blob: 7c979efcc3608b54e32be56ba5e98858d0db1f3a [file] [log] [blame]
Kristian Monsen5ab50182010-05-14 18:53:44 +01001/***************************************************************************
2 * _ _ ____ _
3 * Project ___| | | | _ \| |
4 * / __| | | | |_) | |
5 * | (__| |_| | _ <| |___
6 * \___|\___/|_| \_\_____|
7 *
Alex Deymo486467e2017-12-19 19:04:07 +01008 * Copyright (C) 1998 - 2017, Daniel Stenberg, <daniel@haxx.se>, et al.
Kristian Monsen5ab50182010-05-14 18:53:44 +01009 *
10 * This software is licensed as described in the file COPYING, which
11 * you should have received as part of this distribution. The terms
Alex Deymod15eaac2016-06-28 14:49:26 -070012 * are also available at https://curl.haxx.se/docs/copyright.html.
Kristian Monsen5ab50182010-05-14 18:53:44 +010013 *
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 ***************************************************************************/
22
Bertrand SIMONNETe6cd7382015-07-01 15:39:44 -070023#include "curl_setup.h"
Kristian Monsen5ab50182010-05-14 18:53:44 +010024
Kristian Monsen5ab50182010-05-14 18:53:44 +010025#include "urldata.h"
26#include <curl/curl.h>
Alex Deymo486467e2017-12-19 19:04:07 +010027#include <stddef.h>
28
29#ifdef HAVE_ZLIB_H
30#include <zlib.h>
31#ifdef __SYMBIAN32__
32/* zlib pollutes the namespace with this definition */
33#undef WIN32
34#endif
35#endif
36
37#ifdef HAVE_BROTLI
38#include <brotli/decode.h>
39#endif
40
Kristian Monsen5ab50182010-05-14 18:53:44 +010041#include "sendf.h"
Alex Deymo486467e2017-12-19 19:04:07 +010042#include "http.h"
Kristian Monsen5ab50182010-05-14 18:53:44 +010043#include "content_encoding.h"
Elliott Hughescee03382017-06-23 12:17:18 -070044#include "strdup.h"
Alex Deymo486467e2017-12-19 19:04:07 +010045#include "strcase.h"
Kristian Monsen5ab50182010-05-14 18:53:44 +010046#include "curl_memory.h"
Kristian Monsen5ab50182010-05-14 18:53:44 +010047#include "memdebug.h"
48
Alex Deymo486467e2017-12-19 19:04:07 +010049#define CONTENT_ENCODING_DEFAULT "identity"
50
51#ifndef CURL_DISABLE_HTTP
52
53#define DSIZ CURL_MAX_WRITE_SIZE /* buffer size for decompressed data */
54
55
56#ifdef HAVE_LIBZ
57
Kristian Monsen5ab50182010-05-14 18:53:44 +010058/* Comment this out if zlib is always going to be at least ver. 1.2.0.4
59 (doing so will reduce code size slightly). */
60#define OLD_ZLIB_SUPPORT 1
61
Kristian Monsen5ab50182010-05-14 18:53:44 +010062#define GZIP_MAGIC_0 0x1f
63#define GZIP_MAGIC_1 0x8b
64
65/* gzip flag byte */
66#define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
67#define HEAD_CRC 0x02 /* bit 1 set: header CRC present */
68#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
69#define ORIG_NAME 0x08 /* bit 3 set: original file name present */
70#define COMMENT 0x10 /* bit 4 set: file comment present */
71#define RESERVED 0xE0 /* bits 5..7: reserved */
72
Alex Deymo486467e2017-12-19 19:04:07 +010073typedef enum {
74 ZLIB_UNINIT, /* uninitialized */
75 ZLIB_INIT, /* initialized */
Elliott Hughes0128fe42018-02-27 14:57:55 -080076 ZLIB_INFLATING, /* Inflating started. */
Alex Deymo486467e2017-12-19 19:04:07 +010077 ZLIB_GZIP_HEADER, /* reading gzip header */
Elliott Hughes0128fe42018-02-27 14:57:55 -080078 ZLIB_GZIP_TRAILER, /* reading gzip trailer */
Alex Deymo486467e2017-12-19 19:04:07 +010079 ZLIB_GZIP_INFLATING, /* inflating gzip stream */
80 ZLIB_INIT_GZIP /* initialized in transparent gzip mode */
81} zlibInitState;
82
83/* Writer parameters. */
84typedef struct {
Elliott Hughes0128fe42018-02-27 14:57:55 -080085 zlibInitState zlib_init; /* zlib init state */
86 uInt trailerlen; /* Remaining trailer byte count. */
Alex Deymo486467e2017-12-19 19:04:07 +010087 z_stream z; /* State structure for zlib. */
88} zlib_params;
89
90
Bertrand SIMONNETe6cd7382015-07-01 15:39:44 -070091static voidpf
92zalloc_cb(voidpf opaque, unsigned int items, unsigned int size)
93{
94 (void) opaque;
95 /* not a typo, keep it calloc() */
96 return (voidpf) calloc(items, size);
97}
98
99static void
100zfree_cb(voidpf opaque, voidpf ptr)
101{
102 (void) opaque;
103 free(ptr);
104}
105
Kristian Monsen5ab50182010-05-14 18:53:44 +0100106static CURLcode
107process_zlib_error(struct connectdata *conn, z_stream *z)
108{
Alex Deymoe3149cc2016-10-05 11:18:42 -0700109 struct Curl_easy *data = conn->data;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100110 if(z->msg)
