Loading...
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 */
39static void
40xfs_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
85static void
86xfs_inode_buf_read_verify(
87 struct xfs_buf *bp)
88{
89 xfs_inode_buf_verify(bp, false);
90}
91
92static void
93xfs_inode_buf_readahead_verify(
94 struct xfs_buf *bp)
95{
96 xfs_inode_buf_verify(bp, true);
97}
98
99static void
100xfs_inode_buf_write_verify(
101 struct xfs_buf *bp)
102{
103 xfs_inode_buf_verify(bp, false);
104}
105
106const 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
114const 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 */
128int
129xfs_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
140static 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. */
152struct timespec64
153xfs_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
170int
171xfs_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_set_atime_to_ts(inode,
224 xfs_inode_from_disk_ts(from, from->di_atime));
225 inode_set_mtime_to_ts(inode,
226 xfs_inode_from_disk_ts(from, from->di_mtime));
227 inode_set_ctime_to_ts(inode,
228 xfs_inode_from_disk_ts(from, from->di_ctime));
229
230 ip->i_disk_size = be64_to_cpu(from->di_size);
231 ip->i_nblocks = be64_to_cpu(from->di_nblocks);
232 ip->i_extsize = be32_to_cpu(from->di_extsize);
233 ip->i_forkoff = from->di_forkoff;
234 ip->i_diflags = be16_to_cpu(from->di_flags);
235 ip->i_next_unlinked = be32_to_cpu(from->di_next_unlinked);
236
237 if (from->di_dmevmask || from->di_dmstate)
238 xfs_iflags_set(ip, XFS_IPRESERVE_DM_FIELDS);
239
240 if (xfs_has_v3inodes(ip->i_mount)) {
241 inode_set_iversion_queried(inode,
242 be64_to_cpu(from->di_changecount));
243 ip->i_crtime = xfs_inode_from_disk_ts(from, from->di_crtime);
244 ip->i_diflags2 = be64_to_cpu(from->di_flags2);
245 ip->i_cowextsize = be32_to_cpu(from->di_cowextsize);
246 }
247
248 error = xfs_iformat_data_fork(ip, from);
249 if (error)
250 return error;
251 if (from->di_forkoff) {
252 error = xfs_iformat_attr_fork(ip, from);
253 if (error)
254 goto out_destroy_data_fork;
255 }
256 if (xfs_is_reflink_inode(ip))
257 xfs_ifork_init_cow(ip);
258 return 0;
259
260out_destroy_data_fork:
261 xfs_idestroy_fork(&ip->i_df);
262 return error;
263}
264
265/* Convert an incore timestamp to an ondisk timestamp. */
266static inline xfs_timestamp_t
267xfs_inode_to_disk_ts(
268 struct xfs_inode *ip,
269 const struct timespec64 tv)
270{
271 struct xfs_legacy_timestamp *lts;
272 xfs_timestamp_t ts;
273
274 if (xfs_inode_has_bigtime(ip))
275 return cpu_to_be64(xfs_inode_encode_bigtime(tv));
276
277 lts = (struct xfs_legacy_timestamp *)&ts;
278 lts->t_sec = cpu_to_be32(tv.tv_sec);
279 lts->t_nsec = cpu_to_be32(tv.tv_nsec);
280
281 return ts;
282}
283
284static inline void
285xfs_inode_to_disk_iext_counters(
286 struct xfs_inode *ip,
287 struct xfs_dinode *to)
288{
289 if (xfs_inode_has_large_extent_counts(ip)) {
290 to->di_big_nextents = cpu_to_be64(xfs_ifork_nextents(&ip->i_df));
291 to->di_big_anextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_af));
292 /*
293 * We might be upgrading the inode to use larger extent counters
294 * than was previously used. Hence zero the unused field.
295 */
296 to->di_nrext64_pad = cpu_to_be16(0);
297 } else {
298 to->di_nextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_df));
299 to->di_anextents = cpu_to_be16(xfs_ifork_nextents(&ip->i_af));
300 }
301}
302
303void
304xfs_inode_to_disk(
305 struct xfs_inode *ip,
306 struct xfs_dinode *to,
307 xfs_lsn_t lsn)
308{
309 struct inode *inode = VFS_I(ip);
310
311 to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
312 to->di_onlink = 0;
313
314 to->di_format = xfs_ifork_format(&ip->i_df);
315 to->di_uid = cpu_to_be32(i_uid_read(inode));
316 to->di_gid = cpu_to_be32(i_gid_read(inode));
317 to->di_projid_lo = cpu_to_be16(ip->i_projid & 0xffff);
318 to->di_projid_hi = cpu_to_be16(ip->i_projid >> 16);
319
320 to->di_atime = xfs_inode_to_disk_ts(ip, inode_get_atime(inode));
321 to->di_mtime = xfs_inode_to_disk_ts(ip, inode_get_mtime(inode));
322 to->di_ctime = xfs_inode_to_disk_ts(ip, inode_get_ctime(inode));
323 to->di_nlink = cpu_to_be32(inode->i_nlink);
324 to->di_gen = cpu_to_be32(inode->i_generation);
325 to->di_mode = cpu_to_be16(inode->i_mode);
326
327 to->di_size = cpu_to_be64(ip->i_disk_size);
328 to->di_nblocks = cpu_to_be64(ip->i_nblocks);
329 to->di_extsize = cpu_to_be32(ip->i_extsize);
330 to->di_forkoff = ip->i_forkoff;
331 to->di_aformat = xfs_ifork_format(&ip->i_af);
332 to->di_flags = cpu_to_be16(ip->i_diflags);
333
334 if (xfs_has_v3inodes(ip->i_mount)) {
335 to->di_version = 3;
336 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
337 to->di_crtime = xfs_inode_to_disk_ts(ip, ip->i_crtime);
338 to->di_flags2 = cpu_to_be64(ip->i_diflags2);
339 to->di_cowextsize = cpu_to_be32(ip->i_cowextsize);
