| 1 | /*************************************************************************** |
| 2 | * _ _ ____ _ |
| 3 | * Project ___| | | | _ \| | |
| 4 | * / __| | | | |_) | | |
| 5 | * | (__| |_| | _ <| |___ |
| 6 | * \___|\___/|_| \_\_____| |
| 7 | * |
| 8 | * Copyright (C) 1998 - 2018, 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.haxx.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 | ***************************************************************************/ |
| 22 | |
| 23 | #include "curl_setup.h" |
| 24 | |
| 25 | #include "urldata.h" |
| 26 | #include <curl/curl.h> |
| 27 | #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 | |
| 41 | #include "sendf.h" |
| 42 | #include "http.h" |
| 43 | #include "content_encoding.h" |
| 44 | #include "strdup.h" |
| 45 | #include "strcase.h" |
| 46 | #include "curl_memory.h" |
| 47 | #include "memdebug.h" |
| 48 | |
| 49 | #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 | |
| 58 | /* 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 | |
| 62 | #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 0x04 /* bit 2 set: extra field present */ |
| 69 | #define ORIG_NAME 0x08 /* bit 3 set: original file name present */ |
| 70 | #define 0x10 /* bit 4 set: file comment present */ |
| 71 | #define RESERVED 0xE0 /* bits 5..7: reserved */ |
| 72 | |
| 73 | typedef enum { |
| 74 | ZLIB_UNINIT, /* uninitialized */ |
| 75 | ZLIB_INIT, /* initialized */ |
| 76 | ZLIB_INFLATING, /* inflating started. */ |
| 77 | ZLIB_EXTERNAL_TRAILER, /* reading external trailer */ |
| 78 | , /* reading gzip header */ |
| 79 | ZLIB_GZIP_INFLATING, /* inflating gzip stream */ |
| 80 | ZLIB_INIT_GZIP /* initialized in transparent gzip mode */ |
| 81 | } zlibInitState; |
| 82 | |
| 83 | /* Writer parameters. */ |
| 84 | typedef struct { |
| 85 | zlibInitState zlib_init; /* zlib init state */ |
| 86 | uInt trailerlen; /* Remaining trailer byte count. */ |
| 87 | z_stream z; /* State structure for zlib. */ |
| 88 | } zlib_params; |
| 89 | |
| 90 | |
| 91 | static voidpf |
| 92 | zalloc_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 | |
| 99 | static void |
| 100 | zfree_cb(voidpf opaque, voidpf ptr) |
| 101 | { |
| 102 | (void) opaque; |
| 103 | free(ptr); |
| 104 | } |
| 105 | |
| 106 | static CURLcode |
| 107 | process_zlib_error(struct connectdata *conn, z_stream *z) |
| 108 | { |
| 109 | struct Curl_easy *data = conn->data; |
| 110 | if(z->msg) |
| 111 | failf(data, "Error while processing content unencoding: %s" , |
| 112 | z->msg); |
| 113 | else |
| 114 | failf(data, "Error while processing content unencoding: " |
| 115 | "Unknown failure within decompression software." ); |
| 116 | |
| 117 | return CURLE_BAD_CONTENT_ENCODING; |
| 118 | } |
| 119 | |
| 120 | static CURLcode |
| 121 | exit_zlib(struct connectdata *conn, |
| 122 | z_stream *z, zlibInitState *zlib_init, CURLcode result) |
| 123 | { |
| 124 | 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 | |
| 133 | return result; |
| 134 | } |
| 135 | |
| 136 | static 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 or raw deflate. */ |
| 154 | zp->zlib_init = ZLIB_EXTERNAL_TRAILER; |
| 155 | } |
| 156 | return result; |
| 157 | } |
| 158 | |
| 159 | static CURLcode inflate_stream(struct connectdata *conn, |
| 160 | contenc_writer *writer, zlibInitState started) |
| 161 | { |
| 162 | zlib_params *zp = (zlib_params *) &writer->params; |
| 163 | z_stream *z = &zp->z; /* zlib state structure */ |
| 164 | uInt nread = z->avail_in; |
| 165 | Bytef *orig_in = z->next_in; |
| 166 | bool done = FALSE; |
| 167 | CURLcode result = CURLE_OK; /* Curl_client_write status */ |
| 168 | char *decomp; /* Put the decompressed data here. */ |
| 169 | |
| 170 | /* Check state. */ |
| 171 | if(zp->zlib_init != ZLIB_INIT && |
| 172 | zp->zlib_init != ZLIB_INFLATING && |
| 173 | zp->zlib_init != ZLIB_INIT_GZIP && |
| 174 | zp->zlib_init != ZLIB_GZIP_INFLATING) |
| 175 | return exit_zlib(conn, z, &zp->zlib_init, CURLE_WRITE_ERROR); |
| 176 | |
| 177 | /* Dynamically allocate a buffer for decompression because it's uncommonly |
