Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /** |
2 | * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project. | |
3 | * | |
c002f425 | 4 | * Copyright (c) 2001-2005 Anton Altaparmakov |
1da177e4 LT |
5 | * |
6 | * This program/include file is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License as published | |
8 | * by the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program/include file is distributed in the hope that it will be | |
12 | * useful, but WITHOUT ANY WARRANTY; without even the implied warranty | |
13 | * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program (in the main directory of the Linux-NTFS | |
18 | * distribution in the file COPYING); if not, write to the Free Software | |
19 | * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | */ | |
21 | ||
22 | #include <linux/pagemap.h> | |
23 | #include <linux/buffer_head.h> | |
24 | #include <linux/smp_lock.h> | |
25 | #include <linux/quotaops.h> | |
26 | #include <linux/mount.h> | |
27 | ||
28 | #include "aops.h" | |
29 | #include "dir.h" | |
30 | #include "debug.h" | |
31 | #include "inode.h" | |
32 | #include "attrib.h" | |
33 | #include "malloc.h" | |
34 | #include "mft.h" | |
35 | #include "time.h" | |
36 | #include "ntfs.h" | |
37 | ||
38 | /** | |
39 | * ntfs_test_inode - compare two (possibly fake) inodes for equality | |
40 | * @vi: vfs inode which to test | |
41 | * @na: ntfs attribute which is being tested with | |
42 | * | |
43 | * Compare the ntfs attribute embedded in the ntfs specific part of the vfs | |
44 | * inode @vi for equality with the ntfs attribute @na. | |
45 | * | |
46 | * If searching for the normal file/directory inode, set @na->type to AT_UNUSED. | |
47 | * @na->name and @na->name_len are then ignored. | |
48 | * | |
49 | * Return 1 if the attributes match and 0 if not. | |
50 | * | |
51 | * NOTE: This function runs with the inode_lock spin lock held so it is not | |
52 | * allowed to sleep. | |
53 | */ | |
54 | int ntfs_test_inode(struct inode *vi, ntfs_attr *na) | |
55 | { | |
56 | ntfs_inode *ni; | |
57 | ||
58 | if (vi->i_ino != na->mft_no) | |
59 | return 0; | |
60 | ni = NTFS_I(vi); | |
61 | /* If !NInoAttr(ni), @vi is a normal file or directory inode. */ | |
62 | if (likely(!NInoAttr(ni))) { | |
63 | /* If not looking for a normal inode this is a mismatch. */ | |
64 | if (unlikely(na->type != AT_UNUSED)) | |
65 | return 0; | |
66 | } else { | |
67 | /* A fake inode describing an attribute. */ | |
68 | if (ni->type != na->type) | |
69 | return 0; | |
70 | if (ni->name_len != na->name_len) | |
71 | return 0; | |
72 | if (na->name_len && memcmp(ni->name, na->name, | |
73 | na->name_len * sizeof(ntfschar))) | |
74 | return 0; | |
75 | } | |
76 | /* Match! */ | |
77 | return 1; | |
78 | } | |
79 | ||
80 | /** | |
81 | * ntfs_init_locked_inode - initialize an inode | |
82 | * @vi: vfs inode to initialize | |
83 | * @na: ntfs attribute which to initialize @vi to | |
84 | * | |
85 | * Initialize the vfs inode @vi with the values from the ntfs attribute @na in | |
86 | * order to enable ntfs_test_inode() to do its work. | |
87 | * | |
88 | * If initializing the normal file/directory inode, set @na->type to AT_UNUSED. | |
89 | * In that case, @na->name and @na->name_len should be set to NULL and 0, | |
90 | * respectively. Although that is not strictly necessary as | |
91 | * ntfs_read_inode_locked() will fill them in later. | |
92 | * | |
93 | * Return 0 on success and -errno on error. | |
94 | * | |
95 | * NOTE: This function runs with the inode_lock spin lock held so it is not | |
96 | * allowed to sleep. (Hence the GFP_ATOMIC allocation.) | |
97 | */ | |
98 | static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na) | |
99 | { | |
100 | ntfs_inode *ni = NTFS_I(vi); | |
101 | ||
102 | vi->i_ino = na->mft_no; | |
103 | ||
104 | ni->type = na->type; | |
105 | if (na->type == AT_INDEX_ALLOCATION) | |
106 | NInoSetMstProtected(ni); | |
107 | ||
108 | ni->name = na->name; | |
109 | ni->name_len = na->name_len; | |
110 | ||
111 | /* If initializing a normal inode, we are done. */ | |
112 | if (likely(na->type == AT_UNUSED)) { | |
113 | BUG_ON(na->name); | |
114 | BUG_ON(na->name_len); | |
115 | return 0; | |
116 | } | |
117 | ||
118 | /* It is a fake inode. */ | |
119 | NInoSetAttr(ni); | |
120 | ||
121 | /* | |
122 | * We have I30 global constant as an optimization as it is the name | |
123 | * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC | |
124 | * allocation but that is ok. And most attributes are unnamed anyway, | |
125 | * thus the fraction of named attributes with name != I30 is actually | |
126 | * absolutely tiny. | |
127 | */ | |
128 | if (na->name_len && na->name != I30) { | |
129 | unsigned int i; | |
130 | ||
131 | BUG_ON(!na->name); | |
132 | i = na->name_len * sizeof(ntfschar); | |
133 | ni->name = (ntfschar*)kmalloc(i + sizeof(ntfschar), GFP_ATOMIC); | |
134 | if (!ni->name) | |
135 | return -ENOMEM; | |
136 | memcpy(ni->name, na->name, i); | |
137 | ni->name[i] = 0; | |
138 | } | |
139 | return 0; | |
140 | } | |
141 | ||
142 | typedef int (*set_t)(struct inode *, void *); | |
143 | static int ntfs_read_locked_inode(struct inode *vi); | |
144 | static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi); | |
145 | static int ntfs_read_locked_index_inode(struct inode *base_vi, | |
146 | struct inode *vi); | |
147 | ||
148 | /** | |
149 | * ntfs_iget - obtain a struct inode corresponding to a specific normal inode | |
150 | * @sb: super block of mounted volume | |
151 | * @mft_no: mft record number / inode number to obtain | |
152 | * | |
153 | * Obtain the struct inode corresponding to a specific normal inode (i.e. a | |
154 | * file or directory). | |
155 | * | |
156 | * If the inode is in the cache, it is just returned with an increased | |
157 | * reference count. Otherwise, a new struct inode is allocated and initialized, | |
158 | * and finally ntfs_read_locked_inode() is called to read in the inode and | |
159 | * fill in the remainder of the inode structure. | |
160 | * | |
161 | * Return the struct inode on success. Check the return value with IS_ERR() and | |
162 | * if true, the function failed and the error code is obtained from PTR_ERR(). | |
163 | */ | |
164 | struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no) | |
165 | { | |
166 | struct inode *vi; | |
167 | ntfs_attr na; | |
168 | int err; | |
169 | ||
170 | na.mft_no = mft_no; | |
171 | na.type = AT_UNUSED; | |
172 | na.name = NULL; | |
173 | na.name_len = 0; | |
174 | ||
175 | vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode, | |
176 | (set_t)ntfs_init_locked_inode, &na); | |
f50f3ac5 | 177 | if (unlikely(!vi)) |
1da177e4 LT |
178 | return ERR_PTR(-ENOMEM); |
179 | ||
180 | err = 0; | |
181 | ||
182 | /* If this is a freshly allocated inode, need to read it now. */ | |
183 | if (vi->i_state & I_NEW) { | |
184 | err = ntfs_read_locked_inode(vi); | |
185 | unlock_new_inode(vi); | |
186 | } | |
187 | /* | |
188 | * There is no point in keeping bad inodes around if the failure was | |
189 | * due to ENOMEM. We want to be able to retry again later. | |
190 | */ | |
f50f3ac5 | 191 | if (unlikely(err == -ENOMEM)) { |
1da177e4 LT |
192 | iput(vi); |
193 | vi = ERR_PTR(err); | |
194 | } | |
195 | return vi; | |
196 | } | |
197 | ||
198 | /** | |
199 | * ntfs_attr_iget - obtain a struct inode corresponding to an attribute | |
200 | * @base_vi: vfs base inode containing the attribute | |
201 | * @type: attribute type | |
202 | * @name: Unicode name of the attribute (NULL if unnamed) | |
203 | * @name_len: length of @name in Unicode characters (0 if unnamed) | |
204 | * | |
205 | * Obtain the (fake) struct inode corresponding to the attribute specified by | |
206 | * @type, @name, and @name_len, which is present in the base mft record | |
207 | * specified by the vfs inode @base_vi. | |
208 | * | |
209 | * If the attribute inode is in the cache, it is just returned with an | |
210 | * increased reference count. Otherwise, a new struct inode is allocated and | |
211 | * initialized, and finally ntfs_read_locked_attr_inode() is called to read the | |
212 | * attribute and fill in the inode structure. | |
213 | * | |
214 | * Note, for index allocation attributes, you need to use ntfs_index_iget() | |
215 | * instead of ntfs_attr_iget() as working with indices is a lot more complex. | |
216 | * | |
217 | * Return the struct inode of the attribute inode on success. Check the return | |
218 | * value with IS_ERR() and if true, the function failed and the error code is | |
219 | * obtained from PTR_ERR(). | |
220 | */ | |
221 | struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type, | |
222 | ntfschar *name, u32 name_len) | |
223 | { | |
224 | struct inode *vi; | |
225 | ntfs_attr na; | |
226 | int err; | |
227 | ||
228 | /* Make sure no one calls ntfs_attr_iget() for indices. */ | |
229 | BUG_ON(type == AT_INDEX_ALLOCATION); | |
230 | ||
231 | na.mft_no = base_vi->i_ino; | |
232 | na.type = type; | |
233 | na.name = name; | |
234 | na.name_len = name_len; | |
235 | ||
236 | vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode, | |
237 | (set_t)ntfs_init_locked_inode, &na); | |
f50f3ac5 | 238 | if (unlikely(!vi)) |
1da177e4 LT |
239 | return ERR_PTR(-ENOMEM); |
240 | ||
241 | err = 0; | |
242 | ||
243 | /* If this is a freshly allocated inode, need to read it now. */ | |
244 | if (vi->i_state & I_NEW) { | |
245 | err = ntfs_read_locked_attr_inode(base_vi, vi); | |
246 | unlock_new_inode(vi); | |
247 | } | |
248 | /* | |
249 | * There is no point in keeping bad attribute inodes around. This also | |
250 | * simplifies things in that we never need to check for bad attribute | |
251 | * inodes elsewhere. | |
252 | */ | |
f50f3ac5 | 253 | if (unlikely(err)) { |
1da177e4 LT |
254 | iput(vi); |
255 | vi = ERR_PTR(err); | |
256 | } | |
257 | return vi; | |
258 | } | |
259 | ||
260 | /** | |
261 | * ntfs_index_iget - obtain a struct inode corresponding to an index | |
262 | * @base_vi: vfs base inode containing the index related attributes | |
263 | * @name: Unicode name of the index | |
264 | * @name_len: length of @name in Unicode characters | |
265 | * | |
266 | * Obtain the (fake) struct inode corresponding to the index specified by @name | |
267 | * and @name_len, which is present in the base mft record specified by the vfs | |
268 | * inode @base_vi. | |
269 | * | |
270 | * If the index inode is in the cache, it is just returned with an increased | |
271 | * reference count. Otherwise, a new struct inode is allocated and | |
272 | * initialized, and finally ntfs_read_locked_index_inode() is called to read | |
273 | * the index related attributes and fill in the inode structure. | |
274 | * | |
275 | * Return the struct inode of the index inode on success. Check the return | |
276 | * value with IS_ERR() and if true, the function failed and the error code is | |
277 | * obtained from PTR_ERR(). | |
278 | */ | |
279 | struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name, | |
280 | u32 name_len) | |
281 | { | |
282 | struct inode *vi; | |
283 | ntfs_attr na; | |
284 | int err; | |
285 | ||
286 | na.mft_no = base_vi->i_ino; | |
287 | na.type = AT_INDEX_ALLOCATION; | |
288 | na.name = name; | |
289 | na.name_len = name_len; | |
290 | ||
291 | vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode, | |
292 | (set_t)ntfs_init_locked_inode, &na); | |
f50f3ac5 | 293 | if (unlikely(!vi)) |
1da177e4 LT |
294 | return ERR_PTR(-ENOMEM); |
295 | ||
296 | err = 0; | |
297 | ||
298 | /* If this is a freshly allocated inode, need to read it now. */ | |
299 | if (vi->i_state & I_NEW) { | |
300 | err = ntfs_read_locked_index_inode(base_vi, vi); | |
301 | unlock_new_inode(vi); | |
302 | } | |
303 | /* | |
304 | * There is no point in keeping bad index inodes around. This also | |
305 | * simplifies things in that we never need to check for bad index | |
306 | * inodes elsewhere. | |
307 | */ | |
f50f3ac5 | 308 | if (unlikely(err)) { |
1da177e4 LT |
309 | iput(vi); |
310 | vi = ERR_PTR(err); | |
311 | } | |
312 | return vi; | |
313 | } | |
314 | ||
315 | struct inode *ntfs_alloc_big_inode(struct super_block *sb) | |
316 | { | |
317 | ntfs_inode *ni; | |
318 | ||
319 | ntfs_debug("Entering."); | |
2fb21db2 | 320 | ni = kmem_cache_alloc(ntfs_big_inode_cache, SLAB_NOFS); |
1da177e4 LT |
321 | if (likely(ni != NULL)) { |
322 | ni->state = 0; | |
323 | return VFS_I(ni); | |
324 | } | |
325 | ntfs_error(sb, "Allocation of NTFS big inode structure failed."); | |
326 | return NULL; | |
327 | } | |
328 | ||
329 | void ntfs_destroy_big_inode(struct inode *inode) | |
330 | { | |
331 | ntfs_inode *ni = NTFS_I(inode); | |
332 | ||
333 | ntfs_debug("Entering."); | |
334 | BUG_ON(ni->page); | |
335 | if (!atomic_dec_and_test(&ni->count)) | |
336 | BUG(); | |
337 | kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode)); | |
338 | } | |
339 | ||
340 | static inline ntfs_inode *ntfs_alloc_extent_inode(void) | |
341 | { | |
342 | ntfs_inode *ni; | |
343 | ||
344 | ntfs_debug("Entering."); | |
2fb21db2 | 345 | ni = kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS); |
1da177e4 LT |
346 | if (likely(ni != NULL)) { |
347 | ni->state = 0; | |
