sched: move the __update_rq_clock() call to scheduler_tick()
[usb.git] / fs / ntfs / inode.c
blobb532a730cec2378072a20304fb85d07ed0d529f3
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 "attrib.h"
38 #include "lcnalloc.h"
39 #include "malloc.h"
40 #include "mft.h"
41 #include "time.h"
42 #include "ntfs.h"
44 /**
45 * ntfs_test_inode - compare two (possibly fake) inodes for equality
46 * @vi: vfs inode which to test
47 * @na: ntfs attribute which is being tested with
49 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
50 * inode @vi for equality with the ntfs attribute @na.
52 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
53 * @na->name and @na->name_len are then ignored.
55 * Return 1 if the attributes match and 0 if not.
57 * NOTE: This function runs with the inode_lock spin lock held so it is not
58 * allowed to sleep.
60 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
62 ntfs_inode *ni;
64 if (vi->i_ino != na->mft_no)
65 return 0;
66 ni = NTFS_I(vi);
67 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
68 if (likely(!NInoAttr(ni))) {
69 /* If not looking for a normal inode this is a mismatch. */
70 if (unlikely(na->type != AT_UNUSED))
71 return 0;
72 } else {
73 /* A fake inode describing an attribute. */
74 if (ni->type != na->type)
75 return 0;
76 if (ni->name_len != na->name_len)
77 return 0;
78 if (na->name_len && memcmp(ni->name, na->name,
79 na->name_len * sizeof(ntfschar)))
80 return 0;
82 /* Match! */
83 return 1;
86 /**
87 * ntfs_init_locked_inode - initialize an inode
88 * @vi: vfs inode to initialize
89 * @na: ntfs attribute which to initialize @vi to
91 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
92 * order to enable ntfs_test_inode() to do its work.
94 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
95 * In that case, @na->name and @na->name_len should be set to NULL and 0,
96 * respectively. Although that is not strictly necessary as
97 * ntfs_read_locked_inode() will fill them in later.
99 * Return 0 on success and -errno on error.
101 * NOTE: This function runs with the inode_lock spin lock held so it is not
102 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
104 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
106 ntfs_inode *ni = NTFS_I(vi);
108 vi->i_ino = na->mft_no;
110 ni->type = na->type;
111 if (na->type == AT_INDEX_ALLOCATION)
112 NInoSetMstProtected(ni);
114 ni->name = na->name;
115 ni->name_len = na->name_len;
117 /* If initializing a normal inode, we are done. */
118 if (likely(na->type == AT_UNUSED)) {
119 BUG_ON(na->name);
120 BUG_ON(na->name_len);
121 return 0;
124 /* It is a fake inode. */
125 NInoSetAttr(ni);
128 * We have I30 global constant as an optimization as it is the name
129 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
130 * allocation but that is ok. And most attributes are unnamed anyway,
131 * thus the fraction of named attributes with name != I30 is actually
132 * absolutely tiny.
134 if (na->name_len && na->name != I30) {
135 unsigned int i;
137 BUG_ON(!na->name);
138 i = na->name_len * sizeof(ntfschar);
139 ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
140 if (!ni->name)
141 return -ENOMEM;
142 memcpy(ni->name, na->name, i);
143 ni->name[na->name_len] = 0;
145 return 0;
148 typedef int (*set_t)(struct inode *, void *);
149 static int ntfs_read_locked_inode(struct inode *vi);
150 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
151 static int ntfs_read_locked_index_inode(struct inode *base_vi,
152 struct inode *vi);
155 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
156 * @sb: super block of mounted volume
157 * @mft_no: mft record number / inode number to obtain
159 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
160 * file or directory).
162 * If the inode is in the cache, it is just returned with an increased
163 * reference count. Otherwise, a new struct inode is allocated and initialized,
164 * and finally ntfs_read_locked_inode() is called to read in the inode and
165 * fill in the remainder of the inode structure.
167 * Return the struct inode on success. Check the return value with IS_ERR() and
168 * if true, the function failed and the error code is obtained from PTR_ERR().
170 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
172 struct inode *vi;
173 int err;
174 ntfs_attr na;
176 na.mft_no = mft_no;
177 na.type = AT_UNUSED;
178 na.name = NULL;
179 na.name_len = 0;
181 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
182 (set_t)ntfs_init_locked_inode, &na);
183 if (unlikely(!vi))
184 return ERR_PTR(-ENOMEM);
186 err = 0;
188 /* If this is a freshly allocated inode, need to read it now. */
189 if (vi->i_state & I_NEW) {
190 err = ntfs_read_locked_inode(vi);
191 unlock_new_inode(vi);
194 * There is no point in keeping bad inodes around if the failure was
195 * due to ENOMEM. We want to be able to retry again later.
197 if (unlikely(err == -ENOMEM)) {
198 iput(vi);
199 vi = ERR_PTR(err);
201 return vi;
205 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
206 * @base_vi: vfs base inode containing the attribute
207 * @type: attribute type
208 * @name: Unicode name of the attribute (NULL if unnamed)
209 * @name_len: length of @name in Unicode characters (0 if unnamed)
211 * Obtain the (fake) struct inode corresponding to the attribute specified by
212 * @type, @name, and @name_len, which is present in the base mft record
213 * specified by the vfs inode @base_vi.
215 * If the attribute inode is in the cache, it is just returned with an
216 * increased reference count. Otherwise, a new struct inode is allocated and
217 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
218 * attribute and fill in the inode structure.
220 * Note, for index allocation attributes, you need to use ntfs_index_iget()
221 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
223 * Return the struct inode of the attribute inode on success. Check the return
224 * value with IS_ERR() and if true, the function failed and the error code is
225 * obtained from PTR_ERR().
227 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
228 ntfschar *name, u32 name_len)
230 struct inode *vi;
231 int err;
232 ntfs_attr na;
234 /* Make sure no one calls ntfs_attr_iget() for indices. */
235 BUG_ON(type == AT_INDEX_ALLOCATION);
237 na.mft_no = base_vi->i_ino;
238 na.type = type;
239 na.name = name;
240 na.name_len = name_len;
242 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
243 (set_t)ntfs_init_locked_inode, &na);
244 if (unlikely(!vi))
245 return ERR_PTR(-ENOMEM);
247 err = 0;
249 /* If this is a freshly allocated inode, need to read it now. */
250 if (vi->i_state & I_NEW) {
251 err = ntfs_read_locked_attr_inode(base_vi, vi);
252 unlock_new_inode(vi);
255 * There is no point in keeping bad attribute inodes around. This also
256 * simplifies things in that we never need to check for bad attribute
257 * inodes elsewhere.
259 if (unlikely(err)) {
260 iput(vi);
261 vi = ERR_PTR(err);
263 return vi;
267 * ntfs_index_iget - obtain a struct inode corresponding to an index
268 * @base_vi: vfs base inode containing the index related attributes
269 * @name: Unicode name of the index
270 * @name_len: length of @name in Unicode characters
272 * Obtain the (fake) struct inode corresponding to the index specified by @name
273 * and @name_len, which is present in the base mft record specified by the vfs
274 * inode @base_vi.
276 * If the index inode is in the cache, it is just returned with an increased
277 * reference count. Otherwise, a new struct inode is allocated and
278 * initialized, and finally ntfs_read_locked_index_inode() is called to read
279 * the index related attributes and fill in the inode structure.
281 * Return the struct inode of the index inode on success. Check the return
282 * value with IS_ERR() and if true, the function failed and the error code is
283 * obtained from PTR_ERR().
285 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
286 u32 name_len)
288 struct inode *vi;
289 int err;
290 ntfs_attr na;
292 na.mft_no = base_vi->i_ino;
293 na.type = AT_INDEX_ALLOCATION;
294 na.name = name;
295 na.name_len = name_len;
297 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
298 (set_t)ntfs_init_locked_inode, &na);
299 if (unlikely(!vi))
300 return ERR_PTR(-ENOMEM);
302 err = 0;
304 /* If this is a freshly allocated inode, need to read it now. */
305 if (vi->i_state & I_NEW) {
306 err = ntfs_read_locked_index_inode(base_vi, vi);
307 unlock_new_inode(vi);
310 * There is no point in keeping bad index inodes around. This also
311 * simplifies things in that we never need to check for bad index
312 * inodes elsewhere.
314 if (unlikely(err)) {
315 iput(vi);
316 vi = ERR_PTR(err);
318 return vi;
321 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
323 ntfs_inode *ni;
325 ntfs_debug("Entering.");
326 ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_NOFS);
327 if (likely(ni != NULL)) {
328 ni->state = 0;
329 return VFS_I(ni);
331 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
332 return NULL;
335 void ntfs_destroy_big_inode(struct inode *inode)
337 ntfs_inode *ni = NTFS_I(inode);
339 ntfs_debug("Entering.");
340 BUG_ON(ni->page);
341 if (!atomic_dec_and_test(&ni->count))
342 BUG();
343 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
346 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
348 ntfs_inode *ni;
350 ntfs_debug("Entering.");
351 ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS);
352 if (likely(ni != NULL)) {
353 ni->state = 0;
354 return ni;
356 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
357 return NULL;
360 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
362 ntfs_debug("Entering.");
363 BUG_ON(ni->page);
364 if (!atomic_dec_and_test(&ni->count))
365 BUG();
366 kmem_cache_free(ntfs_inode_cache, ni);
370 * The attribute runlist lock has separate locking rules from the
371 * normal runlist lock, so split the two lock-classes:
373 static struct lock_class_key attr_list_rl_lock_class;
376 * __ntfs_init_inode - initialize ntfs specific part of an inode
377 * @sb: super block of mounted volume
378 * @ni: freshly allocated ntfs inode which to initialize
380 * Initialize an ntfs inode to defaults.
382 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
383 * untouched. Make sure to initialize them elsewhere.
385 * Return zero on success and -ENOMEM on error.
387 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
389 ntfs_debug("Entering.");
390 rwlock_init(&ni->size_lock);
391 ni->initialized_size = ni->allocated_size = 0;
392 ni->seq_no = 0;
393 atomic_set(&ni->count, 1);
394 ni->vol = NTFS_SB(sb);
395 ntfs_init_runlist(&ni->runlist);
396 mutex_init(&ni->mrec_lock);
397 ni->page = NULL;
398 ni->page_ofs = 0;
399 ni->attr_list_size = 0;
400 ni->attr_list = NULL;
401 ntfs_init_runlist(&ni->attr_list_rl);
402 lockdep_set_class(&ni->attr_list_rl.lock,
403 &attr_list_rl_lock_class);
404 ni->itype.index.block_size = 0;
405 ni->itype.index.vcn_size = 0;
406 ni->itype.index.collation_rule = 0;
407 ni->itype.index.block_size_bits = 0;
408 ni->itype.index.vcn_size_bits = 0;
409 mutex_init(&ni->extent_lock);
410 ni->nr_extents = 0;
411 ni->ext.base_ntfs_ino = NULL;
415 * Extent inodes get MFT-mapped in a nested way, while the base inode
416 * is still mapped. Teach this nesting to the lock validator by creating
417 * a separate class for nested inode's mrec_lock's:
419 static struct lock_class_key extent_inode_mrec_lock_key;
421 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
422 unsigned long mft_no)
424 ntfs_inode *ni = ntfs_alloc_extent_inode();
426 ntfs_debug("Entering.");
427 if (likely(ni != NULL)) {
428 __ntfs_init_inode(sb, ni);
429 lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
430 ni->mft_no = mft_no;
431 ni->type = AT_UNUSED;
432 ni->name = NULL;
433 ni->name_len = 0;
435 return ni;
439 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
440 * @ctx: initialized attribute search context
442 * Search all file name attributes in the inode described by the attribute
443 * search context @ctx and check if any of the names are in the $Extend system
444 * directory.
446 * Return values:
447 * 1: file is in $Extend directory
448 * 0: file is not in $Extend directory
449 * -errno: failed to determine if the file is in the $Extend directory
451 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
453 int nr_links, err;
455 /* Restart search. */
456 ntfs_attr_reinit_search_ctx(ctx);
458 /* Get number of hard links. */
459 nr_links = le16_to_cpu(ctx->mrec->link_count);
461 /* Loop through all hard links. */
462 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
463 ctx))) {
464 FILE_NAME_ATTR *file_name_attr;
465 ATTR_RECORD *attr = ctx->attr;
466 u8 *p, *p2;
468 nr_links--;
470 * Maximum sanity checking as we are called on an inode that
471 * we suspect might be corrupt.
