xref: /openbmc/linux/fs/xfs/libxfs/xfs_inode_buf.c (revision 62e59c4e)
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_defer.h"
14 #include "xfs_inode.h"
15 #include "xfs_errortag.h"
16 #include "xfs_error.h"
17 #include "xfs_cksum.h"
18 #include "xfs_icache.h"
19 #include "xfs_trans.h"
20 #include "xfs_ialloc.h"
21 #include "xfs_dir2.h"
22 
23 #include <linux/iversion.h>
24 
25 /*
26  * Check that none of the inode's in the buffer have a next
27  * unlinked field of 0.
28  */
29 #if defined(DEBUG)
30 void
31 xfs_inobp_check(
32 	xfs_mount_t	*mp,
33 	xfs_buf_t	*bp)
34 {
35 	int		i;
36 	int		j;
37 	xfs_dinode_t	*dip;
38 
39 	j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;
40 
41 	for (i = 0; i < j; i++) {
42 		dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize);
43 		if (!dip->di_next_unlinked)  {
44 			xfs_alert(mp,
45 	"Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
46 				i, (long long)bp->b_bn);
47 		}
48 	}
49 }
50 #endif
51 
52 bool
53 xfs_dinode_good_version(
54 	struct xfs_mount *mp,
55 	__u8		version)
56 {
57 	if (xfs_sb_version_hascrc(&mp->m_sb))
58 		return version == 3;
59 
60 	return version == 1 || version == 2;
61 }
62 
63 /*
64  * If we are doing readahead on an inode buffer, we might be in log recovery
65  * reading an inode allocation buffer that hasn't yet been replayed, and hence
66  * has not had the inode cores stamped into it. Hence for readahead, the buffer
67  * may be potentially invalid.
68  *
69  * If the readahead buffer is invalid, we need to mark it with an error and
70  * clear the DONE status of the buffer so that a followup read will re-read it
71  * from disk. We don't report the error otherwise to avoid warnings during log
72  * recovery and we don't get unnecssary panics on debug kernels. We use EIO here
73  * because all we want to do is say readahead failed; there is no-one to report
74  * the error to, so this will distinguish it from a non-ra verifier failure.
75  * Changes to this readahead error behavour also need to be reflected in
76  * xfs_dquot_buf_readahead_verify().
77  */
78 static void
79 xfs_inode_buf_verify(
80 	struct xfs_buf	*bp,
81 	bool		readahead)
82 {
83 	struct xfs_mount *mp = bp->b_target->bt_mount;
84 	xfs_agnumber_t	agno;
85 	int		i;
86 	int		ni;
87 
88 	/*
89 	 * Validate the magic number and version of every inode in the buffer
90 	 */
91 	agno = xfs_daddr_to_agno(mp, XFS_BUF_ADDR(bp));
92 	ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
93 	for (i = 0; i < ni; i++) {
94 		int		di_ok;
95 		xfs_dinode_t	*dip;
96 		xfs_agino_t	unlinked_ino;
97 
98 		dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
99 		unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
100 		di_ok = xfs_verify_magic16(bp, dip->di_magic) &&
101 			xfs_dinode_good_version(mp, dip->di_version) &&
102 			xfs_verify_agino_or_null(mp, agno, unlinked_ino);
103 		if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
104 						XFS_ERRTAG_ITOBP_INOTOBP))) {
105 			if (readahead) {
106 				bp->b_flags &= ~XBF_DONE;
107 				xfs_buf_ioerror(bp, -EIO);
108 				return;
109 			}
110 
111 #ifdef DEBUG
112 			xfs_alert(mp,
113 				"bad inode magic/vsn daddr %lld #%d (magic=%x)",
114 				(unsigned long long)bp->b_bn, i,
115 				be16_to_cpu(dip->di_magic));
116 #endif
117 			xfs_buf_verifier_error(bp, -EFSCORRUPTED,
118 					__func__, dip, sizeof(*dip),
119 					NULL);
