1// SPDX-License-Identifier: GPL-2.0
  2/*
  3 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
  4 * All Rights Reserved.
  5 */
  6#include "xfs.h"
  7#include "xfs_fs.h"
  8#include "xfs_shared.h"
  9#include "xfs_format.h"
 10#include "xfs_log_format.h"
 11#include "xfs_trans_resv.h"
 12#include "xfs_mount.h"
 13#include "xfs_ag.h"
 14#include "xfs_inode.h"
 15#include "xfs_errortag.h"
 16#include "xfs_error.h"
 17#include "xfs_icache.h"
 18#include "xfs_trans.h"
 19#include "xfs_ialloc.h"
 20#include "xfs_dir2.h"
 21
 22#include <linux/iversion.h>
 23
 24/*
 25 * If we are doing readahead on an inode buffer, we might be in log recovery
 26 * reading an inode allocation buffer that hasn't yet been replayed, and hence
 27 * has not had the inode cores stamped into it. Hence for readahead, the buffer
 28 * may be potentially invalid.
 29 *
 30 * If the readahead buffer is invalid, we need to mark it with an error and
 31 * clear the DONE status of the buffer so that a followup read will re-read it
 32 * from disk. We don't report the error otherwise to avoid warnings during log
 33 * recovery and we don't get unnecessary panics on debug kernels. We use EIO here
 34 * because all we want to do is say readahead failed; there is no-one to report
 35 * the error to, so this will distinguish it from a non-ra verifier failure.
 36 * Changes to this readahead error behaviour also need to be reflected in
 37 * xfs_dquot_buf_readahead_verify().
 38 */
 39static void
 40xfs_inode_buf_verify(
 41	struct xfs_buf	*bp,
 42	bool		readahead)
 43{
 44	struct xfs_mount *mp = bp->b_mount;
 
 45	int		i;
 46	int		ni;
 47
 48	/*
 49	 * Validate the magic number and version of every inode in the buffer
 50	 */
 
 51	ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
 52	for (i = 0; i < ni; i++) {
 53		struct xfs_dinode	*dip;
 54		xfs_agino_t		unlinked_ino;
 55		int			di_ok;
 56
 57		dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
 58		unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
 59		di_ok = xfs_verify_magic16(bp, dip->di_magic) &&
 60			xfs_dinode_good_version(mp, dip->di_version) &&
 61			xfs_verify_agino_or_null(bp->b_pag, unlinked_ino);
 62		if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
 63						XFS_ERRTAG_ITOBP_INOTOBP))) {
 64			if (readahead) {
 65				bp->b_flags &= ~XBF_DONE;
 66				xfs_buf_ioerror(bp, -EIO);
 67				return;
 68			}
 69
 70#ifdef DEBUG
 71			xfs_alert(mp,
 72				"bad inode magic/vsn daddr %lld #%d (magic=%x)",
 73				(unsigned long long)xfs_buf_daddr(bp), i,
 74				be16_to_cpu(dip->di_magic));
 75#endif
 76			xfs_buf_verifier_error(bp, -EFSCORRUPTED,
 77					__func__, dip, sizeof(*dip),
 78					NULL);
 79			return;
 80		}
 81	}
 82}
 83
 84
 85static void
 86xfs_inode_buf_read_verify(
 87	struct xfs_buf	*bp)
 88{
 89	xfs_inode_buf_verify(bp, false);
 90}
 91
 92static void
 93xfs_inode_buf_readahead_verify(
 94	struct xfs_buf	*bp)
 95{
 96	xfs_inode_buf_verify(bp, true);
 97}
 98
 99static void
100xfs_inode_buf_write_verify(
101	struct xfs_buf	*bp)
102{
103	xfs_inode_buf_verify(bp, false);
104}
105
106const struct xfs_buf_ops xfs_inode_buf_ops = {
107	.name = "xfs_inode",
108	.magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
109		     cpu_to_be16(XFS_DINODE_MAGIC) },
110	.verify_read = xfs_inode_buf_read_verify,
111	.verify_write = xfs_inode_buf_write_verify,
112};
113
114const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
115	.name = "xfs_inode_ra",
116	.magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
117		     cpu_to_be16(XFS_DINODE_MAGIC) },
118	.verify_read = xfs_inode_buf_readahead_verify,
119	.verify_write = xfs_inode_buf_write_verify,
120};
121
122
123/*
124 * This routine is called to map an inode to the buffer containing the on-disk
125 * version of the inode.  It returns a pointer to the buffer containing the
126 * on-disk inode in the bpp parameter.
 
