Linux Audio

Check our new training course

Loading...
Note: File does not exist in v6.13.7.
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/**
   3 * runlist.c - NTFS runlist handling code.  Part of the Linux-NTFS project.
   4 *
   5 * Copyright (c) 2001-2007 Anton Altaparmakov
   6 * Copyright (c) 2002-2005 Richard Russon
   7 */
   8
   9#include "debug.h"
  10#include "dir.h"
  11#include "endian.h"
  12#include "malloc.h"
  13#include "ntfs.h"
  14
  15/**
  16 * ntfs_rl_mm - runlist memmove
  17 *
  18 * It is up to the caller to serialize access to the runlist @base.
  19 */
  20static inline void ntfs_rl_mm(runlist_element *base, int dst, int src,
  21		int size)
  22{
  23	if (likely((dst != src) && (size > 0)))
  24		memmove(base + dst, base + src, size * sizeof(*base));
  25}
  26
  27/**
  28 * ntfs_rl_mc - runlist memory copy
  29 *
  30 * It is up to the caller to serialize access to the runlists @dstbase and
  31 * @srcbase.
  32 */
  33static inline void ntfs_rl_mc(runlist_element *dstbase, int dst,
  34		runlist_element *srcbase, int src, int size)
  35{
  36	if (likely(size > 0))
  37		memcpy(dstbase + dst, srcbase + src, size * sizeof(*dstbase));
  38}
  39
  40/**
  41 * ntfs_rl_realloc - Reallocate memory for runlists
  42 * @rl:		original runlist
  43 * @old_size:	number of runlist elements in the original runlist @rl
  44 * @new_size:	number of runlist elements we need space for
  45 *
  46 * As the runlists grow, more memory will be required.  To prevent the
  47 * kernel having to allocate and reallocate large numbers of small bits of
  48 * memory, this function returns an entire page of memory.
  49 *
  50 * It is up to the caller to serialize access to the runlist @rl.
  51 *
  52 * N.B.  If the new allocation doesn't require a different number of pages in
  53 *       memory, the function will return the original pointer.
  54 *
  55 * On success, return a pointer to the newly allocated, or recycled, memory.
  56 * On error, return -errno. The following error codes are defined:
  57 *	-ENOMEM	- Not enough memory to allocate runlist array.
  58 *	-EINVAL	- Invalid parameters were passed in.
  59 */
  60static inline runlist_element *ntfs_rl_realloc(runlist_element *rl,
  61		int old_size, int new_size)
  62{
  63	runlist_element *new_rl;
  64
  65	old_size = PAGE_ALIGN(old_size * sizeof(*rl));
  66	new_size = PAGE_ALIGN(new_size * sizeof(*rl));
  67	if (old_size == new_size)
  68		return rl;
  69
  70	new_rl = ntfs_malloc_nofs(new_size);
  71	if (unlikely(!new_rl))
  72		return ERR_PTR(-ENOMEM);
  73
  74	if (likely(rl != NULL)) {
  75		if (unlikely(old_size > new_size))
  76			old_size = new_size;
  77		memcpy(new_rl, rl, old_size);
  78		ntfs_free(rl);
  79	}
  80	return new_rl;
  81}
  82
  83/**
  84 * ntfs_rl_realloc_nofail - Reallocate memory for runlists
  85 * @rl:		original runlist
  86 * @old_size:	number of runlist elements in the original runlist @rl
  87 * @new_size:	number of runlist elements we need space for
  88 *
  89 * As the runlists grow, more memory will be required.  To prevent the
  90 * kernel having to allocate and reallocate large numbers of small bits of
  91 * memory, this function returns an entire page of memory.
  92 *
  93 * This function guarantees that the allocation will succeed.  It will sleep
  94 * for as long as it takes to complete the allocation.
  95 *
  96 * It is up to the caller to serialize access to the runlist @rl.
  97 *
  98 * N.B.  If the new allocation doesn't require a different number of pages in
  99 *       memory, the function will return the original pointer.
 100 *
 101 * On success, return a pointer to the newly allocated, or recycled, memory.
 102 * On error, return -errno. The following error codes are defined:
 103 *	-ENOMEM	- Not enough memory to allocate runlist array.
 104 *	-EINVAL	- Invalid parameters were passed in.
 105 */
 106static inline runlist_element *ntfs_rl_realloc_nofail(runlist_element *rl,
 107		int old_size, int new_size)
 108{
 109	runlist_element *new_rl;
 110
 111	old_size = PAGE_ALIGN(old_size * sizeof(*rl));
 112	new_size = PAGE_ALIGN(new_size * sizeof(*rl));
 113	if (old_size == new_size)
 114		return rl;
 115
 116	new_rl = ntfs_malloc_nofs_nofail(new_size);
 117	BUG_ON(!new_rl);
 118
 119	if (likely(rl != NULL)) {
 120		if (unlikely(old_size > new_size))
 121			old_size = new_size;
 122		memcpy(new_rl, rl, old_size);
 123		ntfs_free(rl);
 124	}
 125	return new_rl;
 126}
 127
 128/**
 129 * ntfs_are_rl_mergeable - test if two runlists can be joined together
 130 * @dst:	original runlist
 131 * @src:	new runlist to test for mergeability with @dst
 132 *
 133 * Test if two runlists can be joined together. For this, their VCNs and LCNs
 134 * must be adjacent.
 135 *
 136 * It is up to the caller to serialize access to the runlists @dst and @src.
 137 *
 138 * Return: true   Success, the runlists can be merged.
 139 *	   false  Failure, the runlists cannot be merged.
 140 */
 141static inline bool ntfs_are_rl_mergeable(runlist_element *dst,
 142		runlist_element *src)
 143{
 144	BUG_ON(!dst);
 145	BUG_ON(!src);
 146
 147	/* We can merge unmapped regions even if they are misaligned. */
 148	if ((dst->lcn == LCN_RL_NOT_MAPPED) && (src->lcn == LCN_RL_NOT_MAPPED))
 149		return true;
 150	/* If the runs are misaligned, we cannot merge them. */
 151	if ((dst->vcn + dst->length) != src->vcn)
 152		return false;
 153	/* If both runs are non-sparse and contiguous, we can merge them. */
 154	if ((dst->lcn >= 0) && (src->lcn >= 0) &&
 155			((dst->lcn + dst->length) == src->lcn))
 156		return true;
 157	/* If we are merging two holes, we can merge them. */
 158	if ((dst->lcn == LCN_HOLE) && (src->lcn == LCN_HOLE))
 159		return true;
 160	/* Cannot merge. */
 161	return false;
 162}
 163
 164/**
 165 * __ntfs_rl_merge - merge two runlists without testing if they can be merged
 166 * @dst:	original, destination runlist
 167 * @src:	new runlist to merge with @dst
 168 *
 169 * Merge the two runlists, writing into the destination runlist @dst. The
 170 * caller must make sure the runlists can be merged or this will corrupt the
 171 * destination runlist.
 172 *
 173 * It is up to the caller to serialize access to the runlists @dst and @src.
 174 */
 175static inline void __ntfs_rl_merge(runlist_element *dst, runlist_element *src)
 176{
 177	dst->length += src->length;
 178}
 179
 180/**
 181 * ntfs_rl_append - append a runlist after a given element
 182 * @dst:	original runlist to be worked on
 183 * @dsize:	number of elements in @dst (including end marker)
 184 * @src:	runlist to be inserted into @dst
 185 * @ssize:	number of elements in @src (excluding end marker)
 186 * @loc:	append the new runlist @src after this element in @dst
 187 *
 188 * Append the runlist @src after element @loc in @dst.  Merge the right end of
 189 * the new runlist, if necessary. Adjust the size of the hole before the
 190 * appended runlist.
 191 *
 192 * It is up to the caller to serialize access to the runlists @dst and @src.
 193 *
 194 * On success, return a pointer to the new, combined, runlist. Note, both
 195 * runlists @dst and @src are deallocated before returning so you cannot use
 196 * the pointers for anything any more. (Strictly speaking the returned runlist
 197 * may be the same as @dst but this is irrelevant.)
 198 *
 199 * On error, return -errno. Both runlists are left unmodified. The following
 200 * error codes are defined:
 201 *	-ENOMEM	- Not enough memory to allocate runlist array.
 202 *	-EINVAL	- Invalid parameters were passed in.
 203 */
 204static inline runlist_element *ntfs_rl_append(runlist_element *dst,
 205		int dsize, runlist_element *src, int ssize, int loc)
 206{
 207	bool right = false;	/* Right end of @src needs merging. */
 208	int marker;		/* End of the inserted runs. */
 209
 210	BUG_ON(!dst);
 211	BUG_ON(!src);
 212
 213	/* First, check if the right hand end needs merging. */
 214	if ((loc + 1) < dsize)
 215		right = ntfs_are_rl_mergeable(src + ssize - 1, dst + loc + 1);
 216
 217	/* Space required: @dst size + @src size, less one if we merged. */
 218	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize - right);
 219	if (IS_ERR(dst))
 220		return dst;
 221	/*
 222	 * We are guaranteed to succeed from here so can start modifying the
 223	 * original runlists.