120 			return;
121 		}
122 	}
123 }
124 
125 
126 static void
127 xfs_inode_buf_read_verify(
128 	struct xfs_buf	*bp)
129 {
130 	xfs_inode_buf_verify(bp, false);
131 }
132 
133 static void
134 xfs_inode_buf_readahead_verify(
135 	struct xfs_buf	*bp)
136 {
137 	xfs_inode_buf_verify(bp, true);
138 }
139 
140 static void
141 xfs_inode_buf_write_verify(
142 	struct xfs_buf	*bp)
143 {
144 	xfs_inode_buf_verify(bp, false);
145 }
146 
147 const struct xfs_buf_ops xfs_inode_buf_ops = {
148 	.name = "xfs_inode",
149 	.magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
150 		     cpu_to_be16(XFS_DINODE_MAGIC) },
151 	.verify_read = xfs_inode_buf_read_verify,
152 	.verify_write = xfs_inode_buf_write_verify,
153 };
154 
155 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
156 	.name = "xfs_inode_ra",
157 	.magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
158 		     cpu_to_be16(XFS_DINODE_MAGIC) },
159 	.verify_read = xfs_inode_buf_readahead_verify,
160 	.verify_write = xfs_inode_buf_write_verify,
161 };
162 
163 
164 /*
165  * This routine is called to map an inode to the buffer containing the on-disk
166  * version of the inode.  It returns a pointer to the buffer containing the
167  * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
168  * pointer to the on-disk inode within that buffer.
169  *
170  * If a non-zero error is returned, then the contents of bpp and dipp are
171  * undefined.
172  */
173 int
174 xfs_imap_to_bp(
175 	struct xfs_mount	*mp,
176 	struct xfs_trans	*tp,
177 	struct xfs_imap		*imap,
178 	struct xfs_dinode       **dipp,
179 	struct xfs_buf		**bpp,
180 	uint			buf_flags,
181 	uint			iget_flags)
182 {
183 	struct xfs_buf		*bp;
184 	int			error;
185 
186 	buf_flags |= XBF_UNMAPPED;
187 	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
188 				   (int)imap->im_len, buf_flags, &bp,
189 				   &xfs_inode_buf_ops);
190 	if (error) {
191 		if (error == -EAGAIN) {
192 			ASSERT(buf_flags & XBF_TRYLOCK);
193 			return error;
194 		}
195 		xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
196 			__func__, error);
197 		return error;
198 	}
199 
200 	*bpp = bp;
201 	*dipp = xfs_buf_offset(bp, imap->im_boffset);
202 	return 0;
203 }
204 
205 void
206 xfs_inode_from_disk(
207 	struct xfs_inode	*ip,
208 	struct xfs_dinode	*from)
209 {
210 	struct xfs_icdinode	*to = &ip->i_d;
211 	struct inode		*inode = VFS_I(ip);
212 
213 
214 	/*
215 	 * Convert v1 inodes immediately to v2 inode format as this is the
216 	 * minimum inode version format we support in the rest of the code.
217 	 */
218 	to->di_version = from->di_version;
219 	if (to->di_version == 1) {
220 		set_nlink(inode, be16_to_cpu(from->di_onlink));
221 		to->di_projid_lo = 0;
222 		to->di_projid_hi = 0;
223 		to->di_version = 2;
224 	} else {
225 		set_nlink(inode, be32_to_cpu(from->di_nlink));
226 		to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
227 		to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
228 	}
229 
230 	to->di_format = from->di_format;
231 	to->di_uid = be32_to_cpu(from->di_uid);
232 	to->di_gid = be32_to_cpu(from->di_gid);
233 	to->di_flushiter = be16_to_cpu(from->di_flushiter);
234 
235 	/*
236 	 * Time is signed, so need to convert to signed 32 bit before
237 	 * storing in inode timestamp which may be 64 bit. Otherwise
238 	 * a time before epoch is converted to a time long after epoch
239 	 * on 64 bit systems.
