• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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_inode.h"
14 #include "xfs_errortag.h"
15 #include "xfs_error.h"
16 #include "xfs_icache.h"
17 #include "xfs_trans.h"
18 #include "xfs_ialloc.h"
19 #include "xfs_dir2.h"
20 
21 #include <linux/iversion.h>
22 
23 /*
24  * If we are doing readahead on an inode buffer, we might be in log recovery
25  * reading an inode allocation buffer that hasn't yet been replayed, and hence
26  * has not had the inode cores stamped into it. Hence for readahead, the buffer
27  * may be potentially invalid.
28  *
29  * If the readahead buffer is invalid, we need to mark it with an error and
30  * clear the DONE status of the buffer so that a followup read will re-read it
31  * from disk. We don't report the error otherwise to avoid warnings during log
32  * recovery and we don't get unnecessary panics on debug kernels. We use EIO here
33  * because all we want to do is say readahead failed; there is no-one to report
34  * the error to, so this will distinguish it from a non-ra verifier failure.
35  * Changes to this readahead error behaviour also need to be reflected in
36  * xfs_dquot_buf_readahead_verify().
37  */
38 static void
xfs_inode_buf_verify(struct xfs_buf * bp,bool readahead)39 xfs_inode_buf_verify(
40 	struct xfs_buf	*bp,
41 	bool		readahead)
42 {
43 	struct xfs_mount *mp = bp->b_mount;
44 	xfs_agnumber_t	agno;
45 	int		i;
46 	int		ni;
47 
48 	/*
49 	 * Validate the magic number and version of every inode in the buffer
50 	 */
51 	agno = xfs_daddr_to_agno(mp, xfs_buf_daddr(bp));
52 	ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
53 	for (i = 0; i < ni; i++) {
54 		int		di_ok;
55 		xfs_dinode_t	*dip;
56 		xfs_agino_t	unlinked_ino;
57 
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(mp, agno, unlinked_ino);
63 		if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
64 						XFS_ERRTAG_ITOBP_INOTOBP))) {
65 			if (readahead) {
66 				bp->b_flags &= ~XBF_DONE;
67 				xfs_buf_ioerror(bp, -EIO);
68 				return;
69 			}
70 
71 #ifdef DEBUG
72 			xfs_alert(mp,
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));
76 #endif
77 			xfs_buf_verifier_error(bp, -EFSCORRUPTED,
78 					__func__, dip, sizeof(*dip),
79 					NULL);
80 			return;
81 		}
82 	}
83 }
84 
85 
86 static void
xfs_inode_buf_read_verify(struct xfs_buf * bp)87 xfs_inode_buf_read_verify(
88 	struct xfs_buf	*bp)
89 {
90 	xfs_inode_buf_verify(bp, false);
91 }
92 
93 static void
xfs_inode_buf_readahead_verify(struct xfs_buf * bp)94 xfs_inode_buf_readahead_verify(
95 	struct xfs_buf	*bp)
96 {
97 	xfs_inode_buf_verify(bp, true);
98 }
99 
100 static void
xfs_inode_buf_write_verify(struct xfs_buf * bp)101 xfs_inode_buf_write_verify(
102 	struct xfs_buf	*bp)
103 {
104 	xfs_inode_buf_verify(bp, false);
105 }
106 
107 const struct xfs_buf_ops xfs_inode_buf_ops = {
108 	.name = "xfs_inode",
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,
113 };
114 
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,
121 };
122 
123 
124 /*
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.
