ext/unicorn/http11/http11.c

   1 /**
   2  * Copyright (c) 2009 Eric Wong (all bugs are Eric's fault)
   3  * Copyright (c) 2005 Zed A. Shaw
   4  * You can redistribute it and/or modify it under the same terms as Ruby.
   5  */
   6 #include "ruby.h"
   7 #include "ext_help.h"
   8 #include <assert.h>
   9 #include <string.h>
  10 #include "http11_parser.h"
  11
  12 #ifndef RSTRING_PTR
  13 #define RSTRING_PTR(s) (RSTRING(s)->ptr)
  14 #endif
  15 #ifndef RSTRING_LEN
  16 #define RSTRING_LEN(s) (RSTRING(s)->len)
  17 #endif
  18
  19 static VALUE mUnicorn;
  20 static VALUE cHttpParser;
  21 static VALUE eHttpParserError;
  22 static VALUE sym_http_body;
  23
  24 #define HTTP_PREFIX "HTTP_"
  25 #define HTTP_PREFIX_LEN (sizeof(HTTP_PREFIX) - 1)
  26
  27 static VALUE global_rack_url_scheme;
  28 static VALUE global_request_method;
  29 static VALUE global_request_uri;
  30 static VALUE global_fragment;
  31 static VALUE global_query_string;
  32 static VALUE global_http_version;
  33 static VALUE global_content_length;
  34 static VALUE global_request_path;
  35 static VALUE global_content_type;
  36 static VALUE global_server_name;
  37 static VALUE global_server_port;
  38 static VALUE global_server_protocol;
  39 static VALUE global_server_protocol_value;
  40 static VALUE global_http_host;
  41 static VALUE global_http_x_forwarded_proto;
  42 static VALUE global_port_80;
  43 static VALUE global_port_443;
  44 static VALUE global_localhost;
  45 static VALUE global_http;
  46
  47 /** Defines common length and error messages for input length validation. */
  48 #define DEF_MAX_LENGTH(N,length) const size_t MAX_##N##_LENGTH = length; const char *MAX_##N##_LENGTH_ERR = "HTTP element " # N  " is longer than the " # length " allowed length."
  49
  50 /** Validates the max length of given input and throws an HttpParserError exception if over. */
  51 #define VALIDATE_MAX_LENGTH(len, N) if(len > MAX_##N##_LENGTH) { rb_raise(eHttpParserError, MAX_##N##_LENGTH_ERR); }
  52
  53 /** Defines global strings in the init method. */
  54 #define DEF_GLOBAL(N, val)   global_##N = rb_obj_freeze(rb_str_new2(val)); rb_global_variable(&global_##N)
  55
  56
  57 /* Defines the maximum allowed lengths for various input elements.*/
  58 DEF_MAX_LENGTH(FIELD_NAME, 256);
  59 DEF_MAX_LENGTH(FIELD_VALUE, 80 * 1024);
  60 DEF_MAX_LENGTH(REQUEST_URI, 1024 * 12);
  61 DEF_MAX_LENGTH(FRAGMENT, 1024); /* Don't know if this length is specified somewhere or not */
  62 DEF_MAX_LENGTH(REQUEST_PATH, 1024);
  63 DEF_MAX_LENGTH(QUERY_STRING, (1024 * 10));
  64 DEF_MAX_LENGTH(HEADER, (1024 * (80 + 32)));
  65
  66 struct common_field {
  67         const signed long len;
  68         const char *name;
  69         VALUE value;
  70 };
  71
  72 /*
  73  * A list of common HTTP headers we expect to receive.
  74  * This allows us to avoid repeatedly creating identical string
  75  * objects to be used with rb_hash_aset().
