/* * Copyright (C) 2007-2009 Sourcefire, Inc. * * Authors: Tomasz Kojm, Trog, Török Edvin * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, * MA 02110-1301, USA. */ #if HAVE_CONFIG_H #include "clamav-config.h" #endif #include #include #include #include #include #include #include #ifndef _WIN32 #include #include #include #include #endif #ifdef HAVE_UNISTD_H #include #endif #include #ifdef C_SOLARIS #include #endif #include "libclamav/clamav.h" #include "shared/output.h" #include "shared/optparser.h" #include "server.h" #include "thrmgr.h" #include "session.h" #include "clamuko.h" #include "others.h" #include "shared.h" #include "libclamav/others.h" #include "libclamav/readdb.h" #include "libclamav/cltypes.h" #define BUFFSIZE 1024 int progexit = 0; pthread_mutex_t exit_mutex = PTHREAD_MUTEX_INITIALIZER; int reload = 0; time_t reloaded_time = 0; pthread_mutex_t reload_mutex = PTHREAD_MUTEX_INITIALIZER; int sighup = 0; static struct cl_stat dbstat; static void scanner_thread(void *arg) { client_conn_t *conn = (client_conn_t *) arg; #ifndef _WIN32 sigset_t sigset; #endif int ret; unsigned virus=0, errors = 0; #ifndef _WIN32 /* ignore all signals */ sigfillset(&sigset); /* The behavior of a process is undefined after it ignores a * SIGFPE, SIGILL, SIGSEGV, or SIGBUS signal */ sigdelset(&sigset, SIGFPE); sigdelset(&sigset, SIGILL); sigdelset(&sigset, SIGSEGV); #ifdef SIGBUS sigdelset(&sigset, SIGBUS); #endif sigdelset(&sigset, SIGTSTP); sigdelset(&sigset, SIGCONT); pthread_sigmask(SIG_SETMASK, &sigset, NULL); #endif ret = command(conn, &virus); if (ret == -1) { pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); errors = 1; } else errors = ret; thrmgr_setactiveengine(NULL); if (conn->filename) free(conn->filename); logg("$Finished scanthread\n"); if (thrmgr_group_finished(conn->group, virus ? EXIT_OTHER : errors ? EXIT_ERROR : EXIT_OK)) { logg("$Scanthread: connection shut down (FD %d)\n", conn->sd); /* close connection if we were last in group */ shutdown(conn->sd, 2); closesocket(conn->sd); } cl_engine_free(conn->engine); free(conn); return; } static int syncpipe_wake_recv_w = -1; void sighandler_th(int sig) { int action = 0; switch(sig) { case SIGINT: case SIGTERM: progexit = 1; action = 1; break; #ifdef SIGHUP case SIGHUP: sighup = 1; action = 1; break; #endif #ifdef SIGUSR2 case SIGUSR2: reload = 1; action = 1; break; #endif default: break; /* Take no action on other signals - e.g. SIGPIPE */ } /* a signal doesn't always wake poll(), for example on FreeBSD */ if (action && syncpipe_wake_recv_w != -1) if (write(syncpipe_wake_recv_w, "", 1) != 1) logg("$Failed to write to syncpipe\n"); } static struct cl_engine *reload_db(struct cl_engine *engine, unsigned int dboptions, const struct optstruct *opts, int do_check, int *ret) { const char *dbdir; int retval; unsigned int sigs = 0; struct cl_settings *settings = NULL; *ret = 0; if(do_check) { if(!dbstat.entries) { logg("No stats for Database check - forcing reload\n"); return engine; } if(cl_statchkdir(&dbstat) == 1) { logg("SelfCheck: Database modification detected. Forcing reload.\n"); return engine; } else { logg("SelfCheck: Database status OK.\n"); return NULL; } } /* release old structure */ if(engine) { /* copy current settings */ settings = cl_engine_settings_copy(engine); if(!settings) logg("^Can't make a copy of the current engine settings\n"); thrmgr_setactiveengine(NULL); cl_engine_free(engine); } dbdir = optget(opts, "DatabaseDirectory")->strarg; logg("Reading databases from %s\n", dbdir); if(dbstat.entries) cl_statfree(&dbstat); memset(&dbstat, 0, sizeof(struct cl_stat)); if((retval = cl_statinidir(dbdir, &dbstat))) { logg("!cl_statinidir() failed: %s\n", cl_strerror(retval)); *ret = 1; if(settings) cl_engine_settings_free(settings); return NULL; } if(!