Import libarchive and bsdtar. The default tar (/usr/bin/tar) can be choosen
[dragonfly.git] / contrib / libarchive / archive_read_support_compression_gzip.c
1 /*-
2  * Copyright (c) 2003-2004 Tim Kientzle
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
16  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18  * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
19  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25  */
26
27 #include "archive_platform.h"
28
29 __FBSDID("$FreeBSD: src/lib/libarchive/archive_read_support_compression_gzip.c,v 1.7 2004/08/14 03:45:45 kientzle Exp $");
30
31
32 #include <errno.h>
33 #include <stdlib.h>
34 #include <string.h>
35 #include <unistd.h>
36 #ifdef HAVE_ZLIB_H
37 #include <zlib.h>
38 #endif
39
40 #include "archive.h"
41 #include "archive_private.h"
42
43 #ifdef HAVE_ZLIB_H
44 struct private_data {
45         z_stream         stream;
46         unsigned char   *uncompressed_buffer;
47         size_t           uncompressed_buffer_size;
48         unsigned char   *read_next;
49         int64_t          total_out;
50         unsigned long    crc;
51         char             header_done;
52 };
53
54 static int      finish(struct archive *);
55 static ssize_t  read_ahead(struct archive *, const void **, size_t);
56 static ssize_t  read_consume(struct archive *, size_t);
57 static int      drive_decompressor(struct archive *a, struct private_data *);
58 #endif
59
60 /* These two functions are defined even if we lack zlib.  See below. */
61 static int      bid(const void *, size_t);
62 static int      init(struct archive *, const void *, size_t);
63
64 int
65 archive_read_support_compression_gzip(struct archive *a)
66 {
67         return (__archive_read_register_compression(a, bid, init));
68 }
69
70 /*
71  * Test whether we can handle this data.
72  *
73  * This logic returns zero if any part of the signature fails.  It
74  * also tries to Do The Right Thing if a very short buffer prevents us
75  * from verifying as much as we would like.
76  */
77 static int
78 bid(const void *buff, size_t len)
79 {
80         const unsigned char *buffer;
81         int bits_checked;
82
83         if (len < 1)
84                 return (0);
85
86         buffer = buff;
87         bits_checked = 0;
88         if (buffer[0] != 037)   /* Verify first ID byte. */
89                 return (0);
90         bits_checked += 8;
91         if (len < 2)
92                 return (bits_checked);
93
94         if (buffer[1] != 0213)  /* Verify second ID byte. */
95                 return (0);
96         bits_checked += 8;
97         if (len < 3)
98                 return (bits_checked);
99
100         if (buffer[2] != 8)     /* Compression must be 'deflate'. */
101                 return (0);
102         bits_checked += 8;
103         if (len < 4)
104                 return (bits_checked);
105
106         if ((buffer[3] & 0xE0)!= 0)     /* No reserved flags set. */
107                 return (0);
108         bits_checked += 3;
109         if (len < 5)
110                 return (bits_checked);
111
112         /*
113          * TODO: Verify more; in particular, gzip has an optional
114          * header CRC, which would give us 16 more verified bits.  We
115          * may also be able to verify certain constraints on other
116          * fields.
117          */
118
119         return (bits_checked);
120 }
121
122
123 #ifndef HAVE_ZLIB_H
124
125 /*
126  * If we don't have zlib on this system, we can't actually do the
127  * decompression.  We can, however, still detect gzip-compressed
128  * archives and emit a useful message.
129  */
130 static int
131 init(struct archive *a, const void *buff, size_t n)
132 {
133         (void)a;        /* UNUSED */
134         (void)buff;     /* UNUSED */
135         (void)n;        /* UNUSED */
136
137         archive_set_error(a, -1,
138             "This version of libarchive was compiled without gzip support");
139         return (ARCHIVE_FATAL);
140 }
141
142
143 #else
144
145 /*
146  * Setup the callbacks.
