Commit | Line | Data |
---|---|---|
1fd7115e DC |
1 | /* |
2 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. | |
3 | * All Rights Reserved. | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or | |
6 | * modify it under the terms of the GNU General Public License as | |
7 | * published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope that it would be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
17 | */ | |
18 | #include "xfs.h" | |
19 | #include "xfs_fs.h" | |
20 | #include "xfs_format.h" | |
70a9883c | 21 | #include "xfs_shared.h" |
1fd7115e DC |
22 | #include "xfs_log.h" |
23 | #include "xfs_trans.h" | |
24 | #include "xfs_sb.h" | |
25 | #include "xfs_ag.h" | |
26 | #include "xfs_mount.h" | |
27 | #include "xfs_bmap_btree.h" | |
28 | #include "xfs_ialloc_btree.h" | |
29 | #include "xfs_dinode.h" | |
30 | #include "xfs_inode.h" | |
31 | #include "xfs_error.h" | |
32 | #include "xfs_cksum.h" | |
33 | #include "xfs_icache.h" | |
34 | #include "xfs_ialloc.h" | |
35 | ||
36 | /* | |
37 | * Check that none of the inode's in the buffer have a next | |
38 | * unlinked field of 0. | |
39 | */ | |
40 | #if defined(DEBUG) | |
41 | void | |
42 | xfs_inobp_check( | |
43 | xfs_mount_t *mp, | |
44 | xfs_buf_t *bp) | |
45 | { | |
46 | int i; | |
47 | int j; | |
48 | xfs_dinode_t *dip; | |
49 | ||
50 | j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog; | |
51 | ||
52 | for (i = 0; i < j; i++) { | |
53 | dip = (xfs_dinode_t *)xfs_buf_offset(bp, | |
54 | i * mp->m_sb.sb_inodesize); | |
55 | if (!dip->di_next_unlinked) { | |
56 | xfs_alert(mp, | |
74ffa796 DC |
57 | "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.", |
58 | i, (long long)bp->b_bn); | |
1fd7115e DC |
59 | } |
60 | } | |
61 | } | |
62 | #endif | |
63 | ||
d8914002 DC |
64 | /* |
65 | * If we are doing readahead on an inode buffer, we might be in log recovery | |
66 | * reading an inode allocation buffer that hasn't yet been replayed, and hence | |
67 | * has not had the inode cores stamped into it. Hence for readahead, the buffer | |
68 | * may be potentially invalid. | |
69 | * | |
70 | * If the readahead buffer is invalid, we don't want to mark it with an error, | |
71 | * but we do want to clear the DONE status of the buffer so that a followup read | |
72 | * will re-read it from disk. This will ensure that we don't get an unnecessary | |
73 | * warnings during log recovery and we don't get unnecssary panics on debug | |
74 | * kernels. | |
75 | */ | |
1fd7115e DC |
76 | static void |
77 | xfs_inode_buf_verify( | |
d8914002 DC |
78 | struct xfs_buf *bp, |
79 | bool readahead) | |
1fd7115e DC |
80 | { |
81 | struct xfs_mount *mp = bp->b_target->bt_mount; | |
82 | int i; | |
83 | int ni; | |
84 | ||
85 | /* | |
86 | * Validate the magic number and version of every inode in the buffer | |
87 | */ | |
88 | ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock; | |
89 | for (i = 0; i < ni; i++) { | |
90 | int di_ok; | |
91 | xfs_dinode_t *dip; | |
92 | ||
93 | dip = (struct xfs_dinode *)xfs_buf_offset(bp, | |
94 | (i << mp->m_sb.sb_inodelog)); | |
95 | di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) && | |
96 | XFS_DINODE_GOOD_VERSION(dip->di_version); | |
97 | if (unlikely(XFS_TEST_ERROR(!di_ok, mp, | |
98 | XFS_ERRTAG_ITOBP_INOTOBP, | |
99 | XFS_RANDOM_ITOBP_INOTOBP))) { | |
d8914002 DC |
100 | if (readahead) { |
101 | bp->b_flags &= ~XBF_DONE; | |
102 | return; | |
103 | } | |
104 | ||
1fd7115e DC |
105 | xfs_buf_ioerror(bp, EFSCORRUPTED); |
106 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_HIGH, | |
107 | mp, dip); | |
108 | #ifdef DEBUG | |
74ffa796 | 109 | xfs_alert(mp, |
1fd7115e DC |
110 | "bad inode magic/vsn daddr %lld #%d (magic=%x)", |
