1 /* gzread.c contains minimal changes required to be compiled with zlibWrapper:
2 * - gz_statep was converted to union to work with -Wstrict-aliasing=1 */
3
4 /* gzread.c -- zlib functions for reading gzip files
5 * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013, 2016 Mark Adler
6 * For conditions of distribution and use, see http://www.zlib.net/zlib_license.html
7 */
8
9 #include "gzguts.h"
10
11 /* fix for Visual Studio, which doesn't support ssize_t type.
12 * see https://github.com/facebook/zstd/issues/1800#issuecomment-545945050 */
13 #if defined(_MSC_VER) && !defined(ssize_t)
14 # include <BaseTsd.h>
15 typedef SSIZE_T ssize_t;
16 #endif
17
18
19 /* Local functions */
20 local int gz_load OF((gz_statep, unsigned char *, unsigned, unsigned *));
21 local int gz_avail OF((gz_statep));
22 local int gz_look OF((gz_statep));
23 local int gz_decomp OF((gz_statep));
24 local int gz_fetch OF((gz_statep));
25 local int gz_skip OF((gz_statep, z_off64_t));
26 local z_size_t gz_read OF((gz_statep, voidp, z_size_t));
27
28 /* Use read() to load a buffer -- return -1 on error, otherwise 0. Read from
29 state.state->fd, and update state.state->eof, state.state->err, and state.state->msg as appropriate.
30 This function needs to loop on read(), since read() is not guaranteed to
31 read the number of bytes requested, depending on the type of descriptor. */
gz_load(state,buf,len,have)32 local int gz_load(state, buf, len, have)
33 gz_statep state;
34 unsigned char *buf;
35 unsigned len;
36 unsigned *have;
37 {
38 ssize_t ret;
39 unsigned get, max = ((unsigned)-1 >> 2) + 1;
40
41 *have = 0;
42 do {
43 get = len - *have;
44 if (get > max)
45 get = max;
46 ret = read(state.state->fd, buf + *have, get);
47 if (ret <= 0)
48 break;
49 *have += (unsigned)ret;
50 } while (*have < len);
51 if (ret < 0) {
52 gz_error(state, Z_ERRNO, zstrerror());
53 return -1;
54 }
55 if (ret == 0)
56 state.state->eof = 1;
57 return 0;
58 }
59
60 /* Load up input buffer and set eof flag if last data loaded -- return -1 on
61 error, 0 otherwise. Note that the eof flag is set when the end of the input
62 file is reached, even though there may be unused data in the buffer. Once
63 that data has been used, no more attempts will be made to read the file.
64 If strm->avail_in != 0, then the current data is moved to the beginning of
65 the input buffer, and then the remainder of the buffer is loaded with the
66 available data from the input file. */
gz_avail(state)67 local int gz_avail(state)
68 gz_statep state;
69 {
70 unsigned got;
71 z_streamp strm = &(state.state->strm);
72
73 if (state.state->err != Z_OK && state.state->err != Z_BUF_ERROR)
74 return -1;
75 if (state.state->eof == 0) {
76 if (strm->avail_in) { /* copy what's there to the start */
77 unsigned char *p = state.state->in;
78 unsigned const char *q = strm->next_in;
79 unsigned n = strm->avail_in;
80 do {
81 *p++ = *q++;
82 } while (--n);
83 }
84 if (gz_load(state, state.state->in + strm->avail_in,
85 state.state->size - strm->avail_in, &got) == -1)
86 return -1;
87 strm->avail_in += got;
88 strm->next_in = state.state->in;
89 }
90 return 0;
91 }
92
93 /* Look for gzip header, set up for inflate or copy. state.state->x.have must be 0.
94 If this is the first time in, allocate required memory. state.state->how will be
95 left unchanged if there is no more input data available, will be set to COPY
96 if there is no gzip header and direct copying will be performed, or it will
97 be set to GZIP for decompression. If direct copying, then leftover input
98 data from the input buffer will be copied to the output buffer. In that
99 case, all further file reads will be directly to either the output buffer or
100 a user buffer. If decompressing, the inflate state will be initialized.
