xref: /openbmc/linux/fs/ntfs/runlist.c (revision b2441318)
1 /**
2  * runlist.c - NTFS runlist handling code.  Part of the Linux-NTFS project.
3  *
4  * Copyright (c) 2001-2007 Anton Altaparmakov
5  * Copyright (c) 2002-2005 Richard Russon
6  *
7  * This program/include file is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU General Public License as published
9  * by the Free Software Foundation; either version 2 of the License, or
10  * (at your option) any later version.
11  *
12  * This program/include file is distributed in the hope that it will be
13  * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
14  * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15  * GNU General Public License for more details.
16  *
17  * You should have received a copy of the GNU General Public License
18  * along with this program (in the main directory of the Linux-NTFS
19  * distribution in the file COPYING); if not, write to the Free Software
20  * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
21  */
22 
23 #include "debug.h"
24 #include "dir.h"
25 #include "endian.h"
26 #include "malloc.h"
27 #include "ntfs.h"
28 
29 /**
30  * ntfs_rl_mm - runlist memmove
31  *
32  * It is up to the caller to serialize access to the runlist @base.
33  */
34 static inline void ntfs_rl_mm(runlist_element *base, int dst, int src,
35 		int size)
36 {
37 	if (likely((dst != src) && (size > 0)))
38 		memmove(base + dst, base + src, size * sizeof(*base));
39 }
40 
41 /**
42  * ntfs_rl_mc - runlist memory copy
43  *
44  * It is up to the caller to serialize access to the runlists @dstbase and
45  * @srcbase.
46  */
47 static inline void ntfs_rl_mc(runlist_element *dstbase, int dst,
48 		runlist_element *srcbase, int src, int size)
49 {
50 	if (likely(size > 0))
51 		memcpy(dstbase + dst, srcbase + src, size * sizeof(*dstbase));
52 }
53 
54 /**
55  * ntfs_rl_realloc - Reallocate memory for runlists
56  * @rl:		original runlist
57  * @old_size:	number of runlist elements in the original runlist @rl
58  * @new_size:	number of runlist elements we need space for
59  *
60  * As the runlists grow, more memory will be required.  To prevent the
61  * kernel having to allocate and reallocate large numbers of small bits of
62  * memory, this function returns an entire page of memory.
63  *
64  * It is up to the caller to serialize access to the runlist @rl.
65  *
66  * N.B.  If the new allocation doesn't require a different number of pages in
67  *       memory, the function will return the original pointer.
68  *
69  * On success, return a pointer to the newly allocated, or recycled, memory.
70  * On error, return -errno. The following error codes are defined:
71  *	-ENOMEM	- Not enough memory to allocate runlist array.
72  *	-EINVAL	- Invalid parameters were passed in.
73  */
74 static inline runlist_element *ntfs_rl_realloc(runlist_element *rl,
75 		int old_size, int new_size)
76 {
77 	runlist_element *new_rl;
78 
79 	old_size = PAGE_ALIGN(old_size * sizeof(*rl));
80 	new_size = PAGE_ALIGN(new_size * sizeof(*rl));
81 	if (old_size == new_size)
82 		return rl;
83 
84 	new_rl = ntfs_malloc_nofs(new_size);
85 	if (unlikely(!new_rl))
86 		return ERR_PTR(-ENOMEM);
87 
88 	if (likely(rl != NULL)) {
89 		if (unlikely(old_size > new_size))
90 			old_size = new_size;
91 		memcpy(new_rl, rl, old_size);
92 		ntfs_free(rl);
93 	}
94 	return new_rl;
95 }
96 
97 /**
98  * ntfs_rl_realloc_nofail - Reallocate memory for runlists
99  * @rl:		original runlist
100  * @old_size:	number of runlist elements in the original runlist @rl
101  * @new_size:	number of runlist elements we need space for
102  *
103  * As the runlists grow, more memory will be required.  To prevent the
104  * kernel having to allocate and reallocate large numbers of small bits of
105  * memory, this function returns an entire page of memory.
106  *
107  * This function guarantees that the allocation will succeed.  It will sleep
108  * for as long as it takes to complete the allocation.
109  *
110  * It is up to the caller to serialize access to the runlist @rl.
111  *
112  * N.B.  If the new allocation doesn't require a different number of pages in
113  *       memory, the function will return the original pointer.
114  *
115  * On success, return a pointer to the newly allocated, or recycled, memory.
116  * On error, return -errno. The following error codes are defined:
117  *	-ENOMEM	- Not enough memory to allocate runlist array.
118  *	-EINVAL	- Invalid parameters were passed in.
119  */
120 static inline runlist_element *ntfs_rl_realloc_nofail(runlist_element *rl,
121 		int old_size, int new_size)
122 {
123 	runlist_element *new_rl;
124 
125 	old_size = PAGE_ALIGN(old_size * sizeof(*rl));
126 	new_size = PAGE_ALIGN(new_size * sizeof(*rl));
127 	if (old_size == new_size)
128 		return rl;
129 
130 	new_rl = ntfs_malloc_nofs_nofail(new_size);
131 	BUG_ON(!new_rl);
132 
133 	if (likely(rl != NULL)) {
134 		if (unlikely(old_size > new_size))
135 			old_size = new_size;
136 		memcpy(new_rl, rl, old_size);
137 		ntfs_free(rl);
138 	}
139 	return new_rl;
140 }
141 
142 /**
143  * ntfs_are_rl_mergeable - test if two runlists can be joined together
144  * @dst:	original runlist
145  * @src:	new runlist to test for mergeability with @dst
146  *
147  * Test if two runlists can be joined together. For this, their VCNs and LCNs
148  * must be adjacent.
149  *
150  * It is up to the caller to serialize access to the runlists @dst and @src.
151  *
152  * Return: true   Success, the runlists can be merged.
153  *	   false  Failure, the runlists cannot be merged.
154  */
155 static inline bool ntfs_are_rl_mergeable(runlist_element *dst,
156 		runlist_element *src)
157 {
158 	BUG_ON(!dst);
159 	BUG_ON(!src);
160 
161 	/* We can merge unmapped regions even if they are misaligned. */
162 	if ((dst->lcn == LCN_RL_NOT_MAPPED) && (src->lcn == LCN_RL_NOT_MAPPED))
163 		return true;
164 	/* If the runs are misaligned, we cannot merge them. */
165 	if ((dst->vcn + dst->length) != src->vcn)
166 		return false;
167 	/* If both runs are non-sparse and contiguous, we can merge them. */
168 	if ((dst->lcn >= 0) && (src->lcn >= 0) &&
169 			((dst->lcn + dst->length) == src->lcn))
170 		return true;
171 	/* If we are merging two holes, we can merge them. */
172 	if ((dst->lcn == LCN_HOLE) && (src->lcn == LCN_HOLE))
173 		return true;
174 	/* Cannot merge. */
175 	return false;
176 }
177 
178 /**
179  * __ntfs_rl_merge - merge two runlists without testing if they can be merged
180  * @dst:	original, destination runlist
181  * @src:	new runlist to merge with @dst
182  *
183  * Merge the two runlists, writing into the destination runlist @dst. The
184  * caller must make sure the runlists can be merged or this will corrupt the
185  * destination runlist.
186  *
187  * It is up to the caller to serialize access to the runlists @dst and @src.
188  */
189 static inline void __ntfs_rl_merge(runlist_element *dst, runlist_element *src)
190 {
191 	dst->length += src->length;
192 }
193 
194 /**
195  * ntfs_rl_append - append a runlist after a given element
196  * @dst:	original runlist to be worked on
197  * @dsize:	number of elements in @dst (including end marker)
198  * @src:	runlist to be inserted into @dst
199  * @ssize:	number of elements in @src (excluding end marker)
200  * @loc:	append the new runlist @src after this element in @dst
201  *
202  * Append the runlist @src after element @loc in @dst.  Merge the right end of
203  * the new runlist, if necessary. Adjust the size of the hole before the
204  * appended runlist.
205  *
206  * It is up to the caller to serialize access to the runlists @dst and @src.
207  *
208  * On success, return a pointer to the new, combined, runlist. Note, both
209  * runlists @dst and @src are deallocated before returning so you cannot use
210  * the pointers for anything any more. (Strictly speaking the returned runlist
211  * may be the same as @dst but this is irrelevant.)
212  *
213  * On error, return -errno. Both runlists are left unmodified. The following
214  * error codes are defined:
215  *	-ENOMEM	- Not enough memory to allocate runlist array.
216  *	-EINVAL	- Invalid parameters were passed in.
217  */
218 static inline runlist_element *ntfs_rl_append(runlist_element *dst,
219 		int dsize, runlist_element *src, int ssize, int loc)
220 {
221 	bool right = false;	/* Right end of @src needs merging. */
222 	int marker;		/* End of the inserted runs. */
223 
224 	BUG_ON(!dst);
225 	BUG_ON(!src);
226 
227 	/* First, check if the right hand end needs merging. */
228 	if ((loc + 1) < dsize)
229 		right = ntfs_are_rl_mergeable(src + ssize - 1, dst + loc + 1);
230 
231 	/* Space required: @dst size + @src size, less one if we merged. */
232 	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize - right);
233 	if (IS_ERR(dst))
234 		return dst;
235 	/*
236 	 * We are guaranteed to succeed from here so can start modifying the
237 	 * original runlists.
