Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0
2 : /*
3 : * Copyright (c) 2000-2006 Silicon Graphics, Inc.
4 : * All Rights Reserved.
5 : */
6 : #include "xfs.h"
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_ag.h"
14 : #include "xfs_inode.h"
15 : #include "xfs_errortag.h"
16 : #include "xfs_error.h"
17 : #include "xfs_icache.h"
18 : #include "xfs_trans.h"
19 : #include "xfs_ialloc.h"
20 : #include "xfs_dir2.h"
21 :
22 : #include <linux/iversion.h>
23 :
24 : /*
25 : * If we are doing readahead on an inode buffer, we might be in log recovery
26 : * reading an inode allocation buffer that hasn't yet been replayed, and hence
27 : * has not had the inode cores stamped into it. Hence for readahead, the buffer
28 : * may be potentially invalid.
29 : *
30 : * If the readahead buffer is invalid, we need to mark it with an error and
31 : * clear the DONE status of the buffer so that a followup read will re-read it
32 : * from disk. We don't report the error otherwise to avoid warnings during log
33 : * recovery and we don't get unnecessary panics on debug kernels. We use EIO here
34 : * because all we want to do is say readahead failed; there is no-one to report
35 : * the error to, so this will distinguish it from a non-ra verifier failure.
36 : * Changes to this readahead error behaviour also need to be reflected in
37 : * xfs_dquot_buf_readahead_verify().
38 : */
39 : static void
40 40197277 : xfs_inode_buf_verify(
41 : struct xfs_buf *bp,
42 : bool readahead)
43 : {
44 40197277 : struct xfs_mount *mp = bp->b_mount;
45 40197277 : int i;
46 40197277 : int ni;
47 :
48 : /*
49 : * Validate the magic number and version of every inode in the buffer
50 : */
51 40197277 : ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
52 1326356624 : for (i = 0; i < ni; i++) {
53 1286164479 : struct xfs_dinode *dip;
54 1286164479 : xfs_agino_t unlinked_ino;
55 1286164479 : int di_ok;
56 :
57 1286164479 : dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
58 1286164385 : unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
59 2572325344 : di_ok = xfs_verify_magic16(bp, dip->di_magic) &&
60 1286164560 : xfs_dinode_good_version(mp, dip->di_version) &&
61 1286161011 : xfs_verify_agino_or_null(bp->b_pag, unlinked_ino);
62 1286164560 : if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
63 : XFS_ERRTAG_ITOBP_INOTOBP))) {
64 5131 : if (readahead) {
65 3158 : bp->b_flags &= ~XBF_DONE;
66 3158 : xfs_buf_ioerror(bp, -EIO);
67 3158 : return;
68 : }
69 :
70 : #ifdef DEBUG
71 1973 : xfs_alert(mp,
72 : "bad inode magic/vsn daddr %lld #%d (magic=%x)",
73 : (unsigned long long)xfs_buf_daddr(bp), i,
74 : be16_to_cpu(dip->di_magic));
75 : #endif
76 1973 : xfs_buf_verifier_error(bp, -EFSCORRUPTED,
77 : __func__, dip, sizeof(*dip),
78 : NULL);
79 1973 : return;
80 : }
81 : }
82 : }
83 :
84 :
85 : static void
86 11822497 : xfs_inode_buf_read_verify(
87 : struct xfs_buf *bp)
88 : {
89 11822497 : xfs_inode_buf_verify(bp, false);
90 11822497 : }
91 :
92 : static void
93 299131 : xfs_inode_buf_readahead_verify(
94 : struct xfs_buf *bp)
95 : {
96 299131 : xfs_inode_buf_verify(bp, true);
97 299131 : }
98 :
99 : static void
100 28075652 : xfs_inode_buf_write_verify(
101 : struct xfs_buf *bp)
102 : {
103 28075652 : xfs_inode_buf_verify(bp, false);
104 28075652 : }
105 :
106 : const struct xfs_buf_ops xfs_inode_buf_ops = {
107 : .name = "xfs_inode",
108 : .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
109 : cpu_to_be16(XFS_DINODE_MAGIC) },
110 : .verify_read = xfs_inode_buf_read_verify,
111 : .verify_write = xfs_inode_buf_write_verify,
112 : };
113 :
114 : const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
115 : .name = "xfs_inode_ra",
116 : .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
117 : cpu_to_be16(XFS_DINODE_MAGIC) },
118 : .verify_read = xfs_inode_buf_readahead_verify,
119 : .verify_write = xfs_inode_buf_write_verify,
120 : };
121 :
122 :
123 : /*
124 : * This routine is called to map an inode to the buffer containing the on-disk
125 : * version of the inode. It returns a pointer to the buffer containing the
126 : * on-disk inode in the bpp parameter.