(engine = cl_engine_new())) { logg("!Can't initialize antivirus engine\n"); *ret = 1; if(settings) cl_engine_settings_free(settings); return NULL; } if(settings) { retval = cl_engine_settings_apply(engine, settings); if(retval != CL_SUCCESS) { logg("^Can't apply previous engine settings: %s\n", cl_strerror(retval)); logg("^Using default engine settings\n"); } cl_engine_settings_free(settings); } if((retval = cl_load(dbdir, engine, &sigs, dboptions))) { logg("!reload db failed: %s\n", cl_strerror(retval)); cl_engine_free(engine); *ret = 1; return NULL; } if((retval = cl_engine_compile(engine)) != 0) { logg("!Database initialization error: can't compile engine: %s\n", cl_strerror(retval)); cl_engine_free(engine); *ret = 1; return NULL; } logg("Database correctly reloaded (%u signatures)\n", sigs); thrmgr_setactiveengine(engine); return engine; } /* * zCOMMANDS are delimited by \0 * nCOMMANDS are delimited by \n * Old-style non-prefixed commands are one packet, optionally delimited by \n, * with trailing \r|\n ignored */ static const char *get_cmd(struct fd_buf *buf, size_t off, size_t *len, char *term, int *oldstyle) { unsigned char *pos; if (!buf->off || off >= buf->off) { *len = 0; return NULL; } *term = '\n'; switch (buf->buffer[off]) { /* commands terminated by delimiters */ case 'z': *term = '\0'; case 'n': pos = memchr(buf->buffer + off, *term, buf->off - off); if (!pos) { /* we don't have another full command yet */ *len = 0; return NULL; } *pos = '\0'; if (*term) { *len = cli_chomp(buf->buffer + off); } else { *len = pos - buf->buffer - off; } *oldstyle = 0; return buf->buffer + off + 1; default: /* one packet = one command */ if (off) return NULL; pos = memchr(buf->buffer, '\n', buf->off); if (pos) { *len = pos - buf->buffer; *pos = '\0'; } else { *len = buf->off; buf->buffer[buf->off] = '\0'; } cli_chomp(buf->buffer); *oldstyle = 1; return buf->buffer; } } struct acceptdata { struct fd_data fds; struct fd_data recv_fds; pthread_cond_t cond_nfds; int max_queue; int commandtimeout; int syncpipe_wake_recv[2]; int syncpipe_wake_accept[2]; }; #define ACCEPTDATA_INIT(mutex1, mutex2) { FDS_INIT(mutex1), FDS_INIT(mutex2), PTHREAD_COND_INITIALIZER, 0, 0, {-1, -1}, {-1, -1}} static void *acceptloop_th(void *arg) { char buff[BUFFSIZE + 1]; size_t i; struct acceptdata *data = (struct acceptdata*)arg; struct fd_data *fds = &data->fds; struct fd_data *recv_fds = &data->recv_fds; int max_queue = data->max_queue; int commandtimeout = data->commandtimeout; pthread_mutex_lock(fds->buf_mutex); for (;;) { /* Block waiting for data to become available for reading */ int new_sd = fds_poll_recv(fds, -1, 0); /* TODO: what about sockets that get rm-ed? */ if (!fds->nfds) { /* no more sockets to poll, all gave an error */ logg("!Main socket gone: fatal\n"); break; } if (new_sd == -1 && errno != EINTR) { logg("!Failed to poll sockets, fatal\n"); pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); break; } /* accept() loop */ for (i=0;i < fds->nfds && new_sd >= 0; i++) { struct fd_buf *buf = &fds->buf[i]; if (!buf->got_newdata) continue; if (buf->fd == data->syncpipe_wake_accept[0]) { /* dummy sync pipe, just to wake us */ if (read(buf->fd, buff, sizeof(buff)) < 0) { logg("^Syncpipe read failed\n"); } continue; } if (buf->got_newdata == -1) { logg("$Acceptloop closed FD: %d\n", buf->fd); shutdown(buf->fd, 2); closesocket(buf->fd); buf->fd = -1; continue; } /* don't accept unlimited number of connections, or * we'll run out of file descriptors */ pthread_mutex_lock(recv_fds->buf_mutex); while (recv_fds->nfds > (unsigned)max_queue) { pthread_mutex_lock(&exit_mutex); if(progexit) { pthread_mutex_unlock(&exit_mutex); break; } pthread_mutex_unlock(&exit_mutex); pthread_cond_wait(&data->cond_nfds, recv_fds->buf_mutex); } pthread_mutex_unlock(recv_fds->buf_mutex); pthread_mutex_lock(&exit_mutex); if(progexit) { pthread_mutex_unlock(&exit_mutex); break; } pthread_mutex_unlock(&exit_mutex); /* listen only socket */ new_sd = accept(fds->buf[i].fd, NULL, NULL); if (new_sd >= 0) { int ret, flags; #ifdef F_GETFL flags = fcntl(new_sd, F_GETFL, 0); if (flags != -1) { if (fcntl(new_sd, F_SETFL, flags | O_NONBLOCK) == -1) { logg("^Can't set socket to nonblocking mode, errno %d\n", errno); } } else { logg("^Can't get socket flags, errno %d\n", errno); } #else logg("^Nonblocking sockets not available!