Elliott Hughes82be86d2017-09-20 17:00:17 -0700111 failf(data, "Error while processing content unencoding: %s",
112 z->msg);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100113 else
Elliott Hughes82be86d2017-09-20 17:00:17 -0700114 failf(data, "Error while processing content unencoding: "
115 "Unknown failure within decompression software.");
Kristian Monsen5ab50182010-05-14 18:53:44 +0100116
117 return CURLE_BAD_CONTENT_ENCODING;
118}
119
120static CURLcode
Alex Deymo486467e2017-12-19 19:04:07 +0100121exit_zlib(struct connectdata *conn,
122 z_stream *z, zlibInitState *zlib_init, CURLcode result)
Kristian Monsen5ab50182010-05-14 18:53:44 +0100123{
Alex Deymo486467e2017-12-19 19:04:07 +0100124 if(*zlib_init == ZLIB_GZIP_HEADER)
125 Curl_safefree(z->next_in);
126
127 if(*zlib_init != ZLIB_UNINIT) {
128 if(inflateEnd(z) != Z_OK && result == CURLE_OK)
129 result = process_zlib_error(conn, z);
130 *zlib_init = ZLIB_UNINIT;
131 }
132
Kristian Monsen5ab50182010-05-14 18:53:44 +0100133 return result;
134}
135
Elliott Hughes0128fe42018-02-27 14:57:55 -0800136static CURLcode process_trailer(struct connectdata *conn, zlib_params *zp)
137{
138 z_stream *z = &zp->z;
139 CURLcode result = CURLE_OK;
140 uInt len = z->avail_in < zp->trailerlen? z->avail_in: zp->trailerlen;
141
142 /* Consume expected trailer bytes. Terminate stream if exhausted.
143 Issue an error if unexpected bytes follow. */
144
145 zp->trailerlen -= len;
146 z->avail_in -= len;
147 z->next_in += len;
148 if(z->avail_in)
149 result = CURLE_WRITE_ERROR;
150 if(result || !zp->trailerlen)
151 result = exit_zlib(conn, z, &zp->zlib_init, result);
152 else {
153 /* Only occurs for gzip with zlib < 1.2.0.4. */
154 zp->zlib_init = ZLIB_GZIP_TRAILER;
155 }
156 return result;
157}
158
159static CURLcode inflate_stream(struct connectdata *conn,
160 contenc_writer *writer, zlibInitState started)
Kristian Monsen5ab50182010-05-14 18:53:44 +0100161{
Alex Deymo486467e2017-12-19 19:04:07 +0100162 zlib_params *zp = (zlib_params *) &writer->params;
Alex Deymo486467e2017-12-19 19:04:07 +0100163 z_stream *z = &zp->z; /* zlib state structure */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100164 uInt nread = z->avail_in;
165 Bytef *orig_in = z->next_in;
166 int status; /* zlib status */
Elliott Hughes0128fe42018-02-27 14:57:55 -0800167 bool done = FALSE;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100168 CURLcode result = CURLE_OK; /* Curl_client_write status */
169 char *decomp; /* Put the decompressed data here. */
170
Elliott Hughes0128fe42018-02-27 14:57:55 -0800171 /* Check state. */
172 if(zp->zlib_init != ZLIB_INIT &&
173 zp->zlib_init != ZLIB_INFLATING &&
174 zp->zlib_init != ZLIB_INIT_GZIP &&
175 zp->zlib_init != ZLIB_GZIP_INFLATING)
176 return exit_zlib(conn, z, &zp->zlib_init, CURLE_WRITE_ERROR);
177
Kristian Monsen5ab50182010-05-14 18:53:44 +0100178 /* Dynamically allocate a buffer for decompression because it's uncommonly
179 large to hold on the stack */
180 decomp = malloc(DSIZ);
Elliott Hughes0128fe42018-02-27 14:57:55 -0800181 if(decomp == NULL)
Alex Deymo486467e2017-12-19 19:04:07 +0100182 return exit_zlib(conn, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100183
184 /* because the buffer size is fixed, iteratively decompress and transfer to
Elliott Hughes0128fe42018-02-27 14:57:55 -0800185 the client via downstream_write function. */
186 while(!done) {
187 done = TRUE;
Alex Deymo486467e2017-12-19 19:04:07 +0100188
Kristian Monsen5ab50182010-05-14 18:53:44 +0100189 /* (re)set buffer for decompressed output for every iteration */
Alex Deymo486467e2017-12-19 19:04:07 +0100190 z->next_out = (Bytef *) decomp;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100191 z->avail_out = DSIZ;
192
Elliott Hughes0128fe42018-02-27 14:57:55 -0800193 status = inflate(z, Z_BLOCK);
194
195 /* Flush output data if some. */
196 if(z->avail_out != DSIZ) {
197 if(status == Z_OK || status == Z_STREAM_END) {
198 zp->zlib_init = started; /* Data started. */
199 result = Curl_unencode_write(conn, writer->downstream, decomp,
200 DSIZ - z->avail_out);
201 if(result) {
202 exit_zlib(conn, z, &zp->zlib_init, result);
203 break;
204 }
Kristian Monsen5ab50182010-05-14 18:53:44 +0100205 }
Kristian Monsen5ab50182010-05-14 18:53:44 +0100206 }