473 p = (u8*)attr + le32_to_cpu(attr->length);
474 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
475 le32_to_cpu(ctx->mrec->bytes_in_use)) {
476 err_corrupt_attr:
477 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
478 "attribute. You should run chkdsk.");
479 return -EIO;
481 if (attr->non_resident) {
482 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
483 "name. You should run chkdsk.");
484 return -EIO;
486 if (attr->flags) {
487 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
488 "invalid flags. You should run "
489 "chkdsk.");
490 return -EIO;
492 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
493 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
494 "name. You should run chkdsk.");
495 return -EIO;
497 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
498 le16_to_cpu(attr->data.resident.value_offset));
499 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
500 if (p2 < (u8*)attr || p2 > p)
501 goto err_corrupt_attr;
502 /* This attribute is ok, but is it in the $Extend directory? */
503 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
504 return 1; /* YES, it's an extended system file. */
506 if (unlikely(err != -ENOENT))
507 return err;
508 if (unlikely(nr_links)) {
509 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
510 "doesn't match number of name attributes. You "
511 "should run chkdsk.");
512 return -EIO;
514 return 0; /* NO, it is not an extended system file. */
518 * ntfs_read_locked_inode - read an inode from its device
519 * @vi: inode to read
521 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
522 * described by @vi into memory from the device.
524 * The only fields in @vi that we need to/can look at when the function is
525 * called are i_sb, pointing to the mounted device's super block, and i_ino,
526 * the number of the inode to load.
528 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
529 * for reading and sets up the necessary @vi fields as well as initializing
530 * the ntfs inode.
532 * Q: What locks are held when the function is called?
533 * A: i_state has I_LOCK set, hence the inode is locked, also
534 * i_count is set to 1, so it is not going to go away
535 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
536 * is allowed to write to them. We should of course be honouring them but
537 * we need to do that using the IS_* macros defined in include/linux/fs.h.
538 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
540 * Return 0 on success and -errno on error. In the error case, the inode will
541 * have had make_bad_inode() executed on it.
543 static int ntfs_read_locked_inode(struct inode *vi)
545 ntfs_volume *vol = NTFS_SB(vi->i_sb);
546 ntfs_inode *ni;
547 struct inode *bvi;
548 MFT_RECORD *m;
549 ATTR_RECORD *a;
550 STANDARD_INFORMATION *si;
551 ntfs_attr_search_ctx *ctx;
552 int err = 0;
554 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
556 /* Setup the generic vfs inode parts now. */
559 * This is for checking whether an inode has changed w.r.t. a file so
560 * that the file can be updated if necessary (compare with f_version).
562 vi->i_version = 1;
564 vi->i_uid = vol->uid;
565 vi->i_gid = vol->gid;
566 vi->i_mode = 0;
569 * Initialize the ntfs specific part of @vi special casing
570 * FILE_MFT which we need to do at mount time.
572 if (vi->i_ino != FILE_MFT)
573 ntfs_init_big_inode(vi);
574 ni = NTFS_I(vi);
576 m = map_mft_record(ni);
577 if (IS_ERR(m)) {
578 err = PTR_ERR(m);
579 goto err_out;
581 ctx = ntfs_attr_get_search_ctx(ni, m);
582 if (!ctx) {
583 err = -ENOMEM;
584 goto unm_err_out;
587 if (!(m->flags & MFT_RECORD_IN_USE)) {
588 ntfs_error(vi->i_sb, "Inode is not in use!");
589 goto unm_err_out;
591 if (m->base_mft_record) {
592 ntfs_error(vi->i_sb, "Inode is an extent inode!");
593 goto unm_err_out;
596 /* Transfer information from mft record into vfs and ntfs inodes. */
597 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
600 * FIXME: Keep in mind that link_count is two for files which have both
601 * a long file name and a short file name as separate entries, so if
602 * we are hiding short file names this will be too high. Either we need
603 * to account for the short file names by subtracting them or we need
604 * to make sure we delete files even though i_nlink is not zero which
605 * might be tricky due to vfs interactions. Need to think about this
606 * some more when implementing the unlink command.
608 vi->i_nlink = le16_to_cpu(m->link_count);
610 * FIXME: Reparse points can have the directory bit set even though
611 * they would be S_IFLNK. Need to deal with this further below when we
612 * implement reparse points / symbolic links but it will do for now.
613 * Also if not a directory, it could be something else, rather than
614 * a regular file. But again, will do for now.
616 /* Everyone gets all permissions. */
617 vi->i_mode |= S_IRWXUGO;
618 /* If read-only, noone gets write permissions. */
619 if (IS_RDONLY(vi))
620 vi->i_mode &= ~S_IWUGO;
621 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
622 vi->i_mode |= S_IFDIR;
624 * Apply the directory permissions mask set in the mount
625 * options.
627 vi->i_mode &= ~vol->dmask;
628 /* Things break without this kludge! */
629 if (vi->i_nlink > 1)
630 vi->i_nlink = 1;
631 } else {
632 vi->i_mode |= S_IFREG;
633 /* Apply the file permissions mask set in the mount options. */
634 vi->i_mode &= ~vol->fmask;
637 * Find the standard information attribute in the mft record. At this
638 * stage we haven't setup the attribute list stuff yet, so this could
639 * in fact fail if the standard information is in an extent record, but
640 * I don't think this actually ever happens.
642 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
643 ctx);
644 if (unlikely(err)) {
645 if (err == -ENOENT) {
647 * TODO: We should be performing a hot fix here (if the
648 * recover mount option is set) by creating a new
649 * attribute.
651 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
652 "is missing.");
654 goto unm_err_out;
656 a = ctx->attr;
657 /* Get the standard information attribute value. */
658 si = (STANDARD_INFORMATION*)((u8*)a +
659 le16_to_cpu(a->data.resident.value_offset));
661 /* Transfer information from the standard information into vi. */
663 * Note: The i_?times do not quite map perfectly onto the NTFS times,
664 * but they are close enough, and in the end it doesn't really matter
665 * that much...
668 * mtime is the last change of the data within the file. Not changed
669 * when only metadata is changed, e.g. a rename doesn't affect mtime.
671 vi->i_mtime = ntfs2utc(si->last_data_change_time);
673 * ctime is the last change of the metadata of the file. This obviously
674 * always changes, when mtime is changed. ctime can be changed on its
675 * own, mtime is then not changed, e.g. when a file is renamed.
677 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
679 * Last access to the data within the file. Not changed during a rename
680 * for example but changed whenever the file is written to.
682 vi->i_atime = ntfs2utc(si->last_access_time);
684 /* Find the attribute list attribute if present. */
685 ntfs_attr_reinit_search_ctx(ctx);
686 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
687 if (err) {
688 if (unlikely(err != -ENOENT)) {
689 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
690 "attribute.");
691 goto unm_err_out;
693 } else /* if (!err) */ {
694 if (vi->i_ino == FILE_MFT)
695 goto skip_attr_list_load;
696 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
697 NInoSetAttrList(ni);
698 a = ctx->attr;
699 if (a->flags & ATTR_COMPRESSION_MASK) {
700 ntfs_error(vi->i_sb, "Attribute list attribute is "
701 "compressed.");
702 goto unm_err_out;
704 if (a->flags & ATTR_IS_ENCRYPTED ||
705 a->flags & ATTR_IS_SPARSE) {
706 if (a->non_resident) {
707 ntfs_error(vi->i_sb, "Non-resident attribute "
708 "list attribute is encrypted/"
709 "sparse.");
710 goto unm_err_out;
712 ntfs_warning(vi->i_sb, "Resident attribute list "
713 "attribute in inode 0x%lx is marked "
714 "encrypted/sparse which is not true. "
715 "However, Windows allows this and "
716 "chkdsk does not detect or correct it "
717 "so we will just ignore the invalid "
718 "flags and pretend they are not set.",
719 vi->i_ino);
721 /* Now allocate memory for the attribute list. */
722 ni->attr_list_size = (u32)ntfs_attr_size(a);
723 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
724 if (!ni->attr_list) {
725 ntfs_error(vi->i_sb, "Not enough memory to allocate "
726 "buffer for attribute list.");
727 err = -ENOMEM;
728 goto unm_err_out;
730 if (a->non_resident) {
731 NInoSetAttrListNonResident(ni);
732 if (a->data.non_resident.lowest_vcn) {
733 ntfs_error(vi->i_sb, "Attribute list has non "
734 "zero lowest_vcn.");
735 goto unm_err_out;
738 * Setup the runlist. No need for locking as we have
739 * exclusive access to the inode at this time.
741 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
742 a, NULL);
743 if (IS_ERR(ni->attr_list_rl.rl)) {
744 err = PTR_ERR(ni->attr_list_rl.rl);
745 ni->attr_list_rl.rl = NULL;
746 ntfs_error(vi->i_sb, "Mapping pairs "
747 "decompression failed.");
748 goto unm_err_out;
750 /* Now load the attribute list. */
751 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
752 ni->attr_list, ni->attr_list_size,
753 sle64_to_cpu(a->data.non_resident.
754 initialized_size)))) {
755 ntfs_error(vi->i_sb, "Failed to load "
756 "attribute list attribute.");
757 goto unm_err_out;
759 } else /* if (!a->non_resident) */ {
760 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
761 + le32_to_cpu(
762 a->data.resident.value_length) >
763 (u8*)ctx->mrec + vol->mft_record_size) {
764 ntfs_error(vi->i_sb, "Corrupt attribute list "
765 "in inode.");
766 goto unm_err_out;
768 /* Now copy the attribute list. */
769 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
770 a->data.resident.value_offset),
771 le32_to_cpu(
772 a->data.resident.value_length));
775 skip_attr_list_load:
777 * If an attribute list is present we now have the attribute list value
778 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
780 if (S_ISDIR(vi->i_mode)) {
781 loff_t bvi_size;
782 ntfs_inode *bni;
783 INDEX_ROOT *ir;
784 u8 *ir_end, *index_end;
786 /* It is a directory, find index root attribute. */
787 ntfs_attr_reinit_search_ctx(ctx);
788 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
789 0, NULL, 0, ctx);
790 if (unlikely(err)) {
791 if (err == -ENOENT) {
792 // FIXME: File is corrupt! Hot-fix with empty
793 // index root attribute if recovery option is
794 // set.
795 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
796 "is missing.");
798 goto unm_err_out;
800 a = ctx->attr;
801 /* Set up the state. */
802 if (unlikely(a->non_resident)) {
803 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
804 "resident.");
805 goto unm_err_out;
807 /* Ensure the attribute name is placed before the value. */
808 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
809 le16_to_cpu(a->data.resident.value_offset)))) {
810 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
811 "placed after the attribute value.");
812 goto unm_err_out;
815 * Compressed/encrypted index root just means that the newly
816 * created files in that directory should be created compressed/
817 * encrypted. However index root cannot be both compressed and
818 * encrypted.
820 if (a->flags & ATTR_COMPRESSION_MASK)
821 NInoSetCompressed(ni);
822 if (a->flags & ATTR_IS_ENCRYPTED) {
823 if (a->flags & ATTR_COMPRESSION_MASK) {
824 ntfs_error(vi->i_sb, "Found encrypted and "
825 "compressed attribute.");
826 goto unm_err_out;
828 NInoSetEncrypted(ni);
830 if (a->flags & ATTR_IS_SPARSE)
831 NInoSetSparse(ni);
832 ir = (INDEX_ROOT*)((u8*)a +
833 le16_to_cpu(a->data.resident.value_offset));
834 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
835 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
836 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
837 "corrupt.");
838 goto unm_err_out;
840 index_end = (u8*)&ir->index +
841 le32_to_cpu(ir->index.index_length);
842 if (index_end > ir_end) {
843 ntfs_error(vi->i_sb, "Directory index is corrupt.");
844 goto unm_err_out;
846 if (ir->type != AT_FILE_NAME) {
847 ntfs_error(vi->i_sb, "Indexed attribute is not "
848 "$FILE_NAME.");
849 goto unm_err_out;
851 if (ir->collation_rule != COLLATION_FILE_NAME) {
852 ntfs_error(vi->i_sb, "Index collation rule is not "
853 "COLLATION_FILE_NAME.");
854 goto unm_err_out;
856 ni->itype.index.collation_rule = ir->collation_rule;
857 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
858 if (ni->itype.index.block_size &
859 (ni->itype.index.block_size - 1)) {
860 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
861 "power of two.",
862 ni->itype.index.block_size);
863 goto unm_err_out;
865 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
866 ntfs_error(vi->i_sb, "Index block size (%u) > "
867 "PAGE_CACHE_SIZE (%ld) is not "
868 "supported. Sorry.",
869 ni->itype.index.block_size,
870 PAGE_CACHE_SIZE);
871 err = -EOPNOTSUPP;
872 goto unm_err_out;
874 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
875 ntfs_error(vi->i_sb, "Index block size (%u) < "
876 "NTFS_BLOCK_SIZE (%i) is not "
877 "supported. Sorry.",
878 ni->itype.index.block_size,
879 NTFS_BLOCK_SIZE);
880 err = -EOPNOTSUPP;
881 goto unm_err_out;
883 ni->itype.index.block_size_bits =
884 ffs(ni->itype.index.block_size) - 1;
885 /* Determine the size of a vcn in the directory index. */
886 if (vol->cluster_size <= ni->itype.index.block_size) {
887 ni->itype.index.vcn_size = vol->cluster_size;
888 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
889 } else {
890 ni->itype.index.vcn_size = vol->sector_size;
891 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
894 /* Setup the index allocation attribute, even if not present. */
895 NInoSetMstProtected(ni);
896 ni->type = AT_INDEX_ALLOCATION;
897 ni->name = I30;
898 ni->name_len = 4;
900 if (!(ir->index.flags & LARGE_INDEX)) {
901 /* No index allocation. */
902 vi->i_size = ni->initialized_size =
903 ni->allocated_size = 0;
904 /* We are done with the mft record, so we release it. */
905 ntfs_attr_put_search_ctx(ctx);
906 unmap_mft_record(ni);
907 m = NULL;
908 ctx = NULL;
909 goto skip_large_dir_stuff;
910 } /* LARGE_INDEX: Index allocation present. Setup state. */
911 NInoSetIndexAllocPresent(ni);
912 /* Find index allocation attribute. */
913 ntfs_attr_reinit_search_ctx(ctx);
914 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
915 CASE_SENSITIVE, 0, NULL, 0, ctx);
916 if (unlikely(err)) {
917 if (err == -ENOENT)
918 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
919 "attribute is not present but "
920 "$INDEX_ROOT indicated it is.");
921 else
922 ntfs_error(vi->i_sb, "Failed to lookup "
923 "$INDEX_ALLOCATION "
924 "attribute.");
925 goto unm_err_out;
927 a = ctx->attr;
928 if (!a->non_resident) {
929 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
930 "is resident.");
931 goto unm_err_out;
934 * Ensure the attribute name is placed before the mapping pairs
935 * array.
