Root/fs/ntfs/inode.c

1/**
2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
3 *
4 * Copyright (c) 2001-2007 Anton Altaparmakov
5 *
6 * This program/include file is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as published
8 * by the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * This program/include file is distributed in the hope that it will be
12 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
13 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program (in the main directory of the Linux-NTFS
18 * distribution in the file COPYING); if not, write to the Free Software
19 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 */
21
22#include <linux/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>
30#include <linux/log2.h>
31
32#include "aops.h"
33#include "attrib.h"
34#include "bitmap.h"
35#include "dir.h"
36#include "debug.h"
37#include "inode.h"
38#include "lcnalloc.h"
39#include "malloc.h"
40#include "mft.h"
41#include "time.h"
42#include "ntfs.h"
43
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
48 *
49 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
50 * inode @vi for equality with the ntfs attribute @na.
51 *
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.
54 *
55 * Return 1 if the attributes match and 0 if not.
56 *
57 * NOTE: This function runs with the inode->i_lock spin lock held so it is not
58 * allowed to sleep.
59 */
60int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
61{
62    ntfs_inode *ni;
63
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;
81    }
82    /* Match! */
83    return 1;
84}
85
86/**
87 * ntfs_init_locked_inode - initialize an inode
88 * @vi: vfs inode to initialize
89 * @na: ntfs attribute which to initialize @vi to
90 *
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.
93 *
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.
98 *
99 * Return 0 on success and -errno on error.
100 *
101 * NOTE: This function runs with the inode->i_lock spin lock held so it is not
102 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
103 */
104static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
105{
106    ntfs_inode *ni = NTFS_I(vi);
107
108    vi->i_ino = na->mft_no;
109
110    ni->type = na->type;
111    if (na->type == AT_INDEX_ALLOCATION)
112        NInoSetMstProtected(ni);
113
114    ni->name = na->name;
115    ni->name_len = na->name_len;
116
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;
122    }
123
124    /* It is a fake inode. */
125    NInoSetAttr(ni);
126
127    /*
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.
133     */
134    if (na->name_len && na->name != I30) {
135        unsigned int i;
136
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;
144    }
145    return 0;
146}
147
148typedef int (*set_t)(struct inode *, void *);
149static int ntfs_read_locked_inode(struct inode *vi);
150static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
151static int ntfs_read_locked_index_inode(struct inode *base_vi,
152        struct inode *vi);
153
154/**
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
158 *
159 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
160 * file or directory).
161 *
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.
166 *
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().
169 */
170struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
171{
172    struct inode *vi;
173    int err;
174    ntfs_attr na;
175
176    na.mft_no = mft_no;
177    na.type = AT_UNUSED;
178    na.name = NULL;
179    na.name_len = 0;
180
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);
185
186    err = 0;
187
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);
192    }
193    /*
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.
196     */
197    if (unlikely(err == -ENOMEM)) {
198        iput(vi);
199        vi = ERR_PTR(err);
200    }
201    return vi;
202}
203
204/**
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)
210 *
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.
214 *
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.
219 *
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.
222 *
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().
226 */
227struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
228        ntfschar *name, u32 name_len)
229{
230    struct inode *vi;
231    int err;
232    ntfs_attr na;
233
234    /* Make sure no one calls ntfs_attr_iget() for indices. */
235    BUG_ON(type == AT_INDEX_ALLOCATION);
236
237    na.mft_no = base_vi->i_ino;
238    na.type = type;
239    na.name = name;
240    na.name_len = name_len;
241
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);
246
247    err = 0;
248
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);
253    }
254    /*
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.
258     */
259    if (unlikely(err)) {
260        iput(vi);
261        vi = ERR_PTR(err);
262    }
263    return vi;
264}
265
266/**
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
271 *
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.
275 *
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.
280 *
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().
284 */
285struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
286        u32 name_len)
287{
288    struct inode *vi;
289    int err;
290    ntfs_attr na;
291
292    na.mft_no = base_vi->i_ino;
293    na.type = AT_INDEX_ALLOCATION;
294    na.name = name;
295    na.name_len = name_len;
296
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);
301
302    err = 0;
303
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);
308    }
309    /*
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.
313     */
314    if (unlikely(err)) {
315        iput(vi);
316        vi = ERR_PTR(err);
317    }
318    return vi;
319}
320
321struct inode *ntfs_alloc_big_inode(struct super_block *sb)
322{
323    ntfs_inode *ni;
324
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);
330    }
331    ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
332    return NULL;
333}
334
335static void ntfs_i_callback(struct rcu_head *head)
336{
337    struct inode *inode = container_of(head, struct inode, i_rcu);
338    INIT_LIST_HEAD(&inode->i_dentry);
339    kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
340}
341
342void ntfs_destroy_big_inode(struct inode *inode)
343{
344    ntfs_inode *ni = NTFS_I(inode);
345
346    ntfs_debug("Entering.");
347    BUG_ON(ni->page);
348    if (!atomic_dec_and_test(&ni->count))
349        BUG();
350    call_rcu(&inode->i_rcu, ntfs_i_callback);
351}
352
353static inline ntfs_inode *ntfs_alloc_extent_inode(void)
354{
355    ntfs_inode *ni;
356
357    ntfs_debug("Entering.");
358    ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS);
359    if (likely(ni != NULL)) {
360        ni->state = 0;
361        return ni;
362    }
363    ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
364    return NULL;
365}
366
367static void ntfs_destroy_extent_inode(ntfs_inode *ni)
368{
369    ntfs_debug("Entering.");
370    BUG_ON(ni->page);
371    if (!atomic_dec_and_test(&ni->count))
372        BUG();
373    kmem_cache_free(ntfs_inode_cache, ni);
374}
375
376/*
377 * The attribute runlist lock has separate locking rules from the
378 * normal runlist lock, so split the two lock-classes:
379 */
380static struct lock_class_key attr_list_rl_lock_class;
381
382/**
383 * __ntfs_init_inode - initialize ntfs specific part of an inode
384 * @sb: super block of mounted volume
385 * @ni: freshly allocated ntfs inode which to initialize
386 *
387 * Initialize an ntfs inode to defaults.
388 *
389 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
390 * untouched. Make sure to initialize them elsewhere.
391 *
392 * Return zero on success and -ENOMEM on error.
393 */
394void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
395{
396    ntfs_debug("Entering.");
397    rwlock_init(&ni->size_lock);
398    ni->initialized_size = ni->allocated_size = 0;
399    ni->seq_no = 0;
400    atomic_set(&ni->count, 1);
401    ni->vol = NTFS_SB(sb);
402    ntfs_init_runlist(&ni->runlist);
403    mutex_init(&ni->mrec_lock);
404    ni->page = NULL;
405    ni->page_ofs = 0;
406    ni->attr_list_size = 0;
407    ni->attr_list = NULL;
408    ntfs_init_runlist(&ni->attr_list_rl);
409    lockdep_set_class(&ni->attr_list_rl.lock,
410                &attr_list_rl_lock_class);
411    ni->itype.index.block_size = 0;
412    ni->itype.index.vcn_size = 0;
413    ni->itype.index.collation_rule = 0;
414    ni->itype.index.block_size_bits = 0;
415    ni->itype.index.vcn_size_bits = 0;
416    mutex_init(&ni->extent_lock);
417    ni->nr_extents = 0;
418    ni->ext.base_ntfs_ino = NULL;
419}
420
421/*
422 * Extent inodes get MFT-mapped in a nested way, while the base inode
423 * is still mapped. Teach this nesting to the lock validator by creating
424 * a separate class for nested inode's mrec_lock's:
425 */
426static struct lock_class_key extent_inode_mrec_lock_key;
427
428inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
429        unsigned long mft_no)
430{
431    ntfs_inode *ni = ntfs_alloc_extent_inode();
432
433    ntfs_debug("Entering.");
434    if (likely(ni != NULL)) {
435        __ntfs_init_inode(sb, ni);
436        lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
437        ni->mft_no = mft_no;
438        ni->type = AT_UNUSED;
439        ni->name = NULL;
440        ni->name_len = 0;
441    }
442    return ni;
443}
444
445/**
446 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
447 * @ctx: initialized attribute search context
448 *
449 * Search all file name attributes in the inode described by the attribute
450 * search context @ctx and check if any of the names are in the $Extend system
451 * directory.
452 *
453 * Return values:
454 * 1: file is in $Extend directory
455 * 0: file is not in $Extend directory
456 * -errno: failed to determine if the file is in the $Extend directory
457 */
458static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
459{
460    int nr_links, err;
461
462    /* Restart search. */
463    ntfs_attr_reinit_search_ctx(ctx);
464
465    /* Get number of hard links. */
466    nr_links = le16_to_cpu(ctx->mrec->link_count);
467
468    /* Loop through all hard links. */
469    while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
470            ctx))) {
471        FILE_NAME_ATTR *file_name_attr;
472        ATTR_RECORD *attr = ctx->attr;
473        u8 *p, *p2;
474
475        nr_links--;
476        /*
477         * Maximum sanity checking as we are called on an inode that
478         * we suspect might be corrupt.
479         */
480        p = (u8*)attr + le32_to_cpu(attr->length);
481        if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
482                le32_to_cpu(ctx->mrec->bytes_in_use)) {
483err_corrupt_attr:
484            ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
485                    "attribute. You should run chkdsk.");
486            return -EIO;
487        }
488        if (attr->non_resident) {
489            ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
490                    "name. You should run chkdsk.");
491            return -EIO;
492        }
493        if (attr->flags) {
494            ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
495                    "invalid flags. You should run "
496                    "chkdsk.");
497            return -EIO;
498        }
499        if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
500            ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
501                    "name. You should run chkdsk.");
502            return -EIO;
503        }
504        file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
505                le16_to_cpu(attr->data.resident.value_offset));
506        p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
507        if (p2 < (u8*)attr || p2 > p)
508            goto err_corrupt_attr;
509        /* This attribute is ok, but is it in the $Extend directory? */
510        if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
511            return 1; /* YES, it's an extended system file. */
512    }
513    if (unlikely(err != -ENOENT))
514        return err;
515    if (unlikely(nr_links)) {
516        ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
517                "doesn't match number of name attributes. You "
518                "should run chkdsk.");
519        return -EIO;
520    }
521    return 0; /* NO, it is not an extended system file. */
522}
523
524/**
525 * ntfs_read_locked_inode - read an inode from its device
526 * @vi: inode to read
527 *
528 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
529 * described by @vi into memory from the device.
530 *
531 * The only fields in @vi that we need to/can look at when the function is
532 * called are i_sb, pointing to the mounted device's super block, and i_ino,
533 * the number of the inode to load.
534 *
535 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
536 * for reading and sets up the necessary @vi fields as well as initializing
537 * the ntfs inode.
538 *
539 * Q: What locks are held when the function is called?
540 * A: i_state has I_NEW set, hence the inode is locked, also
541 * i_count is set to 1, so it is not going to go away
542 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
543 * is allowed to write to them. We should of course be honouring them but
544 * we need to do that using the IS_* macros defined in include/linux/fs.h.
545 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
546 *
547 * Return 0 on success and -errno on error. In the error case, the inode will
548 * have had make_bad_inode() executed on it.
549 */
550static int ntfs_read_locked_inode(struct inode *vi)
551{
552    ntfs_volume *vol = NTFS_SB(vi->i_sb);
553    ntfs_inode *ni;
554    struct inode *bvi;
555    MFT_RECORD *m;
556    ATTR_RECORD *a;
557    STANDARD_INFORMATION *si;
558    ntfs_attr_search_ctx *ctx;
559    int err = 0;
560
561    ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
562
563    /* Setup the generic vfs inode parts now. */
564
565    /*
566     * This is for checking whether an inode has changed w.r.t. a file so
567     * that the file can be updated if necessary (compare with f_version).
568     */
569    vi->i_version = 1;
570
571    vi->i_uid = vol->uid;
572    vi->i_gid = vol->gid;
573    vi->i_mode = 0;
574
575    /*
576     * Initialize the ntfs specific part of @vi special casing
577     * FILE_MFT which we need to do at mount time.
578     */
579    if (vi->i_ino != FILE_MFT)
580        ntfs_init_big_inode(vi);
581    ni = NTFS_I(vi);
582
583    m = map_mft_record(ni);
584    if (IS_ERR(m)) {
585        err = PTR_ERR(m);
586        goto err_out;
587    }
588    ctx = ntfs_attr_get_search_ctx(ni, m);
589    if (!ctx) {
590        err = -ENOMEM;
591        goto unm_err_out;
592    }
593
594    if (!(m->flags & MFT_RECORD_IN_USE)) {
595        ntfs_error(vi->i_sb, "Inode is not in use!");
596        goto unm_err_out;
597    }
598    if (m->base_mft_record) {
599        ntfs_error(vi->i_sb, "Inode is an extent inode!");
600        goto unm_err_out;
601    }
602
603    /* Transfer information from mft record into vfs and ntfs inodes. */
604    vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
605
606    /*
607     * FIXME: Keep in mind that link_count is two for files which have both
608     * a long file name and a short file name as separate entries, so if
609     * we are hiding short file names this will be too high. Either we need
610     * to account for the short file names by subtracting them or we need
611     * to make sure we delete files even though i_nlink is not zero which
612     * might be tricky due to vfs interactions. Need to think about this
613     * some more when implementing the unlink command.