 224	 */
 225
 226	/* First, merge the right hand end, if necessary. */
 227	if (right)
 228		__ntfs_rl_merge(src + ssize - 1, dst + loc + 1);
 229
 230	/* First run after the @src runs that have been inserted. */
 231	marker = loc + ssize + 1;
 232
 233	/* Move the tail of @dst out of the way, then copy in @src. */
 234	ntfs_rl_mm(dst, marker, loc + 1 + right, dsize - (loc + 1 + right));
 235	ntfs_rl_mc(dst, loc + 1, src, 0, ssize);
 236
 237	/* Adjust the size of the preceding hole. */
 238	dst[loc].length = dst[loc + 1].vcn - dst[loc].vcn;
 239
 240	/* We may have changed the length of the file, so fix the end marker */
 241	if (dst[marker].lcn == LCN_ENOENT)
 242		dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
 243
 244	return dst;
 245}
 246
 247/**
 248 * ntfs_rl_insert - insert a runlist into another
 249 * @dst:	original runlist to be worked on
 250 * @dsize:	number of elements in @dst (including end marker)
 251 * @src:	new runlist to be inserted
 252 * @ssize:	number of elements in @src (excluding end marker)
 253 * @loc:	insert the new runlist @src before this element in @dst
 254 *
 255 * Insert the runlist @src before element @loc in the runlist @dst. Merge the
 256 * left end of the new runlist, if necessary. Adjust the size of the hole
 257 * after the inserted runlist.
 258 *
 259 * It is up to the caller to serialize access to the runlists @dst and @src.
 260 *
 261 * On success, return a pointer to the new, combined, runlist. Note, both
 262 * runlists @dst and @src are deallocated before returning so you cannot use
 263 * the pointers for anything any more. (Strictly speaking the returned runlist
 264 * may be the same as @dst but this is irrelevant.)
 265 *
 266 * On error, return -errno. Both runlists are left unmodified. The following
 267 * error codes are defined:
 268 *	-ENOMEM	- Not enough memory to allocate runlist array.
 269 *	-EINVAL	- Invalid parameters were passed in.
 270 */
 271static inline runlist_element *ntfs_rl_insert(runlist_element *dst,
 272		int dsize, runlist_element *src, int ssize, int loc)
 273{
 274	bool left = false;	/* Left end of @src needs merging. */
 275	bool disc = false;	/* Discontinuity between @dst and @src. */
 276	int marker;		/* End of the inserted runs. */
 277
 278	BUG_ON(!dst);
 279	BUG_ON(!src);
 280
 281	/*
 282	 * disc => Discontinuity between the end of @dst and the start of @src.
 283	 *	   This means we might need to insert a "not mapped" run.
 284	 */
 285	if (loc == 0)
 286		disc = (src[0].vcn > 0);
 287	else {
 288		s64 merged_length;
 289
 290		left = ntfs_are_rl_mergeable(dst + loc - 1, src);
 291
 292		merged_length = dst[loc - 1].length;
 293		if (left)
 294			merged_length += src->length;
 295
 296		disc = (src[0].vcn > dst[loc - 1].vcn + merged_length);
 297	}
 298	/*
 299	 * Space required: @dst size + @src size, less one if we merged, plus
 300	 * one if there was a discontinuity.
 301	 */
 302	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize - left + disc);
 303	if (IS_ERR(dst))
 304		return dst;
 305	/*
 306	 * We are guaranteed to succeed from here so can start modifying the
 307	 * original runlist.
 308	 */
 309	if (left)
 310		__ntfs_rl_merge(dst + loc - 1, src);
 311	/*
 312	 * First run after the @src runs that have been inserted.
 313	 * Nominally,  @marker equals @loc + @ssize, i.e. location + number of
 314	 * runs in @src.  However, if @left, then the first run in @src has
 315	 * been merged with one in @dst.  And if @disc, then @dst and @src do
 316	 * not meet and we need an extra run to fill the gap.
 317	 */
 318	marker = loc + ssize - left + disc;
 319
 320	/* Move the tail of @dst out of the way, then copy in @src. */
 321	ntfs_rl_mm(dst, marker, loc, dsize - loc);
 322	ntfs_rl_mc(dst, loc + disc, src, left, ssize - left);
 323
 324	/* Adjust the VCN of the first run after the insertion... */
 325	dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
 326	/* ... and the length. */
 327	if (dst[marker].lcn == LCN_HOLE || dst[marker].lcn == LCN_RL_NOT_MAPPED)
 328		dst[marker].length = dst[marker + 1].vcn - dst[marker].vcn;
 329
 330	/* Writing beyond the end of the file and there is a discontinuity. */
 331	if (disc) {
 332		if (loc > 0) {
 333			dst[loc].vcn = dst[loc - 1].vcn + dst[loc - 1].length;
 334			dst[loc].length = dst[loc + 1].vcn - dst[loc].vcn;
 335		} else {
 336			dst[loc].vcn = 0;
 337			dst[loc].length = dst[loc + 1].vcn;
 338		}
 339		dst[loc].lcn = LCN_RL_NOT_MAPPED;
 340	}
 341	return dst;
 342}
 343
 344/**
 345 * ntfs_rl_replace - overwrite a runlist element with another runlist
 346 * @dst:	original runlist to be worked on
 347 * @dsize:	number of elements in @dst (including end marker)
 348 * @src:	new runlist to be inserted
 349 * @ssize:	number of elements in @src (excluding end marker)
 350 * @loc:	index in runlist @dst to overwrite with @src
 351 *
 352 * Replace the runlist element @dst at @loc with @src. Merge the left and
 353 * right ends of the inserted runlist, if necessary.
 354 *
 355 * It is up to the caller to serialize access to the runlists @dst and @src.
 356 *
 357 * On success, return a pointer to the new, combined, runlist. Note, both
 358 * runlists @dst and @src are deallocated before returning so you cannot use
 359 * the pointers for anything any more. (Strictly speaking the returned runlist
 360 * may be the same as @dst but this is irrelevant.)
 361 *
 362 * On error, return -errno. Both runlists are left unmodified. The following
 363 * error codes are defined:
 364 *	-ENOMEM	- Not enough memory to allocate runlist array.
 365 *	-EINVAL	- Invalid parameters were passed in.
 366 */
 367static inline runlist_element *ntfs_rl_replace(runlist_element *dst,
 368		int dsize, runlist_element *src, int ssize, int loc)
 369{
 370	signed delta;
 371	bool left = false;	/* Left end of @src needs merging. */
 372	bool right = false;	/* Right end of @src needs merging. */
 373	int tail;		/* Start of tail of @dst. */
 374	int marker;		/* End of the inserted runs. */
 375
 376	BUG_ON(!dst);
 377	BUG_ON(!src);
 378
 379	/* First, see if the left and right ends need merging. */
 380	if ((loc + 1) < dsize)
 381		right = ntfs_are_rl_mergeable(src + ssize - 1, dst + loc + 1);
 382	if (loc > 0)
 383		left = ntfs_are_rl_mergeable(dst + loc - 1, src);
 384	/*
 385	 * Allocate some space.  We will need less if the left, right, or both
 386	 * ends get merged.  The -1 accounts for the run being replaced.
 387	 */
 388	delta = ssize - 1 - left - right;
 389	if (delta > 0) {
 390		dst = ntfs_rl_realloc(dst, dsize, dsize + delta);
 391		if (IS_ERR(dst))
 392			return dst;
 393	}
 394	/*
 395	 * We are guaranteed to succeed from here so can start modifying the
 396	 * original runlists.
 397	 */
 398
 399	/* First, merge the left and right ends, if necessary. */
 400	if (right)
 401		__ntfs_rl_merge(src + ssize - 1, dst + loc + 1);
 402	if (left)
 403		__ntfs_rl_merge(dst + loc - 1, src);
 404	/*
 405	 * Offset of the tail of @dst.  This needs to be moved out of the way
 406	 * to make space for the runs to be copied from @src, i.e. the first
 407	 * run of the tail of @dst.
 408	 * Nominally, @tail equals @loc + 1, i.e. location, skipping the
 409	 * replaced run.  However, if @right, then one of @dst's runs is
 410	 * already merged into @src.
 411	 */
 412	tail = loc + right + 1;
 413	/*
 414	 * First run after the @src runs that have been inserted, i.e. where
 415	 * the tail of @dst needs to be moved to.
 416	 * Nominally, @marker equals @loc + @ssize, i.e. location + number of
 417	 * runs in @src.  However, if @left, then the first run in @src has
 418	 * been merged with one in @dst.
 419	 */
 420	marker = loc + ssize - left;
 421
 422	/* Move the tail of @dst out of the way, then copy in @src. */
 423	ntfs_rl_mm(dst, marker, tail, dsize - tail);
 424	ntfs_rl_mc(dst, loc, src, left, ssize - left);
 425
 426	/* We may have changed the length of the file, so fix the end marker. */
 427	if (dsize - tail > 0 && dst[marker].lcn == LCN_ENOENT)
 428		dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
 429	return dst;
 430}
 431
 432/**
 433 * ntfs_rl_split - insert a runlist into the centre of a hole
 434 * @dst:	original runlist to be worked on
 435 * @dsize:	number of elements in @dst (including end marker)
 436 * @src:	new runlist to be inserted
 437 * @ssize:	number of elements in @src (excluding end marker)
 438 * @loc:	index in runlist @dst at which to split and insert @src
 439 *
 440 * Split the runlist @dst at @loc into two and insert @new in between the two
 441 * fragments. No merging of runlists is necessary. Adjust the size of the
 442 * holes either side.
 443 *
 444 * It is up to the caller to serialize access to the runlists @dst and @src.
 445 *
 446 * On success, return a pointer to the new, combined, runlist. Note, both
 447 * runlists @dst and @src are deallocated before returning so you cannot use
 448 * the pointers for anything any more. (Strictly speaking the returned runlist
 449 * may be the same as @dst but this is irrelevant.)
 450 *
 451 * On error, return -errno. Both runlists are left unmodified. The following
 452 * error codes are defined:
 453 *	-ENOMEM	- Not enough memory to allocate runlist array.
 454 *	-EINVAL	- Invalid parameters were passed in.
 455 */
 456static inline runlist_element *ntfs_rl_split(runlist_element *dst, int dsize,
 457		runlist_element *src, int ssize, int loc)
 458{
 459	BUG_ON(!dst);
 460	BUG_ON(!src);
 461
 462	/* Space required: @dst size + @src size + one new hole. */
 463	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize + 1);
 464	if (IS_ERR(dst))
 465		return dst;
 466	/*
 467	 * We are guaranteed to succeed from here so can start modifying the
 468	 * original runlists.