238 	 */
239 
240 	/* First, merge the right hand end, if necessary. */
241 	if (right)
242 		__ntfs_rl_merge(src + ssize - 1, dst + loc + 1);
243 
244 	/* First run after the @src runs that have been inserted. */
245 	marker = loc + ssize + 1;
246 
247 	/* Move the tail of @dst out of the way, then copy in @src. */
248 	ntfs_rl_mm(dst, marker, loc + 1 + right, dsize - (loc + 1 + right));
249 	ntfs_rl_mc(dst, loc + 1, src, 0, ssize);
250 
251 	/* Adjust the size of the preceding hole. */
252 	dst[loc].length = dst[loc + 1].vcn - dst[loc].vcn;
253 
254 	/* We may have changed the length of the file, so fix the end marker */
255 	if (dst[marker].lcn == LCN_ENOENT)
256 		dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
257 
258 	return dst;
259 }
260 
261 /**
262  * ntfs_rl_insert - insert a runlist into another
263  * @dst:	original runlist to be worked on
264  * @dsize:	number of elements in @dst (including end marker)
265  * @src:	new runlist to be inserted
266  * @ssize:	number of elements in @src (excluding end marker)
267  * @loc:	insert the new runlist @src before this element in @dst
268  *
269  * Insert the runlist @src before element @loc in the runlist @dst. Merge the
270  * left end of the new runlist, if necessary. Adjust the size of the hole
271  * after the inserted runlist.
272  *
273  * It is up to the caller to serialize access to the runlists @dst and @src.
274  *
275  * On success, return a pointer to the new, combined, runlist. Note, both
276  * runlists @dst and @src are deallocated before returning so you cannot use
277  * the pointers for anything any more. (Strictly speaking the returned runlist
278  * may be the same as @dst but this is irrelevant.)
279  *
280  * On error, return -errno. Both runlists are left unmodified. The following
281  * error codes are defined:
282  *	-ENOMEM	- Not enough memory to allocate runlist array.
283  *	-EINVAL	- Invalid parameters were passed in.
284  */
285 static inline runlist_element *ntfs_rl_insert(runlist_element *dst,
286 		int dsize, runlist_element *src, int ssize, int loc)
287 {
288 	bool left = false;	/* Left end of @src needs merging. */
289 	bool disc = false;	/* Discontinuity between @dst and @src. */
290 	int marker;		/* End of the inserted runs. */
291 
292 	BUG_ON(!dst);
293 	BUG_ON(!src);
294 
295 	/*
296 	 * disc => Discontinuity between the end of @dst and the start of @src.
297 	 *	   This means we might need to insert a "not mapped" run.
298 	 */
299 	if (loc == 0)
300 		disc = (src[0].vcn > 0);
301 	else {
302 		s64 merged_length;
303 
304 		left = ntfs_are_rl_mergeable(dst + loc - 1, src);
305 
306 		merged_length = dst[loc - 1].length;
307 		if (left)
308 			merged_length += src->length;
309 
310 		disc = (src[0].vcn > dst[loc - 1].vcn + merged_length);
311 	}
312 	/*
313 	 * Space required: @dst size + @src size, less one if we merged, plus
314 	 * one if there was a discontinuity.
315 	 */
316 	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize - left + disc);
317 	if (IS_ERR(dst))
318 		return dst;
319 	/*
320 	 * We are guaranteed to succeed from here so can start modifying the
321 	 * original runlist.
322 	 */
323 	if (left)
324 		__ntfs_rl_merge(dst + loc - 1, src);
325 	/*
326 	 * First run after the @src runs that have been inserted.
327 	 * Nominally,  @marker equals @loc + @ssize, i.e. location + number of
328 	 * runs in @src.  However, if @left, then the first run in @src has
329 	 * been merged with one in @dst.  And if @disc, then @dst and @src do
330 	 * not meet and we need an extra run to fill the gap.
331 	 */
332 	marker = loc + ssize - left + disc;
333 
334 	/* Move the tail of @dst out of the way, then copy in @src. */
335 	ntfs_rl_mm(dst, marker, loc, dsize - loc);
336 	ntfs_rl_mc(dst, loc + disc, src, left, ssize - left);
337 
338 	/* Adjust the VCN of the first run after the insertion... */
339 	dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
340 	/* ... and the length. */
341 	if (dst[marker].lcn == LCN_HOLE || dst[marker].lcn == LCN_RL_NOT_MAPPED)
342 		dst[marker].length = dst[marker + 1].vcn - dst[marker].vcn;
343 
344 	/* Writing beyond the end of the file and there is a discontinuity. */
345 	if (disc) {
346 		if (loc > 0) {
347 			dst[loc].vcn = dst[loc - 1].vcn + dst[loc - 1].length;
348 			dst[loc].length = dst[loc + 1].vcn - dst[loc].vcn;
349 		} else {
350 			dst[loc].vcn = 0;
351 			dst[loc].length = dst[loc + 1].vcn;
352 		}
353 		dst[loc].lcn = LCN_RL_NOT_MAPPED;
354 	}
355 	return dst;
356 }
357 
358 /**
359  * ntfs_rl_replace - overwrite a runlist element with another runlist
360  * @dst:	original runlist to be worked on
361  * @dsize:	number of elements in @dst (including end marker)
362  * @src:	new runlist to be inserted
363  * @ssize:	number of elements in @src (excluding end marker)
364  * @loc:	index in runlist @dst to overwrite with @src
365  *
366  * Replace the runlist element @dst at @loc with @src. Merge the left and
367  * right ends of the inserted runlist, if necessary.
368  *
369  * It is up to the caller to serialize access to the runlists @dst and @src.
370  *
371  * On success, return a pointer to the new, combined, runlist. Note, both
372  * runlists @dst and @src are deallocated before returning so you cannot use
373  * the pointers for anything any more. (Strictly speaking the returned runlist
374  * may be the same as @dst but this is irrelevant.)
375  *
376  * On error, return -errno. Both runlists are left unmodified. The following
377  * error codes are defined:
378  *	-ENOMEM	- Not enough memory to allocate runlist array.
379  *	-EINVAL	- Invalid parameters were passed in.
380  */
381 static inline runlist_element *ntfs_rl_replace(runlist_element *dst,
382 		int dsize, runlist_element *src, int ssize, int loc)
383 {
384 	signed delta;
385 	bool left = false;	/* Left end of @src needs merging. */
386 	bool right = false;	/* Right end of @src needs merging. */
387 	int tail;		/* Start of tail of @dst. */
388 	int marker;		/* End of the inserted runs. */
389 
390 	BUG_ON(!dst);
391 	BUG_ON(!src);
392 
393 	/* First, see if the left and right ends need merging. */
394 	if ((loc + 1) < dsize)
395 		right = ntfs_are_rl_mergeable(src + ssize - 1, dst + loc + 1);
396 	if (loc > 0)
397 		left = ntfs_are_rl_mergeable(dst + loc - 1, src);
398 	/*
399 	 * Allocate some space.  We will need less if the left, right, or both
400 	 * ends get merged.  The -1 accounts for the run being replaced.
401 	 */
402 	delta = ssize - 1 - left - right;
403 	if (delta > 0) {
404 		dst = ntfs_rl_realloc(dst, dsize, dsize + delta);
405 		if (IS_ERR(dst))
406 			return dst;
407 	}
408 	/*
409 	 * We are guaranteed to succeed from here so can start modifying the
410 	 * original runlists.
411 	 */
412 
413 	/* First, merge the left and right ends, if necessary. */
414 	if (right)
415 		__ntfs_rl_merge(src + ssize - 1, dst + loc + 1);
416 	if (left)
417 		__ntfs_rl_merge(dst + loc - 1, src);
418 	/*
419 	 * Offset of the tail of @dst.  This needs to be moved out of the way
420 	 * to make space for the runs to be copied from @src, i.e. the first
421 	 * run of the tail of @dst.
422 	 * Nominally, @tail equals @loc + 1, i.e. location, skipping the
423 	 * replaced run.  However, if @right, then one of @dst's runs is
424 	 * already merged into @src.
425 	 */
426 	tail = loc + right + 1;
427 	/*
428 	 * First run after the @src runs that have been inserted, i.e. where
429 	 * the tail of @dst needs to be moved to.
430 	 * Nominally, @marker equals @loc + @ssize, i.e. location + number of
431 	 * runs in @src.  However, if @left, then the first run in @src has
432 	 * been merged with one in @dst.
433 	 */
434 	marker = loc + ssize - left;
435 
436 	/* Move the tail of @dst out of the way, then copy in @src. */
437 	ntfs_rl_mm(dst, marker, tail, dsize - tail);
438 	ntfs_rl_mc(dst, loc, src, left, ssize - left);
439 
440 	/* We may have changed the length of the file, so fix the end marker. */
441 	if (dsize - tail > 0 && dst[marker].lcn == LCN_ENOENT)
442 		dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
443 	return dst;
444 }
445 
446 /**
447  * ntfs_rl_split - insert a runlist into the centre of a hole
448  * @dst:	original runlist to be worked on
449  * @dsize:	number of elements in @dst (including end marker)
450  * @src:	new runlist to be inserted
451  * @ssize:	number of elements in @src (excluding end marker)
452  * @loc:	index in runlist @dst at which to split and insert @src
453  *
454  * Split the runlist @dst at @loc into two and insert @new in between the two
455  * fragments. No merging of runlists is necessary. Adjust the size of the
456  * holes either side.
457  *
458  * It is up to the caller to serialize access to the runlists @dst and @src.
459  *
460  * On success, return a pointer to the new, combined, runlist. Note, both
461  * runlists @dst and @src are deallocated before returning so you cannot use
462  * the pointers for anything any more. (Strictly speaking the returned runlist
463  * may be the same as @dst but this is irrelevant.)
464  *
465  * On error, return -errno. Both runlists are left unmodified. The following
466  * error codes are defined:
467  *	-ENOMEM	- Not enough memory to allocate runlist array.
468  *	-EINVAL	- Invalid parameters were passed in.
469  */
470 static inline runlist_element *ntfs_rl_split(runlist_element *dst, int dsize,
471 		runlist_element *src, int ssize, int loc)
472 {
473 	BUG_ON(!dst);
474 	BUG_ON(!src);
475 
476 	/* Space required: @dst size + @src size + one new hole. */
477 	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize + 1);
478 	if (IS_ERR(dst))
479 		return dst;
480 	/*
481 	 * We are guaranteed to succeed from here so can start modifying the
482 	 * original runlists.
