aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ntfs/inode.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
committerLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
commit1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch)
tree0bba044c4ce775e45a88a51686b5d9f90697ea9d /fs/ntfs/inode.c
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
Diffstat (limited to 'fs/ntfs/inode.c')
-rw-r--r--fs/ntfs/inode.c2616
1 files changed, 2616 insertions, 0 deletions
diff --git a/fs/ntfs/inode.c b/fs/ntfs/inode.c
new file mode 100644
index 000000000000..31840ba0b38c
--- /dev/null
+++ b/fs/ntfs/inode.c
@@ -0,0 +1,2616 @@
1/**
2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
3 *
4 * Copyright (c) 2001-2004 Anton Altaparmakov
5 *
6 * This program/include file is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as published
8 * by the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * This program/include file is distributed in the hope that it will be
12 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
13 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program (in the main directory of the Linux-NTFS
18 * distribution in the file COPYING); if not, write to the Free Software
19 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 */
21
22#include <linux/pagemap.h>
23#include <linux/buffer_head.h>
24#include <linux/smp_lock.h>
25#include <linux/quotaops.h>
26#include <linux/mount.h>
27
28#include "aops.h"
29#include "dir.h"
30#include "debug.h"
31#include "inode.h"
32#include "attrib.h"
33#include "malloc.h"
34#include "mft.h"
35#include "time.h"
36#include "ntfs.h"
37
38/**
39 * ntfs_test_inode - compare two (possibly fake) inodes for equality
40 * @vi: vfs inode which to test
41 * @na: ntfs attribute which is being tested with
42 *
43 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
44 * inode @vi for equality with the ntfs attribute @na.
45 *
46 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
47 * @na->name and @na->name_len are then ignored.
48 *
49 * Return 1 if the attributes match and 0 if not.
50 *
51 * NOTE: This function runs with the inode_lock spin lock held so it is not
52 * allowed to sleep.
53 */
54int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
55{
56 ntfs_inode *ni;
57
58 if (vi->i_ino != na->mft_no)
59 return 0;
60 ni = NTFS_I(vi);
61 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
62 if (likely(!NInoAttr(ni))) {
63 /* If not looking for a normal inode this is a mismatch. */
64 if (unlikely(na->type != AT_UNUSED))
65 return 0;
66 } else {
67 /* A fake inode describing an attribute. */
68 if (ni->type != na->type)
69 return 0;
70 if (ni->name_len != na->name_len)
71 return 0;
72 if (na->name_len && memcmp(ni->name, na->name,
73 na->name_len * sizeof(ntfschar)))
74 return 0;
75 }
76 /* Match! */
77 return 1;
78}
79
80/**
81 * ntfs_init_locked_inode - initialize an inode
82 * @vi: vfs inode to initialize
83 * @na: ntfs attribute which to initialize @vi to
84 *
85 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
86 * order to enable ntfs_test_inode() to do its work.
87 *
88 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
89 * In that case, @na->name and @na->name_len should be set to NULL and 0,
90 * respectively. Although that is not strictly necessary as
91 * ntfs_read_inode_locked() will fill them in later.
92 *
93 * Return 0 on success and -errno on error.
94 *
95 * NOTE: This function runs with the inode_lock spin lock held so it is not
96 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
97 */
98static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
99{
100 ntfs_inode *ni = NTFS_I(vi);
101
102 vi->i_ino = na->mft_no;
103
104 ni->type = na->type;
105 if (na->type == AT_INDEX_ALLOCATION)
106 NInoSetMstProtected(ni);
107
108 ni->name = na->name;
109 ni->name_len = na->name_len;
110
111 /* If initializing a normal inode, we are done. */
112 if (likely(na->type == AT_UNUSED)) {
113 BUG_ON(na->name);
114 BUG_ON(na->name_len);
115 return 0;
116 }
117
118 /* It is a fake inode. */
119 NInoSetAttr(ni);
120
121 /*
122 * We have I30 global constant as an optimization as it is the name
123 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
124 * allocation but that is ok. And most attributes are unnamed anyway,
125 * thus the fraction of named attributes with name != I30 is actually
126 * absolutely tiny.
127 */
128 if (na->name_len && na->name != I30) {
129 unsigned int i;
130
131 BUG_ON(!na->name);
132 i = na->name_len * sizeof(ntfschar);
133 ni->name = (ntfschar*)kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
134 if (!ni->name)
135 return -ENOMEM;
136 memcpy(ni->name, na->name, i);
137 ni->name[i] = 0;
138 }
139 return 0;
140}
141
142typedef int (*set_t)(struct inode *, void *);
143static int ntfs_read_locked_inode(struct inode *vi);
144static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
145static int ntfs_read_locked_index_inode(struct inode *base_vi,
146 struct inode *vi);
147
148/**
149 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
150 * @sb: super block of mounted volume
151 * @mft_no: mft record number / inode number to obtain
152 *
153 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
154 * file or directory).
155 *
156 * If the inode is in the cache, it is just returned with an increased
157 * reference count. Otherwise, a new struct inode is allocated and initialized,
158 * and finally ntfs_read_locked_inode() is called to read in the inode and
159 * fill in the remainder of the inode structure.
160 *
161 * Return the struct inode on success. Check the return value with IS_ERR() and
162 * if true, the function failed and the error code is obtained from PTR_ERR().
163 */
164struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
165{
166 struct inode *vi;
167 ntfs_attr na;
168 int err;
169
170 na.mft_no = mft_no;
171 na.type = AT_UNUSED;
172 na.name = NULL;
173 na.name_len = 0;
174
175 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
176 (set_t)ntfs_init_locked_inode, &na);
177 if (!vi)
178 return ERR_PTR(-ENOMEM);
179
180 err = 0;
181
182 /* If this is a freshly allocated inode, need to read it now. */
183 if (vi->i_state & I_NEW) {
184 err = ntfs_read_locked_inode(vi);
185 unlock_new_inode(vi);
186 }
187 /*
188 * There is no point in keeping bad inodes around if the failure was
189 * due to ENOMEM. We want to be able to retry again later.
190 */
191 if (err == -ENOMEM) {
192 iput(vi);
193 vi = ERR_PTR(err);
194 }
195 return vi;
196}
197
198/**
199 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
200 * @base_vi: vfs base inode containing the attribute
201 * @type: attribute type
202 * @name: Unicode name of the attribute (NULL if unnamed)
203 * @name_len: length of @name in Unicode characters (0 if unnamed)
204 *
205 * Obtain the (fake) struct inode corresponding to the attribute specified by
206 * @type, @name, and @name_len, which is present in the base mft record
207 * specified by the vfs inode @base_vi.
208 *
209 * If the attribute inode is in the cache, it is just returned with an
210 * increased reference count. Otherwise, a new struct inode is allocated and
211 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
212 * attribute and fill in the inode structure.
213 *
214 * Note, for index allocation attributes, you need to use ntfs_index_iget()
215 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
216 *
217 * Return the struct inode of the attribute inode on success. Check the return
218 * value with IS_ERR() and if true, the function failed and the error code is
219 * obtained from PTR_ERR().
220 */
221struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
222 ntfschar *name, u32 name_len)
223{
224 struct inode *vi;
225 ntfs_attr na;
226 int err;
227
228 /* Make sure no one calls ntfs_attr_iget() for indices. */
229 BUG_ON(type == AT_INDEX_ALLOCATION);
230
231 na.mft_no = base_vi->i_ino;
232 na.type = type;
233 na.name = name;
234 na.name_len = name_len;
235
236 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
237 (set_t)ntfs_init_locked_inode, &na);
238 if (!vi)
239 return ERR_PTR(-ENOMEM);
240
241 err = 0;
242
243 /* If this is a freshly allocated inode, need to read it now. */
244 if (vi->i_state & I_NEW) {
245 err = ntfs_read_locked_attr_inode(base_vi, vi);
246 unlock_new_inode(vi);
247 }
248 /*
249 * There is no point in keeping bad attribute inodes around. This also
250 * simplifies things in that we never need to check for bad attribute
251 * inodes elsewhere.
252 */
253 if (err) {
254 iput(vi);
255 vi = ERR_PTR(err);
256 }
257 return vi;
258}
259
260/**
261 * ntfs_index_iget - obtain a struct inode corresponding to an index
262 * @base_vi: vfs base inode containing the index related attributes
263 * @name: Unicode name of the index
264 * @name_len: length of @name in Unicode characters
265 *
266 * Obtain the (fake) struct inode corresponding to the index specified by @name
267 * and @name_len, which is present in the base mft record specified by the vfs
268 * inode @base_vi.
269 *
270 * If the index inode is in the cache, it is just returned with an increased
271 * reference count. Otherwise, a new struct inode is allocated and
272 * initialized, and finally ntfs_read_locked_index_inode() is called to read
273 * the index related attributes and fill in the inode structure.
274 *
275 * Return the struct inode of the index inode on success. Check the return
276 * value with IS_ERR() and if true, the function failed and the error code is
277 * obtained from PTR_ERR().
278 */
279struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
280 u32 name_len)
281{
282 struct inode *vi;
283 ntfs_attr na;
284 int err;
285
286 na.mft_no = base_vi->i_ino;
287 na.type = AT_INDEX_ALLOCATION;
288 na.name = name;
289 na.name_len = name_len;
290
291 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
292 (set_t)ntfs_init_locked_inode, &na);
293 if (!vi)
294 return ERR_PTR(-ENOMEM);
295
296 err = 0;
297
298 /* If this is a freshly allocated inode, need to read it now. */
299 if (vi->i_state & I_NEW) {
300 err = ntfs_read_locked_index_inode(base_vi, vi);
301 unlock_new_inode(vi);
302 }
303 /*
304 * There is no point in keeping bad index inodes around. This also
305 * simplifies things in that we never need to check for bad index
306 * inodes elsewhere.
307 */
308 if (err) {
309 iput(vi);
310 vi = ERR_PTR(err);
311 }
312 return vi;
313}
314
315struct inode *ntfs_alloc_big_inode(struct super_block *sb)
316{
317 ntfs_inode *ni;
318
319 ntfs_debug("Entering.");
320 ni = (ntfs_inode *)kmem_cache_alloc(ntfs_big_inode_cache,
321 SLAB_NOFS);
322 if (likely(ni != NULL)) {
323 ni->state = 0;
324 return VFS_I(ni);
325 }
326 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
327 return NULL;
328}
329
330void ntfs_destroy_big_inode(struct inode *inode)
331{
332 ntfs_inode *ni = NTFS_I(inode);
333
334 ntfs_debug("Entering.");
335 BUG_ON(ni->page);
336 if (!atomic_dec_and_test(&ni->count))
337 BUG();
338 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
339}
340
341static inline ntfs_inode *ntfs_alloc_extent_inode(void)
342{
343 ntfs_inode *ni;
344
345 ntfs_debug("Entering.");
346 ni = (ntfs_inode *)kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS);
347 if (likely(ni != NULL)) {
348 ni->state = 0;
349 return ni;
350 }
351 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
352 return NULL;
353}
354
355static void ntfs_destroy_extent_inode(ntfs_inode *ni)
356{
357 ntfs_debug("Entering.");
358 BUG_ON(ni->page);
359 if (!atomic_dec_and_test(&ni->count))
360 BUG();
361 kmem_cache_free(ntfs_inode_cache, ni);
362}
363
364/**
365 * __ntfs_init_inode - initialize ntfs specific part of an inode
366 * @sb: super block of mounted volume
367 * @ni: freshly allocated ntfs inode which to initialize
368 *
369 * Initialize an ntfs inode to defaults.
370 *
371 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
372 * untouched. Make sure to initialize them elsewhere.
373 *
374 * Return zero on success and -ENOMEM on error.
