xref: /freebsd/contrib/xz/src/liblzma/common/index.c (revision 7ef62cebc2f965b0f640263e179276928885e33d)
1 ///////////////////////////////////////////////////////////////////////////////
2 //
3 /// \file       index.c
4 /// \brief      Handling of .xz Indexes and some other Stream information
5 //
6 //  Author:     Lasse Collin
7 //
8 //  This file has been put into the public domain.
9 //  You can do whatever you want with this file.
10 //
11 ///////////////////////////////////////////////////////////////////////////////
12 
13 #include "common.h"
14 #include "index.h"
15 #include "stream_flags_common.h"
16 
17 
18 /// \brief      How many Records to allocate at once
19 ///
20 /// This should be big enough to avoid making lots of tiny allocations
21 /// but small enough to avoid too much unused memory at once.
22 #define INDEX_GROUP_SIZE 512
23 
24 
25 /// \brief      How many Records can be allocated at once at maximum
26 #define PREALLOC_MAX ((SIZE_MAX - sizeof(index_group)) / sizeof(index_record))
27 
28 
29 /// \brief      Base structure for index_stream and index_group structures
30 typedef struct index_tree_node_s index_tree_node;
31 struct index_tree_node_s {
32 	/// Uncompressed start offset of this Stream (relative to the
33 	/// beginning of the file) or Block (relative to the beginning
34 	/// of the Stream)
35 	lzma_vli uncompressed_base;
36 
37 	/// Compressed start offset of this Stream or Block
38 	lzma_vli compressed_base;
39 
40 	index_tree_node *parent;
41 	index_tree_node *left;
42 	index_tree_node *right;
43 };
44 
45 
46 /// \brief      AVL tree to hold index_stream or index_group structures
47 typedef struct {
48 	/// Root node
49 	index_tree_node *root;
50 
51 	/// Leftmost node. Since the tree will be filled sequentially,
52 	/// this won't change after the first node has been added to
53 	/// the tree.
54 	index_tree_node *leftmost;
55 
56 	/// The rightmost node in the tree. Since the tree is filled
57 	/// sequentially, this is always the node where to add the new data.
58 	index_tree_node *rightmost;
59 
60 	/// Number of nodes in the tree
61 	uint32_t count;
62 
63 } index_tree;
64 
65 
66 typedef struct {
67 	lzma_vli uncompressed_sum;
68 	lzma_vli unpadded_sum;
69 } index_record;
70 
71 
72 typedef struct {
73 	/// Every Record group is part of index_stream.groups tree.
74 	index_tree_node node;
75 
76 	/// Number of Blocks in this Stream before this group.
77 	lzma_vli number_base;
78 
79 	/// Number of Records that can be put in records[].
80 	size_t allocated;
81 
82 	/// Index of the last Record in use.
83 	size_t last;
84 
85 	/// The sizes in this array are stored as cumulative sums relative
86 	/// to the beginning of the Stream. This makes it possible to
87 	/// use binary search in lzma_index_locate().
88 	///
89 	/// Note that the cumulative summing is done specially for
90 	/// unpadded_sum: The previous value is rounded up to the next
91 	/// multiple of four before adding the Unpadded Size of the new
92 	/// Block. The total encoded size of the Blocks in the Stream
93 	/// is records[last].unpadded_sum in the last Record group of
94 	/// the Stream.
95 	///
96 	/// For example, if the Unpadded Sizes are 39, 57, and 81, the
97 	/// stored values are 39, 97 (40 + 57), and 181 (100 + 181).
98 	/// The total encoded size of these Blocks is 184.
99 	///
100 	/// This is a flexible array, because it makes easy to optimize
101 	/// memory usage in case someone concatenates many Streams that
102 	/// have only one or few Blocks.
103 	index_record records[];
104 
105 } index_group;
106 
107 
108 typedef struct {
109 	/// Every index_stream is a node in the tree of Streams.
110 	index_tree_node node;
111 
112 	/// Number of this Stream (first one is 1)
113 	uint32_t number;
114 
115 	/// Total number of Blocks before this Stream
116 	lzma_vli block_number_base;
117 
118 	/// Record groups of this Stream are stored in a tree.
119 	/// It's a T-tree with AVL-tree balancing. There are
120 	/// INDEX_GROUP_SIZE Records per node by default.
121 	/// This keeps the number of memory allocations reasonable
122 	/// and finding a Record is fast.
123 	index_tree groups;
124 
125 	/// Number of Records in this Stream
126 	lzma_vli record_count;
127 
128 	/// Size of the List of Records field in this Stream. This is used
129 	/// together with record_count to calculate the size of the Index
130 	/// field and thus the total size of the Stream.
131 	lzma_vli index_list_size;
132 
133 	/// Stream Flags of this Stream. This is meaningful only if
134 	/// the Stream Flags have been told us with lzma_index_stream_flags().
135 	/// Initially stream_flags.version is set to UINT32_MAX to indicate
136 	/// that the Stream Flags are unknown.
137 	lzma_stream_flags stream_flags;
138 
139 	/// Amount of Stream Padding after this Stream. This defaults to
140 	/// zero and can be set with lzma_index_stream_padding().
141 	lzma_vli stream_padding;
142 
143 } index_stream;
144 
145 
146 struct lzma_index_s {
147 	/// AVL-tree containing the Stream(s). Often there is just one
148 	/// Stream, but using a tree keeps lookups fast even when there
149 	/// are many concatenated Streams.
150 	index_tree streams;
151 
152 	/// Uncompressed size of all the Blocks in the Stream(s)
153 	lzma_vli uncompressed_size;
154 
155 	/// Total size of all the Blocks in the Stream(s)
156 	lzma_vli total_size;
157 
158 	/// Total number of Records in all Streams in this lzma_index
159 	lzma_vli record_count;
160 
161 	/// Size of the List of Records field if all the Streams in this
162 	/// lzma_index were packed into a single Stream (makes it simpler to
163 	/// take many .xz files and combine them into a single Stream).
