xref: /openbmc/linux/fs/ntfs/inode.c (revision dc6a81c3)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /**
3  * inode.c - NTFS kernel inode handling.
4  *
5  * Copyright (c) 2001-2014 Anton Altaparmakov and Tuxera Inc.
6  */
7 
8 #include <linux/buffer_head.h>
9 #include <linux/fs.h>
10 #include <linux/mm.h>
11 #include <linux/mount.h>
12 #include <linux/mutex.h>
13 #include <linux/pagemap.h>
14 #include <linux/quotaops.h>
15 #include <linux/slab.h>
16 #include <linux/log2.h>
17 
18 #include "aops.h"
19 #include "attrib.h"
20 #include "bitmap.h"
21 #include "dir.h"
22 #include "debug.h"
23 #include "inode.h"
24 #include "lcnalloc.h"
25 #include "malloc.h"
26 #include "mft.h"
27 #include "time.h"
28 #include "ntfs.h"
29 
30 /**
31  * ntfs_test_inode - compare two (possibly fake) inodes for equality
32  * @vi:		vfs inode which to test
33  * @na:		ntfs attribute which is being tested with
34  *
35  * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
36  * inode @vi for equality with the ntfs attribute @na.
37  *
38  * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
39  * @na->name and @na->name_len are then ignored.
40  *
41  * Return 1 if the attributes match and 0 if not.
42  *
43  * NOTE: This function runs with the inode_hash_lock spin lock held so it is not
44  * allowed to sleep.
45  */
46 int ntfs_test_inode(struct inode *vi, ntfs_attr *na)
47 {
48 	ntfs_inode *ni;
49 
50 	if (vi->i_ino != na->mft_no)
51 		return 0;
52 	ni = NTFS_I(vi);
53 	/* If !NInoAttr(ni), @vi is a normal file or directory inode. */
54 	if (likely(!NInoAttr(ni))) {
55 		/* If not looking for a normal inode this is a mismatch. */
56 		if (unlikely(na->type != AT_UNUSED))
57 			return 0;
58 	} else {
59 		/* A fake inode describing an attribute. */
60 		if (ni->type != na->type)
61 			return 0;
62 		if (ni->name_len != na->name_len)
63 			return 0;
64 		if (na->name_len && memcmp(ni->name, na->name,
65 				na->name_len * sizeof(ntfschar)))
66 			return 0;
67 	}
68 	/* Match! */
69 	return 1;
70 }
71 
72 /**
73  * ntfs_init_locked_inode - initialize an inode
74  * @vi:		vfs inode to initialize
75  * @na:		ntfs attribute which to initialize @vi to
76  *
77  * Initialize the vfs inode @vi with the values from the ntfs attribute @na in
78  * order to enable ntfs_test_inode() to do its work.
79  *
80  * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
81  * In that case, @na->name and @na->name_len should be set to NULL and 0,
82  * respectively. Although that is not strictly necessary as
83  * ntfs_read_locked_inode() will fill them in later.
84  *
85  * Return 0 on success and -errno on error.
86  *
87  * NOTE: This function runs with the inode->i_lock spin lock held so it is not
88  * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
89  */
90 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na)
91 {
92 	ntfs_inode *ni = NTFS_I(vi);
93 
94 	vi->i_ino = na->mft_no;
95 
96 	ni->type = na->type;
97 	if (na->type == AT_INDEX_ALLOCATION)
98 		NInoSetMstProtected(ni);
99 
100 	ni->name = na->name;
101 	ni->name_len = na->name_len;
102 
103 	/* If initializing a normal inode, we are done. */
104 	if (likely(na->type == AT_UNUSED)) {
105 		BUG_ON(na->name);
106 		BUG_ON(na->name_len);
107 		return 0;
108 	}
109 
110 	/* It is a fake inode. */
111 	NInoSetAttr(ni);
112 
113 	/*
114 	 * We have I30 global constant as an optimization as it is the name
115 	 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
116 	 * allocation but that is ok. And most attributes are unnamed anyway,
117 	 * thus the fraction of named attributes with name != I30 is actually
118 	 * absolutely tiny.
119 	 */
120 	if (na->name_len && na->name != I30) {
121 		unsigned int i;
122 
123 		BUG_ON(!na->name);
124 		i = na->name_len * sizeof(ntfschar);
125 		ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
126 		if (!ni->name)
127 			return -ENOMEM;
128 		memcpy(ni->name, na->name, i);
129 		ni->name[na->name_len] = 0;
130 	}
131 	return 0;
132 }
133 
134 typedef int (*set_t)(struct inode *, void *);
135 static int ntfs_read_locked_inode(struct inode *vi);
136 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
137 static int ntfs_read_locked_index_inode(struct inode *base_vi,
138 		struct inode *vi);
139 
140 /**
141  * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
142  * @sb:		super block of mounted volume
143  * @mft_no:	mft record number / inode number to obtain
144  *
145  * Obtain the struct inode corresponding to a specific normal inode (i.e. a
146  * file or directory).
147  *
148  * If the inode is in the cache, it is just returned with an increased
149  * reference count. Otherwise, a new struct inode is allocated and initialized,
150  * and finally ntfs_read_locked_inode() is called to read in the inode and
151  * fill in the remainder of the inode structure.
152  *
153  * Return the struct inode on success. Check the return value with IS_ERR() and
154  * if true, the function failed and the error code is obtained from PTR_ERR().
155  */
156 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
157 {
158 	struct inode *vi;
159 	int err;
160 	ntfs_attr na;
161 
162 	na.mft_no = mft_no;
163 	na.type = AT_UNUSED;
164 	na.name = NULL;
165 	na.name_len = 0;
166 
167 	vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode,
168 			(set_t)ntfs_init_locked_inode, &na);
169 	if (unlikely(!vi))
170 		return ERR_PTR(-ENOMEM);
171 
172 	err = 0;
173 
174 	/* If this is a freshly allocated inode, need to read it now. */
175 	if (vi->i_state & I_NEW) {
176 		err = ntfs_read_locked_inode(vi);
177 		unlock_new_inode(vi);
178 	}
179 	/*
180 	 * There is no point in keeping bad inodes around if the failure was
181 	 * due to ENOMEM. We want to be able to retry again later.
182 	 */
183 	if (unlikely(err == -ENOMEM)) {
184 		iput(vi);
185 		vi = ERR_PTR(err);
186 	}
187 	return vi;
188 }
189 
190 /**
191  * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
192  * @base_vi:	vfs base inode containing the attribute
193  * @type:	attribute type
194  * @name:	Unicode name of the attribute (NULL if unnamed)
195  * @name_len:	length of @name in Unicode characters (0 if unnamed)
196  *
197  * Obtain the (fake) struct inode corresponding to the attribute specified by
198  * @type, @name, and @name_len, which is present in the base mft record
199  * specified by the vfs inode @base_vi.
200  *
201  * If the attribute inode is in the cache, it is just returned with an
202  * increased reference count. Otherwise, a new struct inode is allocated and
203  * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
204  * attribute and fill in the inode structure.
205  *
206  * Note, for index allocation attributes, you need to use ntfs_index_iget()
207  * instead of ntfs_attr_iget() as working with indices is a lot more complex.
208  *
209  * Return the struct inode of the attribute inode on success. Check the return
210  * value with IS_ERR() and if true, the function failed and the error code is
211  * obtained from PTR_ERR().
212  */
213 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
214 		ntfschar *name, u32 name_len)
215 {
216 	struct inode *vi;
217 	int err;
218 	ntfs_attr na;
219 
220 	/* Make sure no one calls ntfs_attr_iget() for indices. */
221 	BUG_ON(type == AT_INDEX_ALLOCATION);
222 
223 	na.mft_no = base_vi->i_ino;
224 	na.type = type;
225 	na.name = name;
226 	na.name_len = name_len;
227 
228 	vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
229 			(set_t)ntfs_init_locked_inode, &na);
230 	if (unlikely(!vi))
231 		return ERR_PTR(-ENOMEM);
232 
233 	err = 0;
234 
235 	/* If this is a freshly allocated inode, need to read it now. */
236 	if (vi->i_state & I_NEW) {
237 		err = ntfs_read_locked_attr_inode(base_vi, vi);
238 		unlock_new_inode(vi);
239 	}
240 	/*
241 	 * There is no point in keeping bad attribute inodes around. This also
242 	 * simplifies things in that we never need to check for bad attribute
243 	 * inodes elsewhere.
244 	 */
245 	if (unlikely(err)) {
246 		iput(vi);
247 		vi = ERR_PTR(err);
248 	}
249 	return vi;
250 }
251 
252 /**
253  * ntfs_index_iget - obtain a struct inode corresponding to an index
254  * @base_vi:	vfs base inode containing the index related attributes
255  * @name:	Unicode name of the index
256  * @name_len:	length of @name in Unicode characters
257  *
258  * Obtain the (fake) struct inode corresponding to the index specified by @name
259  * and @name_len, which is present in the base mft record specified by the vfs
260  * inode @base_vi.
261  *
262  * If the index inode is in the cache, it is just returned with an increased
263  * reference count.  Otherwise, a new struct inode is allocated and
264  * initialized, and finally ntfs_read_locked_index_inode() is called to read
265  * the index related attributes and fill in the inode structure.
266  *
267  * Return the struct inode of the index inode on success. Check the return
268  * value with IS_ERR() and if true, the function failed and the error code is
269  * obtained from PTR_ERR().
270  */
271 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
272 		u32 name_len)
273 {
274 	struct inode *vi;
275 	int err;
276 	ntfs_attr na;
277 
278 	na.mft_no = base_vi->i_ino;
279 	na.type = AT_INDEX_ALLOCATION;
280 	na.name = name;
281 	na.name_len = name_len;
282 
283 	vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode,
284 			(set_t)ntfs_init_locked_inode, &na);
285 	if (unlikely(!vi))
286 		return ERR_PTR(-ENOMEM);
287 
288 	err = 0;
289 
290 	/* If this is a freshly allocated inode, need to read it now. */
291 	if (vi->i_state & I_NEW) {
292 		err = ntfs_read_locked_index_inode(base_vi, vi);
293 		unlock_new_inode(vi);
294 	}
295 	/*
296 	 * There is no point in keeping bad index inodes around.  This also
297 	 * simplifies things in that we never need to check for bad index
298 	 * inodes elsewhere.
299 	 */
300 	if (unlikely(err)) {
301 		iput(vi);
302 		vi = ERR_PTR(err);
303 	}
304 	return vi;
305 }
306 
307 struct inode *ntfs_alloc_big_inode(struct super_block *sb)
308 {
309 	ntfs_inode *ni;
310 
311 	ntfs_debug("Entering.");
312 	ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_NOFS);
313 	if (likely(ni != NULL)) {
314 		ni->state = 0;
315 		return VFS_I(ni);
316 	}
317 	ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
318 	return NULL;
319 }
320 
321 void ntfs_free_big_inode(struct inode *inode)
322 {
323 	kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
324 }
325 
326 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
327 {
328 	ntfs_inode *ni;
329 
330 	ntfs_debug("Entering.");
331 	ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS);
332 	if (likely(ni != NULL)) {
333 		ni->state = 0;
334 		return ni;
335 	}
336 	ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
337 	return NULL;
338 }
339 
340 static void ntfs_destroy_extent_inode(ntfs_inode *ni)
341 {
342 	ntfs_debug("Entering.");
343 	BUG_ON(ni->page);
344 	if (!atomic_dec_and_test(&ni->count))
345 		BUG();
346 	kmem_cache_free(ntfs_inode_cache, ni);
347 }
348 
349 /*
350  * The attribute runlist lock has separate locking rules from the
351  * normal runlist lock, so split the two lock-classes:
352  */
353 static struct lock_class_key attr_list_rl_lock_class;
354 
355 /**
356  * __ntfs_init_inode - initialize ntfs specific part of an inode
357  * @sb:		super block of mounted volume
358  * @ni:		freshly allocated ntfs inode which to initialize
359  *
360  * Initialize an ntfs inode to defaults.
361  *
362  * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
363  * untouched. Make sure to initialize them elsewhere.
364  *
365  * Return zero on success and -ENOMEM on error.
366  */
367 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
368 {
369 	ntfs_debug("Entering.");
370 	rwlock_init(&ni->size_lock);
371 	ni->initialized_size = ni->allocated_size = 0;
372 	ni->seq_no = 0;
373 	atomic_set(&ni->count, 1);
374 	ni->vol = NTFS_SB(sb);
375 	ntfs_init_runlist(&ni->runlist);
376 	mutex_init(&ni->mrec_lock);
377 	ni->page = NULL;
378 	ni->page_ofs = 0;
379 	ni->attr_list_size = 0;
380 	ni->attr_list = NULL;
381 	ntfs_init_runlist(&ni->attr_list_rl);
382 	lockdep_set_class(&ni->attr_list_rl.lock,
383 				&attr_list_rl_lock_class);
384 	ni->itype.index.block_size = 0;
385 	ni->itype.index.vcn_size = 0;
386 	ni->itype.index.collation_rule = 0;
387 	ni->itype.index.block_size_bits = 0;
388 	ni->itype.index.vcn_size_bits = 0;
389 	mutex_init(&ni->extent_lock);
390 	ni->nr_extents = 0;
391 	ni->ext.base_ntfs_ino = NULL;
392 }
393 
394 /*
395  * Extent inodes get MFT-mapped in a nested way, while the base inode
396  * is still mapped. Teach this nesting to the lock validator by creating
397  * a separate class for nested inode's mrec_lock's:
398  */
399 static struct lock_class_key extent_inode_mrec_lock_key;
400 
401 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
402 		unsigned long mft_no)
403 {
404 	ntfs_inode *ni = ntfs_alloc_extent_inode();
405 
406 	ntfs_debug("Entering.");
407 	if (likely(ni != NULL)) {
408 		__ntfs_init_inode(sb, ni);
409 		lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
410 		ni->mft_no = mft_no;
411 		ni->type = AT_UNUSED;
412 		ni->name = NULL;
413 		ni->name_len = 0;
414 	}
415 	return ni;
416 }
417 
418 /**
419  * ntfs_is_extended_system_file - check if a file is in the $Extend directory
420  * @ctx:	initialized attribute search context
421  *
422  * Search all file name attributes in the inode described by the attribute
423  * search context @ctx and check if any of the names are in the $Extend system
424  * directory.
425  *
426  * Return values:
427  *	   1: file is in $Extend directory
428  *	   0: file is not in $Extend directory
429  *    -errno: failed to determine if the file is in the $Extend directory
430  */
431 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
432 {
433 	int nr_links, err;
434 
435 	/* Restart search. */
436 	ntfs_attr_reinit_search_ctx(ctx);
437 
438 	/* Get number of hard links. */
439 	nr_links = le16_to_cpu(ctx->mrec->link_count);
440 
441 	/* Loop through all hard links. */
442 	while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
443 			ctx))) {
444 		FILE_NAME_ATTR *file_name_attr;
445 		ATTR_RECORD *attr = ctx->attr;
446 		u8 *p, *p2;
447 
448 		nr_links--;
449 		/*
450 		 * Maximum sanity checking as we are called on an inode that
451 		 * we suspect might be corrupt.
