Linux Audio

Check our new training course

Loading...
v4.17
 
  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}
v5.14.15
  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 */
 38static void
 39xfs_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_ADDR(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->m_sb, 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)bp->b_bn, 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
 86static void
 87xfs_inode_buf_read_verify(
 88	struct xfs_buf	*bp)
 89{
 90	xfs_inode_buf_verify(bp, false);
 91}
 92
 93static void
 94xfs_inode_buf_readahead_verify(
 95	struct xfs_buf	*bp)
 96{
 97	xfs_inode_buf_verify(bp, true);
 98}
 99
100static void
101xfs_inode_buf_write_verify(
102	struct xfs_buf	*bp)
103{
104	xfs_inode_buf_verify(bp, false);
105}
106
107const 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
115const 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 */
129int
130xfs_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
141static 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. */
153struct timespec64
154xfs_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
171int
172xfs_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_sb_version_has_v3inode(&ip->i_mount->m_sb))
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_sb_version_has_v3inode(&ip->i_mount->m_sb)) {
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
258out_destroy_data_fork:
259	xfs_idestroy_fork(&ip->i_df);
260	return error;
261}
262
263/* Convert an incore timestamp to an ondisk timestamp. */
264static inline xfs_timestamp_t
265xfs_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
282void
283xfs_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_sb_version_has_v3inode(&ip->i_mount->m_sb)) {
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
333static xfs_failaddr_t
334xfs_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
341	switch (XFS_DFORK_FORMAT(dip, whichfork)) {
342	case XFS_DINODE_FMT_LOCAL:
343		/*
344		 * no local regular files yet
345		 */
346		if (whichfork == XFS_DATA_FORK) {
347			if (S_ISREG(be16_to_cpu(dip->di_mode)))
348				return __this_address;
349			if (be64_to_cpu(dip->di_size) >
350					XFS_DFORK_SIZE(dip, mp, whichfork))
351				return __this_address;
352		}
353		if (di_nextents)
354			return __this_address;
355		break;
356	case XFS_DINODE_FMT_EXTENTS:
357		if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
358			return __this_address;
359		break;
360	case XFS_DINODE_FMT_BTREE:
361		if (whichfork == XFS_ATTR_FORK) {
362			if (di_nextents > MAXAEXTNUM)
363				return __this_address;
364		} else if (di_nextents > MAXEXTNUM) {
365			return __this_address;
366		}
367		break;
368	default:
369		return __this_address;
370	}
371	return NULL;
372}
373
374static xfs_failaddr_t
375xfs_dinode_verify_forkoff(
376	struct xfs_dinode	*dip,
377	struct xfs_mount	*mp)
378{
379	if (!dip->di_forkoff)
380		return NULL;
381
382	switch (dip->di_format)  {
383	case XFS_DINODE_FMT_DEV:
384		if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
385			return __this_address;
386		break;
387	case XFS_DINODE_FMT_LOCAL:	/* fall through ... */
388	case XFS_DINODE_FMT_EXTENTS:    /* fall through ... */
389	case XFS_DINODE_FMT_BTREE:
390		if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3))
391			return __this_address;
392		break;
393	default:
394		return __this_address;
395	}
396	return NULL;
397}
398
399xfs_failaddr_t
400xfs_dinode_verify(
401	struct xfs_mount	*mp,
402	xfs_ino_t		ino,
403	struct xfs_dinode	*dip)
404{
405	xfs_failaddr_t		fa;
406	uint16_t		mode;
407	uint16_t		flags;
408	uint64_t		flags2;
409	uint64_t		di_size;
410
411	if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
412		return __this_address;
413
414	/* Verify v3 integrity information first */
415	if (dip->di_version >= 3) {
416		if (!xfs_sb_version_has_v3inode(&mp->m_sb))
417			return __this_address;
418		if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
419				      XFS_DINODE_CRC_OFF))
420			return __this_address;
421		if (be64_to_cpu(dip->di_ino) != ino)
422			return __this_address;
423		if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
424			return __this_address;
425	}
426
427	/* don't allow invalid i_size */
428	di_size = be64_to_cpu(dip->di_size);
429	if (di_size & (1ULL << 63))
430		return __this_address;
431
432	mode = be16_to_cpu(dip->di_mode);
433	if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
434		return __this_address;
435
436	/* No zero-length symlinks/dirs. */
437	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
438		return __this_address;
439
440	/* Fork checks carried over from xfs_iformat_fork */
441	if (mode &&
442	    be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
443			be64_to_cpu(dip->di_nblocks))
444		return __this_address;
445
446	if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
447		return __this_address;
448
449	flags = be16_to_cpu(dip->di_flags);
450
451	if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
452		return __this_address;
453
454	/* check for illegal values of forkoff */
455	fa = xfs_dinode_verify_forkoff(dip, mp);
456	if (fa)
457		return fa;
458
459	/* Do we have appropriate data fork formats for the mode? */
460	switch (mode & S_IFMT) {
461	case S_IFIFO:
462	case S_IFCHR:
463	case S_IFBLK:
464	case S_IFSOCK:
465		if (dip->di_format != XFS_DINODE_FMT_DEV)
466			return __this_address;
467		break;
468	case S_IFREG:
469	case S_IFLNK:
470	case S_IFDIR:
471		fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
472		if (fa)
473			return fa;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
474		break;
475	case 0:
476		/* Uninitialized inode ok. */
477		break;
478	default:
479		return __this_address;
480	}
481
482	if (dip->di_forkoff) {
483		fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
484		if (fa)
485			return fa;
 
