src/protocol/proxy.c

   1 /* Proxy handling */
   2
   3 #ifndef _GNU_SOURCE
   4 #define _GNU_SOURCE /* XXX: we _WANT_ strcasestr() ! */
   5 #endif
   6
   7 #ifdef HAVE_CONFIG_H
   8 #include "config.h"
   9 #endif
  10
  11 #include <stdlib.h>
  12 #include <string.h>
  13
  14 #include "elinks.h"
  15
  16 #include "config/options.h"
  17 #include "main/event.h"
  18 #include "network/connection.h"
  19 #include "network/state.h"
  20 #include "protocol/protocol.h"
  21 #include "protocol/proxy.h"
  22 #include "protocol/uri.h"
  23 #include "util/memory.h"
  24 #include "util/string.h"
  25
  26
  27 static int
  28 proxy_probe_no_proxy(unsigned char *url, unsigned char *no_proxy)
  29 {
  30         unsigned char *slash = strchr(url, '/');
  31
  32         if (slash) *slash = '\0';
  33
  34         while (no_proxy && *no_proxy) {
  35                 unsigned char *jumper = strchr(no_proxy, ',');
  36
  37                 skip_space(no_proxy);
  38                 if (jumper) *jumper = '\0';
  39
  40                 if (c_strcasestr(url, no_proxy)) {
  41                         if (jumper) *jumper = ',';
  42                         if (slash) *slash = '/';
  43                         return 1;
  44                 }
  45                 no_proxy = jumper;
  46                 if (jumper) {
  47                         *jumper = ',';
  48                         no_proxy++;
  49                 }
  50         }
  51
  52         if (slash) *slash = '/';
  53         return 0;
  54 }
  55
  56 static struct uri *
  57 proxy_uri(struct uri *uri, unsigned char *proxy,
  58           struct connection_state *error_state)
  59 {
  60         struct string string;
  61
  62         if (init_string(&string)
  63             && string_concat(&string, "proxy://", proxy, "/",
  64                              (unsigned char *) NULL)
  65             && add_uri_to_string(&string, uri, URI_BASE)) {
  66                 /* There is no need to use URI_BASE when calling get_uri()
  67                  * because URI_BASE should not add any fragments in the first
  68                  * place. */
  69                 uri = get_uri(string.source, 0);
  70                 /* XXX: Assume the problem is due to @proxy having bad format.
  71                  * This is a lot faster easier than checking the format. */
  72                 if (!uri)
  73                         *error_state = connection_state(S_PROXY_ERROR);
  74         } else {
  75                 uri = NULL;
  76                 *error_state = connection_state(S_OUT_OF_MEM);
  77         }
  78
  79         done_string(&string);
  80         return uri;
  81 }
  82
  83 static unsigned char *
  84 strip_proxy_protocol(unsigned char *proxy,
  85                      unsigned char *strip1, unsigned char *strip2)
  86 {
  87         assert(proxy && *proxy);
  88
  89         if (!c_strncasecmp(proxy, strip1, strlen(strip1)))
  90                 proxy += strlen(strip1);
  91         else if (strip2 && !c_strncasecmp(proxy, strip2, strlen(strip2)))
  92                 proxy += strlen(strip2);
  93
  94         return proxy;
  95 }
  96
  97 /* TODO: We could of course significantly simplify the calling convention by
  98  * autogenerating most of the parameters from protocol name. Having a function
  99  * exported by protocol/protocol.* dedicated to that would be nice too.