 469	 */
 470
 471	/* Move the tail of @dst out of the way, then copy in @src. */
 472	ntfs_rl_mm(dst, loc + 1 + ssize, loc, dsize - loc);
 473	ntfs_rl_mc(dst, loc + 1, src, 0, ssize);
 474
 475	/* Adjust the size of the holes either size of @src. */
 476	dst[loc].length		= dst[loc+1].vcn       - dst[loc].vcn;
 477	dst[loc+ssize+1].vcn    = dst[loc+ssize].vcn   + dst[loc+ssize].length;
 478	dst[loc+ssize+1].length = dst[loc+ssize+2].vcn - dst[loc+ssize+1].vcn;
 479
 480	return dst;
 481}
 482
 483/**
 484 * ntfs_runlists_merge - merge two runlists into one
 485 * @drl:	original runlist to be worked on
 486 * @srl:	new runlist to be merged into @drl
 487 *
 488 * First we sanity check the two runlists @srl and @drl to make sure that they
 489 * are sensible and can be merged. The runlist @srl must be either after the
 490 * runlist @drl or completely within a hole (or unmapped region) in @drl.
 491 *
 492 * It is up to the caller to serialize access to the runlists @drl and @srl.
 493 *
 494 * Merging of runlists is necessary in two cases:
 495 *   1. When attribute lists are used and a further extent is being mapped.
 496 *   2. When new clusters are allocated to fill a hole or extend a file.
 497 *
 498 * There are four possible ways @srl can be merged. It can:
 499 *	- be inserted at the beginning of a hole,
 500 *	- split the hole in two and be inserted between the two fragments,
 501 *	- be appended at the end of a hole, or it can
 502 *	- replace the whole hole.
 503 * It can also be appended to the end of the runlist, which is just a variant
 504 * of the insert case.
 505 *
 506 * On success, return a pointer to the new, combined, runlist. Note, both
 507 * runlists @drl and @srl are deallocated before returning so you cannot use
 508 * the pointers for anything any more. (Strictly speaking the returned runlist
 509 * may be the same as @dst but this is irrelevant.)
 510 *
 511 * On error, return -errno. Both runlists are left unmodified. The following
 512 * error codes are defined:
 513 *	-ENOMEM	- Not enough memory to allocate runlist array.
 514 *	-EINVAL	- Invalid parameters were passed in.
 515 *	-ERANGE	- The runlists overlap and cannot be merged.
 516 */
 517runlist_element *ntfs_runlists_merge(runlist_element *drl,
 518		runlist_element *srl)
 519{
 520	int di, si;		/* Current index into @[ds]rl. */
 521	int sstart;		/* First index with lcn > LCN_RL_NOT_MAPPED. */
 522	int dins;		/* Index into @drl at which to insert @srl. */
 523	int dend, send;		/* Last index into @[ds]rl. */
 524	int dfinal, sfinal;	/* The last index into @[ds]rl with
 525				   lcn >= LCN_HOLE. */
 526	int marker = 0;
 527	VCN marker_vcn = 0;
 528
 529#ifdef DEBUG
 530	ntfs_debug("dst:");
 531	ntfs_debug_dump_runlist(drl);
 532	ntfs_debug("src:");
 533	ntfs_debug_dump_runlist(srl);
 534#endif
 535
 536	/* Check for silly calling... */
 537	if (unlikely(!srl))
 538		return drl;
 539	if (IS_ERR(srl) || IS_ERR(drl))
 540		return ERR_PTR(-EINVAL);
 541
 542	/* Check for the case where the first mapping is being done now. */
 543	if (unlikely(!drl)) {
 544		drl = srl;
 545		/* Complete the source runlist if necessary. */
 546		if (unlikely(drl[0].vcn)) {
 547			/* Scan to the end of the source runlist. */
 548			for (dend = 0; likely(drl[dend].length); dend++)
 549				;
 550			dend++;
 551			drl = ntfs_rl_realloc(drl, dend, dend + 1);
 552			if (IS_ERR(drl))
 553				return drl;
 554			/* Insert start element at the front of the runlist. */
 555			ntfs_rl_mm(drl, 1, 0, dend);
 556			drl[0].vcn = 0;
 557			drl[0].lcn = LCN_RL_NOT_MAPPED;
 558			drl[0].length = drl[1].vcn;
 559		}
 560		goto finished;
 561	}
 562
 563	si = di = 0;
 564
 565	/* Skip any unmapped start element(s) in the source runlist. */
 566	while (srl[si].length && srl[si].lcn < LCN_HOLE)
 567		si++;
 568
 569	/* Can't have an entirely unmapped source runlist. */
 570	BUG_ON(!srl[si].length);
 571
 572	/* Record the starting points. */
 573	sstart = si;
 574
 575	/*
 576	 * Skip forward in @drl until we reach the position where @srl needs to
 577	 * be inserted. If we reach the end of @drl, @srl just needs to be
 578	 * appended to @drl.
 579	 */
 580	for (; drl[di].length; di++) {
 581		if (drl[di].vcn + drl[di].length > srl[sstart].vcn)
 582			break;
 583	}
 584	dins = di;
 585
 586	/* Sanity check for illegal overlaps. */
 587	if ((drl[di].vcn == srl[si].vcn) && (drl[di].lcn >= 0) &&
 588			(srl[si].lcn >= 0)) {
 589		ntfs_error(NULL, "Run lists overlap. Cannot merge!");
 590		return ERR_PTR(-ERANGE);
 591	}
 592
 593	/* Scan to the end of both runlists in order to know their sizes. */
 594	for (send = si; srl[send].length; send++)
 595		;
 596	for (dend = di; drl[dend].length; dend++)
 597		;
 598
 599	if (srl[send].lcn == LCN_ENOENT)
 600		marker_vcn = srl[marker = send].vcn;
 601
 602	/* Scan to the last element with lcn >= LCN_HOLE. */
 603	for (sfinal = send; sfinal >= 0 && srl[sfinal].lcn < LCN_HOLE; sfinal--)
 604		;
 605	for (dfinal = dend; dfinal >= 0 && drl[dfinal].lcn < LCN_HOLE; dfinal--)
 606		;
 607
 608	{
 609	bool start;
 610	bool finish;
 611	int ds = dend + 1;		/* Number of elements in drl & srl */
 612	int ss = sfinal - sstart + 1;
 613
 614	start  = ((drl[dins].lcn <  LCN_RL_NOT_MAPPED) ||    /* End of file   */
 615		  (drl[dins].vcn == srl[sstart].vcn));	     /* Start of hole */
 616	finish = ((drl[dins].lcn >= LCN_RL_NOT_MAPPED) &&    /* End of file   */
 617		 ((drl[dins].vcn + drl[dins].length) <=      /* End of hole   */
 618		  (srl[send - 1].vcn + srl[send - 1].length)));
 619
 620	/* Or we will lose an end marker. */
 621	if (finish && !drl[dins].length)
 622		ss++;
 623	if (marker && (drl[dins].vcn + drl[dins].length > srl[send - 1].vcn))
 624		finish = false;
 625#if 0
 626	ntfs_debug("dfinal = %i, dend = %i", dfinal, dend);
 627	ntfs_debug("sstart = %i, sfinal = %i, send = %i", sstart, sfinal, send);
 628	ntfs_debug("start = %i, finish = %i", start, finish);
 629	ntfs_debug("ds = %i, ss = %i, dins = %i", ds, ss, dins);
 630#endif
 631	if (start) {
 632		if (finish)
 633			drl = ntfs_rl_replace(drl, ds, srl + sstart, ss, dins);
 634		else
 635			drl = ntfs_rl_insert(drl, ds, srl + sstart, ss, dins);
 636	} else {
 637		if (finish)
 638			drl = ntfs_rl_append(drl, ds, srl + sstart, ss, dins);
 639		else
 640			drl = ntfs_rl_split(drl, ds, srl + sstart, ss, dins);
 641	}
 642	if (IS_ERR(drl)) {
 643		ntfs_error(NULL, "Merge failed.");
 644		return drl;
 645	}
 646	ntfs_free(srl);
 647	if (marker) {
 648		ntfs_debug("Triggering marker code.");
 649		for (ds = dend; drl[ds].length; ds++)
 650			;
 651		/* We only need to care if @srl ended after @drl. */
 652		if (drl[ds].vcn <= marker_vcn) {
 653			int slots = 0;
 654
 655			if (drl[ds].vcn == marker_vcn) {
 656				ntfs_debug("Old marker = 0x%llx, replacing "
 657						"with LCN_ENOENT.",
 658						(unsigned long long)
 659						drl[ds].lcn);
 660				drl[ds].lcn = LCN_ENOENT;
 661				goto finished;
 662			}
 663			/*
 664			 * We need to create an unmapped runlist element in
 665			 * @drl or extend an existing one before adding the
 666			 * ENOENT terminator.
