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
xfs_inode_buf_verify(struct xfs_buf * bp,bool readahead)40 xfs_inode_buf_verify(
41 struct xfs_buf *bp,
42 bool readahead)
43 {
44 struct xfs_mount *mp = bp->b_mount;
45 int i;
46 int ni;
47
48 /*
49 * Validate the magic number and version of every inode in the buffer
50 */
51 ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
52 for (i = 0; i < ni; i++) {
53 struct xfs_dinode *dip;
54 xfs_agino_t unlinked_ino;
55 int di_ok;
56
57 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
58 unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
59 di_ok = xfs_verify_magic16(bp, dip->di_magic) &&
60 xfs_dinode_good_version(mp, dip->di_version) &&
61 xfs_verify_agino_or_null(bp->b_pag, unlinked_ino);
62 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
63 XFS_ERRTAG_ITOBP_INOTOBP))) {
64 if (readahead) {
65 bp->b_flags &= ~XBF_DONE;
66 xfs_buf_ioerror(bp, -EIO);
67 return;
68 }
69
70 #ifdef DEBUG
71 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 xfs_buf_verifier_error(bp, -EFSCORRUPTED,
77 __func__, dip, sizeof(*dip),
78 NULL);
79 return;
80 }
81 }
82 }
83
84
85 static void
xfs_inode_buf_read_verify(struct xfs_buf * bp)86 xfs_inode_buf_read_verify(
87 struct xfs_buf *bp)
88 {
89 xfs_inode_buf_verify(bp, false);
90 }
91
92 static void
xfs_inode_buf_readahead_verify(struct xfs_buf * bp)93 xfs_inode_buf_readahead_verify(
94 struct xfs_buf *bp)
95 {
96 xfs_inode_buf_verify(bp, true);
97 }
98
99 static void
xfs_inode_buf_write_verify(struct xfs_buf * bp)100 xfs_inode_buf_write_verify(
101 struct xfs_buf *bp)
102 {
103 xfs_inode_buf_verify(bp, false);
104 }
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
xfs_imap_to_bp(struct xfs_mount * mp,struct xfs_trans * tp,struct xfs_imap * imap,struct xfs_buf ** bpp)129 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 return xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
136 imap->im_len, XBF_UNMAPPED, bpp,
137 &xfs_inode_buf_ops);
138 }
139
xfs_inode_decode_bigtime(uint64_t ts)140 static inline struct timespec64 xfs_inode_decode_bigtime(uint64_t ts)
141 {
142 struct timespec64 tv;
143 uint32_t n;
144
145 tv.tv_sec = xfs_bigtime_to_unix(div_u64_rem(ts, NSEC_PER_SEC, &n));
146 tv.tv_nsec = n;
147
148 return tv;
149 }
150
151 /* Convert an ondisk timestamp to an incore timestamp. */
152 struct timespec64
xfs_inode_from_disk_ts(struct xfs_dinode * dip,const xfs_timestamp_t ts)153 xfs_inode_from_disk_ts(
154 struct xfs_dinode *dip,
155 const xfs_timestamp_t ts)
156 {
157 struct timespec64 tv;
158 struct xfs_legacy_timestamp *lts;
159
160 if (xfs_dinode_has_bigtime(dip))
161 return xfs_inode_decode_bigtime(be64_to_cpu(ts));
162
163 lts = (struct xfs_legacy_timestamp *)&ts;
164 tv.tv_sec = (int)be32_to_cpu(lts->t_sec);
165 tv.tv_nsec = (int)be32_to_cpu(lts->t_nsec);
166
167 return tv;
168 }
169
170 int
xfs_inode_from_disk(struct xfs_inode * ip,struct xfs_dinode * from)171 xfs_inode_from_disk(
172 struct xfs_inode *ip,
173 struct xfs_dinode *from)
174 {
175 struct inode *inode = VFS_I(ip);
176 int error;
177 xfs_failaddr_t fa;
178
179 ASSERT(ip->i_cowfp == NULL);
180
181 fa = xfs_dinode_verify(ip->i_mount, ip->i_ino, from);
182 if (fa) {
