]>
Commit | Line | Data |
---|---|---|
37b3b4d6 | 1 | // SPDX-License-Identifier: GPL-2.0 |
2bd0ea18 | 2 | /* |
5e656dbb | 3 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. |
da23017d | 4 | * All Rights Reserved. |
2bd0ea18 | 5 | */ |
9c799827 | 6 | #include "libxfs_priv.h" |
b626fb59 DC |
7 | #include "xfs_fs.h" |
8 | #include "xfs_shared.h" | |
9 | #include "xfs_format.h" | |
10 | #include "xfs_log_format.h" | |
11 | #include "xfs_trans_resv.h" | |
12 | #include "xfs_mount.h" | |
f944d3d0 | 13 | #include "xfs_defer.h" |
b626fb59 | 14 | #include "xfs_inode.h" |
56d3fc2b | 15 | #include "xfs_errortag.h" |
b626fb59 DC |
16 | #include "xfs_cksum.h" |
17 | #include "xfs_trans.h" | |
18 | #include "xfs_ialloc.h" | |
0962b5b3 | 19 | #include "xfs_dir2.h" |
2bd0ea18 | 20 | |
5e656dbb BN |
21 | /* |
22 | * Check that none of the inode's in the buffer have a next | |
23 | * unlinked field of 0. | |
24 | */ | |
25 | #if defined(DEBUG) | |
2bd0ea18 NS |
26 | void |
27 | xfs_inobp_check( | |
28 | xfs_mount_t *mp, | |
29 | xfs_buf_t *bp) | |
30 | { | |
31 | int i; | |
32 | int j; | |
33 | xfs_dinode_t *dip; | |
34 | ||
35 | j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog; | |
36 | ||
37 | for (i = 0; i < j; i++) { | |
92acb899 | 38 | dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize); |
46eca962 | 39 | if (!dip->di_next_unlinked) { |
a2ceac1f | 40 | xfs_alert(mp, |
ea6a00d4 DC |
41 | "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.", |
42 | i, (long long)bp->b_bn); | |
2bd0ea18 NS |
43 | } |
44 | } | |
45 | } | |
46 | #endif | |
47 | ||
307ae950 RW |
48 | bool |
49 | xfs_dinode_good_version( | |
50 | struct xfs_mount *mp, | |
51 | __u8 version) | |
52 | { | |
53 | if (xfs_sb_version_hascrc(&mp->m_sb)) | |
54 | return version == 3; | |
55 | ||
56 | return version == 1 || version == 2; | |
57 | } | |
58 | ||
f85fc622 DC |
59 | /* |
60 | * If we are doing readahead on an inode buffer, we might be in log recovery | |
61 | * reading an inode allocation buffer that hasn't yet been replayed, and hence | |
62 | * has not had the inode cores stamped into it. Hence for readahead, the buffer | |
63 | * may be potentially invalid. | |
64 | * | |
736e1cf6 DC |
65 | * If the readahead buffer is invalid, we need to mark it with an error and |
66 | * clear the DONE status of the buffer so that a followup read will re-read it | |
67 | * from disk. We don't report the error otherwise to avoid warnings during log | |
68 | * recovery and we don't get unnecssary panics on debug kernels. We use EIO here | |
69 | * because all we want to do is say readahead failed; there is no-one to report | |
70 | * the error to, so this will distinguish it from a non-ra verifier failure. | |
6daba42a DC |
71 | * Changes to this readahead error behavour also need to be reflected in |
72 | * xfs_dquot_buf_readahead_verify(). | |
f85fc622 | 73 | */ |
a2ceac1f DC |
74 | static void |
75 | xfs_inode_buf_verify( | |
f85fc622 DC |
76 | struct xfs_buf *bp, |
77 | bool readahead) | |
5e656dbb | 78 | { |
a2ceac1f | 79 | struct xfs_mount *mp = bp->b_target->bt_mount; |
2949b467 | 80 | xfs_agnumber_t agno; |
5e656dbb BN |
81 | int i; |
82 | int ni; | |
5e656dbb BN |
83 | |
84 | /* | |
85 | * Validate the magic number and version of every inode in the buffer | |
5e656dbb | 86 | */ |
2949b467 | 87 | agno = xfs_daddr_to_agno(mp, XFS_BUF_ADDR(bp)); |
a2ceac1f | 88 | ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock; |
5e656dbb BN |
89 | for (i = 0; i < ni; i++) { |
90 | int di_ok; | |
91 | xfs_dinode_t *dip; | |
2949b467 | 92 | xfs_agino_t unlinked_ino; |
5e656dbb | 93 | |
92acb899 | 94 | dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog)); |
2949b467 | 95 | unlinked_ino = be32_to_cpu(dip->di_next_unlinked); |
9e26de8f | 96 | di_ok = xfs_verify_magic16(bp, dip->di_magic) && |
2949b467 | 97 | xfs_dinode_good_version(mp, dip->di_version) && |
7a425ea8 | 98 | xfs_verify_agino_or_null(mp, agno, unlinked_ino); |
5e656dbb | 99 | if (unlikely(XFS_TEST_ERROR(!di_ok, mp, |
e2a190dd | 100 | XFS_ERRTAG_ITOBP_INOTOBP))) { |
f85fc622 DC |
101 | if (readahead) { |