164 	///
165 	/// This value together with record_count is needed to calculate
166 	/// Backward Size that is stored into Stream Footer.
167 	lzma_vli index_list_size;
168 
169 	/// How many Records to allocate at once in lzma_index_append().
170 	/// This defaults to INDEX_GROUP_SIZE but can be overridden with
171 	/// lzma_index_prealloc().
172 	size_t prealloc;
173 
174 	/// Bitmask indicating what integrity check types have been used
175 	/// as set by lzma_index_stream_flags(). The bit of the last Stream
176 	/// is not included here, since it is possible to change it by
177 	/// calling lzma_index_stream_flags() again.
178 	uint32_t checks;
179 };
180 
181 
182 static void
183 index_tree_init(index_tree *tree)
184 {
185 	tree->root = NULL;
186 	tree->leftmost = NULL;
187 	tree->rightmost = NULL;
188 	tree->count = 0;
189 	return;
190 }
191 
192 
193 /// Helper for index_tree_end()
194 static void
195 index_tree_node_end(index_tree_node *node, const lzma_allocator *allocator,
196 		void (*free_func)(void *node, const lzma_allocator *allocator))
197 {
198 	// The tree won't ever be very huge, so recursion should be fine.
199 	// 20 levels in the tree is likely quite a lot already in practice.
200 	if (node->left != NULL)
201 		index_tree_node_end(node->left, allocator, free_func);
202 
203 	if (node->right != NULL)
204 		index_tree_node_end(node->right, allocator, free_func);
205 
206 	free_func(node, allocator);
207 	return;
208 }
209 
210 
211 /// Free the memory allocated for a tree. Each node is freed using the
212 /// given free_func which is either &lzma_free or &index_stream_end.
213 /// The latter is used to free the Record groups from each index_stream
214 /// before freeing the index_stream itself.
215 static void
216 index_tree_end(index_tree *tree, const lzma_allocator *allocator,
217 		void (*free_func)(void *node, const lzma_allocator *allocator))
218 {
219 	assert(free_func != NULL);
220 
221 	if (tree->root != NULL)
222 		index_tree_node_end(tree->root, allocator, free_func);
223 
224 	return;
225 }
226 
227 
228 /// Add a new node to the tree. node->uncompressed_base and
229 /// node->compressed_base must have been set by the caller already.
230 static void
231 index_tree_append(index_tree *tree, index_tree_node *node)
232 {
233 	node->parent = tree->rightmost;
234 	node->left = NULL;
235 	node->right = NULL;
236 
237 	++tree->count;
238 
239 	// Handle the special case of adding the first node.
240 	if (tree->root == NULL) {
241 		tree->root = node;
242 		tree->leftmost = node;
243 		tree->rightmost = node;
244 		return;
245 	}
246 
247 	// The tree is always filled sequentially.
248 	assert(tree->rightmost->uncompressed_base <= node->uncompressed_base);
249 	assert(tree->rightmost->compressed_base < node->compressed_base);
250 
251 	// Add the new node after the rightmost node. It's the correct
252 	// place due to the reason above.
253 	tree->rightmost->right = node;
254 	tree->rightmost = node;
255 
256 	// Balance the AVL-tree if needed. We don't need to keep the balance
257 	// factors in nodes, because we always fill the tree sequentially,
258 	// and thus know the state of the tree just by looking at the node
259 	// count. From the node count we can calculate how many steps to go
260 	// up in the tree to find the rotation root.
261 	uint32_t up = tree->count ^ (UINT32_C(1) << bsr32(tree->count));
262 	if (up != 0) {
263 		// Locate the root node for the rotation.
264 		up = ctz32(tree->count) + 2;
265 		do {
266 			node = node->parent;
267 		} while (--up > 0);
268 
269 		// Rotate left using node as the rotation root.
270 		index_tree_node *pivot = node->right;
271 
272 		if (node->parent == NULL) {
273 			tree->root = pivot;
274 		} else {
275 			assert(node->parent->right == node);
276 			node->parent->right = pivot;
277 		}
278 
279 		pivot->parent = node->parent;
280 
281 		node->right = pivot->left;
282 		if (node->right != NULL)
283 			node->right->parent = node;
284 
285 		pivot->left = node;
286 		node->parent = pivot;
287 	}
288 
289 	return;
290 }
291 
292 
293 /// Get the next node in the tree. Return NULL if there are no more nodes.
294 static void *
295 index_tree_next(const index_tree_node *node)
296 {
297 	if (node->right != NULL) {
298 		node = node->right;
299 		while (node->left != NULL)
300 			node = node->left;
301 
302 		return (void *)(node);
303 	}
304 
305 	while (node->parent != NULL && node->parent->right == node)
306 		node = node->parent;
307 
308 	return (void *)(node->parent);
309 }
310 
311 
312 /// Locate a node that contains the given uncompressed offset. It is
313 /// caller's job to check that target is not bigger than the uncompressed
314 /// size of the tree (the last node would be returned in that case still).
315 static void *
316 index_tree_locate(const index_tree *tree, lzma_vli target)
317 {
318 	const index_tree_node *result = NULL;
319 	const index_tree_node *node = tree->root;
320 
321 	assert(tree->leftmost == NULL
322 			|| tree->leftmost->uncompressed_base == 0);
323 
324 	// Consecutive nodes may have the same uncompressed_base.
325 	// We must pick the rightmost one.
326 	while (node != NULL) {
327 		if (node->uncompressed_base > target) {
328 			node = node->left;
329 		} else {
330 			result = node;
331 			node = node->right;
332 		}
333 	}
334 
335 	return (void *)(result);
336 }
337 
338 
339 /// Allocate and initialize a new Stream using the given base offsets.
