Annotation of parser3/src/main/pa_http.C, revision 1.87
1.1 paf 1: /** @file
2: Parser: http support functions.
3:
1.79 moko 4: Copyright (c) 2001-2017 Art. Lebedev Studio (http://www.artlebedev.com)
1.1 paf 5: Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
6: */
7:
8: #include "pa_http.h"
9: #include "pa_common.h"
1.81 moko 10: #include "pa_base64.h"
1.1 paf 11: #include "pa_charsets.h"
12: #include "pa_request_charsets.h"
1.22 misha 13: #include "pa_request.h"
14: #include "pa_vfile.h"
15: #include "pa_random.h"
1.1 paf 16:
1.87 ! moko 17: volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.86 2020/10/11 23:25:10 moko Exp $" IDENT_PA_HTTP_H;
1.59 moko 18:
19: #ifdef _MSC_VER
20: #include <windows.h>
21: #else
22: #define closesocket close
23: #endif
1.53 moko 24:
1.1 paf 25: // defines
26:
1.19 misha 27: #define HTTP_METHOD_NAME "method"
28: #define HTTP_FORM_NAME "form"
29: #define HTTP_BODY_NAME "body"
30: #define HTTP_TIMEOUT_NAME "timeout"
31: #define HTTP_HEADERS_NAME "headers"
1.22 misha 32: #define HTTP_FORM_ENCTYPE_NAME "enctype"
1.19 misha 33: #define HTTP_ANY_STATUS_NAME "any-status"
1.59 moko 34: #define HTTP_OMIT_POST_CHARSET_NAME "omit-post-charset" // ^file::load[...;http://...;$.method[post]] by default adds charset to content-type
1.12 misha 35:
1.1 paf 36: #define HTTP_USER "user"
37: #define HTTP_PASSWORD "password"
38:
1.70 moko 39: #define HTTP_USER_AGENT "user-agent"
1.1 paf 40: #define DEFAULT_USER_AGENT "parser3"
41:
1.59 moko 42: #ifndef INADDR_NONE
43: #define INADDR_NONE ((ulong) -1)
44: #endif
1.1 paf 45:
46: #undef CRLF
47: #define CRLF "\r\n"
48:
1.54 misha 49: // helpers
1.56 misha 50:
1.85 moko 51: bool HTTP_Headers::add_header(const char *line){
1.78 moko 52: const char *value=strchr(line, ':');
53:
54: if(value && value != line){ // we need only headers, not the response code
55: Header header(str_upper(line, value-line), String::Body(value+1).trim(String::TRIM_BOTH, " \t\n\r"));
56:
57: if(header.name == String::Body(HTTP_CONTENT_TYPE_UPPER) && content_type.is_empty())
58: content_type=header.value;
59:
60: if(header.name == String::Body("CONTENT-LENGTH") && content_length==0)
61: content_length=pa_atoul(header.value.cstr(), 10);
62:
63: headers+=header;
64:
65: return true;
66: }
67: return false;
68: }
69:
1.54 misha 70: class Cookies_table_template_columns: public ArrayString {
71: public:
72: Cookies_table_template_columns() {
73: *this+=new String("name");
74: *this+=new String("value");
75: *this+=new String("expires");
76: *this+=new String("max-age");
77: *this+=new String("domain");
78: *this+=new String("path");
79: *this+=new String("httponly");
80: *this+=new String("secure");
81: }
82: };
83:
84:
1.1 paf 85: static bool set_addr(struct sockaddr_in *addr, const char* host, const short port){
1.22 misha 86: memset(addr, 0, sizeof(*addr));
87: addr->sin_family=AF_INET;
88: addr->sin_port=htons(port);
89: if(host) {
1.65 moko 90: struct hostent *hostIP=gethostbyname(host);
91: if(hostIP && hostIP->h_addrtype == AF_INET){
92: memcpy(&addr->sin_addr, hostIP->h_addr, hostIP->h_length);
93: return true;
94: }
95: }
96: return false;
1.1 paf 97: }
98:
1.84 moko 99: class HTTP_response : public PA_Allocated {
1.78 moko 100: public:
101: char *buf;
102: size_t length;
103: size_t buf_size;
104: size_t body_offset;
105:
1.85 moko 106: HTTP_Headers headers;
1.84 moko 107: const String &url;
1.78 moko 108:
1.84 moko 109: HTTP_response(const String& aurl) : buf(NULL), length(0), buf_size(0), body_offset(0), url(aurl){}
1.78 moko 110:
111: void resize(size_t size){
112: buf_size=size;
113: buf=(char *)pa_realloc(buf, size + 1);
114: }
115:
116: bool read(int sock, size_t size){
117: if(length+size>buf_size)
118: resize(buf_size*2 + size);
119: ssize_t received_size=recv(sock, buf + length, size, 0);
120: if(received_size==0)
121: return false;
122: if(received_size<0) {
123: if(int no=pa_socks_errno())
124: throw Exception("http.timeout", &url, "error receiving response body: %s (%d)", pa_socks_strerr(no), no);
125: return false;
126: }
127: length+=received_size;
128: buf[length]='\0';
129: return true;
130: }
131:
1.83 moko 132: size_t first_line(){
1.78 moko 133: char *headers=strchr(buf, '\n');
134: if(!headers)
135: return false;