348 | return ni; | |
349 | } | |
350 | ntfs_error(NULL, "Allocation of NTFS inode structure failed."); | |
351 | return NULL; | |
352 | } | |
353 | ||
354 | static void ntfs_destroy_extent_inode(ntfs_inode *ni) | |
355 | { | |
356 | ntfs_debug("Entering."); | |
357 | BUG_ON(ni->page); | |
358 | if (!atomic_dec_and_test(&ni->count)) | |
359 | BUG(); | |
360 | kmem_cache_free(ntfs_inode_cache, ni); | |
361 | } | |
362 | ||
363 | /** | |
364 | * __ntfs_init_inode - initialize ntfs specific part of an inode | |
365 | * @sb: super block of mounted volume | |
366 | * @ni: freshly allocated ntfs inode which to initialize | |
367 | * | |
368 | * Initialize an ntfs inode to defaults. | |
369 | * | |
370 | * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left | |
371 | * untouched. Make sure to initialize them elsewhere. | |
372 | * | |
373 | * Return zero on success and -ENOMEM on error. | |
374 | */ | |
375 | void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni) | |
376 | { | |
377 | ntfs_debug("Entering."); | |
36763677 | 378 | rwlock_init(&ni->size_lock); |
1da177e4 LT |
379 | ni->initialized_size = ni->allocated_size = 0; |
380 | ni->seq_no = 0; | |
381 | atomic_set(&ni->count, 1); | |
382 | ni->vol = NTFS_SB(sb); | |
383 | ntfs_init_runlist(&ni->runlist); | |
384 | init_MUTEX(&ni->mrec_lock); | |
385 | ni->page = NULL; | |
386 | ni->page_ofs = 0; | |
387 | ni->attr_list_size = 0; | |
388 | ni->attr_list = NULL; | |
389 | ntfs_init_runlist(&ni->attr_list_rl); | |
390 | ni->itype.index.bmp_ino = NULL; | |
391 | ni->itype.index.block_size = 0; | |
392 | ni->itype.index.vcn_size = 0; | |
393 | ni->itype.index.collation_rule = 0; | |
394 | ni->itype.index.block_size_bits = 0; | |
395 | ni->itype.index.vcn_size_bits = 0; | |
396 | init_MUTEX(&ni->extent_lock); | |
397 | ni->nr_extents = 0; | |
398 | ni->ext.base_ntfs_ino = NULL; | |
399 | } | |
400 | ||
401 | inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb, | |
402 | unsigned long mft_no) | |
403 | { | |
404 | ntfs_inode *ni = ntfs_alloc_extent_inode(); | |
405 | ||
406 | ntfs_debug("Entering."); | |
407 | if (likely(ni != NULL)) { | |
408 | __ntfs_init_inode(sb, ni); | |
409 | ni->mft_no = mft_no; | |
410 | ni->type = AT_UNUSED; | |
411 | ni->name = NULL; | |
412 | ni->name_len = 0; | |
413 | } | |
414 | return ni; | |
415 | } | |
416 | ||
417 | /** | |
418 | * ntfs_is_extended_system_file - check if a file is in the $Extend directory | |
419 | * @ctx: initialized attribute search context | |
420 | * | |
421 | * Search all file name attributes in the inode described by the attribute | |
422 | * search context @ctx and check if any of the names are in the $Extend system | |
423 | * directory. | |
424 | * | |
425 | * Return values: | |
426 | * 1: file is in $Extend directory | |
427 | * 0: file is not in $Extend directory | |
428 | * -errno: failed to determine if the file is in the $Extend directory | |
429 | */ | |
430 | static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx) | |
431 | { | |
432 | int nr_links, err; | |
433 | ||
434 | /* Restart search. */ | |
435 | ntfs_attr_reinit_search_ctx(ctx); | |
436 | ||
437 | /* Get number of hard links. */ | |
438 | nr_links = le16_to_cpu(ctx->mrec->link_count); | |
439 | ||
440 | /* Loop through all hard links. */ | |
441 | while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0, | |
442 | ctx))) { | |
443 | FILE_NAME_ATTR *file_name_attr; | |
444 | ATTR_RECORD *attr = ctx->attr; | |
445 | u8 *p, *p2; | |
446 | ||
447 | nr_links--; | |
448 | /* | |
449 | * Maximum sanity checking as we are called on an inode that | |
450 | * we suspect might be corrupt. | |
451 | */ | |
452 | p = (u8*)attr + le32_to_cpu(attr->length); | |
453 | if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec + | |
454 | le32_to_cpu(ctx->mrec->bytes_in_use)) { | |
455 | err_corrupt_attr: | |
456 | ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name " | |
457 | "attribute. You should run chkdsk."); | |
458 | return -EIO; | |
459 | } | |
460 | if (attr->non_resident) { | |
461 | ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file " | |
462 | "name. You should run chkdsk."); | |
463 | return -EIO; | |
464 | } | |
465 | if (attr->flags) { | |
466 | ntfs_error(ctx->ntfs_ino->vol->sb, "File name with " | |
467 | "invalid flags. You should run " | |
468 | "chkdsk."); | |
469 | return -EIO; | |
470 | } | |
471 | if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) { | |
472 | ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file " | |
473 | "name. You should run chkdsk."); | |
474 | return -EIO; | |
475 | } | |
476 | file_name_attr = (FILE_NAME_ATTR*)((u8*)attr + | |
477 | le16_to_cpu(attr->data.resident.value_offset)); | |
478 | p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length); | |
479 | if (p2 < (u8*)attr || p2 > p) | |
480 | goto err_corrupt_attr; | |
481 | /* This attribute is ok, but is it in the $Extend directory? */ | |
482 | if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend) | |
483 | return 1; /* YES, it's an extended system file. */ | |
484 | } | |
485 | if (unlikely(err != -ENOENT)) | |
486 | return err; | |
487 | if (unlikely(nr_links)) { | |
488 | ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count " | |
489 | "doesn't match number of name attributes. You " | |
490 | "should run chkdsk."); | |
491 | return -EIO; | |
492 | } | |
493 | return 0; /* NO, it is not an extended system file. */ | |
494 | } | |
495 | ||
496 | /** | |
497 | * ntfs_read_locked_inode - read an inode from its device | |
498 | * @vi: inode to read | |
499 | * | |
500 | * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode | |
501 | * described by @vi into memory from the device. | |
502 | * | |
503 | * The only fields in @vi that we need to/can look at when the function is | |
504 | * called are i_sb, pointing to the mounted device's super block, and i_ino, | |
505 | * the number of the inode to load. | |
506 | * | |
507 | * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino | |
508 | * for reading and sets up the necessary @vi fields as well as initializing | |
509 | * the ntfs inode. | |
510 | * | |
511 | * Q: What locks are held when the function is called? | |
512 | * A: i_state has I_LOCK set, hence the inode is locked, also | |
513 | * i_count is set to 1, so it is not going to go away | |
514 | * i_flags is set to 0 and we have no business touching it. Only an ioctl() | |
515 | * is allowed to write to them. We should of course be honouring them but | |
516 | * we need to do that using the IS_* macros defined in include/linux/fs.h. | |
517 | * In any case ntfs_read_locked_inode() has nothing to do with i_flags. | |
518 | * | |
519 | * Return 0 on success and -errno on error. In the error case, the inode will | |
520 | * have had make_bad_inode() executed on it. | |
521 | */ | |
522 | static int ntfs_read_locked_inode(struct inode *vi) | |
523 | { | |
524 | ntfs_volume *vol = NTFS_SB(vi->i_sb); | |
525 | ntfs_inode *ni; | |
526 | MFT_RECORD *m; | |
5ae9fcf8 | 527 | ATTR_RECORD *a; |
1da177e4 LT |
528 | STANDARD_INFORMATION *si; |
529 | ntfs_attr_search_ctx *ctx; | |
530 | int err = 0; | |
531 | ||
532 | ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); | |
533 | ||
534 | /* Setup the generic vfs inode parts now. */ | |
535 | ||
536 | /* This is the optimal IO size (for stat), not the fs block size. */ | |
537 | vi->i_blksize = PAGE_CACHE_SIZE; | |
538 | /* | |
539 | * This is for checking whether an inode has changed w.r.t. a file so | |
540 | * that the file can be updated if necessary (compare with f_version). | |
541 | */ | |
542 | vi->i_version = 1; | |
543 | ||
544 | vi->i_uid = vol->uid; | |
545 | vi->i_gid = vol->gid; | |
546 | vi->i_mode = 0; | |
547 | ||
548 | /* | |
549 | * Initialize the ntfs specific part of @vi special casing | |
550 | * FILE_MFT which we need to do at mount time. | |
551 | */ | |
552 | if (vi->i_ino != FILE_MFT) | |
553 | ntfs_init_big_inode(vi); | |
554 | ni = NTFS_I(vi); | |
555 | ||
556 | m = map_mft_record(ni); | |
557 | if (IS_ERR(m)) { | |
558 | err = PTR_ERR(m); | |
559 | goto err_out; | |
560 | } | |
561 | ctx = ntfs_attr_get_search_ctx(ni, m); | |
562 | if (!ctx) { | |
563 | err = -ENOMEM; | |
564 | goto unm_err_out; | |
565 | } | |
566 | ||
567 | if (!(m->flags & MFT_RECORD_IN_USE)) { | |
568 | ntfs_error(vi->i_sb, "Inode is not in use!"); | |
569 | goto unm_err_out; | |
570 | } | |
571 | if (m->base_mft_record) { | |
572 | ntfs_error(vi->i_sb, "Inode is an extent inode!"); | |
573 | goto unm_err_out; | |
574 | } | |
575 | ||
576 | /* Transfer information from mft record into vfs and ntfs inodes. */ | |
577 | vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number); | |
578 | ||
579 | /* | |
580 | * FIXME: Keep in mind that link_count is two for files which have both | |
581 | * a long file name and a short file name as separate entries, so if | |
582 | * we are hiding short file names this will be too high. Either we need | |
583 | * to account for the short file names by subtracting them or we need | |
584 | * to make sure we delete files even though i_nlink is not zero which | |
585 | * might be tricky due to vfs interactions. Need to think about this | |
586 | * some more when implementing the unlink command. | |
587 | */ | |
588 | vi->i_nlink = le16_to_cpu(m->link_count); | |
589 | /* | |
590 | * FIXME: Reparse points can have the directory bit set even though | |
591 | * they would be S_IFLNK. Need to deal with this further below when we | |
592 | * implement reparse points / symbolic links but it will do for now. | |
593 | * Also if not a directory, it could be something else, rather than | |
594 | * a regular file. But again, will do for now. | |
595 | */ | |
596 | /* Everyone gets all permissions. */ | |
597 | vi->i_mode |= S_IRWXUGO; | |
598 | /* If read-only, noone gets write permissions. */ | |
599 | if (IS_RDONLY(vi)) | |
600 | vi->i_mode &= ~S_IWUGO; | |
601 | if (m->flags & MFT_RECORD_IS_DIRECTORY) { | |
602 | vi->i_mode |= S_IFDIR; | |
603 | /* | |
604 | * Apply the directory permissions mask set in the mount | |
605 | * options. | |
606 | */ | |
607 | vi->i_mode &= ~vol->dmask; | |
608 | /* Things break without this kludge! */ | |
609 | if (vi->i_nlink > 1) | |
610 | vi->i_nlink = 1; | |
611 | } else { | |
612 | vi->i_mode |= S_IFREG; | |
613 | /* Apply the file permissions mask set in the mount options. */ | |
614 | vi->i_mode &= ~vol->fmask; | |
615 | } | |
616 | /* | |
617 | * Find the standard information attribute in the mft record. At this | |
618 | * stage we haven't setup the attribute list stuff yet, so this could | |
619 | * in fact fail if the standard information is in an extent record, but | |
620 | * I don't think this actually ever happens. | |
621 | */ | |
622 | err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0, | |
623 | ctx); | |
624 | if (unlikely(err)) { | |
625 | if (err == -ENOENT) { | |
626 | /* | |
627 | * TODO: We should be performing a hot fix here (if the | |
628 | * recover mount option is set) by creating a new | |
629 | * attribute. | |
630 | */ | |
631 | ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute " | |
632 | "is missing."); | |
633 | } | |
634 | goto unm_err_out; | |
635 | } | |
5ae9fcf8 | 636 | a = ctx->attr; |
1da177e4 | 637 | /* Get the standard information attribute value. */ |
5ae9fcf8 AA |
638 | si = (STANDARD_INFORMATION*)((u8*)a + |
639 | le16_to_cpu(a->data.resident.value_offset)); | |
1da177e4 LT |
640 | |
641 | /* Transfer information from the standard information into vi. */ | |
642 | /* | |
643 | * Note: The i_?times do not quite map perfectly onto the NTFS times, | |
644 | * but they are close enough, and in the end it doesn't really matter | |
645 | * that much... | |
646 | */ | |
647 | /* | |
648 | * mtime is the last change of the data within the file. Not changed | |
649 | * when only metadata is changed, e.g. a rename doesn't affect mtime. | |
650 | */ | |
651 | vi->i_mtime = ntfs2utc(si->last_data_change_time); | |
652 | /* | |
653 | * ctime is the last change of the metadata of the file. This obviously | |
654 | * always changes, when mtime is changed. ctime can be changed on its | |
655 | * own, mtime is then not changed, e.g. when a file is renamed. | |
656 | */ | |
657 | vi->i_ctime = ntfs2utc(si->last_mft_change_time); | |
658 | /* | |
659 | * Last access to the data within the file. Not changed during a rename | |
660 | * for example but changed whenever the file is written to. | |
661 | */ | |
662 | vi->i_atime = ntfs2utc(si->last_access_time); | |
663 | ||
664 | /* Find the attribute list attribute if present. */ | |
665 | ntfs_attr_reinit_search_ctx(ctx); | |
666 | err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx); | |
667 | if (err) { | |
668 | if (unlikely(err != -ENOENT)) { | |
669 | ntfs_error(vi->i_sb, "Failed to lookup attribute list " | |
670 | "attribute."); | |
671 | goto unm_err_out; | |
672 | } | |
673 | } else /* if (!err) */ { | |
674 | if (vi->i_ino == FILE_MFT) | |
675 | goto skip_attr_list_load; | |
676 | ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino); | |
677 | NInoSetAttrList(ni); | |
5ae9fcf8 AA |
678 | a = ctx->attr; |
