]>
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" | |
13 | #include "xfs_inode.h" | |
2cf10e4c | 14 | #include "xfs_errortag.h" |
b626fb59 DC |
15 | #include "xfs_trans.h" |
16 | #include "xfs_ialloc.h" | |
0962b5b3 | 17 | #include "xfs_dir2.h" |
2bd0ea18 | 18 | |
998aed52 | 19 | |
5e656dbb BN |
20 | /* |
21 | * Check that none of the inode's in the buffer have a next | |
22 | * unlinked field of 0. | |
23 | */ | |
24 | #if defined(DEBUG) | |
2bd0ea18 NS |
25 | void |
26 | xfs_inobp_check( | |
27 | xfs_mount_t *mp, | |
28 | xfs_buf_t *bp) | |
29 | { | |
30 | int i; | |
2bd0ea18 NS |
31 | xfs_dinode_t *dip; |
32 | ||
4b0fd0f4 | 33 | for (i = 0; i < M_IGEO(mp)->inodes_per_cluster; i++) { |
92acb899 | 34 | dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize); |
46eca962 | 35 | if (!dip->di_next_unlinked) { |
a2ceac1f | 36 | xfs_alert(mp, |
ea6a00d4 DC |
37 | "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.", |
38 | i, (long long)bp->b_bn); | |
2bd0ea18 NS |
39 | } |
40 | } | |
41 | } | |
42 | #endif | |
43 | ||
f85fc622 DC |
44 | /* |
45 | * If we are doing readahead on an inode buffer, we might be in log recovery | |
46 | * reading an inode allocation buffer that hasn't yet been replayed, and hence | |
47 | * has not had the inode cores stamped into it. Hence for readahead, the buffer | |
48 | * may be potentially invalid. | |
49 | * | |
736e1cf6 DC |
50 | * If the readahead buffer is invalid, we need to mark it with an error and |
51 | * clear the DONE status of the buffer so that a followup read will re-read it | |
52 | * from disk. We don't report the error otherwise to avoid warnings during log | |
53 | * recovery and we don't get unnecssary panics on debug kernels. We use EIO here | |
54 | * because all we want to do is say readahead failed; there is no-one to report | |
55 | * the error to, so this will distinguish it from a non-ra verifier failure. | |
6daba42a DC |
56 | * Changes to this readahead error behavour also need to be reflected in |
57 | * xfs_dquot_buf_readahead_verify(). | |
f85fc622 | 58 | */ |
a2ceac1f DC |
59 | static void |
60 | xfs_inode_buf_verify( | |
f85fc622 DC |
61 | struct xfs_buf *bp, |
62 | bool readahead) | |
5e656dbb | 63 | { |
7861ef77 | 64 | struct xfs_mount *mp = bp->b_mount; |
2949b467 | 65 | xfs_agnumber_t agno; |
5e656dbb BN |
66 | int i; |
67 | int ni; | |
5e656dbb BN |
68 | |
69 | /* | |
70 | * Validate the magic number and version of every inode in the buffer | |
5e656dbb | 71 | */ |
2949b467 | 72 | agno = xfs_daddr_to_agno(mp, XFS_BUF_ADDR(bp)); |
a2ceac1f | 73 | ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock; |
5e656dbb BN |
74 | for (i = 0; i < ni; i++) { |
75 | int di_ok; | |
76 | xfs_dinode_t *dip; | |
2949b467 | 77 | xfs_agino_t unlinked_ino; |
5e656dbb | 78 | |
92acb899 | 79 | dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog)); |
2949b467 | 80 | unlinked_ino = be32_to_cpu(dip->di_next_unlinked); |
9e26de8f | 81 | di_ok = xfs_verify_magic16(bp, dip->di_magic) && |
db84c7e8 | 82 | xfs_dinode_good_version(&mp->m_sb, dip->di_version) && |
7a425ea8 | 83 | xfs_verify_agino_or_null(mp, agno, unlinked_ino); |
5e656dbb | 84 | if (unlikely(XFS_TEST_ERROR(!di_ok, mp, |
e2a190dd | 85 | XFS_ERRTAG_ITOBP_INOTOBP))) { |
f85fc622 DC |
86 | if (readahead) { |
87 | bp->b_flags &= ~XBF_DONE; | |
736e1cf6 | 88 | xfs_buf_ioerror(bp, -EIO); |
f85fc622 DC |
89 | return; |
90 | } | |
91 | ||
5e656dbb | 92 | #ifdef DEBUG |
ea6a00d4 | 93 | xfs_alert(mp, |
a2ceac1f DC |
94 | "bad inode magic/vsn daddr %lld #%d (magic=%x)", |
95 | (unsigned long long)bp->b_bn, i, | |
56b2de80 | 96 | be16_to_cpu(dip->di_magic)); |
5e656dbb | 97 | #endif |
43f7ceb5 DW |
98 | xfs_buf_verifier_error(bp, -EFSCORRUPTED, |
99 | __func__, dip, sizeof(*dip), | |
100 | NULL); | |
2949b467 | 101 | return; |
5e656dbb BN |
102 | } |
103 | } | |
a2ceac1f | 104 | } |
5e656dbb | 105 | |
e6d77a21 | 106 | |
a2ceac1f DC |
107 | static void |
108 | xfs_inode_buf_read_verify( | |