\n"); #endif logg("$Got new connection, FD %d\n", new_sd); pthread_mutex_lock(recv_fds->buf_mutex); ret = fds_add(recv_fds, new_sd, 0, commandtimeout); pthread_mutex_unlock(recv_fds->buf_mutex); if (ret == -1) { logg("!fds_add failed\n"); closesocket(new_sd); continue; } /* notify recvloop */ if (write(data->syncpipe_wake_recv[1], "", 1) == -1) { logg("!write syncpipe failed\n"); continue; } } else if (errno != EINTR) { /* very bad - need to exit or restart */ #ifdef HAVE_STRERROR_R strerror_r(errno, buff, BUFFSIZE); logg("!accept() failed: %s\n", buff); #else logg("!accept() failed\n"); #endif /* give the poll loop a chance to close disconnected FDs */ break; } } /* handle progexit */ pthread_mutex_lock(&exit_mutex); if (progexit) { pthread_mutex_unlock(&exit_mutex); break; } pthread_mutex_unlock(&exit_mutex); } pthread_mutex_unlock(fds->buf_mutex); for (i=0;i < fds->nfds; i++) { if (fds->buf[i].fd == -1) continue; logg("$Shutdown: closed fd %d\n", fds->buf[i].fd); shutdown(fds->buf[i].fd, 2); closesocket(fds->buf[i].fd); } fds_free(fds); pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); if (write(data->syncpipe_wake_recv[1], "", 1) < 0) { logg("$Syncpipe write failed\n"); } return NULL; } static const unsigned char* parse_dispatch_cmd(client_conn_t *conn, struct fd_buf *buf, size_t *ppos, int *error, const struct optstruct *opts, int readtimeout) { const unsigned char *cmd = NULL; int rc; size_t cmdlen; char term; int oldstyle; size_t pos = *ppos; /* Parse & dispatch commands */ while ((conn->mode == MODE_COMMAND) && (cmd = get_cmd(buf, pos, &cmdlen, &term, &oldstyle)) != NULL) { const char *argument; enum commands cmdtype; if (conn->group && oldstyle) { logg("$Received oldstyle command inside IDSESSION: %s\n", cmd); conn_reply_error(conn, "Only nCMDS\\n and zCMDS\\0 are accepted inside IDSESSION."); *error = 1; break; } cmdtype = parse_command(cmd, &argument, oldstyle); logg("$got command %s (%u, %u), argument: %s\n", cmd, (unsigned)cmdlen, (unsigned)cmdtype, argument ? argument : ""); if (cmdtype == COMMAND_FILDES) { if (buf->buffer + buf->off <= cmd + strlen("FILDES\n")) { /* we need the extra byte from recvmsg */ conn->mode = MODE_WAITANCILL; buf->mode = MODE_WAITANCILL; /* put term back */ buf->buffer[pos + cmdlen] = term; cmdlen = 0; logg("$RECVTH: mode -> MODE_WAITANCILL\n"); break; } /* eat extra \0 for controlmsg */ cmdlen++; logg("$RECVTH: FILDES command complete\n"); } conn->term = term; buf->term = term; if ((rc = execute_or_dispatch_command(conn, cmdtype, argument)) < 0) { logg("!Command dispatch failed\n"); if(rc == -1 && optget(opts, "ExitOnOOM")->enabled) { pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); } *error = 1; } if (thrmgr_group_need_terminate(conn->group)) { logg("$Receive thread: have to terminate group\n"); *error = CL_ETIMEOUT; break; } if (*error || !conn->group || rc) { if (rc && thrmgr_group_finished(conn->group, EXIT_OK)) { logg("$Receive thread: closing conn (FD %d), group finished\n", conn->sd); /* if there are no more active jobs */ shutdown(conn->sd, 2); closesocket(conn->sd); buf->fd = -1; conn->group = NULL; } else if (conn->mode != MODE_STREAM) { logg("$mode -> MODE_WAITREPLY\n"); /* no more commands are accepted */ conn->mode = MODE_WAITREPLY; /* Stop monitoring this FD, it will be closed either * by us, or by the scanner thread. * Never close a file descriptor that is being * monitored by poll()/select() from another thread, * because this can lead to subtle bugs such as: * Other thread closes file descriptor -> POLLHUP is * set, but the poller thread doesn't wake up yet. * Another client opens a connection and sends some * data. If the socket reuses the previous file descriptor, * then POLLIN is set on the file descriptor too. * When poll() wakes up it sees POLLIN | POLLHUP * and thinks that the client has sent some data, * and closed the connection, so clamd closes the * connection in turn resulting in a bug. * * If we wouldn't have poll()-ed the file descriptor * we closed in another thread, but rather made sure * that we don't put a FD that we're about to close * into poll()'s list of watched fds; then POLLHUP * would be set, but the file descriptor would stay * open, until we wake up from poll() and close it. * Thus a new connection won't be able to reuse the * same FD, and there is no bug. * */ buf->fd = -1; } } /* we received a command, set readtimeout */ time(&buf->timeout_at); buf->timeout_at += readtimeout; pos += cmdlen+1; if (conn->mode == MODE_STREAM) { /* TODO: this doesn't