240 	 */
241 	inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec);
242 	inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec);
243 	inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec);
244 	inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec);
245 	inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec);
246 	inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec);
247 	inode->i_generation = be32_to_cpu(from->di_gen);
248 	inode->i_mode = be16_to_cpu(from->di_mode);
249 
250 	to->di_size = be64_to_cpu(from->di_size);
251 	to->di_nblocks = be64_to_cpu(from->di_nblocks);
252 	to->di_extsize = be32_to_cpu(from->di_extsize);
253 	to->di_nextents = be32_to_cpu(from->di_nextents);
254 	to->di_anextents = be16_to_cpu(from->di_anextents);
255 	to->di_forkoff = from->di_forkoff;
256 	to->di_aformat	= from->di_aformat;
257 	to->di_dmevmask	= be32_to_cpu(from->di_dmevmask);
258 	to->di_dmstate	= be16_to_cpu(from->di_dmstate);
259 	to->di_flags	= be16_to_cpu(from->di_flags);
260 
261 	if (to->di_version == 3) {
262 		inode_set_iversion_queried(inode,
263 					   be64_to_cpu(from->di_changecount));
264 		to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
265 		to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
266 		to->di_flags2 = be64_to_cpu(from->di_flags2);
267 		to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
268 	}
269 }
270 
271 void
272 xfs_inode_to_disk(
273 	struct xfs_inode	*ip,
274 	struct xfs_dinode	*to,
275 	xfs_lsn_t		lsn)
276 {
277 	struct xfs_icdinode	*from = &ip->i_d;
278 	struct inode		*inode = VFS_I(ip);
279 
280 	to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
281 	to->di_onlink = 0;
282 
283 	to->di_version = from->di_version;
284 	to->di_format = from->di_format;
285 	to->di_uid = cpu_to_be32(from->di_uid);
286 	to->di_gid = cpu_to_be32(from->di_gid);
287 	to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
288 	to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
289 
290 	memset(to->di_pad, 0, sizeof(to->di_pad));
291 	to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec);
292 	to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec);
293 	to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec);
294 	to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec);
295 	to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec);
296 	to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec);
297 	to->di_nlink = cpu_to_be32(inode->i_nlink);
298 	to->di_gen = cpu_to_be32(inode->i_generation);
299 	to->di_mode = cpu_to_be16(inode->i_mode);
300 
301 	to->di_size = cpu_to_be64(from->di_size);
302 	to->di_nblocks = cpu_to_be64(from->di_nblocks);
303 	to->di_extsize = cpu_to_be32(from->di_extsize);
304 	to->di_nextents = cpu_to_be32(from->di_nextents);
305 	to->di_anextents = cpu_to_be16(from->di_anextents);
306 	to->di_forkoff = from->di_forkoff;
307 	to->di_aformat = from->di_aformat;
308 	to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
309 	to->di_dmstate = cpu_to_be16(from->di_dmstate);
310 	to->di_flags = cpu_to_be16(from->di_flags);
311 
312 	if (from->di_version == 3) {
313 		to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
314 		to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
315 		to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
316 		to->di_flags2 = cpu_to_be64(from->di_flags2);
317 		to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
318 		to->di_ino = cpu_to_be64(ip->i_ino);
319 		to->di_lsn = cpu_to_be64(lsn);
320 		memset(to->di_pad2, 0, sizeof(to->di_pad2));
321 		uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
322 		to->di_flushiter = 0;
323 	} else {
324 		to->di_flushiter = cpu_to_be16(from->di_flushiter);
325 	}
326 }
327 
328 void
329 xfs_log_dinode_to_disk(
330 	struct xfs_log_dinode	*from,
331 	struct xfs_dinode	*to)
332 {
333 	to->di_magic = cpu_to_be16(from->di_magic);
334 	to->di_mode = cpu_to_be16(from->di_mode);
335 	to->di_version = from->di_version;
336 	to->di_format = from->di_format;
337 	to->di_onlink = 0;
338 	to->di_uid = cpu_to_be32(from->di_uid);
