2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation,
6 This file is part of GNU Wget.
8 GNU Wget is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or
11 (at your option) any later version.
13 GNU Wget is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with Wget. If not, see <http://www.gnu.org/licenses/>.
21 Additional permission under GNU GPL version 3 section 7
23 If you modify this program, or any covered work, by linking or
24 combining it with the OpenSSL project's OpenSSL library (or a
25 modified version of that library), containing parts covered by the
26 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
27 grants you additional permission to convey the resulting work.
28 Corresponding Source for a non-source form of such a combination
29 shall include the source code for the parts of OpenSSL used as well
30 as that of the covered work. */
55 # include "http-ntlm.h"
69 #endif /* def __VMS */
71 extern char *version_string;
75 static char *create_authorization_line (const char *, const char *,
76 const char *, const char *,
77 const char *, bool *, uerr_t *);
78 static char *basic_authentication_encode (const char *, const char *);
79 static bool known_authentication_scheme_p (const char *, const char *);
80 static void ensure_extension (struct http_stat *, const char *, int *);
81 static void load_cookies (void);
84 # define MIN(x, y) ((x) > (y) ? (y) : (x))
88 static bool cookies_loaded_p;
89 static struct cookie_jar *wget_cookie_jar;
91 #define TEXTHTML_S "text/html"
92 #define TEXTXHTML_S "application/xhtml+xml"
93 #define TEXTCSS_S "text/css"
95 /* Some status code validation macros: */
96 #define H_10X(x) (((x) >= 100) && ((x) < 200))
97 #define H_20X(x) (((x) >= 200) && ((x) < 300))
98 #define H_PARTIAL(x) ((x) == HTTP_STATUS_PARTIAL_CONTENTS)
99 #define H_REDIRECTED(x) ((x) == HTTP_STATUS_MOVED_PERMANENTLY \
100 || (x) == HTTP_STATUS_MOVED_TEMPORARILY \
101 || (x) == HTTP_STATUS_SEE_OTHER \
102 || (x) == HTTP_STATUS_TEMPORARY_REDIRECT)
104 /* HTTP/1.0 status codes from RFC1945, provided for reference. */
105 /* Successful 2xx. */
106 #define HTTP_STATUS_OK 200
107 #define HTTP_STATUS_CREATED 201
108 #define HTTP_STATUS_ACCEPTED 202
109 #define HTTP_STATUS_NO_CONTENT 204
110 #define HTTP_STATUS_PARTIAL_CONTENTS 206
112 /* Redirection 3xx. */
113 #define HTTP_STATUS_MULTIPLE_CHOICES 300
114 #define HTTP_STATUS_MOVED_PERMANENTLY 301
115 #define HTTP_STATUS_MOVED_TEMPORARILY 302
116 #define HTTP_STATUS_SEE_OTHER 303 /* from HTTP/1.1 */
117 #define HTTP_STATUS_NOT_MODIFIED 304
118 #define HTTP_STATUS_TEMPORARY_REDIRECT 307 /* from HTTP/1.1 */
120 /* Client error 4xx. */
121 #define HTTP_STATUS_BAD_REQUEST 400
122 #define HTTP_STATUS_UNAUTHORIZED 401
123 #define HTTP_STATUS_FORBIDDEN 403
124 #define HTTP_STATUS_NOT_FOUND 404
125 #define HTTP_STATUS_RANGE_NOT_SATISFIABLE 416
127 /* Server errors 5xx. */
128 #define HTTP_STATUS_INTERNAL 500
129 #define HTTP_STATUS_NOT_IMPLEMENTED 501
130 #define HTTP_STATUS_BAD_GATEWAY 502
131 #define HTTP_STATUS_UNAVAILABLE 503
134 rel_none, rel_name, rel_value, rel_both
141 struct request_header {
143 enum rp release_policy;
145 int hcount, hcapacity;
150 /* Create a new, empty request. Set the request's method and its
151 arguments. METHOD should be a literal string (or it should outlive
152 the request) because it will not be freed. ARG will be freed by
155 static struct request *
156 request_new (const char *method, char *arg)
158 struct request *req = xnew0 (struct request);
160 req->headers = xnew_array (struct request_header, req->hcapacity);
161 req->method = method;
166 /* Return the method string passed with the last call to
167 request_set_method. */
170 request_method (const struct request *req)
175 /* Free one header according to the release policy specified with
176 request_set_header. */
179 release_header (struct request_header *hdr)
181 switch (hdr->release_policy)
198 /* Set the request named NAME to VALUE. Specifically, this means that
199 a "NAME: VALUE\r\n" header line will be used in the request. If a
200 header with the same name previously existed in the request, its
201 value will be replaced by this one. A NULL value means do nothing.
203 RELEASE_POLICY determines whether NAME and VALUE should be released
204 (freed) with request_free. Allowed values are:
206 - rel_none - don't free NAME or VALUE
207 - rel_name - free NAME when done
208 - rel_value - free VALUE when done
209 - rel_both - free both NAME and VALUE when done
211 Setting release policy is useful when arguments come from different
212 sources. For example:
214 // Don't free literal strings!
215 request_set_header (req, "Pragma", "no-cache", rel_none);
217 // Don't free a global variable, we'll need it later.
218 request_set_header (req, "Referer", opt.referer, rel_none);
220 // Value freshly allocated, free it when done.
221 request_set_header (req, "Range",
222 aprintf ("bytes=%s-", number_to_static_string (hs->restval)),
227 request_set_header (struct request *req, const char *name, const char *value,
228 enum rp release_policy)
230 struct request_header *hdr;
235 /* A NULL value is a no-op; if freeing the name is requested,
236 free it now to avoid leaks. */
237 if (release_policy == rel_name || release_policy == rel_both)
238 xfree ((void *)name);
242 for (i = 0; i < req->hcount; i++)
244 hdr = &req->headers[i];
245 if (0 == strcasecmp (name, hdr->name))
247 /* Replace existing header. */
248 release_header (hdr);
249 hdr->name = (void *)name;
250 hdr->value = (void *)value;
251 hdr->release_policy = release_policy;
256 /* Install new header. */
258 if (req->hcount >= req->hcapacity)
260 req->hcapacity <<= 1;
261 req->headers = xrealloc (req->headers, req->hcapacity * sizeof (*hdr));
263 hdr = &req->headers[req->hcount++];
264 hdr->name = (void *)name;
265 hdr->value = (void *)value;
266 hdr->release_policy = release_policy;
269 /* Like request_set_header, but sets the whole header line, as
270 provided by the user using the `--header' option. For example,
271 request_set_user_header (req, "Foo: bar") works just like
272 request_set_header (req, "Foo", "bar"). */
275 request_set_user_header (struct request *req, const char *header)
278 const char *p = strchr (header, ':');
281 BOUNDED_TO_ALLOCA (header, p, name);
283 while (c_isspace (*p))
285 request_set_header (req, xstrdup (name), (char *) p, rel_name);
288 /* Remove the header with specified name from REQ. Returns true if
289 the header was actually removed, false otherwise. */
292 request_remove_header (struct request *req, const char *name)
295 for (i = 0; i < req->hcount; i++)
297 struct request_header *hdr = &req->headers[i];
298 if (0 == strcasecmp (name, hdr->name))
300 release_header (hdr);
301 /* Move the remaining headers by one. */
302 if (i < req->hcount - 1)
303 memmove (hdr, hdr + 1, (req->hcount - i - 1) * sizeof (*hdr));
311 #define APPEND(p, str) do { \
312 int A_len = strlen (str); \
313 memcpy (p, str, A_len); \
317 /* Construct the request and write it to FD using fd_write.
318 If warc_tmp is set to a file pointer, the request string will
319 also be written to that file. */
322 request_send (const struct request *req, int fd, FILE *warc_tmp)
324 char *request_string, *p;
325 int i, size, write_error;
327 /* Count the request size. */
330 /* METHOD " " ARG " " "HTTP/1.0" "\r\n" */
331 size += strlen (req->method) + 1 + strlen (req->arg) + 1 + 8 + 2;
333 for (i = 0; i < req->hcount; i++)
335 struct request_header *hdr = &req->headers[i];
336 /* NAME ": " VALUE "\r\n" */
337 size += strlen (hdr->name) + 2 + strlen (hdr->value) + 2;
343 p = request_string = alloca_array (char, size);
345 /* Generate the request. */
347 APPEND (p, req->method); *p++ = ' ';
348 APPEND (p, req->arg); *p++ = ' ';
349 memcpy (p, "HTTP/1.1\r\n", 10); p += 10;
351 for (i = 0; i < req->hcount; i++)
353 struct request_header *hdr = &req->headers[i];
354 APPEND (p, hdr->name);
355 *p++ = ':', *p++ = ' ';
356 APPEND (p, hdr->value);
357 *p++ = '\r', *p++ = '\n';
360 *p++ = '\r', *p++ = '\n', *p++ = '\0';
361 assert (p - request_string == size);
365 DEBUGP (("\n---request begin---\n%s---request end---\n", request_string));
367 /* Send the request to the server. */
369 write_error = fd_write (fd, request_string, size - 1, -1);
371 logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
373 else if (warc_tmp != NULL)
375 /* Write a copy of the data to the WARC record. */
376 int warc_tmp_written = fwrite (request_string, 1, size - 1, warc_tmp);
377 if (warc_tmp_written != size - 1)
383 /* Release the resources used by REQ. */
386 request_free (struct request *req)
389 xfree_null (req->arg);
390 for (i = 0; i < req->hcount; i++)
391 release_header (&req->headers[i]);
392 xfree_null (req->headers);
396 static struct hash_table *basic_authed_hosts;
398 /* Find out if this host has issued a Basic challenge yet; if so, give
399 * it the username, password. A temporary measure until we can get
400 * proper authentication in place. */
403 maybe_send_basic_creds (const char *hostname, const char *user,
404 const char *passwd, struct request *req)
406 bool do_challenge = false;
408 if (opt.auth_without_challenge)
410 DEBUGP (("Auth-without-challenge set, sending Basic credentials.\n"));
413 else if (basic_authed_hosts
414 && hash_table_contains(basic_authed_hosts, hostname))
416 DEBUGP (("Found %s in basic_authed_hosts.\n", quote (hostname)));
421 DEBUGP (("Host %s has not issued a general basic challenge.\n",
426 request_set_header (req, "Authorization",
427 basic_authentication_encode (user, passwd),
434 register_basic_auth_host (const char *hostname)
436 if (!basic_authed_hosts)
438 basic_authed_hosts = make_nocase_string_hash_table (1);
440 if (!hash_table_contains(basic_authed_hosts, hostname))
442 hash_table_put (basic_authed_hosts, xstrdup(hostname), NULL);
443 DEBUGP (("Inserted %s into basic_authed_hosts\n", quote (hostname)));
448 /* Send the contents of FILE_NAME to SOCK. Make sure that exactly
449 PROMISED_SIZE bytes are sent over the wire -- if the file is
450 longer, read only that much; if the file is shorter, report an error.
451 If warc_tmp is set to a file pointer, the post data will
452 also be written to that file. */
455 body_file_send (int sock, const char *file_name, wgint promised_size, FILE *warc_tmp)
457 static char chunk[8192];
462 DEBUGP (("[writing BODY file %s ... ", file_name));
464 fp = fopen (file_name, "rb");
467 while (!feof (fp) && written < promised_size)
470 int length = fread (chunk, 1, sizeof (chunk), fp);
473 towrite = MIN (promised_size - written, length);
474 write_error = fd_write (sock, chunk, towrite, -1);
480 if (warc_tmp != NULL)
482 /* Write a copy of the data to the WARC record. */
483 int warc_tmp_written = fwrite (chunk, 1, towrite, warc_tmp);
484 if (warc_tmp_written != towrite)
494 /* If we've written less than was promised, report a (probably
495 nonsensical) error rather than break the promise. */
496 if (written < promised_size)
502 assert (written == promised_size);
503 DEBUGP (("done]\n"));
507 /* Determine whether [START, PEEKED + PEEKLEN) contains an empty line.
508 If so, return the pointer to the position after the line, otherwise
509 return NULL. This is used as callback to fd_read_hunk. The data
510 between START and PEEKED has been read and cannot be "unread"; the
511 data after PEEKED has only been peeked. */
514 response_head_terminator (const char *start, const char *peeked, int peeklen)
518 /* If at first peek, verify whether HUNK starts with "HTTP". If
519 not, this is a HTTP/0.9 request and we must bail out without
521 if (start == peeked && 0 != memcmp (start, "HTTP", MIN (peeklen, 4)))
524 /* Look for "\n[\r]\n", and return the following position if found.
525 Start two chars before the current to cover the possibility that
526 part of the terminator (e.g. "\n\r") arrived in the previous
528 p = peeked - start < 2 ? start : peeked - 2;
529 end = peeked + peeklen;
531 /* Check for \n\r\n or \n\n anywhere in [p, end-2). */
532 for (; p < end - 2; p++)
535 if (p[1] == '\r' && p[2] == '\n')
537 else if (p[1] == '\n')
540 /* p==end-2: check for \n\n directly preceding END. */
541 if (p[0] == '\n' && p[1] == '\n')
547 /* The maximum size of a single HTTP response we care to read. Rather
548 than being a limit of the reader implementation, this limit
549 prevents Wget from slurping all available memory upon encountering
550 malicious or buggy server output, thus protecting the user. Define
551 it to 0 to remove the limit. */
553 #define HTTP_RESPONSE_MAX_SIZE 65536
555 /* Read the HTTP request head from FD and return it. The error
556 conditions are the same as with fd_read_hunk.
