848 | case XR_E_INUSE: | |
849 | if (owner >= 0 && | |
850 | owner < mp->m_sb.sb_dblocks) | |
851 | break; | |
852 | do_warn( | |
853 | _("in use block (%d,%d-%d) mismatch in %s tree, state - %d,%" PRIx64 "\n"), | |
854 | agno, b, b + blen - 1, | |
855 | name, state, owner); | |
856 | break; | |
857 | case XR_E_FREE1: | |
858 | case XR_E_FREE: | |
859 | /* | |
860 | * May be on the AGFL. If not, they'll | |
861 | * be caught later. | |
862 | */ | |
863 | break; | |
864 | default: | |
865 | do_warn( | |
866 | _("unknown block (%d,%d-%d) mismatch on %s tree, state - %d,%" PRIx64 "\n"), | |
867 | agno, b, b + blen - 1, | |
868 | name, state, owner); | |
869 | break; | |
870 | } | |
871 | } | |
872 | ||
11b9e510 DW |
873 | struct rmap_priv { |
874 | struct aghdr_cnts *agcnts; | |
875 | struct xfs_rmap_irec high_key; | |
65841734 | 876 | struct xfs_rmap_irec last_rec; |
11b9e510 DW |
877 | xfs_agblock_t nr_blocks; |
878 | }; | |
879 | ||
0f94fa4b DW |
880 | static void |
881 | scan_rmapbt( | |
882 | struct xfs_btree_block *block, | |
883 | int level, | |
884 | xfs_agblock_t bno, | |
885 | xfs_agnumber_t agno, | |
886 | int suspect, | |
887 | int isroot, | |
888 | __uint32_t magic, | |
889 | void *priv) | |
890 | { | |
0f94fa4b DW |
891 | const char *name = "rmap"; |
892 | int i; | |
893 | xfs_rmap_ptr_t *pp; | |
894 | struct xfs_rmap_rec *rp; | |
11b9e510 | 895 | struct rmap_priv *rmap_priv = priv; |
0f94fa4b DW |
896 | int hdr_errors = 0; |
897 | int numrecs; | |
898 | int state; | |
899 | xfs_agblock_t lastblock = 0; | |
11b9e510 | 900 | struct xfs_rmap_key *kp; |
138ce9ff | 901 | struct xfs_rmap_irec key = {0}; |
0f94fa4b DW |
902 | |
903 | if (magic != XFS_RMAP_CRC_MAGIC) { | |
904 | name = "(unknown)"; | |
11b9e510 DW |
905 | hdr_errors++; |
906 | suspect++; | |
907 | goto out; | |
0f94fa4b DW |
908 | } |
909 | ||
910 | if (be32_to_cpu(block->bb_magic) != magic) { | |
911 | do_warn(_("bad magic # %#x in bt%s block %d/%d\n"), | |
912 | be32_to_cpu(block->bb_magic), name, agno, bno); | |
913 | hdr_errors++; | |
914 | if (suspect) | |
11b9e510 | 915 | goto out; |
0f94fa4b DW |
916 | } |
917 | ||
918 | /* | |
919 | * All RMAP btree blocks except the roots are freed for a | |
920 | * fully empty filesystem, thus they are counted towards the | |
921 | * free data block counter. | |
922 | */ | |
923 | if (!isroot) { | |
11b9e510 DW |
924 | rmap_priv->agcnts->agfbtreeblks++; |
925 | rmap_priv->agcnts->fdblocks++; | |
0f94fa4b | 926 | } |
11b9e510 | 927 | rmap_priv->nr_blocks++; |
0f94fa4b DW |
928 | |
929 | if (be16_to_cpu(block->bb_level) != level) { | |
930 | do_warn(_("expected level %d got %d in bt%s block %d/%d\n"), | |
931 | level, be16_to_cpu(block->bb_level), name, agno, bno); | |
932 | hdr_errors++; | |
933 | if (suspect) | |
11b9e510 | 934 | goto out; |
0f94fa4b DW |
935 | } |
936 | ||
937 | /* check for btree blocks multiply claimed */ | |
938 | state = get_bmap(agno, bno); | |
939 | if (!(state == XR_E_UNKNOWN || state == XR_E_FS_MAP1)) { | |
940 | set_bmap(agno, bno, XR_E_MULT); | |
941 | do_warn( | |
942 | _("%s rmap btree block claimed (state %d), agno %d, bno %d, suspect %d\n"), | |
943 | name, state, agno, bno, suspect); | |
11b9e510 | 944 | goto out; |
0f94fa4b DW |
945 | } |
946 | set_bmap(agno, bno, XR_E_FS_MAP); | |
947 | ||
948 | numrecs = be16_to_cpu(block->bb_numrecs); | |
949 | if (level == 0) { | |
950 | if (numrecs > mp->m_rmap_mxr[0]) { | |
951 | numrecs = mp->m_rmap_mxr[0]; | |
952 | hdr_errors++; | |
953 | } | |
954 | if (isroot == 0 && numrecs < mp->m_rmap_mnr[0]) { | |
955 | numrecs = mp->m_rmap_mnr[0]; | |
956 | hdr_errors++; | |
957 | } | |
958 | ||
959 | if (hdr_errors) { | |
960 | do_warn( | |
961 | _("bad btree nrecs (%u, min=%u, max=%u) in bt%s block %u/%u\n"), | |
962 | be16_to_cpu(block->bb_numrecs), | |
963 | mp->m_rmap_mnr[0], mp->m_rmap_mxr[0], | |
964 | name, agno, bno); | |
965 | suspect++; | |
966 | } | |
967 | ||
968 | rp = XFS_RMAP_REC_ADDR(block, 1); | |
969 | for (i = 0; i < numrecs; i++) { | |
970 | xfs_agblock_t b, end; | |
971 | xfs_extlen_t len, blen; | |
972 | int64_t owner, offset; | |
973 | ||
974 | b = be32_to_cpu(rp[i].rm_startblock); | |
975 | len = be32_to_cpu(rp[i].rm_blockcount); | |
976 | owner = be64_to_cpu(rp[i].rm_owner); | |
977 | offset = be64_to_cpu(rp[i].rm_offset); | |
11b9e510 DW |
978 | |
979 | key.rm_flags = 0; | |
980 | key.rm_startblock = b; | |
981 | key.rm_blockcount = len; | |
982 | key.rm_owner = owner; | |
983 | if (libxfs_rmap_irec_offset_unpack(offset, &key)) { | |
984 | /* Look for impossible flags. */ | |
985 | do_warn( | |
986 | _("invalid flags in record %u of %s btree block %u/%u\n"), | |
987 | i, name, agno, bno); | |
988 | continue; | |
989 | } | |
990 | ||
991 | end = key.rm_startblock + key.rm_blockcount; | |
0f94fa4b DW |
992 | |
993 | /* Make sure agbno & len make sense. */ | |
994 | if (!verify_agbno(mp, agno, b)) { | |
995 | do_warn( | |
996 | _("invalid start block %u in record %u of %s btree block %u/%u\n"), | |
997 | b, i, name, agno, bno); | |
998 | continue; | |
999 | } | |
1000 | if (len == 0 || !verify_agbno(mp, agno, end - 1)) { | |
1001 | do_warn( | |
1002 | _("invalid length %u in record %u of %s btree block %u/%u\n"), | |
1003 | len, i, name, agno, bno); | |
1004 | continue; | |
1005 | } | |
1006 | ||
1007 | /* Look for impossible owners. */ | |
1008 | if (!(owner > 0 || (owner > XFS_RMAP_OWN_MIN && | |
1009 | owner <= XFS_RMAP_OWN_FS))) | |
1010 | do_warn( | |
1011 | _("invalid owner in rmap btree record %d (%"PRId64" %u) block %u/%u\n"), | |
1012 | i, owner, len, agno, bno); | |
1013 | ||
a418fe7d DW |
1014 | /* Look for impossible record field combinations. */ |
1015 | if (XFS_RMAP_NON_INODE_OWNER(key.rm_owner)) { | |
1016 | if (key.rm_flags) | |
1017 | do_warn( | |
1018 | _("record %d of block (%u/%u) in %s btree cannot have non-inode owner with flags\n"), | |
1019 | i, agno, bno, name); | |
1020 | if (key.rm_offset) | |
1021 | do_warn( | |
1022 | _("record %d of block (%u/%u) in %s btree cannot have non-inode owner with offset\n"), | |
1023 | i, agno, bno, name); | |
1024 | } | |
1025 | ||
0f94fa4b DW |
1026 | /* Check for out of order records. */ |
1027 | if (i == 0) { | |
1028 | advance: | |
1029 | lastblock = b; | |
0f94fa4b DW |
1030 | } else { |
1031 | bool bad; | |
1032 | ||
1033 | bad = b <= lastblock; | |
1034 | if (bad) | |
1035 | do_warn( | |
1036 | _("out-of-order rmap btree record %d (%u %"PRId64" %"PRIx64" %u) block %u/%u\n"), | |
1037 | i, b, owner, offset, len, agno, bno); | |
1038 | else | |
1039 | goto advance; | |
1040 | } | |
1041 | ||
65841734 | 1042 | /* Is this mergeable with the previous record? */ |
2d273771 | 1043 | if (rmaps_are_mergeable(&rmap_priv->last_rec, &key)) { |
