1 /***************************************************************************
2 * _ _ ____ _
3 * Project ___| | | | _ \| |
4 * / __| | | | |_) | |
5 * | (__| |_| | _ <| |___
6 * \___|\___/|_| \_\_____|
7 *
8 * Copyright (C) Daniel Stenberg, <daniel@haxx.se>, et al.
9 *
10 * This software is licensed as described in the file COPYING, which
11 * you should have received as part of this distribution. The terms
12 * are also available at https://curl.se/docs/copyright.html.
13 *
14 * You may opt to use, copy, modify, merge, publish, distribute and/or sell
15 * copies of the Software, and permit persons to whom the Software is
16 * furnished to do so, under the terms of the COPYING file.
17 *
18 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
19 * KIND, either express or implied.
20 *
21 * SPDX-License-Identifier: curl
22 *
23 ***************************************************************************/
24
25 #include "curl_setup.h"
26
27 #include "urldata.h"
28 #include <curl/curl.h>
29 #include <stddef.h>
30
31 #ifdef HAVE_LIBZ
32 #include <zlib.h>
33 #endif
34
35 #ifdef HAVE_BROTLI
36 #if defined(__GNUC__)
37 /* Ignore -Wvla warnings in brotli headers */
38 #pragma GCC diagnostic push
39 #pragma GCC diagnostic ignored "-Wvla"
40 #endif
41 #include <brotli/decode.h>
42 #if defined(__GNUC__)
43 #pragma GCC diagnostic pop
44 #endif
45 #endif
46
47 #ifdef HAVE_ZSTD
48 #include <zstd.h>
49 #endif
50
51 #include "sendf.h"
52 #include "http.h"
53 #include "content_encoding.h"
54 #include "strdup.h"
55 #include "strcase.h"
56 #include "curl_memory.h"
57 #include "memdebug.h"
58
59 #define CONTENT_ENCODING_DEFAULT "identity"
60
61 #ifndef CURL_DISABLE_HTTP
62
63 #define DSIZ CURL_MAX_WRITE_SIZE /* buffer size for decompressed data */
64
65
66 #ifdef HAVE_LIBZ
67
68 /* Comment this out if zlib is always going to be at least ver. 1.2.0.4
69 (doing so will reduce code size slightly). */
70 #define OLD_ZLIB_SUPPORT 1
71
72 #define GZIP_MAGIC_0 0x1f
73 #define GZIP_MAGIC_1 0x8b
74
75 /* gzip flag byte */
76 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
77 #define HEAD_CRC 0x02 /* bit 1 set: header CRC present */
78 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
79 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
80 #define COMMENT 0x10 /* bit 4 set: file comment present */
81 #define RESERVED 0xE0 /* bits 5..7: reserved */
82
83 typedef enum {
84 ZLIB_UNINIT, /* uninitialized */
85 ZLIB_INIT, /* initialized */
86 ZLIB_INFLATING, /* inflating started. */
87 ZLIB_EXTERNAL_TRAILER, /* reading external trailer */
88 ZLIB_GZIP_HEADER, /* reading gzip header */
89 ZLIB_GZIP_INFLATING, /* inflating gzip stream */
90 ZLIB_INIT_GZIP /* initialized in transparent gzip mode */
91 } zlibInitState;
92
93 /* Deflate and gzip writer. */
94 struct zlib_writer {
95 struct contenc_writer super;
96 zlibInitState zlib_init; /* zlib init state */
97 uInt trailerlen; /* Remaining trailer byte count. */
98 z_stream z; /* State structure for zlib. */
99 };
100
101
102 static voidpf
zalloc_cb(voidpf opaque,unsigned int items,unsigned int size)103 zalloc_cb(voidpf opaque, unsigned int items, unsigned int size)
104 {
105 (void) opaque;
106 /* not a typo, keep it calloc() */
107 return (voidpf) calloc(items, size);
108 }
109
110 static void
zfree_cb(voidpf opaque,voidpf ptr)111 zfree_cb(voidpf opaque, voidpf ptr)
112 {
113 (void) opaque;
114 free(ptr);
115 }
116
117 static CURLcode
process_zlib_error(struct Curl_easy * data,z_stream * z)118 process_zlib_error(struct Curl_easy *data, z_stream *z)
119 {
120 if(z->msg)
121 failf(data, "Error while processing content unencoding: %s",
122 z->msg);
123 else
124 failf(data, "Error while processing content unencoding: "
125 "Unknown failure within decompression software.");
126
127 return CURLE_BAD_CONTENT_ENCODING;
128 }
129
130 static CURLcode
exit_zlib(struct Curl_easy * data,z_stream * z,zlibInitState * zlib_init,CURLcode result)131 exit_zlib(struct Curl_easy *data,
132 z_stream *z, zlibInitState *zlib_init, CURLcode result)
133 {
134 if(*zlib_init == ZLIB_GZIP_HEADER)
135 Curl_safefree(z->next_in);
136
137 if(*zlib_init != ZLIB_UNINIT) {
138 if(inflateEnd(z) != Z_OK && result == CURLE_OK)
139 result = process_zlib_error(data, z);
140 *zlib_init = ZLIB_UNINIT;
141 }
142
143 return result;
144 }
145
process_trailer(struct Curl_easy * data,struct zlib_writer * zp)146 static CURLcode process_trailer(struct Curl_easy *data,
147 struct zlib_writer *zp)
148 {
149 z_stream *z = &zp->z;
150 CURLcode result = CURLE_OK;
151 uInt len = z->avail_in < zp->trailerlen? z->avail_in: zp->trailerlen;
152
153 /* Consume expected trailer bytes. Terminate stream if exhausted.