 
 
 
127 */
128int
129xfs_imap_to_bp(
130	struct xfs_mount	*mp,
131	struct xfs_trans	*tp,
132	struct xfs_imap		*imap,
133	struct xfs_buf		**bpp)
134{
135	return xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
136				   imap->im_len, XBF_UNMAPPED, bpp,
137				   &xfs_inode_buf_ops);
138}
139
140static inline struct timespec64 xfs_inode_decode_bigtime(uint64_t ts)
141{
142	struct timespec64	tv;
143	uint32_t		n;
144
145	tv.tv_sec = xfs_bigtime_to_unix(div_u64_rem(ts, NSEC_PER_SEC, &n));
146	tv.tv_nsec = n;
147
148	return tv;
149}
150
151/* Convert an ondisk timestamp to an incore timestamp. */
152struct timespec64
153xfs_inode_from_disk_ts(
154	struct xfs_dinode		*dip,
155	const xfs_timestamp_t		ts)
156{
157	struct timespec64		tv;
158	struct xfs_legacy_timestamp	*lts;
159
160	if (xfs_dinode_has_bigtime(dip))
161		return xfs_inode_decode_bigtime(be64_to_cpu(ts));
162
163	lts = (struct xfs_legacy_timestamp *)&ts;
164	tv.tv_sec = (int)be32_to_cpu(lts->t_sec);
165	tv.tv_nsec = (int)be32_to_cpu(lts->t_nsec);
 
 
 
 
 
166
167	return tv;
 
 
 
168}
169
170int
171xfs_inode_from_disk(
172	struct xfs_inode	*ip,
173	struct xfs_dinode	*from)
174{
 
175	struct inode		*inode = VFS_I(ip);
176	int			error;
177	xfs_failaddr_t		fa;
178
179	ASSERT(ip->i_cowfp == NULL);
 
180
181	fa = xfs_dinode_verify(ip->i_mount, ip->i_ino, from);
182	if (fa) {
183		xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", from,
184				sizeof(*from), fa);
185		return -EFSCORRUPTED;
186	}
187
188	/*
189	 * First get the permanent information that is needed to allocate an
190	 * inode. If the inode is unused, mode is zero and we shouldn't mess
191	 * with the uninitialized part of it.
192	 */
193	if (!xfs_has_v3inodes(ip->i_mount))
194		ip->i_flushiter = be16_to_cpu(from->di_flushiter);
195	inode->i_generation = be32_to_cpu(from->di_gen);
196	inode->i_mode = be16_to_cpu(from->di_mode);
197	if (!inode->i_mode)
198		return 0;
199
200	/*
201	 * Convert v1 inodes immediately to v2 inode format as this is the
202	 * minimum inode version format we support in the rest of the code.
203	 * They will also be unconditionally written back to disk as v2 inodes.
204	 */
205	if (unlikely(from->di_version == 1)) {
206		set_nlink(inode, be16_to_cpu(from->di_onlink));
207		ip->i_projid = 0;
208	} else {
209		set_nlink(inode, be32_to_cpu(from->di_nlink));
210		ip->i_projid = (prid_t)be16_to_cpu(from->di_projid_hi) << 16 |
211					be16_to_cpu(from->di_projid_lo);
212	}
213
214	i_uid_write(inode, be32_to_cpu(from->di_uid));
215	i_gid_write(inode, be32_to_cpu(from->di_gid));
216
217	/*
218	 * Time is signed, so need to convert to signed 32 bit before
219	 * storing in inode timestamp which may be 64 bit. Otherwise
220	 * a time before epoch is converted to a time long after epoch
221	 * on 64 bit systems.
222	 */
223	inode_set_atime_to_ts(inode,
224			      xfs_inode_from_disk_ts(from, from->di_atime));
225	inode_set_mtime_to_ts(inode,
226			      xfs_inode_from_disk_ts(from, from->di_mtime));
227	inode_set_ctime_to_ts(inode,
228			      xfs_inode_from_disk_ts(from, from->di_ctime));
229
230	ip->i_disk_size = be64_to_cpu(from->di_size);
231	ip->i_nblocks = be64_to_cpu(from->di_nblocks);
232	ip->i_extsize = be32_to_cpu(from->di_extsize);
233	ip->i_forkoff = from->di_forkoff;
234	ip->i_diflags = be16_to_cpu(from->di_flags);
235	ip->i_next_unlinked = be32_to_cpu(from->di_next_unlinked);
 