 667			 */
 668			if (drl[ds].lcn == LCN_ENOENT) {
 669				ds--;
 670				slots = 1;
 671			}
 672			if (drl[ds].lcn != LCN_RL_NOT_MAPPED) {
 673				/* Add an unmapped runlist element. */
 674				if (!slots) {
 675					drl = ntfs_rl_realloc_nofail(drl, ds,
 676							ds + 2);
 677					slots = 2;
 678				}
 679				ds++;
 680				/* Need to set vcn if it isn't set already. */
 681				if (slots != 1)
 682					drl[ds].vcn = drl[ds - 1].vcn +
 683							drl[ds - 1].length;
 684				drl[ds].lcn = LCN_RL_NOT_MAPPED;
 685				/* We now used up a slot. */
 686				slots--;
 687			}
 688			drl[ds].length = marker_vcn - drl[ds].vcn;
 689			/* Finally add the ENOENT terminator. */
 690			ds++;
 691			if (!slots)
 692				drl = ntfs_rl_realloc_nofail(drl, ds, ds + 1);
 693			drl[ds].vcn = marker_vcn;
 694			drl[ds].lcn = LCN_ENOENT;
 695			drl[ds].length = (s64)0;
 696		}
 697	}
 698	}
 699
 700finished:
 701	/* The merge was completed successfully. */
 702	ntfs_debug("Merged runlist:");
 703	ntfs_debug_dump_runlist(drl);
 704	return drl;
 705}
 706
 707/**
 708 * ntfs_mapping_pairs_decompress - convert mapping pairs array to runlist
 709 * @vol:	ntfs volume on which the attribute resides
 710 * @attr:	attribute record whose mapping pairs array to decompress
 711 * @old_rl:	optional runlist in which to insert @attr's runlist
 712 *
 713 * It is up to the caller to serialize access to the runlist @old_rl.
 714 *
 715 * Decompress the attribute @attr's mapping pairs array into a runlist. On
 716 * success, return the decompressed runlist.
 717 *
 718 * If @old_rl is not NULL, decompressed runlist is inserted into the
 719 * appropriate place in @old_rl and the resultant, combined runlist is
 720 * returned. The original @old_rl is deallocated.
 721 *
 722 * On error, return -errno. @old_rl is left unmodified in that case.
 723 *
 724 * The following error codes are defined:
 725 *	-ENOMEM	- Not enough memory to allocate runlist array.
 726 *	-EIO	- Corrupt runlist.
 727 *	-EINVAL	- Invalid parameters were passed in.
 728 *	-ERANGE	- The two runlists overlap.
 729 *
 730 * FIXME: For now we take the conceptionally simplest approach of creating the
 731 * new runlist disregarding the already existing one and then splicing the
 732 * two into one, if that is possible (we check for overlap and discard the new
 733 * runlist if overlap present before returning ERR_PTR(-ERANGE)).
 734 */
 735runlist_element *ntfs_mapping_pairs_decompress(const ntfs_volume *vol,
 736		const ATTR_RECORD *attr, runlist_element *old_rl)
 737{
 738	VCN vcn;		/* Current vcn. */
 739	LCN lcn;		/* Current lcn. */
 740	s64 deltaxcn;		/* Change in [vl]cn. */
 741	runlist_element *rl;	/* The output runlist. */
 742	u8 *buf;		/* Current position in mapping pairs array. */
 743	u8 *attr_end;		/* End of attribute. */
 744	int rlsize;		/* Size of runlist buffer. */
 745	u16 rlpos;		/* Current runlist position in units of
 746				   runlist_elements. */
 747	u8 b;			/* Current byte offset in buf. */
 748
 749#ifdef DEBUG
 750	/* Make sure attr exists and is non-resident. */
 751	if (!attr || !attr->non_resident || sle64_to_cpu(
 752			attr->data.non_resident.lowest_vcn) < (VCN)0) {
 753		ntfs_error(vol->sb, "Invalid arguments.");
 754		return ERR_PTR(-EINVAL);
 755	}
 756#endif
 757	/* Start at vcn = lowest_vcn and lcn 0. */
 758	vcn = sle64_to_cpu(attr->data.non_resident.lowest_vcn);
 759	lcn = 0;
 760	/* Get start of the mapping pairs array. */
 761	buf = (u8*)attr + le16_to_cpu(
 762			attr->data.non_resident.mapping_pairs_offset);
 763	attr_end = (u8*)attr + le32_to_cpu(attr->length);
 764	if (unlikely(buf < (u8*)attr || buf > attr_end)) {
 765		ntfs_error(vol->sb, "Corrupt attribute.");
 766		return ERR_PTR(-EIO);
 767	}
 768	/* If the mapping pairs array is valid but empty, nothing to do. */
 769	if (!vcn && !*buf)
 770		return old_rl;
 771	/* Current position in runlist array. */
 772	rlpos = 0;
 773	/* Allocate first page and set current runlist size to one page. */
 774	rl = ntfs_malloc_nofs(rlsize = PAGE_SIZE);
 775	if (unlikely(!rl))
 776		return ERR_PTR(-ENOMEM);
 777	/* Insert unmapped starting element if necessary. */
 778	if (vcn) {
 779		rl->vcn = 0;
 780		rl->lcn = LCN_RL_NOT_MAPPED;
 781		rl->length = vcn;
 782		rlpos++;
 783	}
 784	while (buf < attr_end && *buf) {
 785		/*
 786		 * Allocate more memory if needed, including space for the
 787		 * not-mapped and terminator elements. ntfs_malloc_nofs()
 788		 * operates on whole pages only.
 789		 */
 790		if (((rlpos + 3) * sizeof(*old_rl)) > rlsize) {
 791			runlist_element *rl2;
 792
 793			rl2 = ntfs_malloc_nofs(rlsize + (int)PAGE_SIZE);
 794			if (unlikely(!rl2)) {
 795				ntfs_free(rl);
 796				return ERR_PTR(-ENOMEM);
 797			}
 798			memcpy(rl2, rl, rlsize);
 799			ntfs_free(rl);
 800			rl = rl2;
 801			rlsize += PAGE_SIZE;
 802		}
 803		/* Enter the current vcn into the current runlist element. */
 804		rl[rlpos].vcn = vcn;
 805		/*
 806		 * Get the change in vcn, i.e. the run length in clusters.
 807		 * Doing it this way ensures that we signextend negative values.
 808		 * A negative run length doesn't make any sense, but hey, I
 809		 * didn't make up the NTFS specs and Windows NT4 treats the run
 810		 * length as a signed value so that's how it is...
 811		 */
 812		b = *buf & 0xf;
 813		if (b) {
 814			if (unlikely(buf + b > attr_end))
 815				goto io_error;
 816			for (deltaxcn = (s8)buf[b--]; b; b--)
 817				deltaxcn = (deltaxcn << 8) + buf[b];
 818		} else { /* The length entry is compulsory. */
 819			ntfs_error(vol->sb, "Missing length entry in mapping "
 820					"pairs array.");
 821			deltaxcn = (s64)-1;
 822		}
 823		/*
 824		 * Assume a negative length to indicate data corruption and
 825		 * hence clean-up and return NULL.
 826		 */
 827		if (unlikely(deltaxcn < 0)) {
 828			ntfs_error(vol->sb, "Invalid length in mapping pairs "
 829					"array.");
 830			goto err_out;
 831		}
 832		/*
 833		 * Enter the current run length into the current runlist
 834		 * element.
 835		 */
 836		rl[rlpos].length = deltaxcn;
 837		/* Increment the current vcn by the current run length. */
 838		vcn += deltaxcn;
 839		/*
 840		 * There might be no lcn change at all, as is the case for
 841		 * sparse clusters on NTFS 3.0+, in which case we set the lcn
 842		 * to LCN_HOLE.
 843		 */
 844		if (!(*buf & 0xf0))
 845			rl[rlpos].lcn = LCN_HOLE;
 846		else {
 847			/* Get the lcn change which really can be negative. */
 848			u8 b2 = *buf & 0xf;
 849			b = b2 + ((*buf >> 4) & 0xf);
 850			if (buf + b > attr_end)
 851				goto io_error;
 852			for (deltaxcn = (s8)buf[b--]; b > b2; b--)
 853				deltaxcn = (deltaxcn << 8) + buf[b];
 854			/* Change the current lcn to its new value. */
 855			lcn += deltaxcn;
 856#ifdef DEBUG
 857			/*
 858			 * On NTFS 1.2-, apparently can have lcn == -1 to
 859			 * indicate a hole. But we haven't verified ourselves
 860			 * whether it is really the lcn or the deltaxcn that is
 861			 * -1. So if either is found give us a message so we
 862			 * can investigate it further!
 863			 */
 864			if (vol->major_ver < 3) {
 865				if (unlikely(deltaxcn == (LCN)-1))
 866					ntfs_error(vol->sb, "lcn delta == -1");
 867				if (unlikely(lcn == (LCN)-1))
 868					ntfs_error(vol->sb, "lcn == -1");
 869			}
 870#endif
 871			/* Check lcn is not below -1. */
 872			if (unlikely(lcn < (LCN)-1)) {
 873				ntfs_error(vol->sb, "Invalid LCN < -1 in "
 874						"mapping pairs array.");
 875				goto err_out;
 876			}
 877			/* Enter the current lcn into the runlist element. */
 878			rl[rlpos].lcn = lcn;
 879		}
 880		/* Get to the next runlist element. */
 881		rlpos++;
 882		/* Increment the buffer position to the next mapping pair. */
 883		buf += (*buf & 0xf) + ((*buf >> 4) & 0xf) + 1;
 884	}
 885	if (unlikely(buf >= attr_end))
 886		goto io_error;
 887	/*
 888	 * If there is a highest_vcn specified, it must be equal to the final
 889	 * vcn in the runlist - 1, or something has gone badly wrong.
 890	 */
 891	deltaxcn = sle64_to_cpu(attr->data.non_resident.highest_vcn);
 892	if (unlikely(deltaxcn && vcn - 1 != deltaxcn)) {
 893mpa_err:
 894		ntfs_error(vol->sb, "Corrupt mapping pairs array in "
 895				"non-resident attribute.");
 896		goto err_out;
 897	}
 898	/* Setup not mapped runlist element if this is the base extent. */
 899	if (!attr->data.non_resident.lowest_vcn) {
 900		VCN max_cluster;
 901
 902		max_cluster = ((sle64_to_cpu(
 903				attr->data.non_resident.allocated_size) +
 904				vol->cluster_size - 1) >>
 905				vol->cluster_size_bits) - 1;
 906		/*
 907		 * A highest_vcn of zero means this is a single extent
 908		 * attribute so simply terminate the runlist with LCN_ENOENT).