128  */
129 int
xfs_imap_to_bp(struct xfs_mount * mp,struct xfs_trans * tp,struct xfs_imap * imap,struct xfs_buf ** bpp)130 xfs_imap_to_bp(
131 	struct xfs_mount	*mp,
132 	struct xfs_trans	*tp,
133 	struct xfs_imap		*imap,
134 	struct xfs_buf		**bpp)
135 {
136 	return xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
137 				   imap->im_len, XBF_UNMAPPED, bpp,
138 				   &xfs_inode_buf_ops);
139 }
140 
xfs_inode_decode_bigtime(uint64_t ts)141 static inline struct timespec64 xfs_inode_decode_bigtime(uint64_t ts)
142 {
143 	struct timespec64	tv;
144 	uint32_t		n;
145 
146 	tv.tv_sec = xfs_bigtime_to_unix(div_u64_rem(ts, NSEC_PER_SEC, &n));
147 	tv.tv_nsec = n;
148 
149 	return tv;
150 }
151 
152 /* Convert an ondisk timestamp to an incore timestamp. */
153 struct timespec64
xfs_inode_from_disk_ts(struct xfs_dinode * dip,const xfs_timestamp_t ts)154 xfs_inode_from_disk_ts(
155 	struct xfs_dinode		*dip,
156 	const xfs_timestamp_t		ts)
157 {
158 	struct timespec64		tv;
159 	struct xfs_legacy_timestamp	*lts;
160 
161 	if (xfs_dinode_has_bigtime(dip))
162 		return xfs_inode_decode_bigtime(be64_to_cpu(ts));
163 
164 	lts = (struct xfs_legacy_timestamp *)&ts;
165 	tv.tv_sec = (int)be32_to_cpu(lts->t_sec);
166 	tv.tv_nsec = (int)be32_to_cpu(lts->t_nsec);
167 
168 	return tv;
169 }
170 
171 int
xfs_inode_from_disk(struct xfs_inode * ip,struct xfs_dinode * from)172 xfs_inode_from_disk(
173 	struct xfs_inode	*ip,
174 	struct xfs_dinode	*from)
175 {
176 	struct inode		*inode = VFS_I(ip);
177 	int			error;
178 	xfs_failaddr_t		fa;
179 
180 	ASSERT(ip->i_cowfp == NULL);
181 	ASSERT(ip->i_afp == NULL);
182 
183 	fa = xfs_dinode_verify(ip->i_mount, ip->i_ino, from);
184 	if (fa) {
185 		xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", from,
186 				sizeof(*from), fa);
187 		return -EFSCORRUPTED;
188 	}
189 
190 	/*
191 	 * First get the permanent information that is needed to allocate an
192 	 * inode. If the inode is unused, mode is zero and we shouldn't mess
193 	 * with the uninitialized part of it.
194 	 */
195 	if (!xfs_has_v3inodes(ip->i_mount))
196 		ip->i_flushiter = be16_to_cpu(from->di_flushiter);
197 	inode->i_generation = be32_to_cpu(from->di_gen);
198 	inode->i_mode = be16_to_cpu(from->di_mode);
199 	if (!inode->i_mode)
200 		return 0;
201 
202 	/*
203 	 * Convert v1 inodes immediately to v2 inode format as this is the
204 	 * minimum inode version format we support in the rest of the code.
205 	 * They will also be unconditionally written back to disk as v2 inodes.
206 	 */
207 	if (unlikely(from->di_version == 1)) {
208 		set_nlink(inode, be16_to_cpu(from->di_onlink));
209 		ip->i_projid = 0;
210 	} else {
211 		set_nlink(inode, be32_to_cpu(from->di_nlink));
212 		ip->i_projid = (prid_t)be16_to_cpu(from->di_projid_hi) << 16 |
213 					be16_to_cpu(from->di_projid_lo);
214 	}
215 
216 	i_uid_write(inode, be32_to_cpu(from->di_uid));
217 	i_gid_write(inode, be32_to_cpu(from->di_gid));
218 
219 	/*
220 	 * Time is signed, so need to convert to signed 32 bit before
221 	 * storing in inode timestamp which may be 64 bit. Otherwise
222 	 * a time before epoch is converted to a time long after epoch
223 	 * on 64 bit systems.