236
237	if (from->di_dmevmask || from->di_dmstate)
238		xfs_iflags_set(ip, XFS_IPRESERVE_DM_FIELDS);
239
240	if (xfs_has_v3inodes(ip->i_mount)) {
241		inode_set_iversion_queried(inode,
242					   be64_to_cpu(from->di_changecount));
243		ip->i_crtime = xfs_inode_from_disk_ts(from, from->di_crtime);
244		ip->i_diflags2 = be64_to_cpu(from->di_flags2);
245		ip->i_cowextsize = be32_to_cpu(from->di_cowextsize);
 
246	}
247
248	error = xfs_iformat_data_fork(ip, from);
249	if (error)
250		return error;
251	if (from->di_forkoff) {
252		error = xfs_iformat_attr_fork(ip, from);
253		if (error)
254			goto out_destroy_data_fork;
255	}
256	if (xfs_is_reflink_inode(ip))
257		xfs_ifork_init_cow(ip);
258	return 0;
259
260out_destroy_data_fork:
261	xfs_idestroy_fork(&ip->i_df);
262	return error;
263}
264
265/* Convert an incore timestamp to an ondisk timestamp. */
266static inline xfs_timestamp_t
267xfs_inode_to_disk_ts(
268	struct xfs_inode		*ip,
269	const struct timespec64		tv)
270{
271	struct xfs_legacy_timestamp	*lts;
272	xfs_timestamp_t			ts;
273
274	if (xfs_inode_has_bigtime(ip))
275		return cpu_to_be64(xfs_inode_encode_bigtime(tv));
276
277	lts = (struct xfs_legacy_timestamp *)&ts;
278	lts->t_sec = cpu_to_be32(tv.tv_sec);
279	lts->t_nsec = cpu_to_be32(tv.tv_nsec);
280
281	return ts;
282}
283
284static inline void
285xfs_inode_to_disk_iext_counters(
286	struct xfs_inode	*ip,
287	struct xfs_dinode	*to)
288{
289	if (xfs_inode_has_large_extent_counts(ip)) {
290		to->di_big_nextents = cpu_to_be64(xfs_ifork_nextents(&ip->i_df));
291		to->di_big_anextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_af));
292		/*
293		 * We might be upgrading the inode to use larger extent counters
294		 * than was previously used. Hence zero the unused field.
295		 */
296		to->di_nrext64_pad = cpu_to_be16(0);
297	} else {
298		to->di_nextents = cpu_to_be32(xfs_ifork_nextents(&ip->i_df));
299		to->di_anextents = cpu_to_be16(xfs_ifork_nextents(&ip->i_af));
300	}
301}
302
303void
304xfs_inode_to_disk(
305	struct xfs_inode	*ip,
306	struct xfs_dinode	*to,
307	xfs_lsn_t		lsn)
308{
 
