x86: disable BTS ptrace extensions for now
[wrt350n-kernel.git] / fs / ntfs / inode.c
blobe9da092e27723a91d5c2bc9bd60a1645b494ca00
1 /**
2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
4 * Copyright (c) 2001-2007 Anton Altaparmakov
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.
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.
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
22 #include <linux/buffer_head.h>
23 #include <linux/fs.h>
24 #include <linux/mm.h>
25 #include <linux/mount.h>
26 #include <linux/mutex.h>
27 #include <linux/pagemap.h>
28 #include <linux/quotaops.h>
29 #include <linux/slab.h>
31 #include "aops.h"
32 #include "attrib.h"
33 #include "bitmap.h"
34 #include "dir.h"
35 #include "debug.h"
36 #include "inode.h"
37 #include "lcnalloc.h"
38 #include "malloc.h"
39 #include "mft.h"
40 #include "time.h"
41 #include "ntfs.h"
43 /**
44 * ntfs_test_inode - compare two (possibly fake) inodes for equality
45 * @vi: vfs inode which to test
46 * @na: ntfs attribute which is being tested with
48 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
49 * inode @vi for equality with the ntfs attribute @na.
51 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
52 * @na->name and @na->name_len are then ignored.
54 * Return 1 if the attributes match and 0 if not.
56 * NOTE: This function runs with the inode_lock spin lock held so it is not
57 * allowed to sleep.
59 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
61 ntfs_inode *ni;
63 if (vi->i_ino != na->mft_no)
64 return 0;
65 ni = NTFS_I(vi);
66 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
67 if (likely(!NInoAttr(ni))) {
68 /* If not looking for a normal inode this is a mismatch. */
69 if (unlikely(na->type != AT_UNUSED))
70 return 0;
71 } else {
72 /* A fake inode describing an attribute. */
73 if (ni->type != na->type)
74 return 0;
75 if (ni->name_len != na->name_len)
76 return 0;
77 if (na->name_len && memcmp(ni->name, na->name,
78 na->name_len * sizeof(ntfschar)))
79 return 0;
81 /* Match! */
82 return 1;
85 /**
86 * ntfs_init_locked_inode - initialize an inode
87 * @vi: vfs inode to initialize
88 * @na: ntfs attribute which to initialize @vi to
90 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
91 * order to enable ntfs_test_inode() to do its work.
93 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
94 * In that case, @na->name and @na->name_len should be set to NULL and 0,
95 * respectively. Although that is not strictly necessary as
96 * ntfs_read_locked_inode() will fill them in later.
98 * Return 0 on success and -errno on error.
100 * NOTE: This function runs with the inode_lock spin lock held so it is not
101 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
103 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
105 ntfs_inode *ni = NTFS_I(vi);
107 vi->i_ino = na->mft_no;
109 ni->type = na->type;
110 if (na->type == AT_INDEX_ALLOCATION)
111 NInoSetMstProtected(ni);
113 ni->name = na->name;
114 ni->name_len = na->name_len;
116 /* If initializing a normal inode, we are done. */
117 if (likely(na->type == AT_UNUSED)) {
118 BUG_ON(na->name);
119 BUG_ON(na->name_len);
120 return 0;
123 /* It is a fake inode. */
124 NInoSetAttr(ni);
127 * We have I30 global constant as an optimization as it is the name
128 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
129 * allocation but that is ok. And most attributes are unnamed anyway,
130 * thus the fraction of named attributes with name != I30 is actually
131 * absolutely tiny.
133 if (na->name_len && na->name != I30) {
134 unsigned int i;
136 BUG_ON(!na->name);
137 i = na->name_len * sizeof(ntfschar);
138 ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
139 if (!ni->name)
140 return -ENOMEM;
141 memcpy(ni->name, na->name, i);
142 ni->name[na->name_len] = 0;
144 return 0;
147 typedef int (*set_t)(struct inode *, void *);
148 static int ntfs_read_locked_inode(struct inode *vi);
149 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
150 static int ntfs_read_locked_index_inode(struct inode *base_vi,
151 struct inode *vi);
154 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
155 * @sb: super block of mounted volume
156 * @mft_no: mft record number / inode number to obtain
158 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
159 * file or directory).
161 * If the inode is in the cache, it is just returned with an increased
162 * reference count. Otherwise, a new struct inode is allocated and initialized,
163 * and finally ntfs_read_locked_inode() is called to read in the inode and
164 * fill in the remainder of the inode structure.
166 * Return the struct inode on success. Check the return value with IS_ERR() and
167 * if true, the function failed and the error code is obtained from PTR_ERR().
169 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
171 struct inode *vi;
172 int err;
173 ntfs_attr na;
175 na.mft_no = mft_no;
176 na.type = AT_UNUSED;
177 na.name = NULL;
178 na.name_len = 0;
180 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
181 (set_t)ntfs_init_locked_inode, &na);
182 if (unlikely(!vi))
183 return ERR_PTR(-ENOMEM);
185 err = 0;
187 /* If this is a freshly allocated inode, need to read it now. */
188 if (vi->i_state & I_NEW) {
189 err = ntfs_read_locked_inode(vi);
190 unlock_new_inode(vi);
193 * There is no point in keeping bad inodes around if the failure was
194 * due to ENOMEM. We want to be able to retry again later.
196 if (unlikely(err == -ENOMEM)) {
197 iput(vi);
198 vi = ERR_PTR(err);
200 return vi;
204 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
205 * @base_vi: vfs base inode containing the attribute
206 * @type: attribute type
207 * @name: Unicode name of the attribute (NULL if unnamed)
208 * @name_len: length of @name in Unicode characters (0 if unnamed)
210 * Obtain the (fake) struct inode corresponding to the attribute specified by
211 * @type, @name, and @name_len, which is present in the base mft record
212 * specified by the vfs inode @base_vi.
214 * If the attribute inode is in the cache, it is just returned with an
215 * increased reference count. Otherwise, a new struct inode is allocated and
216 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
217 * attribute and fill in the inode structure.
219 * Note, for index allocation attributes, you need to use ntfs_index_iget()
220 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
222 * Return the struct inode of the attribute inode on success. Check the return
223 * value with IS_ERR() and if true, the function failed and the error code is
224 * obtained from PTR_ERR().
226 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
227 ntfschar *name, u32 name_len)
229 struct inode *vi;
230 int err;
231 ntfs_attr na;
233 /* Make sure no one calls ntfs_attr_iget() for indices. */
234 BUG_ON(type == AT_INDEX_ALLOCATION);
236 na.mft_no = base_vi->i_ino;
237 na.type = type;
238 na.name = name;
239 na.name_len = name_len;
241 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
242 (set_t)ntfs_init_locked_inode, &na);
243 if (unlikely(!vi))
244 return ERR_PTR(-ENOMEM);
246 err = 0;
248 /* If this is a freshly allocated inode, need to read it now. */
249 if (vi->i_state & I_NEW) {
250 err = ntfs_read_locked_attr_inode(base_vi, vi);
251 unlock_new_inode(vi);
254 * There is no point in keeping bad attribute inodes around. This also
255 * simplifies things in that we never need to check for bad attribute
256 * inodes elsewhere.
258 if (unlikely(err)) {
259 iput(vi);
260 vi = ERR_PTR(err);
262 return vi;
266 * ntfs_index_iget - obtain a struct inode corresponding to an index
267 * @base_vi: vfs base inode containing the index related attributes
268 * @name: Unicode name of the index
269 * @name_len: length of @name in Unicode characters
271 * Obtain the (fake) struct inode corresponding to the index specified by @name
272 * and @name_len, which is present in the base mft record specified by the vfs
273 * inode @base_vi.
275 * If the index inode is in the cache, it is just returned with an increased
276 * reference count. Otherwise, a new struct inode is allocated and
277 * initialized, and finally ntfs_read_locked_index_inode() is called to read
278 * the index related attributes and fill in the inode structure.
280 * Return the struct inode of the index inode on success. Check the return
281 * value with IS_ERR() and if true, the function failed and the error code is
282 * obtained from PTR_ERR().
284 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
285 u32 name_len)
287 struct inode *vi;
288 int err;
289 ntfs_attr na;
291 na.mft_no = base_vi->i_ino;
292 na.type = AT_INDEX_ALLOCATION;
293 na.name = name;
294 na.name_len = name_len;
296 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
297 (set_t)ntfs_init_locked_inode, &na);
298 if (unlikely(!vi))
299 return ERR_PTR(-ENOMEM);
301 err = 0;
303 /* If this is a freshly allocated inode, need to read it now. */
304 if (vi->i_state & I_NEW) {
305 err = ntfs_read_locked_index_inode(base_vi, vi);
306 unlock_new_inode(vi);
309 * There is no point in keeping bad index inodes around. This also
310 * simplifies things in that we never need to check for bad index
311 * inodes elsewhere.
313 if (unlikely(err)) {
314 iput(vi);
315 vi = ERR_PTR(err);
317 return vi;
320 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
322 ntfs_inode *ni;
324 ntfs_debug("Entering.");
325 ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_NOFS);
326 if (likely(ni != NULL)) {
327 ni->state = 0;
328 return VFS_I(ni);
330 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
331 return NULL;
334 void ntfs_destroy_big_inode(struct inode *inode)
336 ntfs_inode *ni = NTFS_I(inode);
338 ntfs_debug("Entering.");
339 BUG_ON(ni->page);
340 if (!atomic_dec_and_test(&ni->count))
341 BUG();
342 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
345 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
347 ntfs_inode *ni;
349 ntfs_debug("Entering.");
350 ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS);
351 if (likely(ni != NULL)) {
352 ni->state = 0;
353 return ni;
355 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
356 return NULL;
359 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
361 ntfs_debug("Entering.");
362 BUG_ON(ni->page);
363 if (!atomic_dec_and_test(&ni->count))
364 BUG();
365 kmem_cache_free(ntfs_inode_cache, ni);
369 * The attribute runlist lock has separate locking rules from the
370 * normal runlist lock, so split the two lock-classes:
372 static struct lock_class_key attr_list_rl_lock_class;
375 * __ntfs_init_inode - initialize ntfs specific part of an inode
376 * @sb: super block of mounted volume
377 * @ni: freshly allocated ntfs inode which to initialize
379 * Initialize an ntfs inode to defaults.
381 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
382 * untouched. Make sure to initialize them elsewhere.
384 * Return zero on success and -ENOMEM on error.
386 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
388 ntfs_debug("Entering.");
389 rwlock_init(&ni->size_lock);
390 ni->initialized_size = ni->allocated_size = 0;
391 ni->seq_no = 0;
392 atomic_set(&ni->count, 1);
393 ni->vol = NTFS_SB(sb);
394 ntfs_init_runlist(&ni->runlist);
395 mutex_init(&ni->mrec_lock);
396 ni->page = NULL;
397 ni->page_ofs = 0;
398 ni->attr_list_size = 0;
399 ni->attr_list = NULL;
400 ntfs_init_runlist(&ni->attr_list_rl);
401 lockdep_set_class(&ni->attr_list_rl.lock,
402 &attr_list_rl_lock_class);
403 ni->itype.index.block_size = 0;
404 ni->itype.index.vcn_size = 0;
405 ni->itype.index.collation_rule = 0;
406 ni->itype.index.block_size_bits = 0;
407 ni->itype.index.vcn_size_bits = 0;
408 mutex_init(&ni->extent_lock);
409 ni->nr_extents = 0;
410 ni->ext.base_ntfs_ino = NULL;
414 * Extent inodes get MFT-mapped in a nested way, while the base inode
415 * is still mapped. Teach this nesting to the lock validator by creating
416 * a separate class for nested inode's mrec_lock's:
418 static struct lock_class_key extent_inode_mrec_lock_key;
420 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
421 unsigned long mft_no)
423 ntfs_inode *ni = ntfs_alloc_extent_inode();
425 ntfs_debug("Entering.");
426 if (likely(ni != NULL)) {
427 __ntfs_init_inode(sb, ni);
428 lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
429 ni->mft_no = mft_no;
430 ni->type = AT_UNUSED;
431 ni->name = NULL;
432 ni->name_len = 0;
434 return ni;
438 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
439 * @ctx: initialized attribute search context
441 * Search all file name attributes in the inode described by the attribute
442 * search context @ctx and check if any of the names are in the $Extend system
443 * directory.
445 * Return values:
446 * 1: file is in $Extend directory
447 * 0: file is not in $Extend directory
448 * -errno: failed to determine if the file is in the $Extend directory
450 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
452 int nr_links, err;
454 /* Restart search. */
455 ntfs_attr_reinit_search_ctx(ctx);
457 /* Get number of hard links. */
458 nr_links = le16_to_cpu(ctx->mrec->link_count);
460 /* Loop through all hard links. */
461 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
462 ctx))) {
463 FILE_NAME_ATTR *file_name_attr;
464 ATTR_RECORD *attr = ctx->attr;
465 u8 *p, *p2;
467 nr_links--;
469 * Maximum sanity checking as we are called on an inode that
470 * we suspect might be corrupt.