375 */
376void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
377{
378 ntfs_debug("Entering.");
379 ni->initialized_size = ni->allocated_size = 0;
380 ni->seq_no = 0;
381 atomic_set(&ni->count, 1);
382 ni->vol = NTFS_SB(sb);
383 ntfs_init_runlist(&ni->runlist);
384 init_MUTEX(&ni->mrec_lock);
385 ni->page = NULL;
386 ni->page_ofs = 0;
387 ni->attr_list_size = 0;
388 ni->attr_list = NULL;
389 ntfs_init_runlist(&ni->attr_list_rl);
390 ni->itype.index.bmp_ino = NULL;
391 ni->itype.index.block_size = 0;
392 ni->itype.index.vcn_size = 0;
393 ni->itype.index.collation_rule = 0;
394 ni->itype.index.block_size_bits = 0;
395 ni->itype.index.vcn_size_bits = 0;
396 init_MUTEX(&ni->extent_lock);
397 ni->nr_extents = 0;
398 ni->ext.base_ntfs_ino = NULL;
399}
400
401inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
402 unsigned long mft_no)
403{
404 ntfs_inode *ni = ntfs_alloc_extent_inode();
405
406 ntfs_debug("Entering.");
407 if (likely(ni != NULL)) {
408 __ntfs_init_inode(sb, ni);
409 ni->mft_no = mft_no;
410 ni->type = AT_UNUSED;
411 ni->name = NULL;
412 ni->name_len = 0;
413 }
414 return ni;
415}
416
417/**
418 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
419 * @ctx: initialized attribute search context
420 *
421 * Search all file name attributes in the inode described by the attribute
422 * search context @ctx and check if any of the names are in the $Extend system
423 * directory.
424 *
425 * Return values:
426 * 1: file is in $Extend directory
427 * 0: file is not in $Extend directory
428 * -errno: failed to determine if the file is in the $Extend directory
429 */
430static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
431{
432 int nr_links, err;
433
434 /* Restart search. */
435 ntfs_attr_reinit_search_ctx(ctx);
436
437 /* Get number of hard links. */
438 nr_links = le16_to_cpu(ctx->mrec->link_count);
439
440 /* Loop through all hard links. */
441 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
442 ctx))) {
443 FILE_NAME_ATTR *file_name_attr;
444 ATTR_RECORD *attr = ctx->attr;
445 u8 *p, *p2;
446
447 nr_links--;
448 /*
449 * Maximum sanity checking as we are called on an inode that
450 * we suspect might be corrupt.
451 */
452 p = (u8*)attr + le32_to_cpu(attr->length);
453 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
454 le32_to_cpu(ctx->mrec->bytes_in_use)) {
455err_corrupt_attr:
456 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
457 "attribute. You should run chkdsk.");
458 return -EIO;
459 }
460 if (attr->non_resident) {
461 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
462 "name. You should run chkdsk.");
463 return -EIO;
464 }
465 if (attr->flags) {
466 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
467 "invalid flags. You should run "
468 "chkdsk.");
469 return -EIO;
470 }
471 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
472 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
473 "name. You should run chkdsk.");
474 return -EIO;
475 }
476 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
477 le16_to_cpu(attr->data.resident.value_offset));
478 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
479 if (p2 < (u8*)attr || p2 > p)
480 goto err_corrupt_attr;
481 /* This attribute is ok, but is it in the $Extend directory? */
482 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
483 return 1; /* YES, it's an extended system file. */
484 }
485 if (unlikely(err != -ENOENT))
486 return err;
487 if (unlikely(nr_links)) {
488 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
489 "doesn't match number of name attributes. You "
490 "should run chkdsk.");
491 return -EIO;
492 }
493 return 0; /* NO, it is not an extended system file. */
494}
495
496/**
497 * ntfs_read_locked_inode - read an inode from its device
498 * @vi: inode to read
499 *
500 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
501 * described by @vi into memory from the device.
502 *
503 * The only fields in @vi that we need to/can look at when the function is
504 * called are i_sb, pointing to the mounted device's super block, and i_ino,
505 * the number of the inode to load.
506 *
507 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
508 * for reading and sets up the necessary @vi fields as well as initializing
509 * the ntfs inode.
510 *
511 * Q: What locks are held when the function is called?
512 * A: i_state has I_LOCK set, hence the inode is locked, also
513 * i_count is set to 1, so it is not going to go away
514 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
515 * is allowed to write to them. We should of course be honouring them but
516 * we need to do that using the IS_* macros defined in include/linux/fs.h.
517 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
518 *
519 * Return 0 on success and -errno on error. In the error case, the inode will
520 * have had make_bad_inode() executed on it.
521 */
522static int ntfs_read_locked_inode(struct inode *vi)
523{
524 ntfs_volume *vol = NTFS_SB(vi->i_sb);
525 ntfs_inode *ni;
526 MFT_RECORD *m;
527 STANDARD_INFORMATION *si;
528 ntfs_attr_search_ctx *ctx;
529 int err = 0;
530
531 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
532
533 /* Setup the generic vfs inode parts now. */
534
535 /* This is the optimal IO size (for stat), not the fs block size. */
536 vi->i_blksize = PAGE_CACHE_SIZE;
537 /*
538 * This is for checking whether an inode has changed w.r.t. a file so
539 * that the file can be updated if necessary (compare with f_version).
540 */
541 vi->i_version = 1;
542
543 vi->i_uid = vol->uid;
544 vi->i_gid = vol->gid;
545 vi->i_mode = 0;
546
547 /*
548 * Initialize the ntfs specific part of @vi special casing
549 * FILE_MFT which we need to do at mount time.
550 */
551 if (vi->i_ino != FILE_MFT)
552 ntfs_init_big_inode(vi);
553 ni = NTFS_I(vi);
554
555 m = map_mft_record(ni);
556 if (IS_ERR(m)) {
557 err = PTR_ERR(m);
558 goto err_out;
559 }
560 ctx = ntfs_attr_get_search_ctx(ni, m);
561 if (!ctx) {
562 err = -ENOMEM;
563 goto unm_err_out;
564 }
565
566 if (!(m->flags & MFT_RECORD_IN_USE)) {
567 ntfs_error(vi->i_sb, "Inode is not in use!");
568 goto unm_err_out;
569 }
570 if (m->base_mft_record) {
571 ntfs_error(vi->i_sb, "Inode is an extent inode!");
572 goto unm_err_out;
573 }
574
575 /* Transfer information from mft record into vfs and ntfs inodes. */
576 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
577
578 /*
579 * FIXME: Keep in mind that link_count is two for files which have both
580 * a long file name and a short file name as separate entries, so if
581 * we are hiding short file names this will be too high. Either we need
582 * to account for the short file names by subtracting them or we need
583 * to make sure we delete files even though i_nlink is not zero which
584 * might be tricky due to vfs interactions. Need to think about this
585 * some more when implementing the unlink command.
586 */
587 vi->i_nlink = le16_to_cpu(m->link_count);
588 /*
589 * FIXME: Reparse points can have the directory bit set even though
590 * they would be S_IFLNK. Need to deal with this further below when we
591 * implement reparse points / symbolic links but it will do for now.
592 * Also if not a directory, it could be something else, rather than
593 * a regular file. But again, will do for now.
594 */
595 /* Everyone gets all permissions. */
596 vi->i_mode |= S_IRWXUGO;
597 /* If read-only, noone gets write permissions. */
598 if (IS_RDONLY(vi))
599 vi->i_mode &= ~S_IWUGO;
600 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
601 vi->i_mode |= S_IFDIR;
602 /*
603 * Apply the directory permissions mask set in the mount
604 * options.
605 */
606 vi->i_mode &= ~vol->dmask;
607 /* Things break without this kludge! */
608 if (vi->i_nlink > 1)
609 vi->i_nlink = 1;
610 } else {
611 vi->i_mode |= S_IFREG;
612 /* Apply the file permissions mask set in the mount options. */
613 vi->i_mode &= ~vol->fmask;
614 }
615 /*
616 * Find the standard information attribute in the mft record. At this
617 * stage we haven't setup the attribute list stuff yet, so this could
618 * in fact fail if the standard information is in an extent record, but
619 * I don't think this actually ever happens.
620 */
621 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
622 ctx);
623 if (unlikely(err)) {
624 if (err == -ENOENT) {
625 /*
626 * TODO: We should be performing a hot fix here (if the
627 * recover mount option is set) by creating a new
628 * attribute.
629 */
630 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
631 "is missing.");
632 }
633 goto unm_err_out;
634 }
635 /* Get the standard information attribute value. */
636 si = (STANDARD_INFORMATION*)((char*)ctx->attr +
637 le16_to_cpu(ctx->attr->data.resident.value_offset));
638
639 /* Transfer information from the standard information into vi. */
640 /*
641 * Note: The i_?times do not quite map perfectly onto the NTFS times,
642 * but they are close enough, and in the end it doesn't really matter
643 * that much...
644 */
645 /*
646 * mtime is the last change of the data within the file. Not changed
647 * when only metadata is changed, e.g. a rename doesn't affect mtime.
648 */
649 vi->i_mtime = ntfs2utc(si->last_data_change_time);
650 /*
651 * ctime is the last change of the metadata of the file. This obviously
652 * always changes, when mtime is changed. ctime can be changed on its
653 * own, mtime is then not changed, e.g. when a file is renamed.
654 */
655 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
656 /*
657 * Last access to the data within the file. Not changed during a rename
658 * for example but changed whenever the file is written to.
659 */
660 vi->i_atime = ntfs2utc(si->last_access_time);
661
662 /* Find the attribute list attribute if present. */
663 ntfs_attr_reinit_search_ctx(ctx);
664 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
665 if (err) {
666 if (unlikely(err != -ENOENT)) {
667 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
668 "attribute.");
669 goto unm_err_out;
670 }
671 } else /* if (!err) */ {
672 if (vi->i_ino == FILE_MFT)
673 goto skip_attr_list_load;
674 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
675 NInoSetAttrList(ni);
676 if (ctx->attr->flags & ATTR_IS_ENCRYPTED ||
677 ctx->attr->flags & ATTR_COMPRESSION_MASK ||
678 ctx->attr->flags & ATTR_IS_SPARSE) {
679 ntfs_error(vi->i_sb, "Attribute list attribute is "
680 "compressed/encrypted/sparse.");
681 goto unm_err_out;
682 }
683 /* Now allocate memory for the attribute list. */
684 ni->attr_list_size = (u32)ntfs_attr_size(ctx->attr);
685 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
686 if (!ni->attr_list) {
687 ntfs_error(vi->i_sb, "Not enough memory to allocate "
688 "buffer for attribute list.");
689 err = -ENOMEM;
690 goto unm_err_out;
691 }
692 if (ctx->attr->non_resident) {
693 NInoSetAttrListNonResident(ni);
694 if (ctx->attr->data.non_resident.lowest_vcn) {
695 ntfs_error(vi->i_sb, "Attribute list has non "
696 "zero lowest_vcn.");
697 goto unm_err_out;
698 }
699 /*
700 * Setup the runlist. No need for locking as we have
701 * exclusive access to the inode at this time.
702 */
703 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
704 ctx->attr, NULL);
705 if (IS_ERR(ni->attr_list_rl.rl)) {
706 err = PTR_ERR(ni->attr_list_rl.rl);
707 ni->attr_list_rl.rl = NULL;
708 ntfs_error(vi->i_sb, "Mapping pairs "
709 "decompression failed.");
710 goto unm_err_out;
711 }
712 /* Now load the attribute list. */
713 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
714 ni->attr_list, ni->attr_list_size,
715 sle64_to_cpu(ctx->attr->data.
