1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
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"
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"
21 #include "xfs_health.h"
23 #include <linux/iversion.h>
26 * If we are doing readahead on an inode buffer, we might be in log recovery
27 * reading an inode allocation buffer that hasn't yet been replayed, and hence
28 * has not had the inode cores stamped into it. Hence for readahead, the buffer
29 * may be potentially invalid.
31 * If the readahead buffer is invalid, we need to mark it with an error and
32 * clear the DONE status of the buffer so that a followup read will re-read it
33 * from disk. We don't report the error otherwise to avoid warnings during log
34 * recovery and we don't get unnecessary panics on debug kernels. We use EIO here
35 * because all we want to do is say readahead failed; there is no-one to report
36 * the error to, so this will distinguish it from a non-ra verifier failure.
37 * Changes to this readahead error behaviour also need to be reflected in
38 * xfs_dquot_buf_readahead_verify().
45 struct xfs_mount *mp = bp->b_mount;
50 * Validate the magic number and version of every inode in the buffer
52 ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
53 for (i = 0; i < ni; i++) {
54 struct xfs_dinode *dip;
55 xfs_agino_t unlinked_ino;
58 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
59 unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
60 di_ok = xfs_verify_magic16(bp, dip->di_magic) &&
61 xfs_dinode_good_version(mp, dip->di_version) &&
62 xfs_verify_agino_or_null(bp->b_pag, unlinked_ino);
63 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
64 XFS_ERRTAG_ITOBP_INOTOBP))) {
66 bp->b_flags &= ~XBF_DONE;
67 xfs_buf_ioerror(bp, -EIO);
73 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
74 (unsigned long long)xfs_buf_daddr(bp), i,
75 be16_to_cpu(dip->di_magic));
77 xfs_buf_verifier_error(bp, -EFSCORRUPTED,
78 __func__, dip, sizeof(*dip),
87 xfs_inode_buf_read_verify(
90 xfs_inode_buf_verify(bp, false);
94 xfs_inode_buf_readahead_verify(
97 xfs_inode_buf_verify(bp, true);
101 xfs_inode_buf_write_verify(
104 xfs_inode_buf_verify(bp, false);
107 const struct xfs_buf_ops xfs_inode_buf_ops = {
109 .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
110 cpu_to_be16(XFS_DINODE_MAGIC) },
111 .verify_read = xfs_inode_buf_read_verify,
112 .verify_write = xfs_inode_buf_write_verify,
115 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
116 .name = "xfs_inode_ra",
117 .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
118 cpu_to_be16(XFS_DINODE_MAGIC) },
119 .verify_read = xfs_inode_buf_readahead_verify,
120 .verify_write = xfs_inode_buf_write_verify,
125 * This routine is called to map an inode to the buffer containing the on-disk
126 * version of the inode. It returns a pointer to the buffer containing the
127 * on-disk inode in the bpp parameter.
131 struct xfs_mount *mp,
132 struct xfs_trans *tp,
133 struct xfs_imap *imap,
134 struct xfs_buf **bpp)
138 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
139 imap->im_len, XBF_UNMAPPED, bpp, &xfs_inode_buf_ops);
140 if (xfs_metadata_is_sick(error))
141 xfs_agno_mark_sick(mp, xfs_daddr_to_agno(mp, imap->im_blkno),
146 static inline struct timespec64 xfs_inode_decode_bigtime(uint64_t ts)
148 struct timespec64 tv;
151 tv.tv_sec = xfs_bigtime_to_unix(div_u64_rem(ts, NSEC_PER_SEC, &n));
157 /* Convert an ondisk timestamp to an incore timestamp. */
159 xfs_inode_from_disk_ts(
160 struct xfs_dinode *dip,
161 const xfs_timestamp_t ts)
163 struct timespec64 tv;
164 struct xfs_legacy_timestamp *lts;
166 if (xfs_dinode_has_bigtime(dip))
167 return xfs_inode_decode_bigtime(be64_to_cpu(ts));
169 lts = (struct xfs_legacy_timestamp *)&ts;
170 tv.tv_sec = (int)be32_to_cpu(lts->t_sec);
171 tv.tv_nsec = (int)be32_to_cpu(lts->t_nsec);
178 struct xfs_inode *ip,
179 struct xfs_dinode *from)
181 struct inode *inode = VFS_I(ip);
185 ASSERT(ip->i_cowfp == NULL);
187 fa = xfs_dinode_verify(ip->i_mount, ip->i_ino, from);
189 xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", from,
191 return -EFSCORRUPTED;
195 * First get the permanent information that is needed to allocate an
196 * inode. If the inode is unused, mode is zero and we shouldn't mess
197 * with the uninitialized part of it.