614     */
615    vi->i_nlink = le16_to_cpu(m->link_count);
616    /*
617     * FIXME: Reparse points can have the directory bit set even though
618     * they would be S_IFLNK. Need to deal with this further below when we
619     * implement reparse points / symbolic links but it will do for now.
620     * Also if not a directory, it could be something else, rather than
621     * a regular file. But again, will do for now.
622     */
623    /* Everyone gets all permissions. */
624    vi->i_mode |= S_IRWXUGO;
625    /* If read-only, no one gets write permissions. */
626    if (IS_RDONLY(vi))
627        vi->i_mode &= ~S_IWUGO;
628    if (m->flags & MFT_RECORD_IS_DIRECTORY) {
629        vi->i_mode |= S_IFDIR;
630        /*
631         * Apply the directory permissions mask set in the mount
632         * options.
633         */
634        vi->i_mode &= ~vol->dmask;
635        /* Things break without this kludge! */
636        if (vi->i_nlink > 1)
637            vi->i_nlink = 1;
638    } else {
639        vi->i_mode |= S_IFREG;
640        /* Apply the file permissions mask set in the mount options. */
641        vi->i_mode &= ~vol->fmask;
642    }
643    /*
644     * Find the standard information attribute in the mft record. At this
645     * stage we haven't setup the attribute list stuff yet, so this could
646     * in fact fail if the standard information is in an extent record, but
647     * I don't think this actually ever happens.
648     */
649    err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
650            ctx);
651    if (unlikely(err)) {
652        if (err == -ENOENT) {
653            /*
654             * TODO: We should be performing a hot fix here (if the
655             * recover mount option is set) by creating a new
656             * attribute.
657             */
658            ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
659                    "is missing.");
660        }
661        goto unm_err_out;
662    }
663    a = ctx->attr;
664    /* Get the standard information attribute value. */
665    si = (STANDARD_INFORMATION*)((u8*)a +
666            le16_to_cpu(a->data.resident.value_offset));
667
668    /* Transfer information from the standard information into vi. */
669    /*
670     * Note: The i_?times do not quite map perfectly onto the NTFS times,
671     * but they are close enough, and in the end it doesn't really matter
672     * that much...
673     */
674    /*
675     * mtime is the last change of the data within the file. Not changed
676     * when only metadata is changed, e.g. a rename doesn't affect mtime.
677     */
678    vi->i_mtime = ntfs2utc(si->last_data_change_time);
679    /*
680     * ctime is the last change of the metadata of the file. This obviously
681     * always changes, when mtime is changed. ctime can be changed on its
682     * own, mtime is then not changed, e.g. when a file is renamed.
683     */
684    vi->i_ctime = ntfs2utc(si->last_mft_change_time);
685    /*
686     * Last access to the data within the file. Not changed during a rename
687     * for example but changed whenever the file is written to.
688     */
689    vi->i_atime = ntfs2utc(si->last_access_time);
690
691    /* Find the attribute list attribute if present. */
692    ntfs_attr_reinit_search_ctx(ctx);
693    err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
694    if (err) {
695        if (unlikely(err != -ENOENT)) {
696            ntfs_error(vi->i_sb, "Failed to lookup attribute list "
697                    "attribute.");
698            goto unm_err_out;
699        }
700    } else /* if (!err) */ {
701        if (vi->i_ino == FILE_MFT)
702            goto skip_attr_list_load;
703        ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
704        NInoSetAttrList(ni);
705        a = ctx->attr;
706        if (a->flags & ATTR_COMPRESSION_MASK) {
707            ntfs_error(vi->i_sb, "Attribute list attribute is "
708                    "compressed.");
709            goto unm_err_out;
710        }
711        if (a->flags & ATTR_IS_ENCRYPTED ||
712                a->flags & ATTR_IS_SPARSE) {
713            if (a->non_resident) {
714                ntfs_error(vi->i_sb, "Non-resident attribute "
715                        "list attribute is encrypted/"
716                        "sparse.");
717                goto unm_err_out;
718            }
719            ntfs_warning(vi->i_sb, "Resident attribute list "
720                    "attribute in inode 0x%lx is marked "
721                    "encrypted/sparse which is not true. "
722                    "However, Windows allows this and "
723                    "chkdsk does not detect or correct it "
724                    "so we will just ignore the invalid "
725                    "flags and pretend they are not set.",
726                    vi->i_ino);
727        }
728        /* Now allocate memory for the attribute list. */
729        ni->attr_list_size = (u32)ntfs_attr_size(a);
730        ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
731        if (!ni->attr_list) {
732            ntfs_error(vi->i_sb, "Not enough memory to allocate "
733                    "buffer for attribute list.");
734            err = -ENOMEM;
735            goto unm_err_out;
736        }
737        if (a->non_resident) {
738            NInoSetAttrListNonResident(ni);
739            if (a->data.non_resident.lowest_vcn) {
740                ntfs_error(vi->i_sb, "Attribute list has non "
741                        "zero lowest_vcn.");
742                goto unm_err_out;
743            }
744            /*
745             * Setup the runlist. No need for locking as we have
746             * exclusive access to the inode at this time.
747             */
748            ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
749                    a, NULL);
750            if (IS_ERR(ni->attr_list_rl.rl)) {
751                err = PTR_ERR(ni->attr_list_rl.rl);
752                ni->attr_list_rl.rl = NULL;
753                ntfs_error(vi->i_sb, "Mapping pairs "
754                        "decompression failed.");
755                goto unm_err_out;
756            }
757            /* Now load the attribute list. */
758            if ((err = load_attribute_list(vol, &ni->attr_list_rl,
759                    ni->attr_list, ni->attr_list_size,
760                    sle64_to_cpu(a->data.non_resident.
761                    initialized_size)))) {
762                ntfs_error(vi->i_sb, "Failed to load "
763                        "attribute list attribute.");
764                goto unm_err_out;
765            }
766        } else /* if (!a->non_resident) */ {
767            if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
768                    + le32_to_cpu(
769                    a->data.resident.value_length) >
770                    (u8*)ctx->mrec + vol->mft_record_size) {
771                ntfs_error(vi->i_sb, "Corrupt attribute list "
772                        "in inode.");
773                goto unm_err_out;
774            }
775            /* Now copy the attribute list. */
776            memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
777                    a->data.resident.value_offset),
778                    le32_to_cpu(
779                    a->data.resident.value_length));
780        }
781    }
782skip_attr_list_load:
783    /*
784     * If an attribute list is present we now have the attribute list value
785     * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
786     */
787    if (S_ISDIR(vi->i_mode)) {
788        loff_t bvi_size;
789        ntfs_inode *bni;
790        INDEX_ROOT *ir;
791        u8 *ir_end, *index_end;
792
793        /* It is a directory, find index root attribute. */
794        ntfs_attr_reinit_search_ctx(ctx);
795        err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
796                0, NULL, 0, ctx);
797        if (unlikely(err)) {
798            if (err == -ENOENT) {
799                // FIXME: File is corrupt! Hot-fix with empty
800                // index root attribute if recovery option is
801                // set.
802                ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
803                        "is missing.");
804            }
805            goto unm_err_out;
806        }
807        a = ctx->attr;
808        /* Set up the state. */
809        if (unlikely(a->non_resident)) {
810            ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
811                    "resident.");
812            goto unm_err_out;
813        }
814        /* Ensure the attribute name is placed before the value. */
815        if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
816                le16_to_cpu(a->data.resident.value_offset)))) {
817            ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
818                    "placed after the attribute value.");
819            goto unm_err_out;
820        }
821        /*
822         * Compressed/encrypted index root just means that the newly
823         * created files in that directory should be created compressed/
824         * encrypted. However index root cannot be both compressed and
825         * encrypted.
826         */
827        if (a->flags & ATTR_COMPRESSION_MASK)
828            NInoSetCompressed(ni);
829        if (a->flags & ATTR_IS_ENCRYPTED) {
830            if (a->flags & ATTR_COMPRESSION_MASK) {
831                ntfs_error(vi->i_sb, "Found encrypted and "
832                        "compressed attribute.");
833                goto unm_err_out;
834            }
835            NInoSetEncrypted(ni);
836        }
837        if (a->flags & ATTR_IS_SPARSE)
838            NInoSetSparse(ni);
839        ir = (INDEX_ROOT*)((u8*)a +
840                le16_to_cpu(a->data.resident.value_offset));
841        ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
842        if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
843            ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
844                    "corrupt.");
845            goto unm_err_out;
846        }
847        index_end = (u8*)&ir->index +
848                le32_to_cpu(ir->index.index_length);
849        if (index_end > ir_end) {
850            ntfs_error(vi->i_sb, "Directory index is corrupt.");
851            goto unm_err_out;
852        }
853        if (ir->type != AT_FILE_NAME) {
854            ntfs_error(vi->i_sb, "Indexed attribute is not "
855                    "$FILE_NAME.");
856            goto unm_err_out;
857        }
858        if (ir->collation_rule != COLLATION_FILE_NAME) {
859            ntfs_error(vi->i_sb, "Index collation rule is not "
860                    "COLLATION_FILE_NAME.");
861            goto unm_err_out;
862        }
863        ni->itype.index.collation_rule = ir->collation_rule;
864        ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
865        if (ni->itype.index.block_size &
866                (ni->itype.index.block_size - 1)) {
867            ntfs_error(vi->i_sb, "Index block size (%u) is not a "
868                    "power of two.",
869                    ni->itype.index.block_size);
870            goto unm_err_out;
871        }
872        if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
873            ntfs_error(vi->i_sb, "Index block size (%u) > "
874                    "PAGE_CACHE_SIZE (%ld) is not "
875                    "supported. Sorry.",
876                    ni->itype.index.block_size,
877                    PAGE_CACHE_SIZE);
878            err = -EOPNOTSUPP;
879            goto unm_err_out;
880        }
881        if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
882            ntfs_error(vi->i_sb, "Index block size (%u) < "
883                    "NTFS_BLOCK_SIZE (%i) is not "
884                    "supported. Sorry.",
885                    ni->itype.index.block_size,
886                    NTFS_BLOCK_SIZE);
887            err = -EOPNOTSUPP;
888            goto unm_err_out;
889        }
890        ni->itype.index.block_size_bits =
891                ffs(ni->itype.index.block_size) - 1;
892        /* Determine the size of a vcn in the directory index. */
893        if (vol->cluster_size <= ni->itype.index.block_size) {
894            ni->itype.index.vcn_size = vol->cluster_size;
895            ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
896        } else {
897            ni->itype.index.vcn_size = vol->sector_size;
898            ni->itype.index.vcn_size_bits = vol->sector_size_bits;
899        }
900
901        /* Setup the index allocation attribute, even if not present. */
902        NInoSetMstProtected(ni);
903        ni->type = AT_INDEX_ALLOCATION;
904        ni->name = I30;
905        ni->name_len = 4;
906
907        if (!(ir->index.flags & LARGE_INDEX)) {
908            /* No index allocation. */
909            vi->i_size = ni->initialized_size =
910                    ni->allocated_size = 0;
911            /* We are done with the mft record, so we release it. */
912            ntfs_attr_put_search_ctx(ctx);
913            unmap_mft_record(ni);
914            m = NULL;
915            ctx = NULL;
916            goto skip_large_dir_stuff;
917        } /* LARGE_INDEX: Index allocation present. Setup state. */
918        NInoSetIndexAllocPresent(ni);
919        /* Find index allocation attribute. */
920        ntfs_attr_reinit_search_ctx(ctx);
921        err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
922                CASE_SENSITIVE, 0, NULL, 0, ctx);
923        if (unlikely(err)) {
924            if (err == -ENOENT)
925                ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
926                        "attribute is not present but "
927                        "$INDEX_ROOT indicated it is.");
928            else
929                ntfs_error(vi->i_sb, "Failed to lookup "
930                        "$INDEX_ALLOCATION "
931                        "attribute.");
932            goto unm_err_out;
933        }
934        a = ctx->attr;
935        if (!a->non_resident) {
936            ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
937                    "is resident.");
938            goto unm_err_out;
939        }
940        /*
941         * Ensure the attribute name is placed before the mapping pairs
942         * array.
943         */
944        if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
945                le16_to_cpu(
946                a->data.non_resident.mapping_pairs_offset)))) {
947            ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
948                    "is placed after the mapping pairs "
949                    "array.");
950            goto unm_err_out;
951        }
952        if (a->flags & ATTR_IS_ENCRYPTED) {
953            ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
954                    "is encrypted.");
955            goto unm_err_out;
956        }
957        if (a->flags & ATTR_IS_SPARSE) {
958            ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
959                    "is sparse.");
960            goto unm_err_out;
961        }
962        if (a->flags & ATTR_COMPRESSION_MASK) {
963            ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
964                    "is compressed.");
965            goto unm_err_out;
966        }
967        if (a->data.non_resident.lowest_vcn) {
968            ntfs_error(vi->i_sb, "First extent of "
969                    "$INDEX_ALLOCATION attribute has non "
970                    "zero lowest_vcn.");
971            goto unm_err_out;
972        }
973        vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
974        ni->initialized_size = sle64_to_cpu(
975                a->data.non_resident.initialized_size);
976        ni->allocated_size = sle64_to_cpu(
977                a->data.non_resident.allocated_size);
978        /*
979         * We are done with the mft record, so we release it. Otherwise
980         * we would deadlock in ntfs_attr_iget().
