NTFS: Add fs/ntfs/attrib.[hc]::ntfs_attr_extend_allocation(), a function to
[linux-block.git] / fs / ntfs / inode.c
CommitLineData
1da177e4
LT
1/**
2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
3 *
c002f425 4 * Copyright (c) 2001-2005 Anton Altaparmakov
1da177e4
LT
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);
f50f3ac5 177 if (unlikely(!vi))
1da177e4
LT
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 */
f50f3ac5 191 if (unlikely(err == -ENOMEM)) {
1da177e4
LT
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);
f50f3ac5 238 if (unlikely(!vi))
1da177e4
LT
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 */
f50f3ac5 253 if (unlikely(err)) {
1da177e4
LT
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);
f50f3ac5 293 if (unlikely(!vi))
1da177e4
LT
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 */
f50f3ac5 308 if (unlikely(err)) {
1da177e4
LT
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.");
2fb21db2 320 ni = kmem_cache_alloc(ntfs_big_inode_cache, SLAB_NOFS);
1da177e4
LT
321 if (likely(ni != NULL)) {
322 ni->state = 0;
323 return VFS_I(ni);
324 }
325 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
326 return NULL;
327}
328
329void ntfs_destroy_big_inode(struct inode *inode)
330{
331 ntfs_inode *ni = NTFS_I(inode);
332
333 ntfs_debug("Entering.");
334 BUG_ON(ni->page);
335 if (!atomic_dec_and_test(&ni->count))
336 BUG();
337 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
338}
339
340static inline ntfs_inode *ntfs_alloc_extent_inode(void)
341{
342 ntfs_inode *ni;
343
344 ntfs_debug("Entering.");
2fb21db2 345 ni = kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS);
1da177e4
LT
346 if (likely(ni != NULL)) {
347 ni->state = 0;
348 return ni;
349 }
350 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
351 return NULL;
352}
353
354static void ntfs_destroy_extent_inode(ntfs_inode *ni)
355{
356 ntfs_debug("Entering.");
357 BUG_ON(ni->page);
358 if (!atomic_dec_and_test(&ni->count))
359 BUG();
360 kmem_cache_free(ntfs_inode_cache, ni);
361}
362
363/**
364 * __ntfs_init_inode - initialize ntfs specific part of an inode
365 * @sb: super block of mounted volume
366 * @ni: freshly allocated ntfs inode which to initialize
367 *
368 * Initialize an ntfs inode to defaults.
369 *
370 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
371 * untouched. Make sure to initialize them elsewhere.
372 *
373 * Return zero on success and -ENOMEM on error.
374 */
375void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
376{
377 ntfs_debug("Entering.");
36763677 378 rwlock_init(&ni->size_lock);
1da177e4
LT
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;
5ae9fcf8 527 ATTR_RECORD *a;
1da177e4
LT
528 STANDARD_INFORMATION *si;
529 ntfs_attr_search_ctx *ctx;
530 int err = 0;
531
532 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
533
534 /* Setup the generic vfs inode parts now. */
535
536 /* This is the optimal IO size (for stat), not the fs block size. */
537 vi->i_blksize = PAGE_CACHE_SIZE;
538 /*
539 * This is for checking whether an inode has changed w.r.t. a file so
540 * that the file can be updated if necessary (compare with f_version).
541 */
542 vi->i_version = 1;
543
544 vi->i_uid = vol->uid;
545 vi->i_gid = vol->gid;
546 vi->i_mode = 0;
547
548 /*
549 * Initialize the ntfs specific part of @vi special casing
550 * FILE_MFT which we need to do at mount time.
551 */
552 if (vi->i_ino != FILE_MFT)
553 ntfs_init_big_inode(vi);
554 ni = NTFS_I(vi);
555
556 m = map_mft_record(ni);
557 if (IS_ERR(m)) {
558 err = PTR_ERR(m);
559 goto err_out;
560 }
561 ctx = ntfs_attr_get_search_ctx(ni, m);
562 if (!ctx) {
563 err = -ENOMEM;
564 goto unm_err_out;
565 }
566
567 if (!(m->flags & MFT_RECORD_IN_USE)) {
568 ntfs_error(vi->i_sb, "Inode is not in use!");
569 goto unm_err_out;
570 }
571 if (m->base_mft_record) {
572 ntfs_error(vi->i_sb, "Inode is an extent inode!");
573 goto unm_err_out;
574 }
575
576 /* Transfer information from mft record into vfs and ntfs inodes. */
577 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
578
579 /*
580 * FIXME: Keep in mind that link_count is two for files which have both
581 * a long file name and a short file name as separate entries, so if
582 * we are hiding short file names this will be too high. Either we need
583 * to account for the short file names by subtracting them or we need
584 * to make sure we delete files even though i_nlink is not zero which
585 * might be tricky due to vfs interactions. Need to think about this
586 * some more when implementing the unlink command.
587 */
588 vi->i_nlink = le16_to_cpu(m->link_count);
589 /*
590 * FIXME: Reparse points can have the directory bit set even though
591 * they would be S_IFLNK. Need to deal with this further below when we
592 * implement reparse points / symbolic links but it will do for now.
593 * Also if not a directory, it could be something else, rather than
594 * a regular file. But again, will do for now.
595 */
596 /* Everyone gets all permissions. */
597 vi->i_mode |= S_IRWXUGO;
598 /* If read-only, noone gets write permissions. */
599 if (IS_RDONLY(vi))
600 vi->i_mode &= ~S_IWUGO;
601 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
602 vi->i_mode |= S_IFDIR;
603 /*
604 * Apply the directory permissions mask set in the mount
605 * options.
606 */
607 vi->i_mode &= ~vol->dmask;
608 /* Things break without this kludge! */
609 if (vi->i_nlink > 1)
610 vi->i_nlink = 1;
611 } else {
612 vi->i_mode |= S_IFREG;
613 /* Apply the file permissions mask set in the mount options. */
614 vi->i_mode &= ~vol->fmask;
615 }
616 /*
617 * Find the standard information attribute in the mft record. At this
618 * stage we haven't setup the attribute list stuff yet, so this could
619 * in fact fail if the standard information is in an extent record, but
620 * I don't think this actually ever happens.
621 */
622 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
623 ctx);
624 if (unlikely(err)) {
625 if (err == -ENOENT) {
626 /*
627 * TODO: We should be performing a hot fix here (if the
628 * recover mount option is set) by creating a new
629 * attribute.
630 */
631 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
632 "is missing.");
633 }
634 goto unm_err_out;
635 }
5ae9fcf8 636 a = ctx->attr;
1da177e4 637 /* Get the standard information attribute value. */
5ae9fcf8
AA
638 si = (STANDARD_INFORMATION*)((u8*)a +
639 le16_to_cpu(a->data.resident.value_offset));
1da177e4
LT
640
641 /* Transfer information from the standard information into vi. */
642 /*
643 * Note: The i_?times do not quite map perfectly onto the NTFS times,
644 * but they are close enough, and in the end it doesn't really matter
645 * that much...
646 */
647 /*
648 * mtime is the last change of the data within the file. Not changed
649 * when only metadata is changed, e.g. a rename doesn't affect mtime.
650 */
651 vi->i_mtime = ntfs2utc(si->last_data_change_time);
652 /*
653 * ctime is the last change of the metadata of the file. This obviously
654 * always changes, when mtime is changed. ctime can be changed on its
655 * own, mtime is then not changed, e.g. when a file is renamed.
656 */
657 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
658 /*
659 * Last access to the data within the file. Not changed during a rename
660 * for example but changed whenever the file is written to.
661 */
662 vi->i_atime = ntfs2utc(si->last_access_time);
663
664 /* Find the attribute list attribute if present. */
665 ntfs_attr_reinit_search_ctx(ctx);
666 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
667 if (err) {
668 if (unlikely(err != -ENOENT)) {
669 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
670 "attribute.");
671 goto unm_err_out;
672 }
673 } else /* if (!err) */ {
674 if (vi->i_ino == FILE_MFT)
675 goto skip_attr_list_load;
676 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
677 NInoSetAttrList(ni);
5ae9fcf8
AA
678 a = ctx->attr;
679 if (a->flags & ATTR_IS_ENCRYPTED ||
680 a->flags & ATTR_COMPRESSION_MASK ||
681 a->flags & ATTR_IS_SPARSE) {
1da177e4
LT
682 ntfs_error(vi->i_sb, "Attribute list attribute is "
683 "compressed/encrypted/sparse.");
684 goto unm_err_out;
685 }
686 /* Now allocate memory for the attribute list. */
5ae9fcf8 687 ni->attr_list_size = (u32)ntfs_attr_size(a);
1da177e4
LT
688 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
689 if (!ni->attr_list) {
690 ntfs_error(vi->i_sb, "Not enough memory to allocate "
691 "buffer for attribute list.");
692 err = -ENOMEM;
693 goto unm_err_out;
694 }
5ae9fcf8 695 if (a->non_resident) {
1da177e4 696 NInoSetAttrListNonResident(ni);
5ae9fcf8 697 if (a->data.non_resident.lowest_vcn) {
1da177e4
LT
698 ntfs_error(vi->i_sb, "Attribute list has non "
699 "zero lowest_vcn.");
700 goto unm_err_out;
701 }
702 /*
703 * Setup the runlist. No need for locking as we have
704 * exclusive access to the inode at this time.
705 */
706 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
5ae9fcf8 707 a, NULL);
1da177e4
LT
708 if (IS_ERR(ni->attr_list_rl.rl)) {
709 err = PTR_ERR(ni->attr_list_rl.rl);
710 ni->attr_list_rl.rl = NULL;
711 ntfs_error(vi->i_sb, "Mapping pairs "
712 "decompression failed.");
713 goto unm_err_out;
714 }
715 /* Now load the attribute list. */
716 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
717 ni->attr_list, ni->attr_list_size,
5ae9fcf8
AA
718 sle64_to_cpu(a->data.non_resident.