65841734 DW |
1044 | do_warn( |
1045 | _("record %d in block (%u/%u) of %s tree should be merged with previous record\n"), | |
1046 | i, agno, bno, name); | |
1047 | rmap_priv->last_rec.rm_blockcount += | |
1048 | key.rm_blockcount; | |
1049 | } else | |
1050 | rmap_priv->last_rec = key; | |
1051 | ||
11b9e510 DW |
1052 | /* Check that we don't go past the high key. */ |
1053 | key.rm_startblock += key.rm_blockcount - 1; | |
1054 | if (!XFS_RMAP_NON_INODE_OWNER(key.rm_owner) && | |
1055 | !(key.rm_flags & XFS_RMAP_BMBT_BLOCK)) | |
1056 | key.rm_offset += key.rm_blockcount - 1; | |
1057 | key.rm_blockcount = 0; | |
1058 | if (rmap_diffkeys(&key, &rmap_priv->high_key) > 0) { | |
1059 | do_warn( | |
1060 | _("record %d greater than high key of block (%u/%u) in %s tree\n"), | |
1061 | i, agno, bno, name); | |
1062 | } | |
1063 | ||
0f94fa4b DW |
1064 | /* Check for block owner collisions. */ |
1065 | for ( ; b < end; b += blen) { | |
1066 | state = get_bmap_ext(agno, b, end, &blen); | |
1067 | process_rmap_rec(mp, agno, b, end, blen, owner, | |
1068 | state, name); | |
1069 | } | |
1070 | } | |
11b9e510 | 1071 | goto out; |
0f94fa4b DW |
1072 | } |
1073 | ||
1074 | /* | |
1075 | * interior record | |
1076 | */ | |
1077 | pp = XFS_RMAP_PTR_ADDR(block, 1, mp->m_rmap_mxr[1]); | |
1078 | ||
1079 | if (numrecs > mp->m_rmap_mxr[1]) { | |
1080 | numrecs = mp->m_rmap_mxr[1]; | |
1081 | hdr_errors++; | |
1082 | } | |
1083 | if (isroot == 0 && numrecs < mp->m_rmap_mnr[1]) { | |
1084 | numrecs = mp->m_rmap_mnr[1]; | |
1085 | hdr_errors++; | |
1086 | } | |
1087 | ||
1088 | /* | |
1089 | * don't pass bogus tree flag down further if this block | |
1090 | * looked ok. bail out if two levels in a row look bad. | |
1091 | */ | |
1092 | if (hdr_errors) { | |
1093 | do_warn( | |
1094 | _("bad btree nrecs (%u, min=%u, max=%u) in bt%s block %u/%u\n"), | |
1095 | be16_to_cpu(block->bb_numrecs), | |
1096 | mp->m_rmap_mnr[1], mp->m_rmap_mxr[1], | |
1097 | name, agno, bno); | |
1098 | if (suspect) | |
11b9e510 | 1099 | goto out; |
0f94fa4b DW |
1100 | suspect++; |
1101 | } else if (suspect) { | |
1102 | suspect = 0; | |
1103 | } | |
1104 | ||
11b9e510 DW |
1105 | /* check the node's high keys */ |
1106 | for (i = 0; !isroot && i < numrecs; i++) { | |
1107 | kp = XFS_RMAP_HIGH_KEY_ADDR(block, i + 1); | |
1108 | ||
1109 | key.rm_flags = 0; | |
1110 | key.rm_startblock = be32_to_cpu(kp->rm_startblock); | |
1111 | key.rm_owner = be64_to_cpu(kp->rm_owner); | |
1112 | if (libxfs_rmap_irec_offset_unpack(be64_to_cpu(kp->rm_offset), | |
1113 | &key)) { | |
1114 | /* Look for impossible flags. */ | |
1115 | do_warn( | |
1116 | _("invalid flags in key %u of %s btree block %u/%u\n"), | |
1117 | i, name, agno, bno); | |
1118 | continue; | |
1119 | } | |
1120 | if (rmap_diffkeys(&key, &rmap_priv->high_key) > 0) | |
1121 | do_warn( | |
1122 | _("key %d greater than high key of block (%u/%u) in %s tree\n"), | |
1123 | i, agno, bno, name); | |
1124 | } | |
1125 | ||
0f94fa4b DW |
1126 | for (i = 0; i < numrecs; i++) { |
1127 | xfs_agblock_t bno = be32_to_cpu(pp[i]); | |
1128 | ||
1129 | /* | |
1130 | * XXX - put sibling detection right here. | |
1131 | * we know our sibling chain is good. So as we go, | |
1132 | * we check the entry before and after each entry. | |
1133 | * If either of the entries references a different block, | |
1134 | * check the sibling pointer. If there's a sibling | |
1135 | * pointer mismatch, try and extract as much data | |
1136 | * as possible. | |
1137 | */ | |
11b9e510 DW |
1138 | kp = XFS_RMAP_HIGH_KEY_ADDR(block, i + 1); |
1139 | rmap_priv->high_key.rm_flags = 0; | |
1140 | rmap_priv->high_key.rm_startblock = | |
1141 | be32_to_cpu(kp->rm_startblock); | |
1142 | rmap_priv->high_key.rm_owner = | |
1143 | be64_to_cpu(kp->rm_owner); | |
1144 | if (libxfs_rmap_irec_offset_unpack(be64_to_cpu(kp->rm_offset), | |
1145 | &rmap_priv->high_key)) { | |
1146 | /* Look for impossible flags. */ | |
1147 | do_warn( | |
1148 | _("invalid flags in high key %u of %s btree block %u/%u\n"), | |
1149 | i, name, agno, bno); | |
1150 | continue; | |
1151 | } | |
1152 | ||
0f94fa4b DW |
1153 | if (bno != 0 && verify_agbno(mp, agno, bno)) { |
1154 | scan_sbtree(bno, level, agno, suspect, scan_rmapbt, 0, | |
1155 | magic, priv, &xfs_rmapbt_buf_ops); | |
1156 | } | |
1157 | } | |
11b9e510 DW |
1158 | |
1159 | out: | |
1160 | if (suspect) | |
1161 | rmap_avoid_check(); | |
0f94fa4b DW |
1162 | } |
1163 | ||
22458187 BF |
1164 | /* |
1165 | * The following helpers are to help process and validate individual on-disk | |
1166 | * inode btree records. We have two possible inode btrees with slightly | |
1167 | * different semantics. Many of the validations and actions are equivalent, such | |
1168 | * as record alignment constraints, etc. Other validations differ, such as the | |
1169 | * fact that the inode chunk block allocation state is set by the content of the | |
1170 | * core inobt and verified by the content of the finobt. | |
1171 | * | |
1172 | * The following structures are used to facilitate common validation routines | |
1173 | * where the only difference between validation of the inobt or finobt might be | |
1174 | * the error messages that results in the event of failure. | |
1175 | */ | |
1176 | ||
1177 | enum inobt_type { | |
1178 | INOBT, | |
1179 | FINOBT | |
1180 | }; | |
1181 | const char *inobt_names[] = { | |
1182 | "inobt", | |
1183 | "finobt" | |
1184 | }; | |
1185 | ||
6ceb5693 | 1186 | static int |
22458187 | 1187 | verify_single_ino_chunk_align( |
6ceb5693 | 1188 | xfs_agnumber_t agno, |
22458187 BF |
1189 | enum inobt_type type, |
1190 | struct xfs_inobt_rec *rp, | |
1191 | int suspect, | |
1192 | bool *skip) | |
6ceb5693 | 1193 | { |
22458187 | 1194 | const char *inobt_name = inobt_names[type]; |
6ceb5693 BN |
1195 | xfs_ino_t lino; |
1196 | xfs_agino_t ino; | |
1197 | xfs_agblock_t agbno; | |
6ceb5693 | 1198 | int off; |
6ceb5693 | 1199 | |
22458187 | 1200 | *skip = false; |
6ceb5693 BN |
1201 | ino = be32_to_cpu(rp->ir_startino); |
1202 | off = XFS_AGINO_TO_OFFSET(mp, ino); | |
1203 | agbno = XFS_AGINO_TO_AGBNO(mp, ino); | |
1204 | lino = XFS_AGINO_TO_INO(mp, agno, ino); | |
1205 | ||
1206 | /* | |
22458187 BF |
1207 | * on multi-block block chunks, all chunks start at the beginning of the |
1208 | * block. with multi-chunk blocks, all chunks must start on 64-inode | |
1209 | * boundaries since each block can hold N complete chunks. if fs has | |
1210 | * aligned inodes, all chunks must start at a fs_ino_alignment*N'th | |
1211 | * agbno. skip recs with badly aligned starting inodes. | |
6ceb5693 BN |
1212 | */ |
1213 | if (ino == 0 || | |