183 xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", from,
184 sizeof(*from), fa);
185 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 if (!xfs_has_v3inodes(ip->i_mount))
194 ip->i_flushiter = be16_to_cpu(from->di_flushiter);
195 inode->i_generation = be32_to_cpu(from->di_gen);
196 inode->i_mode = be16_to_cpu(from->di_mode);
197 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 if (unlikely(from->di_version == 1)) {
206 set_nlink(inode, be16_to_cpu(from->di_onlink));
207 ip->i_projid = 0;
208 } else {
209 set_nlink(inode, be32_to_cpu(from->di_nlink));
210 ip->i_projid = (prid_t)be16_to_cpu(from->di_projid_hi) << 16 |
211 be16_to_cpu(from->di_projid_lo);
212 }
213
214 i_uid_write(inode, be32_to_cpu(from->di_uid));
215 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 inode->i_atime = xfs_inode_from_disk_ts(from, from->di_atime);
224 inode->i_mtime = xfs_inode_from_disk_ts(from, from->di_mtime);
225 inode_set_ctime_to_ts(inode,
226 xfs_inode_from_disk_ts(from, from->di_ctime));
227
228 ip->i_disk_size = be64_to_cpu(from->di_size);
229 ip->i_nblocks = be64_to_cpu(from->di_nblocks);
230 ip->i_extsize = be32_to_cpu(from->di_extsize);
231 ip->i_forkoff = from->di_forkoff;
232 ip->i_diflags = be16_to_cpu(from->di_flags);
233 ip->i_next_unlinked = be32_to_cpu(from->di_next_unlinked);
234
235 if (from->di_dmevmask || from->di_dmstate)
236 xfs_iflags_set(ip, XFS_IPRESERVE_DM_FIELDS);
237
238 if (xfs_has_v3inodes(ip->i_mount)) {
239 inode_set_iversion_queried(inode,
240 be64_to_cpu(from->di_changecount));
241 ip->i_crtime = xfs_inode_from_disk_ts(from, from->di_crtime);
242 ip->i_diflags2 = be64_to_cpu(from->di_flags2);
243 ip->i_cowextsize = be32_to_cpu(from->di_cowextsize);
244 }
245
246 error = xfs_iformat_data_fork(ip, from);
247 if (error)
248 return error;
249 if (from->di_forkoff) {
250 error = xfs_iformat_attr_fork(ip, from);
251 if (error)
252 goto out_destroy_data_fork;
253 }
254 if (xfs_is_reflink_inode(ip))
255 xfs_ifork_init_cow(ip);
256 return 0;
257
258 out_destroy_data_fork:
259 xfs_idestroy_fork(&ip->i_df);
260 return error;
261 }
262
263 /* Convert an incore timestamp to an ondisk timestamp. */
264 static inline xfs_timestamp_t
xfs_inode_to_disk_ts(struct xfs_inode * ip,const struct timespec64 tv)265 xfs_inode_to_disk_ts(
266 struct xfs_inode *ip,
267 const struct timespec64 tv)
268 {
269 struct xfs_legacy_timestamp *lts;
270 xfs_timestamp_t ts;
271
272 if (xfs_inode_has_bigtime(ip))
273 return cpu_to_be64(xfs_inode_encode_bigtime(tv));
274
275 lts = (struct xfs_legacy_timestamp *)&ts;
276 lts->t_sec = cpu_to_be32(tv.tv_sec);
277 lts->t_nsec = cpu_to_be32(tv.tv_nsec);
278
279 return ts;
280 }
281
282 static inline void
xfs_inode_to_disk_iext_counters(struct xfs_inode * ip,struct xfs_dinode * to)283 xfs_inode_to_disk_iext_counters(
284 struct xfs_inode *ip,
285 struct xfs_dinode *to)
286 {
287 if (xfs_inode_has_large_extent_counts(ip)) {
288 to->di_big_nextents = cpu_to_be64(xfs_ifork_nextents(&ip->i_df));
289 to->di_big_anextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_af));
290 /*
291 * We might be upgrading the inode to use larger extent counters
292 * than was previously used. Hence zero the unused field.
