xref: /openbmc/linux/fs/xfs/libxfs/xfs_inode_buf.c (revision 160b8e75)
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)
42 void
43 xfs_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 
64 bool
65 xfs_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  */
90 static void
91 xfs_inode_buf_verify(
92 	struct xfs_buf	*bp,
93 	bool		readahead)
94 {
95 	struct xfs_mount *mp = bp->b_target->bt_mount;
96 	int		i;
97 	int		ni;
98 
99 	/*
100 	 * Validate the magic number and version of every inode in the buffer
101 	 */
102 	ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
103 	for (i = 0; i < ni; i++) {
104 		int		di_ok;
105 		xfs_dinode_t	*dip;
106 
107 		dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
108 		di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
109 			xfs_dinode_good_version(mp, dip->di_version);
110 		if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
111 						XFS_ERRTAG_ITOBP_INOTOBP))) {
112 			if (readahead) {
113 				bp->b_flags &= ~XBF_DONE;
114 				xfs_buf_ioerror(bp, -EIO);
115 				return;
116 			}
117 
118 			xfs_verifier_error(bp, -EFSCORRUPTED, __this_address);
119 #ifdef DEBUG
120 			xfs_alert(mp,
121 				"bad inode magic/vsn daddr %lld #%d (magic=%x)",
122 				(unsigned long long)bp->b_bn, i,
123 				be16_to_cpu(dip->di_magic));
124 #endif
125 		}
126 	}
127 	xfs_inobp_check(mp, bp);
128 }
129 
130 
131 static void
132 xfs_inode_buf_read_verify(
133 	struct xfs_buf	*bp)
134 {
135 	xfs_inode_buf_verify(bp, false);
136 }
137 
138 static void
139 xfs_inode_buf_readahead_verify(
140 	struct xfs_buf	*bp)
141 {
142 	xfs_inode_buf_verify(bp, true);
143 }
144 
145 static void
146 xfs_inode_buf_write_verify(
147 	struct xfs_buf	*bp)
148 {
149 	xfs_inode_buf_verify(bp, false);
150 }
151 
152 const struct xfs_buf_ops xfs_inode_buf_ops = {
153 	.name = "xfs_inode",
154 	.verify_read = xfs_inode_buf_read_verify,
155 	.verify_write = xfs_inode_buf_write_verify,
156 };
157 
158 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
159 	.name = "xxfs_inode_ra",
160 	.verify_read = xfs_inode_buf_readahead_verify,
161 	.verify_write = xfs_inode_buf_write_verify,
162 };
163 
164 
165 /*
166  * This routine is called to map an inode to the buffer containing the on-disk
167  * version of the inode.  It returns a pointer to the buffer containing the
168  * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
169  * pointer to the on-disk inode within that buffer.
170  *
171  * If a non-zero error is returned, then the contents of bpp and dipp are
172  * undefined.
173  */
174 int
175 xfs_imap_to_bp(
176 	struct xfs_mount	*mp,
177 	struct xfs_trans	*tp,
178 	struct xfs_imap		*imap,
179 	struct xfs_dinode       **dipp,
180 	struct xfs_buf		**bpp,
181 	uint			buf_flags,
182 	uint			iget_flags)
183 {
184 	struct xfs_buf		*bp;
185 	int			error;
186 
187 	buf_flags |= XBF_UNMAPPED;
188 	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
189 				   (int)imap->im_len, buf_flags, &bp,
190 				   &xfs_inode_buf_ops);
191 	if (error) {
192 		if (error == -EAGAIN) {
193 			ASSERT(buf_flags & XBF_TRYLOCK);
194 			return error;
195 		}
196 
197 		if (error == -EFSCORRUPTED &&
198 		    (iget_flags & XFS_IGET_UNTRUSTED))
199 			return -EINVAL;
200 
201 		xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
202 			__func__, error);
203 		return error;
204 	}
205 
206 	*bpp = bp;
207 	*dipp = xfs_buf_offset(bp, imap->im_boffset);
208 	return 0;
209 }
210 
211 void
212 xfs_inode_from_disk(
213 	struct xfs_inode	*ip,
214 	struct xfs_dinode	*from)
215 {
216 	struct xfs_icdinode	*to = &ip->i_d;
217 	struct inode		*inode = VFS_I(ip);
218 
219 
220 	/*
221 	 * Convert v1 inodes immediately to v2 inode format as this is the
222 	 * minimum inode version format we support in the rest of the code.