1214 | (inodes_per_block <= XFS_INODES_PER_CHUNK && off != 0) || | |
1215 | (inodes_per_block > XFS_INODES_PER_CHUNK && | |
1216 | off % XFS_INODES_PER_CHUNK != 0) || | |
ca0a799e | 1217 | (fs_aligned_inodes && fs_ino_alignment && |
22458187 | 1218 | agbno % fs_ino_alignment != 0)) { |
6ceb5693 | 1219 | do_warn( |
22458187 BF |
1220 | _("badly aligned %s rec (starting inode = %" PRIu64 ")\n"), |
1221 | inobt_name, lino); | |
6ceb5693 BN |
1222 | suspect++; |
1223 | } | |
1224 | ||
1225 | /* | |
22458187 BF |
1226 | * verify numeric validity of inode chunk first before inserting into a |
1227 | * tree. don't have to worry about the overflow case because the | |
1228 | * starting ino number of a chunk can only get within 255 inodes of max | |
1229 | * (NULLAGINO). if it gets closer, the agino number will be illegal as | |
1230 | * the agbno will be too large. | |
6ceb5693 | 1231 | */ |
22458187 | 1232 | if (verify_aginum(mp, agno, ino)) { |
6ceb5693 | 1233 | do_warn( |
22458187 BF |
1234 | _("bad starting inode # (%" PRIu64 " (0x%x 0x%x)) in %s rec, skipping rec\n"), |
1235 | lino, agno, ino, inobt_name); | |
1236 | *skip = true; | |
6ceb5693 BN |
1237 | return ++suspect; |
1238 | } | |
1239 | ||
1240 | if (verify_aginum(mp, agno, | |
22458187 | 1241 | ino + XFS_INODES_PER_CHUNK - 1)) { |
6ceb5693 | 1242 | do_warn( |
22458187 | 1243 | _("bad ending inode # (%" PRIu64 " (0x%x 0x%zx)) in %s rec, skipping rec\n"), |
6ceb5693 | 1244 | lino + XFS_INODES_PER_CHUNK - 1, |
5d1b7f0f | 1245 | agno, |
22458187 BF |
1246 | ino + XFS_INODES_PER_CHUNK - 1, |
1247 | inobt_name); | |
1248 | *skip = true; | |
6ceb5693 BN |
1249 | return ++suspect; |
1250 | } | |
1251 | ||
22458187 BF |
1252 | return suspect; |
1253 | } | |
1254 | ||
411c5e55 BF |
1255 | /* |
1256 | * Process the state of individual inodes in an on-disk inobt record and import | |
1257 | * into the appropriate in-core tree based on whether the on-disk tree is | |
1258 | * suspect. Return the total and free inode counts based on the record free and | |
1259 | * hole masks. | |
1260 | */ | |
1261 | static int | |
1262 | import_single_ino_chunk( | |
1263 | xfs_agnumber_t agno, | |
1264 | enum inobt_type type, | |
1265 | struct xfs_inobt_rec *rp, | |
1266 | int suspect, | |
1267 | int *p_nfree, | |
1268 | int *p_ninodes) | |
1269 | { | |
1270 | struct ino_tree_node *ino_rec = NULL; | |
1271 | const char *inobt_name = inobt_names[type]; | |
1272 | xfs_agino_t ino; | |
1273 | int j; | |
1274 | int nfree; | |
1275 | int ninodes; | |
1276 | ||
1277 | ino = be32_to_cpu(rp->ir_startino); | |
1278 | ||
1279 | if (!suspect) { | |
1280 | if (XFS_INOBT_IS_FREE_DISK(rp, 0)) | |
1281 | ino_rec = set_inode_free_alloc(mp, agno, ino); | |
1282 | else | |
1283 | ino_rec = set_inode_used_alloc(mp, agno, ino); | |
1284 | for (j = 1; j < XFS_INODES_PER_CHUNK; j++) { | |
1285 | if (XFS_INOBT_IS_FREE_DISK(rp, j)) | |
1286 | set_inode_free(ino_rec, j); | |
1287 | else | |
1288 | set_inode_used(ino_rec, j); | |
1289 | } | |
1290 | } else { | |
1291 | for (j = 0; j < XFS_INODES_PER_CHUNK; j++) { | |
1292 | if (XFS_INOBT_IS_FREE_DISK(rp, j)) | |
1293 | add_aginode_uncertain(mp, agno, ino + j, 1); | |
1294 | else | |
1295 | add_aginode_uncertain(mp, agno, ino + j, 0); | |
1296 | } | |
1297 | } | |
1298 | ||
1299 | /* | |
1300 | * Mark sparse inodes as such in the in-core tree. Verify that sparse | |
1301 | * inodes are free and that freecount is consistent with the free mask. | |
1302 | */ | |
1303 | nfree = ninodes = 0; | |
1304 | for (j = 0; j < XFS_INODES_PER_CHUNK; j++) { | |
1305 | if (ino_issparse(rp, j)) { | |
1306 | if (!suspect && !XFS_INOBT_IS_FREE_DISK(rp, j)) { | |
1307 | do_warn( | |
1308 | _("ir_holemask/ir_free mismatch, %s chunk %d/%u, holemask 0x%x free 0x%llx\n"), | |
1309 | inobt_name, agno, ino, | |
1310 | be16_to_cpu(rp->ir_u.sp.ir_holemask), | |
1311 | be64_to_cpu(rp->ir_free)); | |
1312 | suspect++; | |
1313 | } | |
1314 | if (!suspect && ino_rec) | |
1315 | set_inode_sparse(ino_rec, j); | |
1316 | } else { | |
1317 | /* count fields track non-sparse inos */ | |
1318 | if (XFS_INOBT_IS_FREE_DISK(rp, j)) | |
1319 | nfree++; | |
1320 | ninodes++; | |
1321 | } | |
1322 | } | |
1323 | ||
1324 | *p_nfree = nfree; | |
1325 | *p_ninodes = ninodes; | |
1326 | ||
1327 | return suspect; | |
1328 | } | |
1329 | ||
22458187 BF |
1330 | static int |
1331 | scan_single_ino_chunk( | |
1332 | xfs_agnumber_t agno, | |
1333 | xfs_inobt_rec_t *rp, | |
1334 | int suspect) | |
1335 | { | |
1336 | xfs_ino_t lino; | |
1337 | xfs_agino_t ino; | |
1338 | xfs_agblock_t agbno; | |
1339 | int j; | |
1340 | int nfree; | |
1341 | int ninodes; | |
1342 | int off; | |
1343 | int state; | |
22458187 BF |
1344 | ino_tree_node_t *first_rec, *last_rec; |
1345 | int freecount; | |
1346 | bool skip = false; | |
1347 | ||
1348 | ino = be32_to_cpu(rp->ir_startino); | |
1349 | off = XFS_AGINO_TO_OFFSET(mp, ino); | |
1350 | agbno = XFS_AGINO_TO_AGBNO(mp, ino); | |
1351 | lino = XFS_AGINO_TO_INO(mp, agno, ino); | |
1352 | freecount = inorec_get_freecount(mp, rp); | |
1353 | ||
1354 | /* | |
1355 | * Verify record alignment, start/end inode numbers, etc. | |
1356 | */ | |
1357 | suspect = verify_single_ino_chunk_align(agno, INOBT, rp, suspect, | |
1358 | &skip); | |
1359 | if (skip) | |
1360 | return suspect; | |
1361 | ||
6ceb5693 BN |
1362 | /* |
1363 | * set state of each block containing inodes | |
1364 | */ | |
1365 | if (off == 0 && !suspect) { | |
1366 | for (j = 0; | |
1367 | j < XFS_INODES_PER_CHUNK; | |
1368 | j += mp->m_sb.sb_inopblock) { | |
95650c4d | 1369 | |
c749bd55 BF |
1370 | /* inodes in sparse chunks don't use blocks */ |
1371 | if (ino_issparse(rp, j)) | |
1372 | continue; | |
1373 | ||
1374 | agbno = XFS_AGINO_TO_AGBNO(mp, ino + j); | |
95650c4d | 1375 | state = get_bmap(agno, agbno); |
0f94fa4b DW |
1376 | switch (state) { |
1377 | case XR_E_INO: | |
1378 | break; | |
1379 | case XR_E_UNKNOWN: | |
1380 | case XR_E_INO1: /* seen by rmap */ | |
95650c4d | 1381 | set_bmap(agno, agbno, XR_E_INO); |
0f94fa4b DW |
1382 | break; |
1383 | case XR_E_INUSE_FS: | |
1384 | case XR_E_INUSE_FS1: | |
1385 | if (agno == 0 && | |
1386 | ino + j >= first_prealloc_ino && | |
1387 | ino + j < last_prealloc_ino) { | |
1388 | set_bmap(agno, agbno, XR_E_INO); | |
1389 | break; | |
1390 | } | |
1391 | /* fall through */ | |
1392 | default: | |
1393 | /* XXX - maybe should mark block a duplicate */ | |
6ceb5693 BN |
1394 | do_warn( |
1395 | _("inode chunk claims used block, inobt block - agno %d, bno %d, inopb %d\n"), | |
1396 | agno, agbno, mp->m_sb.sb_inopblock); | |
6ceb5693 BN |
1397 | return ++suspect; |
1398 | } | |
1399 | } | |
1400 | } | |
1401 | ||
1402 | /* | |
1403 | * ensure only one avl entry per chunk | |