127 : */
128 : int
129 710140465 : xfs_imap_to_bp(
130 : struct xfs_mount *mp,
131 : struct xfs_trans *tp,
132 : struct xfs_imap *imap,
133 : struct xfs_buf **bpp)
134 : {
135 1420011739 : return xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
136 710140465 : imap->im_len, XBF_UNMAPPED, bpp,
137 : &xfs_inode_buf_ops);
138 : }
139 :
140 : static inline struct timespec64 xfs_inode_decode_bigtime(uint64_t ts)
141 : {
142 3085436388 : struct timespec64 tv;
143 3085436388 : uint32_t n;
144 :
145 3085436388 : tv.tv_sec = xfs_bigtime_to_unix(div_u64_rem(ts, NSEC_PER_SEC, &n));
146 3085436388 : tv.tv_nsec = n;
147 :
148 3085436388 : return tv;
149 : }
150 :
151 : /* Convert an ondisk timestamp to an incore timestamp. */
152 : struct timespec64
153 3085678186 : xfs_inode_from_disk_ts(
154 : struct xfs_dinode *dip,
155 : const xfs_timestamp_t ts)
156 : {
157 3085678186 : struct timespec64 tv;
158 3085678186 : struct xfs_legacy_timestamp *lts;
159 :
160 6171356372 : if (xfs_dinode_has_bigtime(dip))
161 3085436388 : return xfs_inode_decode_bigtime(be64_to_cpu(ts));
162 :
163 241798 : lts = (struct xfs_legacy_timestamp *)&ts;
164 241798 : tv.tv_sec = (int)be32_to_cpu(lts->t_sec);
165 241798 : tv.tv_nsec = (int)be32_to_cpu(lts->t_nsec);
166 :
167 241798 : return tv;
168 : }
169 :
170 : int
171 392450220 : xfs_inode_from_disk(
172 : struct xfs_inode *ip,
173 : struct xfs_dinode *from)
174 : {
175 392450220 : struct inode *inode = VFS_I(ip);
176 392450220 : int error;
177 392450220 : xfs_failaddr_t fa;
178 :
179 392450220 : ASSERT(ip->i_cowfp == NULL);
180 :
181 392450220 : fa = xfs_dinode_verify(ip->i_mount, ip->i_ino, from);
182 392469088 : if (fa) {
183 1083 : xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", from,
184 : sizeof(*from), fa);
185 1083 : return -EFSCORRUPTED;
186 : }
187 :
188 : /*
189 : * First get the permanent information that is needed to allocate an
190 : * inode. If the inode is unused, mode is zero and we shouldn't mess
191 : * with the uninitialized part of it.
192 : */
193 392468005 : if (!xfs_has_v3inodes(ip->i_mount))
194 2087 : ip->i_flushiter = be16_to_cpu(from->di_flushiter);
195 392468005 : inode->i_generation = be32_to_cpu(from->di_gen);
196 392468005 : inode->i_mode = be16_to_cpu(from->di_mode);
197 392468005 : if (!inode->i_mode)
198 : return 0;
199 :
200 : /*
201 : * Convert v1 inodes immediately to v2 inode format as this is the
202 : * minimum inode version format we support in the rest of the code.
