]> git.ipfire.org Git - thirdparty/xfsprogs-dev.git/blame - libxfs/xfs_inode_buf.c
xfs: move inode extent size hint validation to libxfs
[thirdparty/xfsprogs-dev.git] / libxfs / xfs_inode_buf.c
CommitLineData
2bd0ea18 1/*
5e656dbb 2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
da23017d 3 * All Rights Reserved.
5000d01d 4 *
da23017d
NS
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
2bd0ea18 7 * published by the Free Software Foundation.
5000d01d 8 *
da23017d
NS
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
5000d01d 13 *
da23017d
NS
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2bd0ea18 17 */
9c799827 18#include "libxfs_priv.h"
b626fb59
DC
19#include "xfs_fs.h"
20#include "xfs_shared.h"
21#include "xfs_format.h"
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
24#include "xfs_mount.h"
f944d3d0 25#include "xfs_defer.h"
b626fb59 26#include "xfs_inode.h"
56d3fc2b 27#include "xfs_errortag.h"
b626fb59
DC
28#include "xfs_cksum.h"
29#include "xfs_trans.h"
30#include "xfs_ialloc.h"
0962b5b3 31#include "xfs_dir2.h"
2bd0ea18 32
5e656dbb
BN
33/*
34 * Check that none of the inode's in the buffer have a next
35 * unlinked field of 0.
36 */
37#if defined(DEBUG)
2bd0ea18
NS
38void
39xfs_inobp_check(
40 xfs_mount_t *mp,
41 xfs_buf_t *bp)
42{
43 int i;
44 int j;
45 xfs_dinode_t *dip;
46
47 j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;
48
49 for (i = 0; i < j; i++) {
92acb899 50 dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize);
46eca962 51 if (!dip->di_next_unlinked) {
a2ceac1f 52 xfs_alert(mp,
ea6a00d4
DC
53 "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
54 i, (long long)bp->b_bn);
2bd0ea18
NS
55 }
56 }
57}
58#endif
59
307ae950
RW
60bool
61xfs_dinode_good_version(
62 struct xfs_mount *mp,
63 __u8 version)
64{
65 if (xfs_sb_version_hascrc(&mp->m_sb))
66 return version == 3;
67
68 return version == 1 || version == 2;
69}
70
f85fc622
DC
71/*
72 * If we are doing readahead on an inode buffer, we might be in log recovery
73 * reading an inode allocation buffer that hasn't yet been replayed, and hence
74 * has not had the inode cores stamped into it. Hence for readahead, the buffer
75 * may be potentially invalid.
76 *
736e1cf6
DC
77 * If the readahead buffer is invalid, we need to mark it with an error and
78 * clear the DONE status of the buffer so that a followup read will re-read it
79 * from disk. We don't report the error otherwise to avoid warnings during log
80 * recovery and we don't get unnecssary panics on debug kernels. We use EIO here
81 * because all we want to do is say readahead failed; there is no-one to report
82 * the error to, so this will distinguish it from a non-ra verifier failure.
6daba42a
DC
83 * Changes to this readahead error behavour also need to be reflected in
84 * xfs_dquot_buf_readahead_verify().
