1 /**
2  * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project.
3  *
4  * Copyright (c) 2001-2007 Anton Altaparmakov
5  *
6  * This program/include file is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU General Public License as published
8  * by the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  *
11  * This program/include file is distributed in the hope that it will be
12  * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
13  * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License
17  * along with this program (in the main directory of the Linux-NTFS
18  * distribution in the file COPYING); if not, write to the Free Software
19  * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
20  */
21 
22 #include <linux/buffer_head.h>
23 #include <linux/fs.h>
24 #include <linux/mm.h>
25 #include <linux/mount.h>
26 #include <linux/mutex.h>
27 #include <linux/pagemap.h>
28 #include <linux/quotaops.h>
29 #include <linux/slab.h>
30 #include <linux/log2.h>
31 
32 #include "aops.h"
33 #include "attrib.h"
34 #include "bitmap.h"
35 #include "dir.h"
36 #include "debug.h"
37 #include "inode.h"
38 #include "lcnalloc.h"
39 #include "malloc.h"
40 #include "mft.h"
41 #include "time.h"
42 #include "ntfs.h"
43 
44 /**
45  * ntfs_test_inode - compare two (possibly fake) inodes for equality
46  * @vi:		vfs inode which to test
47  * @na:		ntfs attribute which is being tested with
48  *
49  * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
50  * inode @vi for equality with the ntfs attribute @na.
51  *
52  * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
53  * @na->name and @na->name_len are then ignored.
54  *
55  * Return 1 if the attributes match and 0 if not.
56  *
57  * NOTE: This function runs with the inode->i_lock spin lock held so it is not
58  * allowed to sleep.
59  */
ntfs_test_inode(struct inode * vi,ntfs_attr * na)60 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
61 {
62 	ntfs_inode *ni;
63 
64 	if (vi->i_ino != na->mft_no)
65 		return 0;
66 	ni = NTFS_I(vi);
67 	/* If !NInoAttr(ni), @vi is a normal file or directory inode. */
68 	if (likely(!NInoAttr(ni))) {
69 		/* If not looking for a normal inode this is a mismatch. */
70 		if (unlikely(na->type != AT_UNUSED))
71 			return 0;
72 	} else {
73 		/* A fake inode describing an attribute. */
74 		if (ni->type != na->type)
75 			return 0;
76 		if (ni->name_len != na->name_len)
77 			return 0;
78 		if (na->name_len && memcmp(ni->name, na->name,
79 				na->name_len * sizeof(ntfschar)))
80 			return 0;
81 	}
82 	/* Match! */
83 	return 1;
84 }
85 
86 /**
87  * ntfs_init_locked_inode - initialize an inode
88  * @vi:		vfs inode to initialize
89  * @na:		ntfs attribute which to initialize @vi to
90  *
91  * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
92  * order to enable ntfs_test_inode() to do its work.
93  *
94  * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
95  * In that case, @na->name and @na->name_len should be set to NULL and 0,
96  * respectively. Although that is not strictly necessary as
97  * ntfs_read_locked_inode() will fill them in later.
98  *
99  * Return 0 on success and -errno on error.
100  *
101  * NOTE: This function runs with the inode->i_lock spin lock held so it is not
102  * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
103  */
ntfs_init_locked_inode(struct inode * vi,ntfs_attr * na)104 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
105 {
106 	ntfs_inode *ni = NTFS_I(vi);
107 
108 	vi->i_ino = na->mft_no;
109 
110 	ni->type = na->type;
111 	if (na->type == AT_INDEX_ALLOCATION)
112 		NInoSetMstProtected(ni);
113 
114 	ni->name = na->name;
115 	ni->name_len = na->name_len;
116 
117 	/* If initializing a normal inode, we are done. */
118 	if (likely(na->type == AT_UNUSED)) {
119 		BUG_ON(na->name);
120 		BUG_ON(na->name_len);
121 		return 0;
122 	}
123 
124 	/* It is a fake inode. */
125 	NInoSetAttr(ni);
126 
127 	/*
128 	 * We have I30 global constant as an optimization as it is the name
129 	 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
130 	 * allocation but that is ok. And most attributes are unnamed anyway,
131 	 * thus the fraction of named attributes with name != I30 is actually
132 	 * absolutely tiny.
133 	 */
134 	if (na->name_len && na->name != I30) {
135 		unsigned int i;
136 
137 		BUG_ON(!na->name);
138 		i = na->name_len * sizeof(ntfschar);
139 		ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
140 		if (!ni->name)
141 			return -ENOMEM;
142 		memcpy(ni->name, na->name, i);
143 		ni->name[na->name_len] = 0;
144 	}
145 	return 0;
146 }
147 
148 typedef int (*set_t)(struct inode *, void *);
149 static int ntfs_read_locked_inode(struct inode *vi);
150 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
151 static int ntfs_read_locked_index_inode(struct inode *base_vi,
152 		struct inode *vi);
153 
154 /**
155  * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
156  * @sb:		super block of mounted volume
157  * @mft_no:	mft record number / inode number to obtain
158  *
159  * Obtain the struct inode corresponding to a specific normal inode (i.e. a
160  * file or directory).
161  *
162  * If the inode is in the cache, it is just returned with an increased
163  * reference count. Otherwise, a new struct inode is allocated and initialized,
164  * and finally ntfs_read_locked_inode() is called to read in the inode and
165  * fill in the remainder of the inode structure.
166  *
167  * Return the struct inode on success. Check the return value with IS_ERR() and
168  * if true, the function failed and the error code is obtained from PTR_ERR().
169  */
ntfs_iget(struct super_block * sb,unsigned long mft_no)170 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
171 {
172 	struct inode *vi;
173 	int err;
174 	ntfs_attr na;
175 
176 	na.mft_no = mft_no;
177 	na.type = AT_UNUSED;
178 	na.name = NULL;
179 	na.name_len = 0;
180 
181 	vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
182 			(set_t)ntfs_init_locked_inode, &na);
183 	if (unlikely(!vi))
184 		return ERR_PTR(-ENOMEM);
185 
186 	err = 0;
187 
188 	/* If this is a freshly allocated inode, need to read it now. */
189 	if (vi->i_state & I_NEW) {
190 		err = ntfs_read_locked_inode(vi);
191 		unlock_new_inode(vi);
192 	}
193 	/*
194 	 * There is no point in keeping bad inodes around if the failure was
195 	 * due to ENOMEM. We want to be able to retry again later.
196 	 */
197 	if (unlikely(err == -ENOMEM)) {
198 		iput(vi);
199 		vi = ERR_PTR(err);
200 	}
201 	return vi;
202 }
203 
204 /**
205  * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
206  * @base_vi:	vfs base inode containing the attribute
207  * @type:	attribute type
208  * @name:	Unicode name of the attribute (NULL if unnamed)
209  * @name_len:	length of @name in Unicode characters (0 if unnamed)
210  *
211  * Obtain the (fake) struct inode corresponding to the attribute specified by
212  * @type, @name, and @name_len, which is present in the base mft record
213  * specified by the vfs inode @base_vi.
214  *
215  * If the attribute inode is in the cache, it is just returned with an
216  * increased reference count. Otherwise, a new struct inode is allocated and
217  * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
218  * attribute and fill in the inode structure.
219  *
220  * Note, for index allocation attributes, you need to use ntfs_index_iget()
221  * instead of ntfs_attr_iget() as working with indices is a lot more complex.
222  *
223  * Return the struct inode of the attribute inode on success. Check the return
224  * value with IS_ERR() and if true, the function failed and the error code is
225  * obtained from PTR_ERR().
226  */
ntfs_attr_iget(struct inode * base_vi,ATTR_TYPE type,ntfschar * name,u32 name_len)227 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
228 		ntfschar *name, u32 name_len)
229 {
230 	struct inode *vi;
231 	int err;
232 	ntfs_attr na;
233 
234 	/* Make sure no one calls ntfs_attr_iget() for indices. */
235 	BUG_ON(type == AT_INDEX_ALLOCATION);
236 
237 	na.mft_no = base_vi->i_ino;
238 	na.type = type;
239 	na.name = name;
240 	na.name_len = name_len;
241 
242 	vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
243 			(set_t)ntfs_init_locked_inode, &na);
244 	if (unlikely(!vi))
245 		return ERR_PTR(-ENOMEM);
246 
247 	err = 0;
248 
249 	/* If this is a freshly allocated inode, need to read it now. */
250 	if (vi->i_state & I_NEW) {
251 		err = ntfs_read_locked_attr_inode(base_vi, vi);
252 		unlock_new_inode(vi);
253 	}
254 	/*
255 	 * There is no point in keeping bad attribute inodes around. This also
256 	 * simplifies things in that we never need to check for bad attribute
257 	 * inodes elsewhere.
258 	 */
259 	if (unlikely(err)) {
260 		iput(vi);
261 		vi = ERR_PTR(err);
262 	}
263 	return vi;
264 }
265 
266 /**
267  * ntfs_index_iget - obtain a struct inode corresponding to an index
268  * @base_vi:	vfs base inode containing the index related attributes
269  * @name:	Unicode name of the index
270  * @name_len:	length of @name in Unicode characters
271  *
272  * Obtain the (fake) struct inode corresponding to the index specified by @name
273  * and @name_len, which is present in the base mft record specified by the vfs
274  * inode @base_vi.
275  *
276  * If the index inode is in the cache, it is just returned with an increased
277  * reference count.  Otherwise, a new struct inode is allocated and
278  * initialized, and finally ntfs_read_locked_index_inode() is called to read
279  * the index related attributes and fill in the inode structure.
280  *
281  * Return the struct inode of the index inode on success. Check the return
282  * value with IS_ERR() and if true, the function failed and the error code is
283  * obtained from PTR_ERR().
284  */
ntfs_index_iget(struct inode * base_vi,ntfschar * name,u32 name_len)285 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
286 		u32 name_len)
287 {
288 	struct inode *vi;
289 	int err;
290 	ntfs_attr na;
291 
292 	na.mft_no = base_vi->i_ino;
293 	na.type = AT_INDEX_ALLOCATION;
294 	na.name = name;
295 	na.name_len = name_len;
296 
297 	vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
298 			(set_t)ntfs_init_locked_inode, &na);
299 	if (unlikely(!vi))
300 		return ERR_PTR(-ENOMEM);
301 
302 	err = 0;
303 
304 	/* If this is a freshly allocated inode, need to read it now. */
305 	if (vi->i_state & I_NEW) {
306 		err = ntfs_read_locked_index_inode(base_vi, vi);
307 		unlock_new_inode(vi);
308 	}
309 	/*
310 	 * There is no point in keeping bad index inodes around.  This also
311 	 * simplifies things in that we never need to check for bad index
312 	 * inodes elsewhere.
313 	 */
314 	if (unlikely(err)) {
315 		iput(vi);
316 		vi = ERR_PTR(err);
317 	}
318 	return vi;
319 }
320 
ntfs_alloc_big_inode(struct super_block * sb)321 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
322 {
323 	ntfs_inode *ni;
324 
325 	ntfs_debug("Entering.");
326 	ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_NOFS);
327 	if (likely(ni != NULL)) {
328 		ni->state = 0;
329 		return VFS_I(ni);
330 	}
331 	ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
332 	return NULL;
333 }
334 
ntfs_i_callback(struct rcu_head * head)335 static void ntfs_i_callback(struct rcu_head *head)
336 {
337 	struct inode *inode = container_of(head, struct inode, i_rcu);
338 	kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
339 }
340 
ntfs_destroy_big_inode(struct inode * inode)341 void ntfs_destroy_big_inode(struct inode *inode)
342 {
343 	ntfs_inode *ni = NTFS_I(inode);
344 
345 	ntfs_debug("Entering.");
346 	BUG_ON(ni->page);
347 	if (!atomic_dec_and_test(&ni->count))
348 		BUG();
349 	call_rcu(&inode->i_rcu, ntfs_i_callback);
350 }
351 
ntfs_alloc_extent_inode(void)352 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
353 {
354 	ntfs_inode *ni;
355 
356 	ntfs_debug("Entering.");
357 	ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS);
358 	if (likely(ni != NULL)) {
359 		ni->state = 0;
360 		return ni;
361 	}
362 	ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
363 	return NULL;
364 }
365 
ntfs_destroy_extent_inode(ntfs_inode * ni)366 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
367 {
368 	ntfs_debug("Entering.");
369 	BUG_ON(ni->page);
370 	if (!atomic_dec_and_test(&ni->count))
371 		BUG();
372 	kmem_cache_free(ntfs_inode_cache, ni);
373 }
374 
375 /*
376  * The attribute runlist lock has separate locking rules from the
377  * normal runlist lock, so split the two lock-classes:
378  */
379 static struct lock_class_key attr_list_rl_lock_class;
380 
381 /**
382  * __ntfs_init_inode - initialize ntfs specific part of an inode
383  * @sb:		super block of mounted volume
384  * @ni:		freshly allocated ntfs inode which to initialize
385  *
386  * Initialize an ntfs inode to defaults.
387  *
388  * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
389  * untouched. Make sure to initialize them elsewhere.
390  *
391  * Return zero on success and -ENOMEM on error.
392  */
__ntfs_init_inode(struct super_block * sb,ntfs_inode * ni)393 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
394 {
395 	ntfs_debug("Entering.");
396 	rwlock_init(&ni->size_lock);
397 	ni->initialized_size = ni->allocated_size = 0;
398 	ni->seq_no = 0;
399 	atomic_set(&ni->count, 1);
400 	ni->vol = NTFS_SB(sb);
401 	ntfs_init_runlist(&ni->runlist);
402 	mutex_init(&ni->mrec_lock);
403 	ni->page = NULL;
404 	ni->page_ofs = 0;
405 	ni->attr_list_size = 0;
406 	ni->attr_list = NULL;
407 	ntfs_init_runlist(&ni->attr_list_rl);
408 	lockdep_set_class(&ni->attr_list_rl.lock,
409 				&attr_list_rl_lock_class);
410 	ni->itype.index.block_size = 0;
411 	ni->itype.index.vcn_size = 0;
412 	ni->itype.index.collation_rule = 0;
413 	ni->itype.index.block_size_bits = 0;
414 	ni->itype.index.vcn_size_bits = 0;
415 	mutex_init(&ni->extent_lock);
416 	ni->nr_extents = 0;
417 	ni->ext.base_ntfs_ino = NULL;
418 }
419 
420 /*
421  * Extent inodes get MFT-mapped in a nested way, while the base inode
422  * is still mapped. Teach this nesting to the lock validator by creating
423  * a separate class for nested inode's mrec_lock's:
424  */
425 static struct lock_class_key extent_inode_mrec_lock_key;
426 
ntfs_new_extent_inode(struct super_block * sb,unsigned long mft_no)427 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
428 		unsigned long mft_no)
429 {
430 	ntfs_inode *ni = ntfs_alloc_extent_inode();
431 
432 	ntfs_debug("Entering.");
433 	if (likely(ni != NULL)) {
434 		__ntfs_init_inode(sb, ni);
435 		lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
436 		ni->mft_no = mft_no;
437 		ni->type = AT_UNUSED;
438 		ni->name = NULL;
439 		ni->name_len = 0;
440 	}
441 	return ni;
442 }
443 
444 /**
445  * ntfs_is_extended_system_file - check if a file is in the $Extend directory
446  * @ctx:	initialized attribute search context
447  *
448  * Search all file name attributes in the inode described by the attribute
449  * search context @ctx and check if any of the names are in the $Extend system
450  * directory.