203 : * They will also be unconditionally written back to disk as v2 inodes.
204 : */
205 392466621 : if (unlikely(from->di_version == 1)) {
206 0 : set_nlink(inode, be16_to_cpu(from->di_onlink));
207 0 : ip->i_projid = 0;
208 : } else {
209 392466621 : set_nlink(inode, be32_to_cpu(from->di_nlink));
210 392435636 : ip->i_projid = (prid_t)be16_to_cpu(from->di_projid_hi) << 16 |
211 392435636 : be16_to_cpu(from->di_projid_lo);
212 : }
213 :
214 392435636 : i_uid_write(inode, be32_to_cpu(from->di_uid));
215 392470539 : i_gid_write(inode, be32_to_cpu(from->di_gid));
216 :
217 : /*
218 : * Time is signed, so need to convert to signed 32 bit before
219 : * storing in inode timestamp which may be 64 bit. Otherwise
220 : * a time before epoch is converted to a time long after epoch
221 : * on 64 bit systems.
222 : */
223 392502009 : inode->i_atime = xfs_inode_from_disk_ts(from, from->di_atime);
224 392502009 : inode->i_mtime = xfs_inode_from_disk_ts(from, from->di_mtime);
225 392502009 : inode->i_ctime = xfs_inode_from_disk_ts(from, from->di_ctime);
226 :
227 392502009 : ip->i_disk_size = be64_to_cpu(from->di_size);
228 392502009 : ip->i_nblocks = be64_to_cpu(from->di_nblocks);
229 392502009 : ip->i_extsize = be32_to_cpu(from->di_extsize);
230 392502009 : ip->i_forkoff = from->di_forkoff;
231 392502009 : ip->i_diflags = be16_to_cpu(from->di_flags);
232 392502009 : ip->i_next_unlinked = be32_to_cpu(from->di_next_unlinked);
233 :
234 392502009 : if (from->di_dmevmask || from->di_dmstate)
235 4 : xfs_iflags_set(ip, XFS_IPRESERVE_DM_FIELDS);
236 :
237 392507649 : if (xfs_has_v3inodes(ip->i_mount)) {
238 392506946 : inode_set_iversion_queried(inode,
239 392506946 : be64_to_cpu(from->di_changecount));
240 392506946 : ip->i_crtime = xfs_inode_from_disk_ts(from, from->di_crtime);
241 392506946 : ip->i_diflags2 = be64_to_cpu(from->di_flags2);
242 392506946 : ip->i_cowextsize = be32_to_cpu(from->di_cowextsize);
243 : }
244 :
245 392507649 : error = xfs_iformat_data_fork(ip, from);
246 392470161 : if (error)
247 : return error;
248 392470139 : if (from->di_forkoff) {
249 44343407 : error = xfs_iformat_attr_fork(ip, from);
250 44343356 : if (error)
251 0 : goto out_destroy_data_fork;
252 : }
253 392470088 : if (xfs_is_reflink_inode(ip))
254 94297698 : xfs_ifork_init_cow(ip);
255 : return 0;
256 :
257 : out_destroy_data_fork:
258 0 : xfs_idestroy_fork(&ip->i_df);
259 0 : return error;
260 : }
261 :
262 : /* Convert an incore timestamp to an ondisk timestamp. */
263 : static inline xfs_timestamp_t
264 2213950516 : xfs_inode_to_disk_ts(
265 : struct xfs_inode *ip,
266 : const struct timespec64 tv)
267 : {
268 2213950516 : struct xfs_legacy_timestamp *lts;
269 2213950516 : xfs_timestamp_t ts;
270 :
271 2213950516 : if (xfs_inode_has_bigtime(ip))
272 2213743057 : return cpu_to_be64(xfs_inode_encode_bigtime(tv));
273 :
274 207459 : lts = (struct xfs_legacy_timestamp *)&ts;
275 207459 : lts->t_sec = cpu_to_be32(tv.tv_sec);
276 207459 : lts->t_nsec = cpu_to_be32(tv.tv_nsec);
277 :
278 207459 : return ts;
279 : }
280 :
281 : static inline void
282 553416013 : xfs_inode_to_disk_iext_counters(
283 : struct xfs_inode *ip,
284 : struct xfs_dinode *to)
285 : {
286 553416013 : if (xfs_inode_has_large_extent_counts(ip)) {
287 552779014 : to->di_big_nextents = cpu_to_be64(xfs_ifork_nextents(&ip->i_df));
288 552779014 : to->di_big_anextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_af));
289 : /*
290 : * We might be upgrading the inode to use larger extent counters
291 : * than was previously used. Hence zero the unused field.