102 | bp->b_flags &= ~XBF_DONE; | |
736e1cf6 | 103 | xfs_buf_ioerror(bp, -EIO); |
f85fc622 DC |
104 | return; |
105 | } | |
106 | ||
5e656dbb | 107 | #ifdef DEBUG |
ea6a00d4 | 108 | xfs_alert(mp, |
a2ceac1f DC |
109 | "bad inode magic/vsn daddr %lld #%d (magic=%x)", |
110 | (unsigned long long)bp->b_bn, i, | |
56b2de80 | 111 | be16_to_cpu(dip->di_magic)); |
5e656dbb | 112 | #endif |
43f7ceb5 DW |
113 | xfs_buf_verifier_error(bp, -EFSCORRUPTED, |
114 | __func__, dip, sizeof(*dip), | |
115 | NULL); | |
2949b467 | 116 | return; |
5e656dbb BN |
117 | } |
118 | } | |
a2ceac1f | 119 | } |
5e656dbb | 120 | |
e6d77a21 | 121 | |
a2ceac1f DC |
122 | static void |
123 | xfs_inode_buf_read_verify( | |
124 | struct xfs_buf *bp) | |
125 | { | |
f85fc622 DC |
126 | xfs_inode_buf_verify(bp, false); |
127 | } | |
128 | ||
129 | static void | |
130 | xfs_inode_buf_readahead_verify( | |
131 | struct xfs_buf *bp) | |
132 | { | |
133 | xfs_inode_buf_verify(bp, true); | |
5e656dbb | 134 | } |
2bd0ea18 | 135 | |
a2ceac1f DC |
136 | static void |
137 | xfs_inode_buf_write_verify( | |
138 | struct xfs_buf *bp) | |
56b2de80 | 139 | { |
f85fc622 | 140 | xfs_inode_buf_verify(bp, false); |
56b2de80 DC |
141 | } |
142 | ||
a2ceac1f | 143 | const struct xfs_buf_ops xfs_inode_buf_ops = { |
a3fac935 | 144 | .name = "xfs_inode", |
9e26de8f DW |
145 | .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC), |
146 | cpu_to_be16(XFS_DINODE_MAGIC) }, | |
a2ceac1f DC |
147 | .verify_read = xfs_inode_buf_read_verify, |
148 | .verify_write = xfs_inode_buf_write_verify, | |
149 | }; | |
150 | ||
f85fc622 | 151 | const struct xfs_buf_ops xfs_inode_buf_ra_ops = { |
7563fc9c | 152 | .name = "xfs_inode_ra", |
9e26de8f DW |
153 | .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC), |
154 | cpu_to_be16(XFS_DINODE_MAGIC) }, | |
f85fc622 DC |
155 | .verify_read = xfs_inode_buf_readahead_verify, |
156 | .verify_write = xfs_inode_buf_write_verify, | |
157 | }; | |
158 | ||
159 | ||
2bd0ea18 | 160 | /* |
a2ceac1f DC |
161 | * This routine is called to map an inode to the buffer containing the on-disk |
162 | * version of the inode. It returns a pointer to the buffer containing the | |
163 | * on-disk inode in the bpp parameter, and in the dipp parameter it returns a | |
164 | * pointer to the on-disk inode within that buffer. | |
2bd0ea18 | 165 | * |
a2ceac1f DC |
166 | * If a non-zero error is returned, then the contents of bpp and dipp are |
167 | * undefined. | |
2bd0ea18 NS |
168 | */ |
169 | int | |
a2ceac1f DC |
170 | xfs_imap_to_bp( |
171 | struct xfs_mount *mp, | |
172 | struct xfs_trans *tp, | |
173 | struct xfs_imap *imap, | |
174 | struct xfs_dinode **dipp, | |
175 | struct xfs_buf **bpp, | |
176 | uint buf_flags, | |
177 | uint iget_flags) | |
2bd0ea18 | 178 | { |
a2ceac1f DC |
179 | struct xfs_buf *bp; |
180 | int error; | |
2bd0ea18 | 181 | |
a2ceac1f DC |
182 | buf_flags |= XBF_UNMAPPED; |
183 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno, | |
184 | (int)imap->im_len, buf_flags, &bp, | |
185 | &xfs_inode_buf_ops); | |
186 | if (error) { | |
12b53197 | 187 | if (error == -EAGAIN) { |
a2ceac1f DC |
188 | ASSERT(buf_flags & XBF_TRYLOCK); |
189 | return error; | |
190 | } | |
a2ceac1f DC |
191 | xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.", |
192 | __func__, error); | |
193 | return error; | |
2bd0ea18 | 194 | } |
2bd0ea18 | 195 | |
2bd0ea18 | 196 | *bpp = bp; |
92acb899 | 197 | *dipp = xfs_buf_offset(bp, imap->im_boffset); |
2bd0ea18 NS |
198 | return 0; |
199 | } | |
200 | ||
5000d01d | 201 | void |
1bc6cbe3 DC |
202 | xfs_inode_from_disk( |
203 | struct xfs_inode *ip, | |
138659f1 | 204 | struct xfs_dinode *from) |
2bd0ea18 | 205 | { |
1bc6cbe3 DC |
206 | struct xfs_icdinode *to = &ip->i_d; |
207 | struct inode *inode = VFS_I(ip); | |
208 | ||
f089fc42 DC |
209 | |
210 | /* | |
211 | * Convert v1 inodes immediately to v2 inode format as this is the | |
212 | * minimum inode version format we support in the rest of the code. | |
213 | */ | |
e37bf53c | 214 | to->di_version = from->di_version; |
f089fc42 | 215 | if (to->di_version == 1) { |