199 if (!xfs_has_v3inodes(ip->i_mount))
200 ip->i_flushiter = be16_to_cpu(from->di_flushiter);
201 inode->i_generation = be32_to_cpu(from->di_gen);
202 inode->i_mode = be16_to_cpu(from->di_mode);
207 * Convert v1 inodes immediately to v2 inode format as this is the
208 * minimum inode version format we support in the rest of the code.
209 * They will also be unconditionally written back to disk as v2 inodes.
211 if (unlikely(from->di_version == 1)) {
212 set_nlink(inode, be16_to_cpu(from->di_onlink));
215 set_nlink(inode, be32_to_cpu(from->di_nlink));
216 ip->i_projid = (prid_t)be16_to_cpu(from->di_projid_hi) << 16 |
217 be16_to_cpu(from->di_projid_lo);
220 i_uid_write(inode, be32_to_cpu(from->di_uid));
221 i_gid_write(inode, be32_to_cpu(from->di_gid));
224 * Time is signed, so need to convert to signed 32 bit before
225 * storing in inode timestamp which may be 64 bit. Otherwise
226 * a time before epoch is converted to a time long after epoch
229 inode_set_atime_to_ts(inode,
230 xfs_inode_from_disk_ts(from, from->di_atime));
231 inode_set_mtime_to_ts(inode,
232 xfs_inode_from_disk_ts(from, from->di_mtime));
233 inode_set_ctime_to_ts(inode,
234 xfs_inode_from_disk_ts(from, from->di_ctime));
236 ip->i_disk_size = be64_to_cpu(from->di_size);
237 ip->i_nblocks = be64_to_cpu(from->di_nblocks);
238 ip->i_extsize = be32_to_cpu(from->di_extsize);
239 ip->i_forkoff = from->di_forkoff;
240 ip->i_diflags = be16_to_cpu(from->di_flags);
241 ip->i_next_unlinked = be32_to_cpu(from->di_next_unlinked);
243 if (from->di_dmevmask || from->di_dmstate)
244 xfs_iflags_set(ip, XFS_IPRESERVE_DM_FIELDS);
246 if (xfs_has_v3inodes(ip->i_mount)) {
247 inode_set_iversion_queried(inode,
248 be64_to_cpu(from->di_changecount));
249 ip->i_crtime = xfs_inode_from_disk_ts(from, from->di_crtime);
250 ip->i_diflags2 = be64_to_cpu(from->di_flags2);
251 ip->i_cowextsize = be32_to_cpu(from->di_cowextsize);
254 error = xfs_iformat_data_fork(ip, from);
257 if (from->di_forkoff) {
258 error = xfs_iformat_attr_fork(ip, from);
260 goto out_destroy_data_fork;
262 if (xfs_is_reflink_inode(ip))
263 xfs_ifork_init_cow(ip);
266 out_destroy_data_fork:
267 xfs_idestroy_fork(&ip->i_df);
271 /* Convert an incore timestamp to an ondisk timestamp. */
272 static inline xfs_timestamp_t
273 xfs_inode_to_disk_ts(
274 struct xfs_inode *ip,
275 const struct timespec64 tv)
277 struct xfs_legacy_timestamp *lts;
280 if (xfs_inode_has_bigtime(ip))
281 return cpu_to_be64(xfs_inode_encode_bigtime(tv));
283 lts = (struct xfs_legacy_timestamp *)&ts;
284 lts->t_sec = cpu_to_be32(tv.tv_sec);
285 lts->t_nsec = cpu_to_be32(tv.tv_nsec);
291 xfs_inode_to_disk_iext_counters(
292 struct xfs_inode *ip,
293 struct xfs_dinode *to)
295 if (xfs_inode_has_large_extent_counts(ip)) {
296 to->di_big_nextents = cpu_to_be64(xfs_ifork_nextents(&ip->i_df));
297 to->di_big_anextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_af));
299 * We might be upgrading the inode to use larger extent counters
300 * than was previously used. Hence zero the unused field.