293 */
294 to->di_nrext64_pad = cpu_to_be16(0);
295 } else {
296 to->di_nextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_df));
297 to->di_anextents = cpu_to_be16(xfs_ifork_nextents(&ip->i_af));
298 }
299 }
300
301 void
xfs_inode_to_disk(struct xfs_inode * ip,struct xfs_dinode * to,xfs_lsn_t lsn)302 xfs_inode_to_disk(
303 struct xfs_inode *ip,
304 struct xfs_dinode *to,
305 xfs_lsn_t lsn)
306 {
307 struct inode *inode = VFS_I(ip);
308
309 to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
310 to->di_onlink = 0;
311
312 to->di_format = xfs_ifork_format(&ip->i_df);
313 to->di_uid = cpu_to_be32(i_uid_read(inode));
314 to->di_gid = cpu_to_be32(i_gid_read(inode));
315 to->di_projid_lo = cpu_to_be16(ip->i_projid & 0xffff);
316 to->di_projid_hi = cpu_to_be16(ip->i_projid >> 16);
317
318 to->di_atime = xfs_inode_to_disk_ts(ip, inode->i_atime);
319 to->di_mtime = xfs_inode_to_disk_ts(ip, inode->i_mtime);
320 to->di_ctime = xfs_inode_to_disk_ts(ip, inode_get_ctime(inode));
321 to->di_nlink = cpu_to_be32(inode->i_nlink);
322 to->di_gen = cpu_to_be32(inode->i_generation);
323 to->di_mode = cpu_to_be16(inode->i_mode);
324
325 to->di_size = cpu_to_be64(ip->i_disk_size);
326 to->di_nblocks = cpu_to_be64(ip->i_nblocks);
327 to->di_extsize = cpu_to_be32(ip->i_extsize);
328 to->di_forkoff = ip->i_forkoff;
329 to->di_aformat = xfs_ifork_format(&ip->i_af);
330 to->di_flags = cpu_to_be16(ip->i_diflags);
331
332 if (xfs_has_v3inodes(ip->i_mount)) {
333 to->di_version = 3;
334 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
335 to->di_crtime = xfs_inode_to_disk_ts(ip, ip->i_crtime);
336 to->di_flags2 = cpu_to_be64(ip->i_diflags2);
337 to->di_cowextsize = cpu_to_be32(ip->i_cowextsize);
338 to->di_ino = cpu_to_be64(ip->i_ino);
339 to->di_lsn = cpu_to_be64(lsn);
340 memset(to->di_pad2, 0, sizeof(to->di_pad2));
341 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
342 to->di_v3_pad = 0;
343 } else {
344 to->di_version = 2;
345 to->di_flushiter = cpu_to_be16(ip->i_flushiter);
346 memset(to->di_v2_pad, 0, sizeof(to->di_v2_pad));
347 }
348
349 xfs_inode_to_disk_iext_counters(ip, to);
350 }
351
352 static xfs_failaddr_t
xfs_dinode_verify_fork(struct xfs_dinode * dip,struct xfs_mount * mp,int whichfork)353 xfs_dinode_verify_fork(
354 struct xfs_dinode *dip,
355 struct xfs_mount *mp,
356 int whichfork)
357 {
358 xfs_extnum_t di_nextents;
359 xfs_extnum_t max_extents;
360 mode_t mode = be16_to_cpu(dip->di_mode);
361 uint32_t fork_size = XFS_DFORK_SIZE(dip, mp, whichfork);
362 uint32_t fork_format = XFS_DFORK_FORMAT(dip, whichfork);
363
364 di_nextents = xfs_dfork_nextents(dip, whichfork);
365
366 /*
367 * For fork types that can contain local data, check that the fork
368 * format matches the size of local data contained within the fork.
369 *
370 * For all types, check that when the size says the should be in extent
371 * or btree format, the inode isn't claiming it is in local format.
372 */
373 if (whichfork == XFS_DATA_FORK) {
374 if (S_ISDIR(mode) || S_ISLNK(mode)) {
375 if (be64_to_cpu(dip->di_size) <= fork_size &&
376 fork_format != XFS_DINODE_FMT_LOCAL)
377 return __this_address;
378 }
379
380 if (be64_to_cpu(dip->di_size) > fork_size &&
381 fork_format == XFS_DINODE_FMT_LOCAL)
382 return __this_address;
383 }
384
385 switch (fork_format) {
386 case XFS_DINODE_FMT_LOCAL:
387 /*
388 * No local regular files yet.