483 	 */
484 
485 	/* Move the tail of @dst out of the way, then copy in @src. */
486 	ntfs_rl_mm(dst, loc + 1 + ssize, loc, dsize - loc);
487 	ntfs_rl_mc(dst, loc + 1, src, 0, ssize);
488 
489 	/* Adjust the size of the holes either size of @src. */
490 	dst[loc].length		= dst[loc+1].vcn       - dst[loc].vcn;
491 	dst[loc+ssize+1].vcn    = dst[loc+ssize].vcn   + dst[loc+ssize].length;
492 	dst[loc+ssize+1].length = dst[loc+ssize+2].vcn - dst[loc+ssize+1].vcn;
493 
494 	return dst;
495 }
496 
497 /**
498  * ntfs_runlists_merge - merge two runlists into one
499  * @drl:	original runlist to be worked on
500  * @srl:	new runlist to be merged into @drl
501  *
502  * First we sanity check the two runlists @srl and @drl to make sure that they
503  * are sensible and can be merged. The runlist @srl must be either after the
504  * runlist @drl or completely within a hole (or unmapped region) in @drl.
505  *
506  * It is up to the caller to serialize access to the runlists @drl and @srl.
507  *
508  * Merging of runlists is necessary in two cases:
509  *   1. When attribute lists are used and a further extent is being mapped.
510  *   2. When new clusters are allocated to fill a hole or extend a file.
511  *
512  * There are four possible ways @srl can be merged. It can:
513  *	- be inserted at the beginning of a hole,
514  *	- split the hole in two and be inserted between the two fragments,
515  *	- be appended at the end of a hole, or it can
516  *	- replace the whole hole.
517  * It can also be appended to the end of the runlist, which is just a variant
518  * of the insert case.
519  *
520  * On success, return a pointer to the new, combined, runlist. Note, both
521  * runlists @drl and @srl are deallocated before returning so you cannot use
522  * the pointers for anything any more. (Strictly speaking the returned runlist
523  * may be the same as @dst but this is irrelevant.)
524  *
525  * On error, return -errno. Both runlists are left unmodified. The following
526  * error codes are defined:
527  *	-ENOMEM	- Not enough memory to allocate runlist array.
528  *	-EINVAL	- Invalid parameters were passed in.
529  *	-ERANGE	- The runlists overlap and cannot be merged.
530  */
531 runlist_element *ntfs_runlists_merge(runlist_element *drl,
532 		runlist_element *srl)
533 {
534 	int di, si;		/* Current index into @[ds]rl. */
535 	int sstart;		/* First index with lcn > LCN_RL_NOT_MAPPED. */
536 	int dins;		/* Index into @drl at which to insert @srl. */
537 	int dend, send;		/* Last index into @[ds]rl. */
538 	int dfinal, sfinal;	/* The last index into @[ds]rl with
539 				   lcn >= LCN_HOLE. */
540 	int marker = 0;
541 	VCN marker_vcn = 0;
542 
543 #ifdef DEBUG
544 	ntfs_debug("dst:");
545 	ntfs_debug_dump_runlist(drl);
546 	ntfs_debug("src:");
547 	ntfs_debug_dump_runlist(srl);
548 #endif
549 
550 	/* Check for silly calling... */
551 	if (unlikely(!srl))
552 		return drl;
553 	if (IS_ERR(srl) || IS_ERR(drl))
554 		return ERR_PTR(-EINVAL);
555 
556 	/* Check for the case where the first mapping is being done now. */
557 	if (unlikely(!drl)) {
558 		drl = srl;
559 		/* Complete the source runlist if necessary. */
560 		if (unlikely(drl[0].vcn)) {
561 			/* Scan to the end of the source runlist. */
562 			for (dend = 0; likely(drl[dend].length); dend++)
563 				;
564 			dend++;
565 			drl = ntfs_rl_realloc(drl, dend, dend + 1);
566 			if (IS_ERR(drl))
567 				return drl;
568 			/* Insert start element at the front of the runlist. */
569 			ntfs_rl_mm(drl, 1, 0, dend);
570 			drl[0].vcn = 0;
571 			drl[0].lcn = LCN_RL_NOT_MAPPED;
572 			drl[0].length = drl[1].vcn;
573 		}
574 		goto finished;
575 	}
576 
577 	si = di = 0;
578 
579 	/* Skip any unmapped start element(s) in the source runlist. */
580 	while (srl[si].length && srl[si].lcn < LCN_HOLE)
581 		si++;
582 
583 	/* Can't have an entirely unmapped source runlist. */
584 	BUG_ON(!srl[si].length);
585 
586 	/* Record the starting points. */
587 	sstart = si;
588 
589 	/*
590 	 * Skip forward in @drl until we reach the position where @srl needs to
591 	 * be inserted. If we reach the end of @drl, @srl just needs to be
592 	 * appended to @drl.
593 	 */
594 	for (; drl[di].length; di++) {
595 		if (drl[di].vcn + drl[di].length > srl[sstart].vcn)
596 			break;
597 	}
598 	dins = di;
599 
600 	/* Sanity check for illegal overlaps. */
601 	if ((drl[di].vcn == srl[si].vcn) && (drl[di].lcn >= 0) &&
602 			(srl[si].lcn >= 0)) {
603 		ntfs_error(NULL, "Run lists overlap. Cannot merge!");
604 		return ERR_PTR(-ERANGE);
605 	}
606 
607 	/* Scan to the end of both runlists in order to know their sizes. */
608 	for (send = si; srl[send].length; send++)
609 		;
610 	for (dend = di; drl[dend].length; dend++)
611 		;
612 
613 	if (srl[send].lcn == LCN_ENOENT)
614 		marker_vcn = srl[marker = send].vcn;
615 
616 	/* Scan to the last element with lcn >= LCN_HOLE. */
617 	for (sfinal = send; sfinal >= 0 && srl[sfinal].lcn < LCN_HOLE; sfinal--)
618 		;
619 	for (dfinal = dend; dfinal >= 0 && drl[dfinal].lcn < LCN_HOLE; dfinal--)
620 		;
621 
622 	{
623 	bool start;
624 	bool finish;
625 	int ds = dend + 1;		/* Number of elements in drl & srl */
626 	int ss = sfinal - sstart + 1;
627 
628 	start  = ((drl[dins].lcn <  LCN_RL_NOT_MAPPED) ||    /* End of file   */
629 		  (drl[dins].vcn == srl[sstart].vcn));	     /* Start of hole */
630 	finish = ((drl[dins].lcn >= LCN_RL_NOT_MAPPED) &&    /* End of file   */
631 		 ((drl[dins].vcn + drl[dins].length) <=      /* End of hole   */
632 		  (srl[send - 1].vcn + srl[send - 1].length)));
633 
634 	/* Or we will lose an end marker. */
635 	if (finish && !drl[dins].length)
636 		ss++;
637 	if (marker && (drl[dins].vcn + drl[dins].length > srl[send - 1].vcn))
638 		finish = false;
639 #if 0
640 	ntfs_debug("dfinal = %i, dend = %i", dfinal, dend);
641 	ntfs_debug("sstart = %i, sfinal = %i, send = %i", sstart, sfinal, send);
642 	ntfs_debug("start = %i, finish = %i", start, finish);
643 	ntfs_debug("ds = %i, ss = %i, dins = %i", ds, ss, dins);
644 #endif
645 	if (start) {
646 		if (finish)
647 			drl = ntfs_rl_replace(drl, ds, srl + sstart, ss, dins);
648 		else
649 			drl = ntfs_rl_insert(drl, ds, srl + sstart, ss, dins);
650 	} else {
651 		if (finish)
652 			drl = ntfs_rl_append(drl, ds, srl + sstart, ss, dins);
653 		else
654 			drl = ntfs_rl_split(drl, ds, srl + sstart, ss, dins);
655 	}
656 	if (IS_ERR(drl)) {
657 		ntfs_error(NULL, "Merge failed.");
658 		return drl;
659 	}
660 	ntfs_free(srl);
661 	if (marker) {
662 		ntfs_debug("Triggering marker code.");
663 		for (ds = dend; drl[ds].length; ds++)
664 			;
665 		/* We only need to care if @srl ended after @drl. */
666 		if (drl[ds].vcn <= marker_vcn) {
667 			int slots = 0;
668 
669 			if (drl[ds].vcn == marker_vcn) {
670 				ntfs_debug("Old marker = 0x%llx, replacing "
671 						"with LCN_ENOENT.",
672 						(unsigned long long)
673 						drl[ds].lcn);
674 				drl[ds].lcn = LCN_ENOENT;
675 				goto finished;
676 			}
677 			/*
678 			 * We need to create an unmapped runlist element in
679 			 * @drl or extend an existing one before adding the
680 			 * ENOENT terminator.
681 			 */
682 			if (drl[ds].lcn == LCN_ENOENT) {
683 				ds--;
684 				slots = 1;
685 			}
686 			if (drl[ds].lcn != LCN_RL_NOT_MAPPED) {
687 				/* Add an unmapped runlist element. */
688 				if (!slots) {
689 					drl = ntfs_rl_realloc_nofail(drl, ds,
690 							ds + 2);
691 					slots = 2;
692 				}
693 				ds++;
694 				/* Need to set vcn if it isn't set already. */
695 				if (slots != 1)
696 					drl[ds].vcn = drl[ds - 1].vcn +
697 							drl[ds - 1].length;
698 				drl[ds].lcn = LCN_RL_NOT_MAPPED;
699 				/* We now used up a slot. */
700 				slots--;
701 			}
702 			drl[ds].length = marker_vcn - drl[ds].vcn;
703 			/* Finally add the ENOENT terminator. */
704 			ds++;
705 			if (!slots)
706 				drl = ntfs_rl_realloc_nofail(drl, ds, ds + 1);
707 			drl[ds].vcn = marker_vcn;
708 			drl[ds].lcn = LCN_ENOENT;
709 			drl[ds].length = (s64)0;
710 		}
711 	}
712 	}
713 
714 finished:
715 	/* The merge was completed successfully. */
716 	ntfs_debug("Merged runlist:");
717 	ntfs_debug_dump_runlist(drl);
718 	return drl;
719 }
720 
721 /**
722  * ntfs_mapping_pairs_decompress - convert mapping pairs array to runlist
723  * @vol:	ntfs volume on which the attribute resides
724  * @attr:	attribute record whose mapping pairs array to decompress
725  * @old_rl:	optional runlist in which to insert @attr's runlist
726  *
727  * It is up to the caller to serialize access to the runlist @old_rl.