136:
137: return headers-buf;
138: }
139:
140: const char *status_code(char *status_line, int &result){
141: char* status_start = strchr(status_line, ' ');
142:
143: if(!(status_start++))
144: return status_line;
145:
146: char* status_end=strchr(status_start, ' ');
147:
148: if(!status_end)
149: return status_line;
150:
151: if(status_end==status_start)
152: return status_line;
1.1 paf 153:
1.78 moko 154: const char *result_str=pa_strdup(status_start, status_end-status_start);
1.84 moko 155: result=pa_atoui(result_str, 10);
1.78 moko 156: return result_str;
157: }
1.2 paf 158:
1.78 moko 159: bool body_start(){
160: char *p=buf;
161: while((p=strchr(p, '\n'))) {
162: if(p[1]=='\r' && p[2]=='\n'){ // \r\n\r\n
163: *p='\0';
164: body_offset=p-buf+3;
165: return true;
166: }
167: if(p[1]=='\n') { // \n\n
168: *p='\0';
169: body_offset=p-buf+2;
170: return true;
171: }
172: p++;
173: }
174: return false;
1.2 paf 175: }
1.78 moko 176:
177: void parse_headers(){
178: const String header_block(buf, String::L_TAINTED);
179:
180: ArrayString aheaders;
181: header_block.split(aheaders, 0, "\n");
182:
183: Array_iterator<const String*> i(aheaders);
184: i.next(); // skipping status
185: for(;i.has_next();){
186: const char *line=i.next()->cstr();
187: if(!headers.add_header(line))
188: throw Exception("http.response", &url, "bad response from host - bad header \"%s\"", line);
189: }
1.1 paf 190: }
191:
1.78 moko 192: };
193:
194: enum HTTP_response_state {
195: HTTP_STATUS_CODE,
196: HTTP_HEADERS,
197: HTTP_BODY
198: };
199:
200: static int http_read_response(HTTP_response& response, int sock, bool fail_on_status_ne_200) {
201: HTTP_response_state state=HTTP_STATUS_CODE;
202: int result=0;
203:
204: size_t chunk_size=0x400*16;
205: response.resize(2*chunk_size);
206:
207: while(response.read(sock, chunk_size)){
208: switch(state){
209: case HTTP_STATUS_CODE: {
1.83 moko 210: size_t status_size=response.first_line();
1.78 moko 211: if(!status_size)
212: break;
213:
214: const char *status=response.status_code(pa_strdup(response.buf, status_size), result);
215:
216: if(!result || fail_on_status_ne_200 && result!=200)
217: throw Exception("http.status", status ? new String(status) : &String::Empty, "invalid HTTP response status");
218:
219: state=HTTP_HEADERS;
220: }
221:
222: case HTTP_HEADERS: {
223: if(!response.body_start())
224: break;
225:
226: response.parse_headers();
227:
228: size_t content_length=check_file_size(response.headers.content_length, response.url);
229: if(content_length>0 && (content_length + response.body_offset) > response.length){
230: response.resize(content_length + response.body_offset + 0x400*64);
231: }
232:
233: state=HTTP_BODY;
1.1 paf 234: break;
235: }
1.78 moko 236:
237: case HTTP_BODY: {
238: chunk_size=0x400*64;
1.1 paf 239: break;
240: }
241: }
242: }
1.78 moko 243:
244: if(state==HTTP_STATUS_CODE)
245: throw Exception("http.response", &response.url, "bad response from host - no status found (size=%u)", response.length);
246:
247: if(state==HTTP_HEADERS){
248: response.parse_headers();
249: response.body_offset=response.length;
1.1 paf 250: }
1.78 moko 251:
252: return result;
1.1 paf 253: }
254:
255: /* ********************** request *************************** */
256:
257: #if defined(SIGALRM) && defined(HAVE_SIGSETJMP) && defined(HAVE_SIGLONGJMP)
258: # define PA_USE_ALARM
259: #endif
260:
261: #ifdef PA_USE_ALARM
262: static sigjmp_buf timeout_env;
263: static void timeout_handler(int /*sig*/){
1.22 misha 264: siglongjmp(timeout_env, 1);
1.1 paf 265: }
266: #endif
267:
1.78 moko 268: static int http_request(HTTP_response& response, const char* host, short port, const char* request, size_t request_size, int timeout_secs, bool fail_on_status_ne_200) {
1.1 paf 269: if(!host)
1.73 moko 270: throw Exception("http.host", 0, "zero hostname"); //never
1.1 paf 271:
272: volatile // to prevent makeing it register variable, because it will be clobbered by longjmp [thanks gcc warning]
273: int sock=-1;
274: #ifdef PA_USE_ALARM
275: signal(SIGALRM, timeout_handler);
276: #endif
277: #ifdef PA_USE_ALARM
278: if(sigsetjmp(timeout_env, 1)) {
279: // stupid gcc [2.95.4] generated bad code
280: // which failed to handle sigsetjmp+throw: crashed inside of pre-throw code.