389 */
390 if (S_ISREG(mode) && whichfork == XFS_DATA_FORK)
391 return __this_address;
392 if (di_nextents)
393 return __this_address;
394 break;
395 case XFS_DINODE_FMT_EXTENTS:
396 if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
397 return __this_address;
398 break;
399 case XFS_DINODE_FMT_BTREE:
400 max_extents = xfs_iext_max_nextents(
401 xfs_dinode_has_large_extent_counts(dip),
402 whichfork);
403 if (di_nextents > max_extents)
404 return __this_address;
405 break;
406 default:
407 return __this_address;
408 }
409 return NULL;
410 }
411
412 static xfs_failaddr_t
xfs_dinode_verify_forkoff(struct xfs_dinode * dip,struct xfs_mount * mp)413 xfs_dinode_verify_forkoff(
414 struct xfs_dinode *dip,
415 struct xfs_mount *mp)
416 {
417 if (!dip->di_forkoff)
418 return NULL;
419
420 switch (dip->di_format) {
421 case XFS_DINODE_FMT_DEV:
422 if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
423 return __this_address;
424 break;
425 case XFS_DINODE_FMT_LOCAL: /* fall through ... */
426 case XFS_DINODE_FMT_EXTENTS: /* fall through ... */
427 case XFS_DINODE_FMT_BTREE:
428 if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3))
429 return __this_address;
430 break;
431 default:
432 return __this_address;
433 }
434 return NULL;
435 }
436
437 static xfs_failaddr_t
xfs_dinode_verify_nrext64(struct xfs_mount * mp,struct xfs_dinode * dip)438 xfs_dinode_verify_nrext64(
439 struct xfs_mount *mp,
440 struct xfs_dinode *dip)
441 {
442 if (xfs_dinode_has_large_extent_counts(dip)) {
443 if (!xfs_has_large_extent_counts(mp))
444 return __this_address;
445 if (dip->di_nrext64_pad != 0)
446 return __this_address;
447 } else if (dip->di_version >= 3) {
448 if (dip->di_v3_pad != 0)
449 return __this_address;
450 }
451
452 return NULL;
453 }
454
455 xfs_failaddr_t
xfs_dinode_verify(struct xfs_mount * mp,xfs_ino_t ino,struct xfs_dinode * dip)456 xfs_dinode_verify(
457 struct xfs_mount *mp,
458 xfs_ino_t ino,
459 struct xfs_dinode *dip)
460 {
461 xfs_failaddr_t fa;
462 uint16_t mode;
463 uint16_t flags;
464 uint64_t flags2;
465 uint64_t di_size;
466 xfs_extnum_t nextents;
467 xfs_extnum_t naextents;
468 xfs_filblks_t nblocks;
469
470 if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
471 return __this_address;
472
473 /* Verify v3 integrity information first */
474 if (dip->di_version >= 3) {
475 if (!xfs_has_v3inodes(mp))
476 return __this_address;
477 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
478 XFS_DINODE_CRC_OFF))
479 return __this_address;
480 if (be64_to_cpu(dip->di_ino) != ino)
481 return __this_address;
482 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
483 return __this_address;
484 }
485
486 /* don't allow invalid i_size */
487 di_size = be64_to_cpu(dip->di_size);
488 if (di_size & (1ULL << 63))
489 return __this_address;
490
491 mode = be16_to_cpu(dip->di_mode);
492 if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
493 return __this_address;
494
495 /* No zero-length symlinks/dirs. */
496 if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
497 return __this_address;
498
499 fa = xfs_dinode_verify_nrext64(mp, dip);
500 if (fa)
501 return fa;
502
503 nextents = xfs_dfork_data_extents(dip);
504 naextents = xfs_dfork_attr_extents(dip);
505 nblocks = be64_to_cpu(dip->di_nblocks);
506
507 /* Fork checks carried over from xfs_iformat_fork */
508 if (mode && nextents + naextents > nblocks)
509 return __this_address;
510
511 if (nextents + naextents == 0 && nblocks != 0)
512 return __this_address;
513
514 if (S_ISDIR(mode) && nextents > mp->m_dir_geo->max_extents)
515 return __this_address;
516
517 if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
518 return __this_address;
519
520 flags = be16_to_cpu(dip->di_flags);
521
522 if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
523 return __this_address;
524
525 /* check for illegal values of forkoff */
526 fa = xfs_dinode_verify_forkoff(dip, mp);
527 if (fa)
528 return fa;
529
530 /* Do we have appropriate data fork formats for the mode? */
531 switch (mode & S_IFMT) {
532 case S_IFIFO:
533 case S_IFCHR:
534 case S_IFBLK:
535 case S_IFSOCK:
536 if (dip->di_format != XFS_DINODE_FMT_DEV)
537 return __this_address;
538 break;
539 case S_IFREG:
540 case S_IFLNK:
541 case S_IFDIR:
542 fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
543 if (fa)
544 return fa;
545 break;
546 case 0:
547 /* Uninitialized inode ok. */
548 break;
549 default:
550 return __this_address;
551 }
552
553 if (dip->di_forkoff) {
554 fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
555 if (fa)
556 return fa;
557 } else {
558 /*
559 * If there is no fork offset, this may be a freshly-made inode
560 * in a new disk cluster, in which case di_aformat is zeroed.