981         */
982        ntfs_attr_put_search_ctx(ctx);
983        unmap_mft_record(ni);
984        m = NULL;
985        ctx = NULL;
986        /* Get the index bitmap attribute inode. */
987        bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
988        if (IS_ERR(bvi)) {
989            ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
990            err = PTR_ERR(bvi);
991            goto unm_err_out;
992        }
993        bni = NTFS_I(bvi);
994        if (NInoCompressed(bni) || NInoEncrypted(bni) ||
995                NInoSparse(bni)) {
996            ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
997                    "and/or encrypted and/or sparse.");
998            goto iput_unm_err_out;
999        }
1000        /* Consistency check bitmap size vs. index allocation size. */
1001        bvi_size = i_size_read(bvi);
1002        if ((bvi_size << 3) < (vi->i_size >>
1003                ni->itype.index.block_size_bits)) {
1004            ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
1005                    "for index allocation (0x%llx).",
1006                    bvi_size << 3, vi->i_size);
1007            goto iput_unm_err_out;
1008        }
1009        /* No longer need the bitmap attribute inode. */
1010        iput(bvi);
1011skip_large_dir_stuff:
1012        /* Setup the operations for this inode. */
1013        vi->i_op = &ntfs_dir_inode_ops;
1014        vi->i_fop = &ntfs_dir_ops;
1015    } else {
1016        /* It is a file. */
1017        ntfs_attr_reinit_search_ctx(ctx);
1018
1019        /* Setup the data attribute, even if not present. */
1020        ni->type = AT_DATA;
1021        ni->name = NULL;
1022        ni->name_len = 0;
1023
1024        /* Find first extent of the unnamed data attribute. */
1025        err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1026        if (unlikely(err)) {
1027            vi->i_size = ni->initialized_size =
1028                    ni->allocated_size = 0;
1029            if (err != -ENOENT) {
1030                ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1031                        "attribute.");
1032                goto unm_err_out;
1033            }
1034            /*
1035             * FILE_Secure does not have an unnamed $DATA
1036             * attribute, so we special case it here.
1037             */
1038            if (vi->i_ino == FILE_Secure)
1039                goto no_data_attr_special_case;
1040            /*
1041             * Most if not all the system files in the $Extend
1042             * system directory do not have unnamed data
1043             * attributes so we need to check if the parent
1044             * directory of the file is FILE_Extend and if it is
1045             * ignore this error. To do this we need to get the
1046             * name of this inode from the mft record as the name
1047             * contains the back reference to the parent directory.
1048             */
1049            if (ntfs_is_extended_system_file(ctx) > 0)
1050                goto no_data_attr_special_case;
1051            // FIXME: File is corrupt! Hot-fix with empty data
1052            // attribute if recovery option is set.
1053            ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1054            goto unm_err_out;
1055        }
1056        a = ctx->attr;
1057        /* Setup the state. */
1058        if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1059            if (a->flags & ATTR_COMPRESSION_MASK) {
1060                NInoSetCompressed(ni);
1061                if (vol->cluster_size > 4096) {
1062                    ntfs_error(vi->i_sb, "Found "
1063                            "compressed data but "
1064                            "compression is "
1065                            "disabled due to "
1066                            "cluster size (%i) > "
1067                            "4kiB.",
1068                            vol->cluster_size);
1069                    goto unm_err_out;
1070                }
1071                if ((a->flags & ATTR_COMPRESSION_MASK)
1072                        != ATTR_IS_COMPRESSED) {
1073                    ntfs_error(vi->i_sb, "Found unknown "
1074                            "compression method "
1075                            "or corrupt file.");
1076                    goto unm_err_out;
1077                }
1078            }
1079            if (a->flags & ATTR_IS_SPARSE)
1080                NInoSetSparse(ni);
1081        }
1082        if (a->flags & ATTR_IS_ENCRYPTED) {
1083            if (NInoCompressed(ni)) {
1084                ntfs_error(vi->i_sb, "Found encrypted and "
1085                        "compressed data.");
1086                goto unm_err_out;
1087            }
1088            NInoSetEncrypted(ni);
1089        }
1090        if (a->non_resident) {
1091            NInoSetNonResident(ni);
1092            if (NInoCompressed(ni) || NInoSparse(ni)) {
1093                if (NInoCompressed(ni) && a->data.non_resident.
1094                        compression_unit != 4) {
1095                    ntfs_error(vi->i_sb, "Found "
1096                            "non-standard "
1097                            "compression unit (%u "
1098                            "instead of 4). "
1099                            "Cannot handle this.",
1100                            a->data.non_resident.
1101                            compression_unit);
1102                    err = -EOPNOTSUPP;
1103                    goto unm_err_out;
1104                }
1105                if (a->data.non_resident.compression_unit) {
1106                    ni->itype.compressed.block_size = 1U <<
1107                            (a->data.non_resident.
1108                            compression_unit +
1109                            vol->cluster_size_bits);
1110                    ni->itype.compressed.block_size_bits =
1111                            ffs(ni->itype.
1112                            compressed.
1113                            block_size) - 1;
1114                    ni->itype.compressed.block_clusters =
1115                            1U << a->data.
1116                            non_resident.
1117                            compression_unit;
1118                } else {
1119                    ni->itype.compressed.block_size = 0;
1120                    ni->itype.compressed.block_size_bits =
1121                            0;
1122                    ni->itype.compressed.block_clusters =
1123                            0;
1124                }
1125                ni->itype.compressed.size = sle64_to_cpu(
1126                        a->data.non_resident.
1127                        compressed_size);
1128            }
1129            if (a->data.non_resident.lowest_vcn) {
1130                ntfs_error(vi->i_sb, "First extent of $DATA "
1131                        "attribute has non zero "
1132                        "lowest_vcn.");
1133                goto unm_err_out;
1134            }
1135            vi->i_size = sle64_to_cpu(
1136                    a->data.non_resident.data_size);
1137            ni->initialized_size = sle64_to_cpu(
1138                    a->data.non_resident.initialized_size);
1139            ni->allocated_size = sle64_to_cpu(
1140                    a->data.non_resident.allocated_size);
1141        } else { /* Resident attribute. */
1142            vi->i_size = ni->initialized_size = le32_to_cpu(
1143                    a->data.resident.value_length);
1144            ni->allocated_size = le32_to_cpu(a->length) -
1145                    le16_to_cpu(
1146                    a->data.resident.value_offset);
1147            if (vi->i_size > ni->allocated_size) {
1148                ntfs_error(vi->i_sb, "Resident data attribute "
1149                        "is corrupt (size exceeds "
1150                        "allocation).");
1151                goto unm_err_out;
1152            }
1153        }
1154no_data_attr_special_case:
1155        /* We are done with the mft record, so we release it. */
1156        ntfs_attr_put_search_ctx(ctx);
1157        unmap_mft_record(ni);
1158        m = NULL;
1159        ctx = NULL;
1160        /* Setup the operations for this inode. */
1161        vi->i_op = &ntfs_file_inode_ops;
1162        vi->i_fop = &ntfs_file_ops;
1163    }
1164    if (NInoMstProtected(ni))
1165        vi->i_mapping->a_ops = &ntfs_mst_aops;
1166    else
1167        vi->i_mapping->a_ops = &ntfs_aops;
1168    /*
1169     * The number of 512-byte blocks used on disk (for stat). This is in so
1170     * far inaccurate as it doesn't account for any named streams or other
1171     * special non-resident attributes, but that is how Windows works, too,
1172     * so we are at least consistent with Windows, if not entirely
1173     * consistent with the Linux Way. Doing it the Linux Way would cause a
1174     * significant slowdown as it would involve iterating over all
1175     * attributes in the mft record and adding the allocated/compressed
1176     * sizes of all non-resident attributes present to give us the Linux
1177     * correct size that should go into i_blocks (after division by 512).
1178     */
1179    if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1180        vi->i_blocks = ni->itype.compressed.size >> 9;
1181    else
1182        vi->i_blocks = ni->allocated_size >> 9;
1183    ntfs_debug("Done.");
1184    return 0;
1185iput_unm_err_out:
1186    iput(bvi);
1187unm_err_out:
1188    if (!err)
1189        err = -EIO;
1190    if (ctx)
1191        ntfs_attr_put_search_ctx(ctx);
1192    if (m)
1193        unmap_mft_record(ni);
1194err_out:
1195    ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1196            "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1197    make_bad_inode(vi);
1198    if (err != -EOPNOTSUPP && err != -ENOMEM)
1199        NVolSetErrors(vol);
1200    return err;
1201}
1202
1203/**
1204 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1205 * @base_vi: base inode
1206 * @vi: attribute inode to read
1207 *
1208 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1209 * attribute inode described by @vi into memory from the base mft record
1210 * described by @base_ni.
1211 *
1212 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1213 * reading and looks up the attribute described by @vi before setting up the
1214 * necessary fields in @vi as well as initializing the ntfs inode.
1215 *
1216 * Q: What locks are held when the function is called?
1217 * A: i_state has I_NEW set, hence the inode is locked, also
1218 * i_count is set to 1, so it is not going to go away
1219 *
1220 * Return 0 on success and -errno on error. In the error case, the inode will
1221 * have had make_bad_inode() executed on it.
1222 *
1223 * Note this cannot be called for AT_INDEX_ALLOCATION.
1224 */
1225static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1226{
1227    ntfs_volume *vol = NTFS_SB(vi->i_sb);
1228    ntfs_inode *ni, *base_ni;
1229    MFT_RECORD *m;
1230    ATTR_RECORD *a;
1231    ntfs_attr_search_ctx *ctx;
1232    int err = 0;
1233
1234    ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1235
1236    ntfs_init_big_inode(vi);
1237
1238    ni = NTFS_I(vi);
1239    base_ni = NTFS_I(base_vi);
1240
1241    /* Just mirror the values from the base inode. */
1242    vi->i_version = base_vi->i_version;
1243    vi->i_uid = base_vi->i_uid;
1244    vi->i_gid = base_vi->i_gid;
1245    vi->i_nlink = base_vi->i_nlink;
1246    vi->i_mtime = base_vi->i_mtime;
1247    vi->i_ctime = base_vi->i_ctime;
1248    vi->i_atime = base_vi->i_atime;
1249    vi->i_generation = ni->seq_no = base_ni->seq_no;
1250
1251    /* Set inode type to zero but preserve permissions. */
1252    vi->i_mode = base_vi->i_mode & ~S_IFMT;
1253
1254    m = map_mft_record(base_ni);
1255    if (IS_ERR(m)) {
1256        err = PTR_ERR(m);
1257        goto err_out;
1258    }
1259    ctx = ntfs_attr_get_search_ctx(base_ni, m);
1260    if (!ctx) {
1261        err = -ENOMEM;
1262        goto unm_err_out;
1263    }
1264    /* Find the attribute. */
1265    err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1266            CASE_SENSITIVE, 0, NULL, 0, ctx);
1267    if (unlikely(err))
1268        goto unm_err_out;
1269    a = ctx->attr;
1270    if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1271        if (a->flags & ATTR_COMPRESSION_MASK) {
1272            NInoSetCompressed(ni);
1273            if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1274                    ni->name_len)) {
1275                ntfs_error(vi->i_sb, "Found compressed "
1276                        "non-data or named data "
1277                        "attribute. Please report "
1278                        "you saw this message to "
1279                        "linux-ntfs-dev@lists."
1280                        "sourceforge.net");
1281                goto unm_err_out;
1282            }
1283            if (vol->cluster_size > 4096) {
1284                ntfs_error(vi->i_sb, "Found compressed "
1285                        "attribute but compression is "
1286                        "disabled due to cluster size "
1287                        "(%i) > 4kiB.",
1288                        vol->cluster_size);
1289                goto unm_err_out;
1290            }
1291            if ((a->flags & ATTR_COMPRESSION_MASK) !=
1292                    ATTR_IS_COMPRESSED) {
1293                ntfs_error(vi->i_sb, "Found unknown "
1294                        "compression method.");
1295                goto unm_err_out;
1296            }
1297        }
1298        /*
1299         * The compressed/sparse flag set in an index root just means
1300         * to compress all files.
1301         */
1302        if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1303            ntfs_error(vi->i_sb, "Found mst protected attribute "
1304                    "but the attribute is %s. Please "
1305                    "report you saw this message to "
1306                    "linux-ntfs-dev@lists.sourceforge.net",
1307                    NInoCompressed(ni) ? "compressed" :
1308                    "sparse");
1309            goto unm_err_out;
1310        }
1311        if (a->flags & ATTR_IS_SPARSE)
1312            NInoSetSparse(ni);
1313    }
1314    if (a->flags & ATTR_IS_ENCRYPTED) {
1315        if (NInoCompressed(ni)) {
1316            ntfs_error(vi->i_sb, "Found encrypted and compressed "
1317                    "data.");
1318            goto unm_err_out;
1319        }
1320        /*
1321         * The encryption flag set in an index root just means to
1322         * encrypt all files.
1323         */
1324        if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1325            ntfs_error(vi->i_sb, "Found mst protected attribute "
1326                    "but the attribute is encrypted. "
1327                    "Please report you saw this message "
1328                    "to linux-ntfs-dev@lists.sourceforge."