452 		 */
453 		p = (u8*)attr + le32_to_cpu(attr->length);
454 		if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
455 				le32_to_cpu(ctx->mrec->bytes_in_use)) {
456 err_corrupt_attr:
457 			ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
458 					"attribute. You should run chkdsk.");
459 			return -EIO;
460 		}
461 		if (attr->non_resident) {
462 			ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
463 					"name. You should run chkdsk.");
464 			return -EIO;
465 		}
466 		if (attr->flags) {
467 			ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
468 					"invalid flags. You should run "
469 					"chkdsk.");
470 			return -EIO;
471 		}
472 		if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
473 			ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
474 					"name. You should run chkdsk.");
475 			return -EIO;
476 		}
477 		file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
478 				le16_to_cpu(attr->data.resident.value_offset));
479 		p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length);
480 		if (p2 < (u8*)attr || p2 > p)
481 			goto err_corrupt_attr;
482 		/* This attribute is ok, but is it in the $Extend directory? */
483 		if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
484 			return 1;	/* YES, it's an extended system file. */
485 	}
486 	if (unlikely(err != -ENOENT))
487 		return err;
488 	if (unlikely(nr_links)) {
489 		ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
490 				"doesn't match number of name attributes. You "
491 				"should run chkdsk.");
492 		return -EIO;
493 	}
494 	return 0;	/* NO, it is not an extended system file. */
495 }
496 
497 /**
498  * ntfs_read_locked_inode - read an inode from its device
499  * @vi:		inode to read
500  *
501  * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
502  * described by @vi into memory from the device.
503  *
504  * The only fields in @vi that we need to/can look at when the function is
505  * called are i_sb, pointing to the mounted device's super block, and i_ino,
506  * the number of the inode to load.
507  *
508  * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
509  * for reading and sets up the necessary @vi fields as well as initializing
510  * the ntfs inode.
511  *
512  * Q: What locks are held when the function is called?
513  * A: i_state has I_NEW set, hence the inode is locked, also
514  *    i_count is set to 1, so it is not going to go away
515  *    i_flags is set to 0 and we have no business touching it.  Only an ioctl()
516  *    is allowed to write to them. We should of course be honouring them but
517  *    we need to do that using the IS_* macros defined in include/linux/fs.h.
518  *    In any case ntfs_read_locked_inode() has nothing to do with i_flags.
519  *
520  * Return 0 on success and -errno on error.  In the error case, the inode will
521  * have had make_bad_inode() executed on it.
522  */
523 static int ntfs_read_locked_inode(struct inode *vi)
524 {
525 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
526 	ntfs_inode *ni;
527 	struct inode *bvi;
528 	MFT_RECORD *m;
529 	ATTR_RECORD *a;
530 	STANDARD_INFORMATION *si;
531 	ntfs_attr_search_ctx *ctx;
532 	int err = 0;
533 
534 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
535 
536 	/* Setup the generic vfs inode parts now. */
537 	vi->i_uid = vol->uid;
538 	vi->i_gid = vol->gid;
539 	vi->i_mode = 0;
540 
541 	/*
542 	 * Initialize the ntfs specific part of @vi special casing
543 	 * FILE_MFT which we need to do at mount time.
544 	 */
545 	if (vi->i_ino != FILE_MFT)
546 		ntfs_init_big_inode(vi);
547 	ni = NTFS_I(vi);
548 
549 	m = map_mft_record(ni);
550 	if (IS_ERR(m)) {
551 		err = PTR_ERR(m);
552 		goto err_out;
553 	}
554 	ctx = ntfs_attr_get_search_ctx(ni, m);
555 	if (!ctx) {
556 		err = -ENOMEM;
557 		goto unm_err_out;
558 	}
559 
560 	if (!(m->flags & MFT_RECORD_IN_USE)) {
561 		ntfs_error(vi->i_sb, "Inode is not in use!");
562 		goto unm_err_out;
563 	}
564 	if (m->base_mft_record) {
565 		ntfs_error(vi->i_sb, "Inode is an extent inode!");
566 		goto unm_err_out;
567 	}
568 
569 	/* Transfer information from mft record into vfs and ntfs inodes. */
570 	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
571 
572 	/*
573 	 * FIXME: Keep in mind that link_count is two for files which have both
574 	 * a long file name and a short file name as separate entries, so if
575 	 * we are hiding short file names this will be too high. Either we need
576 	 * to account for the short file names by subtracting them or we need
577 	 * to make sure we delete files even though i_nlink is not zero which
578 	 * might be tricky due to vfs interactions. Need to think about this
579 	 * some more when implementing the unlink command.
580 	 */
581 	set_nlink(vi, le16_to_cpu(m->link_count));
582 	/*
583 	 * FIXME: Reparse points can have the directory bit set even though
584 	 * they would be S_IFLNK. Need to deal with this further below when we
585 	 * implement reparse points / symbolic links but it will do for now.
586 	 * Also if not a directory, it could be something else, rather than
587 	 * a regular file. But again, will do for now.
588 	 */
589 	/* Everyone gets all permissions. */
590 	vi->i_mode |= S_IRWXUGO;
591 	/* If read-only, no one gets write permissions. */
592 	if (IS_RDONLY(vi))
593 		vi->i_mode &= ~S_IWUGO;
594 	if (m->flags & MFT_RECORD_IS_DIRECTORY) {
595 		vi->i_mode |= S_IFDIR;
596 		/*
597 		 * Apply the directory permissions mask set in the mount
598 		 * options.
599 		 */
600 		vi->i_mode &= ~vol->dmask;
601 		/* Things break without this kludge! */
602 		if (vi->i_nlink > 1)
603 			set_nlink(vi, 1);
604 	} else {
605 		vi->i_mode |= S_IFREG;
606 		/* Apply the file permissions mask set in the mount options. */
607 		vi->i_mode &= ~vol->fmask;
608 	}
609 	/*
610 	 * Find the standard information attribute in the mft record. At this
611 	 * stage we haven't setup the attribute list stuff yet, so this could
612 	 * in fact fail if the standard information is in an extent record, but
613 	 * I don't think this actually ever happens.
614 	 */
615 	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
616 			ctx);
617 	if (unlikely(err)) {
618 		if (err == -ENOENT) {
619 			/*
620 			 * TODO: We should be performing a hot fix here (if the
621 			 * recover mount option is set) by creating a new
622 			 * attribute.
623 			 */
624 			ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
625 					"is missing.");
626 		}
627 		goto unm_err_out;
628 	}
629 	a = ctx->attr;
630 	/* Get the standard information attribute value. */
631 	si = (STANDARD_INFORMATION*)((u8*)a +
632 			le16_to_cpu(a->data.resident.value_offset));
633 
634 	/* Transfer information from the standard information into vi. */
635 	/*
636 	 * Note: The i_?times do not quite map perfectly onto the NTFS times,
637 	 * but they are close enough, and in the end it doesn't really matter
638 	 * that much...
639 	 */
640 	/*
641 	 * mtime is the last change of the data within the file. Not changed
642 	 * when only metadata is changed, e.g. a rename doesn't affect mtime.
643 	 */
644 	vi->i_mtime = ntfs2utc(si->last_data_change_time);
645 	/*
646 	 * ctime is the last change of the metadata of the file. This obviously
647 	 * always changes, when mtime is changed. ctime can be changed on its
648 	 * own, mtime is then not changed, e.g. when a file is renamed.
649 	 */
650 	vi->i_ctime = ntfs2utc(si->last_mft_change_time);
651 	/*
652 	 * Last access to the data within the file. Not changed during a rename
653 	 * for example but changed whenever the file is written to.
654 	 */
655 	vi->i_atime = ntfs2utc(si->last_access_time);
656 
657 	/* Find the attribute list attribute if present. */
658 	ntfs_attr_reinit_search_ctx(ctx);
659 	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
660 	if (err) {
661 		if (unlikely(err != -ENOENT)) {
662 			ntfs_error(vi->i_sb, "Failed to lookup attribute list "
663 					"attribute.");
664 			goto unm_err_out;
665 		}
666 	} else /* if (!err) */ {
667 		if (vi->i_ino == FILE_MFT)
668 			goto skip_attr_list_load;
669 		ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
670 		NInoSetAttrList(ni);
671 		a = ctx->attr;
672 		if (a->flags & ATTR_COMPRESSION_MASK) {
673 			ntfs_error(vi->i_sb, "Attribute list attribute is "
674 					"compressed.");
675 			goto unm_err_out;
676 		}
677 		if (a->flags & ATTR_IS_ENCRYPTED ||
678 				a->flags & ATTR_IS_SPARSE) {
679 			if (a->non_resident) {
680 				ntfs_error(vi->i_sb, "Non-resident attribute "
681 						"list attribute is encrypted/"
682 						"sparse.");
683 				goto unm_err_out;
684 			}
685 			ntfs_warning(vi->i_sb, "Resident attribute list "
686 					"attribute in inode 0x%lx is marked "
687 					"encrypted/sparse which is not true.  "
688 					"However, Windows allows this and "
689 					"chkdsk does not detect or correct it "
690 					"so we will just ignore the invalid "
691 					"flags and pretend they are not set.",
692 					vi->i_ino);
693 		}
694 		/* Now allocate memory for the attribute list. */
695 		ni->attr_list_size = (u32)ntfs_attr_size(a);
696 		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
697 		if (!ni->attr_list) {
698 			ntfs_error(vi->i_sb, "Not enough memory to allocate "
699 					"buffer for attribute list.");
700 			err = -ENOMEM;
701 			goto unm_err_out;
702 		}
703 		if (a->non_resident) {
704 			NInoSetAttrListNonResident(ni);
705 			if (a->data.non_resident.lowest_vcn) {
706 				ntfs_error(vi->i_sb, "Attribute list has non "
707 						"zero lowest_vcn.");
708 				goto unm_err_out;
709 			}
710 			/*
711 			 * Setup the runlist. No need for locking as we have
712 			 * exclusive access to the inode at this time.
713 			 */
714 			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
715 					a, NULL);
716 			if (IS_ERR(ni->attr_list_rl.rl)) {
717 				err = PTR_ERR(ni->attr_list_rl.rl);
718 				ni->attr_list_rl.rl = NULL;
719 				ntfs_error(vi->i_sb, "Mapping pairs "
720 						"decompression failed.");
721 				goto unm_err_out;
722 			}
723 			/* Now load the attribute list. */
724 			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
725 					ni->attr_list, ni->attr_list_size,
726 					sle64_to_cpu(a->data.non_resident.
727 					initialized_size)))) {
728 				ntfs_error(vi->i_sb, "Failed to load "
729 						"attribute list attribute.");
730 				goto unm_err_out;
731 			}
732 		} else /* if (!a->non_resident) */ {
733 			if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
734 					+ le32_to_cpu(
735 					a->data.resident.value_length) >
736 					(u8*)ctx->mrec + vol->mft_record_size) {
737 				ntfs_error(vi->i_sb, "Corrupt attribute list "
738 						"in inode.");
739 				goto unm_err_out;
740 			}
741 			/* Now copy the attribute list. */
742 			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
743 					a->data.resident.value_offset),
744 					le32_to_cpu(
745 					a->data.resident.value_length));
746 		}
747 	}
748 skip_attr_list_load:
749 	/*
750 	 * If an attribute list is present we now have the attribute list value
751 	 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
752 	 */
753 	if (S_ISDIR(vi->i_mode)) {
754 		loff_t bvi_size;
755 		ntfs_inode *bni;
756 		INDEX_ROOT *ir;
757 		u8 *ir_end, *index_end;
758 
759 		/* It is a directory, find index root attribute. */
760 		ntfs_attr_reinit_search_ctx(ctx);
761 		err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
762 				0, NULL, 0, ctx);
763 		if (unlikely(err)) {
764 			if (err == -ENOENT) {
765 				// FIXME: File is corrupt! Hot-fix with empty
766 				// index root attribute if recovery option is
767 				// set.
768 				ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
769 						"is missing.");
770 			}
771 			goto unm_err_out;
772 		}
773 		a = ctx->attr;
774 		/* Set up the state. */
775 		if (unlikely(a->non_resident)) {
776 			ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
777 					"resident.");
778 			goto unm_err_out;
779 		}
780 		/* Ensure the attribute name is placed before the value. */
781 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
782 				le16_to_cpu(a->data.resident.value_offset)))) {
783 			ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
784 					"placed after the attribute value.");
785 			goto unm_err_out;
786 		}
787 		/*
788 		 * Compressed/encrypted index root just means that the newly
789 		 * created files in that directory should be created compressed/
790 		 * encrypted. However index root cannot be both compressed and
791 		 * encrypted.
792 		 */
793 		if (a->flags & ATTR_COMPRESSION_MASK)
794 			NInoSetCompressed(ni);
795 		if (a->flags & ATTR_IS_ENCRYPTED) {
796 			if (a->flags & ATTR_COMPRESSION_MASK) {
797 				ntfs_error(vi->i_sb, "Found encrypted and "
798 						"compressed attribute.");
799 				goto unm_err_out;
800 			}
801 			NInoSetEncrypted(ni);
802 		}
803 		if (a->flags & ATTR_IS_SPARSE)
804 			NInoSetSparse(ni);
805 		ir = (INDEX_ROOT*)((u8*)a +
806 				le16_to_cpu(a->data.resident.value_offset));
807 		ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
808 		if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
809 			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
810 					"corrupt.");
811 			goto unm_err_out;
812 		}
813 		index_end = (u8*)&ir->index +
814 				le32_to_cpu(ir->index.index_length);
815 		if (index_end > ir_end) {
816 			ntfs_error(vi->i_sb, "Directory index is corrupt.");
817 			goto unm_err_out;
818 		}
819 		if (ir->type != AT_FILE_NAME) {
820 			ntfs_error(vi->i_sb, "Indexed attribute is not "
821 					"$FILE_NAME.");
822 			goto unm_err_out;
823 		}
824 		if (ir->collation_rule != COLLATION_FILE_NAME) {
825 			ntfs_error(vi->i_sb, "Index collation rule is not "
826 					"COLLATION_FILE_NAME.");
827 			goto unm_err_out;
828 		}
829 		ni->itype.index.collation_rule = ir->collation_rule;
830 		ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
831 		if (ni->itype.index.block_size &
832 				(ni->itype.index.block_size - 1)) {
833 			ntfs_error(vi->i_sb, "Index block size (%u) is not a "
834 					"power of two.",
835 					ni->itype.index.block_size);
836 			goto unm_err_out;
837 		}
838 		if (ni->itype.index.block_size > PAGE_SIZE) {
839 			ntfs_error(vi->i_sb, "Index block size (%u) > "
840 					"PAGE_SIZE (%ld) is not "
841 					"supported.  Sorry.",
842 					ni->itype.index.block_size,
843 					PAGE_SIZE);
844 			err = -EOPNOTSUPP;
845 			goto unm_err_out;
846 		}
847 		if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
848 			ntfs_error(vi->i_sb, "Index block size (%u) < "
849 					"NTFS_BLOCK_SIZE (%i) is not "
850 					"supported.  Sorry.",
851 					ni->itype.index.block_size,
852 					NTFS_BLOCK_SIZE);
853 			err = -EOPNOTSUPP;
854 			goto unm_err_out;
855 		}
856 		ni->itype.index.block_size_bits =
857 				ffs(ni->itype.index.block_size) - 1;
858 		/* Determine the size of a vcn in the directory index. */
859 		if (vol->cluster_size <= ni->itype.index.block_size) {
860 			ni->itype.index.vcn_size = vol->cluster_size;
861 			ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
862 		} else {
863 			ni->itype.index.vcn_size = vol->sector_size;
864 			ni->itype.index.vcn_size_bits = vol->sector_size_bits;
865 		}
866 
867 		/* Setup the index allocation attribute, even if not present. */
868 		NInoSetMstProtected(ni);
869 		ni->type = AT_INDEX_ALLOCATION;
870 		ni->name = I30;
871 		ni->name_len = 4;
872 
873 		if (!(ir->index.flags & LARGE_INDEX)) {
874 			/* No index allocation. */
875 			vi->i_size = ni->initialized_size =
876 					ni->allocated_size = 0;
877 			/* We are done with the mft record, so we release it. */
878 			ntfs_attr_put_search_ctx(ctx);
879 			unmap_mft_record(ni);
880 			m = NULL;
881 			ctx = NULL;
882 			goto skip_large_dir_stuff;
883 		} /* LARGE_INDEX: Index allocation present. Setup state. */
884 		NInoSetIndexAllocPresent(ni);
885 		/* Find index allocation attribute. */
886 		ntfs_attr_reinit_search_ctx(ctx);
887 		err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
888 				CASE_SENSITIVE, 0, NULL, 0, ctx);
889 		if (unlikely(err)) {
890 			if (err == -ENOENT)
891 				ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
892 						"attribute is not present but "
893 						"$INDEX_ROOT indicated it is.");
894 			else
895 				ntfs_error(vi->i_sb, "Failed to lookup "
896 						"$INDEX_ALLOCATION "
897 						"attribute.");
898 			goto unm_err_out;
899 		}
900 		a = ctx->attr;
901 		if (!a->non_resident) {
902 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
903 					"is resident.");
904 			goto unm_err_out;
905 		}
906 		/*
907 		 * Ensure the attribute name is placed before the mapping pairs
908 		 * array.
