1 /*
2 +----------------------------------------------------------------------+
3 | Copyright (c) The PHP Group |
4 +----------------------------------------------------------------------+
5 | This source file is subject to version 3.01 of the PHP license, |
6 | that is bundled with this package in the file LICENSE, and is |
7 | available through the world-wide-web at the following url: |
8 | https://www.php.net/license/3_01.txt |
9 | If you did not receive a copy of the PHP license and are unable to |
10 | obtain it through the world-wide-web, please send a note to |
11 | license@php.net so we can mail you a copy immediately. |
12 +----------------------------------------------------------------------+
13 | Authors: Rasmus Lerdorf <rasmus@php.net> |
14 | Jim Winstead <jimw@php.net> |
15 | Hartmut Holzgraefe <hholzgra@php.net> |
16 | Wez Furlong <wez@thebrainroom.com> |
17 | Sara Golemon <pollita@php.net> |
18 +----------------------------------------------------------------------+
19 */
20
21 #include "php.h"
22 #include "php_globals.h"
23 #include "php_streams.h"
24 #include "php_network.h"
25 #include "php_ini.h"
26 #include "ext/standard/basic_functions.h"
27 #include "zend_smart_str.h"
28
29 #include <stdio.h>
30 #include <stdlib.h>
31 #include <errno.h>
32 #include <sys/types.h>
33 #include <sys/stat.h>
34 #include <fcntl.h>
35
36 #ifdef PHP_WIN32
37 #define O_RDONLY _O_RDONLY
38 #include "win32/param.h"
39 #else
40 #include <sys/param.h>
41 #endif
42
43 #include "php_standard.h"
44
45 #include <sys/types.h>
46 #ifdef HAVE_SYS_SOCKET_H
47 #include <sys/socket.h>
48 #endif
49
50 #ifdef PHP_WIN32
51 #include <winsock2.h>
52 #else
53 #include <netinet/in.h>
54 #include <netdb.h>
55 #ifdef HAVE_ARPA_INET_H
56 #include <arpa/inet.h>
57 #endif
58 #endif
59
60 #if defined(PHP_WIN32) || defined(__riscos__)
61 #undef AF_UNIX
62 #endif
63
64 #if defined(AF_UNIX)
65 #include <sys/un.h>
66 #endif
67
68 #include "php_fopen_wrappers.h"
69
70 #define HTTP_HEADER_BLOCK_SIZE 1024
71 #define PHP_URL_REDIRECT_MAX 20
72 #define HTTP_HEADER_USER_AGENT 1
73 #define HTTP_HEADER_HOST 2
74 #define HTTP_HEADER_AUTH 4
75 #define HTTP_HEADER_FROM 8
76 #define HTTP_HEADER_CONTENT_LENGTH 16
77 #define HTTP_HEADER_TYPE 32
78 #define HTTP_HEADER_CONNECTION 64
79
80 #define HTTP_WRAPPER_HEADER_INIT 1
81 #define HTTP_WRAPPER_REDIRECTED 2
82 #define HTTP_WRAPPER_KEEP_METHOD 4
83
strip_header(char * header_bag,char * lc_header_bag,const char * lc_header_name)84 static inline void strip_header(char *header_bag, char *lc_header_bag,
85 const char *lc_header_name)
86 {
87 char *lc_header_start = strstr(lc_header_bag, lc_header_name);
88 if (lc_header_start
89 && (lc_header_start == lc_header_bag || *(lc_header_start-1) == '\n')
90 ) {
91 char *header_start = header_bag + (lc_header_start - lc_header_bag);
92 char *lc_eol = strchr(lc_header_start, '\n');
93
94 if (lc_eol) {
95 char *eol = header_start + (lc_eol - lc_header_start);
96 size_t eollen = strlen(lc_eol);
97
98 memmove(lc_header_start, lc_eol+1, eollen);
99 memmove(header_start, eol+1, eollen);
100 } else {
101 *lc_header_start = '\0';
102 *header_start = '\0';
103 }
104 }
105 }
106
check_has_header(const char * headers,const char * header)107 static bool check_has_header(const char *headers, const char *header) {
108 const char *s = headers;
109 while ((s = strstr(s, header))) {
110 if (s == headers || *(s-1) == '\n') {
111 return 1;
112 }
113 s++;
114 }
115 return 0;
116 }
117
php_stream_handle_proxy_authorization_header(const char * s,smart_str * header)118 static zend_result php_stream_handle_proxy_authorization_header(const char *s, smart_str *header)
119 {
120 const char *p;
121
122 do {
123 while (*s == ' ' || *s == '\t') s++;
124 p = s;
125 while (*p != 0 && *p != ':' && *p != '\r' && *p !='\n') p++;
126 if (*p == ':') {
127 p++;
128 if (p - s == sizeof("Proxy-Authorization:") - 1 &&
129 zend_binary_strcasecmp(s, sizeof("Proxy-Authorization:") - 1,
130 "Proxy-Authorization:", sizeof("Proxy-Authorization:") - 1) == 0) {
131 while (*p != 0 && *p != '\r' && *p !='\n') p++;
132 smart_str_appendl(header, s, p - s);
133 smart_str_appendl(header, "\r\n", sizeof("\r\n")-1);
134 return SUCCESS;
135 } else {
136 while (*p != 0 && *p != '\r' && *p !='\n') p++;
137 }
138 }
139 s = p;
140 while (*s == '\r' || *s == '\n') s++;
141 } while (*s != 0);
142
143 return FAILURE;
144 }
145
php_stream_url_wrap_http_ex(php_stream_wrapper * wrapper,const char * path,const char * mode,int options,zend_string ** opened_path,php_stream_context * context,int redirect_max,int flags,zval * response_header STREAMS_DC)146 static php_stream *php_stream_url_wrap_http_ex(php_stream_wrapper *wrapper,
