--- parser3/src/main/pa_http.C 2020/10/14 21:22:59 1.96 +++ parser3/src/main/pa_http.C 2024/11/23 00:36:40 1.130 @@ -1,8 +1,8 @@ /** @file Parser: http support functions. - Copyright (c) 2001-2017 Art. Lebedev Studio (http://www.artlebedev.com) - Author: Alexandr Petrosian (http://paf.design.ru) + Copyright (c) 2001-2024 Art. Lebedev Studio (http://www.artlebedev.com) + Authors: Konstantin Morshnev , Alexandr Petrosian */ #include "pa_http.h" @@ -14,14 +14,7 @@ #include "pa_vfile.h" #include "pa_random.h" -volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.96 2020/10/14 21:22:59 moko Exp $" IDENT_PA_HTTP_H; - -#ifdef _MSC_VER -#include -#define socklen_t int -#else -#define closesocket close -#endif +volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.130 2024/11/23 00:36:40 moko Exp $" IDENT_PA_HTTP_H; // defines @@ -49,6 +42,32 @@ volatile const char * IDENT_PA_HTTP_C="$ // helpers +#ifdef HTTPD_DEBUG +void pa_log(const char* fmt, ...); +#define LOG(action) action +#else +#define LOG(action) +#endif + +ssize_t pa_send(int sock, const char *buffer, size_t len){ + size_t total_sent = 0; + while (total_sent < len) { + ssize_t bytes_sent=send(sock, buffer + total_sent, len - total_sent, 0); + if (bytes_sent < 0) { + return bytes_sent; + } else if (bytes_sent == 0) { + // Connection closed by the remote peer? + break; + } + LOG( + if(bytes_sent != len - total_sent) + pa_log("httpd [%d] partial send %d of (%d)", sock, bytes_sent, len - total_sent) + ); + total_sent += bytes_sent; + } + return total_sent; +} + bool HTTP_Headers::add_header(const char *line){ const char *value=strchr(line, ':'); @@ -105,24 +124,23 @@ public: size_t body_offset; HTTP_Headers headers; - const String &url; - HTTP_response(const String& aurl) : buf(NULL), length(0), buf_size(0), body_offset(0), url(aurl){} + HTTP_response() : buf(NULL), length(0), buf_size(0), body_offset(0){} void resize(size_t size){ buf_size=size; buf=(char *)pa_realloc(buf, size + 1); } - bool read(int sock, size_t size){ - if(length+size>buf_size) - resize(buf_size*2 + size); + bool read(SOCKET sock, size_t size){ + if(length + size > buf_size) + resize(buf_size * 2 + size); ssize_t received_size=recv(sock, buf + length, size, 0); - if(received_size==0) + if(received_size == 0) return false; - if(received_size<0) { - if(int no=pa_socks_errno()) - throw Exception("http.timeout", &url, "error receiving response body: %s (%d)", pa_socks_strerr(no), no); + if(received_size < 0) { + if(int no = pa_socks_errno()) + throw Exception("http.timeout", 0, "error receiving response: %s (%d)", pa_socks_strerr(no), no); return false; } length+=received_size; @@ -181,16 +199,16 @@ public: ArrayString aheaders; header_block.split(aheaders, 0, "\n"); - Array_iterator i(aheaders); + ArrayString::Iterator i(aheaders); i.next(); // skipping status - for(;i.has_next();){ + for(;i;){ const char *line=i.next()->cstr(); if(!headers.add_header(line)) - throw Exception("http.response", &url, "bad response from host - bad header \"%s\"", line); + throw Exception("http.response", 0, "bad response from host - bad header \"%s\"", line); } } - int read_response(int sock, bool fail_on_status_ne_200); + int read_response(SOCKET sock, bool fail_on_status_ne_200); }; enum HTTP_response_state { @@ -199,7 +217,7 @@ enum HTTP_response_state { HTTP_BODY }; -int HTTP_response::read_response(int sock, bool fail_on_status_ne_200) { +int HTTP_response::read_response(SOCKET sock, bool fail_on_status_ne_200) { HTTP_response_state state=HTTP_STATUS_CODE; int result=0; @@ -227,7 +245,7 @@ int HTTP_response::read_response(int soc parse_headers(); - size_t content_length=check_file_size(headers.content_length, &url); + size_t content_length=check_file_size(headers.content_length, 0); if(content_length>0 && (content_length + body_offset) > length){ resize(content_length + body_offset + 0x400*64); } @@ -244,7 +262,7 @@ int HTTP_response::read_response(int soc } if(state==HTTP_STATUS_CODE) - throw Exception("http.response", &url, "bad response from host - no status found (size=%u)", length); + throw Exception("http.response", 0, "bad response from host - no status found (size=%u)", length); if(state==HTTP_HEADERS){ parse_headers(); @@ -263,41 +281,44 @@ int HTTP_response::read_response(int soc #ifdef PA_USE_ALARM static sigjmp_buf timeout_env; static void timeout_handler(int /*sig*/){ - siglongjmp(timeout_env, 1); + siglongjmp(timeout_env, 1); } + +#define PA_NO_THREADS (HTTPD_Server::mode != HTTPD_Server::MULTITHREADED) + +#define ALARM(value) if(PA_NO_THREADS) alarm(value) +#else +#define ALARM(value) #endif 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) { if(!host) throw Exception("http.host", 0, "zero hostname"); //never - volatile // to prevent makeing it register variable, because it will be clobbered by longjmp [thanks gcc warning] - int sock=-1; -#ifdef PA_USE_ALARM - signal(SIGALRM, timeout_handler); -#endif + volatile SOCKET sock=INVALID_SOCKET; // to prevent makeing it register variable, because it will be clobbered by longjmp [thanks gcc warning] + #ifdef PA_USE_ALARM - if(sigsetjmp(timeout_env, 1)) { - // stupid gcc [2.95.4] generated bad code - // which failed to handle sigsetjmp+throw: crashed inside of pre-throw code. - // rewritten simplier [athough duplicating closesocket code] - if(sock>=0) - closesocket(sock); - throw Exception("http.timeout", 0, "timeout occurred while retrieving document"); + if(PA_NO_THREADS) signal(SIGALRM, timeout_handler); + if(PA_NO_THREADS && sigsetjmp(timeout_env, 1)) { + // duplicating closesocket to make code more simple for old compilers + if(sock != INVALID_SOCKET) + closesocket(sock); + throw Exception("http.timeout", 0, "timeout occurred while retrieving document"); return 0; // never - } else { - alarm(timeout_secs); + } else #endif + { + ALARM(timeout_secs); try { int result; struct sockaddr_in dest; if(!set_addr(&dest, host, port)) - throw Exception("http.host", 0, "can not resolve hostname \"%s\"", host); + throw Exception("http.host", 0, "cannot resolve hostname \"%s\"", host); - if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) { + if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/)) == INVALID_SOCKET) { int no=pa_socks_errno(); - throw Exception("http.connect", 0, "can not make socket: %s (%d)", pa_socks_strerr(no), no); + throw Exception("http.connect", 0, "cannot make socket: %s (%d)", pa_socks_strerr(no), no); } // To enable SO_DONTLINGER (that is, disable SO_LINGER) @@ -316,31 +337,25 @@ static int http_request(HTTP_response& r if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) { int no=pa_socks_errno(); - throw Exception("http.connect", 0, "can not connect to host \"%s\": %s (%d)", host, pa_socks_strerr(no), no); + throw Exception("http.connect", 0, "cannot connect to host \"%s\": %s (%d)", host, pa_socks_strerr(no), no); } - if(send(sock, request, request_size, 0)!