--- parser3/src/main/pa_http.C 2017/05/17 14:22:12 1.80 +++ parser3/src/main/pa_http.C 2020/12/15 17:10:36 1.109 @@ -1,22 +1,24 @@ /** @file Parser: http support functions. - Copyright (c) 2001-2017 Art. Lebedev Studio (http://www.artlebedev.com) + Copyright (c) 2001-2020 Art. Lebedev Studio (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ #include "pa_http.h" #include "pa_common.h" +#include "pa_base64.h" #include "pa_charsets.h" #include "pa_request_charsets.h" #include "pa_request.h" #include "pa_vfile.h" #include "pa_random.h" -volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.80 2017/05/17 14:22:12 moko Exp $" IDENT_PA_HTTP_H; +volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.109 2020/12/15 17:10:36 moko Exp $" IDENT_PA_HTTP_H; #ifdef _MSC_VER #include +#define socklen_t int #else #define closesocket close #endif @@ -47,7 +49,7 @@ volatile const char * IDENT_PA_HTTP_C="$ // helpers -bool ResponseHeaders::add_header(const char *line){ +bool HTTP_Headers::add_header(const char *line){ const char *value=strchr(line, ':'); if(value && value != line){ // we need only headers, not the response code @@ -57,7 +59,7 @@ bool ResponseHeaders::add_header(const c 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; @@ -95,17 +97,16 @@ static bool set_addr(struct sockaddr_in return false; } -class HTTP_response { +class HTTP_response : public PA_Allocated { public: char *buf; size_t length; size_t buf_size; size_t body_offset; - ResponseHeaders headers; - String &url; + HTTP_Headers headers; - HTTP_response(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; @@ -113,14 +114,14 @@ public: } bool read(int sock, size_t size){ - if(length+size>buf_size) - resize(buf_size*2 + 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; @@ -128,12 +129,12 @@ public: return true; } - size_t status_size(){ - char *headers=strchr(buf, '\n'); - if(!headers) + size_t first_line(){ + char *header=strchr(buf, '\n'); + if(!header) return false; - return headers-buf; + return header-buf; } const char *status_code(char *status_line, int &result){ @@ -151,7 +152,7 @@ public: return status_line; const char *result_str=pa_strdup(status_start, status_end-status_start); - result=atoi(result_str); + ALTER_EXCEPTION_COMMENT(result=pa_atoui(result_str), " for HTTP status"); return result_str; } @@ -184,10 +185,11 @@ public: for(;i.has_next();){ 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); }; enum HTTP_response_state { @@ -196,21 +198,21 @@ 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(int 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.status_size(); + 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"); @@ -219,14 +221,14 @@ 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; @@ -241,11 +243,11 @@ 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; @@ -260,31 +262,31 @@ 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 ALARM(value) 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 int sock=-1; // to prevent makeing it register variable, because it will be clobbered by longjmp [thanks gcc warning] + #ifdef PA_USE_ALARM + signal(SIGALRM, timeout_handler); 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"); + // duplicating closesocket to make code more simple for old compilers + if(sock>=0) + 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; @@ -321,23 +323,17 @@ 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 + ALARM(0); if(sock>=0) closesocket(sock); rethrow; } -#ifdef PA_USE_ALARM } -#endif } #ifndef DOXYGEN @@ -362,9 +358,7 @@ 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(); @@ -406,7 +400,7 @@ static const String* basic_authorization combined< i(response.headers.headers); i.has_next(); ){ - ResponseHeaders::Header header=i.next(); + for(Array_iterator i(response.headers.headers); i.has_next(); ){ + HTTP_Headers::Header header=i.next(); header.transcode(*real_remote_charset, r.charsets.source()); @@ -895,7 +890,7 @@ 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 @@ -915,3 +910,335 @@ File_read_http_result pa_internal_file_r return result; } + +/* ********************** httpd *************************** */ + +enum EscapeState { + Initial, + Default, + EscapeFirst, + EscapeSecond +}; + +static bool check_uri(const char *uri){ + EscapeState state=Initial; + uint escapedValue; + + 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(), method(NULL), uri(NULL){}; + + ssize_t pa_recv(int sockfd, char *buf, size_t len); + + bool