--- parser3/src/targets/cgi/parser3.C 2020/12/07 22:49:59 1.323 +++ parser3/src/targets/cgi/parser3.C 2026/10/28 01:17:38 1.389 @@ -1,11 +1,11 @@ /** @file Parser: scripting and CGI main. - Copyright (c) 2001-2017 Art. Lebedev Studio (http://www.artlebedev.com) - Author: Alexandr Petrosian (http://paf.design.ru) + Copyright (c) 2001-2026 Art. Lebedev Studio (https://www.artlebedev.com) + Authors: Konstantin Morshnev , Alexandr Petrosian */ -volatile const char * IDENT_PARSER3_C="$Id: parser3.C,v 1.323 2020/12/07 22:49:59 moko Exp $"; +volatile const char * IDENT_PARSER3_C="$Id: parser3.C,v 1.389 2026/10/28 01:17:38 moko Exp $"; #include "pa_config_includes.h" @@ -14,19 +14,30 @@ volatile const char * IDENT_PARSER3_C="$ #include "pa_common.h" #include "pa_request.h" #include "pa_version.h" +#include "pa_threads.h" #include "pa_vconsole.h" #include "pa_sapi_info.h" +#include "pa_code_cache.h" +#include "pa_websocket.h" +#include "pa_os.h" #ifdef _MSC_VER #include #include #include + +extern "C" HANDLE WINAPI GC_CreateThread(LPSECURITY_ATTRIBUTES, DWORD, LPTHREAD_START_ROUTINE, LPVOID, DWORD, LPDWORD); + +#else + +extern "C" int GC_pthread_create(pthread_t *, const pthread_attr_t *, void *(*)(void *), void * /* arg */); + #endif // defines // comment remove me after debugging -//#define PA_DEBUG_CGI_ENTRY_EXIT "parser3-debug.log" +//#define PA_DEBUG_CGI_ENTRY_EXIT #if defined(_MSC_VER) && !defined(_DEBUG) # define PA_SUPPRESS_SYSTEM_EXCEPTION @@ -38,96 +49,101 @@ volatile const char * IDENT_PARSER3_C="$ #define PARSER_CONFIG_ENV_NAME "CGI_PARSER_CONFIG" #define PARSER_LOG_ENV_NAME "CGI_PARSER_LOG" +SAPI_Info sapi_console; +SAPI_Info_CGI sapi_cgi; + +static SAPI_Info *sapi_info = &sapi_cgi; +static THREAD_LOCAL SAPI_Info *sapi_info_4log = NULL; // global for correct send error in die() + +const char* parser3_mode = "cgi"; // $status:mode static const char* filespec_to_process = 0; // [file] static const char* httpd_host_port = 0; // -p option static const char* config_filespec = 0; // -f option or from env or next to the executable if exists +static const char* log_filespec = 0; // -l option or from env, full disk path static bool mail_received = false; // -m option? [asked to parse incoming message to $mail:received] -static char* parser3_filespec = 0; // argv[0] +static const char* parser3_filespec = 0; // argv[0] static char** argv_extra = NULL; -// for signal handlers -Request *request=0; - // for error logging -static Request_info request_info; // global for correct log() reporting -static const char* filespec_4log = NULL; // null only if system-wide auto.p used +static THREAD_LOCAL Request_info *request_info_4log = NULL; // global for correct log() reporting +static const char* filespec_4log = NULL; // null only if system-wide auto.p used; full disk path, parser3.log goes next to it -// SAPI +const char *parser3_log_filespec() { // $status:log-filename + const char* slog=log_filespec; -static void log(const char* fmt, va_list args) { - bool opened=false; - FILE *f=0; - - const char* log_by_env=getenv(PARSER_LOG_ENV_NAME); - if(!log_by_env) - log_by_env=getenv(REDIRECT_PREFIX PARSER_LOG_ENV_NAME); - if(log_by_env) { - f=fopen(log_by_env, "at"); - opened=f!=0; - } -#ifdef PA_DEBUG_CGI_ENTRY_EXIT - f=fopen(PA_DEBUG_CGI_ENTRY_EXIT, "at"); - opened=f!=0; -#endif + if(!slog) { + static THREAD_LOCAL char log_spec[MAX_STRING + 12 /* '/parser3.log' */]; + pa_strncpy(log_spec, filespec_4log, MAX_STRING); - if(!opened && filespec_4log) { - char beside_config_path[MAX_STRING]; - strncpy(beside_config_path, filespec_4log, MAX_STRING-1); beside_config_path[MAX_STRING-1]=0; - if(!(rsplit(beside_config_path, '/') || rsplit(beside_config_path, '\\'))) { // strip filename - // no path, just filename - strcpy(beside_config_path, "."); + if(char* log_dir_pos=strrchr(log_spec, '/')){ + strcpy(log_dir_pos, "/parser3.log"); + } else { + // system-wide auto.p used: the current directory + strcpy(log_spec, "./parser3.log"); } + slog=log_spec; + } + return slog; +} - char file_spec[MAX_STRING]; - snprintf(file_spec, MAX_STRING, "%s/parser3.log", beside_config_path); - f=fopen(file_spec, "at"); - opened=f!