147 const char *path, const char *mode, int options, zend_string **opened_path,
148 php_stream_context *context, int redirect_max, int flags,
149 zval *response_header STREAMS_DC) /* {{{ */
150 {
151 php_stream *stream = NULL;
152 php_url *resource = NULL;
153 int use_ssl;
154 int use_proxy = 0;
155 zend_string *tmp = NULL;
156 char *ua_str = NULL;
157 zval *ua_zval = NULL, *tmpzval = NULL, ssl_proxy_peer_name;
158 char location[HTTP_HEADER_BLOCK_SIZE];
159 int reqok = 0;
160 char *http_header_line = NULL;
161 char tmp_line[128];
162 size_t chunk_size = 0, file_size = 0;
163 int eol_detect = 0;
164 zend_string *transport_string;
165 zend_string *errstr = NULL;
166 int have_header = 0;
167 bool request_fulluri = false, ignore_errors = false;
168 struct timeval timeout;
169 char *user_headers = NULL;
170 int header_init = ((flags & HTTP_WRAPPER_HEADER_INIT) != 0);
171 int redirected = ((flags & HTTP_WRAPPER_REDIRECTED) != 0);
172 int redirect_keep_method = ((flags & HTTP_WRAPPER_KEEP_METHOD) != 0);
173 bool follow_location = 1;
174 php_stream_filter *transfer_encoding = NULL;
175 int response_code;
176 smart_str req_buf = {0};
177 bool custom_request_method;
178
179 tmp_line[0] = '\0';
180
181 if (redirect_max < 1) {
182 php_stream_wrapper_log_error(wrapper, options, "Redirection limit reached, aborting");
183 return NULL;
184 }
185
186 resource = php_url_parse(path);
187 if (resource == NULL) {
188 return NULL;
189 }
190
191 ZEND_ASSERT(resource->scheme);
192 if (!zend_string_equals_literal_ci(resource->scheme, "http") &&
193 !zend_string_equals_literal_ci(resource->scheme, "https")) {
194 if (!context ||
195 (tmpzval = php_stream_context_get_option(context, wrapper->wops->label, "proxy")) == NULL ||
196 Z_TYPE_P(tmpzval) != IS_STRING ||
197 Z_STRLEN_P(tmpzval) == 0) {
198 php_url_free(resource);
199 return php_stream_open_wrapper_ex(path, mode, REPORT_ERRORS, NULL, context);
200 }
201 /* Called from a non-http wrapper with http proxying requested (i.e. ftp) */
202 request_fulluri = true;
203 use_ssl = 0;
204 use_proxy = 1;
205 transport_string = zend_string_copy(Z_STR_P(tmpzval));
206 } else {
207 /* Normal http request (possibly with proxy) */
208
209 if (strpbrk(mode, "awx+")) {
210 php_stream_wrapper_log_error(wrapper, options, "HTTP wrapper does not support writeable connections");
211 php_url_free(resource);
212 return NULL;
213 }
214
215 use_ssl = (ZSTR_LEN(resource->scheme) > 4) && ZSTR_VAL(resource->scheme)[4] == 's';
216 /* choose default ports */
217 if (use_ssl && resource->port == 0)
218 resource->port = 443;
219 else if (resource->port == 0)
220 resource->port = 80;
221
222 if (context &&
223 (tmpzval = php_stream_context_get_option(context, wrapper->wops->label, "proxy")) != NULL &&
224 Z_TYPE_P(tmpzval) == IS_STRING &&
225 Z_STRLEN_P(tmpzval) > 0) {
226 use_proxy = 1;
227 transport_string = zend_string_copy(Z_STR_P(tmpzval));
228 } else {
229 transport_string = zend_strpprintf(0, "%s://%s:%d", use_ssl ? "ssl" : "tcp", ZSTR_VAL(resource->host), resource->port);
230 }
231 }
232
233 if (context && (tmpzval = php_stream_context_get_option(context, wrapper->wops->label, "timeout")) != NULL) {
234 double d = zval_get_double(tmpzval);
235 #ifndef PHP_WIN32
236 timeout.tv_sec = (time_t) d;
237 timeout.tv_usec = (size_t) ((d - timeout.tv_sec) * 1000000);
238 #else
239 timeout.tv_sec = (long) d;
240 timeout.tv_usec = (long) ((d - timeout.tv_sec) * 1000000);
241 #endif
242 } else {
243 #ifndef PHP_WIN32
244 timeout.tv_sec = FG(default_socket_timeout);
245 #else
246 timeout.tv_sec = (long)FG(default_socket_timeout);
247 #endif
248 timeout.tv_usec = 0;
249 }
250
251 stream = php_stream_xport_create(ZSTR_VAL(transport_string), ZSTR_LEN(transport_string), options,
252 STREAM_XPORT_CLIENT | STREAM_XPORT_CONNECT,
253 NULL, &timeout, context, &errstr, NULL);
254
255 if (stream) {
256 php_stream_set_option(stream, PHP_STREAM_OPTION_READ_TIMEOUT, 0, &timeout);
257 }
258
259 if (errstr) {
260 php_stream_wrapper_log_error(wrapper, options, "%s", ZSTR_VAL(errstr));
261 zend_string_release_ex(errstr, 0);
262 errstr = NULL;
263 }
264
265 zend_string_release(transport_string);
266
267 if (stream && use_proxy && use_ssl) {
268 smart_str header = {0};
269
270 /* Set peer_name or name verification will try to use the proxy server name */
271 if (!context || (tmpzval = php_stream_context_get_option(context, "ssl", "peer_name")) == NULL) {
272 ZVAL_STR_COPY(&ssl_proxy_peer_name, resource->host);
273 php_stream_context_set_option(PHP_STREAM_CONTEXT(stream), "ssl", "peer_name", &ssl_proxy_peer_name);
274 zval_ptr_dtor(&ssl_proxy_peer_name);