309	struct inode		*inode = VFS_I(ip);
310
311	to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
312	to->di_onlink = 0;
313
314	to->di_format = xfs_ifork_format(&ip->i_df);
315	to->di_uid = cpu_to_be32(i_uid_read(inode));
316	to->di_gid = cpu_to_be32(i_gid_read(inode));
317	to->di_projid_lo = cpu_to_be16(ip->i_projid & 0xffff);
318	to->di_projid_hi = cpu_to_be16(ip->i_projid >> 16);
319
320	to->di_atime = xfs_inode_to_disk_ts(ip, inode_get_atime(inode));
321	to->di_mtime = xfs_inode_to_disk_ts(ip, inode_get_mtime(inode));
322	to->di_ctime = xfs_inode_to_disk_ts(ip, inode_get_ctime(inode));
 
 
 
 
323	to->di_nlink = cpu_to_be32(inode->i_nlink);
324	to->di_gen = cpu_to_be32(inode->i_generation);
325	to->di_mode = cpu_to_be16(inode->i_mode);
326
327	to->di_size = cpu_to_be64(ip->i_disk_size);
328	to->di_nblocks = cpu_to_be64(ip->i_nblocks);
329	to->di_extsize = cpu_to_be32(ip->i_extsize);
330	to->di_forkoff = ip->i_forkoff;
331	to->di_aformat = xfs_ifork_format(&ip->i_af);
332	to->di_flags = cpu_to_be16(ip->i_diflags);
 
 
 
 
333
334	if (xfs_has_v3inodes(ip->i_mount)) {
335		to->di_version = 3;
336		to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
337		to->di_crtime = xfs_inode_to_disk_ts(ip, ip->i_crtime);
338		to->di_flags2 = cpu_to_be64(ip->i_diflags2);
339		to->di_cowextsize = cpu_to_be32(ip->i_cowextsize);
 
340		to->di_ino = cpu_to_be64(ip->i_ino);
341		to->di_lsn = cpu_to_be64(lsn);
342		memset(to->di_pad2, 0, sizeof(to->di_pad2));
343		uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
344		to->di_v3_pad = 0;
345	} else {
346		to->di_version = 2;
347		to->di_flushiter = cpu_to_be16(ip->i_flushiter);
348		memset(to->di_v2_pad, 0, sizeof(to->di_v2_pad));
349	}
 
350
351	xfs_inode_to_disk_iext_counters(ip, to);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
352}
353
354static xfs_failaddr_t
355xfs_dinode_verify_fork(
356	struct xfs_dinode	*dip,
357	struct xfs_mount	*mp,
358	int			whichfork)
359{
360	xfs_extnum_t		di_nextents;
361	xfs_extnum_t		max_extents;
362	mode_t			mode = be16_to_cpu(dip->di_mode);
363	uint32_t		fork_size = XFS_DFORK_SIZE(dip, mp, whichfork);
364	uint32_t		fork_format = XFS_DFORK_FORMAT(dip, whichfork);
365
366	di_nextents = xfs_dfork_nextents(dip, whichfork);
367
368	/*
369	 * For fork types that can contain local data, check that the fork
370	 * format matches the size of local data contained within the fork.
371	 *
372	 * For all types, check that when the size says the should be in extent
373	 * or btree format, the inode isn't claiming it is in local format.
374	 */
375	if (whichfork == XFS_DATA_FORK) {
376		if (S_ISDIR(mode) || S_ISLNK(mode)) {
377			if (be64_to_cpu(dip->di_size) <= fork_size &&
378			    fork_format != XFS_DINODE_FMT_LOCAL)
379				return __this_address;
380		}
381
382		if (be64_to_cpu(dip->di_size) > fork_size &&
383		    fork_format == XFS_DINODE_FMT_LOCAL)
384			return __this_address;
385	}
386
387	switch (fork_format) {
388	case XFS_DINODE_FMT_LOCAL:
389		/*
390		 * No local regular files yet.
391		 */
392		if (S_ISREG(mode) && whichfork == XFS_DATA_FORK)
393			return __this_address;
 