339 	to->di_gid = cpu_to_be32(from->di_gid);
340 	to->di_nlink = cpu_to_be32(from->di_nlink);
341 	to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
342 	to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
343 	memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
344 
345 	to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
346 	to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
347 	to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
348 	to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
349 	to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
350 	to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
351 
352 	to->di_size = cpu_to_be64(from->di_size);
353 	to->di_nblocks = cpu_to_be64(from->di_nblocks);
354 	to->di_extsize = cpu_to_be32(from->di_extsize);
355 	to->di_nextents = cpu_to_be32(from->di_nextents);
356 	to->di_anextents = cpu_to_be16(from->di_anextents);
357 	to->di_forkoff = from->di_forkoff;
358 	to->di_aformat = from->di_aformat;
359 	to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
360 	to->di_dmstate = cpu_to_be16(from->di_dmstate);
361 	to->di_flags = cpu_to_be16(from->di_flags);
362 	to->di_gen = cpu_to_be32(from->di_gen);
363 
364 	if (from->di_version == 3) {
365 		to->di_changecount = cpu_to_be64(from->di_changecount);
366 		to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
367 		to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
368 		to->di_flags2 = cpu_to_be64(from->di_flags2);
369 		to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
370 		to->di_ino = cpu_to_be64(from->di_ino);
371 		to->di_lsn = cpu_to_be64(from->di_lsn);
372 		memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
373 		uuid_copy(&to->di_uuid, &from->di_uuid);
374 		to->di_flushiter = 0;
375 	} else {
376 		to->di_flushiter = cpu_to_be16(from->di_flushiter);
377 	}
378 }
379 
380 static xfs_failaddr_t
381 xfs_dinode_verify_fork(
382 	struct xfs_dinode	*dip,
383 	struct xfs_mount	*mp,
384 	int			whichfork)
385 {
386 	uint32_t		di_nextents = XFS_DFORK_NEXTENTS(dip, whichfork);
387 
388 	switch (XFS_DFORK_FORMAT(dip, whichfork)) {
389 	case XFS_DINODE_FMT_LOCAL:
390 		/*
391 		 * no local regular files yet
392 		 */
393 		if (whichfork == XFS_DATA_FORK) {
394 			if (S_ISREG(be16_to_cpu(dip->di_mode)))
395 				return __this_address;
396 			if (be64_to_cpu(dip->di_size) >
397 					XFS_DFORK_SIZE(dip, mp, whichfork))
398 				return __this_address;
399 		}
400 		if (di_nextents)
401 			return __this_address;
402 		break;
403 	case XFS_DINODE_FMT_EXTENTS:
404 		if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
405 			return __this_address;
406 		break;
407 	case XFS_DINODE_FMT_BTREE:
408 		if (whichfork == XFS_ATTR_FORK) {
409 			if (di_nextents > MAXAEXTNUM)
410 				return __this_address;
411 		} else if (di_nextents > MAXEXTNUM) {
412 			return __this_address;
413 		}
414 		break;
415 	default:
416 		return __this_address;
417 	}
418 	return NULL;
419 }
420 
421 static xfs_failaddr_t
422 xfs_dinode_verify_forkoff(
423 	struct xfs_dinode	*dip,
424 	struct xfs_mount	*mp)
425 {
426 	if (!XFS_DFORK_Q(dip))
427 		return NULL;
428 
429 	switch (dip->di_format)  {
430 	case XFS_DINODE_FMT_DEV:
431 		if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
432 			return __this_address;
433 		break;
434 	case XFS_DINODE_FMT_LOCAL:	/* fall through ... */
435 	case XFS_DINODE_FMT_EXTENTS:    /* fall through ... */
436 	case XFS_DINODE_FMT_BTREE:
437 		if (dip->di_forkoff >= (XFS_LITINO(mp, dip->di_version) >> 3))
438 			return __this_address;
439 		break;
440 	default:
441 		return __this_address;
442 	}
443 	return NULL;
444 }
445 
446 xfs_failaddr_t
447 xfs_dinode_verify(
448 	struct xfs_mount	*mp,
449 	xfs_ino_t		ino,
450 	struct xfs_dinode	*dip)
451 {
452 	xfs_failaddr_t		fa;
453 	uint16_t		mode;
454 	uint16_t		flags;
455 	uint64_t		flags2;
456 	uint64_t		di_size;
457 
458 	if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
459 		return __this_address;
460 
461 	/* Verify v3 integrity information first */
462 	if (dip->di_version >= 3) {
463 		if (!xfs_sb_version_hascrc(&mp->m_sb))