109 | struct xfs_buf *bp) | |
110 | { | |
f85fc622 DC |
111 | xfs_inode_buf_verify(bp, false); |
112 | } | |
113 | ||
114 | static void | |
115 | xfs_inode_buf_readahead_verify( | |
116 | struct xfs_buf *bp) | |
117 | { | |
118 | xfs_inode_buf_verify(bp, true); | |
5e656dbb | 119 | } |
2bd0ea18 | 120 | |
a2ceac1f DC |
121 | static void |
122 | xfs_inode_buf_write_verify( | |
123 | struct xfs_buf *bp) | |
56b2de80 | 124 | { |
f85fc622 | 125 | xfs_inode_buf_verify(bp, false); |
56b2de80 DC |
126 | } |
127 | ||
a2ceac1f | 128 | const struct xfs_buf_ops xfs_inode_buf_ops = { |
a3fac935 | 129 | .name = "xfs_inode", |
9e26de8f DW |
130 | .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC), |
131 | cpu_to_be16(XFS_DINODE_MAGIC) }, | |
a2ceac1f DC |
132 | .verify_read = xfs_inode_buf_read_verify, |
133 | .verify_write = xfs_inode_buf_write_verify, | |
134 | }; | |
135 | ||
f85fc622 | 136 | const struct xfs_buf_ops xfs_inode_buf_ra_ops = { |
7563fc9c | 137 | .name = "xfs_inode_ra", |
9e26de8f DW |
138 | .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC), |
139 | cpu_to_be16(XFS_DINODE_MAGIC) }, | |
f85fc622 DC |
140 | .verify_read = xfs_inode_buf_readahead_verify, |
141 | .verify_write = xfs_inode_buf_write_verify, | |
142 | }; | |
143 | ||
144 | ||
2bd0ea18 | 145 | /* |
a2ceac1f DC |
146 | * This routine is called to map an inode to the buffer containing the on-disk |
147 | * version of the inode. It returns a pointer to the buffer containing the | |
148 | * on-disk inode in the bpp parameter, and in the dipp parameter it returns a | |
149 | * pointer to the on-disk inode within that buffer. | |
2bd0ea18 | 150 | * |
a2ceac1f DC |
151 | * If a non-zero error is returned, then the contents of bpp and dipp are |
152 | * undefined. | |
2bd0ea18 NS |
153 | */ |
154 | int | |
a2ceac1f DC |
155 | xfs_imap_to_bp( |
156 | struct xfs_mount *mp, | |
157 | struct xfs_trans *tp, | |
158 | struct xfs_imap *imap, | |
159 | struct xfs_dinode **dipp, | |
160 | struct xfs_buf **bpp, | |
0cd4b54a | 161 | uint buf_flags) |
2bd0ea18 | 162 | { |
a2ceac1f DC |
163 | struct xfs_buf *bp; |
164 | int error; | |
2bd0ea18 | 165 | |
a2ceac1f DC |
166 | buf_flags |= XBF_UNMAPPED; |
167 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno, | |
168 | (int)imap->im_len, buf_flags, &bp, | |
169 | &xfs_inode_buf_ops); | |
170 | if (error) { | |
820c6548 | 171 | ASSERT(error != -EAGAIN || (buf_flags & XBF_TRYLOCK)); |
a2ceac1f | 172 | return error; |
2bd0ea18 | 173 | } |
2bd0ea18 | 174 | |
2bd0ea18 | 175 | *bpp = bp; |
92acb899 | 176 | *dipp = xfs_buf_offset(bp, imap->im_boffset); |
2bd0ea18 NS |
177 | return 0; |
178 | } | |
179 | ||
3eb685ae | 180 | int |
1bc6cbe3 DC |
181 | xfs_inode_from_disk( |
182 | struct xfs_inode *ip, | |
138659f1 | 183 | struct xfs_dinode *from) |
2bd0ea18 | 184 | { |
1bc6cbe3 DC |
185 | struct xfs_icdinode *to = &ip->i_d; |
186 | struct inode *inode = VFS_I(ip); | |
187 | ||
f089fc42 DC |
188 | /* |
189 | * Convert v1 inodes immediately to v2 inode format as this is the | |
190 | * minimum inode version format we support in the rest of the code. | |
fd82a87e | 191 | * They will also be unconditionally written back to disk as v2 inodes. |
f089fc42 | 192 | */ |
fd82a87e | 193 | if (unlikely(from->di_version == 1)) { |
bcbe04c1 | 194 | set_nlink(inode, be16_to_cpu(from->di_onlink)); |
e6bd76f0 | 195 | to->di_projid = 0; |
f089fc42 | 196 | } else { |
bcbe04c1 | 197 | set_nlink(inode, be32_to_cpu(from->di_nlink)); |
e6bd76f0 CH |
198 | to->di_projid = (prid_t)be16_to_cpu(from->di_projid_hi) << 16 | |
199 | be16_to_cpu(from->di_projid_lo); | |
f089fc42 DC |
200 | } |
201 | ||
5e656dbb | 202 | to->di_format = from->di_format; |
100eb9e9 CH |
203 | i_uid_write(inode, be32_to_cpu(from->di_uid)); |
204 | i_gid_write(inode, be32_to_cpu(from->di_gid)); | |
5e656dbb | 205 | to->di_flushiter = be16_to_cpu(from->di_flushiter); |
1bc6cbe3 DC |
206 | |
207 | /* | |
208 | * Time is signed, so need to convert to signed 32 bit before | |