719 initialized_size)))) {
1da177e4
LT
720 ntfs_error(vi->i_sb, "Failed to load "
721 "attribute list attribute.");
722 goto unm_err_out;
723 }
5ae9fcf8
AA
724 } else /* if (!a->non_resident) */ {
725 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
726 + le32_to_cpu(
727 a->data.resident.value_length) >
1da177e4
LT
728 (u8*)ctx->mrec + vol->mft_record_size) {
729 ntfs_error(vi->i_sb, "Corrupt attribute list "
730 "in inode.");
731 goto unm_err_out;
732 }
733 /* Now copy the attribute list. */
5ae9fcf8
AA
734 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
735 a->data.resident.value_offset),
1da177e4 736 le32_to_cpu(
5ae9fcf8 737 a->data.resident.value_length));
1da177e4
LT
738 }
739 }
740skip_attr_list_load:
741 /*
742 * If an attribute list is present we now have the attribute list value
743 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
744 */
745 if (S_ISDIR(vi->i_mode)) {
f50f3ac5 746 loff_t bvi_size;
1da177e4
LT
747 struct inode *bvi;
748 ntfs_inode *bni;
749 INDEX_ROOT *ir;
5ae9fcf8 750 u8 *ir_end, *index_end;
1da177e4
LT
751
752 /* It is a directory, find index root attribute. */
753 ntfs_attr_reinit_search_ctx(ctx);
754 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
755 0, NULL, 0, ctx);
756 if (unlikely(err)) {
757 if (err == -ENOENT) {
758 // FIXME: File is corrupt! Hot-fix with empty
759 // index root attribute if recovery option is
760 // set.
761 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
762 "is missing.");
763 }
764 goto unm_err_out;
765 }
5ae9fcf8 766 a = ctx->attr;
1da177e4 767 /* Set up the state. */
5ae9fcf8 768 if (unlikely(a->non_resident)) {
1da177e4
LT
769 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
770 "resident.");
771 goto unm_err_out;
772 }
773 /* Ensure the attribute name is placed before the value. */
5ae9fcf8
AA
774 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
775 le16_to_cpu(a->data.resident.value_offset)))) {
1da177e4
LT
776 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
777 "placed after the attribute value.");
778 goto unm_err_out;
779 }
780 /*
781 * Compressed/encrypted index root just means that the newly
782 * created files in that directory should be created compressed/
783 * encrypted. However index root cannot be both compressed and
784 * encrypted.
785 */
5ae9fcf8 786 if (a->flags & ATTR_COMPRESSION_MASK)
1da177e4 787 NInoSetCompressed(ni);
5ae9fcf8
AA
788 if (a->flags & ATTR_IS_ENCRYPTED) {
789 if (a->flags & ATTR_COMPRESSION_MASK) {
1da177e4
LT
790 ntfs_error(vi->i_sb, "Found encrypted and "
791 "compressed attribute.");
792 goto unm_err_out;
793 }
794 NInoSetEncrypted(ni);
795 }
5ae9fcf8 796 if (a->flags & ATTR_IS_SPARSE)
1da177e4 797 NInoSetSparse(ni);
5ae9fcf8
AA
798 ir = (INDEX_ROOT*)((u8*)a +
799 le16_to_cpu(a->data.resident.value_offset));
800 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
801 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1da177e4
LT
802 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
803 "corrupt.");
804 goto unm_err_out;
805 }
5ae9fcf8 806 index_end = (u8*)&ir->index +
1da177e4
LT
807 le32_to_cpu(ir->index.index_length);
808 if (index_end > ir_end) {
809 ntfs_error(vi->i_sb, "Directory index is corrupt.");
810 goto unm_err_out;
811 }
812 if (ir->type != AT_FILE_NAME) {
813 ntfs_error(vi->i_sb, "Indexed attribute is not "
814 "$FILE_NAME.");
815 goto unm_err_out;
816 }
817 if (ir->collation_rule != COLLATION_FILE_NAME) {
818 ntfs_error(vi->i_sb, "Index collation rule is not "
819 "COLLATION_FILE_NAME.");
820 goto unm_err_out;
821 }
822 ni->itype.index.collation_rule = ir->collation_rule;
823 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
824 if (ni->itype.index.block_size &
825 (ni->itype.index.block_size - 1)) {
826 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
827 "power of two.",
828 ni->itype.index.block_size);
829 goto unm_err_out;
830 }
831 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
832 ntfs_error(vi->i_sb, "Index block size (%u) > "
833 "PAGE_CACHE_SIZE (%ld) is not "
834 "supported. Sorry.",
835 ni->itype.index.block_size,
836 PAGE_CACHE_SIZE);
837 err = -EOPNOTSUPP;
838 goto unm_err_out;
839 }
840 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
841 ntfs_error(vi->i_sb, "Index block size (%u) < "
842 "NTFS_BLOCK_SIZE (%i) is not "
843 "supported. Sorry.",
844 ni->itype.index.block_size,
845 NTFS_BLOCK_SIZE);
846 err = -EOPNOTSUPP;
847 goto unm_err_out;
848 }
849 ni->itype.index.block_size_bits =
850 ffs(ni->itype.index.block_size) - 1;
851 /* Determine the size of a vcn in the directory index. */
852 if (vol->cluster_size <= ni->itype.index.block_size) {
853 ni->itype.index.vcn_size = vol->cluster_size;
854 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
855 } else {
856 ni->itype.index.vcn_size = vol->sector_size;
857 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
858 }
859
860 /* Setup the index allocation attribute, even if not present. */
861 NInoSetMstProtected(ni);
862 ni->type = AT_INDEX_ALLOCATION;
863 ni->name = I30;
864 ni->name_len = 4;
865
866 if (!(ir->index.flags & LARGE_INDEX)) {
867 /* No index allocation. */
868 vi->i_size = ni->initialized_size =
869 ni->allocated_size = 0;
870 /* We are done with the mft record, so we release it. */
871 ntfs_attr_put_search_ctx(ctx);
872 unmap_mft_record(ni);
873 m = NULL;
874 ctx = NULL;
875 goto skip_large_dir_stuff;
876 } /* LARGE_INDEX: Index allocation present. Setup state. */
877 NInoSetIndexAllocPresent(ni);
878 /* Find index allocation attribute. */
879 ntfs_attr_reinit_search_ctx(ctx);
880 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
881 CASE_SENSITIVE, 0, NULL, 0, ctx);
882 if (unlikely(err)) {
883 if (err == -ENOENT)
884 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
885 "attribute is not present but "
886 "$INDEX_ROOT indicated it is.");
887 else
888 ntfs_error(vi->i_sb, "Failed to lookup "
889 "$INDEX_ALLOCATION "
890 "attribute.");
891 goto unm_err_out;
892 }
5ae9fcf8
AA
893 a = ctx->attr;
894 if (!a->non_resident) {
1da177e4
LT
895 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
896 "is resident.");
897 goto unm_err_out;
898 }
899 /*
900 * Ensure the attribute name is placed before the mapping pairs
901 * array.
902 */
5ae9fcf8
AA
903 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
904 le16_to_cpu(
905 a->data.non_resident.mapping_pairs_offset)))) {
1da177e4
LT
906 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
907 "is placed after the mapping pairs "
908 "array.");
909 goto unm_err_out;
910 }
5ae9fcf8 911 if (a->flags & ATTR_IS_ENCRYPTED) {
1da177e4
LT
912 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
913 "is encrypted.");
914 goto unm_err_out;
915 }
5ae9fcf8 916 if (a->flags & ATTR_IS_SPARSE) {
1da177e4
LT
917 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
918 "is sparse.");
919 goto unm_err_out;
920 }
5ae9fcf8 921 if (a->flags & ATTR_COMPRESSION_MASK) {
1da177e4
LT
922 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
923 "is compressed.");
924 goto unm_err_out;
925 }
5ae9fcf8 926 if (a->data.non_resident.lowest_vcn) {
1da177e4
LT
927 ntfs_error(vi->i_sb, "First extent of "
928 "$INDEX_ALLOCATION attribute has non "
929 "zero lowest_vcn.");
930 goto unm_err_out;
931 }
5ae9fcf8 932 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1da177e4 933 ni->initialized_size = sle64_to_cpu(
5ae9fcf8 934 a->data.non_resident.initialized_size);
1da177e4 935 ni->allocated_size = sle64_to_cpu(
5ae9fcf8 936 a->data.non_resident.allocated_size);
1da177e4
LT
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. */
f50f3ac5
AA
961 bvi_size = i_size_read(bvi);
962 if ((bvi_size << 3) < (vi->i_size >>
1da177e4
LT
963 ni->itype.index.block_size_bits)) {
964 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
965 "for index allocation (0x%llx).",
f50f3ac5 966 bvi_size << 3, vi->i_size);
1da177e4
LT
967 goto unm_err_out;
968 }
969skip_large_dir_stuff:
970 /* Setup the operations for this inode. */
971 vi->i_op = &ntfs_dir_inode_ops;
972 vi->i_fop = &ntfs_dir_ops;
973 } else {
974 /* It is a file. */
975 ntfs_attr_reinit_search_ctx(ctx);
976
977 /* Setup the data attribute, even if not present. */
978 ni->type = AT_DATA;
979 ni->name = NULL;
980 ni->name_len = 0;
981
982 /* Find first extent of the unnamed data attribute. */
983 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
984 if (unlikely(err)) {
985 vi->i_size = ni->initialized_size =
986 ni->allocated_size = 0;
987 if (err != -ENOENT) {
988 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
989 "attribute.");
990 goto unm_err_out;
991 }
992 /*
993 * FILE_Secure does not have an unnamed $DATA
994 * attribute, so we special case it here.
995 */
996 if (vi->i_ino == FILE_Secure)
997 goto no_data_attr_special_case;
998 /*
999 * Most if not all the system files in the $Extend
1000 * system directory do not have unnamed data
1001 * attributes so we need to check if the parent
1002 * directory of the file is FILE_Extend and if it is
1003 * ignore this error. To do this we need to get the
1004 * name of this inode from the mft record as the name
1005 * contains the back reference to the parent directory.
1006 */
1007 if (ntfs_is_extended_system_file(ctx) > 0)
1008 goto no_data_attr_special_case;
1009 // FIXME: File is corrupt! Hot-fix with empty data
1010 // attribute if recovery option is set.