101 gz_look() will return 0 on success or -1 on failure. */
gz_look(state)102 local int gz_look(state)
103 gz_statep state;
104 {
105 z_streamp strm = &(state.state->strm);
106
107 /* allocate read buffers and inflate memory */
108 if (state.state->size == 0) {
109 /* allocate buffers */
110 state.state->in = (unsigned char *)malloc(state.state->want);
111 state.state->out = (unsigned char *)malloc(state.state->want << 1);
112 if (state.state->in == NULL || state.state->out == NULL) {
113 free(state.state->out);
114 free(state.state->in);
115 gz_error(state, Z_MEM_ERROR, "out of memory");
116 return -1;
117 }
118 state.state->size = state.state->want;
119
120 /* allocate inflate memory */
121 state.state->strm.zalloc = Z_NULL;
122 state.state->strm.zfree = Z_NULL;
123 state.state->strm.opaque = Z_NULL;
124 state.state->strm.avail_in = 0;
125 state.state->strm.next_in = Z_NULL;
126 if (inflateInit2(&(state.state->strm), 15 + 16) != Z_OK) { /* gunzip */
127 free(state.state->out);
128 free(state.state->in);
129 state.state->size = 0;
130 gz_error(state, Z_MEM_ERROR, "out of memory");
131 return -1;
132 }
133 }
134
135 /* get at least the magic bytes in the input buffer */
136 if (strm->avail_in < 2) {
137 if (gz_avail(state) == -1)
138 return -1;
139 if (strm->avail_in == 0)
140 return 0;
141 }
142
143 /* look for gzip magic bytes -- if there, do gzip decoding (note: there is
144 a logical dilemma here when considering the case of a partially written
145 gzip file, to wit, if a single 31 byte is written, then we cannot tell
146 whether this is a single-byte file, or just a partially written gzip
147 file -- for here we assume that if a gzip file is being written, then
148 the header will be written in a single operation, so that reading a
149 single byte is sufficient indication that it is not a gzip file) */
150 if (strm->avail_in > 1 &&
151 ((strm->next_in[0] == 31 && strm->next_in[1] == 139) /* gz header */
152 || (strm->next_in[0] == 40 && strm->next_in[1] == 181))) { /* zstd header */
153 inflateReset(strm);
154 state.state->how = GZIP;
155 state.state->direct = 0;
156 return 0;
157 }
158
159 /* no gzip header -- if we were decoding gzip before, then this is trailing
160 garbage. Ignore the trailing garbage and finish. */
161 if (state.state->direct == 0) {
162 strm->avail_in = 0;
163 state.state->eof = 1;
164 state.state->x.have = 0;
165 return 0;
166 }
167
168 /* doing raw i/o, copy any leftover input to output -- this assumes that
169 the output buffer is larger than the input buffer, which also assures
170 space for gzungetc() */
171 state.state->x.next = state.state->out;
172 if (strm->avail_in) {
173 memcpy(state.state->x.next, strm->next_in, strm->avail_in);
174 state.state->x.have = strm->avail_in;
175 strm->avail_in = 0;
176 }
177 state.state->how = COPY;
178 state.state->direct = 1;
179 return 0;
180 }
181
182 /* Decompress from input to the provided next_out and avail_out in the state.