224 	 */
225 	inode->i_atime = xfs_inode_from_disk_ts(from, from->di_atime);
226 	inode->i_mtime = xfs_inode_from_disk_ts(from, from->di_mtime);
227 	inode->i_ctime = xfs_inode_from_disk_ts(from, from->di_ctime);
228 
229 	ip->i_disk_size = be64_to_cpu(from->di_size);
230 	ip->i_nblocks = be64_to_cpu(from->di_nblocks);
231 	ip->i_extsize = be32_to_cpu(from->di_extsize);
232 	ip->i_forkoff = from->di_forkoff;
233 	ip->i_diflags	= be16_to_cpu(from->di_flags);
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 void
xfs_inode_to_disk(struct xfs_inode * ip,struct xfs_dinode * to,xfs_lsn_t lsn)283 xfs_inode_to_disk(
284 	struct xfs_inode	*ip,
285 	struct xfs_dinode	*to,
286 	xfs_lsn_t		lsn)
287 {
288 	struct inode		*inode = VFS_I(ip);
289 
290 	to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
291 	to->di_onlink = 0;
292 
293 	to->di_format = xfs_ifork_format(&ip->i_df);
294 	to->di_uid = cpu_to_be32(i_uid_read(inode));
295 	to->di_gid = cpu_to_be32(i_gid_read(inode));
296 	to->di_projid_lo = cpu_to_be16(ip->i_projid & 0xffff);
297 	to->di_projid_hi = cpu_to_be16(ip->i_projid >> 16);
298 
299 	memset(to->di_pad, 0, sizeof(to->di_pad));
300 	to->di_atime = xfs_inode_to_disk_ts(ip, inode->i_atime);
301 	to->di_mtime = xfs_inode_to_disk_ts(ip, inode->i_mtime);
302 	to->di_ctime = xfs_inode_to_disk_ts(ip, inode->i_ctime);
303 	to->di_nlink = cpu_to_be32(inode->i_nlink);
304 	to->di_gen = cpu_to_be32(inode->i_generation);
305 	to->di_mode = cpu_to_be16(inode->i_mode);
306 
307 	to->di_size = cpu_to_be64(ip->i_disk_size);
308 	to->di_nblocks = cpu_to_be64(ip->i_nblocks);
309 	to->di_extsize = cpu_to_be32(ip->i_extsize);
310 	to->di_nextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_df));
311 	to->di_anextents = cpu_to_be16(xfs_ifork_nextents(ip->i_afp));
312 	to->di_forkoff = ip->i_forkoff;
313 	to->di_aformat = xfs_ifork_format(ip->i_afp);
314 	to->di_flags = cpu_to_be16(ip->i_diflags);
315 
316 	if (xfs_has_v3inodes(ip->i_mount)) {
317 		to->di_version = 3;
318 		to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
319 		to->di_crtime = xfs_inode_to_disk_ts(ip, ip->i_crtime);
320 		to->di_flags2 = cpu_to_be64(ip->i_diflags2);
321 		to->di_cowextsize = cpu_to_be32(ip->i_cowextsize);
322 		to->di_ino = cpu_to_be64(ip->i_ino);
323 		to->di_lsn = cpu_to_be64(lsn);
324 		memset(to->di_pad2, 0, sizeof(to->di_pad2));
325 		uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
326 		to->di_flushiter = 0;
327 	} else {
328 		to->di_version = 2;
329 		to->di_flushiter = cpu_to_be16(ip->i_flushiter);
330 	}
331 }
332 
333 static xfs_failaddr_t
xfs_dinode_verify_fork(struct xfs_dinode * dip,struct xfs_mount * mp,int whichfork)334 xfs_dinode_verify_fork(
335 	struct xfs_dinode	*dip,
336 	struct xfs_mount	*mp,
337 	int			whichfork)
338 {
339 	uint32_t		di_nextents = XFS_DFORK_NEXTENTS(dip, whichfork);
340 	mode_t			mode = be16_to_cpu(dip->di_mode);
341 	uint32_t		fork_size = XFS_DFORK_SIZE(dip, mp, whichfork);
342 	uint32_t		fork_format = XFS_DFORK_FORMAT(dip, whichfork);
343 
344 	/*
345 	 * For fork types that can contain local data, check that the fork
346 	 * format matches the size of local data contained within the fork.
347 	 *
348 	 * For all types, check that when the size says the should be in extent
349 	 * or btree format, the inode isn't claiming it is in local format.