1011 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1012 goto unm_err_out;
1013 }
5ae9fcf8 1014 a = ctx->attr;
1da177e4 1015 /* Setup the state. */
67bb1037
AA
1016 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1017 if (a->flags & ATTR_COMPRESSION_MASK) {
1018 NInoSetCompressed(ni);
1019 if (vol->cluster_size > 4096) {
1020 ntfs_error(vi->i_sb, "Found "
9451f851
AA
1021 "compressed data but "
1022 "compression is "
1023 "disabled due to "
1024 "cluster size (%i) > "
1025 "4kiB.",
1026 vol->cluster_size);
67bb1037
AA
1027 goto unm_err_out;
1028 }
1029 if ((a->flags & ATTR_COMPRESSION_MASK)
1030 != ATTR_IS_COMPRESSED) {
1031 ntfs_error(vi->i_sb, "Found unknown "
1032 "compression method "
1033 "or corrupt file.");
1034 goto unm_err_out;
1da177e4 1035 }
67bb1037
AA
1036 }
1037 if (a->flags & ATTR_IS_SPARSE)
1038 NInoSetSparse(ni);
1039 }
1040 if (a->flags & ATTR_IS_ENCRYPTED) {
1041 if (NInoCompressed(ni)) {
1042 ntfs_error(vi->i_sb, "Found encrypted and "
1043 "compressed data.");
1044 goto unm_err_out;
1045 }
1046 NInoSetEncrypted(ni);
1047 }
1048 if (a->non_resident) {
1049 NInoSetNonResident(ni);
1050 if (NInoCompressed(ni) || NInoSparse(ni)) {
5ae9fcf8
AA
1051 if (a->data.non_resident.compression_unit !=
1052 4) {
1da177e4 1053 ntfs_error(vi->i_sb, "Found "
67bb1037
AA
1054 "nonstandard "
1055 "compression unit (%u "
1056 "instead of 4). "
1057 "Cannot handle this.",
1058 a->data.non_resident.
1059 compression_unit);
1da177e4
LT
1060 err = -EOPNOTSUPP;
1061 goto unm_err_out;
1062 }
9451f851
AA
1063 ni->itype.compressed.block_clusters = 1U <<
1064 a->data.non_resident.
1065 compression_unit;
1da177e4 1066 ni->itype.compressed.block_size = 1U << (
5ae9fcf8 1067 a->data.non_resident.
1da177e4
LT
1068 compression_unit +
1069 vol->cluster_size_bits);
1070 ni->itype.compressed.block_size_bits = ffs(
9451f851
AA
1071 ni->itype.compressed.
1072 block_size) - 1;
1073 ni->itype.compressed.size = sle64_to_cpu(
1074 a->data.non_resident.
1075 compressed_size);
1da177e4 1076 }
5ae9fcf8 1077 if (a->data.non_resident.lowest_vcn) {
1da177e4
LT
1078 ntfs_error(vi->i_sb, "First extent of $DATA "
1079 "attribute has non zero "
1080 "lowest_vcn.");
1081 goto unm_err_out;
1082 }
1da177e4 1083 vi->i_size = sle64_to_cpu(
5ae9fcf8 1084 a->data.non_resident.data_size);
1da177e4 1085 ni->initialized_size = sle64_to_cpu(
5ae9fcf8 1086 a->data.non_resident.initialized_size);
1da177e4 1087 ni->allocated_size = sle64_to_cpu(
5ae9fcf8 1088 a->data.non_resident.allocated_size);
1da177e4 1089 } else { /* Resident attribute. */
5ae9fcf8
AA
1090 vi->i_size = ni->initialized_size = le32_to_cpu(
1091 a->data.resident.value_length);
1092 ni->allocated_size = le32_to_cpu(a->length) -
1093 le16_to_cpu(
1094 a->data.resident.value_offset);
1095 if (vi->i_size > ni->allocated_size) {
1096 ntfs_error(vi->i_sb, "Resident data attribute "
1097 "is corrupt (size exceeds "
1098 "allocation).");
1099 goto unm_err_out;
1100 }
1da177e4
LT
1101 }
1102no_data_attr_special_case:
1103 /* We are done with the mft record, so we release it. */
1104 ntfs_attr_put_search_ctx(ctx);
1105 unmap_mft_record(ni);
1106 m = NULL;
1107 ctx = NULL;
1108 /* Setup the operations for this inode. */
1109 vi->i_op = &ntfs_file_inode_ops;
1110 vi->i_fop = &ntfs_file_ops;
1111 }
1112 if (NInoMstProtected(ni))
1113 vi->i_mapping->a_ops = &ntfs_mst_aops;
1114 else
1115 vi->i_mapping->a_ops = &ntfs_aops;
1116 /*
1117 * The number of 512-byte blocks used on disk (for stat). This is in so
1118 * far inaccurate as it doesn't account for any named streams or other
1119 * special non-resident attributes, but that is how Windows works, too,
1120 * so we are at least consistent with Windows, if not entirely
1121 * consistent with the Linux Way. Doing it the Linux Way would cause a
1122 * significant slowdown as it would involve iterating over all
1123 * attributes in the mft record and adding the allocated/compressed
1124 * sizes of all non-resident attributes present to give us the Linux
1125 * correct size that should go into i_blocks (after division by 512).
1126 */
9451f851 1127 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1da177e4 1128 vi->i_blocks = ni->itype.compressed.size >> 9;
9451f851
AA
1129 else
1130 vi->i_blocks = ni->allocated_size >> 9;
1da177e4
LT
1131 ntfs_debug("Done.");
1132 return 0;
1133
1134unm_err_out:
1135 if (!err)
1136 err = -EIO;
1137 if (ctx)
1138 ntfs_attr_put_search_ctx(ctx);
1139 if (m)
1140 unmap_mft_record(ni);
1141err_out:
1142 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1143 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1144 make_bad_inode(vi);
1145 if (err != -EOPNOTSUPP && err != -ENOMEM)
1146 NVolSetErrors(vol);
1147 return err;
1148}
1149
1150/**
1151 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1152 * @base_vi: base inode
1153 * @vi: attribute inode to read
1154 *
1155 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1156 * attribute inode described by @vi into memory from the base mft record
1157 * described by @base_ni.
1158 *
1159 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1160 * reading and looks up the attribute described by @vi before setting up the
1161 * necessary fields in @vi as well as initializing the ntfs inode.
1162 *
1163 * Q: What locks are held when the function is called?
1164 * A: i_state has I_LOCK set, hence the inode is locked, also
1165 * i_count is set to 1, so it is not going to go away
1166 *
1167 * Return 0 on success and -errno on error. In the error case, the inode will
1168 * have had make_bad_inode() executed on it.
f6098cf4
AA
1169 *
1170 * Note this cannot be called for AT_INDEX_ALLOCATION.
1da177e4
LT
1171 */
1172static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1173{
1174 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1175 ntfs_inode *ni, *base_ni;
1176 MFT_RECORD *m;
5ae9fcf8 1177 ATTR_RECORD *a;
1da177e4
LT
1178 ntfs_attr_search_ctx *ctx;
1179 int err = 0;
1180
1181 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1182
1183 ntfs_init_big_inode(vi);
1184
1185 ni = NTFS_I(vi);
1186 base_ni = NTFS_I(base_vi);
1187
1188 /* Just mirror the values from the base inode. */
1189 vi->i_blksize = base_vi->i_blksize;
1190 vi->i_version = base_vi->i_version;
1191 vi->i_uid = base_vi->i_uid;
1192 vi->i_gid = base_vi->i_gid;
1193 vi->i_nlink = base_vi->i_nlink;
1194 vi->i_mtime = base_vi->i_mtime;
1195 vi->i_ctime = base_vi->i_ctime;
1196 vi->i_atime = base_vi->i_atime;
1197 vi->i_generation = ni->seq_no = base_ni->seq_no;
1198
1199 /* Set inode type to zero but preserve permissions. */
1200 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1201
1202 m = map_mft_record(base_ni);
1203 if (IS_ERR(m)) {
1204 err = PTR_ERR(m);
1205 goto err_out;
1206 }
1207 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1208 if (!ctx) {
1209 err = -ENOMEM;
1210 goto unm_err_out;
1211 }
1da177e4
LT
1212 /* Find the attribute. */
1213 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1214 CASE_SENSITIVE, 0, NULL, 0, ctx);
1215 if (unlikely(err))
1216 goto unm_err_out;
5ae9fcf8 1217 a = ctx->attr;
67bb1037
AA
1218 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1219 if (a->flags & ATTR_COMPRESSION_MASK) {
1220 NInoSetCompressed(ni);
1221 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1222 ni->name_len)) {
1223 ntfs_error(vi->i_sb, "Found compressed "
1224 "non-data or named data "
1225 "attribute. Please report "
1226 "you saw this message to "
1227 "linux-ntfs-dev@lists."
1228 "sourceforge.net");
1229 goto unm_err_out;
1230 }
1231 if (vol->cluster_size > 4096) {
1232 ntfs_error(vi->i_sb, "Found compressed "
1233 "attribute but compression is "
1234 "disabled due to cluster size "
1235 "(%i) > 4kiB.",
1236 vol->cluster_size);
1237 goto unm_err_out;
1238 }
1239 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1240 ATTR_IS_COMPRESSED) {
1241 ntfs_error(vi->i_sb, "Found unknown "
1242 "compression method.");
1243 goto unm_err_out;
1244 }
1245 }
1246 /*
f6098cf4
AA
1247 * The compressed/sparse flag set in an index root just means
1248 * to compress all files.
67bb1037
AA
1249 */
1250 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1251 ntfs_error(vi->i_sb, "Found mst protected attribute "
1252 "but the attribute is %s. Please "
1253 "report you saw this message to "
1254 "linux-ntfs-dev@lists.sourceforge.net",
1255 NInoCompressed(ni) ? "compressed" :
1256 "sparse");
1257 goto unm_err_out;
1258 }
1259 if (a->flags & ATTR_IS_SPARSE)
1260 NInoSetSparse(ni);
1261 }
1262 if (a->flags & ATTR_IS_ENCRYPTED) {
1263 if (NInoCompressed(ni)) {
1264 ntfs_error(vi->i_sb, "Found encrypted and compressed "
1265 "data.");
1266 goto unm_err_out;
1267 }
1268 /*
1269 * The encryption flag set in an index root just means to
1270 * encrypt all files.
1271 */
1272 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1273 ntfs_error(vi->i_sb, "Found mst protected attribute "
1274 "but the attribute is encrypted. "
1275 "Please report you saw this message "
1276 "to linux-ntfs-dev@lists.sourceforge."