183 On return, state.state->x.have and state.state->x.next point to the just decompressed
184 data. If the gzip stream completes, state.state->how is reset to LOOK to look for
185 the next gzip stream or raw data, once state.state->x.have is depleted. Returns 0
186 on success, -1 on failure. */
gz_decomp(state)187 local int gz_decomp(state)
188 gz_statep state;
189 {
190 int ret = Z_OK;
191 unsigned had;
192 z_streamp strm = &(state.state->strm);
193
194 /* fill output buffer up to end of deflate stream */
195 had = strm->avail_out;
196 do {
197 /* get more input for inflate() */
198 if (strm->avail_in == 0 && gz_avail(state) == -1)
199 return -1;
200 if (strm->avail_in == 0) {
201 gz_error(state, Z_BUF_ERROR, "unexpected end of file");
202 break;
203 }
204
205 /* decompress and handle errors */
206 ret = inflate(strm, Z_NO_FLUSH);
207 if (ret == Z_STREAM_ERROR || ret == Z_NEED_DICT) {
208 gz_error(state, Z_STREAM_ERROR,
209 "internal error: inflate stream corrupt");
210 return -1;
211 }
212 if (ret == Z_MEM_ERROR) {
213 gz_error(state, Z_MEM_ERROR, "out of memory");
214 return -1;
215 }
216 if (ret == Z_DATA_ERROR) { /* deflate stream invalid */
217 gz_error(state, Z_DATA_ERROR,
218 strm->msg == NULL ? "compressed data error" : strm->msg);
219 return -1;
220 }
221 } while (strm->avail_out && ret != Z_STREAM_END);
222
223 /* update available output */
224 state.state->x.have = had - strm->avail_out;
225 state.state->x.next = strm->next_out - state.state->x.have;
226
227 /* if the gzip stream completed successfully, look for another */
228 if (ret == Z_STREAM_END)
229 state.state->how = LOOK;
230
231 /* good decompression */
232 return 0;
233 }
234
235 /* Fetch data and put it in the output buffer. Assumes state.state->x.have is 0.
236 Data is either copied from the input file or decompressed from the input
237 file depending on state.state->how. If state.state->how is LOOK, then a gzip header is
238 looked for to determine whether to copy or decompress. Returns -1 on error,
239 otherwise 0. gz_fetch() will leave state.state->how as COPY or GZIP unless the
240 end of the input file has been reached and all data has been processed. */
gz_fetch(state)241 local int gz_fetch(state)
242 gz_statep state;
243 {
244 z_streamp strm = &(state.state->strm);
245
246 do {
247 switch(state.state->how) {
248 case LOOK: /* -> LOOK, COPY (only if never GZIP), or GZIP */
249 if (gz_look(state) == -1)
250 return -1;
251 if (state.state->how == LOOK)
252 return 0;
253 break;
254 case COPY: /* -> COPY */
255 if (gz_load(state, state.state->out, state.state->size << 1, &(state.state->x.have))
256 == -1)
257 return -1;
258 state.state->x.next = state.state->out;
259 return 0;
260 case GZIP: /* -> GZIP or LOOK (if end of gzip stream) */
261 strm->avail_out = state.state->size << 1;
262 strm->next_out = state.state->out;
263 if (gz_decomp(state) == -1)
264 return -1;
265 }
266 } while (state.state->x.have == 0 && (!state.state->eof || strm->avail_in));
267 return 0;
268 }
269
270 /* Skip len uncompressed bytes of output. Return -1 on error, 0 on success. */
gz_skip(state,len)271 local int gz_skip(state, len)
272 gz_statep state;
273 z_off64_t len;
274 {
275 unsigned n;
276
277 /* skip over len bytes or reach end-of-file, whichever comes first */
278 while (len)
279 /* skip over whatever is in output buffer */
280 if (state.state->x.have) {
281 n = GT_OFF(state.state->x.have) || (z_off64_t)state.state->x.have > len ?