558 To support HTTP/0.9 responses, this function tries to make sure
559 that the data begins with "HTTP". If this is not the case, no data
560 is read and an empty request is returned, so that the remaining
561 data can be treated as body. */
564 read_http_response_head (int fd)
566 return fd_read_hunk (fd, response_head_terminator, 512,
567 HTTP_RESPONSE_MAX_SIZE);
571 /* The response data. */
574 /* The array of pointers that indicate where each header starts.
575 For example, given this HTTP response:
582 The headers are located like this:
584 "HTTP/1.0 200 Ok\r\nDescription: some\r\n text\r\nEtag: x\r\n\r\n"
586 headers[0] headers[1] headers[2] headers[3]
588 I.e. headers[0] points to the beginning of the request,
589 headers[1] points to the end of the first header and the
590 beginning of the second one, etc. */
592 const char **headers;
595 /* Create a new response object from the text of the HTTP response,
596 available in HEAD. That text is automatically split into
597 constituent header lines for fast retrieval using
600 static struct response *
601 resp_new (const char *head)
606 struct response *resp = xnew0 (struct response);
611 /* Empty head means that we're dealing with a headerless
612 (HTTP/0.9) response. In that case, don't set HEADERS at
617 /* Split HEAD into header lines, so that resp_header_* functions
618 don't need to do this over and over again. */
624 DO_REALLOC (resp->headers, size, count + 1, const char *);
625 resp->headers[count++] = hdr;
627 /* Break upon encountering an empty line. */
628 if (!hdr[0] || (hdr[0] == '\r' && hdr[1] == '\n') || hdr[0] == '\n')
631 /* Find the end of HDR, including continuations. */
634 const char *end = strchr (hdr, '\n');
640 while (*hdr == ' ' || *hdr == '\t');
642 DO_REALLOC (resp->headers, size, count + 1, const char *);
643 resp->headers[count] = NULL;
648 /* Locate the header named NAME in the request data, starting with
649 position START. This allows the code to loop through the request
650 data, filtering for all requests of a given name. Returns the
651 found position, or -1 for failure. The code that uses this
652 function typically looks like this:
654 for (pos = 0; (pos = resp_header_locate (...)) != -1; pos++)
655 ... do something with header ...
657 If you only care about one header, use resp_header_get instead of
661 resp_header_locate (const struct response *resp, const char *name, int start,
662 const char **begptr, const char **endptr)
665 const char **headers = resp->headers;
668 if (!headers || !headers[1])
671 name_len = strlen (name);
677 for (; headers[i + 1]; i++)
679 const char *b = headers[i];
680 const char *e = headers[i + 1];
682 && b[name_len] == ':'
683 && 0 == strncasecmp (b, name, name_len))
686 while (b < e && c_isspace (*b))
688 while (b < e && c_isspace (e[-1]))
698 /* Find and retrieve the header named NAME in the request data. If
699 found, set *BEGPTR to its starting, and *ENDPTR to its ending
700 position, and return true. Otherwise return false.
702 This function is used as a building block for resp_header_copy
703 and resp_header_strdup. */
706 resp_header_get (const struct response *resp, const char *name,
707 const char **begptr, const char **endptr)
709 int pos = resp_header_locate (resp, name, 0, begptr, endptr);
713 /* Copy the response header named NAME to buffer BUF, no longer than
714 BUFSIZE (BUFSIZE includes the terminating 0). If the header
715 exists, true is returned, false otherwise. If there should be no
716 limit on the size of the header, use resp_header_strdup instead.
718 If BUFSIZE is 0, no data is copied, but the boolean indication of
719 whether the header is present is still returned. */
722 resp_header_copy (const struct response *resp, const char *name,
723 char *buf, int bufsize)
726 if (!resp_header_get (resp, name, &b, &e))
730 int len = MIN (e - b, bufsize - 1);
731 memcpy (buf, b, len);
737 /* Return the value of header named NAME in RESP, allocated with
738 malloc. If such a header does not exist in RESP, return NULL. */
741 resp_header_strdup (const struct response *resp, const char *name)
744 if (!resp_header_get (resp, name, &b, &e))
746 return strdupdelim (b, e);
749 /* Parse the HTTP status line, which is of format:
751 HTTP-Version SP Status-Code SP Reason-Phrase
753 The function returns the status-code, or -1 if the status line
754 appears malformed. The pointer to "reason-phrase" message is
755 returned in *MESSAGE. */
758 resp_status (const struct response *resp, char **message)
765 /* For a HTTP/0.9 response, assume status 200. */
767 *message = xstrdup (_("No headers, assuming HTTP/0.9"));
771 p = resp->headers[0];
772 end = resp->headers[1];
778 if (end - p < 4 || 0 != strncmp (p, "HTTP", 4))
782 /* Match the HTTP version. This is optional because Gnutella
783 servers have been reported to not specify HTTP version. */
784 if (p < end && *p == '/')
787 while (p < end && c_isdigit (*p))
789 if (p < end && *p == '.')
791 while (p < end && c_isdigit (*p))
795 while (p < end && c_isspace (*p))
797 if (end - p < 3 || !c_isdigit (p[0]) || !c_isdigit (p[1]) || !c_isdigit (p[2]))
800 status = 100 * (p[0] - '0') + 10 * (p[1] - '0') + (p[2] - '0');
805 while (p < end && c_isspace (*p))
807 while (p < end && c_isspace (end[-1]))
809 *message = strdupdelim (p, end);
815 /* Release the resources used by RESP. */
818 resp_free (struct response *resp)
820 xfree_null (resp->headers);
824 /* Print a single line of response, the characters [b, e). We tried
826 logprintf (LOG_VERBOSE, "%s%.*s\n", prefix, (int) (e - b), b);
827 but that failed to escape the non-printable characters and, in fact,
828 caused crashes in UTF-8 locales. */
831 print_response_line(const char *prefix, const char *b, const char *e)
834 BOUNDED_TO_ALLOCA(b, e, copy);
835 logprintf (LOG_ALWAYS, "%s%s\n", prefix,
836 quotearg_style (escape_quoting_style, copy));
839 /* Print the server response, line by line, omitting the trailing CRLF
840 from individual header lines, and prefixed with PREFIX. */
843 print_server_response (const struct response *resp, const char *prefix)
848 for (i = 0; resp->headers[i + 1]; i++)
850 const char *b = resp->headers[i];
851 const char *e = resp->headers[i + 1];
853 if (b < e && e[-1] == '\n')
855 if (b < e && e[-1] == '\r')
857 print_response_line(prefix, b, e);
861 /* Parse the `Content-Range' header and extract the information it
862 contains. Returns true if successful, false otherwise. */
864 parse_content_range (const char *hdr, wgint *first_byte_ptr,
865 wgint *last_byte_ptr, wgint *entity_length_ptr)
869 /* Ancient versions of Netscape proxy server, presumably predating
870 rfc2068, sent out `Content-Range' without the "bytes"
872 if (0 == strncasecmp (hdr, "bytes", 5))
875 /* "JavaWebServer/1.1.1" sends "bytes: x-y/z", contrary to the
879 while (c_isspace (*hdr))
884 if (!c_isdigit (*hdr))
886 for (num = 0; c_isdigit (*hdr); hdr++)
887 num = 10 * num + (*hdr - '0');
888 if (*hdr != '-' || !c_isdigit (*(hdr + 1)))
890 *first_byte_ptr = num;
892 for (num = 0; c_isdigit (*hdr); hdr++)
893 num = 10 * num + (*hdr - '0');
894 if (*hdr != '/' || !c_isdigit (*(hdr + 1)))
896 *last_byte_ptr = num;
901 for (num = 0; c_isdigit (*hdr); hdr++)
902 num = 10 * num + (*hdr - '0');
903 *entity_length_ptr = num;
907 /* Read the body of the request, but don't store it anywhere and don't
908 display a progress gauge. This is useful for reading the bodies of
909 administrative responses to which we will soon issue another
910 request. The response is not useful to the user, but reading it
911 allows us to continue using the same connection to the server.
913 If reading fails, false is returned, true otherwise. In debug
914 mode, the body is displayed for debugging purposes. */
917 skip_short_body (int fd, wgint contlen, bool chunked)
920 SKIP_SIZE = 512, /* size of the download buffer */
921 SKIP_THRESHOLD = 4096 /* the largest size we read */
923 wgint remaining_chunk_size = 0;
924 char dlbuf[SKIP_SIZE + 1];
925 dlbuf[SKIP_SIZE] = '\0'; /* so DEBUGP can safely print it */
927 /* If the body is too large, it makes more sense to simply close the
928 connection than to try to read the body. */
929 if (contlen > SKIP_THRESHOLD)
932 while (contlen > 0 || chunked)
937 if (remaining_chunk_size == 0)
939 char *line = fd_read_line (fd);
944 remaining_chunk_size = strtol (line, &endl, 16);
947 if (remaining_chunk_size == 0)
949 line = fd_read_line (fd);
955 contlen = MIN (remaining_chunk_size, SKIP_SIZE);
958 DEBUGP (("Skipping %s bytes of body: [", number_to_static_string (contlen)));
960 ret = fd_read (fd, dlbuf, MIN (contlen, SKIP_SIZE), -1);
963 /* Don't normally report the error since this is an
964 optimization that should be invisible to the user. */
965 DEBUGP (("] aborting (%s).\n",
966 ret < 0 ? fd_errstr (fd) : "EOF received"));
973 remaining_chunk_size -= ret;
974 if (remaining_chunk_size == 0)
976 char *line = fd_read_line (fd);
984 /* Safe even if %.*s bogusly expects terminating \0 because
985 we've zero-terminated dlbuf above. */
986 DEBUGP (("%.*s", ret, dlbuf));
989 DEBUGP (("] done.\n"));
993 #define NOT_RFC2231 0
994 #define RFC2231_NOENCODING 1
995 #define RFC2231_ENCODING 2
997 /* extract_param extracts the parameter name into NAME.
998 However, if the parameter name is in RFC2231 format then
999 this function adjusts NAME by stripping of the trailing
1000 characters that are not part of the name but are present to
1001 indicate the presence of encoding information in the value
1002 or a fragment of a long parameter value
1005 modify_param_name(param_token *name)
1007 const char *delim1 = memchr (name->b, '*', name->e - name->b);
1008 const char *delim2 = memrchr (name->b, '*', name->e - name->b);
1014 result = NOT_RFC2231;
1016 else if(delim1 == delim2)
1018 if ((name->e - 1) == delim1)
1020 result = RFC2231_ENCODING;
1024 result = RFC2231_NOENCODING;
1031 result = RFC2231_ENCODING;
1036 /* extract_param extract the paramater value into VALUE.
1037 Like modify_param_name this function modifies VALUE by
1038 stripping off the encoding information from the actual value
1041 modify_param_value (param_token *value, int encoding_type )
1043 if (encoding_type == RFC2231_ENCODING)
1045 const char *delim = memrchr (value->b, '\'', value->e - value->b);
1046 if ( delim != NULL )
1048 value->b = (delim+1);
1053 /* Extract a parameter from the string (typically an HTTP header) at
1054 **SOURCE and advance SOURCE to the next parameter. Return false
1055 when there are no more parameters to extract. The name of the
1056 parameter is returned in NAME, and the value in VALUE. If the
1057 parameter has no value, the token's value is zeroed out.
1059 For example, if *SOURCE points to the string "attachment;
1060 filename=\"foo bar\"", the first call to this function will return
1061 the token named "attachment" and no value, and the second call will
1062 return the token named "filename" and value "foo bar". The third
1063 call will return false, indicating no more valid tokens.
1065 is_url_encoded is an out parameter. If not NULL, a boolean value will be
1066 stored into it, letting the caller know whether or not the extracted value is
1067 URL-encoded. The caller can then decode it with url_unescape(), which however
1068 performs decoding in-place. URL-encoding is used by RFC 2231 to support
1069 non-US-ASCII characters in HTTP header values. */
1072 extract_param (const char **source, param_token *name, param_token *value,
1073 char separator, bool *is_url_encoded)
1075 const char *p = *source;
1078 *is_url_encoded = false; /* initializing the out parameter */
1080 while (c_isspace (*p)) ++p;
1084 return false; /* no error; nothing more to extract */
1089 while (*p && !c_isspace (*p) && *p != '=' && *p != separator) ++p;
1091 if (name->b == name->e)
1092 return false; /* empty name: error */
1093 while (c_isspace (*p)) ++p;
1094 if (*p == separator || !*p) /* no value */
1097 if (*p == separator) ++p;
1102 return false; /* error */
1104 /* *p is '=', extract value */
1106 while (c_isspace (*p)) ++p;
1107 if (*p == '"') /* quoted */
1110 while (*p && *p != '"') ++p;
1114 /* Currently at closing quote; find the end of param. */
1115 while (c_isspace (*p)) ++p;
1116 while (*p && *p != separator) ++p;
1117 if (*p == separator)
1120 /* garbage after closed quote, e.g. foo="bar"baz */
1126 while (*p && *p != separator) ++p;
1128 while (value->e != value->b && c_isspace (value->e[-1]))
1130 if (*p == separator) ++p;
1134 param_type = modify_param_name(name);
1135 if (param_type != NOT_RFC2231)
1137 if (param_type == RFC2231_ENCODING && is_url_encoded)
1138 *is_url_encoded = true;
1139 modify_param_value(value, param_type);
1145 #undef RFC2231_NOENCODING
1146 #undef RFC2231_ENCODING
1148 /* Appends the string represented by VALUE to FILENAME */
1151 append_value_to_filename (char **filename, param_token const * const value,
1152 bool is_url_encoded)
1154 int original_length = strlen(*filename);
1155 int new_length = strlen(*filename) + (value->e - value->b);
1156 *filename = xrealloc (*filename, new_length+1);
1157 memcpy (*filename + original_length, value->b, (value->e - value->b));
1158 (*filename)[new_length] = '\0';
1160 url_unescape (*filename + original_length);
1164 #define MAX(p, q) ((p) > (q) ? (p) : (q))
1166 /* Parse the contents of the `Content-Disposition' header, extracting
1167 the information useful to Wget. Content-Disposition is a header
1168 borrowed from MIME; when used in HTTP, it typically serves for
1169 specifying the desired file name of the resource. For example:
1171 Content-Disposition: attachment; filename="flora.jpg"
1173 Wget will skip the tokens it doesn't care about, such as
1174 "attachment" in the previous example; it will also skip other
1175 unrecognized params. If the header is syntactically correct and
1176 contains a file name, a copy of the file name is stored in
1177 *filename and true is returned. Otherwise, the function returns
1180 The file name is stripped of directory components and must not be
1183 Historically, this function returned filename prefixed with opt.dir_prefix,
1184 now that logic is handled by the caller, new code should pay attention,
1185 changed by crq, Sep 2010.