679 | if (a->flags & ATTR_IS_ENCRYPTED || | |
680 | a->flags & ATTR_COMPRESSION_MASK || | |
681 | a->flags & ATTR_IS_SPARSE) { | |
1da177e4 LT |
682 | ntfs_error(vi->i_sb, "Attribute list attribute is " |
683 | "compressed/encrypted/sparse."); | |
684 | goto unm_err_out; | |
685 | } | |
686 | /* Now allocate memory for the attribute list. */ | |
5ae9fcf8 | 687 | ni->attr_list_size = (u32)ntfs_attr_size(a); |
1da177e4 LT |
688 | ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size); |
689 | if (!ni->attr_list) { | |
690 | ntfs_error(vi->i_sb, "Not enough memory to allocate " | |
691 | "buffer for attribute list."); | |
692 | err = -ENOMEM; | |
693 | goto unm_err_out; | |
694 | } | |
5ae9fcf8 | 695 | if (a->non_resident) { |
1da177e4 | 696 | NInoSetAttrListNonResident(ni); |
5ae9fcf8 | 697 | if (a->data.non_resident.lowest_vcn) { |
1da177e4 LT |
698 | ntfs_error(vi->i_sb, "Attribute list has non " |
699 | "zero lowest_vcn."); | |
700 | goto unm_err_out; | |
701 | } | |
702 | /* | |
703 | * Setup the runlist. No need for locking as we have | |
704 | * exclusive access to the inode at this time. | |
705 | */ | |
706 | ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol, | |
5ae9fcf8 | 707 | a, NULL); |
1da177e4 LT |
708 | if (IS_ERR(ni->attr_list_rl.rl)) { |
709 | err = PTR_ERR(ni->attr_list_rl.rl); | |
710 | ni->attr_list_rl.rl = NULL; | |
711 | ntfs_error(vi->i_sb, "Mapping pairs " | |
712 | "decompression failed."); | |
713 | goto unm_err_out; | |
714 | } | |
715 | /* Now load the attribute list. */ | |
716 | if ((err = load_attribute_list(vol, &ni->attr_list_rl, | |
717 | ni->attr_list, ni->attr_list_size, | |
5ae9fcf8 AA |
718 | sle64_to_cpu(a->data.non_resident. |
719 | initialized_size)))) { | |
1da177e4 LT |
720 | ntfs_error(vi->i_sb, "Failed to load " |
721 | "attribute list attribute."); | |
722 | goto unm_err_out; | |
723 | } | |
5ae9fcf8 AA |
724 | } else /* if (!a->non_resident) */ { |
725 | if ((u8*)a + le16_to_cpu(a->data.resident.value_offset) | |
726 | + le32_to_cpu( | |
727 | a->data.resident.value_length) > | |
1da177e4 LT |
728 | (u8*)ctx->mrec + vol->mft_record_size) { |
729 | ntfs_error(vi->i_sb, "Corrupt attribute list " | |
730 | "in inode."); | |
731 | goto unm_err_out; | |
732 | } | |
733 | /* Now copy the attribute list. */ | |
5ae9fcf8 AA |
734 | memcpy(ni->attr_list, (u8*)a + le16_to_cpu( |
735 | a->data.resident.value_offset), | |
1da177e4 | 736 | le32_to_cpu( |
5ae9fcf8 | 737 | a->data.resident.value_length)); |
1da177e4 LT |
738 | } |
739 | } | |
740 | skip_attr_list_load: | |
741 | /* | |
742 | * If an attribute list is present we now have the attribute list value | |
743 | * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes. | |
744 | */ | |
745 | if (S_ISDIR(vi->i_mode)) { | |
f50f3ac5 | 746 | loff_t bvi_size; |
1da177e4 LT |
747 | struct inode *bvi; |
748 | ntfs_inode *bni; | |
749 | INDEX_ROOT *ir; | |
5ae9fcf8 | 750 | u8 *ir_end, *index_end; |
1da177e4 LT |
751 | |
752 | /* It is a directory, find index root attribute. */ | |
753 | ntfs_attr_reinit_search_ctx(ctx); | |
754 | err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE, | |
755 | 0, NULL, 0, ctx); | |
756 | if (unlikely(err)) { | |
757 | if (err == -ENOENT) { | |
758 | // FIXME: File is corrupt! Hot-fix with empty | |
759 | // index root attribute if recovery option is | |
760 | // set. | |
761 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute " | |
762 | "is missing."); | |
763 | } | |
764 | goto unm_err_out; | |
765 | } | |
5ae9fcf8 | 766 | a = ctx->attr; |
1da177e4 | 767 | /* Set up the state. */ |
5ae9fcf8 | 768 | if (unlikely(a->non_resident)) { |
1da177e4 LT |
769 | ntfs_error(vol->sb, "$INDEX_ROOT attribute is not " |
770 | "resident."); | |
771 | goto unm_err_out; | |
772 | } | |
773 | /* Ensure the attribute name is placed before the value. */ | |
5ae9fcf8 AA |
774 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
775 | le16_to_cpu(a->data.resident.value_offset)))) { | |
1da177e4 LT |
776 | ntfs_error(vol->sb, "$INDEX_ROOT attribute name is " |
777 | "placed after the attribute value."); | |
778 | goto unm_err_out; | |
779 | } | |
780 | /* | |
781 | * Compressed/encrypted index root just means that the newly | |
782 | * created files in that directory should be created compressed/ | |
783 | * encrypted. However index root cannot be both compressed and | |
784 | * encrypted. | |
785 | */ | |
5ae9fcf8 | 786 | if (a->flags & ATTR_COMPRESSION_MASK) |
1da177e4 | 787 | NInoSetCompressed(ni); |
5ae9fcf8 AA |
788 | if (a->flags & ATTR_IS_ENCRYPTED) { |
789 | if (a->flags & ATTR_COMPRESSION_MASK) { | |
1da177e4 LT |
790 | ntfs_error(vi->i_sb, "Found encrypted and " |
791 | "compressed attribute."); | |
792 | goto unm_err_out; | |
793 | } | |
794 | NInoSetEncrypted(ni); | |
795 | } | |
5ae9fcf8 | 796 | if (a->flags & ATTR_IS_SPARSE) |
1da177e4 | 797 | NInoSetSparse(ni); |
5ae9fcf8 AA |
798 | ir = (INDEX_ROOT*)((u8*)a + |
799 | le16_to_cpu(a->data.resident.value_offset)); | |
800 | ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length); | |
801 | if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) { | |
1da177e4 LT |
802 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is " |
803 | "corrupt."); | |
804 | goto unm_err_out; | |
805 | } | |
5ae9fcf8 | 806 | index_end = (u8*)&ir->index + |
1da177e4 LT |
807 | le32_to_cpu(ir->index.index_length); |
808 | if (index_end > ir_end) { | |
809 | ntfs_error(vi->i_sb, "Directory index is corrupt."); | |
810 | goto unm_err_out; | |
811 | } | |
812 | if (ir->type != AT_FILE_NAME) { | |
813 | ntfs_error(vi->i_sb, "Indexed attribute is not " | |
814 | "$FILE_NAME."); | |
815 | goto unm_err_out; | |
816 | } | |
817 | if (ir->collation_rule != COLLATION_FILE_NAME) { | |
818 | ntfs_error(vi->i_sb, "Index collation rule is not " | |
819 | "COLLATION_FILE_NAME."); | |
820 | goto unm_err_out; | |
821 | } | |
822 | ni->itype.index.collation_rule = ir->collation_rule; | |
823 | ni->itype.index.block_size = le32_to_cpu(ir->index_block_size); | |
824 | if (ni->itype.index.block_size & | |
825 | (ni->itype.index.block_size - 1)) { | |
826 | ntfs_error(vi->i_sb, "Index block size (%u) is not a " | |
827 | "power of two.", | |
828 | ni->itype.index.block_size); | |
829 | goto unm_err_out; | |
830 | } | |
831 | if (ni->itype.index.block_size > PAGE_CACHE_SIZE) { | |
832 | ntfs_error(vi->i_sb, "Index block size (%u) > " | |
833 | "PAGE_CACHE_SIZE (%ld) is not " | |
834 | "supported. Sorry.", | |
835 | ni->itype.index.block_size, | |
836 | PAGE_CACHE_SIZE); | |
837 | err = -EOPNOTSUPP; | |
838 | goto unm_err_out; | |
839 | } | |
840 | if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) { | |
841 | ntfs_error(vi->i_sb, "Index block size (%u) < " | |
842 | "NTFS_BLOCK_SIZE (%i) is not " | |
843 | "supported. Sorry.", | |
844 | ni->itype.index.block_size, | |
845 | NTFS_BLOCK_SIZE); | |
846 | err = -EOPNOTSUPP; | |
847 | goto unm_err_out; | |
848 | } | |
849 | ni->itype.index.block_size_bits = | |
850 | ffs(ni->itype.index.block_size) - 1; | |
851 | /* Determine the size of a vcn in the directory index. */ | |
852 | if (vol->cluster_size <= ni->itype.index.block_size) { | |
853 | ni->itype.index.vcn_size = vol->cluster_size; | |
854 | ni->itype.index.vcn_size_bits = vol->cluster_size_bits; | |
855 | } else { | |
856 | ni->itype.index.vcn_size = vol->sector_size; | |
857 | ni->itype.index.vcn_size_bits = vol->sector_size_bits; | |
858 | } | |
859 | ||
860 | /* Setup the index allocation attribute, even if not present. */ | |
861 | NInoSetMstProtected(ni); | |
862 | ni->type = AT_INDEX_ALLOCATION; | |
863 | ni->name = I30; | |
864 | ni->name_len = 4; | |
865 | ||
866 | if (!(ir->index.flags & LARGE_INDEX)) { | |
867 | /* No index allocation. */ | |
868 | vi->i_size = ni->initialized_size = | |
869 | ni->allocated_size = 0; | |
870 | /* We are done with the mft record, so we release it. */ | |
871 | ntfs_attr_put_search_ctx(ctx); | |
872 | unmap_mft_record(ni); | |
873 | m = NULL; | |
874 | ctx = NULL; | |
875 | goto skip_large_dir_stuff; | |
876 | } /* LARGE_INDEX: Index allocation present. Setup state. */ | |
877 | NInoSetIndexAllocPresent(ni); | |
878 | /* Find index allocation attribute. */ | |
879 | ntfs_attr_reinit_search_ctx(ctx); | |
880 | err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4, | |
881 | CASE_SENSITIVE, 0, NULL, 0, ctx); | |
882 | if (unlikely(err)) { | |
883 | if (err == -ENOENT) | |
884 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION " | |
885 | "attribute is not present but " | |
886 | "$INDEX_ROOT indicated it is."); | |
887 | else | |
888 | ntfs_error(vi->i_sb, "Failed to lookup " | |
889 | "$INDEX_ALLOCATION " | |
890 | "attribute."); | |
891 | goto unm_err_out; | |
892 | } | |
5ae9fcf8 AA |
893 | a = ctx->attr; |
894 | if (!a->non_resident) { | |
1da177e4 LT |
895 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
896 | "is resident."); | |
897 | goto unm_err_out; | |
898 | } | |
899 | /* | |
900 | * Ensure the attribute name is placed before the mapping pairs | |
901 | * array. | |
902 | */ | |
5ae9fcf8 AA |
903 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
904 | le16_to_cpu( | |
905 | a->data.non_resident.mapping_pairs_offset)))) { | |
1da177e4 LT |
906 | ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name " |
907 | "is placed after the mapping pairs " | |
908 | "array."); | |
909 | goto unm_err_out; | |
910 | } | |
5ae9fcf8 | 911 | if (a->flags & ATTR_IS_ENCRYPTED) { |
1da177e4 LT |
912 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
913 | "is encrypted."); | |
914 | goto unm_err_out; | |
915 | } | |
5ae9fcf8 | 916 | if (a->flags & ATTR_IS_SPARSE) { |
1da177e4 LT |
917 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
918 | "is sparse."); | |
919 | goto unm_err_out; | |
920 | } | |
5ae9fcf8 | 921 | if (a->flags & ATTR_COMPRESSION_MASK) { |
1da177e4 LT |
922 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
923 | "is compressed."); | |
924 | goto unm_err_out; | |
925 | } | |
5ae9fcf8 | 926 | if (a->data.non_resident.lowest_vcn) { |
1da177e4 LT |
927 | ntfs_error(vi->i_sb, "First extent of " |
928 | "$INDEX_ALLOCATION attribute has non " | |
929 | "zero lowest_vcn."); | |
930 | goto unm_err_out; | |
931 | } | |
5ae9fcf8 | 932 | vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); |
1da177e4 | 933 | ni->initialized_size = sle64_to_cpu( |
5ae9fcf8 | 934 | a->data.non_resident.initialized_size); |
1da177e4 | 935 | ni->allocated_size = sle64_to_cpu( |
5ae9fcf8 | 936 | a->data.non_resident.allocated_size); |
1da177e4 LT |
937 | /* |
938 | * We are done with the mft record, so we release it. Otherwise | |
939 | * we would deadlock in ntfs_attr_iget(). | |
940 | */ | |
941 | ntfs_attr_put_search_ctx(ctx); | |
942 | unmap_mft_record(ni); | |
943 | m = NULL; | |
944 | ctx = NULL; | |
945 | /* Get the index bitmap attribute inode. */ | |
946 | bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4); | |
947 | if (IS_ERR(bvi)) { | |
948 | ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); | |
949 | err = PTR_ERR(bvi); | |
950 | goto unm_err_out; | |
951 | } | |
952 | ni->itype.index.bmp_ino = bvi; | |
953 | bni = NTFS_I(bvi); | |
954 | if (NInoCompressed(bni) || NInoEncrypted(bni) || | |
955 | NInoSparse(bni)) { | |
956 | ntfs_error(vi->i_sb, "$BITMAP attribute is compressed " | |
957 | "and/or encrypted and/or sparse."); | |
958 | goto unm_err_out; | |
959 | } | |
960 | /* Consistency check bitmap size vs. index allocation size. */ | |
f50f3ac5 AA |
961 | bvi_size = i_size_read(bvi); |
962 | if ((bvi_size << 3) < (vi->i_size >> | |
1da177e4 LT |
963 | ni->itype.index.block_size_bits)) { |
964 | ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) " | |
965 | "for index allocation (0x%llx).", | |
f50f3ac5 | 966 | bvi_size << 3, vi->i_size); |
1da177e4 LT |
967 | goto unm_err_out; |
968 | } | |
969 | skip_large_dir_stuff: | |
970 | /* Setup the operations for this inode. */ | |
971 | vi->i_op = &ntfs_dir_inode_ops; | |
972 | vi->i_fop = &ntfs_dir_ops; | |
973 | } else { | |
974 | /* It is a file. */ | |
975 | ntfs_attr_reinit_search_ctx(ctx); | |
976 | ||
977 | /* Setup the data attribute, even if not present. */ | |
978 | ni->type = AT_DATA; | |
979 | ni->name = NULL; | |
980 | ni->name_len = 0; | |
981 | ||
982 | /* Find first extent of the unnamed data attribute. */ | |
983 | err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx); | |
984 | if (unlikely(err)) { | |
985 | vi->i_size = ni->initialized_size = | |
986 | ni->allocated_size = 0; | |
987 | if (err != -ENOENT) { | |
988 | ntfs_error(vi->i_sb, "Failed to lookup $DATA " | |
989 | "attribute."); | |
990 | goto unm_err_out; | |
991 | } | |
992 | /* | |
993 | * FILE_Secure does not have an unnamed $DATA | |
994 | * attribute, so we special case it here. | |
995 | */ | |
996 | if (vi->i_ino == FILE_Secure) | |
997 | goto no_data_attr_special_case; | |
998 | /* | |
999 | * Most if not all the system files in the $Extend | |
1000 | * system directory do not have unnamed data | |
1001 | * attributes so we need to check if the parent | |
1002 | * directory of the file is FILE_Extend and if it is | |
1003 | * ignore this error. To do this we need to get the | |
1004 | * name of this inode from the mft record as the name | |
