blob: 247989891b4b6e45cd9f20548a91d202c1cea82e [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);
Panagiotis Issarisf52720c2006-09-27 01:49:39 -0700140 ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 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.");
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800327 ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_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.");
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800352 ni = kmem_cache_alloc(ntfs_inode_cache, GFP_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
Ingo Molnar59345372006-07-03 00:25:18 -0700370/*
371 * The attribute runlist lock has separate locking rules from the
372 * normal runlist lock, so split the two lock-classes:
373 */
374static struct lock_class_key attr_list_rl_lock_class;
375
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376/**
377 * __ntfs_init_inode - initialize ntfs specific part of an inode
378 * @sb: super block of mounted volume
379 * @ni: freshly allocated ntfs inode which to initialize
380 *
381 * Initialize an ntfs inode to defaults.
382 *
383 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
384 * untouched. Make sure to initialize them elsewhere.
385 *
386 * Return zero on success and -ENOMEM on error.
387 */
388void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
389{
390 ntfs_debug("Entering.");
Anton Altaparmakov36763672004-11-18 13:46:45 +0000391 rwlock_init(&ni->size_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 ni->initialized_size = ni->allocated_size = 0;
393 ni->seq_no = 0;
394 atomic_set(&ni->count, 1);
395 ni->vol = NTFS_SB(sb);
396 ntfs_init_runlist(&ni->runlist);
Ingo Molnar4e5e5292006-03-23 16:57:48 +0000397 mutex_init(&ni->mrec_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 ni->page = NULL;
399 ni->page_ofs = 0;
400 ni->attr_list_size = 0;
401 ni->attr_list = NULL;
402 ntfs_init_runlist(&ni->attr_list_rl);
Ingo Molnar59345372006-07-03 00:25:18 -0700403 lockdep_set_class(&ni->attr_list_rl.lock,
404 &attr_list_rl_lock_class);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 ni->itype.index.bmp_ino = NULL;
406 ni->itype.index.block_size = 0;
407 ni->itype.index.vcn_size = 0;
408 ni->itype.index.collation_rule = 0;
409 ni->itype.index.block_size_bits = 0;
410 ni->itype.index.vcn_size_bits = 0;
Ingo Molnar4e5e5292006-03-23 16:57:48 +0000411 mutex_init(&ni->extent_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 ni->nr_extents = 0;
413 ni->ext.base_ntfs_ino = NULL;
414}
415
Ingo Molnar59345372006-07-03 00:25:18 -0700416/*
417 * Extent inodes get MFT-mapped in a nested way, while the base inode
418 * is still mapped. Teach this nesting to the lock validator by creating
419 * a separate class for nested inode's mrec_lock's:
420 */
421static struct lock_class_key extent_inode_mrec_lock_key;
422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
424 unsigned long mft_no)
425{
426 ntfs_inode *ni = ntfs_alloc_extent_inode();
427
428 ntfs_debug("Entering.");
429 if (likely(ni != NULL)) {
430 __ntfs_init_inode(sb, ni);
Ingo Molnar59345372006-07-03 00:25:18 -0700431 lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 ni->mft_no = mft_no;
433 ni->type = AT_UNUSED;
434 ni->name = NULL;
435 ni->name_len = 0;
436 }
437 return ni;
438}
439
440/**
441 * ntfs_is_extended_system_file - check if a file is in the $Extend directory
442 * @ctx: initialized attribute search context
443 *
444 * Search all file name attributes in the inode described by the attribute
445 * search context @ctx and check if any of the names are in the $Extend system
446 * directory.
447 *
448 * Return values:
449 * 1: file is in $Extend directory
450 * 0: file is not in $Extend directory
451 * -errno: failed to determine if the file is in the $Extend directory
452 */
453static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
454{
455 int nr_links, err;
456
457 /* Restart search. */
458 ntfs_attr_reinit_search_ctx(ctx);
459
460 /* Get number of hard links. */
461 nr_links = le16_to_cpu(ctx->mrec->link_count);
462
463 /* Loop through all hard links. */
464 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
465 ctx))) {
466 FILE_NAME_ATTR *file_name_attr;
467 ATTR_RECORD *attr = ctx->attr;
468 u8 *p, *p2;
469
470 nr_links--;
471 /*
472 * Maximum sanity checking as we are called on an inode that
473 * we suspect might be corrupt.
474 */
475 p = (u8*)attr + le32_to_cpu(attr->length);
476 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
477 le32_to_cpu(ctx->mrec->bytes_in_use)) {
478err_corrupt_attr:
479 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
480 "attribute. You should run chkdsk.");
481 return -EIO;
482 }
483 if (attr->non_resident) {
484 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
485 "name. You should run chkdsk.");
486 return -EIO;
487 }
488 if (attr->flags) {
489 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
490 "invalid flags. You should run "
491 "chkdsk.");
492 return -EIO;
493 }
494 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
495 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
496 "name. You should run chkdsk.");
497 return -EIO;
498 }
499 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
500 le16_to_cpu(attr->data.resident.value_offset));
501 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
502 if (p2 < (u8*)attr || p2 > p)
503 goto err_corrupt_attr;
504 /* This attribute is ok, but is it in the $Extend directory? */
505 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
506 return 1; /* YES, it's an extended system file. */
507 }
508 if (unlikely(err != -ENOENT))
509 return err;
510 if (unlikely(nr_links)) {
511 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
512 "doesn't match number of name attributes. You "
513 "should run chkdsk.");
514 return -EIO;
515 }
516 return 0; /* NO, it is not an extended system file. */
517}
518
519/**
520 * ntfs_read_locked_inode - read an inode from its device
521 * @vi: inode to read
522 *
523 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
524 * described by @vi into memory from the device.
525 *
526 * The only fields in @vi that we need to/can look at when the function is
527 * called are i_sb, pointing to the mounted device's super block, and i_ino,
528 * the number of the inode to load.
529 *
530 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
531 * for reading and sets up the necessary @vi fields as well as initializing
532 * the ntfs inode.
533 *
534 * Q: What locks are held when the function is called?
535 * A: i_state has I_LOCK set, hence the inode is locked, also
536 * i_count is set to 1, so it is not going to go away
537 * i_flags is set to 0 and we have no business touching it. Only an ioctl()
538 * is allowed to write to them. We should of course be honouring them but
539 * we need to do that using the IS_* macros defined in include/linux/fs.h.
540 * In any case ntfs_read_locked_inode() has nothing to do with i_flags.
541 *
542 * Return 0 on success and -errno on error. In the error case, the inode will
543 * have had make_bad_inode() executed on it.
544 */
545static int ntfs_read_locked_inode(struct inode *vi)
546{
547 ntfs_volume *vol = NTFS_SB(vi->i_sb);
548 ntfs_inode *ni;
549 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000550 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 STANDARD_INFORMATION *si;
552 ntfs_attr_search_ctx *ctx;
553 int err = 0;
554
555 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
556
557 /* Setup the generic vfs inode parts now. */
558
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 /*
560 * This is for checking whether an inode has changed w.r.t. a file so
561 * that the file can be updated if necessary (compare with f_version).
562 */
563 vi->i_version = 1;
564
565 vi->i_uid = vol->uid;
566 vi->i_gid = vol->gid;
567 vi->i_mode = 0;
568
569 /*
570 * Initialize the ntfs specific part of @vi special casing
571 * FILE_MFT which we need to do at mount time.
572 */
573 if (vi->i_ino != FILE_MFT)
574 ntfs_init_big_inode(vi);
575 ni = NTFS_I(vi);
576
577 m = map_mft_record(ni);
578 if (IS_ERR(m)) {
579 err = PTR_ERR(m);
580 goto err_out;
581 }
582 ctx = ntfs_attr_get_search_ctx(ni, m);
583 if (!ctx) {
584 err = -ENOMEM;
585 goto unm_err_out;
586 }
587
588 if (!(m->flags & MFT_RECORD_IN_USE)) {
589 ntfs_error(vi->i_sb, "Inode is not in use!");
590 goto unm_err_out;
591 }
592 if (m->base_mft_record) {
593 ntfs_error(vi->i_sb, "Inode is an extent inode!");
594 goto unm_err_out;
595 }
596
597 /* Transfer information from mft record into vfs and ntfs inodes. */
598 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
599
600 /*
601 * FIXME: Keep in mind that link_count is two for files which have both
602 * a long file name and a short file name as separate entries, so if
603 * we are hiding short file names this will be too high. Either we need
604 * to account for the short file names by subtracting them or we need
605 * to make sure we delete files even though i_nlink is not zero which
606 * might be tricky due to vfs interactions. Need to think about this
607 * some more when implementing the unlink command.
608 */
609 vi->i_nlink = le16_to_cpu(m->link_count);
610 /*
611 * FIXME: Reparse points can have the directory bit set even though
612 * they would be S_IFLNK. Need to deal with this further below when we
613 * implement reparse points / symbolic links but it will do for now.
614 * Also if not a directory, it could be something else, rather than
615 * a regular file. But again, will do for now.
616 */
617 /* Everyone gets all permissions. */
618 vi->i_mode |= S_IRWXUGO;
619 /* If read-only, noone gets write permissions. */
620 if (IS_RDONLY(vi))
621 vi->i_mode &= ~S_IWUGO;
622 if (m->flags & MFT_RECORD_IS_DIRECTORY) {
623 vi->i_mode |= S_IFDIR;
624 /*
625 * Apply the directory permissions mask set in the mount
626 * options.
627 */
628 vi->i_mode &= ~vol->dmask;
629 /* Things break without this kludge! */
630 if (vi->i_nlink > 1)
631 vi->i_nlink = 1;
632 } else {
633 vi->i_mode |= S_IFREG;
634 /* Apply the file permissions mask set in the mount options. */
635 vi->i_mode &= ~vol->fmask;
636 }
637 /*
638 * Find the standard information attribute in the mft record. At this
639 * stage we haven't setup the attribute list stuff yet, so this could
640 * in fact fail if the standard information is in an extent record, but
641 * I don't think this actually ever happens.
642 */
643 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
644 ctx);
645 if (unlikely(err)) {
646 if (err == -ENOENT) {
647 /*
648 * TODO: We should be performing a hot fix here (if the
649 * recover mount option is set) by creating a new
650 * attribute.
651 */
652 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
653 "is missing.");
654 }
655 goto unm_err_out;
656 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000657 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 /* Get the standard information attribute value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000659 si = (STANDARD_INFORMATION*)((u8*)a +
660 le16_to_cpu(a->data.resident.value_offset));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
662 /* Transfer information from the standard information into vi. */
663 /*
664 * Note: The i_?times do not quite map perfectly onto the NTFS times,
665 * but they are close enough, and in the end it doesn't really matter
666 * that much...
667 */
668 /*
669 * mtime is the last change of the data within the file. Not changed
670 * when only metadata is changed, e.g. a rename doesn't affect mtime.
671 */
672 vi->i_mtime = ntfs2utc(si->last_data_change_time);
673 /*
674 * ctime is the last change of the metadata of the file. This obviously
675 * always changes, when mtime is changed. ctime can be changed on its
676 * own, mtime is then not changed, e.g. when a file is renamed.
677 */
678 vi->i_ctime = ntfs2utc(si->last_mft_change_time);
679 /*
680 * Last access to the data within the file. Not changed during a rename
681 * for example but changed whenever the file is written to.
682 */
683 vi->i_atime = ntfs2utc(si->last_access_time);
684
685 /* Find the attribute list attribute if present. */
686 ntfs_attr_reinit_search_ctx(ctx);
687 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
688 if (err) {
689 if (unlikely(err != -ENOENT)) {
690 ntfs_error(vi->i_sb, "Failed to lookup attribute list "
691 "attribute.");
692 goto unm_err_out;
693 }
694 } else /* if (!err) */ {
695 if (vi->i_ino == FILE_MFT)
696 goto skip_attr_list_load;
697 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
698 NInoSetAttrList(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000699 a = ctx->attr;
Anton Altaparmakov3672b632006-02-24 09:55:07 +0000700 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 ntfs_error(vi->i_sb, "Attribute list attribute is "
Anton Altaparmakov3672b632006-02-24 09:55:07 +0000702 "compressed.");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 goto unm_err_out;
704 }
Anton Altaparmakov3672b632006-02-24 09:55:07 +0000705 if (a->flags & ATTR_IS_ENCRYPTED ||
706 a->flags & ATTR_IS_SPARSE) {
707 if (a->non_resident) {
708 ntfs_error(vi->i_sb, "Non-resident attribute "
709 "list attribute is encrypted/"
710 "sparse.");
711 goto unm_err_out;
712 }
713 ntfs_warning(vi->i_sb, "Resident attribute list "
714 "attribute in inode 0x%lx is marked "
715 "encrypted/sparse which is not true. "
716 "However, Windows allows this and "
717 "chkdsk does not detect or correct it "
718 "so we will just ignore the invalid "
719 "flags and pretend they are not set.",
720 vi->i_ino);
721 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 /* Now allocate memory for the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000723 ni->attr_list_size = (u32)ntfs_attr_size(a);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
725 if (!ni->attr_list) {
726 ntfs_error(vi->i_sb, "Not enough memory to allocate "
727 "buffer for attribute list.");
728 err = -ENOMEM;
729 goto unm_err_out;
730 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000731 if (a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 NInoSetAttrListNonResident(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000733 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 ntfs_error(vi->i_sb, "Attribute list has non "
735 "zero lowest_vcn.");
736 goto unm_err_out;
737 }
738 /*
739 * Setup the runlist. No need for locking as we have
740 * exclusive access to the inode at this time.