154 Issue an error if unexpected bytes follow. */
155
156 zp->trailerlen -= len;
157 z->avail_in -= len;
158 z->next_in += len;
159 if(z->avail_in)
160 result = CURLE_WRITE_ERROR;
161 if(result || !zp->trailerlen)
162 result = exit_zlib(data, z, &zp->zlib_init, result);
163 else {
164 /* Only occurs for gzip with zlib < 1.2.0.4 or raw deflate. */
165 zp->zlib_init = ZLIB_EXTERNAL_TRAILER;
166 }
167 return result;
168 }
169
inflate_stream(struct Curl_easy * data,struct contenc_writer * writer,zlibInitState started)170 static CURLcode inflate_stream(struct Curl_easy *data,
171 struct contenc_writer *writer,
172 zlibInitState started)
173 {
174 struct zlib_writer *zp = (struct zlib_writer *) writer;
175 z_stream *z = &zp->z; /* zlib state structure */
176 uInt nread = z->avail_in;
177 Bytef *orig_in = z->next_in;
178 bool done = FALSE;
179 CURLcode result = CURLE_OK; /* Curl_client_write status */
180 char *decomp; /* Put the decompressed data here. */
181
182 /* Check state. */
183 if(zp->zlib_init != ZLIB_INIT &&
184 zp->zlib_init != ZLIB_INFLATING &&
185 zp->zlib_init != ZLIB_INIT_GZIP &&
186 zp->zlib_init != ZLIB_GZIP_INFLATING)
187 return exit_zlib(data, z, &zp->zlib_init, CURLE_WRITE_ERROR);
188
189 /* Dynamically allocate a buffer for decompression because it's uncommonly
190 large to hold on the stack */
191 decomp = malloc(DSIZ);
192 if(!decomp)
193 return exit_zlib(data, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY);
194
195 /* because the buffer size is fixed, iteratively decompress and transfer to
196 the client via downstream_write function. */
197 while(!done) {
198 int status; /* zlib status */
199 done = TRUE;
200
201 /* (re)set buffer for decompressed output for every iteration */
202 z->next_out = (Bytef *) decomp;
203 z->avail_out = DSIZ;
204
205 #ifdef Z_BLOCK
206 /* Z_BLOCK is only available in zlib ver. >= 1.2.0.5 */
207 status = inflate(z, Z_BLOCK);
208 #else
209 /* fallback for zlib ver. < 1.2.0.5 */
210 status = inflate(z, Z_SYNC_FLUSH);
211 #endif
212
213 /* Flush output data if some. */
214 if(z->avail_out != DSIZ) {
215 if(status == Z_OK || status == Z_STREAM_END) {
216 zp->zlib_init = started; /* Data started. */
217 result = Curl_unencode_write(data, writer->downstream, decomp,
218 DSIZ - z->avail_out);
219 if(result) {
220 exit_zlib(data, z, &zp->zlib_init, result);
221 break;
222 }
223 }
224 }
225
226 /* Dispatch by inflate() status. */
227 switch(status) {
228 case Z_OK:
229 /* Always loop: there may be unflushed latched data in zlib state. */
230 done = FALSE;
231 break;
232 case Z_BUF_ERROR:
233 /* No more data to flush: just exit loop. */
234 break;
235 case Z_STREAM_END:
236 result = process_trailer(data, zp);
237 break;
238 case Z_DATA_ERROR:
239 /* some servers seem to not generate zlib headers, so this is an attempt
240 to fix and continue anyway */
241 if(zp->zlib_init == ZLIB_INIT) {
242 /* Do not use inflateReset2(): only available since zlib 1.2.3.4. */
243 (void) inflateEnd(z); /* don't care about the return code */
244 if(inflateInit2(z, -MAX_WBITS) == Z_OK) {
245 z->next_in = orig_in;
246 z->avail_in = nread;
247 zp->zlib_init = ZLIB_INFLATING;
248 zp->trailerlen = 4; /* Tolerate up to 4 unknown trailer bytes. */
249 done = FALSE;
250 break;
251 }
252 zp->zlib_init = ZLIB_UNINIT; /* inflateEnd() already called. */
253 }
254 result = exit_zlib(data, z, &zp->zlib_init, process_zlib_error(data, z));
255 break;
256 default:
257 result = exit_zlib(data, z, &zp->zlib_init, process_zlib_error(data, z));
258 break;
259 }
260 }
261 free(decomp);
262
263 /* We're about to leave this call so the `nread' data bytes won't be seen
264 again. If we are in a state that would wrongly allow restart in raw mode
265 at the next call, assume output has already started. */
266 if(nread && zp->zlib_init == ZLIB_INIT)
267 zp->zlib_init = started; /* Cannot restart anymore. */
268
269 return result;
270 }
271
272
273 /* Deflate handler. */
deflate_init_writer(struct Curl_easy * data,struct contenc_writer * writer)274 static CURLcode deflate_init_writer(struct Curl_easy *data,
275 struct contenc_writer *writer)
276 {
277 struct zlib_writer *zp = (struct zlib_writer *) writer;
278 z_stream *z = &zp->z; /* zlib state structure */
279
280 if(!writer->downstream)
281 return CURLE_WRITE_ERROR;
282
283 /* Initialize zlib */
284 z->zalloc = (alloc_func) zalloc_cb;
285 z->zfree = (free_func) zfree_cb;
286
287 if(inflateInit(z) != Z_OK)
288 return process_zlib_error(data, z);
289 zp->zlib_init = ZLIB_INIT;
290 return CURLE_OK;
291 }
