xref: /netbsd-src/usr.bin/gzip/unxz.c (revision f57e938466b663d8f44db7e94d48225681221b88)
1 /*	$NetBSD: unxz.c,v 1.9 2024/05/04 13:17:03 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 2011 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Christos Zoulas.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 #include <sys/cdefs.h>
32 __RCSID("$NetBSD: unxz.c,v 1.9 2024/05/04 13:17:03 christos Exp $");
33 
34 #include <stdarg.h>
35 #include <errno.h>
36 #include <stdio.h>
37 #include <unistd.h>
38 #include <lzma.h>
39 
40 static off_t
unxz(int i,int o,char * pre,size_t prelen,off_t * bytes_in)41 unxz(int i, int o, char *pre, size_t prelen, off_t *bytes_in)
42 {
43 	lzma_stream strm = LZMA_STREAM_INIT;
44 	static const int flags = LZMA_TELL_UNSUPPORTED_CHECK|LZMA_CONCATENATED;
45 	lzma_ret ret;
46 	lzma_action action = LZMA_RUN;
47 	off_t bytes_out, bp;
48 	uint8_t ibuf[BUFSIZ];
49 	uint8_t obuf[BUFSIZ];
50 
51 	if (bytes_in == NULL)
52 		bytes_in = &bp;
53 
54 	strm.next_in = ibuf;
55 	memcpy(ibuf, pre, prelen);
56 	strm.avail_in = read(i, ibuf + prelen, sizeof(ibuf) - prelen);
57 	if (strm.avail_in == (size_t)-1)
58 		maybe_err("read failed");
59 	infile_newdata(strm.avail_in);
60 	strm.avail_in += prelen;
61 	*bytes_in = strm.avail_in;
62 
63 	if ((ret = lzma_stream_decoder(&strm, UINT64_MAX, flags)) != LZMA_OK)
64 		maybe_errx("Can't initialize decoder (%d)", ret);
65 
66 	strm.next_out = NULL;
67 	strm.avail_out = 0;
68 	if ((ret = lzma_code(&strm, LZMA_RUN)) != LZMA_OK)
69 		maybe_errx("Can't read headers (%d)", ret);
70 
71 	bytes_out = 0;
72 	strm.next_out = obuf;
73 	strm.avail_out = sizeof(obuf);
74 
75 	for (;;) {
76 		check_siginfo();
77 		if (strm.avail_in == 0) {
78 			strm.next_in = ibuf;
79 			strm.avail_in = read(i, ibuf, sizeof(ibuf));
80 			switch (strm.avail_in) {
81 			case (size_t)-1:
82 				maybe_err("read failed");
83 				/*NOTREACHED*/
84 			case 0:
85 				action = LZMA_FINISH;
86 				break;
87 			default:
88 				infile_newdata(strm.avail_in);
89 				*bytes_in += strm.avail_in;
90 				break;
91 			}
92 		}
93 
94 		ret = lzma_code(&strm, action);
95 
96 		// Write and check write error before checking decoder error.
97 		// This way as much data as possible gets written to output
98 		// even if decoder detected an error.
99 		if (strm.avail_out == 0 || ret != LZMA_OK) {
100 			const size_t write_size = sizeof(obuf) - strm.avail_out;
101 
102 			if (!tflag &&
103 			    write(o, obuf, write_size) != (ssize_t)write_size)
104 				maybe_err("write failed");
105 
106 			strm.next_out = obuf;
107 			strm.avail_out = sizeof(obuf);
108 			bytes_out += write_size;
109 		}
110 
111 		if (ret != LZMA_OK) {
112 			if (ret == LZMA_STREAM_END) {
113 				// Check that there's no trailing garbage.
114 				if (strm.avail_in != 0 || read(i, ibuf, 1))
115 					ret = LZMA_DATA_ERROR;
116 				else {
117 					lzma_end(&strm);
118 					return bytes_out;
119 				}
120 			}
121 
122 			const char *msg;
123 			switch (ret) {
124 			case LZMA_MEM_ERROR:
125 				msg = strerror(ENOMEM);
126 				break;
127 
128 			case LZMA_FORMAT_ERROR:
129 				msg = "File format not recognized";
130 				break;
131 
132 			case LZMA_OPTIONS_ERROR:
133 				// FIXME: Better message?