Elliott Hughes0128fe42018-02-27 14:57:55 -0800207
208 /* Dispatch by inflate() status. */
209 switch(status) {
210 case Z_OK:
211 /* Always loop: there may be unflushed latched data in zlib state. */
212 done = FALSE;
213 break;
214 case Z_BUF_ERROR:
215 /* No more data to flush: just exit loop. */
216 break;
217 case Z_STREAM_END:
218 result = process_trailer(conn, zp);
219 break;
220 case Z_DATA_ERROR:
Kristian Monsen5ab50182010-05-14 18:53:44 +0100221 /* some servers seem to not generate zlib headers, so this is an attempt
222 to fix and continue anyway */
Elliott Hughes0128fe42018-02-27 14:57:55 -0800223 if(zp->zlib_init == ZLIB_INIT) {
224 /* Do not use inflateReset2(): only available since zlib 1.2.3.4. */
225 (void) inflateEnd(z); /* don't care about the return code */
226 if(inflateInit2(z, -MAX_WBITS) == Z_OK) {
227 z->next_in = orig_in;
228 z->avail_in = nread;
229 zp->zlib_init = ZLIB_INFLATING;
230 done = FALSE;
231 break;
232 }
233 zp->zlib_init = ZLIB_UNINIT; /* inflateEnd() already called. */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100234 }
Elliott Hughes0128fe42018-02-27 14:57:55 -0800235 /* FALLTHROUGH */
236 default:
237 result = exit_zlib(conn, z, &zp->zlib_init, process_zlib_error(conn, z));
238 break;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100239 }
240 }
Elliott Hughes0128fe42018-02-27 14:57:55 -0800241 free(decomp);
242
243 /* We're about to leave this call so the `nread' data bytes won't be seen
244 again. If we are in a state that would wrongly allow restart in raw mode
245 at the next call, assume output has already started. */
246 if(nread && zp->zlib_init == ZLIB_INIT)
247 zp->zlib_init = started; /* Cannot restart anymore. */
248
249 return result;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100250}
251
Alex Deymo486467e2017-12-19 19:04:07 +0100252
253/* Deflate handler. */
254static CURLcode deflate_init_writer(struct connectdata *conn,
255 contenc_writer *writer)
Kristian Monsen5ab50182010-05-14 18:53:44 +0100256{
Alex Deymo486467e2017-12-19 19:04:07 +0100257 zlib_params *zp = (zlib_params *) &writer->params;
258 z_stream *z = &zp->z; /* zlib state structure */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100259
Alex Deymo486467e2017-12-19 19:04:07 +0100260 if(!writer->downstream)
261 return CURLE_WRITE_ERROR;
Bertrand SIMONNETe6cd7382015-07-01 15:39:44 -0700262
Alex Deymo486467e2017-12-19 19:04:07 +0100263 /* Initialize zlib */
264 z->zalloc = (alloc_func) zalloc_cb;
265 z->zfree = (free_func) zfree_cb;
266
267 if(inflateInit(z) != Z_OK)
268 return process_zlib_error(conn, z);
269 zp->zlib_init = ZLIB_INIT;
270 return CURLE_OK;
271}
272
273static CURLcode deflate_unencode_write(struct connectdata *conn,
274 contenc_writer *writer,
275 const char *buf, size_t nbytes)
276{
277 zlib_params *zp = (zlib_params *) &writer->params;
278 z_stream *z = &zp->z; /* zlib state structure */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100279
280 /* Set the compressed input when this function is called */
Alex Deymo486467e2017-12-19 19:04:07 +0100281 z->next_in = (Bytef *) buf;
282 z->avail_in = (uInt) nbytes;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100283
284 /* Now uncompress the data */
Elliott Hughes0128fe42018-02-27 14:57:55 -0800285 return inflate_stream(conn, writer, ZLIB_INFLATING);
Alex Deymo486467e2017-12-19 19:04:07 +0100286}
287
288static void deflate_close_writer(struct connectdata *conn,
289 contenc_writer *writer)
290{
291 zlib_params *zp = (zlib_params *) &writer->params;
292 z_stream *z = &zp->z; /* zlib state structure */
293
294 exit_zlib(conn, z, &zp->zlib_init, CURLE_OK);
295}
296
297static const content_encoding deflate_encoding = {
298 "deflate",
299 NULL,
300 deflate_init_writer,
301 deflate_unencode_write,
302 deflate_close_writer,
303 sizeof(zlib_params)
304};
305
306
307/* Gzip handler. */
308static CURLcode gzip_init_writer(struct connectdata *conn,
309 contenc_writer *writer)
310{
311 zlib_params *zp = (zlib_params *) &writer->params;
312 z_stream *z = &zp->z; /* zlib state structure */
313
314 if(!writer->downstream)
315 return CURLE_WRITE_ERROR;
316
317 /* Initialize zlib */
318 z->zalloc = (alloc_func) zalloc_cb;
319 z->zfree = (free_func) zfree_cb;
320
321 if(strcmp(zlibVersion(), "1.2.0.4") >= 0) {
322 /* zlib ver. >= 1.2.0.4 supports transparent gzip decompressing */