=0; - } - // fallback to stderr - if(!opened) - f=stderr; +// SAPI + +#ifndef O_CLOEXEC +#define O_CLOEXEC 0 +#endif +#ifndef STDERR_FILENO +#define STDERR_FILENO 2 +#endif - // use no memory [so that we could log out-of-memory error] - setbuf(f, 0); // stderr stream is unbuffered by default, but still... +static void pa_log(const char* stamp, const char* fmt, va_list args) { + // use no memory (to log out-of-memory error): the line is built on the stack and written at once + char line[64 + MAX_LOG_STRING + 1024 + 32 + 100]; // prefix + message + request info suffix: uri + method and the rest of it + size_t size=0; // prefix - time_t t=time(0); - if(const char* stamp=ctime(&t)) { // never saw that - if(size_t len=strlen(stamp)) // saw once stamp being ="" - fprintf(f, "[%.*s] [%u] ", (int)len-1, stamp, (unsigned int)getpid() ); - } - // message + if(!stamp) + stamp=pa_ctime(time(0)); + if(size_t len=strlen(stamp)) // saw once stamp being ="" + size=pa_snprintf(line, 64, "[%.*s] [%u] ", (int)len-1 /* exclude \n */, stamp, (unsigned int)pa_get_thread_id() ); - char buf[MAX_LOG_STRING]; - size_t size=vsnprintf(buf, MAX_LOG_STRING, fmt, args); - size=remove_crlf(buf, buf+size); - fwrite(buf, size, 1, f); - - if(request_info.method) { - fprintf(f, " [uri=%s, method=%s, cl=%lu]\n", request_info.uri ? request_info.uri : "", request_info.method, (unsigned long)request_info.content_length); - } else - fputs(" [no request info]\n", f); + // message + size+=remove_crlf(line+size, line+size+pa_vsnprintf(line+size, MAX_LOG_STRING, fmt, args)); - if(opened) - fclose(f); + // suffix + if(request_info_4log && request_info_4log->method) + size+=pa_snprintf(line+size, sizeof(line)-size, " [uri=%.1024s, method=%.32s, cl=%lu]\n", request_info_4log->uri ? request_info_4log->uri : "", request_info_4log->method, (unsigned long)request_info_4log->content_length); else - fflush(f); + size+=pa_snprintf(line+size, sizeof(line)-size, " [no request info]\n"); + + const char* slog=parser3_log_filespec(); + int fd=strcmp(slog,"-") ? pa_open(slog, O_WRONLY | O_APPEND | O_CREAT | O_CLOEXEC, 0666) : -1; + if(fd<0) + fd=STDERR_FILENO; // fallback to stderr + PA_UNUSED int ignore_result=write(fd, line, (unsigned int)size); + if(fd!=STDERR_FILENO) + close(fd); } -#ifdef PA_DEBUG_CGI_ENTRY_EXIT -static void log(const char* fmt, ...) { +void pa_log(const char* fmt, ...) { va_list args; va_start(args,fmt); - log(fmt, args); + pa_log(0, fmt, args); + va_end(args); +} + +// for signal handlers, where pa_ctime can't be called: the stamp is prepared in advance +static void pa_log_at(const char* stamp, const char* fmt, ...) { + va_list args; + va_start(args,fmt); + pa_log(stamp, fmt, args); va_end(args); } -#endif // appends to parser3.log located next to the config file if openable, to stderr otherwize void SAPI::log(SAPI_Info&, const char* fmt, ...) { va_list args; va_start(args,fmt); - ::log(fmt, args); + pa_log(0, fmt, args); va_end(args); } @@ -136,19 +152,23 @@ void SAPI::die(const char* fmt, ...) { // logging first, first vsnprintf va_start(args,fmt); - ::log(fmt, args); + pa_log(0, fmt, args); va_end(args); // inform user, second vsnprintf va_start(args, fmt); char message[MAX_STRING]; - vsnprintf(message, MAX_STRING, fmt, args); + pa_vsnprintf(message, MAX_STRING, fmt, args); - SAPI::send_error(*sapiInfo, message); + SAPI::send_error(sapi_info_4log ? *sapi_info_4log : *sapi_info, message); exit(1); // va_end(args); } +void SAPI::send_error(SAPI_Info& info, const char *exception_cstr, const char *status){ + info.send_error(exception_cstr, status); +} + char* SAPI::Env::get(SAPI_Info& info, const char* name) { return info.get_env(name); } @@ -161,78 +181,59 @@ const char* const *SAPI::Env::get(SAPI_I return info.get_env(); } -size_t SAPI::read_post(SAPI_Info& info, char *buf, size_t max_bytes) { +size_t SAPI::read_post(SAPI_Info& info, char* buf, size_t max_bytes) { return info.read_post(buf, max_bytes); } void SAPI::add_header_attribute(SAPI_Info& info, const char* dont_store_key, const