111 | (unsigned long long)bp->b_bn, i, | |
112 | be16_to_cpu(dip->di_magic)); | |
1fd7115e DC |
113 | #endif |
114 | } | |
115 | } | |
116 | xfs_inobp_check(mp, bp); | |
117 | } | |
118 | ||
119 | ||
120 | static void | |
121 | xfs_inode_buf_read_verify( | |
122 | struct xfs_buf *bp) | |
123 | { | |
d8914002 DC |
124 | xfs_inode_buf_verify(bp, false); |
125 | } | |
126 | ||
127 | static void | |
128 | xfs_inode_buf_readahead_verify( | |
129 | struct xfs_buf *bp) | |
130 | { | |
131 | xfs_inode_buf_verify(bp, true); | |
1fd7115e DC |
132 | } |
133 | ||
134 | static void | |
135 | xfs_inode_buf_write_verify( | |
136 | struct xfs_buf *bp) | |
137 | { | |
d8914002 | 138 | xfs_inode_buf_verify(bp, false); |
1fd7115e DC |
139 | } |
140 | ||
141 | const struct xfs_buf_ops xfs_inode_buf_ops = { | |
142 | .verify_read = xfs_inode_buf_read_verify, | |
143 | .verify_write = xfs_inode_buf_write_verify, | |
144 | }; | |
145 | ||
d8914002 DC |
146 | const struct xfs_buf_ops xfs_inode_buf_ra_ops = { |
147 | .verify_read = xfs_inode_buf_readahead_verify, | |
148 | .verify_write = xfs_inode_buf_write_verify, | |
149 | }; | |
150 | ||
1fd7115e DC |
151 | |
152 | /* | |
153 | * This routine is called to map an inode to the buffer containing the on-disk | |
154 | * version of the inode. It returns a pointer to the buffer containing the | |
155 | * on-disk inode in the bpp parameter, and in the dipp parameter it returns a | |
156 | * pointer to the on-disk inode within that buffer. | |
157 | * | |
158 | * If a non-zero error is returned, then the contents of bpp and dipp are | |
159 | * undefined. | |
160 | */ | |
161 | int | |
162 | xfs_imap_to_bp( | |
163 | struct xfs_mount *mp, | |
164 | struct xfs_trans *tp, | |
165 | struct xfs_imap *imap, | |
166 | struct xfs_dinode **dipp, | |
167 | struct xfs_buf **bpp, | |
168 | uint buf_flags, | |
169 | uint iget_flags) | |
170 | { | |
171 | struct xfs_buf *bp; | |
172 | int error; | |
173 | ||
174 | buf_flags |= XBF_UNMAPPED; | |
175 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno, | |
176 | (int)imap->im_len, buf_flags, &bp, | |
177 | &xfs_inode_buf_ops); | |
178 | if (error) { | |
179 | if (error == EAGAIN) { | |
180 | ASSERT(buf_flags & XBF_TRYLOCK); | |
181 | return error; | |
182 | } | |
183 | ||
184 | if (error == EFSCORRUPTED && | |
185 | (iget_flags & XFS_IGET_UNTRUSTED)) | |
186 | return XFS_ERROR(EINVAL); | |
187 | ||
188 | xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.", | |
189 | __func__, error); | |
190 | return error; | |
191 | } | |
192 | ||
193 | *bpp = bp; | |
194 | *dipp = (struct xfs_dinode *)xfs_buf_offset(bp, imap->im_boffset); | |
195 | return 0; | |
196 | } | |
197 | ||
638f4416 | 198 | void |
1fd7115e DC |
199 | xfs_dinode_from_disk( |
200 | xfs_icdinode_t *to, | |
201 | xfs_dinode_t *from) | |
202 | { | |
203 | to->di_magic = be16_to_cpu(from->di_magic); | |
204 | to->di_mode = be16_to_cpu(from->di_mode); | |
205 | to->di_version = from ->di_version; | |
206 | to->di_format = from->di_format; | |
207 | to->di_onlink = be16_to_cpu(from->di_onlink); | |
208 | to->di_uid = be32_to_cpu(from->di_uid); | |
209 | to->di_gid = be32_to_cpu(from->di_gid); | |
210 | to->di_nlink = be32_to_cpu(from->di_nlink); | |
211 | to->di_projid_lo = be16_to_cpu(from->di_projid_lo); | |
212 | to->di_projid_hi = be16_to_cpu(from->di_projid_hi); | |
213 | memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); | |
214 | to->di_flushiter = be16_to_cpu(from->di_flushiter); | |
215 | to->di_atime.t_sec = be32_to_cpu(from->di_atime.t_sec); | |
216 | to->di_atime.t_nsec = be32_to_cpu(from->di_atime.t_nsec); | |
217 | to->di_mtime.t_sec = be32_to_cpu(from->di_mtime.t_sec); | |