bcbe04c1 | 216 | set_nlink(inode, be16_to_cpu(from->di_onlink)); |
f089fc42 DC |
217 | to->di_projid_lo = 0; |
218 | to->di_projid_hi = 0; | |
219 | to->di_version = 2; | |
220 | } else { | |
bcbe04c1 | 221 | set_nlink(inode, be32_to_cpu(from->di_nlink)); |
f089fc42 DC |
222 | to->di_projid_lo = be16_to_cpu(from->di_projid_lo); |
223 | to->di_projid_hi = be16_to_cpu(from->di_projid_hi); | |
224 | } | |
225 | ||
5e656dbb | 226 | to->di_format = from->di_format; |
5e656dbb BN |
227 | to->di_uid = be32_to_cpu(from->di_uid); |
228 | to->di_gid = be32_to_cpu(from->di_gid); | |
5e656dbb | 229 | to->di_flushiter = be16_to_cpu(from->di_flushiter); |
1bc6cbe3 DC |
230 | |
231 | /* | |
232 | * Time is signed, so need to convert to signed 32 bit before | |
233 | * storing in inode timestamp which may be 64 bit. Otherwise | |
234 | * a time before epoch is converted to a time long after epoch | |
235 | * on 64 bit systems. | |
236 | */ | |
237 | inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec); | |
238 | inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec); | |
239 | inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec); | |
240 | inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec); | |
241 | inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec); | |
242 | inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec); | |
6652c253 | 243 | inode->i_generation = be32_to_cpu(from->di_gen); |
e37bf53c | 244 | inode->i_mode = be16_to_cpu(from->di_mode); |
1bc6cbe3 | 245 | |
5e656dbb BN |
246 | to->di_size = be64_to_cpu(from->di_size); |
247 | to->di_nblocks = be64_to_cpu(from->di_nblocks); | |
248 | to->di_extsize = be32_to_cpu(from->di_extsize); | |
249 | to->di_nextents = be32_to_cpu(from->di_nextents); | |
250 | to->di_anextents = be16_to_cpu(from->di_anextents); | |
251 | to->di_forkoff = from->di_forkoff; | |
252 | to->di_aformat = from->di_aformat; | |
253 | to->di_dmevmask = be32_to_cpu(from->di_dmevmask); | |
254 | to->di_dmstate = be16_to_cpu(from->di_dmstate); | |
255 | to->di_flags = be16_to_cpu(from->di_flags); | |
41ce5f36 DC |
256 | |
257 | if (to->di_version == 3) { | |
d5105a89 JL |
258 | inode_set_iversion_queried(inode, |
259 | be64_to_cpu(from->di_changecount)); | |
41ce5f36 DC |
260 | to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec); |
261 | to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec); | |
262 | to->di_flags2 = be64_to_cpu(from->di_flags2); | |
10460994 | 263 | to->di_cowextsize = be32_to_cpu(from->di_cowextsize); |
41ce5f36 | 264 | } |
5e656dbb BN |
265 | } |
266 | ||
267 | void | |
1bc6cbe3 DC |
268 | xfs_inode_to_disk( |
269 | struct xfs_inode *ip, | |
db17aebe DC |
270 | struct xfs_dinode *to, |
271 | xfs_lsn_t lsn) | |
1bc6cbe3 DC |
272 | { |
273 | struct xfs_icdinode *from = &ip->i_d; | |
274 | struct inode *inode = VFS_I(ip); | |
275 | ||
db17aebe | 276 | to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); |
f089fc42 | 277 | to->di_onlink = 0; |
db17aebe | 278 | |
db17aebe | 279 | to->di_version = from->di_version; |
1bc6cbe3 | 280 | to->di_format = from->di_format; |
1bc6cbe3 DC |
281 | to->di_uid = cpu_to_be32(from->di_uid); |
282 | to->di_gid = cpu_to_be32(from->di_gid); | |
1bc6cbe3 DC |
283 | to->di_projid_lo = cpu_to_be16(from->di_projid_lo); |
284 | to->di_projid_hi = cpu_to_be16(from->di_projid_hi); | |
1bc6cbe3 | 285 | |
db17aebe | 286 | memset(to->di_pad, 0, sizeof(to->di_pad)); |
1bc6cbe3 DC |
287 | to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec); |
288 | to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec); | |
289 | to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec); | |
290 | to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec); | |
291 | to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec); | |
292 | to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec); | |
bcbe04c1 | 293 | to->di_nlink = cpu_to_be32(inode->i_nlink); |
6652c253 | 294 | to->di_gen = cpu_to_be32(inode->i_generation); |
e37bf53c | 295 | to->di_mode = cpu_to_be16(inode->i_mode); |
1bc6cbe3 DC |
296 | |
297 | to->di_size = cpu_to_be64(from->di_size); | |
