blob: 4c86b7e1d1eb435addf5822d6fd67132b580604a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/**
2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
3 *
Anton Altaparmakova0646a12006-03-23 15:53:03 +00004 * Copyright (c) 2001-2006 Anton Altaparmakov
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 *
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
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/buffer_head.h>
Anton Altaparmakova778f212006-03-23 16:18:23 +000023#include <linux/fs.h>
24#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/mount.h>
Anton Altaparmakova0646a12006-03-23 15:53:03 +000026#include <linux/mutex.h>
Anton Altaparmakova778f212006-03-23 16:18:23 +000027#include <linux/pagemap.h>
28#include <linux/quotaops.h>
29#include <linux/slab.h>
30#include <linux/smp_lock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32#include "aops.h"
Anton Altaparmakova0646a12006-03-23 15:53:03 +000033#include "attrib.h"
Anton Altaparmakova778f212006-03-23 16:18:23 +000034#include "bitmap.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include "dir.h"
36#include "debug.h"
37#include "inode.h"
38#include "attrib.h"
Anton Altaparmakovdd072332005-10-04 15:39:02 +010039#include "lcnalloc.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070040#include "malloc.h"
41#include "mft.h"
42#include "time.h"
43#include "ntfs.h"
44
45/**
46 * ntfs_test_inode - compare two (possibly fake) inodes for equality
47 * @vi: vfs inode which to test
48 * @na: ntfs attribute which is being tested with
49 *
50 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
51 * inode @vi for equality with the ntfs attribute @na.
52 *
53 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
54 * @na->name and @na->name_len are then ignored.
55 *
56 * Return 1 if the attributes match and 0 if not.
57 *
58 * NOTE: This function runs with the inode_lock spin lock held so it is not
59 * allowed to sleep.
60 */
61int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
62{
63 ntfs_inode *ni;
64
65 if (vi->i_ino != na->mft_no)
66 return 0;
67 ni = NTFS_I(vi);
68 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */
69 if (likely(!NInoAttr(ni))) {
70 /* If not looking for a normal inode this is a mismatch. */
71 if (unlikely(na->type != AT_UNUSED))
72 return 0;
73 } else {
74 /* A fake inode describing an attribute. */
75 if (ni->type != na->type)
76 return 0;
77 if (ni->name_len != na->name_len)
78 return 0;
79 if (na->name_len && memcmp(ni->name, na->name,
80 na->name_len * sizeof(ntfschar)))
81 return 0;
82 }
83 /* Match! */
84 return 1;
85}
86
87/**
88 * ntfs_init_locked_inode - initialize an inode
89 * @vi: vfs inode to initialize
90 * @na: ntfs attribute which to initialize @vi to
91 *
92 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
93 * order to enable ntfs_test_inode() to do its work.
94 *
95 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
96 * In that case, @na->name and @na->name_len should be set to NULL and 0,
97 * respectively. Although that is not strictly necessary as
98 * ntfs_read_inode_locked() will fill them in later.
99 *
100 * Return 0 on success and -errno on error.
101 *
102 * NOTE: This function runs with the inode_lock spin lock held so it is not
103 * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
104 */
105static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
106{
107 ntfs_inode *ni = NTFS_I(vi);
108
109 vi->i_ino = na->mft_no;
110
111 ni->type = na->type;
112 if (na->type == AT_INDEX_ALLOCATION)
113 NInoSetMstProtected(ni);
114
115 ni->name = na->name;
116 ni->name_len = na->name_len;
117
118 /* If initializing a normal inode, we are done. */
119 if (likely(na->type == AT_UNUSED)) {
120 BUG_ON(na->name);
121 BUG_ON(na->name_len);
122 return 0;
123 }
124
125 /* It is a fake inode. */
126 NInoSetAttr(ni);
127
128 /*
129 * We have I30 global constant as an optimization as it is the name
130 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
131 * allocation but that is ok. And most attributes are unnamed anyway,
132 * thus the fraction of named attributes with name != I30 is actually
133 * absolutely tiny.
134 */
135 if (na->name_len && na->name != I30) {
136 unsigned int i;
137
138 BUG_ON(!na->name);
139 i = na->name_len * sizeof(ntfschar);
140 ni->name = (ntfschar*)kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
141 if (!ni->name)
142 return -ENOMEM;
143 memcpy(ni->name, na->name, i);
144 ni->name[i] = 0;
145 }
146 return 0;
147}
148
149typedef int (*set_t)(struct inode *, void *);
150static int ntfs_read_locked_inode(struct inode *vi);
151static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
152static int ntfs_read_locked_index_inode(struct inode *base_vi,
153 struct inode *vi);
154
155/**
156 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
157 * @sb: super block of mounted volume
158 * @mft_no: mft record number / inode number to obtain
159 *
160 * Obtain the struct inode corresponding to a specific normal inode (i.e. a
161 * file or directory).
162 *
163 * If the inode is in the cache, it is just returned with an increased
164 * reference count. Otherwise, a new struct inode is allocated and initialized,
165 * and finally ntfs_read_locked_inode() is called to read in the inode and
166 * fill in the remainder of the inode structure.
167 *
168 * Return the struct inode on success. Check the return value with IS_ERR() and
169 * if true, the function failed and the error code is obtained from PTR_ERR().
170 */
171struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
172{
173 struct inode *vi;
174 ntfs_attr na;
175 int err;
176
177 na.mft_no = mft_no;
178 na.type = AT_UNUSED;
179 na.name = NULL;
180 na.name_len = 0;
181
182 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
183 (set_t)ntfs_init_locked_inode, &na);
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000184 if (unlikely(!vi))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 return ERR_PTR(-ENOMEM);
186
187 err = 0;
188
189 /* If this is a freshly allocated inode, need to read it now. */
190 if (vi->i_state & I_NEW) {
191 err = ntfs_read_locked_inode(vi);
192 unlock_new_inode(vi);
193 }
194 /*
195 * There is no point in keeping bad inodes around if the failure was
196 * due to ENOMEM. We want to be able to retry again later.
197 */
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000198 if (unlikely(err == -ENOMEM)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 iput(vi);
200 vi = ERR_PTR(err);
201 }
202 return vi;
203}
204
205/**
206 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
207 * @base_vi: vfs base inode containing the attribute
208 * @type: attribute type
209 * @name: Unicode name of the attribute (NULL if unnamed)
210 * @name_len: length of @name in Unicode characters (0 if unnamed)
211 *
212 * Obtain the (fake) struct inode corresponding to the attribute specified by
213 * @type, @name, and @name_len, which is present in the base mft record
214 * specified by the vfs inode @base_vi.
215 *
216 * If the attribute inode is in the cache, it is just returned with an
217 * increased reference count. Otherwise, a new struct inode is allocated and
218 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
219 * attribute and fill in the inode structure.
220 *
221 * Note, for index allocation attributes, you need to use ntfs_index_iget()
222 * instead of ntfs_attr_iget() as working with indices is a lot more complex.
223 *
224 * Return the struct inode of the attribute inode on success. Check the return
225 * value with IS_ERR() and if true, the function failed and the error code is
226 * obtained from PTR_ERR().
227 */
228struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
229 ntfschar *name, u32 name_len)
230{
231 struct inode *vi;
232 ntfs_attr na;
233 int err;
234
235 /* Make sure no one calls ntfs_attr_iget() for indices. */
236 BUG_ON(type == AT_INDEX_ALLOCATION);
237
238 na.mft_no = base_vi->i_ino;
239 na.type = type;
240 na.name = name;
241 na.name_len = name_len;
242
243 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
244 (set_t)ntfs_init_locked_inode, &na);
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000245 if (unlikely(!vi))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 return ERR_PTR(-ENOMEM);
247
248 err = 0;
249
250 /* If this is a freshly allocated inode, need to read it now. */
251 if (vi->i_state & I_NEW) {
252 err = ntfs_read_locked_attr_inode(base_vi, vi);
253 unlock_new_inode(vi);
254 }
255 /*
256 * There is no point in keeping bad attribute inodes around. This also
257 * simplifies things in that we never need to check for bad attribute
258 * inodes elsewhere.
259 */
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000260 if (unlikely(err)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 iput(vi);
262 vi = ERR_PTR(err);
263 }
264 return vi;
265}
266
267/**
268 * ntfs_index_iget - obtain a struct inode corresponding to an index
269 * @base_vi: vfs base inode containing the index related attributes
270 * @name: Unicode name of the index
271 * @name_len: length of @name in Unicode characters
272 *
273 * Obtain the (fake) struct inode corresponding to the index specified by @name
274 * and @name_len, which is present in the base mft record specified by the vfs
275 * inode @base_vi.
276 *
277 * If the index inode is in the cache, it is just returned with an increased
278 * reference count. Otherwise, a new struct inode is allocated and
279 * initialized, and finally ntfs_read_locked_index_inode() is called to read
280 * the index related attributes and fill in the inode structure.
281 *
282 * Return the struct inode of the index inode on success. Check the return
283 * value with IS_ERR() and if true, the function failed and the error code is
284 * obtained from PTR_ERR().
285 */
286struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
287 u32 name_len)
288{
289 struct inode *vi;
290 ntfs_attr na;
291 int err;
292
293 na.mft_no = base_vi->i_ino;
294 na.type = AT_INDEX_ALLOCATION;
295 na.name = name;
296 na.name_len = name_len;
297
298 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
299 (set_t)ntfs_init_locked_inode, &na);
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000300 if (unlikely(!vi))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 return ERR_PTR(-ENOMEM);
302
303 err = 0;
304
305 /* If this is a freshly allocated inode, need to read it now. */
306 if (vi->i_state & I_NEW) {
307 err = ntfs_read_locked_index_inode(base_vi, vi);
308 unlock_new_inode(vi);
309 }
310 /*
311 * There is no point in keeping bad index inodes around. This also
312 * simplifies things in that we never need to check for bad index
313 * inodes elsewhere.
314 */
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000315 if (unlikely(err)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 iput(vi);
317 vi = ERR_PTR(err);
318 }
319 return vi;
320}
321
322struct inode *ntfs_alloc_big_inode(struct super_block *sb)
323{
324 ntfs_inode *ni;
325
326 ntfs_debug("Entering.");
Pekka Enberg2fb21db2005-05-25 21:15:34 +0300327 ni = kmem_cache_alloc(ntfs_big_inode_cache, SLAB_NOFS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 if (likely(ni != NULL)) {
329 ni->state = 0;
330 return VFS_I(ni);
331 }
332 ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
333 return NULL;
334}
335
336void ntfs_destroy_big_inode(struct inode *inode)
337{
338 ntfs_inode *ni = NTFS_I(inode);
339
340 ntfs_debug("Entering.");
341 BUG_ON(ni->page);
342 if (!atomic_dec_and_test(&ni->count))
343 BUG();
344 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
345}
346
347static inline ntfs_inode *ntfs_alloc_extent_inode(void)
348{
349 ntfs_inode *ni;
350
351 ntfs_debug("Entering.");
Pekka Enberg2fb21db2005-05-25 21:15:34 +0300352 ni = kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 if (likely(ni != NULL)) {
354 ni->state = 0;
355 return ni;
356 }
357 ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
358 return NULL;
359}
360
361static void ntfs_destroy_extent_inode(ntfs_inode *ni)
362{
363 ntfs_debug("Entering.");
364 BUG_ON(ni->page);
365 if (!atomic_dec_and_test(&ni->count))
366 BUG();
367 kmem_cache_free(ntfs_inode_cache, ni);
368}
369
370/**
371 * __ntfs_init_inode - initialize ntfs specific part of an inode
372 * @sb: super block of mounted volume
373 * @ni: freshly allocated ntfs inode which to initialize
374 *
375 * Initialize an ntfs inode to defaults.
376 *
377 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
378 * untouched. Make sure to initialize them elsewhere.
379 *
380 * Return zero on success and -ENOMEM on error.
381 */
382void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
383{
384 ntfs_debug("Entering.");
Anton Altaparmakov36763672004-11-18 13:46:45 +0000385 rwlock_init(&ni->size_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 ni->initialized_size = ni->allocated_size = 0;
387 ni->seq_no = 0;
388 atomic_set(&ni->count, 1);
389 ni->vol = NTFS_SB(sb);
390 ntfs_init_runlist(&ni->runlist);
Ingo Molnar4e5e5292006-03-23 16:57:48 +0000391 mutex_init(&ni->mrec_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 ni->page = NULL;
393 ni->page_ofs = 0;
394 ni->attr_list_size = 0;
395 ni->attr_list = NULL;
396 ntfs_init_runlist(&ni->attr_list_rl);
397 ni->itype.index.bmp_ino = NULL;
398 ni->itype.index.block_size = 0;
399 ni->itype.index.vcn_size = 0;
400 ni->itype.index.collation_rule = 0;
401 ni->itype.index.block_size_bits = 0;
402 ni->itype.index.vcn_size_bits = 0;
Ingo Molnar4e5e5292006-03-23 16:57:48 +0000403 mutex_init(&ni->extent_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 ni->nr_extents = 0;
405 ni->ext.base_ntfs_ino = NULL;
406}
407
408inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
409 unsigned long mft_no)
410{
411 ntfs_inode *ni = ntfs_alloc_extent_inode();
412
413 ntfs_debug("Entering.");
414 if (likely(ni != NULL)) {
415 __ntfs_init_inode(sb, ni);
416 ni->mft_no = mft_no;
417 ni->type = AT_UNUSED;
418 ni->name = NULL;
419 ni->name_len = 0;
420 }
421 return ni;
422}
423
424/**
425 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
426 * @ctx: initialized attribute search context
427 *
428 * Search all file name attributes in the inode described by the attribute
429 * search context @ctx and check if any of the names are in the $Extend system
430 * directory.