292
deflate_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)293 static CURLcode deflate_unencode_write(struct Curl_easy *data,
294 struct contenc_writer *writer,
295 const char *buf, size_t nbytes)
296 {
297 struct zlib_writer *zp = (struct zlib_writer *) writer;
298 z_stream *z = &zp->z; /* zlib state structure */
299
300 /* Set the compressed input when this function is called */
301 z->next_in = (Bytef *) buf;
302 z->avail_in = (uInt) nbytes;
303
304 if(zp->zlib_init == ZLIB_EXTERNAL_TRAILER)
305 return process_trailer(data, zp);
306
307 /* Now uncompress the data */
308 return inflate_stream(data, writer, ZLIB_INFLATING);
309 }
310
deflate_close_writer(struct Curl_easy * data,struct contenc_writer * writer)311 static void deflate_close_writer(struct Curl_easy *data,
312 struct contenc_writer *writer)
313 {
314 struct zlib_writer *zp = (struct zlib_writer *) writer;
315 z_stream *z = &zp->z; /* zlib state structure */
316
317 exit_zlib(data, z, &zp->zlib_init, CURLE_OK);
318 }
319
320 static const struct content_encoding deflate_encoding = {
321 "deflate",
322 NULL,
323 deflate_init_writer,
324 deflate_unencode_write,
325 deflate_close_writer,
326 sizeof(struct zlib_writer)
327 };
328
329
330 /* Gzip handler. */
gzip_init_writer(struct Curl_easy * data,struct contenc_writer * writer)331 static CURLcode gzip_init_writer(struct Curl_easy *data,
332 struct contenc_writer *writer)
333 {
334 struct zlib_writer *zp = (struct zlib_writer *) writer;
335 z_stream *z = &zp->z; /* zlib state structure */
336
337 if(!writer->downstream)
338 return CURLE_WRITE_ERROR;
339
340 /* Initialize zlib */
341 z->zalloc = (alloc_func) zalloc_cb;
342 z->zfree = (free_func) zfree_cb;
343
344 if(strcmp(zlibVersion(), "1.2.0.4") >= 0) {
345 /* zlib ver. >= 1.2.0.4 supports transparent gzip decompressing */
346 if(inflateInit2(z, MAX_WBITS + 32) != Z_OK) {
347 return process_zlib_error(data, z);
348 }
349 zp->zlib_init = ZLIB_INIT_GZIP; /* Transparent gzip decompress state */
350 }
351 else {
352 /* we must parse the gzip header and trailer ourselves */
353 if(inflateInit2(z, -MAX_WBITS) != Z_OK) {
354 return process_zlib_error(data, z);
355 }
356 zp->trailerlen = 8; /* A CRC-32 and a 32-bit input size (RFC 1952, 2.2) */
357 zp->zlib_init = ZLIB_INIT; /* Initial call state */
358 }
359
360 return CURLE_OK;
361 }
362
363 #ifdef OLD_ZLIB_SUPPORT
364 /* Skip over the gzip header */
365 static enum {
366 GZIP_OK,
367 GZIP_BAD,
368 GZIP_UNDERFLOW
check_gzip_header(unsigned char const * data,ssize_t len,ssize_t * headerlen)369 } check_gzip_header(unsigned char const *data, ssize_t len, ssize_t *headerlen)
370 {
371 int method, flags;
372 const ssize_t totallen = len;
373
374 /* The shortest header is 10 bytes */
375 if(len < 10)
376 return GZIP_UNDERFLOW;
377
378 if((data[0] != GZIP_MAGIC_0) || (data[1] != GZIP_MAGIC_1))
379 return GZIP_BAD;
380
381 method = data[2];
382 flags = data[3];
383
384 if(method != Z_DEFLATED || (flags & RESERVED) != 0) {
385 /* Can't handle this compression method or unknown flag */
386 return GZIP_BAD;
387 }
388
389 /* Skip over time, xflags, OS code and all previous bytes */
390 len -= 10;
391 data += 10;
392
393 if(flags & EXTRA_FIELD) {
394 ssize_t extra_len;
395
396 if(len < 2)
397 return GZIP_UNDERFLOW;
398
399 extra_len = (data[1] << 8) | data[0];
400
401 if(len < (extra_len + 2))
402 return GZIP_UNDERFLOW;
403
404 len -= (extra_len + 2);
405 data += (extra_len + 2);
406 }
407
408 if(flags & ORIG_NAME) {
409 /* Skip over NUL-terminated file name */
410 while(len && *data) {
411 --len;
412 ++data;
413 }
414 if(!len || *data)
415 return GZIP_UNDERFLOW;
416
417 /* Skip over the NUL */
418 --len;
419 ++data;
420 }
421
422 if(flags & COMMENT) {
423 /* Skip over NUL-terminated comment */
424 while(len && *data) {
425 --len;
426 ++data;
427 }
428 if(!len || *data)
429 return GZIP_UNDERFLOW;
430
431 /* Skip over the NUL */
432 --len;
433 }
434
435 if(flags & HEAD_CRC) {
436 if(len < 2)
437 return GZIP_UNDERFLOW;
438
439 len -= 2;
440 }
441
442 *headerlen = totallen - len;
443 return GZIP_OK;
444 }
445 #endif
446
gzip_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)447 static CURLcode gzip_unencode_write(struct Curl_easy *data,
448 struct contenc_writer *writer,
449 const char *buf, size_t nbytes)
450 {
451 struct zlib_writer *zp = (struct zlib_writer *) writer;
452 z_stream *z = &zp->z; /* zlib state structure */
453
454 if(zp->zlib_init == ZLIB_INIT_GZIP) {
455 /* Let zlib handle the gzip decompression entirely */
456 z->next_in = (Bytef *) buf;