281: // rewritten simplier [athough duplicating closesocket code]
282: if(sock>=0)
283: closesocket(sock);
1.80 moko 284: throw Exception("http.timeout", 0, "timeout occurred while retrieving document");
1.1 paf 285: return 0; // never
286: } else {
287: alarm(timeout_secs);
288: #endif
289: try {
290: int result;
291: struct sockaddr_in dest;
292:
293: if(!set_addr(&dest, host, port))
1.73 moko 294: throw Exception("http.host", 0, "can not resolve hostname \"%s\"", host);
1.1 paf 295:
296: if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) {
297: int no=pa_socks_errno();
1.73 moko 298: throw Exception("http.connect", 0, "can not make socket: %s (%d)", pa_socks_strerr(no), no);
1.1 paf 299: }
300:
301: // To enable SO_DONTLINGER (that is, disable SO_LINGER)
302: // l_onoff should be set to zero and setsockopt should be called
303: linger dont_linger={0,0};
304: setsockopt(sock, SOL_SOCKET, SO_LINGER, (const char *)&dont_linger, sizeof(dont_linger));
305:
306: #ifdef WIN32
307: // SO_*TIMEO can be defined in .h but not implemlemented in protocol,
308: // failing subsequently with Option not supported by protocol (99) message
309: // could not suppress that, so leaving this only for win32
310: int timeout_ms=timeout_secs*1000;
311: setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO, (const char*)&timeout_ms, sizeof(timeout_ms));
312: setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, (const char*)&timeout_ms, sizeof(timeout_ms));
313: #endif
314:
315: if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) {
316: int no=pa_socks_errno();
1.78 moko 317: throw Exception("http.connect", 0, "can not connect to host \"%s\": %s (%d)", host, pa_socks_strerr(no), no);
1.1 paf 318: }
1.22 misha 319:
1.1 paf 320: if(send(sock, request, request_size, 0)!=(ssize_t)request_size) {
321: int no=pa_socks_errno();
1.78 moko 322: throw Exception("http.timeout", 0, "error sending request: %s (%d)", pa_socks_strerr(no), no);
1.1 paf 323: }
324:
1.78 moko 325: result=http_read_response(response, sock, fail_on_status_ne_200);
326: closesocket(sock);
1.1 paf 327: #ifdef PA_USE_ALARM
1.78 moko 328: alarm(0);
1.1 paf 329: #endif
330: return result;
331: } catch(...) {
332: #ifdef PA_USE_ALARM
1.78 moko 333: alarm(0);
1.1 paf 334: #endif
1.78 moko 335: if(sock>=0)
336: closesocket(sock);
1.1 paf 337: rethrow;
338: }
339: #ifdef PA_USE_ALARM
340: }
341: #endif
342: }
343:
344: #ifndef DOXYGEN
345: struct Http_pass_header_info {
346: Request_charsets* charsets;
347: String* request;
1.35 misha 348: bool* user_agent_specified;
349: bool* content_type_specified;
350: bool* content_type_url_encoded;
1.1 paf 351: };
352: #endif
1.50 moko 353:
354: char *pa_http_safe_header_name(const char *name) {
355: char *result=pa_strdup(name);
356: char *n=result;
1.52 misha 357: if(!pa_isalpha((unsigned char)*n))
1.50 moko 358: *n++ = '_';
359: for(; *n; ++n) {
1.52 misha 360: if (!pa_isalnum((unsigned char)*n) && *n != '-' && *n != '_')
1.50 moko 361: *n = '_';
362: }
363: return result;
364: }
365:
1.35 misha 366: static void http_pass_header(HashStringValue::key_type aname,
367: HashStringValue::value_type avalue,
1.22 misha 368: Http_pass_header_info *info) {
1.9 misha 369:
1.41 misha 370: const char* name_cstr=aname.cstr();
371:
1.38 misha 372: if(strcasecmp(name_cstr, HTTP_CONTENT_LENGTH)==0)
373: return;
374:
1.50 moko 375: String name=String(pa_http_safe_header_name(capitalize(name_cstr)), String::L_AS_IS);
376: String value=attributed_meaning_to_string(*avalue, String::L_HTTP_HEADER, true);
1.9 misha 377:
1.35 misha 378: *info->request << name << ": " << value << CRLF;
1.1 paf 379:
1.38 misha 380: if(strcasecmp(name_cstr, HTTP_USER_AGENT)==0)
1.35 misha 381: *info->user_agent_specified=true;
1.38 misha 382: if(strcasecmp(name_cstr, HTTP_CONTENT_TYPE)==0){
1.35 misha 383: *info->content_type_specified=true;
1.62 moko 384: *info->content_type_url_encoded=pa_strncasecmp(value.cstr(), HTTP_CONTENT_TYPE_FORM_URLENCODED)==0;
1.35 misha 385: }
1.1 paf 386: }
387:
1.10 misha 388: static void http_pass_cookie(HashStringValue::key_type name,
1.20 misha 389: HashStringValue::value_type value,
390: Http_pass_header_info *info) {
1.10 misha 391:
1.17 misha 392: *info->request << String(name, String::L_HTTP_COOKIE) << "="
1.31 misha 393: << attributed_meaning_to_string(*value, String::L_HTTP_COOKIE, true)
1.10 misha 394: << "; ";
395:
396: }
1.1 paf 397:
398: static const String* basic_authorization_field(const char* user, const char* pass) {
399: if(!user&& !pass)
400: return 0;
401:
402: String combined;
403: if(user)
404: combined<<user;
405: combined<<":";
406: if(pass)
407: combined<<pass;
408:
1.20 misha 409: String* result=new String("Basic ");
1.82 moko 410: *result<<pa_base64_encode(combined.cstr(), combined.length(), Base64Options(false /*no wrap*/));
1.1 paf 411: return result;
412: }
413:
1.73 moko 414: static void form_string_value2string(HashStringValue::key_type key, const String& value, String& result) {
1.30 misha 415: result << String(key, String::L_URI) << "=" << String(value, String::L_URI) << "&";
1.1 paf 416: }
1.20 misha 417:
1.1 paf 418: #ifndef DOXYGEN
419: struct Form_table_value2string_info {
420: HashStringValue::key_type key;
421: String& result;
422:
423: Form_table_value2string_info(HashStringValue::key_type akey, String& aresult):
424: key(akey), result(aresult) {}
425: };
426: #endif
427: static void form_table_value2string(Table::element_type row, Form_table_value2string_info* info) {
428: form_string_value2string(info->key, *row->get(0), info->result);
429: }
1.73 moko 430:
431: static void form_value2string(HashStringValue::key_type key, HashStringValue::value_type value, String* result) {
1.1 paf 432: if(const String* svalue=value->get_string())
433: form_string_value2string(key, *svalue, *result);
434: else if(Table* tvalue=value->get_table()) {
435: Form_table_value2string_info info(key, *result);
436: tvalue->for_each(form_table_value2string, &info);
437: } else
1.73 moko 438: throw Exception(PARSER_RUNTIME, new String(key, String::L_TAINTED),