 909		 */
 910		if (deltaxcn) {
 911			/*
 912			 * If there is a difference between the highest_vcn and
 913			 * the highest cluster, the runlist is either corrupt
 914			 * or, more likely, there are more extents following
 915			 * this one.
 916			 */
 917			if (deltaxcn < max_cluster) {
 918				ntfs_debug("More extents to follow; deltaxcn "
 919						"= 0x%llx, max_cluster = "
 920						"0x%llx",
 921						(unsigned long long)deltaxcn,
 922						(unsigned long long)
 923						max_cluster);
 924				rl[rlpos].vcn = vcn;
 925				vcn += rl[rlpos].length = max_cluster -
 926						deltaxcn;
 927				rl[rlpos].lcn = LCN_RL_NOT_MAPPED;
 928				rlpos++;
 929			} else if (unlikely(deltaxcn > max_cluster)) {
 930				ntfs_error(vol->sb, "Corrupt attribute.  "
 931						"deltaxcn = 0x%llx, "
 932						"max_cluster = 0x%llx",
 933						(unsigned long long)deltaxcn,
 934						(unsigned long long)
 935						max_cluster);
 936				goto mpa_err;
 937			}
 938		}
 939		rl[rlpos].lcn = LCN_ENOENT;
 940	} else /* Not the base extent. There may be more extents to follow. */
 941		rl[rlpos].lcn = LCN_RL_NOT_MAPPED;
 942
 943	/* Setup terminating runlist element. */
 944	rl[rlpos].vcn = vcn;
 945	rl[rlpos].length = (s64)0;
 946	/* If no existing runlist was specified, we are done. */
 947	if (!old_rl) {
 948		ntfs_debug("Mapping pairs array successfully decompressed:");
 949		ntfs_debug_dump_runlist(rl);
 950		return rl;
 951	}
 952	/* Now combine the new and old runlists checking for overlaps. */
 953	old_rl = ntfs_runlists_merge(old_rl, rl);
 954	if (!IS_ERR(old_rl))
 955		return old_rl;
 956	ntfs_free(rl);
 957	ntfs_error(vol->sb, "Failed to merge runlists.");
 958	return old_rl;
 959io_error:
 960	ntfs_error(vol->sb, "Corrupt attribute.");
 961err_out:
 962	ntfs_free(rl);
 963	return ERR_PTR(-EIO);
 964}
 965
 966/**
 967 * ntfs_rl_vcn_to_lcn - convert a vcn into a lcn given a runlist
 968 * @rl:		runlist to use for conversion
 969 * @vcn:	vcn to convert
 970 *
 971 * Convert the virtual cluster number @vcn of an attribute into a logical
 972 * cluster number (lcn) of a device using the runlist @rl to map vcns to their
 973 * corresponding lcns.
 974 *
 975 * It is up to the caller to serialize access to the runlist @rl.
 976 *
 977 * Since lcns must be >= 0, we use negative return codes with special meaning:
 978 *
 979 * Return code		Meaning / Description
 980 * ==================================================
 981 *  LCN_HOLE		Hole / not allocated on disk.
 982 *  LCN_RL_NOT_MAPPED	This is part of the runlist which has not been
 983 *			inserted into the runlist yet.
 984 *  LCN_ENOENT		There is no such vcn in the attribute.
 985 *
 986 * Locking: - The caller must have locked the runlist (for reading or writing).
 987 *	    - This function does not touch the lock, nor does it modify the
 988 *	      runlist.
 989 */
 990LCN ntfs_rl_vcn_to_lcn(const runlist_element *rl, const VCN vcn)
 991{
 992	int i;
 993
 994	BUG_ON(vcn < 0);
 995	/*
 996	 * If rl is NULL, assume that we have found an unmapped runlist. The
 997	 * caller can then attempt to map it and fail appropriately if
 998	 * necessary.
 999	 */
1000	if (unlikely(!rl))
1001		return LCN_RL_NOT_MAPPED;
1002
1003	/* Catch out of lower bounds vcn. */
1004	if (unlikely(vcn < rl[0].vcn))
1005		return LCN_ENOENT;
1006
1007	for (i = 0; likely(rl[i].length); i++) {
1008		if (unlikely(vcn < rl[i+1].vcn)) {
1009			if (likely(rl[i].lcn >= (LCN)0))
1010				return rl[i].lcn + (vcn - rl[i].vcn);
1011			return rl[i].lcn;
1012		}
1013	}
1014	/*
1015	 * The terminator element is setup to the correct value, i.e. one of
1016	 * LCN_HOLE, LCN_RL_NOT_MAPPED, or LCN_ENOENT.
1017	 */
1018	if (likely(rl[i].lcn < (LCN)0))
1019		return rl[i].lcn;
1020	/* Just in case... We could replace this with BUG() some day. */
1021	return LCN_ENOENT;
1022}
1023
1024#ifdef NTFS_RW
1025
1026/**
1027 * ntfs_rl_find_vcn_nolock - find a vcn in a runlist
1028 * @rl:		runlist to search
1029 * @vcn:	vcn to find
1030 *
1031 * Find the virtual cluster number @vcn in the runlist @rl and return the
1032 * address of the runlist element containing the @vcn on success.
1033 *
1034 * Return NULL if @rl is NULL or @vcn is in an unmapped part/out of bounds of
1035 * the runlist.
1036 *
1037 * Locking: The runlist must be locked on entry.
1038 */
1039runlist_element *ntfs_rl_find_vcn_nolock(runlist_element *rl, const VCN vcn)
1040{
1041	BUG_ON(vcn < 0);
1042	if (unlikely(!rl || vcn < rl[0].vcn))
1043		return NULL;
1044	while (likely(rl->length)) {
1045		if (unlikely(vcn < rl[1].vcn)) {
1046			if (likely(rl->lcn >= LCN_HOLE))
1047				return rl;
1048			return NULL;
1049		}
1050		rl++;
1051	}
1052	if (likely(rl->lcn == LCN_ENOENT))
1053		return rl;
1054	return NULL;
1055}
1056
1057/**
1058 * ntfs_get_nr_significant_bytes - get number of bytes needed to store a number
1059 * @n:		number for which to get the number of bytes for
1060 *
1061 * Return the number of bytes required to store @n unambiguously as
1062 * a signed number.
1063 *
1064 * This is used in the context of the mapping pairs array to determine how
1065 * many bytes will be needed in the array to store a given logical cluster
1066 * number (lcn) or a specific run length.
1067 *
1068 * Return the number of bytes written.  This function cannot fail.
1069 */
1070static inline int ntfs_get_nr_significant_bytes(const s64 n)
1071{
1072	s64 l = n;
1073	int i;
1074	s8 j;
1075
1076	i = 0;
1077	do {
1078		l >>= 8;
1079		i++;
1080	} while (l != 0 && l != -1);
1081	j = (n >> 8 * (i - 1)) & 0xff;
1082	/* If the sign bit is wrong, we need an extra byte. */
1083	if ((n < 0 && j >= 0) || (n > 0 && j < 0))
1084		i++;
1085	return i;
1086}
1087
1088/**
1089 * ntfs_get_size_for_mapping_pairs - get bytes needed for mapping pairs array
1090 * @vol:	ntfs volume (needed for the ntfs version)
1091 * @rl:		locked runlist to determine the size of the mapping pairs of
1092 * @first_vcn:	first vcn which to include in the mapping pairs array
1093 * @last_vcn:	last vcn which to include in the mapping pairs array
1094 *
1095 * Walk the locked runlist @rl and calculate the size in bytes of the mapping
1096 * pairs array corresponding to the runlist @rl, starting at vcn @first_vcn and
1097 * finishing with vcn @last_vcn.
1098 *
1099 * A @last_vcn of -1 means end of runlist and in that case the size of the
1100 * mapping pairs array corresponding to the runlist starting at vcn @first_vcn
1101 * and finishing at the end of the runlist is determined.
1102 *
1103 * This for example allows us to allocate a buffer of the right size when
1104 * building the mapping pairs array.
1105 *
1106 * If @rl is NULL, just return 1 (for the single terminator byte).
1107 *
1108 * Return the calculated size in bytes on success.  On error, return -errno.
1109 * The following error codes are defined:
1110 *	-EINVAL	- Run list contains unmapped elements.  Make sure to only pass
1111 *		  fully mapped runlists to this function.
1112 *	-EIO	- The runlist is corrupt.
1113 *
1114 * Locking: @rl must be locked on entry (either for reading or writing), it
1115 *	    remains locked throughout, and is left locked upon return.
1116 */
1117int ntfs_get_size_for_mapping_pairs(const ntfs_volume *vol,
1118		const runlist_element *rl, const VCN first_vcn,
1119		const VCN last_vcn)
1120{
1121	LCN prev_lcn;
1122	int rls;
1123	bool the_end = false;
1124
1125	BUG_ON(first_vcn < 0);
1126	BUG_ON(last_vcn < -1);
1127	BUG_ON(last_vcn >= 0 && first_vcn > last_vcn);
1128	if (!rl) {
1129		BUG_ON(first_vcn);
1130		BUG_ON(last_vcn > 0);
1131		return 1;
1132	}
1133	/* Skip to runlist element containing @first_vcn. */
1134	while (rl->length && first_vcn >= rl[1].vcn)
1135		rl++;
1136	if (unlikely((!rl->length && first_vcn > rl->vcn) ||
1137			first_vcn < rl->vcn))
1138		return -EINVAL;
1139	prev_lcn = 0;
1140	/* Always need the termining zero byte. */
1141	rls = 1;
1142	/* Do the first partial run if present. */
1143	if (first_vcn > rl->vcn) {
1144		s64 delta, length = rl->length;
1145
1146		/* We know rl->length != 0 already. */
1147		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1148			goto err_out;
1149		/*
1150		 * If @stop_vcn is given and finishes inside this run, cap the
1151		 * run length.