209 | * storing in inode timestamp which may be 64 bit. Otherwise | |
210 | * a time before epoch is converted to a time long after epoch | |
211 | * on 64 bit systems. | |
212 | */ | |
213 | inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec); | |
214 | inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec); | |
215 | inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec); | |
216 | inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec); | |
217 | inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec); | |
218 | inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec); | |
6652c253 | 219 | inode->i_generation = be32_to_cpu(from->di_gen); |
e37bf53c | 220 | inode->i_mode = be16_to_cpu(from->di_mode); |
1bc6cbe3 | 221 | |
5e656dbb BN |
222 | to->di_size = be64_to_cpu(from->di_size); |
223 | to->di_nblocks = be64_to_cpu(from->di_nblocks); | |
224 | to->di_extsize = be32_to_cpu(from->di_extsize); | |
225 | to->di_nextents = be32_to_cpu(from->di_nextents); | |
226 | to->di_anextents = be16_to_cpu(from->di_anextents); | |
227 | to->di_forkoff = from->di_forkoff; | |
228 | to->di_aformat = from->di_aformat; | |
229 | to->di_dmevmask = be32_to_cpu(from->di_dmevmask); | |
230 | to->di_dmstate = be16_to_cpu(from->di_dmstate); | |
231 | to->di_flags = be16_to_cpu(from->di_flags); | |
41ce5f36 | 232 | |
fd82a87e | 233 | if (xfs_sb_version_has_v3inode(&ip->i_mount->m_sb)) { |
d5105a89 JL |
234 | inode_set_iversion_queried(inode, |
235 | be64_to_cpu(from->di_changecount)); | |
cb49e9a4 CH |
236 | to->di_crtime.tv_sec = be32_to_cpu(from->di_crtime.t_sec); |
237 | to->di_crtime.tv_nsec = be32_to_cpu(from->di_crtime.t_nsec); | |
41ce5f36 | 238 | to->di_flags2 = be64_to_cpu(from->di_flags2); |
10460994 | 239 | to->di_cowextsize = be32_to_cpu(from->di_cowextsize); |
41ce5f36 | 240 | } |
3eb685ae CH |
241 | |
242 | return xfs_iformat_fork(ip, from); | |
5e656dbb BN |
243 | } |
244 | ||
245 | void | |
1bc6cbe3 DC |
246 | xfs_inode_to_disk( |
247 | struct xfs_inode *ip, | |
db17aebe DC |
248 | struct xfs_dinode *to, |
249 | xfs_lsn_t lsn) | |
1bc6cbe3 DC |
250 | { |
251 | struct xfs_icdinode *from = &ip->i_d; | |
252 | struct inode *inode = VFS_I(ip); | |
253 | ||
db17aebe | 254 | to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); |
f089fc42 | 255 | to->di_onlink = 0; |
db17aebe | 256 | |
1bc6cbe3 | 257 | to->di_format = from->di_format; |
100eb9e9 CH |
258 | to->di_uid = cpu_to_be32(i_uid_read(inode)); |
259 | to->di_gid = cpu_to_be32(i_gid_read(inode)); | |
e6bd76f0 CH |
260 | to->di_projid_lo = cpu_to_be16(from->di_projid & 0xffff); |
261 | to->di_projid_hi = cpu_to_be16(from->di_projid >> 16); | |
1bc6cbe3 | 262 | |
db17aebe | 263 | memset(to->di_pad, 0, sizeof(to->di_pad)); |
1bc6cbe3 DC |
264 | to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec); |
265 | to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec); | |
266 | to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec); | |
267 | to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec); | |
268 | to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec); | |
269 | to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec); | |
bcbe04c1 | 270 | to->di_nlink = cpu_to_be32(inode->i_nlink); |
6652c253 | 271 | to->di_gen = cpu_to_be32(inode->i_generation); |
e37bf53c | 272 | to->di_mode = cpu_to_be16(inode->i_mode); |
1bc6cbe3 DC |
273 | |
274 | to->di_size = cpu_to_be64(from->di_size); | |
275 | to->di_nblocks = cpu_to_be64(from->di_nblocks); | |
276 | to->di_extsize = cpu_to_be32(from->di_extsize); | |
277 | to->di_nextents = cpu_to_be32(from->di_nextents); | |
278 | to->di_anextents = cpu_to_be16(from->di_anextents); | |
279 | to->di_forkoff = from->di_forkoff; | |
280 | to->di_aformat = from->di_aformat; | |
281 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); | |
282 | to->di_dmstate = cpu_to_be16(from->di_dmstate); | |
283 | to->di_flags = cpu_to_be16(from->di_flags); | |
1bc6cbe3 | 284 | |