  76  */
  77 static struct common_field common_http_fields[] = {
  78 # define f(N) { (sizeof(N) - 1), N, Qnil }
  79         f("ACCEPT"),
  80         f("ACCEPT_CHARSET"),
  81         f("ACCEPT_ENCODING"),
  82         f("ACCEPT_LANGUAGE"),
  83         f("ALLOW"),
  84         f("AUTHORIZATION"),
  85         f("CACHE_CONTROL"),
  86         f("CONNECTION"),
  87         f("CONTENT_ENCODING"),
  88         f("CONTENT_LENGTH"),
  89         f("CONTENT_TYPE"),
  90         f("COOKIE"),
  91         f("DATE"),
  92         f("EXPECT"),
  93         f("FROM"),
  94         f("HOST"),
  95         f("IF_MATCH"),
  96         f("IF_MODIFIED_SINCE"),
  97         f("IF_NONE_MATCH"),
  98         f("IF_RANGE"),
  99         f("IF_UNMODIFIED_SINCE"),
 100         f("KEEP_ALIVE"), /* Firefox sends this */
 101         f("MAX_FORWARDS"),
 102         f("PRAGMA"),
 103         f("PROXY_AUTHORIZATION"),
 104         f("RANGE"),
 105         f("REFERER"),
 106         f("TE"),
 107         f("TRAILER"),
 108         f("TRANSFER_ENCODING"),
 109         f("UPGRADE"),
 110         f("USER_AGENT"),
 111         f("VIA"),
 112         f("X_FORWARDED_FOR"), /* common for proxies */
 113         f("X_FORWARDED_PROTO"), /* common for proxies */
 114         f("X_REAL_IP"), /* common for proxies */
 115         f("WARNING")
 116 # undef f
 117 };
 118
 119 /* this function is not performance-critical */
 120 static void init_common_fields(void)
 121 {
 122   int i;
 123   struct common_field *cf = common_http_fields;
 124   char tmp[256]; /* MAX_FIELD_NAME_LENGTH */
 125   memcpy(tmp, HTTP_PREFIX, HTTP_PREFIX_LEN);
 126
 127   for(i = 0; i < ARRAY_SIZE(common_http_fields); cf++, i++) {
 128     /* Rack doesn't like certain headers prefixed with "HTTP_" */
 129     if (!strcmp("CONTENT_LENGTH", cf->name) ||
 130         !strcmp("CONTENT_TYPE", cf->name)) {
 131       cf->value = rb_str_new(cf->name, cf->len);
 132     } else {
 133       memcpy(tmp + HTTP_PREFIX_LEN, cf->name, cf->len + 1);
 134       cf->value = rb_str_new(tmp, HTTP_PREFIX_LEN + cf->len);
 135     }
 136     cf->value = rb_obj_freeze(cf->value);
 137     rb_global_variable(&cf->value);
 138   }
 139 }
 140
 141 static VALUE find_common_field_value(const char *field, size_t flen)
 142 {
 143   int i;
 144   struct common_field *cf = common_http_fields;
 145   for(i = 0; i < ARRAY_SIZE(common_http_fields); i++, cf++) {
 146     if (cf->len == flen && !memcmp(cf->name, field, flen))
 147       return cf->value;
 148   }
 149   return Qnil;
 150 }
 151
 152 static void http_field(void *data, const char *field,
 153                        size_t flen, const char *value, size_t vlen)
 154 {
 155   VALUE req = (VALUE)data;
 156   VALUE v = Qnil;
 157   VALUE f = Qnil;
 158
 159   VALIDATE_MAX_LENGTH(flen, FIELD_NAME);
 160   VALIDATE_MAX_LENGTH(vlen, FIELD_VALUE);
 161
 162   v = rb_str_new(value, vlen);
 163
 164   f = find_common_field_value(field, flen);
 165
 166   if (f == Qnil) {
 167     /*
 168      * We got a strange header that we don't have a memoized value for.
 169      * Fallback to creating a new string to use as a hash key.
 170      *
 171      * using rb_str_new(NULL, len) here is faster than rb_str_buf_new(len)
 172      * in my testing, because: there's no minimum allocation length (and
 173      * no check for it, either), RSTRING_LEN(f) does not need to be
 174      * written twice, and and RSTRING_PTR(f) will already be
 175      * null-terminated for us.