457 z->avail_in = (uInt) nbytes;
458 /* Now uncompress the data */
459 return inflate_stream(data, writer, ZLIB_INIT_GZIP);
460 }
461
462 #ifndef OLD_ZLIB_SUPPORT
463 /* Support for old zlib versions is compiled away and we are running with
464 an old version, so return an error. */
465 return exit_zlib(data, z, &zp->zlib_init, CURLE_WRITE_ERROR);
466
467 #else
468 /* This next mess is to get around the potential case where there isn't
469 * enough data passed in to skip over the gzip header. If that happens, we
470 * malloc a block and copy what we have then wait for the next call. If
471 * there still isn't enough (this is definitely a worst-case scenario), we
472 * make the block bigger, copy the next part in and keep waiting.
473 *
474 * This is only required with zlib versions < 1.2.0.4 as newer versions
475 * can handle the gzip header themselves.
476 */
477
478 switch(zp->zlib_init) {
479 /* Skip over gzip header? */
480 case ZLIB_INIT:
481 {
482 /* Initial call state */
483 ssize_t hlen;
484
485 switch(check_gzip_header((unsigned char *) buf, nbytes, &hlen)) {
486 case GZIP_OK:
487 z->next_in = (Bytef *) buf + hlen;
488 z->avail_in = (uInt) (nbytes - hlen);
489 zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */
490 break;
491
492 case GZIP_UNDERFLOW:
493 /* We need more data so we can find the end of the gzip header. It's
494 * possible that the memory block we malloc here will never be freed if
495 * the transfer abruptly aborts after this point. Since it's unlikely
496 * that circumstances will be right for this code path to be followed in
497 * the first place, and it's even more unlikely for a transfer to fail
498 * immediately afterwards, it should seldom be a problem.
499 */
500 z->avail_in = (uInt) nbytes;
501 z->next_in = malloc(z->avail_in);
502 if(!z->next_in) {
503 return exit_zlib(data, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY);
504 }
505 memcpy(z->next_in, buf, z->avail_in);
506 zp->zlib_init = ZLIB_GZIP_HEADER; /* Need more gzip header data state */
507 /* We don't have any data to inflate yet */
508 return CURLE_OK;
509
510 case GZIP_BAD:
511 default:
512 return exit_zlib(data, z, &zp->zlib_init, process_zlib_error(data, z));
513 }
514
515 }
516 break;
517
518 case ZLIB_GZIP_HEADER:
519 {
520 /* Need more gzip header data state */
521 ssize_t hlen;
522 z->avail_in += (uInt) nbytes;
523 z->next_in = Curl_saferealloc(z->next_in, z->avail_in);
524 if(!z->next_in) {
525 return exit_zlib(data, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY);
526 }
527 /* Append the new block of data to the previous one */
528 memcpy(z->next_in + z->avail_in - nbytes, buf, nbytes);
529
530 switch(check_gzip_header(z->next_in, z->avail_in, &hlen)) {
531 case GZIP_OK:
532 /* This is the zlib stream data */
533 free(z->next_in);
534 /* Don't point into the malloced block since we just freed it */
535 z->next_in = (Bytef *) buf + hlen + nbytes - z->avail_in;
536 z->avail_in = (uInt) (z->avail_in - hlen);
537 zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */
538 break;
539
540 case GZIP_UNDERFLOW:
541 /* We still don't have any data to inflate! */
542 return CURLE_OK;
543
544 case GZIP_BAD:
545 default:
546 return exit_zlib(data, z, &zp->zlib_init, process_zlib_error(data, z));
547 }
548
549 }
550 break;
551
552 case ZLIB_EXTERNAL_TRAILER:
553 z->next_in = (Bytef *) buf;
554 z->avail_in = (uInt) nbytes;
555 return process_trailer(data, zp);
556
557 case ZLIB_GZIP_INFLATING:
558 default:
559 /* Inflating stream state */
560 z->next_in = (Bytef *) buf;
561 z->avail_in = (uInt) nbytes;
562 break;
563 }
564
565 if(z->avail_in == 0) {
566 /* We don't have any data to inflate; wait until next time */
567 return CURLE_OK;
568 }
569
570 /* We've parsed the header, now uncompress the data */
571 return inflate_stream(data, writer, ZLIB_GZIP_INFLATING);
572 #endif
573 }
574
gzip_close_writer(struct Curl_easy * data,struct contenc_writer * writer)575 static void gzip_close_writer(struct Curl_easy *data,
576 struct contenc_writer *writer)
577 {
578 struct zlib_writer *zp = (struct zlib_writer *) writer;
579 z_stream *z = &zp->z; /* zlib state structure */
580
581 exit_zlib(data, z, &zp->zlib_init, CURLE_OK);
582 }
583
584 static const struct content_encoding gzip_encoding = {
585 "gzip",
586 "x-gzip",
587 gzip_init_writer,
588 gzip_unencode_write,
589 gzip_close_writer,
590 sizeof(struct zlib_writer)
591 };
592
593 #endif /* HAVE_LIBZ */
594
595
596 #ifdef HAVE_BROTLI
597 /* Brotli writer. */
598 struct brotli_writer {