728  *
729  * Decompress the attribute @attr's mapping pairs array into a runlist. On
730  * success, return the decompressed runlist.
731  *
732  * If @old_rl is not NULL, decompressed runlist is inserted into the
733  * appropriate place in @old_rl and the resultant, combined runlist is
734  * returned. The original @old_rl is deallocated.
735  *
736  * On error, return -errno. @old_rl is left unmodified in that case.
737  *
738  * The following error codes are defined:
739  *	-ENOMEM	- Not enough memory to allocate runlist array.
740  *	-EIO	- Corrupt runlist.
741  *	-EINVAL	- Invalid parameters were passed in.
742  *	-ERANGE	- The two runlists overlap.
743  *
744  * FIXME: For now we take the conceptionally simplest approach of creating the
745  * new runlist disregarding the already existing one and then splicing the
746  * two into one, if that is possible (we check for overlap and discard the new
747  * runlist if overlap present before returning ERR_PTR(-ERANGE)).
748  */
749 runlist_element *ntfs_mapping_pairs_decompress(const ntfs_volume *vol,
750 		const ATTR_RECORD *attr, runlist_element *old_rl)
751 {
752 	VCN vcn;		/* Current vcn. */
753 	LCN lcn;		/* Current lcn. */
754 	s64 deltaxcn;		/* Change in [vl]cn. */
755 	runlist_element *rl;	/* The output runlist. */
756 	u8 *buf;		/* Current position in mapping pairs array. */
757 	u8 *attr_end;		/* End of attribute. */
758 	int rlsize;		/* Size of runlist buffer. */
759 	u16 rlpos;		/* Current runlist position in units of
760 				   runlist_elements. */
761 	u8 b;			/* Current byte offset in buf. */
762 
763 #ifdef DEBUG
764 	/* Make sure attr exists and is non-resident. */
765 	if (!attr || !attr->non_resident || sle64_to_cpu(
766 			attr->data.non_resident.lowest_vcn) < (VCN)0) {
767 		ntfs_error(vol->sb, "Invalid arguments.");
768 		return ERR_PTR(-EINVAL);
769 	}
770 #endif
771 	/* Start at vcn = lowest_vcn and lcn 0. */
772 	vcn = sle64_to_cpu(attr->data.non_resident.lowest_vcn);
773 	lcn = 0;
774 	/* Get start of the mapping pairs array. */
775 	buf = (u8*)attr + le16_to_cpu(
776 			attr->data.non_resident.mapping_pairs_offset);
777 	attr_end = (u8*)attr + le32_to_cpu(attr->length);
778 	if (unlikely(buf < (u8*)attr || buf > attr_end)) {
779 		ntfs_error(vol->sb, "Corrupt attribute.");
780 		return ERR_PTR(-EIO);
781 	}
782 	/* If the mapping pairs array is valid but empty, nothing to do. */
783 	if (!vcn && !*buf)
784 		return old_rl;
785 	/* Current position in runlist array. */
786 	rlpos = 0;
787 	/* Allocate first page and set current runlist size to one page. */
788 	rl = ntfs_malloc_nofs(rlsize = PAGE_SIZE);
789 	if (unlikely(!rl))
790 		return ERR_PTR(-ENOMEM);
791 	/* Insert unmapped starting element if necessary. */
792 	if (vcn) {
793 		rl->vcn = 0;
794 		rl->lcn = LCN_RL_NOT_MAPPED;
795 		rl->length = vcn;
796 		rlpos++;
797 	}
798 	while (buf < attr_end && *buf) {
799 		/*
800 		 * Allocate more memory if needed, including space for the
801 		 * not-mapped and terminator elements. ntfs_malloc_nofs()
802 		 * operates on whole pages only.
803 		 */
804 		if (((rlpos + 3) * sizeof(*old_rl)) > rlsize) {
805 			runlist_element *rl2;
806 
807 			rl2 = ntfs_malloc_nofs(rlsize + (int)PAGE_SIZE);
808 			if (unlikely(!rl2)) {
809 				ntfs_free(rl);
810 				return ERR_PTR(-ENOMEM);
811 			}
812 			memcpy(rl2, rl, rlsize);
813 			ntfs_free(rl);
814 			rl = rl2;
815 			rlsize += PAGE_SIZE;
816 		}
817 		/* Enter the current vcn into the current runlist element. */
818 		rl[rlpos].vcn = vcn;
819 		/*
820 		 * Get the change in vcn, i.e. the run length in clusters.
821 		 * Doing it this way ensures that we signextend negative values.
822 		 * A negative run length doesn't make any sense, but hey, I
823 		 * didn't make up the NTFS specs and Windows NT4 treats the run
824 		 * length as a signed value so that's how it is...
825 		 */
826 		b = *buf & 0xf;
827 		if (b) {
828 			if (unlikely(buf + b > attr_end))
829 				goto io_error;
830 			for (deltaxcn = (s8)buf[b--]; b; b--)
831 				deltaxcn = (deltaxcn << 8) + buf[b];
832 		} else { /* The length entry is compulsory. */
833 			ntfs_error(vol->sb, "Missing length entry in mapping "
834 					"pairs array.");
835 			deltaxcn = (s64)-1;
836 		}
837 		/*
838 		 * Assume a negative length to indicate data corruption and
839 		 * hence clean-up and return NULL.
840 		 */
841 		if (unlikely(deltaxcn < 0)) {
842 			ntfs_error(vol->sb, "Invalid length in mapping pairs "
843 					"array.");
844 			goto err_out;
845 		}
846 		/*
847 		 * Enter the current run length into the current runlist
848 		 * element.
849 		 */
850 		rl[rlpos].length = deltaxcn;
851 		/* Increment the current vcn by the current run length. */
852 		vcn += deltaxcn;
853 		/*
854 		 * There might be no lcn change at all, as is the case for
855 		 * sparse clusters on NTFS 3.0+, in which case we set the lcn
856 		 * to LCN_HOLE.
857 		 */
858 		if (!(*buf & 0xf0))
859 			rl[rlpos].lcn = LCN_HOLE;
860 		else {
861 			/* Get the lcn change which really can be negative. */
862 			u8 b2 = *buf & 0xf;
863 			b = b2 + ((*buf >> 4) & 0xf);
864 			if (buf + b > attr_end)
865 				goto io_error;
866 			for (deltaxcn = (s8)buf[b--]; b > b2; b--)
867 				deltaxcn = (deltaxcn << 8) + buf[b];
868 			/* Change the current lcn to its new value. */
869 			lcn += deltaxcn;
870 #ifdef DEBUG
871 			/*
872 			 * On NTFS 1.2-, apparently can have lcn == -1 to
873 			 * indicate a hole. But we haven't verified ourselves
874 			 * whether it is really the lcn or the deltaxcn that is
875 			 * -1. So if either is found give us a message so we
876 			 * can investigate it further!
877 			 */
878 			if (vol->major_ver < 3) {
879 				if (unlikely(deltaxcn == (LCN)-1))
880 					ntfs_error(vol->sb, "lcn delta == -1");
881 				if (unlikely(lcn == (LCN)-1))
882 					ntfs_error(vol->sb, "lcn == -1");
883 			}
884 #endif
885 			/* Check lcn is not below -1. */
886 			if (unlikely(lcn < (LCN)-1)) {
887 				ntfs_error(vol->sb, "Invalid LCN < -1 in "
888 						"mapping pairs array.");
889 				goto err_out;
890 			}
891 			/* Enter the current lcn into the runlist element. */
892 			rl[rlpos].lcn = lcn;
893 		}
894 		/* Get to the next runlist element. */
895 		rlpos++;
896 		/* Increment the buffer position to the next mapping pair. */
897 		buf += (*buf & 0xf) + ((*buf >> 4) & 0xf) + 1;
898 	}
899 	if (unlikely(buf >= attr_end))
900 		goto io_error;
901 	/*
902 	 * If there is a highest_vcn specified, it must be equal to the final
903 	 * vcn in the runlist - 1, or something has gone badly wrong.
904 	 */
905 	deltaxcn = sle64_to_cpu(attr->data.non_resident.highest_vcn);
906 	if (unlikely(deltaxcn && vcn - 1 != deltaxcn)) {
907 mpa_err:
908 		ntfs_error(vol->sb, "Corrupt mapping pairs array in "
909 				"non-resident attribute.");
910 		goto err_out;
911 	}
912 	/* Setup not mapped runlist element if this is the base extent. */
913 	if (!attr->data.non_resident.lowest_vcn) {
914 		VCN max_cluster;
915 
916 		max_cluster = ((sle64_to_cpu(
917 				attr->data.non_resident.allocated_size) +
918 				vol->cluster_size - 1) >>
919 				vol->cluster_size_bits) - 1;
920 		/*
921 		 * A highest_vcn of zero means this is a single extent
922 		 * attribute so simply terminate the runlist with LCN_ENOENT).
923 		 */
924 		if (deltaxcn) {
925 			/*
926 			 * If there is a difference between the highest_vcn and
927 			 * the highest cluster, the runlist is either corrupt
928 			 * or, more likely, there are more extents following
929 			 * this one.