275 }
276
277 smart_str_appendl(&header, "CONNECT ", sizeof("CONNECT ")-1);
278 smart_str_appends(&header, ZSTR_VAL(resource->host));
279 smart_str_appendc(&header, ':');
280 smart_str_append_unsigned(&header, resource->port);
281 smart_str_appendl(&header, " HTTP/1.0\r\n", sizeof(" HTTP/1.0\r\n")-1);
282
283 /* check if we have Proxy-Authorization header */
284 if (context && (tmpzval = php_stream_context_get_option(context, "http", "header")) != NULL) {
285 const char *s;
286
287 if (Z_TYPE_P(tmpzval) == IS_ARRAY) {
288 zval *tmpheader = NULL;
289
290 ZEND_HASH_FOREACH_VAL(Z_ARRVAL_P(tmpzval), tmpheader) {
291 if (Z_TYPE_P(tmpheader) == IS_STRING) {
292 s = Z_STRVAL_P(tmpheader);
293 if (php_stream_handle_proxy_authorization_header(s, &header) == SUCCESS) {
294 goto finish;
295 }
296 }
297 } ZEND_HASH_FOREACH_END();
298 } else if (Z_TYPE_P(tmpzval) == IS_STRING && Z_STRLEN_P(tmpzval)) {
299 s = Z_STRVAL_P(tmpzval);
300 if (php_stream_handle_proxy_authorization_header(s, &header) == SUCCESS) {
301 goto finish;
302 }
303 }
304 }
305 finish:
306 smart_str_appendl(&header, "\r\n", sizeof("\r\n")-1);
307
308 if (php_stream_write(stream, ZSTR_VAL(header.s), ZSTR_LEN(header.s)) != ZSTR_LEN(header.s)) {
309 php_stream_wrapper_log_error(wrapper, options, "Cannot connect to HTTPS server through proxy");
310 php_stream_close(stream);
311 stream = NULL;
312 }
313 smart_str_free(&header);
314
315 if (stream) {
316 char header_line[HTTP_HEADER_BLOCK_SIZE];
317
318 /* get response header */
319 while (php_stream_gets(stream, header_line, HTTP_HEADER_BLOCK_SIZE-1) != NULL) {
320 if (header_line[0] == '\n' ||
321 header_line[0] == '\r' ||
322 header_line[0] == '\0') {
323 break;
324 }
325 }
326 }
327
328 /* enable SSL transport layer */
329 if (stream) {
330 if (php_stream_xport_crypto_setup(stream, STREAM_CRYPTO_METHOD_SSLv23_CLIENT, NULL) < 0 ||
331 php_stream_xport_crypto_enable(stream, 1) < 0) {
332 php_stream_wrapper_log_error(wrapper, options, "Cannot connect to HTTPS server through proxy");
333 php_stream_close(stream);
334 stream = NULL;
335 }
336 }
337 }
338
339 if (stream == NULL)
340 goto out;
341
342 /* avoid buffering issues while reading header */
343 if (options & STREAM_WILL_CAST)
344 chunk_size = php_stream_set_chunk_size(stream, 1);
345
346 /* avoid problems with auto-detecting when reading the headers -> the headers
347 * are always in canonical \r\n format */
348 eol_detect = stream->flags & (PHP_STREAM_FLAG_DETECT_EOL | PHP_STREAM_FLAG_EOL_MAC);
349 stream->flags &= ~(PHP_STREAM_FLAG_DETECT_EOL | PHP_STREAM_FLAG_EOL_MAC);
350
351 php_stream_context_set(stream, context);
352
353 php_stream_notify_info(context, PHP_STREAM_NOTIFY_CONNECT, NULL, 0);
354
355 if (header_init && context && (tmpzval = php_stream_context_get_option(context, "http", "max_redirects")) != NULL) {
356 redirect_max = (int)zval_get_long(tmpzval);
357 }
358
359 custom_request_method = 0;
360 if (context && (tmpzval = php_stream_context_get_option(context, "http", "method")) != NULL) {
361 if (Z_TYPE_P(tmpzval) == IS_STRING && Z_STRLEN_P(tmpzval) > 0) {
362 /* As per the RFC, automatically redirected requests MUST NOT use other methods than
363 * GET and HEAD unless it can be confirmed by the user. */
364 if (!redirected || redirect_keep_method
365 || zend_string_equals_literal(Z_STR_P(tmpzval), "GET")
366 || zend_string_equals_literal(Z_STR_P(tmpzval), "HEAD")
367 ) {
368 custom_request_method = 1;
369 smart_str_append(&req_buf, Z_STR_P(tmpzval));
370 smart_str_appendc(&req_buf, ' ');
371 }
372 }
373 }
374
375 if (!custom_request_method) {
376 smart_str_appends(&req_buf, "GET ");
377 }
378
379 /* Should we send the entire path in the request line, default to no. */
380 if (!request_fulluri && context &&
381 (tmpzval = php_stream_context_get_option(context, "http", "request_fulluri")) != NULL) {
382 request_fulluri = zend_is_true(tmpzval);
383 }
384
385 if (request_fulluri) {
386 /* Ask for everything */
387 smart_str_appends(&req_buf, path);
388 } else {
389 /* Send the traditional /path/to/file?query_string */
390
391 /* file */
392 if (resource->path && ZSTR_LEN(resource->path)) {
393 smart_str_appends(&req_buf, ZSTR_VAL(resource->path));
394 } else {
395 smart_str_appendc(&req_buf, '/');
396 }
397
398 /* query string */
399 if (resource->query) {
400 smart_str_appendc(&req_buf, '?');
401 smart_str_appends(&req_buf, ZSTR_VAL(resource->query));
402 }
403 }
404
405 /* protocol version we are speaking */
406 if (context && (tmpzval = php_stream_context_get_option(context, "http", "protocol_version")) != NULL) {
407 char *protocol_version;
408 spprintf(&protocol_version, 0, "%.1F", zval_get_double(tmpzval));