340 to->di_ino = cpu_to_be64(ip->i_ino);
341 to->di_lsn = cpu_to_be64(lsn);
342 memset(to->di_pad2, 0, sizeof(to->di_pad2));
343 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
344 to->di_v3_pad = 0;
345 } else {
346 to->di_version = 2;
347 to->di_flushiter = cpu_to_be16(ip->i_flushiter);
348 memset(to->di_v2_pad, 0, sizeof(to->di_v2_pad));
349 }
350
351 xfs_inode_to_disk_iext_counters(ip, to);
352}
353
354static xfs_failaddr_t
355xfs_dinode_verify_fork(
356 struct xfs_dinode *dip,
357 struct xfs_mount *mp,
358 int whichfork)
359{
360 xfs_extnum_t di_nextents;
361 xfs_extnum_t max_extents;
362 mode_t mode = be16_to_cpu(dip->di_mode);
363 uint32_t fork_size = XFS_DFORK_SIZE(dip, mp, whichfork);
364 uint32_t fork_format = XFS_DFORK_FORMAT(dip, whichfork);
365
366 di_nextents = xfs_dfork_nextents(dip, whichfork);
367
368 /*
369 * For fork types that can contain local data, check that the fork
370 * format matches the size of local data contained within the fork.
371 *
372 * For all types, check that when the size says the should be in extent
373 * or btree format, the inode isn't claiming it is in local format.
374 */
375 if (whichfork == XFS_DATA_FORK) {
376 if (S_ISDIR(mode) || S_ISLNK(mode)) {
377 if (be64_to_cpu(dip->di_size) <= fork_size &&
378 fork_format != XFS_DINODE_FMT_LOCAL)
379 return __this_address;
380 }
381
382 if (be64_to_cpu(dip->di_size) > fork_size &&
383 fork_format == XFS_DINODE_FMT_LOCAL)
384 return __this_address;
385 }
386
387 switch (fork_format) {
388 case XFS_DINODE_FMT_LOCAL:
389 /*
390 * No local regular files yet.
391 */
392 if (S_ISREG(mode) && whichfork == XFS_DATA_FORK)
393 return __this_address;
394 if (di_nextents)
395 return __this_address;
396 break;
397 case XFS_DINODE_FMT_EXTENTS:
398 if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
399 return __this_address;
400 break;
401 case XFS_DINODE_FMT_BTREE:
402 max_extents = xfs_iext_max_nextents(
403 xfs_dinode_has_large_extent_counts(dip),
404 whichfork);
405 if (di_nextents > max_extents)
406 return __this_address;
407 break;
408 default:
409 return __this_address;
410 }
411 return NULL;
412}
413
414static xfs_failaddr_t
415xfs_dinode_verify_forkoff(
416 struct xfs_dinode *dip,
417 struct xfs_mount *mp)
418{
419 if (!dip->di_forkoff)
420 return NULL;
421
422 switch (dip->di_format) {
423 case XFS_DINODE_FMT_DEV:
424 if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
425 return __this_address;
426 break;
427 case XFS_DINODE_FMT_LOCAL: /* fall through ... */
428 case XFS_DINODE_FMT_EXTENTS: /* fall through ... */
429 case XFS_DINODE_FMT_BTREE:
430 if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3))
431 return __this_address;
432 break;
433 default:
434 return __this_address;
435 }
436 return NULL;
437}
438
439static xfs_failaddr_t
440xfs_dinode_verify_nrext64(
441 struct xfs_mount *mp,
442 struct xfs_dinode *dip)
443{
444 if (xfs_dinode_has_large_extent_counts(dip)) {
445 if (!xfs_has_large_extent_counts(mp))
446 return __this_address;
447 if (dip->di_nrext64_pad != 0)
448 return __this_address;
449 } else if (dip->di_version >= 3) {
450 if (dip->di_v3_pad != 0)
451 return __this_address;
452 }
453
454 return NULL;
455}
456
457xfs_failaddr_t
458xfs_dinode_verify(
459 struct xfs_mount *mp,
460 xfs_ino_t ino,
461 struct xfs_dinode *dip)
462{
463 xfs_failaddr_t fa;
464 uint16_t mode;
465 uint16_t flags;
466 uint64_t flags2;
467 uint64_t di_size;
468 xfs_extnum_t nextents;
469 xfs_extnum_t naextents;
470 xfs_filblks_t nblocks;
471
472 if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
473 return __this_address;
474
475 /* Verify v3 integrity information first */
476 if (dip->di_version >= 3) {
477 if (!xfs_has_v3inodes(mp))
478 return __this_address;
479 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
480 XFS_DINODE_CRC_OFF))
481 return __this_address;
482 if (be64_to_cpu(dip->di_ino) != ino)
483 return __this_address;
484 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
485 return __this_address;
486 }
487
488 /* don't allow invalid i_size */
489 di_size = be64_to_cpu(dip->di_size);
490 if (di_size & (1ULL << 63))
491 return __this_address;
492
493 mode = be16_to_cpu(dip->di_mode);
494 if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
495 return __this_address;
496
497 /* No zero-length symlinks/dirs. */
498 if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
499 return __this_address;
500
501 fa = xfs_dinode_verify_nrext64(mp, dip);
502 if (fa)
503 return fa;
504
505 nextents = xfs_dfork_data_extents(dip);
506 naextents = xfs_dfork_attr_extents(dip);
507 nblocks = be64_to_cpu(dip->di_nblocks);
508
509 /* Fork checks carried over from xfs_iformat_fork */
510 if (mode && nextents + naextents > nblocks)
511 return __this_address;
512
513 if (nextents + naextents == 0 && nblocks != 0)