1189 parse_content_disposition (const char *hdr, char **filename)
1191 param_token name, value;
1193 bool is_url_encoded = false;
1194 for ( ; extract_param (&hdr, &name, &value, ';', &is_url_encoded);
1195 is_url_encoded = false)
1197 int isFilename = BOUNDED_EQUAL_NO_CASE ( name.b, name.e, "filename" );
1198 if ( isFilename && value.b != NULL)
1200 /* Make the file name begin at the last slash or backslash. */
1201 const char *last_slash = memrchr (value.b, '/', value.e - value.b);
1202 const char *last_bs = memrchr (value.b, '\\', value.e - value.b);
1203 if (last_slash && last_bs)
1204 value.b = 1 + MAX (last_slash, last_bs);
1205 else if (last_slash || last_bs)
1206 value.b = 1 + (last_slash ? last_slash : last_bs);
1207 if (value.b == value.e)
1211 append_value_to_filename (filename, &value, is_url_encoded);
1214 *filename = strdupdelim (value.b, value.e);
1216 url_unescape (*filename);
1228 /* Persistent connections. Currently, we cache the most recently used
1229 connection as persistent, provided that the HTTP server agrees to
1230 make it such. The persistence data is stored in the variables
1231 below. Ideally, it should be possible to cache an arbitrary fixed
1232 number of these connections. */
1234 /* Whether a persistent connection is active. */
1235 static bool pconn_active;
1238 /* The socket of the connection. */
1241 /* Host and port of the currently active persistent connection. */
1245 /* Whether a ssl handshake has occoured on this connection. */
1248 /* Whether the connection was authorized. This is only done by
1249 NTLM, which authorizes *connections* rather than individual
1250 requests. (That practice is peculiar for HTTP, but it is a
1251 useful optimization.) */
1255 /* NTLM data of the current connection. */
1256 struct ntlmdata ntlm;
1260 /* Mark the persistent connection as invalid and free the resources it
1261 uses. This is used by the CLOSE_* macros after they forcefully
1262 close a registered persistent connection. */
1265 invalidate_persistent (void)
1267 DEBUGP (("Disabling further reuse of socket %d.\n", pconn.socket));
1268 pconn_active = false;
1269 fd_close (pconn.socket);
1274 /* Register FD, which should be a TCP/IP connection to HOST:PORT, as
1275 persistent. This will enable someone to use the same connection
1276 later. In the context of HTTP, this must be called only AFTER the
1277 response has been received and the server has promised that the
1278 connection will remain alive.
1280 If a previous connection was persistent, it is closed. */
1283 register_persistent (const char *host, int port, int fd, bool ssl)
1287 if (pconn.socket == fd)
1289 /* The connection FD is already registered. */
1294 /* The old persistent connection is still active; close it
1295 first. This situation arises whenever a persistent
1296 connection exists, but we then connect to a different
1297 host, and try to register a persistent connection to that
1299 invalidate_persistent ();
1303 pconn_active = true;
1305 pconn.host = xstrdup (host);
1308 pconn.authorized = false;
1310 DEBUGP (("Registered socket %d for persistent reuse.\n", fd));
1313 /* Return true if a persistent connection is available for connecting
1317 persistent_available_p (const char *host, int port, bool ssl,
1318 bool *host_lookup_failed)
1320 /* First, check whether a persistent connection is active at all. */
1324 /* If we want SSL and the last connection wasn't or vice versa,
1325 don't use it. Checking for host and port is not enough because
1326 HTTP and HTTPS can apparently coexist on the same port. */
1327 if (ssl != pconn.ssl)
1330 /* If we're not connecting to the same port, we're not interested. */
1331 if (port != pconn.port)
1334 /* If the host is the same, we're in business. If not, there is
1335 still hope -- read below. */
1336 if (0 != strcasecmp (host, pconn.host))
1338 /* Check if pconn.socket is talking to HOST under another name.
1339 This happens often when both sites are virtual hosts
1340 distinguished only by name and served by the same network
1341 interface, and hence the same web server (possibly set up by
1342 the ISP and serving many different web sites). This
1343 admittedly unconventional optimization does not contradict
1344 HTTP and works well with popular server software. */
1348 struct address_list *al;
1351 /* Don't try to talk to two different SSL sites over the same
1352 secure connection! (Besides, it's not clear that
1353 name-based virtual hosting is even possible with SSL.) */
1356 /* If pconn.socket's peer is one of the IP addresses HOST
1357 resolves to, pconn.socket is for all intents and purposes
1358 already talking to HOST. */
1360 if (!socket_ip_address (pconn.socket, &ip, ENDPOINT_PEER))
1362 /* Can't get the peer's address -- something must be very
1363 wrong with the connection. */
1364 invalidate_persistent ();
1367 al = lookup_host (host, 0);
1370 *host_lookup_failed = true;
1374 found = address_list_contains (al, &ip);
1375 address_list_release (al);
1380 /* The persistent connection's peer address was found among the
1381 addresses HOST resolved to; therefore, pconn.sock is in fact
1382 already talking to HOST -- no need to reconnect. */
1385 /* Finally, check whether the connection is still open. This is
1386 important because most servers implement liberal (short) timeout
1387 on persistent connections. Wget can of course always reconnect
1388 if the connection doesn't work out, but it's nicer to know in
1389 advance. This test is a logical followup of the first test, but
1390 is "expensive" and therefore placed at the end of the list.
1392 (Current implementation of test_socket_open has a nice side
1393 effect that it treats sockets with pending data as "closed".
1394 This is exactly what we want: if a broken server sends message
1395 body in response to HEAD, or if it sends more than conent-length
1396 data, we won't reuse the corrupted connection.) */
1398 if (!test_socket_open (pconn.socket))
1400 /* Oops, the socket is no longer open. Now that we know that,
1401 let's invalidate the persistent connection before returning
1403 invalidate_persistent ();
1410 /* The idea behind these two CLOSE macros is to distinguish between
1411 two cases: one when the job we've been doing is finished, and we
1412 want to close the connection and leave, and two when something is
1413 seriously wrong and we're closing the connection as part of
1416 In case of keep_alive, CLOSE_FINISH should leave the connection
1417 open, while CLOSE_INVALIDATE should still close it.
1419 Note that the semantics of the flag `keep_alive' is "this
1420 connection *will* be reused (the server has promised not to close
1421 the connection once we're done)", while the semantics of
1422 `pc_active_p && (fd) == pc_last_fd' is "we're *now* using an
1423 active, registered connection". */
1425 #define CLOSE_FINISH(fd) do { \
1428 if (pconn_active && (fd) == pconn.socket) \
1429 invalidate_persistent (); \
1438 #define CLOSE_INVALIDATE(fd) do { \
1439 if (pconn_active && (fd) == pconn.socket) \
1440 invalidate_persistent (); \
1448 wgint len; /* received length */
1449 wgint contlen; /* expected length */
1450 wgint restval; /* the restart value */
1451 int res; /* the result of last read */
1452 char *rderrmsg; /* error message from read error */
1453 char *newloc; /* new location (redirection) */
1454 char *remote_time; /* remote time-stamp string */
1455 char *error; /* textual HTTP error */
1456 int statcode; /* status code */
1457 char *message; /* status message */
1458 wgint rd_size; /* amount of data read from socket */
1459 double dltime; /* time it took to download the data */
1460 const char *referer; /* value of the referer header. */
1461 char *local_file; /* local file name. */
1462 bool existence_checked; /* true if we already checked for a file's
1463 existence after having begun to download
1464 (needed in gethttp for when connection is
1465 interrupted/restarted. */
1466 bool timestamp_checked; /* true if pre-download time-stamping checks
1467 * have already been performed */
1468 char *orig_file_name; /* name of file to compare for time-stamping
1469 * (might be != local_file if -K is set) */
1470 wgint orig_file_size; /* size of file to compare for time-stamping */
1471 time_t orig_file_tstamp; /* time-stamp of file to compare for
1476 free_hstat (struct http_stat *hs)
1478 xfree_null (hs->newloc);
1479 xfree_null (hs->remote_time);
1480 xfree_null (hs->error);
1481 xfree_null (hs->rderrmsg);
1482 xfree_null (hs->local_file);
1483 xfree_null (hs->orig_file_name);
1484 xfree_null (hs->message);
1486 /* Guard against being called twice. */
1488 hs->remote_time = NULL;
1493 get_file_flags (const char *filename, int *dt)
1495 logprintf (LOG_VERBOSE, _("\
1496 File %s already there; not retrieving.\n\n"), quote (filename));
1497 /* If the file is there, we suppose it's retrieved OK. */
1500 /* #### Bogusness alert. */
1501 /* If its suffix is "html" or "htm" or similar, assume text/html. */
1502 if (has_html_suffix_p (filename))
1506 /* Download the response body from the socket and writes it to
1507 an output file. The headers have already been read from the
1508 socket. If WARC is enabled, the response body will also be
1509 written to a WARC response record.
1511 hs, contlen, contrange, chunked_transfer_encoding and url are
1512 parameters from the gethttp method. fp is a pointer to the
1515 url, warc_timestamp_str, warc_request_uuid, warc_ip, type
1516 and statcode will be saved in the headers of the WARC record.
1517 The head parameter contains the HTTP headers of the response.
1519 If fp is NULL and WARC is enabled, the response body will be
1520 written only to the WARC file. If WARC is disabled and fp
1521 is a file pointer, the data will be written to the file.
1522 If fp is a file pointer and WARC is enabled, the body will
1523 be written to both destinations.
1525 Returns the error code. */
1527 read_response_body (struct http_stat *hs, int sock, FILE *fp, wgint contlen,
1528 wgint contrange, bool chunked_transfer_encoding,
1529 char *url, char *warc_timestamp_str, char *warc_request_uuid,
1530 ip_address *warc_ip, char *type, int statcode, char *head)
1532 int warc_payload_offset = 0;
1533 FILE *warc_tmp = NULL;
1536 if (opt.warc_filename != NULL)
1538 /* Open a temporary file where we can write the response before we
1539 add it to the WARC record. */
1540 warc_tmp = warc_tempfile ();
1541 if (warc_tmp == NULL)
1542 warcerr = WARC_TMP_FOPENERR;
1546 /* We should keep the response headers for the WARC record. */
1547 int head_len = strlen (head);
1548 int warc_tmp_written = fwrite (head, 1, head_len, warc_tmp);
1549 if (warc_tmp_written != head_len)
1550 warcerr = WARC_TMP_FWRITEERR;
1551 warc_payload_offset = head_len;
1556 if (warc_tmp != NULL)
1564 /* This confuses the timestamping code that checks for file size.
1565 #### The timestamping code should be smarter about file size. */
1566 if (opt.save_headers && hs->restval == 0)
1567 fwrite (head, 1, strlen (head), fp);
1570 /* Read the response body. */
1573 /* If content-length is present, read that much; otherwise, read
1574 until EOF. The HTTP spec doesn't require the server to
1575 actually close the connection when it's done sending data. */
1576 flags |= rb_read_exactly;
1577 if (fp != NULL && hs->restval > 0 && contrange == 0)
1578 /* If the server ignored our range request, instruct fd_read_body
1579 to skip the first RESTVAL bytes of body. */
1580 flags |= rb_skip_startpos;
1581 if (chunked_transfer_encoding)
1582 flags |= rb_chunked_transfer_encoding;
1584 hs->len = hs->restval;
1586 /* Download the response body and write it to fp.
1587 If we are working on a WARC file, we simultaneously write the
1588 response body to warc_tmp. */
1589 hs->res = fd_read_body (hs->local_file, sock, fp, contlen != -1 ? contlen : 0,
1590 hs->restval, &hs->rd_size, &hs->len, &hs->dltime,
1594 if (warc_tmp != NULL)
1596 /* Create a response record and write it to the WARC file.
1597 Note: per the WARC standard, the request and response should share
1598 the same date header. We re-use the timestamp of the request.