292 : */
293 552779014 : to->di_nrext64_pad = cpu_to_be16(0);
294 : } else {
295 636999 : to->di_nextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_df));
296 1273998 : to->di_anextents = cpu_to_be16(xfs_ifork_nextents(&ip->i_af));
297 : }
298 553416013 : }
299 :
300 : void
301 553962593 : xfs_inode_to_disk(
302 : struct xfs_inode *ip,
303 : struct xfs_dinode *to,
304 : xfs_lsn_t lsn)
305 : {
306 553962593 : struct inode *inode = VFS_I(ip);
307 :
308 553962593 : to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
309 553962593 : to->di_onlink = 0;
310 :
311 553962593 : to->di_format = xfs_ifork_format(&ip->i_df);
312 553962593 : to->di_uid = cpu_to_be32(i_uid_read(inode));
313 553755159 : to->di_gid = cpu_to_be32(i_gid_read(inode));
314 553863227 : to->di_projid_lo = cpu_to_be16(ip->i_projid & 0xffff);
315 553863227 : to->di_projid_hi = cpu_to_be16(ip->i_projid >> 16);
316 :
317 553863227 : to->di_atime = xfs_inode_to_disk_ts(ip, inode->i_atime);
318 553863227 : to->di_mtime = xfs_inode_to_disk_ts(ip, inode->i_mtime);
319 553863227 : to->di_ctime = xfs_inode_to_disk_ts(ip, inode->i_ctime);
320 553863227 : to->di_nlink = cpu_to_be32(inode->i_nlink);
321 553863227 : to->di_gen = cpu_to_be32(inode->i_generation);
322 553863227 : to->di_mode = cpu_to_be16(inode->i_mode);
323 :
324 553863227 : to->di_size = cpu_to_be64(ip->i_disk_size);
325 553863227 : to->di_nblocks = cpu_to_be64(ip->i_nblocks);
326 553863227 : to->di_extsize = cpu_to_be32(ip->i_extsize);
327 553863227 : to->di_forkoff = ip->i_forkoff;
328 553863227 : to->di_aformat = xfs_ifork_format(&ip->i_af);
329 553863227 : to->di_flags = cpu_to_be16(ip->i_diflags);
330 :
331 553863227 : if (xfs_has_v3inodes(ip->i_mount)) {
332 553861854 : to->di_version = 3;
333 553861854 : to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
334 553861854 : to->di_crtime = xfs_inode_to_disk_ts(ip, ip->i_crtime);
335 553861854 : to->di_flags2 = cpu_to_be64(ip->i_diflags2);
336 553861854 : to->di_cowextsize = cpu_to_be32(ip->i_cowextsize);
337 553861854 : to->di_ino = cpu_to_be64(ip->i_ino);
338 553861854 : to->di_lsn = cpu_to_be64(lsn);
339 553861854 : memset(to->di_pad2, 0, sizeof(to->di_pad2));
340 553861854 : uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
341 553289297 : to->di_v3_pad = 0;
342 : } else {
343 1373 : to->di_version = 2;
344 1373 : to->di_flushiter = cpu_to_be16(ip->i_flushiter);
345 2746 : memset(to->di_v2_pad, 0, sizeof(to->di_v2_pad));
346 : }
347 :
348 553290670 : xfs_inode_to_disk_iext_counters(ip, to);
349 553439768 : }
350 :
351 : static xfs_failaddr_t
352 305518420 : xfs_dinode_verify_fork(
353 : struct xfs_dinode *dip,
354 : struct xfs_mount *mp,
355 : int whichfork)
356 : {
357 305518420 : xfs_extnum_t di_nextents;
358 305518420 : xfs_extnum_t max_extents;
359 305518420 : mode_t mode = be16_to_cpu(dip->di_mode);
360 305519134 : uint32_t fork_size = XFS_DFORK_SIZE(dip, mp, whichfork);
361 305518420 : uint32_t fork_format = XFS_DFORK_FORMAT(dip, whichfork);
362 :
363 305518420 : di_nextents = xfs_dfork_nextents(dip, whichfork);
364 :
365 : /*
366 : * For fork types that can contain local data, check that the fork
367 : * format matches the size of local data contained within the fork.