716 non_resident.initialized_size)))) {
717 ntfs_error(vi->i_sb, "Failed to load "
718 "attribute list attribute.");
719 goto unm_err_out;
720 }
721 } else /* if (!ctx.attr->non_resident) */ {
722 if ((u8*)ctx->attr + le16_to_cpu(
723 ctx->attr->data.resident.value_offset) +
724 le32_to_cpu(
725 ctx->attr->data.resident.value_length) >
726 (u8*)ctx->mrec + vol->mft_record_size) {
727 ntfs_error(vi->i_sb, "Corrupt attribute list "
728 "in inode.");
729 goto unm_err_out;
730 }
731 /* Now copy the attribute list. */
732 memcpy(ni->attr_list, (u8*)ctx->attr + le16_to_cpu(
733 ctx->attr->data.resident.value_offset),
734 le32_to_cpu(
735 ctx->attr->data.resident.value_length));
736 }
737 }
738skip_attr_list_load:
739 /*
740 * If an attribute list is present we now have the attribute list value
741 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
742 */
743 if (S_ISDIR(vi->i_mode)) {
744 struct inode *bvi;
745 ntfs_inode *bni;
746 INDEX_ROOT *ir;
747 char *ir_end, *index_end;
748
749 /* It is a directory, find index root attribute. */
750 ntfs_attr_reinit_search_ctx(ctx);
751 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
752 0, NULL, 0, ctx);
753 if (unlikely(err)) {
754 if (err == -ENOENT) {
755 // FIXME: File is corrupt! Hot-fix with empty
756 // index root attribute if recovery option is
757 // set.
758 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
759 "is missing.");
760 }
761 goto unm_err_out;
762 }
763 /* Set up the state. */
764 if (unlikely(ctx->attr->non_resident)) {
765 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
766 "resident.");
767 goto unm_err_out;
768 }
769 /* Ensure the attribute name is placed before the value. */
770 if (unlikely(ctx->attr->name_length &&
771 (le16_to_cpu(ctx->attr->name_offset) >=
772 le16_to_cpu(ctx->attr->data.resident.
773 value_offset)))) {
774 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
775 "placed after the attribute value.");
776 goto unm_err_out;
777 }
778 /*
779 * Compressed/encrypted index root just means that the newly
780 * created files in that directory should be created compressed/
781 * encrypted. However index root cannot be both compressed and
782 * encrypted.
783 */
784 if (ctx->attr->flags & ATTR_COMPRESSION_MASK)
785 NInoSetCompressed(ni);
786 if (ctx->attr->flags & ATTR_IS_ENCRYPTED) {
787 if (ctx->attr->flags & ATTR_COMPRESSION_MASK) {
788 ntfs_error(vi->i_sb, "Found encrypted and "
789 "compressed attribute.");
790 goto unm_err_out;
791 }
792 NInoSetEncrypted(ni);
793 }
794 if (ctx->attr->flags & ATTR_IS_SPARSE)
795 NInoSetSparse(ni);
796 ir = (INDEX_ROOT*)((char*)ctx->attr + le16_to_cpu(
797 ctx->attr->data.resident.value_offset));
798 ir_end = (char*)ir + le32_to_cpu(
799 ctx->attr->data.resident.value_length);
800 if (ir_end > (char*)ctx->mrec + vol->mft_record_size) {
801 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
802 "corrupt.");
803 goto unm_err_out;
804 }
805 index_end = (char*)&ir->index +
806 le32_to_cpu(ir->index.index_length);
807 if (index_end > ir_end) {
808 ntfs_error(vi->i_sb, "Directory index is corrupt.");
809 goto unm_err_out;
810 }
811 if (ir->type != AT_FILE_NAME) {
812 ntfs_error(vi->i_sb, "Indexed attribute is not "
813 "$FILE_NAME.");
814 goto unm_err_out;
815 }
816 if (ir->collation_rule != COLLATION_FILE_NAME) {
817 ntfs_error(vi->i_sb, "Index collation rule is not "
818 "COLLATION_FILE_NAME.");
819 goto unm_err_out;
820 }
821 ni->itype.index.collation_rule = ir->collation_rule;
822 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
823 if (ni->itype.index.block_size &
824 (ni->itype.index.block_size - 1)) {
825 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
826 "power of two.",
827 ni->itype.index.block_size);
828 goto unm_err_out;
829 }
830 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
831 ntfs_error(vi->i_sb, "Index block size (%u) > "
832 "PAGE_CACHE_SIZE (%ld) is not "
833 "supported. Sorry.",
834 ni->itype.index.block_size,
835 PAGE_CACHE_SIZE);
836 err = -EOPNOTSUPP;
837 goto unm_err_out;
838 }
839 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
840 ntfs_error(vi->i_sb, "Index block size (%u) < "
841 "NTFS_BLOCK_SIZE (%i) is not "
842 "supported. Sorry.",
843 ni->itype.index.block_size,
844 NTFS_BLOCK_SIZE);
845 err = -EOPNOTSUPP;
846 goto unm_err_out;
847 }
848 ni->itype.index.block_size_bits =
849 ffs(ni->itype.index.block_size) - 1;
850 /* Determine the size of a vcn in the directory index. */
851 if (vol->cluster_size <= ni->itype.index.block_size) {
852 ni->itype.index.vcn_size = vol->cluster_size;
853 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
854 } else {
855 ni->itype.index.vcn_size = vol->sector_size;
856 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
857 }
858
859 /* Setup the index allocation attribute, even if not present. */
860 NInoSetMstProtected(ni);
861 ni->type = AT_INDEX_ALLOCATION;
862 ni->name = I30;
863 ni->name_len = 4;
864
865 if (!(ir->index.flags & LARGE_INDEX)) {
866 /* No index allocation. */
867 vi->i_size = ni->initialized_size =
868 ni->allocated_size = 0;
869 /* We are done with the mft record, so we release it. */
870 ntfs_attr_put_search_ctx(ctx);
871 unmap_mft_record(ni);
872 m = NULL;
873 ctx = NULL;
874 goto skip_large_dir_stuff;
875 } /* LARGE_INDEX: Index allocation present. Setup state. */
876 NInoSetIndexAllocPresent(ni);
877 /* Find index allocation attribute. */
878 ntfs_attr_reinit_search_ctx(ctx);
879 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
880 CASE_SENSITIVE, 0, NULL, 0, ctx);
881 if (unlikely(err)) {
882 if (err == -ENOENT)
883 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
884 "attribute is not present but "
885 "$INDEX_ROOT indicated it is.");
886 else
887 ntfs_error(vi->i_sb, "Failed to lookup "
888 "$INDEX_ALLOCATION "
889 "attribute.");
890 goto unm_err_out;
891 }
892 if (!ctx->attr->non_resident) {
893 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
894 "is resident.");
895 goto unm_err_out;
896 }
897 /*
898 * Ensure the attribute name is placed before the mapping pairs
899 * array.
900 */
901 if (unlikely(ctx->attr->name_length &&
902 (le16_to_cpu(ctx->attr->name_offset) >=
903 le16_to_cpu(ctx->attr->data.non_resident.
904 mapping_pairs_offset)))) {
905 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
906 "is placed after the mapping pairs "
907 "array.");
908 goto unm_err_out;
909 }
910 if (ctx->attr->flags & ATTR_IS_ENCRYPTED) {
911 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
912 "is encrypted.");
913 goto unm_err_out;
914 }
915 if (ctx->attr->flags & ATTR_IS_SPARSE) {
916 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
917 "is sparse.");
918 goto unm_err_out;
919 }
920 if (ctx->attr->flags & ATTR_COMPRESSION_MASK) {
921 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
922 "is compressed.");
923 goto unm_err_out;
924 }
925 if (ctx->attr->data.non_resident.lowest_vcn) {
926 ntfs_error(vi->i_sb, "First extent of "
927 "$INDEX_ALLOCATION attribute has non "
928 "zero lowest_vcn.");
929 goto unm_err_out;
930 }
931 vi->i_size = sle64_to_cpu(
932 ctx->attr->data.non_resident.data_size);
933 ni->initialized_size = sle64_to_cpu(
934 ctx->attr->data.non_resident.initialized_size);
935 ni->allocated_size = sle64_to_cpu(
936 ctx->attr->data.non_resident.allocated_size);
937 /*
938 * We are done with the mft record, so we release it. Otherwise
939 * we would deadlock in ntfs_attr_iget().
940 */
941 ntfs_attr_put_search_ctx(ctx);
942 unmap_mft_record(ni);
943 m = NULL;
944 ctx = NULL;
945 /* Get the index bitmap attribute inode. */
946 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
947 if (IS_ERR(bvi)) {
948 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
949 err = PTR_ERR(bvi);
950 goto unm_err_out;
951 }
952 ni->itype.index.bmp_ino = bvi;
953 bni = NTFS_I(bvi);
954 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
955 NInoSparse(bni)) {
956 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
957 "and/or encrypted and/or sparse.");
958 goto unm_err_out;
959 }
960 /* Consistency check bitmap size vs. index allocation size. */
961 if ((bvi->i_size << 3) < (vi->i_size >>
962 ni->itype.index.block_size_bits)) {
963 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
964 "for index allocation (0x%llx).",
965 bvi->i_size << 3, vi->i_size);
966 goto unm_err_out;
967 }
968skip_large_dir_stuff:
969 /* Setup the operations for this inode. */
970 vi->i_op = &ntfs_dir_inode_ops;
971 vi->i_fop = &ntfs_dir_ops;
972 } else {
973 /* It is a file. */
974 ntfs_attr_reinit_search_ctx(ctx);
975
976 /* Setup the data attribute, even if not present. */
977 ni->type = AT_DATA;
978 ni->name = NULL;
979 ni->name_len = 0;
980
981 /* Find first extent of the unnamed data attribute. */
982 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
983 if (unlikely(err)) {
984 vi->i_size = ni->initialized_size =
985 ni->allocated_size = 0;
986 if (err != -ENOENT) {
987 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
988 "attribute.");
989 goto unm_err_out;
990 }
991 /*
992 * FILE_Secure does not have an unnamed $DATA
993 * attribute, so we special case it here.
994 */
995 if (vi->i_ino == FILE_Secure)
996 goto no_data_attr_special_case;
997 /*
998 * Most if not all the system files in the $Extend
999 * system directory do not have unnamed data
1000 * attributes so we need to check if the parent
1001 * directory of the file is FILE_Extend and if it is
1002 * ignore this error. To do this we need to get the
1003 * name of this inode from the mft record as the name
1004 * contains the back reference to the parent directory.
1005 */
1006 if (ntfs_is_extended_system_file(ctx) > 0)
1007 goto no_data_attr_special_case;
1008 // FIXME: File is corrupt! Hot-fix with empty data
1009 // attribute if recovery option is set.
1010 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1011 goto unm_err_out;
1012 }
1013 /* Setup the state. */
1014 if (ctx->attr->non_resident) {
1015 NInoSetNonResident(ni);
1016 if (ctx->attr->flags & ATTR_COMPRESSION_MASK) {
1017 NInoSetCompressed(ni);
1018 if (vol->cluster_size > 4096) {
1019 ntfs_error(vi->i_sb, "Found "
1020 "compressed data but "
1021 "compression is disabled due "
1022 "to cluster size (%i) > 4kiB.",
1023 vol->cluster_size);
1024 goto unm_err_out;
1025 }
1026 if ((ctx->attr->flags & ATTR_COMPRESSION_MASK)
1027 != ATTR_IS_COMPRESSED) {
1028 ntfs_error(vi->i_sb, "Found "
1029 "unknown compression method or "
1030 "corrupt file.");
1031 goto unm_err_out;
1032 }
1033 ni->itype.compressed.block_clusters = 1U <<
1034 ctx->attr->data.non_resident.