472 p = (u8*)attr + le32_to_cpu(attr->length);
473 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
474 le32_to_cpu(ctx->mrec->bytes_in_use)) {
475 err_corrupt_attr:
476 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
477 "attribute. You should run chkdsk.");
478 return -EIO;
480 if (attr->non_resident) {
481 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
482 "name. You should run chkdsk.");
483 return -EIO;
485 if (attr->flags) {
486 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
487 "invalid flags. You should run "
488 "chkdsk.");
489 return -EIO;
491 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
492 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
493 "name. You should run chkdsk.");
494 return -EIO;
496 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
497 le16_to_cpu(attr->data.resident.value_offset));
498 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
499 if (p2 < (u8*)attr || p2 > p)
500 goto err_corrupt_attr;
501 /* This attribute is ok, but is it in the $Extend directory? */
502 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
503 return 1; /* YES, it's an extended system file. */
505 if (unlikely(err != -ENOENT))
506 return err;
507 if (unlikely(nr_links)) {
508 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
509 "doesn't match number of name attributes. You "
510 "should run chkdsk.");
511 return -EIO;
513 return 0; /* NO, it is not an extended system file. */
517 * ntfs_read_locked_inode - read an inode from its device
518 * @vi: inode to read
520 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
521 * described by @vi into memory from the device.
523 * The only fields in @vi that we need to/can look at when the function is
524 * called are i_sb, pointing to the mounted device's super block, and i_ino,
525 * the number of the inode to load.
527 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
528 * for reading and sets up the necessary @vi fields as well as initializing
529 * the ntfs inode.
531 * Q: What locks are held when the function is called?
532 * A: i_state has I_LOCK set, hence the inode is locked, also
533 * i_count is set to 1, so it is not going to go away
534 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
535 * is allowed to write to them. We should of course be honouring them but
536 * we need to do that using the IS_* macros defined in include/linux/fs.h.
537 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
539 * Return 0 on success and -errno on error. In the error case, the inode will
540 * have had make_bad_inode() executed on it.
542 static int ntfs_read_locked_inode(struct inode *vi)
544 ntfs_volume *vol = NTFS_SB(vi->i_sb);
545 ntfs_inode *ni;
546 struct inode *bvi;
547 MFT_RECORD *m;
548 ATTR_RECORD *a;
549 STANDARD_INFORMATION *si;
550 ntfs_attr_search_ctx *ctx;
551 int err = 0;
553 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
555 /* Setup the generic vfs inode parts now. */
558 * This is for checking whether an inode has changed w.r.t. a file so
559 * that the file can be updated if necessary (compare with f_version).
561 vi->i_version = 1;
563 vi->i_uid = vol->uid;
564 vi->i_gid = vol->gid;
565 vi->i_mode = 0;
568 * Initialize the ntfs specific part of @vi special casing
569 * FILE_MFT which we need to do at mount time.
571 if (vi->i_ino != FILE_MFT)
572 ntfs_init_big_inode(vi);
573 ni = NTFS_I(vi);
575 m = map_mft_record(ni);
576 if (IS_ERR(m)) {
577 err = PTR_ERR(m);
578 goto err_out;
580 ctx = ntfs_attr_get_search_ctx(ni, m);
581 if (!ctx) {
582 err = -ENOMEM;
583 goto unm_err_out;
586 if (!(m->flags & MFT_RECORD_IN_USE)) {
587 ntfs_error(vi->i_sb, "Inode is not in use!");
588 goto unm_err_out;
590 if (m->base_mft_record) {
591 ntfs_error(vi->i_sb, "Inode is an extent inode!");
592 goto unm_err_out;
595 /* Transfer information from mft record into vfs and ntfs inodes. */
596 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
599 * FIXME: Keep in mind that link_count is two for files which have both
600 * a long file name and a short file name as separate entries, so if
601 * we are hiding short file names this will be too high. Either we need
602 * to account for the short file names by subtracting them or we need
603 * to make sure we delete files even though i_nlink is not zero which
604 * might be tricky due to vfs interactions. Need to think about this
605 * some more when implementing the unlink command.
607 vi->i_nlink = le16_to_cpu(m->link_count);
609 * FIXME: Reparse points can have the directory bit set even though
610 * they would be S_IFLNK. Need to deal with this further below when we
611 * implement reparse points / symbolic links but it will do for now.
612 * Also if not a directory, it could be something else, rather than
613 * a regular file. But again, will do for now.
615 /* Everyone gets all permissions. */
616 vi->i_mode |= S_IRWXUGO;
617 /* If read-only, noone gets write permissions. */
618 if (IS_RDONLY(vi))
619 vi->i_mode &= ~S_IWUGO;
620 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
621 vi->i_mode |= S_IFDIR;
623 * Apply the directory permissions mask set in the mount
624 * options.
626 vi->i_mode &= ~vol->dmask;
627 /* Things break without this kludge! */
628 if (vi->i_nlink > 1)
629 vi->i_nlink = 1;
630 } else {
631 vi->i_mode |= S_IFREG;
632 /* Apply the file permissions mask set in the mount options. */
633 vi->i_mode &= ~vol->fmask;
636 * Find the standard information attribute in the mft record. At this
637 * stage we haven't setup the attribute list stuff yet, so this could
638 * in fact fail if the standard information is in an extent record, but
639 * I don't think this actually ever happens.
641 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
642 ctx);
643 if (unlikely(err)) {
644 if (err == -ENOENT) {
646 * TODO: We should be performing a hot fix here (if the
647 * recover mount option is set) by creating a new
648 * attribute.
650 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
651 "is missing.");
653 goto unm_err_out;
655 a = ctx->attr;
656 /* Get the standard information attribute value. */
657 si = (STANDARD_INFORMATION*)((u8*)a +
658 le16_to_cpu(a->data.resident.value_offset));
660 /* Transfer information from the standard information into vi. */
662 * Note: The i_?times do not quite map perfectly onto the NTFS times,
663 * but they are close enough, and in the end it doesn't really matter
664 * that much...
667 * mtime is the last change of the data within the file. Not changed
668 * when only metadata is changed, e.g. a rename doesn't affect mtime.
670 vi->i_mtime = ntfs2utc(si->last_data_change_time);
672 * ctime is the last change of the metadata of the file. This obviously
673 * always changes, when mtime is changed. ctime can be changed on its
674 * own, mtime is then not changed, e.g. when a file is renamed.
676 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
678 * Last access to the data within the file. Not changed during a rename
679 * for example but changed whenever the file is written to.
681 vi->i_atime = ntfs2utc(si->last_access_time);
683 /* Find the attribute list attribute if present. */
684 ntfs_attr_reinit_search_ctx(ctx);
685 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
686 if (err) {
687 if (unlikely(err != -ENOENT)) {
688 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
689 "attribute.");
690 goto unm_err_out;
692 } else /* if (!err) */ {
693 if (vi->i_ino == FILE_MFT)
694 goto skip_attr_list_load;
695 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
696 NInoSetAttrList(ni);
697 a = ctx->attr;
698 if (a->flags & ATTR_COMPRESSION_MASK) {
699 ntfs_error(vi->i_sb, "Attribute list attribute is "
700 "compressed.");
701 goto unm_err_out;
703 if (a->flags & ATTR_IS_ENCRYPTED ||
704 a->flags & ATTR_IS_SPARSE) {
705 if (a->non_resident) {
706 ntfs_error(vi->i_sb, "Non-resident attribute "
707 "list attribute is encrypted/"
708 "sparse.");
709 goto unm_err_out;
711 ntfs_warning(vi->i_sb, "Resident attribute list "
712 "attribute in inode 0x%lx is marked "
713 "encrypted/sparse which is not true. "
714 "However, Windows allows this and "
715 "chkdsk does not detect or correct it "
716 "so we will just ignore the invalid "
717 "flags and pretend they are not set.",
718 vi->i_ino);
720 /* Now allocate memory for the attribute list. */
721 ni->attr_list_size = (u32)ntfs_attr_size(a);
722 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
723 if (!ni->attr_list) {
724 ntfs_error(vi->i_sb, "Not enough memory to allocate "
725 "buffer for attribute list.");
726 err = -ENOMEM;
727 goto unm_err_out;
729 if (a->non_resident) {
730 NInoSetAttrListNonResident(ni);
731 if (a->data.non_resident.lowest_vcn) {
732 ntfs_error(vi->i_sb, "Attribute list has non "
733 "zero lowest_vcn.");
734 goto unm_err_out;
737 * Setup the runlist. No need for locking as we have
738 * exclusive access to the inode at this time.
740 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
741 a, NULL);
742 if (IS_ERR(ni->attr_list_rl.rl)) {
743 err = PTR_ERR(ni->attr_list_rl.rl);
744 ni->attr_list_rl.rl = NULL;
745 ntfs_error(vi->i_sb, "Mapping pairs "
746 "decompression failed.");
747 goto unm_err_out;
749 /* Now load the attribute list. */
750 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
751 ni->attr_list, ni->attr_list_size,
752 sle64_to_cpu(a->data.non_resident.
753 initialized_size)))) {
754 ntfs_error(vi->i_sb, "Failed to load "
755 "attribute list attribute.");
756 goto unm_err_out;
758 } else /* if (!a->non_resident) */ {
759 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
760 + le32_to_cpu(
761 a->data.resident.value_length) >
762 (u8*)ctx->mrec + vol->mft_record_size) {
763 ntfs_error(vi->i_sb, "Corrupt attribute list "
764 "in inode.");
765 goto unm_err_out;
767 /* Now copy the attribute list. */
768 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
769 a->data.resident.value_offset),
770 le32_to_cpu(
771 a->data.resident.value_length));
774 skip_attr_list_load:
776 * If an attribute list is present we now have the attribute list value
777 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
779 if (S_ISDIR(vi->i_mode)) {
780 loff_t bvi_size;
781 ntfs_inode *bni;
782 INDEX_ROOT *ir;
783 u8 *ir_end, *index_end;
785 /* It is a directory, find index root attribute. */
786 ntfs_attr_reinit_search_ctx(ctx);
787 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
788 0, NULL, 0, ctx);
789 if (unlikely(err)) {
790 if (err == -ENOENT) {
791 // FIXME: File is corrupt! Hot-fix with empty
792 // index root attribute if recovery option is
793 // set.
794 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
795 "is missing.");
797 goto unm_err_out;
799 a = ctx->attr;
800 /* Set up the state. */
801 if (unlikely(a->non_resident)) {
802 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
803 "resident.");
804 goto unm_err_out;
806 /* Ensure the attribute name is placed before the value. */
807 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
808 le16_to_cpu(a->data.resident.value_offset)))) {
809 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
810 "placed after the attribute value.");
811 goto unm_err_out;
814 * Compressed/encrypted index root just means that the newly
815 * created files in that directory should be created compressed/
816 * encrypted. However index root cannot be both compressed and
817 * encrypted.