1005 | * contains the back reference to the parent directory. | |
1006 | */ | |
1007 | if (ntfs_is_extended_system_file(ctx) > 0) | |
1008 | goto no_data_attr_special_case; | |
1009 | // FIXME: File is corrupt! Hot-fix with empty data | |
1010 | // attribute if recovery option is set. | |
1011 | ntfs_error(vi->i_sb, "$DATA attribute is missing."); | |
1012 | goto unm_err_out; | |
1013 | } | |
5ae9fcf8 | 1014 | a = ctx->attr; |
1da177e4 | 1015 | /* Setup the state. */ |
5ae9fcf8 | 1016 | if (a->non_resident) { |
1da177e4 | 1017 | NInoSetNonResident(ni); |
9451f851 AA |
1018 | if (a->flags & (ATTR_COMPRESSION_MASK | |
1019 | ATTR_IS_SPARSE)) { | |
1020 | if (a->flags & ATTR_COMPRESSION_MASK) { | |
1021 | NInoSetCompressed(ni); | |
1022 | if (vol->cluster_size > 4096) { | |
1023 | ntfs_error(vi->i_sb, "Found " | |
1024 | "compressed data but " | |
1025 | "compression is " | |
1026 | "disabled due to " | |
1027 | "cluster size (%i) > " | |
1028 | "4kiB.", | |
1029 | vol->cluster_size); | |
1030 | goto unm_err_out; | |
1031 | } | |
1032 | if ((a->flags & ATTR_COMPRESSION_MASK) | |
1033 | != ATTR_IS_COMPRESSED) { | |
1034 | ntfs_error(vi->i_sb, "Found " | |
1035 | "unknown compression " | |
1036 | "method or corrupt " | |
1037 | "file."); | |
1038 | goto unm_err_out; | |
1039 | } | |
1da177e4 | 1040 | } |
9451f851 AA |
1041 | if (a->flags & ATTR_IS_SPARSE) |
1042 | NInoSetSparse(ni); | |
5ae9fcf8 AA |
1043 | if (a->data.non_resident.compression_unit != |
1044 | 4) { | |
1da177e4 LT |
1045 | ntfs_error(vi->i_sb, "Found " |
1046 | "nonstandard compression unit " | |
1047 | "(%u instead of 4). Cannot " | |
1048 | "handle this.", | |
5ae9fcf8 | 1049 | a->data.non_resident. |
1da177e4 LT |
1050 | compression_unit); |
1051 | err = -EOPNOTSUPP; | |
1052 | goto unm_err_out; | |
1053 | } | |
9451f851 AA |
1054 | ni->itype.compressed.block_clusters = 1U << |
1055 | a->data.non_resident. | |
1056 | compression_unit; | |
1da177e4 | 1057 | ni->itype.compressed.block_size = 1U << ( |
5ae9fcf8 | 1058 | a->data.non_resident. |
1da177e4 LT |
1059 | compression_unit + |
1060 | vol->cluster_size_bits); | |
1061 | ni->itype.compressed.block_size_bits = ffs( | |
9451f851 AA |
1062 | ni->itype.compressed. |
1063 | block_size) - 1; | |
1064 | ni->itype.compressed.size = sle64_to_cpu( | |
1065 | a->data.non_resident. | |
1066 | compressed_size); | |
1da177e4 | 1067 | } |
5ae9fcf8 AA |
1068 | if (a->flags & ATTR_IS_ENCRYPTED) { |
1069 | if (a->flags & ATTR_COMPRESSION_MASK) { | |
1da177e4 LT |
1070 | ntfs_error(vi->i_sb, "Found encrypted " |
1071 | "and compressed data."); | |
1072 | goto unm_err_out; | |
1073 | } | |
1074 | NInoSetEncrypted(ni); | |
1075 | } | |
5ae9fcf8 | 1076 | if (a->data.non_resident.lowest_vcn) { |
1da177e4 LT |
1077 | ntfs_error(vi->i_sb, "First extent of $DATA " |
1078 | "attribute has non zero " | |
1079 | "lowest_vcn."); | |
1080 | goto unm_err_out; | |
1081 | } | |
1da177e4 | 1082 | vi->i_size = sle64_to_cpu( |
5ae9fcf8 | 1083 | a->data.non_resident.data_size); |
1da177e4 | 1084 | ni->initialized_size = sle64_to_cpu( |
5ae9fcf8 | 1085 | a->data.non_resident.initialized_size); |
1da177e4 | 1086 | ni->allocated_size = sle64_to_cpu( |
5ae9fcf8 | 1087 | a->data.non_resident.allocated_size); |
1da177e4 | 1088 | } else { /* Resident attribute. */ |
5ae9fcf8 AA |
1089 | vi->i_size = ni->initialized_size = le32_to_cpu( |
1090 | a->data.resident.value_length); | |
1091 | ni->allocated_size = le32_to_cpu(a->length) - | |
1092 | le16_to_cpu( | |
1093 | a->data.resident.value_offset); | |
1094 | if (vi->i_size > ni->allocated_size) { | |
1095 | ntfs_error(vi->i_sb, "Resident data attribute " | |
1096 | "is corrupt (size exceeds " | |
1097 | "allocation)."); | |
1098 | goto unm_err_out; | |
1099 | } | |
1da177e4 LT |
1100 | } |
1101 | no_data_attr_special_case: | |
1102 | /* We are done with the mft record, so we release it. */ | |
1103 | ntfs_attr_put_search_ctx(ctx); | |
1104 | unmap_mft_record(ni); | |
1105 | m = NULL; | |
1106 | ctx = NULL; | |
1107 | /* Setup the operations for this inode. */ | |
1108 | vi->i_op = &ntfs_file_inode_ops; | |
1109 | vi->i_fop = &ntfs_file_ops; | |
1110 | } | |
1111 | if (NInoMstProtected(ni)) | |
1112 | vi->i_mapping->a_ops = &ntfs_mst_aops; | |
1113 | else | |
1114 | vi->i_mapping->a_ops = &ntfs_aops; | |
1115 | /* | |
1116 | * The number of 512-byte blocks used on disk (for stat). This is in so | |
1117 | * far inaccurate as it doesn't account for any named streams or other | |
1118 | * special non-resident attributes, but that is how Windows works, too, | |
1119 | * so we are at least consistent with Windows, if not entirely | |
1120 | * consistent with the Linux Way. Doing it the Linux Way would cause a | |
1121 | * significant slowdown as it would involve iterating over all | |
1122 | * attributes in the mft record and adding the allocated/compressed | |
1123 | * sizes of all non-resident attributes present to give us the Linux | |
1124 | * correct size that should go into i_blocks (after division by 512). | |
1125 | */ | |
9451f851 | 1126 | if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni))) |
1da177e4 | 1127 | vi->i_blocks = ni->itype.compressed.size >> 9; |
9451f851 AA |
1128 | else |
1129 | vi->i_blocks = ni->allocated_size >> 9; | |
1da177e4 LT |
1130 | ntfs_debug("Done."); |
1131 | return 0; | |
1132 | ||
1133 | unm_err_out: | |
1134 | if (!err) | |
1135 | err = -EIO; | |
1136 | if (ctx) | |
1137 | ntfs_attr_put_search_ctx(ctx); | |
1138 | if (m) | |
1139 | unmap_mft_record(ni); | |
1140 | err_out: | |
1141 | ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt " | |
1142 | "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino); | |
1143 | make_bad_inode(vi); | |
1144 | if (err != -EOPNOTSUPP && err != -ENOMEM) | |
1145 | NVolSetErrors(vol); | |
1146 | return err; | |
1147 | } | |
1148 | ||
1149 | /** | |
1150 | * ntfs_read_locked_attr_inode - read an attribute inode from its base inode | |
1151 | * @base_vi: base inode | |
1152 | * @vi: attribute inode to read | |
1153 | * | |
1154 | * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the | |
1155 | * attribute inode described by @vi into memory from the base mft record | |
1156 | * described by @base_ni. | |
1157 | * | |
1158 | * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for | |
1159 | * reading and looks up the attribute described by @vi before setting up the | |
1160 | * necessary fields in @vi as well as initializing the ntfs inode. | |
1161 | * | |
1162 | * Q: What locks are held when the function is called? | |
1163 | * A: i_state has I_LOCK set, hence the inode is locked, also | |
1164 | * i_count is set to 1, so it is not going to go away | |
1165 | * | |
1166 | * Return 0 on success and -errno on error. In the error case, the inode will | |
1167 | * have had make_bad_inode() executed on it. | |
1168 | */ | |
1169 | static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi) | |
1170 | { | |
1171 | ntfs_volume *vol = NTFS_SB(vi->i_sb); | |
1172 | ntfs_inode *ni, *base_ni; | |
1173 | MFT_RECORD *m; | |
5ae9fcf8 | 1174 | ATTR_RECORD *a; |
1da177e4 LT |
1175 | ntfs_attr_search_ctx *ctx; |
1176 | int err = 0; | |
1177 | ||
1178 | ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); | |
1179 | ||
1180 | ntfs_init_big_inode(vi); | |
1181 | ||
1182 | ni = NTFS_I(vi); | |
1183 | base_ni = NTFS_I(base_vi); | |
1184 | ||
1185 | /* Just mirror the values from the base inode. */ | |
1186 | vi->i_blksize = base_vi->i_blksize; | |
1187 | vi->i_version = base_vi->i_version; | |
1188 | vi->i_uid = base_vi->i_uid; | |
1189 | vi->i_gid = base_vi->i_gid; | |
1190 | vi->i_nlink = base_vi->i_nlink; | |
1191 | vi->i_mtime = base_vi->i_mtime; | |
1192 | vi->i_ctime = base_vi->i_ctime; | |
1193 | vi->i_atime = base_vi->i_atime; | |
1194 | vi->i_generation = ni->seq_no = base_ni->seq_no; | |
1195 | ||
1196 | /* Set inode type to zero but preserve permissions. */ | |
1197 | vi->i_mode = base_vi->i_mode & ~S_IFMT; | |
1198 | ||
1199 | m = map_mft_record(base_ni); | |
1200 | if (IS_ERR(m)) { | |
1201 | err = PTR_ERR(m); | |
1202 | goto err_out; | |
1203 | } | |
1204 | ctx = ntfs_attr_get_search_ctx(base_ni, m); | |
1205 | if (!ctx) { | |
1206 | err = -ENOMEM; | |
1207 | goto unm_err_out; | |
1208 | } | |
1da177e4 LT |
1209 | /* Find the attribute. */ |
1210 | err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len, | |
1211 | CASE_SENSITIVE, 0, NULL, 0, ctx); | |
1212 | if (unlikely(err)) | |
1213 | goto unm_err_out; | |
5ae9fcf8 AA |
1214 | a = ctx->attr; |
1215 | if (!a->non_resident) { | |
1da177e4 | 1216 | /* Ensure the attribute name is placed before the value. */ |
5ae9fcf8 AA |
1217 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
1218 | le16_to_cpu(a->data.resident.value_offset)))) { | |
1da177e4 LT |
1219 | ntfs_error(vol->sb, "Attribute name is placed after " |
1220 | "the attribute value."); | |
1221 | goto unm_err_out; | |
1222 | } | |
5ae9fcf8 | 1223 | if (NInoMstProtected(ni) || a->flags) { |
1da177e4 LT |
1224 | ntfs_error(vi->i_sb, "Found mst protected attribute " |
1225 | "or attribute with non-zero flags but " | |
1226 | "the attribute is resident. Please " | |
1227 | "report you saw this message to " | |
1228 | "linux-ntfs-dev@lists.sourceforge.net"); | |
1229 | goto unm_err_out; | |
1230 | } | |
5ae9fcf8 AA |
1231 | vi->i_size = ni->initialized_size = le32_to_cpu( |
1232 | a->data.resident.value_length); | |
1233 | ni->allocated_size = le32_to_cpu(a->length) - | |
1234 | le16_to_cpu(a->data.resident.value_offset); | |
1235 | if (vi->i_size > ni->allocated_size) { | |
9451f851 AA |
1236 | ntfs_error(vi->i_sb, "Resident attribute is corrupt " |
1237 | "(size exceeds allocation)."); | |
5ae9fcf8 AA |
1238 | goto unm_err_out; |
1239 | } | |
1da177e4 LT |
1240 | } else { |
1241 | NInoSetNonResident(ni); | |
1242 | /* | |
1243 | * Ensure the attribute name is placed before the mapping pairs | |
1244 | * array. | |
1245 | */ | |
5ae9fcf8 AA |
1246 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
1247 | le16_to_cpu( | |
1248 | a->data.non_resident.mapping_pairs_offset)))) { | |
1da177e4 LT |
1249 | ntfs_error(vol->sb, "Attribute name is placed after " |
1250 | "the mapping pairs array."); | |
1251 | goto unm_err_out; | |
1252 | } | |
9451f851 AA |
1253 | if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) { |
1254 | if (a->flags & ATTR_COMPRESSION_MASK) { | |
1255 | NInoSetCompressed(ni); | |
1256 | if ((ni->type != AT_DATA) || (ni->type == | |
1257 | AT_DATA && ni->name_len)) { | |
1258 | ntfs_error(vi->i_sb, "Found compressed " | |
1259 | "non-data or named " | |
1260 | "data attribute. " | |
1261 | "Please report you " | |
1262 | "saw this message to " | |
1263 | "linux-ntfs-dev@lists." | |
1264 | "sourceforge.net"); | |
1265 | goto unm_err_out; | |
1266 | } | |
1267 | if (vol->cluster_size > 4096) { | |
1268 | ntfs_error(vi->i_sb, "Found compressed " | |
1269 | "attribute but " | |
1270 | "compression is " | |
1271 | "disabled due to " | |
1272 | "cluster size (%i) > " | |
1273 | "4kiB.", | |
1274 | vol->cluster_size); | |
1275 | goto unm_err_out; | |
1276 | } | |
1277 | if ((a->flags & ATTR_COMPRESSION_MASK) != | |
1278 | ATTR_IS_COMPRESSED) { | |
1279 | ntfs_error(vi->i_sb, "Found unknown " | |
1280 | "compression method."); | |
1281 | goto unm_err_out; | |
1282 | } | |
1283 | } | |
1da177e4 LT |
1284 | if (NInoMstProtected(ni)) { |
1285 | ntfs_error(vi->i_sb, "Found mst protected " | |
1286 | "attribute but the attribute " | |
9451f851 AA |
1287 | "is %s. Please report you " |
1288 | "saw this message to " | |
1da177e4 | 1289 | "linux-ntfs-dev@lists." |
9451f851 AA |
1290 | "sourceforge.net", |
1291 | NInoCompressed(ni) ? | |
1292 | "compressed" : "sparse"); | |
1da177e4 LT |
1293 | goto unm_err_out; |
1294 | } | |
9451f851 AA |
1295 | if (a->flags & ATTR_IS_SPARSE) |
1296 | NInoSetSparse(ni); | |
5ae9fcf8 | 1297 | if (a->data.non_resident.compression_unit != 4) { |
1da177e4 LT |
1298 | ntfs_error(vi->i_sb, "Found nonstandard " |
1299 | "compression unit (%u instead " | |
1300 | "of 4). Cannot handle this.", | |
5ae9fcf8 | 1301 | a->data.non_resident. |
1da177e4 LT |
1302 | compression_unit); |
1303 | err = -EOPNOTSUPP; | |
1304 | goto unm_err_out; | |
1305 | } | |
9451f851 AA |
1306 | ni->itype.compressed.block_clusters = 1U << |
1307 | a->data.non_resident.compression_unit; | |
1da177e4 | 1308 | ni->itype.compressed.block_size = 1U << ( |
5ae9fcf8 | 1309 | a->data.non_resident.compression_unit + |
1da177e4 LT |
1310 | vol->cluster_size_bits); |
1311 | ni->itype.compressed.block_size_bits = ffs( | |
5ae9fcf8 | 1312 | ni->itype.compressed.block_size) - 1; |
9451f851 AA |
1313 | ni->itype.compressed.size = sle64_to_cpu( |
1314 | a->data.non_resident.compressed_size); | |
1da177e4 | 1315 | } |
5ae9fcf8 AA |
1316 | if (a->flags & ATTR_IS_ENCRYPTED) { |
1317 | if (a->flags & ATTR_COMPRESSION_MASK) { | |
1318 | ntfs_error(vi->i_sb, "Found encrypted and " | |
1319 | "compressed data."); | |
1da177e4 LT |
1320 | goto unm_err_out; |
1321 | } | |
1322 | if (NInoMstProtected(ni)) { | |
1323 | ntfs_error(vi->i_sb, "Found mst protected " | |
1324 | "attribute but the attribute " | |
1325 | "is encrypted. Please report " | |