282 (unsigned)len : state.state->x.have;
283 state.state->x.have -= n;
284 state.state->x.next += n;
285 state.state->x.pos += n;
286 len -= n;
287 }
288
289 /* output buffer empty -- return if we're at the end of the input */
290 else if (state.state->eof && state.state->strm.avail_in == 0)
291 break;
292
293 /* need more data to skip -- load up output buffer */
294 else {
295 /* get more output, looking for header if required */
296 if (gz_fetch(state) == -1)
297 return -1;
298 }
299 return 0;
300 }
301
302 /* Read len bytes into buf from file, or less than len up to the end of the
303 input. Return the number of bytes read. If zero is returned, either the
304 end of file was reached, or there was an error. state.state->err must be
305 consulted in that case to determine which. */
gz_read(state,buf,len)306 local z_size_t gz_read(state, buf, len)
307 gz_statep state;
308 voidp buf;
309 z_size_t len;
310 {
311 z_size_t got;
312 unsigned n;
313
314 /* if len is zero, avoid unnecessary operations */
315 if (len == 0)
316 return 0;
317
318 /* process a skip request */
319 if (state.state->seek) {
320 state.state->seek = 0;
321 if (gz_skip(state, state.state->skip) == -1)
322 return 0;
323 }
324
325 /* get len bytes to buf, or less than len if at the end */
326 got = 0;
327 do {
328 /* set n to the maximum amount of len that fits in an unsigned int */
329 n = -1;
330 if (n > len)
331 n = (unsigned)len;
332
333 /* first just try copying data from the output buffer */
334 if (state.state->x.have) {
335 if (state.state->x.have < n)
336 n = state.state->x.have;
337 memcpy(buf, state.state->x.next, n);
338 state.state->x.next += n;
339 state.state->x.have -= n;
340 }
341
342 /* output buffer empty -- return if we're at the end of the input */
343 else if (state.state->eof && state.state->strm.avail_in == 0) {
344 state.state->past = 1; /* tried to read past end */
345 break;
346 }
347
348 /* need output data -- for small len or new stream load up our output
349 buffer */
350 else if (state.state->how == LOOK || n < (state.state->size << 1)) {
351 /* get more output, looking for header if required */
352 if (gz_fetch(state) == -1)
353 return 0;
354 continue; /* no progress yet -- go back to copy above */
355 /* the copy above assures that we will leave with space in the
356 output buffer, allowing at least one gzungetc() to succeed */
357 }
358
359 /* large len -- read directly into user buffer */
360 else if (state.state->how == COPY) { /* read directly */
361 if (gz_load(state, (unsigned char *)buf, n, &n) == -1)
362 return 0;
363 }
364
365 /* large len -- decompress directly into user buffer */
366 else { /* state.state->how == GZIP */
367 state.state->strm.avail_out = n;
368 state.state->strm.next_out = (unsigned char *)buf;
369 if (gz_decomp(state) == -1)
370 return 0;
371 n = state.state->x.have;
372 state.state->x.have = 0;
373 }
374
375 /* update progress */
376 len -= n;
377 buf = (char *)buf + n;
378 got += n;
379 state.state->x.pos += n;
380 } while (len);
381
382 /* return number of bytes read into user buffer */
383 return got;
384 }
385
386 /* -- see zlib.h -- */
gzread(file,buf,len)387 int ZEXPORT gzread(file, buf, len)
388 gzFile file;
389 voidp buf;
390 unsigned len;
391 {
392 gz_statep state;
393
394 /* get internal structure */
395 if (file == NULL)
396 return -1;
397 state.file = file;
398
399 /* check that we're reading and that there's no (serious) error */
400 if (state.state->mode != GZ_READ ||
401 (state.state->err != Z_OK && state.state->err != Z_BUF_ERROR))
402 return -1;
403
404 /* since an int is returned, make sure len fits in one, otherwise return
405 with an error (this avoids a flaw in the interface) */
406 if ((int)len < 0) {