1277 "net");
1278 goto unm_err_out;
1279 }
1280 if (ni->type != AT_DATA) {
1281 ntfs_error(vi->i_sb, "Found encrypted non-data "
1282 "attribute.");
1283 goto unm_err_out;
1284 }
1285 NInoSetEncrypted(ni);
1286 }
5ae9fcf8 1287 if (!a->non_resident) {
1da177e4 1288 /* Ensure the attribute name is placed before the value. */
5ae9fcf8
AA
1289 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1290 le16_to_cpu(a->data.resident.value_offset)))) {
1da177e4
LT
1291 ntfs_error(vol->sb, "Attribute name is placed after "
1292 "the attribute value.");
1293 goto unm_err_out;
1294 }
67bb1037 1295 if (NInoMstProtected(ni)) {
1da177e4 1296 ntfs_error(vi->i_sb, "Found mst protected attribute "
67bb1037
AA
1297 "but the attribute is resident. "
1298 "Please report you saw this message to "
1da177e4
LT
1299 "linux-ntfs-dev@lists.sourceforge.net");
1300 goto unm_err_out;
1301 }
5ae9fcf8
AA
1302 vi->i_size = ni->initialized_size = le32_to_cpu(
1303 a->data.resident.value_length);
1304 ni->allocated_size = le32_to_cpu(a->length) -
1305 le16_to_cpu(a->data.resident.value_offset);
1306 if (vi->i_size > ni->allocated_size) {
9451f851
AA
1307 ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1308 "(size exceeds allocation).");
5ae9fcf8
AA
1309 goto unm_err_out;
1310 }
1da177e4
LT
1311 } else {
1312 NInoSetNonResident(ni);
1313 /*
1314 * Ensure the attribute name is placed before the mapping pairs
1315 * array.
1316 */
5ae9fcf8
AA
1317 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1318 le16_to_cpu(
1319 a->data.non_resident.mapping_pairs_offset)))) {
1da177e4
LT
1320 ntfs_error(vol->sb, "Attribute name is placed after "
1321 "the mapping pairs array.");
1322 goto unm_err_out;
1323 }
f6098cf4 1324 if (NInoCompressed(ni) || NInoSparse(ni)) {
5ae9fcf8 1325 if (a->data.non_resident.compression_unit != 4) {
1da177e4
LT
1326 ntfs_error(vi->i_sb, "Found nonstandard "
1327 "compression unit (%u instead "
1328 "of 4). Cannot handle this.",
5ae9fcf8 1329 a->data.non_resident.
1da177e4
LT
1330 compression_unit);
1331 err = -EOPNOTSUPP;
1332 goto unm_err_out;
1333 }
9451f851
AA
1334 ni->itype.compressed.block_clusters = 1U <<
1335 a->data.non_resident.compression_unit;
1da177e4 1336 ni->itype.compressed.block_size = 1U << (
5ae9fcf8 1337 a->data.non_resident.compression_unit +
1da177e4
LT
1338 vol->cluster_size_bits);
1339 ni->itype.compressed.block_size_bits = ffs(
5ae9fcf8 1340 ni->itype.compressed.block_size) - 1;
9451f851
AA
1341 ni->itype.compressed.size = sle64_to_cpu(
1342 a->data.non_resident.compressed_size);
1da177e4 1343 }
5ae9fcf8 1344 if (a->data.non_resident.lowest_vcn) {
1da177e4
LT
1345 ntfs_error(vi->i_sb, "First extent of attribute has "
1346 "non-zero lowest_vcn.");
1347 goto unm_err_out;
1348 }
5ae9fcf8 1349 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1da177e4 1350 ni->initialized_size = sle64_to_cpu(
5ae9fcf8 1351 a->data.non_resident.initialized_size);
1da177e4 1352 ni->allocated_size = sle64_to_cpu(
5ae9fcf8 1353 a->data.non_resident.allocated_size);
1da177e4 1354 }
1da177e4
LT
1355 /* Setup the operations for this attribute inode. */
1356 vi->i_op = NULL;
1357 vi->i_fop = NULL;
1358 if (NInoMstProtected(ni))
1359 vi->i_mapping->a_ops = &ntfs_mst_aops;
1360 else
1361 vi->i_mapping->a_ops = &ntfs_aops;
67bb1037 1362 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1da177e4 1363 vi->i_blocks = ni->itype.compressed.size >> 9;
9451f851
AA
1364 else
1365 vi->i_blocks = ni->allocated_size >> 9;
1da177e4 1366 /*
67bb1037 1367 * Make sure the base inode does not go away and attach it to the
1da177e4
LT
1368 * attribute inode.
1369 */
1370 igrab(base_vi);
1371 ni->ext.base_ntfs_ino = base_ni;
1372 ni->nr_extents = -1;
1373
1374 ntfs_attr_put_search_ctx(ctx);
1375 unmap_mft_record(base_ni);
1376
1377 ntfs_debug("Done.");
1378 return 0;
1379
1380unm_err_out:
1381 if (!err)
1382 err = -EIO;
1383 if (ctx)
1384 ntfs_attr_put_search_ctx(ctx);
1385 unmap_mft_record(base_ni);
1386err_out:
1387 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1388 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1389 "Marking corrupt inode and base inode 0x%lx as bad. "
1390 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1391 base_vi->i_ino);
1392 make_bad_inode(vi);
1393 make_bad_inode(base_vi);
1394 if (err != -ENOMEM)
1395 NVolSetErrors(vol);
1396 return err;
1397}
1398
1399/**
1400 * ntfs_read_locked_index_inode - read an index inode from its base inode
1401 * @base_vi: base inode
1402 * @vi: index inode to read
1403 *
1404 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1405 * index inode described by @vi into memory from the base mft record described
1406 * by @base_ni.
1407 *
1408 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1409 * reading and looks up the attributes relating to the index described by @vi
1410 * before setting up the necessary fields in @vi as well as initializing the
1411 * ntfs inode.
1412 *
1413 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1414 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1415 * are setup like directory inodes since directories are a special case of
1416 * indices ao they need to be treated in much the same way. Most importantly,
1417 * for small indices the index allocation attribute might not actually exist.
1418 * However, the index root attribute always exists but this does not need to
1419 * have an inode associated with it and this is why we define a new inode type
1420 * index. Also, like for directories, we need to have an attribute inode for
1421 * the bitmap attribute corresponding to the index allocation attribute and we
1422 * can store this in the appropriate field of the inode, just like we do for
1423 * normal directory inodes.
1424 *
1425 * Q: What locks are held when the function is called?
1426 * A: i_state has I_LOCK set, hence the inode is locked, also
1427 * i_count is set to 1, so it is not going to go away
1428 *
1429 * Return 0 on success and -errno on error. In the error case, the inode will
1430 * have had make_bad_inode() executed on it.
1431 */
1432static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1433{
f50f3ac5 1434 loff_t bvi_size;
1da177e4
LT
1435 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1436 ntfs_inode *ni, *base_ni, *bni;
1437 struct inode *bvi;
1438 MFT_RECORD *m;
5ae9fcf8 1439 ATTR_RECORD *a;
1da177e4
LT
1440 ntfs_attr_search_ctx *ctx;
1441 INDEX_ROOT *ir;
1442 u8 *ir_end, *index_end;
1443 int err = 0;
1444
1445 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1446 ntfs_init_big_inode(vi);
1447 ni = NTFS_I(vi);
1448 base_ni = NTFS_I(base_vi);
1449 /* Just mirror the values from the base inode. */
1450 vi->i_blksize = base_vi->i_blksize;
1451 vi->i_version = base_vi->i_version;
1452 vi->i_uid = base_vi->i_uid;
1453 vi->i_gid = base_vi->i_gid;
1454 vi->i_nlink = base_vi->i_nlink;
1455 vi->i_mtime = base_vi->i_mtime;
1456 vi->i_ctime = base_vi->i_ctime;
1457 vi->i_atime = base_vi->i_atime;
1458 vi->i_generation = ni->seq_no = base_ni->seq_no;
1459 /* Set inode type to zero but preserve permissions. */
1460 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1461 /* Map the mft record for the base inode. */
1462 m = map_mft_record(base_ni);
1463 if (IS_ERR(m)) {
1464 err = PTR_ERR(m);
1465 goto err_out;
1466 }
1467 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1468 if (!ctx) {
1469 err = -ENOMEM;
1470 goto unm_err_out;
1471 }
1472 /* Find the index root attribute. */
1473 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1474 CASE_SENSITIVE, 0, NULL, 0, ctx);
1475 if (unlikely(err)) {
1476 if (err == -ENOENT)
1477 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1478 "missing.");
1479 goto unm_err_out;
1480 }
5ae9fcf8 1481 a = ctx->attr;
1da177e4 1482 /* Set up the state. */
5ae9fcf8 1483 if (unlikely(a->non_resident)) {
1da177e4
LT
1484 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1485 goto unm_err_out;
1486 }
1487 /* Ensure the attribute name is placed before the value. */
5ae9fcf8
AA
1488 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1489 le16_to_cpu(a->data.resident.value_offset)))) {
1da177e4
LT
1490 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1491 "after the attribute value.");
1492 goto unm_err_out;
1493 }
67bb1037
AA
1494 /*
1495 * Compressed/encrypted/sparse index root is not allowed, except for
1496 * directories of course but those are not dealt with here.