| 178 | large to hold on the stack */ |
| 179 | decomp = malloc(DSIZ); |
| 180 | if(decomp == NULL) |
| 181 | return exit_zlib(conn, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY); |
| 182 | |
| 183 | /* because the buffer size is fixed, iteratively decompress and transfer to |
| 184 | the client via downstream_write function. */ |
| 185 | while(!done) { |
| 186 | int status; /* zlib status */ |
| 187 | done = TRUE; |
| 188 | |
| 189 | /* (re)set buffer for decompressed output for every iteration */ |
| 190 | z->next_out = (Bytef *) decomp; |
| 191 | z->avail_out = DSIZ; |
| 192 | |
| 193 | #ifdef Z_BLOCK |
| 194 | /* Z_BLOCK is only available in zlib ver. >= 1.2.0.5 */ |
| 195 | status = inflate(z, Z_BLOCK); |
| 196 | #else |
| 197 | /* fallback for zlib ver. < 1.2.0.5 */ |
| 198 | status = inflate(z, Z_SYNC_FLUSH); |
| 199 | #endif |
| 200 | |
| 201 | /* Flush output data if some. */ |
| 202 | if(z->avail_out != DSIZ) { |
| 203 | if(status == Z_OK || status == Z_STREAM_END) { |
| 204 | zp->zlib_init = started; /* Data started. */ |
| 205 | result = Curl_unencode_write(conn, writer->downstream, decomp, |
| 206 | DSIZ - z->avail_out); |
| 207 | if(result) { |
| 208 | exit_zlib(conn, z, &zp->zlib_init, result); |
| 209 | break; |
| 210 | } |
| 211 | } |
| 212 | } |
| 213 | |
| 214 | /* Dispatch by inflate() status. */ |
| 215 | switch(status) { |
| 216 | case Z_OK: |
| 217 | /* Always loop: there may be unflushed latched data in zlib state. */ |
| 218 | done = FALSE; |
| 219 | break; |
| 220 | case Z_BUF_ERROR: |
| 221 | /* No more data to flush: just exit loop. */ |
| 222 | break; |
| 223 | case Z_STREAM_END: |
| 224 | result = process_trailer(conn, zp); |
| 225 | break; |
| 226 | case Z_DATA_ERROR: |
| 227 | /* some servers seem to not generate zlib headers, so this is an attempt |
| 228 | to fix and continue anyway */ |
| 229 | if(zp->zlib_init == ZLIB_INIT) { |
| 230 | /* Do not use inflateReset2(): only available since zlib 1.2.3.4. */ |
| 231 | (void) inflateEnd(z); /* don't care about the return code */ |
| 232 | if(inflateInit2(z, -MAX_WBITS) == Z_OK) { |
| 233 | z->next_in = orig_in; |
| 234 | z->avail_in = nread; |
| 235 | zp->zlib_init = ZLIB_INFLATING; |
| 236 | zp->trailerlen = 4; /* Tolerate up to 4 unknown trailer bytes. */ |
| 237 | done = FALSE; |
| 238 | break; |
| 239 | } |
| 240 | zp->zlib_init = ZLIB_UNINIT; /* inflateEnd() already called. */ |
| 241 | } |
| 242 | /* FALLTHROUGH */ |
| 243 | default: |
| 244 | result = exit_zlib(conn, z, &zp->zlib_init, process_zlib_error(conn, z)); |
| 245 | break; |
| 246 | } |
| 247 | } |
| 248 | free(decomp); |
| 249 | |
| 250 | /* We're about to leave this call so the `nread' data bytes won't be seen |
| 251 | again. If we are in a state that would wrongly allow restart in raw mode |
| 252 | at the next call, assume output has already started. */ |
| 253 | if(nread && zp->zlib_init == ZLIB_INIT) |
| 254 | zp->zlib_init = started; /* Cannot restart anymore. */ |
| 255 | |
| 256 | return result; |
| 257 | } |
| 258 | |
| 259 | |
| 260 | /* Deflate handler. */ |
| 261 | static CURLcode deflate_init_writer(struct connectdata *conn, |
| 262 | contenc_writer *writer) |
| 263 | { |
| 264 | zlib_params *zp = (zlib_params *) &writer->params; |
| 265 | z_stream *z = &zp->z; /* zlib state structure */ |
| 266 | |
| 267 | if(!writer->downstream) |
| 268 | return CURLE_WRITE_ERROR; |
| 269 | |
| 270 | /* Initialize zlib */ |
| 271 | z->zalloc = (alloc_func) zalloc_cb; |
| 272 | z->zfree = (free_func) zfree_cb; |
| 273 | |
| 274 | if(inflateInit(z) != Z_OK) |
| 275 | return process_zlib_error(conn, z); |
| 276 | zp->zlib_init = ZLIB_INIT; |
| 277 | return CURLE_OK; |
| 278 | } |
| 279 | |
| 280 | static CURLcode deflate_unencode_write(struct connectdata *conn, |
| 281 | contenc_writer *writer, |
| 282 | const char *buf, size_t nbytes) |
| 283 | { |
| 284 | zlib_params *zp = (zlib_params *) &writer->params; |
| 285 | z_stream *z = &zp->z; /* zlib state structure */ |
| 286 | |
| 287 | /* Set the compressed input when this function is called */ |
