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 | |
57 | /* The last 3 #include files should be in this order */ |
58 | #include "curl_printf.h" |
59 | #include "curl_memory.h" |
60 | #include "memdebug.h" |
61 | |
62 | #define CONTENT_ENCODING_DEFAULT "identity" |
63 | |
64 | #ifndef CURL_DISABLE_HTTP |
65 | |
66 | #define DSIZ CURL_MAX_WRITE_SIZE /* buffer size for decompressed data */ |
67 | |
68 | |
69 | #ifdef HAVE_LIBZ |
70 | |
71 | /* Comment this out if zlib is always going to be at least ver. 1.2.0.4 |
72 | (doing so will reduce code size slightly). */ |
73 | #define OLD_ZLIB_SUPPORT 1 |
74 | |
75 | #define GZIP_MAGIC_0 0x1f |
76 | #define GZIP_MAGIC_1 0x8b |
77 | |
78 | /* gzip flag byte */ |
79 | #define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */ |
80 | #define HEAD_CRC 0x02 /* bit 1 set: header CRC present */ |
81 | #define 0x04 /* bit 2 set: extra field present */ |
82 | #define ORIG_NAME 0x08 /* bit 3 set: original file name present */ |
83 | #define 0x10 /* bit 4 set: file comment present */ |
84 | #define RESERVED 0xE0 /* bits 5..7: reserved */ |
85 | |
86 | typedef enum { |
87 | ZLIB_UNINIT, /* uninitialized */ |
88 | ZLIB_INIT, /* initialized */ |
89 | ZLIB_INFLATING, /* inflating started. */ |
90 | ZLIB_EXTERNAL_TRAILER, /* reading external trailer */ |
91 | , /* reading gzip header */ |
92 | ZLIB_GZIP_INFLATING, /* inflating gzip stream */ |
93 | ZLIB_INIT_GZIP /* initialized in transparent gzip mode */ |
94 | } zlibInitState; |
95 | |
96 | /* Deflate and gzip writer. */ |
97 | struct zlib_writer { |
98 | struct contenc_writer super; |
99 | zlibInitState zlib_init; /* zlib init state */ |
100 | uInt trailerlen; /* Remaining trailer byte count. */ |
101 | z_stream z; /* State structure for zlib. */ |
102 | }; |
103 | |
104 | |
105 | static voidpf |
106 | zalloc_cb(voidpf opaque, unsigned int items, unsigned int size) |
107 | { |
108 | (void) opaque; |
109 | /* not a typo, keep it calloc() */ |
110 | return (voidpf) calloc(items, size); |
111 | } |
112 | |
113 | static void |
114 | zfree_cb(voidpf opaque, voidpf ptr) |
115 | { |
116 | (void) opaque; |
117 | free(ptr); |
118 | } |
119 | |
120 | static CURLcode |
121 | process_zlib_error(struct Curl_easy *data, z_stream *z) |
122 | { |
123 | if(z->msg) |
124 | failf(data, fmt: "Error while processing content unencoding: %s" , |
125 | z->msg); |
126 | else |
127 | failf(data, fmt: "Error while processing content unencoding: " |
128 | "Unknown failure within decompression software." ); |
129 | |
130 | return CURLE_BAD_CONTENT_ENCODING; |
131 | } |
132 | |
133 | static CURLcode |
134 | exit_zlib(struct Curl_easy *data, |
135 | z_stream *z, zlibInitState *zlib_init, CURLcode result) |
136 | { |
137 | if(*zlib_init == ZLIB_GZIP_HEADER) |
138 | Curl_safefree(z->next_in); |
139 | |
140 | if(*zlib_init != ZLIB_UNINIT) { |
141 | if(inflateEnd(strm: z) != Z_OK && result == CURLE_OK) |
142 | result = process_zlib_error(data, z); |
143 | *zlib_init = ZLIB_UNINIT; |
144 | } |
145 | |
146 | return result; |
147 | } |
148 | |
149 | static CURLcode process_trailer(struct Curl_easy *data, |
150 | struct zlib_writer *zp) |
151 | { |
152 | z_stream *z = &zp->z; |
153 | CURLcode result = CURLE_OK; |
154 | uInt len = z->avail_in < zp->trailerlen? z->avail_in: zp->trailerlen; |
155 | |
156 | /* Consume expected trailer bytes. Terminate stream if exhausted. |
157 | Issue an error if unexpected bytes follow. */ |
158 | |
159 | zp->trailerlen -= len; |
160 | z->avail_in -= len; |
161 | z->next_in += len; |
162 | if(z->avail_in) |
163 | result = CURLE_WRITE_ERROR; |
164 | if(result || !zp->trailerlen) |
165 | result = exit_zlib(data, z, zlib_init: &zp->zlib_init, result); |
166 | else { |
167 | /* Only occurs for gzip with zlib < 1.2.0.4 or raw deflate. */ |
168 | zp->zlib_init = ZLIB_EXTERNAL_TRAILER; |
169 | } |
170 | return result; |