1497 */
5ae9fcf8 1498 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1da177e4
LT
1499 ATTR_IS_SPARSE)) {
1500 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1501 "root attribute.");
1502 goto unm_err_out;
1503 }
5ae9fcf8
AA
1504 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1505 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1da177e4
LT
1506 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1507 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1508 goto unm_err_out;
1509 }
1510 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1511 if (index_end > ir_end) {
1512 ntfs_error(vi->i_sb, "Index is corrupt.");
1513 goto unm_err_out;
1514 }
1515 if (ir->type) {
1516 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1517 le32_to_cpu(ir->type));
1518 goto unm_err_out;
1519 }
1520 ni->itype.index.collation_rule = ir->collation_rule;
1521 ntfs_debug("Index collation rule is 0x%x.",
1522 le32_to_cpu(ir->collation_rule));
1523 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1524 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1525 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1526 "two.", ni->itype.index.block_size);
1527 goto unm_err_out;
1528 }
1529 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1530 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1531 "(%ld) is not supported. Sorry.",
1532 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1533 err = -EOPNOTSUPP;
1534 goto unm_err_out;
1535 }
1536 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1537 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1538 "(%i) is not supported. Sorry.",
1539 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1540 err = -EOPNOTSUPP;
1541 goto unm_err_out;
1542 }
1543 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1544 /* Determine the size of a vcn in the index. */
1545 if (vol->cluster_size <= ni->itype.index.block_size) {
1546 ni->itype.index.vcn_size = vol->cluster_size;
1547 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1548 } else {
1549 ni->itype.index.vcn_size = vol->sector_size;
1550 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1551 }
1552 /* Check for presence of index allocation attribute. */
1553 if (!(ir->index.flags & LARGE_INDEX)) {
1554 /* No index allocation. */
1555 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1556 /* We are done with the mft record, so we release it. */
1557 ntfs_attr_put_search_ctx(ctx);
1558 unmap_mft_record(base_ni);
1559 m = NULL;
1560 ctx = NULL;
1561 goto skip_large_index_stuff;
1562 } /* LARGE_INDEX: Index allocation present. Setup state. */
1563 NInoSetIndexAllocPresent(ni);
1564 /* Find index allocation attribute. */
1565 ntfs_attr_reinit_search_ctx(ctx);
1566 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1567 CASE_SENSITIVE, 0, NULL, 0, ctx);
1568 if (unlikely(err)) {
1569 if (err == -ENOENT)
1570 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1571 "not present but $INDEX_ROOT "
1572 "indicated it is.");
1573 else
1574 ntfs_error(vi->i_sb, "Failed to lookup "
1575 "$INDEX_ALLOCATION attribute.");
1576 goto unm_err_out;
1577 }
5ae9fcf8 1578 if (!a->non_resident) {
1da177e4
LT
1579 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1580 "resident.");
1581 goto unm_err_out;
1582 }
1583 /*
1584 * Ensure the attribute name is placed before the mapping pairs array.
1585 */
5ae9fcf8
AA
1586 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1587 le16_to_cpu(
1588 a->data.non_resident.mapping_pairs_offset)))) {
1da177e4
LT
1589 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1590 "placed after the mapping pairs array.");
1591 goto unm_err_out;
1592 }
5ae9fcf8 1593 if (a->flags & ATTR_IS_ENCRYPTED) {
1da177e4
LT
1594 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1595 "encrypted.");
1596 goto unm_err_out;
1597 }
5ae9fcf8 1598 if (a->flags & ATTR_IS_SPARSE) {
1da177e4
LT
1599 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1600 goto unm_err_out;
1601 }
5ae9fcf8 1602 if (a->flags & ATTR_COMPRESSION_MASK) {
1da177e4
LT
1603 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1604 "compressed.");
1605 goto unm_err_out;
1606 }
5ae9fcf8 1607 if (a->data.non_resident.lowest_vcn) {
1da177e4
LT
1608 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1609 "attribute has non zero lowest_vcn.");
1610 goto unm_err_out;
1611 }
5ae9fcf8 1612 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1da177e4 1613 ni->initialized_size = sle64_to_cpu(
5ae9fcf8
AA
1614 a->data.non_resident.initialized_size);
1615 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1da177e4
LT
1616 /*
1617 * We are done with the mft record, so we release it. Otherwise
1618 * we would deadlock in ntfs_attr_iget().
1619 */
1620 ntfs_attr_put_search_ctx(ctx);
1621 unmap_mft_record(base_ni);
1622 m = NULL;
1623 ctx = NULL;
1624 /* Get the index bitmap attribute inode. */
1625 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1626 if (IS_ERR(bvi)) {
1627 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1628 err = PTR_ERR(bvi);
1629 goto unm_err_out;
1630 }
1631 bni = NTFS_I(bvi);
1632 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1633 NInoSparse(bni)) {
1634 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1635 "encrypted and/or sparse.");
1636 goto iput_unm_err_out;
1637 }
1638 /* Consistency check bitmap size vs. index allocation size. */
f50f3ac5
AA
1639 bvi_size = i_size_read(bvi);
1640 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1da177e4 1641 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
f50f3ac5 1642 "index allocation (0x%llx).", bvi_size << 3,
1da177e4
LT
1643 vi->i_size);
1644 goto iput_unm_err_out;
1645 }
1646 ni->itype.index.bmp_ino = bvi;
1647skip_large_index_stuff:
1648 /* Setup the operations for this index inode. */
1649 vi->i_op = NULL;
1650 vi->i_fop = NULL;
1651 vi->i_mapping->a_ops = &ntfs_mst_aops;
1652 vi->i_blocks = ni->allocated_size >> 9;
1da177e4
LT
1653 /*
1654 * Make sure the base inode doesn't go away and attach it to the
1655 * index inode.
1656 */
1657 igrab(base_vi);
1658 ni->ext.base_ntfs_ino = base_ni;
1659 ni->nr_extents = -1;
1660
1661 ntfs_debug("Done.");
1662 return 0;
1663
1664iput_unm_err_out:
1665 iput(bvi);
1666unm_err_out:
1667 if (!err)
1668 err = -EIO;
1669 if (ctx)
1670 ntfs_attr_put_search_ctx(ctx);
1671 if (m)
1672 unmap_mft_record(base_ni);
1673err_out:
1674 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1675 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1676 ni->name_len);
1677 make_bad_inode(vi);
1678 if (err != -EOPNOTSUPP && err != -ENOMEM)
1679 NVolSetErrors(vol);
1680 return err;
1681}
1682
1683/**
1684 * ntfs_read_inode_mount - special read_inode for mount time use only
1685 * @vi: inode to read
1686 *
1687 * Read inode FILE_MFT at mount time, only called with super_block lock
1688 * held from within the read_super() code path.
1689 *
1690 * This function exists because when it is called the page cache for $MFT/$DATA
1691 * is not initialized and hence we cannot get at the contents of mft records
1692 * by calling map_mft_record*().
1693 *
1694 * Further it needs to cope with the circular references problem, i.e. cannot
1695 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1696 * we do not know where the other extent mft records are yet and again, because
1697 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1698 * attribute list is actually present in $MFT inode.
1699 *
1700 * We solve these problems by starting with the $DATA attribute before anything
1701 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1702 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1703 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1704 * sufficient information for the next step to complete.
1705 *
1706 * This should work but there are two possible pit falls (see inline comments
1707 * below), but only time will tell if they are real pits or just smoke...
1708 */
1709int ntfs_read_inode_mount(struct inode *vi)
1710{
1711 VCN next_vcn, last_vcn, highest_vcn;
1712 s64 block;
1713 struct super_block *sb = vi->i_sb;
1714 ntfs_volume *vol = NTFS_SB(sb);
1715 struct buffer_head *bh;
1716 ntfs_inode *ni;
1717 MFT_RECORD *m = NULL;
5ae9fcf8 1718 ATTR_RECORD *a;
1da177e4
LT
1719 ntfs_attr_search_ctx *ctx;
1720 unsigned int i, nr_blocks;
1721 int err;
1722
1723 ntfs_debug("Entering.");
1724
1725 /* Initialize the ntfs specific part of @vi. */
1726 ntfs_init_big_inode(vi);
1727
1728 ni = NTFS_I(vi);
1729
1730 /* Setup the data attribute. It is special as it is mst protected. */
1731 NInoSetNonResident(ni);
1732 NInoSetMstProtected(ni);
c002f425 1733 NInoSetSparseDisabled(ni);
1da177e4
LT
1734 ni->type = AT_DATA;
1735 ni->name = NULL;
1736 ni->name_len = 0;
1da177e4
LT
1737 /*
1738 * This sets up our little cheat allowing us to reuse the async read io
1739 * completion handler for directories.
1740 */
1741 ni->itype.index.block_size = vol->mft_record_size;
1742 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1743
1744 /* Very important! Needed to be able to call map_mft_record*(). */
1745 vol->mft_ino = vi;
1746
1747 /* Allocate enough memory to read the first mft record. */
1748 if (vol->mft_record_size > 64 * 1024) {
1749 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1750 vol->mft_record_size);
1751 goto err_out;
1752 }
1753 i = vol->mft_record_size;
1754 if (i < sb->s_blocksize)
1755 i = sb->s_blocksize;
1756 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1757 if (!m) {
1758 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1759 goto err_out;
1760 }
1761
1762 /* Determine the first block of the $MFT/$DATA attribute. */
1763 block = vol->mft_lcn << vol->cluster_size_bits >>
1764 sb->s_blocksize_bits;
1765 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1766 if (!nr_blocks)
1767 nr_blocks = 1;
1768
1769 /* Load $MFT/$DATA's first mft record. */
1770 for (i = 0; i < nr_blocks; i++) {
1771 bh = sb_bread(sb, block++);
1772 if (!bh) {
1773 ntfs_error(sb, "Device read failed.");
1774 goto err_out;
1775 }
1776 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1777 sb->s_blocksize);
1778 brelse(bh);
1779 }
1780
1781 /* Apply the mst fixups. */
1782 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1783 /* FIXME: Try to use the $MFTMirr now. */
1784 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1785 goto err_out;
1786 }
1787
1788 /* Need this to sanity check attribute list references to $MFT. */
1789 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1790
1791 /* Provides readpage() and sync_page() for map_mft_record(). */
1792 vi->i_mapping->a_ops = &ntfs_mst_aops;
1793
1794 ctx = ntfs_attr_get_search_ctx(ni, m);
1795 if (!ctx) {
1796 err = -ENOMEM;
1797 goto err_out;
1798 }
1799
1800 /* Find the attribute list attribute if present. */
1801 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1802 if (err) {
1803 if (unlikely(err != -ENOENT)) {
1804 ntfs_error(sb, "Failed to lookup attribute list "
1805 "attribute. You should run chkdsk.");
1806 goto put_err_out;
1807 }
1808 } else /* if (!err) */ {
1809 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1810 u8 *al_end;
1811
1812 ntfs_debug("Attribute list attribute found in $MFT.");
1813 NInoSetAttrList(ni);
5ae9fcf8
AA
1814 a = ctx->attr;
1815 if (a->flags & ATTR_IS_ENCRYPTED ||
1816 a->flags & ATTR_COMPRESSION_MASK ||
1817 a->flags & ATTR_IS_SPARSE) {
1da177e4
LT
1818 ntfs_error(sb, "Attribute list attribute is "
1819 "compressed/encrypted/sparse. Not "
1820 "allowed. $MFT is corrupt. You should "
1821 "run chkdsk.");
1822 goto put_err_out;
1823 }
1824 /* Now allocate memory for the attribute list. */
5ae9fcf8 1825 ni->attr_list_size = (u32)ntfs_attr_size(a);
1da177e4
LT
1826 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1827 if (!ni->attr_list) {
1828 ntfs_error(sb, "Not enough memory to allocate buffer "
1829 "for attribute list.");
1830 goto put_err_out;
1831 }
5ae9fcf8 1832 if (a->non_resident) {
1da177e4 1833 NInoSetAttrListNonResident(ni);
5ae9fcf8 1834 if (a->data.non_resident.lowest_vcn) {
1da177e4
LT
1835 ntfs_error(sb, "Attribute list has non zero "
1836 "lowest_vcn. $MFT is corrupt. "
1837 "You should run chkdsk.");
1838 goto put_err_out;
1839 }
1840 /* Setup the runlist. */
1841 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
5ae9fcf8 1842 a, NULL);
1da177e4
LT
1843 if (IS_ERR(ni->attr_list_rl.rl)) {
1844 err = PTR_ERR(ni->attr_list_rl.rl);
1845 ni->attr_list_rl.rl = NULL;
1846 ntfs_error(sb, "Mapping pairs decompression "
1847 "failed with error code %i.",
1848 -err);
1849 goto put_err_out;
1850 }
1851 /* Now load the attribute list. */
1852 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1853 ni->attr_list, ni->attr_list_size,
5ae9fcf8 1854 sle64_to_cpu(a->data.