belong here */ buf->dumpname = conn->filename; buf->dumpfd = conn->scanfd; logg("$Receive thread: INSTREAM: %s fd %u\n", buf->dumpname, buf->dumpfd); } if (conn->mode != MODE_COMMAND) { logg("$Breaking command loop, mode is no longer MODE_COMMAND\n"); break; } conn->id++; } *ppos = pos; buf->mode = conn->mode; buf->id = conn->id; buf->group = conn->group; buf->quota = conn->quota; if (conn->scanfd != -1 && conn->scanfd != buf->dumpfd) { logg("$Unclaimed file descriptor received, closing: %d\n", conn->scanfd); close(conn->scanfd); /* protocol error */ conn_reply_error(conn, "PROTOCOL ERROR: ancillary data sent without FILDES."); *error = 1; return NULL; } if (!*error) { /* move partial command to beginning of buffer */ if (pos < buf->off) { memmove (buf->buffer, &buf->buffer[pos], buf->off - pos); buf->off -= pos; } else buf->off = 0; if (buf->off) logg("$Moved partial command: %lu\n", (unsigned long)buf->off); else logg("$Consumed entire command\n"); } *ppos = pos; return cmd; } /* static const unsigned char* parse_dispatch_cmd(client_conn_t *conn, struct fd_buf *buf, size_t *ppos, int *error, const struct optstruct *opts, int readtimeout) */ static int handle_stream(client_conn_t *conn, struct fd_buf *buf, const struct optstruct *opts, int *error, size_t *ppos, int readtimeout) { int rc; size_t pos = *ppos; size_t cmdlen; logg("$mode == MODE_STREAM\n"); /* we received a chunk, set readtimeout */ time(&buf->timeout_at); buf->timeout_at += readtimeout; if (!buf->chunksize) { /* read chunksize */ if (buf->off >= 4) { uint32_t cs = *(uint32_t*)buf->buffer; buf->chunksize = ntohl(cs); logg("$Got chunksize: %u\n", buf->chunksize); if (!buf->chunksize) { /* chunksize 0 marks end of stream */ conn->scanfd = buf->dumpfd; conn->term = buf->term; buf->dumpfd = -1; buf->mode = buf->group ? MODE_COMMAND : MODE_WAITREPLY; if (buf->mode == MODE_WAITREPLY) buf->fd = -1; logg("$Chunks complete\n"); buf->dumpname = NULL; if ((rc = execute_or_dispatch_command(conn, COMMAND_INSTREAMSCAN, NULL)) < 0) { logg("!Command dispatch failed\n"); if(rc == -1 && optget(opts, "ExitOnOOM")->enabled) { pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); } *error = 1; } else { pos = 4; memmove (buf->buffer, &buf->buffer[pos], buf->off - pos); buf->off -= pos; *ppos = 0; buf->id++; return 0; } } if (buf->chunksize > buf->quota) { logg("^INSTREAM: Size limit reached, (requested: %lu, max: %lu)\n", (unsigned long)buf->chunksize, (unsigned long)buf->quota); conn_reply_error(conn, "INSTREAM size limit exceeded."); *error = 1; *ppos = pos; return -1; } else { buf->quota -= buf->chunksize; } logg("$Quota: %lu\n", buf->quota); pos = 4; } else return -1; } else pos = 0; if (pos + buf->chunksize < buf->off) cmdlen = buf->chunksize; else cmdlen = buf->off - pos; buf->chunksize -= cmdlen; if (cli_writen(buf->dumpfd, buf->buffer + pos, cmdlen) < 0) { conn_reply_error(conn, "Error writing to temporary file"); logg("!INSTREAM: Can't write to temporary file.\n"); *error = 1; } logg("$Processed %lu bytes of chunkdata\n", cmdlen); pos += cmdlen; if (pos == buf->off) { buf->off = 0; } *ppos = pos; return 0; } int recvloop_th(int *socketds, unsigned nsockets, struct cl_engine *engine, unsigned int dboptions, const struct optstruct *opts) { int max_threads, max_queue, readtimeout, ret = 0; unsigned int options = 0; char timestr[32]; #ifndef _WIN32 struct sigaction sigact; sigset_t sigset; struct rlimit rlim; #endif mode_t old_umask; const struct optstruct *opt; char buff[BUFFSIZE + 1]; pid_t mainpid; int idletimeout; unsigned long long val; size_t i, j, rr_last = 0; pthread_t accept_th; pthread_mutex_t fds_mutex = PTHREAD_MUTEX_INITIALIZER; pthread_mutex_t recvfds_mutex = PTHREAD_MUTEX_INITIALIZER; struct acceptdata acceptdata = ACCEPTDATA_INIT(&fds_mutex, &recvfds_mutex); struct fd_data *fds = &acceptdata.recv_fds; time_t start_time, current_time; unsigned int selfchk; threadpool_t *thr_pool; #ifdef CLAMUKO pthread_t clamuko_pid; pthread_attr_t clamuko_attr; struct thrarg *tharg = NULL; /* shut up gcc */ #endif #ifndef _WIN32 memset(&sigact, 0, sizeof(struct sigaction)); #endif /* set up limits */ if((opt = optget(opts, "MaxScanSize"))->active) { if((ret = cl_engine_set_num(engine, CL_ENGINE_MAX_SCANSIZE, opt->numarg))) { logg("!cl_engine_set_num(CL_ENGINE_MAX_SCANSIZE) failed: %s\n", cl_strerror(ret)); cl_engine_free(engine); return 1; } } val = cl_engine_get_num(engine, CL_ENGINE_MAX_SCANSIZE, NULL); if(val) logg("Limits: Global size limit set to %llu bytes.