171 | } |
172 | |
173 | static CURLcode inflate_stream(struct Curl_easy *data, |
174 | struct contenc_writer *writer, |
175 | zlibInitState started) |
176 | { |
177 | struct zlib_writer *zp = (struct zlib_writer *) writer; |
178 | z_stream *z = &zp->z; /* zlib state structure */ |
179 | uInt nread = z->avail_in; |
180 | Bytef *orig_in = z->next_in; |
181 | bool done = FALSE; |
182 | CURLcode result = CURLE_OK; /* Curl_client_write status */ |
183 | char *decomp; /* Put the decompressed data here. */ |
184 | |
185 | /* Check state. */ |
186 | if(zp->zlib_init != ZLIB_INIT && |
187 | zp->zlib_init != ZLIB_INFLATING && |
188 | zp->zlib_init != ZLIB_INIT_GZIP && |
189 | zp->zlib_init != ZLIB_GZIP_INFLATING) |
190 | return exit_zlib(data, z, zlib_init: &zp->zlib_init, result: CURLE_WRITE_ERROR); |
191 | |
192 | /* Dynamically allocate a buffer for decompression because it's uncommonly |
193 | large to hold on the stack */ |
194 | decomp = malloc(DSIZ); |
195 | if(!decomp) |
196 | return exit_zlib(data, z, zlib_init: &zp->zlib_init, result: CURLE_OUT_OF_MEMORY); |
197 | |
198 | /* because the buffer size is fixed, iteratively decompress and transfer to |
199 | the client via downstream_write function. */ |
200 | while(!done) { |
201 | int status; /* zlib status */ |
202 | done = TRUE; |
203 | |
204 | /* (re)set buffer for decompressed output for every iteration */ |
205 | z->next_out = (Bytef *) decomp; |
206 | z->avail_out = DSIZ; |
207 | |
208 | #ifdef Z_BLOCK |
209 | /* Z_BLOCK is only available in zlib ver. >= 1.2.0.5 */ |
210 | status = inflate(strm: z, Z_BLOCK); |
211 | #else |
212 | /* fallback for zlib ver. < 1.2.0.5 */ |
213 | status = inflate(z, Z_SYNC_FLUSH); |
214 | #endif |
215 | |
216 | /* Flush output data if some. */ |
217 | if(z->avail_out != DSIZ) { |
218 | if(status == Z_OK || status == Z_STREAM_END) { |
219 | zp->zlib_init = started; /* Data started. */ |
220 | result = Curl_unencode_write(data, writer: writer->downstream, buf: decomp, |
221 | DSIZ - z->avail_out); |
222 | if(result) { |
223 | exit_zlib(data, z, zlib_init: &zp->zlib_init, result); |
224 | break; |
225 | } |
226 | } |
227 | } |
228 | |
229 | /* Dispatch by inflate() status. */ |
230 | switch(status) { |
231 | case Z_OK: |
232 | /* Always loop: there may be unflushed latched data in zlib state. */ |
233 | done = FALSE; |
234 | break; |
235 | case Z_BUF_ERROR: |
236 | /* No more data to flush: just exit loop. */ |
237 | break; |
238 | case Z_STREAM_END: |
239 | result = process_trailer(data, zp); |
240 | break; |
241 | case Z_DATA_ERROR: |
242 | /* some servers seem to not generate zlib headers, so this is an attempt |
243 | to fix and continue anyway */ |
244 | if(zp->zlib_init == ZLIB_INIT) { |
245 | /* Do not use inflateReset2(): only available since zlib 1.2.3.4. */ |
246 | (void) inflateEnd(strm: z); /* don't care about the return code */ |
247 | if(inflateInit2(z, -MAX_WBITS) == Z_OK) { |
248 | z->next_in = orig_in; |
249 | z->avail_in = nread; |
250 | zp->zlib_init = ZLIB_INFLATING; |
251 | zp->trailerlen = 4; /* Tolerate up to 4 unknown trailer bytes. */ |
252 | done = FALSE; |
253 | break; |
254 | } |
255 | zp->zlib_init = ZLIB_UNINIT; /* inflateEnd() already called. */ |
256 | } |
257 | result = exit_zlib(data, z, zlib_init: &zp->zlib_init, result: process_zlib_error(data, z)); |
258 | break; |
259 | default: |
260 | result = exit_zlib(data, z, zlib_init: &zp->zlib_init, result: process_zlib_error(data, z)); |
261 | break; |
262 | } |
263 | } |
264 | free(decomp); |
265 | |
266 | /* We're about to leave this call so the `nread' data bytes won't be seen |
267 | again. If we are in a state that would wrongly allow restart in raw mode |
268 | at the next call, assume output has already started. */ |
269 | if(nread && zp->zlib_init == ZLIB_INIT) |
270 | zp->zlib_init = started; /* Cannot restart anymore. */ |