| 288 | z->next_in = (Bytef *) buf; |
| 289 | z->avail_in = (uInt) nbytes; |
| 290 | |
| 291 | if(zp->zlib_init == ZLIB_EXTERNAL_TRAILER) |
| 292 | return process_trailer(conn, zp); |
| 293 | |
| 294 | /* Now uncompress the data */ |
| 295 | return inflate_stream(conn, writer, ZLIB_INFLATING); |
| 296 | } |
| 297 | |
| 298 | static void deflate_close_writer(struct connectdata *conn, |
| 299 | contenc_writer *writer) |
| 300 | { |
| 301 | zlib_params *zp = (zlib_params *) &writer->params; |
| 302 | z_stream *z = &zp->z; /* zlib state structure */ |
| 303 | |
| 304 | exit_zlib(conn, z, &zp->zlib_init, CURLE_OK); |
| 305 | } |
| 306 | |
| 307 | static const content_encoding deflate_encoding = { |
| 308 | "deflate" , |
| 309 | NULL, |
| 310 | deflate_init_writer, |
| 311 | deflate_unencode_write, |
| 312 | deflate_close_writer, |
| 313 | sizeof(zlib_params) |
| 314 | }; |
| 315 | |
| 316 | |
| 317 | /* Gzip handler. */ |
| 318 | static CURLcode gzip_init_writer(struct connectdata *conn, |
| 319 | contenc_writer *writer) |
| 320 | { |
| 321 | zlib_params *zp = (zlib_params *) &writer->params; |
| 322 | z_stream *z = &zp->z; /* zlib state structure */ |
| 323 | |
| 324 | if(!writer->downstream) |
| 325 | return CURLE_WRITE_ERROR; |
| 326 | |
| 327 | /* Initialize zlib */ |
| 328 | z->zalloc = (alloc_func) zalloc_cb; |
| 329 | z->zfree = (free_func) zfree_cb; |
| 330 | |
| 331 | if(strcmp(zlibVersion(), "1.2.0.4" ) >= 0) { |
| 332 | /* zlib ver. >= 1.2.0.4 supports transparent gzip decompressing */ |
| 333 | if(inflateInit2(z, MAX_WBITS + 32) != Z_OK) { |
| 334 | return process_zlib_error(conn, z); |
| 335 | } |
| 336 | zp->zlib_init = ZLIB_INIT_GZIP; /* Transparent gzip decompress state */ |
| 337 | } |
| 338 | else { |
| 339 | /* we must parse the gzip header and trailer ourselves */ |
| 340 | if(inflateInit2(z, -MAX_WBITS) != Z_OK) { |
| 341 | return process_zlib_error(conn, z); |
| 342 | } |
| 343 | zp->trailerlen = 8; /* A CRC-32 and a 32-bit input size (RFC 1952, 2.2) */ |
| 344 | zp->zlib_init = ZLIB_INIT; /* Initial call state */ |
| 345 | } |
| 346 | |
| 347 | return CURLE_OK; |
| 348 | } |
| 349 | |
| 350 | #ifdef OLD_ZLIB_SUPPORT |
| 351 | /* Skip over the gzip header */ |
| 352 | static enum { |
| 353 | GZIP_OK, |
| 354 | GZIP_BAD, |
| 355 | GZIP_UNDERFLOW |
| 356 | } (unsigned char const *data, ssize_t len, ssize_t *) |
| 357 | { |
| 358 | int method, flags; |
| 359 | const ssize_t totallen = len; |
| 360 | |
| 361 | /* The shortest header is 10 bytes */ |
| 362 | if(len < 10) |
| 363 | return GZIP_UNDERFLOW; |
| 364 | |
| 365 | if((data[0] != GZIP_MAGIC_0) || (data[1] != GZIP_MAGIC_1)) |
| 366 | return GZIP_BAD; |
| 367 | |
| 368 | method = data[2]; |
| 369 | flags = data[3]; |
| 370 | |
| 371 | if(method != Z_DEFLATED || (flags & RESERVED) != 0) { |
| 372 | /* Can't handle this compression method or unknown flag */ |
| 373 | return GZIP_BAD; |
| 374 | } |
| 375 | |
| 376 | /* Skip over time, xflags, OS code and all previous bytes */ |
| 377 | len -= 10; |
| 378 | data += 10; |
| 379 | |
| 380 | if(flags & EXTRA_FIELD) { |
| 381 | ssize_t ; |
| 382 | |
| 383 | if(len < 2) |
| 384 | return GZIP_UNDERFLOW; |
| 385 | |
| 386 | extra_len = (data[1] << 8) | data[0]; |
| 387 | |
| 388 | if(len < (extra_len + 2)) |
| 389 | return GZIP_UNDERFLOW; |
| 390 | |
| 391 | len -= (extra_len + 2); |
| 392 | data += (extra_len + 2); |
| 393 | } |
| 394 | |
| 395 | if(flags & ORIG_NAME) { |
| 396 | /* Skip over NUL-terminated file name */ |
| 397 | while(len && *data) { |
| 398 | --len; |
| 399 | ++data; |
| 400 | } |
| 401 | if(!len || *data) |
| 402 | return GZIP_UNDERFLOW; |
| 403 | |
| 404 | /* Skip over the NUL */ |
| 405 | --len; |
| 406 | ++data; |
| 407 | } |
| 408 | |
| 409 | if(flags & COMMENT) { |
| 410 | /* Skip over NUL-terminated comment */ |
| 411 | while(len && *data) { |
| 412 | --len; |
| 413 | ++data; |
| 414 | } |
| 415 | if(!len || *data) |
| 416 | return GZIP_UNDERFLOW; |
| 417 | |
| 418 | /* Skip over the NUL */ |
| 419 | --len; |
| 420 | } |
| 421 | |
| 422 | if(flags & HEAD_CRC) { |