937 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
938 le16_to_cpu(
939 a->data.non_resident.mapping_pairs_offset)))) {
940 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
941 "is placed after the mapping pairs "
942 "array.");
943 goto unm_err_out;
945 if (a->flags & ATTR_IS_ENCRYPTED) {
946 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
947 "is encrypted.");
948 goto unm_err_out;
950 if (a->flags & ATTR_IS_SPARSE) {
951 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
952 "is sparse.");
953 goto unm_err_out;
955 if (a->flags & ATTR_COMPRESSION_MASK) {
956 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
957 "is compressed.");
958 goto unm_err_out;
960 if (a->data.non_resident.lowest_vcn) {
961 ntfs_error(vi->i_sb, "First extent of "
962 "$INDEX_ALLOCATION attribute has non "
963 "zero lowest_vcn.");
964 goto unm_err_out;
966 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
967 ni->initialized_size = sle64_to_cpu(
968 a->data.non_resident.initialized_size);
969 ni->allocated_size = sle64_to_cpu(
970 a->data.non_resident.allocated_size);
972 * We are done with the mft record, so we release it. Otherwise
973 * we would deadlock in ntfs_attr_iget().
975 ntfs_attr_put_search_ctx(ctx);
976 unmap_mft_record(ni);
977 m = NULL;
978 ctx = NULL;
979 /* Get the index bitmap attribute inode. */
980 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
981 if (IS_ERR(bvi)) {
982 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
983 err = PTR_ERR(bvi);
984 goto unm_err_out;
986 bni = NTFS_I(bvi);
987 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
988 NInoSparse(bni)) {
989 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
990 "and/or encrypted and/or sparse.");
991 goto iput_unm_err_out;
993 /* Consistency check bitmap size vs. index allocation size. */
994 bvi_size = i_size_read(bvi);
995 if ((bvi_size << 3) < (vi->i_size >>
996 ni->itype.index.block_size_bits)) {
997 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
998 "for index allocation (0x%llx).",
999 bvi_size << 3, vi->i_size);
1000 goto iput_unm_err_out;
1002 /* No longer need the bitmap attribute inode. */
1003 iput(bvi);
1004 skip_large_dir_stuff:
1005 /* Setup the operations for this inode. */
1006 vi->i_op = &ntfs_dir_inode_ops;
1007 vi->i_fop = &ntfs_dir_ops;
1008 } else {
1009 /* It is a file. */
1010 ntfs_attr_reinit_search_ctx(ctx);
1012 /* Setup the data attribute, even if not present. */
1013 ni->type = AT_DATA;
1014 ni->name = NULL;
1015 ni->name_len = 0;
1017 /* Find first extent of the unnamed data attribute. */
1018 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1019 if (unlikely(err)) {
1020 vi->i_size = ni->initialized_size =
1021 ni->allocated_size = 0;
1022 if (err != -ENOENT) {
1023 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1024 "attribute.");
1025 goto unm_err_out;
1028 * FILE_Secure does not have an unnamed $DATA
1029 * attribute, so we special case it here.
1031 if (vi->i_ino == FILE_Secure)
1032 goto no_data_attr_special_case;
1034 * Most if not all the system files in the $Extend
1035 * system directory do not have unnamed data
1036 * attributes so we need to check if the parent
1037 * directory of the file is FILE_Extend and if it is
1038 * ignore this error. To do this we need to get the
1039 * name of this inode from the mft record as the name
1040 * contains the back reference to the parent directory.
1042 if (ntfs_is_extended_system_file(ctx) > 0)
1043 goto no_data_attr_special_case;
1044 // FIXME: File is corrupt! Hot-fix with empty data
1045 // attribute if recovery option is set.
1046 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1047 goto unm_err_out;
1049 a = ctx->attr;
1050 /* Setup the state. */
1051 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1052 if (a->flags & ATTR_COMPRESSION_MASK) {
1053 NInoSetCompressed(ni);
1054 if (vol->cluster_size > 4096) {
1055 ntfs_error(vi->i_sb, "Found "
1056 "compressed data but "
1057 "compression is "
1058 "disabled due to "
1059 "cluster size (%i) > "
1060 "4kiB.",
1061 vol->cluster_size);
1062 goto unm_err_out;
1064 if ((a->flags & ATTR_COMPRESSION_MASK)
1065 != ATTR_IS_COMPRESSED) {
1066 ntfs_error(vi->i_sb, "Found unknown "
1067 "compression method "
1068 "or corrupt file.");
1069 goto unm_err_out;
1072 if (a->flags & ATTR_IS_SPARSE)
1073 NInoSetSparse(ni);
1075 if (a->flags & ATTR_IS_ENCRYPTED) {
1076 if (NInoCompressed(ni)) {
1077 ntfs_error(vi->i_sb, "Found encrypted and "
1078 "compressed data.");
1079 goto unm_err_out;
1081 NInoSetEncrypted(ni);
1083 if (a->non_resident) {
1084 NInoSetNonResident(ni);
1085 if (NInoCompressed(ni) || NInoSparse(ni)) {
1086 if (NInoCompressed(ni) && a->data.non_resident.
1087 compression_unit != 4) {
1088 ntfs_error(vi->i_sb, "Found "
1089 "non-standard "
1090 "compression unit (%u "
1091 "instead of 4). "
1092 "Cannot handle this.",
1093 a->data.non_resident.
1094 compression_unit);
1095 err = -EOPNOTSUPP;
1096 goto unm_err_out;
1098 if (a->data.non_resident.compression_unit) {
1099 ni->itype.compressed.block_size = 1U <<
1100 (a->data.non_resident.
1101 compression_unit +
1102 vol->cluster_size_bits);
1103 ni->itype.compressed.block_size_bits =
1104 ffs(ni->itype.
1105 compressed.
1106 block_size) - 1;
1107 ni->itype.compressed.block_clusters =
1108 1U << a->data.
1109 non_resident.
1110 compression_unit;
1111 } else {
1112 ni->itype.compressed.block_size = 0;
1113 ni->itype.compressed.block_size_bits =
1115 ni->itype.compressed.block_clusters =
1118 ni->itype.compressed.size = sle64_to_cpu(
1119 a->data.non_resident.
1120 compressed_size);
1122 if (a->data.non_resident.lowest_vcn) {
1123 ntfs_error(vi->i_sb, "First extent of $DATA "
1124 "attribute has non zero "
1125 "lowest_vcn.");
1126 goto unm_err_out;
1128 vi->i_size = sle64_to_cpu(
1129 a->data.non_resident.data_size);
1130 ni->initialized_size = sle64_to_cpu(
1131 a->data.non_resident.initialized_size);
1132 ni->allocated_size = sle64_to_cpu(
1133 a->data.non_resident.allocated_size);
1134 } else { /* Resident attribute. */
1135 vi->i_size = ni->initialized_size = le32_to_cpu(
1136 a->data.resident.value_length);
1137 ni->allocated_size = le32_to_cpu(a->length) -
1138 le16_to_cpu(
1139 a->data.resident.value_offset);
1140 if (vi->i_size > ni->allocated_size) {
1141 ntfs_error(vi->i_sb, "Resident data attribute "
1142 "is corrupt (size exceeds "
1143 "allocation).");
1144 goto unm_err_out;
1147 no_data_attr_special_case:
1148 /* We are done with the mft record, so we release it. */
1149 ntfs_attr_put_search_ctx(ctx);
1150 unmap_mft_record(ni);
1151 m = NULL;
1152 ctx = NULL;
1153 /* Setup the operations for this inode. */
1154 vi->i_op = &ntfs_file_inode_ops;
1155 vi->i_fop = &ntfs_file_ops;
1157 if (NInoMstProtected(ni))
1158 vi->i_mapping->a_ops = &ntfs_mst_aops;
1159 else
1160 vi->i_mapping->a_ops = &ntfs_aops;
1162 * The number of 512-byte blocks used on disk (for stat). This is in so
1163 * far inaccurate as it doesn't account for any named streams or other
1164 * special non-resident attributes, but that is how Windows works, too,
1165 * so we are at least consistent with Windows, if not entirely
1166 * consistent with the Linux Way. Doing it the Linux Way would cause a
1167 * significant slowdown as it would involve iterating over all
1168 * attributes in the mft record and adding the allocated/compressed
1169 * sizes of all non-resident attributes present to give us the Linux
1170 * correct size that should go into i_blocks (after division by 512).
1172 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1173 vi->i_blocks = ni->itype.compressed.size >> 9;
1174 else
1175 vi->i_blocks = ni->allocated_size >> 9;
1176 ntfs_debug("Done.");
1177 return 0;
1178 iput_unm_err_out:
1179 iput(bvi);
1180 unm_err_out:
1181 if (!err)
1182 err = -EIO;
1183 if (ctx)
1184 ntfs_attr_put_search_ctx(ctx);
1185 if (m)
1186 unmap_mft_record(ni);
1187 err_out:
1188 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1189 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1190 make_bad_inode(vi);
1191 if (err != -EOPNOTSUPP && err != -ENOMEM)
1192 NVolSetErrors(vol);
1193 return err;
1197 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1198 * @base_vi: base inode
1199 * @vi: attribute inode to read
1201 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1202 * attribute inode described by @vi into memory from the base mft record
1203 * described by @base_ni.
1205 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1206 * reading and looks up the attribute described by @vi before setting up the
1207 * necessary fields in @vi as well as initializing the ntfs inode.
1209 * Q: What locks are held when the function is called?
1210 * A: i_state has I_LOCK set, hence the inode is locked, also
1211 * i_count is set to 1, so it is not going to go away
1213 * Return 0 on success and -errno on error. In the error case, the inode will
1214 * have had make_bad_inode() executed on it.
1216 * Note this cannot be called for AT_INDEX_ALLOCATION.
1218 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1220 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1221 ntfs_inode *ni, *base_ni;
1222 MFT_RECORD *m;
1223 ATTR_RECORD *a;
1224 ntfs_attr_search_ctx *ctx;
1225 int err = 0;
1227 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1229 ntfs_init_big_inode(vi);
1231 ni = NTFS_I(vi);
1232 base_ni = NTFS_I(base_vi);
1234 /* Just mirror the values from the base inode. */
1235 vi->i_version = base_vi->i_version;
1236 vi->i_uid = base_vi->i_uid;
1237 vi->i_gid = base_vi->i_gid;
1238 vi->i_nlink = base_vi->i_nlink;
1239 vi->i_mtime = base_vi->i_mtime;
1240 vi->i_ctime = base_vi->i_ctime;
1241 vi->i_atime = base_vi->i_atime;
1242 vi->i_generation = ni->seq_no = base_ni->seq_no;
1244 /* Set inode type to zero but preserve permissions. */
1245 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1247 m = map_mft_record(base_ni);
1248 if (IS_ERR(m)) {
1249 err = PTR_ERR(m);
1250 goto err_out;
1252 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1253 if (!ctx) {
1254 err = -ENOMEM;
1255 goto unm_err_out;
1257 /* Find the attribute. */
1258 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1259 CASE_SENSITIVE, 0, NULL, 0, ctx);
1260 if (unlikely(err))
1261 goto unm_err_out;
1262 a = ctx->attr;
1263 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1264 if (a->flags & ATTR_COMPRESSION_MASK) {
1265 NInoSetCompressed(ni);
1266 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1267 ni->name_len)) {
1268 ntfs_error(vi->i_sb, "Found compressed "
1269 "non-data or named data "
1270 "attribute. Please report "
1271 "you saw this message to "
1272 "linux-ntfs-dev@lists."