1326 | "you saw this message to " | |
1327 | "linux-ntfs-dev@lists." | |
1328 | "sourceforge.net"); | |
1329 | goto unm_err_out; | |
1330 | } | |
1331 | NInoSetEncrypted(ni); | |
1332 | } | |
5ae9fcf8 | 1333 | if (a->data.non_resident.lowest_vcn) { |
1da177e4 LT |
1334 | ntfs_error(vi->i_sb, "First extent of attribute has " |
1335 | "non-zero lowest_vcn."); | |
1336 | goto unm_err_out; | |
1337 | } | |
5ae9fcf8 | 1338 | vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); |
1da177e4 | 1339 | ni->initialized_size = sle64_to_cpu( |
5ae9fcf8 | 1340 | a->data.non_resident.initialized_size); |
1da177e4 | 1341 | ni->allocated_size = sle64_to_cpu( |
5ae9fcf8 | 1342 | a->data.non_resident.allocated_size); |
1da177e4 | 1343 | } |
1da177e4 LT |
1344 | /* Setup the operations for this attribute inode. */ |
1345 | vi->i_op = NULL; | |
1346 | vi->i_fop = NULL; | |
1347 | if (NInoMstProtected(ni)) | |
1348 | vi->i_mapping->a_ops = &ntfs_mst_aops; | |
1349 | else | |
1350 | vi->i_mapping->a_ops = &ntfs_aops; | |
9451f851 | 1351 | if (NInoCompressed(ni) || NInoSparse(ni)) |
1da177e4 | 1352 | vi->i_blocks = ni->itype.compressed.size >> 9; |
9451f851 AA |
1353 | else |
1354 | vi->i_blocks = ni->allocated_size >> 9; | |
1da177e4 LT |
1355 | /* |
1356 | * Make sure the base inode doesn't go away and attach it to the | |
1357 | * attribute inode. | |
1358 | */ | |
1359 | igrab(base_vi); | |
1360 | ni->ext.base_ntfs_ino = base_ni; | |
1361 | ni->nr_extents = -1; | |
1362 | ||
1363 | ntfs_attr_put_search_ctx(ctx); | |
1364 | unmap_mft_record(base_ni); | |
1365 | ||
1366 | ntfs_debug("Done."); | |
1367 | return 0; | |
1368 | ||
1369 | unm_err_out: | |
1370 | if (!err) | |
1371 | err = -EIO; | |
1372 | if (ctx) | |
1373 | ntfs_attr_put_search_ctx(ctx); | |
1374 | unmap_mft_record(base_ni); | |
1375 | err_out: | |
1376 | ntfs_error(vol->sb, "Failed with error code %i while reading attribute " | |
1377 | "inode (mft_no 0x%lx, type 0x%x, name_len %i). " | |
1378 | "Marking corrupt inode and base inode 0x%lx as bad. " | |
1379 | "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len, | |
1380 | base_vi->i_ino); | |
1381 | make_bad_inode(vi); | |
1382 | make_bad_inode(base_vi); | |
1383 | if (err != -ENOMEM) | |
1384 | NVolSetErrors(vol); | |
1385 | return err; | |
1386 | } | |
1387 | ||
1388 | /** | |
1389 | * ntfs_read_locked_index_inode - read an index inode from its base inode | |
1390 | * @base_vi: base inode | |
1391 | * @vi: index inode to read | |
1392 | * | |
1393 | * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the | |
1394 | * index inode described by @vi into memory from the base mft record described | |
1395 | * by @base_ni. | |
1396 | * | |
1397 | * ntfs_read_locked_index_inode() maps, pins and locks the base inode for | |
1398 | * reading and looks up the attributes relating to the index described by @vi | |
1399 | * before setting up the necessary fields in @vi as well as initializing the | |
1400 | * ntfs inode. | |
1401 | * | |
1402 | * Note, index inodes are essentially attribute inodes (NInoAttr() is true) | |
1403 | * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they | |
1404 | * are setup like directory inodes since directories are a special case of | |
1405 | * indices ao they need to be treated in much the same way. Most importantly, | |
1406 | * for small indices the index allocation attribute might not actually exist. | |
1407 | * However, the index root attribute always exists but this does not need to | |
1408 | * have an inode associated with it and this is why we define a new inode type | |
1409 | * index. Also, like for directories, we need to have an attribute inode for | |
1410 | * the bitmap attribute corresponding to the index allocation attribute and we | |
1411 | * can store this in the appropriate field of the inode, just like we do for | |
1412 | * normal directory inodes. | |
1413 | * | |
1414 | * Q: What locks are held when the function is called? | |
1415 | * A: i_state has I_LOCK set, hence the inode is locked, also | |
1416 | * i_count is set to 1, so it is not going to go away | |
1417 | * | |
1418 | * Return 0 on success and -errno on error. In the error case, the inode will | |
1419 | * have had make_bad_inode() executed on it. | |
1420 | */ | |
1421 | static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi) | |
1422 | { | |
f50f3ac5 | 1423 | loff_t bvi_size; |
1da177e4 LT |
1424 | ntfs_volume *vol = NTFS_SB(vi->i_sb); |
1425 | ntfs_inode *ni, *base_ni, *bni; | |
1426 | struct inode *bvi; | |
1427 | MFT_RECORD *m; | |
5ae9fcf8 | 1428 | ATTR_RECORD *a; |
1da177e4 LT |
1429 | ntfs_attr_search_ctx *ctx; |
1430 | INDEX_ROOT *ir; | |
1431 | u8 *ir_end, *index_end; | |
1432 | int err = 0; | |
1433 | ||
1434 | ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); | |
1435 | ntfs_init_big_inode(vi); | |
1436 | ni = NTFS_I(vi); | |
1437 | base_ni = NTFS_I(base_vi); | |
1438 | /* Just mirror the values from the base inode. */ | |
1439 | vi->i_blksize = base_vi->i_blksize; | |
1440 | vi->i_version = base_vi->i_version; | |
1441 | vi->i_uid = base_vi->i_uid; | |
1442 | vi->i_gid = base_vi->i_gid; | |
1443 | vi->i_nlink = base_vi->i_nlink; | |
1444 | vi->i_mtime = base_vi->i_mtime; | |
1445 | vi->i_ctime = base_vi->i_ctime; | |
1446 | vi->i_atime = base_vi->i_atime; | |
1447 | vi->i_generation = ni->seq_no = base_ni->seq_no; | |
1448 | /* Set inode type to zero but preserve permissions. */ | |
1449 | vi->i_mode = base_vi->i_mode & ~S_IFMT; | |
1450 | /* Map the mft record for the base inode. */ | |
1451 | m = map_mft_record(base_ni); | |
1452 | if (IS_ERR(m)) { | |
1453 | err = PTR_ERR(m); | |
1454 | goto err_out; | |
1455 | } | |
1456 | ctx = ntfs_attr_get_search_ctx(base_ni, m); | |
1457 | if (!ctx) { | |
1458 | err = -ENOMEM; | |
1459 | goto unm_err_out; | |
1460 | } | |
1461 | /* Find the index root attribute. */ | |
1462 | err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len, | |
1463 | CASE_SENSITIVE, 0, NULL, 0, ctx); | |
1464 | if (unlikely(err)) { | |
1465 | if (err == -ENOENT) | |
1466 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is " | |
1467 | "missing."); | |
1468 | goto unm_err_out; | |
1469 | } | |
5ae9fcf8 | 1470 | a = ctx->attr; |
1da177e4 | 1471 | /* Set up the state. */ |
5ae9fcf8 | 1472 | if (unlikely(a->non_resident)) { |
1da177e4 LT |
1473 | ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident."); |
1474 | goto unm_err_out; | |
1475 | } | |
1476 | /* Ensure the attribute name is placed before the value. */ | |
5ae9fcf8 AA |
1477 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
1478 | le16_to_cpu(a->data.resident.value_offset)))) { | |
1da177e4 LT |
1479 | ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed " |
1480 | "after the attribute value."); | |
1481 | goto unm_err_out; | |
1482 | } | |
1483 | /* Compressed/encrypted/sparse index root is not allowed. */ | |
5ae9fcf8 | 1484 | if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED | |
1da177e4 LT |
1485 | ATTR_IS_SPARSE)) { |
1486 | ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index " | |
1487 | "root attribute."); | |
1488 | goto unm_err_out; | |
1489 | } | |
5ae9fcf8 AA |
1490 | ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset)); |
1491 | ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length); | |
1da177e4 LT |
1492 | if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) { |
1493 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt."); | |
1494 | goto unm_err_out; | |
1495 | } | |
1496 | index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length); | |
1497 | if (index_end > ir_end) { | |
1498 | ntfs_error(vi->i_sb, "Index is corrupt."); | |
1499 | goto unm_err_out; | |
1500 | } | |
1501 | if (ir->type) { | |
1502 | ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).", | |
1503 | le32_to_cpu(ir->type)); | |
1504 | goto unm_err_out; | |
1505 | } | |
1506 | ni->itype.index.collation_rule = ir->collation_rule; | |
1507 | ntfs_debug("Index collation rule is 0x%x.", | |
1508 | le32_to_cpu(ir->collation_rule)); | |
1509 | ni->itype.index.block_size = le32_to_cpu(ir->index_block_size); | |
1510 | if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) { | |
1511 | ntfs_error(vi->i_sb, "Index block size (%u) is not a power of " | |
1512 | "two.", ni->itype.index.block_size); | |
1513 | goto unm_err_out; | |
1514 | } | |
1515 | if (ni->itype.index.block_size > PAGE_CACHE_SIZE) { | |
1516 | ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE " | |
1517 | "(%ld) is not supported. Sorry.", | |
1518 | ni->itype.index.block_size, PAGE_CACHE_SIZE); | |
1519 | err = -EOPNOTSUPP; | |
1520 | goto unm_err_out; | |
1521 | } | |
1522 | if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) { | |
1523 | ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE " | |
1524 | "(%i) is not supported. Sorry.", | |
1525 | ni->itype.index.block_size, NTFS_BLOCK_SIZE); | |
1526 | err = -EOPNOTSUPP; | |
1527 | goto unm_err_out; | |
1528 | } | |
1529 | ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1; | |
1530 | /* Determine the size of a vcn in the index. */ | |
1531 | if (vol->cluster_size <= ni->itype.index.block_size) { | |
1532 | ni->itype.index.vcn_size = vol->cluster_size; | |
1533 | ni->itype.index.vcn_size_bits = vol->cluster_size_bits; | |
1534 | } else { | |
1535 | ni->itype.index.vcn_size = vol->sector_size; | |
1536 | ni->itype.index.vcn_size_bits = vol->sector_size_bits; | |
1537 | } | |
1538 | /* Check for presence of index allocation attribute. */ | |
1539 | if (!(ir->index.flags & LARGE_INDEX)) { | |
1540 | /* No index allocation. */ | |
1541 | vi->i_size = ni->initialized_size = ni->allocated_size = 0; | |
1542 | /* We are done with the mft record, so we release it. */ | |
1543 | ntfs_attr_put_search_ctx(ctx); | |
1544 | unmap_mft_record(base_ni); | |
1545 | m = NULL; | |
1546 | ctx = NULL; | |
1547 | goto skip_large_index_stuff; | |
1548 | } /* LARGE_INDEX: Index allocation present. Setup state. */ | |
1549 | NInoSetIndexAllocPresent(ni); | |
1550 | /* Find index allocation attribute. */ | |
1551 | ntfs_attr_reinit_search_ctx(ctx); | |
1552 | err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len, | |
1553 | CASE_SENSITIVE, 0, NULL, 0, ctx); | |
1554 | if (unlikely(err)) { | |
1555 | if (err == -ENOENT) | |
1556 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " | |
1557 | "not present but $INDEX_ROOT " | |
1558 | "indicated it is."); | |
1559 | else | |
1560 | ntfs_error(vi->i_sb, "Failed to lookup " | |
1561 | "$INDEX_ALLOCATION attribute."); | |
1562 | goto unm_err_out; | |
1563 | } | |
5ae9fcf8 | 1564 | if (!a->non_resident) { |
1da177e4 LT |
1565 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " |
1566 | "resident."); | |
1567 | goto unm_err_out; | |
1568 | } | |
1569 | /* | |
1570 | * Ensure the attribute name is placed before the mapping pairs array. | |
1571 | */ | |
5ae9fcf8 AA |
1572 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
1573 | le16_to_cpu( | |
1574 | a->data.non_resident.mapping_pairs_offset)))) { | |
1da177e4 LT |
1575 | ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is " |
1576 | "placed after the mapping pairs array."); | |
1577 | goto unm_err_out; | |
1578 | } | |
5ae9fcf8 | 1579 | if (a->flags & ATTR_IS_ENCRYPTED) { |
1da177e4 LT |
1580 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " |
1581 | "encrypted."); | |
1582 | goto unm_err_out; | |
1583 | } | |
5ae9fcf8 | 1584 | if (a->flags & ATTR_IS_SPARSE) { |
1da177e4 LT |
1585 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse."); |
1586 | goto unm_err_out; | |
1587 | } | |
5ae9fcf8 | 1588 | if (a->flags & ATTR_COMPRESSION_MASK) { |
1da177e4 LT |
1589 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " |
1590 | "compressed."); | |
1591 | goto unm_err_out; | |
1592 | } | |
5ae9fcf8 | 1593 | if (a->data.non_resident.lowest_vcn) { |
1da177e4 LT |
1594 | ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION " |
1595 | "attribute has non zero lowest_vcn."); | |
1596 | goto unm_err_out; | |
1597 | } | |
5ae9fcf8 | 1598 | vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); |
1da177e4 | 1599 | ni->initialized_size = sle64_to_cpu( |
5ae9fcf8 AA |
1600 | a->data.non_resident.initialized_size); |
1601 | ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size); | |
1da177e4 LT |
1602 | /* |
1603 | * We are done with the mft record, so we release it. Otherwise | |
1604 | * we would deadlock in ntfs_attr_iget(). | |
1605 | */ | |
1606 | ntfs_attr_put_search_ctx(ctx); | |
1607 | unmap_mft_record(base_ni); | |
1608 | m = NULL; | |
1609 | ctx = NULL; | |
1610 | /* Get the index bitmap attribute inode. */ | |
1611 | bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len); | |
1612 | if (IS_ERR(bvi)) { | |
1613 | ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); | |
1614 | err = PTR_ERR(bvi); | |
1615 | goto unm_err_out; | |
1616 | } | |
1617 | bni = NTFS_I(bvi); | |
1618 | if (NInoCompressed(bni) || NInoEncrypted(bni) || | |
1619 | NInoSparse(bni)) { | |
1620 | ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or " | |
1621 | "encrypted and/or sparse."); | |
1622 | goto iput_unm_err_out; | |
1623 | } | |
1624 | /* Consistency check bitmap size vs. index allocation size. */ | |