 176      */
 177     f = rb_str_new(NULL, HTTP_PREFIX_LEN + flen);
 178     memcpy(RSTRING_PTR(f), HTTP_PREFIX, HTTP_PREFIX_LEN);
 179     memcpy(RSTRING_PTR(f) + HTTP_PREFIX_LEN, field, flen);
 180     assert(*(RSTRING_PTR(f) + RSTRING_LEN(f)) == '\0'); /* paranoia */
 181     /* fprintf(stderr, "UNKNOWN HEADER <%s>\n", RSTRING_PTR(f)); */
 182   }
 183
 184   rb_hash_aset(req, f, v);
 185 }
 186
 187 static void request_method(void *data, const char *at, size_t length)
 188 {
 189   VALUE req = (VALUE)data;
 190   VALUE val = Qnil;
 191
 192   val = rb_str_new(at, length);
 193   rb_hash_aset(req, global_request_method, val);
 194 }
 195
 196 static void request_uri(void *data, const char *at, size_t length)
 197 {
 198   VALUE req = (VALUE)data;
 199   VALUE val = Qnil;
 200
 201   VALIDATE_MAX_LENGTH(length, REQUEST_URI);
 202
 203   val = rb_str_new(at, length);
 204   rb_hash_aset(req, global_request_uri, val);
 205 }
 206
 207 static void fragment(void *data, const char *at, size_t length)
 208 {
 209   VALUE req = (VALUE)data;
 210   VALUE val = Qnil;
 211
 212   VALIDATE_MAX_LENGTH(length, FRAGMENT);
 213
 214   val = rb_str_new(at, length);
 215   rb_hash_aset(req, global_fragment, val);
 216 }
 217
 218 static void request_path(void *data, const char *at, size_t length)
 219 {
 220   VALUE req = (VALUE)data;
 221   VALUE val = Qnil;
 222
 223   VALIDATE_MAX_LENGTH(length, REQUEST_PATH);
 224
 225   val = rb_str_new(at, length);
 226   rb_hash_aset(req, global_request_path, val);
 227 }
 228
 229 static void query_string(void *data, const char *at, size_t length)
 230 {
 231   VALUE req = (VALUE)data;
 232   VALUE val = Qnil;
 233
 234   VALIDATE_MAX_LENGTH(length, QUERY_STRING);
 235
 236   val = rb_str_new(at, length);
 237   rb_hash_aset(req, global_query_string, val);
 238 }
 239
 240 static void http_version(void *data, const char *at, size_t length)
 241 {
 242   VALUE req = (VALUE)data;
 243   VALUE val = rb_str_new(at, length);
 244   rb_hash_aset(req, global_http_version, val);
 245 }
 246
 247 /** Finalizes the request header to have a bunch of stuff that's needed. */
 248 static void header_done(void *data, const char *at, size_t length)
 249 {
 250   VALUE req = (VALUE)data;
 251   VALUE server_name = global_localhost;
 252   VALUE server_port = global_port_80;
 253   VALUE temp;
 254
 255   /* set rack.url_scheme to "https" or "http", no others are allowed by Rack */
 256   if ((temp = rb_hash_aref(req, global_http_x_forwarded_proto)) != Qnil &&
 257       RSTRING_LEN(temp) == 5 &&
 258       !memcmp("https", RSTRING_PTR(temp), 5))
 259     server_port = global_port_443;
 260   else
 261     temp = global_http;
 262   rb_hash_aset(req, global_rack_url_scheme, temp);
 263
 264   /* parse and set the SERVER_NAME and SERVER_PORT variables */
 265   if ((temp = rb_hash_aref(req, global_http_host)) != Qnil) {
 266     char *colon = memchr(RSTRING_PTR(temp), ':', RSTRING_LEN(temp));
 267     if (colon) {
 268       server_name = rb_str_substr(temp, 0, colon - RSTRING_PTR(temp));
 269       server_port = rb_str_substr(temp, colon - RSTRING_PTR(temp)+1,
 270                                   RSTRING_LEN(temp));
 271     } else {
 272       server_name = temp;
 273     }
 274   }
 275   rb_hash_aset(req, global_server_name, server_name);
 276   rb_hash_aset(req, global_server_port, server_port);
 277
 278   /* grab the initial body and stuff it into the hash */
 279   rb_hash_aset(req, sym_http_body, rb_str_new(at, length));
 280   rb_hash_aset(req, global_server_protocol, global_server_protocol_value);
 281 }
 282
 283 static void HttpParser_free(void *data) {
 284   TRACE();
 285
 286   if(data) {
 287     free(data);
 288   }
 289 }
 290
 291
 292 static VALUE HttpParser_alloc(VALUE klass)
 293 {
 294   VALUE obj;
 295   http_parser *hp = ALLOC_N(http_parser, 1);
 296   TRACE();
 297   hp->http_field = http_field;
 298   hp->request_method = request_method;
 299   hp->request_uri = request_uri;
 300   hp->fragment = fragment;
 301   hp->request_path = request_path;
 302   hp->query_string = query_string;
 303   hp->http_version = http_version;
 304   hp->header_done = header_done;
 305   http_parser_init(hp);
 306
 307   obj = Data_Wrap_Struct(klass, NULL, HttpParser_free, hp);
 308
 309   return obj;
 310 }
 311
 312
 313 /**
 314  * call-seq:
 315  *    parser.new -> parser
 316  *
 317  * Creates a new parser.