218 | to->di_mtime.t_nsec = be32_to_cpu(from->di_mtime.t_nsec); | |
219 | to->di_ctime.t_sec = be32_to_cpu(from->di_ctime.t_sec); | |
220 | to->di_ctime.t_nsec = be32_to_cpu(from->di_ctime.t_nsec); | |
221 | to->di_size = be64_to_cpu(from->di_size); | |
222 | to->di_nblocks = be64_to_cpu(from->di_nblocks); | |
223 | to->di_extsize = be32_to_cpu(from->di_extsize); | |
224 | to->di_nextents = be32_to_cpu(from->di_nextents); | |
225 | to->di_anextents = be16_to_cpu(from->di_anextents); | |
226 | to->di_forkoff = from->di_forkoff; | |
227 | to->di_aformat = from->di_aformat; | |
228 | to->di_dmevmask = be32_to_cpu(from->di_dmevmask); | |
229 | to->di_dmstate = be16_to_cpu(from->di_dmstate); | |
230 | to->di_flags = be16_to_cpu(from->di_flags); | |
231 | to->di_gen = be32_to_cpu(from->di_gen); | |
232 | ||
233 | if (to->di_version == 3) { | |
234 | to->di_changecount = be64_to_cpu(from->di_changecount); | |
235 | to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec); | |
236 | to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec); | |
237 | to->di_flags2 = be64_to_cpu(from->di_flags2); | |
238 | to->di_ino = be64_to_cpu(from->di_ino); | |
239 | to->di_lsn = be64_to_cpu(from->di_lsn); | |
240 | memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2)); | |
241 | uuid_copy(&to->di_uuid, &from->di_uuid); | |
242 | } | |
243 | } | |
244 | ||
245 | void | |
246 | xfs_dinode_to_disk( | |
247 | xfs_dinode_t *to, | |
248 | xfs_icdinode_t *from) | |
249 | { | |
250 | to->di_magic = cpu_to_be16(from->di_magic); | |
251 | to->di_mode = cpu_to_be16(from->di_mode); | |
252 | to->di_version = from ->di_version; | |
253 | to->di_format = from->di_format; | |
254 | to->di_onlink = cpu_to_be16(from->di_onlink); | |
255 | to->di_uid = cpu_to_be32(from->di_uid); | |
256 | to->di_gid = cpu_to_be32(from->di_gid); | |
257 | to->di_nlink = cpu_to_be32(from->di_nlink); | |
258 | to->di_projid_lo = cpu_to_be16(from->di_projid_lo); | |
259 | to->di_projid_hi = cpu_to_be16(from->di_projid_hi); | |
260 | memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); | |
261 | to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec); | |
262 | to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec); | |
263 | to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec); | |
264 | to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec); | |
265 | to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec); | |
266 | to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec); | |
267 | to->di_size = cpu_to_be64(from->di_size); | |
268 | to->di_nblocks = cpu_to_be64(from->di_nblocks); | |
269 | to->di_extsize = cpu_to_be32(from->di_extsize); | |
270 | to->di_nextents = cpu_to_be32(from->di_nextents); | |
271 | to->di_anextents = cpu_to_be16(from->di_anextents); | |
272 | to->di_forkoff = from->di_forkoff; | |
273 | to->di_aformat = from->di_aformat; | |
274 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); | |
275 | to->di_dmstate = cpu_to_be16(from->di_dmstate); | |
276 | to->di_flags = cpu_to_be16(from->di_flags); | |
277 | to->di_gen = cpu_to_be32(from->di_gen); | |
278 | ||
279 | if (from->di_version == 3) { | |
280 | to->di_changecount = cpu_to_be64(from->di_changecount); | |
281 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec); | |
282 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec); | |
283 | to->di_flags2 = cpu_to_be64(from->di_flags2); | |
284 | to->di_ino = cpu_to_be64(from->di_ino); | |
285 | to->di_lsn = cpu_to_be64(from->di_lsn); | |
286 | memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2)); | |
287 | uuid_copy(&to->di_uuid, &from->di_uuid); | |
288 | to->di_flushiter = 0; | |
289 | } else { | |