147  */
148 static int
149 init(struct archive *a, const void *buff, size_t n)
150 {
151         struct private_data *state;
152         int ret;
153
154         a->compression_code = ARCHIVE_COMPRESSION_GZIP;
155         a->compression_name = "gzip";
156
157         state = malloc(sizeof(*state));
158         if (state == NULL) {
159                 archive_set_error(a, ENOMEM,
160                     "Can't allocate data for %s decompression",
161                     a->compression_name);
162                 return (ARCHIVE_FATAL);
163         }
164         memset(state, 0, sizeof(*state));
165
166         state->crc = crc32(0L, NULL, 0);
167         state->header_done = 0; /* We've not yet begun to parse header... */
168
169         state->uncompressed_buffer_size = 64 * 1024;
170         state->uncompressed_buffer = malloc(state->uncompressed_buffer_size);
171         state->stream.next_out = state->uncompressed_buffer;
172         state->read_next = state->uncompressed_buffer;
173         state->stream.avail_out = state->uncompressed_buffer_size;
174
175         if (state->uncompressed_buffer == NULL) {
176                 archive_set_error(a, ENOMEM,
177                     "Can't allocate %s decompression buffers",
178                     a->compression_name);
179                 free(state);
180                 return (ARCHIVE_FATAL);
181         }
182
183         /*
184          * A bug in zlib.h: stream.next_in should be marked 'const'
185          * but isn't (the library never alters data through the
186          * next_in pointer, only reads it).  The result: this ugly
187          * cast to remove 'const'.
188          */
189         state->stream.next_in = (void *)(uintptr_t)(const void *)buff;
190         state->stream.avail_in = n;
191
192         a->compression_read_ahead = read_ahead;
193         a->compression_read_consume = read_consume;
194         a->compression_finish = finish;
195
196         /*
197          * TODO: Do I need to parse the gzip header before calling
198          * inflateInit2()?  In particular, one of the header bytes
199          * marks "best compression" or "fastest", which may be
200          * appropriate for setting the second parameter here.
201          * However, I think the only penalty for not setting it
202          * correctly is wasted memory.  If this is necessary, it
203          * should probably go into drive_decompressor() below.
204          */
205
206         /* Initialize compression library. */
207         ret = inflateInit2(&(state->stream),
208             -15 /* Don't check for zlib header */);
209         if (ret == Z_OK) {
210                 a->compression_data = state;
211                 return (ARCHIVE_OK);
212         }
213
214         /* Library setup failed: Clean up. */
215         archive_set_error(a, ARCHIVE_ERRNO_MISC,
216             "Internal error initializing %s library", a->compression_name);
217         free(state->uncompressed_buffer);
218         free(state);
219
220         /* Override the error message if we know what really went wrong. */
221         switch (ret) {
222         case Z_STREAM_ERROR:
223                 archive_set_error(a, ARCHIVE_ERRNO_MISC,
224                     "Internal error initializing compression library: "
225                     "invalid setup parameter");
226                 break;
227         case Z_MEM_ERROR:
228                 archive_set_error(a, ENOMEM,
229                     "Internal error initializing compression library: "
230                     "out of memory");
231                 break;
232         case Z_VERSION_ERROR:
233                 archive_set_error(a, ARCHIVE_ERRNO_MISC,
234                     "Internal error initializing compression library: "
235                     "invalid library version");
236                 break;
237         }
238
239         return (ARCHIVE_FATAL);
240 }
241
242 /*
243  * Return a block of data from the decompression buffer.  Decompress more
244  * as necessary.
245  */
246 static ssize_t
247 read_ahead(struct archive *a, const void **p, size_t min)
248 {
249         struct private_data *state;
250         int read_avail, was_avail, ret;
251
252         state = a->compression_data;
253         was_avail = -1;
254         if (!a->client_reader) {
255                 archive_set_error(a, ARCHIVE_ERRNO_PROGRAMMER,
256                     "No read callback is registered?  "
257                     "This is probably an internal programming error.");
258                 return (ARCHIVE_FATAL);
259         }
260
261         read_avail = state->stream.next_out - state->read_next;
262
263         if (read_avail + state->stream.avail_out < min) {
264                 memmove(state->uncompressed_buffer, state->read_next,
265                     read_avail);
266                 state->read_next = state->uncompressed_buffer;
267                 state->stream.next_out = state->read_next + read_avail;
268                 state->stream.avail_out
269                     = state->uncompressed_buffer_size - read_avail;
270         }
271
272         while (was_avail < read_avail &&        /* Made some progress. */
273             read_avail < (int)min &&            /* Haven't satisfied min. */
274             read_avail < (int)state->uncompressed_buffer_size) { /* !full */
275                 if ((ret = drive_decompressor(a, state)) != ARCHIVE_OK)
276                         return (ret);
277                 was_avail = read_avail;
278                 read_avail = state->stream.next_out - state->read_next;
279         }
280
281         *p = state->read_next;
282         return (read_avail);
283 }
284
285 /*
286  * Mark a previously-returned block of data as read.