| 423 | if(len < 2) |
| 424 | return GZIP_UNDERFLOW; |
| 425 | |
| 426 | len -= 2; |
| 427 | } |
| 428 | |
| 429 | *headerlen = totallen - len; |
| 430 | return GZIP_OK; |
| 431 | } |
| 432 | #endif |
| 433 | |
| 434 | static CURLcode gzip_unencode_write(struct connectdata *conn, |
| 435 | contenc_writer *writer, |
| 436 | const char *buf, size_t nbytes) |
| 437 | { |
| 438 | zlib_params *zp = (zlib_params *) &writer->params; |
| 439 | z_stream *z = &zp->z; /* zlib state structure */ |
| 440 | |
| 441 | if(zp->zlib_init == ZLIB_INIT_GZIP) { |
| 442 | /* Let zlib handle the gzip decompression entirely */ |
| 443 | z->next_in = (Bytef *) buf; |
| 444 | z->avail_in = (uInt) nbytes; |
| 445 | /* Now uncompress the data */ |
| 446 | return inflate_stream(conn, writer, ZLIB_INIT_GZIP); |
| 447 | } |
| 448 | |
| 449 | #ifndef OLD_ZLIB_SUPPORT |
| 450 | /* Support for old zlib versions is compiled away and we are running with |
| 451 | an old version, so return an error. */ |
| 452 | return exit_zlib(conn, z, &zp->zlib_init, CURLE_WRITE_ERROR); |
| 453 | |
| 454 | #else |
| 455 | /* This next mess is to get around the potential case where there isn't |
| 456 | * enough data passed in to skip over the gzip header. If that happens, we |
| 457 | * malloc a block and copy what we have then wait for the next call. If |
| 458 | * there still isn't enough (this is definitely a worst-case scenario), we |
| 459 | * make the block bigger, copy the next part in and keep waiting. |
| 460 | * |
| 461 | * This is only required with zlib versions < 1.2.0.4 as newer versions |
| 462 | * can handle the gzip header themselves. |
| 463 | */ |
| 464 | |
| 465 | switch(zp->zlib_init) { |
| 466 | /* Skip over gzip header? */ |
| 467 | case ZLIB_INIT: |
| 468 | { |
| 469 | /* Initial call state */ |
| 470 | ssize_t hlen; |
| 471 | |
| 472 | switch(check_gzip_header((unsigned char *) buf, nbytes, &hlen)) { |
| 473 | case GZIP_OK: |
| 474 | z->next_in = (Bytef *) buf + hlen; |
| 475 | z->avail_in = (uInt) (nbytes - hlen); |
| 476 | zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */ |
| 477 | break; |
| 478 | |
| 479 | case GZIP_UNDERFLOW: |
| 480 | /* We need more data so we can find the end of the gzip header. It's |
| 481 | * possible that the memory block we malloc here will never be freed if |
| 482 | * the transfer abruptly aborts after this point. Since it's unlikely |
| 483 | * that circumstances will be right for this code path to be followed in |
| 484 | * the first place, and it's even more unlikely for a transfer to fail |
| 485 | * immediately afterwards, it should seldom be a problem. |
| 486 | */ |
| 487 | z->avail_in = (uInt) nbytes; |
| 488 | z->next_in = malloc(z->avail_in); |
| 489 | if(z->next_in == NULL) { |
| 490 | return exit_zlib(conn, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY); |
| 491 | } |
| 492 | memcpy(z->next_in, buf, z->avail_in); |
| 493 | zp->zlib_init = ZLIB_GZIP_HEADER; /* Need more gzip header data state */ |
| 494 | /* We don't have any data to inflate yet */ |
| 495 | return CURLE_OK; |
| 496 | |
| 497 | case GZIP_BAD: |
| 498 | default: |
| 499 | return exit_zlib(conn, z, &zp->zlib_init, process_zlib_error(conn, z)); |
| 500 | } |
| 501 | |
| 502 | } |
| 503 | break; |
| 504 | |
| 505 | case ZLIB_GZIP_HEADER: |
| 506 | { |
| 507 | /* Need more gzip header data state */ |
| 508 | ssize_t hlen; |
| 509 | z->avail_in += (uInt) nbytes; |
| 510 | z->next_in = Curl_saferealloc(z->next_in, z->avail_in); |
| 511 | if(z->next_in == NULL) { |
| 512 | return exit_zlib(conn, z, &zp->zlib_init, CURLE_OUT_OF_MEMORY); |
| 513 | } |
| 514 | /* Append the new block of data to the previous one */ |
| 515 | memcpy(z->next_in + z->avail_in - nbytes, buf, nbytes); |
| 516 | |
| 517 | switch(check_gzip_header(z->next_in, z->avail_in, &hlen)) { |
| 518 | case GZIP_OK: |
| 519 | /* This is the zlib stream data */ |
| 520 | free(z->next_in); |
| 521 | /* Don't point into the malloced block since we just freed it */ |
| 522 | z->next_in = (Bytef *) buf + hlen + nbytes - z->avail_in; |
