xref: /freebsd/sys/contrib/zlib/gzread.c (revision 22cf89c938886d14f5796fc49f9f020c23ea8eaf)
1 /* gzread.c -- zlib functions for reading gzip files
2  * Copyright (C) 2004-2017 Mark Adler
3  * For conditions of distribution and use, see copyright notice in zlib.h
4  */
5 
6 #include "gzguts.h"
7 #include <unistd.h>
8 
9 /* Use read() to load a buffer -- return -1 on error, otherwise 0.  Read from
10    state->fd, and update state->eof, state->err, and state->msg as appropriate.
11    This function needs to loop on read(), since read() is not guaranteed to
12    read the number of bytes requested, depending on the type of descriptor. */
13 local int gz_load(gz_statep state, unsigned char *buf, unsigned len,
14                   unsigned *have) {
15     int ret;
16     unsigned get, max = ((unsigned)-1 >> 2) + 1;
17 
18     *have = 0;
19     do {
20         get = len - *have;
21         if (get > max)
22             get = max;
23         ret = read(state->fd, buf + *have, get);
24         if (ret <= 0)
25             break;
26         *have += (unsigned)ret;
27     } while (*have < len);
28     if (ret < 0) {
29         gz_error(state, Z_ERRNO, zstrerror());
30         return -1;
31     }
32     if (ret == 0)
33         state->eof = 1;
34     return 0;
35 }
36 
37 /* Load up input buffer and set eof flag if last data loaded -- return -1 on
38    error, 0 otherwise.  Note that the eof flag is set when the end of the input
39    file is reached, even though there may be unused data in the buffer.  Once
40    that data has been used, no more attempts will be made to read the file.
41    If strm->avail_in != 0, then the current data is moved to the beginning of
42    the input buffer, and then the remainder of the buffer is loaded with the
43    available data from the input file. */
44 local int gz_avail(gz_statep state) {
45     unsigned got;
46     z_streamp strm = &(state->strm);
47 
48     if (state->err != Z_OK && state->err != Z_BUF_ERROR)
49         return -1;
50     if (state->eof == 0) {
51         if (strm->avail_in) {       /* copy what's there to the start */
52             unsigned char *p = state->in;
53             unsigned const char *q = strm->next_in;
54             unsigned n = strm->avail_in;
55             do {
56                 *p++ = *q++;
57             } while (--n);
58         }
59         if (gz_load(state, state->in + strm->avail_in,
60                     state->size - strm->avail_in, &got) == -1)
61             return -1;
62         strm->avail_in += got;
63         strm->next_in = state->in;
64     }
65     return 0;
66 }
67 
68 /* Look for gzip header, set up for inflate or copy.  state->x.have must be 0.
69    If this is the first time in, allocate required memory.  state->how will be
70    left unchanged if there is no more input data available, will be set to COPY
71    if there is no gzip header and direct copying will be performed, or it will
72    be set to GZIP for decompression.  If direct copying, then leftover input
73    data from the input buffer will be copied to the output buffer.  In that
74    case, all further file reads will be directly to either the output buffer or
75    a user buffer.  If decompressing, the inflate state will be initialized.