1404 | */ | |
1ae311d5 | 1405 | find_inode_rec_range(mp, agno, ino, ino + XFS_INODES_PER_CHUNK, |
6ceb5693 BN |
1406 | &first_rec, &last_rec); |
1407 | if (first_rec != NULL) { | |
1408 | /* | |
1409 | * this chunk overlaps with one (or more) | |
1410 | * already in the tree | |
1411 | */ | |
1412 | do_warn( | |
5d1b7f0f | 1413 | _("inode rec for ino %" PRIu64 " (%d/%d) overlaps existing rec (start %d/%d)\n"), |
6ceb5693 BN |
1414 | lino, agno, ino, agno, first_rec->ino_startnum); |
1415 | suspect++; | |
1416 | ||
1417 | /* | |
1418 | * if the 2 chunks start at the same place, | |
1419 | * then we don't have to put this one | |
1420 | * in the uncertain list. go to the next one. | |
1421 | */ | |
1422 | if (first_rec->ino_startnum == ino) | |
1423 | return suspect; | |
1424 | } | |
1425 | ||
6ceb5693 | 1426 | /* |
411c5e55 BF |
1427 | * Import the state of individual inodes into the appropriate in-core |
1428 | * trees, mark them free or used, and get the resulting total and free | |
1429 | * inode counts. | |
c749bd55 | 1430 | */ |
4a7d3215 | 1431 | nfree = ninodes = 0; |
411c5e55 BF |
1432 | suspect = import_single_ino_chunk(agno, INOBT, rp, suspect, &nfree, |
1433 | &ninodes); | |
c749bd55 | 1434 | |
aa6ce1f1 BF |
1435 | if (nfree != freecount) { |
1436 | do_warn( | |
1437 | _("ir_freecount/free mismatch, inode chunk %d/%u, freecount %d nfree %d\n"), | |
1438 | agno, ino, freecount, nfree); | |
6ceb5693 BN |
1439 | } |
1440 | ||
4a7d3215 BF |
1441 | /* verify sparse record formats have a valid inode count */ |
1442 | if (xfs_sb_version_hassparseinodes(&mp->m_sb) && | |
1443 | ninodes != rp->ir_u.sp.ir_count) { | |
1444 | do_warn( | |
1445 | _("invalid inode count, inode chunk %d/%u, count %d ninodes %d\n"), | |
1446 | agno, ino, rp->ir_u.sp.ir_count, ninodes); | |
1447 | } | |
1448 | ||
6ceb5693 BN |
1449 | return suspect; |
1450 | } | |
1451 | ||
1578050f BF |
1452 | static int |
1453 | scan_single_finobt_chunk( | |
1454 | xfs_agnumber_t agno, | |
1455 | xfs_inobt_rec_t *rp, | |
1456 | int suspect) | |
1457 | { | |
1458 | xfs_ino_t lino; | |
1459 | xfs_agino_t ino; | |
1460 | xfs_agblock_t agbno; | |
1461 | int j; | |
1462 | int nfree; | |
4a7d3215 | 1463 | int ninodes; |
1578050f BF |
1464 | int off; |
1465 | int state; | |
3f05ceae | 1466 | ino_tree_node_t *first_rec, *last_rec; |
aa6ce1f1 | 1467 | int freecount; |
22458187 | 1468 | bool skip = false; |
1578050f BF |
1469 | |
1470 | ino = be32_to_cpu(rp->ir_startino); | |
1471 | off = XFS_AGINO_TO_OFFSET(mp, ino); | |
1472 | agbno = XFS_AGINO_TO_AGBNO(mp, ino); | |
1473 | lino = XFS_AGINO_TO_INO(mp, agno, ino); | |
ac9a3f73 | 1474 | freecount = inorec_get_freecount(mp, rp); |
1578050f BF |
1475 | |
1476 | /* | |
22458187 | 1477 | * Verify record alignment, start/end inode numbers, etc. |
1578050f | 1478 | */ |
22458187 BF |
1479 | suspect = verify_single_ino_chunk_align(agno, FINOBT, rp, suspect, |
1480 | &skip); | |
1481 | if (skip) | |
1482 | return suspect; | |
1578050f BF |
1483 | |
1484 | /* | |
1485 | * cross check state of each block containing inodes referenced by the | |
1486 | * finobt against what we have already scanned from the alloc inobt. | |
1487 | */ | |
1488 | if (off == 0 && !suspect) { | |
1489 | for (j = 0; | |
1490 | j < XFS_INODES_PER_CHUNK; | |
1491 | j += mp->m_sb.sb_inopblock) { | |
1492 | agbno = XFS_AGINO_TO_AGBNO(mp, ino + j); | |
1578050f | 1493 | state = get_bmap(agno, agbno); |
3f05ceae BF |
1494 | |
1495 | /* sparse inodes should not refer to inode blocks */ | |
1496 | if (ino_issparse(rp, j)) { | |
1497 | if (state == XR_E_INO) { | |
1498 | do_warn( | |
1499 | _("sparse inode chunk claims inode block, finobt block - agno %d, bno %d, inopb %d\n"), | |
1500 | agno, agbno, mp->m_sb.sb_inopblock); | |
1501 | suspect++; | |
1502 | } | |
1503 | continue; | |
1504 | } | |
1505 | ||
0f94fa4b DW |
1506 | switch (state) { |
1507 | case XR_E_INO: | |
1508 | break; | |
1509 | case XR_E_INO1: /* seen by rmap */ | |
1510 | set_bmap(agno, agbno, XR_E_INO); | |
1511 | break; | |
1512 | case XR_E_UNKNOWN: | |
1578050f BF |
1513 | do_warn( |
1514 | _("inode chunk claims untracked block, finobt block - agno %d, bno %d, inopb %d\n"), | |
1515 | agno, agbno, mp->m_sb.sb_inopblock); | |
1516 | ||
1517 | set_bmap(agno, agbno, XR_E_INO); | |
1518 | suspect++; | |
0f94fa4b DW |
1519 | break; |
1520 | case XR_E_INUSE_FS: | |
1521 | case XR_E_INUSE_FS1: | |
1522 | if (agno == 0 && | |
1523 | ino + j >= first_prealloc_ino && | |
1524 | ino + j < last_prealloc_ino) { | |
1525 | do_warn( | |
1526 | _("inode chunk claims untracked block, finobt block - agno %d, bno %d, inopb %d\n"), | |
1527 | agno, agbno, mp->m_sb.sb_inopblock); | |
1528 | ||
1529 | set_bmap(agno, agbno, XR_E_INO); | |
1530 | suspect++; | |
1531 | break; | |
1532 | } | |
1533 | /* fall through */ | |
1534 | default: | |
1578050f BF |
1535 | do_warn( |
1536 | _("inode chunk claims used block, finobt block - agno %d, bno %d, inopb %d\n"), | |
1537 | agno, agbno, mp->m_sb.sb_inopblock); | |
1538 | return ++suspect; | |
1539 | } | |
1540 | } | |
1541 | } | |
1542 | ||
1543 | /* | |
1544 | * ensure we have an incore entry for each chunk | |
1545 | */ | |
1546 | find_inode_rec_range(mp, agno, ino, ino + XFS_INODES_PER_CHUNK, | |
1547 | &first_rec, &last_rec); | |
1548 | ||
1549 | if (first_rec) { | |
1550 | if (suspect) | |
1551 | return suspect; | |
1552 | ||
1553 | /* | |
1554 | * verify consistency between finobt record and incore state | |
1555 | */ | |
1556 | if (first_rec->ino_startnum != ino) { | |
1557 | do_warn( | |
1558 | _("finobt rec for ino %" PRIu64 " (%d/%u) does not match existing rec (%d/%d)\n"), | |
1559 | lino, agno, ino, agno, first_rec->ino_startnum); | |
1560 | return ++suspect; | |
1561 | } | |
1562 | ||
4a7d3215 | 1563 | nfree = ninodes = 0; |
1578050f BF |
1564 | for (j = 0; j < XFS_INODES_PER_CHUNK; j++) { |
1565 | int isfree = XFS_INOBT_IS_FREE_DISK(rp, j); | |
3f05ceae | 1566 | int issparse = ino_issparse(rp, j); |
1578050f | 1567 | |
4a7d3215 BF |
1568 | if (!issparse) |
1569 | ninodes++; | |
3f05ceae | 1570 | if (isfree && !issparse) |
1578050f BF |
1571 | nfree++; |
1572 | ||
1573 | /* | |
1574 | * inode allocation state should be consistent between | |
1575 | * the inobt and finobt | |
1576 | */ | |
1577 | if (!suspect && | |
1578 | isfree != is_inode_free(first_rec, j)) | |
1579 | suspect++; | |
3f05ceae BF |
1580 | |
1581 | if (!suspect && | |
1582 | issparse != is_inode_sparse(first_rec, j)) | |
1583 | suspect++; | |
1578050f BF |
1584 | } |
1585 | ||
1586 | goto check_freecount; | |
1587 | } | |
1588 | ||
1589 | /* | |
411c5e55 BF |
1590 | * The finobt contains a record that the previous inobt scan never |