read(int 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.timeout", 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, ' '); + + if(!uri_start || uri_start == method_line) + return NULL; + + char* uri_end=strchr(uri_start+1, ' '); + + if(!uri_end || uri_end == uri_start+1) + 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); +}; + +enum HTTPD_request_state { + HTTPD_METHOD, + HTTPD_HEADERS +}; + +ssize_t HTTPD_request::pa_recv(int sockfd, char *buffer, size_t len){ + if(HTTPD_Server::mode == HTTPD_Server::MULTITHREADED) + return recv(sockfd, buffer, len, 0); + +#ifdef PA_USE_ALARM + signal(SIGALRM, timeout_handler); + if(sigsetjmp(timeout_env, 1)) { + throw Exception("httpd.timeout", 0, "timeout occurred while receiving request"); + return 0; // never + } else +#endif + { + ALARM(pa_httpd_timeout); + ssize_t result=recv(sockfd, buffer, len, 0); + ALARM(0); + return result; + } +} + +void HTTPD_request::read_header(int sock) { + enum HTTPD_request_state state = HTTPD_METHOD; + + size_t chunk_size = 0x400*4; + resize(chunk_size); + + while(read(sock, chunk_size)){ + switch(state){ + case HTTPD_METHOD: { + size_t method_size = first_line(); + if(!method_size) + break; + + 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") + ) + throw Exception("httpd.method", new String(method ? method : method_line), "invalid request method"); + state = HTTPD_HEADERS; + } + + case HTTPD_HEADERS: { + if(!body_start()) + break; + + parse_headers(); + return; + } + } + } + + if(state == HTTPD_METHOD) + throw Exception("httpd.request", 0, "bad request from host - no method found (size=%u)", length); + + if(state == HTTPD_HEADERS){ + parse_headers(); + body_offset=length; + } +} + +size_t HTTPD_request::read_post(int 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.timeout", new String(uri), "error receiving request body: %s (%d)", pa_socks_strerr(no), no); + return total_read; + } + total_read += received_size; + } + return total_read; +} + +/* ********************************************************** */ + +Array &HTTPD_Connection::headers() { + return request->headers.headers; +} + +const char *HTTPD_Connection::method() { + return request->method; +} + +const char *HTTPD_Connection::uri() { + return request->uri; +} + +const char *HTTPD_Connection::content_type() { + return request->headers.content_type.cstr(); +} + +uint64_t HTTPD_Connection::content_length(){ + return request->headers.content_length; +} + +void HTTPD_Connection::read_header(){ + request = new HTTPD_request(); + request->read_header(sock); +} + +size_t HTTPD_Connection::read_post(char *body, size_t max_bytes) { + return request->read_post(sock, body, max_bytes); +} + +size_t HTTPD_Connection::send_body(const void *buf, size_t size) { + if(send(sock, (const char*)buf, size, 0) != (ssize_t)size) { + int no=pa_socks_errno(); + throw Exception("httpd.timeout", 0, "error sending response: %s (%d)", pa_socks_strerr(no), no); + } + return size; +} + +HTTPD_Connection::~HTTPD_Connection(){ + if(sock != -1) + closesocket(sock); +} + +static int sock_ready(int fd,int operation,int timeout_value){ + struct timeval timeout = {0, timeout_value * 1000}; + fd_set fds; + FD_ZERO(&fds); + FD_SET(fd, &fds); + 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 */ + } +} + +bool HTTPD_Connection::accept(int 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 false; + throw Exception("httpd.accept", 0, "error waiting for connection: %s (%d)", pa_socks_strerr(no), no); + } + if (ready == 0) + return false; /* Timeout */ + + struct sockaddr_in addr; + socklen_t sock_addr_len = sizeof(struct sockaddr_in); + memset(&addr, 0, sock_addr_len); + + sock = ::accept(server_sock, (struct sockaddr *)&addr, &sock_addr_len); + if(server_sock == -1){ + int no=pa_socks_errno(); + throw Exception("httpd.accept", 0, "error accepting connection: %s (%d)", pa_socks_strerr(no), no); + } + + remote_addr = pa_strdup(inet_ntoa(addr.sin_addr)); + return true; +} + +HTTPD_Server::HTTPD_MODE HTTPD_Server::mode = HTTPD_Server::SEQUENTIAL; + +void HTTPD_Server::set_mode(const String &value){ + if(value == "sequental") mode = SEQUENTIAL; + else if (value == "threaded") mode = MULTITHREADED; +#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 +} + +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, (short)pa_atoui(port))){ + 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*/); + + if(sock < 0){ + 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; +}