1273 "sourceforge.net");
1274 goto unm_err_out;
1276 if (vol->cluster_size > 4096) {
1277 ntfs_error(vi->i_sb, "Found compressed "
1278 "attribute but compression is "
1279 "disabled due to cluster size "
1280 "(%i) > 4kiB.",
1281 vol->cluster_size);
1282 goto unm_err_out;
1284 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1285 ATTR_IS_COMPRESSED) {
1286 ntfs_error(vi->i_sb, "Found unknown "
1287 "compression method.");
1288 goto unm_err_out;
1292 * The compressed/sparse flag set in an index root just means
1293 * to compress all files.
1295 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1296 ntfs_error(vi->i_sb, "Found mst protected attribute "
1297 "but the attribute is %s. Please "
1298 "report you saw this message to "
1299 "linux-ntfs-dev@lists.sourceforge.net",
1300 NInoCompressed(ni) ? "compressed" :
1301 "sparse");
1302 goto unm_err_out;
1304 if (a->flags & ATTR_IS_SPARSE)
1305 NInoSetSparse(ni);
1307 if (a->flags & ATTR_IS_ENCRYPTED) {
1308 if (NInoCompressed(ni)) {
1309 ntfs_error(vi->i_sb, "Found encrypted and compressed "
1310 "data.");
1311 goto unm_err_out;
1314 * The encryption flag set in an index root just means to
1315 * encrypt all files.
1317 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1318 ntfs_error(vi->i_sb, "Found mst protected attribute "
1319 "but the attribute is encrypted. "
1320 "Please report you saw this message "
1321 "to linux-ntfs-dev@lists.sourceforge."
1322 "net");
1323 goto unm_err_out;
1325 if (ni->type != AT_DATA) {
1326 ntfs_error(vi->i_sb, "Found encrypted non-data "
1327 "attribute.");
1328 goto unm_err_out;
1330 NInoSetEncrypted(ni);
1332 if (!a->non_resident) {
1333 /* Ensure the attribute name is placed before the value. */
1334 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1335 le16_to_cpu(a->data.resident.value_offset)))) {
1336 ntfs_error(vol->sb, "Attribute name is placed after "
1337 "the attribute value.");
1338 goto unm_err_out;
1340 if (NInoMstProtected(ni)) {
1341 ntfs_error(vi->i_sb, "Found mst protected attribute "
1342 "but the attribute is resident. "
1343 "Please report you saw this message to "
1344 "linux-ntfs-dev@lists.sourceforge.net");
1345 goto unm_err_out;
1347 vi->i_size = ni->initialized_size = le32_to_cpu(
1348 a->data.resident.value_length);
1349 ni->allocated_size = le32_to_cpu(a->length) -
1350 le16_to_cpu(a->data.resident.value_offset);
1351 if (vi->i_size > ni->allocated_size) {
1352 ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1353 "(size exceeds allocation).");
1354 goto unm_err_out;
1356 } else {
1357 NInoSetNonResident(ni);
1359 * Ensure the attribute name is placed before the mapping pairs
1360 * array.
1362 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1363 le16_to_cpu(
1364 a->data.non_resident.mapping_pairs_offset)))) {
1365 ntfs_error(vol->sb, "Attribute name is placed after "
1366 "the mapping pairs array.");
1367 goto unm_err_out;
1369 if (NInoCompressed(ni) || NInoSparse(ni)) {
1370 if (NInoCompressed(ni) && a->data.non_resident.
1371 compression_unit != 4) {
1372 ntfs_error(vi->i_sb, "Found non-standard "
1373 "compression unit (%u instead "
1374 "of 4). Cannot handle this.",
1375 a->data.non_resident.
1376 compression_unit);
1377 err = -EOPNOTSUPP;
1378 goto unm_err_out;
1380 if (a->data.non_resident.compression_unit) {
1381 ni->itype.compressed.block_size = 1U <<
1382 (a->data.non_resident.
1383 compression_unit +
1384 vol->cluster_size_bits);
1385 ni->itype.compressed.block_size_bits =
1386 ffs(ni->itype.compressed.
1387 block_size) - 1;
1388 ni->itype.compressed.block_clusters = 1U <<
1389 a->data.non_resident.
1390 compression_unit;
1391 } else {
1392 ni->itype.compressed.block_size = 0;
1393 ni->itype.compressed.block_size_bits = 0;
1394 ni->itype.compressed.block_clusters = 0;
1396 ni->itype.compressed.size = sle64_to_cpu(
1397 a->data.non_resident.compressed_size);
1399 if (a->data.non_resident.lowest_vcn) {
1400 ntfs_error(vi->i_sb, "First extent of attribute has "
1401 "non-zero lowest_vcn.");
1402 goto unm_err_out;
1404 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1405 ni->initialized_size = sle64_to_cpu(
1406 a->data.non_resident.initialized_size);
1407 ni->allocated_size = sle64_to_cpu(
1408 a->data.non_resident.allocated_size);
1410 /* Setup the operations for this attribute inode. */
1411 vi->i_op = NULL;
1412 vi->i_fop = NULL;
1413 if (NInoMstProtected(ni))
1414 vi->i_mapping->a_ops = &ntfs_mst_aops;
1415 else
1416 vi->i_mapping->a_ops = &ntfs_aops;
1417 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1418 vi->i_blocks = ni->itype.compressed.size >> 9;
1419 else
1420 vi->i_blocks = ni->allocated_size >> 9;
1422 * Make sure the base inode does not go away and attach it to the
1423 * attribute inode.
1425 igrab(base_vi);
1426 ni->ext.base_ntfs_ino = base_ni;
1427 ni->nr_extents = -1;
1429 ntfs_attr_put_search_ctx(ctx);
1430 unmap_mft_record(base_ni);
1432 ntfs_debug("Done.");
1433 return 0;
1435 unm_err_out:
1436 if (!err)
1437 err = -EIO;
1438 if (ctx)
1439 ntfs_attr_put_search_ctx(ctx);
1440 unmap_mft_record(base_ni);
1441 err_out:
1442 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1443 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1444 "Marking corrupt inode and base inode 0x%lx as bad. "
1445 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1446 base_vi->i_ino);
1447 make_bad_inode(vi);
1448 if (err != -ENOMEM)
1449 NVolSetErrors(vol);
1450 return err;
1454 * ntfs_read_locked_index_inode - read an index inode from its base inode
1455 * @base_vi: base inode
1456 * @vi: index inode to read
1458 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1459 * index inode described by @vi into memory from the base mft record described
1460 * by @base_ni.
1462 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1463 * reading and looks up the attributes relating to the index described by @vi
1464 * before setting up the necessary fields in @vi as well as initializing the
1465 * ntfs inode.
1467 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1468 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1469 * are setup like directory inodes since directories are a special case of
1470 * indices ao they need to be treated in much the same way. Most importantly,
1471 * for small indices the index allocation attribute might not actually exist.
1472 * However, the index root attribute always exists but this does not need to
1473 * have an inode associated with it and this is why we define a new inode type
1474 * index. Also, like for directories, we need to have an attribute inode for
1475 * the bitmap attribute corresponding to the index allocation attribute and we
1476 * can store this in the appropriate field of the inode, just like we do for
1477 * normal directory inodes.
1479 * Q: What locks are held when the function is called?
1480 * A: i_state has I_LOCK set, hence the inode is locked, also
1481 * i_count is set to 1, so it is not going to go away
1483 * Return 0 on success and -errno on error. In the error case, the inode will
1484 * have had make_bad_inode() executed on it.
1486 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1488 loff_t bvi_size;
1489 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1490 ntfs_inode *ni, *base_ni, *bni;
1491 struct inode *bvi;
1492 MFT_RECORD *m;
1493 ATTR_RECORD *a;
1494 ntfs_attr_search_ctx *ctx;
1495 INDEX_ROOT *ir;
1496 u8 *ir_end, *index_end;
1497 int err = 0;
1499 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1500 ntfs_init_big_inode(vi);
1501 ni = NTFS_I(vi);
1502 base_ni = NTFS_I(base_vi);
1503 /* Just mirror the values from the base inode. */
1504 vi->i_version = base_vi->i_version;
1505 vi->i_uid = base_vi->i_uid;
1506 vi->i_gid = base_vi->i_gid;
1507 vi->i_nlink = base_vi->i_nlink;
1508 vi->i_mtime = base_vi->i_mtime;
1509 vi->i_ctime = base_vi->i_ctime;
1510 vi->i_atime = base_vi->i_atime;
1511 vi->i_generation = ni->seq_no = base_ni->seq_no;
1512 /* Set inode type to zero but preserve permissions. */
1513 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1514 /* Map the mft record for the base inode. */
1515 m = map_mft_record(base_ni);
1516 if (IS_ERR(m)) {
1517 err = PTR_ERR(m);
1518 goto err_out;
1520 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1521 if (!ctx) {
1522 err = -ENOMEM;
1523 goto unm_err_out;
1525 /* Find the index root attribute. */
1526 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1527 CASE_SENSITIVE, 0, NULL, 0, ctx);
1528 if (unlikely(err)) {
1529 if (err == -ENOENT)
1530 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1531 "missing.");
1532 goto unm_err_out;
1534 a = ctx->attr;
1535 /* Set up the state. */
1536 if (unlikely(a->non_resident)) {
1537 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1538 goto unm_err_out;
1540 /* Ensure the attribute name is placed before the value. */
1541 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1542 le16_to_cpu(a->data.resident.value_offset)))) {
1543 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1544 "after the attribute value.");
1545 goto unm_err_out;
1548 * Compressed/encrypted/sparse index root is not allowed, except for
1549 * directories of course but those are not dealt with here.
1551 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1552 ATTR_IS_SPARSE)) {
1553 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1554 "root attribute.");
1555 goto unm_err_out;
1557 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1558 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1559 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1560 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1561 goto unm_err_out;
1563 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1564 if (index_end > ir_end) {
1565 ntfs_error(vi->i_sb, "Index is corrupt.");
1566 goto unm_err_out;
1568 if (ir->type) {
1569 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1570 le32_to_cpu(ir->type));
1571 goto unm_err_out;
1573 ni->itype.index.collation_rule = ir->collation_rule;
1574 ntfs_debug("Index collation rule is 0x%x.",
1575 le32_to_cpu(ir->collation_rule));
1576 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1577 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1578 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1579 "two.", ni->itype.index.block_size);
1580 goto unm_err_out;
1582 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1583 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1584 "(%ld) is not supported. Sorry.",
1585 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1586 err = -EOPNOTSUPP;
1587 goto unm_err_out;
1589 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1590 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1591 "(%i) is not supported. Sorry.",
1592 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1593 err = -EOPNOTSUPP;
1594 goto unm_err_out;
1596 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1597 /* Determine the size of a vcn in the index. */
1598 if (vol->cluster_size <= ni->itype.index.block_size) {
1599 ni->itype.index.vcn_size = vol->cluster_size;
1600 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1601 } else {
1602 ni->itype.index.vcn_size = vol->sector_size;
1603 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1605 /* Check for presence of index allocation attribute. */
1606 if (!(ir->index.flags & LARGE_INDEX)) {
1607 /* No index allocation. */
1608 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1609 /* We are done with the mft record, so we release it. */
1610 ntfs_attr_put_search_ctx(ctx);
1611 unmap_mft_record(base_ni);
1612 m = NULL;
1613 ctx = NULL;
1614 goto skip_large_index_stuff;
1615 } /* LARGE_INDEX: Index allocation present. Setup state. */
1616 NInoSetIndexAllocPresent(ni);
1617 /* Find index allocation attribute. */
1618 ntfs_attr_reinit_search_ctx(ctx);
1619 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1620 CASE_SENSITIVE, 0, NULL, 0, ctx);
1621 if (unlikely(err)) {
1622 if (err == -ENOENT)
1623 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1624 "not present but $INDEX_ROOT "
1625 "indicated it is.");
1626 else
1627 ntfs_error(vi->i_sb, "Failed to lookup "
1628 "$INDEX_ALLOCATION attribute.");
1629 goto unm_err_out;
1631 a = ctx->attr;
1632 if (!a->non_resident) {
1633 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1634 "resident.");
1635 goto unm_err_out;
1638 * Ensure the attribute name is placed before the mapping pairs array.