431 *
432 * Return values:
433 * 1: file is in $Extend directory
434 * 0: file is not in $Extend directory
435 * -errno: failed to determine if the file is in the $Extend directory
436 */
437static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
438{
439 int nr_links, err;
440
441 /* Restart search. */
442 ntfs_attr_reinit_search_ctx(ctx);
443
444 /* Get number of hard links. */
445 nr_links = le16_to_cpu(ctx->mrec->link_count);
446
447 /* Loop through all hard links. */
448 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
449 ctx))) {
450 FILE_NAME_ATTR *file_name_attr;
451 ATTR_RECORD *attr = ctx->attr;
452 u8 *p, *p2;
453
454 nr_links--;
455 /*
456 * Maximum sanity checking as we are called on an inode that
457 * we suspect might be corrupt.
458 */
459 p = (u8*)attr + le32_to_cpu(attr->length);
460 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
461 le32_to_cpu(ctx->mrec->bytes_in_use)) {
462err_corrupt_attr:
463 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
464 "attribute. You should run chkdsk.");
465 return -EIO;
466 }
467 if (attr->non_resident) {
468 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
469 "name. You should run chkdsk.");
470 return -EIO;
471 }
472 if (attr->flags) {
473 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
474 "invalid flags. You should run "
475 "chkdsk.");
476 return -EIO;
477 }
478 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
479 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
480 "name. You should run chkdsk.");
481 return -EIO;
482 }
483 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
484 le16_to_cpu(attr->data.resident.value_offset));
485 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
486 if (p2 < (u8*)attr || p2 > p)
487 goto err_corrupt_attr;
488 /* This attribute is ok, but is it in the $Extend directory? */
489 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
490 return 1; /* YES, it's an extended system file. */
491 }
492 if (unlikely(err != -ENOENT))
493 return err;
494 if (unlikely(nr_links)) {
495 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
496 "doesn't match number of name attributes. You "
497 "should run chkdsk.");
498 return -EIO;
499 }
500 return 0; /* NO, it is not an extended system file. */
501}
502
503/**
504 * ntfs_read_locked_inode - read an inode from its device
505 * @vi: inode to read
506 *
507 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
508 * described by @vi into memory from the device.
509 *
510 * The only fields in @vi that we need to/can look at when the function is
511 * called are i_sb, pointing to the mounted device's super block, and i_ino,
512 * the number of the inode to load.
513 *
514 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
515 * for reading and sets up the necessary @vi fields as well as initializing
516 * the ntfs inode.
517 *
518 * Q: What locks are held when the function is called?
519 * A: i_state has I_LOCK set, hence the inode is locked, also
520 * i_count is set to 1, so it is not going to go away
521 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
522 * is allowed to write to them. We should of course be honouring them but
523 * we need to do that using the IS_* macros defined in include/linux/fs.h.
524 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
525 *
526 * Return 0 on success and -errno on error. In the error case, the inode will
527 * have had make_bad_inode() executed on it.
528 */
529static int ntfs_read_locked_inode(struct inode *vi)
530{
531 ntfs_volume *vol = NTFS_SB(vi->i_sb);
532 ntfs_inode *ni;
533 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000534 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 STANDARD_INFORMATION *si;
536 ntfs_attr_search_ctx *ctx;
537 int err = 0;
538
539 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
540
541 /* Setup the generic vfs inode parts now. */
542
543 /* This is the optimal IO size (for stat), not the fs block size. */
544 vi->i_blksize = PAGE_CACHE_SIZE;
545 /*
546 * This is for checking whether an inode has changed w.r.t. a file so
547 * that the file can be updated if necessary (compare with f_version).
548 */
549 vi->i_version = 1;
550
551 vi->i_uid = vol->uid;
552 vi->i_gid = vol->gid;
553 vi->i_mode = 0;
554
555 /*
556 * Initialize the ntfs specific part of @vi special casing
557 * FILE_MFT which we need to do at mount time.
558 */
559 if (vi->i_ino != FILE_MFT)
560 ntfs_init_big_inode(vi);
561 ni = NTFS_I(vi);
562
563 m = map_mft_record(ni);
564 if (IS_ERR(m)) {
565 err = PTR_ERR(m);
566 goto err_out;
567 }
568 ctx = ntfs_attr_get_search_ctx(ni, m);
569 if (!ctx) {
570 err = -ENOMEM;
571 goto unm_err_out;
572 }
573
574 if (!(m->flags & MFT_RECORD_IN_USE)) {
575 ntfs_error(vi->i_sb, "Inode is not in use!");
576 goto unm_err_out;
577 }
578 if (m->base_mft_record) {
579 ntfs_error(vi->i_sb, "Inode is an extent inode!");
580 goto unm_err_out;
581 }
582
583 /* Transfer information from mft record into vfs and ntfs inodes. */
584 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
585
586 /*
587 * FIXME: Keep in mind that link_count is two for files which have both
588 * a long file name and a short file name as separate entries, so if
589 * we are hiding short file names this will be too high. Either we need
590 * to account for the short file names by subtracting them or we need
591 * to make sure we delete files even though i_nlink is not zero which
592 * might be tricky due to vfs interactions. Need to think about this
593 * some more when implementing the unlink command.
594 */
595 vi->i_nlink = le16_to_cpu(m->link_count);
596 /*
597 * FIXME: Reparse points can have the directory bit set even though
598 * they would be S_IFLNK. Need to deal with this further below when we
599 * implement reparse points / symbolic links but it will do for now.
600 * Also if not a directory, it could be something else, rather than
601 * a regular file. But again, will do for now.
602 */
603 /* Everyone gets all permissions. */
604 vi->i_mode |= S_IRWXUGO;
605 /* If read-only, noone gets write permissions. */
606 if (IS_RDONLY(vi))
607 vi->i_mode &= ~S_IWUGO;
608 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
609 vi->i_mode |= S_IFDIR;
610 /*
611 * Apply the directory permissions mask set in the mount
612 * options.
613 */
614 vi->i_mode &= ~vol->dmask;
615 /* Things break without this kludge! */
616 if (vi->i_nlink > 1)
617 vi->i_nlink = 1;
618 } else {
619 vi->i_mode |= S_IFREG;
620 /* Apply the file permissions mask set in the mount options. */
621 vi->i_mode &= ~vol->fmask;
622 }
623 /*
624 * Find the standard information attribute in the mft record. At this
625 * stage we haven't setup the attribute list stuff yet, so this could
626 * in fact fail if the standard information is in an extent record, but
627 * I don't think this actually ever happens.
628 */
629 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
630 ctx);
631 if (unlikely(err)) {
632 if (err == -ENOENT) {
633 /*
634 * TODO: We should be performing a hot fix here (if the
635 * recover mount option is set) by creating a new
636 * attribute.
637 */
638 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
639 "is missing.");
640 }
641 goto unm_err_out;
642 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000643 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 /* Get the standard information attribute value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000645 si = (STANDARD_INFORMATION*)((u8*)a +
646 le16_to_cpu(a->data.resident.value_offset));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647
648 /* Transfer information from the standard information into vi. */
649 /*
650 * Note: The i_?times do not quite map perfectly onto the NTFS times,
651 * but they are close enough, and in the end it doesn't really matter
652 * that much...
653 */
654 /*
655 * mtime is the last change of the data within the file. Not changed
656 * when only metadata is changed, e.g. a rename doesn't affect mtime.
657 */
658 vi->i_mtime = ntfs2utc(si->last_data_change_time);
659 /*
660 * ctime is the last change of the metadata of the file. This obviously
661 * always changes, when mtime is changed. ctime can be changed on its
662 * own, mtime is then not changed, e.g. when a file is renamed.
663 */
664 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
665 /*
666 * Last access to the data within the file. Not changed during a rename
667 * for example but changed whenever the file is written to.
668 */
669 vi->i_atime = ntfs2utc(si->last_access_time);
670
671 /* Find the attribute list attribute if present. */
672 ntfs_attr_reinit_search_ctx(ctx);
673 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
674 if (err) {
675 if (unlikely(err != -ENOENT)) {
676 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
677 "attribute.");
678 goto unm_err_out;
679 }
680 } else /* if (!err) */ {
681 if (vi->i_ino == FILE_MFT)
682 goto skip_attr_list_load;
683 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
684 NInoSetAttrList(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000685 a = ctx->attr;
Anton Altaparmakov3672b632006-02-24 09:55:07 +0000686 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 ntfs_error(vi->i_sb, "Attribute list attribute is "
Anton Altaparmakov3672b632006-02-24 09:55:07 +0000688 "compressed.");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 goto unm_err_out;
690 }
Anton Altaparmakov3672b632006-02-24 09:55:07 +0000691 if (a->flags & ATTR_IS_ENCRYPTED ||
692 a->flags & ATTR_IS_SPARSE) {
693 if (a->non_resident) {
694 ntfs_error(vi->i_sb, "Non-resident attribute "
695 "list attribute is encrypted/"
696 "sparse.");
697 goto unm_err_out;
698 }
699 ntfs_warning(vi->i_sb, "Resident attribute list "
700 "attribute in inode 0x%lx is marked "
701 "encrypted/sparse which is not true. "
702 "However, Windows allows this and "
703 "chkdsk does not detect or correct it "
704 "so we will just ignore the invalid "
705 "flags and pretend they are not set.",
706 vi->i_ino);
707 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 /* Now allocate memory for the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000709 ni->attr_list_size = (u32)ntfs_attr_size(a);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
711 if (!ni->attr_list) {
712 ntfs_error(vi->i_sb, "Not enough memory to allocate "
713 "buffer for attribute list.");
714 err = -ENOMEM;
715 goto unm_err_out;
716 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000717 if (a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 NInoSetAttrListNonResident(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000719 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 ntfs_error(vi->i_sb, "Attribute list has non "
721 "zero lowest_vcn.");
722 goto unm_err_out;
723 }
724 /*
725 * Setup the runlist. No need for locking as we have
726 * exclusive access to the inode at this time.
727 */
728 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000729 a, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 if (IS_ERR(ni->attr_list_rl.rl)) {
731 err = PTR_ERR(ni->attr_list_rl.rl);
732 ni->attr_list_rl.rl = NULL;
733 ntfs_error(vi->i_sb, "Mapping pairs "
734 "decompression failed.");
735 goto unm_err_out;
736 }
737 /* Now load the attribute list. */
738 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
739 ni->attr_list, ni->attr_list_size,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000740 sle64_to_cpu(a->data.non_resident.
741 initialized_size)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 ntfs_error(vi->i_sb, "Failed to load "
743 "attribute list attribute.");
744 goto unm_err_out;
745 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000746 } else /* if (!a->non_resident) */ {
747 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
748 + le32_to_cpu(
749 a->data.resident.value_length) >
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 (u8*)ctx->mrec + vol->mft_record_size) {
751 ntfs_error(vi->i_sb, "Corrupt attribute list "
752 "in inode.");
753 goto unm_err_out;
754 }
755 /* Now copy the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000756 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
757 a->data.resident.value_offset),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 le32_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000759 a->data.resident.value_length));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 }
761 }
762skip_attr_list_load:
763 /*
764 * If an attribute list is present we now have the attribute list value
765 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
766 */
767 if (S_ISDIR(vi->i_mode)) {
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000768 loff_t bvi_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 struct inode *bvi;
770 ntfs_inode *bni;
771 INDEX_ROOT *ir;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000772 u8 *ir_end, *index_end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774 /* It is a directory, find index root attribute. */
775 ntfs_attr_reinit_search_ctx(ctx);
776 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
777 0, NULL, 0, ctx);
778 if (unlikely(err)) {
779 if (err == -ENOENT) {
780 // FIXME: File is corrupt! Hot-fix with empty
781 // index root attribute if recovery option is
782 // set.
783 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
784 "is missing.");
785 }
786 goto unm_err_out;
787 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000788 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 /* Set up the state. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000790 if (unlikely(a->non_resident)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
792 "resident.");
793 goto unm_err_out;
794 }
795 /* Ensure the attribute name is placed before the value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000796 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
797 le16_to_cpu(a->data.resident.value_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
799 "placed after the attribute value.");
800 goto unm_err_out;
801 }
802 /*
803 * Compressed/encrypted index root just means that the newly
804 * created files in that directory should be created compressed/
805 * encrypted. However index root cannot be both compressed and
806 * encrypted.
