--- parser3/src/main/pa_http.C 2020/10/11 23:25:10 1.86 +++ parser3/src/main/pa_http.C 2024/09/07 16:30:26 1.125 @@ -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) 2001-2023 Art. Lebedev Studio (http://www.artlebedev.com) + Authors: Konstantin Morshnev , Alexandr Petrosian */ #include "pa_http.h" @@ -14,13 +14,7 @@ #include "pa_vfile.h" #include "pa_random.h" -volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.86 2020/10/11 23:25:10 moko Exp $" IDENT_PA_HTTP_H; - -#ifdef _MSC_VER -#include -#else -#define closesocket close -#endif +volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.125 2024/09/07 16:30:26 moko Exp $" IDENT_PA_HTTP_H; // defines @@ -58,7 +52,7 @@ bool HTTP_Headers::add_header(const char content_type=header.value; if(header.name == String::Body("CONTENT-LENGTH") && content_length==0) - content_length=pa_atoul(header.value.cstr(), 10); + ALTER_EXCEPTION_COMMENT(content_length=pa_atoul(header.value.cstr()), " for content-length"); headers+=header; @@ -104,24 +98,23 @@ 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; @@ -130,11 +123,11 @@ public: } size_t first_line(){ - char *headers=strchr(buf, '\n'); - if(!headers) + char *header=strchr(buf, '\n'); + if(!header) return false; - return headers-buf; + return header-buf; } const char *status_code(char *status_line, int &result){ @@ -152,7 +145,7 @@ public: return status_line; const char *result_str=pa_strdup(status_start, status_end-status_start); - result=pa_atoui(result_str, 10); + ALTER_EXCEPTION_COMMENT(result=pa_atoui(result_str), " for HTTP status"); return result_str; } @@ -180,15 +173,16 @@ 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(SOCKET sock, bool fail_on_status_ne_200); }; enum HTTP_response_state { @@ -197,21 +191,21 @@ enum HTTP_response_state { HTTP_BODY }; -static int http_read_response(HTTP_response& 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; size_t chunk_size=0x400*16; - response.resize(2*chunk_size); + resize(2*chunk_size); - while(response.read(sock, chunk_size)){ + while(read(sock, chunk_size)){ switch(state){ case HTTP_STATUS_CODE: { - size_t status_size=response.first_line(); + size_t status_size=first_line(); if(!status_size) break; - const char *status=response.status_code(pa_strdup(response.buf, status_size), result); + const char *status=status_code(pa_strdup(buf, status_size), result); if(!result || fail_on_status_ne_200 && result!=200) throw Exception("http.status", status ? new String(status) : &String::Empty, "invalid HTTP response status"); @@ -220,14 +214,14 @@ static int http_read_response(HTTP_respo } case HTTP_HEADERS: { - if(!response.body_start()) + if(!body_start()) break; - response.parse_headers(); + parse_headers(); - size_t content_length=check_file_size(response.headers.content_length, response.url); - if(content_length>0 && (content_length + response.body_offset) > response.length){ - response.resize(content_length + response.body_offset + 0x400*64); + 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); } state=HTTP_BODY; @@ -242,11 +236,11 @@ static int http_read_response(HTTP_respo } if(state==HTTP_STATUS_CODE) - throw Exception("http.response", &response.url, "bad response from host - no status found (size=%u)", response.length); + throw Exception("http.response", 0, "bad response from host - no status found (size=%u)", length); if(state==HTTP_HEADERS){ - response.parse_headers(); - response.body_offset=response.length; + parse_headers(); + body_offset=length; } return result; @@ -261,31 +255,34 @@ static int http_read_response(HTTP_respo #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; @@ -293,7 +290,7 @@ static int http_request(HTTP_response& r if(!set_addr(&dest, host, port)) throw Exception("http.host", 0, "can not 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); } @@ -322,23 +319,17 @@ static int http_request(HTTP_response& r throw Exception("http.timeout", 0, "error sending request: %s (%d)", pa_socks_strerr(no), no); } - result=http_read_response(response, sock, fail_on_status_ne_200); + 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 @@ -363,9 +354,7 @@ 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(); @@ -562,7 +551,8 @@ 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); @@ -605,7 +595,7 @@ 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; @@ -746,7 +736,7 @@ 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, ':'); @@ -862,10 +852,11 @@ File_read_http_result pa_internal_file_r } - HTTP_response response(connect_string); + HTTP_response response; // sending request - int status_code=http_request(response, idna_host, port, request, request_size, timeout_secs, fail_on_status_ne_200); + int status_code; + ALTER_EXCEPTION_SOURCE(status_code=http_request(response, idna_host, port, request, request_size, timeout_secs, fail_on_status_ne_200), &connect_string); // processing results char* raw_body=response.buf + response.body_offset; @@ -884,7 +875,7 @@ 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()); @@ -896,7 +887,7 @@ File_read_http_result pa_internal_file_r } // filling $.cookies - if(Value *vcookies=vtables->hash().get("SET-COOKIE")) + if(vcookies=vtables->hash().get("SET-COOKIE")) result.headers->put(HTTP_COOKIES_NAME, new