1599 The response record should also refer to the uuid of the request. */
1600 bool r = warc_write_response_record (url, warc_timestamp_str,
1601 warc_request_uuid, warc_ip,
1602 warc_tmp, warc_payload_offset,
1603 type, statcode, hs->newloc);
1605 /* warc_write_response_record has closed warc_tmp. */
1611 return RETRFINISHED;
1614 if (warc_tmp != NULL)
1619 /* Error while writing to fd. */
1622 else if (hs->res == -3)
1624 /* Error while writing to warc_tmp. */
1625 return WARC_TMP_FWRITEERR;
1630 hs->rderrmsg = xstrdup (fd_errstr (sock));
1631 return RETRFINISHED;
1635 #define BEGINS_WITH(line, string_constant) \
1636 (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
1637 && (c_isspace (line[sizeof (string_constant) - 1]) \
1638 || !line[sizeof (string_constant) - 1]))
1641 #define SET_USER_AGENT(req) do { \
1642 if (!opt.useragent) \
1643 request_set_header (req, "User-Agent", \
1644 aprintf ("Wget/%s (VMS %s %s)", \
1645 version_string, vms_arch(), vms_vers()), \
1647 else if (*opt.useragent) \
1648 request_set_header (req, "User-Agent", opt.useragent, rel_none); \
1650 #else /* def __VMS */
1651 #define SET_USER_AGENT(req) do { \
1652 if (!opt.useragent) \
1653 request_set_header (req, "User-Agent", \
1654 aprintf ("Wget/%s (%s)", \
1655 version_string, OS_TYPE), \
1657 else if (*opt.useragent) \
1658 request_set_header (req, "User-Agent", opt.useragent, rel_none); \
1660 #endif /* def __VMS [else] */
1662 /* Retrieve a document through HTTP protocol. It recognizes status
1663 code, and correctly handles redirections. It closes the network
1664 socket. If it receives an error from the functions below it, it
1665 will print it if there is enough information to do so (almost
1666 always), returning the error to the caller (i.e. http_loop).
1668 Various HTTP parameters are stored to hs.
1670 If PROXY is non-NULL, the connection will be made to the proxy
1671 server, and u->url will be requested. */
1673 gethttp (struct url *u, struct http_stat *hs, int *dt, struct url *proxy,
1674 struct iri *iri, int count)
1676 struct request *req;
1679 char *user, *passwd;
1683 wgint contlen, contrange;
1690 /* Set to 1 when the authorization has already been sent and should
1691 not be tried again. */
1692 bool auth_finished = false;
1694 /* Set to 1 when just globally-set Basic authorization has been sent;
1695 * should prevent further Basic negotiations, but not other
1697 bool basic_auth_finished = false;
1699 /* Whether NTLM authentication is used for this request. */
1700 bool ntlm_seen = false;
1702 /* Whether our connection to the remote host is through SSL. */
1703 bool using_ssl = false;
1705 /* Whether a HEAD request will be issued (as opposed to GET or
1707 bool head_only = !!(*dt & HEAD_ONLY);
1710 struct response *resp;
1714 /* Declare WARC variables. */
1715 bool warc_enabled = (opt.warc_filename != NULL);
1716 FILE *warc_tmp = NULL;
1717 char warc_timestamp_str [21];
1718 char warc_request_uuid [48];
1719 ip_address *warc_ip = NULL;
1720 off_t warc_payload_offset = -1;
1722 /* Whether this connection will be kept alive after the HTTP request
1726 /* Is the server using the chunked transfer encoding? */
1727 bool chunked_transfer_encoding = false;
1729 /* Whether keep-alive should be inhibited. */
1730 bool inhibit_keep_alive =
1731 !opt.http_keep_alive || opt.ignore_length;
1733 /* Headers sent when using POST. */
1734 wgint body_data_size = 0;
1736 bool host_lookup_failed = false;
1739 if (u->scheme == SCHEME_HTTPS)
1741 /* Initialize the SSL context. After this has once been done,
1742 it becomes a no-op. */
1745 scheme_disable (SCHEME_HTTPS);
1746 logprintf (LOG_NOTQUIET,
1747 _("Disabling SSL due to encountered errors.\n"));
1748 return SSLINITFAILED;
1751 #endif /* HAVE_SSL */
1753 /* Initialize certain elements of struct http_stat. */
1757 hs->rderrmsg = NULL;
1759 hs->remote_time = NULL;
1765 /* Prepare the request to send. */
1768 const char *meth = "GET";
1771 else if (opt.method)
1773 /* Use the full path, i.e. one that includes the leading slash and
1774 the query string. E.g. if u->path is "foo/bar" and u->query is
1775 "param=value", full_path will be "/foo/bar?param=value". */
1778 /* When using SSL over proxy, CONNECT establishes a direct
1779 connection to the HTTPS server. Therefore use the same
1780 argument as when talking to the server directly. */
1781 && u->scheme != SCHEME_HTTPS
1784 meth_arg = xstrdup (u->url);
1786 meth_arg = url_full_path (u);
1787 req = request_new (meth, meth_arg);
1790 request_set_header (req, "Referer", (char *) hs->referer, rel_none);
1791 if (*dt & SEND_NOCACHE)
1793 /* Cache-Control MUST be obeyed by all HTTP/1.1 caching mechanisms... */
1794 request_set_header (req, "Cache-Control", "no-cache, must-revalidate", rel_none);
1796 /* ... but some HTTP/1.0 caches doesn't implement Cache-Control. */
1797 request_set_header (req, "Pragma", "no-cache", rel_none);
1800 request_set_header (req, "Range",
1801 aprintf ("bytes=%s-",
1802 number_to_static_string (hs->restval)),
1804 SET_USER_AGENT (req);
1805 request_set_header (req, "Accept", "*/*", rel_none);
1807 /* Find the username and password for authentication. */
1810 search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
1811 user = user ? user : (opt.http_user ? opt.http_user : opt.user);
1812 passwd = passwd ? passwd : (opt.http_passwd ? opt.http_passwd : opt.passwd);
1814 /* We only do "site-wide" authentication with "global" user/password
1815 * values unless --auth-no-challange has been requested; URL user/password
1816 * info overrides. */
1817 if (user && passwd && (!u->user || opt.auth_without_challenge))
1819 /* If this is a host for which we've already received a Basic
1820 * challenge, we'll go ahead and send Basic authentication creds. */
1821 basic_auth_finished = maybe_send_basic_creds(u->host, user, passwd, req);
1824 /* Generate the Host header, HOST:PORT. Take into account that:
1826 - Broken server-side software often doesn't recognize the PORT
1827 argument, so we must generate "Host: www.server.com" instead of
1828 "Host: www.server.com:80" (and likewise for https port).
1830 - IPv6 addresses contain ":", so "Host: 3ffe:8100:200:2::2:1234"
1831 becomes ambiguous and needs to be rewritten as "Host:
1832 [3ffe:8100:200:2::2]:1234". */
1834 /* Formats arranged for hfmt[add_port][add_squares]. */
1835 static const char *hfmt[][2] = {
1836 { "%s", "[%s]" }, { "%s:%d", "[%s]:%d" }
1838 int add_port = u->port != scheme_default_port (u->scheme);
1839 int add_squares = strchr (u->host, ':') != NULL;
1840 request_set_header (req, "Host",
1841 aprintf (hfmt[add_port][add_squares], u->host, u->port),
1845 if (inhibit_keep_alive)
1846 request_set_header (req, "Connection", "Close", rel_none);
1850 request_set_header (req, "Connection", "Keep-Alive", rel_none);
1853 request_set_header (req, "Connection", "Close", rel_none);
1854 request_set_header (req, "Proxy-Connection", "Keep-Alive", rel_none);
1861 if (opt.body_data || opt.body_file)
1863 request_set_header (req, "Content-Type",
1864 "application/x-www-form-urlencoded", rel_none);
1867 body_data_size = strlen (opt.body_data);
1870 body_data_size = file_size (opt.body_file);
1871 if (body_data_size == -1)
1873 logprintf (LOG_NOTQUIET, _("BODY data file %s missing: %s\n"),
1874 quote (opt.body_file), strerror (errno));
1878 request_set_header (req, "Content-Length",
1879 xstrdup (number_to_static_string (body_data_size)),
1885 /* We need to come back here when the initial attempt to retrieve
1886 without authorization header fails. (Expected to happen at least
1887 for the Digest authorization scheme.) */
1890 request_set_header (req, "Cookie",
1891 cookie_header (wget_cookie_jar,
1892 u->host, u->port, u->path,
1894 u->scheme == SCHEME_HTTPS
1901 /* Add the user headers. */
1902 if (opt.user_headers)
1905 for (i = 0; opt.user_headers[i]; i++)
1906 request_set_user_header (req, opt.user_headers[i]);
1912 char *proxy_user, *proxy_passwd;
1913 /* For normal username and password, URL components override
1914 command-line/wgetrc parameters. With proxy
1915 authentication, it's the reverse, because proxy URLs are
1916 normally the "permanent" ones, so command-line args
1917 should take precedence. */
1918 if (opt.proxy_user && opt.proxy_passwd)
1920 proxy_user = opt.proxy_user;
1921 proxy_passwd = opt.proxy_passwd;
1925 proxy_user = proxy->user;
1926 proxy_passwd = proxy->passwd;
1928 /* #### This does not appear right. Can't the proxy request,
1929 say, `Digest' authentication? */
1930 if (proxy_user && proxy_passwd)
1931 proxyauth = basic_authentication_encode (proxy_user, proxy_passwd);
1933 /* If we're using a proxy, we will be connecting to the proxy
1937 /* Proxy authorization over SSL is handled below. */
1939 if (u->scheme != SCHEME_HTTPS)
1941 request_set_header (req, "Proxy-Authorization", proxyauth, rel_value);
1946 /* Establish the connection. */
1948 if (inhibit_keep_alive)
1952 /* Look for a persistent connection to target host, unless a
1953 proxy is used. The exception is when SSL is in use, in which
1954 case the proxy is nothing but a passthrough to the target
1955 host, registered as a connection to the latter. */
1956 struct url *relevant = conn;
1958 if (u->scheme == SCHEME_HTTPS)
1962 if (persistent_available_p (relevant->host, relevant->port,
1964 relevant->scheme == SCHEME_HTTPS,
1968 &host_lookup_failed))
1970 int family = socket_family (pconn.socket, ENDPOINT_PEER);
1971 sock = pconn.socket;
1972 using_ssl = pconn.ssl;
1974 if (family == AF_INET6)
1975 logprintf (LOG_VERBOSE, _("Reusing existing connection to [%s]:%d.\n"),
1976 quotearg_style (escape_quoting_style, pconn.host),
1980 logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
1981 quotearg_style (escape_quoting_style, pconn.host),
1983 DEBUGP (("Reusing fd %d.\n", sock));
1984 if (pconn.authorized)
1985 /* If the connection is already authorized, the "Basic"
1986 authorization added by code above is unnecessary and
1988 request_remove_header (req, "Authorization");
1990 else if (host_lookup_failed)
1993 logprintf(LOG_NOTQUIET,
1994 _("%s: unable to resolve host address %s\n"),
1995 exec_name, quote (relevant->host));
1998 else if (sock != -1)
2006 sock = connect_to_host (conn->host, conn->port);
2015 return (retryable_socket_connect_error (errno)
2016 ? CONERROR : CONIMPOSSIBLE);
2020 if (proxy && u->scheme == SCHEME_HTTPS)
2022 /* When requesting SSL URLs through proxies, use the
2023 CONNECT method to request passthrough. */
2024 struct request *connreq = request_new ("CONNECT",
2025 aprintf ("%s:%d", u->host, u->port));
2026 SET_USER_AGENT (connreq);
2029 request_set_header (connreq, "Proxy-Authorization",
2030 proxyauth, rel_value);
2031 /* Now that PROXYAUTH is part of the CONNECT request,
2032 zero it out so we don't send proxy authorization with
2033 the regular request below. */
2036 request_set_header (connreq, "Host",
2037 aprintf ("%s:%d", u->host, u->port),
2040 write_error = request_send (connreq, sock, 0);
2041 request_free (connreq);
2042 if (write_error < 0)
2044 CLOSE_INVALIDATE (sock);
2049 head = read_http_response_head (sock);
2052 logprintf (LOG_VERBOSE, _("Failed reading proxy response: %s\n"),
2054 CLOSE_INVALIDATE (sock);
2064 DEBUGP (("proxy responded with: [%s]\n", head));
2066 resp = resp_new (head);
2067 statcode = resp_status (resp, &message);
2070 char *tms = datetime_str (time (NULL));
2071 logprintf (LOG_VERBOSE, "%d\n", statcode);
2072 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"), tms, statcode,
2073 quotearg_style (escape_quoting_style,
2074 _("Malformed status line")));
2079 hs->message = xstrdup (message);
2082 if (statcode != 200)
2085 logprintf (LOG_NOTQUIET, _("Proxy tunneling failed: %s"),
2086 message ? quotearg_style (escape_quoting_style, message) : "?");
2087 xfree_null (message);
2091 xfree_null (message);
2093 /* SOCK is now *really* connected to u->host, so update CONN
2094 to reflect this. That way register_persistent will
2095 register SOCK as being connected to u->host:u->port. */
2099 if (conn->scheme == SCHEME_HTTPS)
2101 if (!ssl_connect_wget (sock, u->host))
2107 else if (!ssl_check_certificate (sock, u->host))
2111 return VERIFCERTERR;
2115 #endif /* HAVE_SSL */
2118 /* Open the temporary file where we will write the request. */
2121 warc_tmp = warc_tempfile ();
2122 if (warc_tmp == NULL)
2124 CLOSE_INVALIDATE (sock);
2126 return WARC_TMP_FOPENERR;
2131 warc_ip = (ip_address *) alloca (sizeof (ip_address));
2132 socket_ip_address (sock, warc_ip, ENDPOINT_PEER);