340 static index_stream *
341 index_stream_init(lzma_vli compressed_base, lzma_vli uncompressed_base,
342 		uint32_t stream_number, lzma_vli block_number_base,
343 		const lzma_allocator *allocator)
344 {
345 	index_stream *s = lzma_alloc(sizeof(index_stream), allocator);
346 	if (s == NULL)
347 		return NULL;
348 
349 	s->node.uncompressed_base = uncompressed_base;
350 	s->node.compressed_base = compressed_base;
351 	s->node.parent = NULL;
352 	s->node.left = NULL;
353 	s->node.right = NULL;
354 
355 	s->number = stream_number;
356 	s->block_number_base = block_number_base;
357 
358 	index_tree_init(&s->groups);
359 
360 	s->record_count = 0;
361 	s->index_list_size = 0;
362 	s->stream_flags.version = UINT32_MAX;
363 	s->stream_padding = 0;
364 
365 	return s;
366 }
367 
368 
369 /// Free the memory allocated for a Stream and its Record groups.
370 static void
371 index_stream_end(void *node, const lzma_allocator *allocator)
372 {
373 	index_stream *s = node;
374 	index_tree_end(&s->groups, allocator, &lzma_free);
375 	lzma_free(s, allocator);
376 	return;
377 }
378 
379 
380 static lzma_index *
381 index_init_plain(const lzma_allocator *allocator)
382 {
383 	lzma_index *i = lzma_alloc(sizeof(lzma_index), allocator);
384 	if (i != NULL) {
385 		index_tree_init(&i->streams);
386 		i->uncompressed_size = 0;
387 		i->total_size = 0;
388 		i->record_count = 0;
389 		i->index_list_size = 0;
390 		i->prealloc = INDEX_GROUP_SIZE;
391 		i->checks = 0;
392 	}
393 
394 	return i;
395 }
396 
397 
398 extern LZMA_API(lzma_index *)
399 lzma_index_init(const lzma_allocator *allocator)
400 {
401 	lzma_index *i = index_init_plain(allocator);
402 	if (i == NULL)
403 		return NULL;
404 
405 	index_stream *s = index_stream_init(0, 0, 1, 0, allocator);
406 	if (s == NULL) {
407 		lzma_free(i, allocator);
408 		return NULL;
409 	}
410 
411 	index_tree_append(&i->streams, &s->node);
412 
413 	return i;
414 }
415 
416 
417 extern LZMA_API(void)
418 lzma_index_end(lzma_index *i, const lzma_allocator *allocator)
419 {
420 	// NOTE: If you modify this function, check also the bottom
421 	// of lzma_index_cat().
422 	if (i != NULL) {
423 		index_tree_end(&i->streams, allocator, &index_stream_end);
424 		lzma_free(i, allocator);
425 	}
426 
427 	return;
428 }
429 
430 
431 extern void
432 lzma_index_prealloc(lzma_index *i, lzma_vli records)
433 {
434 	if (records > PREALLOC_MAX)
435 		records = PREALLOC_MAX;
436 
437 	i->prealloc = (size_t)(records);
438 	return;
439 }
440 
441 
442 extern LZMA_API(uint64_t)
443 lzma_index_memusage(lzma_vli streams, lzma_vli blocks)
444 {
445 	// This calculates an upper bound that is only a little bit
446 	// bigger than the exact maximum memory usage with the given
447 	// parameters.
448 
449 	// Typical malloc() overhead is 2 * sizeof(void *) but we take
450 	// a little bit extra just in case. Using LZMA_MEMUSAGE_BASE
451 	// instead would give too inaccurate estimate.
452 	const size_t alloc_overhead = 4 * sizeof(void *);
453 
454 	// Amount of memory needed for each Stream base structures.
455 	// We assume that every Stream has at least one Block and
456 	// thus at least one group.
457 	const size_t stream_base = sizeof(index_stream)
458 			+ sizeof(index_group) + 2 * alloc_overhead;
459 
460 	// Amount of memory needed per group.
461 	const size_t group_base = sizeof(index_group)
462 			+ INDEX_GROUP_SIZE * sizeof(index_record)
463 			+ alloc_overhead;
464 
465 	// Number of groups. There may actually be more, but that overhead
466 	// has been taken into account in stream_base already.
467 	const lzma_vli groups
468 			= (blocks + INDEX_GROUP_SIZE - 1) / INDEX_GROUP_SIZE;
469 
470 	// Memory used by index_stream and index_group structures.
471 	const uint64_t streams_mem = streams * stream_base;
472 	const uint64_t groups_mem = groups * group_base;
473 
474 	// Memory used by the base structure.
475 	const uint64_t index_base = sizeof(lzma_index) + alloc_overhead;
476 
477 	// Validate the arguments and catch integer overflows.
478 	// Maximum number of Streams is "only" UINT32_MAX, because
479 	// that limit is used by the tree containing the Streams.