134 				msg = "Unsupported compression options";
135 				break;
136 
137 			case LZMA_DATA_ERROR:
138 				msg = "File is corrupt";
139 				break;
140 
141 			case LZMA_BUF_ERROR:
142 				msg = "Unexpected end of input";
143 				break;
144 
145 			case LZMA_MEMLIMIT_ERROR:
146 				msg = "Reached memory limit";
147 				break;
148 
149 			default:
150 				maybe_errx("Unknown error (%d)", ret);
151 				break;
152 			}
153 			maybe_errx("%s", msg);
154 
155 		}
156 	}
157 }
158 
159 #include <stdbool.h>
160 
161 /*
162  * Copied various bits and pieces from xz support code or brute force
163  * replacements.
164  */
165 
166 #define	my_min(A,B)	((A)<(B)?(A):(B))
167 
168 // Some systems have suboptimal BUFSIZ. Use a bit bigger value on them.
169 // We also need that IO_BUFFER_SIZE is a multiple of 8 (sizeof(uint64_t))
170 #if BUFSIZ <= 1024
171 #       define IO_BUFFER_SIZE 8192
172 #else
173 #       define IO_BUFFER_SIZE (BUFSIZ & ~7U)
174 #endif
175 
176 /// is_sparse() accesses the buffer as uint64_t for maximum speed.
177 /// Use an union to make sure that the buffer is properly aligned.
178 typedef union {
179         uint8_t u8[IO_BUFFER_SIZE];
180         uint32_t u32[IO_BUFFER_SIZE / sizeof(uint32_t)];
181         uint64_t u64[IO_BUFFER_SIZE / sizeof(uint64_t)];
182 } io_buf;
183 
184 
185 static bool
io_pread(int fd,io_buf * buf,size_t size,off_t pos)186 io_pread(int fd, io_buf *buf, size_t size, off_t pos)
187 {
188 	// Using lseek() and read() is more portable than pread() and
189 	// for us it is as good as real pread().
190 	if (lseek(fd, pos, SEEK_SET) != pos) {
191 		return true;
192 	}
193 
194 	const size_t amount = read(fd, buf, size);
195 	if (amount == SIZE_MAX)
196 		return true;
197 
198 	if (amount != size) {
199 		return true;
200 	}
201 
202 	return false;
203 }
204 
205 /*
206  * Most of the following is copied (mostly verbatim) from the xz
207  * distribution, from file src/xz/list.c
208  */
209 
210 ///////////////////////////////////////////////////////////////////////////////
211 //
212 /// \file       list.c
213 /// \brief      Listing information about .xz files
214 //
215 //  Author:     Lasse Collin
216 //
217 //  This file has been put into the public domain.
218 //  You can do whatever you want with this file.
219 //
220 ///////////////////////////////////////////////////////////////////////////////
221 
222 
223 /// Information about a .xz file
224 typedef struct {
225 	/// Combined Index of all Streams in the file
226 	lzma_index *idx;
227 
228 	/// Total amount of Stream Padding
229 	uint64_t stream_padding;
230 
231 	/// Highest memory usage so far
232 	uint64_t memusage_max;
233 
234 	/// True if all Blocks so far have Compressed Size and
235 	/// Uncompressed Size fields
236 	bool all_have_sizes;
237 
238 	/// Oldest XZ Utils version that will decompress the file
239 	uint32_t min_version;
240 
241 } xz_file_info;
242 
243 #define XZ_FILE_INFO_INIT { NULL, 0, 0, true, 50000002 }
244 
245 
246 /// \brief      Parse the Index(es) from the given .xz file
247 ///
248 /// \param      xfi     Pointer to structure where the decoded information
249 ///                     is stored.