2136 /* Send the request to server. */
2137 write_error = request_send (req, sock, warc_tmp);
2139 if (write_error >= 0)
2143 DEBUGP (("[BODY data: %s]\n", opt.body_data));
2144 write_error = fd_write (sock, opt.body_data, body_data_size, -1);
2145 if (write_error >= 0 && warc_tmp != NULL)
2147 /* Remember end of headers / start of payload. */
2148 warc_payload_offset = ftello (warc_tmp);
2150 /* Write a copy of the data to the WARC record. */
2151 int warc_tmp_written = fwrite (opt.body_data, 1, body_data_size, warc_tmp);
2152 if (warc_tmp_written != body_data_size)
2156 else if (opt.body_file && body_data_size != 0)
2158 if (warc_tmp != NULL)
2159 /* Remember end of headers / start of payload */
2160 warc_payload_offset = ftello (warc_tmp);
2162 write_error = body_file_send (sock, opt.body_file, body_data_size, warc_tmp);
2166 if (write_error < 0)
2168 CLOSE_INVALIDATE (sock);
2171 if (warc_tmp != NULL)
2174 if (write_error == -2)
2175 return WARC_TMP_FWRITEERR;
2179 logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
2180 proxy ? "Proxy" : "HTTP");
2189 /* Generate a timestamp and uuid for this request. */
2190 warc_timestamp (warc_timestamp_str);
2191 warc_uuid_str (warc_request_uuid);
2193 /* Create a request record and store it in the WARC file. */
2194 warc_result = warc_write_request_record (u->url, warc_timestamp_str,
2195 warc_request_uuid, warc_ip,
2196 warc_tmp, warc_payload_offset);
2199 CLOSE_INVALIDATE (sock);
2204 /* warc_write_request_record has also closed warc_tmp. */
2209 head = read_http_response_head (sock);
2214 logputs (LOG_NOTQUIET, _("No data received.\n"));
2215 CLOSE_INVALIDATE (sock);
2221 logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
2223 CLOSE_INVALIDATE (sock);
2228 DEBUGP (("\n---response begin---\n%s---response end---\n", head));
2230 resp = resp_new (head);
2232 /* Check for status line. */
2234 statcode = resp_status (resp, &message);
2237 char *tms = datetime_str (time (NULL));
2238 logprintf (LOG_VERBOSE, "%d\n", statcode);
2239 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"), tms, statcode,
2240 quotearg_style (escape_quoting_style,
2241 _("Malformed status line")));
2242 CLOSE_INVALIDATE (sock);
2249 if (H_10X (statcode))
2251 DEBUGP (("Ignoring response\n"));
2257 hs->message = xstrdup (message);
2258 if (!opt.server_response)
2259 logprintf (LOG_VERBOSE, "%2d %s\n", statcode,
2260 message ? quotearg_style (escape_quoting_style, message) : "");
2263 logprintf (LOG_VERBOSE, "\n");
2264 print_server_response (resp, " ");
2267 if (!opt.ignore_length
2268 && resp_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
2272 parsed = str_to_wgint (hdrval, NULL, 10);
2273 if (parsed == WGINT_MAX && errno == ERANGE)
2276 #### If Content-Length is out of range, it most likely
2277 means that the file is larger than 2G and that we're
2278 compiled without LFS. In that case we should probably
2279 refuse to even attempt to download the file. */
2282 else if (parsed < 0)
2284 /* Negative Content-Length; nonsensical, so we can't
2285 assume any information about the content to receive. */
2292 /* Check for keep-alive related responses. */
2293 if (!inhibit_keep_alive && contlen != -1)
2295 if (resp_header_copy (resp, "Connection", hdrval, sizeof (hdrval)))
2297 if (0 == strcasecmp (hdrval, "Close"))
2302 chunked_transfer_encoding = false;
2303 if (resp_header_copy (resp, "Transfer-Encoding", hdrval, sizeof (hdrval))
2304 && 0 == strcasecmp (hdrval, "chunked"))
2305 chunked_transfer_encoding = true;
2307 /* Handle (possibly multiple instances of) the Set-Cookie header. */
2311 const char *scbeg, *scend;
2312 /* The jar should have been created by now. */
2313 assert (wget_cookie_jar != NULL);
2315 (scpos = resp_header_locate (resp, "Set-Cookie", scpos,
2316 &scbeg, &scend)) != -1;
2319 char *set_cookie; BOUNDED_TO_ALLOCA (scbeg, scend, set_cookie);
2320 cookie_handle_set_cookie (wget_cookie_jar, u->host, u->port,
2321 u->path, set_cookie);
2326 /* The server has promised that it will not close the connection
2327 when we're done. This means that we can register it. */
2328 register_persistent (conn->host, conn->port, sock, using_ssl);
2330 if (statcode == HTTP_STATUS_UNAUTHORIZED)
2332 /* Authorization is required. */
2334 /* Normally we are not interested in the response body.
2335 But if we are writing a WARC file we are: we like to keep everyting. */
2339 type = resp_header_strdup (resp, "Content-Type");
2340 _err = read_response_body (hs, sock, NULL, contlen, 0,
2341 chunked_transfer_encoding,
2342 u->url, warc_timestamp_str,
2343 warc_request_uuid, warc_ip, type,
2347 if (_err != RETRFINISHED || hs->res < 0)
2349 CLOSE_INVALIDATE (sock);
2351 xfree_null (message);
2357 CLOSE_FINISH (sock);
2361 /* Since WARC is disabled, we are not interested in the response body. */
2362 if (keep_alive && !head_only
2363 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2364 CLOSE_FINISH (sock);
2366 CLOSE_INVALIDATE (sock);
2369 pconn.authorized = false;
2370 uerr_t auth_err = RETROK;
2371 if (!auth_finished && (user && passwd))
2373 /* IIS sends multiple copies of WWW-Authenticate, one with
2374 the value "negotiate", and other(s) with data. Loop over
2375 all the occurrences and pick the one we recognize. */
2377 const char *wabeg, *waend;
2378 char *www_authenticate = NULL;
2380 (wapos = resp_header_locate (resp, "WWW-Authenticate", wapos,
2381 &wabeg, &waend)) != -1;
2383 if (known_authentication_scheme_p (wabeg, waend))
2385 BOUNDED_TO_ALLOCA (wabeg, waend, www_authenticate);
2389 if (!www_authenticate)
2391 /* If the authentication header is missing or
2392 unrecognized, there's no sense in retrying. */
2393 logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
2395 else if (!basic_auth_finished
2396 || !BEGINS_WITH (www_authenticate, "Basic"))
2398 char *pth = url_full_path (u);
2401 auth_stat = xmalloc (sizeof (uerr_t));
2402 *auth_stat = RETROK;
2404 value = create_authorization_line (www_authenticate,
2406 request_method (req),
2411 auth_err = *auth_stat;
2412 if (auth_err == RETROK)
2414 request_set_header (req, "Authorization", value, rel_value);
2416 if (BEGINS_WITH (www_authenticate, "NTLM"))
2418 else if (!u->user && BEGINS_WITH (www_authenticate, "Basic"))
2420 /* Need to register this host as using basic auth,
2421 * so we automatically send creds next time. */
2422 register_basic_auth_host (u->host);
2426 xfree_null (message);
2430 goto retry_with_auth;
2434 /* Creating the Authorization header went wrong */
2439 /* We already did Basic auth, and it failed. Gotta
2444 xfree_null (message);
2447 if (auth_err == RETROK)
2452 else /* statcode != HTTP_STATUS_UNAUTHORIZED */
2454 /* Kludge: if NTLM is used, mark the TCP connection as authorized. */
2456 pconn.authorized = true;
2459 /* Determine the local filename if needed. Notice that if -O is used
2460 * hstat.local_file is set by http_loop to the argument of -O. */
2461 if (!hs->local_file)
2463 char *local_file = NULL;
2465 /* Honor Content-Disposition whether possible. */
2466 if (!opt.content_disposition
2467 || !resp_header_copy (resp, "Content-Disposition",
2468 hdrval, sizeof (hdrval))
2469 || !parse_content_disposition (hdrval, &local_file))
2471 /* The Content-Disposition header is missing or broken.
2472 * Choose unique file name according to given URL. */
2473 hs->local_file = url_file_name (u, NULL);
2477 DEBUGP (("Parsed filename from Content-Disposition: %s\n",
2479 hs->local_file = url_file_name (u, local_file);
2483 /* TODO: perform this check only once. */
2484 if (!hs->existence_checked && file_exists_p (hs->local_file))
2486 if (opt.noclobber && !opt.output_document)
2488 /* If opt.noclobber is turned on and file already exists, do not
2489 retrieve the file. But if the output_document was given, then this
2490 test was already done and the file didn't exist. Hence the !opt.output_document */
2491 get_file_flags (hs->local_file, dt);
2495 xfree_null (message);
2496 return RETRUNNEEDED;
2498 else if (!ALLOW_CLOBBER)
2500 char *unique = unique_name (hs->local_file, true);
2501 if (unique != hs->local_file)
2502 xfree (hs->local_file);
2503 hs->local_file = unique;
2506 hs->existence_checked = true;
2508 /* Support timestamping */
2509 /* TODO: move this code out of gethttp. */
2510 if (opt.timestamping && !hs->timestamp_checked)
2512 size_t filename_len = strlen (hs->local_file);
2513 char *filename_plus_orig_suffix = alloca (filename_len + sizeof (ORIG_SFX));
2514 bool local_dot_orig_file_exists = false;
2515 char *local_filename = NULL;
2518 if (opt.backup_converted)
2519 /* If -K is specified, we'll act on the assumption that it was specified
2520 last time these files were downloaded as well, and instead of just
2521 comparing local file X against server file X, we'll compare local
2522 file X.orig (if extant, else X) against server file X. If -K
2523 _wasn't_ specified last time, or the server contains files called
2524 *.orig, -N will be back to not operating correctly with -k. */
2526 /* Would a single s[n]printf() call be faster? --dan
2528 Definitely not. sprintf() is horribly slow. It's a
2529 different question whether the difference between the two
2530 affects a program. Usually I'd say "no", but at one
2531 point I profiled Wget, and found that a measurable and
2532 non-negligible amount of time was lost calling sprintf()
2533 in url.c. Replacing sprintf with inline calls to
2534 strcpy() and number_to_string() made a difference.
2536 memcpy (filename_plus_orig_suffix, hs->local_file, filename_len);
2537 memcpy (filename_plus_orig_suffix + filename_len,
2538 ORIG_SFX, sizeof (ORIG_SFX));
2540 /* Try to stat() the .orig file. */
2541 if (stat (filename_plus_orig_suffix, &st) == 0)
2543 local_dot_orig_file_exists = true;
2544 local_filename = filename_plus_orig_suffix;
2548 if (!local_dot_orig_file_exists)
2549 /* Couldn't stat() <file>.orig, so try to stat() <file>. */
2550 if (stat (hs->local_file, &st) == 0)
2551 local_filename = hs->local_file;
2553 if (local_filename != NULL)
2554 /* There was a local file, so we'll check later to see if the version
2555 the server has is the same version we already have, allowing us to
2558 hs->orig_file_name = xstrdup (local_filename);
2559 hs->orig_file_size = st.st_size;
2560 hs->orig_file_tstamp = st.st_mtime;
2562 /* Modification time granularity is 2 seconds for Windows, so
2563 increase local time by 1 second for later comparison. */
2564 ++hs->orig_file_tstamp;
2571 hs->statcode = statcode;
2573 hs->error = xstrdup (_("Malformed status line"));
2575 hs->error = xstrdup (_("(no description)"));
2577 hs->error = xstrdup (message);
2578 xfree_null (message);
2580 type = resp_header_strdup (resp, "Content-Type");
2583 char *tmp = strchr (type, ';');
2586 /* sXXXav: only needed if IRI support is enabled */
2587 char *tmp2 = tmp + 1;
2589 while (tmp > type && c_isspace (tmp[-1]))
2593 /* Try to get remote encoding if needed */
2594 if (opt.enable_iri && !opt.encoding_remote)
2596 tmp = parse_charset (tmp2);
2598 set_content_encoding (iri, tmp);
2603 hs->newloc = resp_header_strdup (resp, "Location");
2604 hs->remote_time = resp_header_strdup (resp, "Last-Modified");
2606 if (resp_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
2608 wgint first_byte_pos, last_byte_pos, entity_length;
2609 if (parse_content_range (hdrval, &first_byte_pos, &last_byte_pos,
2612 contrange = first_byte_pos;
2613 contlen = last_byte_pos - first_byte_pos + 1;
2618 /* 20x responses are counted among successful by default. */
2619 if (H_20X (statcode))
2622 if (statcode == HTTP_STATUS_NO_CONTENT)
2624 /* 204 response has no body (RFC 2616, 4.3) */
2626 /* In case the caller cares to look... */
2631 CLOSE_FINISH (sock);
2635 return RETRFINISHED;
2638 /* Return if redirected. */
2639 if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES)
2641 /* RFC2068 says that in case of the 300 (multiple choices)
2642 response, the server can output a preferred URL through
2643 `Location' header; otherwise, the request should be treated
2644 like GET. So, if the location is set, it will be a
2645 redirection; otherwise, just proceed normally. */
2646 if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc)
2650 logprintf (LOG_VERBOSE,
2651 _("Location: %s%s\n"),
2652 hs->newloc ? escnonprint_uri (hs->newloc) : _("unspecified"),
2653 hs->newloc ? _(" [following]") : "");
2655 /* In case the caller cares to look... */
2660 /* Normally we are not interested in the response body of a redirect.