480 	const uint64_t limit = UINT64_MAX - index_base;
481 	if (streams == 0 || streams > UINT32_MAX || blocks > LZMA_VLI_MAX
482 			|| streams > limit / stream_base
483 			|| groups > limit / group_base
484 			|| limit - streams_mem < groups_mem)
485 		return UINT64_MAX;
486 
487 	return index_base + streams_mem + groups_mem;
488 }
489 
490 
491 extern LZMA_API(uint64_t)
492 lzma_index_memused(const lzma_index *i)
493 {
494 	return lzma_index_memusage(i->streams.count, i->record_count);
495 }
496 
497 
498 extern LZMA_API(lzma_vli)
499 lzma_index_block_count(const lzma_index *i)
500 {
501 	return i->record_count;
502 }
503 
504 
505 extern LZMA_API(lzma_vli)
506 lzma_index_stream_count(const lzma_index *i)
507 {
508 	return i->streams.count;
509 }
510 
511 
512 extern LZMA_API(lzma_vli)
513 lzma_index_size(const lzma_index *i)
514 {
515 	return index_size(i->record_count, i->index_list_size);
516 }
517 
518 
519 extern LZMA_API(lzma_vli)
520 lzma_index_total_size(const lzma_index *i)
521 {
522 	return i->total_size;
523 }
524 
525 
526 extern LZMA_API(lzma_vli)
527 lzma_index_stream_size(const lzma_index *i)
528 {
529 	// Stream Header + Blocks + Index + Stream Footer
530 	return LZMA_STREAM_HEADER_SIZE + i->total_size
531 			+ index_size(i->record_count, i->index_list_size)
532 			+ LZMA_STREAM_HEADER_SIZE;
533 }
534 
535 
536 static lzma_vli
537 index_file_size(lzma_vli compressed_base, lzma_vli unpadded_sum,
538 		lzma_vli record_count, lzma_vli index_list_size,
539 		lzma_vli stream_padding)
540 {
541 	// Earlier Streams and Stream Paddings + Stream Header
542 	// + Blocks + Index + Stream Footer + Stream Padding
543 	//
544 	// This might go over LZMA_VLI_MAX due to too big unpadded_sum
545 	// when this function is used in lzma_index_append().
546 	lzma_vli file_size = compressed_base + 2 * LZMA_STREAM_HEADER_SIZE
547 			+ stream_padding + vli_ceil4(unpadded_sum);
548 	if (file_size > LZMA_VLI_MAX)
549 		return LZMA_VLI_UNKNOWN;
550 
551 	// The same applies here.
552 	file_size += index_size(record_count, index_list_size);
553 	if (file_size > LZMA_VLI_MAX)
554 		return LZMA_VLI_UNKNOWN;
555 
556 	return file_size;
557 }
558 
559 
560 extern LZMA_API(lzma_vli)
561 lzma_index_file_size(const lzma_index *i)
562 {
563 	const index_stream *s = (const index_stream *)(i->streams.rightmost);
564 	const index_group *g = (const index_group *)(s->groups.rightmost);
565 	return index_file_size(s->node.compressed_base,
566 			g == NULL ? 0 : g->records[g->last].unpadded_sum,
567 			s->record_count, s->index_list_size,
568 			s->stream_padding);
569 }
570 
571 
572 extern LZMA_API(lzma_vli)
573 lzma_index_uncompressed_size(const lzma_index *i)
574 {
575 	return i->uncompressed_size;
576 }
577 
578 
579 extern LZMA_API(uint32_t)
580 lzma_index_checks(const lzma_index *i)
581 {
582 	uint32_t checks = i->checks;
583 
584 	// Get the type of the Check of the last Stream too.
585 	const index_stream *s = (const index_stream *)(i->streams.rightmost);
586 	if (s->stream_flags.version != UINT32_MAX)
587 		checks |= UINT32_C(1) << s->stream_flags.check;
588 
589 	return checks;
590 }
591 
592 
593 extern uint32_t
594 lzma_index_padding_size(const lzma_index *i)
595 {
596 	return (LZMA_VLI_C(4) - index_size_unpadded(
597 			i->record_count, i->index_list_size)) & 3;
598 }
599 
600 
601 extern LZMA_API(lzma_ret)
602 lzma_index_stream_flags(lzma_index *i, const lzma_stream_flags *stream_flags)
603 {
604 	if (i == NULL || stream_flags == NULL)
605 		return LZMA_PROG_ERROR;
606 
607 	// Validate the Stream Flags.
608 	return_if_error(lzma_stream_flags_compare(
609 			stream_flags, stream_flags));
610 
611 	index_stream *s = (index_stream *)(i->streams.rightmost);
612 	s->stream_flags = *stream_flags;
613 
614 	return LZMA_OK;
615 }
616 
617 
618 extern LZMA_API(lzma_ret)
619 lzma_index_stream_padding(lzma_index *i, lzma_vli stream_padding)
620 {
621 	if (i == NULL || stream_padding > LZMA_VLI_MAX
622 			|| (stream_padding & 3) != 0)
623 		return LZMA_PROG_ERROR;
624 
625 	index_stream *s = (index_stream *)(i->streams.rightmost);
626 
627 	// Check that the new value won't make the file grow too big.
628 	const lzma_vli old_stream_padding = s->stream_padding;
629 	s->stream_padding = 0;
630 	if (lzma_index_file_size(i) + stream_padding > LZMA_VLI_MAX) {
631 		s->stream_padding = old_stream_padding;
632 		return LZMA_DATA_ERROR;
633 	}
634 
635 	s->stream_padding = stream_padding;
636 	return LZMA_OK;
637 }
638 
639 
640 extern LZMA_API(lzma_ret)
641 lzma_index_append(lzma_index *i, const lzma_allocator *allocator,
642 		lzma_vli unpadded_size, lzma_vli uncompressed_size)
643 {
644 	// Validate.
645 	if (i == NULL || unpadded_size < UNPADDED_SIZE_MIN
646 			|| unpadded_size > UNPADDED_SIZE_MAX
647 			|| uncompressed_size > LZMA_VLI_MAX)
648 		return LZMA_PROG_ERROR;
649 
650 	index_stream *s = (index_stream *)(i->streams.rightmost);
651 	index_group *g = (index_group *)(s->groups.rightmost);
652 
653 	const lzma_vli compressed_base = g == NULL ? 0
654 			: vli_ceil4(g->records[g->last].unpadded_sum);
655 	const lzma_vli uncompressed_base = g == NULL ? 0
656 			: g->records[g->last].uncompressed_sum;
657 	const uint32_t index_list_size_add = lzma_vli_size(unpadded_size)
658 			+ lzma_vli_size(uncompressed_size);
659 
660 	// Check that uncompressed size will not overflow.
661 	if (uncompressed_base + uncompressed_size > LZMA_VLI_MAX)
662 		return LZMA_DATA_ERROR;
663 
664 	// Check that the file size will stay within limits.