451  *
452  * Return values:
453  *	   1: file is in $Extend directory
454  *	   0: file is not in $Extend directory
455  *    -errno: failed to determine if the file is in the $Extend directory
456  */
ntfs_is_extended_system_file(ntfs_attr_search_ctx * ctx)457 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
458 {
459 	int nr_links, err;
460 
461 	/* Restart search. */
462 	ntfs_attr_reinit_search_ctx(ctx);
463 
464 	/* Get number of hard links. */
465 	nr_links = le16_to_cpu(ctx->mrec->link_count);
466 
467 	/* Loop through all hard links. */
468 	while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
469 			ctx))) {
470 		FILE_NAME_ATTR *file_name_attr;
471 		ATTR_RECORD *attr = ctx->attr;
472 		u8 *p, *p2;
473 
474 		nr_links--;
475 		/*
476 		 * Maximum sanity checking as we are called on an inode that
477 		 * we suspect might be corrupt.
478 		 */
479 		p = (u8*)attr + le32_to_cpu(attr->length);
480 		if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
481 				le32_to_cpu(ctx->mrec->bytes_in_use)) {
482 err_corrupt_attr:
483 			ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
484 					"attribute. You should run chkdsk.");
485 			return -EIO;
486 		}
487 		if (attr->non_resident) {
488 			ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
489 					"name. You should run chkdsk.");
490 			return -EIO;
491 		}
492 		if (attr->flags) {
493 			ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
494 					"invalid flags. You should run "
495 					"chkdsk.");
496 			return -EIO;
497 		}
498 		if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
499 			ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
500 					"name. You should run chkdsk.");
501 			return -EIO;
502 		}
503 		file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
504 				le16_to_cpu(attr->data.resident.value_offset));
505 		p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
506 		if (p2 < (u8*)attr || p2 > p)
507 			goto err_corrupt_attr;
508 		/* This attribute is ok, but is it in the $Extend directory? */
509 		if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
510 			return 1;	/* YES, it's an extended system file. */
511 	}
512 	if (unlikely(err != -ENOENT))
513 		return err;
514 	if (unlikely(nr_links)) {
515 		ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
516 				"doesn't match number of name attributes. You "
517 				"should run chkdsk.");
518 		return -EIO;
519 	}
520 	return 0;	/* NO, it is not an extended system file. */
521 }
522 
523 /**
524  * ntfs_read_locked_inode - read an inode from its device
525  * @vi:		inode to read
526  *
527  * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
528  * described by @vi into memory from the device.
529  *
530  * The only fields in @vi that we need to/can look at when the function is
531  * called are i_sb, pointing to the mounted device's super block, and i_ino,
532  * the number of the inode to load.
533  *
534  * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
535  * for reading and sets up the necessary @vi fields as well as initializing
536  * the ntfs inode.
537  *
538  * Q: What locks are held when the function is called?
539  * A: i_state has I_NEW set, hence the inode is locked, also
540  *    i_count is set to 1, so it is not going to go away
541  *    i_flags is set to 0 and we have no business touching it.  Only an ioctl()
542  *    is allowed to write to them. We should of course be honouring them but
543  *    we need to do that using the IS_* macros defined in include/linux/fs.h.
544  *    In any case ntfs_read_locked_inode() has nothing to do with i_flags.
545  *
546  * Return 0 on success and -errno on error.  In the error case, the inode will
547  * have had make_bad_inode() executed on it.
548  */
ntfs_read_locked_inode(struct inode * vi)549 static int ntfs_read_locked_inode(struct inode *vi)
550 {
551 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
552 	ntfs_inode *ni;
553 	struct inode *bvi;
554 	MFT_RECORD *m;
555 	ATTR_RECORD *a;
556 	STANDARD_INFORMATION *si;
557 	ntfs_attr_search_ctx *ctx;
558 	int err = 0;
559 
560 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
561 
562 	/* Setup the generic vfs inode parts now. */
563 
564 	/*
565 	 * This is for checking whether an inode has changed w.r.t. a file so
566 	 * that the file can be updated if necessary (compare with f_version).
567 	 */
568 	vi->i_version = 1;
569 
570 	vi->i_uid = vol->uid;
571 	vi->i_gid = vol->gid;
572 	vi->i_mode = 0;
573 
574 	/*
575 	 * Initialize the ntfs specific part of @vi special casing
576 	 * FILE_MFT which we need to do at mount time.
577 	 */
578 	if (vi->i_ino != FILE_MFT)
579 		ntfs_init_big_inode(vi);
580 	ni = NTFS_I(vi);
581 
582 	m = map_mft_record(ni);
583 	if (IS_ERR(m)) {
584 		err = PTR_ERR(m);
585 		goto err_out;
586 	}
587 	ctx = ntfs_attr_get_search_ctx(ni, m);
588 	if (!ctx) {
589 		err = -ENOMEM;
590 		goto unm_err_out;
591 	}
592 
593 	if (!(m->flags & MFT_RECORD_IN_USE)) {
594 		ntfs_error(vi->i_sb, "Inode is not in use!");
595 		goto unm_err_out;
596 	}
597 	if (m->base_mft_record) {
598 		ntfs_error(vi->i_sb, "Inode is an extent inode!");
599 		goto unm_err_out;
600 	}
601 
602 	/* Transfer information from mft record into vfs and ntfs inodes. */
603 	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
604 
605 	/*
606 	 * FIXME: Keep in mind that link_count is two for files which have both
607 	 * a long file name and a short file name as separate entries, so if
608 	 * we are hiding short file names this will be too high. Either we need
609 	 * to account for the short file names by subtracting them or we need
610 	 * to make sure we delete files even though i_nlink is not zero which
611 	 * might be tricky due to vfs interactions. Need to think about this
612 	 * some more when implementing the unlink command.
613 	 */
614 	set_nlink(vi, le16_to_cpu(m->link_count));
615 	/*
616 	 * FIXME: Reparse points can have the directory bit set even though
617 	 * they would be S_IFLNK. Need to deal with this further below when we
618 	 * implement reparse points / symbolic links but it will do for now.
619 	 * Also if not a directory, it could be something else, rather than
620 	 * a regular file. But again, will do for now.
621 	 */
622 	/* Everyone gets all permissions. */
623 	vi->i_mode |= S_IRWXUGO;
624 	/* If read-only, no one gets write permissions. */
625 	if (IS_RDONLY(vi))
626 		vi->i_mode &= ~S_IWUGO;
627 	if (m->flags & MFT_RECORD_IS_DIRECTORY) {
628 		vi->i_mode |= S_IFDIR;
629 		/*
630 		 * Apply the directory permissions mask set in the mount
631 		 * options.
632 		 */
633 		vi->i_mode &= ~vol->dmask;
634 		/* Things break without this kludge! */
635 		if (vi->i_nlink > 1)
636 			set_nlink(vi, 1);
637 	} else {
638 		vi->i_mode |= S_IFREG;
639 		/* Apply the file permissions mask set in the mount options. */
640 		vi->i_mode &= ~vol->fmask;
641 	}
642 	/*
643 	 * Find the standard information attribute in the mft record. At this
644 	 * stage we haven't setup the attribute list stuff yet, so this could
645 	 * in fact fail if the standard information is in an extent record, but
646 	 * I don't think this actually ever happens.
647 	 */
648 	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
649 			ctx);
650 	if (unlikely(err)) {
651 		if (err == -ENOENT) {
652 			/*
653 			 * TODO: We should be performing a hot fix here (if the
654 			 * recover mount option is set) by creating a new
655 			 * attribute.
656 			 */
657 			ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
658 					"is missing.");
659 		}
660 		goto unm_err_out;
661 	}
662 	a = ctx->attr;
663 	/* Get the standard information attribute value. */
664 	si = (STANDARD_INFORMATION*)((u8*)a +
665 			le16_to_cpu(a->data.resident.value_offset));
666 
667 	/* Transfer information from the standard information into vi. */
668 	/*
669 	 * Note: The i_?times do not quite map perfectly onto the NTFS times,
670 	 * but they are close enough, and in the end it doesn't really matter
671 	 * that much...
672 	 */
673 	/*
674 	 * mtime is the last change of the data within the file. Not changed
675 	 * when only metadata is changed, e.g. a rename doesn't affect mtime.
676 	 */
677 	vi->i_mtime = ntfs2utc(si->last_data_change_time);
678 	/*
679 	 * ctime is the last change of the metadata of the file. This obviously
680 	 * always changes, when mtime is changed. ctime can be changed on its
681 	 * own, mtime is then not changed, e.g. when a file is renamed.
682 	 */
683 	vi->i_ctime = ntfs2utc(si->last_mft_change_time);
684 	/*
685 	 * Last access to the data within the file. Not changed during a rename
686 	 * for example but changed whenever the file is written to.
687 	 */
688 	vi->i_atime = ntfs2utc(si->last_access_time);
689 
690 	/* Find the attribute list attribute if present. */
691 	ntfs_attr_reinit_search_ctx(ctx);
692 	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
693 	if (err) {
694 		if (unlikely(err != -ENOENT)) {
695 			ntfs_error(vi->i_sb, "Failed to lookup attribute list "
696 					"attribute.");
697 			goto unm_err_out;
698 		}
699 	} else /* if (!err) */ {
700 		if (vi->i_ino == FILE_MFT)
701 			goto skip_attr_list_load;
702 		ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
703 		NInoSetAttrList(ni);
704 		a = ctx->attr;
705 		if (a->flags & ATTR_COMPRESSION_MASK) {
706 			ntfs_error(vi->i_sb, "Attribute list attribute is "
707 					"compressed.");
708 			goto unm_err_out;
709 		}
710 		if (a->flags & ATTR_IS_ENCRYPTED ||
711 				a->flags & ATTR_IS_SPARSE) {
712 			if (a->non_resident) {
713 				ntfs_error(vi->i_sb, "Non-resident attribute "
714 						"list attribute is encrypted/"
715 						"sparse.");
716 				goto unm_err_out;
717 			}
718 			ntfs_warning(vi->i_sb, "Resident attribute list "
719 					"attribute in inode 0x%lx is marked "
720 					"encrypted/sparse which is not true.  "
721 					"However, Windows allows this and "
722 					"chkdsk does not detect or correct it "
723 					"so we will just ignore the invalid "
724 					"flags and pretend they are not set.",
725 					vi->i_ino);
726 		}
727 		/* Now allocate memory for the attribute list. */
728 		ni->attr_list_size = (u32)ntfs_attr_size(a);
729 		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
730 		if (!ni->attr_list) {
731 			ntfs_error(vi->i_sb, "Not enough memory to allocate "
732 					"buffer for attribute list.");
733 			err = -ENOMEM;
734 			goto unm_err_out;
735 		}
736 		if (a->non_resident) {
737 			NInoSetAttrListNonResident(ni);
738 			if (a->data.non_resident.lowest_vcn) {
739 				ntfs_error(vi->i_sb, "Attribute list has non "
740 						"zero lowest_vcn.");
741 				goto unm_err_out;
742 			}
743 			/*
744 			 * Setup the runlist. No need for locking as we have
745 			 * exclusive access to the inode at this time.
746 			 */
747 			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
748 					a, NULL);
749 			if (IS_ERR(ni->attr_list_rl.rl)) {
750 				err = PTR_ERR(ni->attr_list_rl.rl);
751 				ni->attr_list_rl.rl = NULL;
752 				ntfs_error(vi->i_sb, "Mapping pairs "
753 						"decompression failed.");
754 				goto unm_err_out;
755 			}
756 			/* Now load the attribute list. */
757 			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
758 					ni->attr_list, ni->attr_list_size,
759 					sle64_to_cpu(a->data.non_resident.
760 					initialized_size)))) {
761 				ntfs_error(vi->i_sb, "Failed to load "
762 						"attribute list attribute.");
763 				goto unm_err_out;
764 			}
765 		} else /* if (!a->non_resident) */ {
766 			if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
767 					+ le32_to_cpu(
768 					a->data.resident.value_length) >
769 					(u8*)ctx->mrec + vol->mft_record_size) {
770 				ntfs_error(vi->i_sb, "Corrupt attribute list "
771 						"in inode.");
772 				goto unm_err_out;
773 			}
774 			/* Now copy the attribute list. */
775 			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
776 					a->data.resident.value_offset),
777 					le32_to_cpu(
778 					a->data.resident.value_length));
779 		}
780 	}
781 skip_attr_list_load:
782 	/*
783 	 * If an attribute list is present we now have the attribute list value
784 	 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
785 	 */
786 	if (S_ISDIR(vi->i_mode)) {
787 		loff_t bvi_size;
788 		ntfs_inode *bni;
789 		INDEX_ROOT *ir;
790 		u8 *ir_end, *index_end;
791 
792 		/* It is a directory, find index root attribute. */
793 		ntfs_attr_reinit_search_ctx(ctx);
794 		err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
795 				0, NULL, 0, ctx);
796 		if (unlikely(err)) {
797 			if (err == -ENOENT) {
798 				// FIXME: File is corrupt! Hot-fix with empty
799 				// index root attribute if recovery option is
800 				// set.
801 				ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
802 						"is missing.");
803 			}
804 			goto unm_err_out;
805 		}
806 		a = ctx->attr;
807 		/* Set up the state. */
808 		if (unlikely(a->non_resident)) {
809 			ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
810 					"resident.");
811 			goto unm_err_out;
812 		}
813 		/* Ensure the attribute name is placed before the value. */
814 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
815 				le16_to_cpu(a->data.resident.value_offset)))) {
816 			ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
817 					"placed after the attribute value.");
818 			goto unm_err_out;
819 		}
820 		/*
821 		 * Compressed/encrypted index root just means that the newly
822 		 * created files in that directory should be created compressed/
823 		 * encrypted. However index root cannot be both compressed and
824 		 * encrypted.