368 : *
369 : * For all types, check that when the size says the should be in extent
370 : * or btree format, the inode isn't claiming it is in local format.
371 : */
372 305524484 : if (whichfork == XFS_DATA_FORK) {
373 261185649 : if (S_ISDIR(mode) || S_ISLNK(mode)) {
374 83812015 : if (be64_to_cpu(dip->di_size) <= fork_size &&
375 : fork_format != XFS_DINODE_FMT_LOCAL)
376 33 : return __this_address;
377 : }
378 :
379 261185616 : if (be64_to_cpu(dip->di_size) > fork_size &&
380 : fork_format == XFS_DINODE_FMT_LOCAL)
381 0 : return __this_address;
382 : }
383 :
384 305524451 : switch (fork_format) {
385 112437158 : case XFS_DINODE_FMT_LOCAL:
386 : /*
387 : * No local regular files yet.
388 : */
389 112437158 : if (S_ISREG(mode) && whichfork == XFS_DATA_FORK)
390 22 : return __this_address;
391 112437136 : if (di_nextents)
392 0 : return __this_address;
393 : break;
394 187159654 : case XFS_DINODE_FMT_EXTENTS:
395 187160140 : if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
396 0 : return __this_address;
397 : break;
398 : case XFS_DINODE_FMT_BTREE:
399 11855190 : max_extents = xfs_iext_max_nextents(
400 : xfs_dinode_has_large_extent_counts(dip),
401 : whichfork);
402 5927593 : if (di_nextents > max_extents)
403 0 : return __this_address;
404 : break;
405 : default:
406 44 : return __this_address;
407 : }
408 : return NULL;
409 : }
410 :
411 : static xfs_failaddr_t
412 392443315 : xfs_dinode_verify_forkoff(
413 : struct xfs_dinode *dip,
414 : struct xfs_mount *mp)
415 : {
416 392443315 : if (!dip->di_forkoff)
417 : return NULL;
418 :
419 44343222 : switch (dip->di_format) {
420 50 : case XFS_DINODE_FMT_DEV:
421 50 : if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
422 0 : return __this_address;
423 : break;
424 44343172 : case XFS_DINODE_FMT_LOCAL: /* fall through ... */
425 : case XFS_DINODE_FMT_EXTENTS: /* fall through ... */
426 : case XFS_DINODE_FMT_BTREE:
427 44343183 : if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3))
428 0 : return __this_address;
429 : break;
430 : default:
431 0 : return __this_address;
432 : }
433 : return NULL;
434 : }
435 :
436 : static xfs_failaddr_t
437 392428211 : xfs_dinode_verify_nrext64(
438 : struct xfs_mount *mp,
439 : struct xfs_dinode *dip)
440 : {
441 784856422 : if (xfs_dinode_has_large_extent_counts(dip)) {
442 391690330 : if (!xfs_has_large_extent_counts(mp))
443 0 : return __this_address;
444 391690330 : if (dip->di_nrext64_pad != 0)
445 0 : return __this_address;
446 737881 : } else if (dip->di_version >= 3) {
447 735783 : if (dip->di_v3_pad != 0)
448 0 : return __this_address;
449 : }
450 :
451 : return NULL;
452 : }
453 :
454 : xfs_failaddr_t
455 392493480 : xfs_dinode_verify(
456 : struct xfs_mount *mp,
457 : xfs_ino_t ino,
458 : struct xfs_dinode *dip)
459 : {
460 392493480 : xfs_failaddr_t fa;
461 392493480 : uint16_t mode;
462 392493480 : uint16_t flags;
463 392493480 : uint64_t flags2;
464 392493480 : uint64_t di_size;
465 392493480 : xfs_extnum_t nextents;
466 392493480 : xfs_extnum_t naextents;
467 392493480 : xfs_filblks_t nblocks;
468 :