302 to->di_nrext64_pad = cpu_to_be16(0);
304 to->di_nextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_df));
305 to->di_anextents = cpu_to_be16(xfs_ifork_nextents(&ip->i_af));
311 struct xfs_inode *ip,
312 struct xfs_dinode *to,
315 struct inode *inode = VFS_I(ip);
317 to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
320 to->di_format = xfs_ifork_format(&ip->i_df);
321 to->di_uid = cpu_to_be32(i_uid_read(inode));
322 to->di_gid = cpu_to_be32(i_gid_read(inode));
323 to->di_projid_lo = cpu_to_be16(ip->i_projid & 0xffff);
324 to->di_projid_hi = cpu_to_be16(ip->i_projid >> 16);
326 to->di_atime = xfs_inode_to_disk_ts(ip, inode_get_atime(inode));
327 to->di_mtime = xfs_inode_to_disk_ts(ip, inode_get_mtime(inode));
328 to->di_ctime = xfs_inode_to_disk_ts(ip, inode_get_ctime(inode));
329 to->di_nlink = cpu_to_be32(inode->i_nlink);
330 to->di_gen = cpu_to_be32(inode->i_generation);
331 to->di_mode = cpu_to_be16(inode->i_mode);
333 to->di_size = cpu_to_be64(ip->i_disk_size);
334 to->di_nblocks = cpu_to_be64(ip->i_nblocks);
335 to->di_extsize = cpu_to_be32(ip->i_extsize);
336 to->di_forkoff = ip->i_forkoff;
337 to->di_aformat = xfs_ifork_format(&ip->i_af);
338 to->di_flags = cpu_to_be16(ip->i_diflags);
340 if (xfs_has_v3inodes(ip->i_mount)) {
342 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
343 to->di_crtime = xfs_inode_to_disk_ts(ip, ip->i_crtime);
344 to->di_flags2 = cpu_to_be64(ip->i_diflags2);
345 to->di_cowextsize = cpu_to_be32(ip->i_cowextsize);
346 to->di_ino = cpu_to_be64(ip->i_ino);
347 to->di_lsn = cpu_to_be64(lsn);
348 memset(to->di_pad2, 0, sizeof(to->di_pad2));
349 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
353 to->di_flushiter = cpu_to_be16(ip->i_flushiter);
354 memset(to->di_v2_pad, 0, sizeof(to->di_v2_pad));
357 xfs_inode_to_disk_iext_counters(ip, to);
360 static xfs_failaddr_t
361 xfs_dinode_verify_fork(
362 struct xfs_dinode *dip,
363 struct xfs_mount *mp,
366 xfs_extnum_t di_nextents;
367 xfs_extnum_t max_extents;
368 mode_t mode = be16_to_cpu(dip->di_mode);
369 uint32_t fork_size = XFS_DFORK_SIZE(dip, mp, whichfork);
370 uint32_t fork_format = XFS_DFORK_FORMAT(dip, whichfork);
372 di_nextents = xfs_dfork_nextents(dip, whichfork);
375 * For fork types that can contain local data, check that the fork
376 * format matches the size of local data contained within the fork.
378 if (whichfork == XFS_DATA_FORK) {
380 * A directory small enough to fit in the inode must be stored
381 * in local format. The directory sf <-> extents conversion
382 * code updates the directory size accordingly. Directories
383 * being truncated have zero size and are not subject to this
388 be64_to_cpu(dip->di_size) <= fork_size &&
389 fork_format != XFS_DINODE_FMT_LOCAL)
390 return __this_address;
394 * A symlink with a target small enough to fit in the inode can
395 * be stored in extents format if xattrs were added (thus
396 * converting the data fork from shortform to remote format)
400 if (be64_to_cpu(dip->di_size) <= fork_size &&
401 fork_format != XFS_DINODE_FMT_EXTENTS &&
402 fork_format != XFS_DINODE_FMT_LOCAL)
403 return __this_address;
407 * For all types, check that when the size says the fork should
408 * be in extent or btree format, the inode isn't claiming to be
411 if (be64_to_cpu(dip->di_size) > fork_size &&
412 fork_format == XFS_DINODE_FMT_LOCAL)
413 return __this_address;
416 switch (fork_format) {
417 case XFS_DINODE_FMT_LOCAL:
419 * No local regular files yet.