825 		 */
826 		if (a->flags & ATTR_COMPRESSION_MASK)
827 			NInoSetCompressed(ni);
828 		if (a->flags & ATTR_IS_ENCRYPTED) {
829 			if (a->flags & ATTR_COMPRESSION_MASK) {
830 				ntfs_error(vi->i_sb, "Found encrypted and "
831 						"compressed attribute.");
832 				goto unm_err_out;
833 			}
834 			NInoSetEncrypted(ni);
835 		}
836 		if (a->flags & ATTR_IS_SPARSE)
837 			NInoSetSparse(ni);
838 		ir = (INDEX_ROOT*)((u8*)a +
839 				le16_to_cpu(a->data.resident.value_offset));
840 		ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
841 		if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
842 			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
843 					"corrupt.");
844 			goto unm_err_out;
845 		}
846 		index_end = (u8*)&ir->index +
847 				le32_to_cpu(ir->index.index_length);
848 		if (index_end > ir_end) {
849 			ntfs_error(vi->i_sb, "Directory index is corrupt.");
850 			goto unm_err_out;
851 		}
852 		if (ir->type != AT_FILE_NAME) {
853 			ntfs_error(vi->i_sb, "Indexed attribute is not "
854 					"$FILE_NAME.");
855 			goto unm_err_out;
856 		}
857 		if (ir->collation_rule != COLLATION_FILE_NAME) {
858 			ntfs_error(vi->i_sb, "Index collation rule is not "
859 					"COLLATION_FILE_NAME.");
860 			goto unm_err_out;
861 		}
862 		ni->itype.index.collation_rule = ir->collation_rule;
863 		ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
864 		if (ni->itype.index.block_size &
865 				(ni->itype.index.block_size - 1)) {
866 			ntfs_error(vi->i_sb, "Index block size (%u) is not a "
867 					"power of two.",
868 					ni->itype.index.block_size);
869 			goto unm_err_out;
870 		}
871 		if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
872 			ntfs_error(vi->i_sb, "Index block size (%u) > "
873 					"PAGE_CACHE_SIZE (%ld) is not "
874 					"supported.  Sorry.",
875 					ni->itype.index.block_size,
876 					PAGE_CACHE_SIZE);
877 			err = -EOPNOTSUPP;
878 			goto unm_err_out;
879 		}
880 		if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
881 			ntfs_error(vi->i_sb, "Index block size (%u) < "
882 					"NTFS_BLOCK_SIZE (%i) is not "
883 					"supported.  Sorry.",
884 					ni->itype.index.block_size,
885 					NTFS_BLOCK_SIZE);
886 			err = -EOPNOTSUPP;
887 			goto unm_err_out;
888 		}
889 		ni->itype.index.block_size_bits =
890 				ffs(ni->itype.index.block_size) - 1;
891 		/* Determine the size of a vcn in the directory index. */
892 		if (vol->cluster_size <= ni->itype.index.block_size) {
893 			ni->itype.index.vcn_size = vol->cluster_size;
894 			ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
895 		} else {
896 			ni->itype.index.vcn_size = vol->sector_size;
897 			ni->itype.index.vcn_size_bits = vol->sector_size_bits;
898 		}
899 
900 		/* Setup the index allocation attribute, even if not present. */
901 		NInoSetMstProtected(ni);
902 		ni->type = AT_INDEX_ALLOCATION;
903 		ni->name = I30;
904 		ni->name_len = 4;
905 
906 		if (!(ir->index.flags & LARGE_INDEX)) {
907 			/* No index allocation. */
908 			vi->i_size = ni->initialized_size =
909 					ni->allocated_size = 0;
910 			/* We are done with the mft record, so we release it. */
911 			ntfs_attr_put_search_ctx(ctx);
912 			unmap_mft_record(ni);
913 			m = NULL;
914 			ctx = NULL;
915 			goto skip_large_dir_stuff;
916 		} /* LARGE_INDEX: Index allocation present. Setup state. */
917 		NInoSetIndexAllocPresent(ni);
918 		/* Find index allocation attribute. */
919 		ntfs_attr_reinit_search_ctx(ctx);
920 		err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
921 				CASE_SENSITIVE, 0, NULL, 0, ctx);
922 		if (unlikely(err)) {
923 			if (err == -ENOENT)
924 				ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
925 						"attribute is not present but "
926 						"$INDEX_ROOT indicated it is.");
927 			else
928 				ntfs_error(vi->i_sb, "Failed to lookup "
929 						"$INDEX_ALLOCATION "
930 						"attribute.");
931 			goto unm_err_out;
932 		}
933 		a = ctx->attr;
934 		if (!a->non_resident) {
935 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
936 					"is resident.");
937 			goto unm_err_out;
938 		}
939 		/*
940 		 * Ensure the attribute name is placed before the mapping pairs
941 		 * array.
942 		 */
943 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
944 				le16_to_cpu(
945 				a->data.non_resident.mapping_pairs_offset)))) {
946 			ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
947 					"is placed after the mapping pairs "
948 					"array.");
949 			goto unm_err_out;
950 		}
951 		if (a->flags & ATTR_IS_ENCRYPTED) {
952 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
953 					"is encrypted.");
954 			goto unm_err_out;
955 		}
956 		if (a->flags & ATTR_IS_SPARSE) {
957 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
958 					"is sparse.");
959 			goto unm_err_out;
960 		}
961 		if (a->flags & ATTR_COMPRESSION_MASK) {
962 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
963 					"is compressed.");
964 			goto unm_err_out;
965 		}
966 		if (a->data.non_resident.lowest_vcn) {
967 			ntfs_error(vi->i_sb, "First extent of "
968 					"$INDEX_ALLOCATION attribute has non "
969 					"zero lowest_vcn.");
970 			goto unm_err_out;
971 		}
972 		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
973 		ni->initialized_size = sle64_to_cpu(
974 				a->data.non_resident.initialized_size);
975 		ni->allocated_size = sle64_to_cpu(
976 				a->data.non_resident.allocated_size);
977 		/*
978 		 * We are done with the mft record, so we release it. Otherwise
979 		 * we would deadlock in ntfs_attr_iget().
980 		 */
981 		ntfs_attr_put_search_ctx(ctx);
982 		unmap_mft_record(ni);
983 		m = NULL;
984 		ctx = NULL;
985 		/* Get the index bitmap attribute inode. */
986 		bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
987 		if (IS_ERR(bvi)) {
988 			ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
989 			err = PTR_ERR(bvi);
990 			goto unm_err_out;
991 		}
992 		bni = NTFS_I(bvi);
993 		if (NInoCompressed(bni) || NInoEncrypted(bni) ||
994 				NInoSparse(bni)) {
995 			ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
996 					"and/or encrypted and/or sparse.");
997 			goto iput_unm_err_out;
998 		}
999 		/* Consistency check bitmap size vs. index allocation size. */
1000 		bvi_size = i_size_read(bvi);
1001 		if ((bvi_size << 3) < (vi->i_size >>
1002 				ni->itype.index.block_size_bits)) {
1003 			ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
1004 					"for index allocation (0x%llx).",
1005 					bvi_size << 3, vi->i_size);
1006 			goto iput_unm_err_out;
1007 		}
1008 		/* No longer need the bitmap attribute inode. */
1009 		iput(bvi);
1010 skip_large_dir_stuff:
1011 		/* Setup the operations for this inode. */
1012 		vi->i_op = &ntfs_dir_inode_ops;
1013 		vi->i_fop = &ntfs_dir_ops;
1014 	} else {
1015 		/* It is a file. */
1016 		ntfs_attr_reinit_search_ctx(ctx);
1017 
1018 		/* Setup the data attribute, even if not present. */
1019 		ni->type = AT_DATA;
1020 		ni->name = NULL;
1021 		ni->name_len = 0;
1022 
1023 		/* Find first extent of the unnamed data attribute. */
1024 		err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
1025 		if (unlikely(err)) {
1026 			vi->i_size = ni->initialized_size =
1027 					ni->allocated_size = 0;
1028 			if (err != -ENOENT) {
1029 				ntfs_error(vi->i_sb, "Failed to lookup $DATA "
1030 						"attribute.");
1031 				goto unm_err_out;
1032 			}
1033 			/*
1034 			 * FILE_Secure does not have an unnamed $DATA
1035 			 * attribute, so we special case it here.
1036 			 */
1037 			if (vi->i_ino == FILE_Secure)
1038 				goto no_data_attr_special_case;
1039 			/*
1040 			 * Most if not all the system files in the $Extend
1041 			 * system directory do not have unnamed data
1042 			 * attributes so we need to check if the parent
1043 			 * directory of the file is FILE_Extend and if it is
1044 			 * ignore this error. To do this we need to get the
1045 			 * name of this inode from the mft record as the name
1046 			 * contains the back reference to the parent directory.
1047 			 */
1048 			if (ntfs_is_extended_system_file(ctx) > 0)
1049 				goto no_data_attr_special_case;
1050 			// FIXME: File is corrupt! Hot-fix with empty data
1051 			// attribute if recovery option is set.
1052 			ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1053 			goto unm_err_out;
1054 		}
1055 		a = ctx->attr;
1056 		/* Setup the state. */
1057 		if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1058 			if (a->flags & ATTR_COMPRESSION_MASK) {
1059 				NInoSetCompressed(ni);
1060 				if (vol->cluster_size > 4096) {
1061 					ntfs_error(vi->i_sb, "Found "
1062 							"compressed data but "
1063 							"compression is "
1064 							"disabled due to "
1065 							"cluster size (%i) > "
1066 							"4kiB.",
1067 							vol->cluster_size);
1068 					goto unm_err_out;
1069 				}
1070 				if ((a->flags & ATTR_COMPRESSION_MASK)
1071 						!= ATTR_IS_COMPRESSED) {
1072 					ntfs_error(vi->i_sb, "Found unknown "
1073 							"compression method "
1074 							"or corrupt file.");
1075 					goto unm_err_out;
1076 				}
1077 			}
1078 			if (a->flags & ATTR_IS_SPARSE)
1079 				NInoSetSparse(ni);
1080 		}
1081 		if (a->flags & ATTR_IS_ENCRYPTED) {
1082 			if (NInoCompressed(ni)) {
1083 				ntfs_error(vi->i_sb, "Found encrypted and "
1084 						"compressed data.");
1085 				goto unm_err_out;
1086 			}
1087 			NInoSetEncrypted(ni);
1088 		}
1089 		if (a->non_resident) {
1090 			NInoSetNonResident(ni);
1091 			if (NInoCompressed(ni) || NInoSparse(ni)) {
1092 				if (NInoCompressed(ni) && a->data.non_resident.
1093 						compression_unit != 4) {
1094 					ntfs_error(vi->i_sb, "Found "
1095 							"non-standard "
1096 							"compression unit (%u "
1097 							"instead of 4).  "
1098 							"Cannot handle this.",
1099 							a->data.non_resident.
1100 							compression_unit);
1101 					err = -EOPNOTSUPP;
1102 					goto unm_err_out;
1103 				}
1104 				if (a->data.non_resident.compression_unit) {
1105 					ni->itype.compressed.block_size = 1U <<
1106 							(a->data.non_resident.
1107 							compression_unit +
1108 							vol->cluster_size_bits);
1109 					ni->itype.compressed.block_size_bits =
1110 							ffs(ni->itype.
1111 							compressed.
1112 							block_size) - 1;
1113 					ni->itype.compressed.block_clusters =
1114 							1U << a->data.
1115 							non_resident.
1116 							compression_unit;
1117 				} else {
1118 					ni->itype.compressed.block_size = 0;
1119 					ni->itype.compressed.block_size_bits =
1120 							0;
1121 					ni->itype.compressed.block_clusters =
1122 							0;
1123 				}
1124 				ni->itype.compressed.size = sle64_to_cpu(
1125 						a->data.non_resident.
1126 						compressed_size);
1127 			}
1128 			if (a->data.non_resident.lowest_vcn) {
1129 				ntfs_error(vi->i_sb, "First extent of $DATA "
1130 						"attribute has non zero "
1131 						"lowest_vcn.");
1132 				goto unm_err_out;
1133 			}
1134 			vi->i_size = sle64_to_cpu(
1135 					a->data.non_resident.data_size);
1136 			ni->initialized_size = sle64_to_cpu(
1137 					a->data.non_resident.initialized_size);
1138 			ni->allocated_size = sle64_to_cpu(
1139 					a->data.non_resident.allocated_size);
1140 		} else { /* Resident attribute. */
1141 			vi->i_size = ni->initialized_size = le32_to_cpu(
1142 					a->data.resident.value_length);
1143 			ni->allocated_size = le32_to_cpu(a->length) -
1144 					le16_to_cpu(
1145 					a->data.resident.value_offset);
1146 			if (vi->i_size > ni->allocated_size) {
1147 				ntfs_error(vi->i_sb, "Resident data attribute "
1148 						"is corrupt (size exceeds "
1149 						"allocation).");
1150 				goto unm_err_out;
1151 			}
1152 		}
1153 no_data_attr_special_case:
1154 		/* We are done with the mft record, so we release it. */
1155 		ntfs_attr_put_search_ctx(ctx);
1156 		unmap_mft_record(ni);
1157 		m = NULL;
1158 		ctx = NULL;
1159 		/* Setup the operations for this inode. */
1160 		vi->i_op = &ntfs_file_inode_ops;
1161 		vi->i_fop = &ntfs_file_ops;
1162 	}
1163 	if (NInoMstProtected(ni))
1164 		vi->i_mapping->a_ops = &ntfs_mst_aops;
1165 	else
1166 		vi->i_mapping->a_ops = &ntfs_aops;
1167 	/*
1168 	 * The number of 512-byte blocks used on disk (for stat). This is in so
1169 	 * far inaccurate as it doesn't account for any named streams or other
1170 	 * special non-resident attributes, but that is how Windows works, too,
1171 	 * so we are at least consistent with Windows, if not entirely
1172 	 * consistent with the Linux Way. Doing it the Linux Way would cause a
1173 	 * significant slowdown as it would involve iterating over all
1174 	 * attributes in the mft record and adding the allocated/compressed
1175 	 * sizes of all non-resident attributes present to give us the Linux
1176 	 * correct size that should go into i_blocks (after division by 512).
1177 	 */
1178 	if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1179 		vi->i_blocks = ni->itype.compressed.size >> 9;
1180 	else
1181 		vi->i_blocks = ni->allocated_size >> 9;
1182 	ntfs_debug("Done.");
1183 	return 0;
1184 iput_unm_err_out:
1185 	iput(bvi);
1186 unm_err_out:
1187 	if (!err)
1188 		err = -EIO;
1189 	if (ctx)
1190 		ntfs_attr_put_search_ctx(ctx);
1191 	if (m)
1192 		unmap_mft_record(ni);
1193 err_out:
1194 	ntfs_error(vol->sb, "Failed with error code %i.  Marking corrupt "
1195 			"inode 0x%lx as bad.  Run chkdsk.", err, vi->i_ino);
1196 	make_bad_inode(vi);
1197 	if (err != -EOPNOTSUPP && err != -ENOMEM)
1198 		NVolSetErrors(vol);
1199 	return err;
1200 }
1201 
1202 /**
1203  * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1204  * @base_vi:	base inode
1205  * @vi:		attribute inode to read
1206  *
1207  * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1208  * attribute inode described by @vi into memory from the base mft record
1209  * described by @base_ni.
1210  *
1211  * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1212  * reading and looks up the attribute described by @vi before setting up the
1213  * necessary fields in @vi as well as initializing the ntfs inode.
1214  *
1215  * Q: What locks are held when the function is called?
1216  * A: i_state has I_NEW set, hence the inode is locked, also
1217  *    i_count is set to 1, so it is not going to go away
1218  *
1219  * Return 0 on success and -errno on error.  In the error case, the inode will
1220  * have had make_bad_inode() executed on it.
1221  *
1222  * Note this cannot be called for AT_INDEX_ALLOCATION.