514 return __this_address;
515
516 if (S_ISDIR(mode) && nextents > mp->m_dir_geo->max_extents)
517 return __this_address;
518
519 if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
520 return __this_address;
521
522 flags = be16_to_cpu(dip->di_flags);
523
524 if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
525 return __this_address;
526
527 /* check for illegal values of forkoff */
528 fa = xfs_dinode_verify_forkoff(dip, mp);
529 if (fa)
530 return fa;
531
532 /* Do we have appropriate data fork formats for the mode? */
533 switch (mode & S_IFMT) {
534 case S_IFIFO:
535 case S_IFCHR:
536 case S_IFBLK:
537 case S_IFSOCK:
538 if (dip->di_format != XFS_DINODE_FMT_DEV)
539 return __this_address;
540 break;
541 case S_IFREG:
542 case S_IFLNK:
543 case S_IFDIR:
544 fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
545 if (fa)
546 return fa;
547 break;
548 case 0:
549 /* Uninitialized inode ok. */
550 break;
551 default:
552 return __this_address;
553 }
554
555 if (dip->di_forkoff) {
556 fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
557 if (fa)
558 return fa;
559 } else {
560 /*
561 * If there is no fork offset, this may be a freshly-made inode
562 * in a new disk cluster, in which case di_aformat is zeroed.
563 * Otherwise, such an inode must be in EXTENTS format; this goes
564 * for freed inodes as well.
565 */
566 switch (dip->di_aformat) {
567 case 0:
568 case XFS_DINODE_FMT_EXTENTS:
569 break;
570 default:
571 return __this_address;
572 }
573 if (naextents)
574 return __this_address;
575 }
576
577 /* extent size hint validation */
578 fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
579 mode, flags);
580 if (fa)
581 return fa;
582
583 /* only version 3 or greater inodes are extensively verified here */
584 if (dip->di_version < 3)
585 return NULL;
586
587 flags2 = be64_to_cpu(dip->di_flags2);
588
589 /* don't allow reflink/cowextsize if we don't have reflink */
590 if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
591 !xfs_has_reflink(mp))
592 return __this_address;
593
594 /* only regular files get reflink */
595 if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
596 return __this_address;
597
598 /* don't let reflink and realtime mix */
599 if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
600 return __this_address;
601
602 /* COW extent size hint validation */
603 fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
604 mode, flags, flags2);
605 if (fa)
606 return fa;
607
608 /* bigtime iflag can only happen on bigtime filesystems */
609 if (xfs_dinode_has_bigtime(dip) &&
610 !xfs_has_bigtime(mp))
611 return __this_address;
612
613 return NULL;
614}
615
616void
617xfs_dinode_calc_crc(
618 struct xfs_mount *mp,
619 struct xfs_dinode *dip)
620{
621 uint32_t crc;
622
623 if (dip->di_version < 3)
624 return;
625
626 ASSERT(xfs_has_crc(mp));
627 crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
628 XFS_DINODE_CRC_OFF);
629 dip->di_crc = xfs_end_cksum(crc);
630}
631
632/*
633 * Validate di_extsize hint.
634 *
635 * 1. Extent size hint is only valid for directories and regular files.
636 * 2. FS_XFLAG_EXTSIZE is only valid for regular files.
637 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
638 * 4. Hint cannot be larger than MAXTEXTLEN.
639 * 5. Can be changed on directories at any time.
640 * 6. Hint value of 0 turns off hints, clears inode flags.
641 * 7. Extent size must be a multiple of the appropriate block size.
642 * For realtime files, this is the rt extent size.
643 * 8. For non-realtime files, the extent size hint must be limited
644 * to half the AG size to avoid alignment extending the extent beyond the
645 * limits of the AG.
646 */
647xfs_failaddr_t
648xfs_inode_validate_extsize(
649 struct xfs_mount *mp,
650 uint32_t extsize,
651 uint16_t mode,
652 uint16_t flags)
653{
654 bool rt_flag;
655 bool hint_flag;
656 bool inherit_flag;
657 uint32_t extsize_bytes;
658 uint32_t blocksize_bytes;
659
660 rt_flag = (flags & XFS_DIFLAG_REALTIME);
661 hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
662 inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
663 extsize_bytes = XFS_FSB_TO_B(mp, extsize);
664
665 /*
666 * This comment describes a historic gap in this verifier function.
667 *
668 * For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
669 * function has never checked that the extent size hint is an integer
670 * multiple of the realtime extent size. Since we allow users to set
671 * this combination on non-rt filesystems /and/ to change the rt
672 * extent size when adding a rt device to a filesystem, the net effect
673 * is that users can configure a filesystem anticipating one rt
674 * geometry and change their minds later. Directories do not use the
675 * extent size hint, so this is harmless for them.