271 | |
272 | return result; |
273 | } |
274 | |
275 | |
276 | /* Deflate handler. */ |
277 | static CURLcode deflate_init_writer(struct Curl_easy *data, |
278 | struct contenc_writer *writer) |
279 | { |
280 | struct zlib_writer *zp = (struct zlib_writer *) writer; |
281 | z_stream *z = &zp->z; /* zlib state structure */ |
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 | |
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, started: ZLIB_INFLATING); |
309 | } |
310 | |
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, zlib_init: &zp->zlib_init, result: 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. */ |
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 | /* Initialize zlib */ |
338 | z->zalloc = (alloc_func) zalloc_cb; |
339 | z->zfree = (free_func) zfree_cb; |
340 | |
341 | if(strcmp(s1: zlibVersion(), s2: "1.2.0.4" ) >= 0) { |
342 | /* zlib ver. >= 1.2.0.4 supports transparent gzip decompressing */ |
343 | if(inflateInit2(z, MAX_WBITS + 32) != Z_OK) { |
344 | return process_zlib_error(data, z); |
345 | } |
346 | zp->zlib_init = ZLIB_INIT_GZIP; /* Transparent gzip decompress state */ |
347 | } |
348 | else { |
349 | /* we must parse the gzip header and trailer ourselves */ |
350 | if(inflateInit2(z, -MAX_WBITS) != Z_OK) { |
351 | return process_zlib_error(data, z); |
352 | } |
353 | zp->trailerlen = 8; /* A CRC-32 and a 32-bit input size (RFC 1952, 2.2) */ |
354 | zp->zlib_init = ZLIB_INIT; /* Initial call state */ |
355 | } |
356 | |
357 | return CURLE_OK; |
358 | } |
359 | |
360 | #ifdef OLD_ZLIB_SUPPORT |
361 | /* Skip over the gzip header */ |
362 | static enum { |
363 | GZIP_OK, |
364 | GZIP_BAD, |
365 | GZIP_UNDERFLOW |
366 | } (unsigned char const *data, ssize_t len, ssize_t *) |
367 | { |
368 | int method, flags; |
369 | const ssize_t totallen = len; |
370 | |
371 | /* The shortest header is 10 bytes */ |
372 | if(len < 10) |
373 | return GZIP_UNDERFLOW; |
374 | |
375 | if((data[0] != GZIP_MAGIC_0) || (data[1] != GZIP_MAGIC_1)) |
376 | return GZIP_BAD; |
377 | |
378 | method = data[2]; |
379 | flags = data[3]; |
380 | |
381 | if(method != Z_DEFLATED || (flags & RESERVED) != 0) { |
382 | /* Can't handle this compression method or unknown flag */ |
383 | return GZIP_BAD; |
384 | } |
385 | |
386 | /* Skip over time, xflags, OS code and all previous bytes */ |
387 | len -= 10; |
388 | data += 10; |
389 | |
390 | if(flags & EXTRA_FIELD) { |
391 | ssize_t ; |
392 | |
393 | if(len < 2) |
394 | return GZIP_UNDERFLOW; |
395 | |
396 | extra_len = (data[1] << 8) | data[0]; |
397 | |
398 | if(len < (extra_len + 2)) |
399 | return GZIP_UNDERFLOW; |
400 | |
401 | len -= (extra_len + 2); |
402 | data += (extra_len + 2); |
403 | } |
404 | |
405 | if(flags & ORIG_NAME) { |
406 | /* Skip over NUL-terminated file name */ |
407 | while(len && *data) { |
408 | --len; |
409 | ++data; |
410 | } |
411 | if(!len || *data) |
412 | return GZIP_UNDERFLOW; |
413 | |
414 | /* Skip over the NUL */ |
415 | --len; |
416 | ++data; |
417 | } |
418 | |
419 | if(flags & COMMENT) { |
420 | /* Skip over NUL-terminated comment */ |
421 | while(len && *data) { |
422 | --len; |
423 | ++data; |
424 | } |
425 | if(!len || *data) |
426 | return GZIP_UNDERFLOW; |
427 | |
428 | /* Skip over the NUL */ |
429 | --len; |
430 | } |
431 | |
432 | if(flags & HEAD_CRC) { |
433 | if(len < 2) |
434 | return GZIP_UNDERFLOW; |
435 | |
436 | len -= 2; |
437 | } |
438 | |
439 | *headerlen = totallen - len; |
440 | return GZIP_OK; |
441 | } |
442 | #endif |
443 | |
444 | static CURLcode gzip_unencode_write(struct Curl_easy *data, |
445 | struct contenc_writer *writer, |
446 | const char *buf, size_t nbytes) |
447 | { |
448 | struct zlib_writer *zp = (struct zlib_writer *) writer; |
449 | z_stream *z = &zp->z; /* zlib state structure */ |
450 | |
451 | if(zp->zlib_init == ZLIB_INIT_GZIP) { |