1.63 moko 439: "is %s, " HTTP_FORM_NAME " option value can be string or table only (file is allowed for $." HTTP_METHOD_NAME "[POST] + $." HTTP_FORM_ENCTYPE_NAME "[" HTTP_CONTENT_TYPE_MULTIPART_FORMDATA "])", value->type());
1.1 paf 440: }
1.20 misha 441:
1.5 misha 442: const char* pa_form2string(HashStringValue& form, Request_charsets& charsets) {
1.1 paf 443: String string;
1.3 paf 444: form.for_each<String*>(form_value2string, &string);
1.44 misha 445: return string.untaint_and_transcode_cstr(String::L_URI, &charsets);
1.1 paf 446: }
1.22 misha 447:
448: struct FormPart {
449: Request* r;
450: const char* boundary;
1.48 moko 451: String* string;
1.22 misha 452: Form_table_value2string_info* info;
1.48 moko 453:
454: struct BinaryBlock{
455: const char* ptr;
456: size_t length;
457:
458: BinaryBlock(String* astring, Request* r): ptr(astring->untaint_and_transcode_cstr(String::L_AS_IS, &r->charsets)), length(strlen(ptr)){}
459: BinaryBlock(const char* aptr, size_t alength): ptr(aptr), length(alength){}
460: };
461:
462: Array<BinaryBlock> blocks;
463:
464: FormPart(Request* ar, const char* aboundary): r(ar), boundary(aboundary), string(new String()){}
465:
466: const char *post(size_t &length){
467: if(blocks.count()){
468: blocks+=BinaryBlock(string, r);
469:
470: length=0;
471: for(size_t i=0; i<blocks.count(); i++)
472: length+=blocks[i].length;
473:
474: char *result=(char *)pa_malloc_atomic(length);
475: char *ptr=result;
476:
477: for(size_t i=0; i<blocks.count(); i++){
478: memcpy(ptr, blocks[i].ptr, blocks[i].length);
479: ptr+=blocks[i].length;
480: }
481:
482: return result;
483: } else {
484: BinaryBlock result(string, r);
485: length=result.length;
486: return result.ptr;
487: }
488: }
1.22 misha 489: };
490:
1.73 moko 491: static void form_part_boundary_header(FormPart& part, String::Body name, const char* file_name=0) {
492: *part.string << "--" << part.boundary << CRLF CONTENT_DISPOSITION_CAPITALIZED ": form-data; name=\"" << name << "\"";
1.22 misha 493: if(file_name){
494: if(strcmp(file_name, NONAME_DAT)!=0)
1.48 moko 495: *part.string << "; filename=\"" << file_name << "\"";
496: *part.string << CRLF HTTP_CONTENT_TYPE_CAPITALIZED ": " << part.r->mime_type_of(file_name);
1.22 misha 497: }
1.48 moko 498: *part.string << CRLF CRLF;
1.22 misha 499: }
500:
1.73 moko 501: static void form_string_value2part(HashStringValue::key_type key, const String& value, FormPart& part) {
1.28 misha 502: form_part_boundary_header(part, key);
1.48 moko 503: *part.string << value << CRLF;
1.22 misha 504: }
505:
1.73 moko 506: static void form_file_value2part(HashStringValue::key_type key, VFile& vfile, FormPart& part) {
1.28 misha 507: form_part_boundary_header(part, key, vfile.fields().get(name_name)->as_string().cstr());
1.48 moko 508: part.blocks+=FormPart::BinaryBlock(part.string, part.r);
509: part.blocks+=FormPart::BinaryBlock(vfile.value_ptr(), vfile.value_size());
510: part.string=new String();
511: *part.string << CRLF;
1.22 misha 512: }
513:
514: static void form_table_value2part(Table::element_type row, FormPart* part) {
515: form_string_value2part(part->info->key, *row->get(0), *part);
516: }
517:
1.73 moko 518: static void form_value2part(HashStringValue::key_type key, HashStringValue::value_type value, FormPart& part) {
1.22 misha 519: if(const String* svalue=value->get_string())
520: form_string_value2part(key, *svalue, part);
521: else if(Table* tvalue=value->get_table()) {
1.48 moko 522: Form_table_value2string_info info(key, *part.string);
1.22 misha 523: part.info = &info;
524: tvalue->for_each(form_table_value2part, &part);
1.33 misha 525: } else if(VFile* vfile=static_cast<VFile *>(value->as("file"))){
1.22 misha 526: form_file_value2part(key, *vfile, part);
527: } else
1.73 moko 528: throw Exception(PARSER_RUNTIME, new String(key, String::L_TAINTED), "is %s, " HTTP_FORM_NAME " option value can be string, table or file only", value->type());
1.22 misha 529: }
530:
531: const char* pa_form2string_multipart(HashStringValue& form, Request& r, const char* boundary, size_t& post_size){
1.48 moko 532: FormPart formpart(&r, boundary);
1.22 misha 533: form.for_each<FormPart&>(form_value2part, formpart);
1.48 moko 534: *formpart.string << "--" << boundary << "--";
535: // @todo: return binary blocks here to save memory in pa_internal_file_read_http
536: return formpart.post(post_size);
1.22 misha 537: }
538:
1.54 misha 539: // Set-Cookie: name=value; Domain=docs.foo.com; Path=/accounts; Expires=Wed, 13-Jan-2021 22:23:01 GMT; Secure; HttpOnly
540: static ArrayString* parse_cookie(Request& r, const String& cookie) {
1.64 moko 541: char *current=pa_strdup(cookie.cstr());
1.54 misha 542:
543: const String* name=0;
1.55 moko 544: const String* value=&String::Empty;
545: const String* expires=&String::Empty;
546: const String* max_age=&String::Empty;
547: const String* path=&String::Empty;
548: const String* domain=&String::Empty;
549: const String* httponly=&String::Empty;
550: const String* secure=&String::Empty;
1.54 misha 551:
552: bool first_pair=true;
553:
554: do {
555: if(char *meaning=search_stop(current, ';'))
556: if(char *attribute=search_stop(meaning, '=')) {
557: const String* sname=new String(unescape_chars(attribute, strlen(attribute), &r.charsets.source(), true/*don't convert '"' to space*/), String::L_TAINTED);
558: const String* smeaning=0;
559: if(meaning)
560: smeaning=new String(unescape_chars(meaning, strlen(meaning), &r.charsets.source(), true/*don't convert '"' to space*/), String::L_TAINTED);
561:
562: if(first_pair) {
563: // name + value
564: name=sname;
565: value=smeaning;
566: first_pair=false;
567: } else {
568: const String& slower=sname->change_case(r.charsets.source(), String::CC_LOWER);
569:
570: if(slower == "expires")
571: expires=smeaning;
572: else if(slower == "max-age")
573: max_age=smeaning;
574: else if(slower == "domain")
575: domain=smeaning;
576: else if(slower == "path")
577: path=smeaning;
578: else if(slower == "httponly")
579: httponly=new String("1", String::L_CLEAN);
580: else if(slower == "secure")
581: secure=new String("1", String::L_CLEAN);
582: else {
583: // todo@ ?