=(ssize_t)request_size) { + if(pa_send(sock, request, request_size) < 0) { int no=pa_socks_errno(); throw Exception("http.timeout", 0, "error sending request: %s (%d)", pa_socks_strerr(no), no); } result=response.read_response(sock, fail_on_status_ne_200); closesocket(sock); -#ifdef PA_USE_ALARM - alarm(0); -#endif + ALARM(0); return result; } catch(...) { -#ifdef PA_USE_ALARM - alarm(0); -#endif - if(sock>=0) + ALARM(0); + if(sock != INVALID_SOCKET) closesocket(sock); rethrow; } -#ifdef PA_USE_ALARM } -#endif } #ifndef DOXYGEN @@ -365,9 +380,7 @@ char *pa_http_safe_header_name(const cha return result; } -static void http_pass_header(HashStringValue::key_type aname, - HashStringValue::value_type avalue, - Http_pass_header_info *info) { +static void http_pass_header(HashStringValue::key_type aname, HashStringValue::value_type avalue, Http_pass_header_info *info) { const char* name_cstr=aname.cstr(); @@ -524,7 +537,7 @@ static void form_value2part(HashStringVa Form_table_value2string_info info(key, *part.string); part.info = &info; tvalue->for_each(form_table_value2part, &part); - } else if(VFile* vfile=static_cast(value->as("file"))){ + } else if(VFile* vfile=dynamic_cast(value)){ form_file_value2part(key, *vfile, part); } else 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()); @@ -564,7 +577,8 @@ static ArrayString* parse_cookie(Request if(first_pair) { // name + value name=sname; - value=smeaning; + if(smeaning) + value=smeaning; first_pair=false; } else { const String& slower=sname->change_case(r.charsets.source(), String::CC_LOWER); @@ -607,7 +621,7 @@ static ArrayString* parse_cookie(Request Table* parse_cookies(Request& r, Table *cookies){ Table& result=*new Table(new Cookies_table_template_columns); - for(Array_iterator i(*cookies); i.has_next(); ) + for(Array_iterator i(*cookies); i; ) if(ArrayString* row=parse_cookie(r, *i.next()->get(0))) result+=row; @@ -716,7 +730,7 @@ File_read_http_result pa_internal_file_r if(encode){ if(method_is_get) - throw Exception(PARSER_RUNTIME, 0, "you can not use $." HTTP_FORM_ENCTYPE_NAME " option with method GET"); + throw Exception(PARSER_RUNTIME, 0, "you cannot use $." HTTP_FORM_ENCTYPE_NAME " option with method GET"); multipart=strcasecmp(encode, HTTP_CONTENT_TYPE_MULTIPART_FORMDATA)==0; @@ -726,10 +740,10 @@ File_read_http_result pa_internal_file_r if(vbody){ if(method_is_get) - throw Exception(PARSER_RUNTIME, 0, "you can not use $." HTTP_BODY_NAME " option with method GET"); + throw Exception(PARSER_RUNTIME, 0, "you cannot use $." HTTP_BODY_NAME " option with method GET"); if(form) - throw Exception(PARSER_RUNTIME, 0, "you can not use options $." HTTP_BODY_NAME " and $." HTTP_FORM_NAME " together"); + throw Exception(PARSER_RUNTIME, 0, "you cannot use options $." HTTP_BODY_NAME " and $." HTTP_FORM_NAME " together"); } //preparing request @@ -748,7 +762,7 @@ File_read_http_result pa_internal_file_r throw Exception(PARSER_RUNTIME, &connect_string, "does not start with http://"); //never current+=7; - strncpy(host, current, sizeof(host)-1); host[sizeof(host)-1]=0; + pa_strncpy(host, current, sizeof(host)); char* host_uri=lsplit(host, '/'); uri=host_uri?current+(host_uri-1-host):"/"; char* port_cstr=lsplit(host, ':'); @@ -844,7 +858,7 @@ File_read_http_result pa_internal_file_r } if(request_body) - head << "Content-Length: " << format(post_size, "%u") << CRLF; + head << "Content-Length: " << pa_uitoa(post_size) << CRLF; head << CRLF; @@ -864,7 +878,7 @@ File_read_http_result pa_internal_file_r } - HTTP_response response(connect_string); + HTTP_response response; // sending request int status_code; @@ -887,7 +901,7 @@ File_read_http_result pa_internal_file_r if (!real_remote_charset) real_remote_charset=asked_remote_charset; // never null - for(Array_iterator