f85fc622 85 */
a2ceac1f
DC
86static void
87xfs_inode_buf_verify(
f85fc622
DC
88 struct xfs_buf *bp,
89 bool readahead)
5e656dbb 90{
a2ceac1f 91 struct xfs_mount *mp = bp->b_target->bt_mount;
5e656dbb
BN
92 int i;
93 int ni;
5e656dbb
BN
94
95 /*
96 * Validate the magic number and version of every inode in the buffer
5e656dbb 97 */
a2ceac1f 98 ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
5e656dbb
BN
99 for (i = 0; i < ni; i++) {
100 int di_ok;
101 xfs_dinode_t *dip;
102
92acb899 103 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
a2ceac1f 104 di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
307ae950 105 xfs_dinode_good_version(mp, dip->di_version);
5e656dbb 106 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
e2a190dd 107 XFS_ERRTAG_ITOBP_INOTOBP))) {
f85fc622
DC
108 if (readahead) {
109 bp->b_flags &= ~XBF_DONE;
736e1cf6 110 xfs_buf_ioerror(bp, -EIO);
f85fc622
DC
111 return;
112 }
113
5e656dbb 114#ifdef DEBUG
ea6a00d4 115 xfs_alert(mp,
a2ceac1f
DC
116 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
117 (unsigned long long)bp->b_bn, i,
56b2de80 118 be16_to_cpu(dip->di_magic));
5e656dbb 119#endif
43f7ceb5
DW
120 xfs_buf_verifier_error(bp, -EFSCORRUPTED,
121 __func__, dip, sizeof(*dip),
122 NULL);
5e656dbb
BN
123 }
124 }
5e656dbb 125 xfs_inobp_check(mp, bp);
a2ceac1f 126}
5e656dbb 127
e6d77a21 128
a2ceac1f
DC
129static void
130xfs_inode_buf_read_verify(
131 struct xfs_buf *bp)
132{
f85fc622
DC
133 xfs_inode_buf_verify(bp, false);
134}
135
136static void
137xfs_inode_buf_readahead_verify(
138 struct xfs_buf *bp)
139{
140 xfs_inode_buf_verify(bp, true);
5e656dbb 141}
2bd0ea18 142
a2ceac1f
DC
143static void
144xfs_inode_buf_write_verify(
145 struct xfs_buf *bp)
56b2de80 146{
f85fc622 147 xfs_inode_buf_verify(bp, false);
56b2de80
DC
148}
149
a2ceac1f 150const struct xfs_buf_ops xfs_inode_buf_ops = {
a3fac935 151 .name = "xfs_inode",
a2ceac1f
DC
152 .verify_read = xfs_inode_buf_read_verify,
153 .verify_write = xfs_inode_buf_write_verify,
154};
155
f85fc622 156const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
a3fac935 157 .name = "xxfs_inode_ra",
f85fc622
DC
158 .verify_read = xfs_inode_buf_readahead_verify,
159 .verify_write = xfs_inode_buf_write_verify,
160};
161
162
2bd0ea18 163/*
a2ceac1f
DC
164 * This routine is called to map an inode to the buffer containing the on-disk
165 * version of the inode. It returns a pointer to the buffer containing the
166 * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
167 * pointer to the on-disk inode within that buffer.
2bd0ea18 168 *
a2ceac1f
DC
169 * If a non-zero error is returned, then the contents of bpp and dipp are
170 * undefined.
2bd0ea18
NS
171 */
172int
a2ceac1f
DC
173xfs_imap_to_bp(
174 struct xfs_mount *mp,
175 struct xfs_trans *tp,
176 struct xfs_imap *imap,
177 struct xfs_dinode **dipp,
178 struct xfs_buf **bpp,
179 uint buf_flags,
180 uint iget_flags)
2bd0ea18 181{
a2ceac1f
DC
182 struct xfs_buf *bp;
183 int error;
2bd0ea18 184
a2ceac1f
DC
185 buf_flags |= XBF_UNMAPPED;
186 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
187 (int)imap->im_len, buf_flags, &bp,
188 &xfs_inode_buf_ops);
189 if (error) {
12b53197 190 if (error == -EAGAIN) {
a2ceac1f
DC
191 ASSERT(buf_flags & XBF_TRYLOCK);
192 return error;
193 }
2bd0ea18 194
12b53197 195 if (error == -EFSCORRUPTED &&
a2ceac1f 196 (iget_flags & XFS_IGET_UNTRUSTED))
12b53197 197 return -EINVAL;
2bd0ea18 198
a2ceac1f
DC
199 xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
200 __func__, error);
201 return error;
2bd0ea18 202 }
2bd0ea18 203
2bd0ea18 204 *bpp = bp;
92acb899 205 *dipp = xfs_buf_offset(bp, imap->im_boffset);
2bd0ea18
NS
206 return 0;
207}
208
5000d01d 209void
1bc6cbe3
DC
210xfs_inode_from_disk(
211 struct xfs_inode *ip,
138659f1 212 struct xfs_dinode *from)
2bd0ea18 213{
1bc6cbe3
DC
214 struct xfs_icdinode *to = &ip->i_d;
215 struct inode *inode = VFS_I(ip);
216
f089fc42
DC
217
218 /*
219 * Convert v1 inodes immediately to v2 inode format as this is the
220 * minimum inode version format we support in the rest of the code.