 
 
 
 
394		if (di_nextents)
395			return __this_address;
396		break;
397	case XFS_DINODE_FMT_EXTENTS:
398		if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
399			return __this_address;
400		break;
401	case XFS_DINODE_FMT_BTREE:
402		max_extents = xfs_iext_max_nextents(
403					xfs_dinode_has_large_extent_counts(dip),
404					whichfork);
405		if (di_nextents > max_extents)
406			return __this_address;
 
407		break;
408	default:
409		return __this_address;
410	}
411	return NULL;
412}
413
414static xfs_failaddr_t
415xfs_dinode_verify_forkoff(
416	struct xfs_dinode	*dip,
417	struct xfs_mount	*mp)
418{
419	if (!dip->di_forkoff)
420		return NULL;
421
422	switch (dip->di_format)  {
423	case XFS_DINODE_FMT_DEV:
424		if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
425			return __this_address;
426		break;
427	case XFS_DINODE_FMT_LOCAL:	/* fall through ... */
428	case XFS_DINODE_FMT_EXTENTS:    /* fall through ... */
429	case XFS_DINODE_FMT_BTREE:
430		if (dip->di_forkoff >= (XFS_LITINO(mp) >> 3))
431			return __this_address;
432		break;
433	default:
434		return __this_address;
435	}
436	return NULL;
437}
438
439static xfs_failaddr_t
440xfs_dinode_verify_nrext64(
441	struct xfs_mount	*mp,
442	struct xfs_dinode	*dip)
443{
444	if (xfs_dinode_has_large_extent_counts(dip)) {
445		if (!xfs_has_large_extent_counts(mp))
446			return __this_address;
447		if (dip->di_nrext64_pad != 0)
448			return __this_address;
449	} else if (dip->di_version >= 3) {
450		if (dip->di_v3_pad != 0)
451			return __this_address;
452	}
453
454	return NULL;
455}
456
457xfs_failaddr_t
458xfs_dinode_verify(
459	struct xfs_mount	*mp,
460	xfs_ino_t		ino,
461	struct xfs_dinode	*dip)
462{
463	xfs_failaddr_t		fa;
464	uint16_t		mode;
465	uint16_t		flags;
466	uint64_t		flags2;
467	uint64_t		di_size;
468	xfs_extnum_t		nextents;
469	xfs_extnum_t		naextents;
470	xfs_filblks_t		nblocks;
471
472	if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
473		return __this_address;
474
475	/* Verify v3 integrity information first */
476	if (dip->di_version >= 3) {
477		if (!xfs_has_v3inodes(mp))
478			return __this_address;
479		if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
480				      XFS_DINODE_CRC_OFF))
481			return __this_address;
482		if (be64_to_cpu(dip->di_ino) != ino)
483			return __this_address;
484		if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
485			return __this_address;
486	}
487
488	/* don't allow invalid i_size */
489	di_size = be64_to_cpu(dip->di_size);
490	if (di_size & (1ULL << 63))
491		return __this_address;
492
493	mode = be16_to_cpu(dip->di_mode);
494	if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
495		return __this_address;
496
497	/* No zero-length symlinks/dirs. */
498	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