1035 compression_unit;
1036 if (ctx->attr->data.non_resident.
1037 compression_unit != 4) {
1038 ntfs_error(vi->i_sb, "Found "
1039 "nonstandard compression unit "
1040 "(%u instead of 4). Cannot "
1041 "handle this.",
1042 ctx->attr->data.non_resident.
1043 compression_unit);
1044 err = -EOPNOTSUPP;
1045 goto unm_err_out;
1046 }
1047 ni->itype.compressed.block_size = 1U << (
1048 ctx->attr->data.non_resident.
1049 compression_unit +
1050 vol->cluster_size_bits);
1051 ni->itype.compressed.block_size_bits = ffs(
1052 ni->itype.compressed.block_size) - 1;
1053 }
1054 if (ctx->attr->flags & ATTR_IS_ENCRYPTED) {
1055 if (ctx->attr->flags & ATTR_COMPRESSION_MASK) {
1056 ntfs_error(vi->i_sb, "Found encrypted "
1057 "and compressed data.");
1058 goto unm_err_out;
1059 }
1060 NInoSetEncrypted(ni);
1061 }
1062 if (ctx->attr->flags & ATTR_IS_SPARSE)
1063 NInoSetSparse(ni);
1064 if (ctx->attr->data.non_resident.lowest_vcn) {
1065 ntfs_error(vi->i_sb, "First extent of $DATA "
1066 "attribute has non zero "
1067 "lowest_vcn.");
1068 goto unm_err_out;
1069 }
1070 /* Setup all the sizes. */
1071 vi->i_size = sle64_to_cpu(
1072 ctx->attr->data.non_resident.data_size);
1073 ni->initialized_size = sle64_to_cpu(
1074 ctx->attr->data.non_resident.
1075 initialized_size);
1076 ni->allocated_size = sle64_to_cpu(
1077 ctx->attr->data.non_resident.
1078 allocated_size);
1079 if (NInoCompressed(ni)) {
1080 ni->itype.compressed.size = sle64_to_cpu(
1081 ctx->attr->data.non_resident.
1082 compressed_size);
1083 }
1084 } else { /* Resident attribute. */
1085 /*
1086 * Make all sizes equal for simplicity in read code
1087 * paths. FIXME: Need to keep this in mind when
1088 * converting to non-resident attribute in write code
1089 * path. (Probably only affects truncate().)
1090 */
1091 vi->i_size = ni->initialized_size = ni->allocated_size =
1092 le32_to_cpu(
1093 ctx->attr->data.resident.value_length);
1094 }
1095no_data_attr_special_case:
1096 /* We are done with the mft record, so we release it. */
1097 ntfs_attr_put_search_ctx(ctx);
1098 unmap_mft_record(ni);
1099 m = NULL;
1100 ctx = NULL;
1101 /* Setup the operations for this inode. */
1102 vi->i_op = &ntfs_file_inode_ops;
1103 vi->i_fop = &ntfs_file_ops;
1104 }
1105 if (NInoMstProtected(ni))
1106 vi->i_mapping->a_ops = &ntfs_mst_aops;
1107 else
1108 vi->i_mapping->a_ops = &ntfs_aops;
1109 /*
1110 * The number of 512-byte blocks used on disk (for stat). This is in so
1111 * far inaccurate as it doesn't account for any named streams or other
1112 * special non-resident attributes, but that is how Windows works, too,
1113 * so we are at least consistent with Windows, if not entirely
1114 * consistent with the Linux Way. Doing it the Linux Way would cause a
1115 * significant slowdown as it would involve iterating over all
1116 * attributes in the mft record and adding the allocated/compressed
1117 * sizes of all non-resident attributes present to give us the Linux
1118 * correct size that should go into i_blocks (after division by 512).
1119 */
1120 if (S_ISDIR(vi->i_mode) || !NInoCompressed(ni))
1121 vi->i_blocks = ni->allocated_size >> 9;
1122 else
1123 vi->i_blocks = ni->itype.compressed.size >> 9;
1124
1125 ntfs_debug("Done.");
1126 return 0;
1127
1128unm_err_out:
1129 if (!err)
1130 err = -EIO;
1131 if (ctx)
1132 ntfs_attr_put_search_ctx(ctx);
1133 if (m)
1134 unmap_mft_record(ni);
1135err_out:
1136 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1137 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1138 make_bad_inode(vi);
1139 if (err != -EOPNOTSUPP && err != -ENOMEM)
1140 NVolSetErrors(vol);
1141 return err;
1142}
1143
1144/**
1145 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1146 * @base_vi: base inode
1147 * @vi: attribute inode to read
1148 *
1149 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1150 * attribute inode described by @vi into memory from the base mft record
1151 * described by @base_ni.
1152 *
1153 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1154 * reading and looks up the attribute described by @vi before setting up the
1155 * necessary fields in @vi as well as initializing the ntfs inode.
1156 *
1157 * Q: What locks are held when the function is called?
1158 * A: i_state has I_LOCK set, hence the inode is locked, also
1159 * i_count is set to 1, so it is not going to go away
1160 *
1161 * Return 0 on success and -errno on error. In the error case, the inode will
1162 * have had make_bad_inode() executed on it.
1163 */
1164static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1165{
1166 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1167 ntfs_inode *ni, *base_ni;
1168 MFT_RECORD *m;
1169 ntfs_attr_search_ctx *ctx;
1170 int err = 0;
1171
1172 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1173
1174 ntfs_init_big_inode(vi);
1175
1176 ni = NTFS_I(vi);
1177 base_ni = NTFS_I(base_vi);
1178
1179 /* Just mirror the values from the base inode. */
1180 vi->i_blksize = base_vi->i_blksize;
1181 vi->i_version = base_vi->i_version;
1182 vi->i_uid = base_vi->i_uid;
1183 vi->i_gid = base_vi->i_gid;
1184 vi->i_nlink = base_vi->i_nlink;
1185 vi->i_mtime = base_vi->i_mtime;
1186 vi->i_ctime = base_vi->i_ctime;
1187 vi->i_atime = base_vi->i_atime;
1188 vi->i_generation = ni->seq_no = base_ni->seq_no;
1189
1190 /* Set inode type to zero but preserve permissions. */
1191 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1192
1193 m = map_mft_record(base_ni);
1194 if (IS_ERR(m)) {
1195 err = PTR_ERR(m);
1196 goto err_out;
1197 }
1198 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1199 if (!ctx) {
1200 err = -ENOMEM;
1201 goto unm_err_out;
1202 }
1203
1204 /* Find the attribute. */
1205 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1206 CASE_SENSITIVE, 0, NULL, 0, ctx);
1207 if (unlikely(err))
1208 goto unm_err_out;
1209
1210 if (!ctx->attr->non_resident) {
1211 /* Ensure the attribute name is placed before the value. */
1212 if (unlikely(ctx->attr->name_length &&
1213 (le16_to_cpu(ctx->attr->name_offset) >=
1214 le16_to_cpu(ctx->attr->data.resident.
1215 value_offset)))) {
1216 ntfs_error(vol->sb, "Attribute name is placed after "
1217 "the attribute value.");
1218 goto unm_err_out;
1219 }
1220 if (NInoMstProtected(ni) || ctx->attr->flags) {
1221 ntfs_error(vi->i_sb, "Found mst protected attribute "
1222 "or attribute with non-zero flags but "
1223 "the attribute is resident. Please "
1224 "report you saw this message to "
1225 "linux-ntfs-dev@lists.sourceforge.net");
1226 goto unm_err_out;
1227 }
1228 /*
1229 * Resident attribute. Make all sizes equal for simplicity in
1230 * read code paths.
1231 */
1232 vi->i_size = ni->initialized_size = ni->allocated_size =
1233 le32_to_cpu(ctx->attr->data.resident.value_length);
1234 } else {
1235 NInoSetNonResident(ni);
1236 /*
1237 * Ensure the attribute name is placed before the mapping pairs
1238 * array.
1239 */
1240 if (unlikely(ctx->attr->name_length &&
1241 (le16_to_cpu(ctx->attr->name_offset) >=
1242 le16_to_cpu(ctx->attr->data.non_resident.
1243 mapping_pairs_offset)))) {
1244 ntfs_error(vol->sb, "Attribute name is placed after "
1245 "the mapping pairs array.");
1246 goto unm_err_out;
1247 }
1248 if (ctx->attr->flags & ATTR_COMPRESSION_MASK) {
1249 if (NInoMstProtected(ni)) {
1250 ntfs_error(vi->i_sb, "Found mst protected "
1251 "attribute but the attribute "
1252 "is compressed. Please report "
1253 "you saw this message to "
1254 "linux-ntfs-dev@lists."
1255 "sourceforge.net");
1256 goto unm_err_out;
1257 }
1258 NInoSetCompressed(ni);
1259 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1260 ni->name_len)) {
1261 ntfs_error(vi->i_sb, "Found compressed "
1262 "non-data or named data "
1263 "attribute. Please report "
1264 "you saw this message to "
1265 "linux-ntfs-dev@lists."
1266 "sourceforge.net");
1267 goto unm_err_out;
1268 }
1269 if (vol->cluster_size > 4096) {
1270 ntfs_error(vi->i_sb, "Found compressed "
1271 "attribute but compression is "
1272 "disabled due to cluster size "
1273 "(%i) > 4kiB.",
1274 vol->cluster_size);
1275 goto unm_err_out;
1276 }
1277 if ((ctx->attr->flags & ATTR_COMPRESSION_MASK)
1278 != ATTR_IS_COMPRESSED) {
1279 ntfs_error(vi->i_sb, "Found unknown "
1280 "compression method.");
1281 goto unm_err_out;
1282 }
1283 ni->itype.compressed.block_clusters = 1U <<
1284 ctx->attr->data.non_resident.
1285 compression_unit;
1286 if (ctx->attr->data.non_resident.compression_unit !=
1287 4) {
1288 ntfs_error(vi->i_sb, "Found nonstandard "
1289 "compression unit (%u instead "
1290 "of 4). Cannot handle this.",
1291 ctx->attr->data.non_resident.
1292 compression_unit);
1293 err = -EOPNOTSUPP;
1294 goto unm_err_out;
1295 }
1296 ni->itype.compressed.block_size = 1U << (
1297 ctx->attr->data.non_resident.
1298 compression_unit +
1299 vol->cluster_size_bits);
1300 ni->itype.compressed.block_size_bits = ffs(
1301 ni->itype.compressed.block_size) - 1;
1302 }
1303 if (ctx->attr->flags & ATTR_IS_ENCRYPTED) {
1304 if (ctx->attr->flags & ATTR_COMPRESSION_MASK) {
1305 ntfs_error(vi->i_sb, "Found encrypted "
1306 "and compressed data.");
1307 goto unm_err_out;
1308 }
1309 if (NInoMstProtected(ni)) {
1310 ntfs_error(vi->i_sb, "Found mst protected "
1311 "attribute but the attribute "
1312 "is encrypted. Please report "
1313 "you saw this message to "
1314 "linux-ntfs-dev@lists."
1315 "sourceforge.net");
1316 goto unm_err_out;
1317 }
1318 NInoSetEncrypted(ni);
1319 }
1320 if (ctx->attr->flags & ATTR_IS_SPARSE) {
1321 if (NInoMstProtected(ni)) {
1322 ntfs_error(vi->i_sb, "Found mst protected "
1323 "attribute but the attribute "
1324 "is sparse. Please report "
1325 "you saw this message to "
1326 "linux-ntfs-dev@lists."