f50f3ac5 AA |
1625 | bvi_size = i_size_read(bvi); |
1626 | if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) { | |
1da177e4 | 1627 | ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for " |
f50f3ac5 | 1628 | "index allocation (0x%llx).", bvi_size << 3, |
1da177e4 LT |
1629 | vi->i_size); |
1630 | goto iput_unm_err_out; | |
1631 | } | |
1632 | ni->itype.index.bmp_ino = bvi; | |
1633 | skip_large_index_stuff: | |
1634 | /* Setup the operations for this index inode. */ | |
1635 | vi->i_op = NULL; | |
1636 | vi->i_fop = NULL; | |
1637 | vi->i_mapping->a_ops = &ntfs_mst_aops; | |
1638 | vi->i_blocks = ni->allocated_size >> 9; | |
1da177e4 LT |
1639 | /* |
1640 | * Make sure the base inode doesn't go away and attach it to the | |
1641 | * index inode. | |
1642 | */ | |
1643 | igrab(base_vi); | |
1644 | ni->ext.base_ntfs_ino = base_ni; | |
1645 | ni->nr_extents = -1; | |
1646 | ||
1647 | ntfs_debug("Done."); | |
1648 | return 0; | |
1649 | ||
1650 | iput_unm_err_out: | |
1651 | iput(bvi); | |
1652 | unm_err_out: | |
1653 | if (!err) | |
1654 | err = -EIO; | |
1655 | if (ctx) | |
1656 | ntfs_attr_put_search_ctx(ctx); | |
1657 | if (m) | |
1658 | unmap_mft_record(base_ni); | |
1659 | err_out: | |
1660 | ntfs_error(vi->i_sb, "Failed with error code %i while reading index " | |
1661 | "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino, | |
1662 | ni->name_len); | |
1663 | make_bad_inode(vi); | |
1664 | if (err != -EOPNOTSUPP && err != -ENOMEM) | |
1665 | NVolSetErrors(vol); | |
1666 | return err; | |
1667 | } | |
1668 | ||
1669 | /** | |
1670 | * ntfs_read_inode_mount - special read_inode for mount time use only | |
1671 | * @vi: inode to read | |
1672 | * | |
1673 | * Read inode FILE_MFT at mount time, only called with super_block lock | |
1674 | * held from within the read_super() code path. | |
1675 | * | |
1676 | * This function exists because when it is called the page cache for $MFT/$DATA | |
1677 | * is not initialized and hence we cannot get at the contents of mft records | |
1678 | * by calling map_mft_record*(). | |
1679 | * | |
1680 | * Further it needs to cope with the circular references problem, i.e. cannot | |
1681 | * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because | |
1682 | * we do not know where the other extent mft records are yet and again, because | |
1683 | * we cannot call map_mft_record*() yet. Obviously this applies only when an | |
1684 | * attribute list is actually present in $MFT inode. | |
1685 | * | |
1686 | * We solve these problems by starting with the $DATA attribute before anything | |
1687 | * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each | |
1688 | * extent is found, we ntfs_mapping_pairs_decompress() including the implied | |
1689 | * ntfs_runlists_merge(). Each step of the iteration necessarily provides | |
1690 | * sufficient information for the next step to complete. | |
1691 | * | |
1692 | * This should work but there are two possible pit falls (see inline comments | |
1693 | * below), but only time will tell if they are real pits or just smoke... | |
1694 | */ | |
1695 | int ntfs_read_inode_mount(struct inode *vi) | |
1696 | { | |
1697 | VCN next_vcn, last_vcn, highest_vcn; | |
1698 | s64 block; | |
1699 | struct super_block *sb = vi->i_sb; | |
1700 | ntfs_volume *vol = NTFS_SB(sb); | |
1701 | struct buffer_head *bh; | |
1702 | ntfs_inode *ni; | |
1703 | MFT_RECORD *m = NULL; | |
5ae9fcf8 | 1704 | ATTR_RECORD *a; |
1da177e4 LT |
1705 | ntfs_attr_search_ctx *ctx; |
1706 | unsigned int i, nr_blocks; | |
1707 | int err; | |
1708 | ||
1709 | ntfs_debug("Entering."); | |
1710 | ||
1711 | /* Initialize the ntfs specific part of @vi. */ | |
1712 | ntfs_init_big_inode(vi); | |
1713 | ||
1714 | ni = NTFS_I(vi); | |
1715 | ||
1716 | /* Setup the data attribute. It is special as it is mst protected. */ | |
1717 | NInoSetNonResident(ni); | |
1718 | NInoSetMstProtected(ni); | |
c002f425 | 1719 | NInoSetSparseDisabled(ni); |
1da177e4 LT |
1720 | ni->type = AT_DATA; |
1721 | ni->name = NULL; | |
1722 | ni->name_len = 0; | |
1da177e4 LT |
1723 | /* |
1724 | * This sets up our little cheat allowing us to reuse the async read io | |
1725 | * completion handler for directories. | |
1726 | */ | |
1727 | ni->itype.index.block_size = vol->mft_record_size; | |
1728 | ni->itype.index.block_size_bits = vol->mft_record_size_bits; | |
1729 | ||
1730 | /* Very important! Needed to be able to call map_mft_record*(). */ | |
1731 | vol->mft_ino = vi; | |
1732 | ||
1733 | /* Allocate enough memory to read the first mft record. */ | |
1734 | if (vol->mft_record_size > 64 * 1024) { | |
1735 | ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).", | |
1736 | vol->mft_record_size); | |
1737 | goto err_out; | |
1738 | } | |
1739 | i = vol->mft_record_size; | |
1740 | if (i < sb->s_blocksize) | |
1741 | i = sb->s_blocksize; | |
1742 | m = (MFT_RECORD*)ntfs_malloc_nofs(i); | |
1743 | if (!m) { | |
1744 | ntfs_error(sb, "Failed to allocate buffer for $MFT record 0."); | |
1745 | goto err_out; | |
1746 | } | |
1747 | ||
1748 | /* Determine the first block of the $MFT/$DATA attribute. */ | |
1749 | block = vol->mft_lcn << vol->cluster_size_bits >> | |
1750 | sb->s_blocksize_bits; | |
1751 | nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits; | |
1752 | if (!nr_blocks) | |
1753 | nr_blocks = 1; | |
1754 | ||
1755 | /* Load $MFT/$DATA's first mft record. */ | |
1756 | for (i = 0; i < nr_blocks; i++) { | |
1757 | bh = sb_bread(sb, block++); | |
1758 | if (!bh) { | |
1759 | ntfs_error(sb, "Device read failed."); | |
1760 | goto err_out; | |
1761 | } | |
1762 | memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data, | |
1763 | sb->s_blocksize); | |
1764 | brelse(bh); | |
1765 | } | |
1766 | ||
1767 | /* Apply the mst fixups. */ | |
1768 | if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) { | |
1769 | /* FIXME: Try to use the $MFTMirr now. */ | |
1770 | ntfs_error(sb, "MST fixup failed. $MFT is corrupt."); | |
1771 | goto err_out; | |
1772 | } | |
1773 | ||
1774 | /* Need this to sanity check attribute list references to $MFT. */ | |
1775 | vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number); | |
1776 | ||
1777 | /* Provides readpage() and sync_page() for map_mft_record(). */ | |
1778 | vi->i_mapping->a_ops = &ntfs_mst_aops; | |
1779 | ||
1780 | ctx = ntfs_attr_get_search_ctx(ni, m); | |
1781 | if (!ctx) { | |
1782 | err = -ENOMEM; | |
1783 | goto err_out; | |
1784 | } | |
1785 | ||
1786 | /* Find the attribute list attribute if present. */ | |
1787 | err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx); | |
1788 | if (err) { | |
1789 | if (unlikely(err != -ENOENT)) { | |
1790 | ntfs_error(sb, "Failed to lookup attribute list " | |
1791 | "attribute. You should run chkdsk."); | |
1792 | goto put_err_out; | |
1793 | } | |
1794 | } else /* if (!err) */ { | |
1795 | ATTR_LIST_ENTRY *al_entry, *next_al_entry; | |
1796 | u8 *al_end; | |
1797 | ||
1798 | ntfs_debug("Attribute list attribute found in $MFT."); | |
1799 | NInoSetAttrList(ni); | |
5ae9fcf8 AA |
1800 | a = ctx->attr; |
1801 | if (a->flags & ATTR_IS_ENCRYPTED || | |
1802 | a->flags & ATTR_COMPRESSION_MASK || | |
1803 | a->flags & ATTR_IS_SPARSE) { | |
1da177e4 LT |
1804 | ntfs_error(sb, "Attribute list attribute is " |
1805 | "compressed/encrypted/sparse. Not " | |
1806 | "allowed. $MFT is corrupt. You should " | |
1807 | "run chkdsk."); | |
1808 | goto put_err_out; | |
1809 | } | |
1810 | /* Now allocate memory for the attribute list. */ | |
5ae9fcf8 | 1811 | ni->attr_list_size = (u32)ntfs_attr_size(a); |
1da177e4 LT |
1812 | ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size); |
1813 | if (!ni->attr_list) { | |
1814 | ntfs_error(sb, "Not enough memory to allocate buffer " | |
1815 | "for attribute list."); | |
1816 | goto put_err_out; | |
1817 | } | |
5ae9fcf8 | 1818 | if (a->non_resident) { |
1da177e4 | 1819 | NInoSetAttrListNonResident(ni); |
5ae9fcf8 | 1820 | if (a->data.non_resident.lowest_vcn) { |
1da177e4 LT |
1821 | ntfs_error(sb, "Attribute list has non zero " |
1822 | "lowest_vcn. $MFT is corrupt. " | |
1823 | "You should run chkdsk."); | |
1824 | goto put_err_out; | |
1825 | } | |
1826 | /* Setup the runlist. */ | |
1827 | ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol, | |
5ae9fcf8 | 1828 | a, NULL); |
1da177e4 LT |
1829 | if (IS_ERR(ni->attr_list_rl.rl)) { |
1830 | err = PTR_ERR(ni->attr_list_rl.rl); | |
1831 | ni->attr_list_rl.rl = NULL; | |
1832 | ntfs_error(sb, "Mapping pairs decompression " | |
1833 | "failed with error code %i.", | |
1834 | -err); | |
1835 | goto put_err_out; | |
1836 | } | |
1837 | /* Now load the attribute list. */ | |
1838 | if ((err = load_attribute_list(vol, &ni->attr_list_rl, | |
1839 | ni->attr_list, ni->attr_list_size, | |
5ae9fcf8 | 1840 | sle64_to_cpu(a->data. |
1da177e4 LT |
1841 | non_resident.initialized_size)))) { |
1842 | ntfs_error(sb, "Failed to load attribute list " | |
1843 | "attribute with error code %i.", | |
1844 | -err); | |
1845 | goto put_err_out; | |
1846 | } | |
1847 | } else /* if (!ctx.attr->non_resident) */ { | |
5ae9fcf8 AA |
1848 | if ((u8*)a + le16_to_cpu( |
1849 | a->data.resident.value_offset) + | |
1da177e4 | 1850 | le32_to_cpu( |
5ae9fcf8 | 1851 | a->data.resident.value_length) > |
1da177e4 LT |
1852 | (u8*)ctx->mrec + vol->mft_record_size) { |
1853 | ntfs_error(sb, "Corrupt attribute list " | |
1854 | "attribute."); | |
1855 | goto put_err_out; | |
1856 | } | |
1857 | /* Now copy the attribute list. */ | |
5ae9fcf8 AA |
1858 | memcpy(ni->attr_list, (u8*)a + le16_to_cpu( |
1859 | a->data.resident.value_offset), | |
1da177e4 | 1860 | le32_to_cpu( |
5ae9fcf8 | 1861 | a->data.resident.value_length)); |
1da177e4 LT |
1862 | } |
1863 | /* The attribute list is now setup in memory. */ | |
1864 | /* | |
1865 | * FIXME: I don't know if this case is actually possible. | |
1866 | * According to logic it is not possible but I have seen too | |
1867 | * many weird things in MS software to rely on logic... Thus we | |
1868 | * perform a manual search and make sure the first $MFT/$DATA | |
1869 | * extent is in the base inode. If it is not we abort with an | |
1870 | * error and if we ever see a report of this error we will need | |
1871 | * to do some magic in order to have the necessary mft record | |
1872 | * loaded and in the right place in the page cache. But | |
1873 | * hopefully logic will prevail and this never happens... | |
1874 | */ | |
1875 | al_entry = (ATTR_LIST_ENTRY*)ni->attr_list; | |
1876 | al_end = (u8*)al_entry + ni->attr_list_size; | |
1877 | for (;; al_entry = next_al_entry) { | |
1878 | /* Out of bounds check. */ | |
1879 | if ((u8*)al_entry < ni->attr_list || | |
1880 | (u8*)al_entry > al_end) | |
1881 | goto em_put_err_out; | |
1882 | /* Catch the end of the attribute list. */ | |
1883 | if ((u8*)al_entry == al_end) | |
1884 | goto em_put_err_out; | |
1885 | if (!al_entry->length) | |
1886 | goto em_put_err_out; | |
1887 | if ((u8*)al_entry + 6 > al_end || (u8*)al_entry + | |
1888 | le16_to_cpu(al_entry->length) > al_end) | |
1889 | goto em_put_err_out; | |
1890 | next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry + | |
1891 | le16_to_cpu(al_entry->length)); | |
1892 | if (le32_to_cpu(al_entry->type) > | |
1893 | const_le32_to_cpu(AT_DATA)) | |
1894 | goto em_put_err_out; | |
1895 | if (AT_DATA != al_entry->type) | |
1896 | continue; | |
1897 | /* We want an unnamed attribute. */ | |
1898 | if (al_entry->name_length) | |
1899 | goto em_put_err_out; | |
1900 | /* Want the first entry, i.e. lowest_vcn == 0. */ | |
1901 | if (al_entry->lowest_vcn) | |
1902 | goto em_put_err_out; | |
1903 | /* First entry has to be in the base mft record. */ | |
1904 | if (MREF_LE(al_entry->mft_reference) != vi->i_ino) { | |
1905 | /* MFT references do not match, logic fails. */ | |
1906 | ntfs_error(sb, "BUG: The first $DATA extent " | |
1907 | "of $MFT is not in the base " | |
1908 | "mft record. Please report " | |
1909 | "you saw this message to " | |
1910 | "linux-ntfs-dev@lists." | |
1911 | "sourceforge.net"); | |
1912 | goto put_err_out; | |
1913 | } else { | |
1914 | /* Sequence numbers must match. */ | |
1915 | if (MSEQNO_LE(al_entry->mft_reference) != | |
1916 | ni->seq_no) | |
1917 | goto em_put_err_out; | |
1918 | /* Got it. All is ok. We can stop now. */ | |
1919 | break; | |
1920 | } | |
1921 | } | |
1922 | } | |
1923 | ||
1924 | ntfs_attr_reinit_search_ctx(ctx); | |
1925 | ||
1926 | /* Now load all attribute extents. */ | |
5ae9fcf8 | 1927 | a = NULL; |
1da177e4 LT |
1928 | next_vcn = last_vcn = highest_vcn = 0; |
1929 | while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0, | |
1930 | ctx))) { | |
1931 | runlist_element *nrl; | |
1932 | ||
1933 | /* Cache the current attribute. */ | |
5ae9fcf8 | 1934 | a = ctx->attr; |
1da177e4 | 1935 | /* $MFT must be non-resident. */ |
5ae9fcf8 | 1936 | if (!a->non_resident) { |
1da177e4 LT |
1937 | ntfs_error(sb, "$MFT must be non-resident but a " |
1938 | "resident extent was found. $MFT is " | |
1939 | "corrupt. Run chkdsk."); | |
1940 | goto put_err_out; | |
1941 | } | |
1942 | /* $MFT must be uncompressed and unencrypted. */ | |
5ae9fcf8 AA |