 
 
 
 
 
 
 
486	} else {
487		/*
488		 * If there is no fork offset, this may be a freshly-made inode
489		 * in a new disk cluster, in which case di_aformat is zeroed.
490		 * Otherwise, such an inode must be in EXTENTS format; this goes
491		 * for freed inodes as well.
492		 */
493		switch (dip->di_aformat) {
494		case 0:
495		case XFS_DINODE_FMT_EXTENTS:
496			break;
497		default:
498			return __this_address;
499		}
500		if (dip->di_anextents)
501			return __this_address;
502	}
503
504	/* extent size hint validation */
505	fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
506			mode, flags);
507	if (fa)
508		return fa;
509
510	/* only version 3 or greater inodes are extensively verified here */
511	if (dip->di_version < 3)
512		return NULL;
513
514	flags2 = be64_to_cpu(dip->di_flags2);
515
516	/* don't allow reflink/cowextsize if we don't have reflink */
517	if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
518	     !xfs_sb_version_hasreflink(&mp->m_sb))
519		return __this_address;
520
521	/* only regular files get reflink */
522	if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
523		return __this_address;
524
525	/* don't let reflink and realtime mix */
526	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
527		return __this_address;
528
529	/* COW extent size hint validation */
530	fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
531			mode, flags, flags2);
532	if (fa)
533		return fa;
534
535	/* bigtime iflag can only happen on bigtime filesystems */
536	if (xfs_dinode_has_bigtime(dip) &&
537	    !xfs_sb_version_hasbigtime(&mp->m_sb))
538		return __this_address;
539
540	return NULL;
541}
542
543void
544xfs_dinode_calc_crc(
545	struct xfs_mount	*mp,
546	struct xfs_dinode	*dip)
547{
548	uint32_t		crc;
549
550	if (dip->di_version < 3)
551		return;
552
553	ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
554	crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
555			      XFS_DINODE_CRC_OFF);
556	dip->di_crc = xfs_end_cksum(crc);
557}
558
559/*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
560 * Validate di_extsize hint.
561 *
562 * 1. Extent size hint is only valid for directories and regular files.
563 * 2. FS_XFLAG_EXTSIZE is only valid for regular files.
564 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
565 * 4. Hint cannot be larger than MAXTEXTLEN.
566 * 5. Can be changed on directories at any time.
567 * 6. Hint value of 0 turns off hints, clears inode flags.
568 * 7. Extent size must be a multiple of the appropriate block size.
569 *    For realtime files, this is the rt extent size.
570 * 8. For non-realtime files, the extent size hint must be limited
571 *    to half the AG size to avoid alignment extending the extent beyond the
572 *    limits of the AG.
573 */
574xfs_failaddr_t
575xfs_inode_validate_extsize(
576	struct xfs_mount		*mp,
577	uint32_t			extsize,
578	uint16_t			mode,
579	uint16_t			flags)
580{
581	bool				rt_flag;
582	bool				hint_flag;
583	bool				inherit_flag;
584	uint32_t			extsize_bytes;
585	uint32_t			blocksize_bytes;
586
587	rt_flag = (flags & XFS_DIFLAG_REALTIME);
588	hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
589	inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
590	extsize_bytes = XFS_FSB_TO_B(mp, extsize);
591
592	/*
593	 * This comment describes a historic gap in this verifier function.
594	 *
595	 * For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
596	 * function has never checked that the extent size hint is an integer
597	 * multiple of the realtime extent size.  Since we allow users to set
598	 * this combination  on non-rt filesystems /and/ to change the rt
