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