char* dont_store_value) { - info.add_header_attribute(dont_store_key, dont_store_value); + info.add_header(dont_store_key, dont_store_value); } -void SAPI::send_header(SAPI_Info& info) { - info.send_header(); +void SAPI::send_headers(SAPI_Info& info) { + info.send_headers(); } -size_t SAPI::send_body(SAPI_Info& info, const void *buf, size_t size) { - return info.send_body(buf, size); +void SAPI::clear_headers(SAPI_Info& info) { + info.clear_headers(); } -static void full_disk_path(const char* file_name, char *buf, size_t buf_size) { - if(file_name[0]=='/' -#ifdef WIN32 - || file_name[0] && file_name[1]==':' -#endif - ){ - strncpy(buf, file_name, buf_size-1); buf[buf_size-1]=0; - } else { - char cwd[MAX_STRING]; - snprintf(buf, buf_size, "%s/%s", getcwd(cwd, MAX_STRING) ? cwd : "", file_name); - } -#ifdef WIN32 - back_slashes_to_slashes(buf); -#endif -} - -static void log_signal(const char* signal_name) { - SAPI::log(*sapiInfo, "%s received %s processing request", signal_name, request ? "while" : "before or after"); +size_t SAPI::send_body(SAPI_Info& info, const void *buf, size_t size) { + return info.send_body(buf, size); } -#ifdef SIGUSR1 -static void SIGUSR1_handler(int /*sig*/){ - log_signal("SIGUSR1"); +static const char* full_disk_path(const char* file_name = "") { + const char* path=backslashes_to_slashes(file_name); + if(path[0]=='/' || is_os_absolute_path(path)) // '/' includes the UNC path //server, as there are no web paths here; C: is left for is_os_absolute_path + return path; + char cwd[MAX_STRING]; + return pa_strcat(backslashes_to_slashes(getcwd(cwd, MAX_STRING) ? cwd : ""), "/", path); } -#endif #ifdef SIGPIPE -#define SIGPIPE_NAME "SIGPIPE" -static const String sigpipe_name(SIGPIPE_NAME); -static void SIGPIPE_handler(int /*sig*/){ - Value* sigpipe=0; - if(request) - sigpipe=request->main_class.get_element(sigpipe_name); - if(sigpipe && sigpipe->as_bool()) - log_signal(SIGPIPE_NAME); - - if(request) - request->set_skip(Request::SKIP_INTERRUPTED); -} +// write to a closed socket or pipe fails with EPIPE. Not SIG_IGN, as the ignored signal is inherited by the executed programs +static void SIGPIPE_handler(int /*sig*/){} #endif // requires pa_thread_request() in entry_exists() under Windows -static const char *locate_config(const char *config_filespec_option, const char *executable_path){ +static const char* locate_config(const char* config_filespec_option, const char* executable_path){ filespec_4log=config_filespec_option; if(!filespec_4log) filespec_4log=getenv(PARSER_CONFIG_ENV_NAME); if(!filespec_4log) filespec_4log=getenv(REDIRECT_PREFIX PARSER_CONFIG_ENV_NAME); if(!filespec_4log){ - // next to the executable - char *beside_executable_path = pa_strdup(executable_path); - bool stripped_filename = rsplit(beside_executable_path, '/') || rsplit(beside_executable_path, '\\'); - filespec_4log = pa_strcat(stripped_filename ? beside_executable_path : "." /* no path, just filename */ , "/" AUTO_FILE_NAME); - if(entry_exists(filespec_4log)) - return filespec_4log; + const char* exec_dir_pos = strrchr(executable_path, '/'); +#ifdef SYSTEM_CONFIG_FILE + if(exec_dir_pos){ +#endif + // next to the executable; full path should be used to avoid "parser already configured" + filespec_4log=full_disk_path(!exec_dir_pos || (exec_dir_pos==executable_path+1 && *executable_path=='.') + ? AUTO_FILE_NAME // parser3 or ./parser3 + : pa_strcat(pa_strdup(executable_path, exec_dir_pos - executable_path), "/" AUTO_FILE_NAME)); + if(entry_exists(filespec_4log)) + return filespec_4log; #ifdef SYSTEM_CONFIG_FILE + } if(entry_exists(SYSTEM_CONFIG_FILE)){ filespec_4log=NULL; return SYSTEM_CONFIG_FILE; @@ -240,11 +241,12 @@ static const char *locate_config(const c #endif return NULL; } - return filespec_4log; + // when -f auto.p full path should be used to avoid "parser already configured" + return filespec_4log=full_disk_path(filespec_4log); } #ifdef WIN32 -const char* maybe_reconstruct_IIS_status_in_qs(const char* original) { +static const char* maybe_reconstruct_IIS_status_in_qs(const char* original) { // 