1da177e4
LT
1855 non_resident.initialized_size)))) {
1856 ntfs_error(sb, "Failed to load attribute list "
1857 "attribute with error code %i.",
1858 -err);
1859 goto put_err_out;
1860 }
1861 } else /* if (!ctx.attr->non_resident) */ {
5ae9fcf8
AA
1862 if ((u8*)a + le16_to_cpu(
1863 a->data.resident.value_offset) +
1da177e4 1864 le32_to_cpu(
5ae9fcf8 1865 a->data.resident.value_length) >
1da177e4
LT
1866 (u8*)ctx->mrec + vol->mft_record_size) {
1867 ntfs_error(sb, "Corrupt attribute list "
1868 "attribute.");
1869 goto put_err_out;
1870 }
1871 /* Now copy the attribute list. */
5ae9fcf8
AA
1872 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1873 a->data.resident.value_offset),
1da177e4 1874 le32_to_cpu(
5ae9fcf8 1875 a->data.resident.value_length));
1da177e4
LT
1876 }
1877 /* The attribute list is now setup in memory. */
1878 /*
1879 * FIXME: I don't know if this case is actually possible.
1880 * According to logic it is not possible but I have seen too
1881 * many weird things in MS software to rely on logic... Thus we
1882 * perform a manual search and make sure the first $MFT/$DATA
1883 * extent is in the base inode. If it is not we abort with an
1884 * error and if we ever see a report of this error we will need
1885 * to do some magic in order to have the necessary mft record
1886 * loaded and in the right place in the page cache. But
1887 * hopefully logic will prevail and this never happens...
1888 */
1889 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1890 al_end = (u8*)al_entry + ni->attr_list_size;
1891 for (;; al_entry = next_al_entry) {
1892 /* Out of bounds check. */
1893 if ((u8*)al_entry < ni->attr_list ||
1894 (u8*)al_entry > al_end)
1895 goto em_put_err_out;
1896 /* Catch the end of the attribute list. */
1897 if ((u8*)al_entry == al_end)
1898 goto em_put_err_out;
1899 if (!al_entry->length)
1900 goto em_put_err_out;
1901 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1902 le16_to_cpu(al_entry->length) > al_end)
1903 goto em_put_err_out;
1904 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1905 le16_to_cpu(al_entry->length));
1906 if (le32_to_cpu(al_entry->type) >
1907 const_le32_to_cpu(AT_DATA))
1908 goto em_put_err_out;
1909 if (AT_DATA != al_entry->type)
1910 continue;
1911 /* We want an unnamed attribute. */
1912 if (al_entry->name_length)
1913 goto em_put_err_out;
1914 /* Want the first entry, i.e. lowest_vcn == 0. */
1915 if (al_entry->lowest_vcn)
1916 goto em_put_err_out;
1917 /* First entry has to be in the base mft record. */
1918 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1919 /* MFT references do not match, logic fails. */
1920 ntfs_error(sb, "BUG: The first $DATA extent "
1921 "of $MFT is not in the base "
1922 "mft record. Please report "
1923 "you saw this message to "
1924 "linux-ntfs-dev@lists."
1925 "sourceforge.net");
1926 goto put_err_out;
1927 } else {
1928 /* Sequence numbers must match. */
1929 if (MSEQNO_LE(al_entry->mft_reference) !=
1930 ni->seq_no)
1931 goto em_put_err_out;
1932 /* Got it. All is ok. We can stop now. */
1933 break;
1934 }
1935 }
1936 }
1937
1938 ntfs_attr_reinit_search_ctx(ctx);
1939
1940 /* Now load all attribute extents. */
5ae9fcf8 1941 a = NULL;
1da177e4
LT
1942 next_vcn = last_vcn = highest_vcn = 0;
1943 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
1944 ctx))) {
1945 runlist_element *nrl;
1946
1947 /* Cache the current attribute. */
5ae9fcf8 1948 a = ctx->attr;
1da177e4 1949 /* $MFT must be non-resident. */
5ae9fcf8 1950 if (!a->non_resident) {
1da177e4
LT
1951 ntfs_error(sb, "$MFT must be non-resident but a "
1952 "resident extent was found. $MFT is "
1953 "corrupt. Run chkdsk.");
1954 goto put_err_out;
1955 }
1956 /* $MFT must be uncompressed and unencrypted. */
5ae9fcf8
AA
1957 if (a->flags & ATTR_COMPRESSION_MASK ||
1958 a->flags & ATTR_IS_ENCRYPTED ||
1959 a->flags & ATTR_IS_SPARSE) {
1da177e4
LT
1960 ntfs_error(sb, "$MFT must be uncompressed, "
1961 "non-sparse, and unencrypted but a "
1962 "compressed/sparse/encrypted extent "
1963 "was found. $MFT is corrupt. Run "
1964 "chkdsk.");
1965 goto put_err_out;
1966 }
1967 /*
1968 * Decompress the mapping pairs array of this extent and merge
1969 * the result into the existing runlist. No need for locking
1970 * as we have exclusive access to the inode at this time and we
1971 * are a mount in progress task, too.
1972 */
5ae9fcf8 1973 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
1da177e4
LT
1974 if (IS_ERR(nrl)) {
1975 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
1976 "failed with error code %ld. $MFT is "
1977 "corrupt.", PTR_ERR(nrl));
1978 goto put_err_out;
1979 }
1980 ni->runlist.rl = nrl;
1981
1982 /* Are we in the first extent? */
1983 if (!next_vcn) {
5ae9fcf8 1984 if (a->data.non_resident.lowest_vcn) {
1da177e4
LT
1985 ntfs_error(sb, "First extent of $DATA "
1986 "attribute has non zero "
1987 "lowest_vcn. $MFT is corrupt. "
1988 "You should run chkdsk.");
1989 goto put_err_out;
1990 }
1991 /* Get the last vcn in the $DATA attribute. */
1992 last_vcn = sle64_to_cpu(
5ae9fcf8 1993 a->data.non_resident.allocated_size)
1da177e4
LT
1994 >> vol->cluster_size_bits;
1995 /* Fill in the inode size. */
1996 vi->i_size = sle64_to_cpu(
5ae9fcf8
AA
1997 a->data.non_resident.data_size);
1998 ni->initialized_size = sle64_to_cpu(
1999 a->data.non_resident.initialized_size);
1da177e4 2000 ni->allocated_size = sle64_to_cpu(
5ae9fcf8 2001 a->data.non_resident.allocated_size);
1da177e4
LT
2002 /*
2003 * Verify the number of mft records does not exceed
2004 * 2^32 - 1.
2005 */
2006 if ((vi->i_size >> vol->mft_record_size_bits) >=
2007 (1ULL << 32)) {
2008 ntfs_error(sb, "$MFT is too big! Aborting.");
2009 goto put_err_out;
2010 }
2011 /*
2012 * We have got the first extent of the runlist for
2013 * $MFT which means it is now relatively safe to call
2014 * the normal ntfs_read_inode() function.
2015 * Complete reading the inode, this will actually
2016 * re-read the mft record for $MFT, this time entering
2017 * it into the page cache with which we complete the
2018 * kick start of the volume. It should be safe to do
2019 * this now as the first extent of $MFT/$DATA is
2020 * already known and we would hope that we don't need
2021 * further extents in order to find the other
2022 * attributes belonging to $MFT. Only time will tell if
2023 * this is really the case. If not we will have to play
2024 * magic at this point, possibly duplicating a lot of
2025 * ntfs_read_inode() at this point. We will need to
2026 * ensure we do enough of its work to be able to call
2027 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2028 * hope this never happens...
2029 */
2030 ntfs_read_locked_inode(vi);
2031 if (is_bad_inode(vi)) {
2032 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2033 "failed. BUG or corrupt $MFT. "
2034 "Run chkdsk and if no errors "
2035 "are found, please report you "
2036 "saw this message to "
2037 "linux-ntfs-dev@lists."
2038 "sourceforge.net");
2039 ntfs_attr_put_search_ctx(ctx);
2040 /* Revert to the safe super operations. */
2041 ntfs_free(m);
2042 return -1;
2043 }
2044 /*
2045 * Re-initialize some specifics about $MFT's inode as
2046 * ntfs_read_inode() will have set up the default ones.