323 if(inflateInit2(z, MAX_WBITS + 32) != Z_OK) {
324 return process_zlib_error(conn, z);
325 }
326 zp->zlib_init = ZLIB_INIT_GZIP; /* Transparent gzip decompress state */
327 }
328 else {
Elliott Hughes0128fe42018-02-27 14:57:55 -0800329 /* we must parse the gzip header and trailer ourselves */
Alex Deymo486467e2017-12-19 19:04:07 +0100330 if(inflateInit2(z, -MAX_WBITS) != Z_OK) {
331 return process_zlib_error(conn, z);
332 }
Elliott Hughes0128fe42018-02-27 14:57:55 -0800333 zp->trailerlen = 8; /* A CRC-32 and a 32-bit input size (RFC 1952, 2.2) */
334 zp->zlib_init = ZLIB_INIT; /* Initial call state */
Alex Deymo486467e2017-12-19 19:04:07 +0100335 }
336
337 return CURLE_OK;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100338}
339
340#ifdef OLD_ZLIB_SUPPORT
341/* Skip over the gzip header */
342static enum {
343 GZIP_OK,
344 GZIP_BAD,
345 GZIP_UNDERFLOW
346} check_gzip_header(unsigned char const *data, ssize_t len, ssize_t *headerlen)
347{
348 int method, flags;
349 const ssize_t totallen = len;
350
351 /* The shortest header is 10 bytes */
352 if(len < 10)
353 return GZIP_UNDERFLOW;
354
355 if((data[0] != GZIP_MAGIC_0) || (data[1] != GZIP_MAGIC_1))
356 return GZIP_BAD;
357
358 method = data[2];
359 flags = data[3];
360
361 if(method != Z_DEFLATED || (flags & RESERVED) != 0) {
362 /* Can't handle this compression method or unknown flag */
363 return GZIP_BAD;
364 }
365
366 /* Skip over time, xflags, OS code and all previous bytes */
367 len -= 10;
368 data += 10;
369
370 if(flags & EXTRA_FIELD) {
371 ssize_t extra_len;
372
373 if(len < 2)
374 return GZIP_UNDERFLOW;
375
376 extra_len = (data[1] << 8) | data[0];
377
Alex Deymo486467e2017-12-19 19:04:07 +0100378 if(len < (extra_len + 2))
Kristian Monsen5ab50182010-05-14 18:53:44 +0100379 return GZIP_UNDERFLOW;
380
381 len -= (extra_len + 2);
382 data += (extra_len + 2);
383 }
384
385 if(flags & ORIG_NAME) {
386 /* Skip over NUL-terminated file name */
387 while(len && *data) {
388 --len;
389 ++data;
390 }
391 if(!len || *data)
392 return GZIP_UNDERFLOW;
393
394 /* Skip over the NUL */
395 --len;
396 ++data;
397 }
398
399 if(flags & COMMENT) {
400 /* Skip over NUL-terminated comment */
401 while(len && *data) {
402 --len;
403 ++data;
404 }
405 if(!len || *data)
406 return GZIP_UNDERFLOW;
407
408 /* Skip over the NUL */
409 --len;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100410 }
411
412 if(flags & HEAD_CRC) {
413 if(len < 2)
414 return GZIP_UNDERFLOW;
415
416 len -= 2;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100417 }
418
419 *headerlen = totallen - len;
420 return GZIP_OK;
421}
422#endif
423
Alex Deymo486467e2017-12-19 19:04:07 +0100424static CURLcode gzip_unencode_write(struct connectdata *conn,
425 contenc_writer *writer,
426 const char *buf, size_t nbytes)
Kristian Monsen5ab50182010-05-14 18:53:44 +0100427{
Alex Deymo486467e2017-12-19 19:04:07 +0100428 zlib_params *zp = (zlib_params *) &writer->params;
429 z_stream *z = &zp->z; /* zlib state structure */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100430
Alex Deymo486467e2017-12-19 19:04:07 +0100431 if(zp->zlib_init == ZLIB_INIT_GZIP) {
Kristian Monsen5ab50182010-05-14 18:53:44 +0100432 /* Let zlib handle the gzip decompression entirely */
Alex Deymo486467e2017-12-19 19:04:07 +0100433 z->next_in = (Bytef *) buf;
434 z->avail_in = (uInt) nbytes;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100435 /* Now uncompress the data */
Elliott Hughes0128fe42018-02-27 14:57:55 -0800436 return inflate_stream(conn, writer, ZLIB_INIT_GZIP);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100437 }
438
439#ifndef OLD_ZLIB_SUPPORT
440 /* Support for old zlib versions is compiled away and we are running with
441 an old version, so return an error. */
Alex Deymo486467e2017-12-19 19:04:07 +0100442 return exit_zlib(conn, z, &zp->zlib_init, CURLE_WRITE_ERROR);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100443
444#else
445 /* This next mess is to get around the potential case where there isn't
446 * enough data passed in to skip over the gzip header. If that happens, we
447 * malloc a block and copy what we have then wait for the next call. If
448 * there still isn't enough (this is definitely a worst-case scenario), we
449 * make the block bigger, copy the next part in and keep waiting.