665 	if (index_file_size(s->node.compressed_base,
666 			compressed_base + unpadded_size, s->record_count + 1,
667 			s->index_list_size + index_list_size_add,
668 			s->stream_padding) == LZMA_VLI_UNKNOWN)
669 		return LZMA_DATA_ERROR;
670 
671 	// The size of the Index field must not exceed the maximum value
672 	// that can be stored in the Backward Size field.
673 	if (index_size(i->record_count + 1,
674 			i->index_list_size + index_list_size_add)
675 			> LZMA_BACKWARD_SIZE_MAX)
676 		return LZMA_DATA_ERROR;
677 
678 	if (g != NULL && g->last + 1 < g->allocated) {
679 		// There is space in the last group at least for one Record.
680 		++g->last;
681 	} else {
682 		// We need to allocate a new group.
683 		g = lzma_alloc(sizeof(index_group)
684 				+ i->prealloc * sizeof(index_record),
685 				allocator);
686 		if (g == NULL)
687 			return LZMA_MEM_ERROR;
688 
689 		g->last = 0;
690 		g->allocated = i->prealloc;
691 
692 		// Reset prealloc so that if the application happens to
693 		// add new Records, the allocation size will be sane.
694 		i->prealloc = INDEX_GROUP_SIZE;
695 
696 		// Set the start offsets of this group.
697 		g->node.uncompressed_base = uncompressed_base;
698 		g->node.compressed_base = compressed_base;
699 		g->number_base = s->record_count + 1;
700 
701 		// Add the new group to the Stream.
702 		index_tree_append(&s->groups, &g->node);
703 	}
704 
705 	// Add the new Record to the group.
706 	g->records[g->last].uncompressed_sum
707 			= uncompressed_base + uncompressed_size;
708 	g->records[g->last].unpadded_sum
709 			= compressed_base + unpadded_size;
710 
711 	// Update the totals.
712 	++s->record_count;
713 	s->index_list_size += index_list_size_add;
714 
715 	i->total_size += vli_ceil4(unpadded_size);
716 	i->uncompressed_size += uncompressed_size;
717 	++i->record_count;
718 	i->index_list_size += index_list_size_add;
719 
720 	return LZMA_OK;
721 }
722 
723 
724 /// Structure to pass info to index_cat_helper()
725 typedef struct {
726 	/// Uncompressed size of the destination
727 	lzma_vli uncompressed_size;
728 
729 	/// Compressed file size of the destination
730 	lzma_vli file_size;
731 
732 	/// Same as above but for Block numbers
733 	lzma_vli block_number_add;
734 
735 	/// Number of Streams that were in the destination index before we
736 	/// started appending new Streams from the source index. This is
737 	/// used to fix the Stream numbering.
738 	uint32_t stream_number_add;
739 
740 	/// Destination index' Stream tree
741 	index_tree *streams;
742 
743 } index_cat_info;
744 
745 
746 /// Add the Stream nodes from the source index to dest using recursion.
747 /// Simplest iterative traversal of the source tree wouldn't work, because
748 /// we update the pointers in nodes when moving them to the destination tree.
749 static void
750 index_cat_helper(const index_cat_info *info, index_stream *this)
751 {
752 	index_stream *left = (index_stream *)(this->node.left);
753 	index_stream *right = (index_stream *)(this->node.right);
754 
755 	if (left != NULL)
756 		index_cat_helper(info, left);
757 
758 	this->node.uncompressed_base += info->uncompressed_size;
759 	this->node.compressed_base += info->file_size;
760 	this->number += info->stream_number_add;
761 	this->block_number_base += info->block_number_add;
762 	index_tree_append(info->streams, &this->node);
763 
764 	if (right != NULL)
765 		index_cat_helper(info, right);
766 
767 	return;
768 }
769 
770 
771 extern LZMA_API(lzma_ret)
772 lzma_index_cat(lzma_index *restrict dest, lzma_index *restrict src,
773 		const lzma_allocator *allocator)
774 {
775 	if (dest == NULL || src == NULL)
776 		return LZMA_PROG_ERROR;
777 
778 	const lzma_vli dest_file_size = lzma_index_file_size(dest);
779 
780 	// Check that we don't exceed the file size limits.
781 	if (dest_file_size + lzma_index_file_size(src) > LZMA_VLI_MAX
782 			|| dest->uncompressed_size + src->uncompressed_size
783 				> LZMA_VLI_MAX)
784 		return LZMA_DATA_ERROR;
785 
786 	// Check that the encoded size of the combined lzma_indexes stays
787 	// within limits. In theory, this should be done only if we know
788 	// that the user plans to actually combine the Streams and thus
789 	// construct a single Index (probably rare). However, exceeding
790 	// this limit is quite theoretical, so we do this check always
791 	// to simplify things elsewhere.
792 	{
793 		const lzma_vli dest_size = index_size_unpadded(
794 				dest->record_count, dest->index_list_size);
795 		const lzma_vli src_size = index_size_unpadded(
796 				src->record_count, src->index_list_size);
797 		if (vli_ceil4(dest_size + src_size) > LZMA_BACKWARD_SIZE_MAX)
798 			return LZMA_DATA_ERROR;
799 	}
800 
801 	// Optimize the last group to minimize memory usage. Allocation has
802 	// to be done before modifying dest or src.