819 if (a->flags & ATTR_COMPRESSION_MASK)
820 NInoSetCompressed(ni);
821 if (a->flags & ATTR_IS_ENCRYPTED) {
822 if (a->flags & ATTR_COMPRESSION_MASK) {
823 ntfs_error(vi->i_sb, "Found encrypted and "
824 "compressed attribute.");
825 goto unm_err_out;
827 NInoSetEncrypted(ni);
829 if (a->flags & ATTR_IS_SPARSE)
830 NInoSetSparse(ni);
831 ir = (INDEX_ROOT*)((u8*)a +
832 le16_to_cpu(a->data.resident.value_offset));
833 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
834 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
835 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
836 "corrupt.");
837 goto unm_err_out;
839 index_end = (u8*)&ir->index +
840 le32_to_cpu(ir->index.index_length);
841 if (index_end > ir_end) {
842 ntfs_error(vi->i_sb, "Directory index is corrupt.");
843 goto unm_err_out;
845 if (ir->type != AT_FILE_NAME) {
846 ntfs_error(vi->i_sb, "Indexed attribute is not "
847 "$FILE_NAME.");
848 goto unm_err_out;
850 if (ir->collation_rule != COLLATION_FILE_NAME) {
851 ntfs_error(vi->i_sb, "Index collation rule is not "
852 "COLLATION_FILE_NAME.");
853 goto unm_err_out;
855 ni->itype.index.collation_rule = ir->collation_rule;
856 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
857 if (ni->itype.index.block_size &
858 (ni->itype.index.block_size - 1)) {
859 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
860 "power of two.",
861 ni->itype.index.block_size);
862 goto unm_err_out;
864 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
865 ntfs_error(vi->i_sb, "Index block size (%u) > "
866 "PAGE_CACHE_SIZE (%ld) is not "
867 "supported. Sorry.",
868 ni->itype.index.block_size,
869 PAGE_CACHE_SIZE);
870 err = -EOPNOTSUPP;
871 goto unm_err_out;
873 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
874 ntfs_error(vi->i_sb, "Index block size (%u) < "
875 "NTFS_BLOCK_SIZE (%i) is not "
876 "supported. Sorry.",
877 ni->itype.index.block_size,
878 NTFS_BLOCK_SIZE);
879 err = -EOPNOTSUPP;
880 goto unm_err_out;
882 ni->itype.index.block_size_bits =
883 ffs(ni->itype.index.block_size) - 1;
884 /* Determine the size of a vcn in the directory index. */
885 if (vol->cluster_size <= ni->itype.index.block_size) {
886 ni->itype.index.vcn_size = vol->cluster_size;
887 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
888 } else {
889 ni->itype.index.vcn_size = vol->sector_size;
890 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
893 /* Setup the index allocation attribute, even if not present. */
894 NInoSetMstProtected(ni);
895 ni->type = AT_INDEX_ALLOCATION;
896 ni->name = I30;
897 ni->name_len = 4;
899 if (!(ir->index.flags & LARGE_INDEX)) {
900 /* No index allocation. */
901 vi->i_size = ni->initialized_size =
902 ni->allocated_size = 0;
903 /* We are done with the mft record, so we release it. */
904 ntfs_attr_put_search_ctx(ctx);
905 unmap_mft_record(ni);
906 m = NULL;
907 ctx = NULL;
908 goto skip_large_dir_stuff;
909 } /* LARGE_INDEX: Index allocation present. Setup state. */
910 NInoSetIndexAllocPresent(ni);
911 /* Find index allocation attribute. */
912 ntfs_attr_reinit_search_ctx(ctx);
913 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
914 CASE_SENSITIVE, 0, NULL, 0, ctx);
915 if (unlikely(err)) {
916 if (err == -ENOENT)
917 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
918 "attribute is not present but "
919 "$INDEX_ROOT indicated it is.");
920 else
921 ntfs_error(vi->i_sb, "Failed to lookup "
922 "$INDEX_ALLOCATION "
923 "attribute.");
924 goto unm_err_out;
926 a = ctx->attr;
927 if (!a->non_resident) {
928 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
929 "is resident.");
930 goto unm_err_out;
933 * Ensure the attribute name is placed before the mapping pairs
934 * array.
936 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
937 le16_to_cpu(
938 a->data.non_resident.mapping_pairs_offset)))) {
939 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
940 "is placed after the mapping pairs "
941 "array.");
942 goto unm_err_out;
944 if (a->flags & ATTR_IS_ENCRYPTED) {
945 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
946 "is encrypted.");
947 goto unm_err_out;
949 if (a->flags & ATTR_IS_SPARSE) {
950 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
951 "is sparse.");
952 goto unm_err_out;
954 if (a->flags & ATTR_COMPRESSION_MASK) {
955 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
956 "is compressed.");
957 goto unm_err_out;
959 if (a->data.non_resident.lowest_vcn) {
960 ntfs_error(vi->i_sb, "First extent of "
961 "$INDEX_ALLOCATION attribute has non "
962 "zero lowest_vcn.");
963 goto unm_err_out;
965 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
966 ni->initialized_size = sle64_to_cpu(
967 a->data.non_resident.initialized_size);
968 ni->allocated_size = sle64_to_cpu(
969 a->data.non_resident.allocated_size);
971 * We are done with the mft record, so we release it. Otherwise
972 * we would deadlock in ntfs_attr_iget().
974 ntfs_attr_put_search_ctx(ctx);
975 unmap_mft_record(ni);
976 m = NULL;
977 ctx = NULL;
978 /* Get the index bitmap attribute inode. */
979 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
980 if (IS_ERR(bvi)) {
981 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
982 err = PTR_ERR(bvi);
983 goto unm_err_out;
985 bni = NTFS_I(bvi);
986 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
987 NInoSparse(bni)) {
988 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
989 "and/or encrypted and/or sparse.");
990 goto iput_unm_err_out;
992 /* Consistency check bitmap size vs. index allocation size. */
993 bvi_size = i_size_read(bvi);
994 if ((bvi_size << 3) < (vi->i_size >>
995 ni->itype.index.block_size_bits)) {
996 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
997 "for index allocation (0x%llx).",
998 bvi_size << 3, vi->i_size);
999 goto iput_unm_err_out;
1001 /* No longer need the bitmap attribute inode. */
1002 iput(bvi);
1003 skip_large_dir_stuff:
1004 /* Setup the operations for this inode. */
1005 vi->i_op = &ntfs_dir_inode_ops;
1006 vi->i_fop = &ntfs_dir_ops;
1007 } else {
1008 /* It is a file. */
1009 ntfs_attr_reinit_search_ctx(ctx);
1011 /* Setup the data attribute, even if not present. */
1012 ni->type = AT_DATA;
1013 ni->name = NULL;
1014 ni->name_len = 0;
1016 /* Find first extent of the unnamed data attribute. */
1017 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1018 if (unlikely(err)) {
1019 vi->i_size = ni->initialized_size =
1020 ni->allocated_size = 0;
1021 if (err != -ENOENT) {
1022 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1023 "attribute.");
1024 goto unm_err_out;
1027 * FILE_Secure does not have an unnamed $DATA
1028 * attribute, so we special case it here.
1030 if (vi->i_ino == FILE_Secure)
1031 goto no_data_attr_special_case;
1033 * Most if not all the system files in the $Extend
1034 * system directory do not have unnamed data
1035 * attributes so we need to check if the parent
1036 * directory of the file is FILE_Extend and if it is
1037 * ignore this error. To do this we need to get the
1038 * name of this inode from the mft record as the name
1039 * contains the back reference to the parent directory.
1041 if (ntfs_is_extended_system_file(ctx) > 0)
1042 goto no_data_attr_special_case;
1043 // FIXME: File is corrupt! Hot-fix with empty data
1044 // attribute if recovery option is set.
1045 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1046 goto unm_err_out;
1048 a = ctx->attr;
1049 /* Setup the state. */
1050 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1051 if (a->flags & ATTR_COMPRESSION_MASK) {
1052 NInoSetCompressed(ni);
1053 if (vol->cluster_size > 4096) {
1054 ntfs_error(vi->i_sb, "Found "
1055 "compressed data but "
1056 "compression is "
1057 "disabled due to "
1058 "cluster size (%i) > "
1059 "4kiB.",
1060 vol->cluster_size);
1061 goto unm_err_out;
1063 if ((a->flags & ATTR_COMPRESSION_MASK)
1064 != ATTR_IS_COMPRESSED) {
1065 ntfs_error(vi->i_sb, "Found unknown "
1066 "compression method "
1067 "or corrupt file.");
1068 goto unm_err_out;
1071 if (a->flags & ATTR_IS_SPARSE)
1072 NInoSetSparse(ni);
1074 if (a->flags & ATTR_IS_ENCRYPTED) {
1075 if (NInoCompressed(ni)) {
1076 ntfs_error(vi->i_sb, "Found encrypted and "
1077 "compressed data.");
1078 goto unm_err_out;
1080 NInoSetEncrypted(ni);
1082 if (a->non_resident) {
1083 NInoSetNonResident(ni);
1084 if (NInoCompressed(ni) || NInoSparse(ni)) {
1085 if (NInoCompressed(ni) && a->data.non_resident.
1086 compression_unit != 4) {
1087 ntfs_error(vi->i_sb, "Found "
1088 "non-standard "
1089 "compression unit (%u "
1090 "instead of 4). "
1091 "Cannot handle this.",
1092 a->data.non_resident.
1093 compression_unit);
1094 err = -EOPNOTSUPP;
1095 goto unm_err_out;
1097 if (a->data.non_resident.compression_unit) {
1098 ni->itype.compressed.block_size = 1U <<
1099 (a->data.non_resident.
1100 compression_unit +
1101 vol->cluster_size_bits);
1102 ni->itype.compressed.block_size_bits =
1103 ffs(ni->itype.
1104 compressed.
1105 block_size) - 1;
1106 ni->itype.compressed.block_clusters =
1107 1U << a->data.
1108 non_resident.
1109 compression_unit;
1110 } else {
1111 ni->itype.compressed.block_size = 0;
1112 ni->itype.compressed.block_size_bits =
1114 ni->itype.compressed.block_clusters =
1117 ni->itype.compressed.size = sle64_to_cpu(
1118 a->data.non_resident.
1119 compressed_size);
1121 if (a->data.non_resident.lowest_vcn) {
1122 ntfs_error(vi->i_sb, "First extent of $DATA "
1123 "attribute has non zero "
1124 "lowest_vcn.");
1125 goto unm_err_out;
1127 vi->i_size = sle64_to_cpu(
1128 a->data.non_resident.data_size);
1129 ni->initialized_size = sle64_to_cpu(
1130 a->data.non_resident.initialized_size);
1131 ni->allocated_size = sle64_to_cpu(
1132 a->data.non_resident.allocated_size);
1133 } else { /* Resident attribute. */
1134 vi->i_size = ni->initialized_size = le32_to_cpu(
1135 a->data.resident.value_length);
1136 ni->allocated_size = le32_to_cpu(a->length) -
1137 le16_to_cpu(
1138 a->data.resident.value_offset);
1139 if (vi->i_size > ni->allocated_size) {
1140 ntfs_error(vi->i_sb, "Resident data attribute "
1141 "is corrupt (size exceeds "
1142 "allocation).");
1143 goto unm_err_out;
1146 no_data_attr_special_case:
1147 /* We are done with the mft record, so we release it. */
1148 ntfs_attr_put_search_ctx(ctx);
1149 unmap_mft_record(ni);
1150 m = NULL;
1151 ctx = NULL;
1152 /* Setup the operations for this inode. */
1153 vi->i_op = &ntfs_file_inode_ops;
1154 vi->i_fop = &ntfs_file_ops;
1156 if (NInoMstProtected(ni))
1157 vi->i_mapping->a_ops = &ntfs_mst_aops;
1158 else
1159 vi->i_mapping->a_ops = &ntfs_aops;
1161 * The number of 512-byte blocks used on disk (for stat). This is in so
1162 * far inaccurate as it doesn't account for any named streams or other
1163 * special non-resident attributes, but that is how Windows works, too,
1164 * so we are at least consistent with Windows, if not entirely
1165 * consistent with the Linux Way. Doing it the Linux Way would cause a
1166 * significant slowdown as it would involve iterating over all
1167 * attributes in the mft record and adding the allocated/compressed
1168 * sizes of all non-resident attributes present to give us the Linux
1169 * correct size that should go into i_blocks (after division by 512).
1171 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1172 vi->i_blocks = ni->itype.compressed.size >> 9;
1173 else
1174 vi->i_blocks = ni->allocated_size >> 9;
1175 ntfs_debug("Done.");
1176 return 0;
1177 iput_unm_err_out:
1178 iput(bvi);
1179 unm_err_out:
1180 if (!err)
1181 err = -EIO;
1182 if (ctx)
1183 ntfs_attr_put_search_ctx(ctx);
1184 if (m)
1185 unmap_mft_record(ni);
1186 err_out:
1187 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1188 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1189 make_bad_inode(vi);
1190 if (err != -EOPNOTSUPP && err != -ENOMEM)
1191 NVolSetErrors(vol);
1192 return err;
1196 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1197 * @base_vi: base inode
1198 * @vi: attribute inode to read
1200 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1201 * attribute inode described by @vi into memory from the base mft record
1202 * described by @base_ni.
1204 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1205 * reading and looks up the attribute described by @vi before setting up the
1206 * necessary fields in @vi as well as initializing the ntfs inode.
1208 * Q: What locks are held when the function is called?
1209 * A: i_state has I_LOCK set, hence the inode is locked, also
1210 * i_count is set to 1, so it is not going to go away
1212 * Return 0 on success and -errno on error. In the error case, the inode will
1213 * have had make_bad_inode() executed on it.
1215 * Note this cannot be called for AT_INDEX_ALLOCATION.
1217 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1219 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1220 ntfs_inode *ni, *base_ni;
1221 MFT_RECORD *m;
1222 ATTR_RECORD *a;
1223 ntfs_attr_search_ctx *ctx;
1224 int err = 0;
1226 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1228 ntfs_init_big_inode(vi);
1230 ni = NTFS_I(vi);
1231 base_ni = NTFS_I(base_vi);
1233 /* Just mirror the values from the base inode. */
1234 vi->i_version = base_vi->i_version;
1235 vi->i_uid = base_vi->i_uid;
1236 vi->i_gid = base_vi->i_gid;
1237 vi->i_nlink = base_vi->i_nlink;
1238 vi->i_mtime = base_vi->i_mtime;
1239 vi->i_ctime = base_vi->i_ctime;
1240 vi->i_atime = base_vi->i_atime;
1241 vi->i_generation = ni->seq_no = base_ni->seq_no;
1243 /* Set inode type to zero but preserve permissions. */
1244 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1246 m = map_mft_record(base_ni);
1247 if (IS_ERR(m)) {
1248 err = PTR_ERR(m);
1249 goto err_out;
1251 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1252 if (!ctx) {
1253 err = -ENOMEM;
1254 goto unm_err_out;
1256 /* Find the attribute. */
1257 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1258 CASE_SENSITIVE, 0, NULL, 0, ctx);
1259 if (unlikely(err))
1260 goto unm_err_out;
1261 a = ctx->attr;
1262 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1263 if (a->flags & ATTR_COMPRESSION_MASK) {
1264 NInoSetCompressed(ni);
1265 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1266 ni->name_len)) {
1267 ntfs_error(vi->i_sb, "Found compressed "
1268 "non-data or named data "
1269 "attribute. Please report "
1270 "you saw this message to "
1271 "linux-ntfs-dev@lists."