452 | /* Let zlib handle the gzip decompression entirely */ |
453 | z->next_in = (Bytef *) buf; |
454 | z->avail_in = (uInt) nbytes; |
455 | /* Now uncompress the data */ |
456 | return inflate_stream(data, writer, started: ZLIB_INIT_GZIP); |
457 | } |
458 | |
459 | #ifndef OLD_ZLIB_SUPPORT |
460 | /* Support for old zlib versions is compiled away and we are running with |
461 | an old version, so return an error. */ |
462 | return exit_zlib(data, z, &zp->zlib_init, CURLE_WRITE_ERROR); |
463 | |
464 | #else |
465 | /* This next mess is to get around the potential case where there isn't |
466 | * enough data passed in to skip over the gzip header. If that happens, we |
467 | * malloc a block and copy what we have then wait for the next call. If |
468 | * there still isn't enough (this is definitely a worst-case scenario), we |
469 | * make the block bigger, copy the next part in and keep waiting. |
470 | * |
471 | * This is only required with zlib versions < 1.2.0.4 as newer versions |
472 | * can handle the gzip header themselves. |
473 | */ |
474 | |
475 | switch(zp->zlib_init) { |
476 | /* Skip over gzip header? */ |
477 | case ZLIB_INIT: |
478 | { |
479 | /* Initial call state */ |
480 | ssize_t hlen; |
481 | |
482 | switch(check_gzip_header(data: (unsigned char *) buf, len: nbytes, headerlen: &hlen)) { |
483 | case GZIP_OK: |
484 | z->next_in = (Bytef *) buf + hlen; |
485 | z->avail_in = (uInt) (nbytes - hlen); |
486 | zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */ |
487 | break; |
488 | |
489 | case GZIP_UNDERFLOW: |
490 | /* We need more data so we can find the end of the gzip header. It's |
491 | * possible that the memory block we malloc here will never be freed if |
492 | * the transfer abruptly aborts after this point. Since it's unlikely |
493 | * that circumstances will be right for this code path to be followed in |
494 | * the first place, and it's even more unlikely for a transfer to fail |
495 | * immediately afterwards, it should seldom be a problem. |
496 | */ |
497 | z->avail_in = (uInt) nbytes; |
498 | z->next_in = malloc(z->avail_in); |
499 | if(!z->next_in) { |
500 | return exit_zlib(data, z, zlib_init: &zp->zlib_init, result: CURLE_OUT_OF_MEMORY); |
501 | } |
502 | memcpy(dest: z->next_in, src: buf, n: z->avail_in); |
503 | zp->zlib_init = ZLIB_GZIP_HEADER; /* Need more gzip header data state */ |
504 | /* We don't have any data to inflate yet */ |
505 | return CURLE_OK; |
506 | |
507 | case GZIP_BAD: |
508 | default: |
509 | return exit_zlib(data, z, zlib_init: &zp->zlib_init, result: process_zlib_error(data, z)); |
510 | } |
511 | |
512 | } |
513 | break; |
514 | |
515 | case ZLIB_GZIP_HEADER: |
516 | { |
517 | /* Need more gzip header data state */ |
518 | ssize_t hlen; |
519 | z->avail_in += (uInt) nbytes; |
520 | z->next_in = Curl_saferealloc(ptr: z->next_in, size: z->avail_in); |
521 | if(!z->next_in) { |
522 | return exit_zlib(data, z, zlib_init: &zp->zlib_init, result: CURLE_OUT_OF_MEMORY); |
523 | } |
524 | /* Append the new block of data to the previous one */ |
525 | memcpy(dest: z->next_in + z->avail_in - nbytes, src: buf, n: nbytes); |
526 | |
527 | switch(check_gzip_header(data: z->next_in, len: z->avail_in, headerlen: &hlen)) { |
528 | case GZIP_OK: |
529 | /* This is the zlib stream data */ |
530 | free(z->next_in); |
531 | /* Don't point into the malloced block since we just freed it */ |
532 | z->next_in = (Bytef *) buf + hlen + nbytes - z->avail_in; |
533 | z->avail_in = (uInt) (z->avail_in - hlen); |
534 | zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */ |
535 | break; |
536 | |
537 | case GZIP_UNDERFLOW: |
538 | /* We still don't have any data to inflate! */ |
539 | return CURLE_OK; |
540 | |
541 | case GZIP_BAD: |
542 | default: |
543 | return exit_zlib(data, z, zlib_init: &zp->zlib_init, result: process_zlib_error(data, z)); |
544 | } |