561 * Otherwise, such an inode must be in EXTENTS format; this goes
562 * for freed inodes as well.
563 */
564 switch (dip->di_aformat) {
565 case 0:
566 case XFS_DINODE_FMT_EXTENTS:
567 break;
568 default:
569 return __this_address;
570 }
571 if (naextents)
572 return __this_address;
573 }
574
575 /* extent size hint validation */
576 fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
577 mode, flags);
578 if (fa)
579 return fa;
580
581 /* only version 3 or greater inodes are extensively verified here */
582 if (dip->di_version < 3)
583 return NULL;
584
585 flags2 = be64_to_cpu(dip->di_flags2);
586
587 /* don't allow reflink/cowextsize if we don't have reflink */
588 if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
589 !xfs_has_reflink(mp))
590 return __this_address;
591
592 /* only regular files get reflink */
593 if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
594 return __this_address;
595
596 /* don't let reflink and realtime mix */
597 if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
598 return __this_address;
599
600 /* COW extent size hint validation */
601 fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
602 mode, flags, flags2);
603 if (fa)
604 return fa;
605
606 /* bigtime iflag can only happen on bigtime filesystems */
607 if (xfs_dinode_has_bigtime(dip) &&
608 !xfs_has_bigtime(mp))
609 return __this_address;
610
611 return NULL;
612 }
613
614 void
xfs_dinode_calc_crc(struct xfs_mount * mp,struct xfs_dinode * dip)615 xfs_dinode_calc_crc(
616 struct xfs_mount *mp,
617 struct xfs_dinode *dip)
618 {
619 uint32_t crc;
620
621 if (dip->di_version < 3)
622 return;
623
624 ASSERT(xfs_has_crc(mp));
625 crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
626 XFS_DINODE_CRC_OFF);
627 dip->di_crc = xfs_end_cksum(crc);
628 }
629
630 /*
631 * Validate di_extsize hint.
632 *
633 * 1. Extent size hint is only valid for directories and regular files.
634 * 2. FS_XFLAG_EXTSIZE is only valid for regular files.
635 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
636 * 4. Hint cannot be larger than MAXTEXTLEN.
637 * 5. Can be changed on directories at any time.
638 * 6. Hint value of 0 turns off hints, clears inode flags.
639 * 7. Extent size must be a multiple of the appropriate block size.
640 * For realtime files, this is the rt extent size.
641 * 8. For non-realtime files, the extent size hint must be limited
642 * to half the AG size to avoid alignment extending the extent beyond the
643 * limits of the AG.
644 */
645 xfs_failaddr_t
xfs_inode_validate_extsize(struct xfs_mount * mp,uint32_t extsize,uint16_t mode,uint16_t flags)646 xfs_inode_validate_extsize(
647 struct xfs_mount *mp,
648 uint32_t extsize,
649 uint16_t mode,
650 uint16_t flags)
651 {
652 bool rt_flag;
653 bool hint_flag;
654 bool inherit_flag;
655 uint32_t extsize_bytes;
656 uint32_t blocksize_bytes;
657
658 rt_flag = (flags & XFS_DIFLAG_REALTIME);
659 hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
660 inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
661 extsize_bytes = XFS_FSB_TO_B(mp, extsize);
662
663 /*
664 * This comment describes a historic gap in this verifier function.