223 	 */
224 	to->di_version = from->di_version;
225 	if (to->di_version == 1) {
226 		set_nlink(inode, be16_to_cpu(from->di_onlink));
227 		to->di_projid_lo = 0;
228 		to->di_projid_hi = 0;
229 		to->di_version = 2;
230 	} else {
231 		set_nlink(inode, be32_to_cpu(from->di_nlink));
232 		to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
233 		to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
234 	}
235 
236 	to->di_format = from->di_format;
237 	to->di_uid = be32_to_cpu(from->di_uid);
238 	to->di_gid = be32_to_cpu(from->di_gid);
239 	to->di_flushiter = be16_to_cpu(from->di_flushiter);
240 
241 	/*
242 	 * Time is signed, so need to convert to signed 32 bit before
243 	 * storing in inode timestamp which may be 64 bit. Otherwise
244 	 * a time before epoch is converted to a time long after epoch
245 	 * on 64 bit systems.
246 	 */
247 	inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec);
248 	inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec);
249 	inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec);
250 	inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec);
251 	inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec);
252 	inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec);
253 	inode->i_generation = be32_to_cpu(from->di_gen);
254 	inode->i_mode = be16_to_cpu(from->di_mode);
255 
256 	to->di_size = be64_to_cpu(from->di_size);
257 	to->di_nblocks = be64_to_cpu(from->di_nblocks);
258 	to->di_extsize = be32_to_cpu(from->di_extsize);
259 	to->di_nextents = be32_to_cpu(from->di_nextents);
260 	to->di_anextents = be16_to_cpu(from->di_anextents);
261 	to->di_forkoff = from->di_forkoff;
262 	to->di_aformat	= from->di_aformat;
263 	to->di_dmevmask	= be32_to_cpu(from->di_dmevmask);
264 	to->di_dmstate	= be16_to_cpu(from->di_dmstate);
265 	to->di_flags	= be16_to_cpu(from->di_flags);
266 
267 	if (to->di_version == 3) {
268 		inode_set_iversion_queried(inode,
269 					   be64_to_cpu(from->di_changecount));
270 		to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
271 		to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
272 		to->di_flags2 = be64_to_cpu(from->di_flags2);
273 		to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
274 	}
275 }
276 
277 void
278 xfs_inode_to_disk(
279 	struct xfs_inode	*ip,
280 	struct xfs_dinode	*to,
281 	xfs_lsn_t		lsn)
282 {
283 	struct xfs_icdinode	*from = &ip->i_d;
284 	struct inode		*inode = VFS_I(ip);
285 
286 	to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
287 	to->di_onlink = 0;
288 
289 	to->di_version = from->di_version;
290 	to->di_format = from->di_format;
291 	to->di_uid = cpu_to_be32(from->di_uid);
292 	to->di_gid = cpu_to_be32(from->di_gid);
293 	to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
294 	to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
295 
296 	memset(to->di_pad, 0, sizeof(to->di_pad));
297 	to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec);
298 	to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec);
299 	to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec);
300 	to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec);
301 	to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec);
302 	to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec);
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(from->di_size);
308 	to->di_nblocks = cpu_to_be64(from->di_nblocks);
309 	to->di_extsize = cpu_to_be32(from->di_extsize);
310 	to->di_nextents = cpu_to_be32(from->di_nextents);
311 	to->di_anextents = cpu_to_be16(from->di_anextents);
312 	to->di_forkoff = from->di_forkoff;
313 	to->di_aformat = from->di_aformat;
314 	to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
315 	to->di_dmstate = cpu_to_be16(from->di_dmstate);
316 	to->di_flags = cpu_to_be16(from->di_flags);
317 
318 	if (from->di_version == 3) {