545 | |
546 | } |
547 | break; |
548 | |
549 | case ZLIB_EXTERNAL_TRAILER: |
550 | z->next_in = (Bytef *) buf; |
551 | z->avail_in = (uInt) nbytes; |
552 | return process_trailer(data, zp); |
553 | |
554 | case ZLIB_GZIP_INFLATING: |
555 | default: |
556 | /* Inflating stream state */ |
557 | z->next_in = (Bytef *) buf; |
558 | z->avail_in = (uInt) nbytes; |
559 | break; |
560 | } |
561 | |
562 | if(z->avail_in == 0) { |
563 | /* We don't have any data to inflate; wait until next time */ |
564 | return CURLE_OK; |
565 | } |
566 | |
567 | /* We've parsed the header, now uncompress the data */ |
568 | return inflate_stream(data, writer, started: ZLIB_GZIP_INFLATING); |
569 | #endif |
570 | } |
571 | |
572 | static void gzip_close_writer(struct Curl_easy *data, |
573 | struct contenc_writer *writer) |
574 | { |
575 | struct zlib_writer *zp = (struct zlib_writer *) writer; |
576 | z_stream *z = &zp->z; /* zlib state structure */ |
577 | |
578 | exit_zlib(data, z, zlib_init: &zp->zlib_init, result: CURLE_OK); |
579 | } |
580 | |
581 | static const struct content_encoding gzip_encoding = { |
582 | "gzip" , |
583 | "x-gzip" , |
584 | gzip_init_writer, |
585 | gzip_unencode_write, |
586 | gzip_close_writer, |
587 | sizeof(struct zlib_writer) |
588 | }; |
589 | |
590 | #endif /* HAVE_LIBZ */ |
591 | |
592 | |
593 | #ifdef HAVE_BROTLI |
594 | /* Brotli writer. */ |
595 | struct brotli_writer { |
596 | struct contenc_writer super; |
597 | BrotliDecoderState *br; /* State structure for brotli. */ |
598 | }; |
599 | |
600 | static CURLcode brotli_map_error(BrotliDecoderErrorCode be) |
601 | { |
602 | switch(be) { |
603 | case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_NIBBLE: |
604 | case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_META_NIBBLE: |
605 | case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_ALPHABET: |
606 | case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_SAME: |
607 | case BROTLI_DECODER_ERROR_FORMAT_CL_SPACE: |
608 | case BROTLI_DECODER_ERROR_FORMAT_HUFFMAN_SPACE: |
609 | case BROTLI_DECODER_ERROR_FORMAT_CONTEXT_MAP_REPEAT: |
610 | case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_1: |
611 | case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_2: |
612 | case BROTLI_DECODER_ERROR_FORMAT_TRANSFORM: |
613 | case BROTLI_DECODER_ERROR_FORMAT_DICTIONARY: |
614 | case BROTLI_DECODER_ERROR_FORMAT_WINDOW_BITS: |
615 | case BROTLI_DECODER_ERROR_FORMAT_PADDING_1: |
616 | case BROTLI_DECODER_ERROR_FORMAT_PADDING_2: |
617 | #ifdef BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY |
618 | case BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY: |
619 | #endif |
620 | #ifdef BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET |
621 | case BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET: |
622 | #endif |
623 | case BROTLI_DECODER_ERROR_INVALID_ARGUMENTS: |
624 | return CURLE_BAD_CONTENT_ENCODING; |
625 | case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MODES: |
626 | case BROTLI_DECODER_ERROR_ALLOC_TREE_GROUPS: |
627 | case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MAP: |
628 | case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_1: |
629 | case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_2: |
630 | case BROTLI_DECODER_ERROR_ALLOC_BLOCK_TYPE_TREES: |
631 | return CURLE_OUT_OF_MEMORY; |
632 | default: |
633 | break; |
634 | } |
635 | return CURLE_WRITE_ERROR; |
636 | } |
637 | |
638 | static CURLcode brotli_init_writer(struct Curl_easy *data, |
639 | struct contenc_writer *writer) |
640 | { |
641 | struct brotli_writer *bp = (struct brotli_writer *) writer; |
642 | (void) data; |
643 | |
644 | bp->br = BrotliDecoderCreateInstance(NULL, NULL, NULL); |
645 | return bp->br? CURLE_OK: CURLE_OUT_OF_MEMORY; |
646 | } |
647 | |
648 | static CURLcode brotli_unencode_write(struct Curl_easy *data, |