409
410 smart_str_appends(&req_buf, " HTTP/");
411 smart_str_appends(&req_buf, protocol_version);
412 smart_str_appends(&req_buf, "\r\n");
413 efree(protocol_version);
414 } else {
415 smart_str_appends(&req_buf, " HTTP/1.1\r\n");
416 }
417
418 if (context && (tmpzval = php_stream_context_get_option(context, "http", "header")) != NULL) {
419 tmp = NULL;
420
421 if (Z_TYPE_P(tmpzval) == IS_ARRAY) {
422 zval *tmpheader = NULL;
423 smart_str tmpstr = {0};
424
425 ZEND_HASH_FOREACH_VAL(Z_ARRVAL_P(tmpzval), tmpheader) {
426 if (Z_TYPE_P(tmpheader) == IS_STRING) {
427 smart_str_append(&tmpstr, Z_STR_P(tmpheader));
428 smart_str_appendl(&tmpstr, "\r\n", sizeof("\r\n") - 1);
429 }
430 } ZEND_HASH_FOREACH_END();
431 smart_str_0(&tmpstr);
432 /* Remove newlines and spaces from start and end. there's at least one extra \r\n at the end that needs to go. */
433 if (tmpstr.s) {
434 tmp = php_trim(tmpstr.s, NULL, 0, 3);
435 smart_str_free(&tmpstr);
436 }
437 } else if (Z_TYPE_P(tmpzval) == IS_STRING && Z_STRLEN_P(tmpzval)) {
438 /* Remove newlines and spaces from start and end php_trim will estrndup() */
439 tmp = php_trim(Z_STR_P(tmpzval), NULL, 0, 3);
440 }
441 if (tmp && ZSTR_LEN(tmp)) {
442 char *s;
443 char *t;
444
445 user_headers = estrndup(ZSTR_VAL(tmp), ZSTR_LEN(tmp));
446
447 if (ZSTR_IS_INTERNED(tmp)) {
448 tmp = zend_string_init(ZSTR_VAL(tmp), ZSTR_LEN(tmp), 0);
449 } else if (GC_REFCOUNT(tmp) > 1) {
450 GC_DELREF(tmp);
451 tmp = zend_string_init(ZSTR_VAL(tmp), ZSTR_LEN(tmp), 0);
452 }
453
454 /* Make lowercase for easy comparison against 'standard' headers */
455 zend_str_tolower(ZSTR_VAL(tmp), ZSTR_LEN(tmp));
456 t = ZSTR_VAL(tmp);
457
458 if (!header_init && !redirect_keep_method) {
459 /* strip POST headers on redirect */
460 strip_header(user_headers, t, "content-length:");
461 strip_header(user_headers, t, "content-type:");
462 }
463
464 if (check_has_header(t, "user-agent:")) {
465 have_header |= HTTP_HEADER_USER_AGENT;
466 }
467 if (check_has_header(t, "host:")) {
468 have_header |= HTTP_HEADER_HOST;
469 }
470 if (check_has_header(t, "from:")) {
471 have_header |= HTTP_HEADER_FROM;
472 }
473 if (check_has_header(t, "authorization:")) {
474 have_header |= HTTP_HEADER_AUTH;
475 }
476 if (check_has_header(t, "content-length:")) {
477 have_header |= HTTP_HEADER_CONTENT_LENGTH;
478 }
479 if (check_has_header(t, "content-type:")) {
480 have_header |= HTTP_HEADER_TYPE;
481 }
482 if (check_has_header(t, "connection:")) {
483 have_header |= HTTP_HEADER_CONNECTION;
484 }
485
486 /* remove Proxy-Authorization header */
487 if (use_proxy && use_ssl && (s = strstr(t, "proxy-authorization:")) &&
488 (s == t || *(s-1) == '\n')) {
489 char *p = s + sizeof("proxy-authorization:") - 1;
490
491 while (s > t && (*(s-1) == ' ' || *(s-1) == '\t')) s--;
492 while (*p != 0 && *p != '\r' && *p != '\n') p++;
493 while (*p == '\r' || *p == '\n') p++;
494 if (*p == 0) {
495 if (s == t) {
496 efree(user_headers);
497 user_headers = NULL;
498 } else {
499 while (s > t && (*(s-1) == '\r' || *(s-1) == '\n')) s--;
500 user_headers[s - t] = 0;
501 }
502 } else {
503 memmove(user_headers + (s - t), user_headers + (p - t), strlen(p) + 1);
504 }
505 }
506
507 }
508 if (tmp) {
509 zend_string_release_ex(tmp, 0);
510 }
511 }
512
513 /* auth header if it was specified */
514 if (((have_header & HTTP_HEADER_AUTH) == 0) && resource->user) {
515 /* make scratch large enough to hold the whole URL (over-estimate) */
516 size_t scratch_len = strlen(path) + 1;
517 char *scratch = emalloc(scratch_len);
518 zend_string *stmp;
519
520 /* decode the strings first */
521 php_url_decode(ZSTR_VAL(resource->user), ZSTR_LEN(resource->user));
522
523 strcpy(scratch, ZSTR_VAL(resource->user));
524 strcat(scratch, ":");
525
526 /* Note: password is optional! */
527 if (resource->pass) {
528 php_url_decode(ZSTR_VAL(resource->pass), ZSTR_LEN(resource->pass));
529 strcat(scratch, ZSTR_VAL(resource->pass));
530 }
531
532 stmp = php_base64_encode((unsigned char*)scratch, strlen(scratch));
533
534 smart_str_appends(&req_buf, "Authorization: Basic ");
535 smart_str_appends(&req_buf, ZSTR_VAL(stmp));
536 smart_str_appends(&req_buf, "\r\n");
537
538 php_stream_notify_info(context, PHP_STREAM_NOTIFY_AUTH_REQUIRED, NULL, 0);
539
540 zend_string_free(stmp);
541 efree(scratch);
542 }
543
544 /* if the user has configured who they are, send a From: line */
545 if (!(have_header & HTTP_HEADER_FROM) && FG(from_address)) {
546 smart_str_appends(&req_buf, "From: ");
547 smart_str_appends(&req_buf, FG(from_address));
548 smart_str_appends(&req_buf, "\r\n");
549 }
550
551 /* Send Host: header so name-based virtual hosts work */