676 *
677 * If a directory with a misaligned extent size hint is allowed to
678 * propagate that hint into a new regular realtime file, the result
679 * is that the inode cluster buffer verifier will trigger a corruption
680 * shutdown the next time it is run, because the verifier has always
681 * enforced the alignment rule for regular files.
682 *
683 * Because we allow administrators to set a new rt extent size when
684 * adding a rt section, we cannot add a check to this verifier because
685 * that will result a new source of directory corruption errors when
686 * reading an existing filesystem. Instead, we rely on callers to
687 * decide when alignment checks are appropriate, and fix things up as
688 * needed.
689 */
690
691 if (rt_flag)
692 blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
693 else
694 blocksize_bytes = mp->m_sb.sb_blocksize;
695
696 if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
697 return __this_address;
698
699 if (hint_flag && !S_ISREG(mode))
700 return __this_address;
701
702 if (inherit_flag && !S_ISDIR(mode))
703 return __this_address;
704
705 if ((hint_flag || inherit_flag) && extsize == 0)
706 return __this_address;
707
708 /* free inodes get flags set to zero but extsize remains */
709 if (mode && !(hint_flag || inherit_flag) && extsize != 0)
710 return __this_address;
711
712 if (extsize_bytes % blocksize_bytes)
713 return __this_address;
714
715 if (extsize > XFS_MAX_BMBT_EXTLEN)
716 return __this_address;
717
718 if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
719 return __this_address;
720
721 return NULL;
722}
723
724/*
725 * Validate di_cowextsize hint.
726 *
727 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
728 * The inode does not have to have any shared blocks, but it must be a v3.
729 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
730 * for a directory, the hint is propagated to new files.
731 * 3. Can be changed on files & directories at any time.
732 * 4. Hint value of 0 turns off hints, clears inode flags.
733 * 5. Extent size must be a multiple of the appropriate block size.
734 * 6. The extent size hint must be limited to half the AG size to avoid
735 * alignment extending the extent beyond the limits of the AG.
736 */
737xfs_failaddr_t
738xfs_inode_validate_cowextsize(
739 struct xfs_mount *mp,
740 uint32_t cowextsize,
741 uint16_t mode,
742 uint16_t flags,
743 uint64_t flags2)
744{
745 bool rt_flag;
746 bool hint_flag;
747 uint32_t cowextsize_bytes;
748
749 rt_flag = (flags & XFS_DIFLAG_REALTIME);
750 hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
751 cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
752
753 if (hint_flag && !xfs_has_reflink(mp))
754 return __this_address;
755
756 if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
757 return __this_address;
758
759 if (hint_flag && cowextsize == 0)
760 return __this_address;
761
762 /* free inodes get flags set to zero but cowextsize remains */
763 if (mode && !hint_flag && cowextsize != 0)
764 return __this_address;
765
766 if (hint_flag && rt_flag)
767 return __this_address;
768
769 if (cowextsize_bytes % mp->m_sb.sb_blocksize)
770 return __this_address;
771
772 if (cowextsize > XFS_MAX_BMBT_EXTLEN)
773 return __this_address;
774
775 if (cowextsize > mp->m_sb.sb_agblocks / 2)
776 return __this_address;
777
778 return NULL;
779}
1/*
2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
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
7 * published by the Free Software Foundation.
8 *
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.
13 *
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
17 */
18#include "xfs.h"
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"
25#include "xfs_defer.h"
26#include "xfs_inode.h"
27#include "xfs_errortag.h"
28#include "xfs_error.h"
29#include "xfs_cksum.h"
30#include "xfs_icache.h"
31#include "xfs_trans.h"
32#include "xfs_ialloc.h"
33#include "xfs_dir2.h"
34
35#include <linux/iversion.h>
36
37/*
38 * Check that none of the inode's in the buffer have a next
39 * unlinked field of 0.
40 */
41#if defined(DEBUG)
42void
43xfs_inobp_check(
44 xfs_mount_t *mp,
45 xfs_buf_t *bp)
46{
47 int i;
48 int j;
49 xfs_dinode_t *dip;
50
51 j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;
52
53 for (i = 0; i < j; i++) {
54 dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize);
55 if (!dip->di_next_unlinked) {
56 xfs_alert(mp,
57 "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
58 i, (long long)bp->b_bn);
59 }
60 }
61}
62#endif
63
64bool
65xfs_dinode_good_version(
66 struct xfs_mount *mp,
67 __u8 version)
68{
69 if (xfs_sb_version_hascrc(&mp->m_sb))
70 return version == 3;
71
72 return version == 1 || version == 2;
73}
74
75/*
76 * If we are doing readahead on an inode buffer, we might be in log recovery
77 * reading an inode allocation buffer that hasn't yet been replayed, and hence
78 * has not had the inode cores stamped into it. Hence for readahead, the buffer
79 * may be potentially invalid.
80 *
81 * If the readahead buffer is invalid, we need to mark it with an error and
82 * clear the DONE status of the buffer so that a followup read will re-read it
83 * from disk. We don't report the error otherwise to avoid warnings during log
84 * recovery and we don't get unnecssary panics on debug kernels. We use EIO here
85 * because all we want to do is say readahead failed; there is no-one to report
86 * the error to, so this will distinguish it from a non-ra verifier failure.