1272 "sourceforge.net");
1273 goto unm_err_out;
1275 if (vol->cluster_size > 4096) {
1276 ntfs_error(vi->i_sb, "Found compressed "
1277 "attribute but compression is "
1278 "disabled due to cluster size "
1279 "(%i) > 4kiB.",
1280 vol->cluster_size);
1281 goto unm_err_out;
1283 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1284 ATTR_IS_COMPRESSED) {
1285 ntfs_error(vi->i_sb, "Found unknown "
1286 "compression method.");
1287 goto unm_err_out;
1291 * The compressed/sparse flag set in an index root just means
1292 * to compress all files.
1294 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1295 ntfs_error(vi->i_sb, "Found mst protected attribute "
1296 "but the attribute is %s. Please "
1297 "report you saw this message to "
1298 "linux-ntfs-dev@lists.sourceforge.net",
1299 NInoCompressed(ni) ? "compressed" :
1300 "sparse");
1301 goto unm_err_out;
1303 if (a->flags & ATTR_IS_SPARSE)
1304 NInoSetSparse(ni);
1306 if (a->flags & ATTR_IS_ENCRYPTED) {
1307 if (NInoCompressed(ni)) {
1308 ntfs_error(vi->i_sb, "Found encrypted and compressed "
1309 "data.");
1310 goto unm_err_out;
1313 * The encryption flag set in an index root just means to
1314 * encrypt all files.
1316 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1317 ntfs_error(vi->i_sb, "Found mst protected attribute "
1318 "but the attribute is encrypted. "
1319 "Please report you saw this message "
1320 "to linux-ntfs-dev@lists.sourceforge."
1321 "net");
1322 goto unm_err_out;
1324 if (ni->type != AT_DATA) {
1325 ntfs_error(vi->i_sb, "Found encrypted non-data "
1326 "attribute.");
1327 goto unm_err_out;
1329 NInoSetEncrypted(ni);
1331 if (!a->non_resident) {
1332 /* Ensure the attribute name is placed before the value. */
1333 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1334 le16_to_cpu(a->data.resident.value_offset)))) {
1335 ntfs_error(vol->sb, "Attribute name is placed after "
1336 "the attribute value.");
1337 goto unm_err_out;
1339 if (NInoMstProtected(ni)) {
1340 ntfs_error(vi->i_sb, "Found mst protected attribute "
1341 "but the attribute is resident. "
1342 "Please report you saw this message to "
1343 "linux-ntfs-dev@lists.sourceforge.net");
1344 goto unm_err_out;
1346 vi->i_size = ni->initialized_size = le32_to_cpu(
1347 a->data.resident.value_length);
1348 ni->allocated_size = le32_to_cpu(a->length) -
1349 le16_to_cpu(a->data.resident.value_offset);
1350 if (vi->i_size > ni->allocated_size) {
1351 ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1352 "(size exceeds allocation).");
1353 goto unm_err_out;
1355 } else {
1356 NInoSetNonResident(ni);
1358 * Ensure the attribute name is placed before the mapping pairs
1359 * array.
1361 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1362 le16_to_cpu(
1363 a->data.non_resident.mapping_pairs_offset)))) {
1364 ntfs_error(vol->sb, "Attribute name is placed after "
1365 "the mapping pairs array.");
1366 goto unm_err_out;
1368 if (NInoCompressed(ni) || NInoSparse(ni)) {
1369 if (NInoCompressed(ni) && a->data.non_resident.
1370 compression_unit != 4) {
1371 ntfs_error(vi->i_sb, "Found non-standard "
1372 "compression unit (%u instead "
1373 "of 4). Cannot handle this.",
1374 a->data.non_resident.
1375 compression_unit);
1376 err = -EOPNOTSUPP;
1377 goto unm_err_out;
1379 if (a->data.non_resident.compression_unit) {
1380 ni->itype.compressed.block_size = 1U <<
1381 (a->data.non_resident.
1382 compression_unit +
1383 vol->cluster_size_bits);
1384 ni->itype.compressed.block_size_bits =
1385 ffs(ni->itype.compressed.
1386 block_size) - 1;
1387 ni->itype.compressed.block_clusters = 1U <<
1388 a->data.non_resident.
1389 compression_unit;
1390 } else {
1391 ni->itype.compressed.block_size = 0;
1392 ni->itype.compressed.block_size_bits = 0;
1393 ni->itype.compressed.block_clusters = 0;
1395 ni->itype.compressed.size = sle64_to_cpu(
1396 a->data.non_resident.compressed_size);
1398 if (a->data.non_resident.lowest_vcn) {
1399 ntfs_error(vi->i_sb, "First extent of attribute has "
1400 "non-zero lowest_vcn.");
1401 goto unm_err_out;
1403 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1404 ni->initialized_size = sle64_to_cpu(
1405 a->data.non_resident.initialized_size);
1406 ni->allocated_size = sle64_to_cpu(
1407 a->data.non_resident.allocated_size);
1409 /* Setup the operations for this attribute inode. */
1410 vi->i_op = NULL;
1411 vi->i_fop = NULL;
1412 if (NInoMstProtected(ni))
1413 vi->i_mapping->a_ops = &ntfs_mst_aops;
1414 else
1415 vi->i_mapping->a_ops = &ntfs_aops;
1416 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1417 vi->i_blocks = ni->itype.compressed.size >> 9;
1418 else
1419 vi->i_blocks = ni->allocated_size >> 9;
1421 * Make sure the base inode does not go away and attach it to the
1422 * attribute inode.
1424 igrab(base_vi);
1425 ni->ext.base_ntfs_ino = base_ni;
1426 ni->nr_extents = -1;
1428 ntfs_attr_put_search_ctx(ctx);
1429 unmap_mft_record(base_ni);
1431 ntfs_debug("Done.");
1432 return 0;
1434 unm_err_out:
1435 if (!err)
1436 err = -EIO;
1437 if (ctx)
1438 ntfs_attr_put_search_ctx(ctx);
1439 unmap_mft_record(base_ni);
1440 err_out:
1441 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1442 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1443 "Marking corrupt inode and base inode 0x%lx as bad. "
1444 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1445 base_vi->i_ino);
1446 make_bad_inode(vi);
1447 if (err != -ENOMEM)
1448 NVolSetErrors(vol);
1449 return err;
1453 * ntfs_read_locked_index_inode - read an index inode from its base inode
1454 * @base_vi: base inode
1455 * @vi: index inode to read
1457 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1458 * index inode described by @vi into memory from the base mft record described
1459 * by @base_ni.
1461 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1462 * reading and looks up the attributes relating to the index described by @vi
1463 * before setting up the necessary fields in @vi as well as initializing the
1464 * ntfs inode.
1466 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1467 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1468 * are setup like directory inodes since directories are a special case of
1469 * indices ao they need to be treated in much the same way. Most importantly,
1470 * for small indices the index allocation attribute might not actually exist.
1471 * However, the index root attribute always exists but this does not need to
1472 * have an inode associated with it and this is why we define a new inode type
1473 * index. Also, like for directories, we need to have an attribute inode for
1474 * the bitmap attribute corresponding to the index allocation attribute and we
1475 * can store this in the appropriate field of the inode, just like we do for
1476 * normal directory inodes.
1478 * Q: What locks are held when the function is called?
1479 * A: i_state has I_LOCK set, hence the inode is locked, also
1480 * i_count is set to 1, so it is not going to go away
1482 * Return 0 on success and -errno on error. In the error case, the inode will
1483 * have had make_bad_inode() executed on it.
1485 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1487 loff_t bvi_size;
1488 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1489 ntfs_inode *ni, *base_ni, *bni;
1490 struct inode *bvi;
1491 MFT_RECORD *m;
1492 ATTR_RECORD *a;
1493 ntfs_attr_search_ctx *ctx;
1494 INDEX_ROOT *ir;
1495 u8 *ir_end, *index_end;
1496 int err = 0;
1498 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1499 ntfs_init_big_inode(vi);
1500 ni = NTFS_I(vi);
1501 base_ni = NTFS_I(base_vi);
1502 /* Just mirror the values from the base inode. */
1503 vi->i_version = base_vi->i_version;
1504 vi->i_uid = base_vi->i_uid;
1505 vi->i_gid = base_vi->i_gid;
1506 vi->i_nlink = base_vi->i_nlink;
1507 vi->i_mtime = base_vi->i_mtime;
1508 vi->i_ctime = base_vi->i_ctime;
1509 vi->i_atime = base_vi->i_atime;
1510 vi->i_generation = ni->seq_no = base_ni->seq_no;
1511 /* Set inode type to zero but preserve permissions. */
1512 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1513 /* Map the mft record for the base inode. */
1514 m = map_mft_record(base_ni);
1515 if (IS_ERR(m)) {
1516 err = PTR_ERR(m);
1517 goto err_out;
1519 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1520 if (!ctx) {
1521 err = -ENOMEM;
1522 goto unm_err_out;
1524 /* Find the index root attribute. */
1525 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1526 CASE_SENSITIVE, 0, NULL, 0, ctx);
1527 if (unlikely(err)) {
1528 if (err == -ENOENT)
1529 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1530 "missing.");
1531 goto unm_err_out;
1533 a = ctx->attr;
1534 /* Set up the state. */
1535 if (unlikely(a->non_resident)) {
1536 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1537 goto unm_err_out;
1539 /* Ensure the attribute name is placed before the value. */
1540 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1541 le16_to_cpu(a->data.resident.value_offset)))) {
1542 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1543 "after the attribute value.");
1544 goto unm_err_out;
1547 * Compressed/encrypted/sparse index root is not allowed, except for
1548 * directories of course but those are not dealt with here.
1550 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1551 ATTR_IS_SPARSE)) {
1552 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1553 "root attribute.");
1554 goto unm_err_out;
1556 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1557 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1558 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1559 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1560 goto unm_err_out;
1562 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1563 if (index_end > ir_end) {
1564 ntfs_error(vi->i_sb, "Index is corrupt.");
1565 goto unm_err_out;
1567 if (ir->type) {
1568 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1569 le32_to_cpu(ir->type));
1570 goto unm_err_out;
1572 ni->itype.index.collation_rule = ir->collation_rule;
1573 ntfs_debug("Index collation rule is 0x%x.",
1574 le32_to_cpu(ir->collation_rule));
1575 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1576 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1577 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1578 "two.", ni->itype.index.block_size);
1579 goto unm_err_out;
1581 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1582 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1583 "(%ld) is not supported. Sorry.",
1584 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1585 err = -EOPNOTSUPP;
1586 goto unm_err_out;
1588 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1589 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1590 "(%i) is not supported. Sorry.",
1591 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1592 err = -EOPNOTSUPP;
1593 goto unm_err_out;
1595 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1596 /* Determine the size of a vcn in the index. */
1597 if (vol->cluster_size <= ni->itype.index.block_size) {
1598 ni->itype.index.vcn_size = vol->cluster_size;
1599 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1600 } else {
1601 ni->itype.index.vcn_size = vol->sector_size;
1602 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1604 /* Check for presence of index allocation attribute. */
1605 if (!(ir->index.flags & LARGE_INDEX)) {
1606 /* No index allocation. */
1607 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1608 /* We are done with the mft record, so we release it. */
1609 ntfs_attr_put_search_ctx(ctx);
1610 unmap_mft_record(base_ni);
1611 m = NULL;
1612 ctx = NULL;
1613 goto skip_large_index_stuff;
1614 } /* LARGE_INDEX: Index allocation present. Setup state. */
1615 NInoSetIndexAllocPresent(ni);
1616 /* Find index allocation attribute. */
1617 ntfs_attr_reinit_search_ctx(ctx);
1618 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1619 CASE_SENSITIVE, 0, NULL, 0, ctx);
1620 if (unlikely(err)) {
1621 if (err == -ENOENT)
1622 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1623 "not present but $INDEX_ROOT "
1624 "indicated it is.");
1625 else
1626 ntfs_error(vi->i_sb, "Failed to lookup "
1627 "$INDEX_ALLOCATION attribute.");
1628 goto unm_err_out;
1630 a = ctx->attr;
1631 if (!a->non_resident) {
1632 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1633 "resident.");
1634 goto unm_err_out;
1637 * Ensure the attribute name is placed before the mapping pairs array.