599 struct contenc_writer super;
600 BrotliDecoderState *br; /* State structure for brotli. */
601 };
602
brotli_map_error(BrotliDecoderErrorCode be)603 static CURLcode brotli_map_error(BrotliDecoderErrorCode be)
604 {
605 switch(be) {
606 case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_NIBBLE:
607 case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_META_NIBBLE:
608 case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_ALPHABET:
609 case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_SAME:
610 case BROTLI_DECODER_ERROR_FORMAT_CL_SPACE:
611 case BROTLI_DECODER_ERROR_FORMAT_HUFFMAN_SPACE:
612 case BROTLI_DECODER_ERROR_FORMAT_CONTEXT_MAP_REPEAT:
613 case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_1:
614 case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_2:
615 case BROTLI_DECODER_ERROR_FORMAT_TRANSFORM:
616 case BROTLI_DECODER_ERROR_FORMAT_DICTIONARY:
617 case BROTLI_DECODER_ERROR_FORMAT_WINDOW_BITS:
618 case BROTLI_DECODER_ERROR_FORMAT_PADDING_1:
619 case BROTLI_DECODER_ERROR_FORMAT_PADDING_2:
620 #ifdef BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY
621 case BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY:
622 #endif
623 #ifdef BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET
624 case BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET:
625 #endif
626 case BROTLI_DECODER_ERROR_INVALID_ARGUMENTS:
627 return CURLE_BAD_CONTENT_ENCODING;
628 case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MODES:
629 case BROTLI_DECODER_ERROR_ALLOC_TREE_GROUPS:
630 case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MAP:
631 case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_1:
632 case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_2:
633 case BROTLI_DECODER_ERROR_ALLOC_BLOCK_TYPE_TREES:
634 return CURLE_OUT_OF_MEMORY;
635 default:
636 break;
637 }
638 return CURLE_WRITE_ERROR;
639 }
640
brotli_init_writer(struct Curl_easy * data,struct contenc_writer * writer)641 static CURLcode brotli_init_writer(struct Curl_easy *data,
642 struct contenc_writer *writer)
643 {
644 struct brotli_writer *bp = (struct brotli_writer *) writer;
645 (void) data;
646
647 if(!writer->downstream)
648 return CURLE_WRITE_ERROR;
649
650 bp->br = BrotliDecoderCreateInstance(NULL, NULL, NULL);
651 return bp->br? CURLE_OK: CURLE_OUT_OF_MEMORY;
652 }
653
brotli_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)654 static CURLcode brotli_unencode_write(struct Curl_easy *data,
655 struct contenc_writer *writer,
656 const char *buf, size_t nbytes)
657 {
658 struct brotli_writer *bp = (struct brotli_writer *) writer;
659 const uint8_t *src = (const uint8_t *) buf;
660 char *decomp;
661 uint8_t *dst;
662 size_t dstleft;
663 CURLcode result = CURLE_OK;
664 BrotliDecoderResult r = BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT;
665
666 if(!bp->br)
667 return CURLE_WRITE_ERROR; /* Stream already ended. */
668
669 decomp = malloc(DSIZ);
670 if(!decomp)
671 return CURLE_OUT_OF_MEMORY;
672
673 while((nbytes || r == BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT) &&
674 result == CURLE_OK) {
675 dst = (uint8_t *) decomp;
676 dstleft = DSIZ;
677 r = BrotliDecoderDecompressStream(bp->br,
678 &nbytes, &src, &dstleft, &dst, NULL);
679 result = Curl_unencode_write(data, writer->downstream,
680 decomp, DSIZ - dstleft);
681 if(result)
682 break;
683 switch(r) {
684 case BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT:
685 case BROTLI_DECODER_RESULT_NEEDS_MORE_INPUT:
686 break;
687 case BROTLI_DECODER_RESULT_SUCCESS:
688 BrotliDecoderDestroyInstance(bp->br);
689 bp->br = NULL;
690 if(nbytes)
691 result = CURLE_WRITE_ERROR;
692 break;
693 default:
694 result = brotli_map_error(BrotliDecoderGetErrorCode(bp->br));
695 break;
696 }
697 }
698 free(decomp);
699 return result;
700 }
701
brotli_close_writer(struct Curl_easy * data,struct contenc_writer * writer)702 static void brotli_close_writer(struct Curl_easy *data,
703 struct contenc_writer *writer)
704 {
705 struct brotli_writer *bp = (struct brotli_writer *) writer;
706
707 (void) data;
708
709 if(bp->br) {
710 BrotliDecoderDestroyInstance(bp->br);
711 bp->br = NULL;
712 }
713 }
714
715 static const struct content_encoding brotli_encoding = {
716 "br",
717 NULL,
718 brotli_init_writer,
719 brotli_unencode_write,
720 brotli_close_writer,
721 sizeof(struct brotli_writer)
722 };
723 #endif
724
725
726 #ifdef HAVE_ZSTD
727 /* Zstd writer. */
728 struct zstd_writer {
729 struct contenc_writer super;
730 ZSTD_DStream *zds; /* State structure for zstd. */
731 void *decomp;