1327 "sourceforge.net");
1328 goto unm_err_out;
1329 }
1330 NInoSetSparse(ni);
1331 }
1332 if (ctx->attr->data.non_resident.lowest_vcn) {
1333 ntfs_error(vi->i_sb, "First extent of attribute has "
1334 "non-zero lowest_vcn.");
1335 goto unm_err_out;
1336 }
1337 /* Setup all the sizes. */
1338 vi->i_size = sle64_to_cpu(
1339 ctx->attr->data.non_resident.data_size);
1340 ni->initialized_size = sle64_to_cpu(
1341 ctx->attr->data.non_resident.initialized_size);
1342 ni->allocated_size = sle64_to_cpu(
1343 ctx->attr->data.non_resident.allocated_size);
1344 if (NInoCompressed(ni)) {
1345 ni->itype.compressed.size = sle64_to_cpu(
1346 ctx->attr->data.non_resident.
1347 compressed_size);
1348 }
1349 }
1350
1351 /* Setup the operations for this attribute inode. */
1352 vi->i_op = NULL;
1353 vi->i_fop = NULL;
1354 if (NInoMstProtected(ni))
1355 vi->i_mapping->a_ops = &ntfs_mst_aops;
1356 else
1357 vi->i_mapping->a_ops = &ntfs_aops;
1358
1359 if (!NInoCompressed(ni))
1360 vi->i_blocks = ni->allocated_size >> 9;
1361 else
1362 vi->i_blocks = ni->itype.compressed.size >> 9;
1363
1364 /*
1365 * Make sure the base inode doesn't go away and attach it to the
1366 * attribute inode.
1367 */
1368 igrab(base_vi);
1369 ni->ext.base_ntfs_ino = base_ni;
1370 ni->nr_extents = -1;
1371
1372 ntfs_attr_put_search_ctx(ctx);
1373 unmap_mft_record(base_ni);
1374
1375 ntfs_debug("Done.");
1376 return 0;
1377
1378unm_err_out:
1379 if (!err)
1380 err = -EIO;
1381 if (ctx)
1382 ntfs_attr_put_search_ctx(ctx);
1383 unmap_mft_record(base_ni);
1384err_out:
1385 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1386 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1387 "Marking corrupt inode and base inode 0x%lx as bad. "
1388 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1389 base_vi->i_ino);
1390 make_bad_inode(vi);
1391 make_bad_inode(base_vi);
1392 if (err != -ENOMEM)
1393 NVolSetErrors(vol);
1394 return err;
1395}
1396
1397/**
1398 * ntfs_read_locked_index_inode - read an index inode from its base inode
1399 * @base_vi: base inode
1400 * @vi: index inode to read
1401 *
1402 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1403 * index inode described by @vi into memory from the base mft record described
1404 * by @base_ni.
1405 *
1406 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1407 * reading and looks up the attributes relating to the index described by @vi
1408 * before setting up the necessary fields in @vi as well as initializing the
1409 * ntfs inode.
1410 *
1411 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1412 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1413 * are setup like directory inodes since directories are a special case of
1414 * indices ao they need to be treated in much the same way. Most importantly,
1415 * for small indices the index allocation attribute might not actually exist.
1416 * However, the index root attribute always exists but this does not need to
1417 * have an inode associated with it and this is why we define a new inode type
1418 * index. Also, like for directories, we need to have an attribute inode for
1419 * the bitmap attribute corresponding to the index allocation attribute and we
1420 * can store this in the appropriate field of the inode, just like we do for
1421 * normal directory inodes.
1422 *
1423 * Q: What locks are held when the function is called?
1424 * A: i_state has I_LOCK set, hence the inode is locked, also
1425 * i_count is set to 1, so it is not going to go away
1426 *
1427 * Return 0 on success and -errno on error. In the error case, the inode will
1428 * have had make_bad_inode() executed on it.
1429 */
1430static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1431{
1432 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1433 ntfs_inode *ni, *base_ni, *bni;
1434 struct inode *bvi;
1435 MFT_RECORD *m;
1436 ntfs_attr_search_ctx *ctx;
1437 INDEX_ROOT *ir;
1438 u8 *ir_end, *index_end;
1439 int err = 0;
1440
1441 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1442 ntfs_init_big_inode(vi);
1443 ni = NTFS_I(vi);
1444 base_ni = NTFS_I(base_vi);
1445 /* Just mirror the values from the base inode. */
1446 vi->i_blksize = base_vi->i_blksize;
1447 vi->i_version = base_vi->i_version;
1448 vi->i_uid = base_vi->i_uid;
1449 vi->i_gid = base_vi->i_gid;
1450 vi->i_nlink = base_vi->i_nlink;
1451 vi->i_mtime = base_vi->i_mtime;
1452 vi->i_ctime = base_vi->i_ctime;
1453 vi->i_atime = base_vi->i_atime;
1454 vi->i_generation = ni->seq_no = base_ni->seq_no;
1455 /* Set inode type to zero but preserve permissions. */
1456 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1457 /* Map the mft record for the base inode. */
1458 m = map_mft_record(base_ni);
1459 if (IS_ERR(m)) {
1460 err = PTR_ERR(m);
1461 goto err_out;
1462 }
1463 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1464 if (!ctx) {
1465 err = -ENOMEM;
1466 goto unm_err_out;
1467 }
1468 /* Find the index root attribute. */
1469 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1470 CASE_SENSITIVE, 0, NULL, 0, ctx);
1471 if (unlikely(err)) {
1472 if (err == -ENOENT)
1473 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1474 "missing.");
1475 goto unm_err_out;
1476 }
1477 /* Set up the state. */
1478 if (unlikely(ctx->attr->non_resident)) {
1479 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1480 goto unm_err_out;
1481 }
1482 /* Ensure the attribute name is placed before the value. */
1483 if (unlikely(ctx->attr->name_length &&
1484 (le16_to_cpu(ctx->attr->name_offset) >=
1485 le16_to_cpu(ctx->attr->data.resident.
1486 value_offset)))) {
1487 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1488 "after the attribute value.");
1489 goto unm_err_out;
1490 }
1491 /* Compressed/encrypted/sparse index root is not allowed. */
1492 if (ctx->attr->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1493 ATTR_IS_SPARSE)) {
1494 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1495 "root attribute.");
1496 goto unm_err_out;
1497 }
1498 ir = (INDEX_ROOT*)((u8*)ctx->attr +
1499 le16_to_cpu(ctx->attr->data.resident.value_offset));
1500 ir_end = (u8*)ir + le32_to_cpu(ctx->attr->data.resident.value_length);
1501 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1502 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1503 goto unm_err_out;
1504 }
1505 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1506 if (index_end > ir_end) {
1507 ntfs_error(vi->i_sb, "Index is corrupt.");
1508 goto unm_err_out;
1509 }
1510 if (ir->type) {
1511 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1512 le32_to_cpu(ir->type));
1513 goto unm_err_out;
1514 }
1515 ni->itype.index.collation_rule = ir->collation_rule;
1516 ntfs_debug("Index collation rule is 0x%x.",
1517 le32_to_cpu(ir->collation_rule));
1518 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1519 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1520 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1521 "two.", ni->itype.index.block_size);
1522 goto unm_err_out;
1523 }
1524 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1525 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1526 "(%ld) is not supported. Sorry.",
1527 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1528 err = -EOPNOTSUPP;
1529 goto unm_err_out;
1530 }
1531 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1532 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1533 "(%i) is not supported. Sorry.",
1534 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1535 err = -EOPNOTSUPP;
1536 goto unm_err_out;
1537 }
1538 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1539 /* Determine the size of a vcn in the index. */
1540 if (vol->cluster_size <= ni->itype.index.block_size) {
1541 ni->itype.index.vcn_size = vol->cluster_size;
1542 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1543 } else {
1544 ni->itype.index.vcn_size = vol->sector_size;
1545 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1546 }
1547 /* Check for presence of index allocation attribute. */
1548 if (!(ir->index.flags & LARGE_INDEX)) {
1549 /* No index allocation. */
1550 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1551 /* We are done with the mft record, so we release it. */
1552 ntfs_attr_put_search_ctx(ctx);
1553 unmap_mft_record(base_ni);
1554 m = NULL;
1555 ctx = NULL;
1556 goto skip_large_index_stuff;
1557 } /* LARGE_INDEX: Index allocation present. Setup state. */
1558 NInoSetIndexAllocPresent(ni);
1559 /* Find index allocation attribute. */
1560 ntfs_attr_reinit_search_ctx(ctx);
1561 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1562 CASE_SENSITIVE, 0, NULL, 0, ctx);
1563 if (unlikely(err)) {
1564 if (err == -ENOENT)
1565 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1566 "not present but $INDEX_ROOT "
1567 "indicated it is.");
1568 else
1569 ntfs_error(vi->i_sb, "Failed to lookup "
1570 "$INDEX_ALLOCATION attribute.");
1571 goto unm_err_out;
1572 }
1573 if (!ctx->attr->non_resident) {
1574 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1575 "resident.");
1576 goto unm_err_out;
1577 }
1578 /*
1579 * Ensure the attribute name is placed before the mapping pairs array.
1580 */
1581 if (unlikely(ctx->attr->name_length && (le16_to_cpu(
1582 ctx->attr->name_offset) >= le16_to_cpu(
1583 ctx->attr->data.non_resident.mapping_pairs_offset)))) {
1584 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1585 "placed after the mapping pairs array.");
1586 goto unm_err_out;
1587 }
1588 if (ctx->attr->flags & ATTR_IS_ENCRYPTED) {
1589 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1590 "encrypted.");
1591 goto unm_err_out;
1592 }
1593 if (ctx->attr->flags & ATTR_IS_SPARSE) {
1594 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1595 goto unm_err_out;
1596 }
1597 if (ctx->attr->flags & ATTR_COMPRESSION_MASK) {
1598 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1599 "compressed.");
1600 goto unm_err_out;
1601 }
1602 if (ctx->attr->data.non_resident.lowest_vcn) {
1603 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1604 "attribute has non zero lowest_vcn.");
1605 goto unm_err_out;
1606 }
1607 vi->i_size = sle64_to_cpu(ctx->attr->data.non_resident.data_size);
1608 ni->initialized_size = sle64_to_cpu(
1609 ctx->attr->data.non_resident.initialized_size);
1610 ni->allocated_size = sle64_to_cpu(
1611 ctx->attr->data.non_resident.allocated_size);
1612 /*
1613 * We are done with the mft record, so we release it. Otherwise
1614 * we would deadlock in ntfs_attr_iget().
1615 */
1616 ntfs_attr_put_search_ctx(ctx);
1617 unmap_mft_record(base_ni);
1618 m = NULL;
1619 ctx = NULL;
1620 /* Get the index bitmap attribute inode. */
1621 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1622 if (IS_ERR(bvi)) {
1623 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1624 err = PTR_ERR(bvi);
1625 goto unm_err_out;
1626 }
1627 bni = NTFS_I(bvi);
1628 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1629 NInoSparse(bni)) {
1630 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1631 "encrypted and/or sparse.");
1632 goto iput_unm_err_out;
1633 }
1634 /* Consistency check bitmap size vs. index allocation size. */
1635 if ((bvi->i_size << 3) < (vi->i_size >>
1636 ni->itype.index.block_size_bits)) {
1637 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1638 "index allocation (0x%llx).", bvi->i_size << 3,
1639 vi->i_size);
1640 goto iput_unm_err_out;
1641 }
1642 ni->itype.index.bmp_ino = bvi;
1643skip_large_index_stuff:
1644 /* Setup the operations for this index inode. */
1645 vi->i_op = NULL;
1646 vi->i_fop = NULL;
1647 vi->i_mapping->a_ops = &ntfs_mst_aops;
1648 vi->i_blocks = ni->allocated_size >> 9;
1649
1650 /*
1651 * Make sure the base inode doesn't go away and attach it to the
1652 * index inode.