76    gz_look() will return 0 on success or -1 on failure. */
77 local int gz_look(gz_statep state) {
78     z_streamp strm = &(state->strm);
79 
80     /* allocate read buffers and inflate memory */
81     if (state->size == 0) {
82         /* allocate buffers */
83         state->in = (unsigned char *)malloc(state->want);
84         state->out = (unsigned char *)malloc(state->want << 1);
85         if (state->in == NULL || state->out == NULL) {
86             free(state->out);
87             free(state->in);
88             gz_error(state, Z_MEM_ERROR, "out of memory");
89             return -1;
90         }
91         state->size = state->want;
92 
93         /* allocate inflate memory */
94         state->strm.zalloc = Z_NULL;
95         state->strm.zfree = Z_NULL;
96         state->strm.opaque = Z_NULL;
97         state->strm.avail_in = 0;
98         state->strm.next_in = Z_NULL;
99         if (inflateInit2(&(state->strm), 15 + 16) != Z_OK) {    /* gunzip */
100             free(state->out);
101             free(state->in);
102             state->size = 0;
103             gz_error(state, Z_MEM_ERROR, "out of memory");
104             return -1;
105         }
106     }
107 
108     /* get at least the magic bytes in the input buffer */
109     if (strm->avail_in < 2) {
110         if (gz_avail(state) == -1)
111             return -1;
112         if (strm->avail_in == 0)
113             return 0;
114     }
115 
116     /* look for gzip magic bytes -- if there, do gzip decoding (note: there is
117        a logical dilemma here when considering the case of a partially written
118        gzip file, to wit, if a single 31 byte is written, then we cannot tell
119        whether this is a single-byte file, or just a partially written gzip
120        file -- for here we assume that if a gzip file is being written, then
121        the header will be written in a single operation, so that reading a
122        single byte is sufficient indication that it is not a gzip file) */
123     if (strm->avail_in > 1 &&
124             strm->next_in[0] == 31 && strm->next_in[1] == 139) {
125         inflateReset(strm);
126         state->how = GZIP;
127         state->direct = 0;
128         return 0;
129     }
130 
131     /* no gzip header -- if we were decoding gzip before, then this is trailing
132        garbage.  Ignore the trailing garbage and finish. */
133     if (state->direct == 0) {
134         strm->avail_in = 0;
135         state->eof = 1;
136         state->x.have = 0;
137         return 0;
138     }
139 
140     /* doing raw i/o, copy any leftover input to output -- this assumes that
141        the output buffer is larger than the input buffer, which also assures
142        space for gzungetc() */
143     state->x.next = state->out;
144     memcpy(state->x.next, strm->next_in, strm->avail_in);
145     state->x.have = strm->avail_in;
146     strm->avail_in = 0;
147     state->how = COPY;
148     state->direct = 1;
149     return 0;
150 }
151 
152 /* Decompress from input to the provided next_out and avail_out in the state.
153    On return, state->x.have and state->x.next point to the just decompressed
154    data.  If the gzip stream completes, state->how is reset to LOOK to look for
155    the next gzip stream or raw data, once state->x.have is depleted.  Returns 0
156    on success, -1 on failure. */
157 local int gz_decomp(gz_statep state) {
158     int ret = Z_OK;
159     unsigned had;
160     z_streamp strm = &(state->strm);
161 
162     /* fill output buffer up to end of deflate stream */
163     had = strm->avail_out;
164     do {
165         /* get more input for inflate() */
166         if (strm->avail_in == 0 && gz_avail(state) == -1)
167             return -1;
168         if (strm->avail_in == 0) {
169             gz_error(state, Z_BUF_ERROR, "unexpected end of file");
170             break;
171         }
172 
173         /* decompress and handle errors */
174         ret = inflate(strm, Z_NO_FLUSH);
175         if (ret == Z_STREAM_ERROR || ret == Z_NEED_DICT) {
176             gz_error(state, Z_STREAM_ERROR,
177                      "internal error: inflate stream corrupt");
178             return -1;
179         }
180         if (ret == Z_MEM_ERROR) {
181             gz_error(state, Z_MEM_ERROR, "out of memory");
182             return -1;
183         }
184         if (ret == Z_DATA_ERROR) {              /* deflate stream invalid */
185             gz_error(state, Z_DATA_ERROR,
186                      strm->msg == NULL ? "compressed data error" : strm->msg);
187             return -1;
188         }
189     } while (strm->avail_out && ret != Z_STREAM_END);
190 
191     /* update available output */
192     state->x.have = had - strm->avail_out;
193     state->x.next = strm->next_out - state->x.have;
194 
195     /* if the gzip stream completed successfully, look for another */
196     if (ret == Z_STREAM_END)
197         state->how = LOOK;
198 
199     /* good decompression */
200     return 0;
201 }
202 
203 /* Fetch data and put it in the output buffer.  Assumes state->x.have is 0.