807 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000808 if (a->flags & ATTR_COMPRESSION_MASK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 NInoSetCompressed(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000810 if (a->flags & ATTR_IS_ENCRYPTED) {
811 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 ntfs_error(vi->i_sb, "Found encrypted and "
813 "compressed attribute.");
814 goto unm_err_out;
815 }
816 NInoSetEncrypted(ni);
817 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000818 if (a->flags & ATTR_IS_SPARSE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 NInoSetSparse(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000820 ir = (INDEX_ROOT*)((u8*)a +
821 le16_to_cpu(a->data.resident.value_offset));
822 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
823 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
825 "corrupt.");
826 goto unm_err_out;
827 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000828 index_end = (u8*)&ir->index +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 le32_to_cpu(ir->index.index_length);
830 if (index_end > ir_end) {
831 ntfs_error(vi->i_sb, "Directory index is corrupt.");
832 goto unm_err_out;
833 }
834 if (ir->type != AT_FILE_NAME) {
835 ntfs_error(vi->i_sb, "Indexed attribute is not "
836 "$FILE_NAME.");
837 goto unm_err_out;
838 }
839 if (ir->collation_rule != COLLATION_FILE_NAME) {
840 ntfs_error(vi->i_sb, "Index collation rule is not "
841 "COLLATION_FILE_NAME.");
842 goto unm_err_out;
843 }
844 ni->itype.index.collation_rule = ir->collation_rule;
845 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
846 if (ni->itype.index.block_size &
847 (ni->itype.index.block_size - 1)) {
848 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
849 "power of two.",
850 ni->itype.index.block_size);
851 goto unm_err_out;
852 }
853 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
854 ntfs_error(vi->i_sb, "Index block size (%u) > "
855 "PAGE_CACHE_SIZE (%ld) is not "
856 "supported. Sorry.",
857 ni->itype.index.block_size,
858 PAGE_CACHE_SIZE);
859 err = -EOPNOTSUPP;
860 goto unm_err_out;
861 }
862 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
863 ntfs_error(vi->i_sb, "Index block size (%u) < "
864 "NTFS_BLOCK_SIZE (%i) is not "
865 "supported. Sorry.",
866 ni->itype.index.block_size,
867 NTFS_BLOCK_SIZE);
868 err = -EOPNOTSUPP;
869 goto unm_err_out;
870 }
871 ni->itype.index.block_size_bits =
872 ffs(ni->itype.index.block_size) - 1;
873 /* Determine the size of a vcn in the directory index. */
874 if (vol->cluster_size <= ni->itype.index.block_size) {
875 ni->itype.index.vcn_size = vol->cluster_size;
876 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
877 } else {
878 ni->itype.index.vcn_size = vol->sector_size;
879 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
880 }
881
882 /* Setup the index allocation attribute, even if not present. */
883 NInoSetMstProtected(ni);
884 ni->type = AT_INDEX_ALLOCATION;
885 ni->name = I30;
886 ni->name_len = 4;
887
888 if (!(ir->index.flags & LARGE_INDEX)) {
889 /* No index allocation. */
890 vi->i_size = ni->initialized_size =
891 ni->allocated_size = 0;
892 /* We are done with the mft record, so we release it. */
893 ntfs_attr_put_search_ctx(ctx);
894 unmap_mft_record(ni);
895 m = NULL;
896 ctx = NULL;
897 goto skip_large_dir_stuff;
898 } /* LARGE_INDEX: Index allocation present. Setup state. */
899 NInoSetIndexAllocPresent(ni);
900 /* Find index allocation attribute. */
901 ntfs_attr_reinit_search_ctx(ctx);
902 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
903 CASE_SENSITIVE, 0, NULL, 0, ctx);
904 if (unlikely(err)) {
905 if (err == -ENOENT)
906 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
907 "attribute is not present but "
908 "$INDEX_ROOT indicated it is.");
909 else
910 ntfs_error(vi->i_sb, "Failed to lookup "
911 "$INDEX_ALLOCATION "
912 "attribute.");
913 goto unm_err_out;
914 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000915 a = ctx->attr;
916 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
918 "is resident.");
919 goto unm_err_out;
920 }
921 /*
922 * Ensure the attribute name is placed before the mapping pairs
923 * array.
924 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000925 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
926 le16_to_cpu(
927 a->data.non_resident.mapping_pairs_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
929 "is placed after the mapping pairs "
930 "array.");
931 goto unm_err_out;
932 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000933 if (a->flags & ATTR_IS_ENCRYPTED) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
935 "is encrypted.");
936 goto unm_err_out;
937 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000938 if (a->flags & ATTR_IS_SPARSE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
940 "is sparse.");
941 goto unm_err_out;
942 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000943 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
945 "is compressed.");
946 goto unm_err_out;
947 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000948 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 ntfs_error(vi->i_sb, "First extent of "
950 "$INDEX_ALLOCATION attribute has non "
951 "zero lowest_vcn.");
952 goto unm_err_out;
953 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000954 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000956 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000958 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 /*
960 * We are done with the mft record, so we release it. Otherwise
961 * we would deadlock in ntfs_attr_iget().
962 */
963 ntfs_attr_put_search_ctx(ctx);
964 unmap_mft_record(ni);
965 m = NULL;
966 ctx = NULL;
967 /* Get the index bitmap attribute inode. */
968 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
969 if (IS_ERR(bvi)) {
970 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
971 err = PTR_ERR(bvi);
972 goto unm_err_out;
973 }
974 ni->itype.index.bmp_ino = bvi;
975 bni = NTFS_I(bvi);
976 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
977 NInoSparse(bni)) {
978 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
979 "and/or encrypted and/or sparse.");
980 goto unm_err_out;
981 }
982 /* Consistency check bitmap size vs. index allocation size. */
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000983 bvi_size = i_size_read(bvi);
984 if ((bvi_size << 3) < (vi->i_size >>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 ni->itype.index.block_size_bits)) {
986 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
987 "for index allocation (0x%llx).",
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000988 bvi_size << 3, vi->i_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 goto unm_err_out;
990 }
991skip_large_dir_stuff:
992 /* Setup the operations for this inode. */
993 vi->i_op = &ntfs_dir_inode_ops;
994 vi->i_fop = &ntfs_dir_ops;
995 } else {
996 /* It is a file. */
997 ntfs_attr_reinit_search_ctx(ctx);
998
999 /* Setup the data attribute, even if not present. */
1000 ni->type = AT_DATA;
1001 ni->name = NULL;
1002 ni->name_len = 0;
1003
1004 /* Find first extent of the unnamed data attribute. */
1005 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1006 if (unlikely(err)) {
1007 vi->i_size = ni->initialized_size =
1008 ni->allocated_size = 0;
1009 if (err != -ENOENT) {
1010 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1011 "attribute.");
1012 goto unm_err_out;
1013 }
1014 /*
1015 * FILE_Secure does not have an unnamed $DATA
1016 * attribute, so we special case it here.
1017 */
1018 if (vi->i_ino == FILE_Secure)
1019 goto no_data_attr_special_case;
1020 /*
1021 * Most if not all the system files in the $Extend
1022 * system directory do not have unnamed data
1023 * attributes so we need to check if the parent
1024 * directory of the file is FILE_Extend and if it is
1025 * ignore this error. To do this we need to get the
1026 * name of this inode from the mft record as the name
1027 * contains the back reference to the parent directory.
1028 */
1029 if (ntfs_is_extended_system_file(ctx) > 0)
1030 goto no_data_attr_special_case;
1031 // FIXME: File is corrupt! Hot-fix with empty data
1032 // attribute if recovery option is set.
1033 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1034 goto unm_err_out;
1035 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001036 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 /* Setup the state. */
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001038 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1039 if (a->flags & ATTR_COMPRESSION_MASK) {
1040 NInoSetCompressed(ni);
1041 if (vol->cluster_size > 4096) {
1042 ntfs_error(vi->i_sb, "Found "
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001043 "compressed data but "
1044 "compression is "
1045 "disabled due to "
1046 "cluster size (%i) > "
1047 "4kiB.",
1048 vol->cluster_size);
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001049 goto unm_err_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 }
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001051 if ((a->flags & ATTR_COMPRESSION_MASK)
1052 != ATTR_IS_COMPRESSED) {
1053 ntfs_error(vi->i_sb, "Found unknown "
1054 "compression method "
1055 "or corrupt file.");
1056 goto unm_err_out;
1057 }
1058 }
1059 if (a->flags & ATTR_IS_SPARSE)
1060 NInoSetSparse(ni);
1061 }
1062 if (a->flags & ATTR_IS_ENCRYPTED) {
1063 if (NInoCompressed(ni)) {
1064 ntfs_error(vi->i_sb, "Found encrypted and "
1065 "compressed data.");
1066 goto unm_err_out;
1067 }
1068 NInoSetEncrypted(ni);
1069 }
1070 if (a->non_resident) {
1071 NInoSetNonResident(ni);
1072 if (NInoCompressed(ni) || NInoSparse(ni)) {
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001073 if (NInoCompressed(ni) && a->data.non_resident.
1074 compression_unit != 4) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 ntfs_error(vi->i_sb, "Found "
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001076 "non-standard "
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001077 "compression unit (%u "
1078 "instead of 4). "
1079 "Cannot handle this.",
1080 a->data.non_resident.
1081 compression_unit);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 err = -EOPNOTSUPP;
1083 goto unm_err_out;
1084 }
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001085 if (a->data.non_resident.compression_unit) {
1086 ni->itype.compressed.block_size = 1U <<
1087 (a->data.non_resident.
1088 compression_unit +
1089 vol->cluster_size_bits);
1090 ni->itype.compressed.block_size_bits =
1091 ffs(ni->itype.
1092 compressed.
1093 block_size) - 1;
1094 ni->itype.compressed.block_clusters =
1095 1U << a->data.
1096 non_resident.
1097 compression_unit;
1098 } else {
1099 ni->itype.compressed.block_size = 0;
1100 ni->itype.compressed.block_size_bits =
1101 0;
1102 ni->itype.compressed.block_clusters =
1103 0;
1104 }
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001105 ni->itype.compressed.size = sle64_to_cpu(
1106 a->data.non_resident.
1107 compressed_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001109 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 ntfs_error(vi->i_sb, "First extent of $DATA "
1111 "attribute has non zero "
1112 "lowest_vcn.");
1113 goto unm_err_out;
1114 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 vi->i_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001116 a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001118 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001120 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 } else { /* Resident attribute. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001122 vi->i_size = ni->initialized_size = le32_to_cpu(
1123 a->data.resident.value_length);
1124 ni->allocated_size = le32_to_cpu(a->length) -
1125 le16_to_cpu(
1126 a->data.resident.value_offset);
1127 if (vi->i_size > ni->allocated_size) {
1128 ntfs_error(vi->i_sb, "Resident data attribute "
1129 "is corrupt (size exceeds "
1130 "allocation).");
1131 goto unm_err_out;
1132 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 }
1134no_data_attr_special_case:
1135 /* We are done with the mft record, so we release it. */
1136 ntfs_attr_put_search_ctx(ctx);
1137 unmap_mft_record(ni);
1138 m = NULL;
1139 ctx = NULL;
1140 /* Setup the operations for this inode. */
1141 vi->i_op = &ntfs_file_inode_ops;
1142 vi->i_fop = &ntfs_file_ops;
1143 }
1144 if (NInoMstProtected(ni))
1145 vi->i_mapping->a_ops = &ntfs_mst_aops;
1146 else
1147 vi->i_mapping->a_ops = &ntfs_aops;
1148 /*
1149 * The number of 512-byte blocks used on disk (for stat). This is in so
1150 * far inaccurate as it doesn't account for any named streams or other
1151 * special non-resident attributes, but that is how Windows works, too,
1152 * so we are at least consistent with Windows, if not entirely
1153 * consistent with the Linux Way. Doing it the Linux Way would cause a
1154 * significant slowdown as it would involve iterating over all
1155 * attributes in the mft record and adding the allocated/compressed
1156 * sizes of all non-resident attributes present to give us the Linux
1157 * correct size that should go into i_blocks (after division by 512).
1158 */
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001159 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 vi->i_blocks = ni->itype.compressed.size >> 9;
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001161 else
1162 vi->i_blocks = ni->allocated_size >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 ntfs_debug("Done.");
1164 return 0;
1165
1166unm_err_out:
1167 if (!err)
1168 err = -EIO;
1169 if (ctx)
1170 ntfs_attr_put_search_ctx(ctx);
1171 if (m)
1172 unmap_mft_record(ni);
1173err_out:
1174 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1175 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1176 make_bad_inode(vi);
1177 if (err != -EOPNOTSUPP && err != -ENOMEM)
1178 NVolSetErrors(vol);
1179 return err;
1180}
1181
1182/**
1183 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1184 * @base_vi: base inode
1185 * @vi: attribute inode to read
1186 *
1187 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1188 * attribute inode described by @vi into memory from the base mft record
1189 * described by @base_ni.
1190 *
1191 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1192 * reading and looks up the attribute described by @vi before setting up the
1193 * necessary fields in @vi as well as initializing the ntfs inode.
1194 *
1195 * Q: What locks are held when the function is called?
1196 * A: i_state has I_LOCK set, hence the inode is locked, also
1197 * i_count is set to 1, so it is not going to go away
1198 *
1199 * Return 0 on success and -errno on error. In the error case, the inode will
1200 * have had make_bad_inode() executed on it.