1640 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1641 le16_to_cpu(
1642 a->data.non_resident.mapping_pairs_offset)))) {
1643 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1644 "placed after the mapping pairs array.");
1645 goto unm_err_out;
1647 if (a->flags & ATTR_IS_ENCRYPTED) {
1648 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1649 "encrypted.");
1650 goto unm_err_out;
1652 if (a->flags & ATTR_IS_SPARSE) {
1653 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1654 goto unm_err_out;
1656 if (a->flags & ATTR_COMPRESSION_MASK) {
1657 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1658 "compressed.");
1659 goto unm_err_out;
1661 if (a->data.non_resident.lowest_vcn) {
1662 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1663 "attribute has non zero lowest_vcn.");
1664 goto unm_err_out;
1666 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1667 ni->initialized_size = sle64_to_cpu(
1668 a->data.non_resident.initialized_size);
1669 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1671 * We are done with the mft record, so we release it. Otherwise
1672 * we would deadlock in ntfs_attr_iget().
1674 ntfs_attr_put_search_ctx(ctx);
1675 unmap_mft_record(base_ni);
1676 m = NULL;
1677 ctx = NULL;
1678 /* Get the index bitmap attribute inode. */
1679 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1680 if (IS_ERR(bvi)) {
1681 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1682 err = PTR_ERR(bvi);
1683 goto unm_err_out;
1685 bni = NTFS_I(bvi);
1686 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1687 NInoSparse(bni)) {
1688 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1689 "encrypted and/or sparse.");
1690 goto iput_unm_err_out;
1692 /* Consistency check bitmap size vs. index allocation size. */
1693 bvi_size = i_size_read(bvi);
1694 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1695 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1696 "index allocation (0x%llx).", bvi_size << 3,
1697 vi->i_size);
1698 goto iput_unm_err_out;
1700 iput(bvi);
1701 skip_large_index_stuff:
1702 /* Setup the operations for this index inode. */
1703 vi->i_op = NULL;
1704 vi->i_fop = NULL;
1705 vi->i_mapping->a_ops = &ntfs_mst_aops;
1706 vi->i_blocks = ni->allocated_size >> 9;
1708 * Make sure the base inode doesn't go away and attach it to the
1709 * index inode.
1711 igrab(base_vi);
1712 ni->ext.base_ntfs_ino = base_ni;
1713 ni->nr_extents = -1;
1715 ntfs_debug("Done.");
1716 return 0;
1717 iput_unm_err_out:
1718 iput(bvi);
1719 unm_err_out:
1720 if (!err)
1721 err = -EIO;
1722 if (ctx)
1723 ntfs_attr_put_search_ctx(ctx);
1724 if (m)
1725 unmap_mft_record(base_ni);
1726 err_out:
1727 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1728 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1729 ni->name_len);
1730 make_bad_inode(vi);
1731 if (err != -EOPNOTSUPP && err != -ENOMEM)
1732 NVolSetErrors(vol);
1733 return err;
1737 * The MFT inode has special locking, so teach the lock validator
1738 * about this by splitting off the locking rules of the MFT from
1739 * the locking rules of other inodes. The MFT inode can never be
1740 * accessed from the VFS side (or even internally), only by the
1741 * map_mft functions.
1743 static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
1746 * ntfs_read_inode_mount - special read_inode for mount time use only
1747 * @vi: inode to read
1749 * Read inode FILE_MFT at mount time, only called with super_block lock
1750 * held from within the read_super() code path.
1752 * This function exists because when it is called the page cache for $MFT/$DATA
1753 * is not initialized and hence we cannot get at the contents of mft records
1754 * by calling map_mft_record*().
1756 * Further it needs to cope with the circular references problem, i.e. cannot
1757 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1758 * we do not know where the other extent mft records are yet and again, because
1759 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1760 * attribute list is actually present in $MFT inode.
1762 * We solve these problems by starting with the $DATA attribute before anything
1763 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1764 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1765 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1766 * sufficient information for the next step to complete.
1768 * This should work but there are two possible pit falls (see inline comments
1769 * below), but only time will tell if they are real pits or just smoke...
1771 int ntfs_read_inode_mount(struct inode *vi)
1773 VCN next_vcn, last_vcn, highest_vcn;
1774 s64 block;
1775 struct super_block *sb = vi->i_sb;
1776 ntfs_volume *vol = NTFS_SB(sb);
1777 struct buffer_head *bh;
1778 ntfs_inode *ni;
1779 MFT_RECORD *m = NULL;
1780 ATTR_RECORD *a;
1781 ntfs_attr_search_ctx *ctx;
1782 unsigned int i, nr_blocks;
1783 int err;
1785 ntfs_debug("Entering.");
1787 /* Initialize the ntfs specific part of @vi. */
1788 ntfs_init_big_inode(vi);
1790 ni = NTFS_I(vi);
1792 /* Setup the data attribute. It is special as it is mst protected. */
1793 NInoSetNonResident(ni);
1794 NInoSetMstProtected(ni);
1795 NInoSetSparseDisabled(ni);
1796 ni->type = AT_DATA;
1797 ni->name = NULL;
1798 ni->name_len = 0;
1800 * This sets up our little cheat allowing us to reuse the async read io
1801 * completion handler for directories.
1803 ni->itype.index.block_size = vol->mft_record_size;
1804 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1806 /* Very important! Needed to be able to call map_mft_record*(). */
1807 vol->mft_ino = vi;
1809 /* Allocate enough memory to read the first mft record. */
1810 if (vol->mft_record_size > 64 * 1024) {
1811 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1812 vol->mft_record_size);
1813 goto err_out;
1815 i = vol->mft_record_size;
1816 if (i < sb->s_blocksize)
1817 i = sb->s_blocksize;
1818 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1819 if (!m) {
1820 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1821 goto err_out;
1824 /* Determine the first block of the $MFT/$DATA attribute. */
1825 block = vol->mft_lcn << vol->cluster_size_bits >>
1826 sb->s_blocksize_bits;
1827 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1828 if (!nr_blocks)
1829 nr_blocks = 1;
1831 /* Load $MFT/$DATA's first mft record. */
1832 for (i = 0; i < nr_blocks; i++) {
1833 bh = sb_bread(sb, block++);
1834 if (!bh) {
1835 ntfs_error(sb, "Device read failed.");
1836 goto err_out;
1838 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1839 sb->s_blocksize);
1840 brelse(bh);
1843 /* Apply the mst fixups. */
1844 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1845 /* FIXME: Try to use the $MFTMirr now. */
1846 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1847 goto err_out;
1850 /* Need this to sanity check attribute list references to $MFT. */
1851 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1853 /* Provides readpage() and sync_page() for map_mft_record(). */
1854 vi->i_mapping->a_ops = &ntfs_mst_aops;
1856 ctx = ntfs_attr_get_search_ctx(ni, m);
1857 if (!ctx) {
1858 err = -ENOMEM;
1859 goto err_out;
1862 /* Find the attribute list attribute if present. */
1863 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1864 if (err) {
1865 if (unlikely(err != -ENOENT)) {
1866 ntfs_error(sb, "Failed to lookup attribute list "
1867 "attribute. You should run chkdsk.");
1868 goto put_err_out;
1870 } else /* if (!err) */ {
1871 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1872 u8 *al_end;
1873 static const char *es = " Not allowed. $MFT is corrupt. "
1874 "You should run chkdsk.";
1876 ntfs_debug("Attribute list attribute found in $MFT.");
1877 NInoSetAttrList(ni);
1878 a = ctx->attr;
1879 if (a->flags & ATTR_COMPRESSION_MASK) {
1880 ntfs_error(sb, "Attribute list attribute is "
1881 "compressed.%s", es);
1882 goto put_err_out;
1884 if (a->flags & ATTR_IS_ENCRYPTED ||
1885 a->flags & ATTR_IS_SPARSE) {
1886 if (a->non_resident) {
1887 ntfs_error(sb, "Non-resident attribute list "
1888 "attribute is encrypted/"
1889 "sparse.%s", es);
1890 goto put_err_out;
1892 ntfs_warning(sb, "Resident attribute list attribute "
1893 "in $MFT system file is marked "
1894 "encrypted/sparse which is not true. "
1895 "However, Windows allows this and "
1896 "chkdsk does not detect or correct it "
1897 "so we will just ignore the invalid "
1898 "flags and pretend they are not set.");
1900 /* Now allocate memory for the attribute list. */
1901 ni->attr_list_size = (u32)ntfs_attr_size(a);
1902 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1903 if (!ni->attr_list) {
1904 ntfs_error(sb, "Not enough memory to allocate buffer "
1905 "for attribute list.");
1906 goto put_err_out;
1908 if (a->non_resident) {
1909 NInoSetAttrListNonResident(ni);
1910 if (a->data.non_resident.lowest_vcn) {
1911 ntfs_error(sb, "Attribute list has non zero "
1912 "lowest_vcn. $MFT is corrupt. "
1913 "You should run chkdsk.");
1914 goto put_err_out;
1916 /* Setup the runlist. */
1917 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1918 a, NULL);
1919 if (IS_ERR(ni->attr_list_rl.rl)) {
1920 err = PTR_ERR(ni->attr_list_rl.rl);
1921 ni->attr_list_rl.rl = NULL;
1922 ntfs_error(sb, "Mapping pairs decompression "
1923 "failed with error code %i.",
1924 -err);
1925 goto put_err_out;
1927 /* Now load the attribute list. */
1928 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1929 ni->attr_list, ni->attr_list_size,
1930 sle64_to_cpu(a->data.
1931 non_resident.initialized_size)))) {
1932 ntfs_error(sb, "Failed to load attribute list "
1933 "attribute with error code %i.",
1934 -err);
1935 goto put_err_out;
1937 } else /* if (!ctx.attr->non_resident) */ {
1938 if ((u8*)a + le16_to_cpu(
1939 a->data.resident.value_offset) +
1940 le32_to_cpu(
1941 a->data.resident.value_length) >
1942 (u8*)ctx->mrec + vol->mft_record_size) {
1943 ntfs_error(sb, "Corrupt attribute list "
1944 "attribute.");
1945 goto put_err_out;
1947 /* Now copy the attribute list. */
1948 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1949 a->data.resident.value_offset),
1950 le32_to_cpu(
1951 a->data.resident.value_length));
1953 /* The attribute list is now setup in memory. */
1955 * FIXME: I don't know if this case is actually possible.
1956 * According to logic it is not possible but I have seen too
1957 * many weird things in MS software to rely on logic... Thus we
1958 * perform a manual search and make sure the first $MFT/$DATA
1959 * extent is in the base inode. If it is not we abort with an
1960 * error and if we ever see a report of this error we will need
1961 * to do some magic in order to have the necessary mft record
1962 * loaded and in the right place in the page cache. But
1963 * hopefully logic will prevail and this never happens...
1965 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1966 al_end = (u8*)al_entry + ni->attr_list_size;
1967 for (;; al_entry = next_al_entry) {
1968 /* Out of bounds check. */
1969 if ((u8*)al_entry < ni->attr_list ||
1970 (u8*)al_entry > al_end)
1971 goto em_put_err_out;
1972 /* Catch the end of the attribute list. */
1973 if ((u8*)al_entry == al_end)
1974 goto em_put_err_out;
1975 if (!al_entry->length)
1976 goto em_put_err_out;
1977 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1978 le16_to_cpu(al_entry->length) > al_end)
1979 goto em_put_err_out;
1980 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1981 le16_to_cpu(al_entry->length));
1982 if (le32_to_cpu(al_entry->type) >
1983 const_le32_to_cpu(AT_DATA))
1984 goto em_put_err_out;
1985 if (AT_DATA != al_entry->type)
1986 continue;
1987 /* We want an unnamed attribute. */
1988 if (al_entry->name_length)
1989 goto em_put_err_out;
1990 /* Want the first entry, i.e. lowest_vcn == 0. */
1991 if (al_entry->lowest_vcn)
1992 goto em_put_err_out;
1993 /* First entry has to be in the base mft record. */
1994 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1995 /* MFT references do not match, logic fails. */
1996 ntfs_error(sb, "BUG: The first $DATA extent "
1997 "of $MFT is not in the base "
1998 "mft record. Please report "
1999 "you saw this message to "
2000 "linux-ntfs-dev@lists."