469 392493480 : if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
470 0 : return __this_address;
471 :
472 : /* Verify v3 integrity information first */
473 392493480 : if (dip->di_version >= 3) {
474 392471237 : if (!xfs_has_v3inodes(mp))
475 0 : return __this_address;
476 392471237 : if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
477 : XFS_DINODE_CRC_OFF))
478 27 : return __this_address;
479 392463680 : if (be64_to_cpu(dip->di_ino) != ino)
480 0 : return __this_address;
481 392463680 : if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
482 0 : return __this_address;
483 : }
484 :
485 : /* don't allow invalid i_size */
486 392440196 : di_size = be64_to_cpu(dip->di_size);
487 392440196 : if (di_size & (1ULL << 63))
488 0 : return __this_address;
489 :
490 392440196 : mode = be16_to_cpu(dip->di_mode);
491 392440196 : if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
492 693 : return __this_address;
493 :
494 : /* No zero-length symlinks/dirs. */
495 392438786 : if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
496 88 : return __this_address;
497 :
498 392438698 : fa = xfs_dinode_verify_nrext64(mp, dip);
499 392424760 : if (fa)
500 : return fa;
501 :
502 392436537 : nextents = xfs_dfork_data_extents(dip);
503 392436537 : naextents = xfs_dfork_attr_extents(dip);
504 392436537 : nblocks = be64_to_cpu(dip->di_nblocks);
505 :
506 : /* Fork checks carried over from xfs_iformat_fork */
507 392436537 : if (mode && nextents + naextents > nblocks)
508 0 : return __this_address;
509 :
510 392436537 : if (S_ISDIR(mode) && nextents > mp->m_dir_geo->max_extents)
511 0 : return __this_address;
512 :
513 392436537 : if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
514 0 : return __this_address;
515 :
516 392436537 : flags = be16_to_cpu(dip->di_flags);
517 :
518 392436537 : if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
519 0 : return __this_address;
520 :
521 : /* check for illegal values of forkoff */
522 392436537 : fa = xfs_dinode_verify_forkoff(dip, mp);
523 392405450 : if (fa)
524 : return fa;
525 :
526 : /* Do we have appropriate data fork formats for the mode? */
527 392446554 : switch (mode & S_IFMT) {
528 131282172 : case S_IFIFO:
529 : case S_IFCHR:
530 : case S_IFBLK:
531 : case S_IFSOCK:
532 131282172 : if (dip->di_format != XFS_DINODE_FMT_DEV)
533 176 : return __this_address;
534 : break;
535 261167303 : case S_IFREG:
536 : case S_IFLNK:
537 : case S_IFDIR:
538 261167303 : fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
539 261170832 : if (fa)
540 : return fa;
541 : break;
542 : case 0:
543 : /* Uninitialized inode ok. */
544 : break;
545 : default:
546 0 : return __this_address;
547 : }
548 :
549 392451058 : if (dip->di_forkoff) {
550 44343204 : fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
551 44343247 : if (fa)
552 : return fa;
553 : } else {
554 : /*
555 : * If there is no fork offset, this may be a freshly-made inode
556 : * in a new disk cluster, in which case di_aformat is zeroed.
557 : * Otherwise, such an inode must be in EXTENTS format; this goes
558 : * for freed inodes as well.
