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" | |
70a9883c | 20 | #include "xfs_shared.h" |
239880ef DC |
21 | #include "xfs_format.h" |
22 | #include "xfs_log_format.h" | |
23 | #include "xfs_trans_resv.h" | |
1fd7115e DC |
24 | #include "xfs_sb.h" |
25 | #include "xfs_ag.h" | |
26 | #include "xfs_mount.h" | |
1fd7115e DC |
27 | #include "xfs_inode.h" |
28 | #include "xfs_error.h" | |
29 | #include "xfs_cksum.h" | |
30 | #include "xfs_icache.h" | |
239880ef | 31 | #include "xfs_trans.h" |
a4fbe6ab DC |
32 | #include "xfs_ialloc.h" |
33 | #include "xfs_dinode.h" | |
1fd7115e DC |
34 | |
35 | /* | |
36 | * Check that none of the inode's in the buffer have a next | |
37 | * unlinked field of 0. | |
38 | */ | |
39 | #if defined(DEBUG) | |
40 | void | |
41 | xfs_inobp_check( | |
42 | xfs_mount_t *mp, | |
43 | xfs_buf_t *bp) | |
44 | { | |
45 | int i; | |
46 | int j; | |
47 | xfs_dinode_t *dip; | |
48 | ||
49 | j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog; | |
50 | ||
51 | for (i = 0; i < j; i++) { | |
52 | dip = (xfs_dinode_t *)xfs_buf_offset(bp, | |
53 | i * mp->m_sb.sb_inodesize); | |
54 | if (!dip->di_next_unlinked) { | |
55 | xfs_alert(mp, | |
74ffa796 DC |
56 | "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.", |
57 | i, (long long)bp->b_bn); | |
1fd7115e DC |
58 | } |
59 | } | |
60 | } | |
61 | #endif | |
62 | ||
d8914002 DC |
63 | /* |
64 | * If we are doing readahead on an inode buffer, we might be in log recovery | |
65 | * reading an inode allocation buffer that hasn't yet been replayed, and hence | |
66 | * has not had the inode cores stamped into it. Hence for readahead, the buffer | |
67 | * may be potentially invalid. | |
68 | * | |
69 | * If the readahead buffer is invalid, we don't want to mark it with an error, | |
70 | * but we do want to clear the DONE status of the buffer so that a followup read | |
71 | * will re-read it from disk. This will ensure that we don't get an unnecessary | |
72 | * warnings during log recovery and we don't get unnecssary panics on debug | |
73 | * kernels. | |
74 | */ | |
1fd7115e DC |
75 | static void |
76 | xfs_inode_buf_verify( | |
d8914002 DC |
77 | struct xfs_buf *bp, |
78 | bool readahead) | |
1fd7115e DC |
79 | { |
80 | struct xfs_mount *mp = bp->b_target->bt_mount; | |
81 | int i; | |
82 | int ni; | |
83 | ||
84 | /* | |
85 | * Validate the magic number and version of every inode in the buffer | |
86 | */ | |
87 | ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock; | |
88 | for (i = 0; i < ni; i++) { | |
89 | int di_ok; | |
90 | xfs_dinode_t *dip; | |
91 | ||
92 | dip = (struct xfs_dinode *)xfs_buf_offset(bp, | |
93 | (i << mp->m_sb.sb_inodelog)); | |
94 | di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) && | |
95 | XFS_DINODE_GOOD_VERSION(dip->di_version); | |
96 | if (unlikely(XFS_TEST_ERROR(!di_ok, mp, | |
97 | XFS_ERRTAG_ITOBP_INOTOBP, | |
98 | XFS_RANDOM_ITOBP_INOTOBP))) { | |
d8914002 DC |
99 | if (readahead) { |
100 | bp->b_flags &= ~XBF_DONE; | |
101 | return; | |
102 | } | |
103 | ||
1fd7115e | 104 | xfs_buf_ioerror(bp, EFSCORRUPTED); |
ce5028cf | 105 | xfs_verifier_error(bp); |
1fd7115e | 106 | #ifdef DEBUG |
74ffa796 | 107 | xfs_alert(mp, |
1fd7115e DC |
108 | "bad inode magic/vsn daddr %lld #%d (magic=%x)", |
109 | (unsigned long long)bp->b_bn, i, | |
110 | be16_to_cpu(dip->di_magic)); | |
1fd7115e DC |
111 | #endif |
112 | } | |
113 | } | |
114 | xfs_inobp_check(mp, bp); | |
115 | } | |
116 | ||
117 | ||
118 | static void | |
119 | xfs_inode_buf_read_verify( | |
120 | struct xfs_buf *bp) | |
121 | { | |
d8914002 DC |