741 */
742 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000743 a, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 if (IS_ERR(ni->attr_list_rl.rl)) {
745 err = PTR_ERR(ni->attr_list_rl.rl);
746 ni->attr_list_rl.rl = NULL;
747 ntfs_error(vi->i_sb, "Mapping pairs "
748 "decompression failed.");
749 goto unm_err_out;
750 }
751 /* Now load the attribute list. */
752 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
753 ni->attr_list, ni->attr_list_size,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000754 sle64_to_cpu(a->data.non_resident.
755 initialized_size)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 ntfs_error(vi->i_sb, "Failed to load "
757 "attribute list attribute.");
758 goto unm_err_out;
759 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000760 } else /* if (!a->non_resident) */ {
761 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
762 + le32_to_cpu(
763 a->data.resident.value_length) >
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 (u8*)ctx->mrec + vol->mft_record_size) {
765 ntfs_error(vi->i_sb, "Corrupt attribute list "
766 "in inode.");
767 goto unm_err_out;
768 }
769 /* Now copy the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000770 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
771 a->data.resident.value_offset),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 le32_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000773 a->data.resident.value_length));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 }
775 }
776skip_attr_list_load:
777 /*
778 * If an attribute list is present we now have the attribute list value
779 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
780 */
781 if (S_ISDIR(vi->i_mode)) {
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000782 loff_t bvi_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 struct inode *bvi;
784 ntfs_inode *bni;
785 INDEX_ROOT *ir;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000786 u8 *ir_end, *index_end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
788 /* It is a directory, find index root attribute. */
789 ntfs_attr_reinit_search_ctx(ctx);
790 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
791 0, NULL, 0, ctx);
792 if (unlikely(err)) {
793 if (err == -ENOENT) {
794 // FIXME: File is corrupt! Hot-fix with empty
795 // index root attribute if recovery option is
796 // set.
797 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
798 "is missing.");
799 }
800 goto unm_err_out;
801 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000802 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 /* Set up the state. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000804 if (unlikely(a->non_resident)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
806 "resident.");
807 goto unm_err_out;
808 }
809 /* Ensure the attribute name is placed before the value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000810 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
811 le16_to_cpu(a->data.resident.value_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
813 "placed after the attribute value.");
814 goto unm_err_out;
815 }
816 /*
817 * Compressed/encrypted index root just means that the newly
818 * created files in that directory should be created compressed/
819 * encrypted. However index root cannot be both compressed and
820 * encrypted.
821 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000822 if (a->flags & ATTR_COMPRESSION_MASK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 NInoSetCompressed(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000824 if (a->flags & ATTR_IS_ENCRYPTED) {
825 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 ntfs_error(vi->i_sb, "Found encrypted and "
827 "compressed attribute.");
828 goto unm_err_out;
829 }
830 NInoSetEncrypted(ni);
831 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000832 if (a->flags & ATTR_IS_SPARSE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 NInoSetSparse(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000834 ir = (INDEX_ROOT*)((u8*)a +
835 le16_to_cpu(a->data.resident.value_offset));
836 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
837 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
839 "corrupt.");
840 goto unm_err_out;
841 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000842 index_end = (u8*)&ir->index +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 le32_to_cpu(ir->index.index_length);
844 if (index_end > ir_end) {
845 ntfs_error(vi->i_sb, "Directory index is corrupt.");
846 goto unm_err_out;
847 }
848 if (ir->type != AT_FILE_NAME) {
849 ntfs_error(vi->i_sb, "Indexed attribute is not "
850 "$FILE_NAME.");
851 goto unm_err_out;
852 }
853 if (ir->collation_rule != COLLATION_FILE_NAME) {
854 ntfs_error(vi->i_sb, "Index collation rule is not "
855 "COLLATION_FILE_NAME.");
856 goto unm_err_out;
857 }
858 ni->itype.index.collation_rule = ir->collation_rule;
859 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
860 if (ni->itype.index.block_size &
861 (ni->itype.index.block_size - 1)) {
862 ntfs_error(vi->i_sb, "Index block size (%u) is not a "
863 "power of two.",
864 ni->itype.index.block_size);
865 goto unm_err_out;
866 }
867 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
868 ntfs_error(vi->i_sb, "Index block size (%u) > "
869 "PAGE_CACHE_SIZE (%ld) is not "
870 "supported. Sorry.",
871 ni->itype.index.block_size,
872 PAGE_CACHE_SIZE);
873 err = -EOPNOTSUPP;
874 goto unm_err_out;
875 }
876 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
877 ntfs_error(vi->i_sb, "Index block size (%u) < "
878 "NTFS_BLOCK_SIZE (%i) is not "
879 "supported. Sorry.",
880 ni->itype.index.block_size,
881 NTFS_BLOCK_SIZE);
882 err = -EOPNOTSUPP;
883 goto unm_err_out;
884 }
885 ni->itype.index.block_size_bits =
886 ffs(ni->itype.index.block_size) - 1;
887 /* Determine the size of a vcn in the directory index. */
888 if (vol->cluster_size <= ni->itype.index.block_size) {
889 ni->itype.index.vcn_size = vol->cluster_size;
890 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
891 } else {
892 ni->itype.index.vcn_size = vol->sector_size;
893 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
894 }
895
896 /* Setup the index allocation attribute, even if not present. */
897 NInoSetMstProtected(ni);
898 ni->type = AT_INDEX_ALLOCATION;
899 ni->name = I30;
900 ni->name_len = 4;
901
902 if (!(ir->index.flags & LARGE_INDEX)) {
903 /* No index allocation. */
904 vi->i_size = ni->initialized_size =
905 ni->allocated_size = 0;
906 /* We are done with the mft record, so we release it. */
907 ntfs_attr_put_search_ctx(ctx);
908 unmap_mft_record(ni);
909 m = NULL;
910 ctx = NULL;
911 goto skip_large_dir_stuff;
912 } /* LARGE_INDEX: Index allocation present. Setup state. */
913 NInoSetIndexAllocPresent(ni);
914 /* Find index allocation attribute. */
915 ntfs_attr_reinit_search_ctx(ctx);
916 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
917 CASE_SENSITIVE, 0, NULL, 0, ctx);
918 if (unlikely(err)) {
919 if (err == -ENOENT)
920 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
921 "attribute is not present but "
922 "$INDEX_ROOT indicated it is.");
923 else
924 ntfs_error(vi->i_sb, "Failed to lookup "
925 "$INDEX_ALLOCATION "
926 "attribute.");
927 goto unm_err_out;
928 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000929 a = ctx->attr;
930 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
932 "is resident.");
933 goto unm_err_out;
934 }
935 /*
936 * Ensure the attribute name is placed before the mapping pairs
937 * array.
938 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000939 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
940 le16_to_cpu(
941 a->data.non_resident.mapping_pairs_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
943 "is placed after the mapping pairs "
944 "array.");
945 goto unm_err_out;
946 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000947 if (a->flags & ATTR_IS_ENCRYPTED) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
949 "is encrypted.");
950 goto unm_err_out;
951 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000952 if (a->flags & ATTR_IS_SPARSE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
954 "is sparse.");
955 goto unm_err_out;
956 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000957 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
959 "is compressed.");
960 goto unm_err_out;
961 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000962 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 ntfs_error(vi->i_sb, "First extent of "
964 "$INDEX_ALLOCATION attribute has non "
965 "zero lowest_vcn.");
966 goto unm_err_out;
967 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000968 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000970 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +0000972 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 /*
974 * We are done with the mft record, so we release it. Otherwise
975 * we would deadlock in ntfs_attr_iget().
976 */
977 ntfs_attr_put_search_ctx(ctx);
978 unmap_mft_record(ni);
979 m = NULL;
980 ctx = NULL;
981 /* Get the index bitmap attribute inode. */
982 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
983 if (IS_ERR(bvi)) {
984 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
985 err = PTR_ERR(bvi);
986 goto unm_err_out;
987 }
988 ni->itype.index.bmp_ino = bvi;
989 bni = NTFS_I(bvi);
990 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
991 NInoSparse(bni)) {
992 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
993 "and/or encrypted and/or sparse.");
994 goto unm_err_out;
995 }
996 /* Consistency check bitmap size vs. index allocation size. */
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +0000997 bvi_size = i_size_read(bvi);
998 if ((bvi_size << 3) < (vi->i_size >>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 ni->itype.index.block_size_bits)) {
1000 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
1001 "for index allocation (0x%llx).",
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +00001002 bvi_size << 3, vi->i_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 goto unm_err_out;
1004 }
1005skip_large_dir_stuff:
1006 /* Setup the operations for this inode. */
1007 vi->i_op = &ntfs_dir_inode_ops;
1008 vi->i_fop = &ntfs_dir_ops;
1009 } else {
1010 /* It is a file. */
1011 ntfs_attr_reinit_search_ctx(ctx);
1012
1013 /* Setup the data attribute, even if not present. */
1014 ni->type = AT_DATA;
1015 ni->name = NULL;
1016 ni->name_len = 0;
1017
1018 /* Find first extent of the unnamed data attribute. */
1019 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1020 if (unlikely(err)) {
1021 vi->i_size = ni->initialized_size =
1022 ni->allocated_size = 0;
1023 if (err != -ENOENT) {
1024 ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1025 "attribute.");
1026 goto unm_err_out;
1027 }
1028 /*
1029 * FILE_Secure does not have an unnamed $DATA
1030 * attribute, so we special case it here.
1031 */
1032 if (vi->i_ino == FILE_Secure)
1033 goto no_data_attr_special_case;
1034 /*
1035 * Most if not all the system files in the $Extend
1036 * system directory do not have unnamed data
1037 * attributes so we need to check if the parent
1038 * directory of the file is FILE_Extend and if it is
1039 * ignore this error. To do this we need to get the
1040 * name of this inode from the mft record as the name
1041 * contains the back reference to the parent directory.
1042 */
1043 if (ntfs_is_extended_system_file(ctx) > 0)
1044 goto no_data_attr_special_case;
1045 // FIXME: File is corrupt! Hot-fix with empty data
1046 // attribute if recovery option is set.
1047 ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1048 goto unm_err_out;
1049 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001050 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 /* Setup the state. */
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001052 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1053 if (a->flags & ATTR_COMPRESSION_MASK) {
1054 NInoSetCompressed(ni);
1055 if (vol->cluster_size > 4096) {
1056 ntfs_error(vi->i_sb, "Found "
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001057 "compressed data but "
1058 "compression is "
1059 "disabled due to "
1060 "cluster size (%i) > "
1061 "4kiB.",
1062 vol->cluster_size);
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001063 goto unm_err_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 }
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001065 if ((a->flags & ATTR_COMPRESSION_MASK)
1066 != ATTR_IS_COMPRESSED) {
1067 ntfs_error(vi->i_sb, "Found unknown "
1068 "compression method "
1069 "or corrupt file.");
1070 goto unm_err_out;
1071 }
1072 }
1073 if (a->flags & ATTR_IS_SPARSE)
1074 NInoSetSparse(ni);
1075 }
1076 if (a->flags & ATTR_IS_ENCRYPTED) {
1077 if (NInoCompressed(ni)) {
1078 ntfs_error(vi->i_sb, "Found encrypted and "
1079 "compressed data.");
1080 goto unm_err_out;
1081 }
1082 NInoSetEncrypted(ni);
1083 }
1084 if (a->non_resident) {
1085 NInoSetNonResident(ni);
1086 if (NInoCompressed(ni) || NInoSparse(ni)) {
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001087 if (NInoCompressed(ni) && a->data.non_resident.
1088 compression_unit != 4) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 ntfs_error(vi->i_sb, "Found "
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001090 "non-standard "
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001091 "compression unit (%u "
1092 "instead of 4). "
1093 "Cannot handle this.",
1094 a->data.non_resident.
1095 compression_unit);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 err = -EOPNOTSUPP;
1097 goto unm_err_out;
1098 }
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001099 if (a->data.non_resident.compression_unit) {
1100 ni->itype.compressed.block_size = 1U <<
1101 (a->data.non_resident.
1102 compression_unit +
1103 vol->cluster_size_bits);
1104 ni->itype.compressed.block_size_bits =
1105 ffs(ni->itype.