1329                    "net");
1330            goto unm_err_out;
1331        }
1332        if (ni->type != AT_DATA) {
1333            ntfs_error(vi->i_sb, "Found encrypted non-data "
1334                    "attribute.");
1335            goto unm_err_out;
1336        }
1337        NInoSetEncrypted(ni);
1338    }
1339    if (!a->non_resident) {
1340        /* Ensure the attribute name is placed before the value. */
1341        if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1342                le16_to_cpu(a->data.resident.value_offset)))) {
1343            ntfs_error(vol->sb, "Attribute name is placed after "
1344                    "the attribute value.");
1345            goto unm_err_out;
1346        }
1347        if (NInoMstProtected(ni)) {
1348            ntfs_error(vi->i_sb, "Found mst protected attribute "
1349                    "but the attribute is resident. "
1350                    "Please report you saw this message to "
1351                    "linux-ntfs-dev@lists.sourceforge.net");
1352            goto unm_err_out;
1353        }
1354        vi->i_size = ni->initialized_size = le32_to_cpu(
1355                a->data.resident.value_length);
1356        ni->allocated_size = le32_to_cpu(a->length) -
1357                le16_to_cpu(a->data.resident.value_offset);
1358        if (vi->i_size > ni->allocated_size) {
1359            ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1360                    "(size exceeds allocation).");
1361            goto unm_err_out;
1362        }
1363    } else {
1364        NInoSetNonResident(ni);
1365        /*
1366         * Ensure the attribute name is placed before the mapping pairs
1367         * array.
1368         */
1369        if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1370                le16_to_cpu(
1371                a->data.non_resident.mapping_pairs_offset)))) {
1372            ntfs_error(vol->sb, "Attribute name is placed after "
1373                    "the mapping pairs array.");
1374            goto unm_err_out;
1375        }
1376        if (NInoCompressed(ni) || NInoSparse(ni)) {
1377            if (NInoCompressed(ni) && a->data.non_resident.
1378                    compression_unit != 4) {
1379                ntfs_error(vi->i_sb, "Found non-standard "
1380                        "compression unit (%u instead "
1381                        "of 4). Cannot handle this.",
1382                        a->data.non_resident.
1383                        compression_unit);
1384                err = -EOPNOTSUPP;
1385                goto unm_err_out;
1386            }
1387            if (a->data.non_resident.compression_unit) {
1388                ni->itype.compressed.block_size = 1U <<
1389                        (a->data.non_resident.
1390                        compression_unit +
1391                        vol->cluster_size_bits);
1392                ni->itype.compressed.block_size_bits =
1393                        ffs(ni->itype.compressed.
1394                        block_size) - 1;
1395                ni->itype.compressed.block_clusters = 1U <<
1396                        a->data.non_resident.
1397                        compression_unit;
1398            } else {
1399                ni->itype.compressed.block_size = 0;
1400                ni->itype.compressed.block_size_bits = 0;
1401                ni->itype.compressed.block_clusters = 0;
1402            }
1403            ni->itype.compressed.size = sle64_to_cpu(
1404                    a->data.non_resident.compressed_size);
1405        }
1406        if (a->data.non_resident.lowest_vcn) {
1407            ntfs_error(vi->i_sb, "First extent of attribute has "
1408                    "non-zero lowest_vcn.");
1409            goto unm_err_out;
1410        }
1411        vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1412        ni->initialized_size = sle64_to_cpu(
1413                a->data.non_resident.initialized_size);
1414        ni->allocated_size = sle64_to_cpu(
1415                a->data.non_resident.allocated_size);
1416    }
1417    if (NInoMstProtected(ni))
1418        vi->i_mapping->a_ops = &ntfs_mst_aops;
1419    else
1420        vi->i_mapping->a_ops = &ntfs_aops;
1421    if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1422        vi->i_blocks = ni->itype.compressed.size >> 9;
1423    else
1424        vi->i_blocks = ni->allocated_size >> 9;
1425    /*
1426     * Make sure the base inode does not go away and attach it to the
1427     * attribute inode.
1428     */
1429    igrab(base_vi);
1430    ni->ext.base_ntfs_ino = base_ni;
1431    ni->nr_extents = -1;
1432
1433    ntfs_attr_put_search_ctx(ctx);
1434    unmap_mft_record(base_ni);
1435
1436    ntfs_debug("Done.");
1437    return 0;
1438
1439unm_err_out:
1440    if (!err)
1441        err = -EIO;
1442    if (ctx)
1443        ntfs_attr_put_search_ctx(ctx);
1444    unmap_mft_record(base_ni);
1445err_out:
1446    ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1447            "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1448            "Marking corrupt inode and base inode 0x%lx as bad. "
1449            "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1450            base_vi->i_ino);
1451    make_bad_inode(vi);
1452    if (err != -ENOMEM)
1453        NVolSetErrors(vol);
1454    return err;
1455}
1456
1457/**
1458 * ntfs_read_locked_index_inode - read an index inode from its base inode
1459 * @base_vi: base inode
1460 * @vi: index inode to read
1461 *
1462 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1463 * index inode described by @vi into memory from the base mft record described
1464 * by @base_ni.
1465 *
1466 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1467 * reading and looks up the attributes relating to the index described by @vi
1468 * before setting up the necessary fields in @vi as well as initializing the
1469 * ntfs inode.
1470 *
1471 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1472 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1473 * are setup like directory inodes since directories are a special case of
1474 * indices ao they need to be treated in much the same way. Most importantly,
1475 * for small indices the index allocation attribute might not actually exist.
1476 * However, the index root attribute always exists but this does not need to
1477 * have an inode associated with it and this is why we define a new inode type
1478 * index. Also, like for directories, we need to have an attribute inode for
1479 * the bitmap attribute corresponding to the index allocation attribute and we
1480 * can store this in the appropriate field of the inode, just like we do for
1481 * normal directory inodes.
1482 *
1483 * Q: What locks are held when the function is called?
1484 * A: i_state has I_NEW set, hence the inode is locked, also
1485 * i_count is set to 1, so it is not going to go away
1486 *
1487 * Return 0 on success and -errno on error. In the error case, the inode will
1488 * have had make_bad_inode() executed on it.
1489 */
1490static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1491{
1492    loff_t bvi_size;
1493    ntfs_volume *vol = NTFS_SB(vi->i_sb);
1494    ntfs_inode *ni, *base_ni, *bni;
1495    struct inode *bvi;
1496    MFT_RECORD *m;
1497    ATTR_RECORD *a;
1498    ntfs_attr_search_ctx *ctx;
1499    INDEX_ROOT *ir;
1500    u8 *ir_end, *index_end;
1501    int err = 0;
1502
1503    ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1504    ntfs_init_big_inode(vi);
1505    ni = NTFS_I(vi);
1506    base_ni = NTFS_I(base_vi);
1507    /* Just mirror the values from the base inode. */
1508    vi->i_version = base_vi->i_version;
1509    vi->i_uid = base_vi->i_uid;
1510    vi->i_gid = base_vi->i_gid;
1511    vi->i_nlink = base_vi->i_nlink;
1512    vi->i_mtime = base_vi->i_mtime;
1513    vi->i_ctime = base_vi->i_ctime;
1514    vi->i_atime = base_vi->i_atime;
1515    vi->i_generation = ni->seq_no = base_ni->seq_no;
1516    /* Set inode type to zero but preserve permissions. */
1517    vi->i_mode = base_vi->i_mode & ~S_IFMT;
1518    /* Map the mft record for the base inode. */
1519    m = map_mft_record(base_ni);
1520    if (IS_ERR(m)) {
1521        err = PTR_ERR(m);
1522        goto err_out;
1523    }
1524    ctx = ntfs_attr_get_search_ctx(base_ni, m);
1525    if (!ctx) {
1526        err = -ENOMEM;
1527        goto unm_err_out;
1528    }
1529    /* Find the index root attribute. */
1530    err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1531            CASE_SENSITIVE, 0, NULL, 0, ctx);
1532    if (unlikely(err)) {
1533        if (err == -ENOENT)
1534            ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1535                    "missing.");
1536        goto unm_err_out;
1537    }
1538    a = ctx->attr;
1539    /* Set up the state. */
1540    if (unlikely(a->non_resident)) {
1541        ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1542        goto unm_err_out;
1543    }
1544    /* Ensure the attribute name is placed before the value. */
1545    if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1546            le16_to_cpu(a->data.resident.value_offset)))) {
1547        ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1548                "after the attribute value.");
1549        goto unm_err_out;
1550    }
1551    /*
1552     * Compressed/encrypted/sparse index root is not allowed, except for
1553     * directories of course but those are not dealt with here.
1554     */
1555    if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1556            ATTR_IS_SPARSE)) {
1557        ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1558                "root attribute.");
1559        goto unm_err_out;
1560    }
1561    ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1562    ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1563    if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1564        ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1565        goto unm_err_out;
1566    }
1567    index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1568    if (index_end > ir_end) {
1569        ntfs_error(vi->i_sb, "Index is corrupt.");
1570        goto unm_err_out;
1571    }
1572    if (ir->type) {
1573        ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1574                le32_to_cpu(ir->type));
1575        goto unm_err_out;
1576    }
1577    ni->itype.index.collation_rule = ir->collation_rule;
1578    ntfs_debug("Index collation rule is 0x%x.",
1579            le32_to_cpu(ir->collation_rule));
1580    ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1581    if (!is_power_of_2(ni->itype.index.block_size)) {
1582        ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1583                "two.", ni->itype.index.block_size);
1584        goto unm_err_out;
1585    }
1586    if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1587        ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1588                "(%ld) is not supported. Sorry.",
1589                ni->itype.index.block_size, PAGE_CACHE_SIZE);
1590        err = -EOPNOTSUPP;
1591        goto unm_err_out;
1592    }
1593    if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1594        ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1595                "(%i) is not supported. Sorry.",
1596                ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1597        err = -EOPNOTSUPP;
1598        goto unm_err_out;
1599    }
1600    ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1601    /* Determine the size of a vcn in the index. */
1602    if (vol->cluster_size <= ni->itype.index.block_size) {
1603        ni->itype.index.vcn_size = vol->cluster_size;
1604        ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1605    } else {
1606        ni->itype.index.vcn_size = vol->sector_size;
1607        ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1608    }
1609    /* Check for presence of index allocation attribute. */
1610    if (!(ir->index.flags & LARGE_INDEX)) {
1611        /* No index allocation. */
1612        vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1613        /* We are done with the mft record, so we release it. */
1614        ntfs_attr_put_search_ctx(ctx);
1615        unmap_mft_record(base_ni);
1616        m = NULL;
1617        ctx = NULL;
1618        goto skip_large_index_stuff;
1619    } /* LARGE_INDEX: Index allocation present. Setup state. */
1620    NInoSetIndexAllocPresent(ni);
1621    /* Find index allocation attribute. */
1622    ntfs_attr_reinit_search_ctx(ctx);
1623    err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1624            CASE_SENSITIVE, 0, NULL, 0, ctx);
1625    if (unlikely(err)) {
1626        if (err == -ENOENT)
1627            ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1628                    "not present but $INDEX_ROOT "
1629                    "indicated it is.");
1630        else
1631            ntfs_error(vi->i_sb, "Failed to lookup "
1632                    "$INDEX_ALLOCATION attribute.");
1633        goto unm_err_out;
1634    }
1635    a = ctx->attr;
1636    if (!a->non_resident) {
1637        ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1638                "resident.");
1639        goto unm_err_out;
1640    }
1641    /*
1642     * Ensure the attribute name is placed before the mapping pairs array.
1643     */
1644    if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1645            le16_to_cpu(
1646            a->data.non_resident.mapping_pairs_offset)))) {
1647        ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1648                "placed after the mapping pairs array.");
1649        goto unm_err_out;
1650    }
1651    if (a->flags & ATTR_IS_ENCRYPTED) {
1652        ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1653                "encrypted.");
1654        goto unm_err_out;
1655    }
1656    if (a->flags & ATTR_IS_SPARSE) {
1657        ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1658        goto unm_err_out;
1659    }
1660    if (a->flags & ATTR_COMPRESSION_MASK) {
1661        ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1662                "compressed.");
1663        goto unm_err_out;
1664    }
1665    if (a->data.non_resident.lowest_vcn) {
1666        ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1667                "attribute has non zero lowest_vcn.");
1668        goto unm_err_out;
1669    }
1670    vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1671    ni->initialized_size = sle64_to_cpu(
1672            a->data.non_resident.initialized_size);
1673    ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1674    /*
1675     * We are done with the mft record, so we release it. Otherwise
1676     * we would deadlock in ntfs_attr_iget().
1677     */
1678    ntfs_attr_put_search_ctx(ctx);
1679    unmap_mft_record(base_ni);
1680    m = NULL;
1681    ctx = NULL;
1682    /* Get the index bitmap attribute inode. */
1683    bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1684    if (IS_ERR(bvi)) {
1685        ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1686        err = PTR_ERR(bvi);
1687        goto unm_err_out;
1688    }
1689    bni = NTFS_I(bvi);
1690    if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1691            NInoSparse(bni)) {
1692        ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1693                "encrypted and/or sparse.");
1694        goto iput_unm_err_out;
1695    }
1696    /* Consistency check bitmap size vs. index allocation size. */
1697    bvi_size = i_size_read(bvi);
1698    if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1699        ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1700                "index allocation (0x%llx).", bvi_size << 3,
1701                vi->i_size);
1702        goto iput_unm_err_out;
1703    }
1704    iput(bvi);
1705skip_large_index_stuff:
1706    /* Setup the operations for this index inode. */
1707    vi->i_op = NULL;
1708    vi->i_fop = NULL;
1709    vi->i_mapping->a_ops = &ntfs_mst_aops;
1710    vi->i_blocks = ni->allocated_size >> 9;
1711    /*
1712     * Make sure the base inode doesn't go away and attach it to the
1713     * index inode.