87 * Changes to this readahead error behavour also need to be reflected in
88 * xfs_dquot_buf_readahead_verify().
89 */
90static void
91xfs_inode_buf_verify(
92 struct xfs_buf *bp,
93 bool readahead)
94{
95 struct xfs_mount *mp = bp->b_target->bt_mount;
96 xfs_agnumber_t agno;
97 int i;
98 int ni;
99
100 /*
101 * Validate the magic number and version of every inode in the buffer
102 */
103 agno = xfs_daddr_to_agno(mp, XFS_BUF_ADDR(bp));
104 ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
105 for (i = 0; i < ni; i++) {
106 int di_ok;
107 xfs_dinode_t *dip;
108 xfs_agino_t unlinked_ino;
109
110 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
111 unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
112 di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
113 xfs_dinode_good_version(mp, dip->di_version) &&
114 (unlinked_ino == NULLAGINO ||
115 xfs_verify_agino(mp, agno, unlinked_ino));
116 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
117 XFS_ERRTAG_ITOBP_INOTOBP))) {
118 if (readahead) {
119 bp->b_flags &= ~XBF_DONE;
120 xfs_buf_ioerror(bp, -EIO);
121 return;
122 }
123
124#ifdef DEBUG
125 xfs_alert(mp,
126 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
127 (unsigned long long)bp->b_bn, i,
128 be16_to_cpu(dip->di_magic));
129#endif
130 xfs_buf_verifier_error(bp, -EFSCORRUPTED,
131 __func__, dip, sizeof(*dip),
132 NULL);
133 return;
134 }
135 }
136}
137
138
139static void
140xfs_inode_buf_read_verify(
141 struct xfs_buf *bp)
142{
143 xfs_inode_buf_verify(bp, false);
144}
145
146static void
147xfs_inode_buf_readahead_verify(
148 struct xfs_buf *bp)
149{
150 xfs_inode_buf_verify(bp, true);
151}
152
153static void
154xfs_inode_buf_write_verify(
155 struct xfs_buf *bp)
156{
157 xfs_inode_buf_verify(bp, false);
158}
159
160const struct xfs_buf_ops xfs_inode_buf_ops = {
161 .name = "xfs_inode",
162 .verify_read = xfs_inode_buf_read_verify,
163 .verify_write = xfs_inode_buf_write_verify,
164};
165
166const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
167 .name = "xxfs_inode_ra",
168 .verify_read = xfs_inode_buf_readahead_verify,
169 .verify_write = xfs_inode_buf_write_verify,
170};
171
172
173/*
174 * This routine is called to map an inode to the buffer containing the on-disk
175 * version of the inode. It returns a pointer to the buffer containing the
176 * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
177 * pointer to the on-disk inode within that buffer.
178 *
179 * If a non-zero error is returned, then the contents of bpp and dipp are
180 * undefined.
181 */
182int
183xfs_imap_to_bp(
184 struct xfs_mount *mp,
185 struct xfs_trans *tp,
186 struct xfs_imap *imap,
187 struct xfs_dinode **dipp,
188 struct xfs_buf **bpp,
189 uint buf_flags,
190 uint iget_flags)
191{
192 struct xfs_buf *bp;
193 int error;
194
195 buf_flags |= XBF_UNMAPPED;
196 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
197 (int)imap->im_len, buf_flags, &bp,
198 &xfs_inode_buf_ops);
199 if (error) {
200 if (error == -EAGAIN) {
201 ASSERT(buf_flags & XBF_TRYLOCK);
202 return error;
203 }
204
205 if (error == -EFSCORRUPTED &&
206 (iget_flags & XFS_IGET_UNTRUSTED))
207 return -EINVAL;
208
209 xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
210 __func__, error);
211 return error;
212 }
213
214 *bpp = bp;
215 *dipp = xfs_buf_offset(bp, imap->im_boffset);
216 return 0;
217}
218
219void
220xfs_inode_from_disk(
221 struct xfs_inode *ip,
222 struct xfs_dinode *from)
223{
224 struct xfs_icdinode *to = &ip->i_d;
225 struct inode *inode = VFS_I(ip);
226
227
228 /*
229 * Convert v1 inodes immediately to v2 inode format as this is the
230 * minimum inode version format we support in the rest of the code.
231 */
232 to->di_version = from->di_version;
233 if (to->di_version == 1) {
234 set_nlink(inode, be16_to_cpu(from->di_onlink));
235 to->di_projid_lo = 0;
236 to->di_projid_hi = 0;
237 to->di_version = 2;
238 } else {
239 set_nlink(inode, be32_to_cpu(from->di_nlink));
240 to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
241 to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
242 }
243
244 to->di_format = from->di_format;
245 to->di_uid = be32_to_cpu(from->di_uid);
246 to->di_gid = be32_to_cpu(from->di_gid);
247 to->di_flushiter = be16_to_cpu(from->di_flushiter);
248
249 /*
250 * Time is signed, so need to convert to signed 32 bit before
251 * storing in inode timestamp which may be 64 bit. Otherwise
252 * a time before epoch is converted to a time long after epoch
253 * on 64 bit systems.