909 		 */
910 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
911 				le16_to_cpu(
912 				a->data.non_resident.mapping_pairs_offset)))) {
913 			ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
914 					"is placed after the mapping pairs "
915 					"array.");
916 			goto unm_err_out;
917 		}
918 		if (a->flags & ATTR_IS_ENCRYPTED) {
919 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
920 					"is encrypted.");
921 			goto unm_err_out;
922 		}
923 		if (a->flags & ATTR_IS_SPARSE) {
924 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
925 					"is sparse.");
926 			goto unm_err_out;
927 		}
928 		if (a->flags & ATTR_COMPRESSION_MASK) {
929 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
930 					"is compressed.");
931 			goto unm_err_out;
932 		}
933 		if (a->data.non_resident.lowest_vcn) {
934 			ntfs_error(vi->i_sb, "First extent of "
935 					"$INDEX_ALLOCATION attribute has non "
936 					"zero lowest_vcn.");
937 			goto unm_err_out;
938 		}
939 		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
940 		ni->initialized_size = sle64_to_cpu(
941 				a->data.non_resident.initialized_size);
942 		ni->allocated_size = sle64_to_cpu(
943 				a->data.non_resident.allocated_size);
944 		/*
945 		 * We are done with the mft record, so we release it. Otherwise
946 		 * we would deadlock in ntfs_attr_iget().
947 		 */
948 		ntfs_attr_put_search_ctx(ctx);
949 		unmap_mft_record(ni);
950 		m = NULL;
951 		ctx = NULL;
952 		/* Get the index bitmap attribute inode. */
953 		bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
954 		if (IS_ERR(bvi)) {
955 			ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
956 			err = PTR_ERR(bvi);
957 			goto unm_err_out;
958 		}
959 		bni = NTFS_I(bvi);
960 		if (NInoCompressed(bni) || NInoEncrypted(bni) ||
961 				NInoSparse(bni)) {
962 			ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
963 					"and/or encrypted and/or sparse.");
964 			goto iput_unm_err_out;
965 		}
966 		/* Consistency check bitmap size vs. index allocation size. */
967 		bvi_size = i_size_read(bvi);
968 		if ((bvi_size << 3) < (vi->i_size >>
969 				ni->itype.index.block_size_bits)) {
970 			ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
971 					"for index allocation (0x%llx).",
972 					bvi_size << 3, vi->i_size);
973 			goto iput_unm_err_out;
974 		}
975 		/* No longer need the bitmap attribute inode. */
976 		iput(bvi);
977 skip_large_dir_stuff:
978 		/* Setup the operations for this inode. */
979 		vi->i_op = &ntfs_dir_inode_ops;
980 		vi->i_fop = &ntfs_dir_ops;
981 		vi->i_mapping->a_ops = &ntfs_mst_aops;
982 	} else {
983 		/* It is a file. */
984 		ntfs_attr_reinit_search_ctx(ctx);
985 
986 		/* Setup the data attribute, even if not present. */
987 		ni->type = AT_DATA;
988 		ni->name = NULL;
989 		ni->name_len = 0;
990 
991 		/* Find first extent of the unnamed data attribute. */
992 		err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
993 		if (unlikely(err)) {
994 			vi->i_size = ni->initialized_size =
995 					ni->allocated_size = 0;
996 			if (err != -ENOENT) {
997 				ntfs_error(vi->i_sb, "Failed to lookup $DATA "
998 						"attribute.");
999 				goto unm_err_out;
1000 			}
1001 			/*
1002 			 * FILE_Secure does not have an unnamed $DATA
1003 			 * attribute, so we special case it here.
1004 			 */
1005 			if (vi->i_ino == FILE_Secure)
1006 				goto no_data_attr_special_case;
1007 			/*
1008 			 * Most if not all the system files in the $Extend
1009 			 * system directory do not have unnamed data
1010 			 * attributes so we need to check if the parent
1011 			 * directory of the file is FILE_Extend and if it is
1012 			 * ignore this error. To do this we need to get the
1013 			 * name of this inode from the mft record as the name
1014 			 * contains the back reference to the parent directory.
1015 			 */
1016 			if (ntfs_is_extended_system_file(ctx) > 0)
1017 				goto no_data_attr_special_case;
1018 			// FIXME: File is corrupt! Hot-fix with empty data
1019 			// attribute if recovery option is set.
1020 			ntfs_error(vi->i_sb, "$DATA attribute is missing.");
1021 			goto unm_err_out;
1022 		}
1023 		a = ctx->attr;
1024 		/* Setup the state. */
1025 		if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1026 			if (a->flags & ATTR_COMPRESSION_MASK) {
1027 				NInoSetCompressed(ni);
1028 				if (vol->cluster_size > 4096) {
1029 					ntfs_error(vi->i_sb, "Found "
1030 							"compressed data but "
1031 							"compression is "
1032 							"disabled due to "
1033 							"cluster size (%i) > "
1034 							"4kiB.",
1035 							vol->cluster_size);
1036 					goto unm_err_out;
1037 				}
1038 				if ((a->flags & ATTR_COMPRESSION_MASK)
1039 						!= ATTR_IS_COMPRESSED) {
1040 					ntfs_error(vi->i_sb, "Found unknown "
1041 							"compression method "
1042 							"or corrupt file.");
1043 					goto unm_err_out;
1044 				}
1045 			}
1046 			if (a->flags & ATTR_IS_SPARSE)
1047 				NInoSetSparse(ni);
1048 		}
1049 		if (a->flags & ATTR_IS_ENCRYPTED) {
1050 			if (NInoCompressed(ni)) {
1051 				ntfs_error(vi->i_sb, "Found encrypted and "
1052 						"compressed data.");
1053 				goto unm_err_out;
1054 			}
1055 			NInoSetEncrypted(ni);
1056 		}
1057 		if (a->non_resident) {
1058 			NInoSetNonResident(ni);
1059 			if (NInoCompressed(ni) || NInoSparse(ni)) {
1060 				if (NInoCompressed(ni) && a->data.non_resident.
1061 						compression_unit != 4) {
1062 					ntfs_error(vi->i_sb, "Found "
1063 							"non-standard "
1064 							"compression unit (%u "
1065 							"instead of 4).  "
1066 							"Cannot handle this.",
1067 							a->data.non_resident.
1068 							compression_unit);
1069 					err = -EOPNOTSUPP;
1070 					goto unm_err_out;
1071 				}
1072 				if (a->data.non_resident.compression_unit) {
1073 					ni->itype.compressed.block_size = 1U <<
1074 							(a->data.non_resident.
1075 							compression_unit +
1076 							vol->cluster_size_bits);
1077 					ni->itype.compressed.block_size_bits =
1078 							ffs(ni->itype.
1079 							compressed.
1080 							block_size) - 1;
1081 					ni->itype.compressed.block_clusters =
1082 							1U << a->data.
1083 							non_resident.
1084 							compression_unit;
1085 				} else {
1086 					ni->itype.compressed.block_size = 0;
1087 					ni->itype.compressed.block_size_bits =
1088 							0;
1089 					ni->itype.compressed.block_clusters =
1090 							0;
1091 				}
1092 				ni->itype.compressed.size = sle64_to_cpu(
1093 						a->data.non_resident.
1094 						compressed_size);
1095 			}
1096 			if (a->data.non_resident.lowest_vcn) {
1097 				ntfs_error(vi->i_sb, "First extent of $DATA "
1098 						"attribute has non zero "
1099 						"lowest_vcn.");
1100 				goto unm_err_out;
1101 			}
1102 			vi->i_size = sle64_to_cpu(
1103 					a->data.non_resident.data_size);
1104 			ni->initialized_size = sle64_to_cpu(
1105 					a->data.non_resident.initialized_size);
1106 			ni->allocated_size = sle64_to_cpu(
1107 					a->data.non_resident.allocated_size);
1108 		} else { /* Resident attribute. */
1109 			vi->i_size = ni->initialized_size = le32_to_cpu(
1110 					a->data.resident.value_length);
1111 			ni->allocated_size = le32_to_cpu(a->length) -
1112 					le16_to_cpu(
1113 					a->data.resident.value_offset);
1114 			if (vi->i_size > ni->allocated_size) {
1115 				ntfs_error(vi->i_sb, "Resident data attribute "
1116 						"is corrupt (size exceeds "
1117 						"allocation).");
1118 				goto unm_err_out;
1119 			}
1120 		}
1121 no_data_attr_special_case:
1122 		/* We are done with the mft record, so we release it. */
1123 		ntfs_attr_put_search_ctx(ctx);
1124 		unmap_mft_record(ni);
1125 		m = NULL;
1126 		ctx = NULL;
1127 		/* Setup the operations for this inode. */
1128 		vi->i_op = &ntfs_file_inode_ops;
1129 		vi->i_fop = &ntfs_file_ops;
1130 		vi->i_mapping->a_ops = &ntfs_normal_aops;
1131 		if (NInoMstProtected(ni))
1132 			vi->i_mapping->a_ops = &ntfs_mst_aops;
1133 		else if (NInoCompressed(ni))
1134 			vi->i_mapping->a_ops = &ntfs_compressed_aops;
1135 	}
1136 	/*
1137 	 * The number of 512-byte blocks used on disk (for stat). This is in so
1138 	 * far inaccurate as it doesn't account for any named streams or other
1139 	 * special non-resident attributes, but that is how Windows works, too,
1140 	 * so we are at least consistent with Windows, if not entirely
1141 	 * consistent with the Linux Way. Doing it the Linux Way would cause a
1142 	 * significant slowdown as it would involve iterating over all
1143 	 * attributes in the mft record and adding the allocated/compressed
1144 	 * sizes of all non-resident attributes present to give us the Linux
1145 	 * correct size that should go into i_blocks (after division by 512).
1146 	 */
1147 	if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
1148 		vi->i_blocks = ni->itype.compressed.size >> 9;
1149 	else
1150 		vi->i_blocks = ni->allocated_size >> 9;
1151 	ntfs_debug("Done.");
1152 	return 0;
1153 iput_unm_err_out:
1154 	iput(bvi);
1155 unm_err_out:
1156 	if (!err)
1157 		err = -EIO;
1158 	if (ctx)
1159 		ntfs_attr_put_search_ctx(ctx);
1160 	if (m)
1161 		unmap_mft_record(ni);
1162 err_out:
1163 	ntfs_error(vol->sb, "Failed with error code %i.  Marking corrupt "
1164 			"inode 0x%lx as bad.  Run chkdsk.", err, vi->i_ino);
1165 	make_bad_inode(vi);
1166 	if (err != -EOPNOTSUPP && err != -ENOMEM)
1167 		NVolSetErrors(vol);
1168 	return err;
1169 }
1170 
1171 /**
1172  * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
1173  * @base_vi:	base inode
1174  * @vi:		attribute inode to read
1175  *
1176  * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
1177  * attribute inode described by @vi into memory from the base mft record
1178  * described by @base_ni.
1179  *
1180  * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
1181  * reading and looks up the attribute described by @vi before setting up the
1182  * necessary fields in @vi as well as initializing the ntfs inode.
1183  *
1184  * Q: What locks are held when the function is called?
1185  * A: i_state has I_NEW set, hence the inode is locked, also
1186  *    i_count is set to 1, so it is not going to go away
1187  *
1188  * Return 0 on success and -errno on error.  In the error case, the inode will
1189  * have had make_bad_inode() executed on it.
1190  *
1191  * Note this cannot be called for AT_INDEX_ALLOCATION.
1192  */
1193 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
1194 {
1195 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
1196 	ntfs_inode *ni, *base_ni;
1197 	MFT_RECORD *m;
1198 	ATTR_RECORD *a;
1199 	ntfs_attr_search_ctx *ctx;
1200 	int err = 0;
1201 
1202 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1203 
1204 	ntfs_init_big_inode(vi);
1205 
1206 	ni	= NTFS_I(vi);
1207 	base_ni = NTFS_I(base_vi);
1208 
1209 	/* Just mirror the values from the base inode. */
1210 	vi->i_uid	= base_vi->i_uid;
1211 	vi->i_gid	= base_vi->i_gid;
1212 	set_nlink(vi, base_vi->i_nlink);
1213 	vi->i_mtime	= base_vi->i_mtime;
1214 	vi->i_ctime	= base_vi->i_ctime;
1215 	vi->i_atime	= base_vi->i_atime;
1216 	vi->i_generation = ni->seq_no = base_ni->seq_no;
1217 
1218 	/* Set inode type to zero but preserve permissions. */
1219 	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
1220 
1221 	m = map_mft_record(base_ni);
1222 	if (IS_ERR(m)) {
1223 		err = PTR_ERR(m);
1224 		goto err_out;
1225 	}
1226 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
1227 	if (!ctx) {
1228 		err = -ENOMEM;
1229 		goto unm_err_out;
1230 	}
1231 	/* Find the attribute. */
1232 	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1233 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1234 	if (unlikely(err))
1235 		goto unm_err_out;
1236 	a = ctx->attr;
1237 	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
1238 		if (a->flags & ATTR_COMPRESSION_MASK) {
1239 			NInoSetCompressed(ni);
1240 			if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
1241 					ni->name_len)) {
1242 				ntfs_error(vi->i_sb, "Found compressed "
1243 						"non-data or named data "
1244 						"attribute.  Please report "
1245 						"you saw this message to "
1246 						"linux-ntfs-dev@lists."