250 /// \param      pair    Input file
251 ///
252 /// \return     On success, false is returned. On error, true is returned.
253 ///
254 // TODO: This function is pretty big. liblzma should have a function that
255 // takes a callback function to parse the Index(es) from a .xz file to make
256 // it easy for applications.
257 static bool
parse_indexes(xz_file_info * xfi,int src_fd)258 parse_indexes(xz_file_info *xfi, int src_fd)
259 {
260 	struct stat st;
261 
262 	fstat(src_fd, &st);
263 	if (st.st_size <= 0) {
264 		return true;
265 	}
266 
267 	if (st.st_size < 2 * LZMA_STREAM_HEADER_SIZE) {
268 		return true;
269 	}
270 
271 	io_buf buf;
272 	lzma_stream_flags header_flags;
273 	lzma_stream_flags footer_flags;
274 	lzma_ret ret;
275 
276 	// lzma_stream for the Index decoder
277 	lzma_stream strm = LZMA_STREAM_INIT;
278 
279 	// All Indexes decoded so far
280 	lzma_index *combined_index = NULL;
281 
282 	// The Index currently being decoded
283 	lzma_index *this_index = NULL;
284 
285 	// Current position in the file. We parse the file backwards so
286 	// initialize it to point to the end of the file.
287 	off_t pos = st.st_size;
288 
289 	// Each loop iteration decodes one Index.
290 	do {
291 		// Check that there is enough data left to contain at least
292 		// the Stream Header and Stream Footer. This check cannot
293 		// fail in the first pass of this loop.
294 		if (pos < 2 * LZMA_STREAM_HEADER_SIZE) {
295 			goto error;
296 		}
297 
298 		pos -= LZMA_STREAM_HEADER_SIZE;
299 		lzma_vli stream_padding = 0;
300 
301 		// Locate the Stream Footer. There may be Stream Padding which
302 		// we must skip when reading backwards.
303 		while (true) {
304 			if (pos < LZMA_STREAM_HEADER_SIZE) {
305 				goto error;
306 			}
307 
308 			if (io_pread(src_fd, &buf,
309 					LZMA_STREAM_HEADER_SIZE, pos))
310 				goto error;
311 
312 			// Stream Padding is always a multiple of four bytes.
313 			int i = 2;
314 			if (buf.u32[i] != 0)
315 				break;
316 
317 			// To avoid calling io_pread() for every four bytes
318 			// of Stream Padding, take advantage that we read
319 			// 12 bytes (LZMA_STREAM_HEADER_SIZE) already and
320 			// check them too before calling io_pread() again.
321 			do {
322 				stream_padding += 4;
323 				pos -= 4;
324 				--i;
325 			} while (i >= 0 && buf.u32[i] == 0);
326 		}
327 
328 		// Decode the Stream Footer.
329 		ret = lzma_stream_footer_decode(&footer_flags, buf.u8);
330 		if (ret != LZMA_OK) {
331 			goto error;
332 		}
333 
334 		// Check that the Stream Footer doesn't specify something
335 		// that we don't support. This can only happen if the xz
336 		// version is older than liblzma and liblzma supports
337 		// something new.
338 		//
339 		// It is enough to check Stream Footer. Stream Header must
340 		// match when it is compared against Stream Footer with
341 		// lzma_stream_flags_compare().
342 		if (footer_flags.version != 0) {
343 			goto error;
344 		}
345 
346 		// Check that the size of the Index field looks sane.
347 		lzma_vli index_size = footer_flags.backward_size;
348 		if ((lzma_vli)(pos) < index_size + LZMA_STREAM_HEADER_SIZE) {
349 			goto error;
350 		}
351 
352 		// Set pos to the beginning of the Index.
353 		pos -= index_size;
354 
355 		// Decode the Index.
356 		ret = lzma_index_decoder(&strm, &this_index, UINT64_MAX);
357 		if (ret != LZMA_OK) {
358 			goto error;
359 		}
360 
361 		do {
362 			// Don't give the decoder more input than the
363 			// Index size.