464 			return __this_address;
465 		if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
466 				      XFS_DINODE_CRC_OFF))
467 			return __this_address;
468 		if (be64_to_cpu(dip->di_ino) != ino)
469 			return __this_address;
470 		if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
471 			return __this_address;
472 	}
473 
474 	/* don't allow invalid i_size */
475 	di_size = be64_to_cpu(dip->di_size);
476 	if (di_size & (1ULL << 63))
477 		return __this_address;
478 
479 	mode = be16_to_cpu(dip->di_mode);
480 	if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
481 		return __this_address;
482 
483 	/* No zero-length symlinks/dirs. */
484 	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
485 		return __this_address;
486 
487 	/* Fork checks carried over from xfs_iformat_fork */
488 	if (mode &&
489 	    be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
490 			be64_to_cpu(dip->di_nblocks))
491 		return __this_address;
492 
493 	if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
494 		return __this_address;
495 
496 	flags = be16_to_cpu(dip->di_flags);
497 
498 	if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
499 		return __this_address;
500 
501 	/* check for illegal values of forkoff */
502 	fa = xfs_dinode_verify_forkoff(dip, mp);
503 	if (fa)
504 		return fa;
505 
506 	/* Do we have appropriate data fork formats for the mode? */
507 	switch (mode & S_IFMT) {
508 	case S_IFIFO:
509 	case S_IFCHR:
510 	case S_IFBLK:
511 	case S_IFSOCK:
512 		if (dip->di_format != XFS_DINODE_FMT_DEV)
513 			return __this_address;
514 		break;
515 	case S_IFREG:
516 	case S_IFLNK:
517 	case S_IFDIR:
518 		fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
519 		if (fa)
520 			return fa;
521 		break;
522 	case 0:
523 		/* Uninitialized inode ok. */
524 		break;
525 	default:
526 		return __this_address;
527 	}
528 
529 	if (XFS_DFORK_Q(dip)) {
530 		fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
531 		if (fa)
532 			return fa;
533 	} else {
534 		/*
535 		 * If there is no fork offset, this may be a freshly-made inode
536 		 * in a new disk cluster, in which case di_aformat is zeroed.
537 		 * Otherwise, such an inode must be in EXTENTS format; this goes
538 		 * for freed inodes as well.
539 		 */
540 		switch (dip->di_aformat) {
541 		case 0:
542 		case XFS_DINODE_FMT_EXTENTS:
543 			break;
544 		default:
545 			return __this_address;
546 		}
547 		if (dip->di_anextents)
548 			return __this_address;
549 	}
550 
551 	/* extent size hint validation */
552 	fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
553 			mode, flags);
554 	if (fa)
555 		return fa;
556 
557 	/* only version 3 or greater inodes are extensively verified here */
558 	if (dip->di_version < 3)
559 		return NULL;
560 
561 	flags2 = be64_to_cpu(dip->di_flags2);
562 
563 	/* don't allow reflink/cowextsize if we don't have reflink */
564 	if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
565 	     !xfs_sb_version_hasreflink(&mp->m_sb))
566 		return __this_address;
567 
568 	/* only regular files get reflink */
569 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
570 		return __this_address;
571 
572 	/* don't let reflink and realtime mix */
573 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
574 		return __this_address;
575 
576 	/* don't let reflink and dax mix */
577 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
578 		return __this_address;
579 
580 	/* COW extent size hint validation */
581 	fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
582 			mode, flags, flags2);
583 	if (fa)
584 		return fa;
585 
586 	return NULL;
587 }
588 
589 void
590 xfs_dinode_calc_crc(
591 	struct xfs_mount	*mp,
592 	struct xfs_dinode	*dip)
593 {
594 	uint32_t		crc;
595 
596 	if (dip->di_version < 3)
597 		return;
598 
599 	ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
600 	crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
601 			      XFS_DINODE_CRC_OFF);
602 	dip->di_crc = xfs_end_cksum(crc);
603 }
604 
605 /*
606  * Read the disk inode attributes into the in-core inode structure.