1106 compressed.
1107 block_size) - 1;
1108 ni->itype.compressed.block_clusters =
1109 1U << a->data.
1110 non_resident.
1111 compression_unit;
1112 } else {
1113 ni->itype.compressed.block_size = 0;
1114 ni->itype.compressed.block_size_bits =
1115 0;
1116 ni->itype.compressed.block_clusters =
1117 0;
1118 }
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001119 ni->itype.compressed.size = sle64_to_cpu(
1120 a->data.non_resident.
1121 compressed_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001123 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 ntfs_error(vi->i_sb, "First extent of $DATA "
1125 "attribute has non zero "
1126 "lowest_vcn.");
1127 goto unm_err_out;
1128 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 vi->i_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001130 a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001132 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001134 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 } else { /* Resident attribute. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001136 vi->i_size = ni->initialized_size = le32_to_cpu(
1137 a->data.resident.value_length);
1138 ni->allocated_size = le32_to_cpu(a->length) -
1139 le16_to_cpu(
1140 a->data.resident.value_offset);
1141 if (vi->i_size > ni->allocated_size) {
1142 ntfs_error(vi->i_sb, "Resident data attribute "
1143 "is corrupt (size exceeds "
1144 "allocation).");
1145 goto unm_err_out;
1146 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 }
1148no_data_attr_special_case:
1149 /* We are done with the mft record, so we release it. */
1150 ntfs_attr_put_search_ctx(ctx);
1151 unmap_mft_record(ni);
1152 m = NULL;
1153 ctx = NULL;
1154 /* Setup the operations for this inode. */
1155 vi->i_op = &ntfs_file_inode_ops;
1156 vi->i_fop = &ntfs_file_ops;
1157 }
1158 if (NInoMstProtected(ni))
1159 vi->i_mapping->a_ops = &ntfs_mst_aops;
1160 else
1161 vi->i_mapping->a_ops = &ntfs_aops;
1162 /*
1163 * The number of 512-byte blocks used on disk (for stat). This is in so
1164 * far inaccurate as it doesn't account for any named streams or other
1165 * special non-resident attributes, but that is how Windows works, too,
1166 * so we are at least consistent with Windows, if not entirely
1167 * consistent with the Linux Way. Doing it the Linux Way would cause a
1168 * significant slowdown as it would involve iterating over all
1169 * attributes in the mft record and adding the allocated/compressed
1170 * sizes of all non-resident attributes present to give us the Linux
1171 * correct size that should go into i_blocks (after division by 512).
1172 */
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001173 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 vi->i_blocks = ni->itype.compressed.size >> 9;
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001175 else
1176 vi->i_blocks = ni->allocated_size >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 ntfs_debug("Done.");
1178 return 0;
1179
1180unm_err_out:
1181 if (!err)
1182 err = -EIO;
1183 if (ctx)
1184 ntfs_attr_put_search_ctx(ctx);
1185 if (m)
1186 unmap_mft_record(ni);
1187err_out:
1188 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt "
1189 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino);
1190 make_bad_inode(vi);
1191 if (err != -EOPNOTSUPP && err != -ENOMEM)
1192 NVolSetErrors(vol);
1193 return err;
1194}
1195
1196/**
1197 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1198 * @base_vi: base inode
1199 * @vi: attribute inode to read
1200 *
1201 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1202 * attribute inode described by @vi into memory from the base mft record
1203 * described by @base_ni.
1204 *
1205 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1206 * reading and looks up the attribute described by @vi before setting up the
1207 * necessary fields in @vi as well as initializing the ntfs inode.
1208 *
1209 * Q: What locks are held when the function is called?
1210 * A: i_state has I_LOCK set, hence the inode is locked, also
1211 * i_count is set to 1, so it is not going to go away
1212 *
1213 * Return 0 on success and -errno on error. In the error case, the inode will
1214 * have had make_bad_inode() executed on it.
Anton Altaparmakovf6098cf2005-09-19 09:41:39 +01001215 *
1216 * Note this cannot be called for AT_INDEX_ALLOCATION.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 */
1218static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1219{
1220 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1221 ntfs_inode *ni, *base_ni;
1222 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001223 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 ntfs_attr_search_ctx *ctx;
1225 int err = 0;
1226
1227 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1228
1229 ntfs_init_big_inode(vi);
1230
1231 ni = NTFS_I(vi);
1232 base_ni = NTFS_I(base_vi);
1233
1234 /* Just mirror the values from the base inode. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 vi->i_version = base_vi->i_version;
1236 vi->i_uid = base_vi->i_uid;
1237 vi->i_gid = base_vi->i_gid;
1238 vi->i_nlink = base_vi->i_nlink;
1239 vi->i_mtime = base_vi->i_mtime;
1240 vi->i_ctime = base_vi->i_ctime;
1241 vi->i_atime = base_vi->i_atime;
1242 vi->i_generation = ni->seq_no = base_ni->seq_no;
1243
1244 /* Set inode type to zero but preserve permissions. */
1245 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1246
1247 m = map_mft_record(base_ni);
1248 if (IS_ERR(m)) {
1249 err = PTR_ERR(m);
1250 goto err_out;
1251 }
1252 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1253 if (!ctx) {
1254 err = -ENOMEM;
1255 goto unm_err_out;
1256 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 /* Find the attribute. */
1258 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1259 CASE_SENSITIVE, 0, NULL, 0, ctx);
1260 if (unlikely(err))
1261 goto unm_err_out;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001262 a = ctx->attr;
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001263 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1264 if (a->flags & ATTR_COMPRESSION_MASK) {
1265 NInoSetCompressed(ni);
1266 if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1267 ni->name_len)) {
1268 ntfs_error(vi->i_sb, "Found compressed "
1269 "non-data or named data "
1270 "attribute. Please report "
1271 "you saw this message to "
1272 "linux-ntfs-dev@lists."
1273 "sourceforge.net");
1274 goto unm_err_out;
1275 }
1276 if (vol->cluster_size > 4096) {
1277 ntfs_error(vi->i_sb, "Found compressed "
1278 "attribute but compression is "
1279 "disabled due to cluster size "
1280 "(%i) > 4kiB.",
1281 vol->cluster_size);
1282 goto unm_err_out;
1283 }
1284 if ((a->flags & ATTR_COMPRESSION_MASK) !=
1285 ATTR_IS_COMPRESSED) {
1286 ntfs_error(vi->i_sb, "Found unknown "
1287 "compression method.");
1288 goto unm_err_out;
1289 }
1290 }
1291 /*
Anton Altaparmakovf6098cf2005-09-19 09:41:39 +01001292 * The compressed/sparse flag set in an index root just means
1293 * to compress all files.
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001294 */
1295 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1296 ntfs_error(vi->i_sb, "Found mst protected attribute "
1297 "but the attribute is %s. Please "
1298 "report you saw this message to "
1299 "linux-ntfs-dev@lists.sourceforge.net",
1300 NInoCompressed(ni) ? "compressed" :
1301 "sparse");
1302 goto unm_err_out;
1303 }
1304 if (a->flags & ATTR_IS_SPARSE)
1305 NInoSetSparse(ni);
1306 }
1307 if (a->flags & ATTR_IS_ENCRYPTED) {
1308 if (NInoCompressed(ni)) {
1309 ntfs_error(vi->i_sb, "Found encrypted and compressed "
1310 "data.");
1311 goto unm_err_out;
1312 }
1313 /*
1314 * The encryption flag set in an index root just means to
1315 * encrypt all files.
1316 */
1317 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1318 ntfs_error(vi->i_sb, "Found mst protected attribute "
1319 "but the attribute is encrypted. "
1320 "Please report you saw this message "
1321 "to linux-ntfs-dev@lists.sourceforge."
1322 "net");
1323 goto unm_err_out;
1324 }
1325 if (ni->type != AT_DATA) {
1326 ntfs_error(vi->i_sb, "Found encrypted non-data "
1327 "attribute.");
1328 goto unm_err_out;
1329 }
1330 NInoSetEncrypted(ni);
1331 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001332 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 /* Ensure the attribute name is placed before the value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001334 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1335 le16_to_cpu(a->data.resident.value_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 ntfs_error(vol->sb, "Attribute name is placed after "
1337 "the attribute value.");
1338 goto unm_err_out;
1339 }
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001340 if (NInoMstProtected(ni)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 ntfs_error(vi->i_sb, "Found mst protected attribute "
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001342 "but the attribute is resident. "
1343 "Please report you saw this message to "
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 "linux-ntfs-dev@lists.sourceforge.net");
1345 goto unm_err_out;
1346 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001347 vi->i_size = ni->initialized_size = le32_to_cpu(
1348 a->data.resident.value_length);
1349 ni->allocated_size = le32_to_cpu(a->length) -
1350 le16_to_cpu(a->data.resident.value_offset);
1351 if (vi->i_size > ni->allocated_size) {
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001352 ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1353 "(size exceeds allocation).");
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001354 goto unm_err_out;
1355 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 } else {
1357 NInoSetNonResident(ni);
1358 /*
1359 * Ensure the attribute name is placed before the mapping pairs
1360 * array.
1361 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001362 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1363 le16_to_cpu(
1364 a->data.non_resident.mapping_pairs_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 ntfs_error(vol->sb, "Attribute name is placed after "
1366 "the mapping pairs array.");
1367 goto unm_err_out;
1368 }
Anton Altaparmakovf6098cf2005-09-19 09:41:39 +01001369 if (NInoCompressed(ni) || NInoSparse(ni)) {
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001370 if (NInoCompressed(ni) && a->data.non_resident.
1371 compression_unit != 4) {
1372 ntfs_error(vi->i_sb, "Found non-standard "
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 "compression unit (%u instead "
1374 "of 4). Cannot handle this.",
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001375 a->data.non_resident.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 compression_unit);
1377 err = -EOPNOTSUPP;
1378 goto unm_err_out;
1379 }
Anton Altaparmakova0646a12006-03-23 15:53:03 +00001380 if (a->data.non_resident.compression_unit) {
1381 ni->itype.compressed.block_size = 1U <<
1382 (a->data.non_resident.
1383 compression_unit +
1384 vol->cluster_size_bits);
1385 ni->itype.compressed.block_size_bits =
1386 ffs(ni->itype.compressed.
1387 block_size) - 1;
1388 ni->itype.compressed.block_clusters = 1U <<
1389 a->data.non_resident.
1390 compression_unit;
1391 } else {
1392 ni->itype.compressed.block_size = 0;
1393 ni->itype.compressed.block_size_bits = 0;
1394 ni->itype.compressed.block_clusters = 0;
1395 }
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001396 ni->itype.compressed.size = sle64_to_cpu(
1397 a->data.non_resident.compressed_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001399 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 ntfs_error(vi->i_sb, "First extent of attribute has "
1401 "non-zero lowest_vcn.");
1402 goto unm_err_out;
1403 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001404 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001406 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001408 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 /* Setup the operations for this attribute inode. */
1411 vi->i_op = NULL;
1412 vi->i_fop = NULL;
1413 if (NInoMstProtected(ni))
1414 vi->i_mapping->a_ops = &ntfs_mst_aops;
1415 else
1416 vi->i_mapping->a_ops = &ntfs_aops;
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001417 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 vi->i_blocks = ni->itype.compressed.size >> 9;
Anton Altaparmakov9451f852005-03-03 14:43:43 +00001419 else
1420 vi->i_blocks = ni->allocated_size >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 /*
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001422 * Make sure the base inode does not go away and attach it to the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 * attribute inode.
1424 */
1425 igrab(base_vi);
1426 ni->ext.base_ntfs_ino = base_ni;
1427 ni->nr_extents = -1;
1428
1429 ntfs_attr_put_search_ctx(ctx);
1430 unmap_mft_record(base_ni);
1431
1432 ntfs_debug("Done.");
1433 return 0;
1434
1435unm_err_out:
1436 if (!err)
1437 err = -EIO;
1438 if (ctx)
1439 ntfs_attr_put_search_ctx(ctx);
1440 unmap_mft_record(base_ni);
1441err_out:
1442 ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1443 "inode (mft_no 0x%lx, type 0x%x, name_len %i). "
1444 "Marking corrupt inode and base inode 0x%lx as bad. "
1445 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1446 base_vi->i_ino);
1447 make_bad_inode(vi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 if (err != -ENOMEM)
1449 NVolSetErrors(vol);
1450 return err;
1451}
1452
1453/**
1454 * ntfs_read_locked_index_inode - read an index inode from its base inode
1455 * @base_vi: base inode
1456 * @vi: index inode to read
1457 *
1458 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1459 * index inode described by @vi into memory from the base mft record described
1460 * by @base_ni.
1461 *
1462 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1463 * reading and looks up the attributes relating to the index described by @vi
1464 * before setting up the necessary fields in @vi as well as initializing the
1465 * ntfs inode.