\n", val); else logg("^Limits: Global size limit protection disabled.\n"); if((opt = optget(opts, "MaxFileSize"))->active) { if((ret = cl_engine_set_num(engine, CL_ENGINE_MAX_FILESIZE, opt->numarg))) { logg("!cl_engine_set_num(CL_ENGINE_MAX_FILESIZE) failed: %s\n", cl_strerror(ret)); cl_engine_free(engine); return 1; } } val = cl_engine_get_num(engine, CL_ENGINE_MAX_FILESIZE, NULL); if(val) logg("Limits: File size limit set to %llu bytes.\n", val); else logg("^Limits: File size limit protection disabled.\n"); #ifndef _WIN32 if(getrlimit(RLIMIT_FSIZE, &rlim) == 0) { if(rlim.rlim_cur < (rlim_t) cl_engine_get_num(engine, CL_ENGINE_MAX_FILESIZE, NULL)) logg("^System limit for file size is lower than engine->maxfilesize\n"); if(rlim.rlim_cur < (rlim_t) cl_engine_get_num(engine, CL_ENGINE_MAX_SCANSIZE, NULL)) logg("^System limit for file size is lower than engine->maxscansize\n"); } else { logg("^Cannot obtain resource limits for file size\n"); } #endif if((opt = optget(opts, "MaxRecursion"))->active) { if((ret = cl_engine_set_num(engine, CL_ENGINE_MAX_RECURSION, opt->numarg))) { logg("!cl_engine_set_num(CL_ENGINE_MAX_RECURSION) failed: %s\n", cl_strerror(ret)); cl_engine_free(engine); return 1; } } val = cl_engine_get_num(engine, CL_ENGINE_MAX_RECURSION, NULL); if(val) logg("Limits: Recursion level limit set to %u.\n", (unsigned int) val); else logg("^Limits: Recursion level limit protection disabled.\n"); if((opt = optget(opts, "MaxFiles"))->active) { if((ret = cl_engine_set_num(engine, CL_ENGINE_MAX_FILES, opt->numarg))) { logg("!cl_engine_set_num(CL_ENGINE_MAX_FILES) failed: %s\n", cl_strerror(ret)); cl_engine_free(engine); return 1; } } val = cl_engine_get_num(engine, CL_ENGINE_MAX_FILES, NULL); if(val) logg("Limits: Files limit set to %u.\n", (unsigned int) val); else logg("^Limits: Files limit protection disabled.\n"); #ifndef _WIN32 if (getrlimit(RLIMIT_CORE, &rlim) == 0) { logg("*Limits: Core-dump limit is %lu.\n", (unsigned long)rlim.rlim_cur); } #endif if(optget(opts, "ScanArchive")->enabled) { logg("Archive support enabled.\n"); options |= CL_SCAN_ARCHIVE; if(optget(opts, "ArchiveBlockEncrypted")->enabled) { logg("Archive: Blocking encrypted archives.\n"); options |= CL_SCAN_BLOCKENCRYPTED; } } else { logg("Archive support disabled.\n"); } if(optget(opts, "AlgorithmicDetection")->enabled) { logg("Algorithmic detection enabled.\n"); options |= CL_SCAN_ALGORITHMIC; } else { logg("Algorithmic detection disabled.\n"); } if(optget(opts, "ScanPE")->enabled) { logg("Portable Executable support enabled.\n"); options |= CL_SCAN_PE; } else { logg("Portable Executable support disabled.\n"); } if(optget(opts, "ScanELF")->enabled) { logg("ELF support enabled.\n"); options |= CL_SCAN_ELF; } else { logg("ELF support disabled.\n"); } if(optget(opts, "ScanPE")->enabled || optget(opts, "ScanELF")->enabled) { if(optget(opts, "DetectBrokenExecutables")->enabled) { logg("Detection of broken executables enabled.\n"); options |= CL_SCAN_BLOCKBROKEN; } } if(optget(opts, "ScanMail")->enabled) { logg("Mail files support enabled.\n"); options |= CL_SCAN_MAIL; if(optget(opts, "ScanPartialMessages")->enabled) { logg("Mail: RFC1341 handling enabled.\n"); options |= CL_SCAN_PARTIAL_MESSAGE; } } else { logg("Mail files support disabled.\n"); } if(optget(opts, "ScanOLE2")->enabled) { logg("OLE2 support enabled.\n"); options |= CL_SCAN_OLE2; } else { logg("OLE2 support disabled.\n"); } if(optget(opts, "ScanPDF")->enabled) { logg("PDF support enabled.\n"); options |= CL_SCAN_PDF; } else { logg("PDF support disabled.\n"); } if(optget(opts, "ScanHTML")->enabled) { logg("HTML support enabled.\n"); options |= CL_SCAN_HTML; } else { logg("HTML support disabled.