607  *
608  * For version 5 superblocks, if we are initialising a new inode and we are not
609  * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
610  * inode core with a random generation number. If we are keeping inodes around,
611  * we need to read the inode cluster to get the existing generation number off
612  * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
613  * format) then log recovery is dependent on the di_flushiter field being
614  * initialised from the current on-disk value and hence we must also read the
615  * inode off disk.
616  */
617 int
618 xfs_iread(
619 	xfs_mount_t	*mp,
620 	xfs_trans_t	*tp,
621 	xfs_inode_t	*ip,
622 	uint		iget_flags)
623 {
624 	xfs_buf_t	*bp;
625 	xfs_dinode_t	*dip;
626 	xfs_failaddr_t	fa;
627 	int		error;
628 
629 	/*
630 	 * Fill in the location information in the in-core inode.
631 	 */
632 	error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
633 	if (error)
634 		return error;
635 
636 	/* shortcut IO on inode allocation if possible */
637 	if ((iget_flags & XFS_IGET_CREATE) &&
638 	    xfs_sb_version_hascrc(&mp->m_sb) &&
639 	    !(mp->m_flags & XFS_MOUNT_IKEEP)) {
640 		/* initialise the on-disk inode core */
641 		memset(&ip->i_d, 0, sizeof(ip->i_d));
642 		VFS_I(ip)->i_generation = prandom_u32();
643 		ip->i_d.di_version = 3;
644 		return 0;
645 	}
646 
647 	/*
648 	 * Get pointers to the on-disk inode and the buffer containing it.
649 	 */
650 	error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
651 	if (error)
652 		return error;
653 
654 	/* even unallocated inodes are verified */
655 	fa = xfs_dinode_verify(mp, ip->i_ino, dip);
656 	if (fa) {
657 		xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
658 				sizeof(*dip), fa);
659 		error = -EFSCORRUPTED;
660 		goto out_brelse;
661 	}
662 
663 	/*
664 	 * If the on-disk inode is already linked to a directory
665 	 * entry, copy all of the inode into the in-core inode.
666 	 * xfs_iformat_fork() handles copying in the inode format
667 	 * specific information.
668 	 * Otherwise, just get the truly permanent information.
669 	 */
670 	if (dip->di_mode) {
671 		xfs_inode_from_disk(ip, dip);
672 		error = xfs_iformat_fork(ip, dip);
673 		if (error)  {
674 #ifdef DEBUG
675 			xfs_alert(mp, "%s: xfs_iformat() returned error %d",
676 				__func__, error);
677 #endif /* DEBUG */
678 			goto out_brelse;
679 		}
680 	} else {
681 		/*
682 		 * Partial initialisation of the in-core inode. Just the bits
683 		 * that xfs_ialloc won't overwrite or relies on being correct.
684 		 */
685 		ip->i_d.di_version = dip->di_version;
686 		VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
687 		ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
688 
689 		/*
690 		 * Make sure to pull in the mode here as well in
691 		 * case the inode is released without being used.
692 		 * This ensures that xfs_inactive() will see that
693 		 * the inode is already free and not try to mess
694 		 * with the uninitialized part of it.
695 		 */
696 		VFS_I(ip)->i_mode = 0;
697 	}
698 
699 	ASSERT(ip->i_d.di_version >= 2);
700 	ip->i_delayed_blks = 0;
701 
702 	/*
703 	 * Mark the buffer containing the inode as something to keep
704 	 * around for a while.  This helps to keep recently accessed
705 	 * meta-data in-core longer.