421 if (S_ISREG(mode) && whichfork == XFS_DATA_FORK)
422 return __this_address;
424 return __this_address;
426 case XFS_DINODE_FMT_EXTENTS:
427 if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
428 return __this_address;
430 case XFS_DINODE_FMT_BTREE:
431 max_extents = xfs_iext_max_nextents(
432 xfs_dinode_has_large_extent_counts(dip),
434 if (di_nextents > max_extents)
435 return __this_address;
438 return __this_address;
443 static xfs_failaddr_t
444 xfs_dinode_verify_forkoff(
445 struct xfs_dinode *dip,
446 struct xfs_mount *mp)
448 if (!dip->di_forkoff)
451 switch (dip->di_format) {
452 case XFS_DINODE_FMT_DEV:
453 if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
454 return __this_address;
456 case XFS_DINODE_FMT_LOCAL: /* fall through ... */
457 case XFS_DINODE_FMT_EXTENTS: /* fall through ... */
458 case XFS_DINODE_FMT_BTREE:
459 if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3))
460 return __this_address;
463 return __this_address;
468 static xfs_failaddr_t
469 xfs_dinode_verify_nrext64(
470 struct xfs_mount *mp,
471 struct xfs_dinode *dip)
473 if (xfs_dinode_has_large_extent_counts(dip)) {
474 if (!xfs_has_large_extent_counts(mp))
475 return __this_address;
476 if (dip->di_nrext64_pad != 0)
477 return __this_address;
478 } else if (dip->di_version >= 3) {
479 if (dip->di_v3_pad != 0)
480 return __this_address;
488 struct xfs_mount *mp,
490 struct xfs_dinode *dip)
497 xfs_extnum_t nextents;
498 xfs_extnum_t naextents;
499 xfs_filblks_t nblocks;
501 if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
502 return __this_address;
504 /* Verify v3 integrity information first */
505 if (dip->di_version >= 3) {
506 if (!xfs_has_v3inodes(mp))
507 return __this_address;
508 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
510 return __this_address;
511 if (be64_to_cpu(dip->di_ino) != ino)
512 return __this_address;
513 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
514 return __this_address;
518 * Historical note: xfsprogs in the 3.2 era set up its incore inodes to
519 * have di_nlink track the link count, even if the actual filesystem
520 * only supported V1 inodes (i.e. di_onlink). When writing out the
521 * ondisk inode, it would set both the ondisk di_nlink and di_onlink to
522 * the the incore di_nlink value, which is why we cannot check for
523 * di_nlink==0 on a V1 inode. V2/3 inodes would get written out with
524 * di_onlink==0, so we can check that.
526 if (dip->di_version >= 2) {
528 return __this_address;
531 /* don't allow invalid i_size */
532 di_size = be64_to_cpu(dip->di_size);
533 if (di_size & (1ULL << 63))
534 return __this_address;
536 mode = be16_to_cpu(dip->di_mode);
537 if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
538 return __this_address;
541 * No zero-length symlinks/dirs unless they're unlinked and hence being
544 if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0) {
545 if (dip->di_version > 1) {
547 return __this_address;
550 return __this_address;
554 fa = xfs_dinode_verify_nrext64(mp, dip);
558 nextents = xfs_dfork_data_extents(dip);
559 naextents = xfs_dfork_attr_extents(dip);
560 nblocks = be64_to_cpu(dip->di_nblocks);
562 /* Fork checks carried over from xfs_iformat_fork */
563 if (mode && nextents + naextents > nblocks)
564 return __this_address;
566 if (nextents + naextents == 0 && nblocks != 0)
567 return __this_address;
569 if (S_ISDIR(mode) && nextents > mp->m_dir_geo->max_extents)
570 return __this_address;
572 if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
573 return __this_address;
575 flags = be16_to_cpu(dip->di_flags);
577 if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
578 return __this_address;
580 /* check for illegal values of forkoff */
581 fa = xfs_dinode_verify_forkoff(dip, mp);
585 /* Do we have appropriate data fork formats for the mode? */
586 switch (mode & S_IFMT) {
591 if (dip->di_format != XFS_DINODE_FMT_DEV)
592 return __this_address;
597 fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
602 /* Uninitialized inode ok. */
605 return __this_address;
608 if (dip->di_forkoff) {
609 fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
614 * If there is no fork offset, this may be a freshly-made inode
615 * in a new disk cluster, in which case di_aformat is zeroed.
616 * Otherwise, such an inode must be in EXTENTS format; this goes
617 * for freed inodes as well.
619 switch (dip->di_aformat) {
621 case XFS_DINODE_FMT_EXTENTS:
624 return __this_address;
627 return __this_address;
630 /* extent size hint validation */
631 fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
636 /* only version 3 or greater inodes are extensively verified here */
637 if (dip->di_version < 3)
640 flags2 = be64_to_cpu(dip->di_flags2);
642 /* don't allow reflink/cowextsize if we don't have reflink */
643 if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
644 !xfs_has_reflink(mp))
645 return __this_address;
647 /* only regular files get reflink */
648 if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
649 return __this_address;
651 /* don't let reflink and realtime mix */
652 if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
653 return __this_address;
655 /* COW extent size hint validation */
656 fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
657 mode, flags, flags2);
661 /* bigtime iflag can only happen on bigtime filesystems */
662 if (xfs_dinode_has_bigtime(dip) &&
663 !xfs_has_bigtime(mp))
664 return __this_address;
671 struct xfs_mount *mp,
672 struct xfs_dinode *dip)
676 if (dip->di_version < 3)
679 ASSERT(xfs_has_crc(mp));
680 crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
682 dip->di_crc = xfs_end_cksum(crc);
686 * Validate di_extsize hint.