fd82a87e CH |
285 | if (xfs_sb_version_has_v3inode(&ip->i_mount->m_sb)) { |
286 | to->di_version = 3; | |
d5105a89 | 287 | to->di_changecount = cpu_to_be64(inode_peek_iversion(inode)); |
cb49e9a4 CH |
288 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.tv_sec); |
289 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.tv_nsec); | |
1bc6cbe3 | 290 | to->di_flags2 = cpu_to_be64(from->di_flags2); |
10460994 | 291 | to->di_cowextsize = cpu_to_be32(from->di_cowextsize); |
db17aebe DC |
292 | to->di_ino = cpu_to_be64(ip->i_ino); |
293 | to->di_lsn = cpu_to_be64(lsn); | |
294 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); | |
295 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); | |
1bc6cbe3 DC |
296 | to->di_flushiter = 0; |
297 | } else { | |
fd82a87e | 298 | to->di_version = 2; |
1bc6cbe3 DC |
299 | to->di_flushiter = cpu_to_be16(from->di_flushiter); |
300 | } | |
301 | } | |
302 | ||
303 | void | |
304 | xfs_log_dinode_to_disk( | |
305 | struct xfs_log_dinode *from, | |
306 | struct xfs_dinode *to) | |
5e656dbb BN |
307 | { |
308 | to->di_magic = cpu_to_be16(from->di_magic); | |
309 | to->di_mode = cpu_to_be16(from->di_mode); | |
f089fc42 | 310 | to->di_version = from->di_version; |
5e656dbb | 311 | to->di_format = from->di_format; |
f089fc42 | 312 | to->di_onlink = 0; |
5e656dbb BN |
313 | to->di_uid = cpu_to_be32(from->di_uid); |
314 | to->di_gid = cpu_to_be32(from->di_gid); | |
315 | to->di_nlink = cpu_to_be32(from->di_nlink); | |
22bc10ed AM |
316 | to->di_projid_lo = cpu_to_be16(from->di_projid_lo); |
317 | to->di_projid_hi = cpu_to_be16(from->di_projid_hi); | |
5e656dbb | 318 | memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); |
1bc6cbe3 | 319 | |
5e656dbb BN |
320 | to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec); |
321 | to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec); | |
322 | to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec); | |
323 | to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec); | |
324 | to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec); | |
325 | to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec); | |
1bc6cbe3 | 326 | |
5e656dbb BN |
327 | to->di_size = cpu_to_be64(from->di_size); |
328 | to->di_nblocks = cpu_to_be64(from->di_nblocks); | |
329 | to->di_extsize = cpu_to_be32(from->di_extsize); | |
330 | to->di_nextents = cpu_to_be32(from->di_nextents); | |
331 | to->di_anextents = cpu_to_be16(from->di_anextents); | |
332 | to->di_forkoff = from->di_forkoff; | |
333 | to->di_aformat = from->di_aformat; | |
334 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); | |
335 | to->di_dmstate = cpu_to_be16(from->di_dmstate); | |
336 | to->di_flags = cpu_to_be16(from->di_flags); | |
337 | to->di_gen = cpu_to_be32(from->di_gen); | |
41ce5f36 DC |
338 | |
339 | if (from->di_version == 3) { | |
340 | to->di_changecount = cpu_to_be64(from->di_changecount); | |
341 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec); | |
342 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec); | |
343 | to->di_flags2 = cpu_to_be64(from->di_flags2); | |
10460994 | 344 | to->di_cowextsize = cpu_to_be32(from->di_cowextsize); |
41ce5f36 DC |
345 | to->di_ino = cpu_to_be64(from->di_ino); |
346 | to->di_lsn = cpu_to_be64(from->di_lsn); | |
347 | memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2)); | |
f85fc622 | 348 | uuid_copy(&to->di_uuid, &from->di_uuid); |
fde71651 DC |
349 | to->di_flushiter = 0; |
350 | } else { | |
351 | to->di_flushiter = cpu_to_be16(from->di_flushiter); | |
41ce5f36 DC |
352 | } |
353 | } | |
354 | ||
8ac50baf DC |
355 | static xfs_failaddr_t |
356 | xfs_dinode_verify_fork( | |
357 | struct xfs_dinode *dip, | |
358 | struct xfs_mount *mp, | |
359 | int whichfork) | |
360 | { | |
361 | uint32_t di_nextents = XFS_DFORK_NEXTENTS(dip, whichfork); | |
362 | ||
363 | switch (XFS_DFORK_FORMAT(dip, whichfork)) { | |
364 | case XFS_DINODE_FMT_LOCAL: | |
365 | /* | |
366 | * no local regular files yet | |