221 */
e37bf53c 222 to->di_version = from->di_version;
f089fc42 223 if (to->di_version == 1) {
bcbe04c1 224 set_nlink(inode, be16_to_cpu(from->di_onlink));
f089fc42
DC
225 to->di_projid_lo = 0;
226 to->di_projid_hi = 0;
227 to->di_version = 2;
228 } else {
bcbe04c1 229 set_nlink(inode, be32_to_cpu(from->di_nlink));
f089fc42
DC
230 to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
231 to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
232 }
233
5e656dbb 234 to->di_format = from->di_format;
5e656dbb
BN
235 to->di_uid = be32_to_cpu(from->di_uid);
236 to->di_gid = be32_to_cpu(from->di_gid);
5e656dbb 237 to->di_flushiter = be16_to_cpu(from->di_flushiter);
1bc6cbe3
DC
238
239 /*
240 * Time is signed, so need to convert to signed 32 bit before
241 * storing in inode timestamp which may be 64 bit. Otherwise
242 * a time before epoch is converted to a time long after epoch
243 * on 64 bit systems.
244 */
245 inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec);
246 inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec);
247 inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec);
248 inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec);
249 inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec);
250 inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec);
6652c253 251 inode->i_generation = be32_to_cpu(from->di_gen);
e37bf53c 252 inode->i_mode = be16_to_cpu(from->di_mode);
1bc6cbe3 253
5e656dbb
BN
254 to->di_size = be64_to_cpu(from->di_size);
255 to->di_nblocks = be64_to_cpu(from->di_nblocks);
256 to->di_extsize = be32_to_cpu(from->di_extsize);
257 to->di_nextents = be32_to_cpu(from->di_nextents);
258 to->di_anextents = be16_to_cpu(from->di_anextents);
259 to->di_forkoff = from->di_forkoff;
260 to->di_aformat = from->di_aformat;
261 to->di_dmevmask = be32_to_cpu(from->di_dmevmask);
262 to->di_dmstate = be16_to_cpu(from->di_dmstate);
263 to->di_flags = be16_to_cpu(from->di_flags);
41ce5f36
DC
264
265 if (to->di_version == 3) {
d5105a89
JL
266 inode_set_iversion_queried(inode,
267 be64_to_cpu(from->di_changecount));
41ce5f36
DC
268 to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
269 to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
270 to->di_flags2 = be64_to_cpu(from->di_flags2);
10460994 271 to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
41ce5f36 272 }
5e656dbb
BN
273}
274
275void
1bc6cbe3
DC
276xfs_inode_to_disk(
277 struct xfs_inode *ip,
db17aebe
DC
278 struct xfs_dinode *to,
279 xfs_lsn_t lsn)
1bc6cbe3
DC
280{
281 struct xfs_icdinode *from = &ip->i_d;
282 struct inode *inode = VFS_I(ip);
283
db17aebe 284 to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
f089fc42 285 to->di_onlink = 0;
db17aebe 286
db17aebe 287 to->di_version = from->di_version;
1bc6cbe3 288 to->di_format = from->di_format;
1bc6cbe3
DC
289 to->di_uid = cpu_to_be32(from->di_uid);
290 to->di_gid = cpu_to_be32(from->di_gid);
1bc6cbe3
DC
291 to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
292 to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
1bc6cbe3 293
db17aebe 294 memset(to->di_pad, 0, sizeof(to->di_pad));
1bc6cbe3
DC
295 to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec);
296 to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec);
297 to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec);
298 to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec);
299 to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec);
300 to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec);
bcbe04c1 301 to->di_nlink = cpu_to_be32(inode->i_nlink);
6652c253 302 to->di_gen = cpu_to_be32(inode->i_generation);
e37bf53c 303 to->di_mode = cpu_to_be16(inode->i_mode);
1bc6cbe3
DC
304
305 to->di_size = cpu_to_be64(from->di_size);
306 to->di_nblocks = cpu_to_be64(from->di_nblocks);
307 to->di_extsize = cpu_to_be32(from->di_extsize);