\n"); } if(optget(opts,"PhishingScanURLs")->enabled) { if(optget(opts,"PhishingAlwaysBlockCloak")->enabled) { options |= CL_SCAN_PHISHING_BLOCKCLOAK; logg("Phishing: Always checking for cloaked urls\n"); } if(optget(opts,"PhishingAlwaysBlockSSLMismatch")->enabled) { options |= CL_SCAN_PHISHING_BLOCKSSL; logg("Phishing: Always checking for ssl mismatches\n"); } } if(optget(opts,"HeuristicScanPrecedence")->enabled) { options |= CL_SCAN_HEURISTIC_PRECEDENCE; logg("Heuristic: precedence enabled\n"); } if(optget(opts, "StructuredDataDetection")->enabled) { options |= CL_SCAN_STRUCTURED; if((opt = optget(opts, "StructuredMinCreditCardCount"))->enabled) { if((ret = cl_engine_set_num(engine, CL_ENGINE_MIN_CC_COUNT, opt->numarg))) { logg("!cl_engine_set_num(CL_ENGINE_MIN_CC_COUNT) failed: %s\n", cl_strerror(ret)); cl_engine_free(engine); return 1; } } val = cl_engine_get_num(engine, CL_ENGINE_MIN_CC_COUNT, NULL); logg("Structured: Minimum Credit Card Number Count set to %u\n", (unsigned int) val); if((opt = optget(opts, "StructuredMinSSNCount"))->enabled) { if((ret = cl_engine_set_num(engine, CL_ENGINE_MIN_SSN_COUNT, opt->numarg))) { logg("!cl_engine_set_num(CL_ENGINE_MIN_SSN_COUNT) failed: %s\n", cl_strerror(ret)); cl_engine_free(engine); return 1; } } val = cl_engine_get_num(engine, CL_ENGINE_MIN_SSN_COUNT, NULL); logg("Structured: Minimum Social Security Number Count set to %u\n", (unsigned int) val); if(optget(opts, "StructuredSSNFormatNormal")->enabled) options |= CL_SCAN_STRUCTURED_SSN_NORMAL; if(optget(opts, "StructuredSSNFormatStripped")->enabled) options |= CL_SCAN_STRUCTURED_SSN_STRIPPED; } selfchk = optget(opts, "SelfCheck")->numarg; if(!selfchk) { logg("Self checking disabled.\n"); } else { logg("Self checking every %u seconds.\n", selfchk); } memset(&dbstat, 0, sizeof(dbstat)); /* save the PID */ mainpid = getpid(); if((opt = optget(opts, "PidFile"))->enabled) { FILE *fd; old_umask = umask(0002); if((fd = fopen(opt->strarg, "w")) == NULL) { logg("!Can't save PID in file %s\n", opt->strarg); } else { if (fprintf(fd, "%u", (unsigned int) mainpid)<0) { logg("!Can't save PID in file %s\n", opt->strarg); } fclose(fd); } umask(old_umask); } logg("*Listening daemon: PID: %u\n", (unsigned int) mainpid); max_threads = optget(opts, "MaxThreads")->numarg; max_queue = optget(opts, "MaxQueue")->numarg; acceptdata.commandtimeout = optget(opts, "CommandReadTimeout")->numarg; readtimeout = optget(opts, "ReadTimeout")->numarg; #if !defined(_WIN32) && defined(RLIMIT_NOFILE) if (getrlimit(RLIMIT_NOFILE, &rlim) == 0) { /* don't warn if default value is too high, silently fix it */ unsigned maxrec; int max_max_queue; unsigned warn = optget(opts, "MaxQueue")->active; const unsigned clamdfiles = 6; /* Condition to not run out of file descriptors: * MaxThreads * MaxRecursion + (MaxQueue - MaxThreads) + CLAMDFILES < RLIMIT_NOFILE * CLAMDFILES is 6: 3 standard FD + logfile + 2 FD for reloading the DB * */ #ifdef C_SOLARIS #ifdef F_BADFD if (enable_extended_FILE_stdio(-1, -1) == -1) { logg("^Unable to set extended FILE stdio, clamd will be limited to max 256 open files\n"); rlim.rlim_cur = rlim.rlim_cur > 255 ? 255 : rlim.rlim_cur; } #elif !defined(_LP64) if (rlim.rlim_cur > 255) { rlim.rlim_cur = 255; logg("^Solaris only supports 256 open files for 32-bit processes, you need at least Solaris 10u4, or compile as 64-bit to support more!\n"); } #endif #endif opt = optget(opts,"MaxRecursion"); maxrec = opt->numarg; max_max_queue = rlim.rlim_cur - maxrec * max_threads - clamdfiles + max_threads; if (max_queue < max_threads) { max_queue = max_threads; if (warn) logg("^MaxQueue value too low, increasing to: %d\n", max_queue); } if (max_max_queue < max_threads) { logg("^MaxThreads * MaxRecursion is too high: %d, open file descriptor limit is: %lu\n", maxrec*max_threads, (unsigned long)rlim.rlim_cur); max_max_queue = max_threads; } if (max_queue > max_max_queue) { max_queue = max_max_queue; if (warn) logg("^MaxQueue value too high, lowering to: %d\n", max_queue); } else if (max_queue < 2*max_threads && max_queue < max_max_queue) { max_queue = 2*max_threads; if (max_queue > max_max_queue) max_queue = max_max_queue; /* always warn here */ logg("^MaxQueue is lower than twice MaxThreads, increasing to: %d\n", max_queue); } } #endif logg("*MaxQueue set to: %d\n", max_queue); acceptdata.max_queue = max_queue; if(optget(opts, "ClamukoScanOnAccess")->enabled) #ifdef CLAMUKO { do { if(pthread_attr_init(&clamuko_attr)) break; pthread_attr_setdetachstate(&clamuko_attr, PTHREAD_CREATE_JOINABLE); if(!