732 };
733
zstd_init_writer(struct Curl_easy * data,struct contenc_writer * writer)734 static CURLcode zstd_init_writer(struct Curl_easy *data,
735 struct contenc_writer *writer)
736 {
737 struct zstd_writer *zp = (struct zstd_writer *) writer;
738
739 (void)data;
740
741 if(!writer->downstream)
742 return CURLE_WRITE_ERROR;
743
744 zp->zds = ZSTD_createDStream();
745 zp->decomp = NULL;
746 return zp->zds ? CURLE_OK : CURLE_OUT_OF_MEMORY;
747 }
748
zstd_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)749 static CURLcode zstd_unencode_write(struct Curl_easy *data,
750 struct contenc_writer *writer,
751 const char *buf, size_t nbytes)
752 {
753 CURLcode result = CURLE_OK;
754 struct zstd_writer *zp = (struct zstd_writer *) writer;
755 ZSTD_inBuffer in;
756 ZSTD_outBuffer out;
757 size_t errorCode;
758
759 if(!zp->decomp) {
760 zp->decomp = malloc(DSIZ);
761 if(!zp->decomp)
762 return CURLE_OUT_OF_MEMORY;
763 }
764 in.pos = 0;
765 in.src = buf;
766 in.size = nbytes;
767
768 for(;;) {
769 out.pos = 0;
770 out.dst = zp->decomp;
771 out.size = DSIZ;
772
773 errorCode = ZSTD_decompressStream(zp->zds, &out, &in);
774 if(ZSTD_isError(errorCode)) {
775 return CURLE_BAD_CONTENT_ENCODING;
776 }
777 if(out.pos > 0) {
778 result = Curl_unencode_write(data, writer->downstream,
779 zp->decomp, out.pos);
780 if(result)
781 break;
782 }
783 if((in.pos == nbytes) && (out.pos < out.size))
784 break;
785 }
786
787 return result;
788 }
789
zstd_close_writer(struct Curl_easy * data,struct contenc_writer * writer)790 static void zstd_close_writer(struct Curl_easy *data,
791 struct contenc_writer *writer)
792 {
793 struct zstd_writer *zp = (struct zstd_writer *) writer;
794
795 (void)data;
796
797 if(zp->decomp) {
798 free(zp->decomp);
799 zp->decomp = NULL;
800 }
801 if(zp->zds) {
802 ZSTD_freeDStream(zp->zds);
803 zp->zds = NULL;
804 }
805 }
806
807 static const struct content_encoding zstd_encoding = {
808 "zstd",
809 NULL,
810 zstd_init_writer,
811 zstd_unencode_write,
812 zstd_close_writer,
813 sizeof(struct zstd_writer)
814 };
815 #endif
816
817
818 /* Identity handler. */
identity_init_writer(struct Curl_easy * data,struct contenc_writer * writer)819 static CURLcode identity_init_writer(struct Curl_easy *data,
820 struct contenc_writer *writer)
821 {
822 (void) data;
823 return writer->downstream? CURLE_OK: CURLE_WRITE_ERROR;
824 }
825
identity_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)826 static CURLcode identity_unencode_write(struct Curl_easy *data,
827 struct contenc_writer *writer,
828 const char *buf, size_t nbytes)
829 {
830 return Curl_unencode_write(data, writer->downstream, buf, nbytes);
831 }
832
identity_close_writer(struct Curl_easy * data,struct contenc_writer * writer)833 static void identity_close_writer(struct Curl_easy *data,
834 struct contenc_writer *writer)
835 {
836 (void) data;
837 (void) writer;
838 }
839
840 static const struct content_encoding identity_encoding = {
841 "identity",
842 "none",
843 identity_init_writer,
844 identity_unencode_write,
845 identity_close_writer,
846 sizeof(struct contenc_writer)
847 };
848
849
850 /* supported content encodings table. */
851 static const struct content_encoding * const encodings[] = {
852 &identity_encoding,
853 #ifdef HAVE_LIBZ
854 &deflate_encoding,
855 &gzip_encoding,
856 #endif
857 #ifdef HAVE_BROTLI
858 &brotli_encoding,
859 #endif
860 #ifdef HAVE_ZSTD
861 &zstd_encoding,
862 #endif
863 NULL
864 };
865
866
867 /* Return a list of comma-separated names of supported encodings. */
Curl_all_content_encodings(void)868 char *Curl_all_content_encodings(void)
869 {
870 size_t len = 0;
871 const struct content_encoding * const *cep;
872 const struct content_encoding *ce;
873 char *ace;
874
875 for(cep = encodings; *cep; cep++) {
876 ce = *cep;
877 if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT))
878 len += strlen(ce->name) + 2;
879 }
880
881 if(!len)
882 return strdup(CONTENT_ENCODING_DEFAULT);
883
884 ace = malloc(len);
885 if(ace) {
886 char *p = ace;
887 for(cep = encodings; *cep; cep++) {
888 ce = *cep;
889 if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT)) {
890 strcpy(p, ce->name);
891 p += strlen(p);
892 *p++ = ',';
893 *p++ = ' ';
894 }
895 }
896 p[-2] = '\0';
897 }
898
899 return ace;
900 }
901
902
903 /* Real client writer: no downstream. */
client_init_writer(struct Curl_easy * data,struct contenc_writer * writer)904 static CURLcode client_init_writer(struct Curl_easy *data,
905 struct contenc_writer *writer)
906 {
907 (void) data;