254 */
255 inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec);
256 inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec);
257 inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec);
258 inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec);
259 inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec);
260 inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec);
261 inode->i_generation = be32_to_cpu(from->di_gen);
262 inode->i_mode = be16_to_cpu(from->di_mode);
263
264 to->di_size = be64_to_cpu(from->di_size);
265 to->di_nblocks = be64_to_cpu(from->di_nblocks);
266 to->di_extsize = be32_to_cpu(from->di_extsize);
267 to->di_nextents = be32_to_cpu(from->di_nextents);
268 to->di_anextents = be16_to_cpu(from->di_anextents);
269 to->di_forkoff = from->di_forkoff;
270 to->di_aformat = from->di_aformat;
271 to->di_dmevmask = be32_to_cpu(from->di_dmevmask);
272 to->di_dmstate = be16_to_cpu(from->di_dmstate);
273 to->di_flags = be16_to_cpu(from->di_flags);
274
275 if (to->di_version == 3) {
276 inode_set_iversion_queried(inode,
277 be64_to_cpu(from->di_changecount));
278 to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
279 to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
280 to->di_flags2 = be64_to_cpu(from->di_flags2);
281 to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
282 }
283}
284
285void
286xfs_inode_to_disk(
287 struct xfs_inode *ip,
288 struct xfs_dinode *to,
289 xfs_lsn_t lsn)
290{
291 struct xfs_icdinode *from = &ip->i_d;
292 struct inode *inode = VFS_I(ip);
293
294 to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
295 to->di_onlink = 0;
296
297 to->di_version = from->di_version;
298 to->di_format = from->di_format;
299 to->di_uid = cpu_to_be32(from->di_uid);
300 to->di_gid = cpu_to_be32(from->di_gid);
301 to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
302 to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
303
304 memset(to->di_pad, 0, sizeof(to->di_pad));
305 to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec);
306 to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec);
307 to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec);
308 to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec);
309 to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec);
310 to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec);
311 to->di_nlink = cpu_to_be32(inode->i_nlink);
312 to->di_gen = cpu_to_be32(inode->i_generation);
313 to->di_mode = cpu_to_be16(inode->i_mode);
314
315 to->di_size = cpu_to_be64(from->di_size);
316 to->di_nblocks = cpu_to_be64(from->di_nblocks);
317 to->di_extsize = cpu_to_be32(from->di_extsize);
318 to->di_nextents = cpu_to_be32(from->di_nextents);
319 to->di_anextents = cpu_to_be16(from->di_anextents);
320 to->di_forkoff = from->di_forkoff;
321 to->di_aformat = from->di_aformat;
322 to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
323 to->di_dmstate = cpu_to_be16(from->di_dmstate);
324 to->di_flags = cpu_to_be16(from->di_flags);
325
326 if (from->di_version == 3) {
327 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
328 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
329 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
330 to->di_flags2 = cpu_to_be64(from->di_flags2);
331 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
332 to->di_ino = cpu_to_be64(ip->i_ino);
333 to->di_lsn = cpu_to_be64(lsn);
334 memset(to->di_pad2, 0, sizeof(to->di_pad2));
335 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
336 to->di_flushiter = 0;
337 } else {
338 to->di_flushiter = cpu_to_be16(from->di_flushiter);
339 }
340}
341
342void
343xfs_log_dinode_to_disk(
344 struct xfs_log_dinode *from,
345 struct xfs_dinode *to)
346{
347 to->di_magic = cpu_to_be16(from->di_magic);
348 to->di_mode = cpu_to_be16(from->di_mode);
349 to->di_version = from->di_version;
350 to->di_format = from->di_format;
351 to->di_onlink = 0;
352 to->di_uid = cpu_to_be32(from->di_uid);
353 to->di_gid = cpu_to_be32(from->di_gid);
354 to->di_nlink = cpu_to_be32(from->di_nlink);
355 to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
356 to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
357 memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
358
359 to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
360 to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
361 to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
362 to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
363 to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
364 to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
365
366 to->di_size = cpu_to_be64(from->di_size);
367 to->di_nblocks = cpu_to_be64(from->di_nblocks);
368 to->di_extsize = cpu_to_be32(from->di_extsize);
369 to->di_nextents = cpu_to_be32(from->di_nextents);
370 to->di_anextents = cpu_to_be16(from->di_anextents);
371 to->di_forkoff = from->di_forkoff;
372 to->di_aformat = from->di_aformat;
373 to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
374 to->di_dmstate = cpu_to_be16(from->di_dmstate);
375 to->di_flags = cpu_to_be16(from->di_flags);
376 to->di_gen = cpu_to_be32(from->di_gen);
377
378 if (from->di_version == 3) {
379 to->di_changecount = cpu_to_be64(from->di_changecount);
380 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
381 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
382 to->di_flags2 = cpu_to_be64(from->di_flags2);
383 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
384 to->di_ino = cpu_to_be64(from->di_ino);
385 to->di_lsn = cpu_to_be64(from->di_lsn);
386 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
387 uuid_copy(&to->di_uuid, &from->di_uuid);
388 to->di_flushiter = 0;
389 } else {
390 to->di_flushiter = cpu_to_be16(from->di_flushiter);