(tharg = (struct thrarg *) malloc(sizeof(struct thrarg)))) break; tharg->opts = opts; tharg->engine = engine; tharg->options = options; if(!pthread_create(&clamuko_pid, &clamuko_attr, clamukoth, tharg)) break; free(tharg); tharg=NULL; } while(0); if (!tharg) logg("!Unable to start Clamuko\n"); } #else logg("Clamuko is not available.\n"); #endif #ifndef _WIN32 /* set up signal handling */ sigfillset(&sigset); sigdelset(&sigset, SIGINT); sigdelset(&sigset, SIGTERM); sigdelset(&sigset, SIGSEGV); sigdelset(&sigset, SIGHUP); sigdelset(&sigset, SIGPIPE); sigdelset(&sigset, SIGUSR2); /* The behavior of a process is undefined after it ignores a * SIGFPE, SIGILL, SIGSEGV, or SIGBUS signal */ sigdelset(&sigset, SIGFPE); sigdelset(&sigset, SIGILL); sigdelset(&sigset, SIGSEGV); #ifdef SIGBUS sigdelset(&sigset, SIGBUS); #endif sigdelset(&sigset, SIGTSTP); sigdelset(&sigset, SIGCONT); sigprocmask(SIG_SETMASK, &sigset, NULL); /* SIGINT, SIGTERM, SIGSEGV */ sigact.sa_handler = sighandler_th; sigemptyset(&sigact.sa_mask); sigaddset(&sigact.sa_mask, SIGINT); sigaddset(&sigact.sa_mask, SIGTERM); sigaddset(&sigact.sa_mask, SIGHUP); sigaddset(&sigact.sa_mask, SIGPIPE); sigaddset(&sigact.sa_mask, SIGUSR2); sigaction(SIGINT, &sigact, NULL); sigaction(SIGTERM, &sigact, NULL); sigaction(SIGHUP, &sigact, NULL); sigaction(SIGPIPE, &sigact, NULL); sigaction(SIGUSR2, &sigact, NULL); #endif idletimeout = optget(opts, "IdleTimeout")->numarg; for (i=0;i < nsockets;i++) if (fds_add(&acceptdata.fds, socketds[i], 1, 0) == -1) { logg("!fds_add failed\n"); cl_engine_free(engine); return 1; } if (pipe(acceptdata.syncpipe_wake_recv) == -1 || (pipe(acceptdata.syncpipe_wake_accept) == -1)) { logg("!pipe failed\n"); exit(-1); } syncpipe_wake_recv_w = acceptdata.syncpipe_wake_recv[1]; if (fds_add(fds, acceptdata.syncpipe_wake_recv[0], 1, 0) == -1 || fds_add(&acceptdata.fds, acceptdata.syncpipe_wake_accept[0], 1, 0)) { logg("!failed to add pipe fd\n"); exit(-1); } if ((thr_pool = thrmgr_new(max_threads, idletimeout, max_queue, scanner_thread)) == NULL) { logg("!thrmgr_new failed\n"); exit(-1); } if (pthread_create(&accept_th, NULL, acceptloop_th, &acceptdata)) { logg("!pthread_create failed\n"); exit(-1); } time(&start_time); for(;;) { int new_sd; /* Block waiting for connection on any of the sockets */ pthread_mutex_lock(fds->buf_mutex); fds_cleanup(fds); /* signal that we can accept more connections */ if (fds->nfds <= (unsigned)max_queue) pthread_cond_signal(&acceptdata.cond_nfds); new_sd = fds_poll_recv(fds, selfchk ? (int)selfchk : -1, 1); if (!fds->nfds) { /* at least the dummy/sync pipe should have remained */ logg("!All recv() descriptors gone: fatal\n"); pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); pthread_mutex_unlock(fds->buf_mutex); break; } if (new_sd == -1 && errno != EINTR) { logg("!Failed to poll sockets, fatal\n"); pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); } i = (rr_last + 1) % fds->nfds; for (j = 0; j < fds->nfds && new_sd >= 0; j++, i = (i+1) % fds->nfds) { size_t pos = 0; int error = 0; struct fd_buf *buf = &fds->buf[i]; if (!buf->got_newdata) continue; if (buf->fd == acceptdata.syncpipe_wake_recv[0]) { /* dummy sync pipe, just to wake us */ if (read(buf->fd, buff, sizeof(buff)) < 0) { logg("^Syncpipe read failed\n"); } continue; } if (buf->got_newdata == -1) { if (buf->mode == MODE_WAITREPLY) { logg("$mode WAIT_REPLY -> closed\n"); buf->fd = -1; thrmgr_group_terminate(buf->group); thrmgr_group_finished(buf->group, EXIT_ERROR); continue; } else { logg("$client read error or EOF on read\n"); error = 1; } } if (buf->fd != -1 && buf->got_newdata == -2) { logg("$Client read timed out\n"); mdprintf(buf->fd, "COMMAND READ TIMED OUT\n"); error = 1; } rr_last = i; if (buf->mode == MODE_WAITANCILL) { buf->mode = MODE_COMMAND; logg("$mode -> MODE_COMMAND\n"); } while (!error && buf->fd != -1 && buf->buffer && pos < buf->off && buf->mode != MODE_WAITANCILL) { client_conn_t conn; const unsigned char *cmd = NULL; int rc; /* New data available to read on socket. */ memset(&conn, 0, sizeof(conn)); conn.scanfd = buf->recvfd; buf->recvfd = -1; conn.sd = buf->fd; conn.options = options; conn.opts = opts; conn.thrpool = thr_pool; conn.engine = engine; conn.group = buf->group; conn.id = buf->id; conn.quota = buf->quota; conn.filename = buf->dumpname; conn.mode = buf->mode; conn.term = buf->term; /* Parse & dispatch command */ cmd = parse_dispatch_cmd(&conn, buf, &pos, &error, opts, readtimeout); if (conn.mode == MODE_COMMAND && !cmd) break; if (!error) { if (buf->mode == MODE_WAITREPLY && buf->off) { /* Client is not supposed to send anything more */ logg("^Client sent garbage after last command: %lu bytes\n", (unsigned long)buf->off); buf->buffer[buf->off] = '\0'; logg("$Garbage: %s\n", buf->buffer); error = 1; } else if (buf->mode == MODE_STREAM) { rc = handle_stream(&conn, buf, opts, &error, &pos, readtimeout); if (rc == -1) break; else continue; } } if (error && error != CL_ETIMEOUT) { conn_reply_error(&conn, "Error processing command."); } } if (error) { if (buf->dumpfd != -1) { close(buf->dumpfd); if (buf->dumpname) { cli_unlink(buf->dumpname); free(buf->dumpname); } buf->dumpfd = -1; } thrmgr_group_terminate(buf->group); if (thrmgr_group_finished(buf->group, EXIT_ERROR)) { logg("$Shutting down socket after error (FD %d)\n", buf->fd); shutdown(buf->fd, 2); closesocket(buf->fd); } else logg("$Socket not shut down due to active tasks\n"); buf->fd = -1; } } pthread_mutex_unlock(fds->buf_mutex); /* handle progexit */ pthread_mutex_lock(&exit_mutex); if (progexit) { pthread_mutex_unlock(&exit_mutex); pthread_mutex_lock(fds->buf_mutex); for (i=0;i < fds->nfds; i++) { if (fds->buf[i].fd == -1) continue; thrmgr_group_terminate(fds->buf[i].group); if (thrmgr_group_finished(fds->buf[i].group, EXIT_ERROR)) { logg("$Shutdown closed fd %d\n", fds->buf[i].fd); shutdown(fds->buf[i].fd, 2); closesocket(fds->buf[i].fd); fds->buf[i].fd = -1; } } pthread_mutex_unlock(fds->buf_mutex); break; } pthread_mutex_unlock(&exit_mutex); /* SIGHUP */ if (sighup) { logg("SIGHUP caught: re-opening log file.\n"); logg_close(); sighup = 0; if(!logg_file && (opt = optget(opts, "LogFile"))->enabled) logg_file = opt->strarg; } /* SelfCheck */ if(selfchk) { time(¤t_time); if((current_time - start_time) >= (time_t)selfchk) { if(reload_db(engine, dboptions, opts, TRUE, &ret)) { pthread_mutex_lock(&reload_mutex); reload = 1; pthread_mutex_unlock(&reload_mutex); } time(&start_time); } } /* DB reload */ pthread_mutex_lock(&reload_mutex); if(reload) { pthread_mutex_unlock(&reload_mutex); #ifdef CLAMUKO if(optget(opts, "ClamukoScanOnAccess")->enabled && tharg) { logg("Stopping and restarting Clamuko.\n"); pthread_kill(clamuko_pid, SIGUSR1); pthread_join(clamuko_pid, NULL); } #endif engine = reload_db(engine, dboptions, opts, FALSE, &ret); if(ret) { logg("Terminating because of a fatal error.\n"); if(new_sd >= 0) closesocket(new_sd); break; } pthread_mutex_lock(&reload_mutex); reload = 0; time(&reloaded_time); pthread_mutex_unlock(&reload_mutex); #ifdef CLAMUKO if(optget(opts, "ClamukoScanOnAccess")->enabled && tharg) { tharg->engine = engine; pthread_create(&clamuko_pid, &clamuko_attr, clamukoth, tharg); } #endif } else { pthread_mutex_unlock(&reload_mutex); } } pthread_mutex_lock(&exit_mutex); progexit = 1; pthread_mutex_unlock(&exit_mutex); if (write(acceptdata.syncpipe_wake_accept[1], "", 1) < 0) { logg("^Write to syncpipe failed\n"); } /* Destroy the thread manager. * This waits for all current tasks to end */ logg("*Waiting for all threads to finish\n"); thrmgr_destroy(thr_pool); #ifdef CLAMUKO if(optget(opts, "ClamukoScanOnAccess")->enabled) { logg("Stopping Clamuko.\n"); pthread_kill(clamuko_pid, SIGUSR1); pthread_join(clamuko_pid, NULL); } #endif if(engine) { thrmgr_setactiveengine(NULL); cl_engine_free(engine); } pthread_join(accept_th, NULL); fds_free(fds); close(acceptdata.syncpipe_wake_accept[1]); close(acceptdata.syncpipe_wake_recv[1]); if(dbstat.entries) cl_statfree(&dbstat); logg("*Shutting down the main socket%s.\n", (nsockets > 1) ? "s" : ""); for (i = 0; i < nsockets; i++) shutdown(socketds[i], 2); if((opt = optget(opts, "PidFile"))->enabled) { if(unlink(opt->strarg) == -1) logg("!Can't unlink the pid file %s\n", opt->strarg); else logg("Pid file removed.\n"); } time(¤t_time); logg("--- Stopped at %s", cli_ctime(¤t_time, timestr, sizeof(timestr))); return ret; }