1247 						"sourceforge.net");
1248 				goto unm_err_out;
1249 			}
1250 			if (vol->cluster_size > 4096) {
1251 				ntfs_error(vi->i_sb, "Found compressed "
1252 						"attribute but compression is "
1253 						"disabled due to cluster size "
1254 						"(%i) > 4kiB.",
1255 						vol->cluster_size);
1256 				goto unm_err_out;
1257 			}
1258 			if ((a->flags & ATTR_COMPRESSION_MASK) !=
1259 					ATTR_IS_COMPRESSED) {
1260 				ntfs_error(vi->i_sb, "Found unknown "
1261 						"compression method.");
1262 				goto unm_err_out;
1263 			}
1264 		}
1265 		/*
1266 		 * The compressed/sparse flag set in an index root just means
1267 		 * to compress all files.
1268 		 */
1269 		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1270 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1271 					"but the attribute is %s.  Please "
1272 					"report you saw this message to "
1273 					"linux-ntfs-dev@lists.sourceforge.net",
1274 					NInoCompressed(ni) ? "compressed" :
1275 					"sparse");
1276 			goto unm_err_out;
1277 		}
1278 		if (a->flags & ATTR_IS_SPARSE)
1279 			NInoSetSparse(ni);
1280 	}
1281 	if (a->flags & ATTR_IS_ENCRYPTED) {
1282 		if (NInoCompressed(ni)) {
1283 			ntfs_error(vi->i_sb, "Found encrypted and compressed "
1284 					"data.");
1285 			goto unm_err_out;
1286 		}
1287 		/*
1288 		 * The encryption flag set in an index root just means to
1289 		 * encrypt all files.
1290 		 */
1291 		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
1292 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1293 					"but the attribute is encrypted.  "
1294 					"Please report you saw this message "
1295 					"to linux-ntfs-dev@lists.sourceforge."
1296 					"net");
1297 			goto unm_err_out;
1298 		}
1299 		if (ni->type != AT_DATA) {
1300 			ntfs_error(vi->i_sb, "Found encrypted non-data "
1301 					"attribute.");
1302 			goto unm_err_out;
1303 		}
1304 		NInoSetEncrypted(ni);
1305 	}
1306 	if (!a->non_resident) {
1307 		/* Ensure the attribute name is placed before the value. */
1308 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1309 				le16_to_cpu(a->data.resident.value_offset)))) {
1310 			ntfs_error(vol->sb, "Attribute name is placed after "
1311 					"the attribute value.");
1312 			goto unm_err_out;
1313 		}
1314 		if (NInoMstProtected(ni)) {
1315 			ntfs_error(vi->i_sb, "Found mst protected attribute "
1316 					"but the attribute is resident.  "
1317 					"Please report you saw this message to "
1318 					"linux-ntfs-dev@lists.sourceforge.net");
1319 			goto unm_err_out;
1320 		}
1321 		vi->i_size = ni->initialized_size = le32_to_cpu(
1322 				a->data.resident.value_length);
1323 		ni->allocated_size = le32_to_cpu(a->length) -
1324 				le16_to_cpu(a->data.resident.value_offset);
1325 		if (vi->i_size > ni->allocated_size) {
1326 			ntfs_error(vi->i_sb, "Resident attribute is corrupt "
1327 					"(size exceeds allocation).");
1328 			goto unm_err_out;
1329 		}
1330 	} else {
1331 		NInoSetNonResident(ni);
1332 		/*
1333 		 * Ensure the attribute name is placed before the mapping pairs
1334 		 * array.
1335 		 */
1336 		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1337 				le16_to_cpu(
1338 				a->data.non_resident.mapping_pairs_offset)))) {
1339 			ntfs_error(vol->sb, "Attribute name is placed after "
1340 					"the mapping pairs array.");
1341 			goto unm_err_out;
1342 		}
1343 		if (NInoCompressed(ni) || NInoSparse(ni)) {
1344 			if (NInoCompressed(ni) && a->data.non_resident.
1345 					compression_unit != 4) {
1346 				ntfs_error(vi->i_sb, "Found non-standard "
1347 						"compression unit (%u instead "
1348 						"of 4).  Cannot handle this.",
1349 						a->data.non_resident.
1350 						compression_unit);
1351 				err = -EOPNOTSUPP;
1352 				goto unm_err_out;
1353 			}
1354 			if (a->data.non_resident.compression_unit) {
1355 				ni->itype.compressed.block_size = 1U <<
1356 						(a->data.non_resident.
1357 						compression_unit +
1358 						vol->cluster_size_bits);
1359 				ni->itype.compressed.block_size_bits =
1360 						ffs(ni->itype.compressed.
1361 						block_size) - 1;
1362 				ni->itype.compressed.block_clusters = 1U <<
1363 						a->data.non_resident.
1364 						compression_unit;
1365 			} else {
1366 				ni->itype.compressed.block_size = 0;
1367 				ni->itype.compressed.block_size_bits = 0;
1368 				ni->itype.compressed.block_clusters = 0;
1369 			}
1370 			ni->itype.compressed.size = sle64_to_cpu(
1371 					a->data.non_resident.compressed_size);
1372 		}
1373 		if (a->data.non_resident.lowest_vcn) {
1374 			ntfs_error(vi->i_sb, "First extent of attribute has "
1375 					"non-zero lowest_vcn.");
1376 			goto unm_err_out;
1377 		}
1378 		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1379 		ni->initialized_size = sle64_to_cpu(
1380 				a->data.non_resident.initialized_size);
1381 		ni->allocated_size = sle64_to_cpu(
1382 				a->data.non_resident.allocated_size);
1383 	}
1384 	vi->i_mapping->a_ops = &ntfs_normal_aops;
1385 	if (NInoMstProtected(ni))
1386 		vi->i_mapping->a_ops = &ntfs_mst_aops;
1387 	else if (NInoCompressed(ni))
1388 		vi->i_mapping->a_ops = &ntfs_compressed_aops;
1389 	if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
1390 		vi->i_blocks = ni->itype.compressed.size >> 9;
1391 	else
1392 		vi->i_blocks = ni->allocated_size >> 9;
1393 	/*
1394 	 * Make sure the base inode does not go away and attach it to the
1395 	 * attribute inode.
1396 	 */
1397 	igrab(base_vi);
1398 	ni->ext.base_ntfs_ino = base_ni;
1399 	ni->nr_extents = -1;
1400 
1401 	ntfs_attr_put_search_ctx(ctx);
1402 	unmap_mft_record(base_ni);
1403 
1404 	ntfs_debug("Done.");
1405 	return 0;
1406 
1407 unm_err_out:
1408 	if (!err)
1409 		err = -EIO;
1410 	if (ctx)
1411 		ntfs_attr_put_search_ctx(ctx);
1412 	unmap_mft_record(base_ni);
1413 err_out:
1414 	ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
1415 			"inode (mft_no 0x%lx, type 0x%x, name_len %i).  "
1416 			"Marking corrupt inode and base inode 0x%lx as bad.  "
1417 			"Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
1418 			base_vi->i_ino);
1419 	make_bad_inode(vi);
1420 	if (err != -ENOMEM)
1421 		NVolSetErrors(vol);
1422 	return err;
1423 }
1424 
1425 /**
1426  * ntfs_read_locked_index_inode - read an index inode from its base inode
1427  * @base_vi:	base inode
1428  * @vi:		index inode to read
1429  *
1430  * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
1431  * index inode described by @vi into memory from the base mft record described
1432  * by @base_ni.
1433  *
1434  * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
1435  * reading and looks up the attributes relating to the index described by @vi
1436  * before setting up the necessary fields in @vi as well as initializing the
1437  * ntfs inode.
1438  *
1439  * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
1440  * with the attribute type set to AT_INDEX_ALLOCATION.  Apart from that, they
1441  * are setup like directory inodes since directories are a special case of
1442  * indices ao they need to be treated in much the same way.  Most importantly,
1443  * for small indices the index allocation attribute might not actually exist.
1444  * However, the index root attribute always exists but this does not need to
1445  * have an inode associated with it and this is why we define a new inode type
1446  * index.  Also, like for directories, we need to have an attribute inode for
1447  * the bitmap attribute corresponding to the index allocation attribute and we
1448  * can store this in the appropriate field of the inode, just like we do for
1449  * normal directory inodes.
1450  *
1451  * Q: What locks are held when the function is called?
1452  * A: i_state has I_NEW set, hence the inode is locked, also
1453  *    i_count is set to 1, so it is not going to go away
1454  *
1455  * Return 0 on success and -errno on error.  In the error case, the inode will
1456  * have had make_bad_inode() executed on it.
1457  */
1458 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
1459 {
1460 	loff_t bvi_size;
1461 	ntfs_volume *vol = NTFS_SB(vi->i_sb);
1462 	ntfs_inode *ni, *base_ni, *bni;
1463 	struct inode *bvi;
1464 	MFT_RECORD *m;
1465 	ATTR_RECORD *a;
1466 	ntfs_attr_search_ctx *ctx;
1467 	INDEX_ROOT *ir;
1468 	u8 *ir_end, *index_end;
1469 	int err = 0;
1470 
1471 	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
1472 	ntfs_init_big_inode(vi);
1473 	ni	= NTFS_I(vi);
1474 	base_ni = NTFS_I(base_vi);
1475 	/* Just mirror the values from the base inode. */
1476 	vi->i_uid	= base_vi->i_uid;
1477 	vi->i_gid	= base_vi->i_gid;
1478 	set_nlink(vi, base_vi->i_nlink);
1479 	vi->i_mtime	= base_vi->i_mtime;
1480 	vi->i_ctime	= base_vi->i_ctime;
1481 	vi->i_atime	= base_vi->i_atime;
1482 	vi->i_generation = ni->seq_no = base_ni->seq_no;
1483 	/* Set inode type to zero but preserve permissions. */
1484 	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
1485 	/* Map the mft record for the base inode. */
1486 	m = map_mft_record(base_ni);
1487 	if (IS_ERR(m)) {
1488 		err = PTR_ERR(m);
1489 		goto err_out;
1490 	}
1491 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
1492 	if (!ctx) {
1493 		err = -ENOMEM;
1494 		goto unm_err_out;
1495 	}
1496 	/* Find the index root attribute. */
1497 	err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
1498 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1499 	if (unlikely(err)) {
1500 		if (err == -ENOENT)
1501 			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
1502 					"missing.");
1503 		goto unm_err_out;
1504 	}
1505 	a = ctx->attr;
1506 	/* Set up the state. */
1507 	if (unlikely(a->non_resident)) {
1508 		ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
1509 		goto unm_err_out;
1510 	}
1511 	/* Ensure the attribute name is placed before the value. */
1512 	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1513 			le16_to_cpu(a->data.resident.value_offset)))) {
1514 		ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
1515 				"after the attribute value.");
1516 		goto unm_err_out;
1517 	}
1518 	/*
1519 	 * Compressed/encrypted/sparse index root is not allowed, except for
1520 	 * directories of course but those are not dealt with here.
1521 	 */
1522 	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
1523 			ATTR_IS_SPARSE)) {
1524 		ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
1525 				"root attribute.");
1526 		goto unm_err_out;
1527 	}
1528 	ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
1529 	ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
1530 	if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
1531 		ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
1532 		goto unm_err_out;
1533 	}
1534 	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
1535 	if (index_end > ir_end) {
1536 		ntfs_error(vi->i_sb, "Index is corrupt.");
1537 		goto unm_err_out;
1538 	}
1539 	if (ir->type) {
1540 		ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
1541 				le32_to_cpu(ir->type));
1542 		goto unm_err_out;
1543 	}
1544 	ni->itype.index.collation_rule = ir->collation_rule;
1545 	ntfs_debug("Index collation rule is 0x%x.",
1546 			le32_to_cpu(ir->collation_rule));
1547 	ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
1548 	if (!is_power_of_2(ni->itype.index.block_size)) {
1549 		ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
1550 				"two.", ni->itype.index.block_size);
1551 		goto unm_err_out;
1552 	}
1553 	if (ni->itype.index.block_size > PAGE_SIZE) {
1554 		ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_SIZE "
1555 				"(%ld) is not supported.  Sorry.",
1556 				ni->itype.index.block_size, PAGE_SIZE);
1557 		err = -EOPNOTSUPP;
1558 		goto unm_err_out;
1559 	}
1560 	if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
1561 		ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
1562 				"(%i) is not supported.  Sorry.",
1563 				ni->itype.index.block_size, NTFS_BLOCK_SIZE);
1564 		err = -EOPNOTSUPP;
1565 		goto unm_err_out;
1566 	}
1567 	ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
1568 	/* Determine the size of a vcn in the index. */
1569 	if (vol->cluster_size <= ni->itype.index.block_size) {
1570 		ni->itype.index.vcn_size = vol->cluster_size;
1571 		ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
1572 	} else {
1573 		ni->itype.index.vcn_size = vol->sector_size;
1574 		ni->itype.index.vcn_size_bits = vol->sector_size_bits;
1575 	}
1576 	/* Check for presence of index allocation attribute. */
1577 	if (!(ir->index.flags & LARGE_INDEX)) {
1578 		/* No index allocation. */
1579 		vi->i_size = ni->initialized_size = ni->allocated_size = 0;
1580 		/* We are done with the mft record, so we release it. */
1581 		ntfs_attr_put_search_ctx(ctx);
1582 		unmap_mft_record(base_ni);
1583 		m = NULL;
1584 		ctx = NULL;
1585 		goto skip_large_index_stuff;
1586 	} /* LARGE_INDEX:  Index allocation present.  Setup state. */
1587 	NInoSetIndexAllocPresent(ni);
1588 	/* Find index allocation attribute. */
1589 	ntfs_attr_reinit_search_ctx(ctx);
1590 	err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
1591 			CASE_SENSITIVE, 0, NULL, 0, ctx);
1592 	if (unlikely(err)) {
1593 		if (err == -ENOENT)
1594 			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1595 					"not present but $INDEX_ROOT "
1596 					"indicated it is.");
1597 		else
1598 			ntfs_error(vi->i_sb, "Failed to lookup "
1599 					"$INDEX_ALLOCATION attribute.");
1600 		goto unm_err_out;
1601 	}
1602 	a = ctx->attr;
1603 	if (!a->non_resident) {
1604 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1605 				"resident.");
1606 		goto unm_err_out;
1607 	}
1608 	/*
1609 	 * Ensure the attribute name is placed before the mapping pairs array.