908 return writer->downstream? CURLE_WRITE_ERROR: CURLE_OK;
909 }
910
client_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)911 static CURLcode client_unencode_write(struct Curl_easy *data,
912 struct contenc_writer *writer,
913 const char *buf, size_t nbytes)
914 {
915 struct SingleRequest *k = &data->req;
916
917 (void) writer;
918
919 if(!nbytes || k->ignorebody)
920 return CURLE_OK;
921
922 return Curl_client_write(data, CLIENTWRITE_BODY, (char *) buf, nbytes);
923 }
924
client_close_writer(struct Curl_easy * data,struct contenc_writer * writer)925 static void client_close_writer(struct Curl_easy *data,
926 struct contenc_writer *writer)
927 {
928 (void) data;
929 (void) writer;
930 }
931
932 static const struct content_encoding client_encoding = {
933 NULL,
934 NULL,
935 client_init_writer,
936 client_unencode_write,
937 client_close_writer,
938 sizeof(struct contenc_writer)
939 };
940
941
942 /* Deferred error dummy writer. */
error_init_writer(struct Curl_easy * data,struct contenc_writer * writer)943 static CURLcode error_init_writer(struct Curl_easy *data,
944 struct contenc_writer *writer)
945 {
946 (void) data;
947 return writer->downstream? CURLE_OK: CURLE_WRITE_ERROR;
948 }
949
error_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)950 static CURLcode error_unencode_write(struct Curl_easy *data,
951 struct contenc_writer *writer,
952 const char *buf, size_t nbytes)
953 {
954 char *all = Curl_all_content_encodings();
955
956 (void) writer;
957 (void) buf;
958 (void) nbytes;
959
960 if(!all)
961 return CURLE_OUT_OF_MEMORY;
962 failf(data, "Unrecognized content encoding type. "
963 "libcurl understands %s content encodings.", all);
964 free(all);
965 return CURLE_BAD_CONTENT_ENCODING;
966 }
967
error_close_writer(struct Curl_easy * data,struct contenc_writer * writer)968 static void error_close_writer(struct Curl_easy *data,
969 struct contenc_writer *writer)
970 {
971 (void) data;
972 (void) writer;
973 }
974
975 static const struct content_encoding error_encoding = {
976 NULL,
977 NULL,
978 error_init_writer,
979 error_unencode_write,
980 error_close_writer,
981 sizeof(struct contenc_writer)
982 };
983
984 /* Create an unencoding writer stage using the given handler. */
985 static struct contenc_writer *
new_unencoding_writer(struct Curl_easy * data,const struct content_encoding * handler,struct contenc_writer * downstream,int order)986 new_unencoding_writer(struct Curl_easy *data,
987 const struct content_encoding *handler,
988 struct contenc_writer *downstream,
989 int order)
990 {
991 struct contenc_writer *writer;
992
993 DEBUGASSERT(handler->writersize >= sizeof(struct contenc_writer));
994 writer = (struct contenc_writer *) calloc(1, handler->writersize);
995
996 if(writer) {
997 writer->handler = handler;
998 writer->downstream = downstream;
999 writer->order = order;
1000 if(handler->init_writer(data, writer)) {
1001 free(writer);
1002 writer = NULL;
1003 }
1004 }
1005
1006 return writer;
1007 }
1008
1009 /* Write data using an unencoding writer stack. "nbytes" is not
1010 allowed to be 0. */
Curl_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)1011 CURLcode Curl_unencode_write(struct Curl_easy *data,
1012 struct contenc_writer *writer,
1013 const char *buf, size_t nbytes)
1014 {
1015 if(!nbytes)
1016 return CURLE_OK;
1017 return writer->handler->unencode_write(data, writer, buf, nbytes);
1018 }
1019
1020 /* Close and clean-up the connection's writer stack. */
Curl_unencode_cleanup(struct Curl_easy * data)1021 void Curl_unencode_cleanup(struct Curl_easy *data)
1022 {
1023 struct SingleRequest *k = &data->req;
1024 struct contenc_writer *writer = k->writer_stack;
1025
1026 while(writer) {
1027 k->writer_stack = writer->downstream;
1028 writer->handler->close_writer(data, writer);
1029 free(writer);
1030 writer = k->writer_stack;
1031 }
1032 }
1033
1034 /* Find the content encoding by name. */
find_encoding(const char * name,size_t len)1035 static const struct content_encoding *find_encoding(const char *name,
1036 size_t len)
1037 {
1038 const struct content_encoding * const *cep;
1039
1040 for(cep = encodings; *cep; cep++) {
1041 const struct content_encoding *ce = *cep;
1042 if((strncasecompare(name, ce->name, len) && !ce->name[len]) ||
1043 (ce->alias && strncasecompare(name, ce->alias, len) && !ce->alias[len]))
1044 return ce;
1045 }
1046 return NULL;
1047 }
1048
1049 /* allow no more than 5 "chained" compression steps */
1050 #define MAX_ENCODE_STACK 5
1051
1052 /* Set-up the unencoding stack from the Content-Encoding header value.