1943 | if (a->flags & ATTR_COMPRESSION_MASK || |
1944 | a->flags & ATTR_IS_ENCRYPTED || | |
1945 | a->flags & ATTR_IS_SPARSE) { | |
1da177e4 LT |
1946 | ntfs_error(sb, "$MFT must be uncompressed, " |
1947 | "non-sparse, and unencrypted but a " | |
1948 | "compressed/sparse/encrypted extent " | |
1949 | "was found. $MFT is corrupt. Run " | |
1950 | "chkdsk."); | |
1951 | goto put_err_out; | |
1952 | } | |
1953 | /* | |
1954 | * Decompress the mapping pairs array of this extent and merge | |
1955 | * the result into the existing runlist. No need for locking | |
1956 | * as we have exclusive access to the inode at this time and we | |
1957 | * are a mount in progress task, too. | |
1958 | */ | |
5ae9fcf8 | 1959 | nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl); |
1da177e4 LT |
1960 | if (IS_ERR(nrl)) { |
1961 | ntfs_error(sb, "ntfs_mapping_pairs_decompress() " | |
1962 | "failed with error code %ld. $MFT is " | |
1963 | "corrupt.", PTR_ERR(nrl)); | |
1964 | goto put_err_out; | |
1965 | } | |
1966 | ni->runlist.rl = nrl; | |
1967 | ||
1968 | /* Are we in the first extent? */ | |
1969 | if (!next_vcn) { | |
5ae9fcf8 | 1970 | if (a->data.non_resident.lowest_vcn) { |
1da177e4 LT |
1971 | ntfs_error(sb, "First extent of $DATA " |
1972 | "attribute has non zero " | |
1973 | "lowest_vcn. $MFT is corrupt. " | |
1974 | "You should run chkdsk."); | |
1975 | goto put_err_out; | |
1976 | } | |
1977 | /* Get the last vcn in the $DATA attribute. */ | |
1978 | last_vcn = sle64_to_cpu( | |
5ae9fcf8 | 1979 | a->data.non_resident.allocated_size) |
1da177e4 LT |
1980 | >> vol->cluster_size_bits; |
1981 | /* Fill in the inode size. */ | |
1982 | vi->i_size = sle64_to_cpu( | |
5ae9fcf8 AA |
1983 | a->data.non_resident.data_size); |
1984 | ni->initialized_size = sle64_to_cpu( | |
1985 | a->data.non_resident.initialized_size); | |
1da177e4 | 1986 | ni->allocated_size = sle64_to_cpu( |
5ae9fcf8 | 1987 | a->data.non_resident.allocated_size); |
1da177e4 LT |
1988 | /* |
1989 | * Verify the number of mft records does not exceed | |
1990 | * 2^32 - 1. | |
1991 | */ | |
1992 | if ((vi->i_size >> vol->mft_record_size_bits) >= | |
1993 | (1ULL << 32)) { | |
1994 | ntfs_error(sb, "$MFT is too big! Aborting."); | |
1995 | goto put_err_out; | |
1996 | } | |
1997 | /* | |
1998 | * We have got the first extent of the runlist for | |
1999 | * $MFT which means it is now relatively safe to call | |
2000 | * the normal ntfs_read_inode() function. | |
2001 | * Complete reading the inode, this will actually | |
2002 | * re-read the mft record for $MFT, this time entering | |
2003 | * it into the page cache with which we complete the | |
2004 | * kick start of the volume. It should be safe to do | |
2005 | * this now as the first extent of $MFT/$DATA is | |
2006 | * already known and we would hope that we don't need | |
2007 | * further extents in order to find the other | |
2008 | * attributes belonging to $MFT. Only time will tell if | |
2009 | * this is really the case. If not we will have to play | |
2010 | * magic at this point, possibly duplicating a lot of | |
2011 | * ntfs_read_inode() at this point. We will need to | |
2012 | * ensure we do enough of its work to be able to call | |
2013 | * ntfs_read_inode() on extents of $MFT/$DATA. But lets | |
2014 | * hope this never happens... | |
2015 | */ | |
2016 | ntfs_read_locked_inode(vi); | |
2017 | if (is_bad_inode(vi)) { | |
2018 | ntfs_error(sb, "ntfs_read_inode() of $MFT " | |
2019 | "failed. BUG or corrupt $MFT. " | |
2020 | "Run chkdsk and if no errors " | |
2021 | "are found, please report you " | |
2022 | "saw this message to " | |
2023 | "linux-ntfs-dev@lists." | |
2024 | "sourceforge.net"); | |
2025 | ntfs_attr_put_search_ctx(ctx); | |
2026 | /* Revert to the safe super operations. */ | |
2027 | ntfs_free(m); | |
2028 | return -1; | |
2029 | } | |
2030 | /* | |
2031 | * Re-initialize some specifics about $MFT's inode as | |
2032 | * ntfs_read_inode() will have set up the default ones. | |
2033 | */ | |
2034 | /* Set uid and gid to root. */ | |
2035 | vi->i_uid = vi->i_gid = 0; | |
2036 | /* Regular file. No access for anyone. */ | |
2037 | vi->i_mode = S_IFREG; | |
2038 | /* No VFS initiated operations allowed for $MFT. */ | |
2039 | vi->i_op = &ntfs_empty_inode_ops; | |
2040 | vi->i_fop = &ntfs_empty_file_ops; | |
2041 | } | |
2042 | ||
2043 | /* Get the lowest vcn for the next extent. */ | |
5ae9fcf8 | 2044 | highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn); |
1da177e4 LT |
2045 | next_vcn = highest_vcn + 1; |
2046 | ||
2047 | /* Only one extent or error, which we catch below. */ | |
2048 | if (next_vcn <= 0) | |
2049 | break; | |
2050 | ||
2051 | /* Avoid endless loops due to corruption. */ | |
2052 | if (next_vcn < sle64_to_cpu( | |
5ae9fcf8 | 2053 | a->data.non_resident.lowest_vcn)) { |
1da177e4 LT |
2054 | ntfs_error(sb, "$MFT has corrupt attribute list " |
2055 | "attribute. Run chkdsk."); | |
2056 | goto put_err_out; | |
2057 | } | |
2058 | } | |
2059 | if (err != -ENOENT) { | |
2060 | ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. " | |
2061 | "$MFT is corrupt. Run chkdsk."); | |
2062 | goto put_err_out; | |
2063 | } | |
5ae9fcf8 | 2064 | if (!a) { |
1da177e4 LT |
2065 | ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is " |
2066 | "corrupt. Run chkdsk."); | |
2067 | goto put_err_out; | |
2068 | } | |
2069 | if (highest_vcn && highest_vcn != last_vcn - 1) { | |
2070 | ntfs_error(sb, "Failed to load the complete runlist for " | |
2071 | "$MFT/$DATA. Driver bug or corrupt $MFT. " | |
2072 | "Run chkdsk."); | |
2073 | ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx", | |
2074 | (unsigned long long)highest_vcn, | |
2075 | (unsigned long long)last_vcn - 1); | |
2076 | goto put_err_out; | |
2077 | } | |
2078 | ntfs_attr_put_search_ctx(ctx); | |
2079 | ntfs_debug("Done."); | |
2080 | ntfs_free(m); | |
2081 | return 0; | |
2082 | ||
2083 | em_put_err_out: | |
2084 | ntfs_error(sb, "Couldn't find first extent of $DATA attribute in " | |
2085 | "attribute list. $MFT is corrupt. Run chkdsk."); | |
2086 | put_err_out: | |
2087 | ntfs_attr_put_search_ctx(ctx); | |
2088 | err_out: | |
2089 | ntfs_error(sb, "Failed. Marking inode as bad."); | |
2090 | make_bad_inode(vi); | |
2091 | ntfs_free(m); | |
2092 | return -1; | |
2093 | } | |
2094 | ||
2095 | /** | |
2096 | * ntfs_put_inode - handler for when the inode reference count is decremented | |
2097 | * @vi: vfs inode | |
2098 | * | |
2099 | * The VFS calls ntfs_put_inode() every time the inode reference count (i_count) | |
2100 | * is about to be decremented (but before the decrement itself. | |
2101 | * | |
2102 | * If the inode @vi is a directory with two references, one of which is being | |
2103 | * dropped, we need to put the attribute inode for the directory index bitmap, | |
2104 | * if it is present, otherwise the directory inode would remain pinned for | |
2105 | * ever. | |
2106 | */ | |
2107 | void ntfs_put_inode(struct inode *vi) | |
2108 | { | |
2109 | if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) { | |
2110 | ntfs_inode *ni = NTFS_I(vi); | |
2111 | if (NInoIndexAllocPresent(ni)) { | |
2112 | struct inode *bvi = NULL; | |
2113 | down(&vi->i_sem); | |
2114 | if (atomic_read(&vi->i_count) == 2) { | |
2115 | bvi = ni->itype.index.bmp_ino; | |
2116 | if (bvi) | |
2117 | ni->itype.index.bmp_ino = NULL; | |
2118 | } | |
2119 | up(&vi->i_sem); | |
2120 | if (bvi) | |
2121 | iput(bvi); | |
2122 | } | |
2123 | } | |
2124 | } | |
2125 | ||
2126 | static void __ntfs_clear_inode(ntfs_inode *ni) | |
2127 | { | |
2128 | /* Free all alocated memory. */ | |
2129 | down_write(&ni->runlist.lock); | |
2130 | if (ni->runlist.rl) { | |
2131 | ntfs_free(ni->runlist.rl); | |
2132 | ni->runlist.rl = NULL; | |
2133 | } | |
2134 | up_write(&ni->runlist.lock); | |
2135 | ||
2136 | if (ni->attr_list) { | |
2137 | ntfs_free(ni->attr_list); | |
2138 | ni->attr_list = NULL; | |
2139 | } | |
2140 | ||
2141 | down_write(&ni->attr_list_rl.lock); | |
2142 | if (ni->attr_list_rl.rl) { | |
2143 | ntfs_free(ni->attr_list_rl.rl); | |
2144 | ni->attr_list_rl.rl = NULL; | |
2145 | } | |
2146 | up_write(&ni->attr_list_rl.lock); | |
2147 | ||
2148 | if (ni->name_len && ni->name != I30) { | |
2149 | /* Catch bugs... */ | |
2150 | BUG_ON(!ni->name); | |
2151 | kfree(ni->name); | |
2152 | } | |
2153 | } | |
2154 | ||
2155 | void ntfs_clear_extent_inode(ntfs_inode *ni) | |
2156 | { | |
2157 | ntfs_debug("Entering for inode 0x%lx.", ni->mft_no); | |
2158 | ||
2159 | BUG_ON(NInoAttr(ni)); | |
2160 | BUG_ON(ni->nr_extents != -1); | |
2161 | ||
2162 | #ifdef NTFS_RW | |
2163 | if (NInoDirty(ni)) { | |
2164 | if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino))) | |
2165 | ntfs_error(ni->vol->sb, "Clearing dirty extent inode! " | |
2166 | "Losing data! This is a BUG!!!"); | |
2167 | // FIXME: Do something!!! | |
2168 | } | |
2169 | #endif /* NTFS_RW */ | |
2170 | ||
2171 | __ntfs_clear_inode(ni); | |
2172 | ||
2173 | /* Bye, bye... */ | |
2174 | ntfs_destroy_extent_inode(ni); | |
2175 | } | |
2176 | ||
2177 | /** | |
2178 | * ntfs_clear_big_inode - clean up the ntfs specific part of an inode | |
2179 | * @vi: vfs inode pending annihilation | |
2180 | * | |
2181 | * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode() | |
2182 | * is called, which deallocates all memory belonging to the NTFS specific part | |
2183 | * of the inode and returns. | |
2184 | * | |
2185 | * If the MFT record is dirty, we commit it before doing anything else. | |
2186 | */ | |
2187 | void ntfs_clear_big_inode(struct inode *vi) | |
2188 | { | |
2189 | ntfs_inode *ni = NTFS_I(vi); | |
2190 | ||
2191 | /* | |
2192 | * If the inode @vi is an index inode we need to put the attribute | |
2193 | * inode for the index bitmap, if it is present, otherwise the index | |
2194 | * inode would disappear and the attribute inode for the index bitmap | |
2195 | * would no longer be referenced from anywhere and thus it would remain | |
2196 | * pinned for ever. | |
2197 | */ | |
2198 | if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) && | |
2199 | NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) { | |
2200 | iput(ni->itype.index.bmp_ino); | |
2201 | ni->itype.index.bmp_ino = NULL; | |
2202 | } | |
2203 | #ifdef NTFS_RW | |
2204 | if (NInoDirty(ni)) { | |
2205 | BOOL was_bad = (is_bad_inode(vi)); | |
2206 | ||
2207 | /* Committing the inode also commits all extent inodes. */ | |
2208 | ntfs_commit_inode(vi); | |
2209 | ||
2210 | if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) { | |
2211 | ntfs_error(vi->i_sb, "Failed to commit dirty inode " | |
2212 | "0x%lx. Losing data!", vi->i_ino); | |
2213 | // FIXME: Do something!!! | |
2214 | } | |
2215 | } | |
2216 | #endif /* NTFS_RW */ | |
2217 | ||
2218 | /* No need to lock at this stage as no one else has a reference. */ | |
2219 | if (ni->nr_extents > 0) { | |
2220 | int i; | |
2221 | ||
2222 | for (i = 0; i < ni->nr_extents; i++) | |
2223 | ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]); | |
2224 | kfree(ni->ext.extent_ntfs_inos); | |
2225 | } | |
2226 | ||
2227 | __ntfs_clear_inode(ni); | |
2228 | ||
2229 | if (NInoAttr(ni)) { | |
2230 | /* Release the base inode if we are holding it. */ | |
2231 | if (ni->nr_extents == -1) { | |
2232 | iput(VFS_I(ni->ext.base_ntfs_ino)); | |
2233 | ni->nr_extents = 0; | |
2234 | ni->ext.base_ntfs_ino = NULL; | |
2235 | } | |
2236 | } | |
2237 | return; | |
2238 | } | |
2239 | ||
2240 | /** | |
2241 | * ntfs_show_options - show mount options in /proc/mounts | |
2242 | * @sf: seq_file in which to write our mount options | |
2243 | * @mnt: vfs mount whose mount options to display | |
2244 | * | |
2245 | * Called by the VFS once for each mounted ntfs volume when someone reads | |
2246 | * /proc/mounts in order to display the NTFS specific mount options of each | |
2247 | * mount. The mount options of the vfs mount @mnt are written to the seq file | |
2248 | * @sf and success is returned. | |
2249 | */ | |
2250 | int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt) | |
2251 | { | |
2252 | ntfs_volume *vol = NTFS_SB(mnt->mnt_sb); | |
2253 | int i; | |
2254 | ||
2255 | seq_printf(sf, ",uid=%i", vol->uid); | |
2256 | seq_printf(sf, ",gid=%i", vol->gid); | |
2257 | if (vol->fmask == vol->dmask) | |
2258 | seq_printf(sf, ",umask=0%o", vol->fmask); | |
2259 | else { | |
2260 | seq_printf(sf, ",fmask=0%o", vol->fmask); | |
2261 | seq_printf(sf, ",dmask=0%o", vol->dmask); | |
2262 | } | |
2263 | seq_printf(sf, ",nls=%s", vol->nls_map->charset); | |
2264 | if (NVolCaseSensitive(vol)) | |
2265 | seq_printf(sf, ",case_sensitive"); | |
2266 | if (NVolShowSystemFiles(vol)) | |
2267 | seq_printf(sf, ",show_sys_files"); | |
c002f425 AA |
2268 | if (!NVolSparseEnabled(vol)) |
2269 | seq_printf(sf, ",disable_sparse"); | |
1da177e4 LT |
2270 | for (i = 0; on_errors_arr[i].val; i++) { |
2271 | if (on_errors_arr[i].val & vol->on_errors) | |
2272 | seq_printf(sf, ",errors=%s", on_errors_arr[i].str); | |
2273 | } | |
2274 | seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier); | |
2275 | return 0; | |
2276 | } | |
2277 | ||
2278 | #ifdef NTFS_RW | |
2279 | ||
2280 | /** | |
2281 | * ntfs_truncate - called when the i_size of an ntfs inode is changed | |