1610 	 */
1611 	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
1612 			le16_to_cpu(
1613 			a->data.non_resident.mapping_pairs_offset)))) {
1614 		ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
1615 				"placed after the mapping pairs array.");
1616 		goto unm_err_out;
1617 	}
1618 	if (a->flags & ATTR_IS_ENCRYPTED) {
1619 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1620 				"encrypted.");
1621 		goto unm_err_out;
1622 	}
1623 	if (a->flags & ATTR_IS_SPARSE) {
1624 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
1625 		goto unm_err_out;
1626 	}
1627 	if (a->flags & ATTR_COMPRESSION_MASK) {
1628 		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
1629 				"compressed.");
1630 		goto unm_err_out;
1631 	}
1632 	if (a->data.non_resident.lowest_vcn) {
1633 		ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
1634 				"attribute has non zero lowest_vcn.");
1635 		goto unm_err_out;
1636 	}
1637 	vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
1638 	ni->initialized_size = sle64_to_cpu(
1639 			a->data.non_resident.initialized_size);
1640 	ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
1641 	/*
1642 	 * We are done with the mft record, so we release it.  Otherwise
1643 	 * we would deadlock in ntfs_attr_iget().
1644 	 */
1645 	ntfs_attr_put_search_ctx(ctx);
1646 	unmap_mft_record(base_ni);
1647 	m = NULL;
1648 	ctx = NULL;
1649 	/* Get the index bitmap attribute inode. */
1650 	bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
1651 	if (IS_ERR(bvi)) {
1652 		ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
1653 		err = PTR_ERR(bvi);
1654 		goto unm_err_out;
1655 	}
1656 	bni = NTFS_I(bvi);
1657 	if (NInoCompressed(bni) || NInoEncrypted(bni) ||
1658 			NInoSparse(bni)) {
1659 		ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
1660 				"encrypted and/or sparse.");
1661 		goto iput_unm_err_out;
1662 	}
1663 	/* Consistency check bitmap size vs. index allocation size. */
1664 	bvi_size = i_size_read(bvi);
1665 	if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
1666 		ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
1667 				"index allocation (0x%llx).", bvi_size << 3,
1668 				vi->i_size);
1669 		goto iput_unm_err_out;
1670 	}
1671 	iput(bvi);
1672 skip_large_index_stuff:
1673 	/* Setup the operations for this index inode. */
1674 	vi->i_mapping->a_ops = &ntfs_mst_aops;
1675 	vi->i_blocks = ni->allocated_size >> 9;
1676 	/*
1677 	 * Make sure the base inode doesn't go away and attach it to the
1678 	 * index inode.
1679 	 */
1680 	igrab(base_vi);
1681 	ni->ext.base_ntfs_ino = base_ni;
1682 	ni->nr_extents = -1;
1683 
1684 	ntfs_debug("Done.");
1685 	return 0;
1686 iput_unm_err_out:
1687 	iput(bvi);
1688 unm_err_out:
1689 	if (!err)
1690 		err = -EIO;
1691 	if (ctx)
1692 		ntfs_attr_put_search_ctx(ctx);
1693 	if (m)
1694 		unmap_mft_record(base_ni);
1695 err_out:
1696 	ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
1697 			"inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
1698 			ni->name_len);
1699 	make_bad_inode(vi);
1700 	if (err != -EOPNOTSUPP && err != -ENOMEM)
1701 		NVolSetErrors(vol);
1702 	return err;
1703 }
1704 
1705 /*
1706  * The MFT inode has special locking, so teach the lock validator
1707  * about this by splitting off the locking rules of the MFT from
1708  * the locking rules of other inodes. The MFT inode can never be
1709  * accessed from the VFS side (or even internally), only by the
1710  * map_mft functions.
1711  */
1712 static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
1713 
1714 /**
1715  * ntfs_read_inode_mount - special read_inode for mount time use only
1716  * @vi:		inode to read
1717  *
1718  * Read inode FILE_MFT at mount time, only called with super_block lock
1719  * held from within the read_super() code path.
1720  *
1721  * This function exists because when it is called the page cache for $MFT/$DATA
1722  * is not initialized and hence we cannot get at the contents of mft records
1723  * by calling map_mft_record*().
1724  *
1725  * Further it needs to cope with the circular references problem, i.e. cannot
1726  * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
1727  * we do not know where the other extent mft records are yet and again, because
1728  * we cannot call map_mft_record*() yet.  Obviously this applies only when an
1729  * attribute list is actually present in $MFT inode.
1730  *
1731  * We solve these problems by starting with the $DATA attribute before anything
1732  * else and iterating using ntfs_attr_lookup($DATA) over all extents.  As each
1733  * extent is found, we ntfs_mapping_pairs_decompress() including the implied
1734  * ntfs_runlists_merge().  Each step of the iteration necessarily provides
1735  * sufficient information for the next step to complete.
1736  *
1737  * This should work but there are two possible pit falls (see inline comments
1738  * below), but only time will tell if they are real pits or just smoke...
1739  */
1740 int ntfs_read_inode_mount(struct inode *vi)
1741 {
1742 	VCN next_vcn, last_vcn, highest_vcn;
1743 	s64 block;
1744 	struct super_block *sb = vi->i_sb;
1745 	ntfs_volume *vol = NTFS_SB(sb);
1746 	struct buffer_head *bh;
1747 	ntfs_inode *ni;
1748 	MFT_RECORD *m = NULL;
1749 	ATTR_RECORD *a;
1750 	ntfs_attr_search_ctx *ctx;
1751 	unsigned int i, nr_blocks;
1752 	int err;
1753 
1754 	ntfs_debug("Entering.");
1755 
1756 	/* Initialize the ntfs specific part of @vi. */
1757 	ntfs_init_big_inode(vi);
1758 
1759 	ni = NTFS_I(vi);
1760 
1761 	/* Setup the data attribute. It is special as it is mst protected. */
1762 	NInoSetNonResident(ni);
1763 	NInoSetMstProtected(ni);
1764 	NInoSetSparseDisabled(ni);
1765 	ni->type = AT_DATA;
1766 	ni->name = NULL;
1767 	ni->name_len = 0;
1768 	/*
1769 	 * This sets up our little cheat allowing us to reuse the async read io
1770 	 * completion handler for directories.
1771 	 */
1772 	ni->itype.index.block_size = vol->mft_record_size;
1773 	ni->itype.index.block_size_bits = vol->mft_record_size_bits;
1774 
1775 	/* Very important! Needed to be able to call map_mft_record*(). */
1776 	vol->mft_ino = vi;
1777 
1778 	/* Allocate enough memory to read the first mft record. */
1779 	if (vol->mft_record_size > 64 * 1024) {
1780 		ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
1781 				vol->mft_record_size);
1782 		goto err_out;
1783 	}
1784 	i = vol->mft_record_size;
1785 	if (i < sb->s_blocksize)
1786 		i = sb->s_blocksize;
1787 	m = (MFT_RECORD*)ntfs_malloc_nofs(i);
1788 	if (!m) {
1789 		ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
1790 		goto err_out;
1791 	}
1792 
1793 	/* Determine the first block of the $MFT/$DATA attribute. */
1794 	block = vol->mft_lcn << vol->cluster_size_bits >>
1795 			sb->s_blocksize_bits;
1796 	nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
1797 	if (!nr_blocks)
1798 		nr_blocks = 1;
1799 
1800 	/* Load $MFT/$DATA's first mft record. */
1801 	for (i = 0; i < nr_blocks; i++) {
1802 		bh = sb_bread(sb, block++);
1803 		if (!bh) {
1804 			ntfs_error(sb, "Device read failed.");
1805 			goto err_out;
1806 		}
1807 		memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
1808 				sb->s_blocksize);
1809 		brelse(bh);
1810 	}
1811 
1812 	/* Apply the mst fixups. */
1813 	if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
1814 		/* FIXME: Try to use the $MFTMirr now. */
1815 		ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
1816 		goto err_out;
1817 	}
1818 
1819 	/* Need this to sanity check attribute list references to $MFT. */
1820 	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
1821 
1822 	/* Provides readpage() for map_mft_record(). */
1823 	vi->i_mapping->a_ops = &ntfs_mst_aops;
1824 
1825 	ctx = ntfs_attr_get_search_ctx(ni, m);
1826 	if (!ctx) {
1827 		err = -ENOMEM;
1828 		goto err_out;
1829 	}
1830 
1831 	/* Find the attribute list attribute if present. */
1832 	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
1833 	if (err) {
1834 		if (unlikely(err != -ENOENT)) {
1835 			ntfs_error(sb, "Failed to lookup attribute list "
1836 					"attribute. You should run chkdsk.");
1837 			goto put_err_out;
1838 		}
1839 	} else /* if (!err) */ {
1840 		ATTR_LIST_ENTRY *al_entry, *next_al_entry;
1841 		u8 *al_end;
1842 		static const char *es = "  Not allowed.  $MFT is corrupt.  "
1843 				"You should run chkdsk.";
1844 
1845 		ntfs_debug("Attribute list attribute found in $MFT.");
1846 		NInoSetAttrList(ni);
1847 		a = ctx->attr;
1848 		if (a->flags & ATTR_COMPRESSION_MASK) {
1849 			ntfs_error(sb, "Attribute list attribute is "
1850 					"compressed.%s", es);
1851 			goto put_err_out;
1852 		}
1853 		if (a->flags & ATTR_IS_ENCRYPTED ||
1854 				a->flags & ATTR_IS_SPARSE) {
1855 			if (a->non_resident) {
1856 				ntfs_error(sb, "Non-resident attribute list "
1857 						"attribute is encrypted/"
1858 						"sparse.%s", es);
1859 				goto put_err_out;
1860 			}
1861 			ntfs_warning(sb, "Resident attribute list attribute "
1862 					"in $MFT system file is marked "
1863 					"encrypted/sparse which is not true.  "
1864 					"However, Windows allows this and "
1865 					"chkdsk does not detect or correct it "
1866 					"so we will just ignore the invalid "
1867 					"flags and pretend they are not set.");
1868 		}
1869 		/* Now allocate memory for the attribute list. */
1870 		ni->attr_list_size = (u32)ntfs_attr_size(a);
1871 		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
1872 		if (!ni->attr_list) {
1873 			ntfs_error(sb, "Not enough memory to allocate buffer "
1874 					"for attribute list.");
1875 			goto put_err_out;
1876 		}
1877 		if (a->non_resident) {
1878 			NInoSetAttrListNonResident(ni);
1879 			if (a->data.non_resident.lowest_vcn) {
1880 				ntfs_error(sb, "Attribute list has non zero "
1881 						"lowest_vcn. $MFT is corrupt. "
1882 						"You should run chkdsk.");
1883 				goto put_err_out;
1884 			}
1885 			/* Setup the runlist. */
1886 			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
1887 					a, NULL);
1888 			if (IS_ERR(ni->attr_list_rl.rl)) {
1889 				err = PTR_ERR(ni->attr_list_rl.rl);
1890 				ni->attr_list_rl.rl = NULL;
1891 				ntfs_error(sb, "Mapping pairs decompression "
1892 						"failed with error code %i.",
1893 						-err);
1894 				goto put_err_out;
1895 			}
1896 			/* Now load the attribute list. */
1897 			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
1898 					ni->attr_list, ni->attr_list_size,
1899 					sle64_to_cpu(a->data.
1900 					non_resident.initialized_size)))) {
1901 				ntfs_error(sb, "Failed to load attribute list "
1902 						"attribute with error code %i.",
1903 						-err);
1904 				goto put_err_out;
1905 			}
1906 		} else /* if (!ctx.attr->non_resident) */ {
1907 			if ((u8*)a + le16_to_cpu(
1908 					a->data.resident.value_offset) +
1909 					le32_to_cpu(
1910 					a->data.resident.value_length) >
1911 					(u8*)ctx->mrec + vol->mft_record_size) {
1912 				ntfs_error(sb, "Corrupt attribute list "
1913 						"attribute.");
1914 				goto put_err_out;
1915 			}
1916 			/* Now copy the attribute list. */
1917 			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
1918 					a->data.resident.value_offset),
1919 					le32_to_cpu(
1920 					a->data.resident.value_length));
1921 		}
1922 		/* The attribute list is now setup in memory. */
1923 		/*
1924 		 * FIXME: I don't know if this case is actually possible.
1925 		 * According to logic it is not possible but I have seen too
1926 		 * many weird things in MS software to rely on logic... Thus we
1927 		 * perform a manual search and make sure the first $MFT/$DATA
1928 		 * extent is in the base inode. If it is not we abort with an
1929 		 * error and if we ever see a report of this error we will need
1930 		 * to do some magic in order to have the necessary mft record
1931 		 * loaded and in the right place in the page cache. But
1932 		 * hopefully logic will prevail and this never happens...
1933 		 */
1934 		al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
1935 		al_end = (u8*)al_entry + ni->attr_list_size;
1936 		for (;; al_entry = next_al_entry) {
1937 			/* Out of bounds check. */
1938 			if ((u8*)al_entry < ni->attr_list ||
1939 					(u8*)al_entry > al_end)
1940 				goto em_put_err_out;
1941 			/* Catch the end of the attribute list. */
1942 			if ((u8*)al_entry == al_end)
1943 				goto em_put_err_out;
1944 			if (!al_entry->length)
1945 				goto em_put_err_out;
1946 			if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
1947 					le16_to_cpu(al_entry->length) > al_end)
1948 				goto em_put_err_out;
1949 			next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
1950 					le16_to_cpu(al_entry->length));
1951 			if (le32_to_cpu(al_entry->type) > le32_to_cpu(AT_DATA))
1952 				goto em_put_err_out;
1953 			if (AT_DATA != al_entry->type)
1954 				continue;
1955 			/* We want an unnamed attribute. */
1956 			if (al_entry->name_length)
1957 				goto em_put_err_out;
1958 			/* Want the first entry, i.e. lowest_vcn == 0. */
1959 			if (al_entry->lowest_vcn)
1960 				goto em_put_err_out;
1961 			/* First entry has to be in the base mft record. */
1962 			if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
1963 				/* MFT references do not match, logic fails. */
1964 				ntfs_error(sb, "BUG: The first $DATA extent "
1965 						"of $MFT is not in the base "
1966 						"mft record. Please report "
1967 						"you saw this message to "
1968 						"linux-ntfs-dev@lists."
