xref: /freebsd/contrib/xz/src/liblzma/common/index_hash.c (revision e9ac41698b2f322d55ccf9da50a3596edb2c1800)
1 // SPDX-License-Identifier: 0BSD
2 
3 ///////////////////////////////////////////////////////////////////////////////
4 //
5 /// \file       index_hash.c
6 /// \brief      Validates Index by using a hash function
7 //
8 //  Author:     Lasse Collin
9 //
10 ///////////////////////////////////////////////////////////////////////////////
11 
12 #include "common.h"
13 #include "index.h"
14 #include "check.h"
15 
16 
17 typedef struct {
18 	/// Sum of the Block sizes (including Block Padding)
19 	lzma_vli blocks_size;
20 
21 	/// Sum of the Uncompressed Size fields
22 	lzma_vli uncompressed_size;
23 
24 	/// Number of Records
25 	lzma_vli count;
26 
27 	/// Size of the List of Index Records as bytes
28 	lzma_vli index_list_size;
29 
30 	/// Check calculated from Unpadded Sizes and Uncompressed Sizes.
31 	lzma_check_state check;
32 
33 } lzma_index_hash_info;
34 
35 
36 struct lzma_index_hash_s {
37 	enum {
38 		SEQ_BLOCK,
39 		SEQ_COUNT,
40 		SEQ_UNPADDED,
41 		SEQ_UNCOMPRESSED,
42 		SEQ_PADDING_INIT,
43 		SEQ_PADDING,
44 		SEQ_CRC32,
45 	} sequence;
46 
47 	/// Information collected while decoding the actual Blocks.
48 	lzma_index_hash_info blocks;
49 
50 	/// Information collected from the Index field.
51 	lzma_index_hash_info records;
52 
53 	/// Number of Records not fully decoded
54 	lzma_vli remaining;
55 
56 	/// Unpadded Size currently being read from an Index Record.
57 	lzma_vli unpadded_size;
58 
59 	/// Uncompressed Size currently being read from an Index Record.
60 	lzma_vli uncompressed_size;
61 
62 	/// Position in variable-length integers when decoding them from
63 	/// the List of Records.
64 	size_t pos;
65 
66 	/// CRC32 of the Index
67 	uint32_t crc32;
68 };
69 
70 
71 extern LZMA_API(lzma_index_hash *)
72 lzma_index_hash_init(lzma_index_hash *index_hash,
73 		const lzma_allocator *allocator)
74 {
75 	if (index_hash == NULL) {
76 		index_hash = lzma_alloc(sizeof(lzma_index_hash), allocator);
77 		if (index_hash == NULL)
78 			return NULL;
79 	}
80 
81 	index_hash->sequence = SEQ_BLOCK;
82 	index_hash->blocks.blocks_size = 0;
83 	index_hash->blocks.uncompressed_size = 0;
84 	index_hash->blocks.count = 0;
85 	index_hash->blocks.index_list_size = 0;
86 	index_hash->records.blocks_size = 0;
87 	index_hash->records.uncompressed_size = 0;
88 	index_hash->records.count = 0;
89 	index_hash->records.index_list_size = 0;
90 	index_hash->unpadded_size = 0;
91 	index_hash->uncompressed_size = 0;
92 	index_hash->pos = 0;
93 	index_hash->crc32 = 0;
94 
95 	// These cannot fail because LZMA_CHECK_BEST is known to be supported.
96 	(void)lzma_check_init(&index_hash->blocks.check, LZMA_CHECK_BEST);
97 	(void)lzma_check_init(&index_hash->records.check, LZMA_CHECK_BEST);
98 
99 	return index_hash;
100 }
101 
102 
103 extern LZMA_API(void)
104 lzma_index_hash_end(lzma_index_hash *index_hash,
105 		const lzma_allocator *allocator)
106 {
107 	lzma_free(index_hash, allocator);
108 	return;
109 }
110 
111 
112 extern LZMA_API(lzma_vli)
113 lzma_index_hash_size(const lzma_index_hash *index_hash)
114 {
115 	// Get the size of the Index from ->blocks instead of ->records for
116 	// cases where application wants to know the Index Size before
117 	// decoding the Index.
118 	return index_size(index_hash->blocks.count,
119 			index_hash->blocks.index_list_size);
120 }
121 
122 
123 /// Updates the sizes and the hash without any validation.