2282 | * @vi: inode for which the i_size was changed | |
2283 | * | |
2284 | * We do not support i_size changes yet. | |
2285 | * | |
2286 | * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and | |
2287 | * that the change is allowed. | |
2288 | * | |
2289 | * This implies for us that @vi is a file inode rather than a directory, index, | |
2290 | * or attribute inode as well as that @vi is a base inode. | |
2291 | * | |
2292 | * Returns 0 on success or -errno on error. | |
2293 | * | |
2294 | * Called with ->i_sem held. In all but one case ->i_alloc_sem is held for | |
2295 | * writing. The only case where ->i_alloc_sem is not held is | |
2296 | * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called | |
2297 | * with the current i_size as the offset which means that it is a noop as far | |
2298 | * as ntfs_truncate() is concerned. | |
2299 | */ | |
2300 | int ntfs_truncate(struct inode *vi) | |
2301 | { | |
2302 | ntfs_inode *ni = NTFS_I(vi); | |
2303 | ntfs_volume *vol = ni->vol; | |
2304 | ntfs_attr_search_ctx *ctx; | |
2305 | MFT_RECORD *m; | |
5ae9fcf8 | 2306 | ATTR_RECORD *a; |
1da177e4 LT |
2307 | const char *te = " Leaving file length out of sync with i_size."; |
2308 | int err; | |
2309 | ||
2310 | ntfs_debug("Entering for inode 0x%lx.", vi->i_ino); | |
2311 | BUG_ON(NInoAttr(ni)); | |
2312 | BUG_ON(ni->nr_extents < 0); | |
2313 | m = map_mft_record(ni); | |
2314 | if (IS_ERR(m)) { | |
2315 | err = PTR_ERR(m); | |
2316 | ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx " | |
2317 | "(error code %d).%s", vi->i_ino, err, te); | |
2318 | ctx = NULL; | |
2319 | m = NULL; | |
2320 | goto err_out; | |
2321 | } | |
2322 | ctx = ntfs_attr_get_search_ctx(ni, m); | |
2323 | if (unlikely(!ctx)) { | |
2324 | ntfs_error(vi->i_sb, "Failed to allocate a search context for " | |
2325 | "inode 0x%lx (not enough memory).%s", | |
2326 | vi->i_ino, te); | |
2327 | err = -ENOMEM; | |
2328 | goto err_out; | |
2329 | } | |
2330 | err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len, | |
2331 | CASE_SENSITIVE, 0, NULL, 0, ctx); | |
2332 | if (unlikely(err)) { | |
2333 | if (err == -ENOENT) | |
2334 | ntfs_error(vi->i_sb, "Open attribute is missing from " | |
2335 | "mft record. Inode 0x%lx is corrupt. " | |
2336 | "Run chkdsk.", vi->i_ino); | |
2337 | else | |
2338 | ntfs_error(vi->i_sb, "Failed to lookup attribute in " | |
2339 | "inode 0x%lx (error code %d).", | |
2340 | vi->i_ino, err); | |
2341 | goto err_out; | |
2342 | } | |
5ae9fcf8 | 2343 | a = ctx->attr; |
1da177e4 | 2344 | /* If the size has not changed there is nothing to do. */ |
5ae9fcf8 | 2345 | if (ntfs_attr_size(a) == i_size_read(vi)) |
1da177e4 LT |
2346 | goto done; |
2347 | // TODO: Implement the truncate... | |
2348 | ntfs_error(vi->i_sb, "Inode size has changed but this is not " | |
2349 | "implemented yet. Resetting inode size to old value. " | |
2350 | " This is most likely a bug in the ntfs driver!"); | |
5ae9fcf8 | 2351 | i_size_write(vi, ntfs_attr_size(a)); |
1da177e4 LT |
2352 | done: |
2353 | ntfs_attr_put_search_ctx(ctx); | |
2354 | unmap_mft_record(ni); | |
2355 | NInoClearTruncateFailed(ni); | |
2356 | ntfs_debug("Done."); | |
2357 | return 0; | |
2358 | err_out: | |
2359 | if (err != -ENOMEM) { | |
2360 | NVolSetErrors(vol); | |
2361 | make_bad_inode(vi); | |
2362 | } | |
2363 | if (ctx) | |
2364 | ntfs_attr_put_search_ctx(ctx); | |
2365 | if (m) | |
2366 | unmap_mft_record(ni); | |
2367 | NInoSetTruncateFailed(ni); | |
2368 | return err; | |
2369 | } | |
2370 | ||
2371 | /** | |
2372 | * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value | |
2373 | * @vi: inode for which the i_size was changed | |
2374 | * | |
2375 | * Wrapper for ntfs_truncate() that has no return value. | |
2376 | * | |
2377 | * See ntfs_truncate() description above for details. | |
2378 | */ | |
2379 | void ntfs_truncate_vfs(struct inode *vi) { | |
2380 | ntfs_truncate(vi); | |
2381 | } | |
2382 | ||
2383 | /** | |
2384 | * ntfs_setattr - called from notify_change() when an attribute is being changed | |
2385 | * @dentry: dentry whose attributes to change | |
2386 | * @attr: structure describing the attributes and the changes | |
2387 | * | |
2388 | * We have to trap VFS attempts to truncate the file described by @dentry as | |
2389 | * soon as possible, because we do not implement changes in i_size yet. So we | |
2390 | * abort all i_size changes here. | |
2391 | * | |
2392 | * We also abort all changes of user, group, and mode as we do not implement | |
2393 | * the NTFS ACLs yet. | |
2394 | * | |
2395 | * Called with ->i_sem held. For the ATTR_SIZE (i.e. ->truncate) case, also | |
2396 | * called with ->i_alloc_sem held for writing. | |
2397 | * | |
2398 | * Basically this is a copy of generic notify_change() and inode_setattr() | |
2399 | * functionality, except we intercept and abort changes in i_size. | |
2400 | */ | |
2401 | int ntfs_setattr(struct dentry *dentry, struct iattr *attr) | |
2402 | { | |
2403 | struct inode *vi = dentry->d_inode; | |
2404 | int err; | |
2405 | unsigned int ia_valid = attr->ia_valid; | |
2406 | ||
2407 | err = inode_change_ok(vi, attr); | |
2408 | if (err) | |
2409 | return err; | |
2410 | ||
2411 | /* We do not support NTFS ACLs yet. */ | |
2412 | if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) { | |
2413 | ntfs_warning(vi->i_sb, "Changes in user/group/mode are not " | |
2414 | "supported yet, ignoring."); | |
2415 | err = -EOPNOTSUPP; | |
2416 | goto out; | |
2417 | } | |
2418 | ||
2419 | if (ia_valid & ATTR_SIZE) { | |
2420 | if (attr->ia_size != i_size_read(vi)) { | |
2421 | ntfs_warning(vi->i_sb, "Changes in inode size are not " | |
2422 | "supported yet, ignoring."); | |
2423 | err = -EOPNOTSUPP; | |
2424 | // TODO: Implement... | |
2425 | // err = vmtruncate(vi, attr->ia_size); | |
2426 | if (err || ia_valid == ATTR_SIZE) | |
2427 | goto out; | |
2428 | } else { | |
2429 | /* | |
2430 | * We skipped the truncate but must still update | |
2431 | * timestamps. | |
2432 | */ | |
2433 | ia_valid |= ATTR_MTIME|ATTR_CTIME; | |
2434 | } | |
2435 | } | |
2436 | ||
2437 | if (ia_valid & ATTR_ATIME) | |
2438 | vi->i_atime = attr->ia_atime; | |
2439 | if (ia_valid & ATTR_MTIME) | |
2440 | vi->i_mtime = attr->ia_mtime; | |
2441 | if (ia_valid & ATTR_CTIME) | |
2442 | vi->i_ctime = attr->ia_ctime; | |
2443 | mark_inode_dirty(vi); | |
2444 | out: | |
2445 | return err; | |
2446 | } | |
2447 | ||
2448 | /** | |
2449 | * ntfs_write_inode - write out a dirty inode | |
2450 | * @vi: inode to write out | |
2451 | * @sync: if true, write out synchronously | |
2452 | * | |
2453 | * Write out a dirty inode to disk including any extent inodes if present. | |
2454 | * | |
2455 | * If @sync is true, commit the inode to disk and wait for io completion. This | |
2456 | * is done using write_mft_record(). | |
2457 | * | |
2458 | * If @sync is false, just schedule the write to happen but do not wait for i/o | |
2459 | * completion. In 2.6 kernels, scheduling usually happens just by virtue of | |
2460 | * marking the page (and in this case mft record) dirty but we do not implement | |
2461 | * this yet as write_mft_record() largely ignores the @sync parameter and | |
2462 | * always performs synchronous writes. | |
2463 | * | |
2464 | * Return 0 on success and -errno on error. | |
2465 | */ | |
2466 | int ntfs_write_inode(struct inode *vi, int sync) | |
2467 | { | |
2468 | sle64 nt; | |
2469 | ntfs_inode *ni = NTFS_I(vi); | |
2470 | ntfs_attr_search_ctx *ctx; | |
2471 | MFT_RECORD *m; | |
2472 | STANDARD_INFORMATION *si; | |
2473 | int err = 0; | |
2474 | BOOL modified = FALSE; | |
2475 | ||
2476 | ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "", | |
2477 | vi->i_ino); | |
2478 | /* | |
2479 | * Dirty attribute inodes are written via their real inodes so just | |
2480 | * clean them here. Access time updates are taken care off when the | |
2481 | * real inode is written. | |
2482 | */ | |
2483 | if (NInoAttr(ni)) { | |
2484 | NInoClearDirty(ni); | |
2485 | ntfs_debug("Done."); | |
2486 | return 0; | |
2487 | } | |
2488 | /* Map, pin, and lock the mft record belonging to the inode. */ | |
2489 | m = map_mft_record(ni); | |
2490 | if (IS_ERR(m)) { | |
2491 | err = PTR_ERR(m); | |
2492 | goto err_out; | |
2493 | } | |
2494 | /* Update the access times in the standard information attribute. */ | |
2495 | ctx = ntfs_attr_get_search_ctx(ni, m); | |
2496 | if (unlikely(!ctx)) { | |
2497 | err = -ENOMEM; | |
2498 | goto unm_err_out; | |
2499 | } | |
2500 | err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, | |
2501 | CASE_SENSITIVE, 0, NULL, 0, ctx); | |
2502 | if (unlikely(err)) { | |
2503 | ntfs_attr_put_search_ctx(ctx); | |
2504 | goto unm_err_out; | |
2505 | } | |
2506 | si = (STANDARD_INFORMATION*)((u8*)ctx->attr + | |
2507 | le16_to_cpu(ctx->attr->data.resident.value_offset)); | |
2508 | /* Update the access times if they have changed. */ | |
2509 | nt = utc2ntfs(vi->i_mtime); | |
2510 | if (si->last_data_change_time != nt) { | |
2511 | ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, " | |
8907547d | 2512 | "new = 0x%llx", vi->i_ino, (long long) |
1da177e4 | 2513 | sle64_to_cpu(si->last_data_change_time), |
8907547d | 2514 | (long long)sle64_to_cpu(nt)); |
1da177e4 LT |
2515 | si->last_data_change_time = nt; |
2516 | modified = TRUE; | |
2517 | } | |
2518 | nt = utc2ntfs(vi->i_ctime); | |
2519 | if (si->last_mft_change_time != nt) { | |
2520 | ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, " | |
8907547d | 2521 | "new = 0x%llx", vi->i_ino, (long long) |
1da177e4 | 2522 | sle64_to_cpu(si->last_mft_change_time), |
8907547d | 2523 | (long long)sle64_to_cpu(nt)); |
1da177e4 LT |
2524 | si->last_mft_change_time = nt; |
2525 | modified = TRUE; | |
2526 | } | |
2527 | nt = utc2ntfs(vi->i_atime); | |
2528 | if (si->last_access_time != nt) { | |
2529 | ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, " | |
2530 | "new = 0x%llx", vi->i_ino, | |
8907547d RD |
2531 | (long long)sle64_to_cpu(si->last_access_time), |
2532 | (long long)sle64_to_cpu(nt)); | |
1da177e4 LT |
2533 | si->last_access_time = nt; |
2534 | modified = TRUE; | |
2535 | } | |
2536 | /* | |
2537 | * If we just modified the standard information attribute we need to | |
2538 | * mark the mft record it is in dirty. We do this manually so that | |
2539 | * mark_inode_dirty() is not called which would redirty the inode and | |
2540 | * hence result in an infinite loop of trying to write the inode. | |
2541 | * There is no need to mark the base inode nor the base mft record | |
2542 | * dirty, since we are going to write this mft record below in any case | |
2543 | * and the base mft record may actually not have been modified so it | |
2544 | * might not need to be written out. | |
2545 | * NOTE: It is not a problem when the inode for $MFT itself is being | |
2546 | * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES | |
2547 | * on the $MFT inode and hence ntfs_write_inode() will not be | |
2548 | * re-invoked because of it which in turn is ok since the dirtied mft | |
2549 | * record will be cleaned and written out to disk below, i.e. before | |
2550 | * this function returns. | |
2551 | */ | |
2552 | if (modified && !NInoTestSetDirty(ctx->ntfs_ino)) | |
2553 | mark_ntfs_record_dirty(ctx->ntfs_ino->page, | |
2554 | ctx->ntfs_ino->page_ofs); | |
2555 | ntfs_attr_put_search_ctx(ctx); | |
2556 | /* Now the access times are updated, write the base mft record. */ | |
2557 | if (NInoDirty(ni)) | |
2558 | err = write_mft_record(ni, m, sync); | |
2559 | /* Write all attached extent mft records. */ | |
2560 | down(&ni->extent_lock); | |
2561 | if (ni->nr_extents > 0) { | |
2562 | ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos; | |
2563 | int i; | |
2564 | ||
2565 | ntfs_debug("Writing %i extent inodes.", ni->nr_extents); | |
2566 | for (i = 0; i < ni->nr_extents; i++) { | |
2567 | ntfs_inode *tni = extent_nis[i]; | |
2568 | ||
2569 | if (NInoDirty(tni)) { | |
2570 | MFT_RECORD *tm = map_mft_record(tni); | |
2571 | int ret; | |
2572 | ||
2573 | if (IS_ERR(tm)) { | |
2574 | if (!err || err == -ENOMEM) | |
2575 | err = PTR_ERR(tm); | |
2576 | continue; | |
2577 | } | |
2578 | ret = write_mft_record(tni, tm, sync); | |
2579 | unmap_mft_record(tni); | |
2580 | if (unlikely(ret)) { | |
2581 | if (!err || err == -ENOMEM) | |
2582 | err = ret; | |
2583 | } | |
2584 | } | |
2585 | } | |
2586 | } | |
2587 | up(&ni->extent_lock); | |
2588 | unmap_mft_record(ni); | |
2589 | if (unlikely(err)) | |
2590 | goto err_out; | |
2591 | ntfs_debug("Done."); | |
2592 | return 0; | |
2593 | unm_err_out: | |
2594 | unmap_mft_record(ni); | |
2595 | err_out: | |
2596 | if (err == -ENOMEM) { | |
2597 | ntfs_warning(vi->i_sb, "Not enough memory to write inode. " | |
2598 | "Marking the inode dirty again, so the VFS " | |
2599 | "retries later."); | |
2600 | mark_inode_dirty(vi); | |
2601 | } else { | |
2602 | ntfs_error(vi->i_sb, "Failed (error code %i): Marking inode " | |
2603 | "as bad. You should run chkdsk.", -err); | |
2604 | make_bad_inode(vi); | |
2605 | NVolSetErrors(ni->vol); | |
2606 | } | |
2607 | return err; | |
2608 | } | |
2609 | ||
2610 | #endif /* NTFS_RW */ |