559 : */
560 348107854 : switch (dip->di_aformat) {
561 : case 0:
562 : case XFS_DINODE_FMT_EXTENTS:
563 348107854 : break;
564 : default:
565 0 : return __this_address;
566 : }
567 348107854 : if (naextents)
568 0 : return __this_address;
569 : }
570 :
571 : /* extent size hint validation */
572 392451098 : fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
573 : mode, flags);
574 392442703 : if (fa)
575 : return fa;
576 :
577 : /* only version 3 or greater inodes are extensively verified here */
578 392450964 : if (dip->di_version < 3)
579 : return NULL;
580 :
581 392448877 : flags2 = be64_to_cpu(dip->di_flags2);
582 :
583 : /* don't allow reflink/cowextsize if we don't have reflink */
584 392448877 : if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
585 : !xfs_has_reflink(mp))
586 0 : return __this_address;
587 :
588 : /* only regular files get reflink */
589 392448877 : if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
590 0 : return __this_address;
591 :
592 : /* don't let reflink and realtime mix */
593 392448877 : if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
594 0 : return __this_address;
595 :
596 : /* COW extent size hint validation */
597 392448877 : fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
598 : mode, flags, flags2);
599 392442887 : if (fa)
600 : return fa;
601 :
602 : /* bigtime iflag can only happen on bigtime filesystems */
603 784890550 : if (xfs_dinode_has_bigtime(dip) &&
604 : !xfs_has_bigtime(mp))
605 0 : return __this_address;
606 :
607 : return NULL;
608 : }
609 :
610 : void
611 288965394 : xfs_dinode_calc_crc(
612 : struct xfs_mount *mp,
613 : struct xfs_dinode *dip)
614 : {
615 288965394 : uint32_t crc;
616 :
617 288965394 : if (dip->di_version < 3)
618 : return;
619 :
620 288964109 : ASSERT(xfs_has_crc(mp));
621 288964109 : crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
622 : XFS_DINODE_CRC_OFF);
623 288877029 : dip->di_crc = xfs_end_cksum(crc);
624 : }
625 :
626 : /*
627 : * Validate di_extsize hint.
628 : *
629 : * 1. Extent size hint is only valid for directories and regular files.
630 : * 2. FS_XFLAG_EXTSIZE is only valid for regular files.
631 : * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
632 : * 4. Hint cannot be larger than MAXTEXTLEN.
633 : * 5. Can be changed on directories at any time.
634 : * 6. Hint value of 0 turns off hints, clears inode flags.
635 : * 7. Extent size must be a multiple of the appropriate block size.
636 : * For realtime files, this is the rt extent size.
637 : * 8. For non-realtime files, the extent size hint must be limited
638 : * to half the AG size to avoid alignment extending the extent beyond the
639 : * limits of the AG.
640 : */
641 : xfs_failaddr_t
642 794448137 : xfs_inode_validate_extsize(
643 : struct xfs_mount *mp,
644 : uint32_t extsize,
645 : uint16_t mode,
646 : uint16_t flags)
647 : {
648 794448137 : bool rt_flag;
649 794448137 : bool hint_flag;
650 794448137 : bool inherit_flag;
651 794448137 : uint32_t extsize_bytes;
652 794448137 : uint32_t blocksize_bytes;
653 :
654 794448137 : rt_flag = (flags & XFS_DIFLAG_REALTIME);
655 794448137 : hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
656 794448137 : inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
657 794448137 : extsize_bytes = XFS_FSB_TO_B(mp, extsize);
658 :
659 : /*
660 : * This comment describes a historic gap in this verifier function.
661 : *
662 : * For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
663 : * function has never checked that the extent size hint is an integer
664 : * multiple of the realtime extent size. Since we allow users to set
665 : * this combination on non-rt filesystems /and/ to change the rt
666 : * extent size when adding a rt device to a filesystem, the net effect
667 : * is that users can configure a filesystem anticipating one rt
668 : * geometry and change their minds later. Directories do not use the
669 : * extent size hint, so this is harmless for them.
