xref: /linux/lib/decompress_unxz.c (revision c358f53871605a1a8d7ed6e544a05ea00e9c80cb)
1 /*
2  * Wrapper for decompressing XZ-compressed kernel, initramfs, and initrd
3  *
4  * Author: Lasse Collin <lasse.collin@tukaani.org>
5  *
6  * This file has been put into the public domain.
7  * You can do whatever you want with this file.
8  */
9 
10 /*
11  * Important notes about in-place decompression
12  *
13  * At least on x86, the kernel is decompressed in place: the compressed data
14  * is placed to the end of the output buffer, and the decompressor overwrites
15  * most of the compressed data. There must be enough safety margin to
16  * guarantee that the write position is always behind the read position.
17  *
18  * The safety margin for XZ with LZMA2 or BCJ+LZMA2 is calculated below.
19  * Note that the margin with XZ is bigger than with Deflate (gzip)!
20  *
21  * The worst case for in-place decompression is that the beginning of
22  * the file is compressed extremely well, and the rest of the file is
23  * incompressible. Thus, we must look for worst-case expansion when the
24  * compressor is encoding incompressible data.
25  *
26  * The structure of the .xz file in case of a compressed kernel is as follows.
27  * Sizes (as bytes) of the fields are in parenthesis.
28  *
29  *    Stream Header (12)
30  *    Block Header:
31  *      Block Header (8-12)
32  *      Compressed Data (N)
33  *      Block Padding (0-3)
34  *      CRC32 (4)
35  *    Index (8-20)
36  *    Stream Footer (12)
37  *
38  * Normally there is exactly one Block, but let's assume that there are
39  * 2-4 Blocks just in case. Because Stream Header and also Block Header
40  * of the first Block don't make the decompressor produce any uncompressed
41  * data, we can ignore them from our calculations. Block Headers of possible
42  * additional Blocks have to be taken into account still. With these
43  * assumptions, it is safe to assume that the total header overhead is
44  * less than 128 bytes.
45  *
46  * Compressed Data contains LZMA2 or BCJ+LZMA2 encoded data. Since BCJ
47  * doesn't change the size of the data, it is enough to calculate the
48  * safety margin for LZMA2.
49  *
50  * LZMA2 stores the data in chunks. Each chunk has a header whose size is
51  * a maximum of 6 bytes, but to get round 2^n numbers, let's assume that
52  * the maximum chunk header size is 8 bytes. After the chunk header, there
53  * may be up to 64 KiB of actual payload in the chunk. Often the payload is
54  * quite a bit smaller though; to be safe, let's assume that an average
55  * chunk has only 32 KiB of payload.
56  *
57  * The maximum uncompressed size of the payload is 2 MiB. The minimum
58  * uncompressed size of the payload is in practice never less than the
59  * payload size itself. The LZMA2 format would allow uncompressed size
60  * to be less than the payload size, but no sane compressor creates such
61  * files. LZMA2 supports storing incompressible data in uncompressed form,
62  * so there's never a need to create payloads whose uncompressed size is
63  * smaller than the compressed size.
64  *
65  * The assumption, that the uncompressed size of the payload is never
66  * smaller than the payload itself, is valid only when talking about
67  * the payload as a whole. It is possible that the payload has parts where
68  * the decompressor consumes more input than it produces output. Calculating
69  * the worst case for this would be tricky. Instead of trying to do that,
70  * let's simply make sure that the decompressor never overwrites any bytes
71  * of the payload which it is currently reading.
72  *
73  * Now we have enough information to calculate the safety margin. We need
74  *   - 128 bytes for the .xz file format headers;
75  *   - 8 bytes per every 32 KiB of uncompressed size (one LZMA2 chunk header
76  *     per chunk, each chunk having average payload size of 32 KiB); and
77  *   - 64 KiB (biggest possible LZMA2 chunk payload size) to make sure that
78  *     the decompressor never overwrites anything from the LZMA2 chunk
79  *     payload it is currently reading.
80  *
81  * We get the following formula:
82  *
83  *    safety_margin = 128 + uncompressed_size * 8 / 32768 + 65536
84  *                  = 128 + (uncompressed_size >> 12) + 65536
85  *
86  * For comparison, according to arch/x86/boot/compressed/misc.c, the
87  * equivalent formula for Deflate is this:
88  *
89  *    safety_margin = 18 + (uncompressed_size >> 12) + 32768
90  *
91  * Thus, when updating Deflate-only in-place kernel decompressor to
92  * support XZ, the fixed overhead has to be increased from 18+32768 bytes
93  * to 128+65536 bytes.
94  */
95 
96 /*
97  * STATIC is defined to "static" if we are being built for kernel
98  * decompression (pre-boot code). <linux/decompress/mm.h> will define
99  * STATIC to empty if it wasn't already defined. Since we will need to
100  * know later if we are being used for kernel decompression, we define
101  * XZ_PREBOOT here.