1223  */
ntfs_read_locked_attr_inode(struct inode * base_vi,struct inode * vi)1224 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1225 {
1226 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
1227 	ntfs_inode *ni, *base_ni;
1228 	MFT_RECORD *m;
1229 	ATTR_RECORD *a;
1230 	ntfs_attr_search_ctx *ctx;
1231 	int err = 0;
1232 
1233 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1234 
1235 	ntfs_init_big_inode(vi);
1236 
1237 	ni	= NTFS_I(vi);
1238 	base_ni = NTFS_I(base_vi);
1239 
1240 	/* Just mirror the values from the base inode. */
1241 	vi->i_version	= base_vi->i_version;
1242 	vi->i_uid	= base_vi->i_uid;
1243 	vi->i_gid	= base_vi->i_gid;
1244 	set_nlink(vi, base_vi->i_nlink);
1245 	vi->i_mtime	= base_vi->i_mtime;
1246 	vi->i_ctime	= base_vi->i_ctime;
1247 	vi->i_atime	= base_vi->i_atime;
1248 	vi->i_generation = ni->seq_no = base_ni->seq_no;
1249 
1250 	/* Set inode type to zero but preserve permissions. */
1251 	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
1252 
1253 	m = map_mft_record(base_ni);
1254 	if (IS_ERR(m)) {
1255 		err = PTR_ERR(m);
1256 		goto err_out;
1257 	}
1258 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
1259 	if (!ctx) {
1260 		err = -ENOMEM;
1261 		goto unm_err_out;
1262 	}
1263 	/* Find the attribute. */
1264 	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1265 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1266 	if (unlikely(err))
1267 		goto unm_err_out;
1268 	a = ctx->attr;
1269 	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1270 		if (a->flags & ATTR_COMPRESSION_MASK) {
1271 			NInoSetCompressed(ni);
1272 			if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1273 					ni->name_len)) {
1274 				ntfs_error(vi->i_sb, "Found compressed "
1275 						"non-data or named data "
1276 						"attribute.  Please report "
1277 						"you saw this message to "
1278 						"linux-ntfs-dev@lists."
1279 						"sourceforge.net");
1280 				goto unm_err_out;
1281 			}
1282 			if (vol->cluster_size > 4096) {
1283 				ntfs_error(vi->i_sb, "Found compressed "
1284 						"attribute but compression is "
1285 						"disabled due to cluster size "
1286 						"(%i) > 4kiB.",
1287 						vol->cluster_size);
1288 				goto unm_err_out;
1289 			}
1290 			if ((a->flags & ATTR_COMPRESSION_MASK) !=
1291 					ATTR_IS_COMPRESSED) {
1292 				ntfs_error(vi->i_sb, "Found unknown "
1293 						"compression method.");
1294 				goto unm_err_out;
1295 			}
1296 		}
1297 		/*
1298 		 * The compressed/sparse flag set in an index root just means
1299 		 * to compress all files.
1300 		 */
1301 		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1302 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1303 					"but the attribute is %s.  Please "
1304 					"report you saw this message to "
1305 					"linux-ntfs-dev@lists.sourceforge.net",
1306 					NInoCompressed(ni) ? "compressed" :
1307 					"sparse");
1308 			goto unm_err_out;
1309 		}
1310 		if (a->flags & ATTR_IS_SPARSE)
1311 			NInoSetSparse(ni);
1312 	}
1313 	if (a->flags & ATTR_IS_ENCRYPTED) {
1314 		if (NInoCompressed(ni)) {
1315 			ntfs_error(vi->i_sb, "Found encrypted and compressed "
1316 					"data.");
1317 			goto unm_err_out;
1318 		}
1319 		/*
1320 		 * The encryption flag set in an index root just means to
1321 		 * encrypt all files.
1322 		 */
1323 		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1324 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1325 					"but the attribute is encrypted.  "
1326 					"Please report you saw this message "
1327 					"to linux-ntfs-dev@lists.sourceforge."
1328 					"net");
1329 			goto unm_err_out;
1330 		}
1331 		if (ni->type != AT_DATA) {
1332 			ntfs_error(vi->i_sb, "Found encrypted non-data "
1333 					"attribute.");
1334 			goto unm_err_out;
1335 		}
1336 		NInoSetEncrypted(ni);
1337 	}
1338 	if (!a->non_resident) {
1339 		/* Ensure the attribute name is placed before the value. */
1340 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1341 				le16_to_cpu(a->data.resident.value_offset)))) {
1342 			ntfs_error(vol->sb, "Attribute name is placed after "
1343 					"the attribute value.");
1344 			goto unm_err_out;
1345 		}
1346 		if (NInoMstProtected(ni)) {
1347 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1348 					"but the attribute is resident.  "
1349 					"Please report you saw this message to "
1350 					"linux-ntfs-dev@lists.sourceforge.net");
1351 			goto unm_err_out;
1352 		}
1353 		vi->i_size = ni->initialized_size = le32_to_cpu(
1354 				a->data.resident.value_length);
1355 		ni->allocated_size = le32_to_cpu(a->length) -
1356 				le16_to_cpu(a->data.resident.value_offset);
1357 		if (vi->i_size > ni->allocated_size) {
1358 			ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1359 					"(size exceeds allocation).");
1360 			goto unm_err_out;
1361 		}
1362 	} else {
1363 		NInoSetNonResident(ni);
1364 		/*
1365 		 * Ensure the attribute name is placed before the mapping pairs
1366 		 * array.
1367 		 */
1368 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1369 				le16_to_cpu(
1370 				a->data.non_resident.mapping_pairs_offset)))) {
1371 			ntfs_error(vol->sb, "Attribute name is placed after "
1372 					"the mapping pairs array.");
1373 			goto unm_err_out;
1374 		}
1375 		if (NInoCompressed(ni) || NInoSparse(ni)) {
1376 			if (NInoCompressed(ni) && a->data.non_resident.
1377 					compression_unit != 4) {
1378 				ntfs_error(vi->i_sb, "Found non-standard "
1379 						"compression unit (%u instead "
1380 						"of 4).  Cannot handle this.",
1381 						a->data.non_resident.
1382 						compression_unit);
1383 				err = -EOPNOTSUPP;
1384 				goto unm_err_out;
1385 			}
1386 			if (a->data.non_resident.compression_unit) {
1387 				ni->itype.compressed.block_size = 1U <<
1388 						(a->data.non_resident.
1389 						compression_unit +
1390 						vol->cluster_size_bits);
1391 				ni->itype.compressed.block_size_bits =
1392 						ffs(ni->itype.compressed.
1393 						block_size) - 1;
1394 				ni->itype.compressed.block_clusters = 1U <<
1395 						a->data.non_resident.
1396 						compression_unit;
1397 			} else {
1398 				ni->itype.compressed.block_size = 0;
1399 				ni->itype.compressed.block_size_bits = 0;
1400 				ni->itype.compressed.block_clusters = 0;
1401 			}
1402 			ni->itype.compressed.size = sle64_to_cpu(
1403 					a->data.non_resident.compressed_size);
1404 		}
1405 		if (a->data.non_resident.lowest_vcn) {
1406 			ntfs_error(vi->i_sb, "First extent of attribute has "
1407 					"non-zero lowest_vcn.");
1408 			goto unm_err_out;
1409 		}
1410 		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1411 		ni->initialized_size = sle64_to_cpu(
1412 				a->data.non_resident.initialized_size);
1413 		ni->allocated_size = sle64_to_cpu(
1414 				a->data.non_resident.allocated_size);
1415 	}
1416 	if (NInoMstProtected(ni))
1417 		vi->i_mapping->a_ops = &ntfs_mst_aops;
1418 	else
1419 		vi->i_mapping->a_ops = &ntfs_aops;
1420 	if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1421 		vi->i_blocks = ni->itype.compressed.size >> 9;
1422 	else
1423 		vi->i_blocks = ni->allocated_size >> 9;
1424 	/*
1425 	 * Make sure the base inode does not go away and attach it to the
1426 	 * attribute inode.
1427 	 */
1428 	igrab(base_vi);
1429 	ni->ext.base_ntfs_ino = base_ni;
1430 	ni->nr_extents = -1;
1431 
1432 	ntfs_attr_put_search_ctx(ctx);
1433 	unmap_mft_record(base_ni);
1434 
1435 	ntfs_debug("Done.");
1436 	return 0;
1437 
1438 unm_err_out:
1439 	if (!err)
1440 		err = -EIO;
1441 	if (ctx)
1442 		ntfs_attr_put_search_ctx(ctx);
1443 	unmap_mft_record(base_ni);
1444 err_out:
1445 	ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1446 			"inode (mft_no 0x%lx, type 0x%x, name_len %i).  "
1447 			"Marking corrupt inode and base inode 0x%lx as bad.  "
1448 			"Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1449 			base_vi->i_ino);
1450 	make_bad_inode(vi);
1451 	if (err != -ENOMEM)
1452 		NVolSetErrors(vol);
1453 	return err;
1454 }
1455 
1456 /**
1457  * ntfs_read_locked_index_inode - read an index inode from its base inode
1458  * @base_vi:	base inode
1459  * @vi:		index inode to read
1460  *
1461  * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1462  * index inode described by @vi into memory from the base mft record described
1463  * by @base_ni.
1464  *
1465  * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1466  * reading and looks up the attributes relating to the index described by @vi
1467  * before setting up the necessary fields in @vi as well as initializing the
1468  * ntfs inode.
1469  *
1470  * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1471  * with the attribute type set to AT_INDEX_ALLOCATION.  Apart from that, they
1472  * are setup like directory inodes since directories are a special case of
1473  * indices ao they need to be treated in much the same way.  Most importantly,
1474  * for small indices the index allocation attribute might not actually exist.
1475  * However, the index root attribute always exists but this does not need to
1476  * have an inode associated with it and this is why we define a new inode type
1477  * index.  Also, like for directories, we need to have an attribute inode for
1478  * the bitmap attribute corresponding to the index allocation attribute and we
1479  * can store this in the appropriate field of the inode, just like we do for
1480  * normal directory inodes.
1481  *
1482  * Q: What locks are held when the function is called?
1483  * A: i_state has I_NEW set, hence the inode is locked, also
1484  *    i_count is set to 1, so it is not going to go away
1485  *
1486  * Return 0 on success and -errno on error.  In the error case, the inode will
1487  * have had make_bad_inode() executed on it.
1488  */
ntfs_read_locked_index_inode(struct inode * base_vi,struct inode * vi)1489 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1490 {
1491 	loff_t bvi_size;
1492 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
1493 	ntfs_inode *ni, *base_ni, *bni;
1494 	struct inode *bvi;
1495 	MFT_RECORD *m;
1496 	ATTR_RECORD *a;
1497 	ntfs_attr_search_ctx *ctx;
1498 	INDEX_ROOT *ir;
1499 	u8 *ir_end, *index_end;
1500 	int err = 0;
1501 
1502 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1503 	ntfs_init_big_inode(vi);
1504 	ni	= NTFS_I(vi);
1505 	base_ni = NTFS_I(base_vi);
1506 	/* Just mirror the values from the base inode. */
1507 	vi->i_version	= base_vi->i_version;
1508 	vi->i_uid	= base_vi->i_uid;
1509 	vi->i_gid	= base_vi->i_gid;
1510 	set_nlink(vi, base_vi->i_nlink);
1511 	vi->i_mtime	= base_vi->i_mtime;
1512 	vi->i_ctime	= base_vi->i_ctime;
1513 	vi->i_atime	= base_vi->i_atime;
1514 	vi->i_generation = ni->seq_no = base_ni->seq_no;
1515 	/* Set inode type to zero but preserve permissions. */
1516 	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
1517 	/* Map the mft record for the base inode. */
1518 	m = map_mft_record(base_ni);
1519 	if (IS_ERR(m)) {
1520 		err = PTR_ERR(m);
1521 		goto err_out;
1522 	}
1523 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
1524 	if (!ctx) {
1525 		err = -ENOMEM;
1526 		goto unm_err_out;
1527 	}
1528 	/* Find the index root attribute. */
1529 	err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1530 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1531 	if (unlikely(err)) {
1532 		if (err == -ENOENT)
1533 			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1534 					"missing.");
1535 		goto unm_err_out;
1536 	}
1537 	a = ctx->attr;
1538 	/* Set up the state. */
1539 	if (unlikely(a->non_resident)) {
1540 		ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1541 		goto unm_err_out;
1542 	}
1543 	/* Ensure the attribute name is placed before the value. */
1544 	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1545 			le16_to_cpu(a->data.resident.value_offset)))) {
1546 		ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1547 				"after the attribute value.");
1548 		goto unm_err_out;
1549 	}
1550 	/*
1551 	 * Compressed/encrypted/sparse index root is not allowed, except for
1552 	 * directories of course but those are not dealt with here.
1553 	 */
1554 	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1555 			ATTR_IS_SPARSE)) {
1556 		ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1557 				"root attribute.");
1558 		goto unm_err_out;
1559 	}
1560 	ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1561 	ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1562 	if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1563 		ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1564 		goto unm_err_out;
1565 	}
1566 	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1567 	if (index_end > ir_end) {
1568 		ntfs_error(vi->i_sb, "Index is corrupt.");
1569 		goto unm_err_out;
1570 	}
1571 	if (ir->type) {
1572 		ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1573 				le32_to_cpu(ir->type));
1574 		goto unm_err_out;
1575 	}
1576 	ni->itype.index.collation_rule = ir->collation_rule;
1577 	ntfs_debug("Index collation rule is 0x%x.",
1578 			le32_to_cpu(ir->collation_rule));
1579 	ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1580 	if (!is_power_of_2(ni->itype.index.block_size)) {
1581 		ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1582 				"two.", ni->itype.index.block_size);
1583 		goto unm_err_out;
1584 	}
1585 	if (ni->itype.index.block_size > PAGE_CACHE_SIZE) {
1586 		ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE "
1587 				"(%ld) is not supported.  Sorry.",
1588 				ni->itype.index.block_size, PAGE_CACHE_SIZE);
1589 		err = -EOPNOTSUPP;
1590 		goto unm_err_out;
1591 	}
1592 	if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1593 		ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1594 				"(%i) is not supported.  Sorry.",
1595 				ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1596 		err = -EOPNOTSUPP;
1597 		goto unm_err_out;
1598 	}
1599 	ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1600 	/* Determine the size of a vcn in the index. */
1601 	if (vol->cluster_size <= ni->itype.index.block_size) {
1602 		ni->itype.index.vcn_size = vol->cluster_size;
1603 		ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1604 	} else {
1605 		ni->itype.index.vcn_size = vol->sector_size;
1606 		ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1607 	}
1608 	/* Check for presence of index allocation attribute. */
1609 	if (!(ir->index.flags & LARGE_INDEX)) {
1610 		/* No index allocation. */
1611 		vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1612 		/* We are done with the mft record, so we release it. */
1613 		ntfs_attr_put_search_ctx(ctx);
1614 		unmap_mft_record(base_ni);
1615 		m = NULL;
1616 		ctx = NULL;
1617 		goto skip_large_index_stuff;
1618 	} /* LARGE_INDEX:  Index allocation present.  Setup state. */
1619 	NInoSetIndexAllocPresent(ni);
1620 	/* Find index allocation attribute. */
1621 	ntfs_attr_reinit_search_ctx(ctx);
1622 	err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1623 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1624 	if (unlikely(err)) {
1625 		if (err == -ENOENT)
1626 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1627 					"not present but $INDEX_ROOT "
1628 					"indicated it is.");
1629 		else
1630 			ntfs_error(vi->i_sb, "Failed to lookup "
1631 					"$INDEX_ALLOCATION attribute.");
1632 		goto unm_err_out;
1633 	}
1634 	a = ctx->attr;
1635 	if (!a->non_resident) {
1636 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1637 				"resident.");
1638 		goto unm_err_out;
1639 	}
1640 	/*
1641 	 * Ensure the attribute name is placed before the mapping pairs array.