404;http://servername/page[?param=value...] // ';' should be urlencoded by HTTP standard, so we shouldn't get it from browser // and can consider that as an indication that this is IIS way to report errors @@ -277,95 +279,259 @@ const char* maybe_reconstruct_IIS_status } #define MAYBE_RECONSTRUCT_IIS_STATUS_IN_QS(s) maybe_reconstruct_IIS_status_in_qs(s) -#else + +#else + #define MAYBE_RECONSTRUCT_IIS_STATUS_IN_QS(s) s + #endif +#define RESPONSE_TIMEOUT_GRACE 1 // seconds to finish after the interruption, then the process is killed -class RequestController { +/// $response:timeout: interrupts the request that is not finished in time, the object is created before the request and destroyed after it +class Response_timeout { public: - RequestController(Request* r){ - ::request=r; + static THREAD_LOCAL Response_timeout *current; + + /// @a akill_process: the process will be killed if the request is still running after the grace period + Response_timeout(bool akill_process) : kill_process(akill_process), expired(0), armed(false) { + current=this; + } + ~Response_timeout() { + set(0); + current=0; + } + + /// restarts the timer for the given number of seconds, 0 cancels it + void set(int aseconds) { + if(armed){ + pa_timer_stop(timer); + armed=false; + } + if(!aseconds) + return; + + seconds=aseconds; + expired=0; + if(kill_process) // pa_ctime can't be called from the handler, so preparing for the log in advance + stamp=pa_strdup(pa_ctime(time(0)+seconds+RESPONSE_TIMEOUT_GRACE)); + + armed=pa_timer_start(&timer, seconds, kill_process ? RESPONSE_TIMEOUT_GRACE : 0, handler); + } + +private: + bool kill_process; + volatile int expired; + bool armed; + pa_timer timer; + int seconds; + const char *stamp; + + static void handler(int) { + Response_timeout *timeout=current; + if(!timeout) + return; + + if(timeout->expired && timeout->kill_process){ + // the request is still running after the grace period + pa_log_at(timeout->stamp, "response timeout (%d s) exceeded, process killed", timeout->seconds); +#ifndef _MSC_VER + kill(getpid(), SIGKILL); +#endif + return; + } + + timeout->expired=1; + if(thread_request) + thread_request->interrupt(); } - ~RequestController(){ - ::request=0; +}; + +THREAD_LOCAL Response_timeout *Response_timeout::current=0; + +void SAPI::set_response_timeout(SAPI_Info&, int seconds) { + if(Response_timeout::current) + Response_timeout::current->set(seconds); +} + +class RequestInfoController { +public: + RequestInfoController(Request_info* rinfo, SAPI_Info* sinfo){ + request_info_4log=rinfo; + sapi_info_4log=sinfo; + } + ~RequestInfoController(){ + request_info_4log=0; + sapi_info_4log=0; } }; +/** httpd support */ +static const String httpd_class_name("httpd"); + static void config_handler(SAPI_Info &info) { - char document_root_buf[MAX_STRING]; - full_disk_path("", document_root_buf, sizeof(document_root_buf)); + Request_info request_info; + RequestInfoController ric(&request_info, &info); - request_info.document_root = document_root_buf; + request_info.set_document_root(full_disk_path()); request_info.uri = ""; request_info.argv = argv_extra; // prepare to process request Request r(info, request_info, String::Language(String::L_HTML|String::L_OPTIMIZE_BIT)); - { - // initing ::request ptr for signal handlers - RequestController rc(&r); - // only once - config_filespec = locate_config(config_filespec, parser3_filespec); - // process main auto.p only - r.core(config_filespec, false, String::Empty); - // clearing ::request in RequestController desctructor to prevent signal handlers from accessing invalid memory - } + // only once + config_filespec = locate_config(config_filespec, parser3_filespec); + // process main auto.p only +#ifdef CODE_CACHE + Code_cache_manager::start(); +#endif + r.core(config_filespec, String::Empty); +#ifdef CODE_CACHE + Code_cache_manager::finish(r); +#endif } static void connection_handler(SAPI_Info_HTTPD &info, HTTPD_Connection &connection) { - connection.read_header(); - info.populate_env(); + Request_info request_info; + RequestInfoController ric(&request_info, &info); + + try { + if(!connection.read_header()) + return; // ignore "void" connections + info.populate_env(); + + request_info.set_document_root(full_disk_path()); + request_info.path_translated = filespec_to_process; + request_info.method = connection.method(); + request_info.query_string = connection.query(); + request_info.uri = request_info.strip_absolute_uri(connection.uri()); + request_info.content_type = connection.content_type(); + request_info.content_length = (size_t)connection.content_length(); + request_info.cookie = info.get_env("HTTP_COOKIE"); + request_info.mail_received = false; + request_info.argv = argv_extra; + + // process can be killed on timeout only in parallel mode + Response_timeout response_timeout(HTTPD_Server::mode == HTTPD_Server::PARALLEL); + + // prepare to process request + Request r(info, request_info, String::Language(String::L_HTML|String::L_OPTIMIZE_BIT)); + WebSocket websocket(connection); + r.websocket=&websocket; + // process the request + r.core(config_filespec, main_method_name, &httpd_class_name); + } catch(const Exception& e) { // exception in connection handling + SAPI::log(info, "%s", e.comment()); + const char* status = info.exception_http_status(e.type()); + if(*status) + SAPI::send_error(info, e.comment(), status); + } +} - char document_root_buf[MAX_STRING]; - full_disk_path("", document_root_buf, sizeof(document_root_buf)); +#ifdef _MSC_VER +DWORD WINAPI connection_thread(void *arg){ +#else +static void *connection_thread(void *arg){ +#endif + Temp_gc_references gc_references; // for pointers GC does not see, the very start of the thread - memset(&request_info, 0, sizeof(request_info)); - request_info.document_root = document_root_buf; - request_info.path_translated = filespec_to_process; - request_info.method = connection.method(); - request_info.query_string = connection.query(); - request_info.uri = request_info.strip_absolute_uri(connection.uri()); - request_info.content_type = connection.content_type(); - request_info.content_length = connection.content_length(); - request_info.cookie = info.get_env("HTTP_COOKIE"); - request_info.mail_received = false; - request_info.argv = argv_extra; + HTTPD_Connection &connection=*(HTTPD_Connection*)arg; + SAPI_Info_HTTPD info(connection); - // prepare to process request - Request r(info, request_info, String::Language(String::L_HTML|String::L_OPTIMIZE_BIT)); - { - // initing ::request ptr for signal handlers - RequestController rc(&r); - // process the request - r.core(config_filespec, strcasecmp(request_info.method, "HEAD")==0, String("httpd-main")); - // clearing ::request in RequestController desctructor to prevent signal handlers from accessing invalid memory + try { + connection_handler(info, connection); + } catch(const Exception& e) { // exception in send_error + pa_log("%s", e.comment()); } + + delete(&connection); + return 0; +} + +static void reject_connection(HTTPD_Connection &connection) { + // main thread only + static unsigned long total=0; + static time_t logged=0; + + total++; + time_t now=time(0); + if(now!=logged){ // not more than once a second + logged=now; + pa_log("max-connections limit %d reached, rejecting connection from %s (%lu rejected in total)", HTTPD_Server::max_connections, connection.remote_addr, total); + } + + SAPI_Info_HTTPD info(connection); + SAPI::send_error(info, "too many connections", "503"); } static void httpd_mode() { - int sock = HTTPD_Server::bind(httpd_host_port); + config_handler(*sapi_info); - config_handler(*sapiInfo); + SOCKET sock = HTTPD_Server::bind(httpd_host_port); while(1){ +#ifndef _MSC_VER + pid_t pid=1; +#endif try { HTTPD_Connection connection; - if(!connection.accept(sock, 5)) + bool accepted=connection.accept(sock, 500); +#ifndef _MSC_VER + if(HTTPD_Server::mode == HTTPD_Server::PARALLEL) + while (waitpid((pid_t)(-1), 0, WNOHANG) > 0) + HTTPD_Server::track_connection(-1); // the child has exited, some of them while we were waiting +#endif + if(!accepted) continue; - SAPI_Info_HTTPD info(connection); + if(HTTPD_Server::too_many_connections()){ + reject_connection(connection); // no