290 | to->di_flushiter = cpu_to_be16(from->di_flushiter); | |
291 | } | |
292 | } | |
293 | ||
294 | static bool | |
295 | xfs_dinode_verify( | |
296 | struct xfs_mount *mp, | |
297 | struct xfs_inode *ip, | |
298 | struct xfs_dinode *dip) | |
299 | { | |
300 | if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC)) | |
301 | return false; | |
302 | ||
303 | /* only version 3 or greater inodes are extensively verified here */ | |
304 | if (dip->di_version < 3) | |
305 | return true; | |
306 | ||
307 | if (!xfs_sb_version_hascrc(&mp->m_sb)) | |
308 | return false; | |
309 | if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize, | |
310 | offsetof(struct xfs_dinode, di_crc))) | |
311 | return false; | |
312 | if (be64_to_cpu(dip->di_ino) != ip->i_ino) | |
313 | return false; | |
314 | if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_uuid)) | |
315 | return false; | |
316 | return true; | |
317 | } | |
318 | ||
319 | void | |
320 | xfs_dinode_calc_crc( | |
321 | struct xfs_mount *mp, | |
322 | struct xfs_dinode *dip) | |
323 | { | |
324 | __uint32_t crc; | |
325 | ||
326 | if (dip->di_version < 3) | |
327 | return; | |
328 | ||
329 | ASSERT(xfs_sb_version_hascrc(&mp->m_sb)); | |
330 | crc = xfs_start_cksum((char *)dip, mp->m_sb.sb_inodesize, | |
331 | offsetof(struct xfs_dinode, di_crc)); | |
332 | dip->di_crc = xfs_end_cksum(crc); | |
333 | } | |
334 | ||
335 | /* | |
336 | * Read the disk inode attributes into the in-core inode structure. | |
337 | * | |
338 | * For version 5 superblocks, if we are initialising a new inode and we are not | |
339 | * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new | |
340 | * inode core with a random generation number. If we are keeping inodes around, | |
341 | * we need to read the inode cluster to get the existing generation number off | |
342 | * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode | |
343 | * format) then log recovery is dependent on the di_flushiter field being | |
344 | * initialised from the current on-disk value and hence we must also read the | |
345 | * inode off disk. | |
346 | */ | |
347 | int | |
348 | xfs_iread( | |
349 | xfs_mount_t *mp, | |
350 | xfs_trans_t *tp, | |
351 | xfs_inode_t *ip, | |
352 | uint iget_flags) | |
353 | { | |
354 | xfs_buf_t *bp; | |
355 | xfs_dinode_t *dip; | |
356 | int error; | |
357 | ||
358 | /* | |
359 | * Fill in the location information in the in-core inode. | |
360 | */ | |
361 | error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags); | |
362 | if (error) | |
363 | return error; | |
364 | ||
365 | /* shortcut IO on inode allocation if possible */ | |
366 | if ((iget_flags & XFS_IGET_CREATE) && | |
367 | xfs_sb_version_hascrc(&mp->m_sb) && | |
368 | !(mp->m_flags & XFS_MOUNT_IKEEP)) { | |
369 | /* initialise the on-disk inode core */ | |
370 | memset(&ip->i_d, 0, sizeof(ip->i_d)); | |
371 | ip->i_d.di_magic = XFS_DINODE_MAGIC; | |
372 | ip->i_d.di_gen = prandom_u32(); | |
373 | if (xfs_sb_version_hascrc(&mp->m_sb)) { | |
374 | ip->i_d.di_version = 3; | |
375 | ip->i_d.di_ino = ip->i_ino; | |
376 | uuid_copy(&ip->i_d.di_uuid, &mp->m_sb.sb_uuid); | |
377 | } else | |
378 | ip->i_d.di_version = 2; | |
379 | return 0; | |
380 | } | |
381 | ||
382 | /* | |
383 | * Get pointers to the on-disk inode and the buffer containing it. | |
384 | */ | |
385 | error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags); | |
386 | if (error) | |
387 | return error; | |
388 | ||
389 | /* even unallocated inodes are verified */ | |
390 | if (!xfs_dinode_verify(mp, ip, dip)) { | |
391 | xfs_alert(mp, "%s: validation failed for inode %lld failed", | |
392 | __func__, ip->i_ino); | |
393 | ||
394 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, dip); | |