350 	 */
351 	if (whichfork == XFS_DATA_FORK) {
352 		if (S_ISDIR(mode) || S_ISLNK(mode)) {
353 			if (be64_to_cpu(dip->di_size) <= fork_size &&
354 			    fork_format != XFS_DINODE_FMT_LOCAL)
355 				return __this_address;
356 		}
357 
358 		if (be64_to_cpu(dip->di_size) > fork_size &&
359 		    fork_format == XFS_DINODE_FMT_LOCAL)
360 			return __this_address;
361 	}
362 
363 	switch (fork_format) {
364 	case XFS_DINODE_FMT_LOCAL:
365 		/*
366 		 * No local regular files yet.
367 		 */
368 		if (S_ISREG(mode) && whichfork == XFS_DATA_FORK)
369 			return __this_address;
370 		if (di_nextents)
371 			return __this_address;
372 		break;
373 	case XFS_DINODE_FMT_EXTENTS:
374 		if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
375 			return __this_address;
376 		break;
377 	case XFS_DINODE_FMT_BTREE:
378 		if (whichfork == XFS_ATTR_FORK) {
379 			if (di_nextents > MAXAEXTNUM)
380 				return __this_address;
381 		} else if (di_nextents > MAXEXTNUM) {
382 			return __this_address;
383 		}
384 		break;
385 	default:
386 		return __this_address;
387 	}
388 	return NULL;
389 }
390 
391 static xfs_failaddr_t
xfs_dinode_verify_forkoff(struct xfs_dinode * dip,struct xfs_mount * mp)392 xfs_dinode_verify_forkoff(
393 	struct xfs_dinode	*dip,
394 	struct xfs_mount	*mp)
395 {
396 	if (!dip->di_forkoff)
397 		return NULL;
398 
399 	switch (dip->di_format)  {
400 	case XFS_DINODE_FMT_DEV:
401 		if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
402 			return __this_address;
403 		break;
404 	case XFS_DINODE_FMT_LOCAL:	/* fall through ... */
405 	case XFS_DINODE_FMT_EXTENTS:    /* fall through ... */
406 	case XFS_DINODE_FMT_BTREE:
407 		if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3))
408 			return __this_address;
409 		break;
410 	default:
411 		return __this_address;
412 	}
413 	return NULL;
414 }
415 
416 xfs_failaddr_t
xfs_dinode_verify(struct xfs_mount * mp,xfs_ino_t ino,struct xfs_dinode * dip)417 xfs_dinode_verify(
418 	struct xfs_mount	*mp,
419 	xfs_ino_t		ino,
420 	struct xfs_dinode	*dip)
421 {
422 	xfs_failaddr_t		fa;
423 	uint16_t		mode;
424 	uint16_t		flags;
425 	uint64_t		flags2;
426 	uint64_t		di_size;
427 
428 	if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
429 		return __this_address;
430 
431 	/* Verify v3 integrity information first */
432 	if (dip->di_version >= 3) {
433 		if (!xfs_has_v3inodes(mp))
434 			return __this_address;
435 		if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
436 				      XFS_DINODE_CRC_OFF))
437 			return __this_address;
438 		if (be64_to_cpu(dip->di_ino) != ino)
439 			return __this_address;
440 		if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
441 			return __this_address;
442 	}
443 
444 	/* don't allow invalid i_size */
445 	di_size = be64_to_cpu(dip->di_size);
446 	if (di_size & (1ULL << 63))
447 		return __this_address;
448 
449 	mode = be16_to_cpu(dip->di_mode);
450 	if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
451 		return __this_address;
452 
453 	/* No zero-length symlinks/dirs. */
454 	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
455 		return __this_address;
456 
457 	/* Fork checks carried over from xfs_iformat_fork */
458 	if (mode &&
459 	    be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
460 			be64_to_cpu(dip->di_nblocks))
461 		return __this_address;
462 
463 	if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
464 		return __this_address;
465 
466 	flags = be16_to_cpu(dip->di_flags);
467 
468 	if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
469 		return __this_address;
470 
471 	/* check for illegal values of forkoff */
472 	fa = xfs_dinode_verify_forkoff(dip, mp);
473 	if (fa)
474 		return fa;
475 
476 	/* Do we have appropriate data fork formats for the mode? */
477 	switch (mode & S_IFMT) {
478 	case S_IFIFO:
479 	case S_IFCHR:
480 	case S_IFBLK:
481 	case S_IFSOCK:
482 		if (dip->di_format != XFS_DINODE_FMT_DEV)
483 			return __this_address;
484 		break;
485 	case S_IFREG:
486 	case S_IFLNK:
487 	case S_IFDIR:
488 		fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
489 		if (fa)
490 			return fa;
491 		break;
492 	case 0:
493 		/* Uninitialized inode ok. */
494 		break;
495 	default:
496 		return __this_address;
497 	}
498 
499 	if (dip->di_forkoff) {
500 		fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
501 		if (fa)
502 			return fa;
503 	} else {
504 		/*
505 		 * If there is no fork offset, this may be a freshly-made inode
506 		 * in a new disk cluster, in which case di_aformat is zeroed.