803 	{
804 		index_stream *s = (index_stream *)(dest->streams.rightmost);
805 		index_group *g = (index_group *)(s->groups.rightmost);
806 		if (g != NULL && g->last + 1 < g->allocated) {
807 			assert(g->node.left == NULL);
808 			assert(g->node.right == NULL);
809 
810 			index_group *newg = lzma_alloc(sizeof(index_group)
811 					+ (g->last + 1)
812 					* sizeof(index_record),
813 					allocator);
814 			if (newg == NULL)
815 				return LZMA_MEM_ERROR;
816 
817 			newg->node = g->node;
818 			newg->allocated = g->last + 1;
819 			newg->last = g->last;
820 			newg->number_base = g->number_base;
821 
822 			memcpy(newg->records, g->records, newg->allocated
823 					* sizeof(index_record));
824 
825 			if (g->node.parent != NULL) {
826 				assert(g->node.parent->right == &g->node);
827 				g->node.parent->right = &newg->node;
828 			}
829 
830 			if (s->groups.leftmost == &g->node) {
831 				assert(s->groups.root == &g->node);
832 				s->groups.leftmost = &newg->node;
833 				s->groups.root = &newg->node;
834 			}
835 
836 			assert(s->groups.rightmost == &g->node);
837 			s->groups.rightmost = &newg->node;
838 
839 			lzma_free(g, allocator);
840 
841 			// NOTE: newg isn't leaked here because
842 			// newg == (void *)&newg->node.
843 		}
844 	}
845 
846 	// dest->checks includes the check types of all except the last Stream
847 	// in dest. Set the bit for the check type of the last Stream now so
848 	// that it won't get lost when Stream(s) from src are appended to dest.
849 	dest->checks = lzma_index_checks(dest);
850 
851 	// Add all the Streams from src to dest. Update the base offsets
852 	// of each Stream from src.
853 	const index_cat_info info = {
854 		.uncompressed_size = dest->uncompressed_size,
855 		.file_size = dest_file_size,
856 		.stream_number_add = dest->streams.count,
857 		.block_number_add = dest->record_count,
858 		.streams = &dest->streams,
859 	};
860 	index_cat_helper(&info, (index_stream *)(src->streams.root));
861 
862 	// Update info about all the combined Streams.
863 	dest->uncompressed_size += src->uncompressed_size;
864 	dest->total_size += src->total_size;
865 	dest->record_count += src->record_count;
866 	dest->index_list_size += src->index_list_size;
867 	dest->checks |= src->checks;
868 
869 	// There's nothing else left in src than the base structure.
870 	lzma_free(src, allocator);
871 
872 	return LZMA_OK;
873 }
874 
875 
876 /// Duplicate an index_stream.
877 static index_stream *
878 index_dup_stream(const index_stream *src, const lzma_allocator *allocator)
879 {
880 	// Catch a somewhat theoretical integer overflow.
881 	if (src->record_count > PREALLOC_MAX)
882 		return NULL;
883 
884 	// Allocate and initialize a new Stream.
885 	index_stream *dest = index_stream_init(src->node.compressed_base,
886 			src->node.uncompressed_base, src->number,
887 			src->block_number_base, allocator);
888 	if (dest == NULL)
889 		return NULL;
890 
891 	// Copy the overall information.
892 	dest->record_count = src->record_count;
893 	dest->index_list_size = src->index_list_size;
894 	dest->stream_flags = src->stream_flags;
895 	dest->stream_padding = src->stream_padding;
896 
897 	// Return if there are no groups to duplicate.
898 	if (src->groups.leftmost == NULL)
899 		return dest;
900 
901 	// Allocate memory for the Records. We put all the Records into
902 	// a single group. It's simplest and also tends to make
903 	// lzma_index_locate() a little bit faster with very big Indexes.
904 	index_group *destg = lzma_alloc(sizeof(index_group)
905 			+ src->record_count * sizeof(index_record),
906 			allocator);
907 	if (destg == NULL) {
908 		index_stream_end(dest, allocator);
909 		return NULL;
910 	}
911 
912 	// Initialize destg.
913 	destg->node.uncompressed_base = 0;
914 	destg->node.compressed_base = 0;
915 	destg->number_base = 1;
916 	destg->allocated = src->record_count;
917 	destg->last = src->record_count - 1;
918 
919 	// Go through all the groups in src and copy the Records into destg.
920 	const index_group *srcg = (const index_group *)(src->groups.leftmost);
921 	size_t i = 0;
922 	do {
923 		memcpy(destg->records + i, srcg->records,
924 				(srcg->last + 1) * sizeof(index_record));
925 		i += srcg->last + 1;
926 		srcg = index_tree_next(&srcg->node);
927 	} while (srcg != NULL);
928 
929 	assert(i == destg->allocated);
930 
931 	// Add the group to the new Stream.
932 	index_tree_append(&dest->groups, &destg->node);
933 
934 	return dest;
935 }
936 
937 
938 extern LZMA_API(lzma_index *)
939 lzma_index_dup(const lzma_index *src, const lzma_allocator *allocator)
940 {
941 	// Allocate the base structure (no initial Stream).
942 	lzma_index *dest = index_init_plain(allocator);
943 	if (dest == NULL)
944 		return NULL;
945 
946 	// Copy the totals.
947 	dest->uncompressed_size = src->uncompressed_size;
948 	dest->total_size = src->total_size;
949 	dest->record_count = src->record_count;
950 	dest->index_list_size = src->index_list_size;
951 
952 	// Copy the Streams and the groups in them.
953 	const index_stream *srcstream
954 			= (const index_stream *)(src->streams.leftmost);
955 	do {
956 		index_stream *deststream = index_dup_stream(
957 				srcstream, allocator);
958 		if (deststream == NULL) {
959 			lzma_index_end(dest, allocator);
960 			return NULL;
961 		}
962 
963 		index_tree_append(&dest->streams, &deststream->node);
964 
965 		srcstream = index_tree_next(&srcstream->node);
966 	} while (srcstream != NULL);
967 
968 	return dest;
969 }
970 
971 
972 /// Indexing for lzma_index_iter.internal[]
973 enum {
974 	ITER_INDEX,
975 	ITER_STREAM,
976 	ITER_GROUP,
977 	ITER_RECORD,
978 	ITER_METHOD,
979 };
980 
981 
982 /// Values for lzma_index_iter.internal[ITER_METHOD].s
983 enum {
984 	ITER_METHOD_NORMAL,
985 	ITER_METHOD_NEXT,
986 	ITER_METHOD_LEFTMOST,
987 };
988 
989 
990 static void
991 iter_set_info(lzma_index_iter *iter)
992 {
993 	const lzma_index *i = iter->internal[ITER_INDEX].p;
994 	const index_stream *stream = iter->internal[ITER_STREAM].p;
995 	const index_group *group = iter->internal[ITER_GROUP].p;
996 	const size_t record = iter->internal[ITER_RECORD].s;
997 
998 	// lzma_index_iter.internal must not contain a pointer to the last
999 	// group in the index, because that may be reallocated by
1000 	// lzma_index_cat().