1152		 */
1153		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1154			s64 s1 = last_vcn + 1;
1155			if (unlikely(rl[1].vcn > s1))
1156				length = s1 - rl->vcn;
1157			the_end = true;
1158		}
1159		delta = first_vcn - rl->vcn;
1160		/* Header byte + length. */
1161		rls += 1 + ntfs_get_nr_significant_bytes(length - delta);
1162		/*
1163		 * If the logical cluster number (lcn) denotes a hole and we
1164		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1165		 * zero space.  On earlier NTFS versions we just store the lcn.
1166		 * Note: this assumes that on NTFS 1.2-, holes are stored with
1167		 * an lcn of -1 and not a delta_lcn of -1 (unless both are -1).
1168		 */
1169		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1170			prev_lcn = rl->lcn;
1171			if (likely(rl->lcn >= 0))
1172				prev_lcn += delta;
1173			/* Change in lcn. */
1174			rls += ntfs_get_nr_significant_bytes(prev_lcn);
1175		}
1176		/* Go to next runlist element. */
1177		rl++;
1178	}
1179	/* Do the full runs. */
1180	for (; rl->length && !the_end; rl++) {
1181		s64 length = rl->length;
1182
1183		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1184			goto err_out;
1185		/*
1186		 * If @stop_vcn is given and finishes inside this run, cap the
1187		 * run length.
1188		 */
1189		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1190			s64 s1 = last_vcn + 1;
1191			if (unlikely(rl[1].vcn > s1))
1192				length = s1 - rl->vcn;
1193			the_end = true;
1194		}
1195		/* Header byte + length. */
1196		rls += 1 + ntfs_get_nr_significant_bytes(length);
1197		/*
1198		 * If the logical cluster number (lcn) denotes a hole and we
1199		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1200		 * zero space.  On earlier NTFS versions we just store the lcn.
1201		 * Note: this assumes that on NTFS 1.2-, holes are stored with
1202		 * an lcn of -1 and not a delta_lcn of -1 (unless both are -1).
1203		 */
1204		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1205			/* Change in lcn. */
1206			rls += ntfs_get_nr_significant_bytes(rl->lcn -
1207					prev_lcn);
1208			prev_lcn = rl->lcn;
1209		}
1210	}
1211	return rls;
1212err_out:
1213	if (rl->lcn == LCN_RL_NOT_MAPPED)
1214		rls = -EINVAL;
1215	else
1216		rls = -EIO;
1217	return rls;
1218}
1219
1220/**
1221 * ntfs_write_significant_bytes - write the significant bytes of a number
1222 * @dst:	destination buffer to write to
1223 * @dst_max:	pointer to last byte of destination buffer for bounds checking
1224 * @n:		number whose significant bytes to write
1225 *
1226 * Store in @dst, the minimum bytes of the number @n which are required to
1227 * identify @n unambiguously as a signed number, taking care not to exceed
1228 * @dest_max, the maximum position within @dst to which we are allowed to
1229 * write.
1230 *
1231 * This is used when building the mapping pairs array of a runlist to compress
1232 * a given logical cluster number (lcn) or a specific run length to the minimum
1233 * size possible.
1234 *
1235 * Return the number of bytes written on success.  On error, i.e. the
1236 * destination buffer @dst is too small, return -ENOSPC.
1237 */
1238static inline int ntfs_write_significant_bytes(s8 *dst, const s8 *dst_max,
1239		const s64 n)
1240{
1241	s64 l = n;
1242	int i;
1243	s8 j;
1244
1245	i = 0;
1246	do {
1247		if (unlikely(dst > dst_max))
1248			goto err_out;
1249		*dst++ = l & 0xffll;
1250		l >>= 8;
1251		i++;
1252	} while (l != 0 && l != -1);
1253	j = (n >> 8 * (i - 1)) & 0xff;
1254	/* If the sign bit is wrong, we need an extra byte. */
1255	if (n < 0 && j >= 0) {
1256		if (unlikely(dst > dst_max))
1257			goto err_out;
1258		i++;
1259		*dst = (s8)-1;
1260	} else if (n > 0 && j < 0) {
1261		if (unlikely(dst > dst_max))
1262			goto err_out;
1263		i++;
1264		*dst = (s8)0;
1265	}
1266	return i;
1267err_out:
1268	return -ENOSPC;
1269}
1270
1271/**
1272 * ntfs_mapping_pairs_build - build the mapping pairs array from a runlist
1273 * @vol:	ntfs volume (needed for the ntfs version)
1274 * @dst:	destination buffer to which to write the mapping pairs array
1275 * @dst_len:	size of destination buffer @dst in bytes
1276 * @rl:		locked runlist for which to build the mapping pairs array
1277 * @first_vcn:	first vcn which to include in the mapping pairs array
1278 * @last_vcn:	last vcn which to include in the mapping pairs array
1279 * @stop_vcn:	first vcn outside destination buffer on success or -ENOSPC
1280 *
1281 * Create the mapping pairs array from the locked runlist @rl, starting at vcn
1282 * @first_vcn and finishing with vcn @last_vcn and save the array in @dst.
1283 * @dst_len is the size of @dst in bytes and it should be at least equal to the
1284 * value obtained by calling ntfs_get_size_for_mapping_pairs().
1285 *
1286 * A @last_vcn of -1 means end of runlist and in that case the mapping pairs
1287 * array corresponding to the runlist starting at vcn @first_vcn and finishing
1288 * at the end of the runlist is created.
1289 *
1290 * If @rl is NULL, just write a single terminator byte to @dst.
1291 *
1292 * On success or -ENOSPC error, if @stop_vcn is not NULL, *@stop_vcn is set to
1293 * the first vcn outside the destination buffer.  Note that on error, @dst has
1294 * been filled with all the mapping pairs that will fit, thus it can be treated
1295 * as partial success, in that a new attribute extent needs to be created or
1296 * the next extent has to be used and the mapping pairs build has to be
1297 * continued with @first_vcn set to *@stop_vcn.
1298 *
1299 * Return 0 on success and -errno on error.  The following error codes are
1300 * defined:
1301 *	-EINVAL	- Run list contains unmapped elements.  Make sure to only pass
1302 *		  fully mapped runlists to this function.
1303 *	-EIO	- The runlist is corrupt.
1304 *	-ENOSPC	- The destination buffer is too small.
1305 *
1306 * Locking: @rl must be locked on entry (either for reading or writing), it
1307 *	    remains locked throughout, and is left locked upon return.
1308 */
1309int ntfs_mapping_pairs_build(const ntfs_volume *vol, s8 *dst,
1310		const int dst_len, const runlist_element *rl,
1311		const VCN first_vcn, const VCN last_vcn, VCN *const stop_vcn)
1312{
1313	LCN prev_lcn;
1314	s8 *dst_max, *dst_next;
1315	int err = -ENOSPC;
1316	bool the_end = false;
1317	s8 len_len, lcn_len;
1318
1319	BUG_ON(first_vcn < 0);
1320	BUG_ON(last_vcn < -1);
1321	BUG_ON(last_vcn >= 0 && first_vcn > last_vcn);
1322	BUG_ON(dst_len < 1);
1323	if (!rl) {
1324		BUG_ON(first_vcn);
1325		BUG_ON(last_vcn > 0);
1326		if (stop_vcn)
1327			*stop_vcn = 0;
1328		/* Terminator byte. */
1329		*dst = 0;
1330		return 0;
1331	}
1332	/* Skip to runlist element containing @first_vcn. */
1333	while (rl->length && first_vcn >= rl[1].vcn)
1334		rl++;
1335	if (unlikely((!rl->length && first_vcn > rl->vcn) ||
1336			first_vcn < rl->vcn))
1337		return -EINVAL;
1338	/*
1339	 * @dst_max is used for bounds checking in
1340	 * ntfs_write_significant_bytes().
1341	 */
1342	dst_max = dst + dst_len - 1;
1343	prev_lcn = 0;
1344	/* Do the first partial run if present. */
1345	if (first_vcn > rl->vcn) {
1346		s64 delta, length = rl->length;
1347
1348		/* We know rl->length != 0 already. */
1349		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1350			goto err_out;
1351		/*
1352		 * If @stop_vcn is given and finishes inside this run, cap the
1353		 * run length.
1354		 */
1355		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1356			s64 s1 = last_vcn + 1;
1357			if (unlikely(rl[1].vcn > s1))
1358				length = s1 - rl->vcn;
1359			the_end = true;
1360		}
1361		delta = first_vcn - rl->vcn;
1362		/* Write length. */
1363		len_len = ntfs_write_significant_bytes(dst + 1, dst_max,
1364				length - delta);
1365		if (unlikely(len_len < 0))
1366			goto size_err;
1367		/*
1368		 * If the logical cluster number (lcn) denotes a hole and we
1369		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1370		 * zero space.  On earlier NTFS versions we just write the lcn
1371		 * change.  FIXME: Do we need to write the lcn change or just
1372		 * the lcn in that case?  Not sure as I have never seen this
1373		 * case on NT4. - We assume that we just need to write the lcn
1374		 * change until someone tells us otherwise... (AIA)
1375		 */
1376		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1377			prev_lcn = rl->lcn;
1378			if (likely(rl->lcn >= 0))
1379				prev_lcn += delta;
1380			/* Write change in lcn. */
1381			lcn_len = ntfs_write_significant_bytes(dst + 1 +
1382					len_len, dst_max, prev_lcn);
1383			if (unlikely(lcn_len < 0))
1384				goto size_err;
1385		} else
1386			lcn_len = 0;
1387		dst_next = dst + len_len + lcn_len + 1;
1388		if (unlikely(dst_next > dst_max))
1389			goto size_err;
1390		/* Update header byte. */
1391		*dst = lcn_len << 4 | len_len;
1392		/* Position at next mapping pairs array element. */
1393		dst = dst_next;
1394		/* Go to next runlist element. */
1395		rl++;
1396	}
1397	/* Do the full runs. */
1398	for (; rl->length && !the_end; rl++) {
1399		s64 length = rl->length;
1400
1401		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
1402			goto err_out;
1403		/*
1404		 * If @stop_vcn is given and finishes inside this run, cap the
1405		 * run length.