| 523 | z->avail_in = (uInt) (z->avail_in - hlen); |
| 524 | zp->zlib_init = ZLIB_GZIP_INFLATING; /* Inflating stream state */ |
| 525 | break; |
| 526 | |
| 527 | case GZIP_UNDERFLOW: |
| 528 | /* We still don't have any data to inflate! */ |
| 529 | return CURLE_OK; |
| 530 | |
| 531 | case GZIP_BAD: |
| 532 | default: |
| 533 | return exit_zlib(conn, z, &zp->zlib_init, process_zlib_error(conn, z)); |
| 534 | } |
| 535 | |
| 536 | } |
| 537 | break; |
| 538 | |
| 539 | case ZLIB_EXTERNAL_TRAILER: |
| 540 | z->next_in = (Bytef *) buf; |
| 541 | z->avail_in = (uInt) nbytes; |
| 542 | return process_trailer(conn, zp); |
| 543 | |
| 544 | case ZLIB_GZIP_INFLATING: |
| 545 | default: |
| 546 | /* Inflating stream state */ |
| 547 | z->next_in = (Bytef *) buf; |
| 548 | z->avail_in = (uInt) nbytes; |
| 549 | break; |
| 550 | } |
| 551 | |
| 552 | if(z->avail_in == 0) { |
| 553 | /* We don't have any data to inflate; wait until next time */ |
| 554 | return CURLE_OK; |
| 555 | } |
| 556 | |
| 557 | /* We've parsed the header, now uncompress the data */ |
| 558 | return inflate_stream(conn, writer, ZLIB_GZIP_INFLATING); |
| 559 | #endif |
| 560 | } |
| 561 | |
| 562 | static void gzip_close_writer(struct connectdata *conn, |
| 563 | contenc_writer *writer) |
| 564 | { |
| 565 | zlib_params *zp = (zlib_params *) &writer->params; |
| 566 | z_stream *z = &zp->z; /* zlib state structure */ |
| 567 | |
| 568 | exit_zlib(conn, z, &zp->zlib_init, CURLE_OK); |
| 569 | } |
| 570 | |
| 571 | static const content_encoding gzip_encoding = { |
| 572 | "gzip" , |
| 573 | "x-gzip" , |
| 574 | gzip_init_writer, |
| 575 | gzip_unencode_write, |
| 576 | gzip_close_writer, |
| 577 | sizeof(zlib_params) |
| 578 | }; |
| 579 | |
| 580 | #endif /* HAVE_LIBZ */ |
| 581 | |
| 582 | |
| 583 | #ifdef HAVE_BROTLI |
| 584 | |
| 585 | /* Writer parameters. */ |
| 586 | typedef struct { |
| 587 | BrotliDecoderState *br; /* State structure for brotli. */ |
| 588 | } brotli_params; |
| 589 | |
| 590 | |
| 591 | static CURLcode brotli_map_error(BrotliDecoderErrorCode be) |
| 592 | { |
| 593 | switch(be) { |
| 594 | case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_NIBBLE: |
| 595 | case BROTLI_DECODER_ERROR_FORMAT_EXUBERANT_META_NIBBLE: |
| 596 | case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_ALPHABET: |
| 597 | case BROTLI_DECODER_ERROR_FORMAT_SIMPLE_HUFFMAN_SAME: |
| 598 | case BROTLI_DECODER_ERROR_FORMAT_CL_SPACE: |
| 599 | case BROTLI_DECODER_ERROR_FORMAT_HUFFMAN_SPACE: |
| 600 | case BROTLI_DECODER_ERROR_FORMAT_CONTEXT_MAP_REPEAT: |
| 601 | case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_1: |
| 602 | case BROTLI_DECODER_ERROR_FORMAT_BLOCK_LENGTH_2: |
| 603 | case BROTLI_DECODER_ERROR_FORMAT_TRANSFORM: |
| 604 | case BROTLI_DECODER_ERROR_FORMAT_DICTIONARY: |
| 605 | case BROTLI_DECODER_ERROR_FORMAT_WINDOW_BITS: |
| 606 | case BROTLI_DECODER_ERROR_FORMAT_PADDING_1: |
| 607 | case BROTLI_DECODER_ERROR_FORMAT_PADDING_2: |
| 608 | #ifdef BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY |
| 609 | case BROTLI_DECODER_ERROR_COMPOUND_DICTIONARY: |
| 610 | #endif |
| 611 | #ifdef BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET |
| 612 | case BROTLI_DECODER_ERROR_DICTIONARY_NOT_SET: |
| 613 | #endif |
| 614 | case BROTLI_DECODER_ERROR_INVALID_ARGUMENTS: |
| 615 | return CURLE_BAD_CONTENT_ENCODING; |
| 616 | case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MODES: |
| 617 | case BROTLI_DECODER_ERROR_ALLOC_TREE_GROUPS: |
| 618 | case BROTLI_DECODER_ERROR_ALLOC_CONTEXT_MAP: |
| 619 | case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_1: |
| 620 | case BROTLI_DECODER_ERROR_ALLOC_RING_BUFFER_2: |
| 621 | case BROTLI_DECODER_ERROR_ALLOC_BLOCK_TYPE_TREES: |
| 622 | return CURLE_OUT_OF_MEMORY; |
| 623 | default: |
| 624 | break; |
| 625 | } |
| 626 | return CURLE_WRITE_ERROR; |
| 627 | } |
| 628 | |
| 629 | static CURLcode brotli_init_writer(struct connectdata *conn, |
| 630 | contenc_writer *writer) |
| 631 | { |
| 632 | brotli_params *bp = (brotli_params *) &writer->params; |
| 633 | |