706 	 */
707 	xfs_buf_set_ref(bp, XFS_INO_REF);
708 
709 	/*
710 	 * Use xfs_trans_brelse() to release the buffer containing the on-disk
711 	 * inode, because it was acquired with xfs_trans_read_buf() in
712 	 * xfs_imap_to_bp() above.  If tp is NULL, this is just a normal
713 	 * brelse().  If we're within a transaction, then xfs_trans_brelse()
714 	 * will only release the buffer if it is not dirty within the
715 	 * transaction.  It will be OK to release the buffer in this case,
716 	 * because inodes on disk are never destroyed and we will be locking the
717 	 * new in-core inode before putting it in the cache where other
718 	 * processes can find it.  Thus we don't have to worry about the inode
719 	 * being changed just because we released the buffer.
720 	 */
721  out_brelse:
722 	xfs_trans_brelse(tp, bp);
723 	return error;
724 }
725 
726 /*
727  * Validate di_extsize hint.
728  *
729  * The rules are documented at xfs_ioctl_setattr_check_extsize().
730  * These functions must be kept in sync with each other.
731  */
732 xfs_failaddr_t
733 xfs_inode_validate_extsize(
734 	struct xfs_mount		*mp,
735 	uint32_t			extsize,
736 	uint16_t			mode,
737 	uint16_t			flags)
738 {
739 	bool				rt_flag;
740 	bool				hint_flag;
741 	bool				inherit_flag;
742 	uint32_t			extsize_bytes;
743 	uint32_t			blocksize_bytes;
744 
745 	rt_flag = (flags & XFS_DIFLAG_REALTIME);
746 	hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
747 	inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
748 	extsize_bytes = XFS_FSB_TO_B(mp, extsize);
749 
750 	if (rt_flag)
751 		blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
752 	else
753 		blocksize_bytes = mp->m_sb.sb_blocksize;
754 
755 	if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
756 		return __this_address;
757 
758 	if (hint_flag && !S_ISREG(mode))
759 		return __this_address;
760 
761 	if (inherit_flag && !S_ISDIR(mode))
762 		return __this_address;
763 
764 	if ((hint_flag || inherit_flag) && extsize == 0)
765 		return __this_address;
766 
767 	/* free inodes get flags set to zero but extsize remains */
768 	if (mode && !(hint_flag || inherit_flag) && extsize != 0)
769 		return __this_address;
770 
771 	if (extsize_bytes % blocksize_bytes)
772 		return __this_address;
773 
774 	if (extsize > MAXEXTLEN)
775 		return __this_address;
776 
777 	if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
778 		return __this_address;
779 
780 	return NULL;
781 }
782 
783 /*
784  * Validate di_cowextsize hint.
785  *
786  * The rules are documented at xfs_ioctl_setattr_check_cowextsize().
787  * These functions must be kept in sync with each other.
788  */
789 xfs_failaddr_t
790 xfs_inode_validate_cowextsize(
791 	struct xfs_mount		*mp,
792 	uint32_t			cowextsize,
793 	uint16_t			mode,
794 	uint16_t			flags,
795 	uint64_t			flags2)
796 {
797 	bool				rt_flag;
798 	bool				hint_flag;
799 	uint32_t			cowextsize_bytes;
800 
801 	rt_flag = (flags & XFS_DIFLAG_REALTIME);
802 	hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
803 	cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
804 
805 	if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb))
806 		return __this_address;
807 
808 	if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
809 		return __this_address;
810 
811 	if (hint_flag && cowextsize == 0)
812 		return __this_address;
813 
814 	/* free inodes get flags set to zero but cowextsize remains */
815 	if (mode && !hint_flag && cowextsize != 0)
816 		return __this_address;
817 
818 	if (hint_flag && rt_flag)
819 		return __this_address;
820 
821 	if (cowextsize_bytes % mp->m_sb.sb_blocksize)
822 		return __this_address;
823 
824 	if (cowextsize > MAXEXTLEN)
825 		return __this_address;
826 
827 	if (cowextsize > mp->m_sb.sb_agblocks / 2)
828 		return __this_address;
829 
830 	return NULL;
831 }
832