2047 */
2048 /* Set uid and gid to root. */
2049 vi->i_uid = vi->i_gid = 0;
2050 /* Regular file. No access for anyone. */
2051 vi->i_mode = S_IFREG;
2052 /* No VFS initiated operations allowed for $MFT. */
2053 vi->i_op = &ntfs_empty_inode_ops;
2054 vi->i_fop = &ntfs_empty_file_ops;
2055 }
2056
2057 /* Get the lowest vcn for the next extent. */
5ae9fcf8 2058 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
1da177e4
LT
2059 next_vcn = highest_vcn + 1;
2060
2061 /* Only one extent or error, which we catch below. */
2062 if (next_vcn <= 0)
2063 break;
2064
2065 /* Avoid endless loops due to corruption. */
2066 if (next_vcn < sle64_to_cpu(
5ae9fcf8 2067 a->data.non_resident.lowest_vcn)) {
1da177e4
LT
2068 ntfs_error(sb, "$MFT has corrupt attribute list "
2069 "attribute. Run chkdsk.");
2070 goto put_err_out;
2071 }
2072 }
2073 if (err != -ENOENT) {
2074 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2075 "$MFT is corrupt. Run chkdsk.");
2076 goto put_err_out;
2077 }
5ae9fcf8 2078 if (!a) {
1da177e4
LT
2079 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2080 "corrupt. Run chkdsk.");
2081 goto put_err_out;
2082 }
2083 if (highest_vcn && highest_vcn != last_vcn - 1) {
2084 ntfs_error(sb, "Failed to load the complete runlist for "
2085 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2086 "Run chkdsk.");
2087 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2088 (unsigned long long)highest_vcn,
2089 (unsigned long long)last_vcn - 1);
2090 goto put_err_out;
2091 }
2092 ntfs_attr_put_search_ctx(ctx);
2093 ntfs_debug("Done.");
2094 ntfs_free(m);
2095 return 0;
2096
2097em_put_err_out:
2098 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2099 "attribute list. $MFT is corrupt. Run chkdsk.");
2100put_err_out:
2101 ntfs_attr_put_search_ctx(ctx);
2102err_out:
2103 ntfs_error(sb, "Failed. Marking inode as bad.");
2104 make_bad_inode(vi);
2105 ntfs_free(m);
2106 return -1;
2107}
2108
2109/**
2110 * ntfs_put_inode - handler for when the inode reference count is decremented
2111 * @vi: vfs inode
2112 *
2113 * The VFS calls ntfs_put_inode() every time the inode reference count (i_count)
2114 * is about to be decremented (but before the decrement itself.
2115 *
2116 * If the inode @vi is a directory with two references, one of which is being
2117 * dropped, we need to put the attribute inode for the directory index bitmap,
2118 * if it is present, otherwise the directory inode would remain pinned for
2119 * ever.
2120 */
2121void ntfs_put_inode(struct inode *vi)
2122{
2123 if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
2124 ntfs_inode *ni = NTFS_I(vi);
2125 if (NInoIndexAllocPresent(ni)) {
2126 struct inode *bvi = NULL;
2127 down(&vi->i_sem);
2128 if (atomic_read(&vi->i_count) == 2) {
2129 bvi = ni->itype.index.bmp_ino;
2130 if (bvi)
2131 ni->itype.index.bmp_ino = NULL;
2132 }
2133 up(&vi->i_sem);
2134 if (bvi)
2135 iput(bvi);
2136 }
2137 }
2138}
2139
2140static void __ntfs_clear_inode(ntfs_inode *ni)
2141{
2142 /* Free all alocated memory. */
2143 down_write(&ni->runlist.lock);
2144 if (ni->runlist.rl) {
2145 ntfs_free(ni->runlist.rl);
2146 ni->runlist.rl = NULL;
2147 }
2148 up_write(&ni->runlist.lock);
2149
2150 if (ni->attr_list) {
2151 ntfs_free(ni->attr_list);
2152 ni->attr_list = NULL;
2153 }
2154
2155 down_write(&ni->attr_list_rl.lock);
2156 if (ni->attr_list_rl.rl) {
2157 ntfs_free(ni->attr_list_rl.rl);
2158 ni->attr_list_rl.rl = NULL;
2159 }
2160 up_write(&ni->attr_list_rl.lock);
2161
2162 if (ni->name_len && ni->name != I30) {
2163 /* Catch bugs... */
2164 BUG_ON(!ni->name);
2165 kfree(ni->name);
2166 }
2167}
2168
2169void ntfs_clear_extent_inode(ntfs_inode *ni)
2170{
2171 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2172
2173 BUG_ON(NInoAttr(ni));
2174 BUG_ON(ni->nr_extents != -1);
2175
2176#ifdef NTFS_RW
2177 if (NInoDirty(ni)) {
2178 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2179 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2180 "Losing data! This is a BUG!!!");
2181 // FIXME: Do something!!!
2182 }
2183#endif /* NTFS_RW */
2184
2185 __ntfs_clear_inode(ni);
2186
2187 /* Bye, bye... */
2188 ntfs_destroy_extent_inode(ni);
2189}
2190
2191/**
2192 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2193 * @vi: vfs inode pending annihilation
2194 *
2195 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2196 * is called, which deallocates all memory belonging to the NTFS specific part
2197 * of the inode and returns.
2198 *
2199 * If the MFT record is dirty, we commit it before doing anything else.
2200 */
2201void ntfs_clear_big_inode(struct inode *vi)
2202{
2203 ntfs_inode *ni = NTFS_I(vi);
2204
2205 /*
2206 * If the inode @vi is an index inode we need to put the attribute
2207 * inode for the index bitmap, if it is present, otherwise the index
2208 * inode would disappear and the attribute inode for the index bitmap
2209 * would no longer be referenced from anywhere and thus it would remain
2210 * pinned for ever.
2211 */
2212 if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) &&
2213 NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) {
2214 iput(ni->itype.index.bmp_ino);
2215 ni->itype.index.bmp_ino = NULL;
2216 }
2217#ifdef NTFS_RW
2218 if (NInoDirty(ni)) {
2219 BOOL was_bad = (is_bad_inode(vi));
2220
2221 /* Committing the inode also commits all extent inodes. */
2222 ntfs_commit_inode(vi);
2223
2224 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2225 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2226 "0x%lx. Losing data!", vi->i_ino);
2227 // FIXME: Do something!!!
2228 }
2229 }
2230#endif /* NTFS_RW */
2231
2232 /* No need to lock at this stage as no one else has a reference. */
2233 if (ni->nr_extents > 0) {
2234 int i;
2235
2236 for (i = 0; i < ni->nr_extents; i++)
2237 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2238 kfree(ni->ext.extent_ntfs_inos);
2239 }
2240
2241 __ntfs_clear_inode(ni);
2242
2243 if (NInoAttr(ni)) {
2244 /* Release the base inode if we are holding it. */
2245 if (ni->nr_extents == -1) {
2246 iput(VFS_I(ni->ext.base_ntfs_ino));
2247 ni->nr_extents = 0;
2248 ni->ext.base_ntfs_ino = NULL;
2249 }
2250 }
2251 return;
2252}
2253
2254/**
2255 * ntfs_show_options - show mount options in /proc/mounts
2256 * @sf: seq_file in which to write our mount options
2257 * @mnt: vfs mount whose mount options to display
2258 *
2259 * Called by the VFS once for each mounted ntfs volume when someone reads
2260 * /proc/mounts in order to display the NTFS specific mount options of each
2261 * mount. The mount options of the vfs mount @mnt are written to the seq file
2262 * @sf and success is returned.
2263 */
2264int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2265{
2266 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2267 int i;
2268
2269 seq_printf(sf, ",uid=%i", vol->uid);
2270 seq_printf(sf, ",gid=%i", vol->gid);
2271 if (vol->fmask == vol->dmask)
2272 seq_printf(sf, ",umask=0%o", vol->fmask);
2273 else {
2274 seq_printf(sf, ",fmask=0%o", vol->fmask);
2275 seq_printf(sf, ",dmask=0%o", vol->dmask);
2276 }
2277 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2278 if (NVolCaseSensitive(vol))
2279 seq_printf(sf, ",case_sensitive");
2280 if (NVolShowSystemFiles(vol))
2281 seq_printf(sf, ",show_sys_files");
c002f425
AA
2282 if (!NVolSparseEnabled(vol))
2283 seq_printf(sf, ",disable_sparse");
1da177e4
LT
2284 for (i = 0; on_errors_arr[i].val; i++) {
2285 if (on_errors_arr[i].val & vol->on_errors)
2286 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2287 }
2288 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2289 return 0;
2290}
2291
2292#ifdef NTFS_RW
2293
2294/**
2295 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2296 * @vi: inode for which the i_size was changed
2297 *
2298 * We do not support i_size changes yet.
2299 *
2300 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2301 * that the change is allowed.
2302 *
2303 * This implies for us that @vi is a file inode rather than a directory, index,
2304 * or attribute inode as well as that @vi is a base inode.
2305 *
2306 * Returns 0 on success or -errno on error.
2307 *
2308 * Called with ->i_sem held. In all but one case ->i_alloc_sem is held for
2309 * writing. The only case where ->i_alloc_sem is not held is
2310 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2311 * with the current i_size as the offset which means that it is a noop as far
2312 * as ntfs_truncate() is concerned.
2313 */
2314int ntfs_truncate(struct inode *vi)
2315{
2316 ntfs_inode *ni = NTFS_I(vi);
2317 ntfs_volume *vol = ni->vol;
2318 ntfs_attr_search_ctx *ctx;
2319 MFT_RECORD *m;
5ae9fcf8 2320 ATTR_RECORD *a;
1da177e4
LT
2321 const char *te = " Leaving file length out of sync with i_size.";
2322 int err;
2323
2324 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2325 BUG_ON(NInoAttr(ni));
2326 BUG_ON(ni->nr_extents < 0);
2327 m = map_mft_record(ni);
2328 if (IS_ERR(m)) {
2329 err = PTR_ERR(m);
2330 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2331 "(error code %d).%s", vi->i_ino, err, te);
2332 ctx = NULL;
2333 m = NULL;
2334 goto err_out;
2335 }
2336 ctx = ntfs_attr_get_search_ctx(ni, m);
2337 if (unlikely(!ctx)) {
2338 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2339 "inode 0x%lx (not enough memory).%s",
2340 vi->i_ino, te);
2341 err = -ENOMEM;
2342 goto err_out;
2343 }
2344 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2345 CASE_SENSITIVE, 0, NULL, 0, ctx);
2346 if (unlikely(err)) {
2347 if (err == -ENOENT)
2348 ntfs_error(vi->i_sb, "Open attribute is missing from "
2349 "mft record. Inode 0x%lx is corrupt. "
2350 "Run chkdsk.", vi->i_ino);
2351 else
2352 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2353 "inode 0x%lx (error code %d).",
2354 vi->i_ino, err);
2355 goto err_out;
2356 }
5ae9fcf8 2357 a = ctx->attr;
1da177e4 2358 /* If the size has not changed there is nothing to do. */
5ae9fcf8 2359 if (ntfs_attr_size(a) == i_size_read(vi))
1da177e4
LT
2360 goto done;
2361 // TODO: Implement the truncate...