1969 						"sourceforge.net");
1970 				goto put_err_out;
1971 			} else {
1972 				/* Sequence numbers must match. */
1973 				if (MSEQNO_LE(al_entry->mft_reference) !=
1974 						ni->seq_no)
1975 					goto em_put_err_out;
1976 				/* Got it. All is ok. We can stop now. */
1977 				break;
1978 			}
1979 		}
1980 	}
1981 
1982 	ntfs_attr_reinit_search_ctx(ctx);
1983 
1984 	/* Now load all attribute extents. */
1985 	a = NULL;
1986 	next_vcn = last_vcn = highest_vcn = 0;
1987 	while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
1988 			ctx))) {
1989 		runlist_element *nrl;
1990 
1991 		/* Cache the current attribute. */
1992 		a = ctx->attr;
1993 		/* $MFT must be non-resident. */
1994 		if (!a->non_resident) {
1995 			ntfs_error(sb, "$MFT must be non-resident but a "
1996 					"resident extent was found. $MFT is "
1997 					"corrupt. Run chkdsk.");
1998 			goto put_err_out;
1999 		}
2000 		/* $MFT must be uncompressed and unencrypted. */
2001 		if (a->flags & ATTR_COMPRESSION_MASK ||
2002 				a->flags & ATTR_IS_ENCRYPTED ||
2003 				a->flags & ATTR_IS_SPARSE) {
2004 			ntfs_error(sb, "$MFT must be uncompressed, "
2005 					"non-sparse, and unencrypted but a "
2006 					"compressed/sparse/encrypted extent "
2007 					"was found. $MFT is corrupt. Run "
2008 					"chkdsk.");
2009 			goto put_err_out;
2010 		}
2011 		/*
2012 		 * Decompress the mapping pairs array of this extent and merge
2013 		 * the result into the existing runlist. No need for locking
2014 		 * as we have exclusive access to the inode at this time and we
2015 		 * are a mount in progress task, too.
2016 		 */
2017 		nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2018 		if (IS_ERR(nrl)) {
2019 			ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
2020 					"failed with error code %ld.  $MFT is "
2021 					"corrupt.", PTR_ERR(nrl));
2022 			goto put_err_out;
2023 		}
2024 		ni->runlist.rl = nrl;
2025 
2026 		/* Are we in the first extent? */
2027 		if (!next_vcn) {
2028 			if (a->data.non_resident.lowest_vcn) {
2029 				ntfs_error(sb, "First extent of $DATA "
2030 						"attribute has non zero "
2031 						"lowest_vcn. $MFT is corrupt. "
2032 						"You should run chkdsk.");
2033 				goto put_err_out;
2034 			}
2035 			/* Get the last vcn in the $DATA attribute. */
2036 			last_vcn = sle64_to_cpu(
2037 					a->data.non_resident.allocated_size)
2038 					>> vol->cluster_size_bits;
2039 			/* Fill in the inode size. */
2040 			vi->i_size = sle64_to_cpu(
2041 					a->data.non_resident.data_size);
2042 			ni->initialized_size = sle64_to_cpu(
2043 					a->data.non_resident.initialized_size);
2044 			ni->allocated_size = sle64_to_cpu(
2045 					a->data.non_resident.allocated_size);
2046 			/*
2047 			 * Verify the number of mft records does not exceed
2048 			 * 2^32 - 1.
2049 			 */
2050 			if ((vi->i_size >> vol->mft_record_size_bits) >=
2051 					(1ULL << 32)) {
2052 				ntfs_error(sb, "$MFT is too big! Aborting.");
2053 				goto put_err_out;
2054 			}
2055 			/*
2056 			 * We have got the first extent of the runlist for
2057 			 * $MFT which means it is now relatively safe to call
2058 			 * the normal ntfs_read_inode() function.
2059 			 * Complete reading the inode, this will actually
2060 			 * re-read the mft record for $MFT, this time entering
2061 			 * it into the page cache with which we complete the
2062 			 * kick start of the volume. It should be safe to do
2063 			 * this now as the first extent of $MFT/$DATA is
2064 			 * already known and we would hope that we don't need
2065 			 * further extents in order to find the other
2066 			 * attributes belonging to $MFT. Only time will tell if
2067 			 * this is really the case. If not we will have to play
2068 			 * magic at this point, possibly duplicating a lot of
2069 			 * ntfs_read_inode() at this point. We will need to
2070 			 * ensure we do enough of its work to be able to call
2071 			 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
2072 			 * hope this never happens...
2073 			 */
2074 			ntfs_read_locked_inode(vi);
2075 			if (is_bad_inode(vi)) {
2076 				ntfs_error(sb, "ntfs_read_inode() of $MFT "
2077 						"failed. BUG or corrupt $MFT. "
2078 						"Run chkdsk and if no errors "
2079 						"are found, please report you "
2080 						"saw this message to "
2081 						"linux-ntfs-dev@lists."
2082 						"sourceforge.net");
2083 				ntfs_attr_put_search_ctx(ctx);
2084 				/* Revert to the safe super operations. */
2085 				ntfs_free(m);
2086 				return -1;
2087 			}
2088 			/*
2089 			 * Re-initialize some specifics about $MFT's inode as
2090 			 * ntfs_read_inode() will have set up the default ones.
2091 			 */
2092 			/* Set uid and gid to root. */
2093 			vi->i_uid = GLOBAL_ROOT_UID;
2094 			vi->i_gid = GLOBAL_ROOT_GID;
2095 			/* Regular file. No access for anyone. */
2096 			vi->i_mode = S_IFREG;
2097 			/* No VFS initiated operations allowed for $MFT. */
2098 			vi->i_op = &ntfs_empty_inode_ops;
2099 			vi->i_fop = &ntfs_empty_file_ops;
2100 		}
2101 
2102 		/* Get the lowest vcn for the next extent. */
2103 		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2104 		next_vcn = highest_vcn + 1;
2105 
2106 		/* Only one extent or error, which we catch below. */
2107 		if (next_vcn <= 0)
2108 			break;
2109 
2110 		/* Avoid endless loops due to corruption. */
2111 		if (next_vcn < sle64_to_cpu(
2112 				a->data.non_resident.lowest_vcn)) {
2113 			ntfs_error(sb, "$MFT has corrupt attribute list "
2114 					"attribute. Run chkdsk.");
2115 			goto put_err_out;
2116 		}
2117 	}
2118 	if (err != -ENOENT) {
2119 		ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
2120 				"$MFT is corrupt. Run chkdsk.");
2121 		goto put_err_out;
2122 	}
2123 	if (!a) {
2124 		ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
2125 				"corrupt. Run chkdsk.");
2126 		goto put_err_out;
2127 	}
2128 	if (highest_vcn && highest_vcn != last_vcn - 1) {
2129 		ntfs_error(sb, "Failed to load the complete runlist for "
2130 				"$MFT/$DATA. Driver bug or corrupt $MFT. "
2131 				"Run chkdsk.");
2132 		ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
2133 				(unsigned long long)highest_vcn,
2134 				(unsigned long long)last_vcn - 1);
2135 		goto put_err_out;
2136 	}
2137 	ntfs_attr_put_search_ctx(ctx);
2138 	ntfs_debug("Done.");
2139 	ntfs_free(m);
2140 
2141 	/*
2142 	 * Split the locking rules of the MFT inode from the
2143 	 * locking rules of other inodes:
2144 	 */
2145 	lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
2146 	lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
2147 
2148 	return 0;
2149 
2150 em_put_err_out:
2151 	ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
2152 			"attribute list. $MFT is corrupt. Run chkdsk.");
2153 put_err_out:
2154 	ntfs_attr_put_search_ctx(ctx);
2155 err_out:
2156 	ntfs_error(sb, "Failed. Marking inode as bad.");
2157 	make_bad_inode(vi);
2158 	ntfs_free(m);
2159 	return -1;
2160 }
2161 
2162 static void __ntfs_clear_inode(ntfs_inode *ni)
2163 {
2164 	/* Free all alocated memory. */
2165 	down_write(&ni->runlist.lock);
2166 	if (ni->runlist.rl) {
2167 		ntfs_free(ni->runlist.rl);
2168 		ni->runlist.rl = NULL;
2169 	}
2170 	up_write(&ni->runlist.lock);
2171 
2172 	if (ni->attr_list) {
2173 		ntfs_free(ni->attr_list);
2174 		ni->attr_list = NULL;
2175 	}
2176 
2177 	down_write(&ni->attr_list_rl.lock);
2178 	if (ni->attr_list_rl.rl) {
2179 		ntfs_free(ni->attr_list_rl.rl);
2180 		ni->attr_list_rl.rl = NULL;
2181 	}
2182 	up_write(&ni->attr_list_rl.lock);
2183 
2184 	if (ni->name_len && ni->name != I30) {
2185 		/* Catch bugs... */
2186 		BUG_ON(!ni->name);
2187 		kfree(ni->name);
2188 	}
2189 }
2190 
2191 void ntfs_clear_extent_inode(ntfs_inode *ni)
2192 {
2193 	ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
2194 
2195 	BUG_ON(NInoAttr(ni));
2196 	BUG_ON(ni->nr_extents != -1);
2197 
2198 #ifdef NTFS_RW
2199 	if (NInoDirty(ni)) {
2200 		if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
2201 			ntfs_error(ni->vol->sb, "Clearing dirty extent inode!  "
2202 					"Losing data!  This is a BUG!!!");
2203 		// FIXME:  Do something!!!
2204 	}
2205 #endif /* NTFS_RW */
2206 
2207 	__ntfs_clear_inode(ni);
2208 
2209 	/* Bye, bye... */
2210 	ntfs_destroy_extent_inode(ni);
2211 }
2212 
2213 /**
2214  * ntfs_evict_big_inode - clean up the ntfs specific part of an inode
2215  * @vi:		vfs inode pending annihilation
2216  *
2217  * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
2218  * is called, which deallocates all memory belonging to the NTFS specific part
2219  * of the inode and returns.
2220  *
2221  * If the MFT record is dirty, we commit it before doing anything else.
2222  */
2223 void ntfs_evict_big_inode(struct inode *vi)
2224 {
2225 	ntfs_inode *ni = NTFS_I(vi);
2226 
2227 	truncate_inode_pages_final(&vi->i_data);
2228 	clear_inode(vi);
2229 
2230 #ifdef NTFS_RW
2231 	if (NInoDirty(ni)) {
2232 		bool was_bad = (is_bad_inode(vi));
2233 
2234 		/* Committing the inode also commits all extent inodes. */
2235 		ntfs_commit_inode(vi);
2236 
2237 		if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
2238 			ntfs_error(vi->i_sb, "Failed to commit dirty inode "
2239 					"0x%lx.  Losing data!", vi->i_ino);
2240 			// FIXME:  Do something!!!
2241 		}
2242 	}
2243 #endif /* NTFS_RW */
2244 
2245 	/* No need to lock at this stage as no one else has a reference. */
2246 	if (ni->nr_extents > 0) {
2247 		int i;
2248 
2249 		for (i = 0; i < ni->nr_extents; i++)
2250 			ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
2251 		kfree(ni->ext.extent_ntfs_inos);
2252 	}
2253 
2254 	__ntfs_clear_inode(ni);
2255 
2256 	if (NInoAttr(ni)) {
2257 		/* Release the base inode if we are holding it. */
2258 		if (ni->nr_extents == -1) {
2259 			iput(VFS_I(ni->ext.base_ntfs_ino));
2260 			ni->nr_extents = 0;
2261 			ni->ext.base_ntfs_ino = NULL;
2262 		}
2263 	}
2264 	BUG_ON(ni->page);
2265 	if (!atomic_dec_and_test(&ni->count))
2266 		BUG();
2267 	return;
2268 }
2269 
2270 /**
2271  * ntfs_show_options - show mount options in /proc/mounts
2272  * @sf:		seq_file in which to write our mount options
2273  * @root:	root of the mounted tree whose mount options to display
2274  *
2275  * Called by the VFS once for each mounted ntfs volume when someone reads
2276  * /proc/mounts in order to display the NTFS specific mount options of each
2277  * mount. The mount options of fs specified by @root are written to the seq file
2278  * @sf and success is returned.
2279  */
2280 int ntfs_show_options(struct seq_file *sf, struct dentry *root)
2281 {
2282 	ntfs_volume *vol = NTFS_SB(root->d_sb);
2283 	int i;
2284 
2285 	seq_printf(sf, ",uid=%i", from_kuid_munged(&init_user_ns, vol->uid));
2286 	seq_printf(sf, ",gid=%i", from_kgid_munged(&init_user_ns, vol->gid));
2287 	if (vol->fmask == vol->dmask)
2288 		seq_printf(sf, ",umask=0%o", vol->fmask);
2289 	else {
2290 		seq_printf(sf, ",fmask=0%o", vol->fmask);
2291 		seq_printf(sf, ",dmask=0%o", vol->dmask);
2292 	}
2293 	seq_printf(sf, ",nls=%s", vol->nls_map->charset);
2294 	if (NVolCaseSensitive(vol))
2295 		seq_printf(sf, ",case_sensitive");
2296 	if (NVolShowSystemFiles(vol))
2297 		seq_printf(sf, ",show_sys_files");
2298 	if (!NVolSparseEnabled(vol))
2299 		seq_printf(sf, ",disable_sparse");
2300 	for (i = 0; on_errors_arr[i].val; i++) {
2301 		if (on_errors_arr[i].val & vol->on_errors)
2302 			seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
2303 	}
2304 	seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
2305 	return 0;
2306 }
2307 
2308 #ifdef NTFS_RW
2309 
2310 static const char *es = "  Leaving inconsistent metadata.  Unmount and run "
2311 		"chkdsk.";
2312 
2313 /**
2314  * ntfs_truncate - called when the i_size of an ntfs inode is changed
2315  * @vi:		inode for which the i_size was changed
2316  *
2317  * We only support i_size changes for normal files at present, i.e. not
2318  * compressed and not encrypted.  This is enforced in ntfs_setattr(), see
2319  * below.
2320  *
2321  * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
2322  * that the change is allowed.
2323  *
2324  * This implies for us that @vi is a file inode rather than a directory, index,
2325  * or attribute inode as well as that @vi is a base inode.
2326  *
2327  * Returns 0 on success or -errno on error.
2328  *
2329  * Called with ->i_mutex held.
2330  */
2331 int ntfs_truncate(struct inode *vi)
2332 {
2333 	s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
2334 	VCN highest_vcn;
2335 	unsigned long flags;
2336 	ntfs_inode *base_ni, *ni = NTFS_I(vi);
2337 	ntfs_volume *vol = ni->vol;
2338 	ntfs_attr_search_ctx *ctx;
2339 	MFT_RECORD *m;
2340 	ATTR_RECORD *a;
2341 	const char *te = "  Leaving file length out of sync with i_size.";
2342 	int err, mp_size, size_change, alloc_change;
2343 	u32 attr_len;
2344 
2345 	ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
2346 	BUG_ON(NInoAttr(ni));
2347 	BUG_ON(S_ISDIR(vi->i_mode));
2348 	BUG_ON(NInoMstProtected(ni));
2349 	BUG_ON(ni->nr_extents < 0);
2350 retry_truncate:
2351 	/*
2352 	 * Lock the runlist for writing and map the mft record to ensure it is
2353 	 * safe to mess with the attribute runlist and sizes.