298 | to->di_nblocks = cpu_to_be64(from->di_nblocks); | |
299 | to->di_extsize = cpu_to_be32(from->di_extsize); | |
300 | to->di_nextents = cpu_to_be32(from->di_nextents); | |
301 | to->di_anextents = cpu_to_be16(from->di_anextents); | |
302 | to->di_forkoff = from->di_forkoff; | |
303 | to->di_aformat = from->di_aformat; | |
304 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); | |
305 | to->di_dmstate = cpu_to_be16(from->di_dmstate); | |
306 | to->di_flags = cpu_to_be16(from->di_flags); | |
1bc6cbe3 DC |
307 | |
308 | if (from->di_version == 3) { | |
d5105a89 | 309 | to->di_changecount = cpu_to_be64(inode_peek_iversion(inode)); |
1bc6cbe3 DC |
310 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec); |
311 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec); | |
312 | to->di_flags2 = cpu_to_be64(from->di_flags2); | |
10460994 | 313 | to->di_cowextsize = cpu_to_be32(from->di_cowextsize); |
db17aebe DC |
314 | to->di_ino = cpu_to_be64(ip->i_ino); |
315 | to->di_lsn = cpu_to_be64(lsn); | |
316 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); | |
317 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); | |
1bc6cbe3 DC |
318 | to->di_flushiter = 0; |
319 | } else { | |
320 | to->di_flushiter = cpu_to_be16(from->di_flushiter); | |
321 | } | |
322 | } | |
323 | ||
324 | void | |
325 | xfs_log_dinode_to_disk( | |
326 | struct xfs_log_dinode *from, | |
327 | struct xfs_dinode *to) | |
5e656dbb BN |
328 | { |
329 | to->di_magic = cpu_to_be16(from->di_magic); | |
330 | to->di_mode = cpu_to_be16(from->di_mode); | |
f089fc42 | 331 | to->di_version = from->di_version; |
5e656dbb | 332 | to->di_format = from->di_format; |
f089fc42 | 333 | to->di_onlink = 0; |
5e656dbb BN |
334 | to->di_uid = cpu_to_be32(from->di_uid); |
335 | to->di_gid = cpu_to_be32(from->di_gid); | |
336 | to->di_nlink = cpu_to_be32(from->di_nlink); | |
22bc10ed AM |
337 | to->di_projid_lo = cpu_to_be16(from->di_projid_lo); |
338 | to->di_projid_hi = cpu_to_be16(from->di_projid_hi); | |
5e656dbb | 339 | memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); |
1bc6cbe3 | 340 | |
5e656dbb BN |
341 | to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec); |
342 | to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec); | |
343 | to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec); | |
344 | to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec); | |
345 | to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec); | |
346 | to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec); | |
1bc6cbe3 | 347 | |
5e656dbb BN |
348 | to->di_size = cpu_to_be64(from->di_size); |
349 | to->di_nblocks = cpu_to_be64(from->di_nblocks); | |
350 | to->di_extsize = cpu_to_be32(from->di_extsize); | |
351 | to->di_nextents = cpu_to_be32(from->di_nextents); | |
352 | to->di_anextents = cpu_to_be16(from->di_anextents); | |
353 | to->di_forkoff = from->di_forkoff; | |
354 | to->di_aformat = from->di_aformat; | |
355 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); | |
356 | to->di_dmstate = cpu_to_be16(from->di_dmstate); | |
357 | to->di_flags = cpu_to_be16(from->di_flags); | |
358 | to->di_gen = cpu_to_be32(from->di_gen); | |
41ce5f36 DC |
359 | |
360 | if (from->di_version == 3) { | |
361 | to->di_changecount = cpu_to_be64(from->di_changecount); | |
362 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec); | |
363 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec); | |
364 | to->di_flags2 = cpu_to_be64(from->di_flags2); | |
10460994 | 365 | to->di_cowextsize = cpu_to_be32(from->di_cowextsize); |
41ce5f36 DC |
366 | to->di_ino = cpu_to_be64(from->di_ino); |
367 | to->di_lsn = cpu_to_be64(from->di_lsn); | |
368 | memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2)); | |
f85fc622 | 369 | uuid_copy(&to->di_uuid, &from->di_uuid); |
fde71651 DC |
370 | to->di_flushiter = 0; |
371 | } else { | |
372 | to->di_flushiter = cpu_to_be16(from->di_flushiter); | |
41ce5f36 DC |
373 | } |
374 | } | |
375 | ||
8ac50baf DC |
376 | static xfs_failaddr_t |
377 | xfs_dinode_verify_fork( | |
378 | struct xfs_dinode *dip, | |