930 			 */
931 			if (deltaxcn < max_cluster) {
932 				ntfs_debug("More extents to follow; deltaxcn "
933 						"= 0x%llx, max_cluster = "
934 						"0x%llx",
935 						(unsigned long long)deltaxcn,
936 						(unsigned long long)
937 						max_cluster);
938 				rl[rlpos].vcn = vcn;
939 				vcn += rl[rlpos].length = max_cluster -
940 						deltaxcn;
941 				rl[rlpos].lcn = LCN_RL_NOT_MAPPED;
942 				rlpos++;
943 			} else if (unlikely(deltaxcn > max_cluster)) {
944 				ntfs_error(vol->sb, "Corrupt attribute.  "
945 						"deltaxcn = 0x%llx, "
946 						"max_cluster = 0x%llx",
947 						(unsigned long long)deltaxcn,
948 						(unsigned long long)
949 						max_cluster);
950 				goto mpa_err;
951 			}
952 		}
953 		rl[rlpos].lcn = LCN_ENOENT;
954 	} else /* Not the base extent. There may be more extents to follow. */
955 		rl[rlpos].lcn = LCN_RL_NOT_MAPPED;
956 
957 	/* Setup terminating runlist element. */
958 	rl[rlpos].vcn = vcn;
959 	rl[rlpos].length = (s64)0;
960 	/* If no existing runlist was specified, we are done. */
961 	if (!old_rl) {
962 		ntfs_debug("Mapping pairs array successfully decompressed:");
963 		ntfs_debug_dump_runlist(rl);
964 		return rl;
965 	}
966 	/* Now combine the new and old runlists checking for overlaps. */
967 	old_rl = ntfs_runlists_merge(old_rl, rl);
968 	if (likely(!IS_ERR(old_rl)))
969 		return old_rl;
970 	ntfs_free(rl);
971 	ntfs_error(vol->sb, "Failed to merge runlists.");
972 	return old_rl;
973 io_error:
974 	ntfs_error(vol->sb, "Corrupt attribute.");
975 err_out:
976 	ntfs_free(rl);
977 	return ERR_PTR(-EIO);
978 }
979 
980 /**
981  * ntfs_rl_vcn_to_lcn - convert a vcn into a lcn given a runlist
982  * @rl:		runlist to use for conversion
983  * @vcn:	vcn to convert
984  *
985  * Convert the virtual cluster number @vcn of an attribute into a logical
986  * cluster number (lcn) of a device using the runlist @rl to map vcns to their
987  * corresponding lcns.
988  *
989  * It is up to the caller to serialize access to the runlist @rl.
990  *
991  * Since lcns must be >= 0, we use negative return codes with special meaning:
992  *
993  * Return code		Meaning / Description
994  * ==================================================
995  *  LCN_HOLE		Hole / not allocated on disk.
996  *  LCN_RL_NOT_MAPPED	This is part of the runlist which has not been
997  *			inserted into the runlist yet.
998  *  LCN_ENOENT		There is no such vcn in the attribute.
999  *
1000  * Locking: - The caller must have locked the runlist (for reading or writing).
1001  *	    - This function does not touch the lock, nor does it modify the
1002  *	      runlist.
1003  */
1004 LCN ntfs_rl_vcn_to_lcn(const runlist_element *rl, const VCN vcn)
1005 {
1006 	int i;
1007 
1008 	BUG_ON(vcn < 0);
1009 	/*
1010 	 * If rl is NULL, assume that we have found an unmapped runlist. The
1011 	 * caller can then attempt to map it and fail appropriately if
1012 	 * necessary.
1013 	 */
1014 	if (unlikely(!rl))
1015 		return LCN_RL_NOT_MAPPED;
1016 
1017 	/* Catch out of lower bounds vcn. */
1018 	if (unlikely(vcn < rl[0].vcn))
1019 		return LCN_ENOENT;
1020 
1021 	for (i = 0; likely(rl[i].length); i++) {
1022 		if (unlikely(vcn < rl[i+1].vcn)) {
1023 			if (likely(rl[i].lcn >= (LCN)0))
1024 				return rl[i].lcn + (vcn - rl[i].vcn);
1025 			return rl[i].lcn;
1026 		}
1027 	}
1028 	/*
1029 	 * The terminator element is setup to the correct value, i.e. one of
1030 	 * LCN_HOLE, LCN_RL_NOT_MAPPED, or LCN_ENOENT.
1031 	 */
1032 	if (likely(rl[i].lcn < (LCN)0))
1033 		return rl[i].lcn;
1034 	/* Just in case... We could replace this with BUG() some day. */
1035 	return LCN_ENOENT;
1036 }
1037 
1038 #ifdef NTFS_RW
1039 
1040 /**
1041  * ntfs_rl_find_vcn_nolock - find a vcn in a runlist
1042  * @rl:		runlist to search
1043  * @vcn:	vcn to find
1044  *
1045  * Find the virtual cluster number @vcn in the runlist @rl and return the
1046  * address of the runlist element containing the @vcn on success.
1047  *
1048  * Return NULL if @rl is NULL or @vcn is in an unmapped part/out of bounds of
1049  * the runlist.
1050  *
1051  * Locking: The runlist must be locked on entry.
1052  */
1053 runlist_element *ntfs_rl_find_vcn_nolock(runlist_element *rl, const VCN vcn)
1054 {
1055 	BUG_ON(vcn < 0);
1056 	if (unlikely(!rl || vcn < rl[0].vcn))
1057 		return NULL;
1058 	while (likely(rl->length)) {
1059 		if (unlikely(vcn < rl[1].vcn)) {
1060 			if (likely(rl->lcn >= LCN_HOLE))
1061 				return rl;
1062 			return NULL;
1063 		}
1064 		rl++;
1065 	}
1066 	if (likely(rl->lcn == LCN_ENOENT))
1067 		return rl;
1068 	return NULL;
1069 }
1070 
1071 /**
1072  * ntfs_get_nr_significant_bytes - get number of bytes needed to store a number
1073  * @n:		number for which to get the number of bytes for
1074  *
1075  * Return the number of bytes required to store @n unambiguously as
1076  * a signed number.
1077  *
1078  * This is used in the context of the mapping pairs array to determine how
1079  * many bytes will be needed in the array to store a given logical cluster
1080  * number (lcn) or a specific run length.
1081  *
1082  * Return the number of bytes written.  This function cannot fail.
1083  */
1084 static inline int ntfs_get_nr_significant_bytes(const s64 n)
1085 {
1086 	s64 l = n;
1087 	int i;
1088 	s8 j;
1089 
1090 	i = 0;
1091 	do {
1092 		l >>= 8;
1093 		i++;
1094 	} while (l != 0 && l != -1);
1095 	j = (n >> 8 * (i - 1)) & 0xff;
1096 	/* If the sign bit is wrong, we need an extra byte. */
1097 	if ((n < 0 && j >= 0) || (n > 0 && j < 0))
1098 		i++;
1099 	return i;
1100 }
1101 
1102 /**
1103  * ntfs_get_size_for_mapping_pairs - get bytes needed for mapping pairs array
1104  * @vol:	ntfs volume (needed for the ntfs version)
1105  * @rl:		locked runlist to determine the size of the mapping pairs of
1106  * @first_vcn:	first vcn which to include in the mapping pairs array
1107  * @last_vcn:	last vcn which to include in the mapping pairs array
1108  *
1109  * Walk the locked runlist @rl and calculate the size in bytes of the mapping
1110  * pairs array corresponding to the runlist @rl, starting at vcn @first_vcn and
1111  * finishing with vcn @last_vcn.
1112  *
1113  * A @last_vcn of -1 means end of runlist and in that case the size of the
1114  * mapping pairs array corresponding to the runlist starting at vcn @first_vcn
1115  * and finishing at the end of the runlist is determined.
1116  *
1117  * This for example allows us to allocate a buffer of the right size when
1118  * building the mapping pairs array.
1119  *
1120  * If @rl is NULL, just return 1 (for the single terminator byte).
1121  *
1122  * Return the calculated size in bytes on success.  On error, return -errno.
1123  * The following error codes are defined:
1124  *	-EINVAL	- Run list contains unmapped elements.  Make sure to only pass
1125  *		  fully mapped runlists to this function.
1126  *	-EIO	- The runlist is corrupt.
1127  *
1128  * Locking: @rl must be locked on entry (either for reading or writing), it
1129  *	    remains locked throughout, and is left locked upon return.
1130  */
1131 int ntfs_get_size_for_mapping_pairs(const ntfs_volume *vol,
1132 		const runlist_element *rl, const VCN first_vcn,
1133 		const VCN last_vcn)
1134 {
1135 	LCN prev_lcn;
1136 	int rls;
1137 	bool the_end = false;
1138 
1139 	BUG_ON(first_vcn < 0);
1140 	BUG_ON(last_vcn < -1);
1141 	BUG_ON(last_vcn >= 0 && first_vcn > last_vcn);
1142 	if (!rl) {
1143 		BUG_ON(first_vcn);
1144 		BUG_ON(last_vcn > 0);
1145 		return 1;
1146 	}
1147 	/* Skip to runlist element containing @first_vcn. */
1148 	while (rl->length && first_vcn >= rl[1].vcn)
1149 		rl++;
1150 	if (unlikely((!rl->length && first_vcn > rl->vcn) ||
1151 			first_vcn < rl->vcn))
1152 		return -EINVAL;
1153 	prev_lcn = 0;
1154 	/* Always need the termining zero byte. */
1155 	rls = 1;
1156 	/* Do the first partial run if present. */
1157 	if (first_vcn > rl->vcn) {
1158 		s64 delta, length = rl->length;
1159 
1160 		/* We know rl->length != 0 already. */
1161 		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1162 			goto err_out;
1163 		/*
1164 		 * If @stop_vcn is given and finishes inside this run, cap the
1165 		 * run length.
1166 		 */
1167 		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1168 			s64 s1 = last_vcn + 1;
1169 			if (unlikely(rl[1].vcn > s1))
1170 				length = s1 - rl->vcn;
1171 			the_end = true;
1172 		}
1173 		delta = first_vcn - rl->vcn;
1174 		/* Header byte + length. */
1175 		rls += 1 + ntfs_get_nr_significant_bytes(length - delta);
1176 		/*
1177 		 * If the logical cluster number (lcn) denotes a hole and we
1178 		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1179 		 * zero space.  On earlier NTFS versions we just store the lcn.
1180 		 * Note: this assumes that on NTFS 1.2-, holes are stored with
1181 		 * an lcn of -1 and not a delta_lcn of -1 (unless both are -1).