1053 * See RFC 7231 section 3.1.2.2. */
Curl_build_unencoding_stack(struct Curl_easy * data,const char * enclist,int is_transfer)1054 CURLcode Curl_build_unencoding_stack(struct Curl_easy *data,
1055 const char *enclist, int is_transfer)
1056 {
1057 struct SingleRequest *k = &data->req;
1058 unsigned int order = is_transfer? 2: 1;
1059
1060 do {
1061 const char *name;
1062 size_t namelen;
1063
1064 /* Parse a single encoding name. */
1065 while(ISBLANK(*enclist) || *enclist == ',')
1066 enclist++;
1067
1068 name = enclist;
1069
1070 for(namelen = 0; *enclist && *enclist != ','; enclist++)
1071 if(!ISSPACE(*enclist))
1072 namelen = enclist - name + 1;
1073
1074 /* Special case: chunked encoding is handled at the reader level. */
1075 if(is_transfer && namelen == 7 && strncasecompare(name, "chunked", 7)) {
1076 k->chunk = TRUE; /* chunks coming our way. */
1077 Curl_httpchunk_init(data); /* init our chunky engine. */
1078 }
1079 else if(namelen) {
1080 const struct content_encoding *encoding = find_encoding(name, namelen);
1081 struct contenc_writer *writer;
1082
1083 if(!k->writer_stack) {
1084 k->writer_stack = new_unencoding_writer(data, &client_encoding,
1085 NULL, 0);
1086
1087 if(!k->writer_stack)
1088 return CURLE_OUT_OF_MEMORY;
1089 }
1090
1091 if(!encoding)
1092 encoding = &error_encoding; /* Defer error at stack use. */
1093
1094 if(k->writer_stack_depth++ >= MAX_ENCODE_STACK) {
1095 failf(data, "Reject response due to more than %u content encodings",
1096 MAX_ENCODE_STACK);
1097 return CURLE_BAD_CONTENT_ENCODING;
1098 }
1099 /* Stack the unencoding stage. */
1100 if(order >= k->writer_stack->order) {
1101 writer = new_unencoding_writer(data, encoding,
1102 k->writer_stack, order);
1103 if(!writer)
1104 return CURLE_OUT_OF_MEMORY;
1105 k->writer_stack = writer;
1106 }
1107 else {
1108 struct contenc_writer *w = k->writer_stack;
1109 while(w->downstream && order < w->downstream->order)
1110 w = w->downstream;
1111 writer = new_unencoding_writer(data, encoding,
1112 w->downstream, order);
1113 if(!writer)
1114 return CURLE_OUT_OF_MEMORY;
1115 w->downstream = writer;
1116 }
1117 }
1118 } while(*enclist);
1119
1120 return CURLE_OK;
1121 }
1122
1123 #else
1124 /* Stubs for builds without HTTP. */
Curl_build_unencoding_stack(struct Curl_easy * data,const char * enclist,int is_transfer)1125 CURLcode Curl_build_unencoding_stack(struct Curl_easy *data,
1126 const char *enclist, int is_transfer)
1127 {
1128 (void) data;
1129 (void) enclist;
1130 (void) is_transfer;
1131 return CURLE_NOT_BUILT_IN;
1132 }
1133
Curl_unencode_write(struct Curl_easy * data,struct contenc_writer * writer,const char * buf,size_t nbytes)1134 CURLcode Curl_unencode_write(struct Curl_easy *data,
1135 struct contenc_writer *writer,
1136 const char *buf, size_t nbytes)
1137 {
1138 (void) data;
1139 (void) writer;
1140 (void) buf;
1141 (void) nbytes;
1142 return CURLE_NOT_BUILT_IN;
1143 }
1144
Curl_unencode_cleanup(struct Curl_easy * data)1145 void Curl_unencode_cleanup(struct Curl_easy *data)
1146 {
1147 (void) data;
1148 }
1149
Curl_all_content_encodings(void)1150 char *Curl_all_content_encodings(void)
1151 {
1152 return strdup(CONTENT_ENCODING_DEFAULT); /* Satisfy caller. */
1153 }
1154
1155 #endif /* CURL_DISABLE_HTTP */
1156