1639 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1640 le16_to_cpu(
1641 a->data.non_resident.mapping_pairs_offset)))) {
1642 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1643 "placed after the mapping pairs array.");
1644 goto unm_err_out;
1646 if (a->flags & ATTR_IS_ENCRYPTED) {
1647 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1648 "encrypted.");
1649 goto unm_err_out;
1651 if (a->flags & ATTR_IS_SPARSE) {
1652 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1653 goto unm_err_out;
1655 if (a->flags & ATTR_COMPRESSION_MASK) {
1656 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1657 "compressed.");
1658 goto unm_err_out;
1660 if (a->data.non_resident.lowest_vcn) {
1661 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1662 "attribute has non zero lowest_vcn.");
1663 goto unm_err_out;
1665 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1666 ni->initialized_size = sle64_to_cpu(
1667 a->data.non_resident.initialized_size);
1668 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1670 * We are done with the mft record, so we release it. Otherwise
1671 * we would deadlock in ntfs_attr_iget().
1673 ntfs_attr_put_search_ctx(ctx);
1674 unmap_mft_record(base_ni);
1675 m = NULL;
1676 ctx = NULL;
1677 /* Get the index bitmap attribute inode. */
1678 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1679 if (IS_ERR(bvi)) {
1680 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1681 err = PTR_ERR(bvi);
1682 goto unm_err_out;
1684 bni = NTFS_I(bvi);
1685 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1686 NInoSparse(bni)) {
1687 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1688 "encrypted and/or sparse.");
1689 goto iput_unm_err_out;
1691 /* Consistency check bitmap size vs. index allocation size. */
1692 bvi_size = i_size_read(bvi);
1693 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1694 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1695 "index allocation (0x%llx).", bvi_size << 3,
1696 vi->i_size);
1697 goto iput_unm_err_out;
1699 iput(bvi);
1700 skip_large_index_stuff:
1701 /* Setup the operations for this index inode. */
1702 vi->i_op = NULL;
1703 vi->i_fop = NULL;
1704 vi->i_mapping->a_ops = &ntfs_mst_aops;
1705 vi->i_blocks = ni->allocated_size >> 9;
1707 * Make sure the base inode doesn't go away and attach it to the
1708 * index inode.
1710 igrab(base_vi);
1711 ni->ext.base_ntfs_ino = base_ni;
1712 ni->nr_extents = -1;
1714 ntfs_debug("Done.");
1715 return 0;
1716 iput_unm_err_out:
1717 iput(bvi);
1718 unm_err_out:
1719 if (!err)
1720 err = -EIO;
1721 if (ctx)
1722 ntfs_attr_put_search_ctx(ctx);
1723 if (m)
1724 unmap_mft_record(base_ni);
1725 err_out:
1726 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1727 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1728 ni->name_len);
1729 make_bad_inode(vi);
1730 if (err != -EOPNOTSUPP && err != -ENOMEM)
1731 NVolSetErrors(vol);
1732 return err;
1736 * The MFT inode has special locking, so teach the lock validator
1737 * about this by splitting off the locking rules of the MFT from
1738 * the locking rules of other inodes. The MFT inode can never be
1739 * accessed from the VFS side (or even internally), only by the
1740 * map_mft functions.
1742 static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
1745 * ntfs_read_inode_mount - special read_inode for mount time use only
1746 * @vi: inode to read
1748 * Read inode FILE_MFT at mount time, only called with super_block lock
1749 * held from within the read_super() code path.
1751 * This function exists because when it is called the page cache for $MFT/$DATA
1752 * is not initialized and hence we cannot get at the contents of mft records
1753 * by calling map_mft_record*().
1755 * Further it needs to cope with the circular references problem, i.e. cannot
1756 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1757 * we do not know where the other extent mft records are yet and again, because
1758 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1759 * attribute list is actually present in $MFT inode.
1761 * We solve these problems by starting with the $DATA attribute before anything
1762 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1763 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1764 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1765 * sufficient information for the next step to complete.
1767 * This should work but there are two possible pit falls (see inline comments
1768 * below), but only time will tell if they are real pits or just smoke...
1770 int ntfs_read_inode_mount(struct inode *vi)
1772 VCN next_vcn, last_vcn, highest_vcn;
1773 s64 block;
1774 struct super_block *sb = vi->i_sb;
1775 ntfs_volume *vol = NTFS_SB(sb);
1776 struct buffer_head *bh;
1777 ntfs_inode *ni;
1778 MFT_RECORD *m = NULL;
1779 ATTR_RECORD *a;
1780 ntfs_attr_search_ctx *ctx;
1781 unsigned int i, nr_blocks;
1782 int err;
1784 ntfs_debug("Entering.");
1786 /* Initialize the ntfs specific part of @vi. */
1787 ntfs_init_big_inode(vi);
1789 ni = NTFS_I(vi);
1791 /* Setup the data attribute. It is special as it is mst protected. */
1792 NInoSetNonResident(ni);
1793 NInoSetMstProtected(ni);
1794 NInoSetSparseDisabled(ni);
1795 ni->type = AT_DATA;
1796 ni->name = NULL;
1797 ni->name_len = 0;
1799 * This sets up our little cheat allowing us to reuse the async read io
1800 * completion handler for directories.
1802 ni->itype.index.block_size = vol->mft_record_size;
1803 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1805 /* Very important! Needed to be able to call map_mft_record*(). */
1806 vol->mft_ino = vi;
1808 /* Allocate enough memory to read the first mft record. */
1809 if (vol->mft_record_size > 64 * 1024) {
1810 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1811 vol->mft_record_size);
1812 goto err_out;
1814 i = vol->mft_record_size;
1815 if (i < sb->s_blocksize)
1816 i = sb->s_blocksize;
1817 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1818 if (!m) {
1819 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1820 goto err_out;
1823 /* Determine the first block of the $MFT/$DATA attribute. */
1824 block = vol->mft_lcn << vol->cluster_size_bits >>
1825 sb->s_blocksize_bits;
1826 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1827 if (!nr_blocks)
1828 nr_blocks = 1;
1830 /* Load $MFT/$DATA's first mft record. */
1831 for (i = 0; i < nr_blocks; i++) {
1832 bh = sb_bread(sb, block++);
1833 if (!bh) {
1834 ntfs_error(sb, "Device read failed.");
1835 goto err_out;
1837 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1838 sb->s_blocksize);
1839 brelse(bh);
1842 /* Apply the mst fixups. */
1843 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1844 /* FIXME: Try to use the $MFTMirr now. */
1845 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1846 goto err_out;
1849 /* Need this to sanity check attribute list references to $MFT. */
1850 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1852 /* Provides readpage() and sync_page() for map_mft_record(). */
1853 vi->i_mapping->a_ops = &ntfs_mst_aops;
1855 ctx = ntfs_attr_get_search_ctx(ni, m);
1856 if (!ctx) {
1857 err = -ENOMEM;
1858 goto err_out;
1861 /* Find the attribute list attribute if present. */
1862 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1863 if (err) {
1864 if (unlikely(err != -ENOENT)) {
1865 ntfs_error(sb, "Failed to lookup attribute list "
1866 "attribute. You should run chkdsk.");
1867 goto put_err_out;
1869 } else /* if (!err) */ {
1870 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1871 u8 *al_end;
1872 static const char *es = " Not allowed. $MFT is corrupt. "
1873 "You should run chkdsk.";
1875 ntfs_debug("Attribute list attribute found in $MFT.");
1876 NInoSetAttrList(ni);
1877 a = ctx->attr;
1878 if (a->flags & ATTR_COMPRESSION_MASK) {
1879 ntfs_error(sb, "Attribute list attribute is "
1880 "compressed.%s", es);
1881 goto put_err_out;
1883 if (a->flags & ATTR_IS_ENCRYPTED ||
1884 a->flags & ATTR_IS_SPARSE) {
1885 if (a->non_resident) {
1886 ntfs_error(sb, "Non-resident attribute list "
1887 "attribute is encrypted/"
1888 "sparse.%s", es);
1889 goto put_err_out;
1891 ntfs_warning(sb, "Resident attribute list attribute "
1892 "in $MFT system file is marked "
1893 "encrypted/sparse which is not true. "
1894 "However, Windows allows this and "
1895 "chkdsk does not detect or correct it "
1896 "so we will just ignore the invalid "
1897 "flags and pretend they are not set.");
1899 /* Now allocate memory for the attribute list. */
1900 ni->attr_list_size = (u32)ntfs_attr_size(a);
1901 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1902 if (!ni->attr_list) {
1903 ntfs_error(sb, "Not enough memory to allocate buffer "
1904 "for attribute list.");
1905 goto put_err_out;
1907 if (a->non_resident) {
1908 NInoSetAttrListNonResident(ni);
1909 if (a->data.non_resident.lowest_vcn) {
1910 ntfs_error(sb, "Attribute list has non zero "
1911 "lowest_vcn. $MFT is corrupt. "
1912 "You should run chkdsk.");
1913 goto put_err_out;
1915 /* Setup the runlist. */
1916 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1917 a, NULL);
1918 if (IS_ERR(ni->attr_list_rl.rl)) {
1919 err = PTR_ERR(ni->attr_list_rl.rl);
1920 ni->attr_list_rl.rl = NULL;
1921 ntfs_error(sb, "Mapping pairs decompression "
1922 "failed with error code %i.",
1923 -err);
1924 goto put_err_out;
1926 /* Now load the attribute list. */
1927 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1928 ni->attr_list, ni->attr_list_size,
1929 sle64_to_cpu(a->data.
1930 non_resident.initialized_size)))) {
1931 ntfs_error(sb, "Failed to load attribute list "
1932 "attribute with error code %i.",
1933 -err);
1934 goto put_err_out;
1936 } else /* if (!ctx.attr->non_resident) */ {
1937 if ((u8*)a + le16_to_cpu(
1938 a->data.resident.value_offset) +
1939 le32_to_cpu(
1940 a->data.resident.value_length) >
1941 (u8*)ctx->mrec + vol->mft_record_size) {
1942 ntfs_error(sb, "Corrupt attribute list "
1943 "attribute.");
1944 goto put_err_out;
1946 /* Now copy the attribute list. */
1947 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1948 a->data.resident.value_offset),
1949 le32_to_cpu(
1950 a->data.resident.value_length));
1952 /* The attribute list is now setup in memory. */
1954 * FIXME: I don't know if this case is actually possible.
1955 * According to logic it is not possible but I have seen too
1956 * many weird things in MS software to rely on logic... Thus we
1957 * perform a manual search and make sure the first $MFT/$DATA
1958 * extent is in the base inode. If it is not we abort with an
1959 * error and if we ever see a report of this error we will need
1960 * to do some magic in order to have the necessary mft record
1961 * loaded and in the right place in the page cache. But
1962 * hopefully logic will prevail and this never happens...
1964 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1965 al_end = (u8*)al_entry + ni->attr_list_size;
1966 for (;; al_entry = next_al_entry) {
1967 /* Out of bounds check. */
1968 if ((u8*)al_entry < ni->attr_list ||
1969 (u8*)al_entry > al_end)
1970 goto em_put_err_out;
1971 /* Catch the end of the attribute list. */
1972 if ((u8*)al_entry == al_end)
1973 goto em_put_err_out;
1974 if (!al_entry->length)
1975 goto em_put_err_out;
1976 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1977 le16_to_cpu(al_entry->length) > al_end)
1978 goto em_put_err_out;
1979 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1980 le16_to_cpu(al_entry->length));
1981 if (le32_to_cpu(al_entry->type) >
1982 const_le32_to_cpu(AT_DATA))
1983 goto em_put_err_out;
1984 if (AT_DATA != al_entry->type)
1985 continue;
1986 /* We want an unnamed attribute. */
1987 if (al_entry->name_length)
1988 goto em_put_err_out;
1989 /* Want the first entry, i.e. lowest_vcn == 0. */
1990 if (al_entry->lowest_vcn)
1991 goto em_put_err_out;
1992 /* First entry has to be in the base mft record. */
1993 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1994 /* MFT references do not match, logic fails. */
1995 ntfs_error(sb, "BUG: The first $DATA extent "
1996 "of $MFT is not in the base "
1997 "mft record. Please report "
1998 "you saw this message to "
1999 "linux-ntfs-dev@lists."