1591 | * found. Warn about it and import the inodes into the appropriate | |
1592 | * trees. | |
1593 | * | |
1594 | * Note that this should do the right thing if the previous inobt scan | |
1595 | * had added these inodes to the uncertain tree. If the finobt is not | |
1596 | * suspect, these inodes should supercede the uncertain ones. Otherwise, | |
1597 | * the uncertain tree helpers handle the case where uncertain inodes | |
1598 | * already exist. | |
1578050f BF |
1599 | */ |
1600 | do_warn(_("undiscovered finobt record, ino %" PRIu64 " (%d/%u)\n"), | |
1601 | lino, agno, ino); | |
1602 | ||
4a7d3215 | 1603 | nfree = ninodes = 0; |
411c5e55 BF |
1604 | suspect = import_single_ino_chunk(agno, FINOBT, rp, suspect, &nfree, |
1605 | &ninodes); | |
3f05ceae | 1606 | |
1578050f BF |
1607 | check_freecount: |
1608 | ||
1609 | /* | |
1610 | * Verify that the record freecount matches the actual number of free | |
1611 | * inodes counted in the record. Don't increment 'suspect' here, since | |
1612 | * we have already verified the allocation state of the individual | |
1613 | * inodes against the in-core state. This will have already incremented | |
1614 | * 'suspect' if something is wrong. If suspect hasn't been set at this | |
1615 | * point, these warnings mean that we have a simple freecount | |
1616 | * inconsistency or a stray finobt record (as opposed to a broader tree | |
1617 | * corruption). Issue a warning and continue the scan. The final btree | |
1618 | * reconstruction will correct this naturally. | |
1619 | */ | |
aa6ce1f1 | 1620 | if (nfree != freecount) { |
1578050f BF |
1621 | do_warn( |
1622 | _("finobt ir_freecount/free mismatch, inode chunk %d/%u, freecount %d nfree %d\n"), | |
aa6ce1f1 | 1623 | agno, ino, freecount, nfree); |
1578050f BF |
1624 | } |
1625 | ||
1626 | if (!nfree) { | |
1627 | do_warn( | |
1628 | _("finobt record with no free inodes, inode chunk %d/%u\n"), agno, ino); | |
1629 | } | |
1630 | ||
4a7d3215 BF |
1631 | /* verify sparse record formats have a valid inode count */ |
1632 | if (xfs_sb_version_hassparseinodes(&mp->m_sb) && | |
1633 | ninodes != rp->ir_u.sp.ir_count) { | |
1634 | do_warn( | |
1635 | _("invalid inode count, inode chunk %d/%u, count %d ninodes %d\n"), | |
1636 | agno, ino, rp->ir_u.sp.ir_count, ninodes); | |
1637 | } | |
1638 | ||
1578050f BF |
1639 | return suspect; |
1640 | } | |
6ceb5693 | 1641 | |
2bd0ea18 NS |
1642 | /* |
1643 | * this one walks the inode btrees sucking the info there into | |
1644 | * the incore avl tree. We try and rescue corrupted btree records | |
1645 | * to minimize our chances of losing inodes. Inode info from potentially | |
1646 | * corrupt sources could be bogus so rather than put the info straight | |
1647 | * into the tree, instead we put it on a list and try and verify the | |
1648 | * info in the next phase by examining what's on disk. At that point, | |
1649 | * we'll be able to figure out what's what and stick the corrected info | |
1650 | * into the tree. We do bail out at some point and give up on a subtree | |
1651 | * so as to avoid walking randomly all over the ag. | |
1652 | * | |
1653 | * Note that it's also ok if the free/inuse info wrong, we can correct | |
1654 | * that when we examine the on-disk inode. The important thing is to | |
1655 | * get the start and alignment of the inode chunks right. Those chunks | |
1656 | * that we aren't sure about go into the uncertain list. | |
1657 | */ | |
364a126c | 1658 | static void |
e0607266 | 1659 | scan_inobt( |
b3563c19 | 1660 | struct xfs_btree_block *block, |
2bd0ea18 NS |
1661 | int level, |
1662 | xfs_agblock_t bno, | |
1663 | xfs_agnumber_t agno, | |
1664 | int suspect, | |
364a126c | 1665 | int isroot, |
e0607266 | 1666 | __uint32_t magic, |
364a126c | 1667 | void *priv) |
2bd0ea18 | 1668 | { |
364a126c | 1669 | struct aghdr_cnts *agcnts = priv; |
2bd0ea18 | 1670 | int i; |
2bd0ea18 NS |
1671 | int numrecs; |
1672 | int state; | |
1673 | xfs_inobt_ptr_t *pp; | |
1674 | xfs_inobt_rec_t *rp; | |
2bd0ea18 | 1675 | int hdr_errors; |
aa6ce1f1 | 1676 | int freecount; |
2bd0ea18 | 1677 | |
2bd0ea18 NS |
1678 | hdr_errors = 0; |
1679 | ||
e0607266 | 1680 | if (be32_to_cpu(block->bb_magic) != magic) { |
507f4e33 | 1681 | do_warn(_("bad magic # %#x in inobt block %d/%d\n"), |
5e656dbb | 1682 | be32_to_cpu(block->bb_magic), agno, bno); |
2bd0ea18 NS |
1683 | hdr_errors++; |
1684 | bad_ino_btree = 1; | |
1685 | if (suspect) | |
1686 | return; | |
1687 | } | |
5e656dbb | 1688 | if (be16_to_cpu(block->bb_level) != level) { |
507f4e33 | 1689 | do_warn(_("expected level %d got %d in inobt block %d/%d\n"), |
5e656dbb | 1690 | level, be16_to_cpu(block->bb_level), agno, bno); |
2bd0ea18 NS |
1691 | hdr_errors++; |
1692 | bad_ino_btree = 1; | |
1693 | if (suspect) | |
1694 | return; | |
1695 | } | |
1696 | ||
1697 | /* | |
1698 | * check for btree blocks multiply claimed, any unknown/free state | |
1699 | * is ok in the bitmap block. | |
1700 | */ | |
95650c4d | 1701 | state = get_bmap(agno, bno); |
2bd0ea18 | 1702 | switch (state) { |
0f94fa4b | 1703 | case XR_E_FS_MAP1: /* already been seen by an rmap scan */ |
2bd0ea18 NS |
1704 | case XR_E_UNKNOWN: |
1705 | case XR_E_FREE1: | |
1706 | case XR_E_FREE: | |
95650c4d | 1707 | set_bmap(agno, bno, XR_E_FS_MAP); |
2bd0ea18 NS |
1708 | break; |
1709 | default: | |
95650c4d | 1710 | set_bmap(agno, bno, XR_E_MULT); |
2bd0ea18 | 1711 | do_warn( |
507f4e33 NS |
1712 | _("inode btree block claimed (state %d), agno %d, bno %d, suspect %d\n"), |
1713 | state, agno, bno, suspect); | |
2bd0ea18 NS |
1714 | } |
1715 | ||
5e656dbb | 1716 | numrecs = be16_to_cpu(block->bb_numrecs); |
2bd0ea18 NS |
1717 | |
1718 | /* | |
1719 | * leaf record in btree | |
1720 | */ | |
1721 | if (level == 0) { | |
1722 | /* check for trashed btree block */ | |
1723 | ||
5e656dbb | 1724 | if (numrecs > mp->m_inobt_mxr[0]) { |
2bd0ea18 NS |
1725 | numrecs = mp->m_inobt_mxr[0]; |
1726 | hdr_errors++; | |
1727 | } | |
5e656dbb | 1728 | if (isroot == 0 && numrecs < mp->m_inobt_mnr[0]) { |
2bd0ea18 NS |
1729 | numrecs = mp->m_inobt_mnr[0]; |
1730 | hdr_errors++; | |
1731 | } | |
1732 | ||
1733 | if (hdr_errors) { | |
1734 | bad_ino_btree = 1; | |
507f4e33 | 1735 | do_warn(_("dubious inode btree block header %d/%d\n"), |
2bd0ea18 NS |
1736 | agno, bno); |
1737 | suspect++; | |
1738 | } | |
1739 | ||
b3563c19 | 1740 | rp = XFS_INOBT_REC_ADDR(mp, block, 1); |
2bd0ea18 NS |
1741 | |
1742 | /* | |
1743 | * step through the records, each record points to | |
1744 | * a chunk of inodes. The start of inode chunks should | |
1745 | * be block-aligned. Each inode btree rec should point | |
1746 | * to the start of a block of inodes or the start of a group | |