308 to->di_nextents = cpu_to_be32(from->di_nextents);
309 to->di_anextents = cpu_to_be16(from->di_anextents);
310 to->di_forkoff = from->di_forkoff;
311 to->di_aformat = from->di_aformat;
312 to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
313 to->di_dmstate = cpu_to_be16(from->di_dmstate);
314 to->di_flags = cpu_to_be16(from->di_flags);
1bc6cbe3
DC
315
316 if (from->di_version == 3) {
d5105a89 317 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
1bc6cbe3
DC
318 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
319 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
320 to->di_flags2 = cpu_to_be64(from->di_flags2);
10460994 321 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
db17aebe
DC
322 to->di_ino = cpu_to_be64(ip->i_ino);
323 to->di_lsn = cpu_to_be64(lsn);
324 memset(to->di_pad2, 0, sizeof(to->di_pad2));
325 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
1bc6cbe3
DC
326 to->di_flushiter = 0;
327 } else {
328 to->di_flushiter = cpu_to_be16(from->di_flushiter);
329 }
330}
331
332void
333xfs_log_dinode_to_disk(
334 struct xfs_log_dinode *from,
335 struct xfs_dinode *to)
5e656dbb
BN
336{
337 to->di_magic = cpu_to_be16(from->di_magic);
338 to->di_mode = cpu_to_be16(from->di_mode);
f089fc42 339 to->di_version = from->di_version;
5e656dbb 340 to->di_format = from->di_format;
f089fc42 341 to->di_onlink = 0;
5e656dbb
BN
342 to->di_uid = cpu_to_be32(from->di_uid);
343 to->di_gid = cpu_to_be32(from->di_gid);
344 to->di_nlink = cpu_to_be32(from->di_nlink);
22bc10ed
AM
345 to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
346 to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
5e656dbb 347 memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
1bc6cbe3 348
5e656dbb
BN
349 to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
350 to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
351 to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
352 to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
353 to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
354 to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
1bc6cbe3 355
5e656dbb
BN
356 to->di_size = cpu_to_be64(from->di_size);
357 to->di_nblocks = cpu_to_be64(from->di_nblocks);
358 to->di_extsize = cpu_to_be32(from->di_extsize);
359 to->di_nextents = cpu_to_be32(from->di_nextents);
360 to->di_anextents = cpu_to_be16(from->di_anextents);
361 to->di_forkoff = from->di_forkoff;
362 to->di_aformat = from->di_aformat;
363 to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
364 to->di_dmstate = cpu_to_be16(from->di_dmstate);
365 to->di_flags = cpu_to_be16(from->di_flags);
366 to->di_gen = cpu_to_be32(from->di_gen);
41ce5f36
DC
367
368 if (from->di_version == 3) {
369 to->di_changecount = cpu_to_be64(from->di_changecount);
370 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
371 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
372 to->di_flags2 = cpu_to_be64(from->di_flags2);
10460994 373 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
41ce5f36
DC
374 to->di_ino = cpu_to_be64(from->di_ino);
375 to->di_lsn = cpu_to_be64(from->di_lsn);
376 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
f85fc622 377 uuid_copy(&to->di_uuid, &from->di_uuid);
fde71651
DC
378 to->di_flushiter = 0;
379 } else {
380 to->di_flushiter = cpu_to_be16(from->di_flushiter);
41ce5f36
DC
381 }
382}
383
bc01119d 384xfs_failaddr_t
41ce5f36
DC
385xfs_dinode_verify(
386 struct xfs_mount *mp,
e515cca1 387 xfs_ino_t ino,
41ce5f36
DC
388 struct xfs_dinode *dip)
389{
3cfabff6 390 uint16_t mode;
183537ed
DW
391 uint16_t flags;
392 uint64_t flags2;
ad42e5a1 393 uint64_t di_size;
183537ed 394
41ce5f36 395 if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
bc01119d 396 return __this_address;
41ce5f36 397
98703400
DW
398 /* Verify v3 integrity information first */