1642 	 */
1643 	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1644 			le16_to_cpu(
1645 			a->data.non_resident.mapping_pairs_offset)))) {
1646 		ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1647 				"placed after the mapping pairs array.");
1648 		goto unm_err_out;
1649 	}
1650 	if (a->flags & ATTR_IS_ENCRYPTED) {
1651 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1652 				"encrypted.");
1653 		goto unm_err_out;
1654 	}
1655 	if (a->flags & ATTR_IS_SPARSE) {
1656 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1657 		goto unm_err_out;
1658 	}
1659 	if (a->flags & ATTR_COMPRESSION_MASK) {
1660 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1661 				"compressed.");
1662 		goto unm_err_out;
1663 	}
1664 	if (a->data.non_resident.lowest_vcn) {
1665 		ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1666 				"attribute has non zero lowest_vcn.");
1667 		goto unm_err_out;
1668 	}
1669 	vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1670 	ni->initialized_size = sle64_to_cpu(
1671 			a->data.non_resident.initialized_size);
1672 	ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1673 	/*
1674 	 * We are done with the mft record, so we release it.  Otherwise
1675 	 * we would deadlock in ntfs_attr_iget().
1676 	 */
1677 	ntfs_attr_put_search_ctx(ctx);
1678 	unmap_mft_record(base_ni);
1679 	m = NULL;
1680 	ctx = NULL;
1681 	/* Get the index bitmap attribute inode. */
1682 	bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1683 	if (IS_ERR(bvi)) {
1684 		ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1685 		err = PTR_ERR(bvi);
1686 		goto unm_err_out;
1687 	}
1688 	bni = NTFS_I(bvi);
1689 	if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1690 			NInoSparse(bni)) {
1691 		ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1692 				"encrypted and/or sparse.");
1693 		goto iput_unm_err_out;
1694 	}
1695 	/* Consistency check bitmap size vs. index allocation size. */
1696 	bvi_size = i_size_read(bvi);
1697 	if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1698 		ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1699 				"index allocation (0x%llx).", bvi_size << 3,
1700 				vi->i_size);
1701 		goto iput_unm_err_out;
1702 	}
1703 	iput(bvi);
1704 skip_large_index_stuff:
1705 	/* Setup the operations for this index inode. */
1706 	vi->i_op = NULL;
1707 	vi->i_fop = NULL;
1708 	vi->i_mapping->a_ops = &ntfs_mst_aops;
1709 	vi->i_blocks = ni->allocated_size >> 9;
1710 	/*
1711 	 * Make sure the base inode doesn't go away and attach it to the
1712 	 * index inode.
1713 	 */
1714 	igrab(base_vi);
1715 	ni->ext.base_ntfs_ino = base_ni;
1716 	ni->nr_extents = -1;
1717 
1718 	ntfs_debug("Done.");
1719 	return 0;
1720 iput_unm_err_out:
1721 	iput(bvi);
1722 unm_err_out:
1723 	if (!err)
1724 		err = -EIO;
1725 	if (ctx)
1726 		ntfs_attr_put_search_ctx(ctx);
1727 	if (m)
1728 		unmap_mft_record(base_ni);
1729 err_out:
1730 	ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1731 			"inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1732 			ni->name_len);
1733 	make_bad_inode(vi);
1734 	if (err != -EOPNOTSUPP && err != -ENOMEM)
1735 		NVolSetErrors(vol);
1736 	return err;
1737 }
1738 
1739 /*
1740  * The MFT inode has special locking, so teach the lock validator
1741  * about this by splitting off the locking rules of the MFT from
1742  * the locking rules of other inodes. The MFT inode can never be
1743  * accessed from the VFS side (or even internally), only by the
1744  * map_mft functions.
1745  */
1746 static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
1747 
1748 /**
1749  * ntfs_read_inode_mount - special read_inode for mount time use only
1750  * @vi:		inode to read
1751  *
1752  * Read inode FILE_MFT at mount time, only called with super_block lock
1753  * held from within the read_super() code path.
1754  *
1755  * This function exists because when it is called the page cache for $MFT/$DATA
1756  * is not initialized and hence we cannot get at the contents of mft records
1757  * by calling map_mft_record*().
1758  *
1759  * Further it needs to cope with the circular references problem, i.e. cannot
1760  * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1761  * we do not know where the other extent mft records are yet and again, because
1762  * we cannot call map_mft_record*() yet.  Obviously this applies only when an
1763  * attribute list is actually present in $MFT inode.
1764  *
1765  * We solve these problems by starting with the $DATA attribute before anything
1766  * else and iterating using ntfs_attr_lookup($DATA) over all extents.  As each
1767  * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1768  * ntfs_runlists_merge().  Each step of the iteration necessarily provides
1769  * sufficient information for the next step to complete.
1770  *
1771  * This should work but there are two possible pit falls (see inline comments
1772  * below), but only time will tell if they are real pits or just smoke...
1773  */
ntfs_read_inode_mount(struct inode * vi)1774 int ntfs_read_inode_mount(struct inode *vi)
1775 {
1776 	VCN next_vcn, last_vcn, highest_vcn;
1777 	s64 block;
1778 	struct super_block *sb = vi->i_sb;
1779 	ntfs_volume *vol = NTFS_SB(sb);
1780 	struct buffer_head *bh;
1781 	ntfs_inode *ni;
1782 	MFT_RECORD *m = NULL;
1783 	ATTR_RECORD *a;
1784 	ntfs_attr_search_ctx *ctx;
1785 	unsigned int i, nr_blocks;
1786 	int err;
1787 
1788 	ntfs_debug("Entering.");
1789 
1790 	/* Initialize the ntfs specific part of @vi. */
1791 	ntfs_init_big_inode(vi);
1792 
1793 	ni = NTFS_I(vi);
1794 
1795 	/* Setup the data attribute. It is special as it is mst protected. */
1796 	NInoSetNonResident(ni);
1797 	NInoSetMstProtected(ni);
1798 	NInoSetSparseDisabled(ni);
1799 	ni->type = AT_DATA;
1800 	ni->name = NULL;
1801 	ni->name_len = 0;
1802 	/*
1803 	 * This sets up our little cheat allowing us to reuse the async read io
1804 	 * completion handler for directories.
1805 	 */
1806 	ni->itype.index.block_size = vol->mft_record_size;
1807 	ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1808 
1809 	/* Very important! Needed to be able to call map_mft_record*(). */
1810 	vol->mft_ino = vi;
1811 
1812 	/* Allocate enough memory to read the first mft record. */
1813 	if (vol->mft_record_size > 64 * 1024) {
1814 		ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1815 				vol->mft_record_size);
1816 		goto err_out;
1817 	}
1818 	i = vol->mft_record_size;
1819 	if (i < sb->s_blocksize)
1820 		i = sb->s_blocksize;
1821 	m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1822 	if (!m) {
1823 		ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1824 		goto err_out;
1825 	}
1826 
1827 	/* Determine the first block of the $MFT/$DATA attribute. */
1828 	block = vol->mft_lcn << vol->cluster_size_bits >>
1829 			sb->s_blocksize_bits;
1830 	nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1831 	if (!nr_blocks)
1832 		nr_blocks = 1;
1833 
1834 	/* Load $MFT/$DATA's first mft record. */
1835 	for (i = 0; i < nr_blocks; i++) {
1836 		bh = sb_bread(sb, block++);
1837 		if (!bh) {
1838 			ntfs_error(sb, "Device read failed.");
1839 			goto err_out;
1840 		}
1841 		memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1842 				sb->s_blocksize);
1843 		brelse(bh);
1844 	}
1845 
1846 	/* Apply the mst fixups. */
1847 	if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1848 		/* FIXME: Try to use the $MFTMirr now. */
1849 		ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1850 		goto err_out;
1851 	}
1852 
1853 	/* Need this to sanity check attribute list references to $MFT. */
1854 	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1855 
1856 	/* Provides readpage() and sync_page() for map_mft_record(). */
1857 	vi->i_mapping->a_ops = &ntfs_mst_aops;
1858 
1859 	ctx = ntfs_attr_get_search_ctx(ni, m);
1860 	if (!ctx) {
1861 		err = -ENOMEM;
1862 		goto err_out;
1863 	}
1864 
1865 	/* Find the attribute list attribute if present. */
1866 	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1867 	if (err) {
1868 		if (unlikely(err != -ENOENT)) {
1869 			ntfs_error(sb, "Failed to lookup attribute list "
1870 					"attribute. You should run chkdsk.");
1871 			goto put_err_out;
1872 		}
1873 	} else /* if (!err) */ {
1874 		ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1875 		u8 *al_end;
1876 		static const char *es = "  Not allowed.  $MFT is corrupt.  "
1877 				"You should run chkdsk.";
1878 
1879 		ntfs_debug("Attribute list attribute found in $MFT.");
1880 		NInoSetAttrList(ni);
1881 		a = ctx->attr;
1882 		if (a->flags & ATTR_COMPRESSION_MASK) {
1883 			ntfs_error(sb, "Attribute list attribute is "
1884 					"compressed.%s", es);
1885 			goto put_err_out;
1886 		}
1887 		if (a->flags & ATTR_IS_ENCRYPTED ||
1888 				a->flags & ATTR_IS_SPARSE) {
1889 			if (a->non_resident) {
1890 				ntfs_error(sb, "Non-resident attribute list "
1891 						"attribute is encrypted/"
1892 						"sparse.%s", es);
1893 				goto put_err_out;
1894 			}
1895 			ntfs_warning(sb, "Resident attribute list attribute "
1896 					"in $MFT system file is marked "
1897 					"encrypted/sparse which is not true.  "
1898 					"However, Windows allows this and "
1899 					"chkdsk does not detect or correct it "
1900 					"so we will just ignore the invalid "
1901 					"flags and pretend they are not set.");
1902 		}
1903 		/* Now allocate memory for the attribute list. */
1904 		ni->attr_list_size = (u32)ntfs_attr_size(a);
1905 		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1906 		if (!ni->attr_list) {
1907 			ntfs_error(sb, "Not enough memory to allocate buffer "
1908 					"for attribute list.");
1909 			goto put_err_out;
1910 		}
1911 		if (a->non_resident) {
1912 			NInoSetAttrListNonResident(ni);
1913 			if (a->data.non_resident.lowest_vcn) {
1914 				ntfs_error(sb, "Attribute list has non zero "
1915 						"lowest_vcn. $MFT is corrupt. "
1916 						"You should run chkdsk.");
1917 				goto put_err_out;
1918 			}
1919 			/* Setup the runlist. */
1920 			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1921 					a, NULL);
1922 			if (IS_ERR(ni->attr_list_rl.rl)) {
1923 				err = PTR_ERR(ni->attr_list_rl.rl);
1924 				ni->attr_list_rl.rl = NULL;
1925 				ntfs_error(sb, "Mapping pairs decompression "
1926 						"failed with error code %i.",
1927 						-err);
1928 				goto put_err_out;
1929 			}
1930 			/* Now load the attribute list. */
1931 			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1932 					ni->attr_list, ni->attr_list_size,
1933 					sle64_to_cpu(a->data.
1934 					non_resident.initialized_size)))) {
1935 				ntfs_error(sb, "Failed to load attribute list "
1936 						"attribute with error code %i.",
1937 						-err);
1938 				goto put_err_out;
1939 			}
1940 		} else /* if (!ctx.attr->non_resident) */ {
1941 			if ((u8*)a + le16_to_cpu(
1942 					a->data.resident.value_offset) +
1943 					le32_to_cpu(
1944 					a->data.resident.value_length) >
1945 					(u8*)ctx->mrec + vol->mft_record_size) {
1946 				ntfs_error(sb, "Corrupt attribute list "
1947 						"attribute.");
1948 				goto put_err_out;
1949 			}
1950 			/* Now copy the attribute list. */
1951 			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1952 					a->data.resident.value_offset),
1953 					le32_to_cpu(
1954 					a->data.resident.value_length));
1955 		}
1956 		/* The attribute list is now setup in memory. */
1957 		/*
1958 		 * FIXME: I don't know if this case is actually possible.
1959 		 * According to logic it is not possible but I have seen too
1960 		 * many weird things in MS software to rely on logic... Thus we
1961 		 * perform a manual search and make sure the first $MFT/$DATA
1962 		 * extent is in the base inode. If it is not we abort with an
1963 		 * error and if we ever see a report of this error we will need
1964 		 * to do some magic in order to have the necessary mft record
1965 		 * loaded and in the right place in the page cache. But
1966 		 * hopefully logic will prevail and this never happens...
1967 		 */
1968 		al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1969 		al_end = (u8*)al_entry + ni->attr_list_size;
1970 		for (;; al_entry = next_al_entry) {
1971 			/* Out of bounds check. */
1972 			if ((u8*)al_entry < ni->attr_list ||
1973 					(u8*)al_entry > al_end)
1974 				goto em_put_err_out;
1975 			/* Catch the end of the attribute list. */
1976 			if ((u8*)al_entry == al_end)
1977 				goto em_put_err_out;
1978 			if (!al_entry->length)
1979 				goto em_put_err_out;
1980 			if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1981 					le16_to_cpu(al_entry->length) > al_end)
1982 				goto em_put_err_out;
1983 			next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1984 					le16_to_cpu(al_entry->length));
1985 			if (le32_to_cpu(al_entry->type) > le32_to_cpu(AT_DATA))
1986 				goto em_put_err_out;
1987 			if (AT_DATA != al_entry->type)
1988 				continue;
1989 			/* We want an unnamed attribute. */
1990 			if (al_entry->name_length)
1991 				goto em_put_err_out;
1992 			/* Want the first entry, i.e. lowest_vcn == 0. */
1993 			if (al_entry->lowest_vcn)
1994 				goto em_put_err_out;
1995 			/* First entry has to be in the base mft record. */
1996 			if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1997 				/* MFT references do not match, logic fails. */
1998 				ntfs_error(sb, "BUG: The first $DATA extent "
1999 						"of $MFT is not in the base "
2000 						"mft record. Please report "
2001 						"you saw this message to "
2002 						"linux-ntfs-dev@lists."