1406		 */
1407		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
1408			s64 s1 = last_vcn + 1;
1409			if (unlikely(rl[1].vcn > s1))
1410				length = s1 - rl->vcn;
1411			the_end = true;
1412		}
1413		/* Write length. */
1414		len_len = ntfs_write_significant_bytes(dst + 1, dst_max,
1415				length);
1416		if (unlikely(len_len < 0))
1417			goto size_err;
1418		/*
1419		 * If the logical cluster number (lcn) denotes a hole and we
1420		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
1421		 * zero space.  On earlier NTFS versions we just write the lcn
1422		 * change.  FIXME: Do we need to write the lcn change or just
1423		 * the lcn in that case?  Not sure as I have never seen this
1424		 * case on NT4. - We assume that we just need to write the lcn
1425		 * change until someone tells us otherwise... (AIA)
1426		 */
1427		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
1428			/* Write change in lcn. */
1429			lcn_len = ntfs_write_significant_bytes(dst + 1 +
1430					len_len, dst_max, rl->lcn - prev_lcn);
1431			if (unlikely(lcn_len < 0))
1432				goto size_err;
1433			prev_lcn = rl->lcn;
1434		} else
1435			lcn_len = 0;
1436		dst_next = dst + len_len + lcn_len + 1;
1437		if (unlikely(dst_next > dst_max))
1438			goto size_err;
1439		/* Update header byte. */
1440		*dst = lcn_len << 4 | len_len;
1441		/* Position at next mapping pairs array element. */
1442		dst = dst_next;
1443	}
1444	/* Success. */
1445	err = 0;
1446size_err:
1447	/* Set stop vcn. */
1448	if (stop_vcn)
1449		*stop_vcn = rl->vcn;
1450	/* Add terminator byte. */
1451	*dst = 0;
1452	return err;
1453err_out:
1454	if (rl->lcn == LCN_RL_NOT_MAPPED)
1455		err = -EINVAL;
1456	else
1457		err = -EIO;
1458	return err;
1459}
1460
1461/**
1462 * ntfs_rl_truncate_nolock - truncate a runlist starting at a specified vcn
1463 * @vol:	ntfs volume (needed for error output)
1464 * @runlist:	runlist to truncate
1465 * @new_length:	the new length of the runlist in VCNs
1466 *
1467 * Truncate the runlist described by @runlist as well as the memory buffer
1468 * holding the runlist elements to a length of @new_length VCNs.
1469 *
1470 * If @new_length lies within the runlist, the runlist elements with VCNs of
1471 * @new_length and above are discarded.  As a special case if @new_length is
1472 * zero, the runlist is discarded and set to NULL.
1473 *
1474 * If @new_length lies beyond the runlist, a sparse runlist element is added to
1475 * the end of the runlist @runlist or if the last runlist element is a sparse
1476 * one already, this is extended.
1477 *
1478 * Note, no checking is done for unmapped runlist elements.  It is assumed that
1479 * the caller has mapped any elements that need to be mapped already.
1480 *
1481 * Return 0 on success and -errno on error.
1482 *
1483 * Locking: The caller must hold @runlist->lock for writing.
1484 */
1485int ntfs_rl_truncate_nolock(const ntfs_volume *vol, runlist *const runlist,
1486		const s64 new_length)
1487{
1488	runlist_element *rl;
1489	int old_size;
1490
1491	ntfs_debug("Entering for new_length 0x%llx.", (long long)new_length);
1492	BUG_ON(!runlist);
1493	BUG_ON(new_length < 0);
1494	rl = runlist->rl;
1495	if (!new_length) {
1496		ntfs_debug("Freeing runlist.");
1497		runlist->rl = NULL;
1498		if (rl)
1499			ntfs_free(rl);
1500		return 0;
1501	}
1502	if (unlikely(!rl)) {
1503		/*
1504		 * Create a runlist consisting of a sparse runlist element of
1505		 * length @new_length followed by a terminator runlist element.
1506		 */
1507		rl = ntfs_malloc_nofs(PAGE_SIZE);
1508		if (unlikely(!rl)) {
1509			ntfs_error(vol->sb, "Not enough memory to allocate "
1510					"runlist element buffer.");
1511			return -ENOMEM;
1512		}
1513		runlist->rl = rl;
1514		rl[1].length = rl->vcn = 0;
1515		rl->lcn = LCN_HOLE;
1516		rl[1].vcn = rl->length = new_length;
1517		rl[1].lcn = LCN_ENOENT;
1518		return 0;
1519	}
1520	BUG_ON(new_length < rl->vcn);
1521	/* Find @new_length in the runlist. */
1522	while (likely(rl->length && new_length >= rl[1].vcn))
1523		rl++;
1524	/*
1525	 * If not at the end of the runlist we need to shrink it.
1526	 * If at the end of the runlist we need to expand it.
1527	 */
1528	if (rl->length) {
1529		runlist_element *trl;
1530		bool is_end;
1531
1532		ntfs_debug("Shrinking runlist.");
1533		/* Determine the runlist size. */
1534		trl = rl + 1;
1535		while (likely(trl->length))
1536			trl++;
1537		old_size = trl - runlist->rl + 1;
1538		/* Truncate the run. */
1539		rl->length = new_length - rl->vcn;
1540		/*
1541		 * If a run was partially truncated, make the following runlist
1542		 * element a terminator.
1543		 */
1544		is_end = false;
1545		if (rl->length) {
1546			rl++;
1547			if (!rl->length)
1548				is_end = true;
1549			rl->vcn = new_length;
1550			rl->length = 0;
1551		}
1552		rl->lcn = LCN_ENOENT;
1553		/* Reallocate memory if necessary. */
1554		if (!is_end) {
1555			int new_size = rl - runlist->rl + 1;
1556			rl = ntfs_rl_realloc(runlist->rl, old_size, new_size);
1557			if (IS_ERR(rl))
1558				ntfs_warning(vol->sb, "Failed to shrink "
1559						"runlist buffer.  This just "
1560						"wastes a bit of memory "
1561						"temporarily so we ignore it "
1562						"and return success.");
1563			else
1564				runlist->rl = rl;
1565		}
1566	} else if (likely(/* !rl->length && */ new_length > rl->vcn)) {
1567		ntfs_debug("Expanding runlist.");
1568		/*
1569		 * If there is a previous runlist element and it is a sparse
1570		 * one, extend it.  Otherwise need to add a new, sparse runlist
1571		 * element.
1572		 */
1573		if ((rl > runlist->rl) && ((rl - 1)->lcn == LCN_HOLE))
1574			(rl - 1)->length = new_length - (rl - 1)->vcn;
1575		else {
1576			/* Determine the runlist size. */
1577			old_size = rl - runlist->rl + 1;
1578			/* Reallocate memory if necessary. */
1579			rl = ntfs_rl_realloc(runlist->rl, old_size,
1580					old_size + 1);
1581			if (IS_ERR(rl)) {
1582				ntfs_error(vol->sb, "Failed to expand runlist "
1583						"buffer, aborting.");
1584				return PTR_ERR(rl);
1585			}
1586			runlist->rl = rl;
1587			/*
1588			 * Set @rl to the same runlist element in the new
1589			 * runlist as before in the old runlist.
1590			 */
1591			rl += old_size - 1;
1592			/* Add a new, sparse runlist element. */
1593			rl->lcn = LCN_HOLE;
1594			rl->length = new_length - rl->vcn;
1595			/* Add a new terminator runlist element. */
1596			rl++;
1597			rl->length = 0;
1598		}
1599		rl->vcn = new_length;
1600		rl->lcn = LCN_ENOENT;
1601	} else /* if (unlikely(!rl->length && new_length == rl->vcn)) */ {
1602		/* Runlist already has same size as requested. */
1603		rl->lcn = LCN_ENOENT;
1604	}
1605	ntfs_debug("Done.");
1606	return 0;
1607}
1608
1609/**
1610 * ntfs_rl_punch_nolock - punch a hole into a runlist
1611 * @vol:	ntfs volume (needed for error output)
1612 * @runlist:	runlist to punch a hole into
1613 * @start:	starting VCN of the hole to be created
1614 * @length:	size of the hole to be created in units of clusters
1615 *
1616 * Punch a hole into the runlist @runlist starting at VCN @start and of size
1617 * @length clusters.
1618 *
1619 * Return 0 on success and -errno on error, in which case @runlist has not been
1620 * modified.
1621 *
1622 * If @start and/or @start + @length are outside the runlist return error code
1623 * -ENOENT.
1624 *
1625 * If the runlist contains unmapped or error elements between @start and @start
1626 * + @length return error code -EINVAL.
1627 *
1628 * Locking: The caller must hold @runlist->lock for writing.