122 | xfs_inode_buf_verify(bp, false); |
123 | } | |
124 | ||
125 | static void | |
126 | xfs_inode_buf_readahead_verify( | |
127 | struct xfs_buf *bp) | |
128 | { | |
129 | xfs_inode_buf_verify(bp, true); | |
1fd7115e DC |
130 | } |
131 | ||
132 | static void | |
133 | xfs_inode_buf_write_verify( | |
134 | struct xfs_buf *bp) | |
135 | { | |
d8914002 | 136 | xfs_inode_buf_verify(bp, false); |
1fd7115e DC |
137 | } |
138 | ||
139 | const struct xfs_buf_ops xfs_inode_buf_ops = { | |
140 | .verify_read = xfs_inode_buf_read_verify, | |
141 | .verify_write = xfs_inode_buf_write_verify, | |
142 | }; | |
143 | ||
d8914002 DC |
144 | const struct xfs_buf_ops xfs_inode_buf_ra_ops = { |
145 | .verify_read = xfs_inode_buf_readahead_verify, | |
146 | .verify_write = xfs_inode_buf_write_verify, | |
147 | }; | |
148 | ||
1fd7115e DC |
149 | |
150 | /* | |
151 | * This routine is called to map an inode to the buffer containing the on-disk | |
152 | * version of the inode. It returns a pointer to the buffer containing the | |
153 | * on-disk inode in the bpp parameter, and in the dipp parameter it returns a | |
154 | * pointer to the on-disk inode within that buffer. | |
155 | * | |
156 | * If a non-zero error is returned, then the contents of bpp and dipp are | |
157 | * undefined. | |
158 | */ | |
159 | int | |
160 | xfs_imap_to_bp( | |
161 | struct xfs_mount *mp, | |
162 | struct xfs_trans *tp, | |
163 | struct xfs_imap *imap, | |
164 | struct xfs_dinode **dipp, | |
165 | struct xfs_buf **bpp, | |
166 | uint buf_flags, | |
167 | uint iget_flags) | |
168 | { | |
169 | struct xfs_buf *bp; | |
170 | int error; | |
171 | ||
172 | buf_flags |= XBF_UNMAPPED; | |
173 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno, | |
174 | (int)imap->im_len, buf_flags, &bp, | |
175 | &xfs_inode_buf_ops); | |
176 | if (error) { | |
177 | if (error == EAGAIN) { | |
178 | ASSERT(buf_flags & XBF_TRYLOCK); | |
179 | return error; | |
180 | } | |
181 | ||
182 | if (error == EFSCORRUPTED && | |
183 | (iget_flags & XFS_IGET_UNTRUSTED)) | |
b474c7ae | 184 | return EINVAL; |
1fd7115e DC |
185 | |
186 | xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.", | |
187 | __func__, error); | |
188 | return error; | |
189 | } | |
190 | ||
191 | *bpp = bp; | |
192 | *dipp = (struct xfs_dinode *)xfs_buf_offset(bp, imap->im_boffset); | |
193 | return 0; | |
194 | } | |
195 | ||
638f4416 | 196 | void |
1fd7115e DC |
197 | xfs_dinode_from_disk( |
198 | xfs_icdinode_t *to, | |
199 | xfs_dinode_t *from) | |
200 | { | |
201 | to->di_magic = be16_to_cpu(from->di_magic); | |
202 | to->di_mode = be16_to_cpu(from->di_mode); | |
203 | to->di_version = from ->di_version; | |
204 | to->di_format = from->di_format; | |
205 | to->di_onlink = be16_to_cpu(from->di_onlink); | |
206 | to->di_uid = be32_to_cpu(from->di_uid); | |
207 | to->di_gid = be32_to_cpu(from->di_gid); | |
208 | to->di_nlink = be32_to_cpu(from->di_nlink); | |
209 | to->di_projid_lo = be16_to_cpu(from->di_projid_lo); | |
210 | to->di_projid_hi = be16_to_cpu(from->di_projid_hi); | |
211 | memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); | |
212 | to->di_flushiter = be16_to_cpu(from->di_flushiter); | |
213 | to->di_atime.t_sec = be32_to_cpu(from->di_atime.t_sec); | |
214 | to->di_atime.t_nsec = be32_to_cpu(from->di_atime.t_nsec); | |
215 | to->di_mtime.t_sec = be32_to_cpu(from->di_mtime.t_sec); | |
216 | to->di_mtime.t_nsec = be32_to_cpu(from->di_mtime.t_nsec); | |
217 | to->di_ctime.t_sec = be32_to_cpu(from->di_ctime.t_sec); | |
218 | to->di_ctime.t_nsec = be32_to_cpu(from->di_ctime.t_nsec); | |
219 | to->di_size = be64_to_cpu(from->di_size); | |