287  */
288 static ssize_t
289 read_consume(struct archive *a, size_t n)
290 {
291         struct private_data *state;
292
293         state = a->compression_data;
294         a->file_position += n;
295         state->read_next += n;
296         if (state->read_next > state->stream.next_out)
297                 __archive_errx(1, "Request to consume too many "
298                     "bytes from gzip decompressor");
299         return (n);
300 }
301
302 /*
303  * Clean up the decompressor.
304  */
305 static int
306 finish(struct archive *a)
307 {
308         struct private_data *state;
309         int ret;
310
311         state = a->compression_data;
312         ret = ARCHIVE_OK;
313         switch (inflateEnd(&(state->stream))) {
314         case Z_OK:
315                 break;
316         default:
317                 archive_set_error(a, ARCHIVE_ERRNO_MISC,
318                     "Failed to clean up %s compressor", a->compression_name);
319                 ret = ARCHIVE_FATAL;
320         }
321
322         free(state->uncompressed_buffer);
323         free(state);
324
325         a->compression_data = NULL;
326         if (a->client_closer != NULL)
327                 (a->client_closer)(a, a->client_data);
328
329         return (ret);
330 }
331
332 /*
333  * Utility function to pull data through decompressor, reading input
334  * blocks as necessary.
335  */
336 static int
337 drive_decompressor(struct archive *a, struct private_data *state)
338 {
339         ssize_t ret;
340         int decompressed, total_decompressed;
341         int count, flags, header_state;
342         unsigned char *output;
343         unsigned char b;
344
345         flags = 0;
346         count = 0;
347         header_state = 0;
348         total_decompressed = 0;
349         for (;;) {
350                 if (state->stream.avail_in == 0) {
351                         ret = (a->client_reader)(a, a->client_data,
352                             (const void **)&state->stream.next_in);
353                         if (ret < 0) {
354                                 /*
355                                  * TODO: Find a better way to handle
356                                  * this read failure.
357                                  */
358                                 goto fatal;
359                         }
360                         if (ret == 0  &&  total_decompressed == 0) {
361                                 archive_set_error(a, EIO,
362                                     "Premature end of %s compressed data",
363                                     a->compression_name);
364                                 return (ARCHIVE_FATAL);
365                         }
366                         a->raw_position += ret;
367                         state->stream.avail_in = ret;
368                 }
369
370                 if (!state->header_done) {
371                         /*
372                          * If still parsing the header, interpret the
373                          * next byte.
374                          */
375                         b = *(state->stream.next_in++);
376                         state->stream.avail_in--;
377
378                         /*
379                          * Yes, this is somewhat crude, but it works,
380                          * GZip format isn't likely to change anytime
381                          * in the near future, and header parsing is
382                          * certainly not a performance issue, so
383                          * there's little point in making this more
384                          * elegant.  Of course, if you see an easy way
385                          * to make this more elegant, please let me
386                          * know.. ;-)
387                          */
388                         switch (header_state) {
389                         case 0: /* First byte of signature. */
390                                 if (b != 037)
391                                         goto fatal;
392                                 header_state = 1;
393                                 break;
394                         case 1: /* Second byte of signature. */
395                                 if (b != 0213)
396                                         goto fatal;
397                                 header_state = 2;
398                                 break;
399                         case 2: /* Compression type must be 8. */
400                                 if (b != 8)
401                                         goto fatal;
402                                 header_state = 3;
403                                 break;
404                         case 3: /* GZip flags. */
405                                 flags = b;
406                                 header_state = 4;
407                                 break;
408                         case 4: case 5: case 6: case 7: /* Mod time. */
409                                 header_state++;
410                                 break;
411                         case 8: /* Deflate flags. */
412                                 header_state = 9;
413                                 break;
414                         case 9: /* OS. */
415                                 header_state = 10;
416                                 break;
417                         case 10: /* Optional Extra: First byte of Length. */
418                                 if ((flags & 4)) {
419                                         count = 255 & (int)b;
420                                         header_state = 11;
421                                         break;
422                                 }
423                                 /*
424                                  * Fall through if there is no
425                                  * Optional Extra field.