1466 *
1467 * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1468 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they
1469 * are setup like directory inodes since directories are a special case of
1470 * indices ao they need to be treated in much the same way. Most importantly,
1471 * for small indices the index allocation attribute might not actually exist.
1472 * However, the index root attribute always exists but this does not need to
1473 * have an inode associated with it and this is why we define a new inode type
1474 * index. Also, like for directories, we need to have an attribute inode for
1475 * the bitmap attribute corresponding to the index allocation attribute and we
1476 * can store this in the appropriate field of the inode, just like we do for
1477 * normal directory inodes.
1478 *
1479 * Q: What locks are held when the function is called?
1480 * A: i_state has I_LOCK set, hence the inode is locked, also
1481 * i_count is set to 1, so it is not going to go away
1482 *
1483 * Return 0 on success and -errno on error. In the error case, the inode will
1484 * have had make_bad_inode() executed on it.
1485 */
1486static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1487{
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +00001488 loff_t bvi_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 ntfs_volume *vol = NTFS_SB(vi->i_sb);
1490 ntfs_inode *ni, *base_ni, *bni;
1491 struct inode *bvi;
1492 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001493 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 ntfs_attr_search_ctx *ctx;
1495 INDEX_ROOT *ir;
1496 u8 *ir_end, *index_end;
1497 int err = 0;
1498
1499 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1500 ntfs_init_big_inode(vi);
1501 ni = NTFS_I(vi);
1502 base_ni = NTFS_I(base_vi);
1503 /* Just mirror the values from the base inode. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 vi->i_version = base_vi->i_version;
1505 vi->i_uid = base_vi->i_uid;
1506 vi->i_gid = base_vi->i_gid;
1507 vi->i_nlink = base_vi->i_nlink;
1508 vi->i_mtime = base_vi->i_mtime;
1509 vi->i_ctime = base_vi->i_ctime;
1510 vi->i_atime = base_vi->i_atime;
1511 vi->i_generation = ni->seq_no = base_ni->seq_no;
1512 /* Set inode type to zero but preserve permissions. */
1513 vi->i_mode = base_vi->i_mode & ~S_IFMT;
1514 /* Map the mft record for the base inode. */
1515 m = map_mft_record(base_ni);
1516 if (IS_ERR(m)) {
1517 err = PTR_ERR(m);
1518 goto err_out;
1519 }
1520 ctx = ntfs_attr_get_search_ctx(base_ni, m);
1521 if (!ctx) {
1522 err = -ENOMEM;
1523 goto unm_err_out;
1524 }
1525 /* Find the index root attribute. */
1526 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1527 CASE_SENSITIVE, 0, NULL, 0, ctx);
1528 if (unlikely(err)) {
1529 if (err == -ENOENT)
1530 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1531 "missing.");
1532 goto unm_err_out;
1533 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001534 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 /* Set up the state. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001536 if (unlikely(a->non_resident)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1538 goto unm_err_out;
1539 }
1540 /* Ensure the attribute name is placed before the value. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001541 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1542 le16_to_cpu(a->data.resident.value_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1544 "after the attribute value.");
1545 goto unm_err_out;
1546 }
Anton Altaparmakov67bb1032005-09-08 21:19:45 +01001547 /*
1548 * Compressed/encrypted/sparse index root is not allowed, except for
1549 * directories of course but those are not dealt with here.
1550 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001551 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 ATTR_IS_SPARSE)) {
1553 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1554 "root attribute.");
1555 goto unm_err_out;
1556 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001557 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1558 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1560 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1561 goto unm_err_out;
1562 }
1563 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1564 if (index_end > ir_end) {
1565 ntfs_error(vi->i_sb, "Index is corrupt.");
1566 goto unm_err_out;
1567 }
1568 if (ir->type) {
1569 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1570 le32_to_cpu(ir->type));
1571 goto unm_err_out;
1572 }
1573 ni->itype.index.collation_rule = ir->collation_rule;
1574 ntfs_debug("Index collation rule is 0x%x.",
1575 le32_to_cpu(ir->collation_rule));
1576 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1577 if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) {
1578 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1579 "two.", ni->itype.index.block_size);
1580 goto unm_err_out;
1581 }
1582 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1583 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1584 "(%ld) is not supported. Sorry.",
1585 ni->itype.index.block_size, PAGE_CACHE_SIZE);
1586 err = -EOPNOTSUPP;
1587 goto unm_err_out;
1588 }
1589 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1590 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1591 "(%i) is not supported. Sorry.",
1592 ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1593 err = -EOPNOTSUPP;
1594 goto unm_err_out;
1595 }
1596 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1597 /* Determine the size of a vcn in the index. */
1598 if (vol->cluster_size <= ni->itype.index.block_size) {
1599 ni->itype.index.vcn_size = vol->cluster_size;
1600 ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1601 } else {
1602 ni->itype.index.vcn_size = vol->sector_size;
1603 ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1604 }
1605 /* Check for presence of index allocation attribute. */
1606 if (!(ir->index.flags & LARGE_INDEX)) {
1607 /* No index allocation. */
1608 vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1609 /* We are done with the mft record, so we release it. */
1610 ntfs_attr_put_search_ctx(ctx);
1611 unmap_mft_record(base_ni);
1612 m = NULL;
1613 ctx = NULL;
1614 goto skip_large_index_stuff;
1615 } /* LARGE_INDEX: Index allocation present. Setup state. */
1616 NInoSetIndexAllocPresent(ni);
1617 /* Find index allocation attribute. */
1618 ntfs_attr_reinit_search_ctx(ctx);
1619 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1620 CASE_SENSITIVE, 0, NULL, 0, ctx);
1621 if (unlikely(err)) {
1622 if (err == -ENOENT)
1623 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1624 "not present but $INDEX_ROOT "
1625 "indicated it is.");
1626 else
1627 ntfs_error(vi->i_sb, "Failed to lookup "
1628 "$INDEX_ALLOCATION attribute.");
1629 goto unm_err_out;
1630 }
Anton Altaparmakova778f212006-03-23 16:18:23 +00001631 a = ctx->attr;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001632 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1634 "resident.");
1635 goto unm_err_out;
1636 }
1637 /*
1638 * Ensure the attribute name is placed before the mapping pairs array.
1639 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001640 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1641 le16_to_cpu(
1642 a->data.non_resident.mapping_pairs_offset)))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1644 "placed after the mapping pairs array.");
1645 goto unm_err_out;
1646 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001647 if (a->flags & ATTR_IS_ENCRYPTED) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1649 "encrypted.");
1650 goto unm_err_out;
1651 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001652 if (a->flags & ATTR_IS_SPARSE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1654 goto unm_err_out;
1655 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001656 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1658 "compressed.");
1659 goto unm_err_out;
1660 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001661 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1663 "attribute has non zero lowest_vcn.");
1664 goto unm_err_out;
1665 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001666 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 ni->initialized_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001668 a->data.non_resident.initialized_size);
1669 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 /*
1671 * We are done with the mft record, so we release it. Otherwise
1672 * we would deadlock in ntfs_attr_iget().
1673 */
1674 ntfs_attr_put_search_ctx(ctx);
1675 unmap_mft_record(base_ni);
1676 m = NULL;
1677 ctx = NULL;
1678 /* Get the index bitmap attribute inode. */
1679 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1680 if (IS_ERR(bvi)) {
1681 ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1682 err = PTR_ERR(bvi);
1683 goto unm_err_out;
1684 }
1685 bni = NTFS_I(bvi);
1686 if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1687 NInoSparse(bni)) {
1688 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1689 "encrypted and/or sparse.");
1690 goto iput_unm_err_out;
1691 }
1692 /* Consistency check bitmap size vs. index allocation size. */
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +00001693 bvi_size = i_size_read(bvi);
1694 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
Anton Altaparmakovf50f3ac2004-11-19 22:16:00 +00001696 "index allocation (0x%llx).", bvi_size << 3,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697 vi->i_size);
1698 goto iput_unm_err_out;
1699 }
1700 ni->itype.index.bmp_ino = bvi;
1701skip_large_index_stuff:
1702 /* Setup the operations for this index inode. */
1703 vi->i_op = NULL;
1704 vi->i_fop = NULL;
1705 vi->i_mapping->a_ops = &ntfs_mst_aops;
1706 vi->i_blocks = ni->allocated_size >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 /*
1708 * Make sure the base inode doesn't go away and attach it to the
1709 * index inode.
1710 */
1711 igrab(base_vi);
1712 ni->ext.base_ntfs_ino = base_ni;
1713 ni->nr_extents = -1;
1714
1715 ntfs_debug("Done.");
1716 return 0;
1717
1718iput_unm_err_out:
1719 iput(bvi);
1720unm_err_out:
1721 if (!err)
1722 err = -EIO;
1723 if (ctx)
1724 ntfs_attr_put_search_ctx(ctx);
1725 if (m)
1726 unmap_mft_record(base_ni);
1727err_out:
1728 ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1729 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1730 ni->name_len);
1731 make_bad_inode(vi);
1732 if (err != -EOPNOTSUPP && err != -ENOMEM)
1733 NVolSetErrors(vol);
1734 return err;
1735}
1736
Ingo Molnar59345372006-07-03 00:25:18 -07001737/*
1738 * The MFT inode has special locking, so teach the lock validator
1739 * about this by splitting off the locking rules of the MFT from
1740 * the locking rules of other inodes. The MFT inode can never be
1741 * accessed from the VFS side (or even internally), only by the
1742 * map_mft functions.
1743 */
1744static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
1745
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746/**
1747 * ntfs_read_inode_mount - special read_inode for mount time use only
1748 * @vi: inode to read
1749 *
1750 * Read inode FILE_MFT at mount time, only called with super_block lock
1751 * held from within the read_super() code path.
1752 *
1753 * This function exists because when it is called the page cache for $MFT/$DATA
1754 * is not initialized and hence we cannot get at the contents of mft records
1755 * by calling map_mft_record*().
1756 *
1757 * Further it needs to cope with the circular references problem, i.e. cannot
1758 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1759 * we do not know where the other extent mft records are yet and again, because
1760 * we cannot call map_mft_record*() yet. Obviously this applies only when an
1761 * attribute list is actually present in $MFT inode.
1762 *
1763 * We solve these problems by starting with the $DATA attribute before anything
1764 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each
1765 * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1766 * ntfs_runlists_merge(). Each step of the iteration necessarily provides
1767 * sufficient information for the next step to complete.
1768 *
1769 * This should work but there are two possible pit falls (see inline comments
1770 * below), but only time will tell if they are real pits or just smoke...
1771 */
1772int ntfs_read_inode_mount(struct inode *vi)
1773{
1774 VCN next_vcn, last_vcn, highest_vcn;
1775 s64 block;
1776 struct super_block *sb = vi->i_sb;
1777 ntfs_volume *vol = NTFS_SB(sb);
1778 struct buffer_head *bh;
1779 ntfs_inode *ni;
1780 MFT_RECORD *m = NULL;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001781 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 ntfs_attr_search_ctx *ctx;
1783 unsigned int i, nr_blocks;
1784 int err;
1785
1786 ntfs_debug("Entering.");
1787
1788 /* Initialize the ntfs specific part of @vi. */
1789 ntfs_init_big_inode(vi);
1790
1791 ni = NTFS_I(vi);
1792
1793 /* Setup the data attribute. It is special as it is mst protected. */
1794 NInoSetNonResident(ni);
1795 NInoSetMstProtected(ni);
Anton Altaparmakovc002f422005-02-03 12:02:56 +00001796 NInoSetSparseDisabled(ni);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797 ni->type = AT_DATA;
1798 ni->name = NULL;
1799 ni->name_len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 /*
1801 * This sets up our little cheat allowing us to reuse the async read io
1802 * completion handler for directories.