552 if ((have_header & HTTP_HEADER_HOST) == 0) {
553 smart_str_appends(&req_buf, "Host: ");
554 smart_str_appends(&req_buf, ZSTR_VAL(resource->host));
555 if ((use_ssl && resource->port != 443 && resource->port != 0) ||
556 (!use_ssl && resource->port != 80 && resource->port != 0)) {
557 smart_str_appendc(&req_buf, ':');
558 smart_str_append_unsigned(&req_buf, resource->port);
559 }
560 smart_str_appends(&req_buf, "\r\n");
561 }
562
563 /* Send a Connection: close header to avoid hanging when the server
564 * interprets the RFC literally and establishes a keep-alive connection,
565 * unless the user specifically requests something else by specifying a
566 * Connection header in the context options. Send that header even for
567 * HTTP/1.0 to avoid issues when the server respond with an HTTP/1.1
568 * keep-alive response, which is the preferred response type. */
569 if ((have_header & HTTP_HEADER_CONNECTION) == 0) {
570 smart_str_appends(&req_buf, "Connection: close\r\n");
571 }
572
573 if (context &&
574 (ua_zval = php_stream_context_get_option(context, "http", "user_agent")) != NULL &&
575 Z_TYPE_P(ua_zval) == IS_STRING) {
576 ua_str = Z_STRVAL_P(ua_zval);
577 } else if (FG(user_agent)) {
578 ua_str = FG(user_agent);
579 }
580
581 if (((have_header & HTTP_HEADER_USER_AGENT) == 0) && ua_str) {
582 #define _UA_HEADER "User-Agent: %s\r\n"
583 char *ua;
584 size_t ua_len;
585
586 ua_len = sizeof(_UA_HEADER) + strlen(ua_str);
587
588 /* ensure the header is only sent if user_agent is not blank */
589 if (ua_len > sizeof(_UA_HEADER)) {
590 ua = emalloc(ua_len + 1);
591 if ((ua_len = slprintf(ua, ua_len, _UA_HEADER, ua_str)) > 0) {
592 ua[ua_len] = 0;
593 smart_str_appendl(&req_buf, ua, ua_len);
594 } else {
595 php_error_docref(NULL, E_WARNING, "Cannot construct User-agent header");
596 }
597 efree(ua);
598 }
599 }
600
601 if (user_headers) {
602 /* A bit weird, but some servers require that Content-Length be sent prior to Content-Type for POST
603 * see bug #44603 for details. Since Content-Type maybe part of user's headers we need to do this check first.
604 */
605 if (
606 (header_init || redirect_keep_method) &&
607 context &&
608 !(have_header & HTTP_HEADER_CONTENT_LENGTH) &&
609 (tmpzval = php_stream_context_get_option(context, "http", "content")) != NULL &&
610 Z_TYPE_P(tmpzval) == IS_STRING && Z_STRLEN_P(tmpzval) > 0
611 ) {
612 smart_str_appends(&req_buf, "Content-Length: ");
613 smart_str_append_unsigned(&req_buf, Z_STRLEN_P(tmpzval));
614 smart_str_appends(&req_buf, "\r\n");
615 have_header |= HTTP_HEADER_CONTENT_LENGTH;
616 }
617
618 smart_str_appends(&req_buf, user_headers);
619 smart_str_appends(&req_buf, "\r\n");
620 efree(user_headers);
621 }
622
623 /* Request content, such as for POST requests */
624 if ((header_init || redirect_keep_method) && context &&
625 (tmpzval = php_stream_context_get_option(context, "http", "content")) != NULL &&
626 Z_TYPE_P(tmpzval) == IS_STRING && Z_STRLEN_P(tmpzval) > 0) {
627 if (!(have_header & HTTP_HEADER_CONTENT_LENGTH)) {
628 smart_str_appends(&req_buf, "Content-Length: ");
629 smart_str_append_unsigned(&req_buf, Z_STRLEN_P(tmpzval));
630 smart_str_appends(&req_buf, "\r\n");
631 }
632 if (!(have_header & HTTP_HEADER_TYPE)) {
633 smart_str_appends(&req_buf, "Content-Type: application/x-www-form-urlencoded\r\n");
634 php_error_docref(NULL, E_NOTICE, "Content-type not specified assuming application/x-www-form-urlencoded");
635 }
636 smart_str_appends(&req_buf, "\r\n");
637 smart_str_appendl(&req_buf, Z_STRVAL_P(tmpzval), Z_STRLEN_P(tmpzval));
638 } else {
639 smart_str_appends(&req_buf, "\r\n");
640 }
641
642 /* send it */
643 php_stream_write(stream, ZSTR_VAL(req_buf.s), ZSTR_LEN(req_buf.s));
644
645 location[0] = '\0';
646
647 if (Z_ISUNDEF_P(response_header)) {
648 array_init(response_header);
649 }
650
651 {
652 /* get response header */
653 size_t tmp_line_len;
654 if (!php_stream_eof(stream) &&
655 php_stream_get_line(stream, tmp_line, sizeof(tmp_line) - 1, &tmp_line_len) != NULL) {
656 zval http_response;
657
658 if (tmp_line_len > 9) {
659 response_code = atoi(tmp_line + 9);
660 } else {
661 response_code = 0;
662 }
663 if (context && NULL != (tmpzval = php_stream_context_get_option(context, "http", "ignore_errors"))) {
664 ignore_errors = zend_is_true(tmpzval);
665 }
666 /* when we request only the header, don't fail even on error codes */
667 if ((options & STREAM_ONLY_GET_HEADERS) || ignore_errors) {
668 reqok = 1;
669 }
670
671 /* status codes of 1xx are "informational", and will be followed by a real response
672 * e.g "100 Continue". RFC 7231 states that unexpected 1xx status MUST be parsed,