379 | struct xfs_mount *mp, | |
380 | int whichfork) | |
381 | { | |
382 | uint32_t di_nextents = XFS_DFORK_NEXTENTS(dip, whichfork); | |
383 | ||
384 | switch (XFS_DFORK_FORMAT(dip, whichfork)) { | |
385 | case XFS_DINODE_FMT_LOCAL: | |
386 | /* | |
387 | * no local regular files yet | |
388 | */ | |
389 | if (whichfork == XFS_DATA_FORK) { | |
390 | if (S_ISREG(be16_to_cpu(dip->di_mode))) | |
391 | return __this_address; | |
392 | if (be64_to_cpu(dip->di_size) > | |
393 | XFS_DFORK_SIZE(dip, mp, whichfork)) | |
394 | return __this_address; | |
395 | } | |
396 | if (di_nextents) | |
397 | return __this_address; | |
398 | break; | |
399 | case XFS_DINODE_FMT_EXTENTS: | |
400 | if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork)) | |
401 | return __this_address; | |
402 | break; | |
403 | case XFS_DINODE_FMT_BTREE: | |
404 | if (whichfork == XFS_ATTR_FORK) { | |
405 | if (di_nextents > MAXAEXTNUM) | |
406 | return __this_address; | |
407 | } else if (di_nextents > MAXEXTNUM) { | |
408 | return __this_address; | |
409 | } | |
410 | break; | |
411 | default: | |
412 | return __this_address; | |
413 | } | |
414 | return NULL; | |
415 | } | |
416 | ||
91b32a08 ES |
417 | static xfs_failaddr_t |
418 | xfs_dinode_verify_forkoff( | |
419 | struct xfs_dinode *dip, | |
420 | struct xfs_mount *mp) | |
421 | { | |
422 | if (!XFS_DFORK_Q(dip)) | |
423 | return NULL; | |
424 | ||
425 | switch (dip->di_format) { | |
426 | case XFS_DINODE_FMT_DEV: | |
427 | if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3)) | |
428 | return __this_address; | |
429 | break; | |
430 | case XFS_DINODE_FMT_LOCAL: /* fall through ... */ | |
431 | case XFS_DINODE_FMT_EXTENTS: /* fall through ... */ | |
432 | case XFS_DINODE_FMT_BTREE: | |
433 | if (dip->di_forkoff >= (XFS_LITINO(mp, dip->di_version) >> 3)) | |
434 | return __this_address; | |
435 | break; | |
436 | default: | |
437 | return __this_address; | |
438 | } | |
439 | return NULL; | |
440 | } | |
441 | ||
bc01119d | 442 | xfs_failaddr_t |
41ce5f36 DC |
443 | xfs_dinode_verify( |
444 | struct xfs_mount *mp, | |
e515cca1 | 445 | xfs_ino_t ino, |
41ce5f36 DC |
446 | struct xfs_dinode *dip) |
447 | { | |
316d5a9f | 448 | xfs_failaddr_t fa; |
3cfabff6 | 449 | uint16_t mode; |
183537ed DW |
450 | uint16_t flags; |
451 | uint64_t flags2; | |
ad42e5a1 | 452 | uint64_t di_size; |
183537ed | 453 | |
41ce5f36 | 454 | if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC)) |
bc01119d | 455 | return __this_address; |
41ce5f36 | 456 | |
98703400 DW |
457 | /* Verify v3 integrity information first */ |
458 | if (dip->di_version >= 3) { | |
459 | if (!xfs_sb_version_hascrc(&mp->m_sb)) | |
460 | return __this_address; | |
461 | if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize, | |
462 | XFS_DINODE_CRC_OFF)) | |
463 | return __this_address; | |
464 | if (be64_to_cpu(dip->di_ino) != ino) | |
465 | return __this_address; | |
466 | if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid)) | |
467 | return __this_address; | |
468 | } | |
469 | ||
9d85eb44 | 470 | /* don't allow invalid i_size */ |
ad42e5a1 DW |
471 | di_size = be64_to_cpu(dip->di_size); |
472 | if (di_size & (1ULL << 63)) | |
bc01119d | 473 | return __this_address; |
9d85eb44 | 474 | |
3cfabff6 | 475 | mode = be16_to_cpu(dip->di_mode); |
0962b5b3 | 476 | if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN) |
bc01119d | 477 | return __this_address; |
3cfabff6 AG |
478 | |
479 | /* No zero-length symlinks/dirs. */ | |
ad42e5a1 | 480 | if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0) |
bc01119d | 481 | return __this_address; |
9d85eb44 | 482 | |
ad42e5a1 DW |
483 | /* Fork checks carried over from xfs_iformat_fork */ |
484 | if (mode && | |
485 | be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) > | |
486 | be64_to_cpu(dip->di_nblocks)) | |
487 | return __this_address; | |
488 | ||
489 | if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize) | |
490 | return __this_address; | |
491 | ||
492 | flags = be16_to_cpu(dip->di_flags); | |
493 | ||
494 | if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp) | |