2000 "sourceforge.net");
2001 goto put_err_out;
2002 } else {
2003 /* Sequence numbers must match. */
2004 if (MSEQNO_LE(al_entry->mft_reference) !=
2005 ni->seq_no)
2006 goto em_put_err_out;
2007 /* Got it. All is ok. We can stop now. */
2008 break;
2013 ntfs_attr_reinit_search_ctx(ctx);
2015 /* Now load all attribute extents. */
2016 a = NULL;
2017 next_vcn = last_vcn = highest_vcn = 0;
2018 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
2019 ctx))) {
2020 runlist_element *nrl;
2022 /* Cache the current attribute. */
2023 a = ctx->attr;
2024 /* $MFT must be non-resident. */
2025 if (!a->non_resident) {
2026 ntfs_error(sb, "$MFT must be non-resident but a "
2027 "resident extent was found. $MFT is "
2028 "corrupt. Run chkdsk.");
2029 goto put_err_out;
2031 /* $MFT must be uncompressed and unencrypted. */
2032 if (a->flags & ATTR_COMPRESSION_MASK ||
2033 a->flags & ATTR_IS_ENCRYPTED ||
2034 a->flags & ATTR_IS_SPARSE) {
2035 ntfs_error(sb, "$MFT must be uncompressed, "
2036 "non-sparse, and unencrypted but a "
2037 "compressed/sparse/encrypted extent "
2038 "was found. $MFT is corrupt. Run "
2039 "chkdsk.");
2040 goto put_err_out;
2043 * Decompress the mapping pairs array of this extent and merge
2044 * the result into the existing runlist. No need for locking
2045 * as we have exclusive access to the inode at this time and we
2046 * are a mount in progress task, too.
2048 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2049 if (IS_ERR(nrl)) {
2050 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2051 "failed with error code %ld. $MFT is "
2052 "corrupt.", PTR_ERR(nrl));
2053 goto put_err_out;
2055 ni->runlist.rl = nrl;
2057 /* Are we in the first extent? */
2058 if (!next_vcn) {
2059 if (a->data.non_resident.lowest_vcn) {
2060 ntfs_error(sb, "First extent of $DATA "
2061 "attribute has non zero "
2062 "lowest_vcn. $MFT is corrupt. "
2063 "You should run chkdsk.");
2064 goto put_err_out;
2066 /* Get the last vcn in the $DATA attribute. */
2067 last_vcn = sle64_to_cpu(
2068 a->data.non_resident.allocated_size)
2069 >> vol->cluster_size_bits;
2070 /* Fill in the inode size. */
2071 vi->i_size = sle64_to_cpu(
2072 a->data.non_resident.data_size);
2073 ni->initialized_size = sle64_to_cpu(
2074 a->data.non_resident.initialized_size);
2075 ni->allocated_size = sle64_to_cpu(
2076 a->data.non_resident.allocated_size);
2078 * Verify the number of mft records does not exceed
2079 * 2^32 - 1.
2081 if ((vi->i_size >> vol->mft_record_size_bits) >=
2082 (1ULL << 32)) {
2083 ntfs_error(sb, "$MFT is too big! Aborting.");
2084 goto put_err_out;
2087 * We have got the first extent of the runlist for
2088 * $MFT which means it is now relatively safe to call
2089 * the normal ntfs_read_inode() function.
2090 * Complete reading the inode, this will actually
2091 * re-read the mft record for $MFT, this time entering
2092 * it into the page cache with which we complete the
2093 * kick start of the volume. It should be safe to do
2094 * this now as the first extent of $MFT/$DATA is
2095 * already known and we would hope that we don't need
2096 * further extents in order to find the other
2097 * attributes belonging to $MFT. Only time will tell if
2098 * this is really the case. If not we will have to play
2099 * magic at this point, possibly duplicating a lot of
2100 * ntfs_read_inode() at this point. We will need to
2101 * ensure we do enough of its work to be able to call
2102 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2103 * hope this never happens...
2105 ntfs_read_locked_inode(vi);
2106 if (is_bad_inode(vi)) {
2107 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2108 "failed. BUG or corrupt $MFT. "
2109 "Run chkdsk and if no errors "
2110 "are found, please report you "
2111 "saw this message to "
2112 "linux-ntfs-dev@lists."
2113 "sourceforge.net");
2114 ntfs_attr_put_search_ctx(ctx);
2115 /* Revert to the safe super operations. */
2116 ntfs_free(m);
2117 return -1;
2120 * Re-initialize some specifics about $MFT's inode as
2121 * ntfs_read_inode() will have set up the default ones.
2123 /* Set uid and gid to root. */
2124 vi->i_uid = vi->i_gid = 0;
2125 /* Regular file. No access for anyone. */
2126 vi->i_mode = S_IFREG;
2127 /* No VFS initiated operations allowed for $MFT. */
2128 vi->i_op = &ntfs_empty_inode_ops;
2129 vi->i_fop = &ntfs_empty_file_ops;
2132 /* Get the lowest vcn for the next extent. */
2133 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2134 next_vcn = highest_vcn + 1;
2136 /* Only one extent or error, which we catch below. */
2137 if (next_vcn <= 0)
2138 break;
2140 /* Avoid endless loops due to corruption. */
2141 if (next_vcn < sle64_to_cpu(
2142 a->data.non_resident.lowest_vcn)) {
2143 ntfs_error(sb, "$MFT has corrupt attribute list "
2144 "attribute. Run chkdsk.");
2145 goto put_err_out;
2148 if (err != -ENOENT) {
2149 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2150 "$MFT is corrupt. Run chkdsk.");
2151 goto put_err_out;
2153 if (!a) {
2154 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2155 "corrupt. Run chkdsk.");
2156 goto put_err_out;
2158 if (highest_vcn && highest_vcn != last_vcn - 1) {
2159 ntfs_error(sb, "Failed to load the complete runlist for "
2160 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2161 "Run chkdsk.");
2162 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2163 (unsigned long long)highest_vcn,
2164 (unsigned long long)last_vcn - 1);
2165 goto put_err_out;
2167 ntfs_attr_put_search_ctx(ctx);
2168 ntfs_debug("Done.");
2169 ntfs_free(m);
2172 * Split the locking rules of the MFT inode from the
2173 * locking rules of other inodes:
2175 lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
2176 lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
2178 return 0;
2180 em_put_err_out:
2181 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2182 "attribute list. $MFT is corrupt. Run chkdsk.");
2183 put_err_out:
2184 ntfs_attr_put_search_ctx(ctx);
2185 err_out:
2186 ntfs_error(sb, "Failed. Marking inode as bad.");
2187 make_bad_inode(vi);
2188 ntfs_free(m);
2189 return -1;
2192 static void __ntfs_clear_inode(ntfs_inode *ni)
2194 /* Free all alocated memory. */
2195 down_write(&ni->runlist.lock);
2196 if (ni->runlist.rl) {
2197 ntfs_free(ni->runlist.rl);
2198 ni->runlist.rl = NULL;
2200 up_write(&ni->runlist.lock);
2202 if (ni->attr_list) {
2203 ntfs_free(ni->attr_list);
2204 ni->attr_list = NULL;
2207 down_write(&ni->attr_list_rl.lock);
2208 if (ni->attr_list_rl.rl) {
2209 ntfs_free(ni->attr_list_rl.rl);
2210 ni->attr_list_rl.rl = NULL;
2212 up_write(&ni->attr_list_rl.lock);
2214 if (ni->name_len && ni->name != I30) {
2215 /* Catch bugs... */
2216 BUG_ON(!ni->name);
2217 kfree(ni->name);
2221 void ntfs_clear_extent_inode(ntfs_inode *ni)
2223 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2225 BUG_ON(NInoAttr(ni));
2226 BUG_ON(ni->nr_extents != -1);
2228 #ifdef NTFS_RW
2229 if (NInoDirty(ni)) {
2230 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2231 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2232 "Losing data! This is a BUG!!!");
2233 // FIXME: Do something!!!
2235 #endif /* NTFS_RW */
2237 __ntfs_clear_inode(ni);
2239 /* Bye, bye... */
2240 ntfs_destroy_extent_inode(ni);
2244 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2245 * @vi: vfs inode pending annihilation
2247 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2248 * is called, which deallocates all memory belonging to the NTFS specific part
2249 * of the inode and returns.
2251 * If the MFT record is dirty, we commit it before doing anything else.
2253 void ntfs_clear_big_inode(struct inode *vi)
2255 ntfs_inode *ni = NTFS_I(vi);
2257 #ifdef NTFS_RW
2258 if (NInoDirty(ni)) {
2259 bool was_bad = (is_bad_inode(vi));
2261 /* Committing the inode also commits all extent inodes. */
2262 ntfs_commit_inode(vi);
2264 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2265 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2266 "0x%lx. Losing data!", vi->i_ino);
2267 // FIXME: Do something!!!
2270 #endif /* NTFS_RW */
2272 /* No need to lock at this stage as no one else has a reference. */
2273 if (ni->nr_extents > 0) {
2274 int i;
2276 for (i = 0; i < ni->nr_extents; i++)
2277 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2278 kfree(ni->ext.extent_ntfs_inos);
2281 __ntfs_clear_inode(ni);
2283 if (NInoAttr(ni)) {
2284 /* Release the base inode if we are holding it. */
2285 if (ni->nr_extents == -1) {
2286 iput(VFS_I(ni->ext.base_ntfs_ino));
2287 ni->nr_extents = 0;
2288 ni->ext.base_ntfs_ino = NULL;
2291 return;
2295 * ntfs_show_options - show mount options in /proc/mounts
2296 * @sf: seq_file in which to write our mount options
2297 * @mnt: vfs mount whose mount options to display
2299 * Called by the VFS once for each mounted ntfs volume when someone reads
2300 * /proc/mounts in order to display the NTFS specific mount options of each
2301 * mount. The mount options of the vfs mount @mnt are written to the seq file
2302 * @sf and success is returned.
2304 int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2306 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2307 int i;
2309 seq_printf(sf, ",uid=%i", vol->uid);
2310 seq_printf(sf, ",gid=%i", vol->gid);
2311 if (vol->fmask == vol->dmask)
2312 seq_printf(sf, ",umask=0%o", vol->fmask);
2313 else {
2314 seq_printf(sf, ",fmask=0%o", vol->fmask);
2315 seq_printf(sf, ",dmask=0%o", vol->dmask);
2317 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2318 if (NVolCaseSensitive(vol))
2319 seq_printf(sf, ",case_sensitive");
2320 if (NVolShowSystemFiles(vol))
2321 seq_printf(sf, ",show_sys_files");
2322 if (!NVolSparseEnabled(vol))
2323 seq_printf(sf, ",disable_sparse");
2324 for (i = 0; on_errors_arr[i].val; i++) {
2325 if (on_errors_arr[i].val & vol->on_errors)
2326 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2328 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2329 return 0;
2332 #ifdef NTFS_RW
2334 static const char *es = " Leaving inconsistent metadata. Unmount and run "
2335 "chkdsk.";
2338 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2339 * @vi: inode for which the i_size was changed
2341 * We only support i_size changes for normal files at present, i.e. not
2342 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2343 * below.
2345 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2346 * that the change is allowed.
2348 * This implies for us that @vi is a file inode rather than a directory, index,
2349 * or attribute inode as well as that @vi is a base inode.
2351 * Returns 0 on success or -errno on error.
2353 * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for
2354 * writing. The only case in the kernel where ->i_alloc_sem is not held is
2355 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2356 * with the current i_size as the offset. The analogous place in NTFS is in
2357 * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again
2358 * without holding ->i_alloc_sem.
2360 int ntfs_truncate(struct inode *vi)
2362 s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2363 VCN highest_vcn;
2364 unsigned long flags;
2365 ntfs_inode *base_ni, *ni = NTFS_I(vi);
2366 ntfs_volume *vol = ni->vol;
2367 ntfs_attr_search_ctx *ctx;
2368 MFT_RECORD *m;
2369 ATTR_RECORD *a;
2370 const char *te = " Leaving file length out of sync with i_size.";
2371 int err, mp_size, size_change, alloc_change;
2372 u32 attr_len;
2374 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2375 BUG_ON(NInoAttr(ni));
2376 BUG_ON(S_ISDIR(vi->i_mode));
2377 BUG_ON(NInoMstProtected(ni));
2378 BUG_ON(ni->nr_extents < 0);
2379 retry_truncate:
2381 * Lock the runlist for writing and map the mft record to ensure it is
2382 * safe to mess with the attribute runlist and sizes.