1653 */
1654 igrab(base_vi);
1655 ni->ext.base_ntfs_ino = base_ni;
1656 ni->nr_extents = -1;
1657
1658 ntfs_debug("Done.");
1659 return 0;
1660
1661iput_unm_err_out:
1662 iput(bvi);
1663unm_err_out:
1664 if (!err)
1665 err = -EIO;
1666 if (ctx)
1667 ntfs_attr_put_search_ctx(ctx);
1668 if (m)
1669 unmap_mft_record(base_ni);
1670err_out:
1671 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1672 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1673 ni->name_len);
1674 make_bad_inode(vi);
1675 if (err != -EOPNOTSUPP && err != -ENOMEM)
1676 NVolSetErrors(vol);
1677 return err;
1678}
1679
1680/**
1681 * ntfs_read_inode_mount - special read_inode for mount time use only
1682 * @vi: inode to read
1683 *
1684 * Read inode FILE_MFT at mount time, only called with super_block lock
1685 * held from within the read_super() code path.
1686 *
1687 * This function exists because when it is called the page cache for $MFT/$DATA
1688 * is not initialized and hence we cannot get at the contents of mft records
1689 * by calling map_mft_record*().
1690 *
1691 * Further it needs to cope with the circular references problem, i.e. cannot
1692 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1693 * we do not know where the other extent mft records are yet and again, because
1694 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1695 * attribute list is actually present in $MFT inode.
1696 *
1697 * We solve these problems by starting with the $DATA attribute before anything
1698 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1699 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1700 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1701 * sufficient information for the next step to complete.
1702 *
1703 * This should work but there are two possible pit falls (see inline comments
1704 * below), but only time will tell if they are real pits or just smoke...
1705 */
1706int ntfs_read_inode_mount(struct inode *vi)
1707{
1708 VCN next_vcn, last_vcn, highest_vcn;
1709 s64 block;
1710 struct super_block *sb = vi->i_sb;
1711 ntfs_volume *vol = NTFS_SB(sb);
1712 struct buffer_head *bh;
1713 ntfs_inode *ni;
1714 MFT_RECORD *m = NULL;
1715 ATTR_RECORD *attr;
1716 ntfs_attr_search_ctx *ctx;
1717 unsigned int i, nr_blocks;
1718 int err;
1719
1720 ntfs_debug("Entering.");
1721
1722 /* Initialize the ntfs specific part of @vi. */
1723 ntfs_init_big_inode(vi);
1724
1725 ni = NTFS_I(vi);
1726
1727 /* Setup the data attribute. It is special as it is mst protected. */
1728 NInoSetNonResident(ni);
1729 NInoSetMstProtected(ni);
1730 ni->type = AT_DATA;
1731 ni->name = NULL;
1732 ni->name_len = 0;
1733
1734 /*
1735 * This sets up our little cheat allowing us to reuse the async read io
1736 * completion handler for directories.
1737 */
1738 ni->itype.index.block_size = vol->mft_record_size;
1739 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1740
1741 /* Very important! Needed to be able to call map_mft_record*(). */
1742 vol->mft_ino = vi;
1743
1744 /* Allocate enough memory to read the first mft record. */
1745 if (vol->mft_record_size > 64 * 1024) {
1746 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1747 vol->mft_record_size);
1748 goto err_out;
1749 }
1750 i = vol->mft_record_size;
1751 if (i < sb->s_blocksize)
1752 i = sb->s_blocksize;
1753 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1754 if (!m) {
1755 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1756 goto err_out;
1757 }
1758
1759 /* Determine the first block of the $MFT/$DATA attribute. */
1760 block = vol->mft_lcn << vol->cluster_size_bits >>
1761 sb->s_blocksize_bits;
1762 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1763 if (!nr_blocks)
1764 nr_blocks = 1;
1765
1766 /* Load $MFT/$DATA's first mft record. */
1767 for (i = 0; i < nr_blocks; i++) {
1768 bh = sb_bread(sb, block++);
1769 if (!bh) {
1770 ntfs_error(sb, "Device read failed.");
1771 goto err_out;
1772 }
1773 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1774 sb->s_blocksize);
1775 brelse(bh);
1776 }
1777
1778 /* Apply the mst fixups. */
1779 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1780 /* FIXME: Try to use the $MFTMirr now. */
1781 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1782 goto err_out;
1783 }
1784
1785 /* Need this to sanity check attribute list references to $MFT. */
1786 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1787
1788 /* Provides readpage() and sync_page() for map_mft_record(). */
1789 vi->i_mapping->a_ops = &ntfs_mst_aops;
1790
1791 ctx = ntfs_attr_get_search_ctx(ni, m);
1792 if (!ctx) {
1793 err = -ENOMEM;
1794 goto err_out;
1795 }
1796
1797 /* Find the attribute list attribute if present. */
1798 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1799 if (err) {
1800 if (unlikely(err != -ENOENT)) {
1801 ntfs_error(sb, "Failed to lookup attribute list "
1802 "attribute. You should run chkdsk.");
1803 goto put_err_out;
1804 }
1805 } else /* if (!err) */ {
1806 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1807 u8 *al_end;
1808
1809 ntfs_debug("Attribute list attribute found in $MFT.");
1810 NInoSetAttrList(ni);
1811 if (ctx->attr->flags & ATTR_IS_ENCRYPTED ||
1812 ctx->attr->flags & ATTR_COMPRESSION_MASK ||
1813 ctx->attr->flags & ATTR_IS_SPARSE) {
1814 ntfs_error(sb, "Attribute list attribute is "
1815 "compressed/encrypted/sparse. Not "
1816 "allowed. $MFT is corrupt. You should "
1817 "run chkdsk.");
1818 goto put_err_out;
1819 }
1820 /* Now allocate memory for the attribute list. */
1821 ni->attr_list_size = (u32)ntfs_attr_size(ctx->attr);
1822 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1823 if (!ni->attr_list) {
1824 ntfs_error(sb, "Not enough memory to allocate buffer "
1825 "for attribute list.");
1826 goto put_err_out;
1827 }
1828 if (ctx->attr->non_resident) {
1829 NInoSetAttrListNonResident(ni);
1830 if (ctx->attr->data.non_resident.lowest_vcn) {
1831 ntfs_error(sb, "Attribute list has non zero "
1832 "lowest_vcn. $MFT is corrupt. "
1833 "You should run chkdsk.");
1834 goto put_err_out;
1835 }
1836 /* Setup the runlist. */
1837 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1838 ctx->attr, NULL);
1839 if (IS_ERR(ni->attr_list_rl.rl)) {
1840 err = PTR_ERR(ni->attr_list_rl.rl);
1841 ni->attr_list_rl.rl = NULL;
1842 ntfs_error(sb, "Mapping pairs decompression "
1843 "failed with error code %i.",
1844 -err);
1845 goto put_err_out;
1846 }
1847 /* Now load the attribute list. */
1848 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1849 ni->attr_list, ni->attr_list_size,
1850 sle64_to_cpu(ctx->attr->data.
1851 non_resident.initialized_size)))) {
1852 ntfs_error(sb, "Failed to load attribute list "
1853 "attribute with error code %i.",
1854 -err);
1855 goto put_err_out;
1856 }
1857 } else /* if (!ctx.attr->non_resident) */ {
1858 if ((u8*)ctx->attr + le16_to_cpu(
1859 ctx->attr->data.resident.value_offset) +
1860 le32_to_cpu(
1861 ctx->attr->data.resident.value_length) >
1862 (u8*)ctx->mrec + vol->mft_record_size) {
1863 ntfs_error(sb, "Corrupt attribute list "
1864 "attribute.");
1865 goto put_err_out;
1866 }
1867 /* Now copy the attribute list. */
1868 memcpy(ni->attr_list, (u8*)ctx->attr + le16_to_cpu(
1869 ctx->attr->data.resident.value_offset),
1870 le32_to_cpu(
1871 ctx->attr->data.resident.value_length));
1872 }
1873 /* The attribute list is now setup in memory. */
1874 /*
1875 * FIXME: I don't know if this case is actually possible.
1876 * According to logic it is not possible but I have seen too
1877 * many weird things in MS software to rely on logic... Thus we
1878 * perform a manual search and make sure the first $MFT/$DATA
1879 * extent is in the base inode. If it is not we abort with an
1880 * error and if we ever see a report of this error we will need
1881 * to do some magic in order to have the necessary mft record
1882 * loaded and in the right place in the page cache. But
1883 * hopefully logic will prevail and this never happens...
1884 */
1885 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1886 al_end = (u8*)al_entry + ni->attr_list_size;
1887 for (;; al_entry = next_al_entry) {
1888 /* Out of bounds check. */
1889 if ((u8*)al_entry < ni->attr_list ||
1890 (u8*)al_entry > al_end)
1891 goto em_put_err_out;
1892 /* Catch the end of the attribute list. */
1893 if ((u8*)al_entry == al_end)
1894 goto em_put_err_out;
1895 if (!al_entry->length)
1896 goto em_put_err_out;
1897 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1898 le16_to_cpu(al_entry->length) > al_end)
1899 goto em_put_err_out;
1900 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1901 le16_to_cpu(al_entry->length));
1902 if (le32_to_cpu(al_entry->type) >
1903 const_le32_to_cpu(AT_DATA))
1904 goto em_put_err_out;
1905 if (AT_DATA != al_entry->type)
1906 continue;
1907 /* We want an unnamed attribute. */
1908 if (al_entry->name_length)
1909 goto em_put_err_out;
1910 /* Want the first entry, i.e. lowest_vcn == 0. */
1911 if (al_entry->lowest_vcn)
1912 goto em_put_err_out;
1913 /* First entry has to be in the base mft record. */
1914 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1915 /* MFT references do not match, logic fails. */
1916 ntfs_error(sb, "BUG: The first $DATA extent "
1917 "of $MFT is not in the base "
1918 "mft record. Please report "
1919 "you saw this message to "
1920 "linux-ntfs-dev@lists."
1921 "sourceforge.net");
1922 goto put_err_out;
1923 } else {
1924 /* Sequence numbers must match. */
1925 if (MSEQNO_LE(al_entry->mft_reference) !=
1926 ni->seq_no)
1927 goto em_put_err_out;
1928 /* Got it. All is ok. We can stop now. */
1929 break;
1930 }
1931 }
1932 }
1933
1934 ntfs_attr_reinit_search_ctx(ctx);
1935
1936 /* Now load all attribute extents. */
1937 attr = NULL;
1938 next_vcn = last_vcn = highest_vcn = 0;
1939 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
1940 ctx))) {
1941 runlist_element *nrl;
1942
1943 /* Cache the current attribute. */
1944 attr = ctx->attr;
1945 /* $MFT must be non-resident. */
1946 if (!attr->non_resident) {
1947 ntfs_error(sb, "$MFT must be non-resident but a "
1948 "resident extent was found. $MFT is "
1949 "corrupt. Run chkdsk.");
1950 goto put_err_out;
1951 }
1952 /* $MFT must be uncompressed and unencrypted. */
1953 if (attr->flags & ATTR_COMPRESSION_MASK ||
1954 attr->flags & ATTR_IS_ENCRYPTED ||
1955 attr->flags & ATTR_IS_SPARSE) {
1956 ntfs_error(sb, "$MFT must be uncompressed, "
1957 "non-sparse, and unencrypted but a "
1958 "compressed/sparse/encrypted extent "
1959 "was found. $MFT is corrupt. Run "
1960 "chkdsk.");
1961 goto put_err_out;
1962 }
1963 /*
1964 * Decompress the mapping pairs array of this extent and merge
1965 * the result into the existing runlist. No need for locking
1966 * as we have exclusive access to the inode at this time and we
1967 * are a mount in progress task, too.