thread or process is started + continue; + } - try { // connection try - connection_handler(info, connection); - } catch(const Exception& e) { // exception in connection handling or unhandled exception - SAPI::log(info, "%s", e.comment()); - SAPI::send_error(info, e.comment(), info.exception_http_status(e.type())); + switch (HTTPD_Server::mode) { + case HTTPD_Server::MULTITHREADED: +#ifdef _MSC_VER + if (!GC_CreateThread(0, 0, connection_thread, new HTTPD_Connection(connection), 0, 0)) + throw Exception("httpd.fork", 0, "thread creation failed"); + connection.sock=INVALID_SOCKET; // thread owns the connection now + break; +#else +#ifdef HAVE_TLS + pthread_t thread; + pthread_attr_t attr; + pthread_attr_init(&attr); + pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED); + + if(int result=GC_pthread_create(&thread, &attr, connection_thread, new HTTPD_Connection(connection))) + throw Exception("httpd.fork", 0, "thread creation failed (%d)", result); + connection.sock=INVALID_SOCKET; // thread owns the connection now + break; +#endif + case HTTPD_Server::PARALLEL: + pid=fork(); + if(pid<0) + throw Exception("httpd.fork", 0, "fork failed: %s (%d)", strerror(errno), errno); + if(pid>0){ + closesocket(connection.sock); // parent should close connection.sock as well + connection.sock=INVALID_SOCKET; // child owns the connection now + continue; + } +#endif + case HTTPD_Server::SEQUENTIAL: // and fork child + + SAPI_Info_HTTPD info(connection); + connection_handler(info, connection); } - // closing connection socket in HTTPD_Connection destructor + // SEQUENTIAL or error: closing connection socket in HTTPD_Connection destructor } catch(const Exception& e) { // exception in accept or send_error - SAPI::log(*sapiInfo, "%s", e.comment()); + pa_log("%s", e.comment()); } + +#ifndef _MSC_VER + if(pid==0) // fork child + exit(0); +#endif } } @@ -381,12 +547,15 @@ static void real_parser_handler(bool cgi const char* request_method=getenv("REQUEST_METHOD"); - if(!filespec_to_process) - SAPI::die("Parser/%s", PARSER_VERSION); - - char document_root_buf[MAX_STRING]; + if(!filespec_to_process){ + SAPI::send_error(*sapi_info, "Parser/" PARSER_VERSION_WITH_FEATURES); + exit(1); + } // global request info + Request_info request_info; + RequestInfoController ric(&request_info, sapi_info); + request_info.path_translated = filespec_to_process; request_info.method = request_method ? request_method : "GET"; request_info.query_string = MAYBE_RECONSTRUCT_IIS_STATUS_IN_QS(getenv("QUERY_STRING")); @@ -395,26 +564,25 @@ static void real_parser_handler(bool cgi // obligatory const char* path_info=getenv("PATH_INFO"); if(!path_info) - SAPI::die("CGI: illegal call (missing PATH_INFO)"); + SAPI::die("parser3: illegal CGI call (missing PATH_INFO)"); - request_info.document_root = getenv("DOCUMENT_ROOT"); - if(!request_info.document_root) { + const char* document_root = getenv("DOCUMENT_ROOT"); + if(!document_root) { // IIS or fcgiwrap minimalistic setup ssize_t prefix_len = strlen(filespec_to_process) - strlen(path_info); if(prefix_len < 0 || strcmp(filespec_to_process + prefix_len, path_info) != 0) - SAPI::die("CGI: illegal call (invalid PATH_INFO in reinventing DOCUMENT_ROOT)"); + SAPI::die("parser3: illegal CGI call (invalid PATH_INFO in reinventing DOCUMENT_ROOT)"); - char* document_root = new(PointerFreeGC) char[prefix_len + 1/*0*/]; - memcpy(document_root, filespec_to_process, prefix_len); document_root[prefix_len] = 0; - request_info.document_root = document_root; + document_root = pa_strdup(filespec_to_process, prefix_len); } + request_info.set_document_root(document_root); request_info.uri = request_info.strip_absolute_uri(getenv("REQUEST_URI")); if(request_info.uri) { // apache & others stuck to standards // another obligatory const char* script_name = getenv("SCRIPT_NAME"); if(!script_name) - SAPI::die("CGI: illegal call (missing SCRIPT_NAME)"); + SAPI::die("parser3: illegal CGI call (missing SCRIPT_NAME)"); /* http://parser3/env.html?123 =OK $request:uri=/env.html?123 @@ -431,35 +599,35 @@ static void real_parser_handler(bool