2384 down_write(&ni->runlist.lock);
2385 if (!NInoAttr(ni))
2386 base_ni = ni;
2387 else
2388 base_ni = ni->ext.base_ntfs_ino;
2389 m = map_mft_record(base_ni);
2390 if (IS_ERR(m)) {
2391 err = PTR_ERR(m);
2392 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2393 "(error code %d).%s", vi->i_ino, err, te);
2394 ctx = NULL;
2395 m = NULL;
2396 goto old_bad_out;
2398 ctx = ntfs_attr_get_search_ctx(base_ni, m);
2399 if (unlikely(!ctx)) {
2400 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2401 "inode 0x%lx (not enough memory).%s",
2402 vi->i_ino, te);
2403 err = -ENOMEM;
2404 goto old_bad_out;
2406 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2407 CASE_SENSITIVE, 0, NULL, 0, ctx);
2408 if (unlikely(err)) {
2409 if (err == -ENOENT) {
2410 ntfs_error(vi->i_sb, "Open attribute is missing from "
2411 "mft record. Inode 0x%lx is corrupt. "
2412 "Run chkdsk.%s", vi->i_ino, te);
2413 err = -EIO;
2414 } else
2415 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2416 "inode 0x%lx (error code %d).%s",
2417 vi->i_ino, err, te);
2418 goto old_bad_out;
2420 m = ctx->mrec;
2421 a = ctx->attr;
2423 * The i_size of the vfs inode is the new size for the attribute value.
2425 new_size = i_size_read(vi);
2426 /* The current size of the attribute value is the old size. */
2427 old_size = ntfs_attr_size(a);
2428 /* Calculate the new allocated size. */
2429 if (NInoNonResident(ni))
2430 new_alloc_size = (new_size + vol->cluster_size - 1) &
2431 ~(s64)vol->cluster_size_mask;
2432 else
2433 new_alloc_size = (new_size + 7) & ~7;
2434 /* The current allocated size is the old allocated size. */
2435 read_lock_irqsave(&ni->size_lock, flags);
2436 old_alloc_size = ni->allocated_size;
2437 read_unlock_irqrestore(&ni->size_lock, flags);
2439 * The change in the file size. This will be 0 if no change, >0 if the
2440 * size is growing, and <0 if the size is shrinking.
2442 size_change = -1;
2443 if (new_size - old_size >= 0) {
2444 size_change = 1;
2445 if (new_size == old_size)
2446 size_change = 0;
2448 /* As above for the allocated size. */
2449 alloc_change = -1;
2450 if (new_alloc_size - old_alloc_size >= 0) {
2451 alloc_change = 1;
2452 if (new_alloc_size == old_alloc_size)
2453 alloc_change = 0;
2456 * If neither the size nor the allocation are being changed there is
2457 * nothing to do.
2459 if (!size_change && !alloc_change)
2460 goto unm_done;
2461 /* If the size is changing, check if new size is allowed in $AttrDef. */
2462 if (size_change) {
2463 err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2464 if (unlikely(err)) {
2465 if (err == -ERANGE) {
2466 ntfs_error(vol->sb, "Truncate would cause the "
2467 "inode 0x%lx to %simum size "
2468 "for its attribute type "
2469 "(0x%x). Aborting truncate.",
2470 vi->i_ino,
2471 new_size > old_size ? "exceed "
2472 "the max" : "go under the min",
2473 le32_to_cpu(ni->type));
2474 err = -EFBIG;
2475 } else {
2476 ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2477 "attribute type 0x%x. "
2478 "Aborting truncate.",
2479 vi->i_ino,
2480 le32_to_cpu(ni->type));
2481 err = -EIO;
2483 /* Reset the vfs inode size to the old size. */
2484 i_size_write(vi, old_size);
2485 goto err_out;
2488 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2489 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2490 "supported yet for %s files, ignoring.",
2491 NInoCompressed(ni) ? "compressed" :
2492 "encrypted");
2493 err = -EOPNOTSUPP;
2494 goto bad_out;
2496 if (a->non_resident)
2497 goto do_non_resident_truncate;
2498 BUG_ON(NInoNonResident(ni));
2499 /* Resize the attribute record to best fit the new attribute size. */
2500 if (new_size < vol->mft_record_size &&
2501 !ntfs_resident_attr_value_resize(m, a, new_size)) {
2502 /* The resize succeeded! */
2503 flush_dcache_mft_record_page(ctx->ntfs_ino);
2504 mark_mft_record_dirty(ctx->ntfs_ino);
2505 write_lock_irqsave(&ni->size_lock, flags);
2506 /* Update the sizes in the ntfs inode and all is done. */
2507 ni->allocated_size = le32_to_cpu(a->length) -
2508 le16_to_cpu(a->data.resident.value_offset);
2510 * Note ntfs_resident_attr_value_resize() has already done any
2511 * necessary data clearing in the attribute record. When the
2512 * file is being shrunk vmtruncate() will already have cleared
2513 * the top part of the last partial page, i.e. since this is
2514 * the resident case this is the page with index 0. However,
2515 * when the file is being expanded, the page cache page data
2516 * between the old data_size, i.e. old_size, and the new_size
2517 * has not been zeroed. Fortunately, we do not need to zero it
2518 * either since on one hand it will either already be zero due
2519 * to both readpage and writepage clearing partial page data
2520 * beyond i_size in which case there is nothing to do or in the
2521 * case of the file being mmap()ped at the same time, POSIX
2522 * specifies that the behaviour is unspecified thus we do not
2523 * have to do anything. This means that in our implementation
2524 * in the rare case that the file is mmap()ped and a write
2525 * occured into the mmap()ped region just beyond the file size
2526 * and writepage has not yet been called to write out the page
2527 * (which would clear the area beyond the file size) and we now
2528 * extend the file size to incorporate this dirty region
2529 * outside the file size, a write of the page would result in
2530 * this data being written to disk instead of being cleared.
2531 * Given both POSIX and the Linux mmap(2) man page specify that
2532 * this corner case is undefined, we choose to leave it like
2533 * that as this is much simpler for us as we cannot lock the
2534 * relevant page now since we are holding too many ntfs locks
2535 * which would result in a lock reversal deadlock.
2537 ni->initialized_size = new_size;
2538 write_unlock_irqrestore(&ni->size_lock, flags);
2539 goto unm_done;
2541 /* If the above resize failed, this must be an attribute extension. */
2542 BUG_ON(size_change < 0);
2544 * We have to drop all the locks so we can call
2545 * ntfs_attr_make_non_resident(). This could be optimised by try-
2546 * locking the first page cache page and only if that fails dropping
2547 * the locks, locking the page, and redoing all the locking and
2548 * lookups. While this would be a huge optimisation, it is not worth
2549 * it as this is definitely a slow code path as it only ever can happen
2550 * once for any given file.
2552 ntfs_attr_put_search_ctx(ctx);
2553 unmap_mft_record(base_ni);
2554 up_write(&ni->runlist.lock);
2556 * Not enough space in the mft record, try to make the attribute
2557 * non-resident and if successful restart the truncation process.
2559 err = ntfs_attr_make_non_resident(ni, old_size);
2560 if (likely(!err))
2561 goto retry_truncate;
2563 * Could not make non-resident. If this is due to this not being
2564 * permitted for this attribute type or there not being enough space,
2565 * try to make other attributes non-resident. Otherwise fail.
2567 if (unlikely(err != -EPERM && err != -ENOSPC)) {
2568 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2569 "type 0x%x, because the conversion from "
2570 "resident to non-resident attribute failed "
2571 "with error code %i.", vi->i_ino,
2572 (unsigned)le32_to_cpu(ni->type), err);
2573 if (err != -ENOMEM)
2574 err = -EIO;
2575 goto conv_err_out;
2577 /* TODO: Not implemented from here, abort. */
2578 if (err == -ENOSPC)
2579 ntfs_error(vol->sb, "Not enough space in the mft record/on "
2580 "disk for the non-resident attribute value. "
2581 "This case is not implemented yet.");
2582 else /* if (err == -EPERM) */
2583 ntfs_error(vol->sb, "This attribute type may not be "
2584 "non-resident. This case is not implemented "
2585 "yet.");
2586 err = -EOPNOTSUPP;
2587 goto conv_err_out;
2588 #if 0
2589 // TODO: Attempt to make other attributes non-resident.
2590 if (!err)
2591 goto do_resident_extend;
2593 * Both the attribute list attribute and the standard information
2594 * attribute must remain in the base inode. Thus, if this is one of
2595 * these attributes, we have to try to move other attributes out into
2596 * extent mft records instead.
2598 if (ni->type == AT_ATTRIBUTE_LIST ||
2599 ni->type == AT_STANDARD_INFORMATION) {
2600 // TODO: Attempt to move other attributes into extent mft
2601 // records.
2602 err = -EOPNOTSUPP;
2603 if (!err)
2604 goto do_resident_extend;
2605 goto err_out;
2607 // TODO: Attempt to move this attribute to an extent mft record, but
2608 // only if it is not already the only attribute in an mft record in
2609 // which case there would be nothing to gain.
2610 err = -EOPNOTSUPP;
2611 if (!err)
2612 goto do_resident_extend;
2613 /* There is nothing we can do to make enough space. )-: */
2614 goto err_out;
2615 #endif
2616 do_non_resident_truncate:
2617 BUG_ON(!NInoNonResident(ni));
2618 if (alloc_change < 0) {
2619 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2620 if (highest_vcn > 0 &&
2621 old_alloc_size >> vol->cluster_size_bits >
2622 highest_vcn + 1) {
2624 * This attribute has multiple extents. Not yet
2625 * supported.
2627 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2628 "attribute type 0x%x, because the "
2629 "attribute is highly fragmented (it "
2630 "consists of multiple extents) and "
2631 "this case is not implemented yet.",
2632 vi->i_ino,
2633 (unsigned)le32_to_cpu(ni->type));
2634 err = -EOPNOTSUPP;
2635 goto bad_out;
2639 * If the size is shrinking, need to reduce the initialized_size and
2640 * the data_size before reducing the allocation.
2642 if (size_change < 0) {
2644 * Make the valid size smaller (i_size is already up-to-date).
2646 write_lock_irqsave(&ni->size_lock, flags);
2647 if (new_size < ni->initialized_size) {
2648 ni->initialized_size = new_size;
2649 a->data.non_resident.initialized_size =
2650 cpu_to_sle64(new_size);
2652 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2653 write_unlock_irqrestore(&ni->size_lock, flags);
2654 flush_dcache_mft_record_page(ctx->ntfs_ino);
2655 mark_mft_record_dirty(ctx->ntfs_ino);
2656 /* If the allocated size is not changing, we are done. */
2657 if (!alloc_change)
2658 goto unm_done;
2660 * If the size is shrinking it makes no sense for the
2661 * allocation to be growing.
2663 BUG_ON(alloc_change > 0);
2664 } else /* if (size_change >= 0) */ {
2666 * The file size is growing or staying the same but the
2667 * allocation can be shrinking, growing or staying the same.
2669 if (alloc_change > 0) {
2671 * We need to extend the allocation and possibly update
2672 * the data size. If we are updating the data size,
2673 * since we are not touching the initialized_size we do
2674 * not need to worry about the actual data on disk.
2675 * And as far as the page cache is concerned, there
2676 * will be no pages beyond the old data size and any
2677 * partial region in the last page between the old and
2678 * new data size (or the end of the page if the new
2679 * data size is outside the page) does not need to be
2680 * modified as explained above for the resident
2681 * attribute truncate case. To do this, we simply drop
2682 * the locks we hold and leave all the work to our
2683 * friendly helper ntfs_attr_extend_allocation().
2685 ntfs_attr_put_search_ctx(ctx);
2686 unmap_mft_record(base_ni);
2687 up_write(&ni->runlist.lock);
2688 err = ntfs_attr_extend_allocation(ni, new_size,
2689 size_change > 0 ? new_size : -1, -1);
2691 * ntfs_attr_extend_allocation() will have done error
2692 * output already.
2694 goto done;
2696 if (!alloc_change)
2697 goto alloc_done;
2699 /* alloc_change < 0 */
2700 /* Free the clusters. */
2701 nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2702 vol->cluster_size_bits, -1, ctx);
2703 m = ctx->mrec;
2704 a = ctx->attr;
2705 if (unlikely(nr_freed < 0)) {
2706 ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2707 "%lli). Unmount and run chkdsk to recover "
2708 "the lost cluster(s).", (long long)nr_freed);
2709 NVolSetErrors(vol);
2710 nr_freed = 0;
2712 /* Truncate the runlist. */
2713 err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2714 new_alloc_size >> vol->cluster_size_bits);
2716 * If the runlist truncation failed and/or the search context is no
2717 * longer valid, we cannot resize the attribute record or build the
2718 * mapping pairs array thus we mark the inode bad so that no access to
2719 * the freed clusters can happen.
2721 if (unlikely(err || IS_ERR(m))) {
2722 ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2723 IS_ERR(m) ?
2724 "restore attribute search context" :
2725 "truncate attribute runlist",
2726 IS_ERR(m) ? PTR_ERR(m) : err, es);
2727 err = -EIO;
2728 goto bad_out;
2730 /* Get the size for the shrunk mapping pairs array for the runlist. */
2731 mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2732 if (unlikely(mp_size <= 0)) {
2733 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2734 "attribute type 0x%x, because determining the "
2735 "size for the mapping pairs failed with error "
2736 "code %i.%s", vi->i_ino,
2737 (unsigned)le32_to_cpu(ni->type), mp_size, es);
2738 err = -EIO;
2739 goto bad_out;
2742 * Shrink the attribute record for the new mapping pairs array. Note,
2743 * this cannot fail since we are making the attribute smaller thus by
2744 * definition there is enough space to do so.