399 if (dip->di_version >= 3) {
400 if (!xfs_sb_version_hascrc(&mp->m_sb))
401 return __this_address;
402 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
403 XFS_DINODE_CRC_OFF))
404 return __this_address;
405 if (be64_to_cpu(dip->di_ino) != ino)
406 return __this_address;
407 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
408 return __this_address;
409 }
410
9d85eb44 411 /* don't allow invalid i_size */
ad42e5a1
DW
412 di_size = be64_to_cpu(dip->di_size);
413 if (di_size & (1ULL << 63))
bc01119d 414 return __this_address;
9d85eb44 415
3cfabff6 416 mode = be16_to_cpu(dip->di_mode);
0962b5b3 417 if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
bc01119d 418 return __this_address;
3cfabff6
AG
419
420 /* No zero-length symlinks/dirs. */
ad42e5a1 421 if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
bc01119d 422 return __this_address;
9d85eb44 423
ad42e5a1
DW
424 /* Fork checks carried over from xfs_iformat_fork */
425 if (mode &&
426 be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
427 be64_to_cpu(dip->di_nblocks))
428 return __this_address;
429
430 if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
431 return __this_address;
432
433 flags = be16_to_cpu(dip->di_flags);
434
435 if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
436 return __this_address;
437
438 /* Do we have appropriate data fork formats for the mode? */
439 switch (mode & S_IFMT) {
440 case S_IFIFO:
441 case S_IFCHR:
442 case S_IFBLK:
443 case S_IFSOCK:
444 if (dip->di_format != XFS_DINODE_FMT_DEV)
445 return __this_address;
446 break;
447 case S_IFREG:
448 case S_IFLNK:
449 case S_IFDIR:
450 switch (dip->di_format) {
451 case XFS_DINODE_FMT_LOCAL:
452 /*
453 * no local regular files yet
454 */
455 if (S_ISREG(mode))
456 return __this_address;
457 if (di_size > XFS_DFORK_DSIZE(dip, mp))
458 return __this_address;
459 /* fall through */
460 case XFS_DINODE_FMT_EXTENTS:
461 case XFS_DINODE_FMT_BTREE:
462 break;
463 default:
464 return __this_address;
465 }
466 break;
467 case 0:
468 /* Uninitialized inode ok. */
469 break;
470 default:
471 return __this_address;
472 }
473
474 if (XFS_DFORK_Q(dip)) {
475 switch (dip->di_aformat) {
476 case XFS_DINODE_FMT_LOCAL:
477 case XFS_DINODE_FMT_EXTENTS:
478 case XFS_DINODE_FMT_BTREE:
479 break;
480 default:
481 return __this_address;
482 }
483 }
484
41ce5f36
DC
485 /* only version 3 or greater inodes are extensively verified here */
486 if (dip->di_version < 3)
bc01119d 487 return NULL;
41ce5f36 488
183537ed
DW
489 flags2 = be64_to_cpu(dip->di_flags2);
490
491 /* don't allow reflink/cowextsize if we don't have reflink */
492 if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
493 !xfs_sb_version_hasreflink(&mp->m_sb))
bc01119d 494 return __this_address;
183537ed 495
ad42e5a1
DW
496 /* only regular files get reflink */
497 if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
498 return __this_address;
499
183537ed
DW
500 /* don't let reflink and realtime mix */
501 if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
bc01119d 502 return __this_address;
183537ed 503
378ed4e7
DW
504 /* don't let reflink and dax mix */
505 if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
bc01119d 506 return __this_address;
378ed4e7 507
bc01119d 508 return NULL;
41ce5f36 509}
e6d77a21 510
41ce5f36
DC
511void
512xfs_dinode_calc_crc(
513 struct xfs_mount *mp,
514 struct xfs_dinode *dip)
515{
4a492e72 516 uint32_t crc;
41ce5f36
DC
517
518 if (dip->di_version < 3)
519 return;
520
521 ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
0bb90214 522 crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
3faddc54 523 XFS_DINODE_CRC_OFF);
41ce5f36 524 dip->di_crc = xfs_end_cksum(crc);
2bd0ea18
NS
525}
526
a2ceac1f
DC
527/*
528 * Read the disk inode attributes into the in-core inode structure.