2354 	 */
2355 	down_write(&ni->runlist.lock);
2356 	if (!NInoAttr(ni))
2357 		base_ni = ni;
2358 	else
2359 		base_ni = ni->ext.base_ntfs_ino;
2360 	m = map_mft_record(base_ni);
2361 	if (IS_ERR(m)) {
2362 		err = PTR_ERR(m);
2363 		ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
2364 				"(error code %d).%s", vi->i_ino, err, te);
2365 		ctx = NULL;
2366 		m = NULL;
2367 		goto old_bad_out;
2368 	}
2369 	ctx = ntfs_attr_get_search_ctx(base_ni, m);
2370 	if (unlikely(!ctx)) {
2371 		ntfs_error(vi->i_sb, "Failed to allocate a search context for "
2372 				"inode 0x%lx (not enough memory).%s",
2373 				vi->i_ino, te);
2374 		err = -ENOMEM;
2375 		goto old_bad_out;
2376 	}
2377 	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2378 			CASE_SENSITIVE, 0, NULL, 0, ctx);
2379 	if (unlikely(err)) {
2380 		if (err == -ENOENT) {
2381 			ntfs_error(vi->i_sb, "Open attribute is missing from "
2382 					"mft record.  Inode 0x%lx is corrupt.  "
2383 					"Run chkdsk.%s", vi->i_ino, te);
2384 			err = -EIO;
2385 		} else
2386 			ntfs_error(vi->i_sb, "Failed to lookup attribute in "
2387 					"inode 0x%lx (error code %d).%s",
2388 					vi->i_ino, err, te);
2389 		goto old_bad_out;
2390 	}
2391 	m = ctx->mrec;
2392 	a = ctx->attr;
2393 	/*
2394 	 * The i_size of the vfs inode is the new size for the attribute value.
2395 	 */
2396 	new_size = i_size_read(vi);
2397 	/* The current size of the attribute value is the old size. */
2398 	old_size = ntfs_attr_size(a);
2399 	/* Calculate the new allocated size. */
2400 	if (NInoNonResident(ni))
2401 		new_alloc_size = (new_size + vol->cluster_size - 1) &
2402 				~(s64)vol->cluster_size_mask;
2403 	else
2404 		new_alloc_size = (new_size + 7) & ~7;
2405 	/* The current allocated size is the old allocated size. */
2406 	read_lock_irqsave(&ni->size_lock, flags);
2407 	old_alloc_size = ni->allocated_size;
2408 	read_unlock_irqrestore(&ni->size_lock, flags);
2409 	/*
2410 	 * The change in the file size.  This will be 0 if no change, >0 if the
2411 	 * size is growing, and <0 if the size is shrinking.
2412 	 */
2413 	size_change = -1;
2414 	if (new_size - old_size >= 0) {
2415 		size_change = 1;
2416 		if (new_size == old_size)
2417 			size_change = 0;
2418 	}
2419 	/* As above for the allocated size. */
2420 	alloc_change = -1;
2421 	if (new_alloc_size - old_alloc_size >= 0) {
2422 		alloc_change = 1;
2423 		if (new_alloc_size == old_alloc_size)
2424 			alloc_change = 0;
2425 	}
2426 	/*
2427 	 * If neither the size nor the allocation are being changed there is
2428 	 * nothing to do.
2429 	 */
2430 	if (!size_change && !alloc_change)
2431 		goto unm_done;
2432 	/* If the size is changing, check if new size is allowed in $AttrDef. */
2433 	if (size_change) {
2434 		err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
2435 		if (unlikely(err)) {
2436 			if (err == -ERANGE) {
2437 				ntfs_error(vol->sb, "Truncate would cause the "
2438 						"inode 0x%lx to %simum size "
2439 						"for its attribute type "
2440 						"(0x%x).  Aborting truncate.",
2441 						vi->i_ino,
2442 						new_size > old_size ? "exceed "
2443 						"the max" : "go under the min",
2444 						le32_to_cpu(ni->type));
2445 				err = -EFBIG;
2446 			} else {
2447 				ntfs_error(vol->sb, "Inode 0x%lx has unknown "
2448 						"attribute type 0x%x.  "
2449 						"Aborting truncate.",
2450 						vi->i_ino,
2451 						le32_to_cpu(ni->type));
2452 				err = -EIO;
2453 			}
2454 			/* Reset the vfs inode size to the old size. */
2455 			i_size_write(vi, old_size);
2456 			goto err_out;
2457 		}
2458 	}
2459 	if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2460 		ntfs_warning(vi->i_sb, "Changes in inode size are not "
2461 				"supported yet for %s files, ignoring.",
2462 				NInoCompressed(ni) ? "compressed" :
2463 				"encrypted");
2464 		err = -EOPNOTSUPP;
2465 		goto bad_out;
2466 	}
2467 	if (a->non_resident)
2468 		goto do_non_resident_truncate;
2469 	BUG_ON(NInoNonResident(ni));
2470 	/* Resize the attribute record to best fit the new attribute size. */
2471 	if (new_size < vol->mft_record_size &&
2472 			!ntfs_resident_attr_value_resize(m, a, new_size)) {
2473 		/* The resize succeeded! */
2474 		flush_dcache_mft_record_page(ctx->ntfs_ino);
2475 		mark_mft_record_dirty(ctx->ntfs_ino);
2476 		write_lock_irqsave(&ni->size_lock, flags);
2477 		/* Update the sizes in the ntfs inode and all is done. */
2478 		ni->allocated_size = le32_to_cpu(a->length) -
2479 				le16_to_cpu(a->data.resident.value_offset);
2480 		/*
2481 		 * Note ntfs_resident_attr_value_resize() has already done any
2482 		 * necessary data clearing in the attribute record.  When the
2483 		 * file is being shrunk vmtruncate() will already have cleared
2484 		 * the top part of the last partial page, i.e. since this is
2485 		 * the resident case this is the page with index 0.  However,
2486 		 * when the file is being expanded, the page cache page data
2487 		 * between the old data_size, i.e. old_size, and the new_size
2488 		 * has not been zeroed.  Fortunately, we do not need to zero it
2489 		 * either since on one hand it will either already be zero due
2490 		 * to both readpage and writepage clearing partial page data
2491 		 * beyond i_size in which case there is nothing to do or in the
2492 		 * case of the file being mmap()ped at the same time, POSIX
2493 		 * specifies that the behaviour is unspecified thus we do not
2494 		 * have to do anything.  This means that in our implementation
2495 		 * in the rare case that the file is mmap()ped and a write
2496 		 * occurred into the mmap()ped region just beyond the file size
2497 		 * and writepage has not yet been called to write out the page
2498 		 * (which would clear the area beyond the file size) and we now
2499 		 * extend the file size to incorporate this dirty region
2500 		 * outside the file size, a write of the page would result in
2501 		 * this data being written to disk instead of being cleared.
2502 		 * Given both POSIX and the Linux mmap(2) man page specify that
2503 		 * this corner case is undefined, we choose to leave it like
2504 		 * that as this is much simpler for us as we cannot lock the
2505 		 * relevant page now since we are holding too many ntfs locks
2506 		 * which would result in a lock reversal deadlock.
2507 		 */
2508 		ni->initialized_size = new_size;
2509 		write_unlock_irqrestore(&ni->size_lock, flags);
2510 		goto unm_done;
2511 	}
2512 	/* If the above resize failed, this must be an attribute extension. */
2513 	BUG_ON(size_change < 0);
2514 	/*
2515 	 * We have to drop all the locks so we can call
2516 	 * ntfs_attr_make_non_resident().  This could be optimised by try-
2517 	 * locking the first page cache page and only if that fails dropping
2518 	 * the locks, locking the page, and redoing all the locking and
2519 	 * lookups.  While this would be a huge optimisation, it is not worth
2520 	 * it as this is definitely a slow code path as it only ever can happen
2521 	 * once for any given file.
2522 	 */
2523 	ntfs_attr_put_search_ctx(ctx);
2524 	unmap_mft_record(base_ni);
2525 	up_write(&ni->runlist.lock);
2526 	/*
2527 	 * Not enough space in the mft record, try to make the attribute
2528 	 * non-resident and if successful restart the truncation process.
2529 	 */
2530 	err = ntfs_attr_make_non_resident(ni, old_size);
2531 	if (likely(!err))
2532 		goto retry_truncate;
2533 	/*
2534 	 * Could not make non-resident.  If this is due to this not being
2535 	 * permitted for this attribute type or there not being enough space,
2536 	 * try to make other attributes non-resident.  Otherwise fail.
2537 	 */
2538 	if (unlikely(err != -EPERM && err != -ENOSPC)) {
2539 		ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
2540 				"type 0x%x, because the conversion from "
2541 				"resident to non-resident attribute failed "
2542 				"with error code %i.", vi->i_ino,
2543 				(unsigned)le32_to_cpu(ni->type), err);
2544 		if (err != -ENOMEM)
2545 			err = -EIO;
2546 		goto conv_err_out;
2547 	}
2548 	/* TODO: Not implemented from here, abort. */
2549 	if (err == -ENOSPC)
2550 		ntfs_error(vol->sb, "Not enough space in the mft record/on "
2551 				"disk for the non-resident attribute value.  "
2552 				"This case is not implemented yet.");
2553 	else /* if (err == -EPERM) */
2554 		ntfs_error(vol->sb, "This attribute type may not be "
2555 				"non-resident.  This case is not implemented "
2556 				"yet.");
2557 	err = -EOPNOTSUPP;
2558 	goto conv_err_out;
2559 #if 0
2560 	// TODO: Attempt to make other attributes non-resident.
2561 	if (!err)
2562 		goto do_resident_extend;
2563 	/*
2564 	 * Both the attribute list attribute and the standard information
2565 	 * attribute must remain in the base inode.  Thus, if this is one of
2566 	 * these attributes, we have to try to move other attributes out into
2567 	 * extent mft records instead.
2568 	 */
2569 	if (ni->type == AT_ATTRIBUTE_LIST ||
2570 			ni->type == AT_STANDARD_INFORMATION) {
2571 		// TODO: Attempt to move other attributes into extent mft
2572 		// records.
2573 		err = -EOPNOTSUPP;
2574 		if (!err)
2575 			goto do_resident_extend;
2576 		goto err_out;
2577 	}
2578 	// TODO: Attempt to move this attribute to an extent mft record, but
2579 	// only if it is not already the only attribute in an mft record in
2580 	// which case there would be nothing to gain.
2581 	err = -EOPNOTSUPP;
2582 	if (!err)
2583 		goto do_resident_extend;
2584 	/* There is nothing we can do to make enough space. )-: */
2585 	goto err_out;
2586 #endif
2587 do_non_resident_truncate:
2588 	BUG_ON(!NInoNonResident(ni));
2589 	if (alloc_change < 0) {
2590 		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
2591 		if (highest_vcn > 0 &&
2592 				old_alloc_size >> vol->cluster_size_bits >
2593 				highest_vcn + 1) {
2594 			/*
2595 			 * This attribute has multiple extents.  Not yet
2596 			 * supported.
2597 			 */
2598 			ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
2599 					"attribute type 0x%x, because the "
2600 					"attribute is highly fragmented (it "
2601 					"consists of multiple extents) and "
2602 					"this case is not implemented yet.",
2603 					vi->i_ino,
2604 					(unsigned)le32_to_cpu(ni->type));
2605 			err = -EOPNOTSUPP;
2606 			goto bad_out;
2607 		}
2608 	}
2609 	/*
2610 	 * If the size is shrinking, need to reduce the initialized_size and
2611 	 * the data_size before reducing the allocation.
2612 	 */
2613 	if (size_change < 0) {
2614 		/*
2615 		 * Make the valid size smaller (i_size is already up-to-date).
2616 		 */
2617 		write_lock_irqsave(&ni->size_lock, flags);
2618 		if (new_size < ni->initialized_size) {
2619 			ni->initialized_size = new_size;
2620 			a->data.non_resident.initialized_size =
2621 					cpu_to_sle64(new_size);
2622 		}
2623 		a->data.non_resident.data_size = cpu_to_sle64(new_size);
2624 		write_unlock_irqrestore(&ni->size_lock, flags);
2625 		flush_dcache_mft_record_page(ctx->ntfs_ino);
2626 		mark_mft_record_dirty(ctx->ntfs_ino);
2627 		/* If the allocated size is not changing, we are done. */
2628 		if (!alloc_change)
2629 			goto unm_done;
2630 		/*
2631 		 * If the size is shrinking it makes no sense for the
2632 		 * allocation to be growing.
2633 		 */
2634 		BUG_ON(alloc_change > 0);
2635 	} else /* if (size_change >= 0) */ {
2636 		/*
2637 		 * The file size is growing or staying the same but the
2638 		 * allocation can be shrinking, growing or staying the same.
2639 		 */
2640 		if (alloc_change > 0) {
2641 			/*
2642 			 * We need to extend the allocation and possibly update
2643 			 * the data size.  If we are updating the data size,
2644 			 * since we are not touching the initialized_size we do
2645 			 * not need to worry about the actual data on disk.
2646 			 * And as far as the page cache is concerned, there
2647 			 * will be no pages beyond the old data size and any
2648 			 * partial region in the last page between the old and
2649 			 * new data size (or the end of the page if the new
2650 			 * data size is outside the page) does not need to be
2651 			 * modified as explained above for the resident
2652 			 * attribute truncate case.  To do this, we simply drop
2653 			 * the locks we hold and leave all the work to our
2654 			 * friendly helper ntfs_attr_extend_allocation().
2655 			 */
2656 			ntfs_attr_put_search_ctx(ctx);
2657 			unmap_mft_record(base_ni);
2658 			up_write(&ni->runlist.lock);
2659 			err = ntfs_attr_extend_allocation(ni, new_size,
2660 					size_change > 0 ? new_size : -1, -1);
2661 			/*
2662 			 * ntfs_attr_extend_allocation() will have done error
2663 			 * output already.
2664 			 */
2665 			goto done;
2666 		}
2667 		if (!alloc_change)
2668 			goto alloc_done;
2669 	}
2670 	/* alloc_change < 0 */
2671 	/* Free the clusters. */
2672 	nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
2673 			vol->cluster_size_bits, -1, ctx);
2674 	m = ctx->mrec;
2675 	a = ctx->attr;
2676 	if (unlikely(nr_freed < 0)) {
2677 		ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
2678 				"%lli).  Unmount and run chkdsk to recover "
2679 				"the lost cluster(s).", (long long)nr_freed);
2680 		NVolSetErrors(vol);
2681 		nr_freed = 0;
2682 	}
2683 	/* Truncate the runlist. */
2684 	err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
2685 			new_alloc_size >> vol->cluster_size_bits);
2686 	/*
2687 	 * If the runlist truncation failed and/or the search context is no
2688 	 * longer valid, we cannot resize the attribute record or build the
2689 	 * mapping pairs array thus we mark the inode bad so that no access to
2690 	 * the freed clusters can happen.
2691 	 */
2692 	if (unlikely(err || IS_ERR(m))) {
2693 		ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
2694 				IS_ERR(m) ?
2695 				"restore attribute search context" :
2696 				"truncate attribute runlist",
2697 				IS_ERR(m) ? PTR_ERR(m) : err, es);
2698 		err = -EIO;
2699 		goto bad_out;
2700 	}
2701 	/* Get the size for the shrunk mapping pairs array for the runlist. */
2702 	mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
2703 	if (unlikely(mp_size <= 0)) {
2704 		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2705 				"attribute type 0x%x, because determining the "
2706 				"size for the mapping pairs failed with error "
2707 				"code %i.%s", vi->i_ino,
2708 				(unsigned)le32_to_cpu(ni->type), mp_size, es);
2709 		err = -EIO;
2710 		goto bad_out;
2711 	}
2712 	/*
2713 	 * Shrink the attribute record for the new mapping pairs array.  Note,
2714 	 * this cannot fail since we are making the attribute smaller thus by
2715 	 * definition there is enough space to do so.