367 | */ | |
368 | if (whichfork == XFS_DATA_FORK) { | |
369 | if (S_ISREG(be16_to_cpu(dip->di_mode))) | |
370 | return __this_address; | |
371 | if (be64_to_cpu(dip->di_size) > | |
372 | XFS_DFORK_SIZE(dip, mp, whichfork)) | |
373 | return __this_address; | |
374 | } | |
375 | if (di_nextents) | |
376 | return __this_address; | |
377 | break; | |
378 | case XFS_DINODE_FMT_EXTENTS: | |
379 | if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork)) | |
380 | return __this_address; | |
381 | break; | |
382 | case XFS_DINODE_FMT_BTREE: | |
383 | if (whichfork == XFS_ATTR_FORK) { | |
384 | if (di_nextents > MAXAEXTNUM) | |
385 | return __this_address; | |
386 | } else if (di_nextents > MAXEXTNUM) { | |
387 | return __this_address; | |
388 | } | |
389 | break; | |
390 | default: | |
391 | return __this_address; | |
392 | } | |
393 | return NULL; | |
394 | } | |
395 | ||
91b32a08 ES |
396 | static xfs_failaddr_t |
397 | xfs_dinode_verify_forkoff( | |
398 | struct xfs_dinode *dip, | |
399 | struct xfs_mount *mp) | |
400 | { | |
401 | if (!XFS_DFORK_Q(dip)) | |
402 | return NULL; | |
403 | ||
404 | switch (dip->di_format) { | |
405 | case XFS_DINODE_FMT_DEV: | |
406 | if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3)) | |
407 | return __this_address; | |
408 | break; | |
409 | case XFS_DINODE_FMT_LOCAL: /* fall through ... */ | |
410 | case XFS_DINODE_FMT_EXTENTS: /* fall through ... */ | |
411 | case XFS_DINODE_FMT_BTREE: | |
4de63245 | 412 | if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3)) |
91b32a08 ES |
413 | return __this_address; |
414 | break; | |
415 | default: | |
416 | return __this_address; | |
417 | } | |
418 | return NULL; | |
419 | } | |
420 | ||
bc01119d | 421 | xfs_failaddr_t |
41ce5f36 DC |
422 | xfs_dinode_verify( |
423 | struct xfs_mount *mp, | |
e515cca1 | 424 | xfs_ino_t ino, |
41ce5f36 DC |
425 | struct xfs_dinode *dip) |
426 | { | |
316d5a9f | 427 | xfs_failaddr_t fa; |
3cfabff6 | 428 | uint16_t mode; |
183537ed DW |
429 | uint16_t flags; |
430 | uint64_t flags2; | |
ad42e5a1 | 431 | uint64_t di_size; |
183537ed | 432 | |
41ce5f36 | 433 | if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC)) |
bc01119d | 434 | return __this_address; |
41ce5f36 | 435 | |
98703400 DW |
436 | /* Verify v3 integrity information first */ |
437 | if (dip->di_version >= 3) { | |
db84c7e8 | 438 | if (!xfs_sb_version_has_v3inode(&mp->m_sb)) |
98703400 DW |
439 | return __this_address; |
440 | if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize, | |
441 | XFS_DINODE_CRC_OFF)) | |
442 | return __this_address; | |
443 | if (be64_to_cpu(dip->di_ino) != ino) | |
444 | return __this_address; | |
445 | if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid)) | |
446 | return __this_address; | |
447 | } | |
448 | ||
9d85eb44 | 449 | /* don't allow invalid i_size */ |
ad42e5a1 DW |
450 | di_size = be64_to_cpu(dip->di_size); |
451 | if (di_size & (1ULL << 63)) | |
bc01119d | 452 | return __this_address; |
9d85eb44 | 453 | |
3cfabff6 | 454 | mode = be16_to_cpu(dip->di_mode); |
0962b5b3 | 455 | if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN) |
bc01119d | 456 | return __this_address; |
3cfabff6 AG |
457 | |
458 | /* No zero-length symlinks/dirs. */ | |
ad42e5a1 | 459 | if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0) |
bc01119d | 460 | return __this_address; |
9d85eb44 | 461 | |
ad42e5a1 DW |
462 | /* Fork checks carried over from xfs_iformat_fork */ |
463 | if (mode && | |
464 | be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) > | |
465 | be64_to_cpu(dip->di_nblocks)) | |
466 | return __this_address; | |
467 | ||
468 | if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize) | |
469 | return __this_address; | |
470 | ||
471 | flags = be16_to_cpu(dip->di_flags); | |
472 | ||
473 | if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp) | |
474 | return __this_address; | |
475 | ||
91b32a08 ES |
476 | /* check for illegal values of forkoff */ |