1182 		 */
1183 		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1184 			prev_lcn = rl->lcn;
1185 			if (likely(rl->lcn >= 0))
1186 				prev_lcn += delta;
1187 			/* Change in lcn. */
1188 			rls += ntfs_get_nr_significant_bytes(prev_lcn);
1189 		}
1190 		/* Go to next runlist element. */
1191 		rl++;
1192 	}
1193 	/* Do the full runs. */
1194 	for (; rl->length && !the_end; rl++) {
1195 		s64 length = rl->length;
1196 
1197 		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1198 			goto err_out;
1199 		/*
1200 		 * If @stop_vcn is given and finishes inside this run, cap the
1201 		 * run length.
1202 		 */
1203 		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1204 			s64 s1 = last_vcn + 1;
1205 			if (unlikely(rl[1].vcn > s1))
1206 				length = s1 - rl->vcn;
1207 			the_end = true;
1208 		}
1209 		/* Header byte + length. */
1210 		rls += 1 + ntfs_get_nr_significant_bytes(length);
1211 		/*
1212 		 * If the logical cluster number (lcn) denotes a hole and we
1213 		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1214 		 * zero space.  On earlier NTFS versions we just store the lcn.
1215 		 * Note: this assumes that on NTFS 1.2-, holes are stored with
1216 		 * an lcn of -1 and not a delta_lcn of -1 (unless both are -1).
1217 		 */
1218 		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1219 			/* Change in lcn. */
1220 			rls += ntfs_get_nr_significant_bytes(rl->lcn -
1221 					prev_lcn);
1222 			prev_lcn = rl->lcn;
1223 		}
1224 	}
1225 	return rls;
1226 err_out:
1227 	if (rl->lcn == LCN_RL_NOT_MAPPED)
1228 		rls = -EINVAL;
1229 	else
1230 		rls = -EIO;
1231 	return rls;
1232 }
1233 
1234 /**
1235  * ntfs_write_significant_bytes - write the significant bytes of a number
1236  * @dst:	destination buffer to write to
1237  * @dst_max:	pointer to last byte of destination buffer for bounds checking
1238  * @n:		number whose significant bytes to write
1239  *
1240  * Store in @dst, the minimum bytes of the number @n which are required to
1241  * identify @n unambiguously as a signed number, taking care not to exceed
1242  * @dest_max, the maximum position within @dst to which we are allowed to
1243  * write.
1244  *
1245  * This is used when building the mapping pairs array of a runlist to compress
1246  * a given logical cluster number (lcn) or a specific run length to the minimum
1247  * size possible.
1248  *
1249  * Return the number of bytes written on success.  On error, i.e. the
1250  * destination buffer @dst is too small, return -ENOSPC.
1251  */
1252 static inline int ntfs_write_significant_bytes(s8 *dst, const s8 *dst_max,
1253 		const s64 n)
1254 {
1255 	s64 l = n;
1256 	int i;
1257 	s8 j;
1258 
1259 	i = 0;
1260 	do {
1261 		if (unlikely(dst > dst_max))
1262 			goto err_out;
1263 		*dst++ = l & 0xffll;
1264 		l >>= 8;
1265 		i++;
1266 	} while (l != 0 && l != -1);
1267 	j = (n >> 8 * (i - 1)) & 0xff;
1268 	/* If the sign bit is wrong, we need an extra byte. */
1269 	if (n < 0 && j >= 0) {
1270 		if (unlikely(dst > dst_max))
1271 			goto err_out;
1272 		i++;
1273 		*dst = (s8)-1;
1274 	} else if (n > 0 && j < 0) {
1275 		if (unlikely(dst > dst_max))
1276 			goto err_out;
1277 		i++;
1278 		*dst = (s8)0;
1279 	}
1280 	return i;
1281 err_out:
1282 	return -ENOSPC;
1283 }
1284 
1285 /**
1286  * ntfs_mapping_pairs_build - build the mapping pairs array from a runlist
1287  * @vol:	ntfs volume (needed for the ntfs version)
1288  * @dst:	destination buffer to which to write the mapping pairs array
1289  * @dst_len:	size of destination buffer @dst in bytes
1290  * @rl:		locked runlist for which to build the mapping pairs array
1291  * @first_vcn:	first vcn which to include in the mapping pairs array
1292  * @last_vcn:	last vcn which to include in the mapping pairs array
1293  * @stop_vcn:	first vcn outside destination buffer on success or -ENOSPC
1294  *
1295  * Create the mapping pairs array from the locked runlist @rl, starting at vcn
1296  * @first_vcn and finishing with vcn @last_vcn and save the array in @dst.
1297  * @dst_len is the size of @dst in bytes and it should be at least equal to the
1298  * value obtained by calling ntfs_get_size_for_mapping_pairs().
1299  *
1300  * A @last_vcn of -1 means end of runlist and in that case the mapping pairs
1301  * array corresponding to the runlist starting at vcn @first_vcn and finishing
1302  * at the end of the runlist is created.
1303  *
1304  * If @rl is NULL, just write a single terminator byte to @dst.
1305  *
1306  * On success or -ENOSPC error, if @stop_vcn is not NULL, *@stop_vcn is set to
1307  * the first vcn outside the destination buffer.  Note that on error, @dst has
1308  * been filled with all the mapping pairs that will fit, thus it can be treated
1309  * as partial success, in that a new attribute extent needs to be created or
1310  * the next extent has to be used and the mapping pairs build has to be
1311  * continued with @first_vcn set to *@stop_vcn.
1312  *
1313  * Return 0 on success and -errno on error.  The following error codes are
1314  * defined:
1315  *	-EINVAL	- Run list contains unmapped elements.  Make sure to only pass
1316  *		  fully mapped runlists to this function.
1317  *	-EIO	- The runlist is corrupt.
1318  *	-ENOSPC	- The destination buffer is too small.
1319  *
1320  * Locking: @rl must be locked on entry (either for reading or writing), it
1321  *	    remains locked throughout, and is left locked upon return.
1322  */
1323 int ntfs_mapping_pairs_build(const ntfs_volume *vol, s8 *dst,
1324 		const int dst_len, const runlist_element *rl,
1325 		const VCN first_vcn, const VCN last_vcn, VCN *const stop_vcn)
1326 {
1327 	LCN prev_lcn;
1328 	s8 *dst_max, *dst_next;
1329 	int err = -ENOSPC;
1330 	bool the_end = false;
1331 	s8 len_len, lcn_len;
1332 
1333 	BUG_ON(first_vcn < 0);
1334 	BUG_ON(last_vcn < -1);
1335 	BUG_ON(last_vcn >= 0 && first_vcn > last_vcn);
1336 	BUG_ON(dst_len < 1);
1337 	if (!rl) {
1338 		BUG_ON(first_vcn);
1339 		BUG_ON(last_vcn > 0);
1340 		if (stop_vcn)
1341 			*stop_vcn = 0;
1342 		/* Terminator byte. */
1343 		*dst = 0;
1344 		return 0;
1345 	}
1346 	/* Skip to runlist element containing @first_vcn. */
1347 	while (rl->length && first_vcn >= rl[1].vcn)
1348 		rl++;
1349 	if (unlikely((!rl->length && first_vcn > rl->vcn) ||
1350 			first_vcn < rl->vcn))
1351 		return -EINVAL;
1352 	/*
1353 	 * @dst_max is used for bounds checking in
1354 	 * ntfs_write_significant_bytes().
1355 	 */
1356 	dst_max = dst + dst_len - 1;
1357 	prev_lcn = 0;
1358 	/* Do the first partial run if present. */
1359 	if (first_vcn > rl->vcn) {
1360 		s64 delta, length = rl->length;
1361 
1362 		/* We know rl->length != 0 already. */
1363 		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1364 			goto err_out;
1365 		/*
1366 		 * If @stop_vcn is given and finishes inside this run, cap the
1367 		 * run length.
1368 		 */
1369 		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1370 			s64 s1 = last_vcn + 1;
1371 			if (unlikely(rl[1].vcn > s1))
1372 				length = s1 - rl->vcn;
1373 			the_end = true;
1374 		}
1375 		delta = first_vcn - rl->vcn;
1376 		/* Write length. */
1377 		len_len = ntfs_write_significant_bytes(dst + 1, dst_max,
1378 				length - delta);
1379 		if (unlikely(len_len < 0))
1380 			goto size_err;
1381 		/*
1382 		 * If the logical cluster number (lcn) denotes a hole and we
1383 		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1384 		 * zero space.  On earlier NTFS versions we just write the lcn
1385 		 * change.  FIXME: Do we need to write the lcn change or just
1386 		 * the lcn in that case?  Not sure as I have never seen this
1387 		 * case on NT4. - We assume that we just need to write the lcn
1388 		 * change until someone tells us otherwise... (AIA)
1389 		 */
1390 		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1391 			prev_lcn = rl->lcn;
1392 			if (likely(rl->lcn >= 0))
1393 				prev_lcn += delta;
1394 			/* Write change in lcn. */
1395 			lcn_len = ntfs_write_significant_bytes(dst + 1 +
1396 					len_len, dst_max, prev_lcn);
1397 			if (unlikely(lcn_len < 0))
1398 				goto size_err;
1399 		} else
1400 			lcn_len = 0;
1401 		dst_next = dst + len_len + lcn_len + 1;
1402 		if (unlikely(dst_next > dst_max))
1403 			goto size_err;
1404 		/* Update header byte. */
1405 		*dst = lcn_len << 4 | len_len;
1406 		/* Position at next mapping pairs array element. */
1407 		dst = dst_next;
1408 		/* Go to next runlist element. */
1409 		rl++;
1410 	}
1411 	/* Do the full runs. */
1412 	for (; rl->length && !the_end; rl++) {
1413 		s64 length = rl->length;
1414 
1415 		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1416 			goto err_out;
1417 		/*
1418 		 * If @stop_vcn is given and finishes inside this run, cap the
1419 		 * run length.