1747 | * of INODES_PER_CHUNK (64) inodes. off is the offset into | |
1748 | * the block. skip processing of bogus records. | |
1749 | */ | |
48d49157 | 1750 | for (i = 0; i < numrecs; i++) { |
ac9a3f73 | 1751 | freecount = inorec_get_freecount(mp, &rp[i]); |
aa6ce1f1 | 1752 | |
1578050f BF |
1753 | if (magic == XFS_IBT_MAGIC || |
1754 | magic == XFS_IBT_CRC_MAGIC) { | |
85ed3bc3 BF |
1755 | int icount = XFS_INODES_PER_CHUNK; |
1756 | ||
1757 | /* | |
1758 | * ir_count holds the inode count for all | |
1759 | * records on fs' with sparse inode support | |
1760 | */ | |
1761 | if (xfs_sb_version_hassparseinodes(&mp->m_sb)) | |
1762 | icount = rp[i].ir_u.sp.ir_count; | |
1763 | ||
1764 | agcnts->agicount += icount; | |
aa6ce1f1 BF |
1765 | agcnts->agifreecount += freecount; |
1766 | agcnts->ifreecount += freecount; | |
1578050f BF |
1767 | |
1768 | suspect = scan_single_ino_chunk(agno, &rp[i], | |
1769 | suspect); | |
1770 | } else { | |
1771 | /* | |
1772 | * the finobt tracks records with free inodes, | |
1773 | * so only the free inode count is expected to be | |
1774 | * consistent with the agi | |
1775 | */ | |
aa6ce1f1 | 1776 | agcnts->fibtfreecount += freecount; |
48d49157 | 1777 | |
1578050f BF |
1778 | suspect = scan_single_finobt_chunk(agno, &rp[i], |
1779 | suspect); | |
1780 | } | |
48d49157 | 1781 | } |
2bd0ea18 NS |
1782 | |
1783 | if (suspect) | |
1784 | bad_ino_btree = 1; | |
1785 | ||
1786 | return; | |
1787 | } | |
1788 | ||
1789 | /* | |
1790 | * interior record, continue on | |
1791 | */ | |
5e656dbb | 1792 | if (numrecs > mp->m_inobt_mxr[1]) { |
2bd0ea18 NS |
1793 | numrecs = mp->m_inobt_mxr[1]; |
1794 | hdr_errors++; | |
1795 | } | |
5e656dbb | 1796 | if (isroot == 0 && numrecs < mp->m_inobt_mnr[1]) { |
2bd0ea18 NS |
1797 | numrecs = mp->m_inobt_mnr[1]; |
1798 | hdr_errors++; | |
1799 | } | |
1800 | ||
b3563c19 | 1801 | pp = XFS_INOBT_PTR_ADDR(mp, block, 1, mp->m_inobt_mxr[1]); |
2bd0ea18 NS |
1802 | |
1803 | /* | |
1804 | * don't pass bogus tree flag down further if this block | |
1805 | * looked ok. bail out if two levels in a row look bad. | |
1806 | */ | |
1807 | ||
1808 | if (suspect && !hdr_errors) | |
1809 | suspect = 0; | |
1810 | ||
1811 | if (hdr_errors) { | |
1812 | bad_ino_btree = 1; | |
1813 | if (suspect) | |
1814 | return; | |
1815 | else suspect++; | |
1816 | } | |
1817 | ||
1818 | for (i = 0; i < numrecs; i++) { | |
5e656dbb BN |
1819 | if (be32_to_cpu(pp[i]) != 0 && verify_agbno(mp, agno, |
1820 | be32_to_cpu(pp[i]))) | |
1821 | scan_sbtree(be32_to_cpu(pp[i]), level, agno, | |
e0607266 DC |
1822 | suspect, scan_inobt, 0, magic, priv, |
1823 | &xfs_inobt_buf_ops); | |
2bd0ea18 NS |
1824 | } |
1825 | } | |
1826 | ||
364a126c | 1827 | static void |
2bd0ea18 | 1828 | scan_freelist( |
364a126c DC |
1829 | xfs_agf_t *agf, |
1830 | struct aghdr_cnts *agcnts) | |
2bd0ea18 | 1831 | { |
2bd0ea18 | 1832 | xfs_buf_t *agflbuf; |
1d0df0af | 1833 | xfs_agnumber_t agno; |
2bd0ea18 NS |
1834 | xfs_agblock_t bno; |
1835 | int count; | |
1836 | int i; | |
84232448 | 1837 | __be32 *freelist; |
2bd0ea18 | 1838 | |
1d0df0af BN |
1839 | agno = be32_to_cpu(agf->agf_seqno); |
1840 | ||
2bd0ea18 | 1841 | if (XFS_SB_BLOCK(mp) != XFS_AGFL_BLOCK(mp) && |
1d0df0af BN |
1842 | XFS_AGF_BLOCK(mp) != XFS_AGFL_BLOCK(mp) && |
1843 | XFS_AGI_BLOCK(mp) != XFS_AGFL_BLOCK(mp)) | |
0f94fa4b | 1844 | set_bmap(agno, XFS_AGFL_BLOCK(mp), XR_E_INUSE_FS); |
1d0df0af | 1845 | |
5e656dbb | 1846 | if (be32_to_cpu(agf->agf_flcount) == 0) |
2bd0ea18 | 1847 | return; |
1d0df0af BN |
1848 | |
1849 | agflbuf = libxfs_readbuf(mp->m_dev, | |
1850 | XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)), | |
75c8b434 | 1851 | XFS_FSS_TO_BB(mp, 1), 0, &xfs_agfl_buf_ops); |
2bd0ea18 | 1852 | if (!agflbuf) { |
1d0df0af | 1853 | do_abort(_("can't read agfl block for ag %d\n"), agno); |
2bd0ea18 NS |
1854 | return; |
1855 | } | |
66fc04e0 | 1856 | if (agflbuf->b_error == -EFSBADCRC) |
a76a5a71 DC |
1857 | do_warn(_("agfl has bad CRC for ag %d\n"), agno); |
1858 | ||
84232448 | 1859 | freelist = XFS_BUF_TO_AGFL_BNO(mp, agflbuf); |
a529cc7f | 1860 | i = be32_to_cpu(agf->agf_flfirst); |
7e8e3cce ES |
1861 | |
1862 | if (no_modify) { | |
1863 | /* agf values not fixed in verify_set_agf, so recheck */ | |
1864 | if (be32_to_cpu(agf->agf_flfirst) >= XFS_AGFL_SIZE(mp) || | |
1865 | be32_to_cpu(agf->agf_fllast) >= XFS_AGFL_SIZE(mp)) { | |
1866 | do_warn(_("agf %d freelist blocks bad, skipping " | |
1867 | "freelist scan\n"), i); | |
1868 | return; | |
1869 | } | |
b326e067 | 1870 | } |
7e8e3cce | 1871 | |
2bd0ea18 NS |
1872 | count = 0; |
1873 | for (;;) { | |
84232448 | 1874 | bno = be32_to_cpu(freelist[i]); |
1d0df0af | 1875 | if (verify_agbno(mp, agno, bno)) |
95650c4d | 1876 | set_bmap(agno, bno, XR_E_FREE); |
2bd0ea18 | 1877 | else |
507f4e33 | 1878 | do_warn(_("bad agbno %u in agfl, agno %d\n"), |
1d0df0af | 1879 | bno, agno); |
2bd0ea18 | 1880 | count++; |
5e656dbb | 1881 | if (i == be32_to_cpu(agf->agf_fllast)) |
2bd0ea18 | 1882 | break; |
9440d84d | 1883 | if (++i == XFS_AGFL_SIZE(mp)) |
2bd0ea18 NS |
1884 | i = 0; |
1885 | } | |
5e656dbb | 1886 | if (count != be32_to_cpu(agf->agf_flcount)) { |
507f4e33 | 1887 | do_warn(_("freeblk count %d != flcount %d in ag %d\n"), count, |
1d0df0af | 1888 | be32_to_cpu(agf->agf_flcount), agno); |
2bd0ea18 | 1889 | } |
48d49157 | 1890 | |
364a126c | 1891 | agcnts->fdblocks += count; |
48d49157 | 1892 | |
2bd0ea18 NS |
1893 | libxfs_putbuf(agflbuf); |
1894 | } | |
1895 | ||
48d49157 CH |
1896 | static void |
1897 | validate_agf( | |
1898 | struct xfs_agf *agf, | |
364a126c DC |
1899 | xfs_agnumber_t agno, |
1900 | struct aghdr_cnts *agcnts) | |
48d49157 CH |
1901 | { |
1902 | xfs_agblock_t bno; | |
e0607266 | 1903 | __uint32_t magic; |
48d49157 CH |
1904 | |
1905 | bno = be32_to_cpu(agf->agf_roots[XFS_BTNUM_BNO]); | |
1906 | if (bno != 0 && verify_agbno(mp, agno, bno)) { | |
e0607266 DC |
1907 | magic = xfs_sb_version_hascrc(&mp->m_sb) ? XFS_ABTB_CRC_MAGIC |
1908 | : XFS_ABTB_MAGIC; | |
48d49157 | 1909 | scan_sbtree(bno, be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]), |
e0607266 DC |
1910 | agno, 0, scan_allocbt, 1, magic, agcnts, |
1911 | &xfs_allocbt_buf_ops); | |
48d49157 CH |
1912 | } else { |
1913 | do_warn(_("bad agbno %u for btbno root, agno %d\n"), | |
1914 | bno, agno); | |
1915 | } | |
1916 | ||
1917 | bno = be32_to_cpu(agf->agf_roots[XFS_BTNUM_CNT]); | |
1918 | if (bno != 0 && verify_agbno(mp, agno, bno)) { | |
e0607266 DC |
1919 | magic = xfs_sb_version_hascrc(&mp->m_sb) ? XFS_ABTC_CRC_MAGIC |
1920 | : XFS_ABTC_MAGIC; | |
48d49157 | 1921 | scan_sbtree(bno, be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]), |