407 gz_error(state, Z_STREAM_ERROR, "request does not fit in an int");
408 return -1;
409 }
410
411 /* read len or fewer bytes to buf */
412 len = (unsigned)gz_read(state, buf, len);
413
414 /* check for an error */
415 if (len == 0 && state.state->err != Z_OK && state.state->err != Z_BUF_ERROR)
416 return -1;
417
418 /* return the number of bytes read (this is assured to fit in an int) */
419 return (int)len;
420 }
421
422 /* -- see zlib.h -- */
gzfread(buf,size,nitems,file)423 z_size_t ZEXPORT gzfread(buf, size, nitems, file)
424 voidp buf;
425 z_size_t size;
426 z_size_t nitems;
427 gzFile file;
428 {
429 z_size_t len;
430 gz_statep state;
431
432 /* get internal structure */
433 if (file == NULL)
434 return 0;
435 state.file = file;
436
437 /* check that we're reading and that there's no (serious) error */
438 if (state.state->mode != GZ_READ ||
439 (state.state->err != Z_OK && state.state->err != Z_BUF_ERROR))
440 return 0;
441
442 /* compute bytes to read -- error on overflow */
443 len = nitems * size;
444 if (size && len / size != nitems) {
445 gz_error(state, Z_STREAM_ERROR, "request does not fit in a size_t");
446 return 0;
447 }
448
449 /* read len or fewer bytes to buf, return the number of full items read */
450 return len ? gz_read(state, buf, len) / size : 0;
451 }
452
453 /* -- see zlib.h -- */
454 #if ZLIB_VERNUM >= 0x1261
455 #ifdef Z_PREFIX_SET
456 # undef z_gzgetc
457 #else
458 # undef gzgetc
459 #endif
460 #endif
461
462 #if ZLIB_VERNUM == 0x1260
463 # undef gzgetc
464 #endif
465
466 #if ZLIB_VERNUM <= 0x1250
467 ZEXTERN int ZEXPORT gzgetc OF((gzFile file));
468 ZEXTERN int ZEXPORT gzgetc_ OF((gzFile file));
469 #endif
470
gzgetc(file)471 int ZEXPORT gzgetc(file)
472 gzFile file;
473 {
474 int ret;
475 unsigned char buf[1];
476 gz_statep state;
477
478 /* get internal structure */
479 if (file == NULL)
480 return -1;
481 state.file = file;
482
483 /* check that we're reading and that there's no (serious) error */
484 if (state.state->mode != GZ_READ ||
485 (state.state->err != Z_OK && state.state->err != Z_BUF_ERROR))
486 return -1;
487
488 /* try output buffer (no need to check for skip request) */
489 if (state.state->x.have) {
490 state.state->x.have--;
491 state.state->x.pos++;
492 return *(state.state->x.next)++;
493 }
494
495 /* nothing there -- try gz_read() */
496 ret = (int)gz_read(state, buf, 1);
497 return ret < 1 ? -1 : buf[0];
498 }
499
gzgetc_(file)500 int ZEXPORT gzgetc_(file)
501 gzFile file;
502 {
503 return gzgetc(file);
504 }
505
506 /* -- see zlib.h -- */
gzungetc(c,file)507 int ZEXPORT gzungetc(c, file)
508 int c;
509 gzFile file;
510 {
511 gz_statep state;
512
513 /* get internal structure */
514 if (file == NULL)
515 return -1;
516 state.file = file;
517
518 /* check that we're reading and that there's no (serious) error */
519 if (state.state->mode != GZ_READ ||
520 (state.state->err != Z_OK && state.state->err != Z_BUF_ERROR))
521 return -1;
522
523 /* process a skip request */
524 if (state.state->seek) {
525 state.state->seek = 0;
526 if (gz_skip(state, state.state->skip) == -1)
527 return -1;
528 }
529
530 /* can't push EOF */
531 if (c < 0)
532 return -1;
533
534 /* if output buffer empty, put byte at end (allows more pushing) */
535 if (state.state->x.have == 0) {
536 state.state->x.have = 1;
537 state.state->x.next = state.state->out + (state.state->size << 1) - 1;
538 state.state->x.next[0] = (unsigned char)c;
539 state.state->x.pos--;
540 state.state->past = 0;
541 return c;
542 }
543
544 /* if no room, give up (must have already done a gzungetc()) */
545 if (state.state->x.have == (state.state->size << 1)) {
546 gz_error(state, Z_DATA_ERROR, "out of room to push characters");
547 return -1;
548 }
549
550 /* slide output data if needed and insert byte before existing data */