507 		 * Otherwise, such an inode must be in EXTENTS format; this goes
508 		 * for freed inodes as well.
509 		 */
510 		switch (dip->di_aformat) {
511 		case 0:
512 		case XFS_DINODE_FMT_EXTENTS:
513 			break;
514 		default:
515 			return __this_address;
516 		}
517 		if (dip->di_anextents)
518 			return __this_address;
519 	}
520 
521 	/* extent size hint validation */
522 	fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
523 			mode, flags);
524 	if (fa)
525 		return fa;
526 
527 	/* only version 3 or greater inodes are extensively verified here */
528 	if (dip->di_version < 3)
529 		return NULL;
530 
531 	flags2 = be64_to_cpu(dip->di_flags2);
532 
533 	/* don't allow reflink/cowextsize if we don't have reflink */
534 	if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
535 	     !xfs_has_reflink(mp))
536 		return __this_address;
537 
538 	/* only regular files get reflink */
539 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
540 		return __this_address;
541 
542 	/* don't let reflink and realtime mix */
543 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
544 		return __this_address;
545 
546 	/* COW extent size hint validation */
547 	fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
548 			mode, flags, flags2);
549 	if (fa)
550 		return fa;
551 
552 	/* bigtime iflag can only happen on bigtime filesystems */
553 	if (xfs_dinode_has_bigtime(dip) &&
554 	    !xfs_has_bigtime(mp))
555 		return __this_address;
556 
557 	return NULL;
558 }
559 
560 void
xfs_dinode_calc_crc(struct xfs_mount * mp,struct xfs_dinode * dip)561 xfs_dinode_calc_crc(
562 	struct xfs_mount	*mp,
563 	struct xfs_dinode	*dip)
564 {
565 	uint32_t		crc;
566 
567 	if (dip->di_version < 3)
568 		return;
569 
570 	ASSERT(xfs_has_crc(mp));
571 	crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
572 			      XFS_DINODE_CRC_OFF);
573 	dip->di_crc = xfs_end_cksum(crc);
574 }
575 
576 /*
577  * Validate di_extsize hint.
578  *
579  * 1. Extent size hint is only valid for directories and regular files.
580  * 2. FS_XFLAG_EXTSIZE is only valid for regular files.
581  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
582  * 4. Hint cannot be larger than MAXTEXTLEN.
583  * 5. Can be changed on directories at any time.
584  * 6. Hint value of 0 turns off hints, clears inode flags.
585  * 7. Extent size must be a multiple of the appropriate block size.
586  *    For realtime files, this is the rt extent size.
587  * 8. For non-realtime files, the extent size hint must be limited
588  *    to half the AG size to avoid alignment extending the extent beyond the
589  *    limits of the AG.
590  */
591 xfs_failaddr_t
xfs_inode_validate_extsize(struct xfs_mount * mp,uint32_t extsize,uint16_t mode,uint16_t flags)592 xfs_inode_validate_extsize(
593 	struct xfs_mount		*mp,
594 	uint32_t			extsize,
595 	uint16_t			mode,
596 	uint16_t			flags)
597 {
598 	bool				rt_flag;
599 	bool				hint_flag;
600 	bool				inherit_flag;
601 	uint32_t			extsize_bytes;
602 	uint32_t			blocksize_bytes;
603 
604 	rt_flag = (flags & XFS_DIFLAG_REALTIME);
605 	hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
606 	inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
607 	extsize_bytes = XFS_FSB_TO_B(mp, extsize);
608 
609 	/*
610 	 * This comment describes a historic gap in this verifier function.