599	 * extent size when adding a rt device to a filesystem, the net effect
600	 * is that users can configure a filesystem anticipating one rt
601	 * geometry and change their minds later.  Directories do not use the
602	 * extent size hint, so this is harmless for them.
603	 *
604	 * If a directory with a misaligned extent size hint is allowed to
605	 * propagate that hint into a new regular realtime file, the result
606	 * is that the inode cluster buffer verifier will trigger a corruption
607	 * shutdown the next time it is run, because the verifier has always
608	 * enforced the alignment rule for regular files.
609	 *
610	 * Because we allow administrators to set a new rt extent size when
611	 * adding a rt section, we cannot add a check to this verifier because
612	 * that will result a new source of directory corruption errors when
613	 * reading an existing filesystem.  Instead, we rely on callers to
614	 * decide when alignment checks are appropriate, and fix things up as
615	 * needed.
616	 */
617
618	if (rt_flag)
619		blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
620	else
621		blocksize_bytes = mp->m_sb.sb_blocksize;
622
623	if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
624		return __this_address;
625
626	if (hint_flag && !S_ISREG(mode))
627		return __this_address;
628
629	if (inherit_flag && !S_ISDIR(mode))
630		return __this_address;
631
632	if ((hint_flag || inherit_flag) && extsize == 0)
633		return __this_address;
634
635	/* free inodes get flags set to zero but extsize remains */
636	if (mode && !(hint_flag || inherit_flag) && extsize != 0)
637		return __this_address;
638
639	if (extsize_bytes % blocksize_bytes)
640		return __this_address;
641
642	if (extsize > MAXEXTLEN)
643		return __this_address;
644
645	if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
646		return __this_address;
647
648	return NULL;
649}
650
651/*
652 * Validate di_cowextsize hint.
653 *
654 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
655 *    The inode does not have to have any shared blocks, but it must be a v3.
656 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
657 *    for a directory, the hint is propagated to new files.
658 * 3. Can be changed on files & directories at any time.
659 * 4. Hint value of 0 turns off hints, clears inode flags.
660 * 5. Extent size must be a multiple of the appropriate block size.
661 * 6. The extent size hint must be limited to half the AG size to avoid
662 *    alignment extending the extent beyond the limits of the AG.
663 */
664xfs_failaddr_t
665xfs_inode_validate_cowextsize(
666	struct xfs_mount		*mp,
667	uint32_t			cowextsize,
668	uint16_t			mode,
669	uint16_t			flags,
670	uint64_t			flags2)
671{
672	bool				rt_flag;
673	bool				hint_flag;
674	uint32_t			cowextsize_bytes;
675
676	rt_flag = (flags & XFS_DIFLAG_REALTIME);
677	hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
678	cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
679
680	if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb))
681		return __this_address;
682
683	if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
684		return __this_address;
685
686	if (hint_flag && cowextsize == 0)
687		return __this_address;
688
689	/* free inodes get flags set to zero but cowextsize remains */
690	if (mode && !hint_flag && cowextsize != 0)
691		return __this_address;
692
693	if (hint_flag && rt_flag)
694		return __this_address;
695
696	if (cowextsize_bytes % mp->m_sb.sb_blocksize)
697		return __this_address;
698
699	if (cowextsize > MAXEXTLEN)
700		return __this_address;
701
702	if (cowextsize > mp->m_sb.sb_agblocks / 2)
703		return __this_address;
704
705	return NULL;
706}