2661 But if we are writing a WARC file we are: we like to keep everyting. */
2664 int _err = read_response_body (hs, sock, NULL, contlen, 0,
2665 chunked_transfer_encoding,
2666 u->url, warc_timestamp_str,
2667 warc_request_uuid, warc_ip, type,
2670 if (_err != RETRFINISHED || hs->res < 0)
2672 CLOSE_INVALIDATE (sock);
2678 CLOSE_FINISH (sock);
2682 /* Since WARC is disabled, we are not interested in the response body. */
2683 if (keep_alive && !head_only
2684 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2685 CLOSE_FINISH (sock);
2687 CLOSE_INVALIDATE (sock);
2692 /* From RFC2616: The status codes 303 and 307 have
2693 been added for servers that wish to make unambiguously
2694 clear which kind of reaction is expected of the client.
2696 A 307 should be redirected using the same method,
2697 in other words, a POST should be preserved and not
2698 converted to a GET in that case.
2700 With strict adherence to RFC2616, POST requests are not
2701 converted to a GET request on 301 Permanent Redirect
2702 or 302 Temporary Redirect.
2704 A switch may be provided later based on the HTTPbis draft
2705 that allows clients to convert POST requests to GET
2706 requests on 301 and 302 response codes. */
2709 case HTTP_STATUS_TEMPORARY_REDIRECT:
2710 return NEWLOCATION_KEEP_POST;
2711 case HTTP_STATUS_MOVED_PERMANENTLY:
2712 if (opt.method && strcasecmp (opt.method, "post") != 0)
2713 return NEWLOCATION_KEEP_POST;
2715 case HTTP_STATUS_MOVED_TEMPORARILY:
2716 if (opt.method && strcasecmp (opt.method, "post") != 0)
2717 return NEWLOCATION_KEEP_POST;
2726 /* If content-type is not given, assume text/html. This is because
2727 of the multitude of broken CGI's that "forget" to generate the
2730 0 == strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S)) ||
2731 0 == strncasecmp (type, TEXTXHTML_S, strlen (TEXTXHTML_S)))
2737 0 == strncasecmp (type, TEXTCSS_S, strlen (TEXTCSS_S)))
2742 if (opt.adjust_extension)
2745 /* -E / --adjust-extension / adjust_extension = on was specified,
2746 and this is a text/html file. If some case-insensitive
2747 variation on ".htm[l]" isn't already the file's suffix,
2750 ensure_extension (hs, ".html", dt);
2752 else if (*dt & TEXTCSS)
2754 ensure_extension (hs, ".css", dt);
2758 if (statcode == HTTP_STATUS_RANGE_NOT_SATISFIABLE
2759 || (!opt.timestamping && hs->restval > 0 && statcode == HTTP_STATUS_OK
2760 && contrange == 0 && contlen >= 0 && hs->restval >= contlen))
2762 /* If `-c' is in use and the file has been fully downloaded (or
2763 the remote file has shrunk), Wget effectively requests bytes
2764 after the end of file and the server response with 416
2765 (or 200 with a <= Content-Length. */
2766 logputs (LOG_VERBOSE, _("\
2767 \n The file is already fully retrieved; nothing to do.\n\n"));
2768 /* In case the caller inspects. */
2771 /* Mark as successfully retrieved. */
2774 CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
2775 might be more bytes in the body. */
2777 return RETRUNNEEDED;
2779 if ((contrange != 0 && contrange != hs->restval)
2780 || (H_PARTIAL (statcode) && !contrange))
2782 /* The Range request was somehow misunderstood by the server.
2785 CLOSE_INVALIDATE (sock);
2792 hs->contlen = contlen + contrange;
2798 /* No need to print this output if the body won't be
2799 downloaded at all, or if the original server response is
2801 logputs (LOG_VERBOSE, _("Length: "));
2804 logputs (LOG_VERBOSE, number_to_static_string (contlen + contrange));
2805 if (contlen + contrange >= 1024)
2806 logprintf (LOG_VERBOSE, " (%s)",
2807 human_readable (contlen + contrange, 10, 1));
2810 if (contlen >= 1024)
2811 logprintf (LOG_VERBOSE, _(", %s (%s) remaining"),
2812 number_to_static_string (contlen),
2813 human_readable (contlen, 10, 1));
2815 logprintf (LOG_VERBOSE, _(", %s remaining"),
2816 number_to_static_string (contlen));
2820 logputs (LOG_VERBOSE,
2821 opt.ignore_length ? _("ignored") : _("unspecified"));
2823 logprintf (LOG_VERBOSE, " [%s]\n", quotearg_style (escape_quoting_style, type));
2825 logputs (LOG_VERBOSE, "\n");
2829 /* Return if we have no intention of further downloading. */
2830 if ((!(*dt & RETROKF) && !opt.content_on_error) || head_only)
2832 /* In case the caller cares to look... */
2837 /* Normally we are not interested in the response body of a error responses.
2838 But if we are writing a WARC file we are: we like to keep everyting. */
2841 int _err = read_response_body (hs, sock, NULL, contlen, 0,
2842 chunked_transfer_encoding,
2843 u->url, warc_timestamp_str,
2844 warc_request_uuid, warc_ip, type,
2847 if (_err != RETRFINISHED || hs->res < 0)
2849 CLOSE_INVALIDATE (sock);
2855 CLOSE_FINISH (sock);
2859 /* Since WARC is disabled, we are not interested in the response body. */
2861 /* Pre-1.10 Wget used CLOSE_INVALIDATE here. Now we trust the
2862 servers not to send body in response to a HEAD request, and
2863 those that do will likely be caught by test_socket_open.
2864 If not, they can be worked around using
2865 `--no-http-keep-alive'. */
2866 CLOSE_FINISH (sock);
2868 && skip_short_body (sock, contlen, chunked_transfer_encoding))
2869 /* Successfully skipped the body; also keep using the socket. */
2870 CLOSE_FINISH (sock);
2872 CLOSE_INVALIDATE (sock);
2877 return RETRFINISHED;
2881 For VMS, define common fopen() optional arguments.
2884 # define FOPEN_OPT_ARGS "fop=sqo", "acc", acc_cb, &open_id
2885 # define FOPEN_BIN_FLAG 3
2886 #else /* def __VMS */
2887 # define FOPEN_BIN_FLAG true
2888 #endif /* def __VMS [else] */
2890 /* Open the local file. */
2893 mkalldirs (hs->local_file);
2895 rotate_backups (hs->local_file);
2902 fp = fopen (hs->local_file, "ab", FOPEN_OPT_ARGS);
2903 #else /* def __VMS */
2904 fp = fopen (hs->local_file, "ab");
2905 #endif /* def __VMS [else] */
2907 else if (ALLOW_CLOBBER || count > 0)
2909 if (opt.unlink && file_exists_p (hs->local_file))
2911 int res = unlink (hs->local_file);
2914 logprintf (LOG_NOTQUIET, "%s: %s\n", hs->local_file,
2916 CLOSE_INVALIDATE (sock);
2927 fp = fopen (hs->local_file, "wb", FOPEN_OPT_ARGS);
2928 #else /* def __VMS */
2929 fp = fopen (hs->local_file, "wb");
2930 #endif /* def __VMS [else] */
2934 fp = fopen_excl (hs->local_file, FOPEN_BIN_FLAG);
2935 if (!fp && errno == EEXIST)
2937 /* We cannot just invent a new name and use it (which is
2938 what functions like unique_create typically do)
2939 because we told the user we'd use this name.
2940 Instead, return and retry the download. */
2941 logprintf (LOG_NOTQUIET,
2942 _("%s has sprung into existence.\n"),
2944 CLOSE_INVALIDATE (sock);
2947 return FOPEN_EXCL_ERR;
2952 logprintf (LOG_NOTQUIET, "%s: %s\n", hs->local_file, strerror (errno));
2953 CLOSE_INVALIDATE (sock);
2962 /* Print fetch message, if opt.verbose. */
2963 logprintf (LOG_VERBOSE, _("Saving to: %s\n"),
2964 HYPHENP (hs->local_file) ? quote ("STDOUT") : quote (hs->local_file));
2967 err = read_response_body (hs, sock, fp, contlen, contrange,
2968 chunked_transfer_encoding,
2969 u->url, warc_timestamp_str,
2970 warc_request_uuid, warc_ip, type,
2973 /* Now we no longer need to store the response header. */
2978 CLOSE_FINISH (sock);
2980 CLOSE_INVALIDATE (sock);
2988 /* The genuine HTTP loop! This is the part where the retrieval is
2989 retried, and retried, and retried, and... */
2991 http_loop (struct url *u, struct url *original_url, char **newloc,
2992 char **local_file, const char *referer, int *dt, struct url *proxy,
2996 bool got_head = false; /* used for time-stamping and filename detection */
2997 bool time_came_from_head = false;
2998 bool got_name = false;
3001 uerr_t err, ret = TRYLIMEXC;
3002 time_t tmr = -1; /* remote time-stamp */
3003 struct http_stat hstat; /* HTTP status */
3005 bool send_head_first = true;
3007 bool force_full_retrieve = false;
3010 /* If we are writing to a WARC file: always retrieve the whole file. */
3011 if (opt.warc_filename != NULL)
3012 force_full_retrieve = true;
3015 /* Assert that no value for *LOCAL_FILE was passed. */
3016 assert (local_file == NULL || *local_file == NULL);
3018 /* Set LOCAL_FILE parameter. */
3019 if (local_file && opt.output_document)
3020 *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
3022 /* Reset NEWLOC parameter. */
3025 /* This used to be done in main(), but it's a better idea to do it
3026 here so that we don't go through the hoops if we're just using
3031 /* Warn on (likely bogus) wildcard usage in HTTP. */
3032 if (opt.ftp_glob && has_wildcards_p (u->path))
3033 logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
3035 /* Setup hstat struct. */
3037 hstat.referer = referer;
3039 if (opt.output_document)
3041 hstat.local_file = xstrdup (opt.output_document);
3044 else if (!opt.content_disposition)
3047 url_file_name (opt.trustservernames ? u : original_url, NULL);
3051 if (got_name && file_exists_p (hstat.local_file) && opt.noclobber && !opt.output_document)
3053 /* If opt.noclobber is turned on and file already exists, do not
3054 retrieve the file. But if the output_document was given, then this
3055 test was already done and the file didn't exist. Hence the !opt.output_document */
3056 get_file_flags (hstat.local_file, dt);
3061 /* Reset the counter. */
3064 /* Reset the document type. */
3067 /* Skip preliminary HEAD request if we're not in spider mode. */
3069 send_head_first = false;
3071 /* Send preliminary HEAD request if --content-disposition and -c are used
3073 if (opt.content_disposition && opt.always_rest)
3074 send_head_first = true;
3076 /* Send preliminary HEAD request if -N is given and we have an existing
3077 * destination file. */
3078 if (!opt.output_document)
3079 file_name = url_file_name (opt.trustservernames ? u : original_url, NULL);
3081 file_name = xstrdup (opt.output_document);
3082 if (opt.timestamping && (file_exists_p (file_name)
3083 || opt.content_disposition))
3084 send_head_first = true;
3090 /* Increment the pass counter. */
3092 sleep_between_retrievals (count);
3094 /* Get the current time string. */
3095 tms = datetime_str (time (NULL));
3097 if (opt.spider && !got_head)
3098 logprintf (LOG_VERBOSE, _("\
3099 Spider mode enabled. Check if remote file exists.\n"));
3101 /* Print fetch message, if opt.verbose. */
3104 char *hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3109 sprintf (tmp, _("(try:%2d)"), count);
3110 logprintf (LOG_NOTQUIET, "--%s-- %s %s\n",
3115 logprintf (LOG_NOTQUIET, "--%s-- %s\n",
3120 ws_changetitle (hurl);
3125 /* Default document type is empty. However, if spider mode is
3126 on or time-stamping is employed, HEAD_ONLY commands is
3127 encoded within *dt. */
3128 if (send_head_first && !got_head)
3133 /* Decide whether or not to restart. */
3134 if (force_full_retrieve)
3135 hstat.restval = hstat.len;
3136 else if (opt.start_pos >= 0)
3137 hstat.restval = opt.start_pos;
3138 else if (opt.always_rest
3140 && stat (hstat.local_file, &st) == 0
3141 && S_ISREG (st.st_mode))
3142 /* When -c is used, continue from on-disk size. (Can't use
3143 hstat.len even if count>1 because we don't want a failed
3144 first attempt to clobber existing data.) */
3145 hstat.restval = st.st_size;
3147 /* otherwise, continue where the previous try left off */
3148 hstat.restval = hstat.len;
3152 /* Decide whether to send the no-cache directive. We send it in
3154 a) we're using a proxy, and we're past our first retrieval.
3155 Some proxies are notorious for caching incomplete data, so
3156 we require a fresh get.