584: }
585: }
586: }
587: } while(current);
588:
589: if(!name)
590: return 0;
591:
592: ArrayString* result=new ArrayString(8);
593: *result+=name;
594: *result+=value;
595: *result+=expires;
596: *result+=max_age;
597: *result+=domain;
598: *result+=path;
599: *result+=httponly;
600: *result+=secure;
601:
602: return result;
603: }
604:
1.56 misha 605: Table* parse_cookies(Request& r, Table *cookies){
606: Table& result=*new Table(new Cookies_table_template_columns);
607:
608: for(Array_iterator<Table::element_type> i(*cookies); i.has_next(); )
609: if(ArrayString* row=parse_cookie(r, *i.next()->get(0)))
610: result+=row;
611:
612: return &result;
613: }
614:
1.75 moko 615: void tables_update(HashStringValue& tables, const String::Body name, const String& value){
1.72 moko 616: Table *table;
617: if(Value *valready=tables.get(name)) {
618: // second+ appearence
619: table=valready->get_table();
620: } else {
621: // first appearence
622: Table::columns_type columns=new ArrayString(1);
623: *columns+=new String("value");
624: table=new Table(columns);
625: tables.put(name, new VTable(table));
626: }
627: // this string becomes next row
628: ArrayString& row=*new ArrayString(1);
629: row+=&value;
630: *table+=&row;
631: }
632:
1.1 paf 633: /// @todo build .cookies field. use ^file.tables.SET-COOKIES.menu{ for now
1.72 moko 634: File_read_http_result pa_internal_file_read_http(Request& r, const String& file_spec, bool as_text, HashStringValue *options, bool transcode_text_result) {
1.1 paf 635: File_read_http_result result;
1.20 misha 636: char host[MAX_STRING];
1.66 moko 637: const char *idna_host;
1.1 paf 638: const char* uri;
1.49 moko 639: short port=80;
1.10 misha 640: const char* method="GET";
1.21 misha 641: bool method_is_get=true;
1.1 paf 642: HashStringValue* form=0;
643: int timeout_secs=2;
644: bool fail_on_status_ne_200=true;
1.12 misha 645: bool omit_post_charset=false;
1.1 paf 646: Value* vheaders=0;
1.10 misha 647: Value* vcookies=0;
1.11 misha 648: Value* vbody=0;
1.72 moko 649: Charset* asked_remote_charset=0;
1.58 moko 650: Charset* real_remote_charset=0;
1.1 paf 651: const char* user_cstr=0;
652: const char* password_cstr=0;
1.22 misha 653: const char* encode=0;
654: bool multipart=false;
1.1 paf 655:
656: if(options) {
657: int valid_options=pa_get_valid_file_options_count(*options);
658:
659: if(Value* vmethod=options->get(HTTP_METHOD_NAME)) {
660: valid_options++;
1.21 misha 661: method=vmethod->as_string().change_case(r.charsets.source(), String::CC_UPPER).cstr();
662: method_is_get=strcmp(method, "GET")==0;
1.1 paf 663: }
1.22 misha 664: if(Value* vencode=options->get(HTTP_FORM_ENCTYPE_NAME)) {
665: valid_options++;
666: encode=vencode->as_string().cstr();
667: }
1.1 paf 668: if(Value* vform=options->get(HTTP_FORM_NAME)) {
669: valid_options++;
670: form=vform->get_hash();
671: }
1.11 misha 672: if(vbody=options->get(HTTP_BODY_NAME)) {
1.1 paf 673: valid_options++;
674: }
675: if(Value* vtimeout=options->get(HTTP_TIMEOUT_NAME)) {
676: valid_options++;
677: timeout_secs=vtimeout->as_int();
678: }
1.11 misha 679: if(vheaders=options->get(HTTP_HEADERS_NAME)) {
1.1 paf 680: valid_options++;
681: }
1.11 misha 682: if(vcookies=options->get(HTTP_COOKIES_NAME)) {
1.10 misha 683: valid_options++;
684: }
1.1 paf 685: if(Value* vany_status=options->get(HTTP_ANY_STATUS_NAME)) {
686: valid_options++;
687: fail_on_status_ne_200=!vany_status->as_bool();
1.12 misha 688: }
1.20 misha 689: if(Value* vomit_post_charset=options->get(HTTP_OMIT_POST_CHARSET_NAME)){
1.12 misha 690: valid_options++;
691: omit_post_charset=vomit_post_charset->as_bool();
692: }
1.6 misha 693: if(Value* vcharset_name=options->get(PA_CHARSET_NAME)) {
1.77 moko 694: asked_remote_charset=&pa_charsets.get(vcharset_name->as_string());
1.58 moko 695: }
696: if(Value* vresponse_charset_name=options->get(PA_RESPONSE_CHARSET_NAME)) {
1.61 moko 697: valid_options++;
1.77 moko 698: real_remote_charset=&pa_charsets.get(vresponse_charset_name->as_string());
1.1 paf 699: }
700: if(Value* vuser=options->get(HTTP_USER)) {
701: valid_options++;
702: user_cstr=vuser->as_string().cstr();
703: }
704: if(Value* vpassword=options->get(HTTP_PASSWORD)) {
705: valid_options++;
706: password_cstr=vpassword->as_string().cstr();
707: }
708:
709: if(valid_options!=options->count())
1.46 misha 710: throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION);
1.1 paf 711: }
712: if(!asked_remote_charset) // defaulting to $request:charset