102  */
103 #ifdef STATIC
104 #	define XZ_PREBOOT
105 #endif
106 #ifdef __KERNEL__
107 #	include <linux/decompress/mm.h>
108 #endif
109 #define XZ_EXTERN STATIC
110 
111 #ifndef XZ_PREBOOT
112 #	include <linux/slab.h>
113 #	include <linux/xz.h>
114 #else
115 /*
116  * Use the internal CRC32 code instead of kernel's CRC32 module, which
117  * is not available in early phase of booting.
118  */
119 #define XZ_INTERNAL_CRC32 1
120 
121 /*
122  * For boot time use, we enable only the BCJ filter of the current
123  * architecture or none if no BCJ filter is available for the architecture.
124  */
125 #ifdef CONFIG_X86
126 #	define XZ_DEC_X86
127 #endif
128 #ifdef CONFIG_PPC
129 #	define XZ_DEC_POWERPC
130 #endif
131 #ifdef CONFIG_ARM
132 #	define XZ_DEC_ARM
133 #endif
134 #ifdef CONFIG_IA64
135 #	define XZ_DEC_IA64
136 #endif
137 #ifdef CONFIG_SPARC
138 #	define XZ_DEC_SPARC
139 #endif
140 
141 /*
142  * This will get the basic headers so that memeq() and others
143  * can be defined.
144  */
145 #include "xz/xz_private.h"
146 
147 /*
148  * Replace the normal allocation functions with the versions from
149  * <linux/decompress/mm.h>. vfree() needs to support vfree(NULL)
150  * when XZ_DYNALLOC is used, but the pre-boot free() doesn't support it.
151  * Workaround it here because the other decompressors don't need it.
152  */
153 #undef kmalloc
154 #undef kfree
155 #undef vmalloc
156 #undef vfree
157 #define kmalloc(size, flags) malloc(size)
158 #define kfree(ptr) free(ptr)
159 #define vmalloc(size) malloc(size)
160 #define vfree(ptr) do { if (ptr != NULL) free(ptr); } while (0)
161 
162 /*
163  * FIXME: Not all basic memory functions are provided in architecture-specific
164  * files (yet). We define our own versions here for now, but this should be
165  * only a temporary solution.
166  *
167  * memeq and memzero are not used much and any remotely sane implementation
168  * is fast enough. memcpy/memmove speed matters in multi-call mode, but
169  * the kernel image is decompressed in single-call mode, in which only
170  * memmove speed can matter and only if there is a lot of incompressible data
171  * (LZMA2 stores incompressible chunks in uncompressed form). Thus, the
172  * functions below should just be kept small; it's probably not worth
173  * optimizing for speed.
174  */
175 
176 #ifndef memeq
177 static bool memeq(const void *a, const void *b, size_t size)
178 {
179 	const uint8_t *x = a;
180 	const uint8_t *y = b;
181 	size_t i;
182 
183 	for (i = 0; i < size; ++i)
184 		if (x[i] != y[i])
185 			return false;
186 
187 	return true;
188 }
189 #endif
190 
191 #ifndef memzero
192 static void memzero(void *buf, size_t size)
193 {
194 	uint8_t *b = buf;
195 	uint8_t *e = b + size;
196 
197 	while (b != e)
198 		*b++ = '\0';
199 }
200 #endif
201 
202 #ifndef memmove
203 /* Not static to avoid a conflict with the prototype in the Linux headers. */
204 void *memmove(void *dest, const void *src, size_t size)
205 {
206 	uint8_t *d = dest;
207 	const uint8_t *s = src;
208 	size_t i;
209 
210 	if (d < s) {
211 		for (i = 0; i < size; ++i)
212 			d[i] = s[i];
213 	} else if (d > s) {
214 		i = size;
215 		while (i-- > 0)
216 			d[i] = s[i];
217 	}
218 
219 	return dest;
220 }
221 #endif
222 
223 /*
224  * Since we need memmove anyway, would use it as memcpy too.
225  * Commented out for now to avoid breaking things.
226  */
227 /*
228 #ifndef memcpy
229 #	define memcpy memmove
230 #endif
231 */
232 
233 #include "xz/xz_crc32.c"
234 #include "xz/xz_dec_stream.c"
235 #include "xz/xz_dec_lzma2.c"
236 #include "xz/xz_dec_bcj.c"
237 
238 #endif /* XZ_PREBOOT */
239 
240 /* Size of the input and output buffers in multi-call mode */
241 #define XZ_IOBUF_SIZE 4096
242 
243 /*
244  * This function implements the API defined in <linux/decompress/generic.h>.
245  *
246  * This wrapper will automatically choose single-call or multi-call mode
247  * of the native XZ decoder API. The single-call mode can be used only when
248  * both input and output buffers are available as a single chunk, i.e. when
249  * fill() and flush() won't be used.