364 			strm.avail_in = my_min(IO_BUFFER_SIZE, index_size);
365 			if (io_pread(src_fd, &buf, strm.avail_in, pos))
366 				goto error;
367 
368 			pos += strm.avail_in;
369 			index_size -= strm.avail_in;
370 
371 			strm.next_in = buf.u8;
372 			ret = lzma_code(&strm, LZMA_RUN);
373 
374 		} while (ret == LZMA_OK);
375 
376 		// If the decoding seems to be successful, check also that
377 		// the Index decoder consumed as much input as indicated
378 		// by the Backward Size field.
379 		if (ret == LZMA_STREAM_END)
380 			if (index_size != 0 || strm.avail_in != 0)
381 				ret = LZMA_DATA_ERROR;
382 
383 		if (ret != LZMA_STREAM_END) {
384 			// LZMA_BUFFER_ERROR means that the Index decoder
385 			// would have liked more input than what the Index
386 			// size should be according to Stream Footer.
387 			// The message for LZMA_DATA_ERROR makes more
388 			// sense in that case.
389 			if (ret == LZMA_BUF_ERROR)
390 				ret = LZMA_DATA_ERROR;
391 
392 			goto error;
393 		}
394 
395 		// Decode the Stream Header and check that its Stream Flags
396 		// match the Stream Footer.
397 		pos -= footer_flags.backward_size + LZMA_STREAM_HEADER_SIZE;
398 		if ((lzma_vli)(pos) < lzma_index_total_size(this_index)) {
399 			goto error;
400 		}
401 
402 		pos -= lzma_index_total_size(this_index);
403 		if (io_pread(src_fd, &buf, LZMA_STREAM_HEADER_SIZE, pos))
404 			goto error;
405 
406 		ret = lzma_stream_header_decode(&header_flags, buf.u8);
407 		if (ret != LZMA_OK) {
408 			goto error;
409 		}
410 
411 		ret = lzma_stream_flags_compare(&header_flags, &footer_flags);
412 		if (ret != LZMA_OK) {
413 			goto error;
414 		}
415 
416 		// Store the decoded Stream Flags into this_index. This is
417 		// needed so that we can print which Check is used in each
418 		// Stream.
419 		ret = lzma_index_stream_flags(this_index, &footer_flags);
420 		if (ret != LZMA_OK)
421 			goto error;
422 
423 		// Store also the size of the Stream Padding field. It is
424 		// needed to show the offsets of the Streams correctly.
425 		ret = lzma_index_stream_padding(this_index, stream_padding);
426 		if (ret != LZMA_OK)
427 			goto error;
428 
429 		if (combined_index != NULL) {
430 			// Append the earlier decoded Indexes
431 			// after this_index.
432 			ret = lzma_index_cat(
433 					this_index, combined_index, NULL);
434 			if (ret != LZMA_OK) {
435 				goto error;
436 			}
437 		}
438 
439 		combined_index = this_index;
440 		this_index = NULL;
441 
442 		xfi->stream_padding += stream_padding;
443 
444 	} while (pos > 0);
445 
446 	lzma_end(&strm);
447 
448 	// All OK. Make combined_index available to the caller.
449 	xfi->idx = combined_index;
450 	return false;
451 
452 error:
453 	// Something went wrong, free the allocated memory.
454 	lzma_end(&strm);
455 	lzma_index_end(combined_index, NULL);
456 	lzma_index_end(this_index, NULL);
457 	return true;
458 }
459 
460 /***************** end of copy form list.c *************************/
461 
462 /*
463  * Small wrapper to extract total length of a file
464  */
465 off_t
unxz_len(int fd)466 unxz_len(int fd)
467 {
468 	xz_file_info xfi = XZ_FILE_INFO_INIT;
469 	if (!parse_indexes(&xfi, fd)) {
470 		off_t res = lzma_index_uncompressed_size(xfi.idx);
471 		lzma_index_end(xfi.idx, NULL);
472 		return res;
473 	}
474 	return 0;
475 }
476 
477