477 | fa = xfs_dinode_verify_forkoff(dip, mp); | |
478 | if (fa) | |
479 | return fa; | |
480 | ||
ad42e5a1 DW |
481 | /* Do we have appropriate data fork formats for the mode? */ |
482 | switch (mode & S_IFMT) { | |
483 | case S_IFIFO: | |
484 | case S_IFCHR: | |
485 | case S_IFBLK: | |
486 | case S_IFSOCK: | |
487 | if (dip->di_format != XFS_DINODE_FMT_DEV) | |
488 | return __this_address; | |
489 | break; | |
490 | case S_IFREG: | |
491 | case S_IFLNK: | |
492 | case S_IFDIR: | |
8ac50baf DC |
493 | fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK); |
494 | if (fa) | |
495 | return fa; | |
ad42e5a1 DW |
496 | break; |
497 | case 0: | |
498 | /* Uninitialized inode ok. */ | |
499 | break; | |
500 | default: | |
501 | return __this_address; | |
502 | } | |
503 | ||
504 | if (XFS_DFORK_Q(dip)) { | |
8ac50baf DC |
505 | fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK); |
506 | if (fa) | |
507 | return fa; | |
3060f26a ES |
508 | } else { |
509 | /* | |
510 | * If there is no fork offset, this may be a freshly-made inode | |
511 | * in a new disk cluster, in which case di_aformat is zeroed. | |
512 | * Otherwise, such an inode must be in EXTENTS format; this goes | |
513 | * for freed inodes as well. | |
514 | */ | |
515 | switch (dip->di_aformat) { | |
516 | case 0: | |
517 | case XFS_DINODE_FMT_EXTENTS: | |
518 | break; | |
519 | default: | |
520 | return __this_address; | |
521 | } | |
522 | if (dip->di_anextents) | |
523 | return __this_address; | |
ad42e5a1 DW |
524 | } |
525 | ||
316d5a9f DC |
526 | /* extent size hint validation */ |
527 | fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize), | |
528 | mode, flags); | |
529 | if (fa) | |
530 | return fa; | |
531 | ||
41ce5f36 DC |
532 | /* only version 3 or greater inodes are extensively verified here */ |
533 | if (dip->di_version < 3) | |
bc01119d | 534 | return NULL; |
41ce5f36 | 535 | |
183537ed DW |
536 | flags2 = be64_to_cpu(dip->di_flags2); |
537 | ||
538 | /* don't allow reflink/cowextsize if we don't have reflink */ | |
539 | if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) && | |
316d5a9f | 540 | !xfs_sb_version_hasreflink(&mp->m_sb)) |
bc01119d | 541 | return __this_address; |
183537ed | 542 | |
ad42e5a1 DW |
543 | /* only regular files get reflink */ |
544 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG) | |
545 | return __this_address; | |
546 | ||
183537ed DW |
547 | /* don't let reflink and realtime mix */ |
548 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME)) | |
bc01119d | 549 | return __this_address; |
183537ed | 550 | |
378ed4e7 DW |
551 | /* don't let reflink and dax mix */ |
552 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX)) | |
bc01119d | 553 | return __this_address; |
378ed4e7 | 554 | |
77e17d58 DC |
555 | /* COW extent size hint validation */ |
556 | fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize), | |
557 | mode, flags, flags2); | |
558 | if (fa) | |
559 | return fa; | |
560 | ||
bc01119d | 561 | return NULL; |
41ce5f36 | 562 | } |
e6d77a21 | 563 | |
41ce5f36 DC |
564 | void |
565 | xfs_dinode_calc_crc( | |
566 | struct xfs_mount *mp, | |
567 | struct xfs_dinode *dip) | |
568 | { | |
4a492e72 | 569 | uint32_t crc; |
41ce5f36 DC |
570 | |
571 | if (dip->di_version < 3) | |
572 | return; | |
573 | ||
574 | ASSERT(xfs_sb_version_hascrc(&mp->m_sb)); | |
0bb90214 | 575 | crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize, |
3faddc54 | 576 | XFS_DINODE_CRC_OFF); |
41ce5f36 | 577 | dip->di_crc = xfs_end_cksum(crc); |
2bd0ea18 NS |
578 | } |
579 | ||
a2ceac1f DC |
580 | /* |
581 | * Read the disk inode attributes into the in-core inode structure. | |
ff105f75 DC |
582 | * |
583 | * For version 5 superblocks, if we are initialising a new inode and we are not | |
584 | * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new | |
585 | * inode core with a random generation number. If we are keeping inodes around, | |
586 | * we need to read the inode cluster to get the existing generation number off | |