VTable(parse_cookies(r, vcookies->get_table()))); // output response @@ -919,12 +910,98 @@ File_read_http_result pa_internal_file_r /* ********************** httpd *************************** */ +#ifdef HTTPD_DEBUG +void pa_log(const char* fmt, ...); +#define LOG(action) action +#else +#define LOG(action) +#endif + +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, ' '); @@ -938,10 +1015,15 @@ 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 sock); + + bool read_header(SOCKET); + size_t read_post(SOCKET, char *, size_t); }; enum HTTPD_request_state { @@ -949,7 +1031,43 @@ 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; @@ -965,7 +1083,7 @@ 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")) + if(!valid_http_method(method)) throw Exception("httpd.method", new String(method ? method : method_line), "invalid request method"); state = HTTPD_HEADERS; } @@ -975,11 +1093,16 @@ 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); @@ -987,6 +1110,26 @@ void HTTPD_request::read_header(int sock parse_headers(); body_offset=length; } + + return true; +} + +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 + body_offset, total_read); + + while (total_read < max_bytes){ + 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.read", new String(uri), "error receiving request body: %s (%d)", pa_socks_strerr(no), no); + return total_read; + } + total_read += received_size; + } + return total_read; } /* ********************************************************** */ @@ -1011,86 +1154,130 @@ 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; } -static int sock_on = 1; - -int HTTPD_Server::bind(const char *host_port){ - struct sockaddr_in me; - - const char *port = strchr(host_port, ':'); - const char *host = NULL; - if(port && 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, 10))){ - if (host) - throw Exception("httpd.bind", 0, "can not resolve hostname \"%s\"", host); - me.sin_addr.s_addr=INADDR_ANY; - } - - int sock = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/); +size_t HTTPD_Connection::read_post(char *body, size_t max_bytes) { + return request->read_post(sock, body, max_bytes); +} - if(sock < 0){ +size_t HTTPD_Connection::send_body(const void *buf, size_t size) { + LOG(pa_log("httpd [%d] response %d bytes", sock, size)); + LOG(pa_log("httpd [%d] %s", sock, buf)); + if(send(sock, (const char*)buf, size, 0) != (ssize_t)size) { int no=pa_socks_errno(); - throw Exception("httpd.bind", 0, "can not make socket: %s (%d)", pa_socks_strerr(no), no); + throw Exception("httpd.write", 0, "error sending response: %s (%d)", pa_socks_strerr(no), no); } + return size; +} - if (setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (char *)&sock_on, sizeof(sock_on)) || - setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE, (char *)&sock_on, sizeof(sock_on)) || - ::bind(sock, (struct sockaddr*)&me, sizeof(me)) || - listen(sock, 16)) { - close(sock); - int no = pa_socks_errno(); - throw Exception("httpd.bind", 0, "can not bind socket: %s (%d)", pa_socks_strerr(no), no); +HTTPD_Connection::~HTTPD_Connection(){ + if(sock != INVALID_SOCKET){ + LOG(pa_log("httpd [%d] closed", sock)); + closesocket(sock); } - return sock; } -int 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 */ } } - -HTTPD_Connection *HTTPD_Server::accept(int sock, int timeout_value) { - - int ready = ::ready(sock, 0, 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(); if(no == EINTR) - return NULL; + return false; throw Exception("httpd.accept", 0, "error waiting for connection: %s (%d)", pa_socks_strerr(no), no); } - if (ready == 0) { - /* Timeout */ - return NULL; - } + if (ready == 0) + return false; /* Timeout */ struct sockaddr_in addr; - unsigned int sock_addr_len = sizeof(struct sockaddr_in); + socklen_t sock_addr_len = sizeof(struct sockaddr_in); memset(&addr, 0, sock_addr_len); - int csock = ::accept(sock, (struct sockaddr *)&addr, &sock_addr_len); - if(csock == -1){ + sock = ::accept(server_sock, (struct sockaddr *)&addr, &sock_addr_len); + if(sock == INVALID_SOCKET){ int no=pa_socks_errno(); throw Exception("httpd.accept", 0, "error accepting connection: %s (%d)", pa_socks_strerr(no), no); } - return new HTTPD_Connection(csock, pa_strdup(inet_ntoa(addr.sin_addr))); +// 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; +} + +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; + + port = strchr(host_port, ':'); + const char *host = NULL; + 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, (short)pa_atoui(port))){ + if (host) + throw Exception("httpd.bind", 0, "can not resolve hostname \"%s\"", host); + me.sin_addr.s_addr=INADDR_ANY; + } + + SOCKET sock = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*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); + } + + static int sock_on = 1; + + if (setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (char *)&sock_on, sizeof(sock_on)) || + setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE, (char *)&sock_on, sizeof(sock_on)) || + ::bind(sock, (struct sockaddr*)&me, sizeof(me)) || + 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); + } + return sock; +}