551 if (state.state->x.next == state.state->out) {
552 unsigned char *src = state.state->out + state.state->x.have;
553 unsigned char *dest = state.state->out + (state.state->size << 1);
554 while (src > state.state->out)
555 *--dest = *--src;
556 state.state->x.next = dest;
557 }
558 state.state->x.have++;
559 state.state->x.next--;
560 state.state->x.next[0] = (unsigned char)c;
561 state.state->x.pos--;
562 state.state->past = 0;
563 return c;
564 }
565
566 /* -- see zlib.h -- */
gzgets(file,buf,len)567 char * ZEXPORT gzgets(file, buf, len)
568 gzFile file;
569 char *buf;
570 int len;
571 {
572 unsigned left, n;
573 char *str;
574 unsigned char *eol;
575 gz_statep state;
576
577 /* check parameters and get internal structure */
578 if (file == NULL || buf == NULL || len < 1)
579 return NULL;
580 state.file = file;
581
582 /* check that we're reading and that there's no (serious) error */
583 if (state.state->mode != GZ_READ ||
584 (state.state->err != Z_OK && state.state->err != Z_BUF_ERROR))
585 return NULL;
586
587 /* process a skip request */
588 if (state.state->seek) {
589 state.state->seek = 0;
590 if (gz_skip(state, state.state->skip) == -1)
591 return NULL;
592 }
593
594 /* copy output bytes up to new line or len - 1, whichever comes first --
595 append a terminating zero to the string (we don't check for a zero in
596 the contents, let the user worry about that) */
597 str = buf;
598 left = (unsigned)len - 1;
599 if (left) do {
600 /* assure that something is in the output buffer */
601 if (state.state->x.have == 0 && gz_fetch(state) == -1)
602 return NULL; /* error */
603 if (state.state->x.have == 0) { /* end of file */
604 state.state->past = 1; /* read past end */
605 break; /* return what we have */
606 }
607
608 /* look for end-of-line in current output buffer */
609 n = state.state->x.have > left ? left : state.state->x.have;
610 eol = (unsigned char *)memchr(state.state->x.next, '\n', n);
611 if (eol != NULL)
612 n = (unsigned)(eol - state.state->x.next) + 1;
613
614 /* copy through end-of-line, or remainder if not found */
615 memcpy(buf, state.state->x.next, n);
616 state.state->x.have -= n;
617 state.state->x.next += n;
618 state.state->x.pos += n;
619 left -= n;
620 buf += n;
621 } while (left && eol == NULL);
622
623 /* return terminated string, or if nothing, end of file */
624 if (buf == str)
625 return NULL;
626 buf[0] = 0;
627 return str;
628 }
629
630 /* -- see zlib.h -- */
gzdirect(file)631 int ZEXPORT gzdirect(file)
632 gzFile file;
633 {
634 gz_statep state;
635
636 /* get internal structure */
637 if (file == NULL)
638 return 0;
639 state.file = file;
640
641 /* if the state is not known, but we can find out, then do so (this is
642 mainly for right after a gzopen() or gzdopen()) */
643 if (state.state->mode == GZ_READ && state.state->how == LOOK && state.state->x.have == 0)
644 (void)gz_look(state);
645
646 /* return 1 if transparent, 0 if processing a gzip stream */
647 return state.state->direct;
648 }
649
650 /* -- see zlib.h -- */
gzclose_r(file)651 int ZEXPORT gzclose_r(file)
652 gzFile file;
653 {
654 int ret, err;
655 gz_statep state;
656
657 /* get internal structure */
658 if (file == NULL)
659 return Z_STREAM_ERROR;
660 state.file = file;
661
662 /* check that we're reading */
663 if (state.state->mode != GZ_READ)
664 return Z_STREAM_ERROR;
665
666 /* free memory and close file */
667 if (state.state->size) {
668 inflateEnd(&(state.state->strm));
669 free(state.state->out);
670 free(state.state->in);
671 }
672 err = state.state->err == Z_BUF_ERROR ? Z_BUF_ERROR : Z_OK;
673 gz_error(state, Z_OK, NULL);
674 free(state.state->path);
675 ret = close(state.state->fd);
676 free(state.state);
677 return ret ? Z_ERRNO : err;
678 }
679