2003 						"sourceforge.net");
2004 				goto put_err_out;
2005 			} else {
2006 				/* Sequence numbers must match. */
2007 				if (MSEQNO_LE(al_entry->mft_reference) !=
2008 						ni->seq_no)
2009 					goto em_put_err_out;
2010 				/* Got it. All is ok. We can stop now. */
2011 				break;
2012 			}
2013 		}
2014 	}
2015 
2016 	ntfs_attr_reinit_search_ctx(ctx);
2017 
2018 	/* Now load all attribute extents. */
2019 	a = NULL;
2020 	next_vcn = last_vcn = highest_vcn = 0;
2021 	while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
2022 			ctx))) {
2023 		runlist_element *nrl;
2024 
2025 		/* Cache the current attribute. */
2026 		a = ctx->attr;
2027 		/* $MFT must be non-resident. */
2028 		if (!a->non_resident) {
2029 			ntfs_error(sb, "$MFT must be non-resident but a "
2030 					"resident extent was found. $MFT is "
2031 					"corrupt. Run chkdsk.");
2032 			goto put_err_out;
2033 		}
2034 		/* $MFT must be uncompressed and unencrypted. */
2035 		if (a->flags & ATTR_COMPRESSION_MASK ||
2036 				a->flags & ATTR_IS_ENCRYPTED ||
2037 				a->flags & ATTR_IS_SPARSE) {
2038 			ntfs_error(sb, "$MFT must be uncompressed, "
2039 					"non-sparse, and unencrypted but a "
2040 					"compressed/sparse/encrypted extent "
2041 					"was found. $MFT is corrupt. Run "
2042 					"chkdsk.");
2043 			goto put_err_out;
2044 		}
2045 		/*
2046 		 * Decompress the mapping pairs array of this extent and merge
2047 		 * the result into the existing runlist. No need for locking
2048 		 * as we have exclusive access to the inode at this time and we
2049 		 * are a mount in progress task, too.
2050 		 */
2051 		nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2052 		if (IS_ERR(nrl)) {
2053 			ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2054 					"failed with error code %ld.  $MFT is "
2055 					"corrupt.", PTR_ERR(nrl));
2056 			goto put_err_out;
2057 		}
2058 		ni->runlist.rl = nrl;
2059 
2060 		/* Are we in the first extent? */
2061 		if (!next_vcn) {
2062 			if (a->data.non_resident.lowest_vcn) {
2063 				ntfs_error(sb, "First extent of $DATA "
2064 						"attribute has non zero "
2065 						"lowest_vcn. $MFT is corrupt. "
2066 						"You should run chkdsk.");
2067 				goto put_err_out;
2068 			}
2069 			/* Get the last vcn in the $DATA attribute. */
2070 			last_vcn = sle64_to_cpu(
2071 					a->data.non_resident.allocated_size)
2072 					>> vol->cluster_size_bits;
2073 			/* Fill in the inode size. */
2074 			vi->i_size = sle64_to_cpu(
2075 					a->data.non_resident.data_size);
2076 			ni->initialized_size = sle64_to_cpu(
2077 					a->data.non_resident.initialized_size);
2078 			ni->allocated_size = sle64_to_cpu(
2079 					a->data.non_resident.allocated_size);
2080 			/*
2081 			 * Verify the number of mft records does not exceed
2082 			 * 2^32 - 1.
2083 			 */
2084 			if ((vi->i_size >> vol->mft_record_size_bits) >=
2085 					(1ULL << 32)) {
2086 				ntfs_error(sb, "$MFT is too big! Aborting.");
2087 				goto put_err_out;
2088 			}
2089 			/*
2090 			 * We have got the first extent of the runlist for
2091 			 * $MFT which means it is now relatively safe to call
2092 			 * the normal ntfs_read_inode() function.
2093 			 * Complete reading the inode, this will actually
2094 			 * re-read the mft record for $MFT, this time entering
2095 			 * it into the page cache with which we complete the
2096 			 * kick start of the volume. It should be safe to do
2097 			 * this now as the first extent of $MFT/$DATA is
2098 			 * already known and we would hope that we don't need
2099 			 * further extents in order to find the other
2100 			 * attributes belonging to $MFT. Only time will tell if
2101 			 * this is really the case. If not we will have to play
2102 			 * magic at this point, possibly duplicating a lot of
2103 			 * ntfs_read_inode() at this point. We will need to
2104 			 * ensure we do enough of its work to be able to call
2105 			 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2106 			 * hope this never happens...
2107 			 */
2108 			ntfs_read_locked_inode(vi);
2109 			if (is_bad_inode(vi)) {
2110 				ntfs_error(sb, "ntfs_read_inode() of $MFT "
2111 						"failed. BUG or corrupt $MFT. "
2112 						"Run chkdsk and if no errors "
2113 						"are found, please report you "
2114 						"saw this message to "
2115 						"linux-ntfs-dev@lists."
2116 						"sourceforge.net");
2117 				ntfs_attr_put_search_ctx(ctx);
2118 				/* Revert to the safe super operations. */
2119 				ntfs_free(m);
2120 				return -1;
2121 			}
2122 			/*
2123 			 * Re-initialize some specifics about $MFT's inode as
2124 			 * ntfs_read_inode() will have set up the default ones.
2125 			 */
2126 			/* Set uid and gid to root. */
2127 			vi->i_uid = vi->i_gid = 0;
2128 			/* Regular file. No access for anyone. */
2129 			vi->i_mode = S_IFREG;
2130 			/* No VFS initiated operations allowed for $MFT. */
2131 			vi->i_op = &ntfs_empty_inode_ops;
2132 			vi->i_fop = &ntfs_empty_file_ops;
2133 		}
2134 
2135 		/* Get the lowest vcn for the next extent. */
2136 		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2137 		next_vcn = highest_vcn + 1;
2138 
2139 		/* Only one extent or error, which we catch below. */
2140 		if (next_vcn <= 0)
2141 			break;
2142 
2143 		/* Avoid endless loops due to corruption. */
2144 		if (next_vcn < sle64_to_cpu(
2145 				a->data.non_resident.lowest_vcn)) {
2146 			ntfs_error(sb, "$MFT has corrupt attribute list "
2147 					"attribute. Run chkdsk.");
2148 			goto put_err_out;
2149 		}
2150 	}
2151 	if (err != -ENOENT) {
2152 		ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2153 				"$MFT is corrupt. Run chkdsk.");
2154 		goto put_err_out;
2155 	}
2156 	if (!a) {
2157 		ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2158 				"corrupt. Run chkdsk.");
2159 		goto put_err_out;
2160 	}
2161 	if (highest_vcn && highest_vcn != last_vcn - 1) {
2162 		ntfs_error(sb, "Failed to load the complete runlist for "
2163 				"$MFT/$DATA. Driver bug or corrupt $MFT. "
2164 				"Run chkdsk.");
2165 		ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2166 				(unsigned long long)highest_vcn,
2167 				(unsigned long long)last_vcn - 1);
2168 		goto put_err_out;
2169 	}
2170 	ntfs_attr_put_search_ctx(ctx);
2171 	ntfs_debug("Done.");
2172 	ntfs_free(m);
2173 
2174 	/*
2175 	 * Split the locking rules of the MFT inode from the
2176 	 * locking rules of other inodes:
2177 	 */
2178 	lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
2179 	lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
2180 
2181 	return 0;
2182 
2183 em_put_err_out:
2184 	ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2185 			"attribute list. $MFT is corrupt. Run chkdsk.");
2186 put_err_out:
2187 	ntfs_attr_put_search_ctx(ctx);
2188 err_out:
2189 	ntfs_error(sb, "Failed. Marking inode as bad.");
2190 	make_bad_inode(vi);
2191 	ntfs_free(m);
2192 	return -1;
2193 }
2194 
__ntfs_clear_inode(ntfs_inode * ni)2195 static void __ntfs_clear_inode(ntfs_inode *ni)
2196 {
2197 	/* Free all alocated memory. */
2198 	down_write(&ni->runlist.lock);
2199 	if (ni->runlist.rl) {
2200 		ntfs_free(ni->runlist.rl);
2201 		ni->runlist.rl = NULL;
2202 	}
2203 	up_write(&ni->runlist.lock);
2204 
2205 	if (ni->attr_list) {
2206 		ntfs_free(ni->attr_list);
2207 		ni->attr_list = NULL;
2208 	}
2209 
2210 	down_write(&ni->attr_list_rl.lock);
2211 	if (ni->attr_list_rl.rl) {
2212 		ntfs_free(ni->attr_list_rl.rl);
2213 		ni->attr_list_rl.rl = NULL;
2214 	}
2215 	up_write(&ni->attr_list_rl.lock);
2216 
2217 	if (ni->name_len && ni->name != I30) {
2218 		/* Catch bugs... */
2219 		BUG_ON(!ni->name);
2220 		kfree(ni->name);
2221 	}
2222 }
2223 
ntfs_clear_extent_inode(ntfs_inode * ni)2224 void ntfs_clear_extent_inode(ntfs_inode *ni)
2225 {
2226 	ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2227 
2228 	BUG_ON(NInoAttr(ni));
2229 	BUG_ON(ni->nr_extents != -1);
2230 
2231 #ifdef NTFS_RW
2232 	if (NInoDirty(ni)) {
2233 		if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2234 			ntfs_error(ni->vol->sb, "Clearing dirty extent inode!  "
2235 					"Losing data!  This is a BUG!!!");
2236 		// FIXME:  Do something!!!
2237 	}
2238 #endif /* NTFS_RW */
2239 
2240 	__ntfs_clear_inode(ni);
2241 
2242 	/* Bye, bye... */
2243 	ntfs_destroy_extent_inode(ni);
2244 }
2245 
2246 /**
2247  * ntfs_evict_big_inode - clean up the ntfs specific part of an inode
2248  * @vi:		vfs inode pending annihilation
2249  *
2250  * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2251  * is called, which deallocates all memory belonging to the NTFS specific part
2252  * of the inode and returns.
2253  *
2254  * If the MFT record is dirty, we commit it before doing anything else.
2255  */
ntfs_evict_big_inode(struct inode * vi)2256 void ntfs_evict_big_inode(struct inode *vi)
2257 {
2258 	ntfs_inode *ni = NTFS_I(vi);
2259 
2260 	truncate_inode_pages(&vi->i_data, 0);
2261 	end_writeback(vi);
2262 
2263 #ifdef NTFS_RW
2264 	if (NInoDirty(ni)) {
2265 		bool was_bad = (is_bad_inode(vi));
2266 
2267 		/* Committing the inode also commits all extent inodes. */
2268 		ntfs_commit_inode(vi);
2269 
2270 		if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2271 			ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2272 					"0x%lx.  Losing data!", vi->i_ino);
2273 			// FIXME:  Do something!!!
2274 		}
2275 	}
2276 #endif /* NTFS_RW */
2277 
2278 	/* No need to lock at this stage as no one else has a reference. */
2279 	if (ni->nr_extents > 0) {
2280 		int i;
2281 
2282 		for (i = 0; i < ni->nr_extents; i++)
2283 			ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2284 		kfree(ni->ext.extent_ntfs_inos);
2285 	}
2286 
2287 	__ntfs_clear_inode(ni);
2288 
2289 	if (NInoAttr(ni)) {
2290 		/* Release the base inode if we are holding it. */
2291 		if (ni->nr_extents == -1) {
2292 			iput(VFS_I(ni->ext.base_ntfs_ino));
2293 			ni->nr_extents = 0;
2294 			ni->ext.base_ntfs_ino = NULL;
2295 		}
2296 	}
2297 	return;
2298 }
2299 
2300 /**
2301  * ntfs_show_options - show mount options in /proc/mounts
2302  * @sf:		seq_file in which to write our mount options
2303  * @root:	root of the mounted tree whose mount options to display
2304  *
2305  * Called by the VFS once for each mounted ntfs volume when someone reads
2306  * /proc/mounts in order to display the NTFS specific mount options of each
2307  * mount. The mount options of fs specified by @root are written to the seq file
2308  * @sf and success is returned.
2309  */
ntfs_show_options(struct seq_file * sf,struct dentry * root)2310 int ntfs_show_options(struct seq_file *sf, struct dentry *root)
2311 {
2312 	ntfs_volume *vol = NTFS_SB(root->d_sb);
2313 	int i;
2314 
2315 	seq_printf(sf, ",uid=%i", vol->uid);
2316 	seq_printf(sf, ",gid=%i", vol->gid);
2317 	if (vol->fmask == vol->dmask)
2318 		seq_printf(sf, ",umask=0%o", vol->fmask);
2319 	else {
2320 		seq_printf(sf, ",fmask=0%o", vol->fmask);
2321 		seq_printf(sf, ",dmask=0%o", vol->dmask);
2322 	}
2323 	seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2324 	if (NVolCaseSensitive(vol))
2325 		seq_printf(sf, ",case_sensitive");
2326 	if (NVolShowSystemFiles(vol))
2327 		seq_printf(sf, ",show_sys_files");
2328 	if (!NVolSparseEnabled(vol))
2329 		seq_printf(sf, ",disable_sparse");
2330 	for (i = 0; on_errors_arr[i].val; i++) {
2331 		if (on_errors_arr[i].val & vol->on_errors)
2332 			seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2333 	}
2334 	seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2335 	return 0;
2336 }
2337 
2338 #ifdef NTFS_RW
2339 
2340 static const char *es = "  Leaving inconsistent metadata.  Unmount and run "
2341 		"chkdsk.";
2342 
2343 /**
2344  * ntfs_truncate - called when the i_size of an ntfs inode is changed
2345  * @vi:		inode for which the i_size was changed
2346  *
2347  * We only support i_size changes for normal files at present, i.e. not
2348  * compressed and not encrypted.  This is enforced in ntfs_setattr(), see
2349  * below.
2350  *
2351  * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2352  * that the change is allowed.
2353  *
2354  * This implies for us that @vi is a file inode rather than a directory, index,
2355  * or attribute inode as well as that @vi is a base inode.
2356  *
2357  * Returns 0 on success or -errno on error.
2358  *
2359  * Called with ->i_mutex held.
2360  */
ntfs_truncate(struct inode * vi)2361 int ntfs_truncate(struct inode *vi)
2362 {
2363 	s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2364 	VCN highest_vcn;
2365 	unsigned long flags;
2366 	ntfs_inode *base_ni, *ni = NTFS_I(vi);
2367 	ntfs_volume *vol = ni->vol;
2368 	ntfs_attr_search_ctx *ctx;
2369 	MFT_RECORD *m;
2370 	ATTR_RECORD *a;
2371 	const char *te = "  Leaving file length out of sync with i_size.";
2372 	int err, mp_size, size_change, alloc_change;
2373 	u32 attr_len;
2374 
2375 	ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2376 	BUG_ON(NInoAttr(ni));
2377 	BUG_ON(S_ISDIR(vi->i_mode));
2378 	BUG_ON(NInoMstProtected(ni));
2379 	BUG_ON(ni->nr_extents < 0);
2380 retry_truncate:
2381 	/*
2382 	 * Lock the runlist for writing and map the mft record to ensure it is
2383 	 * safe to mess with the attribute runlist and sizes.