1803 */
1804 ni->itype.index.block_size = vol->mft_record_size;
1805 ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1806
1807 /* Very important! Needed to be able to call map_mft_record*(). */
1808 vol->mft_ino = vi;
1809
1810 /* Allocate enough memory to read the first mft record. */
1811 if (vol->mft_record_size > 64 * 1024) {
1812 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1813 vol->mft_record_size);
1814 goto err_out;
1815 }
1816 i = vol->mft_record_size;
1817 if (i < sb->s_blocksize)
1818 i = sb->s_blocksize;
1819 m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1820 if (!m) {
1821 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1822 goto err_out;
1823 }
1824
1825 /* Determine the first block of the $MFT/$DATA attribute. */
1826 block = vol->mft_lcn << vol->cluster_size_bits >>
1827 sb->s_blocksize_bits;
1828 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1829 if (!nr_blocks)
1830 nr_blocks = 1;
1831
1832 /* Load $MFT/$DATA's first mft record. */
1833 for (i = 0; i < nr_blocks; i++) {
1834 bh = sb_bread(sb, block++);
1835 if (!bh) {
1836 ntfs_error(sb, "Device read failed.");
1837 goto err_out;
1838 }
1839 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1840 sb->s_blocksize);
1841 brelse(bh);
1842 }
1843
1844 /* Apply the mst fixups. */
1845 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1846 /* FIXME: Try to use the $MFTMirr now. */
1847 ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1848 goto err_out;
1849 }
1850
1851 /* Need this to sanity check attribute list references to $MFT. */
1852 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1853
1854 /* Provides readpage() and sync_page() for map_mft_record(). */
1855 vi->i_mapping->a_ops = &ntfs_mst_aops;
1856
1857 ctx = ntfs_attr_get_search_ctx(ni, m);
1858 if (!ctx) {
1859 err = -ENOMEM;
1860 goto err_out;
1861 }
1862
1863 /* Find the attribute list attribute if present. */
1864 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1865 if (err) {
1866 if (unlikely(err != -ENOENT)) {
1867 ntfs_error(sb, "Failed to lookup attribute list "
1868 "attribute. You should run chkdsk.");
1869 goto put_err_out;
1870 }
1871 } else /* if (!err) */ {
1872 ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1873 u8 *al_end;
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001874 static const char *es = " Not allowed. $MFT is corrupt. "
1875 "You should run chkdsk.";
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876
1877 ntfs_debug("Attribute list attribute found in $MFT.");
1878 NInoSetAttrList(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001879 a = ctx->attr;
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001880 if (a->flags & ATTR_COMPRESSION_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 ntfs_error(sb, "Attribute list attribute is "
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001882 "compressed.%s", es);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 goto put_err_out;
1884 }
Anton Altaparmakov3672b632006-02-24 09:55:07 +00001885 if (a->flags & ATTR_IS_ENCRYPTED ||
1886 a->flags & ATTR_IS_SPARSE) {
1887 if (a->non_resident) {
1888 ntfs_error(sb, "Non-resident attribute list "
1889 "attribute is encrypted/"
1890 "sparse.%s", es);
1891 goto put_err_out;
1892 }
1893 ntfs_warning(sb, "Resident attribute list attribute "
1894 "in $MFT system file is marked "
1895 "encrypted/sparse which is not true. "
1896 "However, Windows allows this and "
1897 "chkdsk does not detect or correct it "
1898 "so we will just ignore the invalid "
1899 "flags and pretend they are not set.");
1900 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 /* Now allocate memory for the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001902 ni->attr_list_size = (u32)ntfs_attr_size(a);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1904 if (!ni->attr_list) {
1905 ntfs_error(sb, "Not enough memory to allocate buffer "
1906 "for attribute list.");
1907 goto put_err_out;
1908 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001909 if (a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 NInoSetAttrListNonResident(ni);
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001911 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 ntfs_error(sb, "Attribute list has non zero "
1913 "lowest_vcn. $MFT is corrupt. "
1914 "You should run chkdsk.");
1915 goto put_err_out;
1916 }
1917 /* Setup the runlist. */
1918 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001919 a, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 if (IS_ERR(ni->attr_list_rl.rl)) {
1921 err = PTR_ERR(ni->attr_list_rl.rl);
1922 ni->attr_list_rl.rl = NULL;
1923 ntfs_error(sb, "Mapping pairs decompression "
1924 "failed with error code %i.",
1925 -err);
1926 goto put_err_out;
1927 }
1928 /* Now load the attribute list. */
1929 if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1930 ni->attr_list, ni->attr_list_size,
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001931 sle64_to_cpu(a->data.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 non_resident.initialized_size)))) {
1933 ntfs_error(sb, "Failed to load attribute list "
1934 "attribute with error code %i.",
1935 -err);
1936 goto put_err_out;
1937 }
1938 } else /* if (!ctx.attr->non_resident) */ {
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001939 if ((u8*)a + le16_to_cpu(
1940 a->data.resident.value_offset) +
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 le32_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001942 a->data.resident.value_length) >
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 (u8*)ctx->mrec + vol->mft_record_size) {
1944 ntfs_error(sb, "Corrupt attribute list "
1945 "attribute.");
1946 goto put_err_out;
1947 }
1948 /* Now copy the attribute list. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001949 memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1950 a->data.resident.value_offset),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 le32_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00001952 a->data.resident.value_length));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 }
1954 /* The attribute list is now setup in memory. */
1955 /*
1956 * FIXME: I don't know if this case is actually possible.
1957 * According to logic it is not possible but I have seen too
1958 * many weird things in MS software to rely on logic... Thus we
1959 * perform a manual search and make sure the first $MFT/$DATA
1960 * extent is in the base inode. If it is not we abort with an
1961 * error and if we ever see a report of this error we will need
1962 * to do some magic in order to have the necessary mft record
1963 * loaded and in the right place in the page cache. But
1964 * hopefully logic will prevail and this never happens...
1965 */
1966 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1967 al_end = (u8*)al_entry + ni->attr_list_size;
1968 for (;; al_entry = next_al_entry) {
1969 /* Out of bounds check. */
1970 if ((u8*)al_entry < ni->attr_list ||
1971 (u8*)al_entry > al_end)
1972 goto em_put_err_out;
1973 /* Catch the end of the attribute list. */
1974 if ((u8*)al_entry == al_end)
1975 goto em_put_err_out;
1976 if (!al_entry->length)
1977 goto em_put_err_out;
1978 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1979 le16_to_cpu(al_entry->length) > al_end)
1980 goto em_put_err_out;
1981 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1982 le16_to_cpu(al_entry->length));
1983 if (le32_to_cpu(al_entry->type) >
1984 const_le32_to_cpu(AT_DATA))
1985 goto em_put_err_out;
1986 if (AT_DATA != al_entry->type)
1987 continue;
1988 /* We want an unnamed attribute. */
1989 if (al_entry->name_length)
1990 goto em_put_err_out;
1991 /* Want the first entry, i.e. lowest_vcn == 0. */
1992 if (al_entry->lowest_vcn)
1993 goto em_put_err_out;
1994 /* First entry has to be in the base mft record. */
1995 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1996 /* MFT references do not match, logic fails. */
1997 ntfs_error(sb, "BUG: The first $DATA extent "
1998 "of $MFT is not in the base "
1999 "mft record. Please report "
2000 "you saw this message to "
2001 "linux-ntfs-dev@lists."
2002 "sourceforge.net");
2003 goto put_err_out;
2004 } else {
2005 /* Sequence numbers must match. */
2006 if (MSEQNO_LE(al_entry->mft_reference) !=
2007 ni->seq_no)
2008 goto em_put_err_out;
2009 /* Got it. All is ok. We can stop now. */
2010 break;
2011 }
2012 }
2013 }
2014
2015 ntfs_attr_reinit_search_ctx(ctx);
2016
2017 /* Now load all attribute extents. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002018 a = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 next_vcn = last_vcn = highest_vcn = 0;
2020 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
2021 ctx))) {
2022 runlist_element *nrl;
2023
2024 /* Cache the current attribute. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002025 a = ctx->attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 /* $MFT must be non-resident. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002027 if (!a->non_resident) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 ntfs_error(sb, "$MFT must be non-resident but a "
2029 "resident extent was found. $MFT is "
2030 "corrupt. Run chkdsk.");
2031 goto put_err_out;
2032 }
2033 /* $MFT must be uncompressed and unencrypted. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002034 if (a->flags & ATTR_COMPRESSION_MASK ||
2035 a->flags & ATTR_IS_ENCRYPTED ||
2036 a->flags & ATTR_IS_SPARSE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 ntfs_error(sb, "$MFT must be uncompressed, "
2038 "non-sparse, and unencrypted but a "
2039 "compressed/sparse/encrypted extent "
2040 "was found. $MFT is corrupt. Run "
2041 "chkdsk.");
2042 goto put_err_out;
2043 }
2044 /*
2045 * Decompress the mapping pairs array of this extent and merge
2046 * the result into the existing runlist. No need for locking
2047 * as we have exclusive access to the inode at this time and we
2048 * are a mount in progress task, too.
2049 */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002050 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 if (IS_ERR(nrl)) {
2052 ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2053 "failed with error code %ld. $MFT is "
2054 "corrupt.", PTR_ERR(nrl));
2055 goto put_err_out;
2056 }
2057 ni->runlist.rl = nrl;
2058
2059 /* Are we in the first extent? */
2060 if (!next_vcn) {
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002061 if (a->data.non_resident.lowest_vcn) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062 ntfs_error(sb, "First extent of $DATA "
2063 "attribute has non zero "
2064 "lowest_vcn. $MFT is corrupt. "
2065 "You should run chkdsk.");
2066 goto put_err_out;
2067 }
2068 /* Get the last vcn in the $DATA attribute. */
2069 last_vcn = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002070 a->data.non_resident.allocated_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071 >> vol->cluster_size_bits;
2072 /* Fill in the inode size. */
2073 vi->i_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002074 a->data.non_resident.data_size);
2075 ni->initialized_size = sle64_to_cpu(
2076 a->data.non_resident.initialized_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077 ni->allocated_size = sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002078 a->data.non_resident.allocated_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 /*
2080 * Verify the number of mft records does not exceed
2081 * 2^32 - 1.
2082 */
2083 if ((vi->i_size >> vol->mft_record_size_bits) >=
2084 (1ULL << 32)) {
2085 ntfs_error(sb, "$MFT is too big! Aborting.");
2086 goto put_err_out;
2087 }
2088 /*
2089 * We have got the first extent of the runlist for
2090 * $MFT which means it is now relatively safe to call
2091 * the normal ntfs_read_inode() function.
2092 * Complete reading the inode, this will actually
2093 * re-read the mft record for $MFT, this time entering
2094 * it into the page cache with which we complete the
2095 * kick start of the volume. It should be safe to do
2096 * this now as the first extent of $MFT/$DATA is
2097 * already known and we would hope that we don't need
2098 * further extents in order to find the other
2099 * attributes belonging to $MFT. Only time will tell if
2100 * this is really the case. If not we will have to play
2101 * magic at this point, possibly duplicating a lot of
2102 * ntfs_read_inode() at this point. We will need to
2103 * ensure we do enough of its work to be able to call
2104 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2105 * hope this never happens...
2106 */
2107 ntfs_read_locked_inode(vi);
2108 if (is_bad_inode(vi)) {
2109 ntfs_error(sb, "ntfs_read_inode() of $MFT "
2110 "failed. BUG or corrupt $MFT. "
2111 "Run chkdsk and if no errors "
2112 "are found, please report you "
2113 "saw this message to "
2114 "linux-ntfs-dev@lists."
2115 "sourceforge.net");
2116 ntfs_attr_put_search_ctx(ctx);
2117 /* Revert to the safe super operations. */
2118 ntfs_free(m);
2119 return -1;
2120 }
2121 /*
2122 * Re-initialize some specifics about $MFT's inode as
2123 * ntfs_read_inode() will have set up the default ones.
2124 */
2125 /* Set uid and gid to root. */
2126 vi->i_uid = vi->i_gid = 0;
2127 /* Regular file. No access for anyone. */
2128 vi->i_mode = S_IFREG;
2129 /* No VFS initiated operations allowed for $MFT. */
2130 vi->i_op = &ntfs_empty_inode_ops;
2131 vi->i_fop = &ntfs_empty_file_ops;
2132 }
2133
2134 /* Get the lowest vcn for the next extent. */
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002135 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136 next_vcn = highest_vcn + 1;
2137
2138 /* Only one extent or error, which we catch below. */
2139 if (next_vcn <= 0)
2140 break;
2141
2142 /* Avoid endless loops due to corruption. */
2143 if (next_vcn < sle64_to_cpu(
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002144 a->data.non_resident.lowest_vcn)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 ntfs_error(sb, "$MFT has corrupt attribute list "
2146 "attribute. Run chkdsk.");
2147 goto put_err_out;
2148 }
2149 }
2150 if (err != -ENOENT) {
2151 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2152 "$MFT is corrupt. Run chkdsk.");
2153 goto put_err_out;
2154 }
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002155 if (!a) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2157 "corrupt. Run chkdsk.");
2158 goto put_err_out;
2159 }
2160 if (highest_vcn && highest_vcn != last_vcn - 1) {
2161 ntfs_error(sb, "Failed to load the complete runlist for "
2162 "$MFT/$DATA. Driver bug or corrupt $MFT. "
2163 "Run chkdsk.");
2164 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2165 (unsigned long long)highest_vcn,
2166 (unsigned long long)last_vcn - 1);
2167 goto put_err_out;
2168 }
2169 ntfs_attr_put_search_ctx(ctx);
2170 ntfs_debug("Done.");
2171 ntfs_free(m);
Ingo Molnar59345372006-07-03 00:25:18 -07002172
2173 /*
2174 * Split the locking rules of the MFT inode from the
2175 * locking rules of other inodes:
2176 */
2177 lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
2178 lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
2179
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180 return 0;
2181
2182em_put_err_out:
2183 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2184 "attribute list. $MFT is corrupt. Run chkdsk.");
2185put_err_out:
2186 ntfs_attr_put_search_ctx(ctx);
2187err_out:
2188 ntfs_error(sb, "Failed. Marking inode as bad.");
2189 make_bad_inode(vi);
2190 ntfs_free(m);
2191 return -1;
2192}
2193
2194/**
2195 * ntfs_put_inode - handler for when the inode reference count is decremented
2196 * @vi: vfs inode
2197 *
2198 * The VFS calls ntfs_put_inode() every time the inode reference count (i_count)
2199 * is about to be decremented (but before the decrement itself.