2362 ntfs_error(vi->i_sb, "Inode size has changed but this is not "
2363 "implemented yet. Resetting inode size to old value. "
2364 " This is most likely a bug in the ntfs driver!");
5ae9fcf8 2365 i_size_write(vi, ntfs_attr_size(a));
1da177e4
LT
2366done:
2367 ntfs_attr_put_search_ctx(ctx);
2368 unmap_mft_record(ni);
2369 NInoClearTruncateFailed(ni);
2370 ntfs_debug("Done.");
2371 return 0;
2372err_out:
2373 if (err != -ENOMEM) {
2374 NVolSetErrors(vol);
2375 make_bad_inode(vi);
2376 }
2377 if (ctx)
2378 ntfs_attr_put_search_ctx(ctx);
2379 if (m)
2380 unmap_mft_record(ni);
2381 NInoSetTruncateFailed(ni);
2382 return err;
2383}
2384
2385/**
2386 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2387 * @vi: inode for which the i_size was changed
2388 *
2389 * Wrapper for ntfs_truncate() that has no return value.
2390 *
2391 * See ntfs_truncate() description above for details.
2392 */
2393void ntfs_truncate_vfs(struct inode *vi) {
2394 ntfs_truncate(vi);
2395}
2396
2397/**
2398 * ntfs_setattr - called from notify_change() when an attribute is being changed
2399 * @dentry: dentry whose attributes to change
2400 * @attr: structure describing the attributes and the changes
2401 *
2402 * We have to trap VFS attempts to truncate the file described by @dentry as
2403 * soon as possible, because we do not implement changes in i_size yet. So we
2404 * abort all i_size changes here.
2405 *
2406 * We also abort all changes of user, group, and mode as we do not implement
2407 * the NTFS ACLs yet.
2408 *
2409 * Called with ->i_sem held. For the ATTR_SIZE (i.e. ->truncate) case, also
2410 * called with ->i_alloc_sem held for writing.
2411 *
2412 * Basically this is a copy of generic notify_change() and inode_setattr()
2413 * functionality, except we intercept and abort changes in i_size.
2414 */
2415int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2416{
2417 struct inode *vi = dentry->d_inode;
2418 int err;
2419 unsigned int ia_valid = attr->ia_valid;
2420
2421 err = inode_change_ok(vi, attr);
2422 if (err)
2423 return err;
2424
2425 /* We do not support NTFS ACLs yet. */
2426 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2427 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2428 "supported yet, ignoring.");
2429 err = -EOPNOTSUPP;
2430 goto out;
2431 }
2432
2433 if (ia_valid & ATTR_SIZE) {
2434 if (attr->ia_size != i_size_read(vi)) {
2435 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2436 "supported yet, ignoring.");
2437 err = -EOPNOTSUPP;
2438 // TODO: Implement...
2439 // err = vmtruncate(vi, attr->ia_size);
2440 if (err || ia_valid == ATTR_SIZE)
2441 goto out;
2442 } else {
2443 /*
2444 * We skipped the truncate but must still update
2445 * timestamps.
2446 */
1c7d469d 2447 ia_valid |= ATTR_MTIME | ATTR_CTIME;
1da177e4
LT
2448 }
2449 }
1da177e4 2450 if (ia_valid & ATTR_ATIME)
1c7d469d
AA
2451 vi->i_atime = timespec_trunc(attr->ia_atime,
2452 vi->i_sb->s_time_gran);
1da177e4 2453 if (ia_valid & ATTR_MTIME)
1c7d469d
AA
2454 vi->i_mtime = timespec_trunc(attr->ia_mtime,
2455 vi->i_sb->s_time_gran);
1da177e4 2456 if (ia_valid & ATTR_CTIME)
1c7d469d
AA
2457 vi->i_ctime = timespec_trunc(attr->ia_ctime,
2458 vi->i_sb->s_time_gran);
1da177e4
LT
2459 mark_inode_dirty(vi);
2460out:
2461 return err;
2462}
2463
2464/**
2465 * ntfs_write_inode - write out a dirty inode
2466 * @vi: inode to write out
2467 * @sync: if true, write out synchronously
2468 *
2469 * Write out a dirty inode to disk including any extent inodes if present.
2470 *
2471 * If @sync is true, commit the inode to disk and wait for io completion. This
2472 * is done using write_mft_record().
2473 *
2474 * If @sync is false, just schedule the write to happen but do not wait for i/o
2475 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2476 * marking the page (and in this case mft record) dirty but we do not implement
2477 * this yet as write_mft_record() largely ignores the @sync parameter and
2478 * always performs synchronous writes.
2479 *
2480 * Return 0 on success and -errno on error.
2481 */
2482int ntfs_write_inode(struct inode *vi, int sync)
2483{
2484 sle64 nt;
2485 ntfs_inode *ni = NTFS_I(vi);
2486 ntfs_attr_search_ctx *ctx;
2487 MFT_RECORD *m;
2488 STANDARD_INFORMATION *si;
2489 int err = 0;
2490 BOOL modified = FALSE;
2491
2492 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2493 vi->i_ino);
2494 /*
2495 * Dirty attribute inodes are written via their real inodes so just
2496 * clean them here. Access time updates are taken care off when the
2497 * real inode is written.
2498 */
2499 if (NInoAttr(ni)) {
2500 NInoClearDirty(ni);
2501 ntfs_debug("Done.");
2502 return 0;
2503 }
2504 /* Map, pin, and lock the mft record belonging to the inode. */
2505 m = map_mft_record(ni);
2506 if (IS_ERR(m)) {
2507 err = PTR_ERR(m);
2508 goto err_out;
2509 }
2510 /* Update the access times in the standard information attribute. */
2511 ctx = ntfs_attr_get_search_ctx(ni, m);
2512 if (unlikely(!ctx)) {
2513 err = -ENOMEM;
2514 goto unm_err_out;
2515 }
2516 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
2517 CASE_SENSITIVE, 0, NULL, 0, ctx);
2518 if (unlikely(err)) {
2519 ntfs_attr_put_search_ctx(ctx);
2520 goto unm_err_out;
2521 }
2522 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
2523 le16_to_cpu(ctx->attr->data.resident.value_offset));
2524 /* Update the access times if they have changed. */
2525 nt = utc2ntfs(vi->i_mtime);
2526 if (si->last_data_change_time != nt) {
2527 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
8907547d 2528 "new = 0x%llx", vi->i_ino, (long long)
1da177e4 2529 sle64_to_cpu(si->last_data_change_time),
8907547d 2530 (long long)sle64_to_cpu(nt));
1da177e4
LT
2531 si->last_data_change_time = nt;
2532 modified = TRUE;
2533 }
2534 nt = utc2ntfs(vi->i_ctime);
2535 if (si->last_mft_change_time != nt) {
2536 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
8907547d 2537 "new = 0x%llx", vi->i_ino, (long long)
1da177e4 2538 sle64_to_cpu(si->last_mft_change_time),
8907547d 2539 (long long)sle64_to_cpu(nt));
1da177e4
LT
2540 si->last_mft_change_time = nt;
2541 modified = TRUE;
2542 }
2543 nt = utc2ntfs(vi->i_atime);
2544 if (si->last_access_time != nt) {
2545 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
2546 "new = 0x%llx", vi->i_ino,
8907547d
RD
2547 (long long)sle64_to_cpu(si->last_access_time),
2548 (long long)sle64_to_cpu(nt));
1da177e4
LT
2549 si->last_access_time = nt;
2550 modified = TRUE;
2551 }
2552 /*
2553 * If we just modified the standard information attribute we need to
2554 * mark the mft record it is in dirty. We do this manually so that
2555 * mark_inode_dirty() is not called which would redirty the inode and
2556 * hence result in an infinite loop of trying to write the inode.
2557 * There is no need to mark the base inode nor the base mft record
2558 * dirty, since we are going to write this mft record below in any case
2559 * and the base mft record may actually not have been modified so it
2560 * might not need to be written out.
2561 * NOTE: It is not a problem when the inode for $MFT itself is being
2562 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
2563 * on the $MFT inode and hence ntfs_write_inode() will not be
2564 * re-invoked because of it which in turn is ok since the dirtied mft
2565 * record will be cleaned and written out to disk below, i.e. before
2566 * this function returns.
2567 */
2568 if (modified && !NInoTestSetDirty(ctx->ntfs_ino))
2569 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
2570 ctx->ntfs_ino->page_ofs);
2571 ntfs_attr_put_search_ctx(ctx);
2572 /* Now the access times are updated, write the base mft record. */
2573 if (NInoDirty(ni))
2574 err = write_mft_record(ni, m, sync);
2575 /* Write all attached extent mft records. */
2576 down(&ni->extent_lock);
2577 if (ni->nr_extents > 0) {
2578 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
2579 int i;
2580
2581 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
2582 for (i = 0; i < ni->nr_extents; i++) {
2583 ntfs_inode *tni = extent_nis[i];
2584
2585 if (NInoDirty(tni)) {
2586 MFT_RECORD *tm = map_mft_record(tni);
2587 int ret;
2588
2589 if (IS_ERR(tm)) {
2590 if (!err || err == -ENOMEM)
2591 err = PTR_ERR(tm);
2592 continue;
2593 }
2594 ret = write_mft_record(tni, tm, sync);
2595 unmap_mft_record(tni);
2596 if (unlikely(ret)) {
2597 if (!err || err == -ENOMEM)
2598 err = ret;
2599 }
2600 }
2601 }
2602 }
2603 up(&ni->extent_lock);
2604 unmap_mft_record(ni);
2605 if (unlikely(err))
2606 goto err_out;
2607 ntfs_debug("Done.");
2608 return 0;
2609unm_err_out:
2610 unmap_mft_record(ni);
2611err_out:
2612 if (err == -ENOMEM) {
2613 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
2614 "Marking the inode dirty again, so the VFS "
2615 "retries later.");
2616 mark_inode_dirty(vi);
2617 } else {
2618 ntfs_error(vi->i_sb, "Failed (error code %i): Marking inode "
2619 "as bad. You should run chkdsk.", -err);
2620 make_bad_inode(vi);
2621 NVolSetErrors(ni->vol);
2622 }
2623 return err;
2624}
2625
2626#endif /* NTFS_RW */