649 | struct contenc_writer *writer, |
650 | const char *buf, size_t nbytes) |
651 | { |
652 | struct brotli_writer *bp = (struct brotli_writer *) writer; |
653 | const uint8_t *src = (const uint8_t *) buf; |
654 | char *decomp; |
655 | uint8_t *dst; |
656 | size_t dstleft; |
657 | CURLcode result = CURLE_OK; |
658 | BrotliDecoderResult r = BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT; |
659 | |
660 | if(!bp->br) |
661 | return CURLE_WRITE_ERROR; /* Stream already ended. */ |
662 | |
663 | decomp = malloc(DSIZ); |
664 | if(!decomp) |
665 | return CURLE_OUT_OF_MEMORY; |
666 | |
667 | while((nbytes || r == BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT) && |
668 | result == CURLE_OK) { |
669 | dst = (uint8_t *) decomp; |
670 | dstleft = DSIZ; |
671 | r = BrotliDecoderDecompressStream(bp->br, |
672 | &nbytes, &src, &dstleft, &dst, NULL); |
673 | result = Curl_unencode_write(data, writer->downstream, |
674 | decomp, DSIZ - dstleft); |
675 | if(result) |
676 | break; |
677 | switch(r) { |
678 | case BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT: |
679 | case BROTLI_DECODER_RESULT_NEEDS_MORE_INPUT: |
680 | break; |
681 | case BROTLI_DECODER_RESULT_SUCCESS: |
682 | BrotliDecoderDestroyInstance(bp->br); |
683 | bp->br = NULL; |
684 | if(nbytes) |
685 | result = CURLE_WRITE_ERROR; |
686 | break; |
687 | default: |
688 | result = brotli_map_error(BrotliDecoderGetErrorCode(bp->br)); |
689 | break; |
690 | } |
691 | } |
692 | free(decomp); |
693 | return result; |
694 | } |
695 | |
696 | static void brotli_close_writer(struct Curl_easy *data, |
697 | struct contenc_writer *writer) |
698 | { |
699 | struct brotli_writer *bp = (struct brotli_writer *) writer; |
700 | |
701 | (void) data; |
702 | |
703 | if(bp->br) { |
704 | BrotliDecoderDestroyInstance(bp->br); |
705 | bp->br = NULL; |
706 | } |
707 | } |
708 | |
709 | static const struct content_encoding brotli_encoding = { |
710 | "br" , |
711 | NULL, |
712 | brotli_init_writer, |
713 | brotli_unencode_write, |
714 | brotli_close_writer, |
715 | sizeof(struct brotli_writer) |
716 | }; |
717 | #endif |
718 | |
719 | |
720 | #ifdef HAVE_ZSTD |
721 | /* Zstd writer. */ |
722 | struct zstd_writer { |
723 | struct contenc_writer super; |
724 | ZSTD_DStream *zds; /* State structure for zstd. */ |
725 | void *decomp; |
726 | }; |
727 | |
728 | static CURLcode zstd_init_writer(struct Curl_easy *data, |
729 | struct contenc_writer *writer) |
730 | { |
731 | struct zstd_writer *zp = (struct zstd_writer *) writer; |
732 | |
733 | (void)data; |
734 | |
735 | zp->zds = ZSTD_createDStream(); |
736 | zp->decomp = NULL; |
737 | return zp->zds ? CURLE_OK : CURLE_OUT_OF_MEMORY; |
738 | } |
739 | |
740 | static CURLcode zstd_unencode_write(struct Curl_easy *data, |
741 | struct contenc_writer *writer, |
742 | const char *buf, size_t nbytes) |
743 | { |
744 | CURLcode result = CURLE_OK; |
745 | struct zstd_writer *zp = (struct zstd_writer *) writer; |
746 | ZSTD_inBuffer in; |
747 | ZSTD_outBuffer out; |
748 | size_t errorCode; |
749 | |
750 | if(!zp->decomp) { |
751 | zp->decomp = malloc(DSIZ); |
752 | if(!zp->decomp) |
753 | return CURLE_OUT_OF_MEMORY; |
754 | } |
755 | in.pos = 0; |
756 | in.src = buf; |
757 | in.size = nbytes; |
758 | |
759 | for(;;) { |
760 | out.pos = 0; |
761 | out.dst = zp->decomp; |
762 | out.size = DSIZ; |
763 | |
764 | errorCode = ZSTD_decompressStream(zp->zds, &out, &in); |
765 | if(ZSTD_isError(errorCode)) { |
766 | return CURLE_BAD_CONTENT_ENCODING; |
767 | } |
768 | if(out.pos > 0) { |
769 | result = Curl_unencode_write(data, writer->downstream, |
770 | zp->decomp, out.pos); |
771 | if(result) |
772 | break; |
773 | } |
774 | if((in.pos == nbytes) && (out.pos < out.size)) |
775 | break; |
776 | } |
777 | |
778 | return result; |
779 | } |
780 | |
781 | static void zstd_close_writer(struct Curl_easy *data, |