495 | return __this_address; | |
496 | ||
91b32a08 ES |
497 | /* check for illegal values of forkoff */ |
498 | fa = xfs_dinode_verify_forkoff(dip, mp); | |
499 | if (fa) | |
500 | return fa; | |
501 | ||
ad42e5a1 DW |
502 | /* Do we have appropriate data fork formats for the mode? */ |
503 | switch (mode & S_IFMT) { | |
504 | case S_IFIFO: | |
505 | case S_IFCHR: | |
506 | case S_IFBLK: | |
507 | case S_IFSOCK: | |
508 | if (dip->di_format != XFS_DINODE_FMT_DEV) | |
509 | return __this_address; | |
510 | break; | |
511 | case S_IFREG: | |
512 | case S_IFLNK: | |
513 | case S_IFDIR: | |
8ac50baf DC |
514 | fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK); |
515 | if (fa) | |
516 | return fa; | |
ad42e5a1 DW |
517 | break; |
518 | case 0: | |
519 | /* Uninitialized inode ok. */ | |
520 | break; | |
521 | default: | |
522 | return __this_address; | |
523 | } | |
524 | ||
525 | if (XFS_DFORK_Q(dip)) { | |
8ac50baf DC |
526 | fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK); |
527 | if (fa) | |
528 | return fa; | |
3060f26a ES |
529 | } else { |
530 | /* | |
531 | * If there is no fork offset, this may be a freshly-made inode | |
532 | * in a new disk cluster, in which case di_aformat is zeroed. | |
533 | * Otherwise, such an inode must be in EXTENTS format; this goes | |
534 | * for freed inodes as well. | |
535 | */ | |
536 | switch (dip->di_aformat) { | |
537 | case 0: | |
538 | case XFS_DINODE_FMT_EXTENTS: | |
539 | break; | |
540 | default: | |
541 | return __this_address; | |
542 | } | |
543 | if (dip->di_anextents) | |
544 | return __this_address; | |
ad42e5a1 DW |
545 | } |
546 | ||
316d5a9f DC |
547 | /* extent size hint validation */ |
548 | fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize), | |
549 | mode, flags); | |
550 | if (fa) | |
551 | return fa; | |
552 | ||
41ce5f36 DC |
553 | /* only version 3 or greater inodes are extensively verified here */ |
554 | if (dip->di_version < 3) | |
bc01119d | 555 | return NULL; |
41ce5f36 | 556 | |
183537ed DW |
557 | flags2 = be64_to_cpu(dip->di_flags2); |
558 | ||
559 | /* don't allow reflink/cowextsize if we don't have reflink */ | |
560 | if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) && | |
316d5a9f | 561 | !xfs_sb_version_hasreflink(&mp->m_sb)) |
bc01119d | 562 | return __this_address; |
183537ed | 563 | |
ad42e5a1 DW |
564 | /* only regular files get reflink */ |
565 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG) | |
566 | return __this_address; | |
567 | ||
183537ed DW |
568 | /* don't let reflink and realtime mix */ |
569 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME)) | |
bc01119d | 570 | return __this_address; |
183537ed | 571 | |
378ed4e7 DW |
572 | /* don't let reflink and dax mix */ |
573 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX)) | |
bc01119d | 574 | return __this_address; |
378ed4e7 | 575 | |
77e17d58 DC |
576 | /* COW extent size hint validation */ |
577 | fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize), | |
578 | mode, flags, flags2); | |
579 | if (fa) | |
580 | return fa; | |
581 | ||
bc01119d | 582 | return NULL; |
41ce5f36 | 583 | } |
e6d77a21 | 584 | |
41ce5f36 DC |
585 | void |
586 | xfs_dinode_calc_crc( | |
587 | struct xfs_mount *mp, | |
588 | struct xfs_dinode *dip) | |
589 | { | |
4a492e72 | 590 | uint32_t crc; |
41ce5f36 DC |
591 | |
592 | if (dip->di_version < 3) | |
593 | return; | |
594 | ||
595 | ASSERT(xfs_sb_version_hascrc(&mp->m_sb)); | |
0bb90214 | 596 | crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize, |
3faddc54 | 597 | XFS_DINODE_CRC_OFF); |
41ce5f36 | 598 | dip->di_crc = xfs_end_cksum(crc); |
2bd0ea18 NS |
599 | } |
600 | ||
a2ceac1f DC |
601 | /* |
602 | * Read the disk inode attributes into the in-core inode structure. | |
ff105f75 DC |
603 | * |
604 | * For version 5 superblocks, if we are initialising a new inode and we are not | |
605 | * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new | |
606 | * inode core with a random generation number. If we are keeping inodes around, | |
607 | * we need to read the inode cluster to get the existing generation number off | |