673 * and MAY be ignored. As such, we need to skip ahead to the "real" status*/
674 if (response_code >= 100 && response_code < 200 && response_code != 101) {
675 /* consume lines until we find a line starting 'HTTP/1' */
676 while (
677 !php_stream_eof(stream)
678 && php_stream_get_line(stream, tmp_line, sizeof(tmp_line) - 1, &tmp_line_len) != NULL
679 && ( tmp_line_len < sizeof("HTTP/1") - 1 || strncasecmp(tmp_line, "HTTP/1", sizeof("HTTP/1") - 1) )
680 );
681
682 if (tmp_line_len > 9) {
683 response_code = atoi(tmp_line + 9);
684 } else {
685 response_code = 0;
686 }
687 }
688 /* all status codes in the 2xx range are defined by the specification as successful;
689 * all status codes in the 3xx range are for redirection, and so also should never
690 * fail */
691 if (response_code >= 200 && response_code < 400) {
692 reqok = 1;
693 } else {
694 switch(response_code) {
695 case 403:
696 php_stream_notify_error(context, PHP_STREAM_NOTIFY_AUTH_RESULT,
697 tmp_line, response_code);
698 break;
699 default:
700 /* safety net in the event tmp_line == NULL */
701 if (!tmp_line_len) {
702 tmp_line[0] = '\0';
703 }
704 php_stream_notify_error(context, PHP_STREAM_NOTIFY_FAILURE,
705 tmp_line, response_code);
706 }
707 }
708 if (tmp_line_len >= 1 && tmp_line[tmp_line_len - 1] == '\n') {
709 --tmp_line_len;
710 if (tmp_line_len >= 1 &&tmp_line[tmp_line_len - 1] == '\r') {
711 --tmp_line_len;
712 }
713 } else {
714 // read and discard rest of status line
715 char *line = php_stream_get_line(stream, NULL, 0, NULL);
716 efree(line);
717 }
718 ZVAL_STRINGL(&http_response, tmp_line, tmp_line_len);
719 zend_hash_next_index_insert(Z_ARRVAL_P(response_header), &http_response);
720 } else {
721 php_stream_close(stream);
722 stream = NULL;
723 php_stream_wrapper_log_error(wrapper, options, "HTTP request failed!");
724 goto out;
725 }
726 }
727
728 /* read past HTTP headers */
729
730 while (!php_stream_eof(stream)) {
731 size_t http_header_line_length;
732
733 if (http_header_line != NULL) {
734 efree(http_header_line);
735 }
736 if ((http_header_line = php_stream_get_line(stream, NULL, 0, &http_header_line_length)) && *http_header_line != '\n' && *http_header_line != '\r') {
737 char *e = http_header_line + http_header_line_length - 1;
738 char *http_header_value;
739
740 while (e >= http_header_line && (*e == '\n' || *e == '\r')) {
741 e--;
742 }
743
744 /* The primary definition of an HTTP header in RFC 7230 states:
745 * > Each header field consists of a case-insensitive field name followed
746 * > by a colon (":"), optional leading whitespace, the field value, and
747 * > optional trailing whitespace. */
748
749 /* Strip trailing whitespace */
750 while (e >= http_header_line && (*e == ' ' || *e == '\t')) {
751 e--;
752 }
753
754 /* Terminate header line */
755 e++;
756 *e = '\0';
757 http_header_line_length = e - http_header_line;
758
759 http_header_value = memchr(http_header_line, ':', http_header_line_length);
760 if (http_header_value) {
761 http_header_value++; /* Skip ':' */
762
763 /* Strip leading whitespace */
764 while (http_header_value < e
765 && (*http_header_value == ' ' || *http_header_value == '\t')) {
766 http_header_value++;
767 }
768 } else {
769 /* There is no colon. Set the value to the end of the header line, which is
770 * effectively an empty string. */
771 http_header_value = e;
772 }
773
774 if (!strncasecmp(http_header_line, "Location:", sizeof("Location:")-1)) {
775 if (context && (tmpzval = php_stream_context_get_option(context, "http", "follow_location")) != NULL) {
776 follow_location = zval_is_true(tmpzval);
777 } else if (!((response_code >= 300 && response_code < 304)
778 || 307 == response_code || 308 == response_code)) {
779 /* we shouldn't redirect automatically
780 if follow_location isn't set and response_code not in (300, 301, 302, 303 and 307)
781 see http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.1
782 RFC 7238 defines 308: http://tools.ietf.org/html/rfc7238 */
783 follow_location = 0;
784 }
785 strlcpy(location, http_header_value, sizeof(location));
786 } else if (!strncasecmp(http_header_line, "Content-Type:", sizeof("Content-Type:")-1)) {
787 php_stream_notify_info(context, PHP_STREAM_NOTIFY_MIME_TYPE_IS, http_header_value, 0);
788 } else if (!strncasecmp(http_header_line, "Content-Length:", sizeof("Content-Length:")-1)) {
789 /* https://www.rfc-editor.org/rfc/rfc9110.html#name-content-length */
790 const char *ptr = http_header_value;
791 /* must contain only digits, no + or - symbols */
792 if (*ptr >= '0' && *ptr <= '9') {
793 char *endptr = NULL;
794 size_t parsed = ZEND_STRTOUL(ptr, &endptr, 10);