cgi size_t script_name_len = strlen(script_name); size_t uri_len = strlen(request_info.uri); if(strncmp(request_info.uri, script_name, script_name_len)==0 && script_name_len != uri_len) // under IIS they are the same - SAPI::die("CGI: illegal call (1)"); + SAPI::die("parser3: illegal CGI call (REQUEST_URI starts with SCRIPT_NAME)"); } else { // fcgiwrap minimalistic setup request_info.uri = request_info.query_string && *request_info.query_string ? pa_strcat(path_info, "?", request_info.query_string) : path_info; } } else{ - full_disk_path("", document_root_buf, sizeof(document_root_buf)); - request_info.document_root = document_root_buf; + request_info.set_document_root(full_disk_path()); request_info.uri = ""; } request_info.content_type = getenv("CONTENT_TYPE"); - request_info.content_length = pa_atoul(getenv("CONTENT_LENGTH")); + request_info.content_length = cgi ? (size_t)pa_atoul(getenv("CONTENT_LENGTH")) : 0; // only SAPI_Info_CGI can read POST request_info.cookie = getenv("HTTP_COOKIE"); request_info.mail_received = mail_received; request_info.argv = argv_extra; #ifdef PA_DEBUG_CGI_ENTRY_EXIT - log("request_info: method=%s, uri=%s, q=%s, dr=%s, pt=%s", request_info.method, request_info.uri, request_info.query_string, request_info.document_root, request_info.path_translated); + pa_log("request_info: method=%s, uri=%s, q=%s, dr=%s, pt=%s", request_info.method, request_info.uri, request_info.query_string, request_info.document_root, request_info.path_translated); #endif - // prepare to process request - Request r(*sapiInfo, request_info, cgi ? String::Language(String::L_HTML|String::L_OPTIMIZE_BIT) : String::L_AS_IS); { - // initing ::request ptr for signal handlers - RequestController rc(&r); + // CGI process can be killed on timeout + Response_timeout response_timeout(true); + + // prepare to process request + Request r(*sapi_info, request_info, cgi ? String::Language(String::L_HTML|String::L_OPTIMIZE_BIT) : String::L_AS_IS); // process the request - r.core(locate_config(config_filespec, parser3_filespec), strcasecmp(request_info.method, "HEAD")==0); - // clearing ::request in RequestController destructor to prevent signal handlers from accessing invalid memory + r.core(locate_config(config_filespec, parser3_filespec)); + // then the request is destroyed, before the libraries are unloaded: its pool cleanups call them (memcached) } // finalize libraries @@ -486,7 +654,7 @@ static void call_real_parser_handler__su } __except ( (system_exception=GetExceptionInformation()), EXCEPTION_EXECUTE_HANDLER) { if(system_exception) if(_EXCEPTION_RECORD *er=system_exception->ExceptionRecord) - throw Exception("system", 0, "0x%08X at 0x%08X", er->ExceptionCode, er->ExceptionAddress); + throw Exception("system", 0, "0x%08X at %p", er->ExceptionCode, er->ExceptionAddress); else throw Exception("system", 0, ""); else @@ -502,11 +670,15 @@ static void call_real_parser_handler__su #define REAL_PARSER_HANDLER real_parser_handler #endif -static void usage(const char* program) { +static void usage(const char* message=NULL) { + if(message){ + fprintf(stderr, message, parser3_filespec); + } + printf( "Parser/%s\n" - "Copyright (c) 2001-2017 Art. Lebedev Studio (http://www.artlebedev.com)\n" - "Author: Alexandr Petrosian (http://paf.design.ru)\n" + "Copyright (c) 2001-2026 Art. Lebedev Studio (https://www.artlebedev.com)\n" + "Authors: Konstantin Morshnev , Alexandr Petrosian \n" "\n" "Usage: %s [options] [file]\n" "Options are:\n" @@ -514,95 +686,106 @@ static void usage(const char* program) { " -m Parse mail, put received letter to $mail:received\n" #endif " -f config_file Use this config file (/path/to/auto.p)\n" + " -l log_file Use this log file (/path/to/parser3.log)\n" " -p [host:]port Start web server on this port\n" " -h Display usage information (this message)\n", - PARSER_VERSION, - program); + PARSER_VERSION_WITH_FEATURES, + parser3_filespec ? parser3_filespec : "parser3" ); exit(EINVAL); } +#define ARG_REQUIRED \ + if(c[1] || !