1.22 misha 713: asked_remote_charset=&(r.charsets).source();
714:
715: if(encode){
716: if(method_is_get)
1.72 moko 717: throw Exception(PARSER_RUNTIME, 0, "you can not use $." HTTP_FORM_ENCTYPE_NAME " option with method GET");
1.22 misha 718:
719: multipart=strcasecmp(encode, HTTP_CONTENT_TYPE_MULTIPART_FORMDATA)==0;
720:
721: if(!multipart && strcasecmp(encode, HTTP_CONTENT_TYPE_FORM_URLENCODED)!=0)
1.72 moko 722: throw Exception(PARSER_RUNTIME, 0, "$." HTTP_FORM_ENCTYPE_NAME " option value can be " HTTP_CONTENT_TYPE_FORM_URLENCODED " or " HTTP_CONTENT_TYPE_MULTIPART_FORMDATA " only");
1.22 misha 723: }
1.1 paf 724:
1.11 misha 725: if(vbody){
726: if(method_is_get)
1.72 moko 727: throw Exception(PARSER_RUNTIME, 0, "you can not use $." HTTP_BODY_NAME " option with method GET");
1.11 misha 728:
729: if(form)
1.72 moko 730: throw Exception(PARSER_RUNTIME, 0, "you can not use options $." HTTP_BODY_NAME " and $." HTTP_FORM_NAME " together");
1.11 misha 731: }
1.1 paf 732:
733: //preparing request
1.29 misha 734: String& connect_string=*new String(file_spec);
1.1 paf 735:
1.48 moko 736: const char* request;
737: size_t request_size;
1.1 paf 738: {
739: // influence URLencoding of tainted pieces to String::L_URI lang
1.22 misha 740: Temp_client_charset temp(r.charsets, *asked_remote_charset);
1.1 paf 741:
1.44 misha 742: const char* connect_string_cstr=connect_string.untaint_and_transcode_cstr(String::L_URI, &(r.charsets));
1.1 paf 743:
744: const char* current=connect_string_cstr;
745: if(strncmp(current, "http://", 7)!=0)
1.72 moko 746: throw Exception(PARSER_RUNTIME, &connect_string, "does not start with http://"); //never
1.1 paf 747: current+=7;
748:
749: strncpy(host, current, sizeof(host)-1); host[sizeof(host)-1]=0;
1.34 misha 750: char* host_uri=lsplit(host, '/');
751: uri=host_uri?current+(host_uri-1-host):"/";
752: char* port_cstr=lsplit(host, ':');
1.49 moko 753:
754: if (port_cstr){
755: char* error_pos=0;
756: port=(short)strtol(port_cstr, &error_pos, 10);
757: if(port==0 || *error_pos)
758: throw Exception(PARSER_RUNTIME, &connect_string, "invalid port number '%s'", port_cstr);
759: }
1.1 paf 760:
1.66 moko 761: idna_host=pa_idna_encode(host, r.charsets.source());
762:
1.11 misha 763: // making request head
1.1 paf 764: String head;
1.11 misha 765: head << method << " " << uri;
1.28 misha 766: if(method_is_get && form)
767: head << (strchr(uri, '?')!=0?"&":"?") << pa_form2string(*form, r.charsets);
1.11 misha 768:
1.66 moko 769: head <<" HTTP/1.0" CRLF "Host: "<< idna_host;
1.49 moko 770: if (port != 80)
771: head << ":" << port_cstr;
772: head << CRLF;
1.11 misha 773:
1.71 moko 774: char* boundary= multipart ? get_uuid_boundary() : 0;
1.22 misha 775:
1.35 misha 776: String user_headers;
777: bool user_agent_specified=false;
778: bool content_type_specified=false;
779: bool content_type_url_encoded=false;
780: if(vheaders && !vheaders->is_string()) { // allow empty
781: if(HashStringValue *headers=vheaders->get_hash()) {
782: Http_pass_header_info info={
783: &(r.charsets),
784: &user_headers,
785: &user_agent_specified,
786: &content_type_specified,
787: &content_type_url_encoded};
788: headers->for_each<Http_pass_header_info*>(http_pass_header, &info);
789: } else
1.72 moko 790: throw Exception(PARSER_RUNTIME, 0, "headers param must be hash");
1.35 misha 791: };
792:
1.48 moko 793: const char* request_body=0;
1.22 misha 794: size_t post_size=0;
795: if(form && !method_is_get) {
1.38 misha 796: head << "Content-Type: " << (multipart ? HTTP_CONTENT_TYPE_MULTIPART_FORMDATA : HTTP_CONTENT_TYPE_FORM_URLENCODED);
1.28 misha 797:
798: if(!omit_post_charset)
799: head << "; charset=" << asked_remote_charset->NAME_CSTR();
800:
1.22 misha 801: if(multipart) {
1.28 misha 802: head << "; boundary=" << boundary;
1.48 moko 803: request_body=pa_form2string_multipart(*form, r/*charsets & mime_type needed*/, boundary, post_size/*correct post_size returned here*/);
1.22 misha 804: } else {
1.48 moko 805: request_body=pa_form2string(*form, r.charsets);
806: post_size=strlen(request_body);
1.22 misha 807: }
1.28 misha 808: head << CRLF;
1.35 misha 809: } else if(vbody) {
1.38 misha 810: // $.body was specified
1.35 misha 811: if(content_type_url_encoded){
1.36 misha 812: // transcode + url-encode
1.48 moko 813: request_body=vbody->as_string().untaint_and_transcode_cstr(String::L_URI, &(r.charsets));
1.35 misha 814: } else {
1.36 misha 815: // content-type != application/x-www-form-urlencoded -> transcode only, don't url-encode!