Anton Altaparmakovf6098cf2005-09-19 09:41:39 +01001201 *
1202 * Note this cannot be called for AT_INDEX_ALLOCATION.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 */
1204static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1205{
1206 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1207 ntfs_inode *ni, *base_ni;
1208 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001209 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 ntfs_attr_search_ctx *ctx;
1211 int err = 0;
1212
1213 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1214
1215 ntfs_init_big_inode(vi);
1216
1217 ni = NTFS_I(vi);
1218 base_ni = NTFS_I(base_vi);
1219
1220 /* Just mirror the values from the base inode. */
1221 vi->i_blksize = base_vi->i_blksize;
1222 vi->i_version = base_vi->i_version;
1223 vi->i_uid = base_vi->i_uid;
1224 vi->i_gid = base_vi->i_gid;
1225 vi->i_nlink = base_vi->i_nlink;
1226 vi->i_mtime = base_vi->i_mtime;
1227 vi->i_ctime = base_vi->i_ctime;
1228 vi->i_atime = base_vi->i_atime;
1229 vi->i_generation = ni->seq_no = base_ni->seq_no;
1230
1231 /* Set inode type to zero but preserve permissions. */
1232 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1233
1234 m = map_mft_record(base_ni);
1235 if (IS_ERR(m)) {
1236 err = PTR_ERR(m);
1237 goto err_out;
1238 }
1239 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1240 if (!ctx) {
1241 err = -ENOMEM;
1242 goto unm_err_out;
1243 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 /* Find the attribute. */
1245 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1246 CASE_SENSITIVE, 0, NULL, 0, ctx);
1247 if (unlikely(err))
1248 goto unm_err_out;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001249 a = ctx->attr;
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001250 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1251 if (a->flags & ATTR_COMPRESSION_MASK) {
1252 NInoSetCompressed(ni);
1253 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1254 ni->name_len)) {
1255 ntfs_error(vi->i_sb, "Found compressed "
1256 "non-data or named data "
1257 "attribute. Please report "
1258 "you saw this message to "
1259 "linux-ntfs-dev@lists."
1260 "sourceforge.net");
1261 goto unm_err_out;
1262 }
1263 if (vol->cluster_size > 4096) {
1264 ntfs_error(vi->i_sb, "Found compressed "
1265 "attribute but compression is "
1266 "disabled due to cluster size "
1267 "(%i) > 4kiB.",
1268 vol->cluster_size);
1269 goto unm_err_out;
1270 }
1271 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1272 ATTR_IS_COMPRESSED) {
1273 ntfs_error(vi->i_sb, "Found unknown "
1274 "compression method.");
1275 goto unm_err_out;
1276 }
1277 }
1278 /*
Anton Altaparmakovf6098cf2005-09-19 09:41:39 +01001279 * The compressed/sparse flag set in an index root just means
1280 * to compress all files.
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001281 */
1282 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1283 ntfs_error(vi->i_sb, "Found mst protected attribute "
1284 "but the attribute is %s. Please "
1285 "report you saw this message to "
1286 "linux-ntfs-dev@lists.sourceforge.net",
1287 NInoCompressed(ni) ? "compressed" :
1288 "sparse");
1289 goto unm_err_out;
1290 }
1291 if (a->flags & ATTR_IS_SPARSE)
1292 NInoSetSparse(ni);
1293 }
1294 if (a->flags & ATTR_IS_ENCRYPTED) {
1295 if (NInoCompressed(ni)) {
1296 ntfs_error(vi->i_sb, "Found encrypted and compressed "
1297 "data.");
1298 goto unm_err_out;
1299 }
1300 /*
1301 * The encryption flag set in an index root just means to
1302 * encrypt all files.
1303 */
1304 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1305 ntfs_error(vi->i_sb, "Found mst protected attribute "
1306 "but the attribute is encrypted. "
1307 "Please report you saw this message "
1308 "to linux-ntfs-dev@lists.sourceforge."
1309 "net");
1310 goto unm_err_out;
1311 }
1312 if (ni->type != AT_DATA) {
1313 ntfs_error(vi->i_sb, "Found encrypted non-data "
1314 "attribute.");
1315 goto unm_err_out;
1316 }
1317 NInoSetEncrypted(ni);
1318 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001319 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 /* Ensure the attribute name is placed before the value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001321 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1322 le16_to_cpu(a->data.resident.value_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 ntfs_error(vol->sb, "Attribute name is placed after "
1324 "the attribute value.");
1325 goto unm_err_out;
1326 }
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001327 if (NInoMstProtected(ni)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 ntfs_error(vi->i_sb, "Found mst protected attribute "
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001329 "but the attribute is resident. "
1330 "Please report you saw this message to "
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 "linux-ntfs-dev@lists.sourceforge.net");
1332 goto unm_err_out;
1333 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001334 vi->i_size = ni->initialized_size = le32_to_cpu(
1335 a->data.resident.value_length);
1336 ni->allocated_size = le32_to_cpu(a->length) -
1337 le16_to_cpu(a->data.resident.value_offset);
1338 if (vi->i_size > ni->allocated_size) {
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001339 ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1340 "(size exceeds allocation).");
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001341 goto unm_err_out;
1342 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 } else {
1344 NInoSetNonResident(ni);
1345 /*
1346 * Ensure the attribute name is placed before the mapping pairs
1347 * array.
1348 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001349 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1350 le16_to_cpu(
1351 a->data.non_resident.mapping_pairs_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 ntfs_error(vol->sb, "Attribute name is placed after "
1353 "the mapping pairs array.");
1354 goto unm_err_out;
1355 }
Anton Altaparmakovf6098cf2005-09-19 09:41:39 +01001356 if (NInoCompressed(ni) || NInoSparse(ni)) {
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001357 if (NInoCompressed(ni) && a->data.non_resident.
1358 compression_unit != 4) {
1359 ntfs_error(vi->i_sb, "Found non-standard "
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 "compression unit (%u instead "
1361 "of 4). Cannot handle this.",
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001362 a->data.non_resident.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 compression_unit);
1364 err = -EOPNOTSUPP;
1365 goto unm_err_out;
1366 }
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001367 if (a->data.non_resident.compression_unit) {
1368 ni->itype.compressed.block_size = 1U <<
1369 (a->data.non_resident.
1370 compression_unit +
1371 vol->cluster_size_bits);
1372 ni->itype.compressed.block_size_bits =
1373 ffs(ni->itype.compressed.
1374 block_size) - 1;
1375 ni->itype.compressed.block_clusters = 1U <<
1376 a->data.non_resident.
1377 compression_unit;
1378 } else {
1379 ni->itype.compressed.block_size = 0;
1380 ni->itype.compressed.block_size_bits = 0;
1381 ni->itype.compressed.block_clusters = 0;
1382 }
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001383 ni->itype.compressed.size = sle64_to_cpu(
1384 a->data.non_resident.compressed_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001386 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 ntfs_error(vi->i_sb, "First extent of attribute has "
1388 "non-zero lowest_vcn.");
1389 goto unm_err_out;
1390 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001391 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001393 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001395 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 /* Setup the operations for this attribute inode. */
1398 vi->i_op = NULL;
1399 vi->i_fop = NULL;
1400 if (NInoMstProtected(ni))
1401 vi->i_mapping->a_ops = &ntfs_mst_aops;
1402 else
1403 vi->i_mapping->a_ops = &ntfs_aops;
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001404 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 vi->i_blocks = ni->itype.compressed.size >> 9;
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001406 else
1407 vi->i_blocks = ni->allocated_size >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 /*
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001409 * Make sure the base inode does not go away and attach it to the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 * attribute inode.
1411 */
1412 igrab(base_vi);
1413 ni->ext.base_ntfs_ino = base_ni;
1414 ni->nr_extents = -1;
1415
1416 ntfs_attr_put_search_ctx(ctx);
1417 unmap_mft_record(base_ni);
1418
1419 ntfs_debug("Done.");
1420 return 0;
1421
1422unm_err_out:
1423 if (!err)
1424 err = -EIO;
1425 if (ctx)
1426 ntfs_attr_put_search_ctx(ctx);
1427 unmap_mft_record(base_ni);
1428err_out:
1429 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1430 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1431 "Marking corrupt inode and base inode 0x%lx as bad. "
1432 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1433 base_vi->i_ino);
1434 make_bad_inode(vi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 if (err != -ENOMEM)
1436 NVolSetErrors(vol);
1437 return err;
1438}
1439
1440/**
1441 * ntfs_read_locked_index_inode - read an index inode from its base inode
1442 * @base_vi: base inode
1443 * @vi: index inode to read
1444 *
1445 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1446 * index inode described by @vi into memory from the base mft record described
1447 * by @base_ni.
1448 *
1449 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1450 * reading and looks up the attributes relating to the index described by @vi
1451 * before setting up the necessary fields in @vi as well as initializing the
1452 * ntfs inode.
1453 *
1454 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1455 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1456 * are setup like directory inodes since directories are a special case of
1457 * indices ao they need to be treated in much the same way. Most importantly,
1458 * for small indices the index allocation attribute might not actually exist.
1459 * However, the index root attribute always exists but this does not need to
1460 * have an inode associated with it and this is why we define a new inode type
1461 * index. Also, like for directories, we need to have an attribute inode for
1462 * the bitmap attribute corresponding to the index allocation attribute and we
1463 * can store this in the appropriate field of the inode, just like we do for
1464 * normal directory inodes.
1465 *
1466 * Q: What locks are held when the function is called?
1467 * A: i_state has I_LOCK set, hence the inode is locked, also
1468 * i_count is set to 1, so it is not going to go away
1469 *
1470 * Return 0 on success and -errno on error. In the error case, the inode will
1471 * have had make_bad_inode() executed on it.
1472 */
1473static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1474{
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +00001475 loff_t bvi_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1477 ntfs_inode *ni, *base_ni, *bni;
1478 struct inode *bvi;
1479 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001480 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 ntfs_attr_search_ctx *ctx;
1482 INDEX_ROOT *ir;
1483 u8 *ir_end, *index_end;
1484 int err = 0;
1485
1486 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1487 ntfs_init_big_inode(vi);
1488 ni = NTFS_I(vi);
1489 base_ni = NTFS_I(base_vi);
1490 /* Just mirror the values from the base inode. */
1491 vi->i_blksize = base_vi->i_blksize;
1492 vi->i_version = base_vi->i_version;
1493 vi->i_uid = base_vi->i_uid;
1494 vi->i_gid = base_vi->i_gid;
1495 vi->i_nlink = base_vi->i_nlink;
1496 vi->i_mtime = base_vi->i_mtime;
1497 vi->i_ctime = base_vi->i_ctime;
1498 vi->i_atime = base_vi->i_atime;
1499 vi->i_generation = ni->seq_no = base_ni->seq_no;
1500 /* Set inode type to zero but preserve permissions. */
1501 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1502 /* Map the mft record for the base inode. */
1503 m = map_mft_record(base_ni);
1504 if (IS_ERR(m)) {
1505 err = PTR_ERR(m);
1506 goto err_out;
1507 }
1508 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1509 if (!ctx) {
1510 err = -ENOMEM;
1511 goto unm_err_out;
1512 }
1513 /* Find the index root attribute. */
1514 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1515 CASE_SENSITIVE, 0, NULL, 0, ctx);
1516 if (unlikely(err)) {
1517 if (err == -ENOENT)
1518 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1519 "missing.");
1520 goto unm_err_out;
1521 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001522 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 /* Set up the state. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001524 if (unlikely(a->non_resident)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1526 goto unm_err_out;
1527 }
1528 /* Ensure the attribute name is placed before the value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001529 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1530 le16_to_cpu(a->data.resident.value_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1532 "after the attribute value.");
1533 goto unm_err_out;
1534 }
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001535 /*
1536 * Compressed/encrypted/sparse index root is not allowed, except for
1537 * directories of course but those are not dealt with here.
1538 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001539 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 ATTR_IS_SPARSE)) {
1541 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1542 "root attribute.");
1543 goto unm_err_out;
1544 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001545 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1546 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1548 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1549 goto unm_err_out;
1550 }
1551 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1552 if (index_end > ir_end) {
1553 ntfs_error(vi->i_sb, "Index is corrupt.");
1554 goto unm_err_out;
1555 }
1556 if (ir->type) {
1557 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1558 le32_to_cpu(ir->type));
1559 goto unm_err_out;
1560 }
1561 ni->itype.index.collation_rule = ir->collation_rule;
1562 ntfs_debug("Index collation rule is 0x%x.",
1563 le32_to_cpu(ir->collation_rule));
1564 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1565 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1566 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1567 "two.", ni->itype.index.block_size);
1568 goto unm_err_out;
1569 }
1570 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1571 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1572 "(%ld) is not supported. Sorry.",
1573 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1574 err = -EOPNOTSUPP;
1575 goto unm_err_out;
1576 }
1577 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1578 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1579 "(%i) is not supported. Sorry.",
1580 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1581 err = -EOPNOTSUPP;
1582 goto unm_err_out;
1583 }
1584 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1585 /* Determine the size of a vcn in the index. */
1586 if (vol->cluster_size <= ni->itype.index.block_size) {
1587 ni->itype.index.vcn_size = vol->cluster_size;
1588 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1589 } else {
1590 ni->itype.index.vcn_size = vol->sector_size;
1591 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1592 }
1593 /* Check for presence of index allocation attribute. */
1594 if (!(ir->index.flags & LARGE_INDEX)) {
1595 /* No index allocation. */
1596 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1597 /* We are done with the mft record, so we release it. */
1598 ntfs_attr_put_search_ctx(ctx);
1599 unmap_mft_record(base_ni);
1600 m = NULL;
1601 ctx = NULL;
1602 goto skip_large_index_stuff;
1603 } /* LARGE_INDEX: Index allocation present. Setup state. */
1604 NInoSetIndexAllocPresent(ni);
1605 /* Find index allocation attribute. */
1606 ntfs_attr_reinit_search_ctx(ctx);
1607 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1608 CASE_SENSITIVE, 0, NULL, 0, ctx);
1609 if (unlikely(err)) {
1610 if (err == -ENOENT)
1611 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1612 "not present but $INDEX_ROOT "
1613 "indicated it is.");
1614 else
1615 ntfs_error(vi->i_sb, "Failed to lookup "
1616 "$INDEX_ALLOCATION attribute.");
1617 goto unm_err_out;
1618 }
Anton Altaparmakova778f212006-03-23 16:18:23 +00001619 a = ctx->attr;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001620 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1622 "resident.");
1623 goto unm_err_out;
1624 }
1625 /*
1626 * Ensure the attribute name is placed before the mapping pairs array.