587 | * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode | |
588 | * format) then log recovery is dependent on the di_flushiter field being | |
589 | * initialised from the current on-disk value and hence we must also read the | |
590 | * inode off disk. | |
a2ceac1f DC |
591 | */ |
592 | int | |
593 | xfs_iread( | |
594 | xfs_mount_t *mp, | |
595 | xfs_trans_t *tp, | |
596 | xfs_inode_t *ip, | |
597 | uint iget_flags) | |
598 | { | |
599 | xfs_buf_t *bp; | |
600 | xfs_dinode_t *dip; | |
bc01119d | 601 | xfs_failaddr_t fa; |
a2ceac1f DC |
602 | int error; |
603 | ||
604 | /* | |
605 | * Fill in the location information in the in-core inode. | |
606 | */ | |
607 | error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags); | |
608 | if (error) | |
609 | return error; | |
610 | ||
ff105f75 DC |
611 | /* shortcut IO on inode allocation if possible */ |
612 | if ((iget_flags & XFS_IGET_CREATE) && | |
db84c7e8 | 613 | xfs_sb_version_has_v3inode(&mp->m_sb) && |
ff105f75 | 614 | !(mp->m_flags & XFS_MOUNT_IKEEP)) { |
6652c253 | 615 | VFS_I(ip)->i_generation = prandom_u32(); |
ff105f75 DC |
616 | return 0; |
617 | } | |
618 | ||
a2ceac1f DC |
619 | /* |
620 | * Get pointers to the on-disk inode and the buffer containing it. | |
621 | */ | |
0cd4b54a | 622 | error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0); |
a2ceac1f DC |
623 | if (error) |
624 | return error; | |
625 | ||
41ce5f36 | 626 | /* even unallocated inodes are verified */ |
bc01119d DW |
627 | fa = xfs_dinode_verify(mp, ip->i_ino, dip); |
628 | if (fa) { | |
1d3bac1f DW |
629 | xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip, |
630 | sizeof(*dip), fa); | |
12b53197 | 631 | error = -EFSCORRUPTED; |
a2ceac1f DC |
632 | goto out_brelse; |
633 | } | |
634 | ||
635 | /* | |
636 | * If the on-disk inode is already linked to a directory | |
637 | * entry, copy all of the inode into the in-core inode. | |
5d90ab5a | 638 | * xfs_iformat_fork() handles copying in the inode format |
a2ceac1f DC |
639 | * specific information. |
640 | * Otherwise, just get the truly permanent information. | |
641 | */ | |
642 | if (dip->di_mode) { | |
3eb685ae | 643 | error = xfs_inode_from_disk(ip, dip); |
a2ceac1f DC |
644 | if (error) { |
645 | #ifdef DEBUG | |
646 | xfs_alert(mp, "%s: xfs_iformat() returned error %d", | |
647 | __func__, error); | |
648 | #endif /* DEBUG */ | |
649 | goto out_brelse; | |
650 | } | |
651 | } else { | |
41ce5f36 DC |
652 | /* |
653 | * Partial initialisation of the in-core inode. Just the bits | |
654 | * that xfs_ialloc won't overwrite or relies on being correct. | |
655 | */ | |
6652c253 | 656 | VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen); |
a2ceac1f | 657 | ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter); |
41ce5f36 | 658 | |
a2ceac1f DC |
659 | /* |
660 | * Make sure to pull in the mode here as well in | |
661 | * case the inode is released without being used. | |
662 | * This ensures that xfs_inactive() will see that | |
663 | * the inode is already free and not try to mess | |
664 | * with the uninitialized part of it. | |
665 | */ | |
e37bf53c | 666 | VFS_I(ip)->i_mode = 0; |
a2ceac1f DC |
667 | } |
668 | ||
a2ceac1f DC |
669 | ip->i_delayed_blks = 0; |
670 | ||
671 | /* | |
672 | * Mark the buffer containing the inode as something to keep | |
673 | * around for a while. This helps to keep recently accessed | |
674 | * meta-data in-core longer. | |
675 | */ | |
676 | xfs_buf_set_ref(bp, XFS_INO_REF); | |
677 | ||
678 | /* | |
1d519883 DC |
679 | * Use xfs_trans_brelse() to release the buffer containing the on-disk |
680 | * inode, because it was acquired with xfs_trans_read_buf() in | |
681 | * xfs_imap_to_bp() above. If tp is NULL, this is just a normal | |
a2ceac1f DC |
682 | * brelse(). If we're within a transaction, then xfs_trans_brelse() |
683 | * will only release the buffer if it is not dirty within the | |
684 | * transaction. It will be OK to release the buffer in this case, | |