450 *
451 * This is only required with zlib versions < 1.2.0.4 as newer versions
452 * can handle the gzip header themselves.
453 */
454
Alex Deymo486467e2017-12-19 19:04:07 +0100455 switch(zp->zlib_init) {
Kristian Monsen5ab50182010-05-14 18:53:44 +0100456 /* Skip over gzip header? */
457 case ZLIB_INIT:
458 {
459 /* Initial call state */
460 ssize_t hlen;
461
Alex Deymo486467e2017-12-19 19:04:07 +0100462 switch(check_gzip_header((unsigned char *) buf, nbytes, &hlen)) {
Kristian Monsen5ab50182010-05-14 18:53:44 +0100463 case GZIP_OK:
Alex Deymo486467e2017-12-19 19:04:07 +0100464 z->next_in = (Bytef *) buf + hlen;
465 z->avail_in = (uInt) (nbytes - hlen);
466 zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100467 break;
468
469 case GZIP_UNDERFLOW:
470 /* We need more data so we can find the end of the gzip header. It's
471 * possible that the memory block we malloc here will never be freed if
472 * the transfer abruptly aborts after this point. Since it's unlikely
473 * that circumstances will be right for this code path to be followed in
474 * the first place, and it's even more unlikely for a transfer to fail
475 * immediately afterwards, it should seldom be a problem.
476 */
Alex Deymo486467e2017-12-19 19:04:07 +0100477 z->avail_in = (uInt) nbytes;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100478 z->next_in = malloc(z->avail_in);
479 if(z->next_in == NULL) {
Alex Deymo486467e2017-12-19 19:04:07 +0100480 return exit_zlib(conn, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100481 }
Alex Deymo486467e2017-12-19 19:04:07 +0100482 memcpy(z->next_in, buf, z->avail_in);
483 zp->zlib_init = ZLIB_GZIP_HEADER; /* Need more gzip header data state */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100484 /* We don't have any data to inflate yet */
485 return CURLE_OK;
486
487 case GZIP_BAD:
488 default:
Alex Deymo486467e2017-12-19 19:04:07 +0100489 return exit_zlib(conn, z, &zp->zlib_init, process_zlib_error(conn, z));
Kristian Monsen5ab50182010-05-14 18:53:44 +0100490 }
491
492 }
493 break;
494
495 case ZLIB_GZIP_HEADER:
496 {
497 /* Need more gzip header data state */
498 ssize_t hlen;
Alex Deymo486467e2017-12-19 19:04:07 +0100499 z->avail_in += (uInt) nbytes;
Elliott Hughescee03382017-06-23 12:17:18 -0700500 z->next_in = Curl_saferealloc(z->next_in, z->avail_in);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100501 if(z->next_in == NULL) {
Alex Deymo486467e2017-12-19 19:04:07 +0100502 return exit_zlib(conn, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100503 }
504 /* Append the new block of data to the previous one */
Alex Deymo486467e2017-12-19 19:04:07 +0100505 memcpy(z->next_in + z->avail_in - nbytes, buf, nbytes);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100506
Elliott Hughes82be86d2017-09-20 17:00:17 -0700507 switch(check_gzip_header(z->next_in, z->avail_in, &hlen)) {
Kristian Monsen5ab50182010-05-14 18:53:44 +0100508 case GZIP_OK:
509 /* This is the zlib stream data */
510 free(z->next_in);
511 /* Don't point into the malloced block since we just freed it */
Alex Deymo486467e2017-12-19 19:04:07 +0100512 z->next_in = (Bytef *) buf + hlen + nbytes - z->avail_in;
513 z->avail_in = (uInt) (z->avail_in - hlen);
514 zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100515 break;
516
517 case GZIP_UNDERFLOW:
518 /* We still don't have any data to inflate! */
519 return CURLE_OK;
520
521 case GZIP_BAD:
522 default:
Alex Deymo486467e2017-12-19 19:04:07 +0100523 return exit_zlib(conn, z, &zp->zlib_init, process_zlib_error(conn, z));
Kristian Monsen5ab50182010-05-14 18:53:44 +0100524 }
525
526 }
527 break;
528
Elliott Hughes0128fe42018-02-27 14:57:55 -0800529 case ZLIB_GZIP_TRAILER:
530 z->next_in = (Bytef *) buf;
531 z->avail_in = (uInt) nbytes;
532 return process_trailer(conn, zp);
533
Kristian Monsen5ab50182010-05-14 18:53:44 +0100534 case ZLIB_GZIP_INFLATING:
535 default:
536 /* Inflating stream state */
Alex Deymo486467e2017-12-19 19:04:07 +0100537 z->next_in = (Bytef *) buf;
538 z->avail_in = (uInt) nbytes;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100539 break;
540 }
541
542 if(z->avail_in == 0) {
543 /* We don't have any data to inflate; wait until next time */
544 return CURLE_OK;
545 }
546
547 /* We've parsed the header, now uncompress the data */
Elliott Hughes0128fe42018-02-27 14:57:55 -0800548 return inflate_stream(conn, writer, ZLIB_GZIP_INFLATING);
Kristian Monsen5ab50182010-05-14 18:53:44 +0100549#endif