e0607266 DC |
1922 | agno, 0, scan_allocbt, 1, magic, agcnts, |
1923 | &xfs_allocbt_buf_ops); | |
48d49157 CH |
1924 | } else { |
1925 | do_warn(_("bad agbno %u for btbcnt root, agno %d\n"), | |
1926 | bno, agno); | |
1927 | } | |
1928 | ||
0f94fa4b | 1929 | if (xfs_sb_version_hasrmapbt(&mp->m_sb)) { |
11b9e510 DW |
1930 | struct rmap_priv priv; |
1931 | ||
1932 | memset(&priv.high_key, 0xFF, sizeof(priv.high_key)); | |
1933 | priv.high_key.rm_blockcount = 0; | |
1934 | priv.agcnts = agcnts; | |
65841734 | 1935 | priv.last_rec.rm_owner = XFS_RMAP_OWN_UNKNOWN; |
11b9e510 | 1936 | priv.nr_blocks = 0; |
0f94fa4b DW |
1937 | bno = be32_to_cpu(agf->agf_roots[XFS_BTNUM_RMAP]); |
1938 | if (bno != 0 && verify_agbno(mp, agno, bno)) { | |
1939 | scan_sbtree(bno, | |
1940 | be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]), | |
1941 | agno, 0, scan_rmapbt, 1, XFS_RMAP_CRC_MAGIC, | |
11b9e510 DW |
1942 | &priv, &xfs_rmapbt_buf_ops); |
1943 | if (be32_to_cpu(agf->agf_rmap_blocks) != priv.nr_blocks) | |
1944 | do_warn(_("bad rmapbt block count %u, saw %u\n"), | |
1945 | priv.nr_blocks, | |
1946 | be32_to_cpu(agf->agf_rmap_blocks)); | |
0f94fa4b DW |
1947 | } else { |
1948 | do_warn(_("bad agbno %u for rmapbt root, agno %d\n"), | |
1949 | bno, agno); | |
11b9e510 | 1950 | rmap_avoid_check(); |
0f94fa4b DW |
1951 | } |
1952 | } | |
1953 | ||
364a126c | 1954 | if (be32_to_cpu(agf->agf_freeblks) != agcnts->agffreeblks) { |
48d49157 | 1955 | do_warn(_("agf_freeblks %u, counted %u in ag %u\n"), |
364a126c | 1956 | be32_to_cpu(agf->agf_freeblks), agcnts->agffreeblks, agno); |
48d49157 CH |
1957 | } |
1958 | ||
364a126c | 1959 | if (be32_to_cpu(agf->agf_longest) != agcnts->agflongest) { |
48d49157 | 1960 | do_warn(_("agf_longest %u, counted %u in ag %u\n"), |
364a126c | 1961 | be32_to_cpu(agf->agf_longest), agcnts->agflongest, agno); |
48d49157 CH |
1962 | } |
1963 | ||
1964 | if (xfs_sb_version_haslazysbcount(&mp->m_sb) && | |
364a126c | 1965 | be32_to_cpu(agf->agf_btreeblks) != agcnts->agfbtreeblks) { |
5d1b7f0f | 1966 | do_warn(_("agf_btreeblks %u, counted %" PRIu64 " in ag %u\n"), |
364a126c | 1967 | be32_to_cpu(agf->agf_btreeblks), agcnts->agfbtreeblks, agno); |
48d49157 | 1968 | } |
0f94fa4b | 1969 | |
48d49157 CH |
1970 | } |
1971 | ||
1972 | static void | |
1973 | validate_agi( | |
1974 | struct xfs_agi *agi, | |
364a126c DC |
1975 | xfs_agnumber_t agno, |
1976 | struct aghdr_cnts *agcnts) | |
48d49157 CH |
1977 | { |
1978 | xfs_agblock_t bno; | |
1979 | int i; | |
e0607266 | 1980 | __uint32_t magic; |
48d49157 CH |
1981 | |
1982 | bno = be32_to_cpu(agi->agi_root); | |
1983 | if (bno != 0 && verify_agbno(mp, agno, bno)) { | |
e0607266 DC |
1984 | magic = xfs_sb_version_hascrc(&mp->m_sb) ? XFS_IBT_CRC_MAGIC |
1985 | : XFS_IBT_MAGIC; | |
48d49157 | 1986 | scan_sbtree(bno, be32_to_cpu(agi->agi_level), |
e0607266 DC |
1987 | agno, 0, scan_inobt, 1, magic, agcnts, |
1988 | &xfs_inobt_buf_ops); | |
48d49157 CH |
1989 | } else { |
1990 | do_warn(_("bad agbno %u for inobt root, agno %d\n"), | |
1991 | be32_to_cpu(agi->agi_root), agno); | |
1992 | } | |
1993 | ||
1578050f BF |
1994 | if (xfs_sb_version_hasfinobt(&mp->m_sb)) { |
1995 | bno = be32_to_cpu(agi->agi_free_root); | |
1996 | if (bno != 0 && verify_agbno(mp, agno, bno)) { | |
1997 | magic = xfs_sb_version_hascrc(&mp->m_sb) ? | |
1998 | XFS_FIBT_CRC_MAGIC : XFS_FIBT_MAGIC; | |
1999 | scan_sbtree(bno, be32_to_cpu(agi->agi_free_level), | |
2000 | agno, 0, scan_inobt, 1, magic, agcnts, | |
2001 | &xfs_inobt_buf_ops); | |
2002 | } else { | |
2003 | do_warn(_("bad agbno %u for finobt root, agno %d\n"), | |
2004 | be32_to_cpu(agi->agi_free_root), agno); | |
2005 | } | |
2006 | } | |
2007 | ||
364a126c | 2008 | if (be32_to_cpu(agi->agi_count) != agcnts->agicount) { |
48d49157 | 2009 | do_warn(_("agi_count %u, counted %u in ag %u\n"), |
364a126c | 2010 | be32_to_cpu(agi->agi_count), agcnts->agicount, agno); |
48d49157 CH |
2011 | } |
2012 | ||
364a126c | 2013 | if (be32_to_cpu(agi->agi_freecount) != agcnts->agifreecount) { |
48d49157 | 2014 | do_warn(_("agi_freecount %u, counted %u in ag %u\n"), |
364a126c | 2015 | be32_to_cpu(agi->agi_freecount), agcnts->agifreecount, agno); |
48d49157 CH |
2016 | } |
2017 | ||
1578050f BF |
2018 | if (xfs_sb_version_hasfinobt(&mp->m_sb) && |
2019 | be32_to_cpu(agi->agi_freecount) != agcnts->fibtfreecount) { | |
2020 | do_warn(_("agi_freecount %u, counted %u in ag %u finobt\n"), | |
2021 | be32_to_cpu(agi->agi_freecount), agcnts->fibtfreecount, | |
2022 | agno); | |
2023 | } | |
2024 | ||
48d49157 CH |
2025 | for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++) { |
2026 | xfs_agino_t agino = be32_to_cpu(agi->agi_unlinked[i]); | |
2027 | ||
2028 | if (agino != NULLAGINO) { | |
2029 | do_warn( | |
5d1b7f0f | 2030 | _("agi unlinked bucket %d is %u in ag %u (inode=%" PRIu64 ")\n"), |
48d49157 CH |
2031 | i, agino, agno, |
2032 | XFS_AGINO_TO_INO(mp, agno, agino)); | |
2033 | } | |
2034 | } | |
2035 | } | |
2036 | ||
48d49157 CH |
2037 | /* |
2038 | * Scan an AG for obvious corruption. | |
48d49157 | 2039 | */ |
364a126c | 2040 | static void |
2bd0ea18 | 2041 | scan_ag( |
364a126c DC |
2042 | work_queue_t *wq, |
2043 | xfs_agnumber_t agno, | |
2044 | void *arg) | |
2bd0ea18 | 2045 | { |
364a126c | 2046 | struct aghdr_cnts *agcnts = arg; |
8f657d7e DC |
2047 | struct xfs_agf *agf; |
2048 | struct xfs_buf *agfbuf = NULL; | |
48d49157 | 2049 | int agf_dirty = 0; |
8f657d7e DC |
2050 | struct xfs_agi *agi; |
2051 | struct xfs_buf *agibuf = NULL; | |
48d49157 | 2052 | int agi_dirty = 0; |
8f657d7e DC |
2053 | struct xfs_sb *sb = NULL; |
2054 | struct xfs_buf *sbbuf = NULL; | |
48d49157 | 2055 | int sb_dirty = 0; |
2bd0ea18 | 2056 | int status; |
8f657d7e | 2057 | char *objname = NULL; |
2bd0ea18 | 2058 | |
8bc43a39 | 2059 | sb = (struct xfs_sb *)calloc(BBTOB(XFS_FSS_TO_BB(mp, 1)), 1); |
507f4e33 NS |
2060 | if (!sb) { |
2061 | do_error(_("can't allocate memory for superblock\n")); | |
507f4e33 | 2062 | return; |
dfc130f3 | 2063 | } |
8f657d7e DC |
2064 | |
2065 | sbbuf = libxfs_readbuf(mp->m_dev, XFS_AG_DADDR(mp, agno, XFS_SB_DADDR), | |
2066 | XFS_FSS_TO_BB(mp, 1), 0, &xfs_sb_buf_ops); | |
2067 | if (!sbbuf) { | |
2068 | objname = _("root superblock"); | |
2069 | goto out_free_sb; | |
2070 | } | |
5e656dbb | 2071 | libxfs_sb_from_disk(sb, XFS_BUF_TO_SBP(sbbuf)); |
2bd0ea18 NS |
2072 | |
2073 | agfbuf = libxfs_readbuf(mp->m_dev, | |
9440d84d | 2074 | XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)), |
75c8b434 | 2075 | XFS_FSS_TO_BB(mp, 1), 0, &xfs_agf_buf_ops); |
2bd0ea18 | 2076 | if (!agfbuf) { |
8f657d7e DC |
2077 | objname = _("agf block"); |
2078 | goto out_free_sbbuf; | |
2bd0ea18 NS |
2079 | } |