220 | to->di_nblocks = be64_to_cpu(from->di_nblocks); | |
221 | to->di_extsize = be32_to_cpu(from->di_extsize); | |
222 | to->di_nextents = be32_to_cpu(from->di_nextents); | |
223 | to->di_anextents = be16_to_cpu(from->di_anextents); | |
224 | to->di_forkoff = from->di_forkoff; | |
225 | to->di_aformat = from->di_aformat; | |
226 | to->di_dmevmask = be32_to_cpu(from->di_dmevmask); | |
227 | to->di_dmstate = be16_to_cpu(from->di_dmstate); | |
228 | to->di_flags = be16_to_cpu(from->di_flags); | |
229 | to->di_gen = be32_to_cpu(from->di_gen); | |
230 | ||
231 | if (to->di_version == 3) { | |
232 | to->di_changecount = be64_to_cpu(from->di_changecount); | |
233 | to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec); | |
234 | to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec); | |
235 | to->di_flags2 = be64_to_cpu(from->di_flags2); | |
236 | to->di_ino = be64_to_cpu(from->di_ino); | |
237 | to->di_lsn = be64_to_cpu(from->di_lsn); | |
238 | memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2)); | |
239 | uuid_copy(&to->di_uuid, &from->di_uuid); | |
240 | } | |
241 | } | |
242 | ||
243 | void | |
244 | xfs_dinode_to_disk( | |
245 | xfs_dinode_t *to, | |
246 | xfs_icdinode_t *from) | |
247 | { | |
248 | to->di_magic = cpu_to_be16(from->di_magic); | |
249 | to->di_mode = cpu_to_be16(from->di_mode); | |
250 | to->di_version = from ->di_version; | |
251 | to->di_format = from->di_format; | |
252 | to->di_onlink = cpu_to_be16(from->di_onlink); | |
253 | to->di_uid = cpu_to_be32(from->di_uid); | |
254 | to->di_gid = cpu_to_be32(from->di_gid); | |
255 | to->di_nlink = cpu_to_be32(from->di_nlink); | |
256 | to->di_projid_lo = cpu_to_be16(from->di_projid_lo); | |
257 | to->di_projid_hi = cpu_to_be16(from->di_projid_hi); | |
258 | memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); | |
259 | to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec); | |
260 | to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec); | |
261 | to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec); | |
262 | to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec); | |
263 | to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec); | |
264 | to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec); | |
265 | to->di_size = cpu_to_be64(from->di_size); | |
266 | to->di_nblocks = cpu_to_be64(from->di_nblocks); | |
267 | to->di_extsize = cpu_to_be32(from->di_extsize); | |
268 | to->di_nextents = cpu_to_be32(from->di_nextents); | |
269 | to->di_anextents = cpu_to_be16(from->di_anextents); | |
270 | to->di_forkoff = from->di_forkoff; | |
271 | to->di_aformat = from->di_aformat; | |
272 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); | |
273 | to->di_dmstate = cpu_to_be16(from->di_dmstate); | |
274 | to->di_flags = cpu_to_be16(from->di_flags); | |
275 | to->di_gen = cpu_to_be32(from->di_gen); | |
276 | ||
277 | if (from->di_version == 3) { | |
278 | to->di_changecount = cpu_to_be64(from->di_changecount); | |
279 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec); | |
280 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec); | |
281 | to->di_flags2 = cpu_to_be64(from->di_flags2); | |
282 | to->di_ino = cpu_to_be64(from->di_ino); | |
283 | to->di_lsn = cpu_to_be64(from->di_lsn); | |
284 | memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2)); | |
285 | uuid_copy(&to->di_uuid, &from->di_uuid); | |
286 | to->di_flushiter = 0; | |
287 | } else { | |
288 | to->di_flushiter = cpu_to_be16(from->di_flushiter); | |
289 | } | |
290 | } | |
291 | ||
292 | static bool | |
293 | xfs_dinode_verify( | |
294 | struct xfs_mount *mp, | |