ff105f75
DC
529 *
530 * For version 5 superblocks, if we are initialising a new inode and we are not
531 * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
532 * inode core with a random generation number. If we are keeping inodes around,
533 * we need to read the inode cluster to get the existing generation number off
534 * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
535 * format) then log recovery is dependent on the di_flushiter field being
536 * initialised from the current on-disk value and hence we must also read the
537 * inode off disk.
a2ceac1f
DC
538 */
539int
540xfs_iread(
541 xfs_mount_t *mp,
542 xfs_trans_t *tp,
543 xfs_inode_t *ip,
544 uint iget_flags)
545{
546 xfs_buf_t *bp;
547 xfs_dinode_t *dip;
bc01119d 548 xfs_failaddr_t fa;
a2ceac1f
DC
549 int error;
550
551 /*
552 * Fill in the location information in the in-core inode.
553 */
554 error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
555 if (error)
556 return error;
557
ff105f75
DC
558 /* shortcut IO on inode allocation if possible */
559 if ((iget_flags & XFS_IGET_CREATE) &&
560 xfs_sb_version_hascrc(&mp->m_sb) &&
561 !(mp->m_flags & XFS_MOUNT_IKEEP)) {
562 /* initialise the on-disk inode core */
563 memset(&ip->i_d, 0, sizeof(ip->i_d));
6652c253 564 VFS_I(ip)->i_generation = prandom_u32();
db17aebe 565 if (xfs_sb_version_hascrc(&mp->m_sb))
ff105f75 566 ip->i_d.di_version = 3;
db17aebe 567 else
ff105f75
DC
568 ip->i_d.di_version = 2;
569 return 0;
570 }
571
a2ceac1f
DC
572 /*
573 * Get pointers to the on-disk inode and the buffer containing it.
574 */
575 error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
576 if (error)
577 return error;
578
41ce5f36 579 /* even unallocated inodes are verified */
bc01119d
DW
580 fa = xfs_dinode_verify(mp, ip->i_ino, dip);
581 if (fa) {
1d3bac1f
DW
582 xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
583 sizeof(*dip), fa);
12b53197 584 error = -EFSCORRUPTED;
a2ceac1f
DC
585 goto out_brelse;
586 }
587
588 /*
589 * If the on-disk inode is already linked to a directory
590 * entry, copy all of the inode into the in-core inode.
5d90ab5a 591 * xfs_iformat_fork() handles copying in the inode format
a2ceac1f
DC
592 * specific information.
593 * Otherwise, just get the truly permanent information.
594 */
595 if (dip->di_mode) {
1bc6cbe3 596 xfs_inode_from_disk(ip, dip);
5d90ab5a 597 error = xfs_iformat_fork(ip, dip);
a2ceac1f
DC
598 if (error) {
599#ifdef DEBUG
600 xfs_alert(mp, "%s: xfs_iformat() returned error %d",
601 __func__, error);
602#endif /* DEBUG */
603 goto out_brelse;
604 }
605 } else {
41ce5f36
DC
606 /*
607 * Partial initialisation of the in-core inode. Just the bits
608 * that xfs_ialloc won't overwrite or relies on being correct.
609 */
a2ceac1f 610 ip->i_d.di_version = dip->di_version;
6652c253 611 VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
a2ceac1f 612 ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
41ce5f36 613
a2ceac1f
DC
614 /*
615 * Make sure to pull in the mode here as well in
616 * case the inode is released without being used.
617 * This ensures that xfs_inactive() will see that
618 * the inode is already free and not try to mess
619 * with the uninitialized part of it.