1627 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001628 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1629 le16_to_cpu(
1630 a->data.non_resident.mapping_pairs_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1632 "placed after the mapping pairs array.");
1633 goto unm_err_out;
1634 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001635 if (a->flags & ATTR_IS_ENCRYPTED) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1637 "encrypted.");
1638 goto unm_err_out;
1639 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001640 if (a->flags & ATTR_IS_SPARSE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1642 goto unm_err_out;
1643 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001644 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1646 "compressed.");
1647 goto unm_err_out;
1648 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001649 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1651 "attribute has non zero lowest_vcn.");
1652 goto unm_err_out;
1653 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001654 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001656 a->data.non_resident.initialized_size);
1657 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 /*
1659 * We are done with the mft record, so we release it. Otherwise
1660 * we would deadlock in ntfs_attr_iget().
1661 */
1662 ntfs_attr_put_search_ctx(ctx);
1663 unmap_mft_record(base_ni);
1664 m = NULL;
1665 ctx = NULL;
1666 /* Get the index bitmap attribute inode. */
1667 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1668 if (IS_ERR(bvi)) {
1669 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1670 err = PTR_ERR(bvi);
1671 goto unm_err_out;
1672 }
1673 bni = NTFS_I(bvi);
1674 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1675 NInoSparse(bni)) {
1676 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1677 "encrypted and/or sparse.");
1678 goto iput_unm_err_out;
1679 }
1680 /* Consistency check bitmap size vs. index allocation size. */
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +00001681 bvi_size = i_size_read(bvi);
1682 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +00001684 "index allocation (0x%llx).", bvi_size << 3,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 vi->i_size);
1686 goto iput_unm_err_out;
1687 }
1688 ni->itype.index.bmp_ino = bvi;
1689skip_large_index_stuff:
1690 /* Setup the operations for this index inode. */
1691 vi->i_op = NULL;
1692 vi->i_fop = NULL;
1693 vi->i_mapping->a_ops = &ntfs_mst_aops;
1694 vi->i_blocks = ni->allocated_size >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 /*
1696 * Make sure the base inode doesn't go away and attach it to the
1697 * index inode.
1698 */
1699 igrab(base_vi);
1700 ni->ext.base_ntfs_ino = base_ni;
1701 ni->nr_extents = -1;
1702
1703 ntfs_debug("Done.");
1704 return 0;
1705
1706iput_unm_err_out:
1707 iput(bvi);
1708unm_err_out:
1709 if (!err)
1710 err = -EIO;
1711 if (ctx)
1712 ntfs_attr_put_search_ctx(ctx);
1713 if (m)
1714 unmap_mft_record(base_ni);
1715err_out:
1716 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1717 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1718 ni->name_len);
1719 make_bad_inode(vi);
1720 if (err != -EOPNOTSUPP && err != -ENOMEM)
1721 NVolSetErrors(vol);
1722 return err;
1723}
1724
1725/**
1726 * ntfs_read_inode_mount - special read_inode for mount time use only
1727 * @vi: inode to read
1728 *
1729 * Read inode FILE_MFT at mount time, only called with super_block lock
1730 * held from within the read_super() code path.
1731 *
1732 * This function exists because when it is called the page cache for $MFT/$DATA
1733 * is not initialized and hence we cannot get at the contents of mft records
1734 * by calling map_mft_record*().
1735 *
1736 * Further it needs to cope with the circular references problem, i.e. cannot
1737 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1738 * we do not know where the other extent mft records are yet and again, because
1739 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1740 * attribute list is actually present in $MFT inode.
1741 *
1742 * We solve these problems by starting with the $DATA attribute before anything
1743 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1744 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1745 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1746 * sufficient information for the next step to complete.
1747 *
1748 * This should work but there are two possible pit falls (see inline comments
1749 * below), but only time will tell if they are real pits or just smoke...
1750 */
1751int ntfs_read_inode_mount(struct inode *vi)
1752{
1753 VCN next_vcn, last_vcn, highest_vcn;
1754 s64 block;
1755 struct super_block *sb = vi->i_sb;
1756 ntfs_volume *vol = NTFS_SB(sb);
1757 struct buffer_head *bh;
1758 ntfs_inode *ni;
1759 MFT_RECORD *m = NULL;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001760 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 ntfs_attr_search_ctx *ctx;
1762 unsigned int i, nr_blocks;
1763 int err;
1764
1765 ntfs_debug("Entering.");
1766
1767 /* Initialize the ntfs specific part of @vi. */
1768 ntfs_init_big_inode(vi);
1769
1770 ni = NTFS_I(vi);
1771
1772 /* Setup the data attribute. It is special as it is mst protected. */
1773 NInoSetNonResident(ni);
1774 NInoSetMstProtected(ni);
Anton Altaparmakovc002f422005-02-03 12:02:56 +00001775 NInoSetSparseDisabled(ni);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 ni->type = AT_DATA;
1777 ni->name = NULL;
1778 ni->name_len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 /*
1780 * This sets up our little cheat allowing us to reuse the async read io
1781 * completion handler for directories.
1782 */
1783 ni->itype.index.block_size = vol->mft_record_size;
1784 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1785
1786 /* Very important! Needed to be able to call map_mft_record*(). */
1787 vol->mft_ino = vi;
1788
1789 /* Allocate enough memory to read the first mft record. */
1790 if (vol->mft_record_size > 64 * 1024) {
1791 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1792 vol->mft_record_size);
1793 goto err_out;
1794 }
1795 i = vol->mft_record_size;
1796 if (i < sb->s_blocksize)
1797 i = sb->s_blocksize;
1798 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1799 if (!m) {
1800 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1801 goto err_out;
1802 }
1803
1804 /* Determine the first block of the $MFT/$DATA attribute. */
1805 block = vol->mft_lcn << vol->cluster_size_bits >>
1806 sb->s_blocksize_bits;
1807 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1808 if (!nr_blocks)
1809 nr_blocks = 1;
1810
1811 /* Load $MFT/$DATA's first mft record. */
1812 for (i = 0; i < nr_blocks; i++) {
1813 bh = sb_bread(sb, block++);
1814 if (!bh) {
1815 ntfs_error(sb, "Device read failed.");
1816 goto err_out;
1817 }
1818 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1819 sb->s_blocksize);
1820 brelse(bh);
1821 }
1822
1823 /* Apply the mst fixups. */
1824 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1825 /* FIXME: Try to use the $MFTMirr now. */
1826 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1827 goto err_out;
1828 }
1829
1830 /* Need this to sanity check attribute list references to $MFT. */
1831 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1832
1833 /* Provides readpage() and sync_page() for map_mft_record(). */
1834 vi->i_mapping->a_ops = &ntfs_mst_aops;
1835
1836 ctx = ntfs_attr_get_search_ctx(ni, m);
1837 if (!ctx) {
1838 err = -ENOMEM;
1839 goto err_out;
1840 }
1841
1842 /* Find the attribute list attribute if present. */
1843 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1844 if (err) {
1845 if (unlikely(err != -ENOENT)) {
1846 ntfs_error(sb, "Failed to lookup attribute list "
1847 "attribute. You should run chkdsk.");
1848 goto put_err_out;
1849 }
1850 } else /* if (!err) */ {
1851 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1852 u8 *al_end;
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001853 static const char *es = " Not allowed. $MFT is corrupt. "
1854 "You should run chkdsk.";
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855
1856 ntfs_debug("Attribute list attribute found in $MFT.");
1857 NInoSetAttrList(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001858 a = ctx->attr;
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001859 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 ntfs_error(sb, "Attribute list attribute is "
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001861 "compressed.%s", es);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862 goto put_err_out;
1863 }
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001864 if (a->flags & ATTR_IS_ENCRYPTED ||
1865 a->flags & ATTR_IS_SPARSE) {
1866 if (a->non_resident) {
1867 ntfs_error(sb, "Non-resident attribute list "
1868 "attribute is encrypted/"
1869 "sparse.%s", es);
1870 goto put_err_out;
1871 }
1872 ntfs_warning(sb, "Resident attribute list attribute "
1873 "in $MFT system file is marked "
1874 "encrypted/sparse which is not true. "
1875 "However, Windows allows this and "
1876 "chkdsk does not detect or correct it "
1877 "so we will just ignore the invalid "
1878 "flags and pretend they are not set.");
1879 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 /* Now allocate memory for the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001881 ni->attr_list_size = (u32)ntfs_attr_size(a);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1883 if (!ni->attr_list) {
1884 ntfs_error(sb, "Not enough memory to allocate buffer "
1885 "for attribute list.");
1886 goto put_err_out;
1887 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001888 if (a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 NInoSetAttrListNonResident(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001890 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 ntfs_error(sb, "Attribute list has non zero "
1892 "lowest_vcn. $MFT is corrupt. "
1893 "You should run chkdsk.");
1894 goto put_err_out;
1895 }
1896 /* Setup the runlist. */
1897 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001898 a, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 if (IS_ERR(ni->attr_list_rl.rl)) {
1900 err = PTR_ERR(ni->attr_list_rl.rl);
1901 ni->attr_list_rl.rl = NULL;
1902 ntfs_error(sb, "Mapping pairs decompression "
1903 "failed with error code %i.",
1904 -err);
1905 goto put_err_out;
1906 }
1907 /* Now load the attribute list. */
1908 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1909 ni->attr_list, ni->attr_list_size,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001910 sle64_to_cpu(a->data.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 non_resident.initialized_size)))) {
1912 ntfs_error(sb, "Failed to load attribute list "
1913 "attribute with error code %i.",
1914 -err);
1915 goto put_err_out;
1916 }
1917 } else /* if (!ctx.attr->non_resident) */ {
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001918 if ((u8*)a + le16_to_cpu(
1919 a->data.resident.value_offset) +
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 le32_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001921 a->data.resident.value_length) >
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 (u8*)ctx->mrec + vol->mft_record_size) {
1923 ntfs_error(sb, "Corrupt attribute list "
1924 "attribute.");
1925 goto put_err_out;
1926 }
1927 /* Now copy the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001928 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1929 a->data.resident.value_offset),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 le32_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001931 a->data.resident.value_length));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 }
1933 /* The attribute list is now setup in memory. */
1934 /*
1935 * FIXME: I don't know if this case is actually possible.
1936 * According to logic it is not possible but I have seen too
1937 * many weird things in MS software to rely on logic... Thus we
1938 * perform a manual search and make sure the first $MFT/$DATA
1939 * extent is in the base inode. If it is not we abort with an
1940 * error and if we ever see a report of this error we will need
1941 * to do some magic in order to have the necessary mft record
1942 * loaded and in the right place in the page cache. But
1943 * hopefully logic will prevail and this never happens...
1944 */
1945 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1946 al_end = (u8*)al_entry + ni->attr_list_size;
1947 for (;; al_entry = next_al_entry) {
1948 /* Out of bounds check. */
1949 if ((u8*)al_entry < ni->attr_list ||
1950 (u8*)al_entry > al_end)
1951 goto em_put_err_out;
1952 /* Catch the end of the attribute list. */
1953 if ((u8*)al_entry == al_end)
1954 goto em_put_err_out;
1955 if (!al_entry->length)
1956 goto em_put_err_out;
1957 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1958 le16_to_cpu(al_entry->length) > al_end)
1959 goto em_put_err_out;
1960 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1961 le16_to_cpu(al_entry->length));
1962 if (le32_to_cpu(al_entry->type) >
1963 const_le32_to_cpu(AT_DATA))
1964 goto em_put_err_out;
1965 if (AT_DATA != al_entry->type)
1966 continue;
1967 /* We want an unnamed attribute. */
1968 if (al_entry->name_length)
1969 goto em_put_err_out;
1970 /* Want the first entry, i.e. lowest_vcn == 0. */
1971 if (al_entry->lowest_vcn)
1972 goto em_put_err_out;
1973 /* First entry has to be in the base mft record. */
1974 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1975 /* MFT references do not match, logic fails. */
1976 ntfs_error(sb, "BUG: The first $DATA extent "
1977 "of $MFT is not in the base "
1978 "mft record. Please report "
1979 "you saw this message to "
1980 "linux-ntfs-dev@lists."
1981 "sourceforge.net");
1982 goto put_err_out;
1983 } else {
1984 /* Sequence numbers must match. */
1985 if (MSEQNO_LE(al_entry->mft_reference) !=
1986 ni->seq_no)
1987 goto em_put_err_out;
1988 /* Got it. All is ok. We can stop now. */
1989 break;
1990 }
1991 }
1992 }
1993
1994 ntfs_attr_reinit_search_ctx(ctx);
1995
1996 /* Now load all attribute extents. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001997 a = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 next_vcn = last_vcn = highest_vcn = 0;
1999 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
2000 ctx))) {
2001 runlist_element *nrl;
2002
2003 /* Cache the current attribute. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002004 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 /* $MFT must be non-resident. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002006 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 ntfs_error(sb, "$MFT must be non-resident but a "
2008 "resident extent was found. $MFT is "
2009 "corrupt. Run chkdsk.");
2010 goto put_err_out;
2011 }
2012 /* $MFT must be uncompressed and unencrypted. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002013 if (a->flags & ATTR_COMPRESSION_MASK ||
2014 a->flags & ATTR_IS_ENCRYPTED ||
2015 a->flags & ATTR_IS_SPARSE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016 ntfs_error(sb, "$MFT must be uncompressed, "
2017 "non-sparse, and unencrypted but a "
2018 "compressed/sparse/encrypted extent "
2019 "was found. $MFT is corrupt. Run "
2020 "chkdsk.");
2021 goto put_err_out;
2022 }
2023 /*
2024 * Decompress the mapping pairs array of this extent and merge
2025 * the result into the existing runlist. No need for locking
2026 * as we have exclusive access to the inode at this time and we
2027 * are a mount in progress task, too.