795 /* check whether there was no garbage in the header value and the conversion was successful */
796 if (endptr && !*endptr) {
797 /* truncate for 32-bit such that no negative file sizes occur */
798 file_size = MIN(parsed, ZEND_LONG_MAX);
799 php_stream_notify_file_size(context, file_size, http_header_line, 0);
800 }
801 }
802 } else if (
803 !strncasecmp(http_header_line, "Transfer-Encoding:", sizeof("Transfer-Encoding:")-1)
804 && !strncasecmp(http_header_value, "Chunked", sizeof("Chunked")-1)
805 ) {
806
807 /* create filter to decode response body */
808 if (!(options & STREAM_ONLY_GET_HEADERS)) {
809 bool decode = true;
810
811 if (context && (tmpzval = php_stream_context_get_option(context, "http", "auto_decode")) != NULL) {
812 decode = zend_is_true(tmpzval);
813 }
814 if (decode) {
815 transfer_encoding = php_stream_filter_create("dechunk", NULL, php_stream_is_persistent(stream));
816 if (transfer_encoding) {
817 /* don't store transfer-encodeing header */
818 continue;
819 }
820 }
821 }
822 }
823
824 {
825 zval http_header;
826 ZVAL_STRINGL(&http_header, http_header_line, http_header_line_length);
827 zend_hash_next_index_insert(Z_ARRVAL_P(response_header), &http_header);
828 }
829 } else {
830 break;
831 }
832 }
833
834 if (!reqok || (location[0] != '\0' && follow_location)) {
835 if (!follow_location || (((options & STREAM_ONLY_GET_HEADERS) || ignore_errors) && redirect_max <= 1)) {
836 goto out;
837 }
838
839 if (location[0] != '\0')
840 php_stream_notify_info(context, PHP_STREAM_NOTIFY_REDIRECTED, location, 0);
841
842 php_stream_close(stream);
843 stream = NULL;
844
845 if (transfer_encoding) {
846 php_stream_filter_free(transfer_encoding);
847 transfer_encoding = NULL;
848 }
849
850 if (location[0] != '\0') {
851
852 char new_path[HTTP_HEADER_BLOCK_SIZE];
853 char loc_path[HTTP_HEADER_BLOCK_SIZE];
854
855 *new_path='\0';
856 if (strlen(location)<8 || (strncasecmp(location, "http://", sizeof("http://")-1) &&
857 strncasecmp(location, "https://", sizeof("https://")-1) &&
858 strncasecmp(location, "ftp://", sizeof("ftp://")-1) &&
859 strncasecmp(location, "ftps://", sizeof("ftps://")-1)))
860 {
861 if (*location != '/') {
862 if (*(location+1) != '\0' && resource->path) {
863 char *s = strrchr(ZSTR_VAL(resource->path), '/');
864 if (!s) {
865 s = ZSTR_VAL(resource->path);
866 if (!ZSTR_LEN(resource->path)) {
867 zend_string_release_ex(resource->path, 0);
868 resource->path = ZSTR_INIT_LITERAL("/", 0);
869 s = ZSTR_VAL(resource->path);
870 } else {
871 *s = '/';
872 }
873 }
874 s[1] = '\0';
875 if (resource->path &&
876 ZSTR_VAL(resource->path)[0] == '/' &&
877 ZSTR_VAL(resource->path)[1] == '\0') {
878 snprintf(loc_path, sizeof(loc_path) - 1, "%s%s", ZSTR_VAL(resource->path), location);
879 } else {
880 snprintf(loc_path, sizeof(loc_path) - 1, "%s/%s", ZSTR_VAL(resource->path), location);
881 }
882 } else {
883 snprintf(loc_path, sizeof(loc_path) - 1, "/%s", location);
884 }
885 } else {
886 strlcpy(loc_path, location, sizeof(loc_path));
887 }
888 if ((use_ssl && resource->port != 443) || (!use_ssl && resource->port != 80)) {
889 snprintf(new_path, sizeof(new_path) - 1, "%s://%s:%d%s", ZSTR_VAL(resource->scheme), ZSTR_VAL(resource->host), resource->port, loc_path);
890 } else {
891 snprintf(new_path, sizeof(new_path) - 1, "%s://%s%s", ZSTR_VAL(resource->scheme), ZSTR_VAL(resource->host), loc_path);
892 }
893 } else {
894 strlcpy(new_path, location, sizeof(new_path));
895 }
896
897 php_url_free(resource);
898 /* check for invalid redirection URLs */
899 if ((resource = php_url_parse(new_path)) == NULL) {
900 php_stream_wrapper_log_error(wrapper, options, "Invalid redirect URL! %s", new_path);
901 goto out;
902 }
903
904 #define CHECK_FOR_CNTRL_CHARS(val) { \
905 if (val) { \
906 unsigned char *s, *e; \
907 ZSTR_LEN(val) = php_url_decode(ZSTR_VAL(val), ZSTR_LEN(val)); \
908 s = (unsigned char*)ZSTR_VAL(val); e = s + ZSTR_LEN(val); \
909 while (s < e) { \
910 if (iscntrl(*s)) { \
911 php_stream_wrapper_log_error(wrapper, options, "Invalid redirect URL! %s", new_path); \
912 goto out; \
913 } \
914 s++; \
915 } \
916 } \
917 }
918 /* check for control characters in login, password & path */
919 if (strncasecmp(new_path, "http://", sizeof("http://") - 1) || strncasecmp(new_path, "https://", sizeof("https://") - 1)) {
920 CHECK_FOR_CNTRL_CHARS(resource->user);
921 CHECK_FOR_CNTRL_CHARS(resource->pass);
922 CHECK_FOR_CNTRL_CHARS(resource->path);
923 }
924 int new_flags = HTTP_WRAPPER_REDIRECTED;
925 if (response_code == 307 || response_code == 308) {
926 /* RFC 7538 specifies that status code 308 does not allow changing the request method from POST to GET.