620 */
e37bf53c 621 VFS_I(ip)->i_mode = 0;
a2ceac1f
DC
622 }
623
f089fc42 624 ASSERT(ip->i_d.di_version >= 2);
a2ceac1f
DC
625 ip->i_delayed_blks = 0;
626
627 /*
628 * Mark the buffer containing the inode as something to keep
629 * around for a while. This helps to keep recently accessed
630 * meta-data in-core longer.
631 */
632 xfs_buf_set_ref(bp, XFS_INO_REF);
633
634 /*
1d519883
DC
635 * Use xfs_trans_brelse() to release the buffer containing the on-disk
636 * inode, because it was acquired with xfs_trans_read_buf() in
637 * xfs_imap_to_bp() above. If tp is NULL, this is just a normal
a2ceac1f
DC
638 * brelse(). If we're within a transaction, then xfs_trans_brelse()
639 * will only release the buffer if it is not dirty within the
640 * transaction. It will be OK to release the buffer in this case,
1d519883
DC
641 * because inodes on disk are never destroyed and we will be locking the
642 * new in-core inode before putting it in the cache where other
643 * processes can find it. Thus we don't have to worry about the inode
644 * being changed just because we released the buffer.
a2ceac1f
DC
645 */
646 out_brelse:
647 xfs_trans_brelse(tp, bp);
648 return error;
649}
fd0c360d
DW
650
651/*
652 * Validate di_extsize hint.
653 *
654 * The rules are documented at xfs_ioctl_setattr_check_extsize().
655 * These functions must be kept in sync with each other.
656 */
657xfs_failaddr_t
658xfs_inode_validate_extsize(
659 struct xfs_mount *mp,
660 uint32_t extsize,
661 uint16_t mode,
662 uint16_t flags)
663{
664 bool rt_flag;
665 bool hint_flag;
666 bool inherit_flag;
667 uint32_t extsize_bytes;
668 uint32_t blocksize_bytes;
669
670 rt_flag = (flags & XFS_DIFLAG_REALTIME);
671 hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
672 inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
673 extsize_bytes = XFS_FSB_TO_B(mp, extsize);
674
675 if (rt_flag)
676 blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
677 else
678 blocksize_bytes = mp->m_sb.sb_blocksize;
679
680 if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
681 return __this_address;
682
683 if (hint_flag && !S_ISREG(mode))
684 return __this_address;
685
686 if (inherit_flag && !S_ISDIR(mode))
687 return __this_address;
688
689 if ((hint_flag || inherit_flag) && extsize == 0)
690 return __this_address;
691
692 if (!(hint_flag || inherit_flag) && extsize != 0)
693 return __this_address;
694
695 if (extsize_bytes % blocksize_bytes)
696 return __this_address;
697
698 if (extsize > MAXEXTLEN)
699 return __this_address;
700
701 if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
702 return __this_address;
703
704 return NULL;
705}
706
707/*
708 * Validate di_cowextsize hint.
709 *
710 * The rules are documented at xfs_ioctl_setattr_check_cowextsize().
711 * These functions must be kept in sync with each other.
712 */
713xfs_failaddr_t
714xfs_inode_validate_cowextsize(
715 struct xfs_mount *mp,
716 uint32_t cowextsize,
717 uint16_t mode,
718 uint16_t flags,
719 uint64_t flags2)
720{
721 bool rt_flag;
722 bool hint_flag;
723 uint32_t cowextsize_bytes;
724
725 rt_flag = (flags & XFS_DIFLAG_REALTIME);
726 hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
727 cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
728
729 if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb))
730 return __this_address;
731
732 if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
733 return __this_address;
734
735 if (hint_flag && cowextsize == 0)
736 return __this_address;
737
738 if (!hint_flag && cowextsize != 0)
739 return __this_address;
740
741 if (hint_flag && rt_flag)
742 return __this_address;
743
744 if (cowextsize_bytes % mp->m_sb.sb_blocksize)
745 return __this_address;
746
747 if (cowextsize > MAXEXTLEN)
748 return __this_address;
749
750 if (cowextsize > mp->m_sb.sb_agblocks / 2)
751 return __this_address;
752
753 return NULL;
754}