204    Data is either copied from the input file or decompressed from the input
205    file depending on state->how.  If state->how is LOOK, then a gzip header is
206    looked for to determine whether to copy or decompress.  Returns -1 on error,
207    otherwise 0.  gz_fetch() will leave state->how as COPY or GZIP unless the
208    end of the input file has been reached and all data has been processed.  */
209 local int gz_fetch(gz_statep state) {
210     z_streamp strm = &(state->strm);
211 
212     do {
213         switch(state->how) {
214         case LOOK:      /* -> LOOK, COPY (only if never GZIP), or GZIP */
215             if (gz_look(state) == -1)
216                 return -1;
217             if (state->how == LOOK)
218                 return 0;
219             break;
220         case COPY:      /* -> COPY */
221             if (gz_load(state, state->out, state->size << 1, &(state->x.have))
222                     == -1)
223                 return -1;
224             state->x.next = state->out;
225             return 0;
226         case GZIP:      /* -> GZIP or LOOK (if end of gzip stream) */
227             strm->avail_out = state->size << 1;
228             strm->next_out = state->out;
229             if (gz_decomp(state) == -1)
230                 return -1;
231         }
232     } while (state->x.have == 0 && (!state->eof || strm->avail_in));
233     return 0;
234 }
235 
236 /* Skip len uncompressed bytes of output.  Return -1 on error, 0 on success. */
237 local int gz_skip(gz_statep state, z_off64_t len) {
238     unsigned n;
239 
240     /* skip over len bytes or reach end-of-file, whichever comes first */
241     while (len)
242         /* skip over whatever is in output buffer */
243         if (state->x.have) {
244             n = GT_OFF(state->x.have) || (z_off64_t)state->x.have > len ?
245                 (unsigned)len : state->x.have;
246             state->x.have -= n;
247             state->x.next += n;
248             state->x.pos += n;
249             len -= n;
250         }
251 
252         /* output buffer empty -- return if we're at the end of the input */
253         else if (state->eof && state->strm.avail_in == 0)
254             break;
255 
256         /* need more data to skip -- load up output buffer */
257         else {
258             /* get more output, looking for header if required */
259             if (gz_fetch(state) == -1)
260                 return -1;
261         }
262     return 0;
263 }
264 
265 /* Read len bytes into buf from file, or less than len up to the end of the
266    input.  Return the number of bytes read.  If zero is returned, either the
267    end of file was reached, or there was an error.  state->err must be
268    consulted in that case to determine which. */
269 local z_size_t gz_read(gz_statep state, voidp buf, z_size_t len) {
270     z_size_t got;
271     unsigned n;
272 
273     /* if len is zero, avoid unnecessary operations */
274     if (len == 0)
275         return 0;
276 
277     /* process a skip request */
278     if (state->seek) {
279         state->seek = 0;
280         if (gz_skip(state, state->skip) == -1)
281             return 0;
282     }
283 
284     /* get len bytes to buf, or less than len if at the end */
285     got = 0;
286     do {
287         /* set n to the maximum amount of len that fits in an unsigned int */
288         n = (unsigned)-1;
289         if (n > len)
290             n = (unsigned)len;
291 
292         /* first just try copying data from the output buffer */
293         if (state->x.have) {
294             if (state->x.have < n)
295                 n = state->x.have;
296             memcpy(buf, state->x.next, n);
297             state->x.next += n;
298             state->x.have -= n;
299         }
300 
301         /* output buffer empty -- return if we're at the end of the input */
302         else if (state->eof && state->strm.avail_in == 0) {
303             state->past = 1;        /* tried to read past end */
304             break;
305         }
306 
307         /* need output data -- for small len or new stream load up our output
308            buffer */
309         else if (state->how == LOOK || n < (state->size << 1)) {
310             /* get more output, looking for header if required */
311             if (gz_fetch(state) == -1)
312                 return 0;
313             continue;       /* no progress yet -- go back to copy above */
314             /* the copy above assures that we will leave with space in the
315                output buffer, allowing at least one gzungetc() to succeed */
316         }
317 
318         /* large len -- read directly into user buffer */