782 | struct contenc_writer *writer) |
783 | { |
784 | struct zstd_writer *zp = (struct zstd_writer *) writer; |
785 | |
786 | (void)data; |
787 | |
788 | if(zp->decomp) { |
789 | free(zp->decomp); |
790 | zp->decomp = NULL; |
791 | } |
792 | if(zp->zds) { |
793 | ZSTD_freeDStream(zp->zds); |
794 | zp->zds = NULL; |
795 | } |
796 | } |
797 | |
798 | static const struct content_encoding zstd_encoding = { |
799 | "zstd" , |
800 | NULL, |
801 | zstd_init_writer, |
802 | zstd_unencode_write, |
803 | zstd_close_writer, |
804 | sizeof(struct zstd_writer) |
805 | }; |
806 | #endif |
807 | |
808 | |
809 | /* Identity handler. */ |
810 | static CURLcode identity_init_writer(struct Curl_easy *data, |
811 | struct contenc_writer *writer) |
812 | { |
813 | (void)data; |
814 | (void)writer; |
815 | return CURLE_OK; |
816 | } |
817 | |
818 | static CURLcode identity_unencode_write(struct Curl_easy *data, |
819 | struct contenc_writer *writer, |
820 | const char *buf, size_t nbytes) |
821 | { |
822 | return Curl_unencode_write(data, writer: writer->downstream, buf, nbytes); |
823 | } |
824 | |
825 | static void identity_close_writer(struct Curl_easy *data, |
826 | struct contenc_writer *writer) |
827 | { |
828 | (void) data; |
829 | (void) writer; |
830 | } |
831 | |
832 | static const struct content_encoding identity_encoding = { |
833 | "identity" , |
834 | "none" , |
835 | identity_init_writer, |
836 | identity_unencode_write, |
837 | identity_close_writer, |
838 | sizeof(struct contenc_writer) |
839 | }; |
840 | |
841 | |
842 | /* supported content encodings table. */ |
843 | static const struct content_encoding * const encodings[] = { |
844 | &identity_encoding, |
845 | #ifdef HAVE_LIBZ |
846 | &deflate_encoding, |
847 | &gzip_encoding, |
848 | #endif |
849 | #ifdef HAVE_BROTLI |
850 | &brotli_encoding, |
851 | #endif |
852 | #ifdef HAVE_ZSTD |
853 | &zstd_encoding, |
854 | #endif |
855 | NULL |
856 | }; |
857 | |
858 | |
859 | /* Return a list of comma-separated names of supported encodings. */ |
860 | char *Curl_all_content_encodings(void) |
861 | { |
862 | size_t len = 0; |
863 | const struct content_encoding * const *cep; |
864 | const struct content_encoding *ce; |
865 | char *ace; |
866 | |
867 | for(cep = encodings; *cep; cep++) { |
868 | ce = *cep; |
869 | if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT)) |
870 | len += strlen(s: ce->name) + 2; |
871 | } |
872 | |
873 | if(!len) |
874 | return strdup(CONTENT_ENCODING_DEFAULT); |
875 | |
876 | ace = malloc(len); |
877 | if(ace) { |
878 | char *p = ace; |
879 | for(cep = encodings; *cep; cep++) { |
880 | ce = *cep; |
881 | if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT)) { |
882 | strcpy(dest: p, src: ce->name); |
883 | p += strlen(s: p); |
884 | *p++ = ','; |
885 | *p++ = ' '; |
886 | } |
887 | } |
888 | p[-2] = '\0'; |
889 | } |
890 | |
891 | return ace; |
892 | } |
893 | |
894 | |
895 | /* Deferred error dummy writer. */ |
896 | static CURLcode error_init_writer(struct Curl_easy *data, |
897 | struct contenc_writer *writer) |
898 | { |
899 | (void)data; |
900 | (void)writer; |
901 | return CURLE_OK; |
902 | } |
903 | |
904 | static CURLcode error_unencode_write(struct Curl_easy *data, |
905 | struct contenc_writer *writer, |
906 | const char *buf, size_t nbytes) |
907 | { |
908 | char *all = Curl_all_content_encodings(); |
909 | |
910 | (void) writer; |
911 | (void) buf; |
912 | (void) nbytes; |
913 | |
914 | if(!all) |
915 | return CURLE_OUT_OF_MEMORY; |
916 | failf(data, fmt: "Unrecognized content encoding type. " |
917 | "libcurl understands %s content encodings." , all); |
918 | free(all); |
919 | return CURLE_BAD_CONTENT_ENCODING; |
920 | } |
921 | |
922 | static void error_close_writer(struct Curl_easy *data, |
923 | struct contenc_writer *writer) |
924 | { |
925 | (void) data; |
926 | (void) writer; |
927 | } |
928 | |