1420 		 */
1421 		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1422 			s64 s1 = last_vcn + 1;
1423 			if (unlikely(rl[1].vcn > s1))
1424 				length = s1 - rl->vcn;
1425 			the_end = true;
1426 		}
1427 		/* Write length. */
1428 		len_len = ntfs_write_significant_bytes(dst + 1, dst_max,
1429 				length);
1430 		if (unlikely(len_len < 0))
1431 			goto size_err;
1432 		/*
1433 		 * If the logical cluster number (lcn) denotes a hole and we
1434 		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1435 		 * zero space.  On earlier NTFS versions we just write the lcn
1436 		 * change.  FIXME: Do we need to write the lcn change or just
1437 		 * the lcn in that case?  Not sure as I have never seen this
1438 		 * case on NT4. - We assume that we just need to write the lcn
1439 		 * change until someone tells us otherwise... (AIA)
1440 		 */
1441 		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1442 			/* Write change in lcn. */
1443 			lcn_len = ntfs_write_significant_bytes(dst + 1 +
1444 					len_len, dst_max, rl->lcn - prev_lcn);
1445 			if (unlikely(lcn_len < 0))
1446 				goto size_err;
1447 			prev_lcn = rl->lcn;
1448 		} else
1449 			lcn_len = 0;
1450 		dst_next = dst + len_len + lcn_len + 1;
1451 		if (unlikely(dst_next > dst_max))
1452 			goto size_err;
1453 		/* Update header byte. */
1454 		*dst = lcn_len << 4 | len_len;
1455 		/* Position at next mapping pairs array element. */
1456 		dst = dst_next;
1457 	}
1458 	/* Success. */
1459 	err = 0;
1460 size_err:
1461 	/* Set stop vcn. */
1462 	if (stop_vcn)
1463 		*stop_vcn = rl->vcn;
1464 	/* Add terminator byte. */
1465 	*dst = 0;
1466 	return err;
1467 err_out:
1468 	if (rl->lcn == LCN_RL_NOT_MAPPED)
1469 		err = -EINVAL;
1470 	else
1471 		err = -EIO;
1472 	return err;
1473 }
1474 
1475 /**
1476  * ntfs_rl_truncate_nolock - truncate a runlist starting at a specified vcn
1477  * @vol:	ntfs volume (needed for error output)
1478  * @runlist:	runlist to truncate
1479  * @new_length:	the new length of the runlist in VCNs
1480  *
1481  * Truncate the runlist described by @runlist as well as the memory buffer
1482  * holding the runlist elements to a length of @new_length VCNs.
1483  *
1484  * If @new_length lies within the runlist, the runlist elements with VCNs of
1485  * @new_length and above are discarded.  As a special case if @new_length is
1486  * zero, the runlist is discarded and set to NULL.
1487  *
1488  * If @new_length lies beyond the runlist, a sparse runlist element is added to
1489  * the end of the runlist @runlist or if the last runlist element is a sparse
1490  * one already, this is extended.
1491  *
1492  * Note, no checking is done for unmapped runlist elements.  It is assumed that
1493  * the caller has mapped any elements that need to be mapped already.
1494  *
1495  * Return 0 on success and -errno on error.
1496  *
1497  * Locking: The caller must hold @runlist->lock for writing.
1498  */
1499 int ntfs_rl_truncate_nolock(const ntfs_volume *vol, runlist *const runlist,
1500 		const s64 new_length)
1501 {
1502 	runlist_element *rl;
1503 	int old_size;
1504 
1505 	ntfs_debug("Entering for new_length 0x%llx.", (long long)new_length);
1506 	BUG_ON(!runlist);
1507 	BUG_ON(new_length < 0);
1508 	rl = runlist->rl;
1509 	if (!new_length) {
1510 		ntfs_debug("Freeing runlist.");
1511 		runlist->rl = NULL;
1512 		if (rl)
1513 			ntfs_free(rl);
1514 		return 0;
1515 	}
1516 	if (unlikely(!rl)) {
1517 		/*
1518 		 * Create a runlist consisting of a sparse runlist element of
1519 		 * length @new_length followed by a terminator runlist element.
1520 		 */
1521 		rl = ntfs_malloc_nofs(PAGE_SIZE);
1522 		if (unlikely(!rl)) {
1523 			ntfs_error(vol->sb, "Not enough memory to allocate "
1524 					"runlist element buffer.");
1525 			return -ENOMEM;
1526 		}
1527 		runlist->rl = rl;
1528 		rl[1].length = rl->vcn = 0;
1529 		rl->lcn = LCN_HOLE;
1530 		rl[1].vcn = rl->length = new_length;
1531 		rl[1].lcn = LCN_ENOENT;
1532 		return 0;
1533 	}
1534 	BUG_ON(new_length < rl->vcn);
1535 	/* Find @new_length in the runlist. */
1536 	while (likely(rl->length && new_length >= rl[1].vcn))
1537 		rl++;
1538 	/*
1539 	 * If not at the end of the runlist we need to shrink it.
1540 	 * If at the end of the runlist we need to expand it.
1541 	 */
1542 	if (rl->length) {
1543 		runlist_element *trl;
1544 		bool is_end;
1545 
1546 		ntfs_debug("Shrinking runlist.");
1547 		/* Determine the runlist size. */
1548 		trl = rl + 1;
1549 		while (likely(trl->length))
1550 			trl++;
1551 		old_size = trl - runlist->rl + 1;
1552 		/* Truncate the run. */
1553 		rl->length = new_length - rl->vcn;
1554 		/*
1555 		 * If a run was partially truncated, make the following runlist
1556 		 * element a terminator.
1557 		 */
1558 		is_end = false;
1559 		if (rl->length) {
1560 			rl++;
1561 			if (!rl->length)
1562 				is_end = true;
1563 			rl->vcn = new_length;
1564 			rl->length = 0;
1565 		}
1566 		rl->lcn = LCN_ENOENT;
1567 		/* Reallocate memory if necessary. */
1568 		if (!is_end) {
1569 			int new_size = rl - runlist->rl + 1;
1570 			rl = ntfs_rl_realloc(runlist->rl, old_size, new_size);
1571 			if (IS_ERR(rl))
1572 				ntfs_warning(vol->sb, "Failed to shrink "
1573 						"runlist buffer.  This just "
1574 						"wastes a bit of memory "
1575 						"temporarily so we ignore it "
1576 						"and return success.");
1577 			else
1578 				runlist->rl = rl;
1579 		}
1580 	} else if (likely(/* !rl->length && */ new_length > rl->vcn)) {
1581 		ntfs_debug("Expanding runlist.");
1582 		/*
1583 		 * If there is a previous runlist element and it is a sparse
1584 		 * one, extend it.  Otherwise need to add a new, sparse runlist
1585 		 * element.
1586 		 */
1587 		if ((rl > runlist->rl) && ((rl - 1)->lcn == LCN_HOLE))
1588 			(rl - 1)->length = new_length - (rl - 1)->vcn;
1589 		else {
1590 			/* Determine the runlist size. */
1591 			old_size = rl - runlist->rl + 1;
1592 			/* Reallocate memory if necessary. */
1593 			rl = ntfs_rl_realloc(runlist->rl, old_size,
1594 					old_size + 1);
1595 			if (IS_ERR(rl)) {
1596 				ntfs_error(vol->sb, "Failed to expand runlist "
1597 						"buffer, aborting.");
1598 				return PTR_ERR(rl);
1599 			}
1600 			runlist->rl = rl;
1601 			/*
1602 			 * Set @rl to the same runlist element in the new
1603 			 * runlist as before in the old runlist.
1604 			 */
1605 			rl += old_size - 1;
1606 			/* Add a new, sparse runlist element. */
1607 			rl->lcn = LCN_HOLE;
1608 			rl->length = new_length - rl->vcn;
1609 			/* Add a new terminator runlist element. */
1610 			rl++;
1611 			rl->length = 0;
1612 		}
1613 		rl->vcn = new_length;
1614 		rl->lcn = LCN_ENOENT;
1615 	} else /* if (unlikely(!rl->length && new_length == rl->vcn)) */ {
1616 		/* Runlist already has same size as requested. */
1617 		rl->lcn = LCN_ENOENT;
1618 	}
1619 	ntfs_debug("Done.");
1620 	return 0;
1621 }
1622 
1623 /**
1624  * ntfs_rl_punch_nolock - punch a hole into a runlist
1625  * @vol:	ntfs volume (needed for error output)
1626  * @runlist:	runlist to punch a hole into
1627  * @start:	starting VCN of the hole to be created
1628  * @length:	size of the hole to be created in units of clusters
1629  *
1630  * Punch a hole into the runlist @runlist starting at VCN @start and of size
1631  * @length clusters.
1632  *
1633  * Return 0 on success and -errno on error, in which case @runlist has not been
1634  * modified.
1635  *
1636  * If @start and/or @start + @length are outside the runlist return error code
1637  * -ENOENT.
1638  *
1639  * If the runlist contains unmapped or error elements between @start and @start
1640  * + @length return error code -EINVAL.
1641  *
1642  * Locking: The caller must hold @runlist->lock for writing.