319 		to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
320 		to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
321 		to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
322 		to->di_flags2 = cpu_to_be64(from->di_flags2);
323 		to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
324 		to->di_ino = cpu_to_be64(ip->i_ino);
325 		to->di_lsn = cpu_to_be64(lsn);
326 		memset(to->di_pad2, 0, sizeof(to->di_pad2));
327 		uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
328 		to->di_flushiter = 0;
329 	} else {
330 		to->di_flushiter = cpu_to_be16(from->di_flushiter);
331 	}
332 }
333 
334 void
335 xfs_log_dinode_to_disk(
336 	struct xfs_log_dinode	*from,
337 	struct xfs_dinode	*to)
338 {
339 	to->di_magic = cpu_to_be16(from->di_magic);
340 	to->di_mode = cpu_to_be16(from->di_mode);
341 	to->di_version = from->di_version;
342 	to->di_format = from->di_format;
343 	to->di_onlink = 0;
344 	to->di_uid = cpu_to_be32(from->di_uid);
345 	to->di_gid = cpu_to_be32(from->di_gid);
346 	to->di_nlink = cpu_to_be32(from->di_nlink);
347 	to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
348 	to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
349 	memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
350 
351 	to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
352 	to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
353 	to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
354 	to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
355 	to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
356 	to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
357 
358 	to->di_size = cpu_to_be64(from->di_size);
359 	to->di_nblocks = cpu_to_be64(from->di_nblocks);
360 	to->di_extsize = cpu_to_be32(from->di_extsize);
361 	to->di_nextents = cpu_to_be32(from->di_nextents);
362 	to->di_anextents = cpu_to_be16(from->di_anextents);
363 	to->di_forkoff = from->di_forkoff;
364 	to->di_aformat = from->di_aformat;
365 	to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
366 	to->di_dmstate = cpu_to_be16(from->di_dmstate);
367 	to->di_flags = cpu_to_be16(from->di_flags);
368 	to->di_gen = cpu_to_be32(from->di_gen);
369 
370 	if (from->di_version == 3) {
371 		to->di_changecount = cpu_to_be64(from->di_changecount);
372 		to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
373 		to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
374 		to->di_flags2 = cpu_to_be64(from->di_flags2);
375 		to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
376 		to->di_ino = cpu_to_be64(from->di_ino);
377 		to->di_lsn = cpu_to_be64(from->di_lsn);
378 		memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
379 		uuid_copy(&to->di_uuid, &from->di_uuid);
380 		to->di_flushiter = 0;
381 	} else {
382 		to->di_flushiter = cpu_to_be16(from->di_flushiter);
383 	}
384 }
385 
386 xfs_failaddr_t
387 xfs_dinode_verify(
388 	struct xfs_mount	*mp,
389 	xfs_ino_t		ino,
390 	struct xfs_dinode	*dip)
391 {
392 	uint16_t		mode;
393 	uint16_t		flags;
394 	uint64_t		flags2;
395 	uint64_t		di_size;
396 
397 	if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
398 		return __this_address;
399 
400 	/* Verify v3 integrity information first */
401 	if (dip->di_version >= 3) {
402 		if (!xfs_sb_version_hascrc(&mp->m_sb))
403 			return __this_address;
404 		if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
405 				      XFS_DINODE_CRC_OFF))
406 			return __this_address;
407 		if (be64_to_cpu(dip->di_ino) != ino)
408 			return __this_address;
409 		if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
410 			return __this_address;
411 	}
412 
413 	/* don't allow invalid i_size */
414 	di_size = be64_to_cpu(dip->di_size);
415 	if (di_size & (1ULL << 63))
416 		return __this_address;
417 
418 	mode = be16_to_cpu(dip->di_mode);
419 	if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