391 }
392}
393
394xfs_failaddr_t
395xfs_dinode_verify(
396 struct xfs_mount *mp,
397 xfs_ino_t ino,
398 struct xfs_dinode *dip)
399{
400 uint16_t mode;
401 uint16_t flags;
402 uint64_t flags2;
403 uint64_t di_size;
404
405 if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
406 return __this_address;
407
408 /* Verify v3 integrity information first */
409 if (dip->di_version >= 3) {
410 if (!xfs_sb_version_hascrc(&mp->m_sb))
411 return __this_address;
412 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
413 XFS_DINODE_CRC_OFF))
414 return __this_address;
415 if (be64_to_cpu(dip->di_ino) != ino)
416 return __this_address;
417 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
418 return __this_address;
419 }
420
421 /* don't allow invalid i_size */
422 di_size = be64_to_cpu(dip->di_size);
423 if (di_size & (1ULL << 63))
424 return __this_address;
425
426 mode = be16_to_cpu(dip->di_mode);
427 if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
428 return __this_address;
429
430 /* No zero-length symlinks/dirs. */
431 if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
432 return __this_address;
433
434 /* Fork checks carried over from xfs_iformat_fork */
435 if (mode &&
436 be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
437 be64_to_cpu(dip->di_nblocks))
438 return __this_address;
439
440 if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
441 return __this_address;
442
443 flags = be16_to_cpu(dip->di_flags);
444
445 if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
446 return __this_address;
447
448 /* Do we have appropriate data fork formats for the mode? */
449 switch (mode & S_IFMT) {
450 case S_IFIFO:
451 case S_IFCHR:
452 case S_IFBLK:
453 case S_IFSOCK:
454 if (dip->di_format != XFS_DINODE_FMT_DEV)
455 return __this_address;
456 break;
457 case S_IFREG:
458 case S_IFLNK:
459 case S_IFDIR:
460 switch (dip->di_format) {
461 case XFS_DINODE_FMT_LOCAL:
462 /*
463 * no local regular files yet
464 */
465 if (S_ISREG(mode))
466 return __this_address;
467 if (di_size > XFS_DFORK_DSIZE(dip, mp))
468 return __this_address;
469 if (dip->di_nextents)
470 return __this_address;
471 /* fall through */
472 case XFS_DINODE_FMT_EXTENTS:
473 case XFS_DINODE_FMT_BTREE:
474 break;
475 default:
476 return __this_address;
477 }
478 break;
479 case 0:
480 /* Uninitialized inode ok. */
481 break;
482 default:
483 return __this_address;
484 }
485
486 if (XFS_DFORK_Q(dip)) {
487 switch (dip->di_aformat) {
488 case XFS_DINODE_FMT_LOCAL:
489 if (dip->di_anextents)
490 return __this_address;
491 /* fall through */
492 case XFS_DINODE_FMT_EXTENTS:
493 case XFS_DINODE_FMT_BTREE:
494 break;
495 default:
496 return __this_address;
497 }
498 } else {
499 /*
500 * If there is no fork offset, this may be a freshly-made inode
501 * in a new disk cluster, in which case di_aformat is zeroed.
502 * Otherwise, such an inode must be in EXTENTS format; this goes
503 * for freed inodes as well.
504 */
505 switch (dip->di_aformat) {
506 case 0:
507 case XFS_DINODE_FMT_EXTENTS:
508 break;
509 default:
510 return __this_address;
511 }
512 if (dip->di_anextents)
513 return __this_address;
514 }
515
516 /* only version 3 or greater inodes are extensively verified here */
517 if (dip->di_version < 3)
518 return NULL;
519
520 flags2 = be64_to_cpu(dip->di_flags2);
521
522 /* don't allow reflink/cowextsize if we don't have reflink */
523 if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
524 !xfs_sb_version_hasreflink(&mp->m_sb))
525 return __this_address;
526
527 /* only regular files get reflink */
528 if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
529 return __this_address;
530
531 /* don't let reflink and realtime mix */
532 if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
533 return __this_address;
534
535 /* don't let reflink and dax mix */
536 if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
537 return __this_address;
538
539 return NULL;
540}
541
542void
543xfs_dinode_calc_crc(
544 struct xfs_mount *mp,
545 struct xfs_dinode *dip)
546{
547 uint32_t crc;
548
549 if (dip->di_version < 3)
550 return;
551
552 ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
553 crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
554 XFS_DINODE_CRC_OFF);
555 dip->di_crc = xfs_end_cksum(crc);
556}
557
558/*
559 * Read the disk inode attributes into the in-core inode structure.
560 *
561 * For version 5 superblocks, if we are initialising a new inode and we are not
562 * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
563 * inode core with a random generation number. If we are keeping inodes around,
564 * we need to read the inode cluster to get the existing generation number off
565 * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
566 * format) then log recovery is dependent on the di_flushiter field being
567 * initialised from the current on-disk value and hence we must also read the
568 * inode off disk.
569 */
570int
571xfs_iread(
572 xfs_mount_t *mp,
573 xfs_trans_t *tp,
574 xfs_inode_t *ip,
575 uint iget_flags)
576{
577 xfs_buf_t *bp;
578 xfs_dinode_t *dip;
579 xfs_failaddr_t fa;
580 int error;
581
582 /*
583 * Fill in the location information in the in-core inode.
584 */
585 error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
586 if (error)
587 return error;
588
589 /* shortcut IO on inode allocation if possible */
590 if ((iget_flags & XFS_IGET_CREATE) &&
591 xfs_sb_version_hascrc(&mp->m_sb) &&
592 !(mp->m_flags & XFS_MOUNT_IKEEP)) {
593 /* initialise the on-disk inode core */
594 memset(&ip->i_d, 0, sizeof(ip->i_d));
595 VFS_I(ip)->i_generation = prandom_u32();
596 ip->i_d.di_version = 3;
597 return 0;
598 }
599
600 /*
601 * Get pointers to the on-disk inode and the buffer containing it.