2384 	 */
2385 	down_write(&ni->runlist.lock);
2386 	if (!NInoAttr(ni))
2387 		base_ni = ni;
2388 	else
2389 		base_ni = ni->ext.base_ntfs_ino;
2390 	m = map_mft_record(base_ni);
2391 	if (IS_ERR(m)) {
2392 		err = PTR_ERR(m);
2393 		ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2394 				"(error code %d).%s", vi->i_ino, err, te);
2395 		ctx = NULL;
2396 		m = NULL;
2397 		goto old_bad_out;
2398 	}
2399 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
2400 	if (unlikely(!ctx)) {
2401 		ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2402 				"inode 0x%lx (not enough memory).%s",
2403 				vi->i_ino, te);
2404 		err = -ENOMEM;
2405 		goto old_bad_out;
2406 	}
2407 	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2408 			CASE_SENSITIVE, 0, NULL, 0, ctx);
2409 	if (unlikely(err)) {
2410 		if (err == -ENOENT) {
2411 			ntfs_error(vi->i_sb, "Open attribute is missing from "
2412 					"mft record.  Inode 0x%lx is corrupt.  "
2413 					"Run chkdsk.%s", vi->i_ino, te);
2414 			err = -EIO;
2415 		} else
2416 			ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2417 					"inode 0x%lx (error code %d).%s",
2418 					vi->i_ino, err, te);
2419 		goto old_bad_out;
2420 	}
2421 	m = ctx->mrec;
2422 	a = ctx->attr;
2423 	/*
2424 	 * The i_size of the vfs inode is the new size for the attribute value.
2425 	 */
2426 	new_size = i_size_read(vi);
2427 	/* The current size of the attribute value is the old size. */
2428 	old_size = ntfs_attr_size(a);
2429 	/* Calculate the new allocated size. */
2430 	if (NInoNonResident(ni))
2431 		new_alloc_size = (new_size + vol->cluster_size - 1) &
2432 				~(s64)vol->cluster_size_mask;
2433 	else
2434 		new_alloc_size = (new_size + 7) & ~7;
2435 	/* The current allocated size is the old allocated size. */
2436 	read_lock_irqsave(&ni->size_lock, flags);
2437 	old_alloc_size = ni->allocated_size;
2438 	read_unlock_irqrestore(&ni->size_lock, flags);
2439 	/*
2440 	 * The change in the file size.  This will be 0 if no change, >0 if the
2441 	 * size is growing, and <0 if the size is shrinking.
2442 	 */
2443 	size_change = -1;
2444 	if (new_size - old_size >= 0) {
2445 		size_change = 1;
2446 		if (new_size == old_size)
2447 			size_change = 0;
2448 	}
2449 	/* As above for the allocated size. */
2450 	alloc_change = -1;
2451 	if (new_alloc_size - old_alloc_size >= 0) {
2452 		alloc_change = 1;
2453 		if (new_alloc_size == old_alloc_size)
2454 			alloc_change = 0;
2455 	}
2456 	/*
2457 	 * If neither the size nor the allocation are being changed there is
2458 	 * nothing to do.
2459 	 */
2460 	if (!size_change && !alloc_change)
2461 		goto unm_done;
2462 	/* If the size is changing, check if new size is allowed in $AttrDef. */
2463 	if (size_change) {
2464 		err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2465 		if (unlikely(err)) {
2466 			if (err == -ERANGE) {
2467 				ntfs_error(vol->sb, "Truncate would cause the "
2468 						"inode 0x%lx to %simum size "
2469 						"for its attribute type "
2470 						"(0x%x).  Aborting truncate.",
2471 						vi->i_ino,
2472 						new_size > old_size ? "exceed "
2473 						"the max" : "go under the min",
2474 						le32_to_cpu(ni->type));
2475 				err = -EFBIG;
2476 			} else {
2477 				ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2478 						"attribute type 0x%x.  "
2479 						"Aborting truncate.",
2480 						vi->i_ino,
2481 						le32_to_cpu(ni->type));
2482 				err = -EIO;
2483 			}
2484 			/* Reset the vfs inode size to the old size. */
2485 			i_size_write(vi, old_size);
2486 			goto err_out;
2487 		}
2488 	}
2489 	if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2490 		ntfs_warning(vi->i_sb, "Changes in inode size are not "
2491 				"supported yet for %s files, ignoring.",
2492 				NInoCompressed(ni) ? "compressed" :
2493 				"encrypted");
2494 		err = -EOPNOTSUPP;
2495 		goto bad_out;
2496 	}
2497 	if (a->non_resident)
2498 		goto do_non_resident_truncate;
2499 	BUG_ON(NInoNonResident(ni));
2500 	/* Resize the attribute record to best fit the new attribute size. */
2501 	if (new_size < vol->mft_record_size &&
2502 			!ntfs_resident_attr_value_resize(m, a, new_size)) {
2503 		/* The resize succeeded! */
2504 		flush_dcache_mft_record_page(ctx->ntfs_ino);
2505 		mark_mft_record_dirty(ctx->ntfs_ino);
2506 		write_lock_irqsave(&ni->size_lock, flags);
2507 		/* Update the sizes in the ntfs inode and all is done. */
2508 		ni->allocated_size = le32_to_cpu(a->length) -
2509 				le16_to_cpu(a->data.resident.value_offset);
2510 		/*
2511 		 * Note ntfs_resident_attr_value_resize() has already done any
2512 		 * necessary data clearing in the attribute record.  When the
2513 		 * file is being shrunk vmtruncate() will already have cleared
2514 		 * the top part of the last partial page, i.e. since this is
2515 		 * the resident case this is the page with index 0.  However,
2516 		 * when the file is being expanded, the page cache page data
2517 		 * between the old data_size, i.e. old_size, and the new_size
2518 		 * has not been zeroed.  Fortunately, we do not need to zero it
2519 		 * either since on one hand it will either already be zero due
2520 		 * to both readpage and writepage clearing partial page data
2521 		 * beyond i_size in which case there is nothing to do or in the
2522 		 * case of the file being mmap()ped at the same time, POSIX
2523 		 * specifies that the behaviour is unspecified thus we do not
2524 		 * have to do anything.  This means that in our implementation
2525 		 * in the rare case that the file is mmap()ped and a write
2526 		 * occurred into the mmap()ped region just beyond the file size
2527 		 * and writepage has not yet been called to write out the page
2528 		 * (which would clear the area beyond the file size) and we now
2529 		 * extend the file size to incorporate this dirty region
2530 		 * outside the file size, a write of the page would result in
2531 		 * this data being written to disk instead of being cleared.
2532 		 * Given both POSIX and the Linux mmap(2) man page specify that
2533 		 * this corner case is undefined, we choose to leave it like
2534 		 * that as this is much simpler for us as we cannot lock the
2535 		 * relevant page now since we are holding too many ntfs locks
2536 		 * which would result in a lock reversal deadlock.
2537 		 */
2538 		ni->initialized_size = new_size;
2539 		write_unlock_irqrestore(&ni->size_lock, flags);
2540 		goto unm_done;
2541 	}
2542 	/* If the above resize failed, this must be an attribute extension. */
2543 	BUG_ON(size_change < 0);
2544 	/*
2545 	 * We have to drop all the locks so we can call
2546 	 * ntfs_attr_make_non_resident().  This could be optimised by try-
2547 	 * locking the first page cache page and only if that fails dropping
2548 	 * the locks, locking the page, and redoing all the locking and
2549 	 * lookups.  While this would be a huge optimisation, it is not worth
2550 	 * it as this is definitely a slow code path as it only ever can happen
2551 	 * once for any given file.
2552 	 */
2553 	ntfs_attr_put_search_ctx(ctx);
2554 	unmap_mft_record(base_ni);
2555 	up_write(&ni->runlist.lock);
2556 	/*
2557 	 * Not enough space in the mft record, try to make the attribute
2558 	 * non-resident and if successful restart the truncation process.
2559 	 */
2560 	err = ntfs_attr_make_non_resident(ni, old_size);
2561 	if (likely(!err))
2562 		goto retry_truncate;
2563 	/*
2564 	 * Could not make non-resident.  If this is due to this not being
2565 	 * permitted for this attribute type or there not being enough space,
2566 	 * try to make other attributes non-resident.  Otherwise fail.
2567 	 */
2568 	if (unlikely(err != -EPERM && err != -ENOSPC)) {
2569 		ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2570 				"type 0x%x, because the conversion from "
2571 				"resident to non-resident attribute failed "
2572 				"with error code %i.", vi->i_ino,
2573 				(unsigned)le32_to_cpu(ni->type), err);
2574 		if (err != -ENOMEM)
2575 			err = -EIO;
2576 		goto conv_err_out;
2577 	}
2578 	/* TODO: Not implemented from here, abort. */
2579 	if (err == -ENOSPC)
2580 		ntfs_error(vol->sb, "Not enough space in the mft record/on "
2581 				"disk for the non-resident attribute value.  "
2582 				"This case is not implemented yet.");
2583 	else /* if (err == -EPERM) */
2584 		ntfs_error(vol->sb, "This attribute type may not be "
2585 				"non-resident.  This case is not implemented "
2586 				"yet.");
2587 	err = -EOPNOTSUPP;
2588 	goto conv_err_out;
2589 #if 0
2590 	// TODO: Attempt to make other attributes non-resident.
2591 	if (!err)
2592 		goto do_resident_extend;
2593 	/*
2594 	 * Both the attribute list attribute and the standard information
2595 	 * attribute must remain in the base inode.  Thus, if this is one of
2596 	 * these attributes, we have to try to move other attributes out into
2597 	 * extent mft records instead.
2598 	 */
2599 	if (ni->type == AT_ATTRIBUTE_LIST ||
2600 			ni->type == AT_STANDARD_INFORMATION) {
2601 		// TODO: Attempt to move other attributes into extent mft
2602 		// records.
2603 		err = -EOPNOTSUPP;
2604 		if (!err)
2605 			goto do_resident_extend;
2606 		goto err_out;
2607 	}
2608 	// TODO: Attempt to move this attribute to an extent mft record, but
2609 	// only if it is not already the only attribute in an mft record in
2610 	// which case there would be nothing to gain.
2611 	err = -EOPNOTSUPP;
2612 	if (!err)
2613 		goto do_resident_extend;
2614 	/* There is nothing we can do to make enough space. )-: */
2615 	goto err_out;
2616 #endif
2617 do_non_resident_truncate:
2618 	BUG_ON(!NInoNonResident(ni));
2619 	if (alloc_change < 0) {
2620 		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2621 		if (highest_vcn > 0 &&
2622 				old_alloc_size >> vol->cluster_size_bits >
2623 				highest_vcn + 1) {
2624 			/*
2625 			 * This attribute has multiple extents.  Not yet
2626 			 * supported.
2627 			 */
2628 			ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2629 					"attribute type 0x%x, because the "
2630 					"attribute is highly fragmented (it "
2631 					"consists of multiple extents) and "
2632 					"this case is not implemented yet.",
2633 					vi->i_ino,
2634 					(unsigned)le32_to_cpu(ni->type));
2635 			err = -EOPNOTSUPP;
2636 			goto bad_out;
2637 		}
2638 	}
2639 	/*
2640 	 * If the size is shrinking, need to reduce the initialized_size and
2641 	 * the data_size before reducing the allocation.
2642 	 */
2643 	if (size_change < 0) {
2644 		/*
2645 		 * Make the valid size smaller (i_size is already up-to-date).
2646 		 */
2647 		write_lock_irqsave(&ni->size_lock, flags);
2648 		if (new_size < ni->initialized_size) {
2649 			ni->initialized_size = new_size;
2650 			a->data.non_resident.initialized_size =
2651 					cpu_to_sle64(new_size);
2652 		}
2653 		a->data.non_resident.data_size = cpu_to_sle64(new_size);
2654 		write_unlock_irqrestore(&ni->size_lock, flags);
2655 		flush_dcache_mft_record_page(ctx->ntfs_ino);
2656 		mark_mft_record_dirty(ctx->ntfs_ino);
2657 		/* If the allocated size is not changing, we are done. */
2658 		if (!alloc_change)
2659 			goto unm_done;
2660 		/*
2661 		 * If the size is shrinking it makes no sense for the
2662 		 * allocation to be growing.
2663 		 */
2664 		BUG_ON(alloc_change > 0);
2665 	} else /* if (size_change >= 0) */ {
2666 		/*
2667 		 * The file size is growing or staying the same but the
2668 		 * allocation can be shrinking, growing or staying the same.
2669 		 */
2670 		if (alloc_change > 0) {
2671 			/*
2672 			 * We need to extend the allocation and possibly update
2673 			 * the data size.  If we are updating the data size,
2674 			 * since we are not touching the initialized_size we do
2675 			 * not need to worry about the actual data on disk.
2676 			 * And as far as the page cache is concerned, there
2677 			 * will be no pages beyond the old data size and any
2678 			 * partial region in the last page between the old and
2679 			 * new data size (or the end of the page if the new
2680 			 * data size is outside the page) does not need to be
2681 			 * modified as explained above for the resident
2682 			 * attribute truncate case.  To do this, we simply drop
2683 			 * the locks we hold and leave all the work to our
2684 			 * friendly helper ntfs_attr_extend_allocation().
2685 			 */
2686 			ntfs_attr_put_search_ctx(ctx);
2687 			unmap_mft_record(base_ni);
2688 			up_write(&ni->runlist.lock);
2689 			err = ntfs_attr_extend_allocation(ni, new_size,
2690 					size_change > 0 ? new_size : -1, -1);
2691 			/*
2692 			 * ntfs_attr_extend_allocation() will have done error
2693 			 * output already.
2694 			 */
2695 			goto done;
2696 		}
2697 		if (!alloc_change)
2698 			goto alloc_done;
2699 	}
2700 	/* alloc_change < 0 */
2701 	/* Free the clusters. */
2702 	nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2703 			vol->cluster_size_bits, -1, ctx);
2704 	m = ctx->mrec;
2705 	a = ctx->attr;
2706 	if (unlikely(nr_freed < 0)) {
2707 		ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2708 				"%lli).  Unmount and run chkdsk to recover "
2709 				"the lost cluster(s).", (long long)nr_freed);
2710 		NVolSetErrors(vol);
2711 		nr_freed = 0;
2712 	}
2713 	/* Truncate the runlist. */
2714 	err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2715 			new_alloc_size >> vol->cluster_size_bits);
2716 	/*
2717 	 * If the runlist truncation failed and/or the search context is no
2718 	 * longer valid, we cannot resize the attribute record or build the
2719 	 * mapping pairs array thus we mark the inode bad so that no access to
2720 	 * the freed clusters can happen.
2721 	 */
2722 	if (unlikely(err || IS_ERR(m))) {
2723 		ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2724 				IS_ERR(m) ?
2725 				"restore attribute search context" :
2726 				"truncate attribute runlist",
2727 				IS_ERR(m) ? PTR_ERR(m) : err, es);
2728 		err = -EIO;
2729 		goto bad_out;
2730 	}
2731 	/* Get the size for the shrunk mapping pairs array for the runlist. */
2732 	mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2733 	if (unlikely(mp_size <= 0)) {
2734 		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2735 				"attribute type 0x%x, because determining the "
2736 				"size for the mapping pairs failed with error "
2737 				"code %i.%s", vi->i_ino,
2738 				(unsigned)le32_to_cpu(ni->type), mp_size, es);
2739 		err = -EIO;
2740 		goto bad_out;
2741 	}
2742 	/*
2743 	 * Shrink the attribute record for the new mapping pairs array.  Note,
2744 	 * this cannot fail since we are making the attribute smaller thus by
2745 	 * definition there is enough space to do so.