1714     */
1715    igrab(base_vi);
1716    ni->ext.base_ntfs_ino = base_ni;
1717    ni->nr_extents = -1;
1718
1719    ntfs_debug("Done.");
1720    return 0;
1721iput_unm_err_out:
1722    iput(bvi);
1723unm_err_out:
1724    if (!err)
1725        err = -EIO;
1726    if (ctx)
1727        ntfs_attr_put_search_ctx(ctx);
1728    if (m)
1729        unmap_mft_record(base_ni);
1730err_out:
1731    ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1732            "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1733            ni->name_len);
1734    make_bad_inode(vi);
1735    if (err != -EOPNOTSUPP && err != -ENOMEM)
1736        NVolSetErrors(vol);
1737    return err;
1738}
1739
1740/*
1741 * The MFT inode has special locking, so teach the lock validator
1742 * about this by splitting off the locking rules of the MFT from
1743 * the locking rules of other inodes. The MFT inode can never be
1744 * accessed from the VFS side (or even internally), only by the
1745 * map_mft functions.
1746 */
1747static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
1748
1749/**
1750 * ntfs_read_inode_mount - special read_inode for mount time use only
1751 * @vi: inode to read
1752 *
1753 * Read inode FILE_MFT at mount time, only called with super_block lock
1754 * held from within the read_super() code path.
1755 *
1756 * This function exists because when it is called the page cache for $MFT/$DATA
1757 * is not initialized and hence we cannot get at the contents of mft records
1758 * by calling map_mft_record*().
1759 *
1760 * Further it needs to cope with the circular references problem, i.e. cannot
1761 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1762 * we do not know where the other extent mft records are yet and again, because
1763 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1764 * attribute list is actually present in $MFT inode.
1765 *
1766 * We solve these problems by starting with the $DATA attribute before anything
1767 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1768 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1769 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1770 * sufficient information for the next step to complete.
1771 *
1772 * This should work but there are two possible pit falls (see inline comments
1773 * below), but only time will tell if they are real pits or just smoke...
1774 */
1775int ntfs_read_inode_mount(struct inode *vi)
1776{
1777    VCN next_vcn, last_vcn, highest_vcn;
1778    s64 block;
1779    struct super_block *sb = vi->i_sb;
1780    ntfs_volume *vol = NTFS_SB(sb);
1781    struct buffer_head *bh;
1782    ntfs_inode *ni;
1783    MFT_RECORD *m = NULL;
1784    ATTR_RECORD *a;
1785    ntfs_attr_search_ctx *ctx;
1786    unsigned int i, nr_blocks;
1787    int err;
1788
1789    ntfs_debug("Entering.");
1790
1791    /* Initialize the ntfs specific part of @vi. */
1792    ntfs_init_big_inode(vi);
1793
1794    ni = NTFS_I(vi);
1795
1796    /* Setup the data attribute. It is special as it is mst protected. */
1797    NInoSetNonResident(ni);
1798    NInoSetMstProtected(ni);
1799    NInoSetSparseDisabled(ni);
1800    ni->type = AT_DATA;
1801    ni->name = NULL;
1802    ni->name_len = 0;
1803    /*
1804     * This sets up our little cheat allowing us to reuse the async read io
1805     * completion handler for directories.
1806     */
1807    ni->itype.index.block_size = vol->mft_record_size;
1808    ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1809
1810    /* Very important! Needed to be able to call map_mft_record*(). */
1811    vol->mft_ino = vi;
1812
1813    /* Allocate enough memory to read the first mft record. */
1814    if (vol->mft_record_size > 64 * 1024) {
1815        ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1816                vol->mft_record_size);
1817        goto err_out;
1818    }
1819    i = vol->mft_record_size;
1820    if (i < sb->s_blocksize)
1821        i = sb->s_blocksize;
1822    m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1823    if (!m) {
1824        ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1825        goto err_out;
1826    }
1827
1828    /* Determine the first block of the $MFT/$DATA attribute. */
1829    block = vol->mft_lcn << vol->cluster_size_bits >>
1830            sb->s_blocksize_bits;
1831    nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1832    if (!nr_blocks)
1833        nr_blocks = 1;
1834
1835    /* Load $MFT/$DATA's first mft record. */
1836    for (i = 0; i < nr_blocks; i++) {
1837        bh = sb_bread(sb, block++);
1838        if (!bh) {
1839            ntfs_error(sb, "Device read failed.");
1840            goto err_out;
1841        }
1842        memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1843                sb->s_blocksize);
1844        brelse(bh);
1845    }
1846
1847    /* Apply the mst fixups. */
1848    if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1849        /* FIXME: Try to use the $MFTMirr now. */
1850        ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1851        goto err_out;
1852    }
1853
1854    /* Need this to sanity check attribute list references to $MFT. */
1855    vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1856
1857    /* Provides readpage() and sync_page() for map_mft_record(). */
1858    vi->i_mapping->a_ops = &ntfs_mst_aops;
1859
1860    ctx = ntfs_attr_get_search_ctx(ni, m);
1861    if (!ctx) {
1862        err = -ENOMEM;
1863        goto err_out;
1864    }
1865
1866    /* Find the attribute list attribute if present. */
1867    err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1868    if (err) {
1869        if (unlikely(err != -ENOENT)) {
1870            ntfs_error(sb, "Failed to lookup attribute list "
1871                    "attribute. You should run chkdsk.");
1872            goto put_err_out;
1873        }
1874    } else /* if (!err) */ {
1875        ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1876        u8 *al_end;
1877        static const char *es = " Not allowed. $MFT is corrupt. "
1878                "You should run chkdsk.";
1879
1880        ntfs_debug("Attribute list attribute found in $MFT.");
1881        NInoSetAttrList(ni);
1882        a = ctx->attr;
1883        if (a->flags & ATTR_COMPRESSION_MASK) {
1884            ntfs_error(sb, "Attribute list attribute is "
1885                    "compressed.%s", es);
1886            goto put_err_out;
1887        }
1888        if (a->flags & ATTR_IS_ENCRYPTED ||
1889                a->flags & ATTR_IS_SPARSE) {
1890            if (a->non_resident) {
1891                ntfs_error(sb, "Non-resident attribute list "
1892                        "attribute is encrypted/"
1893                        "sparse.%s", es);
1894                goto put_err_out;
1895            }
1896            ntfs_warning(sb, "Resident attribute list attribute "
1897                    "in $MFT system file is marked "
1898                    "encrypted/sparse which is not true. "
1899                    "However, Windows allows this and "
1900                    "chkdsk does not detect or correct it "
1901                    "so we will just ignore the invalid "
1902                    "flags and pretend they are not set.");
1903        }
1904        /* Now allocate memory for the attribute list. */
1905        ni->attr_list_size = (u32)ntfs_attr_size(a);
1906        ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1907        if (!ni->attr_list) {
1908            ntfs_error(sb, "Not enough memory to allocate buffer "
1909                    "for attribute list.");
1910            goto put_err_out;
1911        }
1912        if (a->non_resident) {
1913            NInoSetAttrListNonResident(ni);
1914            if (a->data.non_resident.lowest_vcn) {
1915                ntfs_error(sb, "Attribute list has non zero "
1916                        "lowest_vcn. $MFT is corrupt. "
1917                        "You should run chkdsk.");
1918                goto put_err_out;
1919            }
1920            /* Setup the runlist. */
1921            ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1922                    a, NULL);
1923            if (IS_ERR(ni->attr_list_rl.rl)) {
1924                err = PTR_ERR(ni->attr_list_rl.rl);
1925                ni->attr_list_rl.rl = NULL;
1926                ntfs_error(sb, "Mapping pairs decompression "
1927                        "failed with error code %i.",
1928                        -err);
1929                goto put_err_out;
1930            }
1931            /* Now load the attribute list. */
1932            if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1933                    ni->attr_list, ni->attr_list_size,
1934                    sle64_to_cpu(a->data.
1935                    non_resident.initialized_size)))) {
1936                ntfs_error(sb, "Failed to load attribute list "
1937                        "attribute with error code %i.",
1938                        -err);
1939                goto put_err_out;
1940            }
1941        } else /* if (!ctx.attr->non_resident) */ {
1942            if ((u8*)a + le16_to_cpu(
1943                    a->data.resident.value_offset) +
1944                    le32_to_cpu(
1945                    a->data.resident.value_length) >
1946                    (u8*)ctx->mrec + vol->mft_record_size) {
1947                ntfs_error(sb, "Corrupt attribute list "
1948                        "attribute.");
1949                goto put_err_out;
1950            }
1951            /* Now copy the attribute list. */
1952            memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1953                    a->data.resident.value_offset),
1954                    le32_to_cpu(
1955                    a->data.resident.value_length));
1956        }
1957        /* The attribute list is now setup in memory. */
1958        /*
1959         * FIXME: I don't know if this case is actually possible.
1960         * According to logic it is not possible but I have seen too
1961         * many weird things in MS software to rely on logic... Thus we
1962         * perform a manual search and make sure the first $MFT/$DATA
1963         * extent is in the base inode. If it is not we abort with an
1964         * error and if we ever see a report of this error we will need
1965         * to do some magic in order to have the necessary mft record
1966         * loaded and in the right place in the page cache. But
1967         * hopefully logic will prevail and this never happens...
1968         */
1969        al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1970        al_end = (u8*)al_entry + ni->attr_list_size;
1971        for (;; al_entry = next_al_entry) {
1972            /* Out of bounds check. */
1973            if ((u8*)al_entry < ni->attr_list ||
1974                    (u8*)al_entry > al_end)
1975                goto em_put_err_out;
1976            /* Catch the end of the attribute list. */
1977            if ((u8*)al_entry == al_end)
1978                goto em_put_err_out;
1979            if (!al_entry->length)
1980                goto em_put_err_out;
1981            if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1982                    le16_to_cpu(al_entry->length) > al_end)
1983                goto em_put_err_out;
1984            next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1985                    le16_to_cpu(al_entry->length));
1986            if (le32_to_cpu(al_entry->type) > le32_to_cpu(AT_DATA))
1987                goto em_put_err_out;
1988            if (AT_DATA != al_entry->type)
1989                continue;
1990            /* We want an unnamed attribute. */
1991            if (al_entry->name_length)
1992                goto em_put_err_out;
1993            /* Want the first entry, i.e. lowest_vcn == 0. */
1994            if (al_entry->lowest_vcn)
1995                goto em_put_err_out;
1996            /* First entry has to be in the base mft record. */
1997            if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1998                /* MFT references do not match, logic fails. */
1999                ntfs_error(sb, "BUG: The first $DATA extent "
2000                        "of $MFT is not in the base "
2001                        "mft record. Please report "
2002                        "you saw this message to "
2003                        "linux-ntfs-dev@lists."
2004                        "sourceforge.net");
2005                goto put_err_out;
2006            } else {
2007                /* Sequence numbers must match. */
2008                if (MSEQNO_LE(al_entry->mft_reference) !=
2009                        ni->seq_no)
2010                    goto em_put_err_out;
2011                /* Got it. All is ok. We can stop now. */
2012                break;
2013            }
2014        }
2015    }
2016
2017    ntfs_attr_reinit_search_ctx(ctx);
2018
2019    /* Now load all attribute extents. */
2020    a = NULL;
2021    next_vcn = last_vcn = highest_vcn = 0;
2022    while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
2023            ctx))) {
2024        runlist_element *nrl;
2025
2026        /* Cache the current attribute. */
2027        a = ctx->attr;
2028        /* $MFT must be non-resident. */
2029        if (!a->non_resident) {
2030            ntfs_error(sb, "$MFT must be non-resident but a "
2031                    "resident extent was found. $MFT is "
2032                    "corrupt. Run chkdsk.");
2033            goto put_err_out;
2034        }
2035        /* $MFT must be uncompressed and unencrypted. */
2036        if (a->flags & ATTR_COMPRESSION_MASK ||
2037                a->flags & ATTR_IS_ENCRYPTED ||
2038                a->flags & ATTR_IS_SPARSE) {
2039            ntfs_error(sb, "$MFT must be uncompressed, "
2040                    "non-sparse, and unencrypted but a "
2041                    "compressed/sparse/encrypted extent "
2042                    "was found. $MFT is corrupt. Run "
2043                    "chkdsk.");
2044            goto put_err_out;
2045        }
2046        /*
2047         * Decompress the mapping pairs array of this extent and merge
2048         * the result into the existing runlist. No need for locking
2049         * as we have exclusive access to the inode at this time and we
2050         * are a mount in progress task, too.