3157 b) caching is explicitly inhibited. */
3158 if ((proxy && count > 1) /* a */
3159 || !opt.allow_cache) /* b */
3160 *dt |= SEND_NOCACHE;
3162 *dt &= ~SEND_NOCACHE;
3164 /* Try fetching the document, or at least its head. */
3165 err = gethttp (u, &hstat, dt, proxy, iri, count);
3168 tms = datetime_str (time (NULL));
3170 /* Get the new location (with or without the redirection). */
3172 *newloc = xstrdup (hstat.newloc);
3176 case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
3177 case CONERROR: case READERR: case WRITEFAILED:
3178 case RANGEERR: case FOPEN_EXCL_ERR:
3179 /* Non-fatal errors continue executing the loop, which will
3180 bring them to "while" statement at the end, to judge
3181 whether the number of tries was exceeded. */
3182 printwhat (count, opt.ntry);
3184 case FWRITEERR: case FOPENERR:
3185 /* Another fatal error. */
3186 logputs (LOG_VERBOSE, "\n");
3187 logprintf (LOG_NOTQUIET, _("Cannot write to %s (%s).\n"),
3188 quote (hstat.local_file), strerror (errno));
3189 case HOSTERR: case CONIMPOSSIBLE: case PROXERR: case SSLINITFAILED:
3190 case CONTNOTSUPPORTED: case VERIFCERTERR: case FILEBADFILE:
3192 /* Fatal errors just return from the function. */
3196 /* A missing attribute in a Header is a fatal Protocol error. */
3197 logputs (LOG_VERBOSE, "\n");
3198 logprintf (LOG_NOTQUIET, _("Required attribute missing from Header received.\n"));
3202 logputs (LOG_VERBOSE, "\n");
3203 logprintf (LOG_NOTQUIET, _("Username/Password Authentication Failed.\n"));
3207 /* A fatal WARC error. */
3208 logputs (LOG_VERBOSE, "\n");
3209 logprintf (LOG_NOTQUIET, _("Cannot write to WARC file.\n"));
3212 case WARC_TMP_FOPENERR: case WARC_TMP_FWRITEERR:
3213 /* A fatal WARC error. */
3214 logputs (LOG_VERBOSE, "\n");
3215 logprintf (LOG_NOTQUIET, _("Cannot write to temporary WARC file.\n"));
3219 /* Another fatal error. */
3220 logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
3224 /* Another fatal error. */
3225 logputs (LOG_VERBOSE, "\n");
3226 logprintf (LOG_NOTQUIET, _("Cannot unlink %s (%s).\n"),
3227 quote (hstat.local_file), strerror (errno));
3231 case NEWLOCATION_KEEP_POST:
3232 /* Return the new location to the caller. */
3235 logprintf (LOG_NOTQUIET,
3236 _("ERROR: Redirection (%d) without location.\n"),
3246 /* The file was already fully retrieved. */
3250 /* Deal with you later. */
3253 /* All possibilities should have been exhausted. */
3257 if (!(*dt & RETROKF))
3262 /* #### Ugly ugly ugly! */
3263 hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3264 logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
3267 /* Fall back to GET if HEAD fails with a 500 or 501 error code. */
3269 && (hstat.statcode == 500 || hstat.statcode == 501))
3274 /* Maybe we should always keep track of broken links, not just in
3276 * Don't log error if it was UTF-8 encoded because we will try
3277 * once unencoded. */
3278 else if (opt.spider && !iri->utf8_encode)
3280 /* #### Again: ugly ugly ugly! */
3282 hurl = url_string (u, URL_AUTH_HIDE_PASSWD);
3283 nonexisting_url (hurl);
3284 logprintf (LOG_NOTQUIET, _("\
3285 Remote file does not exist -- broken link!!!\n"));
3289 logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
3290 tms, hstat.statcode,
3291 quotearg_style (escape_quoting_style, hstat.error));
3293 logputs (LOG_VERBOSE, "\n");
3299 /* Did we get the time-stamp? */
3302 got_head = true; /* no more time-stamping */
3304 if (opt.timestamping && !hstat.remote_time)
3306 logputs (LOG_NOTQUIET, _("\
3307 Last-modified header missing -- time-stamps turned off.\n"));
3309 else if (hstat.remote_time)
3311 /* Convert the date-string into struct tm. */
3312 tmr = http_atotm (hstat.remote_time);
3313 if (tmr == (time_t) (-1))
3314 logputs (LOG_VERBOSE, _("\
3315 Last-modified header invalid -- time-stamp ignored.\n"));
3316 if (*dt & HEAD_ONLY)
3317 time_came_from_head = true;
3320 if (send_head_first)
3322 /* The time-stamping section. */
3323 if (opt.timestamping)
3325 if (hstat.orig_file_name) /* Perform the following
3326 checks only if the file
3328 download already exists. */
3330 if (hstat.remote_time &&
3331 tmr != (time_t) (-1))
3333 /* Now time-stamping can be used validly.
3334 Time-stamping means that if the sizes of
3335 the local and remote file match, and local
3336 file is newer than the remote file, it will
3337 not be retrieved. Otherwise, the normal
3338 download procedure is resumed. */
3339 if (hstat.orig_file_tstamp >= tmr)
3341 if (hstat.contlen == -1
3342 || hstat.orig_file_size == hstat.contlen)
3344 logprintf (LOG_VERBOSE, _("\
3345 Server file no newer than local file %s -- not retrieving.\n\n"),
3346 quote (hstat.orig_file_name));
3352 logprintf (LOG_VERBOSE, _("\
3353 The sizes do not match (local %s) -- retrieving.\n"),
3354 number_to_static_string (hstat.orig_file_size));
3359 force_full_retrieve = true;
3360 logputs (LOG_VERBOSE,
3361 _("Remote file is newer, retrieving.\n"));
3364 logputs (LOG_VERBOSE, "\n");
3368 /* free_hstat (&hstat); */
3369 hstat.timestamp_checked = true;
3374 bool finished = true;
3379 logputs (LOG_VERBOSE, _("\
3380 Remote file exists and could contain links to other resources -- retrieving.\n\n"));
3385 logprintf (LOG_VERBOSE, _("\
3386 Remote file exists but does not contain any link -- not retrieving.\n\n"));
3387 ret = RETROK; /* RETRUNNEEDED is not for caller. */
3394 logprintf (LOG_VERBOSE, _("\
3395 Remote file exists and could contain further links,\n\
3396 but recursion is disabled -- not retrieving.\n\n"));
3400 logprintf (LOG_VERBOSE, _("\
3401 Remote file exists.\n\n"));
3403 ret = RETROK; /* RETRUNNEEDED is not for caller. */
3408 logprintf (LOG_NONVERBOSE,
3409 _("%s URL: %s %2d %s\n"),
3410 tms, u->url, hstat.statcode,
3411 hstat.message ? quotearg_style (escape_quoting_style, hstat.message) : "");
3418 count = 0; /* the retrieve count for HEAD is reset */
3420 } /* send_head_first */
3423 if (opt.useservertimestamps
3424 && (tmr != (time_t) (-1))
3425 && ((hstat.len == hstat.contlen) ||
3426 ((hstat.res == 0) && (hstat.contlen == -1))))
3428 const char *fl = NULL;
3429 set_local_file (&fl, hstat.local_file);
3433 /* Reparse time header, in case it's changed. */
3434 if (time_came_from_head
3435 && hstat.remote_time && hstat.remote_time[0])
3437 newtmr = http_atotm (hstat.remote_time);
3438 if (newtmr != (time_t)-1)
3444 /* End of time-stamping section. */
3446 tmrate = retr_rate (hstat.rd_size, hstat.dltime);
3447 total_download_time += hstat.dltime;
3449 if (hstat.len == hstat.contlen)
3453 bool write_to_stdout = (opt.output_document && HYPHENP (opt.output_document));
3455 logprintf (LOG_VERBOSE,
3457 ? _("%s (%s) - written to stdout %s[%s/%s]\n\n")
3458 : _("%s (%s) - %s saved [%s/%s]\n\n"),
3460 write_to_stdout ? "" : quote (hstat.local_file),
3461 number_to_static_string (hstat.len),
3462 number_to_static_string (hstat.contlen));
3463 logprintf (LOG_NONVERBOSE,
3464 "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
3466 number_to_static_string (hstat.len),
3467 number_to_static_string (hstat.contlen),
3468 hstat.local_file, count);
3471 total_downloaded_bytes += hstat.rd_size;
3473 /* Remember that we downloaded the file for later ".orig" code. */
3474 if (*dt & ADDED_HTML_EXTENSION)
3475 downloaded_file (FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, hstat.local_file);
3477 downloaded_file (FILE_DOWNLOADED_NORMALLY, hstat.local_file);
3482 else if (hstat.res == 0) /* No read error */
3484 if (hstat.contlen == -1) /* We don't know how much we were supposed
3485 to get, so assume we succeeded. */
3489 bool write_to_stdout = (opt.output_document && HYPHENP (opt.output_document));
3491 logprintf (LOG_VERBOSE,
3493 ? _("%s (%s) - written to stdout %s[%s]\n\n")
3494 : _("%s (%s) - %s saved [%s]\n\n"),
3496 write_to_stdout ? "" : quote (hstat.local_file),
3497 number_to_static_string (hstat.len));
3498 logprintf (LOG_NONVERBOSE,
3499 "%s URL:%s [%s] -> \"%s\" [%d]\n",
3500 tms, u->url, number_to_static_string (hstat.len),
3501 hstat.local_file, count);
3504 total_downloaded_bytes += hstat.rd_size;
3506 /* Remember that we downloaded the file for later ".orig" code. */
3507 if (*dt & ADDED_HTML_EXTENSION)
3508 downloaded_file (FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED, hstat.local_file);
3510 downloaded_file (FILE_DOWNLOADED_NORMALLY, hstat.local_file);
3515 else if (hstat.len < hstat.contlen) /* meaning we lost the
3516 connection too soon */
3518 logprintf (LOG_VERBOSE,
3519 _("%s (%s) - Connection closed at byte %s. "),
3520 tms, tmrate, number_to_static_string (hstat.len));
3521 printwhat (count, opt.ntry);
3524 else if (hstat.len != hstat.restval)
3525 /* Getting here would mean reading more data than
3526 requested with content-length, which we never do. */
3530 /* Getting here probably means that the content-length was
3531 * _less_ than the original, local size. We should probably
3532 * truncate or re-read, or something. FIXME */
3537 else /* from now on hstat.res can only be -1 */
3539 if (hstat.contlen == -1)
3541 logprintf (LOG_VERBOSE,
3542 _("%s (%s) - Read error at byte %s (%s)."),
3543 tms, tmrate, number_to_static_string (hstat.len),
3545 printwhat (count, opt.ntry);
3548 else /* hstat.res == -1 and contlen is given */
3550 logprintf (LOG_VERBOSE,
3551 _("%s (%s) - Read error at byte %s/%s (%s). "),
3553 number_to_static_string (hstat.len),
3554 number_to_static_string (hstat.contlen),
3556 printwhat (count, opt.ntry);
3562 while (!opt.ntry || (count < opt.ntry));
3565 if (ret == RETROK && local_file)
3566 *local_file = xstrdup (hstat.local_file);
3567 free_hstat (&hstat);
3572 /* Check whether the result of strptime() indicates success.
3573 strptime() returns the pointer to how far it got to in the string.
3574 The processing has been successful if the string is at `GMT' or
3575 `+X', or at the end of the string.
3577 In extended regexp parlance, the function returns 1 if P matches
3578 "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (which strptime
3579 can return) is considered a failure and 0 is returned. */
3581 check_end (const char *p)
3585 while (c_isspace (*p))
3588 || (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
3589 || ((p[0] == '+' || p[0] == '-') && c_isdigit (p[1])))
3595 /* Convert the textual specification of time in TIME_STRING to the
3596 number of seconds since the Epoch.
3598 TIME_STRING can be in any of the three formats RFC2616 allows the
3599 HTTP servers to emit -- RFC1123-date, RFC850-date or asctime-date,
3600 as well as the time format used in the Set-Cookie header.
3601 Timezones are ignored, and should be GMT.
3603 Return the computed time_t representation, or -1 if the conversion
3606 This function uses strptime with various string formats for parsing
3607 TIME_STRING. This results in a parser that is not as lenient in
3608 interpreting TIME_STRING as I would like it to be. Being based on
3609 strptime, it always allows shortened months, one-digit days, etc.,
3610 but due to the multitude of formats in which time can be
3611 represented, an ideal HTTP time parser would be even more
3612 forgiving. It should completely ignore things like week days and
3613 concentrate only on the various forms of representing years,
3614 months, days, hours, minutes, and seconds. For example, it would
3615 be nice if it accepted ISO 8601 out of the box.
3617 I've investigated free and PD code for this purpose, but none was
3618 usable. getdate was big and unwieldy, and had potential copyright
3619 issues, or so I was informed. Dr. Marcus Hennecke's atotm(),
3620 distributed with phttpd, is excellent, but we cannot use it because
3621 it is not assigned to the FSF. So I stuck it with strptime. */
3624 http_atotm (const char *time_string)
3626 /* NOTE: Solaris strptime man page claims that %n and %t match white
3627 space, but that's not universally available. Instead, we simply
3628 use ` ' to mean "skip all WS", which works under all strptime
3629 implementations I've tested. */
3631 static const char *time_formats[] = {
3632 "%a, %d %b %Y %T", /* rfc1123: Thu, 29 Jan 1998 22:12:57 */
3633 "%A, %d-%b-%y %T", /* rfc850: Thursday, 29-Jan-98 22:12:57 */
3634 "%a %b %d %T %Y", /* asctime: Thu Jan 29 22:12:57 1998 */
3635 "%a, %d-%b-%Y %T" /* cookies: Thu, 29-Jan-1998 22:12:57
3636 (used in Set-Cookie, defined in the
3637 Netscape cookie specification.) */
3639 const char *oldlocale;
3640 char savedlocale[256];
3642 time_t ret = (time_t) -1;
3644 /* Solaris strptime fails to recognize English month names in
3645 non-English locales, which we work around by temporarily setting
3646 locale to C before invoking strptime. */
3647 oldlocale = setlocale (LC_TIME, NULL);
3650 size_t l = strlen (oldlocale) + 1;
3651 if (l >= sizeof savedlocale)
3652 savedlocale[0] = '\0';
3654 memcpy (savedlocale, oldlocale, l);
3656 else savedlocale[0] = '\0';
3658 setlocale (LC_TIME, "C");
3660 for (i = 0; i < countof (time_formats); i++)
3664 /* Some versions of strptime use the existing contents of struct
3665 tm to recalculate the date according to format. Zero it out
3666 to prevent stack garbage from influencing strptime. */
3669 if (check_end (strptime (time_string, time_formats[i], &t)))
3676 /* Restore the previous locale. */
3678 setlocale (LC_TIME, savedlocale);
3683 /* Authorization support: We support three authorization schemes:
3685 * `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
3687 * `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
3688 consisting of answering to the server's challenge with the proper
3691 * `NTLM' ("NT Lan Manager") scheme, based on code written by Daniel
3692 Stenberg for libcurl. Like digest, NTLM is based on a
3693 challenge-response mechanism, but unlike digest, it is non-standard
3694 (authenticates TCP connections rather than requests), undocumented
3695 and Microsoft-specific. */
3697 /* Create the authentication header contents for the `Basic' scheme.