670 : *
671 : * If a directory with a misaligned extent size hint is allowed to
672 : * propagate that hint into a new regular realtime file, the result
673 : * is that the inode cluster buffer verifier will trigger a corruption
674 : * shutdown the next time it is run, because the verifier has always
675 : * enforced the alignment rule for regular files.
676 : *
677 : * Because we allow administrators to set a new rt extent size when
678 : * adding a rt section, we cannot add a check to this verifier because
679 : * that will result a new source of directory corruption errors when
680 : * reading an existing filesystem. Instead, we rely on callers to
681 : * decide when alignment checks are appropriate, and fix things up as
682 : * needed.
683 : */
684 :
685 794448137 : if (rt_flag)
686 115696032 : blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
687 : else
688 678752105 : blocksize_bytes = mp->m_sb.sb_blocksize;
689 :
690 794448137 : if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
691 0 : return __this_address;
692 :
693 794448137 : if (hint_flag && !S_ISREG(mode))
694 0 : return __this_address;
695 :
696 794448137 : if (inherit_flag && !S_ISDIR(mode))
697 0 : return __this_address;
698 :
699 794448137 : if ((hint_flag || inherit_flag) && extsize == 0)
700 0 : return __this_address;
701 :
702 : /* free inodes get flags set to zero but extsize remains */
703 794448137 : if (mode && !(hint_flag || inherit_flag) && extsize != 0)
704 0 : return __this_address;
705 :
706 794448137 : if (extsize_bytes % blocksize_bytes)
707 3 : return __this_address;
708 :
709 794448134 : if (extsize > XFS_MAX_BMBT_EXTLEN)
710 0 : return __this_address;
711 :
712 794448134 : if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
713 0 : return __this_address;
714 :
715 : return NULL;
716 : }
717 :
718 : /*
719 : * Validate di_cowextsize hint.
720 : *
721 : * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
722 : * The inode does not have to have any shared blocks, but it must be a v3.
723 : * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
724 : * for a directory, the hint is propagated to new files.
725 : * 3. Can be changed on files & directories at any time.
726 : * 4. Hint value of 0 turns off hints, clears inode flags.
727 : * 5. Extent size must be a multiple of the appropriate block size.
728 : * 6. The extent size hint must be limited to half the AG size to avoid
729 : * alignment extending the extent beyond the limits of the AG.
730 : */
731 : xfs_failaddr_t
732 822995746 : xfs_inode_validate_cowextsize(
733 : struct xfs_mount *mp,
734 : uint32_t cowextsize,
735 : uint16_t mode,
736 : uint16_t flags,
737 : uint64_t flags2)
738 : {
739 822995746 : bool rt_flag;
740 822995746 : bool hint_flag;
741 822995746 : uint32_t cowextsize_bytes;
742 :
743 822995746 : rt_flag = (flags & XFS_DIFLAG_REALTIME);
744 822995746 : hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
745 822995746 : cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
746 :
747 822995746 : if (hint_flag && !xfs_has_reflink(mp))
748 0 : return __this_address;
749 :
750 822995746 : if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
751 0 : return __this_address;
752 :
753 822995746 : if (hint_flag && cowextsize == 0)
754 0 : return __this_address;
755 :
756 : /* free inodes get flags set to zero but cowextsize remains */
757 822995746 : if (mode && !hint_flag && cowextsize != 0)
758 0 : return __this_address;
759 :
760 822995746 : if (hint_flag && rt_flag)
761 0 : return __this_address;
762 :
763 822995746 : if (cowextsize_bytes % mp->m_sb.sb_blocksize)
764 0 : return __this_address;
765 :
766 822995746 : if (cowextsize > XFS_MAX_BMBT_EXTLEN)
767 0 : return __this_address;
768 :
769 822995746 : if (cowextsize > mp->m_sb.sb_agblocks / 2)
770 0 : return __this_address;
771 :
772 : return NULL;
773 : }
|