2051         */
2052        nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2053        if (IS_ERR(nrl)) {
2054            ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2055                    "failed with error code %ld. $MFT is "
2056                    "corrupt.", PTR_ERR(nrl));
2057            goto put_err_out;
2058        }
2059        ni->runlist.rl = nrl;
2060
2061        /* Are we in the first extent? */
2062        if (!next_vcn) {
2063            if (a->data.non_resident.lowest_vcn) {
2064                ntfs_error(sb, "First extent of $DATA "
2065                        "attribute has non zero "
2066                        "lowest_vcn. $MFT is corrupt. "
2067                        "You should run chkdsk.");
2068                goto put_err_out;
2069            }
2070            /* Get the last vcn in the $DATA attribute. */
2071            last_vcn = sle64_to_cpu(
2072                    a->data.non_resident.allocated_size)
2073                    >> vol->cluster_size_bits;
2074            /* Fill in the inode size. */
2075            vi->i_size = sle64_to_cpu(
2076                    a->data.non_resident.data_size);
2077            ni->initialized_size = sle64_to_cpu(
2078                    a->data.non_resident.initialized_size);
2079            ni->allocated_size = sle64_to_cpu(
2080                    a->data.non_resident.allocated_size);
2081            /*
2082             * Verify the number of mft records does not exceed
2083             * 2^32 - 1.
2084             */
2085            if ((vi->i_size >> vol->mft_record_size_bits) >=
2086                    (1ULL << 32)) {
2087                ntfs_error(sb, "$MFT is too big! Aborting.");
2088                goto put_err_out;
2089            }
2090            /*
2091             * We have got the first extent of the runlist for
2092             * $MFT which means it is now relatively safe to call
2093             * the normal ntfs_read_inode() function.
2094             * Complete reading the inode, this will actually
2095             * re-read the mft record for $MFT, this time entering
2096             * it into the page cache with which we complete the
2097             * kick start of the volume. It should be safe to do
2098             * this now as the first extent of $MFT/$DATA is
2099             * already known and we would hope that we don't need
2100             * further extents in order to find the other
2101             * attributes belonging to $MFT. Only time will tell if
2102             * this is really the case. If not we will have to play
2103             * magic at this point, possibly duplicating a lot of
2104             * ntfs_read_inode() at this point. We will need to
2105             * ensure we do enough of its work to be able to call
2106             * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2107             * hope this never happens...
2108             */
2109            ntfs_read_locked_inode(vi);
2110            if (is_bad_inode(vi)) {
2111                ntfs_error(sb, "ntfs_read_inode() of $MFT "
2112                        "failed. BUG or corrupt $MFT. "
2113                        "Run chkdsk and if no errors "
2114                        "are found, please report you "
2115                        "saw this message to "
2116                        "linux-ntfs-dev@lists."
2117                        "sourceforge.net");
2118                ntfs_attr_put_search_ctx(ctx);
2119                /* Revert to the safe super operations. */
2120                ntfs_free(m);
2121                return -1;
2122            }
2123            /*
2124             * Re-initialize some specifics about $MFT's inode as
2125             * ntfs_read_inode() will have set up the default ones.
2126             */
2127            /* Set uid and gid to root. */
2128            vi->i_uid = vi->i_gid = 0;
2129            /* Regular file. No access for anyone. */
2130            vi->i_mode = S_IFREG;
2131            /* No VFS initiated operations allowed for $MFT. */
2132            vi->i_op = &ntfs_empty_inode_ops;
2133            vi->i_fop = &ntfs_empty_file_ops;
2134        }
2135
2136        /* Get the lowest vcn for the next extent. */
2137        highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2138        next_vcn = highest_vcn + 1;
2139
2140        /* Only one extent or error, which we catch below. */
2141        if (next_vcn <= 0)
2142            break;
2143
2144        /* Avoid endless loops due to corruption. */
2145        if (next_vcn < sle64_to_cpu(
2146                a->data.non_resident.lowest_vcn)) {
2147            ntfs_error(sb, "$MFT has corrupt attribute list "
2148                    "attribute. Run chkdsk.");
2149            goto put_err_out;
2150        }
2151    }
2152    if (err != -ENOENT) {
2153        ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2154                "$MFT is corrupt. Run chkdsk.");
2155        goto put_err_out;
2156    }
2157    if (!a) {
2158        ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2159                "corrupt. Run chkdsk.");
2160        goto put_err_out;
2161    }
2162    if (highest_vcn && highest_vcn != last_vcn - 1) {
2163        ntfs_error(sb, "Failed to load the complete runlist for "
2164                "$MFT/$DATA. Driver bug or corrupt $MFT. "
2165                "Run chkdsk.");
2166        ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2167                (unsigned long long)highest_vcn,
2168                (unsigned long long)last_vcn - 1);
2169        goto put_err_out;
2170    }
2171    ntfs_attr_put_search_ctx(ctx);
2172    ntfs_debug("Done.");
2173    ntfs_free(m);
2174
2175    /*
2176     * Split the locking rules of the MFT inode from the
2177     * locking rules of other inodes:
2178     */
2179    lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
2180    lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
2181
2182    return 0;
2183
2184em_put_err_out:
2185    ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2186            "attribute list. $MFT is corrupt. Run chkdsk.");
2187put_err_out:
2188    ntfs_attr_put_search_ctx(ctx);
2189err_out:
2190    ntfs_error(sb, "Failed. Marking inode as bad.");
2191    make_bad_inode(vi);
2192    ntfs_free(m);
2193    return -1;
2194}
2195
2196static void __ntfs_clear_inode(ntfs_inode *ni)
2197{
2198    /* Free all alocated memory. */
2199    down_write(&ni->runlist.lock);
2200    if (ni->runlist.rl) {
2201        ntfs_free(ni->runlist.rl);
2202        ni->runlist.rl = NULL;
2203    }
2204    up_write(&ni->runlist.lock);
2205
2206    if (ni->attr_list) {
2207        ntfs_free(ni->attr_list);
2208        ni->attr_list = NULL;
2209    }
2210
2211    down_write(&ni->attr_list_rl.lock);
2212    if (ni->attr_list_rl.rl) {
2213        ntfs_free(ni->attr_list_rl.rl);
2214        ni->attr_list_rl.rl = NULL;
2215    }
2216    up_write(&ni->attr_list_rl.lock);
2217
2218    if (ni->name_len && ni->name != I30) {
2219        /* Catch bugs... */
2220        BUG_ON(!ni->name);
2221        kfree(ni->name);
2222    }
2223}
2224
2225void ntfs_clear_extent_inode(ntfs_inode *ni)
2226{
2227    ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2228
2229    BUG_ON(NInoAttr(ni));
2230    BUG_ON(ni->nr_extents != -1);
2231
2232#ifdef NTFS_RW
2233    if (NInoDirty(ni)) {
2234        if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2235            ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2236                    "Losing data! This is a BUG!!!");
2237        // FIXME: Do something!!!
2238    }
2239#endif /* NTFS_RW */
2240
2241    __ntfs_clear_inode(ni);
2242
2243    /* Bye, bye... */
2244    ntfs_destroy_extent_inode(ni);
2245}
2246
2247/**
2248 * ntfs_evict_big_inode - clean up the ntfs specific part of an inode
2249 * @vi: vfs inode pending annihilation
2250 *
2251 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2252 * is called, which deallocates all memory belonging to the NTFS specific part
2253 * of the inode and returns.
2254 *
2255 * If the MFT record is dirty, we commit it before doing anything else.
2256 */
2257void ntfs_evict_big_inode(struct inode *vi)
2258{
2259    ntfs_inode *ni = NTFS_I(vi);
2260
2261    truncate_inode_pages(&vi->i_data, 0);
2262    end_writeback(vi);
2263
2264#ifdef NTFS_RW
2265    if (NInoDirty(ni)) {
2266        bool was_bad = (is_bad_inode(vi));
2267
2268        /* Committing the inode also commits all extent inodes. */
2269        ntfs_commit_inode(vi);
2270
2271        if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2272            ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2273                    "0x%lx. Losing data!", vi->i_ino);
2274            // FIXME: Do something!!!
2275        }
2276    }
2277#endif /* NTFS_RW */
2278
2279    /* No need to lock at this stage as no one else has a reference. */
2280    if (ni->nr_extents > 0) {
2281        int i;
2282
2283        for (i = 0; i < ni->nr_extents; i++)
2284            ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2285        kfree(ni->ext.extent_ntfs_inos);
2286    }
2287
2288    __ntfs_clear_inode(ni);
2289
2290    if (NInoAttr(ni)) {
2291        /* Release the base inode if we are holding it. */
2292        if (ni->nr_extents == -1) {
2293            iput(VFS_I(ni->ext.base_ntfs_ino));
2294            ni->nr_extents = 0;
2295            ni->ext.base_ntfs_ino = NULL;
2296        }
2297    }
2298    return;
2299}
2300
2301/**
2302 * ntfs_show_options - show mount options in /proc/mounts
2303 * @sf: seq_file in which to write our mount options
2304 * @mnt: vfs mount whose mount options to display
2305 *
2306 * Called by the VFS once for each mounted ntfs volume when someone reads
2307 * /proc/mounts in order to display the NTFS specific mount options of each
2308 * mount. The mount options of the vfs mount @mnt are written to the seq file
2309 * @sf and success is returned.
2310 */
2311int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2312{
2313    ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2314    int i;
2315
2316    seq_printf(sf, ",uid=%i", vol->uid);
2317    seq_printf(sf, ",gid=%i", vol->gid);
2318    if (vol->fmask == vol->dmask)
2319        seq_printf(sf, ",umask=0%o", vol->fmask);
2320    else {
2321        seq_printf(sf, ",fmask=0%o", vol->fmask);
2322        seq_printf(sf, ",dmask=0%o", vol->dmask);
2323    }
2324    seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2325    if (NVolCaseSensitive(vol))
2326        seq_printf(sf, ",case_sensitive");
2327    if (NVolShowSystemFiles(vol))
2328        seq_printf(sf, ",show_sys_files");
2329    if (!NVolSparseEnabled(vol))
2330        seq_printf(sf, ",disable_sparse");
2331    for (i = 0; on_errors_arr[i].val; i++) {
2332        if (on_errors_arr[i].val & vol->on_errors)
2333            seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2334    }
2335    seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2336    return 0;
2337}
2338
2339#ifdef NTFS_RW
2340
2341static const char *es = " Leaving inconsistent metadata. Unmount and run "
2342        "chkdsk.";
2343
2344/**
2345 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2346 * @vi: inode for which the i_size was changed
2347 *
2348 * We only support i_size changes for normal files at present, i.e. not
2349 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2350 * below.
2351 *
2352 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2353 * that the change is allowed.
2354 *
2355 * This implies for us that @vi is a file inode rather than a directory, index,
2356 * or attribute inode as well as that @vi is a base inode.
2357 *
2358 * Returns 0 on success or -errno on error.
2359 *
2360 * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for
2361 * writing. The only case in the kernel where ->i_alloc_sem is not held is
2362 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
2363 * with the current i_size as the offset. The analogous place in NTFS is in
2364 * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again
2365 * without holding ->i_alloc_sem.
2366 */
2367int ntfs_truncate(struct inode *vi)
2368{
2369    s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2370    VCN highest_vcn;
2371    unsigned long flags;
2372    ntfs_inode *base_ni, *ni = NTFS_I(vi);
2373    ntfs_volume *vol = ni->vol;
2374    ntfs_attr_search_ctx *ctx;
2375    MFT_RECORD *m;
2376    ATTR_RECORD *a;
2377    const char *te = " Leaving file length out of sync with i_size.";
2378    int err, mp_size, size_change, alloc_change;
2379    u32 attr_len;
2380
2381    ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2382    BUG_ON(NInoAttr(ni));
2383    BUG_ON(S_ISDIR(vi->i_mode));
2384    BUG_ON(NInoMstProtected(ni));
2385    BUG_ON(ni->nr_extents < 0);
2386retry_truncate:
2387    /*
2388     * Lock the runlist for writing and map the mft record to ensure it is
2389     * safe to mess with the attribute runlist and sizes.
2390     */
2391    down_write(&ni->runlist.lock);
2392    if (!NInoAttr(ni))
2393        base_ni = ni;
2394    else
2395        base_ni = ni->ext.base_ntfs_ino;
2396    m = map_mft_record(base_ni);
2397    if (IS_ERR(m)) {
2398        err = PTR_ERR(m);
2399        ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2400                "(error code %d).%s", vi->i_ino, err, te);
2401        ctx = NULL;
2402        m = NULL;
2403        goto old_bad_out;
2404    }
2405    ctx = ntfs_attr_get_search_ctx(base_ni, m);
2406    if (unlikely(!ctx)) {
2407        ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2408                "inode 0x%lx (not enough memory).%s",
2409                vi->i_ino, te);
2410        err = -ENOMEM;
2411        goto old_bad_out;
2412    }
2413    err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2414            CASE_SENSITIVE, 0, NULL, 0, ctx);
2415    if (unlikely(err)) {
2416        if (err == -ENOENT) {
2417            ntfs_error(vi->i_sb, "Open attribute is missing from "
2418                    "mft record. Inode 0x%lx is corrupt. "
2419                    "Run chkdsk.%s", vi->i_ino, te);
2420            err = -EIO;
2421        } else
2422            ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2423                    "inode 0x%lx (error code %d).%s",
2424                    vi->i_ino, err, te);
2425        goto old_bad_out;
2426    }
2427    m = ctx->mrec;
2428    a = ctx->attr;
2429    /*
2430     * The i_size of the vfs inode is the new size for the attribute value.