2001 "sourceforge.net");
2002 goto put_err_out;
2003 } else {
2004 /* Sequence numbers must match. */
2005 if (MSEQNO_LE(al_entry->mft_reference) !=
2006 ni->seq_no)
2007 goto em_put_err_out;
2008 /* Got it. All is ok. We can stop now. */
2009 break;
2014 ntfs_attr_reinit_search_ctx(ctx);
2016 /* Now load all attribute extents. */
2017 a = NULL;
2018 next_vcn = last_vcn = highest_vcn = 0;
2019 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
2020 ctx))) {
2021 runlist_element *nrl;
2023 /* Cache the current attribute. */
2024 a = ctx->attr;
2025 /* $MFT must be non-resident. */
2026 if (!a->non_resident) {
2027 ntfs_error(sb, "$MFT must be non-resident but a "
2028 "resident extent was found. $MFT is "
2029 "corrupt. Run chkdsk.");
2030 goto put_err_out;
2032 /* $MFT must be uncompressed and unencrypted. */
2033 if (a->flags & ATTR_COMPRESSION_MASK ||
2034 a->flags & ATTR_IS_ENCRYPTED ||
2035 a->flags & ATTR_IS_SPARSE) {
2036 ntfs_error(sb, "$MFT must be uncompressed, "
2037 "non-sparse, and unencrypted but a "
2038 "compressed/sparse/encrypted extent "
2039 "was found. $MFT is corrupt. Run "
2040 "chkdsk.");
2041 goto put_err_out;
2044 * Decompress the mapping pairs array of this extent and merge
2045 * the result into the existing runlist. No need for locking
2046 * as we have exclusive access to the inode at this time and we
2047 * are a mount in progress task, too.
2049 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2050 if (IS_ERR(nrl)) {
2051 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2052 "failed with error code %ld. $MFT is "
2053 "corrupt.", PTR_ERR(nrl));
2054 goto put_err_out;
2056 ni->runlist.rl = nrl;
2058 /* Are we in the first extent? */
2059 if (!next_vcn) {
2060 if (a->data.non_resident.lowest_vcn) {
2061 ntfs_error(sb, "First extent of $DATA "
2062 "attribute has non zero "
2063 "lowest_vcn. $MFT is corrupt. "
2064 "You should run chkdsk.");
2065 goto put_err_out;
2067 /* Get the last vcn in the $DATA attribute. */
2068 last_vcn = sle64_to_cpu(
2069 a->data.non_resident.allocated_size)
2070 >> vol->cluster_size_bits;
2071 /* Fill in the inode size. */
2072 vi->i_size = sle64_to_cpu(
2073 a->data.non_resident.data_size);
2074 ni->initialized_size = sle64_to_cpu(
2075 a->data.non_resident.initialized_size);
2076 ni->allocated_size = sle64_to_cpu(
2077 a->data.non_resident.allocated_size);
2079 * Verify the number of mft records does not exceed
2080 * 2^32 - 1.
2082 if ((vi->i_size >> vol->mft_record_size_bits) >=
2083 (1ULL << 32)) {
2084 ntfs_error(sb, "$MFT is too big! Aborting.");
2085 goto put_err_out;
2088 * We have got the first extent of the runlist for
2089 * $MFT which means it is now relatively safe to call
2090 * the normal ntfs_read_inode() function.
2091 * Complete reading the inode, this will actually
2092 * re-read the mft record for $MFT, this time entering
2093 * it into the page cache with which we complete the
2094 * kick start of the volume. It should be safe to do
2095 * this now as the first extent of $MFT/$DATA is
2096 * already known and we would hope that we don't need
2097 * further extents in order to find the other
2098 * attributes belonging to $MFT. Only time will tell if
2099 * this is really the case. If not we will have to play
2100 * magic at this point, possibly duplicating a lot of
2101 * ntfs_read_inode() at this point. We will need to
2102 * ensure we do enough of its work to be able to call
2103 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2104 * hope this never happens...
2106 ntfs_read_locked_inode(vi);
2107 if (is_bad_inode(vi)) {
2108 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2109 "failed. BUG or corrupt $MFT. "
2110 "Run chkdsk and if no errors "
2111 "are found, please report you "
2112 "saw this message to "
2113 "linux-ntfs-dev@lists."
2114 "sourceforge.net");
2115 ntfs_attr_put_search_ctx(ctx);
2116 /* Revert to the safe super operations. */
2117 ntfs_free(m);
2118 return -1;
2121 * Re-initialize some specifics about $MFT's inode as
2122 * ntfs_read_inode() will have set up the default ones.
2124 /* Set uid and gid to root. */
2125 vi->i_uid = vi->i_gid = 0;
2126 /* Regular file. No access for anyone. */
2127 vi->i_mode = S_IFREG;
2128 /* No VFS initiated operations allowed for $MFT. */
2129 vi->i_op = &ntfs_empty_inode_ops;
2130 vi->i_fop = &ntfs_empty_file_ops;
2133 /* Get the lowest vcn for the next extent. */
2134 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2135 next_vcn = highest_vcn + 1;
2137 /* Only one extent or error, which we catch below. */
2138 if (next_vcn <= 0)
2139 break;
2141 /* Avoid endless loops due to corruption. */
2142 if (next_vcn < sle64_to_cpu(
2143 a->data.non_resident.lowest_vcn)) {
2144 ntfs_error(sb, "$MFT has corrupt attribute list "
2145 "attribute. Run chkdsk.");
2146 goto put_err_out;
2149 if (err != -ENOENT) {
2150 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2151 "$MFT is corrupt. Run chkdsk.");
2152 goto put_err_out;
2154 if (!a) {
2155 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2156 "corrupt. Run chkdsk.");
2157 goto put_err_out;
2159 if (highest_vcn && highest_vcn != last_vcn - 1) {
2160 ntfs_error(sb, "Failed to load the complete runlist for "
2161 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2162 "Run chkdsk.");
2163 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2164 (unsigned long long)highest_vcn,
2165 (unsigned long long)last_vcn - 1);
2166 goto put_err_out;
2168 ntfs_attr_put_search_ctx(ctx);
2169 ntfs_debug("Done.");
2170 ntfs_free(m);
2173 * Split the locking rules of the MFT inode from the
2174 * locking rules of other inodes:
2176 lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
2177 lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
2179 return 0;
2181 em_put_err_out:
2182 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2183 "attribute list. $MFT is corrupt. Run chkdsk.");
2184 put_err_out:
2185 ntfs_attr_put_search_ctx(ctx);
2186 err_out:
2187 ntfs_error(sb, "Failed. Marking inode as bad.");
2188 make_bad_inode(vi);
2189 ntfs_free(m);
2190 return -1;
2193 static void __ntfs_clear_inode(ntfs_inode *ni)
2195 /* Free all alocated memory. */
2196 down_write(&ni->runlist.lock);
2197 if (ni->runlist.rl) {
2198 ntfs_free(ni->runlist.rl);
2199 ni->runlist.rl = NULL;
2201 up_write(&ni->runlist.lock);
2203 if (ni->attr_list) {
2204 ntfs_free(ni->attr_list);
2205 ni->attr_list = NULL;
2208 down_write(&ni->attr_list_rl.lock);
2209 if (ni->attr_list_rl.rl) {
2210 ntfs_free(ni->attr_list_rl.rl);
2211 ni->attr_list_rl.rl = NULL;
2213 up_write(&ni->attr_list_rl.lock);
2215 if (ni->name_len && ni->name != I30) {
2216 /* Catch bugs... */
2217 BUG_ON(!ni->name);
2218 kfree(ni->name);
2222 void ntfs_clear_extent_inode(ntfs_inode *ni)
2224 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2226 BUG_ON(NInoAttr(ni));
2227 BUG_ON(ni->nr_extents != -1);
2229 #ifdef NTFS_RW
2230 if (NInoDirty(ni)) {
2231 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2232 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2233 "Losing data! This is a BUG!!!");
2234 // FIXME: Do something!!!
2236 #endif /* NTFS_RW */
2238 __ntfs_clear_inode(ni);
2240 /* Bye, bye... */
2241 ntfs_destroy_extent_inode(ni);
2245 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2246 * @vi: vfs inode pending annihilation
2248 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2249 * is called, which deallocates all memory belonging to the NTFS specific part
2250 * of the inode and returns.
2252 * If the MFT record is dirty, we commit it before doing anything else.
2254 void ntfs_clear_big_inode(struct inode *vi)
2256 ntfs_inode *ni = NTFS_I(vi);
2258 #ifdef NTFS_RW
2259 if (NInoDirty(ni)) {
2260 bool was_bad = (is_bad_inode(vi));
2262 /* Committing the inode also commits all extent inodes. */
2263 ntfs_commit_inode(vi);
2265 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2266 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2267 "0x%lx. Losing data!", vi->i_ino);
2268 // FIXME: Do something!!!
2271 #endif /* NTFS_RW */
2273 /* No need to lock at this stage as no one else has a reference. */
2274 if (ni->nr_extents > 0) {
2275 int i;
2277 for (i = 0; i < ni->nr_extents; i++)
2278 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2279 kfree(ni->ext.extent_ntfs_inos);
2282 __ntfs_clear_inode(ni);
2284 if (NInoAttr(ni)) {
2285 /* Release the base inode if we are holding it. */
2286 if (ni->nr_extents == -1) {
2287 iput(VFS_I(ni->ext.base_ntfs_ino));
2288 ni->nr_extents = 0;
2289 ni->ext.base_ntfs_ino = NULL;
2292 return;
2296 * ntfs_show_options - show mount options in /proc/mounts
2297 * @sf: seq_file in which to write our mount options
2298 * @mnt: vfs mount whose mount options to display
2300 * Called by the VFS once for each mounted ntfs volume when someone reads
2301 * /proc/mounts in order to display the NTFS specific mount options of each
2302 * mount. The mount options of the vfs mount @mnt are written to the seq file
2303 * @sf and success is returned.
2305 int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2307 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2308 int i;
2310 seq_printf(sf, ",uid=%i", vol->uid);
2311 seq_printf(sf, ",gid=%i", vol->gid);
2312 if (vol->fmask == vol->dmask)
2313 seq_printf(sf, ",umask=0%o", vol->fmask);
2314 else {
2315 seq_printf(sf, ",fmask=0%o", vol->fmask);
2316 seq_printf(sf, ",dmask=0%o", vol->dmask);
2318 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2319 if (NVolCaseSensitive(vol))
2320 seq_printf(sf, ",case_sensitive");
2321 if (NVolShowSystemFiles(vol))
2322 seq_printf(sf, ",show_sys_files");
2323 if (!NVolSparseEnabled(vol))
2324 seq_printf(sf, ",disable_sparse");
2325 for (i = 0; on_errors_arr[i].val; i++) {
2326 if (on_errors_arr[i].val & vol->on_errors)
2327 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2329 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2330 return 0;
2333 #ifdef NTFS_RW
2335 static const char *es = " Leaving inconsistent metadata. Unmount and run "
2336 "chkdsk.";
2339 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2340 * @vi: inode for which the i_size was changed
2342 * We only support i_size changes for normal files at present, i.e. not
2343 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2344 * below.
2346 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2347 * that the change is allowed.
2349 * This implies for us that @vi is a file inode rather than a directory, index,
2350 * or attribute inode as well as that @vi is a base inode.
2352 * Returns 0 on success or -errno on error.
2354 * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for
2355 * writing. The only case in the kernel where ->i_alloc_sem is not held is
2356 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2357 * with the current i_size as the offset. The analogous place in NTFS is in
2358 * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again
2359 * without holding ->i_alloc_sem.
2361 int ntfs_truncate(struct inode *vi)
2363 s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2364 VCN highest_vcn;
2365 unsigned long flags;
2366 ntfs_inode *base_ni, *ni = NTFS_I(vi);
2367 ntfs_volume *vol = ni->vol;
2368 ntfs_attr_search_ctx *ctx;
2369 MFT_RECORD *m;
2370 ATTR_RECORD *a;
2371 const char *te = " Leaving file length out of sync with i_size.";
2372 int err, mp_size, size_change, alloc_change;
2373 u32 attr_len;
2375 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2376 BUG_ON(NInoAttr(ni));
2377 BUG_ON(S_ISDIR(vi->i_mode));
2378 BUG_ON(NInoMstProtected(ni));
2379 BUG_ON(ni->nr_extents < 0);
2380 retry_truncate:
2382 * Lock the runlist for writing and map the mft record to ensure it is
2383 * safe to mess with the attribute runlist and sizes.
2385 down_write(&ni->runlist.lock);
2386 if (!NInoAttr(ni))
2387 base_ni = ni;
2388 else
2389 base_ni = ni->ext.base_ntfs_ino;
2390 m = map_mft_record(base_ni);
2391 if (IS_ERR(m)) {
2392 err = PTR_ERR(m);
2393 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2394 "(error code %d).%s", vi->i_ino, err, te);
2395 ctx = NULL;
2396 m = NULL;
2397 goto old_bad_out;
2399 ctx = ntfs_attr_get_search_ctx(base_ni, m);
2400 if (unlikely(!ctx)) {
2401 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2402 "inode 0x%lx (not enough memory).%s",
2403 vi->i_ino, te);
2404 err = -ENOMEM;
2405 goto old_bad_out;
2407 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2408 CASE_SENSITIVE, 0, NULL, 0, ctx);
2409 if (unlikely(err)) {
2410 if (err == -ENOENT) {
2411 ntfs_error(vi->i_sb, "Open attribute is missing from "
2412 "mft record. Inode 0x%lx is corrupt. "
2413 "Run chkdsk.%s", vi->i_ino, te);
2414 err = -EIO;
2415 } else
2416 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2417 "inode 0x%lx (error code %d).%s",
2418 vi->i_ino, err, te);
2419 goto old_bad_out;
2421 m = ctx->mrec;
2422 a = ctx->attr;
2424 * The i_size of the vfs inode is the new size for the attribute value.