124 static void
125 hash_append(lzma_index_hash_info *info, lzma_vli unpadded_size,
126 		lzma_vli uncompressed_size)
127 {
128 	info->blocks_size += vli_ceil4(unpadded_size);
129 	info->uncompressed_size += uncompressed_size;
130 	info->index_list_size += lzma_vli_size(unpadded_size)
131 			+ lzma_vli_size(uncompressed_size);
132 	++info->count;
133 
134 	const lzma_vli sizes[2] = { unpadded_size, uncompressed_size };
135 	lzma_check_update(&info->check, LZMA_CHECK_BEST,
136 			(const uint8_t *)(sizes), sizeof(sizes));
137 
138 	return;
139 }
140 
141 
142 extern LZMA_API(lzma_ret)
143 lzma_index_hash_append(lzma_index_hash *index_hash, lzma_vli unpadded_size,
144 		lzma_vli uncompressed_size)
145 {
146 	// Validate the arguments.
147 	if (index_hash == NULL || index_hash->sequence != SEQ_BLOCK
148 			|| unpadded_size < UNPADDED_SIZE_MIN
149 			|| unpadded_size > UNPADDED_SIZE_MAX
150 			|| uncompressed_size > LZMA_VLI_MAX)
151 		return LZMA_PROG_ERROR;
152 
153 	// Update the hash.
154 	hash_append(&index_hash->blocks, unpadded_size, uncompressed_size);
155 
156 	// Validate the properties of *info are still in allowed limits.
157 	if (index_hash->blocks.blocks_size > LZMA_VLI_MAX
158 			|| index_hash->blocks.uncompressed_size > LZMA_VLI_MAX
159 			|| index_size(index_hash->blocks.count,
160 					index_hash->blocks.index_list_size)
161 				> LZMA_BACKWARD_SIZE_MAX
162 			|| index_stream_size(index_hash->blocks.blocks_size,
163 					index_hash->blocks.count,
164 					index_hash->blocks.index_list_size)
165 				> LZMA_VLI_MAX)
166 		return LZMA_DATA_ERROR;
167 
168 	return LZMA_OK;
169 }
170 
171 
172 extern LZMA_API(lzma_ret)
173 lzma_index_hash_decode(lzma_index_hash *index_hash, const uint8_t *in,
174 		size_t *in_pos, size_t in_size)
175 {
176 	// Catch zero input buffer here, because in contrast to Index encoder
177 	// and decoder functions, applications call this function directly
178 	// instead of via lzma_code(), which does the buffer checking.
179 	if (*in_pos >= in_size)
180 		return LZMA_BUF_ERROR;
181 
182 	// NOTE: This function has many similarities to index_encode() and
183 	// index_decode() functions found from index_encoder.c and
184 	// index_decoder.c. See the comments especially in index_encoder.c.
185 	const size_t in_start = *in_pos;
186 	lzma_ret ret = LZMA_OK;
187 
188 	while (*in_pos < in_size)
189 	switch (index_hash->sequence) {
190 	case SEQ_BLOCK:
191 		// Check the Index Indicator is present.
192 		if (in[(*in_pos)++] != INDEX_INDICATOR)
193 			return LZMA_DATA_ERROR;
194 
195 		index_hash->sequence = SEQ_COUNT;
196 		break;
197 
198 	case SEQ_COUNT: {
199 		ret = lzma_vli_decode(&index_hash->remaining,
200 				&index_hash->pos, in, in_pos, in_size);
201 		if (ret != LZMA_STREAM_END)
202 			goto out;
203 
204 		// The count must match the count of the Blocks decoded.
205 		if (index_hash->remaining != index_hash->blocks.count)
206 			return LZMA_DATA_ERROR;
207 
208 		ret = LZMA_OK;
209 		index_hash->pos = 0;
210 
211 		// Handle the special case when there are no Blocks.