i(response.headers.headers); i.has_next(); ){ + for(Array_iterator i(response.headers.headers); i; ){ HTTP_Headers::Header header=i.next(); header.transcode(*real_remote_charset, r.charsets.source()); @@ -922,12 +936,91 @@ File_read_http_result pa_internal_file_r /* ********************** httpd *************************** */ +enum EscapeState { + Initial, + Default, + EscapeFirst, + EscapeSecond +}; + +static bool check_uri(const char *uri){ + EscapeState state=Initial; + uint escapedValue=0; + + const char *pattern="/../"; + const char *pos=pattern; + + while(*uri){ + uchar c=(uchar)*(uri++); + switch(state) { + case Initial: + if(c!='/') + return false; + state=Default; + break; + case Default: + if(c=='%'){ + state=EscapeFirst; + continue; + } + if(c=='?') + return true; + break; + case EscapeFirst: + if(isxdigit(c)){ + state=EscapeSecond; + escapedValue=hex_value[c] << 4; + continue; + } + return false; + case EscapeSecond: + if(isxdigit(c)){ + state=Default; + c=(uchar)(escapedValue + hex_value[c]); + + // implementing Apache AllowEncodedSlashes Off just in case + if(c=='/' || c=='\\') + return false; + + break; + } + return false; + } + + if(c==*pos || c=='\\' && *pos=='/'){ + if(!*(++pos)) + return false; + } else { + pos=pattern; + } + } + return true; +} + class HTTPD_request : public HTTP_response { public: const char *method; const char *uri; - HTTPD_request() : HTTP_response(String::Empty), method(NULL), uri(NULL){}; + HTTPD_request() : HTTP_response(), method(NULL), uri(NULL){}; + + ssize_t pa_recv(SOCKET sockfd, char *buf, size_t len); + + bool read(SOCKET sock, size_t size){ + if(length + size > buf_size) + resize(buf_size * 2 + size); + ssize_t received_size=pa_recv(sock, buf + length, size); + if(received_size == 0) + return false; + if(received_size < 0) { + if(int no = pa_socks_errno()) + throw Exception("httpd.read", 0, "error receiving request: %s (%d)", pa_socks_strerr(no), no); + return false; + } + length+=received_size; + buf[length]='\0'; + return true; + } const char *extract_method(char *method_line){ char* uri_start = strchr(method_line, ' '); @@ -941,11 +1034,15 @@ public: return NULL; uri=pa_strdup(uri_start+1, uri_end-uri_start-1); + if(!check_uri(uri)) + throw Exception("httpd.request", 0, "invalid uri '%s'", uri); + return str_upper(method_line, uri_start-method_line); } - void read_header(int); - size_t read_post(int, char *, size_t); + + bool read_header(SOCKET); + size_t read_post(SOCKET, char *, size_t); }; enum HTTPD_request_state { @@ -953,7 +1050,43 @@ enum HTTPD_request_state { HTTPD_HEADERS }; -void HTTPD_request::read_header(int sock) { +ssize_t HTTPD_request::pa_recv(SOCKET sockfd, char *buffer, size_t len){ + LOG(pa_log("httpd [%d] recv %d appending to %d ...", sockfd, len, length)); + +#ifdef PA_USE_ALARM + if(PA_NO_THREADS) signal(SIGALRM, timeout_handler); + if(PA_NO_THREADS && sigsetjmp(timeout_env, 1)) { + LOG(pa_log("httpd [%d] recv got %d sec timeout", sockfd, pa_httpd_timeout)); + if(length) // timeout on "void" connection is normal + throw Exception("httpd.timeout", 0, "timeout occurred while receiving request"); + return 0; + } else +#endif + { + ALARM(pa_httpd_timeout); + ssize_t result=recv(sockfd, buffer, len, 0); + ALARM(0); + LOG(pa_log("httpd [%d] recv got %d bytes", sockfd, result)); + LOG(pa_log("httpd [%d] %s", sockfd, buffer)); + return result; + } +} + +static bool valid_http_method(const char * method){ + return method && ( + !strcmp(method, "GET") || + !strcmp(method, "HEAD") || + !strcmp(method, "POST") || + !strcmp(method, "PUT") || + !strcmp(method, "DELETE") || + !strcmp(method, "CONNECT") || + !strcmp(method, "OPTIONS") || + !strcmp(method, "TRACE") || + !strcmp(method, "PATCH") + ); +} + +bool HTTPD_request::read_header(SOCKET sock) { enum HTTPD_request_state state = HTTPD_METHOD; size_t chunk_size = 0x400*4; @@ -969,14 +1102,7 @@ void HTTPD_request::read_header(int sock char *method_line = pa_strdup(buf, method_size); method = extract_method(method_line); - if(!method || - strcmp(method, "GET") && - strcmp(method, "HEAD") && - strcmp(method, "POST") && - strcmp(method, "PUT") && - strcmp(method, "DELETE") && - strcmp(method, "PATCH") - ) + if(!valid_http_method(method)) throw Exception("httpd.method", new String(method ? method : method_line), "invalid request method"); state = HTTPD_HEADERS; } @@ -986,11 +1112,16 @@ void HTTPD_request::read_header(int sock break; parse_headers(); - return; + return true; } } } + if(!length){ // browsers open connections in advance and they will be empty unless user requests more pages + LOG(pa_log("httpd [%d] void request", sock)); + return false; + } + if(state == HTTPD_METHOD) throw Exception("httpd.request", 0, "bad request from host - no method found (size=%u)", length); @@ -998,19 +1129,21 @@ void HTTPD_request::read_header(int sock parse_headers(); body_offset=length; } + + return true; } -size_t HTTPD_request::read_post(int sock, char *body, size_t max_bytes) { +size_t HTTPD_request::read_post(SOCKET sock, char *body, size_t max_bytes) { size_t total_read = min(length - body_offset, max_bytes); - memcpy(body, buf, total_read); + memcpy(body, buf + body_offset, total_read); while (total_read < max_bytes){ - ssize_t received_size = recv(sock, buf + total_read, max_bytes - total_read, 0); + ssize_t received_size = pa_recv(sock, body + total_read, max_bytes - total_read); if(received_size == 0) return total_read; if(received_size < 0) { if(int no = pa_socks_errno()) - throw Exception("httpd.timeout", &url, "error receiving request body: %s (%d)", pa_socks_strerr(no), no); + throw Exception("httpd.read", new String(uri), "error receiving request body: %s (%d)", pa_socks_strerr(no), no); return total_read; } total_read += received_size; @@ -1040,9 +1173,13 @@ uint64_t HTTPD_Connection::content_lengt return request->headers.content_length; } -void HTTPD_Connection::read_header(){ +bool HTTPD_Connection::read_header(){ request = new HTTPD_request(); - request->read_header(sock); + bool result = request->read_header(sock); + LOG(if(result){ + pa_log("httpd [%d] got %s \"%s\"", sock, method(), uri()); + }) + return result; } size_t HTTPD_Connection::read_post(char *body, size_t max_bytes) { @@ -1050,31 +1187,36 @@ size_t HTTPD_Connection::read_post(char } size_t HTTPD_Connection::send_body(const void *buf, size_t size) { - if(send(sock, (const char*)buf, size, 0) != (ssize_t)size) { + LOG(pa_log("httpd [%d] response %d bytes", sock, size)); + LOG(pa_log("httpd [%d] %s", sock, buf)); + if((size=pa_send(sock, (const char*)buf, size)) < 0) { int no=pa_socks_errno(); - throw Exception("httpd.timeout", 0, "error sending response: %s (%d)", pa_socks_strerr(no), no); + throw Exception("httpd.write", 0, "error sending response: %s (%d)", pa_socks_strerr(no), no); } return size; } HTTPD_Connection::~HTTPD_Connection(){ - if(sock != -1) + if(sock != INVALID_SOCKET){ + LOG(pa_log("httpd [%d] closed", sock)); closesocket(sock); + } } -static int sock_ready(int fd,int operation,int timeout_value){ +static int sock_ready(SOCKET fd, int operation, int timeout_value){ struct timeval timeout = {0, timeout_value * 1000}; fd_set fds; FD_ZERO(&fds); FD_SET(fd, &fds); + int nfds = (int)fd + 1; /* typecast as nfds is ignored in MSVC anyway */ switch (operation){ - case 0: return select(fd + 1, &fds, NULL, NULL, &timeout)>0; /* read */ - case 1: return select(fd + 1, NULL, &fds, NULL, &timeout)>0; /* write */ - default: return select(fd + 1, &fds, &fds, NULL, &timeout)>0; /* both */ + case 0: return select(nfds, &fds, NULL, NULL, &timeout)>0; /* read */ + case 1: return select(nfds, NULL, &fds, NULL, &timeout)>0; /* write */ + default: return select(nfds, &fds, &fds, NULL, &timeout)>0; /* both */ } } -bool HTTPD_Connection::accept(int server_sock, int timeout_value) { +bool HTTPD_Connection::accept(SOCKET server_sock, int timeout_value) { int ready = sock_ready(server_sock, 0, timeout_value); if (ready < 0) { int no=pa_socks_errno(); @@ -1090,38 +1232,60 @@ bool HTTPD_Connection::accept(int server memset(&addr, 0, sock_addr_len); sock = ::accept(server_sock, (struct sockaddr *)&addr, &sock_addr_len); - if(server_sock == -1){ + if(sock == INVALID_SOCKET){ int no=pa_socks_errno(); throw Exception("httpd.accept", 0, "error accepting connection: %s (%d)", pa_socks_strerr(no), no); } +// Has no positive performance effect, requires include +// static int sock_on=1; +// setsockopt(sock, IPPROTO_TCP, TCP_NODELAY, (char *)&sock_on, sizeof(sock_on)); + remote_addr = pa_strdup(inet_ntoa(addr.sin_addr)); + LOG(pa_log("httpd [%d] accepted from %s", sock, remote_addr)); return true; } -int HTTPD_Server::bind(const char *host_port){ +HTTPD_Server::HTTPD_MODE HTTPD_Server::mode = HTTPD_Server::SEQUENTIAL; +const char *HTTPD_Server::port=NULL; + +void HTTPD_Server::set_mode(const String &value){ + if(value == "sequental") mode = SEQUENTIAL; +#ifdef HAVE_TLS + else if (value == "threaded") mode = MULTITHREADED; +#endif +#ifdef _MSC_VER + else throw Exception("httpd.mode", &value, "$MAIN:HTTPD.mode must be 'sequental' or 'threaded'"); +#else + else if (value == "parallel") mode = PARALLEL; + else throw Exception("httpd.mode", &value, "$MAIN:HTTPD.mode must be 'sequental', 'parallel' or 'threaded'"); +#endif +} + +SOCKET HTTPD_Server::bind(const char *host_port){ struct sockaddr_in me; - const char *port = strchr(host_port, ':'); + port = strchr(host_port, ':'); const char *host = NULL; - if(port && port > host_port){ - host = pa_strdup(host_port, port - host_port); + if(port){ + if(port > host_port) + host = pa_strdup(host_port, port - host_port); port += 1; } else { port = host_port; } - if(!set_addr(&me, host, pa_atoui(port))){ + if(!set_addr(&me, host, (short)pa_atoui(port))){ if (host) - throw Exception("httpd.bind", 0, "can not resolve hostname \"%s\"", host); + throw Exception("httpd.bind", 0, "cannot resolve hostname \"%s\"", host); me.sin_addr.s_addr=INADDR_ANY; } - int sock = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/); + SOCKET sock = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/); - if(sock < 0){ + if(sock == INVALID_SOCKET){ int no=pa_socks_errno(); - throw Exception("httpd.bind", 0, "can not make socket: %s (%d)", pa_socks_strerr(no), no); + throw Exception("httpd.bind", 0, "cannot make socket: %s (%d)", pa_socks_strerr(no), no); } static int sock_on = 1; @@ -1132,7 +1296,7 @@ int HTTPD_Server::bind(const char *host_ listen(sock, 16)) { closesocket(sock); int no = pa_socks_errno(); - throw Exception("httpd.bind", 0, "can not bind socket: %s (%d)", pa_socks_strerr(no), no); + throw Exception("httpd.bind", 0, "cannot bind socket: %s (%d)", pa_socks_strerr(no), no); } return sock; }