420 		return __this_address;
421 
422 	/* No zero-length symlinks/dirs. */
423 	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
424 		return __this_address;
425 
426 	/* Fork checks carried over from xfs_iformat_fork */
427 	if (mode &&
428 	    be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
429 			be64_to_cpu(dip->di_nblocks))
430 		return __this_address;
431 
432 	if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
433 		return __this_address;
434 
435 	flags = be16_to_cpu(dip->di_flags);
436 
437 	if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
438 		return __this_address;
439 
440 	/* Do we have appropriate data fork formats for the mode? */
441 	switch (mode & S_IFMT) {
442 	case S_IFIFO:
443 	case S_IFCHR:
444 	case S_IFBLK:
445 	case S_IFSOCK:
446 		if (dip->di_format != XFS_DINODE_FMT_DEV)
447 			return __this_address;
448 		break;
449 	case S_IFREG:
450 	case S_IFLNK:
451 	case S_IFDIR:
452 		switch (dip->di_format) {
453 		case XFS_DINODE_FMT_LOCAL:
454 			/*
455 			 * no local regular files yet
456 			 */
457 			if (S_ISREG(mode))
458 				return __this_address;
459 			if (di_size > XFS_DFORK_DSIZE(dip, mp))
460 				return __this_address;
461 			/* fall through */
462 		case XFS_DINODE_FMT_EXTENTS:
463 		case XFS_DINODE_FMT_BTREE:
464 			break;
465 		default:
466 			return __this_address;
467 		}
468 		break;
469 	case 0:
470 		/* Uninitialized inode ok. */
471 		break;
472 	default:
473 		return __this_address;
474 	}
475 
476 	if (XFS_DFORK_Q(dip)) {
477 		switch (dip->di_aformat) {
478 		case XFS_DINODE_FMT_LOCAL:
479 		case XFS_DINODE_FMT_EXTENTS:
480 		case XFS_DINODE_FMT_BTREE:
481 			break;
482 		default:
483 			return __this_address;
484 		}
485 	}
486 
487 	/* only version 3 or greater inodes are extensively verified here */
488 	if (dip->di_version < 3)
489 		return NULL;
490 
491 	flags2 = be64_to_cpu(dip->di_flags2);
492 
493 	/* don't allow reflink/cowextsize if we don't have reflink */
494 	if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
495             !xfs_sb_version_hasreflink(&mp->m_sb))
496 		return __this_address;
497 
498 	/* only regular files get reflink */
499 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
500 		return __this_address;
501 
502 	/* don't let reflink and realtime mix */
503 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
504 		return __this_address;
505 
506 	/* don't let reflink and dax mix */
507 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
508 		return __this_address;
509 
510 	return NULL;
511 }
512 
513 void
514 xfs_dinode_calc_crc(
515 	struct xfs_mount	*mp,
516 	struct xfs_dinode	*dip)
517 {
518 	uint32_t		crc;
519 
520 	if (dip->di_version < 3)
521 		return;
522 
523 	ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
524 	crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
525 			      XFS_DINODE_CRC_OFF);
526 	dip->di_crc = xfs_end_cksum(crc);
527 }
528 
529 /*
530  * Read the disk inode attributes into the in-core inode structure.
531  *
532  * For version 5 superblocks, if we are initialising a new inode and we are not
533  * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
534  * inode core with a random generation number. If we are keeping inodes around,
535  * we need to read the inode cluster to get the existing generation number off
536  * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
537  * format) then log recovery is dependent on the di_flushiter field being
538  * initialised from the current on-disk value and hence we must also read the
539  * inode off disk.
540  */
541 int
542 xfs_iread(
543 	xfs_mount_t	*mp,
544 	xfs_trans_t	*tp,
545 	xfs_inode_t	*ip,
546 	uint		iget_flags)
547 {
548 	xfs_buf_t	*bp;
549 	xfs_dinode_t	*dip;
550 	xfs_failaddr_t	fa;
551 	int		error;
552 
553 	/*
554 	 * Fill in the location information in the in-core inode.