2426 new_size = i_size_read(vi);
2427 /* The current size of the attribute value is the old size. */
2428 old_size = ntfs_attr_size(a);
2429 /* Calculate the new allocated size. */
2430 if (NInoNonResident(ni))
2431 new_alloc_size = (new_size + vol->cluster_size - 1) &
2432 ~(s64)vol->cluster_size_mask;
2433 else
2434 new_alloc_size = (new_size + 7) & ~7;
2435 /* The current allocated size is the old allocated size. */
2436 read_lock_irqsave(&ni->size_lock, flags);
2437 old_alloc_size = ni->allocated_size;
2438 read_unlock_irqrestore(&ni->size_lock, flags);
2440 * The change in the file size. This will be 0 if no change, >0 if the
2441 * size is growing, and <0 if the size is shrinking.
2443 size_change = -1;
2444 if (new_size - old_size >= 0) {
2445 size_change = 1;
2446 if (new_size == old_size)
2447 size_change = 0;
2449 /* As above for the allocated size. */
2450 alloc_change = -1;
2451 if (new_alloc_size - old_alloc_size >= 0) {
2452 alloc_change = 1;
2453 if (new_alloc_size == old_alloc_size)
2454 alloc_change = 0;
2457 * If neither the size nor the allocation are being changed there is
2458 * nothing to do.
2460 if (!size_change && !alloc_change)
2461 goto unm_done;
2462 /* If the size is changing, check if new size is allowed in $AttrDef. */
2463 if (size_change) {
2464 err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2465 if (unlikely(err)) {
2466 if (err == -ERANGE) {
2467 ntfs_error(vol->sb, "Truncate would cause the "
2468 "inode 0x%lx to %simum size "
2469 "for its attribute type "
2470 "(0x%x). Aborting truncate.",
2471 vi->i_ino,
2472 new_size > old_size ? "exceed "
2473 "the max" : "go under the min",
2474 le32_to_cpu(ni->type));
2475 err = -EFBIG;
2476 } else {
2477 ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2478 "attribute type 0x%x. "
2479 "Aborting truncate.",
2480 vi->i_ino,
2481 le32_to_cpu(ni->type));
2482 err = -EIO;
2484 /* Reset the vfs inode size to the old size. */
2485 i_size_write(vi, old_size);
2486 goto err_out;
2489 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2490 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2491 "supported yet for %s files, ignoring.",
2492 NInoCompressed(ni) ? "compressed" :
2493 "encrypted");
2494 err = -EOPNOTSUPP;
2495 goto bad_out;
2497 if (a->non_resident)
2498 goto do_non_resident_truncate;
2499 BUG_ON(NInoNonResident(ni));
2500 /* Resize the attribute record to best fit the new attribute size. */
2501 if (new_size < vol->mft_record_size &&
2502 !ntfs_resident_attr_value_resize(m, a, new_size)) {
2503 unsigned long flags;
2505 /* The resize succeeded! */
2506 flush_dcache_mft_record_page(ctx->ntfs_ino);
2507 mark_mft_record_dirty(ctx->ntfs_ino);
2508 write_lock_irqsave(&ni->size_lock, flags);
2509 /* Update the sizes in the ntfs inode and all is done. */
2510 ni->allocated_size = le32_to_cpu(a->length) -
2511 le16_to_cpu(a->data.resident.value_offset);
2513 * Note ntfs_resident_attr_value_resize() has already done any
2514 * necessary data clearing in the attribute record. When the
2515 * file is being shrunk vmtruncate() will already have cleared
2516 * the top part of the last partial page, i.e. since this is
2517 * the resident case this is the page with index 0. However,
2518 * when the file is being expanded, the page cache page data
2519 * between the old data_size, i.e. old_size, and the new_size
2520 * has not been zeroed. Fortunately, we do not need to zero it
2521 * either since on one hand it will either already be zero due
2522 * to both readpage and writepage clearing partial page data
2523 * beyond i_size in which case there is nothing to do or in the
2524 * case of the file being mmap()ped at the same time, POSIX
2525 * specifies that the behaviour is unspecified thus we do not
2526 * have to do anything. This means that in our implementation
2527 * in the rare case that the file is mmap()ped and a write
2528 * occured into the mmap()ped region just beyond the file size
2529 * and writepage has not yet been called to write out the page
2530 * (which would clear the area beyond the file size) and we now
2531 * extend the file size to incorporate this dirty region
2532 * outside the file size, a write of the page would result in
2533 * this data being written to disk instead of being cleared.
2534 * Given both POSIX and the Linux mmap(2) man page specify that
2535 * this corner case is undefined, we choose to leave it like
2536 * that as this is much simpler for us as we cannot lock the
2537 * relevant page now since we are holding too many ntfs locks
2538 * which would result in a lock reversal deadlock.
2540 ni->initialized_size = new_size;
2541 write_unlock_irqrestore(&ni->size_lock, flags);
2542 goto unm_done;
2544 /* If the above resize failed, this must be an attribute extension. */
2545 BUG_ON(size_change < 0);
2547 * We have to drop all the locks so we can call
2548 * ntfs_attr_make_non_resident(). This could be optimised by try-
2549 * locking the first page cache page and only if that fails dropping
2550 * the locks, locking the page, and redoing all the locking and
2551 * lookups. While this would be a huge optimisation, it is not worth
2552 * it as this is definitely a slow code path as it only ever can happen
2553 * once for any given file.
2555 ntfs_attr_put_search_ctx(ctx);
2556 unmap_mft_record(base_ni);
2557 up_write(&ni->runlist.lock);
2559 * Not enough space in the mft record, try to make the attribute
2560 * non-resident and if successful restart the truncation process.
2562 err = ntfs_attr_make_non_resident(ni, old_size);
2563 if (likely(!err))
2564 goto retry_truncate;
2566 * Could not make non-resident. If this is due to this not being
2567 * permitted for this attribute type or there not being enough space,
2568 * try to make other attributes non-resident. Otherwise fail.
2570 if (unlikely(err != -EPERM && err != -ENOSPC)) {
2571 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2572 "type 0x%x, because the conversion from "
2573 "resident to non-resident attribute failed "
2574 "with error code %i.", vi->i_ino,
2575 (unsigned)le32_to_cpu(ni->type), err);
2576 if (err != -ENOMEM)
2577 err = -EIO;
2578 goto conv_err_out;
2580 /* TODO: Not implemented from here, abort. */
2581 if (err == -ENOSPC)
2582 ntfs_error(vol->sb, "Not enough space in the mft record/on "
2583 "disk for the non-resident attribute value. "
2584 "This case is not implemented yet.");
2585 else /* if (err == -EPERM) */
2586 ntfs_error(vol->sb, "This attribute type may not be "
2587 "non-resident. This case is not implemented "
2588 "yet.");
2589 err = -EOPNOTSUPP;
2590 goto conv_err_out;
2591 #if 0
2592 // TODO: Attempt to make other attributes non-resident.
2593 if (!err)
2594 goto do_resident_extend;
2596 * Both the attribute list attribute and the standard information
2597 * attribute must remain in the base inode. Thus, if this is one of
2598 * these attributes, we have to try to move other attributes out into
2599 * extent mft records instead.
2601 if (ni->type == AT_ATTRIBUTE_LIST ||
2602 ni->type == AT_STANDARD_INFORMATION) {
2603 // TODO: Attempt to move other attributes into extent mft
2604 // records.
2605 err = -EOPNOTSUPP;
2606 if (!err)
2607 goto do_resident_extend;
2608 goto err_out;
2610 // TODO: Attempt to move this attribute to an extent mft record, but
2611 // only if it is not already the only attribute in an mft record in
2612 // which case there would be nothing to gain.
2613 err = -EOPNOTSUPP;
2614 if (!err)
2615 goto do_resident_extend;
2616 /* There is nothing we can do to make enough space. )-: */
2617 goto err_out;
2618 #endif
2619 do_non_resident_truncate:
2620 BUG_ON(!NInoNonResident(ni));
2621 if (alloc_change < 0) {
2622 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2623 if (highest_vcn > 0 &&
2624 old_alloc_size >> vol->cluster_size_bits >
2625 highest_vcn + 1) {
2627 * This attribute has multiple extents. Not yet
2628 * supported.
2630 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2631 "attribute type 0x%x, because the "
2632 "attribute is highly fragmented (it "
2633 "consists of multiple extents) and "
2634 "this case is not implemented yet.",
2635 vi->i_ino,
2636 (unsigned)le32_to_cpu(ni->type));
2637 err = -EOPNOTSUPP;
2638 goto bad_out;
2642 * If the size is shrinking, need to reduce the initialized_size and
2643 * the data_size before reducing the allocation.
2645 if (size_change < 0) {
2647 * Make the valid size smaller (i_size is already up-to-date).
2649 write_lock_irqsave(&ni->size_lock, flags);
2650 if (new_size < ni->initialized_size) {
2651 ni->initialized_size = new_size;
2652 a->data.non_resident.initialized_size =
2653 cpu_to_sle64(new_size);
2655 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2656 write_unlock_irqrestore(&ni->size_lock, flags);
2657 flush_dcache_mft_record_page(ctx->ntfs_ino);
2658 mark_mft_record_dirty(ctx->ntfs_ino);
2659 /* If the allocated size is not changing, we are done. */
2660 if (!alloc_change)
2661 goto unm_done;
2663 * If the size is shrinking it makes no sense for the
2664 * allocation to be growing.
2666 BUG_ON(alloc_change > 0);
2667 } else /* if (size_change >= 0) */ {
2669 * The file size is growing or staying the same but the
2670 * allocation can be shrinking, growing or staying the same.
2672 if (alloc_change > 0) {
2674 * We need to extend the allocation and possibly update
2675 * the data size. If we are updating the data size,
2676 * since we are not touching the initialized_size we do
2677 * not need to worry about the actual data on disk.
2678 * And as far as the page cache is concerned, there
2679 * will be no pages beyond the old data size and any
2680 * partial region in the last page between the old and
2681 * new data size (or the end of the page if the new
2682 * data size is outside the page) does not need to be
2683 * modified as explained above for the resident
2684 * attribute truncate case. To do this, we simply drop
2685 * the locks we hold and leave all the work to our
2686 * friendly helper ntfs_attr_extend_allocation().
2688 ntfs_attr_put_search_ctx(ctx);
2689 unmap_mft_record(base_ni);
2690 up_write(&ni->runlist.lock);
2691 err = ntfs_attr_extend_allocation(ni, new_size,
2692 size_change > 0 ? new_size : -1, -1);
2694 * ntfs_attr_extend_allocation() will have done error
2695 * output already.
2697 goto done;
2699 if (!alloc_change)
2700 goto alloc_done;
2702 /* alloc_change < 0 */
2703 /* Free the clusters. */
2704 nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2705 vol->cluster_size_bits, -1, ctx);
2706 m = ctx->mrec;
2707 a = ctx->attr;
2708 if (unlikely(nr_freed < 0)) {
2709 ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2710 "%lli). Unmount and run chkdsk to recover "
2711 "the lost cluster(s).", (long long)nr_freed);
2712 NVolSetErrors(vol);
2713 nr_freed = 0;
2715 /* Truncate the runlist. */
2716 err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2717 new_alloc_size >> vol->cluster_size_bits);
2719 * If the runlist truncation failed and/or the search context is no
2720 * longer valid, we cannot resize the attribute record or build the
2721 * mapping pairs array thus we mark the inode bad so that no access to
2722 * the freed clusters can happen.
2724 if (unlikely(err || IS_ERR(m))) {
2725 ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2726 IS_ERR(m) ?
2727 "restore attribute search context" :
2728 "truncate attribute runlist",
2729 IS_ERR(m) ? PTR_ERR(m) : err, es);
2730 err = -EIO;
2731 goto bad_out;
2733 /* Get the size for the shrunk mapping pairs array for the runlist. */
2734 mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2735 if (unlikely(mp_size <= 0)) {
2736 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2737 "attribute type 0x%x, because determining the "
2738 "size for the mapping pairs failed with error "
2739 "code %i.%s", vi->i_ino,
2740 (unsigned)le32_to_cpu(ni->type), mp_size, es);
2741 err = -EIO;
2742 goto bad_out;
2745 * Shrink the attribute record for the new mapping pairs array. Note,
2746 * this cannot fail since we are making the attribute smaller thus by
2747 * definition there is enough space to do so.