929 | static const struct content_encoding error_encoding = { |
930 | NULL, |
931 | NULL, |
932 | error_init_writer, |
933 | error_unencode_write, |
934 | error_close_writer, |
935 | sizeof(struct contenc_writer) |
936 | }; |
937 | |
938 | /* Write data using an unencoding writer stack. "nbytes" is not |
939 | allowed to be 0. */ |
940 | CURLcode Curl_unencode_write(struct Curl_easy *data, |
941 | struct contenc_writer *writer, |
942 | const char *buf, size_t nbytes) |
943 | { |
944 | if(!nbytes) |
945 | return CURLE_OK; |
946 | if(!writer) |
947 | return CURLE_WRITE_ERROR; |
948 | return writer->handler->unencode_write(data, writer, buf, nbytes); |
949 | } |
950 | |
951 | /* Find the content encoding by name. */ |
952 | static const struct content_encoding *find_encoding(const char *name, |
953 | size_t len) |
954 | { |
955 | const struct content_encoding * const *cep; |
956 | |
957 | for(cep = encodings; *cep; cep++) { |
958 | const struct content_encoding *ce = *cep; |
959 | if((strncasecompare(name, ce->name, len) && !ce->name[len]) || |
960 | (ce->alias && strncasecompare(name, ce->alias, len) && !ce->alias[len])) |
961 | return ce; |
962 | } |
963 | return NULL; |
964 | } |
965 | |
966 | /* Set-up the unencoding stack from the Content-Encoding header value. |
967 | * See RFC 7231 section 3.1.2.2. */ |
968 | CURLcode Curl_build_unencoding_stack(struct Curl_easy *data, |
969 | const char *enclist, int is_transfer) |
970 | { |
971 | struct SingleRequest *k = &data->req; |
972 | unsigned int order = is_transfer? 2: 1; |
973 | CURLcode result; |
974 | |
975 | do { |
976 | const char *name; |
977 | size_t namelen; |
978 | |
979 | /* Parse a single encoding name. */ |
980 | while(ISBLANK(*enclist) || *enclist == ',') |
981 | enclist++; |
982 | |
983 | name = enclist; |
984 | |
985 | for(namelen = 0; *enclist && *enclist != ','; enclist++) |
986 | if(!ISSPACE(*enclist)) |
987 | namelen = enclist - name + 1; |
988 | |
989 | /* Special case: chunked encoding is handled at the reader level. */ |
990 | if(is_transfer && namelen == 7 && strncasecompare(name, "chunked" , 7)) { |
991 | k->chunk = TRUE; /* chunks coming our way. */ |
992 | Curl_httpchunk_init(data); /* init our chunky engine. */ |
993 | } |
994 | else if(namelen) { |
995 | const struct content_encoding *encoding; |
996 | struct contenc_writer *writer; |
997 | if(is_transfer && !data->set.http_transfer_encoding) |
998 | /* not requested, ignore */ |
999 | return CURLE_OK; |
1000 | |
1001 | encoding = find_encoding(name, len: namelen); |
1002 | if(!encoding) |
1003 | encoding = &error_encoding; /* Defer error at stack use. */ |
1004 | |
1005 | result = Curl_client_create_writer(pwriter: &writer, data, ce_handler: encoding, order); |
1006 | if(result) |
1007 | return result; |
1008 | |
1009 | result = Curl_client_add_writer(data, writer); |
1010 | if(result) { |
1011 | Curl_client_free_writer(data, writer); |
1012 | return result; |
1013 | } |
1014 | } |
1015 | } while(*enclist); |
1016 | |
1017 | return CURLE_OK; |
1018 | } |
1019 | |
1020 | #else |
1021 | /* Stubs for builds without HTTP. */ |
1022 | CURLcode Curl_build_unencoding_stack(struct Curl_easy *data, |
1023 | const char *enclist, int is_transfer) |
1024 | { |
1025 | (void) data; |
1026 | (void) enclist; |
1027 | (void) is_transfer; |
1028 | return CURLE_NOT_BUILT_IN; |
1029 | } |
1030 | |
1031 | CURLcode Curl_unencode_write(struct Curl_easy *data, |
1032 | struct contenc_writer *writer, |
1033 | const char *buf, size_t nbytes) |
1034 | { |
1035 | (void) data; |
1036 | (void) writer; |
1037 | (void) buf; |
1038 | (void) nbytes; |
1039 | return CURLE_NOT_BUILT_IN; |
1040 | } |
1041 | |
1042 | char *Curl_all_content_encodings(void) |
1043 | { |
1044 | return strdup(CONTENT_ENCODING_DEFAULT); /* Satisfy caller. */ |
1045 | } |
1046 | |
1047 | #endif /* CURL_DISABLE_HTTP */ |
1048 | |