1001 	if (group == NULL) {
1002 		// There are no groups.
1003 		assert(stream->groups.root == NULL);
1004 		iter->internal[ITER_METHOD].s = ITER_METHOD_LEFTMOST;
1005 
1006 	} else if (i->streams.rightmost != &stream->node
1007 			|| stream->groups.rightmost != &group->node) {
1008 		// The group is not not the last group in the index.
1009 		iter->internal[ITER_METHOD].s = ITER_METHOD_NORMAL;
1010 
1011 	} else if (stream->groups.leftmost != &group->node) {
1012 		// The group isn't the only group in the Stream, thus we
1013 		// know that it must have a parent group i.e. it's not
1014 		// the root node.
1015 		assert(stream->groups.root != &group->node);
1016 		assert(group->node.parent->right == &group->node);
1017 		iter->internal[ITER_METHOD].s = ITER_METHOD_NEXT;
1018 		iter->internal[ITER_GROUP].p = group->node.parent;
1019 
1020 	} else {
1021 		// The Stream has only one group.
1022 		assert(stream->groups.root == &group->node);
1023 		assert(group->node.parent == NULL);
1024 		iter->internal[ITER_METHOD].s = ITER_METHOD_LEFTMOST;
1025 		iter->internal[ITER_GROUP].p = NULL;
1026 	}
1027 
1028 	// NOTE: lzma_index_iter.stream.number is lzma_vli but we use uint32_t
1029 	// internally.
1030 	iter->stream.number = stream->number;
1031 	iter->stream.block_count = stream->record_count;
1032 	iter->stream.compressed_offset = stream->node.compressed_base;
1033 	iter->stream.uncompressed_offset = stream->node.uncompressed_base;
1034 
1035 	// iter->stream.flags will be NULL if the Stream Flags haven't been
1036 	// set with lzma_index_stream_flags().
1037 	iter->stream.flags = stream->stream_flags.version == UINT32_MAX
1038 			? NULL : &stream->stream_flags;
1039 	iter->stream.padding = stream->stream_padding;
1040 
1041 	if (stream->groups.rightmost == NULL) {
1042 		// Stream has no Blocks.
1043 		iter->stream.compressed_size = index_size(0, 0)
1044 				+ 2 * LZMA_STREAM_HEADER_SIZE;
1045 		iter->stream.uncompressed_size = 0;
1046 	} else {
1047 		const index_group *g = (const index_group *)(
1048 				stream->groups.rightmost);
1049 
1050 		// Stream Header + Stream Footer + Index + Blocks
1051 		iter->stream.compressed_size = 2 * LZMA_STREAM_HEADER_SIZE
1052 				+ index_size(stream->record_count,
1053 					stream->index_list_size)
1054 				+ vli_ceil4(g->records[g->last].unpadded_sum);
1055 		iter->stream.uncompressed_size
1056 				= g->records[g->last].uncompressed_sum;
1057 	}
1058 
1059 	if (group != NULL) {
1060 		iter->block.number_in_stream = group->number_base + record;
1061 		iter->block.number_in_file = iter->block.number_in_stream
1062 				+ stream->block_number_base;
1063 
1064 		iter->block.compressed_stream_offset
1065 				= record == 0 ? group->node.compressed_base
1066 				: vli_ceil4(group->records[
1067 					record - 1].unpadded_sum);
1068 		iter->block.uncompressed_stream_offset
1069 				= record == 0 ? group->node.uncompressed_base
1070 				: group->records[record - 1].uncompressed_sum;
1071 
1072 		iter->block.uncompressed_size
1073 				= group->records[record].uncompressed_sum
1074 				- iter->block.uncompressed_stream_offset;
1075 		iter->block.unpadded_size
1076 				= group->records[record].unpadded_sum
1077 				- iter->block.compressed_stream_offset;
1078 		iter->block.total_size = vli_ceil4(iter->block.unpadded_size);
1079 
1080 		iter->block.compressed_stream_offset
1081 				+= LZMA_STREAM_HEADER_SIZE;
1082 
1083 		iter->block.compressed_file_offset
1084 				= iter->block.compressed_stream_offset
1085 				+ iter->stream.compressed_offset;
1086 		iter->block.uncompressed_file_offset
1087 				= iter->block.uncompressed_stream_offset
1088 				+ iter->stream.uncompressed_offset;
1089 	}
1090 
1091 	return;
1092 }
1093 
1094 
1095 extern LZMA_API(void)
1096 lzma_index_iter_init(lzma_index_iter *iter, const lzma_index *i)
1097 {
1098 	iter->internal[ITER_INDEX].p = i;
1099 	lzma_index_iter_rewind(iter);
1100 	return;
1101 }
1102 
1103 
1104 extern LZMA_API(void)
1105 lzma_index_iter_rewind(lzma_index_iter *iter)
1106 {
1107 	iter->internal[ITER_STREAM].p = NULL;
1108 	iter->internal[ITER_GROUP].p = NULL;
1109 	iter->internal[ITER_RECORD].s = 0;
1110 	iter->internal[ITER_METHOD].s = ITER_METHOD_NORMAL;
1111 	return;
1112 }
1113 
1114 
1115 extern LZMA_API(lzma_bool)
1116 lzma_index_iter_next(lzma_index_iter *iter, lzma_index_iter_mode mode)
1117 {
1118 	// Catch unsupported mode values.