2749 attr_len = le32_to_cpu(a->length);
2750 err = ntfs_attr_record_resize(m, a, mp_size +
2751 le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2752 BUG_ON(err);
2754 * Generate the mapping pairs array directly into the attribute record.
2756 err = ntfs_mapping_pairs_build(vol, (u8*)a +
2757 le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2758 mp_size, ni->runlist.rl, 0, -1, NULL);
2759 if (unlikely(err)) {
2760 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2761 "attribute type 0x%x, because building the "
2762 "mapping pairs failed with error code %i.%s",
2763 vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2764 err, es);
2765 err = -EIO;
2766 goto bad_out;
2768 /* Update the allocated/compressed size as well as the highest vcn. */
2769 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2770 vol->cluster_size_bits) - 1);
2771 write_lock_irqsave(&ni->size_lock, flags);
2772 ni->allocated_size = new_alloc_size;
2773 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2774 if (NInoSparse(ni) || NInoCompressed(ni)) {
2775 if (nr_freed) {
2776 ni->itype.compressed.size -= nr_freed <<
2777 vol->cluster_size_bits;
2778 BUG_ON(ni->itype.compressed.size < 0);
2779 a->data.non_resident.compressed_size = cpu_to_sle64(
2780 ni->itype.compressed.size);
2781 vi->i_blocks = ni->itype.compressed.size >> 9;
2783 } else
2784 vi->i_blocks = new_alloc_size >> 9;
2785 write_unlock_irqrestore(&ni->size_lock, flags);
2787 * We have shrunk the allocation. If this is a shrinking truncate we
2788 * have already dealt with the initialized_size and the data_size above
2789 * and we are done. If the truncate is only changing the allocation
2790 * and not the data_size, we are also done. If this is an extending
2791 * truncate, need to extend the data_size now which is ensured by the
2792 * fact that @size_change is positive.
2794 alloc_done:
2796 * If the size is growing, need to update it now. If it is shrinking,
2797 * we have already updated it above (before the allocation change).
2799 if (size_change > 0)
2800 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2801 /* Ensure the modified mft record is written out. */
2802 flush_dcache_mft_record_page(ctx->ntfs_ino);
2803 mark_mft_record_dirty(ctx->ntfs_ino);
2804 unm_done:
2805 ntfs_attr_put_search_ctx(ctx);
2806 unmap_mft_record(base_ni);
2807 up_write(&ni->runlist.lock);
2808 done:
2809 /* Update the mtime and ctime on the base inode. */
2810 /* normally ->truncate shouldn't update ctime or mtime,
2811 * but ntfs did before so it got a copy & paste version
2812 * of file_update_time. one day someone should fix this
2813 * for real.
2815 if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2816 struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb);
2817 int sync_it = 0;
2819 if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2820 !timespec_equal(&VFS_I(base_ni)->i_ctime, &now))
2821 sync_it = 1;
2822 VFS_I(base_ni)->i_mtime = now;
2823 VFS_I(base_ni)->i_ctime = now;
2825 if (sync_it)
2826 mark_inode_dirty_sync(VFS_I(base_ni));
2829 if (likely(!err)) {
2830 NInoClearTruncateFailed(ni);
2831 ntfs_debug("Done.");
2833 return err;
2834 old_bad_out:
2835 old_size = -1;
2836 bad_out:
2837 if (err != -ENOMEM && err != -EOPNOTSUPP)
2838 NVolSetErrors(vol);
2839 if (err != -EOPNOTSUPP)
2840 NInoSetTruncateFailed(ni);
2841 else if (old_size >= 0)
2842 i_size_write(vi, old_size);
2843 err_out:
2844 if (ctx)
2845 ntfs_attr_put_search_ctx(ctx);
2846 if (m)
2847 unmap_mft_record(base_ni);
2848 up_write(&ni->runlist.lock);
2849 out:
2850 ntfs_debug("Failed. Returning error code %i.", err);
2851 return err;
2852 conv_err_out:
2853 if (err != -ENOMEM && err != -EOPNOTSUPP)
2854 NVolSetErrors(vol);
2855 if (err != -EOPNOTSUPP)
2856 NInoSetTruncateFailed(ni);
2857 else
2858 i_size_write(vi, old_size);
2859 goto out;
2863 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2864 * @vi: inode for which the i_size was changed
2866 * Wrapper for ntfs_truncate() that has no return value.
2868 * See ntfs_truncate() description above for details.
2870 void ntfs_truncate_vfs(struct inode *vi) {
2871 ntfs_truncate(vi);
2875 * ntfs_setattr - called from notify_change() when an attribute is being changed
2876 * @dentry: dentry whose attributes to change
2877 * @attr: structure describing the attributes and the changes
2879 * We have to trap VFS attempts to truncate the file described by @dentry as
2880 * soon as possible, because we do not implement changes in i_size yet. So we
2881 * abort all i_size changes here.
2883 * We also abort all changes of user, group, and mode as we do not implement
2884 * the NTFS ACLs yet.
2886 * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also
2887 * called with ->i_alloc_sem held for writing.
2889 * Basically this is a copy of generic notify_change() and inode_setattr()
2890 * functionality, except we intercept and abort changes in i_size.
2892 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2894 struct inode *vi = dentry->d_inode;
2895 int err;
2896 unsigned int ia_valid = attr->ia_valid;
2898 err = inode_change_ok(vi, attr);
2899 if (err)
2900 goto out;
2901 /* We do not support NTFS ACLs yet. */
2902 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2903 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2904 "supported yet, ignoring.");
2905 err = -EOPNOTSUPP;
2906 goto out;
2908 if (ia_valid & ATTR_SIZE) {
2909 if (attr->ia_size != i_size_read(vi)) {
2910 ntfs_inode *ni = NTFS_I(vi);
2912 * FIXME: For now we do not support resizing of
2913 * compressed or encrypted files yet.
2915 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2916 ntfs_warning(vi->i_sb, "Changes in inode size "
2917 "are not supported yet for "
2918 "%s files, ignoring.",
2919 NInoCompressed(ni) ?
2920 "compressed" : "encrypted");
2921 err = -EOPNOTSUPP;
2922 } else
2923 err = vmtruncate(vi, attr->ia_size);
2924 if (err || ia_valid == ATTR_SIZE)
2925 goto out;
2926 } else {
2928 * We skipped the truncate but must still update
2929 * timestamps.
2931 ia_valid |= ATTR_MTIME | ATTR_CTIME;
2934 if (ia_valid & ATTR_ATIME)
2935 vi->i_atime = timespec_trunc(attr->ia_atime,
2936 vi->i_sb->s_time_gran);
2937 if (ia_valid & ATTR_MTIME)
2938 vi->i_mtime = timespec_trunc(attr->ia_mtime,
2939 vi->i_sb->s_time_gran);
2940 if (ia_valid & ATTR_CTIME)
2941 vi->i_ctime = timespec_trunc(attr->ia_ctime,
2942 vi->i_sb->s_time_gran);
2943 mark_inode_dirty(vi);
2944 out:
2945 return err;
2949 * ntfs_write_inode - write out a dirty inode
2950 * @vi: inode to write out
2951 * @sync: if true, write out synchronously
2953 * Write out a dirty inode to disk including any extent inodes if present.
2955 * If @sync is true, commit the inode to disk and wait for io completion. This
2956 * is done using write_mft_record().
2958 * If @sync is false, just schedule the write to happen but do not wait for i/o
2959 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2960 * marking the page (and in this case mft record) dirty but we do not implement
2961 * this yet as write_mft_record() largely ignores the @sync parameter and
2962 * always performs synchronous writes.
2964 * Return 0 on success and -errno on error.
2966 int ntfs_write_inode(struct inode *vi, int sync)
2968 sle64 nt;
2969 ntfs_inode *ni = NTFS_I(vi);
2970 ntfs_attr_search_ctx *ctx;
2971 MFT_RECORD *m;
2972 STANDARD_INFORMATION *si;
2973 int err = 0;
2974 bool modified = false;
2976 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2977 vi->i_ino);
2979 * Dirty attribute inodes are written via their real inodes so just
2980 * clean them here. Access time updates are taken care off when the
2981 * real inode is written.
2983 if (NInoAttr(ni)) {
2984 NInoClearDirty(ni);
2985 ntfs_debug("Done.");
2986 return 0;
2988 /* Map, pin, and lock the mft record belonging to the inode. */
2989 m = map_mft_record(ni);
2990 if (IS_ERR(m)) {
2991 err = PTR_ERR(m);
2992 goto err_out;
2994 /* Update the access times in the standard information attribute. */
2995 ctx = ntfs_attr_get_search_ctx(ni, m);
2996 if (unlikely(!ctx)) {
2997 err = -ENOMEM;
2998 goto unm_err_out;
3000 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
3001 CASE_SENSITIVE, 0, NULL, 0, ctx);
3002 if (unlikely(err)) {
3003 ntfs_attr_put_search_ctx(ctx);
3004 goto unm_err_out;
3006 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3007 le16_to_cpu(ctx->attr->data.resident.value_offset));
3008 /* Update the access times if they have changed. */
3009 nt = utc2ntfs(vi->i_mtime);
3010 if (si->last_data_change_time != nt) {
3011 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
3012 "new = 0x%llx", vi->i_ino, (long long)
3013 sle64_to_cpu(si->last_data_change_time),
3014 (long long)sle64_to_cpu(nt));
3015 si->last_data_change_time = nt;
3016 modified = true;
3018 nt = utc2ntfs(vi->i_ctime);
3019 if (si->last_mft_change_time != nt) {
3020 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
3021 "new = 0x%llx", vi->i_ino, (long long)
3022 sle64_to_cpu(si->last_mft_change_time),
3023 (long long)sle64_to_cpu(nt));
3024 si->last_mft_change_time = nt;
3025 modified = true;
3027 nt = utc2ntfs(vi->i_atime);
3028 if (si->last_access_time != nt) {
3029 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3030 "new = 0x%llx", vi->i_ino,
3031 (long long)sle64_to_cpu(si->last_access_time),
3032 (long long)sle64_to_cpu(nt));
3033 si->last_access_time = nt;
3034 modified = true;
3037 * If we just modified the standard information attribute we need to
3038 * mark the mft record it is in dirty. We do this manually so that
3039 * mark_inode_dirty() is not called which would redirty the inode and
3040 * hence result in an infinite loop of trying to write the inode.
3041 * There is no need to mark the base inode nor the base mft record
3042 * dirty, since we are going to write this mft record below in any case
3043 * and the base mft record may actually not have been modified so it
3044 * might not need to be written out.
3045 * NOTE: It is not a problem when the inode for $MFT itself is being
3046 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3047 * on the $MFT inode and hence ntfs_write_inode() will not be
3048 * re-invoked because of it which in turn is ok since the dirtied mft
3049 * record will be cleaned and written out to disk below, i.e. before
3050 * this function returns.
3052 if (modified) {
3053 flush_dcache_mft_record_page(ctx->ntfs_ino);
3054 if (!NInoTestSetDirty(ctx->ntfs_ino))
3055 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3056 ctx->ntfs_ino->page_ofs);
3058 ntfs_attr_put_search_ctx(ctx);
3059 /* Now the access times are updated, write the base mft record. */
3060 if (NInoDirty(ni))
3061 err = write_mft_record(ni, m, sync);
3062 /* Write all attached extent mft records. */
3063 mutex_lock(&ni->extent_lock);
3064 if (ni->nr_extents > 0) {
3065 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3066 int i;
3068 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3069 for (i = 0; i < ni->nr_extents; i++) {
3070 ntfs_inode *tni = extent_nis[i];
3072 if (NInoDirty(tni)) {
3073 MFT_RECORD *tm = map_mft_record(tni);
3074 int ret;
3076 if (IS_ERR(tm)) {
3077 if (!err || err == -ENOMEM)
3078 err = PTR_ERR(tm);
3079 continue;
3081 ret = write_mft_record(tni, tm, sync);
3082 unmap_mft_record(tni);
3083 if (unlikely(ret)) {
3084 if (!err || err == -ENOMEM)
3085 err = ret;
3090 mutex_unlock(&ni->extent_lock);
3091 unmap_mft_record(ni);
3092 if (unlikely(err))
3093 goto err_out;
3094 ntfs_debug("Done.");
3095 return 0;
3096 unm_err_out:
3097 unmap_mft_record(ni);
3098 err_out:
3099 if (err == -ENOMEM) {
3100 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
3101 "Marking the inode dirty again, so the VFS "
3102 "retries later.");
3103 mark_inode_dirty(vi);
3104 } else {
3105 ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err);
3106 NVolSetErrors(ni->vol);
3108 return err;
3111 #endif /* NTFS_RW */