395 | error = XFS_ERROR(EFSCORRUPTED); | |
396 | goto out_brelse; | |
397 | } | |
398 | ||
399 | /* | |
400 | * If the on-disk inode is already linked to a directory | |
401 | * entry, copy all of the inode into the in-core inode. | |
402 | * xfs_iformat_fork() handles copying in the inode format | |
403 | * specific information. | |
404 | * Otherwise, just get the truly permanent information. | |
405 | */ | |
406 | if (dip->di_mode) { | |
407 | xfs_dinode_from_disk(&ip->i_d, dip); | |
408 | error = xfs_iformat_fork(ip, dip); | |
409 | if (error) { | |
410 | #ifdef DEBUG | |
411 | xfs_alert(mp, "%s: xfs_iformat() returned error %d", | |
412 | __func__, error); | |
413 | #endif /* DEBUG */ | |
414 | goto out_brelse; | |
415 | } | |
416 | } else { | |
417 | /* | |
418 | * Partial initialisation of the in-core inode. Just the bits | |
419 | * that xfs_ialloc won't overwrite or relies on being correct. | |
420 | */ | |
421 | ip->i_d.di_magic = be16_to_cpu(dip->di_magic); | |
422 | ip->i_d.di_version = dip->di_version; | |
423 | ip->i_d.di_gen = be32_to_cpu(dip->di_gen); | |
424 | ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter); | |
425 | ||
426 | if (dip->di_version == 3) { | |
427 | ip->i_d.di_ino = be64_to_cpu(dip->di_ino); | |
428 | uuid_copy(&ip->i_d.di_uuid, &dip->di_uuid); | |
429 | } | |
430 | ||
431 | /* | |
432 | * Make sure to pull in the mode here as well in | |
433 | * case the inode is released without being used. | |
434 | * This ensures that xfs_inactive() will see that | |
435 | * the inode is already free and not try to mess | |
436 | * with the uninitialized part of it. | |
437 | */ | |
438 | ip->i_d.di_mode = 0; | |
439 | } | |
440 | ||
441 | /* | |
442 | * The inode format changed when we moved the link count and | |
443 | * made it 32 bits long. If this is an old format inode, | |
444 | * convert it in memory to look like a new one. If it gets | |
445 | * flushed to disk we will convert back before flushing or | |
446 | * logging it. We zero out the new projid field and the old link | |
447 | * count field. We'll handle clearing the pad field (the remains | |
448 | * of the old uuid field) when we actually convert the inode to | |
449 | * the new format. We don't change the version number so that we | |
450 | * can distinguish this from a real new format inode. | |
451 | */ | |
452 | if (ip->i_d.di_version == 1) { | |
453 | ip->i_d.di_nlink = ip->i_d.di_onlink; | |
454 | ip->i_d.di_onlink = 0; | |
455 | xfs_set_projid(ip, 0); | |
456 | } | |
457 | ||
458 | ip->i_delayed_blks = 0; | |
459 | ||
460 | /* | |
461 | * Mark the buffer containing the inode as something to keep | |
462 | * around for a while. This helps to keep recently accessed | |
463 | * meta-data in-core longer. | |
464 | */ | |
465 | xfs_buf_set_ref(bp, XFS_INO_REF); | |
466 | ||
467 | /* | |
468 | * Use xfs_trans_brelse() to release the buffer containing the on-disk | |
469 | * inode, because it was acquired with xfs_trans_read_buf() in | |
470 | * xfs_imap_to_bp() above. If tp is NULL, this is just a normal | |
471 | * brelse(). If we're within a transaction, then xfs_trans_brelse() | |
472 | * will only release the buffer if it is not dirty within the | |
473 | * transaction. It will be OK to release the buffer in this case, | |
474 | * because inodes on disk are never destroyed and we will be locking the | |
475 | * new in-core inode before putting it in the cache where other | |
476 | * processes can find it. Thus we don't have to worry about the inode | |
477 | * being changed just because we released the buffer. | |
478 | */ | |
479 | out_brelse: | |
480 | xfs_trans_brelse(tp, bp); | |
481 | return error; | |
482 | } |