1119 	if ((unsigned int)(mode) > LZMA_INDEX_ITER_NONEMPTY_BLOCK)
1120 		return true;
1121 
1122 	const lzma_index *i = iter->internal[ITER_INDEX].p;
1123 	const index_stream *stream = iter->internal[ITER_STREAM].p;
1124 	const index_group *group = NULL;
1125 	size_t record = iter->internal[ITER_RECORD].s;
1126 
1127 	// If we are being asked for the next Stream, leave group to NULL
1128 	// so that the rest of the this function thinks that this Stream
1129 	// has no groups and will thus go to the next Stream.
1130 	if (mode != LZMA_INDEX_ITER_STREAM) {
1131 		// Get the pointer to the current group. See iter_set_inf()
1132 		// for explanation.
1133 		switch (iter->internal[ITER_METHOD].s) {
1134 		case ITER_METHOD_NORMAL:
1135 			group = iter->internal[ITER_GROUP].p;
1136 			break;
1137 
1138 		case ITER_METHOD_NEXT:
1139 			group = index_tree_next(iter->internal[ITER_GROUP].p);
1140 			break;
1141 
1142 		case ITER_METHOD_LEFTMOST:
1143 			group = (const index_group *)(
1144 					stream->groups.leftmost);
1145 			break;
1146 		}
1147 	}
1148 
1149 again:
1150 	if (stream == NULL) {
1151 		// We at the beginning of the lzma_index.
1152 		// Locate the first Stream.
1153 		stream = (const index_stream *)(i->streams.leftmost);
1154 		if (mode >= LZMA_INDEX_ITER_BLOCK) {
1155 			// Since we are being asked to return information
1156 			// about the first a Block, skip Streams that have
1157 			// no Blocks.
1158 			while (stream->groups.leftmost == NULL) {
1159 				stream = index_tree_next(&stream->node);
1160 				if (stream == NULL)
1161 					return true;
1162 			}
1163 		}
1164 
1165 		// Start from the first Record in the Stream.
1166 		group = (const index_group *)(stream->groups.leftmost);
1167 		record = 0;
1168 
1169 	} else if (group != NULL && record < group->last) {
1170 		// The next Record is in the same group.
1171 		++record;
1172 
1173 	} else {
1174 		// This group has no more Records or this Stream has
1175 		// no Blocks at all.
1176 		record = 0;
1177 
1178 		// If group is not NULL, this Stream has at least one Block
1179 		// and thus at least one group. Find the next group.
1180 		if (group != NULL)
1181 			group = index_tree_next(&group->node);
1182 
1183 		if (group == NULL) {
1184 			// This Stream has no more Records. Find the next
1185 			// Stream. If we are being asked to return information
1186 			// about a Block, we skip empty Streams.
1187 			do {
1188 				stream = index_tree_next(&stream->node);
1189 				if (stream == NULL)
1190 					return true;
1191 			} while (mode >= LZMA_INDEX_ITER_BLOCK
1192 					&& stream->groups.leftmost == NULL);
1193 
1194 			group = (const index_group *)(
1195 					stream->groups.leftmost);
1196 		}
1197 	}
1198 
1199 	if (mode == LZMA_INDEX_ITER_NONEMPTY_BLOCK) {
1200 		// We need to look for the next Block again if this Block
1201 		// is empty.
1202 		if (record == 0) {
1203 			if (group->node.uncompressed_base
1204 					== group->records[0].uncompressed_sum)
1205 				goto again;
1206 		} else if (group->records[record - 1].uncompressed_sum
1207 				== group->records[record].uncompressed_sum) {
1208 			goto again;
1209 		}
1210 	}
1211 
1212 	iter->internal[ITER_STREAM].p = stream;
1213 	iter->internal[ITER_GROUP].p = group;
1214 	iter->internal[ITER_RECORD].s = record;
1215 
1216 	iter_set_info(iter);
1217 
1218 	return false;
1219 }
1220 
1221 
1222 extern LZMA_API(lzma_bool)
1223 lzma_index_iter_locate(lzma_index_iter *iter, lzma_vli target)
1224 {
1225 	const lzma_index *i = iter->internal[ITER_INDEX].p;
1226 
1227 	// If the target is past the end of the file, return immediately.
1228 	if (i->uncompressed_size <= target)
1229 		return true;
1230 
1231 	// Locate the Stream containing the target offset.
1232 	const index_stream *stream = index_tree_locate(&i->streams, target);
1233 	assert(stream != NULL);
1234 	target -= stream->node.uncompressed_base;
1235 
1236 	// Locate the group containing the target offset.
1237 	const index_group *group = index_tree_locate(&stream->groups, target);
1238 	assert(group != NULL);
1239 
1240 	// Use binary search to locate the exact Record. It is the first
1241 	// Record whose uncompressed_sum is greater than target.
1242 	// This is because we want the rightmost Record that fulfills the
1243 	// search criterion. It is possible that there are empty Blocks;
1244 	// we don't want to return them.
1245 	size_t left = 0;
1246 	size_t right = group->last;
1247 
1248 	while (left < right) {
1249 		const size_t pos = left + (right - left) / 2;
1250 		if (group->records[pos].uncompressed_sum <= target)
1251 			left = pos + 1;
1252 		else
1253 			right = pos;
1254 	}
1255 
1256 	iter->internal[ITER_STREAM].p = stream;
1257 	iter->internal[ITER_GROUP].p = group;
1258 	iter->internal[ITER_RECORD].s = left;
1259 
1260 	iter_set_info(iter);
1261 
1262 	return false;
1263 }
1264