2080 | agf = XFS_BUF_TO_AGF(agfbuf); | |
2081 | ||
2082 | agibuf = libxfs_readbuf(mp->m_dev, | |
9440d84d | 2083 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
75c8b434 | 2084 | XFS_FSS_TO_BB(mp, 1), 0, &xfs_agi_buf_ops); |
2bd0ea18 | 2085 | if (!agibuf) { |
8f657d7e DC |
2086 | objname = _("agi block"); |
2087 | goto out_free_agfbuf; | |
2bd0ea18 NS |
2088 | } |
2089 | agi = XFS_BUF_TO_AGI(agibuf); | |
2090 | ||
2091 | /* fix up bad ag headers */ | |
2092 | ||
2093 | status = verify_set_agheader(mp, sbbuf, sb, agf, agi, agno); | |
2094 | ||
2095 | if (status & XR_AG_SB_SEC) { | |
2096 | if (!no_modify) | |
2097 | sb_dirty = 1; | |
2098 | /* | |
2099 | * clear bad sector bit because we don't want | |
2100 | * to skip further processing. we just want to | |
2101 | * ensure that we write out the modified sb buffer. | |
2102 | */ | |
2103 | status &= ~XR_AG_SB_SEC; | |
2104 | } | |
2105 | if (status & XR_AG_SB) { | |
507f4e33 NS |
2106 | if (!no_modify) { |
2107 | do_warn(_("reset bad sb for ag %d\n"), agno); | |
2bd0ea18 | 2108 | sb_dirty = 1; |
507f4e33 NS |
2109 | } else { |
2110 | do_warn(_("would reset bad sb for ag %d\n"), agno); | |
2111 | } | |
2bd0ea18 NS |
2112 | } |
2113 | if (status & XR_AG_AGF) { | |
507f4e33 NS |
2114 | if (!no_modify) { |
2115 | do_warn(_("reset bad agf for ag %d\n"), agno); | |
2bd0ea18 | 2116 | agf_dirty = 1; |
507f4e33 NS |
2117 | } else { |
2118 | do_warn(_("would reset bad agf for ag %d\n"), agno); | |
2119 | } | |
2bd0ea18 NS |
2120 | } |
2121 | if (status & XR_AG_AGI) { | |
507f4e33 NS |
2122 | if (!no_modify) { |
2123 | do_warn(_("reset bad agi for ag %d\n"), agno); | |
2bd0ea18 | 2124 | agi_dirty = 1; |
507f4e33 NS |
2125 | } else { |
2126 | do_warn(_("would reset bad agi for ag %d\n"), agno); | |
2127 | } | |
2bd0ea18 NS |
2128 | } |
2129 | ||
2130 | if (status && no_modify) { | |
507f4e33 NS |
2131 | do_warn(_("bad uncorrected agheader %d, skipping ag...\n"), |
2132 | agno); | |
8f657d7e | 2133 | goto out_free_agibuf; |
2bd0ea18 NS |
2134 | } |
2135 | ||
364a126c | 2136 | scan_freelist(agf, agcnts); |
2bd0ea18 | 2137 | |
364a126c DC |
2138 | validate_agf(agf, agno, agcnts); |
2139 | validate_agi(agi, agno, agcnts); | |
2bd0ea18 | 2140 | |
27527004 | 2141 | ASSERT(agi_dirty == 0 || (agi_dirty && !no_modify)); |
8f657d7e DC |
2142 | ASSERT(agf_dirty == 0 || (agf_dirty && !no_modify)); |
2143 | ASSERT(sb_dirty == 0 || (sb_dirty && !no_modify)); | |
2144 | ||
2145 | /* | |
2146 | * Only pay attention to CRC/verifier errors if we can correct them. | |
1fe7b0b0 DC |
2147 | * Note that we can get uncorrected EFSCORRUPTED errors here because |
2148 | * the verifier will flag on out of range values that we can't correct | |
2149 | * until phase 5 when we have all the information necessary to rebuild | |
2150 | * the freespace/inode btrees. We can correct bad CRC errors | |
2151 | * immediately, though. | |
8f657d7e DC |
2152 | */ |
2153 | if (!no_modify) { | |
66fc04e0 DW |
2154 | agi_dirty += (agibuf->b_error == -EFSBADCRC); |
2155 | agf_dirty += (agfbuf->b_error == -EFSBADCRC); | |
2156 | sb_dirty += (sbbuf->b_error == -EFSBADCRC); | |
8f657d7e | 2157 | } |
2bd0ea18 NS |
2158 | |
2159 | if (agi_dirty && !no_modify) | |
2160 | libxfs_writebuf(agibuf, 0); | |
2161 | else | |
2162 | libxfs_putbuf(agibuf); | |
2163 | ||
2bd0ea18 NS |
2164 | if (agf_dirty && !no_modify) |
2165 | libxfs_writebuf(agfbuf, 0); | |
2166 | else | |
2167 | libxfs_putbuf(agfbuf); | |
2168 | ||
2bd0ea18 | 2169 | if (sb_dirty && !no_modify) { |
e8cb94ee BN |
2170 | if (agno == 0) |
2171 | memcpy(&mp->m_sb, sb, sizeof(xfs_sb_t)); | |
19ebedcf | 2172 | libxfs_sb_to_disk(XFS_BUF_TO_SBP(sbbuf), sb); |
2bd0ea18 | 2173 | libxfs_writebuf(sbbuf, 0); |
dfc130f3 | 2174 | } else |
2bd0ea18 | 2175 | libxfs_putbuf(sbbuf); |
dfc130f3 | 2176 | free(sb); |
06fbdda9 | 2177 | PROG_RPT_INC(prog_rpt_done[agno], 1); |
364a126c DC |
2178 | |
2179 | #ifdef XR_INODE_TRACE | |
2180 | print_inode_list(i); | |
2181 | #endif | |
2182 | return; | |
8f657d7e DC |
2183 | |
2184 | out_free_agibuf: | |
2185 | libxfs_putbuf(agibuf); | |
2186 | out_free_agfbuf: | |
2187 | libxfs_putbuf(agfbuf); | |
2188 | out_free_sbbuf: | |
2189 | libxfs_putbuf(sbbuf); | |
2190 | out_free_sb: | |
2191 | free(sb); | |
2192 | ||
2193 | if (objname) | |
2194 | do_error(_("can't get %s for ag %d\n"), objname, agno); | |
2bd0ea18 | 2195 | } |
364a126c DC |
2196 | |
2197 | #define SCAN_THREADS 32 | |
2198 | ||
2199 | void | |
2200 | scan_ags( | |
2201 | struct xfs_mount *mp, | |
2202 | int scan_threads) | |
2203 | { | |
2204 | struct aghdr_cnts *agcnts; | |
2205 | __uint64_t fdblocks = 0; | |
2206 | __uint64_t icount = 0; | |
2207 | __uint64_t ifreecount = 0; | |
0f94fa4b | 2208 | __uint64_t usedblocks = 0; |
364a126c DC |
2209 | xfs_agnumber_t i; |
2210 | work_queue_t wq; | |
2211 | ||
2212 | agcnts = malloc(mp->m_sb.sb_agcount * sizeof(*agcnts)); | |
2213 | if (!agcnts) { | |
2214 | do_abort(_("no memory for ag header counts\n")); | |
2215 | return; | |
2216 | } | |
2217 | memset(agcnts, 0, mp->m_sb.sb_agcount * sizeof(*agcnts)); | |
2218 | ||
f756f80c | 2219 | create_work_queue(&wq, mp, scan_threads); |
364a126c DC |
2220 | |
2221 | for (i = 0; i < mp->m_sb.sb_agcount; i++) | |
2222 | queue_work(&wq, scan_ag, i, &agcnts[i]); | |
2223 | ||
2224 | destroy_work_queue(&wq); | |
2225 | ||
2226 | /* tally up the counts */ | |
2227 | for (i = 0; i < mp->m_sb.sb_agcount; i++) { | |
2228 | fdblocks += agcnts[i].fdblocks; | |
37246b95 | 2229 | icount += agcnts[i].agicount; |
364a126c | 2230 | ifreecount += agcnts[i].ifreecount; |
0f94fa4b | 2231 | usedblocks += agcnts[i].usedblocks; |
364a126c DC |
2232 | } |
2233 | ||
aba29588 ES |
2234 | free(agcnts); |
2235 | ||
364a126c DC |
2236 | /* |
2237 | * Validate that our manual counts match the superblock. | |
2238 | */ | |
2239 | if (mp->m_sb.sb_icount != icount) { | |
5d1b7f0f | 2240 | do_warn(_("sb_icount %" PRIu64 ", counted %" PRIu64 "\n"), |
364a126c DC |
2241 | mp->m_sb.sb_icount, icount); |
2242 | } | |
2243 | ||
2244 | if (mp->m_sb.sb_ifree != ifreecount) { | |
5d1b7f0f | 2245 | do_warn(_("sb_ifree %" PRIu64 ", counted %" PRIu64 "\n"), |
364a126c DC |
2246 | mp->m_sb.sb_ifree, ifreecount); |
2247 | } | |
2248 | ||
2249 | if (mp->m_sb.sb_fdblocks != fdblocks) { | |
5d1b7f0f | 2250 | do_warn(_("sb_fdblocks %" PRIu64 ", counted %" PRIu64 "\n"), |
364a126c DC |
2251 | mp->m_sb.sb_fdblocks, fdblocks); |
2252 | } | |
0f94fa4b DW |
2253 | |
2254 | if (usedblocks && | |
2255 | usedblocks != mp->m_sb.sb_dblocks - fdblocks) { | |
2256 | do_warn(_("used blocks %" PRIu64 ", counted %" PRIu64 "\n"), | |
2257 | mp->m_sb.sb_dblocks - fdblocks, usedblocks); | |
2258 | } | |
364a126c | 2259 | } |