250  */
251 STATIC int INIT unxz(unsigned char *in, long in_size,
252 		     long (*fill)(void *dest, unsigned long size),
253 		     long (*flush)(void *src, unsigned long size),
254 		     unsigned char *out, long *in_used,
255 		     void (*error)(char *x))
256 {
257 	struct xz_buf b;
258 	struct xz_dec *s;
259 	enum xz_ret ret;
260 	bool must_free_in = false;
261 
262 #if XZ_INTERNAL_CRC32
263 	xz_crc32_init();
264 #endif
265 
266 	if (in_used != NULL)
267 		*in_used = 0;
268 
269 	if (fill == NULL && flush == NULL)
270 		s = xz_dec_init(XZ_SINGLE, 0);
271 	else
272 		s = xz_dec_init(XZ_DYNALLOC, (uint32_t)-1);
273 
274 	if (s == NULL)
275 		goto error_alloc_state;
276 
277 	if (flush == NULL) {
278 		b.out = out;
279 		b.out_size = (size_t)-1;
280 	} else {
281 		b.out_size = XZ_IOBUF_SIZE;
282 		b.out = malloc(XZ_IOBUF_SIZE);
283 		if (b.out == NULL)
284 			goto error_alloc_out;
285 	}
286 
287 	if (in == NULL) {
288 		must_free_in = true;
289 		in = malloc(XZ_IOBUF_SIZE);
290 		if (in == NULL)
291 			goto error_alloc_in;
292 	}
293 
294 	b.in = in;
295 	b.in_pos = 0;
296 	b.in_size = in_size;
297 	b.out_pos = 0;
298 
299 	if (fill == NULL && flush == NULL) {
300 		ret = xz_dec_run(s, &b);
301 	} else {
302 		do {
303 			if (b.in_pos == b.in_size && fill != NULL) {
304 				if (in_used != NULL)
305 					*in_used += b.in_pos;
306 
307 				b.in_pos = 0;
308 
309 				in_size = fill(in, XZ_IOBUF_SIZE);
310 				if (in_size < 0) {
311 					/*
312 					 * This isn't an optimal error code
313 					 * but it probably isn't worth making
314 					 * a new one either.
315 					 */
316 					ret = XZ_BUF_ERROR;
317 					break;
318 				}
319 
320 				b.in_size = in_size;
321 			}
322 
323 			ret = xz_dec_run(s, &b);
324 
325 			if (flush != NULL && (b.out_pos == b.out_size
326 					|| (ret != XZ_OK && b.out_pos > 0))) {
327 				/*
328 				 * Setting ret here may hide an error
329 				 * returned by xz_dec_run(), but probably
330 				 * it's not too bad.
331 				 */
332 				if (flush(b.out, b.out_pos) != (long)b.out_pos)
333 					ret = XZ_BUF_ERROR;
334 
335 				b.out_pos = 0;
336 			}
337 		} while (ret == XZ_OK);
338 
339 		if (must_free_in)
340 			free(in);
341 
342 		if (flush != NULL)
343 			free(b.out);
344 	}
345 
346 	if (in_used != NULL)
347 		*in_used += b.in_pos;
348 
349 	xz_dec_end(s);
350 
351 	switch (ret) {
352 	case XZ_STREAM_END:
353 		return 0;
354 
355 	case XZ_MEM_ERROR:
356 		/* This can occur only in multi-call mode. */
357 		error("XZ decompressor ran out of memory");
358 		break;
359 
360 	case XZ_FORMAT_ERROR:
361 		error("Input is not in the XZ format (wrong magic bytes)");
362 		break;
363 
364 	case XZ_OPTIONS_ERROR:
365 		error("Input was encoded with settings that are not "
366 				"supported by this XZ decoder");
367 		break;
368 
369 	case XZ_DATA_ERROR:
370 	case XZ_BUF_ERROR:
371 		error("XZ-compressed data is corrupt");
372 		break;
373 
374 	default:
375 		error("Bug in the XZ decompressor");
376 		break;
377 	}
378 
379 	return -1;
380 
381 error_alloc_in:
382 	if (flush != NULL)
383 		free(b.out);
384 
385 error_alloc_out:
386 	xz_dec_end(s);
387 
388 error_alloc_state:
389 	error("XZ decompressor ran out of memory");
390 	return -1;
391 }
392 
393 /*
394  * This macro is used by architecture-specific files to decompress
395  * the kernel image.
396  */
397 #ifdef XZ_PREBOOT
398 STATIC int INIT __decompress(unsigned char *buf, long len,
399 			   long (*fill)(void*, unsigned long),
400 			   long (*flush)(void*, unsigned long),
401 			   unsigned char *out_buf, long olen,
402 			   long *pos,
403 			   void (*error)(char *x))
404 {
405 	return unxz(buf, len, fill, flush, out_buf, pos, error);
406 }
407 #endif
408