3698 This is done by encoding the string "USER:PASS" to base64 and
3699 prepending the string "Basic " in front of it. */
3702 basic_authentication_encode (const char *user, const char *passwd)
3705 int len1 = strlen (user) + 1 + strlen (passwd);
3707 t1 = (char *)alloca (len1 + 1);
3708 sprintf (t1, "%s:%s", user, passwd);
3710 t2 = (char *)alloca (BASE64_LENGTH (len1) + 1);
3711 base64_encode (t1, len1, t2);
3713 return concat_strings ("Basic ", t2, (char *) 0);
3716 #define SKIP_WS(x) do { \
3717 while (c_isspace (*(x))) \
3721 #ifdef ENABLE_DIGEST
3722 /* Dump the hexadecimal representation of HASH to BUF. HASH should be
3723 an array of 16 bytes containing the hash keys, and BUF should be a
3724 buffer of 33 writable characters (32 for hex digits plus one for
3725 zero termination). */
3727 dump_hash (char *buf, const unsigned char *hash)
3731 for (i = 0; i < MD5_DIGEST_SIZE; i++, hash++)
3733 *buf++ = XNUM_TO_digit (*hash >> 4);
3734 *buf++ = XNUM_TO_digit (*hash & 0xf);
3739 /* Take the line apart to find the challenge, and compose a digest
3740 authorization header. See RFC2069 section 2.1.2. */
3742 digest_authentication_encode (const char *au, const char *user,
3743 const char *passwd, const char *method,
3744 const char *path, uerr_t *auth_err)
3746 static char *realm, *opaque, *nonce, *qop, *algorithm;
3751 { "realm", &realm },
3752 { "opaque", &opaque },
3753 { "nonce", &nonce },
3755 { "algorithm", &algorithm }
3757 char cnonce[16] = "";
3761 param_token name, value;
3764 realm = opaque = nonce = algorithm = qop = NULL;
3766 au += 6; /* skip over `Digest' */
3767 while (extract_param (&au, &name, &value, ',', NULL))
3770 size_t namelen = name.e - name.b;
3771 for (i = 0; i < countof (options); i++)
3772 if (namelen == strlen (options[i].name)
3773 && 0 == strncmp (name.b, options[i].name,
3776 *options[i].variable = strdupdelim (value.b, value.e);
3781 if (qop != NULL && strcmp(qop,"auth"))
3783 logprintf (LOG_NOTQUIET, _("Unsupported quality of protection '%s'.\n"), qop);
3784 xfree_null (qop); /* force freeing mem and return */
3787 else if (algorithm != NULL && strcmp (algorithm,"MD5") && strcmp (algorithm,"MD5-sess"))
3789 logprintf (LOG_NOTQUIET, _("Unsupported algorithm '%s'.\n"), algorithm);
3790 xfree_null (qop); /* force freeing mem and return */
3794 if (!realm || !nonce || !user || !passwd || !path || !method || !qop)
3797 xfree_null (opaque);
3800 xfree_null (algorithm);
3802 *auth_err = UNKNOWNATTR;
3804 *auth_err = ATTRMISSING;
3808 /* Calculate the digest value. */
3811 unsigned char hash[MD5_DIGEST_SIZE];
3812 char a1buf[MD5_DIGEST_SIZE * 2 + 1], a2buf[MD5_DIGEST_SIZE * 2 + 1];
3813 char response_digest[MD5_DIGEST_SIZE * 2 + 1];
3815 /* A1BUF = H(user ":" realm ":" password) */
3816 md5_init_ctx (&ctx);
3817 md5_process_bytes ((unsigned char *)user, strlen (user), &ctx);
3818 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3819 md5_process_bytes ((unsigned char *)realm, strlen (realm), &ctx);
3820 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3821 md5_process_bytes ((unsigned char *)passwd, strlen (passwd), &ctx);
3822 md5_finish_ctx (&ctx, hash);
3824 dump_hash (a1buf, hash);
3826 if (algorithm && !strcmp (algorithm, "MD5-sess"))
3828 /* A1BUF = H( H(user ":" realm ":" password) ":" nonce ":" cnonce ) */
3829 snprintf (cnonce, sizeof (cnonce), "%08x", random_number(INT_MAX));
3831 md5_init_ctx (&ctx);
3832 // md5_process_bytes (hash, MD5_DIGEST_SIZE, &ctx);
3833 md5_process_bytes (a1buf, MD5_DIGEST_SIZE * 2, &ctx);
3834 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3835 md5_process_bytes ((unsigned char *)nonce, strlen (nonce), &ctx);
3836 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3837 md5_process_bytes ((unsigned char *)cnonce, strlen (cnonce), &ctx);
3838 md5_finish_ctx (&ctx, hash);
3840 dump_hash (a1buf, hash);
3843 /* A2BUF = H(method ":" path) */
3844 md5_init_ctx (&ctx);
3845 md5_process_bytes ((unsigned char *)method, strlen (method), &ctx);
3846 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3847 md5_process_bytes ((unsigned char *)path, strlen (path), &ctx);
3848 md5_finish_ctx (&ctx, hash);
3849 dump_hash (a2buf, hash);
3851 if (qop && (!strcmp(qop, "auth") || !strcmp (qop, "auth-int")))
3853 /* RFC 2617 Digest Access Authentication */
3854 /* generate random hex string */
3856 snprintf(cnonce, sizeof(cnonce), "%08x", random_number(INT_MAX));
3858 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" noncecount ":" clientnonce ":" qop ": " A2BUF) */
3859 md5_init_ctx (&ctx);
3860 md5_process_bytes ((unsigned char *)a1buf, MD5_DIGEST_SIZE * 2, &ctx);
3861 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3862 md5_process_bytes ((unsigned char *)nonce, strlen (nonce), &ctx);
3863 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3864 md5_process_bytes ((unsigned char *)"00000001", 8, &ctx); /* TODO: keep track of server nonce values */
3865 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3866 md5_process_bytes ((unsigned char *)cnonce, strlen(cnonce), &ctx);
3867 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3868 md5_process_bytes ((unsigned char *)qop, strlen(qop), &ctx);
3869 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3870 md5_process_bytes ((unsigned char *)a2buf, MD5_DIGEST_SIZE * 2, &ctx);
3871 md5_finish_ctx (&ctx, hash);
3875 /* RFC 2069 Digest Access Authentication */
3876 /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */
3877 md5_init_ctx (&ctx);
3878 md5_process_bytes ((unsigned char *)a1buf, MD5_DIGEST_SIZE * 2, &ctx);
3879 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3880 md5_process_bytes ((unsigned char *)nonce, strlen (nonce), &ctx);
3881 md5_process_bytes ((unsigned char *)":", 1, &ctx);
3882 md5_process_bytes ((unsigned char *)a2buf, MD5_DIGEST_SIZE * 2, &ctx);
3883 md5_finish_ctx (&ctx, hash);
3886 dump_hash (response_digest, hash);
3888 res_size = strlen (user)
3892 + 2 * MD5_DIGEST_SIZE /*strlen (response_digest)*/
3893 + (opaque ? strlen (opaque) : 0)
3894 + (algorithm ? strlen (algorithm) : 0)
3899 res = xmalloc (res_size);
3901 if (qop && !strcmp (qop, "auth"))
3903 res_len = snprintf (res, res_size, "Digest "\
3904 "username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\""\
3905 ", qop=auth, nc=00000001, cnonce=\"%s\"",
3906 user, realm, nonce, path, response_digest, cnonce);
3911 res_len = snprintf (res, res_size, "Digest "\
3912 "username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
3913 user, realm, nonce, path, response_digest);
3918 res_len += snprintf(res + res_len, res_size - res_len, ", opaque=\"%s\"", opaque);
3923 snprintf(res + res_len, res_size - res_len, ", algorithm=\"%s\"", algorithm);
3928 xfree_null (opaque);
3931 xfree_null (algorithm);
3935 #endif /* ENABLE_DIGEST */
3937 /* Computing the size of a string literal must take into account that
3938 value returned by sizeof includes the terminating \0. */
3939 #define STRSIZE(literal) (sizeof (literal) - 1)
3941 /* Whether chars in [b, e) begin with the literal string provided as
3942 first argument and are followed by whitespace or terminating \0.
3943 The comparison is case-insensitive. */
3944 #define STARTS(literal, b, e) \
3946 && ((size_t) ((e) - (b))) >= STRSIZE (literal) \
3947 && 0 == strncasecmp (b, literal, STRSIZE (literal)) \
3948 && ((size_t) ((e) - (b)) == STRSIZE (literal) \
3949 || c_isspace (b[STRSIZE (literal)])))
3952 known_authentication_scheme_p (const char *hdrbeg, const char *hdrend)
3954 return STARTS ("Basic", hdrbeg, hdrend)
3955 #ifdef ENABLE_DIGEST
3956 || STARTS ("Digest", hdrbeg, hdrend)
3959 || STARTS ("NTLM", hdrbeg, hdrend)
3966 /* Create the HTTP authorization request header. When the
3967 `WWW-Authenticate' response header is seen, according to the
3968 authorization scheme specified in that header (`Basic' and `Digest'
3969 are supported by the current implementation), produce an
3970 appropriate HTTP authorization request header. */
3972 create_authorization_line (const char *au, const char *user,
3973 const char *passwd, const char *method,
3974 const char *path, bool *finished, uerr_t *auth_err)
3976 /* We are called only with known schemes, so we can dispatch on the
3978 switch (c_toupper (*au))
3980 case 'B': /* Basic */
3982 return basic_authentication_encode (user, passwd);
3983 #ifdef ENABLE_DIGEST
3984 case 'D': /* Digest */
3986 return digest_authentication_encode (au, user, passwd, method, path, auth_err);
3989 case 'N': /* NTLM */
3990 if (!ntlm_input (&pconn.ntlm, au))
3995 return ntlm_output (&pconn.ntlm, user, passwd, finished);
3998 /* We shouldn't get here -- this function should be only called
3999 with values approved by known_authentication_scheme_p. */
4007 if (!wget_cookie_jar)
4008 wget_cookie_jar = cookie_jar_new ();
4009 if (opt.cookies_input && !cookies_loaded_p)
4011 cookie_jar_load (wget_cookie_jar, opt.cookies_input);
4012 cookies_loaded_p = true;
4019 if (wget_cookie_jar)
4020 cookie_jar_save (wget_cookie_jar, opt.cookies_output);
4026 xfree_null (pconn.host);
4027 if (wget_cookie_jar)
4028 cookie_jar_delete (wget_cookie_jar);
4032 ensure_extension (struct http_stat *hs, const char *ext, int *dt)
4034 char *last_period_in_local_filename = strrchr (hs->local_file, '.');
4036 int len = strlen (ext);
4039 strncpy (shortext, ext, len - 1);
4040 shortext[len - 1] = '\0';
4043 if (last_period_in_local_filename == NULL
4044 || !(0 == strcasecmp (last_period_in_local_filename, shortext)
4045 || 0 == strcasecmp (last_period_in_local_filename, ext)))
4047 int local_filename_len = strlen (hs->local_file);
4048 /* Resize the local file, allowing for ".html" preceded by
4049 optional ".NUMBER". */
4050 hs->local_file = xrealloc (hs->local_file,
4051 local_filename_len + 24 + len);
4052 strcpy (hs->local_file + local_filename_len, ext);
4053 /* If clobbering is not allowed and the file, as named,
4054 exists, tack on ".NUMBER.html" instead. */
4055 if (!ALLOW_CLOBBER && file_exists_p (hs->local_file))
4059 sprintf (hs->local_file + local_filename_len,
4060 ".%d%s", ext_num++, ext);
4061 while (file_exists_p (hs->local_file));
4063 *dt |= ADDED_HTML_EXTENSION;
4071 test_parse_content_disposition(void)
4074 static const struct {
4076 const char *filename;
4079 { "filename=\"file.ext\"", "file.ext", true },
4080 { "attachment; filename=\"file.ext\"", "file.ext", true },
4081 { "attachment; filename=\"file.ext\"; dummy", "file.ext", true },
4082 { "attachment", NULL, false },
4083 { "attachement; filename*=UTF-8'en-US'hello.txt", "hello.txt", true },
4084 { "attachement; filename*0=\"hello\"; filename*1=\"world.txt\"", "helloworld.txt", true },
4087 for (i = 0; i < countof(test_array); ++i)
4092 res = parse_content_disposition (test_array[i].hdrval, &filename);
4094 mu_assert ("test_parse_content_disposition: wrong result",
4095 res == test_array[i].result
4097 || 0 == strcmp (test_array[i].filename, filename)));
4103 #endif /* TESTING */
4106 * vim: et sts=2 sw=2 cino+={s