608 | * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode | |
609 | * format) then log recovery is dependent on the di_flushiter field being | |
610 | * initialised from the current on-disk value and hence we must also read the | |
611 | * inode off disk. | |
a2ceac1f DC |
612 | */ |
613 | int | |
614 | xfs_iread( | |
615 | xfs_mount_t *mp, | |
616 | xfs_trans_t *tp, | |
617 | xfs_inode_t *ip, | |
618 | uint iget_flags) | |
619 | { | |
620 | xfs_buf_t *bp; | |
621 | xfs_dinode_t *dip; | |
bc01119d | 622 | xfs_failaddr_t fa; |
a2ceac1f DC |
623 | int error; |
624 | ||
625 | /* | |
626 | * Fill in the location information in the in-core inode. | |
627 | */ | |
628 | error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags); | |
629 | if (error) | |
630 | return error; | |
631 | ||
ff105f75 DC |
632 | /* shortcut IO on inode allocation if possible */ |
633 | if ((iget_flags & XFS_IGET_CREATE) && | |
634 | xfs_sb_version_hascrc(&mp->m_sb) && | |
635 | !(mp->m_flags & XFS_MOUNT_IKEEP)) { | |
636 | /* initialise the on-disk inode core */ | |
637 | memset(&ip->i_d, 0, sizeof(ip->i_d)); | |
6652c253 | 638 | VFS_I(ip)->i_generation = prandom_u32(); |
d206fcb7 | 639 | ip->i_d.di_version = 3; |
ff105f75 DC |
640 | return 0; |
641 | } | |
642 | ||
a2ceac1f DC |
643 | /* |
644 | * Get pointers to the on-disk inode and the buffer containing it. | |
645 | */ | |
646 | error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags); | |
647 | if (error) | |
648 | return error; | |
649 | ||
41ce5f36 | 650 | /* even unallocated inodes are verified */ |
bc01119d DW |
651 | fa = xfs_dinode_verify(mp, ip->i_ino, dip); |
652 | if (fa) { | |
1d3bac1f DW |
653 | xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip, |
654 | sizeof(*dip), fa); | |
12b53197 | 655 | error = -EFSCORRUPTED; |
a2ceac1f DC |
656 | goto out_brelse; |
657 | } | |
658 | ||
659 | /* | |
660 | * If the on-disk inode is already linked to a directory | |
661 | * entry, copy all of the inode into the in-core inode. | |
5d90ab5a | 662 | * xfs_iformat_fork() handles copying in the inode format |
a2ceac1f DC |
663 | * specific information. |
664 | * Otherwise, just get the truly permanent information. | |
665 | */ | |
666 | if (dip->di_mode) { | |
1bc6cbe3 | 667 | xfs_inode_from_disk(ip, dip); |
5d90ab5a | 668 | error = xfs_iformat_fork(ip, dip); |
a2ceac1f DC |
669 | if (error) { |
670 | #ifdef DEBUG | |
671 | xfs_alert(mp, "%s: xfs_iformat() returned error %d", | |
672 | __func__, error); | |
673 | #endif /* DEBUG */ | |
674 | goto out_brelse; | |
675 | } | |
676 | } else { | |
41ce5f36 DC |
677 | /* |
678 | * Partial initialisation of the in-core inode. Just the bits | |
679 | * that xfs_ialloc won't overwrite or relies on being correct. | |
680 | */ | |
a2ceac1f | 681 | ip->i_d.di_version = dip->di_version; |
6652c253 | 682 | VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen); |
a2ceac1f | 683 | ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter); |
41ce5f36 | 684 | |
a2ceac1f DC |
685 | /* |
686 | * Make sure to pull in the mode here as well in | |
687 | * case the inode is released without being used. | |
688 | * This ensures that xfs_inactive() will see that | |
689 | * the inode is already free and not try to mess | |
690 | * with the uninitialized part of it. | |
691 | */ | |
e37bf53c | 692 | VFS_I(ip)->i_mode = 0; |
a2ceac1f DC |
693 | } |
694 | ||
f089fc42 | 695 | ASSERT(ip->i_d.di_version >= 2); |
a2ceac1f DC |
696 | ip->i_delayed_blks = 0; |
697 | ||
698 | /* | |
699 | * Mark the buffer containing the inode as something to keep | |
700 | * around for a while. This helps to keep recently accessed | |
701 | * meta-data in-core longer. | |
702 | */ | |
703 | xfs_buf_set_ref(bp, XFS_INO_REF); | |
704 | ||
705 | /* | |
1d519883 DC |
706 | * Use xfs_trans_brelse() to release the buffer containing the on-disk |
707 | * inode, because it was acquired with xfs_trans_read_buf() in | |
708 | * xfs_imap_to_bp() above. If tp is NULL, this is just a normal | |
a2ceac1f DC |
709 | * brelse(). If we're within a transaction, then xfs_trans_brelse() |