212 		index_hash->sequence = index_hash->remaining == 0
213 				? SEQ_PADDING_INIT : SEQ_UNPADDED;
214 		break;
215 	}
216 
217 	case SEQ_UNPADDED:
218 	case SEQ_UNCOMPRESSED: {
219 		lzma_vli *size = index_hash->sequence == SEQ_UNPADDED
220 				? &index_hash->unpadded_size
221 				: &index_hash->uncompressed_size;
222 
223 		ret = lzma_vli_decode(size, &index_hash->pos,
224 				in, in_pos, in_size);
225 		if (ret != LZMA_STREAM_END)
226 			goto out;
227 
228 		ret = LZMA_OK;
229 		index_hash->pos = 0;
230 
231 		if (index_hash->sequence == SEQ_UNPADDED) {
232 			if (index_hash->unpadded_size < UNPADDED_SIZE_MIN
233 					|| index_hash->unpadded_size
234 						> UNPADDED_SIZE_MAX)
235 				return LZMA_DATA_ERROR;
236 
237 			index_hash->sequence = SEQ_UNCOMPRESSED;
238 		} else {
239 			// Update the hash.
240 			hash_append(&index_hash->records,
241 					index_hash->unpadded_size,
242 					index_hash->uncompressed_size);
243 
244 			// Verify that we don't go over the known sizes. Note
245 			// that this validation is simpler than the one used
246 			// in lzma_index_hash_append(), because here we know
247 			// that values in index_hash->blocks are already
248 			// validated and we are fine as long as we don't
249 			// exceed them in index_hash->records.
250 			if (index_hash->blocks.blocks_size
251 					< index_hash->records.blocks_size
252 					|| index_hash->blocks.uncompressed_size
253 					< index_hash->records.uncompressed_size
254 					|| index_hash->blocks.index_list_size
255 					< index_hash->records.index_list_size)
256 				return LZMA_DATA_ERROR;
257 
258 			// Check if this was the last Record.
259 			index_hash->sequence = --index_hash->remaining == 0
260 					? SEQ_PADDING_INIT : SEQ_UNPADDED;
261 		}
262 
263 		break;
264 	}
265 
266 	case SEQ_PADDING_INIT:
267 		index_hash->pos = (LZMA_VLI_C(4) - index_size_unpadded(
268 				index_hash->records.count,
269 				index_hash->records.index_list_size)) & 3;
270 		index_hash->sequence = SEQ_PADDING;
271 
272 	// Fall through
273 
274 	case SEQ_PADDING:
275 		if (index_hash->pos > 0) {
276 			--index_hash->pos;
277 			if (in[(*in_pos)++] != 0x00)
278 				return LZMA_DATA_ERROR;
279 
280 			break;
281 		}
282 
283 		// Compare the sizes.
284 		if (index_hash->blocks.blocks_size
285 				!= index_hash->records.blocks_size
286 				|| index_hash->blocks.uncompressed_size
287 				!= index_hash->records.uncompressed_size
288 				|| index_hash->blocks.index_list_size
289 				!= index_hash->records.index_list_size)
290 			return LZMA_DATA_ERROR;
291 
292 		// Finish the hashes and compare them.
293 		lzma_check_finish(&index_hash->blocks.check, LZMA_CHECK_BEST);
294 		lzma_check_finish(&index_hash->records.check, LZMA_CHECK_BEST);
295 		if (memcmp(index_hash->blocks.check.buffer.u8,
296 				index_hash->records.check.buffer.u8,
297 				lzma_check_size(LZMA_CHECK_BEST)) != 0)
298 			return LZMA_DATA_ERROR;
299 
300 		// Finish the CRC32 calculation.
301 		index_hash->crc32 = lzma_crc32(in + in_start,
302 				*in_pos - in_start, index_hash->crc32);
303 
304 		index_hash->sequence = SEQ_CRC32;
305 
306 	// Fall through
307 
308 	case SEQ_CRC32:
309 		do {
310 			if (*in_pos == in_size)
311 				return LZMA_OK;
312 
313 			if (((index_hash->crc32 >> (index_hash->pos * 8))
314 					& 0xFF) != in[(*in_pos)++]) {
315 #ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
316 				return LZMA_DATA_ERROR;
317 #endif
318 			}
319 
320 		} while (++index_hash->pos < 4);
321 
322 		return LZMA_STREAM_END;
323 
324 	default:
325 		assert(0);
326 		return LZMA_PROG_ERROR;
327 	}
328 
329 out:
330 	// Update the CRC32.
331 	//
332 	// Avoid null pointer + 0 (undefined behavior) in "in + in_start".
333 	// In such a case we had no input and thus in_used == 0.
334 	{
335 		const size_t in_used = *in_pos - in_start;
336 		if (in_used > 0)
337 			index_hash->crc32 = lzma_crc32(in + in_start,
338 					in_used, index_hash->crc32);
339 	}
340 
341 	return ret;
342 }
343