602 */
603 error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
604 if (error)
605 return error;
606
607 /* even unallocated inodes are verified */
608 fa = xfs_dinode_verify(mp, ip->i_ino, dip);
609 if (fa) {
610 xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
611 sizeof(*dip), fa);
612 error = -EFSCORRUPTED;
613 goto out_brelse;
614 }
615
616 /*
617 * If the on-disk inode is already linked to a directory
618 * entry, copy all of the inode into the in-core inode.
619 * xfs_iformat_fork() handles copying in the inode format
620 * specific information.
621 * Otherwise, just get the truly permanent information.
622 */
623 if (dip->di_mode) {
624 xfs_inode_from_disk(ip, dip);
625 error = xfs_iformat_fork(ip, dip);
626 if (error) {
627#ifdef DEBUG
628 xfs_alert(mp, "%s: xfs_iformat() returned error %d",
629 __func__, error);
630#endif /* DEBUG */
631 goto out_brelse;
632 }
633 } else {
634 /*
635 * Partial initialisation of the in-core inode. Just the bits
636 * that xfs_ialloc won't overwrite or relies on being correct.
637 */
638 ip->i_d.di_version = dip->di_version;
639 VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
640 ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
641
642 /*
643 * Make sure to pull in the mode here as well in
644 * case the inode is released without being used.
645 * This ensures that xfs_inactive() will see that
646 * the inode is already free and not try to mess
647 * with the uninitialized part of it.
648 */
649 VFS_I(ip)->i_mode = 0;
650 }
651
652 ASSERT(ip->i_d.di_version >= 2);
653 ip->i_delayed_blks = 0;
654
655 /*
656 * Mark the buffer containing the inode as something to keep
657 * around for a while. This helps to keep recently accessed
658 * meta-data in-core longer.
659 */
660 xfs_buf_set_ref(bp, XFS_INO_REF);
661
662 /*
663 * Use xfs_trans_brelse() to release the buffer containing the on-disk
664 * inode, because it was acquired with xfs_trans_read_buf() in
665 * xfs_imap_to_bp() above. If tp is NULL, this is just a normal
666 * brelse(). If we're within a transaction, then xfs_trans_brelse()
667 * will only release the buffer if it is not dirty within the
668 * transaction. It will be OK to release the buffer in this case,
669 * because inodes on disk are never destroyed and we will be locking the
670 * new in-core inode before putting it in the cache where other
671 * processes can find it. Thus we don't have to worry about the inode
672 * being changed just because we released the buffer.
673 */
674 out_brelse:
675 xfs_trans_brelse(tp, bp);
676 return error;
677}
678
679/*
680 * Validate di_extsize hint.
681 *
682 * The rules are documented at xfs_ioctl_setattr_check_extsize().
683 * These functions must be kept in sync with each other.
684 */
685xfs_failaddr_t
686xfs_inode_validate_extsize(
687 struct xfs_mount *mp,
688 uint32_t extsize,
689 uint16_t mode,
690 uint16_t flags)
691{
692 bool rt_flag;
693 bool hint_flag;
694 bool inherit_flag;
695 uint32_t extsize_bytes;
696 uint32_t blocksize_bytes;
697
698 rt_flag = (flags & XFS_DIFLAG_REALTIME);
699 hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
700 inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
701 extsize_bytes = XFS_FSB_TO_B(mp, extsize);
702
703 if (rt_flag)
704 blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
705 else
706 blocksize_bytes = mp->m_sb.sb_blocksize;
707
708 if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
709 return __this_address;
710
711 if (hint_flag && !S_ISREG(mode))
712 return __this_address;
713
714 if (inherit_flag && !S_ISDIR(mode))
715 return __this_address;
716
717 if ((hint_flag || inherit_flag) && extsize == 0)
718 return __this_address;
719
720 if (!(hint_flag || inherit_flag) && extsize != 0)
721 return __this_address;
722
723 if (extsize_bytes % blocksize_bytes)
724 return __this_address;
725
726 if (extsize > MAXEXTLEN)
727 return __this_address;
728
729 if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
730 return __this_address;
731
732 return NULL;
733}
734
735/*
736 * Validate di_cowextsize hint.
737 *
738 * The rules are documented at xfs_ioctl_setattr_check_cowextsize().
739 * These functions must be kept in sync with each other.
740 */
741xfs_failaddr_t
742xfs_inode_validate_cowextsize(
743 struct xfs_mount *mp,
744 uint32_t cowextsize,
745 uint16_t mode,
746 uint16_t flags,
747 uint64_t flags2)
748{
749 bool rt_flag;
750 bool hint_flag;
751 uint32_t cowextsize_bytes;
752
753 rt_flag = (flags & XFS_DIFLAG_REALTIME);
754 hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
755 cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
756
757 if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb))
758 return __this_address;
759
760 if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
761 return __this_address;
762
763 if (hint_flag && cowextsize == 0)
764 return __this_address;
765
766 if (!hint_flag && cowextsize != 0)
767 return __this_address;
768
769 if (hint_flag && rt_flag)
770 return __this_address;
771
772 if (cowextsize_bytes % mp->m_sb.sb_blocksize)
773 return __this_address;
774
775 if (cowextsize > MAXEXTLEN)
776 return __this_address;
777
778 if (cowextsize > mp->m_sb.sb_agblocks / 2)
779 return __this_address;
780
781 return NULL;
782}