1968 */
1969 nrl = ntfs_mapping_pairs_decompress(vol, attr, ni->runlist.rl);
1970 if (IS_ERR(nrl)) {
1971 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
1972 "failed with error code %ld. $MFT is "
1973 "corrupt.", PTR_ERR(nrl));
1974 goto put_err_out;
1975 }
1976 ni->runlist.rl = nrl;
1977
1978 /* Are we in the first extent? */
1979 if (!next_vcn) {
1980 if (attr->data.non_resident.lowest_vcn) {
1981 ntfs_error(sb, "First extent of $DATA "
1982 "attribute has non zero "
1983 "lowest_vcn. $MFT is corrupt. "
1984 "You should run chkdsk.");
1985 goto put_err_out;
1986 }
1987 /* Get the last vcn in the $DATA attribute. */
1988 last_vcn = sle64_to_cpu(
1989 attr->data.non_resident.allocated_size)
1990 >> vol->cluster_size_bits;
1991 /* Fill in the inode size. */
1992 vi->i_size = sle64_to_cpu(
1993 attr->data.non_resident.data_size);
1994 ni->initialized_size = sle64_to_cpu(attr->data.
1995 non_resident.initialized_size);
1996 ni->allocated_size = sle64_to_cpu(
1997 attr->data.non_resident.allocated_size);
1998 /*
1999 * Verify the number of mft records does not exceed
2000 * 2^32 - 1.
2001 */
2002 if ((vi->i_size >> vol->mft_record_size_bits) >=
2003 (1ULL << 32)) {
2004 ntfs_error(sb, "$MFT is too big! Aborting.");
2005 goto put_err_out;
2006 }
2007 /*
2008 * We have got the first extent of the runlist for
2009 * $MFT which means it is now relatively safe to call
2010 * the normal ntfs_read_inode() function.
2011 * Complete reading the inode, this will actually
2012 * re-read the mft record for $MFT, this time entering
2013 * it into the page cache with which we complete the
2014 * kick start of the volume. It should be safe to do
2015 * this now as the first extent of $MFT/$DATA is
2016 * already known and we would hope that we don't need
2017 * further extents in order to find the other
2018 * attributes belonging to $MFT. Only time will tell if
2019 * this is really the case. If not we will have to play
2020 * magic at this point, possibly duplicating a lot of
2021 * ntfs_read_inode() at this point. We will need to
2022 * ensure we do enough of its work to be able to call
2023 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2024 * hope this never happens...
2025 */
2026 ntfs_read_locked_inode(vi);
2027 if (is_bad_inode(vi)) {
2028 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2029 "failed. BUG or corrupt $MFT. "
2030 "Run chkdsk and if no errors "
2031 "are found, please report you "
2032 "saw this message to "
2033 "linux-ntfs-dev@lists."
2034 "sourceforge.net");
2035 ntfs_attr_put_search_ctx(ctx);
2036 /* Revert to the safe super operations. */
2037 ntfs_free(m);
2038 return -1;
2039 }
2040 /*
2041 * Re-initialize some specifics about $MFT's inode as
2042 * ntfs_read_inode() will have set up the default ones.
2043 */
2044 /* Set uid and gid to root. */
2045 vi->i_uid = vi->i_gid = 0;
2046 /* Regular file. No access for anyone. */
2047 vi->i_mode = S_IFREG;
2048 /* No VFS initiated operations allowed for $MFT. */
2049 vi->i_op = &ntfs_empty_inode_ops;
2050 vi->i_fop = &ntfs_empty_file_ops;
2051 }
2052
2053 /* Get the lowest vcn for the next extent. */
2054 highest_vcn = sle64_to_cpu(attr->data.non_resident.highest_vcn);
2055 next_vcn = highest_vcn + 1;
2056
2057 /* Only one extent or error, which we catch below. */
2058 if (next_vcn <= 0)
2059 break;
2060
2061 /* Avoid endless loops due to corruption. */
2062 if (next_vcn < sle64_to_cpu(
2063 attr->data.non_resident.lowest_vcn)) {
2064 ntfs_error(sb, "$MFT has corrupt attribute list "
2065 "attribute. Run chkdsk.");
2066 goto put_err_out;
2067 }
2068 }
2069 if (err != -ENOENT) {
2070 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2071 "$MFT is corrupt. Run chkdsk.");
2072 goto put_err_out;
2073 }
2074 if (!attr) {
2075 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2076 "corrupt. Run chkdsk.");
2077 goto put_err_out;
2078 }
2079 if (highest_vcn && highest_vcn != last_vcn - 1) {
2080 ntfs_error(sb, "Failed to load the complete runlist for "
2081 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2082 "Run chkdsk.");
2083 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2084 (unsigned long long)highest_vcn,
2085 (unsigned long long)last_vcn - 1);
2086 goto put_err_out;
2087 }
2088 ntfs_attr_put_search_ctx(ctx);
2089 ntfs_debug("Done.");
2090 ntfs_free(m);
2091 return 0;
2092
2093em_put_err_out:
2094 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2095 "attribute list. $MFT is corrupt. Run chkdsk.");
2096put_err_out:
2097 ntfs_attr_put_search_ctx(ctx);
2098err_out:
2099 ntfs_error(sb, "Failed. Marking inode as bad.");
2100 make_bad_inode(vi);
2101 ntfs_free(m);
2102 return -1;
2103}
2104
2105/**
2106 * ntfs_put_inode - handler for when the inode reference count is decremented
2107 * @vi: vfs inode
2108 *
2109 * The VFS calls ntfs_put_inode() every time the inode reference count (i_count)
2110 * is about to be decremented (but before the decrement itself.
2111 *
2112 * If the inode @vi is a directory with two references, one of which is being
2113 * dropped, we need to put the attribute inode for the directory index bitmap,
2114 * if it is present, otherwise the directory inode would remain pinned for
2115 * ever.
2116 */
2117void ntfs_put_inode(struct inode *vi)
2118{
2119 if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
2120 ntfs_inode *ni = NTFS_I(vi);
2121 if (NInoIndexAllocPresent(ni)) {
2122 struct inode *bvi = NULL;
2123 down(&vi->i_sem);
2124 if (atomic_read(&vi->i_count) == 2) {
2125 bvi = ni->itype.index.bmp_ino;
2126 if (bvi)
2127 ni->itype.index.bmp_ino = NULL;
2128 }
2129 up(&vi->i_sem);
2130 if (bvi)
2131 iput(bvi);
2132 }
2133 }
2134}
2135
2136static void __ntfs_clear_inode(ntfs_inode *ni)
2137{
2138 /* Free all alocated memory. */
2139 down_write(&ni->runlist.lock);
2140 if (ni->runlist.rl) {
2141 ntfs_free(ni->runlist.rl);
2142 ni->runlist.rl = NULL;
2143 }
2144 up_write(&ni->runlist.lock);
2145
2146 if (ni->attr_list) {
2147 ntfs_free(ni->attr_list);
2148 ni->attr_list = NULL;
2149 }
2150
2151 down_write(&ni->attr_list_rl.lock);
2152 if (ni->attr_list_rl.rl) {
2153 ntfs_free(ni->attr_list_rl.rl);
2154 ni->attr_list_rl.rl = NULL;
2155 }
2156 up_write(&ni->attr_list_rl.lock);
2157
2158 if (ni->name_len && ni->name != I30) {
2159 /* Catch bugs... */
2160 BUG_ON(!ni->name);
2161 kfree(ni->name);
2162 }
2163}
2164
2165void ntfs_clear_extent_inode(ntfs_inode *ni)
2166{
2167 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2168
2169 BUG_ON(NInoAttr(ni));
2170 BUG_ON(ni->nr_extents != -1);
2171
2172#ifdef NTFS_RW
2173 if (NInoDirty(ni)) {
2174 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2175 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2176 "Losing data! This is a BUG!!!");
2177 // FIXME: Do something!!!
2178 }
2179#endif /* NTFS_RW */
2180
2181 __ntfs_clear_inode(ni);
2182
2183 /* Bye, bye... */
2184 ntfs_destroy_extent_inode(ni);
2185}
2186
2187/**
2188 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2189 * @vi: vfs inode pending annihilation
2190 *
2191 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2192 * is called, which deallocates all memory belonging to the NTFS specific part
2193 * of the inode and returns.
2194 *
2195 * If the MFT record is dirty, we commit it before doing anything else.
2196 */
2197void ntfs_clear_big_inode(struct inode *vi)
2198{
2199 ntfs_inode *ni = NTFS_I(vi);
2200
2201 /*
2202 * If the inode @vi is an index inode we need to put the attribute
2203 * inode for the index bitmap, if it is present, otherwise the index
2204 * inode would disappear and the attribute inode for the index bitmap
2205 * would no longer be referenced from anywhere and thus it would remain
2206 * pinned for ever.
2207 */
2208 if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) &&
2209 NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) {
2210 iput(ni->itype.index.bmp_ino);
2211 ni->itype.index.bmp_ino = NULL;
2212 }
2213#ifdef NTFS_RW
2214 if (NInoDirty(ni)) {
2215 BOOL was_bad = (is_bad_inode(vi));
2216
2217 /* Committing the inode also commits all extent inodes. */
2218 ntfs_commit_inode(vi);
2219
2220 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2221 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2222 "0x%lx. Losing data!", vi->i_ino);
2223 // FIXME: Do something!!!
2224 }
2225 }
2226#endif /* NTFS_RW */
2227
2228 /* No need to lock at this stage as no one else has a reference. */
2229 if (ni->nr_extents > 0) {
2230 int i;
2231
2232 for (i = 0; i < ni->nr_extents; i++)
2233 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2234 kfree(ni->ext.extent_ntfs_inos);
2235 }
2236
2237 __ntfs_clear_inode(ni);
2238
2239 if (NInoAttr(ni)) {
2240 /* Release the base inode if we are holding it. */
2241 if (ni->nr_extents == -1) {
2242 iput(VFS_I(ni->ext.base_ntfs_ino));
2243 ni->nr_extents = 0;
2244 ni->ext.base_ntfs_ino = NULL;
2245 }
2246 }
2247 return;
2248}
2249
2250/**
2251 * ntfs_show_options - show mount options in /proc/mounts
2252 * @sf: seq_file in which to write our mount options
2253 * @mnt: vfs mount whose mount options to display
2254 *
2255 * Called by the VFS once for each mounted ntfs volume when someone reads
2256 * /proc/mounts in order to display the NTFS specific mount options of each
2257 * mount. The mount options of the vfs mount @mnt are written to the seq file
2258 * @sf and success is returned.
2259 */
2260int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2261{
2262 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2263 int i;
2264
2265 seq_printf(sf, ",uid=%i", vol->uid);
2266 seq_printf(sf, ",gid=%i", vol->gid);
2267 if (vol->fmask == vol->dmask)
2268 seq_printf(sf, ",umask=0%o", vol->fmask);
2269 else {
2270 seq_printf(sf, ",fmask=0%o", vol->fmask);
2271 seq_printf(sf, ",dmask=0%o", vol->dmask);
2272 }
2273 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2274 if (NVolCaseSensitive(vol))
2275 seq_printf(sf, ",case_sensitive");
2276 if (NVolShowSystemFiles(vol))
2277 seq_printf(sf, ",show_sys_files");
2278 for (i = 0; on_errors_arr[i].val; i++) {
2279 if (on_errors_arr[i].val & vol->on_errors)
2280 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2281 }
2282 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2283 return 0;
2284}
2285
2286#ifdef NTFS_RW
2287
2288/**
2289 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2290 * @vi: inode for which the i_size was changed
2291 *
2292 * We do not support i_size changes yet.
2293 *
2294 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2295 * that the change is allowed.
2296 *
2297 * This implies for us that @vi is a file inode rather than a directory, index,
2298 * or attribute inode as well as that @vi is a base inode.