2431     */
2432    new_size = i_size_read(vi);
2433    /* The current size of the attribute value is the old size. */
2434    old_size = ntfs_attr_size(a);
2435    /* Calculate the new allocated size. */
2436    if (NInoNonResident(ni))
2437        new_alloc_size = (new_size + vol->cluster_size - 1) &
2438                ~(s64)vol->cluster_size_mask;
2439    else
2440        new_alloc_size = (new_size + 7) & ~7;
2441    /* The current allocated size is the old allocated size. */
2442    read_lock_irqsave(&ni->size_lock, flags);
2443    old_alloc_size = ni->allocated_size;
2444    read_unlock_irqrestore(&ni->size_lock, flags);
2445    /*
2446     * The change in the file size. This will be 0 if no change, >0 if the
2447     * size is growing, and <0 if the size is shrinking.
2448     */
2449    size_change = -1;
2450    if (new_size - old_size >= 0) {
2451        size_change = 1;
2452        if (new_size == old_size)
2453            size_change = 0;
2454    }
2455    /* As above for the allocated size. */
2456    alloc_change = -1;
2457    if (new_alloc_size - old_alloc_size >= 0) {
2458        alloc_change = 1;
2459        if (new_alloc_size == old_alloc_size)
2460            alloc_change = 0;
2461    }
2462    /*
2463     * If neither the size nor the allocation are being changed there is
2464     * nothing to do.
2465     */
2466    if (!size_change && !alloc_change)
2467        goto unm_done;
2468    /* If the size is changing, check if new size is allowed in $AttrDef. */
2469    if (size_change) {
2470        err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2471        if (unlikely(err)) {
2472            if (err == -ERANGE) {
2473                ntfs_error(vol->sb, "Truncate would cause the "
2474                        "inode 0x%lx to %simum size "
2475                        "for its attribute type "
2476                        "(0x%x). Aborting truncate.",
2477                        vi->i_ino,
2478                        new_size > old_size ? "exceed "
2479                        "the max" : "go under the min",
2480                        le32_to_cpu(ni->type));
2481                err = -EFBIG;
2482            } else {
2483                ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2484                        "attribute type 0x%x. "
2485                        "Aborting truncate.",
2486                        vi->i_ino,
2487                        le32_to_cpu(ni->type));
2488                err = -EIO;
2489            }
2490            /* Reset the vfs inode size to the old size. */
2491            i_size_write(vi, old_size);
2492            goto err_out;
2493        }
2494    }
2495    if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2496        ntfs_warning(vi->i_sb, "Changes in inode size are not "
2497                "supported yet for %s files, ignoring.",
2498                NInoCompressed(ni) ? "compressed" :
2499                "encrypted");
2500        err = -EOPNOTSUPP;
2501        goto bad_out;
2502    }
2503    if (a->non_resident)
2504        goto do_non_resident_truncate;
2505    BUG_ON(NInoNonResident(ni));
2506    /* Resize the attribute record to best fit the new attribute size. */
2507    if (new_size < vol->mft_record_size &&
2508            !ntfs_resident_attr_value_resize(m, a, new_size)) {
2509        /* The resize succeeded! */
2510        flush_dcache_mft_record_page(ctx->ntfs_ino);
2511        mark_mft_record_dirty(ctx->ntfs_ino);
2512        write_lock_irqsave(&ni->size_lock, flags);
2513        /* Update the sizes in the ntfs inode and all is done. */
2514        ni->allocated_size = le32_to_cpu(a->length) -
2515                le16_to_cpu(a->data.resident.value_offset);
2516        /*
2517         * Note ntfs_resident_attr_value_resize() has already done any
2518         * necessary data clearing in the attribute record. When the
2519         * file is being shrunk vmtruncate() will already have cleared
2520         * the top part of the last partial page, i.e. since this is
2521         * the resident case this is the page with index 0. However,
2522         * when the file is being expanded, the page cache page data
2523         * between the old data_size, i.e. old_size, and the new_size
2524         * has not been zeroed. Fortunately, we do not need to zero it
2525         * either since on one hand it will either already be zero due
2526         * to both readpage and writepage clearing partial page data
2527         * beyond i_size in which case there is nothing to do or in the
2528         * case of the file being mmap()ped at the same time, POSIX
2529         * specifies that the behaviour is unspecified thus we do not
2530         * have to do anything. This means that in our implementation
2531         * in the rare case that the file is mmap()ped and a write
2532         * occurred into the mmap()ped region just beyond the file size
2533         * and writepage has not yet been called to write out the page
2534         * (which would clear the area beyond the file size) and we now
2535         * extend the file size to incorporate this dirty region
2536         * outside the file size, a write of the page would result in
2537         * this data being written to disk instead of being cleared.
2538         * Given both POSIX and the Linux mmap(2) man page specify that
2539         * this corner case is undefined, we choose to leave it like
2540         * that as this is much simpler for us as we cannot lock the
2541         * relevant page now since we are holding too many ntfs locks
2542         * which would result in a lock reversal deadlock.
2543         */
2544        ni->initialized_size = new_size;
2545        write_unlock_irqrestore(&ni->size_lock, flags);
2546        goto unm_done;
2547    }
2548    /* If the above resize failed, this must be an attribute extension. */
2549    BUG_ON(size_change < 0);
2550    /*
2551     * We have to drop all the locks so we can call
2552     * ntfs_attr_make_non_resident(). This could be optimised by try-
2553     * locking the first page cache page and only if that fails dropping
2554     * the locks, locking the page, and redoing all the locking and
2555     * lookups. While this would be a huge optimisation, it is not worth
2556     * it as this is definitely a slow code path as it only ever can happen
2557     * once for any given file.
2558     */
2559    ntfs_attr_put_search_ctx(ctx);
2560    unmap_mft_record(base_ni);
2561    up_write(&ni->runlist.lock);
2562    /*
2563     * Not enough space in the mft record, try to make the attribute
2564     * non-resident and if successful restart the truncation process.
2565     */
2566    err = ntfs_attr_make_non_resident(ni, old_size);
2567    if (likely(!err))
2568        goto retry_truncate;
2569    /*
2570     * Could not make non-resident. If this is due to this not being
2571     * permitted for this attribute type or there not being enough space,
2572     * try to make other attributes non-resident. Otherwise fail.
2573     */
2574    if (unlikely(err != -EPERM && err != -ENOSPC)) {
2575        ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2576                "type 0x%x, because the conversion from "
2577                "resident to non-resident attribute failed "
2578                "with error code %i.", vi->i_ino,
2579                (unsigned)le32_to_cpu(ni->type), err);
2580        if (err != -ENOMEM)
2581            err = -EIO;
2582        goto conv_err_out;
2583    }
2584    /* TODO: Not implemented from here, abort. */
2585    if (err == -ENOSPC)
2586        ntfs_error(vol->sb, "Not enough space in the mft record/on "
2587                "disk for the non-resident attribute value. "
2588                "This case is not implemented yet.");
2589    else /* if (err == -EPERM) */
2590        ntfs_error(vol->sb, "This attribute type may not be "
2591                "non-resident. This case is not implemented "
2592                "yet.");
2593    err = -EOPNOTSUPP;
2594    goto conv_err_out;
2595#if 0
2596    // TODO: Attempt to make other attributes non-resident.
2597    if (!err)
2598        goto do_resident_extend;
2599    /*
2600     * Both the attribute list attribute and the standard information
2601     * attribute must remain in the base inode. Thus, if this is one of
2602     * these attributes, we have to try to move other attributes out into
2603     * extent mft records instead.
2604     */
2605    if (ni->type == AT_ATTRIBUTE_LIST ||
2606            ni->type == AT_STANDARD_INFORMATION) {
2607        // TODO: Attempt to move other attributes into extent mft
2608        // records.
2609        err = -EOPNOTSUPP;
2610        if (!err)
2611            goto do_resident_extend;
2612        goto err_out;
2613    }
2614    // TODO: Attempt to move this attribute to an extent mft record, but
2615    // only if it is not already the only attribute in an mft record in
2616    // which case there would be nothing to gain.
2617    err = -EOPNOTSUPP;
2618    if (!err)
2619        goto do_resident_extend;
2620    /* There is nothing we can do to make enough space. )-: */
2621    goto err_out;
2622#endif
2623do_non_resident_truncate:
2624    BUG_ON(!NInoNonResident(ni));
2625    if (alloc_change < 0) {
2626        highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2627        if (highest_vcn > 0 &&
2628                old_alloc_size >> vol->cluster_size_bits >
2629                highest_vcn + 1) {
2630            /*
2631             * This attribute has multiple extents. Not yet
2632             * supported.
2633             */
2634            ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2635                    "attribute type 0x%x, because the "
2636                    "attribute is highly fragmented (it "
2637                    "consists of multiple extents) and "
2638                    "this case is not implemented yet.",
2639                    vi->i_ino,
2640                    (unsigned)le32_to_cpu(ni->type));
2641            err = -EOPNOTSUPP;
2642            goto bad_out;
2643        }
2644    }
2645    /*
2646     * If the size is shrinking, need to reduce the initialized_size and
2647     * the data_size before reducing the allocation.
2648     */
2649    if (size_change < 0) {
2650        /*
2651         * Make the valid size smaller (i_size is already up-to-date).
2652         */
2653        write_lock_irqsave(&ni->size_lock, flags);
2654        if (new_size < ni->initialized_size) {
2655            ni->initialized_size = new_size;
2656            a->data.non_resident.initialized_size =
2657                    cpu_to_sle64(new_size);
2658        }
2659        a->data.non_resident.data_size = cpu_to_sle64(new_size);
2660        write_unlock_irqrestore(&ni->size_lock, flags);
2661        flush_dcache_mft_record_page(ctx->ntfs_ino);
2662        mark_mft_record_dirty(ctx->ntfs_ino);
2663        /* If the allocated size is not changing, we are done. */
2664        if (!alloc_change)
2665            goto unm_done;
2666        /*
2667         * If the size is shrinking it makes no sense for the
2668         * allocation to be growing.
2669         */
2670        BUG_ON(alloc_change > 0);
2671    } else /* if (size_change >= 0) */ {
2672        /*
2673         * The file size is growing or staying the same but the
2674         * allocation can be shrinking, growing or staying the same.
2675         */
2676        if (alloc_change > 0) {
2677            /*
2678             * We need to extend the allocation and possibly update
2679             * the data size. If we are updating the data size,
2680             * since we are not touching the initialized_size we do
2681             * not need to worry about the actual data on disk.
2682             * And as far as the page cache is concerned, there
2683             * will be no pages beyond the old data size and any
2684             * partial region in the last page between the old and
2685             * new data size (or the end of the page if the new
2686             * data size is outside the page) does not need to be
2687             * modified as explained above for the resident
2688             * attribute truncate case. To do this, we simply drop
2689             * the locks we hold and leave all the work to our
2690             * friendly helper ntfs_attr_extend_allocation().
2691             */
2692            ntfs_attr_put_search_ctx(ctx);
2693            unmap_mft_record(base_ni);
2694            up_write(&ni->runlist.lock);
2695            err = ntfs_attr_extend_allocation(ni, new_size,
2696                    size_change > 0 ? new_size : -1, -1);
2697            /*
2698             * ntfs_attr_extend_allocation() will have done error
2699             * output already.
2700             */
2701            goto done;
2702        }
2703        if (!alloc_change)
2704            goto alloc_done;
2705    }
2706    /* alloc_change < 0 */
2707    /* Free the clusters. */
2708    nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2709            vol->cluster_size_bits, -1, ctx);
2710    m = ctx->mrec;
2711    a = ctx->attr;
2712    if (unlikely(nr_freed < 0)) {
2713        ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2714                "%lli). Unmount and run chkdsk to recover "
2715                "the lost cluster(s).", (long long)nr_freed);
2716        NVolSetErrors(vol);
2717        nr_freed = 0;
2718    }
2719    /* Truncate the runlist. */
2720    err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2721            new_alloc_size >> vol->cluster_size_bits);
2722    /*
2723     * If the runlist truncation failed and/or the search context is no
2724     * longer valid, we cannot resize the attribute record or build the
2725     * mapping pairs array thus we mark the inode bad so that no access to
2726     * the freed clusters can happen.
2727     */
2728    if (unlikely(err || IS_ERR(m))) {
2729        ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2730                IS_ERR(m) ?
2731                "restore attribute search context" :
2732                "truncate attribute runlist",
2733                IS_ERR(m) ? PTR_ERR(m) : err, es);
2734        err = -EIO;
2735        goto bad_out;
2736    }
2737    /* Get the size for the shrunk mapping pairs array for the runlist. */
2738    mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2739    if (unlikely(mp_size <= 0)) {
2740        ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2741                "attribute type 0x%x, because determining the "
2742                "size for the mapping pairs failed with error "
2743                "code %i.%s", vi->i_ino,
2744                (unsigned)le32_to_cpu(ni->type), mp_size, es);
2745        err = -EIO;
2746        goto bad_out;
2747    }
2748    /*
2749     * Shrink the attribute record for the new mapping pairs array. Note,
2750     * this cannot fail since we are making the attribute smaller thus by
2751     * definition there is enough space to do so.
2752     */
2753    attr_len = le32_to_cpu(a->length);
2754    err = ntfs_attr_record_resize(m, a, mp_size +
2755            le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2756    BUG_ON(err);
2757    /*
2758     * Generate the mapping pairs array directly into the attribute record.