| 634 | (void) conn; |
| 635 | |
| 636 | if(!writer->downstream) |
| 637 | return CURLE_WRITE_ERROR; |
| 638 | |
| 639 | bp->br = BrotliDecoderCreateInstance(NULL, NULL, NULL); |
| 640 | return bp->br? CURLE_OK: CURLE_OUT_OF_MEMORY; |
| 641 | } |
| 642 | |
| 643 | static CURLcode brotli_unencode_write(struct connectdata *conn, |
| 644 | contenc_writer *writer, |
| 645 | const char *buf, size_t nbytes) |
| 646 | { |
| 647 | brotli_params *bp = (brotli_params *) &writer->params; |
| 648 | const uint8_t *src = (const uint8_t *) buf; |
| 649 | char *decomp; |
| 650 | uint8_t *dst; |
| 651 | size_t dstleft; |
| 652 | CURLcode result = CURLE_OK; |
| 653 | BrotliDecoderResult r = BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT; |
| 654 | |
| 655 | if(!bp->br) |
| 656 | return CURLE_WRITE_ERROR; /* Stream already ended. */ |
| 657 | |
| 658 | decomp = malloc(DSIZ); |
| 659 | if(!decomp) |
| 660 | return CURLE_OUT_OF_MEMORY; |
| 661 | |
| 662 | while((nbytes || r == BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT) && |
| 663 | result == CURLE_OK) { |
| 664 | dst = (uint8_t *) decomp; |
| 665 | dstleft = DSIZ; |
| 666 | r = BrotliDecoderDecompressStream(bp->br, |
| 667 | &nbytes, &src, &dstleft, &dst, NULL); |
| 668 | result = Curl_unencode_write(conn, writer->downstream, |
| 669 | decomp, DSIZ - dstleft); |
| 670 | if(result) |
| 671 | break; |
| 672 | switch(r) { |
| 673 | case BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT: |
| 674 | case BROTLI_DECODER_RESULT_NEEDS_MORE_INPUT: |
| 675 | break; |
| 676 | case BROTLI_DECODER_RESULT_SUCCESS: |
| 677 | BrotliDecoderDestroyInstance(bp->br); |
| 678 | bp->br = NULL; |
| 679 | if(nbytes) |
| 680 | result = CURLE_WRITE_ERROR; |
| 681 | break; |
| 682 | default: |
| 683 | result = brotli_map_error(BrotliDecoderGetErrorCode(bp->br)); |
| 684 | break; |
| 685 | } |
| 686 | } |
| 687 | free(decomp); |
| 688 | return result; |
| 689 | } |
| 690 | |
| 691 | static void brotli_close_writer(struct connectdata *conn, |
| 692 | contenc_writer *writer) |
| 693 | { |
| 694 | brotli_params *bp = (brotli_params *) &writer->params; |
| 695 | |
| 696 | (void) conn; |
| 697 | |
| 698 | if(bp->br) { |
| 699 | BrotliDecoderDestroyInstance(bp->br); |
| 700 | bp->br = NULL; |
| 701 | } |
| 702 | } |
| 703 | |
| 704 | static const content_encoding brotli_encoding = { |
| 705 | "br" , |
| 706 | NULL, |
| 707 | brotli_init_writer, |
| 708 | brotli_unencode_write, |
| 709 | brotli_close_writer, |
| 710 | sizeof(brotli_params) |
| 711 | }; |
| 712 | #endif |
| 713 | |
| 714 | |
| 715 | /* Identity handler. */ |
| 716 | static CURLcode identity_init_writer(struct connectdata *conn, |
| 717 | contenc_writer *writer) |
| 718 | { |
| 719 | (void) conn; |
| 720 | return writer->downstream? CURLE_OK: CURLE_WRITE_ERROR; |
| 721 | } |
| 722 | |
| 723 | static CURLcode identity_unencode_write(struct connectdata *conn, |
| 724 | contenc_writer *writer, |
| 725 | const char *buf, size_t nbytes) |
| 726 | { |
| 727 | return Curl_unencode_write(conn, writer->downstream, buf, nbytes); |
| 728 | } |
| 729 | |
| 730 | static void identity_close_writer(struct connectdata *conn, |
| 731 | contenc_writer *writer) |
| 732 | { |
| 733 | (void) conn; |
| 734 | (void) writer; |
| 735 | } |
| 736 | |
| 737 | static const content_encoding identity_encoding = { |
| 738 | "identity" , |
| 739 | "none" , |
| 740 | identity_init_writer, |
| 741 | identity_unencode_write, |
| 742 | identity_close_writer, |
| 743 | 0 |
| 744 | }; |
| 745 | |
| 746 | |
| 747 | /* supported content encodings table. */ |
| 748 | static const content_encoding * const encodings[] = { |
| 749 | &identity_encoding, |
| 750 | #ifdef HAVE_LIBZ |
| 751 | &deflate_encoding, |
| 752 | &gzip_encoding, |
| 753 | #endif |
| 754 | #ifdef HAVE_BROTLI |
| 755 | &brotli_encoding, |
| 756 | #endif |
| 757 | NULL |
| 758 | }; |
| 759 | |
| 760 | |
| 761 | /* Return a list of comma-separated names of supported encodings. */ |
| 762 | char *Curl_all_content_encodings(void) |
| 763 | { |
| 764 | size_t len = 0; |
| 765 | const content_encoding * const *cep; |