2746 attr_len = le32_to_cpu(a->length);
2747 err = ntfs_attr_record_resize(m, a, mp_size +
2748 le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2749 BUG_ON(err);
2751 * Generate the mapping pairs array directly into the attribute record.
2753 err = ntfs_mapping_pairs_build(vol, (u8*)a +
2754 le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2755 mp_size, ni->runlist.rl, 0, -1, NULL);
2756 if (unlikely(err)) {
2757 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2758 "attribute type 0x%x, because building the "
2759 "mapping pairs failed with error code %i.%s",
2760 vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2761 err, es);
2762 err = -EIO;
2763 goto bad_out;
2765 /* Update the allocated/compressed size as well as the highest vcn. */
2766 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2767 vol->cluster_size_bits) - 1);
2768 write_lock_irqsave(&ni->size_lock, flags);
2769 ni->allocated_size = new_alloc_size;
2770 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2771 if (NInoSparse(ni) || NInoCompressed(ni)) {
2772 if (nr_freed) {
2773 ni->itype.compressed.size -= nr_freed <<
2774 vol->cluster_size_bits;
2775 BUG_ON(ni->itype.compressed.size < 0);
2776 a->data.non_resident.compressed_size = cpu_to_sle64(
2777 ni->itype.compressed.size);
2778 vi->i_blocks = ni->itype.compressed.size >> 9;
2780 } else
2781 vi->i_blocks = new_alloc_size >> 9;
2782 write_unlock_irqrestore(&ni->size_lock, flags);
2784 * We have shrunk the allocation. If this is a shrinking truncate we
2785 * have already dealt with the initialized_size and the data_size above
2786 * and we are done. If the truncate is only changing the allocation
2787 * and not the data_size, we are also done. If this is an extending
2788 * truncate, need to extend the data_size now which is ensured by the
2789 * fact that @size_change is positive.
2791 alloc_done:
2793 * If the size is growing, need to update it now. If it is shrinking,
2794 * we have already updated it above (before the allocation change).
2796 if (size_change > 0)
2797 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2798 /* Ensure the modified mft record is written out. */
2799 flush_dcache_mft_record_page(ctx->ntfs_ino);
2800 mark_mft_record_dirty(ctx->ntfs_ino);
2801 unm_done:
2802 ntfs_attr_put_search_ctx(ctx);
2803 unmap_mft_record(base_ni);
2804 up_write(&ni->runlist.lock);
2805 done:
2806 /* Update the mtime and ctime on the base inode. */
2807 /* normally ->truncate shouldn't update ctime or mtime,
2808 * but ntfs did before so it got a copy & paste version
2809 * of file_update_time. one day someone should fix this
2810 * for real.
2812 if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2813 struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb);
2814 int sync_it = 0;
2816 if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2817 !timespec_equal(&VFS_I(base_ni)->i_ctime, &now))
2818 sync_it = 1;
2819 VFS_I(base_ni)->i_mtime = now;
2820 VFS_I(base_ni)->i_ctime = now;
2822 if (sync_it)
2823 mark_inode_dirty_sync(VFS_I(base_ni));
2826 if (likely(!err)) {
2827 NInoClearTruncateFailed(ni);
2828 ntfs_debug("Done.");
2830 return err;
2831 old_bad_out:
2832 old_size = -1;
2833 bad_out:
2834 if (err != -ENOMEM && err != -EOPNOTSUPP)
2835 NVolSetErrors(vol);
2836 if (err != -EOPNOTSUPP)
2837 NInoSetTruncateFailed(ni);
2838 else if (old_size >= 0)
2839 i_size_write(vi, old_size);
2840 err_out:
2841 if (ctx)
2842 ntfs_attr_put_search_ctx(ctx);
2843 if (m)
2844 unmap_mft_record(base_ni);
2845 up_write(&ni->runlist.lock);
2846 out:
2847 ntfs_debug("Failed. Returning error code %i.", err);
2848 return err;
2849 conv_err_out:
2850 if (err != -ENOMEM && err != -EOPNOTSUPP)
2851 NVolSetErrors(vol);
2852 if (err != -EOPNOTSUPP)
2853 NInoSetTruncateFailed(ni);
2854 else
2855 i_size_write(vi, old_size);
2856 goto out;
2860 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2861 * @vi: inode for which the i_size was changed
2863 * Wrapper for ntfs_truncate() that has no return value.
2865 * See ntfs_truncate() description above for details.
2867 void ntfs_truncate_vfs(struct inode *vi) {
2868 ntfs_truncate(vi);
2872 * ntfs_setattr - called from notify_change() when an attribute is being changed
2873 * @dentry: dentry whose attributes to change
2874 * @attr: structure describing the attributes and the changes
2876 * We have to trap VFS attempts to truncate the file described by @dentry as
2877 * soon as possible, because we do not implement changes in i_size yet. So we
2878 * abort all i_size changes here.
2880 * We also abort all changes of user, group, and mode as we do not implement
2881 * the NTFS ACLs yet.
2883 * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also
2884 * called with ->i_alloc_sem held for writing.
2886 * Basically this is a copy of generic notify_change() and inode_setattr()
2887 * functionality, except we intercept and abort changes in i_size.
2889 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2891 struct inode *vi = dentry->d_inode;
2892 int err;
2893 unsigned int ia_valid = attr->ia_valid;
2895 err = inode_change_ok(vi, attr);
2896 if (err)
2897 goto out;
2898 /* We do not support NTFS ACLs yet. */
2899 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2900 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2901 "supported yet, ignoring.");
2902 err = -EOPNOTSUPP;
2903 goto out;
2905 if (ia_valid & ATTR_SIZE) {
2906 if (attr->ia_size != i_size_read(vi)) {
2907 ntfs_inode *ni = NTFS_I(vi);
2909 * FIXME: For now we do not support resizing of
2910 * compressed or encrypted files yet.
2912 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2913 ntfs_warning(vi->i_sb, "Changes in inode size "
2914 "are not supported yet for "
2915 "%s files, ignoring.",
2916 NInoCompressed(ni) ?
2917 "compressed" : "encrypted");
2918 err = -EOPNOTSUPP;
2919 } else
2920 err = vmtruncate(vi, attr->ia_size);
2921 if (err || ia_valid == ATTR_SIZE)
2922 goto out;
2923 } else {
2925 * We skipped the truncate but must still update
2926 * timestamps.
2928 ia_valid |= ATTR_MTIME | ATTR_CTIME;
2931 if (ia_valid & ATTR_ATIME)
2932 vi->i_atime = timespec_trunc(attr->ia_atime,
2933 vi->i_sb->s_time_gran);
2934 if (ia_valid & ATTR_MTIME)
2935 vi->i_mtime = timespec_trunc(attr->ia_mtime,
2936 vi->i_sb->s_time_gran);
2937 if (ia_valid & ATTR_CTIME)
2938 vi->i_ctime = timespec_trunc(attr->ia_ctime,
2939 vi->i_sb->s_time_gran);
2940 mark_inode_dirty(vi);
2941 out:
2942 return err;
2946 * ntfs_write_inode - write out a dirty inode
2947 * @vi: inode to write out
2948 * @sync: if true, write out synchronously
2950 * Write out a dirty inode to disk including any extent inodes if present.
2952 * If @sync is true, commit the inode to disk and wait for io completion. This
2953 * is done using write_mft_record().
2955 * If @sync is false, just schedule the write to happen but do not wait for i/o
2956 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2957 * marking the page (and in this case mft record) dirty but we do not implement
2958 * this yet as write_mft_record() largely ignores the @sync parameter and
2959 * always performs synchronous writes.
2961 * Return 0 on success and -errno on error.
2963 int ntfs_write_inode(struct inode *vi, int sync)
2965 sle64 nt;
2966 ntfs_inode *ni = NTFS_I(vi);
2967 ntfs_attr_search_ctx *ctx;
2968 MFT_RECORD *m;
2969 STANDARD_INFORMATION *si;
2970 int err = 0;
2971 bool modified = false;
2973 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2974 vi->i_ino);
2976 * Dirty attribute inodes are written via their real inodes so just
2977 * clean them here. Access time updates are taken care off when the
2978 * real inode is written.
2980 if (NInoAttr(ni)) {
2981 NInoClearDirty(ni);
2982 ntfs_debug("Done.");
2983 return 0;
2985 /* Map, pin, and lock the mft record belonging to the inode. */
2986 m = map_mft_record(ni);
2987 if (IS_ERR(m)) {
2988 err = PTR_ERR(m);
2989 goto err_out;
2991 /* Update the access times in the standard information attribute. */
2992 ctx = ntfs_attr_get_search_ctx(ni, m);
2993 if (unlikely(!ctx)) {
2994 err = -ENOMEM;
2995 goto unm_err_out;
2997 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
2998 CASE_SENSITIVE, 0, NULL, 0, ctx);
2999 if (unlikely(err)) {
3000 ntfs_attr_put_search_ctx(ctx);
3001 goto unm_err_out;
3003 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3004 le16_to_cpu(ctx->attr->data.resident.value_offset));
3005 /* Update the access times if they have changed. */
3006 nt = utc2ntfs(vi->i_mtime);
3007 if (si->last_data_change_time != nt) {
3008 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
3009 "new = 0x%llx", vi->i_ino, (long long)
3010 sle64_to_cpu(si->last_data_change_time),
3011 (long long)sle64_to_cpu(nt));
3012 si->last_data_change_time = nt;
3013 modified = true;
3015 nt = utc2ntfs(vi->i_ctime);
3016 if (si->last_mft_change_time != nt) {
3017 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
3018 "new = 0x%llx", vi->i_ino, (long long)
3019 sle64_to_cpu(si->last_mft_change_time),
3020 (long long)sle64_to_cpu(nt));
3021 si->last_mft_change_time = nt;
3022 modified = true;
3024 nt = utc2ntfs(vi->i_atime);
3025 if (si->last_access_time != nt) {
3026 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3027 "new = 0x%llx", vi->i_ino,
3028 (long long)sle64_to_cpu(si->last_access_time),
3029 (long long)sle64_to_cpu(nt));
3030 si->last_access_time = nt;
3031 modified = true;
3034 * If we just modified the standard information attribute we need to
3035 * mark the mft record it is in dirty. We do this manually so that
3036 * mark_inode_dirty() is not called which would redirty the inode and
3037 * hence result in an infinite loop of trying to write the inode.
3038 * There is no need to mark the base inode nor the base mft record
3039 * dirty, since we are going to write this mft record below in any case
3040 * and the base mft record may actually not have been modified so it
3041 * might not need to be written out.
3042 * NOTE: It is not a problem when the inode for $MFT itself is being
3043 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3044 * on the $MFT inode and hence ntfs_write_inode() will not be
3045 * re-invoked because of it which in turn is ok since the dirtied mft
3046 * record will be cleaned and written out to disk below, i.e. before
3047 * this function returns.
3049 if (modified) {
3050 flush_dcache_mft_record_page(ctx->ntfs_ino);
3051 if (!NInoTestSetDirty(ctx->ntfs_ino))
3052 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3053 ctx->ntfs_ino->page_ofs);
3055 ntfs_attr_put_search_ctx(ctx);
3056 /* Now the access times are updated, write the base mft record. */
3057 if (NInoDirty(ni))
3058 err = write_mft_record(ni, m, sync);
3059 /* Write all attached extent mft records. */
3060 mutex_lock(&ni->extent_lock);
3061 if (ni->nr_extents > 0) {
3062 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3063 int i;
3065 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3066 for (i = 0; i < ni->nr_extents; i++) {
3067 ntfs_inode *tni = extent_nis[i];
3069 if (NInoDirty(tni)) {
3070 MFT_RECORD *tm = map_mft_record(tni);
3071 int ret;
3073 if (IS_ERR(tm)) {
3074 if (!err || err == -ENOMEM)
3075 err = PTR_ERR(tm);
3076 continue;
3078 ret = write_mft_record(tni, tm, sync);
3079 unmap_mft_record(tni);
3080 if (unlikely(ret)) {
3081 if (!err || err == -ENOMEM)
3082 err = ret;
3087 mutex_unlock(&ni->extent_lock);
3088 unmap_mft_record(ni);
3089 if (unlikely(err))
3090 goto err_out;
3091 ntfs_debug("Done.");
3092 return 0;
3093 unm_err_out:
3094 unmap_mft_record(ni);
3095 err_out:
3096 if (err == -ENOMEM) {
3097 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
3098 "Marking the inode dirty again, so the VFS "
3099 "retries later.");
3100 mark_inode_dirty(vi);
3101 } else {
3102 ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err);
3103 NVolSetErrors(ni->vol);
3105 return err;
3108 #endif /* NTFS_RW */