2200 *
2201 * If the inode @vi is a directory with two references, one of which is being
2202 * dropped, we need to put the attribute inode for the directory index bitmap,
2203 * if it is present, otherwise the directory inode would remain pinned for
2204 * ever.
2205 */
2206void ntfs_put_inode(struct inode *vi)
2207{
2208 if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
2209 ntfs_inode *ni = NTFS_I(vi);
2210 if (NInoIndexAllocPresent(ni)) {
2211 struct inode *bvi = NULL;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002212 mutex_lock(&vi->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 if (atomic_read(&vi->i_count) == 2) {
2214 bvi = ni->itype.index.bmp_ino;
2215 if (bvi)
2216 ni->itype.index.bmp_ino = NULL;
2217 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002218 mutex_unlock(&vi->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 if (bvi)
2220 iput(bvi);
2221 }
2222 }
2223}
2224
2225static void __ntfs_clear_inode(ntfs_inode *ni)
2226{
2227 /* Free all alocated memory. */
2228 down_write(&ni->runlist.lock);
2229 if (ni->runlist.rl) {
2230 ntfs_free(ni->runlist.rl);
2231 ni->runlist.rl = NULL;
2232 }
2233 up_write(&ni->runlist.lock);
2234
2235 if (ni->attr_list) {
2236 ntfs_free(ni->attr_list);
2237 ni->attr_list = NULL;
2238 }
2239
2240 down_write(&ni->attr_list_rl.lock);
2241 if (ni->attr_list_rl.rl) {
2242 ntfs_free(ni->attr_list_rl.rl);
2243 ni->attr_list_rl.rl = NULL;
2244 }
2245 up_write(&ni->attr_list_rl.lock);
2246
2247 if (ni->name_len && ni->name != I30) {
2248 /* Catch bugs... */
2249 BUG_ON(!ni->name);
2250 kfree(ni->name);
2251 }
2252}
2253
2254void ntfs_clear_extent_inode(ntfs_inode *ni)
2255{
2256 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2257
2258 BUG_ON(NInoAttr(ni));
2259 BUG_ON(ni->nr_extents != -1);
2260
2261#ifdef NTFS_RW
2262 if (NInoDirty(ni)) {
2263 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2264 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! "
2265 "Losing data! This is a BUG!!!");
2266 // FIXME: Do something!!!
2267 }
2268#endif /* NTFS_RW */
2269
2270 __ntfs_clear_inode(ni);
2271
2272 /* Bye, bye... */
2273 ntfs_destroy_extent_inode(ni);
2274}
2275
2276/**
2277 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode
2278 * @vi: vfs inode pending annihilation
2279 *
2280 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2281 * is called, which deallocates all memory belonging to the NTFS specific part
2282 * of the inode and returns.
2283 *
2284 * If the MFT record is dirty, we commit it before doing anything else.
2285 */
2286void ntfs_clear_big_inode(struct inode *vi)
2287{
2288 ntfs_inode *ni = NTFS_I(vi);
2289
2290 /*
2291 * If the inode @vi is an index inode we need to put the attribute
2292 * inode for the index bitmap, if it is present, otherwise the index
2293 * inode would disappear and the attribute inode for the index bitmap
2294 * would no longer be referenced from anywhere and thus it would remain
2295 * pinned for ever.
2296 */
2297 if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) &&
2298 NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) {
2299 iput(ni->itype.index.bmp_ino);
2300 ni->itype.index.bmp_ino = NULL;
2301 }
2302#ifdef NTFS_RW
2303 if (NInoDirty(ni)) {
Richard Knutssonc49c3112006-09-30 23:27:12 -07002304 bool was_bad = (is_bad_inode(vi));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305
2306 /* Committing the inode also commits all extent inodes. */
2307 ntfs_commit_inode(vi);
2308
2309 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2310 ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2311 "0x%lx. Losing data!", vi->i_ino);
2312 // FIXME: Do something!!!
2313 }
2314 }
2315#endif /* NTFS_RW */
2316
2317 /* No need to lock at this stage as no one else has a reference. */
2318 if (ni->nr_extents > 0) {
2319 int i;
2320
2321 for (i = 0; i < ni->nr_extents; i++)
2322 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2323 kfree(ni->ext.extent_ntfs_inos);
2324 }
2325
2326 __ntfs_clear_inode(ni);
2327
2328 if (NInoAttr(ni)) {
2329 /* Release the base inode if we are holding it. */
2330 if (ni->nr_extents == -1) {
2331 iput(VFS_I(ni->ext.base_ntfs_ino));
2332 ni->nr_extents = 0;
2333 ni->ext.base_ntfs_ino = NULL;
2334 }
2335 }
2336 return;
2337}
2338
2339/**
2340 * ntfs_show_options - show mount options in /proc/mounts
2341 * @sf: seq_file in which to write our mount options
2342 * @mnt: vfs mount whose mount options to display
2343 *
2344 * Called by the VFS once for each mounted ntfs volume when someone reads
2345 * /proc/mounts in order to display the NTFS specific mount options of each
2346 * mount. The mount options of the vfs mount @mnt are written to the seq file
2347 * @sf and success is returned.
2348 */
2349int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt)
2350{
2351 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb);
2352 int i;
2353
2354 seq_printf(sf, ",uid=%i", vol->uid);
2355 seq_printf(sf, ",gid=%i", vol->gid);
2356 if (vol->fmask == vol->dmask)
2357 seq_printf(sf, ",umask=0%o", vol->fmask);
2358 else {
2359 seq_printf(sf, ",fmask=0%o", vol->fmask);
2360 seq_printf(sf, ",dmask=0%o", vol->dmask);
2361 }
2362 seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2363 if (NVolCaseSensitive(vol))
2364 seq_printf(sf, ",case_sensitive");
2365 if (NVolShowSystemFiles(vol))
2366 seq_printf(sf, ",show_sys_files");
Anton Altaparmakovc002f422005-02-03 12:02:56 +00002367 if (!NVolSparseEnabled(vol))
2368 seq_printf(sf, ",disable_sparse");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369 for (i = 0; on_errors_arr[i].val; i++) {
2370 if (on_errors_arr[i].val & vol->on_errors)
2371 seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2372 }
2373 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2374 return 0;
2375}
2376
2377#ifdef NTFS_RW
2378
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002379static const char *es = " Leaving inconsistent metadata. Unmount and run "
2380 "chkdsk.";
2381
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382/**
2383 * ntfs_truncate - called when the i_size of an ntfs inode is changed
2384 * @vi: inode for which the i_size was changed
2385 *
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002386 * We only support i_size changes for normal files at present, i.e. not
2387 * compressed and not encrypted. This is enforced in ntfs_setattr(), see
2388 * below.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 *
2390 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2391 * that the change is allowed.
2392 *
2393 * This implies for us that @vi is a file inode rather than a directory, index,
2394 * or attribute inode as well as that @vi is a base inode.
2395 *
2396 * Returns 0 on success or -errno on error.
2397 *
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002398 * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002399 * writing. The only case in the kernel where ->i_alloc_sem is not held is
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002401 * with the current i_size as the offset. The analogous place in NTFS is in
2402 * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again
2403 * without holding ->i_alloc_sem.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404 */
2405int ntfs_truncate(struct inode *vi)
2406{
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002407 s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2408 VCN highest_vcn;
2409 unsigned long flags;
2410 ntfs_inode *base_ni, *ni = NTFS_I(vi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411 ntfs_volume *vol = ni->vol;
2412 ntfs_attr_search_ctx *ctx;
2413 MFT_RECORD *m;
Anton Altaparmakov5ae9fcf2005-03-02 17:03:24 +00002414 ATTR_RECORD *a;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002415 const char *te = " Leaving file length out of sync with i_size.";
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002416 int err, mp_size, size_change, alloc_change;
2417 u32 attr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418
2419 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2420 BUG_ON(NInoAttr(ni));
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002421 BUG_ON(S_ISDIR(vi->i_mode));
2422 BUG_ON(NInoMstProtected(ni));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423 BUG_ON(ni->nr_extents < 0);
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002424retry_truncate:
2425 /*
2426 * Lock the runlist for writing and map the mft record to ensure it is
2427 * safe to mess with the attribute runlist and sizes.
2428 */
2429 down_write(&ni->runlist.lock);
2430 if (!NInoAttr(ni))
2431 base_ni = ni;
2432 else
2433 base_ni = ni->ext.base_ntfs_ino;
2434 m = map_mft_record(base_ni);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 if (IS_ERR(m)) {
2436 err = PTR_ERR(m);
2437 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2438 "(error code %d).%s", vi->i_ino, err, te);
2439 ctx = NULL;
2440 m = NULL;
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002441 goto old_bad_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442 }
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002443 ctx = ntfs_attr_get_search_ctx(base_ni, m);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444 if (unlikely(!ctx)) {
2445 ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2446 "inode 0x%lx (not enough memory).%s",
2447 vi->i_ino, te);
2448 err = -ENOMEM;
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002449 goto old_bad_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450 }
2451 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2452 CASE_SENSITIVE, 0, NULL, 0, ctx);
2453 if (unlikely(err)) {
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002454 if (err == -ENOENT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455 ntfs_error(vi->i_sb, "Open attribute is missing from "
2456 "mft record. Inode 0x%lx is corrupt. "
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002457 "Run chkdsk.%s", vi->i_ino, te);
2458 err = -EIO;
2459 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460 ntfs_error(vi->i_sb, "Failed to lookup attribute in "
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002461 "inode 0x%lx (error code %d).%s",
2462 vi->i_ino, err, te);
2463 goto old_bad_out;
2464 }
2465 m = ctx->mrec;
2466 a = ctx->attr;
2467 /*
2468 * The i_size of the vfs inode is the new size for the attribute value.
2469 */
2470 new_size = i_size_read(vi);
2471 /* The current size of the attribute value is the old size. */
2472 old_size = ntfs_attr_size(a);
2473 /* Calculate the new allocated size. */
2474 if (NInoNonResident(ni))
2475 new_alloc_size = (new_size + vol->cluster_size - 1) &
2476 ~(s64)vol->cluster_size_mask;
2477 else
2478 new_alloc_size = (new_size + 7) & ~7;
2479 /* The current allocated size is the old allocated size. */
2480 read_lock_irqsave(&ni->size_lock, flags);
2481 old_alloc_size = ni->allocated_size;
2482 read_unlock_irqrestore(&ni->size_lock, flags);
2483 /*
2484 * The change in the file size. This will be 0 if no change, >0 if the
2485 * size is growing, and <0 if the size is shrinking.
2486 */
2487 size_change = -1;
2488 if (new_size - old_size >= 0) {
2489 size_change = 1;
2490 if (new_size == old_size)
2491 size_change = 0;
2492 }
2493 /* As above for the allocated size. */
2494 alloc_change = -1;
2495 if (new_alloc_size - old_alloc_size >= 0) {
2496 alloc_change = 1;
2497 if (new_alloc_size == old_alloc_size)
2498 alloc_change = 0;
2499 }
2500 /*
2501 * If neither the size nor the allocation are being changed there is
2502 * nothing to do.