319         else if (state->how == COPY) {      /* read directly */
320             if (gz_load(state, (unsigned char *)buf, n, &n) == -1)
321                 return 0;
322         }
323 
324         /* large len -- decompress directly into user buffer */
325         else {  /* state->how == GZIP */
326             state->strm.avail_out = n;
327             state->strm.next_out = (unsigned char *)buf;
328             if (gz_decomp(state) == -1)
329                 return 0;
330             n = state->x.have;
331             state->x.have = 0;
332         }
333 
334         /* update progress */
335         len -= n;
336         buf = (char *)buf + n;
337         got += n;
338         state->x.pos += n;
339     } while (len);
340 
341     /* return number of bytes read into user buffer */
342     return got;
343 }
344 
345 /* -- see zlib.h -- */
346 int ZEXPORT gzread(gzFile file, voidp buf, unsigned len) {
347     gz_statep state;
348 
349     /* get internal structure */
350     if (file == NULL)
351         return -1;
352     state = (gz_statep)file;
353 
354     /* check that we're reading and that there's no (serious) error */
355     if (state->mode != GZ_READ ||
356             (state->err != Z_OK && state->err != Z_BUF_ERROR))
357         return -1;
358 
359     /* since an int is returned, make sure len fits in one, otherwise return
360        with an error (this avoids a flaw in the interface) */
361     if ((int)len < 0) {
362         gz_error(state, Z_STREAM_ERROR, "request does not fit in an int");
363         return -1;
364     }
365 
366     /* read len or fewer bytes to buf */
367     len = (unsigned)gz_read(state, buf, len);
368 
369     /* check for an error */
370     if (len == 0 && state->err != Z_OK && state->err != Z_BUF_ERROR)
371         return -1;
372 
373     /* return the number of bytes read (this is assured to fit in an int) */
374     return (int)len;
375 }
376 
377 /* -- see zlib.h -- */
378 z_size_t ZEXPORT gzfread(voidp buf, z_size_t size, z_size_t nitems, gzFile file) {
379     z_size_t len;
380     gz_statep state;
381 
382     /* get internal structure */
383     if (file == NULL)
384         return 0;
385     state = (gz_statep)file;
386 
387     /* check that we're reading and that there's no (serious) error */
388     if (state->mode != GZ_READ ||
389             (state->err != Z_OK && state->err != Z_BUF_ERROR))
390         return 0;
391 
392     /* compute bytes to read -- error on overflow */
393     len = nitems * size;
394     if (size && len / size != nitems) {
395         gz_error(state, Z_STREAM_ERROR, "request does not fit in a size_t");
396         return 0;
397     }
398 
399     /* read len or fewer bytes to buf, return the number of full items read */
400     return len ? gz_read(state, buf, len) / size : 0;
401 }
402 
403 /* -- see zlib.h -- */
404 #ifdef Z_PREFIX_SET
405 #  undef z_gzgetc
406 #else
407 #  undef gzgetc
408 #endif
409 int ZEXPORT gzgetc(gzFile file) {
410     unsigned char buf[1];
411     gz_statep state;
412 
413     /* get internal structure */
414     if (file == NULL)
415         return -1;
416     state = (gz_statep)file;
417 
418     /* check that we're reading and that there's no (serious) error */
419     if (state->mode != GZ_READ ||
420         (state->err != Z_OK && state->err != Z_BUF_ERROR))
421         return -1;
422 
423     /* try output buffer (no need to check for skip request) */
424     if (state->x.have) {
425         state->x.have--;
426         state->x.pos++;
427         return *(state->x.next)++;
428     }
429 
430     /* nothing there -- try gz_read() */
431     return gz_read(state, buf, 1) < 1 ? -1 : buf[0];
432 }
433 
434 int ZEXPORT gzgetc_(gzFile file) {
435     return gzgetc(file);
436 }
437 
438 /* -- see zlib.h -- */
439 int ZEXPORT gzungetc(int c, gzFile file) {
440     gz_statep state;
441 
442     /* get internal structure */
443     if (file == NULL)
444         return -1;
445     state = (gz_statep)file;
446 
447     /* in case this was just opened, set up the input buffer */
448     if (state->mode == GZ_READ && state->how == LOOK && state->x.have == 0)
449         (void)gz_look(state);
450 
451     /* check that we're reading and that there's no (serious) error */
452     if (state->mode != GZ_READ ||
453         (state->err != Z_OK && state->err != Z_BUF_ERROR))
454         return -1;
455 
456     /* process a skip request */
457     if (state->seek) {
458         state->seek = 0;
459         if (gz_skip(state, state->skip) == -1)
460             return -1;
461     }
462 
463     /* can't push EOF */
464     if (c < 0)
465         return -1;
466 