499		return __this_address;
500
501	fa = xfs_dinode_verify_nrext64(mp, dip);
502	if (fa)
503		return fa;
504
505	nextents = xfs_dfork_data_extents(dip);
506	naextents = xfs_dfork_attr_extents(dip);
507	nblocks = be64_to_cpu(dip->di_nblocks);
508
509	/* Fork checks carried over from xfs_iformat_fork */
510	if (mode && nextents + naextents > nblocks)
511		return __this_address;
512
513	if (nextents + naextents == 0 && nblocks != 0)
514		return __this_address;
515
516	if (S_ISDIR(mode) && nextents > mp->m_dir_geo->max_extents)
517		return __this_address;
518
519	if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
520		return __this_address;
521
522	flags = be16_to_cpu(dip->di_flags);
523
524	if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
525		return __this_address;
526
527	/* check for illegal values of forkoff */
528	fa = xfs_dinode_verify_forkoff(dip, mp);
529	if (fa)
530		return fa;
531
532	/* Do we have appropriate data fork formats for the mode? */
533	switch (mode & S_IFMT) {
534	case S_IFIFO:
535	case S_IFCHR:
536	case S_IFBLK:
537	case S_IFSOCK:
538		if (dip->di_format != XFS_DINODE_FMT_DEV)
539			return __this_address;
540		break;
541	case S_IFREG:
542	case S_IFLNK:
543	case S_IFDIR:
544		fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
545		if (fa)
546			return fa;
547		break;
548	case 0:
549		/* Uninitialized inode ok. */
550		break;
551	default:
552		return __this_address;
553	}
554
555	if (dip->di_forkoff) {
556		fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
557		if (fa)
558			return fa;
559	} else {
560		/*
561		 * If there is no fork offset, this may be a freshly-made inode
562		 * in a new disk cluster, in which case di_aformat is zeroed.
563		 * Otherwise, such an inode must be in EXTENTS format; this goes
564		 * for freed inodes as well.
565		 */
566		switch (dip->di_aformat) {
567		case 0:
568		case XFS_DINODE_FMT_EXTENTS:
569			break;
570		default:
571			return __this_address;
572		}
573		if (naextents)
574			return __this_address;
575	}
576
577	/* extent size hint validation */
578	fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
579			mode, flags);
580	if (fa)
581		return fa;
582
583	/* only version 3 or greater inodes are extensively verified here */
584	if (dip->di_version < 3)
585		return NULL;
586
587	flags2 = be64_to_cpu(dip->di_flags2);
588
589	/* don't allow reflink/cowextsize if we don't have reflink */
590	if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
591	     !xfs_has_reflink(mp))
592		return __this_address;
593
594	/* only regular files get reflink */
595	if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
596		return __this_address;
597
598	/* don't let reflink and realtime mix */
599	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
600		return __this_address;
601
 