2503 */
2504 if (!size_change && !alloc_change)
2505 goto unm_done;
2506 /* If the size is changing, check if new size is allowed in $AttrDef. */
2507 if (size_change) {
2508 err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2509 if (unlikely(err)) {
2510 if (err == -ERANGE) {
2511 ntfs_error(vol->sb, "Truncate would cause the "
2512 "inode 0x%lx to %simum size "
2513 "for its attribute type "
2514 "(0x%x). Aborting truncate.",
2515 vi->i_ino,
2516 new_size > old_size ? "exceed "
2517 "the max" : "go under the min",
2518 le32_to_cpu(ni->type));
2519 err = -EFBIG;
2520 } else {
2521 ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2522 "attribute type 0x%x. "
2523 "Aborting truncate.",
2524 vi->i_ino,
2525 le32_to_cpu(ni->type));
2526 err = -EIO;
2527 }
2528 /* Reset the vfs inode size to the old size. */
2529 i_size_write(vi, old_size);
2530 goto err_out;
2531 }
2532 }
2533 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2534 ntfs_warning(vi->i_sb, "Changes in inode size are not "
2535 "supported yet for %s files, ignoring.",
2536 NInoCompressed(ni) ? "compressed" :
2537 "encrypted");
2538 err = -EOPNOTSUPP;
2539 goto bad_out;
2540 }
2541 if (a->non_resident)
2542 goto do_non_resident_truncate;
2543 BUG_ON(NInoNonResident(ni));
2544 /* Resize the attribute record to best fit the new attribute size. */
2545 if (new_size < vol->mft_record_size &&
2546 !ntfs_resident_attr_value_resize(m, a, new_size)) {
2547 unsigned long flags;
2548
2549 /* The resize succeeded! */
2550 flush_dcache_mft_record_page(ctx->ntfs_ino);
2551 mark_mft_record_dirty(ctx->ntfs_ino);
2552 write_lock_irqsave(&ni->size_lock, flags);
2553 /* Update the sizes in the ntfs inode and all is done. */
2554 ni->allocated_size = le32_to_cpu(a->length) -
2555 le16_to_cpu(a->data.resident.value_offset);
2556 /*
2557 * Note ntfs_resident_attr_value_resize() has already done any
2558 * necessary data clearing in the attribute record. When the
2559 * file is being shrunk vmtruncate() will already have cleared
2560 * the top part of the last partial page, i.e. since this is
2561 * the resident case this is the page with index 0. However,
2562 * when the file is being expanded, the page cache page data
2563 * between the old data_size, i.e. old_size, and the new_size
2564 * has not been zeroed. Fortunately, we do not need to zero it
2565 * either since on one hand it will either already be zero due
2566 * to both readpage and writepage clearing partial page data
2567 * beyond i_size in which case there is nothing to do or in the
2568 * case of the file being mmap()ped at the same time, POSIX
2569 * specifies that the behaviour is unspecified thus we do not
2570 * have to do anything. This means that in our implementation
2571 * in the rare case that the file is mmap()ped and a write
2572 * occured into the mmap()ped region just beyond the file size
2573 * and writepage has not yet been called to write out the page
2574 * (which would clear the area beyond the file size) and we now
2575 * extend the file size to incorporate this dirty region
2576 * outside the file size, a write of the page would result in
2577 * this data being written to disk instead of being cleared.
2578 * Given both POSIX and the Linux mmap(2) man page specify that
2579 * this corner case is undefined, we choose to leave it like
2580 * that as this is much simpler for us as we cannot lock the
2581 * relevant page now since we are holding too many ntfs locks
2582 * which would result in a lock reversal deadlock.
2583 */
2584 ni->initialized_size = new_size;
2585 write_unlock_irqrestore(&ni->size_lock, flags);
2586 goto unm_done;
2587 }
2588 /* If the above resize failed, this must be an attribute extension. */
2589 BUG_ON(size_change < 0);
2590 /*
2591 * We have to drop all the locks so we can call
2592 * ntfs_attr_make_non_resident(). This could be optimised by try-
2593 * locking the first page cache page and only if that fails dropping
2594 * the locks, locking the page, and redoing all the locking and
2595 * lookups. While this would be a huge optimisation, it is not worth
2596 * it as this is definitely a slow code path as it only ever can happen
2597 * once for any given file.
2598 */
2599 ntfs_attr_put_search_ctx(ctx);
2600 unmap_mft_record(base_ni);
2601 up_write(&ni->runlist.lock);
2602 /*
2603 * Not enough space in the mft record, try to make the attribute
2604 * non-resident and if successful restart the truncation process.
2605 */
2606 err = ntfs_attr_make_non_resident(ni, old_size);
2607 if (likely(!err))
2608 goto retry_truncate;
2609 /*
2610 * Could not make non-resident. If this is due to this not being
2611 * permitted for this attribute type or there not being enough space,
2612 * try to make other attributes non-resident. Otherwise fail.
2613 */
2614 if (unlikely(err != -EPERM && err != -ENOSPC)) {
2615 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2616 "type 0x%x, because the conversion from "
2617 "resident to non-resident attribute failed "
2618 "with error code %i.", vi->i_ino,
2619 (unsigned)le32_to_cpu(ni->type), err);
2620 if (err != -ENOMEM)
2621 err = -EIO;
2622 goto conv_err_out;
2623 }
2624 /* TODO: Not implemented from here, abort. */
2625 if (err == -ENOSPC)
2626 ntfs_error(vol->sb, "Not enough space in the mft record/on "
2627 "disk for the non-resident attribute value. "
2628 "This case is not implemented yet.");
2629 else /* if (err == -EPERM) */
2630 ntfs_error(vol->sb, "This attribute type may not be "
2631 "non-resident. This case is not implemented "
2632 "yet.");
2633 err = -EOPNOTSUPP;
2634 goto conv_err_out;
2635#if 0
2636 // TODO: Attempt to make other attributes non-resident.
2637 if (!err)
2638 goto do_resident_extend;
2639 /*
2640 * Both the attribute list attribute and the standard information
2641 * attribute must remain in the base inode. Thus, if this is one of
2642 * these attributes, we have to try to move other attributes out into
2643 * extent mft records instead.
2644 */
2645 if (ni->type == AT_ATTRIBUTE_LIST ||
2646 ni->type == AT_STANDARD_INFORMATION) {
2647 // TODO: Attempt to move other attributes into extent mft
2648 // records.
2649 err = -EOPNOTSUPP;
2650 if (!err)
2651 goto do_resident_extend;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652 goto err_out;
2653 }
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002654 // TODO: Attempt to move this attribute to an extent mft record, but
2655 // only if it is not already the only attribute in an mft record in
2656 // which case there would be nothing to gain.
2657 err = -EOPNOTSUPP;
2658 if (!err)
2659 goto do_resident_extend;
2660 /* There is nothing we can do to make enough space. )-: */
2661 goto err_out;
2662#endif
2663do_non_resident_truncate:
2664 BUG_ON(!NInoNonResident(ni));
2665 if (alloc_change < 0) {
2666 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2667 if (highest_vcn > 0 &&
2668 old_alloc_size >> vol->cluster_size_bits >
2669 highest_vcn + 1) {
2670 /*
2671 * This attribute has multiple extents. Not yet
2672 * supported.
2673 */
2674 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2675 "attribute type 0x%x, because the "
2676 "attribute is highly fragmented (it "
2677 "consists of multiple extents) and "
2678 "this case is not implemented yet.",
2679 vi->i_ino,
2680 (unsigned)le32_to_cpu(ni->type));
2681 err = -EOPNOTSUPP;
2682 goto bad_out;
2683 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002684 }
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002685 /*
2686 * If the size is shrinking, need to reduce the initialized_size and
2687 * the data_size before reducing the allocation.
2688 */
2689 if (size_change < 0) {
2690 /*
2691 * Make the valid size smaller (i_size is already up-to-date).
2692 */
2693 write_lock_irqsave(&ni->size_lock, flags);
2694 if (new_size < ni->initialized_size) {
2695 ni->initialized_size = new_size;
2696 a->data.non_resident.initialized_size =
2697 cpu_to_sle64(new_size);
2698 }
2699 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2700 write_unlock_irqrestore(&ni->size_lock, flags);
2701 flush_dcache_mft_record_page(ctx->ntfs_ino);
2702 mark_mft_record_dirty(ctx->ntfs_ino);
2703 /* If the allocated size is not changing, we are done. */
2704 if (!alloc_change)
2705 goto unm_done;
2706 /*
2707 * If the size is shrinking it makes no sense for the
2708 * allocation to be growing.
2709 */
2710 BUG_ON(alloc_change > 0);
2711 } else /* if (size_change >= 0) */ {
2712 /*
2713 * The file size is growing or staying the same but the
2714 * allocation can be shrinking, growing or staying the same.
2715 */
2716 if (alloc_change > 0) {
2717 /*
2718 * We need to extend the allocation and possibly update
2719 * the data size. If we are updating the data size,
2720 * since we are not touching the initialized_size we do
2721 * not need to worry about the actual data on disk.
2722 * And as far as the page cache is concerned, there
2723 * will be no pages beyond the old data size and any
2724 * partial region in the last page between the old and
2725 * new data size (or the end of the page if the new
2726 * data size is outside the page) does not need to be
2727 * modified as explained above for the resident
2728 * attribute truncate case. To do this, we simply drop
2729 * the locks we hold and leave all the work to our
2730 * friendly helper ntfs_attr_extend_allocation().
2731 */
2732 ntfs_attr_put_search_ctx(ctx);
2733 unmap_mft_record(base_ni);
2734 up_write(&ni->runlist.lock);
2735 err = ntfs_attr_extend_allocation(ni, new_size,
2736 size_change > 0 ? new_size : -1, -1);
2737 /*
2738 * ntfs_attr_extend_allocation() will have done error
2739 * output already.
2740 */
2741 goto done;
2742 }
2743 if (!alloc_change)
2744 goto alloc_done;
2745 }
2746 /* alloc_change < 0 */
2747 /* Free the clusters. */
2748 nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2749 vol->cluster_size_bits, -1, ctx);
2750 m = ctx->mrec;
2751 a = ctx->attr;
2752 if (unlikely(nr_freed < 0)) {
2753 ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2754 "%lli). Unmount and run chkdsk to recover "
2755 "the lost cluster(s).", (long long)nr_freed);
2756 NVolSetErrors(vol);
2757 nr_freed = 0;
2758 }
2759 /* Truncate the runlist. */
2760 err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2761 new_alloc_size >> vol->cluster_size_bits);
2762 /*
2763 * If the runlist truncation failed and/or the search context is no
2764 * longer valid, we cannot resize the attribute record or build the
2765 * mapping pairs array thus we mark the inode bad so that no access to
2766 * the freed clusters can happen.
2767 */
2768 if (unlikely(err || IS_ERR(m))) {
2769 ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2770 IS_ERR(m) ?
2771 "restore attribute search context" :
2772 "truncate attribute runlist",
2773 IS_ERR(m) ? PTR_ERR(m) : err, es);
2774 err = -EIO;
2775 goto bad_out;
2776 }
2777 /* Get the size for the shrunk mapping pairs array for the runlist. */
2778 mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2779 if (unlikely(mp_size <= 0)) {
2780 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2781 "attribute type 0x%x, because determining the "
2782 "size for the mapping pairs failed with error "
2783 "code %i.%s", vi->i_ino,
2784 (unsigned)le32_to_cpu(ni->type), mp_size, es);
2785 err = -EIO;
2786 goto bad_out;
2787 }
2788 /*
2789 * Shrink the attribute record for the new mapping pairs array. Note,
2790 * this cannot fail since we are making the attribute smaller thus by
2791 * definition there is enough space to do so.
2792 */
2793 attr_len = le32_to_cpu(a->length);
2794 err = ntfs_attr_record_resize(m, a, mp_size +
2795 le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2796 BUG_ON(err);
2797 /*
2798 * Generate the mapping pairs array directly into the attribute record.
2799 */
2800 err = ntfs_mapping_pairs_build(vol, (u8*)a +
2801 le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2802 mp_size, ni->runlist.rl, 0, -1, NULL);
2803 if (unlikely(err)) {
2804 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2805 "attribute type 0x%x, because building the "
2806 "mapping pairs failed with error code %i.%s",
2807 vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2808 err, es);
2809 err = -EIO;
2810 goto bad_out;
2811 }
2812 /* Update the allocated/compressed size as well as the highest vcn. */
2813 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2814 vol->cluster_size_bits) - 1);
2815 write_lock_irqsave(&ni->size_lock, flags);
2816 ni->allocated_size = new_alloc_size;
2817 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2818 if (NInoSparse(ni) || NInoCompressed(ni)) {
2819 if (nr_freed) {
2820 ni->itype.compressed.size -= nr_freed <<
2821 vol->cluster_size_bits;
2822 BUG_ON(ni->itype.compressed.size < 0);
2823 a->data.non_resident.compressed_size = cpu_to_sle64(
2824 ni->itype.compressed.size);
2825 vi->i_blocks = ni->itype.compressed.size >> 9;
2826 }
2827 } else
2828 vi->i_blocks = new_alloc_size >> 9;
2829 write_unlock_irqrestore(&ni->size_lock, flags);
2830 /*
2831 * We have shrunk the allocation. If this is a shrinking truncate we
2832 * have already dealt with the initialized_size and the data_size above
2833 * and we are done. If the truncate is only changing the allocation
2834 * and not the data_size, we are also done. If this is an extending
2835 * truncate, need to extend the data_size now which is ensured by the
2836 * fact that @size_change is positive.
2837 */
2838alloc_done:
2839 /*
2840 * If the size is growing, need to update it now. If it is shrinking,
2841 * we have already updated it above (before the allocation change).