| 766 | const content_encoding *ce; |
| 767 | char *ace; |
| 768 | |
| 769 | for(cep = encodings; *cep; cep++) { |
| 770 | ce = *cep; |
| 771 | if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT)) |
| 772 | len += strlen(ce->name) + 2; |
| 773 | } |
| 774 | |
| 775 | if(!len) |
| 776 | return strdup(CONTENT_ENCODING_DEFAULT); |
| 777 | |
| 778 | ace = malloc(len); |
| 779 | if(ace) { |
| 780 | char *p = ace; |
| 781 | for(cep = encodings; *cep; cep++) { |
| 782 | ce = *cep; |
| 783 | if(!strcasecompare(ce->name, CONTENT_ENCODING_DEFAULT)) { |
| 784 | strcpy(p, ce->name); |
| 785 | p += strlen(p); |
| 786 | *p++ = ','; |
| 787 | *p++ = ' '; |
| 788 | } |
| 789 | } |
| 790 | p[-2] = '\0'; |
| 791 | } |
| 792 | |
| 793 | return ace; |
| 794 | } |
| 795 | |
| 796 | |
| 797 | /* Real client writer: no downstream. */ |
| 798 | static CURLcode client_init_writer(struct connectdata *conn, |
| 799 | contenc_writer *writer) |
| 800 | { |
| 801 | (void) conn; |
| 802 | return writer->downstream? CURLE_WRITE_ERROR: CURLE_OK; |
| 803 | } |
| 804 | |
| 805 | static CURLcode client_unencode_write(struct connectdata *conn, |
| 806 | contenc_writer *writer, |
| 807 | const char *buf, size_t nbytes) |
| 808 | { |
| 809 | struct Curl_easy *data = conn->data; |
| 810 | struct SingleRequest *k = &data->req; |
| 811 | |
| 812 | (void) writer; |
| 813 | |
| 814 | if(!nbytes || k->ignorebody) |
| 815 | return CURLE_OK; |
| 816 | |
| 817 | return Curl_client_write(conn, CLIENTWRITE_BODY, (char *) buf, nbytes); |
| 818 | } |
| 819 | |
| 820 | static void client_close_writer(struct connectdata *conn, |
| 821 | contenc_writer *writer) |
| 822 | { |
| 823 | (void) conn; |
| 824 | (void) writer; |
| 825 | } |
| 826 | |
| 827 | static const content_encoding client_encoding = { |
| 828 | NULL, |
| 829 | NULL, |
| 830 | client_init_writer, |
| 831 | client_unencode_write, |
| 832 | client_close_writer, |
| 833 | 0 |
| 834 | }; |
| 835 | |
| 836 | |
| 837 | /* Deferred error dummy writer. */ |
| 838 | static CURLcode error_init_writer(struct connectdata *conn, |
| 839 | contenc_writer *writer) |
| 840 | { |
| 841 | (void) conn; |
| 842 | return writer->downstream? CURLE_OK: CURLE_WRITE_ERROR; |
| 843 | } |
| 844 | |
| 845 | static CURLcode error_unencode_write(struct connectdata *conn, |
| 846 | contenc_writer *writer, |
| 847 | const char *buf, size_t nbytes) |
| 848 | { |
| 849 | char *all = Curl_all_content_encodings(); |
| 850 | |
| 851 | (void) writer; |
| 852 | (void) buf; |
| 853 | (void) nbytes; |
| 854 | |
| 855 | if(!all) |
| 856 | return CURLE_OUT_OF_MEMORY; |
| 857 | failf(conn->data, "Unrecognized content encoding type. " |
| 858 | "libcurl understands %s content encodings." , all); |
| 859 | free(all); |
| 860 | return CURLE_BAD_CONTENT_ENCODING; |
| 861 | } |
| 862 | |
| 863 | static void error_close_writer(struct connectdata *conn, |
| 864 | contenc_writer *writer) |
| 865 | { |
| 866 | (void) conn; |
| 867 | (void) writer; |
| 868 | } |
| 869 | |
| 870 | static const content_encoding error_encoding = { |
| 871 | NULL, |
| 872 | NULL, |
| 873 | error_init_writer, |
| 874 | error_unencode_write, |
| 875 | error_close_writer, |
| 876 | 0 |
| 877 | }; |
| 878 | |
| 879 | /* Create an unencoding writer stage using the given handler. */ |
| 880 | static contenc_writer *new_unencoding_writer(struct connectdata *conn, |
| 881 | const content_encoding *handler, |
| 882 | contenc_writer *downstream) |
| 883 | { |
| 884 | size_t sz = offsetof(contenc_writer, params) + handler->paramsize; |
| 885 | contenc_writer *writer = (contenc_writer *) calloc(1, sz); |
| 886 | |
| 887 | if(writer) { |
| 888 | writer->handler = handler; |
| 889 | writer->downstream = downstream; |
| 890 | if(handler->init_writer(conn, writer)) { |
| 891 | free(writer); |
| 892 | writer = NULL; |
| 893 | } |
| 894 | } |
| 895 | |
| 896 | return writer; |
| 897 | } |
| 898 | |
| 899 | /* Write data using an unencoding writer stack. */ |
| 900 | CURLcode Curl_unencode_write(struct connectdata *conn, contenc_writer *writer, |