295 | struct xfs_inode *ip, | |
296 | struct xfs_dinode *dip) | |
297 | { | |
298 | if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC)) | |
299 | return false; | |
300 | ||
301 | /* only version 3 or greater inodes are extensively verified here */ | |
302 | if (dip->di_version < 3) | |
303 | return true; | |
304 | ||
305 | if (!xfs_sb_version_hascrc(&mp->m_sb)) | |
306 | return false; | |
307 | if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize, | |
533b81c8 | 308 | XFS_DINODE_CRC_OFF)) |
1fd7115e DC |
309 | return false; |
310 | if (be64_to_cpu(dip->di_ino) != ip->i_ino) | |
311 | return false; | |
312 | if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_uuid)) | |
313 | return false; | |
314 | return true; | |
315 | } | |
316 | ||
317 | void | |
318 | xfs_dinode_calc_crc( | |
319 | struct xfs_mount *mp, | |
320 | struct xfs_dinode *dip) | |
321 | { | |
322 | __uint32_t crc; | |
323 | ||
324 | if (dip->di_version < 3) | |
325 | return; | |
326 | ||
327 | ASSERT(xfs_sb_version_hascrc(&mp->m_sb)); | |
328 | crc = xfs_start_cksum((char *)dip, mp->m_sb.sb_inodesize, | |
533b81c8 | 329 | XFS_DINODE_CRC_OFF); |
1fd7115e DC |
330 | dip->di_crc = xfs_end_cksum(crc); |
331 | } | |
332 | ||
333 | /* | |
334 | * Read the disk inode attributes into the in-core inode structure. | |
335 | * | |
336 | * For version 5 superblocks, if we are initialising a new inode and we are not | |
337 | * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new | |
338 | * inode core with a random generation number. If we are keeping inodes around, | |
339 | * we need to read the inode cluster to get the existing generation number off | |
340 | * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode | |
341 | * format) then log recovery is dependent on the di_flushiter field being | |
342 | * initialised from the current on-disk value and hence we must also read the | |
343 | * inode off disk. | |
344 | */ | |
345 | int | |
346 | xfs_iread( | |
347 | xfs_mount_t *mp, | |
348 | xfs_trans_t *tp, | |
349 | xfs_inode_t *ip, | |
350 | uint iget_flags) | |
351 | { | |
352 | xfs_buf_t *bp; | |
353 | xfs_dinode_t *dip; | |
354 | int error; | |
355 | ||
356 | /* | |
357 | * Fill in the location information in the in-core inode. | |
358 | */ | |
359 | error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags); | |
360 | if (error) | |
361 | return error; | |
362 | ||
363 | /* shortcut IO on inode allocation if possible */ | |
364 | if ((iget_flags & XFS_IGET_CREATE) && | |
365 | xfs_sb_version_hascrc(&mp->m_sb) && | |
366 | !(mp->m_flags & XFS_MOUNT_IKEEP)) { | |
367 | /* initialise the on-disk inode core */ | |
368 | memset(&ip->i_d, 0, sizeof(ip->i_d)); | |
369 | ip->i_d.di_magic = XFS_DINODE_MAGIC; | |
370 | ip->i_d.di_gen = prandom_u32(); | |
371 | if (xfs_sb_version_hascrc(&mp->m_sb)) { | |
372 | ip->i_d.di_version = 3; | |
373 | ip->i_d.di_ino = ip->i_ino; | |
374 | uuid_copy(&ip->i_d.di_uuid, &mp->m_sb.sb_uuid); | |
375 | } else | |
376 | ip->i_d.di_version = 2; | |
377 | return 0; | |
378 | } | |
379 | ||
380 | /* | |
381 | * Get pointers to the on-disk inode and the buffer containing it. | |
382 | */ | |
383 | error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags); | |
384 | if (error) | |
385 | return error; | |
386 | ||
387 | /* even unallocated inodes are verified */ | |
388 | if (!xfs_dinode_verify(mp, ip, dip)) { | |
389 | xfs_alert(mp, "%s: validation failed for inode %lld failed", | |
390 | __func__, ip->i_ino); | |
391 | ||
392 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, dip); | |
b474c7ae | 393 | error = EFSCORRUPTED; |