1.72 moko 816: const String &sbody=vbody->as_string();
817: request_body=Charset::transcode(String::C(sbody.cstr(), sbody.length()), r.charsets.source(), *asked_remote_charset).str;
1.35 misha 818: }
1.48 moko 819: post_size=strlen(request_body);
1.1 paf 820: }
821:
822: // http://www.ietf.org/rfc/rfc2617.txt
823: if(const String* authorization_field_value=basic_authorization_field(user_cstr, password_cstr))
1.38 misha 824: head << "Authorization: " << *authorization_field_value << CRLF;
1.1 paf 825:
1.35 misha 826: head << user_headers;
827:
1.1 paf 828: if(!user_agent_specified) // defaulting
1.38 misha 829: head << "User-Agent: " DEFAULT_USER_AGENT CRLF;
1.1 paf 830:
1.12 misha 831: if(form && !method_is_get && content_type_specified) // POST + form + content-type was specified
1.72 moko 832: throw Exception(PARSER_RUNTIME, 0, "$.content-type can't be specified with method POST");
1.12 misha 833:
1.11 misha 834: if(vcookies && !vcookies->is_string()){ // allow empty
1.10 misha 835: if(HashStringValue* cookies=vcookies->get_hash()) {
1.37 misha 836: head << "Cookie: ";
1.35 misha 837: Http_pass_header_info info={&(r.charsets), &head, 0, 0, 0};
1.10 misha 838: cookies->for_each<Http_pass_header_info*>(http_pass_cookie, &info);
839: head << CRLF;
840: } else
1.72 moko 841: throw Exception(PARSER_RUNTIME, 0, "cookies param must be hash");
1.10 misha 842: }
843:
1.48 moko 844: if(request_body)
1.38 misha 845: head << "Content-Length: " << format(post_size, "%u") << CRLF;
1.48 moko 846:
847: head << CRLF;
848:
849: const char *request_head=head.untaint_and_transcode_cstr(String::L_URI, &(r.charsets));
1.1 paf 850:
1.48 moko 851: if(request_body){
852: size_t head_size = strlen(request_head);
853: request_size=post_size + head_size;
854: char *ptr=(char *)pa_malloc_atomic(request_size);
855: memcpy(ptr, request_head, head_size);
856: memcpy(ptr+head_size, request_body, post_size);
857: request=ptr;
858: } else {
859: request_size=strlen(request_head);
860: request=request_head;
861: }
1.1 paf 862: }
863:
1.78 moko 864:
865: HTTP_response response(connect_string);
1.22 misha 866:
1.28 misha 867: // sending request
1.78 moko 868: int status_code=http_request(response, idna_host, port, request, request_size, timeout_secs, fail_on_status_ne_200);
869:
1.72 moko 870: // processing results
1.78 moko 871: char* raw_body=response.buf + response.body_offset;
872: size_t raw_body_size=response.length - response.body_offset;
873:
1.1 paf 874: result.headers=new HashStringValue;
875: VHash* vtables=new VHash;
1.72 moko 876: result.headers->put("tables", vtables);
877:
1.78 moko 878: if (!real_remote_charset && !response.headers.content_type.is_empty())
879: real_remote_charset=detect_charset(response.headers.content_type.cstr());
1.1 paf 880:
1.72 moko 881: if(as_text)
1.77 moko 882: real_remote_charset=pa_charsets.checkBOM(raw_body, raw_body_size, real_remote_charset);
1.72 moko 883:
884: if (!real_remote_charset)
885: real_remote_charset=asked_remote_charset; // never null
886:
1.85 moko 887: for(Array_iterator<HTTP_Headers::Header> i(response.headers.headers); i.has_next(); ){
888: HTTP_Headers::Header header=i.next();
1.72 moko 889:
890: header.transcode(*real_remote_charset, r.charsets.source());
891:
892: String &header_value=*new String(header.value, String::L_TAINTED);
893:
894: tables_update(vtables->hash(), header.name, header_value);
895: result.headers->put(header.name, new VString(header_value));
1.16 misha 896: }
897:
1.72 moko 898: // filling $.cookies
899: if(Value *vcookies=vtables->hash().get("SET-COOKIE"))
900: result.headers->put(HTTP_COOKIES_NAME, new VTable(parse_cookies(r, vcookies->get_table())));
901:
1.1 paf 902: // output response
903: String::C real_body=String::C(raw_body, raw_body_size);
1.16 misha 904:
905: if(as_text && transcode_text_result && raw_body_size) { // raw_body_size must be checked because transcode returns CONST string in case length==0, which contradicts hacking few lines below
1.22 misha 906: real_body=Charset::transcode(real_body, *real_remote_charset, r.charsets.source());
1.1 paf 907: }
908:
909: result.str=const_cast<char *>(real_body.str); // hacking a little
910: result.length=real_body.length;
1.16 misha 911:
1.22 misha 912: if(as_text && result.length)
913: fix_line_breaks(result.str, result.length);
914:
1.1 paf 915: result.headers->put(file_status_name, new VInt(status_code));
1.16 misha 916:
1.1 paf 917: return result;
918: }
1.84 moko 919:
920: /* ********************** httpd *************************** */
921:
922: class HTTPD_request : public HTTP_response {
923: public:
924: const char *method;
925: const char *uri;
926:
927: HTTPD_request() : HTTP_response(String::Empty), method(NULL), uri(NULL){};
928:
929: const char *extract_method(char *method_line){
930: char* uri_start = strchr(method_line, ' ');
931:
932: if(!uri_start || uri_start == method_line)
933: return NULL;
934:
935: char* uri_end=strchr(uri_start+1, ' ');
936:
937: if(!uri_end || uri_end == uri_start+1)
938: return NULL;
939:
940: uri=pa_strdup(uri_start+1, uri_end-uri_start-1);
941: return str_upper(method_line, uri_start-method_line);
942: }
943:
1.87 ! moko 944: void read_header(int);
! 945: size_t read_post(int, char *, size_t);
1.84 moko 946: };
947:
948: enum HTTPD_request_state {
949: HTTPD_METHOD,
950: HTTPD_HEADERS
951: };
952:
953: void HTTPD_request::read_header(int sock) {
954: enum HTTPD_request_state state = HTTPD_METHOD;
955:
956: size_t chunk_size = 0x400*4;
957: resize(chunk_size);
958:
959: while(read(sock, chunk_size)){
960: switch(state){
961: case HTTPD_METHOD: {
962: size_t method_size = first_line();
963: if(!method_size)
964: break;
965:
966: char *method_line = pa_strdup(buf, method_size);
967: method = extract_method(method_line);
968:
969: if(!method || strcmp(method, "GET") && strcmp(method, "HEAD") && strcmp(method, "POST") && strcmp(method, "PUT") && strcmp(method, "DELETE"))
970: throw Exception("httpd.method", new String(method ? method : method_line), "invalid request method");
971: state = HTTPD_HEADERS;
972: }
973:
974: case HTTPD_HEADERS: {
975: if(!body_start())
976: break;
977:
978: parse_headers();
979: return;
980: }
981: }
982: }
983:
984: if(state == HTTPD_METHOD)
985: throw Exception("httpd.request", 0, "bad request from host - no method found (size=%u)", length);
986:
987: if(state == HTTPD_HEADERS){
988: parse_headers();
989: body_offset=length;
990: }
991: }
992:
1.87 ! moko 993: size_t HTTPD_request::read_post(int sock, char *body, size_t max_bytes) {
! 994: size_t total_read = min(length - body_offset, max_bytes);
! 995: memcpy(body, buf, total_read);
! 996:
! 997: while (total_read < max_bytes){
! 998: ssize_t received_size = recv(sock, buf + total_read, max_bytes - total_read, 0);
! 999: if(received_size == 0)
! 1000: return total_read;
! 1001: if(received_size < 0) {
! 1002: if(int no = pa_socks_errno())
! 1003: throw Exception("httpd.timeout", &url, "error receiving request body: %s (%d)", pa_socks_strerr(no), no);
! 1004: return total_read;
! 1005: }
! 1006: total_read += received_size;
! 1007: }
! 1008: return total_read;
! 1009: }
! 1010:
1.84 moko 1011: /* ********************************************************** */
1012:
1.85 moko 1013: Array<HTTP_Headers::Header> &HTTPD_Connection::headers() {
1.84 moko 1014: return request->headers.headers;
1015: }
1016:
1017: const char *HTTPD_Connection::method() {
1018: return request->method;
1019: }
1020:
1021: const char *HTTPD_Connection::uri() {
1022: return request->uri;
1023: }
1024:
1025: const char *HTTPD_Connection::content_type() {
1026: return request->headers.content_type.cstr();
1027: }
1028:
1029: uint64_t HTTPD_Connection::content_length(){
1030: return request->headers.content_length;
1031: }
1032:
1033: void HTTPD_Connection::read_header(){
1034: request = new HTTPD_request();
1035: request->read_header(sock);
1036: }
1037:
1.87 ! moko 1038: size_t HTTPD_Connection::read_post(char *body, size_t max_bytes) {
! 1039: return request->read_post(sock, body, max_bytes);
! 1040: }
! 1041:
1.84 moko 1042: static int sock_on = 1;
1043:
1.86 moko 1044: int HTTPD_Server::bind(const char *host_port){
1.84 moko 1045: struct sockaddr_in me;
1046:
1.86 moko 1047: const char *port = strchr(host_port, ':');
1048: const char *host = NULL;
1049: if(port && port > host_port){
1050: host = pa_strdup(host_port, port - host_port);
1051: port += 1;
1052: } else {
1053: port = host_port;
1054: }
1055:
1056: if(!set_addr(&me, host, pa_atoui(port, 10))){
1.84 moko 1057: if (host)
1058: throw Exception("httpd.bind", 0, "can not resolve hostname \"%s\"", host);
1059: me.sin_addr.s_addr=INADDR_ANY;
1060: }
1061:
1062: int sock = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/);
1063:
1064: if(sock < 0){
1065: int no=pa_socks_errno();
1066: throw Exception("httpd.bind", 0, "can not make socket: %s (%d)", pa_socks_strerr(no), no);
1067: }
1068:
1069: if (setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (char *)&sock_on, sizeof(sock_on)) ||
1070: setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE, (char *)&sock_on, sizeof(sock_on)) ||
1071: ::bind(sock, (struct sockaddr*)&me, sizeof(me)) ||
1072: listen(sock, 16)) {
1073: close(sock);
1074: int no = pa_socks_errno();
1075: throw Exception("httpd.bind", 0, "can not bind socket: %s (%d)", pa_socks_strerr(no), no);
1076: }
1077: return sock;
1078: }
1079:
1080: int ready(int fd,int operation,int timeout_value){
1081: struct timeval timeout = {0, timeout_value * 1000};
1082: fd_set fds;
1083: FD_ZERO(&fds);
1084: FD_SET(fd, &fds);
1085: switch (operation){
1086: case 0: return select(fd + 1, &fds, NULL, NULL, &timeout)>0; /* read */
1087: case 1: return select(fd + 1, NULL, &fds, NULL, &timeout)>0; /* write */
1088: default: return select(fd + 1, &fds, &fds, NULL, &timeout)>0; /* both */
1089: }
1090: }
1091:
1092: HTTPD_Connection *HTTPD_Server::accept(int sock, int timeout_value) {
1093: int ready = ::ready(sock, 0, timeout_value);
1094: if (ready < 0) {
1095: int no=pa_socks_errno();
1096: if(no == EINTR)
1097: return NULL;
1098: throw Exception("httpd.accept", 0, "error waiting for connection: %s (%d)", pa_socks_strerr(no), no);
1099: }
1100: if (ready == 0) {
1101: /* Timeout */
1102: return NULL;
1103: }
1104:
1105: struct sockaddr_in addr;
1106: unsigned int sock_addr_len = sizeof(struct sockaddr_in);
1107: memset(&addr, 0, sock_addr_len);
1108:
1109: int csock = ::accept(sock, (struct sockaddr *)&addr, &sock_addr_len);
1110: if(csock == -1){
1111: int no=pa_socks_errno();
1112: throw Exception("httpd.accept", 0, "error accepting connection: %s (%d)", pa_socks_strerr(no), no);
1113: }
1114:
1115: return new HTTPD_Connection(csock, pa_strdup(inet_ntoa(addr.sin_addr)));
1116: }
1117:
E-mail:
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to pa_http.C CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [parser3project] / parser3 / src / main