927 * RFC 7231 does the same for status code 307.
928 * To keep consistency between POST and PATCH requests, we'll also not change the request method from PATCH to GET, even though it's allowed it's not mandated by the RFC. */
929 new_flags |= HTTP_WRAPPER_KEEP_METHOD;
930 }
931 stream = php_stream_url_wrap_http_ex(
932 wrapper, new_path, mode, options, opened_path, context,
933 --redirect_max, new_flags, response_header STREAMS_CC);
934 } else {
935 php_stream_wrapper_log_error(wrapper, options, "HTTP request failed! %s", tmp_line);
936 }
937 }
938 out:
939
940 smart_str_free(&req_buf);
941
942 if (http_header_line) {
943 efree(http_header_line);
944 }
945
946 if (resource) {
947 php_url_free(resource);
948 }
949
950 if (stream) {
951 if (header_init) {
952 ZVAL_COPY(&stream->wrapperdata, response_header);
953 }
954 php_stream_notify_progress_init(context, 0, file_size);
955
956 /* Restore original chunk size now that we're done with headers */
957 if (options & STREAM_WILL_CAST)
958 php_stream_set_chunk_size(stream, (int)chunk_size);
959
960 /* restore the users auto-detect-line-endings setting */
961 stream->flags |= eol_detect;
962
963 /* as far as streams are concerned, we are now at the start of
964 * the stream */
965 stream->position = 0;
966
967 /* restore mode */
968 strlcpy(stream->mode, mode, sizeof(stream->mode));
969
970 if (transfer_encoding) {
971 php_stream_filter_append(&stream->readfilters, transfer_encoding);
972 }
973
974 /* It's possible that the server already sent in more data than just the headers.
975 * We account for this by adjusting the progress counter by the difference of
976 * already read header data and the body. */
977 if (stream->writepos > stream->readpos) {
978 php_stream_notify_progress_increment(context, stream->writepos - stream->readpos, 0);
979 }
980 }
981
982 return stream;
983 }
984 /* }}} */
985
php_stream_url_wrap_http(php_stream_wrapper * wrapper,const char * path,const char * mode,int options,zend_string ** opened_path,php_stream_context * context STREAMS_DC)986 php_stream *php_stream_url_wrap_http(php_stream_wrapper *wrapper, const char *path, const char *mode, int options, zend_string **opened_path, php_stream_context *context STREAMS_DC) /* {{{ */
987 {
988 php_stream *stream;
989 zval headers;
990
991 ZVAL_UNDEF(&headers);
992
993 zval_ptr_dtor(&BG(last_http_headers));
994 ZVAL_UNDEF(&BG(last_http_headers));
995
996 stream = php_stream_url_wrap_http_ex(
997 wrapper, path, mode, options, opened_path, context,
998 PHP_URL_REDIRECT_MAX, HTTP_WRAPPER_HEADER_INIT, &headers STREAMS_CC);
999
1000 if (!Z_ISUNDEF(headers)) {
1001 ZVAL_COPY(&BG(last_http_headers), &headers);
1002
1003 if (FAILURE == zend_set_local_var_str(
1004 "http_response_header", sizeof("http_response_header")-1, &headers, 0)) {
1005 zval_ptr_dtor(&headers);
1006 }
1007 }
1008
1009 return stream;
1010 }
1011 /* }}} */
1012
php_stream_http_stream_stat(php_stream_wrapper * wrapper,php_stream * stream,php_stream_statbuf * ssb)1013 static int php_stream_http_stream_stat(php_stream_wrapper *wrapper, php_stream *stream, php_stream_statbuf *ssb) /* {{{ */
1014 {
1015 /* one day, we could fill in the details based on Date: and Content-Length:
1016 * headers. For now, we return with a failure code to prevent the underlying
1017 * file's details from being used instead. */
1018 return -1;
1019 }
1020 /* }}} */
1021
1022 static const php_stream_wrapper_ops http_stream_wops = {
1023 php_stream_url_wrap_http,
1024 NULL, /* stream_close */
1025 php_stream_http_stream_stat,
1026 NULL, /* stat_url */
1027 NULL, /* opendir */
1028 "http",
1029 NULL, /* unlink */
1030 NULL, /* rename */
1031 NULL, /* mkdir */
1032 NULL, /* rmdir */
1033 NULL
1034 };
1035
1036 PHPAPI const php_stream_wrapper php_stream_http_wrapper = {
1037 &http_stream_wops,
1038 NULL,
1039 1 /* is_url */
1040 };
1041