2028 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002029 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 if (IS_ERR(nrl)) {
2031 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2032 "failed with error code %ld. $MFT is "
2033 "corrupt.", PTR_ERR(nrl));
2034 goto put_err_out;
2035 }
2036 ni->runlist.rl = nrl;
2037
2038 /* Are we in the first extent? */
2039 if (!next_vcn) {
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002040 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 ntfs_error(sb, "First extent of $DATA "
2042 "attribute has non zero "
2043 "lowest_vcn. $MFT is corrupt. "
2044 "You should run chkdsk.");
2045 goto put_err_out;
2046 }
2047 /* Get the last vcn in the $DATA attribute. */
2048 last_vcn = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002049 a->data.non_resident.allocated_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050 >> vol->cluster_size_bits;
2051 /* Fill in the inode size. */
2052 vi->i_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002053 a->data.non_resident.data_size);
2054 ni->initialized_size = sle64_to_cpu(
2055 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002057 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 /*
2059 * Verify the number of mft records does not exceed
2060 * 2^32 - 1.
2061 */
2062 if ((vi->i_size >> vol->mft_record_size_bits) >=
2063 (1ULL << 32)) {
2064 ntfs_error(sb, "$MFT is too big! Aborting.");
2065 goto put_err_out;
2066 }
2067 /*
2068 * We have got the first extent of the runlist for
2069 * $MFT which means it is now relatively safe to call
2070 * the normal ntfs_read_inode() function.
2071 * Complete reading the inode, this will actually
2072 * re-read the mft record for $MFT, this time entering
2073 * it into the page cache with which we complete the
2074 * kick start of the volume. It should be safe to do
2075 * this now as the first extent of $MFT/$DATA is
2076 * already known and we would hope that we don't need
2077 * further extents in order to find the other
2078 * attributes belonging to $MFT. Only time will tell if
2079 * this is really the case. If not we will have to play
2080 * magic at this point, possibly duplicating a lot of
2081 * ntfs_read_inode() at this point. We will need to
2082 * ensure we do enough of its work to be able to call
2083 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2084 * hope this never happens...
2085 */
2086 ntfs_read_locked_inode(vi);
2087 if (is_bad_inode(vi)) {
2088 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2089 "failed. BUG or corrupt $MFT. "
2090 "Run chkdsk and if no errors "
2091 "are found, please report you "
2092 "saw this message to "
2093 "linux-ntfs-dev@lists."
2094 "sourceforge.net");
2095 ntfs_attr_put_search_ctx(ctx);
2096 /* Revert to the safe super operations. */
2097 ntfs_free(m);
2098 return -1;
2099 }
2100 /*
2101 * Re-initialize some specifics about $MFT's inode as
2102 * ntfs_read_inode() will have set up the default ones.
2103 */
2104 /* Set uid and gid to root. */
2105 vi->i_uid = vi->i_gid = 0;
2106 /* Regular file. No access for anyone. */
2107 vi->i_mode = S_IFREG;
2108 /* No VFS initiated operations allowed for $MFT. */
2109 vi->i_op = &ntfs_empty_inode_ops;
2110 vi->i_fop = &ntfs_empty_file_ops;
2111 }
2112
2113 /* Get the lowest vcn for the next extent. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002114 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 next_vcn = highest_vcn + 1;
2116
2117 /* Only one extent or error, which we catch below. */
2118 if (next_vcn <= 0)
2119 break;
2120
2121 /* Avoid endless loops due to corruption. */
2122 if (next_vcn < sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002123 a->data.non_resident.lowest_vcn)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124 ntfs_error(sb, "$MFT has corrupt attribute list "
2125 "attribute. Run chkdsk.");
2126 goto put_err_out;
2127 }
2128 }
2129 if (err != -ENOENT) {
2130 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2131 "$MFT is corrupt. Run chkdsk.");
2132 goto put_err_out;
2133 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002134 if (!a) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2136 "corrupt. Run chkdsk.");
2137 goto put_err_out;
2138 }
2139 if (highest_vcn && highest_vcn != last_vcn - 1) {
2140 ntfs_error(sb, "Failed to load the complete runlist for "
2141 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2142 "Run chkdsk.");
2143 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2144 (unsigned long long)highest_vcn,
2145 (unsigned long long)last_vcn - 1);
2146 goto put_err_out;
2147 }
2148 ntfs_attr_put_search_ctx(ctx);
2149 ntfs_debug("Done.");
2150 ntfs_free(m);
2151 return 0;
2152
2153em_put_err_out:
2154 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2155 "attribute list. $MFT is corrupt. Run chkdsk.");
2156put_err_out:
2157 ntfs_attr_put_search_ctx(ctx);
2158err_out:
2159 ntfs_error(sb, "Failed. Marking inode as bad.");
2160 make_bad_inode(vi);
2161 ntfs_free(m);
2162 return -1;
2163}
2164
2165/**
2166 * ntfs_put_inode - handler for when the inode reference count is decremented
2167 * @vi: vfs inode
2168 *
2169 * The VFS calls ntfs_put_inode() every time the inode reference count (i_count)
2170 * is about to be decremented (but before the decrement itself.
2171 *
2172 * If the inode @vi is a directory with two references, one of which is being
2173 * dropped, we need to put the attribute inode for the directory index bitmap,
2174 * if it is present, otherwise the directory inode would remain pinned for
2175 * ever.
2176 */
2177void ntfs_put_inode(struct inode *vi)
2178{
2179 if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
2180 ntfs_inode *ni = NTFS_I(vi);
2181 if (NInoIndexAllocPresent(ni)) {
2182 struct inode *bvi = NULL;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002183 mutex_lock(&vi->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 if (atomic_read(&vi->i_count) == 2) {
2185 bvi = ni->itype.index.bmp_ino;
2186 if (bvi)
2187 ni->itype.index.bmp_ino = NULL;
2188 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002189 mutex_unlock(&vi->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190 if (bvi)
2191 iput(bvi);
2192 }
2193 }
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_clear_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_clear_big_inode(struct inode *vi)
2258{
2259 ntfs_inode *ni = NTFS_I(vi);
2260
2261 /*
2262 * If the inode @vi is an index inode we need to put the attribute
2263 * inode for the index bitmap, if it is present, otherwise the index
2264 * inode would disappear and the attribute inode for the index bitmap
2265 * would no longer be referenced from anywhere and thus it would remain
2266 * pinned for ever.
2267 */
2268 if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) &&
2269 NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) {
2270 iput(ni->itype.index.bmp_ino);
2271 ni->itype.index.bmp_ino = NULL;
2272 }
2273#ifdef NTFS_RW
2274 if (NInoDirty(ni)) {
2275 BOOL was_bad = (is_bad_inode(vi));
2276
2277 /* Committing the inode also commits all extent inodes. */
2278 ntfs_commit_inode(vi);
2279
2280 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2281 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2282 "0x%lx. Losing data!", vi->i_ino);
2283 // FIXME: Do something!!!
2284 }
2285 }
2286#endif /* NTFS_RW */
2287
2288 /* No need to lock at this stage as no one else has a reference. */
2289 if (ni->nr_extents > 0) {
2290 int i;
2291
2292 for (i = 0; i < ni->nr_extents; i++)
2293 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2294 kfree(ni->ext.extent_ntfs_inos);
2295 }
2296
2297 __ntfs_clear_inode(ni);
2298
2299 if (NInoAttr(ni)) {
2300 /* Release the base inode if we are holding it. */
2301 if (ni->nr_extents == -1) {
2302 iput(VFS_I(ni->ext.base_ntfs_ino));
2303 ni->nr_extents = 0;
2304 ni->ext.base_ntfs_ino = NULL;
2305 }
2306 }
2307 return;
2308}
2309
2310/**
2311 * ntfs_show_options - show mount options in /proc/mounts
2312 * @sf: seq_file in which to write our mount options
2313 * @mnt: vfs mount whose mount options to display
2314 *
2315 * Called by the VFS once for each mounted ntfs volume when someone reads
2316 * /proc/mounts in order to display the NTFS specific mount options of each
2317 * mount. The mount options of the vfs mount @mnt are written to the seq file
2318 * @sf and success is returned.
2319 */
2320int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2321{
2322 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2323 int i;
2324
2325 seq_printf(sf, ",uid=%i", vol->uid);
2326 seq_printf(sf, ",gid=%i", vol->gid);
2327 if (vol->fmask == vol->dmask)
2328 seq_printf(sf, ",umask=0%o", vol->fmask);
2329 else {
2330 seq_printf(sf, ",fmask=0%o", vol->fmask);
2331 seq_printf(sf, ",dmask=0%o", vol->dmask);
2332 }
2333 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2334 if (NVolCaseSensitive(vol))
2335 seq_printf(sf, ",case_sensitive");
2336 if (NVolShowSystemFiles(vol))
2337 seq_printf(sf, ",show_sys_files");
Anton Altaparmakovc002f422005-02-03 12:02:56 +00002338 if (!NVolSparseEnabled(vol))
2339 seq_printf(sf, ",disable_sparse");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340 for (i = 0; on_errors_arr[i].val; i++) {
2341 if (on_errors_arr[i].val & vol->on_errors)
2342 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2343 }
2344 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2345 return 0;
2346}
2347
2348#ifdef NTFS_RW
2349
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002350static const char *es = " Leaving inconsistent metadata. Unmount and run "
2351 "chkdsk.";
2352
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353/**
2354 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2355 * @vi: inode for which the i_size was changed
2356 *
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002357 * We only support i_size changes for normal files at present, i.e. not
2358 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2359 * below.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360 *
2361 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2362 * that the change is allowed.
2363 *
2364 * This implies for us that @vi is a file inode rather than a directory, index,
2365 * or attribute inode as well as that @vi is a base inode.
2366 *
2367 * Returns 0 on success or -errno on error.
2368 *
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002369 * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002370 * writing. The only case in the kernel where ->i_alloc_sem is not held is
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002372 * with the current i_size as the offset. The analogous place in NTFS is in
2373 * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again
2374 * without holding ->i_alloc_sem.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375 */
2376int ntfs_truncate(struct inode *vi)
2377{
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002378 s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2379 VCN highest_vcn;
2380 unsigned long flags;
2381 ntfs_inode *base_ni, *ni = NTFS_I(vi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382 ntfs_volume *vol = ni->vol;
2383 ntfs_attr_search_ctx *ctx;
2384 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002385 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386 const char *te = " Leaving file length out of sync with i_size.";
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002387 int err, mp_size, size_change, alloc_change;
2388 u32 attr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389
2390 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2391 BUG_ON(NInoAttr(ni));
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002392 BUG_ON(S_ISDIR(vi->i_mode));
2393 BUG_ON(NInoMstProtected(ni));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394 BUG_ON(ni->nr_extents < 0);
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002395retry_truncate:
2396 /*
2397 * Lock the runlist for writing and map the mft record to ensure it is
2398 * safe to mess with the attribute runlist and sizes.
2399 */
2400 down_write(&ni->runlist.lock);
2401 if (!NInoAttr(ni))
2402 base_ni = ni;
2403 else
2404 base_ni = ni->ext.base_ntfs_ino;
2405 m = map_mft_record(base_ni);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406 if (IS_ERR(m)) {
2407 err = PTR_ERR(m);
2408 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2409 "(error code %d).%s", vi->i_ino, err, te);
2410 ctx = NULL;
2411 m = NULL;
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002412 goto old_bad_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002413 }
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002414 ctx = ntfs_attr_get_search_ctx(base_ni, m);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002415 if (unlikely(!ctx)) {
2416 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2417 "inode 0x%lx (not enough memory).%s",
2418 vi->i_ino, te);
2419 err = -ENOMEM;
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002420 goto old_bad_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421 }
2422 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2423 CASE_SENSITIVE, 0, NULL, 0, ctx);
2424 if (unlikely(err)) {
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002425 if (err == -ENOENT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 ntfs_error(vi->i_sb, "Open attribute is missing from "
2427 "mft record. Inode 0x%lx is corrupt. "
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002428 "Run chkdsk.%s", vi->i_ino, te);
2429 err = -EIO;
2430 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002432 "inode 0x%lx (error code %d).%s",
2433 vi->i_ino, err, te);
2434 goto old_bad_out;
2435 }
2436 m = ctx->mrec;
2437 a = ctx->attr;
2438 /*
2439 * The i_size of the vfs inode is the new size for the attribute value.
2440 */
2441 new_size = i_size_read(vi);
2442 /* The current size of the attribute value is the old size. */
2443 old_size = ntfs_attr_size(a);
2444 /* Calculate the new allocated size. */
2445 if (NInoNonResident(ni))
2446 new_alloc_size = (new_size + vol->cluster_size - 1) &
2447 ~(s64)vol->cluster_size_mask;
2448 else
2449 new_alloc_size = (new_size + 7) & ~7;
2450 /* The current allocated size is the old allocated size. */
2451 read_lock_irqsave(&ni->size_lock, flags);
2452 old_alloc_size = ni->allocated_size;
2453 read_unlock_irqrestore(&ni->size_lock, flags);
2454 /*
2455 * The change in the file size. This will be 0 if no change, >0 if the
2456 * size is growing, and <0 if the size is shrinking.