2759     */
2760    err = ntfs_mapping_pairs_build(vol, (u8*)a +
2761            le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2762            mp_size, ni->runlist.rl, 0, -1, NULL);
2763    if (unlikely(err)) {
2764        ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2765                "attribute type 0x%x, because building the "
2766                "mapping pairs failed with error code %i.%s",
2767                vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2768                err, es);
2769        err = -EIO;
2770        goto bad_out;
2771    }
2772    /* Update the allocated/compressed size as well as the highest vcn. */
2773    a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2774            vol->cluster_size_bits) - 1);
2775    write_lock_irqsave(&ni->size_lock, flags);
2776    ni->allocated_size = new_alloc_size;
2777    a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2778    if (NInoSparse(ni) || NInoCompressed(ni)) {
2779        if (nr_freed) {
2780            ni->itype.compressed.size -= nr_freed <<
2781                    vol->cluster_size_bits;
2782            BUG_ON(ni->itype.compressed.size < 0);
2783            a->data.non_resident.compressed_size = cpu_to_sle64(
2784                    ni->itype.compressed.size);
2785            vi->i_blocks = ni->itype.compressed.size >> 9;
2786        }
2787    } else
2788        vi->i_blocks = new_alloc_size >> 9;
2789    write_unlock_irqrestore(&ni->size_lock, flags);
2790    /*
2791     * We have shrunk the allocation. If this is a shrinking truncate we
2792     * have already dealt with the initialized_size and the data_size above
2793     * and we are done. If the truncate is only changing the allocation
2794     * and not the data_size, we are also done. If this is an extending
2795     * truncate, need to extend the data_size now which is ensured by the
2796     * fact that @size_change is positive.
2797     */
2798alloc_done:
2799    /*
2800     * If the size is growing, need to update it now. If it is shrinking,
2801     * we have already updated it above (before the allocation change).
2802     */
2803    if (size_change > 0)
2804        a->data.non_resident.data_size = cpu_to_sle64(new_size);
2805    /* Ensure the modified mft record is written out. */
2806    flush_dcache_mft_record_page(ctx->ntfs_ino);
2807    mark_mft_record_dirty(ctx->ntfs_ino);
2808unm_done:
2809    ntfs_attr_put_search_ctx(ctx);
2810    unmap_mft_record(base_ni);
2811    up_write(&ni->runlist.lock);
2812done:
2813    /* Update the mtime and ctime on the base inode. */
2814    /* normally ->truncate shouldn't update ctime or mtime,
2815     * but ntfs did before so it got a copy & paste version
2816     * of file_update_time. one day someone should fix this
2817     * for real.
2818     */
2819    if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2820        struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb);
2821        int sync_it = 0;
2822
2823        if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2824            !timespec_equal(&VFS_I(base_ni)->i_ctime, &now))
2825            sync_it = 1;
2826        VFS_I(base_ni)->i_mtime = now;
2827        VFS_I(base_ni)->i_ctime = now;
2828
2829        if (sync_it)
2830            mark_inode_dirty_sync(VFS_I(base_ni));
2831    }
2832
2833    if (likely(!err)) {
2834        NInoClearTruncateFailed(ni);
2835        ntfs_debug("Done.");
2836    }
2837    return err;
2838old_bad_out:
2839    old_size = -1;
2840bad_out:
2841    if (err != -ENOMEM && err != -EOPNOTSUPP)
2842        NVolSetErrors(vol);
2843    if (err != -EOPNOTSUPP)
2844        NInoSetTruncateFailed(ni);
2845    else if (old_size >= 0)
2846        i_size_write(vi, old_size);
2847err_out:
2848    if (ctx)
2849        ntfs_attr_put_search_ctx(ctx);
2850    if (m)
2851        unmap_mft_record(base_ni);
2852    up_write(&ni->runlist.lock);
2853out:
2854    ntfs_debug("Failed. Returning error code %i.", err);
2855    return err;
2856conv_err_out:
2857    if (err != -ENOMEM && err != -EOPNOTSUPP)
2858        NVolSetErrors(vol);
2859    if (err != -EOPNOTSUPP)
2860        NInoSetTruncateFailed(ni);
2861    else
2862        i_size_write(vi, old_size);
2863    goto out;
2864}
2865
2866/**
2867 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2868 * @vi: inode for which the i_size was changed
2869 *
2870 * Wrapper for ntfs_truncate() that has no return value.
2871 *
2872 * See ntfs_truncate() description above for details.
2873 */
2874void ntfs_truncate_vfs(struct inode *vi) {
2875    ntfs_truncate(vi);
2876}
2877
2878/**
2879 * ntfs_setattr - called from notify_change() when an attribute is being changed
2880 * @dentry: dentry whose attributes to change
2881 * @attr: structure describing the attributes and the changes
2882 *
2883 * We have to trap VFS attempts to truncate the file described by @dentry as
2884 * soon as possible, because we do not implement changes in i_size yet. So we
2885 * abort all i_size changes here.
2886 *
2887 * We also abort all changes of user, group, and mode as we do not implement
2888 * the NTFS ACLs yet.
2889 *
2890 * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also
2891 * called with ->i_alloc_sem held for writing.
2892 */
2893int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2894{
2895    struct inode *vi = dentry->d_inode;
2896    int err;
2897    unsigned int ia_valid = attr->ia_valid;
2898
2899    err = inode_change_ok(vi, attr);
2900    if (err)
2901        goto out;
2902    /* We do not support NTFS ACLs yet. */
2903    if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2904        ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2905                "supported yet, ignoring.");
2906        err = -EOPNOTSUPP;
2907        goto out;
2908    }
2909    if (ia_valid & ATTR_SIZE) {
2910        if (attr->ia_size != i_size_read(vi)) {
2911            ntfs_inode *ni = NTFS_I(vi);
2912            /*
2913             * FIXME: For now we do not support resizing of
2914             * compressed or encrypted files yet.
2915             */
2916            if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2917                ntfs_warning(vi->i_sb, "Changes in inode size "
2918                        "are not supported yet for "
2919                        "%s files, ignoring.",
2920                        NInoCompressed(ni) ?
2921                        "compressed" : "encrypted");
2922                err = -EOPNOTSUPP;
2923            } else
2924                err = vmtruncate(vi, attr->ia_size);
2925            if (err || ia_valid == ATTR_SIZE)
2926                goto out;
2927        } else {
2928            /*
2929             * We skipped the truncate but must still update
2930             * timestamps.
2931             */
2932            ia_valid |= ATTR_MTIME | ATTR_CTIME;
2933        }
2934    }
2935    if (ia_valid & ATTR_ATIME)
2936        vi->i_atime = timespec_trunc(attr->ia_atime,
2937                vi->i_sb->s_time_gran);
2938    if (ia_valid & ATTR_MTIME)
2939        vi->i_mtime = timespec_trunc(attr->ia_mtime,
2940                vi->i_sb->s_time_gran);
2941    if (ia_valid & ATTR_CTIME)
2942        vi->i_ctime = timespec_trunc(attr->ia_ctime,
2943                vi->i_sb->s_time_gran);
2944    mark_inode_dirty(vi);
2945out:
2946    return err;
2947}
2948
2949/**
2950 * ntfs_write_inode - write out a dirty inode
2951 * @vi: inode to write out
2952 * @sync: if true, write out synchronously
2953 *
2954 * Write out a dirty inode to disk including any extent inodes if present.
2955 *
2956 * If @sync is true, commit the inode to disk and wait for io completion. This
2957 * is done using write_mft_record().
2958 *
2959 * If @sync is false, just schedule the write to happen but do not wait for i/o
2960 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2961 * marking the page (and in this case mft record) dirty but we do not implement
2962 * this yet as write_mft_record() largely ignores the @sync parameter and
2963 * always performs synchronous writes.
2964 *
2965 * Return 0 on success and -errno on error.
2966 */
2967int __ntfs_write_inode(struct inode *vi, int sync)
2968{
2969    sle64 nt;
2970    ntfs_inode *ni = NTFS_I(vi);
2971    ntfs_attr_search_ctx *ctx;
2972    MFT_RECORD *m;
2973    STANDARD_INFORMATION *si;
2974    int err = 0;
2975    bool modified = false;
2976
2977    ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2978            vi->i_ino);
2979    /*
2980     * Dirty attribute inodes are written via their real inodes so just
2981     * clean them here. Access time updates are taken care off when the
2982     * real inode is written.
2983     */
2984    if (NInoAttr(ni)) {
2985        NInoClearDirty(ni);
2986        ntfs_debug("Done.");
2987        return 0;
2988    }
2989    /* Map, pin, and lock the mft record belonging to the inode. */
2990    m = map_mft_record(ni);
2991    if (IS_ERR(m)) {
2992        err = PTR_ERR(m);
2993        goto err_out;
2994    }
2995    /* Update the access times in the standard information attribute. */
2996    ctx = ntfs_attr_get_search_ctx(ni, m);
2997    if (unlikely(!ctx)) {
2998        err = -ENOMEM;
2999        goto unm_err_out;
3000    }
3001    err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
3002            CASE_SENSITIVE, 0, NULL, 0, ctx);
3003    if (unlikely(err)) {
3004        ntfs_attr_put_search_ctx(ctx);
3005        goto unm_err_out;
3006    }
3007    si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3008            le16_to_cpu(ctx->attr->data.resident.value_offset));
3009    /* Update the access times if they have changed. */
3010    nt = utc2ntfs(vi->i_mtime);
3011    if (si->last_data_change_time != nt) {
3012        ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
3013                "new = 0x%llx", vi->i_ino, (long long)
3014                sle64_to_cpu(si->last_data_change_time),
3015                (long long)sle64_to_cpu(nt));
3016        si->last_data_change_time = nt;
3017        modified = true;
3018    }
3019    nt = utc2ntfs(vi->i_ctime);
3020    if (si->last_mft_change_time != nt) {
3021        ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
3022                "new = 0x%llx", vi->i_ino, (long long)
3023                sle64_to_cpu(si->last_mft_change_time),
3024                (long long)sle64_to_cpu(nt));
3025        si->last_mft_change_time = nt;
3026        modified = true;
3027    }
3028    nt = utc2ntfs(vi->i_atime);
3029    if (si->last_access_time != nt) {
3030        ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3031                "new = 0x%llx", vi->i_ino,
3032                (long long)sle64_to_cpu(si->last_access_time),
3033                (long long)sle64_to_cpu(nt));
3034        si->last_access_time = nt;
3035        modified = true;
3036    }
3037    /*
3038     * If we just modified the standard information attribute we need to
3039     * mark the mft record it is in dirty. We do this manually so that
3040     * mark_inode_dirty() is not called which would redirty the inode and
3041     * hence result in an infinite loop of trying to write the inode.
3042     * There is no need to mark the base inode nor the base mft record
3043     * dirty, since we are going to write this mft record below in any case
3044     * and the base mft record may actually not have been modified so it
3045     * might not need to be written out.
3046     * NOTE: It is not a problem when the inode for $MFT itself is being
3047     * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3048     * on the $MFT inode and hence ntfs_write_inode() will not be
3049     * re-invoked because of it which in turn is ok since the dirtied mft
3050     * record will be cleaned and written out to disk below, i.e. before
3051     * this function returns.
3052     */
3053    if (modified) {
3054        flush_dcache_mft_record_page(ctx->ntfs_ino);
3055        if (!NInoTestSetDirty(ctx->ntfs_ino))
3056            mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3057                    ctx->ntfs_ino->page_ofs);
3058    }
3059    ntfs_attr_put_search_ctx(ctx);
3060    /* Now the access times are updated, write the base mft record. */
3061    if (NInoDirty(ni))
3062        err = write_mft_record(ni, m, sync);
3063    /* Write all attached extent mft records. */
3064    mutex_lock(&ni->extent_lock);
3065    if (ni->nr_extents > 0) {
3066        ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3067        int i;
3068
3069        ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3070        for (i = 0; i < ni->nr_extents; i++) {
3071            ntfs_inode *tni = extent_nis[i];
3072
3073            if (NInoDirty(tni)) {
3074                MFT_RECORD *tm = map_mft_record(tni);
3075                int ret;
3076
3077                if (IS_ERR(tm)) {
3078                    if (!err || err == -ENOMEM)
3079                        err = PTR_ERR(tm);
3080                    continue;
3081                }
3082                ret = write_mft_record(tni, tm, sync);
3083                unmap_mft_record(tni);
3084                if (unlikely(ret)) {
3085                    if (!err || err == -ENOMEM)
3086                        err = ret;
3087                }
3088            }
3089        }
3090    }
3091    mutex_unlock(&ni->extent_lock);
3092    unmap_mft_record(ni);
3093    if (unlikely(err))
3094        goto err_out;
3095    ntfs_debug("Done.");
3096    return 0;
3097unm_err_out:
3098    unmap_mft_record(ni);
3099err_out:
3100    if (err == -ENOMEM) {
3101        ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
3102                "Marking the inode dirty again, so the VFS "
3103                "retries later.");
3104        mark_inode_dirty(vi);
3105    } else {
3106        ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err);
3107        NVolSetErrors(ni->vol);
3108    }
3109    return err;
3110}
3111
3112#endif /* NTFS_RW */
3113

Archive Download this file



interactive