*(++carg)){ \ + fprintf(stderr, "%s: option '%c' requires an argument\n", parser3_filespec, *c); \ + usage(); \ + } int main(int argc, char *argv[]) { + Temp_gc_references gc_references; // for pointers GC does not see + #ifdef PA_DEBUG_CGI_ENTRY_EXIT - log("main: entry"); + pa_log("main: entry"); #endif - parser3_filespec = argv[0]; + if(!argc || !argv[0]) + usage(); + parser3_filespec = backslashes_to_slashes(argv[0]); + umask(2); // were we started as CGI? bool cgi=(getenv("SERVER_SOFTWARE") || getenv("SERVER_NAME") || getenv("GATEWAY_INTERFACE") || getenv("REQUEST_METHOD")) && !getenv("PARSER_VERSION"); - sapiInfo = cgi ? new SAPI_Info_CGI() : new SAPI_Info(); + if(!cgi){ + sapi_info = &sapi_console; + parser3_mode = "console"; + } -#ifdef SIGUSR1 - if(signal(SIGUSR1, SIGUSR1_handler)==SIG_ERR) - SAPI::die("Can not set handler for SIGUSR1"); -#endif #ifdef SIGPIPE - if(signal(SIGPIPE, SIGPIPE_handler)==SIG_ERR) - SAPI::die("Can not set handler for SIGPIPE"); + signal(SIGPIPE, SIGPIPE_handler); #endif - - char *raw_filespec_to_process = NULL; + char* raw_filespec_to_process = NULL; if(cgi) { raw_filespec_to_process=getenv("PATH_TRANSLATED"); argv_extra=argv + 1; } else { - int optind=1; - while(optind < argc){ - char *carg = argv[optind]; - if(carg[0] != '-') + char** carg = argv + 1; + for(;*carg; carg++){ + if((*carg)[0] != '-') break; - for(size_t k = 1; k < strlen(carg); k++){ - char c = carg[k]; - switch (c) { + for(char* c=(*carg)+1; *c; c++){ + switch (*c) { case 'h': - usage(argv[0]); + usage(); break; case 'f': - if(optind < argc - 1){ - optind++; - config_filespec=argv[optind]; - } + ARG_REQUIRED; + config_filespec=*carg; + break; + case 'l': + ARG_REQUIRED; + log_filespec=*carg; break; case 'p': - if(optind < argc - 1){ - optind++; - httpd_host_port=argv[optind]; - } + ARG_REQUIRED; + httpd_host_port=*carg; + parser3_mode="httpd"; break; #ifdef WITH_MAILRECEIVE case 'm': mail_received=true; + parser3_mode="mail"; break; #endif default: - fprintf(stderr, "%s: invalid option '%c'\n", argv[0], c); - usage(argv[0]); + fprintf(stderr, "%s: invalid option '%c'\n", parser3_filespec, *c); + usage(); break; } } - optind++; } - if (optind > argc - 1) { - if(!httpd_host_port) { - fprintf(stderr, "%s: file not specified\n", argv[0]); - usage(argv[0]); - } + if (*carg) { + raw_filespec_to_process=*carg; } else { - raw_filespec_to_process=argv[optind]; + if(!httpd_host_port) + usage("%s: file not specified\n"); } - if (httpd_host_port && mail_received) { - fprintf(stderr, "%s: -p and -m options should not be used together\n", argv[0]); - usage(argv[0]); - } + if (httpd_host_port && mail_received) + usage("%s: -p and -m options should not be used together\n"); - argv_extra=argv + optind; + argv_extra=carg; } + // prepare full disk path log_filespec here, as logging must not allocate + if(!log_filespec) + log_filespec=getenv(PARSER_LOG_ENV_NAME); + if(!log_filespec) + log_filespec=getenv(REDIRECT_PREFIX PARSER_LOG_ENV_NAME); + if(log_filespec && strcmp(log_filespec, "-")) // "-" is stderr + log_filespec=full_disk_path(log_filespec); + #ifdef _MSC_VER setmode(fileno(stdin), _O_BINARY); setmode(fileno(stdout), _O_BINARY); @@ -624,19 +807,18 @@ int main(int argc, char *argv[]) { #endif try { // global try - char filespec_to_process_buf[MAX_STRING]; if(raw_filespec_to_process && *raw_filespec_to_process){ - full_disk_path(raw_filespec_to_process, filespec_to_process_buf, sizeof(filespec_to_process_buf)); - filespec_to_process=filespec_to_process_buf; + filespec_to_process=full_disk_path(raw_filespec_to_process); } REAL_PARSER_HANDLER(cgi); - } catch(const Exception& e) { // exception in unhandled exception - SAPI::die("%s", e.comment()); + } catch(const Exception& e) { // exception in config_handler + SAPI::log(*sapi_info, "%s", e.comment()); + SAPI::send_error(*sapi_info, e.comment(), strcmp(e.type(), "file.missing") ? "500" : "404"); } #ifdef PA_DEBUG_CGI_ENTRY_EXIT - log("main: successful return"); + pa_log("main: successful return"); #endif - return sapiInfo->http_response_code < 100 ? sapiInfo->http_response_code : 0; + return sapi_info->http_response_code < 100 ? sapi_info->http_response_code : 0; }