2746 	 */
2747 	attr_len = le32_to_cpu(a->length);
2748 	err = ntfs_attr_record_resize(m, a, mp_size +
2749 			le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2750 	BUG_ON(err);
2751 	/*
2752 	 * Generate the mapping pairs array directly into the attribute record.
2753 	 */
2754 	err = ntfs_mapping_pairs_build(vol, (u8*)a +
2755 			le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2756 			mp_size, ni->runlist.rl, 0, -1, NULL);
2757 	if (unlikely(err)) {
2758 		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2759 				"attribute type 0x%x, because building the "
2760 				"mapping pairs failed with error code %i.%s",
2761 				vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2762 				err, es);
2763 		err = -EIO;
2764 		goto bad_out;
2765 	}
2766 	/* Update the allocated/compressed size as well as the highest vcn. */
2767 	a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2768 			vol->cluster_size_bits) - 1);
2769 	write_lock_irqsave(&ni->size_lock, flags);
2770 	ni->allocated_size = new_alloc_size;
2771 	a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2772 	if (NInoSparse(ni) || NInoCompressed(ni)) {
2773 		if (nr_freed) {
2774 			ni->itype.compressed.size -= nr_freed <<
2775 					vol->cluster_size_bits;
2776 			BUG_ON(ni->itype.compressed.size < 0);
2777 			a->data.non_resident.compressed_size = cpu_to_sle64(
2778 					ni->itype.compressed.size);
2779 			vi->i_blocks = ni->itype.compressed.size >> 9;
2780 		}
2781 	} else
2782 		vi->i_blocks = new_alloc_size >> 9;
2783 	write_unlock_irqrestore(&ni->size_lock, flags);
2784 	/*
2785 	 * We have shrunk the allocation.  If this is a shrinking truncate we
2786 	 * have already dealt with the initialized_size and the data_size above
2787 	 * and we are done.  If the truncate is only changing the allocation
2788 	 * and not the data_size, we are also done.  If this is an extending
2789 	 * truncate, need to extend the data_size now which is ensured by the
2790 	 * fact that @size_change is positive.
2791 	 */
2792 alloc_done:
2793 	/*
2794 	 * If the size is growing, need to update it now.  If it is shrinking,
2795 	 * we have already updated it above (before the allocation change).
2796 	 */
2797 	if (size_change > 0)
2798 		a->data.non_resident.data_size = cpu_to_sle64(new_size);
2799 	/* Ensure the modified mft record is written out. */
2800 	flush_dcache_mft_record_page(ctx->ntfs_ino);
2801 	mark_mft_record_dirty(ctx->ntfs_ino);
2802 unm_done:
2803 	ntfs_attr_put_search_ctx(ctx);
2804 	unmap_mft_record(base_ni);
2805 	up_write(&ni->runlist.lock);
2806 done:
2807 	/* Update the mtime and ctime on the base inode. */
2808 	/* normally ->truncate shouldn't update ctime or mtime,
2809 	 * but ntfs did before so it got a copy & paste version
2810 	 * of file_update_time.  one day someone should fix this
2811 	 * for real.
2812 	 */
2813 	if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2814 		struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb);
2815 		int sync_it = 0;
2816 
2817 		if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2818 		    !timespec_equal(&VFS_I(base_ni)->i_ctime, &now))
2819 			sync_it = 1;
2820 		VFS_I(base_ni)->i_mtime = now;
2821 		VFS_I(base_ni)->i_ctime = now;
2822 
2823 		if (sync_it)
2824 			mark_inode_dirty_sync(VFS_I(base_ni));
2825 	}
2826 
2827 	if (likely(!err)) {
2828 		NInoClearTruncateFailed(ni);
2829 		ntfs_debug("Done.");
2830 	}
2831 	return err;
2832 old_bad_out:
2833 	old_size = -1;
2834 bad_out:
2835 	if (err != -ENOMEM && err != -EOPNOTSUPP)
2836 		NVolSetErrors(vol);
2837 	if (err != -EOPNOTSUPP)
2838 		NInoSetTruncateFailed(ni);
2839 	else if (old_size >= 0)
2840 		i_size_write(vi, old_size);
2841 err_out:
2842 	if (ctx)
2843 		ntfs_attr_put_search_ctx(ctx);
2844 	if (m)
2845 		unmap_mft_record(base_ni);
2846 	up_write(&ni->runlist.lock);
2847 out:
2848 	ntfs_debug("Failed.  Returning error code %i.", err);
2849 	return err;
2850 conv_err_out:
2851 	if (err != -ENOMEM && err != -EOPNOTSUPP)
2852 		NVolSetErrors(vol);
2853 	if (err != -EOPNOTSUPP)
2854 		NInoSetTruncateFailed(ni);
2855 	else
2856 		i_size_write(vi, old_size);
2857 	goto out;
2858 }
2859 
2860 /**
2861  * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2862  * @vi:		inode for which the i_size was changed
2863  *
2864  * Wrapper for ntfs_truncate() that has no return value.
2865  *
2866  * See ntfs_truncate() description above for details.
2867  */
ntfs_truncate_vfs(struct inode * vi)2868 void ntfs_truncate_vfs(struct inode *vi) {
2869 	ntfs_truncate(vi);
2870 }
2871 
2872 /**
2873  * ntfs_setattr - called from notify_change() when an attribute is being changed
2874  * @dentry:	dentry whose attributes to change
2875  * @attr:	structure describing the attributes and the changes
2876  *
2877  * We have to trap VFS attempts to truncate the file described by @dentry as
2878  * soon as possible, because we do not implement changes in i_size yet.  So we
2879  * abort all i_size changes here.
2880  *
2881  * We also abort all changes of user, group, and mode as we do not implement
2882  * the NTFS ACLs yet.
2883  *
2884  * Called with ->i_mutex held.
2885  */
ntfs_setattr(struct dentry * dentry,struct iattr * attr)2886 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2887 {
2888 	struct inode *vi = dentry->d_inode;
2889 	int err;
2890 	unsigned int ia_valid = attr->ia_valid;
2891 
2892 	err = inode_change_ok(vi, attr);
2893 	if (err)
2894 		goto out;
2895 	/* We do not support NTFS ACLs yet. */
2896 	if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2897 		ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2898 				"supported yet, ignoring.");
2899 		err = -EOPNOTSUPP;
2900 		goto out;
2901 	}
2902 	if (ia_valid & ATTR_SIZE) {
2903 		if (attr->ia_size != i_size_read(vi)) {
2904 			ntfs_inode *ni = NTFS_I(vi);
2905 			/*
2906 			 * FIXME: For now we do not support resizing of
2907 			 * compressed or encrypted files yet.
2908 			 */
2909 			if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2910 				ntfs_warning(vi->i_sb, "Changes in inode size "
2911 						"are not supported yet for "
2912 						"%s files, ignoring.",
2913 						NInoCompressed(ni) ?
2914 						"compressed" : "encrypted");
2915 				err = -EOPNOTSUPP;
2916 			} else
2917 				err = vmtruncate(vi, attr->ia_size);
2918 			if (err || ia_valid == ATTR_SIZE)
2919 				goto out;
2920 		} else {
2921 			/*
2922 			 * We skipped the truncate but must still update
2923 			 * timestamps.
2924 			 */
2925 			ia_valid |= ATTR_MTIME | ATTR_CTIME;
2926 		}
2927 	}
2928 	if (ia_valid & ATTR_ATIME)
2929 		vi->i_atime = timespec_trunc(attr->ia_atime,
2930 				vi->i_sb->s_time_gran);
2931 	if (ia_valid & ATTR_MTIME)
2932 		vi->i_mtime = timespec_trunc(attr->ia_mtime,
2933 				vi->i_sb->s_time_gran);
2934 	if (ia_valid & ATTR_CTIME)
2935 		vi->i_ctime = timespec_trunc(attr->ia_ctime,
2936 				vi->i_sb->s_time_gran);
2937 	mark_inode_dirty(vi);
2938 out:
2939 	return err;
2940 }
2941 
2942 /**
2943  * ntfs_write_inode - write out a dirty inode
2944  * @vi:		inode to write out
2945  * @sync:	if true, write out synchronously
2946  *
2947  * Write out a dirty inode to disk including any extent inodes if present.
2948  *
2949  * If @sync is true, commit the inode to disk and wait for io completion.  This
2950  * is done using write_mft_record().
2951  *
2952  * If @sync is false, just schedule the write to happen but do not wait for i/o
2953  * completion.  In 2.6 kernels, scheduling usually happens just by virtue of
2954  * marking the page (and in this case mft record) dirty but we do not implement
2955  * this yet as write_mft_record() largely ignores the @sync parameter and
2956  * always performs synchronous writes.
2957  *
2958  * Return 0 on success and -errno on error.
2959  */
__ntfs_write_inode(struct inode * vi,int sync)2960 int __ntfs_write_inode(struct inode *vi, int sync)
2961 {
2962 	sle64 nt;
2963 	ntfs_inode *ni = NTFS_I(vi);
2964 	ntfs_attr_search_ctx *ctx;
2965 	MFT_RECORD *m;
2966 	STANDARD_INFORMATION *si;
2967 	int err = 0;
2968 	bool modified = false;
2969 
2970 	ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2971 			vi->i_ino);
2972 	/*
2973 	 * Dirty attribute inodes are written via their real inodes so just
2974 	 * clean them here.  Access time updates are taken care off when the
2975 	 * real inode is written.
2976 	 */
2977 	if (NInoAttr(ni)) {
2978 		NInoClearDirty(ni);
2979 		ntfs_debug("Done.");
2980 		return 0;
2981 	}
2982 	/* Map, pin, and lock the mft record belonging to the inode. */
2983 	m = map_mft_record(ni);
2984 	if (IS_ERR(m)) {
2985 		err = PTR_ERR(m);
2986 		goto err_out;
2987 	}
2988 	/* Update the access times in the standard information attribute. */
2989 	ctx = ntfs_attr_get_search_ctx(ni, m);
2990 	if (unlikely(!ctx)) {
2991 		err = -ENOMEM;
2992 		goto unm_err_out;
2993 	}
2994 	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
2995 			CASE_SENSITIVE, 0, NULL, 0, ctx);
2996 	if (unlikely(err)) {
2997 		ntfs_attr_put_search_ctx(ctx);
2998 		goto unm_err_out;
2999 	}
3000 	si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
3001 			le16_to_cpu(ctx->attr->data.resident.value_offset));
3002 	/* Update the access times if they have changed. */
3003 	nt = utc2ntfs(vi->i_mtime);
3004 	if (si->last_data_change_time != nt) {
3005 		ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
3006 				"new = 0x%llx", vi->i_ino, (long long)
3007 				sle64_to_cpu(si->last_data_change_time),
3008 				(long long)sle64_to_cpu(nt));
3009 		si->last_data_change_time = nt;
3010 		modified = true;
3011 	}
3012 	nt = utc2ntfs(vi->i_ctime);
3013 	if (si->last_mft_change_time != nt) {
3014 		ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
3015 				"new = 0x%llx", vi->i_ino, (long long)
3016 				sle64_to_cpu(si->last_mft_change_time),
3017 				(long long)sle64_to_cpu(nt));
3018 		si->last_mft_change_time = nt;
3019 		modified = true;
3020 	}
3021 	nt = utc2ntfs(vi->i_atime);
3022 	if (si->last_access_time != nt) {
3023 		ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
3024 				"new = 0x%llx", vi->i_ino,
3025 				(long long)sle64_to_cpu(si->last_access_time),
3026 				(long long)sle64_to_cpu(nt));
3027 		si->last_access_time = nt;
3028 		modified = true;
3029 	}
3030 	/*
3031 	 * If we just modified the standard information attribute we need to
3032 	 * mark the mft record it is in dirty.  We do this manually so that
3033 	 * mark_inode_dirty() is not called which would redirty the inode and
3034 	 * hence result in an infinite loop of trying to write the inode.
3035 	 * There is no need to mark the base inode nor the base mft record
3036 	 * dirty, since we are going to write this mft record below in any case
3037 	 * and the base mft record may actually not have been modified so it
3038 	 * might not need to be written out.
3039 	 * NOTE: It is not a problem when the inode for $MFT itself is being
3040 	 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3041 	 * on the $MFT inode and hence ntfs_write_inode() will not be
3042 	 * re-invoked because of it which in turn is ok since the dirtied mft
3043 	 * record will be cleaned and written out to disk below, i.e. before
3044 	 * this function returns.
3045 	 */
3046 	if (modified) {
3047 		flush_dcache_mft_record_page(ctx->ntfs_ino);
3048 		if (!NInoTestSetDirty(ctx->ntfs_ino))
3049 			mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3050 					ctx->ntfs_ino->page_ofs);
3051 	}
3052 	ntfs_attr_put_search_ctx(ctx);
3053 	/* Now the access times are updated, write the base mft record. */
3054 	if (NInoDirty(ni))
3055 		err = write_mft_record(ni, m, sync);
3056 	/* Write all attached extent mft records. */
3057 	mutex_lock(&ni->extent_lock);
3058 	if (ni->nr_extents > 0) {
3059 		ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3060 		int i;
3061 
3062 		ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3063 		for (i = 0; i < ni->nr_extents; i++) {
3064 			ntfs_inode *tni = extent_nis[i];
3065 
3066 			if (NInoDirty(tni)) {
3067 				MFT_RECORD *tm = map_mft_record(tni);
3068 				int ret;
3069 
3070 				if (IS_ERR(tm)) {
3071 					if (!err || err == -ENOMEM)
3072 						err = PTR_ERR(tm);
3073 					continue;
3074 				}
3075 				ret = write_mft_record(tni, tm, sync);
3076 				unmap_mft_record(tni);
3077 				if (unlikely(ret)) {
3078 					if (!err || err == -ENOMEM)
3079 						err = ret;
3080 				}
3081 			}
3082 		}
3083 	}
3084 	mutex_unlock(&ni->extent_lock);
3085 	unmap_mft_record(ni);
3086 	if (unlikely(err))
3087 		goto err_out;
3088 	ntfs_debug("Done.");
3089 	return 0;
3090 unm_err_out:
3091 	unmap_mft_record(ni);
3092 err_out:
3093 	if (err == -ENOMEM) {
3094 		ntfs_warning(vi->i_sb, "Not enough memory to write inode.  "
3095 				"Marking the inode dirty again, so the VFS "
3096 				"retries later.");
3097 		mark_inode_dirty(vi);
3098 	} else {
3099 		ntfs_error(vi->i_sb, "Failed (error %i):  Run chkdsk.", -err);
3100 		NVolSetErrors(ni->vol);
3101 	}
3102 	return err;
3103 }
3104 
3105 #endif /* NTFS_RW */
3106