550}
551
Alex Deymo486467e2017-12-19 19:04:07 +0100552static void gzip_close_writer(struct connectdata *conn,
553 contenc_writer *writer)
554{
555 zlib_params *zp = (zlib_params *) &writer->params;
556 z_stream *z = &zp->z; /* zlib state structure */
557
558 exit_zlib(conn, z, &zp->zlib_init, CURLE_OK);
559}
560
561static const content_encoding gzip_encoding = {
562 "gzip",
563 "x-gzip",
564 gzip_init_writer,
565 gzip_unencode_write,
566 gzip_close_writer,
567 sizeof(zlib_params)
568};
569
570#endif /* HAVE_LIBZ */
571
572
573#ifdef HAVE_BROTLI
574
575/* Writer parameters. */
576typedef struct {
577 BrotliDecoderState *br; /* State structure for brotli. */
578} brotli_params;
579
580
581static CURLcode brotli_map_error(BrotliDecoderErrorCode be)
582{
583 switch(be) {
584 case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_NIBBLE:
585 case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_META_NIBBLE:
586 case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_ALPHABET:
587 case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_SAME:
588 case BROTLI_DECODER_ERROR_FORMAT_CL_SPACE:
589 case BROTLI_DECODER_ERROR_FORMAT_HUFFMAN_SPACE:
590 case BROTLI_DECODER_ERROR_FORMAT_CONTEXT_MAP_REPEAT:
591 case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_1:
592 case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_2:
593 case BROTLI_DECODER_ERROR_FORMAT_TRANSFORM:
594 case BROTLI_DECODER_ERROR_FORMAT_DICTIONARY:
595 case BROTLI_DECODER_ERROR_FORMAT_WINDOW_BITS:
596 case BROTLI_DECODER_ERROR_FORMAT_PADDING_1:
597 case BROTLI_DECODER_ERROR_FORMAT_PADDING_2:
Elliott Hughes0128fe42018-02-27 14:57:55 -0800598#ifdef BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY
Alex Deymo486467e2017-12-19 19:04:07 +0100599 case BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY:
Elliott Hughes0128fe42018-02-27 14:57:55 -0800600#endif
601#ifdef BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET
Alex Deymo486467e2017-12-19 19:04:07 +0100602 case BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET:
Elliott Hughes0128fe42018-02-27 14:57:55 -0800603#endif
Alex Deymo486467e2017-12-19 19:04:07 +0100604 case BROTLI_DECODER_ERROR_INVALID_ARGUMENTS:
605 return CURLE_BAD_CONTENT_ENCODING;
606 case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MODES:
607 case BROTLI_DECODER_ERROR_ALLOC_TREE_GROUPS:
608 case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MAP:
609 case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_1:
610 case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_2:
611 case BROTLI_DECODER_ERROR_ALLOC_BLOCK_TYPE_TREES:
612 return CURLE_OUT_OF_MEMORY;
613 default:
614 break;
615 }
616 return CURLE_WRITE_ERROR;
617}
618
619static CURLcode brotli_init_writer(struct connectdata *conn,
620 contenc_writer *writer)
621{
622 brotli_params *bp = (brotli_params *) &writer->params;
623
624 (void) conn;
625
626 if(!writer->downstream)
627 return CURLE_WRITE_ERROR;
628
629 bp->br = BrotliDecoderCreateInstance(NULL, NULL, NULL);
630 return bp->br? CURLE_OK: CURLE_OUT_OF_MEMORY;
631}
632
633static CURLcode brotli_unencode_write(struct connectdata *conn,
634 contenc_writer *writer,
635 const char *buf, size_t nbytes)
636{
637 brotli_params *bp = (brotli_params *) &writer->params;
638 const uint8_t *src = (const uint8_t *) buf;
639 char *decomp;
640 uint8_t *dst;
641 size_t dstleft;
642 CURLcode result = CURLE_OK;
Elliott Hughes0128fe42018-02-27 14:57:55 -0800643 BrotliDecoderResult r = BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT;
Alex Deymo486467e2017-12-19 19:04:07 +0100644
645 if(!bp->br)
646 return CURLE_WRITE_ERROR; /* Stream already ended. */
647
648 decomp = malloc(DSIZ);
649 if(!decomp)
650 return CURLE_OUT_OF_MEMORY;
651
Elliott Hughes0128fe42018-02-27 14:57:55 -0800652 while((nbytes || r == BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT) &&
653 result == CURLE_OK) {
Alex Deymo486467e2017-12-19 19:04:07 +0100654 dst = (uint8_t *) decomp;
655 dstleft = DSIZ;
656 r = BrotliDecoderDecompressStream(bp->br,
657 &nbytes, &src, &dstleft, &dst, NULL);
658 result = Curl_unencode_write(conn, writer->downstream,
659 decomp, DSIZ - dstleft);
660 if(result)
661 break;
662 switch(r) {
663 case BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT:
664 case BROTLI_DECODER_RESULT_NEEDS_MORE_INPUT:
665 break;
666 case BROTLI_DECODER_RESULT_SUCCESS:
667 BrotliDecoderDestroyInstance(bp->br);
668 bp->br = NULL;
669 if(nbytes)