1643  */
1644 int ntfs_rl_punch_nolock(const ntfs_volume *vol, runlist *const runlist,
1645 		const VCN start, const s64 length)
1646 {
1647 	const VCN end = start + length;
1648 	s64 delta;
1649 	runlist_element *rl, *rl_end, *rl_real_end, *trl;
1650 	int old_size;
1651 	bool lcn_fixup = false;
1652 
1653 	ntfs_debug("Entering for start 0x%llx, length 0x%llx.",
1654 			(long long)start, (long long)length);
1655 	BUG_ON(!runlist);
1656 	BUG_ON(start < 0);
1657 	BUG_ON(length < 0);
1658 	BUG_ON(end < 0);
1659 	rl = runlist->rl;
1660 	if (unlikely(!rl)) {
1661 		if (likely(!start && !length))
1662 			return 0;
1663 		return -ENOENT;
1664 	}
1665 	/* Find @start in the runlist. */
1666 	while (likely(rl->length && start >= rl[1].vcn))
1667 		rl++;
1668 	rl_end = rl;
1669 	/* Find @end in the runlist. */
1670 	while (likely(rl_end->length && end >= rl_end[1].vcn)) {
1671 		/* Verify there are no unmapped or error elements. */
1672 		if (unlikely(rl_end->lcn < LCN_HOLE))
1673 			return -EINVAL;
1674 		rl_end++;
1675 	}
1676 	/* Check the last element. */
1677 	if (unlikely(rl_end->length && rl_end->lcn < LCN_HOLE))
1678 		return -EINVAL;
1679 	/* This covers @start being out of bounds, too. */
1680 	if (!rl_end->length && end > rl_end->vcn)
1681 		return -ENOENT;
1682 	if (!length)
1683 		return 0;
1684 	if (!rl->length)
1685 		return -ENOENT;
1686 	rl_real_end = rl_end;
1687 	/* Determine the runlist size. */
1688 	while (likely(rl_real_end->length))
1689 		rl_real_end++;
1690 	old_size = rl_real_end - runlist->rl + 1;
1691 	/* If @start is in a hole simply extend the hole. */
1692 	if (rl->lcn == LCN_HOLE) {
1693 		/*
1694 		 * If both @start and @end are in the same sparse run, we are
1695 		 * done.
1696 		 */
1697 		if (end <= rl[1].vcn) {
1698 			ntfs_debug("Done (requested hole is already sparse).");
1699 			return 0;
1700 		}
1701 extend_hole:
1702 		/* Extend the hole. */
1703 		rl->length = end - rl->vcn;
1704 		/* If @end is in a hole, merge it with the current one. */
1705 		if (rl_end->lcn == LCN_HOLE) {
1706 			rl_end++;
1707 			rl->length = rl_end->vcn - rl->vcn;
1708 		}
1709 		/* We have done the hole.  Now deal with the remaining tail. */
1710 		rl++;
1711 		/* Cut out all runlist elements up to @end. */
1712 		if (rl < rl_end)
1713 			memmove(rl, rl_end, (rl_real_end - rl_end + 1) *
1714 					sizeof(*rl));
1715 		/* Adjust the beginning of the tail if necessary. */
1716 		if (end > rl->vcn) {
1717 			delta = end - rl->vcn;
1718 			rl->vcn = end;
1719 			rl->length -= delta;
1720 			/* Only adjust the lcn if it is real. */
1721 			if (rl->lcn >= 0)
1722 				rl->lcn += delta;
1723 		}
1724 shrink_allocation:
1725 		/* Reallocate memory if the allocation changed. */
1726 		if (rl < rl_end) {
1727 			rl = ntfs_rl_realloc(runlist->rl, old_size,
1728 					old_size - (rl_end - rl));
1729 			if (IS_ERR(rl))
1730 				ntfs_warning(vol->sb, "Failed to shrink "
1731 						"runlist buffer.  This just "
1732 						"wastes a bit of memory "
1733 						"temporarily so we ignore it "
1734 						"and return success.");
1735 			else
1736 				runlist->rl = rl;
1737 		}
1738 		ntfs_debug("Done (extend hole).");
1739 		return 0;
1740 	}
1741 	/*
1742 	 * If @start is at the beginning of a run things are easier as there is
1743 	 * no need to split the first run.
1744 	 */
1745 	if (start == rl->vcn) {
1746 		/*
1747 		 * @start is at the beginning of a run.
1748 		 *
1749 		 * If the previous run is sparse, extend its hole.
1750 		 *
1751 		 * If @end is not in the same run, switch the run to be sparse
1752 		 * and extend the newly created hole.
1753 		 *
1754 		 * Thus both of these cases reduce the problem to the above
1755 		 * case of "@start is in a hole".
1756 		 */
1757 		if (rl > runlist->rl && (rl - 1)->lcn == LCN_HOLE) {
1758 			rl--;
1759 			goto extend_hole;
1760 		}
1761 		if (end >= rl[1].vcn) {
1762 			rl->lcn = LCN_HOLE;
1763 			goto extend_hole;
1764 		}
1765 		/*
1766 		 * The final case is when @end is in the same run as @start.
1767 		 * For this need to split the run into two.  One run for the
1768 		 * sparse region between the beginning of the old run, i.e.
1769 		 * @start, and @end and one for the remaining non-sparse
1770 		 * region, i.e. between @end and the end of the old run.
1771 		 */
1772 		trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 1);
1773 		if (IS_ERR(trl))
1774 			goto enomem_out;
1775 		old_size++;
1776 		if (runlist->rl != trl) {
1777 			rl = trl + (rl - runlist->rl);
1778 			rl_end = trl + (rl_end - runlist->rl);
1779 			rl_real_end = trl + (rl_real_end - runlist->rl);
1780 			runlist->rl = trl;
1781 		}
1782 split_end:
1783 		/* Shift all the runs up by one. */
1784 		memmove(rl + 1, rl, (rl_real_end - rl + 1) * sizeof(*rl));
1785 		/* Finally, setup the two split runs. */
1786 		rl->lcn = LCN_HOLE;
1787 		rl->length = length;
1788 		rl++;
1789 		rl->vcn += length;
1790 		/* Only adjust the lcn if it is real. */
1791 		if (rl->lcn >= 0 || lcn_fixup)
1792 			rl->lcn += length;
1793 		rl->length -= length;
1794 		ntfs_debug("Done (split one).");
1795 		return 0;
1796 	}
1797 	/*
1798 	 * @start is neither in a hole nor at the beginning of a run.
1799 	 *
1800 	 * If @end is in a hole, things are easier as simply truncating the run
1801 	 * @start is in to end at @start - 1, deleting all runs after that up
1802 	 * to @end, and finally extending the beginning of the run @end is in
1803 	 * to be @start is all that is needed.
1804 	 */
1805 	if (rl_end->lcn == LCN_HOLE) {
1806 		/* Truncate the run containing @start. */
1807 		rl->length = start - rl->vcn;
1808 		rl++;
1809 		/* Cut out all runlist elements up to @end. */
1810 		if (rl < rl_end)
1811 			memmove(rl, rl_end, (rl_real_end - rl_end + 1) *
1812 					sizeof(*rl));
1813 		/* Extend the beginning of the run @end is in to be @start. */
1814 		rl->vcn = start;
1815 		rl->length = rl[1].vcn - start;
1816 		goto shrink_allocation;
1817 	}
1818 	/*
1819 	 * If @end is not in a hole there are still two cases to distinguish.
1820 	 * Either @end is or is not in the same run as @start.
1821 	 *
1822 	 * The second case is easier as it can be reduced to an already solved
1823 	 * problem by truncating the run @start is in to end at @start - 1.
1824 	 * Then, if @end is in the next run need to split the run into a sparse
1825 	 * run followed by a non-sparse run (already covered above) and if @end
1826 	 * is not in the next run switching it to be sparse, again reduces the
1827 	 * problem to the already covered case of "@start is in a hole".
1828 	 */
1829 	if (end >= rl[1].vcn) {
1830 		/*
1831 		 * If @end is not in the next run, reduce the problem to the
1832 		 * case of "@start is in a hole".
1833 		 */
1834 		if (rl[1].length && end >= rl[2].vcn) {
1835 			/* Truncate the run containing @start. */
1836 			rl->length = start - rl->vcn;
1837 			rl++;
1838 			rl->vcn = start;
1839 			rl->lcn = LCN_HOLE;
1840 			goto extend_hole;
1841 		}
1842 		trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 1);
1843 		if (IS_ERR(trl))
1844 			goto enomem_out;
1845 		old_size++;
1846 		if (runlist->rl != trl) {
1847 			rl = trl + (rl - runlist->rl);
1848 			rl_end = trl + (rl_end - runlist->rl);
1849 			rl_real_end = trl + (rl_real_end - runlist->rl);
1850 			runlist->rl = trl;
1851 		}
1852 		/* Truncate the run containing @start. */
1853 		rl->length = start - rl->vcn;
1854 		rl++;
1855 		/*
1856 		 * @end is in the next run, reduce the problem to the case
1857 		 * where "@start is at the beginning of a run and @end is in
1858 		 * the same run as @start".
1859 		 */
1860 		delta = rl->vcn - start;
1861 		rl->vcn = start;
1862 		if (rl->lcn >= 0) {
1863 			rl->lcn -= delta;
1864 			/* Need this in case the lcn just became negative. */
1865 			lcn_fixup = true;
1866 		}
1867 		rl->length += delta;
1868 		goto split_end;
1869 	}
1870 	/*
1871 	 * The first case from above, i.e. @end is in the same run as @start.
1872 	 * We need to split the run into three.  One run for the non-sparse
1873 	 * region between the beginning of the old run and @start, one for the
1874 	 * sparse region between @start and @end, and one for the remaining
1875 	 * non-sparse region, i.e. between @end and the end of the old run.
1876 	 */
1877 	trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 2);
1878 	if (IS_ERR(trl))
1879 		goto enomem_out;
1880 	old_size += 2;
1881 	if (runlist->rl != trl) {
1882 		rl = trl + (rl - runlist->rl);
1883 		rl_end = trl + (rl_end - runlist->rl);
1884 		rl_real_end = trl + (rl_real_end - runlist->rl);
1885 		runlist->rl = trl;
1886 	}
1887 	/* Shift all the runs up by two. */
1888 	memmove(rl + 2, rl, (rl_real_end - rl + 1) * sizeof(*rl));
1889 	/* Finally, setup the three split runs. */
1890 	rl->length = start - rl->vcn;
1891 	rl++;
1892 	rl->vcn = start;
1893 	rl->lcn = LCN_HOLE;
1894 	rl->length = length;
1895 	rl++;
1896 	delta = end - rl->vcn;
1897 	rl->vcn = end;
1898 	rl->lcn += delta;
1899 	rl->length -= delta;
1900 	ntfs_debug("Done (split both).");
1901 	return 0;
1902 enomem_out:
1903 	ntfs_error(vol->sb, "Not enough memory to extend runlist buffer.");
1904 	return -ENOMEM;
1905 }
1906 
1907 #endif /* NTFS_RW */
1908