426                                  */
427                         case 11: /* Optional Extra: Second byte of Length. */
428                                 if ((flags & 4)) {
429                                         count = (count << 8) | (255 & (int)b);
430                                         header_state = 12;
431                                         break;
432                                 }
433                                 /*
434                                  * Fall through if there is no
435                                  * Optional Extra field.
436                                  */
437                         case 12: /* Optional Extra Field: counted length. */
438                                 if ((flags & 4)) {
439                                         --count;
440                                         if (count == 0) header_state = 13;
441                                         else header_state = 12;
442                                         break;
443                                 }
444                                 /*
445                                  * Fall through if there is no
446                                  * Optional Extra field.
447                                  */
448                         case 13: /* Optional Original Filename. */
449                                 if ((flags & 8)) {
450                                         if (b == 0) header_state = 14;
451                                         else header_state = 13;
452                                         break;
453                                 }
454                                 /*
455                                  * Fall through if no Optional
456                                  * Original Filename.
457                                  */
458                         case 14: /* Optional Comment. */
459                                 if ((flags & 16)) {
460                                         if (b == 0) header_state = 15;
461                                         else header_state = 14;
462                                         break;
463                                 }
464                                 /* Fall through if no Optional Comment. */
465                         case 15: /* Optional Header CRC: First byte. */
466                                 if ((flags & 2)) {
467                                         header_state = 16;
468                                         break;
469                                 }
470                                 /* Fall through if no Optional Header CRC. */
471                         case 16: /* Optional Header CRC: Second byte. */
472                                 if ((flags & 2)) {
473                                         header_state = 17;
474                                         break;
475                                 }
476                                 /* Fall through if no Optional Header CRC. */
477                         case 17: /* First byte of compressed data. */
478                                 state->header_done = 1; /* done with header */
479                                 state->stream.avail_in++;
480                                 state->stream.next_in--;
481                         }
482
483                         /*
484                          * TODO: Consider moving the inflateInit2 call
485                          * here so it can include the compression type
486                          * from the header?
487                          */
488                 } else {
489                         output = state->stream.next_out;
490
491                         /* Decompress some data. */
492                         ret = inflate(&(state->stream), 0);
493                         decompressed = state->stream.next_out - output;
494
495                         /* Accumulate the CRC of the uncompressed data. */
496                         state->crc = crc32(state->crc, output, decompressed);
497
498                         /* Accumulate the total bytes of output. */
499                         state->total_out += decompressed;
500                         total_decompressed += decompressed;
501
502                         switch (ret) {
503                         case Z_OK: /* Decompressor made some progress. */
504                                 if (decompressed > 0)
505                                         return (ARCHIVE_OK);
506                                 break;
507                         case Z_STREAM_END: /* Found end of stream. */
508                                 /*
509                                  * TODO: Verify gzip trailer
510                                  * (uncompressed length and CRC).
511                                  */
512                                 return (ARCHIVE_OK);
513                         default:
514                                 /* Any other return value is an error. */
515                                 goto fatal;
516                         }
517                 }
518         }
519         return (ARCHIVE_OK);
520
521         /* Return a fatal error. */
522 fatal:
523         archive_set_error(a, ARCHIVE_ERRNO_MISC, "%s decompression failed",
524             a->compression_name);
525         return (ARCHIVE_FATAL);
526 }
527
528 #endif /* HAVE_ZLIB_H */