670 result = CURLE_WRITE_ERROR;
671 break;
672 default:
673 result = brotli_map_error(BrotliDecoderGetErrorCode(bp->br));
674 break;
675 }
676 }
677 free(decomp);
678 return result;
679}
680
681static void brotli_close_writer(struct connectdata *conn,
682 contenc_writer *writer)
683{
684 brotli_params *bp = (brotli_params *) &writer->params;
685
686 (void) conn;
687
688 if(bp->br) {
689 BrotliDecoderDestroyInstance(bp->br);
690 bp->br = NULL;
691 }
692}
693
694static const content_encoding brotli_encoding = {
695 "br",
696 NULL,
697 brotli_init_writer,
698 brotli_unencode_write,
699 brotli_close_writer,
700 sizeof(brotli_params)
701};
702#endif
703
704
705/* Identity handler. */
706static CURLcode identity_init_writer(struct connectdata *conn,
707 contenc_writer *writer)
708{
709 (void) conn;
710 return writer->downstream? CURLE_OK: CURLE_WRITE_ERROR;
711}
712
713static CURLcode identity_unencode_write(struct connectdata *conn,
714 contenc_writer *writer,
715 const char *buf, size_t nbytes)
716{
717 return Curl_unencode_write(conn, writer->downstream, buf, nbytes);
718}
719
720static void identity_close_writer(struct connectdata *conn,
721 contenc_writer *writer)
722{
723 (void) conn;
724 (void) writer;
725}
726
727static const content_encoding identity_encoding = {
728 "identity",
Elliott Hughescac39802018-04-27 16:19:43 -0700729 "none",
Alex Deymo486467e2017-12-19 19:04:07 +0100730 identity_init_writer,
731 identity_unencode_write,
732 identity_close_writer,
733 0
734};
735
736
737/* supported content encodings table. */
738static const content_encoding * const encodings[] = {
739 &identity_encoding,
740#ifdef HAVE_LIBZ
741 &deflate_encoding,
742 &gzip_encoding,
743#endif
744#ifdef HAVE_BROTLI
745 &brotli_encoding,
746#endif
747 NULL
748};
749
750
751/* Return a list of comma-separated names of supported encodings. */
752char *Curl_all_content_encodings(void)
753{
754 size_t len = 0;
755 const content_encoding * const *cep;
756 const content_encoding *ce;
757 char *ace;
758 char *p;
759
760 for(cep = encodings; *cep; cep++) {
761 ce = *cep;
762 if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT))
763 len += strlen(ce->name) + 2;
764 }
765
766 if(!len)
767 return strdup(CONTENT_ENCODING_DEFAULT);
768
769 ace = malloc(len);
770 if(ace) {
771 p = ace;
772 for(cep = encodings; *cep; cep++) {
773 ce = *cep;
774 if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT)) {
775 strcpy(p, ce->name);
776 p += strlen(p);
777 *p++ = ',';
778 *p++ = ' ';
779 }
780 }
781 p[-2] = '\0';
782 }
783
784 return ace;
785}
786
787
788/* Real client writer: no downstream. */
789static CURLcode client_init_writer(struct connectdata *conn,
790 contenc_writer *writer)
791{
792 (void) conn;
793 return writer->downstream? CURLE_WRITE_ERROR: CURLE_OK;
794}
795
796static CURLcode client_unencode_write(struct connectdata *conn,
797 contenc_writer *writer,
798 const char *buf, size_t nbytes)
799{
800 struct Curl_easy *data = conn->data;
801 struct SingleRequest *k = &data->req;
802
803 (void) writer;
804
805 if(!nbytes || k->ignorebody)
806 return CURLE_OK;
807
808 return Curl_client_write(conn, CLIENTWRITE_BODY, (char *) buf, nbytes);
809}
810
811static void client_close_writer(struct connectdata *conn,
812 contenc_writer *writer)
813{
814 (void) conn;
815 (void) writer;
816}
817
818static const content_encoding client_encoding = {
819 NULL,
820 NULL,
821 client_init_writer,
822 client_unencode_write,
823 client_close_writer,
824 0
825};
826
827
828/* Deferred error dummy writer. */
829static CURLcode error_init_writer(struct connectdata *conn,
830 contenc_writer *writer)
831{
832 (void) conn;
833 return writer->downstream? CURLE_OK: CURLE_WRITE_ERROR;
834}
835
836static CURLcode error_unencode_write(struct connectdata *conn,
837 contenc_writer *writer,
838 const char *buf, size_t nbytes)
839{
840 char *all = Curl_all_content_encodings();
841
842 (void) writer;
843 (void) buf;
844 (void) nbytes;
845
846 if(!all)
847 return CURLE_OUT_OF_MEMORY;
848 failf(conn->data, "Unrecognized content encoding type. "
849 "libcurl understands %s content encodings.", all);
850 free(all);
851 return CURLE_BAD_CONTENT_ENCODING;
852}
853
854static void error_close_writer(struct connectdata *conn,
855 contenc_writer *writer)
856{
857 (void) conn;
858 (void) writer;
859}
860
861static const content_encoding error_encoding = {
862 NULL,
863 NULL,
864 error_init_writer,
865 error_unencode_write,