467     /* if output buffer empty, put byte at end (allows more pushing) */
468     if (state->x.have == 0) {
469         state->x.have = 1;
470         state->x.next = state->out + (state->size << 1) - 1;
471         state->x.next[0] = (unsigned char)c;
472         state->x.pos--;
473         state->past = 0;
474         return c;
475     }
476 
477     /* if no room, give up (must have already done a gzungetc()) */
478     if (state->x.have == (state->size << 1)) {
479         gz_error(state, Z_DATA_ERROR, "out of room to push characters");
480         return -1;
481     }
482 
483     /* slide output data if needed and insert byte before existing data */
484     if (state->x.next == state->out) {
485         unsigned char *src = state->out + state->x.have;
486         unsigned char *dest = state->out + (state->size << 1);
487         while (src > state->out)
488             *--dest = *--src;
489         state->x.next = dest;
490     }
491     state->x.have++;
492     state->x.next--;
493     state->x.next[0] = (unsigned char)c;
494     state->x.pos--;
495     state->past = 0;
496     return c;
497 }
498 
499 /* -- see zlib.h -- */
500 char * ZEXPORT gzgets(gzFile file, char *buf, int len) {
501     unsigned left, n;
502     char *str;
503     unsigned char *eol;
504     gz_statep state;
505 
506     /* check parameters and get internal structure */
507     if (file == NULL || buf == NULL || len < 1)
508         return NULL;
509     state = (gz_statep)file;
510 
511     /* check that we're reading and that there's no (serious) error */
512     if (state->mode != GZ_READ ||
513         (state->err != Z_OK && state->err != Z_BUF_ERROR))
514         return NULL;
515 
516     /* process a skip request */
517     if (state->seek) {
518         state->seek = 0;
519         if (gz_skip(state, state->skip) == -1)
520             return NULL;
521     }
522 
523     /* copy output bytes up to new line or len - 1, whichever comes first --
524        append a terminating zero to the string (we don't check for a zero in
525        the contents, let the user worry about that) */
526     str = buf;
527     left = (unsigned)len - 1;
528     if (left) do {
529         /* assure that something is in the output buffer */
530         if (state->x.have == 0 && gz_fetch(state) == -1)
531             return NULL;                /* error */
532         if (state->x.have == 0) {       /* end of file */
533             state->past = 1;            /* read past end */
534             break;                      /* return what we have */
535         }
536 
537         /* look for end-of-line in current output buffer */
538         n = state->x.have > left ? left : state->x.have;
539         eol = (unsigned char *)memchr(state->x.next, '\n', n);
540         if (eol != NULL)
541             n = (unsigned)(eol - state->x.next) + 1;
542 
543         /* copy through end-of-line, or remainder if not found */
544         memcpy(buf, state->x.next, n);
545         state->x.have -= n;
546         state->x.next += n;
547         state->x.pos += n;
548         left -= n;
549         buf += n;
550     } while (left && eol == NULL);
551 
552     /* return terminated string, or if nothing, end of file */
553     if (buf == str)
554         return NULL;
555     buf[0] = 0;
556     return str;
557 }
558 
559 /* -- see zlib.h -- */
560 int ZEXPORT gzdirect(gzFile file) {
561     gz_statep state;
562 
563     /* get internal structure */
564     if (file == NULL)
565         return 0;
566     state = (gz_statep)file;
567 
568     /* if the state is not known, but we can find out, then do so (this is
569        mainly for right after a gzopen() or gzdopen()) */
570     if (state->mode == GZ_READ && state->how == LOOK && state->x.have == 0)
571         (void)gz_look(state);
572 
573     /* return 1 if transparent, 0 if processing a gzip stream */
574     return state->direct;
575 }
576 
577 /* -- see zlib.h -- */
578 int ZEXPORT gzclose_r(gzFile file) {
579     int ret, err;
580     gz_statep state;
581 
582     /* get internal structure */
583     if (file == NULL)
584         return Z_STREAM_ERROR;
585     state = (gz_statep)file;
586 
587     /* check that we're reading */
588     if (state->mode != GZ_READ)
589         return Z_STREAM_ERROR;
590 
591     /* free memory and close file */
592     if (state->size) {
593         inflateEnd(&(state->strm));
594         free(state->out);
595         free(state->in);
596     }
597     err = state->err == Z_BUF_ERROR ? Z_BUF_ERROR : Z_OK;
598     gz_error(state, Z_OK, NULL);
599     free(state->path);
600     ret = close(state->fd);
601     free(state);
602     return ret ? Z_ERRNO : err;
603 }
604