710 | * will only release the buffer if it is not dirty within the | |
711 | * transaction. It will be OK to release the buffer in this case, | |
1d519883 DC |
712 | * because inodes on disk are never destroyed and we will be locking the |
713 | * new in-core inode before putting it in the cache where other | |
714 | * processes can find it. Thus we don't have to worry about the inode | |
715 | * being changed just because we released the buffer. | |
a2ceac1f DC |
716 | */ |
717 | out_brelse: | |
718 | xfs_trans_brelse(tp, bp); | |
719 | return error; | |
720 | } | |
fd0c360d DW |
721 | |
722 | /* | |
723 | * Validate di_extsize hint. | |
724 | * | |
725 | * The rules are documented at xfs_ioctl_setattr_check_extsize(). | |
726 | * These functions must be kept in sync with each other. | |
727 | */ | |
728 | xfs_failaddr_t | |
729 | xfs_inode_validate_extsize( | |
730 | struct xfs_mount *mp, | |
731 | uint32_t extsize, | |
732 | uint16_t mode, | |
733 | uint16_t flags) | |
734 | { | |
735 | bool rt_flag; | |
736 | bool hint_flag; | |
737 | bool inherit_flag; | |
738 | uint32_t extsize_bytes; | |
739 | uint32_t blocksize_bytes; | |
740 | ||
741 | rt_flag = (flags & XFS_DIFLAG_REALTIME); | |
742 | hint_flag = (flags & XFS_DIFLAG_EXTSIZE); | |
743 | inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT); | |
744 | extsize_bytes = XFS_FSB_TO_B(mp, extsize); | |
745 | ||
746 | if (rt_flag) | |
747 | blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog; | |
748 | else | |
749 | blocksize_bytes = mp->m_sb.sb_blocksize; | |
750 | ||
751 | if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode))) | |
752 | return __this_address; | |
753 | ||
754 | if (hint_flag && !S_ISREG(mode)) | |
755 | return __this_address; | |
756 | ||
757 | if (inherit_flag && !S_ISDIR(mode)) | |
758 | return __this_address; | |
759 | ||
760 | if ((hint_flag || inherit_flag) && extsize == 0) | |
761 | return __this_address; | |
762 | ||
206bef93 ES |
763 | /* free inodes get flags set to zero but extsize remains */ |
764 | if (mode && !(hint_flag || inherit_flag) && extsize != 0) | |
fd0c360d DW |
765 | return __this_address; |
766 | ||
767 | if (extsize_bytes % blocksize_bytes) | |
768 | return __this_address; | |
769 | ||
770 | if (extsize > MAXEXTLEN) | |
771 | return __this_address; | |
772 | ||
773 | if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2) | |
774 | return __this_address; | |
775 | ||
776 | return NULL; | |
777 | } | |
778 | ||
779 | /* | |
780 | * Validate di_cowextsize hint. | |
781 | * | |
782 | * The rules are documented at xfs_ioctl_setattr_check_cowextsize(). | |
783 | * These functions must be kept in sync with each other. | |
784 | */ | |
785 | xfs_failaddr_t | |
786 | xfs_inode_validate_cowextsize( | |
787 | struct xfs_mount *mp, | |
788 | uint32_t cowextsize, | |
789 | uint16_t mode, | |
790 | uint16_t flags, | |
791 | uint64_t flags2) | |
792 | { | |
793 | bool rt_flag; | |
794 | bool hint_flag; | |
795 | uint32_t cowextsize_bytes; | |
796 | ||
797 | rt_flag = (flags & XFS_DIFLAG_REALTIME); | |
798 | hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE); | |
799 | cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize); | |
800 | ||
801 | if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb)) | |
802 | return __this_address; | |
803 | ||
804 | if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode))) | |
805 | return __this_address; | |
806 | ||
807 | if (hint_flag && cowextsize == 0) | |
808 | return __this_address; | |
809 | ||
206bef93 ES |
810 | /* free inodes get flags set to zero but cowextsize remains */ |
811 | if (mode && !hint_flag && cowextsize != 0) | |
fd0c360d DW |
812 | return __this_address; |
813 | ||
814 | if (hint_flag && rt_flag) | |
815 | return __this_address; | |
816 | ||
817 | if (cowextsize_bytes % mp->m_sb.sb_blocksize) | |
818 | return __this_address; | |
819 | ||
820 | if (cowextsize > MAXEXTLEN) | |
821 | return __this_address; | |
822 | ||
823 | if (cowextsize > mp->m_sb.sb_agblocks / 2) | |
824 | return __this_address; | |
825 | ||
826 | return NULL; | |
827 | } |