 
 
 
602	/* COW extent size hint validation */
603	fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
604			mode, flags, flags2);
605	if (fa)
606		return fa;
607
608	/* bigtime iflag can only happen on bigtime filesystems */
609	if (xfs_dinode_has_bigtime(dip) &&
610	    !xfs_has_bigtime(mp))
611		return __this_address;
612
613	return NULL;
614}
615
616void
617xfs_dinode_calc_crc(
618	struct xfs_mount	*mp,
619	struct xfs_dinode	*dip)
620{
621	uint32_t		crc;
622
623	if (dip->di_version < 3)
624		return;
625
626	ASSERT(xfs_has_crc(mp));
627	crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
628			      XFS_DINODE_CRC_OFF);
629	dip->di_crc = xfs_end_cksum(crc);
630}
631
632/*
633 * Validate di_extsize hint.
634 *
635 * 1. Extent size hint is only valid for directories and regular files.
636 * 2. FS_XFLAG_EXTSIZE is only valid for regular files.
637 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
638 * 4. Hint cannot be larger than MAXTEXTLEN.
639 * 5. Can be changed on directories at any time.
640 * 6. Hint value of 0 turns off hints, clears inode flags.
641 * 7. Extent size must be a multiple of the appropriate block size.
642 *    For realtime files, this is the rt extent size.
643 * 8. For non-realtime files, the extent size hint must be limited
644 *    to half the AG size to avoid alignment extending the extent beyond the
645 *    limits of the AG.
646 */
647xfs_failaddr_t
648xfs_inode_validate_extsize(
649	struct xfs_mount		*mp,
650	uint32_t			extsize,
651	uint16_t			mode,
652	uint16_t			flags)
653{
654	bool				rt_flag;
655	bool				hint_flag;
656	bool				inherit_flag;
657	uint32_t			extsize_bytes;
658	uint32_t			blocksize_bytes;
659
660	rt_flag = (flags & XFS_DIFLAG_REALTIME);
661	hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
662	inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
663	extsize_bytes = XFS_FSB_TO_B(mp, extsize);
664
665	/*
666	 * This comment describes a historic gap in this verifier function.
667	 *
668	 * For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
669	 * function has never checked that the extent size hint is an integer
670	 * multiple of the realtime extent size.  Since we allow users to set
671	 * this combination  on non-rt filesystems /and/ to change the rt
672	 * extent size when adding a rt device to a filesystem, the net effect
673	 * is that users can configure a filesystem anticipating one rt
674	 * geometry and change their minds later.  Directories do not use the
675	 * extent size hint, so this is harmless for them.
676	 *
677	 * If a directory with a misaligned extent size hint is allowed to
678	 * propagate that hint into a new regular realtime file, the result
679	 * is that the inode cluster buffer verifier will trigger a corruption
680	 * shutdown the next time it is run, because the verifier has always
681	 * enforced the alignment rule for regular files.
682	 *
683	 * Because we allow administrators to set a new rt extent size when
684	 * adding a rt section, we cannot add a check to this verifier because
685	 * that will result a new source of directory corruption errors when
686	 * reading an existing filesystem.  Instead, we rely on callers to
687	 * decide when alignment checks are appropriate, and fix things up as
688	 * needed.
689	 */
690
691	if (rt_flag)
692		blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
693	else
694		blocksize_bytes = mp->m_sb.sb_blocksize;
695
696	if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
697		return __this_address;
698
699	if (hint_flag && !S_ISREG(mode))
700		return __this_address;
701
702	if (inherit_flag && !S_ISDIR(mode))
703		return __this_address;
704
705	if ((hint_flag || inherit_flag) && extsize == 0)
706		return __this_address;
707
708	/* free inodes get flags set to zero but extsize remains */
709	if (mode && !(hint_flag || inherit_flag) && extsize != 0)
710		return __this_address;
711
712	if (extsize_bytes % blocksize_bytes)
713		return __this_address;
714
715	if (extsize > XFS_MAX_BMBT_EXTLEN)
716		return __this_address;
717
718	if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
719		return __this_address;
720
721	return NULL;
722}
723
724/*
725 * Validate di_cowextsize hint.
726 *
727 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
728 *    The inode does not have to have any shared blocks, but it must be a v3.
729 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
730 *    for a directory, the hint is propagated to new files.
731 * 3. Can be changed on files & directories at any time.
732 * 4. Hint value of 0 turns off hints, clears inode flags.
733 * 5. Extent size must be a multiple of the appropriate block size.
734 * 6. The extent size hint must be limited to half the AG size to avoid
735 *    alignment extending the extent beyond the limits of the AG.
736 */
737xfs_failaddr_t
738xfs_inode_validate_cowextsize(
739	struct xfs_mount		*mp,
740	uint32_t			cowextsize,
741	uint16_t			mode,
742	uint16_t			flags,
743	uint64_t			flags2)
744{
745	bool				rt_flag;
746	bool				hint_flag;
747	uint32_t			cowextsize_bytes;
748
749	rt_flag = (flags & XFS_DIFLAG_REALTIME);
750	hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
751	cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
752
753	if (hint_flag && !xfs_has_reflink(mp))
754		return __this_address;
755
756	if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
757		return __this_address;
758
759	if (hint_flag && cowextsize == 0)
760		return __this_address;
761
762	/* free inodes get flags set to zero but cowextsize remains */
763	if (mode && !hint_flag && cowextsize != 0)
764		return __this_address;
765
766	if (hint_flag && rt_flag)
767		return __this_address;
768
769	if (cowextsize_bytes % mp->m_sb.sb_blocksize)
770		return __this_address;
771
772	if (cowextsize > XFS_MAX_BMBT_EXTLEN)
773		return __this_address;
774
775	if (cowextsize > mp->m_sb.sb_agblocks / 2)
776		return __this_address;
777
778	return NULL;
779}