1629 */
1630int ntfs_rl_punch_nolock(const ntfs_volume *vol, runlist *const runlist,
1631		const VCN start, const s64 length)
1632{
1633	const VCN end = start + length;
1634	s64 delta;
1635	runlist_element *rl, *rl_end, *rl_real_end, *trl;
1636	int old_size;
1637	bool lcn_fixup = false;
1638
1639	ntfs_debug("Entering for start 0x%llx, length 0x%llx.",
1640			(long long)start, (long long)length);
1641	BUG_ON(!runlist);
1642	BUG_ON(start < 0);
1643	BUG_ON(length < 0);
1644	BUG_ON(end < 0);
1645	rl = runlist->rl;
1646	if (unlikely(!rl)) {
1647		if (likely(!start && !length))
1648			return 0;
1649		return -ENOENT;
1650	}
1651	/* Find @start in the runlist. */
1652	while (likely(rl->length && start >= rl[1].vcn))
1653		rl++;
1654	rl_end = rl;
1655	/* Find @end in the runlist. */
1656	while (likely(rl_end->length && end >= rl_end[1].vcn)) {
1657		/* Verify there are no unmapped or error elements. */
1658		if (unlikely(rl_end->lcn < LCN_HOLE))
1659			return -EINVAL;
1660		rl_end++;
1661	}
1662	/* Check the last element. */
1663	if (unlikely(rl_end->length && rl_end->lcn < LCN_HOLE))
1664		return -EINVAL;
1665	/* This covers @start being out of bounds, too. */
1666	if (!rl_end->length && end > rl_end->vcn)
1667		return -ENOENT;
1668	if (!length)
1669		return 0;
1670	if (!rl->length)
1671		return -ENOENT;
1672	rl_real_end = rl_end;
1673	/* Determine the runlist size. */
1674	while (likely(rl_real_end->length))
1675		rl_real_end++;
1676	old_size = rl_real_end - runlist->rl + 1;
1677	/* If @start is in a hole simply extend the hole. */
1678	if (rl->lcn == LCN_HOLE) {
1679		/*
1680		 * If both @start and @end are in the same sparse run, we are
1681		 * done.
1682		 */
1683		if (end <= rl[1].vcn) {
1684			ntfs_debug("Done (requested hole is already sparse).");
1685			return 0;
1686		}
1687extend_hole:
1688		/* Extend the hole. */
1689		rl->length = end - rl->vcn;
1690		/* If @end is in a hole, merge it with the current one. */
1691		if (rl_end->lcn == LCN_HOLE) {
1692			rl_end++;
1693			rl->length = rl_end->vcn - rl->vcn;
1694		}
1695		/* We have done the hole.  Now deal with the remaining tail. */
1696		rl++;
1697		/* Cut out all runlist elements up to @end. */
1698		if (rl < rl_end)
1699			memmove(rl, rl_end, (rl_real_end - rl_end + 1) *
1700					sizeof(*rl));
1701		/* Adjust the beginning of the tail if necessary. */
1702		if (end > rl->vcn) {
1703			delta = end - rl->vcn;
1704			rl->vcn = end;
1705			rl->length -= delta;
1706			/* Only adjust the lcn if it is real. */
1707			if (rl->lcn >= 0)
1708				rl->lcn += delta;
1709		}
1710shrink_allocation:
1711		/* Reallocate memory if the allocation changed. */
1712		if (rl < rl_end) {
1713			rl = ntfs_rl_realloc(runlist->rl, old_size,
1714					old_size - (rl_end - rl));
1715			if (IS_ERR(rl))
1716				ntfs_warning(vol->sb, "Failed to shrink "
1717						"runlist buffer.  This just "
1718						"wastes a bit of memory "
1719						"temporarily so we ignore it "
1720						"and return success.");
1721			else
1722				runlist->rl = rl;
1723		}
1724		ntfs_debug("Done (extend hole).");
1725		return 0;
1726	}
1727	/*
1728	 * If @start is at the beginning of a run things are easier as there is
1729	 * no need to split the first run.
1730	 */
1731	if (start == rl->vcn) {
1732		/*
1733		 * @start is at the beginning of a run.
1734		 *
1735		 * If the previous run is sparse, extend its hole.
1736		 *
1737		 * If @end is not in the same run, switch the run to be sparse
1738		 * and extend the newly created hole.
1739		 *
1740		 * Thus both of these cases reduce the problem to the above
1741		 * case of "@start is in a hole".
1742		 */
1743		if (rl > runlist->rl && (rl - 1)->lcn == LCN_HOLE) {
1744			rl--;
1745			goto extend_hole;
1746		}
1747		if (end >= rl[1].vcn) {
1748			rl->lcn = LCN_HOLE;
1749			goto extend_hole;
1750		}
1751		/*
1752		 * The final case is when @end is in the same run as @start.
1753		 * For this need to split the run into two.  One run for the
1754		 * sparse region between the beginning of the old run, i.e.
1755		 * @start, and @end and one for the remaining non-sparse
1756		 * region, i.e. between @end and the end of the old run.
1757		 */
1758		trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 1);
1759		if (IS_ERR(trl))
1760			goto enomem_out;
1761		old_size++;
1762		if (runlist->rl != trl) {
1763			rl = trl + (rl - runlist->rl);
1764			rl_end = trl + (rl_end - runlist->rl);
1765			rl_real_end = trl + (rl_real_end - runlist->rl);
1766			runlist->rl = trl;
1767		}
1768split_end:
1769		/* Shift all the runs up by one. */
1770		memmove(rl + 1, rl, (rl_real_end - rl + 1) * sizeof(*rl));
1771		/* Finally, setup the two split runs. */
1772		rl->lcn = LCN_HOLE;
1773		rl->length = length;
1774		rl++;
1775		rl->vcn += length;
1776		/* Only adjust the lcn if it is real. */
1777		if (rl->lcn >= 0 || lcn_fixup)
1778			rl->lcn += length;
1779		rl->length -= length;
1780		ntfs_debug("Done (split one).");
1781		return 0;
1782	}
1783	/*
1784	 * @start is neither in a hole nor at the beginning of a run.
1785	 *
1786	 * If @end is in a hole, things are easier as simply truncating the run
1787	 * @start is in to end at @start - 1, deleting all runs after that up
1788	 * to @end, and finally extending the beginning of the run @end is in
1789	 * to be @start is all that is needed.
1790	 */
1791	if (rl_end->lcn == LCN_HOLE) {
1792		/* Truncate the run containing @start. */
1793		rl->length = start - rl->vcn;
1794		rl++;
1795		/* Cut out all runlist elements up to @end. */
1796		if (rl < rl_end)
1797			memmove(rl, rl_end, (rl_real_end - rl_end + 1) *
1798					sizeof(*rl));
1799		/* Extend the beginning of the run @end is in to be @start. */
1800		rl->vcn = start;
1801		rl->length = rl[1].vcn - start;
1802		goto shrink_allocation;
1803	}
1804	/* 
1805	 * If @end is not in a hole there are still two cases to distinguish.
1806	 * Either @end is or is not in the same run as @start.
1807	 *
1808	 * The second case is easier as it can be reduced to an already solved
1809	 * problem by truncating the run @start is in to end at @start - 1.
1810	 * Then, if @end is in the next run need to split the run into a sparse
1811	 * run followed by a non-sparse run (already covered above) and if @end
1812	 * is not in the next run switching it to be sparse, again reduces the
1813	 * problem to the already covered case of "@start is in a hole".
1814	 */
1815	if (end >= rl[1].vcn) {
1816		/*
1817		 * If @end is not in the next run, reduce the problem to the
1818		 * case of "@start is in a hole".
1819		 */
1820		if (rl[1].length && end >= rl[2].vcn) {
1821			/* Truncate the run containing @start. */
1822			rl->length = start - rl->vcn;
1823			rl++;
1824			rl->vcn = start;
1825			rl->lcn = LCN_HOLE;
1826			goto extend_hole;
1827		}
1828		trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 1);
1829		if (IS_ERR(trl))
1830			goto enomem_out;
1831		old_size++;
1832		if (runlist->rl != trl) {
1833			rl = trl + (rl - runlist->rl);
1834			rl_end = trl + (rl_end - runlist->rl);
1835			rl_real_end = trl + (rl_real_end - runlist->rl);
1836			runlist->rl = trl;
1837		}
1838		/* Truncate the run containing @start. */
1839		rl->length = start - rl->vcn;
1840		rl++;
1841		/*
1842		 * @end is in the next run, reduce the problem to the case
1843		 * where "@start is at the beginning of a run and @end is in
1844		 * the same run as @start".
1845		 */
1846		delta = rl->vcn - start;
1847		rl->vcn = start;
1848		if (rl->lcn >= 0) {
1849			rl->lcn -= delta;
1850			/* Need this in case the lcn just became negative. */
1851			lcn_fixup = true;
1852		}
1853		rl->length += delta;
1854		goto split_end;
1855	}
1856	/*
1857	 * The first case from above, i.e. @end is in the same run as @start.
1858	 * We need to split the run into three.  One run for the non-sparse
1859	 * region between the beginning of the old run and @start, one for the
1860	 * sparse region between @start and @end, and one for the remaining
1861	 * non-sparse region, i.e. between @end and the end of the old run.
1862	 */
1863	trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 2);
1864	if (IS_ERR(trl))
1865		goto enomem_out;
1866	old_size += 2;
1867	if (runlist->rl != trl) {
1868		rl = trl + (rl - runlist->rl);
1869		rl_end = trl + (rl_end - runlist->rl);
1870		rl_real_end = trl + (rl_real_end - runlist->rl);
1871		runlist->rl = trl;
1872	}
1873	/* Shift all the runs up by two. */
1874	memmove(rl + 2, rl, (rl_real_end - rl + 1) * sizeof(*rl));
1875	/* Finally, setup the three split runs. */
1876	rl->length = start - rl->vcn;
1877	rl++;
1878	rl->vcn = start;
1879	rl->lcn = LCN_HOLE;
1880	rl->length = length;
1881	rl++;
1882	delta = end - rl->vcn;
1883	rl->vcn = end;
1884	rl->lcn += delta;
1885	rl->length -= delta;
1886	ntfs_debug("Done (split both).");
1887	return 0;
1888enomem_out:
1889	ntfs_error(vol->sb, "Not enough memory to extend runlist buffer.");
1890	return -ENOMEM;
1891}
1892
1893#endif /* NTFS_RW */