 318  */
 319 static VALUE HttpParser_init(VALUE self)
 320 {
 321   http_parser *http = NULL;
 322   DATA_GET(self, http_parser, http);
 323   http_parser_init(http);
 324
 325   return self;
 326 }
 327
 328
 329 /**
 330  * call-seq:
 331  *    parser.reset -> nil
 332  *
 333  * Resets the parser to it's initial state so that you can reuse it
 334  * rather than making new ones.
 335  */
 336 static VALUE HttpParser_reset(VALUE self)
 337 {
 338   http_parser *http = NULL;
 339   DATA_GET(self, http_parser, http);
 340   http_parser_init(http);
 341
 342   return Qnil;
 343 }
 344
 345
 346 /**
 347  * call-seq:
 348  *    parser.execute(req_hash, data) -> true/false
 349  *
 350  * Takes a Hash and a String of data, parses the String of data filling
 351  * in the Hash returning a boolean to indicate whether or not parsing
 352  * is finished.
 353  *
 354  * This function now throws an exception when there is a parsing error.
 355  * This makes the logic for working with the parser much easier.  You
 356  * will need to wrap the parser with an exception handling block.
 357  */
 358
 359 static VALUE HttpParser_execute(VALUE self, VALUE req_hash, VALUE data)
 360 {
 361   http_parser *http;
 362   char *dptr = RSTRING_PTR(data);
 363   long dlen = RSTRING_LEN(data);
 364
 365   DATA_GET(self, http_parser, http);
 366
 367   if (http->nread < dlen) {
 368     http->data = (void *)req_hash;
 369     http_parser_execute(http, dptr, dlen);
 370
 371     VALIDATE_MAX_LENGTH(http->nread, HEADER);
 372
 373     if (!http_parser_has_error(http))
 374       return http_parser_is_finished(http) ? Qtrue : Qfalse;
 375
 376     rb_raise(eHttpParserError, "Invalid HTTP format, parsing fails.");
 377   }
 378   rb_raise(eHttpParserError, "Requested start is after data buffer end.");
 379 }
 380
 381 void Init_http11()
 382 {
 383
 384   mUnicorn = rb_define_module("Unicorn");
 385
 386   DEF_GLOBAL(rack_url_scheme, "rack.url_scheme");
 387   DEF_GLOBAL(request_method, "REQUEST_METHOD");
 388   DEF_GLOBAL(request_uri, "REQUEST_URI");
 389   DEF_GLOBAL(fragment, "FRAGMENT");
 390   DEF_GLOBAL(query_string, "QUERY_STRING");
 391   DEF_GLOBAL(http_version, "HTTP_VERSION");
 392   DEF_GLOBAL(request_path, "REQUEST_PATH");
 393   DEF_GLOBAL(content_length, "CONTENT_LENGTH");
 394   DEF_GLOBAL(content_type, "CONTENT_TYPE");
 395   DEF_GLOBAL(server_name, "SERVER_NAME");
 396   DEF_GLOBAL(server_port, "SERVER_PORT");
 397   DEF_GLOBAL(server_protocol, "SERVER_PROTOCOL");
 398   DEF_GLOBAL(server_protocol_value, "HTTP/1.1");
 399   DEF_GLOBAL(http_host, "HTTP_HOST");
 400   DEF_GLOBAL(http_x_forwarded_proto, "HTTP_X_FORWARDED_PROTO");
 401   DEF_GLOBAL(port_80, "80");
 402   DEF_GLOBAL(port_443, "443");
 403   DEF_GLOBAL(localhost, "localhost");
 404   DEF_GLOBAL(http, "http");
 405
 406   eHttpParserError = rb_define_class_under(mUnicorn, "HttpParserError", rb_eIOError);
 407
 408   cHttpParser = rb_define_class_under(mUnicorn, "HttpParser", rb_cObject);
 409   rb_define_alloc_func(cHttpParser, HttpParser_alloc);
 410   rb_define_method(cHttpParser, "initialize", HttpParser_init,0);
 411   rb_define_method(cHttpParser, "reset", HttpParser_reset,0);
 412   rb_define_method(cHttpParser, "execute", HttpParser_execute,2);
 413   sym_http_body = ID2SYM(rb_intern("http_body"));
 414   init_common_fields();
 415 }