| 901 | const char *buf, size_t nbytes) |
| 902 | { |
| 903 | if(!nbytes) |
| 904 | return CURLE_OK; |
| 905 | return writer->handler->unencode_write(conn, writer, buf, nbytes); |
| 906 | } |
| 907 | |
| 908 | /* Close and clean-up the connection's writer stack. */ |
| 909 | void Curl_unencode_cleanup(struct connectdata *conn) |
| 910 | { |
| 911 | struct Curl_easy *data = conn->data; |
| 912 | struct SingleRequest *k = &data->req; |
| 913 | contenc_writer *writer = k->writer_stack; |
| 914 | |
| 915 | while(writer) { |
| 916 | k->writer_stack = writer->downstream; |
| 917 | writer->handler->close_writer(conn, writer); |
| 918 | free(writer); |
| 919 | writer = k->writer_stack; |
| 920 | } |
| 921 | } |
| 922 | |
| 923 | /* Find the content encoding by name. */ |
| 924 | static const content_encoding *find_encoding(const char *name, size_t len) |
| 925 | { |
| 926 | const content_encoding * const *cep; |
| 927 | |
| 928 | for(cep = encodings; *cep; cep++) { |
| 929 | const content_encoding *ce = *cep; |
| 930 | if((strncasecompare(name, ce->name, len) && !ce->name[len]) || |
| 931 | (ce->alias && strncasecompare(name, ce->alias, len) && !ce->alias[len])) |
| 932 | return ce; |
| 933 | } |
| 934 | return NULL; |
| 935 | } |
| 936 | |
| 937 | /* Set-up the unencoding stack from the Content-Encoding header value. |
| 938 | * See RFC 7231 section 3.1.2.2. */ |
| 939 | CURLcode Curl_build_unencoding_stack(struct connectdata *conn, |
| 940 | const char *enclist, int maybechunked) |
| 941 | { |
| 942 | struct Curl_easy *data = conn->data; |
| 943 | struct SingleRequest *k = &data->req; |
| 944 | |
| 945 | do { |
| 946 | const char *name; |
| 947 | size_t namelen; |
| 948 | |
| 949 | /* Parse a single encoding name. */ |
| 950 | while(ISSPACE(*enclist) || *enclist == ',') |
| 951 | enclist++; |
| 952 | |
| 953 | name = enclist; |
| 954 | |
| 955 | for(namelen = 0; *enclist && *enclist != ','; enclist++) |
| 956 | if(!ISSPACE(*enclist)) |
| 957 | namelen = enclist - name + 1; |
| 958 | |
| 959 | /* Special case: chunked encoding is handled at the reader level. */ |
| 960 | if(maybechunked && namelen == 7 && strncasecompare(name, "chunked" , 7)) { |
| 961 | k->chunk = TRUE; /* chunks coming our way. */ |
| 962 | Curl_httpchunk_init(conn); /* init our chunky engine. */ |
| 963 | } |
| 964 | else if(namelen) { |
| 965 | const content_encoding *encoding = find_encoding(name, namelen); |
| 966 | contenc_writer *writer; |
| 967 | |
| 968 | if(!k->writer_stack) { |
| 969 | k->writer_stack = new_unencoding_writer(conn, &client_encoding, NULL); |
| 970 | |
| 971 | if(!k->writer_stack) |
| 972 | return CURLE_OUT_OF_MEMORY; |
| 973 | } |
| 974 | |
| 975 | if(!encoding) |
| 976 | encoding = &error_encoding; /* Defer error at stack use. */ |
| 977 | |
| 978 | /* Stack the unencoding stage. */ |
| 979 | writer = new_unencoding_writer(conn, encoding, k->writer_stack); |
| 980 | if(!writer) |
| 981 | return CURLE_OUT_OF_MEMORY; |
| 982 | k->writer_stack = writer; |
| 983 | } |
| 984 | } while(*enclist); |
| 985 | |
| 986 | return CURLE_OK; |
| 987 | } |
| 988 | |
| 989 | #else |
| 990 | /* Stubs for builds without HTTP. */ |
| 991 | CURLcode Curl_build_unencoding_stack(struct connectdata *conn, |
| 992 | const char *enclist, int maybechunked) |
| 993 | { |
| 994 | (void) conn; |
| 995 | (void) enclist; |
| 996 | (void) maybechunked; |
| 997 | return CURLE_NOT_BUILT_IN; |
| 998 | } |
| 999 | |
| 1000 | CURLcode Curl_unencode_write(struct connectdata *conn, contenc_writer *writer, |
| 1001 | const char *buf, size_t nbytes) |
| 1002 | { |
| 1003 | (void) conn; |
| 1004 | (void) writer; |
| 1005 | (void) buf; |
| 1006 | (void) nbytes; |
| 1007 | return CURLE_NOT_BUILT_IN; |
| 1008 | } |
| 1009 | |
| 1010 | void Curl_unencode_cleanup(struct connectdata *conn) |
| 1011 | { |
| 1012 | (void) conn; |
| 1013 | } |
| 1014 | |
| 1015 | char *Curl_all_content_encodings(void) |
| 1016 | { |
| 1017 | return strdup(CONTENT_ENCODING_DEFAULT); /* Satisfy caller. */ |
| 1018 | } |
| 1019 | |
| 1020 | #endif /* CURL_DISABLE_HTTP */ |
| 1021 | |