866 error_close_writer,
867 0
868};
869
870/* Create an unencoding writer stage using the given handler. */
871static contenc_writer *new_unencoding_writer(struct connectdata *conn,
872 const content_encoding *handler,
873 contenc_writer *downstream)
874{
875 size_t sz = offsetof(contenc_writer, params) + handler->paramsize;
Elliott Hughes1ef06ba2018-05-30 15:43:58 -0700876 contenc_writer *writer = (contenc_writer *) calloc(1, sz);
Alex Deymo486467e2017-12-19 19:04:07 +0100877
878 if(writer) {
Alex Deymo486467e2017-12-19 19:04:07 +0100879 writer->handler = handler;
880 writer->downstream = downstream;
881 if(handler->init_writer(conn, writer)) {
882 free(writer);
883 writer = NULL;
884 }
885 }
886
887 return writer;
888}
889
890/* Write data using an unencoding writer stack. */
891CURLcode Curl_unencode_write(struct connectdata *conn, contenc_writer *writer,
892 const char *buf, size_t nbytes)
893{
894 if(!nbytes)
895 return CURLE_OK;
896 return writer->handler->unencode_write(conn, writer, buf, nbytes);
897}
898
899/* Close and clean-up the connection's writer stack. */
Kristian Monsen5ab50182010-05-14 18:53:44 +0100900void Curl_unencode_cleanup(struct connectdata *conn)
901{
Alex Deymoe3149cc2016-10-05 11:18:42 -0700902 struct Curl_easy *data = conn->data;
Kristian Monsen5ab50182010-05-14 18:53:44 +0100903 struct SingleRequest *k = &data->req;
Alex Deymo486467e2017-12-19 19:04:07 +0100904 contenc_writer *writer = k->writer_stack;
905
906 while(writer) {
907 k->writer_stack = writer->downstream;
908 writer->handler->close_writer(conn, writer);
909 free(writer);
910 writer = k->writer_stack;
911 }
Kristian Monsen5ab50182010-05-14 18:53:44 +0100912}
913
Alex Deymo486467e2017-12-19 19:04:07 +0100914/* Find the content encoding by name. */
915static const content_encoding *find_encoding(const char *name, size_t len)
916{
917 const content_encoding * const *cep;
918 const content_encoding *ce;
919
920 for(cep = encodings; *cep; cep++) {
921 ce = *cep;
922 if((strncasecompare(name, ce->name, len) && !ce->name[len]) ||
923 (ce->alias && strncasecompare(name, ce->alias, len) && !ce->alias[len]))
924 return ce;
925 }
926 return NULL;
927}
928
929/* Set-up the unencoding stack from the Content-Encoding header value.
930 * See RFC 7231 section 3.1.2.2. */
931CURLcode Curl_build_unencoding_stack(struct connectdata *conn,
932 const char *enclist, int maybechunked)
933{
934 struct Curl_easy *data = conn->data;
935 struct SingleRequest *k = &data->req;
936
937 do {
938 const char *name;
939 size_t namelen;
940
941 /* Parse a single encoding name. */
942 while(ISSPACE(*enclist) || *enclist == ',')
943 enclist++;
944
945 name = enclist;
946
947 for(namelen = 0; *enclist && *enclist != ','; enclist++)
948 if(!ISSPACE(*enclist))
949 namelen = enclist - name + 1;
950
951 /* Special case: chunked encoding is handled at the reader level. */
952 if(maybechunked && namelen == 7 && strncasecompare(name, "chunked", 7)) {
953 k->chunk = TRUE; /* chunks coming our way. */
954 Curl_httpchunk_init(conn); /* init our chunky engine. */
955 }
956 else if(namelen) {
957 const content_encoding *encoding = find_encoding(name, namelen);
958 contenc_writer *writer;
959
960 if(!k->writer_stack) {
961 k->writer_stack = new_unencoding_writer(conn, &client_encoding, NULL);
962
963 if(!k->writer_stack)
964 return CURLE_OUT_OF_MEMORY;
965 }
966
967 if(!encoding)
968 encoding = &error_encoding; /* Defer error at stack use. */
969
970 /* Stack the unencoding stage. */
971 writer = new_unencoding_writer(conn, encoding, k->writer_stack);
972 if(!writer)
973 return CURLE_OUT_OF_MEMORY;
974 k->writer_stack = writer;
975 }
976 } while(*enclist);
977
978 return CURLE_OK;
979}
980
981#else
982/* Stubs for builds without HTTP. */
983CURLcode Curl_build_unencoding_stack(struct connectdata *conn,
984 const char *enclist, int maybechunked)
985{
986 (void) conn;
987 (void) enclist;
988 (void) maybechunked;
989 return CURLE_NOT_BUILT_IN;
990}
991
992CURLcode Curl_unencode_write(struct connectdata *conn, contenc_writer *writer,
993 const char *buf, size_t nbytes)
994{
995 (void) conn;
996 (void) writer;
997 (void) buf;
998 (void) nbytes;
999 return CURLE_NOT_BUILT_IN;
1000}
1001
1002void Curl_unencode_cleanup(struct connectdata *conn)
1003{
1004 (void) conn;
1005}
1006
1007char *Curl_all_content_encodings(void)
1008{
1009 return strdup(CONTENT_ENCODING_DEFAULT); /* Satisfy caller. */
1010}
1011
1012#endif /* CURL_DISABLE_HTTP */