611 	 *
612 	 * For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
613 	 * function has never checked that the extent size hint is an integer
614 	 * multiple of the realtime extent size.  Since we allow users to set
615 	 * this combination  on non-rt filesystems /and/ to change the rt
616 	 * extent size when adding a rt device to a filesystem, the net effect
617 	 * is that users can configure a filesystem anticipating one rt
618 	 * geometry and change their minds later.  Directories do not use the
619 	 * extent size hint, so this is harmless for them.
620 	 *
621 	 * If a directory with a misaligned extent size hint is allowed to
622 	 * propagate that hint into a new regular realtime file, the result
623 	 * is that the inode cluster buffer verifier will trigger a corruption
624 	 * shutdown the next time it is run, because the verifier has always
625 	 * enforced the alignment rule for regular files.
626 	 *
627 	 * Because we allow administrators to set a new rt extent size when
628 	 * adding a rt section, we cannot add a check to this verifier because
629 	 * that will result a new source of directory corruption errors when
630 	 * reading an existing filesystem.  Instead, we rely on callers to
631 	 * decide when alignment checks are appropriate, and fix things up as
632 	 * needed.
633 	 */
634 
635 	if (rt_flag)
636 		blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
637 	else
638 		blocksize_bytes = mp->m_sb.sb_blocksize;
639 
640 	if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
641 		return __this_address;
642 
643 	if (hint_flag && !S_ISREG(mode))
644 		return __this_address;
645 
646 	if (inherit_flag && !S_ISDIR(mode))
647 		return __this_address;
648 
649 	if ((hint_flag || inherit_flag) && extsize == 0)
650 		return __this_address;
651 
652 	/* free inodes get flags set to zero but extsize remains */
653 	if (mode && !(hint_flag || inherit_flag) && extsize != 0)
654 		return __this_address;
655 
656 	if (extsize_bytes % blocksize_bytes)
657 		return __this_address;
658 
659 	if (extsize > MAXEXTLEN)
660 		return __this_address;
661 
662 	if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
663 		return __this_address;
664 
665 	return NULL;
666 }
667 
668 /*
669  * Validate di_cowextsize hint.
670  *
671  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
672  *    The inode does not have to have any shared blocks, but it must be a v3.
673  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
674  *    for a directory, the hint is propagated to new files.
675  * 3. Can be changed on files & directories at any time.
676  * 4. Hint value of 0 turns off hints, clears inode flags.
677  * 5. Extent size must be a multiple of the appropriate block size.
678  * 6. The extent size hint must be limited to half the AG size to avoid
679  *    alignment extending the extent beyond the limits of the AG.
680  */
681 xfs_failaddr_t
xfs_inode_validate_cowextsize(struct xfs_mount * mp,uint32_t cowextsize,uint16_t mode,uint16_t flags,uint64_t flags2)682 xfs_inode_validate_cowextsize(
683 	struct xfs_mount		*mp,
684 	uint32_t			cowextsize,
685 	uint16_t			mode,
686 	uint16_t			flags,
687 	uint64_t			flags2)
688 {
689 	bool				rt_flag;
690 	bool				hint_flag;
691 	uint32_t			cowextsize_bytes;
692 
693 	rt_flag = (flags & XFS_DIFLAG_REALTIME);
694 	hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
695 	cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
696 
697 	if (hint_flag && !xfs_has_reflink(mp))
698 		return __this_address;
699 
700 	if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
701 		return __this_address;
702 
703 	if (hint_flag && cowextsize == 0)
704 		return __this_address;
705 
706 	/* free inodes get flags set to zero but cowextsize remains */
707 	if (mode && !hint_flag && cowextsize != 0)
708 		return __this_address;
709 
710 	if (hint_flag && rt_flag)
711 		return __this_address;
712 
713 	if (cowextsize_bytes % mp->m_sb.sb_blocksize)
714 		return __this_address;
715 
716 	if (cowextsize > MAXEXTLEN)
717 		return __this_address;
718 
719 	if (cowextsize > mp->m_sb.sb_agblocks / 2)
720 		return __this_address;
721 
722 	return NULL;
723 }
724