2457 */
2458 size_change = -1;
2459 if (new_size - old_size >= 0) {
2460 size_change = 1;
2461 if (new_size == old_size)
2462 size_change = 0;
2463 }
2464 /* As above for the allocated size. */
2465 alloc_change = -1;
2466 if (new_alloc_size - old_alloc_size >= 0) {
2467 alloc_change = 1;
2468 if (new_alloc_size == old_alloc_size)
2469 alloc_change = 0;
2470 }
2471 /*
2472 * If neither the size nor the allocation are being changed there is
2473 * nothing to do.
2474 */
2475 if (!size_change && !alloc_change)
2476 goto unm_done;
2477 /* If the size is changing, check if new size is allowed in $AttrDef. */
2478 if (size_change) {
2479 err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2480 if (unlikely(err)) {
2481 if (err == -ERANGE) {
2482 ntfs_error(vol->sb, "Truncate would cause the "
2483 "inode 0x%lx to %simum size "
2484 "for its attribute type "
2485 "(0x%x). Aborting truncate.",
2486 vi->i_ino,
2487 new_size > old_size ? "exceed "
2488 "the max" : "go under the min",
2489 le32_to_cpu(ni->type));
2490 err = -EFBIG;
2491 } else {
2492 ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2493 "attribute type 0x%x. "
2494 "Aborting truncate.",
2495 vi->i_ino,
2496 le32_to_cpu(ni->type));
2497 err = -EIO;
2498 }
2499 /* Reset the vfs inode size to the old size. */
2500 i_size_write(vi, old_size);
2501 goto err_out;
2502 }
2503 }
2504 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2505 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2506 "supported yet for %s files, ignoring.",
2507 NInoCompressed(ni) ? "compressed" :
2508 "encrypted");
2509 err = -EOPNOTSUPP;
2510 goto bad_out;
2511 }
2512 if (a->non_resident)
2513 goto do_non_resident_truncate;
2514 BUG_ON(NInoNonResident(ni));
2515 /* Resize the attribute record to best fit the new attribute size. */
2516 if (new_size < vol->mft_record_size &&
2517 !ntfs_resident_attr_value_resize(m, a, new_size)) {
2518 unsigned long flags;
2519
2520 /* The resize succeeded! */
2521 flush_dcache_mft_record_page(ctx->ntfs_ino);
2522 mark_mft_record_dirty(ctx->ntfs_ino);
2523 write_lock_irqsave(&ni->size_lock, flags);
2524 /* Update the sizes in the ntfs inode and all is done. */
2525 ni->allocated_size = le32_to_cpu(a->length) -
2526 le16_to_cpu(a->data.resident.value_offset);
2527 /*
2528 * Note ntfs_resident_attr_value_resize() has already done any
2529 * necessary data clearing in the attribute record. When the
2530 * file is being shrunk vmtruncate() will already have cleared
2531 * the top part of the last partial page, i.e. since this is
2532 * the resident case this is the page with index 0. However,
2533 * when the file is being expanded, the page cache page data
2534 * between the old data_size, i.e. old_size, and the new_size
2535 * has not been zeroed. Fortunately, we do not need to zero it
2536 * either since on one hand it will either already be zero due
2537 * to both readpage and writepage clearing partial page data
2538 * beyond i_size in which case there is nothing to do or in the
2539 * case of the file being mmap()ped at the same time, POSIX
2540 * specifies that the behaviour is unspecified thus we do not
2541 * have to do anything. This means that in our implementation
2542 * in the rare case that the file is mmap()ped and a write
2543 * occured into the mmap()ped region just beyond the file size
2544 * and writepage has not yet been called to write out the page
2545 * (which would clear the area beyond the file size) and we now
2546 * extend the file size to incorporate this dirty region
2547 * outside the file size, a write of the page would result in
2548 * this data being written to disk instead of being cleared.
2549 * Given both POSIX and the Linux mmap(2) man page specify that
2550 * this corner case is undefined, we choose to leave it like
2551 * that as this is much simpler for us as we cannot lock the
2552 * relevant page now since we are holding too many ntfs locks
2553 * which would result in a lock reversal deadlock.
2554 */
2555 ni->initialized_size = new_size;
2556 write_unlock_irqrestore(&ni->size_lock, flags);
2557 goto unm_done;
2558 }
2559 /* If the above resize failed, this must be an attribute extension. */
2560 BUG_ON(size_change < 0);
2561 /*
2562 * We have to drop all the locks so we can call
2563 * ntfs_attr_make_non_resident(). This could be optimised by try-
2564 * locking the first page cache page and only if that fails dropping
2565 * the locks, locking the page, and redoing all the locking and
2566 * lookups. While this would be a huge optimisation, it is not worth
2567 * it as this is definitely a slow code path as it only ever can happen
2568 * once for any given file.
2569 */
2570 ntfs_attr_put_search_ctx(ctx);
2571 unmap_mft_record(base_ni);
2572 up_write(&ni->runlist.lock);
2573 /*
2574 * Not enough space in the mft record, try to make the attribute
2575 * non-resident and if successful restart the truncation process.
2576 */
2577 err = ntfs_attr_make_non_resident(ni, old_size);
2578 if (likely(!err))
2579 goto retry_truncate;
2580 /*
2581 * Could not make non-resident. If this is due to this not being
2582 * permitted for this attribute type or there not being enough space,
2583 * try to make other attributes non-resident. Otherwise fail.
2584 */
2585 if (unlikely(err != -EPERM && err != -ENOSPC)) {
2586 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2587 "type 0x%x, because the conversion from "
2588 "resident to non-resident attribute failed "
2589 "with error code %i.", vi->i_ino,
2590 (unsigned)le32_to_cpu(ni->type), err);
2591 if (err != -ENOMEM)
2592 err = -EIO;
2593 goto conv_err_out;
2594 }
2595 /* TODO: Not implemented from here, abort. */
2596 if (err == -ENOSPC)
2597 ntfs_error(vol->sb, "Not enough space in the mft record/on "
2598 "disk for the non-resident attribute value. "
2599 "This case is not implemented yet.");
2600 else /* if (err == -EPERM) */
2601 ntfs_error(vol->sb, "This attribute type may not be "
2602 "non-resident. This case is not implemented "
2603 "yet.");
2604 err = -EOPNOTSUPP;
2605 goto conv_err_out;
2606#if 0
2607 // TODO: Attempt to make other attributes non-resident.
2608 if (!err)
2609 goto do_resident_extend;
2610 /*
2611 * Both the attribute list attribute and the standard information
2612 * attribute must remain in the base inode. Thus, if this is one of
2613 * these attributes, we have to try to move other attributes out into
2614 * extent mft records instead.
2615 */
2616 if (ni->type == AT_ATTRIBUTE_LIST ||
2617 ni->type == AT_STANDARD_INFORMATION) {
2618 // TODO: Attempt to move other attributes into extent mft
2619 // records.
2620 err = -EOPNOTSUPP;
2621 if (!err)
2622 goto do_resident_extend;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002623 goto err_out;
2624 }
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002625 // TODO: Attempt to move this attribute to an extent mft record, but
2626 // only if it is not already the only attribute in an mft record in
2627 // which case there would be nothing to gain.
2628 err = -EOPNOTSUPP;
2629 if (!err)
2630 goto do_resident_extend;
2631 /* There is nothing we can do to make enough space. )-: */
2632 goto err_out;
2633#endif
2634do_non_resident_truncate:
2635 BUG_ON(!NInoNonResident(ni));
2636 if (alloc_change < 0) {
2637 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2638 if (highest_vcn > 0 &&
2639 old_alloc_size >> vol->cluster_size_bits >
2640 highest_vcn + 1) {
2641 /*
2642 * This attribute has multiple extents. Not yet
2643 * supported.
2644 */
2645 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2646 "attribute type 0x%x, because the "
2647 "attribute is highly fragmented (it "
2648 "consists of multiple extents) and "
2649 "this case is not implemented yet.",
2650 vi->i_ino,
2651 (unsigned)le32_to_cpu(ni->type));
2652 err = -EOPNOTSUPP;
2653 goto bad_out;
2654 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655 }
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002656 /*
2657 * If the size is shrinking, need to reduce the initialized_size and
2658 * the data_size before reducing the allocation.
2659 */
2660 if (size_change < 0) {
2661 /*
2662 * Make the valid size smaller (i_size is already up-to-date).
2663 */
2664 write_lock_irqsave(&ni->size_lock, flags);
2665 if (new_size < ni->initialized_size) {
2666 ni->initialized_size = new_size;
2667 a->data.non_resident.initialized_size =
2668 cpu_to_sle64(new_size);
2669 }
2670 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2671 write_unlock_irqrestore(&ni->size_lock, flags);
2672 flush_dcache_mft_record_page(ctx->ntfs_ino);
2673 mark_mft_record_dirty(ctx->ntfs_ino);
2674 /* If the allocated size is not changing, we are done. */
2675 if (!alloc_change)
2676 goto unm_done;
2677 /*
2678 * If the size is shrinking it makes no sense for the
2679 * allocation to be growing.
2680 */
2681 BUG_ON(alloc_change > 0);
2682 } else /* if (size_change >= 0) */ {
2683 /*
2684 * The file size is growing or staying the same but the
2685 * allocation can be shrinking, growing or staying the same.
2686 */
2687 if (alloc_change > 0) {
2688 /*
2689 * We need to extend the allocation and possibly update
2690 * the data size. If we are updating the data size,
2691 * since we are not touching the initialized_size we do
2692 * not need to worry about the actual data on disk.
2693 * And as far as the page cache is concerned, there
2694 * will be no pages beyond the old data size and any
2695 * partial region in the last page between the old and
2696 * new data size (or the end of the page if the new
2697 * data size is outside the page) does not need to be
2698 * modified as explained above for the resident
2699 * attribute truncate case. To do this, we simply drop
2700 * the locks we hold and leave all the work to our
2701 * friendly helper ntfs_attr_extend_allocation().
2702 */
2703 ntfs_attr_put_search_ctx(ctx);
2704 unmap_mft_record(base_ni);
2705 up_write(&ni->runlist.lock);
2706 err = ntfs_attr_extend_allocation(ni, new_size,
2707 size_change > 0 ? new_size : -1, -1);
2708 /*
2709 * ntfs_attr_extend_allocation() will have done error
2710 * output already.
2711 */
2712 goto done;
2713 }
2714 if (!alloc_change)
2715 goto alloc_done;
2716 }
2717 /* alloc_change < 0 */
2718 /* Free the clusters. */
2719 nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2720 vol->cluster_size_bits, -1, ctx);
2721 m = ctx->mrec;
2722 a = ctx->attr;
2723 if (unlikely(nr_freed < 0)) {
2724 ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2725 "%lli). Unmount and run chkdsk to recover "
2726 "the lost cluster(s).", (long long)nr_freed);
2727 NVolSetErrors(vol);
2728 nr_freed = 0;
2729 }
2730 /* Truncate the runlist. */
2731 err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2732 new_alloc_size >> vol->cluster_size_bits);
2733 /*
2734 * If the runlist truncation failed and/or the search context is no
2735 * longer valid, we cannot resize the attribute record or build the
2736 * mapping pairs array thus we mark the inode bad so that no access to
2737 * the freed clusters can happen.
2738 */
2739 if (unlikely(err || IS_ERR(m))) {
2740 ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2741 IS_ERR(m) ?
2742 "restore attribute search context" :
2743 "truncate attribute runlist",
2744 IS_ERR(m) ? PTR_ERR(m) : err, es);
2745 err = -EIO;
2746 goto bad_out;
2747 }
2748 /* Get the size for the shrunk mapping pairs array for the runlist. */
2749 mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2750 if (unlikely(mp_size <= 0)) {
2751 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2752 "attribute type 0x%x, because determining the "
2753 "size for the mapping pairs failed with error "
2754 "code %i.%s", vi->i_ino,
2755 (unsigned)le32_to_cpu(ni->type), mp_size, es);
2756 err = -EIO;
2757 goto bad_out;
2758 }
2759 /*
2760 * Shrink the attribute record for the new mapping pairs array. Note,
2761 * this cannot fail since we are making the attribute smaller thus by
2762 * definition there is enough space to do so.
2763 */
2764 attr_len = le32_to_cpu(a->length);
2765 err = ntfs_attr_record_resize(m, a, mp_size +
2766 le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2767 BUG_ON(err);
2768 /*
2769 * Generate the mapping pairs array directly into the attribute record.