2299 *
2300 * Returns 0 on success or -errno on error.
2301 *
2302 * Called with ->i_sem held. In all but one case ->i_alloc_sem is held for
2303 * writing. The only case where ->i_alloc_sem is not held is
2304 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2305 * with the current i_size as the offset which means that it is a noop as far
2306 * as ntfs_truncate() is concerned.
2307 */
2308int ntfs_truncate(struct inode *vi)
2309{
2310 ntfs_inode *ni = NTFS_I(vi);
2311 ntfs_volume *vol = ni->vol;
2312 ntfs_attr_search_ctx *ctx;
2313 MFT_RECORD *m;
2314 const char *te = " Leaving file length out of sync with i_size.";
2315 int err;
2316
2317 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2318 BUG_ON(NInoAttr(ni));
2319 BUG_ON(ni->nr_extents < 0);
2320 m = map_mft_record(ni);
2321 if (IS_ERR(m)) {
2322 err = PTR_ERR(m);
2323 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2324 "(error code %d).%s", vi->i_ino, err, te);
2325 ctx = NULL;
2326 m = NULL;
2327 goto err_out;
2328 }
2329 ctx = ntfs_attr_get_search_ctx(ni, m);
2330 if (unlikely(!ctx)) {
2331 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2332 "inode 0x%lx (not enough memory).%s",
2333 vi->i_ino, te);
2334 err = -ENOMEM;
2335 goto err_out;
2336 }
2337 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2338 CASE_SENSITIVE, 0, NULL, 0, ctx);
2339 if (unlikely(err)) {
2340 if (err == -ENOENT)
2341 ntfs_error(vi->i_sb, "Open attribute is missing from "
2342 "mft record. Inode 0x%lx is corrupt. "
2343 "Run chkdsk.", vi->i_ino);
2344 else
2345 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2346 "inode 0x%lx (error code %d).",
2347 vi->i_ino, err);
2348 goto err_out;
2349 }
2350 /* If the size has not changed there is nothing to do. */
2351 if (ntfs_attr_size(ctx->attr) == i_size_read(vi))
2352 goto done;
2353 // TODO: Implement the truncate...
2354 ntfs_error(vi->i_sb, "Inode size has changed but this is not "
2355 "implemented yet. Resetting inode size to old value. "
2356 " This is most likely a bug in the ntfs driver!");
2357 i_size_write(vi, ntfs_attr_size(ctx->attr));
2358done:
2359 ntfs_attr_put_search_ctx(ctx);
2360 unmap_mft_record(ni);
2361 NInoClearTruncateFailed(ni);
2362 ntfs_debug("Done.");
2363 return 0;
2364err_out:
2365 if (err != -ENOMEM) {
2366 NVolSetErrors(vol);
2367 make_bad_inode(vi);
2368 }
2369 if (ctx)
2370 ntfs_attr_put_search_ctx(ctx);
2371 if (m)
2372 unmap_mft_record(ni);
2373 NInoSetTruncateFailed(ni);
2374 return err;
2375}
2376
2377/**
2378 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2379 * @vi: inode for which the i_size was changed
2380 *
2381 * Wrapper for ntfs_truncate() that has no return value.
2382 *
2383 * See ntfs_truncate() description above for details.
2384 */
2385void ntfs_truncate_vfs(struct inode *vi) {
2386 ntfs_truncate(vi);
2387}
2388
2389/**
2390 * ntfs_setattr - called from notify_change() when an attribute is being changed
2391 * @dentry: dentry whose attributes to change
2392 * @attr: structure describing the attributes and the changes
2393 *
2394 * We have to trap VFS attempts to truncate the file described by @dentry as
2395 * soon as possible, because we do not implement changes in i_size yet. So we
2396 * abort all i_size changes here.
2397 *
2398 * We also abort all changes of user, group, and mode as we do not implement
2399 * the NTFS ACLs yet.
2400 *
2401 * Called with ->i_sem held. For the ATTR_SIZE (i.e. ->truncate) case, also
2402 * called with ->i_alloc_sem held for writing.
2403 *
2404 * Basically this is a copy of generic notify_change() and inode_setattr()
2405 * functionality, except we intercept and abort changes in i_size.
2406 */
2407int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2408{
2409 struct inode *vi = dentry->d_inode;
2410 int err;
2411 unsigned int ia_valid = attr->ia_valid;
2412
2413 err = inode_change_ok(vi, attr);
2414 if (err)
2415 return err;
2416
2417 /* We do not support NTFS ACLs yet. */
2418 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2419 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2420 "supported yet, ignoring.");
2421 err = -EOPNOTSUPP;
2422 goto out;
2423 }
2424
2425 if (ia_valid & ATTR_SIZE) {
2426 if (attr->ia_size != i_size_read(vi)) {
2427 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2428 "supported yet, ignoring.");
2429 err = -EOPNOTSUPP;
2430 // TODO: Implement...
2431 // err = vmtruncate(vi, attr->ia_size);
2432 if (err || ia_valid == ATTR_SIZE)
2433 goto out;
2434 } else {
2435 /*
2436 * We skipped the truncate but must still update
2437 * timestamps.
2438 */
2439 ia_valid |= ATTR_MTIME|ATTR_CTIME;
2440 }
2441 }
2442
2443 if (ia_valid & ATTR_ATIME)
2444 vi->i_atime = attr->ia_atime;
2445 if (ia_valid & ATTR_MTIME)
2446 vi->i_mtime = attr->ia_mtime;
2447 if (ia_valid & ATTR_CTIME)
2448 vi->i_ctime = attr->ia_ctime;
2449 mark_inode_dirty(vi);
2450out:
2451 return err;
2452}
2453
2454/**
2455 * ntfs_write_inode - write out a dirty inode
2456 * @vi: inode to write out
2457 * @sync: if true, write out synchronously
2458 *
2459 * Write out a dirty inode to disk including any extent inodes if present.
2460 *
2461 * If @sync is true, commit the inode to disk and wait for io completion. This
2462 * is done using write_mft_record().
2463 *
2464 * If @sync is false, just schedule the write to happen but do not wait for i/o
2465 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2466 * marking the page (and in this case mft record) dirty but we do not implement
2467 * this yet as write_mft_record() largely ignores the @sync parameter and
2468 * always performs synchronous writes.
2469 *
2470 * Return 0 on success and -errno on error.
2471 */
2472int ntfs_write_inode(struct inode *vi, int sync)
2473{
2474 sle64 nt;
2475 ntfs_inode *ni = NTFS_I(vi);
2476 ntfs_attr_search_ctx *ctx;
2477 MFT_RECORD *m;
2478 STANDARD_INFORMATION *si;
2479 int err = 0;
2480 BOOL modified = FALSE;
2481
2482 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2483 vi->i_ino);
2484 /*
2485 * Dirty attribute inodes are written via their real inodes so just
2486 * clean them here. Access time updates are taken care off when the
2487 * real inode is written.
2488 */
2489 if (NInoAttr(ni)) {
2490 NInoClearDirty(ni);
2491 ntfs_debug("Done.");
2492 return 0;
2493 }
2494 /* Map, pin, and lock the mft record belonging to the inode. */
2495 m = map_mft_record(ni);
2496 if (IS_ERR(m)) {
2497 err = PTR_ERR(m);
2498 goto err_out;
2499 }
2500 /* Update the access times in the standard information attribute. */
2501 ctx = ntfs_attr_get_search_ctx(ni, m);
2502 if (unlikely(!ctx)) {
2503 err = -ENOMEM;
2504 goto unm_err_out;
2505 }
2506 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
2507 CASE_SENSITIVE, 0, NULL, 0, ctx);
2508 if (unlikely(err)) {
2509 ntfs_attr_put_search_ctx(ctx);
2510 goto unm_err_out;
2511 }
2512 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
2513 le16_to_cpu(ctx->attr->data.resident.value_offset));
2514 /* Update the access times if they have changed. */
2515 nt = utc2ntfs(vi->i_mtime);
2516 if (si->last_data_change_time != nt) {
2517 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
2518 "new = 0x%llx", vi->i_ino,
2519 sle64_to_cpu(si->last_data_change_time),
2520 sle64_to_cpu(nt));
2521 si->last_data_change_time = nt;
2522 modified = TRUE;
2523 }
2524 nt = utc2ntfs(vi->i_ctime);
2525 if (si->last_mft_change_time != nt) {
2526 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
2527 "new = 0x%llx", vi->i_ino,
2528 sle64_to_cpu(si->last_mft_change_time),
2529 sle64_to_cpu(nt));
2530 si->last_mft_change_time = nt;
2531 modified = TRUE;
2532 }
2533 nt = utc2ntfs(vi->i_atime);
2534 if (si->last_access_time != nt) {
2535 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
2536 "new = 0x%llx", vi->i_ino,
2537 sle64_to_cpu(si->last_access_time),
2538 sle64_to_cpu(nt));
2539 si->last_access_time = nt;
2540 modified = TRUE;
2541 }
2542 /*
2543 * If we just modified the standard information attribute we need to
2544 * mark the mft record it is in dirty. We do this manually so that
2545 * mark_inode_dirty() is not called which would redirty the inode and
2546 * hence result in an infinite loop of trying to write the inode.
2547 * There is no need to mark the base inode nor the base mft record
2548 * dirty, since we are going to write this mft record below in any case
2549 * and the base mft record may actually not have been modified so it
2550 * might not need to be written out.
2551 * NOTE: It is not a problem when the inode for $MFT itself is being
2552 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
2553 * on the $MFT inode and hence ntfs_write_inode() will not be
2554 * re-invoked because of it which in turn is ok since the dirtied mft
2555 * record will be cleaned and written out to disk below, i.e. before
2556 * this function returns.
2557 */
2558 if (modified && !NInoTestSetDirty(ctx->ntfs_ino))
2559 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
2560 ctx->ntfs_ino->page_ofs);
2561 ntfs_attr_put_search_ctx(ctx);
2562 /* Now the access times are updated, write the base mft record. */
2563 if (NInoDirty(ni))
2564 err = write_mft_record(ni, m, sync);
2565 /* Write all attached extent mft records. */
2566 down(&ni->extent_lock);
2567 if (ni->nr_extents > 0) {
2568 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
2569 int i;
2570
2571 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
2572 for (i = 0; i < ni->nr_extents; i++) {
2573 ntfs_inode *tni = extent_nis[i];
2574
2575 if (NInoDirty(tni)) {
2576 MFT_RECORD *tm = map_mft_record(tni);
2577 int ret;
2578
2579 if (IS_ERR(tm)) {
2580 if (!err || err == -ENOMEM)
2581 err = PTR_ERR(tm);
2582 continue;
2583 }
2584 ret = write_mft_record(tni, tm, sync);
2585 unmap_mft_record(tni);
2586 if (unlikely(ret)) {
2587 if (!err || err == -ENOMEM)
2588 err = ret;
2589 }
2590 }
2591 }
2592 }
2593 up(&ni->extent_lock);
2594 unmap_mft_record(ni);
2595 if (unlikely(err))
2596 goto err_out;
2597 ntfs_debug("Done.");
2598 return 0;
2599unm_err_out:
2600 unmap_mft_record(ni);
2601err_out:
2602 if (err == -ENOMEM) {
2603 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
2604 "Marking the inode dirty again, so the VFS "
2605 "retries later.");
2606 mark_inode_dirty(vi);
2607 } else {
2608 ntfs_error(vi->i_sb, "Failed (error code %i): Marking inode "
2609 "as bad. You should run chkdsk.", -err);
2610 make_bad_inode(vi);
2611 NVolSetErrors(ni->vol);
2612 }
2613 return err;
2614}
2615
2616#endif /* NTFS_RW */