2842 */
2843 if (size_change > 0)
2844 a->data.non_resident.data_size = cpu_to_sle64(new_size);
2845 /* Ensure the modified mft record is written out. */
2846 flush_dcache_mft_record_page(ctx->ntfs_ino);
2847 mark_mft_record_dirty(ctx->ntfs_ino);
2848unm_done:
2849 ntfs_attr_put_search_ctx(ctx);
2850 unmap_mft_record(base_ni);
2851 up_write(&ni->runlist.lock);
2852done:
2853 /* Update the mtime and ctime on the base inode. */
Christoph Hellwig870f4812006-01-09 20:52:01 -08002854 /* normally ->truncate shouldn't update ctime or mtime,
2855 * but ntfs did before so it got a copy & paste version
2856 * of file_update_time. one day someone should fix this
2857 * for real.
2858 */
2859 if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2860 struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb);
2861 int sync_it = 0;
2862
2863 if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2864 !timespec_equal(&VFS_I(base_ni)->i_ctime, &now))
2865 sync_it = 1;
2866 VFS_I(base_ni)->i_mtime = now;
2867 VFS_I(base_ni)->i_ctime = now;
2868
2869 if (sync_it)
2870 mark_inode_dirty_sync(VFS_I(base_ni));
2871 }
2872
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002873 if (likely(!err)) {
2874 NInoClearTruncateFailed(ni);
2875 ntfs_debug("Done.");
2876 }
2877 return err;
2878old_bad_out:
2879 old_size = -1;
2880bad_out:
Anton Altaparmakova778f212006-03-23 16:18:23 +00002881 if (err != -ENOMEM && err != -EOPNOTSUPP)
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002882 NVolSetErrors(vol);
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002883 if (err != -EOPNOTSUPP)
2884 NInoSetTruncateFailed(ni);
2885 else if (old_size >= 0)
2886 i_size_write(vi, old_size);
2887err_out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002888 if (ctx)
2889 ntfs_attr_put_search_ctx(ctx);
2890 if (m)
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002891 unmap_mft_record(base_ni);
2892 up_write(&ni->runlist.lock);
2893out:
2894 ntfs_debug("Failed. Returning error code %i.", err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002895 return err;
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002896conv_err_out:
Anton Altaparmakova778f212006-03-23 16:18:23 +00002897 if (err != -ENOMEM && err != -EOPNOTSUPP)
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002898 NVolSetErrors(vol);
Anton Altaparmakovdd072332005-10-04 15:39:02 +01002899 if (err != -EOPNOTSUPP)
2900 NInoSetTruncateFailed(ni);
2901 else
2902 i_size_write(vi, old_size);
2903 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904}
2905
2906/**
2907 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2908 * @vi: inode for which the i_size was changed
2909 *
2910 * Wrapper for ntfs_truncate() that has no return value.
2911 *
2912 * See ntfs_truncate() description above for details.
2913 */
2914void ntfs_truncate_vfs(struct inode *vi) {
2915 ntfs_truncate(vi);
2916}
2917
2918/**
2919 * ntfs_setattr - called from notify_change() when an attribute is being changed
2920 * @dentry: dentry whose attributes to change
2921 * @attr: structure describing the attributes and the changes
2922 *
2923 * We have to trap VFS attempts to truncate the file described by @dentry as
2924 * soon as possible, because we do not implement changes in i_size yet. So we
2925 * abort all i_size changes here.
2926 *
2927 * We also abort all changes of user, group, and mode as we do not implement
2928 * the NTFS ACLs yet.
2929 *
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002930 * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also
Linus Torvalds1da177e2005-04-16 15:20:36 -07002931 * called with ->i_alloc_sem held for writing.
2932 *
2933 * Basically this is a copy of generic notify_change() and inode_setattr()
2934 * functionality, except we intercept and abort changes in i_size.
2935 */
2936int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2937{
2938 struct inode *vi = dentry->d_inode;
2939 int err;
2940 unsigned int ia_valid = attr->ia_valid;
2941
2942 err = inode_change_ok(vi, attr);
2943 if (err)
Anton Altaparmakove9438252005-10-04 16:01:06 +01002944 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002945 /* We do not support NTFS ACLs yet. */
2946 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2947 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2948 "supported yet, ignoring.");
2949 err = -EOPNOTSUPP;
2950 goto out;
2951 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002952 if (ia_valid & ATTR_SIZE) {
2953 if (attr->ia_size != i_size_read(vi)) {
Anton Altaparmakove9438252005-10-04 16:01:06 +01002954 ntfs_inode *ni = NTFS_I(vi);
2955 /*
2956 * FIXME: For now we do not support resizing of
2957 * compressed or encrypted files yet.
2958 */
2959 if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2960 ntfs_warning(vi->i_sb, "Changes in inode size "
2961 "are not supported yet for "
2962 "%s files, ignoring.",
2963 NInoCompressed(ni) ?
2964 "compressed" : "encrypted");
2965 err = -EOPNOTSUPP;
2966 } else
2967 err = vmtruncate(vi, attr->ia_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002968 if (err || ia_valid == ATTR_SIZE)
2969 goto out;
2970 } else {
2971 /*
2972 * We skipped the truncate but must still update
2973 * timestamps.
2974 */
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002975 ia_valid |= ATTR_MTIME | ATTR_CTIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002976 }
2977 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002978 if (ia_valid & ATTR_ATIME)
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002979 vi->i_atime = timespec_trunc(attr->ia_atime,
2980 vi->i_sb->s_time_gran);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002981 if (ia_valid & ATTR_MTIME)
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002982 vi->i_mtime = timespec_trunc(attr->ia_mtime,
2983 vi->i_sb->s_time_gran);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002984 if (ia_valid & ATTR_CTIME)
Anton Altaparmakov1c7d4692005-09-08 21:15:09 +01002985 vi->i_ctime = timespec_trunc(attr->ia_ctime,
2986 vi->i_sb->s_time_gran);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002987 mark_inode_dirty(vi);
2988out:
2989 return err;
2990}
2991
2992/**
2993 * ntfs_write_inode - write out a dirty inode
2994 * @vi: inode to write out
2995 * @sync: if true, write out synchronously
2996 *
2997 * Write out a dirty inode to disk including any extent inodes if present.
2998 *
2999 * If @sync is true, commit the inode to disk and wait for io completion. This
3000 * is done using write_mft_record().
3001 *
3002 * If @sync is false, just schedule the write to happen but do not wait for i/o
3003 * completion. In 2.6 kernels, scheduling usually happens just by virtue of
3004 * marking the page (and in this case mft record) dirty but we do not implement
3005 * this yet as write_mft_record() largely ignores the @sync parameter and
3006 * always performs synchronous writes.
3007 *
3008 * Return 0 on success and -errno on error.
3009 */
3010int ntfs_write_inode(struct inode *vi, int sync)
3011{
3012 sle64 nt;
3013 ntfs_inode *ni = NTFS_I(vi);
3014 ntfs_attr_search_ctx *ctx;
3015 MFT_RECORD *m;
3016 STANDARD_INFORMATION *si;
3017 int err = 0;
Richard Knutssonc49c3112006-09-30 23:27:12 -07003018 bool modified = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003019
3020 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
3021 vi->i_ino);
3022 /*
3023 * Dirty attribute inodes are written via their real inodes so just
3024 * clean them here. Access time updates are taken care off when the
3025 * real inode is written.
3026 */
3027 if (NInoAttr(ni)) {
3028 NInoClearDirty(ni);
3029 ntfs_debug("Done.");
3030 return 0;
3031 }
3032 /* Map, pin, and lock the mft record belonging to the inode. */
3033 m = map_mft_record(ni);
3034 if (IS_ERR(m)) {
3035 err = PTR_ERR(m);
3036 goto err_out;
3037 }
3038 /* Update the access times in the standard information attribute. */
3039 ctx = ntfs_attr_get_search_ctx(ni, m);
3040 if (unlikely(!ctx)) {
3041 err = -ENOMEM;
3042 goto unm_err_out;
3043 }
3044 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
3045 CASE_SENSITIVE, 0, NULL, 0, ctx);
3046 if (unlikely(err)) {
3047 ntfs_attr_put_search_ctx(ctx);
3048 goto unm_err_out;
3049 }
3050 si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3051 le16_to_cpu(ctx->attr->data.resident.value_offset));
3052 /* Update the access times if they have changed. */
3053 nt = utc2ntfs(vi->i_mtime);
3054 if (si->last_data_change_time != nt) {
3055 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
Randy Dunlap89075472005-03-03 11:19:53 +00003056 "new = 0x%llx", vi->i_ino, (long long)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003057 sle64_to_cpu(si->last_data_change_time),
Randy Dunlap89075472005-03-03 11:19:53 +00003058 (long long)sle64_to_cpu(nt));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003059 si->last_data_change_time = nt;
Richard Knutssonc49c3112006-09-30 23:27:12 -07003060 modified = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003061 }
3062 nt = utc2ntfs(vi->i_ctime);
3063 if (si->last_mft_change_time != nt) {
3064 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
Randy Dunlap89075472005-03-03 11:19:53 +00003065 "new = 0x%llx", vi->i_ino, (long long)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003066 sle64_to_cpu(si->last_mft_change_time),
Randy Dunlap89075472005-03-03 11:19:53 +00003067 (long long)sle64_to_cpu(nt));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003068 si->last_mft_change_time = nt;
Richard Knutssonc49c3112006-09-30 23:27:12 -07003069 modified = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070 }
3071 nt = utc2ntfs(vi->i_atime);
3072 if (si->last_access_time != nt) {
3073 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3074 "new = 0x%llx", vi->i_ino,
Randy Dunlap89075472005-03-03 11:19:53 +00003075 (long long)sle64_to_cpu(si->last_access_time),
3076 (long long)sle64_to_cpu(nt));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003077 si->last_access_time = nt;
Richard Knutssonc49c3112006-09-30 23:27:12 -07003078 modified = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003079 }
3080 /*
3081 * If we just modified the standard information attribute we need to
3082 * mark the mft record it is in dirty. We do this manually so that
3083 * mark_inode_dirty() is not called which would redirty the inode and
3084 * hence result in an infinite loop of trying to write the inode.
3085 * There is no need to mark the base inode nor the base mft record
3086 * dirty, since we are going to write this mft record below in any case
3087 * and the base mft record may actually not have been modified so it
3088 * might not need to be written out.
3089 * NOTE: It is not a problem when the inode for $MFT itself is being
3090 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3091 * on the $MFT inode and hence ntfs_write_inode() will not be
3092 * re-invoked because of it which in turn is ok since the dirtied mft
3093 * record will be cleaned and written out to disk below, i.e. before
3094 * this function returns.
3095 */
Anton Altaparmakov20fdcf12006-03-23 16:21:02 +00003096 if (modified) {
3097 flush_dcache_mft_record_page(ctx->ntfs_ino);
Ingo Molnar4e5e5292006-03-23 16:57:48 +00003098 if (!NInoTestSetDirty(ctx->ntfs_ino))
Anton Altaparmakov20fdcf12006-03-23 16:21:02 +00003099 mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3100 ctx->ntfs_ino->page_ofs);
3101 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003102 ntfs_attr_put_search_ctx(ctx);
3103 /* Now the access times are updated, write the base mft record. */
3104 if (NInoDirty(ni))
3105 err = write_mft_record(ni, m, sync);
3106 /* Write all attached extent mft records. */
Ingo Molnar4e5e5292006-03-23 16:57:48 +00003107 mutex_lock(&ni->extent_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003108 if (ni->nr_extents > 0) {
3109 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3110 int i;
3111
3112 ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3113 for (i = 0; i < ni->nr_extents; i++) {
3114 ntfs_inode *tni = extent_nis[i];
3115
3116 if (NInoDirty(tni)) {
3117 MFT_RECORD *tm = map_mft_record(tni);
3118 int ret;
3119
3120 if (IS_ERR(tm)) {
3121 if (!err || err == -ENOMEM)
3122 err = PTR_ERR(tm);
3123 continue;
3124 }
3125 ret = write_mft_record(tni, tm, sync);
3126 unmap_mft_record(tni);
3127 if (unlikely(ret)) {
3128 if (!err || err == -ENOMEM)
3129 err = ret;
3130 }
3131 }
3132 }
3133 }
Ingo Molnar4e5e5292006-03-23 16:57:48 +00003134 mutex_unlock(&ni->extent_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003135 unmap_mft_record(ni);
3136 if (unlikely(err))
3137 goto err_out;
3138 ntfs_debug("Done.");
3139 return 0;
3140unm_err_out:
3141 unmap_mft_record(ni);
3142err_out:
3143 if (err == -ENOMEM) {
3144 ntfs_warning(vi->i_sb, "Not enough memory to write inode. "
3145 "Marking the inode dirty again, so the VFS "
3146 "retries later.");
3147 mark_inode_dirty(vi);
3148 } else {
Anton Altaparmakova778f212006-03-23 16:18:23 +00003149 ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003150 NVolSetErrors(ni->vol);
3151 }
3152 return err;
3153}
3154
3155#endif /* NTFS_RW */