2716 	 */
2717 	attr_len = le32_to_cpu(a->length);
2718 	err = ntfs_attr_record_resize(m, a, mp_size +
2719 			le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2720 	BUG_ON(err);
2721 	/*
2722 	 * Generate the mapping pairs array directly into the attribute record.
2723 	 */
2724 	err = ntfs_mapping_pairs_build(vol, (u8*)a +
2725 			le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2726 			mp_size, ni->runlist.rl, 0, -1, NULL);
2727 	if (unlikely(err)) {
2728 		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
2729 				"attribute type 0x%x, because building the "
2730 				"mapping pairs failed with error code %i.%s",
2731 				vi->i_ino, (unsigned)le32_to_cpu(ni->type),
2732 				err, es);
2733 		err = -EIO;
2734 		goto bad_out;
2735 	}
2736 	/* Update the allocated/compressed size as well as the highest vcn. */
2737 	a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2738 			vol->cluster_size_bits) - 1);
2739 	write_lock_irqsave(&ni->size_lock, flags);
2740 	ni->allocated_size = new_alloc_size;
2741 	a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2742 	if (NInoSparse(ni) || NInoCompressed(ni)) {
2743 		if (nr_freed) {
2744 			ni->itype.compressed.size -= nr_freed <<
2745 					vol->cluster_size_bits;
2746 			BUG_ON(ni->itype.compressed.size < 0);
2747 			a->data.non_resident.compressed_size = cpu_to_sle64(
2748 					ni->itype.compressed.size);
2749 			vi->i_blocks = ni->itype.compressed.size >> 9;
2750 		}
2751 	} else
2752 		vi->i_blocks = new_alloc_size >> 9;
2753 	write_unlock_irqrestore(&ni->size_lock, flags);
2754 	/*
2755 	 * We have shrunk the allocation.  If this is a shrinking truncate we
2756 	 * have already dealt with the initialized_size and the data_size above
2757 	 * and we are done.  If the truncate is only changing the allocation
2758 	 * and not the data_size, we are also done.  If this is an extending
2759 	 * truncate, need to extend the data_size now which is ensured by the
2760 	 * fact that @size_change is positive.
2761 	 */
2762 alloc_done:
2763 	/*
2764 	 * If the size is growing, need to update it now.  If it is shrinking,
2765 	 * we have already updated it above (before the allocation change).
2766 	 */
2767 	if (size_change > 0)
2768 		a->data.non_resident.data_size = cpu_to_sle64(new_size);
2769 	/* Ensure the modified mft record is written out. */
2770 	flush_dcache_mft_record_page(ctx->ntfs_ino);
2771 	mark_mft_record_dirty(ctx->ntfs_ino);
2772 unm_done:
2773 	ntfs_attr_put_search_ctx(ctx);
2774 	unmap_mft_record(base_ni);
2775 	up_write(&ni->runlist.lock);
2776 done:
2777 	/* Update the mtime and ctime on the base inode. */
2778 	/* normally ->truncate shouldn't update ctime or mtime,
2779 	 * but ntfs did before so it got a copy & paste version
2780 	 * of file_update_time.  one day someone should fix this
2781 	 * for real.
2782 	 */
2783 	if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
2784 		struct timespec64 now = current_time(VFS_I(base_ni));
2785 		int sync_it = 0;
2786 
2787 		if (!timespec64_equal(&VFS_I(base_ni)->i_mtime, &now) ||
2788 		    !timespec64_equal(&VFS_I(base_ni)->i_ctime, &now))
2789 			sync_it = 1;
2790 		VFS_I(base_ni)->i_mtime = now;
2791 		VFS_I(base_ni)->i_ctime = now;
2792 
2793 		if (sync_it)
2794 			mark_inode_dirty_sync(VFS_I(base_ni));
2795 	}
2796 
2797 	if (likely(!err)) {
2798 		NInoClearTruncateFailed(ni);
2799 		ntfs_debug("Done.");
2800 	}
2801 	return err;
2802 old_bad_out:
2803 	old_size = -1;
2804 bad_out:
2805 	if (err != -ENOMEM && err != -EOPNOTSUPP)
2806 		NVolSetErrors(vol);
2807 	if (err != -EOPNOTSUPP)
2808 		NInoSetTruncateFailed(ni);
2809 	else if (old_size >= 0)
2810 		i_size_write(vi, old_size);
2811 err_out:
2812 	if (ctx)
2813 		ntfs_attr_put_search_ctx(ctx);
2814 	if (m)
2815 		unmap_mft_record(base_ni);
2816 	up_write(&ni->runlist.lock);
2817 out:
2818 	ntfs_debug("Failed.  Returning error code %i.", err);
2819 	return err;
2820 conv_err_out:
2821 	if (err != -ENOMEM && err != -EOPNOTSUPP)
2822 		NVolSetErrors(vol);
2823 	if (err != -EOPNOTSUPP)
2824 		NInoSetTruncateFailed(ni);
2825 	else
2826 		i_size_write(vi, old_size);
2827 	goto out;
2828 }
2829 
2830 /**
2831  * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
2832  * @vi:		inode for which the i_size was changed
2833  *
2834  * Wrapper for ntfs_truncate() that has no return value.
2835  *
2836  * See ntfs_truncate() description above for details.
2837  */
2838 #ifdef NTFS_RW
2839 void ntfs_truncate_vfs(struct inode *vi) {
2840 	ntfs_truncate(vi);
2841 }
2842 #endif
2843 
2844 /**
2845  * ntfs_setattr - called from notify_change() when an attribute is being changed
2846  * @dentry:	dentry whose attributes to change
2847  * @attr:	structure describing the attributes and the changes
2848  *
2849  * We have to trap VFS attempts to truncate the file described by @dentry as
2850  * soon as possible, because we do not implement changes in i_size yet.  So we
2851  * abort all i_size changes here.
2852  *
2853  * We also abort all changes of user, group, and mode as we do not implement
2854  * the NTFS ACLs yet.
2855  *
2856  * Called with ->i_mutex held.
2857  */
2858 int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
2859 {
2860 	struct inode *vi = d_inode(dentry);
2861 	int err;
2862 	unsigned int ia_valid = attr->ia_valid;
2863 
2864 	err = setattr_prepare(dentry, attr);
2865 	if (err)
2866 		goto out;
2867 	/* We do not support NTFS ACLs yet. */
2868 	if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
2869 		ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
2870 				"supported yet, ignoring.");
2871 		err = -EOPNOTSUPP;
2872 		goto out;
2873 	}
2874 	if (ia_valid & ATTR_SIZE) {
2875 		if (attr->ia_size != i_size_read(vi)) {
2876 			ntfs_inode *ni = NTFS_I(vi);
2877 			/*
2878 			 * FIXME: For now we do not support resizing of
2879 			 * compressed or encrypted files yet.
2880 			 */
2881 			if (NInoCompressed(ni) || NInoEncrypted(ni)) {
2882 				ntfs_warning(vi->i_sb, "Changes in inode size "
2883 						"are not supported yet for "
2884 						"%s files, ignoring.",
2885 						NInoCompressed(ni) ?
2886 						"compressed" : "encrypted");
2887 				err = -EOPNOTSUPP;
2888 			} else {
2889 				truncate_setsize(vi, attr->ia_size);
2890 				ntfs_truncate_vfs(vi);
2891 			}
2892 			if (err || ia_valid == ATTR_SIZE)
2893 				goto out;
2894 		} else {
2895 			/*
2896 			 * We skipped the truncate but must still update
2897 			 * timestamps.
2898 			 */
2899 			ia_valid |= ATTR_MTIME | ATTR_CTIME;
2900 		}
2901 	}
2902 	if (ia_valid & ATTR_ATIME)
2903 		vi->i_atime = attr->ia_atime;
2904 	if (ia_valid & ATTR_MTIME)
2905 		vi->i_mtime = attr->ia_mtime;
2906 	if (ia_valid & ATTR_CTIME)
2907 		vi->i_ctime = attr->ia_ctime;
2908 	mark_inode_dirty(vi);
2909 out:
2910 	return err;
2911 }
2912 
2913 /**
2914  * ntfs_write_inode - write out a dirty inode
2915  * @vi:		inode to write out
2916  * @sync:	if true, write out synchronously
2917  *
2918  * Write out a dirty inode to disk including any extent inodes if present.
2919  *
2920  * If @sync is true, commit the inode to disk and wait for io completion.  This
2921  * is done using write_mft_record().
2922  *
2923  * If @sync is false, just schedule the write to happen but do not wait for i/o
2924  * completion.  In 2.6 kernels, scheduling usually happens just by virtue of
2925  * marking the page (and in this case mft record) dirty but we do not implement
2926  * this yet as write_mft_record() largely ignores the @sync parameter and
2927  * always performs synchronous writes.
2928  *
2929  * Return 0 on success and -errno on error.
2930  */
2931 int __ntfs_write_inode(struct inode *vi, int sync)
2932 {
2933 	sle64 nt;
2934 	ntfs_inode *ni = NTFS_I(vi);
2935 	ntfs_attr_search_ctx *ctx;
2936 	MFT_RECORD *m;
2937 	STANDARD_INFORMATION *si;
2938 	int err = 0;
2939 	bool modified = false;
2940 
2941 	ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
2942 			vi->i_ino);
2943 	/*
2944 	 * Dirty attribute inodes are written via their real inodes so just
2945 	 * clean them here.  Access time updates are taken care off when the
2946 	 * real inode is written.
2947 	 */
2948 	if (NInoAttr(ni)) {
2949 		NInoClearDirty(ni);
2950 		ntfs_debug("Done.");
2951 		return 0;
2952 	}
2953 	/* Map, pin, and lock the mft record belonging to the inode. */
2954 	m = map_mft_record(ni);
2955 	if (IS_ERR(m)) {
2956 		err = PTR_ERR(m);
2957 		goto err_out;
2958 	}
2959 	/* Update the access times in the standard information attribute. */
2960 	ctx = ntfs_attr_get_search_ctx(ni, m);
2961 	if (unlikely(!ctx)) {
2962 		err = -ENOMEM;
2963 		goto unm_err_out;
2964 	}
2965 	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
2966 			CASE_SENSITIVE, 0, NULL, 0, ctx);
2967 	if (unlikely(err)) {
2968 		ntfs_attr_put_search_ctx(ctx);
2969 		goto unm_err_out;
2970 	}
2971 	si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
2972 			le16_to_cpu(ctx->attr->data.resident.value_offset));
2973 	/* Update the access times if they have changed. */
2974 	nt = utc2ntfs(vi->i_mtime);
2975 	if (si->last_data_change_time != nt) {
2976 		ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
2977 				"new = 0x%llx", vi->i_ino, (long long)
2978 				sle64_to_cpu(si->last_data_change_time),
2979 				(long long)sle64_to_cpu(nt));
2980 		si->last_data_change_time = nt;
2981 		modified = true;
2982 	}
2983 	nt = utc2ntfs(vi->i_ctime);
2984 	if (si->last_mft_change_time != nt) {
2985 		ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
2986 				"new = 0x%llx", vi->i_ino, (long long)
2987 				sle64_to_cpu(si->last_mft_change_time),
2988 				(long long)sle64_to_cpu(nt));
2989 		si->last_mft_change_time = nt;
2990 		modified = true;
2991 	}
2992 	nt = utc2ntfs(vi->i_atime);
2993 	if (si->last_access_time != nt) {
2994 		ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
2995 				"new = 0x%llx", vi->i_ino,
2996 				(long long)sle64_to_cpu(si->last_access_time),
2997 				(long long)sle64_to_cpu(nt));
2998 		si->last_access_time = nt;
2999 		modified = true;
3000 	}
3001 	/*
3002 	 * If we just modified the standard information attribute we need to
3003 	 * mark the mft record it is in dirty.  We do this manually so that
3004 	 * mark_inode_dirty() is not called which would redirty the inode and
3005 	 * hence result in an infinite loop of trying to write the inode.
3006 	 * There is no need to mark the base inode nor the base mft record
3007 	 * dirty, since we are going to write this mft record below in any case
3008 	 * and the base mft record may actually not have been modified so it
3009 	 * might not need to be written out.
3010 	 * NOTE: It is not a problem when the inode for $MFT itself is being
3011 	 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
3012 	 * on the $MFT inode and hence ntfs_write_inode() will not be
3013 	 * re-invoked because of it which in turn is ok since the dirtied mft
3014 	 * record will be cleaned and written out to disk below, i.e. before
3015 	 * this function returns.
3016 	 */
3017 	if (modified) {
3018 		flush_dcache_mft_record_page(ctx->ntfs_ino);
3019 		if (!NInoTestSetDirty(ctx->ntfs_ino))
3020 			mark_ntfs_record_dirty(ctx->ntfs_ino->page,
3021 					ctx->ntfs_ino->page_ofs);
3022 	}
3023 	ntfs_attr_put_search_ctx(ctx);
3024 	/* Now the access times are updated, write the base mft record. */
3025 	if (NInoDirty(ni))
3026 		err = write_mft_record(ni, m, sync);
3027 	/* Write all attached extent mft records. */
3028 	mutex_lock(&ni->extent_lock);
3029 	if (ni->nr_extents > 0) {
3030 		ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
3031 		int i;
3032 
3033 		ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
3034 		for (i = 0; i < ni->nr_extents; i++) {
3035 			ntfs_inode *tni = extent_nis[i];
3036 
3037 			if (NInoDirty(tni)) {
3038 				MFT_RECORD *tm = map_mft_record(tni);
3039 				int ret;
3040 
3041 				if (IS_ERR(tm)) {
3042 					if (!err || err == -ENOMEM)
3043 						err = PTR_ERR(tm);
3044 					continue;
3045 				}
3046 				ret = write_mft_record(tni, tm, sync);
3047 				unmap_mft_record(tni);
3048 				if (unlikely(ret)) {
3049 					if (!err || err == -ENOMEM)
3050 						err = ret;
3051 				}
3052 			}
3053 		}
3054 	}
3055 	mutex_unlock(&ni->extent_lock);
3056 	unmap_mft_record(ni);
3057 	if (unlikely(err))
3058 		goto err_out;
3059 	ntfs_debug("Done.");
3060 	return 0;
3061 unm_err_out:
3062 	unmap_mft_record(ni);
3063 err_out:
3064 	if (err == -ENOMEM) {
3065 		ntfs_warning(vi->i_sb, "Not enough memory to write inode.  "
3066 				"Marking the inode dirty again, so the VFS "
3067 				"retries later.");
3068 		mark_inode_dirty(vi);
3069 	} else {
3070 		ntfs_error(vi->i_sb, "Failed (error %i):  Run chkdsk.", -err);
3071 		NVolSetErrors(ni->vol);
3072 	}
3073 	return err;
3074 }
3075 
3076 #endif /* NTFS_RW */
3077