688 * 1. Extent size hint is only valid for directories and regular files.
689 * 2. FS_XFLAG_EXTSIZE is only valid for regular files.
690 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
691 * 4. Hint cannot be larger than MAXTEXTLEN.
692 * 5. Can be changed on directories at any time.
693 * 6. Hint value of 0 turns off hints, clears inode flags.
694 * 7. Extent size must be a multiple of the appropriate block size.
695 * For realtime files, this is the rt extent size.
696 * 8. For non-realtime files, the extent size hint must be limited
697 * to half the AG size to avoid alignment extending the extent beyond the
701 xfs_inode_validate_extsize(
702 struct xfs_mount *mp,
710 uint32_t extsize_bytes;
711 uint32_t blocksize_bytes;
713 rt_flag = (flags & XFS_DIFLAG_REALTIME);
714 hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
715 inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
716 extsize_bytes = XFS_FSB_TO_B(mp, extsize);
719 * This comment describes a historic gap in this verifier function.
721 * For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
722 * function has never checked that the extent size hint is an integer
723 * multiple of the realtime extent size. Since we allow users to set
724 * this combination on non-rt filesystems /and/ to change the rt
725 * extent size when adding a rt device to a filesystem, the net effect
726 * is that users can configure a filesystem anticipating one rt
727 * geometry and change their minds later. Directories do not use the
728 * extent size hint, so this is harmless for them.
730 * If a directory with a misaligned extent size hint is allowed to
731 * propagate that hint into a new regular realtime file, the result
732 * is that the inode cluster buffer verifier will trigger a corruption
733 * shutdown the next time it is run, because the verifier has always
734 * enforced the alignment rule for regular files.
736 * Because we allow administrators to set a new rt extent size when
737 * adding a rt section, we cannot add a check to this verifier because
738 * that will result a new source of directory corruption errors when
739 * reading an existing filesystem. Instead, we rely on callers to
740 * decide when alignment checks are appropriate, and fix things up as
745 blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
747 blocksize_bytes = mp->m_sb.sb_blocksize;
749 if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
750 return __this_address;
752 if (hint_flag && !S_ISREG(mode))
753 return __this_address;
755 if (inherit_flag && !S_ISDIR(mode))
756 return __this_address;
758 if ((hint_flag || inherit_flag) && extsize == 0)
759 return __this_address;
761 /* free inodes get flags set to zero but extsize remains */
762 if (mode && !(hint_flag || inherit_flag) && extsize != 0)
763 return __this_address;
765 if (extsize_bytes % blocksize_bytes)
766 return __this_address;
768 if (extsize > XFS_MAX_BMBT_EXTLEN)
769 return __this_address;
771 if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
772 return __this_address;
778 * Validate di_cowextsize hint.
780 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
781 * The inode does not have to have any shared blocks, but it must be a v3.
782 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
783 * for a directory, the hint is propagated to new files.
784 * 3. Can be changed on files & directories at any time.
785 * 4. Hint value of 0 turns off hints, clears inode flags.
786 * 5. Extent size must be a multiple of the appropriate block size.
787 * 6. The extent size hint must be limited to half the AG size to avoid
788 * alignment extending the extent beyond the limits of the AG.
791 xfs_inode_validate_cowextsize(
792 struct xfs_mount *mp,
800 uint32_t cowextsize_bytes;
802 rt_flag = (flags & XFS_DIFLAG_REALTIME);
803 hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
804 cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
806 if (hint_flag && !xfs_has_reflink(mp))
807 return __this_address;
809 if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
810 return __this_address;
812 if (hint_flag && cowextsize == 0)
813 return __this_address;
815 /* free inodes get flags set to zero but cowextsize remains */
816 if (mode && !hint_flag && cowextsize != 0)
817 return __this_address;
819 if (hint_flag && rt_flag)
820 return __this_address;
822 if (cowextsize_bytes % mp->m_sb.sb_blocksize)
823 return __this_address;
825 if (cowextsize > XFS_MAX_BMBT_EXTLEN)
826 return __this_address;
828 if (cowextsize > mp->m_sb.sb_agblocks / 2)
829 return __this_address;