1d519883 DC |
685 | * because inodes on disk are never destroyed and we will be locking the |
686 | * new in-core inode before putting it in the cache where other | |
687 | * processes can find it. Thus we don't have to worry about the inode | |
688 | * being changed just because we released the buffer. | |
a2ceac1f DC |
689 | */ |
690 | out_brelse: | |
691 | xfs_trans_brelse(tp, bp); | |
692 | return error; | |
693 | } | |
fd0c360d DW |
694 | |
695 | /* | |
696 | * Validate di_extsize hint. | |
697 | * | |
698 | * The rules are documented at xfs_ioctl_setattr_check_extsize(). | |
699 | * These functions must be kept in sync with each other. | |
700 | */ | |
701 | xfs_failaddr_t | |
702 | xfs_inode_validate_extsize( | |
703 | struct xfs_mount *mp, | |
704 | uint32_t extsize, | |
705 | uint16_t mode, | |
706 | uint16_t flags) | |
707 | { | |
708 | bool rt_flag; | |
709 | bool hint_flag; | |
710 | bool inherit_flag; | |
711 | uint32_t extsize_bytes; | |
712 | uint32_t blocksize_bytes; | |
713 | ||
714 | rt_flag = (flags & XFS_DIFLAG_REALTIME); | |
715 | hint_flag = (flags & XFS_DIFLAG_EXTSIZE); | |
716 | inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT); | |
717 | extsize_bytes = XFS_FSB_TO_B(mp, extsize); | |
718 | ||
719 | if (rt_flag) | |
720 | blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog; | |
721 | else | |
722 | blocksize_bytes = mp->m_sb.sb_blocksize; | |
723 | ||
724 | if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode))) | |
725 | return __this_address; | |
726 | ||
727 | if (hint_flag && !S_ISREG(mode)) | |
728 | return __this_address; | |
729 | ||
730 | if (inherit_flag && !S_ISDIR(mode)) | |
731 | return __this_address; | |
732 | ||
733 | if ((hint_flag || inherit_flag) && extsize == 0) | |
734 | return __this_address; | |
735 | ||
206bef93 ES |
736 | /* free inodes get flags set to zero but extsize remains */ |
737 | if (mode && !(hint_flag || inherit_flag) && extsize != 0) | |
fd0c360d DW |
738 | return __this_address; |
739 | ||
740 | if (extsize_bytes % blocksize_bytes) | |
741 | return __this_address; | |
742 | ||
743 | if (extsize > MAXEXTLEN) | |
744 | return __this_address; | |
745 | ||
746 | if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2) | |
747 | return __this_address; | |
748 | ||
749 | return NULL; | |
750 | } | |
751 | ||
752 | /* | |
753 | * Validate di_cowextsize hint. | |
754 | * | |
755 | * The rules are documented at xfs_ioctl_setattr_check_cowextsize(). | |
756 | * These functions must be kept in sync with each other. | |
757 | */ | |
758 | xfs_failaddr_t | |
759 | xfs_inode_validate_cowextsize( | |
760 | struct xfs_mount *mp, | |
761 | uint32_t cowextsize, | |
762 | uint16_t mode, | |
763 | uint16_t flags, | |
764 | uint64_t flags2) | |
765 | { | |
766 | bool rt_flag; | |
767 | bool hint_flag; | |
768 | uint32_t cowextsize_bytes; | |
769 | ||
770 | rt_flag = (flags & XFS_DIFLAG_REALTIME); | |
771 | hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE); | |
772 | cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize); | |
773 | ||
774 | if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb)) | |
775 | return __this_address; | |
776 | ||
777 | if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode))) | |
778 | return __this_address; | |
779 | ||
780 | if (hint_flag && cowextsize == 0) | |
781 | return __this_address; | |
782 | ||
206bef93 ES |
783 | /* free inodes get flags set to zero but cowextsize remains */ |
784 | if (mode && !hint_flag && cowextsize != 0) | |
fd0c360d DW |
785 | return __this_address; |
786 | ||
787 | if (hint_flag && rt_flag) | |
788 | return __this_address; | |
789 | ||
790 | if (cowextsize_bytes % mp->m_sb.sb_blocksize) | |
791 | return __this_address; | |
792 | ||
793 | if (cowextsize > MAXEXTLEN) | |
794 | return __this_address; | |
795 | ||
796 | if (cowextsize > mp->m_sb.sb_agblocks / 2) | |
797 | return __this_address; | |
798 | ||
799 | return NULL; | |
800 | } |