 100  * --pasky */
 101 static unsigned char *
 102 get_protocol_proxy(unsigned char *opt,
 103                    unsigned char *env1, unsigned char *env2,
 104                    unsigned char *strip1, unsigned char *strip2)
 105 {
 106         unsigned char *proxy;
 107
 108         proxy = get_opt_str(opt, NULL);
 109         if (!*proxy) proxy = getenv(env1);
 110         if (!proxy || !*proxy) proxy = getenv(env2);
 111
 112         if (proxy && *proxy) {
 113                 proxy = strip_proxy_protocol(proxy, strip1, strip2);
 114         }
 115
 116         return proxy;
 117 }
 118
 119 static struct uri *
 120 get_proxy_worker(struct uri *uri, unsigned char *proxy,
 121                  struct connection_state *error_state)
 122 {
 123         unsigned char *protocol_proxy = NULL;
 124
 125         if (proxy) {
 126                 if (*proxy) {
 127                         proxy = strip_proxy_protocol(proxy, "http://", "ftp://");
 128
 129                         return proxy_uri(uri, proxy, error_state);
 130                 }
 131
 132                 /* "" from script_hook_get_proxy() */
 133                 return get_composed_uri(uri, URI_BASE);
 134         }
 135
 136         switch (uri->protocol) {
 137         case PROTOCOL_HTTP:
 138                 protocol_proxy = get_protocol_proxy("protocol.http.proxy.host",
 139                                                     "HTTP_PROXY", "http_proxy",
 140                                                     "http://", NULL);
 141                 break;
 142
 143         case PROTOCOL_HTTPS:
 144                 /* As Timo Lindfors explains, the communication between ELinks
 145                  * and the proxy server is never encrypted, altho the proxy
 146                  * might be used to transfer encrypted data between Web client
 147                  * and Web server. (Some proxy servers might allow encrypted
 148                  * communication between the Web client and the proxy
 149                  * but ELinks does not support that.) */
 150                 /* So, don't check whether the URI for the proxy begins
 151                  * with "https://" but rather check for "http://".
 152                  * Maybe we should allow either -- ELinks uses HTTP
 153                  * to communicate with the proxy when we use it for FTP, but we
 154                  * check for "ftp://" below; and what about 'be liberal in what
 155                  * you accept' (altho that is usually applied to data received
 156                  * from remote systems, not to user input)?  -- Miciah */
 157                 protocol_proxy = get_protocol_proxy("protocol.https.proxy.host",
 158                                                     "HTTPS_PROXY", "https_proxy",
 159                                                     "http://", NULL);
 160                 break;
 161
 162         case PROTOCOL_FTP:
 163                 protocol_proxy = get_protocol_proxy("protocol.ftp.proxy.host",
 164                                                     "FTP_PROXY", "ftp_proxy",
 165                                                     "ftp://", "http://");
 166                 break;
 167         }
 168
 169         if (protocol_proxy && *protocol_proxy) {
 170                 unsigned char *no_proxy;
 171                 unsigned char *slash = strchr(protocol_proxy, '/');
 172
 173                 if (slash) *slash = 0;
 174
 175                 no_proxy = get_opt_str("protocol.no_proxy", NULL);
 176                 if (!*no_proxy) no_proxy = getenv("NO_PROXY");
 177                 if (!no_proxy || !*no_proxy) no_proxy = getenv("no_proxy");
 178
 179                 if (!proxy_probe_no_proxy(uri->host, no_proxy))
 180                         return proxy_uri(uri, protocol_proxy, error_state);
 181         }
 182
 183         return get_composed_uri(uri, URI_BASE);
 184 }
 185
 186 struct uri *
 187 get_proxy_uri(struct uri *uri, struct connection_state *error_state)
 188 {
 189         if (uri->protocol == PROTOCOL_PROXY) {
 190                 return get_composed_uri(uri, URI_BASE);
 191         } else {
 192 #ifdef CONFIG_SCRIPTING
 193                 unsigned char *tmp = NULL;
 194                 static int get_proxy_event_id = EVENT_NONE;
 195
 196                 set_event_id(get_proxy_event_id, "get-proxy");
 197                 trigger_event(get_proxy_event_id, &tmp, struri(uri));
 198
 199                 uri = get_proxy_worker(uri, tmp, error_state);
 200                 mem_free_if(tmp);
 201                 return uri;
 202 #else
 203                 return get_proxy_worker(uri, NULL, error_state);
 204 #endif
 205         }
 206 }
 207
 208 struct uri *
 209 get_proxied_uri(struct uri *uri)
 210 {
 211         if (uri->protocol == PROTOCOL_PROXY)
 212                 return get_uri(uri->data, URI_BASE);
 213
 214         return get_composed_uri(uri, URI_BASE);
 215 }