1fd7115e DC |
394 | goto out_brelse; |
395 | } | |
396 | ||
397 | /* | |
398 | * If the on-disk inode is already linked to a directory | |
399 | * entry, copy all of the inode into the in-core inode. | |
400 | * xfs_iformat_fork() handles copying in the inode format | |
401 | * specific information. | |
402 | * Otherwise, just get the truly permanent information. | |
403 | */ | |
404 | if (dip->di_mode) { | |
405 | xfs_dinode_from_disk(&ip->i_d, dip); | |
406 | error = xfs_iformat_fork(ip, dip); | |
407 | if (error) { | |
408 | #ifdef DEBUG | |
409 | xfs_alert(mp, "%s: xfs_iformat() returned error %d", | |
410 | __func__, error); | |
411 | #endif /* DEBUG */ | |
412 | goto out_brelse; | |
413 | } | |
414 | } else { | |
415 | /* | |
416 | * Partial initialisation of the in-core inode. Just the bits | |
417 | * that xfs_ialloc won't overwrite or relies on being correct. | |
418 | */ | |
419 | ip->i_d.di_magic = be16_to_cpu(dip->di_magic); | |
420 | ip->i_d.di_version = dip->di_version; | |
421 | ip->i_d.di_gen = be32_to_cpu(dip->di_gen); | |
422 | ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter); | |
423 | ||
424 | if (dip->di_version == 3) { | |
425 | ip->i_d.di_ino = be64_to_cpu(dip->di_ino); | |
426 | uuid_copy(&ip->i_d.di_uuid, &dip->di_uuid); | |
427 | } | |
428 | ||
429 | /* | |
430 | * Make sure to pull in the mode here as well in | |
431 | * case the inode is released without being used. | |
432 | * This ensures that xfs_inactive() will see that | |
433 | * the inode is already free and not try to mess | |
434 | * with the uninitialized part of it. | |
435 | */ | |
436 | ip->i_d.di_mode = 0; | |
437 | } | |
438 | ||
439 | /* | |
263997a6 DC |
440 | * Automatically convert version 1 inode formats in memory to version 2 |
441 | * inode format. If the inode is modified, it will get logged and | |
442 | * rewritten as a version 2 inode. We can do this because we set the | |
443 | * superblock feature bit for v2 inodes unconditionally during mount | |
444 | * and it means the reast of the code can assume the inode version is 2 | |
445 | * or higher. | |
1fd7115e DC |
446 | */ |
447 | if (ip->i_d.di_version == 1) { | |
263997a6 DC |
448 | ip->i_d.di_version = 2; |
449 | memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad)); | |
1fd7115e DC |
450 | ip->i_d.di_nlink = ip->i_d.di_onlink; |
451 | ip->i_d.di_onlink = 0; | |
452 | xfs_set_projid(ip, 0); | |
453 | } | |
454 | ||
455 | ip->i_delayed_blks = 0; | |
456 | ||
457 | /* | |
458 | * Mark the buffer containing the inode as something to keep | |
459 | * around for a while. This helps to keep recently accessed | |
460 | * meta-data in-core longer. | |
461 | */ | |
462 | xfs_buf_set_ref(bp, XFS_INO_REF); | |
463 | ||
464 | /* | |
465 | * Use xfs_trans_brelse() to release the buffer containing the on-disk | |
466 | * inode, because it was acquired with xfs_trans_read_buf() in | |
467 | * xfs_imap_to_bp() above. If tp is NULL, this is just a normal | |
468 | * brelse(). If we're within a transaction, then xfs_trans_brelse() | |
469 | * will only release the buffer if it is not dirty within the | |
470 | * transaction. It will be OK to release the buffer in this case, | |
471 | * because inodes on disk are never destroyed and we will be locking the | |
472 | * new in-core inode before putting it in the cache where other | |
473 | * processes can find it. Thus we don't have to worry about the inode | |
474 | * being changed just because we released the buffer. | |
475 | */ | |
476 | out_brelse: | |
477 | xfs_trans_brelse(tp, bp); | |
478 | return error; | |
479 | } |