555 	 */
556 	error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
557 	if (error)
558 		return error;
559 
560 	/* shortcut IO on inode allocation if possible */
561 	if ((iget_flags & XFS_IGET_CREATE) &&
562 	    xfs_sb_version_hascrc(&mp->m_sb) &&
563 	    !(mp->m_flags & XFS_MOUNT_IKEEP)) {
564 		/* initialise the on-disk inode core */
565 		memset(&ip->i_d, 0, sizeof(ip->i_d));
566 		VFS_I(ip)->i_generation = prandom_u32();
567 		if (xfs_sb_version_hascrc(&mp->m_sb))
568 			ip->i_d.di_version = 3;
569 		else
570 			ip->i_d.di_version = 2;
571 		return 0;
572 	}
573 
574 	/*
575 	 * Get pointers to the on-disk inode and the buffer containing it.
576 	 */
577 	error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
578 	if (error)
579 		return error;
580 
581 	/* even unallocated inodes are verified */
582 	fa = xfs_dinode_verify(mp, ip->i_ino, dip);
583 	if (fa) {
584 		xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
585 				sizeof(*dip), fa);
586 		error = -EFSCORRUPTED;
587 		goto out_brelse;
588 	}
589 
590 	/*
591 	 * If the on-disk inode is already linked to a directory
592 	 * entry, copy all of the inode into the in-core inode.
593 	 * xfs_iformat_fork() handles copying in the inode format
594 	 * specific information.
595 	 * Otherwise, just get the truly permanent information.
596 	 */
597 	if (dip->di_mode) {
598 		xfs_inode_from_disk(ip, dip);
599 		error = xfs_iformat_fork(ip, dip);
600 		if (error)  {
601 #ifdef DEBUG
602 			xfs_alert(mp, "%s: xfs_iformat() returned error %d",
603 				__func__, error);
604 #endif /* DEBUG */
605 			goto out_brelse;
606 		}
607 	} else {
608 		/*
609 		 * Partial initialisation of the in-core inode. Just the bits
610 		 * that xfs_ialloc won't overwrite or relies on being correct.
611 		 */
612 		ip->i_d.di_version = dip->di_version;
613 		VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
614 		ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
615 
616 		/*
617 		 * Make sure to pull in the mode here as well in
618 		 * case the inode is released without being used.
619 		 * This ensures that xfs_inactive() will see that
620 		 * the inode is already free and not try to mess
621 		 * with the uninitialized part of it.
622 		 */
623 		VFS_I(ip)->i_mode = 0;
624 	}
625 
626 	ASSERT(ip->i_d.di_version >= 2);
627 	ip->i_delayed_blks = 0;
628 
629 	/*
630 	 * Mark the buffer containing the inode as something to keep
631 	 * around for a while.  This helps to keep recently accessed
632 	 * meta-data in-core longer.
633 	 */
634 	xfs_buf_set_ref(bp, XFS_INO_REF);
635 
636 	/*
637 	 * Use xfs_trans_brelse() to release the buffer containing the on-disk
638 	 * inode, because it was acquired with xfs_trans_read_buf() in
639 	 * xfs_imap_to_bp() above.  If tp is NULL, this is just a normal
640 	 * brelse().  If we're within a transaction, then xfs_trans_brelse()
641 	 * will only release the buffer if it is not dirty within the
642 	 * transaction.  It will be OK to release the buffer in this case,
643 	 * because inodes on disk are never destroyed and we will be locking the
644 	 * new in-core inode before putting it in the cache where other
645 	 * processes can find it.  Thus we don't have to worry about the inode
646 	 * being changed just because we released the buffer.
647 	 */
648  out_brelse:
649 	xfs_trans_brelse(tp, bp);
650 	return error;
651 }
652