665 *
666 * For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
667 * function has never checked that the extent size hint is an integer
668 * multiple of the realtime extent size. Since we allow users to set
669 * this combination on non-rt filesystems /and/ to change the rt
670 * extent size when adding a rt device to a filesystem, the net effect
671 * is that users can configure a filesystem anticipating one rt
672 * geometry and change their minds later. Directories do not use the
673 * extent size hint, so this is harmless for them.
674 *
675 * If a directory with a misaligned extent size hint is allowed to
676 * propagate that hint into a new regular realtime file, the result
677 * is that the inode cluster buffer verifier will trigger a corruption
678 * shutdown the next time it is run, because the verifier has always
679 * enforced the alignment rule for regular files.
680 *
681 * Because we allow administrators to set a new rt extent size when
682 * adding a rt section, we cannot add a check to this verifier because
683 * that will result a new source of directory corruption errors when
684 * reading an existing filesystem. Instead, we rely on callers to
685 * decide when alignment checks are appropriate, and fix things up as
686 * needed.
687 */
688
689 if (rt_flag)
690 blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
691 else
692 blocksize_bytes = mp->m_sb.sb_blocksize;
693
694 if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
695 return __this_address;
696
697 if (hint_flag && !S_ISREG(mode))
698 return __this_address;
699
700 if (inherit_flag && !S_ISDIR(mode))
701 return __this_address;
702
703 if ((hint_flag || inherit_flag) && extsize == 0)
704 return __this_address;
705
706 /* free inodes get flags set to zero but extsize remains */
707 if (mode && !(hint_flag || inherit_flag) && extsize != 0)
708 return __this_address;
709
710 if (extsize_bytes % blocksize_bytes)
711 return __this_address;
712
713 if (extsize > XFS_MAX_BMBT_EXTLEN)
714 return __this_address;
715
716 if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
717 return __this_address;
718
719 return NULL;
720 }
721
722 /*
723 * Validate di_cowextsize hint.
724 *
725 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
726 * The inode does not have to have any shared blocks, but it must be a v3.
727 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
728 * for a directory, the hint is propagated to new files.
729 * 3. Can be changed on files & directories at any time.
730 * 4. Hint value of 0 turns off hints, clears inode flags.
731 * 5. Extent size must be a multiple of the appropriate block size.
732 * 6. The extent size hint must be limited to half the AG size to avoid
733 * alignment extending the extent beyond the limits of the AG.
734 */
735 xfs_failaddr_t
xfs_inode_validate_cowextsize(struct xfs_mount * mp,uint32_t cowextsize,uint16_t mode,uint16_t flags,uint64_t flags2)736 xfs_inode_validate_cowextsize(
737 struct xfs_mount *mp,
738 uint32_t cowextsize,
739 uint16_t mode,
740 uint16_t flags,
741 uint64_t flags2)
742 {
743 bool rt_flag;
744 bool hint_flag;
745 uint32_t cowextsize_bytes;
746
747 rt_flag = (flags & XFS_DIFLAG_REALTIME);
748 hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
749 cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
750
751 if (hint_flag && !xfs_has_reflink(mp))
752 return __this_address;
753
754 if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
755 return __this_address;
756
757 if (hint_flag && cowextsize == 0)
758 return __this_address;
759
760 /* free inodes get flags set to zero but cowextsize remains */
761 if (mode && !hint_flag && cowextsize != 0)
762 return __this_address;
763
764 if (hint_flag && rt_flag)
765 return __this_address;
766
767 if (cowextsize_bytes % mp->m_sb.sb_blocksize)
768 return __this_address;
769
770 if (cowextsize > XFS_MAX_BMBT_EXTLEN)
771 return __this_address;
772
773 if (cowextsize > mp->m_sb.sb_agblocks / 2)
774 return __this_address;
775
776 return NULL;
777 }
778