2770 */
2771 err = ntfs_mapping_pairs_build(vol, (u8*)a +
2772 le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2773 mp_size, ni->runlist.rl, 0, -1, NULL);
2774 if (unlikely(err)) {
2775 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2776 "attribute type 0x%x, because building the "
2777 "mapping pairs failed with error code %i.%s",
2778 vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2779 err, es);
2780 err = -EIO;
2781 goto bad_out;
2782 }
2783 /* Update the allocated/compressed size as well as the highest vcn. */
2784 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2785 vol->cluster_size_bits) - 1);
2786 write_lock_irqsave(&ni->size_lock, flags);
2787 ni->allocated_size = new_alloc_size;
2788 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2789 if (NInoSparse(ni) || NInoCompressed(ni)) {
2790 if (nr_freed) {
2791 ni->itype.compressed.size -= nr_freed <<
2792 vol->cluster_size_bits;
2793 BUG_ON(ni->itype.compressed.size < 0);
2794 a->data.non_resident.compressed_size = cpu_to_sle64(
2795 ni->itype.compressed.size);
2796 vi->i_blocks = ni->itype.compressed.size >> 9;
2797 }
2798 } else
2799 vi->i_blocks = new_alloc_size >> 9;
2800 write_unlock_irqrestore(&ni->size_lock, flags);
2801 /*
2802 * We have shrunk the allocation. If this is a shrinking truncate we
2803 * have already dealt with the initialized_size and the data_size above
2804 * and we are done. If the truncate is only changing the allocation
2805 * and not the data_size, we are also done. If this is an extending
2806 * truncate, need to extend the data_size now which is ensured by the
2807 * fact that @size_change is positive.
2808 */
2809alloc_done:
2810 /*
2811 * If the size is growing, need to update it now. If it is shrinking,
2812 * we have already updated it above (before the allocation change).
2813 */
2814 if (size_change > 0)
2815 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2816 /* Ensure the modified mft record is written out. */
2817 flush_dcache_mft_record_page(ctx->ntfs_ino);
2818 mark_mft_record_dirty(ctx->ntfs_ino);
2819unm_done:
2820 ntfs_attr_put_search_ctx(ctx);
2821 unmap_mft_record(base_ni);
2822 up_write(&ni->runlist.lock);
2823done:
2824 /* Update the mtime and ctime on the base inode. */
Christoph Hellwig870f4812006-01-09 20:52:01 -08002825 /* normally ->truncate shouldn't update ctime or mtime,
2826 * but ntfs did before so it got a copy & paste version
2827 * of file_update_time. one day someone should fix this
2828 * for real.
2829 */
2830 if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2831 struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb);
2832 int sync_it = 0;
2833
2834 if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2835 !timespec_equal(&VFS_I(base_ni)->i_ctime, &now))
2836 sync_it = 1;
2837 VFS_I(base_ni)->i_mtime = now;
2838 VFS_I(base_ni)->i_ctime = now;
2839
2840 if (sync_it)
2841 mark_inode_dirty_sync(VFS_I(base_ni));
2842 }
2843
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002844 if (likely(!err)) {
2845 NInoClearTruncateFailed(ni);
2846 ntfs_debug("Done.");
2847 }
2848 return err;
2849old_bad_out:
2850 old_size = -1;
2851bad_out:
Anton Altaparmakova778f212006-03-23 16:18:23 +00002852 if (err != -ENOMEM && err != -EOPNOTSUPP)
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002853 NVolSetErrors(vol);
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002854 if (err != -EOPNOTSUPP)
2855 NInoSetTruncateFailed(ni);
2856 else if (old_size >= 0)
2857 i_size_write(vi, old_size);
2858err_out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859 if (ctx)
2860 ntfs_attr_put_search_ctx(ctx);
2861 if (m)
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002862 unmap_mft_record(base_ni);
2863 up_write(&ni->runlist.lock);
2864out:
2865 ntfs_debug("Failed. Returning error code %i.", err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002866 return err;
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002867conv_err_out:
Anton Altaparmakova778f212006-03-23 16:18:23 +00002868 if (err != -ENOMEM && err != -EOPNOTSUPP)
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002869 NVolSetErrors(vol);
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002870 if (err != -EOPNOTSUPP)
2871 NInoSetTruncateFailed(ni);
2872 else
2873 i_size_write(vi, old_size);
2874 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002875}
2876
2877/**
2878 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2879 * @vi: inode for which the i_size was changed
2880 *
2881 * Wrapper for ntfs_truncate() that has no return value.
2882 *
2883 * See ntfs_truncate() description above for details.
2884 */
2885void ntfs_truncate_vfs(struct inode *vi) {
2886 ntfs_truncate(vi);
2887}
2888
2889/**
2890 * ntfs_setattr - called from notify_change() when an attribute is being changed
2891 * @dentry: dentry whose attributes to change
2892 * @attr: structure describing the attributes and the changes
2893 *
2894 * We have to trap VFS attempts to truncate the file described by @dentry as
2895 * soon as possible, because we do not implement changes in i_size yet. So we
2896 * abort all i_size changes here.
2897 *
2898 * We also abort all changes of user, group, and mode as we do not implement
2899 * the NTFS ACLs yet.
2900 *
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002901 * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902 * called with ->i_alloc_sem held for writing.
2903 *
2904 * Basically this is a copy of generic notify_change() and inode_setattr()
2905 * functionality, except we intercept and abort changes in i_size.
2906 */
2907int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2908{
2909 struct inode *vi = dentry->d_inode;
2910 int err;
2911 unsigned int ia_valid = attr->ia_valid;
2912
2913 err = inode_change_ok(vi, attr);
2914 if (err)
Anton Altaparmakove9438252005-10-04 16:01:06 +01002915 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002916 /* We do not support NTFS ACLs yet. */
2917 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2918 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2919 "supported yet, ignoring.");
2920 err = -EOPNOTSUPP;
2921 goto out;
2922 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002923 if (ia_valid & ATTR_SIZE) {
2924 if (attr->ia_size != i_size_read(vi)) {
Anton Altaparmakove9438252005-10-04 16:01:06 +01002925 ntfs_inode *ni = NTFS_I(vi);
2926 /*
2927 * FIXME: For now we do not support resizing of
2928 * compressed or encrypted files yet.
2929 */
2930 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2931 ntfs_warning(vi->i_sb, "Changes in inode size "
2932 "are not supported yet for "
2933 "%s files, ignoring.",
2934 NInoCompressed(ni) ?
2935 "compressed" : "encrypted");
2936 err = -EOPNOTSUPP;
2937 } else
2938 err = vmtruncate(vi, attr->ia_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002939 if (err || ia_valid == ATTR_SIZE)
2940 goto out;
2941 } else {
2942 /*
2943 * We skipped the truncate but must still update
2944 * timestamps.
2945 */
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002946 ia_valid |= ATTR_MTIME | ATTR_CTIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002947 }
2948 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002949 if (ia_valid & ATTR_ATIME)
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002950 vi->i_atime = timespec_trunc(attr->ia_atime,
2951 vi->i_sb->s_time_gran);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002952 if (ia_valid & ATTR_MTIME)
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002953 vi->i_mtime = timespec_trunc(attr->ia_mtime,
2954 vi->i_sb->s_time_gran);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002955 if (ia_valid & ATTR_CTIME)
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002956 vi->i_ctime = timespec_trunc(attr->ia_ctime,
2957 vi->i_sb->s_time_gran);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002958 mark_inode_dirty(vi);
2959out:
2960 return err;
2961}
2962
2963/**
2964 * ntfs_write_inode - write out a dirty inode
2965 * @vi: inode to write out
2966 * @sync: if true, write out synchronously
2967 *
2968 * Write out a dirty inode to disk including any extent inodes if present.
2969 *
2970 * If @sync is true, commit the inode to disk and wait for io completion. This
2971 * is done using write_mft_record().
2972 *
2973 * If @sync is false, just schedule the write to happen but do not wait for i/o
2974 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
2975 * marking the page (and in this case mft record) dirty but we do not implement
2976 * this yet as write_mft_record() largely ignores the @sync parameter and
2977 * always performs synchronous writes.
2978 *
2979 * Return 0 on success and -errno on error.
2980 */
2981int ntfs_write_inode(struct inode *vi, int sync)
2982{
2983 sle64 nt;
2984 ntfs_inode *ni = NTFS_I(vi);
2985 ntfs_attr_search_ctx *ctx;
2986 MFT_RECORD *m;
2987 STANDARD_INFORMATION *si;
2988 int err = 0;
2989 BOOL modified = FALSE;
2990
2991 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2992 vi->i_ino);
2993 /*
2994 * Dirty attribute inodes are written via their real inodes so just
2995 * clean them here. Access time updates are taken care off when the
2996 * real inode is written.
2997 */
2998 if (NInoAttr(ni)) {
2999 NInoClearDirty(ni);
3000 ntfs_debug("Done.");
3001 return 0;
3002 }
3003 /* Map, pin, and lock the mft record belonging to the inode. */
3004 m = map_mft_record(ni);
3005 if (IS_ERR(m)) {
3006 err = PTR_ERR(m);
3007 goto err_out;
3008 }
3009 /* Update the access times in the standard information attribute. */
3010 ctx = ntfs_attr_get_search_ctx(ni, m);
3011 if (unlikely(!ctx)) {
3012 err = -ENOMEM;
3013 goto unm_err_out;
3014 }
3015 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
3016 CASE_SENSITIVE, 0, NULL, 0, ctx);
3017 if (unlikely(err)) {
3018 ntfs_attr_put_search_ctx(ctx);
3019 goto unm_err_out;
3020 }
3021 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3022 le16_to_cpu(ctx->attr->data.resident.value_offset));
3023 /* Update the access times if they have changed. */
3024 nt = utc2ntfs(vi->i_mtime);
3025 if (si->last_data_change_time != nt) {
3026 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
Randy Dunlap89075472005-03-03 11:19:53 +00003027 "new = 0x%llx", vi->i_ino, (long long)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003028 sle64_to_cpu(si->last_data_change_time),
Randy Dunlap89075472005-03-03 11:19:53 +00003029 (long long)sle64_to_cpu(nt));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003030 si->last_data_change_time = nt;
3031 modified = TRUE;
3032 }
3033 nt = utc2ntfs(vi->i_ctime);
3034 if (si->last_mft_change_time != nt) {
3035 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
Randy Dunlap89075472005-03-03 11:19:53 +00003036 "new = 0x%llx", vi->i_ino, (long long)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003037 sle64_to_cpu(si->last_mft_change_time),
Randy Dunlap89075472005-03-03 11:19:53 +00003038 (long long)sle64_to_cpu(nt));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003039 si->last_mft_change_time = nt;
3040 modified = TRUE;
3041 }
3042 nt = utc2ntfs(vi->i_atime);
3043 if (si->last_access_time != nt) {
3044 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3045 "new = 0x%llx", vi->i_ino,
Randy Dunlap89075472005-03-03 11:19:53 +00003046 (long long)sle64_to_cpu(si->last_access_time),
3047 (long long)sle64_to_cpu(nt));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003048 si->last_access_time = nt;
3049 modified = TRUE;
3050 }
3051 /*
3052 * If we just modified the standard information attribute we need to
3053 * mark the mft record it is in dirty. We do this manually so that
3054 * mark_inode_dirty() is not called which would redirty the inode and
3055 * hence result in an infinite loop of trying to write the inode.
3056 * There is no need to mark the base inode nor the base mft record
3057 * dirty, since we are going to write this mft record below in any case
3058 * and the base mft record may actually not have been modified so it
3059 * might not need to be written out.
3060 * NOTE: It is not a problem when the inode for $MFT itself is being
3061 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3062 * on the $MFT inode and hence ntfs_write_inode() will not be
3063 * re-invoked because of it which in turn is ok since the dirtied mft
3064 * record will be cleaned and written out to disk below, i.e. before
3065 * this function returns.
3066 */
Anton Altaparmakov20fdcf12006-03-23 16:21:02 +00003067 if (modified) {
3068 flush_dcache_mft_record_page(ctx->ntfs_ino);
Ingo Molnar4e5e5292006-03-23 16:57:48 +00003069 if (!NInoTestSetDirty(ctx->ntfs_ino))
Anton Altaparmakov20fdcf12006-03-23 16:21:02 +00003070 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3071 ctx->ntfs_ino->page_ofs);
3072 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003073 ntfs_attr_put_search_ctx(ctx);
3074 /* Now the access times are updated, write the base mft record. */
3075 if (NInoDirty(ni))
3076 err = write_mft_record(ni, m, sync);
3077 /* Write all attached extent mft records. */
Ingo Molnar4e5e5292006-03-23 16:57:48 +00003078 mutex_lock(&ni->extent_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003079 if (ni->nr_extents > 0) {
3080 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3081 int i;
3082
3083 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3084 for (i = 0; i < ni->nr_extents; i++) {
3085 ntfs_inode *tni = extent_nis[i];
3086
3087 if (NInoDirty(tni)) {
3088 MFT_RECORD *tm = map_mft_record(tni);
3089 int ret;
3090
3091 if (IS_ERR(tm)) {
3092 if (!err || err == -ENOMEM)
3093 err = PTR_ERR(tm);
3094 continue;
3095 }
3096 ret = write_mft_record(tni, tm, sync);
3097 unmap_mft_record(tni);
3098 if (unlikely(ret)) {
3099 if (!err || err == -ENOMEM)
3100 err = ret;
3101 }
3102 }
3103 }
3104 }
Ingo Molnar4e5e5292006-03-23 16:57:48 +00003105 mutex_unlock(&ni->extent_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003106 unmap_mft_record(ni);
3107 if (unlikely(err))
3108 goto err_out;
3109 ntfs_debug("Done.");
3110 return 0;
3111unm_err_out:
3112 unmap_mft_record(ni);
3113err_out:
3114 if (err == -ENOMEM) {
3115 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
3116 "Marking the inode dirty again, so the VFS "
3117 "retries later.");
3118 mark_inode_dirty(vi);
3119 } else {
Anton Altaparmakova778f212006-03-23 16:18:23 +00003120 ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003121 NVolSetErrors(ni->vol);
3122 }
3123 return err;
3124}
3125
3126#endif /* NTFS_RW */